Focus on Germanic Typology 9783050084336, 9783050041063

Citation preview

Focus on Germanic Typology

Studia typologica EDITED BY THOMAS STOLZ

Beihefte / Supplements STUF - Sprachtypologie und Universalienforschung Language Typology and Universals VOLUME 6

Focus on Germanic Typology Edited by Werner Abraham

Akademie Verlag

ISBN 3-05-004106-4 ISSN 1617-2957 © Akademie Verlag GmbH, Berlin 2004 Das eingesetzte Papier ist alterungsbeständig nach DIN/ISO 9706. Alle Rechte, insbesondere die der Übersetzung in andere Sprachen, vorbehalten. Kein Teil des Buches darf ohne Genehmigung des Verlages in irgendeiner Form - durch Photokopie, Mikroverfilmung oder irgendein anderes Verfahren - reproduziert oder in eine von Maschinen, insbesondere von Datenverarbeitungsmaschinen, verwendbare Sprache übertragen oder übersetzt werden. All rights reserved (including those of translation into another languages). No part of this book may be reproduced in any form - by photoprinting, microfilm, or any other means - nor transmitted or translated into a machine language without written permission from the publishers. Druck und Bindung: MB Medienhaus Berlin GmbH Printed in the Federal Republic of Germany

Table of contents

WERNER ABRAHAM & G E R T W E B E L H U T H

Words of dedication for Hartmut Czepluch

7

WERNER ABRAHAM

Introduction

9

JOHN H . M C W H O R T E R

What happened to English?

19

HALLDOR ÄRMANN SIGURDSSON

Agree and agreement - evidence from Germanic

61

JOHANNA BARDDAL

The semantics of the impersonal construction in Icelandic, German, and Faroese: beyond thematic roles

105

CEDRIC BOECKX & KLEANTHES K . GROHMANN

Left dislocation in Germanic

139

C . JAC CONRADIE

Verb sequence and placement: Afrikaans and Dutch compared

153

HARTMUT C Z E P L U C H T

Reflections on the form and function of passives in English and German

169

MOLLY DIESING

The upper functional domain in Yiddish

195

BRIDGET DRINKA

Präteritumschwund: evidence for areal diffusion

211

6

Table of contents

WERNER ABRAHAM

The European demise of the simple past and the emergence of the periphrastic perfect: areal diffusion or natural, autonomous evolution under parsing facilitation?

241

LÄSZLO MOLNÄRFI

Some remarks on the formal typology of pronouns in West Germanic

273

ROLF THIEROFF

The subjunctive mood in German and in the Germanic languages List of authors

315 359

Words of dedication for Hartmut Czepluch WERNER ABRAHAM

(Wien) & GERT WEBELHUTH (Göttingen)

The present collection of essays on systematic properties of Germanic languages is dedicated to HARTMUT CZEPLUCH. Hartmut died, quite unexpectedly for all of us, on May 7, 2003. GERT WEBELHUTH, chair of English Linguistics at the University of Göttingen, writes this: "The members of the Chair of English Linguistics mourn the premature death of their friend and colleague Dr HARTMUT CZEPLUCH, who died on May 7 after serious illness at the age of 58. Hartmut had been associated with the Chair of English Linguistics at Göttingen for many years and completed both his dissertation and his Habilitation here. When he died, he was employed as a permanent member of its teaching and research staff. Through his work as a formal syntactician and later in life about the relationship between grammar and discourse, Hartmut became well known internationally. His work on Case and the double object construction has been widely cited in the research literature and standard introductory textbooks to syntactic theory. We will all miss Hartmut's gentle friendship, his humble dependability and professionalism, and most of all the infecting enthusiasm with which he dedicated himself to the linguistic profession." GERT WEBELHUTH

(for all members of the Chair of English Linguistics)

WERNER ABRAHAM, long-time friend of Hartmut's and colleague in the field, mourns the loss of a rare independent intellectual: "Ich traure um HARTMUT CZEPLUCH, nicht nur weil ich zutiefst erschrocken bin, ihn noch als Weggefährten sehe, nicht so viele Jahre jünger. Ich erblickte in ihm auch einen geistig Nahen, der sich als einer der ganz wenigen im modernen Fach nicht selbstprofilierenden Scheinproblemen oder oberflächlichen Fleißaufgaben verschrieb, der nicht nur Thesen nummerierte, sich dabei auf bloße Vordenker bezog und aus deren gut geölten Gedankenspuren nur unwesentlich ausscherte. Er gehörte weder zu denen, deren intellektuelle Evolution nie über das Sammeln hinausführte, noch zu jenen, die sich nur in purer theoretischer Nachdenkerschaft geborgen fühlten. Er war vielmehr einer, der als gelernter Anglist und Kenner generativer und anderer formallinguistischer Denkmuster das Deutsche gerade nicht in den Schnürleib von Analysen preßte, die bloß dem Englischen angemessen waren, und damit dem

8

WERNER ABRAHAM & GERT WEBELHUTH

Deutschen empirisch-typologischen Zwang antat. CZEPLUCH sah - wie sein in diesem Band aufgenommener Aufsatz zeigt - wie anders das Deutsche im Konzert der germanischen Sprachen analytisch anzugehen ist: was dessen Diskurskonfigurationalität und grundsätzliche Vertextungstaktik betrifft, was die Kasussystematik etwa im Unterschied zum gleicherart konservativen Isländischen betrifft. Er entwickelte unabhängig davon, ob dies zu Mainstreamakzeptanzen führen konnte, originelle Lösungsansätze zu den Ditransitiven, die internationale Beachtung und Anerkennung erfuhren. Seine berufliche Karriere war geprägt von gelassener intellektueller Unnachgiebigkeit gegenüber seinen Beurteilern, von anspruchsvoller Unpragmatik. Trotzdem wollte er nie nachtragend oder bitter sein. Dieses kompromisslose Verhalten führte ihn durch viele Institutionen im Lande. Ich mußte im Laufe der Jahre meine Adressenkartei unter seinem Namen oft ändern. Was ihm dabei Kraft verlieh, gestand er mir einmal, war sein Zuhause in Göttingen. Es war ihm noch beschert, dort seine letzte Stelle zu finden, vor dem eben entfalteten Horizont in seinem akademischen und privaten Zuhause seine Laufbahn zu beenden. Hartmuts letzte Nachricht erreichte mich am 30.8.2002. Sie betraf seinen Beitrag im vorliegenden Studio Typologica-Band: «Lieber Werner, natürlich habe ich den Beitrag nicht vergessen. Anbei schicke ich Dir die Version, die ich bis Ende Oktober fertig gestellt hatte. Im November bin ich leider nicht mehr dazu gekommen, noch weitere Änderungen einzubauen (wobei ich eigentlich v.a. die Zusammenfassung noch prägnanter hätte fassen wollen). Ich bin am Donnerstag gerade mal wieder aus der Klinik entlassen worden - vorläufig. Zu meiner Diabetes und Gastritis hat sich herausgestellt, daß auch die Bauchspeicheldrüse und eventuell die Leber nicht in Ordnung sind. Seitdem hatte ich fast ausschließlich mit diversen ambulanten Untersuchungen und zwei einwöchigen Klinikaufenthalten zu tun. Und da die Ärzte immer noch keine eindeutigen Befunde haben, wird das wohl noch eine Zeit so weiter gehen. Nach jeder Untersuchung fühle ich mich ein wenig kränker, so daß ich völlig damit ausgelastet bin, wenigstens meine Lehrveranstaltungen einigermaßen aufrechtzuerhalten. Ich hoffe, Du kannst den Beitrag auch in der vorliegenden Form akzeptieren. Aber das wirst Du besser beurteilen können. Für heute: Herzliche Grüße, Hartmut.»" WERNER ABRAHAM

Introduction WERNER ABRAHAM (Wien)

The papers contained in this issue were convoked in January 2002 upon invitation extended to me by the editor-in-chief of Studia Typologica. The topic remained unchanged: Focus on Germanic typology. The idea was not to describe typologically, in a modern quasi-formalized fashion, a whole language. Such books were on the market already, viz. VAN DER AUWERA & KÖNIG 1995. Rather, the double issue was to feature specific phenomena characteristic of one Germanic language in question and set these phenomena off against functional or structural equivalents in one or more other Germanic languages. Diachronic perspectives could be included, but should not be the focus of the contribution. As an example I quoted an investigation of my own on prepositional indirect objects in the Scandinavian languages, in English, and in Dutch compared with non-prepositional equivalents in German (ABRAHAM 2000, 2004). The general insight developed was that the non-prepositional equivalents in German are intensionally wider than their prepositional parallels in the other Germanic languages. However, the respective prepositions are not on a par with respect to intensional weight. English to can be said to be a grammaticalized indirect object preposition with a lot less intensional-semantic weight than other prepositions in English. There are interesting, non-trivial links between preposition and lexical intension of the P-governing verb, all of which are lost in non-prepositional German. Generally, this point of comparison belongs to differential case marking including most prominently subject marking in Icelandic (quirky subjects) vs. equally rich case-morphological German. How come oblique case subjects can be subjects in the first place on an explanatory, theoretical level? It was shown that the decisive criterion is the base structure difference between Icelandic and German. German projects the functional domain inside VP, whereas Icelandic does so outside of VP (ABRAHAM 2004). The present collection pursues this course in great detail for specific aspects of the following languages: Afrikaans, Danish, Dutch, English, Faeroese, German, Icelandic, Norwegian, Swedish, and Yiddish. To some extent also regiolects and dialectal phenomena (such as South German, i.e. Austrian, Bavarian, Swiss German) were incorporated. There are two large domains which are looked into in great detail: morphology and its syntactic functions; and syntax proper and its relation to semantics. In What happened to English?, JOHN H. McWHORTER compares both morphological and constructional properties of almost all Germanic languages to support the claim that, although the loss of inflection is considerable in comparison, it may just be the tip of the iceberg in terms of Germanic features that English has shed.

10

WERNER ABRAHAM

What MCWHORTER shows also is that this morphological demise is complemented by many other losses unconnected to morphological analyticity. Overall, a comparison with its sisters reveals English to be significantly less overspecified semantically and less complexified syntactically. Thus, beside the point that English departs considerably from the Germanic template, the author establishes that this general picture was not one of chance. The gist of the paper is the rebuttal of the much discussed claim that English became what it is as a consequence of creolization between Scandinavian, Anglo-Saxon, and Romance. A discussion of historical facts reveals that a contact-based, linguistically external explanation does not provide a principled account for the relevant facts. Rather, the morphological, and as a consequence the functional, erosion took place already in the source languages of Early Modern English. HALLD0R ÄRMANN SlGURBSSON's Agree and agreement - evidence from Germanic surveys morphological agreement phenomena in the Germanic languages. In addition it studies the relation of overt agreement with the underlying LF relation of Agree. It is argued that CHOMSKY'S Probe-Goal Approach is not well suited to account for the nature of abstract Agree, although it is descriptively adequate for some instances of morphological agreement. The central claim of the paper is that Agree reduces to Merge, i.e. it is a precondition on Merge (and an integrated part of it). Thus, whenever Merge applies, the possibility of agreement arises, i.e. a language has to make a parametric choice whether or not to signal each instance of Merge/Agree morphologically, hence the extreme variation of agreement across languages, even within a relatively limited and a closely related group of languages, such as the Germanic ones. - The claim that Agree reduces to Merge is coined as the Agree Condition on Merge. It is claimed that this simple condition is a law of nature, hence of language. In addition, it is suggested that Move is driven by the needs of Merge/Agree, moving features to the edge of a category Y, such that the edge matches some features of the object with which Y merges. The approach pursued also supports the view that labeling and X'-theoretic conceptions are theoretical artifacts that should be dispensed with. JOHANNA BARDDAL compares in The semantics of the impersonal construction in Icelandic, German, and Faroese: beyond thematic roles the semantics of verbal predicates selecting for non-nominative (logical) subjects in Icelandic, German and Faroese. Hitherto, the explanation employed to account for their case assignment relies on the concept of thematic roles, such as Experiencer and Goal (Beneficiary). A more thorough investigation of the semantics of these predicates reveals subclasses expressing "events" not based on internal experience at all. The analysis presented in this paper, however, relies on concepts such as affectedness, shared semantics, pragmatic inferences and subjectification. In order to account for the case marking of impersonal predicates in Germanic two levels of relations need to be assumed: one profiling the semantic relation between the (logical) subject and the predicate, and another profiling the emphatic relation between the speaker and his/her attitudes towards the content of the proposition. As to CEDRIC BOECKX & KLEANTHES K . GROHMANN'S Left dislocation in Germanic·. Left dislocation constructions seem to come in very different shapes with very different properties across languages. Concentrating on the shapes occurring

Introduction

11

in Germanic a unified movement analysis is argued for which distinguishes contrastive left dislocation from hanging topic left dislocation in that only the former involves Agree on top of Match between the left dislocate and the resuming pronominal. Under such an approach, the variety in shapes and properties observed breaks down to a tight similarity from which the diverging patterns fall out in a straightforward fashion. As to C. JAC CONRADlE's Verb sequence and placement: Afrikaans and Dutch compared: Though Afrikaans still maintains clause-final clustering of verbs, as is the case in Dutch, and is therefore still basically an OV language, verbs are increasingly found in earlier sentence positions, apart from VI and V2. Some of the causes are, as in Dutch, but probably to a larger extent, that the final string in general usage is often penetrated by non-verbal material, complementizers previously selecting strict OV order are increasingly followed by main clause order, and linking and main verbs often "fuse" to co-occur in VI and V2 positions without any clause-final stranding of verbs. However, since clause-final verb strings are also maintained in Afrikaans and a typological change to VO order has not been completed, the rift between OV and VO orders seems to have deepened. While the clustering of verbs in clause-final position is still prevalent in Afrikaans in subordinate clauses as well as - barring Vl/2 positions - in main clauses, and the partial grammaticalization of the clause-final auxiliary het 'have' would suggest a basic OV order, several processes bring about a looser spread of verbs throughout the clause than is the case in Dutch, viz. the increasing use of VO subordinate clauses with or without complementizers, main clause order in embedded WH-clauses, the far-reaching penetration of final verb clusters by non-verbal material and, finally, the process of verbal fusion. This would imply that the structural rift between VO and OV orders, also present in Dutch and German, has widened in the case of Afrikaans. HARTMUT CZEPLUCH's Reflections on the form and function ofpassives in English and German takes up passivization from both formal and functional perspectives in order to shed new light on the issue with interesting results for comparative linguistics. In the bulk of the literature, passivization in English and German is treated as more or less the same grammatical phenomenon. If this is the appropriate cross-linguistic view on passivization, one would expect that passives have the same textual functions (more or less) in different languages. Nevertheless it is wellknown from various textual investigations that passives are not evenly distributed across languages: there are quite a few situations were one language prefers the passive over the active, while another language shows the reversed preference. English and German are two such languages. The paper discusses some aspects of why it might be the case that passives have different textual distributions in the two languages. MOLLY DiESlNG's The upper functional domain in Yiddish examines the structure of the left edge of the clause in Yiddish and its consequences for theories of clausal structure and functional categories. Data from verb-second, wA-questions, and topicalization structures motivate an approach which aligns itself with claims that the functional structure of the clause may in fact vary cross-linguistically.

12

WERNER ABRAHAM

In BRIDGET DRINKA'S Präteritumschwund: evidence for areal diffusion it is argued that the development of the demise of the simple preterit in Southern German and several other languages of Western Europe was not primarily motivated by principles of parsing and discourse, as argued by ABRAHAM & CONRADIE (2001), but rather by the areal diffusion of the innovation from Parisian French to contiguous languages and dialects. Evidence of contemporaneous transmission of other cultural artifacts is presented, as well as French and German documentary evidence from the 13th century showing similarly high frequencies of the perfect as opposed to preterit on both sides of the French-German linguistic border. This Sprachbund position is rebutted in some detail by WERNER ABRAHAM in The European demise of the simple past and the emergence of the periphrastic perfect: areal diffusion or natural, autonomous evolution under parsing facilitation? Of all European languages, the Germanic representatives have been seen to form the core of diffusing developments: French, German, Dutch, and Northern Italian (the latter is only a dialectal area!); a less central area, comprising English, Swedish, Standard Italian, Czech, which nevertheless shares with the core many of the features subsumed under the S(tandard) A(verage) E(uropean) characteristics; and, finally, a peripheral area adopting the fewest of these common features (Russian, Hungarian, Basque). While any areal Sprachbund diffusion is a necessary prerequisite for any such assumption, ABRAHAM contests that by showing that the spread of the analytic past has heterogeneous bases. Simple pasts have disappeared in widely adjoining, in particular the southern and eastern parts of Europe (South Germanic, Romance, Slavic), and periphrastic preterits ("present anterior") have taken over their function. Yet the consequences in their respective languages are different. One can distinguish four types of preterit-loss languages and dialectal areas within these. The classifying criteria are: ((0) present anterior different from simple past: Scandinavian, English, a.o.s) (1) present anterior has the same function as the simple preterit: Standard German and Italian; (2) the simple preterit has become very marginal: Serbo-Croatian, Rumanian, French; (3) no simple preterit exists any longer; its function is taken over fully by the present anterior: South German, Yiddish, Slavic. The latter group of languages can be subdivided according to whether or not they have a paradigmatic representation of the periphrastic future perfect and/or a modal usage linked to the present anterior: South German vs. (4) Slavic. While all of this is suggestive of areal spreading, the source being those languages which have advanced the farthest on this cline of preterit demise and newly acquired properties of the periphrastic perfect (which is Russian), ABRAHAM shows that what is at the bottom of the rise of the complex perfect at cost of the simple preterit is the facilitation of parsing and online semantic interpretation in exclusively oral coding and under dense discourse linking. In Some remarks on the formal typology of pronouns in West Germanic, L Ä S Z L 0 argues against the idea that a formal typology of pronouns in West Germanic should be based on the notion of structural deficiency. Comparing the pronominal system of Dutch and Afrikaans, the author argues that (a) deficient pronouns do not constitute a language universal, and (b) deficiency is a property that is discourse-functionally, rather than case-determined. The derived status of deficient pronouns is explained by their category-inherently defocused status, MOLNÄRFI

13

Introduction

which must be licensed in a non-trivial discourse functional chain. Weakness and strength of pronouns are only phonologically relevant distinctions, which may be

partly grammaticalized in some languages. ROLF THIEROFF'S The subjunctive mood in German and in the Germanic languages takes up a much neglected topic in the literature on Germanic: the subjunctive, or more precisely, subjunctive 1 and subjunctive in Icelandic, Norwegian, Danish, Swedish, Dutch and English, notably for the function of the "conditional" and "reported speech". A special feature unique to the Scandinavian languages is presented. There can be no doubt that all of Germanic has undergone severe changes if we take Old Norse, Runic Germanic, Gothic as well as the oldest states of West Germanic (Old English, Old High German) as codes of comparison. It is equally appropriate to say that among all of Germanic, German has remained the most faithful to the source states of the modern languages. Icelandic is no exception to this claim, since, while it has retained rich verbal and nominal inflection, its syntax is very "modern" in the sense that it has no replica in the old Indo-European source tongues. Quirky subject agreement is one of the assets of Icelandic which supports this claim. What distinguishes the most conservative Germanic language, German, from the rest, however, is locatable in its dense mixture of SVOV-configurationality and discourse prominence. No other representative among Germanic shares this characteristic with German - not even its closest relative, Dutch. In what follows, in all due brevity, I will sketch this conservative, and yet highly flexible, typological idiosyncrasy of discourse prominent German. The key process is "scrambling" (HAIDER & ROSENGREN 1 9 9 8 ) - i.e., the word order flexibility for the purpose of (1) dislocating categorically thematic material (such as (clitic) pronouns and definite nouns) to the left of VP/vP, and (2) dislocating other, non-thematic, material out of its base (generated) position for purposes of discourse or text fit. No other language among the Germanic does this with comparable ease. In this sense, then, no other Germanic language but German disposes of scrambling. Scrambling is always executed in accompaniment with refocusing. German, thus, provides its speakers with a keen notion of contrastive stress - which, in turn, provides the hearer with an online interpretability of otherwise untriggered semantic presuppositions. To all appearances, the typological gist of this property is due to German's SVOV structure, which again is not shared by any of the other Germanic languages - at least not to the same undivided extent, if we consider Dutch and West Frisian. The generalization in this introduction is that languages are not alike with respect to which descriptive module accent and topic vs. focus are determined. This is what we called the typological criterion of discourse marking. The result in the present paper corroborates the views unfolded in MOLNÄRFI (2002) and ABRAHAM & MOLNARFI (2002), which extend the typological classification with German being of the "anti-focus type". Hungarian, under this view, is of the "focus type" (KISS 1998). In ABRAHAM ( 1 9 9 7 )

following sense:

it was argued that German is discourse prominent in the

14

WERNER ABRAHAM

(1)

There is a basic word order in German (and Dutch and West Frisian) with rhematic (informationally new) material in VP. The hermeneutic identification of this word order is "grammatical clausal accent" (GA). Any deviation from this basic word order changes the accent status of the clause from GA to CA ("contrastive clausal accent"). In particular, scrambling as well as topicalization yieldsrefocusingidentified by CA The distribution of clausal accent diagnosed by GA vs. CA and triggered by movement is syntactically unambiguous by recourse to ClNQUE's Clause Accent Null Hypothesis: Unless conditioned otherwise, mainly by movement for discourse accommodation purposes (i.e. to change Thema and/or Rhema), the default accent in a German clause lies on the head of the most deeply embedded constituent in the clause (CINQUE 1993; see also ABRAHAM 1995, chapters 13 and 14).

(2) (3) (4)

The default distribution of focus is shown in the following table. Thema or Rhema1 CoOrd

[cp/ip

-

Thema

Thema

Comp/Infl

AgrOp

[vp

[vp

v° ]]]

aber 'but'Modal Particle auch 'also'Modal Particle denn 'then'Modal Particle denn 'then'Modal Particle

PÜNKTLICH

Extraposition heute! 'today'

SpecCp/IP i

Du 'you'

ii

Aber 'but'

Wieso 'why'

iii

iv

Aber 'but'

wieso 'why'

kommst 'come'

'on time' PÜNKTLICH

kommen! 'come'

'on time' PÜNKTLICH

kommen? 'come'

'on time' PÜNKTLICH

'on time'

kommen 'come'

heute? 'today'

Table 1: The link between the clausa: structure and discourse functions in German ['Thema/rhema' in the 3rd column means that topicalized clausal members can be in both discourse functions simultaneously. Capitals mark focus/rhema: a focus constituent or its head may be GA-marked if the focused element is in its base position; it is CA-marked if it has been moved out of its base position. Base positions: inside VP for indefinites, outside for definites and pronominals; the definite subject is the only constituent which is unmarked in topic position.]

The default domain for pronouns and their clitic forms is indicated by default accent - i.e. accent distribution which triggers the fewest set of presuppositions, which simultaneously means that this default accent goes with the largest possible set of contextual links. See (5) below for characteristics of typical fillers of the TH-slot: 1

Subjects appear as TH, non-subjects as RH in this position

15

Introduction (5)

Categorially determined definiteness (pronouns and their clitically weakened forms) as well as definite NPs scrambled out of VP, where they picked up case. This implies: (5a) unaccented and cliticized pronouns are inherently thematic; (5b) consequently, the default position of unaccented and cliticized pronouns is outside - i.e. to the left o f - VP; (5c) (clitic) pronouns cannot surface as rhematic elements unless they receive contrastive stress - in German with the exception of pronominal es 'it', which never carries contrastive stress and, consequently, has to stay to the left of VP, i.e. in the thematic domain. Furthermore, the following distributions hold between positions and accent distinctions (the abbreviation GA = "grammatical accent" in ClNQUE's (1993) "Accent distributional Null Hypothesis" or "Accent default linking rule"): (6a) (6b) (6c) (6d) (6e)

Wir haben sie gestern GESEHEN we have her yesterday seen 'We saw her yesterday' * Wir haben gestern sie gesehen Wir haben gestern SIE gesehen 'It was her that we saw yesterday' SIE haben wir gestern gesehen 'It was her that we saw yesterday' *Sie haben wir gestern gesehen

...

GA

... ... CA ... CA ...

-

The ungrammatical versions - i.e. those without a GA/CA-appreciation - can be deduced directly from ClNQUE's GA-Linking-Rule. Table 2 shows all version of GA since the adverb is outside of VP and does not reach the structural depth required for the assignment of default stress. Thema/ Rhema C7I° WP for AgrOP [VP v° ]]] [v? [cp/ip (clitic) Spec pronouns CP/IP -

Coad

i

Du

wirst

'you'

'will' 'him/him'

ihn/'n

heute

pünkt-

sehen

'see' 'today' lich 'on time'

ii Aber 'but'

'm

auch

'it him'

'also'

iii

Wieso

es ihm

denn

'why'

'it him'

'then'

pünktlich

ABGEBEN!

Thema GA* CA Extraposition heute! GA 'today' GA

'tum in'

'on time' pünktlich

ABGEBEN?

GA

'tum in'

'on time' iv Aber wieso 'm (s denn sofort zeigen heute? GA 'but' 'why' 'him it' 'then' 'at once' 'show' 'today' Tab le2: Field topology: Verbal bracket, discourse categories and t ie distri jution of default accent (GA) [WP = Wackemagel position] -

16

WERNER ABRAHAM

See the illustrations in (7) below accommodating both intransitive and transitive presentational sentences in German. (7a) Es übertölpeltei [VP prinzipiell [vp der IGEL den HASEN EXPL fooled generally the hedgehog the hare (7b) Aber es Übertölpeini [w jedesmal [VP IGEL HASEN tj] but EXPL fooled each time hedgehogs hares -

-

Coord

SpeoCP

-

-

C

U SpecIP Es ExPL-'it'

η

Aber 'but' -

-

-

V0]]]

derlGEL

ti

übertölpeltei prinzipiell 'fool' 'in principle'

es ExPL-'it'

übertölpelnj 'fool'

There

appeared

jedesmal 'every time' -

den HASEN

'the hedgehog the hare' Igel Hasen 'hedgehogs hares' SOME

dogs

IN

THE GARDEN

ti

(position not available

Tab e 3: All-rhematic presentational/existential clauses Since presentational/existential sentences are fully rhematic it follows that they reach no further than IP; no topic is projected in terms of a definite nominal - see the expletive es in SpecIP. For sure, neither lexical subject nor lexical object NP moves out of VP. The verb retains its rhematic function on account of its prederivational base position with agreement and tense/mood assigned only as grammatical information. We stipulate that discourse prominence is the type of link between discourse functions and syntactic structure illustrated in Tables 1-3. Any clausal accent distribution, CA or GA, is accounted for unambiguously through (non-derivative) movement out of the base position. It is concluded that for a discourse prominent language such as German moved positions are always focused. Vice versa, focused elements appear in non-basic positions moved for discourse functional reasons, i.e. for reasons of context accommodation. By this strategy, it is always possible to distinguish CA from GA, even if the element in question appears to be in identical surface position. See also ABRAHAM & MOLNARFI ( 2 0 0 2 ) as well as MOLNARFI (2002). In particular, focused subjects can surface in clause-topical position. In other words, wA-subjects, which are focused for discourse functional reasons, can appear in clause initial position as in many other languages, including English. In line with CINQUE 1993 and ABRAHAM (1992; see also ABRAHAM 1997, MOLNARFI 2 0 0 2 , ABRAHAM & MOLNARFI 2 0 0 2 ) the identification of discourse functions occurs in terms of position and accent designation alone and the definition of base position of the elements involved. Nothing else is involved. In other words, word order as in (6a-e) is driven by prosody. The basic idea is pursued that features of the peripheral system may encode instructions to both the core derivational component and the interface levels. This claim is substantiated by empirical data

Introduction

17

from German and modern spoken Afrikaans (MOLNÄRFI 2002), both languages showing strong prosodic and, in the case of Afrikaans, also morphological correlates of word order variation. Following MOLNÄRFI (2002) and ABRAHAM & MOLNÄRFI (2002), such interdependences will be taken to indicate the existence of a formal distressing license, which drives the movement of referential definites out of the focus domain of VP. This trigger has been called "Antifocus" (AF). It is taken to be represented by its own maximal projection in the tree. Such movement may or may not induce phonological effects. The apparent optionality of object shift emerges, as after spell-out it is either the head or the base position of the scrambling chain that is lexicalized. In this scenario, subsequent PF-deletion of the remaining lower or higher copy is determined by external legilibility conditions, which are directly relevant for the checking of the destressing feature and for the prosodic and discourse interpretation of the shifted object (MOLNÄRFI 2002). It is suggested that the Germanic languages be typologically gauged on the basis of the criteria developed here for German. This concludes my typological sketch of German as densely fusing grammatical SVOV-configurationality and explicit discourse structure. Notice that this may be explained as a particular characteristic of a complex type of language, yet also a highly redundancy-avoiding language.2 No doubt, the triggering feature is V2+ Vlast in German, which likewise is at the bottom of other idiosyncrasies of German (such as verbal clustering and clause union as well as coherent infinitival embeddings in German and (differently) Dutch; see recently H A I D E R 2003).

2

DAHL (2003) develops a scenario of "linguistic maturity" in which avoidance of redundancy is a major criterion. Notice, however, that rich agreement within the German DP (including, thus, agreement on adjectival attributes in dependence on whether the noun has definite or indefinite reference) does not apply to the same characteristic of "redundancy avoidance". The latter characteristic can be expressed more formally in the following way. Let us assume that the output of the lexicon consists of inflected forms, potentially underspecified for morphosyntactic features, which compete for association with a target node. Morphological blocking effects can be modeled in Optimality Theory in terms of constraints that enforce faithfulness of inflected forms to the morphosyntactic requirements of the containing syntactic representation. This could take care of agreement in DP generally. However, in morphology-rich Icelandic and German it may be possible for a less faithful inflected form to overwrite a more faithful form under certain circumstances, as is well-known. As a result, it can be argued that the constraints that form the basis of the interface in most languages need to be augmented to account for the Germanic phenomena. Icelandic and German, thus, provide evidence for a constraint that seeks to maximize the expression of morphosyntactic features across the NP, possibly at the expense of maximal faithful expression of features associated with a given terminal node within the NP. In addition and other than Icelandic, German (and partly, but less consistently, Dutch) would seem to require an economy constraint on the expression of morphosyntactic features within the complex Adjectival Phrase.

18

WERNER ABRAHAM

References (1992): Überlegungen zur satzgrammatischen Begründung der Diskursfunktionen Thema und Rhema, in: Folia Linguistica XXVI/1-2,197-231. ABRAHAM, WERNER (1995): Deutsche Syntax im Sprachenvergleich. Grundlegung einer typologischen Syntax des Deutschen. (Studien zur deutschen Grammatik 41).Tübingen: Gunter Narr. ABRAHAM, WERNER (1997): The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative functional categories, in: ABRAHAM, WERNER & VAN GELDEREN, E . (eds.), German: syntactic problems - problematic syntax. (Linguistische Arbeiten 374). Tübingen: Niemeyer, 11—42. ABRAHAM, WERNER (2000): Zwischen extensionaler Ökonomie und intensionaler lokalistischer Präzisierung: Dativobjekte im Deutschen und ihre Wiedergabe im kasuslosen Niederländischen, in: BITTNER, Α.; BITTNER, D. & KÖPCKE, K.-M. (eds.), Angemessene Strukturen: Systemorganisation in Phonologie, Morphologie und Syntax. Hildesheim u.a.: Georg Olms, 299-316. ABRAHAM, WERNER (2003): The syntactic link between Thema and Rhema: the syntax-discourse interface, in: ABRAHAM, W . & MOLNÄRFI, L. (guest editors), Optionality in syntax and discourse structure - aspects of word order variation in (West) Germanic and other languages. Folia Linguistica. Acta Societatis Linguisticae Europaeae XXXVII/1-2,1-34. ABRAHAM, WERNER (2004): Bare and prepositional differential case marking: the exotic case of German (and Icelandic) among all of Germanic, in: KULKOV, L.; MALCHUKOV, A. & DE SWART, P. (eds.), Case, valency, and transitivity. (Studies in Language Companion Series). Amsterdam: J. Benjamins. ABRAHAM, WERNER & MOLNÄRFI, LÄSZLÖ (2003): The German clause under discourse functional weight: Focus and Antifocus, in: ABRAHAM, W . & ZWART, C . J . - W . (eds.), Issues in comparative German(ic) typology. (Linguistik Aktuell/Linguistics Today 45). Amsterdam: Benjamins, 1-43. CINQUE, GUGLIELMO(1993): A null theory of phrase and compound stress, in: Linguistic Inquiry24/2:239-298. DAHL, OSTEN ( 2 0 0 3 ) : The growth and maintenance of linguistic complexity. Ms. Univ. of Stockholm. HAIDER, HUBERT & ROSENGREN, INGER ( 1 9 9 8 ) : Scrambling. (Sprache und Pragmatik 4 9 ) . Lund. HAIDER, HUBERT ( 2 0 0 3 ) : V-clustering and clause-union: causes and effects, in: SEUREN, P.A.M. & KEMPEN, G. (eds.), Verb constructions in German and Dutch. (Current Issues in Linguistic Theory 2 4 2 ) . Amsterdam-Philadelphia: Benjamins, 9 1 - 1 2 6 . Kiss, KATALIN E. (1998): Identificational Focus vs. Information Focus, in: Language 2. MOLNÄRFI, LÄSZLÖ ( 2 0 0 2 ) : Focus and antifocus in modern Afrikaans and West Germanic, in: Linguistics 4 0 / 6 , 1 1 0 7 - 1 1 6 0 . MOLNÄRFI, LÄSZLÖ (2003): On optional movement and feature checking in West Germanic, in: ABRAHAM, W. & MOLNÄRFI, L. (guest editors), Optionality in syntax and discourse structure aspects of word order variation in (West) Germanic and other languages. Folia Linguistica. Acta Societatis Linguisticae Europaeae XXXVII/1-2,129-162. ABRAHAM, WERNER

What happened to English? * JOHN H. MCWHORTER (Berkeley)

1. Introduction Since linguistic investigations as early as PAUL (1880), it has been well known that in the evolution of grammars, simplification and complexification play complementary roles. I will argue that in the emergence of Modern English, simplification dominated complexification to a greater extent than in any other Germanic language, in a fashion suggesting the operations of a sociohistorical factor interceding with the full transmission of the grammar across generations at a certain point in history. This paper will not revive the hypothesis of BAILEY & MAROLDT (1977) (followed by DOMINGUE 1977, POUSSA 1982) that Middle English was a Creole that developed when Norman French invaders learned English imperfectly and expanded their reduced English into a full language. This hypothesis was motivated partly by the heavy admixture from other languages in English's lexicon and derivational apparatus. But lexical mixture itself does not equate with creolization. Languages can borrow massive amounts of lexicon and even morphology without evidencing any traits that would suggest the label Creole to any linguist, such as many languages of Australia (HEATH 1981), and "mixed" or "intertwined" languages like Michif (BARKER 1997) and Media Lengua (MUYSKEN 1997). BAILEY & MAROLDT and their followers also based their argument on English's notorious paucity of inflection in comparison to other Germanic languages. However, THOMASON & KAUFMAN (1988: 263-342) refute this thesis. The inflectional loss had proceeded considerably before the Norman Invasion, and even in dialects not in contact with French. In addition, today's Mainland Scandinavian languages are little more inflected than English. Moreover, during the Norman rule, numerically French speakers were but a small elite, whose rendition of English can hardly have had impact on a vast majority of monolingual English speakers. These and other arguments will be taken as conclusive in this paper. THOMASON & KAUFMAN'S argument, however, entails that English's heavy inflectional loss was due simply to its being less "conservative" than its sisters, imSALLY THOMASON, IRMENGARD RAUCH, ANTHONY GRANT, ST6PHANE GOYETTE, GARY HOLLAND, JAMES MATISOFF, PETER TIERSMA, JARICH HOEKSTRA, ANDREW GARRETT, MARTIN HASPELMATH,

and my once-and-future mentor ELIZABETH TRAUGOTT have helped me in innumerable crucial ways in venturing this paper. Sincere thanks to all of them, none of whom are responsible, of course, for remaining flaws. A more condensed version of this paper appeared in Diachrortica 19 (2002). SALLY THOMASON, WERNER ABRAHAM,

20

JOHN Η . MCWHORTER

plying that no external factor distinguished English's development. Meanwhile, to the extent that English manifests other features we might treat as less marked than their equivalents in its sister languages, the generative historical linguistics tradition tends to ascribe these to effects of inflectional loss that modern syntactic theory would predict. Examples include the rich literature on the loss of OV and V2 word order, and other features such as obligatory postposing of particles to the verb (e.g. PLATZACK 1986, VAN KEMENADE & VINCENT 1997, FISCHER et al. 2000). The common consensus among specialists on the history of English is that features suggesting a break in transmission or unusual simplification in the English timeline are mere trompe I'oeil's, having in fact emerged by ordinary processes of change. However, the focus on certain abstract syntactic features that the generative enterprise conditions has perhaps narrowed our purview, in a fashion neglecting other aspects of grammar that suggest a larger story. I will propose that loss of inflection is but the tip of the iceberg in terms of Germanic features that English has shed, complemented by many other losses unconnected with analyticity. Overall, a comparison with its sisters reveals English to be significantly less overspecified semantically and complexified syntactically. Some scholars, such as LASS (1987: 3 1 7 - 3 3 2 ) , recognize that English departs considerably from the Germanic template, but leave aside the question as to why, with the implication that the issue was a matter of chance. However, I will argue that a contact-based, external explanation provides a principled account for the relevant facts.

2. The nature of overspecification and complexification 2.1. Overspecification Any given language gives overt manifestation to particular underlying semantic or syntactic distinctions which are left unmarked in many other languages. The fact that many or most languages operate without marking the distinctions in question entails that to mark the feature at all is ornamental, rather than necessary, to human language. As such, in a given area, one language can be considered overspecified in comparison to another. For example, the Northern Californian American Indian language Karok has grammaticalized different verbal suffixes for various containment mediums: ρα:θkirih 'throw into fire', pa: θ-kürih 'throw into water', pa: θ-rüprih 'throw in through a solid' (BRIGHT 1957: 9 8 , 1 0 2 ) . (These morphemes are not perceivable reflexes of the words for 'fire', 'water', or 'solid' respectively.) Most of the world's grammars do not happen to have grammaticalized such fine-grained overt expressions of containment mediums, and it would be impossible to argue that Universal Grammar specifies such. On the contrary, as useful as these suffixes are in Karok grammar, their emergence was due to a chance elaboration within a particular semantic area, not communicative necessity. Thus: in the area of marking of containment mediums, Karok is overspecified in comparison to English, just as in its grammaticalized marking of definiteness and indefiniteness of NPs, English is overspecified in comparison to Karok. A similar argument could be made for other languages re

What happened to English?

