Insights in Germanic Linguistics: I Methodology in Transition 9783110810868, 9783110143591

Citation preview

Insights in Germanic Linguistics I

W G DE

Trends in Linguistics Studies and Monographs 83

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Insights in Germanic Linguistics I Methodology in Transition

edited by

Irmengard Rauch Gerald F. Carr

Mouton de Gruyter Berlin · New York

1995

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin.

® Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress

Cataloging-in-Publication-Data

Insights in Germanic linguistics : methodology in transition / edited by Irmengard Rauch, Gerald F. Carr. p. cm. - (Trends in linguistics. Studies and monographs ; 83) Includes bibliographical references and index. ISBN 3-11-014359-3 (acid-free paper) 1. Germanic languages—Grammar. 2. Linguistics— Methodology. I. Rauch, Irmengard. II. Carr, Gerald F. III. Series. PD101.I57 1995 430—dc20 95-4819 CIP

Die Deutsche Bibliothek -

Cataloging-in-Publication-Data

Insights in Germanic linguistics / ed. by Irmengard Rauch ; Gerald F. Carr. — Berlin ; New York : Mouton de Gruyter. NE: Rauch, Irmengard [Hrsg.] 1. Methodology in transition. - 1995 (Trends in linguistics : Studies and monographs ; 83) ISBN 3-11-014359-3 NE: Trends in linguistics / Studies and monographs

© Copyright 1995 by Walter de Gruyter & Co., D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting and printing: Arthur Collignon GmbH, Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

Contents

Germanic linguistics in the Post-Modern Age Irmengard Rauch

1

Infinitival subject sentences in Gothic Stephen A. Berard

5

Lexicality and versification of Johann Heinrich Yoß: Observations on prosodic feature analysis David Chisholm

47

backen > buk/backen > backte: A study in grammatical variation Nada M. Cook

67

To the rescue of time in German tense Richard d'Alquen

87

On syntactic and pragmatic features of speech acts in Wulfstan's homilies Eugene Green

109

Argument-predicate structure in grammar and performance: A comparison of English and German John A. Hawkins

127

The definition of a grammatical category: Gothic absolute constructions Gregor Hens

145

Subject sich and expletive pro: Impersonal reflexive passives in German Robert G. Hoeing

161

Diminutive formation in Yiddish: A syllable-based account Neil G. Jacobs

169

vi

Contents

Grammar and grammars in seventeenth-century Germany: The case of Christian Gueintz Robert L. Kyes

185

The beginning and end of the (great) vowel shifts in Late Germanic Anatoly Liberman 203 Nonlinear phonology and the development of post-consonantal resonants in word-final position in West Scandinavian and Germanic B. Richard Page 231 Diachronic stratification of the Germanic vocabulary Edgar C. ΡοΙοιηέ

243

Formal and less formal rules Irmengard Rauch

265

BAG IV: Phonological Irmengard Rauch et al interference

275

Semantic motivation vs. arbitrariness in grammar: Toward a more general account of the DAT/ACC contrast with German two-way prepositions Michael B. Smith 293 The Double Object construction in the Germanic languages: Some synchronic and diachronic notes Rex A. Sprouse

325

Index

343

Germanic linguistics in the Post-Modern Age Irmengard Rauch

How do we gauge the current pulse of the pursuit of Germanic linguistics, if not by a diverse but well-integrated set of papers attesting a venerable discipline which, however, has made the transition into the Post-Modern linguistic world. A predominant hallmark of this new linguistic order is, of course, pluralism. While the centralization of Germanic linguistics, i.e., the research unity of East, West, and North Germanic languages, whether historical or contemporary, has become a fact, its centripetalism captures more and more Germanic languages and it does so with increasing methodological foci. Thus, in this set of papers, Sprouse appeals to English, Danish, Icelandic, Dutch, and German data in his study of "The Double Object construction in the Germanic languages ..." which he finds to be fundamental, whereas the oblique construction is due to the lexical specification of given verbs, within the Chomskyan Principles and Parameters approach. Three further contributions attest to the sustained intense interest in syntactic studies: Hawkins continues his contrastive studies of Modern German with Modern English in his contribution "Argument-predicate structure in grammar and performance ..." He explains how canonical verb position in English and German is accountable for left-to-right temporary ambiguity in the former, and lack of such ambiguity in the latter, in sentence processing. Modern German is also the object of Hoeing's "Subject sich and expletive pro ...", in which he shows the inherent cöntradiction between pro, a base-generated form, and expletive, a more surface-like phenomenon, in refuting an expletive subject for impersonal passives contrary to the Extended Projection Principle. The syntax of Gothic gains prominence in Berard's "Infinitival subject sentences in Gothic", in which he addresses evidence for the fact that in non-configurational Gothic, embedded subject sentences are not extraposed or verb complements as they are in configurational languages. Several morphosyntactic contributions reflect the richness of syntactic investigations and the incorporation of morphology into syntax/semantics. Dealing with the Gothic language, Hens views "... Gothic absolute

2

Irmengard

Rauch

constructions" with the aim of arriving at a prototype definition for the Gothic absolute data. In showing how the Bayesian inverse-probability theorem can be compatible with Lakoffian encyclopedic judgment, he derives features for the Gothic absolute which are more prototypical than others and which are not necessarily statistically predominant. Dual (two-way) prepositions of Modern German are depicted within prototype theory by Smith's "Semantic motivation vs. arbitrariness in grammar ..." He displays the range of meanings relative to a central change: no-change dichotomy depicted by the accusative and dative, repectively. The trajector or the landmark can mutate through the extensions induced by the accusative. Nada Cook taps Modern German informants in ascertaining the variable occurrence of backen>buklbacken > backte and similar verbs with dual inflections. She finds that while the speakers' selection of the strong or the weak variant reflects sensitivity to semantic-syntactic features, to the language register and to informant age, it evinces no direct correlation to verb frequency or to informant sex. "To the rescue of time in German tense" by d'Alquen challenges recent arguments that non-past time is not the unmarked feature but rather that various features such as probability, factuality, personal assurance, and objective assurance are primary in the case of the present and future tense morphology. In the case of the preterite and perfect tense morphology, time too is maintained as the Standard German speaker chooses from the northern and southern past morphology relative to varying discourse and textual features. Historical languages use forms the focal point in Green's "On syntactic and pragmatic features of speech acts in Wulfstan's homilies". He demonstrates the illocutionary and the perlocutionary forces indicative of Wulfstan's art of persuasion which seeks self-reflection on the part of his Old English auditors. Two morphological articles appeal to phonological criteria. In "Formal and less formal rules", Rauch approaches the morphology of the Old Saxon pronominal and strong adjective inflections, asking whether a segmentation into monophonic and "rootless" pronominal forms has psychological reality. Appeal to phonological criteria such as the facts that the Old Saxon pronoun is typically under weak stress but does not show the ravages of weak stress, that the pronoun does not alliterate, and that it can be found in enclitic position speak for its cliticization salience. By means of a Lexical Phonological approach to "Diminutive formation in Yiddish ...", Jacobs offers a solution to the cruxes generated by suffixal diminutive /, such as d-insertion in tautosyllabic nl and resistance to di-

Germanic linguistics in the Post-Modern Age

3

minutive formation by a word such as mojl 'ritual circumciser' through syllable-based cyclic rules. Four contributions center on phonology. In "Nonlinear phonology and the development of post-consonantal resonants . . . " Page determines that the Modern Icelandic split between voiceless nonsyllabic post-stop η or / (e.g., takn 'sign') and syllabic r which engenders epenthesis after stop (e.g., akur 'field'), a split mirrored also by Old Norse and Old High German, respectively, is due on the one hand, to the resonants' position in the sonority scale, and, on the other hand, to language-specific Goldsmithean phonotactic syllable licensing rules. In " B A G IV: Phonological interference", Rauch et al. study a critical set of German sound features in data elicited from native German adults and first-generation GermanAmerican children residing in the San Francisco Bay area. Relative to English, German umlaut, vowel quality and length, the /I/ and /r/ sounds, initial /s/ and final devoicing are targeted; production ability for vowels under primary stress is an adult feature, while perception ability to distinguish acoustically near-homophonous sets of German—English words characterizes the "Kinderlect". Liberman demonstrates in "The beginning and end of the (great) vowel shifts in Late Germanic" that the English Great Vowel shift and similar shifts of bimoric vowels in Germanic sister languages had no beginning in the sense that such vowels always had diphthongal variants. The vowel shifts were in reaction to prosodic end-syllable weakening and resulted in glided diphthongs with no monophthongal alternates. Prosodic features also play a major role in Chisholm's "Lexicality and the versification of Johann Heinrich Voß . . . " In comparing the poetry of Voß with his eighteenth-century German contemporaries such as Klopstock, Goethe, and Hölderlin, he observes Voß' peculiar violation of the verbal material licensed in trochaic sequences as well as his preference for lexical constraints over stress constraints in hexameters. The Germanic lexicon is addressed by Polome. His "Diachronic stratification of Germanic vocabulary" offers insights into the vast area of the Germanic lexicon over several millennia. Most ancient is the Germanic vocabulary shared with Balto-Slavic; the Germanic-Celtic-Italic shared lexical isoglosses attest to a Northwest-Indo-European continuum in which the Italic group may have been situated territorially between the other two peoples. Finally, Kyes highlights several grammatical components in a given text. His "Grammar and grammars in seventeenth-century Germany . . . " provides a historiographical approach to the development of German

4

Irmengard

Rauch

grammars in the person of Christian Gueintz. He shows this early lackluster German grammarian to have concentrated strongly on the form and meaning of German words, without implication to universal grammar so avidly espoused by non-German grammarians of that era. In the seventeen contributions which comprise this volume we recognize that the era of the linguistic study of singular components of the grammar, of uniquely recognized methods, of frequently studied languages, and/or of language periods has yielded to an era which tolerates, indeed encourages, pluralism in grammar components, methods, languages, and time frames within an article itself and across the seventeen studies. Crucially for the science that is linguistics, this Post-Modern trend has in no way exacted less research rigor. On the contrary, the opening-up of Germanic linguistics to diverse foci is resulting in an unprecedented richness and solidification of the research on languages spoken by people in most parts of the world.

Infinitival subject sentences in Gothic Stephen A. Berard

1. Syntactic representation This paper takes as its basis the modifications to the Government-andBinding (GB) framework suggested in Berard (1993). It is assumed throughout that the linking of arguments to their grammatical functions (GFs), the θ-roles (GF-Θ), in Biblical Gothic is canonically morphological and only exceptionally structural. GF-determinant morphological systems, as opposed to noncriterial morphological features redundantly embedded in structural systems (S-systems), are called M-systems. S-systems and M-systems have divergent ways of realizing the Fundamental Function (FF). Fundamental Function is that which is accomplished by a GF/ θ-linking system, i. e., the establishment of "a mapping between lexically stored thematic information and the syntactic exponents of the GFs" (Haider 1989: 186). In M-systems, the category INFL(ECTION) is projected directly onto the verb and its arguments in the form of morphological markers. Rather than structurally mediating between [ r NP] (the subject) and [v qipandei: pu is sunus meins sa liuba, in puzei waila galeikada. [Noninfinitival translations of infinitives are in roman.]

Except for Type C, which involves functionally obligatory topicalization, all main subtypes of infinitival subject Ss exhibit invertibility of the canonical order of matrix-level infinitive motivator (X) and embedded infinitival subject S (S2), as the examples in (4) indicate. When X is split the status of constituent ordering is here determined on the basis of the position of the matrix (X) verb. In no case does this purely FSP-adjustment affect the literal meaning. Even in the case of Type C, inversion, if permitted by FSP criteria, would cause no more than a shift in emphasis, as the inverted versions of two representative examples of that type, in (5), show. (4) a.

Structural distribution of infinitival subject S 2 s by type: Type A: Head verb copulative, predicate adjective in matrix complement Canonical Si > XS 2 unte [S1 [χ ni gop ist] [S2 niman hlaib for not good is to-take bread 5 barne]] (I Thes i. 6) of-children Exceptional Si > S 2 X [si [s2 du wisan in leika] [xpaurftozo in to be in body more-necessary because izware]]6 (Phil. i. 24) you

12

Stephen A. Berard

b.

Type B: Head verb copulative, predicate nominal in matrix complement Canonical (?)7 S! > S 2 X [si [sip0 samona izwis meljan] [x mis]] the same-things to-you to-write to-me ... [si [x ni latei]]8 (Phil. iii. 1) not bother Exceptional (?) Sj > XS 2

c.

d.

[so ni [S1 [x wulwa ]] [x [so rahnida ]] [S1 [S2 wisan sik not robbery he-thought to-be himself 9 galeiko guda ]]] (Phil. ii. 6) like God Type C: Head verb copulative, predicate pronominal in matrix complement Pragmatically Obligatory? Si > XS 2 [si [x hwa ist] [S2pata us daupaim what is the from dead 10 usstandanl]] [Mk. ix. 10) to-rise Type D: Head verb not copulative Canonical Sy > XS 2 [si [x warp]]... [S1 [S2 gaswiltan] [(X) pamma happened to-die to-the unledin ]] n (L. xvi. 22) poor-(man) Exceptional Sj > S 2 X

e.

swaei [si[(X)Ww] [S2>nais faginon] so-that to me more to-rejoice 12 [x warp]] (II Cor. vii. 7) happened Type E: Raising to subject 13 Canonical Si > XS 2 [si]x dugann Iesus [S2 gipan paim manageim bi began Jesus to-speak to-the crowds about 14 Iohannen]] (M. xi. 7) John Exceptional Si > S 2 X

Infinitival

subject

sentences

in Gothic

13

unte [si [s2gaunon jah gretan] for(conj.) to-mourn and to-weep 14 [x duginnid]] (L. vi. 25) you-begin (5)

a. b.

Ipata us the/that from IXristu dalap Christ down

daupaim usstandan ist hwal15 (Mk. ix. 10) (the)-dead to-rise is what? attiuhan ist pata16 (Rom. x. 6) to-bring is that

2.2 Arguments for subject sentences in Gothic through comparison with some other languages Burzio's (1986: 85-119) exposition of what he considers to be extraposition in Modern English and Modern Italian is particularly complete and lucid, builds on much recent work, and provides a useful touchstone for identifying the individual causal relationships within constructions of this sort. In very general terms, if one follows Burzio's reasoning, extraposition does in fact seem to exist in Modern Italian and perhaps Modern English (and presumably in other configurational languages), and the distributional limitations pointed out in treatments such as those of Emonds (1970) and Köster (1978) can be understood as resulting from certain exigencies, proper to S-system modalities, which block extraposed Ss from appearing in certain positions. After showing that anaphors in an extraposed S 2 are not coindexed with surface structure subjects in Si, Burzio deduces that this is due to the existence of a coindexational relationship between pleonastic subjects and extraposed S2s, since, under the Accessible Subject Condition, 17 both the pleonastic and the INFLj which governs it 18 would be disallowed as accessible subjects if they were coindexed with the category (S2) of which the anaphor in question was a constituent (Burzio 1986: 94-95). Burzio demonstrates that extraposition is parallel to, and subject to, the same principles as (subject) inversion. By comparing consistent patterns of complementary distribution (see below) among inverted subjects (i-subjects), he establishes that there are two different types of i-subjects: those like that in example (6a) which have a direct object θ-role (because the verb is unaccusative 19 ), and those like that in example (6b) which have a subject θ-role (examples from Burzio [1986: 96]). According to Burzio, the sentence in (6a) is base-generated. The Empty Category (EC; shown as e in phrase markers), which is properly governed and assigned Nomi-

14

Stephen A. Berard

native Case by pronominal (i.e., Pro-Drop) INFL, is a variable analogous to a pleonastic. 20 (In a language like Modern English, overt pleonastics are used to fill such positions: there arrived a busload of girls.) It differs from a Trace in that it is not created by movement. It transmits Nominative Case to the i-subject (which is not a real subject), from which it in turn receives a θ-role ("Theme"). This type of inversion will be referred to as "Inversion", i.e., in quotation marks. (6)

a. b.

[e;] arriva Giovanni\ 'Giovanni arrives' [ij] telefona Giovannij 'Giovanni phones'

The sentence in (6b) is created by movement. Its i-subject has a subject θ-role. The EC is a Trace ( 0 which transmits nominative case to the isubject. This type of inversion (true inversion) will be called Inversion. Burzio demonstrates that the same consistent distributional restrictions which reveal "Inversion'VInversion apply equally to the pair "Extraposition'VExtraposition, as is exemplified in the sentences in (7) (taken from Burzio 1986: 100). Sentence (7a) is an example of "Extraposition" because its i-subject bears a verb complement θ-role in Modern Italian. Conversely, example (7b) is generated by means of Extraposition because its i-subject bears a subject θ-role. (7)

a

b.

(8)

mi e capitato [di rivedere Maria (it) to-me is happened of to see-again Maria Ί happened to see Maria again' [?;] mi ha seccato [rivedere Maria\ (it) me has bothered to see-again Maria 'It bothered me to see Maria again'

a b.

Extraposition: It was expected [that John would leave ] *[e] was expected [that John would leave]

c. d.

Extraposition: It bothers me [that John left ] *[/] bothers me [that John left]21

Since "Extraposition" and Extraposition are always accompanied by pleonastics in Modern English (because, according to Burzio, in Modern English I N F L can never properly govern, i. e., can never have a pronominal function), as exemplified in the sentences in (8), it is easy to use contrastive evidence from Modern English when arguing against the likelihood that postposed subjects in Gothic form chains with preverb cate-

Infinitival subject sentences in Gothic

15

gories. Such a tactic, however, is not extremely enlightening because it is not entirely appropriate. It is clear that Biblical Gothic does not use pleonastics, but then no one would argue that Gothic treats subject Ss the way a non-Pro-Drop configurational language does. For similar reasons, it would probably be relatively unfruitful to contrast it in this regard with Modern German, which is non-Pro-Drop and usually requires at least informal pleonastics (see below). What we need is some method, beyond pleonastics, of detecting i-subjects in Gothic. A review of the early Germanic dialects other than Gothic would seem somewhat more promising. In general we find that the major branches (represented by Old English, Old High German, Old Saxon, and Old Norse) use pleonastics optionally, with usage increasing in the course of time, especially in West Germanic, where it quickly and firmly establishes itself. In North Germanic the development of pleonastics was considerably more sluggish, as evidenced by Old Norse (see Ogura [1986: 32]). To judge from the evidence, it seems logical to assume that the development of +CS structure at the level of the daughters of S proceeds, in informal terms, more or less as follows. Despite de jure lateral mobility of the Δ-nodes, functional factors tend to promote certain configurations in certain situations and to suppress others until these tendencies come to be perceived to some extent as de facto prescriptions. The formal lateral mobility of the Δ-nodes may, in intermediate stages, still in principle be formally present, but speakers avail themselves of this freedom less and less. Gradually, speakers begin to behave as though it were more than mere functional appropriateness which establishes certain categories as "belonging" at a particular position at CS. In languages with FSP canons favoring SXV/SVX-type ordering, one hallmark of this behavior is the tendency to avoid beginning a simple declarative sentence with a verb. When functional canons have established postverbal position for subject categories in certain stereotypical situations in declarative sentences, the most convenient way to avoid the result of an initial verb is to use short, frequently occurring words as fillers or "place holders". This practice is already fairly well developed, though not by any means obligatory, in early Old English. In "impersonal"-type expressions (see below) (h)it is the most common filler word. In other situations pä predominates, although several types of expressions may serve this function. 22 The tendency toward complementary distribution between the orderings of expressions of the morgen com 'morning came' type (see Ogura [1986: 48-50]) and those of the extremely common hit wees dceg 'it was day' type is illustrative of the alternation between

16

Stephen A. Berard

subjects and subject place-holders. We observe a pattern [a/S V ß/S®] (where "S" = "subject" and both "X/X" and "X® ... X®" = mutually exclusive alternatives) recurring at high relative percentages. 23 Another apparent symptom of incipient S-modalities at the (former) Δ level is the converse of the tendency toward lexical place-holders. It is what might be informally described as speech habits which reflect an attitude that, when categories are not realized in their canonical positions, these positions are nevertheless in some underlying way still present and to be taken into account. In this way, pleonastics and Traces become observable realities. Clearly, speech habits reflecting + C S L S do not arise uniformly and unanimously. The early Germanic dialects seem to present strong evidence for the view that the transition is gradual and uneven and that in the transitional period M-systems and S-systems alternate with and supplete one another. A study of the evolution of A'-dependencies in Germanic (or in Indo-European) would be extremely valuable. As it is, though, sufficient work has not been done in this area to show us any more subtle signs of nascent configurational alternatives in Germanic. 24 As mentioned above, Biblical Gothic patently lacks formal pleonastics, and the examples provided in Berard (1993) of Gothic deviation from the Surface Structure of the Vorlage make it seem certain that the same was true of natural Gothic. "Informal pleonastics" are fronted lexical categories, of generically specified type, which trigger verb-second order (that is, they stand in the place vacated by a noncanonically positioned subject) in declarative sentences but are not coindexed and thus do not form a chain. Informal pleonastics play an important part in some branches of West Germanic and, insofar as they support obligatory configurations, are important for the study of the evolution of FSP-templates and/or S-modalities. However, since the surface order of arguments and adjuncts in Biblical Gothic mostly follows that of the Greek, we are not in a good position to observe whether or not there were strong tendencies in natural Gothic toward an alternation between subject-initial order and informal pleonasm. It does not seem possible to tell, for example, whether the surface order in a sentence such as (9a) violated a prevalent FSP bias, which in natural Gothic could already have been taking on the quality of a nascent -I-CS prescription, but which could nevertheless be violated, especially in exceptional language (as, for example in a Bible translation), because the FSP bias had not yet made the full transition to an S-system rule. It appears that deviations from the Greek original, like that occurring in

Infinitival

subject

sentences

in Gothic

17

(9b), are probably due to a rule requiring pre-S (= VI") position of certain cojuncts, like paruh, and not to a prohibition against verb-initial declarative sentences, since the latter are allowed, for example, in the presence of the postpositive auk, as is seen in (9c). Also, because the order of Δcategories in the Gothic Bible follows that of the Vorlage, the methods used in a primarily word-order-based study like Zaenen, Maling, and Thrainsson's (1985) examination of θ-tracking systems in Icelandic are for the most part inapplicable to Biblical Gothic. (9)

a.

b.

c.

andhof Iesus jah qap du im (J. vi. 29) άπεκρίθη Ίησοΰς και είπεν αύτοις ... 'Jesus answered and said to them' E>aruh wildedun ina niman in skip (J. vi. 21) ήθελον ούν λαβείν αύτόν εις τό πλοΐον 'And so they wanted to take him into the ship' qi]3a auk izwispatei ... (M. v.20) λέγω γαρ ύμΐν δτι ... 'For I say to you that ...'

Before going on to section 3, which considers the problem of "impersonals", and section 4, which concentrates on arguments internal to Gothic (with some reference to the Vorlage), we will briefly return to Burzio's treatment of Modern Italian and take a look at some of the particular phenomena which facilitate his detection of i-subjects. Although Modern Italian is not a Germanic language, it is nevertheless a Pro-Drop configurational language, and this, presumably, is the general typological group towards or into which a configurationalist devil's advocate would have to contend that fourth-century Gothic may already have moved. We have already mentioned that Burzio claims that both "extraposed" and extraposed subject Ss are coindexed with subject-position ECs, with which they form chains. Chomsky (1988: 215) uses example (10a) to show that the pleonastic subject it must be coindexed with the extraposed S 3 due to the Accessible Subject Condition (see note 17). The Modern Italian translation of this sentence, example (10b), does not contain an anaphor, 25 but sentences like (10c) point to the same conclusion as does (10a). In order to confirm independently the existence and structural position of non-lexical ("covert") pleonastic subjects in Modern Italian, Burzio (1986: 97-98) takes a sentence with an extraposed S 2 , (10d) 26 and inserts it into a larger matrix which triggers infinitive morphology in the verb whose subject the putative pleonastic is. This sentence (lOe) as well as its analogs (see for example (lOf-g) from Burzio [1986: 97], which

18

Stephen A. Berard

have overt NPs) are consistently disallowed. According to Burzio, this constitutes evidence that the subject position is filled with a covert pleonastic which receives Case. Since, of course, infinitival (as opposed to finite) INFL in Modern Italian, as in Modern English, never assigns the subject Case, it is therefore incapable of governing a pleonastic subject, whether overt or covert. (10)

a.

[si They, think [S2 it is a pity [S3 that [pictures of each other,] are hanging on the wall]] b. [si Ciascuno pensa [s-2 che [S2 e peccato [S'3 che each-one thinks that (it) is (a) shame that [ss cifi penda [un quaddro del altro] „attaccato al (there) hangs a picture of-the other attached to-the muro ]]]i]]] wall c. [si Luij pensa s-2 che [S2 tu e peccato [s-3 che [53 ci^ he thinks that (it) is (a) shame that there penda [un quaddro di sejjjj attaccato al muro ]]]ü]]] hangs a picture of himself attached to-the wall d. t\Mi[[ e capitato exx di nuovo [di [vedere Maria ]]; (it) to me is happened again of to see Maria e. *[si La probabilitä [ S ' 2 di [S2 h capitarmi di nuovo the likelihood of (it) to happen-to-me again [s'3 di [ S3 vedere Maria]]j]] e scarsa] of to-see Maria is scanty f. *La speranza [{di) Giovanni arrivare] e svanita the hope (of) Giovanni to-arrive is vanished g. *La speranza [(di) arrivare Giovanni] e svanita the hope (of) to-arrive Giovanni is vanished

Turning now to Gothic, we take a sentence like (11a), which contains a postverbal subject sentence (S3), and ask whether it is to be interpreted as in (lib), as we have been assuming all along, or whether it may be analyzed as hypothesized in (11c), that is, as in Modern Italian. The fact that Gothic seems to employ an M-system implies that infinitive INFL in Biblical Gothic, unlike configurational INFL, can assign Case. The accusative lexical anaphor sik (subject of S 3 ) would seem to confirm this observation. Apparently it is cased accusative by infinitival INFL according to Default Case Marking. 27 On the other hand, the postverbal clausal subject of S 2 , i.e., S 3 , must receive accusative case from the matrix verb rahnida, according to Exceptional Case Marking. 28 The predicate nominal wulwa must then also be accusative.

Infinitival

(11)

a.

b. c.

subject

sentences

in Gothic

19

Phil. ii. 6: ni wulwa rahnida , wisan sik galeiko guda not stolen goods (he)-thought to-be himself like God 'He did not think (it) to be a prize (to be appropriated to himself) (for) himself to be equal to God' [si [vri &i ( N ) ] ni [ 6 S 2 [v'2-2 wulwa] [ V r 2 -i (wisan)]] rahnida [5S2 [s3(vi"2> wisan [Vi"3 i , //: i (A) ] galeiko guda ]]] ?[si [vr'i / ? W I ( N ) ] ni [ S 2 [ N P 2 'Ü] [VP2 wulwa (wisan)]] rahnida [S2 [s3 [^iü(A)] wisan [sik^]^ galeiko guda]n]]

If we attempt to follow the + C S interpretation in (11c), we arrive at a configuration in which sik, which would have to be an i-subject within S 3 , as indicated, is coindexable with pro because the EC [i;j] (presumably a Trace) in the position of subject of S 2 is coindexed with S 3 , of which sik is a constituent. The (principal) problem with this interpretation, however, is that, if we are assuming that arguments in Gothic are assigned grammatical functions by means of an S-system, which we must assume in order to allow the projection of these EC heads of Α-chains, it does not seem that we can simultaneously allow infinitival IN FL to case-mark such ECs, which it must do if such ECs are to form chains with an isubject N P (sik). According to the state of our knowledge, S-system infinitival I N F L does not case-mark the subject, except in Icelandic, where it nevertheless does not properly govern it (see Sigurösson 1991), in contrast to the way Gothic I N F L properly governs sik here. Sentence (12), on the other hand, presents one of the many examples (see note 27) from Biblical Gothic of a postposed S 2 whose subject is morphologically casemarked by infinitival INFL. This apparent divergence from languages like Modern English and Modern Italian seems to be due to the fact that an M-system is operative in Gothic. (12)

L. xvi. 17: φ [si azetizo ist [S2 [vr2 him in(A) jah αϊφα(A)] hindarleipan ]] 'But it is easier for heaven and earth to disappear')

In summary, the absence in Biblical Gothic of that syntactic property (i.e., non-properly governing infinitival INFL) which Burzio employs to establish the likelihood that Modern Italian i-subjects form chains with pleonastic ECs, not only deprives us of a diagnostic tool for the determination of the properties of potential covert Variables coindexed with isubjects, it also constitutes evidence that Biblical Gothic utilizes an Msystem, which in turn seems to exclude the possibility of Gothic i-subjects

20

Stephen A. Berard

altogether. As we observe in other languages with a Δ-level, such as Latin and Ancient Greek, infinitival S 2 s are unordered Δ-nodes at D- and Sstructure (see Chomsky 1988: 127-135) and are subject to no adjacency conditions for the purpose of assigning Case to their subjects. That is, 1) the mobility of infinitival S 2 s at CS and 2) the status of M-type infinitival I N F L as a lexical (pronominal) category, are corollaries of one another. Remaining with our comparison of Modern Italian and Gothic, we note that Biblical Gothic lacks not only all signs of both overt and covert pleonastics, it also lacks any "subjectless" reflexive constructions which could be associated with i-subject formation and "Extraposition'VExtraposition. Gothic seinalsislsik, for example, is never used in any way resembling the use of Modern Italian clitic subject si, which forms chains with "extraposed'Vextraposed S 2 s (see Burzio [1986: 42-53, 106-115]). Finally, there is no sign in Biblical Gothic surface strings of alternations which would reveal an underlying distinction between basegenerated and non-base-generated i-subject NPs or extraposed S 2 s. In Modern Icelandic, for example, base-generated referential pad (pad]) and pleonastic pad (pad2) have demonstrably different distributions as shown by Thräinsson (1979: chapter 4). pad2 "seems to have no meaning or reference"; pad ι can occur in subject position immediately before an S 2 whereas pad2 cannot; and pad2 does not undergo subject-verb inversion whereas pad} does (Thräinsson 1979: 184). Turning to Modern Italian, we see that, as Burzio (1986: 20-27, 86-164) demonstrates, the syntax observes clear formal distinctions between "Inversion" and Inversion. In the case of the former, the auxiliary in the past perfect tense is essere and the past participle is cosuperscripted with the (ostensible) subject. 29 In the case of the latter, the auxiliary in the past perfect is avere and the past participle is not cosuperscripted with the subject.

3. Impersonal expressions In Biblical Gothic there is a very limited number of expressions which have only a minimally filled subject position. That is, the verbs in these expressions refer to events or states to which no personal agency can be assigned. They tend to refer to acts of nature, to existential facts, and to mental events which appear not to be within human control. These "true impersonals" describe situations which are beyond the realm of human agency. When used in the passive voice, they can indicate actions whose agents are unknown, unspecific, or unimportant. In Gothic, true imper-

Infinitival

subject

sentences

in Gothic

21

sonals take third person singular morphology. This is probably so because the third person singular has the most remote pragmatic reference, if we assume that plural number is semantically more marked than singular. The verbal expressions occurring in true impersonal constructions in Gothic are andhaitan 'to profess' in the passive voice, galeikan 'to please', fraqistjan 'to destroy, kill', giban 'to give' in the passive voice, rignjan 'to rain', and puthaurnjan 'to sound', i. e., referring to the last trumpet. The class of true impersonals is represented in Biblical Gothic by the sentences in (13). (13)

a.

gibad, jah gibada izwis (L. vi. 38) Give, and (it)-will-be-given to-you 'Give and you shall receive' b. rignida swibla(D) jah funin(D) '(It) rained sulfur and fire us himina jah fraqistida30 allaim(D) (L. xvii. 29) out of heaven and (it)-destroyed everyone' c. unte pishwammeh(D) saei habaip, gibada imma(D) (Mk. iv. 25) 'For (to)-whomsoever (who) has, shall (it)-be-given' to-him = 'for he who has shall be given more' d. hairto(N) auk galaubeip du garaihtipai, (the)-heart for believes to righteousness, ip munpa(D) andhaitada du ganistai (R. x. 10) but with-(the)-mouth (it)-is-promised to salvation 'For we believe in our hearts and are put right with God; we declare with our lips and are saved' 31 e. puthaurneip auk, jah daupans usstandand unriurjai, jah weis inmaidjanda (I Cor. xv. 52) For (It)-shall-be-trumpeted, and (the)-dead (shall)-rise incorruptible, and we (shall)-be-changed 'For a trumpet shall sound, and the dead shall rise incorruptible and we shall be changed' f. in pizei mis galeikaip in siukeim, in anamahtein, in naupim, in wrekeim, in preihslam faur Xristu (II Cor. xii. 10) Because-of this to me (it)-seems-good [i.e., I am content] in sicknesses, in maltreatments, in persecutions, in necessities, in hardships for-the-sake-of-Christ 'For this reason I am content in sickness, in maltreatment, in necessity, in persecution, and in hardships for the sake of Christ'

22

Stephen A.

