208 62 9MB
German Pages 152 Year 1979
Linguistische Arbeiten
71
Herausgegeben von Herbert E. Brekle, Hans Jürgen Heringer, Christian Rohrer, Heinz Vater und Otmar Werner
Jürgen Klausenburger
Morphologization : Studies in Latin and Romance Morphophonology
Max Niemeyer Verlag Tübingen 1979
CIP-Kurztitelaufnähme der Deutschen Bibliothek Klauaenbuiger, Jürgen: Morphologization, studies in Latin and Romance morphophonology / Jürgen Klausenburger. - Tübingen : Niemeyer, 1979. (Linguistische Arbeiten ; 71) ISBN 3-484-10332-9
ISBN 3-484-10332-9
ISSN 0344-6727
© Max Niemeyer Verlag Tübingen 1979 Alle Rechte vorbehalten. Ohne ausdrückliche Genehmigung des Verlages ist es auch nicht gestattet, dieses Buch oder Teile daraus auf photomechanischem Wege zu vervielfältigen. Printed in Germany
Lynn, Paul, Christiant
Marcus
VII
CONTENTS
Preface
IX
I. THEORY 1. Current proposals concerning norpho logi zation
1
1.1. Within transformational generative phonology 1.2. Within natural generative phonology
1 4
1.3. The role of the phoneme
9
2. Kruszewski's Ueber die Lautabweahslung
12
2.1. Background 2.2. Synchronic analysis
12 13
2.3. Diachronic analysis
17
3. A comprehensive theory ofraorphologi zation
23
3.1. Data and rules 3.2. Theoretical interpretation 3.3. Historical (morpho)phonology
23 29 35
II. ILLUSTRATIONS 4. Latin
37
4.1. Rhotacism 4.2. Lachmann's Law
37 42
4.3. Nasal infixation
49
5. French
55
5.1. Linking phenomena 5.1.1. Introduction 5.1.2. The transformational generative analysis 5.1.3. Some current trends 5.1.4. A revised analysis 5.1.5. Morphologization 5.2. The mute /e/ 5.3. Nasalization
55 55 58 60 62 65 68 74
Vili
6. Romance 6.1. Pluralization in Western Romance 6.1.1. Old French 6.1.2. Catalan 6.1.3. Spanish 6.1.4. Portuguese 6.2. Diphthongization 6.2.1. Spanish 6.2.2. French 6.3. Velar-palatalization 6.3.1. Italian 6.3.2. Old French
84 84 84 88 92 96 101 101 105 108 108 113
III. IMPLICATIONS A. Dissection of Part II
115
B. Motivation
118
C. What is morphologization?
124
JUST OF TABLES
127
REFERENCES
128
INDEX OF AUTHORS
136
IX
PREFACE
Within the last half decade, phonological theory has undergone great upheavals, the paradigm established in Chomsky and Halle's Sound Pattern of English (1968), also represented in Harms (1968), Postal (1968), and Lightner (1971), clearly giving way to a post-paradigm period of differentiation and variation, a development which is in accord with Kuhn's (1969) conception of the "structure of scientific revolutions," Percival's (1976) skeptical view of the applicability of paradigm analysis to linguistics notwithstanding. Textbooks on phonology since SPE, such as Schane (1973), Artier son (1974), Hyman (1975), Kenstcwicz and Kisseberth (1977), and Sloat et al. (1978), do not, however, amount to revolutionary breaks frcm transformational generative theory. The one exception in this respect is Hooper (1976), a (probably preliminary) codification of the most premising thrust in post-transformational generative phonology, natural generative phonology, as developed by both Vennemann and Hooper.
It is, perhaps, too
early to call natural generative phonology a new paradigm of phonology. The apparently decisive differences between transfomational generative and natural generative approaches could, in other words, still be viewed intra-paradigmatically, and not necessarily inter-paradigmatically (cf. Botha 1971, 1973; Derwing 1973). The theoretical apparatus of this monograph aligns itself with natural generative phonology in its most essential aspects.
It is entirely devoted to mor-
phologization, a concept that plays a crucial role in natural generative theory, since the dichotomy of phonetically conditioned phonological rules and morphologically conditioned rrorphophonemic rules constitutes its cornerstone. Other issues separating transformtional frcm natural generative phonology, such as rule ordering controversies, the status of archi-phonemes, and the question of words vs. morphemes in the lexicon, will not be gone into here. The reader is referred to Hooper (1976), who presents the most up-to-date review of these questions.
X The general structure of the book is as follows. Part I develops a theory of norphologization, campariiig and contrasting it to extant formulations in the literature. One novel component of this section is the detailed analysis of a little-known work of the last century, M. Kruszewski's Ueber die Lautabweohslung (1881), to which are attributed theoretical ideas reminiscent of most recent claims in post-transformational generative phonology. Part II applies the new theory to selected aspics of Latin and Rcmance morphophonology. No diachronic relationship between the Latin and Rcmance data exists, nor is it intended to give a complete synchronic description of any of the languages. The canton thread, one which, it is hoped, produces a coherent whole of so many disparate subjects, should be seen in the morphologization aspect that each demonstrates to some degree, further explored for its general irrplications in Part III. The choice of topics remains, of course, arbitrary, since others offer themselves readily, and their number could be extended with ease. I am grateful to the University of Washington for sabbatical leave, granted to me during Winter and Spring Quarters 1977, which permitted undisturbed work on the pre-final version of the manuscript.
Jürgen Klausenburger Seattle, July 1978
1
I.
THEORY
1.
Current proposals concerning morphologization
1.1.
Within transformational generative phonology
Morphologization plays almost no role in the early version of generative historical linguistics, as codified in King (1969). German umlaut, which, everyone will agree now, constitutes perhaps the classic and best-known case of an erstwhile phonetically conditioned rule becoming morphologized, is reviewed by King (1969:92-101) diachronically without any mention of the term morphologization. This does not mean, however, that King did not, in fact, describe such a process. He presents German umlaut as a case history of "a sound change from innovation to restructuring" (p. 92). Throughout his analysis, he is keenly aware of how long the rule is one of overtly marked umlaut, that is, up to what point the original rule addition (p. 94),
- back long^>
- cons + high - back
retains its phonetic conditioning. Restructuring set in during Old High German Stage IV, when an umlaut-causing final vowel like [i] was reduced to schwa, /gasti/ -»• /geste/ -* [gesta] (p. 101) . King's analysis of Modern German umlaut remains, however, ambiguous, as he considers only seme occurrences of umlaut as phonemic. He would account for a derivational alternation like Mann, männlich by a synchronic recapitulation of the historical sound change, since [i] still appears in the suffix of the adjective. King is not explicit as to how one should relate Gast to (restructured) Gäste today. It is the for-
2
mulatian of a morphologized rule, roughly, a
-r e
/
plural,
vAiich is absent in his account.*' Kiparsky (1971) classifies Germanic unlaut as claque (ii), caused by the elimination of the conditioning [i] and [y], "the initial stage of the process of morphologization" (1971:634). formulated as "a rule A
The concept of opacity, originally
-r B / C _ D is opaque to the extent that there are
surface representations of the form (i) A in environment C _ D, or (ii) B in environment other than C _ D" (1971:621-2), later expanded to "(iia) B derived by P in environment other than C _ D, and (iib) B not derived by P (i.e. underlying or derived by another process) in environment C _ D" (Kiparsky 1976: 179), thus may serve as a motivation for morphologization.
In addition to
unlaut, Kiparsky also suggests that Hebrew spirantization has became morphologized, since it is opaque (ii) (1971:629)(cf. Barkai 1974:373).
Furthermore,
Latin rhotacism is given two synchronic morphological formulations, either s
•+ r / naninative singular, or
r
-» s / derivations like honestus (p. 598).
Although this example is discussed under
PARADIGM CONDITIONS and under PARA-
DIGMATIC COHERENCE (Kiparsky 1972:206-8), the criterion of opacity may also be
1
Interestingly enough, a totally morphological account of German umlaut was proposed within transformational generative theory as early as Zwicky (1967), "an example of a generative treatment of a synchronic phonological phenomenon for which no purely phonological solution has been found" (p. 44). Bach and King (1970) defend the (abstract) phonological version, on the other hand, an analysis criticized by Wurzel (1970:105-69), who also opts for a morphological solution. Most recently, Robinson (1975) has given a concrete historical analysis of umlaut, assuming the onset of morphologization "after its original phonological environment had been phonetically neutralized" (pp. 26-7). As a consequence, "the umlaut rule has become a rule schema with a large number of disjunct morphological environments" (p. 5). See also Hyman (1975:175-6).