21

grammaticalized evidential markers, or the regular overt distinction of alienable from inalienable possession. The Tibeto-Burman language Lahu provides another example. The verb Ιά 'to come' has evolved into various uses, one of which is a benefactive particle indicating that the verbal action is for the benefit of either first- or second-person but not third - cho lä 'chop for me/us/you' (MATISOFF 1991: 396); the verb pi 'to give' is used in the third person. Marking the distinction between discourse participants and referents via pronouns is apparently a language universal, but to do so via suppletive verb stems as well is hardly the usual case cross-linguistically. Lahu is thus overspecified in this area in comparison to most languages. Along these lines, my comparison of English with its sisters suggests that overall, English manifests less of this kind of overspecification than its sisters. Throughout the paper, my delineations of overspecification refer to the above characterization.

2.2.

Complexification

Complexity in syntax is certainly a difficult concept, upon which there is only fitful agreement among linguists working in various paradigms. For the purposes of this thesis, however, we will stipulate that within a given area of grammar, a language's syntax is more complex than another's where it requires the operation of more rules. Thus a grammar involving two kinds of alignment rather than one (i.e. ergative/absolutive and nominative/accusative) can be considered more complex than a strictly nominative/accusative one (cf. the approach of HENRY 1995, HENRY & TANGNEY 1999). Similarly, a language with a V21 rule, requiring that the verb move to C for its surface expression (DEN BESTEN 1983, FISCHER et al. 2000: 1 ΙΟΙ 14) can be considered more complex in this regard than a language with no such requirement of the verb. This conception is admittedly preliminary, but I feel that it will serve adequately to frame, elucidate, and support the particular thesis I wish to present. Importantly, this conception of complexity addresses grammars solely as systems viewed from the outside, as products of millennia of drift and its attendant elaborations upon the rootstuff of the human linguistic capacity. Complexity under this metric is not indexed with relative difficulty of production or processing; this metric takes as a given that all languages are acquired with ease by native learners. Our assumption is that human cognition is capable of effortlessly processing great degrees of overspecification and complexification in language (cf. TRUDGILL 1999). While there is no reason to suppose that differentials in ease of production or processing do not exist, these issues are ultimately of little import to the thesis explored here, a diachronic and comparative one.

1

Abbreviations: V2 = verb in clause-second position; C = Complementizer position (subjunction or finite V in V2-languages).

22

JOHN Η. MCWHORTER

2.3. Epistemological caveat Importantly, my argument is not that English is in any sense radically underspecified or simplified in a cross-linguistic sense. Its do-support, subtle subdivision of the semantic space of the future between four constructions (7 leave tomorrow, I'm going to leave tomorrow, I'm leaving tomorrow, I will leave tomorrow), the subtle interplay of definiteness and referentiality underlying the use of its definite and indefinite articles, and other features are fatal to any argument that English is somehow a "simple" language. Thus my argument proceeds in full acknowledgment of such features, but is predicated upon a relative argument: that overall, English is less overspecified and complexified than its sisters, to an extent suggesting something other than unbroken internal development. My intention can be illustrated with an informal but heuristically useful observation. For both English speakers learning another Germanic language and speakers of other Germanic languages learning English, much of the acquisition task entails learning alternate, but in no sense more complex, strategies for expressing concepts. An example would be the arbitrary differences in semantic space that prepositions cover: pale with fear versus German blass vor Furcht, etc. However, it is my impression that for the English speaker, most of the task beyond this entails learning to attend to things English does not mark, while for the speaker of another Germanic language, most of the task entails learning not to attend to features their native language does mark. The centrality of inflection to modern syntactic theory leads to an impression that this contrast is largely a matter of English's relatively sparse declensional and conjugational paradigms. This, however, is in broad view an artifact of the interests that happen to dominate linguistic inquiry in our times: the contrast to which I refer encompasses a great deal beyond inflection.

3. Examples 3.1. Inherent reflexivity marking Germanic languages overtly mark what is often called "inherent reflexivity". These differ from literal reflexives in that while these refer to an event involving two participants of which both happen, contrary to general expectation, to be the same entity (He shot himself), inherent reflexives entail a perception of one participant, performing upon itself an action whose reflexivity is the expected case rather than an anomaly (He bathed) (HAIMAN 1983, KEMMER 1993). Inherent reflexives are the product of the grammaticalization and bleaching of the reflexive element in verbs connoting inherently reflexive actions, such that in many languages, inherent reflexivity is marked in motion verbs (German sich beeilen 'to hurry'), psych-verbs (sich erinnern 'to remember'), and verbs of social behavior (sich benehmen 'to behave'); this is a cross-linguistic developmental tendency (cf. KEMMER 1993, PEITSARA 1997).

23

What happened to English?

Inherent reflexivity marking is common in all of the Germanic languages but English. In this and similar tables, for reasons of space, Mainland Scandinavian is represented by Swedish, the feature having also been identified in Danish and Norwegian:2 German

sich rasieren 'to shave', sich beeilen 'to hurry', sich erinnern 'to remember' Dutch zieh scheren 'to shave', zieh bewegen 'to move', zieh herinneren 'to remember' (203-204) Frisian hy skeart him 'he shaves', ik skamje my Ί am embarrassed', ik stel myfoar Ί imagine' (66,147) Afrikaans hy bevind horn 'he is situated (at)', hy roer horn 'he gets going', hy herinner horn 'he remembers' (288-291) Scandinavian raka sig 'to shave', rörasig 'to move' kännasig 'to feel' (105-106) Icelandic koma 'to comtlkomast 'to get to, reach', snua ser 'to turn around', skammast sin 'to be ashamed' (105,143) Faroese raka seer 'to shave', sruigva seer 'to turn', eetla seer 'to intend' (117-118) Yiddish bukn zikh 'to bow', shlaykhn zikh 'to sneak', shemen zikh 'to be ashamed* (89-90) Table 1: Inherent reflexives in Germanic Note that a distinct reflexive pronominal form is not necessary to inherent reflexive marking. The absence of a reflex of sich is not local to English, but was already the case in Ingvaeonic, such that Frisian also lacks a sich reflex; Afrikaans does as well. In all Germanic languages, inherent reflexivity is marked in the first and second persons with the corresponding accusative or oblique pronoun (ich rasiere mich, du rasierst dich Ί shave, you shave'), and those without a sich reflex mark reflexivity in the third person with an accusative or oblique third person pronoun, e.g. Frisian hy skeart him 'he shaves'. In Old English, inherent reflexivity was marked with either the dative or the accusative pronoun: (1)

tha

beseah

hehine

toanum

his manna

and cweed

then look.PAST he him.DAT to one.DAT his man-PL.GEN and say .PAST 'Then he looked at one of his men and said' [VISSER 1963: 14]

(2)

Reste

öcet folc

hit

rest.PAST

the

it.CL on

people

on

'The people rested on the seventh day'

dam

seofothan

the.DAT seventh

dcege day

[ibid. 147]

Even at this early date its use was optional (0 where the reflexive pronoun would occur): 2

Page numbers refer to the following sources: Dutch: DONALDSON 1997; Swedish: HOLMES & HINCHLIFFE 1997; D a n i s h : THOMAS 1991; N o r w e g i a n : STRANDSKOGEN & STRANDSKOGEN 1986;

Afrikaans: PONELIS 1993; Frisian: TIERSMA 1985; Icelandic: KRESS 1982; Faroese: LOCKWOOD

1955; Yiddish: LOCKWOOD 1995. Throughout the paper, where not cited, the source of the data is these (e.g. in cases of negative evidence such as the absence of a feature in a grammar, which is often impossible to refer to by page).

24

JOHN Η . MCWHORTER

(3)

se sylfa Drihten wolde 0 of heofenum on eordan beseon he self Lord want.PAST from heaven.DAT on earth.ACC see.LNF 'The Lord himself wanted to look upon the earth from heaven' [ibid. 146]

Throughout the Middle English period, however, inherent reflexivity is marked increasingly less (MUSTANOJA 1960: 431), likely preferred as a metrical device (FISCHER 1992: 2 3 9 ) . As early as Old English, texts suggest only vague semantic distinction between a given verb's usage with and without the reflexive pronoun (MITCHELL 1985: 114, VISSER 1963: 3 2 2 , RISSANEN 1999: 2 5 6 ) .

By the Early Modern English period, PEITSARA (1997: 303) finds inherent reflexive marking in only a third of potential cases from 1500 to 1570, and in less than a sixth from 1570 to 1640. By the latter period, PEITSARA finds the marking only in a limited number of verbs, including ones of motion, posture, self-care and equipment; of psych-verbs only fear retains it (optionally) and among social ones, commend. Eventually, inherent reflexive marking with simple pronouns is eliminated completely, except for in scattered frozen archaisms (Now I lay me down to sleep). As inherent reflexive marking declines, self which begins in Old English expressing emphatic reflexivity (among other uses: cf. MITCHELL 1985: 115, FALTZ 1985: 1 8 - 1 9 , 35), increases in frequency in Middle English; by the fifteenth century its use had bleached semantically into compatibility with verbs that previously took just a simple pronoun (PEITSARA 1997: 3 2 0 - 3 2 3 ) . Thus where WYCLIFFE in the fourteenth century has Adam and his wijf hidden hem fro the face of the Lord God (Old Testament, Genesis 3:8), TYNDALE in the sixteenth has And Adam hyd hymselfe and his wyfe also from the face of the LORde God (Five Books of Moses, Genesis 3:8). However, instead of replicating the full range that the bare-pronoun reflexive strategy once covered, ^//-pronouns settle into the modern pattern, largely marking only literal reflexivity. Today, reflexivity is usually marked where operation upon the self is emphasized for clarity (bathe oneself) or stylistic purpose (they hid themselves). With some verbs, the reflexive usage has conventionalized into a particular meaning (to behave versus to behave oneself), while in only a limited number of verbs is the marking obligatory (pride oneself, perjure oneself). The result of this process was that as PEITSARA neatly puts it, English became unique among Germanic languages in "an individual tendency to treat overt reflexivity as redundant, unless marked for practical or stylistic reasons" ( 1 9 9 7 : 3 3 7 ) . This cannot be attributed simply to the fact that inherent reflexivity marking was already optional in Old English, given that such an account begs the question as to why English did not instead choose to conventionalize the initially optional usage rather than eliminate it. In Gothic, for instance, inherent reflexives were at first used almost exclusively with humans, and even then, were optional, the variability in their usage having ambiguous semantic import, as we can see within a single sentence in ni idreigo mik, jah jabai idreigoda Ί do not repent, although (s)he repented' (DAL 1966: 155). However, while an analogous variability in Old English drifted into the gradual disappearance of the usage, German, despite the generalization of the feature being a gradual process (ibid. 1 5 5 - 1 5 6 ) , obligatorified it; CURME ( 1 9 5 2 : 3 3 1 ) notes "German is usually tenacious of reflexive form even after its meaning has changed".

What happened to English?

25

Obviously this was also the case in the other Germanic languages. In Scandinavian, the grammaticalization went so far as that the reflexive pronoun has eroded into a mere suffix on many verbs, creating deponents such as minna-s 'to remember'; Icelandic and Faroese's -si suffix is similar. The English situation must also be seen within the context of a similar generalization of inherent reflexive marking across Europe as a whole. HASPELMATH (1998: 276) for example, describes the development of anticausative marking with the reflexive pronoun as a pan-European Sprachbund feature. Along those lines, the grammaticalization of inherent reflexives is obviously connected to the marking not only of shades of reflexivity, but to distinctions of valence (transitivity) and mood (passive and middle voice). These distinctions are commonly related cross-linguistically (e.g. LYONS 1968: 3 7 3 - 3 7 5 ) , and the Scandinavian -5-marked verbs, for example, also encode passivity (bakas 'to be baked'). In this light, the disappearance in English of inherent reflexive marking can be seen as one symptom of the general drift towards "transitivization" that VlSSER (1963: 127-135) describes, where the overt distinction between transitive and intransitive use of verbs erodes. Under this analysis, the eclipse of inherent reflexive marking was part of a general process which also included the disappearance of the ge-prefix that once distinguished transitive verbs (cernan 'to run', gecernan 'to reach, attain by running'). While we might be tempted to suppose, as ViSSER's treatment implies, that this "transitivization" drift was the fortuitous result of various independent processes that just happened upon a single result, the question we must ask is why similar processes did not converge upon the same result elsewhere in Germanic. Scandinavian also lost its ge-prefix, and the coalescence of vowels that once marked causative distinctions (Old English sincan versus sencari) is hardly unique to English. Yet if there is a tendency for a grammar to fill in "open spaces" in syntax as well as phonetic inventories as VlSSER surmises (ibid. 135), then we might ask why English did not, as its sisters did, seek to compensate for the ravages of phonetic erosion by obligatorifying its usage of reflexives to retain overt signaling of transitivity and passiveness. Put another way, why did English not submit these pronouns to "exaptational" usage, in LASS' (1990) conception borrowed from evolutionary biology, recruiting reflexive pronouns as valence and mood markers as dozens of other languages were concurrently doing across Europe? Instead, English became a grammar markedly less overspecifled than its sisters in this area, leaving inherent reflexivity, transitivity, and causativity to context to an extent unique in its subfamily, and unusual in the Indo-European family as a whole. 3.2. External possessor marking When a possessed object falls into a semantic class roughly definable as inalienable, Germanic languages typically mark the possessor with a case separate from that of the possessed entity, as in German's use of the dative in Die Mutter wäscht dem Kind die Haare 'The mother washes the child's hair'. The result is that the possessee is encoded as a distinct constituent rather than as a constituent of the pos-

26

JOHN Η . MCWHORTER

sessum NP. While this construction is sometimes termed the "sympathetic dative", I follow VERGNAUD & ZUBIZARRETA (1992) and KÖNIG & HASPELMATH (1997)

in referring to this construction as "external possessor marking", as the term captures the larger generalization that the "external" constituent can be marked with cases other than accusative. Indeed, this feature is found in all of the Germanic languages but English except Afrikaans; in Mainland Scandinavian and Icelandic the marking is locative rather than dative, while Yiddish has the dative for pronouns and the locative for nouns. External possessor marking generally refers to animate possessors, and applies to body parts, relatives, clothing, and sometimes even emotional conditions, and habitations: the motivating factor is membership in the "personal sphere" (KÖNIG & HASPELMATH 1 9 9 7 : 5 3 0 - 5 3 3 ) . (4)

(5)

(6)

German: Die Mutter wäscht dem the mother wash.3s the-DAT 'The mother washes the child's hair.' Dutch: Men heeft hem IMP have.3s 3S-OBJ 'They broke his arm.'

die

Haare.

the-PL

hair.PL

[KÖNIG & HASPELMATH 1 9 9 7 : 5 5 4 ]

zijn his

Frisian: Ik stompte

my

de

I

me. DAT

the

bump.PAST

Kind child

arm

gebroken.

arm

break-PART

[JARICH HOEKSTRA, July 2001 p.c.]

holle. head

Ί bumped my head.' (7)

(8)

(9)

Scandinavian: Nägon bröt someone break.PAST 'Someone broke his arm.'

[KÖNIG & HASPELMATH 1 9 9 7 : 5 5 9 ]

armen arm.DEF

Icelandic: Han nuddadi a he massage.PAST on 'He massaged her legs.' Faroese: Eg hoyrdi I

hear.PAST

pa

honom.

PREP

3S.OBJ

[ibid.] henni

fcetur-na.

her.DAT

leg.PL-DEF:ACC [LOCKWOOD

r0ddina voice

ά on

1955:105]

honum. 3s.DAT

Ί heard his voice.' (10) Yiddish: gevasht mame hot em (10a) Di the mother have.3s 3S.DAT wash.PART 'The mother washed his hair.' mame hot gevasht di (10b) Di the mother have.3s wash.PART the 'The mother washed the child's hair.'

[JIM MATISOFF, p.c.]

di the

hor hair

hor. hair fam for.the

kind. child

27

What happened to English?

Old English had external possessor marking, as in: (11)

Old English:

[MITCHELL 1985: 1 2 5 ]

tha

cnitton

hi

then

tie.PAST t h e y

rapas

hire

to handum

rope-PL

her.DAT t o hand.DAT.PL a n d

and

fotum foot.DAT.PL

'then they tied ropes.. .to her hands and feet' However, the construction, already optional in Old English (MITCHELL 1985: 126), decreases in frequency throughout the Old English period (ibid. 1 2 6 - 1 2 7 , AHLGREN 1 9 4 6 ) . VISSER ( 1 9 6 3 : 6 3 3 ) notes it as "common" but not obligatory in Middle English, but almost completely obsolete by the Modern English period. Modern English retains but sparse remnants of the earlier construction, as in She looked him in the eyes and She died on me (KÖNIG & HASPELMATH 1997: 5 5 4 , 560). AHLGREN ( 1 9 4 6 : 2 0 1 - 2 0 2 ) suggests that English may have lost external possessor marking due to the collapse of the dative and the accusative in English casemarking. This early explanation has its echo in the emphasis in modem treatments on tracing historical developments in English to loss of inflectional distinctions. But this surely cannot serve as an explanation for the loss of external possessor marking when Dutch and Scandinavian have experienced the same collapse of dative and accusative in pronouns and yet retain the feature. It is also germane that even a language retaining the dative/accusative contrast robustly, Icelandic, has nevertheless shed dative-marked external possessor marking for marking it with the locative. Obviously the collapse was not a causal factor (cf. HASPELMATH 1999: 125). Nor can AHLGREN's suggestion ( 1 9 4 6 : 2 1 0 - 2 1 6 ) that Latin was a deciding factor stand, when French and other Romance languages retain dative external possessor marking (il m'a frappe la main 'he hit my hand') despite Latin playing as influential a role as a language of scholarship in their lifespans as it did in that of English. Thus the question that arises is why it is ungrammatical in English to say They broke him his arm, when Dutch has Men heeft hem zijn arm gebroken, or why English does not have Someone broke the arm on him as the Scandinavian languages do. Proposing a pathway of semantic evolution of prepositions upon which external possession falls at a highly evolved point, HASPELMATH ( 1 9 9 9 : 1 3 0 - 1 3 1 ) suggests that English does not use the preposition to in the function because it has yet to abstractualize to this extent. This is well-taken in explaining why *They broke the leg to him is ungrammatical, but does not account for why English does not instead encode external possession with oblique pronouns as Dutch does, or locative ones like Scandinavian. Under any internally-based account, it is indeed "difficult to find a proper explanation", as ViSSER ( 1 9 6 3 : 6 3 3 ) puts it, of the absence of external possessor marking in English; like so many, ViSSER is left to simply describe the change. KÖNIG & HASPELMATH ( 1 9 9 7 : 5 8 3 ) note that the general tendency throughout Europe is for dative external possessor marking to recede, with only Baltic, Slavic, and Albanian preserving it as robustly as in early documents. Yet our question regarding English must be why it has lost the feature so quickly and thoroughly: today, in Europe Welsh and Breton are the only other languages that lack the con-

28

JOHN Η . MCWHORTER

struction (and Turkish if we count it as a European language), with even Finnish and Hungarian having picked up reflections of it, presumably through Sprachbund effects (ibid. 587-588). Indeed, the loss of this construction in English must be viewed against concurrent cross-linguistic tendencies that are the very source of constructions such as these. Despite the broad trend over millennia to dilute external possessor marking, for instance, Romanian has extended the semantic boundaries of the "personal sphere" somewhat: (12) Ne-am läsat bagajele in autocar. lPL.DAT-have.LPL leave-PART baggage.PL in coach. 'We left our baggage in the coach.' [DELETANT& ALEXANDRESCU 1992:108] Also germane here is what Icelandic linguists term "dative sickness", whereby over time, the marking of experiencers as dative rather than accusative is increasing over time rather than decreasing, as in: (13a) Mig lS.ACC (13b) Mer IS.DAT

brestur lack.3s brestur lack.3s

Ί lack courage.'

kjark. courage.ACC kjarkur. courage.NOM [SMITH 1992: 291]

SMITH identifies dative sickness as symptomatic of a general diachronic tendency for case marking to decrease in what he terms "abstractness"; under his definition, abstractness decreases as the linking of grammatical relations (general) to semantic roles (specific) becomes more explicit. Thus the pan-European tendency to dilute dative possessor marking co-exists with a countervailing possibility that a grammar may also drift into increasing the overt marking of particular semantic roles applicable to a given general grammatical relation. English shunned the pathway that Romanian and Icelandic have taken, so decisively that external possessor marking vanished completely. Finally, it is relevant that the only other Germanic language lacking external possessor marking is Afrikaans, whose structure is now agreed upon to have been decisively impacted by extensive acquisition of Dutch as a second language by people of various ethnicities. This is one suggestion that the situation in English is not simply a matter of internally-driven "business as usual". English, then, is unique among its European sisters in having chosen not to mark an inalienable, "personal" shade of experiencerhood. While just beyond the island both northward and eastward other Germanic languages have developed and retained a degree of inalienable possessive marking, English makes no grammaticalized differentiation between He grabbed my hand and He grabbed my folder, and as in the inherent reflexive marking case, is quite unusual in this even among European languages as a whole.

What happened to English?

29

3.3. Grammatical gender marking on the determiner The mechanics of the loss of grammatical gender in English have been well covered, recent examples including THOMASON & KAUFMAN (1988) and LASS (1992: 103-116). By the end of the twelfth century, grammatical gender was already all but lost in northern dialects; two centuries later, it had all but disappeared even in the south (cf. STRANG 1970: 265). However, in the emphasis on the phonological predictability of these changes, it has been less acknowledged that this change left English the only Indo-European language in Europe with no grammatical gender marking - despite the obvious vigor of phonetic erosion and analogy in its sisters. Certainly, as THOMASON & KAUFMAN emphasize, English is hardly unique in having lost grammatical gender marking on nouns themselves - Mainland Scandinavian, Dutch and Frisian have only remnants of segmental indication of gender on nouns. Even Old English had already moved considerably in this direction, the emergence of multiple homophonies via phonetic change having already rendered the nominal morphology "relatively inexpressive and ambiguous", as LASS (1992: 104) puts it and so many others have noted. The Germanic syllable-initial stress system is well-known for having encouraged the erosion of unaccented word-final segments, which left nominal morphology especially vulnerable. But nominal inflections were but a subset of the grammatical gender marking apparatus. To quote LASS again, in Old English "the richest and most distinctive marking for nominal categories is on determiners, in the strong adjective declension, and in pronouns" (LASS 1992: 106). Here, the determiners are particularly important. Concurrently with the erosion of the nominal inflections, an initial threegender distinction in the demonstrative/determiner - se (masc.), seo (fem.) and pcet (neut.) - collapses into the gender neutral the. Scholars on the history of English typically subsume the inflectional erosion and the collapse of the determiner's case distinctions under a general "shift" from a grammatical to a natural gender marking system, such as STRANG (1970: 265, 268), and LASS, who describes this as a "cumulative weighting of 'decisions' in favour of natural gender". However, in a broader view, the "drift" characterization can serve only as a description rather than as an explanation. This is because it begs the question as to why all of the other Germanic languages of Europe, despite the erosion of the nominal inflections, maintained a grammatical gender distinction in the determiners. Surely, Dutch, Frisian and Mainland Scandinavian collapse the original Germanic masculine and feminine into a common gender contrasting with a neuter. However, it is unclear why English could not have done the same: it would even seem to have been phonetically plausible for se and seo to collapse into, perhaps, se, with pcet remaining as a neuter marker. STRANG (1970: 268) states that "Gender, as a grammatical system, can hardly survive the transformation of the personal pronoun system" - but since Dutch, Frisian and Mainland Scandinavian underwent similar collapses in their pronouns, it is unclear that the disappearance of grammatical gender in English was so foreordained. (Note also that some nonstandard Mainland Scandinavian dialects retain all three genders [e.g. HABERLAND 1994: 324].) Similarly, LASS' observation on "cumulative weightings" is obviously cor-

30

JOHN Η . MCWHORTER

rect in itself, but reveals English speakers to have been unique among Europe's Germanic languages in this regard. Is it possible that the shapes of Old Norse's articles led to the collapse of gender distinctions in Old English's cognate items as English speakers intermingled with Scandinavian settlers? It would seem that this explanation will only take us so far. The definite articles English speakers would have heard Old Norse speakers using were pcen (masc.), pe (fem.), and pcet (neut.) ( G O R D O N 1927: 3 0 2 ) . It is logical that English speakers in contact with Old Norse might have replaced the initial consonant of se and seo with [Θ], the varieties being so typologically close that small adjustments like these could have gone a long way in easing communication. Yet contact qua se gives us no reason to assume that the immediate result would have been a single gender-neutral item the. When bilingualism between Western Danish and Low German was common in the thirteenth to fifteenth centuries, Danish lost much inflection ( H A U G E N 1981) but retained gender distinction in its definite article (rather than reducing all forms to, for example, [da] on the phonetic model of what was common to the German definite articles). On the contrary, one plausible scenario is that English would have developed, for instance, a common gender item such as [Be] and preserved its neuter pcet. If such had occurred, there are no grounds for assuming that phonetic erosion would inevitably have eliminated the distinction in final consonant between the two forms. Faroese is a living demonstration, with its cognate configuration, masculine and feminine tann and neuter tad. After centuries of regular use, the two remain distinct, partly due to the inherent conservativity of heavily used items. Only in English was this tendency overridden in favor of eliminating gender entirely. Two broader observations highlight that English's lack of grammatical gender is a more "interesting" fact than generally assumed. First, among Germanic languages, again English's only parallel is Afrikaans, a language whose history was heavy with second-language acquisition. Second, it has been seldom remarked that in its lack of any kind of grammatical gender within the noun phrase, English is unique not only among European Germanic languages, but among all the standard languages of Europe. As we would predict from the tendency for erosion and analogy to erase word-final morphology, there are scattered instances recorded in Europe of the loss of grammatical gender: but this is only in particular nonstandard dialects, such as Western Danish (HABERLAND 1994), Ostrobothnian Swedish, Tamian Latvian (MATTHEWS 1956), and Mandres Albanian (HAMP 1965). Crucially, in the last two cases, gender was lost not through internal change but because of language contact (with Livonian and Turkish respectively) - significant for our argument that English was crucially affected by contact. To be sure, the two cases of internal loss are both Germanic. Yet the question remains: why is English the only European language in which all dialects have lost grammatical gender, such that today English speakers are the only Europeans who encounter grammatical gender marking as a new concept

What happened to English?

31

when acquiring another European language?3 Or, to view this from another angle,

English is the only language in Europe where loss of grammatical gender occurred so quickly and completely as to be a fait accompli across all dialects by the time European languages were being standardized in the middle of the last millennium. Thus English contrasts with its sisters in lacking a particular type of overspecification: the obligatory marking of noun phrases according to categories generally only marginally correspondent with any real-world distinctions. Some linguists are uncomfortable with the idea of grammatical gender marking as "useless", and indeed once present, it can be useful in reference tracking (e.g. HEATH 1975, FOLEY & VAN VALIN 1984: 3 2 6 ) . But TRUDGELL ( 1 9 9 9 ) rather conclusively refutes the notion that the gender marking arose in order to serve such functions, it being rather clear that the markers can be "exapted" into this function along the nonteleological lines of KELLER'S ( 1 9 9 4 ) "invisible hand" concept. As such I am inclined to classify grammatical gender as equivalent to, in LASS' deft phrasing, "linguistic male nipples" (1997: 13). Yet even if one is inclined to disagree, the point stands. Whether English was the only Germanic language disinclined to preserve a useless feature, or the only one that shed the feature despite its being useful to communication, it has been unique in its "streamlining" orientation.

3.4. Derivational morphology As is well known, English is unusual among Germanic languages in the volume of original Germanic derivational morphology that it has lost. The typical account focuses on the frequent replacement of Germanic derivational affixes with French ones. But some analysts note that the loss appears to have predated significant contact with French (STRANG 1970: 191; DALTON-PUFFER 1995: 39); while French lexical items often only appear in texts after the Norman occupation, HlLTUNEN (1983: 92) describes the derivational loss as virtually complete as soon as Middle English texts begin. This means that in the strict sense, Old English apparently simply let a great deal of its derivational apparatus go; the French replacements were a later consequence of geopolitical developments. 3.4.1. Verb prefixes Of course, in the case of many of the Old English prefixed verbs, we cannot speak properly of loss given that they were simply replaced by equivalent phrasal verbs, as HlLTUNEN describes: toberstan > 'to break apart', inlcedan > 'to bring in', etc. In these cases, however, this only happened where the prefix either co-existed with a free preposition (e.g. in) or was of semantics robust and discrete enough to be readily substituted by an equivalent free word: down for nither-, around for ymb-, up and out for the intensificational for-, as in forbcernan > to burn up. 3

I take the liberty of assuming that speakers of the aforementioned nonstandard dialects have generally had enough exposure to the standard one to be familiar with the concept of gender, especially in these times when dialects of this kind are so often threatened by standard varieties and geopolitically dominant languages.

32

JOHN Η . MCWHORTER

This process would then fall under the rubric of the general loss of morphology in the inflectional realm, where similarly, distinctions coded by affixation are often replaced by ones coded by free morphemes (e.g. prepositions) or word order. But all of the derivational losses cannot be subsumed under a simple substitution of the analytical for the synthetic, as certain prefixes were instead eliminated from the grammar without any substitution. Namely, the prefixes with semantics leaning more towards the grammatical, abstract pole simply disappeared except for fossilized remnants. Thus English lost its transitivizing be- (seon 'to see', beseon 'to look at'), and ge-, alternatively described as transitivizing (VISSER 1963: 127) or perfectivizing (MITCHELL & ROBINSON 1986: 58), as in ceman 'to run', gecernan 'to reach, attain by running' and winnan 'to toil', gewinnan 'to conquer'. Another question arises with another use of for-. Grammars typically describe its relatively compositional uses, such as the intensificational one. However, a survey of its uses across the Old English lexicon shows that its contribution had often bleached to the point that there was little or no perceptible difference of meaning between the bare verb and its conjunction with for-, as in helanlforhelan 'to conceal' and polian/forpolian 'to lack, be deprived o f . This is also the case in German, Dutch, Afrikaans, Yiddish, Mainland Scandinavian, and Frisian, where there are doublets of this type whose differentiation of usage is, at best, highly subtle and sometimes register-bound (German wamen/verwarnen 'to warn'). It is likely that this was a step towards the reanalysis of the/or-cognate in these languages as simply a marker of verbhood, connoting transformation, applicable to nouns and adjectives as well as verbs: German verlängern 'to make longer', Dutch vernederlandsen 'to Dutchify', Afrikaans verafrikaans 'to Afrikaansify', Swedish förgifia 'to poison', etc. Even heavy contact has not hindered this development, as we see in Afrikaans as well as Pennsylvania German, threatened by English, where nevertheless the cognate prefix is used to create new verbs: [farbotft] 'all botched up' (VAN NESS 1 9 9 4 : 4 3 3 ) . Old English could, theoretically, have replaced the compositional uses of for- with phrasal verb particles, but retained for- as a verbalizer of this kind. But instead, for- disappeared as a productive morpheme. In short, we are interested in the loss of overt specification that Old English's loss of prefixes entailed. Generally, borrowed affixes simply replace native ones with no general loss in amount of morphology, an example being the borrowing of many Uzbek suffixes into Tadzhik dialects as the result of heavy intermarriage ( C O M R I E 1 9 8 1 : 4 1 , 5 1 ) ; these dialects have not left Tadzhik less morphologically elaborated than its relatives overall. As it happens, Icelandic and Faroese have also opted for phrasal verbs to the virtual exclusion of the Germanic verbal prefixes, including losing be-, ge- and forcognates. My thesis hardly rules out that a given language might have shed a given feature, due either to developments elsewhere in the grammar or to sheer chance. However, my thesis is indeed that to couch the developments in English as unremarkable results of a "trend towards analyticity" misses a larger point. In that light, we must first note how very much inflectional morphology Icelandic and Faroese nevertheless retain: both, for instance, have various allomorphs of the past participial suffix (e.g. in Icelandic, -inn and -tt/ddur) where English retains only -en (and this only on a subset of verbs), and meanwhile retain the various other overspeci-

What happened to English?

33

fied features discussed thus far. Moreover, it is also relevant that even Afrikaans, despite its heavy inflectional loss, has retained the verbal prefixes be- (beslis 'to decide'), ont- (ontken 'to deny') and ver- (verpletter 'to smash'), and not in fossilized form, but with a certain degree of productivity, as demonstrated by neologisms such as beplan 'to plan', ontlont 'to defuse', and verafrikaans 'to Afrikaansify' (PONELIS 1993: 556-557). When the derivational prefixes were lost in two languages otherwise heavy with morphology, and meanwhile preserved in a language that otherwise underwent major reduction in morphology, the loss in English appears traceable to something other than random phonetic erosion - something more specific was at work. 3.4.2. Suffixes The losses of various derivational suffixes in Old English is overall less indicative of any trend against overspecification. Old English documents capture the language at a stage where some suffixes are robustly productive while others are falling by the wayside, some obscured from perception by the transformations of the root they caused (most notably -the, as in foul/filthe, young/you(ng)the), others losing out in Darwinian competition with ones filling similar spaces (-reden and -lac versus -ness). In any language we see affixes at various points on the cline between glory and oblivion, and thus it is not necessarily the case that these losses in themselves suggest that Middle English was a "recessive" language as DALTON-PUFFER (1995) proposes. However, there is one case here which again points to a trend towards underspecification. English loses its infinitive marker -(e)n by the end of the 1400s (LASS 1 9 9 2 : 9 8 ) . Other than English, only Afrikaans, Germanic's contact language par excellence, has lost this marker completely. But in itself, the loss in English entailed no loss of overspecification. Already in Old English, there was a semantic difference between bare infinitives and those occurring with to, the former's conjunction with the preceding verb connoting more transitive actions processed as one event (CALLAWAY 1 9 1 3 cited in FISCHER 1 9 9 7 ) . When the infinitive marker wore off and all infinitives were then marked with to, this distinction was not lost: -mg-marked verbs came in to connote the more transitive relationship with infinitives now processed as less so (e.g. I saw him doing it versus I wanted to do it) (MITTWOCH 1 9 9 0 , FISCHER 1 9 9 7 : 1 2 6 ) .

For the purposes of this argument, however, it is important to note that English lost not only the verb-marking reflex of this morpheme, but also those that served derivationally as word-building devices. In Old English, -sian, -ettan, and -Icecan served to make nouns and adjectives into verbs (ricsian 'to reign', licettan 'pretend', geanlcecan 'to unite' [MITCHELL & ROBINSON 1 9 8 6 : 6 0 ] ) . The loss of the infinitive marker flushed these away as well, despite the fact that they were more phonetically robust than the -(e)n marker alone - and this must be seen against the fact that, again, every other Germanic language in Europe held on to at least the vocalic segment of the infinitival marker itself. Yet English did not take the route of, for example, letting -(e)n erode but retaining markers -s, -ett, and -Icec as derivational equipment. Instead, English chose the pathway towards its now notorious

34

JOHN Η . MCWHORTER

fondness for zero-denominal and -deadjectival derivation - in German, I telephoniere\ in English I just telephone. 3.4.3. Description versus explanation The causes traditionally adduced for this rather striking sloughing away of derivational apparatus within a few centuries leave more questions than answers. The idea that lack of stress rendered the morphemes uniquely vulnerable addresses a tendency rather than an inevitable death sentence. German and others, after all, retain, for example, past participle marker ge-, and even southwestern English dialects retained its cognates y- and a- (cf. BARNES 1 8 8 6 : 2 7 - 2 8 on Dorset) as a participial marker until pressure from the standard rendered them extinct. This last questions MARCHAND's ( 1 9 6 9 : 1 3 0 - 1 3 1 ) suggestion that the vowel-initial prefixes in particular were uniquely vulnerable, as well as the idea that erosion is the sole reason that the almost three dozen Old English verbs transitivized by ge- that VlSSER ( 1 9 6 3 : 1 2 7 ) lists were shorn of their valence markers. Authors also sometimes suppose that an affix was ripe for elimination because it had many meanings (DALTON-PUFFER 1996: 179 on adjectival marker -ly), or because its contributions to many stems were no longer semantically predictable (MARCHAND 1 9 6 9 : 1 3 0 - 1 3 1 on for-). However, in any language, a given affix may remain in productive use in a core meaning while its contributions to myriad roots have drifted into noncompositionality. The noncompositional uses are not evidence of imminent demise of the affix, but merely indications that the affix has been in use for a long time. The German ver- is a useful example. One usage conveys the notion of 'away': jagen 'to hunt', verjagen 'to chase away'. There are extended meanings from this one, such as error ('away' from the right path) as in antonymy: lernen 'to learn', verlernen 'to forget'. Meanwhile, many uses of ver- are unattributable to any of these meanings and must be learned by rote, such as nehmen 'to take', vernehmen 'to perceive'. Yet this is not taken to signal that ver- is on its way out of the grammar; on the contrary, it is used productively to create new verbs (verschlagworten 'to file under a subject heading' [INGO P L A G , p.c.]), with the non- and semi-compositional results of its historical legacy simply dragged along by speakers. Thus the question is why English does not drag along noncompositional cases like forbid and forgive at the same time as creating words likeforenglish to mean to Englishify. Finally, there are explanations such as VLSSER's ( 1 9 6 3 : 1 3 4 ) that a given affix disappeared because a great number of the words displaying it "dropped into disuse". The implication would seem to be that the massive incursion of French words eliminated so much of the original Germanic lexicon that in some cases too few uses of a given affix remained to be processible by speakers. Yet it is well known that French words often took their place alongside Germanic words to create synonyms, often occupying different registers. VISSER'S list of now lost words where be- was affixed to roots still used in Modern English is worth citing in full: bebark, bedwell, bechirp, beflow, befly, begaze, beglide, beglitter, bego, behoot, beleap, belie, bemew, berain, beride, berow, beshite, beshriek, besit, bescramble, bescratch, besparkle, beswink. First, why could a healthy subset of words of this kind not have persisted alongside French equiva-

What happened to English?

