Grammaticalization Scenarios: Volume 2 Grammaticalization Scenarios from Africa, the Americas, and the Pacific 9783110712735, 9783110712643, 9783110712797, 9783110712681, 2020942621

Citation preview

Walter Bisang and Andrej Malchukov (Eds.) Grammaticalization Scenarios: Cross-linguistic Variation and Universal Tendencies Vol. 2

Comparative Handbooks of Linguistics

Edited by Edith Moravcsik and Andrej Malchukov

Volume 4.2

Grammaticalization Scenarios: Cross-linguistic Variation and Universal Tendencies

Volume 2: Grammaticalization Scenarios from Africa, the Americas, and the Pacific Edited by Walter Bisang and Andrej Malchukov

ISBN 978-3-11-071264-3 (Vol. 4.2) e-ISBN (PDF) 978-3-11-071273-5 (Vol. 4.2) e-ISBN (EPUB) 978-3-11-071279-7 (Vol. 4.2) ISBN 978-3-11-071268-1 (Set volume 4.1 & 4.2) ISSN 2364-4354 Library of Congress Control Number: 2020942621 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2020 Walter de Gruyter GmbH, Berlin/Boston Cover image: Jupiterimages/PHOTOS.com/thinkstock Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark Printing and binding: CPI books GmbH, Leck www.degruyter.com

Contents Volume 2 Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun ix Measuring Grammaticalization: A questionnaire Martine Vanhove 15 Grammaticalization in Cushitic, with special reference to Beja Denis Creissels 16 Grammaticalization in Manding languages

659

695

Ronald P. Schaefer and Francis O. Egbokhare 729 17 Grammaticalization in Emai Denis Creissels 18 Grammaticalization in Tswana

769

Christian Lehmann 19 Grammaticalization in Yucatec Maya

803

Zarina Estrada-Fernández 20 Grammaticalization in Uto-Aztecan languages from northwestern 853 Mexico Johannes Helmbrecht 21 Grammaticalizations in Hoocąk

903

Marianne Mithun 22 Grammaticalization and polysynthesis: Iroquoian

943

Willem F. H. Adelaar 23 Grammaticalization in the Quechuan and Aymaran languages of the Central 977 Andes Sebastian Fedden 24 Grammaticalization in Mountain Ok (Papua New Guinea)

1007

Nikolaus P. Himmelmann 25 Grammaticisation processes and reanalyses in Sulawesi languages

1043

vi

Contents

William B. McGregor 26 Grammaticalization patterns in Nyulnyulan language

1077

Susanne Maria Michaelis and Martin Haspelmath 27 Grammaticalization in creole languages: Accelerated functionalization and 1109 semantic imitation Language index Subject index

1129 1137

Contents

vii

Volume 1 Acknowledgements List of authors

v

vii

Walter Bisang, Andrej Malchukov, and the Mainz Grammaticalization Project team (Iris Rieder, Linlin Sun, Marvin Martiny, Svenja Luell) 1 1 Position paper: Universal and areal patterns in grammaticalization Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun 89 2 Measuring Grammaticalization: A questionnaire Damaris Nübling and Luise Kempf 3 Grammaticalization in the Germanic languages

105

Michela Cennamo 4 Mechanisms and paths of grammaticalization and reanalysis 165 in Romance Björn Wiemer 5 Grammaticalization in Slavic

249

Timur Maisak 6 Grammaticalization in Lezgic (East Caucasian)

309

Juha Janhunen 7 Grammaticalization in Uralic as viewed from a general Eurasian 361 perspective Andrej L. Malchukov 8 Grammaticalization in Ewen (North-Tungusic) in a comparative 399 perspective Edward Vajda 9 Areal features in Yeniseian grammaticalization Agnes Korn 10 Grammaticalization and reanalysis in Iranian

433

465

Annie Montaut 11 Grammaticalization in standard Hindi/Urdu and Hindi dialects

499

viii

Contents

Guillaume Jacques 12 Grammaticalization in Japhug

539

Seongha Rhee 13 Grammaticalization in Korean

575

Linlin Sun and Walter Bisang 14 Grammaticalization changes in Chinese

609

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

Measuring Grammaticalization: A questionnaire 1 Grammaticalization Parameters 1.1 Parameters: an overview Our questionnaire measures grammaticalization in terms of eight parameters. The first six parameters correspond to Lehmann’s (1995) parameters as summarized in Table 1:

Tab. 1: Parameters for measuring autonomy (Lehmann 1995). Paradigmatic

Syntagmatic

Weight

1. Semantic integrity 2. Phonetic reduction

Structural scope

Cohesion

3. Paradigmaticity

4. Bondedness

Variability

5. Paradigmatic variability

6. Syntagmatic variability

The parameters as they are shown in Table 1 and as they are used in our questionnaire deviate from Lehmann’s (1995) parameters in the following way: – We split Paradigmatic Weight into the two logically independent parameters of Semantic Integrity (reduction of semantic weight, desemanticization) and Phonetic Reduction (loss of phonetic integrity, phonetic attrition). – We do not use structural scope because it proved to be theoretically and empirically most challenging (cf. Tabor and Traugott 1998; Lehmann 2004; Diewald 2010; Norde 2012). We add the following two parameters, which are frequently discussed in the literature on grammaticalization: – Parameter 7: Decategorization (cf. Hopper and Traugott 2003) – Parameter 8: Allomorphy (thus, we single out allomorphy from paradigmaticity, parameter 3) Additional remarks on our methodology: 1. The above eight parameters are logically independent. Even if all of them instantiate loss of autonomy in Lehmann’s (1995) framework, there does not seem to https://doi.org/10.1515/9783110712735-203

x

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

be covariation in the sense that the change of one parameter automatically entails the change of all other parameters. In fact, this project expects interesting cross-linguistic variation. Thus, the extent to which there are correlations and the extent to which these correlations are subject to cross-linguistic variation will be one of the results of our database. 2. There are four values for each parameter, starting from 1 for “lowest value” to 4 for “highest value”. The details will be explained separately for each parameter in section 1.2). 3. In principle, these values can be assigned to a linguistic sign in an absolute and in a relative way: – Absolute assignment: Here, the value refers to the value of that sign in its target function. – Relative assignment: Here, we look at whether the value of the sign has changed from source to target. This distinction is relevant for finding out if a given target has changed its value in the process of change from source to target but it is irrelevant for the definition of the parameters. While the definition of parameters might suggest different perspectives (6 and 7 defined relatively, other parameters in absolute terms), actually, the binary values (‘+/-‘) represent the binary perspective while ‘level’ values (1, 2, 3, 4) represent the absolute perspective, as explained below. The values which we need for our statistics consist of two parts, a value (1, 2, 3 or 4) plus ‘+’ or a ‘–’ sign. (i) The assignment of the value is absolute. We look at the properties of the target and we assign it one of the possible values as defined for each parameter. If a linguistic sign is an agglutinative affix, it will get the value 3 for parameter 4 ‘bondedness’. (ii) The assignment of the ‘+’ / ‘–’ sign depends on whether there was a change of value between the source and the target. Thus, the assignment of the ‘+’ / ‘–’ sign is based on a relative perspective: – If there is no change from source to target, the ‘–’ sign is written in front of the relevant value of the target: e.g., –3 for parameter 4 ‘bondedness’, if the source is an agglutinative affix and the target remains an agglutinative affix. – If there is a change in value, a ‘+’ sign will be added to the value of the target, e.g., +4. This means that the parameter value for the target is 4 and that the value for the source was lower than 4. Notice that we do not indicate a concrete value for the source concept in our statistics. We are only interested in whether there is a change from source to target (+) or not (–). 4. If markers develop different values in different slots of a paradigm, the marker with the highest value will be selected. This can be illustrated by an example from Beja (Cushitic): Table 2 shows the perfective (pfv) forms of ‘say’ and the suffixes for the imperfective derived from it:

Measuring Grammaticalization: A questionnaire

xi

Tab. 2: From the paper of Vanhove (this volume, Table 2). ‘say’ pfv

Imperfective marker (v2)

1sg

a-ni

-ani

2sg.m

ti-ni-ja

-tnija

2sg.f

ti-ni:

-tini:

3sg.m

i-ni

-i:ni

3sg.f

ti-ni

-tini

1pl

ni-di

-nej/-naj

2pl

ti-di:-na

-te:n(a)

3pl

e:-n(a)

-e:n(a)

A look at parameter 2 (Phonetic Reduction) shows that there is no change from 1sg.pfv > 1sg.ipfv (value 1), while we have “loss of phonetic substance with effects on syllable structure” (value 3) in the cases of 2sg.m, 1pl and 2pl. Since the latter case has a higher grammaticalization value, this value is chosen +2 (the markers in question are syllabic and not subsyllabic). 5. In many cases, grammaticalization does not affect a single source item but a more complex construction consisting of more than one item. In such cases, our measurement will be focused only on the nucleus of that construction, which we define as the most “lexical” component of the source concept (cf. the notion of by a ‘construction marker’ in Himmelmann [2005: 80]). Thus, the measurement of the English future marker be going to will be focused on the (lexical) nucleus go (rather than –ing or to). Of course, what is assessed is the target be going to but for assessing the change from source to concept (the + or – values), it is necessary to integrate the properties of the source as well. 6. The distinction between tokens and types of sources and targets is of no relevance for this project, since the values (e.g., +3) remain the same, irrespective of whether we evaluate a token or a type.

1.2 The individual parameters Parameter 1: Semantic Integrity A grammatical marker is more general to the extent that it is semantically compatible with more lexical host items (Bybee [1985] on semantic generality). 1 The linguistic sign has a lexical meaning. 2 The linguistic sign has an abstract meaning, which is however referential/denotational rather than relational (e.g., ‘people’, existential verb, pronouns).

xii 3 4

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

The linguistic sign has an abstract non-denotational or relational meaning (e.g., prepositions, auxiliaries, numeral classifiers). The linguistic sign only has a syntactic function.

The whole continuum from 1 to 4 can be exemplified by the common development from body part > locative case through a number of intermediate stages (Lehmann 1995; Heine and Kuteva 2002): e.g., ‘head’ (value 1) > ‘top’ (2) > on (3) > locative case (4). The distinction between ‘referential’ and ‘relational’ (cf. values 2 and 3) should be understood in the conventional sense that lexical categories have denotations, while grammatical categories do not. Lehmann (2002: 139) refers to this distinction also as autosemantic vs. synsemantic. He notes that any full lexical item can signify by itself a certain concept (object, process, property) and independently refer to a certain class of such concepts, but when grammaticalized, it loses this ability and depends on another item for expressing certain distinctions. “A number or gender markers does not signify a number or gender concepts as such, but only insofar as these are features of other concepts” (Lehmann 2002: 139).1 The difference between values 3 and 4 corresponds to the distinction between “wide” vs. “narrow” definitions of grammatical (inflectional) meaning. The wide perspective (value 3) covers semantically meaningful inflectional categories (like tense in verbs; cf. “inherent inflectional morphemes” in terms of Anderson [1985]2). Narrow inflectional meaning (value 4) is characterized by the reduction to purely syntactic functions (case on nouns; agreement on verbs, adjectives, etc.). Parameter 2: Phonetic reduction 1 The linguistic sign consists of 2 or more syllables. 2 The linguistic sign is (i) a monosyllabic word or (ii) a full syllable with no word status. 3 The linguistic sign is a subsyllabic morpheme. 4 The linguistic sign is (reduced to) a suprasegmental feature or is lost. Parameter 3: Paradigmaticity (paradigmatic cohesion) This parameter is concerned with the size of the paradigm and its degree of homogeneity. 1 If the distinction between the values 2 and 3 is not clear for individual expressions, a somewhat simplified version of the tests designed by Boye and Harder (2012) can be used to distinguish between lexical vs grammatical categories (cf. addressability and focalizability). Thus, only lexical (‘referential’, ‘autosemantic’) but not grammatical (‘synsemantic’) items can be questioned or focused: cf. demonstratives vs. the definite article in English: (a) Which book? – This. Vs. *The; (b) It is this I like vs. *It is the I like. 2 Anderson (1985) distinguishes the following three types of inflectional forms: relational inflectional morphemes, agreement morphemes, inherent inflectional morphemes.

Measuring Grammaticalization: A questionnaire

1 2 3 4

xiii

The linguistic sign belongs to a major (open) word class. The linguistic sign is an element of a minor (closed) word class. The linguistic sign expresses a grammatical category but is not fully paradigmatic. The linguistic sign is a member of a small and homogeneous paradigm (it occurs in the same morphological slot as the other members of the same grammatical domain, its phonological shape in terms of number of phonemes or syllable structure is similar to at least most of the other members of the paradigm).

Examples of value 1 are elements that belong to a major word class (N, V, A), while elements with value 2 belong to a minor word class (adpositions, auxiliaries, but also pronouns). The linguistic expressions with value 3 express a grammatical category but are not fully integrated formally and/or semantically as, e.g., analytic tense forms in languages with synthetic tenses (cf. English Has done, German werde gehen [will go]). Similarly, aspect forms in Russian showing semantic and formal idiosyncrasies are not fully paradigmatic. The same holds for gender marking on nouns, which does not form a paradigm for individual lexical items, i.e., a specific noun does not occur in different gender forms in most cases (exceptions are cases like Spanish muchach-a ‘girl’ vs. muchach-o ‘boy’. Differentiation of markers with values 2 vs. 3 is not always straightforward, but can be decided for individual cases on the basis of frequency of individual markers, functional coherence and the number of grammaticalized markers (high frequency, semantic coherence and few markers involved speak for 3, that is for an emerging paradigm). Finally, value 4 will be reserved for canonical inflectional paradigms (tense on verbs, case on nouns), where individual lexemes have at least two forms in a paradigm. On this view, gender has value 4 only on adjectives (which have agreement), but not on nouns. Parameter 4: Bondedness (syntagmatic cohesion) This parameter measures the degree of cohesion/fusion between the host and the linguistic sign that undergoes grammaticalization. 1 The linguistic sign is a free morpheme or is the lexical root of a word. 2 The linguistic sign is a clitic (its use is not limited to a single word class). 3 The linguistic sign is an agglutinative affix (affixed to individual words which are members of a single word class). 4 The linguistic sign is part of a porte-manteau morpheme or is a suprasegmental (e.g., tonal marker) or a process morpheme (Ablaut, …), or a zero morpheme. Parameter 5: Paradigmatic variability Degree of obligatoriness: 1 No obligatoriness: The use of a linguistic sign is not imposed by the system, which is true of all lexical items, but also of grammatical items at early stages of grammaticalization.

xiv 2 3 4

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

The obligatoriness of a linguistic sign is restricted to a small set of clearly defined constructions in certain contexts (cf. examples3). The linguistic sign is obligatory in most contexts (e.g., an article is obligatory for indefinites except when used as nominal predicates). The linguistic sign is generally obligatory (i.e., it always cooccurs with the relevant host) and the speaker has to select one value out of a set of markers expressing the relevant category (e.g., tense marking system with three values: PRS, PST, FUT).

Clarification on obligatoriness based on Lehmann’s (1995) view: By this we mean the freedom of the language user with regard to the paradigm as a whole. The paradigm represents a certain grammatical category, and its members, the subcategories (or values) of that category. There may then be a certain freedom in either specifying the category by using one of its subcategories or leaving the whole category unspecified. To the extent that the latter option becomes constrained and finally impossible, the category becomes obligatory. (Lehmann 1995: 124)

Generally, the opposite values (not obligatory: 1 vs. obligatory: 4) are least controversial. Intermediate values relate to cases in which a category is obligatory in a small number of instances (value: 2), or in a large number of instances (value: 3). A good example are the different degrees of grammaticalization in plural marking: Not grammaticalized (value: 1), grammaticalized only on pronouns and kin terms (value: 2), grammaticalized/obligatory for all nouns except for non-specific ones (value: 3), and finally grammaticalized for all (count) nouns (value: 4). For the values 3 and 4, we define individual obligatoriness domains as for example: – Tense: Occurrence in an independently utterable declarative clause. – Aspect: Like tense. – Evidentials: Like tense.

3 (i) Accusative in some languages with differential object marking is only found on pronouns, but not on (common) nouns. (ii) The selection of passive voice due to syntactic reasons as in the case of conjunction reduction (Equi NP deletion). Here the verb occurs in two forms and the speaker is forced into selecting either the active voice with the coreference of agent and intransitive S as in (1) or the passive form with the coreference of patient and intransitive S as in (2): (1) Johni sees the dog and øi runs away; (2) The dogi was seen by John and øi run away. Thus, English passives get the value 2. In contrast, symmetric voice systems as we find them in Austronesian get the value 4 because the selection of a specific voice marker out of a set of voice markers is compulsory in a finite assertive clause. (iii) The obligatory expression of the subject in the subordinate clause of Chinese control constructions if the matrix-clause subject differs from the subordinate subject. E.g., Lĭ Líni yào tāk măi shū [Li Lin want s/he buy book] ‘Li Lini wants him/herk to buy a book’ vs. Lĭ Líni yào øk măi shū [Li Lin want buy book] ‘Li Lini wants øi to buy a book’.

Measuring Grammaticalization: A questionnaire

– – – –

xv

Person: Like tense. Number: Occurrence with all count nouns, or like tense in the case of verb agreement. Definiteness/indefiniteness: Occurrence with all count nouns. Gender: Occurrence with all nouns (±including proper names) plus agreement.

A remark on the independence of the parameters with regard to paradigmaticity (parameter 3) and obligatoriness: A grammatical category can be obligatory but not (fully) paradigmatic. This can be seen from the case of gender marking, which tends to be maximally 3 (on parameter 3) if marked on the noun even though it is obligatory (according to parameter 5). There also exists the opposite case in which a category is paradigmatic but not obligatory. For example, different slots in polysynthetic languages might be seen as paradigmatic (accommodating different aspectual, valency and other categories which are mutually exclusive for a certain position class) even though not all of them are obligatory. In fact, the minimal verb form in such languages might include very few slots which are obligatory (often person, number and some subcategories of TAM). Thus, paradigmaticity is clearly related to obligatoriness, but there is no one-to-one relation between the two. At a more general level, paradigmaticity relates to individual markers (e.g., perfect tense), while obligatoriness relates to a category as a whole.

Parameter 6: Syntagmatic variability The syntagmatic variability of a sign is the ease with which it can be shifted around in its context. In the case of a grammaticalized sign, this concerns mainly its positional mutability with respect to those constituents with which it enters into construction. Syntagmatic variability decreases with increasing grammaticalization. (Lehmann 1995: 140) 1 Word order is not constrained, i.e., it is as free as the lexical items of a language to which the source concept belongs. 2 The position is more constrained, but is syntactically transparent in the sense that it corresponds to the normal syntactic position of the source concept. 3 The linguistic sign is assigned to a position that is no longer transparent (undergoes ‘positional adjustment’ in terms of Lehmann [1995: 169] or ‘permutation’ in terms of Heine [Heine and Reh 1984: 132]). 4 The linguistic sign becomes bound (is assigned a morphologically fixed position in a morphological template). The distinction between values 2 and 3 depends on whether the grammatical marker retains its position or undergoes restructuring (in terms of position or type [nonbound vs. bound, overt vs. zero]). A good example of restructuring is the grammaticalization of participles (deinen Vorschlägen entsprechend [your propositions according]) to prepositions (entsprechend deinen Vorschlägen [according your propo-

xvi

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

sitions]) (cf. Lehmann 1995: 142). In this case, entsprechend in its function as a preposition gets the value +3. Another example of restructuring are auxiliaries in Romance languages. While the preferred word order for the verb in Latin was clause final (epistulam scriptam habeo [letter written have]) or clause initial (habeo epistulam scriptam [have letter written]), it is now in a fixed position immediately preceding the lexical verb, which was a rather marked position in Latin: Ital. Ho scritto una lettera (= habeo scriptam epistulam) (cf. Lehmann 1995: 141). For that reason, the Italian auxiliary gets again the value +3. If the source is already a bound morpheme as in many cases of secondary grammaticalization (i.e., a “grammatical marker becoming more grammatical”, along the aforementioned parameters), the value of the target is -4.

Parameter 7: Decategorization 1 No change of categorial properties of the source word class (in particular, inflectional categories of nouns and verbs). 2 Partial absence of certain categorial properties (for example, German kann [3.sg:prs:can] does not have non-finite forms, but retains tense and person distinctions). 3 Absence of most categorial properties (for example, English modal verbs like can retain tense distinctions, but lose person). 4 Total decategorization.

Parameter 8: Allomorphy 1 No allomorphy. 2 The linguistic sign shows moderate (phonologically conditioned) allomorphy. 3 There is morphologically conditioned allomorphy. 4 There is lexically conditioned allomorphy. In Bybee, Pagliuca, and Perkins’s approach (1994: 110–113), allomorphy contributes to the loss of autonomy of a sign, and phonologically conditioned allomorphy is regarded as instantiating less dependency than morphologically or lexically conditioned allomorphy. We also adopt a similar approach here albeit in a simplified way.

2 Source concepts to be evaluated Before going into the details about the features for measuring grammaticalization, indicate which of the following 30 source concepts (a selection from Heine and Kuteva [2002]) undergo grammaticalization in the language you analyze (Yes vs. No). Please, also note the target function(s).

Measuring Grammaticalization: A questionnaire

Tab. 3: 30 source concepts (Heine and Kuteva 2002). No

Source Concept

Target function(s)

1.

arrive

1) succeed 2) ability 3) allative 4) until (temporal)

2.

back (body part)

1) after 2) behind 3) cause 4) earlier 5) then 6) up (spatial)

3.

body

1) intensive-refl 2) middle 3) reciprocal 4) reflexive

4.

child

1) classifier 2) diminutive 3) partitive

5.

come

1) consecutive 2) continuous 3) hortative 4) venitive

6.

copula

1) avertive 2) conditional 3) consecutive 4) focus 5) future 6) obligation

7.

demonstrative

1) complementizer 2) conjunction 3) copula

Yes

Target functions Other

xvii

xviii

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

Tab. 3 (continued) No

Source Concept

Target function(s)

Yes

Target functions Other

4) definite 5) focus 6) person pron, third 7) relative 8) subordinator 8.

do

1) causative 2) continuous 3) emphasis 4) obligation 5) pro-verb

9.

fall

1) down 2) passive

10.

finish

1) after 2) already 3) completive 4) consecutive4 5) perfective

11.

follow

1) according to 2) behind 3) comitative

12.

get

1) ability 2) change-of-state 3) obligation 4) passive 5) past 6) permissive 7) h-possessive5

4 Nonsubordinating clause-linker: S1, (and) then S2. 5 That is, predicative possessive (marker of possessive ‘have’-constructions; Heine and Kuteva [2002]).

Measuring Grammaticalization: A questionnaire

Tab. 3 (continued) No

Source Concept

Target function(s)

8) possiblity 9) succeed 13.

give

1) benefactive 2) causative 3) concern 4) dative 5) purpose

14.

go

1) andative 2) change-of-state 3) consecutive 4) continuous 5) distal demonstrative 6) habitual 7) hortative 8) immediate future

15.

hand (body part)

1) agent 2) five 3) locative 4) h-possesive

16.

head (body part)

1) front 2) intensive-refl 3) middle 4) reflexive 5) up

17.

here

1) cause 2) demonstrative 3) person-pron 4) relative

18.

leave

1) ablative 2) completive

Yes

Target functions Other

xix

xx

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

Tab. 3 (continued) No

Source Concept

Target function(s)

3) egressive 4) hortative 5) negation 6) permissive 19.

live

1) continuous 2) habitual 3) locative copula 4) exist

20.

love

1) avertive 2) future 3) intention 4) proximative

21.

man

1) classifier 2) exlamation 3) indefinite pronoun 4) male 5) third pers-pron

22.

one

1) alone 2) indefinite 3) indefinite pronoun 4) only 5) other 6) same 7) singulative 8) some 9) together

23.

say

1) cause 2) complementizer 3) conditional 4) evidential

Yes

Target functions Other

Measuring Grammaticalization: A questionnaire

xxi

Tab. 3 (continued) No

Source Concept

Target function(s)

Yes

Target functions Other

5) purpose 6) quotative 7) simile 8) subordinator 24.

see

1) allative 2) passive

25.

side

1) beside 2) locative 3) near

26.

sit

1) continous 2) copula 3) habitual

27.

take

1) causative 2) comitative 3) completive 4) future 5) instrument 6) patient 7) h-possessive

28.

thing

1) complementizer 2) indefinite pronoun 3) a-possessive6

29.

want

1) avertive 1) future 2) proximative

30.

woman

1) classifier 2) female

6 That is, attributive possessive, for example, expressed by genitive or an adposition (Heine and Kuteva 2002).

xxii

Walter Bisang, Andrej Malchukov, Iris Rieder, and Linlin Sun

We ask the contributors to check whether any of grammaticalization paths listed above (in our 30-concept list in Table 3) is found in the language of your expertise. If there is a path of grammaticalization for a given source concept, please provide an example of grammaticalized structures, as well as evaluate the grammaticalization path in accordance with the parameters (and their values), as specified above. In addition, we ask our contributors to check the tables documenting for grammaticalization paths extracted from individual papers for accuracy of value assignment (especially, cases where the author is explicitly asked to provide feedback but also elsewhere). Form for analyzing the grammaticalization value of individual markers/ linguistic signs Language: SOURCE concept: TARGET concept:

Parameter

1

2

3

4

5

6

7

8

Value

Explanations and examples 1. Semantic integrity: 2. Phonetic reduction: 3. Paradigmaticity (paradigmatic cohesion) 4. Bondedness (syntagmatic cohesion) 5. Paradigmatic variability 6. Syntagmatic variability 7. Decategorization 8. Allomorphy

References Anderson, Stephen R. 1985. Inflectional morphology. In Timothy Shopen (ed.), Language typology and syntactic description, Vol. III: Grammatical categories and the lexicon, 150–201. Cambridge: Cambridge University Press. Boye, Kasper & Peter Harder. 2012. A usage-based theory of grammatical status and grammaticalization. Language 88. 1–44. Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form (Typological Studies in Language 9). Amsterdam & Philadelphia: John Benjamins.

Measuring Grammaticalization: A questionnaire

xxiii

Bybee, Joan L., Revere D. Perkins & William Pagliuca. 1994. The evolution of grammar: tense, aspect, and modality in the languages of the World. Chicago IL: University of Chicago Press. Diewald, Gabriele. 2010. On some problem areas in grammaticalization theory. In Katerina Stathi, Elke Gehweiler & Ekkehard König (eds.), Grammaticalization: Current views and issues, 17–50. Amsterdam & Philadelphia: John Benjamins. Heine, Bernd & Mechtild Reh. 1984. Grammaticalization and reanalysis in African languages. Hamburg: Buske Verlag. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Himmelmann, Nikolaus P. 2005. Gram, construction, and class formation. In Clemens Knobloch & Burkhard Schaeder (eds.), Wortarten und Grammatikalisierung, 79–93. Berlin & New York: Mouton de Gruyter. Hopper, Paul J. & Elizabeth Closs Traugott. 2003. Grammaticalization (2nd edn). Cambridge: Cambridge University Press. Lehmann, Christian. 1995 [1982]. Thoughts on grammaticalization. A programmatic sketch. Munich: Lincom Europa. Lehmann, Christian. 2002, Thoughts on grammaticalization (2nd revised edn). Erfurt: Seminar für Sprachwissenschaft der Universität (ASSidUE, 9). Lehmann, Christian. 2004. Theory and method in grammaticalization. Zeitschrift für germanistische Linguistik 32. 152–187. Norde, Muriel. 2012. Lehmann’s parameters revisited. In Kristin Davidse, Tine Breban, Lot Brems & Tanja Mortelmans (eds.), Grammaticalization and language change: New reflections, 73–110. Amsterdam: John Benjamins. Tabor, Whitney & Elizabeth C. Traugott. 1998. Structural scope expansion and grammaticalization. In Anna Giacalone Ramat & Paul J. Hopper (eds.), The limits of grammaticalization (Typological Studies in Language 37), 229–272. Amsterdam & Philadelphia: John Benjamins.

Martine Vanhove

15 Grammaticalization in Cushitic, with special reference to Beja Abstract: Reconstructible grammaticalization processes in Cushitic (Afroasiatic) concern mainly the pronominal and verbal domains, markers of subordination, adpositions, question words and discourse particles. Starting with the sole representative of its Northern branch, Beja, this chapter investigates the various sources, targets and paths of 70 grammaticalization processes, seeks for comparable phenomena in three other branches of Cushitic – Central, Highland East and Lowland East – and compares them with the list provided in Heine and Kuteva’s (2002) lexicon of the languages of the world, or with other literature about the languages of the Horn of Africa where Cushitic languages are spoken. Four main features of grammaticalization in Cushitic emerge from this investigation: (i) the nominal domain can be a source of grammaticalization, but this not (or hardly) the case for targets; (ii) auxiliaries, in particular the quotative verb which shows a vast array of functions, are often the source of the renewal and enrichment of the verbal system, but this is far less the case for verbless clauses as opposed to other Afroasiatic families, namely Semitic and Egyptian (Cohen 1984); (iii) quotative verbs have a strong tendency to grammaticalize at different levels of the language structure, verbs, complex sentences, discourse, including functions unattested in other genetic stocks; (iv) Cushitic languages show a pervasive semantic link between ‘say’ and ‘purpose’ at large. Keywords: Cushitic, Beja, grammaticalization, auxiliary, light verb construction

1 Introduction In this chapter, grammaticalization processes (some of them still on-going) are viewed from a Beja perspective as a starting point (Sections 2–5). They are briefly compared to similar, and sometimes different, processes in a sample of languages from three other branches, Central Cushitic, Lowland-East Cushitic and HighlandEast Cushitic (Sections 3–6), which form the closest (or rather less distant) relatives of Beja (the main divisions and geographical localizations of the Cushitic groups are displayed on the map below). A comparative summary of the grammaticalization paths in all four branches and a discussion of their peculiarities is proposed in Section 7.

https://doi.org/10.1515/9783110712735-001

660

Martine Vanhove

Fig. 1: Map of Cushitic Languages.

1.1 General information about Beja The North-Cushitic (NC) branch of the Afroasiatic phylum consists of only one language, Beja, locally named beɖawijeː=t. It is lexically and grammatically quite distant from its closest Lowland East and Central Cushitic relatives, Afar-Saho and Agaw, and is considered as peripheral within the Cushitic family, but the affiliation of Beja to the Cushitic branch is no longer disputed (see Hetzron [1980]; Lamberti [1991]; Zaborski [1997]; Tosco [2000]; Appleyard [2004] for a discussion). It is spoken in the northernmost part of the Cushitic-speaking area, mainly in Eastern Sudan between the Red Sea and the river Atbara by approximately 1,100,000 speakers (1993 census, probably even more now), and in Northern Eritrea (approx. 60,000 speakers). It used to be spoken in Southern Egypt, but it seems that (almost?) all speakers have now shifted to Arabic. The first very fragmentary attestations of the language (mainly a few anthroponyms) date back to the 3rd millennium BC until the 7th century AD (see Rilly [2014] for details), but the first comprehensive scientific study and data concerning Beja date back to the second half of the 19th century (Almkvist 1881–1885).

Grammaticalization in Cushitic, with special reference to Beja

661

Beja dialectology is ill-known, but dialects do not seem much differentiated. There are three main dialectal zones: North, Centre and South, which have further local and tribal-based subdivisions. These divisions are mainly based on vocalic isoglosses and some lexical peculiarities. In Sudan, bilingualism with Sudanese Arabic is widespread and expanding, but discredited for women who lead a cloistered life. Beja speakers have a strong awareness of a hierarchy of speech related to rules of honour, politeness, and taboos, correlated to a strong inclination towards allusive speech. Beja is unwritten in Sudan and school education is in Arabic.

1.2 Typological characteristics Within the consonantal phonemes, the alveolar order predominates and is represented by six phonemes while the others have fewer members: 3 bilabials, 3 palatals, 2 retroflex, 2 velars, 2 labiovelars, 2 laryngeals, and 1 labiodental. The voiceless retroflex consonant ʈ is rare, and there are none of the two pharyngeal fricatives which are reconstructed for proto-Cushitic (Cohen 1988: 251) and still exist in neighboring Cushitic languages. The vowel inventory opposes short and long vowels in an unbalanced way: 4 short vowels vs. 5 long ones. Stress assignment mostly depends on the syllabic structure, and there are no phonological tones. Beja has a rich and complex morphology, flectional and derivational, both in the nominal and verbal domains. It is partly organized through the interplay of consonantal roots and patterns involving vocalic ablaut in the stem, and also reduplication and affixes in some instances, for verb inflection and derivation, noun formation, verb-noun derivation, adjective and plural formation (this is similar to Arabic, but unlike most Cushitic languages which have at best traces of this system, often limited to noun plural formation). Beja is also partly agglutinative, with suffixes and enclitics (the majority), as well as prefixes and proclitics, which are often portmanteau morphemes. It is a marked nominative language with three nominal cases for lexical items, two for the verb core arguments, nominative and accusative, and one for noun phrases, genitive. Case of core arguments is marked on determiners by vocalic ablaut (or a consonant in one marginal case), and by vocalic suffixes on the dependent noun for the genitive. Pronouns have in addition dative and ablative sets. There are two genders, masculine and feminine, and two sets of articles, a proclitic definite one, which partly neutralizes number, gender and case, and an enclitic indefinite one which varies for gender (m and f) and number (sg and pl), and marginally for case, which surfaces only in some phonological contexts in the masculine. In the noun phrase, the number and gender of both the possessor and the possessed are indexed on the possessor.

662

Martine Vanhove

Nouns, pronouns, adjectives, the manner deictics, some interrogative words, and verbs can all be predicates. Verb paradigms have dedicated sets of flectional morphemes, while the other word categories are conjugated with an enclitic copula which varies for number, person, and partially for gender. Verbs have both finite and non-finite forms. Finite forms are organized along a three-term aspectual system, which distinguishes through flectional morphemes and apophony in the stem a Perfective, an Imperfective and an Aorist which index also the person, number and gender (only in 2 and 3 sg) of S. There are two morphological verb classes. V1 verbs have prefixed paradigms (except plural indices which are suffixes), or an infix for disyllabic verbs. This is historically the oldest class. V2 verbs have only suffixes, and represent a common Cushitic innovation. Example (1) illustrates V1 verbs for the three paradigms (‘kneel’, ‘have’, ‘kill’), and V2 for the Aorist (‘come’). (1)

i-ganif-na=eːb oː=doːr jhaːm dhaːj 3m-kneel\mid.pfv-pl=rel.m def.sg.m.acc=time leopard dir jʔ-i=t eː=kam come-aor.3sg.m=coord def.pl.m.acc=camel\pl ji=iː-biri-n=eː=naː=jeːb kaːm hoːj dari rel.pl.m=3-have\aor-pl=rel=thing=loc.pl camel abl.3 kill[3sg.m] ‘When they stopped (pfv), a leopard came (aor) towards (them) and killed (ipfv) one of the camels that they had (aor).’

There is, in addition, a rich system of semantic and voice derivation involving ablaut, reduplication and affixal devices (pluractional, intensive, middle, passive, reciprocal, causative, and marginally double causative). The non-finite forms amount to four converbs labelled General, Sequential, Simultaneity and Manner converbs. They are invariable, except the Manner converb which may vary for gender in some specific syntactic environments. Only the Simultaneity converb can be negated, with the function of a Privative converb (‘without’). This might be because it is the sole converb that can be traced back to a finite form (it has the same morphological pattern as the actual Prohibitive and negative Optative forms). For the other converbs, either coordinate clauses or balanced subordinate clauses are used in negative utterances. The converbs are used in deranked subordinate clauses and as auxiliated forms in complex predicates which will be dealt with in detail in Section 3. Syntactically, Beja is predominantly head-final, the canonical constituent order is (X)(S)(O)V, and dependent clause – main clause. Constituent order is not particularly rigid and may vary for pragmatic reasons (for further details, see Vanhove [2017]). In the noun phrase, the constituent order is often conditioned by information structure, particularly in relative clauses where the head noun is often topicalized and thus precedes the relative clause.

Grammaticalization in Cushitic, with special reference to Beja

663

2 Grammaticalization of nominal categories In most instances, I was not able to trace back the morphology of nominal categories to grammaticalization processes. The clitic feminine morpheme t and the plural devices are pan-Afroasiatic, the genitive suffix -i goes back to Proto-Cushitic, the indefinite accusative article =b cannot be etymologically related to any free form, nor to another grammatical marker, and the vowels of the nominative and accusative cases are of an unknown origin. We only know that they do not go back to Proto-Cushitic (see Appleyard [2004] for a detailed discussion). The origin of the definite articles is disputed as either a reduction or a reinforcement of former demonstratives (Appleyard 2004: 178–181). Nonetheless, the nominal domain presents three transparent grammaticalization processes that are worth mentioning.

2.1 The origin of independent personal pronouns The second and third person independent personal pronouns (nominative and accusative) are a Beja innovation as compared to other Cushitic languages. They have developed from a verbo-nominal base, *bar- in the masculine, *ba(r)=t- in the feminine ‘*possession(s)’, which does not survive as such in the language, and to which the possessive enclitic suffixes were added in the appropriate case. This form derives from the verb of possession biri ‘have’ (Bechhaus-Gerst 1985; Appleyard 2004). A new grammaticalization process is taking place: all six persons, including first persons, can be reinforced with the proximal demonstratives that often procliticize to the pronouns and whose long vowels are often reduced to short ones, as shown in Table 1. Tab. 1: Beja independent personal pronouns (Vanhove 2017). nom

1 2m 2f 3m 3f

acc

sg

pl

sg

pl

(un=)ani (um=)bar-uː-k (um=)bat-uː-k (um=)bar-uː (um=)bat-uː

(an=)hinin (am=)bar-aː-k(na) (am=)bat-aː-k(na) (am=)bar-aː (am=)bat-aː

(on=)aneːb (om=)bar-oː-k (om=)bat-oː-k (om=)bar-oː (om=)bat-oː

(en=)hinin (em=)bar-eː-k(na) (em=)bat-eː-k(na) (em=)bar-eː (em=)bat-eː

2.2 The origin of the emphatic reflexive pronouns Among the three devices that form emphatic reflexive pronouns (intensifiers), one is a borrowing from Arabic, but the other two clearly go back to nouns.

664

Martine Vanhove

The first source is the possessive noun meaning ‘owner’ kina (kna after the definite article), as, e.g., in Manding languages (Creissels this volume), but unlike Manding, it is not limited to the third person and occurs for all three persons in the singular and the plural. It is always used with the definite article and is marked for gender and number. It has three main functions. Its first use is that of an adnominal intensifier, often for subject and object topicalization. In this case kina has, in addition, an enclitic possessive pronoun referring to the person if the topic is a pronoun (2) but not if the topic is a noun or a demonstrative (3). (2)

ani i=kna=ji bak 1sg.nom def.m=pro.refl=poss.1sg.nom deict.mnr daː-s-eːti a-ndʔi be_there-caus-cvb.gnrl 1sg-do\ipfv ‘As for me, I would let the matter drop.’1

(3)

toːt toː=kna neːw-eːti prox.sg.f.acc def.sg.f.acc=pro.refl insult-cvb.gnrl ‘As for her, since he insults her …’

The second adnominal function is attributive. It marks possession in one’s own right. It occurs after a noun and bears no possessive pronoun. (4)

i=liːli aː=kna gʷid-a na def.m=eye def.pl.m.nom=pro.refl be_numerous-cvb.mnr thing rh-jaːn=heːb see-pfv.3pl=obj.1sg ‘My own eyes (Lit. the eyes themselves) saw many things for me.’

The third function is adverbial, and kina intensifies the meaning of the predicate or the entire clause, and bears no possessive pronoun either. (5)

toː=kna a-naw=hoːb def.sg.f.acc=pro.refl 1sg-lack\mid.pfv=when ‘When I missed her altogether …’

1 All the Beja examples come from data I collected in Sudan between 2000 and 2011 and in Paris in 2014 and 2016. Most of them are available online at http://corporan.huma-num.fr/Archives/corpus.php.

Grammaticalization in Cushitic, with special reference to Beja

665

A second set of emphatic reflexive pronouns, less frequent, is built on the same structural principle and goes back to a body part noun biji ‘member’ still recorded with this meaning in Reinisch (1895: 54), but lost today.

2.3 The origin of the indefinite pronoun As in many languages of the world, the indefinite pronoun ‘some, someone, some people’ has grammaticalized from the numeral ‘one’. In Beja, the plural form may also be used (in addition to the singular). (6)

oːn i=kiraːj ah-iːt gali=b gaw prox.sg.m.acc def.m=payment take-cvb.seq one\pl=indf.m.acc house ‘Having rented a house from someone …’

3 Grammaticalization of verbal categories The verbal domain is the most prolific as far as grammaticalization is concerned. The use of periphrastic constructions of various morphological make-ups is well developed for the expression of several TAM, in light verb constructions and in directional constructions. The auxiliary strategy has been the basis of the renewal of verb morphology in the whole Cushitic family. Noteworthy is the fact that in Beja the speech verb meaning ‘say’ has been recruited to encode a large variety of grammatical and semantic categories, beyond well-known cross-linguistic developments (for Africa, see Güldemann [2008]).

3.1 Renewal of paradigms and light verb constructions As mentioned in the introduction, the V2 verb class conjugated with suffixes is a recent innovation in Beja, as in other Cushitic languages. It is usually argued (but see contra Banti [2004], Section 6.2.1 below) that this verb class is the result of the evolution of an auxiliary construction made of a verbo-nominal base and a verb n/j ‘say’ (or ‘be’ in other languages) conjugated with prefixes, as Table 2 shows.2 This evolution is usually explained by the intermediate stage of a light verb construction as it still exists in Beja and other Cushitic languages, and also as one

2 The original ‘say’ verb has partially been renewed with another root d in contemporary Beja (present in the Perfective and the Aorist for all persons), hence the non-concordance of forms in 1 and 2 pl.

666

Martine Vanhove

Tab. 2: Comparison of paradigm of di ‘say’ and V2 flectional morphemes.

1sg 2sg.m 2sg.f 3sg.m 3sg.f 1pl 2pl 3pl

‘say’ PFV

V2 IPFV

a-ni ti-ni-ja ti-niː i-ni ti-ni ni-di ti-diː-na eː-n(a)

-ani -tnija -tiniː -iːni -tini -nej/-naj -teːn(a) -eːn(a)

of the areal features in the languages of North-East Africa (see, e.g., Güldemann 2005). In Beja, the verb ‘say’ can be in the base form (for transitive and intransitive verbs) or in the double-causative and intensive derivation (for transitive verbs only, and onomatopoeias).

(7)

faras=t=iː bak nhal diː-tiːt pierce\n.loc=indf.f=abl.sg deict.mnr go_up say-cvb.seq ‘After he had gone up quickly through the opening …’

(8)

nʔa-ti daː-s i-siːs-joːdi-n=hoːb be_down-cvb.gnrl be_there-caus 3sg.m-dbl.caus-say\int.pfv-pl=when ‘When they put it down …’

As can be seen from Table 2, what is still the Perfective paradigm of di ‘say’ has become the inflection of the Imperfective of V2 verbs. This is linked to the introduction of a new Imperfective paradigm in Beja, probably prior to the introduction of V2 verb class, which was also partly based on a light verb construction with ‘say’ (for the singular only) and which resulted in a dramatic change in the aspectual system of Beja via a push-chain: the former Perfective became an Aorist (a form which mainly occurs in dependent clauses and as a past habitual) while the former Imperfective became a Perfective as summed up in Table 3.

Tab. 3: The evolution of the TAM system in Beja. Beja today

Reconstructed forms and values

Imperfective Perfective Aorist

*Verbo-nominal + ‘say’ (new paradigm) *Imperfective (“old present”) *Perfective (“old past”)

Grammaticalization in Cushitic, with special reference to Beja

667

Tab. 4: Paradigms of the Imperfective of mono- and disyllabic V1 verbs. V1 IPFV

1sg 2sg.m 2sg.f 3sg.m 3sg.f

Monosyll.

Disyll.

an-CiːC tin-CiːC-a tin-CiːC-i in-CiːC tin-CiːC

a-CaCiːC CaCiːC-a CaCiːC-i CaCiːC CaCiːC

The arguments in favour of such a change can be summarized as follows: (i) The vowel of the Perfective flectional morphemes is -a, while it is that of the Imperfective in most Cushitic languages; (ii) the vowel of the Perfective flectional morphemes is -i in some archaic verbs, as in most Cushitic languages; (iii) some verbs have only two paradigms, the Imperfective and the Aorist (e.g., biri ‘have’, faj ‘be there’); the Present progressive (unknown today) is formed with the Perfective of the core verb and the auxiliaries haj or faj ‘be there’ in the Imperfective. As Table 4 below shows, the prefix and infix morphemes of respectively monosyllabic and disyllabic stems of V1 verbs also include *n ‘say’, with clear traces of pan-Afroasiatic prefixes for the monosyllabic stems: 1sg ʔa- (> a- in Beja), 2sg.m, 2sg.f, 3sg.f ti-, 3sg.m ji- (> i- in Beja). A detailed discussion of the evolution is found in Cohen (1972, 1973) and Zaborski (1975).

3.2 Valency In Beja, voice and valency changing mechanisms are mainly expressed by morphological devices that cannot be reconstructed as the result of grammaticalization. There is only one valency increasing grammaticalized construction which pertains to the domain of applicative, namely a construction with a benefactive auxiliary. It is based on the verb hi ‘give’, as in many languages of the world, but it is a rare device in Afroasiatic. The core verb is either a non-finite form, the General converb, or a finite form, the Perfective (sometimes followed by a linker -i), or the Imperative if the auxiliary is also in the Imperative. (9)

i=ʔaraːw=uːn dhaːj tikʷ-ija-i def.m=friend=poss.1pl.nom dir descend-pfv.3sg.m-l i-hi=heːb 3sg.m-give\pfv=obj.1sg ‘Our friend came down to him for me.’

668

Martine Vanhove

3.3 Aspect 3.3.1 Progressive A Progressive aspect marked by a ‘be there’ auxiliary is reported in Reinisch (1893– 1894) for Northern Beja and in Roper (1928) for Central Beja. It is not attested any longer in my Central and Southern data.

3.3.2 Perfect The Perfect aspect developed through the predication of the Manner converb with the copula in main and independent clauses, or the verb ak ‘become, be’ in negative utterances and dependent clauses. (10) kak jʔ-aː=b=wa? how come-cvb.mnr=indf.m.acc=cop.2sg.m ‘How have you come?’ 3.3.3. Phasal Continuative The Phasal Continuative aspect is the result of the grammaticalization of a spatial construction. The verb of centrifugal movement baj ‘go’ is preceded by the postposition harʔa ‘after’ and by the core verb in the Simultaneity converb form. (11) ʔadaroː-ja=ka=t aka-jeː harʔa i-beː-n=eːb=ka red-pl=cmpr=indf.f become-cvb.smlt after 3-go\int.pfv-pl=rel.m=distr ‘It kept becoming redder and redder all the time.’ 3.3.4. Phasal Terminative The Phasal Terminative is encoded with a posture verb, bʔa ‘lie down’. The core verb may be in the same TAM as the auxiliary or in the form of the General converb. (12) dirar-ti bʔ-eːn=eːb oː=doːr dine\mid-cvb.gnrl lie_down-ipfv.3pl=rel.m def.sg.m.acc=time ‘When they finish dining …’

Grammaticalization in Cushitic, with special reference to Beja

669

3.4 Modality 3.4.1 Intention and volition Intention and Volition are expressed with the Manner converb and the auxiliary verb di ‘say’. (13) baroːk tam-aː=hoːk eː-d-na 2sg.m.acc eat-cvb.mnr=obj.2sg 3-say\ipfv-pl ‘They want/intend to eat you.’

3.4.2 Potential The potential construction comes from the grammaticalization of the verb of centripetal movement jʔ ‘come’ and the core verb with a multifunctional verbo-nominal suffix -at/-it (only used in complex predicates). It often expresses epistemic modalities of possibility, inference or almost certainty. The core verb is followed by the auxiliary, more or less phonologically reduced depending on the person. The full form of jʔ (a highly irregular verb) is used in the 3sg, 1pl and 3pl, it is reduced to flectional morphemes in the 2sg and 2pl, and omitted in the 1sg. (14) dijar-an=eːk ka=a-kan dabal=had be_tired-pfv.1sg=if neg.ipfv=1sg-know\mid.pfv small=until fiːn-at=aj rest-pot.[1sg]=csl ‘I am really exhausted (Lit. I don’t know if I am tired), so I might rest for a while.’ It can also sometimes be interpreted as a future, with a pragmatic connotation of threat. (15) hinin rhi-is-at=oːk eːnej 1pl.nom see-caus-pot=poss.2sg.acc come\ipfv.1pl ‘We’ll show you!’

3.4.3 Obligation The deontic modality of obligation goes back to a bi-clausal construction that brings into play the expression of identity via the copula or the verb ak ‘become, be’, and a marker of complement clauses, which itself goes back to a relative clause with a dummy head noun na ‘thing’: =eːt toː=na (Lit. rel.f + the thing). It is reminiscent

670

Martine Vanhove

of cleft or pseudo-cleft constructions which are quite frequent in Cushitic languages, but absent in Beja except in an explicative function. (16) hankaːj hoːj sajjas-iːn=eːt toː=naː=t=i before abl.3 think-aor.3pl=rel.f def.sg.f.acc=thing=indf.f=cop.3sg ‘They should have thought about it before.’ (Lit. It is the thing that they had thought about before)

3.5 Tense 3.5.1. Future Once again, the verb meaning ‘say’ is used as an auxiliary. In order to express the Future tense, di ‘say’, in the Imperfective as in (17) or in the appropriate paradigm requested by the syntax, is associated to frozen forms of the 1st persons sg and pl of the Aorist of the core verb (glossed as fut and not aor in the examples). Number concord is often preserved, but not systematically and the singular form of the core verb is gaining over the domain of the plural, more systematically in the Southern variety of Beja than in the Central one. (17) kak iː-wr=heːb i-jad-na how fut.sg-do=obj.1sg 3-say\ipfv-pl ‘What are they going to do to me?’

3.5.2 Immediate future/Prospective immediate future is made of the posture verb gad ‘stop, stand’ in a directional construction with the allative postposition dhaːj/=da/=d ‘towards, to’. The core verb is in the form of the masdar (or action noun) in the genitive case. (18) ʔabuːk-i dhaːj=ka i-ngadi=jaːt=ka catch\n.ac-gen dir=distr 3sg.m-stop\pfv=coord=distr ‘Each time he was about to catch it …’

3.6 Emphatic constructions Two auxiliary constructions are used to emphasize an event in order to reinforce either its truth value or its property. The auxiliaries are not interchangeable and are associated with semantic classes of verbs. The first auxiliary dʔi ‘do’ and the second (highly irregular) one ah ‘take’ are used either with a General or a Manner converb or with a finite verb form.

Grammaticalization in Cushitic, with special reference to Beja

671

(19) fidig-a dʔi-jaː=b=u untie-cvb.mnr do-cvb.mnr=indf.m.acc=cop.1sg ‘I did let it go.’ (20) haːj gab gab-eːti i-niːn com be_rich\n.ac be_rich-cvb.gnrl 3sg.m-take\ipfv ‘He becomes really rich with that.’

3.7 Contrastive negation Rules of honour and politeness have led to an overuse of negative forms. For sake of clarification and insistence on the negation, Beja makes use of the auxiliary verb rib ‘refuse’ together with a non-finite core verb with the multifunctional suffix -at/ -it (here glossed vn). For details, see Hamid Ahmed and Vanhove (2004). (21) t=ʔaba=t=i dh=eː tʔi-it def.f=valley=indf.f=poss.1sg.nom dir=poss.1sg.acc resemble-vn ti-rib 3sg.f-refuse\pfv ‘To me, my valley did not look like it.’

3.8 Directional constructions With movement verbs, Beja can (but does not need to) specify the trajectory of the movement on the vertical axes by using two posture verbs ʔas ‘be up’ and nʔa ‘be down’ as auxiliaries. Contrary to the other complex verb constructions, these verbs precede the core verb, a trace of their bi-clausal origin where converb clauses usually precede the main clause. The posture verbs can be in the General converb form, or conjugated in the same TAM as the core verb or, in the Southern variety, have a special form with an -i suffix which only occurs in this context. (22) ʔas-ti far-iːni be_up-cvb.gnrl jump-ipfv.3sg.m ‘He jumps upwards.’ (23) luːl nʔ-i gid-a rope be_down-suf throw-imp.sg.m ‘Throw the rope down!’ On the horizontal axis two different grammaticalized strategies are used. A centrifugal movement is expressed by similar means as the ones on the vertical axis, namely by the General converb of the verb fidin ‘go away’, which is fol-

672

Martine Vanhove

lowed by the core verb (24). The two elements of the complex predicate may be separated by an ablative pronoun indicating the reference point (25). The bare stem of fidin is used with posture core verbs (26). (24) fidin-ti ɖaːb-eːtiːt go_away-cvb.gnrl run-cvb.seq ‘After having run away …’ (25) t=ʔaːrbi fidin-ti hoːsoː gid-ti=hoːb def.f=draw\n.agn go_away-cvb.gnrl abl.3 throw-aor.3sg.f=when ‘When water collecting people were throwing (stones) (to chase him) away from them …’ (26) fidin hoːsi i-sʔa go_away abl.1sg 3sg.m-sit\mid.pfv ‘He stayed away from me.’ For a centripetal movement towards a reference point, a nominal construction is involved, rather than a converb. It makes use of a proximal demonstrative and the noun mari ‘direction’ in the genitive case, without an article, an omission which would be ungrammatical in non-fossilized expressions. (27) w=ʔawi oːn mari-i far-ja def.sg.m=stone prox.sg.m.acc direction-gen jump-pfv.3sg.m ‘He jumped over the rock, in our direction.’ (Lit. in this direction)

4 Grammaticalization of complex constructions Leaving aside deranked complement and adverbial clauses with converbs, most complex syntactic constructions of Beja are the result of various nominalization and relativization processes.

4.1 Relative clauses Apart from a general embedding morpheme =eː, the main relative clause markers are clearly cognate with determiners, the definite and indefinite articles, and they agree in gender, number and (usually) definiteness with their head. Indefinite articles3 may be followed by the dummy noun na ‘thing’ and may combine with the 3 For the sake of clarity, the articles functioning as markers of relative and complement clauses are glossed rel and not def or indf in this section.

Grammaticalization in Cushitic, with special reference to Beja

673

general embedding morpheme. Both types of articles may combine as well, as in (29). Demonstratives are not involved as is commonly the case in African languages (Heine 2011: 706), but it cannot be excluded (although not proven) that the Beja articles ultimately come from former demonstratives. (28) takat miskiːn=t areː [ʔarit mhaj=t tiː-biri=t] woman poor=indf.f then girl\pl three=indf.f 3sg.f-have\aor=rel.f tiː-fi 3sg.f-be_there\aor ‘There was a poor woman then, who had three daughters.’ (29) toːt ti=takat [ti=waw-ti=t] prox.sg.f.acc def.f=woman rel.f=weep-aor.3sg.f=rel.f rh-i=hoːb see-aor.3sg.m=when ‘When he saw this woman who was crying …’ (30) oː=mhiːn [w=ʔani a-ngaːd def.sg.m.acc=place rel.sg.m=1sg.nom 1sg-stop\pfv iː-kti=jeːb] 3sg.m-become\aor=rel.m ‘The place where I had stopped …’

4.2 Complement clauses Complement clauses basically use the same markers as those that form relative clauses, i.e., articles. There are, in addition, dedicated markers for balanced complement clauses: the feminine complex marker =eːt, which goes back to =eː + the feminine indefinite article =t (more rarely just the general embedding morpheme =eː) is followed by the dummy noun in its definite accusative form, toː=na ‘the thing’. na may also be used on its own as a complementizer after the simulative verb, but in a deranked construction (ex. 32) with the manner converb in the complement clause. (31) [waʤʤa-i gaw iː-kti=jeːt toː=na] meeting-gen house 3sg.m-become\aor=rel.f def.sg.f.acc=thing a-kan 1sg-know\mid.pfv ‘I knew it was a meeting house.’

674

Martine Vanhove

(32) dajjara-a na tannʔi be_tired-cvb.mnr thing ressemble\ipfv.3sg ‘He pretends to be tired.’

4.3 Adverbial clauses 4.3.1 Temporal clauses The most frequent temporal conjunction comes from a temporal noun hoːb ‘time’, which survives in the noun phrase oːn hoːb ‘at this present time’, and has become a temporal marker ‘when’, which is directly enclitic to finite verb forms. (33) a-dif=hoːb biri dh=eː i-jaːm 1sg-leave\ipfv=when rain dir=poss.1sg.acc 3sg.m-pour_water\int.pfv ‘When I left, it rained over me.’ Another marker of temporal adverbial clauses is also based on a temporal masculine noun, doːr ‘turn (order of succession), time’, an Arabic loan dawr ‘period, turn’, in its definite form. The construction goes back to a relative clause where doːr behaved as the head. doːr may follow or precede (rarely in my data and always with an additional proximal demonstrative) the relative clause. (34) oː=tak ʃoːb-an=eːb oː=doːr def.sg.m.acc=man greet-pfv.1sg=rel.m def.sg.m.acc=time ‘When I greeted the man …’ (35) oːn oː=doːr wi=kʷiɖ-jaːn=eːb prox.sg.m.acc def.sg.m.acc=time rel.m=disappear-pfv.3pl=rel.m ‘When they disappeared …’ The following two temporal markers are terminative markers which express the end point of a period of time, ‘until’. They also go back to relativized nouns. The first is =eːb hadiːd (Central Beja) where the indefinite masculine noun hadiːd is an Arabic loan meaning ‘limit’, and the second one is =eːb oː=kiːk (Southern Beja), where the definite masculine noun oː=kiːk means ‘the equivalent’. The terminative clause generally follows the main clause. (36) oːn i=kaːm=oːk heː=heːb prox.sg.m.acc def.m=camel=poss.2sg.acc give\imp.[sg.m]=obj.1sg baruːk oː=buːn gʷʔa-ti=eːb hadiːd 2sg.m.nom def.sg.m.acc=coffee drink-aor.2sg.m=rel.m limit ‘Give me your camel until you have drunk the coffee.’

Grammaticalization in Cushitic, with special reference to Beja

675

(37) i=jam-i dar ti-ngad=eːb oː=kiːk def.m=water\pl-gen side 2sg.f-stop\pfv=rel.m def.m.acc=equivalent ‘Until she stopped by the water.’

4.3.2 Conditional clauses Conditional clauses have a dedicated enclitic marker =eːk but when the object of the verb of the protasis is a pronoun, Beja may use the enclitic set of possessive suffixes instead of the object set. This conditional construction is obligatory if the verb is in the Aorist, but optional if it is in the Perfective, for which the possessive set alternates with the =eːk marker and object pronouns. For both verb forms, the use of the possessive set excludes the use of the marker of conditional clauses. Thus, the possessive pronouns in this clause type may be thought of as grammaticalized (portmanteau) markers of conditional clauses. (38) w=haˈwaːd jʔ-i=juːk majʔa rha-tnija def.sg.m=night come-aor.3sg.m=poss.2sg.nom light see-ipfv.2sg.m ‘If night falls upon you, you’ll see light’ (39) aː=jas j-ʔaʃiʃ-n=uːk ʃiːʃik fif-a def.pl.m.nom=dog\pl 3m-face\pfv-pl=poss.2sg.nom thorn pour-imp.sg.m hi-ja give-imp.sg.m ‘If you meet the dogs (Lit. the dogs faced yours), pour thorns for them!’

4.3.3 Concessive clauses Concession is expressed with several polyfunctional devices. Most of them are quite frequent crosslinguistically and none of them is fully grammaticalized as a concessive marker in the language. For instance, Beja may use the conditional enclitic morpheme =eːk ‘if ’ or a temporal device, the Simultaneity converb, which may be followed by the additive focus particle han ‘also’. (40) t=ittifaːgijaːj haː-jeː han ʔaːsiː-jiːni def.f=agreement be_there-cvb.smlt also disobey-ipfv.3sg.m ‘Even if there was an agreement, he could disobey.’ Another possibility is to use the enclitic morpheme =ka, which is a universal quantifier, a distributive marker and a temporal iterative marker when there is no case marker, and a comparative marker when licensing the genitive case. In a concessive

676

Martine Vanhove

sense, in the first construction with =ka the predicate of the adverbial clause is a masdar, while the second construction is inserted in a relative clause. These constructions are not fully grammaticalized and heavily depend on the context for their semantic interpretation. (41) ani haːjloː i-keː-n sar-oːj=ka baroːk 1sg.nom dm 3-become\pfv-pl be_awake-n.ac=distr 2sg.m.acc ti-kati=hoːb geː w=ʔadil doːr 2sg.m-become\ipfv=when dm def.sg.m=reconciliation time hi-it=oːk=aj give-pot.[1sg]=poss.2sg.acc=csl ‘Anyway, in spite of the sleepless nights (Lit. all being awake), since it is you, I grant you the reconciliation.’ (42) geː suːr sak-an=eː=naː=t-i=ka tʔa diːseːt dʔam hoːj dm before do-pfv.1sg=rel=thing=indf.f-gen=cmpr so slowly taste abl.3 a-mri 1sg-find\pfv ‘Anyway, in spite of what I did before, I slowly regained confidence.’ (Lit. compared to the thing that I had done before I slowly found taste from it)

4.3.4 Reason clauses There is a dedicated enclitic morpheme which marks reason clauses, but oblique relative clauses may also express a causal relation when followed by a locative, directional or ablative postposition. Again, this semantic interpretation heavily depends on the context. (43) oːn oː=firʔa prox.sg.m.acc def.sg.m.acc=go_out\n.ac [wi=ti-firʔa=jeː=joː=na-ji=da] rel.m=2sg.m-go_out\pfv=rel=poss.1sg.acc=thing-gen=dir ‘Because you made me go out (I’ll give you something.)’ (Lit. this way out towards which you took me out)

4.3.5 Purpose clauses Two different nominalized constructions form purpose clauses. The first one makes use of the directional postposition dhaːj ‘towards’ which licenses the genitive case. The predicate of the purpose clause is a masdar (or action noun).

Grammaticalization in Cushitic, with special reference to Beja

677

(44) xadaːr mi-jaːj-i dhaːj jʔ-an vegetable n.ac-take-gen dir come-pfv.1sg ‘I came to buy vegetables.’ The second nominalized construction has grammaticalized the masdar of the verb ‘say’, mi-jaːd, as a purpose clause marker. The verb of the purpose clause is either the Manner converb or the Optative finite paradigm. The clause order is most often the reverse of the canonical one. (45) nʔi ti-gdha t=ʔoːr eː=jam be_down 3sg.f-descend\pfv def.f=child def.pl.m.acc=water baj-a mi-jaːd go\int-cvb.mnr n.ac-say ‘The girl went down in order to go to the water.’

4.4 Insubordination Insubordination is part of the origin of most auxiliary constructions where a converb is involved (see Section 3.3). Insubordination is also attested for the expression of an oath and a wish, in particular in exclamatory contexts. The clause is formally marked as a headless relative clause by a relative marker, but the verb of the main clause is omitted in the first case (46), or replaced by a particle in the second one (47). In the case of an oath, the verbs baːʃ or gilad ‘swear’ are regularly omitted by the speakers. (46) [hinin toː=tiji t=asarama girma-ji 1pl.nom def.sg.f.acc=snake def.f=seven head-gen dir-a=b ni-kati=eːb] kill-cvb.mnr=indf.m.acc 1pl-become\ipfv=rel.m ‘(We swear) that we have killed the seven-headed snake!’ For the expression of a wish, the non-predicative discourse particle baːbija, which can roughly be translated as ‘if only’ and only occurs in clause initial position, is used instead of a desiderative verb, followed by the complement clause. (47) baːbija [dibi-a=b tiː-kti=ji] ptcl keep-cvb.mnr=indf.m.acc 3sg.f-become\aor=rel ‘If only you had kept it!’

678

Martine Vanhove

5 Other patterns of grammaticalization and reanalysis 5.1 Postpositions Like in many languages of the world, most postpositions grammaticalized from body parts and spatial nouns: geːb ‘next to’ < ‘side’, suːr ‘before’ < ‘precedence’, dabaːj ‘forward’ < ‘face, front’ + gen, whiː ‘under’ < ‘depth’, dhaːj ‘towards’ < ‘proximity’ + gen, harʔiː ‘behind’ < ‘back part’, bitka ‘between’ < ‘space, gap’, kalawaj ‘inside’ < ‘interior’ + gen. They are all common cross-linguistically. One postposition with an allative meaning grammaticalized from a General converb, that of the movement verb baj ‘go’ > beːti ‘towards’. (48) oː=rba beːti hireːr-eː def.sg.m.acc=mountain towards walk-cvb.smlt ‘While walking towards the mountain …’

5.2 Question words The polar question marker has been recruited from the scalar additive particle han ‘also’. (49) w=handi-i=t nʔandaː =t=iːb ti-ngadi-ja han def.sg.m=tree-gen=indf.f shadow=indf.f=loc.sg 2-stop\ipfv-sg.m q.plr ‘Do you stay in the shadow of the tree?’ The selective, non-selective and causal interrogatives come from the feminine dummy noun na ‘thing’, a process which is also attested, e.g., in Yucatec Maya (Lehmann this volume) for the non-selective interrogative. The non-selective interrogative pronoun (also used as a quantitative interrogative adverb ‘how much?’) is either naːn ‘what?’ (probably a former reduplicated form of the dummy noun ‘thing’ as the following form shows), or a reduplicated form with the feminine indefinite article =t and the copula na~naː=t=i. The selective interrogative pronoun is just the dummy noun in its bare form for masculine with a lengthening of the vowel, naː ‘which?’, and with the indefinite article for feminine, naː=t. In the feminine it is also used as a quantitative (“how much?”) or locative (‘where?’) interrogative adverb. The causal interrogative adverb is also based on the reduplicated form of the selective interrogative but combined with a spatial postposition, either the ablative one naː~naː=t=iː, or the directional one, naː~naːt-i=da ‘why?’4 4 na, in its indefinite form na=t, has also given rise to a negative indefinite pronoun ‘nothing’.

Grammaticalization in Cushitic, with special reference to Beja

679

5.3 Discourse particles 5.3.1 2nd person dative independent pronoun The 2nd person dative independent pronoun hoːk (or =hoːk, the enclitic object pronoun, after the similative morpheme) is used at discourse level as an adverbial intensifier. (50) i=xartuːm=wa winneːt hoːk madiːna=t def.m=Khartoum=coord plenty 2sg.dat town=indf.f bi=t-kaːj=i doːr neg.opt=3sg.f-become\opt=rel time ‘Khartoum and (big cities like that), was not a big city at that time.’ (51) oːn suːr-n=eːt=hoːk miskiːn ak-eːtiːt prox.sg.m.acc before-l=simil.pl=obj.2sg poor become-cvb.seq ‘After he had become poor as before …’

5.3.2 Perfective 3pl of di ‘say’ The quotative verb di ‘say’ in the Perfective 3rd person plural, eːn, is frequently used to mark the end of a discursive unit (of whatever length) in narratives and poetry. (52) uːn uː=tak kʷakʷar=t tambalʔaːj prox.sg.m.nom def.sg.m.nom=man adder=indf.f lizard sankʷi=t rhi-iːni eːn chase\ipfv.[3sg.m]=indf.f see-ipfv.3sg.m say\pfv.3pl ‘This man sees an adder that is chasing a lizard, (Lit.) they said.’ Japanese also uses a quotative in a comparable way, as a ‘sentence-final’/’turn-final’ particle (Thompson and Suzuki 2011: 676). The verb ‘say’ is also integrated in various complex sentence-final particles in Korean (Rhee 2011: 768).

6 Comparative outlook A third of the above-mentioned grammaticalization patterns found in North-Cushitic are at least sporadically attested in other Cushitic languages. On the other hand, Beja lacks other grammaticalizations that occurred in the rest of the family and, beyond that, in the languages of North-East Africa from other families and phyla

680

Martine Vanhove

where several convergence phenomena are attested (for Ethio-Semitic and Highland East Cushitic, see Crass and Meyer [2008]). It is impossible to list them all for lack of space, but in what follows, I will briefly outline the commonalities and most salient differences between North-Cushitic and a sample of languages from its closest relatives, Lowland East Cushitic (LEC: Afar, Arbore, Dhaasanac, Konso, Oromo, Saho, Southern Somali [= Tuni]), Highland East Cushitic (HEC: Alaaba, Hadiyya, K’abeena, Kambaata, Sidaama), and Central Cushitic (Agaw) (CC: Awngi, Xamtanga), on the basis of available descriptions and Cushitic comparative studies.

6.1 Nominal categories None of the above three nominal categories in Beja have grammaticalized from the same source in other Cushitic languages. Cushitic languages have undergone various innovations for personal pronouns, but none of them involves a have-possessive construction. Demonstratives may sometime have played a role in the renewal process of personal pronouns but differently from Beja. This is hypothesized for HEC Alaaba by Schneider-Blum (2007: 169), for which she assumes that the demonstratives were attached to pronouns to become case/gender markers. If this analysis is correct, then it is not restricted to personal pronouns, but it also concerns nouns (Y. Treis, personal communication). Noteworthy from a crosslinguistic perspective are the new plural forms in HEC Kambaata which resulted from the reinforcement of the old plural pronouns by an associative plural morpheme (Treis 2008a: 333). This is also the case in CC Xamtanga (Darmon 2015: 96). As often in the languages of the world, and notably in African languages (Heine 2011: 700–702; König and Siemund 2013), reflexive emphatic pronouns stem from :various body parts such as ‘neck’ (CC Xamtanga, Darmon [2015: 101]), ‘hand’ or ‘head’ (LEC Konso, Ongaye 2013: 134), or from ‘person’ gagá < ‘very close person, person of one’s kin-group, person who is like oneself ’ (e.g., HEC Kambaata, Treis [2008a: 217–218], and Alaaba, Schneider-Blum [2007: 188]). Treis (p.c.) mentions that gaga also means ‘owner’ in Kambaata, as in the Beja emphatic pronouns. The source ‘member’ is not attested elsewhere in Cushitic in the literature I consulted, nor is it attested in Arabic, Tigre and Nubian, the contact languages of Beja. There is rarely any information about indefinite pronouns in the grammars I consulted. In HEC Kambaata (Treis 2008a: 395) and K’abeena (Crass 2005: 134) they stem from the numeral ‘one’ as in Beja. K’abeena has in addition ‘people’, and Kambaata interrogatives. In CC Xamtanga (Darmon 2015: 108), interrogatives (for negation only) are also a source for indefinite pronouns as well as the universal quantifier + a proximal demonstrative.

Grammaticalization in Cushitic, with special reference to Beja

681

6.2 Verbal categories 6.2.1 Renewal of paradigms and light verb constructions As mentioned in Section 3.1, it is generally considered that the renewal of the finite verb paradigms with flectional suffixes in other Cushitic languages also partly go back to a light verb construction with a verb meaning ‘say’ conjugated with prefixes (for a different hypothesis based on phonological and analogical processes of inherited cognates of the Afroasiatic stative conjugation, see Banti [2004]). According to Banti (2004: 33–39), in most HEC and CC languages, the suffix verb type of Beja, which he supposes was inherited from the Afroasiatic stative, does not occur in independent or main clauses but can be traced back in converb forms. Cushitic languages also use light verb constructions (LVC) with pragmatic values which vary to some extent from one language to another (for detailed surveys in Cushitic and beyond see Appleyard [2001]; Cohen, Simeone-Senelle, and Vanhove [2002]; Güldemann [2005]; Meyer [2009]). ‘Say’ LVC-s may be used with ideophones (e.g., LEC Konso, Ongaye [2013: 247–253]; HEC K’abeena, Crass [2005: 229]) and borrowings, or with ideophones and nouns, then extended to intransitive verbs (transitive verbs are often used in LVC-s with a transitive auxiliary such as ‘do’ or ‘put’ for instance), or may be attested for (almost) all parts of speech, including transitive verbs, sometimes together with derived forms of ‘say’ (e.g., CC Xamtanga, Darmon [2015: 289]). They can also be more integrated phonologically and morphologically into the verbal complex for certain word classes, as e.g., in LEC Saho with ideophones and nouns (Banti and Vergari 2005: 108). According to Güldemann, what makes the constructions in Cushitic languages, as well as Semitic and Nilo-Saharan languages of North-East Africa, different from similar constructions elsewhere in the world is the co-occurrence of a set of features that are reproduced hereafter: 1. A semantically generic verb serves as the inflectional basis of complex predicates; 2. The respective verb is usually also used in reported discourse; 3. The range of content signs converted into predicates/verbs is fairly wide; 4. The auxiliary occurs after the content sign; 5. The complex predicates tend to merge to a one-word sign. (Güldemann 2005: 141)

6.2.2 Valency ‘Give’ as an applicative auxiliary with a benefactive value is also attested in one CC language, Xamtanga (Darmon 2015: 194–196), but, contrary to Beja, it belongs to a three-term applicative system which has recruited two other auxiliaries meaning ‘leave, abandon’ (with a malefactive reading) and ‘say’. The ‘give’ strategy, frequent

682

Martine Vanhove

cross-linguistically, is absent in the other Cushitic languages of my sample. Instead, they often have a benefactive verbal derivation or have recourse to a periphrastic construction with a converb of ‘say’ and a copula, an areal feature in languages of Ethiopia, as for instance in HEC K’abeena and Libido (Crass and Meyer 2008: 243).

6.3 Aspect 6.3.1 Progressive Progressive is often grammaticalized in Cushitic languages, stemming from a construction with subordinate core verbs and ‘be’ verbs (LEC Afar, Bliese 1981: 117; Konso, Ongaye 2013: 161), or copulas (CC Xamtanga – together with a durative morpheme, Darmon [2015: 209]; HEC Alaaba, Schneider-Blum [2007: 113, 220]).

6.3.2 Perfect In Cushitic languages that have developed a new Perfect aspect (sometimes in addition to an already existing Perfect), constructions with converbs and auxiliaries have been recruited. One of the CC Xamtanga constructions is similar to Beja, namely the ‘generic’ Perfect formed with a converb and an existential semi-copula (Darmon 2015: 210). The other construction which marks the “experiential Perfect” is based on a converb and the verb ‘know’ (Darmon 2015: 212), as in HEC K’abeena (Crass 2005: 192) or Alaaba (Schneider-Blum 2007: 253) for instance, an areal feature of the Ethiopian area (Crass and Meyer 2008: 244). In all these languages, the “experiential Perfect” co-exists with a non-periphrastic Perfect (in opposition to Perfective and Imperfective). In LEC Afar, the Perfect is made of an auxiliary meaning ‘have’ and a converb (Bliese 1981: 119).

6.3.3 Phasal Terminative Only one grammar mentions a grammaticalized Phasal Terminative construction. In HEC K’abeena (Crass 2005: 190) (and also in HEC Kambaata, Y. Treis p.c.), the auxiliary construction is built from a verb meaning ‘stand up, get up’, the antonym of Beja ‘lie down’.

6.4 Modality 6.4.1 Intention and volition CC Xamtanga (Darmon 2015: 283–284) and HEC K’abeena (Crass 2005: 321) are the sole other Cushitic languages where I could find mention of ‘say’ as a device to

Grammaticalization in Cushitic, with special reference to Beja

683

express intention and volition. In Xamtanga, the formal means involve a purposive subordinate or a jussive verb form introduced by the converb form of ‘say’, while in K’abeena these modalities are just contextual and not grammaticalized.

6.4.2 Obligation The expression of deontic obligation by a bi-clausal (pseudo-)cleft construction with an identificational copula or ‘be’ verb is also attested in two LEC languages, Afar (Bliese 1981: 55) and Somali (Tosco 1997: 136). Some other languages have grammaticalized obligation via ‘exist’, or verbs with an intermediate stage of possessive ‘have’ constructions with a verbal noun (but not with a cleft or relative clause), such as HEC K’abeena (Crass and Meyer 2008: 245), or ‘be’ verbs with a purposive paradigm as in CC Xamtanga (Darmon 2015: 223).

6.5 Tense 6.5.1 Immediate future/Prospective None of the other branches of Cushitic has grammaticalized the verb ‘say’ as a Future auxiliary. HEC Kambaata, Alaaba, K’abeena, and Sidaama (Treis 2012) all have same subject purpose-cleft constructions that are used to express imminent and/or intended future events. Other future constructions are based on dative-marked verbal nouns which have also given rise to either imminent or intended future, or both. This is the case in one LEC language, Oromo, and one HEC language. It involves a verbonominal form with a dative case and a copula (Crass and Meyer 2008: 241–242).

6.6 Complex constructions 6.6.1 Relative clauses and cleft constructions Relative clauses or cleft constructions involving nominalizations via the dummy noun ‘thing’ are also known in LEC Sidaama (Kawachi 2007: 620) and Afar (Simeone-Senelle, Vanhove, and Houmedgaba 2000). The determiner strategy is attested in LEC Dhaasanac (Tosco 2001: 282) for restrictive relative clauses and CC Awngi (Hetzron 1969: 17) for cleft sentences. HEC Alaaba (Schneider-Blum 2007: 363), and Kambaata (Treis 2008b: 165–166) have only prosodic cues. In Kambaata, a stress shift to the final position is accompanied by the voicing of the final otherwise devoiced vowel of the verb in main clauses

684

Martine Vanhove

for verb forms ending in a vowel, and for some particular forms by the loss of submorphemic glottal appendices. In all branches there are often specialized verb forms to mark relative clauses. In CC Agaw for instance these paradigms go back to auxiliary verbs meaning ‘be’ or ‘say’ (Appleyard 2002). But in HEC Sidaama, Hadiyya and Kambaata, Hudson (1976: 269) a final vowel is dropped (+ in Hadiyya the 2nd subject agreement marker).

6.6.2 Complement clauses Complement clauses are built upon (headless) relative clauses in LEC Arbore (Hayward 1984: 319–320), Somali (Tosco 1997: 132), Afar (Bliese 1981: 13), and Dhaasanac (Tosco 2001: 286), where the verb of the complement clause is nominalized as a verbal noun, and in CC Xamtanga (Darmon 2015: 239). Similar to Beja, Afar also involves a suffix whose origin is a dummy noun ‘thing’ (Bliese 1981: 13). In HEC Alaaba, Schneider-Blum’s (2007: 391) interpretation of the conjugated converb form of ‘say’ as a complementizer after verbs of propositional attitude (‘think’, ‘doubt’ and ‘hope’) is debatable. It resembles a pan East-Cushitic use of direct reported speech in this context introduced by the quotative verb at a converb form which is not analyzed as a complementizer for other HEC languages (Y. Treis, p.c.). More frequent in Cushitic is the similative morpheme ‘like’ as a source of markers of complement clauses (Treis 2017).

6.6.3 Adverbial clauses Usually alongside other devices, adverbial clauses derived from relative clauses or complement clauses are attested in all three Cushitic branches, e.g., in HEC Arbore (Hayward 1984: 319–320), and Kambaata (Treis 2012: 189), in LEC Somali (Tosco 1997: 136), Dhaasanac (Tosco 2001: 282–286), and Konso (Ongaye 2013: 231–241), and in CC Xamtanga (Darmon 2015: 237).

6.6.3.1 Temporal clauses As in Beja, temporal clauses involve a head noun meaning ‘time’ or ‘period’ (not cognate with the Beja nouns) in HEC Alaaba (Schneider-Blum 2007: 375), and Kambaata (Treis 2012: 191), in LEC Somali (Tosco 1997: 136), Afar (Bliese 1981: 62), Arbore (Hayward 1984: 569), and Dhaasanac (Tosco 2001: 284), and in CC Xamtanga (Darmon 2015: 248). In e.g., LEC Konso a relative pronoun without a head noun may just also be used (Ongaye 2013: 238). The marker of Terminative temporal clauses in HEC Alaaba (Schneider-Blum 2007: 376, n. 178) stems from a boundary verb meaning ‘reach’.

Grammaticalization in Cushitic, with special reference to Beja

685

6.6.3.2 Reason clauses Reason clauses may involve spatial and/or directional devices and relative clauses in HEC, LEC and CC. This is the case, e.g., in HEC Alaaba (Schneider-Blum 2007: 373) with ‘place’ and a dative case. HEC Hadiyya, uses a head noun meaning ‘side’ (Sim 1989: 315), and LEC Dhaasanach (Tosco 2001: 286) also. HEC Kambaata (alongside with two other constructions) has a relative construction with the dative case of a noun meaning ‘size, extent, amount; capability, ability’ (cognate with Hadiyya ‘side’) (Treis 2012: 196). In CC Awngi (Hetzron 1969: 20–21), there is no head noun but a dative/ instrumental case suffix on the verb of the relative clause. In LEC Afar, a relative clause and a suffix originating from the dummy noun ‘thing’ is used, but, unlike Beja, without a directional or spatial device (Bliese 1981: 27). 6.6.3.3 Concessive clauses As Beja, Kambaata marks concessive clauses by a (scalar) additive particle (which consists in the lengthening of the final vowel) (Treis 2015). It is also probably the case in HEC Alaaba (Y. Treis p.c.). Schneider-Blum (2007: 372) analyzes this lengthening as an “emphatic suffix”. The additive strategy does not occur in the other sources I consulted. 6.6.3.4 Purpose clauses Cushitic languages frequently use several strategies to encode purpose clauses. As Beja and many languages of the world, most of them use a directional/dative strategy. Treis (2011: 4) mentions LEC Sidaama, Oromo, HEC Kambaata, Alaaba, Kabeena, Hadiyya, probably also HEC Gedeo and Burji. This is also the case in LEC Somali (Tosco 1997: 136) with an adposition meaning ‘to’, and in CC Xamtanga (Darmon 2015: 255) with a dative case. Most Cushitic languages, especially HEC ones, also have dedicated verb forms for purposive adverbial clauses (see, e.g., Treis [2010] for Kambaata). Alternatively, as Treis (2017) shows, what is massively recruited for the expression of purpose clauses in Cushitic languages (sometimes limited to negative purpose clauses) is the similative marker (often derived from a manner deictic), e.g., in CC Awngi and Xamtanga, in HEC Alaaba, Hadiyya, K’abeena, Kambaata, and Sidaama, in LEC Oromo, and Somali, and beyond, in Semitic and Omotic languages of the area.

686

Martine Vanhove

6.7 Others 6.7.1 Adpositions All languages have adpositions whose sources are body parts terms, a frequent grammaticalization process cross-linguistically. Depending on the language and the adposition, they may be more or less grammaticalized. ‘Back’, ‘mouth’, chest’, ‘head’, ‘belly’, ‘bottom’, ‘eyes’, ‘waist’, ‘body’, ‘forehead’ have often been recruited as relational devices, but also locational nouns such as ‘top’, ‘ground’, ‘direction’, ‘inside’, ‘front’, ‘side’, ‘middle’, ‘proximity’, ‘distance’, as well as, in CC Xamtanga (Darmon 2015: 45), ‘sky, up’ and ‘down’. ‘Opening’ is found as a grammaticalized locative postposition in LEC Saho (Banti and Vergari 2005: 124) but with a meaning different from Beja ‘gap’ > ‘between’, namely ‘before’. HEC Kambaata, according to Treis’s (2008a: 202) analysis, has no adpositions, but spatial nouns (in addition to cases) are used to express spatial relations between figure and ground, among them a noun meaning ‘middle, center, space between’ to express ‘between, in the middle of ’ (for a general overview of this path in African languages, see Heine [2011: 698– 699]). A few adpositions come, or may come, from verbs or copulas: in LEC Afar, the comitative postposition comes from the verb ‘have’ (Bliese 1981: 74), and in CC Xamtanga two locative postpositions meaning ‘in, inside’, and ‘on, in’ may come from a copula for the former, and a verb meaning ‘draw near’ for the latter (Darmon 2015: 45). None of the languages of the sample seem to have grammaticalized the verb ‘go’ into a postposition.

6.7.2 Question words Most Cushitic languages of the sample have only recourse to intonation to mark polar questions. When a polar question marker exists, it does not seem to be related to an additive particle. As for content question words, no other Cushitic language seem to have grammaticalized a dummy noun ‘thing’ in this function.

7 Discussion and conclusion Table 5 below compares the grammaticalization paths discussed for all four branches of Cushitic studied in this chapter, organized in alphabetical order of sources. + indicates the presence of a particular path in at least one language of the group, – its absence. Empty cells indicate lack of information in the literature I consulted. The right-hand column mentions the presence (Y) or absence (N) in either Heine

687

Grammaticalization in Cushitic, with special reference to Beja

Tab. 5: Comparison of grammaticalization paths in four branches of Cushitic. Source

Target

NC

CC

LEC

HEC

H&K

2nd person pronoun Additive particle

Discourse particle Polar question marker Concessive clause Purpose clause Plural (in independent personal pronouns) Progressive Relative verbs Deontic obligation Vertical verbal periphrasis Vertical verbal periphrasis Reflexive emphatic Pronouns/ intensifiers Adpositions Temporal Terminative clause Potentional, future Adverbial clauses Perfect paradigm Progressive immediate future/Prospective immediate future/Prospective Future Deontic obligation Purpose clause Centripetal verbal periphrasis

+ + + + –

– – – +

– – +

– + + +

N N Y Y

(+) – + + + +

– + – – – +

+ – + – – +

– + – – – +

Y N N N N Y

+ + + + + – – – – – – +

+

+

– + + + – + + + +

– +

+ + – +

Y Y Y

– +

+ +

– –

– –

+ –

+ –

+ – –

– – – + – –

– – + – – –

+ – – – – –

– – – –

– + + –

– – – +

Allative adposition Associative Be Be + relative Be down Be up Body parts

Boundary/reach Come Complement clauses Copula Copula + dative Copula + purposive

Dative Demonstrative + direction Demonstratives Determiners DEF & INDF Do Equivalent Exist > have Gap, opening Give Go Go away Have-possession

Insubordination

Independent personal pronouns Relative marker complementizer Emphatic auxiliary Temporal Terminative clause Deontic obligation Adposition ‘between’ Adposition ‘before’ Benefactive Phasal Continuative Allative adposition Centrifugal verbal periphrasis Independent personal pronouns Perfect Comitative Perfect paradigm Auxiliary constructions with converbs Perfect paradigm Wish Oath

+ + + + + – + – + + + + + – – – + + + +

Y Y Y Y Y Y N N Y Y Y Y N Y N N Y Y Y N N Y N Y

688

Martine Vanhove

Tab. 5 (continued) Source

Target

NC

CC

LEC

Interrogative Know Lie One Owner

Indefinite pronoun Perfect Phasal Terminative Indefinite pronoun Reflexive emphatic pronouns/ intensifiers Conditional clause Contrastive negation Adverbial clauses Quotative Light verb construction Verbal flectional morphemes Relative verbs Intention Volition Future Purpose clause marker Discourse marker Benefactive Purpose clause Phasal terminative immediate future/Prospective Emphatic auxiliary Complementizer Relativizer Question words Negative indefinite pronoun Temporal clause Temporal clause

– – + + +

+ + –

– –

–

+ + + + + + – + + + + + – – – + + + + + + + +

Possessive pronouns Refuse Relative clauses Say

Similative Stand Stand + after Take Thing

Time Time + relative marker

HEC

H&K

–

+ + – + –

Y N N Y Y

– – + + + + + + + – –

– – + + + + – – – – –

– – + + + + – – – – –

N N

+ + – –

+ + – –

+ + + –

– – –

– + –

– – –

– +

+ +

– +

Y [Y] [Y] N N N N Y [Y] N N N N N Y N Y Y Y

and Kuteva’s (2002) lexicon,5 or in other literature about the languages of the Horn of Africa when in between square brackets [N]. An empty cell in this column just means that Heine and Kuteva’s lexicon is not concerned by these grammaticalization processes. A striking feature of Cushitic is the vast array of grammaticalizations (ten in Table 5) displayed by the ‘say’ verbs in all four branches. The co-occurrence of cross-linguistically attested ones (quotative index, light verb constructions) has already been noticed in the literature on Cushitic and the Ethiopian area and related

5 Heine and Kuteva, together with their Chinese colleagues, are preparing an augmented version of their dictionary, which was not issued at the time of the writing of this paper, thus some of the N in the right hand column might have to be changed to Y.

Grammaticalization in Cushitic, with special reference to Beja

689

to other features (renewal of verb paradigms, position of the light verb, large combinatorial properties, degrees of grammaticalization) as typical of the North-East African area (see Section 6.2.1), including also Semitic and Nilo-Saharan languages. In particular, the renewal of verb paradigms with a ‘say’ verb is also attested in Nubian (Armbruster 1960), a language of Sudan in contact with Beja. Nonetheless these two grammaticalization paths are among the few that are shared by all four Cushitic branches. Moreover, there are additional grammaticalizations that seem particular to this genetic stock and to some extent to Ethio-Semitic (unrecorded in Heine and Kuteva 2002), and they concern different levels of language structure. Three of them occur only in Beja at the verb level (future, unrecorded in other languages as far as I know but clearly linked with the intentional function), at the complex sentence level (marker of purpose clause), and at discourse level (end of discourse unit). Two concerning verbal modality are shared by Beja with CC (volition and intention), one only concerns CC for verb paradigms and complex sentences (relative verbs), and one (benefactive auxiliary) concerns three branches except NC. Semantically, there is in the Cushitic languages of the sample a pervasive link of ‘say’ with ‘purpose’ at large (intention and volition included), particularly developed in Beja and Xamtanga. As Darmon (2015: 221, 290–284) suggests, this development is ultimately linked to the domain of internal awareness which represents the semantic starting point of the grammaticalization path. As far as the benefactive target (not necessarily with a deputative reading) is concerned, one should note that it is not recorded in Heine and Kuteva (2002), making this feature a good candidate for a largely pan-Cushitic innovation. The list provided in Table 5 actually does not cover all the functions of ‘say’ verbs in Cushitic languages, and for instance CC Xamtanga (Darmon 2015: 280) has, in addition to those of naming, reported evidence, explanation, negative thoughts attribution, formation of manner deictics, formation of manner/circumstantial phrases, and a valency-changing mechanism: imperative causation. Noteworthy is the fact that it seems that no Cushitic language testifies of a complementizer stage. Of course the grammaticalization of ‘say’ is not the whole story (like any other grammaticalization), and the constructions in which both the speech verb and verbo-nominal forms occur also play a role in their extensions to other functional domains. In an Afroasiatic perspective, it is important to mention that in Cushitic, auxiliaries have often been recruited as the source of the renewal and enrichment of the verbal system, but this is far less the case for verbless clauses as opposed to some other Afroasiatic families such as Semitic or Egyptian (Cohen 1984). Fifteen (sometimes only ongoing) grammaticalization paths of Beja are unrecorded in Heine and Kuteva (2002) as well as in other Cushitic languages, relisted below for sake of convenience. 1. Additive particle ‘also’ > polar question marker 2. ‘Be up’ (stative) > trajectory on vertical axis 3. ‘Be down’ (stative) > trajectory on vertical axis

690 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Martine Vanhove

Demonstrative + ‘direction’ > centripetal trajectory on the horizontal axis ‘Gap’ > adposition ‘between’ ‘Go away’ > centrifugal trajectory on the horizontal axis ‘Equivalent’ > marker of terminative adverbial clause Have-possession > personal pronouns ‘Lie down’ > phasal terminative auxiliary Possessive pronouns on verbs > conditional ‘Refuse’ > contrastive negation ‘Say’ > future tense ‘Say’ > marker of end of discourse unit ‘Stand’ ‘+ after’ > immediate future ‘Take’ > emphatic auxiliary

The functional extensions of most of them are quite transparent semantically, sometimes closely related to other grammaticalization paths mentioned in the literature such as genitive-possession > pronouns, or the incorporation of spatial terms in verb paradigms, and deserve no further comment. More unexpected are the (ongoing) grammaticalizations of ‘also’ as a marker of polar questions, ‘take’, and possessive pronouns for which I can offer no explanation yet. Among the 70 grammaticalizations in Cushitic languages studied in this chapter, 18 occur only in NC, and a further 12 cannot be assessed for lack of information on these domains in the grammars consulted. Among the remaining 40, Beja shares 21 of them with at least one language in any of the other three Cushitic branches. Eight concern all four branches. There are more shared features between NC, LEC and CC (11, 5 NC+LEC, 4 NC+CC, 2 NC+CC+LEC) than between NC and HEC (4), which mirrors lexico-statistic studies on the distance between the various subgroups of Cushitic. ‘Be, exist’ verbs and copulas are often recruited as markers of deontic obligation including in CC and HEC, but it is noteworthy that only NC and LEC have recourse in addition to a relative/cleft construction for this purpose. It was not possible within the space constraint of this chapter to review all grammaticalization paths attested in the other three branches of Cushitic and those, that do not occur in NC, but some of them seem to be good candidates for shared innovations, either genetically or areally, and to form a bundle of features that, taken together, could be diagnostic for further research on diachronic evolutions and convergence phenomenon in North-East Africa. Particular instances of these, that seem to be (almost?) restricted to these languages and area, are the recruitment of ‘be’ and ‘say’ verbs for the creation of relative verb paradigms, ‘know’ for the perfect aspect, ‘say’ as a benefactive marker, and similative morphemes as purpose clause markers (without a quotative marker intermediate stage).

Grammaticalization in Cushitic, with special reference to Beja

691

Acknowledgements My thanks are due to the editors of the volume and to my colleague Yvonne Treis for their valuable and constructive comments. I also gratefully acknowledge the financial support of the Llacan (UMR 8135, CNRS – INALCO – Université Sorbonne Paris-Cité), the ANR projects CorpAfroAs and CorTypo (PI Amina Mettouchi), and the Labex EFL Empirical Foundations of Linguistics ANR-10-LABX-0083) for my research on the Beja language.

Abbreviations abl = ablative, acc = accusative, aor = aorist, caus = causative, cmpr = comparative, coord = coordination, cop = copula, csl = causal, cvb = converb, dbl.caus = double causative, def = definite, deict = deictic, dir = directional, distr = distributive, dm = discourse marker, f = feminine, fut = future, gen = genitive, gnrl = general, imp = imperative, indf = indefinite, int = intensive, ipfv = imperfective, l = linker, loc = locative, m = masculine, mid = middle, mnr = manner, n.ac = action noun (masdar), n.agn = agent noun, nom = nominative, obj = object, pfv = perfective, pl = plural, poss = possessive, pot = potential, pro.refl = emphatic reflexive pronoun, prox = proximal, q.plr = polar question marker, rel = relator, seq = sequential, sg = singular, smlt = simultaneity, suf = suffix, vn = verbo-nominal.

References Almkvist, Herman. 1881–1885. Die Bischari-Sprache Tuu-Beḍaawie in Nord-ost-Afrika. Uppsala: Kaiserliche Akademie der Wissenschaften. Appleyard, David. 2001. The verb “to say” as a verb “recycling device” in Ethiopian languages. In Andrzej Zaborski (ed.), New data and new methods in Afroasiatic Linguistics: Robert Hetzron in memoriam, 1–11. Wiesbaden: Harrassowitz. Appleyard, David. 2002. The morphology of main and subordinate verb form marking, with special reference to Ethiopian Semitic and Agaw. Afrikanistische Arbeitspapiere 7. 9–31. Appleyard, David. 2004. Beja as a Cushitic Language. In Gabor Takács (ed.), Egyptian and SemitoHamitic (Afro-Asiatic) Studies in Memoriam W. Vycichl, 173–194. Leiden: Brill. Armbruster, Charles H. 1960. Dongolese Nubian, a grammar. Cambridge: Cambridge University Press. Banti, Giorgio. 2004. New perspectives on the Cushitic verbal system. In Andrew Simpson (ed.), Proceedings of the Twenty-Seventh Annual Meeting of the Berkeley Linguistics Society (March 22–25, 2001) – Special Session on Afroasiatic Languages, 1–48. Berkeley: Berkeley Linguistics Society. Banti, Giorgio & Moreno Vergari. 2005. A sketch of Saho grammar. Journal of Eritrean Studies 4. 100–131. Bechhaus-Gerst, Marianne. 1985. ‘Du bist, was du hast’ – Zur Enstehung neuer Personalpronomen im Tu Bedawye (Beja). Afrikanistische Arbeitspapiere 1. 125–129. Bliese, Loren F. 1981. A generative grammar of Afar. Arlington: The Summer Institute of Linguistics and the University of Texas.

692

Martine Vanhove

Cohen, David. 1972. La mutation aspectivo-temporelle dans quelques langues couchitiques et le système verbal chamito-sémitique. In Jacqueline Thomas & Lucien Bernot (eds.), Langues et techniques, nature et société, 57–63. Paris: Klincksieck. Cohen, David. 1973. Le renouvellement de l’inaccompli en bédawiye. Comptes Rendus du GLECS 14. 69–78. Cohen, David. 1984. La phrase nominale et l’évolution du système verbal en sémitique. Etude de syntaxe historique. Leuven: Peeters. Cohen, David. 1988. Couchitique-omotique. In David Cohen (ed.), Les langues dans le monde ancien et moderne. Langues Chamito-Sémitiques, 243–269. Paris: Editions du CNRS. Cohen, David, Marie-Claude Simeone-Senelle & Martine Vanhove. 2002. The Grammaticalization of ‘say’ and ‘do’: An areal phenomenon in the Horn of Africa. In Tom Güldemann & Manfred von Roncador (eds.), Reported speech: A meeting ground for different linguistic domains, 227–251. Amsterdam: John Benjamins. Crass, Joachim. 2005. Das K’abeena. Deskriptive Grammatik einer hochlandost-kuschitischen Sprache. Köln: Köppe. Crass, Joachim & Ronny Meyer. 2008. Ethiopia. In Bernd Heine & Derek Nurse (eds.), A linguistic geography of Africa, 228–250. Cambridge: Cambridge University Press. Creissels, Denis. This Volume. Grammaticalization in Manding languages. In Walter Bisang & Andrej Malchukov (eds.), Grammaticalization Scenarios. Areal Patterns and Cross-Linguistic Variation. A Comparative Handbook. Berlin: De Gruyter, Mouton. Darmon, Chloé. 2015. A morphosyntactic description of Xamtanga: An Agaw (Central Cushitic) language of the northern Ethiopian highlands. Lyon, France: Université Lyon 2 dissertation. Güldemann, Tom. 2005. Complex predicates based on generic auxiliaries as an areal feature in Northeast Africa. In F. K. Erhard Voeltz (ed.), Studies in African linguistic typology, 131–154. Amsterdam: Benjamins. Güldemann, Tom. 2008. Quotative indexes in African languages: A synchronic and diachronic survey. Berlin: Mouton de Gruyter. Hamid Ahmed, Mohamed-Tahir & Martine Vanhove. 2004. Contrastive negation in Beja: the auxiliary verb rib. Afrika und Übersee 85. 81–98. Hayward, Dick. 1984. The Arbore language. A first investigation. Including a vocabulary. Hamburg: Buske. Heine, Bernd. 2011. Grammaticalization in African languages. In Bernd Heine & Heiko Narrog (eds.). The Oxford handbook of grammaticalization, 696–707. Oxford: Oxford University Press. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Hetzron, Robert. 1969. The verbal system of Southern Agaw. Berkeley, Los Angeles: University of California Press. Hetzron, Robert. 1980. The limits of Cushitic. Sprache und Geschichte in Afrika 2. 7–126. Hudson, Grover. 1976. Highland East Cushitic. In Lionel M. Bender (ed.), The Non-Semitic languages of Ethiopia, 232–277. East Lansing: African Studies Center, Michigan State University. Kawachi, Kazuhiro. 2007. A grammar of Sidaama (Sidamo). A Cushitic language of Ethiopia. New York: Department of Linguistics, University of Buffalo, the State University of New York dissertation. König, Ekkehard & Peter Siemund (with Stephan Töpper). 2013. Intensifiers and reflexive pronouns. In Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http:// wals.info/chapter/47 (accessed 16 February 2017). Lamberti, Marcello. 1991. Cushitic and its classifications. Anthropos 86. 552–561.

Grammaticalization in Cushitic, with special reference to Beja

693

Lehmann, Christian. This Volume. Grammaticalization in Yucatec Maya. In Walter Bisang & Andrej Malchukov (eds.), Grammaticalization Scenarios. Areal Patterns and Cross-Linguistic Variation. A Comparative Handbook. Berlin: De Gruyter, Mouton. Meyer, Ronny. 2009. The quotative verb in Ethiosemitic languages and in Oromo. In Joachim Crass & Rony Meyer (eds.), Language contact and language change in Ethiopia, 17–42. Köln: Köppe. Ongaye, Oda Orkaydo. 2013. A grammar of Konso. Utrecht: LOT Publications. Reinisch, Leo. 1893–1894. Die Beḍauye-Sprache in Nordost-Afrika, 3 vol. Wien: Sitzungs Berichte der Kais. Akademie der Wissenschaften in Wien L, Band CXXVIII. Reinisch, Leo. 1895. Wörterbuch der Beḍawiye-Sprache. Wien: Hölder. Rhee, Seongha. 2011. Grammaticalization in Korean. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of grammaticalization, 764–774. Oxford: Oxford University Press. Rilly, Claude. 2014. Language and ethnicity in ancient Sudan. In Julie R. Anderson & Derek Welsby (eds.), The fourth cataract and beyond. Proceedings of the 12th International Conference for Nubian Studies, 1169–1188. Leuven: Peeters. Roper, E. M. 1928. Tu Beḍawie. An elementary handbook for the use of Sudan Government officials. Hertford: Austin. Schneider-Blum, Gertrud. 2007. A grammar of Alaaba. Köln: Köppe. Simeone-Senelle, Marie-Claude, Martine Vanhove & Makki Houmedgaba. 2000. La focalisation en afar. In Bernard Caron (ed.), Topicalisation et focalisation dans les langues africaines, 289– 309. Louvain: Peeters. Thompson, Sandra A. & Ryoko Suzuki. 2011. The grammaticalization of final particles. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of grammaticalization, 668–680. Oxford: Oxford University Press. Tosco, Mauro. 1997. Af Tunni: grammar, texts, and glossary of a southern Somali dialect. Köln: Köppe. Tosco, Mauro. 2000. Cushitic overview. Journal of Ethiopian Studies 33(2). 87–121. Tosco, Mauro. 2001. The Dhaasanac language: Grammar, texts, vocabulary of a Cushitic language of Ethiopia. Köln: Köppe. Treis, Yvonne. 2008a. A grammar of Kambaata. Part 1: Phonology, nominal morphology, and nonverbal predication. Köln: Köppe. Treis, Yvonne. 2008b. Relativization in Kambaata (Cushitic). In Zygmunt Frajzyngier & Erin Shay (eds.), Interaction of morphology and syntax. Case studies in Afroasiatic, 161–206. Amsterdam: Benjamins. Treis, Yvonne. 2010. Purpose-encoding strategies in Kambaata. Afrika und Übersee 91. 1–38. Treis, Yvonne. 2011. Purpose clauses in Ethiopian Languages. Paper presented at the 14th Italian meeting of Afroasiatic linguistics, Torino, 15–18 June. Treis, Yvonne. 2012. Expressing future time reference in Kambaata. Nordic journal of African studies 20(2). 132–149. Treis, Yvonne. 2015. The additive -V in Kambaata. Paper presented at the 48th annual meeting of the SLE, workshop “Additives across languages”, Leiden, 2–5 September. Treis, Yvonne. 2017. Similative morphemes as purpose clause marker in Ethiopia and beyond. In Yvonne Treis & Martine Vanhove (eds.), Similative and equative constructions. A typological perspective, 91–142. Amsterdam: John Benjamins. Vanhove, Martine (avec la collaboration de Mohamed-Tahir Hamid Ahmed et Ahmed Abdallah Mohamed-Tahir). 2017. Le bedja. Leuven: Peeters. Zaborski, Andrzej. 1975. The verb in Cushitic. Krakow: Nakladem Uniwersytetu Jagiellonskiego. Zaborski. Andrzej. 1997. The position of Cushitic and Berber within Hamito-Semitic dialects. In Alessandro Bausi & Mauro Tosco (eds.), Afroasiatica Neapolitana. Contributi presentati all’ 8° Incontro di Linguistica Afroasiatica (Camito-Semitica). Napoli, 25–26 Gennaio 1996, 49–59. Napoli: Istituto Universitario Orientale.

Denis Creissels

16 Grammaticalization in Manding languages 1 Introduction 1.1 The aim of the present chapter In this chapter, I review grammaticalization processes that can be reconstructed in the history of Manding languages on the basis of internal reconstruction and/or comparative data (both within Manding languages and between Manding languages and languages belonging to other branches of the Mande language family), and also phenomena found in present-day Manding languages that can be analyzed as grammaticalization in progress.

1.2 The Manding languages: inventory, location, and genetic affiliation The Manding languages are a group of closely related languages (sometimes considered dialects of a single macro-language) included in the Central sub-branch of the Western branch of the Mande language family.1 Manding is the second largest dialect cluster / macro-language of West Africa after Hausa, with a total number of speakers exceeding 20 million. It is a typical dialectal continuum where sharp linguistic boundaries are rare, which makes it impossible to decide on the basis of purely linguistic criteria how many Manding languages should be distinguished. The most important and best-known Manding varieties are Bambara (aka Bamanan, the most widely spoken language in Mali), Maninka-mori (a major language of Guinea),2 Mandinka (the main language of Gambia, also spoken in Senegal and Guinea Bissau), and Jula (or Dyula, a lingua franca of Northern Ivory Coast and Western Burkina Faso, linguistically close to Bambara). The other Manding varieties mentioned in the present article are Baninko Bambara (Southern Mali), Kita Maninka

1 The Mande language family was included by Greenberg in the Niger-Congo phylum, but the evidence for a Niger-Congo affiliation of Mande is rather slim, and, for example, Dimmendaal (2011) argues that Mande is best treated as an independent language family. It is commonly admitted that the time distance between the most ancient branches of the Mande language family exceeds 5 millenia, whereas the time depth of the Manding dialect cluster does not exceed 8 centuries. On the classification of Mande languages, see Vydrin (2009). 2 Maninka or Malinke is a cover term for several Manding varieties that are not particularly close together. For example, linguistically, Niokolo Maninka is much closer to Gambian Mandinka than to Maninka-mori or to Kita Maninka. https://doi.org/10.1515/9783110712735-002

696

Denis Creissels

(Western Mali), Korokan (Central Ivory Coast), Koyaga (Central Ivory Coast), Marka (Burkina-Faso), Maukakan (Western Ivory Coast), Niokolo Maninka (Eastern Senegal), and Xasonga (Western Mali).3

1.3 The Manding languages: typological characteristics 1.3.1 Phonology As a rule, Manding languages have typologically unremarkable phoneme inventories including 5 to 7 vowels, plus nasality and/or length contrasts (depending on the individual varieties), and about 20 consonants. Manding languages are tonal languages with complex systems of combinatory rules responsible for frequent discrepancies between underlying tones and surface tones (see for example Creissels [2009: 19–39] on the tonal system of Kita Maninka). In this connection, the reader is invited to keep in mind that, whenever the same word with exactly the same gloss occurs across examples with different tones on its last syllable (as Mandinka ‘money.D’: kódóo in Ex. [2], kódòo in Ex. [3d]), this variation results from automatic tone sandhi rules and has no morphological significance.

1.3.2 Word classes and constructional morphology Manding languages have rich systems of derivational morphology (mainly suffixal). Compounding is also very productive. By contrast, inflectional morphology is very limited, and most grammaticalized semantic distinctions are expressed by grammatical words or clitics rather than affixation or other morphological operations. In Manding languages, nouns cannot be freely converted into verbs; by contrast, as a rule, verbal lexemes can also be used as event nouns (or in the case of transitive verbs, as the second formative of compound nouns whose first formative is a nominal lexeme representing their object argument) without any specific morphological marking.

3 In the present paper, Manding forms quoted without further specification are Standard Bambara forms as recorded in Dumestre’s reference dictionary (Dumestre 2011). The dialectal origin of other forms is systematically indicated by the following abbreviations: BBm = Baninko Bambara, Ju = Jula, KMnk = Kita Maninka, Kor = Korokan, Koy = Koyaga, Mdk = Mandinka, Mnkm = Maninkamori, Mau = Maukakan, Xs = Xasonga. When the context requires it, Standard Bambara forms are indicated as Bm.

Grammaticalization in Manding languages

697

1.3.3 Nouns and noun phrases Manding languages have no grammaticalized system of nominal classification. The structure of Manding noun phrases can be schematized as follows, with two possible positions for determiners:4 (Gen) (Det₁) N (Attr) (Num) (Det₂) All Manding varieties have a clitic determiner (glossed d) that can be characterized semantically as a default determiner, since in most contexts it carries no particular semantic specification, and must simply be present if the speaker does not consider useful to select a determiner with a more specific meaning. The default determiner occupies the position Det₂ in the template above, and its combination with nouns tends to behave as the default form of nouns, whereas its absence must be licensed by otherwise overtly expressed grammatical features of the noun phrase or of the clause in which it is included. There are, however, contexts in which the default determiner still contrasts with its absence, in particular negative contexts. (1)

Mandinka a. Mùsôo máŋ nǎa. (mùsôo < mùsú + ò) woman.d cpl.neg come ‘The woman did not come.’ b. Mùsú máŋ nǎa. woman cpl.neg come ‘No woman came.’

As indicated in (1a), in Mandinka, the default determiner is underlyingly -ò, but in many Manding varieties it is reduced to a floating tone added to the inherent tonal melody of its host (compare Mdk básà ‘lizard’ + d → básòo with KMnk básá ‘lizard’ + d → básà). Manding languages have no agreement mechanism between head nouns and their dependents, and more generally, head-dependent relationships within NPs are not morphologically marked, with the only exception of indirect (or ‘alienable’) possession, in which the genitival dependent is marked by a postposition. Relative clauses are not constructed as modifiers in a noun phrase headed by the noun they modify. The commonest relativization strategy in Manding languages is a correlative strategy in which the relative clause precedes the matrix clause; the semantic head of the relative clause occupies the relativized position; it is marked

4 Gen = genitival modifier, Det = determiner, n = head noun, Attr = attributive adjective, Num = numeral; on relative clauses, see 8.1.

698

Denis Creissels

by a relativizer, and resumed in the matrix clause by a demonstrative or personal pronoun – Ex. (2). (2)

Mandinka [Mùsôo yè kèwôoi mîŋ ná kódóo tǎa], ŋ́ níŋ wǒoi woman.d cpl.tr man.d rel gen money.d take 1sg with dem běn-tà. meet-cpl.intr ‘I met the man whose money was taken by the woman.’ lit. something like ‘The woman took which man’s money, I met that one.’

1.3.4 Verbs and clauses Verbal clauses are characterized by a particularly rigid constituent order. Their construction can be schematized as S pm (O) V X, with a so-called ‘predicative marker’ (a kind of auxiliary) immediately after the subject NP. Most grammaticalized TAM and polarity distinctions are expressed by the predicative markers. Verbal inflection in independent clauses is very limited. As a rule, the only verbal affix that expresses a TAM / polarity value independently, in a construction including no overt predicative marker, is a verbal suffix encoding ‘completive, positive’ in intransitive predication – Ex. (3a). Note that this verbal suffix is in complementary distribution with a predicative marker expressing the same value in transitive predication – Ex. (3c). As illustrated in Ex. (3b) and (3d), as a rule, the grammaticalized TAM and polarity values other than ‘completive, positive’ are expressed in the same way in transitive and intransitive predication. (3)

Mandinka a. Yíróo bòyí-tà sílòo kâŋ. tree.d fall-cpl.intr road.d on ‘The tree fell down on the road.’ b. Nèwóo kà kómôŋ jíyòo kónò. iron.d incpl rust water.d in ‘Iron rusts in water.’ c. Wùlôo yè díndíŋó tàŋkàndí dìmbâa má. dog.d cpl.tr child.d save fire.d postp ‘The dog saved the child from the fire.’ d. Kèwóo kà à téerímáa máakóyì kódòo tó. man.d incpl 3sg friend help money.d loc ‘The man helps his friend financially.’

Grammaticalization in Manding languages

699

As can be seen from Ex. (3), subjects and objects are neither flagged nor indexed, and obliques are commonly encoded as postpositional phrases, although some prepositions can also be found. Prepositions seem to be mostly recent innovations, with the exception of the comitative preposition ní, which has cognates in other branches of the Mande language family. Personal pronouns are found in the same positions as ordinary NPs, and have the same form in all their possible functions. With the only exception of the implicit 2nd person singular subject of imperative clauses, in verbal predication, the subject and object slots cannot be left empty. As argued in Creissels (2015a), constructions in which one of the core arguments of a potentially transitive verb is left unexpressed must not be analyzed as transitive constructions with a null subject or object, but as intransitive constructions with the remaining argument in subject function, and the possibility of such intransitive constructions of potentially transitive verbs depends on the valency properties of individual verbs. In this respect, it is noteworthy that Manding languages have very limited classes of A-labile verbs (i.e., verbs used transitively or intransitively with the same semantic role assigned to their subject), whereas P-lability is pervasive. Moreover, as illustrated by Ex. (4), Manding languages do not have only P-lability of the cross-linguistically common causative / anticausative type, but also active / passive lability, and are even, according to Cobbinah and Lüpke (2009), a particularly clear case of languages with morphologically unmarked passive constructions. (4)

Bambara a. Sékù má nὲgὲsô dílán. Sékou cpl.neg bicycle repair S pm O V ‘Sékou did not repair the bicycle.’ b. Nὲgὲsô má dílán Sékù fὲ. bicycle cpl.neg repair Sékou by S pm V X ‘The bicycle was not repaired by Sékou.’

Causative is the only valency operation encoded by derivational affixes. Morphologically unmarked predication is quite marginal in Manding languages, and copulas are systematically used in non-verbal predication.

2 Grammaticalization of nominal categories 2.1 Class/gender As already mentioned, Manding languages have no grammaticalized system of nominal classification of any type, and there is no evidence that they might have

700

Denis Creissels

had such systems in the past. They don’t show phenomena that might be interpreted as emergent nominal classification either.

2.2 Number 2.2.1 The origin of ordinary plural markers Western and Southwestern Manding languages have a plural marker -lú ~ -lí, probably cognate among others with Soninke -nú ~ -ní and Bobo -re ~ -ne ~ -ri ~ -ni. In the Manding varieties that have this suffix, it is used with all nouns without exception, whereas in Soninke, -nú ~ -ní is used with some common nouns only, but at the same time combines with proper names as an associative plural marker. In Bobo, -re ~ -ne ~ -ri ~ -ni is mainly found with kinship terms and human nouns. Since associative plural markers ni are found in Southwestern Mande languages and in Mano (South Mande), a Proto-Mande associative plural marker *ni can be reconstructed, and it seems reasonable to assume that Manding -lú ~ -lí is a reflex of this associative plural marker which was reanalyzed as an ordinary plural marker and subsequently generalized to all nouns. Moreover, comparative data suggest that the Proto-Mande associative plural marker *ni might have resulted from the grammaticalization of a comitative preposition *ni also used for NP additive coordination, whose reflex in Manding is ní. The diachronic scenario (supported by data from languages as diverse as Sara languages [Nilo-Saharan] and Basque) could be the reanalysis of an additive coordination marker as an associative plural marker in a coordinative construction in which the elision of the second conjunct was possible (N and Ø → ‘N and other persons’) – Creissels (2015b). Bambara and some other Manding varieties have a plural suffix -ú whose relationship with -lú ~ -lí is problematic, since there is no regular l ~ Ø correspondence between Bambara and the varieties in which the plural marker is -lú ~ -lí. I am aware of no plausible etymology for this plural suffix, which however may be cognate with the plural suffix -u found in Soninke. The resemblance with the Bambara 3rd person plural pronoun ǔ has sometimes been emphasized, but the rising tone of the 3rd person plural pronoun rather suggests that it results from the contraction of a dissyllabic form with a LH tonal melody, which casts some doubt on the hypothesis that the plural suffix -ú might straightforwardly originate from the 3rd person plural pronoun ǔ.5

5 Outside Manding, a full coincidence between the plural marker of nouns and the 3rd person plural pronoun is found in Soso, but with the entirely different form è.

Grammaticalization in Manding languages

701

2.2.2 The grammaticalization of associative plural markers In most Manding languages, the associative plural is expressed analytically as lit. ‘those of N’. Mandinka has an associative plural marker -ñòlú (as in Mdk Fàatú-ñòlú ‘Fatou and other persons with her’). The univerbation of ‘those of N’ is a possible etymology, since ñù-lú is attested as a variant of the plural of the Mdk demonstrative ñǐŋ. However, the univerbation of ‘the likes of N’ is another possible etymology, since the Mandinka equivalent of ‘like (N)’ is ñóo ~ ñóŋ.

2.3 Possession Some Manding varieties mark inalienable possession by means of a postposition (Mdk lá, KMnk yé) also used to mark oblique terms of predicative constructions. According to Grégoire (1984), the construction of alienable genitives as attested in present-day Manding languages results from the grammaticalization of a construction that initially described the relationship between the possessor and the possessee as a spatial relationship with the possessor in the role of ground and the possessee in the role of figure. The scenario put forward by Grégoire explains the coincidence between postpositions and genitival linkers marking alienable possession, since there is ample evidence that even the postpositions that synchronically have no locative use started as spatial postpositions (see Section 2.5.1). However, in some Manding languages, the genitival linker does not coincide with a postposition marking oblique terms in predicative constructions, but with the proprietive pronoun tá. The meaning of this pronoun it that the dependent in a genitival construction must be identified with a discursively salient notion, as in (5b) and (6b), and its grammaticalization as a genitival linker is illustrated in (6a), to be compared with (5a). The origin of (6a) is an appostive construction with the literal meaning ‘that of the woman, money’. (5)

Mandinka a. mùsôo lá kódòo woman.d gen money.d ‘the money of the woman’ b. mùsôo tâa woman.d propr ‘that of the woman’

(6)

Jula a. mùsô tá wárî woman.d gen money.d ‘the money of the woman’

702

Denis Creissels

b. mùsô tá woman.d propr ‘that of the woman’

2.4 Determiners 2.4.1 The default determiner Given the coincidence between the full form ò of the default determiner and the demonstrative ǒ (see Ex. [1] in Section 1.3.3), the default determiner can be analyzed, within the frame of the definiteness cycle put forward by Greenberg (1978), as a reflex of this demonstrative which probably functioned as a definite article at some stage in the history of Manding languages, but is now at an advanced stage of a semantic bleaching process. Some varieties also have a determiner resulting from the evolution of the demonstrative nǐn. For example, Mdk ŋˋ is an optional variant of the default determiner for nouns combined with the demonstrative ñǐŋ (the definite form of mùsú ‘woman’ is mùsôo, but ñǐŋ mùsôo ‘this woman’ is in free variation with ñǐŋ mùsû ŋ).

2.4.2 Third person singular possessive > definiteness marker In Manding languages, nouns combined with the third person singular pronoun à in genitive function can be found sporadically with the meaning ‘the aforementioned N’, ‘the N in question’ instead of their usual meaning ‘the N in the personal sphere of x (where x must be identified to a discursively salient entity)’. Synchronically, this particular use of the third person singular pronoun can be analyzed as an instance of ‘vague reference’ comparable to the use of third person singular pronouns in meteorological expressions like ‘it rains’: the third person pronoun in genitive function can be interpreted as referring to a situation taken as a whole rather than to a particular entity, and consequently ‘its N’ can be interpreted as ‘the N of the situation in question’ > ‘the N in question’. Diachronically, this can be viewed as the first step in a grammaticalization process whose outcome could be the creation of a definiteness marker.

2.4.3 The grammaticalization of *do ‘one’ The Proto-Mande reconstruction for the numeral ‘one’ is *do. In Manding languages, the numeral kélén ‘one’ is clearly not a reflex of *do, and probably comes from an adjective meaning something like ‘single’, ‘isolated’, but the indefinite determiner dɔ́ ‘some’ is the obvious reflex of Proto-Mande *do ‘one’.

Grammaticalization in Manding languages

703

2.4.4 The grammaticalization of fέn ‘thing’ Fέn ô fέn ‘everything’ (also used quite regularly with a free choice meaning – ‘anything’, and in negative contexts with a negative meaning – ‘nothing’) is the distributive form of the noun fέn ‘thing’. In some Manding varieties, fέn ô fέn has grammaticalized as a distributive / free choice / negative determiner that can indiscriminately combine with animate as well as inanimate nouns, as in Mandinka í díŋò féŋ wóo fèŋ ‘any of your children’, lit. ‘anything of your children’. The same grammaticalization path is attested with Soninke hó wó hò ‘everything’. In Soninke, hóonú ‘things’ has similarly grammaticalized as a pronoun meaning ‘some of them’, without any distinction between animate and inanimate referents, but as far as I know, this evolution is not attested in Manding languages.

2.4.5 Sí ‘kind’ > negative determiner In Manding languages, it is always possible to express negative determination by using bare nouns in negative contexts – see Ex. 1 in Section 1.3.3. However, in addition to a distributive / free choice / negative determiner originating from the distributive form of fέn ‘thing’ (see Section 2.4.3), some Manding varieties also have a negative determiner originating from the noun sí ‘seed’, ‘kind’. For example, the etymological meaning of Bm Tùbàbù sí má nà ‘No European came.’ is ‘Kind of European did not come’ (where má is a completive negative marker). As a rule, the use of these overt negative determiners is perceived as more emphatic than the mere use of bare nouns in negative contexts.

2.5 Case/adpositions Manding languages do not have morphological case, but have rich inventories of adpositions (mainly postpositions).

2.5.1 The grammaticalization of postpositions Some postpositions have quite obviously grammaticalized from a nominal lexeme still attested with the same form: – Mdk bálà ‘in contact with’ < bálà ‘body’; – Mnkm bárá ‘at N’s place’ (French ‘chez’) < bárá ‘home’; – cέ ‘between’ < cέ ‘waist’, ‘middle’; – bóló ‘under the responsibility of ’, ‘in the personal sphere of ’, also used to mark agent phrases in passive constructions < bóló ‘hand, arm’;

704 – – –

– – –

Denis Creissels

kɔ́ ‘behind’, ‘after’ < kɔ́ ‘back’; kɔ́nɔ́ ‘in’ < kɔ́nɔ́ ‘belly’, ‘pregnancy’; kɔ́rɔ́ ‘under’ is probably cognate with kɔ́rɔ́ ‘meaning’: a reasonable hypothesis is that the original meaning of this noun (maintained by the postposition) was something like ‘bottom’ or ‘underneath’, and its present-day meaning results from a metaphorical shift; kǔn ‘on’ < kǔn ‘head’; ɲέ ‘in front of ’, ‘before’ < ɲέ ‘eye’, ‘face’; Mdk yǎa ‘at N’s place’ (French ‘chez’) < yǎa ‘home’.