3 used: rhotacism is doubly opaque, and thus a morphologized rule. In Kiparsky (1972), both PARADIOftTIC COHERENCE and CONSERVATION OF FUNCTIONAL DISTINCTIONS are Introduced as forces behind morphologization, a term which, however, is not enployed in this article. The second principle describes a morphological conditioning which "characteristically originates as a blocking of rules in environments in which their free application would wipe out morphological distinctions on the surface" (p. 197). This is illustrated in seme American dialects by a word-final t-deletion rule which does not apply when the [t] happens to be the only marker of the past tense: keep, kep' but steep, steeped (p. 197).
Harris (1974:19) proposes the following diphthongization rule for Chicano Spanish: (a)
[+ stress] X [+ back] C 0 1 conj
verb
(b)
Part (b) constitutes the morphological conditioning in this dialect, co-existing with the phonetic conditioning in part (a). Chicano diphthongization is opposed to the Standard Spanish rule (p. 12): e o
ye r
[+ stress]
Evidence for the morphologized version cones from a verbal paradigm like vuelar 'to fly,' which displays the diphthong [we] throughout, contrary to Standard Spanish, which has the monophthong [o] in the 1st and 2nd persons of the plural. Actually, it would seem that no rule at all is needed in Chicano, due to the levelling. However, Harris points to derivatives like volador 'flyer,' 2
See Schindler (1974:2) on rhotacism and below, 4.1. Schindler equates an opaque (i) rule to morphologization (p. 6, fn. 2), while an opaque (iib) rule may be described as a (morphologized) rule inversion (p. 3). See also Thomason (1976). Kiparsky (1975:193-4) seems to retract his 1971 position when he calls "unmotivated ... the assumption that rules which are
4 which is assumed to alternate with the verb. One may also question the phonological nature of Standard Spanish diphthongization. Harris does not hesitate to bring up the "countless instances of stressed, undiphthongized /e/ and /o/" (p. 11), and the "many examples of diphthongs, both stressed and unstressed, that do not alternate with /e/ and /o/, or anything else" (p. 12). These facts render diphthongization in Standard Spanish doubly opaque, thus, presumably, norphologized. Kiparsky's criterion for morphologization is mentioned by Harris (pp. 8-9), yet, curiously, not applied to his own example. Harris, in the final analysis, attributes the morphologization of Chicano diphthongization to "paradigmatic uniformity" (p. 20).^ 1.2.
Within natural generative phonology
tforphologization plays a key role in the theory of natural generative phonology, as developed by Vennemann and Hooper, since the most striking difference between transformational and natural generative phonology resides in the latter' s "dividing up of the group of processes that TCP calls phonological into three types: phonological rules, morphophonemic rules, and via-rules" (Hooper 1976:84). The synchronic division presupposes that morphophonemic rules (and via-rules, which are irerely lexical relation equations) evolve historically from phonological rules because of "a desire on the part of the speaker to interpret alternations as meaningful" (1976:87). This general motivating principle behind morphologi zation is made irore specific in connection with individual case histories. Vennemann (1972a:188-9) considers the morphologization of Yiddish devoicing,
3
opaque or have exceptions are by that fact alone to be treated as morphologically conditioned." Cf. Koefoed (1974), King (1972), and Kiparsky (1974a, b). Crothers and Shibatani (1975:507) place opaque rules into MORPHOPHONOLOGY and transparent (non-opaque) rules into PHONOLOGY. Cf. Crothers (1971). For a detailed discussion of Spanish diphthongization, and a different view of the Chicano situation, see Anderson (1975), and below, 6.2.
5
[- sonorant]
[- voice] /
into [- sonorant] •* [- voice]
morphosyntactic information
as the consequence of an apocope rule, which had placed new voiced obstruents in final position: tok, tegi'day' (sg., pi.)
tok, teg. Of course, this
morphologization also led to the loss of devoicing, resulting in paradigm levelling: tog, teg. Vennemann makes the important claim that "rule loss is always caused by conceptual analogy (Humboldt's Universal)" (p. 188), one of whose implementations is morphologization. Similarly, Verner's Law's original phonetic nature. + obstruent
[+ voice]
+ syllabic ([+ voice])
[+ voice]
- accent
+ continuant
was modified morphosyntactically into (p. 189) + obstruent [+ voice] + continuant
[+ voice]
+ + + + + +
verb past ablaut plural subjunctive participle
[+ voice]
changing phonetically predictable forms like wirQana (infinitive), warOa (past sg. indie.), and wr%ana (past participle) 'to become,' into the phonetically unaccountable series wirQana, warQa, wriana. Again, the final result is the loss of the morphologized version: Gothic wairrfan, ward, waurrfans (p. 190). Vennemann includes morphologization as part of "a future classification of phonological change" (p. 202):
6
I. Phonetically motivated simplification
II. Conceptually motivated simplification
1. Addition of natural rule
1. Rule loss
2. Rule generalization
2. Relexicalization, with a. a resulting lexical redundancy rule b. a morphologized rule c. an inverse rule
3. Rule unordering
For the first time in the contemporary literature, morphologization plays an 4
integral part in a theory of (norpho)phonological change. Vennemann (1972b, 1974) introduced into current theory the notion of rule inversion, a special type of morphologization. What is involved is not only the change in conditioning frcm purely phonetic to (partially) morphological, but also a reversal of the input and output of the rule, the original derived segment (or form) becoming basic: | A -» B / phonetic conditioning B -» A / morphological conditioning.^ Hie fundamental motivation of rule inversion is "the principle of the dominance of semantically primitive categories," (p. 237) according to which a derived segment can plausibly be considered basic for a subsequent generation if it happens to be part of morphological categories like singular, nominative, present tense, and active voice. In the inverted rule, the output would belong to more "complex" categories, such as plural, oblique case, past tense, and passive voice. Vennemann is aware of the difficulties with such a principle, its inpreciseness and arbitrariness (1974:139). Therefore, in practice, other criteria must be used for deciding whether rule inversion has occurred, such as the relative frequency of alternants, order in language acquisition, and historical data. Hius, in motivating the inverse synchronic rule for the Eng4 5
The place of morphologization in a post-TG historical linguistics will be taken up in detail in 3. Hyman (^97^:177) mentions the case of an inversion without morphologization from Fe' fe' : . d -» 1 / # # _ * 1 •* d / n _ In fact, the inverted rule may be considered more "natural" than the original, historical, one, "since a [+ cont] segment such as /l/ can assimilate to the [- cont] specification of a preceding homorganic nasal, thereby becoming /d/." See also Vennemann (1972b:239).
7 lish indefinite article, a -r an / _V, as opposed to the historical an -» a / _ C, Vennemann cites both relative frequency (a is basic because it occurs before a consonant, a presumably more frequent environment than _ V), and language acquisition (children have a long before an, thus *a apple, *a other one) (1972b:213). A second hiatus rule example of inversion concerns certain English dialect alternations like the water is [wjtar iz], the water may [votjmey], the water [wjt>], which are explained historically by an r-deletion rule,
7
- - rj' synchronically, however, due to the existence of intrusive r's in cases where they are etymologically not justified, as in the idea -r- is [aydijjr iz], r-insertion before vowels seems to be motivated (p. 216).^ Hooper (1974, 1976:32-41) traces in some detail the morphologization of the plural marker in the Eastern Andalusian dialect of Granada. Along with other Spanish dialects, this one had a vowel laxing rule, conditioned by checked syllables: [klase] sg., [klasEs] pi. 'class' - the capital letter signifies laxness (p. 161). However, syllable-final [s] may be weakened to aspiration, [h], or lost entirely, yielding a plural like [klasE(h)]. Vowel laxing has been morphologized, V
-»• [- tense] /
pluralj
(h) #
#
(p. 164) ,
since /s/ is no longer the plural marker; rather, "the plural is signalled by an open vowel followed optionally by aspiration: the morpherre that is added after consonants is A*/, after vowels / h / or 0" (p. 164). At first glance, it vrould appear that this norphologization can be explained by assuming that it took place BECAUSE the phonetic motivation for a rule was destroyed or obscured (by s-aspiration or deletion). Hooper contributes this significant 6
The essence of rule inversion can be traced back as far as Paul (1909: 119). Cf. Plank (1975). French liaison is analyzed as rule inversion in Klausenburger (1974a) and will be discussed as part of French linking processes below, in 5.1.