35

lents, as help persisted alongside aid, etc.? Certainly we would expect some to vanish by the sheer dictates of serendipity - but so many that today the prefix occurs on too few words to be processible to the modern speaker? Even if all of the words on VlSSER's list did for some reason "drop into disuse" by chance, why did speakers not come to apply the native root to borrowed ones, as they did some native suffixes (,speakable, bondage [DALTON-PUFFER 1996: 221], or today,faxable)l An alternate interpretation of the disappearance of the words is that it was the affix that speakers were rejecting, not the words themselves - especially when they so often retained the root itself (i.e. in reference to VlSSER's list, bark, dwell, chirp, flow, etc.). Thus the loss of derivational morphology in English takes it place alongside the loss of inherent reflexive marking, external dative possessor marking, and grammatical gender marking in rendering the language less accreted with overspecification. In this case, English shed overt marking of transitivity with the jettisoning of be- and one use of ge-, and overt marking of noun- and adjectivehood with its shedding of derivational uses of reflexes of the infinitive marker -(e)n and the prefix for-.

3.5. Directional adverbs Germanic languages typically distinguish forms of adverbs of place according to location, motion towards, and motion away from (Swedish här 'here', hit 'to here', and härifrän 'from here' HOLMES & HLNCHLLFFE 1997: 115-116). Old English did so as well (e.g. her, hider, heonan), but the system was already fragile, with her often used for motion towards, -arc-suffixed forms losing their sense of 'from' and being used as mere locationals, etc. (MITCHELL 1985: 476, and MERONEY 1945: 386, cited in MITCHELL). This uncertainty and variability can even be taken as a suggestion that in the spoken language the distinctions were even more fitfully observed. In any case, even in its written form the system was already what MERONEY (1945: 386) describes as "a stage of compromise between Germanic and Modern English", and by the latter stage, English had become the sole Germanic language not to attend regularly to this distinction. Given that these forms were widely used in high literary English into the 1800s, it is difficult to place exactly when they passed out of spoken English. However, all would agree that they are no longer current today and have not been for, at the very least, the last two centuries.4 The loss of the "motion away from" forms did not in itself lead to a loss in encoded meaning, since the word from was recruited to serve the same purpose: heonan became from here, etc. However, motion towards a destination is often contained within the semantics of a verb of motion, and in these cases, English, as so often elsewhere, took the route of leaving the nuance to context:5

4

5

Shortly after writing this I noticed thence used in nonfiction prose here and there. However, the English /o-forms hither and thither are definitely impossible in Modern English beyond the ironic or deliberately archaic. Pages for data in sources listed in fn. 2: Dutch: 125, Yiddish: 59, Swedish: 115-116, Icelandic: 97, Faroese: 57-61; Afrikaans data from EKSTEEN (1997).

36

JOHN Η . MCWHORTER

Eng here

OE

Ger

Du

Fr6

Yi

Sc7

Ic

Fa

Afr

her hider

hier heri-)

hjir hjirhinne

hi aher

här hit

there

thcer thider

dort m-)

der derhinne

dort ahin

där dit

where

hwcer hwider

wo wohin

hier hiernaartoe daar daamaartoe waar waar..heen

wer werhime

vu vuhin

var vart

hir hingath thar thangath hvar hvert

her higar har hagar hvar

hier heima -toe daar daarnatoe waar waarheen

Table 2: Directional adverbs in Germanic Certainly the distinction is not entirely foreign to English, as in in the house versus into the house. Moreover, colloquial Englishes often make the distinction variably with where as in Where is she at? versus A: We're going now. B: Where to? But this is hardly the case with most applicable adverbs, and the point remains that it is obligatory in the standard not to mark the distinction on any adverb: hither, thither, and whither are strictly archaic words foreign to even the highest registers of Modern English. Moreover, arguably, the absence of the distinction is grammatical to all English speakers - it is difficult to imagine any Anglophone context on earth where asking Where? rather than Where to? would sound nonnative or clumsy. Other Germanic languages differ slightly in the degree to which the distinctions are obligatory ( T F F I R S M A ' s Frisian grammar has the use of hjir with the come-verb as grammatical, and Faroese even has an outright gap with hvar), to their scope of application within the grammar (German's conventionalization of the conjunction of her- and hin- with prepositions and verb particles being an extreme), and in where particular instances fall in terms of register. But in all of the languages, the distinction applies to a wide array of adverbs, and is a robust aspect of their grammars (to my knowledge, even nonstandard varieties). We cannot simply classify this loss as a mere symptom of the erosion of morphology in Old English. For one, the "motion" forms in Old English differed in shape from the locational reflexes far beyond the mere addition of an affix. If morphological loss were the smoking gun here, then we might expect, for example, hid and heon to have resulted, still distinct from her (> here). In any case, too often Germanic languages have maintained this distinction despite vast morphological losses: Afrikaans is the most pointed demonstration, followed by Mainland Scandinavian and Dutch.

3.6. Be with past participles A hallmark of Germanic (and Romance) is the use of the verb to be with a large subset of intransitive verbs in the perfect: German er hat gegessen 'he has eaten', 'he ate', er ist gekommen 'he has arrived', 'he arrived'. Of course, the precise domain of intransitives to which be applies varies across the languages, but the basic 6 7

Frisian data in this table from PETER TIERSMA (July 2001 p.c.) The Mainland Scandinavian varieties differ in the fashions and extents to which they indicate direction in their adverbs, but the distinction is overall very much alive in both Danish and Norwegian.

37

What happened to English?

distinction is retained even in Afrikaans (Ze zijrt vertrokken 'They have left' [PONELIS

1993: 444]) and Yiddish Ikh bin geblibn Ί stayed' [LOCKWOOD 1995:

83]). Accordingly, Old English marked this distinction with the verbs beon and wesan: (14) hu sio lar Lcedengediodes cer dissum how the learning Latin-language.GEN before this.DAT afeallen wees fall-away.PART

was

'How the learning of Latin was fallen away before this' [MITCHELL & ROBINSON 1986: i l l ]

Yet as so often with typical Germanic constructions, already in Old English the usage was apparently in flux, with habban encroaching on the domain of beon and wesan. MITCHELL (1985: 302-304) suggests that none of the attempts over the years to delineate a principled semantic distinction between the use of a verb with habban as opposed to beon or wesan withstand scrutiny, and questions whether the documentation even indicates a grammaticalized fee-perfect, as opposed to a typical use of a fee-verb with Stative adjectivals. By the 1500s, the use of be in the perfect had largely shrunk to the change-of-state class of intransitives such as come, become, arrive, enter, run, and grow (TRAUGOTT 1972: 144, RLSSANEN 1999: 213): (15) And didst thou not, when she was gone downstairs, desire me to be no more so familiarity with such poor people? [Henry IV, II.i.96 cited in TRAUGOTT 1972] By the early 1800s, the fee-perfect was largely limited to become (STRANG 1970: 100), and finally vanished. The pathway English followed is striking given that the development of feeperfects was an innovation in Germanic and Romance rather than an inheritance from Proto-Indo-European, and in many languages the domain of fee has spread rather than contracted over time - cf. the varying extents of its application across Western European languages in SORACE (2000). English instead reversed this pathway of grammaticalization, thus joining the elimination of inherent reflexive marking and the derivational prefixes fee- and ge- in rendering English the Germanic language with the least overt marking of valence. Obviously we cannot lay this change at the feet of inflectional loss, nor can it be subsumed under the rubric of the drift towards analyticity. Typically, the disappearance of the fee-perfect is attributed to the recruitment of fee as a marker of the passive (MUSTANOJA 1960: 501, TRAUGOTT 1972: 145, MITCHELL 1985: 299, RlSSANEN 1999: 213). There is even comparative support for this explanation, in

the fact that Swedish, the only Germanic language other than English that lacks the fee-perfect (Vi har rest till Spanien förr 'We have gone to Spain before' [HOLMES & HINCHLIFFE 1997: 100]), has also recruited its fee-verb vara to mark the passive (cf. RlSSANEN 1999: 215).

But the causal relationship here is not absolute. Icelandic, too, forms its passives with its fee-verb vera (eg vor barinn Ί was hit') and in the perfect uses have with both transitives and intransitives (eg hef komid Ί have come') (KRESS 1982: 148-

38

JOHN Η . MCWHORTER

149). Yet Icelandic also uses vera with intransitive verbs of motion and change-ofstate to connote the resultative: ig er kominn Ί am come, I am here', and the class of verbs used this way is large (ibid. 152-153) - the strategy is by no means the recessive, marginal archaism that it was, for example, by Middle English. It also bears mentioning that Swedish is unique even in Mainland Scandinavian in lacking a fee-perfect. Danish and Norwegian retain it: Danish Barnet er kommen 'The child has come' (THOMAS 1911: 133-134), Norwegian Han er reist 'He has left' (STRANDSKOGEN & STRANDSKOGEN 1986: 21). This highlights the general tenacity of this feature in a language once it arises (although it happens to be receding in Norwegian). Yet note that English parallels the Swedish exception - as always, if a Germanic language other than English happens to opt for context where family tradition calls for being explicit, English will have done the same (the Faroese lack of a ίο-marked where being another example). 3.7. Passive marking with become Another Germanic tribal marker is the use of a verb "become" to form the passive (German Die Tür wird geschlossen 'The door is (beeing) closed'), this including Afrikaans with its word (Die trui word gebere 'The jersey is put away'). While Swedish indeed uses its fee-verb vara in the passive, it does so in conjunction with its verb 'to become' bli: vara conveys a "stative" passiveness in line with its semantics: Himlen är täckt αν moln 'The sky is covered in cloud', while bli conveys more perfective semantics: Han blev päkörd αν en bil 'He was run down by a car' ( H O L M E S & H I N C H L I F F E 1997: 109). Icelandic, too, retains its verba along with its use of vera 'to be' in the passive, in a division of labor in which verba is the marked, but hardly marginal, member ( K R E S S 1982: 150). As frequently, Old English followed the Germanic pattern, but already in rather atrophied fashion. Beon and wesan were already easing out weorpan in the passive, and MITCHELL (1985: 3 2 4 - 3 3 5 ) rather spiritedly refutes common claims that this was instead a regularized distinction between perfective semantics conveyed by weorpan and stative ones by beon and wesan (along the lines of Swedish's vara and bli). By Middle English, the weorpan-passive is not just recessive, but nonexistent (RISSANEN 1999: 325).

Modern English has innovated the marking of passive with get {He got hit) and have (He had his hair cut), but both are pragmatically constrained, encoding especial activeness on the part of the subject, with the Aave-passive essentially a causative. Overall, English remains the only Germanic language without a lexical item dedicated to expressing a pragmatically neutral manifestation of the passive. Only in Icelandic is it even grammatical to use the fee-verb to say He was kicked, and even it has retained verba alongside; meanwhile, properly speaking Swedish has recruited vara into a subdomain of the passive, retaining bli to distinguish the "true" passive as opposed to its more stative manifestations. Once again, English opts for underspecification where its sisters insist on dotting the i's and crossing the f s .

39

What happened to English?

3.8. V2 All Germanic languages but English have verb-second word order, including Afrikaans. The languages differ in their particular manifestations of the phenomenon, often classified as "asymmetric" when V2 occurs only in root clauses and "symmetric" when V2 occurs in both root and subordinate clauses. Which type of V2 Old English manifested is disputed (VAN KEMENADE 1987 versus PLNTZUK 1991), but it is uncontested that it was a V2 language: (16)

On twam in

pingum

hceafde

two.DAT thing.DAT have.PAST

God

pees

marines

God

this.GEN

man.GEN

sowie gegodod. soul endow.PART 'God had endowed this man's soul with two things.' [FISCHER et al. 2000: 107]

V2 in English begins a decline in the fifteenth century, and is essentially dead by the seventeenth (JACOBSSON 1951, NEVALAINEN 1997). T h e question obviously

arises as to why. One current consensus links the loss to the erosion of verbal inflectional morphology. The general assumption is that V2 results from verb movement, specifically to C (DEN BESTEN 1983), and inflection-based accounts of V2 loss suppose that the erosion of verbal morphology led to the verb staying in place rather than moving upwards in its clause (e.g. FISCHER et al. 2000: 135-136). But overall, the explanations offered in this case lack explanatory power or falsifiability. For example, an inflection-based account of the loss of V2 presumes that the very small difference in degree of verbal inflection between Mainland Scandinavian and English determined that the former would preserve V2 while the latter would lose it. Yet this difference consists only of the fact that Mainland Scandinavian marks the present in all persons and numbers with -r (Swedish jag arbetar Ί work') while English inflects in the present only the third person singular. This would appear to attribute a profound configurational transformation to a rather minor discrepancy, especially given that discourse studies show that the third person singular is by far the most frequent in speech (e.g. GREENBERG 1966: 45), such that the inflected form in English constitutes a disproportional component of input to learners. Where is the cut-off point that determines how "weak" inflection must be before it conditions a change in movement rules? This question is all the more pressing given that in reference to a related process, ROBERTS (1993), ROHRBACHER (1994) and others have argued that loss of verbal inflection in both English and Mainland Scandinavian led to the loss of verb movement to I in subordinate clauses (the movement taken to be subsumed by Vto-C movement in matrix clauses in these languages). A demonstration case is English, in which previously the verb moved ahead of the negator, adverbs, and other elements, as in ...if I gave not this accompt to you from 1557 (GÖRLACH 1991: 223).

40

JOHN Η . MCWHORTER

But there appear to be no principled accounts to date which motivate the differing fates of V-to-I and V-to-C movement in English: precisely why did inflectional loss preserve V2 but eliminate V-to-I in Mainland Scandinavian, while eliminating both in English? We might be tempted to suppose that for some reason, the small difference in degree of inflection was indeed responsible for the very specific effect of preserving V2 but not V-to-I movement in Mainland Scandinavian. But then the latest evidence suggests that the decisive causal link is solely between "strong" inflection and verb movement; when inflection is "weak", then the verb may or may not move (ROBERTS 1999: 292). The Kronoby Swedish dialect preserves V-to-I despite the inflectional erosion (PLATZACK & HOLMBERG 1989). Meanwhile, KROCH & TAYLOR (1997) argue that when verbal inflection eroded in English dialects in the north (under Scandinavian influence), the result was not the loss of V2, but a mere change in its configuration, from symmetric (the authors assume PINTZUK'S analysis of "native" Old English V2) to asymmetric. BAPTISTA (2000) shows that there is evidence of verb movement in Cape Verdean Creole despite its having but a single verbal inflection. ROHRBACHER (1994) notes that Faroese verbs do not move to I despite robust plural inflection on verbs. He thereby surmises that "strong" inflection entails overt marking of the first and second persons in at least one number of at least one tense. But this stipulation is rather ad hoc, contradicting the centrality of the third person singular in discourse, and would seem to have been invalidated by dialects like Kronoby Swedish. It would appear that, simply, the correlation between inflection and verb movement is a rather loose one. While work on the relationship of overt morphology to verb movement continues to be refined (e.g. BOBALIJK & THRAINNSON 1997), it seems clear that the link is too weak in itself to offer a conclusive explanation for what happened in English. LlGHTFOOT (1997: 268-269) argues that this is not problematic for generative diachronic syntacticians, given that their enterprise is to use language change to illuminate the effects of synchronic parameters; under this constrained conception, the reasons for the changes are irrelevant: "Sometimes the concern with explanation is excessive.. .such things happen for various reasons which are often of no particular interest to grammarians". Thus LlGHTFOOT prefers to simply chart changes like the loss of V2 in terms of input gradually depriving learners of "triggers" motivating the setting of the appropriate parameters. However, the reason for the disappearance of the "trigger" for V2 is crucial to this thesis, and I suggest that it is less obscure, or "contingent", as LlGHTFOOT (1997) has it, than it might seem. In becoming the only Germanic language without V2, English opted for what can be argued to be the less complex syntactic configuration. Despite its air of "linguocentricity" when argued by an Anglophone, there is a great deal of evidence that SVO is a universally unmarked order. KAYNE (1994) is an articulate generative demonstration. Pidginization and creolization data also support SVO as a "universal" order. Most of the creators of Kenya Pidgin Swahili spoke closely related Bantu languages with preverbal object inflections, but the pidgin replaces these with free pronouns placed after the verb: Standard Swahili nita-m-piga (I-FUT-him-hit) Ί will hit him', Kenya Pidgin Swahili mimi na-pigaye (I AORlST-hit him) (HEINE 1975: 9). Creoles tend to be SVO regardless of the word

What happened to English?

41

order of their substrate languages, such as Berbice Dutch Creole, formed between speakers of Dutch and the SOV Niger-Congo language Ijo (KOUWENBERG 1994). In eliminating V2, English eliminated a feature requiring the operation of a rule

moving the verb to C, a feature whose supplementary character in general is illustrated by the typological rarity of the V2 feature beyond Germanic. Inflectioncentered, syntax-internal accounts have yielded stimulating explanations proper for various variations upon the manifestation of V2 (such as the observation that the obligatoriness of the filling of COMP is crucial in determining word order asymmetries between matrix and subordinate clauses). But for the complete elimination of verb movement, the best they have provided to date are loose correlations. I suggest, along with DANCHEV (1997), that only a larger, contact-based explanation can surpass this obstacle. 3.9. The disappearance o/thou By the 1700s, the originally plural you had replaced thou in Standard English. Recent research on court documents from the northeast suggest that in spoken English, you was already the conventional second-person singular form as early as the late 1500s, with thou used only in particular marked contexts (HOPE 1994). HOPE suggests that the wider use of thou in literary sources such as Shakespeare may have been a conservatism that spoken English had moved beyond. This development is typically discussed within the larger context of the use of second-person plural pronouns in formal address to single persons across Europe. But STRANG'S (1970: 139) comment that "Such a use, once introduced, must snowball", while obviously apt, does not explain why the "snowballing" went so far in English as to leave it the only Germanic language which lost a distinct second-person singular pronoun altogether. The usual result of the well-known development of "T-V" forms was for the V form to encroach ever more upon the realm of familiarity - but all of the other Germanic languages nevertheless retain the familiar form. If anything, the modern development has been towards the reassertion of the Τ form within the democratizing ideological tendencies of the post-Enlightenment age. Yet during just this period, English relegated thou to the archaism of the religious register. As STRANG notes (1970: 140), it might not have persisted even here if the King James Bible had not happened to reproduce to such an extent the usage of TYNDALE, who wrote in the early 1500s when thou was still in current use. Clearly, neither inflectional loss nor a drift towards analyticity were related to a change which did not transpire even in Afrikaans, and which even highly isolating languages rarely display. Even the inquisitive undergraduate is often given to ask, when exposed to the T-V pronoun issue, why English went as far as to eliminate thou entirely while German retained its du, etc. Often the professor can only offer an articulate shrug. But in fact, this development correlates with the eight we have seen so far in rendering English less overspecified than its sisters. Obviously the lack of a number distinction in the second person occasions no significant communicational difficulties. However, every single Germanic language but English has preserved this distinction, frill though it is, to the present day. To be sure, Dutch

42

JOHN Η . MCWHORTER

and Frisian have lost the du-cognate itself, but have nevertheless "exapted" other material to maintain a T-V distinction. English stands as unique among its sisters in having eschewed the nuance altogether. 3.10. The disappearance of man English began with the usual Germanic endowment of an indefinite pronoun man, grammaticalized enough to have eroded phonetically to me by Middle English: (17) Ac me ne auh to bien hersum bute of gode. but one NEG ought to be obedient except in good 'But one should not be obedient except in good things.' [RLSSANEN 1997: 520]

But the form rapidly disappeared, essentially gone by the late fourteenth century. The original form man had split off to connote 'a human being': (18) thanne man forgiet that he seien sholde. thanne bed when one forget what he say .INF should, then be.3s his tunge alse hit cleued were. his tongue as it stuck were 'When a person forgets what he should say, his tongue is as if it were stuck.' [RISSANEN 1997: 520]

Yet this usage as well did not eventually survive, and Modern English recruits you, they, and people in the function once served by its birthright man. Meanwhile, the only Germanic languages lacking a distinct indefinite pronoun are Icelandic and Faroese (Icelandic recruits madur 'men' [KRESS 1982: 113], while Faroese uses man but only as a Danicism, preferring to use tu 'you' and teir 'they' like English [LOCKWOOD 1955: 125]). Afrikaans does not retain Dutch's men, but instead uses the colloquial Dutch 'n mens (< 'a person') - and the tendency is to shorten this to mens, creating what PONELIS (1993: 224) analyzes as a new pronoun. RISSANEN (1997) surmises that the disappearance of me was due to two factors. If I read him correctly, his proposition that me "was too weak for the subject position" leads us to ask why similarly weak forms survive across Germanic, such as the unemphatic forms in Dutch (je for jij, etc.).8 His other suggestion is that homonymy with the oblique me, especially with impersonal verbs (me semeth), was a factor. Yet RISSANEN himself elsewhere notes that "admittedly, homonymy and disambiguation offer only a shaky argument for the loss of forms", and this is especially a propos here. Crucially, it is likely that oblique me and indefinite me were not homophones: the former retained a long vowel while the vowel in me was likely a weakened one such as schwa (MEIER 8

To the extent that we analyze forms of this kind as clitics, we might ask why indefinite me could not have simply evolved into one rather than simply disappearing.

What happened to English?

43

1953: 179-182 cited in RLSSANEN). Thus especially given that we are dealing with phonetically similar but hardly identical forms, the homonymy argument is weakened in view of, for example, det ([de:]) 'it' and de 'they' in Swedish, which like oblique and indefinite me in Middle English occur with the same verbal ending in the present (since Swedish has but one across person and number). While one might argue that the increasing prevalence of dom as 'they' in Swedish responds to this homonymy, note that dej is now established in colloquial Swedish in the second person singular, creating yet another near-homonymy even in the absence of conjugational allomorphs to signal distinction. Then beyond Germanic, there are of course cases like lei used both as 'she' and as a term of polite address in Italian (both with the same third-person verbal ending), and the absence of a number distinction in third person pronouns in spoken French (illils, elle/elles), despite there being no verb endings to distinguish them. And meanwhile, obviously neither of these two explanations would apply to the more robust form man that persisted alongside its more deeply grammaticalized descendant me. One might decide that the disappearance of the distinct indefinite pronoun "just happened". Yet as STRANG (1970: 267) puts it, "No satisfactory allpurpose substitute for it has ever been found", thus leaving the absence of a single indefinite pronoun to combine with the concurrent disappearance of thou in rendering the English pronominal array the most context-dependent of any Germanic language. Of course, in larger view, this is but one of the many developments we have seen that are all symptomatic of a clear trend towards underspecification - far beyond the realm of inflectional loss.

4. Implications That one language can be overspecified in a particular area compared to another one is clear to all analysts. However, linguists often claim that overall, languages balance out in terms of complexity of this kind (e.g. EDWARDS 1994: 90; BLCKERTON 1 9 9 5 : 6 7 ; O'GRADY e t al. 1 9 9 7 : 6 ; CRYSTAL 1 9 8 7 : 6 - 7 ) . H o w e v e r , t h i s a s -

sumption has rarely been actively investigated, and those examining the issue more closely have tended to conclude that, more properly, all languages are complex to a considerable, but not equal, degree (CROWLEY 2000, GIL 2001). In this light, it must be reiterated that while I do not claim that English is a "simple" language cross-linguistically speaking, I do claim that English is significantly less complex than its sisters in the senses of overspecification and complexity presented in section 2. The contrast I refer to is illustrated by this comparative table.

44

JOHN Η. MCWHORTER

Inherent reflexives External possessors Gender beyond noun Loss of prefixes Directional adverbs be-perfect Passive become verb V2 Singular you Indefinite pronoun

G X X X X X X X X X X

Du X X X X X X X X X X

Y X X X X X X X X X X

Fr X X X X X X X X X X

s X X X X X X X X X

Ν X X X X X X X X X X

Da X X X X X X X X X X

I X X X X X X X X X

Fa X X X X X X X X X

A X

X X X X X X X

OE X X X X X X X X X X

Ε

Table 3: Losses in English compared to other Germanic languages Certainly English has developed features of overspecification and complexity that its sisters have not. My thesis is that there was a significant disruption in the transmission of English at one point in its history, but this scenario nothing less than requires that after this, English would naturally drift into its own elaborations as all languages do. Thus English is unique amidst Germanic in its do-support, in its grammaticalization of the present participle with be as an obligatory present tense marker, thereby rendering the bare verb zero-marked for habituality, in its distinction of shades of futurity with will, going to, and be + present participle. However, there are four further observations: 1. It is difficult to imagine any complex of original Germanic features that would yield a chart where any Germanic language but English proved to have lost all of the features while the other languages - including English - had retained all or most of them. See M C W H O R T E R (2002: 248-249) for demonstration via Proto-Germanic reconstructions. 2. It is my impression that English has not developed so many innovative overspecifications and complexifications as to "balance out" with its sisters. 3. This is especially the case given that in addition to retaining so much more of the original Germanic heritage, the other languages have also, like English, developed overspecifications and complexities of their own, such as the highly conventionalized modal particles in German, Dutch, Frisian, and Mainland Scandinavian, tone in Swedish, noun incorporation in Frisian, etc. 4. It is significant that the only Germanic language even approaching English in how much original structure it has shed is Afrikaans, whose history was decisively impacted by widespread second-language acquisition of Dutch. LASS

(1987: 318) finds related results in an analogous table:

45

What happened to English?

Ε Grammatical gender Rich case marking 3-person verb morph. Sing./pl. verb morph. Subjunctive Strong/weak verbs Strong/weak adjectives V2 Long/short vowels Robust umlaut

X

X X

Fr Χ Χ Χ χ

Du Χ

χ χ χ χ

Χ Χ Χ Χ

Α

Χ Χ

Χ Χ Χ

G Χ Χ Χ Χ Χ Χ Χ Χ χ χ

Υ χ χ χ

χ χ χ

I χ χ χ χ χ χ χ χ χ χ

Fa Χ Χ χ χ χ χ χ χ χ χ

Ν Χ

Da Χ

S χ

χ χ χ χ χ

Χ Χ Χ Χ Χ

χ χ χ χ χ

Table 4: Losses in English (LASS 1987) L A S S ( 1 9 8 7 : 3 3 2 ) also makes a comparison of English with Afrikaans and Yiddish specifically, treating these three as uniquely impacted by contact. He again shows that English surpasses the other two in loss of Germanic features (German and Dutch are included for contrast):

Ε Loss Loss Loss Loss Loss Loss

of past vs. perfect of infinitive suffix of sentence-brace9 ofSOV of gender of verb marking

X X X X X

A X X

Y X

G

Du

X X X X

Table 5: Losses in the "contact heavy" Germanic languages (LASS 1987) In LASS' first table, English and Afrikaans are even, and are close to it in the second. However, LASS focuses on inflectional features, which automatically brings English and Afrikaans to the fore since both have been shorn of so much inflectional morphology. My treatment, in expanding our perspective to features beyond inflection, demonstrates that Afrikaans is in fact much more conservative than English overall. Importantly, the relative innovativeness of English has consisted not only of the transformation of original materials, but of simply shedding much of it where none of its sisters have.

5. Reassessing the Scandinavian impact One response to the contrast in Table 3 might be to suppose that English just shed these features by chance. Certain members of a given family have long been observed to be more innovative than others. There are often no apparent "reasons" 9

This refers to the separation of auxiliary and main verb by objects, adverbials, etc. (Ich habe den Brief gestern geschrieben Ί wrote the letter yesterday').

46

JOHN Η . MCWHORTER

for this, with linguists treating the difference in conservativity as parents might treat personality differences among their children - they just "are". While authors such as TRUDGILL (1989, 1996) have noted that widespread acquisition as a second language often renders a language less phonetically and morphologically complex than its relatives, CROWLEY (2000) gives clear demonstration of a case where a complexity differential is unlikely to trace to contact effects. The Oceanic language Ura is less complex morphophonemically and inflectionally, and marks fewer categories overtly with its inflections, than its sister Sye despite Ura having no documented history of use as a lingua franca. Many might prefer to treat English as the "Ura" of Germanic. However, it is my impression that the relative prevalence of contact in English's history (while hardly presuming that its sisters developed in isolation) offers a compelling suggestion that English is instead closer to being the "Swahili" of Germanic. Swahili is unusual among Bantu languages in lacking tone, having shed a number of noun class categories and verbal inflections, and simplified its nonverbal predication system, this being due to its being used more as a second language than as a first over the past several centuries (MCWHORTER 1994). While hardly a Creole, Swahili gives clear evidence of a slight but decisive break in transmission in its timeline, and few Bantuists would classify it as "innovative" due strictly to internal factors. We will reject two other possible explanations. It is unlikely that English's departure from the Germanic template was a function of its isolation on an island generally, this kind of isolation is associated with relative conservativity, Icelandic and Faroese being the obviously pertinent cases here. Furthermore, it cannot be upheld that the standardization of English was the culprit. While one might propose that koineization between speakers of various dialects of English in the London area led to a streamlining of the language, this would leave the question as to why similar processes did not leave Standard French (developed amidst contact between dialects brought to the Ile-de-France region) or Russian (developed amidst contact between several dialects in Moscow), similarly simplified in comparison to their sisters. In any case, most of the features I have covered are defunct or on the ropes long before the 1400s. It would seem that the task facing us is to choose between possible contact explanations.

5.1. Evaluating the alternatives To investigate where a contact-induced break in transmission might have occurred in English, we are faced with a choice between three potential culprits. I accept the arguments of T H O M A S O N & K A U F M A N (1988) and others that the Norman occupation cannot have caused a significant break in the transmission of English. Not only were invaders so slight in number and so removed from the general population incapable of creolizing English, but I would propose more specifically that they could not even have had any significant impact on the language beyond the lexical (and thus, also derivational affixes).

What happened to English?

47

I also find it implausible that Low Dutch varieties10, imported from 1150 to about 1700 by Flemish immigrants and agents of the Hanseatic League (VERECK 1993, THOMASON & KAUFMAN 1988: 321-325), had any significant impact on English

grammar. There is nothing in accounts of the Hanseatic League like PAGEL (1983) to suggest that Low German speakers were thick enough on the ground in any one place to influence general speech patterns; on the contrary, such sources indicate that the Hanseatic agents were generally housed in their own quarters of town. Moreover, if Low Dutch speakers' non-native English had influenced the language, we would also expect that their lexical contribution would extend into the grammatical realm. Yet it did not to any significant or conclusive extent (BENSE 1939), which would include the -kin derivational suffix noted by THOMASON & KAUFMAN (1988: 325). These authors (ibid. 323) also propose that Low Dutch lent several dialects of English an enclitic object form for she and they, /as/, but VOSS (1995) is rather compellingly skeptical of this account, as well as others concerning sound changes. Overall, despite VIERECK'S (1993) and DANCHEV'S (1997: 101) useful

call to consider that Low Dutch had greater impact on English than traditionally thought, I suspect that in the final analysis, we will find that this impact was no more significant than that of Dutch upon English in the New Amsterdam colony in early America. Large numbers of settlers almost always impact the lexicon of the language they encounter, but their numbers alone do not entail transformation of the grammar. This leaves us with the Scandinavian invaders, who arrived in England in the late ninth and early tenth centuries; roughly speaking, Danes settled in the northeast area that came to be called the Danelaw while Norwegians later settled in the northwest. The Scandinavians settled among the general population rather than ruling from afar as an elite as the Norman French did, often marrying Anglo-Saxon women. The lexical impact of this contact is hardly in dispute. The Normans imposed their language mainly "from above", lending mainly content words, generally hewing towards the formal realm. In contrast, the Scandinavian legacy extended to grammatical words such as they, their, them, though, both, same, against, and others since lost, and content words of even the homeliest nature (neck, window, knife, skirt, happy, etc.). This alone indicates a highly intimate contact scenario. Our question, then, is whether the Scandinavian impact upon English went even deeper. I propose that there are indeed indications that the Scandinavian invasions were responsible for the very decrease in overspecification and complexity that I presented in section 3.

5.2. Support for Scandinavian influence: timing The first piece of evidence pointing specifically to the Scandinavian impact is evidence that many of these features persisted longest in regions where Scandinavians did not settle, or in those where place-names suggest that they were less 10

The term includes Low German, which some contemporary observations suggest was processed as "the same language" at the time (PETERS 1987: 80).

48

JOHN Η . MCWHORTER

robustly represented. Under this analysis, we might assume that these features would be represented more vividly in these regions even today if a particular dialect that emerged in the Danelaw - Standard English - had not diluted the dialects in question to such a degree. To put a point on it, if England had remained a preindustrial society where literacy was largely limited to elites, then we might hypothesize that English varieties outside of the Danelaw would remain "cardcarrying" Germanic dialects. The evidence here is hardly conclusive, given the paucity of information on regional grammars (as opposed to lexicon and phonology) after Standard English came to prevail in writing. However, it serves as one plank in my general argument. a) Inherent reflexive marking. UPTON et al. (1994: 488) show sit thee down and variants persisting in nonstandard dialects throughout England; this can perhaps be seen as due to these varieties retaining a variability typical of less "focused" speech varieties in contrast to the standard. However, it is perhaps notable that the other attestations (ibid. 488-489) are attested only in regions where Scandinavian settlement was relatively thin according to WAKELIN (1972: 20): laid him down and laid her down in Cheshire, Derbyshire, and Staffordshire; and they play(en) them (they disport themselves) in Lancashire and Derbyshire. b) External possessor marking. UPTON et al. (1994: 488) record wring the neck of him - rather analogous to the locative Scandinavian configuration of this strategy in Derbyshire and, pointedly, Cornwall, where there was no Scandinavian settlement at all. (It may also be relevant that they record he's pulling that chap his leg in Yorkshire, the western region of which was rather thinly settled by Scandinavians.) c) Grammatical gender. Scandinavian settlement was concentrated in the northeast; LASS (1992: 113) notes that "loss of inflection [in the noun phrase] is earliest in the east and north, the south and west generally remaining more conservative". The loss of gender on the definite determiner began in the north and then was attested only variably in the southwest Midlands in late 1200s (LASS 1992: 113); meanwhile, the old three-way distinction persisted in the south at this time (STRANG 1970: 267), and traces of gender marking hung on in Kent as late as 1340 (LASS 1992: 113).

Indicatively, of the regions where UPTON et al. (1994: 486-487) record the use of he/him and/or she/her to refer to objects (the relevant objects being of the same range for both genders), sixteen out of twenty are outside of Scandinavian concentration (most south of the Danelaw). There is also evidence that remnants of the gender distinction persisted in especially grammaticalized form in the Viking-free Southwest. In the early 1200s, there is occasional gender marking of inanimates in documents (STRANG 1970: 265), and this was still attested in the late nineteenth century in BARNES* description of Dorset dialect. Here, BARNES (1886: 17-18) describes what we might expect to have evolved in English, a distinction between a common and neuter genders, which he terms "personal" and "impersonal". The "personal" class includes "full shapen things, or things to which the Almighty or man has given a shape for an end" and includes people, living things, and tools: thus of a tree one said He's α-cut down

What happened to English?

49

but for water, It's α-dried up. The distinction extended to demonstratives (theäse vs. this, thik vs. that). In reference to inflection in general, UPTON et al. (1994: 4 9 0 ) record the infinitive marker -(e)n in Derbyshire, Westmorland, and Lancashire, areas of thin Scandinavian settlement (with the exception of putten, which occurs in Yorkshire and Lincolnshire as well). They record infinitives ending in [-i] only in Monmouthshire, Kent, Cornwall, Somerset, Devon, Lancashire, Derbyshire, Shropshire, and Staffordshire.11 d) Directional adverbs. UPTON et al. ( 1 9 9 4 : 9 2 ) find variations on come hither only in regions south of the Danelaw except Lincolnshire; they find Where to is it? or Where is it to? only in Monmouthshire, Somerset, Wiltshire, Cornwall, Devon, and Dorset (ibid. 502). e) V2. For one, V2 persists in Kentish documents while eroding elsewhere in English (KROCH & TAYLOR 1997: 312). These authors' analysis even offers more finegrained evidence that the erosion of V2 was caused by Scandinavian. They argue (1997: 3 1 8 - 3 2 0 ) that the transition from "symmetric" (verb movement in both matrix and subordinate clauses) to asymmetric (verb movement only in the matrix clause) V2 was one occasioned by the inflectional loss that Scandinavian settlers' incomplete acquisition of English led to. Under the assumption that complete loss of V2 would only be possible with sharp diminution of inflection, the implication of KROCH & TAYLOR'S analysis is that the transition to asymmetric V 2 was an intermediate stage between the original configuration and today's. By extension, this implies that the disappearance of V2 was initiated by language contact in the Danelaw.

5.3. Support for Scandinavian influence: transfer THOMASON & KAUFMAN'S (1988: 3 0 2 - 3 0 3 )

verdict on the evidence of transfer

from Scandinavian in English is the following: "The Norse influence on English was pervasive, in the sense that its results are found in all parts of the language; but it was not deep, except in the lexicon. Norse influence could not have modified the basic typology of English because the two were highly similar in the first place." This conclusion is justified for their masterful argument on the basis of the features they treat as part of their "Notification package". However, other evidence suggests that the Norse influence was indeed deep, and that THOMASON & KAUFMAN'S Norsification package, comprising mostly phonological and morphological

11

If by chance any of these attestations of [-i] were actually remnants of the southwestern -y iterative marker (Idle chap, He'll do nothln but fishy [spend his time in fishing] [BARNES 1886: 25]), then even this - the preservation of an overt aspectual marker alien to English dialects northward supports the idea that contact with Old Norse left English shorn of overspecification beyond what would have been the case amidst internal evolution.