The origin of the other postpositions is less obvious. For some of them, an etymology can however be put forward on the basis of comparative data: – the multifunction postposition lá, particularly productive in instrumental function, is also productive in locative function in the Manding varieties that do not have the dedicated locative postposition rɔ́ (Mdk tó, Mnkm dɔ́). As argued by Grégoire (1984), the comparison with cognate forms in other Mande languages supports the hypothesis that it has its origin in the grammaticalization of the noun ‘mouth’ (found in Manding as dá) into a spatial postposition which subsequently acquired non-spatial uses, and in some Manding varieties has been replaced by rɔ́ in concrete spatial function;6 – Western and southwestern Manding varieties have a postposition (Mdk tí, Mnkm dí) productive in similative and functive-transformative uses only; however, comparison with Soninke ˋdí ‘in’ and Bobo tí ‘place’ suggests that this postposition originated from a locational noun that first grammaticalized into a spatial postposition which subsequently acquired non-spatial uses and lost its original spatial uses; – synchronically, the meaning of the postposition mǎ is difficult to define, since this postposition is mainly used to mark oblique arguments whose semantic role is implied by the lexical meaning of the verb, but it is probably cognate with a noun mǎa ‘body’ found in Sangala Jalonka, and other internal and comparative evidence confirms that, as commonly assumed, it started as a spatial postposition expressing ‘in contact with’. The postposition kǎn ‘on’ may be cognate with the Soninke postposition kànmá ‘on’, and consequently with the Soninke noun kànmê ‘sky’, since in Soninke, the root of kànmê ‘sky’ is kàn-. The problem is however that ‘sky’ is found in Manding languages as sán, which may also be cognate with Soninke kànmê ‘sky’, since there is a regular sound correspondence Manding s ~ other West Mande languages k. It is

6 Rɔ́ is one of the postpositions for which I am aware of no evidence of a lexical origin.

Grammaticalization in Manding languages

705

difficult to explain how the same Soninke root might have a Manding cognate with an initial s, and another with an initial k.7 As regards the benefactive postposition yé, the only possible cognate I have been able to find is the Bobo proprietive pronoun yì.

2.5.2 The grammaticalization of prepositions Among prepositions, kó ‘like’ results from the grammaticalization of the quotative kó ‘say’ – see 4.1. Yànní and sání ‘before’ can be decomposed as yǎn ‘here’ + ní ‘with’ and sá ‘now’ + ní ‘with’ respectively.8 The fact that the preposition jànkó ‘even more so’ is found as sákó or sókó in some Manding varieties suggests a similar decomposition, with the quotative kó as the second element. The infinitives kà bɔ́ < bɔ́ ‘come from’ and kà táa < táa ‘go’ have uses in which they arguably function as prepositions meaning ‘starting from’ and ‘until’, respectively. Kàbí(ní) (Mdk (kà)bíríŋ) is a preposition expressing ‘since’. Its form suggests that it originates from an infinitive form, but I am aware of no verb, either in Manding or in other Mande languages, that could provide a plausible etymology. It is however intriguing to observe that Xasonga has a predicative marker bíríŋ expressing recent past, which might well be cognate with this preposition.

2.6 Comitative and NP additive coordination (‘and’) Like many other Subsaharan languages, Manding languages express NP additive coordination by means of the same morpheme (the preposition ní) as comitative adjunction. It is commonly assumed that the coordinative use of such morphemes develops from their comitative use, but in the case of Manding languages, I am aware of no evidence that the evolution occurred in this direction rather than in the opposite one. See Section 2.2.1 for another grammaticalization process in which the ancestor of this comitative preposition might have been involved.

7 Dumestre (2003: 262) proposes kán ‘neck’ as the lexical origin of kǎn ‘on’. Comparative data make this hypothesis doubtful, since in Soninke, kànmá ‘on’ is quite obviously cognate with kànmê ‘sky’ and not with qánnè ‘neck’. 8 Soso béenùŋ ‘before’ and Soninke sádò ‘before’ lend themselves to similar decompositions.

706

Denis Creissels

2.7 Nominal derivation Several derivational suffixes used to form nouns have a plausible lexical origin: – -bá augmentative is probably cognate with bá ‘mother’; – -bálí privative is probably cognate with bàlí ‘hinder’, ‘fail’; – -lá ‘place occupied by …’ (as in Fàràjὲlá ‘Europe’ < Fàràjέ ‘white-skinned person’) comes probably from the same noun ‘mouth’ as the postposition lá (cf. 2.5.2); – -lén resultative may be cognate with dén ‘child’; – -nín diminutive is probably cognate with dén ‘child’.9

2.8 Personal pronouns 2.8.1 Reflexive pronoun > 2nd person pronoun In Manding languages, the 2nd person singular pronoun í is homonymous with a residual reflexive pronoun used for object reflexivization with a limited set of verbs. Since for example Soninke has a productive reflexive pronoun í distinct from the 2nd person singular pronoun (Soninke án), one may imagine that perhaps the Manding 2nd person singular pronoun resulted from the reanalysis of a reflexive pronoun (a relatively common grammaticalization scenario). Unfortunately, I am aware of no additional evidence supporting this hypothesis, and it must be added that, if 2nd person í developed from reflexive í, this evolution must have occurred very early in the history of Mande languages, since potential reflexes of both 2nd person *í and reflexive *í can be found in various branches of the Mande language family (Valentin Vydrin, p.c.)

2.8.2 Intensive pronouns In Manding languages, intensive pronouns are formed by combining a personal pronoun in genitive function with an element comparable to English self: Mdk ŋ́ fáŋò ‘myself ’, í fáŋò ‘yourself ’, etc. I am aware of no possible etymology for Mdk fáŋò ‘self ’, but Bm yὲrέ ‘self ’ is problably cognate with Soninke yèrú ‘same’.

9 In most Manding varieties, the resemblance between ‘child’ and the diminutive suffix is much greater than in Bambara; cf. for example Mdk díŋ ‘child and -ríŋ ~ -díŋ ~ -ndíŋ diminutive suffix.

Grammaticalization in Manding languages

707

2.8.3 ‘Owner’ > emphatic third person pronoun Several Manding varieties have an emphatic third person pronoun whose transparent etymology is ‘its owner’. For example, in Bambara, depending on the context, à tìgî can be interpreted with its literal meaning ‘its owner’, or as ‘the person in question’.

3 Grammaticalization of verbal categories 3.1 Valency 3.1.1 Passive Manding languages do not have passive morphology, but have very productive morphologically unmarked passive constructions – see Section 1.3.4. Interestingly, as illustrated in (7), causative-anticausative lability is also common in Manding languages (although never as general as active-passive lability). (7)

Mandinka a. Kèwôo yè mùróo jòlóŋ bàŋkôo tó. man.d cpl.tr knife.d drop ground.d loc ‘The man dropped the knife on the ground.’ b. Máŋkóo jòlôn-tá bàŋkôo tó. mango.d fall-cpl.intr ground.d loc ‘The mango fell on the ground.’

Since cross-linguistically, unmarked anticausative constructions are much more widespread than unmarked passive constructions, and the reanalysis of anticausatives as passives is a widely attested process, one may assume that the fully productive active-passive lability that characterizes present-day Manding languages developed historically from the reanalysis of causative-anticausative lability.

3.1.2 Causative The westernmost Manding varieties have a causative suffix (Mdk -ndí) that has cognates in Soninke (-ndí) and Bozo (-ni). In (Creissels, forthcoming), I argue that the etymology of this suffix is a verb reconstructable at Proto-West-Mande level as *ti(n) ‘do, make’.

708

Denis Creissels

In the other Manding varieties, as illustrated in (8), causative verbs are formed by means of a preverb lá-. (8)

Bambara a. Dénmísɛ̂n bέ sírán wùlû ñέ. child.d incpl be_afraid dog.d before ‘The child is afraid of the dog.’ b. Wùlû bέ dénmísɛ̂n lá-sírán. dog.d incpl child.d caus-be_afraid ‘The dog frightens the child.’

This causative prefix is probably cognate with the postposition lá (see section 2.5.1), since in Soso-Jalonka (a language belonging to another branch of West Mande), a similar coincidence is observed between a causative preverb rá- and a postposition rá. Unfortunately, given the extreme rigidity of word / constituent order in Mande languages, it is difficult to imagine how a postposition and a causative preverb may have developed from a common ancestor. It is, however, interesting to observe that (a) lá-causatives expressing indirect causation sometimes contrast with morphologically unmarked causative constructions expressing direct causation, and (b) in some of the Manding varieties that have the causative preverb lá-, its productivity is very low, and many verbs that are strictly intransitive in other varieties can be found in morphologically unmarked causative constructions. This suggests that the causative prefix lá- started as an optional marker of indirect causation in morphologically unmarked causative constructions. However, this does not completely solve the problem.

3.1.3 Permissive causation In Manding languages, bìlá ‘leave’ is regularly used to express permissive causation.

3.1.4 Reflexive Manding languages have a reflexive pronoun whose productivity is rather limited. Reflexivity is more commonly expressed by means of intensive pronouns (see Section 2.8.2).

3.1.5 Reciprocal The reciprocal pronoun ɲɔ́gɔ́n is cognate with the noun ɲɔ ̀gɔ́n ‘the like of ’, and also with a suffix that derives nouns of co-participants from verbs (as in sìgì-ɲɔ́gɔ́n ‘neighbor’ < sìgí ‘sit, settle’).

Grammaticalization in Manding languages

709

3.1.6 Antipassive Mandinka has an antipassive marker -rí (with the allomorph -dírí in combination with stems ending with a nasal), but this suffix does not straightforwardly convert transitive verbs into intransitive ones, and its precise status in the Mandinka system of word formation is not easy to define. Its identification as an antipassive marker follows from the fact that it is found exclusively in combination with transitive verbal lexemes in constructions in which the P argument is left unexpressed, cannot be identified to the referent of a noun phrase included in the same construction, and is interpreted as non-specific. However, Mdk dómò ‘eat’ is the only verb whose antipassive form dómórì can be used as the verbal predicate of finite clauses. With other transitive verbs, the antipassive form can only be used as an antipassive event noun, as in Ex. (9), or as a stem to which typically verbal suffixes can be attached: the agent nominalization suffix -láa ~ -náa, the instrument nominalization suffix -ráŋ ~ -láŋ ~ -dáŋ, etc. (9)

Mandinka a. Mùsôo bé màanì-túwòo lá. woman.d loccop rice-pound.d postp lit. ‘The woman is at the rice-pound(ing).’ → ‘The woman is pounding rice.’ (màaní ‘rice’ saturates the P valency of tǔu ‘pound’, and the subject of the copula is identified to the unexpressed A argument) b. Màanôo bé tùwôo lá. rice.d loccop pound.d postp lit. ‘The rice is at the pound(ing).’ → ‘The rice is being pounded.’ (if none of the arguments of tǔu ‘pound’ is expressed, in the absence of the antipassive suffix, the subject of the copula is identified to the unexpressed P argument) c. Mùsôo bè tùu-rôo lá. woman.d loccop pound-antip.d postp lit. ‘The woman is at the pound(ing).antip.’ → ‘The woman is pounding.’ (the antipassive suffix saturates the P valency of tǔu ‘pound’, and the subject of the copula is identified to the unexpressed A argument)

In other Manding varieties, the suffix cognate with this atypical antipassive marker (Bm -lí, etc.) cannot be analyzed as an antipassive marker, and can only be analyzed as an overt event nominalization marker in competition with morphologically un-

710

Denis Creissels

marked event nominalization, since it can be found with intransitive verbs and is compatible with the expression of the object argument of transitive verbs. In (Creissels, forthcoming), I argue that this suffix is cognate with the Soninke antipassive marker -ndì, and originates from the same verb *ti(n) ‘do’ as the causative suffix -ndí also found in Mandinka and Soninke, which implies the following grammaticalization path: ‘do’ verb following another verb used nominally in object role > antipassive marker > antipassive nominalization marker > nominalization marker Manding languages also have antipassive uses of the reflexive pronoun, but the grammaticalization of the reflexive pronoun as an antipassive marker is limited to a very small number of verbs, typically including mǐn ‘drink’.

3.2 Aspect 3.2.1 Incompletive, habitual, progressive The use of the locational copula as a progressive or incompletive aspect marker is pervasive in Manding, and Manding languages provide ample evidence that progressive periphrases involving the locational copula tend to evolve toward a less specific incompletive meaning. In many Manding varieties, a former progressive periphrasis that has undergone this evolution and synchronically can express meanings such as habitual or future coexists with a (formally distinct) younger periphrasis specifically expressing a progressive meaning. Mandinka also has an incompletive aspect marker kà mainly used with reference to habitual situations. The plausible origin of this marker is a habitual auxiliary kàrí ~ kàrá also found in Mandinka, which to the best of my knowledge has no cognates in other Manding varieties.

3.2.2 Completive As a rule, for the completive positive, Manding languages have two semantically equivalent markers in complementary distribution: a verbal suffix (Mdk -tá, Mnkm -dá, Bm -rá) in intransitive predication, and a predicative marker in post-subjectal position (Mdk yè, Bm yé, KMnk dí, etc.) in transitive predication. I am aware of no evidence suggesting a plausible etymology for the completive positive suffix. As regards the completive positive predicative marker, it is interesting to observe that it has very different forms across Manding varieties, which, however, have in common to be homonymous with postpositions (or at least to have the same segmental

Grammaticalization in Manding languages

711

form as postpositions). Ex. (10) illustrates the quasi-homonymy between the completive positive marker and the benefactive postposition in Mandinka, and Ex. (11) illustrates the same phenomenon with the completive positive marker and the similative-functive-transformative postposition in Kita Maninka. (10) Mandinka Kèwôo yè bàtáayòo sáfée à díŋò yé. man.d cpl.tr letter.d write 3sg son.d ben ‘The man wrote a letter to his son.’ (11) Kita Maninka Jínè dí mìnán yèlèmá tǔn dì. jinn.d cpl.tr antelope.d change anthill.d funct ‘The jinn changed the antelope into an anthill.’ As developed in Creissels (1997a), a possible explanation is that, in Manding languages, the completive positive markers result from the reanalysis of postpositions marking the agent phrase in passive constructions with a topicalized agent phrase. See however Idiatov (2020) for a criticism of this hypothesis, and an alternative proposal according to which they might result from auxiliarization processes.

3.2.3 Perfect A distinction between perfect and narrative past can be found in several Manding varieties. In Kita Maninka, the perfect has a clear resultative origin, since the intransitive perfect is simply marked by a suffix -nín whose cognates can be found in all Manding varieties as resultative derivational suffixes, whereas the transitive perfect is marked by a predicative marker nànín resulting from the grammaticalization of nànín, resultative form of nǎ ‘come’. A predicative marker bárá ~ bádá expressing perfect is found in Maninka-mori, and more or less obvious cognates of this perfect marker can be found not only in other Southern Manding varieties (Kor báa, Koy wá, Mau wέɛ, etc.), but also in Koranko (wára), Soso (bárà) and Jalonka (bántà). Since -rá ~ -dá ~ -tá is a verbal suffix marking the completive aspect in Manding languages, a plausible hypothesis is that this perfect marker originates from the completive form of bán ‘finish’.

3.2.3 ‘Already’ Across Manding languages, bán ‘finish’ is more commonly grammaticalized, either in its bare form bán or in the infinitive form kà bán, as a particle occurring in postverbal position with the meaning ‘already’.

712

Denis Creissels

3.2.4 ‘Fall’ or ‘move towards’ > inchoative As illustrated by Ex. (12) and (13), the verbs bǐn ‘fall’ and sín ‘move toward’ are used as inchoative auxiliaries. (12) Mandinka a. Yíróo bòyí-tà sílòo kâŋ. tree.d fall-cpl.intr road.d on ‘The tree fell on the road.’ b. Mùsóo bòyí-tá kùmbôo-lá. woman.d fall-cpl.intr cry-inf ‘The woman burst into tears.’ (13) Bambara a. Mùsô sín-ná só mà. woman.d move_toward-cpl.intr house.d postp ‘The woman moved toward the house.’ b. Mùsô sín-ná kà kàsí. woman.d move_toward-cpl.intr inf cry ‘The woman burst into tears.’

3.2.5 ‘Want’ or ‘search’ > ‘almost do’, ‘come close to’ The verb ɲíní ‘search’ is used as an auxiliary with the meaning ‘come close to’, as in Ù bέ ɲíní kà sà lit. ‘They are looking for dying > ‘They are about to die’. Mdk làfí ‘want’ is used as an auxiliary with the same meaning, as in À làfí-tá fǎa-là lit. ‘He wanted to die’ > ‘He almost died’.

3.2.6 ‘Return’ > repetitive The verb sègín ‘return’ is used as a repetitive auxiliary, as in À sègìnná kà kùlò lit. ‘He returned to scream’ > ‘He screamed again’.

3.2.7 ‘Remain’ > continuative or habitual The verb tó ‘remain’ is used as a continuative or habitual auxiliary.

Grammaticalization in Manding languages

713

3.2.8 ‘Separate, escape’ > ‘eventually’ The verb tílá ‘separate, escape’ is used as an auxiliary with the meaning ‘eventually’, as in Ù tílálá kà sábálí lit. ‘They escaped to calm down’ > ‘They eventually calmed down’.

3.2.9 ‘Say’ > ‘be about to’ In Manding languages, the quotative kó ‘say’ (cf. Section 4.1) is commonly used to express imminence (as in Mdk Yíròo kó à bé bòyílà ‘The tree is about/close to fall’, lit. ‘The tree says it is falling’).

3.3 Modal categories 3.3.1 Hortative/subjunctive The hortative/subjunctive markers found in Manding languages can be divided into four etymologically distinct subsets represented by Bm ká, Mdk yè, Kor yá, and BBm lá. Three of these hortative/subjunctive markers coincide (at least in their segmental form) with completive markers attested either in the same varieties, or in other varieties, and all of them coincide (at least in their segmental form) with postpositions. Ex. (14) shows that, in Mandinka, the benefactive postposition yé is almost homonymous not only with the completive positive marker, but also with the subjunctive positive marker, used in independent clauses with a hortative meaning. (14) Mandinka Díndíŋò yè táa kàràmbúŋò tó! child.d subj go school.d loc ‘Let the child go to school!’ As proposed in (Creissels 1997b), a possible explanation of these coincidences is that not only completive markers, but also hortative markers may have resulted from the reanalysis of postpositions. In the case of the hortative/subjunctive markers, constructions of the type illustrated by English ‘(it is) up to N to V’ or French ‘à N de Vinf ’ constitute a plausible context for such a reanalysis. In the particular case of Bm ká, the resemblance with Infinitive kà may suggest other possible connections, discussed in (Vydrin 2014).

3.3.2 ‘Don’t come’ > prohibitive marker Most Manding varieties have a dissyllabic prohibitive marker (Mdk kánàa, Bm kànâ), but Southern varieties have monosyllabic forms such as Koy ká in free varia-

714

Denis Creissels

tion with the dissyllabic form, which suggests that kànâ and similar forms were originally the prohibitive form of nǎ ‘come’.

3.3.3 Possibility The verb sé ‘reach’ is also used as a modal auxiliary expressing possibility. Mandinka has a predicative marker sí ~ sé labeled ‘potential’ by Creissels and Sambou (2013), which probably resulted from the further grammaticalization of this verb. Mandinka also has a verb nǒo ‘overcome’, ‘master’ that can be used as a modal auxiliary expressing possibility.

3.3.4 Obligation In Manding languages, kán ‘be equal’ has grammaticalized as a modal auxiliary of obligation. This grammaticalization path, also attested in Soninke, can be explained with reference to an abstract meaning ‘have an affinity with’ which, depending on the nature of the complement, may concretize either as ‘be equal to’ or ‘be obliged to’. See Creissels (2017a) for a more detailed presentation of cross-linguistic data supporting this analysis, and a discussion.

3.3.5 Intention In Manding languages, the quotative kó ‘say’ (cf. Section 4.1) is commonly used to express not only imminence (as already mentioned in Section 3.2.9), but also intention (as in Koyaga, where the etymological meaning of À kó mabri làda ‘He wants to repair the car’ is something likes ‘He says [he will] repair the car’).10

3.4 Tense 3.4.1 Past Three etymologically distinct past markers can be found in Manding languages. Mdk nǔŋ (also found, outside Manding, in Koranko) is cognate with a nominal root *núŋ still attested as the first formative of the adverb núntò ‘previously’ and of compounds such as númmòolú ‘people of ancient times’. 10 To be precise, in Koyaga, kó as an intentional marker does not have the same tonal properties as the quotative kó, and a precise hypothesis about the construction in which the reanalysis occurred (a question that I leave open) should account for this fact.

Grammaticalization in Manding languages

715

A plausible etymology of KMnk tèré is the completive form of tèré ‘find / be found’, since the same grammaticalization path is also attested for example in Soninke, where the completive of ˋñí ‘find / be found’ is used as a past auxiliary. As regards the past marker of Bambara tǔn, I am not convinced by Dumestre’s proposal according to which it would be cognate with túgún ‘again’, but I am aware of no evidence of a more convincing etymology.

3.4.2 Future The use of nǎ ‘come’ in the incompletive aspect as a future auxiliary is common in Manding languages, and the future predicative marker ná found among others in Bambara results from the further grammaticalization of this verb. In Mandinka, the potential marker sè ~ sì (cognate with a verb found in Mandinka as síi ‘reach’) has uses in which its meaning is not very different from that of a future marker. Similarly, in Koyaga, the intentional marker that grammaticalized from the quotative marker kó (see 3.3.5 above) can be interpreted as expressing future rather than intention, depending on the nature of the subject.

3.5 Subject/object agreement As a rule, Manding languages have neither subject nor object agreement. However, a phenomenon analyzable as the emergence of a (still very marginal) mechanism of subject agreement can be observed in Mandinka. In Mandinka, the completive positive marker in the transitive construction is yè, with the variant ŋà in immediate contact with the first person non-emphatic pronouns ŋ́ (1sg) and ŋ̀ (1pl); Interestingly, at least some speakers use ŋà cpl.tr.1pl and ŋá cpl.tr.1sg instead of yè in constructions with an emphatic 1st person pronoun in subject function, even if this 1st person pronoun is not in immediate contact with the predicative marker, as in (15b). (15) Mandinka a. Ń-tè lè yè fòolèesúwòo dádâa. 1sg-emph foc cpl.tr bicycle.d repair ‘It’s me who repaired the bicycle.’ b. Ń-tè lè ŋá fòolèesúwòo dádâa. 1sg-emph foc cpl.tr.1sg bicycle.d repair same meaning as (15a)

716

Denis Creissels

3.6 Preverbs Most Manding languages have two or three productive preverbs, cognate with postpositions: lá-, mǎ-, plus tɔ́- ~ dɔ́- ~ rɔ́- in the western and south-western varieties that have a homonymous locative postposition. The problem of the causative preverb lá- has been discussed in Section 3.1.2. The semantic analysis of the preverb mǎ- is particularly difficult, since it is lexicalized to a considerable extent. A meaning of superficial or attenuated action is however discernible, and this is consistent with the hypothesis that the postposition mǎ developed from a noun meaning ‘body’ which initially grammaticalized with the meaning ‘in contact with’. As regards rɔ́-, its contribution to the meaning of the verbs it derives is generally not difficult to relate to the locative meaning expressed by the corresponding postposition, but as already mentioned, I am aware of no evidence suggesting a plausible lexical origin.

3.7 Non-finite verb forms 3.7.1 Postposition > non-finite verb form marker Mandinka and other Western Manding varieties have a non-finite verb form used in the complementation of modal verbs and in combination with the locational copula in incompletive auxiliary function. This form has no nominal property, but the coincidence between the suffix -lá marking this form and the postposition lá suggests the reanalysis of a construction in which the verb used nominally combined with the postposition lá.

3.7.2 ‘Place’ > purpose converb (supine) marker Compounds consisting of a verbal lexeme plus the noun yɔ́rɔ́ ‘place’ (or its equivalent in Manding varieties other than Bambara: Mdk dùláa, etc.) are frequently found in combination with movement verbs, in a construction whose literal meaning is ‘move to the place dedicated to V-ing’, but which is commonly interpreted as ‘move in order to do V-ing’, without necessary reference to a place specifically dedicated to the activity in question. This can be viewed as the first step in an evolution by which forms whose initial meaning is ‘V-ing place’ could be converted into purpose converbs.11

11 Perekhvalskaya (2016) analyzes similar facts in South Mande languages.

Grammaticalization in Manding languages

717

4 Grammaticalization of complex constructions 4.1 Complement clauses: quotative > complementizer Manding languages have a quotative marker kó. This quotative marker is an invariable word used to introduce reported speech in a construction in which it is followed by a quotation and preceded by an NP representing the person to which the quotation is attributed. A postpositional phrase representing the addressee may be inserted between kó and the quotation, in which case the quotative is optionally repeated, as in Ex. (16). (16) Mandinka Kèwôo kó díndíŋò yé (kó) “ŋ́ kòntôŋ!” man.d quot child.d ben quot 1sg greet ‘The man told the child to greet him.’ Ex. (17) illustrates the use of the quotative kó in complementizer function. The complement clause follows the matrix clause, within which it is represented by a cataphoric pronoun occupying the position that corresponds to its role in the argument structure of the main verb (in this example, the object position between the predicative marker and the verb). (17) Mandinka Ŋ́ ŋá ài lôŋ [kó à bé dòokúwóo sòtó-lá ŋ́ yè]i 1sg cpl.tr 3sg know quot 3sg loccop work.d get-inf 1sg ben ‘I know that he will get work for me.’

4.2 Relative clauses 4.2.1 Demonstrative > relativizer All Manding languages have a relativizer mɛ̂n or mîn, typically used in the construction illustrated by Ex. (2) in Section 1.3.3. In most Manding varieties, this relativizer has no other possible function, but in the varieties spoken in Ivory Coast, it is homonymous with a demonstrative. Among the Central Mande languages that have a relatively close relationship with Manding, a demonstrative mɛ is also found among others in Vai and in Jeli. This suggests that the Manding relativizer resulted from the grammaticalization of a demonstrative, and that in most Manding varieties, this demonstrative lost its original demonstrative function after grammaticalizing as a relativizer.

718

Denis Creissels

4.2.2 The grammaticalization of the distinction between stage level and individual level properties in Mandinka relative clauses As explained in more detail in Creissels and Sambou (2013: 472–473) and Creissels and al. (2015), the construction of Mandinka relative clauses is sensitive to the distinction between stage level and individual level properties. The construction also found in other Manding varieties tends to be reserved to the expression of stage level properties, for example ‘the tree whose fruits we ate yesterday’ (Ex. [18a]), whereas relative clauses referring to a permanent characteristic of the referent they describe, such as ‘a tree whose fruits are eaten by monkeys’ (Ex. [18b]), have a special construction which in this particular case is lit. ‘a tree which you know that monkeys eat its fruits’. (18) Mandinka a. ŋ̀ ŋà yír-òo mîŋ díŋ-ò-lú dómò kúnùŋ 1sg cpl.tr tree.d rel fruit.d-pl eat yesterday ‘the tree whose fruits we ate yesterday’ b. yír-òo mîŋ í yè à lôŋ kó sùlôo-lú kà à díŋò-lú tree.d rel 2sg cpl.tr 3sg know that monkey.d-pl incpl 3sg fruit.d-pl dómò eat ‘a tree whose fruits are eaten by monkeys’ This grammaticalization of an expression whose literal meaning is ‘which you know that’ as a relativizer implying that the relative clause refers to an individual level property can be analyzed as an instance of contact-induced grammaticalization, since the same phenomenon is pervasive in the Atlantic languages spoken in Senegal – Creissels et al. (2015).

4.3 Adverbial clauses 4.3.1 ‘When’, ‘since’ The preposition kàbíní ‘since’ (cf. 2.5.2) is also used as a subordinating conjunction. In this use, it has a broader temporal meaning, and depending on the context can be interpreted as ‘since’ or ‘when’.

4.3.2 ‘As soon as’ Dɔ́rɔ́n occurs in independent clauses as an adnominal or clause-final particle whose meaning is equivalent to English ‘only’. Postposed to the first clause in a sequence of two clauses, it expresses the temporal relationship ‘as soon as’.

Grammaticalization in Manding languages

719

(19) Mandinka À yè ŋ́ háyínâŋ dóróŋ, à yè ŋ́ súutê e. 3sg cpl.tr 1sg see only 3sg cpl.tr 1sg recognize ‘As soon as he saw me, he recognized me.’ In Mandinka, the temporal relationship expressed by the construction illustrated in (19) can also be expressed by means of the manner relativizer ñáamìŋ ‘the way in which’ occupying the same position as dóróŋ in a sequence of two clauses.

4.3.2 ‘Before’ The preposition yànní or sání ‘before’ (see Section 2.5.2) is also used as a temporal conjunction with the same meaning.

4.3.3 ‘It has found that’ > ‘whereas’ Inserted at the beginning of the second clause in a biclausal sequence, Mdk à yè à tàrá, lit. ‘it has found that’, expresses the same contrast as English whereas. The same phenomenon is observed not only in the other Manding languages, but also elsewhere in the Mande language family.

4.3.4 ‘If’ and ‘and’ In Manding, the conjunction introducing hypothetical and conditional clauses is homonymous with the comitative preposition ní, also used as an additive coordination marker (‘and’) – Ex. (20). (20) Bambara a. Ń yé Sékù ní Fàtú wélé. 1sg cpl.tr Sékou and Fatou call ‘I called Sékou and Fatou.’ b. Ní Fàtú nà-ná, wárî dí à mà! if Fatou come-cpl.intr money.d give 3sg to ‘If Fatou comes, give her the money!’ The fact that the same coincidence is found in a number of unrelated languages suggests the existence of a grammaticalization path ‘and’ > ‘if ’, whose details are however unclear to me.

720

Denis Creissels

4.3.5 Purpose > sequential Adverbial subordination with a meaning of purpose is expressed in Manding by means of biclausal constructions in which the second clause (the subordinate purpose clause) is in the subjunctive, and can be reduced to an infinitival phrase if its subject is co-referent with that of the matrix clause. However, this construction can also express sequentiality without any purpose implication, and depending on the context, the sequential reading may be the only one available.

4.3.6 The quotative kó in adverbial subordination The quotative kó can be optionally added at the beginning of subjunctive clauses expressing purpose. Alone or in combination with ní ‘if ’, the quotative kó can introduce similative clauses (‘as if ’). Inserted at the beginning of the second clause in a biclausal sequence, the quotative kó can be interpreted as ‘saying that’, but also as ‘under the pretext that’, ‘on the ground that’.

4.3.7 Causal conjunctions Inserted at the beginning of the second clause in a biclausal sequence, the noun sábú ‘cause’ (borrowed from Arabic) serves as a conjunction expressing a causal relationship. Among the other causal conjunctions, kàtúgú ‘because’ might have originated from an infinitive, since kà is an infinitive marker. There is a verb túgú ‘do something deliberately’, ‘insist on doing something’, but this etymology is semantically dubious.

4.4 Insubordination 4.4.1 Infinitival phrases as interrogative clauses In some Manding languages, infinitival phrases can be used as independent interrogative clauses whose function is to check or confirm a command (for example, in Bambara, kà táa, infinitive of táa ‘go’, uttered with an interrogative intonation, is interpreted as ‘Should I / we go?’).

Grammaticalization in Manding languages

721

4.4.2 ‘Until’ > obligation marker In Manding languages, fó is a preposition/conjunction used in particular to introduce subordinate clauses with the temporal meaning ‘until’,12 but also introduces independent subjunctive clauses expressing obligation. (21) Mandinka a. Kèlôo ké-tà fó tìlóo bòyí-tà. fight.d take_place-cpl.intr until sun.d fall-cpl.intr ‘The fight took place until the sun went down.’ b. Fó í yè bóoróo mǐŋ. oblig 2sg subj medecine.d drink ‘You must drink the medecine.’ This might be a mere coincidence, or perhaps the result of the borrowing of French (il) faut, but the same coincidence is observed in Soninke and in Bozo with completely different forms (Soninke mà, Bozo kara), which suggests that this is rather the result of a grammaticalization process. The most plausible explanation is the insubordination of until-clauses in a construction in which the matrix clause was perhaps an imperative clause with a verb like ‘strive’, ‘do one’s best’, or something like that.

5 Other patterns of grammaticalization and reanalysis 5.1 Associated motion Mandinka has a centripetal marker nǎŋ whose probable origin is the grammaticalization of nǎa jǎŋ ‘come here’ – Creissels (2014). This kind of marker is not found in the other Manding languages, and its development in Mandinka is probably a consequence of contact with Atlantic languages.

5.2 Manner adverbs 5.2.1 ‘Matter, manner’ > manner adverb marker Most Manding varieties have a small number of manner adverbs in which an adjective is preceded by a first formative kó- (kú- in Kita Maninka and other Western 12 On fó ‘until’, see also Sections 5.5 and 5.6.

722

Denis Creissels

varieties), such as kóɲùmán ‘well’ < ɲùmán ‘good’. The lexical origin of this formative kó- is probably the noun kó. In most Manding varieties, kó is only attested with the meaning ‘matter’, which is not a very good candidate as the lexical source of a manner adverb marker, but this noun is also clearly attested at least in Marka with the meaning ‘manner’.

5.2.2 ‘Do’ > manner adverb marker Mandinka does not have manner adverbs of the type dealt with in 5.2.1, but has a relatively productive suffix -ké, as in kéndè-ké ‘well’ > kéndè ‘good, healthy’ – Ex. (22). (22) Mandinka Fàatú yè búŋò fítà kéndè-ké. Fatou cpl.tr room.d sweep good-man ‘Fatou swept the room well.’ As explained in more detail in Creissels and Sambou (2013: 319–320), synchronically, words such as kéndèké can only be analyzed as de-adjectival adverbs, but some of their properties (in particular, but not only, their tonal behavior) suggest that they result from the reanalysis of a construction in which the object of ké ‘do’ was a verb used nominally and combined with an adjective. In the case of Ex. (22), this means that, before this reanalysis, the adjective kéndè modified the verb used nominally (fítá-kéndè ‘good sweeping’), the noun which is synchronically in object function fulfilled the function of genitival modifier (búŋò fítá-kéndè ‘good sweeping of the room’), and the whole phrase was the object of ké ‘do’, subsequently reanalyzed as a suffix converting adjectives into manner adverbs.

5.3 Copulas 5.3.1 Imperative of ‘see / look’ > copula As illustrated by Ex. (23), most Manding languages have an equative and/or locational copula yé homonymous with the verb yé ‘see’. (23) Kita Maninka a. Ń dí Sékù yé kúnùn. 1sg cpl.tr Sékou see yesterday ‘I saw Sékou yesterday.’

Grammaticalization in Manding languages

723

b. Kóngò yè ǹ nà. hunger.d cop 1sg postp ‘I am hungry.’ (lit. ‘Hunger is in me’) In (Creissels 2017b), I review evidence from languages belonging to various branches of West Mande (including Manding languages) supporting the reconstruction of a grammaticalization path Imperative of ‘see / look’ > ostensive marker > copula. This evolution explains not only the creation of this copula yé in the history of Manding languages, but also the existence of sporadic copular-like uses of the ostensive marker fílέ ‘here is’ < fílέ ‘look’.

5.3.2 Others The other copulas found in Manding languages are mú (positive equative copula),13 bέ (positive locational copula), and tέ (negative copula). I am aware of no evidence of a plausible lexical origin of mú. As regards bέ, Tröbs (2003) evokes the possibility that it originates from a Mande demonstrative found in Bobo as bě. Finally, the negative copula might be etymologically a negative form of a verb ‘be’ found in Bobo as tī, whose all other form would have been lost in Manding languages.

5.4 ‘Limit’ > ‘only’ Across Manding varieties, ‘only’ is most commonly expressed as dɔ́rɔ́n (whose lexical origin is not known), but dǎn ‘only’, whose lexical origin is the noun dǎn ‘limit’, is also attested in Bambara and Marka.

5.5 ‘Until’ > ‘except’ In Manding languages, the preposition/conjunction fó ‘until’ (already mentioned in Section 4.4.2 for its grammaticalization as an obligation marker) also expresses ‘except’. Interestingly, in other languages (cf. for example Italian fino (a)), ‘until’ has grammaticalized as a scalar additive particle (‘even’), i.e., with an opposite meaning. A plausible explanation is that, in its original spatio-temporal meaning, ‘until x’ refers to the right limit of an interval but leaves open the choice between a rightopen interval …x[ (> ‘except x’) and a right-closed interval …x] (> ‘even x’).

13 In Bambara, the equative copula mú is not attested as such, but Bambara has an identificational copula dǒn whose plausible etymology is dě (focus marker) + mú (equative copula).

724

Denis Creissels

5.6 Alternative question marker > yes/no question marker Fó, already mentioned in sections 4.4.2 and 5.5 with the meanings ‘until’, ‘except’, and ‘it is necessary that’, has also uses in interrogative clauses whose relationship to its other meanings is not clear to me. What is however clear is that its use as a yes/no question marker, as in (24b), can be explained as resulting from the conventionalization of the ellipsis of the first conjunct in the construction illustrated in (24a), in which it serves as a marker of alternative question. (24) Mandinka a. Í yé sùbôo bóndì, fó wùlôo yè à tǎa lè? 2sg cpl.tr meat.d put_aside altern dog.d cpl.tr 3sg take foc ‘Did you put aside the meat, or did the dog take it?’ b. Fó wùlôo yè sùbóo tǎa? q dog.d cpl.tr meat.d take ‘Did the dog take the meat?’

6 Summary In this paper, I have tried to show that Manding languages show evidence of the following grammaticalization/reanalysis paths, either as emerging processes in present-day Manding languages, or as processes having operated in the history of the Mande language family. In this enumeration, small capitals highlight relatively little-known grammaticalization scenarios for which at the same time the evidence from Manding languages is particularly convincing. – ‘and’ > associative plural marker > ordinary plural marker (2.2.1) – locative postposition > alienable possession marker (2.3) – proprietive pronoun > alienable possession marker (2.3) – demonstrative > definiteness marker > default determiner (2.4.1) – 3rd person singular possessive > definiteness marker (2.4.2) – ‘one’ > indefinite pronoun/determiner (2.4.3) – ‘thing’ > indefinite pronoun/determiner (2.4.4) – ‘kind’ > negative determiner (2.4.5) – noun (in particular body part noun) > postposition (2.5.1) – proprietive pronoun > benefactive postposition (2.5.1) – ‘say’ > similative preposition or conjunction (2.5.2, 4.3.6) – infinitive of ‘come from’ and ‘go’ > prepositions ‘starting from’ and ‘until’ (2.5.2) – ‘mother’ > augmentative derivational affix (2.7) – ‘hinder, fail’ > privative derivational affix (2.7) – ‘mouth/side’ > derivational affix meaning ‘place occupied by…’ (2.7)

Grammaticalization in Manding languages

– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

725

‘child’ > resultative derivational affix (2.7) ‘child’ > diminutive derivational affix (2.7) reflexive pronoun > 2nd person pronoun (2.8.1) ‘owner’ > emphatic 3rd person pronoun (2.8.3) anticausative construction > passive construction (3.1.1) ‘do’ > causative marker (3.1.2) ‘leave’ > permissive causation marker (3.1.3) ‘the like of ’ > reciprocal pronoun (3.1.5) ‘the like of ’ > affix forming names of co-participants from verbs (3.1.5) ‘do’ > antipassive marker > antipassive nominalization marker > nominalization marker (3.1.6) locational copula > progressive marker > incompletive marker (3.2.1) postposition > completive aspect marker (3.2.2) resultative marker > perfect marker (3.2.3) ‘finish’ > perfect marker (3.2.2) ‘finish’ > ‘already’ (3.2.3) ‘fall’ > inchoative auxiliary (3.2.4) ‘move toward’ > inchoative auxiliary (3.2.4) ‘want’ or ‘search’ > auxiliary ‘almost do’, ‘come close to’ (3.2.5) ‘return’ > repetitive auxiliary (3.2.6) ‘remain’ > continuative or habitual auxiliary (3.2.7) ‘separate, escape’ > auxiliary ‘eventually do’ (3.2.8) ‘say’ / quotative > auxiliary expressing intention or imminence > future marker (3.2.9, 3.3.5, 3.4.2) postposition > hortative marker > subjunctive marker (3.3.1) ‘don’t come’ > prohibitive marker (3.3.2) ‘reach’ > potential marker > future marker (3.3.3, 3.4.2) ‘overcome, master’ > potential marker (3.3.3) ‘be equal’ > obligative auxiliary (3.3.4) ‘find’ > past marker (3.4.1) ‘come’ > future marker (3.4.2) personal pronoun > person agreement marker (3.5) ‘body’ > spatial postposition ‘in contact with’ / preverb with a meaning of attenuated or superficial action (3.6) postposition > non-finite verb form marker (3.7.1) ‘place’ > purpose converb (supine) marker (3.7.2) ‘say’ > complementizer (4.1) demonstrative > relativizer (4.2.1) ‘which you know that’ > relativizer implying reference to individual level properties (4.2.2) ‘only’ (restrictive particle) > ‘as soon as’ (conjunction) (4.3.2) ‘it has found that’ > ‘whereas’ (4.3.3)

726 – – – – – – – – – – – – –

Denis Creissels

‘and’ > ‘if ’ (4.3.4) purposive construction > sequential construction (4.3.5) ‘say’ > ‘under the pretext that’, ‘on the ground that’ (4.3.6) ‘cause (noun) > because (conjunction) infinitival phrase > interrogative clause (4.4.1) ‘until’ > obligation marker (4.4.2) ‘come here’ > centripetal marker (5.1) ‘matter, manner’ > manner adverb marker (5.2.1) ‘do’ > manner adverb marker (5.2.2) imperative of ‘see/look’ > ostensive predicator > copula (5.3.1) ‘limit’ (noun) > ‘only’ (restrictive particle) (5.4) ‘until’ > ‘except’ (5.5) alternative question marker > yes/no question marker (5.6)

Abbreviations altern = alternative question marker, antip = antipassive, ben = benefactive, caus = causative, cop = copula, cpl = completive, d = default determiner, dem = demonstrative, emph = emphatic, foc = focus marker, funct = functive, gen = genitival linker, incpl = incompletive, inf = infinitive, intr = intransitive, loc = locative, loccop = locational copula, man = manner adverb marker, neg = negative, oblig = obligative, postp = postposition, propr = proprietive pronoun, q = yes/ no question marker, quot = quotative, rel = relativizer, sg = singular, subj = subjunctive, tr = transitive

References Cobbinah, Alexander & Friederike Lüpke. 2009. Not cut to fit – zero coded passives in African languages. In Matthias Brenzinger & Anne-Maria Fehn (eds.), Proceedings of the 6th World Congress of African Linguistics, 153–165. Cologne: Rüdiger Köppe. Creissels, Denis. 1997a. Postpositions as a possible origin of certain predicative markers in Mande. Afrikanistische Arbeitspapiere 50. 5–17. Creissels, Denis. 1997b. Une tentative d’explication de particularités typologiques de la négation en mandingue. Mandenkan 32. 3–21. Creissels, Denis. 2009. Le malinké de Kita. Cologne: Rüdiger Köppe. Creissels, Denis. 2013. Le maninka du Niokolo (Sénégal Oriental): esquisse phonologique et morphosyntaxique, liste lexicale, textes glosés. Mandenkan 49. 1–218. Creissels, Denis. 2014. Le développement d’un marqueur de déplacement centripète en mandinka. In Carole de Féral, Maarten Kossman & Mauro Tosco (eds.) In and out of Africa, languages in question, In Honor of Robert Nicolaï, Vol. 2 Language contact and language change in Africa, 95–102. Louvain-la-Neuve & Walpole MA: Peeters. Creissels, Denis. 2015a. Valency properties of Mandinka verbs. In Andrej Malchukov & Bernard Comrie (eds.), Valency classes in the world’s languages, vol. 1, 221–260. Berlin: Mouton De Gruyter.

Grammaticalization in Manding languages

727

Creissels, Denis. 2015b. Polysemy patterns involving non-scalar additive particles in Subsaharan languages: the coordinative connection. Paper presented at the 48th annual meeting of the SLE, Leiden, 2–5 September. Creissels, Denis. 2017a. Similarity, suitability, and non-epistemic modalities (volitionality, ability, and obligation). In Yvonne Treis & Martine Vanhove (eds.), Similative and equative constructions: A cross-linguistic perspective, 79–89. Amsterdam: John Benjamins. Creissels, Denis. 2017b. Copulas originating from the imperative of see / look verbs in Mande languages. In Walter Bisang & Andrej Malchukov (eds.), Unity and diversity in grammaticalization scenarios. Berlin: Language Science Press. 45–66. Creissels, Denis. Forthcoming. Reconstructing the origin of the Soninke voice system. Creissels, Denis & Pierre Sambou. 2013. Le mandinka: phonologie, grammaire, textes. Paris: Karthala. Creissels, Denis, Sokhna Bao Diop, Alain-Christian Bassène, Mame Thierno Cissé, Alexander Cobbinah, El Hadji Dieye, Dame Ndao, Sylvie Nouguier-Voisin, Nicolas Quint, Marie Renaudier, Adjaratou Sall & Guillaume Segerer. 2015. L’impersonnalité dans les langues de la région sénégambienne. Africana Linguistica 21. 29–86. Dimmendaal, G. 2011. Historical linguistics and the comparative study of African languages. Amsterdam & Philadelphia: John Benjamins. Dumestre, Gérard. 2003. Grammaire fondamentale du bambara. Paris: Karthala. Dumestre, Gérard. 2011. Dictionnaire bambara-français. Paris: Karthala. Grégoire, Claire. 1984. Le syntagme déterminatif en mandé nord. Journal of African Languages and Linguistics. 6(2). 173–193. Greenberg, Joseph. 1978. How does a language acquire gender markers? In Joseph Greenberg (ed.), Universals of human language Volume 3: Word structure, 49–81. Stanford: Stanford University Press. Idiatov, Dmitry. 2020. Perfective marking conditioned by transitivity status in Western Mande: Constructional competition, specialization and merger. Diachronica 37(1). 43–82. Perekhvalskaya, Elena. 2016. “Oni prišli iz mesta ubyvanija kozy”: ob odnoj strategii grammatikalizacii v južnyx mande [From locative construction to relative time: On a strategy of grammaticalisation in South Mande]. Vestnik Sankt-Peterburgskogo universiteta. Serija 9. 2016. N° 1. 89–99. Tröbs, Holger. 2003. On the origin of some predicative markers in imperfective constructions in Manding. Mandenkan 38. 1–14. Vydrin, Valentin. 2009. On the problem of the Proto-Mande homeland. Вопросы языкового родства – Journal of Language Relationship 1. 107–142. Vydrin, Valentin. 2014. Infinitiv v jazyke bamana [The infinitive in Bambara]. In Mikhail Daniel, Ekaterina Ljutikova, Vladimir Plungian, Sergej Tatevosov & Olga Fedorova (eds.), Jazyk. Konstanty. Peremennyje. Pam’ati Aleksandra Jevgenjevicha Kibrika [Language. Constants. Variables. In memoria Alexandre Kibrik], 653–676. St. Petersburg: Aleteia.

Ronald P. Schaefer and Francis O. Egbokhare

17 Grammaticalization in Emai 1 Introduction Our principal objective in this paper is to identify some of the primary grammaticalization channels evident in Emai, an Edoid (Niger Congo phylum) language of southern Nigeria (Elugbe 1989; Williamson and Blench 2000). In essence, grammaticalization can be viewed as a process whereby lexical items or phrases as well as their meanings become reinterpreted as structures expressing semantically related grammatical functions. Although the literature has set forth a broad developmental outline for grammaticalization, encompassing initial subtle changes of form use that eventually result in phonetic reduction (Heine, Claudi, and Hunnemeyer 1991; Heine and Kuteva 2002; Heine and Song 2010; Kuteva 2001), our attention rests on early-stage grammaticalization (see Table 1). Our survey of Emai will consider grammatical functions evident today in its noun phrase, verb phrase, postverbal particle phrase, and preverb phrase. In each of these syntactic domains, functions have emerged that can be traced to lexical items or phrases, particularly verb phrases, and associated meanings. Functions have emerged in other domains as well but, due to time and space limitations, we leave for subsequent analysis a more comprehensive discussion.

2 Background By way of background, we note here some of Emai’s key structural features relevant to understanding its grammaticalization processes. Typologically, Emai exhibits canonical SVO word order with few prepositions and minimal segmental inflection. Basic word order is affected by noun phrase repositioning in focus or topic position in line with corresponding considerations of new and shared information. Furthermore, syntactic phrase boundaries as well as tone, both lexical and grammatical, play a fundamental role in constraining its phonological system. Tone in particular is integral to Emai inflection as well as auxiliary, preverb and verb expression. Emai distinguishes clause level tense-aspect. Perfective, imperfective and future each correlates with a distinct tonal pattern in the verb phrase. Contrasting tonal patterns in the subject phrase index event time relative to utterance time or a discourse determined reference time. Athough imperfective and future tense are segmentally marked, perfective is not. Monotransitive structures in the perfective illustrate the contrasting tonal patterns for the past perfective (pap) and the present perfective (prp). For past perfective (pap), verb tone is high downstep and subject https://doi.org/10.1515/9783110712735-003

730

Ronald P. Schaefer and Francis O. Egbokhare

phrase tone is right edge high. Present perfective (prp) shows a high tone verb (é ‘eat’) and a right edge low tone subject phrase. (1)

a. ólí ómóhé é' ólí émàè.1 the man pap.eat the food ‘The man ate the food.’ b. ólí ómòhè é ólí émàè. the man prp.eat the food ‘The man has eaten the food.’

Preceding the verb in clause structure are auxiliary and preverb categories in addition to segmental aspectual forms. Auxiliaries code modality (deontic and epistemic) as well as relative tense. Class members include deontic hortative í (2a), the epistemic markers certaintive za (2b) and hypothetical kha (2c), as well as relative tense anterior ke (2d). (2)

a. àlèkè í è ólí émàè Aleke hor eat the food ‘Aleke should eat the food.’ b. àlèkè záà é ólí émàè. Aleke prp.ded eat the food ‘Aleke must have eaten the food.’ c. áléké khà é ólí émàè. Aleke pap.hyp eat the food ‘Aleke would have eaten the food (but she didn’t).’ d. áléké ké' é émà. Aleke pap.ant eat yam ‘Aleke ate yam from then on.’

A constituent type that follows auxiliaries but precedes a matrix clause verb consists of preverbs. There are six preverb category types in Emai. They include aspectual, evaluative, temporal, manner demonstrative, subject attributive and quantity. Of these, we illustrate aspectual conative òó and repetitive che, evaluative duu, temporal bóbo and distal manner demonstrative ìyó. The remaining two categories, subject attributive and quantity, are discussed in Section 6.

1 Orthographic conventions for Emai are generally consistent with those in Schaefer and Egbokhare (1999), Schaefer and Egbokhare (2007) as well as Schaefer and Egbokhare (2017), where o represents a lax mid back vowel, e a lax mid front vowel, and vb a voiced bilabial approximant. With respect to tone, acute accent marks high, grave accent signals low, and acute accent followed by an apostrophe designates high downstep.

Grammaticalization in Emai

(3)

731

a. àlèkè ó ó' è ólí émàè.2 Aleke sc con eat the food ‘Aleke has gone to eat the food.’ b. áléké ché' é émà. Aleke pap.rep eat yam ‘Aleke again ate yam.’ c. àlèkè dúù gbé ólí ófè. Aleke prp.eval kill the rat ‘Aleke has killed the rat for no reason.’ d. áléké bó'bò é ólí émàè Aleke pap.promptly eat the food ‘Aleke ate the food promptly.’ e. áléké íyó' gbé ólí ákhè á. Aleke pap.dmd break the pot cs ‘Aleke broke the pot that way.’

Additional dimensions of Emai grammatical structure are evident in noun and verb phrases. Noun phrases have an internal structure consisting of nominal modifiers positioned as predeterminers or postdeterminers. Relative to head noun, post-determiners manifest the following linear order: head-genitive-adjective-kindred-numeralrelative clause-demonstrative. All but cardinal numerals induce high tone spread, right to left, on their head noun. Among postdeterminers, only genitive and adjective are marked by linkers, respectively, ísì/ésì and li. The most prominent predeterminer is the definite article. It expresses number by inflectional prefix (ó-lì ~ é-lì), as do a relatively small number of head nouns (á-khè ~ é-khè ‘ceramic pot sg ~ pl’). Verb phrases are characterized by simple and complex predicates. Simple predicates are linked to intransitive, transitive and ditransitive verbs (Schaefer and Egbokhare 2015). Complex predicates profile event participants or event aspectual character. They do so via two primary complex predication types: verbs in series and verb plus postverbal particle. Verb series predications identify various participant relations, among them ablative, addressee, instrument, locative goal and standard of comparison. Instrument, marked by verb re, illustrates a serial verb predication articulating an event participant. (4)

údò gbé ákhè á. ólí óvbèkhàn ré the youth prp.take stone break pot cs ‘The youth has used a stone to break a pot.’

Participant relations are also expressed by verb-particle predications (Schaefer and Egbokhare 2010). Postverbal particles code applicative, change of location and pro-

732

Ronald P. Schaefer and Francis O. Egbokhare

jected adherence participants. Applicative li and change of location o predications illustrate two primary event participant functions: recipient for applicative (5a) and proximal displacement for change of location (5b). (5)

émà lí ònwìmè. a. ólì òkpòsò shén the woman prp.sell yam app farmer ‘The woman has sold yam to a farmer.’ émà ó vbí úkpódè. b. ólì òkpòsò nwú the woman prp.carry yam cl loc road ‘The woman has put yam onto the road.’

As well, event aspectual character is conveyed by verb series or verb- particle predications. Verbs in series articulate, among others, telic end state, confirmation of event fulfillment / result, and manner of directional change of locative state. An initial impression of manner of directional change is (6), where manner verb sua ‘push’ precedes the location change verb o ‘enter.’ (6)

ókò ó vbí ékóà. ólí ómòhè súá the man prp.push mortar enter loc room ‘The man has pushed a mortar into the room.’

Confirmation of a lexical implication is illustrated in (7a–b). Some Emai verbs in simple predications leave unspecified an implication of fulfillment / result. In (7a– b), neither initial verb in series, hoo ‘search for’ or hoo ‘wash’, confirms its fulfillment end state. That is, neither the found state of the searched for object nor the cleaned state of the washed object is asserted. The lexical implications of these verbs remain unexpressed in simple predications. In complex predications, like (7a– b), confirmation of fulfillment is articulated by a following verb in series, respectively, mie ‘find’ and fuan ‘clean.’ Just as lexical implications can be confirmed, they can also be cancelled, as will be illustrated in Section 5. (7)

a. òjè hóó ólí ákhè míé. Oje prp.search.for the pot find ‘Oje has searched for the pot and found it.’ ólí úkpùn fúán. b. òjè hóó Oje prp.wash the cloth clean ‘Oje has washed the cloth clean.’

Event aspectual character is also coded by postverbal particles in complex predicates. Relevant particles are change of state (cs) a and temporal perspective (temp) lee. cs a occurs with change of material state verbs and conveys a telic end state. It

Grammaticalization in Emai

733

is obligatory with some verbs, e.g., gbe, and optional with others, e.g., guogho, where cs a presence confirms a telic end state. (8)

a. òjè gbé ólí ákhè á/ *gbé ólí ákhè. Oje prp.break the pot cs prp.break the pot ‘Oje has broken the pot.’ ólí úkpóràn. b. òjè gúóghó Oje prp.break the stick ‘Oje has broken the stick (disjointed but not separated).’ ólí úkpóràn á. c. òjè gúóghó Oje prp.break the stick cs ‘Oje has broken the stick apart (disjointed and separated).’

The postverbal particle for temporal perspective is lee. It brings into the discourse foreground an event temporal boundary, either inception or termination. Depending on clause level aspect, lee highlights inception or termination. In the imperfective with continuous aspect (9), where inception is foregrounded, lee sense is ‘already.’ In the perfective, where termination is profiled, lee identifies a terminal boundary as ‘already completed, finished.’ Furthermore, it should be noted that postverbal particles, unlike postverbal adverbs, do not induce high tone spread in preceding constituents under perfective aspect (compare 9b émàè vs 9c émáé). (9)

a. òjè ò ó è ólí émáé lèé. Oje sc c eat the food temp ‘Oje is already eating the food.’ b. òjè é ólí émàè léé. Oje prp.eat the food temp ‘Oje has already eaten / finished off the food.’ c. òjè é ólí émáé èghéènà. Oje prp.eat the food recently ‘Oje has eaten the food recently.’

3 Grammaticalization in the noun phrase Grammaticalization in Emai has led to functions that impact the structure of its noun phrase. Bound pronominals for reflexive and reciprocal functions have developed to replace a noun phrase in discourse. As well, relational nominals preceding or following a noun phrase have emerged to express functions related to spatial relations and emphatic exclusivity. There is also a correlative conjunction that has

734

Ronald P. Schaefer and Francis O. Egbokhare

evolved. Each of these functions will be illustrated and related to a source noun or, in the instance of correlative conjunction, a copula. Grammaticalization channels illustrated by Emai noun phrases have been recognized in the linguistics literature. In the World Lexicon of Grammaticalization (Heine and Kuteva 2002), the lexical item for the human body is recognized as a source for reflexive and reciprocal pronoun functions. Similarly, body-part and landscape terms as a source domain for static spatial relation functions are well recognized. And a cardinal numeral as source for an emphatic exclusive marker is cited in the World Lexicon of Grammaticalization. Not cited there is a lexical source for correlative conjunction.

3.1 Bound pronouns Emai has grammaticalized two bound pronoun forms. Each is syntactically limited by a clause internal antecedent relation. Pronouns for reflexive and reciprocal functions have as their source the bodypart term égbè ‘body.’ In its reciprocal function, égbè ‘each other, one another’ occurs in non-subject position and requires a plural antecedent. égbèi . (10) élí ívbèkhàni gbé the youths prp.beat rec ‘The youths have beaten each other.’ In its object reflexive function, égbè requires an immediately following pronoun marked as accusative (égbé mè / é / óì [self my your him/her] ‘myself / yourself / herself ’). The pronoun must agree in person and number with its antecedent. Reflexive égbè is limited to non-subject position. As will be shown in Section 6, object reflexive égbè contrasts with emphatic subject reflexive dobó. (11) ólí ómòhèi sá égbé óìi . the man prp.shoot self him ‘The man has shot himself.’ As a body-part term, égbè occurs as noun phrase head in various grammatical relations, including subject. It can occur with a modifier, for example cardinal numeral èvá (11) or Emai’s associative morpheme ísì: égbé ísì àlèkè ‘body of Aleke / Aleke’s body.’ (12) égbè èvá ríì vbì ìtébù. body two prp.be loc table ‘Two bodies are on the table.’

Grammaticalization in Emai

735

3.2 Noun phrase orientation Grammaticalization has led to relational nominals providing a spatial or discourse orientation for an entire noun phrase. These nominals either precede or follow a noun phrase (i.e., a lexical head and any predeterminers or postdeterminers) or a personal pronoun. Those that precede a noun phrase express spatial orientation. Following a noun phrase is a nominal that orients participant identity in discourse and information structure. Members in each class have derived from the grammaticalization of a noun that otherwise occurs as a noun phrase head. A relational nominal with a discourse orientation is òkpá ‘alone, only.’ It conveys an exhaustive or exclusionary function regarding class membership for an argument in focus position (PF) or topic position. òkpá modifies proper nouns (òjè òkpá ‘Oje alone’), emphatic personal pronouns, which are limited to focus or topic position regardless of number (wèwè / vbàvbà òkpá ‘you sg/ you pl alone), and common nouns modified by a determiner (ólí éwé lí óbín' òkpá [the goat R black alone] ‘the black goat alone’). vbí égbóà. (13) a. ólí éwé lí óbín' òkpá lí ó ríì the goat r black alone pf it pap.be loc backyard ‘It was the black goat alone that was in the backyard.’ vbí égbóà. b. ólí éwè òkpá lí ó ríì the goat alone pf it pap.be loc backyard ‘It was the goat alone that was in the backyard.’ Exhaustive marker òkpá is derived from cardinal numeral òkpá ‘one.’ Like its cardinal counterpart, it fails to induce high tone spread on its head noun, in contrast to other non-cardinal numerals as well as other postdeterminers. Relational nominals referring to spatial relations constitute a substantial class. Members express topological and directional orientation functions relative to a noun phrase referent. Relational nominals precede an object pronoun, e.g., óì, or a head noun and any predeterminer. (14) ólì ùgín ríì vbí ékéín àkpótì/ óì. the basket prp.be loc inside box it ‘The basket is inside the box / it.’ Each spatial nominal has grammaticalized from a source noun denoting a body part or a landscape, as with ékéìn ‘inside’ from ékéìn ‘belly.’ Other body-part terms serving as source for a counterpart spatial relation include ùòkhò ‘back’ (úókhó úgbà ‘behind the fence’); égbè ‘body’ (égbé ìwè ‘beside the house’); ídámà ‘chest’ (ídámá úkpódè ‘middle of the road’); àgbàn ‘jaw’ (ágbán édà ‘edge of the river’); and ègèìn ‘crotch’ (égéín ìtèbù ‘under the table’). A landscape term as source for a spatial

736

Ronald P. Schaefer and Francis O. Egbokhare

relation is òkhùnmì ‘sky’ (ókhúnmí ókòó ‘top of the hill’). For some spatial relations, partial reduplication is obligatory, e.g., òtòì ‘ground, soil’ (ótóòtóí édà ‘bottom of the river’). For most other terms, reduplication is optional and reveals intensification or attenuation of the distance between spatial entities. Only one spatial form (ésèsé in ésèsé úkpódè [center road] ‘center/ midst of the road’) has no body-part or landscape term as source.

3.3 Correlative conjunction Grammaticalization is evident for one Emai conjunction type. Conjunction khi precedes each noun phrase standing in a correlative relation. Restricted to topic position, correlatives require a plural anaphor (íyàìn ‘them’) in their corresponding grammatical relation in a matrix clause. In addition, khi exhibits low tone and induces high tone spread (right to left) in its following noun phrase. Thus, each noun phrase following a correlative marker exhibits lowering of left edge high tone in a sequence of high tones. émàè ní íyàìn. (15) khì ólí áwá khì ólì òlógbó, òjè nwú crc the dog crc the cat Oje prp.carry food app them ‘As for the dog and the cat, Oje has given food to them.’ Correlative conjunction khi has grammaticalized from equational identity copula khi. One of five ‘be’ forms (Schaefer and Egbokhare 2006), khi re-aligns the discourse identity of two noun phrases as referentially conjoint, as opposed to referentially disjoint. A speaker might thus employ khi when there is evidence that a hearer holds to a disjoint interpretation. In its copula frame, khi precedes a simple noun phrase but follows a noun phrase in focus position marked by li. Subject for this copula is identity pronoun í, which shows an inherent high tone. As a consequence, khi’s tone is low unless preceded by an auxiliary or preverb (16a–b). (16) a. èkpèn lí í khì ójé ísì éànmì. leopard pf id be chief ass animal ‘It is the leopard who is chief of the animals.’ b. èkpèn lí í mà khí ójé ísì éànmì. leopard pf id cer be chief ass animal ‘It is the leopard who is certainly chief of the animals.’

Grammaticalization in Emai

737

4 Grammaticalization in the verb phrase Verbs serve as source for several functions occurring in Emai complex predicates, where we find two or more verbs in series. Verbs in series convey a single event, not a series of events. For discussion, we select several of these functions, namely allative, momentary impact, sustained contact, and receiver for message transfer. Uniformly, forms expressing these functions are positioned after an initial verb in series. Source verbs for these functions tend to be intransitive or transitive; many derive from a spatial motion or coincidence / contact domain. Grammaticalization of some of these functions has been recognized in the linguistics literature. In the World Lexicon of Grammaticalization, for example, lexical items with meanings related to arrive, go or see are identified as source for an allative function. Less often, if at all, is there a source noted for functions related to momentary or sustained contact, or receiver for message transfer. Across these functions, the source for a grammaticalization outcome is a lexical item expressing spatial direction, spatial coincidence or a similative condition. For several of these forms, grammaticalization is evident less in a complete phonological, semantic and syntactic change than in subtle lexical or collocational changes. We thus take as indicative of grammaticalization, for example, a verb shift from transitive to intransitive or an argument collocational shift from a place noun to a human noun. Several of the grammaticalization channels advanced in this section can be thought of as a ‘snapshot’ in time of a process yet to be completed.

4.1 Allative In Emai, an allative function has grammaticalized from the change of locative state verb ye ‘move toward, to.’ The allative function as it pertains to the domain of possession change identifies non-permanent change of possession. Allative predications receive a ‘take to/for’ interpretation with the noun phrase following ye coding event recipient. Recipient position requires a human noun and permits pronouns that are accusative (e.g., óì) but not locative (*èvbò ‘there yonder’). In complex serial verb predications, allative ye follows a manner verb, e.g., nwu ‘carry’, ze ‘scoop’, or roo ‘pick out.’ (17) a. òjè nwú émà yé ònwìmè. Oje prp.carry yam all farmer ‘Oje has taken yam to a farmer.’ b. òjè zé èkèn yé òhí. Oje prp.scoop sand all Ohi ‘Oje has taken sand to Ohi.’

738

Ronald P. Schaefer and Francis O. Egbokhare

Allative marked predications contrast with applicative structures marked by postverbal particle li (Section 2). The latter expresses permanent possession transfer. On the other hand, possession transfer coded by allative ye is construed as short term or temporary. Allative ye in complex predicates fails to co-occur with verbs of permanent possession transfer, e.g., shen ‘sell’ (18). As a consequence, permanent possession verbs express transfer only with applicative li. émà lí ònwìmè. (18) a. ólì òkpòsò shén the woman prp.sell yam app farmer ‘The woman has sold yam to a farmer.’ émà yé ònwìmè. b. *ólì òkpòsò shén the woman prp.sell yam all farmer ‘The woman has sold yam to a farmer.’ As a simple predicate, ye expresses directional motion toward a locative reference point. It takes a direct object of place; a human noun direct object is unacceptable. Instead, a human noun as reference point is syntactically elaborated. Either it is embedded as subject in a pronominally headed relative clause of place with ri ‘be located’ (19b). Or, it is marked as a possessor by ésì in an associative phrase where the possessum of place has been ellipted (19c). Likewise, direct object position of ye permits pronouns which are locative (èvbò) and not accusative (*óì). (19) a. ólí ómòhè yé áfúzé'/ èvbò. the man prp.move.toword Afuze there.yonder ‘The man has moved toward Afuze / there.’ b. ólì òkpòsò yé ébé' ólí óvbékhán ríì. the woman prp.move.toward where the youth PAP.be ‘The woman has moved toward where the youth is.’ c. ólí ómòhè yé ésì ólí óvbèkhàn. the man prp.move.toward ass the youth ‘The man has moved toward the youth’s place.’

4.2 Message receiver Emai has grammaticalized distinct receiver functions for contrasting directions of message transfer. Regarding the latter, a message from a sender to a receiver may be simply unidirectional, carrying no expectation of a response. Or, it may be bidirectional, expecting thereby a response from the receiver. In Emai, message transfer employs the verb ye ‘send someone’, which in simple transitive predications permits

Grammaticalization in Emai

739

a direct object that is human (but not an inanimate one that denotes a message, e.g., úhùnmì ‘message’). (20) òjè yé óvbèkhàn/ *úhùnmì. Oje prp.send youth message ‘Oje has sent a youth.’ But ‘send’ ye also occurs in complex predications. These distinguish between unidirectional and bidirectional messaging by contrastive marking of message receiver. Complex predications show ye as an initial verb. Its subject serves as sender and its direct object as message, or perhaps more precisely, message carrier. As the second verb in series, either allative ye or ree codes receiver. When allative ye appears as second verb in series, the messaging act is viewed as unidirectional. It can be roughly translated as ‘send a message to someone.’ There is no expectation of a reply. Each verb in series takes a direct object. Series initial ye ‘send’ permits nouns in this position that are message carriers, either human or abstract (limited to úhùnmì ‘message’). It also allows accusative pronouns (e.g., óì). On the other hand, allative ye direct object permits human nouns (òkpòsò in [21a]), human nouns as possessors marked by ésì in associative phrases where possessum of place has been ellipted (ésí ólì òkpòsò in [21b]) or proper nouns of place (áfúzé' in [21c]). As with possession transfer, the direct object of allative ye allows accusative pronouns (óì) but not pronouns that are dative (*áìn) or locative (*èvbò ‘there yonder’). (21) a. òjè yé óvbèkhàn / úhùnmì yé ólì òkpòsò. Oje prp.send youth message all the woman ‘Oje has sent a youth / a message to the woman.’ b. òjè yé óvbèkhàn / úhùnmì yé ésì ólì òkpòsò. Oje prp.send youth message all ass the woman ‘Oje has sent a youth / a message to the woman’s place.’ c. òjè yé óvbèkhàn / úhùnmì yé áfúzé'. Oje prp.send youth message all Afuze ‘Oje has sent a youth / a message to Afuze.’ When ree occurs as second verb in series, the messaging act is viewed as bidirectional. It can be roughly interpreted as ‘send a message after someone.’ There is in essence the expectation of a reply. Again, each verb in series takes a direct object. That of series initial ye behaves as it did in allative ye complex predications. As for ree, its direct object allows only human nouns or an accusative pronoun (óì), not a dative pronoun (*áìn).

740

Ronald P. Schaefer and Francis O. Egbokhare

(22) a. òjè yé óvbèkhàn réé ólì òkpòsò. Oje prp.send youth mr the woman ‘Oje has sent a youth after the woman.’ b. òjè yé úhùnmì réé ólì òkpòsò. Oje prp.send message mr the woman ‘Oje has sent a message after the woman.’ Grammaticalization of a receiver function has already been partially discussed under allative ye in Section 4.1. There, we saw that directional ye’s unbounded movement toward a spatial goal was semantically reanalyzed as temporary, not permanent possession change in ‘take’ predications. In a similar fashion, ye in the message domain is associated with unidirectional rather than bidirectional communication. As for ree, it occurs synchronically in a simple intransitive predication. It expresses a similative relation of physical resemblance between human entities. Its subject allows only human nouns, as does its immediately following similative phrase marked by preposition bí ‘like.’ Without bí, ree never appears as a transitive verb expressing physical resemblance or any other meaning. Its grammaticalization course must therefore have included a shift in verb subcategorization from intransitive to transitive as well as a semantic shift from the similative to the communicative domain. bí érá óì. (23) òjè réè Oje prp.take.after like father his ‘Oje physically resembles / takes after / appears like his father.’

4.3 Momentary impact The verb gbe with meanings related to ‘hit’ has grammaticalized in complex predicates to code a momentary impact ‘against’ function. It specifies the limited duration of impact between a moving entity relative to a stationary object. Entity movement derives from the action of the verb in series preceding gbe. Relative to its momentary contact function, gbe takes an accusative pronoun (e.g., óì), and not a locative pronoun (e.g., *èvbò). (24) ólì òkpòsò sán-nó àmè gbé ólí óvbèkhàn. the woman prp.spray-ds water mi the youth ‘The woman has splashed water against the youth.’ That gbe alone is not sufficient to express momentary impact of event participants is revealed by the causative/inchoative alternation. In the inchoative (25b, 26b), gbe

Grammaticalization in Emai

741

must stand in series with either the erstwhile causative verb, e.g., gbulu ‘roll’, or a semantically related surrogate, e.g., for fi ‘throw’ that would be de ‘fall.’ (25) a. òjè gbúlú ólí ókò gbé ìmátò. Oje prp.roll the mortar mi car ‘Oje has rolled the mortar against the car.’ b. ólí ókò gbúlú gbé ìmátò. the mortar prp.roll mi car ‘The mortar has rolled against the car.’ (26) a. ólì òkpòsò fí úkpóràn gbé ùdékèn. the woman prp.throw stick mi wall ‘The woman has thrown a stick against the wall.’ b. ólí úkpóràn dé gbé ùdékèn. the stick prp.fall mi wall ‘The stick has hit / fallen against the wall.’ As a simple transitive predicate, gbe and its direct object exhibit senses related to momentary contact with impact. With a percussion instrument as direct object, gbe sense is ‘beat, play;’ with a human noun, it is ‘beat, hit;’ and with plural body-part term ábò ‘hands’, it is ‘clap.’ Significantly, simple gbe predications do not permit a place noun as direct object, e.g., *òjè gbé ùdékèn ‘Oje hit the wall’, as do corresponding complex predications with gbe (26). This suggests that collocational restrictions on gbe (from inanimate, non-place to inanimate including place) have been lifted through the process of grammaticalization. (27) a. òjè gbé ólì ìbè. Oje prp.beat the drum ‘Oje has beaten / played the drum.’ b. òjè ò ó gbè ólí óvbèkhàn. Oje sc c beat the youth ‘Oje is beating / hitting the youth.’ c. òjè ò ó gbè ábò. Oje sc c hit hands ‘Oje is clapping his hands.’

4.4 Sustained contact Several complex predications express a variety of functions related to sustained contact. These include oblique contact (oc), suspended attachment (sa), affixed attachment (aa), restrained contact (rc) and inclusive attachment (ia).

742

Ronald P. Schaefer and Francis O. Egbokhare

Sustained contact predications articulate an end state in which an entity coded as direct object assumes a position of prolonged contiguity with a referent marked by locative preposition vbi. The particular nature of the sustained contact is specified by the second verb in series, translated variously in English as against, on, in, onto, into, to. Movement of the referent of the direct object is initiated by the action of the first verb in series, although this initial verb is not the exclusive source for the sense of the complex predication, e.g., lean against, hang in, press onto, squeeze against, and pound in/into. (28) a. òjè róó úkpóràn gón vbí ùdékèn. Oje prp.select pole oc loc wall ‘Oje has leaned the pole against the wall.’ b. òjè nwú ólì èkpà khúán vbí ólí óràn. Oje prp.carry the bag sa loc the tree ‘Oje has hung the bag in the tree.’ c. òjè tú ólí ébè báá vbì ùdékèn. Oje prp.spread the paper aa loc wall ‘Oje has pressed / stuck the paper onto the wall.’ ólí úkpún nyè vbì ùdékèn. d. òjè ò ó lùghù Oje sc c squeeze the cloth rc loc wall ‘Oje is squeezing the cloth against the wall.’ e. òjè kháán ìshé zé vbí úkhùèdè. Oje prp.pound nail ia loc door ‘Oje has pounded / stuck a nail in the door.’ Regarding the second verb in series, gon, khuan, baa, nye and ze, each varies in its degree of grammaticalization. None occurs as a simple predicate with an obligatory locative marked argument. Except for gon, none accepts the erstwhile direct object from its corresponding complex predicate as subject in a simple predicate. Moreover, there is one sustained contact form, ze, which no longer occurs as a simple predicate. It appears only as a semantically related reduplicated verb, as will be shown shortly. Grammaticalization in the sustained contact domain is thus reflected in a variety of linguistic indices: lifting of collocational restrictions on subject and direct object, permitting a locative marked argument and expunging a non-reduplicated form from the lexicon. Each sustained contact function has its source in an intransitive or transitive verb. Additional restrictions on these functions may occur. To start, complex predicates coded by gon express an oblique contact function (28a). This function appears in complex predicates that employ verbs such as roo ‘select’ and nwu ‘carry.’

Grammaticalization in Emai

743

As a simple intransitive predicate, gon exhibits two primary senses. One is the sense ‘become off-center, crooked, bent, bowed.’ In this function, it does not take an oblique argument marked by locative preposition vbi. (29) a. ólí úkpóràn/ áwé ísì òjè gón-ì. the stick legs ass Oje prp.become.crooked-f ‘The stick / The legs of Oje have gotten bent / bowed / crooked.’ vbí ùdékèn. b. *ólí úkpóràn gón-í the stick prp.become.crooked-f loc wall ‘The stick has gotten bent onto the wall.’ Also, as a simple intransitive, gon conveys the sense ‘lean’ under specific collocational and subcategorization conditions. It must occur in series with verb re ‘take’ whose direct object is the body-part noun égbè ‘body.’ Its subject must be a human noun, and it must show an argument of place marked by locative preposition vbi. At the very least, gon grammaticalization has lifted the re ‘take’ subcategorization restriction. (30) òjè ré égbè gón vbí óràn. Oje prp.take body lean loc wood ‘Oje has leaned on the wood with his body.’ Complex predicates with khuan articulate a suspended attachment function (28b). Besides nwu ‘carry’ in series, khuan occurs with verbs like fi ‘throw/toss’ and ze ‘scoop.’ As a simple transitive predicate, khuan has the sense ‘position by suspending / pulling apart / bending open.’ It takes a direct object that is restricted to a noun referring to trapping devices. Each device features a suspended fixture. khuan does not convey the complex predicate sense ‘hang’; it is rather a trap victim that assumes a vertically suspended position when the noun ífì ‘spring trap’ occurs as khuan direct object. Even more to the point, erstwhile complex predicate direct objects (e.g., èkpà ‘bag’) never appear as a khuan argument in simple predicates. Reanalysis of collocational restrictions thus underlies the grammaticalization of khuan.

(31) a. òjè khúán ífì. Oje prp.suspend spring.trap ‘Oje has set a spring trap (with a trip wire) / suspended a trap.’ b. òjè khúán ìkpàkúté'. Oje prp.suspend trap ‘Oje has set a snare trap (with metal teeth).’

744

Ronald P. Schaefer and Francis O. Egbokhare

Complex predicates with baa convey an affixed attachment function (28c). In addition to tu ‘spread’, baa complex predicates permit a verb like teen ‘press’ as the initial verb in series. In a simple predication as an intransitive or transitive form, baa fails to occur. It is restricted to complex predicates with the change of location (cl) particle o, where the sense is ‘insert’, or with a verb series featuring the verb de ‘reach’ to convey ‘join, participate.’ The ‘join, add to’ sense appears to be the semantic source for the complex predicate function ‘onto/against’ in ‘stick onto/against’ and ‘press onto/against.’ (32) a. òjè báá ìshé/ àgbèdé ó vbì òtòì. Oje prp.join nail needle cl loc ground ‘Oje has inserted a nail / a needle into the ground.’ b. òjè dé báá élí ívbèkhàn. Oje PRP.reach join the youths ‘Oje has joined / participated with the youths.’ In neither of these uses, however, does baa take an erstwhile direct object from its affixed attachment complex predicate (e.g., ólí ébè ‘the paper’) as its direct object. Again, it would appear that collocational restrictions have been adjusted through grammaticalization, as have subcategorization restrictions on the type of complex predicate in which baa can occur. (33) *òjè báá ólí ébè ó vbì òtòì. Oje PRP.join the paper cl loc ground ‘Oje has inserted the paper into the ground.’ Complex predicates with nye express a function of restrained contact (28d). Besides lughu ‘squeeze’, restrained contact predications permit, as initial verb in series, forms like gba ‘tie’, teen ‘press’, and khaan ‘pound.’ As a simple transitive predicate, nye has the sense ‘restrain, constrain, strain.’ Its direct object is collocationally restricted; it can only be a human noun or a bodypart noun. As (28d) showed, it must also allow inanimate objects as either understood subject or understood direct object. Furthermore, nye as a simple predicate does not permit a locative argument marked by preposition vbi (34d). It is thus restrictions on collocation and subcategorization of nye that are lifted through grammaticalization. (34) a. ólí ómò ò ó nyè ényè. the child sc c restrain breast ‘The infant is suckling a breast / is breast feeding.’

Grammaticalization in Emai

745

égbè. b. òjè ò ó nyè Oje sc c restrain body ‘Oje is straining his body.’ (as when relieving oneself) ábò nyényé òhí. c. òjè kpén Oje prp.position hands restrain Ohi ‘Oje has prodded Ohi with his hands. / Oje has pushed Ohi along.’ égbè vbí ùdékèn. d. *òjè ò ó nyè Oje sc c restrain body loc wall ‘Oje is restraining his body at / against the wall.’ Complex predicates with ze express an inclusive attachment function (28e). In addition to khaan ‘pound’ as the initial verb in series, predications with this function allow verbs like khuye ‘enclose someone in.’ The form ze no longer appears as a simple predicate. Instead, ze appears as a reduplicated shape, zeze, with the senses ‘congeal’, ‘become stiff ’, and ‘become arduous’ (35a–c), depending on intransitive subject. Although meanings associated with reduplicated ze express the coming together of elements into a single entity, which is required by ze complex predicates, there is no extant lexical verb with the required meaning. As well, none of the simple predicates conveying the ‘coming together’ sense allows a locative argument marked by preposition vbi (35d) Under this circumstance, grammaticalization is best viewed by assuming that ze, as a verb with a sense related to inclusive attachment, has been expunged from the Emai lexicon and that reduplicated forms with related senses are its relatives. zézé'-ì. (35) a. ólì àkàsán the maize.pap pap.congeal-f ‘The maize pap has congealed.’ b. ábó ísì òjè zézé-ì. hand ass Oje prp.become.stiff-f ‘The hand of Oje has become stiff.’ c. ólì òbìà zézé-ì. the work prp.become.arduous-f ‘The work has become arduous.’ zézé'-í vbí úkhùèdè. d. *ólì àkàsán the maize.pap pap.congeal-f loc door ‘The maize pap has congealed against the door.’

746

Ronald P. Schaefer and Francis O. Egbokhare

5 Grammaticalization of postverbal particles Verbs, in some cases multiple verbs, serve as source for functions realized in Emai complex predicates where we find a verb combining with a postverbal particle of change (Section 2). For discussion here, we select several of these functions, namely excessive quantitative achievement, terminal achievement, venitive spatial achievement, as well as event non-achievement. Uniformly, the forms conveying these functions are positioned after an initial verb and are final in their verb phrase. Source for these functions tend to be a single transitive verb or a multi-constituent verb whose significance is related to some sort of achievement. Grammaticalization along few of these channels has been recognized in the linguistics literature. In the World Lexicon of Grammaticalization, lexical items with meanings related to English come are recognized as source for a venitive function. More often, though, the literature has failed to mention functions related to quantitative achievement (excessive or sufficient), spatial coincidence achievement (relative to a deictic center or a terminal point), or event non-achievement (of a lexically implied end state or of an event intended but not realized). For these functions, a source lexical item in Emai is most often a verb of quantitative expression, spatial coincidence or moderated visual impression. Also noteworthy is that grammaticalization along these channels is evident less in a complete change of form and function for a lexical item than in subtle syntactic or semantic changes. For example, we find shifts in subcategorization from transitive to intransitive or a collocational shift from a place noun argument to a human one. Most of the grammaticalization channels in this section, too, can be thought of as a ‘snapshot’ of a change in process.

5.1 Excessive (quantitative) achievement Grammaticalization is evident in complex predicates with an excessive accomplishment function. Articulated by postverbal particle gbe, this function is translated in English by the quantitative adverbial ‘too much’. gbe has grammaticalized fully for this function, as reflected in reanalysis of subcategorization and collocational restrictions. In complex predicates, gbe follows transitive verbs such as gbe ‘beat’, da ‘drink’ and e ‘eat’, as well as intransitives like voon ‘fill’. Each of these verbs in simple predications denotes an incremental quantitative change. In complex predications, this change is signaled as excessive. The level of achievement marked by gbe is assessed relative to a preceding transitive direct object or intransitive subject. (36) a. òjè gbé òlólò gbé. Oje prp.beat Ololo ea ‘Oje has beaten Ololo too much.’

Grammaticalization in Emai

747

b. òjè dá ényò gbé. Oje prp.drink wine ea ‘Oje has drunk too much wine.’ c. òjè é émà gbé. Oje prp.eat yam ea ‘Oje has eaten too much yam.’ gbé'. d. ólì èkpà vóón the bag prp.be.full ea ‘The bag is too full.’ Postverbal particle gbe has evolved from a verb expressing quantitative achievement. As a simple predicate, gbe takes either a direct object or a double object complement. These simple predications have the sense ‘become filthy, infested’. Their common semantic thread (quantitative acquisition to a presumably undesireable level) with the complex predicate sense ‘too much’ is quantitative accumulation. Nonetheless, through grammaticalization gbe in complex predicates has lost its subcategorization privileges (transitive > Ø) and therefore its standing as a verb. ínwà. (37) a. ólí úkpùn gbé the cloth accumulate dirt ‘The cloth has become filthy.’ b. òjè gbé ínwà. Oje accumulate dirt ‘Oje has become filthy.’ c. òjè gbé ólí úkpún ínwà. Oje accumulate the cloth dirt ‘Oje has accumulated dirt on the cloth / made the cloth filthy.’ gbé ólí éékhò. d. èkhòì maggots accumulate the garden.egg ‘Maggots have infested the garden egg.’ It should be pointed out that excessiveness is not the only quantitative achievement recognized in Emai complex predications. There are serial verb predications that express less severe quantitative achievement, i.e., sufficiency (38a) and satisfaction (38b). Each occurs with a preceding verb that asserts incremental quantitative change without specifying an end state of quantitative achievement. (38) a. òjè dá ényò sé. Oje prp.drink wine be.sufficient ‘Oje has drunk enough / sufficient wine.’

748

Ronald P. Schaefer and Francis O. Egbokhare

b. òjè é émà khóón. Oje prp.eat yam be.satisfied ‘Oje has eaten yam to satisfaction.’

5.2 Terminal (spatial) achievement Grammaticalization is similarly pronounced in complex predicates of terminal achievement with so. English translation of this function relies on prepositional phrases of achievement such as ‘to the end / top’. In complex predications, so is preceded by verbs such as transitive heen ‘climb’ along with its direct object or intransitive shan ‘proceed / move to’. Each verb asserts an incremental change of locative state but leaves unspecified any terminal achievement for that change. For transitive verbs, terminal achievement is assessed relative to the direct object referent, while for intransitive verbs it is measured against movement of the referent of the grammatical subject. ólí óràn. (39) a. òjè héén Oje prp.climb the tree ‘Oje has climbed the tree.’ ólí óràn só. b. òjè héén Oje prp.climb the tree ta ‘Oje has climbed to the top of the tree.’ (40) a. ólí óvbèkhàn shán. the youth prp.proceed ‘The youth has proceeded.’ só'. b. ólí óvbèkhàn shán the youth prp.proceed ta ‘The youth has proceeded to the end.’ The terminal achievement sense of so is perhaps even more transparent in English translation when expressing non-motion events. When so is preceded by verb re ‘take’ and a direct object that is an abstract noun, the terminal achievement function is expressed by ‘to the end’. (41) òjè ré ólì èmòì só. Oje take the matter ta ‘Oje has taken the matter to its end.’ Terminal achievement as a function is expressed by erstwhile transitive verb so. It manifests two primary senses, both of which refer to an achievement of contact

Grammaticalization in Emai

749

between physical objects. When its direct object is úgùà ‘bone joint’, so sense is ‘set, join bone pieces that were fragmented’ (42a). When direct object is body-part noun óbò ‘hand’, so sense is ‘shake hands’ so long as the grammatical subject is either plural (élí ímòhè ‘the men’ in 42b) or, if singular, the direct object is accompanied by an external possessor (áléké in [42c]). When its direct object is a noun of place, so has a forceful achievement sense ‘collide with, crash into’. (42) a. ójé ó ò sò úgùà. Oje sc h join bone.joint ‘Oje joins bones / sets bones.’ óbò. b. élí ímòhè só the men prp.join hands ‘The men have shaken hands.’ c. òjè só áléké óbò. Oje prp.join Aleke hands ‘Oje has shaken Aleke’s hand / hands with Aleke.’ d. ójé só' ùdékèn. Oje pap.collide.with wall ‘Oje collided with a wall / crashed into a wall.’ Each sense of so is thus expressed by a transitive predication. However, in its terminal achievement function, expressed by a complex predicate with only one direct object, so is never followed by a direct object. Nor is it understood as taking a direct object. When so follows an initial intransitive verb, there is no direct object (40b). It appears, therefore, that through grammaticalization so has undergone a subcategory shift from transitive > Ø. Moreover, its sense has been reanalyzed from ‘join x and y’ to ‘join x to terminus of x’, a sense shift that appears reflexive in nature.

5.3 Venitive Emai has grammaticalized a venitive function in complex predicates for verb re ‘arrive, rise’. This function assumes an entity undergoing a change in material state or change in locative state relative to a deictic center, that is, space shared with the speaker. The verb preceding re denotes either an incremental change of state with an implied end state that is unspecified, but is open to venitive specification, or a manner activity open to venitive specification, that is a change of location. Translation equivalents engage English forms of emergence: prepositions out, up or verbs arise, bring. The venitive function of re is evident with a range of intransitive and transitive verbs in series. In simple predications, intransitive ze ‘grow’ and transitive filo

750

Ronald P. Schaefer and Francis O. Egbokhare

‘thrust, throw’, assert an incremental change of state on the vertical axis. Under these conditions, each verb implies but does not assert a perceptible end state, such as would be coded by English out. In series with re, where ze and filo precede re, the resulting complex predicate asserts a perceptible end state, e.g., English ‘out’, with the speaker / hearer world as deictic center. (43) a. ólí ókà zé-ì./ the maize prp.grow-ven ‘The maize has grown.’

ólí ókà ò ó zé. the maize sc c grow ‘The maize is growing.’

b. ólí ókà zé ré'. the maize prp.grow ven ‘The maize has sprouted / grew out.’ (44) a. ólí ókà ò ó fì-lò íké'. the maize sc c thrust-ds shoot ‘The maize is growing / developing shoots.’ b. ólí ókà fí-ló íké' ré. the maize prp.thrust-ds shoot ven ‘The maize has sprouted / thrusted out shoots.’ Other change predicates reveal a similar situational condition. The simple transitive predicate shoo ‘awaken’ expresses an incremental change with an implied end state that is unspecified. In complex predicates (45b), the end state is specified by venitive re, which leads to the meanings ‘arise, get up’. (45) a. ólì òkpòsò shóó òjè. the woman prp.awaken Oje ‘The woman has awakened Oje.’ b. òjè shóó ré'. Oje prp.awaken ven ‘Oje has gotten up / arisen.’ (from sleep) In complex predicates with a manner verb in series initial position, a venitive function is also evident. When verbs nwu ‘carry’, ze ‘scoop’ and others appear in simple transitive predications, they assert event manner along with movement of a possessed object. In complex predicates, these verbs in series with re assert an end state where subject and direct object referents of the manner verb come to occupy a deictic center space. English translation is ‘bring’, regardless of manner verb. (46) a. ólì òkpòsò nwú émà. the woman prp.carry yam ‘The woman has carried yam / *brought yam.’

Grammaticalization in Emai

751

émà ré. b. ólí ókpósó nwú' the woman pap.carry yam ven ‘The woman brought yam / *the woman carried yam and arrived.’ èkèn. (47) a. ólí ókpósó zé' the woman pap.scoop sand ‘The woman scooped sand.’ b. ólí ókpósó zé' èkèn ré. the woman pap.scoop sand ven ‘The woman brought sand.’ Simple predicates with re manifest two senses. Each is realized by a transitive verb that takes a direct object of place. Note that this contrasts with expression of the venitive function, where re is obligatorily intransitive. One meaning of verb re is ‘rise to a position on the vertical plane’. This sense requires a subject that is inanimate and a direct object of place. Should the direct object noun not denote place, it must be syntactically elaborated for spatial reference, for example by the spatial orientation nominal ókhúnmí ‘top of ’. (48) ólí údúkpù ré ókhúnmí àmè/ *àmè. the coconut prp.rise top water water ‘The coconut has risen to the top of the water.’ In addition, transitive re has the sense ‘arrive at a position on the horizontal plane’. It takes a subject that is human and a direct object of place. In particular, direct object position requires a locative noun (ègùàì ‘court’) or a syntactically elaborated non-locative noun (ógúí ébàbò ‘place of divination’). As direct object, re accepts neither a deictic locative pronoun (e.g., *èvbò) nor an accusative pronoun (e.g., *óì). Grammaticalization of the venitive function appears to have followed a pathway somewhat akin to that of terminal achievement so. That is, an erstwhile verb taking a direct object has undergone a subcategorization shift from transitive > Ø. (49) a. ólí évbóó ré' ègùàì. the village pap.arrive court ‘The village (people) arrived at court.’ b. ólí ómòhè ré ógúí ébàbò. the man prp.arrive place.of divination ‘The man has arrived to perform divination.’

5.4 Event non-achievement Emai has grammaticalized two functions related to event non-achievement or deactivation. Each is coded by a multi-verb constituent within a complex predicate. In a

752

Ronald P. Schaefer and Francis O. Egbokhare

general way, these functions pertain to event non-achievement or deactivation in the shared universe of discourse. More specifically, though, forms that code each function impact distinct components of event structure. In the instance of ba kun, a lexical implication of fulfillment or result for an activated or realized event is deactivated in discourse. In translation, ba kun is rendered by English adverbial expressions such as in vain, without success, or unsuccessfully. Relative to constituent fee ghoo, it is the intention to undertake an event that is conceptually highlighted. Either an intended event is declared to be unrealized / to have not transpired, or the actual occurrence of an event is declared to be disingenuous. In either case, a stipulation of event “occurrence” is voided. English translation tends to rely on verbs ‘try’, ‘feign’ or ‘pretend’. We now take up each of these forms and their functions as postverbal particles undergoing grammaticalization. The constituent ba kun in complex predicates deactivates or suspends an implication of event fulfillment provided by a preceding verb. It follows change of state verbs like hoo ‘wash’ and hoo ‘search for’. For these verbs in simple transitive predications an implicated state of fulfillment is unspecified. However, the implication is specifiable via context or in a complex predicate by verbs such as fuan ‘clean’ and mie ‘find’ (Section 2). Simple predications of unspecified fulfillment admit ba kun, thus leading to the sense ‘in vain, unsuccessfully’. (50) a. òjè hóó ólí úkpùn bá kùn. Oje prp.wash the cloth naf ‘Oje has washed the cloth without success.’ b. òjè hóó ólí ómòhè bá kùn. Oje prp.search.for the man naf ‘Oje has searched for the man in vain.’ When we individuate the components of ba kun to assess their verb status, we meet with partial success, revealing thereby the onset of grammaticalization. As a simple predicate, ba has the sense ‘stalk, spy on and so to pursue’. Its meaning is dependent on exercising the visual mode of perception. Syntactically, ba kun limits subject and direct object position arguments to nouns that are animate. In addition, ba shows a transitive / detransitive alternation that allows for direct object omission. (51) a. òjè ò ó bá ínì. Oje sc c stalk elephant ‘Oje is stalking an elephant’ b. òjè ò ó bá. Oje sc c stalk ‘Oje is stalking / hunting.’

Grammaticalization in Emai

753

As for kun, it fails synchronically to appear as an intransitive or transitive verb. It combines only with ba to convey the sense ‘stalk, spie on, pursue without desired effect, in vain’. ólí ófè kún. (52) ólí áwà bá the dog prp.stalk the rat in.vain ‘The dog has stalked the rat in vain.’ The form kun is not semantically empty however. In a predication with ba, it is necessary for the function of deactivating event fulfillment. By extension, kun is also fundamental to the deactivation function of the ba kun constituent in complex predicates, where it follows verbs conveying an implication of unfulfilled end state. From these conditions, we assume that grammaticalization has eroded the syntactic standing of kun as a verb. Concomitant with this erosion has been the loss of subcategorization properties, since ba kun in complex predicates no longer takes direct objects, not even an understood direct object. The constituent fee ghoo in complex predicates has a two-fold function. It declares that an event has not transpired, even though it was intended, or it asserts that the intention that motivated event occurrence is disingenuous. fee ghoo follows transitive (khuae ‘raise’) or intransitive (u ‘die’) change of state verbs, rejecting activity and stative verbs. Its predication sense is ‘try / feign, pretend’. (53) a. òjè khúáé ólì ùkòdò féé ghòò. Oje prp.raise the pot nai ‘Oje has tried to raise the pot (but did not).’ b. ójè ló ù féé ghòò. Oje pred die nai ‘Oje will feign death / Oje pretends to die.’ As a predicate, fee ghoo has two distinct senses, both related to the domain of sensory experience. With a subject that is a human noun and a direct object that is also human, it has the visual perception sense ‘inspect/ examine’ (54a). With a subject that is a human noun and a direct object that is an inanimate noun, fee ghoo realizes the gustatory sense ‘taste’ (54b). ólí óvbèkhàn ghóó. (54) a. òjè féé Oje prp.weaken the youth look.at ‘Oje has inspected / examined the youth.’ b. òjè féé òmì ghóó. Oje prp.weaken soup look.at ‘Oje has tasted the soup. / Oje has tried the soup.’

754

Ronald P. Schaefer and Francis O. Egbokhare

We now individuate the components of fee ghoo to assess their verb status. This exercise, like that undertaken with ba kun, leads to only partial success. Still, evidence from syntactic and semantic erosion suggests the onset of grammaticalization. Relative to fee ghoo’s expression of two distinct sensory modes, it is important to recognize that in simple predicates ghoo articulates events for only the visual mode. ghoo in simple predicates occurs as a transitive verb with the visual sense ‘look at, watch, perceive in full view’. It takes a subject that is animate and a direct object that is inanimate or animate. (55) a. ólí áwà ò ó ghòò ólì òkpòsò. the dog sc c look.at the woman ‘The dog is looking at the woman.’ b. ójé ghóó ìtèlìvíshón òdè. Oje pap.look.at television yesterday ‘Oje watched television yesterday.’ c. ójé ghóó úgbèá ísì óvbíóìmí. Oje pap.look.at killing ass orphan ‘Oje watched the killing of the orphan.’ Turning to fee, it fails to appear as a synchronic verb. Despite this, the form fee is not devoid of meaning, as suggested by phonologically related forms. Synchronically, forms phonologically related to fee occur as lexical adjectives and adverbs. Several have a polysyllabic shape with identical consonants of high sonority intervening between vowels. Under these conditions in Edoid languages, it is not uncommon for highly sonorous consonants to become effaced (Elugbe 1989: 111–113), especially in lexical items of high frequency. Furthermore, these phonologically related forms exhibit sense relations akin to that of the fee ghoo constituent. In toto, these forms are suggestive of an earlier semantic reality, if not syntactic verb status. Forms related to fee of the fee ghoo constituent have in common a sense that bears on a weakened state or attenuated condition. This commonality applies to tactile and visual sensory modes of experience. For example, a form féé in proverbial statements occurs as an adjective with the sense ‘gentle or weak’ as applied to a fire and its production of heat. (56) a. éráín féé lí ó ò gbè ókhònmì. fire weak pf sc h kill sick.person ‘It is a weak fire that kills a sick person.’ A second form, féléfélé, appears as an attributive and predicative adjective. It manifests a meaning related to the tactile sense, since it refers to items of very fine and thin texture, like silk.

Grammaticalization in Emai

755

féléfélé. (57) a. ólí úkpùn ú the cloth prp.be very.thin/fine ‘The cloth is very fine.’ b. úkpún lì féléfélé cloth r very.thin/fine ‘very fine cloth’ There is also a postverbal adverb with the shape féléfélé. Its meaning relates to object proximity or distance. In effect, it denotes a condition of extreme closeness, near coincidence of spatial entities. (58) ókhá sí kéé égbé ótóí féléfélé. cotton.tree move near side ground very.close ‘The cotton tree drew very near to the ground.’ Each of these fee related forms refers to an attenuated or weakened condition for a dimension with more robust possibilities. It seems worthwhile in this circumstance to assume that the fee of fee ghoo is semantically associated with a process of attenuation. Relative to ghoo and its ‘look at, perceive in full’ perceptual sense, we take fee to have a mitigating effect. That is, it reduces the full visual impression otherwise denoted by ghoo to one that is less so, i.e., a partial impression. It is this “less than full” notion that is consistent with the ‘inspect / examine’ and ‘taste’ senses of fee ghoo as a predicate. One can easily assume that some part of a whole rather than the whole itself is perceived in any given act of inspection / examination or tasting. We thus conclude that fee moderates the fuller sensory impression stipulated by ghoo. To translate this moderating effect, at least with regard to visual impression, we rely on English verbs ‘weaken’ or ‘diminish’. Similar to what happened with the ba kun constituent, grammaticalization of fee ghoo has eroded the verb status of fee. In conjunction with this shift in syntactic category, there has been a loss of subcategorization properties, for fee ghoo in complex predicates no longer takes a direct object, neither overtly nor covertly.

6 Grammaticalization of preverbs Via grammaticalization, a number of verbs and verb-noun collocations have been reanalyzed relative to their syntactic category, semantically adjusted and realigned as preverbs. The grammaticalized forms function synchronically as preverbs, of which there are six classes (Schaefer and Egbokhare 2000). Two of these classes reveal extensive grammaticalization. Moreover, it appears that preverb classes which have the potential to be farther from the verb and closer to auxiliaries, such

756

Ronald P. Schaefer and Francis O. Egbokhare

as aspectuals and evaluatives, either tend not to manifest grammaticalization or exhibit significantly greater opaqueness of their grammaticalization process. In contrast, those preverb classes ordered closer to the verb, quantity and subject attributive, as shown below, are shaped exclusively by grammaticalization. Grammaticalization along these channels has been little discussed in the linguistics literature. In the World Lexicon of Grammaticalization, the source for an emphatic reflexive function for grammatical subject is a lexical item for the human body or a nominal meaning owner or possessor. As well, the source for a function labeled “together” is cited as the cardinal numeral for one. Neither of these grammaticalization sources applies to Emai preverbs. Instead, a verb noun collocation employing a noun for ‘hand / arm’ is the source for an emphatic subject reflexive. A verb of binding is the source for a collective subject “together” function. Sources for other Emai preverb functions are less often, if at all, cited in the grammaticalization literature.

6.1 Quantity preverbs Four preverbs characterize one or another quantitative dimension applicable to intransitive subject or transitive direct object. Each has grammaticalized from a verb or a verb-noun collocation. Mutually exclusive, these preverbs fall into two classes: emphatic number (dobo, gba) and quantitative intensification (zemi, zeze). Only intensification forms, of all preverbs, correspond to an information question.

6.1.1 Emphatic number preverbs Grammaticalization has shaped Emai emphatic number preverbs. Both instances involve a phrasal shift from verb to preverb. There are two of these preverbs: collective gba and emphatic subject reflexive dobo + pronoun. They exhibit contrasting number restrictions with respect to their grammatical subject. Grammaticalized gba serves a collective function. With transitive or intransitive verbs, it requires a subject that is plural. gba appears minimally constrained by verb semantic type, since it occurs with change of state (59a) and stative (59b) predicates. (59) a. élí ímóhé gbá' híán ólí óràn. the men pap.col cut the wood ‘The men cut the wood together.’ b. élí ímóhé gbá' rì vbì ìwè. the men pap.col be loc house ‘The men were together in the house.’

Grammaticalization in Emai

757

Preverb gba has grammaticalized from a verb with the same phonological shape. As a transitive or ditransitive verb, gba has the sense ‘tie, tether, bind’. (60) a. òjè gbá ólí ómòhè/ ólí éwè Oje prp.tie the man the goat ‘Oje has tied the man / tethered the goat.’ b. è gbá ólí ómóhé íì. they prp.tie the man ropes ‘They have bound the man with ropes.’ A second emphatic preverb that has grammaticalized is dobó. It occurs as a constituent with a personal pronoun in the accusative. This preverb manifests an emphatic subject reflexive function. The pronoun must agree in number as well as person with the grammatical subject of its clause. As with other preverbs, the dobó emphatic reflexive is limited to pre-verbal position. It need not immediately follow the subject, as shown by a dobó phrase following predictive future ló, a tense auxiliary. (61) ójè ló dòbó óì híán ólí óràn. Oje pred refl him cut the wood ‘Oje will by himself cut the wood.’ Emphatic reflexive dobó is a grammaticalization of do óbò [bake hand] meaning ‘fortify one’s hand’. In simple predicates, verb do has the sense ‘bake / fire in a kiln’ (62a). It also occurs in figurative expressions with a kinship term as direct object (62b). For do óbò, it is this related figurative sense, ‘fortify / strengthen one’s hand’, that serves as source for the emphatic subject reflexive dobó. (62) a. yàn dó ólí ákhè. they prp.fire the pot ‘They have fired the pot.’ b. òjè dó óvbí óì. Oje prp.fortify offspring his ‘Oje has fortified his son with rituals.’ lit. Oje has fired his son.

6.1.2 Quantitative intensification preverbs Grammaticalization has led to quantitative intensification preverbs. Central to grammaticalization in this domain is the syntactic reanalysis of a verb phrase constituent, a verb-noun collocation, as a preverb. Contrasting conditions for a quantitative

758

Ronald P. Schaefer and Francis O. Egbokhare

dimension correlate with contrasting functions for the two intensification preverbs. Each selects a gradable or quantitative dimension within the meaning of its predicate and assigns one or another position on this dimension to either intransitive subject or transitive direct object. zemi ‘a lot, very, much’, for instance, exhibits an absolute function relative to quantitative intensification. It asserts that an entity has attained a quantitative dimension’s maximum value. zémì énghén. (63) a. ólì ògèdè the plantain prp.absi please ‘The plantain is very tasty / very pleasing.’ shén émà. b. ólì òkpòsò zémì the woman prp.absi sell yam ‘The woman has sold a lot of yam.’ Absolute intensifier zemi has as its source the verb-noun collocation ze émì ‘select/ choose things’, implying preparation for imminent departure. Its subject is limited to nouns that are human, and it inflects only for imperfective continuous. (64) àlèkè ò ó zè émì. Aleke sc c select things ‘Aleke is selecting her things. / Aleke is preparing to depart.’ Aside from the collocation with émì, ze has the sense ‘select, choose, nominate, elect’. It participates in transitive / detransitive alternations, the latter with an omitted direct object. (65) a. òjè zé úkpún lí óbín'. Oje prp.choose cloth r dark ‘Oje has chosen the dark cloth.’ b. òjè zé-ì. Oje prp.choose-f ‘Oje has chosen.’ Non-absolute intensifier (nabi) zeze signals that an intransitive subject or transitive direct object has realized a value exceeding a quantitative dimension’s minimum. Nonetheless, the maximum value remains unattained. It is perhaps best translated by English ‘not quite, a bit’. shén émà. (66) a. ólí ómóhé zé'zè the man pap.nabi sell yam ‘The man sold a bit / portion of his yam.’

Grammaticalization in Emai

759

dá. b. ólì òkpòsò zézè the woman prp.nabi be.tall ‘The woman is not quite tall.’ Non-absolute zeze has its source in reduplication of the verb ze ‘select, choose’. In this regard, we note that for many polar dimensions encoded by Emai verbs or adjectives, reduplication of form tends toward the negative pole, thus decreasing rather than increasing property value, as happens with verb bi ‘shift’ compared to its reduplicated form bibi ‘stagger’ or verb ka ‘become dry’ relative to kaka ‘to become hard’ (79). Partial reduplication with adjective forms reveals a similar effect, kéré ‘small’ but kékéré ‘very small’. Other lexical examples are noun ògbògbò ‘pneumonia’ from verb gbo ‘shiver’. (67) a. òjè bí égbè. Oje prp.shift body ‘Oje has shifted his body / moved his body aside.’ b. òjè ò ó bì-bí. Oje sc c shift-shift ‘Oje is staggering / moving in zig-zag fashion.’ Unlike all other preverbs, quantification intensifiers correspond to an information interrogative. Each answers a question of extent framed by the forms ébé'…í…se. (68) a. ébé' ólí ómóhé í' dà sé? how the man pap.man be.tall reach ‘How tall is the man?’ b. ò zémì dá./ he prp.absi be.tall. ‘He is very tall.’

ò zézè dá. he prp.nabi be.tall ‘He is not quite tall.’

6.2 Subject attributive preverbs Grammaticalization has resulted in four preverbs ascribing an attribute or quality to the grammatical subject. They reflect two dimensions: intentional and attitudinal. Regardless of dimension, members of the subject attributive class are mutually exclusive. They derive from erstwhile verb plus direct object collocations where bodypart or body-related nouns served as direct object. In clauses, subject attributives restrict inflection and auxiliaries. Most of these forms fail to accept imperfective aspect and future tense as well as deontic modality. As to the grammaticalization of these forms, we note the frequency of body-part terms as the direct object noun of the shift from V-N collocation > preverb. Moreover,

760

Ronald P. Schaefer and Francis O. Egbokhare

it appears that figurative use of such V-body part collocations may further solidify their standing as a unit amenable to grammaticalization. Relative to a completed event, intentional forms express subject motivation or ability. Subclass members are dábo ‘deliberately’, dóbo ‘accidentally, mistakenly’, miti ‘able to’, and dúda ‘defiantly’, as illustrated in (69). (69) a. ójé dá'bò fí ólì èkpà fí à. Oje pap.deliberately threw the bag extend cs ‘Oje deliberately threw the bag away.’ é ólí émàè. b. ójé dó'bò Oje pap.mistakenly eat the food ‘Oje mistakenly/accidentally ate the food.’ c. ójé mí'tì gbé ólí éwè. Oje pap.able kill the goat ‘Oje was able to kill the goat.’ d. ójé dú'dà é ólí émàè. Oje pap.defiantly eat the food ‘Oje defiantly ate the food.’ Collocational sources for intentional preverbs include the following: dábo ‘deliberately’ from the verb phrase daa ábò ‘prop hands’; dóbo ‘mistakenly’ from dee óbò ‘lower hand;’ miti ‘able’ from moe ítìn ‘have breath / spirit;’ and de ùdà from ‘reach defiance’. Some of these collocations exist as figurative expressions, while others do not. (70) a. ójé dáá' ábò. Oje pap.prop arms ‘Oje propped / raised his arms.’ (to do something) b. ójé déé' óbò. Oje pap.lower hand ‘Oje committed a mistake.’ lit. ‘Oje lowered his hand.’ étìn. c. ójé mòè Oje pap.have breath ‘Oje was strong.’ lit. ‘Oje had breadth.’ d. ójé dé' ùdà. Oje pap.reach defiance ‘Oje became defiant.’ lit. ‘Oje reached a defiant condition.’ Relative to their grammatical subject referents, attitudinal attributive preverbs characterize their mental attitude toward an event. Members of this subclass are dégbe

Grammaticalization in Emai

761

‘carefully, cautiously’, gbúdu ‘courageously’, totóbo ‘intensely’, and kakégbe ‘perseveringly’. (71) a. òjè ò ó dègbé è ólí émàè. Oje sc c carefully eat the food ‘Oje is carefully eating the food.’ b. òjè gbúdù é ólí émàè. Oje prp.courageously eat the food ‘Oje has courageously eaten the food.’ nwú ólì èkpà móé. c. ójé tótó'bò Oje pap.intensely take the bag have ‘Oje held the bag with great intensity.’ d. ójé káké'gbè ón ólì àgbó. Oje pap.by.persevering drink the potion ‘Oje drank the malaria potion with perseverance.’ Target-source relationships for these attitudinal preverbs are dégbe ‘carefully’ from dee égbè ‘lower body, duck down;’ gbúdu ‘courageously’ from gbe ùdù ‘position heart;’ totóbo ‘intensely’ from toto óbò ‘strengthen hand;’ and kakégbe ‘perseveringly’ from kaka égbè ‘stiffen body’. As with intentional sources, some attitudinal constituents still exist as figurative expressions. (72) a. ójé déé' égbè. Oje pap.lower body ‘Oje ducked down.’ lit. ‘Oje lowered his body.’ b. ójé gbé' ùdù. Oje pap.position heart ‘Oje got his courage.’ lit. ‘Oje positioned his heart.’ c. ójé tótó óbò. Oje pap.intensify hand ‘Oje got intense.’ lit. ‘Oje intensified his hand.’ d. ójé káká égbè. Oje pap.stiffen body ‘Oje stiffened his resolve.’ lit. ‘Oje stiffened his body.’

6.3 Attenuation of temporal rate Development of temporal rate preverbs through grammaticalization of verb-noun collocations appears to be ongoing as well. For several verb-based constituents of

762

Ronald P. Schaefer and Francis O. Egbokhare

event rate or event manner, it is incomplete. Each constituent in question precedes a main clause verb and exhibits a diminished range of properties associated with verb phrases. A straightforward category shift from verb-noun collocation to preverb for these constituents has not yet been achieved. A constituent with a partially grammaticalized character has emerged for a temporal rate function. It consists of the collocation fuen égbè re, which has a multiverb predicate, fuen and re, and a body-part noun, égbè ‘body’, as direct object. In tandem, these constituents express attenuation of event rate, i.e., ‘slowly, gently’. In complex predicates, fuen égbè re occurs in series with non-stative verbs open to temporal rate specification. The predication sense is ‘slowly, gently’. (73) a. ójé fúén' égbè ré tá étà. Oje pap.slow body ven speak words ‘Oje spoke slowly.’ égbè ré híán ólí óràn. b. ójé fúén' Oje pap.slow body ven cut the wood ‘Oje slowly cut the wood.’ As a stand-alone predicate, fuen égbè re is highly constrained. It takes as subject only a noun that is human and inflects only for imperfect habitual. As a verb, the form fuen does not occur synchronically without re, regardless of aspect. Its direct object noun is obligatorily égbè, no other body-part term is acceptable. In addition, égbè allows no syntactic elaboration by pre- or post-determiners and no substitution by, for example, an accusative pronoun (e.g., *óì). (74) ójé ó ò fùèn égbé ré. Oje sc h slow body ven ‘Oje moves slowly / carefully.’ Two additional constituents occur between the auxiliary phrase and the simple verb phrase. They do not attenuate rate so much as characterize event manner, and perhaps thereby rate. Their manner function is expressed syntactically as one or another form preceding a clausal verb. Relatively few of these manner constituents exist in Emai. One is soso ‘wandering fashion;’ it never occurs as sole verb in a simple predication. shán. (75) a. àlèkè ò ó sòsó Aleke sc c wander move.about ‘Aleke is wandering about / moving about in a wandering fashion.’ b. àlèkè ò ó shán. Aleke sc c move.about ‘Aleke is moving about / proceeding.’

Grammaticalization in Emai

763

Another constituent with a similar function but whose source has a different grammatical standing is bibi ‘stagger’. It appears in complex predicates, also with verb shan ‘move about’, and contributes the sense ‘stagger’. However, bibi appears in simple predicates without shan and expresses simply the sense ‘stagger’. (76) a. òjè ò ó bì-bí shán. Oje sc c shift-shift move.about ‘Oje is staggering about / is moving about in a staggering fashion.’ b. òjè ò ó bì-bí. Oje sc c shift-shift ‘Oje is staggering, moving in a zig-zag fashion.’ The constituent bibi is derived via reduplication from bi ‘shift something aside or to another position’. As a simple predicate, bi collocates with the body-part term égbè ‘body’ as direct object to convey the sense ‘shift aside, shift to one side’. (77) a. òjè bí égbè. Oje prp.shift body ‘Oje has shifted aside / shifted his body aside.’ b. òjè bí égbè lí òhí. Oje prp.shift body app Ohi ‘Oje has shifted aside for Ohi / evaded Ohi.’ Taking a direct object that is an inanimate, non-body part term, bi occurs only in a complex predicate with the change of locative state verb ye ‘move toward’. (78) òjè bí ólí úkpùn yé èfòkpá. Oje prp.shift the cloth move.toward one.side ‘Oje has shifted the cloth toward one side.’ These rate and manner expressions, fuen re, soso and bibi, seem to be in the early stages of grammaticalization. For each, a shift of subcategorization properties is becoming evident through their positioning in the word order of a clause, restrictions of verb direct object to a body-part term, and the relatively rigid syntactic properties of this term.

7 Conclusion In this paper, we have attempted to align grammatical functions in the Edoid language Emai with their etymological sources. Several grammaticalization channels have become evident, although we have discussed them only with respect to their

764

Ronald P. Schaefer and Francis O. Egbokhare

source and target outcome. Nonetheless, one can begin to appreciate how grammaticalization has impacted the structural character of Emai, particularly phrases for its nouns, verbs, postverbal particles and preverbs. Two features of our analysis are perhaps noteworthy and merit closer scrutiny in future investigations of grammaticalization. One concerns the frequency of particular phrase types as source for particular grammaticalization outcomes. In this regard, verb-noun collocations were frequent as the source for preverbs. Also, we saw that multi-verb predicates served as the source for ongoing evolution of some postverbal particles. A second feature concerns the frequency of body-part terms, especially as they occur in idiomatic expressions, as source for the evolution of preverbs. It is not occurrence of these features that is particularly surprising. Body part terms, for instance, frequently appear as source for various grammatical functions (Heine and Kuteva 2002). Rather, it is a question of the extent to which a given source type dominates in a specific grammaticalization pathway. Perhaps it is time to ask if there are pathways that fail to rely on body-part terms for example. Or perhaps we should ask if there are pathways where verb-noun collocations, instead of simply verbs, are the sole or predominant source, as was the case for preverbs for example. As for the frequency of body-part terms as source, there are several intriguing questions that come to mind. We know from this study that body-part terms were active as source for outcomes affecting noun phrases, postverbal particles and preverbs, though not verbs. In response to this observation, one could simply say that body parts are time stable nouns, while verbs tend to be far less time stable. On the other hand, it might be that body-part nouns constitute a part of speech unattractive to verb formation for other reasons. Relative to body-part terms themselves, it might prove revealing to determine whether some never serve as source for any grammaticalization pathway. A final observation concerns the role of various lexical properties in the grammaticalization process. These would include category type, subcategory delineations like transitivity, and collocational restrictions such as the shift from place to human nouns for direct object. One has the sense that these property types are not equally exploited across the various pathways to which they might apply. If this is so, it would be useful to know the relative standing of these (and other) properties, one to another, in the grammaticalization process affecting particular pathways and pathways in general. Along these same lines, our analysis revealed the frequent impact of grammaticalization on constituents that could express event fulfillment of a lexical implication. Further investigation of this issue may affect our understanding of event coding not only in Emai but also in Niger Congo, where coding by multiple verbs within a basic clause has plagued our understanding of word order types (Gensler 1997). In Table 1, we identify Emai functions reflecting grammaticalization and their etymological sources.

Grammaticalization in Emai

765

Tab. 1: Grammaticalization channel sources by phrase type (noun phrase, verb phrase, postverbal particle and preverb phrase) aligned with their respective outcome functions. Noun phrase body-part terms ékéìn, ùòkhò, égbè, ídámà, àgbàn, ègèìn,

>

relational nominal for spatial orientation ‘inside’, ‘behind’, ‘beside’, ‘middle of’, ‘edge of’, ‘under’

égbè ‘body’ égbè ‘body’ khi copula

> > >

reciprocal reflexive self correlative conjunction ‘both … and’

landscape terms òkhùnmì, òtòì

>

relational nominal for spatial orientation ‘top of’, ‘bottom of’

òkpá ‘one’

>

exhaustive, exclusive ‘only, alone’

> > > > > > > > > >

affixed attachment ‘to’ momentary impact ‘against’ oblique contact ‘against’ suspended attachment ‘onto’ restrained contact ‘against, to’ receiver for bidirectional messaging ‘after’ addressee, nonverbal transfer ‘to’ receiver for unidirectional messaging ‘to’ allative ‘to’ inclusive attachment ‘in, into’

> > > > > >

deactivate event fulfillment ‘in vain’ cancel event occurrence ‘try, feign’ excessive quantitative achievement ‘too much’ venitive sufficient quantitative achievement ‘enough’ terminal achievent ‘end’

>

‘deliberately’

>

emphatic subject reflexive

>

‘defiantly’

>

‘carefully’

>

‘mistakenly’

>

attenuate temporal rate, ‘slowly’

Verb phrase baa ‘join, add to’ gbe ‘hit’ gon ‘bend, bowed’ khuan ‘suspend’ nye ‘strain, restrain’ ree ‘take after’ vbiee ‘be apparent’ ye ‘move toward’ ye ‘move toward’ ze ‘include within Postverbal particle phrase ba kun ‘stalk in vain’ fee ghoo ‘inspect, taste’ gbe ‘accumulate’ re ‘rise, arrive’ se ‘amount to’ so ‘join, collide Preverb phrase daa ábò ‘prepare self’ lit. ‘prop/raise hands’ de óbò ‘strengthen self’ ‘fortify hand’ de ùdà ‘become defiant’ lit. ‘reach defiance’ dee égbè ‘duck down’ lit. ‘lower body’ dee óbò ‘committ mistake’ lit. ‘lower hand’ fuen égbè re ‘slow down’

766

Ronald P. Schaefer and Francis O. Egbokhare

Tab. 1 (continued) Preverb phrase gba ‘tie’ gbe ùdù ‘have courage lit. ‘reposition heart’ kaka égbè ‘persevere’ lit ‘stiffen body’ moe étìn ‘be strong’ lit. ‘have breadth’ toto óbò ‘become strong’ lit. ‘strengthen hand’ ze émì ‘choose things’ ze ‘choose, select’

> >

collective ‘together’ ‘courageously’

>

‘perseveringly’

>

‘able to’

>

‘intensely’

> >

absolute quantitative intensification, ‘a lot’ non-absolute intensification, ‘not quite, a bit’

Acknowledgements Data incorporated in this paper derive from research sponsored by the U. S. National Science Foundation, (BNS #9011338 and SBR #9409552), the U. S. Department of State (College and University Affiliations Program grant ASJY 1333), and the U. S. National Endowment for the Humanities (PD-50004-06). We also benefited from support extended by Southern Illinois University Edwardsville and the University of Ibadan, which allowed for discussion of many structures considered here. We thank these institutions for their support, while not extending to them any responsibility for data interpretation.