8 observation to our understanding of morphologization in disagreeing with such an assessment:"... the existence of both vowel harmony and optional final aspiration at one stage suggests that ... the phonetic environment is allowed to be lost BECAUSE the rule has been morphologized" (p. 167). (The vowel harmony mentioned is illustrated in these data: (p. 164) singular
plural
[peetaQo] [kabe9a]
[pEdAGO] [kAbEGA]
'piece' 'head'
[lobo] [grupo]
[10b0h] [grUpO^1]
'wolf' 'group'
It is obvious that the distinction between tense and lax vowels is entirely a question of morphological category: both stressed and unstressed, word-final and non-word-final vowels are affected). The Granada pluralization development is interesting since no opaque rule is involved, yet morphologization has taken place: the presence of (optional) aspiration maintains the transparent nature of vowel laxing vrord-finally. (cf. also Hooper 1976:87). Hooper (1976:88-9) also cites the celebrated Maori case (Hale 1973) as morphologization of a transparent rule: verb
passive
awhi hopu
awhitia hopukia
aru tohu mau
arumia tohunia mauria
'to embrace' 'to catch' 'to follow' 'to point out 'to carry'
The historical consonant deletion rule, C -* & / _ , seen in the verb column, has been morphologized in that the empirical evidence suggests "that the basic form of the passive suffix is /-tia/, and the other forms, /-kia/, /-ria/, etc. 7 are allottorphs" (p. 89). However, Hooper contends, the original C-deletion was not opacated, as it could still exist in Maori, outside the verbal parag digms listed. 7
Hyman(1975:184) proposes the
(morphologized) rule
inversion,
8
?S k,m,n,r / passive (gerund) - . Kenstowicz and Kisseberth (1977:74-7) discuss cases of morphologization under the term "grammaticalization."
9 1.3.
The role of the phoneme
Skousen (1975b) concludes his discussion of the norphologization of a number of rules of Finnish thus: "[This] desire for surface regularity is the motivating force that morphologizes phonetic rules ..." (p. 125). The attribution of irorphologization to surface exceptions is, of course, reminiscent of Kiparsky's opacity. However, in Skousen's analysis, a novel detail appears, as he argues that "if the output of an historical, low-level phonetic rule (or a series of such rules) creates a segment that is interpretable as a phoneme which is different from the one it originally caite from, then speakers will tend to account for the effects of that phonetic rule (or rules) by a morphological rule" (p. 101).
(Cf. also Skousen 1975a:223).
Skousen first considers the historical rule of consonant gradation of Finnish which, in short, closed syllables, had these effects: (p. 106) P
b
t -d k -» §
PP - P tt -» t kk -* k
The second set clearly produced phonemes which already existed in the language, while the first resulted in allophonic fricatives. A synchronic phonetic rule of short consonant gradation seems, thus, plausible. But because both long and short consonant gradation occurred simultaneously, an original ktttettOn 'without a hand1 yields [kcLdetOn] , in which form the obligatory environment for a synchronic t -* d rule (closed syllable) is no longer met. "Gradation was an example of a phonetic rule that caused its own death" (p. 110). That speakers are, in fact, unable to capture short consonant gradation phonetically is proven rather conclusively by the replacement of the second part of the historical alternations p, b; t, d; and k, g; by phonemes: p, v; t, r; t, 1; t, d; t, 0; k, 0 (p. 108).9 Two other historical rules of Finnish, t •* s / _ i, and e -»i / _ also are morphologized, according to Skousen. In both instances, the outputs, 9
Skousen (1975b:59-64) shows that synchronically consonant gradation is not productive and is thus not captured phonetically by the speaker, even though the environment necessary, closed syllable, remains in the language. See also Hooper (1976:87-8) for tliis problem.
10 /s/ and /i/, were phonemes in the language at the time of the addition of the rules. Although the e -» i change applied without exception, the e - i alternation, due to the particular morphological restrictions in distribution, was really accounted for in terms of "whether or not a suffix was added to a stem" (p. 120).
It is important to stress that Skousen considers the phonemic sta-
tus of /i/ to be instrumental in this morphologization.
Similarly, the t - s
alternation (with phonemic /s/) came to be associated with the past tense marker. Substantive evidence for the latter can be found in seme dialects, where si (instead of correct i) is added to form, for exanple, sanosi 'he said' (p. 121). In Skousen's theory of morphology (1974), the phoneme is specifically assured to be the unit of concrete phonetic representation, at which level morphological relationships are expressed. Three characteristics of phonemes are given (pp. 318-9): (1) They are distinguished frcm each other by the principle of "maximal acoustic distinctiveness;" (2) this principle fully specifies all the distinctive features for each phoneme, disallowing archi-phonemes; (3) phonemes are initially extracted frcm phonetic representation alone, and thus no "grairmatical prerequisites to phonemic analysis" exist. Skousen proposes several principles of morphology, which are said to be valid for both adult speech and child language: arbitrariness, identity, induction, extension, and separation (p. 318). Although based in synchrony, his theory of morphology may also serve as a model for itDrphologization: When originally "phonetic differences are psychologically interpreted as semantically contrastive [i.e. when they become phonemes, JK] speakers learning the language will attempt to account for such alternations morphologically" (p. 327).^ Schane (1971:512) anticipates, to a certain degree, natural generative phonology in recognizing that the rules of generative phonology "have basically two different effects or functions - what I shall call MORPHOPHONEMIC and PHONETIC." Among the first he lists the rule that accounts for electric, electricity in English, because both /k/ and /s/ are underlying segments, and they 10
Scott (1975:53-7) defends a similar view, listing evidence for the claim that children learn alternations morphologically, and not phonologically.
11 may both occur before /i/. Nasalization in French constitutes a second type of irorphophonemic rule, one where a derived feature, vocalic nasality, became contrastive on the surface because the conditioning environment was destroyed (by prepausal and preconsonantal N-deletion): [bo] 'good' (masc.), [bo] 'beautiful' (masc.).
Third, rules with morphological conditioning are included,
like the English verb-noun stress minimal pair perm-it - permit. Traditional allophony, of course, is the locus of Schane's phonetic rules. Schane's purpose was to establish which surface phonetic contrasts are "phonemic."
Distinguishing tvro categories of rules makes this possible:
"Only those phonetic differences resulting frcm morphophonemic rules are phonemic" (p. 514). We may turn this statement around and say that phonemic differences produce norphophonemic rules.
Thus, Schane may be said to pro-
pose, indirectly, and never mentioned in the article, a theory of morphologization based on the role of the phoneme.^
11
Linell (1976a:266, 270) argues at length for the phonemic level of representation, which "characterizes our way of making little-detailed perceptual structurings and ... conditions our way of perceiving and structuring speech."
12
2.
Kruszewski's Ueber die Lautabweohslung
2.1.
Background
In spite of repeated references to the contribution of M. Kruszewski to linguistic theory by Roman Jakobson and others, this Polish linguist of the second half of the last century has been almost totally ignored in the linguistic literature during this century, being even more unknown than his teacher at the University of Kazan, J. Baudouin de Courtenay.
Yet it is Kruszewski
who "per la prima volta nella letteratura linguistica propone m a teoria e una classificazione delle alternanze fonetiche" (Jakobson 1965:14).
This the-
ory of phonetic alternation was published in the slim, 41-page, monograph entitled Uebev die Lautabweohslung in 1881.
It is the German translation of
Kruszewski's theoretical introduction to his master's dissertation, done under Baudouin at Kazan, On the question of Guna: an investigation in the field of Old Slavic vooalism. The translation's purpose was to reach scholars in western Europe, but Kruszewski had to pay publishing costs at Kazan, since his work had been rejected by a German publisher, calling it more "methodology" than linguistic science (Kilbury 1974:237). It is unfortunate that recent phonological studies, in particular, make no reference at all to his work.
A careful reading of it leads one to agree
whole-heartedly with Jakobson's assessment of Kruszewski's writings, "truly full of mighty anticipations in the theory of language, fresh and fertile working hypotheses and penetrating observation" (Jakobson 1971:404).
In fact,
Ueber die Lautabweohslung reads incredibly modern, containing in essence, and often in detail, a sketch of synchronic morphophonology, reminiscent of recent "discoveries," and, I shall claim, a theory of morphologization.
Due to its
importance, and due to its unfamiliarity in the current literature, a detailed
12 critical analysis of Kruszewski's theory of alternations has long been overdue. 12 Kilbury (1974:250-2, 1976:18-20) presents an English translation of Kruszewski's classification of alternations, but only a very cursory analysis.
13 2.2.