Berard

In Old English, true impersonals, such as the OE him wlatep 'he feels disgusted' could be used alone or with a causative complement such as, to use Visser's (1963: 43) examples, dees cetes 'of the food' or for disum mete 'because of this food'. Sentential complements were also possible, such as, to quote Visser (1963: 43) again, pcet ic swa pince 'that I think so' or to tellenne hit 'to tell it'. Since a formal subject with impersonals, like OE (h)it, is not present in Gothic, is lacking in Beowulf and the Heliand,32 relatively rare in early poetry, 33 rare in Old Norse, 34 and only takes the upper hand in later early West Germanic and in middle West Germanic, 35 we can probably assume that it was not a part of the original Germanic syntactic apparatus. The usage developed more or less concurrently in Romance and Germanic, 36 with mutual reinforcement no doubt occurring. 37 Impersonal (h)it in Old English and iz in Old High German first gain the upper hand primarily in expressions of weather, time, and the like.38 From this center, they seem to spread gradually outward. Expressions such as mcel is me to feran 'time is for-me to fare' (Beowulf 316) and us in mycel dearf to witenne 'for-us is very necessary to know' ('it is very necessary that we know') (Bückling Homilies, 63, 5, in Morris [1967]) eventually give way to the type it is time to go. Visser (1963: 4 6 - 4 7 ) points out the evolution of the type it is good (for) us to be here to the detriment of the type god is us to wunienne her 'good is us to be/ live here'. With the passage of time and with the disappearance of dativemarked infinitives, pleonastic (originally impersonal) it is seemingly reinterpreted as being coreferential with the infinitival clause, that is, if the extraposition theory is the correct one for at least the earlier stages. In Modern German (see Ebert 1978: 56) {'Wie klar ist es, daß er es ist?') and somewhat in Modern English ( 7 want it to be nice for him to be at home'), the itles pleonastic comes to be analyzed as generated at a deeper level and follows movement rules like a real argument. In effect what seems to have happened, then, is that, due to the proliferation of it, expressions which are originally derived from true impersonals (with infinitival causal complements), such as it irks him to see that, have become indistinguishable from expressions which never went through a true impersonal stage, that is, expressions in which it was never a real subject but was always coindexed with the subject S, if we decide to follow Burzio's (1986) analysis. However we choose to interpret these expressions, the coincidence of the two types in our own languages has often led those of us studying M-systems to confuse true impersonals in those languages with merely postposed subject Ss.39 In Modern English, Modern German, and Modern French, both types are translated with it!

Infinitival subject sentences

in Gothic

23

eslil (and in Modern Italian there is enough collateral evidence to point to a covert pleonastic here), although in the M-systems impersonal expressions and postposed subject Ss may in fact have nothing more in common than that they both usually begin with a part of the predicate and often involve the use of infinitives.

4. Some internal arguments against the impersonal interpretation of subject sentences in Biblical Gothic To the arguments already made in section 2 in favor of the theory of postposed subject Ss (rather than extraposed subject Ss or complementary infinitival Ss), we will now examine some further syntactic arguments which are directed against the "impersonal" interpretation in general, without regard to the question of how the term "impersonal" is specifically to be understood. Unlike the previous arguments, the following do not rely principally on direct comparison with other languages, although, naturally, the Greek Vorlage will be kept in mind and some contrastive examples will be given from other languages, mostly Modern English. 4.1 Resumption, prolepsis,

and

apposition

Resumptive demonstratives are coindexed with elements which are located to their left in S-structure. They are surrogates of their antecedent within an argument structure which the antecedent lacks within is own immediate environment. In a M-system they transmit a θ-role to arguments which are not themselves located at an identifiable Δ-level because they are fronted into the upper (+CS) Spec level (S^). In M-systems it is also common for demonstratives to refer ahead (to the right) to elements with which they are coindexed but which, likewise, are not directly assigned a θ-role because they are adjoined. This function is called prolepsis. The use of proleptic and resumptive pronominals, in concert with movement of their referents out of the Δ-level, helps functionally to focus these referents while at the same time disambiguating their position within the argument structure. For the latter reason, they are helpful in analyzing the θ-roles of their referents. In example (14a), there is a sentence containing what appears to be a subject S embedded within a matrix which could be interpreted "impersonally" if the word order were more nearly canonical. In this sentence,

24

Stephen A. Berard

a resumptive demonstrative pronominal coindexed with the infinitival S 2 is assigned subject status in Gothic (as well as in the Greek). In (14b) I hypothesize a more canonical schema, one which bears the superficial similarity to the Modern English "impersonal" construction (see the Modern English translation). The resumptive pronoun in (14a), however, clearly points to the subject sentence interpretation. I say "clearly points to", rather than "vindicates" because it is at least conceivable that pata could be either 1) a verb complement with a Goal θ-role or 2) a nonargument. For the reasons given below, however, neither of these two possibilities, once investigated, seems even remotely likely. (14)

a.

Phil. i. 22: φ jabai [liban in leika ];, pataj mis akran waurstwis ist, ... ει δέ [τό ζην έν σακρί];, τοΰτό; μοι καρπός έργου, ... 'But > if [to-live in (the) body];, thatj is worth the effort for me

b.

Ίφ jabai akran waurstwis ist mis liban in leika, ... ?ει δέ μοι καρπός έργου έστίν ζην έν σαρκί, ... 'But if (it) is worth the effort for-me to-live in (the) body, ...'

Alternative # / : If Gothic "subject" Ss were in fact verb complements, then we would expect pata in pata mis akran waurstwis ist to be a verb complement, rather than a subject. That is, using Old English for comparison, we would expect jsata mis akran waurstwis ist 'that to-me fruit of-labor is' to be to ?liban in leika mis akran waurstwis ist 'to-live in body to-me fruit of-labor is' as something like ?jDäer/lmr tö/tö Jjsm him wlatep '?thereto to-him (it) is-disgusting' might be to OE him wlatep to tellenne hit 'to-him (it) is-disgusting to tell it' or as something like ?{)äer/j3ar to/ to ^aem is us mycel dearf '?thereto (it) is for-us very necessary' might be to OE us is mycel dearf to witenne 'for-us (it) is very necessary to know' (see above). However, there is no other context where such an interpretation of pata is likely or even possible. If it were, we would apparently be dealing with some sort of accusative of goal. On the other hand, we could attempt to claim that pata has an aggregate reference. That is, we could say that its referential quotient may be a generalized incorporation of a complex of syntactic relationships, as with das in 'Sind wir in großer Gefahr? Ja, das sind wir!' Are we in great danger? Aye, that we are!'. In this case the morphosyntactic expression of "goal" would not need to be explicitly present in the predicative demonstrative pronominal. This would still leave us, though, with the burden of having to find signs in "subject" infinitival clauses of features

Infinitival

subject

sentences

in Gothic

25

which would mark them as a complements rather than as a subjects. Gothic lacks the dative infinitive available to Old English. The origin of the Gothic infinitive as an accusative-marked deverbal nominal, like the dative and locative origins of Greek and Latin infinitives (see Coleman [1985]), was apparently so remote as to be synchronically irrelevant to Biblical Gothic. This is clear, most of all, from the fact that the infinitival constructions in Gothic most likely to be interpreted with Goal semantics, that is, infinitival adjuncts with "final" semantics, make the heaviest use of the preposition/particle du. Since all Germanic forms corresponding to Gothic du take, if anything, dative case, the use of du with infinitives to express Goal seems to constitute conclusive evidence that the Gothic infinitive had lost its accusative force. Besides the accusative case, the only other available assigner of "goal" semantics is the particle du itself. Whereas du is fairly reliable as a marker of "goal" semantics in verb adjuncts, it is only rarely used with "subject" and complement infinitives, and its use with these seems to have virtually no relationship to "goal" semantics. The use of du with subjects and verb complements is quite desultory and is related more closely to attributes of the Greek Vorlage (i. e., to the use of articular infinitives) than to any other consistent principle. Of the few (nine) occurrences of du with "subject" infinitival Ss, three (two in Mk xii. 33 and one in Mk x. 40) are "articular" infinitives φαΐα du + INF) in which the article/demonstrative pata seems to have an unambiguous subject θ-role. 40 Only three examples of an apparent subject infinitive (Mk ii. 9,1 Thes iii. 4, and Sk. i. 18) have a spontaneous occurrence of du, i.e., an occurrence of du where there is no model Greek articular infinitive, and one of these (Sk. i. 18) occurs in the Skeireins in an S 2 with DCM, that is, an accusative subject of the infinitive. 41 There is a second alternative. If Gothic "subject" Ss were in fact extraposed, the demonstrative pata, which is resumptively substituted for this S 2 , would then have to occupy an non-argument position ("A'-position"), just as do the Modern Italian and Modern Spanish demonstratives in (15b) which are substituted for the extraposed subject clauses in (15a). However, there is no evidence in Gothic for A'-position pronominals in Gothic, whether standing for infinitival or for finite clauses. If such were possible, one would expect a configuration such as that in (15c) to occur canonically, or at least occasionally, but, although it is not technically disallowed, in never occurs. When examining this alternative, we should not let ourselves be confused by non-Pro-Drop equivalents. In (15d-e) we see that Standard

26

Stephen A. Berard

Modern English,42 like Gothic, disallows A'-position demonstratives, although it does exhibit Extraposition. In Modern English resumption (13f) and in the nearest Modern English equivalents of prolepsis (15g), the demonstrative bears a subject θ-role, just as in Gothic. We have already pointed out above that Gothic is, at the very least, a Pro-Drop language and that, from a typological point of view, it can sometimes be more appropriate to compare it with Modern Italian and Modern Spanish than with Modern English or Modern German. The specific reason for this discrimination in the present discussion are the disparate origins of "impersonal" expressions with embedded sentences (whether we consider them to involve verb complementation or Extraposition) in Modern English (and Modern Gerean), on the one hand, and in Modern Italian and Modern Spanish on the other. As we have seen, "subject sentences" in Modern English originated as a three-part expression [(NP(*)) (Pred) (S2)], in which the NP was initially covert but gradually acquired obligatorily overt status and the S2 was a verb complement. Extraposition in Modern Italian and Modern Spanish, however, never went through a three-part stage. The Latin precursor was the two-place expression [(Pred) (S2)], seen in (15h). When Modern Italian and Modern Spanish made the transition to S-linking of arguments, postverbal positions became associated with complements and adjuncts, thus causing the standard configuration, represented by the fourth of the Latin alternatives in (15h), i.e., the one adopted at FSP as canonical to be reanalyzed as extraposition. The external (overt) form [(Pred) (S2)], however, remained the prescribed pattern, which may explain from a historical point of view why the configurations in (15i) turn out to be disallowed in Modern Italian and Modern Spanish. Lexicalization of the covert subject would be equally impermissible. One might say that, despite syntactic reanalysis, S-structure remains superficially faithful to the construction's origins. When the subject S2 in Modern Spanish and Modern Italian is substituted with a demonstrative, the resulting configuration must conform to the corresponding historical (Latin) pattern, shown in (15j), which consisted of a maximum of two (and a minimum of one) overt places: [(NP/ 5e) + Pred]. Hence the permissibility of (15b) and (15k), as against the tripartite pattern [(DemPro) (Pred) (S2)] in (15i). A similar historical explanation suggests itself for the impermissibility in standard Modern English of the tripartite configuration in (15e), as against the permissible bipartite one in (151). The prototype for Extrapos-

Infinitival subject sentences in Gothic

27

ition was tripartite, but the prototype for the version with demonstrative pronominal substitution, i.e., simple predication, was bipartite. The essential difference, then, between the versions with demonstrative substitution in Modern English, on the one hand, and Modern Italian and Modern Spanish, on the other, is that in the latter languages the ProDrop-positive setting allows insertion of the A'-position demonstrative with retention of the bipartite overt Suface Structure, as in (15b), whereas the Pro-Drop-negative setting in Modern English does not permit insertion of the A'-position demonstrative, since this would violate the overt bipartite model for predication. (15)

a.

b.

c. d. e. f. g. h.

i. j.

k. 1.

tx [E sano ] [mangiare del tofü\ tx saludable ] [comer tofit 'It's healthy to eat tofu' t\ [E sano ] [ciö ]; ti [£"5 saludable ] [eso ]j '(It) is healthy that' 'That's healthy' *ip jabai ^ ist mis akran waurstwispatau ... (*ει δέ t[ μοί έστι καρπός έργου τοΰτο;, ...) It\ is healthy [to eat tofu\ *Iti is healthy thatx = Standard Modern English [To eat tofu\, thatx is healthy That(l) is healthy, that is, / i. e., [to eat tofu [vr dolere] [vr malum est] (Cicero, De finibus, v. 84) = [VI" dolere] [Vr est malum] = [vi- malum est] [VI» dolere] = [vr est malum] [Vr dolere] 'It is bad to suffer' *Ciö\ t[ e sano [mangiare del tofu], *EsO[ ti es saludable [comer tofu ]i [vr Id/öe] [vr est salutare] = [vr Idlde] [Vr salutare est] = [vr salutare est] [vr id/Se] = [vr est salutare] [ V r id/Se] 'It/that is healthy' Ciö I pro e sano Esolpro es saludable That is healthy

28

Stephen A.

Berard

m. Appan gastauida [pata(A> silbo (A)]j at mis, [ei aftra in saurgai ni qimau^OPT^ at izwis\ (II Cor. ii. 1) "Εκρινα δέ έμαυτω [τοΰτο (Α) ]ί, [τό μή π ά λ ι ν ( Ι Ν Ρ ) έν λύπη έλθε ΐ ν προς ύμας]ί 'But I made up my mind about this: I would not come to you again in sadness' There are, then, two main reasons why we would expect Gothic to have more in common with Modern Italian and Modern Spanish with respect to the constructions in question, than with Modern English. First of all, Gothic has null subjects. Secondly, there is no evidence that Gothic "infinitive subjects" had originated relatively recently as verb complements. (It was briefly indicated above that evidence from dw-usage is lacking, i.e., that du + I N F did not, synchronically in Biblical Gothic, have a dative or Goal function which would have predisposed it to be a complement.) Consequently, if there were indeed Extraposition or "Extraposition" in Gothic, we would expect, at some level of the Gothic language, to find something like what we see in ( 1 5 b - c ) or some sign that such a construction were possible, but we do not. Sentence (15m) is a contrastive example. It involves a proleptic demonstrative coindexed with an S 2 . In this case the θ-role assigned to the demonstrative, and thus also to the S 2 , is clearly that of direct object. Apposed elements, that is elements directly cojoined with arguments, are, like elements coindexed with resumptive and proleptic pronominals, disambiguated, with respect to their θ-role, by the element to which they are apposed. In the example in (16), the N P hwa is the subject of the sentence hwa skuld ist sabbato dagam? 'what is permissible on the days of Sabbath?', but after the verb there is also a series of apposed subject S 2 s, only the first two of which are given here and which are cojoined with the N P hwa. The superordinate subject level is marked VI"2-o> the level at which the subjects of S 2 are cojoined; the N P subject is marked VI"2_i, the level of the first subject of S 2 ; and the postposed alternative pairs of apposed subject infinitival clauses are marked S 3 , S 4 . The cojoined relationship between N P and embedded S subjects indicates that it would be wrong to interpret the infinitival S 2 unpiup taujan as a verb complement in sentences like ni skuld ist lagjan pans in kaurbaunanloiiK έξεστιν βαλεΐν αύτά εις τον κορβοναν 'It is not permitted to put them into the temple fund' (M. xxvii. 6) or ... unte atpaim gahwairbam frakunnan si skuld ist '... for it is not seemly that there be contempt among comrades', since the latter type of sentence is completely parallel to (16)

Infinitival subject sentences in Gothic

29

and no new significant variables seem to have been introduced, to judge both from this specific sentence and from all sentences of this type. (The same arguments which were given above against a verb complement or extraposed status for resumptive pata will also apply to hwa here.) The translator could, for example, have easily substituted the infinitival S 2 with finite S 2 s in the optative mood, as he did in (15m). (16)

L. vi. 9: [SI Sefraihna izwis [s2 [vi"2-o [vi"2-i [NP hwa^[[w y2 skuld (I) ask you what permitted U ist sabbato dagam ], [vi"2-o [vr'2-2 [s3 P^ P taujan (1NF> is (on)-Sabbath days good to-do (INF) pau [S4 unpiup taujan ] etc. ...]]]]? or evil to-do Ί ask you what is permitted on the Sabbath, to do good or to do evil?' [S1 He έπε ρωτήσω ύμας [S2 [νι-2-ο [vi-2-i [np τί]] (I) ask what what [vi-2 έξεστιν τοις is-permitted on-the σάββασιν] [Vr2-o [vi"2-3 [s3 άγαθοποιήσαι (1ΝΡ) ] ή Sabbaths, to-do-good or (ΙΝΡ) [S4 κακοποιήσαι ] etc. ...]]]]? to-do-evil?

Contrast (16) and the above counterexamples (ni skuld ist lagjan pans in kaurbaunan and unte atpaim gahwairbam frakunnan ni skuld ist) with the Modern English pair: a) what is permitted on the Sabbath? and b) it is not permitted to do evil on the Sabbath. When compared with parallels from languages like Modern English, example (16) thus becomes the basis for an argument based upon the lack of pleonastics in Gothic. Comparison with Modern English also supplies an argument from constituent ordering. This argument is weaker than the foregoing, because it is based on the distinction between canonical and non-canonical ordering. It runs as follows. Literal Modern English translations such as what is permitted on Sabbath days, to do good or to do evil? sound somewhat stilted and are avoided by all of the translations which I have consulted. 43 The presumable reason for this is that the cojoined, i. e., structurally equated items, what and to do good or to do evil, are not felt to be exactly structurally parallel, what is permitted on Sabbath days, to do good or to

30

Stephen A. Berard

do evil? implies an answer like to do good on the Sabbath is permitted, which involves a noncanonically fronted infinitive. Whether we accept Burzio's arguments in favor of Extraposition/"Extraposition" or Emonds' (1970) and Koster's (1978) arguments in favor of verb complement status for "subject" infinitival clauses in Modern English, it is clear in either case why these clauses are only very exceptionally fronted in Modern English. The petrification of ordering patterns occasioned by advanced S-systems has caused Modern English infinitives to be very strongly associated with the postverb position, which has thus caused them to lose much of their nominal quality, specifically their NP-style mobility. Gerunds in -ing have largely taken over the lost functions. There are probably two reasons why the literal type of translation does not, nevertheless, seem actually "incorrect". First, preposed subject infinitives (either with an arbitrary subject NP*, as here, or with [for + NP subject]), though noncanonical, are not absolutely incorrect, presumably for the reason that they are in fact logical subjects, although they do sound stilted. Second, although the infinitival clauses are technically cojoined to the subject, they are in fact located to the right of the verb, which gives them a certain familiar ring at Phonetic Form (PF). An apparent added problem with the Extraposition interpretation is that cojoining (of what and S 2 ) has replaced coindexation, because interrogative pronouns cannot corefer. It may be, however, that cojoining without explicit coindexation may be permitted here because at Logical Form (LF) there is in fact a potential retroactive corerefentiality between what and S 2 . If the Gothic version of L. vi. 9, example (16), then, copies the Greek, rather than altering the means of expression somewhat in order to achieve greater idiomaticity, then we have here a style-based argument to add to the above syntactic argument in favor of the subject status of the postposed infinitival S2s. This example of apposition in (16), furthermore, is not an isolated exception but is in conformity with the example of resumption in (14a) and with all the other types of postposed infinitives for which subjecthood has thus far been established or will be established below. To summarize the arguments from apposition, then, 1) the example in (16) is clearly not a verb complement, and 2) the example in (16) is in conformity with the general treatment which logical-subject infinitives receive in Gothic, whereas its literal Modern English counterpart,by contrast, is noncanonical.

Infinitival subject sentences in Gothic

31

4.2 Subject distribution by type Subject NPs and subject Ss are not in complementary distribution in Biblical Gothic (or Greek). It was noted in section 2.1 that clausal Postposing is sometimes made functionally obligatory by semantics or by the length of the subject S. This is typical, for example, of the usage of the verb wairpan, which, for pragmatic reasons precedes all but the shortest subject Ss. The verb wairpan is used for introducing narratives, and it sounds anticlimatic to place warp 'happened' at the end of a long subject S 2 , since the S 2 is a Theme and should not be upstaged by a dramatically suspended matrix verb which, logically, is only a Rheme. However, beyond such purely functional considerations, there is no observable formal distinction made between the two types of subject at the categorial levels of the syntax. In (17) examples of subject sentences are given in order to illustrate that N P and S 2 subjects are treated alike in comparable situations. The sentential subject of binah in (17a) is preposed just as is the NP subject of the same predicate in (17b) and just as is the NP subject of the related predicate binauht ist in (17c). That the infinitive hwopan in (17a) is not a substantive can be shown from the context (the subject of S 2 is raised to the position of indirect internal argument of the matrix, de (D) ). Further, although all long and complex sentential subjects of wairpan are postposed, it is to be noted that a complex NP subject of wairpan, such as that in (17d), is also postposed. 44 Nevertheless, just as in (17a), with the head verb binauhan, so also can simple infinitival S 2 s be preposed even when wairpan is the head verb, as shown in (17e). Example (17f) has, for comparison, a preposed NP subject of wairpan. In (17g) we see a example of a subject S which is more than one word long but which is nevertheless preposed, presumably because FSP canons do not block Preposing here. The postposed predicate paurftozo in izwara is in fact a Theme here and thus can easily bear the weight caused by its suspension until the end. (17)

a.

b.

II Cor. xii. 1: [S2 St-, hwopan^NF)] öe) εξεστιν everything to-me is-allowed L. i.45: wairpip, [vvustauhts(N) pize (there will) come about (the) completion of-those-(things) rodidane izai fram fraujin ] said to-her by (the) Lord (Ν) εσται [ v r τελείωσις ] τοις λελαλημένοις αύτη παρά κυρίου (there) will be fulfillment to those (things) said to her by the Lord II. Cor. vii. 7: swaei misx mais [Vr e>/j faginon(INF) warp so-that to-me more to-rejoice came-about '... so that I rejoiced all the more' ώστε με μάλλον χαρήναι ( Ι Ν Ρ ) so-that me more to-rejoice L. xv. 7: swa [γι- faheds(N)] wairpip in himina thus joy (shall)-come-about in heaven 'Thus there shall be joy in heaven.' ούτως [vr- χαρά ( Ν ) ] έσται έν τω ούρανω thus joy shall be in (the) heaven Phil. i. 24: appan [γι- 8ex du vm««(INF) in leika ] paurftozo(N NT S) but to be in body more-necessary (D) δβ\ in izwara (for-me) because-of you 'Because of you it is more necessary for me to remain in the body.' [s"2 [s'2 τό) δέ [s"2 [s'2 [s2 the but to-remain

έπιμένειν ( Ι Ν Ρ ) έν τη σαρκι]]] in the flesh

Infinitival subject sentences in Gothic

33

άναγκαιότερον (N-NT-S) δβ\ δι' ύμας more-necessary (forme) because-of you It should be mentioned that wairpip sometimes translates έσται 'shall be' / 'shall come about' (see Schulze 1840: 4 0 5 ^ 0 9 ) , which, in the initial position, as in (17d), has "existential" semantic force. The fundamental meaning of this verb in Gothic (as evolved from the basic meaning of PIE wert, 'to turn, revolve, come around') is "in Erscheinung, ins Dasein treten" (Streitberg 1950, 2: 166), just as with the Greek γί(γ)νεσθαι (see Bauer [1957:157]). The associated meaning "stattfinden, geschehen" (1957: 157) is similarly absolute. Only the derived meaning werden need require a predicate. The tendency of expressions like wairpip/εσται to have existential/absolute semantic force is reflected at FSP by their tendency to be canonically preposed. This is to be contrasted to the preposing of νναφ/έγένετο which, as mentioned above, occurs not because warpleyevexο is emphatic but rather because the alternative, suspension until the end of the S 2 , is even more so, due to the odd functional effect arising therefrom (i.e., '[S2 ···] happened'). Both types of Preposing, the emphatic existential type and the unemphatic narrative type, can create confusion with modern "impersonal" expressions such as it happened / shall happen that, es geschah / wird geschehen, daß, and ilpassa /passera que, a confusion which comparison of the parallel usage of S 2 and N P subjects should clear up. 4 5

4.3 Predicate

distribution

by

type

The predicates of subject infinitival Ss are of all types and involve both noncopulative verbs (such as wairpan or galeikan) and copulative verbs with adjective or N P predicate complements. Just as was observed with regard to placement of N P and S 2 subjects, so also there appear to be no patterns of complementary distribution between N P and S 2 subjects with regard to types of predicates. The copulative type with predicate adjective complements in Si are the most common, but N P predicate complements like akran waurstwis ist 'it is the fruit of activity' = 'it is worth the trouble', seen above in (14a), occasionally occur (see Phil. i21 and 22, ii. 6, and iii. 1; Mk ix. 10; Rom. x. 6 and 7; and I Cor. v. 12). In (18a), rewritten in (18b), there are two cojoined NPs predicated of the subject S: latei and pwastipa. Each can be translated impersonally in Modern English: ni latei 'it is not bothersome', and pwastipa 'it is safe'. We can be fairly sure that the use of N P s here is idiomatic, or at least somewhat so, since the Greek original has not nouns but adjectives for both.

34

Stephen A. Berard

In (18c), rewritten as (18d), there are a pair of cojoined sentences, each with a subject S and each with a predicate nominal in the complement. What is interesting here is that only the second predicate nominal, gawaurki, is nonpersonal, and thus potentially compatible with an impersonal interpretation. The first predicate nominal, Xristus, is personal. If S 2 were indeed impersonal, then we would have an instance of zeugma here, both in the Gothic and in the Greek. There are, however, a number of very cogent reasons to disqualify the possibility of zeugma in this instance. Firstly, zeugma is exceedingly rare in St. Paul (see Blass and Debrunner [1961: §479, 2]). Secondly, it would seem irreverent, not to mention quite inelegant, intentionally to create a grammatical situation in which there were a danger of treating the phrase Xristus ist as an impersonal construction: *To me is Christ to live and advantageous to die. Third, if this were a case of zeugma, one would at least expect the author to pair the verb ist with gawaurki rather than with Xristus, or to repeat ist, so as to obviate confusion of the personal with the impersonal construction. (18)

Phil, iii.l: a. po(A) samona(A) those same-(things) ni latei,(N) ip not (a) bother, but

izwis(D) meljan(INF) mis(D} swepauh to-you to-write for-me indeed (is) (D) (N) izwis pwastipa for you (it is) security

'It doesn't bother me to repeat what I have written before, and it will add to your safety.' Good News 1971 τά (Α) αύτά (Α) γράφειν ( Ι Ν Ρ ) ύμΐν(Ε>) έμοί(Ε>) the same (things) to-write to-you (is) to-me μεν ούκ όκνηρόν (Ν) , ύμΐν ( Ι Ν Ρ ) δέ on-the-one-hand not bothersome, for-you however άσφαλές (Ν) . (it is) secure b.

[si [vrP° samona izwis 5t[ meljan] [νιata du frijon nehwundjan swe sik silban managizo ist allaim pa im alabrunstim jah saudim [και τό άγαπάν αύτόν έξ όλης της καρδίας και έξ ολης της συνέσεως και έξ όλης της ψυχής έξ όλης της ισχύος και τό άγαπάν τόν πλησίον ώς έαυτόν πλεϊόν έστιν πάντων των ολοκαυτωμάτων καί θυσιών; Mk x. 40: ip ]>ata du sitan af taihswon meinai aippau af hleidumein nist mein du giban ... [rö δέ καθίσαι έκ δεξιών μου καί εύωνύμων ούκ έστιν έμόν δοΰναι ...]. 41. Mk ii.9: hwapar ist azetizo du qi[?anpamma uslipin ...? [τί έστιν εύκοπώτερον ειπείν τω παραλυτικφ ...]; I Thes. iii.4: fauraqepun izwis patei anawairp was uns du winnan aglipos [προελέγομεν ύμΐν ότι μέλλομεν θλίβεσθαι]; Sk. i. 18: gadob nu was maispans swesamma wiljin ufhausjandans diabulau du ufargaggan anabusn gudis]. (All references to the Skeireins are taken from Streitberg [1950].) 42. I am told that in some British dialects the surface string in (13e) is normal. This would appear to involve an A'-position demonstrative pronoun. The demonstrative in senten-

42

Stephen A. Berard

ces like this is always projected to the right of the VP. This does not hold for Gothic pata. In cases where pata could possibly be construed as occupying an A'-position, Biblical Gothic follows the lead of the Greek, locating pata, when standing alone (for τοΰτο, εκείνο, etc.), freely as a Δ-level argument according to FSP dictates, rather than in any way which could be interpreted as characteristic of an A'-category. Sentences like jah pata izwis taikns (L. ii. 12), for example, with the demonstrative in canonical subject position, do not suggest a translation such as 'and it is a sign to you this'. 43. Good News (1971) has "I ask you: What does our Law allow us to do on the Sabbath? To help or to harm?" King James has: "I will ask you one thing: Is it lawful on the sabbath days to do good, or to do evil?" The New English Bible has: "I put the question to you: is it permitted to do good or to do evil on the Sabbath?" Close translations which still sound fairly idiomatic to me are: 'What is permitted for us to do on Sabbath days, good or evil?', 'What are we permitted to do on the Sabbath, good or evil?', and 'What is permitted on the Sabbath, that one do good or (that one do) evil?' 44. The material which makes up the genitive participial NP-Complement of ustauhts in the Gothic version corresponds to a dative participial clause verb complement with a role of the general indirect object type (whatever the exact θ-role might be determined to be) in the Greek. Thus, the Greek subject NP is uncomplemented. 45. The parallelism noted here between nominative-marked NP subjects and infinitival sentential subjects is not at variance with the established derivation of the particular type of Germanic infinitive used in Gothic from the accusative form of verbal nouns (see Szemerenyi [1970: 300] and the derivation of Greek infinitives from dative and locative verbal nouns (see Szemerenyi [1970: 299] and Coleman [1985]). Coleman (1985: 311313) points out certain usage patterns in both early Greek and early Germanic which seem to involve these origins in some vestigial way, but, as Coleman (1985: 311) also points out, infinitives underwent system-wide reanalysis in both Hellenic and Germanic. Greek realigned case distinctions as markers of aspect, tense, and voice; Gothic infinitives lost all case morphology distinctions due to reduction to an Einheitsform. Aspect distinctions were provided by the prefix ga-; infinitive tense and voice were expressed analytically. 46. Neither covert nor overt pleonastics are represented in this and the following configurational representations, whether well-formed or ill-formed, as base-generated to the left of the verb and then subsequently moved to their postverb position in S-structure. It is a generally accepted principle of θ-theory that categories with Case block contraction. The canonical subject position, when supposedly vacated due to verb-subject "inversion", does not block contraction, as is seen in the Modern Italian cos'el (w/i-movement) in (17h) and in Modern English sentences like What's an endogen? The answer to the problem of the D-structure status of subjects in - R S / + C S may lie in the solution suggested by Pesetsky (1982) and elaborated by Lasnik (1988). They assume, in Lasnik's (1988: 27) words, "that no principle of the grammar can say specifically where a subject must be in D-Structure. But at S-structure there are certain Case requirements that do in general force the subject to be in a particular position." Lasnik uses this assumption to explain why PRO does not block "wanna contraction", whereas NP-Trace does. He argues that since PRO is not subject to Case requirements, it is free to occur in a postVP position. This application of the base-unspecific subject assumption does not apply to our problem with contraction under subject-verb inversion, since a variable left by pro in the subject position would have to be a Trace. However, although Modern English seems to be a Specifier-first language, we might posit an S-structure rule whereby

Infinitival subject sentences in Gothic

43

Specifiers are optionally fixed to the right of Heads and Complements when they are in competition for pre-I' position with certain other types of categories (compare His fiddlers three), among which would be w/i-categories. Affix Hopping (from I to V) would still apply. Of course, sentences like What's he doing to youl, involving only INFL-subject inversion, would not fall under this category (but would, rather, complement it), since movement across INFL does not create a Trace. These two principles together would explain contraction over the subject position. (Naturally, we have to assume that A'-categories are not part of the VP.)