50

JOHN Η . MCWHORTER

traits, various grammatical items, and lexical ones, constitutes but a subset of the relevant evidence. According to tradition, T H O M A S O N & K A U F M A N assume that Old English and Old Norse were too similar for structural transfer to be particularly relevant to analyzing their effect upon one another in contact. Yet a closer look at Old Norse reveals grammatical differences crucial to this thesis. Specifically, no fewer than six of our ten losses in English have parallels in aspects of Old Norse hitherto overlooked, to my knowledge, in studies of the Danelaw situation. a) Inherent reflexives. In Old Norse, the reflexive use of the first person and third person pronouns had eroded and affixed to the verb as a suffix, the latter used in all persons but the first singular: bindomk Ί tie myself, bysk 'you arm yourself, staksk 'he stabbed himself (HEUSLER 1950: 107). This extended to inherent reflexives: peir setiask nipr 'they sat down', er hefnezk ά honom 'you revenge yourselves upon him', petta felsk honom vel ί skap 'that felt good to him, agreed with him' (ibid. 1 3 7 - 1 3 8 ) . Use of free pronouns in the reflexive was not unknown, but was largely restricted to dative forms (hann brn sir 'he wandered'), but even here was variable (hann brask was also grammatical) (ibid. 138). The variability of inherent reflexive marking by Middle English may have been the result of a tendency for Old Norse speakers to omit the inherent reflexive pronouns in speaking English. Use of the full pronoun was the marked case in their native language, and meanwhile, English lacked an equivalent of the reflexive inflection. English specialists often note that inflections could be shed in the Danelaw because they were incidental to communication; this would also have been true of overt inflectional marking of inherent reflexivity. Thus Old Norse speakers would have been comfortable refraining from marking the distinction when speaking English - j u s t as Modern English speakers are. b) External possessors. Scandinavian is unique in Germanic in encoding external possession with the locative rather than the dative (cf. (7), (8), (9)). Faced with this disjunction between external possessors encoding in the two languages, Scandinavians may have taken the choice of eliminating the distinction altogether, given that it was not vital to the expression of the relevant concepts. This is a common process in the development of koines, for example, where often the koine lacks features which were present in most or all of the source varieties but expressed with different morphemes or strategies (cf. discussion in 5.4. below). c) Derivational prefixes. The absence of the core Germanic verbal prefixes in Icelandic and Faroese traces back to Old Norse (HEUSLER 1950: 4 0 ) . It could be that the rapid eclipse of these prefixes in English was due to the absence of cognates in Old Norse speakers' native language. d) fee-perfect. In Old Norse, as in Modern Icelandic, the fee-perfect largely connoted the resultative and the passive (HEUSLER 1950: 136). Its use as a true perfect was limited to a few intransitive verbs such as 'to go': ek em gengenn Ί have gone' (BRENNER 1882: 129). This may have been the spur for the disappearance of the feature in English, including the possibility MITCHELL ( 1 9 8 5 : 3 0 2 - 3 0 4 ) notes that

What happened to English?

51

what has been analyzed as a be-perfect in Old English may have actually been only a resultative construction. e) Äecome-passive. In Old Norse, the passive was usually expressed with vera 'to be'. Werpa 'to become' was relatively restricted in meaning, encoding roughly the saliently active semantics of Modern English's gei-passive: (19) Peer saker skal fyrst d0ma, er fyrra sumar varp the.PL issue.PL should first adjudge.iNF that last summer become eige um d0mt not to judgement 'First the complaints should be decided upon that didn't manage to get to 12 judgment last summer.' [HEUSLER 1 9 5 0 : 1 3 7 ; translation mine] This restricted usage may have been a cause of the otherwise mysterious absence of English's weorpan after Old English. f) Indefinite man. Icelandic and Faroese lack a man-cognate, and already in Old Norse it was recessive (HEUSLER 1 9 5 0 : 1 4 7 ) , generally replaced by impersonal verb constructions or third person verbs without pronouns. This may possibly have set in motion a de-emphasis on the use of Old English's man-cognate that eventually resulted in its disappearance early in Middle English. English is traditionally considered closest to Dutch and Frisian. But in many aspects where Old Norse and its descendants depart from the Germanic pattern, they parallel English, although having overall retained a great deal more of their Germanic legacy. No other West Germanic language offers anything approaching this congruency with English in as many as six out of our ten features, which the author did not even expect to find upon beginning this investigation. This suggests a specific effect from Scandinavian. To wit, in reference to THOMASON & KAUFMAN'S statement that "Norse influence could not have modified the basic typology of English because the two were highly similar in the first place", the similarity dilutes somewhat on closer examination, to an extent that may well have had significant impact on the development of Modern English.

5.4. Support for Scandinavian influence: reduction Generally, discussion of the Scandinavian impact on English is largely restricted to transfer effects: sound changes and lexical borrowings. I have attempted to add to this argument with structural transfers. As such, however, it might be objected that my attempt to expand our conception of the Scandinavian impact is hindered where Old English and Old Norse have parallel structures, under the assumption that these features would be retained. But as THOMASON & KAUFMAN ( 1 9 8 8 : 1 2 9 , passim) note, when a population shifts to a new language and their rendition of the language ousts the original na12

Original German: Die Klagen soll man zuerst aburteilen, über die man im vorigen Sommer nicht zum Urteil gelangte.

52

JOHN Η. MCWHORTER

tive one, then transfer effects often occur alongside evidence of incomplete acquisition - that is, outright reduction rather than transfer. Importantly, these effects often occur even where the languages in question have parallel or cognate structures. 5.4.1. Traditionally accepted: inflection and grammatical morphemes Scholars of English's history traditionally recognize this in the area of inflection, where it is often reconstructed that when Old English and Old Norse speakers were confronted with equivalent but phonetically differing inflections, they simply shed them to ease communication. Yet for our purposes we must refine the conception somewhat. Most writers appear to suppose that this must have been a two-way affair, communication occurring as speakers of both languages shaved away their native inflections, leaving an analytic "common denominator" comprehensible to both. However, it is questionable that this would have allowed significant communication. LASS ( 1 9 8 7 : 5 2 ) notes that claims that Old English and Old Norse were virtually mutually intelligible are exaggerated, and I would agree - this claim strikes me as reflecting the written medium, allowing us to "wrap our heads around" phonetic correspondences at leisure, and concealing the obstacles to comprehension effected by accent, intonation, and morphophonemics. Analyzed side by side, Old English and Old Norse suggest no more mutual intelligibility than that between, for example, Serbo-Croatian and Bulgarian or Spanish and Italian. Moreover, mere contact between closely related varieties hardly entails the dilution of infection: in many cases, the local language acquires substitutes for native inflections without the overall amount of inflection decreasing, as in the case of Rusyn, formed through contact between dialects of Ukrainian, Slovak, and SerboCroatian (HARASOWSKA 1999), or the borrowing of inflections in Australian cases like Warndarang from Nunggubuyu described in HEATH ( 1 9 7 8 ) . Inflectional loss is symptomatic of a more specific phenomenon: non-native acquisition. Specifically it is more likely that the inflectional loss resulted from Old Norse speakers' incomplete renditions of a new language (cf. DANCHEV 1 9 9 7 : 9 0 ) . THOMASON & KAUFMAN'S ( 1 9 8 8 : 1 1 9 - 1 2 0 ) stipulation that interference effects are likely to be stronger in cases of rapid shift over "one or two generations" is also relevant, given that these authors are of the opinion that Norse was no longer spoken in England after 1100 (ibid. 282), disappearing within one or two generations of the reintegration of the Danelaw into the English polity (ibid. 284, 286). The grammatical morphemes make a further case for this. A language generally only borrows grammatical, as opposed to lexical, items in contexts of bilingualism (amidst at least a subset of the population). The question here is whether bilingual acquisition took place more in one direction than another. In that light, the Danes and Norwegians were newcomers, who were largely illiterate and thus did not impose their language in writing or in government, and eventually gave it up. Obviously the impulse towards bilingualism would have been much stronger among the Vikings than among the English. In this light, items such as they/themJtheir, both, same, etc. would stand as remnants of Scandinavian brought into the English spoken by, first, immigrants, and then just as plausibly by generations bilingual in Old Norse and Old English. It is rarely if ever acknowledged that after the first genera-

What happened to English?

53

tion, descendants of the Scandinavian invaders may well have begun to speak English as well as Norse even among themselves, as is typical of shifting speakers, with Norse and English perhaps taking their place in a kind of diglossia. As such, for Norse descendants born in England, an English sprinkled with the occasional Norse grammatical item would be comprehensible and even a marker of, if we may, "ethnic" kinship. 5.4.2. Extending the paradigm These observations, then, suggest that English could easily have shed even features it shared with Old Norse, if its fate was determined by non-native speakers' proficient yet approximate rendition, as the timing and transfer evidence suggest. My proposal is that we extend this conception beyond lexicon and morphology. The relevant comparison is with koines, such as SIEGEL'S ( 1 9 8 7 : 1 8 5 - 2 1 0 ) description of the koine Hindustani of Fiji, developed amidst contact between speakers of a range of divergent varieties of the Hindustani dialect complex. Certainly the koine has picked lexical and morphological features "cafeteria" style from assorted dialects. However, the koine is not on balance as elaborated as any of these dialects, instead being markedly simpler in the formal sense than any of these, even when the dialects all display the feature in question. This includes the elimination of the three-way formality distinction in second-person pronouns ( S E G E L 1987: 199), a general tendency towards replacing synthetic with analytic forms, and a strong tendency to replace SOV order with SVO (ibid. 198). (This last is especially indicative regarding the loss of V2 in English.) A less well-known example is the Riau dialect of Indonesian described by DAVID GIL ( 1 9 9 4 ) . While developed amidst speakers of languages closely related to Indonesian and to one another, Riau Indonesian has vastly simplified Indonesian's valence-marking morphological apparatus - of much greater semantic and syntactic import than mere gender or person/number markers - and other grammatical features with close cognates in the languages spoken by its creators, such as Minangkabau. These cases demonstrate that even when languages in contact are closely related, reduction can play as significant a part in the outcome as exchange of materials.13 In processes of linguistic accommodation, speakers often contribute a less overspecified and complexified rendition of their language. The extreme manifestation here is Foreigner Talk; a less radical manifestation would be the tendency for crea13

In both the Fiji and Riau cases, the scenario is complicated somewhat by the possible impact of pidgin varieties in the contact situations: here, Pidgin Hindustani and Bazaar Malay, respectively. However, this may be an artifact of our chronological proximity to cases such as these. It is hardly inconceivable that there was a "pidgin English" spoken by the first wave of Scandinavians, reflecting the limitations of adult language-learning capabilities. (This is especially the case given the observed fact that Old English and Old Norse were not essentially dialects of the same language as is often claimed.) It may have been the progeny of these invaders who acquired a more proficient English, nevertheless pervaded with Norse features. Our microsociolinguistic knowledge of how interference through language shift (in THOMASON & KAUFMAN'S terminology) proceeds is currently limited, most cases having occurred beyond the purview of written history. Certainly this is, and will likely remain, the case with the Notification of Bnglish over a thousand years ago.

54

JOHN Η . MCWHORTER

tors of Creoles to contribute a "streamlined" version of their native grammars to the new language, such as among Oceanic speakers (KEESING 1988, SIEGEL et al. 2000). Koine scenarios exhibit an analogous process, and there is no theoretical reason that this would not have been the case in the Danelaw - even beyond inflection. This reductive tendency could have operated in three ways: 1) In many cases, the corresponding feature in Old Norse was already variable or marginal: inherent reflexives, the £>e-perfect, the become-passive and the indefinite man. When Old Norse utilized a feature as a marked alternate, the general tendency against overspecified features in non-native acquisition could have led these features to be marginalized even further in Old Norse speakers' rendition of English. There is no stipulation that the features would vanish immediately. Instead, non-native acquisition would initiate a decline in frequency of occurrence which would eventually weaken and eliminate the sufficient "trigger", in LLGHTFOOT's terms, for its transmission to new generations. 2) In other cases, already in Old English features were ripe for marginalization in a contact situation, because they occurred only variably. This is the case with inherent reflexives, external possessors, directional adverbs, the be-perfect, V2, and indefinite man. It is tempting to hypothesize that the optionality of these features in Old English was itself symptomatic of Scandinavian contact. But this is unlikely given that most Old English documentation from 900 A.D. onward is in the West Saxon dialect, outside of Scandinavian settlement. Crucially, however, this hardly means that these features were used in more tightly conventionalized fashion beyond West Saxon. On the contrary, this kind of variability is attributable to speech varieties that are primarily spoken, as was true of Old English - while it comes down to us in writing, it was spoken in a society where literacy was largely limited to an elite. The especial prevalence of "unfocused" conventions is familiar to any linguist working on an unwritten indigenous language, and even characteristic of other early Germanic varieties like Gothic and Old Norse itself. As such, we can assume that the "softness" of the relevant rules was typical of English dialects in general, not just West Saxon. In any case, this would mean that such features were especially vulnerable to falling below the line of acquirability, in the sense that LLGHTFOOT has outlined, in a setting of widespread non-native acquisition. A parallel would be an aspect of the Irish Gaelic spoken by second-language learners in Ireland (HENRY & TANGNEY 1999). Equative predications in Irish Gaelic mark the subject as accusative, place it clause-finally, and use a distinct best,rb (Is dochtuir e [be doctor him] 'He is a doctor'). English-dominant learners replace this quirk with a construction using the more commonly encountered be-verb ta, which occurs sentence-initially as verbs typically do in this VSO language, and takes a nominative subject: Ta se dochtuir [be he doctor]. 3) Finally, other features may have been eliminated even when robust in both languages. That this was the case with grammatical gender on nouns appears to be common consensus; I suggest that it was also the case with the collapse of gen-

What happened to English?

55

der distinctions on the determiners. A non-native speaker of English, confronted with three forms of the determiner corresponding to gender assignments that often conflicted between Old English and Old Norse, would plausibly have made do with a single gender-neutral marker. I also suggest that the gradual disappearance of V2 falls under the same rubric, an analogy being the weakness of SOV order in Fiji's Hindustani koine, created by people all of whom spoke languages where SOV was obligatory. The eclipse of thou admittedly occurred a tad on the late side (the 1500s at the earliest), but the uniqueness of the development - and the implausibility of linking it to any cultural factor local to England - suggest an external explanation. 5.5. The nature of evidence Obviously at this point we yearn for particulars of the sociolinguistic terrain in the Danelaw. However, this was a largely illiterate setting of primarily oral communication lost to the ages; the writing that has come down to us was, in broad view, a marginal activity aimed at a small elite. Even the numerical size of the Scandinavian presence remains controversial - and most likely unknowable. THOMASON & KAUFMAN (1988: 276, 3 6 1 - 3 6 3 ) cite SAWYER'S (1971) arguments that there is no evidence that vast streams of peasant settlers followed in the wake of the invaders as often supposed. However, I propose that the clustering of the changes in the Danelaw region, combined with the correspondences between the losses in English and features absent or marked in Old Norse, implicate the Scandinavian invasions despite the lacunae in the socio-historical record. SAWYER may be correct that this impact was due more to prestige than brute numbers, and the questions here naturally evoke an interest in the precise mechanics of how a language is transformed by widespread second-language acquisition. But at the end of the day, a pre-literate society 1200 years removed from us is unlikely to ever shed much light on this question. To serve our interest in the on-the-ground aspects of this kind of process, there are various living or amply recorded analogues to what I propose happened in the Danelaw (e.g. see LACROIX 1967 on vehicular Fula or SIEGEL 1987 on pidginized and koineized Hindustani). My aim has been to demonstrate that even if we will likely never know precisely how the transmission of English was temporarily diluted in northeastern England, there is evidence allowing us to know that this did happen. Clearly, my hypothesis can be vastly strengthened with closer engagement with the document literature, corpus-based techniques being particularly germane. However, I propose that such engagement would be a worthwhile pursuit, as it offers a possible answer to a contrast which must otherwise stand as a pressing anomaly to wit, Table 3.14 14

As such, the contrast that English presents with its sisters in broad view suggests that it would be more elegant not to surmise, contra THOMASON & KAUFMAN (1988: 303), that the inflectional simplifications in northerly English dialects were already taking place to a significant degree before Norse contact.

56

JOHN Η . MCWHORTER

6. Conclusion My aim in this paper has been to show that the difference between English and its sister languages comprises a much larger array of features than merely loss of inflection or a "tendency towards analyticity", and that the larger awareness that English has moved towards a "different typology" than its relatives (e.g. LASS 1987: 317-332) is traceable to a causal factor rather than being a random "uninteresting" development. As many readers may have noticed, my list of features is hardly complete. Radiating outward from the core of losses that leave English unique or close to it in Germanic, there are other losses that English shares with a few Germanic languages, their interest being that where a subset of Germanic languages have departed sharply from the original Germanic "typology", English never fails to be a member. Examples include subjunctive marking (lost or marginal in Mainland Scandinavian, Dutch, Frisian, and Afrikaans) and verb-final word order in subordinate clauses (lost in Mainland Scandinavian, Icelandic and Faroese and Yiddish). English is also alone in Germanic in lacking a strong-weak distinction in adjectival inflection (even Afrikaans retains this based largely on syllable count), which I omitted from my presentation to detract from the traditional focus on inflectional loss. I must reiterate that my claim is neither that English is a Creole nor that elaborations have been alien to its life-cycle. There would be no reason to suppose that the tendency towards structural reduction that the Scandinavian impact caused would persist indefinitely. On the contrary, after this partial break in transmission, English "got back on its feet" and restored the tendency towards drift into elaborations that is typical of natural languages. The result was developments such as do-support, marking of aspectual distinctions unknown in its sisters across the waters, etc. However, our full cognizance of these features must be seen in conjunction with the striking losses that English suffered in the centuries during and after the Scandinavian invasions, which left it much more underspecified than even Afrikaans. To assume that these changes were merely a matter of chance would seem to require that the slings and arrows of outrageous fortune had led at least one other Germanic language to shed, rather than transform, so much of the Germanic legacy. Otherwise, the conclusion would seem almost unavoidable that the English timeline was decisively influenced by what TRUDGILL (2001) has termed, in apt and savory fashion, "the lousy language-learning abilities of the human adult". Abbreviations ACC CL DAT DEF FUT GEN IMP INF

accusative clitic pronoun dative definite future tense genitive impersonal infinitive

NEG NOM OBJ PART PAST PL PREP S

negation nominative object participle past tense plural preposition singular

What happened to English?

57

References On the use of the definite article with nouns of possession in English. Uppsala: Appelbergs Boktiyckeriaktiebolag. BAILEY, CHARLES-JAMES & MAROLDT, KARL ( 1 9 7 7 ) : The French lineage of English, in: MEISEL, JÜRGEN (ed.), Pidgins, Creoles. Languages in contact. Tübingen: Narr, 21-53 BAKKER, PETER (1997): A language of our own. New York: Oxford University Press. BAPTISTA, MARLYSE ( 2 0 0 0 ) : Verb movement in four Creole languages: A comparative analysis, in: MCWHORTER, JOHN (ed.), Language change and language contact in pidgins and Creoles. Amsterdam: Benjamins, 1-33. BARNES, WILLIAM ( 1 8 8 6 ) : A glossary of the Dorset dialect with a grammar of its word shapening and wording. London: Trübner & Co. BENSE, J.F. (1939): A dictionary of the Low-Dutch element in the English vocabulary. The Hague: Martinus Nijhoff. BESTEN, HANS DEN ( 1 9 8 3 ) : On the interaction of root transformations and lexical deletive rules. On the formal syntax of the Westgermania, in: ABRAHAM, WERNER (ed.), Papers from the Third Groningen Grammar Talks, Groningen, January 1 9 8 1 . Amsterdam: Benjamins, 47—131. BICKERTON, DEREK ( 1 9 9 5 ) : Language and human behavior. Seattle: University of Washington Press. BOBALUK, JONATHAN & THRAINNSON, HOSKI ( 1 9 9 7 ) : Two heads aren't always better than one, in: Syntax 1, 3 7 - 7 1 . BRENNER, OSCAR (1882): Altnordisches Handbuch. Leipzig: T . O . Weigel. BRIGHT, WILLIAM ( 1 9 5 7 ) : The Karok language. (University of California Publications in Linguistics 13). Berkeley: University of California Press. COMRIE, BERNARD (1981): The languages of the Soviet Union. Cambridge: Cambridge University Press. CROWLEY, TERRY ( 2 0 0 0 ) : Simplicity, complexity, emblematicity and grammatical change, in: SIEGEL, JEFF, 175-193. CRYSTAL, DAVID ( 1 9 8 7 ) : The Cambridge Encyclopedia of Language. Cambridge: Cambridge University Press. CURME, GEORGE O. (1952): A grammar of the German language. New York: Frederick Ungar. DAL, INGERID (1966): Kurze deutsche Syntax. Tübingen: Max Niemeyer. DALTON-PUFFER, CHRISTIANE ( 1 9 9 5 ) : Middle English as a Creole and its opposite: on the value of plausible speculation, in: FISIAK, JACEK (ed.), 3 5 - 5 0 . DALTON-PUFFER, CHRISTIANE ( 1 9 9 6 ) : The French influence on Middle English morphology: a corpusbased study of derivation. Berlin: Mouton de Gruyter. DANCHEV, ANDREI ( 1 9 9 7 ) : The Middle English creolization hypothesis revisited, in: FISIAK, JACEK (ed.), 7 9 - 1 0 8 . DEGRAFF, MICHEL (ed.) (1999): Language creation and language change. Cambridge, M A : The M I T Press. DELETANT, DENNIS & ALEXANDRESCU, YVONNE ( 1 9 9 2 ) : Teach yourself Romanian. Lincolnwood, IL: NTC Publishing Group. DOMINGUE, NICOLE Z . ( 1 9 7 7 ) : Middle English: another Creole?, in: Journal of Creole Studies 1, 8 9 - 1 0 0 . DONALDSON, BRUCE ( 1 9 9 7 ) : Dutch: a comprehensive grammar. London: Routledge. EDWARDS, JOHN (1994): Multilingualism. London: Penguin. EKSTEEN, L.C. (1997): Major dictionary/Groot woordeboek. Cape Town: Pharos. FALTZ, LEONARD Μ. (1985): Reflexivization: a study in universal syntax. New York: Garland. FISCHER, OLGA (1992): Syntax, in: BLAKE, NORMAN (ed.), The Cambridge history of the English language. Vol. 2. Cambridge: Cambridge University Press, 207-408. FISCHER, OLGA ( 1 9 9 7 ) : Infinitive marking in Late Middle English: transitivity and changes in the English system of case, in: FISIAK, JACEK (ed.), 1 0 9 - 1 3 4 . FISCHER, OLGA; VAN KEMENADE, ANS; KOOPMAN, WILLEM & VAN DER WURFF, W I M (eds.) (2000): The syntax of early English. Cambridge: Cambridge University Press. FISIAK, JACEK (ed.) ( 1 9 9 5 ) : Linguistic change under contact conditions. Berlin: Mouton de Gruyter. FISIAK, JACEK (ed.) ( 1 9 9 7 ) : Studies in Middle English linguistics. Berlin: Mouton de Gruyter. FOLEY, WILLIAM A . & VAN VALIN, ROBERT ( 1 9 8 4 ) : Functional syntax and universal grammar. Cambridge: Cambridge University Press. GIL, DAVID ( 1 9 9 4 ) : The structure of Riau Indonesian, in: Nordic Journal of Linguistics 1 7 , 1 7 9 - 2 0 0 . AHLGREN, ARTHUR ( 1 9 4 6 ) :

58

JOHN Η . MCWHORTER

(2001): Creoles, complexity, and Riau Indonesian, in: Linguistic Typology. Introduction to Early Modern English. Cambridge: Cambridge University Press. GORDON, E . V . ( 1 9 2 7 ) : An introduction to Old Norse. Oxford: Clarendon. GREENBERG, JOSEPH ( 1 9 6 6 ) : Language universals, with particular reference to feature hierarchies. The Hague: Mouton. HABERLAND, HARTMUT ( 1 9 9 4 ) : Danish, in: KÖNIG & VAN DER AUWERA (eds.), 3 1 3 - 3 4 8 . HAIMAN, JOHN (1983): Iconic and economic motivation, in: Language 59, 781-819. HAMP, ERIC (1965): The Albanian dialect of Mändres, in: Die Sprache 11,137-54. HARASOWSKA, MARTA ( 1 9 9 9 ) : Morphophonemic variability, productivity, and change: the case of Rusyn. Berlin: Mouton de Gruyter. HASPELMATH, MARTIN ( 1 9 9 8 ) : How young is Standard Average European?, in: Language Sciences GIL, DAVID

GÖRLACH, MANFRED ( 1 9 9 L J :

20, 2 7 1 - 2 8 7 . HASPELMATH, MARTIN ( 1 9 9 9 ) : External possession in a L. & BARSHI, IMMANUEL (ed.), External possession.

European areal perspective, in: PAYNE, DORIS Amsterdam: Benjamins, 1 0 9 - 1 3 5 . HAUGEN, EINAR (1981): Language fragmentation in Scandinavia: revolt of the minorities, in: HAUGEN, EINAR (ed.), Minority languages today. Edinburgh: University Press, 100-119. HARASOWKA, MARTA ( 1 9 9 9 ) : Morphophonemic variability, productivity, and change: The case of Rusyn. Berlin: Mouton de Gruyter. HEATH, JEFFREY ( 1 9 7 5 ) : Some functional relationships in grammar, in: Language 5 1 , 8 9 - 1 0 4 . HEATH, JEFFREY (1978): Linguistic diffusion in Arnhem Land. Canberra: Australian Institute of Aboriginal Studies. HEATH, JEFFREY (1981): A case of intensive lexical diffusion, in: Language 57,335-367. HEINE, BERND (1975): Some generalizations of African-based pidgins. Paper presented at the International Conference on Pidgins and Creoles, Honolulu. HENRY, ALISON ( 1 9 9 5 ) : Belfast English and Standard English: dialect variation and parameter setting. New York: Oxford University Press. HENRY, ALISON & TANGNEY, DENISE ( 1 9 9 9 ) : Functional categories and parameter setting in the second-language acquisition of Irish in early childhood, in: DEGRAFF, MICHEL (ed.), 2 3 9 - 2 5 3 . HEUSLER, ANDREAS ( 1 9 5 0 ) : Altisländisches Elementarbuch. Heidelberg: Carl Winter. HILTUNEN, RISTO ( 1 9 8 3 ) : The decline of prefixes and the beginnings of the English phrasal verb. Turku: Turun Yliopisto. HOLMES, PHILIP & HINCHLIFFE, IAN ( 1 9 9 7 ) : Swedish: an essential grammar. London: Routledge. HOPE, JONATHAN ( 1 9 9 4 ) : The use of thou and you in Early Modern spoken English: evidence from depositions in the Durham ecclesiastical court records, in: KASTOVSKY, DIETER (ed.), Studies in Early Modern English. Berlin: Mouton de Gruyter, 1 4 1 - 1 5 1 . JACOBSSON, BENGT ( 1 9 5 1 ) : Inversion in English: with special reference to the Early Modern English period. Uppsala: Almqvist & Wiksell. KAYNE, RICHARD (1994): The antisymmetry of syntax. Cambridge, MA: MIT Press. KEESING, ROGER ( 1 9 8 8 ) : Melanesian Pidgin and the Oceanic substrate. Palo Alto, CA: Stanford University Press. KELLER, RUDI ( 1 9 9 4 ) : On language change: the invisible hand in language. London: Routledge. KEMMER, SUZANNE (1993): The middle voice. Amsterdam: Benjamins. KÖNIG, EKKEHARD & VAN DER AUWERA, JOHAN (eds.) (1994): The Germanic languages. London: Routledge. KÖNIG, EKKEHARD & HASPELMATH, MARTIN ( 1 9 9 7 ) : Les constructions Ä possesseur externe dans les languages d'Europe, in: FEUILLET, JACK (ed.), Actance et valence dans les languages de I'Europe. Berlin: Mouton de Gruyter, 525-606. KOUWENBERG, SILVIA (1994): A grammar of Berbice Dutch Creole. Berlin: Mouton de Gruyter. KRESS, BRUNO ( 1 9 8 2 ) : Isländische Grammatik. München: Max Hüber. KROCH, ANTHONY & TAYLOR, ANN ( 1 9 9 7 ) : Verb movement in Old and Middle English: dialect variation and language contact, in: VAN KEMENADE & VINCENT (eds.), 2 9 7 - 3 2 5 . LACROIX, PIERRE-FRANCOIS ( 1 9 6 7 ) : Quelques aspects de la disintegration d'un systeme classificatoire (peul du sud de l'Adamawa). La classification nominale dans les langues negro-africaines.

What happened to English?

59

(Colloques intemationaux du Centre National de la Recherche Scientifique). Paris: Editions du Centre National de la Recherche Scientifique. LASS, ROGER ( 1 9 8 7 ) : The shape of English. London: J . M . Dent & Sons. LASS, ROGER ( 1 9 9 0 ) : How to do things with junk: exaptation in language evolution, in: Journal of Linguistics 2 6 , 7 9 - 1 0 2 . LASS, ROGER ( 1 9 9 2 ) : Phonology and morphology, in: BLAKE, NORMAN (ed.), The Cambridge history of the English language (Vol. 2 ) . Cambridge: Cambridge University Press, 2 3 - 1 5 5 . LIGHTFOOT, DAVID ( 1 9 9 7 ) : Shifting triggers and diachronic reanalyses, in: VAN KEMENADE & VINCENT (eds.), 2 5 3 - 2 7 2 . LOCKWOOD, W.B. (1955): An introduction to Modern Faroese. Copenhagen: Munksgaard. LOCKWOOD, W . B . ( 1 9 9 5 / - Lehrbuch der modernen jiddischen Sprache. Hamburg: Buske. LYONS, JOHN (1968): Introduction to theoretical linguistics. Cambridge: Cambridge University Press. MARCHAND, HANS ( 1 9 6 9 ) . The categories and types of present-day English word-formation: a synchronic-diachronic approach. München: Beck. MATISOFF, JAMES A. (1991): Areal and universal dimensions of grammaticization in Lahu, in: TRAUGOTT, ELIZABETH C . & HEINE, BERND (eds.), Approaches to grammaticalization (Volume 2). Amsterdam: Benjamins, 383-453. MATTHEWS, WILLIAM K. (1956): The Livonian element in modern Latvian, in: WOLTNER, MARGARETE & BRÄUER, HERBERT (eds.), Festschrift fiir Max Vasmerzur 70. Geburtstag. Berlin: Veröffentlichungen der Abteilung für slawische Sprachen und Literaturen des Osteuropa-Instituts an der Freien Universität Berlin, 307-318. MCWHORTER, JOHN H . ( 1 9 9 4 ) : From focus marker to copula in Swahili. Special session on historical issues in African Languages, in: MOORE, KEVIN E . ; PETERSON, DAVID A. & WENTUM, COMFORT (eds.), Proceedings of the 2tfh annual meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, 57-66. MCWHORTER, JOHN H . ( 2 0 0 2 ) : What happened to English?, in: Diachronica 1 9 , 2 1 7 - 2 7 2 . MERONEY, HOWARD M . ( 1 9 4 5 ) : The early history of "down" as an adverb, in: Journal of English and Germanic Philology 4 4 , 3 7 8 - 3 8 6 . MITCHELL, BRUCE ( 1 9 8 5 ) : Old English syntax, Vol. 1. Oxford: Clarendon Press. MITCHELL, BRUCE & ROBINSON, FRED C . ( 1 9 8 6 ) : A guide to Old English. Oxford: Blackwell. MITTWOCH, ANITA ( 1 9 9 0 ) : On the distribution of bare infinitive complements in English, in: Journal of Linguistics 2 6 , 1 0 3 - 1 3 1 . MUSTANOJA, TAUNO F. (1960): A Middle English syntax, Vol. 1. Helsinki: Soci6t6 Nfophilologique. MUYSKEN, PIETER ( 1 9 9 7 ) : Media Lengua, in: THOMASON, SARAH G . (ed.), Contact languages: a wider perspective. Amsterdam: Benjamins, 365-426. NEVALAINEN, TERTTU ( 1 9 9 7 ) : Recycling inversion: the case of initial adverbs and negators in Early Modem English. A Festschrift for Roger Lass on his sixtieth birthday, in: Studia Anglica Posnaniensa 3 2 , 2 0 3 - 2 1 4 . O'GRADY, WILLIAM; DOBROVOLSKY, MICHAEL & ARONOFF, MARK ( 1 9 9 7 ) : Contemporary linguistics: an introduction (third edition). New York: St. Martin's Press. PAGEL, KARL ( 1 9 8 3 ) : Die Hanse. Braunschweig: Westermann. PAUL, HERMANN (1880): Principien der Sprachgeschichte. Halle: Niemeyer PEITSARA, KIRSTI ( 1 9 9 7 ) : The development of reflexive strategies in English, in: RISSANEN, MATTI; KYTO, MERJA & HEIKKONEN, KJRSI (eds.), Grammaticalization at work. Berlin: Mouton de Gruyter, 2 7 7 - 3 7 0 . PETERS, ROBERT ( 1 9 8 7 ) :

Das Mittelniederdeutsche als Sprache der Hanse, in: URELAND, STURE (ed.), Sprachkontakt in der Hanse: Aspekte des Sprachausgleichs im Ostsee- und Nordseeraum. Tübingen: Niemeyer, 6 5 - 8 8 . PINTZUK, SUSAN ( 1 9 9 1 ) : Phrase structures in competition: variation and change in Old English word order. University of Pennsylvania PhD dissertation. PLATZACK, CHRISTER ( 1 9 8 6 ) : C O M P , I N F L , and Germanic word order, in: HELLAN, LARS & KOCH CHRISTENSEN, KIRSTI (eds.), Topics in Scandinavian syntax. Dordrecht: Reidel, 1 8 5 - 2 3 4 . PLATZACK, CHRISTER & HOLMBERG, ANDERS (1989): The role of AGR and finiteness in Germanic V O languages, in: Working Papers in Scandinavian Syntax 43,51-76. PONELIS, FRITZ (1993): The development of Afrikaans. Frankfurt: Peter Lang.

60

JOHN Η . MCWHORTER

The evolution of early Standard English: the creolization hypothesis, in: Studia Anglica Posnaniensia 1 4 , 6 9 - 8 5 . RISSANEN, MATTI ( 1 9 9 7 ) : Whatever happened to the Middle English indefinite pronouns?, in: FISIAK, JACEK (ed.), 5 1 3 - 5 2 9 . RISSANEN, MATTI (1999): Syntax, in: LASS, ROGER (ed.), The Cambridge history of the English language. Vol. 1. Cambridge: Cambridge University Press, 187-331. ROBERTS, IAN ( 1 9 9 3 ) : Verbs and diachronic syntax: a comparative history of English and French. Dordrecht: Kluwer. ROBERTS, IAN ( 1 9 9 9 ) : Verb movement and markedness, in: DEGRAFF, MICHEL (ed.), 2 8 7 - 3 2 7 . ROHRBACHER, Β . ( 1 9 9 4 ) : The Germanic languages and the full paradigm: a theory ofVto I raising. University of Massachusetts, Amherst PhD dissertation. SAWYER, P.H. ( 1 9 7 1 ) : The age of the Vikings. London: Edward Arnold. SIEGEL, JEFF (1987): Language contact in a plantation environment. Cambridge: CUP. SIEGEL, JEFF ( 2 0 0 0 ) : Processes of language contact: studies from Australia and the South Pacific. Quebec: AGMV Marquis. SIEGEL, JEFF; SANDEMAN, BARBARA & CORNE, CHRIS ( 2 0 0 0 ) : Predicting substrate influence: tensemodality-aspect marking in Tayo, in: SIEGEL, JEFF, 75-97. SMITH, HENRY ( 1 9 9 2 ) : "Dative sickness" and abstractness, in: RAUCH, IRMENGARD; CARR, GERALD F . & KYES, ROBERT L. (eds.), On Germanic linguistics: issues and methods. Berlin: Mouton de Gruyter, 2 8 3 - 2 9 7 . SORACE, ANTONELLA ( 2 0 0 0 ) : Gradients in auxiliary selection with intransitive verbs, in: Language POUSSA, PATRICIA ( 1 9 8 2 ) :

76, 859-890. STRANDSKOGEN, ASE-BERIT & STRANDSKOGEN, ROLF ( 1 9 8 6 ) :

Norwegian: an essential grammar. London: Routledge. STRANG, BARBARA M.H. ( 1 9 7 0 ) : A history of English. London: Methuen. THOMAS, EDWARD JOSEPH ( 1 9 9 1 ) : Danish conversation-grammar. Heidelberg: Julius Groos. THOMASON, SARAH GREY & KAUFMAN, TERRENCE ( 1 9 8 8 ) : Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. TLERSMA, PIETERM. ( 1 9 8 5 ) : Frisian reference grammar. Dordrecht: Foris. TRAUGOTT, ELIZABETH C . ( 1 9 7 2 ) : A history of English syntax: a transformational approach to the history of English sentence structure. New York: Holt, Rhinehart & Winston. TRUDGILL, PETER ( 1 9 8 9 ) : Contact and evolution in linguistic change, in: BREVIK, LEIV EGIL & JAHR, ERNST HAKON (eds.), Language change: contributions to the study of its causes. Berlin: Mouton de Gruyter, 227-237. TRUDGILL, PETER ( 1 9 9 6 ) : Dialect typology: Isolation, social network and phonological structure. Towards a,social science of language, in: GUY, GREGORY; FEAGIN, CRAWFORD; SCHIFFRIN, DEBORAH & BAUGH, JOHN (eds.), Variation and change in language and society. Amsterdam: Benjamins, 3-21. TRUDGILL, PETER ( 1 9 9 9 ) : Language contact and the function of linguistic gender, in: Posnan Studies in Contemporary Linguistics 3 5 , 1 3 3 - 1 5 2 . TRUDGILL, PETER (2001): Reply to McWhorter, in: Linguistic Typology 5,3. UPTON, CLIVE; PARRY, DAVID & WIDDOWSON, J. D . A. (eds.) (1994): Survey of English dialects: the dictionary and grammar. London: Routledge. VAN KEMENADE, ANS ( 1 9 8 7 ) : Syntactic case and morphological case in the history of English. Dordrecht: Foris. VAN KEMENADE, ANS & VINCENT, NIGEL (eds.) (1997): Parameters of morphosyntactic change. Cambridge: Cambridge University Press. VAN NESS, SILKE (1994): Pennsylvania German, in: KÖNIG & VAN DER AUWERA (eds.), 420-438. VERGNAUD, JEAN-ROGER & ZUBIZARRETA, MARIE-LOUISE ( 1 9 9 2 ) : The definite determiner and the inalienable constructions in French and English, in: Linguistic Inquiry 2 3 , 5 9 5 - 6 5 2 . VIERECK, WOLFGANG ( 1 9 9 3 ) : The medieval European common market and its impact on Middle English, in: Neuphilologische Mitteilungen 9 4 , 6 9 - 7 8 . VISSER, F. ΤΗ. (1963): An historical syntax of the English language, Vol. 1. Leiden: E.J. Brill. Voss, MANFRED (1995): Kent and the Low Countries revisited, in: FISIAK, JACEK (ed.), 327-363. WAKELIN, MARTYN ( 1 9 7 2 ) : English dialects: an introduction. London: Athlone Press.