Synchronic analysis
Before undertaking the presentation of phonetic alternations, Kruszewski situates not only phonology but the whole of language within a physiologicalpsychological tension described thus: (p. 6) [My own, rather literal, translation follows each German excerpt] "Die Sprache bildet ein Etwas, welches in der Natur ganz abgesondert dasteht: sie ist eine Kombination von physiologisch-akustischen Erscheinungen, regulirt durch physische Gesetze, mit unbewusstpsychischen Erscheinungen, regulirt durch Gesetze völlig anderer Art. Daraus entspringt eine der wichtigsten Fragen: in welchem Verhältnis stehen zu einander in der Sprache überhaupt und speciell in der Phonetik diese zwei verschiedenen Prinzipien, d.h. das physische und das unbewusstpsychische?" [Language forms (a) something which exists totally apart in nature: it is a combination of physiologically-acoustic occurrences, regulated by physical laws, with subconsciously psychological occurrences, regulated by laws of a completely different kind. Out of this arises one of the most important questions: what relationship obtains between these two different principles, i.e. the physical and the subconsciously psychological, in language in general and in phonetics in particular?] This quotation is indicative of Kruszewski's (very modern) definition of linguistics, which for him, but not for the contemporary neogranmarians, consisted of the "unveiling of laws" (Enthüllung von Gesetzen, p. 3), valid for all languages, valid for all aspects of a language, both diachronically and synchronically. The concept Lautabwechslung is introduced as synchronically "incomprehensible" (unbegreiflich, p. 8) in three ways: (1) concerning the "anthropophonic (= phonetic) relationship" {anthropophonisahe Beziehung) of alternating sounds.
Kruszewski points out that two al-
ternating elements may be so dissimilar phonetically that "die Annahme eines unmittelbaren Ueberganges des einen Lautes in einen anderen dan gesunden Menschenverstände zuwider ist" (p. 8) [
that the assumption of
a direct change-over of one sound into another would go counter healthy Firth (1957:1-2) contains a succinct description of Ueber die Lautabwechslung with reference to a definition of the phoneme. Koerner has, like Jakobson, repeatedly pointed out the importance of Kruszewski (Koerner 1972, 1973b:16-8), mentioning also his (probable) influence on Saussure (1973a: 148-65). Stankiewicz (1972) includes an English translation of Baudouin's Versuch einer Theorie phonetischer Alternationen, which is very comparable to Kruszewki's monograph, but which, by Baudouin's own admission, is less "logical," and whose comprehension is hampered by the author's diffuse style
14 common sense.]13 (2) concerning the causes (Ursachen) or conditions (Bedingungen) of the alternation; (3) concerning the necessity (Nothwendingkeit) of the alternation. All three factors are then utilized in the classification of three principal types of phonetic alternation. FIRST CATEGORY (pp. 9-12) As an example of the first type, Kruszewski offers the German alternation [s] - [z] (HauSj Seele, g ewe sen, unser) which is (a) physiologically (phonetically) conditioned and exceptionless, and (b) has no connection with morphological categories, and (c) applies in the pronunciation of foreign words as well: 14
Simbirsk,
Conser-
vato%re, etc. (p. 9) .
In the context of another case of first category alternation, the Russian palatalization of /t/ to [t1] before front vowels /e/ and /i/, Kruszewski makes the important point that the [t1] before /e/ and /i/ does not mean that a Russian is unable to pronounce a non-palatalized /t/ in that position, only that he does not do it. Such "abnormal" utterances can only be produced "under (the) condition of the15 participation of the will" (untev der Bedingung der Willenstheilnahme,
p. 10).
Using x to symbolize a condition under which s, a given sound, is possible, and x^ as a condition incompatible with s, but compatible with s^, Kruszewski presents the following succinct, yet clear and complete, distinguishing characteristics of the FIRST CATEGORY:
13
14
15
and multitude of new technical terms with "vacillating semantics" (Koerner 1973a:144). An edition of Kruszewski's writings, including an English translation of Ueber die Lautabwechslung by Robert Austerlitz, ed. by E.F. K. Koerner, has been announced. One is reminded of recent abstract underlying representations offered in some transformational generative phonologies which, to paraphrase Kruszewski, certainly "violate common sense" ideas about phonological systems. We could today formulate this (allophonic) rule as, roughly,
This notion is taken up by Stampe (19/3) as the "suppression of natural processes." The latter are "expressions of the language-innocent speech capacity, and they are not overcome by the language learner unless they confront counterinstances in the language he is learning" (p. 27).
15
(1) "Unmittelbare Bestimmbarkeit und Vorhandensein der Ursachen der Lautabwechslung. Bei jeder Abwechslung irgend welcher Laute s // s^ treffen wir irgend welche Abwechslung von phonetischen Ursachen oder Bedingungen x // x^ an." [immediate determinacy and existence of the causes of the alternation. With each alternation of some sounds s // s^ we encounter some alternation of phonetic causes or conditions x // x^.] (2) "Allgemeinheit der Lautabwechslung. Die Lautabwechslung s // s^ kommt allgemein vor, d.h. die Laute wechseln bei erwähnten Bedingungen in allen Worten ab, ohne irgend welche Beziehung darauf zu welchen morphologischen Kategorien sie auch gehären mögen." [Generality of the alternation. The alternation s // s^ occurs generally, i.e. the sounds alternate, under the conditions mentioned, in all words without any reference to which morphological categories they may belong.] (3) "Notwendigkeit der Lautabwechslung. Die Lautabwechslung s // s unter den Bedingungen x // x^ ist notwendig, und lässt absolut keine Ausnahmen zu, d.h. die Erscheinung s neben x und s^ neben x ist unmöglich." [Necessity of the alternation. The alternation s // s^ under the conditions x // x^ is necessary and permits absolutely no exceptions, i.e. the occurrence of s alongside x^ and that of s^ alongside x is impossible . ] (4) "Nahe anthropophonische Verwandtschaft der abwechselnden Laute. Die abwechselnden Laute s // s sind anthropophonisch sehr nahe verwandt, oder, genauer, sie sind Modifikationen eines und desselben Lautes." [Close anthropophonic relationship of the alternating sounds. The alternating sounds s // s^ are anthropophonically very closely related, or, more exactly, they are modifications of one and the same sound.] Kruszewski concludes this section by stressing that criteria (1) - (3) are decisive for the FIRST CATEGORY; however, one of the three would be sufficient, since it entails, by definition, the others.
Following Baudouin de Courtenay,
Kruszewski labels the members of such an alternation "divergents" (Divergenten). SECOND CATEGORY (pp. 12-18) In this category, Kruszewski considers German rhotacism, illustrated by the pair war - gewesen.
He innediately states that the cause of a possible syn-
chronic rule s -»r is no longer present in the language, but can only be determined historically.
Etymologically, /s/ is primary, /r/ secondary, having
developed intervocalically fron /s/.
In the modern language, however, both
intervocalic /s/ and non-intervocalic /r/ exist, creating many exceptions which are to be explained "as having arisen by secondary phonetic and morphological processes" (als durch seoundäre phonetische und morphologische Processe entstanden zu erklären) (p. 13).
Kruszewski introduces the term phoneme (Pho-
16 The FIRST CATEGORY is directly comparable to natural generative phonology's P-rules which "describe processes governed by the physical properties of the vocal tract ... predicted on universal principles ... all P-rules are natural rules" (Hooper 1976:16-7). 17 German rhotacism, in Kiparsky's terminology, would be opaque (i) and (ii).
16
nem) for an alternating mereber of the SECOND CATEGORY, while the word sound (Laut) is reserved for the FIRST CATEGORY
(p. 14, fn. 1).
A parallel example
from Russian, the alternation k -* c (and g •» z), receives a similar explanation
(pp. 16-7), i.e. as non-phonetically conditioned.