References Abraham, Werner 1986 "Word order in the middle field of the German sentence", in: Werner Abraham - S. de Mey (eds.) 94-111. Abraham, Werner - S. de Mey (eds.) 1986 Topic, focus, and configurationality. Amsterdam: Benjamins. Allen, Joseph, Henry - James Bradstreet Greenough - George Lyman Kittredge - Albert Andrew Howard - Benjjamin Leonard D'Ooge (eds.) 1988 New Latin grammar. New Rochelle, New York: Caratzas [reprint], Aoun, Joseph 1985 A grammar of Anaphora. Cambridge, Mass: MIT. Bauer, William F. 1957 Griechisch-Deutsches Wörterbuch zu den Schriften des Neuen Testaments und der übrigen urchristlichen Literatur. Berlin 1952, 1958. [1955] [A Greek-English lexicon of the New Testament and other early Christian literature, translation and adaptation by William F. Arndt and Felix Wilbur Gingrich from the 4th edition. Cambridge - Chicago: Chicago University Press.] Bean, Marian C. 1983 The development of word order patterns in Old English. London — Totowa, N.J.: Croom Helm - Barnes & Noble. Behagel, Otto (ed.) 1984 Heliand und Genesis. (9th edition.) Revised by Burkhard Taeger. Tübingen: Max Niemeyer. Beowulf (See Klaeber) Berard, Stephen A. 1991 "Some empty categories associated with infinitival constructions in Greek, Latin and Gothic", Selecta 12: 1-7. 1993 "Biblical Gothic and the configurationality parameter", American Journal of Germanic Linguistics and Literatures, 5-2: 111-162. Blass, Friedrich — Albert Debrunner 1961 A Greek grammar of the New Testament and other early Christian literature. Chicago: Chicago University Press. Burzio, Luigi 1986 Italian syntax: A Government Binding approach. Dordrecht: Reidel. Chomsky, Noam 1986 Barriers. Cambridge, Mass.: MIT Press. 1988 Lectures on Government and Binding. (5th edition.) Dordrecht: Foris.

44

Stephen A. Berard

Cicero, Marcus Tullius [1967] De finibus bonorum et malorum. (2nd edition.) Cambridge, Mass.: Harvard University Press, van Coetsem, Frans - Herbert L. Kufner (eds.) 1972 Toward a grammar of Proto-Germanic. Tübingen: Max Niemeyer. Coleman, Robert 1985 "The Indo-European origins and Latin development of the accusative with infinitive construction", in: Christian Touratier (ed.), 307-341. Ebert, Robert P. 1978 Historische Syntax des Deutschen. Stuttgart: Metzlersche Verlagsbuchhandlung. Emonds, Joseph E. 1970 Root and structure-preserving transformations. [Mimeographed paper circulated by the Indiana University Linguistics Club.] Gippert, Jost 1978 Zur Syntax der infinitivischen Bildungen in den indogermanischen Sprachen. Frankfurt am Main: Peter Lang. Good News for Modern Man: The New Testament in today's English version 1971 New York: American Bible Society. Goodwin, William W. 1965 A Greek grammar. New York: Macmillan. Haider, Hubert 1989 "θ-tracking systems - evidence from German", in: Läszlö Maräcz - Pieter Muysken (eds.) 85-116. Heliand und Genesis (See Behagel) Heusler, Andreas 1962 Altisländisches Elementarbuch. Heidelberg: Carl Winter Universitätsverlag. Johannessohn, M. 1926 "Das biblische και έγένετο und seine Geschichte", Zeitschrift für vergleichende Sprachforschung 53: 161-212. Johansson, K. A. S. 1968 Studies in the history and development of the English language. [Unpublished dissertation, Indiana University.] Keyser, Samuel Jay (ed.) 1978 Recent transformational studies in European languages. (Linguistic Inquiry Monograph 3.). Cambridge Mass.: MIT. King James Bible (n. d.) Oxford and Cambridge: Oxford and Cambridge University Presses. Klaeber, Friedrich (ed.) 1950 Beowulf and the fight at Finnsburg. Lexington, Mass.: D. C. Heath. Köhler, Artur 1867 "Der syntaktische Gebrauch des Infinitivs im Gothischen", Germania 12: 421-462. Koster, Jan 1978 "Why subject sentences don't exist", in: Samuel Jay Keyser (ed.) Kress, Bruno 1982 Isländische Grammatik. Leipzig: Max Huber.

Infinitival subject sentences in Gothic

45

Lasnik, Howard - Juan Uriagereka 1988 A course in GB syntax: Lectures on Binding and Empty Categories. Cambridge, Mass.: MIT Press. Lehmann, Winfred P. 1972 "Proto-Germanic syntax", in: Frans van Coetsem — Herbert L. Kufner (eds.), 239-268. Liddell, Henry George - Robert Scott 1968 A Greek-English Lexicon. Oxford: Clarendon. Maracz, Läszlo — Pieter Muysken (eds.) 1989 Configurationality: The typology of asymmetries. Dordrecht: Foris. Mitchell, Bruce 1985 Old English syntax. 2 vols. Oxford: Clarendon. Morris, Richard (ed.) 1874—1880 The Blickling Homilies of the tenth century, N. Trübner & Co. [1967] [Reprint. Oxford University Press: London.] The New English Bible: New Testament 1962 Oxford and Cambridge: Oxford and Cambridge University Presses. Ogura, Michiko 1986 Old English "impersonal" verbs and expressions. (Anglistica 24.) Copenhagen: Rosenhilde and Bagger. Ostafin, David Mark 1986 Studies in Latin word order: A transformational approach. [Unpublished dissertation, University of Connecticut.] Perlmutter, David M. 1978 "Impersonal passives and the unaccusative hypothesis", Berkeley Linguistic Society 4: 157-189. Pernot, Hubert 1927 Etudes sur la langue des Evangiles. Paris: Lille. Pesetsky, David Michael 1982 Paths and Categories. [Unpublished doctoral dissertation, MIT.] Rosenbaum, Peter S. 1967 The grammar of English predicate complement constructions. Cambridge, Mass.: MIT. Saitz, Robert L. 1955 Functional word-order in Old English subject object patterns. [Unpublished dissertation, University of Wisconsin.] Schulze, Ernst 1840 Gothisches Glossar. Magdeburg: Verlag der Gebrüder Baensch. Sigurösson, Halldor Α. 1991 "Icelandic case-marked PRO and the licensing of lexical Α-positions", Natural Language and Linguistic Theory 9: 327-363. Smyth, Herbert Weir 1984 Greek Grammar. (20th edition.) Cambridge, Mass.: Harvard University Press. Streitberg, Wilhelm 1920 Gothisches Elementarbuch. Heidelberg: Carl Winters Universitätsbuchhandlung. Streitberg, Wilhelm (ed.) 1950 Die gothische Bibel. Heidelberg: Carl Winters Universitätsbuchhandlung.

46

Stephen A. Berard

Szemerenyi, Oswald 1970 Einführung in die vergleichende Sprachwissenschaft. Darmstadt: Wissenschaftliche Buchgesellschaft. Thräinsson, Höskuldur 1979 On complementation in Icelandic. New York: Garland. Touratier, Christian (ed.) 1985 Syntax et Latin: Actes du Ilme Congres International de Linguistique Latine. Aix-en-Provence, 28-31 Mars 1983. Aix en Provence: Publications Universite de Provence. Visser, Fredericus Theodorus 1963 An historical syntax of the English language, Part I: Syntactical units with one verb. Leiden: E. J. Brill. Wright, Joseph 1975 Grammar of the Gothic language. Oxford: Clarendon [reprint]. Zaenen, Annie - Joan Maling - Höskuldur Thräinsson 1985 "Case and grammatical functions: The Icelandic passive", Natural Language and Linguistic Theory 3: 4 4 1 - 4 8 3 .

Lexicality and the versification of Johann Heinrich Yoß: Observations on prosodic feature analysis David

Chisholm

In human languages there is a natural tendency to emphasize in some way those words or syllables which carry the most semantic information. In Germanic languages, the so-called "content words" (nouns, verbs, adjectives, and some adverbs), which belong to lexical categories, bear in most instances a greater degree of stress than the essentially nonlexical "function-words" (such as prepositions, conjunctions, and determiners), which comprise grammatical categories. Occasionally, however, due to the historical development of the language, a conflict arises between its accentual patterns and its lexical structure, as in the following English and German sentences (in which lexical morphemes not bearing primary accent are printed in roman type): (1)

Jack fell 'down in the 'owifield. She is an 'outgoing person. Sie nahm '.^schied und ging 'hin. Unter keinen ' t/mständen werde ich 'zwsagen.

Such conflicts between the accentual and lexical structure of a language have significant consequences in verse, where these linguistic characteristics can conceivably occur in varying relations to the underlying abstract metrical pattern. Of interest in this regard are the prosodic theories and the poetic practice of late eighteenth and early nineteenth century German poets and theorists, in particular that of Johann Heinrich Voß, a poet and prosodist who exerted considerable influence on his contemporaries. Voß has frequently been criticized by twentieth century German theorists and prosodists, as well as by some of his contemporaries, for attempting to impose Greek verse forms on the German language without sufficient regard for the natural accentual patterns of German. Andreas Heusler, for example, refers to Voß' imitations of the Classical hexameter line as "un-German" distortions which violate the linguistic givens of the German language (Heusler 1956, 3: 271).1 Wolfgang Kayser (1965: 86) uses the word "unGerman" in referring to the spondees in the hexameters of Voß, Schlegel

48

David

Chisholm

and Platen, and agrees with Heusler that these poets infected German verse with what he called a foreign "spondee-sickness". 2 These criticisms, however, are based almost exclusively on accentual patterns (word and phrase accent), and rarely consider other linguistic features, such as word onset and lexicality, which might also be distinctive in verse. Within the framework of prosodic feature analysis, I shall examine the actual poetic practice of Johann Heinrich Voß more closely in an attempt to determine whether his lines of hexameter verse actually represent "unGerman distortions" of the language, and if so, what constitutes these "distortions" and where they occur. In order to determine how Voß' poetic practice is related to the German verse tradition, his verse will be compared on occasion with that of his more famous contemporaries Klopstock, Goethe, and Hölderlin, as well as with some prosodic conventions previously established by German poets. A discussion of Voß' poetic practice would be incomplete without an examination of his most important theoretical work on prosody, which bears the revealing title Die Zeitmessung der deutschen Sprache. This book, which first appeared in 1802 with a second edition in 1831, exerted a considerable influence on contemporary poets in Germany (Voß 1802). As the title of the work implies, Voß classifies the syllable types of German in terms of length rather than accent; he distinguishes three groups which he calls lang, mittelzeitig, and kurz. The second group is subdivided into types which he calls fastlang, schwebend, and fastkurz. This hierarchy is shown with examples below: (2) a. b.

c.

Syllable Type lang: mittelzeitig: (i) fastlang: (ii) schwebend: (iii) fastkurz:

kurz:

Examples Baum, blüht, klein -fach, -haft, -heit, -los, -schaft, -tum, -voll -bar, -sam, -sal, -in, -ing, -nis, -lei, -lein an, auf, aus, bei, durch; und, auch, denn, so, daß; ich, du, er, man, sich; -ich, -ig, -isch, -lieh, -ling, -ung ein, der, die, das; so, zu; Otto, /da, Dante, Mannes, Drittel, hölzern; Gehölz, zerwühlen, durch haun, um kreisen

The examples which he gives for "long" syllables are all nouns, verbs or adjectives, and those which he refers to as "almost long" are primarily "heavy" derivational suffixes such as -fach, -schaft, -tum, and so forth.

Lexicality

and the versification

of Johann Heinrich

Voß

49

The syllable types to which Voß assigned the label schwebend ("hovering" or "vacillating" contain suffixes spelled in a, ei, and i, and his "almost short" group contains monosyllabic prepositions, conjunctions, pronouns and "light" derivational suffixes spelled in i or u. Voß' examples for what he calls "short" syllables are mostly determiners, unstressed syllables of proper names and loan words, unstressed prefixes, and all syllables containing the reduced vowel "schwa". Though he consistently speaks of "length" in setting up this hierarchy, his classification correlates to a considerable extent with the opposition between lexical and nonlexical categories in the language. Those syllable types which he labels "long" are all lexical categories (i.e., nouns, nonauxiliary verbs, or adjectival adverbs) and most of those which he labels "almost long" are either modal auxiliary verbs (e. g., darf, kann, soll) or "heavy" suffixes such as -heit, -haft, -schaft, -tum). An interesting historical characteristic of these so-called "heavy" suffixes is that they still existed as independent lexical categories (heit, haft, schaft, etc.) at least as recently as the Middle High German period, whereas the "light" suffixes (-ig, -isch, -ich, -ling, -ung, etc.) already lost whatever independence they may have had either before or during the Old High German period. With the possible exception of -bar, the syllable types which Voß classifies as "vacillating", "almost short" and "short" either lost their independence before the Middle High German period, or are not known to have ever existed as independent lexical units. 3 Voß appears to be aware that the distinctions he draws in terms of length bear some correlation to the lexical history of the language, for he states that the length of "archaic root syllables" depends in part on the degree of their obsolescence: Mittelzeitig sind die veralteten Stammsilben hinter Benennungen ... Ob sie zur Länge oder zur Kürze geneigter sein, das hängt theils vom Grade der Veraltung, theils von der Schwere der Buchstaben ab ... (Voß 1802: 54-55)

But he refers to these archaic root syllables only in passing, and as his title "Die Ze^messung der deutschen Sprache" indicates, he views the prosodic function of words entirely in terms of the duration of each of their syllables. Thus, in Voß' terms, the second syllable of words such as Kirschbäume, hingehen, and heillose occurs in a metrically prominent position (hereafter labelled x) not because of its morphological or lexical characteristics, but rather because it is "longer" than the second syllable in words like heilige, himmlische, and herrliche, which occurs in a metrically nonprominent position (hereafter labelled o):

50

David

Chisholm

Ο Χ Ο Kirsch-bäu-me hingeh- en heil- lo- se

χ ο χ hei- Ii- ge himm-li-sche herr- Ii- che

In fact, however, these distinctions are essentially lexical and morphological as well as accentual. As a framework for this investigation I use a modified version of a theory of prosody originally proposed by Karl Magnuson and Frank G. Ryder (Magnuson and Ryder 1970: 789-820; 1971: 198-216; Magnuson 1974: 143-154). Applying the concept of complementary distribution to a large corpus of German iambic verse from the seventeenth through the early twentieth century, they identified a set of features which they found to be prosodically distinctive for German. These features either reinforce or disrupt the underlying metrical pattern, depending on their relation to it. In a "perfectly metrical" line of verse, all these prosodic features would reinforce the meter in every position. This hypothetical condition of absolute metricality, which represents the lowermost level of verse structure, indicates for each feature what is normal or expected. In terms of the underlying meter, all prosodic features either "affirm" (reinforce) or "disaffirm" (disrupt) a given position of the meter. The prosodic feature STRESS, for example, which is assigned to all German monosyllabic nouns, main (i. e., nonauxiliary) verbs, adjectival adverbs, and to all syllables bearing primary word stress, reinforces or "affirms" the prominent positions of the meter and disrupts or "disaffirms" the nonprominent positions. Fundamental to the study of verse structure, however, are not only the prosodic characteristics of individual syllables, but also the relationships between sequences of syllables. This fact was stated many years ago by Otto Jespersen, who observed that the stress level of a syllable can only be determined by comparison with a contiguous syllable: Our ear does not really perceive stress relations with any degree of certainty except when the syllables concerned are contiguous. If two syllables are separated by a series of other syllables, it is extremely difficult even for the expert to tell which of them is the stronger ... What is decisive when words have to be used in verse is everywhere the surroundings: the metrical value of a syllable depends on what comes before and what follows after it. (Jespersen 1933: 249-274)

In Germanic languages, metered verse regularizes a tendency in the languages themselves toward an alternation of prominent and non-promi-

Lexicality

and the versification

of Johann Heinrich

Voß

51

nent features. This principle of alternation states that whenever a prosodic feature of the verbal material occurs in a nonaffirming relation to the meter, that same feature must occur in the immediately following position (where it affirms the meter). It may be formalized as follows: 4 (3)

[F] -

[Fj] / [F,]

The brackets enclose a metrical position, and F represents any feature or cluster of features within that position. The subscript 1 indicates that the nonaffirming feature and the immediately following affirming feature are identical. The feature or cluster of features to the right of the diagonal slash always disaffirms the meter and must therefore be followed (in the position represented by the horizontal line) by at least one affirming feature. Any feature or feature cluster to the left of the arrow also dissafirms the meter. The arrow means that the syllable to its left must at least contain the feature or feature cluster to its immediate right. The latter always affirms the meter. The slash means "in the environment". Thus the rule in (3) states that any feature which occurs in a nonaffirming relation to the meter must be followed in the next metrical position by that same feature (where it affirms the meter). In an attempt to determine the manner in which both accent and lexicality function in the poetry of Johann Heinrich Voß, we shall examine his long hexameter poem Luise, first published in 1795, as well as a collection of eighteen Idyls written between 1774 and 1802 (Voß [1969]). These poems consist predominantly of Classical hexameter verse, as well as a few lines written in other meters. The metrical pattern for the Classical hexameter line is shown in (4). Each line contains six metrically prominent positions, which are always monosyllabic, alternating with six metrically non-prominent positions, the first four of which may be either monosyllabic or disyllabic. In the hexameters of most poets, the fifth nonprominent position must be disyllabic,5 and the sixth nonprominent position is always monosyllabic: (4)

1 2 3 4 5 6 7 8 9 10 11 12 χ ο (ο) χ ο (ο) χ ο (ο) χ ο (ο) χ ο ο χ ο

In the pattern, the odd numbers represent prominent positions of the meter and the even numbers represent nonprominent positions. In the latter, the parentheses enclose an optional second syllable. The linguistic realization of this metrical pattern is illustrated by the following lines from Voß' Idyls:

52

(5a)

David

Chisholm

x o x o o x o o x o x o o x o Plötzlich rief aus dem Bette mit leiser Stimme die Mutter χ o x o o χ ο ο χ ο χ ο ο Hell die Decke von Licht, und die Wang' in rosigem χ ο Schimmer

(5b)

χ o Laß dein χ ο Schatten

x o ο χ ο χ ο ο Hütchen nur schief. Kühl wehts in dem ο χ ο der Pappel

χ ο χ ο χ ο ο χ ο ο xo ο Freies Werk schafft Segen und Fröhlichkeit! Michel, du χ ο bringest Earlier it was observed that accented or stressed syllables affirm the prominent positions of the meter and disrupt the nonprominent positions. In (5a) the stressed syllables affirm the meter, fulfilling the expectations of the reader/listener in all positions: all x-positions contain the feature + STRESS (abbreviated [+ST], where the square brackets represent a metrical position) while all o-positions lack this feature and are therefore -STRESS (abbreviated [-ST]. In (5b), however, the stressed syllables kühl and schafft occur in o-positions and therefore disrupt or "disaffirm" the meter. Here the presence of the nonaffirming feature +STRESS in an o-position places certain constraints on the actualization of the immediately following x-position. This temporary disruption of the metrical pattern can be compensated for by a prosodic rule which stipulates that the immediately following position must affirm the meter: (6)

[F] -

X [+ST]

/ 0 [+ST]

This rule states that any occurrence of + STRESS in an o-position must be followed by +STRESS (i.e., the same feature) in the immediately following x-position. Implicit in this rule is the assumption that stressed syllables affirm the meter when they occur in prominent positions and disrupt it when they occur in nonprominent positions. The following lines represent more complex actualizations of the Classical hexameter line than those given in (5):

Lexicality

(7)

and the versification

of Johann Heinrich

Voß

53

x o o x o o x o x o x o o x o Jesus der große Prophet, kommt in die Gräber hernieder (Klopstock, Der Messias II) χ οχ ο χ ο ο χ ο ο χ ο Lahm gebläut! Sehn will er vergnügt, wie die Welt sich οχ ο verändert (Voß, Idyllen) X o x o o x o x o x o o Und die künft'ge Gefahr hielt nicht die grimmige χ ο Wut auf (Goethe, Hermann und Dorothea VI) x o x o o χ ο x o o x o o Also sagt' ich und jetzt kehr' ich an den Rhein, in die χ ο Heimath (Hölderlin (1951: 81)

In each of these lines, a stressed syllable in a nonprominent position (o[+ST]) is followed by an unstressed syllable in a prominent position (x[-ST]); thus each contains a violation of the prosodic rule given in (6). Lines such as the following, however, which also contain the sequence o[+ST] x[-ST], do not occur in German hexameter verse: (8)

x x χ ο χ χ *Also sagt' ich und jetzt kehren wir zum Rhein, in die χ ο Heimath

The occurrence of the sequence kehr' ich and the absence of the sequence kehren in the metrical relation o-x can be accounted for in terms of word boundaries: ich begins a new word, while -ren is a noninitial syllable. This reflects the fact that not only stress, but also word boundaries determine whether certain sequences may occur in o-x in German hexameter verse. By adding word onset [WO] as a prosodic feature to the rule given in (6), it can be rewritten as follows: (9)

X [-ST]

-

[+WO] / 0 [+ST]

This rule states that, after a stressed syllable in an o-position, any unstressed syllable in the following x-position must also be word-initial (i. e.,

54

David

Chisholm

must be [+WO]). This rule allows occurring sequences like kehr' ich and excludes non-occurring sequences like kehren from the metrical relation o-x.

There is, however, a disyllabic word-type in which the initial stressed syllable does occur occasionally in o-x in German hexameter verse. Examples are given below: (10)

ο

χ

Oft um Mitternacht wehklagt die bebende Lippe (Klopstock "Die künftige Geliebte") ο χ Legete, daß auch der Junker verstört aussah in dem Kirchstuhl (Voß, Idyllen) ο χ Selbst hinging nach Paris und bald den schrecklichen Tod fand (Goethe, Hermann und Dorothea IV) ο χ All' die gesammelte Kraft aufflammt in üppigem Frühling (Hölderlin, "Der Wanderer") In the hexameter verse of most German poets, such disyllabic "metrical inversions" (i.e., the sequence o[+ST] x[-ST, - W O ] ) occur much less frequently than monosyllabic inversions (i.e., o[+ST] x[-ST, +WO]). When they do occur, however, the second syllable is always lexical, i. e., it is a syllable (such as klagt, sah, ging in wehklagt, aussah, hinging) which occurs as an independent word in the German lexicon. When these lexical syllables occur as unstressed elements of compound words (7»"« ging, 'aussah, etc.) they still have the potential to be used in German hexameter verse as if they were stress-bearing lexical monosyllables; in words such as gehen, heilig, himmlisch, on the other hand, the unstressed nonlexical syllables do not have this potential. 6 This suggests that, even though stress is the primary factor, both word boundaries and lexicality play secondary and tertiary roles in the prosodic systems of most German poets. The occurrence of verse lines of the type given in (10) can be accounted for by incorporating lexicality as a prosodic feature into the o-x rule: (11)

X [-ST,

-WO] -

[+LX] / 0 [+ST]

Lexicality and the versification of Johann Heinrich Voß

55

This rule states that, after a stressed syllable in an o-position, any unstressed, noninitial syllable in the immediately following x-position must be lexical [+LX], In the context of these rules of prosody it is instructive to compare the poetic practice of Johann Heinrich Voß with that of some of his contemporaries. Table 1 gives the frequency and distribution of two types of o-x sequences in hexameter verse by Klopstock, Voß, Goethe, and Hölderlin. The first and second rows for each poet represent respectively sequences of the types kommt in and Wehmut. As previously mentioned, the second syllable of the latter sequence (o[+ST] x[-ST, —WO]) is always lexical in German hexameter verse.7 For Klopstock, Goethe, and Hölderlin, the frequencies of these two prosodic sequences in o-x are what we might expect: those of the type kommt in (stressed lexical monosyllable + unstressed nonlexical monosyllable) occur much more frequently than those of the type Wehmut, reflectTable 1. Positional distribution of the sequence o[+ST] x[—ST] in German hexameter verse o-x sequences 2-3

4-5

6-7

8-9

Total

Lines

Ratio

10-11

Klopstock, Messias I and II (1799 edition) o[+ST] x[-ST, +WO] o[+ST] x[-ST, - W O ]

7 1

4 1

6 0

2 2

0 0

19 4

1617 1617

0.0116 0.0025

Voß, Luise I, II, III (1795) o[+ST] x[—ST, +WO] o[+ST] x[-ST, - W O ]

1 6

1 17

0 18

1 7

0 0

3 48

1766 1766

0.0017 0.0272

Goethe, Hermann und Dorothea (1797) o[+ST] x[-ST, +WO] o[+ST] x[-ST, - W O ]

0 1

0 0

4 0

1 0

0 0

5 1

2034 2034

0.0025 0.0005

Hölderlin, Gedichte (1796-1805) o[+ST] x[-ST, +WO] o[+ST] x[-ST, - W O ]

1 0

2 1

7 4

0 0

0 0

10 5

967 967

0.0103 0.0052

Examples: o[+ST] x[-ST, +WO]: Kommt in, ging von o[+ST] x[-ST, - W O ] : Wehmut, hinging ( + W O = word-initial syllable; —WO = non-initial syllable)

56

David

Chisholm

ing the fact that word-initial syllables tend to affirm the meter in x-positions, while noninitial syllables tend to disrupt it. In the hexameter lines of Voß, however, exactly the opposite is the case: sequences of the type Wehmut have a high frequency of occurrence that is far out of proportion to that of the type kommt in. This disproportionality is a result of Voß' attempt to create in German the effect of a classical spondee (two successive long syllables in Greek and Latin versification) by placing words which he considered to consist of two "long" syllables in the metrical relation o-x. As we have seen, the German syllable types which he considered "long" ("lang" and "fastlang" in his terminology) are essentially lexical. Since there are relatively few disyllabic words of the type Jahr'zehnt or vollzieht in German (in which an initial unstressed syllable is also lexical), Voß turned to the more frequently occurring compounds of the type hinging, Festtag, Sturmnacht. It is this attempt to recreate the spondee in German which led Andreas Heusler to coin terms like Spondeenjagd, Spondeensucht, Spondeenkrankheit, and falsche Spondeen in referring to the hexameters of Voß, A. W. Schlegel, Platen, and others. There is another type of compound word which occurs not only in hexameter verse, but also in other meters used by German poets, namely trisyllabic words consisting of two stressed root morphemes followed by one or more inflectional morphemes. Examples are given in (12): (12)

a.

Classical hexameter lines: ο χ ο Ringsum durch den unendlichen Raum nachahmend ergießet (Klopstock, Messias I) ο χ ο ο Seiner Luise zur Lust, hausväterlich prangend im Schlafrock (Voß, Luise I) ο χ ο Aber Mama, sanftlächelnd der wohlbekannten Erzählung (Voß, Luise I) ο χ ο Keine Spur nachlassend von seiner lebendigen Wirkung (Goethe, Hermann und Dorothea III) ο χ ο Rollt der König den Blick; irrlächelnd über den Ausgang (Hölderlin, "Der Archipelagus")

Lexicality

b.

and the versification

of Johann Heinrich

Voß

57

Iambic verse: ο χ ο Dein Wort in mir arglistig zu versehren (Gryphius, "Auf den Sonntag ...") ο χ ο Als Diamante spielt und ohn' Aufhören brennt (Gryphius, "An die Sternen") ο χ ο χ Und Körper, die die Kraft gleichfallender Gewichter (Gryphius, "Auf seinen Geburtstag") ο χ ο Wie viel andächtig schwärmen leichter, als (Lessing, Nathan der Weise 1,2) ο χ ο Mich besser kennt. — Schatzmeister bin ich bei (Nathan 1,3) ο χ ο Vorteile, die das Volk nicht kennt, kennst du. {Nathan 111,5) ο χ ο Anmutig Tal! du immergrüner Hain! (Goethe, "Ilmenau") ο χ ο Ihr Kinder, glaubt: ohnmächtig bleibt der Wille (Goethe, Sonette XIV)

Words of this type {nachahmend, aufhören), which are quite common in German iambic verse since the seventeenth century, almost always begin in an o-position, despite the fact that the stress level of the second syllable is less than that of the first. In contrast to disyllabic compounds {Anmut, aufhört), these compounds are consistently used by German poets from the seventeenth through nineteenth centuries as if they consisted of two separate words (e. g., nach lassend, sanft lächelnd, gleich fallend). They are accepted by the o-x rule given in (6), since the second syllable, which bears a greater degree of stress than the third syllable, is [+ST] in the binary feature system. Voß and other German poets used these trisyllabic compounds, as well as disyllabic compounds, to create "spondees" in Classical meters. Thus far discussion of the hexameter verse of Voß and his contemporaries has been limited to the metrical sequence o-x. But this represents

58

David

Chisholm

only part of the metrical structure of the Classical hexameter. The other parts of this structure are the sequences x-o and, in disyllabic nonprominent positions, the sequence o-o. German poets frequently attempted to create spondees in x-o, for example in imitation of the "fermata" in the "anceps" syllable at the end of the line: (13)

χ ο Seelenruh, und Ernst, und Erbarmung, als er vor Gott stand (Klopstock, Messias I) χ ο Hielt der redliche Pfarrer von Grünau heiter ein Gastmahl (Voß, Luise I) χ ο Trieb nach Paris zu gehn, dahin, wo er Kerker und Tod fand (Goethe, Hermann und Dorothea IX) Friedsam geht aus dem Walde der Hirsch ans freundliche Tagslicht (Hölderlin, "Der Wanderer")

The final syllable in these lines, though it does not bear primary stress, is in each case an independent lexical item which bears more semantic "weight" than nonlexical unstressed syllables. The following lines illustrate some of the ways in which poets attempted to create the effect of two "long" syllables in the metrical relation x-o in various positions of the hexameter line: (14)

χ ο Schimmern sie, Vorbilder der gottversöhnten Gemeinen! (Klopstock, Der Messias I) Und braunkolbiges Ried; Seelilien jezo durchrauscht' er, Die gelb blühten und weiß, breit blättrig; jezo den Vorgrund (Voß, Luise I) Denn kennt jemand den Herrn, so kann er ihm leichter genug tun (Goethe, Hermann und Dorothea VIII) Die hoch herzig ein Mädchen vollbrachte, die treffliche Jungfrau (Goethe, Hermann und Dorothea VI)

These x-o sequences are essentially of two types: (1) those which consist of two [+ST] syllables (e.g., Ried; See-, weiß, breit-) and (2) those consisting of a [—ST] syllable followed by a [+ST] syllable (e.g., und braun-, die gelb). In terms of the stress levels of these sequences, the first

Lexicality

and the versification

of Johann Heinrich

Voß

59

([+ST] [+ST]) creates the effect of a "spondee" in German. It is questionable, however, whether the second sequence ([-ST] [+ST]) should be considered a "spondee" at all, since the first (non-lexical) syllable is clearly unstressed. Consider the following stress rule: (15)

[F] — 0 [ - S T ] / X [-ST]

This rule accepts sequences of the first type {weiß, breit-, Kind! hell, Nacht schuf, Tal geht), as in the following lines: (16)

χ ο Schützt vor der blendenden Lampe das Kind! hell glänzet die Windel (Voß, "Das erste Gefühl") χ ο Die Nacht schuf tausend Ungeheuer (Goethe, "Willkommen und Abschied", iambic tetrameter line) χ ο Als wir nun aber den Weg, der quer durchs Tal geht, erreichten (Goethe, Herrmann und Dorothea)

and rejects sequences of the second type (und braun-, die gelb). When the sequence x[—ST] o[+ST] does occur, however, it is usually followed by a compensatory stressed syllable in the immediately following prominent position. This stressed syllable helps to re-establish the meter after the disruption of metrical equilibrium in the two preceding positions of the line: (17)

χ ο χ Die gelb blühten ... Die hochherzig ...