Agree and agreement: evidence from Germanic* H A L L D 0 R Ä R M A N N SLGURDSSON

(Lund)

1. Theoretical background CHOMSKY (2001a: 3) formulates his understanding of Agree as follows: "We therefore have a relation Agree holding between a and β, where a has interpre-table inflectional features and /?has uninterpretable ones, which delete under Agree." In a clause like the Icelandic (1), the finite verb and the logical subject agree in number:1 (1)

Pad hafa komiö hingaö prir mälvisindamenn.2 it have.3.PL come here three linguists.N.PL 'Three linguists have arrived here.'

The logical subject has interpretable number, whereas the number feature of the finite verb is uninterpretable, as it only repeats the number information of the logical subject and thus does not add anything to the interpretation of the clause. Hence, the number feature of the verb may not be visible at the conceptual interface, where the clause is interpreted, i.e. it must be deleted prior to interpretation (although it must not be eliminated at the articulatory or 'phonological' interface). This deletion or 'neutralization' is achieved by agreement with the postverbal nominative argument. The underlying relation is that of Agree and the 'participants' or elements of the relation are referred to as the probe and the goal: The

1

2

A preliminary version of this work was published in Working Papers in Scandinavian Syntax ( 2 0 0 2 ) 7 0 , 1 0 1 - 1 5 6 . Many warm thanks to CHRISTER PLATZACK for insightful comments on that first version. For numerous and very fruitful discussions on agreement over the years, many thanks to ANDERS HOLMBERG. Thanks also to CECILIA FALK, CEDRIC BOECKX, GUNLÖG JOSEFSSON, JÖHANNA BARBDAL, JOSEF BEYER, LARS-OLOF DELSING, VERNER EGERLAND. Finally, many thanks to WERNER ABRAHAM for his helpful comments and important support. I alone am responsible for the shortcomings of this work. They are also both in the third person. However, third person is 'absence of person', [-lp, -2p], not 'true person' and hence no person agreement is involved (on Icelandic from this point of view, see SIGURBSSON 1 9 9 6 and subsequent work). The participle komiö 'come' takes a non-agreeing, default form when selected by hafa 'have' (see section 2 . 2 . 1 ) . The default form is hompohonous with the agreeing N/A.N.SG form, but I refrain from giving irrelevant morphological information in the glosses; such redundant information only makes it harder for the reader to process the examples, and can easily be found in any Icelandic grammar (e.g. EINARSSON 1 9 4 5 ) .

62

HALLDÖR ARMANN SIGURBSSON

finite verb is a probe, searching for a goal, and when it finds a goal with matching features (the nominative argument), Agree is established and the uninterpretable features of the probe are deleted for the purpose of successful interpretation (CHOMSKY 2 0 0 0 , 2 0 0 1 A ) . 3

However, agreement is an amazingly varied phenomenon, much more varied than this 'simple' approach could ever account for. This is true of even closely related languages like the Germanic ones. Consider the variation illustrated below for English, German, Swedish and Icelandic (agreeing forms are boldface): (2a) They would würden (2b) Sie they would.3PL (2c) De skulle they would (2d) Peir mundu they.N.M.PL would.3PL

be reich rich vara be vera be

rich. sein. be rika. rich.PL rikir. rich.N.M.PL

In view of the fact that some languages are similar to English or Chinese in largely or even completely lacking agreement, it is clear that morphological agreement as such is not a property of Universal Grammar (although its underlying syntactic relation or relations are taken to be universal in the present approach). At the other extreme we have agreement-spreading languages like Swahili and other Bantu languages.4 Consider the Swahili example in ( 3 ) (from C O R B E T T 1 9 9 1 : 4 3 ) : (3)

Kikapu kikubwa class7.basket AGR.large 'One large basket fell.'

kimoja AGR. one

kilianguka. AGR. fell

Any Swahili noun has a prefix that marks its number (SG vs. PL) and noun class (or 'gender', 7 or 12 in all, depending on whether or not singular and plural are classified together). Ki- is one of those markers, taken by e.g. the root -kapu in the singular (class 7). In the plural, it takes vi-, yielding vikapu 'baskets' (class 8). In both cases, certain other elements of the clause must agree with the class/number marker, that is, the marker 'spreads'. Consider ( 4 ) (from C O R B E T T 1 9 9 1 : 4 4 ) : 5 (4)

Vikapu vikubwa class8 .basket AGR.large 'Three large baskets fell.'

vitatu AGR.three

vilianguka. AGR. fell

A somewhat different kind of agreement spreading is found in some Australian languages, for instance Yukulta and (the closely related) Kayardild, two nearly or

3 4

5

Cf. the Principle of Full Interpretation in CHOMSKY 1995. The term 'agreement spreading' is a bit unfortunate, as it makes an implication I'm not really in a position to provide, namely that the agreement features percolate (an implication that is not obviously correct and would in any case need to be substantiated). I have not managed to come up with any better descriptive term. For a discussion of 'defective' agreement in Bantu languages (arising in VS orders as opposed to SV orders), see VAN GELDEREN (1997: 26-28) and the references cited there.

63

Agree and agreement: evidence from Germanic

already extinct Pama-Nyungan languages (ETHNOLOGUE). The Kayardild example in (5) is from BLAKE (2001: 108): (5)

Makuntha

yalawujarrantha

woman.AGR

caught-ntha

dangkakarranguninaantha man's, with. AGR

yakurinaantha fish.AGR

mijilnguninantha. net.AGR

'The woman must have caught fish with the man's net.' Although the suffix -ntha is originally a marker of oblique case, it has developed into a marker of inferential modality, and "should probably be treated as a feature of the verb which spreads to the dependents via concord" (BLAKE 2001: 109). The Swahili and the Kayardild types of agreement seem to be basically of the same nature, although they take place within different parts of the clause. The Swahili type may be sketched in the following manner: (6)

AGRT-[CLASSrbasket AGRRlarge AGR1-one] AGRJ-fell

That is, two processes are presumably involved: 1) DP-intemal concord within the subject phrase. 2) Agreement of the predicate with the subject phrase as a whole. This is thus a similar relation as that between DP-internal number concord in the Germanic languages and agreement of finite verbs and predicative adjectives with the number of a subject DP as a whole. In the Kayardild type, on the other hand, the agreement seems to 'stem from' a modal clausal feature or a 'head' (inferential modality) that presumably commands the whole clause at some level of representation (later attracting the verb): (7)

MODj [woman-AGRj caught-AGRj .. ,]-AGRj

The modal head agrees with its complement (the rest of the clause) and, in addition, the complement shows constituent-internal concord, like the subject-DP in Swahili. As we shall see in sections 2.2. and 2.3 below, constituent-internal concord or agreement spreading is found within both predicates and DPs in some of the Germanic languages as well. Facts of this sort illustrate that agreement cannot be reduced to feature matching of DPs and TPs (as in PESETSKY & TORREGO 2001) or matching between extended V-projections and extended N-projections (initially plausible as that approach may seem). Agreement is much more general than that. Reconsider the Icelandic example in (1), with the partial structure in (8): (8)

[there have-AGR| [come [here [three linguistsi]")]]

It is clear that the finite verb somehow has access to the featural information of the nominative argument, even though the latter is deeply embedded in its predicate. One way to account for this access is to assume that the finite verb can probe

64

HALLDÖR ÄRMANN SIGURDSSON

into its complement, as suggested by C H O M S K Y (2000, 2001a).6 Suppose, however, that more elements than just the verb agree with the 'late' nominative, as in (9): (9a) [there have-AGRj [been [elected-AGRj three linguists!-!]] (9b) [there have-AGRi [been [believed-AGRj [elected-AGR1 three linguistsillll In cases like these, not only the finite verb but also the participles show agreement with the logical subject and hence something more than just probing of the finite verb is going on. As we shall see, multiple agreement of a very similar sort is attested within Germanic, in particular in Icelandic. Given the universalist view of the internal, mental system of language,7 the natural working hypothesis is that non-agreement in languages like English and Chinese is in some sense a 'null-exponent' of the same underlying relations that are signalled by the many kinds of complex and multiple agreement in languages like Swahili and Icelandic.8 In view of the immense variation seen in agreement languages, however, it would seem impossible to generalize over all the various agreement phenomena (see e.g. C I N Q U E 1999; J U L I E N 2002: 258), let alone to also generalize over silent or 'covert agreement'. However, this is exactly what a coherent theory of agreement must do. But, as we shall see, the Probe-Goal Approach cannot be maintained, as a probe-goal relation is much too specific and could thus not be the common denominator. An extremely general or 'atomic' approach is needed, an Occam's razor. I therefore suggest that Agree reduces to Merge and possibly also to Move, inasmuch as Move can be seen as an instantiation of Merge ( S T A R K E 2001, C H O M S K Y 2001b).9 It seems to me, however, that unifying Move and Merge is an oversimplification. Move (or 'internal Merge') does not really add information to the structure in the same way as true ('external') Merge does. Consider the 'pre-movement' structure in (10a) and the corresponding 'post-movement' structure in (10b). F is an element that has been merged to XP, thereby creating the structure FP, α is a category that moves to the edge of this structure and -a is its copy or 'trace':10 (10a)

[FP

(10b) Α

F [χρ ...

Α

... ]]

[FPF[XP...«...]]

However, on standard assumptions, the movement of a must have been triggered by some feature or property in the 'pre-movement' structure in (10a), say by some requirement of F, call it f . If so, the movement-information is already there, in a 6

This is basically the same idea as the case-path idea developed in SIGURDSSON 1989, 1990-1991 (by which the case path opened up a way for the pAi-features of the subject DP 'back to' Infi). For closely related approaches, see PESETSKY & TORREGO 2001, PLATZACK 2001, 2002.

7 8

9

10

FNL or the "faculty of language in the narrow sense" in the terminology of HAUSER et al. 2002. One way of implementing this is to say that agreement is a parametric property of the lexical items involved in an agreement construction. I shall however not detail about technicalities here (but see section 3 for some speculations). In footnote 6 in Derivation by phase, CHOMSKY (2001a: 42) suggests that Concord is a similar relation as that of Agree, albeit a distinct one, "involving Merge alone". I propose, in contrast, that Agree and Concord are the same relation, both involving Merge and feature matching. 'Edge' is here used in an intuitive, non-technical sense (roughly the sense defined in CHOMSKY 2001a: 13). Edge phenomena will be discussed in more detail in sections 2.1.5 and 3.

Agree and agreement: evidence from Germanic

65

sense (by merger of F or f), that is, the movement is predestined by Merge and does, accordingly, not add any information to the structure. It would thus seem that Move is vacuous, and in a sense it is: Typically, movement or word order differences between languages are semantically unimportant.11 - In another sense, however, Move is 'meaningful': movement of an element or a feature a to the edge of FP renders FP amenable to external Merge, as will be discussed in section 3. That is, Move does not reduce to Merge, but it is driven or motivated by its 'needs'. Given that Move is always predestined or motivated by Merge, we can restrict our attention to the latter. Thus, my central hypothesis may be stated as the Agree Condition on Merge: (11) Two objects or elements, X and Y, may be merged only if the relation of Agree holds between them. It is more accurate to say that elements match each other's requirements prior to Merge and agree as a result of Merge. When the need arises, I will use this more accurate terminology, otherwise using Agree as a general term for both matching and Agree itself. Being a condition on Merge, Agree is necessarily a sisterhood relation, not a probe-goal relation. However, there is more to this than just sisterhood: the properties or features of the merged objects also matter, that is, they must agree in a sense to be explicated. Defining the relation of Agree amounts to defining the fundamental problem of compositionality:12 How can two distinct units ever make up a whole that, in turn, may function as a part of another, larger whole? This is the very 'mystery' of language structure, stated in extremely simple terms. It is useful to think of Merge/Agree like chemical reactions and chemical bonds. Atoms cannot merge to form chemical bonds unless they 'agree' in a sense, that is, the negatively charged electrons of one atom must be attractable by the positively charged nucleus of another atom and vice versa. The result is a bond, a larger unit where the subunits share some of their outer shell or 'edge' electrons ('agreement features'). Adopting this 'atom metaphor', I also find it profitable to replace the X'-theoretic notions 'head' and 'specifier' with nucleus and edge. In view of the fact that usual labelling is problematic and must arguably be dispensed with (see STARKE 2001, COLLINS 2002), this is a timely step. See further below and section 2.1.5. Reconsider 'distant agreement' as in (1) and the simple (12): (12) Pad there

hafa komiö mälvisindamenn. have.3.PL come linguists.N.M.PL

For sake of expository ease, let us assume that this clause only has the extremely simple structure in (13): (13) [there [Agr have [come [linguists]]]] 11

12

With respect to grammaticality within any given language, however, movement is of course not vacuous. Parametrized EPP effects may be needed. In a more basic sense than usually considered. The focus here is not on the effect of composition (as in many semantic approaches), but on the more fundamental question of how composition itself is possible in the first place.

66

HALLDÖR ÄRMANN SIGURDSSON

Let us also make the pedagogical (but wrong) assumption that (13) illustrates the exact compositionality of the syntactic structure, that is, reflects its derivational history: the main verb first merging with the nominative argument, yielding [come linguists], the second cycle of Merge simply yielding [have come linguists] and the final cycle yielding [there have come linguists]. What the Agree Condition on Merge says, then, is that the relation of Agree holds not between have and linguists, but between every two objects that have been merged, that is:13 (14a) [come] & [linguists] (14b) [have] & [come linguists] (14c) [there] & [have come linguists] However, the only morphological agreement is that between have and linguists. Thus, there is no straightforward correlation between Agree and visible or audible agreement. More often than not, Merge/Agree is 'silent', in fact, the result of a parametric choice not to express a particular instance of Merge in morphology. Secondly, even when Merge/Agree is not silent, it is often only 'partial' in that it only involves some (phonological) subpart of a merged element. Both these aspects of Merge/Agree need to be understood and accounted for (for some observations, see sections 2.1.5 and 3). Before we start describing and discussing facts from the Germanic languages, a few words on the basic clause structure assumed here are in place. CHOMSKY (1991, 1993) suggested that the Infi head of the clause splits into 'subject agreement' or AgrS, tense or Τ and 'object agreement' or AgrO, the corresponding maximal projections being AgrSP, TP and AgrOP (AgrSP corresponding to IP in older systems). While many have adopted this system, CHOMSKY himself quickly abandoned it (1995: 349ff.): Agr does not have any content at LF and must thus be eliminated from the system. The clausal functional heads in CHOMSKY'S approach have since been only T(ense) and the so-called little v, the simple IP-clause being merely a TP dominating a predication phrase or an aspect phrase, referred to as vP. It seems plausible, if not trivially obvious, that there can be no complex agreement or inflectional elements at the basic syntactico-semantic structure of language, LF: what would they be doing there? In contrast, being meaningful, the subparts of the Infi of the earlier work of CHOMSKY (e.g. 1981, 1986) must obviously 'be there somewhere'. These subparts are not only tense but minimally also mood, person and number.14 Rather than eliminating AgrS, I thus assume that it splits into Pers(on) and Num(ber), thereby adopting the approach in SlGURBSSON 2000 and subsequent work.15 Similarly, the tense 'head' or nucleus must split into 13

14

15

This is only meant to illustrate the general point. Of course, these are not the real objects of combination, since we don't take any functional structure into account. In CHOMSKY'S approach, the relevant elements are situated within Τ and Ν (in a similar fashion as they were situated within Infi in earlier approaches, see e.g. SlGURBSSON 1989: 25). Closely related ideas are developed for Icelandic in HAEBERLI 2002, and have also been suggested for other languages, for instance by SHLONSKY 1989 (according to HAEBERLI 2002: 294, 296), EGERLAND 1996b (who, however, assumes both PersP and AgrSP), and, in a different fashion, POLETTO 2000. For expository reasons, I leave object features out of consideration. - A conceptually related but a different approach explored in PLATZACK 2002 is that DPs contain PersP.

67

Agree and agreement: evidence from Germanic

at least M(ood) and T(ense). This gives us - minimally - the clausal structure in (15).16 (15)

[Cp

c [PERSP Pers [

NumP

Num [MP Μ [ w Τ [vP ν

[VP V]]]]]]

I conceive of all complex structures as molecular, such that for instance the CP is a 'C molecule'. In the interest of readability, however, I keep the conventional phrasal notation, CP, VP and so on (without all single bar categories, though). It is important to notice that this does not imply labelling or projections in the present approach. As for Icelandic, the plural verbal inflection renders rather impressive evidence in favor of this radical splitting of 'Infi' (see SlGURBSSON 2001 for some discussion, also of the more opaque singular forms). This is illustrated in (16) for the plural past tense indicative and subjunctive forms of ganga 'walk'. The subjunctive marker -i- is mostly invisible, but it is sometimes indirectly visible as a palatalization effect on a preceding -g- or -k-, indicated in spelling by the letter -j(boldface below):17 (16) (16a) (16b) (16c) (16d) (16e) (16f)

V geng geng geng geng geng geng

Τ

Μ i i i

Num u u u u u u

Pers m ö -

m ö -

= gengum (PAST.IND.PL.L) = gengud (PAST.IND.PL.2) = gengu (PAST.IND.PL.3)

= gengum (PAST.SUBJ.PL.L) = gengjuö (PAST.SUBJ.PL.2) = gengju (PAST.SUBJ.PL.3)

As argued by RlZZl (1997), the C-domain must presumably split as well, and, as in CINQUE (1999), JULIEN ( 2 0 0 2 ) and much related work, the Μ , Τ and Ν domains each also splits into many subdomains. However, the simple structure in (15) is sufficiently articulated for most of our purposes (but for some extentions, see section 3).

16

17

See also e.g. SlGURBSSON (forthcoming) and the works cited there. Arguably, Gender or Class is also a clausal element, visible in inflected predicates in the Scandinavian languages, as discussed in section 3 (whereas it seems to be much higher in the clausal structure in languages where it is visible on the finite verb, such as Arabic and Hindi/Urdu). I do not assume that Person should be split into a Speaker and a Hearer (as in e.g. POLETTO 2000). The Speaker and the Hearer are necessarily 'silent', in the Speech Phrase, SP, a LF-structure that dominates CP (i.e. the personal pronouns and pronominal clitics do not directly represent the speech situation participants, although they represent event participants that may happen to be coreferential with the speech situation participants). For some discussion, see SlGURBSSON (forthcoming) and section 3. Arguably, it is also the trigger of a widespread i-Umlaut in past subjunctive forms, even synchronically, but I shall not pursue this here. Strong verbs like ganga have, of course, no tense suffix, marking tense with Ablaut of the stem vowel. Presumably, the null tense 'marker' triggers or matches Ablaut when it immediately c-commands the verb stem (prior to Verb Raising) in much the same fashion as the mood marker triggers/matches Umlaut (whenever the stem has a vowel amenable to Umlaut).

68

HALLDÖR ÄRMANN SIGURDSSON

2. Germanic agreement phenomena Let us now turn to agreement phenomena in the Germanic languages and see how they bear on the theoretical issues raised in section 1. Icelandic is the richest agreement language among the Germanic languages (with the exception of complementizer agreement, which it lacks). I shall therefore largely base my description on that language, only complementing it now and then by adding comparative data from the other Germanic languages. I shall describe and discuss the following phenomena: I II III IV V VI

Finite verb agreement Primary predicate agreement Secondary predicate agreement DP-internal agreement or concord Noun-possessor agreement Complementizer agreement

My discussion of finite verb agreement (section 2.1) and of primary predicate agreement (section 2.2) is rather thorough, whereas I only give a sketchy overview of the other four phenomena (in section 2.3).18 The purpose of the following presentation is threefold. First, I wish to present a typological overview of Germanic agreement phenomena, since such an overview is lacking and since the range and complexity of the data are not generally appreciated. Second, by presenting these complex data, I underpin my claim that the Probe-Goal Approach to Agree is much too specific: it cannot account for agreement variation in even only Germanic. Third, some of the presented data form a basis for the concluding discussion in section 3. Readers who are not interested in the details of the data ought to be able to go to the concluding discussion in section 3 after only a quick glance at the highlights of this chapter, above all the discussion in 2.1.4-2.1.5. Conversely, readers who are primarily interested in the data might wish to only quickly browse through section 3.

2.1. Finite verb agreement 2.1.1. Subject-verb Agreement vs. Reverse Agreement The Icelandic finite verb shows agreement in person (1, 2, 3) and number (SG, PL), as illustrated by the present tense indicative paradigms in (17). Only some of the numerous inflectional paradigms are represented:

18

Possibly, the structural genitive in languages like English and Mainland Scandinavian (as opposed to the morphological genitive in Icelandic and German) can be analyzed as an agreement marker. In addition, stem-final thematic vowels and so-called binding vowels and binding consonants in compounds can be thought of as word formation agreement markers, that is, mere 'signs of compositionality' below the word-level.

69

Agree and agreement: evidence from Germanic

purfa β 'need' 'get' fee patf LSG parft foerö 2SG fcer 3SG Parf putfum fawn 1PL purfid föiö 2PL purfa 3PL ß Most paradigms have five distinct forms in the indicative, present and past, and in the past subjunctive, but only four in the present subjunctive. Strong verbs with a stem in a vowel, like/ά 'get', however, show six forms in the present indicative. So-called 'mediopassive' verbs in -st, like heyrast 'get heard', fast 'be available', etc., have fewer distinct forms, mostly four in the present indicative but down to even only two in the present subjunctive. Most other Germanic varieties have simplified the system to a varying extent. Consider the overview in (18) of the 'normal present indicative paradigm' in some of the languages in question (with hear and its cognates as an illustrating example). The information is taken from VLKNER (1997-1998: 90-91, 94) and THRAINSSON (2001: 18): byrja 'begin' byrja byrjar byrjar byrjum byrjid byrja

(18) Icelandic German Frisian Yiddish Dutch Faeroese English Swedish

heyra 'hear' heyri heyrir heyrir heyrum heyrid heyra

telja 'count' tel telur telur teljum teljiö telja

Singular forms (1-2-3): heyri - heyrir - heyrir höre - hörst - hört hear - hearst - heart her - herst - hert hoor - hoort - hoort hoyri - hoyrir - hoyrir hear - hear - hears hör - hör - hör

Plural forms (1-2-3): heyrum - heyrid - heyra hören - hört - hören hearre - hearre - hearre hern - hert - hern hören - hören - hören hoyra - hoyra - hoyra hear - hear - hear hör - hör - hör

# 5 4 4 4 3 3 2 1

This illustrates only a part of the variation, though. Other varieties of Germanic include Danish, Norwegian and Afrikaans, which are like Swedish in having only one present tense form, West Flemish and the Norwegian dialect Hallingmilet, both with two forms (albeit with very different functions), Pennsylvania German with four forms and the Swedish dialect or language Älvdalsmälet, also with four forms, that is, a common singular form and three distinct plural forms (see VlKNER 1995: 152; 1997-1998; V A N N E S S 1994; DONALDSON 1994; ROHRBACKER 1994: 105ff.). High Alemannic has up to five forms and Austrian-Bavarian varieties may even have six forms (WERNER ABRAHAM, p.c.). Interestingly, Icelandic (and German) finite verb agreement is not just simple subject-verb agreement.19 To be sure, plain subject-verb agreement is the 'usual' and the 'normal' finite verb agreement type, illustrated in (19):

19

Much of the following discussion is to a varying extent based on earlier work, in particular SIGURDSSON 1989, 1990-1991, 1993, 1996, 2000, 2002. I shall not try the reader's patience by pointing this out each time.

70 (19)

HALLDÖR ÄRMANN SIGURBSSON

LSG

2SG 3SG 1PL 2PL 3PL

Eg si bökina. I see book.the Ί see the book.' Ρύ serd bökina. Ηύη ser bökina. Viö sjäum bökina. Piö sjäid bökina. Peir sjä bökina.

hef seö Eg I have seen Ί have seen the book.' Ρύ hefur seö hefur Hun seö höfum seö Viö hafid Piö seö Peir hafa seö

bökina. book.the bökina. bökina. (Hün=F) bökina. bökina. bökina. (Peir=M)21

As these examples indicate, subjects most commonly precede the finite verb (in non-inverted orders) and trigger its obligatory agreement in both number and person. However, Icelandic (and German) also has several types of reverse agreement, that is, the finite verb sometimes agrees with a nominative that is to its right, internal to the predicate.21 First, Icelandic has the well-known 'late subject' type illustrated in (1) above and in (20). As seen, the type allows but does not require the presence of the expletive pad 'there, it' (in clause-initial position only): (20a) Pad hafa komid hingaö prir mälvisindamenn. it have.3PL come here three linguists.N 'Three linguists have arrived here.' (20b) Pess vegna komu hingaö stundum prir mälvisindamenn. that for came.3PL here sometimes three linguists.N 'Therefore, three linguists sometimes came here.' Second, the finite verb also shows number agreement with nominative objects, as in (21): (21a) Henni hafa alltaf leidst brceöur sinir. her.D have.3PL always bored brothers.N her.N.REFLEXlVE 'She has always found her (own) brothers boring.' (21b) Henni likudu sennilega ekki athugasemdirnar. her.D liked.3PL probably not comments.the.N 'She probably didn't like the comments.' Third, the verb may also agree 'downwards' in number with third person complement subjects (of infinitives or small clauses), as in (22): (22a) Henni fundust brceöurnir gäfaöir. her.D found.3PL brothers.the.N intelligent 'She considered the brothers to be intelligent.'

20 21

Icelandic has gender distinctions in plural, third person pronominals, as well as in singular ones. Reverse agreement is often 'defective', not allowing the full range of agreement (see Universal 33 in GREENBERG 1 9 6 6 and, f o r instance, VAN GELDEREN 1 9 9 7 , BENMAMOUN 2 0 0 0 , PLATZACK 2 0 0 2 ) .

A related phenomenon that I shall not discuss here is defective or incomplete agreement with coordinated nominatives. FRIDJÖNSSON (1990-1991) contains a useful discussion of many of the agreement complications that arise in Icelandic coordinated structures.

71

Agree and agreement: evidence from Germanic

(22b) Henni mundu virdast brcedurnir hafa verid her.D would.3PL seem brothers.the.N have been "The brothers would seem to her to have been intelligent.'

gäfadir. intelligent

This typologically rather unusual construction is referred to as Dative and Nominative with Infinitive, D/NcI, in SlGURBSSON 1989 and I shall adopt that term here.22 Fourth, the verb agrees with NP predicates in clauses with either demonstrative petta 'this' or demonstrative pad 'it' as a subject: (23a) Petta hafa liklega bara verid peir. this have.3PL probably only been they.N.M.PL 'It has probably only been them.' (23b) Eru pad pd kannski bara peer? are.3PL it then perhaps only they.N.F.PL 'Is it then perhaps only them?' A similar phenomenon is found in German, albeit less prominently (Wir sind es (nur) '(only) we are it' is usually preferred over Es sind (nur) wir 'it is (only) us'). 2.1.2. Reverse Agreement: a closer look For convenience, let us label the above mentioned reverse agreement types as follows: I Π ΙΠ IV

Late Subject Agreement ['therefore have come here linguists'] Nominative Object Agreement ['her have liked these comments'] D/NcI Agreement ['her have seemed they be intelligent'] Reverse Predicate Agreement ['it are only theyΊ

Late subjects are subject to the definiteness effect:23 (24a) *Pess that (24b) *Pess that

vegna for vegna for

hafa have.3PL hafa have.3PL

stundum sometimes stundum sometimes

komid come komid come

hingad here hingad here

mennirnir. men.the.N peir. they.N

It follows that the nominative in the Late Subject Construction can never be a personal pronoun of any sort. Nominative Object Agreement and D/NcI Agreement are subject to a slightly less severe restriction, in that they are confined to third person. Thus, in the Nominative Object Construction, we get the following grammaticality patterns: 22

23

In line with the traditional terms Accusative with Infinitive and Nominative with Infinitive, Acl and Ncl, respectively. All three terms, Acl, Ncl, and D/NcI, are slightly misleading, for two reasons: First, the complement is often a small clause rather than an infinitive. Second, the argument that is normally structurally case-marked, as either accusative (in Acl) or nominative (in Ncl and D/NcI), may bear inherent 'quirky' case in Icelandic (depending on the predicate of the infinitive or the small clause). Default third person singular 'agreement' does not rescue these examples. On the contrary, it only makes them still worse.

72

(25a) Honum mundu alltaf lika him.D would.3PL always like 'He would always like them.' (25b) *Honum mundud alltaf lika him.D would.2PL always like [i.e. 'He would always like you'] (25c) *Honum mundum alltaf lika him.D would. 1 PL always like [i.e. 'He would always like us'.]

HALLDÖR Ä R M A N N SIGURBSSON

peir. they.N

OK

pid.

*2P.AGR-2P.NOM

3P.AGR-3P.NOM

you.N.PL

viö. we.N

*1P.AGR-1P.NOM

The fact that (25a) is grammatical, in contrast to (24b), suggests that raising of the dative in (25) serves the same 'purpose' as potential raising of the nominative would do in (24).24 That, in turn, highlights the urgency of the question of why (25b, c) are ungrammatical. A similar pattern is found for D/NcI: (26a) Henni mundu pä virdast peir vera her.D would.3PL then seem they.N be 'It would then seem to her that they are here.' (26b) *Henni mundud pä virdast pid vera her.D would.2PL then seem you.N.PL be [i.e. 'It would then seem to her that you are here.'] (26c) *Henni mundum pä virdast viö vera her.D would.2PL then seem we.N be [i.e. 'It would then seem to her that we are here.']

herna. here hirna. here hirna. here

The constructions are alike in that they both have a dative 'quirky' subject and a nominative argument that is a potential agreement controller: D A T - V ( G R ) - N O M . However, there are also sharp differences between the constructions. First, the Nominative Object Construction is monoclausal while D/NcI is complex or 'biclausal' in the sense that it contains not only a simple main clause but also a subordinate predication (an infinitive or a small clause). Thus, the nominative has a different status in these constructions, as sketched below: 25

A

( 2 7 a ) DAT— [V—NOM] ( 2 7 b ) D A T - [ V - [ N O M + PREDICATE]]

Second, first and second person nominatives are fully grammatical in D/NcI, provided that they do not control agreement. That is, the finite verb is allowed to show up in the default third person singular (= non-person), but not in the agreeing person. This is illustrated (in part only) in (28), where the infinitival complement is within brackets: (28a) Mir mundi/*mundud pä virdast [pid me.D would.3SG/2PL then seem you.N.PL 'It would then seem to me that you are here.' 24 25

vera be

herna]. here

Namely, the 'purpose' of person matching, see below and SIGURBSSON (2003a), (forthcoming). That is to say, when the complement subject is not 'quirky'. 'V' in the formula stands for the finite verb (the predicate may of course be more complex).

73

Agree and agreement: evidence from Germanic

(28b) Henni mundi/*mundum pä viröast her.D would.3SG/lPL then seem 'It would then seem to her that we are here.' (28c) Okkur mundi/*mundir pä viröast us.D would.3SG/2SG then seem 'It would then seem to us that you are here.'

[vid we.N.PL

vera be

hirna], here

[pu you.N.SG

vera be

hirna], here

In contrast, first and second person nominatives are degraded in the Nominative Object Construction, even when they do not trigger agreement. This is illustrated in (29), where '?*' indicates a slightly less sharp ungrammaticality than the plain star: (29a) Henni

?*mundi/*mundud

alltaf

her.D would.3SG/2PL always [i.e. 'She would always like you.']

(29b) Henni

?*mundi/*mundum

alltaf

her.D would.3SG/LPL always [i.e. 'She would always like us.']

(29c) Henni

??mundi/*mundir

alltaf

her.D would.3SG/2SG always [i.e. 'She would always like us.']

lika like

pid.

lika like

vid. we.N

lika like

Ρύ.

you.N.PL

you.N.SG

For first or second person nominatives, the agreement differences between the constructions may thus be sketched as below; Ί / 2 ' simply stands for 'either first or second person': (30) NOMINATIVE OBJECT: ['me like(s) NOM'] *DAT-V1/2-1/2.N0M ?*DAT-V 3S G-1/2P.NOM

D/Ncl: ['me seem(s) NOM + Predicate'] *DAT-VI/2-1/2.NOM + PRED (= full agreement) OK DAT-v -l/2.NOM + PRED (= 3SG'agreement') 3SG

For third person plural nominatives the general pattern is as follows: (31) NOMINATIVE OBJECT: ['me like(s) NOM'] OK

DAT-V3pL-3PL.NOM

?*DAT-V 3SG -3PL.NOM

D/Ncl: ['me seem(s) NOM + Predicate'] OK DAT-V3pL -3PL.NOM + PRED (= fall agreement) OK DAT-v3SG -3PL.NOM + PRED ( = 3SG 'agreement')

The generalizations for the constructions can be summarized as follows: (32a) THE NOMINATIVE OBJECT CONSTRUCTION: (32a 1) Third person nominatives are licensed, normally requiring number agreement. (32a2) First and second person nominatives are prohibited (and fall agreement is even worse than default 3SG 'agreement'). (32b) THE D/Ncl CONSTRUCTION: (32b 1) Nominatives in all persons are licensed. (32b2) Third person nominatives allow but do not require number agreement. (32b3) First and second person nominatives sharply disallow fall agreement (i.e. they are grammatical as such, but only with a default 3SG form of the finite matrix verb).

74

HALLDÖR ARMANN SIGURBSSON

It would thus seem that the person feature of the nominative must be matched in the simple Nominative Object Construction (and the ungrammaticality of the construction for first and second person nominatives stems from the fact that obligatory person matching is blocked), whereas the nominative in D/NcI apparently survives without person matching.26 Common to both constructions and to finite clauses in general is that only nominatives may control agreement (see further below). In addition, however, the relation of the nominative and the finite verb plays a role. While a nominative subject may be in any person and triggers obligatory full agreement, nominatives that are either direct objects or subjects of a non-finite complement (an infinitive or a small clause) are subject to complex restrictions, both with regard to their person selection and with regard to agreement triggered by the selected person. However, the distinction between licensing and disallowing the full range of persons and person agreement is not simply that between (matrix) subjects and nonsubjects. Thus, the Reverse Predicate Agreement Construction shows no restrictions on person selection or person agreement: (33a) Petta erum bara vid. this are. 1 PL only we.N "This/It is only us.' (33b) Erud pad piτ bara pid? are.2PL it then 'Is it then only you?'

(33c) Ert are.2SG

only

petta

pu?

this/it

you.N.SG

you.N.PL

'Is this/it you?' It follows that some other feature or property than just 'structure' in the straightforward configurational sense of most syntactic theories decides whether or not nominative non-subjects may control full agreement. Compare the Nominative Object Construction and the Reverse Predicate Agreement Construction; the 'Subject-Verb Inversion', typical of the Germanic Verb-Second languages, underscores that henni 'her.D' in (34) and petta 'this/it' in (35) are the syntactic subjects in these constructions. The grammaticality judgements are robust: (34a) *Ρά hafid henni sennilega aldrei then have.2PL her.D probably never [i.e., 'Then she has probably never liked you.'] (34b) Ρά hafa henni sennilega aldrei then have.3PL her.D probably never 'Then she has probably never liked them.'

likad liked

pid, you.N.PL

likad liked

peir. they.N

(35a) Ρά hafid petta sennilega bara vend then have.2PL this probably only been 'Then this/it has probably only been you.' 26

pid. you.N.PL

Possibly, however, the infinitival or small clause complement in D/NcI has a silent Pers element (Infi in SIGURBSSON 1 9 8 9 , 1 9 9 6 ) with which the nominative null-agrees.

75

Agree and agreement: evidence from Germanic

(35b) Ρά hafa petta sennilega bar a verid then have.3PL this probably only been 'Then this/it has probably only been them.'

peir. they.Ν

The finite verb has access to the person and number features of the nominative NP or DP in (34b) and (35), thus being able to agree with it. In (34a), on the other hand, the finite verb cannot 'reach' the second person object, that is, something blocks the correlation, intervenes between the finite verb and the person of the object. Arguably, it is the dative that is the interfering factor (BOECKX 2000; SIGURBSSON 1996, 2000). This Dative-Intervention is not found in DAT-NOM constructions e.g. in German or Russian (on the latter, see SIGURBSSON 2002: 719f.; on the former, SLGURDSSON forthcoming). Compare the Icelandic (25b, c) above, = (36), and the German (37): (36a) *Honum mundud alltaf lika him.D would.2PL always like [i.e. 'He would always like you.'] (36b) *Honum mundurn alltaf lika him.D would. 1 PL always like [i.e. 'He would always like us.']

pid. you.N.PL vid. we.N

(37a) Ihm würdet ihr immer gefallen. him.D would.2PL you.N.PL always like 'He would always like you.' (37b) Ihm würden yvir immer gefallen. him.D would. 1 PL we.N always like 'He would always like us.' This is a most striking dichotomy. I shall return to it in section 2.1.4. 2.1.3. The Nominative Restriction As mentioned above, only nominatives may ever control finite verb agreement in Icelandic and German. This Nominative Restriction is a very familiar phenomenon, cross-linguistically. For Icelandic, however, this is in a way a remarkable restriction, since Icelandic is renowned for having numerous constructions with nonnominative 'true' (syntactic) subjects. We have already seen some examples of this, in the Dat-Nom constructions discussed above. A few more examples, where the oblique subjects are underlined: (38a) Hana vantadi peninga. her.A lacked money.A ' She lacked money.' (38b) Henni var oglatt. her.D was un-joyous 'She was nauseated.' (38c) Henni var engin vorkunn. her.D was no pity.N 'There was no reason to pity her.'

76

HALLDÖR ÄRMANN SIGURDSSON

(38d) Hennar gcetti ekkert ά fundinum. her.G noticed not-at-all at meeting.the 'She was not at all noticeable at the meeting.' Icelandic non-nominative arguments of this sort behave syntactically like usual nominative subjects in the language (in contrast to similar arguments in e.g. German). That is, they trigger reflexivization in a similar fashion as nominative subjects, can be represented as 'silent' subjects of infinitives (PRO) and as the subject gap in Conjunction Reduction structures, participate in 'Subject-Verb Inversion', undergo raising and so on. In view of this, it is interesting that these subjects do not trigger agreement (see the discussion in e.g. SlGURBSSON 1996,2003a, forthcoming). Clauses with a non-nominative subject most commonly have a default third person singular finite verb, as in (39): (39a) Okkur

leiddist

ad

hugsa

um

us.D bored.3SG to think about 'We found it boring to think about the boys.'