Using now s // z to
symbolize an alternation, Kruszewski outlines the distinguishing features of the SECOND CATEGORY: (1) "Unmöglichkeit einer unmittelbaren Bestimmbarkeit der Abwechslungsursachen (-bedingungen) und Möglichkeit ihres Nichtvorhandenseins in einzelnen Fällen. Bei der Abwechslung der Laute s // z können die Ursachen (Bedingungen) x // x nur auf dem Wege historischer Untersuchung ermittelt werden. Ausserdem kann in den Worten, in welchen die Lautabwechslung s // z vorkommt, dasjenige nicht Vorhandensein, was die historische Untersuchung als Ursache dieser Abwechslung ermitteln wird." [impossibility of an immediate determination of the causes (conditions) of the alternation and possibility of their non-existence in individual cases. In the alternation of the sounds s // z the causes (conditions) x // x^ can be established only by means of historical examination. In addition, words in which the alternation s // z occurs may not contain that (element) which the historical examination will establish as the cause of this alternation.] (2) "Nichtvorhandensein der Nothwendigkeit der Abwechslung. Das Auftreten des Lautes s bei der Bedingung x^ und des Lautes z bei der Bedingung x ist möglich." [Non-existence of the necessity of the alternation. The appearance of the sound s alongside condition x^ and that of the sound z alongside condition x is possible.] (3) "Nichtvorhandensein der Allgemeinheit der Abwechslung. Die Abwechslung s // z erscheint theilweise mit gewissen morphologischen Kategorien (f // f ) verknüpft." [Non-existence of the generality of the alternation. The alternation s // z appears partly connected with certain morphological categories (f // f^).] (4) "Entfernte anthropophonische Verwandtschaft der abwechselnden Laute. Die abwechselnden Laute s // z stehen meistentheils in einer entfernten anthropophonischen Verwandtschaft zu einander." [Distant anthropophoThe alternating sounds nic relationship of the alternating sounds. s // z have mostly a distant anthropophonic relationship to one another.] The first two characteristics are decisive, less important the third, least important the fourth.
In fact, (1) and
(2) are "inseparable"
(unzertrenn-
lich) . THIRD CATEGORY
(pp. 18-21)
German umlaut best exemplifies
the third type of alternation, for which
Kruszewski reccrrmends the term Fovmabweohslung
instead of
Lautabweohslung.
The original cause of the alternation has been totally obscured, recoverable only historically.
Yet this alternation of forms may be alive, and unavoidab-
le, in certain nouns: a plural of Abraham
would be Äbrahämer
The alternation is displayed thus: (p. 19)
(Viennese dialect).
17 Subst.
a
Subst.
Stamm
o
Staitm
[noun
u
[noun
stem] \au
stem]
ä 0
+ erj
V äu
The THIRD CATEGORY has features (1), (2), and (4) in canton with the SECOND CATEGORY.
Characteristics (3) and (5) are:
(3) "Die Abwechslung der Laute s // z ist mit der Abwechslung der morphologischen Kategorien f // f verbunden." [The alternation of the sounds s // z is tied to the alternation of the morphological categories f // f, -] (5) "Die Erscheinung des Lautes s in der Form f oder des Lautes z in der Form f ist unmöglich." [The occurrence of the sound s in form f or that of sound z in form f is impossible.] The members of both the SECOND and the THIRD CATEGORIES are termed "correlatives" (Korrelativen). It is clear that for Kruszewski types two and three are more closely related than either one is to type one.
Yet, the THIRD CA-
TEGORY is crucially distinguished from the SECOND CATEGORY by having exclusively morphological conditioning, which co-exists with phonetic conditioning in the SECOND CATEGORY, i.e. characteristic (3), and by having functional cha18 racter, i.e. characteristic (5). 2.3.
Diachronic analysis
The establishment of three synchronic categories of morphophonological alternation presupposes a diachronic connection, not unexpectedly going from (1) to (2) to (3).
Kruszewski gives a very explicit description of this evoluti-
on, thus proposing, one may say without exaggeration, a theory of morphologization (pp. 22-3). "Irgend ein Laut s ist mit der Bedingung x verträglich, aber mit der Bedingung x nicht verträglich: diese Bedingung afficiert den Laut s, d. h. er erhält irgend eine geringe Veränderung: verwandelt sich in s^ (kombinatorischer Wandel). Es verschwinden also in allen Worten die Combinationen s // x , und wir erhalten nur die Combinationen s // x und s // x , wobei die Verbindung zwischen s und x eine kausale 18
Morphophonemic rules of natural generative phonology " change phonological features in environments described in morpho-syntactic or lexical terms. These rules must refer to morphological or syntactic categories, such as plural, past, or noun, verb, or the rules refer to arbitrary lexical categories, such as conjugation classes or classes designated arbitrarily by diacritics" (Hooper 1976:15). MP-rules may be equated to both of Kruszewski* s categories two and three; the latter adds the subtle distinction between "partially morphological" and "totally morphological" alternations.
18 sein wird, d.h. die Bedingung x^ verwandelt den Laut s in den Laut s^ (Daraus ist leicht ersichtlich, weshalb die Abwechslung s // s keine Ausnahme zulässt). Dies wird für Worte, die die betrachteten Combinationen aufweisen, die erste Periode ausmachen. Die Laute sind aber auch einem spontanen Lautwandel unterworfen, der sich besonders stark und schnell in denjenigen Lauten äussert, die schon irgend einer Afficierung unterlegen sind. Deshalb verwandelt sich der Laut s , sich ständig verändernd, in irgend einen Laut z: fl " *l z Die in der ersten Periode entstandenen Worte werden also in dieser zweiten Periode folgende Combinationen haben: s // x, z // x^. Die Verbindung zwischen der Bedingung x und dem Laute z ist schon keine kausale mehr, d.h. diese Bedingung kann nicht den Laut s in den Laut z verwandeln; sie ist eine Verbindung der Coexistenz. Da die Verbindung zwischen dem Laute s und dem Laute z zufolge der bedeutenden Degeneration des letzteren nicht mehr gefühlt werden kann, und da die Verbindung dieses Lautes mit der Bedingung x anthropophonisch möglich ist, so können verschiedene unbewusst psychische Processe noch die Verbindung z // x hervorrufen. In solch einem Zustande werden sich nur die sich in der ersten Periode gebildeten Worte befinden, und somit sehen wir in ihnen den Anfang der Verbindung von phonetischen Erscheinungen [Some sound s is compatible with conmit morphologischen Kategorien." dition x, but incompatible with condition x^: this condition alters sound s, i.e. it takes on some minimal change: it changes into s^ (combinatory change). Thus combinations s // x^ disappear in all words, and we obtain only combinations s // x and s^ // x^, where the combination between s^ and x will be causal, i.e. condition x changes sound s into sound s^ (Hence it is easily understood why the alternation s // s^ permits no exception). This will constitute the first phase for words which display the combinations discussed. But sounds are also subjected to spontaneous sound change, which is expressed especially strongly and quickly in those sounds which have already undergone some alteration. Therefore, sound s^, changing constantly, transforms itself into some sound z:
Jl
7/ X
1
z Those words which were created during the first phase will thus have the following combinations in the second phase: s // x, z // x^. The connection between condition x^ and sound z is no longer of a causal nature, i.e. this condition cannot change sound s into sound z; it is a connection of coexistence. Since the connection between sound s and sound z can no longer be felt, due to the important degeneration of the latter, and since the connection of this sound with condition x is anthropophonically possible, various subconsciously psychological processes may still conjure up the connection z // x. Only those words formed during the first phase will be found in such a state, and thus we discover in them the beginning of the connection of phonetic occurrences with morphological categories.]
The last sentence is a clear statement on incipient morphologization, probably better expressed by Kruszewski than in modern accounts.
The complete e-
volution described in the preceding text is visually reinforced in Tafel 1, which follows p. 36:
19
Kruszewski: Tafel 1 Divergence: The sphere of the phonetic law German
Slavic [k] // [k'l
[g] // [g']
20 When the sound changes enter the area of the ellipse, morphologization takes place.
Phonetically, they have passed through one and a half phases, i.e.
the combinatory phase (Phase one) and the quantitative half of Phase two, the spontaneous phase.
The quantitative-qualitative division of the spontaneous
phase is added in the text only later (pp. 33-4).
It is difficult to imagine
a more specific pin-pointing of the boundary between phonetic and (partially) 19 norphological conditioning. The change-over frcxn the SECOND CATEGORY, reached after both the combinatory and spontaneous processes, to the THIRD CATEGORY can be ascribed, according to Kruszewski, to the essential "instability" (Unbeständigkeit) of type two correlatives, which is due to: (pp. 23-4) (1) the absence of anthropophonic connection between sound z and condition x^; (2) the absence of anthropophonic connection between sound s and sound z; (3) the absence of generality in the sense of the independence of a vrord belonging to a certain norphological category. But, Kruszewski continues, this instability in only apparent and a transition to the third diachronic phase: (p. 24) "Alles, was aufgehört hat anthropophonisch unumgänglich nothwendig zu sein, erliegt der Wirkung unbewusst psychischer Factoren, die in Beziehung zu den Korrelativen (d.h. Lauten, die mit einander nur durch eine Coexistenzverbindung vereinigt sind) entweder zerstörender oder befestigender Natur sein können; aber sowohl zerstörend als auch befestigend ringen diese Factoren zur Einsetzung der vollkommensten Ordnung und Einfachheit in der Sprache, als zu ihrem Endziel." [Everything which has ceased to be inevitably anthropophonically necessary succumbs to the effect of subconsciously psychological factors, which in relation to the correlatives (i.e. sounds which are united with one another only by a coexistence connection) may be either of a destructive or a reinforcing nature; but whether destructive or reinforcing, these factors strive for the establishment of a most complete order and simplicity in language as their final goal.] 19
Natural generative phonology makes the claim that all "new alternations in a language arise in phonetic environments" (Hooper 1976:84) although "the implementation of a sound change is subject to diverse influences that create complex situations" (1976:110). Schematically, historical change in NGP may be displayed as follows (Hooper 1976:86): via rules morphophonemic rules phonological rules grammar new rules universal phonetic tendencies
21
By the "destructive" aspect, also called (morphological) assimilation (Assimilation), Kruszewski means what would today be referred to as paradigmatic levelling, illustrated by the Russian paradigm 'to bake,' in which in the standard version [k] alternates with [c], while colloquially the [k] has been regularized, and 'to be able,' which shifts fron a [g] - [z] alternation to levelled [z] in a dialect.