Abbreviations Abbreviations for grammatical morphemes used throughout this paper include: aa = affixed attachment, absi = absolute intensification, all = allative, ant = anterior, anti = anticipative, app = applicative, ass = associative, c = continuous, cer = certaintive, cl = change of location, col = collective, con = conative, cop = copula, crc = correlative conjunction, cs = change of state, ded = deductive, dmd = distal manner deictic, ds = distributive, eval = evaluative, ea = excessive achievement, f = factative, h = habitual, hor = hortative, hyp = hypothetical, id = identity pronoun, ia = inclusive attachment, loc = locative, man = manner, mi = momentary impact, mr = message receiver, nabi = non-absolute intensification, naf = non-achieved fulfillment, nai = nonachieved intention, oc = oblique contact, pap = past perfective, pf = positive focus, prp = present perfective, pred = predictive, rc = restrained contact, rec = reciprocal, refl = subject reflexive, rep = repetitive, sa = suspended attachment, sc = subject concord, self = object reflexive, ta = terminal achievement, temp = temporal perspective, and ven = venitive.

Grammaticalization in Emai

767

References Elugbe, Ben. 1989. Comparative Edoid: Phonology and lexicon. Port Harcourt: University of Port Harcourt Press. Gensler, Orin. 1997. Grammaticalization typology and Niger Elugbe Congo word order. Progress on a still-unsolved problem. Journal of African Languages and Linguistics 18(1). 57–93. Heine, Bernd, Ulrike Claudi & Friederike Hunnemeyer (eds.). 1991. Grammaticalization: A conceptual framework. Chicago: The University of Chicago Press. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. New York: Cambridge Univeristy Press. Heine, Bernd & Kyung-an Song. 2010. On the genesis of personal pronouns: Some conceptual sources. Language and Cognition 2(1). 117–147. Kuteva, Tania. 2001. auxiliation: An inquiry into the nature of grammaticalization. New York: Oxford University Press. Schaefer, Ronald P. & Francis O. Egbokhare. 1999. Oral tradition narratives of the Emai people, Part I and Part II. Hamburg: LIT Verlag. Schaefer, Ronald P. & Francis O. Egbokhare. 2000. Emai preverb order. In Ekkehard Wolff & Orin Gensler (eds.), Proceedings from the 2nd world congress of African linguistics, 733–746. Köln: Rüdiger Köppe Verlag. Schaefer, Ronald P. & Francis O. Egbokhare. 2006. Toward a typological perspective for Emai’s BE constructions. In F. K. Erhard Voeltz (ed.), Studies in African linguistic typology, 377–396. Philadelphia: John Benjamins. Schaefer, Ronald P. & Francis O. Egbokhare. 2007. A dictionary of Emai: An Edoid language of Nigeria. Köln: Rüdiger Köppe Verlag. Schaefer, Ronald P. & Francis O. Egbokhare. 2010. On Emai ditransitive constructions. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), On the typology of ditransitive constructions. (Mouton Grammar Library Series), 1–30. Berlin: Mouton de Gruyter. Schaefer, Ronald P. & Francis O. Egbokhare. 2015. Emai valency classes and their alternations. In Andrej Malchukov & Bernard Comrie (eds.), Toward a typology of valency classes. 261–298. Mouton Grammar Library Series. Berlin: Mouton de Gruyter. Schaefer, Ronald P. & Francis O. Egbokhare. 2017. A grammar of Emai. (Mouton Grammar Library Series). Berlin: Mouton de Gruyter. Williamson, Kay & Roger Blench. 2000. Niger Congo. In Bernd Heine & Derek Nurse (eds.), African languages: An introduction. 11–42. New York: Cambridge University Press.

Denis Creissels

18 Grammaticalization in Tswana 1 Introduction 1.1 The aim of the present paper In the present paper, I review grammaticalization processes that can be reconstructed in the history of Tswana on the basis of internal reconstruction and/or comparative data (including regular relationships between Tswana forms and Proto-Bantu reconstructions), and also phenomena observable in present-day Tswana that can be analyzed as grammaticalization in progress.

1.2 Tswana: location, and genetic affiliation Tswana (aka Setswana), S31 in Guthrie’s nomenclature of Bantu languages, is a Southern Bantu language with approximately 6 million speakers. In Botswana, ethnic Batswana constitute 80 % of the population, estimated at 2.3 million in 2019. In South Africa, Tswana is dominant in the Northwest Province and in some districts of the Free State Province, and the number of its speakers is estimated at 4 million. The closest relatives of Tswana are Pedi and Southern Sotho. These three languages, with a total number of at least 16 million speakers, are so close to each other that, from a strictly linguistic point of view, they should be considered as three varieties of a single language. Pedi is commonly designated as Northern Sotho, but this term is ambiguous, since it is also used with reference to lects (Lobedu, Tswapong, etc.) that, linguistically, are better considered languages distinct from SothoTswana proper.

1.3 The basics of Tswana morphosyntax Typologically, Tswana is in almost every respect a typical Bantu language.1 Its locative system, very different from that found in Central Bantu languages, is the main exception to this generalization, but this is an areal feature, shared by the other Southern Bantu languages.2

1 On the typological profile of Bantu languages, see Creissels (Forthcoming). 2 For an overall presentation of Tswana, see Cole (1955), Creissels (2003). On Southern Sotho, see Doke and Mofokeng (1957). On the changes in the organization of the locative system that affected Southern Bantu languages, see Marten (2010), Creissels (2011). https://doi.org/10.1515/9783110712735-004

770

Denis Creissels

Nouns are characterized by a gender system in which gender markers and number markers are completely fused and show no evidence of ever having been dissociated. This system is manifested in noun prefixes, obligatory head-dependent agreement in the noun phrase, and obligatory agreement of free pronouns and bound pronominal morphemes (subject indexes, object indexes) with the noun they represent. Noun forms divide into 12 agreement classes, and genders can be defined as sets of nominal lexemes that have the same agreement behavior both in the singular and the plural. For example, the singular form mosadi [mʊ̀ sádí] ‘woman’ belongs to the agreement class traditionally labeled class 1, the corresponding plural basadi [bàsádí] ‘women’ belongs to class 2, and consequently mosadi [mʊ̀ sádí] as a lexeme belongs to gender 1–2.3 Similarly, the singular form motse [mʊ̀ -tsɩ̀] ‘village’ belongs to class 3 (in which nouns have the same prefix as in class 1, but distinct agreement properties), the corresponding plural form metse [mɩ̀-tsɩ̀] belongs to class 4, and consequently motse [mʊ̀ -tsɩ̀] as a lexeme belongs to gender 3–4. The major genders in Tswana are 1–2 (mosadi [mʊ̀ sádí] ‘woman’ pl. basadi [bàsádí]), 3–4 (motse [mʊ̀ -tsɩ̀] ‘village’ pl. metse [mɩ̀-tsɩ̀]), 5–6 (lee [lɩ̀-ɩ ́] ‘egg’ pl. mae [mà-ɩ ́]), 7–8/10 (selepe [sɩ̀-lέpέ] ‘axe’ pl. dilepe [dì-lέpέ]),4 9–8/10 (podi [pʊ́ dí] ‘goat’ pl. dipodi [dì-pʊ́ dí]), 11–6 (losea [lʊ̀ -sɩ ́á] ‘baby’ pl. masea [mà-sɩ ́á]), 11–8/10 (loso [lʊ̀ -sɔ̀] ‘spoon’ pl. dintsho [dì-ǹts hɔ̀]), and 14–6 (bothata [bʊ̀ -t hátá] ‘problem’ pl. mathata [mà-t hátá]). Noun phrases are head-initial. Verbs show a rich system of morphological variations, including TAM markers, negation markers, obligatory subject indexes, optional object indexes, valency operators (reflexive, reciprocal, decausative, causative, passive, and applicative), and conjoint/disjoint markers5 – see Creissels (2006). The tonal morphology of the verb is particularly complex, even by Bantu standards – see Creissels (1999), Creissels, Chebanne, and Nkhwa (1997), Creissels (2017a). 3 In this article Tswana forms are systematically quoted both according to current Tswana orthography (in italics), and in broad phonetic transcription (within brackets). The reason is that current orthography may be quite misleading in a linguistic analysis, since it distinguishes only 5 vowels and does not indicate tones at all, whereas Tswana has 9 vowel phonemes, and tones are crucial for morphological analyses. Moreover, many morphemes that are unquestionably prefixes (in particular, subject indexes and object indexes) are written as if they were separate words. The correct word division is given in the phonetic transcription. 4 Tswana has conflated the reflexes of the Proto-Bantu classes 8 and 10. 5 In the context of Bantu studies, a conjoint verb form is a verb form that cannot be found in sentence-final position and cannot be separated from the following phrase by a pause. A disjoint verb form does not have this limitation, but is not excluded from non-final contexts either, and when in non-final sentence position, is not necessarily followed by a perceptible pause. There is some cross-linguistic variation in the function of the conjoint/disjoint contrast – Van der Wal and Hyman (2017). In Tswana, it marks the distinction between phrases in post-verbal position that form part of the verb phrase and contribute to the comment expressed by the verb, and phrases in post-verbal position that fulfill the discourse function of afterthought (alias antitopic) – Creissels (2017a).

Grammaticalization in Tswana

771

Basic verbal clauses have a rigid constituent order subject – verb – objects – obliques. Alignment is consistently nominative/accusative: there is no exception to the rule according to which the coding frames through which verbs express their argument structure must include a term whose coding properties are identical to those of the agent of prototypical transitive verbs. There is no case contrast, but the distinction between subjects, objects, and obliques, is manifested in indexation: with only the exception of the infinitive and the imperative, verb forms include an obligatory subject index, even in the presence of a subject noun phrase; objects can be indexed too, but object indexes occupy a distinct position in the verbal template, and occur only if the clause includes no noun phrase representing the same argument. As illustrated in (1), if a co-referent NP is present, the subject index expresses class agreement with it. In the absence of a co-referent NP, subject indexes that do not belong to 1st or 2nd person are interpreted anaphorically, triggering the identification of the argument they represent with a contextually salient referent compatible with the class expressed by the subject index. (1)

a. Ngwana o thubile mae. màː-ɩ ́.] [ŋʷ-àná ʊ́ -t hùbílé cl1-child s.cl1-break:prf:cj cl6-egg ‘The child broke the eggs.’ b. Ngwana o tsile. [ŋʷ-àná !ʊ́ -tsîːlè.] cl1-child s.cl1-come:prf:dj ‘The child came.’ c. O thubile mae. màː-ɩ ́.] [ʊ́ -t hùbílé s.cl1-break:prf:cj cl6-egg ‘He/She broke the eggs.’ d. O tsile. [ʊ́ -tsîːlè.] s.cl1-come:prf:dj ‘He/she came.’ e. *Ngwana thubile mae. f. *Ngwana tsile.

As illustrated by examples (1a) and (1c), the object of transitive verbs is not obligatorily indexed on the verb form, but topical objects whose precise description is considered superfluous by the speaker are represented by object indexes prefixed to

772

Denis Creissels

verbs. Object indexes immediately precede the verb stem and may be separated from subject indexes by TAM or negation markers as in Ex. (2a–b). (2)

a. Ngwana o a thubile. [ŋʷ-àná ʊ́ -à-t húbîːlè.] cl1-child s.cl1-o.cl6-break:prf:dj ‘The child broke them (the eggs).’ b. Ngwana o tlaa a thuba. [ŋʷ-àná !ʊ́ -t ɬáà-á-t hûːbà.] cl1-child s.cl1-fut-o.cl6-break:dj ‘The child will break them (the eggs).’

Tswana has very productive multiple-object constructions: non-derived verbs may have two non-coordinated objects, and valency-increasing derivations (causative and applicative) may result in constructions with three objects – Ex. (3). In multipleobject constructions, the syntactic differences between the objects are minimal: each object can be converted into the subject of a passive construction, or represented by an object index. The linear order of the objects is rigid, and depends crucially on Animacy Hierarchy. Distinctions in the semantic role of the objects (for example: beneficiaries precede recipients) are taken into account only in the case of objects that do not differ in animacy.6 (3)

a. Ngwana o nole maši. mâː-ʃì.] [ŋʷ-àná !ʊ́ -nʊ́ lé cl1-child s.cl1-drink:prf:cj cl6-milk ‘The child drank milk.’ b. Ke nositse ngwana maši. [kɩ̀-nʊ́ sítsé ŋʷ-àná mâː-ʃì.] s.1sg-drink:caus:prf:cj cl1-child cl6-milk ‘I made the child drink milk.’ c. Ke noseditse Dimpho ngwana maši. ŋʷ-àná mâː-ʃì.] [kɩ̀-nʊ́ sédítsé dím̀ p hɔ́ s.1sg-drink:caus:appl:prf:cj (cl1)Dimpho cl1-child cl6-milk ‘I made the child drink milk in Dimpho’s place.’

6 This may lead to ambiguities. For example, bolaisa motho ntša [bʊ́ láísà mʊ̀ t hʊ̀ ɲ̀tʃá] (kill.CAUS CL1.man CL9.dog) has two possible readings: ‘make the man kill the dog’ or ‘make the dog kill the man’; since ‘man’ has a higher degree of animacy than ‘dog’, in a multiple-object construction, ‘man’ must precede ‘dog’, irrespective of their semantic role.

Grammaticalization in Tswana

773

As in other Southern Bantu languages (and in contrast to the situation observed in Central Bantu languages), locative phrases do not have access to the subject function. They have the internal structure of noun phrases, from which they differ only in that (a) they are headed by a locative, i.e., a nominal form to which the locative prefix go [χʊ́ -] or the locative suffix -ng [-ŋ̀] has been added, and (b) they optionally combine with one of the three locative prepositions ko [kó] (relative remoteness), fa [fá] (relative proximity), or mo [mó] (interiority, contact). The choice between the locative prefix go [χʊ́ -] and the locative suffix -ng [-ŋ̀] is entirely determined by the grammatical nature of the head of the locative-marked noun phrase. Toponyms and a handful of common nouns are used as locatives without locative marking. Neither locative affixes nor locative prepositions specify the distinction between static location, source of movement, or direction of movement. Tswana has a limited inventory of simple non-locative prepositions or prepositional clitics: le [lɩ ́-] (comitative, also used for additive coordination of NPs, and as an additive particle corresponding to English ‘also’, ‘too’, or even’), ka [ká] (instrumental, also used for manner and time adjuncts), and ke [kɩ ́] (used almost exclusively for the agent phrase in passive constructions).

2 Grammaticalization of nominal categories 2.1 Class/gender Bantu ‘noun class’ systems are reflexes of a gender system that grammaticalized very early in the history of Niger-Congo, and nothing concrete can be said about their ultimate origin. Apart from the modifications that affected the relationship between noun classes and locative marking (see Creissels 2011), the Tswana system of noun classes differs from that of Proto-Bantu in minor details only: some classes (12 and 13) have been lost, others (8 and 10) have merged, but the only change that can be analyzed in terms of grammaticalization is the creation of a plural prefix for the nouns of gender 1/2 (human gender) that have a zero prefix in the singular – see 2.2. An interesting detail is however that, in Tswana, the Proto-Bantu class 12 (singular of diminutives) has disappeared as such, but the nominal prefix of this class (ga- [χà-] < Proto-Bantu *kà-) has subsisted as a derivational prefix expressing the number of times an action is carried out, as in gabedi [χà-bèdí] ‘twice’ < -bedi [-bèdí] ‘two’.

2.2 Number As usual in Bantu languages, the class prefixes of nouns that express plural are obvious reflexes of reconstructed Proto-Bantu class prefixes, with the only excep-

774

Denis Creissels

tion being the so-called class 2a prefix, i.e., the plural prefix for those of the nouns belonging to gender 1/2 (human gender) that have a zero prefix in the singular. It is widely acknowledged that it is not possible to reconstruct a class 2a marker in Proto-Bantu, and the question of the possible origins of the class 2a markers found across Bantu languages remains an open question. Van de Velde (2006) tackles this question with reference to Eton (A71). The analysis proposed in Creissels (2016) for Tswana is different, but does not contradict Van de Velde’s analysis, since there are very important differences between the Eton and Tswana data. More generally, it seems reasonable to posit that the class 2a prefixes found across the Bantu family do not result from a grammaticalization path shared by all Bantu languages, but rather from different grammaticalization paths whose only common point is that their input included some marker or pronoun belonging to class 2 (human plural). The grammaticalization path proposed for the class 2a prefix of Tswana (bo[bó-]) is based on the observation that this prefix can combine with proper names as an associative plural marker (as in boKitso [bó-kítsɔ́] ‘Kitso and associates’). With kinship terms, bo [bó-] is ambiguous between an associative plural and a plain plural reading (as in bomalome [bó-màlʊ́ mέ] ‘my uncles’, or ‘my uncle and associates’), whereas it is unambiguously a plain plural marker with the other semantic types of nouns found in gender 1/2 that have a zero prefix in the singular. This supports the hypothesis that bo- [bó-] was originally an associative plural marker whose use as a plain plural marker was extended to other nouns having no overt prefix in the singular. Moreover, Creissels (2016) discusses evidence that the ultimate origin of this associative plural marker is a combination of two genitival prefixes, ba [bá-] (genitival prefix of the human plural class: ‘the people of …’) plus ga [χá-] (genitival prefix of locative class: ‘the place of …’). According to this hypothesis, the plural marker bo- [bó-] results from the contraction of ba ga [bá-χà-] ‘the people at X’s place’, reanalyzed as ‘X and associates’. Note that this phenomenon is related to that evoked in Section 2.7 below.

2.3 Definiteness In many Bantu languages, the gender-number prefix of nouns is preceded by a formative (often reduced to a floating H tone) referred to as the augment, or preprefix. It is commonly assumed that definiteness or referentiality marking was the original function of the augment, which however has been reanalyzed as a case marker in the Bantu languages spoken in a compact geographical area from Gabon to Angola – see Blanchon (1999), Schadeberg (1986). In Tswana, the augment has simply been lost, and no grammaticalized expression of referentiality / definiteness has been created. In Tswana, exactly as in Latin or Russian, bare nouns are freely used in contexts in which other languages would require definiteness marking.

Grammaticalization in Tswana

775

2.4 Case/adpositions Tswana does not have case in the sense of a morphological category of nouns whose functions include the marking of distinctions between core arguments.

2.4.1 Locative marking As usual among Southern Bantu languages, Tswana has lost the system of locative marking by adding locative class prefixes to noun forms already including their inherent prefixes. This loss has been compensated by the creation of two locative affixes in complementary distribution that, synchronically, are not related to the noun class system. One of them, found in Tswana as -ng [-ŋ̀], suffixes to the first word of locative expressions. It is commonly admitted that it is etymologically related to Proto-Bantu *-įni ‘liver’ – see Samsom and Schadeberg (1994), Güldemann (1999). The other one, found in Tswana as go [χʊ́ -], prefixes to the first word of locative expressions. It is analyzed by Grégoire (1975) as originating from a nonfinite form of the copular verb including an index of locative class (something like ‘there-being’). Tswana also uses the genitival linker of locative class ga [χá-] ‘(place) of ’ as a locative marker introducing locative arguments or adjuncts with the meaning ‘at/ to/from X’s place’ (X a human noun) – Ex. (4). (4)

Ke tlaa ya ga Mpho kamoso. [kɩ̀-t ɬàà-jà χá-m̀ p hɔ́ kámʊ̀ ːsɔ́.] s.1sg-fut-go:cj cl17:gen-(cl1)Mpho tomorrow ‘I will go to Mpho’s place tomorrow.’

In addition to these locative affixes, in Tswana (but not in the other Southern Bantu languages, not even in the other Sotho-Tswana varieties), the demonstratives of locative classes, still used in Tswana as deictic locative adverbs (fa [fá] ‘here’, ko [kó] ‘there’, and mo [mó] ‘therein’), also have grammaticalized as locative prepositions – see Creissels (1997, 2011). As indicated in Ex. (5), the grammaticalization of these prepositions is not complete in the sense that, in their locative uses, they can always be dispensed with. However, they are rarely omitted in spontaneous texts. (5)

a. Ke isitse ngwana (ko) ngakeng. [kɩ̀-ìsítsé ŋw-àná (!kó) ŋákèː-ŋ̀] s.1sg-go:caus:prf:cj cl1-child loc (cl9.)doctor-loc ‘I have sent the child to the doctor’s.’ b. Ke tswa (ko) go malome. [kɩ̀-tswà (kó) χʊ́ -màlʊ̂ :mὲ.] s.1sg-come_from:prs:cj loc loc-(cl1.)uncle:1sg ‘I am coming from my uncle’s.’

776

Denis Creissels

Moreover, as illustrated by Ex. (6), one of these three recently created locative prepositions (mo [mó] ‘in’), has already acquired non-locative uses – see Creissels (2013). In its non-locative uses, mo [mó] cannot be omitted. (6)

a. Ga ke reke dikgomo mo legodung. [χà-kɩ ́-rékɩ ́ dí-q hòmʊ́ !mó lɩ ́-χòdùː-ŋ̀.] neg-s.1sg-buy:cj cl8/10-cow loc cl5-thief-loc ‘I don’t buy cows from a thief.’ b. Re aga maraka go sireletsa dikgomo mo dibataneng. [rɩ̀-áχá má-ràká χʊ̀ -sírélétsà dì-q hòmʊ́ !mó s.1pl-build:prs:cj cl6-kraal inf-protect cl8/10-cow loc dí-bàtànèː-ŋ̀.] cl8/10-beast_of_prey-loc ‘We build kraals to protect the cows from the beasts of prey.’

Tswana also has several words that were originally nouns belonging to one of the three Proto-Bantu locative classes 16–17–18 (such as morago [mʊ̀ ráχʊ́ ] ‘behind’, originally a class 18 noun, or godimo [χʊ̀ dímʊ́ ] ‘above’, originally a class 17 noun) that have completely lost the ability to be used as nouns, but are still used as adverbs or prepositions. Their first syllable was originally a class prefix, but it does not act as such anymore. In the prepositional use of these former locative class nouns, their complement is marked by ga [χá-], etymologically the genitival linker of class 17, even for those that originally belonged to another locative class: morago ga [mʊ̀ ráχʊ̀ χá-] ‘behind, after’, godimo ga [χʊ̀ -dímʊ̀ χá-] ‘on top of ’. In Tswana, the corresponding nouns have been re-created by adding prefixes of non-locative classes to a stem which, by itself, is of adverbial nature: bo-morago [bʊ̀ -mʊ̀ ráχʊ́ ] (cl. 14) ‘back’, legodimo [lɩ̀-χʊ̀ dímʊ́ ] (cl. 5) ‘top’. Among former locative class nouns used as prepositions, the development of non-locative uses is attested with ntle [ǹt ɬέ] ‘exterior’ > ntle ga / le [ǹt ɬέ χá- / lɩ ́-] ‘without’ – Ex. (7). (7)

Ke tlaa dira jang ntle le wena? [kɩ̀-t ɬàà-dìrà jáŋ́ ǹt ɬέ lɩ ́-wὲːná?] s.1sg-fut-do:cj how outside with-2sg.pro ‘How will I do without you?’

2.4.2 Non-locative prepositions Tswana has only vestiges of the Bantu comitative preposition na, and productively uses a comitative proclitic le [lɩ ́-]. I am aware of no evidence of a possible etymological analysis of this comitative marker.

Grammaticalization in Tswana

777

Tswana also has an instrumental preposition ka [ká], which however also has more or less marginal uses that can hardly be explained as extensions of its instrumental meaning. Its use to introduce the complement of kana [kàná] ‘of the same size (as)’, illustrated by Ex. (8), suggests a similative meaning (‘like’) which is not productive anymore, but can be viewed as evidence of derivation from the ProtoBantu verb root *-ngà ‘be like’, since in Tswana, k is the regular reflex of ProtoBantu *ng. The details of a hypothetical grammaticalization path similative > instrumental are, however, unclear. (8)

Tshephe e kana ka podi e tona. ! [ts hέp hέ ɩ ́-kàná ká pʊ́ d ì é (cl9.)springbok s.cl9-of_the_same_size as (cl9.)goat:cstr cl9:lk ! tʊ̀ ːnà.] (cl9.)big ‘The springbok is of the same size as a big goat.’

The third simple non-locative preposition, ke [kɩ ́] (used almost exclusively with agent phrases in passive constructions, as in (9)) is etymologically transparent, since ke [kɩ ́] is also an identificational copula (i.e., the translation equivalent of English ‘this is’), and the grammaticalization of identificational copulas into prepositions marking agent phrases in passive constructions is well-attested across Bantu.7 Negative passive clauses, in which the ke-phrase falls under the scope of negation morphologically expressed within the passive verb form, provide decisive evidence that the originally bi-clausal construction has been reanalyzed as a monoclausal construction in which the former identificational clause acts as an adjunct modifying the passive verb form – Ex. (9b). The reanalysis is confirmed by the fact that passive verb forms immediately followed by a ke-phrase must be in the conjoint form, since the disjoint form would be required if there were still a clause boundary between the verb form and ke [kɩ ́]. (9)

a. Lekau le beditswe ke mapodisi [lɩ̀-káú lɩ ́-bɩ̀dítswè kɩ ́ má-pòdîːsì.] cl5-boy s.cl5-hit:pass:prf:cj by cl6-policeman ‘The boy was hit by the policemen.’ lit. ‘The boy was hit, it is the policemen.’

7 Across Bantu, agent phrases in passive constructions may also show comitative marking, instrumental marking, locative marking, or no marking at all. The grammaticalization of an identificational copula as a preposition marking agent phrases is typically found among Southern Bantu languages (Fleisch 2005).

778

Denis Creissels

b. Lekau ga le a bediwa ke mapodisi. ! [lɩ̀-káú χà-lɩ ́-á-bɩ̀díwá kɩ ́ má-pòdîːsì.] cl5-boy neg-s.cl5-prf:neg-hit:pass:cj by cl6-policeman ‘The boy was not hit by the policemen.’ Ex. (10) shows that ke [kɩ ́] ‘it is’ reanalyzed as a preposition marking the agent phrase in passive constructions can be found with a similar meaning in non-passive constructions. (10) Ba sule ke letsapa. ! [bá-sùlé kɩ ́ lɩ ́-tsàːpà.] s.cl2-die:prf:cj from cl5-tiredness ‘They are exhausted.’ lit. ‘They died, it is tiredness.’

2.4.3 Preposition-like uses of words belonging to other categories (fa) e sa le [(fá) ɩ ́-sà-lɩ ́] ‘since’ is etymogically the persistive form of le [lɩ̀] ‘be’ with an expletive subject of class 9, optionally preceded by fa [fá] ‘if ’ (‘(if) it is still’). (11) Fa e sa le phuthego ele ga a robala sentle. [fá ɩ ́-sà-lɩ ́ p hút hὲχɔ̀ é-lé χà-á-rɔ̀bálɩ ́ if s.cl9-pers-be (cl9.)meeting cl9:dem-dist neg-s.cl1-sleep:prs:cj sɩ ́-ǹːt ɬὲ.] cl7-good ‘Since that meeting he does not sleep properly.’ fa e le [fá !ɩ ́-lɩ ́] ‘regarding’, lit. ‘if it is’, is another preposition-like expression involving a regular form of the copular verb le [lɩ̀] ‘be’ preceded by fa [fá] ‘if ’. (12) Fa e le nna, ga ke itse sepe. ! [fá !ɩ ́-lɩ ́ ǹná, χà-kɩ ́-ítsɩ ́ sɩ ̂ː-pὲ.] if s.cl9-be 1sg:pro neg-s.1sg-know:prs:cj cl7-nothing ‘Personally (lit. if it’s me), I don’t know anything.’ Similarly, fa e se [fá !ɩ ́-sɩ ́] ‘except’, lit. ‘if it is not’, derives etymologically from the negative copula se [sɩ̀] ‘not to be’. (13) Basimane botlhe ba na le baesekele, fa e se nna. [bà-símànɩ ́ !b-ót ɬ hé bá-nà lɩ ́-báɩ̀sɩ ́kɩ̀ːlɩ ́, fá !ɩ ́-sɩ ́ ǹːná.] cl2-boy cl2-all s.cl2-be with-(cl9.)bicycle if s.cl9-not_to_be pro:1sg ‘All the boys have a bicycle, except me (lit. if it’s not me).’

Grammaticalization in Tswana

779

e seng [ɩ ́-sɩ̀-ŋ́] is etymologically very similar, since it can be identified as a form of the negative copula with an expletive subject index of class 9 and the relative suffix (‘which is not’). It is also used to introduce NPs, but with a meaning of contrast, as in (14). (14) Bitsa Mpho, e seng Kitso! ! ɩ ́-sɩ̀-ŋ́ kîːtsɔ̀!] [bítsá m̀ p hɔ́, call.imp (cl1.)Mpho s.cl9-not_to_be-rel (cl1.)Kitso ‘Call Mpho, not Kitso!’ Several infinitive forms are also used as prepositions: – go bapa le [χʊ̀ -bàpà lɩ ́-] ‘be side by side with’ > ‘near’ – go feta [χʊ̀ -fɩ̀tà] ‘pass’ > ‘more than’ – Ex. (15) – go fitlhela [χʊ̀ -fɩ̀t ɬ hὲlà] ‘arrive’ > ‘until’ – Ex. (16) – go lebagana le [χʊ̀ -lɩ ́báχánà lɩ ́-] ‘be face to face with’ > ‘in front of ’ – go na le [χʊ̀ -nà lɩ ́-] ‘be with’, ‘have’ > ‘rather than’ – Ex. (17) – go simolola [χʊ̀ -símʊ́ lʊ́ là] ‘begin’ > ‘from’ – go tloga [χʊ̀ -t ɬʊ̀ χà] ‘leave’ > ‘from’ – go tswa [χʊ̀ -tswà] ‘come out’, ‘come from’ > ‘from’ – Ex. (16) – go ya ka [χʊ̀ -jà ká] ‘go with’ > ‘according to’ – Ex. (18) (15) Neo o montle go feta Dimpho. ! ʊ́ -mʊ́ -ǹt ɬὲ χʊ̀ -fɩ̀tà dìm̀ ːp hɔ́.] [nέɔ ́ (cl1.)Neo s.cl1-cl1-beautiful inf-pass (cl1.)Dimpho ‘Neo is more beautiful than Dimpho.’ (16) go tswa letsatsi leo go fitlhela gompieno χʊ́ -fìt ɬ hὲlà χʊ́ m̀píèːnʊ́ .] [χʊ̀ -tswà lɩ̀-tsàtsí !lé-ʊ́ inf-come_from cl5-day cl5.dem-dist inf-arrive today ‘since that day until today’ (17) O rata madi go na le batho. [ʊ̀ -rátá mà-dí χʊ́ -nà lɩ ́-bâː-t hʊ̀ .] s.2sg-love:prs:cj cl6-money inf-be with-cl2-person ‘You love money rather than people.’ (18) go ya ka ngwao ya rona [χʊ̀ -jà ká ŋwáɔ̀ já-rʊ̀ ná] inf-go with (cl9.)tradition cl9:gen-pro.1pl ‘according to our tradition’ Several words morphologically identifiable as locatives have non-locative prepositional uses in which their complement is introduced by le [lɩ ́-] ‘with’:

780 – –

Denis Creissels

kgatlhanong le [q hàt ɬ hànò-ŋ̀ lɩ ́-] with the locative form of kgatlhano [q hàt ɬ hànɔ̀] ‘meeting’ > ‘against’ malebang le [mà-lɩ ́bà-ŋ́ lɩ ́-] or tebang le [tɩ ́bà-ŋ́ lɩ ́-] with the locative form of a nominalization of leba [lɩ ́bá] ‘look’ > ‘about’, ‘with reference to’8

In mabapi le [mà-bàpí lɩ ́-] ‘about’, ‘with reference to’, mabapi [mà-bàpí], nominalized form of bapa [bàpà] ‘be side by side’, is devoid of locative marking. In ka ntata ya [ká ńtàtá já-] ‘because of ’, ka [ká] is the instrumental preposition, ya [já-] is the genitival linker of class 9, and ntata [ǹtàtá] is a nominal form that has no independent existence but is cognate with lata [làtà] ‘go and fetch’ and latela [làtὲlà] ‘follow’. ka ntlha ya [ká ǹt ɬ há já-] ‘on account of ’, ‘because of ’, ‘due to’ has the same structure, but ntlha [ǹt ɬ há] is attested as a noun with the meaning ‘point’ – Ex. (19). (19) Ga ba a tla ka ntlha ya pula. jà-pûːlà.] [χà-bá-à-t ɬá ká ǹt ɬ há neg-s.cl2-prf:neg-come:cj with (cl9.)point cl9:gen-(cl9.)rain ‘They didn’t come due to the rain.’ In comparative constructions, the standard of comparison can be introduced by go feta [χʊ̀ -fɩ̀tà], as in (15) above, but it can also be expressed as a locative, either alone or preceded by bogolo [bʊ̀ χʊ́ lʊ́ ] ‘especially’. Bogolo [bʊ̀ χʊ́ lʊ́ ], etymologically the class 14 form of the adjective -golo [-χʊ́ lʊ́ ] ‘big’, is used nominally with the meaning ‘bigness’, but has also developed an adverbial use in which it can be glossed ‘especially’. The literal meaning of (20) is something like ‘A horse is fast (especially) beside a donkey’. (20) Pitse e lobelo (bogolo) mo tonking. ! ɩ ́-lʊ́ bɩ ́lɔ̀ (bʊ̀ χʊ́ lʊ́ ) !mó !tóŋ́kîː-ŋ̀.] [pìtsɩ ́ (cl9.)horse s.cl9-fast especially loc donkey-loc ‘A horse is faster than a donkey.’

2.5 Construct form of nouns Construct forms of nouns (i.e., special noun forms used in combination with given types of modifiers) are not common at all in Bantu languages, but in Tswana, the … HL tone pattern shown by … HH nouns when immediately followed by certain

8 Note, however, that there is a problem with this etymology, since the locative suffix normally has a low tone, and nothing can explain its replacement by a H tone in malebang le [mà-lɩ ́bà-ŋ́ lɩ ́-] or tebang le [tɩ ́bà-ŋ́ lɩ ́-].

Grammaticalization in Tswana

781

modifiers is not phonologically conditioned, and consequently must be analyzed as a construct form marker, since in phonologically similar but syntactically different contexts, the nouns in question show their inherent … HH pattern – see Creissels (2009). For example, in (21a), Setswana [sɩ̀tswáná] ‘Tswana culture, language, etc.’ is the head of the NP Setswana se ba se buang [sɩ̀tswánà sé básɩ̀búàŋ́] ‘the Tswana they speak > the way they speak Tswana’, and consequently, the contact with the linker se [sé] introducing the relative clause triggers the use of the construct form [sɩ̀tswánà]. In (21b), Setswana [sɩ̀tswáná] is in contact with the same linker, but the linker introduces a dependent of setilo [sɩ̀tílɔ́] ‘chair’,9 not of Setswana [sɩ̀tswáná]; in (21b), Setswana [sɩ̀tswáná] has no dependent, and consequently the construct form would not be correct. (21) a. Ga ke rate Setswana se ba se buang. [χà-kɩ ́-rátɩ ́ sɩ̀-tswánà sé bá-sɩ̀-búàː-ŋ́.] neg-s.1sg-like:prs:cj cl7-Tswana:cstr cl7:lk s.cl2-o.cl7-speak:prs-rel ‘I don’t like the Tswana they speak.’ > the way they speak Tswana b. Ga ke rate setilo sa Setswana se Mpho a se rekileng. [χà-kɩ ́-rátɩ ́ sɩ̀-tílɔ̀ sá-sɩ̀-tswáná sé m̀ p hɔ́ neg-s.1sg-like:prs:cj cl7-chair cl7:gen-cl7-Tswana cl7:lk (cl1)Mpho á-sɩ̀-rékílèː-ŋ́.] s.cl1-o.cl7-buy:prf-rel ‘I don’t like the Tswana chair that Mpho bought.’ A plausible explanation of this tonally marked construct form of nouns is that it resulted from the morphologicization of tonal alternations originally attributable to tonal sandhi.

2.6 Noun-modifier linkers Noun-modifier linkers are grammatical words or clitics whose role is to enable a word/phrase to act as a noun modifier. The noun-modifier linkers found in Bantu languages also express agreement with the head noun, and are maintained in elliptical constructions in which the head-noun is omitted. Linkers are particularly common in the genitival construction: most Bantu languages (including Tswana) have a proclitic genitival linker inherited from ProtoBantu, consisting of an invariable and etymologically opaque element -a- preceded by a class agreement marker, attached to the left of NPs in genitive role. Linkers used specifically to introduce relative clauses are common too. 9 The construct form setilo [sɩ̀tílɔ̀] is licensed by the genitival dependent sa Setswana [sásɩ̀tswáná].

782

Denis Creissels

In Tswana and other southern Bantu languages, the linkers found in the noun + relative clause construction also occur in noun + attributive adjective constructions. Internal evidence and comparison with other Bantu languages show that these linkers are former demonstratives that, when inserted between a head noun and some types of modifiers, have lost their deictic value and acquired a purely syntactic function. For example, yo [jó] is both the linker of class 1 and the basic form of the demonstrative of class 1 (to which deictic particles can be added: yono [jó-nʊ́ ] (proximal), yole [jó-lé] (distal)). In mosadi yo moleele [mʊ̀ -sádì jó mʊ̀ -léèlé] ‘tall woman’, yo [jó] is a linker, semantically void but obligatorily inserted between the construct form of mosadi [mʊ̀ -sádí] ‘woman’ and the class 1 form of the adjective [-léèlé] ‘tall’; yo [jó] as a demonstrative can be added after either the head noun, or the adjective: mosadi yo(no/le) yo moleele [mʊ̀ -sádì jó(nʊ́ /lé) jó mʊ̀ -léèlé] or monna yo moleele yo(no/le) [mʊ̀ -sádì jó mʊ̀ -léèlé jó(nʊ́ /lé)] ‘this/that tall woman’. In Ex. (22a), the same linker is repeated before the second adjective ‘black’, and the relative clause. Ex. (22b) illustrates the same construction with a head noun belonging to another class. (22) a. mosadi yo moleele yo montsho yo o opelang yole [mʊ̀ -sádì jó mʊ̀ -léèlé !jó mʊ́ -ǹts hʊ̀ jó cl1-woman:cstr cl1:lk cl1-tall cl1:lk cl1-black cl1:lk ! ! ʊ́ -ɔ́pέlà-ŋ́ jó-lé] s.cl1-sing:prs-rel cl1:dem-dist ‘that tall woman with dark complexion who is singing’ b. lekau le leleele le lentsho le le opelang lele [lɩ̀-káù lé lɩ̀-léèlé !lé lɩ ́-ǹts hʊ̀ lé cl5-boy:cstr cl5:lk cl5-tall cl5:lk cl5-black cl5:lk ! ! lɩ ́-ɔ́pέlà-ŋ́ lé-lé] s.cl5-sing:prs-rel cl5:dem-dist ‘that tall boy with dark complexion who is singing’ Tswana also has a special linker, etymologically the circumstantial form of the verb ‘be’, in noun + numeral constructions – Ex. (23). (23) basadi bale babedi ! [bà-sádí bá-lɩ ́ bá-bèdí] cl2-woman s.cl2-be:circ cl2-two ‘two women’, lit. ‘women they-being two’

2.7 Genitive and possessives In the genitival construction of Tswana, a genitival linker expressing agreement with the head noun is prefixed to the genitival modifier, as ba [b-á-] in bana ba

Grammaticalization in Tswana

783

motsomi [b-àná b-á-mʊ̀ -tsʊ́ mí] (cl2-child cl2-gen-cl1-hunter) ‘the children of the hunter’. However, when the genitival modifier is a personal name or a kinship term, a variant -a ga [-áχá-] of the genitival linker must be used (as in bana ba ga Kitso [b-àná b-áχà-kítsɔ́] ‘Kitso’s children’). Synchronically, this is just a variant of the genitival linker, in complementary distribution with the plain variant -a [-á], but historically, it results from the grammaticalization of a sequence of two genitival linkers: ba [b-á-] (genitival linker of the human plural class: ‘the people of …’) plus ga [χ-á-] (genitival linker of locative class: ‘the place of …’). In other words, the etymological meaning of bana ba ga Kitso [b-àná b-áχà-kítsɔ́] ‘Kitso’s children’ is ‘the children at Kitso’s place’.

2.8 Nominal derivation In Bantu languages, verb-to-noun derivation is very productive, yielding action nouns, result nouns, agent nouns, instrument nouns, place-of-action nouns, and manner nouns, but the affixes involved in verb-to-noun derivation are etymologically opaque. Noun-to-noun derivation mainly relies on gender alternations that leave the stem unchanged. As a rule, affixal noun-to-noun derivation is relatively limited, and results from relatively recent (and still transparent) grammaticalization processes. This is in particular the case in Tswana. Tswana, like the other Southern Bantu languages, does not form diminutives by gender shift, and has a diminutive suffix cognate with the noun ‘child’ (in present-day Tswana, ngwana [ŋw-àná] pl. bana [b-àná]), as in tawana [tàw-áná] ‘lion cub’ < tau [tàú] ‘lion’.10 This diminutive suffix also combines with adjectives, but in combination with color adjectives applied to domestic animals (including cattle, sheep, goats, horses, dogs, and fowls), it has further grammaticalized into a feminine suffix. In this context, the basic form of the adjective signifies the masculine gender, whereas the diminutive form with suffix -ana [-áná] signifies the feminine. (24) a. namane e ntsho [nàmánɩ̀ é ńts hʊ̀ ] (cl9.)calf cl9:lk (cl9.)black ‘a black male calf ’ b. namane e ntshwana [nàmánɩ̀ é ǹts hw-áná] (cl9)calf cl9:lk (cl9)black-dim ‘a black female calf ’ 10 Interestingly, in other Bantu languages, the same noun grammaticalized as a diminutive prefix, which is more consistent with the typological profile of Bantu languages. On possible explanations of the creation of suffixes in Eastern and Southern Bantu, see Güldemann (1999).

784

Denis Creissels

A little productive feminine / augmentative suffix -gadi [-χádí] < Proto-Bantu *kádí ‘woman’ is also found in Tswana, as in kgosi [q hósí] ‘king, chief ’ > kgosigadi > [q hósíχádì] ‘queen’, ‘chief ’s wife’, or tau [tàú] ‘lion’ > taugadi [tàú-χádí] ‘lioness’ or ‘big lion’. Note that the augmentative use of this suffix is productive in some other Southern Bantu languages (Zulu, Southern Sotho), but very marginal in Tswana. Tswana also has a prefix ra- [rá-] (from rra [r̀rá] ‘father’) ‘owner of ’, ‘responsible for’, ‘expert on’, as in ramotse [rá-mʊ́ tsɩ ́] ‘mayor’ < motse [mʊ̀ tsɩ̀] ‘town’, ramotlakase [rá-mʊ́ t ɬàkásɩ ́] ‘electrician’ < motlakase [mʊ̀ t ɬàkásɩ ́] ‘electricity’. The status of this element as a derivational prefix is however problematic, since derivational prefixes are expected to attach to stems, whereas ra- [rá-] combines with inflected noun forms (in motse [mʊ̀ tsɩ̀] and motlakase [mʊ̀ t ɬàkásɩ ́], mo- [mʊ̀ -] is the prefix of class 3; in rameno [rá-mὲnɔ́] ‘dentist, ra- [rá-] combines with meno [mὲnɔ́] ‘teeth’, plural form of leino [lɩ̀ínɔ́] ‘tooth). The same problem arises with the similative prefix seka [sɩ̀-ká-] (where se- [sɩ̀-] is the prefix of class 7, typically associated with the meaning ‘thing’), equivalent of the English suffixes -oid or -like: in seka motho [sɩ̀-ká-mʊ́ -t hʊ́ ] ‘something resembling a human being’, seka [sɩ̀-ká-] does not combine with the stem -tho [-t hʊ̀ ] ‘human being’, but with the singular form motho [mʊ̀ -t hʊ̀ ]. What is however uncontroversial is that -ka [-ká-] can be analyzed as a reflex of the reconstructed Proto-Bantu verb root *ngà ‘be like’, since Tswana k is the regular reflex of Proto-Bantu *ng.

3 Grammaticalization of verbal categories 3.1 Verbal derivation and valency Compound verbs are marginal in Tswana, and noun-to-verb derivation (as bogale [bʊ̀ -χálɩ ́] ‘fierceness’ > galefa [χálɩ ́-f-á] ‘become angry or fierce’) is of limited productivity, whereas verb-to-verb derivation is very productive. Tswana, like most Bantu languages, has a rich inventory of ‘verbal extensions’, suffixes used to derive verbs from verbs with a variety of functions: they may increase the valency, decrease the valency, (re-)orient the action, or introduce aspectual specifications. Reduplication is also attested with an iterative meaning. The verbal extensions found in Tswana are reflexes of suffixes reconstructed at the Proto-Bantu level (with more or less obvious cognates in other branches of Niger-Congo), and their etymology is opaque, with the only exception being the associative-reciprocal suffix -an- [-an-], which has been claimed to be cognate with the comitative preposition reconstructed as *na ‘with’ (Schladt 1998). Among the evolutions that may have affected the reflexes of Proto-Bantu suffixes encoding operations on valency, the development of antipassive uses of the associative-reciprocal suffix is well-attested among Bantu languages (Bostoen, Dom, and Segerer 2015; Dom, Segerer, and Bostoen 2015), but is only marginally found in

Grammaticalization in Tswana

785

Tswana. The same can be said about the use of causative derivation to license object NPs referring to instruments: in some other Bantu languages (for example, Kinyarwanda – see Kimenyi [1980]), this use of causative derivation is fully productive, whereas in Tswana, it is only marginally attested (Creissels 2002). Some other Bantu languages attest evolutions such as reflexive > reciprocal (Bostoen 2010), reciprocal > middle, or renewal of passive morphology via reanalysis of impersonal constructions as passive (Kawasha 2007), but nothing similar can be found in Tswana. An interesting but so far unsolved question is that of the developments responsible for the variety observed in the functions of applicative derivation across Bantu. Creissels (2004) describes uses of applicative derivation in Tswana that do not fit the standard definition of applicative as encoding the addition of an object phrase representing a participant that could not be encoded as the object of the non-derived form of the same verb, and similar observations have been made on many other Bantu languages. Synchronically, the benefactive use of the Bantu applicative is particularly prominent, but a plausible hypothesis is that the original function of applicative derivation was rather the licensing of locative complements – see Jerro (2016).

3.2 The grammaticalization of TAM Bantu languages are famous for the complexity of their TAM-marking systems. It is widely acknowledged that a major source of the markers placed immediately after the subject index in the verbal template (which is the place where most TAM markers are found) is the univerbation of constructions ‘auxiliary + lexical verb’ in which the former auxiliary expressed subject agreement, and the lexical verb was in some dependent form. In Tswana, this scenario can be reconstructed for the future marker tlaa [-t ɬàà-] and the potential marker ka [-ká-]. The Tswana future results from the univerbation of a construction in which the future auxiliary was followed by the auxiliated verb in the infinitive, and the future marker tlaa [-t ɬàà-] is the reflex of a former sequence tla go [-t ɬà χʊ̀ -], where tla [-t ɬà-] is the verb ‘come’ in future auxiliary function, and go [χʊ̀ -] is the infinitive prefix. As regards the potential marker ka [-ká-], Nurse (2008) observes that modality markers that could be reflexes of a root reconstructable as *nga, variously labeled ‘potential’, ‘conditional’, or ‘concessive’, are widely attested among Bantu languages. Since k is the regular Tswana reflex of Proto-Bantu *ng, Tswana ka [-ká-] belongs to this family of modality markers, which according to Creissels (2017b) results from the grammaticalization of the Proto-Bantu reconstructed root *ngà ‘be like’ (or its reflexes at some stage between Proto-Bantu and the present-day Bantu languages). Tswana also has fully grammaticalized TAM auxiliaries, i.e., words with a morphological structure similar to that of finite verb forms, which however can only be

786

Denis Creissels

used in combination with other verbs, and whose only possible function is to add TAM specifications to the verbs with which they combine. Some of them have no obvious etymology. Those that are etymologically transparent are cognate with tla [t ɬà] ‘come’, ba [bà] ‘be’ (not attested in Tswana, but productively used in the other Sotho-Tswana varieties), or nna [ńná] ‘be’, or are reflexes of the Proto-Bantu reconstructed root *ngà ‘be like’ (and are consequently cognate with the potential marker ka [-ká-]). In addition to the fully grammaticalized TAM markers and TAM auxiliaries, Tswana also has a number of lesser-grammaticalized auxiliary verbs expressing meanings commonly taken up by adverbial expressions in European languages. The following verbs, in addition to the lexical meaning they express in combination with NPs only, are also used as TAM auxiliaries in some of their regularly inflected forms: –

aga [áχá] ‘build’ > ‘do something continually’

(25) Mhero o aga o tlhoga. [m̀ -hὲrɔ̀ ʊ́ -áχà ʊ́ -t ɬ hʊ̀ ːχà.] cl3-weed s.cl3-build:prs:cj s.cl3-grow:circ:prs ‘Weed doesn’t stop growing.’ –

atisa [àtìsà] ‘increase’ > ‘do something often, or in great quantity’

(26) Re atisa go ya kerekeng. [rɩ̀-àtìsà χʊ̀ -jà kérèkêː-ŋ̀.] s.1pl-increase:prs:cj inf-go (CL9.)church-loc ‘We often go to church.’ –

batla [bàt ɬà] ‘look for, want’ > ‘need to do something’ or ‘seem to be about to do something’

(27) Setlhare se batla go nosediwa. χʊ̀ -nʊ́ sédîːwà.] [sɩ̀-t ɬ hàrɩ̀ sɩ ́-bát ɬá cl7-tree s.cl7-want:prs:cj inf-water:pass ‘The tree needs to be watered.’ (28) Mabadi a batla a fola. á-fɔ̀ːlá.] [mà-bádí !á-bát ɬá cl6-wound s.cl6-want:prs:cj s.cl6-heal:circ:prs ‘The wounds seem to be about to heal.’ –

batlile [bàt ɬílé] perfect of [bàt ɬà] ‘look for, want’ > ‘have nearly done something’

Grammaticalization in Tswana

787

(29) Ke batlile go bolawa. [kɩ̀-bàt ɬílé χʊ̀ -bʊ́ lâːwà.] s.1sg-want:prf:cj inf-kill:pass ‘I was almost killed.’ –

boa [bʊ́ á] ‘return’ > ‘do something again’

(30) Tshimo e boile ya lengwa. ! [ts hímʊ́ ɩ ́-bʊ́ ílé já-lɩ̀ːŋwà.] ̀ (cl9.)field s.cl9-return:prf:cj s.cl9:seq-cultivate:pass ‘The field was cultivated again.’ –

dikile [díkílé] perfect of dika [díká] ‘surround’, ‘revolve’ > ‘have done something in the past year’

(31) Tshimo e dikile e lengwa. ! [ts hímʊ́ ɩ ́-díkílè ɩ ́-lɩ ̂ːŋwà.] (cl9.)field s.cl9-surround:prf:cj s.cl9-cultivate:pass:circ:prs ‘The field was cultivated last year.’ –

ga…nne [χà…ńnɩ ́] negative form of nna [ńná] ‘be’ > ‘never do something’

(32) Ga ba nne ba nthusa. [χà-bá-ńnɩ ́ bá-ǹ-t hûːsà.] neg-s.cl2-be:prs:cj s.cl2-o.1sg-help:circ:prs ‘They never help me.’ –

ga…tlhole [χá…t ɬ hɔ̀lɩ ́] negative form of tlhola [t ɬ hɔ̀là] ‘spend the day somewhere’ > ‘have ceased doing something’

(33) Ga ke tlhole ke lema tshimo e. ! kɩ ́-lɩ ́má ts hímʊ̀ ː é.] [χà-kɩ ́-t ɬ hòlɩ ́ neg-s.1sg-spend_the_day:prs:cj s.1sg-cultivate:circ:prs (cl9.)field cl9:dem ‘I do not cultivate this field anymore.’ –

letse [létsɩ ́] perfect of lala [lálá] ‘spend the night somewhere’ > ‘have done something the day before’

(34) Ke letse ke kwala ditlhatlhobo. [kɩ̀-létsɩ ́ kɩ ́-kwàlá dì-t ɬ hát ɬ hʊ̂ ːbɔ̀.] s.1sg-spend_the_night:prf:cj s.1sg-write:circ:prs cl8/10-examination ‘I took an examination yesterday.’

788 –

Denis Creissels

losetsa [lʊ̀ sètsà] ‘struggle for something’ > ‘seem to be about to do something’

(35) Pula e losetsa go na. ! ɩ ́-lʊ́ sétsá χʊ̂ ː-nà.] [púlá (cl9.)rain s.cl9-struggle_for:prs:cj inf-rain ‘It seems to be about to rain.’ –

na le [nà lɩ ́-] ‘be with’ > ‘do something usually’, or ‘be obliged to do something’

(36) Ba na le go ya ngakeng. [bá-nà lɩ ́-χʊ́ -já ŋákèː-ŋ̀.] s.cl2-be with-inf-go (cl9.)doctor-loc ‘They usually go to the doctor.’ or ‘They must go to the doctor.’ –

nama [nàmà] ‘spread, stretch’ > ‘continue to do something’

(37) O name o bereka. [ʊ́ -námὲ ʊ́ -bὲrɛ̂ːkà.] s.2sg-spread:subj s.2sg-work:circ:prs ‘You should continue to work.’ –

namile [nàmílé] perfect of nama [nàmà] ‘spread, stretch’ > ‘have eventually done something’

(38) Ke namile ke kgona ditlhatlhobo. dì-t ɬ hát ɬ hʊ̂ ːbɔ̀.] [kɩ̀-nàmílè kɩ ́-q hɔ́ná s.1sg-spread:prf:cj s.1sg-overcome:circ:prs cl8/10-test ‘I eventually passed the test.’ –

nna [ńná] ‘be’ > ‘be still doing something’

(39) Pula e nnela e na. ! ɩ ́-ńnέlà ɩ ̂ː-nà.] [púlá (cl9)rain s.cl9-be:appl:prs:cj s.cl9-rain:circ:prs ‘It doesn’t stop raining.’ –

ntse [ńtsé] perfect of nna [ńná] ‘be’ > ‘do something continually’, or ‘be still doing something’

(40) Ke ntse ke a bereka. [kɩ̀-ńtsɩ ́ kɩ ́-à-bέrɛ̂ːkà.] s.1sg-be:prf:cj s.1sg-dj-work:prs ‘I am still working.’

Grammaticalization in Tswana

–

789

ratile [rátílé] perfect of rata [rátá] ‘like, want’ > ‘have nearly done something’

(41) Ke ratile go wa. [kɩ̀-rátílé χʊ̂ ː-wà.] s.1sg-like:prf:cj inf-fall_down ‘I almost fell down.’ –

sa le [sà-lɩ ́] persistive form of le [lɩ ́] ‘be’ > ‘have done something long ago, or for the last time’

(42) Ke sa le ka mmona ka laboraro. [kɩ̀-sà-lɩ ́ ká-m̀ -mɔ́nà ká lábʊ̀ râːrʊ̀ .] s.1sg-pers-be s.1sg:seq-o.cl1-see with wednesday ‘The last time I saw him was Wednesday.’ –

sa tswa [sá-tswá] persistive form of tswa [tswá] ‘go out’, ‘come from’ > ‘have just done something’

(43) Ke sa tswa go ja. [kɩ̀-sá-tswá χʊ̀ ː-dʒá.] s.1sg-pers-come_from:cj inf-eat ‘I have just eaten.’ –

setse [sétsɩ ́] perfect of sala [sálá] ‘remain’ > ‘be already doing something’

(44) Ba setse ba bua. [bá-sétsɩ̀ bá-bùːá.] s.cl2-remain:prf:cj s.cl2-speak:circ:prs ‘They are already speaking.’ –

senkile [sɩ ́ŋ́kílé] perfect of senka [sɩ ́ŋ́ká] ‘look for’ > ‘have nearly done something’

(45) Ke senkile go swa. [kɩ̀-sɩ ́ŋ́kílé χʊ̀ ː-swá.] s.1sg-look_for:prf:cj inf-die ‘I almost died.’ –

tlaa dika [t ɬàà-díká] future of dika [díká] ‘surround’, ‘revolve’ > ‘do something during the next year’

(46) Tshimo e tlaa dika e lengwa. ! [ts hímʊ́ ɩ ́-t ɬáà-díkà ɩ ́-lɩ ̂ːŋwà.] (cl9.)field s.cl9-fut-surround:cj s.cl9-cultivate:pass:circ:prs ‘The field will be cultivated next year.’

790 –

Denis Creissels

tlaa lala [t ɬàà-lálá] future of lala [lálá] ‘spend the night somewhere’ > ‘do something before the end of the day’

(47) Pula e tlaa lala e nele. ! [púlá ɩ ́-t ɬáà-lálà ɩ ́-nɩ̀ːlé.] cl9-rain s.cl9-fut-spend_the_night:cj s.cl9-rain:circ:prf ‘It will rain before the end of the day.’ –

tlaa tsoga [tɬàà-tsʊ́ χá] future of tsoga [tsʊ́ χá] ‘get up’ > ‘do something the next day’

(48) Tshimo e tlaa tsoga e lengwa. ! ɩ ́-t ɬáà-tsʊ́ χà ɩ ́-lɩ ̂ːŋwà.] [ts hímʊ́ (cl9.)field s.cl9-fut-get_up:cj s.cl9-plough:pass:circ:prs ‘The field will be ploughed tomorrow.’ –

tlhola [t ɬ hɔ̀là] ‘spend the day somewhere’ > ‘do something continually’

(49) Ngwana yo o tlhola o ntshwenya. ! [ŋw-àná !jó ʊ́ -t ɬ hɔ́là á-ǹ-ts hwêːɲà.] cl1-child cl1:dem s.cl1-sprend_the_time:prs:cj s.cl1-o.1sg-bother:circ:prs ‘This child doesn’t stop bothering me.’ –

tloga [t ɬʊ̀ χà] ‘leave’ > ‘be close to doing something’

(50) Kgomo e tloga go tsala. χʊ̀ -tsâːlà.] [q hòmʊ́ !ɩ ́-t ɬʊ́ χá (cl9.)cow s.cl9-leave:prs:cj inf-calve ‘The cow is close to calving.’ –

tsamaela [tsàmàὲlà] applicative form of tsamaya [tsàmàjà] ‘walk’ > ‘be close to doing something’

(51) Tshipi e tsamaela go fela. ! ɩ ́-tsámáὲlà χʊ̀ -fɛ̂ːlà.] [ts hípí (cl9.)week s.cl9-walk:appl:prs:cj inf-end ‘The week is coming close to its end.’ –

tsamaya [tsàmàjà] ‘walk’ > ‘do something continually’

(52) Ngwana yo o tsamaya a bua maaka. ! ʊ́ -tsámájà á-bùá má-àːká.] [ŋw-àná !jó cl1-child cl1:dem s.cl1-sprend_the_time:prs:cj s.cl1-tell:circ:prs cl6-lie ‘This child keeps telling lies.’

Grammaticalization in Tswana

–

791

tshela [ts hɩ ́lá] ‘live’ > ‘do something continually’

(53) Konopo e e tshela o kgaogile. ! [kʊ́ nʊ́ pɔ̀ é ɩ ́-ts hɩ ́là ɩ ́-q háʊ́ χîːlè.] (cl9.)knob cl9:dem s.cl9-live:prs:cj s.cl9-break:circ:prf ‘This knob is always broken.’ –

tshogana [ts hʊ̀ χànà] reciprocal form of tshoga [ts hʊ̀ χà] ‘fear’ > ‘do something unexpectedly’

(54) Re tshoganye re utlwa lentswe la gagwe. [rɩ̀-ts hʊ̀ χáɲɩ̀ rɩ ́-ùt ɬwá lɩ ́-ǹtswɩ ́ !lá-χâːχwὲ.] s.1pl-fear:prf:cj s.1pl-hear:prs:cj cl5-voice cl5:gen-cl1:pro ‘All of a sudden, we heard his voice.’ –

tshwanetse [ts hwánétsɩ ́] perfect of tshwanela [ts hwánέlá] ‘suit’, ‘fit’ > ‘be obliged to do something’

(55) Re tshwanetse go bereka. [rɩ̀-ts hwánétsɩ ́ χʊ̀ -bέrɛ̂ːkà.] s.1pl-suit:prf:cj inf-work ‘We must work.’ –

tsile [tsìlé] perfect of tla [t ɬà] ‘come’ > ‘seem to be close to do something’

(56) Ntlo e tsile go wa. [ǹt ɬʊ̀ ɩ ́-tsìlé χʊ̂ -wà.] (cl9.)house s.cl9-come:prf:cj inf-fall ‘The house seems to be about to collapse.’ –

tsogile [tsʊ́ χílé] perfect of tsoga [tsʊ́ χá] ‘get up’ > ‘have done something in the morning’

(57) Re tsogile re ile toropong. [rɩ̀-tsʊ́ χílè rɩ ́-ìlé tʊ̀ rópôː-ŋ̀.] s.1pl-get_up:prf:cj s.1pl-go:circ:prf (cl9.)town-loc ‘We went to town this morning.’ –

ya [jà] ‘go’ > ‘be close to doing something’

(58) Ke ya go mo nyala. [kɩ̀-jà χʊ̀ -mʊ̀ -ɲâːlà.] s.1sg-go:prs:cj inf-o.cl1-marry ‘I am going to marry her.’

792

Denis Creissels

3.3 Negation Like most Bantu languages, Tswana expresses negation through verb inflection, and negation marking interferes with TAM marking in a complex way. Among the morphemes involved in negation marking, the only one etymologically transparent is se [-sɩ̀-] (used among others for the negation of the imperative), reflex of the ProtoBantu negative copula *sɩ̀. In addition to negation marking through verb inflection, gana [χáná] ‘refuse’ can be used as a negative auxiliary, as evidenced by its compatibility with inanimate subjects, i.e., with subjects that exclude interpreting gana [χáná] ‘refuse’ in its literal meaning. (59) Pula e gana go wa. ! ɩ ́-χáná χʊ̂ ː-wà.] [púlá cl9-rain s.cl9-refuse:prs:cj s.cl9-rain:circ:prf ‘It is not raining.’, ‘It is not going to rain.’

3.4 Argument indexation The argument indexation system found in Tswana is not different from that reconstructed for Proto-Bantu. In some Bantu languages, the object indexes have become obligatory agreement markers with some categories of objects (human, or definite), but this is not the case in Tswana. The object indexes of Tswana are always syntactically optional, and the choice between object indexes immediately prefixed to the verb stem and free pronouns in the same syntactic slot as objects NPs depends exclusively on information structure.

3.5 Conjoint/disjoint marking In Tswana, the distinction between conjoint and disjoint verb forms (see section 1.2) is marked tonally, with only the exception of the present positive, in which the disjoint marker a [-à-] occurs immediately after the subject index. Nothing concrete can be said about the origin of the tonal marking of the conjoint / disjoint distinction, for lack of clear comparative data, although the morphologicization of tonal alternations due to phonological phrasing and tonal sandhi is a plausible scenario. The disjoint marker a [-à-] has cognates in other Bantu languages, and the ProtoBantu reconstruction *da ‘disjunct’ was proposed by Meeussen (1967: 109). The most plausible scenario is that it developed from an auxiliary expressing verb focalization (as do in English). Güldemann (2003: 355) speculates that its ultimate origin was a verb ‘take’. In the same article, Güldemann analyzes the relationship between mark-

Grammaticalization in Tswana

793

ers of predication focus and present progressive attested in some Bantu languages (but not in Tswana).

3.6 Dependent/non-finite verb forms Among the morphemes involved in the formation of dependent or non-finite verb forms in Tswana, the suffix -ng [-ŋ́] marking relative verb forms is the only one for which an etymological analysis can be considered. In the verbal template, this suffix occupies the ‘postfinal’ slot, typically occupied across Bantu by pronominal clitics. Since in Tswana, -ng [-ŋ́] is also the cliticized form of the interrogative pronoun eng [ɩ̀ŋ́] ‘what?’, a grammaticalization scenario interrogative > relative can be considered. Note however that this scenario is not confirmed by the relative suffixes found in the other Southern Bantu languages (including the closest relatives of Tswana), since either their etymology is unclear, or they seem to rather have a locative origin.

4 Grammaticalization of complex constructions 4.1 Complement clauses: quotative > complementizer Tswana uses various inflected forms of the quotative verb re [rɩ̀] ‘say’ in complementizer function. The most commonly used is the infinitive gore [χʊ̀ -rɩ̀]. (60) Ke bona gore o na le tiro e ntsi. [kɩ̀-bɔ́ná χʊ́ rɩ̀ ʊ̀ -nà lɩ ́-tírɔ̀ é n̂ːtsì.] s.1sg-see:prs:cj that s.2sg-be with-(cl9.)work cl9.lk (cl9.)much ‘I see that you have much work.’

4.2 Relative clauses In Tswana, relative clauses are introduced by a linker that grammaticalized from a demonstrative (see Section 2.6), and verb forms heading relative clauses are marked by a relative suffix that may be cognate with the interrogative pronoun eng [ɩ̀ŋ́] ‘what?’ (see Section 3.6).

4.3 Adverbial clauses Several of the conjunctions used to introduce adverbial clauses have a more or less transparent etymology.

794

Denis Creissels

ka [ká] ‘since, as’ is probably cognate with the reconstructed Bantu verb root *ngà ‘be like’. (61) Ka o itse tsela, ke tlaa go roma. tsɩ ̂ːlà, kɩ̀-t ɬàà-χʊ̀ -rʊ̂ ːmà.] [ká !ʊ́ -ítsɩ ́ as s.2sg-know:prs:cj (cl9.)way s.1sg-fut-o.2sg-send ‘Since you know the way, I will send you.’ ka gore [ká χʊ́ rɩ̀] ‘because’, where gore [χʊ̀ rɩ̀] is etymologically the infinitive of re [rɩ̀] ‘say’ (also used in complementizer function, see 4.1), is the commonest way to express causal subordination. (62) Ba tsile kwano ka gore ba batla go go bona. ! kwánʊ́ !ká χʊ́ rɩ̀ bá-bát ɬá χʊ́ -χʊ̀ -bɔ̂ːnà.] [bá-tsɩ̀lé s.cl2-come:prf:cj here as that s.cl2-want:prs:cj inf-o.2sg-see ‘They came here because they want to see you.’ The same meaning ‘because’ can be carried by gobane [χʊ̀ bánɩ ́] or gobo [χʊ̀ bó], optionally preceded by the instrumental preposition ka [ká]. These forms cannot be analyzed in a strictly synchronic perspective, but they are probably cognate with the infinitive of ba [bà] ‘be’. Similarly, (ka) gonne [(ká) χʊ̀ ńné] ‘because’ cannot be analyzed as a regular form of nna [ńná] ‘be’, but is certainly cognate with the regular infinitive go nna [χʊ̀ -ńná]. fa [fá] ‘when, if ’ originates from the demonstrative of locative class 16 fa [fá]. Like the other demonstratives of locative classes, fa [fá] grammaticalized first as a linker introducing locative relatives (‘(at the place) where …’), and its use was subsequently extended to the marking of temporal and conditional/hypothetical subordination. (63) Fa o rata go nyala, nka go senkela mosadi. [fá ʊ́ -ràtá χʊ̀ -ɲâːlà, ŋ́-ká-χʊ̀ -sɩ ́ŋ́kέlà if s.2sg-like:circ:prs inf-marry s.1sg-pot-o.2sg-look_for:appl:cj mʊ̀ -sâːdì.] cl1-woman ‘If you want to marry, I can look for a wife for you.’ Temporal subordination is also commonly expressed by means of a construction in which the temporal clause is introduced by an inflected form of re [rɩ̀] ‘say’ with an expletive subject index of class 9, for example in the perfect e rile [ɩ̀-r-ílè] lit. ‘it said (that)’. However, formally, in this complex construction, re [rɩ̀] ‘say’ unquestionably acts as the main verb, as evidenced among others by its behavior in negation. For example, ‘When we went to town, he did not accompany us’ is expressed literally

Grammaticalization in Tswana

795

as ‘It did not say that, we going to town, he accompanied us’ – Ex. (64). Consequently, this construction unquestionably involves a grammaticalized use of re [rɩ̀] ‘say’, and the reanalysis of the inflected forms of re [rɩ̀] ‘say’ it involves as temporal conjunctions would be the normal outcome of this process, but that stage has not yet been reached. (64) Ga e a re re ya toropong, ene a re pata. ! [χà-ɩ ́-à-rɩ ́ rɩ ́-já tʊ̀ rópôː-ŋ̀, ὲnέ neg-s.cl9-prf:neg-say:cj s.1pl-go:circ:prs (cl9.)town cl1.pro à-rɩ ́-pâːtà.] s.cl1:seq-o.1pl-accompany ‘When we went to town, he did not accompany us.’ go fitlhela [χʊ̀ -fìt ɬhὲlà] infinitive of fitlhela [fìt ɬhὲlà] ‘come upon’ has a conjunctionlike use with the meaning ‘until’. (65) Ke tlaa ema nao go fitlhela ke swa. [kɩ̀-t ɬàà-έmà ná-ʊ̀ χʊ̀ -fìt ɬhὲlà kɩ ́-ɩ̀ː-swá.] s.1sg-fut-stand:cj with-2sg inf-come_upon s.1sg-circ:prs-die ‘I will stay with you until I die.’ e sa le [ɩ ́-sà-lɩ ́] ‘since’, etymologically the persistive form of le [lɩ̀] ‘be’ with an expletive subject of class 9 (‘it being still’), already mentioned in Section 2.4.2 as a preposition, is also used as a conjunction. Concessive subordination is commonly expressed by means of e tswa [ɩ ́-tswá] ‘although’, etymologically a form of the verb tswa [tswà] ‘come out, come from’ with an expletive subject of class 9 (‘it coming out’). (66) O lobelo e tswa a le monnye ! [ʊ́ -lʊ́ -bɩ ́lɔ̀ ɩ ́-tswá á-lɩ ́ mʊ̀ -ɲ̂ ːɲὲ.] s.cl1-cl11-speed s.cl9-come_from:circ:prs s.cl1-be:circ cl1-small ‘He is fast although he is small.’ Purposive subordination is marked by gore [χʊ̀ rɩ̀] (etymologically the infinitive of ‘say’, see 4.1) followed by the subjunctive. (67) Ngwana o lelela gore o mo fe dijo. [ŋw-àná !ʊ́ -lɩ ́lέlà χʊ̀ rɩ̀ ʊ́ -mʊ̀ -fέ dìː-dʒɔ́.] cl1-child s.cl1-cry:appl:prs:cj that s.2sg-o.cl1-give:subj cl8/10-food ‘The child is crying so that you give him food.’ Apprehensive subordination is typically expressed by means of e se re gongwe [ɩ ́-sɩ̀rɩ ́ χʊ́ -ŋwɩ̀] ‘lest’, etymologically the subjunctive negative of re [rɩ̀] ‘say’ with an expletive subject of class 9 (‘let it not be said’), plus gongwe [χʊ̀ -ŋwɩ̀] ‘somewhere’.

796

Denis Creissels

(68) Tsaya mofago e se re gongwe wa bolawa ke tlala. [tsájá mʊ̀ -fáχɔ́ ɩ ́-sɩ̀-rɩ ́ χʊ́ ŋwɩ̀ wà-bʊ́ láwà kɩ ́ take.imp cl3-food s.cl9-neg-say:subj somewhere s.2sg:seq-kill:pass by t ɬâːlà.] (cl9.)hunger ‘Take some food lest you should suffer from hunger.’

4.4 Sequentiality, simultaneity, coordination of clauses Like the other Bantu languages, Tswana has dedicated ‘sequential’ verb forms characterizing non-initial clauses in clause chains reflecting a chronological representation of events, and uses the subjunctive in sequential function if the first clause of the chain is in the imperative. An emphatic expression of sequentiality can be obtained by introducing the non-initial clauses by a sequential or subjunctive form of tla [t ɬà[ ‘come’ or feta [fɩ̀tà] ‘pass’ acting as conjunctions expressing sequentiality (‘and then’). (69) Ke ile Tlhabane ka tla ka ya Gauteng. ká-t ɬà kà-jà χàútêːŋ̀.] [kɩ̀-ìlé t ɬ hàbánɩ ́ s.1sg-go:prf:cj Rustenburg s.1sg:seq-come s.1sg:seq-go Johannesburg ‘I went to Rustenburg, and then I went to Johannesburg.’ lit. ‘… I came and I went …’ (70) Tlhapang diatla lo fite lo je! dì-àt ɬà lʊ́ -fítὲ lʊ̂ ː-dʒὲ!] [t ɬ hàpá-ŋ̀ wash:imp-pl cl8/10-hand 2pl-pass:subj 2pl-eat:subj ‘Wash your hands, and then eat!’ lit. ‘… pass and eat!’ Simultaneity between the events encoded by two successive clauses can be expressed by simply putting the second verb in the circumstantial form of the present, but the second clause may also be introduced by a circumstantial form of tla [t ɬà] ‘come’ or eta [ὲtà] ‘travel’ (< Proto-Bantu *genda ‘go’) acting as a conjunction expressing simultaneity (‘and at the same time’). (71) O apaya a tla a tlhatswa. ! á-t ɬhâːtswà.] [ʊ́ -ápájà á-t ɬá s.cl1-cook:prs:cj s.cl1-come:circ:prs s.cl1-wash:circ:prs ‘She does the cooking, and at the same time she does the washing.’ (72) Go ja o eta o bua ga go a siama. ʊ́ -bùá [χʊ̀ -dʒá !ʊ́ -έtá inf-eat s.2sg-travel:circ:prs s.2sg-speak:circ:prs

Grammaticalization in Tswana

797

χà-χʊ́ -à-síâːmà.] neg-s.cl15-prf:neg-be_good:dj ‘Eating and speaking at the same time is not good.’ The additive coordination of clauses can be marked by e bile [ɩ ́bìlé] ‘moreover’, etymologically a perfect form of ba [bà] ‘be’ with an expletive subject of class 9 (‘it having been’). (73) a. Ke bua Setswana e bile ke a se kwala. [kɩ̀-búá sɩ̀-tswáná ɩ ́bìlé s.1sg-speak:prs:cj cl7-Tswana it_having_been > and kɩ ́-à-sɩ ́-kwâːlà.] s.1sg-dj-o.cl7-write:prs ‘I speak Tswana and I write it.’ b. Ke rekile diaparo e bile ke rekile ditlhako. [ kɩ̀-rékílé dí-àpàrɔ̀ ɩ ́bìlé kɩ̀-rékílé s.1sg-buy:prf:cj cl8/10-cloth it_having_been > and s.1sg-buy:prf:cj dí-t ɬ hàːkʊ́ .] cl8/10-shoe ‘I bought clothes and (I bought) shoes.’

5 Summary In this paper, I have tried to show that Tswana shows evidence of the grammaticalization/reanalysis paths listed below, either as emerging processes in present-day Tswana or as processes having operated in the history of the Tswana (and sometimes shared by some other Bantu languages). In this enumeration, small capitals highlight relatively little-known and/or semantically non-trivial grammaticalization scenarios for which at the same time the evidence from Tswana is particularly convincing. Questions marks signal unclear cases. – ‘the people at X’s place’ > associative plural marker > plain plural marker (2.2) – ‘liver’ > locative marker (2.4.1) – ‘being there’ > locative marker (2.4.1) – genitival linker of locative class (‘place of ’) > locative marker (2.4.1) – deictic locatives (‘here’, ‘there’, ‘therein’) > locative adpositions (2.4.1) – locative adpositions > adpositions also fulfilling non-locative functions (2.4.1) – locational nouns > adpositions (2.4.1) – (?) ‘be like’ > instrumental adposition (2.4.2) – identification copula (‘it is’) > adposition marking the agent phrase in passive constructions (2.4.2)

798 – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

Denis Creissels

‘if it is still’ > temporal adposition ‘since’ (2.4.2) ‘if it is’ > adposition ‘regarding’ (2.4.2) ‘if it is not’ > adposition ‘except’ (2.4.2) ‘which is not’ > adposition marking a contrast between the NP it combines with and another NP present in the same construction (2.4.2) ‘be side by side with’ > adposition ‘near’ (2.4.2) ‘pass’ > adposition ‘more than’ (2.4.2) ‘arrive’ > adposition ‘until’ (2.4.2) ‘be face to face with’ > adposition ‘in front of ’ (2.4.2) ‘be with’, ‘have’ > adposition ‘rather than’ (2.4.2) ‘begin’ > adposition ‘from’ (2.4.2) ‘leave’ > adposition ‘from’ (2.4.2) ‘come out/from’ > adposition ‘from’ (2.4.2) ‘go with’ > adposition ‘according to’ (2.4.2) nominalized form of ‘meet’ > adposition ‘against’ (2.4.2) nominalized form of ‘look’ > adposition ‘about’, ‘with reference to’ (2.4.2) nominalized form of ‘be side by side’ > adposition ‘about’, ‘with reference to’ (2.4.2) nominalized form of ‘go and fetch’, ‘follow’ > adposition ‘because of ’ (2.4.2) ‘point’ > adposition ‘on account of ’, ‘because of ’ (2.4.2) locative marker > standard of comparison marker (2.4.2) demonstrative > noun-modifier linker (2.6) dependent form of ‘be’ > noun-modifier linker (2.6) ‘at X’s place’ > genitive marker (2.7) ‘child’ > diminutive affix (2.8) diminutive > feminine (2.8) ‘woman’ > feminine affix (2.8) feminine > augmentative (2.8) ‘father’ > derivational affix ‘owner of’, ‘responsible for’, ‘expert on’ (2.8) ‘be like’ > similative affix (2.8) ‘come’ > future’ (3.2) ‘be like’ > potential (3.2) ‘build’ > auxiliary ‘do something continually’ (3.2) ‘increase’ > auxiliary ‘do something often’ (3.2) ‘look for, want’ > auxiliary ‘need to do something’ or ‘seem to be about to do something’ (3.2) perfect of ‘look for, want’ > auxiliary ‘have nearly done something’ (3.2) ‘return’ > auxiliary ‘do something again’ (3.2) perfect of ‘surround’, ‘revolve’ > auxiliary ‘have done something in the past year’ (3.2) ‘not to be’ > auxiliary ‘never do something’ (3.2)

Grammaticalization in Tswana

– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

799

‘not to spend the day somewhere’ > auxiliary ‘have ceased doing something’ (3.2) perfect of ‘spend the night somewhere’ > auxiliary ‘have done something the day before’ (3.2) ‘struggle for something’ > auxiliary ‘seem to be about to do something’ (3.2) ‘be with’ > auxiliary ‘do something usually’, or ‘be obliged to do something’ (3.2) ‘spread, stretch’ > auxiliary ‘continue to do something’ (3.2) perfect of ‘spread, stretch’ > auxiliary ‘have eventually done something’ (3.2) ‘be’ > auxiliary ‘do something continually’, or ‘be still doing something’ (3.2) perfect of ‘like, want’ > auxiliary ‘have nearly done something’ (3.2) persistive form of ‘be’ (‘still be’) > auxiliary ‘have done something long ago, or for the last time’ (3.2) ‘go out’, ‘come from’ > auxiliary ‘have just done something’ (3.2) perfect of ‘remain’ > auxiliary ‘be already doing something’ (3.2) perfect of ‘look for’ > auxiliary ‘have nearly done something’ (3.2) future of ‘surround’, ‘revolve’ > auxiliary ‘do something during the next year’ (3.2) future of ‘spend the night somewhere’ > auxiliary ‘do something before the end of the day’ (3.2) future of ‘get up’ > auxiliary ‘do something the next day’ (3.2) ‘spend the day somewhere’ > auxiliary ‘do something continually’ (3.2) ‘leave’ > auxiliary ‘be close to doing something’ (3.2) applicative form of ‘walk’ > auxiliary ‘be close to doing something’ (3.2) ‘walk’ > auxiliary ‘do something continually’ (3.2) ‘live’ > auxiliary ‘do something continually’ (3.2) ‘fear’ > auxiliary ‘do something unexpectedly’ (3.2) ‘suit’, ‘fit’ > auxiliary ‘be obliged to do something’ (3.2) perfect of ‘come’ > auxiliary ‘seem to be close to do something’ (3.2) perfect of ‘get up’ > auxiliary ‘have done something in the morning’ (3.2) ‘go’ > auxiliary ‘be close to doing something’ (3.2) ‘refuse’ > negative auxiliary (3.3) (?) interrogative pronoun ‘what?’ > relativizer (3.7) quotative verb > complementizer (4.1) ‘be like’ > conjunction ‘since, as’ (4.3) ‘at the place where’ > ‘when’, ‘if ’ (4.3) ‘come upon’ > conjunction ‘until’ (4.3) ‘be still > conjunction ‘since’ (4.3) ‘come out/from’ > conjunction ‘although’ (4.3) ‘let it not be said somewhere’ > conjunction ‘lest’ (4.3)

800 – – – – –

Denis Creissels

sequential form of ‘come’ > conjunction ‘and then’ (4.4) sequential form of ‘pass’ > conjunction ‘and then’ (4.4) simultaneous form of ‘come’ > conjunction ‘and at the same time’ (4.4) simultaneous form of ‘go’ > conjunction ‘and at the same time’ (4.4) ‘it having been’ > conjunction expressing the additive coordination of clauses (4.4)

Abbreviations appl = applicative, caus = causative, circ = circumstantial, cj = conjoint, cl = noun class, cstr = construct, dem = demonstrative, dim = diminutive, dist = distal, dj = disjoint, fut = future, gen = genitival linker, imp = imperative, inf = infinitive, lk = linker, loc = locative, neg = negative, o = object index, pass = passive, pers = persistive, pl = plural, pot = potential, prf = perfect, pro = pronoun, prs = present, rel = relative, s = subject index, seq = sequential, sg = singular, subj = subjunctive

References Blanchon, Jean. 1999. Tone cases in Bantu group B40. In Jean Blanchon & Denis Creissels (eds.), Issues in Bantu Tonology, 37–82. Cologne: Rüdiger Köppe. Bostoen, Koen. 2010. Reflexive-reciprocal polysemy in South-Western Bantu: Typology and origins. Paper presented at Syntax of the World’s Languages IV Conference, Lyon, 23–26 September. Bostoen, Koen, Sebastian Dom & Guillaume Segerer. 2015. The antipassive in Bantu. Linguistics 53(4). 731–772. Cole, D. T. 1955. An introduction to Tswana grammar. Cape Town: Longman. Creissels, Denis. 1997. Prépositions issues de déictiques locatifs: L’exemple du tswana. Faits de langues 9. 71–80. Creissels, Denis. 1999. The role of tone in the conjugation of Setswana. In Jean Blanchon & Denis Creissels (eds.), Issues in Bantu Tonology, 109–152. Cologne: Rüdiger Köppe. Creissels, Denis. 2002. Valence verbale et voix en tswana. Bulletin de la Société de Linguistique de Paris 97(1). 371–416. Creissels, Denis. 2003. Présentation du tswana. Lalies. 5–128. Creissels, Denis. 2004. Non-canonical applicatives and focalization in Tswana. Paper presented at the Syntax of the World’s Languages conference. Leipzig, 5–8 August. Creissels, Denis. 2006. Tswana verb morphology and the Lexical Integrity Principle. Lingue e linguaggio 5(1). 49–66. Creissels, Denis. 2009. Construct forms of nouns in African languages. In Peter K. Austin, Oliver Bond, Monik Charette, David Nathan & Peter Sells (eds.), Proceedings of Conference on Language Documentation & Linguistic Theory 2, 73–82. London: SOAS University of London. Creissels, Denis. 2011. Tswana locatives and their status in the inversion construction. Africana Linguistica 17. 33–52. Creissels, Denis. 2013. Les prépositions simples du tswana. In Jesse Teng (ed.), Prépositions et postpositions: approches typologiques et formelles, 17–56. Paris: Lavoisier.

Grammaticalization in Tswana

801

Creissels, Denis. 2016. Additive coordination, comitative adjunction, and associative plural in Tswana. Linguistique et langues africaines 2. 11–42. Creissels, Denis. 2017a. The conjoint-disjoint distinction in the tonal morphology of Tswana. In Jenneke Van der Wal & Larry Hyman (eds.), The conjoint / disjoint alternation in Bantu, 200– 238. Berlin: Mouton de Gruyter. Creissels, Denis. 2017b. Similarity, suitability, and non-epistemic modalities (volitionality, ability, and obligation). In Yvonne Treis & Martine Vanhove (eds.), Similative and equative constructions: A cross-linguistic perspective, 79–89. John Benjamins. Creissels, Denis. Forthcoming. Bantu languages: Typology and variation. In Lutz Marten, Ellen Hurst, Nancy Kula & Jochen Zeller (eds.), The Oxford guide to the Bantu languages. Oxford University Press. Creissels, Denis, Anderson Chebanne & Heather Nkhwa. 1997. Tonal morphology of the Setswana verb. Munich: Lincom Europa. Doke, C. M. & S. M. Mofokeng. 1957. Textbook of Southern Sotho grammar. Longman South Africa. Dom, Sebastian, Guillaume Segerer & Koen Bostoen. 2015. Antipassive / associative polysemy in Cilubà (Bantu, L31a): A plurality of relations analysis. Studies in Language 39(2). 354–385. Fleisch, Axel. 2005. Agent phrases in Bantu passives. In F. K. Erhard Voeltz (ed.), Studies in African linguistic typology, 93–111. Amsterdam: John Benjamins. Grégoire, Claire. 1975, Les locatifs en bantou. Tervuren, Belgium: Musée Royal de l’Afrique Centrale. Güldemann, Tom. 1999. Head-initial meets head-final: Nominal suffixes in eastern and southern Bantu from a historical perspective. Studies in African Linguistics 28(1). 49–91. Güldemann, Tom. 2003. Present progressive vis-à-vis predication focus in Bantu: A verbal category between semantics and pragmatics. Studies in Language 27(2). 323–360. Jerro, Kyle Joseph. 2016. The syntax and semantics of applicative morphology in Bantu. Austin, USA: University of Texas at Austin dissertation. Kawasha, B. 2007. Passivization in Lunda. Journal of African Languages and Linguistics 28(1). 37–56. Kimenyi, Alexandre. 1980. A relational grammar of Kinyarwanda. Berkeley: University of California Press. Marten, Lutz. 2010. The great siSwati locative shift. In A. Breitbarth, C. Lucas, S. Watts & D. Willis (eds.), Continuity and change in grammar, 249–268. Amsterdam: John Benjamins. Meeussen, A. E. 1967. Bantu grammatical reconstructions. Annales du Musée Royal de l’Afrique Centrale 61. Nurse, Derek. 2008. Tense and aspect in Bantu. Oxford: Oxford University Press. Samsom, R. & T. C. Schadeberg. 1994. Kiinimacho cha mahali: kiambishi tamati cha mahali -ni [The locative suffix in Swahili]. Swahili Forum 1 = Afrikanistische Arbeitspapiere 37. 127–138. Schadeberg, Thilo. 1986. Tone cases in Umbundu. Africana Linguistica 10(1). 423–447. Schladt, M. 1998. Reciprocals in Bantu languages. Afrikanistische Arbeitspapiere 53. 5–251. Van der Wal, Jenneke & Larry Hyman (eds.). 2017. The conjoint / disjoint alternation in Bantu. De Gruyter Mouton. Van de Velde, Mark. 2006. The alleged class 2a prefix bɔ̀ in Eton, a plural word. In Rebecca T. Cover & Yuni Kim (eds.), Proceedings of the Berkeley Linguistics Society 31st annual meeting. Special Session on the Languages of West Africa, 119–130. Berkeley, CA: Berkeley Linguistics Society.

Christian Lehmann

19 Grammaticalization in Yucatec Maya 1 Introduction 1.1 Organization The following is an overview of those areas of the grammar of Yucatec Maya for which grammaticalization processes can be documented historically or at least be reconstructed. Thus, several areas of grammar remain out of discussion because nothing is known on their grammaticalization. These include nominal and verbal number as well as mood. Agreement of the verb with subject and object is reflected in Section 2.5.2. Among subordinate clauses, only adverbial clauses will be treated, while complement and relative clauses are left out of consideration. In Section 2.5.2, Mayan languages outside the Yucatecan branch are included in the comparison in order to reconstruct an origin for a grammaticalization process.

1.2 Prehistory and history of Yucatec Maya Mayan languages are spoken in Guatemala and countries bordering on it. Their current geographical distribution, shown in Figure 1, does not differ too much from their distribution in Pre-Columbian times. Figure 2 shows the genealogical tree, based on Kaufman (1990: 62) with some emendations. With the exception of the Wastecan outlier, Mayan languages share most of their grammatical structure, much of their lexicon, but considerably less of their inventory of grammatical morphemes, which underwent much renewal in the specific languages. In the space delineated by Figure 1, Section 2.5.2.3 will distinguish two main groups by shared grammatical properties. The Lowland group comprises the Yucatecan and Greater Tzeltalan branches with the only exclusion of Tzotzil, as well as Mocho’ and Ixil. The Highland group includes the complementary subset. The Yucatecan languages comprise Yucatec Maya, Lacandón, Itzá and Mopán. They are spoken on the peninsula of Yucatan and in more southern regions of the lowland in Belize, the Petén region of Guatemala and the Mexican state of Chiapas. The internal subdivision of this branch may have started a thousand years ago. While Mopán is clearly a distinct language, Yucatec, Lacandón and Itzá are better regarded as dialects of a single language, with the grading shown in Figure 2. The history of Yucatec Maya may be roughly subdivided into Classical (i.e., PreColumbian) Maya, Colonial Yucatec Maya from 1516 on and Modern Yucatec Maya from 1750 on. Colonial Yucatec is relatively well attested, chiefly in the Diccionario de Motul (Ciudad Real 1577) and the three grammars of Coronel (1620), San Buenahttps://doi.org/10.1515/9783110712735-005

804

Christian Lehmann

Fig. 1: Current distribution of Mayan languages (from Law 2014: 78).

ventura (1684) and Beltrán de Santa Rosa (1746). Especially the latter is a linguistic work of high quality.1

1 Colonial orthography is very poor. Examples have nevertheless been quoted literally from the sources, even though in some cases their word division is clearly wrong. The only additions are the morphological boundary symbols ‘-’ and ‘=’.

Grammaticalization in Yucatec Maya

805

Fig. 2: Genealogy of the Mayan languages, based on Kaufman (1990: 62).

Consequently, the changes described below for Yucatec Maya are evidenced by a corpus of 450 years of documented history. In some cases, grammarians have described different stages of this history, sometimes even noting explicitly grammaticalization phenomena observable at their time. Colonial Yucatec Maya was more synthetic than contemporary Maya, which is more analytic. Many conjugation categories and formatives coded synthetically in

806

Christian Lehmann

Colonial Yucatec are lost in Modern Yucatec. Since the endpoint of a grammaticalization path is loss, several such cases could be mentioned below as illustrating grammaticalization down to zero. Among them are a vocative in -eh and a subjunctive in -Vb (illustrated by [82]). With one exception, such losses will not be described here, simply because they teach us little about grammaticalization.

1.3 Typological sketch of Yucatec Maya Yucatecan languages have grammaticalized semantic relationality, opposing in the grammar relational to absolute nouns, transitive to intransitive verbs and prepositions to adverbs. Dependency constructions are head-marking: indexes cross-reference the subject, direct object, the possessor in a nominal construction and the complement of a preposition. The index is obligatory, the nominal dependent is optional. The two sets of indexes are dubbed A and B; this will be explained in Section 2.5.2.1. Ergativity is essentially restricted to a split in the index paradigm of the intransitive predicate conditioned by a conjugation category called status which properly includes aspect and will be explained more fully in Section 3.2. Morphology is mildly synthetic and predominantly agglutinative. Conjugation is much more complex than declension. Word order is essentially right-branching, with some left-branching constructions in the noun phrase and the verbal complex. Sentence syntax involves left-dislocation of a topicalized constituent and clauseinitial position of the focused constituent, often with a mark on the verb of the extrafocal clause (see Section 5).

2 Grammaticalization of nominal categories 2.1 Numeral classification Like other Mayan languages,2 Yucatec Maya has numeral classification: the numeral is combined with a classifier into one word. Currently, about 21 numeral classifiers are in use.3 Those which have an etymology are listed in Table 1. They are clearly based on nouns, several among these of deverbal origin (Briceño Chel 1997).

2 But not all of them: (26) below from Tz'utujil features a numeral without classifier. 3 For more than one reason, the exact size of the paradigm cannot be determined: 1) There is much dialectal and diachronic variation. 2) The criteria keeping mensuratives – which number more than a hundred – out of the set of numeral classifiers are insufficiently sharp to force a decision in every case. In particular, several of those words (including láak’ and tíich of Table 1) almost never cooccur with a noun being counted or measured.

Grammaticalization in Yucatec Maya

807

Tab. 1: Origin of some Modern Yucatec numeral classifiers. classifier

class

origin

héek’ láak’

branch, twig person located beside another extended object bird inanimate object limb sheet fork [of tree or branch]

héek’ ‘branch, twig’ láak’ ‘companion’

páach’ píich’ p’éel tíich tsíil xa’y

pach’ line [in writing] pich’ ‘grackle, melodious blackbird’ nomen patientis of p’el ‘shell, shuck’ > grain (of corn) nomen patientis of tich ‘extrude’ nomen patientis of tsil ‘tear off’ xa’y ‘fork of tree or branch’

With the exception of p’éel, the most basic numeral classifiers are absent from Table 1 because they are no longer transparent. The formation of the majority of those listed reduces to the conversion of a noun into a numeral classifier by combining it directly with a numeral. So far, no grammaticalization is involved. Numeral classifiers are in use from the earliest sources. However, Colonial Yucatec yet allows occurrence of numerals without a classifier. In (1a), the numeral combines directly with an abstract noun.4 In (1b), the numeral occupies the focus position generally occupied by adverbs, so a more literal reading might be ‘by three’ (German zu dritt). In (1c) finally, the numeral takes a suffix denoting time distance to form a temporal adverb. (1)

Colonial Yucatec a. Yan=h-ebal hun olal tex. exist=cop-subj one mind you.pl ‘May you be of one mind.’ or: ‘May there be concord among you.’ (San Buenaventura 1684: 41v) b. ox cul-ic cħicħ ca in hul-ah=e three sit-dep bird conj a.1.sg shoot-cmpl=r3 ‘three birds were sitting together and I shot them’ (Coronel 1620: 46) c. ox-eh – ox-he three-days.future three-days.past ‘in three days – three days ago’ (Coronel 1620: 65)

4 Coronel (1620: 38) even has hun Cruz ‘a crucifix’, although there a religious notion of unclassifiability may have bent Mayan grammar.

808 (2)

Christian Lehmann

Modern Yucatec a. hun-p’éel óolal one-cl.inan mind ‘concord’ b. óox-túul kul-ik ch’í’ch’ three-cl.an sit-dep bird ‘three birds were sitting’ c. ox-p'éel k'in-ak three-cl.inan day-past ‘three days ago’

All of the formations of (1) are ungrammatical in Modern Yucatec. Instead of (1a), Modern Yucatec has (2a). (1b) would require the classifier of (2b). The second expression in (1c) would have to be (2c) (with a variant of the temporal suffix). In short, in Modern Yucatec, numerals do not occur without a following classifier. They are, in fact, prefixes to classifiers and mensuratives (Lehmann 2010). Transparadigmatic obligatorification here is the main criterion for grammaticalization. Moreover, some of the classifiers underwent semantic bleaching and phonological erosion, as is clearly recognizable in p’éel cl.inan, mostly pronounced p’é. The set of numeral classifiers has been gradually shrinking throughout the centuries. Of the maximum set of 21 items, most can be neutralized in favor of p’éel. The active vocabulary of most speakers born in the 20th century comprises little more than p’éel and túul (as already observed in Tozzer [1921: 103]), and occasionally p’éel is the only classifier used. In sum, grammaticalization of the numeral complex involves morphological reduction of the numeral to a prefix and obligatoriness of a classifier chosen from an ever-smaller paradigm.

2.2 Possession Yucatecan languages share a relational noun aal ‘child, offspring’. This forms an irregular compound with ti’ loc to yield ti’aal id., yet preserved with this meaning in Lacandón (Perez 2003: s. v. tiaar). The other languages have grammaticalized the word into a dummy possessum, as e.g., in Modern Yucatec in ti’a’l (a.1.sg property) ‘mine’; cf. (3a). Structurally, such a dummy head noun is needed as a prop for the clitic indexes of set A. To the extent that it appears from the sources, this construction has made a very fast career in this function: There is no trace of it in Coronel and San Buenaventura; in Beltrán, however, the third person singular complex is even further grammaticalized to function as a preposition ‘for’ (3b) and a conjunction ‘in order to’ (3c).

Grammaticalization in Yucatec Maya

(3)

809

Modern Yucatec a. a ti'a'l le taak'in=o' a.2 property(b.3) dem money=r2 ‘the money is yours’ (hala'ch_008.3)5 b. bix ken u ch'a' áalimeentoh u_ti'a'l u winkl-il=e' how sr.fut a.3 take(subj)(b.3) food for a.3 body-rel=r3 ‘… how he would get food for his body’ (Hijo_044.3) c. túun pompom-chek' u_ti'a'l u háa's-k u yóol le prog\a.3 stamp-foot(incmpl) for a.3 frighten-incmpl a.3 mind dem peek'o' dog=r2 ‘he is stamping in order to frighten the dog’ (EMB_0265) d. u tih-il ma'loob u_tia(')l u mees-il áabril=e'… a.3 dry-incmpl good for a.3 month-rel April=top ‘(that) it dries well until the month of April …’ (Trab_009f)

The grammaticalization of ti’a’l has obviously been guided and accelerated by the model of Spanish para. This is suggested by (3d), where ti’a’l serves as a translation equivalent of Spanish para even in what is, in cross-linguistic perspective, a highly idiosyncratic sense of a benefactive preposition.

2.3 Determiners Colonial Yucatec has no article and does not signal (in-)definiteness as such. The semantic definiteness of the NPs in (4)f is not coded; and the NPs in (6) may be semantically definite or indefinite. (4)

Colonial Yucatec (Coronel 1620: 57) lic y-il-ic-ob ek ox-tul ahau-ob ipfv a.3-see-cmpl-pl star three-cl.an chief-pl ‘the Three Kings saw the star’

5 Abridged references of this structure refer to an unpublished corpus of modern Yucatec which is accessible online: http://linguistik.uni-regensburg.de/yuclex/diccionario/dic_index.html.

810

Christian Lehmann

(5)

Colonial Yucatec (Coronel 1620: 43) Tabx yan cah licil ca ben-el where exist village ipfv a.1.pl go-incmpl ‘where is the village that we are going to?’

(6)

Colonial Yucatec (Coronel 1620: 46) chee cul-ic chhichh tree sit-dep bird ‘on a/the tree a/the bird is sitting’

In the period of Spanish contact, Yucatec has been acquiring a definite and indefinite article paradigm. For definiteness, the discontinuous determiner described in Section 2.5.1 is grammaticalized. The combination of the demonstrative lay with the referential particles (RP) =la’, =lo’ and =le’ is not found in the 17th cent. grammars. Beltrán has hints of it, e.g., in (7). (7)

Colonial Yucatec (Beltrán 1746, § 194) ma Batab lay lo neg chief dem r2 ‘he is not the/a chief ’ (lit.: ‘not chief is that one’)

In the short span since Beltrán’s time, the construction [ le XNom YRP ]NP becomes the standard way of marking definiteness of an NP, with the l of the enclitic RP getting lost in most positions, as illustrated by (8). (8) Modern Yucatec a. le nah=a’ ‘this house’ b. le nah=o’ ‘that house’ c. le nah=e’ ‘the aforementioned house’ On the one hand, =o’ is unmarked as against =a’ and often used if no deixis is involved. This kind of definite determination is even used with proper names, e.g., le Hwaan=o’ ‘John’, going in this beyond the degree of grammaticalization of the Spanish definite article. On the other hand, the referential particle may be omitted, as in (9). (9)

Modern Yucatec tsik le u suuhuy nook'-il u lu'm-il Meehikoh! respect(imp.2.sg)(b.3) dem a.3 virgin cloth-rel a.3 earth-rel Mexico ‘respect the flag of Mexico!’ (ACC_0138)

Grammaticalization in Yucatec Maya

811

Except for the variation mentioned, this amounts to a complete grammaticalization of a demonstrative to a definite article. At the same time, the numeral ‘one’ gets grammaticalized, too. All the examples of hun ‘one’ found in the Colonial Yucatec texts are nothing else than a numeral. Since then, however, hun provided with a numeral classifier has been increasingly used for indefinite determination. For instance, the title of the fable of the tortoise and the hare reads as in (10). (10) Modern Yucatec hun-túul áak yéetel hun-túul kéeh one-cl.an tortoise and one-cl.an deer ‘a tortoise and a deer’ (AAK_000) In this way, Yucatec Maya has acquired a paradigm of definite and indefinite article within two and a half centuries. Influence from Spanish must have accelerated the process, the more so as the formatives le and hun resemble Spanish el and un not only in their syntagmatic position, but even phonologically. In fact, hun has acquired an allomorph un during the 20th century, which cannot be obtained by any rule of phonology.

2.4 Case relators The Mayan languages lack the category of case throughout. What may be described in this place is the grammaticalization of prepositions. From Pre-Columbian times, there is one fully grammaticalized preposition ti’ (Loc), with a prevocalic allomorph t-, which marks both complements and adjuncts in local function and indirect objects. Most local prepositions are based on relational nouns designating regions of space. One of the two most common constructions of such denominal prepositions is [ [ ti’ XA.i YN.rel ]P ZNP.i ]PP, where Y is the slot for the relational noun to be grammaticalized to a preposition and Z is its lexical complement as cross-referenced by X, as illustrated in (11)f. (11) Colonial Yucatec (Coronel 1620: 59) mán-i t=u yam vinic-ob pass-cmpl.b.3 loc=a.3 interspace man-pl ‘he passed among the people’ (12) Modern Yucatec (Yaxley_0191) t=u yáam le nah-o'b=e' hach loob loc=a.3 interspace dem house-pl=r3 very weedy ‘between the houses it is very weedy’

812

Christian Lehmann

Tab. 2: Modern Yucatec denominal local prepositions. noun

meaning

complex preposition

meaning

iknal táan paach óok'ol áanal tséel

proximity front back top bottom side

ti’ X iknal Z ti’ X táan Z ~ táanil ti’ Z ti’ X paach Z ~ paachil ti’ Z ti’ X óok’ol Z ti’ X áanal Z ti’ X tséel Z

‘by, near Z’ ‘in front of Z’ ‘in back of Z’ ‘on, above Z’ ‘under Z’ ‘beside Z’

As shown by the examples, there is no change in this respect from Colonial to Modern Maya. Table 2 enumerates some local prepositions forming this and an alternative construction. From Beltrán (1746) on, the simple preposition ti’, which converts the relational noun construction into a prepositional phrase, becomes optional and is usually omitted in casual speech with the third person, as in y-okol-ob (a.3-on-pl) ‘on them’ (Beltrán 1746: 145). Moreover, several among these prepositions undergo semantic and phonological erosion. Thus, paach also designates company. Óok’ol gets reduced to o’l before a vowel and also takes on more abstract meanings like ‘on account of ’. On the grammaticalization of the benefactive preposition see Section 2.2, and of prepositions to conjunctions see Section 4.1.

2.5 Pronouns and indexes 2.5.1 Demonstratives Colonial Yucatec has some free demonstrative formatives which partially combine to yield discontinuous deictic marking, as follows: First, there is a set of demonstrative words which only code demonstration, but no particular deictic value. This set includes importantly lay ‘that’, which is also used in anaphoric function, as in (13). In Coronel (1620: 5), it is adduced as the third person member of the paradigm of free personal pronouns. (13) Colonial Yucatec (Coronel 1620: 41) lay v chun bin-ci padre that a.3 ground go-dep.cmpl father ‘that is the reason why the reverend went’ This pronoun still sounds lay in Beltrán 1746, but since then has been reduced to le (dem). Apart from fossilized uses, it no longer occurs as a substantival pronoun, but has become a determiner; see Section 2.3.

Grammaticalization in Yucatec Maya

813

On the other hand, Colonial Yucatec has two deictic pronouns, la ‘this’ and lo ‘that’. These may occur independently of a preceding demonstrative word, as in (14) and (15). (14) Colonial Yucatec (Coronel 1620: 105) t=u dza-ah ten=x lo prfv=a.3 give-cmpl me=also r2 ‘he also gave me that’ (15) Colonial Yucatec (Coronel 1620: 68) ma úuchac in beelt-ic lo=e neg possible a.1.sg do-incmpl r2=r3 ‘I cannot do that’ In the colonial grammars, the two deictic pronouns form a plural, e.g., lo-ob (r2-pl). Furthermore, they may determine a noun and then follow it, as in (16). (16) Colonial Yucatec (Coronel 1620: 27) tup cib lo extinguish(imp) candle r2 ‘put that candle out’ Now a demonstrative word is optionally followed, in the same clause, by one of the deictic pronouns to fix the former’s deictic value, as in (17) and (18). (17) Colonial Yucatec (San Buenaventura 1684: 9 Cr) Te v bin-el huun t=in=menel la=e there a.3 go-incmpl letter loc=a.1.sg=doing r1=r3 ‘there goes this letter at my instigation’ (18) Colonial Yucatec (Coronel 1620: 104) he lic tanl-e Dios=e, lay bin bót-ab-ac lo indif ipfv serve-subj God=top that fut pay-pass-subj r2 ‘whoever serves god, that one will be paid’ This discontinuous combination becomes obligatory in Modern Yucatec: Any of the demonstrative words triggers an occurrence of one of the deictic pronouns; and these, in turn, almost only occur when preceded by one of the demonstrative words. Examples are in (8) above. Second, the binary deictic paradigm is joined by the enclitic element =e’, which follows a left-dislocated topical constituent, as in (18). Initially, =e’ may even follow a deictic pronoun, as in (15) and (17); but once it joins the paradigm of the latter, it

814

Christian Lehmann

no longer combines syntagmatically with them. Consequently, this paradigm becomes semantically wider, since its primary subdivision is [±deictic], and it is only for [+deictic] that we get the binary contrast of [±proximal]. Third, the deictic formatives are no longer used as free forms, as in (14)f, but exclusively as enclitics, preferably at the end of a clause or of a left-dislocated constituent, as in (74). They become invariable, no longer taking the plural suffix. At the same time, they preserve their initial /l/ only in continuous combination with certain demonstrative words and otherwise forfeit it. As a consequence, the paradigm of referential clitics becomes formally quite homogeneous: =a’, =o’, =e’. The entire process is an exemplary instance of paradigmaticization. In conclusion, grammaticalization turns one of the demonstrative words, viz. lay, into a definite determiner and forms a paradigm of two deictic clitics and a referential particle.

2.5.2 Personal pronominal elements 2.5.2.1 Paradigms and syntagmatic positions All Mayan languages have at least three sets of personal pronominal formatives. All but one of these paradigms are clitic or bound and function as cross-reference indexes; the last is a set of independent personal pronouns. The main paradigms of bound indexes are called set A and set B in Mayan linguistics. The functions of the pronominal sets are as follows: – Indexes of set A cross-reference the possessor of a nominal group and the actor of the transitive verb. Moreover, in the split-subject marking languages like Yucatec, they cross-reference the subject of an intransitive verb in some verbal statuses (Section 3.2). – Indexes of set B cross-reference the subject of a non-verbal clause and the undergoer of the transitive verb. In the split-subject marking languages, set B also cross-references the intransitive subject in the complementary subset of statuses. – The free pronouns are reinforced forms of set B forms. They appear as left-dislocated topic and in focus position. Some languages including Yucatec Maya have enclitic variants which function as indirect object (as, e.g., in [86]). These sets of functions are heterogeneous, which is why we have to stick to the mnemonically unhelpful labels ‘set A’ and ‘set B’. Table 3 shows the reconstructed forms of sets A and B (according to Kaufman [1990: 71f]). For the first person plural, the exclusive form is given. All of these pronominal elements are free forms at the stage of Proto-Maya. In all Mayan languages, the set A index precedes the possessed nominal, crossreferencing the possessor. In some of them, it has become a prefix to the possessed

Grammaticalization in Yucatec Maya

815

Tab. 3: Proto-Mayan pronominal paradigms. A

B

(n)w

iin

2

aa(w)

at

3

u~r

Ø

Pl. 1

q(a)

o’nh

2

ee(r)

ex

3

k(i)

eb’

Sg. 1

nominal.6 In the following, we will be concerned with the syntagmatic position and the autonomy of set A and B indexes vis-à-vis the verb. (19)f provide a set of illustrative examples. (19) Q'eqchi' a. x-at=b'e [pfv-b.2.sg]=go ‘you went’ b. x-at=x-b'oq [pfv-b.2.sg]=[a.3-call] ‘he7 called you’ (Dayley 1990: 344) (20) Modern Yucatec a. h bin-ech pfv go-b.2.sg ‘you went’ t'an-ech b. t=u pfv=a.3 call(cmpl)-b.2.sg ‘he called you’

6 To be precise: Set A has a subset of allomorphs which appear if two conditions are fulfilled: First, the index directly precedes its head (i.e., no other dependent of the head intervenes), and second, this head starts with a vowel. In this case, the prevocalic allomorph appears; and in contradistinction to the other allomorph, this is prefixal in all Mayan languages under all circumstances. This prefixation is an automatic consequence of the two conditions: nothing separates the index from its grammatical host, and the prevocalic allomorph forms a canonical syllable with the latter’s onset. Examples are in (4) (3rd sg.) and (26) (1st sg.), among others. 7 Mayan languages lack the category of gender and have little or no sex marking, either.

816

Christian Lehmann

The examples are in the completive status, which conditions ergative alignment in all Mayan languages. Both in (19) and in (20), the set A index immediately precedes the transitive verb, although it appears to attach to it only in (19b). A greater difference, however, lies in the position of the set B index: In (19), it precedes the verb, while in (20), it follows it. The position of the indexes is often described uncomprehendingly in the literature. A diachronic perspective will clarify matters. We will review the fate of each of the two sets in turn. 2.5.2.2 Indexes in possessive, ergative and subject function There is a recurrent notion in the literature that “the Set A person markers are invariably prefixes in all Mayan languages” (Grinevald and Peake 2012: 13).8 That is not so.9 Set A pronouns were free in Proto-Maya and became clitic in the specific languages. Since they always precede their head, it seems indeed natural for them to become proclitic. Now the first thing to note here is that all Mayan languages have at least some enclisis; and some of them, including the languages of the Ch’olan and Yucatecan branches, are decidedly averse to proclisis. In these languages, the phonological bond of set A indexes clashes with their syntactic dependency. This leads to complicated grammaticalization processes, but to prefixation only under special circumstances. Although the set A index bears the same syntactic relation to its nominal and to its verbal head (prepositions may be left out of consideration here), there is an important difference between the two constructions: In the nominal construction, the structural slot immediately preceding the pronominal index may be occupied by a determiner. Although in such cases the index cliticizes to the determiner, no systematic processes of coalescence have been described. In most occurrences, the pronominal index is the initial subconstituent of the noun phrase, anyway. In the verbal construction, the structural slot immediately preceding the pronominal index is occupied by some tense/aspect/mood formative, whose status and grammaticalization will be analyzed in Section 3.4 and which will here simply be called an auxil-

8 Same claim in Dayley (1990: 344); likewise Hofling (1991: 18) for Itzá and Coon (2010: § 3.1) for Ch'ol. 9 Information on the clitic or affixal status of grammatical formatives in the Mayan languages is notoriously unreliable. For instance, some Mayanists declare Yucatec auxiliaries and set A indexes to be prefixes of the full verb. Dayley (1990: 376) spells: “t-in-tz’on-aj-Ø” (‘I shot it’; same structure as [22b]; likewise Vapnarsky, Becquey, and Monod Becquelin [2012]), which is clearly wrong in the face of data like (24b). Several authors declare the enclitic set A indexes to be proclitic; thus, e.g., Bruce S. (1968: 40, 45) for Lacandón, despite his own examples (e.g., 1968: 37, 45) to prove the contrary. Shklovsky (2005: ch. 3) corrects earlier analyses of Tzeltal set A indexes as prefixes, showing that they are enclitic. In Tzotzil, set A indexes precede the verb, but in second position of the template of the verb form, whose first position is obligatorily occupied by an aspectual formative. For the other Mayan languages adduced below, esp. (31)f, no explicit analysis is available, so one has to rely on the orthography of the sources.

Grammaticalization in Yucatec Maya

817

iary. This yields the sequence ‘Auxiliary Set_A_index Verb’. Now the auxiliary is present in verbal clauses with much more regularity than the determiner in a noun phrase. As a result, the pronominal index cliticizes to the auxiliary and may form irregular portmanteau morphs with it. However, here too, there are some verbal clauses not preceded by an auxiliary and therefore introduced by the pronominal index. The initial state of set A indexes is as words which stand independently if they are the first element in their sentence, but which form a phonological word with whatever precedes them in the same clause (for the possessive index, this is mostly a preposition or a verb form). The former position is illustrated in (21) and (86) below from Modern Yucatec, in (35) from Colonial Yucatec and in (23a) from Ch’ol. The (#a) examples of (21)–(23) show the set A index in possessive function, while the (#b) examples show it in ergative function. Attachment of the clitic to whatever precedes it is shown by (22) for Modern Yucatec and by (23b) for Ch’ol. (21) Modern Yucatec a. In ho’l=e túun k'i'nam. a.1.sg head=r3 prog\a.3 ache ‘My head aches.’ wil-eh? b. In a.1.sg see-subj(.b.3) ‘(Do you want) me to meet him?’ (22) Modern Yucatec ho’l a. t=in loc=a.1.sg head ‘on my head’ wil-ah b. t=in pfv=a.1.sg see-cmpl ‘I met him’ (23) Ch’ol jol a. c a.1.sg head ‘my head’ (Aulie & Aulie 1998: 232) taj-a b. tsa=c pfv=a.1.sg meet-cmpl ‘I met him’ (Aulie & Aulie 1998: 243)

818

Christian Lehmann

Although these indexes precede their grammatical head, they are not bound to it. Instead, they are enclitic, just as all the clitics in these languages are. In syntactic terms, they cliticize to the wrong side.10 Moreover, modifiers of the head may follow the clitic, as shown in (24),11 again with the clitic in adnominal (#a) and adverbal (#b) position. (24) Modern Yucatec a. ichil =u11 seen p'uha'nil=e' within a.3 very upset:rsltv:abstr=top ‘in all her fury …’ (Hnazario_012.4) hach ka’n-al-o’b b. ts’=u term=a.3 very tired-incmpl-pl ‘they are already very tired’ (Hk’an_487.2) There are, thus, in these languages both phonological and syntactic obstacles to a morphological union of the clitic with its grammatical head. On a cline from independent word via clitic to affix, these elements are closer to the word than to the affix. They differ, however, from free pronouns in their inability to constitute a sentence. The union of the auxiliary with the enclitic set A index illustrated by (24b) is further analyzed in Section 3.4. While this complex remains independent in Yucatecan, Ch’olan and Tzeltal, it univerbates with the following full verb in other Mayan languages. Tzotzil is one of these, as illustrated by (25). (25) Tzotzil l-i-bat pfv-a.1.sg-go ‘I went’ In yet other Mayan languages, it appears that the set A clitics did change their prosodic orientation and directly became prefixes to their host. Their prevocalic allomorphs (fn. 6), which had always been prefixes, may have acted as analogical models here. (26) is from Tz'utujil, thus, from the Quichean branch (Dayley 1985:

10 S. Klavans 1985 on mismatches between syntactic constituency and prosodic orientation of clitics. 11 By means of similar examples, Shklovsky (2005, ch. 3) shows that set A indexes are no prefixes in Tzeltal. 12 In constructions like (24a), the index is enclitic, too. The same goes for (51) below, where the standard pronunciation is [ʦ’ɔ̰klu …] (with syncope). In such constructions, all existent orthographies respect the clitic’s word status.

Grammaticalization in Yucatec Maya

819

ch. 3.1.2). (26a) shows the (prevocalic) set A index in possessive function, while (26b) shows it in ergative function, both as prefixes of their head. Here the set A indexes possess the maximum degree of grammaticalization reached in the Mayan family. (26) Tz'utujil a. K'o jun w-oochooch. exist one a.1.sg-house ‘I have a house.’ (o. c. 142) b. x-at=w-aajo’ pfv-b.2.sg=a.1.sg-love ‘I loved you’ (o. c. 107) To conclude, while the set A indexes can and have become prefixes of their head in several Mayan languages, they are stopped from doing so at least in the Ch’olan and Yucatecan branches and in Tzeltal. The lesson from this is that structural properties of a construction may stop the progress of grammaticalization of its grammatical formative.

2.5.2.3 Indexes in absolutive and subject function Now for the set B indexes, these are described in the literature as being alternately prefixes or suffixes to the verb (Dayley 1990: 74, 344; Grinevald and Peake 2012: 13). Actually, at the stage of Proto-Maya, they were words with a tendency to enclisis. There was, therefore, no direct passage to prefixes from the beginning. Similarly as the enclitic set A indexes of the Ch’olan and Yucatecan branches, the enclitic set B indexes of Tz'utujil and Huastec (Dayley 1990, Section 4.7) may occur independently if no preceding host is available. The nominal clause of (27a) and the perfect clauses of (27b) and (27c) as well as the transitive clause of (28) lack an initial auxiliary. (27) Tz'utujil aachi. a. In b.1.sg man ‘I am a man.’ (Dayley 1985: 298) wa’-naq b. in b.1.sg eat-prf ‘I have eaten’ (Dayley 1985: 75)

820

Christian Lehmann

c. in ki-ch'ey-oon b.1.sg a.3.pl-hit-prf ‘they have hit me’ (Dayley 1985: 75) (28) Huastec (Dayley 1990: 380) tin kw’iy-a-al b.1.sg&a.3 scold-thm-incmpl ‘he scolds me’ Always judging from the orthography of the sources (cf. note 9), this does not entail that the pronominal index is prefixed to the verb. The proper diachronic analysis would seem to be that this is a continuation of the original free status of set B. In most other Mayan languages, set B elements are always left-bound. If the predicate is nominal, as in (27a), they follow it. In Jacaltec, set B indexes are still clitic (rather than affixal) when following a non-verbal predicate, as in (29). (29) Jacaltec (Grinevald and Peake 2012: 19) winaj hach man b.2.sg ‘you are a man’ In a verbal clause, an enclitic set B index may, in principle, follow either the full verb or the auxiliary. This is visualized in Figure 3.

Lowland variant AUX

Highland variant

=set B

set A

stem -deriv -status

=set B

Fig. 3: Syntagmatic position of set B.

The first alternative, called ‘Lowland variant’ in Figure 3, was already illustrated by the Yucatec (20). While the enclitic set A index (of 20b) coalesces with the auxiliary, the set B index – likewise leftbound – attaches to the full verb. The Highland variant, viz. the set B morpheme attaching to the auxiliary, is illustrated by (30) and (31). (31a) shows the set B index cross-referencing the intransitive subject, while in (31b), it cross-references the undergoer and is followed by a deverbal directional. (30) Awacatec (Dayley 1990: 360) n-b’iy ja kxh pfv b.2.sg a.1.sg-hit ‘I hit you’

Grammaticalization in Yucatec Maya

821

(31) Mam kamb’aa-n xjaal a. n-chi prog-b.3.pl win-ap person ‘the people are winning’ t-b’iyo-’n Cheep ka’ xjaal b. ma ch-ok pfv b.3.pl-ill a.3-hit-dir Joe two person ‘Joe hit a pair of people’ (Dayley 1990: 356) It should be observed that the grouping determined by the position of the set B index roughly corresponds to the distribution of the set A index, attaching to the left in the Lowland group, but to the right in the Highland group. The variation shown in Figure 3 appears synchronically as an alternative position of the pronominal index. In a diachronic perspective, the set B index always occupied one position, viz. enclitic to the verb. The synchronic variation lies elsewhere: In a complex sentence with a matrix verb A and a dependent verb B (see Section 3.4.2), A may be impersonal and disallow any pronominal indexes. Or else it may be personal or accept the personal indexes of B by raising. If A then gets grammaticalized to an auxiliary, the index remains in situ under either alternative. The first alternative produces the Lowland variant, the second, the Highland variant. Most other languages of the Highland group, including K’iche’, Q’eqchi’, Chuj and several more, go one step further and univerbate the entire complex consisting of the auxiliary, the indexes and the full verb. The resulting morphological structure was already illustrated by (19). It appears to be flat synchronically, but diachronically it is layered as suggested by the bracketing: In a first step, the set B index gets suffixed to the auxiliary, while the set A index gets prefixed to the full verb. In a second step, the auxiliary complex agglutinates to the full verb complex. The result looks as if set B and set A morphemes formed a prefix sequence13 in front of the verb. Diachronically, however, the set B marker was never prefixed to the full verb.14 Most Mayan languages opt for one of the alternatives shown in Figure 3.15 The Lowland variant, where the set B morpheme follows the full verb, is chosen by the Lowland group introduced in Section 1.2. The Highland variant, where the morpheme attaches to the auxiliary, is chosen by the Highland group. A few languages including Tzotzil have both variants depending on verbal categories. Given the spe-

13 Some languages, including Chuj and Mam, have an additional morphological slot between the set B prefix and the set A prefix; cf. (31b). 14 Vinogradov (2017) shows the diachronic sequence postulated here to match historical evidence in the case of Qeqchi’. It is also consonant with the observation of Grinevald and Peak (2012: 13): “If they [the set B indexes] are prefixes, there are no separate pre-consonantal and pre-vocalic forms as there are for the set A prefixes.” 15 On the areal distribution of the variant positions of set B indexes, cf. Law (2014: ch. 4.2.4).