With regard to these sequences, it is instructive to compare the poetic practice of Voß with that of Klopstock, Goethe, and Hölderlin. Table 2 shows the frequency and distribution of violations of the x-o rule given in (15) in the hexameters of these four writers. For each poet these x-o sequences are classified into two types: those which are followed by the metrically affirming feature x[+ST], and those followed by the nonaffirming feature x[—ST]. For all four poets, these data show a clear tendency for the hexameter line to regain metrical equilibrium immediately after a violation of the xo rule. They also show that this rule is violated much more frequently by Voß than any of the other poets considered. As expected, these violations

60

David Chisho Im

Table 2. Frequency and positional distribution of the sequence x[~ST] o[+ST] in German hexameter verse x-o Sequences Total Lines Ratio 1-2

3-4

5-6

7-8

9-10

11-12

Klopstock, Messias I and II (1799 edition) x[—ST] o[+ST] x[+ST] x[—ST] o[+ST] x [ - S T ]

24 4

8 0

4 0

10 0

1 0

0 0

47 4

1617 1617

0.029 0.002

Voß, Luise I, II, III (1795) x [ - S T ] o[+ST] x[+ST] x [ - S T ] o[+ST] x [ - S T ]

62 3

14 4

17 0

12 2

0 0

0 0

105 9

1766 1766

0.059 0.005

Goethe, Hermann und Dorothea (1797) x [ - S T ] o[+ST] x[+ST] x[—ST] o[+ST] x [ - S T ]

18 2

1 0

1 0

1 0

0 0

0 0

21 2

2034 2034

0.010 0.001

Hölderlin, Gedichte (1796-18050 x [ - S T ] o[+ST] x[+ST] x [ - S T ] o[+ST] o [ - S T ]

3 0

2 2

1 5

7 0

0 0

0 0

13 7

967 967

0.013 0.007

Examples: x[-ST] of+STJ x[—ST]

of+STJ

x[+ST] - Als voll (wurde), die gelb (blühten), die hoch(herzig) mit lehr(reichem) x [ - S T ] - Die vor(längst), dann nimmt sie

occur most frequently at the beginning of the line (positions 1-2), decrease in the middle (positions 2-3, 5-6, 7-8) and do not occur at all at the end of the line (positions 9-10, 11-12). This reflects a general tendency characteristic of all meters in the German verse tradition: disruptions of the meter occur most frequently at the onset of the verse line, and are filtered out toward the end, as metrical equilibrium is regained and the line thereby stabilized. In 62 of the x-o-x sequences in positions 1 -3 in Voß' Luise, the violation of the x-o stress rule is followed by a stressed syllable, while only 3 are followed by an unstressed syllable. No less than 53 of these 62 lines are of the following type: (18)

χ ο χ ο Mit lehrreichem Gespräch zu erfreun, und mancher Erzählung {Luise I)

Lexicality

and the versification

of Johann Heinrich

Voß

61

The other nine lines are of the type: (19)

χ ο χ ο Wann heiß werden die Tag, und die blühende Bohne betäubet {Luise / )

It is the overabundance of sequences such as these which Andreas Heusler criticizes as "one of the sickest distortions which one can demand from Germanic linguistic material" (Heusler 1968; 3: 271). In a reference to Voß' previously mentioned Zeitmessung der deutschen Sprache, Heusler points out the faulty reasoning which led poets such as Voß, A. W. Schlegel, and Platen to use such sequences in their misguided attempts to recreate the Classical spondee in German: Voß setzt den langlebigen Irrtum auf den Thron: das bloße 'Accentuiren' sei eine kindliche Roheit gewesen; nun gelte es, der deutschen Kunst das Längemessen beizubringen. Jede Silbe hat ihr Zeitmaß. Danach addiert man die Silben eines Taktes; zusammen müssen sie die richtige Summe geben. Ist die gehobene Silbe zu kurz, muß die gesenkte um so länger sein: : der hoch : (donnernde) ist ein möglicher Hexametertakt, denn hoch bringt ein, was der zu wenig hat. (In Wahrheit: einer kräftigen Senkung kann sich auch die schwächere Hebung überordnen. Was man sogleich als Widersinn erkennt.) (Heusler 1968, 3: 85)

Note that the problem here is not with words of the type lehrreichem per se — as assumed by Wolfgang Kayser (1965: 86, 313) 8 — but rather with the metrical sequence x-o, a sequence which, in both German and English verse, is much more highly constrained than o-x. It is above all the violations of the x-o rule given in (15) which represent severe distortions of the German language. Any discussion of German hexameter verse would be incomplete without including the disyllabic o-positions and the kinds of verbal material which may or may not occur there. These positions are marked by roman type in the following lines, which illustrate various possibilites: (20)

χ o x o o x o o x o o χ o o x o Sing, unsterbliche Seele, der swidigen Menschen Erlösung (Klopstock, Messias I) Sorglos saß nach dem Mahle der Greis fort, sich und die andern (Voß, Luise I) C/nrat sich häufet und Unrat auf allen Gassen herumliegt ... Künftig die Piaterstadt selbst, so klein sie auch sei zu verzieren (Goethe, Herrmann und Dorothea III)

62

David Chisholm

Leuchtest du von /z/mmlischem Glanz, und so, wie sie wandeln (Hölderlin, "Der Archipelagus") Lwgend in ärmlicher Küche kopf.se/iwrtelnd in halbleere Speisen v46passend dann die Ersc/iö/?ften am Gatter der Gruben und Werften Viel davon hörtet ihr. Dies aber ist, was die Klassiker sagen Siegreich kommt eine zurück Glocken

und die andere läutet die

Leibeigne wurden zu i^hZi/bürgern. Hinter den s/wrmsichern Pfählern (Brecht, Das Manifest) Disyllabic o-positions, whether in Classical meters or in pure dactylic verse, tend to be leicht und fließend. A possible explanation for this tendency is the need to maintain a strong contrast between prominent and nonprominent positions of the meter. When an o-position contains two syllables, these must be especially light, so that their combined "weight" does not disrupt the contrast with the syllables in x-positions. Thus late eighteenth century German poets tended to avoid using "heavy" syllables in disyllabic o-positions. Voß accordingly avoided using syllables in these positions which he considered to be "lang" or "fastlang". Table 3 shows the distribution of stressed syllables in disyllabic o-positions in hexameters by Klopstock, Voß, Goethe, Hölderlin and, for contrast, Bertolt Brecht. 9 Whereas two stressed syllables may occur in succession in the metrical sequences o-x and x-o, this is never the case, even in the hexameters of Brecht, within disyllabic o-positions. Thus there are no lines of the type: Table 3. Distribution of [+ST] in disyllabic o-positions in German hexameter lines

syllables 1 2 1 2 Klopstock Voß Goethe Hölderlin Brecht

2 0 2 1 39

10 0 3 0 20

1 1 3 0 31

1 2 7

0 10 0 6

1 0 0 0 35

4 0 0 0 14

1 2 1 0 0 0 26

1 2 1

0 5 2 11

2 1 5 0 32

10 0 5 0 5

1 2 7 2 10 1 163

32 0 23 2 56

1617 1766 2034 967 379

0.0241 0.0011 0.0162 0.0031 0.5778

Lexicality

(21)

and the versification

of Johann Heinrich

Voß

63

χ ο ο χ ο ο χ ο ο χ ο ο χ * Schön, weit, hoch, herrlich der Blick! rings ins Leben hinein, ο ο χ ο vom Gebirge [construct]

since both weit and hoch are stressed in this hypothetical line. The absence of the sequence of o[+ST +ST]o (where "o[" and "]o" represent respectively the first and second syllables of the disyllabic o-position) can be expressed by two rules: (22)

0 [F

—• 0 [ ~ S T / +ST] 0 F] 0 - - S T ] 0 / 0 [ + S T

The left and right brackets define respectively the first and second syllables of the disyllabic o-position. The first rule states that, if the second syllable in an o-position is stressed, the first must be unstressed; conversely the second rule states that if the first syllable is stressed, the second must be unstressed. The disyllabic o-positions in the hexameters of Voß and Hölderlin are even more highly governed: with very few exceptions, stressed syllables do not occur there at all. Furthermore, there is another prosodic feature which distinguishes Voß' poetic practice from that of the other poets; not only stressed syllables, but also all lexical syllables are excluded from disyllabic o-positions in his hexameters. Lines such as the following (in which the lexical syllables in these positions are emphasized) do not occur in either Luise or the Idyllen: (23)

χ ο ο χ ο Dir nur ist es bekannt, mit was vor Einmuth wir damals (Klopstock, Messias I) χ ο ο χ Wenig Edelmut zeigt' er uns da. Ihr wißt es, mein König! (Goethe, Reineke Fuchs, zehnter Gesang) χ ο ο χ Warm undfröhlich wie einst, Vater\andserde den Sohn (Hölderlin, "Der Wanderer", pentameter line of distich)

The following line from Voß' Idyllen illustrates the degree to which his hexameters are bound by lexical constraints: (24)

χ ο χ Hin zum Johannsbeerstrauch, wo jeglichen Morgen des Hänflings

64

David Chisholm

While Klopstock, Goethe and even Hölderlin could have accommodated the third and fourth syllables of the word χ

ο

ο

χ

Johannisbeerstrauch in a disyllabic o-position, Voß felt constrained to eliminate the third syllable so that the lexical syllable -beer- would occur in a monosyllabic oposition. For Voß, the rules governing disyllabic o-positions therefore have the following highly restrictive form: (25)

F] —» — ST, - LX] 0 / 0 [F [F - 0 ['-ST, - L X /

F] 0

The approach to meter and language outlined here reveals clear prosodic differences in the poetic practice of the late eighteenth century poets Klopstock, Voß, Goethe, and Hölderlin. It also demonstrates the importance of distinguishing "iambic" and "trochaic" sequences (o-x and x-o), as well as the "dactylic" sequences in disyllabic o-positions (o-o). The poems analyzed show that a characteristic of iambic and trochaic verse in Germanic languages also holds for hexameter verse: the metrical sequence x-o is much more highly constrained than the sequence o-x. The frequent "weak spots" in the prosodic system of Johann Heinrich Voß what Heusler refers to as distortions and a painful outrage against the German mother tongue - are almost all violations of the rule which stipulates what kind of verbal material may occur in x-o sequences. The data also show that the hexameters of Voß, in contrast to those of the other poets, are less constrained by the stress rules of the language, but more highly governed by its lexical structure. Notes 1. Heusler writes as follows about the hexameters of Voß, A. W. Schlegel, and Platen: Wie in der Ode, so hat man im Hexameter unter dem Schutze des kräftigen Spondeus die tonschwache vorgeneigte Silbe (die letzte von ihrem Kolongipfel) in die Hebung befördert: eine der übelsten Verzerrungen, die man germanischem Sprachstoffe zumuten kann: mit fuchs/?e/ze verbrämt; zum e'mträchtigen Tanz; als voll wurde, das Jahr, aber es hat Knechtschaff usw. ..." (1956: 271). Im Versbau äußert sich Vossens Strenge zumeist darin, daß er - ν und - für eine im Deutschen nachahmbare Zweiheit hält und daraus die Folgen zieht bis zu schmerzlicher Kränkung der Muttersprache (1956: 215). 2. Kayser makes the following comment about the spondees of Voß, A. W. Schlegel and Platen: "Ihre Spondeen klingen uns undeutsch, und Heusler hat das Wort von der 'Spondeenkrankheit' geprägt, die durch sie in die deutschen Verse gekommen sei" (1986: 86).

Lexicality and the versification of Johann Heinrich Voß

65

3. For a more detailed discussion of the distinctions between these two suffix-types, see Chisholm (1973: 27-38). 4. With some modifications, I follow here the approach to poetic meter first proposed by Magnuson and Ryder 1970. 5. The fifth non-prominent position is occasionally monosyllabic in the hexameters of Klopstock and Voß. 6. Elsewhere (Chisholm 1973) I have shown that in German poetic practice from the seventeenth to the early nineteenth century, the "cut-off point" for this potential is the Middle High German period: unstressed derivational suffixes which still existed as independent words in Middle High German (e. g., heit, haft) can occur in an x-position in words like Freiheit, herzhaft. 7. In German iambic verse, on the other hand, there are occasional exceptions to this rule, particularly in the poems of Rainer Maria Rilke. 8. Kayser fails to distinguish the severity of the disruption of the meter in the following xo-x sequences: χ o x (1) die Freiheit [ - ST] [+ST] [ - ST] - most disruptive (2) noch weghalten [ - S T ] [+ST] [+ST] - disruptive (3) Raub darstellte [+ST] [+ST] [+ST] - only mildly disruptive 9. The figures for the first four poets are based on the same corpus as Tables 1 and 2. The figures for Brecht are based on the 379 hexameter lines in his uncompleted poem Das Manifest.

References Allen, W. Sidney 1973 Accent and rhythm: Prosodic features of Latin and Greek. Cambridge: Cambridge University Press. Attridge, Derek 1982 The rhythms of English poetry. London: Longman. Bockelmann, Eske 1991 Propädeutik einer endlich gültigen Theorie von den deutschen Versen. Tübingen: Niemeyer. Brecht, Bertolt 1967 Gesammelte Gedichte. Vol. 3. Frankfurt. Suhrkamp. Chisholm, David 1973 "Lexicality and German derivational suffixes: A contribution to the Magnuson-Ryder theory of prosody", Language and Style 6: 27-38). 1977 "Generative prosody and English verse", Poetics 6: 111-153. Goethe, Johann Wolfgang 1977 Hermann und Dorothea. Stuttgart: Reclam. Hellmuth, Hans Heinrich and Joachim Schröder (eds.). 1976 Die Lehre von der Nachahmung der antiken Versmaße im Deutschen. München: Fink. Hensen, Walther 1965 Deutsche Wortbildung. Tübingen: Niemeyer. Heusler, Andreas 1917 Deutscher und antiker Vers: Der falsche Spondeus und angrenzende Fragen. Straßburg: Trübner.

66

David Chisholm

1929 Deutsche Versgeschichte. Vol. 3. Berlin: Walter de Gruyter. [1956] [Reprint. Berlin: Walter de Gruyter.] Hölderlin, Friedrich 1943 Sämtliche Werke. Vol. 1. Edited by Friedrich Beissner. Stuttgart: Cotta. 1952 Sämtliche Werke. Vol. 2. Edited by Friedrich Beissner. Stuttgart: W. Kohlhammer. Humboldt, Wilhelm von 1976 "Über Johann Heinrich Vossens metrische Ansichten", in: Hans Heinrich Hellmuth and Joachim Schröder (eds.), 277 278. Jespersen, Otto 1933 "Notes on metre", Linguistica: Selected papers in English, French and German. Copenhagen: Levin and Munksgaard, 249-274. Kayser, Wolfgang 1965 Das sprachliche Kunstwerk. Bern: Francke. Klopstock, Friedrich Gottlob 1974 Der Messias. Vol. 1. Edited by Elisabeth Höpker-Herberg. Berlin: Walter de Gruyter. Küper, Christoph 1988 Sprache und Metrum: Semiotik und Linguistik des Verses. Tübingen: Niemeyer. Lessing, Gotthold Ephraim 1964 Nathan der Weise. Stuttgart: Reclam. Magnuson, Karl 1974 "Rules and observations in prosody", Poetics 12: 143-154. Magnuson, Karl and Frank G. Ryder 1970 "The study of English Prosody: An alternative proposal", College English 31: 789-820. 1971 "Second thoughts on English prosody", College English 33: 198-216. Voß, Johann Heinrich 1802 Die Zeitmessung der deutschen Sprache. Beilage zu den Oden und Elegien. Königsberg: Nicholovius. 1802 Sämtliche Gedichte. Königsberg: Nicholovius. [1969] [Reprint. Bern: Herbert Lang.] Wagenknecht, Christian 1981 Deutsche Metrik: eine historische Einführung. München: Beck.

backen > buk/backen > backte: Α study in grammatical variation Nada Μ. Cook

1. Introduction A set of verbs in German displays variable forms in the preterit and the past participle. With only a few exceptions, the verbs in this set are strong and have an inflection that is marked by gradation of the root vowel. Belonging to a system whose members by and large conform to a different inflectional pattern, i. e., a suffixal one, they have begun to shift to fit the mold of the predominant paradigm. The alternating inflections to which they have adapted and the divided usage of both inflectional forms are by all accounts manifestations of restructuring and constitutive changes affecting the German verb system. Standard reference sources have traditionally argued in favor of the semantic and functional synonymy of such forms. Some recent studies (Quirk 1970, Rauch 1987, Haber 1975) examining the distribution of verbs with alternating inflections cite evidence, however, that the incidence of their alternants may not be arbitrary. They contend, rather, that their usage may be conditioned by specific constraints.

2. Method To test the hypothesis that verbs with dual inflections have a motivated distribution, an experiment was conducted with a set of eleven verbs known for their variable simple past and past participles: backen bersten bewegen brennen (er) bleichen einladen ( er) schallen

buk/backte barstlber stete bewog/bewegte brannte! brennte (er) blich! ( er) bleichte lud ein/ladete ein ( er) scholl! ( er) schallte

gebacken!gebackt geborsten!geberstet bewogen/bewegt gebrannt!gebrenn t (er) blichen! (er) bleicht einge laden!eingelade t ( er ) schollen! ( er ) schallt

68

Neida Μ. Cook

melken saugen schmelzen weben

molk/melkte sog/saugte schmolz!schmelzte wob! webte

gemolken!gemelkt gesogen!gesaugt geschmolzen!geschmelzt gewoben!gewebt

Of the several linguistic and paralinguistic factors which may account for the variation, the following were explored in this study: 1) syntactic/ semantic properties of the verb; i.e., aspect, transitivity function, concreteness of the designated action, type of accompanying NP; 2) degree of formality of contextual setting; 3) verb frequency; 4) age and gender of the speaker. The experiment focused exclusively on the distribution of inflectional variants in spoken German and made no attempt to investigate its representation in writing. Accordingly, the method deemed most appropriate was one which depended solely on auditory cues. For this purpose, some hundred sentence pairs, each pair containing first the suffixal and then the ablaut variant of a verb, were presented orally to a group of informants. Interviewed individually, the subjects were asked to make paired comparison judgements and select one version of each sentence they would be most likely to use in a given situation. Although asked to opt for one sentence in the pair, the subjects had the choice to accept or reject both versions, if they found both (in)appropriate. They were also encouraged to go beyond a mechanical response and comment on their choices. In all, 48 native speakers of German, their ages ranging from 15 to 85, were tested for their usage of the eleven verbs with variable simple past and past participles. The subjects, 21 male and 27 female, were high school and university students, currently employed or retired professionals, all with at least a high school education. Although with diverse regional and dialectal backgrounds, all the subjects spoke and communicated principally in standard German.

Table 1. Subjects grouped by age and gender

Total men women

Group I

Group II

Group III

Group IV

Group V

14-21

22-35

36-50

51-65

66 ^

14 8

15 7

4

13 5

6

8

4

8

2

Total

48 21

27

backen>buk!backen>backte:

Grammatical

variation

69

3. Data analysis The analysis of the subjects' responses to the corpus examples, which were designed to provide linguistic environments where possible patterns in usage could be detected, yields results (Table 2) which clearly show that the degree of stability in the paradigm varies among the individual verbs in the set. The statistics reveal as well a greater consistency in the inflectional patterns for the participles than for the simple past forms. In the preliminary attempt to seek out a pattern in usage of verbs with variable inflections, the analysis necessarily had to proceed from the data containing some evidence of variation. Excluded, therefore, were subjects' responses which offered no such evidence. Because the speakers differed in their treatment of the simple past as opposed to the participial forms, in that they may have alternated between both inflections in the past tense, but not necessarily in the perfect tenses, a separate data count had to be kept for both in this experiment. As shown in Tables 3 and 4, approximately 52% of those tested relied on both inflectional variants in the simple past (Table 3) and approximately 46% alternated between both inflections in their usage of the participles (Table 4). It should be noted, however, that the cumulative count in Table 4 includes not only instances where participles complement auxiliary verbs in the perfect tenses. It encompasses as well nominal phrase constructions, where participles are used attributively, e. g., frisch gemolkene!gemelkte Milch 'freshly drawn milk', die eingeladenen/einge lade ten Gäste 'the invited guests'. Without such uses some verbs offer considerably less evidence of variation in environments containing participial constructions: e. g., 86% of the alternation between weak and strong variant forms for einladen, 17% for backen, 12% for schmelzen, 11% for bersten and 9% for melken were recorded in test frames where participles complement nouns. Considering that in this test battery nominal phrase constructions were far less frequent than those where participles appear in the perfect tenses (by a ratio of approximately one to four), the statistical evidence becomes even more compelling. The relative stability of the participles and what appears to be their lesser engagement in variation, i.e., language change, may be attributed to their comparably lower functional load: void of signalling grammatical distinctions, i.e., person and number, they pose a lesser demand on the system to facilitate restructuring. However, once drawn into the dynamic process of change they are likely to offer less resistance and conform more readily to the new paradigm. The findings presented in Table 2

70

Nada Μ. Cook

α

so s

a !2 § ι α

ag § I t; so Ο

§ι-, I«Ν Ü CS α 3 Ο Ui oc ο 2

Ο. (Ν § i Ii Ο w

O O ^ O

Tf Ο -ί ο

(Ν Ο Ό

r- ο so ο

Ο (S Μ «

Ο Ο t Ο

•^•ΟΓΛΟ

—(ΝΟΌ

Γ- Ο so Ο

Ο fN fN •—'

c> (Ν - ο — Γ- (Ν

m m —
buk/backen>backte:

Grammatical variation

83

appear in the weak inflection, while those referring to non-concrete, abstract actions tended to correlate with the strong inflectional variants. Although this tendency appears throughout the corpus, in many cases its influence is overshadowed by the interplay of various other factors. In the following examples, however, the effect of a figurative usage of the verbs clearly comes into play. Moreover, in some instances, i.e., bewegen, there is a perception that both variant forms are in fact separate lexical items: (18)

Sie bewog28lbewegte4

ihn, das Haus zu verkaufen.

'She persuaded him to sell the house.' (19)

Er hat sich diese Behauptung saugt 6.

aus den Fingern

gesogen2jlge-

'He pulled this claim right out of thin air.' (20)

Wie eine Spinne wob22^webte3 sie ihr Netz von Intrigen.

'Like a spider, she spun her net of intrigue.' 4.4.6 Sty lis tics

The last case of distinction involves examples which show that, in their selection of variant forms, speakers make at times various kinds of stylistic judgments: (1) In some instances, informants objected to the usage of the past tense form berstete because of its reduplicated dental. To emphasize its awkwardness, one informant volunteered the example Die Tür berstete tatsächlich 'The door actually burst'. (2) In the test f r a m e Der Luftdruck

soglsaugte

uns immer mehr in die

Tiefe 'The air pressure pulled us deeper and deeper into the depths', two informants opted for the strong variant because of its association with der Sog, and in Plötzlich hat die Haustürklingel

geschollenlgeschallt

'Sud-

denly the door bell rang', one informant expressed his preference for geschallt because of its proximity to geschellKschellen. In the speakers' minds, the chosen forms rendered the meaning more aptly and came closer to what they considered the intended reading of each sentence. For the same reason, namely appropriateness of expression and proximity to lexical items within the respective semantic field, several informants expressed a reservation about selecting the strong alternants in

84

Neida Μ. Cook

Sein Ruf ist weit und breit erschollen/erschallt 'His reputation has resounded far and wide' and Er ist erblichen!erbleicht 'He grew pale' because of their association to verschollen and verblichen.

5. Conclusion In sum, when asked to judge the acceptability of sentence frames containing verbs with variable inflections, native speakers of German are found to be sensitive to changes in linguistic environment; accordingly, they employ various strategies in selecting the alternating verb forms. Their selection is not random. It is governed by several principles, linguistic as well as nonlinguistic, and takes place in response to some stylistic or functional shift that is an inherent part of the process of analogical levelling. This process is gradual and it presupposes intermediate stages whose ever-changing conditions encourage or discourage the incidence of the alternating forms. The various correlations which come to light under such conditions, although recurring with some regularity, are not systematic in the sense that they adhere to the respective patterns precisely. Such correlations are for the most part merely a by-product of changes affecting the language. As such, they are just another piece of evidence of language change in progress. References Behaghel, Otto 1924

Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 2: Die Wortklassen und Wortformen. Heidelberg: Winter. Curme, George O. 1960 A grammar of the German language. (2nd revised edition) New York: Ungar. Engel, Ulrich 1988 Deutsche Grammatik. Heidelberg: Groos. Götze, Alfred (ed.) 1939 Trübners Deutsches Wörterbuch. Vols. 1-8. Berlin: Walter de Gruyter. Haber, Lyn R. 1975 "Leaped and leapt: A theoretical account of linguistic variation", Papers from the Eleventh Regional Meetings — Chicago Linguistic Society, April 18-20: 211-238. Heidolph, Karl Erich-Walter Flämig-Wolfgang Mötsch (eds.) 1980 Grundzüge einer deutschen Grammatik. (2nd edition) Berlin: Akademie Verlag. Jung, Walter 1971 Grammatik der deutschen Sprache. (4th edition) Leipzig: Bibliographisches Institut.

backen>buklbacken>backte:

Grammatical

variation

85

Klappenbach, R u t h - W o l f g a n g Steinitz (eds.) 1980 Wörterbuch der deutschen Gegenwartssprache. (10th edition) Berlin: Akademie Verlag. Kluge, Friedrich 1967 Etymologisches Wörterbuch der deutschen Sprache. Berlin: Walter de Gruyter. Lederer, Herbert 1969 Reference grammar of the German language. New York: Charles Scribner's Sons. Manczak, Witold 1979 "Frequenz und Sprachwandel". In Helmut Lüdtke (ed.) Kommunikationstheoretische Grundlagen des Sprachwandels, 3 7 - 7 9 . Meier, Helmut 1964 Deutsche Sprachstatistik I and II. Hildesheim: Olms. Morgan, Bavard Quincy 1928 German frequency word order (based on Kaeding's Häuf igkeits Wörterbuch der deutschen Sprache). New York: Macmillan. Paul, Hermann 1960 Deutsches Wörterbuch. (7th edition, by Alfred Schirmer). Halle: Niemeyer. Quirk, Randolph 1970 "Aspect and variant inflection in English verbs", Language 46: 300-311. Rauch, Irmengard, et al. 1988 "Is there an aspect distinction in certain German strong/weak verb alternation? Evidence from German in the San Francisco Bay Area", in Semper Idem et Novus. A Festschrift for Frank Banta. Kümmerle: Göppingen. 433-443. Rosengren, Inger 1972 Ein Frequenzwörterbuch der deutschen Zeitungssprache. Vols. 1-2. Lund: C. W. K. Gleerup. Scherer, George A. C. 1965 Word frequency in the Modern German short story. Boulder: University of Colorado Press. Theobald, Elke 1992 Sprachwandel bei deutschen Verben. Flexionsklassenschwankungen starker und schwacher Verben. Tübingen: Gunter Narr. Wahrig, Gerhard 1977 Wahrig Deutsches Wörterbuch. Berlin: Gütersloh.

To the rescue of time in German tense Richard d'Alquen

My first inkling that time needed rescuing arose when I was putting some notes together for a new course on "Aspects of German Grammar". It seemed entirely natural to begin by consulting Duden. On tense I found the following remarkable opening statement. Wenn wir beim Verb von Zeitformen (Präsens = Gegenwart usw.) sprechen, dann darf man damit keine Vorstellungen von objektiven Zeitverhältnissen verbinden, weil der Tempusbereich unserer Sprache nach eigenen Sehweisen gegliedert ist. (Duden [1973: § 164]) 'When talking about tense forms of the verb (present = present time etc.), one must not associate them with objective time relationships, because the area of tense in our language is organized according to its own way of seeing things.'

Such a sentence must instill a sense of despair in the hearts of those who, trustful in the Duden tradition of bringing the light of higher learning to the general public, find a warning against objectivity and an almost mystical hint at ineffable, language-specific ways of seeing things. The description of tense in that section is a listing, which, while containing many insights and reflecting some recent research on tense, does not really succeed in providing a coherent framework for the tense system as a whole. The traditional tense forms are subdivided according to usage and details of substitution with other tense forms (e.g., 'Present;', 'Present^), but without an underlying principle, the presentation is difficult to follow. The danger of confusion is such that (unless I too am its victim) there are two errors in the summary. 1 Is the German tense system really so unique? Why is there such distrust of the concept of time? Part of the answer lies in the evidence of sentences like the examples below, which show that time reference by tenses is not a simple one-to-one relationship with the time range of the same name. (1) a. b.

Future tense Es wird wohl jetzt erhältlich sein 'It is probably obtainable now' Ein guter Vater wird alle seine Kinder lieben Ά good father will love all his children'

88

Richard

c. (2) a. b. c. d. e. f. (3) a. b. c.

d. e.

(4) a. b. (5) a.

d'Alquen

Wirst du jetzt ruhig sein! 'Will you be quiet now!' [intensified command] Present tense Bist du jetzt ruhig! 'Be quiet now!' [intensified command] Kinder nehmen täglich eine halbe Tablette 'Children take half a tablet daily' Käthe holt uns dann mit dem Auto ab 'Then Cathy will pick us up with the car' Fallender Baum tötet Autofahrer Talling tree kills car driver' Im 9. Jahrhundert greifen die Wikinger wiederholt an 'In the ninth century the Vikings attack repeatedly' Rotkäppchen nimmt den Korb und macht sich auf den Weg 'Little Red Riding Hood takes the basket and sets off Preterite tense Wir übernachteten in einem billigen Hotel, das ich von einer früheren Reise her schon kannte We spent the night in a cheap hotel that I already knew from an earlier journey' Kellner unsicher: Sie bekamen doch den Obstsalat 'Waiter in doubt: It was you, wasn't it, who was getting the fruit salad?' Gab's nicht den Faust morgen im Theater? 'Wasn't Faust being shown at the theatre tomorrow?' Fritz dachte nach: Wenn morgen Weihnachten war, hatte er sehr wenig Zeit. 'Fred reflected: if tomorrow was Christmas, he had very little time' Perfect tense Bis drei Uhr habe ich das Auto repariert 'By three o'clock I shall have repaired the car' Ein Unglück ist schnell geschehen 'An accident happens quickly' [German uses perfect.] Future perfect tense Er wird wohl gescherzt haben Ί expect he was joking'

To the rescue of time in German tense

(6) a. b.