From German, Kruszewski cites the switch from
QHG lesan, las, larun to MG lesen, laSj gelesen 'to read,' and verliuse, Verio Sj verlorn to verliere, verlor, verloren 'to lose.'
Both languages show the
expansion of both the etymologically primary segment ([k] and [s]) and the his*
torically secondary segment ([z] and [r]).
20
The "destructive" development of correlatives does not, quite obviously, lead to the THIRD CATEGORY, but rather eliminates type two alternations, making the at first glance strange term an aptly chosen one.
It is rather the
"reinforcing" effect of morphological factors, also called "differentiation" (Differentiation) which must be chosen.
German umlaut (in the plural) main-
tains and functionalizes the alternation into the THIRD CATEGORY: (p. 26) (f) Mann
// (f^) Männer
Loch
//
Löcher
Buch
//
Bücher
While Tafel 1 gave a schematization of how a phonetic change becomes morphologized, i.e. how it enters the "magic circle" (Zauberkreis), Tafel 2 sketches the further evolution of a morphological alternation, i.e. how it leaves the , 21 iragic circle:
20
21
Kruszewski's theory is more specific, implying the role of the phoneme in morphologization, since the sound after the quantitative step of phase two seems to be equatable to the phoneme, a term used by Kruszewski to designate a member of the SECOND CATEGORY. Cf. Baudouin de Courtenay (1895:66; 110 ff.) In terms of rule loss, we may explain the developments as the loss of the (non-inverted) rules k -» c and s r, and the loss of the (inverted) rules z -» g and r -»s. Cf. 3, below. New divergents are created by the loss of correlation, since the phonetically caused German [s]-[z] and Russian [kl-fk") alternations persist in the respective languages.
22
Kruszewski: Tafel 2
23
3.
A comprehensive theory of norphologization
3.1.
Data and rules
Table I (p. 24) displays some of the most important sound changes in historical French phonology (cf. Klausenburger 1974b:19-25; 57-8). The interpretation of the data for a theory of morphologization will depend crucially on the division into ALDCMDRPHS and UNIQUE MORPHEME, which are, as can be seen, not always overlapping loci. The exanples under UNIQUE MORPHEME illustrate best the sound changes, considered here in the traditional sense, free of contamination with synchronic morphophonological alternation. They may be sunrarized itDst succinctly without distinctive feature notation as follows:22 LENITION: (1) Latin - Old French: p -» v / V _ V (2) Old French - Modern French: no additional change VELAR PALATALIZATION: (1) Latin - Old French: k
c /_a
(2) Old French - Modern French: c
s / context-free
DEVOICING: (1) Latin - Old French: v f/_ # (2) Old French - Modern French: no additional change DEGEMINATION: (1) Latin - Old French: tt -» t / V _ V (2) Old French - Modem French: no additional change NASALIZATION: (1) Latin - Old French: 3 - O / _ N (2) Old French - Modem French: 3 -»3/ _ NV (denasalization, not
DIPHTHONGIZATION (a): (1) Latin - Old French: a
ay /
shown in pont) stress | open syllable
(2) Old French - Modem French: ay -» £ / context-free (Monophthongization) 22
Two diachronic spans. Late Latin to Old French, and Old French to Modern French, will suffice for our purposes here. In addition, for this presentation, the changes are given for each segment specifically, ignoring the over-all processes. For example, lenition, of course, affects all voiceless (and voiced) stops, not just /p/.
24
Table I Data to illustrate a theory of morphologization 21
ALLOMORPHS Mod. Fr.
Old Fr.
UNIQUE MORPHEME L. Latin
Mod. Fr.
Old Fr.
L. Latin
savoir
saveir
sapere
[savwar] [saveyr] [sapere] sec [stk]
sec
secco
[stk]
[sekko]
champ
champ
sèche
seche
secca
[£â]
[camp]
[st£]
[seca]
neuf
nuef
novo
[nwef]
[nívo]
neuve
nueve
nova
oeuf
uef
ovo
[wef]
[avo]
mettre
mettre
mettere
[m)# #p5u(r))# #P»U(1))#
/
[p»u(i>) ]
#pa #pi #p i #pa
+ l + s# + I + s# + l +i# + i + ;#
[pSfj]
98
By what criteria derivations like these can be characterized as "simple" boggles the imagination! We will not consider such historical recapitulations as serious synchronic analyses. Let us, therefore, focus on the concrete approaches. Lipski (1973a:76) motivates an /e/-epenthesis rule, 0 -*e / C _ s#, for nouns like No. 4-6 in the table by phonotactic considerations, claiming that "Portuguese phonology tolerates no word-final consonant clusters (p. 75)." A second group of plurals, No. 7-9, 24, are accounted for by the rule 1 -» i / V _ s#
(p. 78),
the same rule, plus one of shortening, affecting No. 10 also; No. 11, however, is considered "irregular," as is the most notorious set. No. 15-21, for which no atterrpt at rule governedness is made, as "the plural of a word in -ao is completely unpredictable, and such forms are learned as a list by native speakers and foreign students alike (p. 79)." Leite's concrete solution includes, in addition to the actual pluralization rule, (1) i-insertion (2) i-lowering (3) lateral vocalization (4) vowel contraction for No. 4-11 in our data (p. 77), and two morphological rules, a major one, -ao -» -oes, and a minor one, -ao -» -aes (p. 78). The first one would account for No. 15 and 16, the second for No. 19 and 20. No. 17 and 18 are marked [- rule -ao -t -oes], as they sinply add an /s/, like No. 1-3, 13, 14, 22, 23, 25. The choice of rrajor and minor rules for the -ao plurals is based on a clear tendency to prefer the -oes ending, being used frequently instead of the correct -aes, -aos. Within the theoretical apparatus of this study, Leite's rules are very appropriate, written specifically for the data given as follows: (1) i-insertion (Zi - i /
_ s
(SM-I)
(2) i-lcwering [/stress]0-1
3
(S
^NI)
(3) lateral vocalization (and glide formation) 1 -* y / _ s
(SM-NI)
(4) vowel contraction ii
i
(P-NI)
Three simple derivations are the following: (a) No. 7: (carpare with the abstract derivation, above) /kanal + s/ (3)
kanay + s [k»nays]
(pi.)
(b) No. 11: /fisil + s/ (2) (3)
fisel + s fasey + s [fsseys]
(c) No. 10:
i /funil + s/ (3)
funii + s
(4)
funi
f
+ s
[funis] To avoid the extrinsic rule ordering of (2) before (3) one may propose an alternate set of rules: (1') sane (2') sane (3') laberpial vocalization (and glide fornation)
^y/f-Mgh]-3 (4') lateral deletion L deletion
s
(SM
"NI)
(SM-NI)
/
[+ high] Now, derivations (b) and (c) would be: (b)
/fasil + s/ (2") (3')
fasel + s fssey + s [faseys]
(c)
/funil + s/ (4*)
funi
+ s
[funis] 53
These derivations omit various phonetic details of the actual pronunciation, like the two surface representations of /r/, [r], used her«for the apical flap, and [R] , employed in the table for a velar trill, or the other dia-
100
As can be seen, Portuguese pluralization offers another deletion (apocope) vs. epenthesis (insertion) choice reminiscent of Spanish, in this case, however, involving the vowel /i/. Epenthesis is to be preferred since (a) many exceptions to an i-apocope rule exist (No. 13, 14, and 23, for example), and (b) foreign borrowings behave as predicted by the i-insertion rule: club, alubes [klubis] bar, bares [baris] The rule for the nasal plurals, (a) -ao -» -oe(s) (b) -ao -» -ae(s) , nay be classified as either totally morphological or as via-rules, their inverted or non-inverted nature not being easily determinable.