822

Christian Lehmann

cific syntactic preconditions for the Highland variant mentioned before, this one may constitute the innovation.16 The lesson from these changes is the following: It is possible for etymologically identical paradigms to emerge as prefixes in one language and as suffixes in another. However, given their advanced degree of grammaticalization, affixes do no “hop”. Neither is it necessary, if a verb has a sequence of affix slots, that the innermost affix agglutinated first and the outermost last. Instead, what is synchronically a complex morphological structure may have more than one center. In more concrete terms: If there is a sequence of affixes on a verb some of which are indexes while others are tense/aspect/mood markers, then it is probable that the latter once were auxiliaries which first attracted some indexes and then coalesced with the full verb, adding to it a whole bunch of morphemes in one go. 2.5.2.4 Free personal pronouns Colonial Yucatec lacked a neutral anaphoric pronoun; the demonstrative lay ‘that’ (Section 2.5.1) took its place. By erosion, this becomes le in Modern Yucatec. In certain uses in focus position, it survives in substantival function, but now mainly functions as the deictically neutral prenominal determiner discussed in Section 2.3. The paradigm of the contemporary free personal pronouns is in Table 4:

Tab. 4: Modern Yucatec free personal pronouns. person

number singular

plural

1

teen

to’n

 

2

teech

te’x

 

3

leti’

leti’o’b

 

The forms are composed of ti’ (loc) plus a set B morpheme (which is zero in third person singular). The third person forms are, furthermore, combined with the demonstrative le. As a consequence, the latter gets reinforced in its function as third person pronoun. These pronouns have clitic variants, which take the postverbal position in indirect object function and whose third person forms lack the le. Thus, the paradigm

16 The innovation may be one change occurred in a genetic subgroup, in which case it may be necessary to adjust the genealogical tree of Figure 2. Alternatively, it may be due to areal contact among the languages concerned. The latter option is, in principle, argued in Law (2014), although there an alternate positioning of these indexes is assumed to be the innovation.

Grammaticalization in Yucatec Maya

823

of Table 4 is the result of a grammaticalization process involving the loss of the semantic role of the locative preposition and of the demonstrative feature.

2.5.3 Reflexive pronouns Yucatecan languages share a relational noun bah ‘counterpart, replica’, whose original meaning is preserved in Lacandón. (32) Lacandón (Perez 2003 s. v. böh) a. in böh-o a.1.sg counterpart-pl ‘my companions’ b. u böh u txé in tèt a.3 counterpart a.3 tree a.1.sg father ‘the same tree as my father’s’ Already in Colonial Maya, this root had been grammaticalized to a reflexive marker ‘self ’. This combines with set A clitics in a regular way to yield a complex reflexive pronoun. From the earliest sources to this day, there is a tendency to treat this construction as one morphological complex, writing it jointly, as in (33). (33) Colonial Yucatec (Coronel 1620: 4) in cimç-ah inbáh a.1.sg kill-cmpl(b.3) a.1.sg:self ‘I killed myself ’ To this day, the reflexive pronoun behaves syntactically as an NP. It can also depend on a preposition, as in (34). (34) Modern Yucatec (Hk’an 046) t=u ya'l-ah t=u báah prfv=a.3 say-cmpl loc=a.3 self ‘he said to himself ’ In contemporary non-standard writing, the complex reflexive pronoun in direct object function (as in [33]) is generally attached to its verb, being considered a suffix.

2.5.4 Interrogative proforms Colonial Yucatec interrogative proforms have roots as shown in the first column of Table 5 (the forms of the first two columns are from Coronel [1620]). The root for ‘who’ is grammaticalized from máak ‘person, human being’, and the root for ‘what’,

824

Christian Lehmann

Tab. 5: Yucatec interrogative proforms. Colonial Yucatec root

suffixed form

mac bal tab17 bic bahun kun

macx ba(l)x tabx bic(i)x bahunx kunx

Modern Yucatec

meaning

máax ba’x tu’x bix bahux kux

‘who’ ‘what’ ‘where’ ‘how’ ‘how much’ ‘how about’

from ba’l ‘thing’. To judge from the examples in Coronel 1620: 78, the original meaning of bahun may have been ‘quantity’. All of these roots can be used as such or can optionally be reinforced with a suffix -x, which may be cognate to the enclitic ix ‘also’ seen in (14). The combination yields the forms of the second column. The basic and reinforced forms are apparently in free variation in Colonial Yucatec. Given the basic meanings of mac, bal and bahun, it appears that the first-column words acquired interrogative force in the focus position of the interrogative sentence; there is no other mark distinguishing it from a declarative sentence. The combinations with the focusing suffix -x got then grammaticalized in interrogative function. In Modern Yucatec, interrogative pronouns do not occur in their root form; they bear the -x suffix, and a subset of them alternatively bears an -n suffix, like ba’n ‘what’, tu’n ‘where’. Further grammaticalization leads to their use as relative pronouns in headless relative clauses.

3 Grammaticalization of verbal categories 3.1 Verbal clause structure With some simplification, the Yucatecan verbal complex consists of a verb form inflected by set B indexes and preceded by set A indexes; s. Figure 4 below. The elementary independent verbal clause consists of a verbal complex in completive status and its dependents, as in (35). (35) Colonial Yucatec (Motul: s. v. kamnicte) u kam-ah nicte in mehen a.3 get-cmpl-b.3 flower a.1.sg son ‘my son got the flower (i.e., got married)’

17 Preserved in Lacandón to ‘where’ (Perez 2003: s. v. toan).

Grammaticalization in Yucatec Maya

825

Already in Colonial Maya, many verbal clauses are introduced by an element which codes tense, aspect or mood and which will be called an auxiliary (see Section 3.4.1 for discussion of the appropriateness of this term). In Modern Yucatec, this is the default for independent verbal clauses. At this stage, the verbal complex with its dependents as illustrated by (35) only forms a clause core, while an independent declarative verbal clause generally – except in the perfect – requires an auxiliary in front of it. (36) illustrates this with the terminative auxiliary. (36) Modern Yucatec (HA'N_043.1) ts'o'k u mux-ik le ta'b=o' term a.3 grind-incmpl(b.3) dem salt=r2 ‘he has ground the salt’

3.2 Status In all Mayan languages, the verb has a suffixal slot for a category called status, which comprises the subcategories of dependent status proper, aspect and mood. These suffixes belong to the earliest layer of the diachrony and are completely grammaticalized. The Colonial Yucatec paradigm is as shown in Table 6; material in parentheses is optional or confined to pausa position. The incompletive (aspect) derives from a nominalization conditioned by embedding of the full verb clause core under matrix auxiliaries, a construction to be analyzed in Section 3.4.2. In intransitive verbs, status conditions the choice between set A and set B indexes to cross-reference the subject; see Section 2.5.2.1. The status paradigm is alive to this day, but given its high degree of grammaticalization, it is fragile. Several endings appear only in pausa and are syncopated

Tab. 6: Colonial Yucatec status conjugation. stem class intransitive status

aspect/mood

plain

subjunctive completive

dependent

\

basic

derived

basic

-Vc

-n-ac

-Vb

(-i)

-n(-ah)(-i)

subjunctive completive incompletive

transitive

-ebal -ci

(-e) -ah -ic

-n-ici -Vl

derived

-(i)ci / -i18 -ic

18 The allomorph -i appears if the subject is the focus constituent of a cleft-construction.

826

Christian Lehmann

otherwise (Beltrán 1746: §§ 135–147). Some of the allomorphy is utterly complicated, syncretistic and constantly exposed to variation. All in all, this paradigm is unproductive and unstable. Partly preceding, but mainly within the period from Colonial Yucatec to Modern Yucatec, the inherited aspect system is largely renewed by a paradigm of auxiliaries analyzed in Section 3.4. These mark relatively fine distinctions not only of aspectual, but also of temporal and modal categories. They stem from different categorical sources and form different syntactic constructions with the clause core. The latter, in turn, condition different status categories on the full verb. Conditioning them, they render them largely redundant.

3.3 Voice/Valency Yucatecan languages share a passive suffix -b, shown in (37) for Itzá. (37) Itzá (Hofling 1991: 33) k=u tz’on-b-ol ipfv=a.3 shoot-pass-incmpl ‘it is shot’ The suffix appears, inter alia, as a submorphemic component of a passive participle in -bil, shared at least among the Yucatecan, Tzeltalan and Ch’olan branches (cf. Vapnarsky, Becquey, and Monod Becquelin 2012). A passive morpheme -b (of unknown provenience) can therefore safely be reconstructed for some prehistorical stage of Yucatecan. In Mopán and Itzá, this remains a suffix. In Yucatec and Lacandón, it develops an infixal allomorph, as illustrated in (38). (38) Modern Yucatec k=u ts’on-ol ipfv=a.3 shoot-incmpl ‘it is shot’ Forms such as these are already attested in the earliest colonial grammars, so the infixing must have occurred in Pre-Columbian times, but after Itzá split off from Yucatec-Lacandón. Passive is marked by the /ʔ/ infix in primitive transitive roots, i.e., those of the canonical phonotactic shape CVC. Other transitive stems form the passive by a set of allomorphs conditioned by the status category, but sharing a /b/. We appear to be faced with a change of a suffix to an infix by metathesis: [ʔ] is a systematic allophone of /b/ in coda position in all historical phases of Yucatec. Once the /b/ crosses the right morpheme and syllable boundary of the root, it turns into [ʔ].

Grammaticalization in Yucatec Maya

827

Now for forms like the one seen in (37), metathesis cannot be motivated by phonology. Possibly primitive transitive roots ending in /b/, of which there are quite a few, played a triggering role here. Modern Yucatec verbs such as ts’a’ ‘put, give’, cha’ ‘let’, ch’a’ ‘fetch’, he’ ‘open’, chi’ ‘bite’ and several more form their incompletive passive as ts’a’b-al, cha’b-al, ch’a’b-al, he’b-el, chi’b-il. In Classical Maya, these roots had themselves ended in /b/.19 By phonological rule, the sequence of two /b/s yielded [ʔb]. Given that the lexical form of the root ended in /b/, the passive form was reinterpreted as infixation of the passive morph /b/ > [ʔ]. Given the central position and high token frequency of the verbs in question, they may have served as an analogical model for other primitive transitive roots, maybe fostering a metathesis or just the infixation of a glottal stop. The details of passive morphology are complicated. Here it suffices to mention that transitive stems that do not correspond to the canonical phonotactic shape of primitive transitive roots – including all derived stems – infix the glottal stop into the status suffix instead of in the root, like il-a’l (see-incmpl\pass) ‘be seen’. This may be the only known instance of what appears synchronically as an infix inside an affix.20 What matters in the present connection is that the entire process described is an instance of grammaticalization. Not only does agglutinative morphology become fusional morphology. The passive also plays an essential role in the verbal system and in information structure and is very frequent in texts of all genres.

3.4 Aspect The following sections on tense/aspect/mood marking are an abridged version of Lehmann (2017).

3.4.1 Periphrastic aspects The clause has an initial syntactic position preceding the clause core which may be taken by elements and constituents of very different nature. Since the material in this structural position is so heterogeneous, its relation to the rest of the clause differs accordingly, and consequently the constructions are syntactically different. The differences are reflected morphologically on the full verb, which depending on the construction is in the dependent incompletive, the completive or the subjunctive status. As we will be concentrating on such constructions in which the element

19 Among the evidence is the Ch’ol cognate of Yucatecan *ch’ab, which is ch’äm. 20 Diachronically, such forms may originate in il-ab-al (see-pass-incmpl) etc. Such forms are reported in the colonial grammars (e.g. Coronel 1620: 32). While in this context, the /b/ would not become a [ʔ], it would do so after syncope applied to the last syllable of the form, i.e. in the context ilabl …

828

Christian Lehmann

verbal clause verbal clause core verbal complex finitive verb auxiliary

index A

verb stem

-status

-index B

dependents

referential clitic

conditions

Fig. 4: Syntagmatic relation between auxiliary and status.

occupying the initial position gets grammaticalized to an auxiliary, the result is that the auxiliary conditions the status. Figure 4 visualizes this dependency. The first thing to be noted about Figure 4 is that the full verb is finite. This is a peculiarity of Yucatec periphrastic constructions whose diachronic explanation will become clear in the following sections. As we saw in Section 2.5.2.2f, in the Yucatecan languages, the pronominal indexes do not combine with the auxiliary, but with the full verb. Thus, the auxiliary deserves its name only insofar as it carries tense/ aspect/mood information. Person and number, instead, are marked on the full verb, and consequently it is indeed finite. There is in Yucatec a large variety of tenses, aspects and moods that are coded in the initial position of Figure 4.21 While the construction of the initial auxiliary is inherited, most of the auxiliaries actually in use are at best in an incipient stage of grammaticalization in Colonial Yucatec. The set is large and heterogeneous, and only some of them will be described in the following subsections. We will see that all the auxiliation constructions come about by grammaticalization, but that they start off from different sources. Some of the statuses have the intransitive subject represented by a set A index, while others have it represented by a set B index. This is the alignment split already mentioned in Section 2.5. Although it is not the main object of the ensuing analyses, these will nevertheless contribute to its understanding.

3.4.2 auxiliation based on complementation 3.4.2.1 Basics Of importance for auxiliation is a kind of complex construction consisting of a main clause core and a complement clause core occupying an argument position of the governing predicate, most importantly, the subject position of an intransitive verb.

21 An extensive list of relevant markers appears in Briceño Chel (2006: ch. 1.2).

Grammaticalization in Yucatec Maya

829

Being nominalized, the complement clause is in dependent status. This construction is the model for a number of auxiliaries. The clause-initial slot of Figure 4 attracts not only intransitive verbs, but also verboids, nouns and denominal adverbs. The construction, however, remains essentially the same: in all the constructions of Section 3.4.2, the clause core depends on the initial element. 3.4.2.2 From habitual to imperfective aspect The inherited imperfective was renewed in Colonial Yucatec Maya. At the beginning of this process, there is a set of words, apparently denominal in origin, which compete for the auxiliary position. Three of these appear in (39), listed as synonymous in the colonial grammar. One of these is lic, which must be a root with the meaning ‘this time span’. (39) Colonial Yucatec (Coronel 1620: 57) cim-ç-ab-i in yum die-caus-cmpl.pass-b.3.sg a.1.sg father ti=lic / tamuk / ti válac v han-ál loc=this.span / while / loc this.time a.3 eat-incmpl ‘my father was killed while eating’ The preposition subordinating lic may be omitted, as in (40). (40) Colonial Yucatec (Motul s. v. ca6 ) lic u dzoc-ol a han-al ca ta-c-ech uay=e span a.3 end-incmpl a.2 eat-incmpl conj come-subj-b.2.sg here=r3 ‘when you have eaten, you should come here’ While Yucatecan languages possess some types of indisputably subordinate clauses (among them those in dependent and subjunctive status), the coding of temporal and tactic relations is essentially left to the tactic potential of the aspectual system (Bohnemeyer 2003; [45] below is a typical example). Formatives like the three shown in (39) reinforce this tactic relation more than subordinating their clause. At any rate, the clause introduced by lic may also be independent; and then the temporal construction may have a habitual sense (cf. Coronel 1620: 67), clearly visible in (41). (41) Colonial Yucatec (Motul s. v. lic2 ) lic in uen-el tamuk in han-al hab a.1.sg sleep-incompl while a.1.sg eat-incompl ‘I usually fall asleep while eating’ By further grammaticalization, the morpheme functions as a mere imperfective auxiliary, as in (38), (43) and as also possible in (42).

830

Christian Lehmann

(42) Colonial Yucatec (Motul s. v. lici lic) lic bin a haɔ-ic a paal-il tu men u tuz. ipfv quot a.2 beat-incmpl a.2 child-rel because a.3 lie\introv ‘They say you (habitually) beat your boy because he lies.’ Here lic is yet a word. However, there already exists a shortened variant c(i), apparently in free variation, as in the dialogue of (43): (43) Colonial Yucatec a. bal c=a uok-t-ic? what ipfv=a.2 weep-trr-incmpl ‘What are you crying for?’ b. in kéban lic u-ok-t-ic. a.1.sg sin ipfv a.1.sg-weep-trr-incmpl ‘It is for my sins that I am crying.’ (Coronel 1620: 67) In (43a), the particle already univerbates with the enclitic A index, as compared with the (43b) version. Today, the auxiliary only survives in its one-phoneme form k, obligatorily univerbates with the set A index and carries aspectual information only in contrast with more specific auxiliaries, as in (38) above.

3.4.2.3 Progressive aspect The progressive itself is a Proto-Mayan category. In Colonial Yucatec Maya, it is based on the relational noun tan,22 illustrated in (44) in its lexical meaning ‘front, middle’. (44) Colonial Yucatec (San Buenaventura 1684: 39v) t=u tan Dios loc=a.3 front god ‘in front of god’ (44) shows the regular syntactic construction for a noun designating a spatial region, presented in Section 2.4. The same configuration is also at the source of its aspectual use. Already in the earliest sources, the preposition ti’ is normally dropped. (45) – obviously a variant of (39) – illustrates the incipient progressive function for an intransitive verb.

22 The progressive function of this morpheme may be inherited from Proto-Maya; some languages, including K’aqchiquel, have plausible cognates.

Grammaticalization in Yucatec Maya

831

(45) Colonial Yucatec (Coronel 1620: 57) vtán v han-ál in yum, ca cim-ç-ab-i a.3:middle a.3 eat-incmpl a.1.sg father conj die-caus-pass-cmpl(b.3.sg) ‘my father was in the middle of eating when he was killed’ or: ‘while my father was eating, he was killed’ The original construction with the subordinating ti’ and its further evolution are, at any rate, completely analogous to the imperfective ti’ lik seen in (39): It follows the pattern where the full verb of the complement clause is in dependent status. Initially, the new auxiliary is typically used in complex sentences, where the progressive clause provides the background for the event of the main clause, as clearly shown by (45). However, and again like the imperfective, the progressive also appears in monoclausal sentences. (46) features, already at Coronel’s time, a further reduced form of the auxiliary, where the original possessive clitic preceding táan is no longer there. (46) Colonial Yucatec (Coronel 1620: 34) ma tan a túb-ul ten neg prog a.2 escape-incmpl me ‘I am not going to forget you’ Beltrán (1746) only uses the reduced form tan, as in (47). (47) Colonial Yucatec (Beltrán 1746, § 262) tan in tzeec, ca lub ku-na prog a.1.sg preach(incmpl) conj fall(cmpl.b.3) god-house ‘I was preaching, there the church collapsed’ As may be seen, this is now just a progressive aspect. By the positioning of the Wackernagel-enclitic interrogative particle, (48) shows that in Modern Yucatec, táan may still be a word. (48) Modern Yucatec (Hnazario_406) táan wáah a bin? prog int a.2 go(incmpl) ‘Are you going (leaving)? In its further development, and again in analogy with the development of the imperfective auxiliary, the progressive auxiliary coalesces with the set A index which most often follows it, with the full form surviving essentially in writing and, in the oral mode, in cases like (48). This is a process in two phases. At first, the product of the merger of táan with the three singular indexes in, a, u is tíin, táan, túun, as illustrat-

832

Christian Lehmann

ed by (3c) above. This is, however, just a transitional stage rarely represented in writing. In the end, these forms are shortened to tin, ta, tu (cf. Briceño Chel 2006: 24f), as in (49). (49) Modern Yucatec (Hts'oon_310.1) t=u sáas-tal prog=a3 dawn-fient.incmpl káa h téek líik' y-ich hun-túul le peek'=o' conj pfv for.a.moment get.up(cmpl.b.3) a.3-eye one-cl.an dem dog=r2 ‘It was dawning when one of the dogs suddenly rose his glance’ In this configuration, the progressive clause specifies a situation holding in the background, into which the event described by the following main clause is incident. This is functionally equivalent with the – much older – combination of a nominalized clause subordinated by ti’ serving as background information for the main clause (Lehmann 2017, § 4.3). In fact, since the products of the merger of the preposition and of the progressive auxiliary with the following set A index are homonymous, the two constructions are not easily distinguished. It may be assumed that the model of the nominalized construction played a role in the rather radical reduction of the auxiliary complex. 3.4.2.4 Terminative aspect One important subclass of intransitive predicators to fill the clause-initial position are phase verbs. The central phase verbs are ho'p' ‘start’ and a set of verbs including ts'o'k, all meaning ‘end’. They are normally impersonal (Coronel 1620: 34f). Personal use is possible with a few of them, but does not generate auxiliaries. In the impersonal construction, actancy is coded on the dependent verb. (50) illustrates the construction for ts'o'k. The dependent verb is in dependent status. (50) Colonial Yucatec (San Buenaventura 1684: 17r) ɔoc-i incan-ic end-compl(b.3.sg) a.1.sg:learn-dep.incmpl ‘I finished learning / have learnt it’ The phase verb ts’o’k combines with aspect auxiliaries just like any full lexical verb, e.g., in (40). It continues this life up to the present day. (51) displays a symptom of grammaticalization: the phase verb is in the incompletive, but lacks both the introductory imperfective auxiliary and the set A index. This suggests that even in the construction at hand, where the main clause comprises more than just the phase verb, the latter fulfills an auxiliary function, with the form kuxtal in its subject not just being a noun, but rather the verbal predicate (live:incmpl) of the dependent clause core.

Grammaticalization in Yucatec Maya

833

(51) Modern Yucatec (chem_ppuzoob_011) beey túun ts'o'k-ol u kuxtal le p'us-o'b=o' thus then finish-incmpl a.3 life dem hunchback-pl=r2 ‘This then was the end of the life of the P'uz.’ This grammaticalization process starts in the colonial period. The 17th century grammars adduce the phase verbs ɔoc ‘finish’ and hopp ‘begin’ only in order to mention their regular impersonal or personal construction as illustrated by (40) above. Only Beltrán (1746, §§ 85f) observes the expansion of the use of ts’o’k as auxiliary in vogue at his time. The form of this verb which occupies the clauseinitial position, becoming, thus, a component of the grammaticalization path, is the completive form triggered by perfective aspect, as in (52). (52) Modern Yucatec (Briceño Chel 2000b: 84) (h) ts’o’k in meyah pfv finish(cmpl:b.3) a.1.sg work ‘my work ended = I finished working’ In the sequel, the perfective auxiliary is omitted. In fact, by the evidence of (50), the grammaticalization of ts’o’k probably started at a time when the completive alone could make a perfective clause. Otherwise, however, the new auxiliary can maintain its full form even in the colloquial style. There is, however, a reduced form in addition to the full form, although not as widely used as the reduced form of the progressive auxiliary. The auxiliary then shrinks to its initial consonant and coalesces with the set A clitic, as shown by (24b) above (cf. Briceño Chel 2000b: 87f).

3.4.2.5 From debitive to future tense The canonical construction coding ascription of possession consists in inserting a possessed nominal in the central actant position of the existential verboid yaan (cf. [5]), as in (53). (53) Colonial Yucatec (Motul s. v. yan) yan wa a yum? exist int a.2 father ‘do you have a father?’ Once a nominalized verbal complex is substituted for the possessum of the ascription of possession, a debitive construction results. Just as the possessum is ascribed to its possessor, so the obligation is ascribed to the actor of the nominalized verbal complex, as in (54).

834

Christian Lehmann

(54) Modern Yucatec (Hala'ch_084) ba'l=e' yan a bo'l-t-ik-en however=top deb a.2 pay-trr-incmpl-b.1.sg ‘however, you must pay me’ This use is not found in the colonial sources and is documented only in the modern Yucatecan languages. The most recent development, only documented in the 20th century oral register, is a pure future without debitive connotations, as in (55), where the speaker articulates what he thinks will certainly happen. (55) Modern Yucatec (Hnazario_402.1) yan u kaxt-ik-ech a taatah deb a.3 search-incmpl-b.2.sg a.2 father ‘your father will search you’ This construction is currently ousting the (much older) predictive future (Section 3.5.2), which gets pushed back into the formal register.

3.5 Tense A Modern Yucatec tense not to be dealt with here is the assurative future with he’(l) (featured in [75]), because the semantics obtaining at its origin remain to be investigated.

3.5.1 From hodiernal past to perfective As explained in Section 3.2 and illustrated by (35), in Colonial Yucatec, completive status is the only one beside the perfect that a simple independent declarative clause may exhaustively be based on (i.e., without the need for an auxiliary). This means, at the same time, that such clauses have little marking in comparison with all other tense/aspect/mood categories appearing in independent sentences. Moreover, the completive has zero allomorphs in several contexts. The result is that many completive verbal complexes occurring in texts reduce to verb stems provided with indexes. In Colonial Yucatec, the completive clause can be marked for hodiernal completive by the particle ti' ‘there’ (or its prevocalic bound allomorph t-), a deictically neutral local demonstrative adverb which occupies the same position as the auxiliary in Figure 4. (56) shows the simple plain completive for an intransitive (56a) and a transitive (56b) verb. The two parts form minimal pairs with (57a) and (57b), which show the hodiernal completive.

Grammaticalization in Yucatec Maya

835

(56) Colonial Yucatec a. Bin-i Fiscal ti y-otoch ku, go-cmpl(b.3) inspector loc a.3-house god ‘The inspector went to the church’ b. ca vhaɔ-ah pal-alob conj a.3:beat-cmpl child-pl ‘and beat the children’ (San Buenaventura 1684: 23r–v) (57) Colonial Yucatec a. ti bin-i padre hod go-cmpl(b.3) father ‘the reverend went today / has gone’ b. t=in haɔ-ah paal hod=a.1.sg beat-compl(b.3) child ‘I beat the child today / have beaten the child’ (San Buenaventura 1684: 35r) The hodiernal completive is already highly grammaticalized in Colonial Yucatec Maya. Already in Coronel (1620), some completive examples introduced by ti’ are translated as simple past. In Beltrán's (1746) examples – e.g., §§ 264f (t) luben – the completive aspect appears variously with and without the aspect auxiliary t-, with the same Spanish translation caí ‘I fell’ and no comment on any semantic difference. In § 36, he admits that, in front of intransitive verbs, the t is “semipronunciada”, and establishes the variation taken up below. Apparently, the hodiernal component has disappeared, and what we now have is a perfective auxiliary, reduced to the phoneme t, as in (58) below, and therefore regularly univerbated with the following enclitic set A index in front of transitive verbs, as evidenced by (57b). This is, then, the only auxiliary which has already lost word status at the stage of Colonial Yucatec and become a bound morpheme. In Modern Yucatec, the perfective auxiliary has become obligatory with transitive verbs in completive status. Intransitive completive verbs get a set B index suffixed, as seen, i.a., in (20a). The monophonematic auxiliary therefore hits directly on the verb, which may start with a consonant, as in (58). Yucatec has a phonological rule converting /t/ into /h/ in front of /t/. An extended version of this rule may have applied to the perfective auxiliary. At any rate, this auxiliary has an allomorph h with intransitive verbs. A preconsonantal /h/, however, generally disappears in Yucatecan. The h to be seen in (20a) and (52) is optional both in speaking and in writing, but is mostly absent, as it is in (56a). One may speculate that what manifests itself in such cases is an uninterrupted continuation of the plain completive of Colonial Yucatec. This may be hard to settle. At any rate, since the hodiernal feature present at the beginning

836

Christian Lehmann

disappears, the result of the entire grammaticalization process is a weak reinforcement of the inherited completive status. The perfective is the only auxiliary of the Yucatecan branch that cooccurs with completive status. It remains a loner as regards both the source of the auxiliary and the status conditioned (or rather, conserved) by it on the verb. However, the more recent grammaticalization paths converge with it into a common paradigm.

3.5.2 Predictive future The motion-cum-purpose construction is a regular syntactic construction in the Yucatecan branch. It is a complex clause core starting with an oriented motion verb followed by a verbal clause core in the subjunctive, the latter coding the purpose. As long as no focus precedes the motion verb, the core verb is in plain status subjunctive. The bracketing in (58) marks off the motion and the purpose parts. (58) Colonial Yucatec (Beltrán 1746: § 110) t bin-én in cim-ez uacax [pfv go(cmpl)-b.1.sg] [a.1.sg die-caus(subj) cow] ‘I went to kill cows’ The central verbs of oriented motion (‘go’, ‘come’, ‘pass’) become irregular in their conjugation on their way to Modern Yucatec. The reduced forms of bin ‘go’ appear both with the lexical meaning and as auxiliary. The motion-cum-purpose construction with bin as motion verb is grammaticalized to a future in the Yucatecan branch. Coronel (1620) already calls it “futuro” and provides examples of it. Beltrán (1746 § 299: 128) lists bin as ‘partícula de futuro’, giving examples (59)f for the intransitive and transitive construction, resp. (59) Colonial Yucatec (Beltrán 1746: 149) bin bol-nac-én dzedzetàc fut pay-subj-b.1.sg little.by.little ‘I shall pay little by little’ (60) Colonial Yucatec (Beltrán 1746: 129) caix u tancoch in hanal=e, bin in zi-ib tech although a.3 half a.1.sg meal=r3 fut a.1.sg present-subj(b.3.sg) you ‘although it is half of my meal, I’ll give it to you’ The core verb keeps the subjunctive of the source construction. The motion verb complex has been reduced to the root of the motion verb. This becomes impersonal like all the other auxiliaries and, in Yucatec and Lacandón, undergoes an idiosyn-

Grammaticalization in Yucatec Maya

837

cratic phonological change: the vowel of the auxiliary bin (not of the lexical verb!) is lengthened and gets high tone in Yucatec. This may be due to analogy with the progressive auxiliary táan, but may also be regarded as the expression counterpart of the grammatical change. At any rate, the impersonalization and morphological impoverishment of the auxiliary comes under paradigmaticization and may be ascribed to analogical pressure from the older auxiliation constructions analyzed in Section 3.4.2. (61) illustrates the construction for both an intransitive and a transitive verb. (61) Modern Yucatec (Xipaal_032) Bíin suu-nak yéetel bíin in wil-eh. fut return-subj(b.3) and fut a.1.sg see-subj(b.3) ‘He will come back and I will see him.’ This future construction finds its place in the tense/aspect/mood paradigm at the side of three other futures, viz. the debitive future (Section 3.4.2.5), the immediate future (Section 3.5.3) and an assurative future not analyzed here. It does not become an immediate future, as so many futures based on the motion-cum-purpose construction do in other languages. Instead, it bears a feature of neutral, objective prediction, which may be related to the impersonality of its auxiliary. We find the predictive future at an intermediate stage of grammaticalization. On the one hand, the reduction process mentioned above proves that it is grammaticalized to some extent already at the stage of Colonial Yucatec. On the other hand, it retains its word status to this day. The origin of the predictive future construction is the motion-cum-purpose construction. It differs from the other tense/aspect/mood auxiliaries analyzed in Sections 3.5.1f in that the emerging marker – the verb ‘go’ grammaticalized to a future marker – does not originally occupy the clause-initial position described at the beginning of Section 3.4.1 and instead is the remnant of a complete superordinate clause. However, the canonical model for an auxiliary construction requires that the auxiliary be monomorphematic, impersonal and occupy the clause-initial position. In its grammaticalization, the motion-cum-purpose construction is forced into the Procrustean bed of the verbal clause expanded by an initial position, which is the template for the auxiliary construction. This is, thus, a clear example of grammaticalization guided by analogy.

3.5.3 Immediate future The auxiliation strategy leading to the immediate future originates in a verb-focus construction. From there, we get to the immediate future in two steps: First, on the basis of the verb ‘go’ in focus, a focused progressive is formed. Second, this strategy

838

Christian Lehmann

applies to the ‘go’ verb of the motion-cum-purpose construction to form the immediate future of its purpose component. Putting the lexical main verb of a clause into its focus position requires filling the gap that it leaves in the extrafocal clause by a verb meaning ‘do’. For this purpose, Colonial Yucatec used a verb cib ‘do’, which is totally irregular and defective. The verb is rarely found with this meaning in a simple transitive clause; the lexicon contains other verbs meaning ‘do’. Instead, it is used almost exclusively in focus constructions and there takes the form cah. At the stage of Colonial Yucatec, the verb is indispensable as a pro-verb in the verb focus construction, illustrated in (62). (62) Colonial Yucatec (Coronel 1620: 71f) han-ál v cah eat-incmpl a.3 do ‘he is eating’ As suggested by the translation, this construction functions as a progressive in Colonial Yucatec Maya. As a matter of fact, it figures much more prominently in colonial grammars than the simpler progressive with the auxiliary táan (Section 3.4.2.3). The 17th century grammars start their account of the conjugation with the periphrastic construction based on cah, calling it the “presente”. (63) completes the exemplification with a transitive verb. (63) Colonial Yucatec (Coronel 1620: 72) cámbeç-ah in cah ti pál-alob teach-introv(incmpl) a.1.sg do loc child-pl ‘I am teaching the children’ While (62) and (63) are focus constructions, there are a number of peculiarities. Focusing of a finite verb requires its nominalization. Therefore, the focused verbs show the nominalizing suffixes which join the status paradigm as incompletive aspect. This process is relatively complicated with transitive focused verbs, as in (63). The purpose of the construction is to put the verb into focus. Consequently, its dependents remain in the extrafocal clause. Therefore, the verb is detransitivized before it is nominalized. The internal syntax of the extrafocal clause is adjusted, too: what was the direct object of the focused verb becomes a prepositional object (Beltrán 1746: § 172). The verb focus construction is, thus, marked with plurivalent verbs. The progressive aspect views what the verb designates as an ongoing situation that the referent of the subject is in. Consequently, the functional locus of the progressive aspect is in intransitive verbs. The verb focus construction is therefore well suited to get grammaticalized into a progressive aspect. The resulting construction may be dubbed focused progressive. It is grammaticalized in Colonial Yucatec to the extent that it is a member of the conjugation paradigm.

Grammaticalization in Yucatec Maya

839

On its way into the modern Yucatecan languages, the pro-verb cib is fossilized; only the form cah/ka’h occurs in a couple of contexts. It is ousted from its function as a pro-verb in regular verb-focus constructions by the lexical verb beet/meent ‘make’. Ka’h survives in this function only in formulas. Neither is the focused progressive with ka’h further grammaticalized to a plain progressive. As we have seen in Section 3.4.2.3, the progressive construction which gets established involves a different auxiliary. Instead, verb focusing is applied to the motion-cum-purpose construction analyzed in Section 3.5.2. What is put into focus position is the verb benel/binel/bin ‘go’, while the purpose part of the construction is left behind in the extrafocal clause core. The resultant specific construction is, thus, a merger of the focused progressive with the motion-cum-purpose construction. (64)f illustrate it with an intransitive and transitive full verb, resp. (64) Colonial Yucatec (Coronel 1620: 50) ben-el in cah ti han-al go-incmpl a.1.sg do loc eat-incmpl ‘I am going to eat’ (65) Colonial Yucatec (San Buenaventura 1684: 9Br) Bin-el in cah incam-bez pal-alob go-incmpl a.1.sg do a.1.sg:learn-caus(subj.b.3) child-pl ‘I am going to teach the children.’ At the stage of Colonial Yucatec, there is already an asymmetry conditioned by the transitivity of the full verb: a transitive full verb is in the subjunctive motivated by the motion-cum-purpose construction, as in (65), while an intransitive full verb, as in (64), is subordinated by the preposition ti and therefore nominalized, marked by incompletive status. At the same stage, the binary contrast between bin ‘go’ and tal ‘come’ is yet maintained in their grammaticalization, as proved by (66): (66) Colonial Yucatec (Coronel 1620: 69) tal(-el) v cah in bot-ic in ppax come-incmpl a.3 do a.1.sg pay-incmpl a.1.sg debt ‘I am coming to pay my debt’ Further reduction of the paradigm, however, leads to the consequence that the only verb possible in the focus construction with ka’h of Modern Yucatec is bin, and it only survives in the modern immediate future, illustrated by (67). (67) Modern Yucatec a. bin in ka'h xíimbal ti' le chaan kaah …=e' imm.fut a.1.sg do walk(incmpl) loc dem little village=r3 ‘I am going to walk to that little village’ (Hts'on_016)

840

Christian Lehmann

b. bin in ka'h in xíimba-t yuum ahaw imm.fut a.1.sg do a.1.sg walk-trr(subj.b.3) master chief ‘I am going to visit the chief ’ (Hts'on_020) The preposition ti’ no longer shows up in this construction in Modern Yucatec. Furthermore, the full verb no longer needs to be an agentive verb, as shown by (68). (68) Modern Yucatec (FCP 395) bin in ka'h kíim-il go a.1.sg do die-incmpl ‘I am going to die’ By desemanticization, the semantic component of motion has disappeared, and what remains is only the direct tie between present topic time and future event time. Bin … ka’h is now a complex auxiliary with the value of immediate future. Bin … ka’h is the only discontinuous auxiliary of the language. And while bin is impersonal like all its other auxiliaries, ka’h is the only one with personal inflection. As a consequence, with transitive full verbs, the subject is cross-referenced twice (Briceño Chel 1998: 82), as is apparent from examples like (67b). There is, consequently, much redundancy in this auxiliation. In the colloquial register of Modern Yucatec, the full forms are rarely used. They are normally reduced in phonologically irregular ways, and there is currently much variation in this respect. The fusion of bin in/a/u ka’h into nika’h/naka’h/nuka’h (noted in Briceño Chel 1998: 82, 2000a: 88f, 2006: §§ 1.2f) is illustrated in (69a). If the full verb is transitive and therefore preceded by a set A index, the ka’h of the auxiliary coalesces with it, as in (69b). (69) Modern Yucatec a. Ni-ka’h meyah t=in kool. imm.fut\a.1.sg-do work loc=a.1.sg milpa ‘I am going to work on my cornfield.’ (Briceño Chel 2000a: 88) b. Ni-k=in hant bak’ imm.fut\a.1.sg-do=a.1.sg eat:trr(subj) meat ‘I am going to eat meat’ (Briceño Chel 2000a: 99) Other idiosyncratic mergers occur in a variant of the construction in which the ka’h component takes set B indexes. Using this variant with a transitive verb leads to cross-referencing the subject three times. The reduction processes applied in this

Grammaticalization in Yucatec Maya

841

context disguise this to a certain extent. Thus, the first syllable of the complex auxiliary in (70) contains the vowel of the first person singular set A clitic. (70) Modern Yucatec (FCP_043) mi-ka'h-en in wa'l te'x … imm.fut\a.1.sg-do-b.1.sg a.1.sg say(subj) you.pl ‘I’m going to tell you …’ However, contractions with clitics of other persons may also contain an i, so that the interim result of these changes is an auxiliary which takes set B suffixes to crossreference the subject of the clause core. In cases like (70) it leads to doubling, quite untypical of the language. The only comment one may make on the situation is that before a construction reaches the endpoint of its grammaticalization path, much variation occurs. The following steps compose this complex grammaticalization process: – The motion verb bin ‘go’ is semantically bleached; the movement component disappears. – The incompletive or subjunctive verb remaining in the extrafocal clause is reinterpreted as the main verb. – The internal structure of the complex ‘bin set_A_index ka’h’ is blurred. By being forced into the Procrustean bed of the initial position, it is reanalyzed as a discontinuous immediate future auxiliary with internal inflection. – The whole sentence ceases to be complex; it is reinterpreted as a single clause. – Whatever may have remained of the focal emphasis on the initial verb vanishes; the construction becomes open to different information structures that may be superimposed. The model of this complex reanalysis is the structure of the simple fully finite clause, in which the initial auxiliary combines with the enclitic subject pronoun and is followed by the verbal complex (as, e.g., in [57b]). The result of the change conforms to that model to the extent possible for a discontinuous auxiliary.

3.6 Copula Mayan languages generally lack a copula proper. Instead, they have derivational processes of verbalization which are regularly used if a non-verbal predicate needs to inflect for verbal categories, i.e., needs to be marked for some aspect or mood. While most of these derivational operators contribute some aktionsart shade to the meaning, Colonial Yucatec does have a pure copula. The root h- is an inactive intransitive verb which serves as an all-purpose verbalizer. Its free use is limited to its function as an impersonal auxiliary (Section 3.4.2.1) in the modal construction illustrated by (71).

842

Christian Lehmann

(71) Colonial Yucatec (Coronel 1620: 75) h-ij xi-c-en ti Ho cop-cmpl(b.3) go-subj-b.1.sg loc Mérida ‘I would have gone to Mérida’ Unless this formative is in initial position, it is enclitic. Then it combines with any nonverbal predicate and conjugates through the persons and numbers, as in (72) and (73). (72) Colonial Yucatec (Coronel 1620: 52) makol=h-en in cámbeç in mehén-ob lazy=cop(cmpl)-b.1.sg a.1.sg teach(subj) a.1.sg son-pl ‘I was lazy in teaching my sons’ (73) Colonial Yucatec (Coronel 1620: 66) ca-ten=h-i in kaç-ic çukin two-times=cop-cmpl(b.3) a.1.sg break-dep fast ‘(it was) twice (that) I broke fast’ This formative survives in contemporary Maya only in the third person singular form of the completive aspect of a handful of non-verbs, like yaanhih ‘there was’, sáashih ‘it was at dawn’. In all other cases, the non-verbal predicate either is verbalized by one of the other verb-creating derivations (cf. [49]) or accompanied by a temporal adverb. Compare the first clause of (74) with (72). (74) Modern Yucatec (Xipaal_043) le ka'ch hach chichan-en=e' beey u yúuch-ul teen=o' dem back.then really small-b.1.sg=top thus a.3 happen-incmpl me=r2 ‘when I was very young, that happened to me’ The modal meaning coded by h- in (71) is now also coded by the temporal adverb ka’ch, as in (75). (75) Modern Yucatec (ACC_0529) he'l ak bin-o'n ka'ch wáah ma'loob k'iin=e' ass.fut a.1.pl go(incmpl)-1.pl back.then if good sun/day=r3 ‘we would go if the weather were fine’ This is, then, historical evidence of the loss of a copula.

Grammaticalization in Yucatec Maya

843

4 Grammaticalization of complex constructions 4.1 Adverbial clauses Adverbial clauses are very loosely integrated, and their dependency is not easily distinguished from parataxis. They are introduced by coordinative or subordinative conjunctions. These may be denominal or deverbal in origin. Denominal conjunctions start out as complex prepositions. Occupying the position of the relational noun in the construction of the denominal preposition of Section 2.4, the abstract noun ok’lal ‘reason, sake’23 forms a complex preposition meaning ‘on account of, because of ’, as in (76). The component tu is preserved in written style and usually dropped in casual speech. (76) Modern Yucatec (Maglah Canul 2014 s. v. xotk’iin) xotk’int-a’b t=u yo’lal u si’pil sentence-cmpl.pass(b.3) loc=a.3 account a.3 sin ‘he was sentenced on account of his crime’ The complement may be a clause. Of the numerous occurrences in the corpus, all are purpose clauses, as in (77); none is a causal clause s.s. (77) Modern Yucatec (RMC_2092) no'xt-eh yo'lal mu'n lúub-ul wedge-imp so.that neg:a.3 fall-incmpl ‘wedge it so it doesn’t tip over’ In colonial times, the complex preposition men(el) Z ‘at Z’s instigation, by Z’s agency’ is rather freely adjoined to verbal predications, as in (78) (cf. also [17] above). (78) Colonial Yucatec (Coronel 1620: 68) úchuc t=in men possible loc=a.1.sg doing ‘I can (do it)’ This prepositional construction got fixed in the function of an agent phrase in passive constructions. This resulted in internal inflection, like tinmen ‘by me’, tamen ‘by you’ (Coronel 1620: 33). The third person form tumen then got lexicalized and now is underway to an invariable preposition taking free personal pronouns as its

23 Possibly based on the spatial region noun óok’ol ‘top’ of the paradigm treated in Section 2.4.

844

Christian Lehmann

complement, like tumen teen ‘by me’. Finally, it has also become a causal conjunction, as in (79). (79) Modern Yucatec (Maglah Canul 2014 s. v. chi’chnak) le chanpaal=a’ chi’chnak tumen wi’h dem baby=r1 ill-tempered because hungry ‘this baby is ill-tempered because he is hungry’ Summarizing the preceding two cases, a construction based on an abstract noun is grammaticalized to a subordinative conjunction. In Colonial Yucatec, conditional clauses are introduced by the all-purpose conjunction ca (káa) (conj), other uses of which are illustrated in (1b), (40) and (82). Alternatively, they may be introduced by wáah ‘if ’, which originates in an interrogative marker, viz. a particle marking polar interrogative sentence-type in independent sentences and subordinating dependent polar interrogatives, as in (48) and (53). In Modern Yucatec, only this latter conjunction survives in conditional function. It is optionally reinforced by tumen, yielding wáah tumen, with no change to the conditional sense.

4.2 Insubordination and promotion In the 17th cent. grammars, the use of what were to become the imperfective (Section 3.4.2.2) and the progressive (Section 3.4.2.3) auxiliaries is described in connection with complex sentences. Examples are (40) for imperfective lic and (45) for progressive tan. There is also evidence for licil serving as a temporal conjunction, viz. Yucatec ka’likil ‘at the time, while’ and Itzá kil ‘when’ (Hofling 1991: 26). Thus, it is possible that these formatives first introduced a temporal clause serving as the background for an adjacent main clause, and in a second step, the former got detached and independent. Historical evidence for such a diachronic hypothesis is, however, unknown. The case of the Colonial Yucatec comitative preposition etel (i.e., éetel) ‘with’ may be conceived as one of promotion to a higher syntactic level. The preposition is based on the relational noun éet ‘companion’. The derivation in -Vl must have yielded an abstract relational noun ‘company’, which however is not attested as such. The word only occurs in the construction of denominal prepositions shown in Section 2.4, so that the basic meaning of ti’ Ai éetel Zi (with A as the possessive clitic) is ‘in Z’s company’.24 By the start of the colonial history, this preposition is already more grammaticalized than the other denominal prepositions. One of the

24 There is, in Colonial Yucatec, a related preposition étun ‘at, by’ (French chez, German bei).

Grammaticalization in Yucatec Maya

845

relevant clues is the fact that the subordinating ti’ is generally – though not always – omitted. The construction is illustrated by the examples in (80).

(80) Colonial Yucatec (Coronel 1620: 90) v étel – au étel – y-etel Juan a.1.sg with a.2 with a.3-with John ‘with me – with thee – with John

– x-en y-etel ob go-imp a.3-with pl – go with them’

Contexts of use of this preposition expand in two directions. On the one hand, the structural asymmetry inherent in an adposition was already relaxed in Classical Maya. The inherited Yucatecan neutral conjunction is ca (káa), mentioned in Section 4.1. It introduces clauses with no distinction as to coordination or subordination and no information on the interpropositional relation, but it does not coordinate phrases. Phrases are coordinated in Colonial Yucatec by (y)etel, as in (81).

(81) Colonial Yucatec (Coronel 1620: 28) yan=h-om uil chicil ti kin yetel ti v exist=cop-fut(b.3) dub sign loc sun and loc moon ‘there may be signs on the sun and on the moon’

The second member of such a coordinative construction is neither first nor second person, so the preposition fossilizes in its third person form, yéetel. This form develops into an all-purpose symmetric conjunction largely exchangeable with the Spanish loan y ‘and’. This use can already be found in Colonial Yucatec, as in (82).

(82) Colonial Yucatec (San Buenaventura 1684: 25v) tibil ca a can-ab yetel ca xi-c-ech just conj a.2 learn-subj(b.3) and conj go-subj-b.2.sg ‘it would be befitting that you learn it and go’

The change of the syntactic function of the syntagma introduced by yéetel from governed to independent is a (diachronic) promotion to a higher syntactic function. On the other hand, the selection restriction on animate complements is dropped, so that the comitative preposition develops an instrumental use. Traditionally, the instrumental function had been coded by the all-purpose preposition ti’. In the two 17th cent. grammars, éetel is found in comitative, but not in instrumental function. Beltrán (1746, § 20) notes that ti’ in instrumental function is increasingly ousted by yéetel. Thus, (83a) would be an example of traditional usage, whereas (83b) would illustrate then fashionable usage.

846

Christian Lehmann

(83) Colonial Yucatec a. lox tî chê box(imp) loc(b.3) wood ‘beat him with a club’ b. lom y-etél hulté stab(imp) a.3-with spear ‘stab him with a spear’ This is, in sum, a case of polygrammaticalization.

4.3 Discourse-level conjunction The phase verb ts’o’k ‘end’ seen in Section 3.4.2.4 provides another example of polygrammaticalization. In the imperfective aspect, it is the grammaticalization source of a conjunction in a paratactic construction of the form [P; ku ts’o’kole’, Q], literally: ‘P; (now) this ends, (and) Q’, as in (84). (84) Modern Yucatec (Chakwaah_03f) K=u ts'o'k-ol=e' k=in p'o'-ik; ipfv=a.3 finish-incmpl=top ipfv=a.1.sg wash-incmpl ‘Then I wash it;’ k=u ts'o'k-ol=e' ipfv=a.3 finish-incmpl=top k=in ts'a'-ik t=eh k'áak'=o’ … ipfv=a.1.sg put/give-incmpl loc=dem fire=r2 ‘then I put it on fire …’ Q is anaphorically connected with P both by the index u (a.3) and by the referential clitic e’ (r3 or Top) composing the emergent conjunction. It is very widely used in the colloquial register, as witnessed by the monotonous repetition in (84). The phrase ku ts’o’kole’ is commonly reduced to ts’o’(h)le’, the loss of the auxiliary complex being due to grammaticalization, while the shrinking of the verb form follows regular phonological processes.

5 Main constituent order change Let the basic main constituent order be that order of the main constituents of an independent declarative clause which involves the least grammatical apparatus. In

Grammaticalization in Yucatec Maya

847

the verbal clause of Yucatecan languages and several other Mayan languages, this order is VOS. (85) shows it for Modern Yucatec; cf. (35) for Colonial Yucatec. (85) Modern Yucatec (Hijo_074.1) káa t=u t'an-ah hun-túul éenkargaadoh le taatah-tsil=a' conj pfv=a.3 call-cmpl one-cl.an employee dem father-absol=r1 ‘and this father called an employee’ Needless to say, such structurally complete configurations are not particularly frequent in the texts. Within the confines of the simple clause, there is, in Yucatec, no alternative to the verb as first main constituent. In independent sentences, this is not necessarily the sentence-initial position, because as seen in Section 3.4, a verbal sentence generally starts with an auxiliary. Marked information structure provides for two sentence-initial positions to be occupied by main constituents, viz. the position of the left-dislocated topic and the focus position. The maximum configuration was dubbed LIPOC (language-independent preferred order of constituents) in Dik (1981: 189 ff) and may be represented by Figure 5. (86) is an example.

[ left-dislocated topic

[ focus extrafocal clause ]]

Fig. 5: Extended sentence structure.

(86) Modern Yucatec (ACC_0039) le chaan lak he'l=a' in kiik síih-mah-il ten dem little bowl prsv=r1 a.1.sg elder.sister give.as.present-prf-dep(b.3) me ‘this little clay bowl, my elder sister gave it to me’ The left-dislocated topic is marked by a referential enclitic, R1 in (86). The paradigm contains an element (R3, last seen in [84]) which functions as a topicalizer if no deixis is involved. The focus itself is not marked, but the extrafocal clause is marked by a dependent status suffix, -il in (86) (s. Section 3.2). Note that the construction of (86) does not count as an instance of SVO order, since the S is not in the same clause as the V. Both left-dislocation and preverbal focusing are frequent in the texts. Certain categories, especially adverbs (43b) and adverbialized verbs, have a preference for the preverbal position. To the extent that their position becomes fixed, the emphatic potential of the construction as a focus construction is undermined. This is a correlate of grammaticalization (Lehmann 2008). While the marking of constructions with preverbal main constituents remains relatively stable in the other Yucatecan languages, it gets lost in Lacandón. Part of

848

Christian Lehmann

the relevant change is a morphological change: the paradigm of the referential clitics including the topicalizer and the dependent status on the verb get lost. As a result, we find sentences such as (87). (87) Lacandón (Bruce S. 1974: 19) K’akoch t=u me(n)t-a wolol k’uh. K’akoch pfv=a.3 make-cmpl all god ‘K’akoch made all the gods.’ (87) is a chapter heading and therefore lacks information structure. Nevertheless, the subject is in preverbal position, and there is no mark of left-dislocation or focusing. This structure characterizes the majority of independent declaratives in the Lacandón corpus and is the default order according to Bruce S. 1968: 104–106. Thus, the original information structural profile of this construction gets lost together with its outward marks. It is not even possible to ascertain whether the construction illustrated by (87) has been grammaticalized from a left-dislocation or from a focus construction. What remains is a new basic constituent order, viz. SVO. The same phenomenon may be observed in other Mayan languages, esp. Ch’orti’. In Yucatec Maya, too, this order has become more frequent recently; and sometimes, esp. in the speech of younger generations, the marks of topicalization and focusing are missing. This syntactic change is an instance of the grammaticalization of a construction which can easily occur by itself and has occurred in many languages. However, in the sociolinguistic setting in which the Mayan languages are found, it seems plausible that the change is fostered by contact with Spanish. While this is indeed plausible for all the other Mayan languages, it is more problematic in the case of Lacandón, as the Lacandones are known to have avoided Spanish contact as far as possible up to the end of the twentieth century.

6 Comparative outlook Some of the grammaticalization phenomena reviewed above recur in other languages or are even widespread. These include: – the reduction and fossilization of the numeral classifier paradigm (Section 2.1) – the development of a definite and indefinite article system on the basis of a demonstrative and the numeral ‘one’ (Section 2.3) – the transformation of spatial region and other relational nouns into adpositions (Section 2.4), and the use of the latter as subordinative conjunctions (Section 4.1) – the reduction of demonstratives to personal pronouns and to determiners and their reinforcement (Sections 2.5.1f)

Grammaticalization in Yucatec Maya

– – – – –

849

the agglutination of an auxiliary, fully equipped with finite morphology, to the full verb (Sections 2.5.2.3f) the formation of interrogative proforms from indefinite proforms used in a focus construction (Section 2.5.4) the obtainment of a future from a motion-cum-purpose construction, and of a progressive aspect from a verb-focus construction (Section 3.5.3) the desemanticization of a comitative into an instrumental adposition, and the use of such an adposition as a coordinator (Section 4.2) the acquisition of a new main constituent order by leveling of the specific information structure associated with a dislocation or clefting construction (Section 5).

Other changes seem to be rarer and may be typical of Mayan or even just of Yucatecan languages: – the recruitment of a noun meaning ‘counterpart, replica’ as a reflexive prop (Section 2.5.3) – the full grammaticalization of passive morphology down to an internal modification process and to an infix in a suffix (Section 3.3)25 – auxiliation based on impersonal complement constructions (Section 3.4.2), although this is common for modals – double cross-referencing of a verbal actant in a periphrastic construction.26 As mentioned at the outset, the Mayan family is rather uniform in its syntactic and morphological categories. Counterparts of most of the above-mentioned changes may be found in most or all members of the family. However, the Yucatecan branch stands out by the low degree of grammaticalization of the possessive/ergative pronominal indexes, of the sentence-initial auxiliaries and the syntagmatic complexes formed by these two classes of formatives. Several other Mayan languages have gone much further in their grammaticalization, turning them into slot fillers of a template of verbal prefixes. Contrariwise, other Mayan languages, with Chuj, Jacaltec and Mam among them, have grammaticalized a class of directionals in preverbal position, of which there is no trace in Yucatecan. It is conceivable that some of the more idiosyncratic grammaticalization paths may be used to typologically characterize a language.

7 Discussion and conclusion In internal comparison, it strikes the eye that some changes occurred rather quickly within the period of Modern Yucatec, while others have been underway since Pre25 The Latin r-passive is completely grammaticalized, too, but structurally only reaches the stage of affixation. 26 Andrej Malchukov mentions that this may be found in Ket and in Athapaskan, too.

850

Christian Lehmann

Columbian times. The former comprise the article system, the preposition ‘for’, clause coordination with ‘and’ and the main constituent order SVO. These are just the changes that follow a Spanish model. In comparative perspective, it appears that certain kinds of change are principally available in a language whose structure provides the necessary conditions, but that they may be triggered and accelerated by contact with another language. In such cases, one may speak of loan grammaticalization.

Acknowledgements Thanks are due to Andrej Malchukov and Björn Wiemer for helpful comments.

Abbreviations 1, 2, 3 = 1st, 2nd, 3rd person, a = possessive/subject function, absol = absolutive morpheme, abstr = abstract nominalizer, advr = adverbializer, an = animate, ap = antipassive, ass = assurative, aux = auxiliary or modal verboid, b = absolutive function, caus = causative, cfp = clause final particle, cl = numeral classifier, cmpl = completive, conj = all-purpose conjunction, cop = copula, deb = debitive, dem = neutral demonstrative, dep = dependent status, dir = directional, dub = dubitative, exist = existence, fient = fientive, fut = future, hab = habitual, hod = hodiernal past, ill = illative, imm = immediate, imp = imperative, inan = inanimate, incmpl = incompletive, indif = indifferent, int = interrogative, introv = introversive, ipfv = imperfective, loc = locative, neg = negator, pass = passive, pfv = perfective, pl = plural, prf = perfect, prsv = presentative, prog = progressive, quot = quotative, r1, r2 = referential clitic of 1st, 2nd ps. deixis, r3 = nondeictic referential clitic, rel = relational morpheme, rsltv = resultative, sg = singular, sr = subordinator, subj = subjunctive, term = terminative, thm = thematic element, top = topicalizer, trr = transitivizer. N.rel = relational noun, Nom = nominal (group), np = noun phrase, p = preposition, pp = prepositional phrase, rp = referential particle.

References Aulie, H. Wilbur & Evelyn W. de Aulie. 1998. Diccionario ch'ol de Tumbalá, Chiapas, con variaciones dialectales de Tila y Sabanilla. Reeditado por Emily F. Scharfe de Stairs. Coyoacán, D.F.: Instituto Lingüístico de Verano (Serie de vocabularios y diccionarios indígenas “Mariano Silva y Aceves”, 121). Beltrán de Santa Rosa, Pedro. 1746. Arte del idioma maya, reducido a sucinctas reglas, y semilexicon yucateco. México D.F.: Viuda de d.J. Bernardo de Hogal. Bohnemeyer, Jürgen. 2003. The grammar of time reference in Yukatek Maya. München & Newcastle: LINCOM Europa (LINCOM Studies in Native American Languages, 41). Briceño Chel, Fidencio. 1997. La lexicalización de clasificadores numerales en el maya yucateco. In María Cecilia Lara Cebada (ed.), Identidades sociales en Yucatán, 255–275. Mérida, Yucatán: Facultad de Ciencias Antropológicas.

Grammaticalization in Yucatec Maya

851

Briceño Chel, Fidencio. 1998. La gramaticalización del verbo bin (“ir”) en el maya yucateco. In Zarina Estrada Fernández (ed.), Memorias del IV encuentro internacional de lingüística en el noroeste. 4 volúmenes (Serie Lingüística, 1), 69–93. México D.F.: Instituto Nacional de Antropología e Historia. Briceño Chel, Fidencio. 2000a. La gramaticalización del verbo “ir” en las lenguas mayas de la rama yukateka. In Zarina Estrada Fernández (ed.), Memorias del V encuentro internacional de lingüística en el noroeste. 3 volúmenes, 85–101. Hermosillo, México: UniSon. Briceño Chel, Fidencio. 2000b. La gramaticalización del verbo ‘terminar’ (ts'o'okol) en maya yucateco. Lingüística Mexicana 1. 79–91. Briceño Chel, Fidencio. 2006. Los verbos del maya yucateco actual. Investigación, clasificación y sistemas conjugacionales. México D.F.: Instituto Nacional de Lenguas Indígenas. Bruce S., Roberto D. 1968. Gramática del lacandón (Departamento de Investigaciones Antropológicas, Publicaciones, 21). México D.F.: Instituto Nacional de Antropología e Historia. Bruce S., Roberto D. 1974. El libro de Chan K'in. México D.F.: Instituto Nacional de Antropologia e Historia, Departamento de Lingüística. Ciudad Real, Antonio de. 1577. Gran diccionario o Calepino de la lengua maya de Yucatán. Motul, Yucatán: Convento Franciscano (unpublished; first printed in Juan Martínez Hernández (ed.) 1929). Coon, Jessica. 2010. Rethinking split ergativity in Chol. International Journal of American Linguistics 76(2). 207–253. Coronel, Juan. 1620. Arte en lengua de maya. México D.F.: D. Garrido. Reprinted in Juan Martínez Hernández (ed.). 1929: 2–55. Dayley, Jon P. 1985. Tzutujil grammar. Berkeley etc.: University of California Press. Dayley, Jon P. 1990. Voz y ergatividad en idiomas mayas. In Nora C. England & Stephen R. Elliott (eds.), Lecturas sobre la lingüística maya, 335–398. South Woodstock, VT: CIRMA & Plumstock Mesoamerican Studies. Dik, Simon C. 1981. Functional grammar (3rd edn). Dordrecht & Cinnaminson: Foris. Grinevald, Colette & Marc Peake. 2012. Ergativity in Mayan languages: a functional-typological approach. In Gilles Authier & Katharina Haude (eds.), Ergativity, valency and voice, 15–50. Berlin: Mouton de Gruyter. Hofling, Charles A. 1991. Itzá Maya texts with a grammatical overview. Salt Lake City: University of Utah Press. Kaufman, Terrence. 1990. Algunos rasgos estructurales de los idiomas mayances con referencia especial al k'iche'. In Nora C. England & Stephen R. Elliott (eds.), Lecturas sobre la lingüística maya, 59–114. South Woodstock, VT: CIRMA & Plumstock Mesoamerican Studies. Klavans, Judith L. 1985. The independence of syntax and phonology in clitization. Language 61. 95–120. Law, Danny. 2014. Language contact, inherited similarity and social difference: The story of linguistic interaction in the Maya Lowlands (Current Issues in Linguistic Theory, 328). Amsterdam: John Benjamins. Lehmann, Christian. 2008. Information structure and grammaticalization. In Elena Seoane & María José López Couso (eds.), Theoretical and empirical issues in grammaticalization, 207– 229. Amsterdam & Philadelphia: J. Benjamins. Lehmann, Christian. 2010. On the function of numeral classifiers. In Franck Floricic (ed.), Essais de typologie et de linguistique générale. Mélanges offerts à Denis Creissels, 435–445. Lyon: École Normale Supérieure. Lehmann, Christian. 2017. Grammaticalization of tense/aspect/mood marking in Yucatec Maya. In Walter Bisang & Andrej Malchukov (eds.), Unity and diversity in grammaticalization scenarios, 173–237. Berlin: Language Science Press. Maglah Canul, Gaspar. 2014. Diccionario maya-español. Kantunilkin: unpublished ms.

852

Christian Lehmann

Martínez Hernández, Juan (ed.). 1929. ‘Diccionario de Motul maya español’ atribuido a Fray Antonio de Ciudad Real y ‘Arte de lengua maya’ por Fray Juan Coronel. Mérida, Yucatán: Compañia Tipográfica Yucateca. Perez, Patrick. 2003. Lexique lacandon. (https://halshs.archives-ouvertes.fr/halshs-00265386). San Buenaventura, Gabriel de. 1684. Arte de la lengua maya. México D.F.: Viuda de B. Calderón. (Pagination in this edition is in disorder, which is why references contain doubled page numbers and specification of recto (r) and verso (v).) Shklovsky, Kirill. 2005. Person marking in Petalcingo Tzeltal. Portland, OR: Reed College BA thesis. Tozzer, Alfred M. 1921. A Maya grammar, with bibliography and appraisement of the works noted. Cambridge, MA: Peabody Museum of Archaeology and Ethnology, Harvard University. Vapnarsky, Valentina, Cedric Becquey & Aurore Monod Becquelin. 2012. Ergativity and the passive in three Mayan languages. In Gilles Authier & Katharina Haude (eds.), Ergativity, valency and voice, 51–110. Berlin: Mouton de Gruyter. Vinogradov, Igor. 2017. From enclitic to prefix: diachrony of personal absolutive markers in Q’eqchi’. Morphology 27. 105–122.

Zarina Estrada-Fernández

20 Grammaticalization in Uto-Aztecan languages from northwestern Mexico 1 Introduction 1.1 Grammaticalization Canonical approaches to grammaticalization (Lehmann 2002; Heine, Claudi and Hünnemeyer 1991; Heine and Kuteva 2007) assume that lexical elements change into grammaticalized elements, which usually have shorter linguistic forms. In contrast, the data discussed in this paper shows that in Uto-Aztecan languages (UA) from northwestern Mexico, grammaticalization starts out from less grammaticalized elements to more grammaticalized elements. The languages under study belong to two distinct branches: the Taracahitan branch, which includes Yaqui (YQ), Guarijio also called Warihío (GU), and Tarahumara (TAR); and the Tepiman branch, with Tohono O’odham (O’O), Pima Bajo (PB), Northern Tepehuan (NT) and Southeastern Tepehuan (ST) as the representative languages (see a map and a simplified genealogical tree [Figure 1 and Figure 2]).1 While discussing data from these languages, sometimes I also mention two extinct languages Névome, and Ópata. Névome is a historical ancestor of Pima Bajo and Ópata is part of the Ópatan-Eudeve branch (Dakin 2004).2 Data from these languages show that the four parameters of grammaticalization, as defined in Heine and Kuteva (2007: 33–46), extension, desemantization, decategorialization and erosion, cannot always be supported by the phenomena of change observed in them, or that the application of these parameters is only partially valid. This paper aims to identify three different processes of grammaticalization: (i) inherited, or shared grammaticalization, in which linguistic forms can be clearly traced back to the protolanguage (i.e., linguistic change can be explained in terms of the historical evolution of the languages within a linguistic family); (ii) contactinduced processes of grammaticalization, in which processes of change or grammaticalization are due to language contact as a result of a socio-politically complex

1 Guarijio (Warihío), Glottocode huar1255, ISO 639–3var; Northern Tepehuan, Glottocode nort2959, ISO 639–3ntp; Pima Bajo, Glottocode pima1248, ISO 639–3pia; Southeastern Tepehuan, Glottocode sout2975, ISO 639–3stp; Tohono O’odham, Glottocode toho1245, ISO 639–3ood; Tarahumara, Glottocode cent2131, ISO 639–3tar, and Yaqui, Glottocode yaqu1251, ISO 639–3yaq; Ópata, Glottocode opat1246, ISO 639–3opt and Névome, Glottocode pima1248, ISO 639–3nev. 2 Most of the data from Pima Bajo Bajo, Tarahumara, Northern Tepehuan and Yaqui, come from my own field notes. For ethical reasons, the orthographic conventions accepted by the Yaqui and the Tarahumara communities are used on data provided for these languages. https://doi.org/10.1515/9783110712735-006

854

Zarina Estrada-Fernández

Fig. 1: Map of Uto-Aztecan.

event (i.e., colonization), or as a result of sharing boundaries with languages that are spoken within the same geographical area even though they originated from different families, and finally, (iii) processes of change that cannot be attributed to any of the two previously mentioned process and thus are considered to be the result of universal principles of language change. These last types of processes of linguistic change differ from the inherited process in that the latter may be observed in several languages of the same family, while the former are not necessarily attested within languages of a particular language family.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

855

Fig. 2: Simplified genealogical tree of Uto-Aztecan languages.

1.2 Sonoran languages: valley, desert and mountains The languages for this study were chosen for their geographic location and traditional geographical settlements, which is the northwestern region of Mexico. The speakers of Tohono O’odham and Akimel O’odham (formerly Papago and Pima) inhabit desert areas in southern Arizona in the US, while the speakers of the Mexican variety of O’otam live in north-central Sonora in Mexico, also in desert lands. The Tarahumara, Guarijio, Northern Tepehuan, and Pima Bajo occupy the highlands of the Sierra Tarahumara (Northern area of the Sierra Madre Occidental), while the Southeastern Tepehuan speakers inhabit the southern and southeastern region of the south of Durango highlands of the Sierra Madre Occidental. In contrast, the traditional settlements of the Yaqui are valley agricultural lands of central Sonora. The Guarijio is commonly referred to with that name if the speakers live in

856

Zarina Estrada-Fernández

the highlands in the State of Chihuahua, Mexico, but with the name of Warihío if they inhabit the valley lands of Sonora. Concerning the number of speakers, the most recent data3 mention that there are about 73,856 speakers of Tarahumara, 9,568 Northern Tepehuan, 2.088 Guarijio/Warihío, 743 Pima Bajo, 112 O’otam,4 20,340 Yaqui and 36,343 Southeastern Tepehuan.5

1.3 Basic typological properties and differences Uto-Aztecan languages from northwestern Mexico show a certain degree of diversity and can’t be conceived as a group of languages with recurrent similar typological features. In this section, I will only refer to a handful of typological properties in order to provide a glance at the internal differences that characterize the group of languages addressed in this paper. As for basic word order, the proto-Uto-Aztecan is reconstructed as SOV by Langacker (1977), and this order is well attested in Yaqui, but O’otam (as well as Tohono O’odham), Pima Bajo and Tarahumara tend to have a certain degree of free word order . In contrast, Northern Tepehuan and Southeastern Tepehuan show VSO word order, and Guarijio (Warihío) is changing from an SOV to a VSO language. case marking on nouns and adjectives is only found in Yaqui. Moreover, Yaqui shares the property of having case marking (nominative and accusative) on determiners and demonstratives a behavior that is also observed within Pima Bajo. All these languages also have oblique case marking morphology, as is discussed in section 2.3. All languages have distinct pronominal forms for marking subject vs. non-subject. Yaqui is the only language with a more fine-grained pronominal system marking subject, direct object, indirect object and reflexive vs. possessive pronouns. Finally, only Yaqui, Tarahumara, and Guarijio have passive morphology on the verb.

2 Grammaticalization of nominal categories 2.1 Number Langacker (1977: 80) observes that UA languages distinguish between singular and plural nouns, most of them animate, by means of two distinct patterns: the Proto 3 INEGI (2015). 4 The number of speakers considers only the O’otam people living in Mexico, not in the US. Moseley (2007) mentions about 15,000 speakers of Tohono O’odham and Akimel O’odham, in the US. 5 Southeastern Tepehuan speakers inhabit lands that are on the boundaries of two distinct cultural areas, Aridamerica and Mesoamerica. Traditionally, linguistic studies have neglected the distinction between these two different cultural areas, and Aridamerica is considered either part of Mesoamerica or of North-America.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

857

Uto-Aztecan (PUA) suffix *mi, and reduplication. Later on, I will only discuss the suffix marker, since this marker is relevant for the topic of grammaticalization. UA languages from northwestern Mexico in general show the following scenarios of change: (i) Yaqui still has the inherited suffix -(i)m; (ii) Pima Bajo uses this suffix only in a handful of collective (or mass) cultural nouns, as well as speech act participants (SAP), and third-person pronouns; (iii) Tarahumara has no systematic marker for pluralizing nouns (Brambila 1953; Alvarado García 2007),6 (iv) Northern and Southeastern Tepehuan, including Pima Bajo, pluralize nouns by reduplication, a property which is also observed in UA languages from the US, and finally, (v) Northern Tepehuan shows plural marking also on SAP pronouns, but not on nouns. Thus, the route of grammaticalization of the plural suffix -(i)m, seems to obey the hierarchy of animacy, that is, it started on the pronouns, as in (1).7 Observe that for the first person plural pronoun, the suffix -(i)m is redundant, since the hypothesized basic form, *aat, is plural in nature. (1) a. Pima Bajo *aat 1pl aap 2sg hig 3sg b. Northern Tepehuan * aat 1pl aapi 2sg

aat-im aap-im higa-m

1pl 2pl 3pl

aatimi aapimi

1pl 2pl

The second stage of grammaticalization is observed when the plural suffix -(i)m is extended to plural animate nouns, like those in (2a), and finally to inanimate mass nouns, abstract or collective cultural nouns as in (2b–c). (2) a. Yaqui b. Yaqui

usi-m jamuch-im achi-m baajipo’oko-im aina-(i)m aros-im

‘children’ ‘women’ (sg. jamut) laughter’ ‘ritual beverage’ ‘wheat’ (Sp. harina) ‘rice’ (Sp. arroz)

6 According to Givón (2009), Ute, a Uto-Aztecan language from the Numic branch spoken in the US, marks plural in nouns by three mechanisms: the suffixes -u, -mu, and reduplication. Hill (2005: 177) also identifies the suffix -im, in Cupeño. For Chemehuevi, Press (1979) mentions three productive plural mechanisms for pluralizing nouns: the suffixes -wi and -mi, the later limited to animate nouns, and reduplication. 7 Daniel (2005) mentions that plurality occurred first in personal pronouns and later it spread to other nominals.

858

Zarina Estrada-Fernández

c. Pima Bajo

‘tortilla’ ‘ribs’

tɨmit-im hon-im

The grammaticalization chain in (3) illustrates the three different stages presented above: (3) Plural SAP pronouns

>

Plural animate nouns

>

Mass/Collective nouns

2.2 Possession Miller (1980) and Langacker (1977) describe two distinct strategies for expressing possession, a lexical and a morphological one. The former is based on a possessive ‘pet’ classifier and the latter on different markers derived from the suffixes *-wa and *-ya in PUA, whose origins can be traced back to several verbs within “the semantic range of be/have/do” (Langacker 1977: 39).

2.2.1 The possessive ‘pet’ classifier in UA A PUA classifier for ‘pet’ possession, *puku ‘pet, slave’ or ‘adopted person’, was identified by Miller (1980) in several UA languages. A cognate form of this classifier occurs in all the Taracahitan languages (i.e., Tarahumara, Guarijio and Yaqui). The Tepiman languages, in particular, Pima Bajo, Northern Tepehuan, and Southeastern Tepehuan, show the same kind of classifier, although with a different lexical form. For these languages, I assume a cognate *so’/soi as the basic form. The interesting topic for the discussion of grammaticalization is whether the classifier derives from a noun or from a verb. Comparative evidence in the Uto-Aztecan languages from northwestern Mexico support a path of grammaticalization with two possible results, a classifier or a noun, as in (7). I will first discuss data from the Taracahitan languages (Yaqui, Guarijio/Warihío, and Tarahumara), and then from the Tepiman languages (Pima Bajo, Northern and Southeastern Tepehuan). In Yaqui, the cognate form of the pet possession’s classifier appears with a verb meaning ‘to raise/have cattle’, as in (4), but it can also denote a young creature, either a human or an animate being who needs care, if it is nominalized, as in (5):8 (4)

Yaqui Joan wakas-im buke. John cow-pl raise.cattle ‘John raises cows.’

8 The root buk- also occurs in some derived nouns, as in buk-wa-me ‘domestical animal’, from have-imprs-agtvz.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

(5)

859

a. U buk-i bwia-po yeewe. det.sg.nom young.creature-nmlz dirt-loc play.prs ‘The young (human) creature plays in the dirt.’ b. U kaba’i buk-i kora-po weyek. det.sg.nom horse young.creature-nmlz corral-loc play.sg.prs ‘The pony plays in the corral.’

In Guarijio/Warihío, although the cognate form is present, its occurrence shows that this element is losing its function since it is currently possible to convey the same meaning without using the classifier. Examples in (6) illustrate the ‘pet’ classifier in (6a) and the alternative construction without it in (6b):

(6)

Guarijio/Warihío (Félix Armendáriz 2007: 51) a. no’ó puhku-wá čuhčúri. 1sg.nsbj class-poss dog ‘my dog’ b. no’ó čuhčúri-wa. 1sg.nsbj dog-poss ‘my dog’

Thus, Miller’s (1980) proposal can be reoriented to claim that the pet classifier is the result of a chain of grammaticalization that started from an old verb of possession, as attested in Yaqui, and after that, the verb lost this function (i.e., decategorialized) to change either into a noun as in Yaqui or a classifier, as in Guarijio/Warihío. The grammaticalization chain in (7) supports both paths of change:

(7) raise/have (cattle)

>

classifier for animal possession noun

Data from Tarahumara (Guadalaxara 1683, Steffel 1799) confirm this view. In this language, both uses of this element, either as a verb or as a noun occurred in the language. In Tarahumara, the cognate form for *puku is buku, with a voiced labial stop, which means ‘one’s property, slave’ and usually functions as a noun. Moreover, as a noun, it is restricted (specified) to refer to any domestic animal that a person possesses when the exact denotation is clear from the discourse context. According to this view, buku has been restricted to name a ‘beast’ (domestic animal), in particular, a synonym for ‘horse’. Thus, Tarahumara goes a step ahead in comparison to Yaqui and Guarijio/Warihío.

860

Zarina Estrada-Fernández

(8) Tarahumara bukú a. bucú bucúcu bupúgu bucúrúga buquéigamec bucúrú ~ bucurúqui (Guadalaxara 1683)

‘slave’9 ‘horse’ (-cu is the stative suffix) ‘horses’ (plural form) ‘beast’ (-ruga is the instrumental suffix) ‘owner’, ‘sir’ (a nominalized complex form) ‘to take as a slave’, ‘to take as a beast’ (i.e., animal)

b. bucú generic name for ‘beast’, ‘animal’ Pedro bucúla ‘el caballo de Pedro’ (‘Peter’s horse’)

(Steffel 1799) (Guadalaxara 1683)

c. we risió rútzane kému bukú Lírio asegá.10 int tired ride 3sg.poss horse Lirio say ‘Lirio says that he gets very tired from riding his horse.’ (Brambila 1976: 93) The languages of the Tepiman branch show a different scenario. Although the languages do have a classifier for pet possession, the basic lexical form is different. In (9) an example from Pima Bajo is provided, and (10) comes from Northern Tepehuan. In the last language, the classifier for pet possession has been lexicalized adding the alienable suffix -g(a). Thus, the classifier in Pima Bajo occurs as soig, and in Northern Tepehuan as soiga.11 (9)

Pima Bajo soig: huaan gogos in soig mua. John dog 1sg.nsbj class kill.pfv ‘John killed my pet dog.’

(10) Northern Tepehuan soiga: gin soiga gogooši muu. 1sg.poss class dog die.pfv ‘My pet dog died.’ Finally, data from Southeastern Tepehuan from Willett (1991) in (11) confirm the presence of the classifier for pet possession as well in this language:

9 Data from Guadalaxara occur in the traditional orthography of colonial documents, where /k/ is written as . 10 Glosses of the example from Brambila are mine. 11 I have no relevant examples for the classifier ṣoiga from Tohono O’odham, but see Bahr (1986).

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

861

(11) Southeastern Tepehuan (Willett 1991: 16) a. hum ’šoiʔ.12 2sg.poss class ‘Your domesticated animal’ b. hiñ šoiʔ. 1sg.poss class ‘My domesticated animal’ Note that the main argument to claim the verbal origin of the classifier comes from data provided by García (p.c.). The example in (12) shows a verb so’m ‘take care, possess’, with a final consonant /m/ which is probably associated to a durative basic verbal form.13 (12) Southeastern Tepehuan Bhai’ ñich ja’p ba-tibhiich gu ta~toxkolh dir 1sg.sbj dir compl-guide art pl~pigs tu-so’m-dhi-dha’-iñ. dur-take.care-appl-cont-1sg.sbj ‘I guide the pigs in such a direction (since) I used to take care of them.’ The chain of grammaticalization provided in (7) concentrates the two grammaticalization processes observed in the Uto-Aztecan languages from northwestern Mexico, that is, a verb that changes into a classifier or a verb that turns into a noun; its realization, however, is based on two different verbs: (i) Inherited grammaticalization where the proto-form is preserved: Tarahumara, Guarijio, and Yaqui. (ii) Semantic specification with a distinct form: Pima Bajo, Northern Tepehuan, and Southeastern Tepehuan.

2.2.2 Possessive suffixes *wa and *ya (Langacker 1977: 88) Two relevant markers of possession for the languages addressed here are *wa and *ya/*yi. It is possible to hypothesize that these two makers descend from different verbs of possession or existence and that later, they became reduced and restricted

12 Willett (1991: 13) recognizes that the consonant [š] is an allomorph of /s/. 13 Willett and de Willett (2013: 154) document the verb soi’nta’ ‘to tame’ as well as three other alternative tempo-aspectual forms: saso’inta ‘completive.pl’, soi’nta ‘preterite’; so’intar ~ saso’intar ‘present’. The verb form so’m, documented by García (p.c.), has dropped the vowel /i/ and changes the final consonant from a nasal-alveolar /n/ into a bilabial /m/.

862

Zarina Estrada-Fernández

to mark, respectively, alienable and inalienable possession (Dakin 1991: 298). However, Alvarez Gonzalez and Muchembled (2013: 16) show that the distinctive functions of these markers are currently only observed in languages of the Tepiman branch. Contrastingly, the Taracahitan languages have lost such distinct functions. Thus, Tarahumara uses both markers for either, alienable or inalienable possession, and Yaqui, having lost the marker ya/yi, has grammaticalized the marker wa meaning inalienable possession. In the following subsections, I will address each marker separately. 2.2.2.1 The PUA suffix *wa Data from the Guarijio/Warihío, Pima Bajo, Northern and Southeastern Tepehuan validate the grammaticalization pathway that starts in a hypothetical verb of possession and later grammaticalizes into a possessive alienable marker as in (13): (13) possess > possessive alienable marker Examples in (14) illustrate the suffix -wa encoding alienable possession in Guarijio/ Warihío: (14) Guarijio/Warihío (Félix Armendáriz 2007) a. no’o čuhčúri-wa 1sg.nsbj dog-al ‘my dog’ b. no’o kari-wa weruma 1sg.nsbj house-al big ‘my big house’ In the Tepiman languages, the cognate form *wa occurs as -ga, and is also restricted to mark alienable possession. The suffix -wa supports the grammaticalization change described in (13), where a hypothesized verb of possession grammaticalizes into a marker of alienable possession. Examples in (15) are from Pima Bajo and Northern Tepehuan. No good examples of Southeastern Tepehuan are available to illustrate the alienable possessive marker -ga, although Willett (1991: 53) recognizes this function for the suffix -ga.14 (15) a. Pima Bajo Huaan kii-ga John house-al ‘John’s house’

14 Willett (1991: 53) mentions that the suffix -ga “denotes a special status analogous, but not identical to alienability”.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

863

b. Northern Tepehuan (Carrillo 2005) go tuurra-ga. det axe-al ‘His axe’ In Tarahumara, the suffix ga is currently absent, but historically, the language has converged both historical markers, *wa and *ya/yi, into a complex element guara,15 which encodes possession. The example in (16) shows a genitive phrase documented by Guadalaxara, a Jesuit’s grammar published in 1683, where the complex possessive marker is included. The complex element guara results from the combination of the PUA *wa and the reflex of the marker of possession *ya/yi, which in Tarahumara is -ra: (16) Tarahumara (Guadalaxara 1683: book 1, chapter 1) Iuan xunùguara. ‘John’s corn’ In Yaqui, the reflex of the PUA’s marker of possession, *wa, marks inalienable possession as can be seen in (17). Therefore, the pathway of grammaticalization in (13) needs to be modified to include what is observed in Yaqui. The revisited grammaticalization chain is shown in (18): (17) Yaqui a. aapo mala-wa ibakta-k. 3sg.sbj woman-inal abandon-pfv ‘He embraced his mother.’ b. aapo jamut-wa jimaa-k. 3sg.sbj woman-inal abandon-pfv ‘He abandoned his wife.’ c. U koyo tek-wa jeoko. det.sg.sbj oyster flesh-inal be_slimy.impf ‘The oyster’s flesh is slimy.’ (18)

possessive alienable marker possess possessive inalienable marker

Further examples of all the UA languages from northwestern Mexico, including O’otam are given in Table 1.

15 Documented by Steffel (1799).

864

Zarina Estrada-Fernández

Tab. 1: The currently reflexes of the PUA possessive marker *wa. The PUA suffix –wa O’O

PB

NT

ST

YQ

GU

TAR

-ga ‘alienable’: ñ-gogs-ga ‘my-dog-al’

-ga ‘alienable’: ki-ga ‘houseal’, baiba-ga ‘necklace-al’

-ga ‘alienable’: tupúrra-ga ‘axe-al’; giñ báakiga ‘my house-al’

-ga’n ‘alienable’:16 sasoi-ga-’n ‘animals’ pieesta-ga’n ‘its fiesta (party)’17

-wa ‘possessed object’, majka-wa ‘maskinal’

-wa ‘possessed object’: kari-wa ‘houseposs’, weé-wa ‘my land’, čuhčúri-wa ‘dog.poss’

The suffix is currently absent. The language has a complex particle guara18

2.2.2.2 suffix *yi /*ya Contrasting with the PUA suffix *wa, the second marker of possession *yi/*ya, shows a different grammaticalization chain. In (19), the hypothesized verb of possession changes first into a possessive marker in nouns, and after that, it is restricted to encode inalienability (inherent possession, observed in body-parts, kin names and in whole-part relations). Later, in some of the languages under study, the suffix is restricted to third person only. Table 2 illustrates the different cognates of the possessive marker *yi/ *ya in the Taracahitan and Tepiman languages. The hypothesized grammaticalization pathway of this marker is the following: (19) possess > possessive marker > inalienable/inherent possession > 3rd person kin terms possessive body-parts whole-part relation In Guarijio/Warihío, a whole-part relation occurs with the reflex -ra of the PUA possessive marker *yi/*ya, in (20a), while marking inalienable possessive marker in (20b–c):

16 According to Willett (1991: 211), the possessor suffix -d, which derives from the suffix *ya (§ 2.2.2.2), changes to -’n in final-syllable position. This author (1991: 204, ex. 596), also refers to this suffix as ‘status relation’. In Pima Bajo the suffix -di changes first into a palatal [j] as is observed in O’odham and later by rhotacism into an -r (Estrada-Fernández 2014). 17 García-Salido p.c. 18 Estrada-Fernández and Grageda Bustamante (2014) pointed out that Guadalaxara (1683) gives two distinct genitive forms: ra and guara. However, Steffel (1799) divides the last one into two different suffixes, gua and rá, the latter with an alternative allomorph lá.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

865

(20) Guarijio/Warihío (Félix Armendáriz 2007) a. weruma ko’ore tehte-ra big fence stone-inal ‘the stone of the big fence’ b. wakasi aawa-ra cow horn-inal ‘the horns of the cow’ c. hustina čuhčuri-wa mo’o-ra Agustina dog-poss head-inal ‘Agustina’s dog’s head’ In Tarahumara the reflex of the PUA marker of possession *yi/*ya is -ra or -la, it means either alienable or inalienable (inherent) possession. (21) Tarahumara a. nejé karí-ra kítira na natá-ri. 1sg house-poss about loc think-pfv ‘I thought about my house.’ b. nejé raícha-nare nejé raíchaa-ra 1sg talk-want.pfv 1sg language-poss ‘I want to talk in my language.’ c. okó sawá-la pine leaves-poss ‘Pine’s leaves.’ Inalienable possession, in contrast, is conveyed by using niwara ‘property, to have’, as in (22): (22) Tarahumara nejé niwára sunú 1sg property/to have corn ‘I do have corn.’ In relation to Yaqui, recall that in section 2.2.2.1., I mentioned that this language has lost the reflex of the marker of possession *yi/*ya, and that it is the marker -wa, which has been grammaticalized to signal inalienable possession, as shown in (17). Two of the Tepiman languages, Pima Bajo, in (23), and Northern Tepehuan, in (24), show the reflex -di/-r encoding inherent or inalienable possession. Furthermore, in at least three of the languages of this branch, Pima Bajo, Northern Tepehuan and Tohono O’odam, the suffix is now being used as a general marker of

866

Zarina Estrada-Fernández

possession, so it is common to find it placed after the alienable suffix -ga. The new change is observed in (25) as well in some of the examples provided in Table 2. (23) Pima Bajo a. toskil naaka-r pig nose-inal ‘the pig’s nose’ b. divor tiika-r land owner-inal ‘the owner of the land’ In Northern Tepehuan the PUA marker of possession emerges as -dyi: (24) Northern Tepehuan a. igai baaki-d yi dem house-inal ‘her/his house’ b. igai bakuri-d yi dem breeches-inal ‘her/his breeches’ (25) Pima Bajo Marii iip-ga-r Mary skirt-al-poss ‘Mary’s skirt’ Finally, concerning the inalienable possessive marker, in Southeastern Tepehuan, the suffix -’n is the reflex of the PUA marker *yi/*ya, and according to Willett (1991: 203) it means the obligatory inalienable possession, as in (26): (26) Southeastern Tepehuan (Willett 1991: 204, ex. 594) Mi’ pui’-p jum-avasar gu mara-’n ja’l na gu táta-’n.20 there thus-rep rflx-behave art offspring-inal like sub art father-inal ‘The child behaves like his father.’ Table 2 summarizes the current situation in the Uto-Aztecan languages from northwestern Mexico.

19 The suffix in Northern Tepehuan has palatalized the consonant /d/. 20 I have adapted some of Willett’s glosses.

867

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

Tab. 2: The possessive inalienable PUA suffix *yi/*ya. The PUA suffix *yi/*ya O’O

PB

-j ‘3sg.poss’: -di > -r wisilo-ga-j ‘possessive’: ‘calf-al-poss’ mo’o-r ‘its head’ ki-ga-r ‘house.al. poss’

NT

ST

YQ

GU

TAR

-di ‘3sg.poss’: mo-motiriga-di ‘metate-alposs’, bonamuga-di ‘hatal-poss’

-r ‘inherent possessive’: mo’o-r-am head-possloc ‘on its head’

No occurrence of this suffix

-ra ‘third person possessive’ or a partwhole relation and inalienable possesion: mo’ó ‘head’, mo’ó-ra ‘headposs’, ka’aká ‘sandals’, ka’aká-ra ‘sandalsposs’

-ra ‘possessive’: sa’apa ‘flesh’, ‘meat’, sa’apa-ra ‘dog’s flesh’. kaši ‘leg’, kaši-ra ‘leg.poss’

The summary of the grammaticalization of the PUA possessive marker *yi/*ya is the following: (i) Inherited grammaticalization; the suffix reflex is used as a general possessive/ genitive marker: Pima Bajo and Tarahumara (ii) Inalienable possession: Tohono O’odham, Pima Bajo, Northern Tepehuan, Southeastern Tepehuan (iii) Specification (semantic reduction), restricted to mark third person possession: Guarijio, and Tohono O’odham (iv) Suffix loss: Yaqui

2.3 Case Langacker (1977: 82) identifies three different accusative suffixes in PUA languages: (i) an archaic accusative suffix *-ci, (ii) a “newer but still archaic” accusative suffix *-kV, and (iii) a regular accusative suffix *-a. He also identifies the sequence *-t-a (an absolutive21 suffix followed by the regular accusative suffix) which in several

21 The term absolutive within the UA linguistic literature refers to the absolute form of a lexical item that is not in its possessive form. Cognates of the absolutive suffix are found in Nahuatl, -tl ~ -t ~ -l, in Guarijio -t, and in the case suffix in Yaqui -ta.

868

Zarina Estrada-Fernández

UA languages has been analysed as an accusative suffix -ta. Among UA languages from northwestern Mexico, only Yaqui, whose object case suffix is -ta, has inherited the accusative morphology. No other language possesses case markers in nouns, although, as will be mentioned below, a cognate form of the *-kV suffix may occur in object/accusative determiners or adjectives. Concerning the processes of change that I discuss in this section, some of them may be considered to be cases of specification, that is, a semantic change, where a sort of specificity affects the meaning of a linguistic element, and, as Kuteva (2004: 36–37) suggests, I also believe that specification should not be excluded from grammaticalization.

2.3.1 The PUA suffix *-ci The PUA suffix *-ci, is still present in Tarahumara and Guarijio (Warihío), where it functions as a locative marker, as in examples in (27). (27) a. Central Tarahumara échi rejói sunú apeéru kabái-chi. det man corn load.impf horse-loc ‘The man loads up the corn on the horse.’ b. Guarijio/Warihío (Félix Armendáriz 2007: 16) no’ó čapaká mesa-čí pói. 1sg-nsbj leg table-loc lay_down.sg ‘My leg is on the table.’ A cognate form of this suffix is also observed in Yaqui, where the suffix -chi (in its orthographic form) also encodes a locative function, as in (28): (28) Yaqui Joan jabe-tat nooka?… ae-chi. John inter-loc talk.impf 3sg.acc-loc ‘Who is Johni talking about? […] About himj’. Burgess (1984: 59) has also documented the suffix *-ci in Western Tarahumara. Although Burgess labels the suffix as an accusative case maker, the suffix encodes different functions. Examples in (29) show the suffix -’či marking a patient, as in (29a), a recipient, as in (29b), a possessor, and with a minor phonological change a locative, as in (29c).

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

869

(29) a. Western Tarahumara čanigá go’-mé olá mué ne-’čí. why eat-fut do 2sg.sbj 1sg-acc ‘Why are you going to eat me?’ b. ne-’čí rari-gí. 1sg-acc sell-imp ‘Sell it to me!’ c. nabá ne-’čí behte-la-čí. let’s 1sg-acc live-inst-loc ‘Let’s go to my house!’ Further evidence from Tarahumara of Pahuírachi (Central Tarahumara), illustrate that the suffix -chi (in its orthographic representation) marks an oblique argument of the verb majawá ‘to fear’, as in (30). Although, the most frequent function of this suffix is the locative. (30) Tarahumara (Estrada-Fernández 2011–2012) échi rejói nalioó-chi majawá.22 det man wolf -obl fear ‘The man fears the/at/of wolf.’ The data related to the suffix *-ci, argue for a grammaticalization chain starting from a locative that changes into an oblique case marker including the primary object (accusative/dative), as is currently observed in Tarahumara, in (31): (31) Locative > oblique case marker23

2.3.2 The PUA *-t-a case marking sequence The PUA *-t-a case marking sequence, where *-t is the absolutive suffix and *a, the accusative, occurs in Hopi (Hill 1998: 871), an Uto-Aztecan language that is spoken in Arizona, in the US. In Hopi, the suffix -(t)a marks the accusative/object case: (32) a. manna sipmasmi-t tuwa. girl silver_bracelet-acc found ‘The girl found a silver bracelet.’ 22 The use of this suffix to encode an accusative-object needs to be tested with other semantic verb forms, since it does not always occur with all semantic classes of transitive verbs. 23 The oblique case marker may include something similar to an ablative, a dative, a primary object and probably other oblique grammatical relations.

870

Zarina Estrada-Fernández

b. kur nu’ ung hìi-ta tuuvingta. please I you something-acc ask ‘Let me ask you something.’ Among the Uto-Aztecan languages from northwestern Mexico, the suffix -ta is only present in Yaqui, and characterized by several authors as an accusative case marker (Dedrick and Casad 1999; Guerrero and Van Valin 2004; Tubino Blanco 2010), as in (33): (33) Yaqui -ta: kuta ‘stick’, kuta-ta ‘stick-acc’, chu’u ‘dog’, chu’u-ta ‘dog-acc’ However, further analyses of this language allow us to claim that the suffix is in fact, polyfunctional and that such property is the result of a gradual process of grammaticalization. Examples in (34) show the different functions of the suffix -ta, among them: (i) as an accusative case marker (34a), (ii) as a double object marker (34b) – a behaviour that is observed only in a limited number of verbs: majta ‘teach’, maka ‘give’, miika ‘gift’, reuwa ‘lend’, and u’ura ‘remove’; (iii) as a genitive or a possessor marker in possessive constructions, as in (34c–e), and finally, (iv) as an object of a postposition (34f–g). (34) Yaqui a. Accusative case marker U uusi sakusi-ta bwa’e. det.sg.nom child.nom pinole-acc eat.prs ‘The child is eating pinole.’ (Estrada-Fernández, Tubino, and Villalpando 2015) b. Double object case marker Joan soto’i-ta Maria-ta maka-k. Juan.nom pot-acc Maria-acc give-pfv ‘Juan gave Maria the pot.’ (Estrada-Fernández, Tubino, and Villalpando 2015) c. Dependent marker in possessive constructions jamut-ta tajo’ori woman-gen cloth ‘woman’s cloth?’ (Martínez Martínez 2015: 63) d. totoi-ta kabam hen-gen eggs ‘chicken’s eggs’ (Estrada-Fernández, Tubino, and Villalpando 2015)

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

871

e. chu’u-ta ota dog-gen bone ‘dog’s bone’ f. Object of a postposition U yoeme juya_kowi-ta in-sai-ta-mak aamu-k. det.sg.nom man wild_boar-acc 1sg.pos-brother-obj.pp-com hunt-pfv ‘The man hunted a wild boar with my brother.’ g. juya_miisi juya-ta betuk bo’oka koche wildcat tree-obj.pp under lie.stat sleep.prs ‘The wildcat sleeps under the tree.’ Evidence from Yaqui is useful to support a grammaticalization chain for the suffix case marker *ta. This path of grammaticalization begins to favour a change where the suffix first starts as a genitive case marker, then it covers some oblique case domains to end as an accusative marker. This chain is shown in (35): (35) genitive case marker marker

>

oblique case marker

>

patient (accusative) case

2.3.3 The PUA accusative suffix *-kV This case marker occurs in Yaqui either as -ka in accusative determiners or as -k, in adjectives. (36) Yaqui inepo ju-ka siali-k bicha-k 1sg.nom det-acc green-acc see-pfv ‘I saw the green one.’ For Guarijio/Warihío, Félix Armendáriz (2007) also documents the suffix -ka in accusative determiners, as in (37): (37) Guarijio/Warihío (Félix Armendáriz 2007) ihí čučúri ki’kí-ru pu’-ká kuitá. d.p dog bite-pfv.evi d.d-acc child ‘This dog bit that child.’ However, in Pima Bajo, the case suffix is only present in the non-subject form of the determiner, where it occurs phonologically reduced. The example in (38) illustrates the contrast between the subject and the non-subject determiners:

872

Zarina Estrada-Fernández

(38) Pima Bajo kil ik peloot giv. ig det.sbj man det.nsbj ball hit.pfv ‘The man hit the ball.’ Concerning the grammaticalization of the case marking element *kV, the functions of -ka suggest the grammaticalization pathway provided in (39). It supports a change where the case marker -ka occurred first with determiners, and after that in adjectives. Arguments for the grammaticalization chain offered in (39) come from the Head Preference Principle supported by Gelderen (2004: 11), which favours that determiners as heads have certain preferences, such as to bear case marking. A second argument comes from the parameter of erosion (Heine and Kuteva 2007) since it is helpful to demonstrate that the eroded form is the one that occurs with adjectives, as in Yaqui in (36). For the determiner of Pima Bajo, as in (38), the reduced form -k in the accusative or non-subject determiner seems to be natural, since, within this language, case marking is restricted to the determiners and never occurs with adjectives: (39) Determiner case marker

>

adjective accusative case marker

2.4 Demonstratives PUA languages have from two to four demonstrative forms. The demonstratives are distinguished based on the proximal vs. distal distinction (prefixed or independent forms). The proximal demonstrative is recognized as *i, and the distal demonstrative as *u (Langacker 1977: 98). Langacker (1977: 99) also refers to another set of demonstratives, which according to him probably marked a distinction between animate vs. inanimate: *pi-ma and *a-ma (*ma ‘one’). According to Langacker (1977), in PUA both proximal and distal demonstratives were inflected for case *kV acc or number *mi pl. Pima Bajo has a pair of determiners, one for the subject and another for the object, where an old PUA distinction between the i proximate vs. i distal is observed, so the organization of the set of determiners is as in (40): (40) Pima Bajo ig ‘this.sbj’ ik ‘this.obj/acc’ id ‘that.sbj’ ik ‘that.obj/acc’ The elements in (40) are diachronically derived from the demonstratives higai sg/ higa-m pl, which either function as emphatic demonstratives or as a third-person nominative personal pronouns. Some speakers pronounce the non-subject form of

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

873

the emphatic demonstrative as hikam. A reduced form of the emphatic demonstrative functions as a third person singular pronoun, (h)ig. In Northern Tepehuan, Bascom (1982: 323) identifies five demonstrative pronouns, which function also as third-person pronouns, subject or object, singular or plural. Their deictic meanings range from proximal to distal. The system is quite complex since the basic inherited demonstrative occurs with other deictic elements, as for example in the distal, gam ~ gamon, to fulfil pragmatic differentiation. Within the continuum of demonstratives in (41), I have arranged the different demonstratives forms from left to right based on the deictic function: (41) proximal id yi ‘this one’ ‘these’

>

(nearby) govai ‘that one’ ‘those’

>

(far) igai > ‘that one’ ‘those’

distal (g)imai > ‘the other one’

(g)aa ‘others’

Recursively, the basic demonstratives may derive a newer set of complex demonstratives when combined with the distal makers -gimai, and -gaa: (42) Northern Tepehuan id yi-gímai ‘this other one’ go-gimai ‘that other one’ í-gimai ‘that other one’

id yi-gaa go-gaa í-gaa

‘these others’ ‘those others’ ‘those others’

Observe that the old PUA demonstratives, i-, go- and i- occur as the basic form of the determiners. According to Miller (1996: 234), Guarijio has two subject demonstratives: ihí ‘this’, and pu’ ‘that’, in (43). Both forms can receive an oblique case suffix -ka, as in the case of pu’ka, also in (43). The oblique (non-subject) forms show their stress in the final syllable: iɁká ‘this.obl’ and puɁká ‘that.obl’ which can be lost if the demonstrative is cliticized on the previous noun. Félix Armendáriz (2007) observes the demonstrative pu without the glottal stop in Warihío del Río, as in (44a–b).24 (43) Guarijio ihí čučúri ki’kí-ru pu’-ká kuitá. d.p dog bite-pfv.evi d.d-obl child ‘This dog bit that child.’

24 Cora, a Uto-Aztecan language from the Corachol branch, also has a cognate form, pu ‘3sg.sbj’ which, according to Casad (1984: 297), is the reduced form of a causal subject definite pronoun, pu’u.

874

Zarina Estrada-Fernández

(44) Warihío del Río a. yačah-pá-sa muú=pu si-má=mu=pu tekihpána-mia. set-inch-sub 2sg.sbj=d.d go-fut=2sg.sbj=d.d work-fut ‘If you have recovered, you will go to work.’ b. no’nó nane-ré=pu no’nó eikó wa’á ohóe-re=pu. father know-pfv=d.d father then there live-pfv=d.d ‘… my father knew it because my father lived there.’ Recall that the cognate forms of the suffix -ka (Section 2.3.3) suggest a grammaticalization pathway from oblique to accusative. This marker occurs in different nominal modifiers, e.g., numerals, demonstratives, adjectives, and quantifiers (see Yaqui examples). In Guarijio/Warihío, there are two other demonstratives, but they are restricted to the oblique case. One is an alternative form of the oblique proximate demonstrative aha-, and the other the oblique distal demonstrative uɁká. Tarahumara currently has only one demonstrative, échi, which also functions as a third person singular or plural pronoun. However, in a grammar written in the seventeenth century, Guadalaxara (1683) identifies a form pu which means ‘your, yours, the, this’, and grammaticalizes as part of a PUA form *ma, which functions as a general subordinator, (45). (45) Tarahumara mapugitae ximara, ‘for them to leave’ (Guadalaxara 1683) b. mapu-kíte ko’á-ma kípi kúchu-wa. sub-purp eat-irr 3sg.poss offspring-poss ‘in order to feed her children.’ (Estrada-Fernández, 2011–2012) Yaqui grammaticalizes the determiner preserving the subject vs. object distinction: (46) a. u into wakago’ochi […] det.sg.nom and grasshopper ‘and that grasshopper …’ b. bea uka ya’utra-ta then det.sg.acc government-acc ‘thus to that government …’ Although information is not available for all the languages considered in the present study, the analysis of Pima Bajo shows a complex chain of grammaticaliza-

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

875

tion where the demonstrative is involved. In (47) the demonstrative reduces its phonological form first by losing its final two vowel segments and after that by losing its initial aspirate consonant. The reduced form ig can occur, either as a determiner or as a definite third-person pronoun. (47) higai

> (h)ig

> ig

The grammaticalization path provided in (48) shows that the demonstrative is the source of two emerging elements: a 3rd person pronoun, and a definite determiner: (48) Demonstrative

3rd person pronoun definite determiner

Whether this grammaticalization pathway occurs also in other languages remains a topic for future research. In Pima Bajo, it is possible to attest another grammaticalization route, where the demonstrative higai has changed into a relativizer -kig after being suffixed to a stative or adjectival marker -ka (Estrada-Fernández 2012: 137– 140). Example (49) shows a nominalized relative construction ikis vakin-ka ‘washing clothes’ where the verb is nominalized by means of the old oblique > accusative case marker. (50) shows the relativizer suffix -kiga, which results from the grammaticalization of -ka + higai > -kiga. (49) okosi ikis vakin-ka higai si’ gig. woman clothes wash-nmlz dem int big ‘The woman washing clothes, she is very big.’ (50) ig kilia gogos in-mua’a-kiga vuusa dem man dog 1sg.nsbj-kill-rel leave.pfv ‘The man who killed me the dog left.’ In (51) the grammaticalization pathway is given. (51) Demonstrative

>

relative clause marker

2.5 Indefinite pronouns According to Lehmann (2002: 44), indefinite pronouns have emerged from a diverse source of elements, among them interrogatives, the numeral ‘one’, or the combination one of these two types with other kinds of elements. He also documents an example from Nahuatl where the noun for ‘thing’ was combined with the old PUA demonstrative i:itlaa > tlaa ‘something’ (2002: 45). The tlaa ~ tla form is now lost

876

Zarina Estrada-Fernández

in UA languages from northwestern Mexico. Givón (1981) gives one of the earliest discussions about the grammaticalization of the numeral one into an indefinite article. The grammaticalization pathway in (52) is attested for most of the languages discussed in this study, as in (53): (52) Numeral/one

>

indefinite pronoun

(53) Guarijio (Warihío) Tarahumara Pima Bajo Northern Tepehuan Tohono O’odham Névome (†)

piipi ‘one’ bilé ‘one’ himak ‘one imóóko ‘one’ hemako ‘one’ maco ‘one’

> > > > > >

pii ‘a’ (indef.) bilé ‘a’ (indef.) hímak ‘a’ (indef.) imó, umó ‘a’ (indef.) hema ‘a’ (indef.) maco ‘a’ (indef.)

Yaqui is an exception; in this language, the indefinite pronoun comes from a grammaticalized form of an interrogative: hábe ‘anyone’, and not from the numeral seenu ‘one, another’.

3 Grammaticalization of verbal categories The morphological domain of the verbs in UA shows very few pathways of grammaticalization.

3.1 Voice/valency markers The analyses of the causative voice in languages from northwestern Mexico show that it is possible to have at least two different sources for causative suffixes. One is an utterance verb, the other a manipulative verb. The first grammaticalization chain is shown in (54). The utterance verb and source of this change could be related to a lexical verb tíídai ‘tell’, from Northern Tepehuan, in (55). (54) Utterance verb/say

>

causative

(55) Northern Tepehuan Buana pari-gami giɲ ti-tiidai Juan lazy-nmlz 1sg.sbj it-tell.impf ‘Juan calls me lazy.’ After being phonologically reduced and the suffix changes one of its vowels, a causative suffix -tuda emerges in this language, as in (56):

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

877

(56) Northern Tepehuan imo βonamui gin tai=go iʃ sabil-tuda=ni 1sg.nsbj ask.pfv=enf sub buy-caus=1sg.sbj det hat ‘He asked me to buy me to buy a hat.’ The same hypothetical verb was probably the source of the causative suffix -tar in Pima Bajo, where the hypothetical reduced form *tudai may become *tadi by losing the vowel /u/ and after that by rhoticism ended up in its current phonological form. An example of Pima Bajo with the causative suffix is provided in (57): (57) Pima Bajo Maria lii oob kooks-im-tar Maria dim person sleep-cont-caus ‘Maria is making the boy sleep.’ The causative suffix from Tarahumara -ri, illustrated in (58), may as well have its origin in the same hypothetical verb. The suffix in Tarahumara shows a still more reduced phonological form which adapts its phonological form to a common change that applies in this language by which the consonant /t/ changes into an /r/. (58) Tarahumara Juan Maria kachi-ri. Juan Maria laugh-caus ‘Juan makes Maria laugh.’ (59) summarizes the phonological changes that might have given birth to different causative suffixes that I have described for Northern Tepehuan, Pima Bajo, and Tarahumara: -ri TAR [t] ~ [r] (59) tiidai ‘tell’ NT

-tuda NT -tar PB

The second source of a causative suffix, for Pima Bajo and Yaqui, may have its origin in the manipulative verb *tuda, as it occurs in Northern Tepehuan, in (56). A reduced phonological form of that verb occurs in Pima Bajo as an independent verb of manipulation (obligation) in an analytical causative, as in (60a), or as a morphological causative, as in (60b): (60) a. Pima Bajo in-tua gogos ig 3sg.sbj 1sg.nsbj-obligate.pfv dog ‘He forced me to kill my dog.’ (Estrada-Fernández 1998: 102)

878

Zarina Estrada-Fernández

b. Pima Bajo ig timiti.m maatim-kad nan-tua det.sbj tortilla.pl nixtamal-inst prepare-caus ‘The tortillas are obligatorily prepared with nixtamal.’ (Estrada-Fernández 1998: 74) The grammaticalization of a causative suffix -tua is also observed in Yaqui where the canonical causative makes use of this suffix, as in (61): (61) a. Yaqui (Estrada-Fernández, Tubino, and Villalpando 2015: 1369) Joan Maria-ta jin-tua-k. Juan.nom Maria-acc cover-caus-pfv ‘Juan covered Maria.’ b. Yaqui (Estrada et al. 2004: 326) junak beja aman kubaji-m jijiu-tua-wa-k … until then there drum-pl it~play-caus-imprs-pfv ‘Then until the drums were made to play there …’ The grammaticalization chain which describes the source of the suffix -tua, is shown in (62): (62) Modal obligation verb

>

causative

Furthermore, in Section 3.5, I will show that the verb of manipulation *tuda also grammaticalizes in Pima Bajo and Yaqui into an evidential particle (cf. examples [92] and [93]). The passive in the UA languages of northwestern Mexico clearly divides these languages into two branches: the Taracahitan, with passive morphology at the verb, the suffix -wa in Yaqui and Tarahumara, respectively in (63a–b), or with the suffix -tu, in Guarijio/Warihío in (63c). (63) a. Yaqui (Estrada-Fernández and Félix Armendáriz 2010: 98) Ini’i waasa et-wa-k. dem.nom land.nom plant-pass-pfv ‘This land was planted.’ b. Tarahumara (Estrada-Fernández, 2011–2012) a’ bilá ko machi-wa echi na comunidad-chi ba echi bile rejoi aff evi emph know-pass det loc comunity-loc pdp det a man ‘Thus, (when he) is truly known that man by the community …’

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

879

c. Guarijio/Warihío (Félix Armendáriz 2007) kuú tehpuna-ré-tu tapaná (maniwíri-e) tree cut-pfv-pass yesterday Manuel-ins ‘The tree was cut yesterday (by Manuel).’ In contrast, Tepiman languages have no passive morphology. Tohono O’odham and Northern Tepehuan demote the agent argument by changing word order position of the arguments, that is, by changing the order position from APV into PAV or only PV. Pima Bajo and Southeastern Tepehuan have a diathetic mechanism: a non-subject pronoun intransitivizes one- or two-argument constructions. The strategy is usually labelled middle or middle-passive. In Pima Bajo, the middle-passive arises when a 3rd person object (non-subject) pronoun, either singular or plural, a= in (64), extends its function to mark middle events (cf. Kemmer 1993), for the entire pronominal paradigm. (64) Pima Bajo (Estrada-Fernández 2005) aan a=himi-va. 1sg.sbj mid=go-compl ‘I’m leaving.’ The generalized use of a non-subject (object) pronoun to express many of the middle situations in two of the Tepiman languages seems to be the result of a contactinduced change through Spanish influence, a kind of grammaticalization change that, according to Heine and Nomachi (2013: 68), is highly selective. The grammaticalization pathway provided in (65) is a variant of the reflexive > middle marker described by Heine and Kuteva (2002: 333): (65) 3rd person object/reflexive pronoun

>

middle marker

Moreover, although Southeastern Tepehuan shows a similar process, it is a 2nd person object singular pronoun, -jum, in (66a–b), that is the source of the middlepassive, except for the first person, (66c), in which case, the object/reflexive pronoun is used (cf. García Salido and Estrada-Fernández 2009). (66) Southeastern Tepehuan a. Juan jum-ti’n-cho. Juan mid-remember-term ‘John remembers.’ b. dhi’ cham jum-ga’ra. dem neg mid-sell ‘This doesn’t sell.’ (Sp. esto no se vende)

880

Zarina Estrada-Fernández

c. jiñ-gaxbo-iñ. mid.1sg-comb-1sg.sbj ‘I combed.’ The grammaticalization chain in (65) should be adapted for Southeastern Tepehuan as in (67): (67) 2nd singular object/reflexive pronoun

>

middle marker

3.2 Agreement object markers There are no agreement markers in UA languages from northwestern Mexico. However, a right extra-position of an object noun phrase carries the use of a 3rd person singular object/accusative pronoun a (orthographically written as an independent element in Yaqui) as an agreement marker. The canonical use of the 3rd person singular object/accusative pronoun marking the object/accusative pronominal argument in Yaqui is given in (68a), and the agreement marker in (68b): (68) Yaqui a. Aapo a bicha-k. 3sg.sbj 3sg.acc see-pfv ‘He saw him.’ b. Ju jamut a bicha-k ju-ka o’ou-ta. det.nom woman agr see-pfv det-acc man-acc ‘The woman saw him, the man.’ The same grammaticalization pathway is observed in Pima Bajo where a numberneutral, singular/plural 3rd person non-subject pronoun a= is used both as a pronominal argument marking (69a), and as agreement marker (69b). In both cases, the pronoun cliticizes to the left edge of the verb. The agreement marking is probably motivated by the pragmatic function of focus. (69) Pima Bajo a. kova uus-kad a=son-ia. neg.enf stick-inst 3sg.obj=hit-prob ‘Don’t hit it with the stick.’ b. aan a=dɨɨn-va ɨk viv. 1sg.sbj agr=smoke-compl det.nsbj tobacco ‘I have already smoked tobacco.’

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

881

The grammaticalization process of these pronouns is shown in (70): (70) 3sg/pl object pronoun

>

agreement marker

3.3 Tense-Aspect The functional domain of tense-aspect in UA languages from the highlands of northwestern Mexico has been poorly described. Moreover, the descriptions are diverse, since some authors describe these markers as tense, others as aspect. This situation makes it difficult to provide a general overview (at-a-glance) of what is happening in these languages in terms of grammaticalization. In order to show the contrasting patterns of grammaticalization observed among these languages, a brief comparative analysis of the Aztecan/Nahuatl language vs. the Sonoran languages was prepared, in particular concerning the form and locus of the T/A markers. The scenario for Nahuatl is the following: present tense is signalled by the absence of a suffix (zero marking). This is a highly pervasive pattern in other UA languages; however, present tense is also assumed to be the unmarked form, and it can be referred to as an infinitive, a nominalized form, or the imperfective aspect. The past imperfective tense in Nahuatl is conveyed by the suffix -ya, as in ninemi-ya ‘I was living’, ‘I used to live’ (Sullivan 1988: 49). Tarahumara is unique in showing a suffix -ya apparently encoding imperfective/durative aspect, but this suffix also occurs in nominalized verbal complements. The past perfect tense in Classical Nahuatl is encoded by means of four mechanisms (Sullivan 1988: 50). Examples in (71) illustrate one of the mechanisms that consists of dropping the stem-final vowel; in all the mechanisms, the perfect prefix o-, occurs in front of the verb stem. Examples in (71) show the subject pronominal prefixes preceding the verb stem, coch ‘to sleep’, with ni- ‘1sg.sbj’, ti- ‘2sg.sbj’, ti- ‘1pl.sbj’, and an ‘2pl.sbj’, and zero for third singular or plural persons. The plural pronouns show the suffix que in final position. (71) Classical Nahuatl onicoch ‘I slept’ oticoch ‘you slept’ ococh ‘he/she slept’ oticochque ‘we slept’ oancochque ‘you (pl.) slept’ ocochque ‘they slept’ None of the UA languages from northwestern Mexico, except Southeastern Tepehuan, uses prefixes to encode tense-aspect. However, in Tohono O’odham, the future tense occurs as a second position clitic, which may also encode person, number, modality or polarity. The future marker is written as an independent particle in (72).

882

Zarina Estrada-Fernández

Tab. 3: Tense-aspect morphology.

pfv/past

O’O

PB

NT

ST

YQ

Truncation25

Truncation Truncation Truncation -k pfv, -n past -cont

remote/ -d, -da imperfective

-kad/ -tad/ -tat

-tadai, -katadai

compl

va/-va

=ta, =t yki, =tyi

(–)va real -su

-im

-mi

-im

cont/prog

-him

durative

stative

-ka

-ka

-ka

inc/become

-li

-tu, -t yu

prospective/ wo, o future

-(h)a(g)

-mu

probability

-ia

GU

TAR

-re

-ri, -li, -ki

-i (-ka + -i) rem-stat

-sim

-da

-tu pastdur or inch

-ca26

-ka -ka/-ga impf.prtc

-a’, -’

-tu

-pa

-nee

-mea ~ -mia

-ka

-ma ~ -mea

(72) Tohono O’odham (Papago in Zepeda 1983: 73)27 Hegai o’odham ’at o jeːñad g sigal. dem person aux fut smoke det cigarette ‘That person will be smoking the cigarettes.’ Névome or Pima Bajo from the 18th century (Smith 1862) shows a preverbal particle va, (73a), to encode the completive aspect (translated as ‘already’). This marker may also occur suffixed on the verb, as in (73b). (73) Névome (Smith 1862; Shaul 1986) a. va=t'-igui usi-ab-cada io t'-igui padre divia. already=pfv-mod plant-time-rem fut pfv-mod priest arrive ‘It will already be time for planting when the priest has arrived.’

25 Truncation stands for a reduced form of the verb, by which the last syllable of the verbal root is usually omitted. 26 Following a Spanish convention, Willett (1991) writes this suffix as ca. 27 Glosses are provided by me.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

883

b. ni-coco-guia ni-gubu-daga hukio-va-himu 1sg.nsbj-illness-prog 1sg.nsbj-drop-nmlz finish-compl-cont ‘My illness surely is finishing me completely off.’ The rich TAM morphology in UA languages from northwestern Mexico is summarized in Table 3. Not much is possible to propose in terms of grammaticalization. However, a few pathways of grammaticalization for two of the morphemes in Table 3 (signalled in boldface), provide support for typological trends; these are discussed next.

3.3.1 Verb of movement to aspect In Tohono O’odham and Pima Bajo, the verb him(ia) ‘to go’ grammaticalizes as the continuous aspect -im, and in Southeastern Tepehuan the cognate -m is currently a desiderative suffix: (74) a. Verb of movement/go > continuous/progressive aspect >

desiderative

b. Tohono O’odham (Saxton 1982: 162) him-him28 go-prog ‘ambling, wandering’ c. Pima Bajo aan vihig-im-va. 1sg.sbj starve-cont-compl ‘I’m totally hungry.’ d. Southeastern Tepehuan (Willett 1991: 172) súsac-añ xi-’ádau no’-ñ jix-chu-juana-m. sandal-1sbj int-put_on cond-1sbj atr-ext-work-des ‘I (always) put on sandals when I want to work.’ In Taracahitan languages, in particular Tarahumara and Yaqui, there is a verb that expresses general movement, siime ‘to go’. The example in (75a) illustrates the basic verb in Yaqui, and (75b) shows the verb cha’asisime ‘to float’, where siime ‘to go’ is grammaticalized as a reduplicated form. Meanwhile, examples in Tarahumara show the basic, although reduplicated form, encoding the iterative routine action of the boy, in (75c), and the grammaticalized verb as part of another verb that expresses continuous action, in (75d):

28 I preserve the way Saxton glosses the example.

884

Zarina Estrada-Fernández

(75) Yaqui a. U ili uusi Bikam-meu siime. det.nom dim boy Vícam-dir go.prs ‘The small boy is going to Vícam.’ b. Kuta ba’a-po cha’asisime. stick water-loc float.prs ‘The stick floats on the water.’ Tarahumara c. towi tienda sisimi boy store it~go ‘The child goes to the store.’ d. … echi na bila m-uchu-wasima ri ke echi perasi det loc evi loc-crouched_down-go_to_put evi dem.obj det pear riki-busi-ka ba … top-down-st pdp ‘… where he (certainty) crouched down to put the pears while (he) gets them down …’

3.3.2 Posture verb to aspect (stative or perfect) One of the most common aspect markers in UA languages is the stative suffix -ka (cf. Table 4). There are at least two possible grammaticalization pathways associated with this element; the first one is mentioned by Givón (2011: 102), the other by Estrada-Fernández (2014). Givón’s analysis considers the source of the stative -ku to be an old accusative case marker (see Section 2.3.2). The grammaticalization chain is provided in (76), and data from Ute supporting the different stages of such path of change are given in (77): (76) Accusative case suffix

>

complementizer

>

subordinator

(77) a. Ute (Givón 2011: 102) su-i-s ‘one’ (subject) su-ku-s ‘one’ (object) b. Equi-object (manipulation verb): mama-chi aapa-chi wúuka-vaa-ku may-kya. woman-sbj boy-obj work-irr-comp tell-ant ‘The woman wants the boy to work’ c. … kukwi-kwa-puga-y-ku, uwa-rugwa-puga-y-ku … … fill-go-rem-obj-sub him.obj-give-rem-obj-it ‘… when it (the bowl) filled up, (he) gave it to him …’

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

885

However, Estrada-Fernández (2019) argues that the grammaticalization of -ka may have had its origin in the posture verb kaat, ‘be lying’ (like an animal lying on the floor). The posture verb grammaticalizes as a stative aspect marker, which qualifies the event as non-dynamic. Later, the stative suffix yields a common strategy for nominalizing clauses (participle): (78) Posture verb

>

stative marker (participle)

>

nominalizer

(79) a. Tarahumara riwi-sa-ka anee-ri mapu kina sim-ea. see-cond-stat tell-pfv sub here come-pros ‘Once you see (him), tell him to come.’ b. Pima Bajo (Estrada-Fernández 1998: 43) Vis-ka-(a)n am ikoma, si’ du’ud; tua’a-ka tumk all-stat-irr loc gray int rain outside-stat first kig-di-a-kada, good-appl-pros-temp ‘(When) all is cloudy, there is a lot of rain; first it is good to be outside …’ (being outside …) c. Northern Tepehuan (Bascom 1982: 376) ši=ká maatí-ka-mu aapími. sub=already know-stat-fut you.pl ‘… that you will know.’ Pima Bajo and Yaqui have examples where the stative marker is considered to nominalize a verbal element: (80) a. Pima Bajo mul-tam huaha-ka bihk. mule-loc carry-nmlz take.pfv ‘(They) took him carrying on the mule.’ b. Yaqui (Estrada-Fernández et al. 2004: 327) chuku-la yejte-ka bea … later-adjvz sit-nmlz then ‘Then later while (he) was sitting …’ Another change related to the stative marker is the perfective aspect. Example (81) gives the grammaticalization pathway and (82) an example from Yaqui: (81) Stative suffix

>

perfective

886

Zarina Estrada-Fernández

(82) Yaqui (Estrada-Fernández et al. 2004: 334) Junu-ka’a yeu sim-su-ka-’a-po dem.sg-acu dir go-term-pfv-link-loc ‘Then after that happened there …’ Still later, this suffix may combine with a locative postposition -am to form a suffix -kam, which adjectivizes the verb (participle). This suffix is also observed in Tohono O’odham and Northern Tepehuan. (83) Pima Bajo a. kig dah-kam. good be.sitting-nmlz ‘sitting/behaving correctly’ b. aap si’ oam-kam. 2sg.sbj int be.yellow-nmlz ‘You are yellowish.’ (like when a person is scared) c. la’ali oob vihig-kam. pl~dim person starve-nmlz ‘The children are hungry.’