89

Pluperfect tense Nach langjähriger Arbeit hatten sie das Problem gelöst 'After years of work they had solved the problem' So, eine Absage! Und dabei hatte sie mir fest versprochen zu kommen 'So, she regrets. And this time she had truly promised to come'

Although it has long been recognized that the so-called future, present and past tenses do not refer exclusively and respectively to the time ranges of the same name, it was not till well into the twentieth century that time became threatened as the chief factor in tense selection. With regard to German tense, Heinrich Hempel is credited with the first attempt to put modalities at the centre of attention, rivalling temporality. In "Über Bedeutung und Ausdruckswert der deutschen Vergangenheitstempora" 'About meaning and expressive value of the German past tenses', Hempel (1932) finds in the German preterite an "objektive und imaginative Hingabe ans Vergangene ... und so kann das Präteritum gerade distanzierend wirken" 'an objective and imaginative involvement with what is past ... and thus the preterite has the effect of making things quite remote'. Hempel says of the perfect that it conveys a "gefühlsmäßige, wertende, urteilende Stellungnahme zum Vergangenen" 'an emotional, evaluative, judgemental attitude to what is past'. After the war, among other attempts to rethink German grammar, one could pick out Käthe Hamburger's Logik der Dichtung (1975) as a milestone, because it was so well received. A major finding was that the epic preterite did not refer to past time at all, but merely marked off the successive "now-points" of the narrative. Although she was clearly writing about literary narration, the book was a boost to the movement that was moving towards eliminating time from tense analysis in general. The ultimate timeless tense grammar was achieved in 1964 in Harald Weinrich's Tempus: Besprochene und erzählte Welt. Here was another case of a work that, judging by the torrent of mainly positive commentary, clearly resonated with the linguistic Zeitgeist. For German, Weinrich claimed that tense choice depended on a primary dichotomy between two speaker attitudes characterized by the following polarities: 1

Dialogue {besprochen) excited (gespannt) subjective

2

Narrative (erzählt) calm (ungespannt) objective

90

Richard

d'Alquen

Tenses are divided into: Vorschau Null-Tempus Rückschau

Group 1 WIRD SINGEN SINGT HAT GESUNGEN

Group 2 W Ü R D E SINGEN SANG HATTE GESUNGEN

Weinrich (1964: 72) claims: "Das Präsens gibt überhaupt keine zeitliche Orientierung ... Das Präteritum bezeichnet insbesondere keinerlei Vergangenheit, sondern nur Erzählung, nichts weiter." 'The present tense gives absolutely no orientation in time ... The preterite in particular indicates no kind of past, but only narration, nothing more.' Faced with the probable counterclaim that Vorschau and Rückschau are just renamings o f ' f u t u r e ' and 'past', Weinrich (1964: 74) states that in the Vorschau mode "etwas fordert als Verheißung oder Drohung unser Sorgen und Besorgen heraus" 'something by way of promise or threat elicits our anxiety and caring', and that in the Rückschau we are thinking of something that is "verpflichtend in Bindungen und Vorentscheidungen" 'something that commits us through obligations and precedents'. Information about time is left to adverbs and inferential deduction. The decade clearly belongs to Weinrich, for ten years later in 1974, Hermann Gelhaus (1974: 122) can state: "Tempora drücken keine Zeit aus und sind demnach nicht imstande Zeitverhältnisse zu bezeichnen" 'Tenses do not express time and are therefore unable to denote time relationships'. In the intervening years, the topic has not rested, and while there have been discussions that did not avoid the concept of time, it seems to me still necessary to deal with the challenge of the atemporalists, to meet their objections head-on and refine rather than discard the traditional analysis, in which time is central. Let us begin with the present and future tenses. Both of these can refer to the entire non-past time range and thus contrast temporally with the preterite. (7)

Am Freitag wird sie da sein 'On Friday she will be there'

future time

(8)

Am Freitag ist sie da O n Friday she is [= will be] there'

future time

(9)

Am Freitag war sie da O n Friday she was there'

past time

To the rescue of time in German tense

91

Freitag in (7) and (8) lies in future time, in (9) it lies in the past. The only changes made in these sentences are in the verb, which then must be responsible for the difference in meaning between (9) and the other two sentences. Now we add an adverb referring to the moment of speech: (10)

Jetzt wird sie da sein 'She'll be there now'

present time

(11)

Jetzt ist sie da 'She is there now'

present time

(12)

* Jetzt war sie da *'She was there now'

Jetzt indicating present time, co-occurs logically with the present and future tenses, but in (12) creates an illogicality with war, because war indicates past time. If this were not part of a conversation, but rather a narrated thought (erlebte Rede), it would be acceptable, as in (13). (13)

Hans dachte nach: Jetzt war sie da 'John reflected: she was there now'

However, this war is a tense shift from an underlying *ist and follows a "sequence of tenses" rule. Compare: (14)

Hans denkt nach: Jetzt ist sie da 'John reflects: she is there now'

A similar shift takes place in English. The question whether present and future tenses are semantically tied to present and future time respectively must be answered in the negative. Yet I shall argue that a limited separation should be recognized. Reading (15) and (16), (15)

Wir werden in Köln wohnen 'We shall live in Cologne'

Wir wohnen in Köln 'We live in Cologne'

(16)

Er wird bestimmt da sein 'He will certainly be there'

Er ist bestimmt da 'He is certainly there'

we feel a preference - but no more than a preference — for future interpretation of the future tense and present interpretation of the present tense. The larger context in which the sentences are spoken ultimately determines present or future interpretation. However, if we introduce a

92

Richard

d'Alquen

modal component of meaning [probability], interpretation of the future tense swings towards present time. (17)

Wenn ich mich nicht irre, wird er wohl da sein

'If I'm not mistaken, he'll be there' [i.e., 'is there'] Now, this is a well known use of the future tense to express probability in the present. As such, it is a special, non-basic function of the future tense requiring a modal component [probability], which is normally used for time reference to the present. This leads to the tentative positing of primary and secondary time ranges within the non-past limits. We divide non-past time into present and future, stipulating that for the future tense, future time is primary and present time is secondary. Conversely, for the present tense, present time is primary and future time secondary. primary time reference

secondary time reference

future tense

F U T U R E TIME

PRESENT TIME [PROBABILITY]

present tense

PRESENT TIME

F U T U R E TIME [MODALITY?]

If we search for a reason why the future tense in present time reference should be accompanied by [probability], we cannot fail to hit upon the idea that future time, the primary reference of the future tense, is inherently attended by uncertainty, and this characteristic, in the form of a modal component [probability], is brought to bear on the present. In this way, the primary time range leaves its modal imprint on the secondary. This in turn raises the question whether the present tense, when referring to its secondary time range, the future, does not likewise bring with it a modal component based on an inherent quality of its primary reference, present time. One can make a good case for just that. (18)

Morgen unterrichtet

Inge Deutsch

'Tomorrow Inge teaches German' (19)

Nächste

Woche fahren wir wieder nach Hause

'Next week we go home' Both of these sentences seem to refer to an accepted fact rather than a matter of personal conviction. The naming of a future date evokes the expectation of a plan or fixed arrangement, and thus the future event is

To the rescue of time in German tense

93

guaranteed by something more than the assurance of the speaker. This could well be the certainty that is inherent in present observation, a presently existing plan or inevitable recurrence. Such inherent certainty would then be transferred to the secondary time range of the present tense, namely the future, in the form of a modal component. In view of the fact that this certainty is based on something other than the speaker, we could name the component [objective assurance] and expand the table as below. primary time reference

secondary time reference

future tense

FUTURE TIME [MODALITY?]

PRESENT TIME [PROBABILITY]

present tense

PRESENT TIME [MODALITY?]

F U T U R E TIME [OBJECTIVE ASSURANCE]

Duden (1973: § 187) says of the construction werden + infinitive referring to future time that it has several modal components: [ Voraussage] ['prediction'], [Ankündigung] ['announcement'], [ Versicherung] ['assurance'], [Entschluß] ['decision'] and even [Befehl] ['command']. For another common use of this construction, Duden (1973: § 188) records the component [Vermutung] ['presumption']. In the first and second persons particularly, Duden (1984: §231) holds that [Entschluß] ['decision'] or [feste Absicht] ['firm intention'] is discernable. I would prefer to set aside [ Voraussage] and [Ankündigung] as circumlocutions for future time, but the remaining components have in common a personal commitment or emotional involvement that we might express in a single component [personal assurance], thus balancing [objective assurance]. When speakers refer to future time, they have to choose between these balanced components. This may seem at first acquaintaince a rather unlikely restriction, but when one reflects that all sincere predictions must have some kind of justification for the implied belief, one can accept that the justifications may be divided into one group that emphasizes extra-personal, objective criteria, and the rest, which then perforce are dependent on personal conviction, determination, faith and similar mental states. Let us consider, for example: (20)

*Nächsten Samstag gewinnt unsere Mannschaft *'Next Saturday our team wins'

The sentence implies that the speaker knows a reason why the team cannot avoid winning. Apart from freakish circumstances such as bribing the

94

Richard

d'Alquen

referee or drugging the other side, the sentence is not correct. Similarly in English (21)

"Oilers beat Kings next Saturday

is not a well-formed headline, because the present tense in future context bears the component [objective assurance] and no such modality is possible here, where the outcome cannot be part of a generally accepted plan or a predictable recurrence. Both sentences need [personal assurance] and the future tense. (22)

Nächsten Samstag wird unsere Mannschaft gewinnen

(23)

Oilers will beat Kings next Saturday

If [personal assurance] and [objective assurance] are balanced opposites in a polarity, then perhaps [probability] is at one end of another polarity. The opposite component would be in the nature of non-speculative fact. Let us call it [factuality] and thus complete the table. primary time reference

secondary time reference

future tense

F U T U R E TIME [PERSONAL ASSURANCE]

PRESENT TIME [PROBABILITY]

present tense

PRESENT TIME [FACTUALITY]

F U T U R E TIME [OBJECTIVE ASSURANCE]

Using this double coverage of non-past time with balanced and mutually exclusive modal components, we can now explain uniformly something that was explained variously and not really convincingly in Duden with regard to the following three sentences.2 (24)

Viele Fabriken verseuchen die Flüsse, und sie werden es weiterhin tun, wenn es der Gesetzgeber nicht unterbindet 'Many factories are polluting the rivers, and they will continue to do so, unless the legislator puts a stop to it'

(25)

Wir werden ja sehen, wie es weitergeht 'We shall see how things develop'

(26)

Einige haben bereits seine Abdankung gefordert, und solche Stimmen werden sich jetzt mehren 'Some have already demanded his resignation, and such voices will now increase in number'

To the rescue of time in German tense

95

In the above sentences, Duden notes that the future tense cannot be replaced by a present tense. Thus, future time demands the future tense and excludes the present tense. In other words, the claim of the atemporalists that present and future tenses never refer exclusively to present and future time respectively would seem to be disproved by these examples. Yet that is not the conclusion I would draw. In (24) for example, to say in the present tense und sie tun es weiterhin would be a factual statement like the first clause, but the context requires a shift to an unguaranteeable prediction about the future. There is no set plan to continue polluting rivers; it is not an inevitable recurrence. Rather, the speaker is trying to persuade an audience of the validity of his own much healthier intentions for the future. Hence, there is no case for [objective assurance], but a very strong one for [personal assurance]. Neither present time with [factuality] nor future time with [objective assurance] is appropriate in the context. Therefore the present tense is not usable here. What is required is future time reference with [personal assurance], which, in turn, requires the future tense. Similar arguments can be made for (25) and (26). This supports the notion of primary and secondary time reference. Since there are occasions when the future is necessarily in doubt, there will be occasions when [objective assurance] cannot be justified and the present tense will be excluded in that future context. Conversely, since there are occasions when the present situation is not in doubt, there will be occasions when [probability] cannot be justified and the future tense will be excluded. This implies on the one hand what we have seen in (2426), where future time demands future tense, and, on the other, what is demonstrated in (27) and (28), namely that present time contexts sometimes demand exclusively present tense: (27)

Ich bin dreißig Jahre alt Ί am thirty years old'

(28)

Er sieht gesund aus 'He looks healthy'

A change to future tense in (27) or (28), without a change to future time reference, would bring with it the component [probability], which is nonsensical in the given sentences. Thus, there are cases of present time reference where only the primary tense, the present tense, is possible. However, the present tense in its secondary time reference (future time) can always be replaced by the future tense. The reason for this is that for

96

Richard

d'Alquen

future time, the speaker can always add [personal assurance], even if the event is part of a set plan and needs no assurance from the speaker. (29)

Der Autobus kommt um zwei or 'The bus comes at two'

or

Der Autobus wird um zwei kommen 'The bus will come at two'

Conversely, a future tense referring with [probability] to present time, its secondary reference, can always be changed to a present tense (primary) plus an adverb expressing [probability] (30)

Jetzt wird sie zuhause sein

or

'She'll be at home now'

or

Jetzt ist sie vermutlich zuhause 'She is presumably at home now'

The positing of a "primary" time reference of each of our two tenses is justified by the exceptionless applicability of those tenses to their primary time ranges; "secondary" status is justified by the not infrequent inapplicability of the same tenses to their secondary time ranges. The use of the words "present" and "future" in the tense names is therefore not without demonstrable support, but the ultimate control by modality in the choice between these two tenses is something that should be stressed in a refinement of the traditional system. [non-past] time {objective} [objective assurance] (31) Morgen arbeiten sie

[factuality]

{subjective} [personal assurance]

[probability]

(32) (33) Jetzt arbeiten sie Morgen werden sie arbeiten

(34) Jetzt werden sie (wohl) arbeiten

Translation: (31)

'They are working tomorrow'

(32)

'They are working now'

(33)

'They will work tomorrow'

(34)

'They'll be working now'

To the rescue of time in German tense

97

The highest node (outside above tree) dominates [past] and [non-past], a choice based on time. The division of non-past time is modal. The two modal components in wavy brackets are of a higher order than the others, which take on the nature of "allomodalities", selected according to contextual or inferential indications of present or future time. We may agree with the critics of the traditional system that reference to future time does not necessarily result in the choice of a future tense, nor present time in a present tense. Two salient points emerge as important for any teaching grammar. One is that the fundamental tense distinction is past versus non-past, and the second is that this temporal distinction is integrated with two pairs of modal components. Another area of difficulty in teaching German to English speakers is the distinction between the preterite and the perfect. In the scheme set up by Weinrich, the distinction between these two tenses is one of speaker attitude (Sprechereinstellung). As indicated in the earlier diagram, the perfect belongs to the means of conveying besprochen 'dialogue attitude', and the preterite belongs to the means of conveying erzählt 'narrative attitude', but the statistics presented by Hauser-Suida and Hoppe-Beugel (1973: 79) do not bear this out.

In dialogue In non-dialogue

besprochen perfect 47,23% 10,16%

erzählt preterite 50,10% 80,77%

pluperfect 2,67% 9,07%

Dialogue is ideal for the besprochen speaker attitude, yet the preterite, which belongs under erzählt, actually has a higher frequency in dialogue (50,10%) than the perfect (47,23%). While it is true that in non-dialogue the proportion of perfects drops to 10,16%, and that of the preterite rises to over 80%, the figures could never justify the setting up of a grammatical rule. It is, of course, a well known fact that, in general, spoken German prefers the perfect and that narrative prefers the preterite. It turns out to be possible to capture this generality but yet take account of the contrary examples, and we do this by reflecting the history of these two tenses in the choice of their modal components. The loss of the preterite in Upper German has been well investigated (Waterman 1956: 107) and the majority opinion is that it was caused by the loss of unstressed endings in the weak verb. When sagte became sagt, the preterite of all such verbs became homophonous with the present and consequently went out of use. The loss of the preterite indicative in the

98

Richard d'Alquen

regular weak verbs led to its replacement by the perfect indicative, causing an irreversible structural change, which spread by analogy to other verb classes. The only surviving preterite is war. Now, this restructuring occurred in the spoken language in Upper German and southern parts of Middle German, while northern Middle and all Low German speakers continued to pronounce final unstressed vowels and saw no reason to give up the simple past tense. The emerging standard language was a written, conservative medium, based on slowly converging chancery norms sensitive to precedent and to the model of Latin, which possesses a (different) contrast between imperfect and perfect. The result was a learned, written style that retained the form of the preterite serving a language community of which half had no dialect-derived understanding of this verb form. The semantic confusion of the two tenses must have spread to the north through the written language, which became the basis of educated speech. The modern standard has inherited a very complex and even fluid set of guidelines, that represent a kind of compromise between northern sensitivity to the preterite-perfect contrast and southern usage that recognizes only a stylistic difference between two past tenses. The diagram below is a first step towards clarification of the modern situation through consideration of the historical process. South German

North German

MHG

preterite

perfect

preterite

perfect

ENHG

perfect

perfect

preterite

perfect

NHG

perfect varies with preterite

obligatory perfect

* M H G = Middle High German, ( E ) N H G = (Early) New High German North and south agree on the use of the perfect in those applications where north German has always used the perfect. This gives rise in Standard German to the so called "obligatory perfect". However, where the later south German perfect represents an earlier preterite, there the con-

To the rescue of time in German tense

99

flict with north German produced an area of uncertainty or stylistic variation in the modern standard. What is an obligatory perfect? An obligatory perfect cannot be replaced by the preterite. (35)

Da wir jetzt gegessen haben, müssen wir zahlen 'Since we have now eaten, we must pay'

(36)

Nachdem wir gegessen haben, müssen wir zahlen 'After we have eaten, we must pay'

The type of action or event for which the perfect is used can be illustrated in a time-line diagram and contrasted with the preterite type of event.

Moment of speech

ance

(c)

(a)

(b)

Moment of speech

effect

• > » iTime - > » Obligatory Perfect

Preterite or Upper German Perfect

The obligatory or true perfect is the resultative type. The action — essen in (35) and (36) — is seen as completed but having a result or effect on a subsequent point in time that we shall call the "point of relevance". The point of relevance is either the moment of speech - jetzt in (35) — or a future point in time - nachdem in (36). The event may be as in (a) contiguous with, or as in (b) prior to the point of relevance. The tense time is that of the point of relevance, not the event. On the other hand, the preterite or the "Upper German" perfect does not imply any impact of the event on the present or a future time point. Any such message would have to be conveyed by situation or verbal context. Events of this type are represented in (c). History explains why the German obligatory perfect is obligatory and why the variation between perfect and preterite occurs where it does. It further explains why dialogue is conducive to the use of the perfect: it was in spoken German that the perfect replaced the preterite. Conversely, it tells us why the preterite sounds more learned and is used more com-

100

Richard

d'Alquen

monly in writing: the preterite was preserved in the written language even in the south. The Weinrich contrast in speaker attitude between besprochen and erzählt is partly the result of the historical development of these two tenses in conflict as conveyers of the meaning [past]. Significantly, the historical perspective also explains all the types of irregular or unexpected uses of perfect and preterite quoted in the literature, which fall into groups as follows. First, a group of preterites that have a different lexical meaning as perfects (Latzel 1977: 77). (37)

a. b. c. d. e. f. g. h. i.

Plötzlich gähnte vor uns ein Abgrund 'Suddenly a precipice yawned before us' Der Blick der Stube ging auf die gestuften Gärten 'The living room looked out over the terraced gardens' Das Geschoß heulte durch die Luft The shell howled through the air' Der Diamant spielte in vielen Farben 'The diamond flashed [literally 'played'] in many colors' "Ich weiss noch mehr, "fuhr der Alte fort "I know more yet," the old man continued' Der Weg lief an der Grenze entlang 'The track ran along the frontier' Pfingsten fiel auf den 18. Mai 'Whitsun fell on the 18th of May' Die Goten sassen an der Weichsel 'The Goths were settled along the Vistula' Die Schuld traf den Fahrer 'The driver was held responsible' [Literally: 'The guilt hit the driver']

When these figurative preterites are converted to the perfect, the meaning becomes literal. This strongly suggests that the figurative meaning was developed in literary style for narrative purposes, where the perfect was not called for. By contrast, in everyday speech, where the perfect dominated, the literal meaning was the norm. The result is that when we hear the perfect of gähnen 'yawn', we react to its literal meaning. Next, a group of perfect examples, culled from secondary sources, lacking any possibility of being converted to the preterite (Latzel [1977: 79] and Hauser-Suide-Hoppe-Beugel [1973: 7]). (38)

a.

Es hat sich ausgeschätzelt 'It's all over with that darling stuff' [literally 'It has darlinged itself out']

To the rescue of time in German tense

b. c.

d.

e.

f.

101

Er hat's endlich gefressen 'He's got the message [literally 'eaten it'] at last' Sie hat die Weisheit mit dem Löffel gefressen 'She thinks she's so smart' [literally 'She's eaten wisdom with a spoon'] Du bist nicht auf den Mund gefallen 'You're not exactly speechless [literally 'You didn't fall on your mouth'] Der Mantel hat ausgedient 'The coat is worn out' [literally 'has completed its period of service'] Der Sturm hat ausgetobt 'The storm has blown itself out' [literally 'raged out']

These are clear cases of an obligatory perfect in the above sense: the action has been terminated, and there is an effect on the present. In addition, the style is very informal dialogue, which is another inhibitor of the preterite. The examples under (38) with aus as a prefix in the sense of [completion] {hat ausgedient, etc.) are similarly unable to form preterites because [completion] of the event and the consequent enduring state are emphasized. Next is a group known as "aesthetic preterites". They are situationally suited to obligatory perfects, but the preterite seems to raise the stylistic level (Duden 1973: § 172). A radio announcer might well say: (39)

Sie hörten eben die Fünfte Symphonie von Beethoven 'You just heard Beethoven's Fifth Symphony'

The high cultural tone of symphonies is more in keeping with the learned written tradition of the preterite than the unlettered dialect associations of the perfect. Fourth, a group of verbs that, regardless of the type of event, statistically show strong preference for the preterite: sein, haben, modals, passives, verba dicendi. In (40) to (43), Bisher indicates the point of relevance. (40)

a. b. c.

Bisher war nur einer da 'So far only one has been/come' Bisher hatte ich keine Gelegenheit, dorthin zu kommen 'Up to now I haven't had the opportunity to go there' Bisher konnte die Polizei den Täter nicht verhaften 'So far the police haven't been able to arrest the culprit'

102

Richard

d.

d'Alquen

Bisher wurde auf dem Gebiet nichts geleistet ' U p to now nothing has been done in this area'

Latzel (1974: 195) thought the reason for avoidance of the perfect might lie in the low information content of these verbs. It is, of course, undeniable that if there is nothing going on, there is nothing to terminate and there can be no effect on the present, so that the true perfect does not seem applicable. On the other hand, the translations into English do show the present relevance principle through the use of the perfect. The case of war (40) may be special in that it was the only surviving preterite in the south. Since this left it grammatically isolated and unable to maintain a systematic contrast with the perfect, it became simply an alternate form for ist gewesen. Verba dicendi, as mere markers of the source or the utterance, could also be regarded as low in informational meaning, but the statistical preference for the preterite is in line with the frequency of verbs of saying in reporting speech in narrative, where the preterite is the characteristic tense. (41)

"Nein," sagte der Oberförster, "ich muß meine Pflicht tun." ' "No," said the Chief Forest Warden, "I must do my duty."'

Nevertheless, a sentence could be devised in which a completed event of saying is of "present relevance". (42)

Bisher sagten Sie doch, das Gewehr habe auf dem Tisch gelegen or Bisher haben Sie doch gesagt, ... ' U p to now you have been saying the rifle lay on the table'

The decision is no longer so clear-cut, but German would tend to be less sensitive to "present relevance" than English and thus more likely to use the preterite. Some verb phrases with zu also avoid the perfect (Latzel 1977: 76, 170). (43)

a. b. c. d.

... daß ihr die Kräfte zu versagen drohten!schienen '... that her strength threatened/seemed to give out' Das Wetter versprach, schön zu werden 'The weather promised to become fine' Kant pflegte täglich spazierenzugehen 'Kant was accustomed to take a walk daily' ... die wir zu umreißen suchten '... which we tried to outline'

To the rescue of time in German tense

103

One could argue for the first example that drohten or schienen are semantically unsuitable to combine with perfect meaning, for English also rejects the perfect. After all, the enduring effect of a completed event of "seeming" is difficult to imagine. Similarly with the other examples, the meaning of the finite verb is not compatible with perfect meaning as defined by "completion" and "present relevance". This incompatibility excludes the obligatory perfect and leaves the choice open to stylistic considerations whether to use the "Upper German" perfect or a preterite. There seems to be nothing to gain from choosing the longer form. However, a za-verb phrase might be devised suitable in meaning for an obligatory perfect. (44)

Da Sie soeben gewagt haben, vor mir zu erscheinen, ... 'Since you have just dared to appear before me, ...

Here the "present relevance" is strong and the daring can be seen as completed, which justifies a perfect. Thus, although added complexity may well be a deterrent to using the perfect in a zw-verb phrase, the major factor is probably the presence or absence of a situation that calls for an obligatory perfect. Examples such as these appear at first glance to be unsystematic and to throw doubt on the traditional tense analysis, but on closer examination they are explicable as obligatory perfects opposed to preterites, or as "Upper German" perfects varying with preterites, the variation being the result over time of a conflict of northern and southern tense systems. There is no need to explain everything by speaker attitude, although this is clearly a large part of the motivation for choosing between an "Upper German" perfect and a preterite. Now we should be able to exploit the historical perspective further to build an explanatory model of the use of the perfect and preterite tenses. First, we must distinguish between two perfect tenses: the obligatory perfect and what we shall continue to call the Upper German perfect. The obligatory perfect is part of a subsystem of true perfects. The basic contrast is between past and non-past, where the time is that of the point of relevance, not the time of the event. (45)

a.

Bis 1994 wird man das Haus gebaut haben 'By 1994 they will have built the house'

[non-past: future]

104

Richard d'Alquen

b.

Bis 1994 hat man das Haus gebaut 'By 1994 they will have built the house'

c.

Bis 1994 hatte man das Haus gebaut 'By 1994 they had built the house'

[non-past: future] [Same as above but using present perfect] [past]

This set of perfects is parallel to the "absolute" set, in which the time reference is that of the event, not a point of relevance. (46)

a. b.

c.

1994 wird man das Haus bauen 'In 1994 they will build the house' 1994 baut man das Haus 'In 1994 they will build the house' 1990 baute man das Haus 'In 1990 they built the house'

[non-past: future] [non-past: future] [Same as above but using present [past]

This would be a classically simple system, had it not been for the original sin of the Bavarians in dropping schwa from unstressed syllables, for this led to the use of the perfect in the function of the preterite and in effect created for the Pan-German standard a new tense identical in form with the obligatory perfect, so that one can now vary (46 c) with an Upper German perfect. (47)

1990 hat man das Haus gebaut 'In 1990 they built the house'

[past]

It is essential to recognize two homophonous perfect tenses, one of which is not a true perfect. One way to express this duality is to imagine the standard language as having a northern and a southern subsystem, with Standard German speakers switching from one to the other in the same way as between registers. The diagram fills in some details. The northern subsystem, occupying the upper part of the diagram, has the obligatory (or true) perfect with the inherent components, so called because they are original to the tense forms and do not vary with context. These are: (a) [non-past], referring to a point in time affected by the event (point of relevance); (b) [relevance], a modal component; and (c) [completion], an aspectual component. The northern subsystem also has the preterite with the inherent temporal component [past], referring to the event.

To the rescue of time in German tense

INHERENT COMPONENTS

PERFECT (OBLIGATORY) [non-past] [relevance] [completion]

northern subsystem

[past]

southern subsystem

UPPER G E R M A N PERFECT

standard subsystem DERIVED COMPONENTS

[dialogue] [colloquial] [topical] [subjective] [isolated]

105

PRETERITE

[past]

[written] [formal] [remote] [objective] [sequential]

The southern subsystem, occupying the middle part of the diagram, has only the Upper German perfect with the inherent component [past], referring to the event. I class this as inherent because, though not original to the tense form, it does not vary with context. The combined Standard German system consists of more than the lowest third of the diagram, for it includes, according to contextual requirement, either the northern or the southern subsystem. It incorporates — if contextually appropriate - the obligatory perfect, because there is no resistance from the southern subsystem, but when the speaker has to choose between a northern preterite and southern perfect, the conflict is resolved by the "derived" components, so called because they stem from the associations arising from the history of the tenses in the standard language. The modal components come in pairs of opposites. They do not necessarily make themselves felt with equal strength at every occurrence. UPPER GERMAN PERFECT [dialogue]

PRETERITE

HISTORY OF ORIGIN

[written]

[colloquial]

[formal]

Preterite lost in Upper German speech, preserved in writing, Preterite, being the written, learned form, was associated with correct and formal behavior.

106

Richard

d'Alquen

[topical]

[remote]

[subjective]

[objective]

[isolated]

[sequential]

Matters of immediate interest tend to be communicated orally, and the spoken Upper German past tense was/is the perfect. The spoken utterance is normally less reflective than the written statement. Preterite is natural tense of narrative, where events are sequential. In dialogue, events are typically related in isolation and in the Upper German system only in the perfect.

With this analysis of the preterite and the perfect, we are able to maintain the primacy of time as the chief criterion of choice by setting up the point of relevance as the time referred to by the obligatory perfect in contrast to the preterite, which refers to the time of the event. We are able to explain examples of the perfect varying with the preterite as "Upper German" perfects, that is, a general past tense with the form of the present perfect on the model of southern dialects. A primary temporal contrast [past] versus [non-past] is maintained, combined with a set of components either inherent in the old perfect/preterite contrast, or derived from the historical associations of these tenses after they came into competition with each other following the southern German preterite loss. Apparently irrational groups of obligatory preterites and obligatory perfects were explained as consistent with the theory, though unpredictable as individual examples. In the first part of this paper we looked at the present and future tenses. Here too it was found possible to set up a primary [non-past] / [past] contrast. This was supplemented by the incorporation of modal contrasts [probability] versus [factuality] and [personal assurance] versus [objective assurance] to explain the usage of the present and future tenses. If we combine the results of the first and second parts of this paper, we have a proposal for the groundwork of a rational, teachable system that is a compromise between older and newer views. While a primary temporal contrast is retained, it is not the traditional three-way division, but a simple cut into [non-past] versus [past]. It also incorporates many of the insights of more recent work on German tense by systematizing a workable number of non-temporal components, which, however vital, are not sufficient by themselves to supply all the components of a theoretical model of German tense: old-fashioned time needed to be rescued.

To the rescue of time in German tense

107

Notes 1. Duden Grammatik (1973: §§ 191, 111 [b]: for "Präteritumü" read "Perfekt i i i ( a ) " and in § 191 IV (b) for "2. F u t u r " read "2. Futur n ". The term "Perfekt i i i ( a ) " is an addition by the present author to cover the function mentioned in § 175. 2. Duden Grammatik (1973: § 170,5): the phrase "bei bestimmten Tempuskombinationen" 'with certain tense combinations' is used to explain (24), idiomaticity to explain (25), and temporal misunderstanding to explain (26). It is far from clear that these serve as explanations.

References Duden Grammatik der deutschen Gegenwartssprache 1973 (3rd edition by Paul Grebe.) M a n n h e i m - W i e n - Z ü r i c h : Bibliographisches Institut, Dudenverlag. Duden Grammatik der deutschen Gegenwartssprache 1984 (4th edition by Günther Dosdrowski.) M a n n h e i m - W i e n - Z ü r i c h : Bibliographisches Institut, Dudenverlag. Gelhaus, Hermann 1974 Studien zum Tempusgebrauch im Deutschen. Mannheim: Institut für deutsche Sprache. Hamburger, Käthe 1957 Die Logik der Dichtung. Stuttgart: Klett. Hempel, Heinrich 1932 "Über Bedeutung und Ausdruckswert der deutschen Vergangenheitstempora", Hermaea 31 (Festgabe, Philipp Strauch ..., edition by Georg Baesecke and Ferdinand Joseph Schneider): 1-29. Hauser-Suida, Ulrika-Gabriele Hopper-Beugel 1973 Die Vergangenheitstempora in der deutschen geschriebenen Sprache der Gegenwart. (Heutiges Deutsch 1,4). Munich: Hueber. Latzel, Sigbert 1974 "Perfekt und Präteritum in Ein-Satz-Äußerungen der geschriebenen deutschen Sprache", in: Hermann Gelhaus and Sigbert Latzel (eds.) Studien zum Tempusgebrauch in Deutschen. (Forschungsbericht des Instituts für deutsche Sprache 15.) Tübingen: Narr. 1977 Die Deutschen Tempora Perfekt and Präteritum (Heutiges Deutsch III: Linguistisch-didaktische Untersuchungen des Goethe-Instituts, 2.) Munich: Hueber. Waterman, John T. 1956 "The preterite and perfect tenses in German. Α study in functional determinants", The Germanic Review 31: 104-114. Weinrich, Harald 1964 Tempus: Besprochene und erzählte Welt. (Sprache und Literatur 16). Stuttgart: Kohlhammer.

On syntactic and pragmatic features of speech acts in Wulfstan's homilies Eugene

Green

1. From theory to method 1.1 Theoretical

preliminaries

Theories of speech acts generally identify them as an integral aspect of communicative exchange. Whatever utterance passes as a speech act conforms to the linguistic rules of a language, has a discernible reference to a possible world, and conveys an illocutionary force. The term "illocutionary force" concerns what a speaker intends in an exchange: the speaker conveys by means of an utterance an intention that he wants his listener to recognize (Recanati 1987: 211). Such an intention, however, need not be directly linked to an element, at any level, in the linguistic structure of the utterance. Nothing in a linguistic structure itself identifies a particular utterance as a speech act, let alone what illocutionary force the speaker intends. Instead, recognizing a speaker's intention stems in good part from a capacity to grasp how syntactic and pragmatic features influence one another. In short, pragmatics, as applicable to linguistics, centers on "the part of sentence meaning related to sentence use" (Recanati 1987: 19). When Wulfstan translates Matthew iii.2 as Dod dcedbote eowra synna 7 eow sona wyrd heofona duru rade untyned ' D o penance for your sins and soon enough the door to heaven will readily be opened for you', the promise does not arise merely from linguistic elements (Bethurum 1957: 206). Instead, the linguistic elements that shape the utterance, together with the uses they have (the pragmatic context), enable auditors to recognize Wulfstan's intention. That the utterance comprises a sequence of an imperative clause and a principal clause of future time gives it a force typically associated with conditional promises. Moreover, the pronominal eowra and eow help to stipulate what part the audience must have if the promise is to succeed. Yet even if elaborated with a fuller, more detailed account of the syntactic and pragmatic features in Wulfstan's scriptural utterance, the analysis would fall short.