(39b) Okkur likadi vel ad hugsa um us.D liked.3SG well to think about 'We liked attending to the horses.' (39c) Okkur var sagt frä atburöunum. us.D was.3SG told from events.the.D 'We were told about the events.'

sträkana. boys.the.A

hestana. horses.the.A

As we have seen, however, verbs with dative subjects agree in number with a third person nominative object. Thus, if we 'exchange' the infinitival and prepositional complements in (39) with direct nominative objects, as in (40), we get plural agreement on the verb: 7 (40a) Okkur leiddust sträkarnir. us.D bored.3PL boys.the.N.PL 'We found the boys boring.' (40b) Okkur likuöu hestarnir vel. us.D liked.3PL horses.the.N.PL well 'We liked the horses.' (40c) Okkur voru sagdir atburöirnir. us.D were.3PL told events.the.N.PL 'We were told the events.' In short, non-nominative subjects never control finite verb agreement, but clauses with such subjects may show finite verb number agreement with a nominative third person object. Notice that passives behave like active predicates in this respect, as seen in (40c). I shall return to the Nominative Restriction, arguing that it constitutes an argument against the Probe-Goal Approach to Agree. Third person singular is also the form of the finite verb in various types of impersonal clauses, with the expletive pad 'there, it' in first position, as in (41):

27

Mostly obligatorily, but sometimes only optionally. See

SIGURDSSON

1996.

Agree and agreement: evidence from Germanic

77

(41a) Pad rignir stundum mikid ί Rom. it rains.3SG sometimes much in Rome 'It sometimes rains much in Rome.' (41b) Pad er oft leidinlegt herna ά kvöldin. it is.3SG often boring here in evenings.the 'It is often boring here in the evenings.' (41c) Pad er yflrleitt ekki talad um petta mal her. it is.3SG generally not talked about this matter here 'This matter is generally not discussed here.' (4Id) Pad me ekki reylg'a hdr. it may.3SG not smoke here 'One is not allowed to smoke here.' Expletive pad is not a syntactic subject, but a placeholder of the first position in finite clauses (THRÄINSSON 1979 and many others, e.g. PLATZACK 1987, SlGURBSSON 1 9 8 9 , MAGNÜSSON 1 9 9 0 , HOLMBERG & PLATZACK 1 9 9 5 , VlKNER 1995, J0NSSON 1996). If some other constituent takes the first position, then pad must

not be spelled out, that is, the type 'sometimes rains it in Rome' is ungrammatical; instead, one has to say 'sometimes rains in Rome', without the expletive. Icelandic has various impersonal, subjectless and pad-loss clause types of this sort, some of which are illustrated in (42). As seen, the finite verb shows up in the default third person singular: (42a) Stundum rignir mikid ί Rom. sometimes rains.3SG much in Rome 'Sometimes, it rains much in Rome.' (42b) Oft er leidinlegt herna ά kvöldin. often is.3SG boring here in evenings.the 'It is often boring here in the evenings.' (42c) Um petta τηάΐ er yflrleitt ekki talad hir. about this matter is.3SG generally not talked here 'This matter is generally not discussed here.' (42d) Her ma ekki reylg'a. here may.3SG not smoke 'One is not allowed to smoke here.' The expletive must also be left out in the 'corresponding' questions, as in (43), and, again, the finite verb is in the third person singular: (43a) Rignir stundum mikid ί Rom? rains.3SG sometimes much in Rome 'Does it sometimes rain much in Rome?' (43b) Er oft leidinlegt herna ά kvöldin? is.3SG often boring here in evenings.the 'Is it often boring here in the evenings?' (43c) Er yflrleitt ekki talad um petta mal her? is.3SG generally not talked about this matter here 'Is this matter generally not discussed here?'

78

HALLDÖR ARMANN SIGURDSSON

(43d) Μά ekki reykja her? may.3SG not smoke here 'Is one not allowed to smoke here?' Similar facts, albeit much less frequently, are found in German: (44a) Über diese Sache about this matter (44b) Darüber muss there-about must.3SG (45)

Wird is.3SG

hier here

wird hier meistens nicht gesprochen. is.3SG here generally not talked doch gesprochen werden. though talked be

darüber there-about

nicht not

gesprochen? talked

Clauses of this sort are often analyzed as having a 'silent' subject, so-called expletive pro, as sketched in (46): (46a) [about this matter is pro generally not talked here] (46b) [is pro not talked about it here?] In order to account for the 3SG form of the verb, one could say that pro, the 'silent subject', is inherently third person singular - and that the verb simply agrees with the 3SG features of pro. Alternatively, one could say that there is no real subject in structures of this sort and that the 3SG form is default, signalling the absence of specified person and number values. 2.1.4. Dative Intervention: Icelandic vs. German Both the alternatives just mentioned (3SG agreement vs. default non-agreement) could be extended to German 'impersonal' constructions, with a 'prominent' nonnominative argument, as in (47): (47a) Mir me.D (47b) Mir me.D

ist is.3SG wurde was.3SG

kalt. cold geholfen. helped

Spelling out the expletive es in structures of this sort is grammatical (Es ist mir kalt, Es wurde mir geholfen, etc., as discussed in SIGURDSSON 1989 and HAEBERLI 2002), suggesting that the types in (47) might involve a silent subject (alternatively spelled out as es). This analysis is sketched in (48): (48a) DAT is pro cold (48b) DAT was pro helped

The status of pro (and PRO) is unclear and quite problematic in the minimalist program, but let us assume this analysis of the German facts, for the sake of the argument. The important point for our purposes is that this analysis could however not be extended to Icelandic. In that language, the dative is arguably a subject, mostly behaving like a usual nominative subject (with respect to a host of phenomena that are commonly referred to as 'subjecthood tests', such as binding, word order phenomena,

79

Agree and agreement: evidence from Germanic

raising to subject and so on). Thus, the spelling out of the expletive pad in examples of

this sort is sharply ungrammatical: (49a) Mir me.D (49b) Mir me.D

er is.3SG war was.3SG

kalt. cold hjalpaö. helped

/ / / /

*Ραό it *Ραό it

er is var was

mer me mir me

kalt. cold hjälpad. helped

The 3SG form of the verb in Icelandic examples of this sort, then, should arguably not be analyzed as agreeing with a silent pro subject. On the other hand, there are also reasons to doubt that the 3SG form is simply a non-agreeing default form. That is, there is evidence that the third person of the verb in clauses with a non-nominative subject is actually due to 'null-agreement' with the oblique subject: a syntactically active but a morphologically non-active or invisible matching correlation. The evidence in question comes from the deficient agreement patterns in the Nominative Object Construction and in the D/NcI Construction, discussed in section 2.1.2. Reconsider (25) = (50): (50a) Honum

mundu

alltaf

lika

him.D would.3PL always like 'He would always like them.' (50b) *Honum mundud alltaf lika him.D would.2PL always like [i.e. 'He would always like you'] (50c) *Honum mundum alltaf lika him.D would. 1 PL always like [i.e. 'He would always like us'.]

f>eir.

ok

3P.AGR-3P.NOM

they.N feiö. you.N

*2P.AGR-2P.NOM

vid. we.N

'lP.AGR-lP.NOM

We have an account of the ungrammaticality of the person agreement with the nominative object in (50b, c) if the person feature of the finite verb complex of the clause is already 'engaged' in (invisibly) matching the person of the dative subject and is thus blocked from agreeing in person with the nominative object as well. At some stage of the derivation, both arguments are within the predicate (vP in the notation of C H O M S K Y 2000), hence c-commanded by both the Pers(on) and the Num(ber) feature of the finite verb complex: (51)

[cp

c

[PERSP

Pers [NumP Num . . .

[ DAT... NOM ...]]]]

The dative subject evidently raises into the vicinity of the Pers head. The nominative also raises, but to a lower position, an Object Shift type of raising (see e.g. C O L L I N S & T H R Ä I N S S O N 1 9 9 6 ; for ease of exposition however, I do not take Object Shift into account here).28 Assume that the dative raises to Edge,NumP, such that it is immediately 'c-commanded' by Pers:29

28 29

The position in question is presumably at the edge of EP in the system discussed in section 3. As already mentioned, X'-theoretic labels must be dispensed with (cf. also e.g. STARKE 2001, COLLINS 2002). Edges and nuclei are thus only separated by a comma.

80 (52)

HALLDÖR ARMANN SIGURBSSON

[Cp C [persP Pers [ N u m P DAT,Num ... [ΒΑϊ...

NOM ...]]]]

1 I I I As indicated by the 'matching paths', Num does not intervene between Pers and the dative subject, and hence the latter may match 'silently'. Similarly, having been raised, the dative does not intervene between Num and the nominative object and they may accordingly agree. This accounts for the grammaticality of (50a): Pers 'agrees silently' with the dative (third person being an unspecified person or 'no' person), and Num agrees (with only the number of) the nominative object. In contrast, Pers cannot agree with the nominative object since, first, it is already engaged in a matching relation with the dative, and, second, the dative intervenes between the two (a Minimal Link Condition violation, cf. CHOMSKY 2001a: 16). Hence, the ungrammatically of (50b, c).30 Now, recall that German radically differs in this respect from Icelandic, as do many other languages.31 Reconsider (37) = (53): (53a) Ihm würdet ihr him.D would.2PL you.N 'He would always like you.' (53b) Ihm würden wir him.D would. 1 PL we.N 'He would always like us.'

immer always

gefallen. like

immer always

gefallen. like

These facts suggest that the dative does not intervene between Pers and the nominative at the relevant level of representation. One way to accommodate this is to assume that the dative raises out of the agreement scope of Pers prior to agreement, to Edge,PersP. Abstracting away from the word order differences between German and Icelandic, this gives us the structure in (54): (54)

[CP C [persP DAT, Pers [NumP Num ... [BA¥ ... NOM...]]]]

On the assumption that the German Pers parametrically differs from the Icelandic one in being unable to match non-nominatives,32 raising of the dative to a position where it would enter into potential person matching is blocked. Such raising would serve no purpose (i.e. it would be ruled out by economy), and hence the matching position is not an available landing site for the dative. It follows that no dativeintervention arises, and both Pers and Num can access the nominative argument without violating the Minimal Link Condition.

30

31 32

On dative intervention, see BOECKX 2 0 0 0 , SIGURBSSON 2 0 0 0 and, for a different set of data, HOLMBERG & HR0ARSD0TTIR 2 0 0 2 . - In view of the fact that nominative (first and second person) subjects match both Num and Pers, the question arises why non-nominative subjects do not 'nullagree' with Num (resulting in a 'singular' form) as well as with Pers (yielding a 'third person' form). This is accounted for if the relevant movement and feature matching processes apply prior to morphological agreement. See SIGURBSSON 2002 on Russian, SUBBARAO 2001 on Telugu and e.g. DAVISON (forthcoming) on Hindi. Alternatively, the difference might boil down to morphological differences between Icelandic and German DPs, as suggested in SIGURBSSON 1 9 9 4 .

81

Agree and agreement: evidence from Germanic

By abstracting away from the VO/OV dichotomy, I am not claiming that the underlying order of elements is uniform across languages. There may very well be a universal order of elements (KAYNE 1994; CINQUE 1999), but, as far as I can see, this would then be so by some evolutionary 'accident' rather than by conceptual necessity.33 On the other hand, we need to assume that basic functional elements, 'visible' or not, are the same in all languages, or else there could be no Universal Grammar. Rather than postulating a universal order of elements, I am assuming that the order differences between German and Icelandic, underlying or superficial, are irrelevant with respect to the phenomena under discussion. This assumption might turn out to be wrong, but, it remains to be disproved and is thus the null-hypothesis at present.34 Now, recall the Nominative Restriction. In spite of having non-nominative subjects entering into a matching relation with Pers, Icelandic is like German and many other languages in only permitting overt morphological agreement of the finite verb with nominative arguments. It is important that we do not just take this restriction for granted but try to understand it and account for it: It is by no means the case that language categorically forbids agreement controlled by non-nominatives. Thus, we need to develop an account of the Nominative Restriction on finite verb agreement in Icelandic and many other languages. As I have argued elsewhere (in e.g. SlGURBSSON 2003a), finite verb-oblique agreement seems to be blocked by economy: Inherent morphological case is an agreement morphology in itself, such that e.g. the dative of the DP of a dative-taking item X is in morphological agreement with the selectional requirements of X. If the DP were to agree also with the finite verb, it would be simultaneously involved in two morphological agreement relations. Overt 'polygamy' of this sort is not forbidden in language, but it is commonly avoided, arguably for reasons of economy. On this account, morphological agreement is not merely a reflection of a single underlying syntactic relation, Agree: it is crucially affected by other factors as well. This is an important observation. It illustrates that morphological agreement cannot be interpreted as bearing on Agree in any straightforward or simple manner. Both phenomena, Agree and agreement, have to be carefully studied in their own right. 2.1.5. Nucleus-Edge Matching and Stylistic

Fronting

Notice that the edge of a category does not check features of its own 'head' or nucleus in this approach. Rather, a nucleus is in a potential checking or matching

33 34

Given the theoretical premises assumed here, at least. For a very different approach to the difference between Icelandic and German with respect to nonnominative subjects, ultimately relating it to radically different structures of the languages and the V O / O V dichotomy, see HAIDER, e.g. 1997, 2001. As discussed in SIGURBSSON (forthcoming), however, Tamil, an SOV language, seems to be like Icelandic in having 'true' non-nominative subjects. Also, postulating radically different functional structures in different languages, as HAIDER does, is not an option in a minimalist approach, like the present one (as it amounts to abandoning the Uniformity Principle of CHOMSKY 2001a: 2).

82

HALLDÖR ÄRMANN SIGURDSSON

relation with the edge of its sister, Nucleus-Edge Matching.35 Thus, as illustrated in (52) above, Icelandic oblique subjects match Pers when situated in Edge,NumP, not Edge,PersP.36 This approach therefore predicts that Edge,PersP is available as a checking or a matching site for some feature of the C-domain: (55)

[CP ... C ... [persp X(P), ΡβΓ5 [NumP ...

This is borne out by Icelandic Stylistic Fronting, a much discussed process that fronts various elements into the position immediately c-commanded by C, traditionally referred to as Spec,IP.37 Three examples are given in (56). For ease of presentation, copies are shown as traces: (56a) Petta er mal [sem mikidj hefur verid talad U um], this is matter that much has.3SG been talked about 'This is a matter that has been discussed a lot.' (56b) Eg vissi ekki [ad taladi hefdi veriö /,· um petta mäl\. I knew not that talked had been about this matter Ί didn't know that this matter had been discussed.' (56c) Talad hefur veriö tt urn ad fara. talked has been about to go 'People have talked about going.' The C feature that is checked or matched by Stylistic Fronting to Edge,PersP is an EPP feature in the approach of HOLMBERG 2000, but the Fin (finiteness) feature in SIGURDSSON 2003b. The feature in question is most commonly matched by a (nominative or an oblique) subject, leaving a copy in Edge,NumP, as illustrated in (57): (57)

[CP ... Fin [PereP SUB*, Pers [NumP t;, Num . . . [ . . . tj ...]]]]

It follows that Stylistic Fronting of a non-subject can never take place in the presence of a regular (phonologically spelled-out, non-delayed) subject, a restriction known since MALING 1980 as the 'Subject Gap Condition'. This is illustrated in (58). For clarity, the subject is underlined, whereas the Stylistically Fronted participle is boldface: (58a) Petta er mäl [sem harm hefur talad this is matter that he has.3SG talked "This is a matter he has talked much about.'

35

36 37

38

mikid much

um], about

That is, the ECM or Acl type of correlation is not exceptional, after all. - The underlying assumption here is that for instance an object of a verb is not its direct complement, but an edge element of a silent functional 'head' or nucleus in a small clause complement. See BRODY 2000 for an elaboration of this insight (albeit with different implications for matching theory). I return to this issue in section 3. Nominative subjects are likewise in or at Edge,NumP when they match Pers. See for instance MALING 1980, RÖGNVALDSSON & THRÄINSSON 1990, JÖNSSON 1991, SIGURBSSON 1997, HOLMBERG 2000, Βοδκονιό 2001. The Fin feature is silent by necessity in main clauses, but it is matched under Nucleus-Edge Matching by a lexical element in Edge,PersP (SIGURDSSON 2003b).

83

Agree and agreement: evidence from Germanic

(58b) *Petta er mal [sent harm mikidi hefur has.3SG much this is matter that he hefur (58c) * Petta er mal [sem mikidi hann has.3SG this is matter that much he verid (58d) Petta er mil [sem mikidi hefur this is matter that much has.3SG been 'This is a matter that has been discussed a lot.'

talad ti talked talad ti talked talad ti talked

um]. about um]. about um], = (56a) about

In the absence of a regular subject, Stylistic Fronting may apply to the closest possible candidate (in a sense explicated in HOLMBERG 2000). In the presence of a regular subject, on the other hand, the subject itself must be stylistically fronted, from Edge,NumP to Edge,PersP. That is, Stylistic Fronting to Edge,PersP of a non-subject across an overt subject in Edge,NumP would violate the Minimal Link Condition (or Relativized Minimality, RlZZI 1990). On this account, Edge,PersP is not an Ά position' (see the discussion in RÖGNVALDSSON & THRAINSSON 1990).39 Nucleus-Edge Matching is very similar to CHOMSKY'S probe-goal mechanism. However, it is not tantamount to Merge/Agree itself. Rather, it is its specific implementation as well as its prerequisite. In order for X to be able to Merge/Agree with Y, an edge feature of X has to match a nuclear feature of Y. For instance: a NumP that merges with Pers has an edge element (e.g. a dative subject) containing a feature that matches a nuclear feature of Pers, a PersP that merges with Fin has an edge element, e.g. a stylistically fronted category, that contains a feature that matches a nuclear feature of Fin, and so on. - Notice that X and Y with features x}, x2, X3 ... a n d y i , y2,}>3 ... may possibly Merge/Agree in more than one way, such that different features are involved in each instance of Nucleus-Edge Matching. Variation of this sort is evidently severely limited, but the logical possibility arises, underscoring the fact that we need to distinguish between feature matching and Agree itself. I shall return to Nucleus-Edge Matching in section 3, where I suggest that it offers a new understanding of Move as driven by the needs of Merge/Agree. 2.1.6. Non-intervening subjects Finally, recall that demonstrative petta 'this' and demonstrative pad 'it' differ from non-nominative subjects in that they do not block person agreement with a late or a low nominative, in the Reverse Predicate Agreement Construction. Reconsider (33) and (35a), repeated here as (59) and (60): (59a) Petta erum bara this are. 1 PL only 'This/It is only us.' (59b) Erud pad bara are.2PL it only 'Is it then only you?'

39

vid. we.N pid? you.N.PL

Also, of course, a distinction between 'maximal' and 'non-maximal' objects of movement makes no sense in the present approach.

84

HALLDÖR ÄRMANN SIGURDSSON

(59c) Ert

petta

are.2SG this/it ' I s this/it you?'

pü? you.N.SG

(60) Ρά hafid petta sennilega bara verid then have.2PL this probably only been 'Then this/it has probably only been you.'

pid. you.N.PL

Prior to NP-movement, the relevant structure of e.g. (60) is as shown in (61): (61)

[Cp C [persp Pers [NumP Num ... [ &ETTA ... NOM]]]]

Petta (as well as pad in 59b) is a default form ('neuter singular nominative/accusative'), that is, it is void of all featural content, except for its phonological features, its D-feature and its demonstrative force (see the discussion in SIGURDSSON 1996).40 In particular, it has no values for 'true' person, number and gender, thus cannot intervene between Pers/Num and the predicative nominative. Arguably, it moves directly to Edge,PersP, by Stylistic Fronting, matching the Fin feature, like other instances of Stylistic Fronting. 1 Thus, paö and petta are like oblique 'prominent' arguments in German in not 'counting' as interveners (being unable to match Pers), and the question of intervention or not is not simply a question of parametric distinctions between distinct languages, but a more fine-grained issue. Again, the data illustrate that there is more to morphological agreement than just a simple probe-goal relation. This concludes our discussion of finite verb agreement in Germanic: Simple subject-verb agreement, reverse agreement phenomena (four major types), the nominative restriction, dative intervention and Nucleus-Edge Matching. Even if we were to restrict our attention to only finite verb agreement, the data are so complex and varied that a simple Probe-Goal Approach to Agree could not account for but a fraction of the facts. Moreover, there are additional, quite complex and varied agreement facts within Germanic that any general theory of Agree and agreement must account for. We shall take a look at (many of) these data in the next two subsections. 2.2. Primary predicate agreement 2.2.1. The facts in Icelandic In Icelandic, both predicative adjectives and passive past participles agree with their nominative subject, as illustrated in (62)-(63): Predicative adjective agreement in finite clauses: (62a) Hann he.N.M.SG 40

var was

liklega probably

rikur. rich.N.M.SG

Plausibly, the demonstrative force is given or default in the absence of any features specifying the D-feature. 41 ~ For a discussion of the surprising and problematic Dutch pattern Het waren hurt 'It were.PL them.A', see VAN GELDEREN (1997: 154). I have nothing to add to her discussion.

85

Agree and agreement: evidence from Germanic

(62b) Hun she.N.F.SG (62c) Pad it.N.N.SG (62d) Peir they.N.M.PL (62e) Peer they.N.F.PL (62f) Pau they.N.N.PL

var was var was vom were vom were vom were

rik. rich.N.F.SG rikt. rich.N.N.SG rikir. rich.N.M.PL rikar. rich.N.F.PL rik. rich.N.N.PL

liklega probably liklega probably liklega probably liklega probably liklega probably

Past participle agreement in finite clauses: (63a) Hann he.N.M.SG (63b) Hun she.N.F.SG (63c) Pad it.N.N.SG (63d) Peir they.N.M.PL (63e) Peer they.N.F.PL (63 f) Pau they.N.N.PL

var was var was var was vom were vom were vom were

liklega probably liklega probably liklega probably liklega probably liklega probably liklega probably

kosinn. elected.N.M.SG kosin. elected.N.F.SG kosid. elected.N.N.SG kosnir. elected.N.M.PL kosnar. elected.N.F.PL kosin. elected.N.N.PL

As seen, the agreeing features are case, gender and number.42 The same pattern is seen in most infinitives and small clauses (as shown only partially in (64)-(65)). For a discussion of the case-theoretical implications of this and related facts, see SlGURDSSON 1991 and subsequent work (the infinitival clauses are marked offby parentheses): Predicate agreement in 'nominative infinitives': (64a) Peir vonudust til [ad verda they.N.M.PL hoped for to be(come) 'They hoped to get rich.' (64b) Peir vonudust til [ad verda they.N.M.PL hoped for to be 'They hoped to be elected (by someone).'

rikir]. rich.N.M.PL kosnir]. elected.N.M.PL

(65a) Mir höfdu virst [peir vera rikir]. me.D had seemed they.N.M.PL be rich.N.M.PL 'They had seemed to me to be rich.' (65b) Mer höfdu virst [peir vonast til [ad verda kosnir]]. me.D had seemed they.N.M.PL hope for to be elected.N.M.PL 'It had seemed to me that they hoped to be elected.' 42

Definiteness is an agreement feature in DP-internal concord in the Germanic languages (see section 2.3), but not in predicative agreement.

86

HALLDÖR ÄRMANN SIGURDSSON

In addition, the same patterns are seen in ECM or Acl constructions (both infinitives and small clauses), the only difference being that the agreeing case is accusative instead of nominative on both the 'subject' and the predicate: Predicate agreement in 'accusative infinitives': (66a) Eg mundi pvi telja [pa vera I would thus believe them.A.M.PL be 'I would thus believe them to be rich.' (66b) Eg mundi pvi telja [pa hafa vend I would thus believe them.A.M.PL have been Ί would thus believe them to have been elected.'

rika]. rich.A.M.PL kosna]. elected.A.M.PL

Moreover, multiple predicate agreement is found in both nominative and accusative infinitival (and small clause) structures: Multiple predicate agreement: (67a) Peir mundu vera taldir vera sagdir they.N.M.PL would be believed.N.M.PL be said.N.M.PL hafa vend kosttir. have been elected.N.M.PL 'They would be believed to be said to have been elected.' (67b) Eg mundi telja pa vera sagda hafa I would believe them.A.M.PL be said.A.M.PL have vend kosna. been elected.A.M.PL Ί would believe them to be said to have been elected.' Multiple agreement of this sort is problematic for the simple Probe-Goal Approach to Agree (cf. the discussion in CHOMSKY 2001a: 18, and in FRAMPTON & GUTMAN 2000).43 As discussed in section 2.1.3, only nominatives may ever control finite verb agreement. Much the same restriction applies to primary predicate agreement. More exactly, the following generalizations hold: I Π

Only nominatives may control primary predicate agreement in finite clauses. Only those accusatives that correspond to nominatives in finite clauses may control primary predicate agreement in ECM or Acl constructions.

This is rather neatly demonstrated by adjectival predicates such as 'cold, freezing', 'hot, warm', 'ill, bad', 'good, well', that may either take a nominative (theme) or a dative (experiencer) subject. When the subject is nominative (or accusative in ECM), agreement is obligatory, but impossible when the subject is dative. Consider this for 'cold, freezing': 43 44

They are also problematic for the approach in SlGURBSSON 1993, of course. While the dative construction is semantically narrow (referring only to animates sensing coldness), the nominative construction has many meanings, of which I only give the most central one.

87

Agree and agreement: evidence from Germanic

(68a) Peir vom they.N.M.PL were.3PL 'They were cold (to touch).' (68b) Peim var them.D.PL was.3SG 'They were cold/freezing.'

kaldir. cold.N.M.PL kalt. cold.N/A.N.SG

(69a) Eg taldi jxi vera I believed them.A.M.PL be Ί believed them to be cold (to touch).' (69b) Eg taldi peim vera I believed them.D.PL be

kalda. cold.A.M.PL kalt. cold.N/A.N.SG

Active participles selected by vera 'be' and verda 'become' show the same agreement properties as do passive participles, whereas active participles selected by hafa 'have' (and by the modal geta 'can') never agree, instead showing up in a default N/A.N.SG (supine) form:45 (70a) Sträkarnir boys.the (70b) Sträkarnir boys.the

vorn were.3PL höföu had.3PL

horfnir disappeared horfid disappeared

/ / / /

farnirl gone / farid / gone /

byrjadir = N.M.PL begun byrjad = N/A.N.SG begun

Active participles selected by 'have' agree with their raised clitic objects in Italian and French (e.g. E G E R L A N D 1996a: 164ff.). This is illustrated for French in (71a):46 (71a) Paul Paul (71b) Paul Paul

les them a has

a repeintes. has repainted.F.PL repeint/*repeintes les repainted the

chaises. chairs.F.PL

Active participle agreement of this sort was also sporadically found in Old Norse. In contrast to Italian and French active participle agreement, however, the Old Norse agreement was neither contingent on the position nor the content of the object, as illustrated in (72):47 (72a) brätt hefi ek ykkr brenda soon have I you.A.PL burned.A.M.PL ' Soon, I will have burned you.' (72b) svä haföi Helgi hroedda gerva fiändr so had Helgi afraid made.A.M.PL enemies 'Helgi had made all his enemies so afraid (that...).'

sina... his.A.M.PL

It is not clear, to say the least, how one would account for the fact that Icelandic has lost this type of agreement in terms of the Probe-Goal Approach to Agree. Rather, it seems that we have to analyse the historical change in terms of 'shallow' 45

Icelandic perfect types are discussed in e.g. FRIBJÖNSSON (1989), SIGURBSSON (1989: 322FF.), and JÖNSSON ( 1 9 9 2 ) .

46

From KAYNE (1989: 85).

47

F r o m NYGAARD (1906: 188). S e e also SIGURDSSON ( 1 9 9 3 : 4 8 ) .

88

HALLDÖR ÄRMANN SIGURBSSON

morpho-phonological agreement processes. Notice that these processes do not involve or reduce to a parametric choice of whether a participle agrees or not. The historical change applies not to past participles in general but specifically to those participles that are selected by hafa 'have'. Once again, then, the Probe-Goal Approach does not bear on the morphological agreement variation manifested within the Germanic languages. Contradictory as it may seem, we need both a more specific account and a more general one. That is, more specific or microscopic analyses of various morphological agreement phenomena are called for, and, simultaneously, we need a much more general notion of abstract Agree. 2.2.2. The Germanic variation (or at least much of it) In this section, I shall present (much of) the German variation with respect to primary predicate agreement (PRED-AGR), in view of the Icelandic facts just described. Let us start out by reviewing the central aspects of the Icelandic system: I II ΙΠ IV V

Lexical items: Icelandic PRED-AGR applies to both adjectives and past participles. Features: Icelandic PRED-AGR involves agreement in (structural) case, gender and number. Clause types: Icelandic PRED-AGR applies (to predicative adjectives and participles) in finite clauses as well as in both 'nominative infinitives' (and small clauses) and 'accusative infinitives' (and small clauses). Multiplicity: Icelandic PRED-AGR can 'multiply' or apply repeatedly in certain infinitival and small clause structures (the ECM type of structures and their passive counterparts). Case restrictions: Icelandic PRED-AGR is blocked by inherent case of the potential agreement controller, that is, only structurally case-marked DPs may ever control PRED-AGR (accusative DPs in ECM or Acl structures, otherwise nominative DPs). As we have seen, much the same restriction applies to finite verb agreement - it may only be controlled or triggered by a nominative.

Faeroese normally has the 'Icelandic type' of agreement (in case, gender and number) of both predicative adjectives and past participles, at least in finite clauses.48 The 'dramatic' distinction within Germanic with respect to predicate agreement is that between North- and West-Germanic. The West-Germanic languages plainly have no predicate agreement at all. Thus, even German, which has both finite verb agreement and DP-internal concord, has no predicate agreement. Consider the invariant predicative forms in (73) (reminiscent of the so-called short forms in Russian, cf. e.g. FRANKS 1995): 48

There are some interesting differences between the languages with respect to the distribution of inherent and structural cases (BARNES 1986; PETERSEN et al. 1998), but I shall not discuss these here, since they only affect the preconditions on, and hence the distribution of agreement, not the mechanism of agreement itself. In addition, however, there is some tendency in spoken Faeroese to develop non-agreeing 'short forms' of at least predicative adjectives (PETERSEN et al. 1998: 70).

89

Agree and agreement: evidence from Germanic

(73a) Ein a (73b) Eine a (73c) Ein a

guter good gute good gutes good

Mann man Frau woman Kind child

ist is ist is ist is

immer always immer always immer always

gut/*guter. good gut/*gute. good gut/*gutes. good

Past participles also have an invariable, non-agreeing form: (74a) Der the (74b) Die the (74c) Die the (74d) Die the

Mann man Frau woman Männer men Frauen women

wurde was wurde was wurden were wurden were

gewählt/*gewählter. elected gewählt/*gewählte. elected gewählt/*gewählte. elected gewählt/*gewählte. elected

Within the North-Germanic or the Scandinavian branch, the major parameters of variation are the following (a distinction has to be made between three major varieties of Norwegian: 1, 2 and 3): I II III

IV

V

The mainland Scandinavian languages differ from Icelandic and Faeroese in not having any agreement in case (as they only have pronominal case and structural genitive case). All the mainland Scandinavian languages are like Icelandic and Faeroese in having agreement of predicative adjectives. A variety of Norwegian, found mainly along the south and the west coast, is also like Icelandic and Faeroese in having general participle agreement with the logical subject of the predicate. Following HOLMBERG (2002: lOOff.), I refer to this variety as Norwegian 3. Swedish and another variety of Norwegian, including the nynorsk standard, show variation with respect to participle agreement, sometimes having agreement with an expletive subject, sometimes with a non-expletive subject. HOLMBERG (2002) refers to the Norwegian variety or varieties is question as Norwegian 2. Danish and most varieties of Norwegian, including the bokmäl standard, lack participle agreement altogether. These varieties are called Norwegian 1 by HOLMBERG (2002).

We can thus sketch the Germanic predicate agreement variation as follows: (75a) West-Germanic: (75b) Norwegian 1, Danish:

No PRED-AGR Adjectival agreement, no participle agreement (75c) Norwegian 2, Swedish: Adjectival and varied participle agreement (75d) Norwegian 3, Faeroese, Icelandic: Adjectival and non-varied participle agreement

90

HALLDÖR ÄRMANN SIGURDSSON

The distinction between adjectival and participial predicates in Danish and Norwegian 1 is illustrated by the Norwegian examples in (76), from CHRISTENSEN & TARALDSEN (1989: 81, endnote 1): (76a) Gjestene guests.the (76b) Gjestene

er are er

fulle/*full. drunk.PL/SG komet/*komne.

guests.the

are

come.SG/PL

'The guests have arrived.' In Norwegian 3, on the other hand, the participle must agree with the nonexpletive (logical) subject, as shown in (77), from CHRISTENSEN & TARALDSEN (1989: 58). These varieties of Norwegian have the locative der 'there' as an expletive.

(77a) Gjestene guests.the

(77b) Der

er

er

nett

komne/*kome.

are

just

arrived.PL/SG

nett

komne/*kome

there are just come.PL/SG 'Some guests have just arrived.'

nokre

giester.

some

guests

Finally, in Norwegian 2, the participle shows varied agreement, agreeing with either an expletive (pronominal det 'it'), or with the logical subject, in case it raises across the participle, as in (78a), also from CHRISTENSEN & TARALDSEN (1989: 58):49 (78a) Gjestene er nett komne/*kome. guests.the are just arrived.PL/SG (78b) Det er nett kome/*komne nokre it are just come.SG/PL some 'Some guests have just arrived.'

gjester. guests

The same applies to Swedish: It has neuter singular agreement with the expletive if the logical subject is to the right of the participle, but agreement with the logical subject in case it moves to the left of the participle:50 (79a) Tre böcker blev skrivna/*skrivet. three books were written.PL/N.SG (79b) Det blev tre böcker skrivna/*skrivet. it were three books written.PL/N.SG 'There were three books written.' 49

50

convincingly argue that the singular form of the participle actually agrees with the expletive neuter singular det in cases like (78b), whereas the expletive der in (77b) cannot control agreement, not having any pronominal pAi-features (number, gender, ...). See also HOLMBERG (2002: lOOff.). This is the general pattern. There are, however, some speakers that accept agreement with the logical subject also in examples like (79c) - but the pattern of these speakers differs from that found in Icelandic in that the agreement is not obligatory, but rather a bit degraded. At the other extreme, certain Northern Swedish dialects are like Danish and Norwegian 1 in not having any participle agreement (LARS-OLOF DELSING, p.c.). - Examples like (79b) have variable acceptance, i.e. some speakers find them degraded or unacceptable (CECILIA FALK, p.c.).

CHRISTENSEN & TARALDSEN

91

Agree and agreement: evidence from Germanic

(79c) Det blev skrivet/*skrivna tre it were written.N.SG/PL three 'There were three books written.'

böcker. books

HOLMBERG (2002: 106ff.) argues that the above described Scandinavian participle agreement variation can be accounted for if the phase notion is parametrized (the Participle Phrase being a phase in Swedish and Norwegian 2 but not in the other Scandinavian varieties). I propose instead that the variation in question is a 'shallow' morphophonological phenomenon, not reflecting any 'deep' (narrow) syntax differences between the closely related languages in question. More specifically, I suggest that it relates, at least partly, to the 'morphological strength' of gender. See further the discussion in section 3. A somewhat less central but no less interesting variation arises in Scandinavian passives of double object verbs. Consider the dichotomy in (80) between Swedish (80a) and Icelandic (80b), discussed by HOLMBERG (1994, 2002). As before, the agreement controller is underlined, whereas the relevant agreeing element is boldface:51 (80a) Det it 'There (80b) *Ραό it

blev givet pojken presenter. was given.N.SG boy.the presents were presents given to the boy.' voru gefnar einhverjum sträkum were.3PL given.N.F.PL

priär

bcekur.

three

books.N.F.PL

some

boys.D.M.PL

Icelandic expletive pad 'it, there' is like demonstrative pad 'it' and petta 'this', discussed in section 2.1.6, in that it is void of 'true' pronominal features and hence cannot control agreement, in contrast to Swedish det. The participle must thus resort to agreeing with the nominative object prjar bcekur 'three books' in both gender and number, but this is blocked by the intervening dative stmknum 'the boy'. If the dative is moved to the left of the participle, such that it does not intervene between the (gender of the) participle and the (gender of the) nominative object, the structure becomes grammatical (for a different interpretation of this fact, see HOLMBERG 2002: 99):

(81) Pad it

voru

einhverjum

sträkum

gefnar

were.3PL

some

boys.D.M.PL

given.N.F.PL

pridr

bcekur.

three

books.N.F.PL

'Some boys were given three books.'

51

The Swedish example in (80a) is from HOLMBERG (2002: 87). Many speakers find it questionable, but allow the same word order for certain other predicates than ge 'give', for instance tilldela 'confer, award', erbjuda 'offer'.

92

HALLDÖR ÄRMANN SIGURDSSON

2.3. Some notes on other agreement/concord types Four other agreement/concord types deserve mention here: I Π ΠΙ IV

Secondary predicate agreement DP-internal concord Noun-possessor agreement Complementizer agreement

I. Secondary predicate agreement crucially differs from simple primary predicate agreement in two respects (see the discussion in SIGURDSSON 2002: 708ff.). A

Secondary predicate agreement crosses predication boundaries, that is, it involves a predicate (adjective or participle) that agrees with a DP located in another (matrix) predication.

Β

While primary predicate agreement is blocked by inherent case, as discussed above, secondary predicate agreement is not subject to any such case restrictions. In other words, secondary predicates may agree with a DP irrespective of its case, whereas primary predicates can only agree with structurally case marked DPs (this follows from the blocking of 'agreement polygamy', discussed in section 2.1.4).

Some simple Icelandic examples are given in (82) (from SlGURBSSON 2002: 709). As usual, the agreement controller is underlined whereas the relevant agreeing predicate is boldface: (82a) Sträkarnir hittu kennarann drukknir. boys.the.N.M.PL met teacher.the. A.M.SG drunk.N.M.PL 'The boys met the teacher drunk (i.e. the boys were drunk).' (82b) Sträkarnir hittu kennarann drukkinn. boys.the.N.M.PL met teacher.the.A.M.SG drunk.A.M.SG 'The boys met the teacher drunk (i.e. the teacher was drunk).' (82c) Sträkarnir syndu kennaranum ovirdingu drukknum. boy.the.N.M.PL showed teacher.the.D.M.SG disrespect drunk.D.M.SG 'The boys showed the teacher disrespect (when he was) drunk.' In these examples, the secondary predicate is a small clause adjective, but it may also be in certain other types of adjuncts or in infinitival complements. Such examples may or may not involve overt case agreement (as discussed in THRÄINSSON 1979 and in SlGURBSSON 2002).52 Π. DP-internal concord involves agreement of nominal modifiers (adjectives, demonstratives, determiners, quantifiers, semi-predicates) in number, gender, case, and, less centrally, definiteness. This is briefly illustrated for definite DPs in (83) and (84), and for indefinite ones in (85)—(86):

52

That is, they may either show up in the nominative or in the same case as their overt agreement controller. - Gender and number agreement with the overt controller are always obligatory.