3.4 Mood Mood in UA languages shows a rich but uncertain scenario since it is probably the least described, or most under-described, domain. Some markers, as is the case of the irrealis, which has also been described as future or potential, clearly show cognate suffixes: irr1 *-an/-na, and irr2 *-mera. Other moods, like the imperative, show a different scenario. The morphology we have documented is presented in Table 4.

3.4.1 Irrealis to future Taking into consideration both tense-aspect and mood, two grammaticalization pathways can be proposed. The first one describes the change from an irrealis suffix, either -ma or -na, grammaticalized into future; the pathway of grammaticalization is provided in (84) with data from Tarahumara and Guarijio in (85a–b). (84) Irrealis

>

future

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

887

Tab. 4: Mood morphology. O’O

PB

NT

ST

YQ

GU

irrealis/ future/pot

cem (as a particle)29

-an

-na

-a fut

-Ɂean subj/ -ka-Ɂean pas-subj/ -ne ~ -nee fut

-mera ~ -mea -mera ~ -mia ~ -ma ~ -mea ~ fut/pot/irr -ma fut/irr

imperative

-iñ ~ -ñ -in (simple imperative) i- (directional imperative)

negative imperative conditional

-iñ xi-

kova

ṣa ‘if’ conditional

TAR

-ka, -sa sg, -si pl

katé

Sp. sea > sia subj, no cond

-sói ~ -yó fut/subj

-sa

(85) a. Tarahumara (Estrada-Fernández 2011–2012) mapu ka’ra tibu-ma echi kawi pa … sub good take_care-irr dem land dpd ‘… so they may take care of the land …’ b. Guarijio/Warihío (Félix Armendáriz 2005: 56) ehpé tukaó koči-má=ne mehká pete-čí. now night sleep-fut=1sg.sbj far.away house-loc ‘Tonight I will sleep outside the house.’

3.4.2 Conditional to future The second path of grammaticalization is observed when the conditional suffix changes into a future marker, (86), a change observed in Tarahumara, -sa, in (87), and also observed in the future from Nahuatl, -s, as in (88): (86) Conditional

29 Hale (1969).

>

future

888

Zarina Estrada-Fernández

(87) Tarahumara mapu jepuka ala ni-sa repa … sub who good aux-cond above … if there is one who is good to be in charge (above, as governor) … (88) a. Nahuatl (future -s) (Peregrina Llanes 2015: 48) ach-ni-kin-tla-mik-ti-li-s. neg-1sg.sbj-3pl.obj-unsp.obj-die-caus-appl-fut ‘I will not make someone die on behalf of them.’ b. Nahuatl (future -s) (Peregrina Llanes 2015: 81) na ni-tla-(i)štlahua-s. 1sg 1sg.sbj-undf.obj-pay-fut ‘I will pay it.’

3.5 Evidentials The evidential reportative as [aš] in Pima Bajo is derived diachronically from the verb kait [kaiš] ‘hear’, ‘understand’, ‘say’, as in (89) and (90). It is probably related to the affirmative a’ or the conditional-irrealis suffix -sa, (cf. Table 5), both from Tarahumara. Willett (1991: 161) points out that one of the evidentials from Southeastern Tepehuan is sap, ‘reported to the speaker, unknown to the hearer’. This element is probably related to the reportative as from Pima Bajo, where the change from sap to as is the result of metathesis, one of the most common processes of change in the Tepiman languages. The grammaticalization pathway is provided in (89); an example from Pima Bajo with the full verb is provided in (90), and (91a) and (91b) illustrate the reportative, respectively, in Pima Bajo and Southeastern Tepehuan: (89) Say

>

evidential/reportative/quotative

(90) Pima Bajo ko aap is. higai kait 3sg.sbj say.pfv sub 2sg.sbj steal.pfv ‘She said that you stole (her).’ (91) a. Pima Bajo niid. siv aat as iv-is now 1pl.sbj rep respect-ss see.impf ‘(and) now, it is said that we respect [obey] and see (it).’

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

889

b. Southeastern Tepehuan (Willett 1991: 164) ma’n mu-pai’ sap quio gu ma’ncam. one there-where reu live art person ‘(It is told that) there once lived a man in a certain place.’ Recall that in Section 3.1, I mentioned that a second source of evidentials is a modal verb of obligation attested in Pima Bajo, tua ‘obligate’ (cf. 61a), which is grammaticalized as a modal particle in Yaqui, in (92), and as an evidential in Pima Bajo, in (93): (92) Yaqui (epistemic particle) (Estrada-Fernández 2009: 304) jiba tua tebaure-ka weye always cert hungry-nmlz come-prs ‘(He) always comes hungry.’ (93) Pima Bajo tua am-taan? ik inter evi 2sg.nsbj-ask.pfv ‘What did (he) truly ask you?’ The grammaticalization chain is as in (94): (94) Modal verb

>

modal/evidential

4 Grammaticalization of complex constructions 4.1 Complement clauses Complement clauses in UA languages from northwestern Mexico are encoded by different devices, but it is possible to organize them in two main groups: (i) strategies that nominalize a clause and (ii) invariable words functioning as subordinators. Further properties are also relevant; for example, TAM morphology of the verb, which may determine whether the verb can be considered finite or non-finite, the encoding of the subject of the complement clause as genitive, accusative or zero (anaphora zero), or accusative/possessive case marking of the complement. From this perspective, the grammaticalization pathways of these devices are obviously diverse: nominalizing suffixes usually have their origin in lexical nouns, but connectives or subordinators (complementizers), according to Heine and Kuteva (2002, 2007), can develop from verbs, nouns or demonstratives, among other lexical sources. In the UA languages, the source of all these elements requires further research.

890

Zarina Estrada-Fernández

Table 5 organizes the main nominalizing suffixes, if any is present, in the languages considered in this contribution. Yaqui is unique in that the complement clause may carry an accusative marker -ta. For some languages, the suffixes are not restricted to complement clauses but sometimes occur only with relative clauses or with adverbial adjuncts or clauses. For some of these suffixes, their sources are still unclear (see also Section 5.2 for the use of some of the nominalizing suffixes in relative clauses).

Tab. 5: Nominalizing suffixes. Nominalizing suffixes

*-ka

O’O

-ka, -kam

PB

-ka, -kam

*-ame/-eme

*-dam(a)

NT

-ka -me

-dama

-dam

ST

-ka

YQ

GU

TAR

-a

-a

-a

-ka

-ka

-ka

-ame > -me > -m

-ame

-ame

-dami -a’u

Examples that illustrate complement clauses marked with nominalizing suffixes are the following. (95) a. Tohono O’odham (Bahr et al. 1974: 236) …t ’oi a o ’ai ’am ’i hím-dam… impf fut exist later loc go-nmlz ‘… and then eventually it [sickness] will reach him later on …’ b. Pima Bajo aan tukgi-dam tadag. 1sg.sbj faint-nmlz feel.impf ‘I feel like fainting.’/‘I feel like I’m going to faint.’/‘I feel faint.’ c. Northern Tepehuan idir go kiili šianki t=iitini isaa-dami. dem det man sub 1pl.sbj=find.pfv plant-nmlz ‘This is the man who we saw planting.’

30 I have not been able to gloss all the data.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

891

d. Tarahumara (Estrada-Fernández 2011–2012) neje bene-ri kina kuch-wa weeri-a. 1sg.sbj learn-pfv poss children-poss take_care-nmlz ‘I learned to take care of my children.’ e. Yaqui Junum jo’a-po weye-m-ta=ne bicha-k. dem house-loc walk-nmlz-acc=1sg.sbj see-pfv ‘I saw what was happening (going on) in that house.’ (lit. I saw the happening in that house) f. Guarijio/Warihío aapóe pepéne-na wikahtá-ka kawé. 3pl.sbj pl~know-prs sing-nmlz well ‘They know (how) to sing well.’ (lit. they know good singing) Subordinators, or complementizers, in these languages, are also diverse. A subordinator ma is attested in Tohono O’odham, as in (96a), and in Tarahumara, as in (96b). Frequently, in this language, the subordinator combines with an old demonstrative pu, as in (96c). In Southeastern Tepehuan the subordinator is the allomorph na, as in (96d): (96) a. Tohono O’odham/Pima (Bahr et al. 1974: 40–41) ñé:, m ’ant o sisél ’a: mo g húawǐ hab cú’ig, cú:wǐ look sub I fut straightly tell sub the deer like.this named, rabbit hab cú’ig […] like.this named, ‘Look, as I will explain in detail how the deer is, the rabbit is’ […] b. Tarahumara (Brambila 1953: 58) akiná tumu poši ma tumu arewere pa loc 2pl.sbj bring sub 2pl.sbj leave pdp31 ‘Bring here what you have left!’ c. Tarahumara (Estrada-Fernández 2011–2012) ne maye-i mapu towi raja-ri. 1sg.sbj believe-pfv sub child burn-pfv ‘I believed that the boy burned.’ d. Southeastern Tepehuan (Willett 1991: 246) na-ch-ich ya-'ay. Cu-p-ich cham nii'ñ na mi' pu-quic sc-2sbj-pfv neg see sub there sim-stand sub-1pl-pfv here-arrive ‘You didn’t see him standing there when we came?’ 31 Glosses for Brambila’s example are provided by me.

892

Zarina Estrada-Fernández

For Pima Bajo, we can hypothesize that the connective ko has its origin in a denial verb kova, which is synonymous with the vetative kova, and also expresses surprise or emphasis. When this element loses phonological material, it reduces its form to ko and is associated with three different functions: (i) the marking of sequential events, i.e., coordinating conjunction, (ii) the expression of a contrastive situation or different subject, and (iii) the marking of adverbial subordination or complementation in complement clauses. The last function is usually an instance of the following grammaticalization pathway: (97) Subordinator

>

inflected subordinator

In other languages of the world, the marking of agreement within a subordinator has been described as complementizer agreement or inflected subordinators (De Groot 2000). Steele et al. (1981) considered this type of element an aux. The grammaticalization pathway in (97) is also observed in some of the Tepiman languages, thus, while (96d) illustrates the occurrence of an inflected subordinator in Southeastern Tepehuan, na-ch-ich ‘sub-1pl-pfv’, (98) illustrates the inflected subordinator in Pima Bajo: (98) Pima Bajo Peier mat k(o)-at kav mua. Pedro know.pfv sub-1pl.sbj horse kill.pfv ‘Peter knew that we killed the horse.’ Inflected subordinators are the result of the “attraction position” described by Bisang (1996). In other words, the phenomenon is motivated by Wackernagel’s position. This mechanism of change is also observed in Tarahumara, where the general subordinator ma changes into mapu when the demonstrative pu attaches to its right, as illustrated in (96c).

4.2 Relative clauses One of the most frequent sources of a relativizer is an interrogative pronoun. The change is shown in (99): (99) Interrogative pronoun

>

relativizer

The grammaticalization of an interrogative pronoun in relative clauses is observed in Northern Tepehuan and Pima Bajo. The first language has two types of relative clauses, finite and non-finite (Bascom 1982: 377). The encoding strategies differ accordingly. Finite relative clauses show a relativizer that is grammaticalized from an

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

893

interrogative pronoun, calquing the Spanish pattern: =máákidi ‘which’ in (100a), or =anki ‘who’, in (100b), which cliticized to the subordinator ši ~ iš: (100) Northern Tepehuan a. káši=a=n=t íí ááni go-pastí ya iš=máákidi giñ-máá ígai. already=b=i=compl drank I the-pill sub=which me-gave he ‘I already took the pill which he gave me.’ b. s=aid yi ka duaad yímu i-óóki ši=anki kóóyi gi-nááto. quot=then already scared the-woman sub=who snake rflx-became ‘Then the woman who had become a snake was frightened.’ Pima Bajo also shows the grammaticalization of an interrogative pronoun in relative constructions, although it is not the most frequent strategy: (101) Pima Bajo (Estrada-Fernández 2012: 142, 144) kil vuus ki-tav ibiga gogos mua. ig det.sbj man go_out.pfv house-dir where dog die.pfv ‘The man left the house where the dog died.’ The interrogative element may also be combined with a subordinator as is observed in example (102) from Pima Bajo: (102) Pima Bajo (Estrada-Fernández 2012: 144) kil mua gogos aita-ko kiik-im-tad. ig det.sbj man kill.pfv dog inter-sub bit-cont-rem ‘The man killed the dog that was biting him.’ Another strategy to relativize clauses is nominalization; the grammaticalization pathway of nominalization is provided in (103): (103) Nominalized relative constructions

>

finite subordinate constructions

Northern Tepehuan shows the use of the inherited nominalizing suffix -kame (cf. Table 5) in non-finite relative clauses (originally glossed by Bascom [1982: 378] as a participial suffix, prtc): imó gi-vííd ya-kami. (104) bir-ími-i this.way-come-prs one rflx-be.rich-nmlz ‘There comes a rich man.’ (lit. there comes the one that is rich)

894

Zarina Estrada-Fernández

Miller (1996) for Guarijio, and Félix Armendáriz (2007) for Warihío, register a related suffix -(ya)me (Table 5) for the encoding of a relative construction, although this strategy is restricted to subject relative clauses only: (105) Guarijio/Warihío a. tihoé tetewá-ka-me wačó. man see-pst-nmlz soldier ‘The man who saw the soldier.’ b. tihoé karí mete-yáme ko’korépa-re. man house build-nmlz get.sick-pfv ‘The man who builds houses got sick.’ In object relative clauses the suffix is -a: (106) tihoé amó me’eyá-ri-a. man 2sg.nsbj kill-pfv-nmlz ‘The man that you killed.’ The same situation is mentioned by Alvarez Gonzalez (2012: 80) and Guerrero (2012: 102–103) for Yaqui, where the language shows two suffixes: -me, which is used in subject relatives, as in (107a), and -’u, which occurs in non-subject relatives, as in (107b). (107) Yaqui (Guerrero 2012: 99) a. U o’ou enchi bicha-ka-me siika. det man.nom 2sg.acc see-pfv-nmlz go.sg.pfv ‘The man who saw you, left.’ b. U-me o’ou-im em bicha-ka-’u saja-k. det-pl man-pl 2sg.gen see-pfv-clm go.pl-pfv ‘The meni who you saw left.’ In both Yaqui and Guarijio, the subject in object relative clauses appears encoded with an accusative-genitive-non-subject pronoun. In contrast, Tarahumara no longer shows a genitive/possessive or accusative subject in nominalized constructions. The language only preserves the nominalizing suffix -a, and the subordinator mapu. Examples in (108a–c) show a verb which is half-way to become a finite verb, since it shows a reduced perfective suffix -r, but also a nominalizer suffix -a. (108) Tarahumara (Estrada-Fernández, 2011–2012) a. sebali namuti mapu iyeri ich-a Rarámuri … some things sub foc sew-nmlz Rarámuri ‘some of those things that the Raramuri sew …’

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

895

b. ’a na rewa-ri echi kochi mapu Juan mea-ri. aff loc see-pfv dem dog sub John kill-pfv ‘I saw the dog that John killed.’ c. echi nali wikara-ami ba mapu wili-r-a a’li chema la dem loc sing-nmlz pdp sub bring-pfv-nmlz sim celebration evi ba … pdp ‘The singer who carries it (the rattle) during the celebration …’ Modern Pima Bajo has only one nominalizing suffix to encode this type of construction, -kig(a). A subject relative clause is shown in (109a) and an object relative clause in (109b). The subject in an object relative clause must be encoded as an accusative-genitive/possessive-non-subject pronoun, am- ‘2sg.nsbj’. kil hink-im-kig. (109) a. gogos kiia dog bite.pfv man shout-impf-rel ‘The dog bit the man that was shouting.’ si’ lii. b. gogos am-niar-kig dog 2sg.nsbj-buy-rel int small ‘The dog that you bought is small.’ The nominalizing suffix -kig has grammaticalized from an old stative nominalizing suffix -ka and the demonstrative higai. It seems that the use of a demonstrative might be a contact-induced change since Seri also shows this type of strategy as shown in (110). A similar development has been attested in Coahuilteco, an extinct language from Texas, by Troike (2010), as in (111). Data from Coahulteco and Seri32 open the possibility that such a peculiar strategy of relativization must be considered as an areal strategy shared by some languages spoken in the southwest of the US and in northwestern Mexico. (110) Seri (Larios, p.c.)33 xo xiica qu-iistoj izaaj … sub things nmlz.sbj-breathe.sg dem ‘But those things that used to breathe …’

32 Coahuilteco and Seri are, according to Troike (2010) and Marlett (2007), isolated languages, but Campbell (1997) mentions that these two languages were considered part of the Hokan-Siouan group by Sapir (1949 [1929]). 33 Maria Alfonsa Larios’ fieldwork notes of Seri.

896

Zarina Estrada-Fernández

(111) Coahuilteco (Troike, 2010: 118)34 pin (xami·n) e xa-p-o·wxča·lak pitapa-m. thing (you) obj 2>3-sub-steal dem-c ‘that thing (which) (you) stole.’ This pathway of grammaticalization is an instance of the one provided in (51) above (Section 2.4)

5 Discussion In this contribution, I have discussed several grammaticalization pathways observed in UA languages from northwestern Mexico. For some grammatical domains, the analyses are not easy since these languages are under-described, or the descriptions of the categories use different terms to refer to the same or related notions. In this section, I have summarized the major grammaticalization pathways observed in these languages. Eleven of them – signalled at the end of the line by the abbreviations Heine and Kuteva in the following list – occurred in Heine and Kuteva (2002). Two other pathways, with minor annotations, may be also considered to be in Heine and Kuteva (2002): – Plural SAP pronouns > Plural animate nouns > Mass/Collective nouns (§ 2.1) possessive class (§ 2.2.1) – raise/have (cattle) noun (§ 2.2.1) – possess > possessive alienable marker (§ 2.2.2.1) possessive alienable marker > – possess possessive inalienable marker (§ 2.2.2.1) – possess > possessive marker > inalienable/inherent possession > 3rd person possessive body-parts whole-part relation (§ 2.2.2.2) – Locative > oblique case marker (§ 2.3.1)35 – Genitive case marker > oblique case marker > patient (accusative) case marker (§ 2.3.2) – Determiner case marker > adjective accusative case marker (§ 2.3.3) 3rd person pronoun (§ 2.4) (Heine and Kuteva 2002: 330) – Demonstrative definite-determiner (§ 2.4) (Heine and Kuteva 2002: 320) – Demonstrative > relative clause marker (§ 2.4) (Heine and Kuteva 2002: 335) – Numeral/one > indefinite pronoun (§ 2.5) (Heine and Kuteva 2002: 332) – Utterance verb/say > cause/causative (§ 3.1) (Heine and Kuteva 2002: 328)

34 Referential indexes marked by the author are not included. 35 This grammaticalization pathway may be considered equivalent to the ablative > dative/accusative listed in Heine and Kuteva (2002: 217).

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

– – – – – – – – – – – – – – –

897

Modal obligation verb > cause/causative (§ 3.1) 3rd person object/reflexive pronoun > middle marker (§ 3.1) (Heine and Kuteva 2002: 333)36 2nd singular object/reflexive pronoun > middle marker (§ 3.1) (From Givón 2011: 102) 3rd person object pronoun > agreement marker (§ 3.2) (Heine and Kuteva 2002: 327) Verb of movement/go > continuous/progressive aspect > desiderative (§ 3.3.1) (Heine and Kuteva 2002: 330) Accusative case suffix > complementizer > subordinator (§ 3.3.2) (From Givón 2011: 102) Posture verb > stative marker (participle) > nominalizer (§ 3.3.2) Stative suffix > perfective (§ 3.3.2) Irrealis > future (§ 3.4.1) Conditional > future (§ 3.4.2) Say > evidential/reportative/quotative (§ 3.5) (Heine and Kuteva 2002: 335) Modal obligation verb > modal/evidential (§ 3.5) (Heine and Kuteva 2002: 320)37 Subordinator > pronoun > inflected subordinator (§ 4.1) Interrogative pronoun > relativizer (§ 4.2) (Heine and Kuteva 2002: 335) Nominalized relative constructions > finite subordinate constructions (§ 4.2)

The analysis of these grammaticalization pathways is directly related to three major tendencies: A. Conservative inherited strategy (languages use what is available from their families) B. Mixed strategy: languages make use of their own resources, but universal-typological tendencies may cooperate as well C. Contact induced-grammaticalization: languages are influenced by what occurs in neighbouring languages. The fact that only 37.9 % of the grammaticalization pathways were mentioned by Heine and Kuteva (2002) leaves open the possibility of further studies of grammaticalization in other languages of the world at the same time as it imposes a challenge for the study of linguistic diversity.

36 The grammaticalization pathway in Heine and Kuteva without mentioning the 3rd or 2nd person. 37 Heine and Kuteva (2002: 320) describes one of the possible changes of modals like deontic modaliity > (1 epistemic modality).

898

Zarina Estrada-Fernández

Acknowledgements This paper was partially supported by a research grant approved by Conacyt (CB2013–220328), which is gratefully acknowledged. I thank the editors for their insightful comments and helpful suggestions on earlier drafts of this paper.

Notes Most of the data from Pima Bajo, Yaqui, Northern Tepehuan, Tarahumara and Yaqui come from my own fieldwork. References show the source of other examples; in these cases, some of the glosses have been adapted.

Abbreviations acc = accusative, aff = affirmative, agr = agreement, al = alienable, appl = applicative, art = article, atr = attributive, c = subject person concord suffix, class = classifier, clm = complementizer, compl = completive, cond = conditional, cont = continuous, d.d. = distal demonstrative, des = desiderative, det = determiner, dir = directional, dist = distributive, d.p. = proximate demonstrative, dur = durative, emph = emphatic, evi = evidential, foc = focus, fut = future, gen = genitive, imp = imperative, impf = imperfective, inch = incoative, inst = instrument, int = intensive, inter = interrogative, irr = irrealis, it = iterative, link = phonetic linker, loc = locative, mir = mirative, mod = modal, n = noun, neg = negative, nom = nominative, nmlz = nominalizer, nsbj = non-subject, obj = object, obl = oblique, pass = passive, p.c. = personal communication, pdp = prosodic discourse particle, pfv = perfect, pl = plural, pobj = primary object, poss = possessive, pp = postposition, pred = predicative, prog = progressive, pros = prospective, prs = present, prtc = participial, pst = past, purp = purpose, quot = quotative, real = realization, rel = relativizer, rem = remote, rep = reportative, reu = reported to speaker, unknown to hearer, rflx = reflexive, sbj = subject, sc = strong command, sens = sensorial, sg = singular, sim = simultaneity, ss = same subject, st = stative, s.t. = something, sub = subordinator, subj = subjunctive, term = terminative, tr = transitivizer, unsp = unspecified, v = verb

References Alvarado García, Maribel. 2007. Los procesos morfofonológicos y supletividad como mecanismos para expresar la categoría gramatical de número en tarahumara. Dimensión Antropológica. 14(40). 29–51. Alvarez Gonzalez, Albert. 2012. Relative clauses and nominalizations in Yaqui. In Bernard Comrie & Zarina Estrada-Fernández (eds.), Relative clauses in languages of the Americas. A typological overview, 67–95. Amsterdam: John Benjamins. Alvarez Gonzalez, Albert & Fany Muchembled. 2013. Les classificateurs possessifs en langues uto-aztèques: catégorisation et évolutions. La Linguistique 49(2). 11–31.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

899

Bahr, Donald M. 1986. Pima-Papago -ga. Alienability. International Journal of American Linguistics 52(2). 161–171. Bahr, Donald M., Juan Gregorio, David I. Lopez & Albert Alvarez. 1974. Piman shamanism and staying sickness. (Ká:cim Múmkidag). Tucson: The University of Arizona. Bascom, Burt. 1982. Northern Tepehuan. In Ronald W. Langacker (ed.), Studies in Uto-Aztecan grammar. Volume 3. Uto-Aztecan grammatical sketches, 267–393. The Summer Institute of Linguistics/The University of Texas at Arlington. Bisang, Walter. 1996. Areal typology and grammaticalization: processes of grammaticalization based on Mainland South East Asian languages. Studies in language 20(3). 519–597. Brambila, David. 1953. Gramática rarámuri. México: Editorial Buena Prensa. Brambila, David. 1976. Diccionario rarámuri-castellano (tarahumar). México: Buena Prensa. Burgess, Don. 1984. Western Tarahumara. In Ronald W. Langacker (ed.), Studies in Uto-Aztecan grammar. Volume 4. Southern Uto-Aztecan grammatical sketches, 1–149. The Summer Institute of Linguistics/The University of Texas at Arlington. Campbell, Lyle. 1997. American Indian languages. The historical linguistics of Native American. Oxford: Oxford Studies in Anthropological Linguistics. Carrillo, Araceli. 2005. Posesión alienable e inalienable en tepehuano del norte. Talk at the Primer Foro Nacional de Estudiantes de Lingüística y Literatura. Hermosillo, Sonora, 14, 15 y 16 de abril. Casad, Eugene. 1984. Cora. In Ronald W. Langacker (ed.), Studies in Uto-Aztecan grammar. Volume 4. Southern Uto-Aztecan grammatical sketches, 154–459. The Summer Institute of Linguistics/The University of Texas at Arlington. Dakin, Karen. 1991. Nahuatl direct and mediated possession: A historical explanation for irregularities. International Journal of American Linguistics 57(3). 298–329. Dakin, Karen. 2004. Prologo. In Zarina Estrada-Fernández, Crescencio Buitimea Valenzuela, Adriana Elizabeth Gurrola Camacho, María Elena Castillo Celaya & Anabela Carlón Flores (eds.), Diccionario yaqui-español y textos: Obra de preservación lingüística. México: Editorial Plaza y Valdez-Universidad de Sonora. 13–20. Daniel, Michael. 2005. Plurality in independent personal pronouns. In Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/35 (accessed 19 July 2018). Dedrick, John M. & Eugene H. Casad. 1999. Sonora Yaqui Language Structures. Tucson: University of Arizona Press. De Groot, Casper. 2000. Minor word classes. In Geert Booij, Christian Lehmann, Joachim Mugdan in collaboration with Wolfang Kesselheim & Stavros Skopeteas (eds.), Morphologie. Morphology. Ein internationals Handbuch zur Flexion und Wortbildung. An International Handbook on Inflection and Word Formation. 820–831. Berlin – New York: Walter de Gruyter. Estrada-Fernández, Zarina. 1998. Pima bajo de Yepachi, Chihuahua. México: Archivo de lenguas indígenas de México. El Colegio de México. Estrada-Fernández, Zarina. 2005. The pronominal form a- in Pima Bajo as a middle marker. International Journal of American Linguistics 71(3). 277–302. Estrada-Fernández, Zarina. 2009. Modalidad en yaqui. In Ramón Arzápalo Marín & José Luis Moctezuma Zamarrón (eds.), Lingüística Amerindia. Aportaciones recientes, 290–310. México: Instituto de Investigaciones Antropológicas. Universidad Nacional Autónoma de México. Estrada-Fernández, Zarina. 2011–2012. Fieldnotes from Tarahumara. Guachochi, Chihuahua. Estrada-Fernández, Zarina. 2012. From demonstrative to relative marker to clause linker: Relative clause formation in Pima Bajo. In Bernard Comrie & Zarina Estrada-Fernández (eds.), Relative clauses in languages of the Americas. A typological overview, 127–146. Amsterdam: John Benjamins.

900

Zarina Estrada-Fernández

Estrada-Fernández, Zarina. 2014. Gramática de referencia del pima bajo. Volume 1. Hermosillo: Editorial Universidad de Sonora. Estrada-Fernández, Zarina. 2019. Syntactic nominalizations in Pima Bajo: Diachronic Diversity. In Albert Alvarez Gonzalez, Zarina Estrada-Fernández & Claudine Chamoreau (eds.), Diverse Scenarios of Syntactic Complexity. 167–189. Amsterdam: John Benjamins. Estrada Fernández, Zarina, Crescencio Buitimea Valenzuela, Adriana Elizabeth Gurrola Camacho, María Elena Castillo Celaya & Anabela Carlón Flores. 2004. Diccionario yaqui-español y textos: Obra de preservación lingüística. México: Editorial Plaza y Valdez-Universidad de Sonora. Estrada-Fernández, Zarina & Rolando Félix Armendáriz. 2010. Typological differences among middle constructions in some Uto-Aztecan languages. In Claudine Chamoreau, Zarina Estrada-Fernández & Yolanda Lastra (eds.), A new look at language contact in Amerindian languages, 89–117. Lincom-Europa: München. Estrada-Fernández, Zarina & Aarón A. Grageda Bustamante. 2014. Compendio del arte de la lengua de los tarahumares y Grammatica linguae tarahumaricae: Análisis comparativo de dos obras anteriores al siglo XIX sobre el tarahumara. ICHOLS. August 25–29th, Vila Real, Portugal. Estrada-Fernández, Zarina, Mercedes Tubino & Jesús Villalpando. 2015. 32 Valency classes in Yaqui. In Andrej Malchukov & Bernard Comrie (eds.), Valency classes in the world’s languages: Vol. 2: Case studies from New Guinea, Australia, and the Americas, and theoretical outlook, 1359–1389. Berlin: Mouton de Gruyter. Félix Armendáriz, Rolando. 2005. A Grammar of River Warihío. Houston, USA: Rice University dissertation. Félix Armendáriz, Rolando. 2007. A Grammar of River Warihío. München: Lincom. García Salido, Gabriela & Zarina Estrada-Fernández. 2009. La voz reflexiva, recíproca y media: El caso del tepehuano del sur. In Susana Cuevas Suárez (ed.), La lengua y la antropología para un conocimiento global del hombre. Homenaje a Leonardo Manrique. 253–266. México: Instituto Nacional de Antropología. Gelderen, Elly van. 2004. Grammaticalization as economy. Amsterdam: John Benjamins. Givón, Talmy. 1981. On the development of the numeral ‘one’ as an indefinite marker. Folia Lingüística Historica. Societas Linguistica Europaea (2). 35–63. Givón, T. 2009. The genesis of syntactic complexity. Amsterdam: John Benjamins. Givón, T. 2011. Ute reference grammar. Amsterdam: John Benjamins. Guadalaxara, Thomas de. 1683. Compendio del arte de la lengua de los tarahumares, y guazápares. Puebla de los Ángeles: Diego Fernández de León. Guerrero, Lilián. 2012. On relative clauses and related constructions in Yaqui. In Bernard Comrie & Zarina Estrada-Fernández (eds.), Relative clauses in languages of the Americas. A typological overview, 97–126. Amsterdam: John Benjamins. Guerrero, Lilián & Robert D. Van Valin, Jr. 2004. Yaqui and the analysis of primary object languages. International Journal of American Linguistics 70(3). 290–319. Hale, Kenneth L. 1969. Papago cəm. International Journal of American Linguistics 35(2). 203–212. Heine, Bernd, Ulrike Claudi & Frederike Hünnemeyer. 1991. Grammaticalization. A conceptual framework. Chicago: University of Chicago Press. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd & Tania Kuteva. 2007. The genesis of language. A reconstruction. Oxford: Oxford University Press. Heine, Bernd & Motoki Nomachi. 2013. Contact-induced replication. Some diagnostics. In Martine Robbeets & Hubert Cuyckens (eds.), Shared grammaticalization with special focus on the Transeurasian languages, 67–100. Amsterdam: John Benjamins.

Grammaticalization in Uto-Aztecan languages from northwestern Mexico

901

Hill, Jane H. 2005. A grammar of Cupeño (Linguistics, volume 136). Berkeley: University of California Publications. Hill, Kenneth (editor-in-chief). 1998. Hopi Dictionary / Hopìikwa Lavàytutuveni: A Hopi-English dictionary of the Third Mesa dialect. Hopi Dictionary Project, Emory Sekaquaptewa, Mary Elizabeth Black. Tucson: University of Arizona Press. Instituto Nacional de Estadística y Geografía (INEGI). 2015. Encuesta Intercensal 2015. http://www.inegi.org.mx/est/contenidos/Proyectos/encuestas/hogares/especiales/ei2015/ (accessed 07 October 2015) Kemmer, Suzanne. 1993. The middle voice. Amsterdam: John Benjamins. Kuteva, Tania. 2004. Auxiliariation: An enquiry into the nature of grammaticalization. Oxford: Oxford University Press. Langacker, Ronald W. 1977. Studies in Uto-Aztecan grammar. Volume I. An overview of Uto-Aztecan grammar. Dallas: Summer Institute of Linguistics/University of Texas at Arlington. Lehmann, Christian. 2002. Thoughts on grammaticalization. Erfurt: Arbeitspapiere des Seminars für Sprachwissenchaft der Universität Erfurt. Marlett, Stephen A. 2007. La relación entre lenguas “hokanas” en México: ¿Cuál es la evidencia? In Cristina Buenrostro, Samuel Herrera Castro, Yolanda Lastra, Juan José Rendón, Otto Schumann, Leopoldo Valiñas & María Aydeé Vargas Monroy (eds.), Memorias del III Coloquio Internacional de Lingüística, Mauricio Swadesh, 165–192. México City: Universidad Nacional Autónoma de México and Instituto Nacional de Lenguas Indígenas. Martínez Martínez, Denisse Fiordalizo. 2015. El sufijo -ta en yaqui: redireccionando su caracterización. Master in Linguistics Thesis. University of Sonora. Miller, Wick R. 1980. The possession of pets, dogs, and other beasts (both animal and human) in Southern California, and surrounding areas. AAA Annual Meeting, Washington, D.C. Miller, Wick R. 1996. Guarijío: Gramática, textos y vocabulario. México: Universidad Nacional Autónoma de México. Moseley, Christopher (ed.). 2007. Encyclopedia of the World’s Endangered Languaes. London and New York: Routledge. Taylor & Francis Group. Peregrina Llanes, Manuel. 2015. Cuentos en náhuatl Huasteca veracruzana. Hermosillo: Universidad de Sonora. Press, Margaret L. 1979. Chemehuevi. A grammar and lexicon. Berkeley: University of California Press. Sapir, Edward. 1949 [1929]. Central and North American languages. In David G. Mandelbaum (ed.), Selected writings in language, culture, and personality, 69–178. Berkeley: University of California Press. Saxton, Dean. 1982. Papago. In Ronald W. Langacker (ed.), Studies in Uto-Aztecan grammar. Volume 3. Uto-Aztecan grammatical sketches, 93–266. Dallas: Summer Institute of Linguistics/University of Texas at Arlington. Shaul, David Leedom. 1986. Topics in Nevome syntax. Berkeley: University of California Publications. Linguistics Volume 109. University of California Press. Smith, Buckingham (ed.). 1862. Grammar of the Pima or Névome, a language of Sonora, from a manuscript of the XVIII century. New York: Cramoisy Press. [AMS Press, INC. 1970]. [Arte de la lengua névome, que se dice pima, propia de Sonora; con la doctrina chistiana y confesionario añadidos. San Agustín de la Florida.] Steffel, Matthäus. 1799. Gramaticae Linguae Tarahumaricae. Archiv Města Brna, file v3, Knihova Mittrovsky, Sign. A62. Steele, Susan, Adrian Akmajian, Richard Demers, Eloise Jelinek, Chisato Kitagawa, R. T. Oehrle & Thomas Wasow. 1981. An Encyclopedia of AUX. A Study in Cross-Linguistic Equivalence. (Linguistic Inquiry Monographs). Cambridge, MA.: MIT Press Sullivan, Thelma D. 1988. Thelma D. Sullivan’s compendium of Nahuatl grammar. Salt Lake City: University of Utah Press.

902

Zarina Estrada-Fernández

Troike, Rudolph C. 2010. The typology of relative clauses in Coahuilteco, an Indian language of Texas. Southwestern Journal of Linguistics 29(1). 111–126. Tubino Blanco, Mercedes. 2010. Contrasting causatives: a minimalist approach. Amsterdam: John Benjamins. Willett, Thomas Leslie. 1991. A Reference Grammar of Southeastern Tepehuan. Dallas: The Summer Institute of Linguistics and The University of Texas at Arlington. Willett, Thomas L. & Elizabeth R. de Willett (comp). 2013. Diccionario tepehuano de Santa María Ocotán, Durango. https://www.sil.org/resources/archives/63913 (accessed September 24, 2019) Zepeda, Ofelia. 1983. Papago Grammar. Tucson, AZ.: The University of Arizona Press.

Johannes Helmbrecht

21 Grammaticalizations in Hoocąk 1 Introduction The goal of the present study is to present an overview on the grammaticalization processes that led to the historical development of certain grammatical forms and constructions in Hoocąk. Since there are no diachronic data for Hoocąk, the reconstruction of grammaticalizations in Hoocąk depends largely on the historical comparison of the related languages of the Siouan language family. Since the historical linguistics of Siouan is limited for certain grammatical domains, only a selection of forms and historical process is investigated in some detail. Nevertheless, it is a rewarding and interesting endeavor to look on the grammar of Hoocąk from a grammaticalization perspective. Hoocąk is a language with a rich morphology, and in particular, it is the morphology of the verb that is so different from what we know and with what we are familiar in European languages. The Hoocąk verb has about 10 slots for prefixes and the best way to represent these prefixes is a template as will be done below. On the one hand, these prefixes fill the core arguments of the verb, and on the other hand, they change the argument structure of the verb either by increasing or reducing or rearranging the valency of the verb. It is interesting to see where these forms come from and how they are different from the ones in the neighboring Siouan languages. The prefix domain of the Hoocąk verb is strongly grammaticalized so that much of it can be traced back to a common Siouan origin. There are two areas of the grammar that are much less grammaticalized, the suffix and clitic domain of the verb and the noun phrase. The suffixes and enclitics express mostly mood/modality, tense and aspect/aktionsarten meanings. There is also a group of verbal enclitics that mark subordination and focus. Many of the forms in this area cannot be traced back to Proto-Siouan, but are innovation in Hoocąk or in the more closely related Siouan languages. The lexical sources in this area are easier to identify. The same holds for the noun. There is no noun morphology (such as case, number, and gender); categories that are relevant to the noun have to be expressed on the noun phrase level by means of modifiers and determiners, or on the verb level. It is particular rewarding to see where the determiners (articles and demonstratives) came from and which additional functions they developed in Hoocąk. I will give a brief overview of the genetic classification of Hoocąk in the subsequent section (§ 1.1) followed by a brief typological profile of the language (§ 1.2) and some remarks on the method and the data (§ 1.3).

https://doi.org/10.1515/9783110712735-007

904

Johannes Helmbrecht

1.1 Hoocąk and the Siouan languages The Siouan language family consists of about 17 languages that were originally spoken in a large area covering most of the Great Plains expanding from the Southeast of the US to the Northwest into Southern Canada. The genetic sub-classification of the Siouan languages is generally considered as summarized in Table 1 (cf. Rood 1979; Mithun 1999: 501; Parks and Rankin 2001). That Catawba is genetically related to Siouan has been claimed for a long time and it is widely accepted by now among Siouanists (see, for instance, Siebert 1945, Speck [1934] 1969: X; Gatschet 1900). The first split in the genetic tree of the Siouan language family is between Catawba and the rest of Siouan. Three subgroups of Siouan can be distinguished, the Northwestern Missouri Valley Siouan languages, the Central or Mississippi Valley Siouan languages, and the Southeastern Ohio Valley Siouan languages. Mandan is not assigned to any of these groups. Hoocąk belongs to the Mississippi Valley Siouan group and is closely related to Chiwere, a cover term for a dialect continuum consisting of Iowa, Oto, and Missouria. There are two politically independent Hoocąk tribes, one in Wisconsin and the other in Nebraska. Linguistically, there is not much of a difference, but there is no systematic investigation of potential dialectal differences. As far as I know, there are lexical

Tab. 1: Genetic classification of the Siouan languages (cf. Rood 1979; Parks and DeMallie 1992; Mithun 1999: 501; Parks and Rankin 2001). Catawba

Catawba † Woccon †

Siouan

Missouri River Siouan (MRS)

Crow Hidatsa

Mandan Mississippi Valley Siouan (MVS) Central Siouan

Dakotan

Dhegiha

ChiwereHoocąk Ohio Valley Siouan (OVS) Southeastern Siouan

Biloxi † Ofo † Tutelo †

Sioux (dialects: Teton-Lakota, Santee-Yankton-Dakota) Assiniboine Stoney Omaha-Ponca Osage Kansa † Quapaw † Chiwere (dialects: Iowa, Oto, Missouria) Hoocąk/Winnebago

†

Grammaticalizations in Hooca˛k

905

differences. Hoocąk is still spoken at various scattered places in central Wisconsin and by a very few speakers in the Winnebago Reservation in Nebraska. It is estimated that there are less than 160 Native speakers left, all older than 60 years of age. All Siouan languages, except the Dakotan languages and perhaps Crow, are highly endangered languages and are on the verge of extinction or are already extinct (indicated by little crosses in Table 1). Catawba and the Ohio Valley Siouan (OVS) languages have been extinct for a long time now, and for this reason these languages are not well documented or described.

1.2 Typological profile of Hoocąk Morphologically, Hoocąk is a highly synthetic and fusional language with regard to the verbal prefixes, and a mildly agglutinating language with regard to the suffixes and enclitics of the verb. Syntactically, Hoocąk is a head marking language on the clause level. Two, and occasionally three, arguments of the clause are indexed by means of pronominal affixes on the verb. There are two sets of pronominal prefixes, an undergoer series and an actor series that are used to index the person/number of the subject of active and inactive intransitive verbs, respectively. Combinations of basically the same undergoer and actor prefixes are used to mark the actor and undergoer of transitive verbs. Hoocąk is thus a typical example of a language with an active/inactive alignment type (also called Split-S marking type). The other Siouan languages are similar in this respect. Hoocąk has no passive or antipassive construction. The undergoer argument of a transitive verb cannot be brought into actor role by backgrounding or suppressing the former actor. Likewise, the actor argument of a transitive verb cannot be brought into an undergoer role by backgrounding or suppressing the former undergoer at the same time. In addition, there is no passive participle form of the verb and no way to detransitivize a transitive clause as it is required by an ordinary passive construction. Third persons do not show the active/ inactive split as first and second persons do, but have an accusative alignment. Since third person singulars are always zero marked, this accusative alignment is visible only with regard to the third person plural marking. What I said for the first and second persons is also true for the third persons. There is no passive and antipassive construction. One could make a point that the impersonal use of the third person plural subject marker -ire is a kind of passive, since it is a kind of backgrounding of the actor. However, there is no passive marking on the verb and it is never possible to add an oblique agent as in English with a by phrases. The -ire sbj.3pl marker always expresses the actor, even if this actor is impersonal and thus not salient in discourse. The very same holds for the third person plural object marker wa- obj.3pl. Even if this pronominal prefix has an impersonal referent, it never reduced the valency of the verb. Not surprisingly, the morphology of the Hoocąk verb is quite complex. Besides the paradigms for the indexing of syntactic-semantic relations of the clause, there

906

Johannes Helmbrecht

are various derivational means for increasing or decreasing the valence of the verb. These means include different kinds of applicatives (locatives, instrumental, benefactive) that insert an undergoer argument with the respective semantic roles into the argument frame of the verb. Likewise, there are morphological reflexive and reciprocal markers that tend to decrease the valency of the transitive verb. A large inventory of tense, mood and aspect marking enclitics follow the verb. However, Hoocąk has no infinite verb forms such as infinitives, gerunds, and participles. Nouns, on the other hand, are morphologically quite simple. There is no nounspecific morphology such as case or number. In addition, nouns are not classified in the lexicon according to gender. Word order in the noun phrase is generally head initial. The great majority of nominal modifiers, determiners and all quantifiers follow the head noun in that order. On the clause level, constituent order is consistently verb final. Argument NPs, adverbials, and subordinate clauses usually precede the verbal predicate. Deviations from this order are pragmatically marked.

1.3 Methodology and data sources One of the most important data sources is the Comparative Siouan Dictionary (CSD) that is available now as an online resource (cf. the URL http://csd.clld.org/; Rankin et al. 2015). The CSD contains cognate sets for lexical and grammatical items in Siouan and their reconstructions either for the sub-branches of Siouan, and for Proto-Siouan, and for Proto-Siouan-Catawba if possible. The minimum condition for inclusion is that a form has cognates in at least two branches of Siouan. Other sources for this study are grammatical descriptions of the individual Siouan languages. There are modern grammars for many of the Siouan languages, representing all branches. The OVS languages in the south have been extinct for a long time now so that their level of documentation and description is significantly lower than that of the other languages. Data for Hoocąk come from descriptive studies of the author and the digital text corpus of Hoocąk that was compiled during the years 2003–2008 by a large documentation project (DOBES) of Hoocąk.1 The identification of grammaticalizations in Hoocąk presupposes that the grammatical forms in Hoocąk are compared to cognate forms in other Siouan languages. One has to find a reconstructed Proto-Siouan lexical item that can be shown convin-

1 See the website of the DOBES funding initiative the Volkswagen Foundation (http://dobes. mpi.nl//). The glossed texts and audio and video files of the Hoocąk documentation project are stored in the digital archive of the Max-Planck-Institute for Psycholinguistics called “The Language Archive”; the corresponding URL is: https://archive.mpi.nl/islandora/object/lat%3A1839_00_ 0000_0000_0001_367F_7. The website of the DOBES project “Documentation of the Hoocąk Language” led by Johannes Helmbrecht and Christian Lehmann at the University of Erfurt, Germany can be found under the following URL: https://web.archive.org/web/20050319201047fw_/http:// www.uni-erfurt.de:80/sprachwissenschaft/Vgl_SW/Hocank/cult_frames.html.

Grammaticalizations in Hooca˛k

907

cingly to be the source of a specific grammatical form in Hoocąk. The other way to identify grammaticalizations is to show that certain grammatical forms in Hoocąk are, in fact, historical compositions of grammatical forms that are still used synchronically, or that can be traced back to reconstructed material in Siouan.

2 Grammaticalization of nominal categories The noun in Hoocąk lacks most of the category-specific morphology that is found in European languages. There is no number marking, no case marking, and gender is not a category of the noun. There are, however, three derivational prefixes (cf. waa- / woo- / wii-) that derive nouns from verbs, which will not be discussed further. These prefixes are originally verbal prefixes (and still are) that derive, in the majority of cases, words that designate objects or, broadly speaking, nominal concepts in the lexicon. Since the underlying verbal prefixes of these forms are found in all Siouan languages (with slight variations) not much can be said about their grammaticalization. As will be seen below, the same holds for a number of other verbal prefixes in Hoocąk. The plurality of nominal referents is indicated lexically by means of numerals and other modifying quantifiers, or by means of third person plural prefixes on the verb. Gender marking is absent with one minor exception. There is a natural gender marking enclitic =wį that indicates female sex of the referent of personal names; cf. the examples in (1) and (2). (1) wipamą́kerewįga wipamąkere-wį=ga rainbow-fem=prop ‘Rainbow’ (a female proper name of the Thunderbird Clan) (2)

Ahusgáwįga ahu sgáa-wį=ga Wing White-fem=prop ‘White Wing’ (a female proper name of the Bird Clan)

Both traditional Hoocąk names indicate that the name bearer is female. This marker is by no means obligatory. Many personal names for women lack it. This form is also reported to be used with a few nouns designating animals, such as cee=wį ‘female cow’, hinąnek=wį ‘field mouse (female)’ (cf. Lipkind 1945: 52). However, no more examples of this type exist in the DOBES lexical database. This suffix is a reduced and grammaticalized form of Proto-Siouan *wįhe ‘woman, female’ and oc-

908

Johannes Helmbrecht

curs as a female marker also in neighboring Chiwere and the Dhegiha languages of MVS (Cf. Rankin et al. [2015] entry ‘female, woman’). Another derivational enclitic that occurs typically with nouns, but is also found with all other parts of speech such as verbs, adverbs, property words, particles, and even pronouns in Hoocąk, is the diminutive marker =nįk / =įk; cf. the example in (3). (3)

GRIZ006 … hagoréižą nįįšą́nąkįk niske 'eeja hirahígają. hagoreižą nįįšąnąk=įk nįįsge 'eeja hirahi=gają Sometime stream=dim vague there go.to.find(sbj.3sg)=seq ‘… at some point he arrived at a little stream.’

The diminutive marker in Hoocąk (as well as in closely related Chiwere) is historí ‘child’. In the Dakotan languages of cally derived from a Proto-Siouan noun *yįkE MVS, the descendants of PSI *yįḱ E ‘child’ still have the nominal meaning; for instance, in Lakota čhįčá ‘child’. Since there is no real nominal morphology in Hoocąk, I will focus on the grammaticalization processes that are associated with the various determiners that are found in the Hoocąk NP. In particular, I will summarize what can be said about the origins and grammaticalizations of the indefinite and definite article (cf. § 2.1), the two paradigms of demonstratives (cf. § 2.2), and the incipient development of a general locative postposition (cf. § 2.3). Before I start with the discussion of the grammaticalization of determiners in Hoocąk, some information on the structural and distributional properties of the NP is necessary. In Table 2, these properties of the Hoocąk NP with a common noun, a personal name, and a pronominal as head are briefly summarized. The first line in Table 2 gives a template representation of the NP in Hoocąk with a common noun as head. There are basically two determiner and two modifier positions in this template. The first modifier position (Modifier1 ) can be filled only by an object word (noun). Action words (verbs) and property words (adjectives) are not allowed in this position; see also Helmbrecht (2002, 2015b). The second modifier position (Modifier2 ) after the head noun can be filled only by a property word (adjective); see also Helmbrecht (2006a). The first determiner position (det1 ) can be filled only by one of the “old” demonstratives and these constructions are rather rare. The “old” demonstratives are called “old”, because this paradigm of demonstratives can be traced back to ProtoSiouan. The morphosyntactic and semantic properties of the members of this paradigm are very different from the ones in the “new” paradigm of demonstratives (see below (§ 2.2) for a detailed treatment). The second determiner position (det2 ) can be filled with the definite and indefinite article, and the members of both paradigms of demonstratives. The forms in this slot are mutually exclusive. The last position right of det2 is reserved for numerals and quantifiers.

909

Grammaticalizations in Hooca˛k

Tab. 2: Structural properties of the NP in Hoocąk. det1

Modifier1

Head

Modifier2

det2

num

“old” demonstratives

(noun) object word

common noun

(adjective) property word

definite article

numeral

indefinite article

quantifier

“new” demonstratives “old” demonstratives not attested not attested

personal name

not attested not attested

pronominals not attested not attested (“old” demonstratives; interrogative pronouns; free personal pronouns)

not attested proper name marker

not attested not attested

The second line in Table 2 summarizes the distributional possibilities of an NP with a personal name as head. Distributionally, personal names are quite distinct from common nouns and more closely resemble pronominals. They do not take modifiers or determiners other than the personal name marker that will be discussed in more detail below in § 2.2.1. Similarly, pronominals are devoid of all determining and modifying elements.

2.1 Definite and indefinite article Structurally, the definite article =ra and the indefinite article =(h)ižą occupy the det2 position in the NP. In general, the definite article marks the NP as definite, but there are also many cases, where the definite article only marks specificity. This is particularly obvious in combinations with the indefinite article that functions in these combinations rather as the numeral ‘one’. Compare the example in (4). (4)

CHT001b ni ̨i ̨kją́ griižą́ gixetéregiží, hookárakų́i ̨ráanąga, ni ̨i ̨kją́k=ra=hižą́ gi-xeté-ree=giží [child=def=one] appl.ben-be.big-start.going=top hookų-ire=ánąga teach-sbj.3pl=and ‘When a child was growing up, they taught him, and …’

910

Johannes Helmbrecht

The ‘child’ in (4) is not a contextually given, i.e., definite child, but a specific one. The entire text following this stretch of discourse tells the story of this specific example child; actually, this protagonist will be specified later on as a boy that undergoes the initiating rite. Hoocąk =ra must be analyzed as an enclitic, since it appears at the right edge of the NP no matter which word type (noun, property word, verb) precedes it. Only quantifiers may follow the definite article in the NP. In addition, =ra is used to nominalize clauses, thus forming subordinate clauses. In these cases, =ra is attached to the last word in the subordinate clause, which is usually a verb or auxiliary. In these cases, =ra always marks the end of the clause. The double function of =ra as a determiner of a NP and as subordinator can also be found with demonstratives. The diachronic origins of =ra remain opaque, however. This determiner is not cognate to the definite articles in Lakhota and the other Siouan languages, nor to the definite articles in Omaha-Ponca and the other Dhegiha languages. It seems that these languages developed definite articles independently and from different sources. A plausible hypothesis is that Hoocąk =ra is cognate to the Lakhota emphatic topic marker čha (< Proto-Siouan *ya). The Hoocąk form is not listed in this dictionary entry, but the function and the sound correspondence would support such an idea. Proto-Siouan /*y/ gives Dakotan /čh/ and sometimes Hoocąk /r/. Hoocąk merges /*y/ and /*r/ to /r/ (cf. Rory Larson 2016 and p.c.). Besides the topicalizing function, Lakhota čha also subordinates clauses, a function it shares with Hoocąk =ra. Cognate forms are also identified in Tutelo and perhaps in Biloxi (Ohio Valley Siouan). The indefinite article =(h)ižą in Hoocąk is used to mark indefiniteness of a referent and is hence used to introduce new referents in discourse. See, for example, the sentence in (5). (5)

GRIZ002 'Eegi hosgáižą 'eeja hakurucé wa'ų́ąki hagoréižą 'eeja nąxgų́gaja. 'eegi hosga=ižą 'eeja haruce and.then open.land=one there cross(sbj.3sg) wa'ų-'ąk=gi hagoreižą 'eeja nąąxgų=gają do/be(sbj.3sg)-pos.hor=top sometime? there hear(sbj.3sg)=seq ‘As he was crossing an opening, he heard something.’

The referent ‘opening’ is marked as indefinite in this story, because it has not mentioned before. This is a clear indefinite usage of =(h)ižą. Another usage of hižą is as an indefinite or impersonal pronoun often translated as ‘someone’; cf. (6). (6)

RL2014 hižą meegųni ̨ge 'eeja hapagihigiži … hižą mee-hegų ni ̨ge 'eeja hapahi=giži one this-that.way somewhere there go.toward=top ‘and now, if someone goes that way …’

Grammaticalizations in Hooca˛k

911

However, =(h)ižą is also identical to the word for ‘one’ and is often used simply as a numeral in a NP. In counting, a long form of this numeral hižąkira ‘one’ is used. There is a Proto-Siouan numeral ‘one’ *rųų (cf. Rankin et al. 2015: entry ‘one [1]’) that is attested in all Siouan languages except MVS, which innovated another numeral.2 The reconstruction of MVS ‘one’ (Rankin et al. 2015 entry ‘one [2]’) is not fully convincing, however. The Hoocąk form hižąkira ‘one’ is certainly cognate to the Chiwere form iyą́ khi, which is shortened to =yą / =iyą if it is used as indefinite article (cf. Greer 2015). To summarize this section, the definite article in Hoocąk is certainly not grammaticalized from a demonstrative pronoun as one would expect. Unfortunately, the source cannot be identified. It is possible that this form stems from a topic marker that is attested in Lakhota (and in the Ohio Valley Siouan languages Biloxi and Tutelo). The indefinite article in Hoocąk is polyfunctionally used as a marker of indefiniteness, as a numeral ‘one’, and as an impersonal pronoun. This is a quite frequent path of grammaticalization in the languages of the world (cf. Heine and Kuteva 2002: 220 ff).

2.2 Demonstratives Hoocąk has two paradigms of demonstratives with very different morphosyntactic and semantic properties. The respective forms are summarized in Table 3. The old paradigm of demonstratives in the upper part of the table shows a basic distinction between proximal, translated in English as ‘this’, and distal, translated as ‘that’. All (except ga’a ‘that’) have two forms, a long form with a glottal stop, e.g., te'e ‘this’, and a shorter form with simply a long vowel, e.g., tee ‘this’. These alternatives are in free variation. In addition, there are two forms that are glossed ‘this’, te'e and me'e. The second proximal demonstrative me'e ‘this’ is mentioned in older sources (cf. Lipkind 1945: 52) and occurs occasionally in the DOBES corpus. The origins of this /m/ initial form are unclear. This form cannot be traced back to one of the two Proto-Siouan proximal demonstratives on the basis of the known sound laws from Proto-Siouan to Hoocąk. As far as I can tell, there is no meaning difference between them. The distal form ga'a ‘that’ is marked with a question mark in Table 3 because it has become out of use. It is mentioned in older sources (cf. Lipkind 1945: 52) as a distal demonstrative, but it does not occur as such once in the DOBES corpus of Hoocąk texts. The “new” demonstratives are an innovation in Hoocąk. These forms are exclusively adnominal. They cannot be used pronominally like the “old” demonstratives. They are compositions of a positional auxiliary plus a deictic particle. The positional auxiliaries are verbs of being that distinguish three bodily postures of the subject,

2 Rory Larson pointed out to me that this Proto-Siouan numeral ‘one’ is also missing in Mandan.

912

Johannes Helmbrecht

Tab. 3: Paradigms of “old” and “new” demonstratives. paradigm

form

meaning

glossed in the Hoocąk DOBES corpus as

“old” paradigm

tée / te'é

‘this’

this

mée / me'é

‘this’

this

žee / že'é

‘that’ (near you)

that

? ga'á

‘that’ (near him)

that

=nąka (3-have.as.grandchild=pl ‘I have them as grandchild’ b. onkw- ar-a’se’=okon’a ‘our cousin-s’ 1pl.pat-refl-be.cousin=pl ‘we all are cousins to each other’

The enclitics on these plurals originated as distributives. On inanimates, they distribute entities over kinds. (There are several distributive forms across the Five Nations languages. Cognates of the shòn:’a are used with persons in related languages.) (8)

Mohawk distributives a. onén:ia’ onenia’=shòn:’a b. otsikhè:ta’ otsikhe’ta’=shòn:’a c. ierakewáhtha’ ierakewahtha’=shòn:’a

‘rock(s)’ ‘various rocks’ ‘sugar’, candy’ ‘various candies’ ‘towel’ ‘paper products’

948

Marianne Mithun

Onenia’shòn:’a refers to various assorted rocks, probably of different shapes and colors. Otsikhe’ta’shòn:’a could not refer to a row of candy canes: it would be appropriate in a candy store selling various kinds of candies. Ierakewahtha’shòn:’a was considered for a grocery store label for the aisle with a variety of paper products: it could not be used for a display of identical packages of paper towels. These distributives specify variety, but they imply plurality. It appears that distributives were used particularly frequently in mentions of multiple human beings, perhaps because humans are generally perceived as inherently individual. Ultimately, they became routinized with very common forms referring to three or more humans: boys, girls, men, women, grandchildren, cousins, etc. They are not used with terms referring to humanity in general.

2.2 The Augmentative The Northern Iroquoian languages all contain a productive augmentative enclitic that can be added to referring expressions to derive new lexical items. Those in the Lake languages are cognate. (9)

Mohawk augmentative a. kahonwé:ia’ kahonweia’=kó:wa b. otsinó:wen otsinowenh=kó:wa c. otsi’eróhta’ otsi’erohta’=kó:wa

=kó:wa ‘boat’ ‘ship’ ‘mouse’ ‘rat’ ‘crab’ ‘lobster’

It is added to nominals (referring expressions), not just morphological nouns. (10) Mohawk augmentative a. katsi’nónhtaks ka-tsi’nonht-ak-s neuter.agt-bug-eat-hab ‘it eats bugs’ = ‘monkey’ b. katsi’nonhtaks=kó:wa ‘gorilla’ The source of this enclitic is clear: it comes from a full verb, which can be traced back to Proto-Iroquoian. (11) Augmentative source kowá:nen. ka-owan-en neuter-be.big-st ‘it is big’

Grammaticalization and polysynthesis: Iroquoian

949

The development shows the marks of common grammaticalization processes, with semantic shift from concrete to abstract (large size > larger type), loss of lexical autonomy (word to clitic) and reduction of phonetic substance (kowá:nen > ko:wa).

2.3 Determiners As noted earlier, the Tuscarora (along with the Nottoway) were the first to separate from the main Northern Iroquoian branch. Tuscarora has an article that does not match those in the other Northern languages, ha’. At first glance it appears to be a definite article similar to English ‘the’. (12) Tuscarora article: (Howard Hill, speaker p.c.) Iskáh ú’te arakyé:’na:k wahęk’tikęhrę́:’nehƟ not arrow would I have I forgot ‘I didn’t have any arrows; I forgot [to bring them up].’ Wahrę́hrę:’, “Ęktí:’a:k, He said, “I’ll shoot’, tisnę’ kanyú’ yu’nękuhnęhá:’nye’ ha’ ú’te ęhšyé:nę’. and when it is going by the arrow you’ll catch it and when the arrow goes by, you’ll catch it.” In English, definite articles are used when the speaker believes the hearer can identify the referent. This identifiability can come from various kinds of circumstances. (13) Identifiability Extralinguistic context Shared knowledge Uniqueness Prior mention Association with prior mention

The garden looks great. I already fed the dog. Don’t look at the sun. I wore a coat and scarf. The coat was red. I bought a car. The paint’s not great.

Tuscarora ha’ can be used for referents identifiable from extralinguistic context. (14) Tuscarora place of speech act (Elton Greene, speaker p.c.) (‘They reached the land.’) Há:ne:’ ha’ kę́:ne:’ kyè:ní:kę: tyu’wna’kę́hra’r. that art here this point of land ‘That’s this land here (North America).’

950

Marianne Mithun

(15) Tuscarora time of speech act (Elton Greene, speaker p.c.) Čwé:’n ha’ čhę́’ kayetá:kre:’ ękwahę̀:we, . . still the right now they live real person ‘And they are still living here right now.’ But shared knowledge and uniqueness are not sufficient to warrant the article. It is used with proper names only if their referent has already been mentioned in the discourse. At the first mention of Abraham, in example (16), there was no ha’. (16) Tuscarora shared knowledge, uniqueness (Elton Greene, speaker p.c.) Wa’ka’rihę́:tyę’ uhyatęhstatukę́hti’, I word made holy book ‘I read in (the) Bible urihwaká:kę’ rayá:θęh Abraham, old time he is named Abraham about a man named Abraham, in old times, …’ Once Abraham had been introduced into this discourse, however, the speaker used the article before his name. It means more specifically ‘the aforementioned’. (17) Tuscarora article (Elton Greene, speaker p.c.) Há:ne:’ hè:ní:kę: ha’ Abraham wahrę́hrę:’, “…” that one that the Abraham he said ‘Then (the aforementioned) Abraham said, “…” It is the referent that must have been mentioned, not the particular lexical item. At the first mention of the Great Spirit in the passage below there was no article, but at the second there was, even though a different lexical item was used to refer to him. And again, there is an article before the previously mentioned Abraham. (18) Tuscarora article (Elton Greene, speaker p.c.) (‘Where did he get the idea to sacrifice, to give thanks?’) Stá:kwi’ ru’nękuhnáhkę. ‘From (the) Great Spirit. Hé’thu thru:yę́hw’ę kyè:ní:kę:. That’s where he learned this, ha’ Abraham ra’nyę̀:wáhnę ha’ uhę́’nę’ rawę̀:ní:yu:’. the Abraham he sacrifices the in front his word is great from Abraham sacrificing before (the aforementioned) God.’

Grammaticalization and polysynthesis: Iroquoian

951

The Tuscarora article is also used for a referent identifiable by association with a previously mentioned entity. The headman below was identifiable from the mention of the group. (19) Tuscarora article (Elton Greene, speaker p.c.) (A large group was walking on the ice, near starvation, running out of food.) Wa’kayę’na’natkáhri’θ ha’ uhę́’nę’ ru’nę́’ę, they told him the in front he had become ‘They told the headman, “…” In sum, the Tuscarora article ha’ is used for referents identifiable from extralinguistic context, prior mention, or association with some other previously mentioned entity. The likely diachronic origin of this particle still survives in the language: the anaphoric demonstrative há:ne:’ ‘that one’. In (20) it refers back to the incorporated noun ‘land’. (20) Tuscarora demonstrative source (Elton Greene, speaker p.c.) Yahwa’kayehéhnara’. they land reached ‘They reached the land. Há:ne:’ ha’ kę́:ne: kyè:ní:kę: tyu’wna’kę́hra’r. that the here this point of land ‘That’s this land here.’ Diessel (1999: 95–96) provides a useful discussion of anaphoric demonstratives, coreferential with a noun or noun phrase in the previous discourse. Unlike exophoric demonstratives, used to orient the hearer in the outside world (Put it there), anaphoric demonstratives serve a language-internal function: they track participants through discourse. It is easy to see how such a demonstrative could develop into an article meaning ‘the aforementioned’. The development of demonstratives into definite articles is a well-known phenomenon, famously discussed by Greenberg (1978). It involves the kinds of processes frequently cited under grammaticalization: the functional development from concrete to abstract to textual meaning, and phonological reduction (há:ne:’ > ha’). This is in fact quite similar to the well-known path seen in Romance (Epstein 1993; Hawkins 2004; Heine and Kuteva 2006). There were no articles in Classical Latin. The Latin demonstratives ille (masculine) and illa (feminine) ‘that one’, developed into the Old French articles le, la ‘the aforementioned’. These finally developed into the Modern French articles le, la, which do not require previous mention. The Tuscarora system is not unlike that of Old French. Other Northern Iroquoian languages have also developed articles, perhaps from the same source. Mohawk, like related languages, contains a discourse anaphor né:

952

Marianne Mithun

‘that one’. An example of its use to refer to a whole previous discussion can be seen below. (The entire discussion was in Mohawk.) (21) Mohawk discourse anaphor né (Watshenní:ne Sawyer, speaker p.c.) (‘The Mohawk men would go to work building the bridge, and the women and children would go over to the island. There the women would cook and do laundry, and we kids would swim. When the men finished work at the end of the day, they would come over and wash up, and all the families would eat together.’) Né: ki: thí:ken thò:ne Then that this that ‘That was when shontahontáhsawen thi: áhskwa’. when they started that bridge they started that bridge.’ The anaphoric demonstrative has developed into the shorter, unstressed article ne ‘the aforementioned’, which is further reduced before a vowel to just n. In the example below, at first mention of the men, there was no article, but later on, at second mention, the article appeared. (22) Mohawk ne ‘the aforementioned’ (Watshenní:ne Sawyer, speaker p.c.) a. First mention: no article Sok onkwehón:we wahotiio’ténhsera’. so real.person they.went.to.work ‘So (the) Mohawk men went to work.’ b. Later ne ‘the aforementioned.’ [‘At the end of their workday] na’tathón:ne’ ne ronón:kwe … they came back the aforementioned male persons the men came back …’ It might be noted that neither onkwehón:we ‘Native’ nor ronón:kwe ‘men’ carries a distributive here. This speaker was talking about earlier traditions and generic groups of people, not highlighting individual men. Some younger speakers, who are bilingual, are now extending the use of ne to contexts similar to those of English the, which simply marks referents as identifiable. This may be an effect of contact or simply a regular pathway of change similar to that documented for French.

Grammaticalization and polysynthesis: Iroquoian

953

3 Verbs The verbal morphology of the Iroquoian languages has remained remarkably stable. Most can be reconstructed for Proto-Iroquoian, perhaps spoken several thousand years ago. The innovations in individual languages since that time have occurred primarily at the edges of the template.

PRE - PRO PREFIXES

PRO PREFIX

REFLEXIVE MIDDLE PREFIX

NOUN STEM

VERB ROOT

DERIVATIONAL SUFFIXES

ASPECT SUFFIXES

FINAL SUFFIXES

Fig. 3: Verb Template.

The only Southern language, Cherokee, has added an initial conditional prefix y-, and the Northern Iroquoian languages have added an initial contrastive prefix th-. Cherokee has added final mode suffixes, and the Northern languages have added final tense-aspect suffixes. But other grammaticalization processes have continued within the existing morphological structure, much of it similar to developments observed elsewhere in the world.

3.1 Middles A middle prefix can be reconstructed for Proto-Iroquoian. (23) Iroquoian middle *-atCherokee -at-, -ali-, -ataaTuscarora -at, -a’nFive Nations -at-, -ar-, -anAs in other languages, it indicates that the semantic agency and affectedness are merged in a single participant. (Cherokee examples drawn from Montgomery-Anderson [2008] are from Oklahoma; those drawn from King [1975] and Cook [1979] are from North Carolina.) (24) Cherokee middle (Montgomery-Anderson 2008: 347, 348) a. Hiiyawóʔa. hii-awóʔa 2sg.agt> 3anim-bathe.pres ‘You are bathing him/her’

954

Marianne Mithun

b. Hatawóʔa. hi-at-awóʔa 2sg.agt-middle-bathe.pres ‘You’re bathing, swimming’ (25) Cherokee middle (Montgomery-Anderson 2008: 366) a. Akeéhya aàsteeliíta. akeéhya a-steeliíta woman 3sg.agt-help.pres ‘He’s helping the woman.’ b. Aànehlti uutaastehlti. a-anehltiʔa uu-ataa-stehlt-i 3sg.agt-try-pres3sg.agt-middle-help-nmlz ‘He’s trying to help.’ (26) Cherokee middle (Montgomery-Anderson 2008: 370) a. Taàstluyska. tee-a-stluska dist-3sg.agt-split.pres ‘He is splitting it.’ b. Taàlstluska. tee-a-al-stluska dist-3sg.agt-middle-split.pres ‘It is splitting.’ Verb stems containing middle markers have a strong tendency to lexicalize as single chunks. In many cases, middle markers continue as part of lexicalized verb stems, in which the semantic contribution of the middle can still be discerned, but the root no longer exists on its own. (27) Cherokee lexicalized middles (Montgomery-Anderson 2008: 375) -alvvteehíha ‘faint’ -alistuhvska ‘to bud’ -atóòkhtíha ‘stagger’ -ateehv́ska ‘be born’ -atanasiíni ‘crawl’ -atiikháha ‘urinate’ -alihiha ‘fight’ -alihyvv́sánaàwstíha ‘snort’ Northern Iroquoian middles serve similar functions.

Grammaticalization and polysynthesis: Iroquoian

(28) Mohawk middles a. Basic transitive Middle b. Basic transitive Middle c. Basic transitive Middle

955

s-áhseht ‘Hide it!’ s-at-áhseht ‘Hide!’ s-éta’ ‘Put it in!’ s-at-íta’ ‘Get in!’ s-kétsko ‘Raise it!’ s-at-kétsko ‘Get up!’

Middle markers are pervasive through the lexicon. (29) Mohawk middles katá:tis tkatáweia’t katá:wenhs katéhens katè:kwahs katkáhthohs katkèn:se’ katkón:hens kató:rats katshó:rihs katshwà:tha’ tkáttiks kattó:kahs kattókha’ katonhkária’ks katón:ron katónrie’ iekatshnénhtha’ wakathón:te’ wakatì:wen wakatétshen wakatshennón:ni

‘I talk’ ‘I come in’ ‘I swim, bathe’ ‘I am embarrassed’ ‘I escape’ ‘I look’ ‘I taste it’ ‘I put my head down’ ‘I hunt’ ‘I slurp’ ‘I play’ ‘I vomit’ ‘I feel’ ‘I am knowledgeable’ ‘I am hungry’ ‘I am adept’ ‘I breathe’ ‘I descend’ ‘I hear’ ‘I am skinny’ ‘I am greedy’ ‘I am happy’

As in other languages, there is a tendency for middle stems to lexicalize, and for the boundary between the middle prefix and the root to fade. (30) Mohawk middles -ataweia’t ‘enter’ -atehen ‘be embarrassed’ -atek ‘be burning’ -atenion ‘stretch’

956

Marianne Mithun

-atetsh-ate’kw -atkahrit -atkatston -atkennis -atorishen -ato -atikonhen -athontat

‘be greedy’ ‘escape’ ‘play’ ‘make soup’ ‘assemble (intr)’ ‘rest’ ‘swell’ ‘put one’s head down’ ‘obey’

In some such stems, the semantic element contributed by the middle marker can still be imagined, and in some, the root still occurs on its own. (31) Mohawk middles with identifiable roots a. -at-o ‘swell’ -o ‘be in water’ b. -at-hont-at -hont-at

‘obey’ (prick up one’s ears) ‘ear’ ‘stand’

In many others, however, the root no longer exists independently. Middle stems, like other lexical items, can continue to evolve semantically on their own. One might wonder, for example, why a verb ‘sing’ would contain a middle marker: -ate-rennoten. A closer look at its internal structure provides an answer. (32) Mohawk lexicalization Saterennó:ten! ‘Sing!’ s-ate-renn-oten 2sg.agt-middle-song-stand ‘Stand up your song!’ The source of the middle prefix can still be discerned: as in many other languages, it originated as a reflexive. Some verbs with this prefix are still close to reflexives in meaning. (33) Mohawk Wa’katkonhsóhare’. wa’-k-at-konhs-ohar-e’ factual-1sg.agt-refl-face-wash-pfv ‘I face-washed myself ’ = ‘I washed my face.’

Grammaticalization and polysynthesis: Iroquoian

957

In part because of the tendency of middle stems to lexicalize and evolve semantically, the original function of the marker can become obscured. Already in Proto-Iroquoian, the reflexive had been renewed by reduplication: *-atat-. (34) Cherokee reduplicated reflexive (Montgomery-Anderson 2008: 344, 377) a. Aàwataakhthoósthánv awvv́sa. akw-ataat-akahthoósthán-vv́ʔi aki-vv́sa 1sg.pat-refl-look.at-exp 1sg.pat-self ‘I looked at myself.’ b. Aàtaatoohlvvska. a-ataat-oohlvvska 3sg.agt-refl-make.pres ‘It is regenerating itself’ (e.g., a lizard) (35) Mohawk reduplicated reflexive a. wa’kká:ri’ ‘I bit (it)’ wa’k-atat-ká:ri’ ‘I bit myself ’ b. wa’tkáwe’ste’ wa’tk-atat-áwe’ste’

‘I pricked (it), pierced (it)’ ‘I pricked myself’

c. wa’kòn:reke’ wa’k-atat-kòn:reke’

‘I hit (it) (as with a hammer)’ ‘I hit myself’

With some verbs, the reduplicated form still has an emphatic meaning, like Mohawk wa’k-atat-konhsóhare’ ‘I washed my own face.’

3.2 Applicatives Several Iroquoian derivational suffixes add an argument to the core. Benefactive applicatives transitivize verbs by adding a beneficiary, which is coded as a grammatical patient. The forms of the benefactive suffixes are lexically conditioned and vary with aspect. The Mohawk suffixes have the shapes -awi, -ni, -s, -hs, -hahs, -enni, -en ‘for, to’. (36) Mohawk benefactive applicatives a. ke-natà:re-’s ‘I visit’ khe-natahren-á:wi-hs ‘I visit her’ khe-natahren-á:wi-’ ‘I have visited her’ b. Se-ri’wanón:ton She-ri’wanontón:-ni-’

‘You have asked.’ ‘You’ve asked/are asking her’

habitual ben.habitual ben.stative stative ben.st

958

Marianne Mithun

c. Se-nà:ton She-na’tón:-ni-’

‘You’ve pointed it out.’ stative ‘You’ve shown/are showing it to her’. ben.stative

These suffixes still have discernible sources in verb roots. The benefactive -awi is, as in many languages, descended from a verb root ‘give to’. (37) Mohawk verb root -awi ‘give to’ Kheiá:wi. khei-awi 1sg>f.sg-give.to.st ‘I have given it to her.’ The verb root shows the same idiosyncratic allomorphic alternations for aspect as the benefactive applicative: -awi in habitual and stative aspects, and -on in perfectives and imperatives. The benefactive -ni is descended from a verb root ‘lend to’. (38) Mohawk verb root -ni ‘lend to’ Wahákeni’. wa-hak-ni-’ factual-m.sg>1sg-lend.to-pfv ‘He lent it to me.’ Both the verb roots and their descendant applicative suffixes can be reconstructed for Proto-Northern-Iroquoian. The languages also contain instrumental applicatives, among them -(h)st. (39) Mohawk instrumental applicative a.

iera’wistakarháthohs ie-ra’wist-a-karhat-ho-hs generic.agt-pancake-linker-turn-caus-habitual ‘she/one flips pancakes’

b. iera’wistakarhathó-hst-ha’ ie-ra’wist-a-karhat-ho-hst-ha’ generic.agt-pancake-linker-turn-caus-ins.appl-habitual ‘one flips pancakes with it’ = ‘spatula’ Iroquoian languages never show conversion at the root or stem level. Only noun stems can serve as the foundation of morphological nouns, only verb stems serve as the foundation of morphological verbs, and there is no overlap in the two stem

Grammaticalization and polysynthesis: Iroquoian

959

categories. There is, however, extensive conversion at the word level: morphological verbs often come to be used as referring expressions without further derivation. The instrumental applicative -hst is pervasive in Mohawk in terms used to designate the function of an object. (40) Mohawk instrumental applicative in referring expressions iontenaktorók-st-ha’ ‘one covers one’s bed with it’ = ‘bedspread’ iehwista’ék-st-ha’ ‘one strikes metal with it’ = ‘bell’ iene’konhré(k)-hst-ha’ ‘one pounds with it’ = ‘hammer’ ienawa’tstarhó-st-ha’ ‘one spreads mud with it’ = ‘trowel’ ieristarhó-st-ha’ ‘one covers with metal with it’ = ‘(flat)iron’ iontekhwahra’tsherorók-st-ha’ ‘one table covers with it’ = ‘tablecloth’ iontkon’sera’tsherorók-st-ha’ ‘one pillow covers with it’ = ‘pillowcase’ iakotà:-st-ha’ ‘one sleeps with it’ =‘pajama(s)’ ionrahsi’tohrók-st-ha’ ‘one foot inserts with it’ = ‘slipper(s)’ ieksata’á-hst-ha’ ‘one dish inserts with it’ = ‘china cupboard’ This applicative also has a discernible origin in a verb root. (41) Proto-Northern-Iroquoian *-hst ‘use’ a. Tuscarora kíhstha’ k-ihst-ha’ 1sg.agent-use-hab ‘I use it’ b. Mohawk kátstha’ k-át-st-ha’ 1sg.agent-middle-use-hab ‘I use it.’ Another Northern Iroquoian instrumental applicative suffix has the shape *-hkw. (42) Mohawk instrumental applicative -hkw a. khenà:tonhs ‘I call out to her’ khena’tón-hkhw-a’ ‘I call her by name with it’ = ‘I call her X’ b. wákhthare’ wakhthará-hkw-en

‘I am talking’ ‘I am talking about it’

c. kónhnhe’ konhnhé-hkw-en

‘I am alive’ ‘I live on it’

960

Marianne Mithun

This applicative also appears pervasively in terms for objects defined in terms of their uses. (43) Mohawk instrumental applicative -hkw iakenheion’taráhkhwa’ ‘one inserts the dead in with it’ = ‘casket’ ienentstakwararáhkhwa’ ‘one inserts dirty laundry with it’ = ‘hamper’ iontahkwenniaráhkhwa’ ‘one puts one’s clothes in with it’ = ‘closet, trunk’ iontahkweniharáhkhwa’ ‘one hangs one’s clothes with it’ = ‘hanger’ teiontska’hónhkhwa’ ‘one dines with it’ = ‘kitchen, restaurant’ tsiera’wistáhkhwa’ ‘one pierces with it’ = ‘fork, pitchfork’ ienon’tawerontáhkhwa’ ‘one pours milk with it’ = ‘milk pitcher’ ienontarotsienhtáhkhwa’ ‘one fetches soup with it’ = ‘ladle’ teie’nhonhsawenrie’táhkhwa’ ‘one stirs eggs with it’ = ‘egg beater’ teionthiohsahráhkhwa’ ‘one sets one’s elbows with it’ = ‘small armchair’ The source of this applicative also still persists as a verb root in the modern languages and can be reconstructed for Proto-Northern-Iroquoian. (44) Mohawk verb root -hkw ‘pick up’ té:sehkw ‘Pick it up!’ tekéhkhwa’ ‘I pick it up’ ‘I will pick it up’ tén:kehkwe’ tewakéhkwen ‘I have picked it up’

imperative habitual perfective stative

The allomorphs of the aspect markers that occur with this verb root are the same as those that occur with verb stems ending in the -hkw instrumental applicative. (The duplicative prefix te- that occurs with this verb indicates a change of position.) The source of an instrumental applicative in a verb meaning ‘pick up’ is not surprising: one picks something up to use it. There is evidence that all of these applicatives developed from roots that had already become bound, in verb-verb compounds, before developing into derivational suffixes. Modern Iroquoian languages no longer have productive verb-verb compounding, though a few modern roots show traces of earlier compound structure, such as -oha ‘put in water’, probably composed of the verb roots -o- ‘be in water’ and -ha ‘put, take’. The languages do, however, show noun-verb compounding, or noun incorporation, which can be reconstructed for Proto-Iroquoian. It is no longer productive in Cherokee, but it remains highly productive and pervasive in the Northern languages. These compound constructions show a particular phonological pattern: if the initial (noun) stem ends in a consonant, and the following verb root ends in a consonant, an epenthetic vowel or linker -a- is inserted to break up the

Grammaticalization and polysynthesis: Iroquoian

961

potential consonant cluster. The same vowel appears between verb roots ending in a consonant and a following instrumental applicative: -(a)hst- and -(a)hkw. We know that neither of the two source roots, -hst ‘use’ and -hkw ‘pick up’ contains this vowel. The various languages show epenthetic vowels in other contexts, but they are different vowels: i or e.

3.3 Causatives One of the instrumental applicatives, -(a)hst, has been extended to add a related kind of argument: a cause. This is not altogether surprising: instruments can be seen as secondary causers. What is slightly surprising is that while applicatives add a grammatical patient argument, causatives add a grammatical agent. (45) Mohawk causative use a. Teiótsion te-io-i’tsion dv-neuter.pat-be.dirty ‘It is dirty.’ b. Tsí’tsion-hst ts-i’tsion-hst 2sg.agt.imper-be.dirty-caus ‘Get it dirty!’ c. Wentó:re’ ‘It is difficult, it is hard.’ Sentó:r-a-hst ‘Make it difficult!’ d. Watò:roks ‘It shrinks.’ Satò:rok-st ‘Shrink it!’ e. Ioneniò:kwano Seneniò:kwano-hst Seweién:te’ Satáweien-hst

‘It is frozen.’ ‘Freeze it!’ ‘You know how.’ ‘Study! Learn! Practice!’

(46) Cherokee causative use (Montgomery-Anderson 2008: 359–360) a. Jiíyu uùnoóyv́`v je jiíyu uu-noóyv́v`j-éʔi boat 3sg.pat-sink.pfv-non.exp ‘The boat sank.’ b. Jiíyu Iìtiinoóyv́sta. jiíyu iìtii-noóyv́-st-a boat 1pl.agt-sink-caus-imm ‘We sank the boat.’

962

Marianne Mithun

c. Atsv:’ka. Atsv:’-ihst-iha.

‘It is spilling’ ‘He is pouring it out’

d. Akhiyvstéestiiha. Ahyvvtlatiistiisko.

‘He’s getting me drunk.’ ‘It makes it cold.’

e. Uùneestaltiistanv. Toòkakoohvstanv.

‘He froze it.’ ‘We burned them.’

The meaning of the lexical source of the suffix -hst, ‘use’, suggests that the instrumental applicative use developed before the causative use. The second applicative suffix -(a)hkw shows an incipient use as a causative in one of the Northern Five Nations languages, Onondaga. (47) Onondaga causative (Woodbury 2003: 407) -ęnihnyę(t)‘act crazy, be silly’ -ęnihnyęt-hkw‘drive someone crazy’ teshakonihnyętahkwih te-shako-ęnihnye-a-hkw-ih dv-3m.sg>3 f.sg-act.crazy-caus-stative ‘He drives her crazy.’ The lexical origin of the suffix -hkw- in a verb root meaning ‘pick up’, a common source of instrumental applicatives cross-linguistically, coupled with the robust use of the suffix throughout the Northern Iroquoian languages as an applicative but incipient use as a causative, in only one of the languages, again indicates a development from applicative to causative. Another pervasive causative suffix throughout the family has the shape -(ᴧ)ht. (48) Mohawk causative Ionà:nawen. Senà:nawen-ht.

‘It is wet.’ ‘Moisten it!’

(49) Cherokee causative (Cook 1979: 138–139) Koho:hska. ‘It (round) is falling.’ Koh-vht-iha. ‘He is dropping it (something round).’ The source of this causative suffix can still be seen in a Cherokee verb root: -vht ‘use’. (50) Cherokee verb -vht ‘use’ (King 1975: 89) kvhtihskv:ki ka-vht-ihsk-v:ki 3sg.agt-use-ipfv-assertive ‘he was using it’

Grammaticalization and polysynthesis: Iroquoian

963

In the Northern languages, the causative -(a)ht shows the same phonological behavior as the instrumental applicatives -(a)hst and -(a)hkw. If it follows a consonantfinal verb root, an epenthetic linker vowel -a- is inserted, again suggesting that the root ‘use’ was compounded with other verb roots before further developing into a suffix. The meaning of the source root, ‘use’, suggests that this causative is the result of a development like the others, from an instrumental applicative to a causative. We can thus add an alternative direction of development to that observed by Shibatani and Pardeshi (2002: 119), where causatives developed into applicatives.

3.4 Tense The development of some tense markers within Iroquoian also shows patterns of grammaticalization observed elsewhere in the world, but here the developments occurred in markers that were already fused within the verb and reduced phonologically. Cherokee contains a future tense prefix t(a)(51) Cherokee future (Montgomery-Anderson 2008: 329, 330) a. Nihi thihwahtvvhi. nihi t-hi-hwahtvvh-i 2 future-2sg.agt-find-mot ‘You will find it.’ b. Takvv`thaniisáhni. ta-ji-vhthan-iisáhn-i future-1sg.agt-use-pfv-mot ‘I’m going to use it again.’ The source of this tense prefix is clear: a cislocative prefix ‘hither’. The cislocative persists with this meaning in all languages of the family including Cherokee, and it can be reconstructed with this meaning for Proto-Iroquoian. It developed into a future only in Cherokee, with no change in its position within the verbal template. (52) Cherokee cislocative (Montgomery-Anderson 2008: 328) a. Takinatansiinooheéli. ta-kinii-atansiinoo-heéli cislocative-1pl.pat-crawl-appl-pres ‘He’s crawling to us.’ b. Katoòhv tiìjakhthahv́vse? katoòhv ti-iijii-akahthahv́vs-éʔi why cislocative-2pl.agt-turn.back-non.exp.past ‘Why did you turn back (this way)?’

964

Marianne Mithun

The semantic development is the familiar one from space to time. Heine and Kuteva (2002: 308–309) cite the development of independent venitive particles into future tense markers in two other languages, Iraqw and Maa, noting that the two languages are from different language phyla but may be areally related. The first is a Cushitic language spoken in Tanzania, and the second an Eastern Nilotic language spoken in Northern Tanzania and Southern Kenya. The Iraqw ni signals near future; the Maa -ʊ(n) developed first into an inchoative and then to a future with verbs of state (Tucker and Mpaayei 1955: 141; König 1993: 294–316, cited in Heine and Kuteva 2006: 309). The Cherokee development adds a genealogically and areally unrelated example of such a pathway of grammaticalization, but one in which the marker has already undergone considerable grammaticalization, becoming attached to the verb before this semantic development. A different development from space to time can be seen in the Northern languages. Here all verbs except imperatives are inflected for aspect: habitual, stative, or perfective. Perfectives are further inflected for tense/modality: future, irrealis, or past (often termed factual in the literature). (53) Mohawk perfectives en-kká:we’ ‘I’ll paddle’ a:-kká:we’ ‘I should/might/could paddle’ wa’-kká:we’ ‘I paddled’

future irrealis past

There is a discernible source for this past tense prefix within the morphology itself. It is descended from the Proto-Iroquoian translocative prefixes *w-/*h-/*y- ‘thither, in a distant place’. (The different forms correspond to allomorphs in the modern languages distributed according to aspect and varying across the languages.) The translocatives remain in place in all branches of the family. (54) Cherokee translocative (Montgomery-Anderson 2008: 307–312) Waàwatéeka w-aki-atéeka transloc-1sg.pat-throw.pres ‘I’m throwing it there, away’ (55) Tuscarora translocatives (Elton Greene, p.c.) wa’-θúha ‘Put it in water!’ we-hrúhahs ‘he puts it in water’ we-hraúhę ‘he has put it in water’ yah-wahrúha’ ‘he put it in water’ y-ęhrúha’ ‘he’ll put it in water’ y-ahrúha’ ‘he should put it in water’

imperative habitual stative perfective: past perfective: future perfective: irrealis

Grammaticalization and polysynthesis: Iroquoian

965

In the modern Northern languages, the past tense prefix is waʔ-/wa-/w- in Tuscarora and the eastern languages Onondaga, Oneida, and Mohawk, with initial loss of the w in the western languages Cayuga and Seneca and Huron, part of a general pattern of loss of initial glides. In Mohawk and Oneida the translocative is now generally yaʔ-/ya-/y-, and in Onondaga, Cayuga, Seneca, and Huron it is haʔ-/ha-/h-. There are, however, relict forms of what now appear to be the past tense prefix in w- with translocative meaning. The verb -e- ‘go’ appears with this prefix even when the verb is not in the past. (56) Northern Iroquoian relicts: Mohawk Wá:s ‘Go! Get away! Get out of here!’ Ka’ wá-hse’? ‘Where are you going?’ Tiohtià:ke wà:ke’? ‘I’m going to Montreal.’ This prefix also occurs on verbs containing an andative suffix ‘going to’, whether or not the tense is past. These prefixes which retain the directional sense show the same allomorphy as the modern past tense. The development of this past tense prefix in the Northern languages from a translocative directional prefix shows the kind of semantic development frequently observed across languages: from concrete space to more abstract time.

4 Beyond the morphology Though the Iroquoian languages are generally highly polysynthetic, showing effects of pervasive grammaticalization processes, they also show grammatical developments beyond the boundaries of the word. Three such developments are traced here, one pertaining to locative constructions, one to modality, and one to the development of complement constructions.

4.1 Place Iroquoian languages contain well-developed constructions that look at first glance like locative adverbs. (57) Mohawk locatives a. Saréka ohròn:wakon ia’onkeniia’tèn:’en. sareka o-hronhw-akon i-a’-onkeni-ia’t-en-’-en almost neuter-ditch-in transloc-factual-1pl.pat-body-lie-inch-st ‘We almost fell into the ditch.’

966

Marianne Mithun

b. Nen ki’ onhwéntsiakon nen ia’akení:ien’. onen ki’ o-onwents-akon onen i-a’-akeni-ien-’ then in.fact neuter-earth-in then transloc-factual-1du.agt-set-pfv ‘We put them in the basement.’ c. Kanónhsakon ia’katáweia’te’. ka-nonhs-akon i-a’-k-at-aweia-’t-e’ neuter-house-in transloc-factual-1sg.agt-mid-be.in-caus-pfv ‘I went into the house.’ d. Kaná:takon ieiákene’skwe’. ka-nat-akon i-etaken-e-’s-kwe’ neuter-town-in transloc-1incl.du.agt-go-st.dist-past ‘We were in town.’ It appears at first that -akon is a locative suffix ‘in’ added to nouns in place of the basic noun suffix: compare the noun ohròn:wa’ ‘ditch’. There are other similar markers. (58) Mohawk locatives kanákta’ ‘bed’ kanakt-à:ke ‘on the bed’ kanákt-akon ‘in the bed (crib)’ kanakt-ó:kon ‘under the bed’ kanakt-ákta’ ‘next to the bed’ The origins of these constructions can still be recovered. They are descended from verbs with incorporated nouns, with the same structure as that below. (59) Mohawk noun incorporation Wahatihwistarò:roke’. wa-hati-hwist-a-rohrok-e’ factual-m.pl.agt-money-linker-gather-pfv ‘They collected money.’ In all noun incorporation, only the noun stem is incorporated, not the full noun word. The stem in the above verb is -hwist-, without the neuter gender prefix o- or the noun suffix -a’. (60) Mohawk noun ohwísta’ o-hwist-a’ neuter-money-noun.suffix ‘money’

Grammaticalization and polysynthesis: Iroquoian

967

As seen earlier, when an incorporated noun stem ends in a consonant, and the incorporating verb begins in a consonant, an epenthetic linker vowel -a- is inserted between the two to break the potential consonant cluster. This linker can be seen in ‘they money-collected’. Stress in Proto-Northern Iroquoian was basically penultimate. The linker was added sometime after this basic stress pattern was established: if it occurred in an open syllable, it did not count in the determination of stress. As noted above in section 3.2, there is no conversion at the root or stem level in Iroquoian languages, but there is extensive conversion at the word level: morphological verbs are frequently used as referring expressions without further derivation. (61) Mohawk morphological verb ionthnenhsotarhókstha’ ie-at-hnenhs-otarhok-hst-ha’ generic.agt-middle-shoulder-drape-ins.appl-hab ‘one drapes one’s shoulders with it’ = ‘suspenders’ If a morphological verb stem is to be incorporated, it must be overtly nominalized. The word for ‘table’ is a morphological verb, but it is used as a referring expression, as a syntactic nominal. (It itself contains an incorporated noun.) (62) Mohawk morphological verb atekhwà:ra’ o-ate-khw-a-hra’ neuter-middle-food-linker-set ‘it is food set on’ = ‘table’ When it is incorporated, it occurs with the nominalizer suffix -’tsher-. (63) Mohawk incorporated nominalization watekhwhahra’tsherí:io w-ate-khw-a-hra-’tsher-iio neuter-middle-food-linker-set-nominalizer-be.good ‘it is a good table’ We are now in a position to see the origin of the place constructions illustrated above: they originated as stative verbs with incorporated nouns. (64) Mohawk locative verb roots -(a)kon ‘be inside’ -okon ‘be under’ -akta’ ‘be next to’ -ihen ‘be in the middle of ’

968

Marianne Mithun

-(a)here’ -(a)ti

‘be on top of ’ ‘be beyond’

The incorporated nouns indicate the point of reference. In (57), cited earlier and repeated below, the locative ‘be in’ is preceded by the epenthetic linker vowel -a-, the same linker that occurs between incorporated nouns and verbs like that in ‘they money-collected’ above; even though it appears to be in the penultimate syllable, it is not stressed. (57) Mohawk locatives a. Saréka ohròn:wakon ia’onkeniia’tèn:’en. sareka o-hronhw-akon i-a’-onkeni-ia’t-en-’-en almost neuter-ditch-in transloc-factual-1pl.pat-body-lie-inch-st ‘We almost fell into the ditch.’ Furthermore, the nouns in this construction must be morphological noun stems: if they are not, they must be overtly nominalized, just as in other noun incorporation. (65) Mohawk nominalized location atekhwahra’tsheró:kon o-ate-khw-a-hra-’tsher-ok-on neuter-mid-food-linker-set.on-nmlz-be.under-st ‘it is table-under’ = ‘(place) under the table’ Like other verbs, these have come to be used, by conversion, to designate places. They differ subtly from nouns with locative case endings in other languages. They are not relational: they designate places, so they most often situate an event or state in space, but they can also serve as core arguments of a clause. (66) Mohawk place Né: ki’k ní:’ kana:takon nitienté:ri. ne ; ki’=k n=i:’ ka-nat-a-kon ni-k-ienteri it.is in.fact=only the=1 neuter-town-linker-be.in prt-1sg.agt-know ‘I myself just know the village.’ These place constructions originated as clauses, full finite adverbial clauses, but they have undergone several grammaticalization processes. As seen earlier, both nouns and verbs contain prefixes: those on nouns indicate (neuter) gender, and those on verbs indicate the gender, person, number, and role of the core argument(s). The neuter prefixes on nouns are quite similar to those on verbs, with similar patterns of allomorphy, but they do not match perfectly. Those on nouns

Grammaticalization and polysynthesis: Iroquoian

969

lack initial glides. (The letter i represents the palatal glide [j] before a vowel in the standard Mohawk orthography.) (67) Mohawk neuter prefixes Noun ka-nákta’ Verb ka-nóhares

‘bed’ ‘it washes it’

Noun Verb

à:rya’ w-akátste’

‘fish hook’ ‘it is strong, enduring’

Noun Verb

o-hnénhsa’ io-’tarihen

‘shoulder’ ‘it is hot’

Noun Verb

aw-en’náhsa’ iaw-ékon

‘tongue’ ‘it is delicious’

The prefixes on the place words described above, descended from verbs with incorporated nouns, show incipient decategorialization. They have the prefix shapes typical of nouns, without the glides. The pronominal prefixes on verbs distinguish grammatical agent and patient roles. The alternation between the agent or patient paradigm is to a certain extent semantically motivated. Those initiating and controlling events and states are represented by agent forms, and those affected by them but not in control are represented by patient forms. Furthermore, agent prefixes (ka-, w-) tend to appear with inherent states, while patient prefixes (io-, iaw-) tend to appear with resultant states: ka-hòn:tsi ‘it is black’, ió:ri ‘it is ripe’. But the paradigms are lexicalized with each verb stem. When there is an incorporated noun, it is normally still the incorporating verb that determines the agent or patient pronominal paradigm, not the incorporated noun. (68) Mohawk prefixes Noun ó-shes ‘syrup’ Verb ka-hòn:tsi ‘it is black, dark’ Verb ka-shest-a-hòn:tsi ‘it is dark syrup’ = ‘molasses’ With the place constructions, however, it is now the noun stem that determines the prefix category. (69) Mohawk place expressions Noun ka-nákt-a’ ka-nákt-akon ka-nakt-ó:kon Noun

o-’neróhkw-a’ o-’neróhkw-akon o-’nerohk-ó:kon

‘bed’ ‘(place) inside the bed (crib)’ ‘(place) under the bed’ ‘box’ ‘(place) inside the box’ ‘(place) under the box’

970

Marianne Mithun

Furthermore, the place expressions can no longer predicate on their own. To predicate a location, an additional verb must be used. (70) Mohawk locative predication Tanon’ wahsekhró:ri’ ‘And you told me ka’ nón: ohròn:wakon ka’ nonwe o-hronhw-a-k-on what place neuter-ditch-lk-be.inside-st where place inside ditch where the ditch was.’

ítken. i-t-ka-i proth-cisloc-neuter.agt-be.in.st it is there

The development of the Iroquoian place words thus shows a number of familiar types of grammaticalization processes. They began as verbs with incorporated noun stems; they show the same features as other noun incorporation constructions. The incorporated nouns must be formal noun stems, either noun roots or overtly nominalized stems. They show the epenthetic linker vowel -a- typical of incorporation constructions. But these constructions show decategorization at both the morphological and lexical levels. The erstwhile incorporating verb roots have undergone decategorization to become locative nominalizers, deriving terms for places. The pronominal prefix paradigm is now determined by the noun rather than the erstwhile verb, and the prefixes now have the (glide-less) forms typical of nouns. The full constructions can no longer predicate on their own.

4.2 Modality A type of modality often termed dynamic possibility or agent-internal ability pertains to the ability of the controlling participant to effect an action (Goossens 1985; Mithun 2016; Nuyts 2016). The functional development of this distinction within the Iroquoian family matches that traced for other languages with rich philological records, such as those of the Germanic branch of Indo-European. In the Northern Iroquoian languages, agent-internal ability is expressed, not surprisingly, with the verb -kweni ‘be able’. (71) Mohawk agent-internal ability a. A:kkwé:ni’ ahsontakwé:kon a:khthá:ren’. I would be able all night I would talk ‘I could talk all night long.’ b. Iáh tha:hatikwé:ni’ ni:’ taionkhe’nikonhrhá:ron. not would they be able 1 they would mind set us ‘They won’t be able to bother us.’

Grammaticalization and polysynthesis: Iroquoian

971

The modal meaning of this verb has developed from a more concrete origin: ‘beat (in a fight, game, etc.)’ With a middle voice prefix, the stem means ‘win’. (72) Mohawk stem ‘win’ Iah teionkhró:ri nahò: wa’katkwé:ni’. iah te-ionk-hró:ri naho’ten wa’-k-at-kwé:ni-’ not neg-indef>1sg-tell what factual-1sg.agt-middle-beat ‘They hadn’t told me what I won.’ The development of this verb shows a familiar grammaticalization process, from concrete to more abstract. This same verb shows a further semantic development in the western Five Nations languages Seneca, Cayuga, and Onondaga, from agent-internal ability to general possibility. When it indicates agent-internal ability as in the first example below, the pronominal prefix refers to the agent. When it indicates general possibility as in the second, the pronominal prefix is neuter, referring to the event qualified. (73) Seneca (Wallace Chafe, p.c.) a. Agent-internal ability: Agent pronoun ke- ‘I’ tę́ kę̀:hta:t ta’á:öh à:khät. A-ke-kwenyǫ:h ni:’ I would be able myself I will run not possible he would catch me ‘I can run so he can’t catch me.’ b. > General possibility: Neuter pronoun o- ‘it’ ta’áǫh ne’hoh na:yawęh. Ki’shęh o-kwe:nyǫ:h ̨ maybe it is possible not possible there so it would happen ‘It might be possible it wouldn’t happen that way.’ Interestingly, replica grammaticalization of the full pathway can be seen in Tuscarora. As noted earlier, the Tuscarora were the first group to separate from the Northern Iroquoian branch. They were initially encountered by Europeans in the Southeast, but at the beginning of the 18th century they moved north, ultimately settling in Western New York State on land offered to them by the Seneca. The Tuscarora root -kweni is easily identified as a loan from the Five Nations Iroquois languages, because it does not show the characteristic development of Proto-Northern-Iroquoian *n > t. (If it were native to Tuscarora, it would have the form **-kweti.) Nevertheless, this Tuscarora borrowed verb root shows all of the stages of semantic development detectable in its neighbor Seneca: concrete action, agent-internal ability, and agentexternal possibility. (74) Tuscarora -kweni (Elton Greene, speaker p.c.) Ękakwè:niʔ it will be possible

972

Marianne Mithun

ȩkayętęhnì:nę́:hek, they will continue to sell kyè:ní:kę: haʔ tà:wę́:te kayakyetì:yahs This the what they make ‘It will be possible [for them to sell what they make].’

4.3 Complement constructions Finally, effects of some of the same processes of development of complex sentence constructions seen elsewhere in the world can be observed in Iroquoian languages. In the Northern Iroquoian languages, there are no non-finite verb forms. All verbs can constitute full grammatical clauses in themselves, complete with arguments (in the pronominal prefixes) and aspect inflection. Complement clauses need not carry any overt marking beyond their prosody. The matrix always precedes the complement. (75) Mohawk unmarked complement Ne: ki’k wa’kehià:ra’ne’ kí:ken: ‘I was just thinking tsi niió:re’ tsi kiótkon ionkwahskéhnhen how we are always trying hard [onkwehonwehnéha’ aonsetewatá:ti’]. onkwe=honwe=hneha’ a-ons-etewa-atati-’ person=real=style irr-rep-1pl.incl.agt-talk-pfv [to speak Mohawk again]. The final line above, the complement, is a fully finite clause that could serve as a grammatical sentence in isolation: ‘We would speak Mohawk again.’ Mohawk does contain complement clauses with overt marking. (76) Mohawk marked complement: ne ‘So, our seniors group found a way for it to happen. It’s really hard, because we’ve been changed.’ Iakwate’niénhtha’ ‘We try [ne akwé:kon onkwehón:we ne akwekon onkwe=honwe the all person=real the all real person [to speak only Mohawk].’

a:iakwatewennón:tahkwe’]. aa-iakwa-ate-wenn-ont-a-hkw-e’ irr-1pl.exc.agt-mid-word-insert-lk-ins-pfv we would words put in with it

Grammaticalization and polysynthesis: Iroquoian

973

The particle ne is the article described in section 2.3, which is descended from the demonstrative né: ‘that aforementioned’ and which precedes basic nouns: ‘the aforementioned’. Its use has been extended from use before nouns to aforementioned clausal arguments: complements. The complements are treated like other referring expressions. The marker is only used, however, before complement clauses referring to ideas that have already been under discussion in the discourse. Tuscarora shows a parallel development. As in Mohawk, complement constructions need not be marked by more than prosody. (77) Tuscarora unmarked complement Ękakwè:ni’ it will be possible [ȩkayętęhnì:nę́:hek, ę-kaye-atęhninę-hek fut-3pl.agt-sell-cont they will continue to sell kyè:ní:kę: ha’ tà:wę́:te kayakyetì:yahs] kyenikę: ha’ tawęte kayak-ętiy-ahs this the what 3pl.agt-make-hab ‘It will be possible [for them to sell [what they make]].’ Here, too, the complement clause could stand alone as a fully grammatical, complete sentence: ‘They will continue to sell what they make’. Also, as in Mohawk, complements can be marked by an article ha’, descended from the demonstrative há:ne:’ ‘that’. In Tuscarora however, unlike in Mohawk, the complement no longer needs to be aforementioned. In the sentence below, the anaphoric demonstrative há:ne:’ ‘that aforementioned’ can be seen, as well as its two descendants, the article ha’ before the nominal ‘my life’, and the complementizer ha’ before the clause ‘my life will go out’. (78) Tuscarora marked complement: ha’ Há:ne:’ wę́hte’ hane:’ w-ęht-e’ that neuter.agt-mean-st [ha’ ękayá:kę’ ha’ ę-ka-yakę-’ the fut-neuter.agt-exit-pfv ‘That means [that my life will go out].’

haʔ akwanę́nhehkt ha’ akw-an-ęnhe-hk.] the 1sg.poss-mid-live-ins (‘I will die.’)

974

Marianne Mithun

5 Conclusion The Iroquoian languages show that functional and formal processes of grammaticalization can be intertwined. Their common ancestor, Proto-Iroquoian, was already quite polysynthetic, with elaborate verb morphology already showing the effects of repeated processes of univerbation and phonological reduction. Comparative evidence across the daughter languages indicates that for the most part, there has been relatively little further phonological reduction, but a number of functional developments can be charted, many strikingly parallel to grammaticalization processes observed in languages elsewhere in the world. Within nominals, plural enclitics have developed on certain terms referring to persons from erstwhile distributives. An augmentative clitic has developed from a verb meaning ‘it is large’ (semantic abstraction), with concomitant phonological reduction. Articles meaning ‘the aforementioned’ have developed from anaphoric demonstratives, not unlike developments documented in Romance languages. Within the verbal morphology, a middle prefix has developed from a reflexive prefix. The strong tendency for middle constructions to become lexicalized, and the boundary between the middle marker and the root to fade, apparently obscured the original reflexive function of the marker, stimulating a renewal of the reflexive via reduplication. Benefactive applicative suffixes have developed from verbs meaning ‘give to’ and ‘lend to’, and instrumental applicative suffixes have developed from verbs meaning ‘use’ and ‘pick up’. The formal grammaticalization processes involved in these developments apparently preceded the functional ones, however; they began with the compounding of full verb roots (univerbation) which retained their original lexical meanings, followed by subsequent functional developments of semantic abstraction. Tense prefixes have developed from directional markers already within the morphology, in different directions in the two main branches of the family. In Southern Iroquoian (Cherokee), the cislocative (hither) has developed into a future, and in Northern Iroquoian, the translocative (thither) has developed into a past. In none of these cases have the functional developments been followed by further formal reduction. The affixes descended from roots via compounding have retained the full forms of their root sources, and those descended from other affixes have retained the same forms as their affix sources. Overall, the Iroquoian developments indicate that certain functional and formal processes of grammaticalization tend to recur cross-linguistically, but their relative sequencing can vary.

Notes Mohawk examples are presented in the conventional orthography. Most symbols have their approximate IPA values. Nasalized vowels are represented as sequences:

Grammaticalization and polysynthesis: Iroquoian

975

[ᴧ̨ ] by and [ų ] by = [ų ]. The glide [j], which occurs only before vowels, is represented by . Glottal stop [ʔ] is represented by an apostrophe . Vowel length is indicated by a colon , and contrastive tone by accents: acute for high tone (on short vowels) or rising tone (on long vowels), and grave for a tone contour that rises then falls sharply.

Abbreviations Abbreviations follow the Leipzig glossing rules. Additional abbreviations include: and = andative, appl = applicative, cisloc = cislocative, contr = contrastive, dv = duplicative, exp = experienced, lk = linker, mot = motion, proth = prothetic, transloc = translocative

References Cook, William. 1979. A grammar of North Carolina Cherokee. New Haven, USA: Yale University dissertation. Diessel, Holger. 1999. Demonstratives: Form, function, and grammaticalization. Amsterdam: John Benjamins. Epstein, Richard. 1993. The definite article: early stages of development. In Jaap van Marle (ed.), Historical linguistics 1991, 111–134. Amsterdam: John Benjamins. Goossens, Louis. 1985. Modality and the modals. In M. Bolkestein, C. De Groot & L. Mackenzie (eds.), Predicates and terms in functional grammar, 203–217. Dordrecht: Foris. Greenberg, Joseph. 1978. How does a language acquire gender markers? In Joseph Greenberg (ed.), Universals of human language, vol. 3: 47–81. Stanford: Stanford University Press. Hawkins, John 2004. Efficiency and complexity in grammars. Oxford: Oxford University Press. Heine, Bernd & Tania Kuteva. 2002. The world lexicon of grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd & Tania Kuteva. 2006. The changing languages of Europe. Oxford: Oxford University Press. King, Duane. 1975. A grammar and dictionary of the Cherokee language. Athens, USA: University of Georgia dissertation. König, Christa. 1993. Aspekt im Maa. (Afrikanistische Monographien, vol. 3). Cologne: University of Cologne. Mithun, Marianne. 2016. The shaping of modality: Views from Iroquoian. In Jan Nuyts & Johan van der Auwera (eds.), Oxford handbook of mood and modality, 223–257. Oxford: Oxford University Press. Montgomery-Anderson, Brad 2008. A reference grammar of Oklahoma Cherokee. Kansas, USA: University of Kansas dissertation. Nuyts, Jan. 2016. Analyses of the modal meanings. In Jan Nuyts & Johan van der Auwera, (eds.), Oxford Handbook of Mood and Modality, 31–49. Oxford: Oxford University Press. Shibatani, Masayoshi & Prashant Pardeshi. 2002. The causative continuum. In Masayoshi Shibatani (ed.), The grammar of causation and interpersonal manipulation, 85–126. Amsterdam: John Benjamins. Tucker, Archibald & J. T. ole Tompo Mpaayei. 1995. On the application of the notion of grammaticalization to West African Pidgin English. In Anna Giacalone Ramat & Paul Hopper (eds.), The limits of grammaticalization, 273–88. Amsterdam: John Benjamins.

976

Marianne Mithun

Woodbury, Hanni. 2003. Onondaga-English/English-Onondaga dictionary. Toronto: University of Toronto Press.

Willem F. H. Adelaar

23 Grammaticalization in the Quechuan and Aymaran languages of the Central Andes 1 Introduction Quechuan and Aymaran are two numerically important language families located in the Central Andean region, which coincides more or less with the former Inka Empire at its greatest expansion (before it was overthrown by European invaders in 1532). Together they comprise most of the indigenous languages and language varieties still spoken today in a discontinuous area situated in the highlands and Pacific foothills of Bolivia, Ecuador and Peru. Quechuan, the most widely distributed language group, also has extensions in the Amazonian lowlands of these three nations and in Northwestern Argentina, where it has replaced other native languages mainly in the tracks of colonization conducted by the Spaniards. Originally, the Central Andean region was characterized by a much higher level of linguistic diversity, which was still locally robust at the time of the Conquest. Nevertheless, with the exception of the still dominant Quechuan and Aymaran languages and three minor languages (Chipaya in the Bolivian highlands, and Cha’palaachi and Tsafiki in the Ecuadorian coastal region), the native languages of the Central Andean coast and highlands have all disappeared. Only a few of these lost languages have been documented to a more or less significant degree (Cholón, Esmeraldeño, Kunza, Lule, Mochica and Puquina), whereas others can only be identified through toponyms, family names and a few recorded lexical items. In the Amazonian lowlands adjacent to the Central Andes, including the eastern slopes of the Andean mountain range, several indigenous languages are still thriving, and some of them show traces of interaction with the highland languages (for instance, Cahuapanan, Jivaroan, and the Arawak languages of the Campa sub-group, as well as Amuesha). From a genealogical point of view, Quechuan and Aymaran do not share a demonstrable common origin, nor have they been shown to be related to any other existing language family in or outside the Americas. Nevertheless, their grammatical structure and reconstructed sound systems are strikingly similar, and they feature more than 20 % of identical shared lexicon at the level of the reconstructed proto-languages (Proto-Quechua and Proto-Aymara). Since the once popular hypothesis of a Quechumaran language stock (Orr and Longacre 1968) is now generally discarded, the observed similarities must be attributed to a process of linguistic convergence that may have taken place as early as two millennia ago. This time depth has to be assumed in order to leave room for the considerable internal differentiation that occurred within each of the two families after the initial convergence was completed (cf. Adelaar 2012). Nowadays, the Aymaran family comprises two, possibly three, living languages (Aymara and Jaqaru/Kawki), whereas Quechuan is https://doi.org/10.1515/9783110712735-009

978

Willem F. H. Adelaar

internally divided into a large number of local varieties that are alternatively referred to as ‘dialects’ or as ‘languages’, depending on the historical perspective or criteria of the observer. Two main dialect branches, Quechua I and Quechua II (following Torero’s [1964] terminology), can be discerned, and both are characterized by a specific set of phonological and grammatical innovations. Even within each of these branches, mutual intelligibility between the local varieties is often lacking due to an ongoing process of dialectal diversification which can be attributed to both internally and externally motivated changes. The latter include changes due the effects of language contact or substrate languages (including, once more, varieties of Aymaran). Conservative Quechuan language varieties are found in the highlands of central and central-northern Peru, which appear to have been the original focus of expansion of the family. All the Quechua I languages (spoken in the departments of Ancash, Huánuco, Junín, Pasco and parts of Lima), as well as Quechua II varieties located in neighboring areas such as Ayacucho, Cajamarca and Yauyos (in the southeast of the department of Lima) have remained close to the Proto-Quechua prototype. In the remaining Quechua II languages, situated in areas that are more distant from central Peru (viz. Argentina, Bolivia, Ecuador, the Amazonian lowlands, and parts of northern and southern Peru), and where the effects of language contact have been substantial, changes have generally been more conspicuous. In some areas, the original Quechuan prototype was largely abandoned, as in the highly creolized varieties, locally called Quichua, that are spoken in the highlands and Amazonian lowlands of Ecuador. By contrast, Quechuan varieties in Bolivia show a tendency towards developing more complex morphological structures than those found in the original Proto-Quechua prototype (cf. Adelaar 1995). The Aymaran languages are both very conservative and have hardly lost anything of their original morphological complexity. Nonetheless, Jaqaru exhibits a considerable influence of neighboring Quechuan varieties and possibly of unidentified extinct local languages as well (Emlen 2017: 337). Its morphological and morphophonemic complexity are extreme. Due to their structural similarity and dominant position in the Central Andes, Quechuan and Aymaran have often been highlighted as the prototypical representatives of an alleged “Andean” language type, which would be strikingly different from the language types found in lowland South America and in Mesoamerica. However, more systematic investigations of whatever remains of the Andean languages that do not belong to the Quechuan and Aymaran families have shown that the overall structure of these languages can be rather different. Even though there may be some general Andean features such as case marking, dependent marking, and a distinct preference for suffixation, the overall picture remains diffuse, and there are many exceptions to these tendencies. The conclusion is that in South America, Quechuan and Aymaran stand alone in their language type, which is likely to be a product of the millenary process of convergence delineated above.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

979

2 Overall structure of Quechuan and Aymaran The following general characterization holds for Aymaran and for the more conservative varieties of Quechuan, which are closest to the Quechuan prototype. Innovative Quechuan languages can be divergent or exceptional in some respects, but this will not be specified here for each case. Both Quechuan and Aymaran feature a complex morphological structure which is agglutinative and predominantly suffixing, including some cases of vowel lengthening (in Aymaran and in Quechua I) and distinctive stress assignment (in the divergent Quechuan variety of Pacaraos). There are no productive prefixes at all. The Aymaran languages furthermore feature a highly unusual system of vowel suppression which occurs before specific affixes. From a synchronic point of view, this vowel suppression does not appear to be phonologically predictable, and its historical background is still mostly unexplained. The preferred constituent order in Quechuan and Aymaran is verb-final. SV/ AOV is the obligatory order in dependent clauses. Both language groups are strictly Nominative − Accusative. Subjects do not receive case affixes, but objects are normally marked for Accusative, except immediately before a nominalized verb in Quechuan, where it is often omitted. This variation is illustrated in the following examples from Northern Junín Quechua: (1)

a. papa-ta tarpu-q-ra:-č̣i aywa-ru-šaq wara potato-acc sow/plant-agent-incompl-conj go-perf-1subj.fut tomorrow ‘I guess I will still have to go and plant the potatoes tomorrow.’ b. papa qura-q-ra aywa-ru-šaq potato weed-agent-incompl go-perf-1subj.fut ‘I will still have to go and weed the potatoes.’

Within hierarchically structured nominal phrases, modifiers (adjectives, possessors) normally precede their heads. In Quechuan, nominalized verbs used as modifiers also precede their heads, except when they are part of a relative clause construction, in which case it is not unusual to find them after their antecedents. The different orders are illustrated in examples from Tarma Quechua, a sub-variety of Northern Junín Quechua: (2)

a. pišta-na wa:ga slaughter-fut.nmlz cow ‘A cow to be slaughtered.’ b. wa:ga pišta-nxa-: cow slaughter-non.agent.nmlz-1subj ‘The cow that I slaughtered.’

980

Willem F. H. Adelaar

Tab. 1: Personal endings in Ayacucho Quechua.

1st person 2nd person 3rd person inclusive person

Subject

Possessor (V__#)

Possessor (C__#)

-ni -nki -n -nčik

-y -(y)ki -n -nčik

-ni-y -ni-ki -ni-n -ni-nčik

The distinction between transitive and intransitive verbs does not play a significant role in Quechuan and Aymaran. Person agreement in the verb does not preferably target the object but a human (SAP) recipient or addressee (3). There is no sign of ergativity, and there is no passive in the classic sense.1 (3)

Pacaraos Quechua (Adelaar 1987) miku-y ka-nqa-n-ta qu-šu-nki eat-inf be.there-non.agent.nmlz-3subj-acc give-inv-2fut ‘He will give you the food there is.’

Nominal and verbal morphology are clearly distinct, although there is a certain amount of formal coincidence between the way the grammatical person of a nominal possessor and a verbal subject are indicated in Quechuan. Compare the nonfuture verbal subject markers of the indicative mood 2 and the nominal possessive markers in Ayacucho Quechua (Table 1). Nominal morphology comprises case marking, number marking (only plural) and possessor marking, which includes the grammatical person of a possessor (‘my’, ‘your’, etc.) or a possessor in general (‘owner of ’, ‘entity characterized by’). There is no gender distinction, nor any other distinction in terms of noun classes. In a possessive phrase possession must be marked both on the head (Person-ofPossessor) and on the modifier (Genitive case), whenever the modifier is expressed lexically. Person agreement is required between a possessor and the possessed. Number-of-Person agreement is not required. Compare the following (constructed) examples from Ayacucho Quechua (4) and Aymara (5): (4)

inti-pa wasi-n sun-gen house-3poss ‘The house of the sun.’

ñuqa-nčik-pa wasi-nčik 1pers-43-gen house-4poss ‘Our (inclusive) house.’

1 A passive derivation with the root tuku- ‘to become’, ‘to change to’ attached to an Infinitive in -y has been attested in colonial writing from Southern Peru, as in apa-y-tuku- ‘to be carried by animal or person’ (González Holguín 1608). For modern varieties it has been reported in Ecuador (cf. Mugica 1967: 64). 2 The indicative subject markers for 2nd person are identical in future and non-future.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

(5)

wil yka-n[a] uta-pa sun-gen house-3poss ‘The house of the sun.’

981

hiwasa-n[a] uta-sa 4pers-gen house-4poss ‘Our (inclusive) house.’

Verbal derivation is highly elaborate and formally regular but semantically not always predictable. Combinations of verbal roots and derivational affixes often acquire idiosyncratic meaning, a device which is heavily exploited in order to make up for the relative poverty of the verbal root lexicon. Verbal derivation includes valence-changing categories (causative, assistive, reflexive, reciprocal, applicative, non-control, etc.), spatial and directional markers, affixes with pragmatic functions, affixes with aspectual meanings in the broadest interpretation (interruption, trial, repetition, state, etc.), and several others. The Aymaran derivational inventory is more elaborate than that of Quechuan due to the presence of a more elaborate set of spatial markers. By contrast, verbal derivational affixes in Quechuan are more often polyvalent, and their meanings and functions can be elusive. Nevertheless, it should be noted that many affixes tend to coincide in meaning and use in both language families. This is most probably a consequence of the convergence indicated above. For a presentation of the numerous similarities see Cerrón-Palomino (2008). Verbal inflection in the restricted sense occupies the rightmost position in the verb (excluding clitics) and centers around Tense, Mood and Person (in Quechua II also Number). The paradigms concerned often contain portmanteau affixes and other irregularities that appear to be at odds with the agglutinative principle. Aspect, Number (in Quechua I and Aymaran), Speaker orientation (Motion towards speaker, Back to place of origin), Beneficiary, Person-of-object and partly also Causative are located in an area between the truly derivational affixes and the verbal inflection. The rules for suffix ordering can be quite strict in this intermediate area and not always in agreement with what one might logically expect, especially in Quechua I. Compare the (constructed) examples from North Junín Quechua in (6). (6) waqa-či-pa:ku-n cry-caus-pl-3subj ‘They make (her/him/them) cry.’

waqa-rka-ya:-či-n cry-pl-progr-caus-3subj ‘They are making (her/him/them) cry.’