110

Eugene Green

1.1.1

Pragmatic

contexts

For the pragmatic context in homilies such as Wulfstan's reaches beyond the sentence both to the status of the speaker and his relation to his auditors. As for the speaker's status, Wulfstan's authority derives from his office as bishop and from his legitimacy in quoting scripture, here the Old English translation of an utterance in Matthew. Without his authority, the promise he offers of a world elsewhere cannot bear serious consideration. Secondly, much of Wulfstan's authority to use scriptural promises in a credible manner is due to an institutional support that spells out his commitments and those of his audience. For in Wulfstan's homilies, as elsewhere in Catholicism, one aim of the preacher is to teach and recommend — hardly to demand — acts of confession and penance (Frantzen 1983: 154). What the Church expects is a free choice either to embrace its sacraments or to suffer penalty. And this contrast time and again invests Wulfstan's utterances with a pragmatic significance that counts as hortatory (urging compliance) and monitory (a cautionary warning). As Wulfstan makes plain, penance is an act that communicants choose to do: Ac do nu manna gehwylc swa him mycel pearf is, geswice yfeles 7 bete his misdceda pa hwile pe he mage 7 mote. God is swyde mildheort 7 wile swyde gemildsian 7 mycel forgyfan pam pe mid inwerdre heortan yfeles geswicad 7 geornlice betad. 'But let each person now do what is for him very needful, let him refrain from evil and make amends for his misdeeds while he is able. G o d is very compassionate and will be very merciful and greatly forgive those who sincerely in their hearts refrain from evil and eagerly make amends.' (Bethurum 1957: 163)

Plainly hortatory in themselves, these utterances nonetheless suggest in the quality of illocutionary force they support a range that includes monitory possibilities. For if the subjunctive mood of do ... manna gehwylc exemplifies an exhortation mildly but firmly expressed, the subsequent clause specifying a limit to time also implies the penalty of non-compliance. What appears evident, then, in Wulfstan's exhortation is that the force of speech acts often has a composite significance. If outright demand undermines the requisite that communicants freely choose to declare their faith and to act as the Church urges, then homilists need to shape their utterances accordingly. Such utterances preserve the sense that listeners recognize a speaker's intention, yet are free to invest what they hear with different meanings. The same utterance can carry antithetic significance, can admit of contrastive, pragmatic possibilities.

Speech acts in Wulfstan's homilies

111

1.1.2 Linguistics, pragmatics, and the idea of the composite This further aspect of the analysis, the possibility of antithetic significance, enriches the linkages between linguistic structures, pragmatic uses, and illocutionary force. If many speech acts in the homilies have a straightforward illocutionay force, the evidence also indicates frequent instances of utterances that one can legitimately construe in different ways. This substantial group, then, has composite implications. The term "composite" suggests that the linkages — linguistic, pragmatic, illocutionary — are sufficiently flexible to assure varied deontic meanings for the same utterance. To return to the scriptural promise translated from Matthew is to see it as just such an instance of a speech act with composite effects. Although Wulfstan's scriptural utterance appears a straightforward formation of a speech act from grammatical and pragmatic elements, its force can stand as a threat as well as a promise. For the imperative form of the first clause implies circumstances that, if not satisfied, actually make the utterance assume the status of a warning. As the basis of a threat the imperative clause as hypothetical in function entertrains a possible refusal of the audience to comply, to resist the penance urged. If listeners do not take the imperative clause directed at them by dod and eowra as hortatory, the promise lapses and the utterance constitutes a warning of some consequence. That such an alternative — promise or warning — emerges from the utterance depends on its institutional context as much as on the syntactic and pragmatic features of the utterance. Although studies of speech acts offer instances of an utterance expressing different illocutionary forces in varying contexts, this yoking of syntactic and pragmatic features for composite purposes still invites inquiry. To sum up, linguistic structures have no inherent property that identifies them immediately with particular speech acts. The scriptural promise, for example, emerges from a conjunction of clauses that need not always support the same illocutionary force. So the following utterance comprises, in a possible reading, two principal clauses, one in the imperative, one in the indicative, yet does not convey a sense of promise: Leofan men, understandad pcet Crist is cristenra heafod, 7 ealle cristene men syndon to Cristes limum getealde. 'Dear people, understand that Christ is the head of Christians, and all Christian people are counted as limbs of Christ.' (Bethurum 1957: 180)

Further, the pragmatic functions of Wulfstan's homilies include his status as bishop, the functions of forms in the immediate contexts of his utter-

112

Eugene Green

ances, and the institutional premises of the Church. Yet none of these, alone, stipulates what force a speech act is to have. Despite this richness of resources, Wulfstan's grammar and pragmatic devices reveal that his usage is sufficiently consistent to characterize his way of enunciating straightforward and composite speech acts.

1.2

Methodology

1.2.1 Selection of data The data for this study include the sentences as punctuated in the homilies of the Bethurum edition. For the most part, her practice is helpful in identifying specific syntactic patterns that underlie a single or composite speech act. Thus the opening of Homily XX begins with Wulfstan's urging his listeners to reflect: Leofan men, gecnawad facet sod is: peos world is on ofste, 7 hit nealced pam ende, 7 pi hit is on worlde a swa lengc swa wirse; 7 swa hit sceal nyde cer Anticristes tocyme yfelian swide. 'Dear men, recognize what the truth is: this world is speeding apace, and it is nearing the end, and, always, so much longer the world goes on, the worse it is; so that before the coming of Antichrist it shall necessarily become very much worse.' (Bethurum 1957: 261)

Although Wulfstan juxtaposes several assertions, grouped together after Bethurum's colon, the opening imperative, followed by a clause, expresses a governing, hortatory speech act. Now it is possible to say that the imperative mood applies to four separate speech acts, one associated with each assertive clause after the colon. Yet Wulfstan's use of swa after the semi-colon indicates that his arrangement of clauses supports a climactic sequence, ending in a resultative clause. This climactic pattern, then, stands as an appositive for sod and counts as part of an exhortation. For the most part the utterances marked by Bethurum as sentences stand as examples of a speech act, whether straightforward or composite. Two major variants occur as exemplified in the following: (1)

Ac utan warnian us georne 7 geearnian to Gode pcet he us gescylde swa his willa 'But let us take heed earnestly and labor for God that he defend us as he so desires.' (Bethurum 1957: 133)

Speech acts in Wulfs tan's homilies

(2)

113

Fordam mid pam odrum sceal celc cristen man hine to Gode gebiddan, 7 mid pam odrum geswutelian rihtne geleafan. 'Therefore with the first (the Our Father) each Christian person should pray to God and with the second (the Creed) make manifest true belief.' (Bethurum 1957: 157)

Each variant has either utan or a form of a modal auxiliary governing more than one principal verb, and so both sentences, as punctuated, count as having two speech acts. There are 130 such occurrences in the homilies. 1.2.2 Linguistic data The chief linguistic element in identifying the character of an utterance expressive of a speech act is the verb phrase. Verb phrases and their inflections generally account for the utterances that underlie speech acts: performatives for direct and reported speech acts, and many possible verb phrases for indirect speech acts. In detail, the homilies have 16 instances of direct speech acts and 26 that are reported. The performative verbs for direct and reported speech acts include the following: bebeodan 'command', behatan 'promise', beodan 'command', biddan 'exhort, admonish', cwedan 'declare', dihtan 'direct', forbeodan 'forbid', hatan 'command', and Iceran 'exhort, admonish'. As for indirect speech acts, the following patterns of verb occur: 85 inflected as imperatives; 7 of an inflected beon with infinitive, conveying a sense of obligation (Mitchell 1987, 1: 394); 150 marked with forms of the modal auxiliaries agan, *motan, *sculan, bearing a sense of obligation (Mitchell 1987, 1: 422—426); 55 in the present tense, indicative mood; 263 in the present tense, subjunctive mood; 111 that have the form utan 'let us'. There are also 47 instances of a periphrastic phrase as in is mycel pearf, is mycel neod, is nydpearf 'it is (very) necessary', as well as bid gebeorhlic 'it is fitting', gedafenad 'it befits', and riht is 'it is proper'. Seven speech acts begin with the exclamatory wa. 1.2.3 Pragmatic devices The analysis that follows, directed at the scope of straightforward and composite speech acts, assumes an institutional presence throughout Wulfstan's homilies. This presence, already seen in the example above of a scriptural promise (see 1.1), issues from a spiritual and moral principle apparent in a contrast between hortatory encouragement and monitory

114

Eugene Green

warning. In the context of this principle many of Wulfstan's speech acts bespeak, say, either a hortatory or monitory intention, whereas other utterances are composite in force. To sort out the differences between straightforward and composite speech acts requires examining the pragmatic devices as well as the syntactic structures Wulfstan uses. Several pragmatic devices help to determine whether particular speech acts are straightforward or composite in their illocutionary force. For straightforward uses of illocutionary force, a pragmatic device that promotes a sense of unanimity in recognizing Wulfstan's intention is that of inclusiveness. Inclusiveness pertains to whatever linguistic elements Wulfstan uses to achieve a sense of harmony, comradery, joint purpose or the like with his auditors. Section 2.1.1 discusses the linguistic elements that help to advance the pragmatic purpose of attaining inclusiveness. Wulfstan also relies on intensifiers to underscore his intentions, to make them clear in an utterance; intensification (2.1.2) is thus a second pragmatic device for identifying straightforward speech acts. A less reliable pragmatic device for identifying straightforward speech acts is that of quoting biblical and patristic sources. If the scriptural promise already discussed (1.1) exemplifies a speech act, composite in intention, most of the quoted instances are straightforward (2.1.4). For speech acts composite in intention, Wulfstan is similarly systematic in his use of pragmatic devices. Inclusiveness is a device for supporting unanimity in straightforward speech acts, but selective inclusiveness (2.2.1) assumes distinctions between communicants committed to Wulfstan's parenetic efforts and those whom he means to admonish. Just as calls for penance enlist the hypothetical imperative (1.1) to encourage and warn, so Wulfstan has other pragmatic devices for speaking at once to groups differently disposed among his audience. Selective inclusiveness, then, is a pragmatic device to approach groups in an audience varying in moral and spiritual commitment and therefore differing in a willingness to subscribe to Wulfstan's teachings. A second pragmatic device accompanying the composite intentions of a speech act is that of contrastive emphasis (2.2.2). Contrastive emphasis generally relies on elements of negation to distinguish what to embrace from what to shun. In speech acts containing such negative elements the composite intentions are unmistakable, especially intentions hortatory or monitory. A third pragmatic device affects speech acts that are composites of assertion and parenesis (2.3). For these, the over-all force is typically assertive, a proposition expressed about the state of affairs, yet the structures and linkages of clauses promote parenetic effects. Assertion and parenesis so joined

Speech acts in Wuljstart's homilies

115

enables an auditor to regard an utterance as constative (as saying something about a world) or to recognize it also as having a deontic purpose. With this third pragmatic device for speech acts of composite intention, Wulfstan appears to imply that those who have ears to listen will take heed.

2. Analysis and results 2.1 Straightforward 2.1.1

speech acts

Inclusiveness

Inclusiveness is a term indicating a desired concurrence between a speaker and an audience. Applied to speech acts, the term suggests that the speaker's intention is to win the agreement or compliance of those addressed. What counts is not a demonstrable resolution - a request satisfied, a command executed, a promise fulfilled — but a sense that the speaker intends his audience inclusively to concur. 2.1.1.1 Performative verbs Generally, a use of performative verbs, either in direct or reported speech acts, implies that the speaker has an inclusive intention. All of Wulfstan's uses of performative verbs in straightforward direct and indirect speech acts are compatible with inclusive propositions (for performative verbs in speech acts of composite intention, see 2.2.1.1). Here are two selfevident examples: For Godes lufan we biddad manna gehwylcne that he bepence georne hine sylfne. 'For the love of God, we request that each person earnestly consider himself.' (Bethurum 1957: 231); [Christ] Icerde peak poet man to wacmod ponne ne wurde. 'Christ urged, however, that one not become then too faint-hearted.' (Bethurum 1957: 120).

In this sense a use of performative verbs, either in direct or reported speech acts, implies that the speaker aims at solidarity with his audience. 2.1.1.2

utan

Wulfstan's speech acts often feature the form utan 'let us' a linguistic element that works pragmatically to advance inclusiveness. The form utan refers both to the speaker and to the audience, as in the following:

116

Eugene Green

And utan warnian us eac swa wiei his unlara nu swyöe 7 God celmihtigne georne biddan peet he us gescylde wid peene peodscadan. 'And let us moreover take much heed now against his bad teaching and earnestly bid God Almighty that he protect us against that enemy of the people.' (Bethurum 1957: 118)

The predicate utan here governs two speech acts, each hortatory, the first urging precaution, the second entreaty; unfortunately, Mitchell (1987, 1: 384) introduces his section on utan by labeling it a form expressive of a command or wish — an oversimplification. As in the 111 instances of speech acts with utan in the corpus, these two admit of no qualifying phrase or clause that without caviling implies anything other than exhortation. In homiletic utterances, for the speaker to count himself within the scope of a speech act argues that his intention is to be as straightforward as possible. 2.1.1.3 The first person plural pronoun Wulfstan has three other ways to combine inclusiveness and exhortation, two by means of the nominative pronoun, the other by means of the oblique pronoun. The nominative pronoun appears in groupings like we sculon 'we ought to' and don we 'let us do', the first person plural present subjunctive form. As for the modal auxiliary (also including we agan 'we ought' in the eleventh century), one of the very few in the corpus occurs in the homily on the Christian life: Pas beboda 7 fela hertoeacan we sculon healdan, be pam pe us Cristes bee gelomliee Icerad, gif we Gode willap rihtlice hyran, swa swa us pearf is ... 'These commands and many besides we ought to observe, about which Christ's books repeatedly teach us, if we want to obey God rightly, as is needful for us ...' (Bethurum 1957: 201)

If the sixteen instances of this pattern are infrequent, Wulfstan has but two examples of the inclusive, subjunctive form: 7 se de pcet nelle, ehte we his ealle mid woroldlicre steore, pe Ices pe we habban cenigne gemana nu heora synna 7 eft heora wita. 'And he who will not (love God), let us pursue all of his with worldly punishment, lest we have any share now of their sins and thereafter of their guilt.' (Bethurum 1957: 277)

That these two utterances stand as straightforward speech acts, unaffected by composite intentions, is hardly as certain as for those introduced by utan. The hypothetical gif clause of the we sculon utterance is indicative of an anxiety about harmonious inclusiveness (see 2.2.1.1), and the negative pe ICES pe of the subjunctive utterance is all the more. What

Speech acts in Wulfs tan's homilies

117

accounts for these sporadic utterances is an evident, troubling contradiction between the inclusiveness of we and the admonitory purpose of a subordinate clause introduced by, say, pe Ices pe. The oblique use of the first person plural pronoun appears, for example, at the opening of a homily for Lent: Leofan men, us is swyde mycel pearf on oelcne timan poet we gemyndige beon peera pinga pe us for gode to dearfe magan. 'Dear men, the need for us is very great at every opportunity that we be mindful of the things that for the good of us are advantageous.' (Bethurum 1957: 233)

The pronoun us, stressed sufficiently after the vocative, contributes to a sense of inclusiveness, because of its relation also to the first person pronouns of the nominal clause. The sequencing of leofan men and us, indeed, effects a union between speaker and audience as the utterance itself unfolds. Although this pattern appears sparingly, nine instances, its position in the homilies, all but twice at the opening or close or at the onset of a paragraph, is strategic. What also adds to the emphasis of the speech act is Wulfstan's use of the intensifying phrase mycel pearf. Indeed, the use of intensifiers is a pragmatic characteristic of speech acts that have a straightforward, illocutionary force. 2.1.1.4 Inclusiveness and negation: The third person In contrast to the first person pronoun, Wulfstan achieves through the use of the third person an inclusiveness by a pragmatic device of delimiting. All the communicants, as he addresses them monitorially, linking the stipulative adjective aenig to forms of negation, need to resist sinful acts that would consign them to the damned. In more than fifty such admonitions, his efforts to forewarn assume a pattern like that found in this speech act: Ne cenig man oderne on unriht ne fordeme. 'Nor should any one condemn another unjustly.' (Bethurum 1957: 204)

The categorical force of such an utterance suggests that together with inclusiveness Wulfstan often relies on intensification. 2.1.2

Intensification

If inclusiveness binds speaker and listeners, the intensifier has an analogous effect of making the intention of a speech act plain. The premise is that an intensifier such as mycel in mycel pearf helps both to gain the listener's recognition and to induce a commitment to the act proposed.

118

Eugene Green

Wulfstan's pragmatic use of intensifiers, a characteristic found in more than 200 of his straightforward speech acts, nonetheless works in a delimited group of grammatical patterns. In some instances his practice is to include intensifiers with patterns already identified under the rubric inclusiveness: thus intensifiers accompany utan (51 speech acts) and the oblique pronoun us (8 times). In contrast, those speech acts expressed with the first person plural subjunctive or modal auxiliary (25 in sum) have but one instance of an intensifier. Here follow representative speech acts (the total number for each group set in parentheses), most marked with an intensifier, exemplifying a somewhat different grammatical pattern: 2.1.2.1 Intensifier with third person subjunctive Gyf hwa agylte, preage man pcene 7 steore him georne, 7 pa geonglingas swinge man swype. 'If someone sins, one ought to chastise him and reprove him zealously, and one ought to chastise youths severely.' (Bethurum 1957: 193) (53). And do swa ic leere, hyran pa gingran georne heora yldran, 7 lufian 7 leeran pa yldran heora gingran. 'And do as I urge, let the young earnestly obey their elders, and love, and let the elders instruct their youths.' (Bethurum 1957: 209) (12)

In the first example, the pronouns (hwa, man) have distinct, pragmatic functions: hwa, part of a gif clause, has an altogether indefinite reference, whereas man is distributive throughout Wulfstan's auditors. The auditor recognizes Wulfstan's emphatic intentions, and himself as included within the scope of the speech act, yet not directly singled out. The second example combines in its imperative and subjunctive moods the effects of inclusiveness and intensification. 2.1.2.2 Intensifier with third person modal auxiliary Ac pas tide man sceal mid mycelre forhcefednesse healdan, 7 sceal manna gehwylc hine sylfne georne bedencan 7 don swa him pearf is, andettan pa misdceda pe he mid him sylfum wat 7 georne betan. 'And one should observe this season with much self-restraint, and each person should earnestly examine himself and do what is needful for him, confess the sins that he knows about himself and zealously atone.' (Bethurum 1957: 234)(19) Bisceopas syndon bydelas 7 Godes läge lareowas, 7 hy scylon georne oft 7 gelome clypian to Criste 7 for eall cristen folc pingian georne; 7 hy scylan georne Godes riht bodian 7 ceghwylc unriht georne forbeodan. 'Bishops are preachers and teachers of God's law, and they should earnestly call to Christ time and again and intercede earnestly for all Christian folk; and they should zealously teach the law of God and zealously forbid every wrong act.' (Bethurum 1957: 243) (4).

Speech acts in Wulf start's homilies

119

2.1.2.3 Intensifier with imperative mood Leofan menn, understandad swyde georne peel ge cefre habban rihtne geleafan on cenne celmihtigne God. 'Dear people, understand very earnestly that you should always have a righteous belief in one almighty God.' (Bethurum 1957: 158) (22)

2.1.2.4 Intensifier with auxiliary and infinitive construction Se is to lufianne 7 to weordianne ofer ealle oder ding. 'He is to be loved and exalted over all other creatures.' (Bethurum 1957: 207)(4)

Since Se refers to God, the absence of a pronoun elsewhere in the utterance actually has an encompassing effect, supported by the intensifier. 2.1.3

Miscellaneous

Nearly forty speech acts in Wulfstan's homilies are straightforward exhortations, unmarked by forms of inclusiveness or intensification. If none of these speech acts has any distinguishing grammatical pattern, they each share other pragmatic attributes that pertain to a straightforward speech act. As a whole they urge communicants to reflect soberly, to attend to rituals and sacraments, to perform virtuous deeds, and to exercise self-restraint. That these speech acts constitute an integral group requires but a contrast with one similarly structured yet composite in its effect: (1)

(2)

And se pe ne cunne poet Leden understanden, hlyste nu on Englisc be suman dcele hwcet poet Leden cwede ... 'And he who cannot understand Latin, let him now listen in English to some extent to what the Latin says ...' (Bethurum 1957: 252) And se de on unriht abysgode hine sylfne, se on halgum gebedum abysgie hine symle. 'And he who has engaged himself in sin, let him engage himself in holy prayer.' (Bethurum 1957: 207)

The first example uses se pe as a pragmatic formula of politeness, for the homily immediately continues in English, and so everyone in attendance is urged to listen and to reflect. The second example also relies on se pe as a formula of politeness (sin touching everyone), yet its force is both hortatory and admonitory, depending on an auditor's disposition. The difference between the two examples lies in an institutional premise of moral accountability (see 1.1.1). If one fails to attend to the homily (example one), nothing in the speech act implies a penalty. If one should

120

Eugene Green

neglect the exhortation to pray (example two), then the implicit consequence is severe. The upshot is, then, that the speech act of composite, deontic significance is institutionally inherent to orthodox, moral teaching (see 1.1.2). 2.1.4

Direct

quotations

Throughout Wulfstan's homilies, there occur more than eighty instances of speech acts, directly quoted, sometimes immediately following a Latin source. Many of these, except for about fifteen promises (see, for instance, 1.1 above), are straightforward in illocutionary force. What makes these quotations worthwhile for analysis is that although translated, they count in a sense as possessed of greater, pertinent authority than Wulfstan's many own speech acts. Moreover, these quoted speech acts have a distinctive pragmatic device of their own, associated with the second person pronoun. If Wulfstan has twelve examples of the second person pronoun, singular or plural, in his own speech acts, forty percent of those quoted have a form of pu or ge. The citations following illustrate a command, an exhortation, a promise, and a warning: Ne weorda pu fremde godas. 'Do not worship strange gods.' (Bethurum 1957: 201) Ailmihtig God sod lice pus cwced: Adweaö eow, ic leere, 7 clcensiad eow georne 7 afyrsiad of minre gesyhde pa ungedanc eowra heortena. 'Almighty God truly says this: Wash yourselves, I exhort, and earnestly cleanse yourselves, and banish from my sight the evil thoughts in your hearts.' (Bethurum 1957: 219) Swa hwcet swa ge gebindad her ofer eordan fcestum bealubendum for yfelum gewyrhtum, eall hit wyrd on heofenon swa swa on eordan mid Godes yrre gebunden swiöe faste, butan ge lidian 7 da bendas alynian. 'Whatsoever you bind here on earth with secure, pernicious bonds for wicked deeds, it will all be bound quite securely in heaven as on earth with God's anger, unless you are mild and loosen the bonds.' (Bethurum 1957: 244) 7 gif pu sinfullan nelt scealt pa sawle bitere 'And if you will not crimes of the wicked, 240)

synna gestiran 7 pam manfullan mandceda cypan, pu forgildan. reprove the misdeeds of the sinful and expose the you shall requite the soul bitterly.' (Bethurum 1957:

2.2 Speech acts of composite intention The two categories — inclusiveness and intensification — that apply to many straightforward speech acts have comparable, pragmatic categories

Speech acts in Wulfs tan's homilies

121

— selective inclusiveness and contrastive emphasis - under speech acts of composite intention. These comparable categories characterize many of the 257 speech acts, composite in intention, throughout Wulfstan's homilies. A third category of pragmatic devices refers both to patterns of principal and subordinate clauses and to cited utterances, themselves counterparts of the direct quotations listed under straightforward speech acts. 2.2.1 Selective

inclusiveness

Whereas inclusiveness is a category that joins the homilist to his audience, selective inclusiveness as a term pertains to differences in inference regarding what intent a speech act has. These differences mitigate against inclusiveness. So the same speech act, in the same context, addressed to all auditors, finds them divided on its illocutionary force and at odds, even if implicitly, with one another. 2.2.1.1 Hypothetical clauses One pragmatic device that creates selective inclusiveness is the hypothetical clause that conditions the force of a speech act. In the homilies Wulfstan typically relies on two linguistic elements to introduce the hypothetical condition: the clause with its verb inflected as an imperative as part of a promise (see 1.1) and the clause introduced by the subordinate conjunction gif. The first of these linguistic elements occurs twice, the second in twenty speech acts. In almost all of these, auditors need to decide whether to accept the condition specified; if some do not, their refusal or resistance separates them into divergent groups. Wulfstan uses gif pragmatically in quoting authority and in speaking for himself. Of the two examples offered here, one is an instance of a reported speech act: Fordam God sylfa behet synfullum mannum pcet he wolde miltsian, gyf hi woldon earnian. 'Therefore G o d himself promised the sinful that he would show mercy if they would merit it.' (Bethurum 1957: 210); 7 hcefd nu jrnrh his gyfe manna gehwylc, gif he geearnian wylle, heofona rice. 'And through his gift every person now has, if he wishes to merit it, the kingdom of heaven.' (Bethurum 1957: 228)

An earlier instance of the hypothetical (see 2.1.1.3) reveals how much aware Wulfstan himself was of the tensions that the hypothetical poses for inclusiveness; in that example the gif clause — gif we Gode willap rihtlice hyran — immediately precedes an impersonal clause of necessity.

122

Eugene Green

2.2.1.2 Sepe A feature frequent in speech acts of composite force is that of the pronominal combination se de and its analogs (67 instances). Although as already seen (2.1.3), a speech act characterized by se de is neither predictably straightforward nor composite, Wulfstan combines these pronouns with verbs inflected differently for varying effects. If se as subject takes subjunctive forms of the verb (2.1.3) for purposes of compliance or those of exhortation and admonishment, its taking the indicative (20 instances) has other effects: And witodlice se man fuefd wisdom purh Godes gyfe pe wislice leofad 7 symle smead hu he Gode gecweman mage. 'And truly that person has wisdom through the gift of God who wisely believes and always considers how he may serve God.' (Bethurum 1957: 186)

Here the force of the utterance is largely assertive, yet Wulfstan yokes on a hortatory implication, and maybe an admonitory nuance at the periphery. Other inflectional patterns or forms of verb phrase as predicate for se de, se man...se, and the like are sporadic. If se pe as a pragmatic device has a measure of politeness (the listener is presumably free to beg off), Wulfstan sometimes also uses generic nouns that compel selective inclusiveness. So in a homily on the Creed, after a warning about hell, there follows this catalogue (partially given): Dyder sculan mannslagan, 7 dider sculan manswican; dider sculan mvbrecan 7 da fulan forlegenan...7, hrcedest to secganne, ealle pa manfullan pe cer yfel worhton 7 noldan geswican ne wip God pingian. 'Thither shall murderers, and thither shall traitors; thither shall adulterers and the foul fornicators ... and, summarily said, all the wicked who have ever done evil and would not cease nor conciliate with God.' (Bethurum 1957: 163)

The advance to the word manfullan casts any audience at all apprehensive into a mode of self-examination, each noun pragmatically effecting a review of one's innocence or vulnerability. The exclamatory wa appears with both analogs of se de and with generic nouns of selective inclusiveness. So in a homily that closely follows Isaiah, Wulfstan writes. Wa manfullan, he cwced, for his misdcedan; edlean his weorca eal he sceal habban. 'Woe to the wicked, he said, for his sins; he shall have full compensation for his works.' (Bethurum 1957: 215)

Speech acts in Wulf start's homilies

123

Here wa is a pragmatic device to introduce a generic class, warned against, yet broad enough to touch many in a church audience. From the Book of Isaiah, Wulfstan also has Wa eow, he cwced, pe lecgad togcedere hams 7 cehta on unriht begytene on ceghwilc healfe. 'Woe to you, he said, who join together dwellings and property wrongly begotten on each side.' (Bethurum 1957: 216)

This use of wa has an affinity, too, with contrastive emphasis, for as a device for a monitory speech act it implicitly encourages reform. 2.2.2 Contrastive emphasis Contrastive emphasis invites auditors to examine the alternatives between hortatory and monitory possibilities. The issue concerns less the identity or group to which one belongs than the moral choices that the homilist contrasts as virtuous or sinful. One pragmatic device that helps to emphasize this distinction in nearly 100 speech acts is a negative particle of some kind. Thus negative particles, together with a use of the modal auxiliary *sculan, underscore differences between what one ought to do and ought to avoid in this combined admonishment and assertion: Sume hy wurdon cet nyhstan swa purh deofol ahyrde pat hi ncefdon to Gode naöer ne lufe ne ege swa swa hy scoldan, ac durh deofles lare unriht lufedon ealles to swyde. 'Some there were who became at last so hardened by the devil that they have had neither love nor awe for God as they should, but because of the devil's teaching have embraced vice all too much.' (Bethurum 1957: 149)

The evident contrast does not have the direct force of straightforward admonition; it may instead, at least for some, have the effect of arousing one's own sense of guilt. A second pattern of contrast enlists the pronominal se de and contrastive clauses in the subjunctive mood, as in Se pe ware gifre, weorde se syfre... 'He who has been greedy, let him become temperate...' (Bethurum 1957: 206)

This example of an implied admonishment and a hortatory encouragement is one in a sustained cluster in a homily on the Christian life. Here the pragmatic device depends on the juxtaposing of antithetic elements in the relative and principal clauses. Otherwise the modal auxiliary *sculan more frequently contributes to the contrastive emphasis of speech acts yoked closely together.

124

Eugene Green

2.2.3 Pragmatic devices, assertions, parenesis, and promise For some homilies, especially in Wulfstan's closing paragraphs, utterances have a force comprising the expression at once of several speech acts. In such instances, the pragmatic devices in the utterance offset one another to offer the listener a concluding statement at once assertive and deontic. The effect is to have an auditor recall the theme of a homily by means at least of a recapitulative summation. What is distinctive about these summarizing statements is that they appear in the indicative mood, commonly associated with assertion, and imply, rather than enunciate, a hortatory speech act. Here is one example: And facet is swide rihtlic 7 faearflic dom 7 Gode gecweme, 7 se fae facene dom gehealded, he geeamad him sylfum witodlice dom facene betstan at faam fae on heofenum leofad 7 ricxad in ealra worolda woroid a butan ende, amen. 'And that is a judgment, both quite right and profitable, and pleasing to God, and he who upholds that judgment, shall himself merit truly the best judgment from those who live and rule in heaven in world without end, amen.' (Bethurum 1957: 250)

The hortatory tone, unsurprisingly, follows the use of se de and gains support from the intensifier witodlice. What is new, however, is the sense of a promise, indirectly rendered, not quite articulated in distinct terms. A final pattern of utterance in this subcategory is that of a quotation that stands as an indirect speech act; like the example just quoted, this one also appears in a closing paragraph: Godes agen bebod isfacetman lufige God sylfne mid eallum mode 7 mid eallum mcegene, 7 oder is facet man lufige his nehstan swa swa hine sylfne. 'God's own command is that one love God himself with all heart and with all strength, and another is that man love his neighbor as himself.' (Bethurum 1957: 174)

The citing of a command rather than its direct enunciation turns a directive into an assertion and permits an opportunity for hortatory appeal. Insofar as words like bebod have this double effect of transforming one kind of speech act into another (command into assertion), they comprise, too, an instance of a pragmatic device. The device attains its effectiveness in that clauses after bebod do not take the imperatives of commands (say, lufa or lufiad) but subjunctives appropriate for noun clauses. What such transformation enhances, as well, is the possibility of free commitment. The term bebod, alluding to da bebodu 'the ten commandments', nonetheless admits in the context of Wulfstan's utterance scope for consideration and for the use of hortatory persuasiveness.