93

Agree and agreement: evidence from Germanic

jjorir four.N.M.PL fjora four.A.M.PL fjorum four.D.PL fjögurra four.G.PL

frcegu famous. PL.DEF frcegu famous.PL.DEF frcegu famous. PL.DEF frcegu famous. PL.DEF

karlar men.N.M.PL karla men.A.M.PL körlum men.D.M.PL karla men.G.M.PL

pessar fjorar (84a) Allar all.N/A.F.PL these.N/A.F.PLfour.N/A.F.PL (84b) Ollum pessum fjorum all.D.PL these.D.F.PL four.D.PL (84c) Allra pessara fjögurra all.G.PL these.G.F.PL four.G.PL

frcegu famous.PL.DEF frcegu famous.PL.DEF frcegu famous.PL.DEF

konur women.N/AF.PL konum women.D.F.PL kvenna women.G.F.PL

(83a) Allir all.N.M.PL (83b) Alia all.A.M.PL (83c) Öllutn all.D.PL (83d) Allra all.G.PL

(85a) Allir all.N.M.PL (85b) Alia all.A.M.PL (85c) Öllum all.D.PL (85d) Allra all.G.PL

pessir these.N.M.PL pessa these.A.M.PL pessum these.D.PL pessara these.G.PL

frcegir famous.N.M.PL.iNDEF frcega famous.A.M.PL.!NDEF frcegum famous.D.PL.lNDEF frcegra famous.G.PL.INDEF

(86a) Allar frcegar all.N/A.F.PL famous.N/A.F.PL.iNDEF (86b) Ollum frcegum all.D.PL famous.D.PL.lNDEF (86c) Allra frcegra all.G.PL famous.G.PL.INDEF

karlar men.N.M.PL karla men.A.M.PL körlum men.D.M.PL karla men.G.M.PL konur women.N/A.F.PL konum women.D.F.PL kvenna women.G.F.PL

Various types of disjoint or distant agreement are probably best analyzed as subtypes of DP-internal concord. For instance: (87a) Karlarnir firu allir ä fundinn. men.the.N.M.PL went all.N.M.PL to meeting.the (87b) Konurnar foru allar ά fundinn. women.the.N/A.F.PL went all. N/A.F.PL to meeting.the ΠΙ. Noun-possessor agreement. Possessive pronouns agree with their noun. In the Scandinavian languages, however, this is restricted to first and second person and to the reflexive third person possessives, i.e. third person non-reflexive possessives have an invariable genitive form. A few examples from Icelandic:53

53

The agreement has gone lost in first and second person plural in eveiyday Icelandic (GUBMUNDSSON 1972) and largely in Faeroese too (PETERSEN et al. 1998).

94

HALLDÖR ARMANN SIGURDSSON

(88a) Brcedur minir brothers my (88b) Broedum minum (88c) Systur minor sisters my

/ / / / /

Brcedur brothers Broeörum Systur sisters

pinir your pinum pinar your

/ / / / /

Brcedur brothers Brcedrum Systur sisters

(89) Brcedur hans / Brcedrum hans / brothers.N his / brothers.D his, /

sinir = N . M . P L his/her/their sinum = D . M . P L sinar = N / A . F . P L his/her/their

Systur hans sisters.N/A his

IV. Complementizer agreement. Some West-Germanic dialects have complementizer agreement. At least: West-Flemish, Groningen, South Hollandic, East Netherlandic, Luxemburgish, Bavarian, Brabantish (ZWART 1993a, 1993b and many others). This is illustrated in (90) for West Flemish and in (91) for Munic Bavarian (both from ZWART 1993a, 318-319). In the latter dialect, complementizer agreement is accompanied by pro-drop: (90a) dank that

(90b) datj that

ik

kom(e)n

I

come.lSG

Ü he

werkt works.3SG

(91a) damidds kommds there-with come.2PL 'So that you will come.' (91b) damid ich kom there-with I come. 1SG 'So that I will come.'

This concludes our descriptive overview of some of the central aspects of agreement and concord phenomena in the Germanic languages.

3. Concluding discussion Several agreement phenomena that are well-known from other language families are not found in Germanic. For instance: Preposition-complement agreement, verb agreement with more than one argument simultaneously, various types of 'external' complementizer agreement (as opposed to the West-Germanic 'internal' or rightward complementizer agreement). However, it seems fair to say that the richness of Germanic agreement phenomena, their complexity and variety, is impressive for such a limited and closely related group of languages. Reconsider the Germanic variation in (2) above, sketched again in (92): (92a) (92b) (92c) (92d)

Theywould-0berich-0. They would-AGR be rich-0. They would-0 be rich-AGR. They would-AGR be rich-AGR

English German Swedish Icelandic

Agree and agreement: evidence from Germanic

95

On standard assumptions, both agreement types are reflections of a deeper relation at the underlying syntactico-semantic level of LF, Agree in the terminology of CHOMSKY (e.g. 2000, 2001a). Given that all languages have the same basic structure, Universal Grammar, the relation in question must be universal, that is, it must 'be there' in a language like English, in spite of not being reflected in the morphology ofthat particular language, a 'silent' parametric choice. A structure that is void of morphological agreement, like the English one in (92a), clearly does not have 'agreement' in any plausible sense in common with a rich agreement structure, like the Icelandic one in (92d). Rather, the semantic content of the agreement features is common to both languages and so is their combination, Merge. Reconsider the Agree Condition on Merge, ACM, in(11) = (93): (93) Two objects or elements, X and Y, may be merged only if the relation of Agree holds between them. This condition is arguably, if not trivially, a law of nature, hence of the biological/physical system of language. Recall also multiple participle agreement as in (67) = (94): (94a) Peir mundu vera taldir vera sagdir they.N.M.PL would be believed.N.M.PL be said.N.M.PL hafa veriö kostiir. have been elected.N.M.PL 'They would be believed to be said to have been elected.' (94b) Egmundi telja ]xi vera sagda hafa I would believe them.A.M.PL be said.A.M.PL have veriö kosna. been elected.A.M.PL Ί would believe them to be said to have been elected.' The problem posed by data of this sort to the Probe-Goal Approach now vanishes. This is just an ordinary instance of agreement under repeated application of Merge/Agree. Nothing mysterious is involved. The same is true of the Swahili and Kayardild examples in (3)-(5). When two linguistic 'atoms' or units X and Y merge, they do not only become sisters, they also match by necessity. That is, X matches the requirements of Y and vice versa. 4 As discussed in section 2.1.5, matching is Nucleus-Edge Matching in the present approach.55 This entails that, for instance, objects of verbs are not really 54

As discussed at the end of section 2.1.4, inherent case-marking of 'complements' is, for instance, an agreement relation, where the case selection feature of a verbal or a prepositional nucleus matches the case feature of a DP. - A similar (albeit a more specific) approach to Merge is pursued by CONTRERAS & MASULLO 2001, and FRAMPTON & GUTMAN 2000 develop an approach to agree-

55

ment (as feature sharing) that bears some resemblance to the approach advocated here. CHOMSKY'S Probe-Goal Approach, based on Icelandic data, allows for a more general matching strategy, head-X matching, where X is the closest possible candidate for matching with the head (observing minimality). See further below.

96

HALLDÖR ARMANN SIGURDSSON

complements but edge-elements in 'complements' that normally have a silent 'head' or nucleus, F: (95) [vp V [FP DP, F]] V and FP merge/agree, but the relevant matching features of FP are situated at its edge, in the DP. Notice that Move follows in the present approach as motivated by the needs of Merge/Agree: If X is to be merged with a selecting Y, its edge features must match the requirements of Y and this is accomplished by movement of an element containing the matching features to the edge of X (in addition, Move relates the 'inner spheres' of X to its edge). It follows that any main clause must be selected by a silent LF element, Fin (or Speech Time/Place, ST/p), 'heading' a Speech Phrase, SP (see below). It is conceivable that V in (95) is itself an edge element of a silent nucleus, H, and that it is in fact this nucleus that agrees with FP and matches the features of DP:57 (96)

[HP

V, Η [pp DP, F]]

The aspectual semantics of Icelandic case, recently discussed by SVENONIUS (2001, 2002), thus reside within the minimal shell that contains the DP and its selector (i.e. we need not postulate a higher aspectual little ν to account for these semantic facts - plausibly in contrast to aspectual case semantics in e.g. Finnish and the Slavic languages). On this General Shell Approach, all items of at least the major word classes come as shells, with a nucleus and some edge substance (whereas functional elements might be different). This would seem to fit neatly with DP-internal concord, for instance of an adjective and a noun: The nucleus Η is then not invisible, as in (95), but instead visible as an Agr element in a structure like (97): (97) [HP A, H agr [fp N, F]] However, I shall not speculate further here. The important generalization is this: Whenever two linguistic objects X and Y merge, they agree by necessity, making up a featural bond X-Y. A featural bond may or may not be made visible in PF. That is, whenever Merge applies, the possibility of agreement arises, and a language has to make a choice whether or not to morphologically 'signal' it. Only by viewing Agree as being a condition on and an integrated part of Merge itself can we begin to understand the extreme variability of agreement across languages: Merge is the only process that is general enough to encompass all the variation observed within even such a limited and a closely related group of languages as the Germanic ones. The relations between syntactic Merge/Agree and morphological agreement are almost chaotically heterogeneous, being conditioned by syntactic as well as morphological (PF) factors. Probably by far the most common strategy is not to signal 56

57

For a conceptually related approach to particle verbs, where the particle is a head within the complement of the verb, see SVENONIUS 1996. Perhaps, all phonological material is edge substance, 'surrounding' a silent nucleus (consider CINQUE 1999 on lexical adverbs as 'specifiers' of silent 'heads').

97

Agree and agreement: evidence from Germanic

or visualize Merge/Agree at all, as in the English example in (92a) (and as, for instance, for merger of prepositions and DPs in all Germanic). The underlying Merge/Agree relation is evidently visible and making it 'extra' visible by morphological means is costly. Uneconomical strategies of this sort are obviously not banned in language, but there is a tendency to avoid them. As has been discussed by many, there is a strong relationship between movement and agreement in many constructions across languages. See for instance the overview in VAN GELDEREN 1 9 9 7 and some of the facts discussed in sections 2 . 1 . 2 and 2.2.2 above. Many cases of this sort can be understood in terms of visibility and economy. The typical pattern is as in (98), where morphological non-agreement is indicated by ' 0 ' , and where α is a feature or a property that attracts Y (the copy of Y is shown as a trace): (98a) a ... X-0 [ Y ] ( 9 8 b ) YJ-Α . . . X-AGRJ [ TI ]

Given our understanding of Agree, the merger of X and Y in (98a) turns them into a featural bond X-Y, which is 'locally visible'. Movement of Y across X, as in (98b), makes the structural relationship opaque, but this is compensated for by agreement of X, that is, agreement makes the relationship 'non-locally visible'. An alternative way of looking at this issue is to view Move or 'external Merge' as leading to a 'reinforced Agree'. Recall, from the discussion around (10), that Move is predestined by Merge, that is, it is 'triggered' by some property or feature that has already been merged, such as α in (98). However, the merger of the triggering property in the first place is an instance of Merge/Agree in the present approach, that is, a in (98a) agrees with the rest of the structure and the visible agreement of X in (98b) can thus be seen as an instance of Nucleus-Edge Matching, where a is the nucleus and X is the edge (with respect to a, simultaneously as being a nucleus with respect to the rest of the structure). There are many instances of non-reduced rightward, local agreement, but there is however a clear tendency to avoid such 'extravagance'. Many such seemingly 'unmotivated' cases of agreement might in fact be motivated by movement. Consider Icelandic participle agreement: (99) Pad voru skrifadar it were.PL written.N.F.PL 'There were three books written.'

priär three

bcekur. books.N.F.PL

Suppose that the predicate is embedded under an Event Phrase, EP, containing Event Tense, E T (as in SLGURBSSON forthcoming; E T resembles the low Τ in PLATZACK 2001, 2002). Assume also that the EP in turn contains a gender and a number nuclei, Gend and Num (distinct from the higher Num of the finite verb complex), serving the purpose of specifying potential event participants: (100) Τ ... [ E p... E T [Gendp three books.N.F.PL ... [NumP written.N.F.PL ...

58

In lack of a better term, and a deeper understanding, the triggering property is often referred to as ΈΡΡ'.

98

HALLDÖR ÄRMANN SIGURDSSON

The underlying object, prjär bcekur 'three books', has been moved to (or 'internally merged' in/at) Edge,GendP, thereby motivating morphological agreement of the participle skrifadar 'written' in NumP. Subsequently, the verb moves to Edge, EP, where it enters into Nucleus-Edge Matching with Τ (which, in turn, must match or link to Speech Tense, ST).59 Compare (99) to the Swedish (79c), repeated here in a slightly simplified form as (101): (101) Pet blev skrivet tre böcker. it were written.N.SG three books 'There were three books written.' Swedish, like other mainland Scandinavian varieties, has a much reduced gender agreement system, as compared to Icelandic. That is: I. Singular: neuter vs. non-neuter, e.g. skrivet vs. skriven 'written' II. Plural: no gender distinction, e.g. skrivna 'written' Suppose, therefore, that the Swedish Gend is weak in the sense that GendP cannot contain any phonological material.60 If so, the DP 'three books' does not raise to Edge,GendP, across the participle, and hence participle agreement is unmotivated.61 Reconsider the problems raised by Reverse Agreement facts, that is, facts of the sort that lead to the development of the Probe-Goal Approach in the first place. Late Subject Agreement is a case in point. Consider the derivation of the simple (102). (102) Pad komu prir mdlvisindamenn. it came.PL three linguists.N.M.PL 'There arrived three linguists.' Let us tentatively assume that the EP here differs from the structure in (100) in that it contains neither GendP nor PartP. The nominative then raises to the (low) NumP of the EP, the verb raises to Edge,EP and Τ is merged, giving roughly the structure in (103): (103) T - [ E p came.PL [ Num p three linguists.N.M.PL . . . ]] The number feature of the verb matches the mumber feature of the nominative, but this is obviously not the end of the derivation. Tense relates the event tense, Εχ, to the speech tense, ST, hence to the speaker, but additional speaker related features must be merged as well, or else the predication gets no truth value nor would it be interpretable in other respects (as it would not be anchored in any center of deixis and consciousness). The features in question are minimally mood, (higher) number and person. After merger of M(ood) and (the higher) Num, and after raising of the verb to Edge,MP, we have the following structure. The two number features are

59 60

61

Notice that I am abstracting away from the question of where the auxiliary 'be' in (99) is merged. On this approach, the question of whether a feature is strong or not boils down to lexicalization, including movement, rather than exclusively to movement. However, when the DP raises across the participle, as in (79a,b), it triggers agreement of the participle. The intermediate position it raises to, as in (79b), is presumably a higher position than Edge,GendP, but I shall not pursue the issue here.

99

Agree and agreement: evidence from Germanic

distinguished as Num/ev (event related number) and Num/sp (speech situation related number): (104) Num/sp-[Mp came.PL... [num/evP three linguists.N.M.PL... ]] Num/sp and MP agree abstractly, or else they could not have merged, and, in addition, the plural value of Num/sp enters into a local Nuclear-Edge Matching relation with the verb (which subsequently raises further, to Num/spP and PersP). It follows that the morphological agreement relation of the finite verb and the 'late subject' is only apparently non-local. The same is possibly true of more complex structures, as in (105), although it is a non-trivial task to find evidence in favor of that suggestion: (105) Pad mundu pä sennilega [Mp hafa komiö prir mälvisindamenn\. it would.PL then probably have come three linguists.N.M.PL Alternatively, one could relativize the notion of edge with respect to individual features, such that the nominative argument counts as being the edge element of MP with respect to number (being its 'outermost' element containing number), hence entering into a local Nucleus-Edge Matching with Num/sp. In this case, then, our Nucleus-Edge Matching would be empirically non-distinct from the ProbeGoal Approach, augmented or conditioned by the Minimal Link Condition. It is, by the way, a remarkable property of compositionality that Merger/Agree of X and Y, resulting in the bond X-Y, need only involve local matching of a single feature of Y, even in case Y is a highly complex structure. Finally, notice that formal features are neither interpretable nor uninterpretable in any absolute sense. Clausal derivation is not driven by inherent uninterpretablity of some defect features, such as the structural cases.62 Rather, the computation crucially relates features of the Event Phrase to (necessarily silent) features of the Speech Phrase. Informally put: We need to be able to say e.g. 'She hit me' and 'They hit us' instead of merely saying 'The hitter hit the hittee'. Person, as partially also number and gender, is a 'device' for this purpose of relating the event participants to the speech situation, thereby making the proposition interpretable and giving it a truth value. This relating of event features to speech situation features is brought about by Merger/Agree and Move. Abbreviations 1 2 3 A, ACC D, DAT G Ν, ΝΟΜ AGR DEF

62

first person second person third person accusative dative genitive nominative unspecified agreement definite

F IND INDEF Μ Ν PL SG SUBJ

feminine indicative indefinite masculine neuter plural singular subjunctive

See SIGURDSSON (forthcoming) on the relative interpretability of the structural cases.

100

HALLDÖR ARMANN SIGURBSSON

References BARNES, MICHAEL

P. (1986): Subject, nominative and oblique case in Faeroese, in: Scripta Islandica

37, 13-46. The structure ofjunctional categories, a comparative study of Arabic dialects. Oxford: Oxford University Press. BLAKE, BARRY J. (2001): Case. Second edition. Cambridge: Cambridge University Press. BOECKX, CEDRIC ( 2 0 0 0 ) : Quirky agreement, in: Studia Linguistica 5 4 , 3 5 4 - 3 8 0 . Βοδκονιύ, 2EUKO (2001): PF merger in Scandinavian: Stylistic fronting and object shift, in: Working Papers in Scandinavian Syntax 68, 75-115. BRODY, MICHAEL ( 2 0 0 0 ) : Mirror theory: syntactic representation in perfect syntax, in: Linguistic Inquiry 3 1 , 2 9 - 5 6 . CHOMSKY, NOAM ( 1 9 8 1 ) : Lectures on government and binding. Dordrecht: Foris. CHOMSKY, NOAM (1986): Barriers. Cambridge, Massachusetts: MIT Press. CHOMSKY, NOAM ( 1 9 9 1 ) : Some notes on economy of derivation and representation, in: FREIDIN, ROBERT (ed.), Principles and parameters in comparative grammar. Cambridge, Massachusetts: MIT Press, 4 1 7 - 4 5 4 . CHOMSKY, NOAM ( 1 9 9 3 ) : A minimalist program for linguistic theory, in: HALE, KENNETH & KEYSER, SAMUEL JAY (ed.), The view from building 20. Cambridge, Massachusetts: MIT Press, 1 - 5 2 . CHOMSKY, NOAM (1995): The minimalist program. Cambridge, Massachusetts: MIT Press. CHOMSKY, NOAM ( 2 0 0 0 ) : Minimalist inquiries: the framework, in: MARTIN, ROGER; MICHAELS, DAVID & URIAGAREKA, JUAN (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, Massachusetts: MIT Press, 89—155. CHOMSKY, NOAM (2001a): Derivation by phase, in: KENSTOWICZ, MICHAEL (ed.), Ken Hale: A life in language. Cambridge, Massachusetts: MIT Press, 1-52. CHOMSKY, NOAM (2001b): Beyond explanatory adequacy. Ms., MIT. CHRISTENSEN, KIRSTI KOCH & TARALDSSEN, KNUT TARALD ( 1 9 8 9 ) : Expletive chain formation and past participle agreement in Scandinavian dialects, in: BENINCÄ, PAOLA (ed.), Dialect variation and the theory ofgrammar. Dordrecht: Foris, 5 3 - 8 3 . CINQUE, GUGLIELMO (1999): Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. COLLINS, CHRIS (2002): Eliminating labels, in: EPSTEIN, SAMUEL DAVID & SEELY, Τ. DANIEL (eds.), Derivation and explanation in the minimalist program. Oxford: Blackwell Publishing, 42-64. COLLINS, CHRIS & THRAINSSON, HÖSKULDUR (1996): VP-internal structure and object shift in Icelandic, in: Linguistic Inquiry 27, 391-444. CONTRERAS, HELES & MASULLO, PASCUAL JOSE ( 2 0 0 1 ) : Motivating Merge. Ms., University of Washington and Universidad Nacional del Comahue. CORBEXR, GREVILLE ( 1 9 9 1 ) : Gender. Cambridge: Cambridge University Press. DAVISON, ALICE (forthcoming): Non-nominative subjects in Hindi/Urdu, V P structure and case parameters, in: BHASKARARAO, PERI & SUBBARAO, K . V . (eds.), Non-nominative subjects. Amsterdam: John Benjamins. DONALDSON, BRUCE ( 1 9 9 4 ) : Afrikaans, in: KÖNIG, EKKEHARD & AUWERA, JOHAN VAN DER (eds.), The Germanic languages. London: Routledge, 4 7 8 - 5 0 4 . EGERLAND, VERNER (1996a): The syntax ofpast participles, a generative study of nonfinite constructions in Ancient and Modern Italian. Lund: Lund University Press. EGERLAND, VERNER (1996b): On pronoun positions in Swedish and Italian, antisymmetry and the Person Phrase, in: University of Venice Working Papers in Linguistics 6.2,65-104. EINARSSON, STEFAN ( 1 9 4 5 ) : Icelandic. Grammar, texts, glossary. Baltimore: The Johns Hopkins Press. ETHNOLOGUE: Languages of the World [www.ehtnologue.com]. FRAMPTON, JOHN & GUTMAN, SAM (2000): Agreement is feature sharing. Ms., Norhteastern University. FRANKS, STEVEN (1995): Parameters of Slavic morphosyntax. Oxford: Oxford University Press. FRIDJÖNSSON, JÖN ( 1 9 8 9 ) : Samsettar myndir sagna [Complex verbal categories], Reykjavik: Institute of Linguistics. FRIDJÖNSSON, JÖN ( 1 9 9 0 - 1 9 9 1 ) : Beygingarsamraemi meö samsettu frumlagi. [Agreement with a complex subject], in: Islenskt mal og almenn mälfreedi 1 2 - 1 3 , 7 9 - 1 0 3 . BENMAMOUN, ELABBAS (2000):

Agree and agreement: evidence from Germanic

101

Verbal agreement and the grammar behind its breakdown: minimalist feature checking. Tübingen: Niemeyer. GREENBERG, JOSEPH ( 1 9 6 6 ) : Some universale of grammar with particular reference to the order of meaningful elements, in: JOSPEH GREENBERG (ed.), Universals of language. Second edition. Cambridge, Massachusetts: MIT Press, 61-113. GUBMUNDSSON, HELGI (1972): The pronominal dual in Icelandic. Reykjavik: Instititue of Nordic Linguistics. HAEBERLI, ERIC ( 2 0 0 2 ) : Features, categories and the syntax ofΑ-positions. Cross-linguistic variation in the Germanic languages. Dordrecht: Kluwer. HAIDER, HUBERT (1997): Projective economy - on the minimal functional structure of the German clause, in: ABRAHAM, WERNER & VAN GELDEREN, ELLY (eds.), German syntactic problems, problematic syntax. Tübingen: Niemeyer, 83-103. HAIDER, HUBERT (2001): How to stay accusative in insular Germanic, in: Working Papers in Scandinavian Syntax 68,1-14. HAUSER, MARC D . ; CHOMSKY, NOAM & FrrcH TECUMSEH, W . ( 2 0 0 2 ) : The faculty of language: what is it, who has it, and how did it evolve?, in: Science 2 9 8 , 1 5 6 9 - 1 5 7 9 . HOLMBERG, ANDERS ( 1 9 9 4 ) : The pros and cons of agreement in Scandinavian impersonals, in: CINQUE, GUGLIELMO et al. (eds.), Paths towards universal grammar, studies in honor of Richard S. Kayne. Washington, D . C . : Georgetown University Press, 2 1 7 - 2 3 6 . HOLMBERG, ANDERS (2000): Scandinavian stylistic fronting: how any category can become an expletive, in: Linguistic Inquiry 31,445-483. HOLMBERG, ANDERS ( 2 0 0 2 ) : Expletives and agreement in Scandinavian passives, in: Journal of Comparative Germanic Linguistics 4 , 8 5 - 1 2 8 . HOLMBERG, ANDERS & HRÖARSDÖTTIR, THORBJÖRG (2002): Agreement and movement in Icelandic raising constructions, in: Working Papers in Scandinavian Syntax 69,147-168. HOLMBERG, ANDERS & PLATZACK, CHRISTER ( 1 9 9 5 ) : The role of inflection in Scandinavian syntax. Oxford: Oxford University Press. J0NSSON, JÖHANNES GISLI ( 1 9 9 1 ) : Stylistic fronting in Icelandic, in: Working Papers in Scandinavian Syntax 48, 1-43. J0NSSON, JÖHANNES GISLI ( 1 9 9 2 ) : The two perfects of Icelandic, in: Islenskt mil og almenn mälfrceöi GELDEREN, ELLY VAN ( 1 9 9 7 ) :

14,129-145.

JÖNSSON, JÖHANNES GfSLi (1996): Clausal architecture and case in Icelandic. Doctoral dissertation, University of Massachusetts. JULIEN, MARIT (2002): Syntactic heads and wordformation. Oxford: Oxford University Press. KAYNE, RICHARD S. ( 1 9 8 9 ) : Facets of Romance past participle agreement, in: BENINCÄ, PAOLA (ed.), Dialect variation and the theory ofgrammar. Dordrecht: Foris, 8 5 - 1 0 3 . KAYNE, RICHARD S. (1994): The antisymmetry of syntax. Cambridge, Massachusetts: MIT Press. MAGNÜSSON, FRIDRIK ( 1 9 9 0 ) : Kjarnafxrsla ogfyaö-innskoti aukasetningum i islensku [Topicalization and/»ad-insertion in subordinate clauses in Icelandic], Reykjavik: Institute of Linguistics. MALING, JOAN ( 1 9 8 0 ) : Inversion in embedded clauses in Icelandic, in: Islenskt mäl og almenn mälfrceöi 2 , 1 7 5 - 1 9 3 . [Reprinted in MALING, JOAN & ZAENEN, ANNIE (eds.) ( 1 9 9 0 ) : Modern Icelandic syntax. San Diego: Academic Press, 7 1 - 9 1 . ] NESS, SILKE VAN ( 1 9 9 4 ) : Pennsylvania German, in: KÖNIG, EKKEHARD & AUWERA, JOHAN VAN DER (eds.), The Germanic languages. London: Routledge, 4 2 0 - 4 3 8 . NYGAARD, MARIUS (1906): Norren syntax. Oslo: H. Aschehoug & Co (W. Nygaard). PESETSKY, DAVID & TORREGO, ESTHER ( 2 0 0 1 ) : Tense-to-C movement: causes and consequences, in: KENSTOWICZ, MICHAEL (ed.), Ken Hale. A life in linguistics. Cambridge, Massachusetts: MIT Press, 3 5 5 - 4 2 6 . PETERSEN, HJALMAR P . ; JACOBSEN, JÖGVAN Ί LON; HANSEN, ZAKARIAS S. & THRÄINSSON, HÖSKULDUR

(1998): Faeroese. An overview for students and researchers. Ms., Torshavn and Reykjavik. The Scandianvian languages and the null-subject parameter, in: Natural Language and Linguistic Theory 5, 3 7 7 - 4 0 1 . PLATZACK, CHRISTER ( 2 0 0 1 ) : The computational system as a minimal feature driven device and the tripartite TP/VP-hypothesis of the universal clause, in: GLOW Newsletter 46,46-47. PLATZACK, CHRISTER (2002): Agreement and the person phrase hypothesis. Ms., Lund University. POLETTO, CECILIA ( 2 0 0 0 ) : The higher functional field. Oxford: Oxford University Press. PLATZACK, CHRISTER ( 1 9 8 7 ) :

102

HALLDÖR ARMANN SIGURDSSON

RIZZI, LuiGi (1990): Relativized minimality. Cambridge, Massachusetts: MIT Press. Rizzi, LUIGI (1997): The fine structure of the left periphery, in: HAEGEMAN, LILIANE (ed.), Elements of grammar. Handbook in generative syntax. Dordrecht: Kluwer, 281-337. ROGNVALDSSON, EMKUR & THRÄINSSON, HOSKULDUR (1990): On Icelandic word order once more, in: MALING, JOAN & ZAENEN, ANNIE (eds.), Modem Icelandic syntax. San Diego: Academic Press, 3 - 4 0 . ROHRBACKER, BERNHARD (1994): The Germanic VO languages and the full paradigm: a theory ofV-to-I raising. Doctoral dissertation, University of Massachusetts, Amherst SHLONSKY, UR (1989): The hierarchical representation of subject verb agreement Ms., Haifa University. SIGURBSSON, HALLDÖR ÄRMANN ( 1 9 8 9 ) : Verbal syntax and case in Icelandic. Doctoral dissertation, University of Lund. [Published in Reykjavik 1992 by Institute of Linguistics]. SIGURDSSON, HALLDÖR ARMANN (1990-1991): Beygingarsamraemi [Agreement], in: Islenskt mäl og almenn mälfrcedi 12-13, 31-77. SIGURDSSON, HALLDÖR ÄRMANN (1991): Icelandic case-marked PRO and the licensing of lexical arguments, in: Natural Language and Linguistic Theory 9,327-363. SIGURDSSON, HALLDÖR ARMANN (1993): Agreement as head visible feature government, in: Studia Linguistica 47,32-56. SIGURDSSON, HALLDÖR ÄRMANN ( 1 9 9 4 ) : Morphological case, abstract case and licensing, in: HEDLUND, CECILIA & HOLMBERG, ANDERS (eds.), Proceedings of the XIV* Scandinavian Conference of Linguistics ... Special Session on Scandinavian Syntax. Gothenburg: Department of Linguistics, 1 4 5 - 1 5 9 . SIGURDSSON, HALLDÖR ÄRMANN ( 1 9 9 6 ) : Icelandic finite veib agreement, in: Working Papers in Scandinavian SyntaxSl, 1-46. SIGURDSSON, HALLDÖR ÄRMANN ( 1 9 9 7 ) : Stylisticfronting.Paper presented at the conference 'Subjects, Explitives, and the EPP\ Tromse, June 6-7 1997. SIGURDSSON, HALLDÖR ÄRMANN (2000): The locus of case and agreement, in: Working Papers in Scandinavian Syntax 65,65-108. SIGURDSSON, HALLDÖR ÄRMANN ( 2 0 0 1 ) : Inflectional features and clausal structure, in: NIEMI, JUSSI & HEIKKINEN, JANNE (eds.), Nordic and Baltic morphology: papers from a NorFA course, Tartu, June 2000. (Studies in Languages 36). Joenssu: University; Faculty of Humanities, 99-111. SIGURDSSON, HALLDÖR ÄRMANN (2002): To be an oblique subject: Russian vs. Icelandic, in: Natural Language and Linguistic Theory 20,691-724. SIGURDSSON, HALLDÖR ÄRMANN (2003a): Case: abstract vs. moiphological, in: BRANDNER, ELLEN & ZINZMEISTER, HEIKE (eds.), New perspectives on case theory. Stanford: CSLI Publications, 223-268. SIGURDSSON, HALLDÖR ARMANN (2003b): Icelandic left edge phenomena: Fin matches +/- displacement. Paper presented at the 4th Null-Subject Workshop, Durham, May 20,2003. SIGURDSSON, HALLDÖR ÄRMANN (forthcoming): Icelandic non-nominative subjects: facts and implications, in: BHASKARARAO, PERI & SUBBARAO, K.V. (eds.), Non-nominative subjects. Amsterdam: John Benjamins. STARKE, MICHAL (2001): Move dissolves into merge: a theory of locality. Doctoral dissertation, University of Geneve. SVENONIUS, PETER (1996): The optionality of particle shift, in: Working Papers in Scandinavian Syntax 57,47-75. SVENONIUS, PETER ( 2 0 0 1 ) : Case and event structure, in: ZAS Papers in Linguistics 26 [http://www. zas.gwz-berlin.de/papers/zaspil/infos/]. SVENONIUS, PETER (2002): Icelandic case and the structure of events, in: Journal of Comparative Germanic Linguistics 6, 197-225. SUBBARAO, Κ. V. (2001): EPP, agreement and non-nominative subjects in Telugu. Ms., ILCAA Tokyo and University of Delhi. THRÄINSSON, HÖSKULDUR (1979): On complementation in Icelandic. New York: Garland. THRÄINSSON, HÖSKULDUR ( 2 0 0 1 ) : U M sagnbeygingu, sagnfaerslu og setningagerö Ί faereysku og fleiri mälum [On verb-inflection, verb-movement and clause structure in Faeroese and some other languages], in: Islenskt mäl og almenn mälfrcedi 23, 7-70. VIKNER, STEN (1995): Verb movement and expletive subjects in the Germanic languages. Oxford: Oxford University Press.

Agree and agreement: evidence from

Germanic

103

VIKNER, STEN (1997-1998): V°-to-I° flytning og personfleksion i alle tempore ['V°-to-I° movement and inflection for person in all tenses'], in: Islenskt mal og almenn mälfrceöi 19-20, 85-132. ZWART, JAN-WOUTER (1993a): Verb movement and complementizer agreement, in: BOBAUDC, JONA-

THAN D. & PHILLIPS, COLIN (eds.), Papers on case and agreement I (MIT Working Papers in Linguistics 18). Cambridge, Massachusetts: ΜΓΓ, 297-340. ZWART, JAN-WOUTER (1993b): Clues from dialect syntax: complementizer agreement, in: ABRAHAM, WERNER & BAYER, JOSEF (eds.), Dialektsyntax. (Linguistische Berichte Sonderheft 5). Opladen: Westdeutscher Verlag, 246-270.

The semantics of the impersonal construction in Icelandic, German and Faroese: beyond thematic roles* J0HANNA BARDDAL (Lund & Bergen)

1. Introduction Traditionally, dative case marking of (logical) subjects in Germanic, as well as in other languages, has been explained by referring to the fact that such predicates are experience-based predicates and, thus, that the (logical) subject is an Experiencer, or, in the case of benefactive predicates, a Beneficiary (Goal).1 Such an approach has been adapted by linguists irrespective of theoretical frameworks (see ZAENEN et al. 1985, SlGURBSSON 1989, CROFT 1993, FlLIP 1996, FALK 1997, various papers in AlKHENVALD et al. 2001, and many more). This correlation between thematic roles and dative case marking is shown for Icelandic in examples (1-2) below: (1) (la)

(lb)

*

1

Experiencer Mir likar ägcetlega vid hann. me.DAT likes.3P.SG well with him.ACC Ί like him.' Honum brd pegar hann heyrdi him.DAT was-startled.3P.SG when he.NOM heard 'He was startled when he heard the news.'

frettirnar. news-the.ACC

I am indebted to LENA EKBERG, THÖRHALLUR EYTHÖRSSON, WILLEM HOLLMANN, CARJTA PARADIS and SUSAN REED for discussions and comments, to WERNER ABRAHAM and UTE BOHNACKER for help with the German data, to CARITA PARADIS for helping me find adequate terminology to describe the data discussed in section 4, and to the audience at the "10th Postgraduate Linguistics Conference at the University of Manchester", 31 March 2001, and "Semantik i .fokus" in Lund, 17-18 October 2002, where earlier versions of this work have been presented. Finally, I am immensely grateful to the editor of this volume, WERNER ABRAHAM, for revision suggestions which have substantially improved this paper. Accusative, dative and genitive logical subjects have been shown to behave syntactically as nominative subjects in Modem Icelandic and Faroese (first established in ANDREWS 1 9 7 6 , and later worked out in detail by ZAENEN et al. 1 9 8 5 , SIGURBSSON 1 9 8 9 , JÖNSSON 1 9 9 6 for Icelandic, and BARNES 1 9 8 6 for Faroese). The standard analysis, however, for Modern German is that these are syntactic objects. This assumption has recently been challenged by BARDDAL (2002), BARDDAL & EYTHÖRSSON (2003a, 2004) and EYTHÖRSSON & BARDDAL (2003) on basis of examples involving reflexivization, conjunction reduction and control infinitives. For the purpose of this paper, however, I will use the descriptive term "logical subject" since the semantic analysis of the impersonal construction does not hinge upon the syntactic status of the nominals in question.

106

J0HANNA BARDDAL

(2) Goal (2a) Henni

barst

pakki

ί

her.DAT received.3P.SG package.NOM in 'She received a package this morning.'

morgan. morning.ACC

(2b) Henni äskotnadist pessi lampi ί her.DAT got-possession.3P.SG this.NOM lamp.NOM in 'She got (by good luck) possession of this lamp last year.'

fyrra. last-year.ACC

Recently, however, J0NSSON (1997-98) has pointed out that a dative subject in Icelandic can also exhibit the thematic role of a Patient (3a-b) or a Theme (4a-b): (3) (3a)

Patient Sveini batnaöi. Sveinn.DAT got-better.3P.SG 'Sveinn recovered (from the illness).' (3b) Sveini hnignadi. Sveinn.DAT deteriorated.3P.SG 'Sveinn deteriorated.' (4) Theme (4a) Bätnum hvolfdi. boat-the.DAT turned-over.3P.SG 'The boat capsized.' (4b) Snjonum kyngir nidur. snow-the.DAT swallows.3P.SG down 'The snow falls thick.' These data clearly show that the concept of thematic roles/participant roles such as Experiencer and Beneficiary do not capture the facts of non-canonical case marking of (logical) subjects, at least not in Icelandic. J0NSSON, thus, argues that dative subject predicates in Modern Icelandic should be divided into four major classes depending on the thematic role of the subject. Such an analysis is based on the now generally accepted idea that thematic roles/participant roles are not semantic primitives but derivatives of verbal semantics (cf. JACKENDOFF 1990, GOLDBERG 1995), with verbal semantics including both the causal and aspectual structure of predicates as well as their lexical content (CROFT 1998, in preparation). In the same vein as J0NSSON, I have elsewhere (BARDDAL 2001a: 59ff., 2003a) analyzed dative subjects of active verbs in Icelandic as carrying the thematic roles of Experiencers, Cognizers, Perceivers, Stations, Beneficiaries and Themes (see the publications cited above for definitions). This analysis is based on a more thorough categorization of dative subject predicates into semantic classes, including predicates which are stative but not experience-based. My conclusions are in conformity with JÖNSSON's that dative subject predicates in Icelandic, compared with nominative subject predicates, are better described as simply being non-agentive than having the thematic roles of Experiencers and/or Beneficiaries (BARDDAL 2001a: 103).