Different types of nominalization and verbalization are generally indexed by affixes and play a major role in the grammar of Quechuan and Aymaran. Nominalizers occupy the same location as tense affixes and are not necessarily derivational in character. They are frequently found with case suffixes in semi-grammaticalized combinations underlying the formation of complement clauses that indicate pur-

3 We use ‘4’ as a gloss for ‘Inclusive person’ in accordance with the structure of the four-person system of Quechuan and Aymaran (see 3.1.1. below).

982

Willem F. H. Adelaar

pose, temporal relations, motivation, avoidance, etc. In Quechuan, some of these nominalized constructions maintain the possibility of specifying both person-ofsubject and person-of-object (7), whereas Aymara nominalizations behave more like nouns as they are only compatible with the set of possessive affixes (8). (7)

Pacaraos Quechua tapu-ma:-na-n-rayku qati-pa-yka:-ma-n ask-1obj-fut.nmlz-3subj-case:motive follow-appl-progr-1obj-3subj ‘He is pursuing me for the sake of asking me questions.’

(8)

apa-ni-:ta-w[a] naya-n[a] makina-m-Ø 4 machine-2poss-acc learn-cisl-2subj.fut-decl 1pers-gen uñ.x̣ata-ña-ha-taki examine-nmlz-1poss-ben ‘You will bring your machine so that I can have a look at it.’ (Yapita 1991: 122)

Verbalization is mainly derivational in Quechuan, as the affixes concerned are directly attached to a nominal base and occupy a slot before the affixes that denote verb-to-verb derivation and/or inflection. In the example (9) from Tarma Quechua -ča- represents verbalization, whereas -ka(:)- ‘non-control’ and -či- ‘causative’ are cases of verb-to-verb derivation. The combination ali-ča-ka(:)- is a lexicalized verbal derivation, which can be translated as ‘to recover from illness’. (9) Tarma, North Junín Quechua ali-ča-ka:-či-n good-make-non.control-caus-3subj ‘(S)he/it cures people.’ In Aymaran, verbalization is not exclusively derivational as some of the verbalizing affixes make up for the absence of a verb ‘to be’, and hence are located in an inflectional environment (see below 3.2.2 and 3.3.1). Subordinate verbs in the Quechuan languages are subject to a switch-reference system, which in most varieties can be combined with full person marking (subject and object). It creates the possibility to keep track of a topic in discourse without referring to it explicitly in each new sentence. Tail-head linkage is a common device. In present-day Aymara, switch-reference occupies a more peripheral position. A switch-reference construction with non-identical subjects (suffix -pti-) is exemplified in (10) in an utterance from Tapo (North Junín) Quechua.

4 In Aymara the Accusative case is marked by the suppression of the final vowel of the nominal base, here indicated with the Ø symbol.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

983

(10) mana čiwaku pla:nu-nči-ta aspi-pti-n-qa tapu-š limaq not blackbird map-4poss-acc erase-subor:ds-3subj-top Tapo-rep Lima ka-n-man ka-ra be-3subj-pot be-past.3subj ‘It is said that if the blackbird had not erased our map, Tapo could have been the capital (Lima).’ Both Quechuan and Aymaran have a class of affixes that operate at a higher level than the word to which they are attached, although they are completely integrated from a phonological point of view. These clitic-like elements include evidential markers (in Quechuan), interrogation and negation markers, topic markers, completive (‘already’) and incompletive (‘still’), additive markers (‘also’, ‘even’), as well as other similar categories. True clitics which are not fully integrated within the wordform to which they are attached also occur. Verbal composition (combinations of verb roots or nominal incorporation) does not occur in Quechuan and Aymaran. Nominal composition is weekly developed. Juxtaposition of nouns with maintenance of the phonological properties of free forms is common, especially in Quechuan, for instance, in cases such as rumi čaka ‘stone bridge’, ‘a bridge made of stone’ or wasi punku ‘house door’, ‘the entrance of a house’.

3 Patterns of grammaticalization Quechuan and Aymaran languages are notoriously resistant to the identification of grammaticalization patterns underlying their highly elaborate morphology. Especially in their verbal morphology formal similarity between bound morphemes (affixes), on the one hand, and roots or free forms, on the other, seems to be lacking in most cases, even where they represent closely related meanings. In Quechuan, for instance, there is a formal resemblance, if not identity, between specific paradigms denoting grammatical person in nouns (person of possessor) and in verbs (person of subject), but none at all with the corresponding personal pronouns that operate as free forms and refer to the same personal categories. The forms at issue are shown in Table 2 for Ayacucho Quechua. The divergent affixes for Future tense and Imperative are not included. This does not mean, however, that no cases of grammaticalization can be identified at all. These are mainly found in the nominal morphology, rather than in the very opaque verbal morphology, suggesting that the former is younger than the latter. From a phonological perspective, the phonotactic shape of affixes in both Quechuan and Aymaran features options that are never found in word-initial position. Affixes often begin with a consonant cluster, whereas free forms do not nor-

984

Willem F. H. Adelaar

Tab. 2: Free forms of singular pronouns and the corresponding bound person markers in Ayacucho Quechua.

1st person 2nd person 3rd person

free pronoun

possessive affix

subject affix

ñuqa qam pay

-y, -ni-y -yki, -ni-ki -n, -ni-n

-ni, -y -nki, -yki -n, -Ø

mally do so; e.g., Quechuan -nki ‘2nd person subject Present and Future’, -pti- ‘Different subjects subordinator’; Aymara -mpi, Jaqaru -wšqa ‘Instrumental case’, etc.5

3.1 Grammaticalization of nominal categories As observed before, categories of class and gender and the associated agreement phenomena are completely absent from the Quechuan and Aymaran languages, although (non-native) gender indications occasionally occur in loanwords that reflect Spanish gender. In what follows, we will focus on possession (3.1.1), number (3.1.2), determiners (3.1.3), and case (3.1.4).

3.1.1 Possession Substantives, some pronouns (mainly quantifiers), and most nominalized verbs can take affixes that indicate the grammatical person of a possessor. The notion of possession should be understood in the broadest possible way as any form of hierarchical relatedness between two nominal entities. There is no grammatical distinction between alienable and inalienable possession. The Aymaran languages and part of the Quechuan languages (mainly Quechua I) exhibit a system of four persons in which number does not play a role: a. Person of speaker (1); b. Person of addressee (2); c. Person of neither speaker nor addressee (3); d. Person of speaker and addressee combined (4). In accordance with the obvious symmetry of this system, the Inclusive category is often referred to as the ‘4th person’, a term which has been used for other purposes in the context of North American languages. In this chapter we shall use the term ‘inclusive person’ instead, in order to avoid misunderstanding, except in the glosses where we use ‘4’.

5 The main exception is constituted by some finite forms of the verb sa- ‘to say’ in Aymara, e.g., s-ta ‘you say’. However, in such cases an epenthetic element hi- ~ si-, is often added in order to avoid the anomalous situation of a consonant cluster in initial position, hence hi-s-ta or si-s-ta ‘you say’.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

985

Tab. 3: Structure of the personal reference system in Aymaran and in Quechua I (cf. Hardman, Vásquez, and Yapita 1988: 18).

1st person 2nd person 3rd person Inclusive person (4)

Speaker

Addressee

+ – – +

– + – +

Tab. 4: Aymara free pronouns and examples of the possessive affixes. 1st person 2nd person 3rd person Inclusive person (4)

naya ‘I’ huma ‘you’ hupa ‘she/he’ hiwasa ‘we’, ‘you and me’

uta-ha, uta-x ̣a ‘my house’ uta-ma ‘your house’ uta-pa ‘her/his house’ uta-sa ‘our house’

The Aymaran languages, in contrast with the Quechuan languages, exhibit a somewhat regular coincidence between their nominal possessive affixes and the corresponding free pronouns, as illustrated in Table 4. Since three of the four pronouns (1st person is the exception) appear to contain a shared pronominal root element *hu- / *hiw(a)-, it may be assumed that the latter acquired the possessive affixes, rather than that these affixes should be viewed as abbreviated versions of the corresponding pronouns.6 Meanwhile, the relationship between free pronouns and possessive affixes, on the one hand, and verbal affixes referring to person-of-subject or person-of-object, on the other, is diffuse. For instance, a 2nd person marker -m(a) has a limited distribution in verbal paradigms such as the Imperative, and in some non-finite categories, but in most indicative tenses it is mainly found as part of complex personal endings involving a 2nd person object (cf. Yapita [1991]; čura- ‘to give’): (11) čura-m čur-tam čur-sma

‘Give it (to him/her).’ ‘He gives it to you.’ ‘I give it to you.’

3.1.2 Number Nominal number in Quechuan and Aymaran is limited to an optional indication of plurality on substantives and other nouns, which may be human, non-human or 6 Personal pronouns consisting of a specialized root combined with personal reference affixes are found in several other Amerindian languages, e.g., in Yucatec and other Mayan languages.

986

Willem F. H. Adelaar

inanimate without any distinction. Plural marking often has an individualizing function as it refers to a set of countable individuals or entities, rather than to a coherent group or category. As in several other South American languages, the nominal plural marker in Quechuan and Aymaran consists of a disyllabic form: -naka in Aymara and -kuna in Quechuan.7 The form -naka is shared with the unrelated Chipaya language, probably as a result of borrowing, whereas -kuna is again found in the (Aymaran) Jaqaru language (cf. Hardman 1966), presumably as a loan affix from Quechuan. Due to their disyllabic shape, these markers are reminiscent of free forms, although there are no counterparts or related forms in the lexicon from which they might have been derived in a transparent manner. Number marking is also found in association with possessive affixes in order to pluralize a possessor, but only in Quechua II, one of the two main branches of the Quechuan family (according to the classification established in Torero [1964]). In most Quechua II varieties, a suffix -čik or any of its local variants is used to pluralize 2nd person markers; it may have developed from a free form čika referring to ‘size’ or ‘amount’.8 By contrast, 1st and 3rd person markers are pluralized by means of a suffix -ku, which may hark back to an abbreviated version of the nominal pluralizer -kuna (see above). See the following (constructed) examples from Ayacucho Quechua with wasi ‘house’: (12) wasi-y wasi-ki wasi-n

‘my house’ ‘your house’ ‘her/his house’

wasi-y-ku wasi-ki-čik wasi-n-ku

‘our (exclusive) house’ ‘your (plural) house’ ‘their house’

In principle, these forms can be followed by the pluralizing suffix -kuna with the proviso that a sequence *-ku-ku- is generally avoided. (13) wasi-ki-čik-kuna house-2.poss-2.pl-pl ‘Your (plural) houses’9 Aymaran and the Quechua I varieties (as well as some conservative Quechua II varieties) have no dedicated affixes for the pluralization of possessive person markers. Therefore, plurality of possessor can only be made explicit periphrastically. There can hardly be any doubt that possessive number marking in Quechua II is

7 See, for instance, Puquina -gata, Atacameño -ckota, Guaraní -kwera, and several others. 8 It should be noted, however, that the element -čik can also be discerned in -nčik, the most widely used marker for the Inclusive person. Since the latter can be reconstructed for Proto-Quechua, its use must be older than that of -čik as a pluralizing element. 9 A double -kk- sequence is generally pronounced as a single consonant.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

987

the result of innovations. Originally, neither Quechuan (14), nor Aymaran (15) had such a possibility. The Quechuan example is from the northern Junín variety. (14) nuqa-:-kuna-pa wayi-: 1pers-1poss-pl-gen house-1poss ‘Our (exclusive) house’ (15) na-naka-n[a] uta-ha 1pers-pl-gen house-1poss ‘Our (exclusive) house’ In Quechua II varieties spoken in northern Peru, the plurality of affixes indicating person-of-possessor are generally indicated in a different way. For example, speakers of Cajamarca Quechua (Quesada 1976) use alternative suffixes that are derived from the Quechuan roots *l yapa- ‘all’ and *sapa ‘each’, ‘unique’, ‘sole’, as illustrated in (16). The superficial similarity between these two forms is suggestive of a shared origin, although there is no regular correspondence that connects them synchronically. It should be noted that the same suffixes that are used as possessive pluralizers with nouns are also found with parallel functions in the verbal system (see 3.2.1 below for examples from Ayacucho Quechua). (16) wasi-y-ʤapa house-1poss-pl ‘Our (exclusive) house’

3.1.3 Determiners Quechuan and Aymaran determiners do not normally participate in processes of grammaticalization. In accordance with a general rule pertaining to these languages (cf. § 2), determiners must precede the head they modify, whereas the resistance of the Quechuan and Aymaran languages against any kind of prefixing prevents the determiners from developing into prefixes. Consequently, instances of grammaticalization are restricted to the domain of suffixes. An interesting and rather exceptional example of a suffixed determiner is found in the subgroup of Quechua I known as Huanca (or Wanka). The Huanca varieties are spoken in the Mantaro river basin in the southeastern part of the department of Junín in Central Peru. The Huanca suffixed determiner indicates definiteness. It is therefore semantically similar to the definite article in western European languages, which might be interpreted as an indication of contact influence from Spanish. However, its positional characteristics

988

Willem F. H. Adelaar

are incompatible with those of the Spanish definite article, and the same holds for the evolutionary history of the affix in question, as we will see below.10 The Huanca marker of definiteness is historically derived from *ka-q, the Active participle or Agentive nominalization (ending -q) of the verb ‘to be’ (ka- ‘to be’). In Quechuan languages in general, ka-q, literally ‘the one that is’, ‘(the) one being’, is frequently used as an individualizer or a substantivizing element following other nouns, such as adjectives, numerals, or nominalized verbs, its right-hand position being dictated by the verb-final character of the language. This is illustrated in the example (17) from Pacaraos Quechua: (17) mana-ran lapa-n-su saki r ̃akta ka-q-kuna-q[a] not-incompl.ass all-3poss-irr dry thick be-agent-pl-top ‘The thicker ones are not all dry yet.’ Notwithstanding its incipient grammaticalization, the element ka-q is not a bound form. It maintains the prosodic characteristics of a phonological word, because it is not integrated within the (penultimate) stress pattern of the preceding word form. In the Huanca varieties, however, the element ka-q developed into a fully grammaticalized suffix *-kaq, which can be attached to nominal bases. Contrary to the older ka-q form of other Quechuan varieties, this suffix became part of the resulting word form. This can be concluded from the fact that regular stress is assigned after the affixation of *-ka-q to a nominal base. The semantic change that accompanied this process is that of a marker of individuation or countability to one of definiteness. The following examples are from the Jauja dialect of Huanca, where *-kaq is realized as -kah (Wroughton 1996: 34, 98). (18) wayapa-kah atuh-kah wayapa-def fox-def ‘the man’ ‘the fox’ In the Huanca variety of the area of Huancayo unique phonological changes affected the shape of the element *-kaq. The post-velar stop */q/ changed into a phonemic glottal stop /ʔ/ (except for the word-initial position where it coincided with zero). Subsequently, in a geographically more restricted area the sequence /Vʔ/ developed into a long vowel /V:/. The latter situation obtains in the Chongos Bajo variety to the southwest of Huancayo, which was furthermore affected by a non-regular optional loss of the sequence -ka- in word-internal position after bases ending in a short vowel. As a result, the Definite suffix in Chongos Bajo Quechua appears in three different shapes determined by their position in the word. Just vowel length -:-

10 The Huanca definite suffix is more reminiscent of affixed determiners in Basque, Rumanian or Swedish. Contact with these languages can be excluded.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

989

(or -ka:-) obtains in word-internal position after bases ending in a short vowel; the full form -ka:- is required word-internally after any other base; and -ka is found word-finally (author’s fieldwork 1967; see also Cerrón-Palomino 1989: 44–45). This morphophonemic variation is illustrated below by means of the following (constructed) examples: (19) a. wamla-ka girl-def ‘the girl’ b. walaṣ̌-ka boy-def ‘the boy’ c.

wamla-kuna-ka girl-pl-def ‘the girls’

wamla-(ka):-wan girl-def-com ‘with the girl’ walaṣ̌-ka:-wan boy-def-com ‘with the boy’ wamla-kuna-(ka):-wan girl-pl-def-com ‘with the girls’

This complex allomorphic variation of the marker of definiteness in Chongos Bajo is illustrative of the phonological transformation undergone by the element *ka-q and the extent to which it became grammaticalized. Evidently, considering the conservative Jauja form, these changes were posterior to the semantic specialization of ka-q into an indicator of definiteness, which is characteristic of the Huanca dialect area in its totality.

3.1.4 Case Both the Quechuan and the Aymaran languages mark case relations on the noun by means of specialized affixes. Most of these affixes cannot be derived from any existing lexical root. In nominal phrases with an internal hierarchical structure case is normally indicated only once, namely, on the final constituent of the construction (usually the head of the noun phrase). In Quechuan some case markers present a more or less stable form throughout the family, but others may vary according to region, which may be an indication of a relatively recent origin.11 Combinations of case markers with preservation of their original functions occur, in particular, those with Genitive and Comitative-Instrumental case, as illustrated in the following examples from Pacaraos Quechua:

11 It may be observed that in the Andean and sub-Andean regions case markers are often subject to borrowing, even among unrelated languages. Such borrowing can be accompanied by a switch in meaning. For instance, the Quechuan Instrumental marker -wan is found with the function of Accusative in (Barbacoan) Guambiano (Vásquez 1988) in the Cauca region, southern Colombia. Although this particular case may be coincidental, it happens frequently enough to consider it a trend.

990

Willem F. H. Adelaar

(20) a. mana-s[u] yač̣a-nki uyša-pa-kta-q[a]-a: not-irr know-2subj sheep-gen-acc-top-emph ‘Do not you know that [song] of the sheep?’ b. aska-m=aw ñuqa-paq-wan-č̣[a]-a: much-ass=tag 1pers-ben-com-conj-emph ‘It is a lot, is not it? There must be for me as well (not just for you).’ Historically, the Quechua II Ablative case marker -manta appears to be the product of a grammaticalized combination of the Allative suffix -man and the Accusative suffix -(k)ta.12 Its Quechua I counterpart -pita may originally contain the Quechua II Locative marker -pi followed by the same Accusative suffix (although -piq and -piqta are also found with the Ablative function). The Quechuan Benefactive case marker -paq may be historically related to the Genitive case marker -p(a).13 As a matter of fact, some Quechua II varieties use these two case markers interchangeably in order to denote the Genitive. It is tempting to identify the final element -q in -paq (and the one in -piq(ta)) with the Agentive nominalizing suffix -q, but this cannot be confirmed on the basis of the existing evidence. The shape of the Locative case marker in Quechuan is of particular interest. Instead of -pi, which is found in most Quechua II varieties, a different Locative marker -č̣aw (or any of its local reflexes) is found in Quechua I. This suffix is obviously related to č̣awpi (čawpi in most modern varieties), the general Quechuan noun for ‘middle’ or ‘center’, which may preserve the same element -pi that indicates the Locative in Quechua II. As it appears, -č̣aw derives its origin from a hypothetical lexical root *č̣aw, which may have had the same meaning as č̣awpi in present-day Quechuan.14 The Aymaran set of dedicated case markers is less elaborate and comprises Ablative (-tha ~ -ta), Accusative (vowel suppression or -ha), Allative-Dative (-ru), Benefactive (-taki), Comitative-Instrumental (Aymara -mpi, Jaqaru -wšqa), GenitiveLocative (-na). None of them has a transparent grammaticalization history. In addition to its set of dedicated case affixes, Quechuan has a number of specialized affixes, derived from nominal or verbal roots, which can occasionally be used as case markers. The suffix -kama marks Limitative case (‘till’, ‘until’, ‘as long as’) and, in some Quechuan varieties, also Interactive case (‘among’), along with

12 Accusative -kta is found after (short) vowels in 16th century Quechua II and in at least two modern varieties (Pacaraos and Huanca); otherwise only -ta occurs. 13 Genitive -p is found after (short) vowels in 16th century Quechua II and in modern Huanca varieties; reflexes of -p subsist in Cuzco Quechua and Bolivian Quechua; otherwise -pa occurs. 14 The existence of such a hypothetical root is furthermore supported by its presence in the Quechua II word punčaw (< *punč̣aw; Cuzco Quechua p’unčaw, p’unčay) ‘day’, which was originally also used for ‘sun’ (viz., ‘center of the day’). The element pun ‘day’ is still found in Huanca (CerrónPalomino 1976a: 106) and in a compound expression hukpun ‘the other day’ in northern Junín Quechua (Adelaar 1977: 501).

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

991

another non-case related function such as Distributive plural (‘each of them’, ‘each time’). It was probably derived from the verb root kama- ‘to create’, ‘to inspire’, ‘to fit’. As a noun, kama refers to a task or administrative burden. The case marking role of -kama can be established on the basis of its occurrence in the same affix slot as other spatial case markers, as illustrated in (21) and (22) from Pacaraos Quechua. (21) r ̃ate:ru-m kunan-pa č̣a:-ri-mu-n kay-kama thief-ass now-add arrive-pl-cisl-3subj this(place)-limit ‘Thieves now come as far as here too.’ (22) kay-man-pa šamu-n=a: ka:r ̃u-wan this(place)-all-add come-3subj=emph car-ins ‘They come to this place as well, by car.’ Another Quechuan case marker derived from a verb or noun is -rayku ‘because of, for the sake of ’. Although -rayku is widespread in Quechuan, a lexical counterpart has only been attested in the Quechua I variety of northern Junín, where raygu refers to ‘reason’ or ‘motive’, and the derived verb raygu-ku- is translated as ‘to be a guarantor’, ‘to be responsible of ’. Both -kama and -layku (< *rayku; Jaqaru -nayku15) are also found as borrowed case markers in Aymaran. As an illustration of -rayku, we recapitulate the earlier example (7) from Pacaraos Quechua in (23): (23) tapu-ma:-na-n-rayku qati-pa-yka:-ma-n ask-1obj-fut.nmlz-3subj-case:motive follow-appl-progr-1obj-3subj ‘He is pursuing me for the sake of asking me questions.’ Some other specialized case markers, including occasional loans from Spanish, are best analyzed as phonological free forms, rather than as affixes. They show that in Quechuan case markers constitute a semi-open category that can be expanded through an ongoing process of grammaticalization when necessary.

3.2 Grammaticalization of verbal categories As indicated before, the verbal morphology of Quechuan and Aymaran lacks transparency for as far as the identification of affixes or word-parts with semantically related lexical forms is concerned. There is no nominal incorporation, nor verb root composition. In both language groups a finite verb may consist of (a) a verbal base (which does not normally occur on its own), (b) one or more non-final affixes that

15 Cf. Belleza Castro (1995).

992

Willem F. H. Adelaar

Tab. 5: Overview of Quechuan verbal structure. (a)

(b1 )

(b2 )

(c)

verbal root

lexicalizing non-final affixes

nonlexicalizing non-final affixes (person-ofobject) (number)

tense/mood/ subordination (person-of-subject) (number)

nominal verbaliroot zing affix

nominalizing affix (person-ofsubject) (number)

classindependent affixes

clitics

case

must be followed by other affixes, (c) one or more affixes, including portmanteau markers, that refer to Person, Number, Tense and Mood (also Evidentiality in Aymara). Note that Aspect marking (b) is separate from Tense and Mood (c). Additionally, optional affixes (or clitics) with a function at the sentence or clause level may be added. These are not properly verbal as they occur with noun phrases, subordinate clauses and other constituents as well. It follows that a verb root is minimally accompanied by an affix of category (c). Nominalization and subordination are indicated in the same area as the affixes of category (c), in most cases with retention of the possibility to express grammatical person (but not Tense and Mood); see example (23) in the previous section. Verbalization is found at the root level in Quechuan, but plays a more central role in Aymaran where even complex nominalized constructions can (re-)verbalize. The verbal affixes pertaining to category (b) constitute a highly heterogeneous set with very different functions and meanings, which furthermore exhibit a lot of geographical variation across both families. In the literature they are alternatively referred to as derivational affixes, modal affixes, verbal extensions, or post-base morphology. Their functions include valence change (causative, reflexive, reciprocal, applicative), spatial specifications, modality, pragmatic considerations, and much more. Most of these affixes are indeed derivational in nature as they tend to form lexicalized bases derived from specific verb roots with often unpredictable meanings (b1 ), but others convey meanings and functions that are not subject to lexical specialization, such as Number, Aspect, Speaker orientation, Beneficiary, Person-of-object, etc. (b2 ). Affixes of the latter type cluster on the right side of the area covered by (b), whereas lexicalizing affixes remain closer to the root and tend to cluster on the left side of it. The following template provides a representation of a Quechuan verbal structure as outlined above.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

993

Few of the verbal affixes, including the nominalizing and subordinating affixes, show significant affinities with free forms in the lexicon. In this context, we will briefly discuss the category of Number-of-person (3.2.1), possible manifestations of the verb ‘to be’ in Quechuan and Aymaran verbs (3.2.2), and developments concerning a set of verbal directional affixes in Quechuan in general (3.2.3).

3.2.1 Number-of-person in Quechua II Verbal number is expressed in very different ways, a fact which suggests that there may not have been any verbal number marking either in Proto-Quechua, or in ProtoAymara. The Quechua I languages and Aymara make use of non-final affixes located in the intermediate area of the affix chain (b2 in Table 5). These affixes are closely intertwined with the Aspect markers, which are also non-final and separate by function and distribution from the Tense and Mood markers. Verbal plurality usually implies that the subject or agent of the verb is plural, but it can also apply to a direct object or recipient whenever the context permits it, as in example (24) from Tarma (North Junín) Quechua: (24) čay-la-ta-m ni-ba:ku-x lapa-y-ta that-only-acc-ass say-pl-1subj.2obj all-2poss-acc ‘I say only that to all of you.’ The Aymara verbal Plural marker -p-x̣(a)- is illustrated in (25). Note that the Completive suffix -x̣(a)- is obligatory in affirmative statements in La Paz Aymara (Yapita 1991: 108). Some historical sources suggest that this Plural marker may have had its origin in the incorporation of the numeral base pisqa ‘five’ (Cerrón-Palomino 2000: 104), which would make it an obvious case of grammaticalization. (25) laru-si-p-x̣-t-wa laugh-refl-pl-compl-1subj-decl ‘We laugh at him.’ In Quechua II, verbal plurality is indicated by the same devices as those that indicate plurality of a possessor on nouns (see 3.1.2 above for their derivational background). Again, plural marking can refer to the subject or agent of the verb, but also to an object or a recipient. However, the fact that in most of Quechua II there are two pluralization markers, one for 1st and 3rd person (-ku) and one for 2nd person (-čik ) (cf. § 3.1.2), makes it possible to specify which of the two SAP participants involved is to be interpreted as plural. The system with two markers distinguished for person manifests its usefulness in combination with the fact that in transitive constructions both the subject and a SAP patient are identified for person

994

Willem F. H. Adelaar

on the verb.16 See, for instance, the following (constructed) examples from Ayacucho Quechua (26): (26) riku-su-nki see-inv-2subj ‘(S)he sees you.’

riku-su-nki-ku see-inv-2subj-1/3pl ‘They see you.’

riku-su-nki-čik see-inv-2subj-2pl ‘(S)he sees you (plural).’

The addition of -ku in riku-su-nki-ku pluralizes the 3rd person actor, whereas the addition of -čik in riku-su-nki-čik pluralizes a 2nd person, which is the patient in this case. In most Quechuan varieties, it is not possible to explicitly pluralize both the subject and the object. The number status of the unspecified participant remains unexpressed and can only be derived from the context or by the addition of a lexical pronoun.

3.2.2 Derivational affixes and the verb ‘to be’ Most of the non-final verbal affixes in Quechuan and Aymaran fail to show convincing links with any lexical component in the language. A possible exception in Quechuan are suffixes which contain the syllable -ka(:)- (or any reflex of it), which may be identified with the verb ‘to be’, ‘to exist’.17 This element -ka(:)- is found in suffixes that indicate Progressive aspect, such as Ayacucho Quechua -čka-, a form which also underlies reflexes such as -ška-, -sya-, -ša- and -sa- in other varieties of Southern Peru and Bolivia, and -yka(:)- (often reduced to -ya(:)-), which is found with the same function in Quechua I and several varieties of the subgroups Quechua IIA and IIB (e.g., Cajamarca and San Martín Quechua). (27) Ayacucho ri-čka-n go-progr-3subj ‘He is going.’

Pacaraos ri-yka-n [ˈri:kaŋ] go-progr-3subj ‘He is going.’

It is tempting to analyze -yka(:)- as a combination of the verb ‘to be’ with the Infinitive nominalization, which is indexed by a suffix -y-, even though such a combination does not reflect any productive synchronic procedure in the language. For -čkaan explanation is more difficult to find because there is no known nominalization

16 Such constructions are called ‘transiciones’ in the Spanish colonial grammar tradition, a term which is still in use today. 17 In Quechua I and Cajamarca Quechua a verbal suffix -ka(:)- is found, which indicates ‘Noncontrol’ (cf. example (9). It may in some cases be interpreted as a pseudo-passive ‘to get … ed’. A relation with the verb ‘to be’ is possible but not proven.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

995

or verb base that ends in -č or anything of that kind. Note that a semantically related suffix indicating an anticipated progressive action has been attested as -č̣ka- in Cajamarca Quechua (Quesada 1976: 133–134) and as -čka(:)- in Huanca Quechua (Cerrón-Palomino 1976b: 212–213). It suggests that-yka(:)- and -č̣ka- (~-čka(:)-) may have coexisted originally as two different but semantically related categories, as is still the case in these two geographically separated varieties. The Aymaran languages use morphological devices in order to express the notion ‘to be’. In Aymara it consists in lengthening the final vowel of a base;18 Jaqaru has a suffix -w- for this purpose. In addition, Aymara also features another verbalizing element -k(a)- which is exclusively used to indicate existence or location in combination with the Genitive / Locative case marker -n(a) in expressions such as huma-n-k(a)- ‘to be yours’, ‘to belong to you’ and aka-n-k(a)-‘to be here’, ‘to belong to this place’. Cerrón-Palomino (2008: 200) advanced the hypothesis that the Aymara existential and copula verbalizers both hark back to the verb ka- ‘to be’, which would have been shared by the two language groups Aymaran and Quechuan. Both verbalizations co-occur separated by the Agentive nominalizer -iri in example (28), illustrating the high degree of polysynthetic flexibility proper to the Aymaran languages: (28) aka-n-k-iri-:-t-wa this(place)-loc-loc.vrbz-agent-cop.vrbz-1subj-decl ‘I am from here.’, or: ‘I belong here.’

3.2.3 Development of directional affixes In the Aymaran languages there exists a set of four derivational affixes referring to the directional notions ‘in’, ‘out’, ‘up’ and ‘down’, in Southern Aymara, respectively, -nta-, -su-, -ta- and -qa-. As in the case of most derivational affixes in Aymaran, these elements cannot easily be derived from any root or lexical item found in the language. A similar set of affixes has been reconstructed for the Quechua I group (Parker 1973: 22–23). It consists of *-yku- ‘in’, *-rqu- ‘out’, -rku- ‘up’ and -rku- ‘down’. Only the last two affixes are productively used with the abovementioned meanings in present-day Quechua I. The first two are widely used both in Quechua I and Quechua II but have been affected by processes of semantic and functional shift in which their directional meanings were gradually transformed or bleached. Whereas pragmatic applications dominate in the case of the original Inward suffix -yku(~ -yu-), the Outward suffix *-rqu- (~-ru-) has developed meanings such as Recent

18 In the Aymara variety of Muylaque (Moquegua, Peru), vowel length was lost and the morpheme ‘to be’ has no other reflex than the preservation (or rather non-suppression) of a preceding vowel (Coler 2014: 472–477).

996

Willem F. H. Adelaar

past (Ancash Quechua), Punctual or Perfective aspect (North Junín and Huanca Quechua, Pacaraos Quechua), and a connotation of Urgency (Southern Peruvian Quechua II).19 The probable sequence of functional and semantic transformations was: Outward movement > Urgency/emergency > Punctual/Resultative > Perfective aspect > Recent past. Nevertheless, the four suffixes with their original meanings can be recognized in verb sets that are still actively used in Quechua I, such as the one exemplified in (29): (29) yaykuyarquyarkuyarpu-

‘to enter’ ‘to go out’, ‘to leave’ ‘to go upwards’, ‘to climb’ ‘to go downwards’, ‘to descend’

Only the verb yayku- is in general use throughout present-day Quechuan, with some local variation. It is hayku- in Cuzco Quechua. It appears attractive to speculate about a lexical origin for the directional affixes. This seems most promising in the case of the Inward direction suffix *-yku-, which may be associated with the Quechua II word uku ‘interior’, ‘inside’, ‘deep’ (Cuzco Quechua ukhu ~ uxu). Note that this notion is consistently expressed as ruri (with variations) in Quechua I. Nevertheless, in Laraos (Yauyos) Quechua (Taylor 1990) and in colonial sources a form yawku- (instead of yayku-) has occasionally been found for ‘to enter’ (cf. also Durston and Urioste 2013: 408–410). It may give support to the origin suggested above for the suffix -yku-. Drawing the speculation somewhat further, one may look for a similar lexical association in the case of the Outward direction suffix *-rqu-. A possible connection may be sought in the word urqu ‘mountain’, ‘male’, which would represent the open outside world as opposed to the hiddenness of uku. Lexical forms that resemble the Upward and Downward direction suffixes -rku- and -rpu- are, respectively, urku ‘forehead’ and urpu ‘jar’, ‘jug’, ‘pitcher’. Although the former connection seems suggestive, the latter is less likely to represent a significant link. Incidental similarities can be found between other verbal derivational affixes and specific lexical forms, such as -ri- Inchoative (and several other meanings) with the Quechua II verb ri- ‘to go’, and -tya(:)- Interrupted action with tiya- (QII)/ *taya(QI) ‘to sit’. If there should be any semantic relationship at all, it cannot be demonstrated with certainty. For most verbal derivational affixes in both Aymaran and Quechuan a lexical identification of their origin remains illusory.

19 For a presentation of a chain of formal and semantic transformations undergone by the four directional affixes in Northern Junín Quechua see Adelaar (2006).

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

997

3.3 Grammaticalization and reanalysis of complex constructions In Quechuan, grammaticalization of complex constructions is frequently found in combinations of nominalized verbs (participles) with a copula or existential verb (ka- ‘to be’ in both readings). Similar constructions occur in the Aymaran languages with the difference that a morphological device (affix or vowel lengthening) is used for the expression of the notion ‘to be’. The modern Aymaran languages lack a lexical verb ‘to be’, both in the existential and in the copula reading. Other grammaticalized verbal constructions in both language groups can include verbs of motion ‘to go’, ‘to come’, the verb ‘to say’, and a set of verbs that can function as higher verbs complemented by an Infinitive in the Accusative case. Compound tenses in Quechuan are used for the expression of grammatical tenses other than the basic Present, Past and Future. They are comparable to the compound tenses of Western European languages. In spite of their compositional nature, in which the two components may appear as separate phonological words, compound verbs also display behavior pertaining to synthetic verb forms, which makes it clear that they cannot be considered as the mere sum of two free lexical elements. In a further stage of development, a complete formal merger may occur, in which the root of the auxiliary verb ‘to be’ (ka-) is erased and its affixes directly attached to the nominalized verb form. Both stages of development are found in varieties of the Quechuan language family.

3.3.1 Past Habitual in Quechuan and Aymaran One of the most widespread compound tenses is the Past Habitual, which in Quechuan consists of an Agentive nominalization (or Active participle) in -q accompanied by the Present tense form of the verb ‘to be’ in its copula interpretation, as in Ayacucho Quechua riku-q ka-ni ‘I used to see’. Since the 3rd person form of the copula, ka-n, is obligatorily deleted when it carries no further affixes, the 3rd person subject form of the Past Habitual is simply riku-q ‘(s)he used to see’. The 3rd person plural marker -ku, which normally requires the presence of a 3rd person subject affix, can be attached directly to the nominalized verb form as in riku-q-ku ‘they used to see’. In other words, the plural marker -ku reacts to the Agentive form as if it contained a (silent) 3rd person subject marker. When a Past Habitual form consists of two separate components, complex suffix sequences referring to combinations of personal categories can either cluster on the higher (auxiliary) verb or be distributed over the two components, thus showing that the latter jointly operate as a single word form (although phonologically they are not). For example, in Southern Peruvian Quechua (Ayacucho, Cuzco) the 3rd person subject / 2nd person object combination for Present and Future tense (see also section 3.2.1 above) is concurrently indexed by the affixes -su- and -nki (when

998

Willem F. H. Adelaar

used alone, -nki refers to a 2nd person subject). Note that the suffix -su- does not have a meaning of its own other than signaling inversion. In Ayacucho Quechua (Parker 1969: 49) the two affixes are distributed over the two component verb forms (30). (30) riku-su-nki see-inv-2subj ‘He sees you.’

riku-su-q ka-nki20 see-inv-agent be-2subj ‘He used to see you.’

By contrast, in Cuzco Quechua Cusihuamán (1976: 174) reports the following example (31) with the same affixes clustering on the higher verb. (31) yača-či-q ka-su-nki learn-caus-agent be-inv-2subj ‘He used to teach you.’ The Habitual construction in Aymara is based on the Agentive nominalization in -iri followed by the vowel length that conveys the meaning of the copula ‘to be’ (cf. § 3.2.1). (32) aruma-ki-w hut.x̣-iri-:-ta night-only-decl come.from-agent-cop.vrbz-1subj ‘I used to come home only at night.’ (Yapita 1991: 89) With a 3rd person subject, as in Quechuan, the Agentive nominalization suffices, and there is no re-verbalization by means of the copula vowel length. (33) č’iqa tuqi-n-k-iri-wa left side-loc-loc.vrbz-agent-decl ‘He is habitually on the left side.’ (Yapita 1991: 89)

3.3.2 Other compound tenses with ka- ‘to be’ in Quechuan More compound tenses in the Quechuan languages are formed on the basis of the Non-agentive (Past participle) nominalization in -šqa (~ -sqa). Interestingly, the combination with ka- ‘to be’ can function both as a transitive and an intransitive

20 The form *riku-su-q by itself has no meaning.

Grammaticalization in the Quechuan and Aymaran languages of the Central Andes

999

construction. Apparently, in compound tenses the verb ka- may be subject to semantic bleaching, and its resulting function is merely that of a semantically unspecified auxiliary. In Pacaraos Quechua, the combination of the -šqa nominalization with ka- ‘to be’ yields a Perfect tense, which can also be used to denote a Past experience, as in (34). (34) qunqa-šqa-m ka-y forget-past.part-ass be-1subj ‘I have forgotten it.’ In Ayacucho Quechua the -sqa nominalization is used without the element ka- and the endings for 1st, 2nd and Inclusive person subject are attached directly to the nominalized form. The resulting paradigm is a Narrative past with a mirative overtone (35). (35) čay sipas-ta muča-sqa-ni that girl-acc kiss-narrative-1subj ‘I kissed that girl [I mistook her for another].’ (Parker 1969: 49) Part of the Quechua I varieties feature a tense that indicates ‘sudden discovery’ or ‘revelation’, usually in the past (cf. Adelaar 2013). It is marked with an element -ñaq, accompanied by SAP forms of the verb ‘to be’, in varieties such as Huanca and Pacaraos Quechua (36). Historically, this element appears to be a combination of a Completive independent affix *-ña- (which otherwise is not normally incorporated in the verb form) and the Agentive nominalization ending *-q. In the 3rd person subject form the verb ‘to be’ is omitted. (36) a. say mana-š ar ̃os-ta mika-ri-ñaq-su that not-rep rice-acc eat-pl-mir.3subj-irr ‘It turned out, reportedly, that they did not eat rice.’ b. rikapa-rqa-mu-ñaq ka-y look-perf-2obj-mir be-1subj ‘I had been looking at you (without realizing it).’ In Northern Junín Quechua the SAP personal endings are attached directly to the element -na- ( personal pronoun and article The development of pronouns and articles from demonstratives is a well-known process (Heine and Kuteva 2002: 109–111, 112–113). Once the old demonstratives got bleached and ended up as pronouns, presumably in pre-Mountain Ok times, the languages were in need of a new set of demonstratives, which were formed by reinforcing the old demonstrative with a deictic element; again a well-known process typologically. I will now sketch the development of the contemporary Mountain Ok demonstratives.

2.2.2 Demonstratives While the Mountain Ok languages show a relatively homogenous picture in their personal pronoun forms there is more variation in the demonstratives. Historically, the current demonstratives were presumably formed by reinforcing the old demonstratives with an additional deictic element. Synchronically, the forms are not analyzable in this fashion anymore. The analysis of Mountain Ok demonstratives is hampered by the fact that we have reliable data only for three languages: Tifal, Telefol and Mian. For Bimin, only a series of definite articles has been identified (Weber 1997: 47), which prima facie look like forms parallel to the Telefol distal demonstratives (see below). For Faiwol, Mecklenburg and Mecklenburg (1977) do not mention a separate series of demonstratives.7 The forms of the proximal and distal demonstrative roots in Tifal, Telefol, and Mian are set out in Tables 4 and 5. Both sets can be used pronominally and adnominally, e.g., Telefol beéyó [dem.dist.sg.m] ‘that’ and tanum beéyó [man dem.dist.sg.m] ‘that man’. Note that the Telefol forms do not make a number distinction, a point I will take up briefly below. The Telefol and Tifal forms are bound roots which have to take a suffix; the Mian forms are free roots, i.e., they can occur independently or take a suffix. In the following I will sketch the presumed diachronic development of the demonstratives. The forms given will be approximations; particularly vowel length and quality will be finessed. The proximal demonstratives in the three languages share a cognate part. Mian loses initial /k/ conditioned by the quality of the following vowel, where Telefol and Tifal keep it (Loughnane and Fedden 2011: 16; Timothy Usher, p.c.) and there is a sound correspondence Mian /ɛ/ : Tifal /a/ : Telefol /a/ and Mian /o/ : Tifal /u/ : Telefol /u/. This would make *kV ‘here’ a possible etymon

7 There is a form bela ‘this’ in Mecklenburg (1974), which appears once in the phrase bela kinim dukum [this man big] ‘this bigman’ (Mecklenburg 1974: 164). This is too little information to say much about this form.

Grammaticalization in Mountain Ok (Papua New Guinea)

1015

Tab. 4: Proximal demonstrative roots across Mountain Ok (‘this’). Person

Number

Gender

Tifal

Mian

Telefol

3

sg

m

kalaa-

élé

keé-

f

kuluu-

óló

koó-

kalii-

élí

pl

m

keé-

f

koó-

Tab. 5: Distal demonstrative roots across Mountain Ok (‘that’). Person

Number

Gender

Tifal

Mian

Telefol

3

sg

m

kaa-

yé

beé-

f

kuu-

yó

boó-

kii-

yéi

pl

m f

beéboó-

of the proximal element. Note that the Tifal and Mian forms involve an additional element -lV. Likely reflexes of *kV are: ka in the Tifal word kawu ‘here’, Mian é‘here’ and Telefol ka ‘here’. At some point the deictic element would have been collocated with a gendered pronominal element *e, *u or *i, following the pattern [here+3sg.m] to yield a new series of proximal demonstratives, i.e., *kV(-lV)-e, *kV(-lV)-u, *kV(-lV)-i. The feminine forms are characterized by ablaut, e.g., Tifal kuluu- and Mian óló. As Telefol monophthongizes all vowel sequences (Healey 1964b: 2), whatever the exact vowel sequence was the result in Telefol would have been a long like vowel. The masculine singular and plural forms would have collapsed due to phonologically regular convergence and then the singular feminine was extended to the plural by analogy (Timothy Usher, p.c.). Now we turn to the distal forms, which are more heterogenous. The first thing to note is that the Tifal distal forms are actually cognate with the Telefol proximal forms. For Mian the distal forms are historically collocations of a distal deictic element *yV and one of the gendered pronominal elements *e, *u or *i, following the pattern [there+3sg.m] to yield a new series of distal demonstratives, i.e., *yV-e, *yV-o, and *yV-i. The relation of the distal deictic element *yV to the present-day Mian word yé ‘there’ is obvious. It is very likely that the Mian forms are old. The evidence for this is that the personal pronouns *ye ‘he’, *yu ‘she’, and *yi ‘they’ can be reconstructed for Lowland Ok (Healey 1964a: 67; Usher 2014). These seem to be cognate with the Mian distal demonstratives, but have developed into personal pronouns in LOk, replacing the old Ok pronouns *e, *u, and *i. The Telefol forms probably in-

1016

Sebastian Fedden

volve a different element ba- (< *mbV). For Bimin, Weber (1997) does not mention a separate series of demonstratives but analyzes the language as having a set of definite articles consisting of /b-/ followed by the third person pronoun (Weber 1997: 47), thus be ‘def.sg.m’, bu ‘def.sg.f’, bi ‘def.pl’, e.g., kunum be [man def.sg.m] ‘the man’; waneng atite bu [woman beautiful def.sg.f] ‘the beautiful woman’; kunum bi [man def.pl] ‘the men’. We can also find these forms in Tifal: bee /be:/ ‘that person (male), that thing’, buu /bu:/ ‘that person (female), that thing’ (Healey and Steinkraus 1972: 140). It is likely that these are the same /b-/ elements that appear in the Telefol distal demonstratives.

2.3 Gender All Mountain Ok languages have grammatical gender. Gender systems are not typical of Trans New Guinea languages (Wurm 1982: 80; Wälchli 2018). Gender in Mountain Ok is an old phenomenon. The gendered pronouns in the second and third person singular go back to proto Ok times: *kep ‘you (m)’ vs. *kup ‘you (f)’; *e ‘he’ vs. *u ‘she’ (Usher 2014). For more details on pronominal forms, see section 3.5 on agreement. The rest of this section briefly describes the gender systems of the present-day Mountain Ok languages. Agreement targets are pronouns, including demonstratives, articles and verbs. Telefol, Tifal, Faiwol and Bimin are analyzed as having two genders, masculine and feminine. In Telefol, humans and higher animals are assigned gender according to biological sex, males are masculine and females feminine, whereas lower animals and inanimates are assigned gender according to size (small > masculine, large > feminine), with the possibility of flexible gender for nouns that commonly have large or small referents, as in (5). We have no information about gender assignment in Tifal, Faiwol or Bimin. (5)

Telefol a. tuúm íyó stone 3sg.m ‘the (small) stone, pebble’ b. tuúm úyó stone 3sg.f ‘the (large) stone, rock’

Mian has a more complex gender system in which assignment for humans and higher animals is the same as in the other Mountain Ok languages. Other animals have a conventionalized gender, either masculine or feminine, but without size being a predictor of assignment, and inanimates are assigned one of two inanimate genders. Table 6 gives the forms of the third person pronoun.

Grammaticalization in Mountain Ok (Papua New Guinea)

1017

Tab. 6: Forms of the Mian third person pronouns. Singular Masculine

é

Feminine

ó

Neuter 1

é

Neuter 2

ó

Plural í

ó

Mian has four controller genders, i.e., genders defined by the agreements required by the controller. The forms show an interesting pattern of syncretism: all Mian genders are nonautonomous values (Zaliznjak [1973] 2002: 69–74), which means that they have no agreement forms which are unique to them. Consequently, the number of gender distinctions that can be found on targets, i.e., the number of target genders, is smaller, namely two: =e and =o in the singular, and =i and =o in the plural. An alternative way of looking at the Mian gender system would be to say that there is a masculine-feminine distinction in the singular and an animateinanimate distinction in the plural. For a detailed description of the Mian gender system, see Fedden (2007), Fedden (2011: ch. 4), Corbett and Fedden (2016), Corbett, Fedden, and Finkel (2017). There is not enough information about the agreement patterns of plural inanimates in the other Mountain Ok languages to exclude the possibility that the gender system found in Mian, in which animates pattern differently from inanimates, is actually more widespread in Mountain Ok.

2.4 Spatial and temporal markers All Mountain Ok languages have grammaticalized the noun for ‘interior, hole’, Mian tem, into a spatial marker meaning in and they grammaticalized the noun for ‘top (surface)’, Mian dim, into a spatial marker meaning on. For Mian tem, there is a homophonous nominal form with the meaning ‘interior, hole’, cf. amǔn-tem [bellyhole] ‘abdominal cavity’ rather than *‘in(to) the belly’. Mian dim is attested in the compound amgolim [am-go-dim house-head-top] ‘roof ’, rather than *‘on(to) the roof ’. The spatial marker on has also been extended to temporal contexts in the following interrogative collocations: Mian fâtnamin dim, Tifal kanimin diim, both [what on] ‘when’, and in temporal adverbials, e.g., Mian fraide dim [Friday on] ‘on Friday’. A further grammaticalization of the noun for ‘top (surface)’, e.g., Mian dim, into a marker of predicative possession is discussed in section 3.4 below. These are the only two simple postpositions for which there is a clear noun origin. At least Mian has a set of some 14 complex postpositions which consist of a noun (often a body-part noun) and a locative suffix -daa, which itself is potentially

1018

Sebastian Fedden

a grammaticalization from the verb daa ‘abide’, e.g., ase ablan-daa [tree.the underside-at] ‘at the underside of the tree’. For more examples, see Fedden (2011: 240– 242). (6) Mountain Ok interior, hole > in top (surface) > on (spatial) > on (temporal) The development of relational nouns (including nouns referring to body parts) into spatial or temporal grammatical markers is a common grammaticalization channel (Heine and Kuteva 2002: 182, 299–300).

3 Verbal categories In this section I discuss the grammaticalization of the categories aspect (3.1) and tense (3.2). Then I turn to the use of the verb ‘give’ as a valence-increasing device for the introduction of an additional argument with dative-like semantics (3.3), which is also used in predicative possession (3.4). Finally, I present the Mountain Ok agreement system (3.5). Since modal categories are expressed by affixes whose etymologies are obscure I will not say more about them here.

3.1 Aspect Aspect is a central grammatical category in all Mountain Ok languages. The languages allow us to observe several different stages of a grammaticalization cycle. At the outermost layer there are constructions in which the verbs that contribute the aspectual meaning are still a phonologically independent word. Mian uses the verb be, which means ‘move’, as a lexical verb (7), with a persistive meaning ‘keep V-ing’ (8). (7)

Mian inǎb=e yé be-b-e=be snake=sg.m there move-ipfv-3sg.m.sbj=decl ‘A snake is moving there.’

(8)

Mian yé genam be-b-i=be there roll.ipfv persistive-ipfv-1sg.sbj=decl ‘I keep rolling (a cigarette).’

Grammaticalization in Mountain Ok (Papua New Guinea)

1019

Other verbs which are used with an aspectual function in a serial verb construction are deibô ‘leave’ in Mian with a completive meaning, and une ‘go’ in Tifal with a habitual meaning (Boush 1979: 2). Many TNG languages (and Papuan languages more widely) express aspect distinctions by means of serial verb constructions with verbs like ‘hold’, ‘take’ or ‘hit’ for perfective aspect and ‘stay’, ‘stand’, ‘lie’, or ‘do’ for continuous aspect (Foley 1986: 145). Consider example (9) from the Madang language Kalam (Pawley 1993; 2008), where the verb md- ‘stay’ expresses a persisting state of affairs.8 (9)

Kalam b yob ag md-p-ay man big sound stay-prs-3pl ‘The big men are still talking/continue to talk.’

The high incidence of serial verb constructions in the expression of aspect is related to the high frequency of this construction type in Papuan languages in general (Foley 1986: 143). Other TNG languages in which serial verb constructions play a role in aspect marking are Enga (Lang 1975), Fore (Scott 1978) and Dani (Bromley 1981). Outside of TNG this is the case for example in Iatmul (Sepik; Staalsen 1972) and Vanimo (Skou; Ross 1980). Returning to Mountain Ok, at the next layer further inward we find a continuous suffix -bi which grammaticalized from the existential verb ‘exist, stay, remain’, whose forms are: Mian bi, Telefol bi, Tifal bV, Faiwol bi, and Bimin b and which can be reconstructed as *bi for proto Mountain Ok. This development has happened in all Mountain Ok languages except Bimin. Compare examples (10) and (11). (10) Mian é yé bi-s-e=be 3sg.m there exist-rpst-3sg.m.sbj=decl ‘He was there.’ (11) Mian é yé fu-bi-s-e=be 3sg.m there cook-cont-rpst-3sg.m.sbj=decl ‘He was cooking there.’ The oldest layer is a highly irregular set of markers of a perfective-imperfective contrast. A large number of verbs take part in this system of aspect marking, e.g., 71 %

8 In southern New Guinea such serial verb constructions are rare in the expression of aspect. Instead a combination of alternate verb stems and specialized aspectual inflections are used (Drabbe 1955; Siegel 2014; Evans 2015; Döhler 2016).

1020

Sebastian Fedden

Tab. 7: Stem aspect in Mian. Process

Perfective

Imperfective

Gloss

Suffixation

fa

fa-ka

‘make fire’

ge un dei-lâ’ nge-la halbu-a uli-lò me-le he-na

ge-n un-ê dei nge-n halbû’ uli me-n he-n

‘build’ ‘go’ ‘remove hair’ ‘beg’ ‘fold’ ‘roll thread’ ‘touch’ ‘seek’

Apophony

ifa ge

ifu ga

‘serve food’ ‘say’

(Near-) Suppletion

dowôn’ baa

wen o

‘eat’ ‘say’

Tab. 8: Stem aspect in Telefol. Process

Perfective

Imperfective

Gloss

Suffixation

daá

daá-ka

‘put’

un

ún-e

‘go’

bóko

báka

‘say’

ífo

ífú

‘serve food’

undú

wee

‘prepare’

Apophony

Suppletion

of verbs in Mian (Fedden 2015). While the continuous suffix -bi (cf. 12 above) combines with a wide range of verbs the suffixes currently under discussion differ depending on the verb. None of the subtypes contain more than a dozen verbs, some subtypes only have one or two verbs. For some verbs the aspectual distinction is signalled by apophony or (near-)suppletion. Examples from Mian are given in Table 7. Suprasegmental changes in the form of tone changes do occur but they are never the sole exponent of aspect. Every language also has a subset of biaspectual verbs for which one and the same form can have a perfective or an imperfective meaning. In Mian, a large number of verbs, i.e., 132 out of 456 verbs (or 29 %), are biaspectual (Fedden 2015). Examples from Telefol of the use of suffixation, apophony and suppletion in aspect marking are given in Table 8 (Healey 1964a: 68). Given the layers of grammaticalization in the Mountain Ok aspect system it is plausible to assume that the innermost aspect markers are etymologically verbs

Grammaticalization in Mountain Ok (Papua New Guinea)

1021

which have grammaticalized into suffixes. These verbs would originally have been used in a serial verb construction consisting of a lexical verb followed by a functional verb expressing aspect. This sequence would then have undergone desemanticization and coalescence, resulting in a single verb with an aspect suffix. Erstwhile verbal etyma are hardly recognizable anymore. For at least one Mian suffix a potential verbal origin is still apparent: -lo marking perfective aspect, which presumably comes from the extant verb -lô ‘hit’.9 There is a tendency for certain aspect categories to be expressed either periphrastically or inflectionally (Dahl 1985). For instance, a periphrastic construction is typically used for the progressive, whereas perfectives and imperfectives are typically expressed with a bound form. This form-meaning correlation is related to the degree of grammaticalization. A gradual generalization of meaning is paralleled by a gradual reduction in form and fusion with the verb (Bybee and Dahl 1989: 56). Progressives, completives and habituals as less general meanings show less grammaticalization of form, whereas perfective and imperfective are more abstract meanings and hence show a stronger degree of grammaticalization of form. Furthermore, the perfective-imperfective distinction is typically expressed by more complex means than are found in other areas of morphology, including other areas of the tenseaspect system of a language (Dahl 2000: 16). The degree of lexical idiosyncrasy is high and it is often not predictable from one verb to another how the opposition is realized. This is exactly the situation we find in Mountain Ok, as far as the perfective-imperfective contrast is concerned. (12) Mountain Ok exist > continuous (except Bimin) verb (in a serial verb construction) > perfective, imperfective (13) Mian move > persisitve leave > completive hit > perfective (14) Tifal go > habitual Grammaticalization processes where verbs acquire an aspectual meaning are by and large well-known and are mentioned in Heine and Kuteva (2002) in many places. More specifically, for the development of ‘exist’ > cont, see Heine and Kuteva (2002:

9 This is a transitive verb which indexes its object with a prefix (see section 3.5 below). I assume that undergoing coalescence with the lexical verb it lost this prefix. The loss of the prefix is amply attested in Mian verb-verb compounds in which the second verb on its own would index its object with a prefix.

1022

Sebastian Fedden

127), for ‘leave’ > compl, see Heine and Kuteva (2002: 189–190), and for ‘go’ > hab, see Heine and Kuteva (2002: 159). The grammaticalization of ‘hit’ onto markers of perfectivity is not explicitly stated in Heine and Kuteva (2002), but is well-known from Papuan languages (Foley 1986: 145). The only lesser-known channel is the development of a very general ‘move’-verb (other than ‘go’) into a marker of persistence.

3.2 Tense The Mountain Ok languages have a number of suffixally marked tense categories, all with obscure etymologies. Mian shows interesting, more recent developments in its tense system which cannot be found in the other languages. There are two additional tense categories which are both marked with a suffix in a slot after the subject suffix. These are -bio ‘General past’ and -so ‘Hesternal past’. An example of each is given in (15) and (16), respectively. (15) Mian fu-n-eb-bio=be cook-real-2sg.sbj-gpst=decl ‘you cooked’ (16) Mian (sintalo) fu-n-eb-so=be (yesterday) cook-real-2sg.sbj-hpst=decl ‘you cooked yesterday’ The construction from which the general past suffix -bio grammaticalized, probably consisted of a fully inflected lexical verb in a serialization with the existential verb bi, whose function was to signal a time interval between the event and the moment of speaking. The existential verb would have been inflected with a 3sg.f subject suffix -o, which is here used to indicate an impersonal subject. This is illustrated in the hypothetical (hence starred) Mian example in (17). (17) Mian *fu-n-eb bi-o cook-real-2sg.sbj exist-3sg.f.sbj ‘you cooked and then there was staying’ Over time, this construction would have undergone coalescence and the morphologically complex verb bi-o would have been reanalyzed as a tense suffix -bio, locating the event in the past.

Grammaticalization in Mountain Ok (Papua New Guinea)

1023

The second tense category which Mian innovated essentially in the same way is the hesternal past marker -so (see [16] above). Here the underlying construction seems to be a serialization with the verb s ‘sleep’ (Fedden 2011: 453) with a similar impersonal inflection, illustrated in the hypothetical (starred) example (18): (18) Mian *fu-n-eb s-o cook-real-2sg.sbj sleep-3sg.f.sbj ‘you cooked and then there was sleeping’ Again, this construction would have undergone coalescence and the morphologically complex verb s-o would have been reanalyzed as a tense suffix -so for an event that took place yesterday (or the day before yesterday). (19) Mian exist + impersonal inflection > general past sleep + impersonal inflection > hesternal past The verb s ‘sleep’ is a plausible etymon for a hesternal past. While this is not a common grammaticalization channel, and it goes unmentioned in Heine and Kuteva (2002), it has been found in at least two Bantu languages (Fyam and Kgalagadi); see Dahl (2013: 43–44) and references there. The development of ‘yesterday’ into a past marker, however, is mentioned in Heine and Kuteva (2002: 315–316), but it occurs infrequently and is restricted to Africa.10

3.3 Verb ‘give’ as a valence-increasing device The use of the verb ‘give’ as a valence-increasing device is widespread in TNG languages (Foley 1986, 2000; Reesink 2013). Mian, Telefol and Tifal use their ‘give’ verbs -û b’-, -ub- and -ub-, respectively, to add an argument onto which a range of peripheral semantic roles can be mapped, such as recipient, benefactive, malefactive, and goal of ballistic motion. Bimin and Faiwol diverge from the other Mountain Ok languages in that a valence-increasing element derived from ‘give’ cannot be discerned synchronically.

10 Given that the Mian word for ‘yesterday’ is sintalo I cannot exclude that the hesternal past is actually a development from this adverb involving some heavy phonetic erosion (sintalo > -so), rather than a development from the verb s ‘sleep’. The reason why I consider the latter scenario more likely is that it would make the developments of the general past -bio and the hesternal past -so parallel. Both suffixes would go back to verbs, namely bi and s, respectively. There is no plausible adverbial etymon for the general past suffix -bio.

1024

Sebastian Fedden

Example (20) illustrate the use of ‘give’ in Mian as a lexical verb. Examples (21) and (22) shows its use as a valence increasing device, adding a benefactive argument to an intransitive and a transitive verb, respectively. I will call this the benconstruction. The grammaticalization of ‘give’ into dative or benefactive meanings is well-known and common cross-linguistically (Newman 1996; Heine and Kuteva 2002: 149‒151, 153‒154). See Fedden (2010) for a detailed description of this construction in Mian. (20) Mian né naka=e ěil=o om-û b’-a-n-i=be 1sg man=sg.m pig=sg.f 3sg.f_cl.obj-give.pfv-3sg.m.r.pfv-real-1sg.sbj=decl ‘I gave the sow to the man.’ (21) Mian un-û b’-ke-n-amab-i=be go.pfv-ben-2sg.r.pfv-aux.pfv-irr.nanpl.sbj-1sg.sbj=decl ‘I will go for you.’ (22) Mian imen=e fu-b’-ke-Ø-i-bio=be taro=sg.n1 cook-ben-2sg.r.pfv-real-1sg.sbj-gpst=decl ‘I cooked a taro for you.’ Like Mian, Telefol and Tifal use this construction to introduce a further argument. This is illustrated for the Telefol verb -kó ‘hit’ in (23), and the Tifal verb boko ‘say, talk’ in (24). (23) Telefol (Healey 1965b: 12, my segmentation and glosses) beéyó mán í-tá dem.dist.sg.m boy 3sg.m-foc a-ng kó-b-nee-l-á 3sg.m.obj-hit.pfv-ben-1sg.r-pfv-3sg.m.sbj.imm.pst ‘He hit the boy for me.’ (24) Tifal (Boush 1979: 12, my segmentation and glosses) boko-b-na-laam-sad-a-lee wookwookmin uyo kee-mokom-i talk-ben-1sg.r-imm-ds-3sg.m.sbj-conn work it do-irr-1sg.sbj ‘Immediately after he talks to me I will do his work.’ In Mian, Telefol and Tifal, the ben-construction is restricted to perfective aspect. In the imperfective the person suffix is directly appended to the verb stem (see Fedden [2010: 464, 468], and examples in Boush [1979: 17]).

Grammaticalization in Mountain Ok (Papua New Guinea)

1025

There are clear signs of grammaticalization in this construction. First, there is incipient phonetic erosion in ben, which never applies to lexical ‘give’. The former is reduced to just -b’- after a vowel, e.g., Mian fu-b’-ke [cook-ben-2sg.r] ‘cook for you’. Further, there is considerable desemanticization and expansion from lexical ‘give’ to the semantically much more general meaning of ben, which is not restricted to recipients. (25) Mountain Ok (except Faiwol and Bimin) give > benefactive While not well-known from European languages this grammaticalization channel is attested in Africa, South East Asia, Oceania, and also in some creole languages (Heine and Kuteva 2002: 149–151).

3.4 Predicative possession Lacking a verb ‘have’, Mian uses the existential verb bi ‘exist, stay, remain’ as a marker of predicative possession. There are two distinct constructions. One is the ben-construction (see section 3.3 above), in which the possessor is expressed as the (benefactive) object and the possessed is expressed as the subject (26). The other is an adverbial construction involving dimo ‘on’ (< dim ‘top (surface)’ (see section 2.4 above) for the possessor; it encodes the possessed as the subject (27). (26) Mian bi-Ø-e=be né fǔt=e bi-t’-ne11 1sg tobacco=sg.n1 exist-ben-1sg.r exist-ipfv-3sg.n1.sbj=decl ‘I have tobacco.’ (lit. ‘The tobacco is for me’) (27) Mian kéb dim=o fǔt=e bi=a? 2sg.m.poss on=art tobacco=sg.n1 exist.ipfv.3sg.n1.sbj=pq ‘Do you have tobacco?’ (Lit. ‘Is the tobacco on you?’) The latter example nicely illustrates that in the process of grammaticalization decategorialization can happen after desemanticization and extension. This means that a word can be used in a new context with a new meaning but it can still show the morphosyntactic structures associated with its original part of speech. In (27), dim still carries the clitic article =o and the preceding pronoun is in its possessive form

11 /b/ > [t] in (26) is a regular phonological process that applies to ‘give’ as well as to ben.

1026

Sebastian Fedden

(kē b ‘your’). Although noun phrase structure is retained, the meaning is clearly not ‘(top) surface’ anymore, but rather dim has assumed a relational meaning here. Whether either of these predicative possession constructions is used in any of the other Mountain Ok languages remains unclear. Given that at least Mian, Telefol and Tifal have the ben-construction, I think it likely that parallel constructions to the one in (26) exist in these languages, but they are not in evidence in the sources. (28) Mian exist-for > possessive top (surface) > on > possessive According to Heine and Kuteva (2002: 105–106) an important source for possessives are datives. This type of grammaticalization is described as involving the Goal Schema in Heine (1997). Although the benefactive construction, which is the source of the possessive in Mian, is not a dative, it fits here because its meanings are crosslinguistically often found in datives, namely recipient, benefactive and goal of ballistic motion. The development of locatives into possessives, i.e., the second channel that Mian has for the grammaticalization of possessives, is typologically common (Heine and Kuteva 2002: 205).

3.5 Agreement The bound agreements on the Mountain Ok verb are most likely pronominal in origin and were incorporated into the verb in pre-Mountain Ok times. I will highlight plausible etymologies given the present-day forms, but I am aware that the data do not demonstrate the shift from free pronouns to bound (agreement) forms.

3.5.1 Pronouns and person affixes in Mian All finite verbs in Mian agree with their subject by means of a suffix. Table 9 sets out the forms of the personal pronoun roots and the subject suffixes. In contrast to the pronoun roots, which are all high-toned, subject suffixes – having undergone phonetic erosion – are toneless. In the suffixes, the gender contrast is neutralized in the second person singular, so is the clusivity distinction in the first person plural. The etymological relations are rather obvious for the second person singular, the third person singular, and the second person plural. It seems that the second person plural form was extended to mean third person plural as well, creating syncretism for all verbs. Cross-linguistically, this pattern of syncretism is common (Baerman, Brown, and Corbett 2005: 59), and it is particlarly frequent in TNG. There is no obvious relation between the subject suffixes and the pronoun roots of the first person forms.

Grammaticalization in Mountain Ok (Papua New Guinea)

1027

Tab. 9: Mian pronoun roots and subject suffixes. Person

Number

Gender

Clusivity

1 2

Pronoun root

Subject suffix

né

-i

m

kéb-

f

ób-

m

é

-e

f

ó

-o

-eb sg 3

1

excl

ní

incl

níb-

-ob pl 2

íb-

3

í

-ib

Tab. 10: Mian pronoun roots and object prefixes. Person

Number

Gender

Clusivity

1 2

Pronoun root

Object prefix

né

na-

m

kéb-

f

ób-

m

é

a-

f

ó

wa-

kasg 3

1

excl

ní

incl

níb-

pl

i- ~ y(a)-

2

íb-

3

í

Mian has a small subset of transitive verbs which agree with the object by means of a prefix. Again, for most forms the relation to the free pronouns is rather clear. Table 10 compares the forms of the pronoun roots with the object prefixes. Like subject suffixes, object prefixes are toneless and have no gender contrast in the second person singular. The relations are clear for the first and second persons in the singular; a plausible source for the plural object prefix i- ~ y(a)- is the third person plural pronoun.

1028

Sebastian Fedden

3.5.2 Pronouns across Mountain Ok We find similar patterns of relationship between personal pronouns and agreement affixes in the other Mountain Ok languages. Table 11 sets out pronoun roots for the other Mountain Ok languages. There is only some variation in the vowels of the pronoun roots and the second person singular feminine in Mian apparently lost the initial /k/. None of the other Mountain Ok languages have a clusivity distinction, unlike Mian which shares this property with the Lowland Ok languages (Healey 1964a: 67; Fedden 2011: 5). The inclusive pronoun níb- was probably formed by compounding ní ‘we (excl)’ and íb ‘you (pl)’ to yield a first person plural inclusive pronoun.

Tab. 11: Mountain Ok pronoun roots (based on Healey 1964a). Pers

Num

Gen

Clus

Mian

Telefol

Tifal

Faiwol

Bimin

né

ní- ~ ná-

na-

na-

ne

m

kéb-

káb-

kab-

kab-

kab-

f

ób-

kúb-

kub-

kub-

ku

m

é

í- ~ yá-

a-

a-

e

f

ó

ú- ~ ó-

u-

u-

u

nú- ~ nó-

nu-

nu-

nu

1 2 sg 3

1

excl

ní

incl

níb-

pl 2

íb-

íb-

ib-

ib-

yua

3

í

í-

i-

i-

i

a I consider this to be the Bimin reflex of proto Mountain Ok *ib- ‘2pl’, rather than a Bimin innovation.

3.5.3 Person affixes across Mountain Ok Tables 12 and 13 show the subject suffixes and object prefixes across Mountain Ok, which are all cognate and likely related to the personal pronouns in the same way as described above for Mian.12

12 All Mountain Ok languages have a third series of suffixes which index the recipient of ‘give’ or the additional argument in the ben-construction. For reasons of space they cannot be discussed here in detail. I refer the reader to Healey (1964a: 74). For a detailed discussion of this set of suffixes in Mian, see Fedden (2010).

1029

Grammaticalization in Mountain Ok (Papua New Guinea)

Tab. 12: Mountain Ok subject suffixes (based on Healey [1964a]; except Tifal 1pl from Boush [1975] because of an empty paradigm cell in Healey [1964a]). Pers

Num

Gen

Mian

Telefol

Tifal

Faiwol

Bimin

-i

-i

-i

-i

-i

-eb

-ab

-ab

-ab

-eb

m

-e

-a

-a

-a

-e

f

-o

-u

-u

-u

-u

-ob

-ub

-ub

-ub

-ub

-ib

-ib

-ib

-ib

-ib

1 2 sg 3

1 2

pl

3

Tab. 13: Mountain Ok object prefixes (based on Healey [1964a]; except Tifal 2sg and pl forms from Boush [1975] because of empty paradigm cells in Healey [1964a]). Pers

Num

Gen

Mian

Telefol

Tifal

Faiwol

Bimin

na-

na-

ni-

na-

ne-

ka-

ka-

ti-

ka-

ka-

m

a-

a-

a-

a-

a-

f

wa-

wa-

u-

wa-

wa-

i- ~ y(a)-

ya-

i-

ya-

ye-

1 2 sg 3

1 2

pl

3

While subject suffixes are straightforward in their distribution – every finite verb has one – object prefixes show a more complex distribution. There are two complications: first, there are two different sets of prefixes, and second, there is sporadic agreement (Corbett 2006: 17). I will discuss these in turn. The Mountain Ok languages have verbs which use a different set of prefixes for object agreement. In the descriptions of Telefol, Tifal, Faiwol and Bimin this is treated as a difference in inflection class. A verb that agrees with its object either belongs to inflection classes (IC) 1 (Table 13 above) or 2 (Table 14 below).13 For the sake of completeness the Mian forms are included in Table 13; more on the situation in Mian below.

13 For Telefol see Healey (1965b: 12), for Faiwol see Mecklenburg and Mecklenburg (1977: 27), for Tifal see Boush (1975: 13–14), and for Bimin see Weber (1997: 28–29).

1030

Sebastian Fedden

Tab. 14: Mountain Ok object prefixes – IC2 (based on Healey 1964a). Pers

Num

Gen

Mian

Telefol

Tifal

Faiwol

Bimin

nem-

na-

nam-

namn-

nam-

kem-

ka-

tan-

kamn-

kam-

m

dob-

du-

da(b)-

dub-

dab-

f

om-

ku-

um-

kub-

um-

dol-

i-

yim-

imn-

yam-

1 2 sg 3

1 2

pl

3

Conspicuously many of the forms end in /m/, which might be an old O-indexing verb, which together with its prefix was at some point reanalyzed as a separate prefix series. In Mian this set of prefixes forms a separate system rather than a different inflection class (Corbett, Fedden, and Finkel 2017). The Mian system works on an absolutive basis in terms of alignment. It includes further semantic distinctions in the third person, such as ‘long object’ or ‘bundle’, which is a Mian innovation. While the other Mountain Ok languages indeed seem to lack these distinctions, at least in Telefol there is evidence that this second set of prefixes also works on an absolutive basis, cf. ku-men [3sg.f.sbj-fall] ‘it falls’ (Healey 1966: 7), which would translate into om-měin ‘it falls’ in Mian, where the prefix indexes the intransitive subject (S) and not the object (O). This makes an analysis of the prefix set in Table 14 as a distinct IC less likely, but before we can be sure, further research into the alignment of argument marking in the other Mountain Ok languages is required. The first and second person forms of this second set of prefixes are also obviously related to the personal pronouns, so are the plural forms for all Mountain Ok languages except Mian. In all of these cases there are additional nasal segments involved, whose origin remains unclear. For the remaining forms a pronominal origin is less plausible. The second complication is that all Mountain Ok languages have sporadic agreement. Only a subset of transitive verbs agree with their object, e.g., in Mian -nâ’ ‘hit, kill’ agrees with its object (IC 1), whereas bou ‘swat’ does not. With the descriptive material available on the other Mountain Ok languages besides Mian it is not possible to identify exhaustive lists of verbs with object prefixes so that we do not know to which degree the agreeing and non-agreeing subsets differ. The only agreeing verb that we can be certain is shared between all Mountain Ok languages is ‘see’, cf. Mian -têm’, Telefol -tám, Tifal -ta, Faiwol -tam, and Bimin -tem, which can be reconstructed as proto Mountain Ok *-tVm (Healey 1964a: 151).

Grammaticalization in Mountain Ok (Papua New Guinea)

1031

To sum up this section on agreement, at least some of the present-day agreement affixes could plausibly be phonetically reduced (toneless) personal pronouns incorporated into the verb. (29) Mountain Ok free personal pronoun > agreement There are cases where this relationship is less clear, e.g., proto Mountain Ok *nV ‘1sg’, but *-i for subject in all Mountain Ok languages. In these cases the non-etymological suffix can typically be reconstructed for proto Mountain Ok.

4 The Mountain Ok reference-tracking system A reference-tracking system (often called switch-reference system in the literature) is a discourse tracking device which monitors subjects by indicating through verbal morphology whether the subject of some other clause is the same (SS) or different (DS); see Haiman and Munro (1983); Roberts (1997); Stirling (1993); and more recently Van Gijn and Hammond (2016). In Papuan languages, reference-tracking morphology typically occurs in clause chaining constructions (Stirling 1993: 16) and is calculated with respect to the subject of the succeeding clause (cf. Foley and Van Valin 1984). A pair of examples from the Madang language Usan illustrates this. (30) Usan (Reesink 1983: 217 and p. c.) ye nam su-ab is-omei 1sg tree cut-ss go.down-1sg.fpst ‘I cut the tree and (I) went down.’ (31) Usan (Reesink 1983: 218 and p. c.) ye nam su-ine is-orei 1sg tree cut-1sg.ds go.down-3sg.fpst ‘I cut the tree (and it went) down.’ Both examples consist of a medial clause with a medial verb, which has a samesubject marker or a different-subject marker, and a final clause with a final tenseinflected verb. It is typical of clause chains in TNG languages of eastern New Guinea, like the Madang language Usan and Kalam, that medial verbs are different from final verbs, in particular: (i) not all medial verbs are marked for subject person/ number, (ii) medial verbs often cannot be marked for tense or mood categories, and (iii) the markers that medial and final verbs allow are formally and functionally distinct (i.e., reference-tracking function in medial verbs and TAM meanings in final verbs).

1032

Sebastian Fedden

Mian, Telefol, Tifal and Faiwol have a reference-tracking system which operates in clause chains, but unlike the languages of the eastern Highlands, their systems are less grammaticalized,14 which is also characteristic of languages of the West Papuan Highlands, e.g., Dani (Bromley 1981) and Kapauku (Ekagi) (Bill Foley, p. c.). There are generally few differences between medial and final verbs; more strikingly the forms which express tense and aspect meanings in final verbs are also used in medial verbs to convey reference-tracking information. This is illustrated for Mian -b, which means ipfv in final verbs, in (32), and (ds.)sim, in (33). (32) Mian ngaan-b-e=be call.ipfv-ipfv-3sg.m.sbj=decl ‘He is calling.’ (33) Mian ngaan-b-e=a naka=i wentê-n-ib=a call.ipfv-(ds.)sim-3sg.m.sbj=med man=pl.an hear.pfv-seq-3pl.an.sbj=med ‘While he was calling, the men heard (him), and then …’ [Dafinau] The meaning ds appears in brackets because it is only an implicature based on the temporal structure of events and not a fully grammaticalized and entailed meaning. The formal identity and the semantic similarity of the two -b suffixes suggest a historical connection. In final verbs, -b signals that an event is on-going at a specific temporal reference point, typically the moment of speaking (its origin being the existential verb). In medial verbs, -b indicates that an action is on-going when the next event is taking place, i.e., the two actions are overlapping and are at least partially simultaneous. The temporal structure of simultaneity invites an implicature of ds according to the assumption that “… we can normally expect that actions in succession are performed by the same person, while actions that overlap are performed by different people” (Longacre 1983: 198). However, this is an implicature in the Mian reference-tracking system, as there are examples of -b in medial verbs being compatible with conjoint reference as long as events are simultaneous (see Fedden 2012). This happens infrequently, but it means that -b comes with a defeasible implicature of ds. An example is given in (34).

14 The terse description in Weber (1997: 71–72) suggests that Bimin is between these two types of system. It is like Usan in that ss-medial verbs are only inflected with a single suffix and in that there is less overlap between the sets of forms used in medial and in final verbs than in the other Mountain Ok languages. ds-medial verbs on the other hand are morphologically more elaborate than in the eastern Highlands languages, which is more reminiscent of Mountain Ok.

Grammaticalization in Mountain Ok (Papua New Guinea)

1033

(34) Mian ke-ha-b-e=a kimâa’-bi-Ø-e bita do-3sg.n1.r-(ds.)sim-3sg.m.sbj=med care_for.pfv-cont-ipfv-3sg.m.sbj until ‘While cleaning it he was caring for it (a plant) until, …’ [Afoksitgabaam] That ds is not part of the meaning of the cognate form in Tifal, not even through an implicature, can be seen from the fact that it is compatible with speech act participants. In (35) the subject of all three clauses is first person singular and there is no question as to whether this is a ss or a ds context. (35) Tifal (Boush 1975: 28; glosses mine) albita-bad-i kee-b-i lang diki-bad-i15 garden work-sim-1sg.sbj weed-sim-1sg.sbj do-prs-1sg.sbj ‘I worked and weeded the garden.’ These examples suggest that the meaning of -b in medial verbs in Mountain Ok is sim, which makes sense if -b comes from a marker of imperfectivity. There is evidence that ds is an implicature in Mian. For a thorough analysis of the Mian reference-tracking system, see Fedden (2011: ch. 11, 2012). (36) Mountain Ok (except Bimin) exist > continuous > simultaneous (> different subject) While the development of ‘exist’ into a continuous marker is well-attested (Heine and Kuteva 2002: 127) reference-tracking markers with a relation to aspectual morphology do not seem to be common. Existing theories of the origin of such markers in Papuan languages trace them back either to gapping under identity, giving rise to zero marking for SS (Haiman 1983), or to a pronominal or deictic origin. See Roberts (1997: 190‒192) for a condensed summary of these theories of origin and an evaluation of their plausibility. Homophony between aspect and reference-tracking markers is not well attested typologically either, but see Jacobsen (1983: 174‒177) on some formal similarities of reference-tracking and aspectual morphology in North American languages.

5 Reciprocals In Mian reciprocals, a clause-chaining construction has grammaticalized into a reciprocal suffix -sese. For reasons of space I will only present key points here. For a

15 Note that the Tifal medial verb suffix contains some additional material /Vd/ of unknown origin. The situation in Telefol and Faiwol is similar. Some scholars treat this additional material as be-

1034

Sebastian Fedden

detailed account of the development of Mian reciprocal constructions, see Fedden (2013). An example of the sese-construction is (37): (37) Mian ya-têm’-sese bl-Ø-io=be pl.an.obj-see.pfv-recp exist-ipfv-3pl.an.sbj=decl ‘They are glancing at each other.’ The participants in a reciprocal event occupy the same set of argument positions they would in a normal clause, i.e., they have to be (i) subjects and (ii) objects. The existential verb agrees with the whole set of reciprocants and the set of reciprocants is cross-referenced on the verb with an object prefix, cf. ya- in (37) above. However, example (37) is strongly dispreferred for situations with just two participants. In such a situation, the non-subject reciprocant is cross-referenced by a suffix in the singular, so Mian merges the agents through plural subject marking, whereas it keeps the patients apart through singular object marking. (38) Mian a-têm’-sese bl-Ø-io=be 3sg.m.obj-see.pfv-recp exist-ipfv-3pl.an.sbj=decl ‘They (two men) are glancing at each other.’ There is a constructional variant which is syntactically quite different and looks more like a clause chain with different-subject suffixes (-s, also expressing event sequentiality) on the verbs which describe the reciprocal subevents. However, the constructional variant still has dedicated reciprocal semantics. Example (39) illustrates this constructional variant: (39) Mian a-têm’-s-e a-têm’-s-e 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj bl-Ø-io=be exist-ipfv-3pl.an.sbj=decl ‘They (two men) are glancing at each other.’ As in the sese-construction above, the existential verb is in the plural summarizing the reciprocal action as a whole. The different-subject suffix -s behaves in an unusual non-linear or circular fashion. In clause chaining constructions, -s is only anticipatory. Evans (2010) coined the term “unified zigzag construction” for such con-

longing to the subject suffix rather than the reference-tracking suffix (cf. Mecklenburg and Mecklenburg [1977: 17] for Faiwol). This issue needs further research.

Grammaticalization in Mountain Ok (Papua New Guinea)

1035

structions where “a complex form of verb chaining [is] zigzagging between subevents (successive transitive verbs, each marked with a different subject marker, and agreeing with one actor in person and number followed by an intransitive summary auxiliary agreeing with the whole set)” (Evans 2010: 17). A similar reciprocal construction can be found in the Papuan languages Amele (Roberts 1987) and Hua (Haiman 1980). The development from the zigzag-construction (V-s-e V-s-e) to the sese-construction (V-sese) was possibly due to phonological reduction. The deleted material is struck through in (40), which otherwise is an exact repetition of (39) above: (40) Mian a-têm’-s-e a-têm’-s-e 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj bl-Ø-io=be exist-ipfv-3pl.an.sbj=decl ‘They (two men) are glancing at each other.’ If this is true the origin of the zigzag-construction was a biclausal description of the reciprocal situation, consisting of two medial clauses chained together, which express the bounded subevents making up the reciprocal situation. The predicate that expresses the reciprocal action is the same in both clauses but the argument positions are reversed, according to the template NPj V-s NPk and then NPk V-s NPj, where j≠k. Each participant is once encoded as the starting point and once as the endpoint of the reciprocal subevent. Consider example (41): (41) Mian a-têm’-s-e=a 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj=med a-têm’-s-e=a 3sg.m.obj-see.pfv-ds.seq-3sg.m.sbj=med ‘hej glances at himk and hek glances at himj, and then …’ (where j≠k) OR ‘hej glances at himm and hek glances at himm, and then …’ (where j≠k≠m) In the reciprocal interpretation of (41) there are two male referents and two sequential glancing events which are expressed in a clause chaining construction. Note that reciprocal semantics are not entailed here. A non-reciprocal reading is possible, where there are three men and two of them (indexed as j and k) glance at the third man (indexed as m). (42) Mian clause chain (involving -s ‘ds.seq’) > reciprocal

1036

Sebastian Fedden

Mian reciprocals are grammaticalizations from clause chains which were originally used to describe both subevents of a reciprocal situation involving switch-reference morphology. Although at least Amele and Hua have the zigzag-construction with a similar origin as in Mian, only Mian seems to have gone the additional step and fused the two clauses together resulting in a single reciprocal suffix -sese. This type of grammaticalization of reciprocals remains unique.

6 Conclusion A first study in grammaticalization in a family is necessarily selective. In this chapter it was my aim to present the most important grammaticalization channels in Mountain Ok languages, both the ones that are well-established in the literature and the less familiar ones. I will begin this conclusion with a summary of the grammaticalization channels that we find in Mountain Ok. (43) a. demonstrative > personal pronoun and article b. interior, hole > in; top (surface) > on (spatial) > on (temporal) c. move > persisitve (Mian) d. leave > completive (Mian) e. go > habitual (Tifal) f. exist > continuous (except Bimin) g. verb (in a serial verb construction) > perfective, imperfective h. hit > perfective (Mian) i. free personal pronoun > agreement j. give > benefactive (except Faiwol and Bimin) k. exist-for > possessive (Mian) l. top (surface) > on > possessive (Mian) m. sleep + impersonal inflection > hesternal past (Mian) n. exist > imperfective > simultaneous (> different subject) (except Bimin) o. clause chain (involving -s ‘ds.seq’) > reciprocal (Mian) The Mountain Ok languages show a range of grammaticalization patterns which are well-known the world over. In the nominal domain we find the widely attested development of demonstratives into both pronouns and articles (43a). Widespread is also the development of nouns into spatial terms and (in the case of on) into temporal ones (43b). In the verbal domain, aspect is of central importance. Here we find layers of grammaticalization: (43c‒e) use of independent verbs with an aspectual function, e.g., Tifal une ‘go’ for habitual meaning, (43f) productive suffixation with -bi ‘continuous’, and (43g‒h) unproductive aspect suffixes, which plausibly come from verbs but whose verbal etyma can in most cases not be identified anymore. The development of the agreement system through the incorporation of pro-

Grammaticalization in Mountain Ok (Papua New Guinea)

1037

nouns into the verb (43i) is plausible, but cannot be proven with Mountain Ok data because the shift from free pronouns to bound (agreement) forms would have happened in pre-Mountain Ok times. Less well-known from European languages but still amply attested in other parts of the world is the grammaticalization of the verb ‘give’ into a benefactive (43j) introducing a further argument with dative-like semantics. This construction feeds into the development of possessives (43k) and fits into Heine’s (1997) Goal Schema. The development of locatives into possessives is typologically well-attested (43l). On the other hand, there are some less well-established grammaticalization channels. First, Mian seems to have grammaticalized the verb s ‘sleep’ into a marker of the hesternal past (43m), which is a rare grammaticalization channel, though semantically plausible. Second, the reference-tracking systems in clause chaining constructions of the Mountain Ok languages are less strongly grammaticalized than in the languages of the eastern Highlands. The meaning of event simultaneity developed from an aspect suffix. At least in Mian the meaning of different subject is a defeasible implicature (43n). The development of a switch-reference marker from an aspect marker is cross-linguistically rare. Third, and possibly most rare, there is evidence that reciprocals in Mian are grammaticalizations from clause chains which describe both subevents of a reciprocal situation (43o). The study of the Mountain Ok branch of the Ok family from Papua New Guinea corroborates the validity of many well-known grammaticalization channels, but it also lets us discover new pathways, including the unique development of the Mian reciprocal from a clause chain to the zigzag-construction to a suffix.

Acknowledgements I thank the editors and one anonymous reviewer for helpful comments and suggestions. I would also like to thank Matthew Baerman, Grev Corbett, Nick Evans, František Kratochvíl, Bruno Olsson, Andy Pawley, Malcolm Ross, Edgar Suter and Bernhard Wälchli for reading and commenting on a previous version of this chapter. Many thanks to Edgar Suter, Timothy Usher and Paul Widmer for detailed discussion of the diachrony of Mountain Ok pronouns, articles and demonstratives (in section 2.2). All remaining errors and inconsistencies are my own. Versions of this paper were presented in the Newcastle Linguistics Seminar Series at the University of Newcastle (Australia), October 2016; at the 9th Austronesian and Papuan Languages and Linguistics conference (APLL 9), Paris, June 2017; and at the Surrey Morphology Group 25th Anniversary Workshop (September 2017). I thank the respective audiences for helpful discussion and comments.

1038

Sebastian Fedden

Abbreviations 1 = first person, 2 = second person, 3 = third person, an = animate, aux = auxiliary, ben = benefactive, conn = connective, cont = continuous, decl = declarative, def = definite, dem = demonstrative, dist = distal, ds = different subject, f = feminine, f_cl = f-classifier, foc = focus, fpst = far past, gpst = general past, hpst = hesternal past, imm = immediate, ipfv = imperfective, irr = irrealis, m = masculine, med = medial, n1 = neuter 1, nanpl = not animate plural, neg = negation, obj = object, pfv = perfective, pl = plural, poss = possessive, pq = polar question, prs = present, pst = past, r = recipient, real = realis, recp = reciprocal, rpst = remote past, sbj = subject, seq = sequential, sg = singular, sim = simultaneous, ss = same subject, vn = verbal noun.

References Baerman, Matthew, Dunstan Brown & Greville G. Corbett. 2005. The syntax-morphology interface: A study of syncretism. Cambridge: Cambridge University Press. Bisang, Walter. 2011. Grammaticalization and linguistic typology. In Heiko Narrog & Bernd Heine (eds.), The Oxford handbook of grammaticalization, 105‒117. Oxford: Oxford University Press. Boush, Alfred. 1975. Tifal grammar essentials. Ms., Ukarumpa, Summer Institute of Linguistics. Boush, Alfred. 1979. Aspect on Tifal final and medial verbs. Ms., Ukarumpa, Summer Institute of Linguistics. Boush, Alfred & Susan Boush. 1974. Tifal phonology. Ms., Ukarumpa, Summer Institute of Linguistics. Bowden, John. 1992. Behind the preposition: Grammaticalisation of locatives in Oceanic languages. Canberra: Pacific Linguistics. Bromley, H. Myron. 1981. A grammar of Lower Grand Valley Dani. (Pacific Linguistics C-63). Canberra: Australian National University. Bybee, Joan L. & Östen Dahl. 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13(1). 51‒103. Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge University Press. Corbett, Greville G. & Sebastian Fedden. 2016. Canonical gender. Journal of Linguistics 52. 495‒531. Corbett, Greville G., Sebastian Fedden & Raphael A. Finkel. 2017. Single versus concurrent feature systems: nominal classification in Mian. Linguistic Typology 21. 209‒260. Dahl, Östen. 1985. Tense and aspect systems. Cambridge: Cambridge University Press. Dahl, Östen. 2000. The tense and aspect systems of European languages in a typological perspective. In Östen Dahl (ed.), Tense and aspect in the languages of Europe, 3‒25. Berlin: Mouton de Gruyter. Dahl, Östen. 2013. Tense-aspect-mood-evidentiality and the organization of human memory. In Karina Veronica Molsing & Ana Maria Tramunt Ibaños (eds.), Time and TAME in language, 22‒53. Newcastle: Cambridge Scholars Publishing. Diessel, Holger. 1999. Demonstratives: Form, function, and grammaticalization (Typological Studies in Language 42). Amsterdam: John Benjamins. Döhler, Christian. 2016. A grammar of Komnzo. Canberra, Australia: Australian National University dissertation. https://openresearch-repository.anu.edu.au/bitstream/1885/107178/1/ Doehler%20Thesis%202016.pdf (accessed 23 September 2019). Donohue, Mark. 2008. Pronouns, clitics, orders and grammaticalization in Tukang Besi. Studies in Philippine Languages and Cultures 17. 1–12.

Grammaticalization in Mountain Ok (Papua New Guinea)

1039

Drabbe, Petrus. 1955. Spraakkunst van het Marind: Zuidkust Nederlands Nieuw-Guinea [A Grammar of Marind: South coast of Dutch New Guinea]. Wien-Mödling: Drukkerij van het Missiehuis St. Gabriël. Evans, Nicholas. 2015. Inflection in Nen. In Matthew Baerman (ed.), The Oxford handbook of inflection, 543‒575. Oxford: Oxford University Press. Fedden, Sebastian. 2007. Women, houses, and plural objects? – Homophony in the Mian gender system. In Robyn Loughnane, Cara P. Williams & Jana Verhoeven (eds.), In between wor(l)ds. Transformation and translation (School of Languages and Linguistics Postgraduate Research Papers on Language and Literature, vol. 6), 183‒198. Melbourne: University of Melbourne, School of Language and Linguistics. Fedden, Sebastian. 2010. Ditransitives in Mian. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in ditransitive constructions: A comparative handbook, 456‒ 485. Berlin: Mouton de Gruyter. Fedden, Sebastian. 2011. A grammar of Mian (Mouton Grammar Library 55). Berlin: Mouton de Gruyter. Fedden, Sebastian. 2012. Switch-reference and temporal reference in Mian. In Volker Gast & Holger Diessel (eds.), Clause combining in cross-linguistic perspective, 393‒413. Berlin: Mouton de Gruyter. Fedden, Sebastian. 2013. Reciprocals in Mian. Studies in Language 37(1). 58‒93. Fedden, Sebastian. 2015. Verb stem aspect in Mian. In Sandra Augendre, Graziella CouasnonTorlois, Déborah Lebon, Clément Michard, Gilles Boyé & Fabio Montermini (eds.), Proceedings of the Décembrettes 8th International conference on morphology, Bordeaux, 6‒7 December 2012, Carnets de Grammaire, rapports internes de CLLE-ERSS, report 22, 99‒135. University of Toulouse. http://w3.erss.univ-tlse2.fr/textes/publications/CarnetsGrammaire/ carnGram22.pdf (accessed 23 September 2019). Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Foley, William A. 2000. The languages of New Guinea. Annual Review of Anthropology 29. 357‒404. Foley, William A. & Robert Van Valin. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press. Greenberg, Joseph H. 1978. How does a language acquire gender markers? In Joseph H. Greenberg (ed.), Universals of human language. Vol. 3: Word structure, 47‒82. Stanford: Stanford University Press. Haiman, John. 1980. Hua: A Papuan language of the Eastern Highlands of New Guinea. Amsterdam: John Benjamins. Haiman, John. 1983. On some origins of switch reference marking. In John Haiman & Pamela Munro (eds.), Switch reference and universal grammar, 105‒128. Amsterdam: John Benjamins. Haiman, John & Pamela Munro (eds.). 1983. Switch reference and universal grammar. Amsterdam: John Benjamins. Hammarström, Harald, Robert Forkel & Martin Haspelmath. 2017. Glottolog 3.0. Jena: Max Planck Institute for the Science of Human History. Healey, Alan. 1964a. A survey of the Ok family of languages, reconstructing Proto-Ok. Canberra, Australia: Australian National University dissertation. Healey, Alan. 1964b. Telefol phonology. (B-3.) Canberra: Pacific Linguistics. Healey, Phyllis. 1965a. Telefol noun phrases. (B-4.) Canberra: Pacific Linguistics. Healey, Phyllis. 1965b. Telefol clause structure. In Linguistic Circle of Canberra Publications. (A-5.), 1‒26. Canberra: Pacific Linguistics. Healey, Phyllis. 1965c. Telefol verb phrases. In Linguistic Circle of Canberra Publications. (A-5.), 27‒53. Canberra: Pacific Linguistics.

1040

Sebastian Fedden

Healey, Phyllis. 1966. Levels and chaining in Telefol sentences. (B-5.) Canberra: Pacific Linguistics. Healey, Phyllis & Walter Steinkraus. 1972. A preliminary vocabulary of Tifal with grammar notes. Santa Ana, CA: Summer Institute of Linguistics. Healey, Phyllis & Alan Healey. 1977. Telefol dictionary. (C-46.) Canberra: Pacific Linguistics. Heine, Bernd. 1997. Possession: Sources, forces, and grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Himmelmann, Nikolaus P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Jacobsen, William H. 1983. Typological and genetic notes on switch-reference systems in North American Indian languages. In John Haiman & Pamela Munro (eds.), Switch reference and universal grammar, 151‒183. Amsterdam: John Benjamins. Lahiri, Aditi. 2000. Hierarchical restructuring in the creation of verbal morphology in Bengali and Germanic: Evidence from phonology. In Aditi Lahiri (ed.), Analogy, levelling, markedness: Principles of change in phonology and morphology (Trends in Lingustics 127), 71‒123. Berlin: Mouton de Gruyter. Lang, Adrienne. 1975. The semantics of classificatory verbs in Enga (and other Papua New Guinea languages). Pacific Linguistics B-39, 1–246. Lichtenberk, František. 1991. Semantic change and heterosemy in grammaticalization. Language 67. 474–509. Longacre, Robert E. 1983. Switch Reference systems in two distinct linguistic areas: Wojokeso (Papua New Guinea) and Guanano (Northern South America). In John Haiman & Pamela Munro (eds.), Switch Reference and universial grammar, 185‒207. Amsterdam: John Benjamins. Loughnane, Robyn. 2009. A grammar of Oksapmin. Melbourne, Australia: University of Melbourne dissertation. https://minerva-access.unimelb.edu.au/bitstream/handle/11343/35153/115303_ loughnane.pdf?sequence=1 (accessed 23 September 2019). Loughnane, Robyn & Sebastian Fedden. 2011. Is Oksapmin Ok? A study of the genetic relatedness of Oksapmin and the Ok languages. Australian Journal of Linguistics 31. 1‒40. Mecklenburg, Charlotte. 1974. Phonology of Faiwol. In Richard Loving (ed.), Workpapers in Papua New Guinea languages: Studies in languages of the Ok family, vol. 7. 143‒165. Ukarumpa, PNG: Summer Institute of Linguistics. Mecklenburg, Frank & Charlotte Mecklenburg. 1977. A brief grammar of Faiwol. Ms. Ukarumpa, Summer Institute of Linguistics. Newman, John. 1996. Give: A cognitive linguistic study. Berlin: Mouton de Gruyter. Palmer, Bill. 2002. Absolute spatial reference and the grammaticalisation of perceptually salient phenomena. In Giovanni Bennardo (ed.), Representing space in Oceania: Culture in language and mind (Pacific Linguistics 523), 107–157. Canberra: Pacific Linguistics. Pawley, Andrew K. 1993. A language which defies description by ordinary means. In William A. Foley (ed.), The role of theory in language description, 87–129. Berlin: Mouton de Gruyter. Pawley, Andrew K. 2005. The chequered career of the Trans New Guinea hypothesis: Recent research and its implications. In Andrew Pawley, Robert Attenborough, Jack Golson & Robin Hide (eds.), Papuan pasts: Cultural, linguistic and biological histories of Papuan-speaking peoples, 67‒108. Canberra: Pacific Linguistics. Pawley, Andrew K. 2008. Compact versus narrative serial verb constructions in Kalam. In Gunter Senft (ed.), Serial verb constructions in Austronesian and Papuan languages, 171–202. Canberra: Pacific Linguistics. Reesink, Ger P. 1983. Switch reference and topicality hierarchies. Studies in Language 2. 215–246. Reesink, Ger P. 2013. Expressing the GIVE event in Papuan languages: A preliminary survey. Linguistic Typology 17(2). 217–266.

Grammaticalization in Mountain Ok (Papua New Guinea)

1041

Roberts, John R. 1987. Amele (Croom Helm descriptive grammar series). London: Croom Helm. Roberts, John R. 1997. Switch reference in PNG. In Andrew K. Pawley (ed.), Papers in Papuan linguistics, 101–241. Canberra: Australian National University. Ross, Malcolm D. 1980. Some elements of Vanimo, a New Guinea tone language. Pacific Linguistics A-56, 77–109. Ross, Malcolm D. 2004a. The grammaticization of directional verbs in Oceanic languages. In Isabelle Bril & Françoise Ozanne-Rivierre (eds.), Complex predicates in Oceanic languages: Studies in the dynamics of binding and boundness, 297–330. Berlin: Mouton de Gruyter. Ross, Malcolm D. 2004b. Demonstratives, local nouns and directionals in Oceanic languages: A diachronic perspective. In Gunter Senft (ed.), Deixis and demonstratives in Oceanic languages (Pacific Linguistics 562), 175–204. Canberra: Pacific Linguistics. Ross, Malcolm D. 2005. Pronouns as a preliminary diagnostic for grouping Papuan languages. In Andrew Pawley, Robert Attenborough, Jack Golson & Robin Hide (eds.), Papuan pasts: Cultural, linguistic and biological histories of Papuan-speaking peoples, 15‒66. Canberra: Pacific Linguistics. Scott, Graham. 1978. The Fore language of New Guinea. (B-47.) Canberra: Pacific Linguistics. Siegel, Jeff. 2014. The morphology of tense and aspect in Nama, a Papuan language of southern New Guinea. Open Linguistics 1(1). 211–231. Simons, Gary F. & Charles D. Fennig (eds.). 2017. Ethnologue: Languages of the World, 20th edn. Dallas, Texas: SIL International. Staalsen, Philip. 1972. Clause relationships in Iatmul. Pacific Linguistics A-31, 45‒69. Steinkraus, Walter. 1963. Tifal tone. Ms. Ukarumpa, Summer Institute of Linguistics. Steinkraus, Walter. 1969. Tifal phonology showing vowel and tone neutralization. Kivung 2. 57‒66. Stirling, Lesley. 1993. Switch-reference and discourse representation. Cambridge: Cambridge University Press. Usher, Timothy. 2014. Ok. A section of “Newguineaworld”: Classification and reconstruction of Papuan language families. Online manuscript. https://sites.google.com/site/ newguineaworld/families/trans-new-guinea/central-west-new-guinea/digul-river-ok/ok (accessed 23 September 2019). Usher, Timothy & Edgar Suter. 2015. The Anim languages of southern New Guinea. Oceanic Linguistics, 54(1). 110‒142. Van Gijn, Rik & Jeremy Hammond (eds.). 2016. Switch reference 2.0 (Typological Studies in Language 114). Amsterdam: John Benjamins. Weber, Thomas. 1997. Bimin grammar essentials. Ms., Ukarumpa, Summer Institute of Linguistics. Wälchli, Bernhard. 2018. The rise of gender in Nalca (Mek, Tanah Papua): The drift towards the canonical gender attractor. In Sebastian Fedden, Jenny Audring & Greville G. Corbett (eds.), Non-canonical gender systems. 68‒99. Oxford: Oxford University Press. Wurm, Stephen. 1982. Papuan languages of Oceania. Tübingen: Narr. Zaliznjak, Andrej A. 1973. O ponimanii termina ‘padež’ v lingvističeskix opisanijax [Interpreting the term ‘case’ in linguistic descriptions]. In Andrej A. Zaliznjak (ed.), Problemy grammatičeskogo modelirovanija, 53–87. Moscow: Nauka. [Reprinted in Andrej A. Zaliznjak. 2002. Russkoe imennoe slovoizmenenie: s priloženiem izbrannyx rabot po sovremennomu russkomu jazyku i obščemu jazykoznaniju, 613–647. Moscow: Jazyki slavjanskoj kuľtury.]

Nikolaus P. Himmelmann

25 Grammaticisation processes and reanalyses in Sulawesi languages 1 Introduction 1.1 Why Sulawesi? The Indonesian island of Sulawesi is home to more than one hundred languages belonging to eleven different low-level subgroups of the Western Malayo-Polynesian (WMP) linkage of Austronesian. The three north-easternmost of these groups have been argued by Blust (1991) to belong to the Greater Central Philippines subgroup. These are the Sangiric group, represented here by Ratahan, the Minahasan group, to which Tondano and Totemboan belong, and the Gorontalo-Mongondow group, represented here by Buol. With the exception of the extensive South Sulawesi group, which includes Buginese and Makassarese, the remaining subgroups have been argued to belong to a Celebic ‘supergroup’ by Mead (2003), revising an earlier proposal by van den Berg (1996). This proposal, however, has not yet been widely accepted. Proceeding from the northwest to the southeast, the seven subgroups of the putative Celebic supergroup are (member languages referred to in this chapter given in parentheses): Tomini-Tolitoli, the group the author is most familiar with (Lauje, Tajio, Totoli),1 Kaili-Pamona (Da’a, Uma), Saluan-Banggai (Balantak), Bunggku-Tolaki (Mori, Padoe, Tolaki), Wotu-Wolio, Muna-Buton (Muna) and Tukang Besi (Tukang-Besi). Sulawesi belongs to the most linguistically diverse islands in the Austronesian world with regard to morphosyntactic structures (much less so with regard to phonology). More specifically, Sulawesi languages provide evidence for some of the more dramatic restructuring processes that have occurred in the Austronesian family, while at the same time also witnessing grammaticisation and reanalysis processes found in other branches of the family. Thus, to a certain extent, Sulawesi languages may be considered to be representative for such processes in Austronesian more generally. However, it should be clearly understood that Sulawesi languages do not provide a comprehensive picture in this regard. In fact, within the confines of the present chapter, it will not even be possible to cover comprehensively all relevant processes attested in these languages. In line with many Austronesian languages, most Sulawesi languages share a preference for verb-initial clause structures. They differ as to whether they also

1 Data on these languages are from the author’s field notes, if no other source is indicated. Totoli examples are all from the online corpus of Totoli natural speech, see Leto et al. (2005–2010). https://doi.org/10.1515/9783110712735-011

1044

Nikolaus P. Himmelmann

make regular use of an alternative construction where the subject argument appears in clause-internal preverbal position (which needs to be distinguished from a clause-external topic position available at the left edge in all languages). This is illustrated by the Totoli examples in (1) which also illustrate the fact that many Sulawesi languages allow for two (or more) transitive constructions of more or less equal status. The prototypical systems of this type are widely known as symmetrical voice systems (formerly also as Philippine-type ‘focus’ systems), which are attested throughout the Philippines and the northern half of Sulawesi. Such systems contrast an actor voice, where the actor argument is the subject (Ali in [1a]), with one or more undergoer voices, where the undergoer is the subject (deuk in [1b]). (1)

Totoli: Actor voice vs. undergoer voice a. i Ali nambalung deuk itu i Ali non2-mbalung deuk itu hon av.rls-throw:at dog dist ‘Ali threw (stones) at the dog.’ b. deuk itu ni-mbalung i Ali dog dist uv.rls-throw:at hon Ali ‘Ali threw (stones) at the dog.’

Unlike passive alternations, these alternations are symmetrical in that they are both marked by voice morphology (e.g., in [1a] by the actor voice realis prefix non-, and in [1b] by the undergoer voice realis prefix ni-) and case marking patterns are identical across the two constructions (in the case of Totoli, both NPs are not marked for case). Further, subject and non-subject arguments respectively show identical behavioural properties. So, for example, only the subjects (= the pre-verbal arguments) in (1a) and (1b) could function as heads of relative clauses, and only the non-subject arguments (= the post-verbal arguments in [1a] and [1b]) form a constituent with the verb. Riesberg (2014a) provides a detailed analysis of such systems and shows that the degree of symmetry varies significantly across the family, with Totoli showing almost perfect symmetry. A common trait of Sulawesi languages is the occurrence of person markers (affixes or clitics) on the verb, with massive variation in terms of the number and completeness of person marking paradigms and their functions (see section 3.3). Related to the variation in person marking is considerable variation in alignment systems, with Sulawesi languages having been analysed as nominative-accusative, ergative or symmetrical voice languages. Further widely attested verbal morphology

2 Capital n represents a homorganic nasal which assimilates to or substitutes for the base-initial consonant; it most often substitutes for the base-initial consonant if it is voiceless (see Newman [1984] and Pater [2001] for further details).

Grammaticisation processes and reanalyses in Sulawesi languages

1045

includes applicatives and reciprocals. Verbal mode marking (realis vs. non-realis) is also widespread, but does not occur in the South Sulawesi subgroup. Nominal modifiers other than numerals typically follow the head noun, which is normally unmarked but may be preceded by determiner-like or case-marking clitics (see Section 2.4). Numeral classifiers are widespread and typically precede the head noun together with the numeral, with post-nominal position often being a secondary option.

1.2 Introductory observations on grammaticisation in Austronesian To date, most of the work on grammaticisation in Austronesian languages relates to the extensive Oceanic subgroup (> 500 languages), while relatively little has been done on the western half of the family, which next to the WMP linkage and the Formosan languages also includes the Central Malayo-Polynesian (CMP) and the South Halmahera-West New Guinea (SHWNG) languages. Work on Oceanic deals with the grammaticisation of articles and possessive markers, directionals/prepositions/serial verbs, TAM marking, or reason and cause markers, inter alia, as perhaps best exemplified in the work of the late František Lichtenberk (e.g., 1991, 2013a, 2013b). The reason why there is so much more work on grammaticisation in Oceanic is that the history of this subfamily is reasonably well understood. There is a relatively clear and widely agreed reconstruction of Proto-Oceanic grammar and lexicon as well as ample and transparent variation among closely related languages, relatively many of which have been carefully described and documented. Hence, both source items and source constructions are relatively easy to identify, and there is little doubt about the directionality of the attested changes. The situation in the remainder of the family is different for a number of reasons. As for CMP and SHWNG, good comprehensive descriptions are still few, and hence the synchronic basis for identifying grammaticisations is insufficient. Relatively speaking, there are more and better descriptions for Formosan and WMP languages (but still less than for Oceanic), but here the main problem pertains to the fact that much of the core grammatical morphology reconstructed to the proto-level persists in present-day languages or has been lost without leading to major new grammaticisations (exceptions to be noted in the following sections). That is, what is amply attested is the reorganisation of a grammatical system centring on voice morphology and determiner-like elements. These changes, however, may not be instances of grammaticisation in the strict sense of an (essentially self-propelled) development along a unidirectional cline, but rather may instantiate reanalyses of various types, including analogical extensions. More importantly perhaps, more often than not the directionality of the attested changes is unclear (is construction X the predecessor of construction Y or the other way around?), as further exemplified below. Finally,

1046

Nikolaus P. Himmelmann

Tab. 1: The development of the actor voice paradigms in Bikol, R- = reduplication of stem-initial syllable (cp. Lobel 2013: 47).3 Old Bikol (C. 1610)

Middle Bikol (C. 1879)

Modern Bikol

-um-

mag-

-um-

mag-

mag-

Non-realis

-um-

mag-

-um-

mag-

mag-

Realis

-imin-

nag-

-imin ~ -umin- nag-

nag-

Realis imperfective

na- ~mina-

nag-R-

mina-

nag-R-

nag-R-

Non-realis imperfective

ma-

mag-R-

ma-

mag-R-

ma:-

Subjunctive

ø-

pag-

---

---

---

Subjunctive imperfective

R-

pag-R-

---

---

---

Tab. 2: Proclitic person markers in Central Sulawesi Languages. Uma Kaili-Pamona

Da’a Kaili-Pamona

Pendau Tomini

Lauje Tomini

1s

ku-

ku-

'u-

'u-

2s

nu-

mu-

mu-

-

3s

na-

-

-

-

1p.in

ta-

-

-

-

1p.ex

ki-

-

-

-

2p

ni-

-

-

-

3p

ra-

-

-

-

apart from Malay and Javanese, which are intermittently attested from the 7th and 9th century AD onward, historical records for individual languages are scarce and, with few exceptions, date to the colonial period at the earliest. Table 1 provides a typical example for the kind of changes attested in historical records. It shows the reorganisation of the aspect-mode paradigm for the actor voice infix -um- (inserted after the first consonant of the stem) and the actor voice prefix mag- in the Philippine language Bikol. These affixes functionally correspond to the Totoli prefix noN- in example (1a) above (there are no major functional differences

3 Lobel uses tense-based terminology (e.g., past) instead of the mode-aspect based terminology (e.g., realis) preferred in this chapter.

Grammaticisation processes and reanalyses in Sulawesi languages

1047

between infix and prefix, the choice between them largely depending on the stem). The development is essentially one of reduction and consolidation, with one mode (subjunctive) and the infix -um- being lost. Table 2 illustrates the problem of determining the direction of change. It lists proclitic person markers in a number of Central Sulawesi languages, the paradigms varying significantly with regard to the number of occupied cells. Van den Berg (1996) argues that Lauje represents the endpoint of a reduction of a formerly complete paradigm, while Himmelmann (1996) believes that Lauje shows the beginning of the build-up of a person-marking paradigm seen in full in Uma.

2 Grammaticisation of nominal categories 2.1 Numeral classifiers In Austronesian languages quite generally, weakly grammaticised numeral classifiers are often obligatory in count constructions (a major exception being Philippine and Formosan languages). Numerals precede the classifier, and the numeral classifier-unit in turn precedes the head noun. Post-nominal position, however, is often also an option. In example (2) from Tajio, lower numerals are prefixed to the classifier. (2)

Tajio: Classifier construction (Mayani 2013) a. so-bua te=motor one-clf.piece art=motorbike ‘one motorbike’ b. te=oto ro-bua art=car two-clf.piece ‘two cars’

The inventories of classifiers are usually relatively small (< a dozen) and include formatives which also function as independent nouns (illustrated by Tajio in Table 3). In cases where the forms of the noun and the related classifier differ, the difference is usually minimal (cp. puu vs. puung in Table 3). There is usually one general default classifier (often the word for ‘fruit’, ‘seed’ or ‘stone’) for all kinds of objects, but excluding animates. Humans, and sometimes more generally animates, can often be counted directly, i.e., they do not require the classifier construction. Classifier constructions are usually not distinguishable from measure constructions (arguably, then, there is only a single classifier/measure construction in Tajio).

1048

Nikolaus P. Himmelmann

Tab. 3: Tajio Classifiers (Mayani 2013). Classifier

Literal meaning

used in counting

too baang bua vuu puung ndaang lae peka kolo

‘person’ ‘tail’ ‘piece’ ‘seed; bone’ puu ‘tree’ ‘branch’ ‘sheet’ ‘plank’ –

humans, including kin non-human animates including large fish default for objects small round objects (including fruit), small fish, cigar-like shapes Trees Leaves thin and flat objects flat and hard objects means of transport (boat, car, motorbike), rarely used

The latter make use of the same form of the numerals and the same overall construction pattern, with measure nouns taking the place of the classifying nouns.4 (3)

Tajio: Measure construction (Mayani 2013) lima-m-pulu karung te=uli five-lk-ten sack art=skin ‘fifty sacks of peelings (lit. skin)’

2.2 Number Nouns are generally not marked for number in Sulawesi languages. A major exception is Muna, where nouns may be suffixed with -hi for plural (e.g., sau ‘tree’, sauhi ‘trees’, van den Berg [1989: 288]), but this suffix occurs in all parts of speech and has a wide range of functions, its origin being unknown. Doubling of roots to express plurality (e.g., guru-guru ‘teachers’) may sporadically be found, but this does not seem to be a fully grammaticised marking option, but rather an occasional calque of a Malayic pattern. The Dutch colonial linguist Esser notes in this regard that “[w]hen use is made of reduplication with substantives to indicate plurality, this is a detestable Malayism” (Esser and Mead 2011: 70). Number is regularly distinguished in pronouns and occasionally also in articles, demonstratives, interrogatives, relativisers and the like. Plural articles and pronouns are used to form associative plurals with personal names and kin terms, as in Balantak ari Herman (ari = plural article) ‘Herman and his companions’ (van den

4 In all examples cited from the literature, morpheme analysis and glossing have been adapted to the conventions used throughout this chapter. Free translations are from the original source, unless noted otherwise.

Grammaticisation processes and reanalyses in Sulawesi languages

1049

Berg and Busenitz 2012: 46) or in Mori ondae i Tansumawi (3p hon Tansumawi) ‘Tansumawi and those of his’ (Esser and Mead 2011: 128). Minahasan Totemboan and Tonsawang make use of the plural word manga, which is also widely found in other parts of the Austronesian world, including the Philippines. Sneddon (1978: 104) analyses it as a prefix in Tonsawang, e.g., mangawene (pl-woman) ‘women’.

2.3 Possession In possessive constructions, possessum and possessor (in this order) are usually connected by a kind of linking element that is probably of a demonstrative origin and often formally similar or identical to the third person possessive pronoun, cp. Muna roo-no sau (leaf-lk tree; -no is also the third person possessive suffix) ‘leaf of a tree’ (van den Berg 1989: 86).5 Languages which make consistent use of articles typically have a special genitive form of the article which begins with a dental or alveolar nasal and, historically speaking, in all likelihood is a combination of the possessive linker and the article as, for example, in Tajio topombalu nu=bau (seller gen=fish) ‘seller of fish’. This is regularly the case for the honorific article si or i used with personal names and more generally in polite reference to humans, as in Balantak tama-ni Aman (father-gen.hon Aman) ‘Aman’s father’, cp. i Aman ‘Aman’ (van den Berg and Busenitz 2012: 62). The nasal linking element is often also visible in the possessive pronoun series (suffixes or enclitics), for example in Balantak lima-ngku ‘my hand’ or tama-nta ‘our (incl.) father’ (van den Berg and Busenitz 2012: 11).

2.4 Determiners and case Use of article-like elements is widespread in Sulawesi languages. The exact function of these elements is controversial, as is apparent from the wide range of terms in use for designating them. Next to ‘article’ and ‘determiner’, these include ‘case’, ‘phrase’ or ‘noun (phrase) marker’ (cp. Reid 2002; Blust 2015: 437–439). Three features contribute to their cross-linguistically remarkable profile. First, they typically occur in paradigms of clitic (rarely affixal) elements which combine determiner and case functions. Second, the paradigms always include a distinction between honorific and non-honorific forms, the grammatical properties of these two series diverg-

5 Grammars vary as to whether these linkers are represented as clitics or affixes and whether they form a phonological unit with the preceding head noun (enclitc or suffix), or with the possessor noun (proclitic). It is unclear whether these differences in representation reflect a substantial difference or are simply due to different analytical preferences of the authors.

1050

Nikolaus P. Himmelmann

Tab. 4: Phrase markers in Ratahan (Minahasan), Tondano (Sangiric), and Tagalog. Ratahan

Article gen/poss loc/dat

Tondano

Tagalog Common

hon

ang ng [naŋ] sa

si ni kay

Common

hon

Inanimate Animate

(nu) su

i ni si

(N=) -

si (sg) ni (sg) -

sè (pl) nè (pl)

Tab. 5: Phrase markers in three Tomini languages. Lauje

Article gen/poss loc/dat

Pendau

Tajio

Common

hon

Common

honor

Common

hon

(nu1 ) nu2 li

si ni (li)

(u) nu ri

si ni (ri)

te nu (r)i

si ni (r)i

ing slightly. Third, with few exceptions, they do not signal definiteness, but rather specificity or they lack any clearly specifiable semanto-pragmatic function. In fact, Sulawesi languages show a particularly large variability in the factors governing their use, providing ample evidence for the final stages in the grammaticisation of articles (noun markers in the terminology of Greenberg [1978]). Regarding the first point, Table 4 and Table 5 illustrate typical paradigms from a number of Sulawesi languages and, for comparative purposes, Tagalog. While the labels for the genitive/possessive and locative/dative rows are widely agreed, labels for the row article vary widely and include nominative, subject, topic, and many more. It is very likely that the elements for common nouns in this row historically derive from demonstratives and have a determiner-like function, even though other factors may conspire to restrict their use only to core arguments and possibly only to a single core argument, thus giving them the appearance of case or grammatical function markers.6 A major factor in this regard is the complementary distribution of the elements in each column. The genitive/possessive markers are typically transparently related to the article, consisting of the nasal ligature already discussed in the preceding section and the article, as in Tagalog nang < na+ang or Pendau nu < n+u. Article and genitive/possessive markers are without exception in complementary distribution: a nominal expression can only be preceded by one of these mark-

6 While there may be a superficial similarity with other instances where (often already casemarked) indexicals provide the source for nominal case markers (cp., for example, McGregor 2008), the appearance of a case-marking function in the case of WMP languages essentially depends on paradigmatic interactions and works very differently from the cases reviewed by McGregor.

Grammaticisation processes and reanalyses in Sulawesi languages

1051

ers, never by both. This makes sense, as it were, inasmuch as the genitive markers include an article, historically speaking. The locative/dative markers are usually not transparently related to the article, but often reflect a form ri/di/li or sV. Although this is less clear than in the case of the articles, it is possible that these forms also go back to deictics (and more specifically a deictic adverbial). Importantly, these elements are often the only preposition-like element in a given language (though today the inventory is often extended with loans from Malay). In many languages, complementarity of locative/dative markers with the article and the genitive/possessive marker is also without exception, but in some languages the units marked by the locative marker can be preceded by the articles as in Tagalog ang sa bahay (spec loc house) ‘the one in the house’ (cp. Himmelmann 2015). But the reverse order (i.e., *sa ang X) is not possible. The example of Tajio is particularly instructive in this regard, because in this language vowel-initial nouns are obligatorily marked with the common-noun article te unless preceded by a preposition or the genitive marker nu, as exemplified in example (4). (4)

Tajio (Tomini) (Mayani 2013) te=asu nonavuao te=wani yami puu nu=ayu art=dog av.rls:fall:appl art=wasp from tree gen=wood ‘The dog caused the wasp (nest) to fall down from the tree.’

Use of the common noun article te after yami (*yami tepuu) or nu (*nuteayu) is ungrammatical. One possible explanation for this state of affairs is that grammaticisation of articles is blocked in prepositional phrases (cp. Himmelmann [1998] for further evidence and discussion). Whether such blocking is particularly likely to occur when the inventory of (primary) prepositions is small and prepositions themselves are derived from deictic elements is a matter for further investigation. Note that Tajio te is remarkable in that Tajio is the only Tomini language making use of te, the other eight Tomini languages using (n)u instead. While te in article-like functions is attested elsewhere in Sulawesi (e.g., Banggai, Tukang Besi), it is missing in the area where Tajio is spoken. Turning now to our second point, Table 4 and Table 5 clearly show that articles in Austronesian come in two series, the honorific series for respected humans, which typically include older kin and anthropomorphised protagonists in folktales, and the common noun series for all other nouns. The two series are somewhat unequal in three regards. First, the honorific series is more widely attested than the common noun series, with (s)i as the honorific marker in almost all languages. That is, quite a few languages only have an honorific article but no common-noun article, standard varieties of Malay being the most prominent example. Second, the honorific article tends to be more strongly grammaticised than the common noun article, as it is often inseparable from its host. This is very common for personal pronouns, many of which include (s)i as an obligatory initial syllable. Compare the following Tajio forms: sia’u 1s, sio’o 2s, siia 3s, siita 1p.in, siami 1p.ex, simiu 2p, and sisia 3p.