It is in this
set of nouns that phonological changes have really created havoc, as a brief comparison with Spanish cognates demonstrates: Latin ending
Spanish
Portuguese
-CNE (S)
leaoion
leaaiones
liqao
-ANE (S)
oapitan
capitanes
aapitao oapitaes
Uqoes
-ANO (S)
hermano
hermanos
irmao
irmaos
Historically, apocope of final /e/ in the singular, but not in the plural, explains the merger of all singulars into -ao, and the retention of the threeway distinction in the plural.
Intervocalic /n/-deletion, in addition, des-
troyed the conditioning of a possible rule of /e/ (or /i/) epenthesis, comparable to the Spanish ^ - » e / n _ s , in the first tvro plurals.
Since the
Spanish rule qualifies as an SM-I, we are probably justified in tagging the completely morphological or via relationship of the Portuguese cognates as M-I's or V-I's.
lectal possibilities, as outlined in Leite (1974:85), /l/ as [w] word-finally, /s/ as [z] intervocalically and [s] pre-pausally, and nasalization.
101
6.2.
Diphthongization
6.2.1. Spanish The historical diphthongization of Latin open mid vowels U] and [5] to [ye] and [we], respectively, is synchronically retained in ffodern Spanish in verbal stems of the 1st and 2nd conjugations (cf. Table XXV, p. 102), since the conditioning factor, stress, opposes the singular and 3rd p. pi. forms to the 1st (and 2nd) plural. It is uncontroversial to use such paradigmatic data for the illustration of diphthongization, since they represent a "productive" relationship, while other evidence, as the one from derivational morphology (cf. Harris 1974:10-11), could be challenged with respect to rule-governedness, the concept of the via-rule being excluded from the latter. The major work on Spanish phonology, Harris (1969), says surprisingly little about diphthongization in verbal alternations (pp. 161-3). In a later article, Harris (1974) (see now also Harris 1977a) fills this gap, comparing the diphthongization rule of Standard Spanish, e 0
-*
ye" / we
[+ stress]
MINOR RULE (p. 12),
to the norphologized Chicano inversion. Harris recognizes the minor nature of the rule, on account of the many exceptions, duly pointed out (pp. 11-12). (In Table XXV, the most glaring violation of the rule is illustrated by the verb 'to sew' (C)). Yet, Harris does not reach the necessary conclusion, that the Standard Spanish rule, not (only) the Chicano one, has morphologized! (Cf. the connection of opacity and morphologization in this case, mentioned in 1, 1.1., above). Within the theoretical frame^work of this study, Standard Spanish diphthongization is classified as an SM-NI, something like: e 0
-r
ye we
/
1
[+ stress]
morphological information (verbs like oooer, etc.)
(cf. Anderson 1975:111) The levelling of alternations in Chicano Spanish is illustrated in the table by (E). Three structural differences separate this dialect from Standard Spanish (Saltarelli 1975:124):
102
54 Table XXV: Data for Spanish diphthongization Indicative
Subjunctive
1
cal[yé]nto
cal [ye] rite
cal [ e ] n t a r
2
cal[ye]ntas
cal[ye]ntes
A ' to heat'
3
cal[yé]nta
cal[ye]nte
4
cal [e] ntamos
cal[e]ntemos
6
cal[yé]ntan
cal[ye]nten
1
c[wé]zo
c[we]za
c[o]cer
2
c[wé]ces
c[we]zas
B ' t o cook'
3
c[wé]ce
c [we] za
4
c [o] cemos
c[o]zamos
6
c[wé]cen
c[we]zan
1
c [ o ] so
c [ o ] sa
c[o]ser
2
c [ o ] ses
c[o] sas
C ' t o sew'
3
c [ o ] se
c [ o ] sa
4
c[o]semos
c [o] samos
6
c[ó]sen
c[o]san
1
v[we]lo
v[we]le
v[o]Jar
2
v[we]las
v[we]les
D
3
v[wé]la
v[we]le
'to
4
v[o] lamos
v[o]lemos
Standard
6
v[we]lan
v[we]len
1
v[we] lo
#i r v[wejle
v[we]lar
2
v[we]las
v[we]les
E
3
v[we]la
4
v[we]lamos
v[we] le r 'l vlwejlenos
Chicano
6
54
Gloss
v[we]lan
r ' i v[wejlen
'to
fly'
fly'
The data are based on Harris ( 1 9 7 4 ) . Since not a l l speakers use the 2nd p. p i . ( 5 ) , t h i s form i s omitted. I t does not e x i s t in Chicano Spanish. Present paradigms o f the 1 s t and 2nd c o n j u g a t i o n s are included o n l y , the g l o s s t r a n s l a t i n g the i n f i n i t i v e . 3rd c o n j u g a t i o n analyses may be found in Harris (1977b) and Hooper (1976: Chapter 8 ) .
103
Standard
Chicano
Diphthongs
alternating
non-alternating
Stress (subj.)
alternating
non-alternating
First plural (subj.) -mos
-nos 55 Our nein interest lies in the diphthongs, of course." Despite the conplete disappearance of an alternation [o] - [we] in Chicano, Harris posits a minor, morphologized, rule of diphthongization for this dialect, which continues /o/ as underlying (1974:19): (a)
[+ stress]
(
X
[+ back] C 0 1 conj
Stem
Verb
(b)
Hie evidence from derivational morphology, like vuelar - volador 'flier,' which does show an alternation in Chicano, is weak, for reasons repeatedly discussed in this study. Within the verbal paradigms, there is no question that restructuring due to rule loss has occurred.
But it cannot be the loss
since this would yield to]'s throughout.56 Rather,
of the SM-NI, o -»we / the loss of the SM-I,
I [- stress] 1 ( morphological information \ ' 57 accounts for the Chicano levelling. we —o
55
56
57
/
The innovations in the present subjunctive are explained as one phenomenon by Saltarelli (1975:128-9) by assuming a syntactic re-analysis of the 1st p. pi. into two words, v[we]le#nos (Saltarelli actually uses trabajetinos), suggested by an existing "inclusive" exhortation like vamonos. Anderson (1975:111-12) describes an Argentinian dialect where such a loss occurred. Marsian Italian, described in Saltarelli (1977), shows a comparable situation. The historical changes, 11 fay! . ("+ stress "J (followed by /a/ in the next uj [awJ L+ open syllablej syllable), represent the SM-NI's in verbal and nominal alternations for older speakers (over 60) of this dialect, while the young generation (under 30) has levelled to [l,U], a development explainable by the loss of the SM-NI. One sample alternation: Standard Italian Marsian Gloss Older generation Younger generation nera - nero [nayra] - [nlr>] [nlra] - [nlra] 'black' Anderson (1975:114, fn. 3) cites Espinosa for forms from child language like piensamos, which would show the loss of the SM-I ye -re / ... Saltarelli (1975:128) also lists muerir 'to die' (Standard morir) and juegar (Standard jugar). 'to play.'