Speech acts in Wulfstan 's homilies

125

3. Discussion The occurrence of pragmatic devices, systematically related, is not surprising in Wulfstan's homilies. Since his purpose is to persuade auditors to commit themselves to their faith and to live by its precepts, his dependence on pragmatic devices in Old English is inescapable. That these devices should readily subsume themselves under such categories as inclusiveness, intensification, and contrastive emphasis stems from the task Wulfstan aims to realize through the genre of the homily. The homilist needs, after all, to gain the allegiance of his auditors, to win as many as possible to the faith: hence the need for inclusiveness. One way to attain allegiance to the faith is for the homilist to demonstrate his own conviction: hence the value of intensification. What is singularly important, however, is that the commitment of the communicant be a free choice: hence the indispensability of devices enabling contrasts between the virtuous and the vicious. This last device, that promotes contrast, underlies the basis for speech acts of composite intention. If a willing subscription to the faith is one goal of a homily, then auditors need to weigh alternatives, bolstered by hortatory and monitory suasion. As many of the instances provided reveal, Wulfstan continuously challenges his auditors to examine their own motives and their responsibilities. What makes the challenge all the more effective is that although Wulfstan's utterances have composite effects, they draw upon patterns, linguistic and pragmatic, familiar to his auditors. In effect, as auditors explore themselves and their motives, they enjoy the support of linguistic patterns and pragmatic devices they have heard and used repeatedly. What might be new is that Wulfstan marshals these patterns and devices for a genre distinctively its own in AngloSaxon England. It is one thing to know implicitly patterns of language and pragmatic devices of communication; it is quite another to draw upon them for the purposes of examining oneself. References Bethurum, Dorothy 1957 The homilies of Wulfstan. Oxford: Clarendon. Frantzen, Allen J. 1983 The literature of penance in Anglo-Saxon England. N e w Brunswick: Rutgers University Press. Mitchell, Bruce 1987 Old English syntax, vol. 1. Oxford: Clarendon. Recanati, Frangois 1987 Meaning and force: The pragmatics of performative utterances. Cambridge: Cambridge University Press.

Argument-predicate structure in grammar and performance: A comparison of English and German John A. Hawkins

1. Some major grammatical contrasts between English and German In this paper I would like to return to a traditional problem in Germanic linguistics and try to pour some new wine into an old bottle: What are the consequences, for the grammar and for language performance, of putting the verb in different positions in a clause? How are verb-final and SVO structures processed in performance? And can we attribute any major grammatical differences between English and German to their verbposition contrasts? In order to answer this, I shall draw on and develop some ideas from several subdisciplines of linguistics, specifically generative grammar, language typology, and psycholinguistics. I shall propose a hypothesis about the consequences of these different verb positions for argument-predicate structure in grammar and performance, and I will test this hypothesis on data from English and German. I will then test it further on some data from other languages. The major focus of this paper will accordingly be the grammar and use of argument-predicate relations. I begin with the observation that there are three well-known differences between German and English with regard to the surface coding of these relations, exemplified by a pair of sentences such as Der Mann wird den Hund schlagen and The man will beat the dog: (1)

German has surface cases, English does not (outside of pronouns);

(2)

Subject (SU) and direct object (DO) have fixed positions in English, but not in German (Den Hund wird der Mann schlagen is also grammatical in German);

(3)

The position of the main or lexical verb is regularly final in German, except when moved to second or first position in

128

John A.

Hawkins

main clauses that lack an auxiliary, whereas English has no corresponding verb-final structures. There are also many subtle contrasts between English and German in the actual grammar of argument-predicate relations, illustrated in (4—10) (cf. Hawkins [1986, 1992] for detailed discussion). The relevant rules or properties are systematically more productive in English than in German: (4)

a.

The noise ceased to get on his nerves. [Subject-to-Subject Raising]

b.

?Der Lärm hörte auf, ihn aufzuregen.

(5)

a. b.

Literature is boring to study. [Tough Movement] * Die Literatur ist langweilig zu studieren.

(6)

a.

I believe the farmer to have killed the cow. [Subject-to-Object Raising] *Ich glaube den Bauern, die Kuh geschlachtet zu haben.

b. (7)

a. b.

What did you assume that we would not bring? [more WHmovement] * Was hast du angenommen, daß wir nicht mitbringen würden?

(8)

a. b.

My guitar broke a string, [more semantic diversity of GRs] * Meine Gitarre hat eine Saite zerrissen.

(9)

a.

John broke the branch/broke the rope, [broader selectional restrictions] Johann brach den AstIzerriß das Seil.

b. (10)

a. b. c.

This door will open (new possibilities), [broader subcategorization possibilities] Diese Tür wird sich öffnen. * Diese Tür wird neue Möglichkeiten öffnen.

One interesting question raised by these facts is this: Are the contrasts of (1-3) and (4-10) related? Do the contrasts in grammar follow in any way from the contrasts in surface coding? For example, does the existence of surface cases have any consequences for the grammatical processes that refer to NPs bearing these cases? In my book (Hawkins 1986) I argued that surface cases may indeed have such consequences. More generally, I argued for a descriptive generalization underlying all of (1-10): English regularly allows greater ambiguity (or vagueness) of surface forms and structures (e. g., Equi-Raising ambiguities), more "argument-trespassing"

Argument-predicate

structure in grammar

and performance

129

(e. g., more WH-extraction), and more argument deletions than G e r m a n , all of which introduce a more distant a n d less transparent m a p p i n g between surface form and meaning. In this paper I want to pursue a different hypothesis (following H a w kins 1992). The position of the verb (i. e., contrast [3]) will now be argued to be crucial in explaining both the grammatical contrasts of (4-10) and the other surface coding contrasts of (1-2). The significance of verb position as a predictor of, and an explanation for, m a n y other typological properties does not follow f r o m grammatical considerations alone, however, but f r o m a theory of performance.

2. A major performance contrast between English and German Frazier (1985) points out that there are m a n y "temporary ambiguity" and "garden p a t h " sentences in English. Temporary ambiguity means that two or more grammatical analyses can be assigned to a given string of constituents in parsing, before subsequent material in the parse string disambiguates. A garden path means that the first reading selected (if any) turns out subsequently to be the wrong one, forcing a reanalysis when the later material is encountered. Relevant examples are (11-14), with Frazier's parsing principles given in square brackets: (11)

Bill knew the answer was correct. [Minimal Attachment]

(12)

While Mary was reading the b o o k fell down. [Late Closure]

(13)

The candidate expected to win lost. [Minimal Attachment]

(14)

Mary helped the boy and the girl ran away. [Minimal Attachment]

The portions in r o m a n type are all temporarily ambiguous a n d gardenpath the hearer. N o n e of these sentences would be temporarily ambiguous, Frazier observes, if English possessed various grammatical devices to force the intended reading on-line, for example an obligatory that complementizer in (11): (11')

Bill knew that the answer was correct.

130

John A.

Hawkins

verb-final order in (12): (12')

a. b.

While Mary the book reading was, ... While Mary reading was, the book fell down.

or an unambiguous morphology and/or syntax for active/passive ambiguities such as (13), or a case system disambiguating the different arguments to be assigned to different verbs in (14). What is striking is that exactly these devices are present in German, and the corresponding sentence types are regularly unambiguous on-line: (11")

a. b.

Bill wußte, daß die Antwort richtig war. Bill wußte die Antwort.

(12")

a. b.

Während Maria das Buch las, ... Während Maria las, fiel das Buch herunter.

(13')

a. b.

Der Kandidat erwartete, daß er gewinnen würde. Der Kandidat, von dem erwartet wurde, daß er gewinnen würde

(14')

a. b.

Maria half dem Jungen und dem Mädchen. Maria half dem Jungen, und das Mädchen lief weg.

Moreover, the contrasting structures of (4—10) also regularly introduce temporary ambiguities and permit garden paths in English (cf. the portions in roman type in these sentences), while corresponding sentence types and temporary ambiguities are either nonexistent or much less extensive in German. Overall we can say that the grammar of English appears to be systematically tolerant of structures that produce temporary ambiguities in performance, whereas the grammar of German is not and structures involving temporary ambiguities and possible garden paths are a major source of contrast between the two languages. Why should this be? In order to answer this question, we need to examine performance in greater detail. The temporary ambiguity contrasts suggest that there are different performance strategies for the processing of basic clause structure in the two languages.

3. The source of temporary ambiguity Temporary ambiguities and garden paths arise because alternative "predicate frames" or valency structures for one and the same verb are compatible with the on-line data of the parse string and can be activated by the

Argument-predicate

structure in grammar

and performance

131

verb in processing (cf. Hawkins 1992). For example, the verbs break and believe in English can co-occur with a number of different argument types and are listed in the lexicon alongside many predicate frames, including those of (15) and (16) (for ease of exposition grammatical details are kept deliberately to a minimum in these examples): (15)

break

Frame

1. NP V - NP [Agent] [Patient] 2. NP - V [Patient]

John broke my guitar

3. N P — V — P P

A string broke on my guitar

My guitar

broke

[Patient] [Locative] 4. N P - V — N P

My guitar broke a string

[Locative] [Patient] 5. N P — V — N P

My guitar broke a world record

[Instrument] [Patient] (16)

believe

Frame

1. NP - V - NP [Agent] [Dative] 2. NP - V - NP [Agent] [Patient] 3. NP - V - S

I believe the farmer I believe this report I believe (that) killed the cow

the

[Agent] [Object of belief] 4. NP - V - NP - I believe the farmer VP killed the cow [Agent] [Agent] When sequences such as My guitar broke...

farmer

to

have

and I believe the farmer...

are

parsed, these words are compatible with a plurality of predicate frames for each of the verbs, and it is this plurality that gives rise to a temporary ambiguity and possible garden path. It is plausible to assume that these alternative predicate frames are activated by the predicate (V) in processing, and not by the arguments of the predicate, an assumption that has been supported in Tanenhaus and Carlson (1989). Agent and Patient, etc. do not select for, and predict, particular predicates, because they are each compatible with too many. But a predicate can activate a finite and predictable set of predicate frames on-line, and if it occurs first before its arguments, these latter will gradually select one from the set of possibilit-

132

John A.

Hawkins

ies that have been activated. On the other hand, if the arguments occur first, before the predicate, they will not enable the language user to predict which predicate or predicate type is going to occur. As a result, alternative predicate frames will not be activated until the predicate is encountered, and the processing of arguments will take place in advance of the activation of these alternatives. Hence, there will not be a temporary ambiguity on-line between predicate frames, and predicate frame activation and selection will take place simultaneously at the verb. More generally, it seems that the relative order of predicate and arguments can be highly significant in performance and that the observed temporary ambiguity and garden path contrasts between English and German point to a much deeper contrast in the processing of argumentpredicate relations, crucially involving the position of the (lexical) verb.

4. Outline of a processing model for different verb positions For a language such as English, with an early verb position, I assume that there are three basic left-to-right stages in the processing of argument-predicate structure within a clause, shown in (17): (17)

English

Β A: look-ahead (for argument-predicate structure) B: predicate frame activation C: resolution of temporary ambiguity Stage A is a look-ahead stage, as far as argument-predicate structure is concerned at least. Syntactic information can be processed at this point, for example, syntactic categories such as NP and PP can be recognized, but not argument-predicate structure (i. e., the full set of syntactic and semantic information relevant to the grammar of argument-predicate relations, including grammatical relations, semantic or theta-roles). Β is the point of predicate frame activation. If several competing frames are compatible with the material in A and B, then a temporary ambiguity arises. Either a preferred reading may be selected to the exclusion of

Argument-predicate

structure in grammar

and performance

133

dispreferred ones (in accordance with Frazier's [1985] First Analysis Constraint), or multiple analyses may be held in parallel (in accordance with other psycholinguistic theories, cf. Hawkins 1992 for discussion), or there may be both a multiple analysis and a preference for one reading. During stage C the temporary ambiguity must be resolved. If a particular frame was initially preferred, the preference must be confirmed or disconfirmed. If it is disconfirmed, a hitherto dispreferred reading will be selected, and we will get the kind of garden-path effect illustrated above. If no preference existed earlier, a selection will now be made from among the multiple analyses activated at B. The processing of a verb-final German clause proceeds very differently, however, as shown in (18): (18)

German V Ä

~' Β C

A: look-ahead (for argument-predicate structure) Β and C: predicate frame activation and immediate selection of intended frame The look-ahead stage for German is now much longer, again as far as argument-predicate structure is concerned. A lot of syntactic processing can be done prior to the verb, but the assignment of arguments to their appropriate predicates, with their appropriate grammatical relations and semantic or theta-roles, etc., is crucially dependent on the verb. When the verb is then encountered, stages Β and C for English are, in effect, collapsed. The predicate frames for a given predicate are activated, and immediately the intended frame is selected, based on the preceding arguments within A. If we compare the manner in which basic sentences are processed in English and German, therefore, we see some very significant differences. It is these differences that explain the difficulties that English or American learners of German have when trying to adjust to the verb position of German. Speakers of a language in which predicate frames are activated early (cf. stage B) have a hard time delaying that activation and thereby extending the look-ahead period. If you are used to processing your arguments against the background of a psychologically activated

134

John A.

Hawkins

argument-predicate structure, you will feel that you are having to wait too long for the verb. On the other hand, it may be easier for German learners of English to get used to shortening the look-ahead stage and bringing forward stage B. Their difficulties will probably relate to the post-verbal domain in English and will involve temporary ambiguities and garden path resolutions. It would be interesting to examine the relative ease of going from the one language to the other from this performance perspective. In general, when we compare English and German we can see both advantages and disadvantages in each of the verb-early and verb-late strategies that these languages exemplify. The negative for German consists in the extensive look-ahead period, during which decision-making about argument-predicate structure is necessarily delayed. The positive for German is the regular avoidance of temporary ambiguities and garden paths, both of which are known to increase processing load and cause difficulty. The negative for English consists in the fact that there are extensive temporary ambiguities and many possible garden paths. The positive for English is the avoidance of an extensive look-ahead period, resulting in earlier decision-making about argument-predicate structure.

5. The consequences of verb-final position The collapsing of stages Β and C in a verb-final language (cf. [18]) has, I believe, profound consequences both for the parser and for the grammar. For the parser we have already mentioned one consequence: the predicted absence or infrequency of temporary ambiguity in verb-final structures. The arguments that select the appropriate predicate frame activated by a verb have already been presented prior to that verb in (18). Hence, whereas these arguments will resolve a temporary ambiguity when they follow the verb in stage C of a verb-early structure (cf. [17]), they should preclude it when they precede the verb. The second consequence for the parser is even more fundamental. In a verb-final structure, competing predicate frames are activated at V, and the intended frame must then be selected immediately and correctly before the parser moves on to the next clause. Since the verb is the last element of its clause, and the very next item to be parsed belongs in a different clause, with a different predicate and different predicate frame options, it is important that the intended frame can be recognized at the processing of the verb. Verb-final languages do not have the luxury of

Argument-predicate

structure in grammar

and performance

135

the decision course characteristic of verb-early languages, during which alternative possibilities are held unresolved, or one reading is first selected and subsequently rejected. If the intended predicate frame is not correctly recognized at the verb, it will typically not be recognized at all, since the parser will have moved on. In verb-early structures, on the other hand, immediacy and correctness of predicate frame recognition is not necessary, since there is an extensive decision course (stage C), nor is it even possible when arguments have yet to be processed that will determine the final selection. Consider the on-line decisions that must be made in parsing a verbfinal language in a little more detail. Imagine that the parse string consists of a number of arguments, labelled here XP, followed by one or more adjacent verbs (i. e., if there is more than one, the first will be subordinate to the second), and that these syntactic categories have to be mapped onto a given set or sets of predicate frames, as illustrated in (19): (19)

...XP,, XP 2 , XP 3 ... V (V)

Frame 1. [Agent] [Patient] Predicate Frame 2. [Agent] [Dative] Predicate Frame 3. [Experiencer] Predicate

If there is just one verb in the string, the parser must decide which particular predicate frame is intended for that verb, and which XP in the syntax corresponds to which argument in the predicate frame. If there is more than one verb, parsing decisions increase exponentially: which XP belongs to which V in the syntax and semantics? which particular predicate frame is intended for each V? and which XP in the syntax corresponds to which argument in each predicate frame? In a verb-early language, the parser can gradually eliminate such competing possibilities in stage C of (17). But verb finality makes it desirable to correctly and immediately recognize the intended predicate frame and the intended XP-argument correspondences as each verb is processed, by gathering in the preceding arguments and integrating them into their appropriate argument-predicate structure. This need for immediacy and correctness of decision-making as the verb is processed has, I would argue, numerous consequences for the grammar itself. The grammar of a verb-final language has to help the parser by reducing the search space, i.e., by reducing the number of options for alternative predicate frames and argument assignments that the parser has to consider and select from. More options require more

136

John A.

Hawkins

computational resources, more time, and result in more possibilities for error. But in contrast to the parsing of a verb-early language, the parser does not have the time and luxury to subsequently correct errors when the verb is the last element in the clause and the very next (non-verbal) item to be considered belongs in a different clause with different arguments, predicates and predicate frames. So how can the options be reduced? There would appear to be a number of ways: various surface coding devices can be grammaticalized that permit rapid and correct argument-predicate recognition, such as case-marked NPs; certain rule types can be avoided (like raising rules) that reposition arguments in such a way that argument-predicate structure is distorted and the mapping to the surface syntax of the parse string is complicated; subcategorization restrictions can be made tighter, so that arguments can be paired with their predicates more easily; additional selectional restrictions can be imposed; and so on. Specifically, this performance theory makes the following predictions for the grammar and the lexicon of a verb-final language such as German. First, there will be greater predicate-frame differentiation, as shown in (20): (20)

i) ii)

iii)

iv)

v) vi)

Greater predicate-frame differentiation e. g., fewer competing predicate frames associated with each predicate less semantic diversity of basic GRs (fewer predicate frames with non-agentive SUs, cf. my guitar broke a string [8]); fewer subcategorization ambiguities (verbs are often uniquely transitive or intransitive in German where in English they may be ambiguous, cf. [10]); tighter selectional restrictions (English break vs. German brechen!zerreissen, i. e., arguments are more uniquely and unambiguously associated with their predicates than in English, cf- [9]); fewer (and perhaps no) Raising and Tough frames for German predicates corresponding to the relevant predicates of English {believe), cf. [4], [5], [6]; fewer active/passive ambiguities for one and the same predicate form, [13]; greater differentiation of predicate frames achieved through case system: ich glaube dirlieh glaube das, cf. [16].

All of these contrasts result either in fewer alternative predicate frames that need to be accessed and considered for each verb (cf. [i], [ii], [iv] and

Argument-predicate

structure in grammar

and performance

137

[v]); or in fewer sets of arguments that are compatible with each predicate (cf. [iii]); or in fewer predicate frames that match the explicit surface coding devices of the clause from among the set activated by the verb (cf. [vi]). These quantitative reductions can make the recognition of the intended predicate frame both more rapid and more correct. Second, we expect greater argument differentiation, as illustrated in (21):

(21)

Greater argument differentiation i. e., clearer and more unambiguous XP-argument matching, cf. (19) i) argument differentiation in German relies crucially on the case system: der Mann die Rosen den Nelken vorziehen; ii) argument differentiation will still not be achieved unless there is consistent mapping of argument subtypes onto surface cases encoding SU and DO (e. g., Nom and Acc): Nom-marked NPs = generally semantic agents in German (when a clause has an agent), Acc-marked NPs = generally semantic patients. In English, SU and DO can express a greater variety of semantic roles, cf. my guitar broke a string, the key opened the door, I am cold, (8); iii) OBL argument roles (locative, instrumental, dative) are assigned with greater consistency to PPs and OBL cases (e. g., Dative), not to SU and DO; iv) SU of passive Vs in German is more tightly differentiated and limited to the semantic patient (originally an Acc): the apple was given to mell was given the äpple; der Apfel wurde mir gegebenl* Ich wurde den Apfel gegeben

Again, German has more differentiated arguments, in the sense defined and illustrated here. Third, our theory predicts less argument trespassing in German (cf. Hawkins 1986: 97), as shown in (22): (22)

Less argument trespassing i. e., German avoids processes that position an NP where it will be assigned to an immediate predicate of which it is not an argument i) less WH-movement, cf. (7); ii) fewer (or no) Raising and Tough structures, cf. (4 6); iii) less daß deletion, cf. (11);

138

John A.

Hawkins

iv) case system often serves to distinguish among predicates to which an argument can be attached, cf. (14'); v) more pied piping (e. g., VP Pied Piping). German does indeed exhibit less argument trespassing in all of these structural types. Hence, I would argue that the grammar of German has responded to the parser's need for greater predicate-frame differentiation, greater argument differentiation, and less argument trespassing, as a consequence of its basic verb-final word order. A broad range of contrasts between the two languages can be unified when we adopt a processing, or performance, perspective, and an explanation can be given for the typological clustering of these properties. (For discussion of the grammatical details that are summarized in (20-22), the reader is referred to Hawkins 1986.)

6. Some testable typological predictions If this account of the correlating properties (1-14) in German is correct, it should hold for verb-final languages in general. It would be a clear counter-example to our theory if a verb-final language were discovered that had less predicate-frame differentiation, less argument differentiation, and more argument trespassing than English. Verb-final languages are therefore predicted to exhibit a consistent type along these dimensions. The individual grammatical properties that are selected in any given language to achieve argument differentiation may vary of course (e. g., there may be different types of case systems). But the general direction of contrast with English should not. For verb-early (VSO, VOS, SVO) languages, however, no such consistency is predicted. These languages have less need to grammaticalize devices that facilitate immediate and correct predicate-frame recognition, first because the parser still has the remainder of the clause in which to complete its predicate-frame selection, and second because the intended predicate frame can often not be recognized at the verb anyway since relevant arguments have not yet been presented that are needed for frame selection. As a result, the need for immediate and correct decision-making, and the consequent need for clear predicate-frame differentiation, argument differentiation, and clear argument-predicate matching will impose much weaker requirements on the grammars and lexicons of such languages. These requirements may then be overridden by the kinds of functional considerations that motivated, for example, morphological

Argument-predicate

structure in grammar

and performance

139

syncretism in the history of English (i. e., the loss of the case system), the expansion of Equi surface structures to include Raising, and the greater productivity of WH-movement. Typologically, we therefore expect to see verb-final languages with a consistent set of German-type properties, whereas verb-early languages are predicted to be more variable. More precisely, we can formulate the implicational predictions defined in (23), where the precise motivation(s) for the prediction is/are given in square brackets: (23)

i)

ii)

iii)

iv) v) vi)

if SOV, then case system (or fixed SOV) [predicate-frame differentiation; argument differentiation; less argument trespassing] if SOV, then narrow semantic range of basic GRs (e. g., Nom = Agent, Acc = Pat) [predicate-frame differentiation, argument differentiation] if SOV, then more consistent assignment of OBL semantic roles to grammatically OBL arguments (rather than to SU and DO): OBL roles coded by e. g., OBL cases or adpositions [predicate-frame differentiation, argument differentiation] if SOV, then no (or limited) raising [predicate-frame differentiation, less argument trespassing] if SOV, then no (or limited) WH-extraction out of clauses [less argument trespassing] if SOV, then clear clause boundaries (e. g., no complementizer deletion) [less argument trespassing]

The first five of these predictions are tested in Table 1, using data taken from Müller-Gotama (1991). It should be apparent that the predictions are quite strikingly confirmed. The languages with a basic SOV order are strongly consistent: They all have a case system; the semantic range of their basic grammatical relations is narrow, which is understood here to mean that there is a close correlation between the coding of subjecthood and semantic agents, and the coding of direct objecthood and semantic patients; they exhibit no raising, and either no, or limited WH-extraction. The non-SOV languages, on the other hand, are not a consistent type and exhibit the full range of predicate-frame differentiation, argument differentiation, and so on. English is at the opposite end of this non-SOV typological continuum from a language like Russian, which is fundamentally German-like, and hence English provides us with a particularly clear contrast between the SOV type on the one hand, and the structural possibilities that are available to a non-SOV language.

140

John A.

Hawkins

The distribution of properties in Table 1 is also interesting from a diachronic perspective. It confirms the point made in Hawkins (1986) that case syncretism is not a necessary cause of the shift from SOV to SVO, since SVO can co-occur with a case system. Case syncretism can be a sufficient cause, however, since all the languages without cases are nonSOV, in this sample. More generally, none of the uniquely non-SOV properties can be said to be a necessary cause of the SOV to SVO shift, since each can in principle be absent from an SVO language. But each may be sufficient, and hence a broadening of the semantic range of syntactically defined grammatical relations, or the acquisition of extensive raisings or WH-extractions, may result in a verb shift, according to this typological clustering of properties. This may shed some light on languages like Icelandic, in which a shift from the Proto-Germanic SOV has clearly preceded any case syncretism (cf. Askedal 1993). On the other hand, the implicational dependencies of (23) will undoubtedly need a certain refining, in order to take account of the fact that there are often disjunctive ways of satisfying what are ultimately parsing requirements for argument differentiation. Both a case system and fixed subject-before-object order may be a means of distinguishing subject from object in an SOV language. The loss of a case system may not even be a sufficient cause for the SOV to SVO shift, therefore, if argument differentiation can be achieved another way. At any rate, this kind of typology plus the proposed psycholinguistic motivation for it provides a richer synchronic framework against which to view diachronic change, compared to the now well-worn case-syncretism account which has proven to be both too strong and too weak. It is too strong in that it appears to provide an appropriate sufficient cause for verb shift in English, but not for Dutch and Pennsylvania German (cf. Louden 1992), both of which have undergone case syncretism without undergoing the shift to SVO. It is too weak in that there are clearly other sufficient causes of verb shift as well, witness Icelandic.

7. Conclusions I began this paper by pointing out that English and German contrast with regard to the phenomenon of temporary ambiguity (cf. section 2). I then argued that these contrasts are the result of different strategies for processing argument-predicate structure in the two languages, which are in turn a consequence of different verb positions (cf. sections 3 and 4).

Argument-predicate

structure in grammar

and performance

141

The early position of the verb in English means that the parse string will often be compatible with several predicate frames at the time the verb is encountered and processed, with the intended frame being selected subsequently on the basis of material that follows the verb. Hence, temporary ambiguity and also possible garden paths. The late position of the German verb, however, forces immediate and correct decision-making at the verb position, with an extensive look-ahead prior to that. Numerous structural and semantic consequences were then argued to follow from these contrasting verb positions (cf. section 5). The grammar of German was predicted to require greater predicate-frame differentiation, greater argument differentiation, and less argument trespassing than English, and these predictions were shown to be confirmed. A set of implicational universale was then defined with SOV as the antecedent property, and their predictions were tested on a larger set of SOV and non-SOV languages analyzed by Müller-Gotama (1991) (cf. section 6). Verb-final languages were predicted to exhibit a consistent set of German-type properties, whereas verb-early languages were predicted to be variable. These predictions were shown to be correct. I conclude, first, that there is indeed an underlying unity to a broad range of English-German contrasts (cf. section 1), which I would now describe in terms of the three interrelated phenomena of predicate-frame differentiation, argument differentiation, and (no) argument trespassing. Second, I now believe that the ultimate explanation for this unity is one of language processing, or performance, and I would modify the predictions of Hawkins (1986) accordingly. The typology that emerges from the English-German contrasts relates to the expression of argument-predicate structure across languages. The significant parameters of variation involve: the number of competing predicate frames for individual predicates, more generally the degree of predicate frame differentiation; the degree of argument differentiation; and the amount of permitted argument trespassing; but not other ambiguities or deletions. This typology is more limited than that of my book in some respects, but more extensive in others, since it now includes verb position as a central component, rather than an incidental one as it was before. Third, I conclude that the proper explanation for many grammatical phenomena across languages reduces to principles of language performance, and not competence, and more generally that grammars have the rules and properties that they do both because they incorporate the universal primitives of a competence theory and because they have responded to principles of performance, as illustrated here (cf. further Hawkins 1988 and 1994).

142

John A. Hawkins

Table 1. Grammatical correlations with verb position (from Müller-Gotama [1991]) Language

Basic V-position

Morphological Case System

Semantic Range of Basic GRs

Raising

WHextraction

Korean Japanese Malayalam Turkish German

SOV SOV SOV SOV SOV

yes yes yes yes yes

narrow narrow narrow narrow 0 narrow

no a no b no (no) d no a

no no no no limited

Russian Chinese Sawu Hebrew Indonesian Jacaltec English

SVO SVO V-initial SVO SVO

yes no no e no f no no no

narrow broader narrow narrow broader narrow broadest

noab no a no (ηο) ε limited11 yes yes

limited no no yes yes yes yes

vso SVO

Notes a In Korean, German, Russian, and Chinese there are structures that appear on the surface to have undergone Tough Movement. For Korean, Müller-Gotama (1991: 97-100) argues that this is not the correct analysis; Comrie and Matthews (1990) argue against Tough Movement in German and Russian; for Chinese, Müller-Gotama (1991: 277-278) also argues against a Tough Movement analysis, following Shi (1990). b Japanese and Russian have apparent S-O Raising structures in which the embedded verb must be be. Müller-Gotama (1991: 211—212) argues against a raising analysis in these cases, citing relevant unpublished work on Russian by Comrie. c Müller-Gotama (1991) provides semantic range data only for DOs in Turkish. d There may be some limited S-0 Raising in certain dialects of Turkish, cf. Müller-Gotama (1991: 219). e Sawu does not have case morphology, but it does have a set of case prepositions, including one for ergative NPs. f Modern Hebrew lacks morphological case markers. A case particle et serves as a marker of definite direct objects, however, and this leads Müller-Gotama (1991: 295) to assign "yes" to Hebrew for case. However, by this criterion, many other languages should also be assigned "yes" which lack morphological case and have free-standing case particles or prepositions. I have accordingly converted his "yes" to "no" here. 8 Gundel et al. (1988) have argued against S-S and S-O Raising in Hebrew. Müller-Gotama (1991: 287—288) argues that there may be limited evidence for S-S Raising, though not for S-O Raising and Tough Movement, but he considers its existence "doubtful". h S-O Raising exists in Indonesian, cf. Chung (1976) and Müller-Gotama (1991: 137-128). There is also a limited process of S-S Raising, involving obligatory passivization in the dependent clause, and referred to by Chung (1976) as Derived Subject Raising.

Argument-predicate

structure in grammar and performance

143

References Askedal, John O. 1993 "Configurationality in language typology and diachronic syntax: Evidence from Germanic", Norsk Lingvistisk Tidsskrift 11: 125—134. Chung, Sandra 1976 "An object-creating rule in Bahasa Indonesia", Linguistic Inquiry 7: 41-87. Comrie, B e r n a r d - S t e p h e n Matthews 1990 "Prolegomena to a typology of Tough Movement", in: William C r o f t - K e i t h D e n n i n g - S u z a n n e Kemmer (eds.), 43-58. Croft, William-Keith Denning—Suzanne Kemmer (eds.) 1990 Studies in typology and diachrony. Amsterdam: John Benjamins. Dowty, D a v i d - L a u r i K a r t t u n e n - Arnold Zwicky (eds.) 1985 Natural language parsing: Psychological, computational, and theoretical perspectives. Cambridge: Cambridge University Press. Frazier, Lyn 1985 "Syntactic complexity", in: David D o w t y - L a u r i K a r t t u n e n - A r n o l d Zwicky (eds.), 129-189. Gundel, Jeanette Κ.—Kathleen Houlihan—Gerald Sanders 1988 "On the function of marked and unmarked terms", in: Michael Hammond—Edith A. Moravcsik-Jessica Wirth (eds.), 285-301. Hammond, Michael—Edith A. Moravcsik-Jessica Wirth (eds.) 1988 Studies in syntactic typology. Amsterdam: John Benjamins. Hawkins, John A. 1986 A comparative typology of English and German: Unifying the contrasts. Austin: University of Texas Press—London: Routledge (Croom Helm). 1992 "A performance approach to English/German contrasts", in: Christian M a i r - M a n f r e d Markus (eds.), 115-136. 1994 A performance theory of order and constituency. Cambridge: Cambridge University Press. Hawkins, John A. (ed.) 1988 Explaining language universals. Oxford: Basil Blackwell. Louden, Mark L. 1992 "Language contact and the relationship between form and meaning in English and German", in: Irmengard R a u c h - G e r a l d F. C a r r - R o b e r t L. Kyes (eds.), 217-231. Mair, C h r i s t i a n - M a n f r e d Markus (eds.) 1992 New departures in contrastive linguistics. Vol. 1. Innsbruck: Innsbrucker Beiträge zur Kulturwissenschaft, Anglistische Reihe Band 4, University of Innsbruck. Marslen-Wilson, William D. (ed.) 1989 Lexical representation and process. Cambridge Mass: MIT. Müller-Gotama, Franz 1991 A typology of the syntax-semantics interface. [Ph. D. dissertation, University of Southern California.] Now published in revised from as: 1994 Grammatical relations: A cross-linguistic perspective on their syntax and semantics. (Empirical Approaches to Language Typology II.) B e r l i n - N e w York: Mouton de Gruyter.

144

John A.

Hawkins

Rauch, I r m e n g a r d - G e r a l d F. C a r r - R o b e r t L. Kyes (eds.) 1992 On Germanic linguistics: Issues and methods. Berlin: Mouton de Gruyter. Shi, Dingxu 1990 "Is there object-to-subject raising in Chinese?" Proceedings of the Berkeley Linguistics Society 16. Berkeley: University of California Press. Tanenhaus, Michael K.—Gregory Carlson 1989 "Lexical structure and language comprehension", in: William D. MarslenWilson (ed.), 529-561.