1052

Nikolaus P. Himmelmann

Third, despite the fact that both common and honorific articles have a strong tendency to grammaticise further, it is rather rare that the kind of gender or nounclass systems predicted by Greenberg (1978) arise. Some northern Sulawesi languages are possible exceptions. A well-documented example is Tondano, a language that appears to have innovated the system both formally and semantically (cp. Sneddon 1975: 114–117, Brickell 2014: 344–357). Formally, the ‘honorific’ series is identical to the third person pronominal proclitics and accordingly distinguishes a singular and plural form (cp. Table 4). The ‘common noun’ series is formally identical to the linking nasal (N). Consequently, the functions of marking nominal phrases (article) and linking the elements within a phrase (linking element) are not distinguished. Semantically, use of the ‘honorific’ series has been broadened to include all kinds of animates, whereas the linking nasal only marks inanimates, illustrated in the following two examples. (5)

Tondano (Minahasan) (Brickell 2014: 346) mèi kumaan sèko’ko wo sètièi èi kaan sè=ko’ko’ wo sè=tièy come eat an.pl=chicken and an.pl=pig ‘(he) comes to eat chicken and pork’

(6)

Tondano (Minahasan) (Brickell 2014: 354) empistol sikimirong7 si=kirong N=pistol 3s=conceal inan=pistol ‘he hid a pistol (in his clothes)’

Similar developments appear to have occurred in other Minahasan languages (Sneddon 1978: 172) and in Buol (Zobel 2005: 632f). The Buol and Totemboan data point to the possibility that definiteness may have played a role in the extension of the honorific noun marker to animates. In Buol, the honorific marker ti (no number distinction!) is used with personal names and definite human nouns, though not consistently with the latter. In most other Sulawesi languages, no such semantic changes have happened. What we can observe in these languages is that use of the honorific articles is more or less obligatory with personal names and respected (elder) kin and that they are often obligatorily incorporated into the independent personal pronoun series, as illustrated with the Tajio free pronoun forms above. The common-noun series, on the other hand, tends to become phonologised in the final stages of its grammaticisation, thereby not leading to semantically motivated subclasses in the nominal lexicon, as illustrated shortly.

7 The source has sikimiring, but this appears to be a mistake.

Grammaticisation processes and reanalyses in Sulawesi languages

1053

In terms of function, the article-like elements in Sulawesi languages, and in western Austronesian more generally, do not mark definiteness, with the exception of the honorific noun-markers used for definite reference to humans in a few northern Sulawesi languages, as noted previously. The only clear semantic contrast manifest throughout the area is between honorific and common-noun phrases. Otherwise, the use of these elements appears to be largely determined by the grammatical construction. Given that they are usually blocked from occurring in prepositional phrases (cp. example [4] above), article-like elements commonly occur in phrases with core argument functions (subject, object), in topic expressions, in possessor phrases and in nominal predicates. Among these contexts, there is usually only a real choice with regard to the nominal predicate function. The use of the article here correlates with the difference between a referential and a non-referential reading, as illustrated by the following example from Tagalog: (7)

Tagalog a. ma-lakí ang kwarto-ng iyon. st-bigness spec room-lk dist ‘That room is big.’ b. ang ma-lakí ang kwarto-ng iyon. spec st-bigness spec room-lk dist ‘That room is the big one.’

Otherwise, use of the common-noun phrase markers is either obligatory for nonprepositional arguments and topics, or it is semantically rather elusive. In Tajio, for example, the common-noun article te is optional in subject and object position for all consonant-initial bases. Compare: (8)

Tajio (Tomini) (Mayani 2013) (te=)saping neng-inang (te=)gugus art=cow av.rls-eat art=grass ‘Cows feed on grass./The cows feed on (the) grass.’

In Tajio spontaneous discourse, use of the article appears to be the default case, and reasons why it is occasionally dropped with consonant-initial stems are poorly understood. But it would appear very likely that the functional load with regard to signalling distinctions of accessibility (given, new) is rather low or absent. This is further supported by the fact that the article is obligatory for vowel-initial bases occurring in subject or object functions (in example [4], it must be te=asu; asu on its own would be ungrammatical). Note that all languages in the area typically have alternative constructions for introducing new referents into the discourse world and for non-referential men-

1054

Nikolaus P. Himmelmann

tions. These constructions make use of an existential quantifier, as illustrated by the following Tajio example. (9)

Tajio (Tomini) (Mayani 2013) amai te=alaiong i ndaang nu=ayu exist art=owl loc branch gen=wood ‘There was an owl on the branch of the tree.’

Many Sulawesi languages, especially in the southern half of the island, show advanced stages in the grammaticisation of the common-noun article. The usage conditions for these advanced stages are partly syntactic, partly phonotactic, as van den Berg (2012) adeptly illustrated in his survey of ‘elusive articles’ in Sulawesi. In Tolaki, for example, the article is used with disyllabic bases only: (10) Tolaki (Bungku-Tolaki) (van den Berg 2012: 218) o=piso o=donga *o=kaluku *o=laika8 ‘the/a knife’ ‘the/a deer’ ‘the/a coconut’ ‘the/a house’ In Balantak, the article only occurs with subject arguments, provided that these are in post-verbal position. (11) Balantak (Saluan-Banggai) (van den Berg 2012: 210) a. Ma-polos tuu' a sengke'-ku. intr-hurt very art back-1s.poss b. Sengke'-ku ma-polos tuu'. back-1s.poss intr-hurt very ‘My back really hurts.’ Tukang Besi has a similar restriction in that only one core argument can be marked with the article na= in postverbal position. But here all other core arguments are marked with another article, te=, including all arguments in preverbal position, regardless of their syntactic function. (12) Tukang Besi (Donohue 2009: 771) a. No-'ita te kadadi na wowine. 3s.rls-see art bird art woman b. Te wowine no-'ita te kadadi. art woman 3s.rls-see art bird ‘The woman watched birds.’

8 In Sulawesi, vowel sequences are rarely diphthongised. This is a trisyllabic root.

1055

Grammaticisation processes and reanalyses in Sulawesi languages

Tab. 6: Comparison of the article in Balantak, Muna and Tolaki (van den Berg 2012: 223). Balantak a

Muna o

Tolaki o=

semantic

independent meaning

no

no

no

syntax

Category

common nouns

common nouns

common nouns

grammatical function of NP

subjects only

any

any

position in clause

post-predicate

a. clause-initial b. free-standing

any

with possessive suffixes

yes

no

no

with bound numerals

yes

no

no

after prepositions

no

no

no

shape of noun root

any

any

disyllabic only

clitic position

enclitic

proclitic

proclitic

position in intonation phrase

medial

only initial

anywhere

prosody

Table 6 from van den Berg (2012) provides further details on the factors determining the use of articles in three Sulawesi languages. These clearly show that the articles in question do not convey semantic functions and that even though they are restricted to common nouns, it is not likely that they are developing into common noun markers as predicted in Greenberg (1978). To complete this survey on article developments in Sulawesi, a major innovation has occurred in some of the South Sulawesi languages, in particular Makassarese and Bugis. These languages show phrase-final articles in common-noun phrases. These are obligatory for definite NPs, but can also be used together with the numeral ‘one’ to express ‘a certain X’, and as nominalisers (Jukes 2006: 153). For personal names, the ‘standard’ honorific i is used, in the form of a prefix or proclitic depending on the language and the analysis. (13) Makassarese (South Sulawesi) (Jukes 2006: 335) Bambang=i allo-a. hot=3 day-def ‘The day is hot.’ (14) Makassarese (South Sulawesi) (Jukes 2006: 152) Ku=kanre=i unti-a. 1=eat=3 banana-def ‘I eat the bananas.’

1056

Nikolaus P. Himmelmann

Jukes (2006: 151) suggests that it “seems likely that there is a historical connection” between the definite marker -a and the 3rd person pronoun ia (used for singular and plural). There are articles following their heads also in other western Austronesian languages outside of Sulawesi, which typically derive from the third person possessive pronoun and are often in fact formally indistinguishable from them (i.e., they present cases of incipient grammaticisation). Himmelmann (1997: 219–221, 2001a: 839) calls these possessive articles and speculates that their grammaticisation follows a somewhat different trajectory than the trajectory of demonstrative pronouns turning into articles. Specifically, he proposes that they arise from an extension of what is widely referred to as associative anaphora in the definiteness literature, where definite reference is based on the fact that the newly introduced referent belongs to the frame of a preceding referent (mentioning a house allows for definite reference to its door(s) and window(s), for example). The following example from Malay illustrates such an extended use, which, however, can probably still be accounted for in terms of the semantics of possessive pronouns: (15) Malay air-nya water-art? (room service knocking at the door:) the (drinking) water (you asked for/that is found as standard in a hotel room) Balinese provides an example of a more strongly grammaticised possessive article. In Balinese, anaphoric reference requires this article, which historically derives from the cognate of Malay -nya. Compare: (16) Balinese =(n)e (Mayani, p.c.) Tiang ngelah kuluk ajak meong. Kuluk=e putih meong=e selem. 1s av:own dog and cat dog=art white cat=art black ‘I have a dog and a cat. The dog is white, the cat is black.’ The possessive article may occur in all syntactic functions, including prepositional phrases. In fact, there is a tendency–the details of which are not yet well understood–to use it in prepositional phrases even in those instances where it is not clear what the referential anchor for a definite mention could be, as in the following example: (17) Balinese (Clynes 1995: 241) montor-ne tomplok-ang=a ke tembok=e car-3s.poss collide-appl=3s to wall=art ‘He crashed his car into a wall.’

Grammaticisation processes and reanalyses in Sulawesi languages

1057

3 Grammaticisation of verbal categories Quite a few verbal categories receive clitic or affixal expression in Sulawesi languages. However, it is here that the problems in identifying grammaticisation paths in western Austronesian mentioned in Section 1.2 are most noticeable. Following the pioneering work of Wolff (1973), it has been widely assumed that Proto-Austronesian clause structure and verb morphology looked very much like current MesoPhilippine languages. Ross (1995, 2009) has shifted this focus somewhat towards Formosan languages, and in particular Puyuma, but in essence the reconstruction still very much resembles Philippine-type languages. Importantly, there are no serious speculations as to the origins of the roughly twenty affixes that are generally reconstructed at the proto level. Blust, in his detailed survey of Proto Austronesian affixes (Blust 2013, Chapter 6), provides hypotheses on likely meanings and functions of these affixes, but in no instance does he mention possible origins. Furthermore, the suggested meanings and functions often do not differ significantly from the ones attested in the current daughter languages (exceptions to be discussed below). Hence, not much development can be observed, other than the loss of many (or sometimes all) of these affixes. The lone dissenting voice of Lemaréchal (2010), who argues for a scenario in which South Sulawesi languages such as Bugis and Makassarese are deemed to be representative of the oldest stages of the proto-language, does not fundamentally change this overall picture. That is, Lemaréchal works with essentially the same set of reconstructed affixes (sometimes adding additional morphological structure). The major difference to the ‘standard’ view is that he reconstructs a complete set of person markers of the kind illustrated in Table 2 at the highest proto level. Furthermore, he postulates in part different meanings for the reconstructed affixes and offers a few not very well-supported speculations as to the origin of some of them (mostly verbs meaning ‘take, put’, [cp. Lemaréchal 2010: 287–306]).9 The following exposition is largely framed within the widely shared hypothesis that the general direction of historical developments affecting Austronesian affixes

9 Lemaréchal’s ideas also boldly contradict general assumptions regarding phonological and lexical developments, thus leading him to fundamentally question the widely shared out-of-Taiwan hypothesis for the early stages of the spread of Austronesian. However, even if one agrees with this hypothesis and the bulk of the proposed lexical and phonological reconstructions, this does not necessarily mean that Formosan and Philippine languages are the most conservative with regard to the morphosyntactic level, as appears to be widely believed in the relevant literature (most reconstructions, perhaps inadvertently, imply that the structures of most Philippine and some of the Formosan languages have hardly changed over the last 5000–6000 years). To my mind, unless and until it can be shown that the morphosyntactic systems attested throughout the Austronesian family can be properly accounted for in terms of sequences of plausible changes of the reconstructed system (and possibly contact interferences), the jury on the main direction of change is still out. See Foley (2014) for a fuller assessment of Lemaréchal’s proposals.

1058

Nikolaus P. Himmelmann

is from close to Philippine-type to Sulawesi and beyond (i.e., inasmuch as Sulawesi languages diverge from Philippine patterns, they represent innovations). However, it may be useful to keep in mind that the reverse direction has also been proposed and that it is, in fact, not straightforward to decide between competing views at our current stage of knowledge.

3.1 Voice/valency As much of the currently attested voice, and more generally valency, morphology is reconstructed at the PAN level, nothing much can be said as to how this came about. As to further developments, there are only very few where the available data and analyses allow for reasonably well-supported hypotheses. One concerns the prefix ni- which occurs in example (1b) above, where it marks realis mode in undergoer voice. This affix is widely attested throughout the Austronesian world and can safely be reconstructed at the proto level (cp. Blust [2013: 385–389] for details). It is very likely that the primary shape of this affix was an infix -in-, with the prefixal allomorph ni- developing in later stages, first before bases beginning with a liquid (r, l), then becoming the most frequent allomorph in some languages such as Totoli. In terms of function, it seems very likely that it was primarily an aspect or mode marker (perfective or realis) which, importantly, was not restricted to undergoer voices, but also occurred in the actor voice, as is still the case in Ratahan (-im- < *-in-um-): (18) Ratahan (Himmelmann and Wolff 1999) te isé tumpa e then 3s jump.down cpl ‘Then he jumped down.’ In the majority of Philippine and Sulawesi languages, however, -in-/ni- is today restricted to the undergoer voices. In languages which no longer distinguish realis from non-realis voice-marked forms, -in-/ni- typically becomes a simple undergoer voice marker, as seen in the following example from Mori: (19) Mori (Esser and Mead 2011: 458) in-ala-mu ke wunta andio? uv-get-2s.gen Q book prx ‘Is this book the one (one of the ones) fetched by you?’ Esser and Mead label this construction a passive, which, strictly speaking, it is not, because the actor argument can still appear in a core argument position (as a genitive clitic/suffix, the standard in undergoer voice constructions). As the translation indicates, however, this construction has a strongly nominal character, the predi-

Grammaticisation processes and reanalyses in Sulawesi languages

1059

cate phrase being interpretable as a headless relative clause. In many south-eastern Sulawesi languages, this construction is in fact restricted to appear in relative clause constructions, as illustrated by the following example from Tolaki: (20) Tolaki (Mead 2002: 160) o gandu solongako-ro i tonga m-bada art corn pour.out-3p.gen at middle lk-field ‘the corn which had been poured out by them in the middle of the field’ Unlike in Mori, use of this formation as a main clause predicate in Tolaki is obligatorily agentless. Hence, a gloss as a passive here is entirely appropriate (the pronominal prefix no- here refers to the undergoer ‘this’): (21) Tolaki (Mead 2002: 161) Ni'ino, iamo no-in-ala. prx neg.imp 3s-pass-take ‘This must not be taken.’ In some South Sulawesi languages, ni- is used only as a passive marker in an innovative construction in which the actor argument is no longer a core argument. Rather, it is introduced by the general locative preposition ri in a construction roughly similar to English by-passives as in: (22) Makassarese (Jukes 2006: 259) Ruku' ni-kanre ri tedong grass pass-eat loc buffalo ‘Grass was eaten by the buffalo.’ The current exposition suggests that the passive and relative-clause uses of ni- illustrated in the preceding examples may be developments from the earlier realis or perfective marker -in- (possibly providing another instance of the well-known development of resultatives to passives, cp. Nedjalkov [1988], among many other contributions). However, it may very well be that the trajectory of these developments is more closely linked to the nominalising uses of ni-/-in-, which are also widely attested and reconstructible to the PAN level. The most widely attested use here is in deverbal object nouns such as Tagalog sina'ing ‘boiled rice’ from the root sa'ing ‘boil rice’. Note also denominal derivations such as Ilokano binunga ‘child’ from bunga ‘fruit’, or Tagalog in-anak ‘descendant’ from anak ‘child’ (cp. Blust 2013: 387).

3.2 Tense, aspect, mode The realis/non-realis distinction mentioned in the preceding section is closely interlinked with the voice morphology and, in those languages that have it, with person

1060

Nikolaus P. Himmelmann

marking (see next section). This distinction has also been analysed as a tense-related distinction, in which case the affix -in- discussed in the preceding section is analysed as a past or past-perfect marker. Many western Austronesian languages make regular use of reduplication to express imperfective, habitual, or iterative events. Non-reduplicated verbs, then, by default implicate a perfective state of affairs. In addition to reduplication, many Sulawesi languages mark a similar aspectual distinction by (cognates of) the enclitics =mo ‘completive (also called perfective)’ and =po ‘incompletive (or continuative)’, which are also very widespread in Philippine languages. They roughly correspond to English ‘already’ (in combination with a negator ‘no more’) and ‘still’ (in combination with a negator ‘not yet’), respectively. (23) Tajio (Mayani 2013) te=bau ni-ita=mu=mo art=fish uv.rls-see=2s.gen=cpl ‘You have already seen the fish.’ These clitics are usually not obligatory (i.e., not every verbal predication has to include them), but they are very frequent. More importantly, their function does not appear to be restricted to the aspectual domain. The clitic =mo, for example, is also very common in imperative constructions, usually for more polite variants. (24) Tajio tuut=mo sia’u follow=cpl 1s ‘Please follow me!’ (without =mo this would be a brusquer ‘follow me’; Mayani 2013) The clitic =po is often a regular part of comparative constructions: (25) Tajio (Mayani 2013) te=vonua=’u na-basag=po pa te=vonua=mu art=house=1s.gen st.rls-big=icpl than gen=house=2s.gen ‘My house is bigger than your house.’ No sources for these clitics have been suggested, but in Tajio the enclitic =po is in complementary distribution with the particle ompo ‘still’, which seems to be a likely source of the clitic. (26) Tajio siia nonggabu=po/ ompo 3s av.rls:cook=icpl / still ‘He is still cooking.’

Grammaticisation processes and reanalyses in Sulawesi languages

1061

3.3 Person marking Sulawesi languages are remarkable for their great variety of verbal person-marking systems. The major innovation is the occurrence of one or two prefix series of person markers marking either transitive actors (A in the widely used letter code for semanto-syntactic role labelling) or subjects (i.e., transitive actors and the single core argument of an intransitive verb [S/A]). As already noted with regard to Table 5 above, a major parameter of variation here is the number of prefixed person markers. In Lauje, only the first person singular is marked by a prefix, as illustrated in example (27) where second and third person singular actors are marked instead by suffixes. Furthermore, in Tomini languages such as Lauje, as well as in the neighbouring Kaili languages, the person-marking prefixes are restricted to non-realis mode. Example (28) shows that in realis mode, first person agents are also marked by suffixes. The nexus between person marking and non-realis mode will recur throughout this section and provides important evidence for the source construction(s) of these person markers. (27) Lauje (Tomini) láupe 'u-otoi / no-'ootoi-im? / no-'ootoí-ny not.yet 1s-know nrls.uv-know-2s.gen nrls.uv-know-3s.gen ‘I don’t know yet / You don’t know yet? / She doesn’t know yet.’ (28) Lauje (Tomini) 'alolongoome binee'e unga'e. 'alolong-O-Vme bee-'u unga-'u rope-epv-2s.gen give-1s.gen child-1s.gen ‘I gave your rope to my child.’ As also illustrated by these examples, the shape of the prefix series usually mirrors the suffixed series used to express actors in the undergoer voice (as in [28]), which here is glossed as gen(itive) because this series usually functions both for marking possessors in possessive constructions (cp. Lauje unga'e in [28]) and for non-subject actors in undergoer voice constructions. Such a genitive series occurs in most western Austronesian languages and it is always suffixal (or enclitic, the difference often being difficult to diagnose).10 Hence, it seems very likely that the preposed person markers found in many Sulawesi languages historically derive from the enclitic

10 Cp. Himmelmann (2001b: 91f) for an argument as to why the Lauje person markers in examples (27) and (28) can be analysed as suffixes and Kaufman (2010) for a thorough analysis of clitics in WMP languages, with a special focus on Tagalog and Sulawesi languages.

1062

Nikolaus P. Himmelmann

Tab. 7: Pronominal clitics in Makassarese (South Sulawesi, Jukes 2006: 143f).11

1 2 familiar lp.in/2 polite 3

Proclitics

Enclitics

Possessive

ku= nu= ki= na=

=a' =ko =ki' =i

=ku =nu =ta =na

genitive series, as suggested by Jonker (1911),12 Haaksma (1933) and Wolff (1996), among others. We will return to this issue below. First, however, the variation attested in person-marking systems in Sulawesi languages is illustrated further. South Sulawesi person-marking systems show a major innovation in that the former pronominal clitics/affixes have become proper agreement (or cross-reference) markers. That is, they may co-occur with co-referential lexical NPs, but they do not have to be accompanied by them. Furthermore, South Sulawesi languages show both a proclitic and an enclitic series, as illustrated with Makassarese pronouns in Table 7. Their pronoun paradigms are somewhat reduced compared to other western Austronesian languages as they do not systematically distinguish singular and plural forms. The proclitic series is again clearly related to the genitive (possessive) enclitics (same shape except for the possessive first person inclusive =ta). The enclitic series is related to the subject form of the pronouns, which often allows for both clitic and free uses in most Philippine and a few Sulawesi languages (the latter often only allow free form pronouns in subject function). Examples (29) and (30) show that both the proclitic and the enclitic series may cross-refer to full NPs in the same clause. The enclitics cross-reference the single core argument of intransitive predicates (S) and the undergoer argument of transitive predicates (P). The proclitics cross-reference actors of transitive arguments. (30) also shows that no strict ordering rules apply to lexical arguments. (29) Makassarese (South Sulawesi) (Jukes 2006: 333) Tinro=i i Ali. sleep=3 hon Ali ‘Ali is sleeping.’

11 The rarely used archaic form =kang for first plural exclusive has been omitted. Jukes (2006) analyses the possessive series as ‘affixal clitics’ in order to account for the fact that these markers show evidence for both clitic and affixal status. 12 Jonker (1911) reconstructs the prefix series to the Proto-Malayo-Polynesian level, in this regard resembling the approach developed in Lemaréchal (2010). If this is assumed to be correct, the phenomena reviewed here would need to be accounted for in terms of restructuring and loss.

Grammaticisation processes and reanalyses in Sulawesi languages

1063

(30) Makassarese (South Sulawesi) (Jukes 2006: 340) Na=cinik=i tedong=ku i Ali. 3=see=3 buffalo=1.poss hon Ali ‘Ali sees my buffalo / my buffalo sees Ali.’ On first sight, the distribution of the pro- and enclitic series may look like a straightforward ergative pattern, with transitive actors (A) being cross-referenced by proclitics, and transitive undergoers (P) and single core arguments of intransitive predicates (S) by enclitics. However, this is not quite as straightforward as it appears, since in negated sentences the single core argument of an intransitive predicate is marked by the supposedly ergative proclitic, as seen in the following example: (31) Makassarese (South Sulawesi) (Jukes 2006: 320) Tena na=tinro. neg 3=sleep ‘She doesn’t/didn’t sleep.’ In transitive clauses, there are two possibilities. Either the negated clause pattern is identical to the non-negated one (cp. [32]), or both actor and undergoer are represented by the ‘ergative’ proclitics, as in (33). (32) Makassarese (South Sulawesi) (Jukes 2006: 322) Ku=buntul=uk=ki'. Tena (na) ku=buntul=uk=ki'. 1=meet=epv=2.pol neg comp 1=meet=epv=2.pol ‘I met you.’ I didn’t meet you.’ (33) Makassarese (South Sulawesi) (Jukes 2006: 322) Tena (na) ki=ku=buntul=u'. neg comp 2.pol=1=meet=epv ‘I didn’t meet you.’ In both instances, the negator may optionally be followed by the complementiser na. This suggests that, at least historically speaking, we are dealing here with a kind of complement construction in which a matrix predicate (the negator in the preceding examples) governs an embedded predicate. In such a construction, what now appear to be proclitics may also be realised as enclitics on a preceding constituent. In fact, some speakers actually allow just that, i.e., an enclitic realisation of the proclitics: (34) Makassarese (South Sulawesi) (Jukes 2006: 324) ri allo=nna arabaia tette' sampulo‒asse're na=ku … a'-lampa loc day=3.poss Wednesday o’clock ten-lk.one comp=1 intr-go ammekang ri Bulukumba. tr:hook loc Bulukumba ‘On Wednesday at 11 o’clock I went fishing at Bulukumba.’

1064

Nikolaus P. Himmelmann

Tab. 8: Verbal agreement in Tukang Besi (Donohue 1999). Position:

Pre-root

Post-root

Independent

Role:

S, A

O

(any)

(mode):

Realis

Irrealis

1s 2s 3(sg, pl) 1 paucal 1 plural 2p

kuˈuno- / okotoi-

kukona- / akataki-

=aku =ko =ˈe =kami =kita =komini

iaku ikoʻo ia, amai ikami ikita ikomiu

We will return to possible scenarios for the historical development below. First, another major parameter of variation needs to be added, primarily attested in southeastern Sulawesi. Here, it is typical to find two series of proclitics/prefixes, as illustrated for Tukang Besi in Table 8. The two series mark the modal distinction characteristic of symmetrical voice systems; that is, realis vs. non-realis eventualities (example [35]). They are usually analysed as agreement markers, meaning they can co-occur with co-referential full NPs, which, however, are not grammatically required (example [36]). Finally, the prefixes cross-reference the single core argument of intransitive clauses and the actor argument of transitive clauses (example [37]) (i.e., they clearly show nominative alignment). (35) Tukang Besi (Donohue 1999: 153) no-baiara-'e na-baiara-'e 3s.rls-pay-3s 3s.nrls-pay-3s ‘She has paid it.’ ‘She is going to pay it.’ (36) Tukang Besi (Donohue 1999: 164) no-tapa-'e na ana te wowine 3s.rls-slap-3s art child art woman ‘The woman slapped the child.’ (37) Tukang Besi (Donohue 1999: 152) Jari no-'eka di wunua=no. so 3s.rls-climb loc house=3.poss ‘So she went up to her house.’ A full account of the historical development of the different person-marking systems reviewed above must provide for the different alignment options, the different number of paradigms (up to two prefix series and one suffix series), and the propensity

Grammaticisation processes and reanalyses in Sulawesi languages

1065

of one prefix series to mark non-realis mode, if modal distinctions are marked at all on the verb. While many details are still missing for such a comprehensive account, it seems very likely that the prefixal series has its origins in a construction with the following features: an auxiliary-like matrix predicate followed by a subordinate predicate usually in a special subjunctive mode. The clitic pronoun or pronouns specifying arguments of the subordinate predicate are second-position clitics and thus occur in between the matrix and subordinate predicates. This construction type is widely attested in Philippine languages. The following example is from Cebuano and illustrates the regular negation construction, in which the negator waʔ functions as the matrix predicate governing the subjunctive form of the embedded predicate: (38) Cebuano (Zorc 1977: 151) waʔ niya saky-í ang taksi neg.past 3s.gen ride.on-lv.sbj spec taxi ‘He did not ride in the taxi.’ In this kind of construction, the actor clitic precedes its predicate in a non-realis construction. However, more steps are required to get from here to the kind of construction illustrated from Lauje in (27). Most importantly, the genitive actor enclitic (niya in [38]) has to become a proclitic and thus loses its status as a second position clitic. To the best of my knowledge, the factors and steps involved here are not yet well understood. Furthermore, in order to get to the state of affairs illustrated by Makassarese, one needs a construction that allows for the co-occurrence of a pronominal clitic/ affix and a coreferential, phonologically independent expression (full noun or pronoun). Constructions that instantiate such a scenario are occasionally attested. The following example illustrates the standard construction in Tajio for expressing volitive modality. Here again, an auxiliary-like element (seelu) functions as matrix predicate: (39) Tajio (Tomini) (Mayani 2013) sia’u seelu=’u mom-bava te=paku boi kua=’u mom=bava 1s want=1s av.nrls-take art=nail but not.want=1s av.nrls-take te=gola art =sugar ‘I wanted to take the nail but I did not want to take the sugar.’ The two constructions in (38) and (39) are not identical with regard to the grammatical status of the actor clitic. In (38) we are dealing with a second-position clitic (in Philippine languages, clitics usually occur in second position regardless of the function of the first constituent of the clause), while in (39) we are dealing with a possessive clitic on an auxiliary-like element (there are no second-position clitics in Tajio). Hence, there are two possible source constructions for the prefixal person markers

1066

Nikolaus P. Himmelmann

we have reviewed in this chapter, which can be schematically represented as follows: (40) Possible source constructions for Sulawesi person markers a. [(XPi) [predicate/aux=pron.cliticgeni [verbsbj YP]]] → [proclitici=verb XPi YP] b. [(XPi) [aux=2nd position clitici verbsbj YP]] → [proclitici=verb XPi YP] In fact, it is very well possible that both these constructions played a role in the emergence of the proclitic person markers. The great diversity of person-marking systems suggests a scenario in which a number of independent parallel developments occurred across the different Sulawesi subgroups. The probably most complex outcome of such developments is found in the Bungku-Tolaki subgroup, where a language such as Padoe shows two enclitic series alongside one proclitic series plus a phonologically free form immediately preceding the verb, which is used for non-realis and future events (see Mead [2002: 161–167] for details). The following examples illustrate one of the two enclitic series (used for transitive actors and intransitive subjects), the proclitic series (also used for arguments in S/A function), and the ‘future’ pronoun series (on this cp. also Esser and Mead [2011: 114 passim]). Note that in example (43) it is primarily the form of the pronoun that indicates future time reference. (41) Padoe (Bungku-Tolaki) (Mead 2002: 162) Mo-nahu-aku-to inehu. tr1-cook-1s-perf vegetables ‘I cooked vegetables.’ (42) Padoe (Bungku-Tolaki) (Mead 2002: 163) Inehu mbio au-po-nahu. vegetable what 2s-tr1-cook ‘What vegetables are you cooking?’ (43) Padoe (Bungku-Tolaki) (Mead 2002: 164) Lo’iro mo-nahu inehu. 3p tr1-cook vegetables ‘They will cook vegetables.’

3.4 Applicatives Two suffixes that are widely attested as integral parts of Philippine-type symmetrical voice systems, -an and -i, have clearly applicative functions in many of those Sulawesi languages in which the voice systems have been reanalysed and reorga-

Grammaticisation processes and reanalyses in Sulawesi languages

1067

nised. A particularly complex case, which possibly shows a system in transition, is found in the Central Sulawesi language Totoli, discussed in detail in Himmelmann and Riesberg (2013). The following two Totoli examples illustrate typical uses of the two applicatives widely attested in Sulawesi, one benefactive/instrumental and the other goal/recipient (cp. also Donohue 2001). (44) Totoli (Himmelmann and Riesberg 2013: 401) i Rinto manaipan aku taipang i Rinto mon-taip-an aku taipang hon Rinto av-peel-appl1 1s mango ‘Rinto peels mangos for me.’ (45) Totoli (Himmelmann and Riesberg 2013: 418) Aco anu ku-been-i hadiah rel 1s.act-give-appl2 present ‘Aco, who I will give a present to.’ However, a newly innovated suffix, probably cognate with the extant Malay preposition akan ‘about, regarding, for’, is widely attested throughout western Indonesia, including the southern half of Sulawesi. While also covering benefactive and instrumental applicative functions, this suffix usually covers a very broad range of functions and is often simply characterised as ‘transitivising’. Malay -kan, for example, is used for causative derivations from adjectives and nouns, as well as instrumentals and benefactive applicatives and a host of other functions. A particularly complex example from Sulawesi is Mori -ako discussed in detail by Esser and Mead (2011: 482–503, see also Mead 2005: 702–705). It is likely that this applicative suffix arose through a process called ‘preposition capture’ in the Austronesianist literature, referring to the reanalysis of a preposition first as a postverbal clitic and then a suffix (cp. Sirk [1996] for a discussion relating to Sulawesi languages). Adelaar (2011) convincingly argues that current attestations of this suffix may reflect contact-based diffusion in areas strongly dominated by Malay, rather than reflecting the common heritage of a single proto-language.

3.5 Directionals A widespread feature in northern Sulawesi languages is the occurrence of directional particles, which tend to be part of almost every verbal construction (cp., for example, Himmelmann and Wolff 1999: 72–82; Brickell 2014: 258–264; Lobel 2015: 423). The indicated directionality is often transparent, as in the following Ratahan example:

1068

Nikolaus P. Himmelmann

(46) Ratahan (Himmelmann and Wolff 1999: 81) nang-ule mai su Watulinei av.pst-return ven loc Watulinei ‘(The ones who) returned from over in Watulinei.’ In other cases, it is less clear how the reported eventuality involves directionality. In the following example, the directional possibly indicates that the crying of the baby in the kitchen can be heard by speaker and hearer sitting next door. (47) Totoli ntonggiismai Bibi dei abu no-tonggiis=mo=ai Bibi dei abu av.rls-scream=cpl=ven loc kitchen ‘Bibi is crying in the kitchen’ The directionals may be used quite subtly to keep track of who is doing something to whom in narratives where participants are not explicitly mentioned for extended sequences of predicates, despite the fact that it is often not quite clear who is actor and who undergoer. (48) Totoli no-kudut ni-kket-an=na=ko ita-i sia st.rls-break rls-laugh-uv2=3s.gen =and see-appl2 3s no-keket=ai av.rls--laugh=ven ‘(she said the sandals are) broken, she laughed at (him), seeing him, she laughed (was shaken by laughter/laughter came to her)’ The directionals tend to be weakly grammaticised as they may often still be used as main predicates in some languages: (49) Ratahan (Himmelmann and Wolff 1999: 75) roku um-intu-intu atau roku mai nu apa neg.imp av-rdp-descend or neg.imp come gen where ‘(If the soldiers come) don’t go down, don’t go anywhere.’ That is, the directionals are often transparently related to words meaning ‘come’, ‘go’, ‘climb’, ‘cross’ or ‘descend’. While usable as main-clause predicates, these words tend to be defective verbs in that they do not show the morphology typical for ‘proper’ verbs (in the languages of the northern half of Sulawesi, this would be voice morphology in particular).

Grammaticisation processes and reanalyses in Sulawesi languages

1069

4 Grammaticisation of complex constructions The clearest example of subordinate clause constructions in many Sulwesi languages are relative clauses. These quite often involve an initial relative clause marker deriving from the word for ‘person’, as for example Tajio to (< too ‘person’) or Balantak men (< mian ‘person’). (50) Balantak (van den Berg and Busenitz 2012: 222) mian [men mang-asok rombia'] … person rel av.irr-plant sago.tree ‘People who plant sago trees …’ Alternatively, the relative-clause marker is derived from, or identical to, the word for ‘(some)thing’, as with Totoli anu (illustrated in [45] above). However, other strategies, including simple juxtaposition, are also possible. In south-eastern Sulawesi, use of a ‘participial’ verb form involving the affixes -um- or -in-/ni- is widespread, cp. the Tolaki example in (20) (see also van den Berg 1989: 231 ff; Donohue 1999: 367–387). In Makassarese, relative clauses are marked by adding the definite enclitic =a to the predicate of the relative clause, which otherwise is not distinguished from a main clause (Jukes 2006: 238–242). Alternatively, constructions that tend to involve subordinate clause constructions, including adverbial or complement clauses, are in many languages often expressed by simple juxtaposition or by the use of nominalised verb forms. Subordinating conjunctions are widely borrowed from Malay, but the clauses thus introduced typically do not show any other signs of subordination.

5 Comparative outlook Many of the developments discussed in the preceding sections are also found outside of Sulawesi and thus can, to some degree, be deemed representative for the larger Austronesian family. There are certainly other developments not discussed here, in part simply because there is still very little known about grammaticisation phenomena proper in the western (non-Oceanic) half of the family, as mentioned in the introduction. This section is limited to discussing a single development that has occurred outside Sulawesi, as it would appear to be highly unusual also from a wider cross-linguistic perspective. A quite unusual grammaticisation path has been proposed for the passive prefix di- found in many Malayic varieties, which has given rise to a number of different hypotheses conveniently summarised in Adelaar (2005, 2009). The following example illustrates the active/passive alternation marked by this prefix. Note that this is a proper passive alternation where the passive counterpart of the transitive active

1070

Nikolaus P. Himmelmann

construction is intransitive, the actor occurring in a prepositional phrase (but see Riesberg 2014b). This differs from the symmetrical voice alternations discussed in earlier sections where both voices involve transitive constructions. (51) Malay a. Anak saya me-lihat orang itu. child 1s av-see person dist ‘My child saw that person.’ b. Orang itu di-lihat oleh anak saya. person dist pass-see by 1s 1s ‘That person was seen by my child.’ Adelaar (2005) argues that this prefix derives from the highly general, originally probably locative preposition di. The proposed source construction involves a semantically transitive verb immediately preceded by an actor phrase marked by the preposition di, as seen in the Salako example (52a). As (52b) shows, the noun complement of the preposition can actually be omitted in Salako, the ‘preposition’ then becoming a proclitic to the verb. (52) Salako (Kendayan) (Adelaar 2005: 129) a. Bini-e dah mati di kayo munuh. wife-3.poss perf dead by enemy rls:kill ‘His wife was killed by the enemy.’ b. Bini-e dah mati di=munuh. wife-3.poss perf dead pass=rls:kill ‘His wife was killed.’ Adelaar (2009) shows that an essentially similar development may have happened in Javanese, where the modern passive prefix di- ultimately derives from a noun de (< *day/*dai) ‘cause, reason, action, way, manner’. Importantly, the verb form munuh in the above examples13 is undergoer-oriented in the Salako source construction (52a). Thus, the passive meaning of the overall construction has its origins in the overall construction including, in particular, the verb form, not solely in the preposition. Hence, it would be misleading to declare this an instance of a preposition being the source for a passive construction, as the

13 Adelaar (2005: 129) observes with regard to this form: ‘note that in this language nasalisation also applies in passive constructions denoting a completed action’. Strictly speaking, the form munuh is multiply ambiguous, as it could be the realis form of the verb bunuh in either actor or undergoer voice, or the non-realis form in actor voice. This multiple ambiguity may have been a factor in the reanalysis of di as a marker for undergoer voice.

Grammaticisation processes and reanalyses in Sulawesi languages

1071

gram-based grammaticalisation literature would have it. It would not make sense to postulate a grammaticisation path from locative to passive meanings similar to the kind of paths postulated by Bybee, Perkins, and Pagliuca (1994) for the tenseaspect-mode domain.

6 Conclusion The current chapter has reviewed a number of grammaticisation phenomena found in Sulawesi languages which in many ways can be deemed to be representative of grammaticisations widely attested in (western) Austronesian languages. Most of the developments are, in principle, well-known from the general grammaticisation literature. To wit: – numeral classifiers from words for ‘person’, ‘branch’, ‘fruit’ etc. (Section 2.1) – associative plurals from plural pronouns (Section 2.2) – articles from demonstratives and third person possessive pronouns (Section 2.4) – the development of passive markers from an originally perfective- or realismarking infix (Section 3.1) – imperfectives marked by reduplication (Section 3.2) – completive and incompletive markers from particles meaning ‘already’ and ‘still’, respectively (Section 3.2): this is a development that may be more restricted to Sulawesi and the Philippines. – verbal person markers from clitic pronouns (Section 3.3) – applicatives from prepositions (Section 3.4) – directionals from verbs meaning ‘come’, ‘go’, ‘climb’, ‘descend’ (Section 3.5) – relative-clause markers from words for ‘person’ or ‘thing’ (Section 4) Still, some of the developments are somewhat unusual from a larger cross-linguistic perspective (though not necessarily from the narrower Austronesian context). While articles derive from demonstratives, they form tightly integrated paradigms with (locative) prepositional elements, resulting in cross-linguistically highly unusual paradigms of noun-phrase markers (Section 3.3). Furthermore, the late stages in the grammaticisation of articles involve phonological factors to date not widely noted in the literature (also Section 3.3). Similarly, the grammaticisation of personal pronouns from free forms and second position clitics to verbal affixes/clitics, while widely attested in the languages of the world, leads to cross-linguistically unusual results due to the specific context in which it appears to have taken place. Proclitic and prefixed person markers in Sulawesi languages – and more generally in the Austronesian world – are not simply pronominal copies of topicalised constituents as proposed in Givón (1976). Rather, they arise in the context of subordinating constructions conveying nonrealis meanings.

1072

Nikolaus P. Himmelmann

A final, somewhat unusual development that has occurred outside of Sulawesi in Malayic and other languages in western Indonesian is the development of a preposition to a passive marker, as briefly discussed in section 5.

Acknowledgements I am grateful to an anonymous reviewer and Andrej Malchukov as well as Kurt Malcher-Moreno and Yuko Kitada for helpful comments on an earlier version of this chapter. Special thanks to Katherine Walker for thoroughly checking English grammar and style as well as examples and references.

Abbreviations 1 = first person, 2 = second person, 3 = third person, act = actor, an = animate, and = andative (directional), appl = applicative, art = article, av = actor voice, cmp = Central Malayo-Polynesian, clf = classifier, comp = complementiser, cpl = completive, dat = dative, def = definite article, dist = distal, epv = epenthetic vowel, ex = exclusive, exist = existential, gen = genitive, hon = honorific (article), icpl = incompletive, imp = imperative, in = inclusive, inan = inanimate, intr = intransitive, irr = irrealis, lk = linker/linking nasal, loc = locative, lv = locative voice, mot = motion by actor, n = nasal prefix, neg = negation, nrls = non-realis, p / pl = plural, pass = passive, past = past, perf = perfective, pol = polite, poss = possessive, prx = proximal, rdp = reduplication, rel = relative, rls = realis, s / sg = singular, sbj = subjunctive, sf = stem-forming affix, shwng = South Halmahera-West New Guinea, spec = specific article, st = stative, tr = transitive prefix, uv = undergoer voice, ven = venitive (directional), wmp = Western Malayo-Polynesian

References Adelaar, K. Alexander. 2005. Much ado about di-. Bijdragen tot de Taal-, Land- en Volkenkunde. 161(1). 127–142. Adelaar, K. Alexander. 2009. The various origins of the passive prefix di-. In K. Alexander Adelaar & Andrew Pawley (eds.), Austronesian historical linguistics and culture history: A festschrift for Robert Blust, 129–142. Canberra: Pacific Linguistics. Adelaar, K. Alexander. 2011. Javanese -aké and -akən: A Short History. Oceanic Linguistics 50(2). 338–350. Berg, René van den. 1989. A grammar of the Muna language. Dordrecht: Providence. Berg, René van den. 1996. The demise of focus and the spread of conjugated verbs in Sulawesi. In Steinhauer (ed.), Papers in Austronesian linguistics No. 3 (Pacific Linguistics, Series B, 84), 89–114. Canberra: Pacific Linguistics. Berg, René van den. 2012. Elusive articles in Sulawesi: between syntax and prosody. In Niclas Burenhult, Arthur Holmer, Anastasia Karlsson, Håkan Lundström & Jan-Olof Svantesson (eds.), Language documentation and description, vol 10: Special issue on humanities of the

Grammaticisation processes and reanalyses in Sulawesi languages

1073

lesser-known: New directions in the description, documentation and typology of endangered languages and musics, 208–227. London: SOAS. [http://www.elpublishing.org/PID/119]. Berg, René van den & Robert L. Busenitz. 2012. A grammar of Balantak a language of Eastern Sulawesi. SIL e-Book 40. Blust, Robert A. 1991. The Greater Central Philippines Hypothesis. Oceanic Linguistics 30(2). 73–129. Blust, Robert A. 2013. The Austronesian Languages. Revised Edition. Canberra: Asia-Pacific Linguistics. [http://hdl.handle.net/1885/10191]. Blust, Robert A. 2015. The case-markers of Proto-Austronesian. Oceanic Linguistics 54(2). 436–491. Brickell, Timothy C. 2014. A grammatical description of the Tondano (Toundano) language. Melbourne, Australia: La Trobe University dissertation. [http://hdl.handle.net/1959.9/ 516057]. Bybee, Joan L., Revere Perkins & William Pagliuca. 1994. The evolution of grammar. Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press. Clynes, Adrian. 1995. Topics in the phonology and morphosyntax of Balinese, based on the dialect of Singaraja, North Bali. Canberra, Australia: The Australian National University dissertation. Donohue, Mark. 1999. A grammar of Tukang Besi. Berlin: Mouton De Gruyter. Donohue, Mark. 2001. Coding choices in argument structure. Austronesian applicatives in texts. Studies in Language 25(2). 217–254. Donohue, Mark. 2009. Case in an Austronesian Language: Distinguishing Case Functions in Tukang Besi. In Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case, 770–778. New York: Oxford University Press. Esser, Samuel J. & David Mead. 2011. Phonology and morphology of Mori. SIL ebook 27. [https:// www.sil.org/resources/publications/entry/43176]. [Original: S. J. Esser. 1927–1933. Klanken Vormleer van het Morisch. Leiden: Vros (1927) (Part 1) and Bandoeng: Nix (1933) (Part 2). Translated and updated by D. Mead]. Foley, William A. 2014. Comparative grammar and typology: Essays on the historical grammar of the Austronesian languages by Alain Lemaréchal (review). Oceanic Linguistics 53(2). 517–523. Givón, Talmy. 1976. Topic, pronoun, and grammatical agreement. In Charles N. Li (ed.), Subject and topic, 149–185. New York: Academic Press. Greenberg, Joseph H. 1978. How does a language acquire gender markers? In Joseph H. Greenberg, Charles A. Ferguson & Edith Moravcsik (eds.), Universals of human language, vol. III, 47–82. Stanford: Stanford University Press. Haaksma, Rémy. 1933. Inleiding tot de Studie der Vervoegde Vormen in de Indonesische Talen. Leiden: Brill. Himmelmann, Nikolaus P. 1996. Person marking and grammatical relations in Sulawesi. In Steinhauer (ed.), Papers in Austronesian linguistics No. 3 (Pacific Linguistics, Series B, 84), 115–136. Himmelmann, Nikolaus P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Himmelmann, Nikolaus P. 1998. Regularity in irregularity: Article use in adpositional phrases. Linguistic Typology 2(3). 315–353. Himmelmann, Nikolaus P. 2001a. Articles. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Language typology and language universals, 831– 841. Berlin: de Gruyter. Himmelmann, Nikolaus P. 2001b. Sourcebook on Tomini-Tolitoli languages. General information and word lists. Canberra: Pacific Linguistics. Himmelmann, Nikolaus P. 2015. Notes on ‘noun phrase structure’ in Tagalog. In Jens Fleischhauer, Anja Latrouite & Rainer Osswald (eds.), Explorations of the syntax-semantics interface, 315–337. Düsseldorf: dup. Himmelmann, Nikolaus P. & Sonja Riesberg. 2013. Symmetrical voice and applicative alternations: Evidence from Totoli. Oceanic Linguistics 52(2). 396–422.

1074

Nikolaus P. Himmelmann

Himmelmann, Nikolaus P. & John U. Wolff. 1999. Toratán (Ratahan). München: Lincom. Jonker, J. C. G. 1911. Over de ‘Vervoegde’ Werkwoordsvormen in de Maleisch-Polynesische Talen. [On the ‘conjugated’ verb forms in the Malayo-Polynesian languages.] Bijdragen tot de Taal-, Land- en Volkenkunde 65(1). 266–333. Jukes, Anthony. 2006. Makasar. Melbourne, Australia: University of Melbourne dissertation. [http://www.oxis.org/theses/jukes-2006.pdf] Kaufman, Daniel A. 2010. The morphosyntax of Tagalog clitics: A typologically driven approach. New York, USA: Cornell University dissertation. Lemaréchal, Alain. 2010. Comparative grammar and typology: Essays on the historical grammar of the Austronesian languages (Orbis Supplementa 35). Louvain: Peeters. Leto, Claudia, Winarno S. Alamudi, Nikolaus P. Himmelmann, Jani Kuhnt-Saptodewo, Sonja Riesberg & Hasan Basri. 2005–2010. DoBeS Totoli Documentation. Online: [https:// hdl.handle.net/1839/00-0000-0000-0005-24BE-F] Lichtenberk, František. 1991. Semantic change and heterosemy in grammaticalization. Language 67(3). 475–509. Lichtenberk, František. 2013a. The rise and demise of possessive classifiers in Austronesian. In Ritsuko Kikusawa & Lawrence A. Reid (eds.), Historical linguistics 2011: Selected papers from the 20th international conference on historical linguistics, 199–225. Amsterdam & Philadelphia: Benjamins. Lichtenberk, František. 2013b. Development of reason and cause markers in Oceanic. Oceanic Linguistics 52(1). 86–105. Lobel, Jason W. 2013. Philippine and North Bornean languages: Issues in description, subgrouping, and reconstruction. Honolulu, USA: University of Hawai’i dissertation. [http:// www.ling.hawaii.edu/graduate/Dissertations/JasonLobelFinal.pdf] Lobel, Jason W. 2015. Ponosakan: A dying language of Northeastern Sulawesi. Oceanic Linguistics 54(2). 396–435. Mayani, Luh Anik. 2013. A grammar of Tajio. Cologne, Germany: University of Cologne dissertation. McGregor, William B. 2008. Indexicals as sources of case markers in Australian languages. In Folke Josephson & Ingmar Söhrman (eds.), Interdependence of diachronic and synchronic analyses, 299–321. Amsterdam: Benjamins. Mead, David. 2002. Proto Celebic focus revisited. In Fay Wouk & Malcolm Ross (eds.), The history and typology of western Austronesian voice systems, 143–180. Canberra: Pacific Linguistics. Mead, David. 2003. Evidence for a Celebic supergroup. In John Lynch (ed.), Issues in Austronesian historical phonology, 115–141. Canberra: Pacific Linguistics. Mead, David. 2005. Mori Bawah. In K. Alexander Adelaar & Nikolaus P. Himmelmann (eds.), The Austronesian languages of Asia and Madagascar, 683–708. London: Routledge. Nedjalkov, Vladimir P. (ed.). 1988. Typology of resultative constructions. Amsterdam: Benjamins. Newman, John. 1984. Nasal replacement in Western Austronesian: An Overview. Philippine Journal of Linguistics 15(2). 1–17. Pater, Joe. 2001. Austronesian nasal substitution revisited. In Linda Lombardi (ed.), Segmental phonology in Optimality Theory: Constraints and representations, 159–182. Cambridge: Cambridge University Press. Reid, Lawrence A. 2002. Determiners, nouns or what? Problems in the analysis of some commonly occurring forms in Philippine languages. Oceanic Linguistics 41(2). 295–309. Riesberg, Sonja. 2014a. Symmetrical voice and linking in Western Austronesian languages (Pacific Linguistics). Boston: de Gruyter. Riesberg, Sonja. 2014b. Passive actors are not adjuncts – Consequences for the distinction between symmetrical and asymmetrical voice alternations. In I Wayan Arka & N. L. K. Mas Indrawati (eds.), Argument realisations and related constructions in Austronesian languages: Papers from 12-ICAL, Volume 2, 281–302. Canberra: Asia-Pacific Linguistics.

Grammaticisation processes and reanalyses in Sulawesi languages

1075

Ross, Malcolm D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds.), Austronesian studies relating to Taiwan (Symposium Series of the Institute of History and Philology No. 3), 727–791. Taipei: Academia Sinica. Ross, Malcolm D. 2009. Proto Austronesian verbal morphology: A reappraisal. In K. Alexander Adelaar & Andrew Pawley (eds.), Austronesian historical linguistics and culture history: A festschrift for Robert Blust, 295–326. Canberra: Pacific Linguistics. Sirk, Ülo. 1996. On the history of transitive verb suffixes in the languages of western Indonesia. In Hein Steinhauer (ed.), Papers in Austronesian linguistics No. 3 (Pacific Linguistics, Series B, 84), 191–205. Canberra: Pacific Linguistics. Sneddon, James N. 1975. Tondano phonology and grammar (Pacific Linguistics, Series B, 38). Canberra: Pacific Linguistics. Sneddon, James N. 1978. Proto-Minahasan: Phonology, morphology and wordlist (Pacific Linguistics, Series B, 54). Canberra: Pacific Linguistics. Wolff, John U. 1973. Verbal Inflection in Proto-Austronesian. In Andrew B. Gonzales (ed.), Parangal kay Cecilio Lopez, 71–91. Quezon City: Linguistic Society of the Philippines. Wolff, John U. 1996. The development of the passive verb with pronominal prefix in Western Austronesian languages. In Bernd Nothofer (ed.), Reconstruction, classification, description – festschrift in honor of Isidore Dyen, 15–40. Hamburg: Abera-Verlag. Zobel, Erik. 2005. Buol. In K. Alexander Adelaar & Nikolaus P. Himmelmann (eds.), The Austronesian Languages of Asia and Madagascar, 625–648. London: Routledge. Zorc, R. David. 1977. The Bisayan dialects of the Philippines: Subgrouping and reconstruction. Canberra: Pacific Linguistics.

William B. McGregor

26 Grammaticalization patterns in Nyulnyulan language Abstract: The Nyulnyulan languages of the Dampier Land peninsula and western Kimberley region of Western Australia form a small genetic family of about ten languages that are both genetically close and typologically quite similar. A number of grammaticalization tendencies are shared among the languages. Some of these probably began in pre-proto-Nyulnyulan times, although they have developed in somewhat different directions, and have reached different end points. They include, among others: development of indexicals into case markers, developments within the compound verb construction, and development of case markers into verbal markers of subordination and applicatives. There are a few instances of lexical items beginning on the road to grammaticalization. For instance, the nominal buru ‘place’ has begun development into a derivational morpheme in some languages, and into a relational marker in a few. Some of these grammaticalizations appear to be regionally peculiar to Nyulnyulan languages, while others are shared with nearby languages.

1 Introduction Narrog and Heine (2011: 13) aver that “relatively little information is available on grammaticalization processes … in Papua New Guinea or Australia”. This is an overly negative construal of the situation. Even in the late nineteenth century possible origins of grammatical markers were being mooted, and recent decades have seen a considerable body of research on grammaticalization in Australian languages; little of this has, however, make it into the handbooks. In this paper I attempt to partly remedy this situation by drawing together previous research on grammaticalization in Nyulnyulan languages, and situating these grammaticalization patterns within the wider context of nearby families and Australian languages generally. The Nyulnyulan languages were spoken in the Dampier Land peninsula and adjacent parts of the western Kimberley region of Western Australia – see Figure 1. They form a small non-Pama-Nyungan family of some ten languages (in the political sense) that are genetically close and typologically similar. Figure 2 shows the structure of the family based on application of the comparative method. All Nyulnyulan languages are highly endangered or moribund – none

https://doi.org/10.1515/9783110712735-012

1078

William B. McGregor

Fig. 1: Map of Nyulnyulan and nearby languages (redrawn from Stokes and McGregor 2003: 32).

has more than 50 fluent speakers, and the majority have none, though there may be a number of part speakers and/or “rememberers” who know some words.1 Nyulnyulan languages can be broadly characterised as agglutinating (like the majority of Australian languages) with some fusional characteristics. Like most nonPama-Nyungan languages, they are primarily prefixing – they have prefixes as well as suffixes (Capell 1940).

1 In this paper I will use present tense in reference to the situations in the languages at the time of their demise (for most, sometime in the twentieth century), the latest modern forms of the languages; past tense will be used for times prior to this.

Grammaticalization patterns in Nyulnyulan language

1079

Fig. 2: Family tree for Nyulnyulan languages (after McGregor 2012: 5).

Phonologically Nyulnyulan languages are unexceptional for Australian languages. They have typical inventories of phonemes, although some languages show unusual phonotactic features: Nyulnyul has a large number of words ending in consonants, while Bardi has a relatively high proportion of words beginning with vowels. case marking is carried out by means of clitics that normally attach to the end of the first word of a noun phrase, in Wackernagel’s position; there are around a dozen of these markers. Like the neighbouring languages to the east and south, Nyulnyulan languages are morphologically ergative, intransitive subjects and transitive objects (respectively the NPs kamba jarndu ‘that woman’ in (1) and milimili ‘paper’ in (2) being unmarked, while transitive subjects (kamma ‘that’ in [2]) are usually marked by the ergative postposition, a reflex of proto-Nyulnyulan (pNN) *-nima. Ergative marking is however optional in all Nyulnyulan languages, and the marker may optionally be omitted from the subject in (2) (see, e.g., McGregor 2006, 2010a). (1)

Yawuru (Hosokawa 1991: 223) malju-kaja i-nga-rn kamba jarndu laugh-intens 3nom-be-imp that woman ‘The woman is laughing.’

(2)

Yawuru (Hosokawa 1991: 251) kamma-ni i-na-kunba-nda milimili baku that-erg 3nom-cm-send-pfv paper hither ‘He sent the letter here.’

Noun classes are not distinguished, and no class of adjectives separate from nouns exists in any language. Free pronouns distinguish four persons (1, 1 & 2, 2, 3) and make a minimal (smallest possible number for the category – one for 1, 2 and 3, two for 1 & 2) vs. augmented (one or more in addition) contrast in number; in some Eastern Nyulnyu-

1080

William B. McGregor

lan (ENN) languages further number distinctions are made between unit augmented (augmented by one) and plain augmented. Bound pronouns, by contrast, mostly show an Assiniboine system: a three person system in which the first person only distinguishes three forms, a singular, a dual (covering just the speaker-hearer dyad) and a plural (including all other first person groupings) (Greenberg 1988; McGregor 1989). As is typical of languages of the region (McGregor 2002) two categories of lexical items correspond to verbs of languages such as English: preverbs (PVs), which admit little morphological modification; and inflecting verbs (IVs), which inflect for tense-mood-aspect as well as subject person and number. PVs generally comprise open classes of several hundred items; IVs form closed classes of between fifty or so (ENN) and a couple of hundred (Western Nyulnyulan – WNN). PVs usually occur in combination with an IV in a compound verb construction (CVC) (McGregor 2002). Most, if not all, inflected forms of inflecting verbs have the potential of free occurrence. Before we begin, a few general remarks are in order to contextualise the paper. First, as in most regions of the world, the written record in Australia (including Nyulnyulan) has a short history, and grammaticalization studies suffer from the lack of this important source of evidence. This is partly compensated for by the fact that some historical-comparative reconstruction has been done on the Nyulnyulan languages (Stokes and McGregor 2003; Bowern 2004, 2008). Second, the main sources of information on the languages are as follows. Comprehensive modern grammars are available for four Nyulnyulan languages, two for each of the primary subgroups: ENN (Nyikina – Stokes 1982; Yawuru – Hosokawa 1991) and WNN (Bardi – Bowern 2012; Nyulnyul – McGregor 2012). In addition, a sketch grammar is available for Warrwa (McGregor 1994); a comprehensive grammar is currently in preparation. For the other languages, information is less systematic and somewhat limited in scope and variety. A certain amount of lexical information is available for most of the languages, and some grammatical information is available on Jawi, Jukun, and Jabirrjabirr. Nekes and Worms (1953) (published as Nekes and Worms 2006) is an important additional source of information, especially on the now moribund languages. Third, to situate this description in the context of the present handbook, I highlight some features more or less characteristic of grammaticalization in Australian (including Nyulnyulan) languages: – Prosodic changes do not always accompany grammaticalization, and thus bound items deriving from separate words may retain status as full phonological words; – Recategoralization is common, though often with retention of (some or all) morphological properties of the item: that is, it is not decategorialized; – Only a small fraction of the grammatical morphemes of a typical language can be traced back to anything else with any degree of certainty;

Grammaticalization patterns in Nyulnyulan language

– –

1081

There is overall limited evidence of lexical sources for grammatical markers – evidence is generally best in the verbal domain; Metaphoric transfer plays a limited role in grammaticalization.

2 Grammaticalization of nominal categories In the nominal domain we overview some instances of grammaticalization of case markers (§ 2.1), the expression of inalienable possession (§ 2.2), and one instance of incipient grammaticalization of a nominal lexeme (§ 2.3).

2.1 case markers It is possible to reconstruct a system of eight case-marking postpositions for pNN, as shown in Table 1. Most modern languages have around a dozen postpositions; thus, new ones have developed since pNN, some by borrowing, some perhaps by processes of grammaticalization. Borrowings have to some extent resulted in the restructuring of the case systems: for instance, borrowing of the comitative postposition -nyirr from nearby non-Pama-Nyungan languages has resulted in the reflex of *-ngany com becoming an instrumental marker in WNN languages and Warrwa (McGregor 1997b). I have argued in various places (e.g., McGregor 2006, 2008, 2017) that the ergative postposition derives from an indexical item, a determiner or pronominal in prepNN. In brief, the scenario begins with the determiner as a marker of prominence in a cleft-like construction resembling The farmer det kissed the duckling ‘it was the farmer who kissed the duckling’ and The duckling det died ‘it was the duckling that died’. Agent NPs in transitive clauses are prototypically given (Du Bois 1987), and when new were made prominent by presentation in this cleft construction (McGregor 2017). New NPs in other grammatical roles were less frequently presented in this

Tab. 1: Proto-Nyulnyulan case-marking postpositions (Stokes and McGregor 2003: 46). *-nima *-kun *-ngana *-kabu *-junu *-ji *-ngany *-marru

erg loc all abl1 abl2 dat com per

1082

William B. McGregor

construction, and only when in contrastive focus. The construction thus came to be strongly associated with new Agents. Over time, and with increasing use, the focal value of the indexical item weakened and the association with the Agent role strengthened, until eventually the erstwhile determiner was reanalysed as an ergative marker, albeit optional. Consistent with this scenario, some relics of the earlier focus marking remain, e.g., in the occasional marking of subjects of intransitive clauses when in strong contrastive focus (McGregor 2007b). Some Nyulnyulan languages show a more recent and less speculative development of an indexical item into a case marker: the development of a genitive marker from the third person minimal oblique pronominal. This grammaticalization process is incomplete, and the form in question retains its indexical value. Most WNN languages lack a dat postposition – and if they have one, do not use it in expressions of NP-internal possession. Instead they use a construction involving an oblique pronoun, which is often positioned between the PR and PM NPs, as in (3) – roughly, ‘X his/her/its/their/… Y’. In this construction the PR and PM NPs can occur in either order. (3)

Jabirrjabirr, Nyulnyul (Nekes and Worms 1953: 857)2 walaŋg djen ibal djān walangk jin ibal jan spear 3min.obl father 1min.obl ‘my father’s spear’

The same construction is found in ENN languages, alongside of the more typically Australian construction involving the dat postposition, as shown by examples (4) and (5) respectively; the semantic basis of this opposition is uncertain. (4)

Nyikina (Stokes 1982: 386) karnanganyja mangarriy kinyjina emu vegetable.food 3min.obl ‘emu’s food’

(5)

Nyikina (Stokes 1982: 69) marnin-ji yila woman-dat dog ‘the woman’s dog’

There are signs that the oblique pronominals in ENN are losing their independent status in certain contexts, where they are in the process of becoming bound mor2 Examples taken from Nekes and Worms (1953) are presented in a four line format, with the original transcription in the first line and a transcription in a modern phonemic orthography in the second line.

Grammaticalization patterns in Nyulnyulan language

1083

phemes. This appears to have gone furthest in Yawuru, where jina 3min.obl has reduced to an enclitic that serves as a genitive marker on animate possessors (Hosokawa 1991: 246). Thus -jina 3min.obl occurs exclusively on the PR regardless of its position in the NP. (6)

Yawuru (Hosokawa 1991: 246) linyju-jina durrkurrang policeman-3min.obl car ‘the policeman’s car’

(7)

Yawuru (Hosokawa 1991: 246) minyaw-jina makarra cat-3min.obl tail ‘cat’s tail’

According to Hosokawa (1991: 250), -jina has almost – though not quite – achieved the status of a case-marker. Thus, it is rarely used, and only partly productive; it overlaps semantically with the dative, although the genitive -jina is largely restricted to non-specific PRs (Hosokawa 1991: 247).3 The source construction in which the pronominal began to grammaticalize into a case marker was presumably an NP internal possessive construction in which the oblique pronominal served as a copula (see further § 2.2 below). In this construction the obl pronominal may have invariably followed the PR NP; this process could have been restricted to the third person, since for the other person categories the oblique pronominal itself carries as much information as the cardinal pronoun, and full expressions involving both case forms were presumably infrequent. (See however the next subsection.) Grammaticalizations of oblique pronominals to possessive markers are not peculiar to Nyulnyulan languages. Similar processes may have occurred in Jarrakan languages, which are discontinuous with Nyulnyulan (McConvell 2003: 88), and Worrorran languages, which abut Warrwa, but are not adjacent to Yawuru (see further McGregor 2008: 306–307) – and wider afield, e.g., Dutch and Afrikaans (Burridge 1990). As far as I am aware, however, there are no parallel developments in nearby Pama-Nyungan (adjacent to Yawuru) or Bunuban languages (adjacent to Warrwa and Nyikina).

3 It is not impossible that the pNN postposition *-ji dat derives from the same source, namely the third person minimal oblique pronominal form of pre-pNN, with loss of the final syllable, and that the modern change in ENN is a repetition of an earlier pan-Nyulnyulan grammaticalization. This is consistent with the fact that no nearby languages show a similar dative form that might be a source of borrowing.

1084

William B. McGregor

2.2 Bound pronominals and inalienability For pNN it is possible to reconstruct a system of pronominal prefixes to a class of inalienably possessed nominals (Stokes and McGregor 2003; McGregor 1995). These nominals – a closed class of fewer than one hundred members – represent the most central aspects of the person, that are most defining of the person. They are mostly body part nominals, but include also a few other items such as ‘reflection’, ‘shadow’, ‘soul’ and the like. McGregor (1995) proposes that this system of prefixes arose in pre-pNN from an external possession construction (EPC) typical of Australian languages in which the PR and PM were represented by separate NPs in the same case (Hale 1981; McGregor 1985). The majority of instances of this EPC may reasonably be presumed to have involved the PR NP in either the intransitive subject or transitive object role, and thus to have been morphologically unmarked. If, as seems reasonable, the PR NP was typically a pronominal and typically immediately preceded the PM, the sequence of NPs might well have reduced to a single NP, and ultimately to a single word, the pronominal PR becoming bound to the inalienably possessed nominal. This system of nominal prefixing is viable in WNN languages – at least those for which we have sufficient evidence, including Nyulnyul, Jabirrjabirr, Bardi, and Jawi. In ENN languages the system is marginal: it is not in evidence in Yawuru or Nyikina, and is minimally attested in Warrwa. Capell (1952/1953) mentions the system in Warrwa; by the time of my fieldwork in the late 1980s it had virtually disappeared. One speaker never used it, using instead the phrasal possessive constructions; the other speaker used it inconsistently on just eleven body part nominals: -(u)ngu ‘stomach’, -lirr ‘mouth’, -liwa ‘ear’, -alma ‘head’, -mala ‘hand’, -(m)barrma ‘armpit’, -midi ‘leg’, -(ng)kurinykuriny ‘navel’, -yambala ‘foot’, -yangalany ‘tongue’ and -nyji ‘back’. The marginal position of the system in the late twentieth century is presumably a consequence of language attrition, though the Warrwa system of the 1940s may well have been reduced vis-à-vis the proto-ENN (and pNN) system. A system of nominal suffixes to inalienably possessed nominals developed in ENN. All ENN languages show a set of inalienable nominals (mostly body part terms) that involve the final segments jina, the elsewhere allomorph of the 3min.obl pronoun (Stokes 1982: 49), as in, e.g., kurdijina ‘shoulder’ (Yawuru and Nyikina). Moreover, the Big Nyikina dialect has a system of productive possessive suffixing of inalienably possessed items and kinterms; the suffixes are bound forms of the oblique pronominals (with some morphophonemic alternations) (Stokes 1982: 49– 52). The person and number distinctions of the oblique pronominal system are maintained (four persons and minimal-augmented numbers, not Assiniboine as for the bound pronominal prefixes). In Warrwa a few nominals for parts of the body take pronominal suffixes: ngunii- ‘nose’, kurndi- ‘shoulder’ and balngany- ‘thigh’. However, just four such suffixes are attested, all derived from the corresponding oblique pronominals: -yanu ~ -janu 1min; -yiya ~ -jiya 2min; -yina ~ -jina 3min (also

Grammaticalization patterns in Nyulnyulan language

1085

used for other persons and numbers); and -yirra 3aug. Whether this system arose in proto-ENN and followed the trajectory of the pronominal prefixes, or arose at a later time in Big Nyikina and/or Warrwa is impossible to tell. One guesses that the development was stimulated in the first place by the loss of the system of pronominal prefixing to inalienable nominals, but this is consistent with both scenarios.

2.3 Incipient grammaticalization of a nominal lexeme It is sometimes suggested that grammaticalization always begins with lexical items, and that all grammatical items have ultimate origins in lexical items. The evidential basis for this claim is weak, and in the case of Australian languages we often do not have sufficient evidence to permit reconstruction of a source lexeme – if there ever was one. Thus, the forms are short and frequently of low distinctiveness, and plausible resemblances exist to many different lexemes; insufficient historical-comparative work has been done to shore up claims. Moreover, even if lexical sources can be substantiated, it is not clear that they have anything to do with subsequent grammaticalizations. The clearest instances of grammaticalization of lexemes in Australian languages – including Nyulnyulan languages – are in the verbal domain: from verb to verb classifier (see § 3.1) and from verb to marker of a TAM category (McGregor 2013b). In the Nyulnyulan context there are just a few instances of non-verbal lexical items that appear to be beginning to grammaticalize. Perhaps most interesting of these is the nominal bur(u) ‘place’, which is attested in all Nyulnyulan languages, and is a reflex of pNN *buru ‘place’. This lexeme is generic in meaning and shows a wide range of senses: – Spatial location: a place (ground, earth, locality, country, dwelling place, camping place, nest, generalised sense of place, e.g., where something happened; space occupied by a person); – Material comprising or associated with a place (dirt, dust, what is above or below the surface of the earth – vegetation, nature of the ground at any level of granularity); – Abstract location – the position or place of something in an abstract ontological space; – The tide; – Temporal senses (often in compounds with adverbials or nominals): season, stretch of time (of restricted or arbitrary length), time of day (but not the ‘day’ sense as in many Australian languages, which is expressed by ‘sun’). It would take us too far from present concerns to discuss each of these senses. I illustrate here just the two most abstract uses of the nominal. First, buru ‘place’ can be used in the temporal sense, as in, e.g., Warrwa miliyarrinyjunu buru (long:ago place) ‘long ago’ and example (8).

1086 (8)

William B. McGregor

Bardi (Aklif 1991) i-ng-gard-ij booroo 3min.nom-en-enter-pst place ‘It’s evening time.’

Second, it can be used roughly as in the English ‘take the place of ’ or ‘put (something) in the place of ’, as in (9) and (10). The grammatical analysis of these examples is not entirely clear due to the paucity of tokens; presumably bur(u) ‘place’ is either a type of cognate object or a PV forming a lexicalised collocation with the following IV. (9) Nimanburru, Nyulnyul (Nekes and Worms 1953: 409) jesus bör ̣ i-m-bandj-an jesus bur i-m-barnj-an jesus place 3min-en-exchange-pst ‘Jesus sacrificed himself.’ (10) Nimanburru, Nyulnyul (Nekes and Worms 1953: 410) bör ̣ ŋa-n-m-en warindjer yēl bur nga-n-m-in warinyjirr yil place 1min.nom-cm-put-pst one dog ‘I sacrificed one dog.’ There is evidence of incipient grammaticalization of this nominal to: (i) a dummy of place; (ii) a marker of interrogatives of place; (iii) a derivational morpheme; and (iv) a relational morpheme. We briefly outline each of these developments in turn. (i) There are a number of possibly dummy uses of bur(u) ‘place’ where the lexeme appears to be used primarily to satisfy grammatical requirements of the language, namely that there be a subject NP. Thus, it sometimes occurs as the subject of ‘weather’ clauses as in (11). There are also perhaps dummy object (or cognate object) uses of bur(u) ‘place’ in many Nyulnyulan languages in clauses of vision, as in (12) – although it is possible that in such examples bur(u) ‘place’ is used as a PV, as suggested by, e.g., Aklif (1999: 33) for Bardi. (11) Bardi (Aklif 1991) moola gij i-n-j-ij booroo hot very 3min.nom-cm-say-pst place ‘The weather has got hot.’ (12) Nyulnyul bin wamb bambur arri bur i-la-jal this man blind not place 3nom-irr-see ‘This man is blind; he can’t see.’

Grammaticalization patterns in Nyulnyulan language

1087

(ii) In WNN languages bur(u) ‘place’ is sometimes used with an interrogative to specify a content interrogative of spatial location, direction, etc. For instance, in Nyulnyul angka bur (who place) ‘where’ is the most commonly used spatial interrogative in my corpus. Placement of postpositions suggests that this is a compound – e.g., Bardi jana-booroo-ngan (where-place-all) ‘where to’, where the allative is attached to the end of the phrase, not in its usual place at the end of the first word. This use of bur(u) ‘place’ is not attested in ENN. (iii) bur(u) ‘place’ is found in second position in compounds with nominals, specifying types of place, country, rock, etc. and compounds with certain adverbials specifying times. In some circumstances the compounding process is quite regular, and bur(u) ‘place’ is used (almost) derivationally. In perhaps all Nyulnyulan languages a toponym with bur(u) ‘place’ can be used metonymically as a personal name, as in, e.g., the personal names winawal-bur and jilany-bur in Nyulnyul and ngarrigoon-booroo in Bardi (Bowern 2012: 260).4 Interestingly, there is one attestation of buru ‘place’ in an expression resembling Warlpiri kuman(y)jayi, a euphemism for the name of a recently deceased person, whose name is tabooed. (Bates nd) cites Jukun koorabooroo: “when a namesake (Broome) dies, the living one is “koorabooroo”.” In WNN languages bur(u) ‘place’ occurs in the ordinal adverbial yalirr-bur ‘first (place)’ in Nyulnyul and Jabirrjabirr, and aloo-booroo ‘first (place)’ in Bardi. This form is also used in reference to birth order: yalerbor ‘the elder, the first’ (Nekes and Worms 1953: 920). In Warrwa, buru ‘place’ occurs in a compound with layin ‘alone’ meaning ‘by selves’, as in (13). (13) Warrwa kudii ø-ngi-rra-rnda-na-da kinya layin-buru run 3nom-nfut-aug-go-pst-hab this alone-place ‘They ran around by themselves.’ In ENN languages we find what appear to be derivational affixes involving the bisyllabic buru and meaning ‘thing associated with’. Thus in Yawuru -jinaburu “is attached to a stem and derives a noun which means ‘something utilised for … (ing)’” (Hosokawa 1991: 213),5 as in kurlin-jinaburu (sleep-NZ) ‘bed’, dujul-jinaburu (hit-NZ) ‘hammer’, and nganka-jinaburu (talk-NZ) ‘radio’. Nyikina has a similar derivational suffix (Stokes 1982: 54–55), while in Warrwa the corresponding -yina(-)buru is an incipient grammaticalization. Semantically this development is not implausible and

4 Bowern (2012: 260) suggests that in this context the construction in Bardi is not a compound, and booroo ‘place’ has become a suffix. 5 Presumably this is historically a compound form with -jina 3min.obl.

1088

William B. McGregor

may involve metonymic transfer from the general “space” occupied by an event or entity to a thing existing in that space, and central to it.6 Warrwa has the dyadic kinterm suffix -ngurruburu, as in babaala-ngurruburu (brother-DY) ‘pair of brothers’ and yaaku-ngurruburu (brother.in.law-DY) ‘pair of brother-in-laws’. This form may perhaps be etymologically -ngurru-buru. The second component is doubtless buru ‘place’; perhaps the first is a bound form of ngurra ‘place, camp’ borrowed from a nearby Pama-Nyungan language. (iv) According to Nekes and Worms (1953: 473) Nyikina and Yawuru have a type of temporal allative postposition -djibor ~ -iboro ‘unto’ – see also Stokes (1982: 110); they also mention Jabirrjabirr -bor ‘unto’ (Nekes and Worms 1953: 473). Examples are given in (14) and (15). As in these examples, this suffix seems to be associated with resultative secondary predicates. It is thus semantically not too implausible to link the marker to -ji ~ -yi dat plus bur(u) ‘place’. Thus bur(u) may have grammaticalized into (part of) a relational postposition. (14) Yawuru (Nekes and Worms 1953: 473) yēla-ni i-na-m-bondan wamba galbo-iboro balo-gon yila-ni i-na-m-bundan wamba kalbo-iburu balu-gun dog-erg 3nom-cm-en-bit man up-res tree-in ‘The dog followed the man biting him until he climbed a tree.’ More literally, ‘The dog bit the man until he was up in a tree.’ (15) Nyikina (Stokes 1982: 110) wakalman dalaw yi-m-barnji-na-da kunbulu-yibur blow.to.head cut 3min.nom-en-exchange-pst-hab blood-res ‘He used to hack himself around the head until he bled.’ Such a grammaticalization of a generic nominal ‘place’ might seem unsurprising for an Australian language, given the well-known significance of place, and the attested extensions from space to time (e.g., Boroditsky and Gaby 2010). However, there are no similar grammaticalizations in nearby Bunuban, or, as far as I am aware, Worrorran languages. There is a possible parallel grammaticalization in Karajarri (Pama-Nyungan): according to Nekes and Worms (1953: 808) this language has a suffix -ŋura with the temporal ‘until’ sense, as in ŋura-ŋura (night-until) ‘till night’. This suffix may well be etymologically ngurra ‘place’. More speculatively, we might

6 A possible parallel development in WNN languages involved not buru ‘place’ but ngurra ‘place’, borrowed from a Pama-Nyungan language to the south. Thus, Nekes and Worms (1953: 624, 807) refer to an agentive derivational suffix -gor ~ -goro ~ -ŋor ~ -ŋoro in Bardi, Nyulnyul, Jabirrjabirr, and Yawuru. This is unproductive in the modern languages, and in most instances appears to be synchronically part of the root, although there are some exceptions: bindan-ŋoro (bush-NZ) ‘bushman’, djalŋga-ŋor(o) ~ djalŋga-gor(o) (magic.power-NZ) ‘doctor’.

Grammaticalization patterns in Nyulnyulan language

1089

suggest a link with suffixes in a number of nearby Pama-Nyungan languages with the similar shape -wurra ~ -gurra: e.g., the Walmajarri resultative secondary predicate marker -wurra (Hudson 1978: 34), the Jaru resultative/purposive -gurra ~ -wurra, the Nyangumarta temporal enclitic -kurra ‘while’ (Sharp 2004: 291), and the Warlpiri complementiser -kurra.

3 Grammaticalization of verbal categories The verbal constructions in Nyulnyulan languages have been the site of a number of grammaticalization processes, including the binding of previously independent pronominals as affixes or enclitics to IVs. These grammaticalizations, however, must have occurred in pre-pNN. Few of the remaining grammemes in the IV have obvious sources outside of the IV. In this section we briefly outline three independent sets of grammaticalization, one concerning the development of the CVC, the other two concerning developments within the IV.

3.1 Verb classification McGregor (2002) argues that the northern portion of the Australian continent forms a large verb-classification sprachbund. This system of classification has its locus in the CVC, which involves a morphologically virtually inert PV in collocation with an IV, as in examples (13) and (15) above. Languages differ in the extent and frequency to which this system is employed and the degree to which it has been systematised. At one extreme are languages in which it appears to have only recently begun to emerge (e.g., Wangkajunga); at the other extreme are languages (such as Gooniyandi) where the erstwhile IVs have completely lost their lexical status and have become verb classifiers. The CVCs of Nyulnyulan languages appear to be in a fairly early stage of development (McGregor 2002: 150). The systems in WNN languages show evidence of being less entrenched and pervasive than in ENN languages (see McGregor 2002: 150). Thus, while in all Nyulnyulan languages all IVs have the potential for free occurrence, WNN languages have several hundred IVs whereas ENN languages have fewer than a hundred. CVCs make up a smaller proportion of the verbal construction tokens in texts in WNN languages (a bit over a quarter of verbal expressions in the Nyulnyul textual corpus) than in ENN languages (about 45 % of verbal expressions in a sample of 21 Warrwa texts). Presumably, this is a consequence of the more significant contact of ENN languages with nearby unrelated languages that have more grammaticalized CVC systems. Indeed, there is evidence that certain pairings of PVs and IVs in Warrwa may be calques on Bunuban ones; such calques could be

1090

William B. McGregor

at least partly responsible for driving the expansion of the CVC. The WNN languages are more isolated, and do not directly abut languages of any other family. Although CVCs are more entrenched in ENN languages than in WNN and considerably more frequent in usage, there seems to be no significant difference in the degree of grammaticalization of the systems. Regardless of the absolute number of IVs in a given Nyulnyulan language, around ten to twenty may occur in CVCs, and about a dozen are productive in CVCs; the remainder are restricted to simple verb constructions (SVCs) comprising just an IV. The CVC systems in all modern languages are verb classifying systems, in which the IV root classifies the PV (McGregor 2002: 266–275). As shown in McGregor (2002: 354–358), these classification systems are semantically based, with schematic meanings relating to three dimensions: Aktionsart, valency and vectorial configuration (an abstract system of vectors comprising the trajectory of the event). These schematic meanings relate in obvious ways to the meanings of the lexical IVs themselves. This can be seen from the data presented in McGregor (2018), which compares the ten primary and productive verbal categories in Nyulnyul and Warrwa, representatives of the WNN and ENN subgroups respectively. It will be observed that eight of the categories are shared between the languages and show virtually identical semantics, with no difference in the degree of schematicity of their meanings; seven of these are marked by cognate IVs. In each language around ten additional IVs occasionally occur in CVCs; these show less commonality across the languages. I have argued (McGregor 2013a, 2018) that – contrary to Bowern (2008) – the system of verb classification in the modern languages can be traced back to pNN. The evidence for this comes from the existence of consistent collocations across many languages of certain PVs and IVs that are not entirely predictable – that are to some extent “arbitrary” – and of collocations involving PVs and/or IVs that are not cognate, but that are replacements of previously cognate forms. Moreover, these reconstructed collocations differ from the corresponding collocations in languages of nearby families (Bunuban, Worrorran, Jarrakan, Pama-Nyungan). Whether or not this was a system of verb classification is uncertain. Presumably the systems of the modern languages are more entrenched than in the proto language; whether they are more grammaticalized remains uncertain. It seems that at least eight IVs in pNN were employed in CVCs: *-JU ‘say’, *-NI ‘sit, be’, *-MA ‘put’, *-BANJI ‘exchange’, *-KA ‘carry’, *-WA ‘give’, *-NYA ‘get’, and ‘go’ (whatever its form may have been). These IVs have broadly tended to retain their relative frequencies of occurrence in CVC types. However, in Nyikina the ‘say’ verb became more promiscuous, while in Yawuru and Warrwa the ‘sit, be’ IV expanded its collocational potential; the ‘give’ IV went out of use in CVCs in Nyikina and Yawuru. Possible innovations of proto-ENN and proto-WNN, respectively, are of *-NGARA ‘become’ and *-KAL ‘wander’. It seems reasonable to assume that sometime in pre-pNN language there were only IVs, and that CVCs had emerged by the time of pNN. Either by then or some

Grammaticalization patterns in Nyulnyulan language

1091

Tab. 2: Semantic characteristics of categories defined by the 10 primary classifying IVs in Nyulnyul and Warrwa (source: McGregor 2018: 340). (Note that IVs are cited in the text in capitals to distinguish them from other categories of lexeme.) Nyulnyul Atelic -N ‘be’ [stative] -JID ‘go’ [activity progresses over time] -KAL ‘wander’ [action not uniquely directed towards a specific goal] – -K ‘carry’ [move something by constantly applied force to new location]

Warrwa

-NI ~ -WANI ‘be’ [stative] -ARNDA ‘go’ [activity progresses over time] – -NGARA ‘become’ [inchoative] -KA ‘carry’ [move something by constantly applied force to new location]

Telic -R ‘pierce’ [action taking place in a straight line, impacting on something at a point] – -W ‘give’ [action directed outwards from actor, making contact with something] -NY ‘get’ [acquire or achieve an entity or condition by active means] -M ‘put’ [induce something to enter new state, condition, or location]

-NKA ‘hit’ [impact violently on something] -WA ~ -ø ‘give’ [action directed outwards from actor, making contact with something] -NDI ‘get’ [acquire or achieve an entity or condition by active means] -MA ‘put’ [induce something to enter new state, condition, or location]

±Telic -BARNJ ‘exchange’ [reflexive/reciprocal action] -J ‘say, do’ [dynamic activity]

-BANYJI ‘exchange’ [reflexive/reciprocal action] -JI ‘say, do’ [dynamic activity]

–

later time the CVC system grammaticalized to a verb classification system in which IVs served as classifiers, while they retained their full lexical uses. CVCs involve compounding of two verbal elements, and show a degree of unpredictability. One might argue that the relevant diachronic processes are processes of lexicalisation not grammaticalization (e.g., Schultze-Berndt 2000, 2006). In my view, however, the systems fall at the boundary of lexicon and grammar, and show sufficient consistency to permit them to be classified as grammatical systems. Indeed, IVs in CVCs clearly do not serve in their ordinary lexical functions, but rather classify events, and as such, serve grammatical functions as category markers.

3.2 Conjugation classes IVs in Nulnyulan languages fall into two primary conjugation classes, the na-class and the ø-class, distinguished by two partly distinct sets of nominative pronominal prefixes and by the presence or absence of a marker (na- ~ ni- ~ nu- ~ n- ~ a-) in a position somewhere between the initial pronominal prefix and the verb stem or root.

1092

William B. McGregor

For instance, in Warrwa the IV -RA ‘poke’ belongs to the na-class and shows the first person minimal forms nga-na-ra-ny (1min.nom.nfut-cm-poke-pfv) ‘I poked it’ in the past and ka-na-ngka-ra (1min.nom.fut-cm-fut-poke) ‘I will poke it’ in the future. The IV -ARNDA ‘go’ belongs to the ø-class, and the corresponding forms are nga-rnda-ny (1min.nom.nfut-go-pfv) ‘I went’ and ka-rnda (1min.nom.fut-go) ‘I will go’.7 There are usually also some irregular verbs that do not fall into these two major conjugation classes, and in Yawuru there are two ø conjugations, differing only in the second person minimal forms in the future. See McGregor (2002: 214– 219) for further details. There is a good (though imperfect) correspondence between conjugation class and IV valency.8 In Yawuru, for instance, 39 of the 40 known na-class IVs are bivalent, while the two ø-classes are about two-thirds monovalent (Hosokawa 1991: 121– 123). There are also a small number of IVs – a score or so – in each NN language that belong to both conjugation classes; in most cases the valency correlation for these ambicategorial IVs is even better than for IVs generally. Conjugated in the naclass the IV is almost always bivalent, and in the ø-class usually monovalent, or at least has the possibility of occurring in an intransitive clause. For instance, a reflex of *-MARRA ‘burn’ is found in most Nyulnyulan languages, and it is consistently ambicategorial: inflected in the na- pattern it is bivalent ‘burn something’; in the øpattern it is used in the monovalent ‘something burns’, possibly also in the bivalent ‘get burnt by something’. This is illustrated by the following Nyikina examples from Stokes (1982): transitive (16) shows the IV -MARRA ‘burn’ conjugated in the naclass; intransitive and medio-active (McGregor 1999) (17) and (18) show the same IV conjugated in the ø-class. (16) Nyikina wamba-ni wa-n-a-marra wali man-erg 3min.nom.fut-cm-fut-burn meat ‘The man will cook the meat.’ (17) Nyikina jungku yi-marra-n fire 3nom-burn-prs ‘The fire is burning.’

7 As these examples show, there may be other differences between the conjugation classes, here in terms of how future tense is realised. We need not go into the morphological complexities here. 8 By IV valency I refer to the number of inherent roles in the clause type in which the IV occurs in a simple verb construction. In clauses with CVCs the correlation with valency is weaker and not entirely predictable.

Grammaticalization patterns in Nyulnyulan language

1093

(18) Nyikina manyja baba yi-rr-marra-ny jungku-ni many child 3nom-aug-burn-pfv fire-erg ‘Many children were burned by fire.’ McGregor (2002: 361) suggests that the conjugation marker na- is a reflex of a pronominal prefix, a third person minimal accusative prefix cross-referencing the object. There is circumstantial evidence to support this suggestion. First, assuming this origin, the conjugation marker occurs in a not unexpected position in bivalent IVs, following the subject pronominal prefix – in languages of nearby Worrorran and Bunuban families the IV contains pronominals cross-referencing both subject and object (though they are often fused together in portmanteau forms). Second, there is a third person singular pronominal or demonstrative form nu ~ na that is widespread in non-Pama-Nyungan languages (Blake 1988: 7, 11; Evans 2000), and which represents a plausible source for the proposed accusative pronominal prefix of pre-pNN – cf. the third person minimal prefix to prefix-taking nominals (§ 2.2 above). Apparently, something similar happened in the IV across the Nyulnyulan family as happened in most ENN languages to the pronominal prefixes to nominals: the third person minimal form replaced all of the distinct person forms. In the nominal the result was that the third person minimal marker became fused with the old nominal root which it was adjacent to, forming the new root (cf. Watkins’s law); in the IV it became a conjugation marker, forming a conjugation class strongly associated with bivalent IVs. No accusative pronominal prefix was ever present in monovalent IVs, explaining the correlation with the ø-class. The ambicategorial IVs are evidently reflexes of IVs that could take or not take accusative pronominal prefixes depending on the clause type they occurred in. Perhaps as the system of accusative pronominal prefixes was breaking down, new bivalent IVs incorporated into the language (e.g., by borrowing) simply followed the inflectional pattern of monovalent IVs but employed an enclitic pronoun cross-referencing the object. This would explain the weaker valency correlation of ø-class IVs.

3.3 The applicative and other post-IV stem grammemes In many Nyulnyulan languages a smallish number of grammemes that occur in post-stem position in the IV are identical in form with postpositions, or at least an allomorph. These items show somewhat more restricted behaviour in verbal constructions than in NPs, where (as mentioned in § 1) they occur on the first unit regardless of its part of speech membership. By contrast, they are by and large restricted to IVs in both SVCs and CVCs. It is generally presumed in Australian linguistics that such “bivalent” grammemes began as nominal case-markers, and subsequently extended to marking verbal categories. This is consistent with the fact that

1094

William B. McGregor

the meanings in verbal contexts appear to be developments from the meanings in nominal contexts rather than the reverse. All Nyulnyulan languages for which adequate information is available have an applicative marker that occurs somewhere between the IV stem and the pronominal enclitic(s). Its form is identical with the instrumental or comitative postposition that is the reflex of pNN *-ngany com. For instance, Bardi has -ng(a) (Bowern 2012: 364, 366), as in (19). The applicative marker is formally identical with the instrumental postposition -ng(a) (Bowern 2012: 206), as in (20). (19) Bardi baawa-nim i-n-jiidi-ng jan ooldoobal child-erg 3nom-cm-go-app 1min.obl things ‘The child is touching my things.’ (20) Bardi jardinkool a-rr-a-n=irr bornko bardaga-ng irrolo-ng round.up 1nom-aug-say-cnt=3aug.acc around stick-ins spear-ins ‘We round them [the fish] up [in the fish trap] with sticks and spears.’ The development of an applicative marker from a comitative is not implausible, especially for an applicative of the ENN type. For instance, Warrwa has two applicative constructions, a comitative and an instrumental (McGregor 1998). Instrumental applicatives do not normally affect the transitivity of the clause, and the instrumental NP retains the instrumental postposition. Crucially, instrumental applicatives assert a close association between the instrument and an entity in another clausal role: the object if there is one, otherwise the subject. Thus, the instrumental applicative is possible in (22) since there is close physical contiguity between the ant eggs (in object role) and the coolamon (in instrument role) throughout the event. For an ordinary event of spearing, as predicted, the instrumental applicative was rejected by a speaker, and just the ordinary transitive with an instrumental NP was accepted, (23). (21) Bardi baawa-nim i-n-jiidi-ng jan ooldoobal child-erg 3nom-cm-go-app 1min.obl things ‘The child is touching my things.’ (22) Warrwa biina jirrwadaj ø-ja-n-ngany/ bakal-ngany/ ant:eggs yandy 3min.nom-say-prs-app coolamon-ins ‘She is yandying the ant eggs in a coolamon.’

Grammaticalization patterns in Nyulnyulan language

1095

(23) Warrwa ngul nga-na-ma-ny waya-ngany spear 1min.nom-cm-put-pfv wire:spear-ins ‘I speared it with a wire-tipped spear.’ The Warrwa comitative applicative usually has an effect on the transitivity of the clause, as shown by the second clause in (24), a ditransitive clause with two object NPs, jina wanangarri ‘his money’ and ngayu ‘me’. But it does more than this: it also asserts a close association between the object corresponding to a comitative marked NP (here referring to the speaker) in the agnate ordinary clause and an NP in some other role in the clause, either the subject or another object (i.e., the money in [24]). See also Yawuru example (31) below, where the applicative with the IV ‘be’ is the conventional expression of ‘have’; here the close association is between the subject (PR) and object (PM). (24) Warrwa yaab ngi-rnda-ny jina wanangarri away 3min.nom-go-pfv 3min.obl stone ø-na-ma-ny-ngany-ngayu 3min.nom-cm-put-pfv-app-1min.acc ‘He went away, having left his money with me.’ In WNN languages the applicative seems to be more entrenched grammatically, and is not restricted to instrumentals and comitatives. Thus, Nyulnyul also has benefactive and locative applicatives (McGregor 2012: 591–597), while Bardi has goal applicatives, which perhaps correspond to both benefactive and locative applicatives of Nyulnyul, as well as instrumental and comitative applicatives (Bowern 2012: 488– 496). As in ENN languages the applicative does not invariably affect clausal transitivity. My best guess is that the wider uses of the applicative in WNN languages are extensions from the comitative and instrumental uses, and express the directedness of the action to the beneficiary or location/goal which is closely associated with some key role in the clause. Thus, in the locative applicative (25) it seems reasonable to construe a close physical relation between the speaker and their coolamon as a consequence of the event. This story needs to be further developed and substantiated, however. (25) Nyulnyul (Nekes and Worms 1953: 558) wan-ag-aŋ ŋai bindjen djān wa-na-k-ang-ngay binjin jan 2min.nom-cm-carry-app-1min.acc coolamon 1min.obl ‘Bring me my coolamon!’

1096

William B. McGregor

It seems to be rare in Australian Aboriginal languages for applicative markers to develop from case markers, and none of the applicative markers discussed in Austin (1997) appear to have such a source. Moreover, applicatives are lacking in nearby languages. However, sources in case marking adpositions have been suggested for Arawakan languages (Danielsen 2011) and Navajo (Mithun 2001); Haspelmath and Müller-Bardey (2004) suggest that applicatives sometimes have sources in oblique case markers. Baker (1988) analyses applicatives in various Bantu and Austronesian languages as involving preposition incorporation; diachronically this can be interpreted as grammaticalization of applicatives from adpositions. In the ENN languages Yawuru and Nyikina a verbal morpheme with the same shape as the dative postposition (or one of its allomorphs) occurs as a marker of subordinate clauses of purpose (Hosokawa 1991: 151–152; Stokes 1982: 327–330). In Nyikina the dative marked clause may be insubordinated (Evans and Watanabe 2016: 2), in which case the clause indicates a situation that is expected to occur, either from the point of view of the doer of the action or someone else, e.g., the speaker. In Yawuru the same form is found in IVs in main clauses where according to Hosokawa (1991: 141–142) it is used on non-third person future IVs to intensify the directive or imperative function – a use which is evidently closely related to the Nyikina main clause use. Both of these uses are consistent with development from an original dative postposition that marked nominal purpose (among other things); this grammaticalization has occurred in a number of other Australian languages (see further Evans and Watanabe 2016). Nyulnyul has a set of complementisers that are formally identical with postpositions and mark various types of subordinate clause. These include locative, allative, ablative, and temporal postpositions (McGregor 2012: 162), which mark various types of adverbial subordinate clauses of time, place, condition, purpose, apprehension, etc. (McGregor 2012: 652, 660–667). In Bardi a form identical with the allative is also used as a purposive clause subordinator and the semblative as a temporal clause subordinator (Bowern 2012: 633). It seems probable that the complementiser functions are diachronic developments from the earlier case marking functions. There are a number of other complementisers the sources and developmental pathways of which are either opaque or less certain. This is the case for the two Bardi sentential clitics =b(a) and =(j)amb (Bowern 2012: 367). The ENN languages show a subordinate clause marker with shapes -jarri ~ -yarri ~ -arri ~ -ja, marking conditional, temporal clauses and sometimes relative clauses. There is no homophonous postposition, though there is a comitative postposition -nyarri in Warrwa which could perhaps be the historical source. This is not implausible semantically. Intriguingly, Nyulnyul shows, according to Nekes and Worms (1953), a somewhat similar form, which they represent -djer – possibly -jirr – which again closely resembles the Nyulnyul comitative –nyirr.

Grammaticalization patterns in Nyulnyulan language

1097

4 Grammaticalization of predicative possession constructions Nyulnyulan languages show both verbal and verbless predicative possession constructions, that code a possessive relation between two NPs (McGregor 2001a, 2001b). We discuss verbal and verbless predicative possession in order below.

4.1 Verbal predicative possession Most Nyulnyulan languages have a ‘have’ IV that is used in the expression of predicative possession. Its forms differ across the family, although most are probably cognate: -LABA, -LEBANBAD -BAKAND -BIKA ~ -BI -BA

Bardi Nyulnyul, Jabirrjabirr, Nimanburru Nyikina Warrwa

Except for Bardi -LABA, these IVs belong to the ø-class. All, however, occur in transitive clauses, with the PR cross-referenced by a nominative pronominal prefix and optionally represented by an NP marked by the ergative postposition, as shown in (26)–(27). (26) Nyikina (Stokes 1982: 322) nga-m-bika-ny-jarr-irr manyja yila 1min.nom-en-have-pst-rel-3aug.acc many dog ‘I used to have lots of dogs.’ (27) Nyulnyul kinimirr-in i-rr-bakand-in karrj jarringk shark-erg 3nom-pl-have-prs sharp tooth ‘Sharks have very sharp teeth.’ (28) Bardi (Bowern 2012: 359) jarri=gidi aarli i-rr-a-laba-n Boyo agal Lance-nim this=then fish 3-aug-cm-have-cnt Boyo and Lance-erg ‘Here are Boyo and Lance: they have a fish.’ -BAKAND, -LABA, -BIKA ~ -BI, and -BA are generally glossed in the sources as ‘have’, ‘own’, and ‘possess’. However, -BAKAND in Nyulnyul can also mean ‘hold’

1098

William B. McGregor

and ‘keep’ (as in [29]) and for Nyikina -BIKA ~ -BI the ‘keep’ sense is attested (as in [30]). For Warrwa -BA both ‘hold’ and ‘keep’ senses are attested. Only in Bardi are these senses not attested for -LABA – interestingly according to Bowern (2012: 360) most examples of this IV are from earlier sources; in the Bardi she recorded -INYA ‘catch’ was used in the senses ‘have’, ‘possess’, ‘hold’, and ‘grab’. (29) Nyulnyul yirr-in i-ngi-rr-bakand ngarrij-ang wangal-in dub i-li-ny-an 3aug-erg 3nom-pst-pl-have hard-ins wind-erg blow 3nom-irr-catch-imp ‘They held it down tightly, lest the wind blow it away.’ (30) Nyikina (Stokes 1982: 63) kuwaniya-ni yi-m-bika-na jungku jimbin wil-an crocodile-erg 3min.nom-en-have-pst fire inside water-loc ‘The crocodile kept the fire inside in (i.e., under) the water.’ It seems likely that the ‘have’ IV in at least some Nyulnyulan languages – possibly in pNN – began life as a ‘hold, grasp’ IV, and that this IV has not fully developed into a grammatical marker. I have argued (McGregor 2001b) that clauses with this verb still denote situations that happen or take place in the world. As such, they are typically used in circumstances where possession is not inalienable, and is more or less temporary, non-intrinsic, and/or non-characteristic, as in (26)–(28) above. (Recall that sharks’ teeth are regenerative.) Yawuru is the one Nyulnyulan language that lacks a ‘have’ verb (Hosokawa 1991). Instead, it uses a comitative applicative construction with the ‘be’ IV to express predicative possession, as in (31). (31) Yawuru (Hosokawa 1991: 257) marru-kun nga-nga-ny-ngany jurru head-loc 1min.nom-be-imp-app louse ‘My head is lousy.’ (or ‘I have head lice.’) Like the ‘have’ verb in other Nyulnyulan languages, this verb is not a purely grammatical marker, and is also used in the ‘hold’ sense (Hosokawa 1991: 180). Presumably Yawuru previously had a ‘have’ verb cognate with the ‘have’ verb of Nyikina and Warrwa, and that this went out of use, and was replaced by the applicative construction with ‘be’. Significantly, there remains evidence of the expected meanings of this applicative form, ‘sit with’, ‘stay with’, ‘live with’, etc. (Hosokawa 1991: 180) – and these are the only meanings associated with this construction in the other ENN languages. It is widely presumed that Australian languages lack a ‘have’ verb. However, examination of available materials reveals that a number of Australian languages

Grammaticalization patterns in Nyulnyulan language

1099

do have one or more verbs admitting a ‘have’ interpretation. These include, among others, non-Pama-Nyungan languages Gooniyandi, Jaminjung, Wambaya, Ndjébbana, Mangarrayi, Murriny-Patha, and Warray as well as Pama-Nyungan languages Jaru, Karajarri, Mangala, Arrernte and Baakandji. Usually this verb also admits the stative ‘hold’ sense. It thus appears that the situation for most Nyulnyulan languages – in which a ‘hold’ verb has begun to lose some of its lexical meaning and expresses possession in the absence of the sense of holding – is not unique. It has not become a fully grammatical marker however, and in the language in which it may perhaps be most developed grammatically, Bardi, it seems that a more lexical expression involving a ‘catch’ IV is preferred.

4.2 Verbless predicative possession constructions 4.2.1 ‘Belong to’ constructions Three Nyulnyulan languages – Nyulnyul, Bardi and Warrwa – show a more unusual non-verbal predication construction involving use of an oblique free pronominal as a possessive copula. (32) shows the basic shape of this construction. (32) pm + obl:pro + pr As indicated, the PM NP is usually initial, the PR NP final, with the oblique pronoun – which cross-references the PR – typically between them. The construction normally translates as ‘belongs to’, and thus represents a converse of the verbal predicative possession construction. Examples are:9 (33) Nyulnyul ina bur jirr kujarr wamb this camp 3aug.obl two man ‘This camp belongs to the two men.’ (34) Warrwa yangki jina wanangarri kinya who 3min.obl money this ‘Who does this money belong to?’

9 This construction differs in key respects from the well-known mihi est-type possessive construction. Formally, neither PM nor PR NP is case-marked, and the copula is not a verb. And semantically, the construction expresses ‘belong to’ not ‘have’. There is no reason to believe that this construction predates the ‘have’ construction.

1100

William B. McGregor

What has happened in this construction is that the erstwhile free oblique pronominal has become a copula. In the languages with this construction oblique possessive pronouns are used to mark attributive possession, suggesting that the clausal copula usage developed from the NP-internal usage. This diachronic development is not evident in nearby languages, including those languages in which oblique pronouns have become NP-internal possessive copulas. On the other hand, grammaticalization of attributing and identifying copulas from determiners is well attested in languages of the world.

4.2.2 A negative verbless predicative possession construction Many Nyulnyulan languages also have a verbless predicative possession construction that is dedicated to the expression of the notion that the erstwhile PR lacks possessions of a specified type (McGregor 2010b: 219–222). As shown by examples (35)–(36) in WNN languages this construction typically involves an oblique pronominal cross-referencing the PR that is attached to the negative word, which normally occurs in initial position; this is followed by an NP indicating the lacking PM. However, in Bardi occasionally the third person minimal form of the oblique pronominal is employed regardless of the person and number of the PR – see (37). (35) Nyulnyul arri-jan wilamay not-1min.obl food ‘I’ve got no food.’ (36) Bardi arra-jana goolboo not-1min.obl stone ‘I have no money.’ (37) Bardi (Aklif 1999: 21) arra-jina goolboo not-3min.obl stone ‘I have no money.’ Nyikina and Yawuru have constructions like (37) in which an invariant negator is found, albeit one that apparently incorporates an instance of the 3min.obl pronominal, as in (38). In Yawuru the negator is morphologically analysable into the negative particle plus the oblique pronominal; in Nyikina the negator is a cranberry morph, with the initial piece formally similar to the negative particle. Warrwa, like Bardi, shows both patterns, the invariant negator (etymologically the third person

Grammaticalization patterns in Nyulnyulan language

1101

minimal form) being used by one speaker, the person-variable negator by the other speaker – as in examples (39) and (40) respectively. (38) Yawuru (Nekes and Worms 1953: 695) maral nga-nga-n marlu-jina ngamarri desirous 1min.nom-sit-prs not-3min.obl tobacco ‘I am in need of tobacco.’ (More literally, ‘I want tobacco; I have none.’) (39) Warrwa ngayu maliyina wanangarri 1min nothing stone ‘I’ve got no money.’ (40) Warrwa mali-yarra wanangarri nothing-1 & 2aug.obl stone ‘We’ve got no money.’ The negative predicative possessive construction with neg-obl.pro can plausibly be reconstructed for pNN. It survives in WNN languages. In ENN languages Yawuru and Nyikina the person-number marking system reduced to just the 3min form – as happened in various other contexts in these two languages (see above). In Warrwa, this reduction most likely happened as a result of grammatical attrition accompanying the obsolescence of the language. Nyulnyul alone shows a negative existential construction that asserts the nonpresence of entities of the specified type in a particular location, as shown by (41) and (42).10 As these examples indicate, the oblique pronominal cross-references the locative NP, which specifies the search domain in which the entities are claimed to be absent. (41) Nyulnyul bin irrjiwar-uk bucket arri-jirr wul that three-loc bucket not-3aug.obl water ‘There is no water in those three buckets.’ (42) Nyulnyul arri-jin wilamay bur-uk jan not-3min.obl food camp-loc 1min.obl ‘There’s no food in my home.’

10 There are some hints of the emergence of something like this construction in Warrwa and Bardi, but no new structure appears to have yet emerged.

1102

William B. McGregor

How did these constructions develop? Given that diachronic development of possessives from existentials is attested in a number of languages (e.g., Heine and Kuteva 2002: 127–128), it might be suggested that the negative possessive construction is a development from the negative existential. There are however problems with this. First, it is not clear why the oblique pronominal would have arisen in the putative source construction where the search domain is normally an inanimate, and hence unlikely to be referred to by a pronominal: generally, the only locations cross-referenced by bound oblique pronominals are human ones. Thus, the source construction is not a plausible one. Second, why has the locative marker been lost in the negative possessive construction? Third, according to this scenario the negative existential would have had to have been lost in all but one Nyulnyulan language; Occam’s razor suggests it is more likely that it emerged in just one language. McGregor (2010b) advocates the alternative view that the negative existential of Nyulnyul developed from the negative predicative possession construction. In particular, he suggests that the development began with a subtype of negative predicative possession in which the possessive relation was inalienable, and a locative NP specified a body part locus of the (lacking) inalienable possession. Effectively, the source subconstruction was like they have no hair on their heads. This was reinterpreted semantically as a negative existential ‘there is no hair on their heads’, and formally as a same case marked EPC (see § 2.2), roughly there is no hair (on them) on their heads. From this the negative non-EPC existential was backformed, as in there is no food in my house. The negative predicative possession construction arose earlier from the ‘belong to’ construction discussed in § 4.2.1. One type of negation of this construction involved an indefinite PM, and expressed the notion that no entity of the specified type belonged to the PR. Since the PR was a more likely topic, the PR NP tended to occur initially and the negative construction was reanalysed as a negated ‘have’. In the process, the copula, the oblique pronominal, fused with the negator which it typically followed. For fuller discussion of these scenarios see McGregor (2010b: 224–228).

5 Conclusions This paper discusses a selection of grammaticalization processes that have likely occurred in Nyulnyulan languages from pre-pNN times to the present. Few grammemes in the modern languages have obvious sources in lexical items or in other grammemes. This may be due in part to the age of the items in question; it is also almost certainly partly due to extensive borrowing of grammemes – as has been known since Heath (1978), borrowing of bound grammemes is not uncommon in Australian languages. There is also evidence of borrowing of larger constructions into Nyulnyulan languages – e.g., McGregor (2007a) argues that a particular type of desiderative construction in Warrwa was probably calqued on a desiderative con-

Grammaticalization patterns in Nyulnyulan language

1103

struction in Worrorran languages. There may thus be an unusually large number of grammemes and constructions in the modern languages that did not arise via independent processes of grammaticalization – although of course their subsequent development may have been in accordance with grammaticalization processes, shared or independent. The CVC is an example: it was doubtless areally spread over a large part of Australia in earlier times, and subsequently developed in Nyulnyulan by diachronic processes that were also areally shared. It is possible to trace the diachronic development of predicative possession constructions in Nyulnyulan languages. A few other clausal and clause complex constructions have been touched on in passing, in the discussion of the diachrony of the grammemes that mark them: the applicative marker and complementisers. We have not investigated the developments of these constructions in detail. Except for the development of complementisers, all of these diachronic developments are, in the linguistic region, peculiar to Nyulnyulan. The grammaticalization processes discussed in this paper are of two main types. First are processes whereby grammatical markers of various types – case markers, copulas, and bound pronominals – develop from pronominals or determiners. A notable characteristic of grammaticalization in Nyulnyulan is a tendency for such indexical items to lose their primary referential-indexical features and become markers, in line with the suggestions of McGregor (1997a) that these two functions are of the same fundamental type. No nearby non-Nyulnyulan languages show such a propensity for this type of grammaticalization. In the wider scheme of things, such grammaticalizations are not especially unusual, although they have tended to be downplayed in the grammaticalization literature. In particular, indexical items are perhaps the most underrated sources of case markers (McGregor 2017). Of the grammaticalization targets of indexical items mentioned in this chapter, only two are mentioned in Heine and Kuteva (2002) (see also Diessel 2011): demonstrative > third person pronoun (which perhaps occurred in pre-pNN) and either of these > copula (Heine and Kuteva 2002: 108–109, 235). However, what these authors refer to is an identifying or attributing copula, not a possessive one, and the other person forms of the pronouns are not involved, as they are in Nyulnyulan. None of the other seven targets of demonstratives or five for third person pronouns mentioned in Heine and Kuteva (2002) are in evidence in Nyulnyulan.11 Second are processes whereby lexical items have developed, or have begun to develop, grammatical functions. The clearest instances of these processes come from the CVC, where IVs have grammaticalized into category markers. This grammaticalization process, however, has happened within a matrix of lexicalisation processes. Just one instance of the incipient grammaticalization of a nominal is

11 Heine and Kuteva (2002: 234) mention the development of third person pronouns into agreement markers, but the examples they give are of grammemes that clearly do not show agreement in person and number features.

1104

William B. McGregor

known, that involving the generic place nominal bur(u); there is also an instance of incomplete grammaticalization of an IV into a ‘have’ verb – a particularly interesting example since the most grammaticalized instance of this IV appears to have been recently replaced by another IV. The only one of these that is anything like the grammaticalizations discussed in Heine and Kuteva (2002) is the development of a ‘have’ verb from a lexical verb meaning something like ‘get’ or ‘keep’ – the meaning of the pNN source verb is impossible to pin down with certainty. Development of verbs into category markers is not mentioned at all in Heine and Kuteva (2002), and the nearest hit is the development of verbs into markers of aspect – recall that Aktionsart is one of the primary features characterising the verbal categories. The developmental pathway for bur(u) ‘place’ shows little in common with the pathways from ‘place’ identified in Heine and Kuteva (2002: 239–240). The PV usage of this nominal with a meaning something like ‘sacrifice, take the place of ’ is reminiscent of place > instead, though quite clearly Nyulnyulan bur(u) ‘place’ has not developed into a grammeme ‘instead’. Derivational uses of ‘place’ are not mentioned in Heine and Kuteva (2002), and nor is a temporal relational use ‘until’ on resultative secondary predicates. In addition to these two main types there are also a small number of instances of grammemes that have extended their domains of use and/or senses, as in the case of a few postpositions which may have extended their domain from NPs to the IV. Heine and Kuteva (2002: 37) mention development of complementisers from allatives and datives; both are attested in Nyulnyulan. The grammaticalization of comitative to applicative is not mentioned specifically in this source, although it does emerge in passing in the development of comitatives to ‘have’ possessives (Heine and Kuteva 2002: 88). Heine and Kuteva (2002: 84–86) mention the grammaticalization of comitative marker to instrument marker. It is not made clear in this source precisely how this is to be understood, but usually this development is regarded as involving extension via metaphoric transfer. As far as I can tell, this is unlikely to have been the case in the developments from *-ngany com. Most probably this was in pNN a generalised comitative that also covered instrumental senses; borrowing of another comitative postposition into Nyulnyulan languages (presumably around the time of the split into Eastern and Western branches) led to the restriction of the reflexes of *-ngany com to an instrumental marker (in most languages) or comitative (Yawuru) – as discussed in McGregor (1997b). In sum, few of the grammaticalization pathways identified for Nyulnyulan languages are widely recognised in the grammaticalization literature or included in Heine and Kuteva (2002). Processes of metaphoric transfer play a minimal role in Nyulnyulan grammaticalizations. For instance, although there is evidence of the cross-linguistically widespread metaphoric transfer from the domain of space to that of time, this has but very limited relevance to Nyulnyulan grammaticalizations. The appearance of bur(u) ‘place’ as a part of a resultative marker (see § 2.3) is one of the few instances I am aware of. I am aware of no convincing cases of metaphori-

Grammaticalization patterns in Nyulnyulan language

1105

cally based processes of grammaticalization in the domain of case markers, or of tense, mood and aspect markers in Nyulnyulan languages (see also McGregor 2013b).

Abbreviations abl = ablative, acc = accusative, all = allative, app = applicative, aug = augmented, cm = conjugation marker, cnt = continuous, com = comitative, CVC = compound verb construction, dat = dative, en = epenthetic nasal, ENN = Eastern Nyulnyulan, EPC = external possession construction, erg = ergative, fut = future, hab = habitual, imp = imperfective, ins = instrumental, intens = intensifier, irr = irrealis, IV = inflecting verb, loc = locative, min = minimal, nfut = non-future, nom = nominative, NP = noun phrase, nz = nominaliser, obl = oblique, per = perlative, pfv = perfective, pl = plural, PM = possessum, pNN = proto-Nyulnyulan, PR = possessor, pro = pronominal, prs = present, pst = past, PV = preverb, rel = relative, res = resultative, SVC = simple verb construction, then = then, WNN = Western Nyulnyulan. IVs are cited in capitals to distinguish them from other categories of lexeme. 1, 2, and 3 indicate speaker, hearer and other respectively, where these are assumed to represent single individuals.

References Aklif, Gedda. 1991. Bardi material. Canberra: AIATSIS Aboriginal Studies Electronic Data Archive (ASEDA). Aklif, Gedda. 1999. Ardiyooloon Bardi ngaanka: One Arm Point Bardi dictionary. Halls Creek: Kimberley Language Resource Centre. Austin, Peter. 1997. Causatives and applicatives in Australian Aboriginal languages. In Kazuto Matsumura & Tooru Hayasi (eds.), Dative and related phenomena, 165–225. Tokyo: Hitsuji Shobo. Baker, Mark 1988. Incorporation: a theory of grammatical function changing. Chicago: University of Chicago Press. Bates, Daisy M. nd. Native vocabularies – Broome district. Unpublished manuscript. Canberra: Australian National Library. Blake, Barry. 1988. Redefining Pama-Nyungan: towards the prehistory of Australian languages. Aboriginal Linguistics 1. 1–90. Boroditsky, Lera & Alice Gaby. 2010. Remembrances of times east: absolute spatial representations of time in an Australian Aboriginal community. Psychological Science 21(11). 1635–1639. Bowern, Claire. 2004. Diagnostic similarities and differences between Nyulnyulan and neighbouring languages. In Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method, 269–290. Amsterdam: John Benjamins. Bowern, Claire. 2008. The reconstruction of Nyulnyulan complex predication. Diachronica 25(2). 186–212. Bowern, Claire. 2012. A grammar of Bardi. Berlin: De Gruyter Mouton. Burridge, Kate. 1990. Sentence datives and grammaticalization of the dative possessive: evidence from Germanic. La Trobe Working Papers in Linguistics 3. 29–47.

1106

William B. McGregor

Capell, Arthur. 1940. The classification of languages in north and north-west Australia. Oceania 10. 241–272, 404–433. Capell, Arthur. 1952/1953. Notes on the Njigina and Warwa tribes, N. W. Australia. Mankind 4. 351–360, 450–496. Danielsen, Swintha. 2011. The grammaticalization of adpositions into valency increasing verbal affixes in Arawakan languages. Paper presented to SSILA annual meeting, Pittsburgh. Diessel, Holger. 2011. Where do grammatical morphemes come from? On the development of grammatical markers from lexical expressions, demonstratives, and question words. Unpublished. http://www.personal.uni-jena.de/~x4diho/Grammaticalization%20NRG4.pdf (accessed 30 June 2017.) Du Bois, John W. 1987. The discourse basis of ergativity. Language 63. 805–855. Evans, Nicholas. 2000. Sprung from what source? – How our proto-Australian reconstructions affect what we look for in the rest of Sahul. Paper presented at Workshop on Sahul Languages, Max Planck Institute for Psycholinguistics, Nijmegen. Evans, Nicholas & Honoré Watanabe. 2016. The dynamics of insubordination: an overview. In Nicholas Evans & Honoré Watanabe (eds.), Insubordination, 1–37. Amsterdam and Philadelphia: John Benjamins. Greenberg, Joseph. 1988. The first person inclusive dual as an ambiguous category. Studies in Language 12(1). 1–18. Hale, Ken. 1981. Preliminary remarks on the grammar of part-whole relations in Warlpiri. In John Hollyman & Andrew Pawley (eds.), Studies in Pacific linguistics in honor of Bruce Biggs, 333– 344. Auckland: Linguistic Society of New Zealand. Haspelmath, Martin & Thomas Müller-Bardey. 2004. Valence change. In Geert Booij, Christian Lehmann & Joachim Mugdan (eds.), Morphology: a handbook on inflection and word formation, vol. 2, 1130–1145. Berlin: de Gruyter Mouton. Heath, Jeffrey. 1978. Linguistic diffusion in Arnhem Land. Canberra: Australian Institute of Aboriginal Studies. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Hosokawa, Komei. 1991. The Yawuru language of West Kimberley: a meaning-based description. Canberra: Australian National University PhD thesis. Hudson, Joyce. 1978. The core of Walmatjari grammar. Canberra: Australian Institute of Aboriginal Studies. McConvell, Patrick. 2003. Headword migration: a Kimberley counter-example. In Nicholas Evans (ed.), The non-Pama-Nyungan languages of northern Australia: comparative studies of the continent’s most linguistically complex region, 75–92. Canberra: Pacific Linguistics. McGregor, William B. 1985. Body parts in Kuniyanti clause grammar. Australian Journal of Linguistics 5(2). 209–232. McGregor, William B. 1989. Greenberg on the first person inclusive dual: evidence from some Australian languages. Studies in Language 13(2). 437–451. McGregor, William B. 1994. Warrwa. München & Newcastle: Lincom Europa. McGregor, William B. 1995. Nominal prefixing in Nyulnyul. In Hilary Chappell & William B. McGregor (eds.), The grammar of inalienability: a typological perspective on body part terms and the part-whole relation, 251–292. Berlin: Mouton de Gruyter. McGregor, William B. 1997a. Semiotic grammar. Oxford: Clarendon Press. McGregor, William B. 1997b. The story of -ngany in Nyulnyulan languages. Paper presented at First International Workshop on Australian Linguistics, University of Melbourne. McGregor, William B. 1998. Applicative constructions in Warrwa. In Anna Siewierska & Jae Jong Song (eds.), Case, typology, and grammar: in honour of Barry J. Blake, 171–199. Amsterdam: John Benjamins.