104 Tile essence of the explanation given here, making use of rule inversion, choosing one of the surface allomorphs, the historically derived one, as basic, is corrpatible with the conception of underlying forms of MP rules in "early" natural generative phonology (Hooper 1973). Hooper (1976:127-31) [discussed above,3, 3.1., fn.], now defends the suppletion model of Hudson (1975), labeled "disjunctive" by Harris (1977b). Spanish diphthongization is handled by Hooper's disjunctive approach and Harris' (1977b) "conjunctive" = diacritic approach in the following ways: Harris
e
ye we
o + D
/
[+ stress]
Hooper
(cf. exanples A and B of the table) /calj^] nt-/
Harris triggers diphthongization by the diacritic [D], while Hooper, following Hudson, adds "distribution" rules to the lexical representation given. Both solutions are abstract, since neither underlying forms are "pronounceable," a factor which weighs more heavily against Hooper than Harris, since the former advocates maximum concreteness in underlying representations, as expressed by the True Generalization Condition (1976:13)• In addition, both approaches handle the Chicano levelling badly. For Hooper (1976:168) "the levelling is accounted for by the mere loss of the mid vowel case in each alternation, (ye, we/ ye, we [+ s t r e s s ]
(
e
'°l
e
' ° c. Modern French From Table XXXI it can be gleaned that all the categories which irake up the norphologization scheme presented in 3, 3.2., are represented in Part II, ranging from the P-NI conception of French mute /e/ to the demorphologization of Latin nasal infixation (loss of M-I). The exact break-down is the following: (i) P-NI
1
(ii) P-I (iii) SM-NI (iv) SM-I (v) M-NI
1 8 10 3
(vi) M-I (vii) V-NI (viii) V-I (ix) Loss of SM-NI (x) Loss of SM-I (xi) Loss of M-NI (xii) Loss of M-I
4 3 2 4 6 0 2
118 These statistics include, of course, the irultiple accounts, which for some processes remained irreducible to one principal rule. The most important piece of information to be retained frcm this listing is that SM's constitute more than half (18 our of 32) of the formulations, a result in harmony with the premises of the theory of Part I. Another interesting detail concerns the almost even split between inversions (17) and non-inversions (15). Of the 12 losses registered (P's are not eligible, V's were considered only marginally), it may be surprising that 10 aire SM's, since loss seems a priori more likely for M's. Again, however, the theory proposed allows for loss at any stage of morphologi zation, when phonetic conditioning co-exists with morphological conditioning, as well as when morphological ccmditioning alone remains (cf. Dressier' s explication of rule loss, discussed below in C). An essential ingredient of the theory of Part I was the incorporation of Kruszewski's scheme of morphologization, his exact pin-pointing of the latter in the middle of the "spontaneous process" (cf. 2, 2.3.). It turns out that this insight was only infrequently applicable in the data analyzed in Part II. There is a sinple reason for this: Kruszewski's conceptualization enters, by definition, only in cases of "gradual" phonetic changes, for which in our data only Latin rhotacism, French nasalization, and Italian (Old French) velar palatalization qualify. The other diachronic sources involve either (1) "abrupt" phonetic changes, such as the consonant deletions for liaison and Old French pluralization, vowel deletions (apocope, syncope) for Catalan, Spanish, and Portuguese plurals and the French mute /e/, and diphthongizations, or (2) morphology from inception (Lachmann's Law and nasal infixation in Latin). The inappropriateness of Kruszewski's diachronic explanation to most of the data does not, hcwever, entail the same for his synchronic classification into three categories: these clearly may be equated to our P, SM, and M trichotomy. B.
Motivation
Two basic assumptions about allcmorphic alternation constituted the foundation of the theory of morphologization introduced in 3: (a) Phonetic and morphological conditioning co-exist, creating a redundancy; (b) the historically original or the historically derived variant may be underlying during a particular synchronic stage.
119
Building on these premises, we reason that the speaker may account for a given morphophonological alternation in as many as four different ways, containing itorphological conditioning singly, or phonetic and morphological conditioning together, with either non-inversion or inversion (P-NI and P-I are excluded from allomorphic variation). Each account being visaed as a potential abduction, theoretically equally possible, no explanation needs to be forthcoming as to "why" any of the four does occur. In fact, all rules must be considered equally plausible until some additional factor, which allows for a decision on the most frequent abduction, is discovered. Nevertheless, a hierarchy, ranking morphological conditioning ahead of phonetic conditioning, mist be established, the major motivation for it being contained in the direction of change within the four categories of rules, showing two one-way paths: Path (1): (P) SM -» M (V) Path (2) : NI I Path (1) is naturally attributed to the loss of phonetic conditioning (due to new sound change), thus subordinating the latter to morphology. Path (2) became detectable only if the levelling of an alternation in the direction of the derived segment took place. The complete synchronic and diachronic picture, including the potential loss of each type of rule, may be displayed as follows:
SM-NI Path
»
Path (1) (1) Path
Loss «L
TLOSS
M-NI
(2)
31-1 loss'
M-I Path (1)
•
A theory of "multiple abductions" does not, and cannot, predict that a particular alternation will undergo development (1) or (2) , or both (1) and (2). One nay propose that semantic criteria could enable us to irake predictions in this matter. Let us, therefore, now reconsider the dynamics of the data of Part II in the light of semantic properties. The most easily appreciable evolution of rrorphophonological processes is one that results in paradigmatic levelling of the alternation, described in this study exclusively by rule loss. Twelve cases of such levelling were investigated in Part II:
120
1. Latin rhotacism: r ... s -* r ... r Exarrple: honos honor honoris honoris 2. Latin nasal infixation: N ç6 -» N Example: rumpo runpo rupi -• *runpi ruptan *rumptum
N
3. Modem French liaison: (a) 0 C ç( 0 (b) C ... jz* C ... C Exarrples: (a) pas encore pa^ encore (b) cin^ -* cinq 4. Modem French denasalization: V ... V •* V ... V Example: m[a]n ami •* m[3]n ami 5. Chicano Spanish diphthongization: Diphtnong Monophthong -* D Exairple: vuela vuela volamos vuelairos 6. Old French diphthongization : (a) D ... M -» D ... D Exanples: (a) aime amans (b) lieve
(b) M ... D -» M ... M aime aimons lève
levons levons 7. Italian velar palatalization: (a) k ... c -» k ... k n-\
*
»
*
(b) j ... g j ... j Exanples: (a) baroo banco *banci banchi (b) fungo fungio fungi fungi 8. Old French velar palatalization: (a) g (k) ... j -» g (0) ... g (b) g (k) ... 3 ^ z ... z Examples: (a) lore longe (b) lare large
long longue large large
D
121
These exairples happen to illustrate some of the rrcst important semantic (morphological) categories recurrent in all languages (case, tense, person, nunber, gender), all of which, fortunately, allcw fairly uncontroversial determination of a "semantically primitive" (unmarked) vs. "semantically secondary" (marked) dichotom/. Table XXXII (p. 122) summarizes the levellings making use of these considerations (the nunbers refer to the listing given). They can be grouped into three types: A. Cases of levelling in the direction of the semantically primitive allomorph: 5 (3 inversions, 2 non-inversions) No. 2 - Latin nasal infixation No. 5 - Chicano Spanish diphthongization No. 6a - Old French diphthongization No. 7a - Italian velar palatalization No. 8a - Old French velar palatalization B. Cases of levelling in the direction of the serrantically secondary alloirorph: 4 (3 inversions, 1 non-inversion) No. No. No. No.
1 - Latin rhotacism 6b - Old French diphthongization 7b - Italian velar palatalization 8b - Old French velar palatalization
C. Cases of levelling with no obvious seirantic criteria involved: 3 (2 inversions, 1 non-inversion) No. 3a - Modern French liaison No. 3b - same No. 4 - Modern French denasalization These results do not point to a crucial role of semantic criteria, often emphasized in traditional studies of analogy (cf. Kurylowicz 1947; for a complete "state-of-the-art survey," see new Anttila 1977) and also recently (Vennerrann 1972b, Baxter 1975, Hooper 1978) . In addition, a correlation between unmarked morphological categories and rule inversion is not apparent. What is particularly m/sterious in the table is the double direction of levelling in the same process three different times (6 - 8)! Such evidence may clinch the case for the non-role of semantics in morphological levelling. It is only fair to add, hewever, that the choice of data influences the decision on the role of serrantics. In the data analyzed in the case of Modern French liaison, for instance, no obvious morphological-seirantic factors could be isolated. However, other evidence undoubtedly relevant to liaison, such as
122
rH l In O í 0) c •rH C rH •H iH 1 HI C 0
>
UH 0
Ul
>
G c 0 0 •H - H •P ui
C 0 •H Ul
a)
c 0 •H U1 h a)
C 0 •H Ul H a)
c •H
c -H
c •H
>
>
>
c 0 •rH ui >h ai
>
c •H 1 C 0 c
c 0 •H Ul h a)
C 0 •H to
c 0 •H Ul
ai
0)
c •H
c •rH
c •H
c •rH P C 0 c
Ol m
Ul c 0 Ul >H o
>
U
>
^
>
>
ai
>
c 0 •H Ul IH ai
>
c •H
c 0 •H Ul IH 11
>
C
•ri 1 C 0 c
c 0 •rH Ul VH Cl
>
c •H
c CJ ai h 0 •rH c •H O •H ui
>
rJ m O in ü> c
o •H +J
Ol c -H rH rH CD
>
ai i-i
U -H UH P C 0 C 0 •H p o 0) n •H
S e CJ Ul
&
o,
I
U -H P C 1 e 0 P 0 .c u -H 13
ui ai m ra Ü
Ul 01 U) m u n 0) •fi p o
w a) ui c
•H P 10 C -H E 0 c
•p c ai ui a u
ft
u a s: p o
!H 10 -H 3 Oi C •H Ul
«—i io IH 3 rH
ft
ai c •rH rH 3 U Ul
c •rH C •rH g