The definition of a grammatical category: Gothic absolute constructions Gregor Hens

1. Introduction This paper examines theoretical issues concerning the fault line between cognitive and generative grammar as it divides the field of linguistics today. The investigation centers around the attempt to define a term of the grammatical metalanguage, "Absolute Construction", as it is applied to the data of Wulfilian Gothic. Theoretical considerations involving the predictability of linguistic structures in generative grammar have traditionally focused on matters of syntax. The model of language generation posited to predict all possible utterances given a complete set of rules is necessarily of a mathematical nature. Semanticians in the meantime have grappled with the limitations of mathematical systems for the description of meaning. Lexical semantics practiced within the theoretical framework of generative grammar explains only one node in a model of language generation. In such a model the lexicon feeds into a placeholder node without affecting the grammatical structure of the generated utterances. Compared with this approach to semantics, cognitive grammar makes a truly revolutionary claim by extending semantic description to the realm of grammar: Grammatical constructions themselves are understood to be meaningful. Lakoff's (1987: 462-585) case study of thereconstructions, Taylor's (1989: 183 196) discussion of grammatical categories and especially of parts of speech, and Smith's (1987) dissertation on the meanings associated with the accusative and dative cases in German are only a few examples of efforts to categorize grammatical constructions according to meaning. Semantics as understood in cognitive grammar thus contradicts and directly interferes with models of generative procedure and provokes a theoretical clash that has not yet been resolved. Definition is at the heart of every mathematical, and by implication generative, system. In Lakoff's (1987: xii) terminology the definitions of traditional grammar are based on an "objectivist view" of categorization

146

Gregor

Hens

which relies on conjunctions of necessary and sufficient, binary and primitive features. Although this general procedure of defining, i. e., delineating, a term is most familiar in traditional lexical semantics, it is implicit in all levels of grammar in the reliance on grammatical terms of the metalanguage. The studies mentioned above show that even for rather "simple" terms it is not always possible to unambiguously decide on the membership of an instance in a category, without some degree of arbitrary demarcation. In cognitive linguistics membership within a class can be a matter of degree, measured as distance from a prototype or viewed in reference to the theory to which a classification refers. Such classification, however, is purely descriptive, and the main goal of a mathematical model of language, namely the predictability of all possible linguistic structures, is simply abandoned. If the definition of a linguistic phenomenon is based on a set of necessary and sufficient features, then the generativist can identify and classify data by referring to those features alone, rather than by revising the classification procedure with the arrival of new and relevant data. The cognitive linguist on the other hand cannot claim to explain data yet to be encountered with the descriptive apparatus developed on the basis of the data at hand at some fixed stage. Instead, a certain amount of ad-hoc-ness is to be expected in describing and explaining data that introduce new parameters relevant to the classification. Thus it seems at times that the cognitive linguist can explain everything and predict nothing. As Lakoff (1987: 96) observes, the means and motivations of a category's extension "make sense of a system", but they do not "generate, or predict" it. The linguistic method of cognitive grammar instead gives guidelines concerning sensible and permitted combinations of characteristics of a category, but it cannot, and does not claim to, replace the diagnostic, i. e., predictive, function that a so-called objectivist view of definition has in decision procedure.

2. Grammatical features for the class of Gothic dative absolute constructions Latin ablative absolutes, Greek genitive absolutes, and Gothic dative absolutes typically are participial constructions with subjects independent of the main clauses in which they are embedded. 1 In Gothic we also find some instances in the accusative, genitive, and possibly nominative cases (cf. Van Helten 1909 and Beer 1911), all of which could possibly be classi-

Gothic absolute constructions

147

fied as free adjuncts. However, unlike the Greek accusative absolute, which occurs with impersonal verbs only, Gothic does not show grammatical restrictions on case variation. In fact, there is no correlation between case and any grammatical property (such as transitivity, aspect, tense) of the participating verb form. Although the typical characteristics of absolute constructions mentioned are important for the task of identifying all instances in the corpus, and thus have prognostic value in determining the membership of new candidates, it is highly unsatisfactory to assign to them the status of necessary and sufficient features in view of the considerable disagreements among scholars concerning the construction and indeed the wide variety of features that has been invoked to classify such constructions in Gothic. Scholars have tallied absolute constructions in the Gothic Bible at ca. fifty to ca. one hundred. To produce closer agreement, one has to select the instances in a principled way. For an exact count, one might establish an exhaustive listing of all instances and weigh arguments for or against each candidate to arrive at a widely accepted compromise, or else one could formulate yet another set of necessary and sufficient features to delineate the field. For both approaches an underlying hypothesis is needed that amounts to a maximum generalization over a large percentage of possible candidates, and that at the same time discards marginal cases for the sake of including some substantial information about the instances selected to form the category. If, for example, [+ DATIVE] were made a necessary feature of the construction, all [ - DATIVE] cases would be left unaccounted for, but the diagnostic value of [+ DATIVE], combined for example with a participle or an independent subject, would be rather high. It is obvious that the recognition of no one feature can suffice to classify a candidate, and that the features themselves are not all equally salient.

3. Definition as a diagnostic tool To what purpose should we define the phenomenon if not to give us a diagnostic tool for the analysis of new data? If the grammarian subscribes to the "generative" rather than the "cognitive" commitment as described in Lakoff (1990), a definition of a grammatical category in combination with a set of rules should yield a "vocabulary" and a "grammar" (corresponding to a set of individuals and well-formedness rules in formal logic) that allow for certain inferences to be drawn by the linguist and

148

Gregor

Hens

the language learner. A biconditional definition of the form (1) below allows for the diagnostic analysis of data according to their membership in some class X, but conversely the observation of data in view of the definition does not indicate the usefulness or appropriateness of the definition. Notice also that some prior assumptions in the form of a working hypothesis underlie the definition, which in itself represents a rather authoritative act of delineation: (1)

χ belongs to the class X if and only if χ has the features a\ and a2 and α 3 ...α η .

According to Lakoff (1990), one has to give priority to either the "generative" or the "cognitive" commitment, yet they are not necessarily mutually exclusive. If we disregard Lakoff's recommendation to choose one over the other, we can modify the form of the definition (1) above to formalize the disjunctive definition needed to account for family resemblance structures, i. e., categories whose members lack a set of common, essential features. Note that I use "or" in its inclusive sense, i. e., the statement "a or b" is true if one or both disjuncts are true: (2)

λ: belongs to the class X if and only if λ: has the features a\ or a2 or a3...an.

If we further abandon the notion that the features (a) are primitive, certain conditions can be linked, i. e., a feature (a x ) might consist of a conjunction of subfeatures. Although this definition appears to be just a minor modification of (1), a definition of the format (2) is less powerful. Even the introduction of a statistical element will not suffice to salvage the clear-cut decision procedure that the classical definition (1) provides. The frequency of a feature in the observed data alone will not allow a probabilistic inference concerning new data that exhibit this feature. The features have to be weighted in order to determine whether certain combinations found in the data exceed some arbitrary threshold set to delineate the class. The weight of a feature depends on its salience in the known data and not on its frequency. Thus, a purely statistical method of determining the weights of the features does not make the above definition a tool for the classification of new data. Instead, it becomes probabilistic only if a general assessment of all available known data produces agreement on the weights of the features according to their salience in the known data. Given a corpus of data and a preliminary classification of a phenomenon, one can assess the probability of membership according to the fea-

Gothic absolute constructions

149

ture weights based on the given data. Thus, we arrive at the conditional probability (3): (3)

Given the data about a\, a2, α 3 ...α η , the probability of a new instance χ belonging to class X is p.

Such an assessment is clearly probabilistic in that it only provides a tool to evaluate newly encountered instances on the basis of the already known corpus. Scientific reasoning according to the Bayesian approach, as outlined in Howson and Urbach (1989) and Earman (1992), accounts for the updating of the given corpus. It provides a formula according to which new information becomes part of the known information, which in turn provides the basis for further evaluation of incoming data. If the ideal learner or scholar reasons according to the probability calculus, the primary and most consequential application of this calculus is Bayes' inverse-probability theorem, given as published by the Enlightenment mathematician Thomas Bayes and subsequently proven by La Salle: (4)

=

P(e/h)p(h) P(e)

It states that the posterior probability of a hypothesis (h), given new evidence (e), equals the probability of the data given the proposed hypothesis (p[e/h]) multiplied by the prior probability of the hypothesis (p[h]), the result then divided by the probability of the evidence (p[e\). This means that, first, we revise our judgment about a hypothesis (here a decision procedure according to which we classify individual candidates of a category) when we encounter new data, a mechanism the traditional definition (1) does not incorporate; and second, that the degree of belief, a subjective probability assignment, changes more radically the more unexpected the new data, or evidence, is. The ideal scholar thus takes into account the probability assessments of all previous candidates and their sets of features when judging a new case. Hence, the hypothesis is flexible and subject to revision and ceases to be a definition in the traditional sense. Instead, it yields a diagnostic procedure that is maximally informed at any given stage in the analysis of the data. 2 In the following I will summarize various approaches to the delineation of the field of Gothic absolute constructions, corresponding to the forms of definition, or "diagnostic tools", given above in (1-3). After briefly examining the decision criteria invoked in various older studies that rely on a conjunction of necessary and sufficient features, as in (1), I will show that purely statistical methods of defining a category according to

150

Gregor

Hens

the form (2) above fail to produce a tool for diagnosing new data unless Bayesian reasoning, as in (3), is invoked to evaluate the salience of features. Bayesian principles provide for an objective assessment of diagnostic value without reliance on statistical methods. Edward, Lindmann and Savage claim that "two people with widely divergent opinions but reasonably open minds will be forced into arbitrarily close agreement about future observations by a sufficient amount of data" (cited in Howson-Urbach [1989: 236]). If this holds, then an authoritative, socalled Aristotelean definition as part of a mathematical system might not be needed for the establishment of "objective" grammatical concepts; indeed, its price might be too high.

4. Gothic absolute constructions: Traditional accounts All Gothic grammars in some way acknowledge the existence of absolute constructions, although there is no traditional place for their treatment. Both Grimm in the Deutsche Grammatik (1837: 893 919) and Van der Meer (1901: 116-120) discuss such constructions in appendices following their discussions of the prepositions that govern oblique cases. Wright in his Grammar of the Gothic language (1954: 194) devotes half a paragraph at the end of his description of the syntax to dative absolutes only. Streitberg (1900: 142) emphasizes the participle rather than the case when he refers to "die Partizipialkonstruktionen, die man mit dem Namen Dativus Absolutus bezeichnet". In view of such an account we are surprised to find that Metlen (1938: 632) does not hold the participle to be a necessary feature of the construction at all. Ever since Jacob Grimm classified the constructions he considered absolute, scholars of Gothic have listed all occurrences and then argued about particular marginal cases, such as nominative or genitive absolutes. There seems to be agreement on a method that assigns central position to Gothic instances which somehow correspond to Greek or Latin equivalents or stereotypes, and then measures the marginal cases according to their approximation of the thus established "good" examples. Wrede (1920: 382) reads under "Participia": "Wie das Lateinische seinen Abl.[ativus] absol.[utus], das Griechische seinen Gen.[itivus] absol.[utus], hat das Gotische seinen Dat.[ivus] absol.[utus]." This is followed by Gothic examples and their Greek originals. Nowhere do we find a definition that bases class membership on consistent criteria in light of the given data. Three closely related reasons can account for this fact:

Gothic absolute constructions

151

1. The notion we have of the nature of the construction is most likely to have originated in encounters with Latin, usually before we become acquainted with Germanic dialects. (5) and (6) are typical textbook examples. (7) is one of the less controversial cases in Gothic. It is not introduced by the preposition at, the subject of the construction (im 'they', [DATIVE]) is overt and differs from that of the main clause ('he') and both the subject and the participle have dative case. Only the position of the construction (following rather than preceding the main clause) is unusual and deserves further notice. 3 (5)

Stellis visis, nautae gaudent. 'The stars having been seen, the sailors rejoiced.'

(6)

Cicerone consule, magnum erat periculum Romae. 'When Cicero was consul, the danger at Rome was great.'

(7)

Jah bipe warp twalibwintrus, usgaggandam pan im in Iairusaulyma bi biuhtja dulpais. (Lk. 2, 42) 'And when he was twelve years old, they went up to Jerusalem after the custom of the feast.' 4

2. Because cases tend to be unmarked and subject nominals missing, Modern English and German equivalents, as for example (8-10), are not easily identifiable (cf. also Grimm 1837: 908-911). It is, indeed, an issue whether they are comparable at all. (8)

Gesagt, getan. (Erben 1980: 308)

(9)

Stehenden Fußes, ...

(10)

Walking home, Joe found a dollar.

3. The classification is a hybrid of folk and expert category. A beginning student of Latin will use a given definition to identify members of a category he or she considers real and natural, while it is the expert's task to provide the essence of a number of instances grouped together for purposes of simplification and generalization. The Latin student's impression, then, reflects a mixture of preliminary rules based on observed data with a simplified version of the expert category, a phenomenon reminiscent of Selinker's (1972) concept of "interlanguage" in foreign-language pedagogy. Holland cites Serbat's so-called "grammatical vulgate", which is compiled from Latin textbooks, to sum up the folk theory used to identify absolute constructions in Latin:

152

Gregor

Hens

a) they are used only when the subject of the absolute construction is different from that of the main clause, i. e., there cannot be any overt pronominal reference between the absolute construction and the main clause; b) the absolute construction is equivalent to a circumstantial subordinate clause, usually with temporal or causal meaning, but often too with concessive meaning; c) the term 'absolute' is understood as 'not being grammatically connected with the rest of the sentence'; d) all absolute constructions are grouped together and discussed in the same manner, whether they are active or passive, present or past. (Holland 1986: 165)

While it is rather difficult to extrapolate decision criteria or singly necessary and jointly sufficient features, we recognize similar notions in both old and modern descriptions. As a correlate of (c) above, for example, Grimm refers to "...nomina [, die] sich participia zugesellen und mit ihnen in den Satz, ohne von dessen construction berührt zu werden, eintreten, ..." He continues with a correlate of the equivalence condition (b) above: "... im ersten fall ergeht die Verwandlung in den absoluten casus weit leichter [...] im anderen fall hat der absolute casus größere kühnheit und stärke." (Grimm 1837: 894, my emphasis). Metlen (1938) and Costello (1980) both stress that an absolute construction is an abridged adverbial clause that "always can be expanded into a fullfledged dependent clause, usually of time" (Metlen 1938: 632), but they do not explore the function associated with such an abbreviation (cf. note 3). Such general characteristics are then called upon to decide membership in cases that share some features with the stereotype while they lack some other salient feature, usually the dative case. Metlen, for example, considers a rare genitive absolute: (11)

J ah in wisandins sabbat e dagis Mar ja so Magdalene jah Mar ja so Iakobis jah Salome usbauhtedun aromata,... (Mk. 16, 1) 'And when the sabbath was past, Mary Magdalene, and Mary the mother of James, and Salome, had bought sweet spices,...'

Metlen argues for the inclusion of (11) on the following grounds: And there is to my mind no valid reason to doubt the absolute nature of the Gothic construction for it has all the characteristics of an absolute construction by (1) being separated grammatically from the rest of the sentence, (2) having the meaning of an adverbial modifier of the main clause, and (3) being readily turned into a dependent clause of time. (Metlen 1938: 635-636)

Gothic absolute constructions

153

However, he leaves open what constitutes grammatical separation when he states: "The active subject of the a.[bsolute] c o n s t r u c t i o n ] may be identical with that of the main clause,..." (Metlen 1938: 631). Furthermore, while the notion prevails that a participle is a near-necessary part of the construction, as Grimm (1837: 893) had claimed earlier, Metlen (1938: 632) cites a Latin example similar to (6) above and points out: "Sometimes a noun may take the place of a participle, as in Caesar consule..." Grimm (1837: 894) insists on an overt subject to govern the participle and holds only two constructions (Lk. 7, 42 as in [12] and Mk. 8, 1 as in [13]) to be exceptions because they both have a dative nominal group in the main clause such that a dative pronominal form, lost in scribal tradition, could be posited as a subject of the absolute construction. 5 Both Metlen and Van der Meer, however, rule out any such case on the grounds that they lack grammatical independence. Instead, (12) and (13) are considered appositions: (12)

Ni habandam pan hwapro usgebeina, baim [DATIVE] fragaf. (Lk. 7, 42) 'And when they had nothing to pay, he frankly forgave them both.'

(13)

In jainaim pan dagam aftra at filu managai wisandein jah ni habandam hwa matidedeina, athaitands siponjans qapuh du im [DATIVE]. (Mk. 8, 1) 'In those days the multitude being very great, and having nothing to eat, Jesus called his disciples unto him, and sayeth unto them.'

According to my own listing there are exactly fourteen such cases, where we could, with the same justification, posit a dative pronoun, as Grimm does for (12). He supplies im ('they' [DATIVE]) (Grimm 1837: 897). All these cases would have to be discarded if an overt subject were made a singly necessary condition for absolute constructions. Metlen (1938: 631) rightly observes that "some critics waver as to where to draw the exact demarcation line..." (my emphasis).

5. Progress in categorization Returning to the theoretical considerations outlined in the beginning, I would like to show how one would proceed to establish a definition based

154

Gregor

Hens

on a disjunction of features. I will conclude that the statistical information known about grammatical features of the construction at any given stage will not suffice to allow classification of a candidate. Instead, scholars and language learners arrive at an objective assessment of the salience of a feature which is not directly tied to its frequency in the data. A set of features thus assessed will then serve as a guide for the classification of new data. Such assessments, by different scholars tend to concur with an increasing amount of data. The actual treatment of the absolute constructions in grammars and articles, summarized above, indicates, however, that the process of theory formation according to the laws of probability is a highly idealized notion which is not always visibly at work here. In fact, if one reads the scholarly discussion on the topic since Grimm, it is difficult to see that much progress has been made at all, and there seem to be relapses into older errors. What we would like to see, of course, is a steady progression of the selection principles for the class of absolute constructions whenever new data become available, i. e., in this case, whenever a scholar points out a relevant candidate that had not before been noticed. This progression would lead to a categorization of maximum specificity and scope, i. e., to a substantial generalization over a class of instances meaningful enough to justify its having a name. Such a balance is maximally efficient for scholarly discourse. 6 In sciences, categorization proceeds in a similar way, although it is frequently claimed that the objective nature of a scientific classification is derived from natural classes rather than shaped to aid efficient communication among scientists. (For a discussion of biological species cf. Lakoff [1987: 185-195]). The only phenomenon clearly in accordance with a Bayesian view of theory formation is the fact that Grimm's analysis takes the biggest step towards establishing the classification, since he had the most new data to work with; but the linear development towards a more efficient concept "absolute construction" does not proceed as expected for the following reasons. As pointed out above, Latin and Greek stereotypes and the "grammatical vulgate" interfere with rational theory formation based on the given data. Accordingly, listings which try to classify the Gothic data vary in a way that precludes the extraction of a core of singly necessary properties. Eda (1988: 56), for example, lists ninety-four absolute constructions, Grimm (1837: 893-919) seventy-five, while Van der Meer (1901: 118-120) does not recognize the class at all and instead classifies such constructions as (1) "dativus als persoonlijk object" [dative as personal object]; (2) "De

Gothic absolute constructions

155

meer vrije datief' [the freer dative]; (3) "De dativus als loc. en instrum". [the dative as locative and instrumental]. The following example (14) shows just how little extremely marginal cases have in common with what Metlen (1938: 633) calls "the rank and file of absolute constructions": (14)

habai mik faurqipanana (Lk. 14, 18) Ί pray thee hold me excused'

Eda (1988: 54) includes this imperative in her statistics, although she does not accept it in her discussion, on the grounds that "the participle ... can in no way be considered equivalent to that of the participle in other absolute constructions..."

6. Weighted features Let us proceed, then, by asking whether our intuitive confidence in terminological and classificatory progress is justified at all. In order to translate the discussion surrounding these constructions into a model of an idealized process of theory formation, it is necessary to identify more exactly how Bayesian updating applies to the problem. I therefore consider an approach of a purely statistical nature that fails to establish a definition agreeing with our intuition about an efficient categorization, and thus does not provide a valid diagnostic tool. I then discuss ways to salvage the diagnostic value of this approach. In Categories and concepts (1981), the authors Smith and Medin summarize early attempts to explain prototype effects in semantics, among them a weighted feature analysis, which they call probabilistic. Assigning values to properties within a given class based on frequency, they fail to clearly distinguish between the frequency of a feature and its salience: "The features that represent the concept are salient ones that have a substantial probability of occurring in instances of the concept." (Smith-Medin 1981: 62). The reason for this is that Smith and Medin consider the concept as prior to the evaluation of the features. They too presuppose a natural class. If, however, features are to be evaluated to establish the concept, they are initially as salient as they are frequent and their frequency alone does not suffice to assess the probabilities of their occurrence in new instances. I would like to show that the two factors, frequency and salience, are clearly distinct. Relying on an immeasurable amount of additional information, a community of scholars arrives at the correct diagnostic value of a feature that is clearly distinct from its

156

Gregor Hens

frequency. In Table 1 I list some major diagnostic attributes of the grammatical category "dative absolute construction", the number of occurrences and their relative frequencies. I have counted ninety-eight cases proposed in the literature that have a participle and an implicit or explicit subject that differs from the subject of the main clause. Notice the high frequency of the attribute 3 [ACTIVE] versus the lower frequency of attribute 6 [DATIVE]. I find no grammatical reason in the data to exclude such obviously genre-dependent features as [3RD PERSON] and [SINGULAR], Table 1. Features of Gothic absolute constructions and their relative frequencies feature

relative frequency

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

1.0 1.0 .93 .91 .89 .86 .86 .82 .79 .72 .71 .62 .57

PARTICIPLE I N D E P E N D E N T SEMANTIC SUBJECT ACTIVE TEMPORAL, CAUSAL 3 R D PERSON DATIVE OVERT GRAMMATICAL SUBJECT PRESENT TENSE INTRANSITIVE NO PREPOSITION (at) SINGULAR NO CASE COHESION WITH MAIN CLAUSE PERFECTIVE ASPECT

Serious doubts about such a listing stem less from the mixture of semantic and grammatical features or their dubious status as nonpromitives than from their different salience. After all, we frequently find the name "Dative Absolute" for the whole category, but never "Active Absolute". Given a quantitative weighted feature analysis, the definition chosen to represent the essence of the class would be evaluated as the sum of the relative frequencies of its features, representing specificity, multiplied by an adjusted factor that would designate its scope. In practice, we can establish a function that leads from the scope of the definition to its maximum specificity, i. e., ideally this function would provide the optimal combination of features for any set number of cases that we wish to cover with the definition. A conjunction of characteristics 1, 2, 3, 4 and 6 in Table 1, for example, covers seventy-five cases, and, although other combinations of features (of lower frequency) might also

Gothic absolute constructions

157

define a class of seventy-five cases, this one is the most meaningful according to the ranking, i. e., according to the sum of the relative frequencies of these five features. However, the ranking of the features listed according to frequency in Table 1 does not correspond to our intuition about their salience, because our assessments are based on an immeasurable amount of information (including knowledge about the genre of the text, comparison with similar phenomena in other languages, knowledge of the Greek source and more, all available to scholars as a community) that cannot easily be statistically evaluated and is updated with any incoming data.

7. Salience assessments of features Smith and Medin's (1981) analysis of weighted features rests on an objectivist statistical method which relies on arbitrary thresholds. As an alternative the Bayesian account proposes that the updating of a hypothesis of the type (3) above, given new evidence, proceeds according to probability calculus. It ignores, or sanctions, the subjective nature of the prior probabilities entering the equation to prove a mechanistic system of updating that is rational and objective, and in the long run predictable, given certain evidence. As Howson and Urbach (1989) show, bookmakers and their clients behave exactly according to such a mechanism of updating based on originally subjective assessments of odds and quotients. Again considering the grammatical characteristics in Table 1 we can now state that while, for example, the frequency of attribute 3 [ACTIVE] is much higher than that of attribute 6 [DATIVE], we can still consider [DATIVE] to be essential rather than [ACTIVE], because in our judgment of salience we rely on a large amount of evidence, including background knowledge, context, and a history of updated hypotheses. Tversky and Kahneman (1983: 294) propose that Bayesian probability assessments compete with other factors in the majority of judgments made in everyday life, which they call "natural assessments". These factors include "computations of similarity and representativeness, attributions of causality and evaluations of the availability of associations and exemplars". Such an exemplar would be the stereotype of Latin ablative absolutes which should ideally not interfere with scientific reasoning. Lakoflf, following Tversky and Kahneman, concludes that objective salience evaluation is affected by subjective factors: "In short, a cognitive model may function to allow a salient example to stand metonymically

158

Gregor Hens

for a whole category. In such cases, our probability judgments about the category are affected." (Lakoff 1987: 90). Thus while he discards rational intuition, or Bayesian reasoning, he nevertheless relies on it in the form of betting when forming a hypothesis about a category in Dyirbal: "If I had to make a bet (a small bet) I would bet that boys are more central than bats." (Lakoff 1987: 102). While rational decision-making on a subjective basis, for example when placing a bet, seems highly unsatisfactory, it is, in fact, the principle that governs much of our theory-forming.

8. Conclusion I set out to find the most meaningful definition of the category "Gothic absolute constructions" as an instrument for prediction, yet I find myself in a dilemma: Bayesian reasoning allows me to trust my intuition in evaluating diagnostic attributes for the classification of specific instances of a field of data. If rational, I will, step by step, arrive at an efficient categorization while bypassing a purely quantitative analysis of the data. Rational updating, however, is based on calculus that, according to Lakoff (1987: 175-84) is not legitimate in linguistic investigation, and that, according to Tversky and Kahneman (1983: 313 and elsewhere) has little if any relevance for human cognition. The theoretical clash that becomes apparent in this attempt to design a diagnostic tool for future data while abandoning the form of definition given under (1) above is the clash of the "cognitive" vs the "generative" commitment (Lakoff 1990), or the methodological clash between social vs "hard" sciences. It seems largely irreconcilable. I hope to have shown, however, that the "softening" of science in recent years (exemplified by Bayesian reasoning) acknowledges a human, subjective factor in so-called scientific objectivism that recommends it to the linguist who is interested in generalization beyond pure description. Notes 1. With a lower frequency than in the Gothic data, Old High German also shows some dative-absolute constructions. 2. It is by no means uncontroversial to discuss linguistic method in terms of scientific reasoning. However, one of the plausibility arguments in favor of Bayesianism asserts that scientific investigation does not substantially differ from other rational endeavors. Earman (1992: 204) writes: "... methodologists are wasting their time in searching for a demarcation criterion that will draw a bright red line between science and nonscience in

Gothic absolute constructions

159

terms of the methodology of belief formation and validation, for it is just all Bayesianism through and through, whether the setting is the laboratory or the street. What does demarcate science as it is now practiced is the professionalized character of its quest for well-founded belief." Kuhn's ([1970]) approach to the philosophy of science still dominates the discussion about linguistics as a science. Specifically, there is much talk about the incommensurability of the generative and cognitive "paradigms". While Earman (1992: 187-205) convincingly argues for the incompatibility of Kuhn's philosophy of science with Bayesianism, even Kuhn's followers should reconsider the labelling of schools of linguistics as paradigms. According to Kuhn's characterization of the pre-paradigm stage of science, linguistics still is in this stage, much like "the study of motion before Aristotle and of statistics before Archimedes, the study of heat before Black, of chemistry before Boyle and Boerhaave" (Kuhn [1970]: 15). "The pre-paradigm period, in particular, is regularly marked by frequent and deep debates over legitimate methods, problems, and standards of solution, though these serve rather to define schools than to produce agreement." (Kuhn [1970]: 47^18) Kuhn in his famous postscript ([1970]: 208) also hesitates to endorse fully his followers' attempts to apply his theory to "other fields". 3. Twenty-seven out of the twenty-nine absolute constructions I consider in the Luke passages alone precede the main clause. They mostly serve as temporal or causal backgrounding (cf. my Table 1), and thus belong to the theme in the information structure. This "meaning" of the construction is a candidate to be considered when identifying a group of typical instances. 4. English translations are taken from the authorized Bible edition commonly known as the King James Version. 5. "... nur ausnahmsweise dulden einige lat. und deutsche formein die weglassung des subjects, z.b. comperto ..., nhd gesetzt, kaum gesagt, anders ausgedrückt." (Grimm 1837: 894-895) 6. Such "efficient" categorization calls to mind basic level effects as summarized in Lakoff (1987: 46-48). In the case of grammatical categories the balance between specificity and scope would be determined by the needs of the scholarly community to identify an instance quickly and with a low rate of error.

References Beer, Anton 1911 "Gab es einen Gotischen Nominativus Absolutus?", Beiträge zur Geschichte der Deutschen Sprache (PBB) 37: 169-171. Costello, John R. 1980 "The absolute construction in Gothic", Word 31: 91-104. Earman, John 1992 Bayes or bust? Cambridge, Mass.: MIT Press. Eda, Yoko 1988 "A study of Gothic absolute constructions", Gengo Kenkyu 93: 39-60. Erben, Johannes 1980 Deutsche Grammatik. München: Hueber. Grimm, Jacob 1837 Deutsche Grammatik. Vol. III (Syntax). Göttingen: Dietrichsche Buchhandlung.

160

Gregor Hens

Holland, Gary Β. 1986

"Nominal sentences and the origin of absolute constructions in Indo-European", Zeitschrift für vergleichende Sprachforschung 99: 163-193. Howson, Colin—Peter Urbach 1989 Scientific reasoning. The Bayesian approach. La Salle: Open Court. Kuhn, Thomas S. 1962 The structure of scientific revolutions. Chicago: University Press. [1970] [Second edition, enlarged. Chicago: University Press] Lakoff, George 1987 Women, fire, and dangerous things. What categories reveal about the mind. Chicago: University Press. 1990 "The Invariance Hypothesis: Is abstract reason based on image schemas?", Cognitive Linguistics 1: 39-74. Metlen, Michael 1938 "Absolute constructions in the Gothic Bible", PMLA 53: 631-644. Selinker, Larry 1972 "Interlanguage", International Review of Applied Linguistics in Language Teaching 10: 209-231. Smith, Edward E . - Douglas L. Medin 1981 Categories and concepts. Cambridge: Harvard University Press. Smith, Michael B. 1987 The semantics of dative and accusative in German: An investigation in cognitive grammar. [Unpublished Ph. D. Dissertation, University of California, San Diego.] Streitberg, Wilhelm 1900 Gotisches Elementarbuch. Heidelberg: Carl Winter. Taylor, John R. 1989 Linguistic categorization. Prototypes in linguistic theory. Oxford: Clarendon. Tversky, A m o s - D a n i e l Kahneman 1972 "Subjective probability: A judgment of representativeness", Cognitive Psychology 3: 430-454. 1983 "Extensional versus intuitive reasoning: The Conjunction Fallacy in probability judgment", Psychological Review 90 (4): 293-315. Van der Meer, Marten Jan 1901 Gotische Casus-Syntaxis 1. Leiden: Brill. Van Helten, Willem Lodewijk 1909 "Gab es einen Got. Nominativus Absolutus?", Beiträge zur Geschichte der Deutschen Sprache (PBB) 35: 310-311. Wrede, Ferdinand (ed.) 1920 Stamm-Heynes Ulfilas oder die Uns Erhaltenen Denkmäler der Gotischen Sprache. Paderborn: Schöningh. Wright, Joseph 1954 Grammar of the Gothic language. Oxford: Clarendon.

Subject sich and expletive pro: Impersonal reflexive passives in German Robert G. Hoeing

1. Impersonal passives and the expletive empty category The Extended Projection Principle of Government-Binding theory stipulates that every clause must have a subject. 1 Impersonal passives such as (1)

Gestern wurde (*es) getanzt Yesterday was (*it) danced 'There was dancing yesterday' Wurde (*es) gestern getanzt? 'Was there dancing yesterday?' daß (*es) gestern getanzt wurde 'that there was dancing yesterday'

apparently contradict the Extended Projection Principle, since the expletive es can occur only sentence-initially, in the specifier-position (SPEC) of C" — hence in a nonargument, non-O-position, as one would expect, and never in subject position: (2)

Es wurde getanzt [C"[SPEC