Grammaticalization patterns in Nyulnyulan language

1107

McGregor, William B. 1999. The medio-active construction in Nyulnyulan languages. Studies in Language 23(3). 531–567. McGregor, William B. 2001a. Non-verbal predicative possession in Nyulnyulan languages. In Jane Simpson, David Nash, Mary Laughren, Peter Austin & Barry Alpher (eds.), Forty years on: Ken Hale and Australian languages, 337–352. Canberra: Pacific Linguistics. McGregor, William B. 2001b. The verb HAVE in Nyulnyulan languages. In Irène Baron & Michael Herslund (eds.), Dimensions of possession, 67–84. Amsterdam: John Benjamins. McGregor, William B. 2002. Verb classification in Australian languages. Berlin & New York: Mouton de Gruyter. McGregor, William B. 2006. Focal and optional ergative marking in Warrwa (Kimberley, Western Australia). Lingua 116(4). 393–423. McGregor, William B. 2007a. A desiderative complement construction in Warrwa. In Jeff Siegel, John Lynch & Diana K. Eades (eds.), Language description, history and development: linguistic indulgence in memory of Terry Crowley, 27–40. Amsterdam: John Benjamins. McGregor, William B. 2007b. Ergative marking of intransitive subjects in Warrwa. Australian Journal of Linguistics 27(2). 201–229. McGregor, William B. 2008. Indexicals as sources of case markers in Australian languages. In Folke Josephson & Ingmar Söhrman (eds.), Interdependence of diachronic and synchronic analyses, 299–321. Amsterdam & Philadelphia: John Benjamins. McGregor, William B. 2010a. Optional ergative case marking systems in a typological-semiotic perspective. Lingua 120(7). 1610–1636. McGregor, William B. 2010b. The semantics, pragmatics and evolution of two verbless negative constructions in Nyulnyul. Oceanic Linguistics 49(1). 205–232. McGregor, William B. 2012. The Nyulnyul language of Dampier Land, Western Australia. Canberra: Pacific Linguistics. McGregor, William B. 2013a. Comparing linguistic systems of categorisation. In Lars Borin & Anju Saxena (eds.), Approaches to measuring linguistic differences, 387–427. Berlin: De Gruyter Mouton. McGregor, William B. 2013b. Grammaticalization of verbs as temporal and modal markers in Australian languages. In Folke Josephson & Ingmar Söhrman (eds.), Diachronic and typological perspectives on verbs, 107–132. Amsterdam: John Benjamins. McGregor, William B. 2017. Grammaticalization of ergative case marking. In Jessica Coon, Diane Massam & Lisa deMena Travis (eds.), The Oxford handbook of ergativity, 447–464. Oxford: Oxford University Press. McGregor, William B. 2018. The history of verb classification in Nyulnyulan languages. In William B. McGregor & Søren Wichmann (eds.), The diachrony of classification systems, 315–352. Amsterdam: John Benjamins. Mithun, Marianne. 2001. Understanding and explaining applicatives. In Mary Andronis, Christopher Ball, Heidi Elston & Sylvain Neuvel (eds.), Proceedings of the Thirty-seventh Meeting of the Chicago Linguistic Society: Functionalism and formalism in Linguistic Theory, vol. 2, 73–98. Chicago: Chicago Linguistic Society. Narrog, Heiko & Bernd Heine. 2011. Introduction. In Heiko Narrog & Bernd Heine (eds.), The Oxford handbook of grammaticalization, 1–16. Oxford: Oxford University Press. Nekes, Hermann & Ernest A. Worms. 1953. Australian languages. Fribourg: Anthropos Institut. Nekes, Hermann & Ernest A. Worms. 2006. Australian languages. (Edited by William B. McGregor.) Berlin & New York: Mouton de Gruyter. Schultze-Berndt, Eva. 2000. Simple and complex verbs in Jaminjung: a study of event categorisation in an Australian language. Nijmegen, Netherlands: Catholic University of Nijmegen dissertation.

1108

William B. McGregor

Schultze-Berndt, Eva. 2006. Taking a closer look at function verbs: lexicon, grammar, or both? In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching language: the standing challenge of grammar writing, 359–391. Berlin & New York: Mouton de Gruyter. Sharp, Janet. 2004. Nyangumarta: a language of the Pilbara region of Western Australia. Canberra: Pacific Linguistics. Stokes, Bronwyn. 1982. A description of Nyigina: a language of the West Kimberley, Western Australia. Canberra, Australia: Australian National University dissertation. Stokes, Bronwyn & William B. McGregor. 2003. Classification and subclassification of the Nyulnyulan languages. In Nicholas Evans (ed.), The non-Pama-Nyungan languages of northern Australia: comparative studies of the continent’s most linguistically complex region, 29–74. Canberra: Pacific Linguistics.

Susanne Maria Michaelis and Martin Haspelmath

27 Grammaticalization in creole languages: Accelerated functionalization and semantic imitation 1 Overview In this paper, we discuss grammaticalization in creole languages, basing ourselves to a large extent on the data from the Atlas of Pidgin and Creole Languages Structures (APiCS, Michaelis et al. 2013a, 2013b). We start out from the idea that grammaticalization can be said to involve three main processes (i) semantic change, which often results in (ii) functionalization (content items become function items), and is then followed by (iii) coalescence (cliticization, agglutination, fusion of the function item). One central observation of this paper is that English-based, French-based and Ibero-Romance-based creoles show a lot of functionalization, but not much coalescence (though the former content items are often reduced). The claims can probably be extended to all creole languages, though we know a lot less about the prehistory of other kinds of creoles, so they will not play an important role in this paper. In the literature of the last decade, creoles have often been said to be special in that they exhibit great transparency and simplicity (McWhorter 2001; Parkvall 2008; Bakker et al. 2011), but here we would like to stress that they can also be seen as showing accelerated functionalization, when compared to their lexifiers (English, French, Spanish, Portuguese). This is not a synchronic characterization of creoles, but it describes some key aspects of their earlier diachronic development. We think that this idea fits quite well with the extravagance account of grammaticalization (Haspelmath 1999), and less well with the traditional life-cycle view of creolization, according to which creoles arise from pidgins via nativization when children acquire the pidgin as native speakers and therefore expand it grammatically. But of course, creoles also tend to show a very substantial amount of grammatical influence from their substrate languages (so much so that they have sometimes been said to be relexified versions of these, as in Lefebvre [1998, 2015]). This substrate influence (called “apparent grammaticalization” in Bruyn [1996, 2009]) is often hard to tell apart from internal grammaticalization. We examine some of the relevant phenomena and argue that most cases of substrate influence that look similar to grammaticalization can be seen as instances of semantic imitation. This interacts with grammaticalization, but we do not see it as its cause. In the next section (§ 2), we will cite a wide range of new function items in creole languages from 13 different domains, to substantiate our claim that creoles show accelerated functionalization. Then in § 3, we will propose our explanation of this in terms of extra transparency, and in § 4, we discuss different kinds of semantic imitation (re-creation, calquing, and polysemy copying). https://doi.org/10.1515/9783110712735-013

1110

Susanne Maria Michaelis and Martin Haspelmath

2 New function morphemes A striking observation of comparative creole studies is that creole languages often show new function items (morphemes, words, particles, affixes) in various domains that did not exist in the lexifier when the languages arose (typically in the 17th and 18th centuries). The term lexifier is used in creole studies for the language of the colonizers that provided the bulk of content items in the creole languages. Thus, Spanish is the lexifier of Chabacano, French is the lexifier of Seychelles Creole, and English is the lexifier of Jamaican. Of course, the lexifiers have been undergoing grammaticalization processes themselves over the last few centuries, but in the creoles, we see much more of this, and in general it appears that grammaticalization is accelerated in creolization. This was sometimes observed in the literature, e.g., in Plag (2002), and Bakker (2008: 43) notes that in creoles, “it is often easier to trace grammaticalization patterns, and this can help shed light on grammaticalization in general”. However, the idea of accelerated grammaticalization in creoles that we would like to advance here has not been prominent in the literature on creoles. In the following sections, we list a substantial number of examples of new function morphemes in creole languages, mostly based on the chapters of APiCS. The new function items derive from a diverse range of content items (and sometimes from function items with more concrete meanings): nouns, verbs, quantifiers, adverbs, demonstratives, independent personal pronouns, and concrete adpositions. After each subsection heading in this section, we give the number of the relevant APiCS chapter in parentheses (all the information can also be accessed in the electronic version, at http://apics-online.info).

2.1 Definite articles (APiCS 28, 9) New definite articles arise from demonstratives (English that) or deictic adverbs (French là), for example: (1) Sranan a (e.g., a pikin ‘the child’) < da < English that (Bruyn 2009) (2) Kriol thet/thad (e.g., thad lif ‘the leaf ’) < English that: (2.i) Thad lif pat bla mukarra, im gud-wan bla so. dem leaf part poss river.pandanus 3sg good-adj dat sore ‘The leaf of the river pandanus is good for sores.’ (Schultze-Berndt and Angelo 2013) (3) Haitian Creole =la (e.g., nouvel=la ‘the news) < French là ‘there’

Grammaticalization in creole languages

1111

2.2 Indefinite articles (APiCS 29, 10) New indefinite articles arise from the numeral ‘one’, or from the word ‘other’, for example: (4) Sranan wan < English one (Bruyn 2009), also in other English-lexified creoles (5)

Guinea-Bissau Kriyol utru < Portuguese outro ‘other’

(5.i) utru omi musulmanu a man Muslim ‘a Muslim man’ (Intumbo, Inverno, and Holm 2013)

2.3 Plural markers (APiCS 22, 23) New plural markers arise from a word for ‘group’ or a word for ‘all’: (6) Seychelles Creole bann < French bande ‘group’ (6.i) Tou sa bann landrwa mon 'n ale. all dem pl place 1sg prf go ‘It’s to all these places that I have been.’ (Michaelis and Rosalie 2013) (7) Tok Pisin ol < English all (8) Diu Indo Portuguese tud < Portuguese tudo ‘all’ (also Tayo tule < French tous les)

2.4 Genitive markers (APiCS 38, 37) New genitive markers arise from beneficiary prepositions (‘for’) or from the verb ‘belong’: (9) Vincentian Creole fo ‘of ’ < English for (also other English-lexified Caribbean creoles) (9.i) di pikni fo di woman art child for art woman ‘the woman’s child(ren)’ (Prescod 2013)

1112

Susanne Maria Michaelis and Martin Haspelmath

(10) Seychelles Creole pour, Tayo pu < French pour ‘for’ (11) Tok Pisin bilong < (that) belong (to) (also Bislama blong, Kriol bla)

2.5 Personal pronouns in subject or possessor function (APiCS 62) New person forms for subject and possessor function are introduced on the basis of independent personal pronouns, as seen in (12)–(13) for subject function and in (14)–(15) for possessor function. The independent personal pronouns are not really “content items”, like the more typical sources of new function items, but they are forms that can occur independently, and the development is in other ways analogous. (12) Santome obligatory subject person forms (cf. Portuguese optional subject pronouns) (12.i) Bô na sêbê kuma bô so kota mu mon fa? 2sg neg know comp 2sg foc cut 1sg.obj hand neg ‘Don’t you know that it was you who cut my hand off?’ (Hagemeijer 2013) (13) Palenquero obligatory subject person forms (cf. Spanish optional subject pronouns) (13.i) Ele e ta trabahá. [he he is working] (Schwegler 2013) (14) Diu Indo-Portuguese d-el ‘his’, lit. ‘of him’ (cf. Portuguese possessive pronoun seu/sua) (15) Guadeloupean Creole timoun an mwen [child of me] ‘my child(ren)’ (cf. French preposed possessive pronoun mon/ma)

2.6 Accusative markers (APiCS 57) New accusative markers arise from adpositions meaning ‘with’ or ‘for’: (16) Batavia Creole kung, Papiá Kristang ku (< Portuguese com ‘with’), Ternate Chabacano con (< Spanish con ‘with’)

Grammaticalization in creole languages

1113

(17) Sri Lanka Portuguese -pa (< Portuguese para ‘for’) (cf. also Afrikaans vir < Dutch voor ‘for’) (17.i) eev vosa kuɲaadu-pa kada ɔɔra ki-lembraa 1sg 2sg.gen brother.in.law-acc every time hab-think.of ‘I often think of your brother-in-law.’ (Smith 2013)

2.7 Dative markers (APiCS 60, 61) New dative markers arise from adpositions meaning ‘along’, ‘with’ or ‘for’. (18) Bislama long (< English along), cf. also Kriol langa, la (19) Mauritian Creole avek/ek (< French avec ‘with’) (19.i) (av)ek ki sanla to'n don larzaṅ la? with who that.one 2sg.pfv give money def ‘To whom have you given the money?’ (Baker and Kriegel 2013) (20) Diu Indo-Portuguese pe (< Portuguese para ‘for’) (21) Papiá Kristang ku, Batavia Creole kung, Chabacano con/kon (cf. § 2.6)

2.8 Future tense markers (cf. APiCS 43, 44) New future tense markers have diverse sources: a verb meaning ‘go’ (as in Negerhollands), a temporal adverbial expression (as in Tok Pisin), or a purposive expression. (22) Negerhollands lo < loo ‘go’ < Dutch lopen ‘run’ (22.i) Morək mi lō lō . tomorrow 1sg fut go ‘Tomorrow I will go.’ (van Sluijs 2013) (23) Tok Pisin bai < English by and by (24) Seychelles Creole pou < French (être) pour

1114

Susanne Maria Michaelis and Martin Haspelmath

2.9 Past tense (or anterior) markers (APiCS 45) New past tense markers tend to come from auxiliaries meaning ‘be’, but a temporal adverbial may also be the source, as in Batavia Creole. (25) Seychelles Creole ti < French était ‘was’ (26) Jamaican wehn < English been (also in many other English-lexified creoles) (27) Principense tava < Portuguese estava ‘was’ (28) Batavia Creole dja (perfective marker) < Portuguese já ‘already’ (28.i) fala kung ile ki eo dja teng aki tell obj 3sg comp 1sg pfv be here ‘tell him that I have been here’ (Maurer 2013a)

2.10 Imperfective aspect markers (APiCS 46, 47, 48) New imperfective markers have diverse sources: spatial expressions like ‘there’ or ‘near’, and verbs like ‘stay’, ‘do’ and ‘know’. (29) Early Sranan de (prog) < English there (also in other Atlantic English-lexified creoles) (29.i) Hangri de killi mi. hunger prog kill 1sg ‘I am hungry (lit. Hunger is killing me).’ (van den Berg and Bruyn 2013) (30) Seychelles Creole pe, Haitian Creole ap (prog) < French (être) après ‘near, about (to do)’ (31) Tok Pisin i stap (prog) < English stop (in the sense ‘stay’) (32) Palenquero asé (hab) < Spanish hacer ‘do’ (cf. Gullah duhz) (33) Haitian Creole konn (hab) < French connaître ‘know’

Grammaticalization in creole languages

1115

2.11 Completive aspect markers markers (cf. APiCS 43, 44, 50) New completive markers come from verbs meaning ‘finish’ or ‘done’. (34) Seychelles Creole fin, (i)n < French finir (35) Bislama finis < English finish (36) Negerhollands ka < kaba (< Portuguese acabar ‘finish’) (37) African American English done (37.i) That snake done bit me again. ‘That snake has bitten me again.’ (Green 2013)

2.12 Reciprocal/reflexive pronouns (APiCS 87, 88, 89) New reflexive and reciprocal pronouns come from nouns meaning ‘companion, mate’ or ‘body’; in one case a reciprocal marker comes from a verb ‘play’ (‘act together’). (38) Guyanais kompannyen ‘each other’ < French compagnon; Seychelles Creole kanmarad ‘each other’ < French camarade (39) Creolese matii (< English mate?); (but many English-based varieties have continuations of one another) (40) Ternate Chabacano hugá (act together) < Spanish jugar ‘play’ (40.i) Ta hugá keré lótro dos. ipfv play.recp love 3pl two ‘The two of them love each other.’ (Sippola 2013) (41) Tok Pisin em yet ‘himself ’ (but most English-based creoles continue self) (42) Angolar ôngê ‘body, self ’ (but most Romance-based creoles continue mesmo/mismo/même) (42.i) Ê kôntê ôngê rê me. 3sg hate body his self ‘He hates himself.’ (Maurer 2013b)

1116

Susanne Maria Michaelis and Martin Haspelmath

2.13 Causative construction New causative constructions come from verbs meaning ‘do, make’. (43) Juba Arabic ámulu (‘make’), replacing the Arabic synthetic causative pattern (43.i) ána ámulu úo kásulu wésa 1sg make 3sg wash face ‘I made him wash his face.’ (Manfredi and Petrollino 2013) (44) Seychelles Creole fer < French faire ‘do’ Mon fer Zan manze vs French Je fais manger Jean (Michaelis and Rosalie 2013) In this last example, the new function item fer is not different from the French faire, but the word order shows that the Seychelles Creole construction is new and does not continue the French construction. Thus, here we have a new construction, using a function item that happens to have the same origin as the function item used in French (faire ‘make, do’, which always existed as a regular verb alongside the causative use). This counts as a new functionalization, because the construction was clearly created anew and does not continue the old construction. All the examples from creole languages that we have provided in this section show accelerated functionalization, in the sense that they show the development of new function items that did not develop simultaneously in the lexifiers. English could have developed a new dative marker (similar to Tok Pisin long, § 2.7), and French could have developed a new plural marker (similar to Seychelles Creole bann, § 2.3). But they did not, and we are trying to understand why. The lexifiers are also developing and undergoing grammaticalization, of course, but apparently much more slowly.

3 Accelerated functionalization of content items The examples given in the last section are a very strong indication that functionalization of content items in general happens more quickly in creoles (and apparently also in pidgins) than in other languages. While this is not a view that is currently widespread in creole linguistics, it is not a completely new observation either. Thus, Peter Bakker noted: Creoles show accelerated processes of grammaticalization (Bakker 2008: 43) The speed in which a whole grammatical system develops and grammaticalizes may be unique for this category of languages. (Bakker 2008: 46)

Grammaticalization in creole languages

1117

However, it does not seem correct to claim that “grammaticalization” in general is accelerated in creoles, because grammaticalization is usually understood more broadly, comprising both functionalization and coalescence. Roughly speaking, grammaticalization is the gradual drift in a grammar towards tighter structures and less freedom in the use of linguistic expressions. Content words become function words, free morphemes become clitics or affixes, and syntactic constructions become subject to tighter constraints. (Plag 2002: 229; cf. also Haspelmath 1998: 318)

It is only functionalization, the first stage of a grammaticalization cycle, that seems to occur more often in creoles than in the languages that served as their lexifiers and with which we compared them in § 2. The next steps in the process, which may ultimately lead to tightly coalesced grammatical markers, does not seem to be equally quick: Creole languages appear quite “analytic” (cf. Haspelmath and Michaelis 2017), not synthetic like the older Indo-European languages. So why is the functionalization process faster? We think that Plag and Bakker are on the right track when they point to the special communicative conditions in the history of creoles: The discrepancies in time span between certain processes in creoles and non-creoles strongly indicate that creoles develop much faster than non-creoles … The difference can of course be explained in terms of the discontinuity of transmission and the communicative pressure that exists in the situation in which the new creole language emerges. (Plag 2002: 239; emphasis ours) the speed of the process may be greater in Creoles, undoubtedly because of the communicative demands in the contact situation (Bakker 2008: 48; emphasis ours)

However, we think that the connection between the special communicative and social circumstances under which creole languages developed and the grammatical developments needs to be made more precise. Before advancing our proposal, we should note that there are reasons to think that the more extensive functionalization in creole languages is just a special case of a more general phenomenon, namely that such developments are favoured in language contact situations. This is in line with the observation that within Romance languages, we can observe a “grammaticalization cline” as in (45), reflecting the different “pace of grammaticalization” in the different languages (Carlier, De Mulder, and Lamiroy 2012). (45) French > Italian > Spanish (different pace of grammaticalization) This is based on a range of grammatical patterns where French seems to show the most advanced stage of grammaticalization, while Spanish is the most conservative language. In order to explain the differences, the authors appeal to language contact, among other factors:

1118

Susanne Maria Michaelis and Martin Haspelmath

The more languages spread over large populations and involve frequent language contact between individuals who are related to each other by weak ties, the faster languages may evolve by regularizing mechanisms, ultimately also reducing their morphological and grammatical systems. (Carlier, De Mulder, and Lamiroy 2012: 292)

Carlier et al. do not go into any details, but they also cite Lupyan and Dale (2010) and Trudgill (2011) for like-minded approaches. These authors also note that languages which have undergone more contact influence tend to show less morphological complexity. So how does faster functionalization relate to language contact? Here we would like to contrast two different hypotheses that might explain why more new function items arise in situations of more intensive language contact. Those are the Lossand-Replacement Hypothesis in (46) and the Extra-Transparency Hypothesis in (47). We will end up favouring the second hypothesis, but we should consider the first hypothesis, too. (46) Loss-and-Replacement Hypothesis In the transmission bottleneck of pidginization, inflectional and other non-salient grammatical markers are lost, because they cannot be acquired (e.g., Good 2012). This leaves a void, and when pidgins turn into full-fledged languages again, they need to fill the gaps by new material deriving from content words. This hypothesis is similar to the therapeutic view of grammaticalization, where it is claimed that new grammatical markers (function items) arise because the old ones have become reduced phonetically and are no longer distinctive enough. This view has a venerable tradition, but it has been shown to be wrong for grammaticalization in general (cf. Lehmann [1985]; Haspelmath [2000] gives five different arguments against what he calls the “reduction first” view of grammaticalization). Perhaps it is somewhat more plausible for creoles, because in the life-cycle view of creolization, one sees creoles as continuations of pidgins, and pidgins are indeed generally thought to be deficient, lacking many grammatical markers that full-fledged languages have. However, the life-cycle view of creolization is not the only possibility, and we favour a hypothesis that is compatible with the view that creoles do not (necessarily) arise from pidgins: (47) Extra-Transparency Hypothesis In social situations with many (or even mostly) adult second-language speakers, people need to make an extra effort to make themselves understood. This naturally leads to the overuse of content items for grammatical meanings, which may become fixed when more and more speakers adopt the innovative uses.

Grammaticalization in creole languages

1119

This is similar to the extravagance-based view of grammaticalization, which in our view offers the best account of unidirectionality (Haspelmath 1999). Grammaticalization in ordinary situations is explained as due to occasional extravagant language use, when speakers use a stronger or more vivid expression for a grammatical function where an ordinary form would have sufficed (e.g., in back of instead of behind). In high-contact situations, no appeal to extravagance is necessary, because the need for extra clarity or transparency can explain the greater tendency to use stronger, more vivid expressions for grammatical functions and thus to functionalize content items. It should be noted that extra transparency is not needed in all situations of language contact – when a society is generally multilingual, so that even children acquire more than one language, adult second-language acquisition does not take place and no such effect is expected. In recent years, creolists have tended to focus on transparency (Seuren and Wekker 1986; Leufkens 2013), simplification (McWhorter 2001, 2007; Parkvall 2008; Siegel 2008) or on the uniqueness of creole languages (Bakker et al. 2011), rather than on particularly fast functionalization, so there has not been any widespread discussion of these issues. However, grammaticalization in creoles has been discussed to some extent, especially by Bruyn (1996, 2009). Bruyn distinguished between “ordinary grammaticalization” and “instantaneous grammaticalization” (and also “apparent grammaticalization”, discussed in the next section). This distinction may be motivated by the specific Sranan phenomena she discusses, but we are not sure whether the cases seen in § 2 can generally be classified into these two groups. Even “ordinary grammaticalization” must be pretty fast in creoles, given that these languages are not older than 500 years, and often younger. A very important phenomenon that interacts with grammaticalization in creole languages is “semantic imitation” of substrate languages. This will be the topic of the next section.

4 Semantic imitation As is well-known, creole languages show striking influences from their substrate languages (e.g., Alleyne 1980; Keesing 1988; Parkvall 2000), and these are important for the kinds of grammaticalization/functionalization processes that we have seen. The substrate languages of a creole are the languages spoken by the indigenous populations that were involved in the creation of the creole. Thus, the substrate languages of Haitian Creole were Gbe languages and other languages of West Africa, and the substrate languages of Chabacano were the Austronesian languages of the Philippines. In quite a few cases, new function items seem to imitate (or be based on) function items in the substrate languages. A widely cited example is the distinction between mipela (first person plural exclusive) and yumi (first person plural inclusive),

1120

Susanne Maria Michaelis and Martin Haspelmath

and between yu (you, singular), yupela (you, plural), and yutupela (you, dual) in Tok Pisin, which was influenced by its Oceanic substrate languages in this way (Keesing 1991). The key phenomenon here is semantic imitation: Speakers of the recipient language innovate forms that can easily convey meanings of the donor language. This happens because it is easier to manage two languages when they differ only in the pronunciation, but not in their semantic structure. Literal translation is easy when the languages construe the speakers’ experience in analogous ways. Like extra transparency, semantic imitation also has to do with ease of processing, but it seems that it primarily helps the speaker, whereas extra transparency is mainly an aid for the hearer. The imitation is thus not a conscious process – the speakers are not trying to speak more like the donor language. On the contrary, they find it difficult to suppress the influence of the donor language and unconsciously carry over patterns of the donor language to the recipient language. We distinguish two broad classes of semantic imitation: re-creation (where a meaning is borrowed and is rendered by an independent form, § 4.1) and calquing (where the meaning of a composite form is rendered by an analogous composite form, § 4.2).

4.1 Re-creation Re-creation is the situation where speakers of a recipient language create a new expression to render a meaning that occurs in a donor language, and where this new expression is not related to the way in which the meaning is expressed in the donor language. Lexical re-creation is well-known from the literature on lexical borrowing (e.g., Haspelmath 2009) and is called loan creation. Examples are German Umwelt [around-world] for French milieu (originally mi-lieu [mid-place]), or German Kraftwagen [power-carriage] for automobile. Is there also a grammatical, constructional analog to lexical loan creation? According to Heine and Kuteva (2003, 2005), grammatical constructions are sometimes re-created in this sense, i.e., grammaticalization happens on the model of a pattern in a donor language. Some examples are given in (48). (48) a. The German definite article der/die/das re-creates the Romance (French/ Italian) definite article, using the anaphoric demonstrative b. The Tayo personal pronoun dual forms (nu-de, u-de, le-de) re-create the dual distinction in the New Caledonian languages, using the numeral de (French deux) c. The Bislama progressive aspect (stap V ‘be V-ing’) re-creates the progressive aspect of the Vanuatu languages, using the verb stap ‘stay, exist, be’

Grammaticalization in creole languages

1121

We agree that such developments are probably common, though they are hard to prove (just as it is not easy to prove that a new lexical creation like Umwelt is due to the borrowing of a meaning). Heine and Kuteva call this process “ordinary contactinduced grammaticalization”. We would simply call it constructional re-creation (or perhaps re-creative grammaticalization), because it seems that constructional calquing (§ 4.2.2) is much more common, so the qualification “ordinary” is quite odd.

4.2 Calquing In calquing, the meaning of a composite form in the donor language is rendered by an analogous composite form in the recipient language. Thus, while it is possible to re-create the meanings of simple forms, only complex forms can be calqued. We begin by illustrating lexical calquing, before moving to the more relevant constructional calquing.

4.2.1 Lexical calquing Lexical calquing is well-known from cases like English flea market (calqued from French marché aux puces), English loanword (calqued from German Lehnwort), and German Gesichtspunkt (calqued from French point de vue). In calquing (or loan translation), there is a donor pattern Xd+Yd that has a meaning ‘Z’ that is not fully predictable from ‘X’ and ‘Y’. The recipient language has elements Xr and Yr which correspond to Xd and Yd outside of the pattern Xd+Yd: donor recipient

Xd (marché) + Yd (aux puces) Xr (market) + Yr (flea)

English flea corresponds to French puce also outside of this construction, as well as market to marché.1

4.2.2 Constructional calquing Constructional calquing is a very widespread phenomenon in creole languages. Clear cases not involving grammaticalization are (49)–(50) (and see also [29.i] above from Sranan). 1 One could perhaps alternatively say that English borrowed a new, special meaning of flea from French by polysemy copying (see § 4.3), but since this new meaning is confined to this special context, calquing is a better term.

1122

Susanne Maria Michaelis and Martin Haspelmath

(49) Seychelles Creole (Michaelis and Rosalie 2013) Lapli pe tombe. rain prog fall ‘It is raining.’ (Lit. ‘Rain is falling.’) (50) Nigerian Pidgin (Faraclas 2013) He̱dpen dè du mì. headache ipfv do 1sg.obj ‘I’ve got a headache.’ (Lit. ‘Headache is doing/affecting me.’) The substrate languages (here Swahili [51] and Ewe [52]) have constructions expressing raining and headache in which the rain and the headache are in subject position, and this is imitated by the creole languages (cf. Michaelis 2019). (51) Swahili (Bantu; Vitale 1981: 57 f.) (Mvua) i-na-nyesha. (rain) CL9-prs-fall ‘It is raining.’ (lit. ‘Rain is falling.’) (52) Ewe (Kwa; Ameka 1990: 165) ta vé kofí head pain Kofi ‘Kofi had a headache.’ (lit. ‘head pained Kofi’) When the constructional calque involves a grammatical marker, we get a pattern that can be seen as a special case of grammaticalization. The examples in (53) and (54) illustrate what Heine and Kuteva (2005) call “replica grammaticalization”. Irish English has calqued the Irish ‘after’ perfect (53c), and Molisean Croatian has calqued the Italian comitative-like expression of instruments (54c). (53) a. Standard English He has gone. b. Irish English He is after going. ‘He has gone.’ c. Irish Tá sé tar éis imeacht. is he after going ‘He has gone.’ (Lit. ‘He is after going.’) (54) a. Standard Croatian nožem ‘with a knife’ (lit. knife.instr) b. Molisean Croatian s nožem ‘with a knife’ c. Italian con un coltello ‘with a knife’ (cf. con l’amico) Constructional calquing is particularly relevant in the present context because it can easily be shown that quite a few cases of functionalization in creole languages have a substrate model, i.e., they involve constructional calquing, for example:

Grammaticalization in creole languages

– – – – –

1123

reflexive pronouns based on ‘body’ (see, e.g., example 42) complementizers based on verbs of speaking (Sranan taki < English ‘talk’) postpositions based on spatial nouns (Sranan tapu < English ‘top’) (cf. Bruyn 1996, Yakpo and Bruyn 2015) plural markers based on 3rd person pronouns (Haitian Creole yo < French eux) coordinators based on a ‘with’ word (Seychelles Creole ek, Principense ki; [Michaelis and Rosalie 2000])

Bruyn (1996, 2009) therefore calls these cases “apparent grammaticalization”, i.e., she does not regard them as “real grammaticalization”. For Heine and Kuteva, they are instances of “replica grammaticalization”, but this notion is defined in an odd way that seems to presuppose that the grammaticalization process is a conscious one and the history of the construction in the donor language is known to the recipient language speakers (cf. Gast and van der Auwera [2012: 387–388] for justified criticism). Clearly, these cases involve semantic imitation, but at the same time they do involve functionalization of content items, like purely internal functionalization. For this reason, we have not excluded such cases from our examples of § 2. Just like functionalizations that do not imitate anything in a substrate language, these cases can be explained by the motivation of extra transparency. The only difference is that the choice of content item that was the basis for the functionalization was inspired by a substrate language. Lefebvre (1998, 2015) subsumes these cases under relabeling, i.e., a process where the semantics and syntax remain entirely intact, and only the phonological shape (the label) of the substrate language is replaced by a label from the lexifier. However, we think that constructional calquing is based on more general imitation rather than on strict syntax copying. In cases like Gesichtspunkt (< point de vue) and He’s after going, we can easily see that the model pattern was imitated, but not strictly copied, because the word order is not that of the model language.

4.3 Polysemy copying Another related concept is polysemy copying, which considers the semantic imitation of a substrate language expression from the perspective of a single form. For example, Seychelles Creole ek comes from French avec ‘with’, but it is polysemous, meaning both ‘with’ and ‘and’. This polysemy is very similar to the polysemy of the form na found in many Bantu languages (e.g., Swahili), so it is clear that the substrate influence was crucial here (Michaelis and Rosalie 2000).2 2 There are contexts in French where a NP conjunctional reading is possible, as in Marie avec Jésus meaning ‘Mary with/and Jesus’ (under a picture or statue representing the two). But if one were to continue the sentence, one would choose the verb in the singular clearly indicating that the subject

1124

Susanne Maria Michaelis and Martin Haspelmath

Heine and Kuteva (2003: 556) argue that polysemy copying is not “the primary strategy used to transfer grammatical categories from one language to another”, because the copying is often incomplete, i.e., not all meanings of a donor form are copied into a recipient language. But if we think of these cases in terms of imitation, this is not a problem, because the speakers of the recipient language did not blindly copy the donor language structure, but were merely inspired by it, so to say. For us, polysemy copying is not really different from constructional calquing; it is just a different perspective on the same phenomenon. In all the examples of § 4.2.2, we could have said that the function item acquires a new meaning inspired by the donor language (e.g., Irish English after comes to have the perfect meaning, and so on). But since the new meanings are generally restricted to specific constructions, a description in terms of constructional calquing seems generally more insightful.3 When polysemy copying concerns elements that also have a grammatical meaning, there is a strong tendency for new functional meanings to be added to forms that earlier had only content meaning. For example, English talk acquired the additional meaning ‘that (complementizer)’ in Sranan, and not the other way around. Thus, Sranan now has taki ‘talk; that’ (Bruyn 1996), and it did not add the content meaning ‘talk’ to its form for ‘that’ (dati ‘that’). Likeweise, Seychelles Creole ek ‘with’ acquired the additional meaning ‘and’, and not the other way around, i.e., the form e ‘and’ (from French et) did not acquire the additional meaning ‘with’. This was highlighted by Gast and van der Auwera (2012) (see also Matras and Sakel 2007: 834): [Polysemy copying] seems to work, in the majority of cases, in the direction of “more grammatical”. This may be due to the fact that translational equivalence is typically established at the more concrete end of the “semantic space” covered by a linguistic sign. (Gast and van der Auwera 2012: 384–385)

Thus, polysemy copying is one way in which content items develop function meanings and thus become function items. However, occasionally we do find cases where polysemy copying works the other way around, i.e., function items add another content meaning, or a more concrete meaning: (55) a. Solomon Islands Pijin fastaem ‘before; in front of ’ (< English first time ‘before’) (rather than the expected #infrontof ‘in front of; before) (Keesing 1991)

is 3rd person singular and not 3rd person plural. Therefore, we do not consider such special cases as counterexamples to a substrate interpretation. 3 However, lexical polysemy copying may, of course, happen outside of particular constructions. Thus, German Kopf ‘head’ acquired the additional meaning ‘head of a syntactic phrase’ in the 1980s, as a result of polysemy copying, regardless of any constructional context.

Grammaticalization in creole languages

1125

b. Guadeloupean Creole épi ‘and; with’ (< French et puis ‘and (then)’) (rather than the expected #avek ‘with; and’) (Michaelis & Rosalie 2000) c. Principense ningê ‘anyone/someone; person’ (< Portuguese ninguém ‘no one, anyone’) (rather than the expected #pessoa ‘person; someone/anyone’) (Maurer 2009) Thus, polysemy copying involving elements that also have a grammatical meaning does not have to lead to functionalization, even though it apparently most often does. We do not regard grammatical polysemy copying as a phenomenon distinct from grammaticalization, and not even distinct from constructional calquing, but rather just a different perspective on semantic changes in situations of constructional calquing.

5 Conclusion We have tried to show in this paper that one of the most striking phenomena in the development of the grammars of creole languages is the creation of new function items out of content items. This observation can be called accelerated functionalization. By contrast, coalescence of function items is not salient in creoles, and does not seem to be accelerated in any way. Maybe coalescence generally takes a long time, and not enough time has passed for this to happen in creoles (cf. McWhorter [2001] on creoles as “young” languages).4 We have argued that the accelerated functionalization is the result of a need for extra transparency in social situations with many adult second-language speakers who benefit from extensive use of more concrete items for grammatical meanings. Another striking phenomenon is the transfer of substrate features into the grammars of creole languages. We have argued that this results from semantic imitation of the substrate patterns in the speech of second-language speakers of the emerging creoles. Most commonly, we observe constructional calquing, though constructional re-creation also seems to play a role. From the perspective of the individual item, constructional calquing can be regarded as polysemy copying, and there are good reasons why this generally leads to the addition of grammatical meanings. Thus, constructional calquing can be seen as one way in which grammaticalization in

4 Bisang (2008) claims that coalescence (“coevolution of form and meaning”) is found to a much lesser extent in Southeast and East Asia than in Western Eurasia. It is conceivable that the limited coalescence in creole languages is not unrelated to this phenomenon, but we feel that the time factor is probably sufficient to explain why we do not see much coalescence (yet) in creoles.

1126

Susanne Maria Michaelis and Martin Haspelmath

creoles happens, and it need not be seen in opposition to internal (non-contactinduced) grammaticalization.

Acknowledgment The support of the European Research Council (ERC Advanced Grant 670985, Grammatical Universals) is gratefully acknowledged.

References Alleyne, Mervyn C. 1980. Comparative Afro-American: An historical-comparative study of Englishbased Afro-American dialects of the New World. Ann Arbor: Karoma Publishers. Ameka, Felix. 1990. The grammatical packaging of experiencers in Ewe: A study in the semantics of syntax. Australian Journal of Linguistics 10. 139–181. Baker, Philip & Sibylle Kriegel. 2013. Mauritian Creole structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/55. Bakker, Peter. 2008. The development of tense, mood and aspect in creole languages and the typology of affix order. In Folke Josephson & Ingmar Söhrman (eds.), Interdependence of diachronic and synchronic analyses (Studies in Language Companion Series, 103.), 43–59. Amsterdam: John Benjamins. Bakker, Peter, Aymeric Daval-Markussen, Mikael Parkvall & Ingo Plag. 2011. Creoles are typologically distinct from non-creoles. Journal of Pidgin and Creole Languages 26(1). 5–42. Bisang, Walter. 2008. Grammaticalization and the areal factor: The perspective of East and Mainland South East Asian languages. In María José López-Couso & Elena Seoane (eds.), Rethinking grammaticalization (Typological Studies in Language 76), 13–35. Amsterdam: John Benjamins. Bruyn, Adrienne. 1996. On identifying instances of grammaticalization in creole languages. In Philip Baker & Anand Syea (eds.), Changing meanings, changing functions: Papers relating to grammaticalization in contact languages, 29–46. London: University of Westminster. Bruyn, Adrienne. 2009. Grammaticalization in creoles: Ordinary and not-so-ordinary cases. Studies in Language 33(2). 312–337. Carlier, Anne, Walter De Mulder & Béatrice Lamiroy. 2012. Introduction: The pace of grammaticalization in a typological perspective. Folia Linguistica 46(2). 287–301. Faraclas, Nicholas. 2013. Nigerian Pidgin structure dataset. In APiCS Online (Michaelis et al. 2013b). https://apics-online.info/contributions/17 Gast, Volker & Johan van der Auwera. 2012. What is “contact-induced grammaticalization”? Examples from Mayan and Mixe-Zoquen languages. In Björn Wiemer, Bernhard Wälchli & Björn Hansen (eds.), Grammatical replication and borrowability in language contact, 381– 426. Berlin: Mouton de Gruyter. Good, Jeff. 2012. Typologizing grammatical complexities, or: Why creoles may be paradigmatically simple but syntagmatically average. Journal of Pidgin and Creole Languages 27(1). 1–47. Green, Lisa. 2013. African American English structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/14. Hagemeijer, Tjerk. 2013. Santome structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/35.

Grammaticalization in creole languages

1127

Haspelmath, Martin. 1998. Does grammaticalization need reanalysis? Studies in Language 22(2). 315–351. Haspelmath, Martin. 1999. Why is grammaticalization irreversible? Linguistics 37(6). 1043–1068. Haspelmath, Martin. 2000. The relevance of extravagance: A reply to Bart Geurts. Linguistics 38(4). 789–798. Haspelmath, Martin. 2009. Lexical borrowing: Concepts and issues. In Martin Haspelmath & Uri Tadmor (eds.), Loanwords in the world’s languages: A comparative handbook, 35–54. Berlin: De Gruyter Mouton. Haspelmath, Martin & Susanne Maria Michaelis. 2017. Analytic and synthetic: Typological change in varieties of European languages. In Isabelle Buchstaller & Beat Siebenhaar (eds.), Language variation – European perspectives VI: Selected papers from the 8th International Conference on Language Variation in Europe (ICLaVE 8), Leipzig 2015, 3–22. Amsterdam: John Benjamins. Heine, Bernd & Tania Kuteva. 2003. On contact-induced grammaticalization. Studies in Language 27(3). 529–572. Heine, Bernd & Tania Kuteva. 2005. Language contact and grammatical change. Cambridge: Cambridge University Press. Intumbo, Incanha, Liliana Inverno & John Holm. 2013. Guinea-Bissau Kriyol structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/33. Keesing, Roger M. 1988. Melanesian Pidgin and the Oceanic substrate. Stanford: Stanford University Press. Keesing, Roger M. 1991. Substrates, calquing and grammaticalization in Melanesian Pidgin. In Bernd Heine & Elizabeth C. Traugott (eds.), Approaches to grammaticalization, vol. 1, 315– 342. Amsterdam: John Benjamins. Lefebvre, Claire. 1998. Creole genesis and the acquisition of grammar: The case of Haitian Creole. Cambridge: Cambridge University Press. Lefebvre, Claire. 2015. Relabeling in language genesis. New York: Oxford University Press. Lehmann, Christian. 1985. Grammaticalization: Synchronic variation and diachronic change. Lingua e Stile 20(3). 303–318. Leufkens, Sterre. 2013. The transparency of creoles. Journal of Pidgin and Creole Languages 28(2). 323–362. Lupyan, G. & R. Dale. 2010. Language structure is partly determined by social structure. PLoS ONE 5(1). e8559. Manfredi, Stefano & Sara Petrollino. 2013. Juba Arabic structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/64. Matras, Yaron & Jeanette Sakel (eds.). 2007. Grammatical borrowing in cross-linguistic perspective. Berlin: Mouton de Gruyter. Maurer, Philippe. 2009. Principense: Grammar, texts, and vocabulary of the Afro-Portuguese Creole of the island of Príncipe, Gulf of Guinea. London: Battlebridge. Maurer, Philippe. 2013a. Angolar structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/36. Maurer, Philippe. 2013b. Batavia Creole structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/43. McWhorter, John H. 2001. The world’s simplest grammars are creole grammars. Linguistic Typology 5(2–3). 125–166. McWhorter, John H. 2007. Language interrupted: Signs of non-native acquisition in standard language grammars. Oxford: Oxford University Press. Michaelis, Susanne & Marcel Rosalie. 2000. Polysémie et cartes sémantiques: le relateur (av)ek en créole seychellois. Études Créoles 23(2). 79–100. Michaelis, Susanne Maria & Marcel Rosalie. 2013. Seychelles Creole dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/56.

1128

Susanne Maria Michaelis and Martin Haspelmath

Michaelis, Susanne Maria, Philippe Maurer, Martin Haspelmath & Magnus Huber (eds.). 2013a. The Atlas of Pidgin and Creole Language Structures. Oxford: Oxford University Press. Michaelis, Susanne Maria, Philippe Maurer, Martin Haspelmath & Magnus Huber (eds.). 2013b. Atlas of Pidgin and Creole Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://apics-online.info/ Michaelis, Susanne Maria. 2019. World-wide comparative evidence for calquing of valency patterns in creoles. Journal of Language Contact 12(1). 191–231. Parkvall, Mikael. 2000. Out of Africa: African influences in Atlantic Creoles. London: Battlebridge Publications. Parkvall, Mikael. 2008. The simplicity of creoles in a cross-linguistic perspective. In Matti Miestamo, Kaius Sinnemäki & Fred Karlsson (eds.), Language complexity: Typology, contact, change, 265–285. Amsterdam: John Benjamins. Plag, Ingo. 2002. On the role of grammaticalization in creolization. In Glenn Gilbert (ed.), Pidgin and creole linguistics in the 21st century, 229–246. New York: Peter Lang. Prescod, Paula. 2013. Vincentian Creole structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/7. Schultze-Berndt, Eva & Denise Angelo. 2013. Kriol structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/25. Schwegler, Armin. 2013. Palenquero structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/48. Seuren, Pieter & Herman Wekker. 1986. Semantic transparency as a factor in creole genesis. In Pieter Muysken & Norval Smith (eds.), Substrata versus universals in creole genesis, 57–70. Amsterdam: John Benjamins. Siegel, Jeff. 2008. The emergence of pidgin and creole languages. Oxford: Oxford University Press. Sippola, Eeva. 2013. Ternate Chabacano structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/44. Smith, Ian R. 2013. Sri Lanka Portuguese structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/41. Trudgill, Peter. 2011. Sociolinguistic typology: Social determinants of linguistic complexity. Oxford: Oxford University Press. van den Berg, Margot C. & Adrienne Bruyn. 2013. Early Sranan structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/1. van Sluijs, Robbert. 2013. Negerhollands structure dataset. In APiCS Online (Michaelis et al. 2013b). http://apics-online.info/contributions/27. Vitale, Anthony. 1981. Swahili syntax. Dordrecht: Foris. Yakpo, Kofi & Adrienne Bruyn. 2015. Trans-Atlantic patterns: The relexification of locative constructions in Sranan. In Pieter C. Muysken & Norval Smith (eds.), Surviving the Middle Passage: The West Africa-Surinam Sprachbund, 135–174. Berlin: De Gruyter.

Language index Abruzzese 190, 191, 209, 229 Afrikaans 106, 109, 110, 127, 128, 137, 138, 147, 1083, 1113 Agul 15, 71, 309–314, 316–322, 324, 326–330, 332–349, 352 Altaic languages 63, 364, 366, 385, 388, 399, 406, 409, 413, 420, 424, 425, 428 Archi 309–314, 316–318, 324, 326, 327, 331– 334, 336, 338, 339, 347, 348, 352 Austronesian 10, 56, 94, 135, 1007, 1037, 1043, 1045, 1047, 1049, 1051, 1053, 1056–1058, 1060–1062, 1069, 1071, 1096, 1119 Avestan 465–467, 469, 481 Awacatec 820 Awadhi 500, 501, 503, 506–508, 511, 512, 514, 515, 517, 519, 532, 536 Aymara 10, 25, 26, 977, 980, 982, 984–986, 990, 992, 993, 995, 998, 1000, 1001 Aymaran languages 977–979, 983–985, 987, 989, 995, 997, 1002 Bactrian 465, 467, 469–471, 474–476, 491, 492, 494 Balantak 1043, 1048, 1049, 1054, 1055, 1069 Balkan languages 176, 193, 249, 251–253, 255, 256, 266, 272, 276, 277, 284, 286, 290, 291, 294 Balochi 15, 63, 465–468, 470, 472, 473, 475– 480, 483–485, 489–491, 494 Bambara 26, 39, 695, 696, 699, 700, 703, 706–710, 712, 713, 715, 716, 719, 720, 723 Bantu 10, 769, 770, 773–777, 780–785, 792– 794, 796, 797, 1023, 1096, 1122, 1123 – Proto-Bantu 769, 770, 773–777, 781, 784– 786, 792, 796 Bardi 1079, 1080, 1084, 1086–1088, 1094– 1101 Bashkardi 465, 466, 479–481, 484 Batavia Creole 1112–1114 Beja 9, 10, 16, 19–21, 34, 71, 90, 659–668, 670–675, 679–682, 684–686, 689–691, 878 Bengali 500, 501, 506, 517, 523, 529 Bhojpuri 499–501, 503, 505–508, 510–512, 515–517, 523–525, 532, 536 Bikol 1046 Biloxi 904, 910, 911, 913, 914, 930–932, 934, 936 https://doi.org/10.1515/9783110712735-014

Bimin 1007–1010, 1012–1014, 1016, 1019, 1021, 1023, 1025, 1028–1030, 1032, 1033, 1036 Bislama 1112, 1113, 1115, 1120 Budugh 309, 312, 313, 320, 322, 323, 336, 347 Bulgarian 251, 253, 254, 258, 265, 266, 269, 277, 278, 280, 281, 283–285, 290, 295, 506 Catalan 62, 167, 170, 172, 174–180, 182, 184, 186, 187, 189, 190, 192, 193, 196, 197, 200, 201, 204–206, 208, 213, 215–219, 226, 227, 229–231 Catawba 904–906, 926, 927, 931 Caucasian Albanian 310, 313, 317, 322, 323, 327, 339, 351 Caucasian Tat 465, 478, 480, 486, 488, 489 Cebuano 59, 1065 Chabacano 1110, 1112, 1113, 1115, 1119 Chattisgarhi 499–501, 511, 512, 523, 536 Cherokee 24, 72, 943, 944, 953, 954, 957, 960–964, 974 Chinese 9, 10, 17, 21, 27–32, 34, 51, 94, 366, 399, 404, 406, 477, 575, 576, 579, 580, 601–603, 609–621, 624–626, 628, 633, 634, 637, 640, 642, 645–648, 650, 651, 688 – Classical Chinese 609, 618–620, 645, 646, 648–651 – Mandarin 17, 21, 27, 32, 51, 366, 404, 609, 610, 619–621, 628, 645, 647 – Middle Chinese 609, 619, 622, 623, 626– 628, 630–632, 636, 638–641, 643 – Modern Chinese 609, 613, 626, 629–633, 636, 639, 641, 646, 648 – Old Chinese 601, 609, 622, 623, 626–628, 632, 634–636, 638–644, 646 – Standard Chinese 609, 647 Chiwere 904, 908, 911, 913, 915, 920, 925, 926, 930, 931, 934, 936 Ch’ol 65, 816–819, 826, 827 Chorasmian 56, 465, 467, 471, 473, 482, 485, 486, 492, 494 creole languages 9, 10, 29, 30, 35, 69, 167, 226–228, 516, 533, 1025, 1109, 1110, 1116–1126 Croatian 250, 251, 255, 257, 258, 263, 264, 270, 272, 273, 277, 281, 283, 286, 287, 295, 1122

1130

Language index

Crow 904, 905, 913, 914, 916, 924, 930, 932, 934, 936 Cushitic languages 19, 34, 659, 661, 663, 665, 667, 670, 679–682, 685, 686, 689, 690 Czech 132, 133, 135, 153, 250, 251, 259, 263– 266, 270, 272, 273, 280, 282, 283, 291

Dakhini 501, 503, 507, 516, 517, 527–529, 536 Danish 105, 106, 109, 112, 115–118, 121–124, 126, 127, 130, 132, 133, 135, 148 Dravidian 34, 320, 428, 501, 502, 504, 507, 523, 525, 527–529, 532 Dutch 105, 107–110, 112, 120, 121, 123–135, 137, 147, 266, 1048, 1083, 1113

East and mainland Southeast Asian languages (EMSEA) 4, 12, 31, 32, 35, 49, 51, 53, 55, 58, 64, 67–69, 406 Emai 9, 10, 16, 17, 21, 23, 729–734, 736–738, 745–747, 749, 751, 756, 759, 762–764 English 4, 18, 25, 65, 70, 91, 92, 94, 96, 105, 106, 108–110, 112–116, 119–121, 123–131, 133–135, 137, 138, 141, 142, 144, 147, 149, 150, 152, 205, 220, 222, 249, 264, 289, 318, 346, 355, 366, 473, 487, 501, 514, 522, 523, 563, 576, 592, 596, 625, 634, 638, 642, 647, 706, 713, 718, 719, 742, 746, 748–750, 752, 755, 758, 773, 777, 784, 792, 905, 911, 917, 931, 932, 943, 949, 952, 1013, 1059, 1060, 1072, 1080, 1086, 1109–1111, 1113–1116, 1121–1124 – Old English 135, 136, 141, 142, 145, 366 Ewen 46, 64, 399–409, 411–423, 425–427 Ewenki 399, 401, 404, 409–414, 422

Faiwol 1007–1010, 1012–1014, 1016, 1019, 1023, 1025, 1028–1030, 1032–1034, 1036 Faroese 106, 107, 109, 111, 112, 116, 117, 120– 122, 133, 148, 152 Finnic 252, 272, 275, 276, 281, 282, 361–364, 366–369, 371–373, 375–377, 380, 383– 389, 393 – Proto-Finnic 367, 369, 376, 380 Finnish 275, 367, 368, 370–377, 380–390, 393, 394, 575 French 34, 77, 147, 149, 153, 166, 169, 170, 172–177, 179–185, 187–209, 211–213, 215– 220, 222–231, 270, 271, 346, 487, 565,

703, 704, 713, 721, 844, 930, 951, 952, 1109–1117, 1120, 1121, 1123–1125 – old French 134, 180, 211, 213, 951 Frisian 108, 109, 114, 121, 126, 127, 130, 131, 133–135, 148 – Fering 148, 149, 152 – Fering-Öömrang 114 Friulan 172, 193, 205, 209 Galician 167, 189, 195–197, 205, 213, 223 Garhwali 500, 501, 503, 508–512, 515, 519, 525, 529, 534, 536 German 10, 26, 30, 34, 65, 73, 93, 96, 105– 112, 115, 116, 118–121, 123–130, 132–141, 143–147, 149–153, 190, 211, 222, 250, 251, 258, 260, 263, 264, 266, 271, 273, 276, 282, 283, 285, 286, 291, 292, 294, 295, 469, 807, 844, 1120, 1121, 1124 – Alemannic 146, 147 – Bavarian 119 – Early New High German 127, 132, 145, 147, 149, 150 – Middle High German 108, 128, 149 – Middle Low German 118, 122 – New High German 138, 262 – Old High German 115, 118, 124, 135–137, 141, 142, 145, 146, 149, 150 – Standard German 119, 128, 146, 152 – Swiss German 131, 132 Germanic 9, 20, 31, 32, 34, 51, 56–58, 65, 67, 70, 72, 105–109, 111–116, 118, 120–124, 126–131, 133–136, 140–145, 147–149, 151– 153, 201, 218, 255, 266, 271, 275, 276, 280, 376, 473, 474, 970 – North Germanic 105, 107, 142, 148, 153 – Old Norse 72, 107, 108, 116, 117, 145, 148, 150 – Proto-Germanic 70, 105–109, 111, 114, 116, 121, 122, 127, 129, 135, 141, 145, 152 – West-Germanic 32, 107, 116, 118 Gothic 105, 106, 135, 142, 145 Guarijio 17, 21, 28, 853, 855, 856, 858, 859, 861, 862, 864, 865, 867, 868, 871, 873, 874, 876, 878, 879, 886, 887, 891, 894 Gyalrong languages 539, 542, 547–550, 552, 553, 556, 558, 559, 566–570 – proto-Gyalrong 548, 551, 556, 558, 567 Haitian Creole 167, 1110, 1114, 1119, 1123 Hindi 9, 10, 15, 16, 18, 20, 21, 26, 27, 29, 67, 75, 499–505, 507–531, 533, 534, 536

Language index

Hmong 31 Hoocąk 9, 10, 16, 18, 21, 26, 70, 71, 903–928, 930–936 Huastec 819, 820 Hungarian 218, 279, 361–364, 366, 367, 371, 373–378, 381, 383–388, 390, 392–394 Icelandic 72, 106–109, 111, 114–117, 120–122, 124–126, 129, 130, 132, 133, 137, 148, 149, 152, 153 Indo-Aryan languages 67, 75, 499, 502, 503, 505, 507 Indo-European languages 9, 10, 24, 69, 165, 249, 255, 259, 266, 295, 312, 345, 362, 363, 377, 465, 478, 486, 490, 499, 501, 970, 1007, 1117 – Proto-Indo-European 106, 107, 109, 111, 124, 135, 142, 252, 258, 269, 276, 295, 362, 363, 466, 472, 474, 488, 495 Iranian languages 9, 10, 16–19, 21, 26, 29, 33, 34, 56, 63, 327, 342, 465–479, 481–489, 491–494 Iroquoian 9, 69, 72, 943–945, 953, 957–960, 963, 965, 967, 970–972, 974 – Northern Iroquoian 72, 943, 944, 948, 949, 951, 953, 954, 959, 962, 965, 967, 970– 972, 974 – Proto-Iroquoian 943, 948, 953, 957, 960, 963, 964, 974 – Southern Iroquoian 72, 943, 944, 974 Italian 62, 96, 167, 168, 170–206, 208, 209, 212–232, 251, 255, 258, 278, 282, 285, 723, 1117, 1120, 1122 – old Italian 178, 190, 216 Itzá 803, 816, 826, 844 Jabirrjabirr 1080, 1082, 1084, 1087, 1088, 1097 Jacaltec 820, 849 Japanese 33, 69, 80, 222, 400, 408, 544, 575, 599, 601, 602, 679 Japhug 9, 10, 16, 21, 24, 25, 62, 539–542, 544, 545, 547–559, 561–570 Jukun 1080, 1087 Jula 695, 696, 701 Kalam 19, 1019, 1031 Karajarri 1088, 1099 Kashubian 262–264, 280, 286, 295 Kayardild 56, 57

1131

Ket 16, 17, 60, 63, 70, 72, 369, 409, 433–445, 447–461, 470, 849, 946 – Modern Ket 17, 440, 441, 445, 449 Khmer 10, 29–32, 51 Khotanese 465, 467, 468, 472, 473, 478, 481, 485, 494 Kita Maninka 39, 62, 695–697, 701, 710, 711, 713, 715, 721, 722 Korean 9, 10, 33, 34, 60, 63, 66, 69, 70, 80, 364, 400, 575–594, 596–604, 679 Korokan 696, 711, 713 Kott 433, 434, 436, 440, 442, 443, 445, 447– 450, 454, 455 Koyaga 39, 696, 711, 713–715 Kryz 309, 312, 313, 322–324, 326, 327, 341 Kumaoni 500, 501, 503, 509–511, 515, 516, 518, 519, 525, 530, 531, 534, 536 Kurdish 465, 469–471, 483, 494 – Kurmanji 465, 467, 471, 474, 476, 478, 479 – Sorani 18, 465, 467, 471, 472, 476–479, 494

Lacandón 803, 808, 816, 823, 824, 826, 836, 847, 848 Ladin 168, 172, 205, 211 Lakhota 910, 911, 913, 920, 924, 925, 930, 936 Laki 465, 477 Latin 96, 105, 133, 147, 149, 151, 165–176, 178–183, 185–187, 190–192, 199, 201, 205, 208–213, 215–220, 223, 226, 228, 229, 231, 474, 532, 774, 849, 951, 1013 Lauje 59, 1043, 1046, 1047, 1050, 1061, 1065 Lezgic languages 9, 10, 19, 21, 29, 30, 71, 72, 309–316, 319–321, 323, 325–328, 330– 333, 335, 336, 338–340, 342, 345–348, 350–353 – Proto-Lezgic 310, 313–317, 335, 338 Lithuanian 220, 272, 273, 282, 295 Luxembourgish 105, 108–112, 120, 121, 123– 128, 133, 134, 152, 153

Macedonian 251, 253–256, 258, 272, 277, 278, 280–282, 284–286, 290, 291, 295 Magahi 500, 503, 505, 506, 508, 517, 524, 525, 536 Maithili 34, 499, 500, 505, 517, 524, 536 Makassarese 1043, 1055, 1057, 1059, 1062, 1063, 1065, 1069 Malay 1046, 1051, 1056, 1067, 1069, 1070

1132

Language index

Malayo-Polynesian 10, 69, 1043, 1045, 1062, 1072 Mam 278, 821, 849 Manchu 10, 33, 399–402, 404, 406, 408–412, 414, 419, 420, 428 Manding languages 664, 695–697, 699–708, 710, 711, 713–717, 719–724 Mandinka 34, 695–698, 701–707, 709–719, 721, 722, 724 Maninka-mori 39, 695, 696, 703, 704, 710, 711 Mariic 361–363, 367, 371, 372, 378, 382, 387, 393 Marwari 500, 501, 503, 510, 517, 524, 536 Maukakan 696, 711 Mauritian Creole 1113 Mayan languages 65, 67, 803–806, 811, 814– 816, 818–821, 825, 841, 847–849, 985 – Proto-Maya 814, 816, 819, 830 Mazenderani 481 Mian 21, 22, 61, 405, 1007–1011, 1013–1030, 1032–1037, 1069 Mohawk 943, 944, 946–948, 951, 952, 955– 962, 964–974 Mongolian 33, 65, 71, 380, 402, 409, 415, 428 Mongolic languages 402, 419, 428, 433 Mori 10, 1043, 1049, 1058, 1059, 1067 Mountain Ok 10, 62, 63, 1007–1023, 1025, 1026, 1028–1033, 1036, 1037 Muna 1002, 1043, 1048, 1049, 1055 Mundari 34, 524 Munji 465, 468, 485

Na-Dene 433, 436, 443–446, 454, 460 Nahuatl 867, 875, 881, 887, 888 – Classical Nahuatl 881 Nanai 46, 57, 64, 65, 399–401, 403, 404, 411– 419, 424, 426 Neapolitan 171, 173, 181, 184, 189, 194, 209, 214 Negidal 399, 401, 406, 416 Nenets 25, 366, 375–378, 381–383, 385–392, 394 Névome 17 Nimanburru 1086, 1097 non-Pama-Nyungan 1077, 1078, 1081, 1093, 1099 Northern Tepehuan 17, 853, 855–858, 860– 863, 865–867, 873, 876, 877, 879, 885, 886, 890, 892, 893, 898

Norwegian 105, 106, 110, 116–118, 123, 127, 130, 131, 133, 135 – Bokmål 106, 109, 110, 122, 123, 135 – Nynorsk 106, 109, 122 Nyangumarta 1089 Nyikina 1080, 1082–1085, 1087, 1088, 1090, 1092, 1093, 1096–1098, 1100, 1101 Nyulnyul 1079, 1080, 1082, 1084, 1086–1091, 1095–1102 Nyulnyulan 10, 23, 24, 58, 62, 71, 1077–1080, 1082, 1083, 1085–1087, 1089, 1090, 1092–1094, 1097–1100, 1102–1105 – Eastern Nyulnyulan 1080, 1082–1085, 1087, 1089, 1090, 1093–1096, 1098, 1101, 1105 – pre-pNN 24, 1077, 1083, 1084, 1089, 1090, 1093, 1102, 1103 – proto-Nyulnyulan 1077, 1079, 1081, 1083– 1085, 1090, 1094, 1098, 1101, 1104, 1105 – Western Nyulnyulan 1080–1082, 1084, 1087–1090, 1095, 1100, 1101, 1105 Occitan 167, 169, 172, 174–177, 179, 190, 192, 193, 209, 211, 226–229, 232 Old Church Slavonic 249, 253, 255, 259–261, 270, 271, 278, 283, 285, 295 old French 180, 182, 211, 226, 227 Onondaga 944, 962, 965, 971 O’otam 855, 856, 863 Orok 399, 415 Ossetic 337, 465, 467, 468, 471, 472, 474, 477–479, 482, 484, 494 Palenquero 1112, 1114 Pama-Nyungan 1083, 1088–1090, 1099 Parthian 465, 467, 469, 472, 474–476, 478, 485, 486, 491, 494 Pashto 465, 467, 471, 474–479, 494 Persian 18, 34, 311, 328, 339, 345, 353, 465, 468, 470–472, 476, 478–480, 482–485, 487, 489, 492, 494, 501, 505, 506, 526 – Early Judeo-Persian 479, 487 – Middle Persian 465, 467–469, 471, 472, 474, 476, 478, 479, 482–484, 486–489 – New Persian 467, 469–473, 478, 479, 483, 487, 489, 490 – Old Persian 465–467, 485 – Tajiki Persian 482 Pima 853, 855, 891 Pima Bajo 17, 21, 853, 855–858, 860–862, 864–867, 871, 872, 874–880, 882, 883, 885, 886, 888–890, 892, 893, 895, 898

Language index

Pima Bajo Bajo 25 Polish 250, 251, 256, 258–267, 269–273, 276, 278–280, 282, 284, 285, 287, 288, 291– 295 Portuguese 62, 147, 167, 170, 172–178, 181, 183, 185–188, 190, 192–197, 201, 203, 206, 211, 215–219, 223–227, 230–232, 1109, 1111–1115, 1125 – Brazilian Portuguese 173, 176, 177, 181, 184, 185, 190, 196, 198, 202, 203, 205, 206, 212, 217, 223, 226, 227, 229, 231 – European Portuguese 177, 183–185, 189, 190, 196, 197, 205, 206, 223 Principense 1114, 1123, 1125 Q'eqchi' 815 Quechuan 54, 977–998, 1000–1003 – Ayacucho Quechua 980, 983, 984, 986, 987, 994, 997–1001 – Cuzco Quechua 990, 996, 998 – Ecuadorian Quichua 66, 71, 1000, 1001 – Huanca 987–990, 995, 996, 999 – Northern Junín Quechua 979, 990, 996, 999 – Pacaraos Quechua 980, 982, 988, 989, 991, 996, 999 – Proto-Quechua 977, 978, 986, 993 – Quechua I 978, 979, 981, 984–987, 990, 991, 993–996, 999 – Quechua II 978, 981, 986, 987, 990, 993– 996 – Tarma Quechua 979, 982, 999, 1002 Ræto-Romance 168, 186, 189 Rajasthani 500, 505, 508, 509 Ratahan 1043, 1050, 1058, 1067, 1068 Romance 10, 21, 32, 34, 51, 60, 62, 96, 105, 114, 134, 147, 149, 153, 165–181, 183–185, 187, 188, 190–192, 194–196, 198–205, 207–213, 215–220, 222, 223, 225–231, 255, 266, 278, 280, 287, 473, 951, 974, 1013, 1115, 1117, 1120 – Ibero-Romance 168, 173, 174, 182, 189, 193, 195, 197, 212, 215, 1109 – Medieval Romance 168 – northern Romance 168 – old Romance 211, 213 – southern Romance 168 Romanian 62, 144, 167, 170, 171, 173–183, 185–190, 193, 195, 198–201, 205, 206, 208, 209, 213, 216–218, 226–230, 232

1133

Romansh 168, 169, 172, 190, 193, 205, 211 Russian 93, 249, 250, 252, 259–263, 265, 266, 269, 271–275, 278, 280, 281, 283, 287– 290, 294, 295, 309, 313, 347, 433, 442, 453, 456, 458–461, 479, 774 Rutul 309, 312–314, 323, 324, 327, 330, 331, 336, 338, 342, 343, 345–347 Saamic 361–365, 367–369, 371, 372, 375–377, 379, 382–386, 392, 393 Sadani 501, 506, 516, 523, 536 Samoyedic 25, 361–364, 366, 368, 369, 371, 373, 375–379, 381–388, 390, 392, 393, 406, 433 – Proto-Samoyedic 368, 369, 377, 386 Sanskrit 27, 499, 501, 502, 504, 506–518, 521– 523, 525, 526, 531, 532, 534, 536 Sardinian 167, 168, 172, 174, 188, 190, 193, 194, 198, 203, 205, 211, 213 Semnani 465, 467, 475, 476 Seneca 944, 965, 971 Serbian 251, 253, 255, 258, 263, 264, 272, 273, 277, 281, 283, 286, 287, 295 Seri 895 Seychelles Creole 1110–1116, 1122–1124 Sicilian 188, 194, 208, 228 Sinitic 9, 32, 68, 69, 366, 399, 406, 409, 428, 609, 610, 614, 645, 647 Sino-Tibetan 9, 501, 506, 539, 547, 551, 553, 568, 569 Siouan, Siouan languages 903–908, 910, 911, 914, 916, 920, 921, 923–926, 928, 932, 934–937, 939 – Proto-Siouan 903, 906–908, 910, 911, 913– 916, 920, 923, 924, 926, 927, 930, 934– 936 Situ 548–550, 552, 558, 559, 561, 567, 568, 647, 821 Slavic 10, 23, 30, 51, 58, 70, 77, 145, 147, 153, 184, 195, 201, 218, 249–252, 255–272, 274–276, 278–287, 289–291, 293–295, 337, 479 – Balkan Slavic 251–253, 255, 256, 266, 272, 277, 284, 290, 291, 294 – Common Slavic 249–252, 259, 266–269, 271, 275, 276, 278, 279, 281, 285, 290, 291, 295 – East Slavic 249–252, 256, 259, 265, 269, 271, 278–280, 282, 283, 289 – North Slavic 251, 259, 276, 278

1134

Language index

– South Slavic 249–251, 253, 255, 257, 259, 264, 269, 276, 278–280, 283–286, 294 – Western Slavic 34 Slovak 263, 264, 273, 282, 291, 295 Slovene 251, 259, 264–266, 272, 276, 281– 285, 295 Sogdian 18, 19, 465, 467, 469, 471–475, 477, 478, 481–483, 485, 488, 489, 491, 494 Soninke 700, 703–707, 710, 714, 715, 721 Sorbian 250, 256, 257, 262–264, 285, 286, 290, 294, 295 Southeastern Tepehuan 853, 855–858, 860– 862, 866, 879–881, 883, 888, 889, 891, 892 Southern Min 32, 647, 648 Spanish 34, 93, 135, 167, 170, 173–178, 181– 183, 185–187, 189, 190, 192–199, 201– 203, 205, 206, 208–213, 215–221, 223– 230, 232, 625, 809–811, 835, 845, 848, 850, 857, 879, 882, 887, 893, 984, 987, 988, 991, 994, 1013, 1109, 1110, 1112, 1114, 1115, 1117 Sranan 1110, 1111, 1114, 1119, 1121, 1123, 1124 Standard Average European (SAE) 31, 145, 148, 152, 218, 228, 250, 651 Sulawesi 10, 21, 27, 56, 59, 66, 69, 72, 1043– 1064, 1066–1069, 1071, 1072 Swedish 112, 113, 115–117, 123, 132, 149, 153 – Middle Swedish 113 – Old Swedish 112, 113 Tabasaran 309, 311–315, 318, 319, 330, 335– 343, 345, 346 Tagalog 1050, 1051, 1053, 1059, 1061 Tajio 1043, 1047–1054, 1060, 1065, 1069 Taleshi 465, 468, 475, 480 Tamil 502, 512, 528 Taracahitan languages 858, 862, 883 Tarahumara 17, 24, 853, 855–863, 865, 867– 869, 874, 876–878, 881, 883–888, 891, 892, 894, 898 Telefol 1007–1016, 1019, 1020, 1023, 1024, 1026, 1028–1030, 1032, 1033 Tepiman languages 858, 862, 864, 865, 879, 888, 892 Thai 10, 29–32, 51, 511, 615 Tifal 1007–1017, 1019, 1021, 1023, 1024, 1026, 1028–1030, 1032, 1033, 1036 Tohono O’odham 17, 853, 855, 856, 860, 867, 876, 879, 881–883, 886, 890, 891

Tok Pisin 1111–1116, 1120 Tolaki 1043, 1054, 1055, 1059, 1066, 1069 Tondano 1043, 1050, 1052 Totoli 1043, 1044, 1046, 1058, 1067–1069 Tsakhur 309, 311–315, 317, 318, 321, 322, 327, 328, 336, 338, 339, 343–345 Tshobdun 548–551, 561, 566–568 Tswana 10, 15, 21, 22, 30, 769, 770, 772–777, 780–786, 792, 793, 796, 797 Tukang Besi 1043, 1051, 1054, 1064 Tungusic languages 46, 64, 399–401, 403– 406, 410, 412–416, 418, 420, 422, 424, 425 – East-Tungusic languages 46, 79, 399–401, 416, 418, 419 – North-Tungusic languages 46, 399–401, 416, 419 Turkic 18, 33, 312, 352, 363, 364, 367, 372, 374, 376, 385, 388, 400, 433, 441, 448, 457, 460, 477, 487, 601 Turkish 33, 275, 283, 287, 477, 488 Tuscarora 944, 949–951, 953, 959, 964, 965, 971, 973 Tzotzil 803, 816, 818, 821 Tz'utujil 65, 67, 806, 818, 819 Udi 310–316, 318–320, 322, 323, 325–330, 336, 339–341, 345, 347–353 Udihe 399, 406, 414, 415 Uralic 369, 373, 375, 382, 383 – Pre-Proto-Uralic 361, 366, 369, 373, 375, 378, 386, 391–393 – Proto-Uralic 362, 368, 369, 378, 386 Urdu 9, 499, 501–503, 505–507, 509, 512–518, 521–523, 525–528, 530, 531, 536 Usan 1031, 1032 Ute 719, 857, 884 Uto-Aztecan 10, 17, 24, 853–858, 861, 866, 869, 870, 873 – Proto Uto-Aztecan 857, 858, 862–869, 871– 875 Vafsi 465, 475, 476 Vietnamese 31 Wakhi 465, 468, 472, 477 Warihío 853, 856, 858, 859, 862, 864, 865, 868, 871, 874, 876, 878, 879, 887, 891, 894 Warihío del Río 873, 874

Language index

Warlpiri 1087, 1089 Warrwa 1080, 1081, 1083–1085, 1087–1092, 1094–1102 Western Tarahumara 868, 869 Worrorran 1083, 1088, 1090, 1093, 1103 Xasonga 696, 705 Yaghnobi 18, 19, 465, 467, 472, 475, 481, 482, 485, 491, 492 Yaqui 18, 853, 855–859, 861–863, 865, 867, 868, 870–872, 874, 876–878, 880, 883– 886, 889–891, 894, 898 Yawuru 15, 1079, 1080, 1083, 1084, 1087, 1088, 1090, 1092, 1095, 1096, 1098, 1100, 1101, 1104 Yeniseian 10, 17, 433–436, 438–446, 448, 453–456, 459–462 – Pre-Proto-Yeniseian 448

1135

– Proto-Yeniseian 17, 433, 436, 440, 441, 443, 445, 447, 448, 454, 456, 460 Yiddish 109, 111, 128, 130, 133, 137, 139, 140, 143, 146, 147, 149, 152, 153 Yucatec Maya 10, 16–18, 29, 30, 66, 67, 678, 803, 805, 806, 809–811, 814, 816, 848, 985 – Colonial Yucatec 803, 805–807, 809–813, 823, 824, 829–833, 835, 836, 838, 839, 842, 843, 845 – Modern Yucatec 16, 18, 803, 806–813, 815, 817, 818, 822–827, 831–837, 839–844, 846, 847, 849 Yucatecan languages 803, 806, 808, 823, 826, 828, 829, 834, 839, 847, 849 Yugh 433, 434, 436, 442–448, 450, 451, 455, 456, 461 Zazaki 465, 467, 470, 472, 475, 476

Subject index absolutive 310, 311, 315, 316, 318, 319, 324, 325, 328, 341, 346, 348, 349, 355, 819, 850, 867, 869, 924, 1030 accelerated functionalization 1109, 1116, 1125 accusative marker 25, 94, 120, 121, 178, 179, 181, 183–209, 223–225, 260, 286, 383, 385, 387, 391–394, 509, 510, 871, 890 action nominal 63, 271, 312, 322, 339, 342, 354, 380, 442, 443, 450, 451, 453, 456– 458, 460–462, 509, 525, 553–556, 670, 676, 677, 691 adessive 252, 263, 316, 321, 394, 438–440, 462 agglutination 37, 38, 46, 56, 70, 90, 93, 115, 250, 251, 259, 278, 279, 284, 294, 309, 310, 399, 438, 460–462, 469, 576, 661, 806, 821, 822, 827, 849, 905, 979, 981, 1003, 1078, 1109 agreement 25, 34, 63, 67, 92, 93, 95, 110, 122, 124, 144, 167–169, 171, 177, 178, 180, 184, 188, 189, 194, 204–211, 249, 255, 260, 272, 277, 281, 284, 294, 310, 311, 313, 314, 317, 324, 340–342, 345, 346, 348, 350–353, 372, 375, 379, 404, 406, 414, 416–418, 420–422, 425–427, 435, 436, 442–444, 446–449, 451, 453, 454, 460– 462, 475, 476, 488, 492, 503, 505, 507, 515–518, 522, 524, 539, 545, 566, 596, 603, 684, 697, 715, 725, 770, 771, 781, 782, 792, 803, 880, 881, 892, 897, 898, 980, 981, 984, 1007, 1009, 1016–1018, 1026–1031, 1036, 1037, 1062, 1064, 1103 aktionsart 381, 409, 466, 472, 478, 479, 841, 903, 1090, 1104 alienability 114, 862 alignment 252, 310, 311, 324, 448, 452–454, 461, 505, 516, 517, 547, 771, 816, 828, 905, 1030, 1044, 1064 allomorphic variability 262 allomorphy 36, 38–40, 42–48, 51, 55, 56, 58, 78, 89, 96, 102, 249, 268, 403, 414, 826, 923, 944, 965, 968 analogical extension 69, 77, 266, 290, 1045 analogy-driven grammaticalization 65 anticausative 122, 185, 187, 250, 261, 262, 323, 453, 699, 707, 725 antipassive 323, 552–554, 569, 709, 710, 725, 726, 784, 850, 905 https://doi.org/10.1515/9783110712735-015

antiresultative 290, 589 applicative 15, 75, 259, 452, 552, 554–556, 667, 681, 731, 732, 738, 766, 770, 772, 785, 790, 799, 800, 898, 906, 920–923, 936, 938, 957–963, 974, 975, 981, 992, 1045, 1066, 1067, 1072, 1077, 1093–1096, 1098, 1103–1105 apprehensive 594, 595, 603, 795 areality 1, 3, 4, 7, 8, 10–12, 20, 24, 28–35, 49–55, 57, 64, 67–69, 76–78, 80, 134, 250, 255, 262, 265, 269, 275, 276, 279, 282, 284, 294, 295, 353, 362–364, 366, 393, 400, 401, 404, 406, 409, 427, 433, 436, 454, 456, 457, 462, 477, 575, 601, 666, 682, 690, 769, 821, 822, 895, 964, 1103 – areal clustering 28, 30, 76–78, 255, 272 – areal effects 8, 12, 29–31, 35 – areal feature 24, 31, 68, 404, 666, 682, 769 – areal pattern 32, 67, 80, 353, 456 – areality in scenarios of grammaticalization 49 – grammaticalization areas 32, 118 article 18, 31–34, 56–58, 66, 72, 73, 76, 92, 94, 107, 109, 111, 112, 114–120, 126, 128, 139, 145, 149, 152, 153, 168, 169, 171, 174–177, 180–182, 228, 251, 254–258, 293, 316–318, 352, 393, 471, 492, 494, 503, 506, 576, 663, 664, 672, 673, 678, 702, 731, 809–811, 848, 850, 876, 903, 908–911, 913, 915, 918, 920, 935, 949– 952, 973, 974, 987, 988, 1009–1014, 1016, 1025, 1036, 1037, 1045, 1048–1056, 1071, 1110, 1111, 1120 aspect 18–20, 23, 32, 34, 51, 58, 63, 93–95, 111, 122, 124–128, 185–188, 191–194, 196, 206–208, 214, 215, 230, 252, 262–270, 272, 273, 275, 276, 278, 292, 295, 311, 312, 318, 323, 325, 329, 337, 344, 346, 352, 364, 371, 378, 379, 381, 382, 407, 409–411, 413–415, 428, 443–445, 447, 449–451, 461, 465, 466, 475, 478–480, 492, 501, 504, 516, 518–521, 525, 552, 557, 576, 578, 587–589, 599, 609, 620–634, 645, 651, 652, 666, 668, 682, 690, 710, 711, 715, 725, 729, 730, 733, 759, 762, 784, 806, 816, 822, 825–835, 837, 838, 841, 842, 846, 849, 861, 881–886, 897,

1138

Subject index

903, 906, 917, 919, 927, 928, 933–935, 939, 944, 945, 953, 957, 958, 960, 964, 972, 981, 992–994, 996, 1007, 1009, 1018–1021, 1024, 1032, 1033, 1036, 1037, 1046, 1058–1060, 1071, 1080, 1104, 1105, 1114, 1115, 1120 associated motion 551, 568, 569, 721, 1001 associative plural 680, 700, 701, 724, 774, 797, 1048, 1071 augmentative 706, 724, 784, 798, 948, 974 auxiliation 35, 56, 57, 70, 263, 270, 273, 282, 294, 329, 352, 411, 519, 828, 837, 840, 849 avertive 97, 100, 101, 588–590, 603 backformation 557, 568 Balkanism 291 benefactive 15, 31, 33, 75, 99, 124, 403, 645, 646, 667, 681, 682, 687–690, 705, 711, 713, 724, 726, 785, 809, 812, 906, 922, 923, 938, 957, 958, 974, 990, 1023–1026, 1036–1038, 1067, 1095 bi-clausal 631, 669, 671, 683, 777 Bondedness (BD) 2, 4, 5, 35–48, 51, 54, 56– 58, 62, 66–70, 72, 73, 78, 79, 89, 90, 93, 102, 115, 255, 294 borrowing 149, 352, 373, 433, 436, 450, 458, 460, 461, 501, 506, 512, 567, 579, 580, 601, 663, 681, 721, 986, 989, 1081, 1083, 1093, 1102, 1104, 1120, 1121 – MAT-borrowing 252, 263 – PAT-borrowing 262, 263, 265, 270, 271, 283, 286 bracketing construction 138–140 calquing 893, 1109, 1120–1125 canonical typology 79 case marker 16, 25, 57, 62, 70, 167, 181, 182, 209, 223, 224, 310, 330, 338, 339, 366, 372, 375, 388, 392, 393, 401, 403, 404, 407, 423, 424, 460, 467–469, 475, 492, 508, 510, 512, 539, 567, 576, 583, 602, 675, 774, 856, 868–872, 875, 884, 889, 896, 907, 978, 980, 989–991, 995, 1000, 1044, 1045, 1050, 1077, 1079, 1081–1083, 1093, 1096, 1103, 1105 case relators 811 causative 31, 32, 98, 99, 101, 169, 223–225, 230, 251, 259, 264–266, 283, 294, 321– 323, 334, 350, 352, 353, 371, 373, 386,

389, 395, 407–409, 428, 451, 453, 472, 493, 501, 513, 514, 552, 554, 555, 561, 569, 586, 600, 603, 622, 626, 637, 641, 646, 662, 666, 691, 699, 707, 708, 710, 716, 725, 726, 740, 741, 770, 772, 785, 800, 850, 876–878, 896, 897, 961–963, 981, 982, 992, 1067, 1116 causative construction 223–225, 321, 708, 725, 1116 chunking 57, 73, 79 cislocative 24, 72, 551, 552, 559, 570, 963, 974, 975 classifier 21, 27, 28, 97, 100, 101, 109–111, 405, 406, 445, 446, 493, 506, 534, 578– 580, 601–603, 609–617, 651, 806–808, 811, 848, 858–861, 898, 1038, 1045, 1047, 1048, 1071, 1072, 1085, 1089, 1091 – numeral classifier 92, 405, 406, 579, 601, 609–617, 641, 651, 806–808, 811, 848, 850, 1045, 1047, 1048, 1071 – possessive classifier 21, 28, 858–861 clitic 16, 34, 37, 38, 56, 59, 65–67, 70, 73, 93, 106, 107, 112, 113, 115, 116, 120, 122, 129, 130, 135, 139, 152, 168, 169, 176, 177, 179, 181, 183, 184, 204, 206, 208–211, 214, 225, 229, 250, 251, 253–256, 278, 279, 281, 284, 285, 319, 320, 328, 340, 341, 345, 348, 350–353, 355, 378, 379, 383, 393, 427, 437, 440, 448, 451, 452, 457, 461, 470, 471, 474–477, 484, 487–489, 491, 492, 494, 495, 531, 539, 567, 663, 696, 697, 773, 781, 793, 808, 814, 816– 818, 820, 822, 823, 831, 833, 841, 844, 846, 848, 850, 873, 880, 881, 893, 903, 914, 916, 918, 924, 949, 974, 981, 983, 992, 1011, 1012, 1025, 1044, 1045, 1049, 1055, 1057, 1058, 1060–1063, 1065–1067, 1071, 1079, 1096, 1117 – clitic doubling 208–210, 224, 229, 251, 284, 294 coalescence 35, 65, 69, 152, 198, 201, 218, 219, 223–225, 251, 259, 282, 283, 294, 443, 557, 816, 1021–1023, 1109, 1117, 1125 cognitive factors 69 comitative 24, 32, 98, 101, 173, 178, 217, 311, 315, 316, 353, 355, 393, 395, 403, 404, 439, 440, 452, 462, 510, 512, 535, 539, 542, 543, 569, 686, 687, 699, 700, 705, 719, 773, 776, 777, 784, 844, 845, 849, 989, 990, 1081, 1094–1096, 1098, 1104, 1105, 1122

Subject index

comparative 32, 60, 148, 256–258, 313, 339, 354, 374, 376, 395, 404, 533, 543, 544, 582, 599, 645, 675, 691, 780, 1060 competing motivations 11, 55, 68 complement clause 64, 140–142, 199, 202, 212, 213, 218, 269, 286, 389, 390, 423, 425, 487, 489, 562, 597, 669, 672, 673, 677, 684, 687, 717, 793, 828, 829, 831, 889, 890, 892, 916, 972, 973, 981, 1069 complementation 168, 213–215, 285–287, 293, 334, 458, 459, 716, 828, 892 complementizer/complementiser 19, 31, 32, 34, 39, 62, 66, 97, 100, 101, 138, 141–143, 149, 150, 168, 193, 199, 200, 202, 210, 212–216, 218, 220, 222, 229, 269, 285– 288, 294, 342, 343, 352, 354, 355, 457– 459, 487, 493, 494, 505, 526, 528–530, 533, 535, 593, 597, 600, 603, 647, 648, 651, 673, 684, 687–689, 717, 725, 793, 794, 799, 884, 889, 891, 892, 897, 898, 973, 1063, 1072, 1089, 1096, 1103, 1104, 1123, 1124 completive 33, 66, 98, 99, 101, 479, 588, 631– 633, 651, 698, 703, 710, 711, 713, 715, 725, 726, 816, 824, 825, 827, 833–836, 842, 850, 861, 882, 898, 983, 993, 999, 1019, 1021, 1036, 1060, 1071, 1072, 1115 compound verb construction (CVC) 1077, 1080, 1089–1091, 1103 conjugation class 413, 414, 1091–1093 conjugation marker 57, 414, 444, 446, 1093, 1105 conjunction 15, 62, 94, 97, 126, 134, 139, 143, 148–150, 152, 153, 187, 193, 198, 210, 214–216, 218–220, 222, 274, 285, 312, 342, 352, 419, 459, 460, 486, 487, 527, 528, 531, 536, 562, 563, 609, 628, 647, 649, 674, 718–721, 723–726, 733, 734, 736, 755, 765, 766, 793, 795, 796, 799, 800, 808, 812, 843–846, 848, 850, 892, 1069 construct form 780–782 construction marker 25, 38, 76, 91 constructionalization 3, 4, 55, 58 contact-induced degrammaticalization 63 contact-induced grammaticalization 3, 63, 218, 252, 282, 718, 1121 converb 33, 60, 71, 312, 316, 324–329, 331, 333, 338, 339, 342–344, 353–355, 372, 381–383, 385, 391, 395, 399, 404, 409–

1139

412, 414, 416, 417, 420, 422–425, 427– 429, 491, 566, 576, 583, 586–589, 593, 601, 603, 662, 667–673, 675, 677, 678, 681–684, 687, 691, 716, 725 convergence 27, 33, 75, 147, 260, 295, 351, 406, 510, 680, 690, 977, 978, 981, 1015 convergent developments 20, 76 coordination 149, 459, 525, 529, 623, 691, 700, 705, 719, 773, 796, 797, 800, 845, 850 copula 13–15, 24, 26, 33, 66, 71, 74, 75, 97, 100, 101, 122, 123, 130, 132, 134, 138, 187, 194, 205, 252, 262, 270, 283, 284, 290, 311, 318, 319, 324–328, 342, 344, 345, 352–355, 369, 376, 377, 379, 383, 395, 410, 411, 422, 426, 447, 455, 456, 473, 474, 477–482, 493, 494, 514, 515, 517, 518, 520, 523, 526, 535, 562, 594, 598, 603, 609, 642–644, 648, 651, 652, 662, 668, 669, 678, 682, 683, 686, 687, 690, 691, 699, 709, 710, 716, 722, 723, 725, 726, 734, 736, 765, 766, 777–779, 792, 797, 841, 842, 850, 917, 928, 930, 935, 995, 997, 998, 1083, 1099, 1100, 1102, 1103 coverb 323, 346–349, 354, 404, 525, 529, 535, 609, 638, 641, 645, 651 covert grammaticalization 44, 55, 57, 59, 62 creolization 1109, 1110, 1118 culture-specific factors 27 cycle 18, 33, 77, 118, 127, 226–228, 230, 373, 374, 394, 424, 702, 1018, 1109, 1117, 1118 dative marker 75, 181, 403, 509, 511, 512, 938, 1013, 1051, 1116 decategorialization 5, 141, 142, 203, 228, 229, 289, 853, 969, 1025 declension 108, 109, 112, 117, 169, 170, 172, 252, 259, 292, 293, 372, 381, 384, 385, 393, 504, 806 definite article 58, 73, 92, 107, 109, 111, 114– 119, 145, 149, 174, 180, 181, 254–256, 317, 471, 503, 506, 663, 664, 702, 731, 810, 811, 908–911, 913, 915, 918, 920, 935, 949, 951, 987, 988, 1012–1014, 1016, 1072, 1110, 1120 definiteness 95, 117, 118, 168, 173–175, 182, 189, 228, 229, 231, 256, 259, 393, 469, 672, 702, 724, 774, 809, 810, 987–989, 1011, 1050, 1052, 1053, 1056

1140

Subject index

degrammaticalization 7, 11, 62, 63, 109, 112, 113, 191, 223, 225, 228–230, 494, 539, 541, 567 demonstrative 13, 14, 20, 21, 24, 26, 62, 92, 97, 99, 115–119, 142, 145, 149, 152, 174– 176, 189, 228, 256, 317, 339, 353, 369, 375, 457, 459, 468, 471, 477, 494, 547, 578, 581, 585, 602, 634, 643, 644, 648, 649, 652, 663, 664, 672, 673, 680, 687, 690, 698, 701, 702, 717, 723–726, 730, 731, 775, 782, 793, 794, 798, 800, 810– 814, 822, 823, 834, 848, 850, 856, 872– 875, 889, 891, 892, 895, 896, 898, 903, 908–920, 926, 928, 930–932, 935, 936, 951, 952, 973, 974, 1010, 1011, 1013–1016, 1036–1038, 1048–1050, 1056, 1071, 1093, 1103, 1110, 1120 derivation 23, 51, 122, 179, 250, 266–270, 276, 294, 295, 310, 312, 316, 321–323, 334, 336, 337, 346, 347, 349, 351, 352, 364, 371, 373–376, 379, 381, 386, 389, 390, 403, 409, 414, 442, 460, 471, 501, 506, 539, 542, 552–556, 561, 569, 578, 661, 662, 666, 682, 696, 699, 706, 711, 724, 725, 772, 773, 777, 783–785, 798, 841, 842, 844, 906–908, 920, 921, 923, 944, 957, 959, 960, 967, 980–982, 992, 993, 995, 996, 1001–1003, 1059, 1067, 1077, 1086–1088 desemanticization 4, 5, 42, 51, 57, 58, 89, 110, 142, 182, 198, 203, 412, 840, 849, 867, 1002, 1021, 1025 determiner 20, 21, 26, 32, 69, 73, 118, 169, 172, 174–178, 181, 212, 216, 230, 258, 316, 317, 353, 503, 504, 527, 546, 577, 578, 581, 661, 672, 683, 697, 702, 703, 724, 726, 731, 735, 762, 810, 812, 814, 816, 817, 822, 848, 856, 868, 871–875, 896, 898, 903, 906, 908–910, 914, 918, 919, 984, 987, 988, 1007, 1009, 1011, 1045, 1049, 1050, 1081, 1082, 1100, 1103 differential object marking 94, 393, 502 differential subject marking 502 diminutive 63, 97, 250, 471, 516, 535, 706, 725, 773, 783, 798, 800, 908, 935, 936 directional 626, 849, 1045, 1067, 1068, 1071 distributive 319, 321, 409, 602, 675, 691, 703, 766, 898, 947, 948, 952, 974, 991 ditransitive 123, 139, 187, 260, 265, 541, 561, 646, 731, 757, 1009, 1095

ditropic clitics – clitic 73, 437, 448, 452, 461 donor language 132, 1120, 1121, 1123, 1124 double determination 32, 117, 118 drift 8, 33, 34, 68, 149, 192, 195, 196, 1117 dummy noun 25, 63, 406, 672, 673, 678, 683– 686 egophoric 21, 25, 558–560, 569, 570 enclitic 18, 46, 57–59, 63, 64, 73, 178, 205, 250, 253–256, 259, 277, 278, 284, 294, 317–319, 326, 328, 332, 340–342, 350, 353, 372, 402, 403, 417–419, 437, 438, 447, 448, 451, 452, 470, 488, 491, 495, 567, 661–664, 674–676, 679, 810, 813, 814, 816, 818–821, 824, 830, 831, 835, 841, 842, 847, 903, 905–908, 910, 920, 922, 927–933, 947, 948, 974, 1011, 1049, 1055, 1060–1063, 1065, 1066, 1069, 1083, 1089, 1093, 1094 endocentric grammaticalization 25, 76 ergative 21, 26, 27, 310, 311, 315, 316, 319, 321, 324, 330, 334, 339–341, 355, 384, 386–388, 391, 392, 467, 474–476, 478, 492, 502, 505, 509–512, 515, 524, 527, 535, 543, 544, 547, 561, 562, 564, 567, 816, 817, 819, 849, 1044, 1063, 1079, 1081, 1082, 1097, 1105 erosion 5, 42, 66, 67, 70, 152, 294, 440, 446, 447, 503, 597, 600, 753, 754, 808, 812, 822, 853, 872, 1011–1013, 1023, 1025, 1026 evidentiality 25, 100, 201–203, 220, 251, 281, 284, 287, 291, 294, 312, 326, 329, 331, 332, 352, 355, 415, 416, 429, 516, 529, 535, 557, 558, 560, 570, 592, 593, 603, 878, 888, 889, 897, 898, 983, 992, 1000, 1085 – hearsay 19, 284, 286, 294, 326, 332, 333, 342, 352, 354, 529, 530, 592 – inferential 68, 287, 290, 331, 557, 558, 592– 594, 602, 936 – reportive 202, 272, 273, 284, 287, 290, 291 exaptation 62, 70, 77, 114, 290, 293 exocentric grammaticalization 76, 77 expansion-based approach 56 expletive construction 136 factitive 261, 265, 266, 371, 395 feminine 62, 109–114, 120, 169–172, 178, 208, 228, 314, 317, 355, 435, 436, 451, 460,

Subject index

462, 469, 494, 504, 515, 521, 661, 663, 673, 678, 691, 783, 784, 798, 951, 1011, 1015–1017, 1028, 1038 form-function covariation 8, 68, 72 functionalization 35, 56, 57, 69, 137, 1109, 1116–1119, 1122, 1123, 1125 future 4, 15, 18–21, 24, 25, 27, 39, 65, 71, 72, 74, 75, 91, 97, 100, 101, 106, 127–134, 139, 151–153, 186, 192, 193, 196, 198, 199, 202, 203, 219, 229, 262, 265, 275, 276, 278, 279, 282, 289, 290, 294, 320, 325, 327–331, 353–355, 363, 377, 381, 383, 385, 410, 413, 414, 416, 418, 425, 426, 447, 479, 482, 483, 486, 493, 494, 504, 514, 516–518, 521, 525, 526, 535, 593, 595, 600, 603, 620, 635–637, 648, 651, 652, 669, 670, 683, 687–691, 710, 715, 725, 757, 764, 785, 789, 790, 798–800, 833, 834, 836, 837, 840, 849, 850, 875, 881, 882, 886–888, 897, 898, 920, 928, 930, 936, 963, 964, 974, 980, 983, 984, 997, 1001, 1066, 1092, 1096, 1105 gender 63, 92, 93, 95, 108–112, 117, 124, 144, 169–172, 177, 180, 204, 209, 216–218, 228, 252, 255, 258, 260, 279, 281, 284, 294, 310, 311, 313, 314, 317, 318, 324, 348, 351, 355, 433, 460, 470, 475, 502, 503, 505, 508, 515, 517, 524, 525, 533, 579, 661, 662, 664, 672, 680, 699, 770, 773, 774, 783, 815, 903, 906, 907, 926, 935, 944, 966, 968, 980, 984, 1007, 1009–1013, 1015–1017, 1026, 1027, 1052 genitive marker 112, 618, 798, 867, 1051, 1082, 1083, 1111 gerund 141, 196, 197, 208, 231, 250, 293, 442, 508, 517, 525, 535, 906 grammaticalization from above 48, 72, 73 grammaticalization from below 48, 57, 72, 73, 79 grammaticalization parameters 1–8, 10, 11, 35– 38, 40–48, 50–59, 61, 62, 66, 76, 78–80, 89, 93, 95, 96, 102, 115, 206, 250, 252, 270, 272, 276, 279, 282, 290, 853, 872 – correlations between parameters 44, 46, 48, 54 – form-related parameters 35, 36, 40, 41, 45, 47, 56, 78 – function-related parameters 8, 41, 45, 47, 56, 78

1141

grammaticalization paths 1–3, 7–14, 20–22, 24–31, 35, 37, 42, 53, 54, 67, 68, 74, 75, 77–79, 102, 129, 130, 132, 141–143, 150, 152, 165, 177, 180, 191, 195, 199, 205, 216, 221, 227, 229, 251, 313, 316, 320, 326, 331, 343, 352, 399, 400, 404, 409, 411, 428, 429, 437, 440, 526, 534, 542, 543, 565, 566, 570, 659, 686, 687, 689, 690, 703, 710, 714, 715, 719, 723, 764, 774, 777, 806, 833, 836, 841, 849, 862, 864, 872, 874–876, 879, 880, 884–886, 888, 889, 892, 893, 896, 897, 920, 935, 1104 grammaticalization scenarios (scenarios of grammaticalization) 1, 3, 4, 7, 8, 11, 35, 48–50, 53–55, 64, 67–69, 71, 72, 75–78, 80, 399, 706, 724, 793, 797 granularity 7, 12, 20, 22, 24, 78, 1085 heatmap 7, 37, 42, 44, 46, 47, 58, 78, 79 hierarchy 5, 42, 48, 51, 56, 78, 110, 118, 146, 147, 188, 209, 218, 260, 285, 502, 549, 585, 611, 661, 772, 857 honorification 576, 585, 594, 596, 603 hortative 97, 99, 100, 250, 283, 284, 355, 576, 597, 603, 713, 725, 730, 766, 928 ideophone 347, 546, 578, 600, 681 illocutionary force 268, 283, 1009 immediate future 20, 21, 99, 129, 131, 447, 670, 687, 688, 690, 837–841 imperfective aspect 122, 124, 250, 261, 264, 478, 846, 881 inalienability hierarchy 253, 255 inchoative 123, 127, 130–132, 193, 262, 275, 276, 376, 409, 410, 472, 478, 520, 535, 587, 589, 712, 725, 740, 939, 964, 996, 1091 incompletive 825, 983, 1060 incorporation 322, 448, 451, 556, 569, 970, 983, 991, 993, 1036, 1096 indefinite article 18, 31, 114–116, 139, 152, 174, 255, 257, 316, 318, 471, 672, 673, 678, 909–911, 1111 indefinite pronoun 100, 101, 255, 258, 318, 352–354, 493, 665, 678, 680, 688, 724, 875, 876, 896 individuation 108, 110, 174, 188, 228, 611, 615, 988 infinitive 15, 25, 27, 63, 126, 132, 141, 142, 153, 168, 169, 188, 192, 193, 196–198, 201,

1142

Subject index

202, 208, 214, 215, 220, 223–225, 230, 251, 265, 267, 268, 272, 273, 276–278, 283, 295, 312, 321–323, 325, 327, 330, 333, 334, 338, 339, 342, 343, 347, 352– 355, 376, 380, 382, 391, 395, 412, 442, 479, 480, 486, 487, 489, 494, 517, 520, 525, 527, 528, 551, 565, 566, 578, 705, 711, 713, 720, 724, 726, 771, 779, 785, 793–795, 800, 881, 906, 980, 994, 997, 1001, 1002 instrumental 21, 27, 120, 225, 261, 263, 316, 355, 372, 384, 387, 388, 392, 393, 401, 402, 404, 423, 439, 440, 453, 462, 468, 494, 509–511, 515, 517, 523, 535, 603, 685, 704, 773, 777, 780, 794, 797, 845, 849, 860, 906, 921, 923, 926, 927, 936, 939, 958–963, 974, 984, 989, 990, 1067, 1081, 1094, 1095, 1104, 1105 insubordination 33, 60, 61, 76, 211, 219, 220, 230, 326, 389, 390, 405, 414, 416, 424– 428, 530, 596, 599, 602, 603, 677, 687, 720, 721, 844 intention 19, 100, 129–132, 412, 592, 636, 637, 669, 682, 683, 688, 689, 714, 715, 725, 752, 753, 766, 928, 1001 intraparadigmatic variability 256, 263 inverse 277, 454, 513, 547, 549–551, 556, 560 irrealis 15, 136, 141–143, 198–201, 210, 212, 213, 215, 229, 272, 285, 287, 288, 328, 329, 354, 447, 557, 558, 886–888, 897, 898, 930, 964, 1038, 1064, 1072, 1105 isolating 4, 55, 68, 71 language contact 3, 12, 32–34, 63, 106, 114, 151, 252, 366, 399, 461, 477, 479, 494, 569, 853, 978, 1117–1119 lexicalization 22–24, 36, 58, 73, 79, 175, 210, 355, 402, 404, 619, 621, 917, 956 lexifier 1109, 1110, 1116, 1117 light verb 19, 66, 183, 187, 192, 194, 229, 322, 323, 346–351, 353–355, 410, 443, 444, 446–448, 472, 473, 478, 492, 494, 513, 535, 569, 583, 600, 602, 659, 665, 666, 681, 688, 689 locative 16, 21, 27, 32, 67, 76, 92, 99–101, 179, 182, 204–206, 214, 218, 271, 310, 312, 313, 315–317, 321, 324, 327, 330, 335– 339, 347, 349, 352–355, 366, 384, 387, 388, 395, 399, 401–404, 423, 424, 439– 441, 457, 462, 468, 480, 493, 494, 509,

510, 512, 523, 524, 528, 535, 542, 544, 564, 567–569, 581, 626, 633, 634, 645, 652, 676, 678, 686, 691, 701, 704, 716, 724, 726, 731, 732, 737–740, 742–745, 748, 749, 751, 763, 766, 769, 773–777, 779, 780, 783, 785, 793, 794, 797, 798, 800, 823, 850, 868, 869, 886, 896, 898, 906, 908, 920–922, 928, 931, 935, 936, 939, 965–968, 970, 990, 995, 1017, 1026, 1037, 1050, 1051, 1059, 1070–1072, 1095, 1096, 1101, 1102, 1105 macro-areas 12, 30, 31, 50, 54, 78 Mainz Grammaticalization Project (MAGRAM) V, 7, 9–11, 80 masdar (see also action nominal) meaning/form coevolution 4, 8, 11, 18, 19, 35, 38, 40–43, 68, 73, 78, 651, 1125 Meaning First Hypothesis 4, 5, 7, 9, 35, 36, 41, 46, 56, 57, 72, 78, 80 mermaid construction 34, 585, 598, 599, 601– 603 metaphoric transfer (metaphorical extension, metaphorical shift) 27, 54, 560, 629, 636, 704, 1081, 1104 metathesis 443, 446, 460, 461, 826, 827, 888 mirative 290, 447, 516, 535, 898, 999 modal obligation verb 878, 897 modality 27, 129, 136, 141, 148, 151, 192, 198, 199, 201–203, 208, 220, 251, 268, 271– 273, 278, 312, 331, 371, 412, 484, 493, 521, 523, 557, 576, 578, 590–592, 602, 609, 634, 636, 669, 682, 689, 730, 785, 881, 897, 903, 927, 964, 965, 970, 992, 1065 – boulomaic/volition 19, 129, 130, 199, 201, 592, 634, 636, 637, 669, 682, 683, 689 – circumstantial 265, 271, 273, 294, 366, 689, 782, 796, 800, 1093 – dispositional 271, 273 – dynamic 3, 121–123, 131, 185, 186, 188, 201– 203, 261, 262, 265, 271, 336, 412, 428, 514, 526, 552, 561, 582, 623, 633, 634, 636, 637, 885, 970, 1003, 1091 – epistemic 27, 31, 132, 198, 201–203, 220, 271, 272, 278, 287, 288, 328, 330, 331, 355, 493, 591, 592, 634–637, 651, 669, 730, 889, 897, 930 – obligation 26, 31, 97, 98, 129–131, 198, 199, 201, 202, 278, 291, 294, 493, 517, 522,

Subject index

525, 535, 590, 591, 599, 634, 669, 683, 687, 690, 714, 721, 723, 726, 833, 877, 889, 928 morphological reanalysis 62, 67, 69, 172 morphologization 46, 67–72, 294, 316, 411, 414 negative marker 65, 226–228, 230, 319, 703 nominalization 33, 142, 251, 312, 371, 378– 380, 390, 391, 417, 459, 549, 551, 553– 555, 564, 584, 649, 652, 683, 709, 710, 725, 780, 825, 838, 893, 915, 918, 967, 981, 982, 988, 992, 994, 997–1000 nominalizer 21, 25, 27, 295, 372, 376, 382, 383, 385, 386, 390, 391, 395, 414, 462, 509, 548, 550, 584, 597, 598, 602, 603, 850, 885, 894, 897, 898, 967, 970, 981, 995 nominalizing suffix 370, 372, 376, 437, 838, 889, 890, 893–895, 990 noun class 108, 435–437, 460, 539, 773, 775, 800, 980, 1079 noun incorporation 960, 966, 970 nucleus 15, 20, 23, 25, 26, 38, 75, 76, 78, 91 nucleus mismatch 12, 14, 15, 20, 22, 23, 25, 26, 38, 63, 76 oblique 146, 147, 173, 178, 181, 216–218, 260, 263, 266, 272, 292, 310, 311, 314, 316, 340, 355, 380, 384, 388, 390, 392, 395, 420, 423, 427, 467, 468, 470, 475, 476, 491, 495, 504, 508, 527, 565, 566, 676, 699, 701, 704, 741–743, 765, 766, 771, 856, 869, 871, 873–875, 896, 898, 905, 1082–1084, 1096, 1099–1102, 1105 Paradigmatic Variability (PV) 2, 5, 36–49, 51, 56, 63, 78, 89, 93, 102 Paradigmaticity (PM) 2, 4, 5, 36–49, 51, 53, 56, 58, 62, 78, 89, 92, 95, 102 Parallel Reduction Hypothesis 3, 4, 7, 9, 35, 36, 40, 41, 68, 78, 80 Parameter (see also grammaticalization parameters) 102, 206 passive 21, 31, 59, 94, 98, 101, 107, 121–123, 132, 134, 137, 139, 152, 153, 183–189, 192, 201–204, 229, 259–266, 269, 272, 273, 282, 292, 294, 295, 323, 334, 386, 395, 407–409, 428, 453, 473, 478, 483, 492, 494, 513–515, 517, 518, 522, 523, 526, 535,

1143

550, 552, 554, 555, 561, 586, 587, 589, 603, 609, 622, 637–640, 642, 646, 651, 662, 699, 703, 707, 711, 725, 770, 772, 773, 777, 778, 785, 797, 800, 826, 827, 843, 849, 850, 856, 878, 879, 898, 905, 980, 994, 1044, 1058, 1059, 1069–1072 past 15, 21, 22, 24, 31, 71, 72, 75, 98, 122, 125, 127, 132, 169, 186, 192–195, 198, 205, 219, 229, 251, 266, 268, 279, 280, 283, 284, 290–292, 294, 323, 325, 326, 328, 331, 354, 381–383, 390, 410, 413, 415, 416, 427, 444, 446, 447, 451, 472–476, 478, 480, 483–485, 491, 494, 504, 515, 516, 519, 520, 525, 535, 557, 558, 588, 590, 593, 620, 666, 705, 711, 714, 715, 725, 834, 835, 881, 882, 920, 964, 965, 974, 996–999, 1007, 1022, 1023, 1036, 1037, 1046, 1060, 1072, 1092, 1114 paths (see also grammaticalization paths) perfect 22, 31, 34, 71, 95, 106, 123–125, 127, 128, 138–140, 152, 168, 169, 194, 195, 251, 252, 263, 266, 275, 276, 279–282, 290, 291, 294, 325, 326, 328, 329, 337, 344, 347, 354, 355, 383, 414, 415, 421– 423, 426, 470, 473, 475, 478, 482, 485, 491, 493, 495, 502, 503, 515, 518, 567, 593, 621, 627–631, 668, 682, 687, 688, 690, 711, 725, 729, 730, 733, 766, 786– 789, 791, 794, 797–800, 819, 825, 834, 850, 881, 884, 898, 999, 1044, 1060, 1122, 1124 perfective 32, 33, 66, 71, 74, 90, 98, 124, 125, 127, 128, 131, 188, 192, 194, 196, 203, 250, 261–264, 267, 268, 294, 312, 320, 323, 325, 326, 328, 331, 333, 337, 339, 343, 344, 354, 355, 381, 390, 395, 422, 445–447, 461, 462, 475, 478, 479, 518, 520, 525, 535, 557, 558, 603, 621, 629– 631, 645, 652, 662, 665–667, 675, 679, 682, 691, 833–836, 850, 885, 894, 897, 958, 960, 964, 996, 1009, 1019–1021, 1024, 1036, 1038, 1058–1060, 1071, 1072, 1105, 1114 periphrastic construction 70, 127, 129, 132, 321, 325, 379, 383, 466, 665, 682, 828, 838, 849, 1021 permissive 98, 100, 223, 225, 250, 265, 266, 283, 284, 407, 493, 513, 535, 591, 646, 708, 725 personal pronoun 25, 63, 107, 147, 182, 183, 340, 341, 352, 354, 369, 370, 378, 386–

1144

Subject index

388, 392, 405, 406, 585, 663, 680, 687, 690, 698, 706, 725, 735, 757, 812, 814, 822, 843, 848, 857, 872, 915, 926, 930, 983, 1010, 1011, 1013–1015, 1026, 1028, 1030, 1031, 1036, 1051, 1052, 1071, 1110, 1112, 1120 Phonetic Reduction (PR) 4–6, 8, 36–48, 51, 54, 56–58, 60, 62, 66, 68, 69, 73, 78, 79, 89, 91, 92, 102, 729, 1012 phonological reduction 72, 349, 473, 492, 556, 631, 951, 974, 1000, 1035 plural marker 112, 172, 191, 229, 314, 355, 378, 379, 385, 400, 401, 404, 406, 407, 417, 442, 455, 461, 469, 505, 506, 539–541, 544, 580, 602, 700, 701, 724, 774, 797, 924, 928, 935, 986, 993, 997, 1111, 1116, 1123 polygrammaticalization 12, 14, 15, 18–21, 25, 74–76, 145, 411–414, 428, 846 polysemy copying 262, 264, 1109, 1121, 1123– 1125 polysynthesis 73, 461, 943, 944 portmanteau 249, 252, 258, 445, 449, 547, 549–551, 559, 661, 675, 817, 925, 981, 992, 1093 possession 15, 74, 112, 173, 180, 188, 194, 198, 201, 203, 229, 252, 253, 255, 280, 285, 294, 316, 353, 370, 378, 379, 400, 403, 437, 470, 508, 539, 540, 581, 663, 664, 687, 690, 697, 701, 737–740, 808, 833, 858–867, 896, 923, 980, 984, 1007, 1017, 1018, 1025, 1049, 1081, 1082, 1097– 1100, 1102, 1105 – alienable possession 173, 180, 697, 701, 724, 862, 984 – inalienable possession 173, 253, 255, 257, 294, 470, 540, 701, 862–867, 896, 984, 1081, 1084, 1102 possessor 112, 146, 173, 180, 217, 218, 252– 255, 315, 316, 387, 402, 403, 405, 437, 467, 507, 534, 540, 541, 553, 581, 661, 701, 738, 739, 749, 756, 806, 814, 833, 864, 868, 870, 944, 979, 980, 983, 984, 986, 993, 1011, 1025, 1049, 1053, 1061, 1083, 1105, 1112 – external 253, 254, 749 – inalienable 253 – NP-internal 254 postposition 17, 60, 70, 316, 339, 385, 392, 393, 402–405, 433–435, 438–441, 447,

448, 456, 457, 460, 467, 468, 501, 504, 508, 510–512, 517, 527, 528, 530, 531, 543, 544, 567, 568, 575, 576, 581–583, 601– 603, 618, 619, 645, 647, 668, 670, 676, 678, 686, 697, 699, 701, 703–706, 708, 710, 711, 713, 716, 717, 724–726, 870, 871, 886, 898, 908, 920, 921, 935, 1009, 1017, 1079, 1081–1083, 1087, 1088, 1093, 1094, 1096, 1097, 1104, 1123 postpositional 457 posture verb 14, 18, 21, 25, 26, 668, 670, 671, 884, 885, 897, 935 preposition 16, 31, 32, 57, 59, 73, 92, 95, 96, 107, 120, 141, 147, 149, 152, 173, 178, 180–182, 192, 193, 197, 198, 203, 209, 210, 212, 217, 228, 230, 231, 273, 468, 501, 609, 618, 628, 634, 637, 638, 645, 646, 648, 651, 652, 699, 700, 705, 718, 719, 721, 723, 724, 729, 740, 742–745, 749, 773, 775–780, 784, 794, 795, 806, 808, 809, 811, 812, 816, 817, 823, 829, 830, 832, 839, 840, 843–845, 850, 1045, 1051, 1059, 1067, 1070–1072, 1096, 1111 preverb 23, 255, 279, 295, 335, 336, 338, 355, 472, 479, 480, 708, 716, 725, 729, 730, 736, 755–762, 764, 765, 1080, 1086, 1089, 1090, 1104, 1105 proclitic 59, 120, 254–256, 277, 281, 285, 437, 451, 661, 663, 776, 781, 816, 921, 924, 1046, 1047, 1049, 1052, 1055, 1062–1064, 1066, 1070, 1071 progressive 18, 20, 33, 71, 117, 126, 195–197, 229, 266, 270, 271, 326, 327, 354, 366, 383, 409, 410, 413, 428, 470, 479, 480, 483, 492, 494, 504, 514, 518–520, 549, 557, 560, 561, 587–589, 621, 632–634, 651, 652, 667, 668, 682, 687, 710, 725, 793, 830–833, 837–839, 844, 849, 850, 883, 897, 898, 994, 995, 1001, 1021, 1120 prohibitive 269, 283, 355, 590, 603, 662, 713, 714, 725 proper name marker 21, 909, 915, 917, 935, 936 proprietive 404, 542, 543, 701, 705, 724, 726 prototype 79, 978, 979

question marker 218, 533, 536, 595, 678, 686, 687, 689, 724, 726, 916 questionnaire 2, 10, 11, 19, 25, 35, 45, 80, 89

Subject index

quotative 19, 32, 101, 312, 333, 352, 354, 417, 488, 495, 529, 535, 592, 647, 648, 659, 679, 684, 688, 690, 705, 713–715, 717, 720, 725, 726, 793, 799, 850, 888, 897, 898, 914–916, 935 reanalysis 3, 6, 9, 18, 54, 55, 58–64, 69, 72– 74, 76, 77, 109, 110, 124, 135, 142, 148, 165, 169–171, 173, 180, 182, 183, 185, 199, 211, 221, 223, 228–230, 282, 290, 292, 293, 326, 348, 351, 352, 361, 389, 390, 402, 404, 406, 409, 412, 414, 417, 420, 423–429, 433, 442–444, 446, 451, 456, 460, 461, 465, 466, 476, 484, 513, 517, 521, 526, 529, 530, 543, 548, 550, 553– 557, 559, 590, 599, 603, 623, 631, 639, 646, 650, 651, 678, 700, 706, 707, 711, 713, 714, 716, 721, 722, 724, 743, 746, 757, 777, 785, 795, 797, 841, 997, 1043, 1067, 1070 recipient language 1120, 1121, 1123, 1124 reciprocal pronoun 324, 353, 708, 725, 734, 1115 reflexive 21, 97, 99, 122, 178, 183–185, 187, 189, 223–225, 229, 250, 251, 254, 259, 261, 265, 269, 272, 323, 324, 345, 353, 386, 402, 423–425, 478, 502, 506–508, 534, 552, 556, 569, 585, 604, 680, 706, 733, 734, 749, 756, 757, 765, 766, 770, 785, 823, 849, 856, 879, 898, 906, 922, 923, 935, 936, 944, 956, 957, 974, 981, 992, 1091 reflexive possessive 255, 423–425, 923, 937, 939 reflexive pronoun 107, 183, 345, 447, 507, 663, 665, 691, 706, 708, 710, 725, 823, 879, 880, 897, 936, 1115, 1123 relational noun 16, 17, 32, 56, 70, 402, 609, 617, 619, 627, 649, 651, 808, 811, 812, 823, 830, 843, 844, 848, 850, 920, 921, 1018 relative clause 31, 34, 145, 147, 148, 152, 208, 216, 218, 320, 344, 345, 420, 422, 459, 471, 487–490, 501, 562, 649, 662, 669, 672, 674, 676, 677, 683–685, 697, 718, 731, 738, 781, 782, 793, 803, 824, 875, 890, 892–896, 919, 979, 1044, 1059, 1069, 1096 relativizer 134, 146, 148, 149, 216, 217, 459, 462, 535, 597, 598, 688, 698, 717–719, 725, 726, 799, 875, 892, 897, 898

1145

renewal 67, 77, 120, 401, 406, 425, 426, 428, 513, 514, 517, 521, 659, 665, 680, 681, 689, 785, 803, 974 repetitive 310, 337, 338, 349, 352, 354, 355, 587, 712, 725, 730, 766, 939 resultative 18, 71, 75, 124, 125, 186, 188, 189, 194, 251, 252, 262, 264, 265, 275, 279– 282, 290, 293, 294, 326, 354, 409, 415, 421, 445–447, 449, 453, 478, 482, 515, 517, 555, 588, 589, 593, 602, 604, 621– 627, 629, 631–633, 640, 706, 711, 725, 850, 996, 1059, 1088, 1089, 1104, 1105 resumptive pronoun 144, 145, 147, 217, 218, 345, 510, 527, 530, 531, 535, 536 scenarios (see also grammaticalization scenarios) secondary grammaticalization 51, 58, 72, 96 semantic bleaching 116, 123, 180, 203, 223, 225, 228, 229, 285, 440, 584, 615, 702, 808, 999 semantic change 6, 35, 60, 61, 445, 637, 746, 868, 988, 1001, 1052, 1109, 1125 semantic imitation 1109, 1119, 1120, 1123, 1125 Semantic Integrity (SI) 2, 8, 36–49, 51, 56–58, 62, 78, 80, 89, 91, 102 semantic reduction (see also desemanticization, semantic bleaching, semantic integrity) source concepts 1, 3, 10, 12–16, 19, 26–30, 90, 91, 95–97, 102, 129, 611, 649 stance 380, 587, 588, 602 stress-based language 68, 152 Structural Scope 2, 5, 36, 59, 89, 201, 609 subjunctive 20, 26, 59, 74, 75, 106, 132–134, 141, 190, 193, 198–200, 202, 213, 219, 220, 222, 230, 250, 288, 327, 329, 354, 415, 418, 422, 427, 429, 479, 483, 484, 486, 494, 495, 517, 521, 524, 535, 713, 720, 721, 725, 726, 795, 796, 800, 806, 825, 827, 829, 836, 839, 841, 850, 898, 1046, 1047, 1065, 1072 substrate 35, 281, 384, 433, 512, 529, 978, 1109, 1119, 1120, 1122–1125 suppletion 1020 Syntagmatic Variability (SV) 2, 4, 5, 36–49, 51, 56, 58, 62, 70, 78, 89, 95, 102, 139, 256, 294 tense 18, 27, 68, 70, 92–96, 111, 122, 123, 125, 127, 129, 131, 132, 191, 192, 196, 198, 199,

1146

Subject index

251, 252, 262, 263, 266–269, 272, 273, 275–279, 283, 284, 287, 291, 292, 294, 311, 312, 318, 325, 327–329, 331, 333, 339, 344, 346, 352, 353, 371, 377, 378, 380–383, 385, 390, 407, 412–415, 417, 426, 444, 446, 447, 449, 451, 461, 473, 475, 478, 480, 481, 484, 486, 491, 492, 504, 505, 514–516, 524, 525, 535, 557, 576–578, 593, 594, 596, 620, 631, 636, 637, 645, 652, 670, 683, 690, 714, 730, 816, 822, 825, 827, 828, 833, 834, 837, 881, 882, 886, 903, 927, 928, 944, 953, 963–965, 974, 981, 983, 992, 993, 997, 999, 1007, 1009, 1018, 1021–1023, 1031, 1032, 1046, 1059, 1060, 1071, 1078, 1080, 1092, 1105, 1113, 1114 terminative aspect 520, 668, 682, 684, 687, 688, 832 tertiary grammaticalization 62 topic 60, 135–137, 175, 181, 182, 211, 213, 222, 325, 341, 345, 348, 391, 529, 531, 535, 539, 545, 546, 583, 587, 599, 602, 604, 642–644, 647, 662, 664, 711, 729, 735, 736, 771, 806, 813, 814, 840, 847, 848, 910, 911, 918, 920, 928, 931, 932, 935, 936, 982, 983, 1044, 1050, 1053, 1071, 1102 topic-comment construction 643 translocative 72, 551, 964, 965, 974, 975 transparadigmatic variability 270 transparency (extra transparency) 1109, 1118– 1129 unidirectionality 3, 7, 230, 1119 univerbation 57, 70–73, 79, 151, 152, 288, 348, 352, 355, 402, 410, 428, 701, 785, 943, 974

valency 11, 34, 95, 183, 212, 213, 215, 259, 268, 275, 321, 322, 348, 407–409, 445, 452, 478, 513, 561, 569, 586, 603, 667, 681, 689, 699, 707, 709, 770, 772, 784, 826, 876, 903, 905, 906, 921, 923, 981, 992, 1001, 1007, 1018, 1023, 1024, 1058, 1090, 1092, 1093 values 8, 10, 11, 35, 37, 38, 40–47, 49, 51, 53, 55, 56, 58, 62, 90–96, 102 verb serialization 19, 22, 57, 273, 552, 587– 589, 593, 623, 631, 641, 648, 731, 737, 747, 926, 1009, 1019, 1021, 1036, 1045 verbalization 60, 61, 371, 389, 390, 416, 420, 422, 424, 426, 427, 841, 981, 982, 992, 995, 998 verb-first construction (V1-construction) 126, 135–137, 140, 152, 168, 211, 662, 667, 917 verbless 659, 689, 1097, 1099, 1100 verb-second construction (V2-construction) 58, 60, 91, 126, 134, 135, 137, 152, 168, 211, 426, 662, 665, 666 voice 69, 94, 121, 122, 127, 183–185, 192, 229, 251, 259–261, 264, 273, 280, 321, 323, 324, 371, 407, 428, 452, 472, 478, 492, 504, 514, 535, 539, 552, 554, 556, 586, 639, 661, 662, 667, 826, 876, 1044–1046, 1057–1059, 1061, 1066, 1068, 1070, 1072

Wackernagel (position, properties, rule) 254, 255, 277, 288, 831 word order 4, 11, 55, 58, 64, 65, 69, 96, 134, 135, 137, 139, 152, 153, 168, 169, 184, 194, 211, 212, 250, 253, 309, 312, 363, 364, 384, 399, 569, 575, 576, 640, 646, 729, 763, 764, 856, 879, 1116, 1123