The semantics of the impersonal construction in Icelandic, German and Faroese

107

However, the basic problem with all accounts relying on the concept of thematic

roles is that they do not capture the semantic similarities which can be discerned to hold between the various subclasses of impersonal predicates. That is, it does not follow from each thematic role whether or not there are semantic similarities between the classes, or how the particular predicate types within one category of thematic roles are related to other types within other categories of thematic roles. It is, thus, not clear from this approach what verbs like lika 'like' and hvolfa 'capsize' have in common (see section 3.4 below). In this paper I offer an account which deals with these issues. Moreover, certain dative selecting predicates in Icelandic, German and Faroese are of such a nature that it demands that we look beyond the concept of thematic roles in order to provide a satisfactory motivation for the non-canonical case marking of the (logical) subject. I raise the question why the (logical) subject of experience-based predicates should be non-canonically case marked to begin with, and I offer a solution based on cognitive linguistics, more precisely on notions such as affectedness, shared semantics and family resemblance. As stated above, I illustrate that there are predicates in the Germanic languages which do not express any internal experience of the referent denoted by the (logical) subject, but are still non-canonically case marked, including both stative non-experience-based predicates and more "eventive" non-experience-based predicates. I argue that this is motivated by shared semantics in general and pragmatic inferences in particular, including a process of subjectification resulting in a change of meaning of the (logical) subject. Thus, the non-canonical case marking of the (logical) subject has come to be an expression of the attitudes/judgments of the speaker and not an expression of the attitude of the referent denoted by the (logical) subject. I therefore conclude that in order to account for the non-canonical case marking of (logical) subjects of impersonal predicates it is essential that we recognize two levels of relations: a) the semantic relation holding between the referent denoted by the (logical) subject and the "event" denoted by the predicate (the semi-factual level) b) the emphatic relation holding between the speaker and his/her attitudes towards the content of the proposition encoded in the utterance (the subjective level) I begin by giving a definition of the impersonal construction used in this paper (section 2), before I proceed to the semantics and distribution of impersonal predicates in the three modern Germanic languages in which such predicates have been maintained, i.e. Icelandic, German and Faroese. These three languages show the same pattern of distribution, of which Icelandic has the most extensive and elaborated category of impersonal constructions. On the basis of the distribution of the construction across (partially) related semantic fields a conceptual network will be suggested, with various subconstructions of the impersonal construction occupying adjacent regions in semantic space (section 3). I then discuss a small subclass of performance predicates, of which the case marking of the (logical) subject cannot be explained without assuming that the function of the form of this argument has

108

J0HANNA BARDDAL

changed from being subject oriented to becoming speaker oriented (section 4). Finally, in section 5,1 summarize the main findings of this paper.

2. The impersonal construction: a definition Impersonal predicates are traditionally defined as predicates not selecting for a nominative argument, as, for instance, weather verbs which in many languages occur without a nominal argument (5a), and predicates which do select for nominal arguments, however, none in the nominative case (5b-c): (5) Icelandic (5a) I goer rigndi. in yesterday rained.3P.SG 'Yesterday, it rained.' (5b) Mig dreymdi ömmu. me.ACC dreamt.3P.SG grandma.ACC Ί dreamt about grandma.'

(5c) Mer

er

me.DAT is.3P.SG

kalt. cold.3P.SG

'I'm cold.'

The object of my investigation, however, is predicates which select for (logical) subjects that are non-canonically case marked as dative. Therefore, in this study I include predicates that take (logical) objects case marked as nominative, since they always occur together with non-canonically case-marked (logical) subjects (6a-c). (6a)

Icelandic Mir likar

pessi

me.DAT likes.3P.SG this.NOM

Ί like this book.' (6b) German Mir gefällt

dieses

me.DAT likes.3P.SG this.NOM

(6c)

Ί like this book.' Faroese Honum trytur

bok. book.NOM

Buch. book.NOM

[PETERSEN 2002: 70]

peningur.

him.DAT lacks.3P.SG money.NOM

'He lacks money.' Observe that these predicates are traditionally not defined as "impersonal" since one of their arguments is in nominative case. They are, however, included in this research since the logical subject is non-canonically case marked as dative. I also include predicates which have developed expletive uses in German, since the logical subject of these is non-canonically case marked as accusative or dative (7a-b):

109

The semantics of the impersonal construction in Icelandic, German and Faroese (7) German (7a) Mir mangelt

es

an... /Es

me.DAT lacks.3P.SG it.EXPL of... /it.EXPL

' H a c k . . . ' / ' I lack...' (7b) Mich fröstelt. me.ACC shivers.3P.SG

/Es

mangelt

mir

an...

lacks.3P.SG me.DAT of...

fröstelt

/it.EXPL shivers.3P.SG

mich. me.ACC

Ί shiver.' / Ί shiver.' To summarize, the predicates under discussion in this paper all select for a noncanonically case-marked (logical) subject. Some of these predicates can also occur in alternative constructions, such as the nominative subject construction, the expletive construction, or the DAT-NOM construction. However, the common denominator is that these predicates can select for a non-canonically case-marked (logical) subject, which is the property under investigation here. For the sake of convenience, however, I will use the term "impersonal" throughout this paper to refer to predicates of all the types discussed above and presented in examples (5-7). Impersonal predicates selecting for non-nominative (logical) subjects are numerous in Icelandic: Approximately 700 predicates select for dative subjects. These can be of various predicate structures, ranging from simple verbs to complex predicates involving the verb vera 'be' together with an adjective, noun or a prepositional phrase (see BERN0DUSSON 1982, SLGURBSSON 1989 and subsequent work, BARBDAL 2001b). There are approximately 200 predicates which select for accusative subjects in Icelandic, and presumably not more than 10 that take genitive subjects (see BARBDAL 2001a: 136, based on a list compiled by J0NSSON 1998). These counts include different entries for the same lexical verbs. In present-day German and Faroese, however, many impersonal predicates can be constructed with a nominative subject as well as with an accusative or a dative (logical) subject (see BARNES 1986, JONAS 2002, PETERSEN 2002, EYTH0RSSON & J0NSSON 2003 for Faroese; SEEFRANZ-MONTAG 1983, 1984, SMITH 1994, 1996

for German). A similar change of accusative (logical) subjects either taking on nominative case, or more often dative case, has also been documented in Icelandic (SVAVARSD0TTIR 1982, RÖGNVALDSSON 1983, SVAVARSD0TTIR et al. 1984,

HALLD0RSSON 1982, EYTHORSSON 2000, 2002, BARBDAL 2001a: 134-137, 200, BARBDAL & EYTHORSSON 2003b: 465-468). This change is much less pervasive in Icelandic than in Faroese and German in which non-nominative logical subjects are close to being extinct. The verb class which best seems to have maintained the case marking of non-nominative (logical) subjects in German and Faroese are twoplace DAT-NOM predicates, of the type gefallen 'like, please', shown in (6) above. Another difference between Icelandic and the two other Germanic languages is that while Icelandic has maintained most impersonal predicates, many of these have fallen into disuse in both Faroese and German (SEEFRANZ-MONTAG 1983, 1984 for German; JONAS 2002, PETERSEN 2002 and EYTH0RSSON & J0NSSON 2003 for Faroese). It is also true for English, Dutch and Mainland Scandinavian, which have completely lost the impersonal construction, that the loss of the construction is concomitant in part with the loss of the relevant vocabulary items.

110

J0HANNA BARDDAL

Subsequently, since the same impersonal predicate can occur with a (logical) subject in nominative, accusative or dative, the two classes of impersonal predicates selecting for accusative or dative (logical) subjects are not easily kept apart as separate categories in Faroese and German. Also, the verb classes themselves have shrunk in size to approximately 50-60 predicates in Faroese and perhaps 80-100 in German. The analysis presented here is based on two lists of impersonal predicates in Faroese, found in PETERSEN (2002) and EYTH0RSSON & J0NSSON (2003), and a list of German predicates selected from various reference grammars of Modern High German, compiled for the purpose of this paper (see Appendix). Because of space limitations I confine this study to predicates selecting for dative subjects in Icelandic, leaving the other large class of impersonal predicates, i.e. accusative subject predicates, to future research. I include, however, in this study all impersonal predicates in German and Faroese that I know of, since in these languages the basis for a classification into accusative and dative selecting verbs is not well founded. Therefore, I subsume these into one class of oblique selecting predicates.

3. The distribution of the impersonal construction A thorough study of the lexical semantics of the 700 impersonal predicates in Icelandic reveals that the following major predicate classes can be distinguished: (8)

Verbs denoting Perception, Cognition, Idiosyncratic Attitudes, Emotions, Bodily States, Changes in Bodily States, Decline, Personal Properties, Failing /Mistaking, Success/Performance, Existence, Social Interaction, and Gain.

Since impersonal predicates are so numerous in Icelandic, it is impossible to list them all in an overview paper like the present one (see, however, J0NSSON 199798, 1998). Therefore, I will provide some examples from Icelandic of the predicates in each class, as well as examples from Faroese and German when relevant, together with an outline of my analysis of the case marking of the (logical) subject and its motivations. Following BYBEE (1985: ch. 5) I assume that grammar is organized in such a way that there are similarity links connecting items which are semantically and/or formally similar. There are several psycholinguistic motivations for assuming such similarity links between items: One motivation stems from slips of the tongue, namely from the subclass of speech errors which involves word blends, i.e. mixtures of two lexically or functionally synonymous words (see, for instance, LAUBSTEIN 1999 on speech errors). The production of word blends in spontaneous speech shows that speakers can access synonymous words simultaneously in their minds (cf. ROELOFS 1992). A second motivation for assuming similarity links is the emergence of suppletive paradigms in languages. Such development entails that a paradigm attracts a form from another paradigm, providing that the lemmas from both paradigms are similar with respect to their lexical meaning (go vs. went). A third motivation comes from the analysis of error patterns in psycholinguistic ex-

The semantics of the impersonal construction in Icelandic, German and Faroese

111

periments, but such patterns show that errors deviating from the default are usually motivated by similarity between the tested item and an existing item. This can be semantic similarity, formal similarity or both (for further motivations and references see BYBEE 1985, CROFT & CRUSE 2004: ch. 10). In my own work, which specifically involves morphological case, I have pointed out that both new verbs entering a case language like Icelandic and nonce verbs in experimental settings (which have a priori been given semantic content) are, at least in part, assigned argument structure and morphological case by speakers on the basis of the semantic content of the verb (BARBDAL 2000, 2001a: ch. 5). This is not to say that there is an inherent relation between case marking and certain semantics but rather that there are similarity links connecting synonymous verbs, and more generally verbs with similar lexical content, which again motivates the similarity in form. In other words, it seems that similarity in meaning motivates similarity in form. These facts about the productivity of morphological case in Icelandic provide further psycholinguistic evidence for the assumption that a grammatical network makes use of similarity links between items. Similarity links form the basis for conceptual networks of the type illustrated in Figures 2, 3 and 4 in section 3.4 below.2 Moreover, I assume that constructions are stored at different levels of schematicity, ranging from more abstract schematic constructions to concrete lexically-filled instantiations of constructions (cf. BYBEE 1985, 1995, LANGACKER 1988, BARBDAL 2000,2001b, 2002, CROFT 2001: 25-28,2003). I take lexical verbs to be stored as low-level verb-specific constructions, with links between the various instantiations of the same lexical predicate, including both the general constructions these predicates occur in as well as the more specific idiomatic constructions. This is illustrated for the Icelandic verb falla 'fall' in Figure 1 below. The figure shows different levels of schematicity for the impersonal construction, with two instantiations of the more specific idiomatic low-level falla at the bottom level, and links to other instantiations of this particular verb.

2

For a somewhat different approach to network links, see GOLDBERG (1995: ch. 3), in which several types of inheritance links are discussed. The difference between the inheritance hierarchy advocated in GOLDBERG and the conceptual/semantic space model fleshed out here can be regarded as following: The inheritance hierarchy takes formal similarity between various syntactic constructions to be its point of departure when accounting for the relations between constructions in an economic way. My aim, however, is to organize the vocabulary in a semantically coherent manner which is consistent with the similarities and differences in predicate and argument structure found with these lexical items. Needless to say, an ideal model of grammar must incorporate both these views.

112

J0HANNA BARDDAL

Fig. 1. The organization of constructions at different levels of schematicity At the top-most level in Figure 1 we find the overarching schematic impersonal construction without any specifications on the lexical material which can instantiate it, or of the particular form the impersonal construction can have. The second highest level in Figure 1 specifies that the impersonal construction can have dative, accusative or genitive subjects in Icelandic. The intermediate level shows that dative subject predicates can be either simple verbs or complex predicates. The nextlowest level hosts the particular lexical predicates which can occur in the impersonal construction, giving examples of four lexical verbs in Icelandic. The lowest level provides two concrete lexically-filled instantiations of the impersonal construction, both with the verb falla 'fall' (see 9 below). Finally, Figure 1 also shows that the lexical predicate falla 'fall' can occur both as an impersonal and as a personal predicate, with links from the lexical verb falla to its different instantiations. The broken lines show that there may be some intervening levels in the hierarchy, not relevant for the present discussion, which are omitted accordingly. Various lexical verbs occurring in the impersonal construction in Icelandic are similar to light verbs or function verbs in that their conventional meaning is not necessarily a part of the combined meaning of the whole. Again, falla 'fall' is such a verb. Consider the following examples: (9) Icelandic (9a) Mer fellur me.DAT falls.3P.SG 'Ilike this.'

petta

vel

ί

ged.

this.NOM

well

in

psyche.ACC

The semantics of the impersonal construction in Icelandic, German and Faroese

113

(9b) Mer fellust hendur. me.DAT fell-5i.3P.PL hands.NOM Ί gave up.' (9c) Mer fill petta ί skaut. me.DAT fell.3P.SG this.NOM in lap.ACC 'It got it by a coincidence.' In this case, the same lexical item,falla 'fall', is shared by three different predicates of the impersonal construction, i.e. 'like', 'give up' and 'get'. There is, thus, a similarity in form between the three types, and, consequently, a formal similarity link. Similarity links which are based on both meaning and form contribute to an increased salience or entrenchment of a construction (cf. BYBEE 1985: 117-119). There are a few more verbs in Icelandic which are recurrently used as lexical material in different predicate types of the impersonal construction. These are, for instance, renna 'run', halda 'hold', liggja 'lie', standa 'stand', koma 'come', leika 'move, play', vaxa 'grow', veita(st) 'give/be given', sld 'hit', bregda 'change', ganga 'go', vera 'be' and verda 'become'. These verbs are more or less evenly scattered across the various semantic classes of impersonal predicates (with the exception of vera and verdä), and thereby provide formal similarity links between the different classes, in addition to the shared semantic content. Moreover, some of the individual lexical predicates are polysemous, thus falling into more than one of the relevant verb classes. Consider the examples in (10) below: (10a) heyrast '(think to) hear', 'gather' (10b) koma i hug 'get an idea', 'remember' (10c) vera efst ί huga 'have in mind', 'be important' In accordance with my categorization in (8) above of impersonal predicates into semantic verb classes, to be described in more detail in next section, the verb heyrast can either be a Perception verb ('hear') or a Cognition verb ('gather'). The complex predicate koma ί hug is a Cognition predicate which can occupy different points in the semantic spectrum within cognition ('get an idea' and 'remember'), whereas vera efst ί huga can either be a Cognition verb ('have in mind') or a verb of Attitudes ('be important'). In addition to the family resemblance between the various predicates of the impersonal construction, to be discussed in the subsections below, certain lexical fields are more richly represented by the construction than others. Two such fields include verbs of saying and verbs of seeing, to mention only a few fields. (11a) Saying: segjast 'speak, tell', moelast 'speak, tell', talast 'speak, tell', vera lidugt um malbeinid 'talk a lot', vera ordlett 'be talkative', vera litt um mäl 'speak with ease', vera tungutamt 'mention sth. often', vera munntamt 'say sth. often', verda ekki oröa vant 'not be at loss for words', veröa svarafätt 'have no answers', veröa ordafatt 'have no answers', verda ordfall 'be at loss for words', vefjast tunga um tonn 'have problems speaking', vera stirt um mäl 'not be a good speaker', hrjöta af vörum 'accidentally speak', verda ad ordi 'happen to say', verda mismceli 'make a slip of the tongue', farast ord 'say',

114

J0HANNA BARDDAL

liggja ord til e-s 'speak about sb.', verda jjölrcett um 'have long discussions', verda tiörcett urn 'have long discussions', vera skafdrjugt 'have long discussions', verda ad ordum 'start to quarrel', verda sunduroröa 'start to quarrel', talast til 'negotiate' (lib) Seeing: verda starsynt ά 'stare', verda litid ά 'happen to look', missynast 'see wrongly', yfirsjäst 'overlook', synast 'appear', koma Jyrir sjonir 'appear for one's eyes', deprast sjon 'see worse', förlast syn 'see worse', glapna syn 'see worse', sortna Jyrir augum 'see everything black, lose consciousness', glyja Jyrir augum 'be blinded', vitrast syn 'appear in vision', vera eftirsjä '(lit.'lookback')'regret' At least in Icelandic, many impersonal verbs are either real- or near-synonyms. In BARDDAL (2001b: 55-56) I enlist many such synonyms for DAT-NOM/NOM-DAT predicates, and for the purpose of explication I want to add here the following class of verbs of success. Some are used about particular events which are being successfully carried out, others about success in general: (12a) Success (eventive): heppnast, länast, takast, audnast, lukkast, scelg'ast, verda ägengt (12b) Success (general): länast, audnast, lukkast, farnast, vegna, ganga, reida af, verda ägengt J0NSSON (1997-98: 28) observes that there seems to be only one verb of success in Icelandic which takes a nominative subject (ηά 'make, obtain'), while the remainder of the verbs in this semantic class take a dative subject. This is, however, expected since synonymous verbs often select for the same argument structure construction, and hence the same morphological case (cf. BRAINE et al. 1990, BARDDAL 2000, 2001a: ch. 5, 2001b). Observe that a subset of the predicates in (12), for instance, länast 'succeed, manage', audnast 'succeed, manage' and verda ägengt 'make progress', are enlisted in both usage domains, i.e. the specific (eventive) domain and the more general (non-eventive) domain. However, there is only a partial overlap between the verb sets used in the two domains: The verb heppnast 'succeed' is either used about a product (cf. the play was a success) or the producer of a particular product (cf. they succeeded with the play), but not to express that somebody is doing well/badly in life in general. The verb vegna 'do well/badly', on the other hand, is more or less only used about people and their general and social level of achievement in life, however that is defined. To summarize, by shared semantics I mean synonymy, near-synonymy, adjacent points in semantic spectra, semantic/lexical subfields, family resemblance, and metaphorical and pragmatic extensions of one usage domain to another. I will now give an overview of the various semantic subtypes of the impersonal construction in Icelandic, German and Faroese. I start with the most salient subconstruction, i.e. the one instantiated by verbs of Affectedness, and then I discuss the other less central subconstruction, which is lexically-filled by verbs that can be subsumed under the label Happenstance predicates.

The semantics of the impersonal construction in Icelandic, German and Faroese

115

3.1. Experience-based predicates 3.1.1. Emotive verbs The two largest classes of predicates in both Icelandic and German are the classes which I have chosen to call verbs of Emotion and Idiosyncratic Attitudes. Verbs of Emotion comprise predicates meaning 'feel good/bad (Ic. Ηόα vel/illa, G. gut/schlecht sein)', 'feel relief (Ic. letta vi'd)', 'be happy (Ic. vera kätt ί hug)', 'be in a bad mood (Ic. vera pungt ί skapi)\ 'be sad (verapröngt um hjartad)', 'be sorry (Ic. pykja leitt, G. verdrießen)', 'be angry (Ic. vera gramt ί gedi)\ 'like (Ic. lika, falla ί ged, G. gefallen, F. damd)\ 'dislike (Ic. mislika, G. missfallen)', 'be disgusted (Ic. bjööa viö, G. ekeln)', 'shudder (Ic. hrjosa hugur, G. grauen, F. nita)\ 'regret (Ic. vera eftirsjä, G. reuen, F. harma)\ 3.1.2. Idiosyncratic Attitudes The verb class I have labeled Idiosyncratic Attitudes contains predicates denoting attitudes and judgments, with some instances being fairly idiosyncratic. Prototypical such verbs are, for instance, 'be important to sb. (Ic. vera umhugad urn)', 'be against sth. (Ic. vera mötgerö ί e-u)\ 'be easy/difficult for sb. (Ic. vera audvelt/erfitt, G. schwer/leichtfalleny, 'be (impossible for sb. (Ic. vera (ekki) auöi&)\ 'be a mystery to sb. (Ic. vera räögäta)\ It is not a given that Emotive verbs and verbs of Attitudes are two separate verb classes and not one class, since attitudes and beliefs are mental states and as such they are intimately related to emotions and feelings (cf. VLBERG 1 9 8 3 , SWEETSER 1990).

3.1.3. Cognition verbs Verbs of Cognition, Perception and Bodily States are also closely akin to verbs of Emotion and Idiosyncratic Attitudes, since emotions usually take expression in the body, and physical well being affects our emotional states (cf. SWEETSER 1 9 9 0 ) . Our cognitive abilities are rooted in the mind which is associated with the head in most western cultures. Perception is sensing through physical, and thus bodily, organs, such as the eyes, ears and the skin. These domains of experience are associated with the body which again is central to how we conceptualize and construe our internal experiences (cf. LAKOFF & JOHNSON 1 9 8 0 : ch. 6, 1 9 9 9 , GIBBS 1 9 9 7 ) .

Typical Cognition verbs which select for a dative (logical) subject are verbs meaning 'suspect (Ic. bjööa ί grun, G. schwanen, F. gruna)\ 'have in mind (Ic. vera efst ί huga, G. vorschweben, F. vera innari)', 'remember (Ic. vera ί fersku minni, F. minnast)\ 'forget (Ic. liöa iir minni)' and 'understand (Ic. skiljast, F. skiljasty. 3.1.4. Perception verbs Perception verbs selecting for dative (logical) subjects are relatively few in Icelandic, although verbs meaning 'taste (Ic. smakkast, G. schmecken)', 'appear in

116

J0HANNA BARDDAL

vision (Ic. birtast)\ 'wrongly hear/see (Ic. missynast/misheyrasty are included in this class. 3.1.5. Bodily States Verbs of Bodily States are also few in number in Icelandic. Apparently, there is a division of labor between the dative and the accusative in Icelandic since a higher percentage of accusative selecting impersonals are verbs of Bodily States. Examples from this verb class involve verbs meaning 'feel pain (Ic. blceöa, G. weh tun)\ 'feel sick (Ic. vera illt, G. übel sein, F. standast vid)' and 'feel cold/warm (Ic. vera heitt/kalt, G. warm/kalt sein)'. 3.1.6. Changes in Bodily States There are plenty of predicates in both Icelandic and German which express changes in Bodily States. These are, of course, complex predicates involving the inchoative form of vera 'be', i.e. veröa 'become', together with adjectives expressing physical sensations: 'become sick (Ic. veröa illt, G. übel werden)' and 'become cold/warm (Ic. veröa heitt/kalt, G. warm/kalt werden)'. Simple verbs, used about changes in health in general, as for instance 'get better/worse from an illness (Ic. batna/versna)' also belong to this class. 3.1.7. Affectedness Together, the classes of Emotion verbs, verbs of Attitudes, Cognition verbs, Perception verbs, verbs of Bodily States and verbs denoting changes in Bodily States are the largest semantically coherent predicate class which selects for nonnominative (logical) subjects in Icelandic, German and Faroese. At this point the question arises why such predicates should be non-canonically case marked to begin with, thus deviating from the prototypical nominative case marking of (logical) subjects. Given the fact that Icelandic, German and Faroese only have four morphological case categories the choice of non-canonical case marking is limited by this small set to accusative, dative or genitive. In fact, the (logical) subject of impersonal predicates in Icelandic can be case marked with any of these three morphological case forms. However, since accusative, dative and genitive are the typical case forms used on objects, they also express affectedness to a much higher degree than the nominative (cf. LANGACKER 1991: 409-413, SMITH 2001). Logical subjects of experience-based predicates share this meaning of affectedness with prototypical objects of transitive predicates, since they typically involve conceptualizations construing individuals as entities being subject to external forces and happenings. Experience-based predicates of the type discussed in the present paper lack the causal structure associated with the transitive construction of a transmission of force from an initiator to an endpoint (cf. CROFT 1993, 1998, BARBDAL 2001b, SMITH 2001). Instead, they activate a construal profiling the affected endpoint itself.

The semantics of the impersonal construction in Icelandic, German and Faroese

117

3.2. Happenstance predicates 3.2.1. Gain A fairly small class of impersonal predicates in Icelandic and German are verbs of Gain. These include predicates meaning 'receive (Ic. berast, F. lutast)\ 'benefit (Ic. greiöast arfur, G. frommen)', 'have luck (Ic. veröa til happs)\ 'be offered sth. (Ic. bjodasty and 'have one's wishes come true (Ic. veröa ad 0sksinni)\ Verbs of Gain resemble passives of ditransitive verbs in that many ditransitives entail giving or transfer of possession. Thereby, the passive of ditransitives often entails some sort of gain for the dative referent in languages like Icelandic, German and Faroese which have maintained the dative form of the indirect object in the passive construction. Thus, dative passives of ditransitive verbs, with their semantic entailment, provide additional motivation for the maintaining of the impersonal construction, and vice versa, that the impersonal construction is an important factor contributing to the maintaining of the dative passive construction in these languages. A bidirectional relation like this one contributes to the increased entrenchment of both the impersonal and the dative passive construction. 3.2.2. Personal properties Verbs denoting Personal properties can be divided into the following subclasses: physical properties (Ic. liggja hdtt romur 'have a loud voice'), mental properties (Ic. vera e-d edlislcegt, G. angeboren sein 'be natural for sb.'), general inclination (Ic. hcetta til 'be inclined', vera flökurgjarnt 'prone to feeling sick', vera liöugt um mälbeinid 'be talkative', G. eigen sein 'be typical for sb.') and verbs of resemblance (Ic. svipa til 'resemble'). Property verbs typically invoke a construal of the body, like verbs of Bodily States, except that the inclination component, mentioned in the previous paragraph, is missing with verbs denoting Bodily States. In fact, Property verbs are similar to experience-based predicates in general in that Property verbs are inherently stative and experience-based predicates denote transitory states (see, for instance, CROFT in preparation, on the distinction between inherent and transitory states). Property verbs and experience-based predicates, thus, have their stative profile in common. 3.2.3. Existence Verbs of Existence are typically verbs expressing ontological states of affairs, such as 'be in a particular manner (Ic. vera hättaö, vera fariö)' and 'differ (Ic. muna)\ Observe that verbs of Existence are inherently stative like Property verbs. They differ from Property verbs, however, in that the (logical) subject of Property verbs is always a human being, while the (logical) subject of Existence verbs is not. 3.2.4. Decline A substantial amount of predicates expressing changes in Bodily States (section 3.1.6 above) are also used about changes in general (Ic. hnigna 'deteriorate') and changes in weather or time (Ic. afletta 'subside'), thereby invoking a construal of

118

J0HANNA BARDDAL

decline. In this case, the same lexical predicate can be used within more than one conceptual domain. Another subclass of impersonal predicates falling into the semantic class of verbs of Decline are predicates denoting 'heavenly emissions' (in MALlNG's 2002 terminology), as for instance verbs denoting 'snowing', 'raining' and 'thundering', which all select for dative (logical) subjects in Icelandic. These predicates can also be used metaphorically about people getting together (Ic. snjoa aö), invoking a construal of a scene where, for instance, individuals are viewed in terms of snow flings which 'fall thick on the ground'. A similar construal is found with the verb rigna 'rain' in utterances like 'It is raining cats and dogs' which describes a situation of an extensive crowd. The Icelandic verb rigna takes a dative when used in this manner. 3.2.5. Failing/Mistaking Verbs of Failing/Falling/Mistaking are closely related to verbs of Decline since falling denotes physical movement downwards. Many predicates belonging to this class indicate accidental falling (Ic. veröa fötaskortur, F. berast ä), thus providing a link to verbs denoting accidental turning upside-down or capsizing (Ic. hvolfa), accidental mistakes (Ic. mistakast, G. missglücken), such as, for instance, overlooking (Ic. yfirsjast), be unable to speak (Ic. veröa svarafätt), meet with an accident (Ic. hankast ä) and get hindrance (Ic. seinka, F. berast frä). 3.2.6. Performance/Success A fairly large verb class in Icelandic are Performance verbs, expressing events such as speaking (Ic. talast), working (Ic. vinnast) and performing acts (Ic. takast upp). Verbs of Success and the lack thereof are also a prominent subclass of Performance verbs (Ic. heppnast/misheppast). The notion of success often entails completion, thus completive and terminative predicates are also found in this class (Ic. takast 'manage, succeed', ljuka 'finish'). More generally, verbs of advancement, both physical (Ic. skola ά land 'wash ashore') and mental (Ic. fara fram, F. Πόα 'make progress'), are semantically/metaphorically related to verbs of Success. Verbs denoting an increase (Ic. vaxa fylgi) are subject to the same construal. Both German and Faroese have dative selecting predicates expressing success (G. glücken, F. lukkast) and failure (G. missglücken, F. mislukkast). These last predicates, therefore, can be categorized as belonging to both verbs of Failing/Mistaking and verbs of Performance/Success, thus providing a link between the two verb classes. 3.2.7. Social interaction Finally, the last category of verbs which can be subsumed under one and the same heading are verbs which refer to Social interaction, such as 'not getting along (Ic. lenda saman)', 'have arguments (Ic. veröa sunduroröa)', 'be friends (Ic. vera vel til vina)\ 'have discussions (Ic. veröa tiörcett um)' and 'negotiate (Ic. talast til)\

The semantics of the impersonal construction in Icelandic, German and Faroese

119

3.3. Less prototypical instances The overview so far suggests that all impersonal predicates in the Germanic languages can neatly be classified into different semantic categories. In reality, things are not so clear-cut. Certain predicates do not seem to belong to the core instances of any of the semantic classes which I have discussed here, but can either be regarded as less prototypical instances, or possibly as belonging to more than one class. Predicates expressing meanings like 'lack (Ic. verda skortur, G. mangeln, F. trota)', 'need (Ic. lengjast um, F. nytast)' and 'give up (Ic. fallast hendur)' are clearly mental state verbs. The question is whether they are Emotive verbs, verbs of Attitudes or a subclass of their own. Also, fallast hendur 'give up' is similar to Property verbs since it is construed in terms of physical appearance and changes in physical appearance (i.e. in this case that the hands fall down) and Property verbs typically activate such construals. Verbs expressing judgments like 'be of use (Ic. nytast, G. nützen, F. nytta)', and more modal-like instances such as 'be obliged (Ic. bera, G. obliken''), 'be allowed (to do sth.) (Ic. leyfast, vera velkomid, G. willkommen sein)' and 'get away with sth. (Ic. lidasty are not easily categorized either. Verbs with the meanings 'estimate (Ic. reiknast)\ 'agree (Ic. koma saman um, G. einleuchten)' and 'learn (Ic. Icerasty are more cognition-like. Irrespective of the details in the classification of the less prototypical predicates under discussion, they clearly express obligation, mental states, attitudes and/or judgments, and as such they pattern with the core class of impersonal predicates in Icelandic, German and Faroese. 3.4. Summary In sum, I have discussed various motivations for the non-canonical case marking of (logical) subjects in Icelandic, German and Faroese. I have argued that the core function of this kind of case marking is to signal affectedness, since the largest verb class in all three languages under discussion is experience-based, including Emotive verbs, verbs of Attitudes, Cognition verbs, Perception verbs, and verbs of (changes in) Bodily States. Other verb classes are related to the central class through shared semantics. Property verbs and experience-based predicates are both stative: Property verbs are inherently stative, while experience-based predicates denote transitory states. Existence predicates are also inherently stative, except that they do not require the subject to be human. Thereby, they differ from Property verbs which always take a human subject. Verbs expressing changes in Bodily States differ from verbs denoting Bodily States only with regard to the inception component, which is present in the former verb class, but missing in the latter class. Therefore, the former indicates processes whereas the latter indicates states. Moreover, the verb class expressing changes in Bodily States contains several predicates used about changes in health, and/or changes in general. These are often verbs expressing motion either upwards or downwards. Again, they provide a similarity link to predicates which denote movement downwards, such as snowing, raining, thundering, etc., and verb classes denoting movement forward, including both

120

J0HANNA BARDDAL

physical and mental advancement and progress. Another class of predicates is the one expressing physical falling, yet another type of motion downwards. These are, further, semantically/metaphorically related to predicates which involve failure in general, mistakes and accidental or non-volitional actions. Again, these last verb classes are similar to verbs denoting success and/or good performance in that they only differ with respect to the quality of the performance in question. Furthermore, the very concept of success is rooted in our attitudes, thus supplying a link to verbs of Attitudes. Also, success often entails completion, and completion entails termination. Finally, the verb class denoting Social interaction contains predicates that express speaking and discussing, which are thus similar to some of the Property verbs which denote an inclination to speak and express oneself. A subclass of Interaction verbs has the meaning 'decide', which is also found with some Cognition verbs. Yet another subclass of Interaction verbs has to do with whether people are getting along well or not, thus providing a link to verbs of good/bad Performance/Success and verbs of Failing. The relations and interrelations between impersonal predicates in Icelandic can, thereby, be represented as a network model based on the semantics of these predicates and the (different types of) similarity between them. For the sake of simplicity the network representation in Figure 2 below shows the relations and interrelations between the various subconstructions. Such a model is, of course, an abstraction over the numerous impersonal predicates which exist in Icelandic. I believe that it is equally accurate to view the links as existing between the various low-level lexically-filled constructions, as well as between the somewhat fewer more abstract intermediate subconstructions. However, a detailed lexical network representing all the 700 predicates is bound to be too complicated and messy to be presentable here.

Eventive

Stative

Perception

Bodily States

. Changes in Bodily States

Decline Deteriorating

Cognition

Emotion

. Performance Success

Failing/Falling Mistaking

1 Attitudes

. Soc. Int.

1 Gain Fig. 2. A semantic map of the subconstructions of the impersonal construction in Icelandic

The semantics of the impersonal construction in Icelandic, German and Faroese

121

Each subconstruction of the impersonal construction is represented as a box in Figure 2. The labels in the boxes correspond to the semantic classification in (8) above. The size of the boxes corresponds to the actual size of the verb class in Icelandic. The largest class of predicates is the experience-based predicates, i.e. Emotive verbs, Cognition verbs, Perception verbs, verbs of Bodily States, and verbs of Attitudes. Together these are marked with a bold line. All other verb classes can either be regarded as related to the core class via shared lexical/aspectual content, via metaphorical extensions, or finally via pragmatic implicature, such as subjectiflcation (which I discuss in more detail in section 4 below). Therefore, not all the verb classes are extensions of the most central category but may as well be extensions from other classes. One such is the class of Existence verbs which is semantically more similar to the class of Property verbs, as both are inherently Stative, than to the central class since experience-based predicates denote transitory states. Also, verbs of general Decline are more similar to verbs denoting changes in Bodily States than to verbs of Bodily States. Not all possible links between the classes are shown in Figures 2, 3 and 4, only the most salient ones which I have discussed here. The category of the impersonal construction is thus a radial category in the sense discussed in LAKOFF (1987), i.e. it consists of a central class with other classes related to it through family resemblance, including metaphorical and inference-based extensions. That explains the uniformity in the case marking of the syntactic subject of seemingly unrelated verbs like Uka 'like' and hvolfa 'capsize' in Icelandic. These verbs are a part of a semantic/conceptual network in which the members are like family members; some members share certain properties with each other, whereas other members share other properties. All the members, however, can be shown to share properties with some member in the network. The semantic maps for German and Faroese are miniatures of the map for Icelandic, with some of the more peripheral verb classes missing altogether and fewer predicates instantiating the remaining classes. German, with its ca. 80-100 impersonal predicates (see Appendix), has all but three subconstructions of the impersonal construction, with verbs of Existence, Social Interaction and Decline missing. The semantic subclasses of the impersonal construction in German are thus the following: (13) Perception, Cognition, Idiosyncratic Attitudes, Emotion, Bodily States, Changes in Bodily States, Personal Properties, Failing/Mistaking, Success/Performance, and Gain

122

J0HANNA BARDDAL Stative

ι

Eventive

Fig. 3. A semantic map of the subconstructions of the impersonal construction in German The difference in size between Figures 2 and 3 is meant to reflect the differences in the size of the verb classes in the two languages, and thereby also the semantic scope of the construction and its different degrees of entrenchment in the language system as a whole. In Faroese, the size and the scope of the construction has been even further reduced, with only the more central semantic/conceptual regions of the construction still being covered, and only about 50-60 predicates instantiating the construction (see Appendix). This is reflected in the size of Figure 4 below. The semantic subclasses of the impersonal construction in Faroese are the following: (14) Cognition, Emotion, Bodily States, Failing/Mistaking, Success/Performance, and Gain Stative

ι

Eventive

B. States Cognition Emotion . Perform. Failing Gain Fig. 4. A semantic map of the subconstructions of the impersonal construction in Faroese A semantic map representation based on similarity makes it possible to give an account of the distribution of impersonal predicates across semantic fields, whereas approaches based on thematic roles are predestined to miss at least some of the semantic relations and regularities which I have discussed in this paper. Semantic maps also lend themselves suitable for modeling typological similarities and dif-

The semantics of the impersonal construction in Icelandic, German and Faroese

123

ferences between languages, as in this case between Icelandic, German and Faroese. A comparison of the semantic maps for all three languages reveals that the semantics of the construction in Faroese is a proper subset of the semantics of the German construction, and the semantics of the German construction is a proper subset of the semantics of the construction in Icelandic. This fact supports the traditional assumption that the construction is a common Germanic inheritance. If the impersonal construction had emerged independently in the individual languages, this complete overlap would not necessarily be expected, since the region of all possible non-agentive, or low-transitive, meanings in natural languages is much greater in semantic space than the small region occupied by the impersonal construction, and illustrated in Figures 2, 3 and 4 above. Rather, it would be expected that the construction be more evenly scattered across the non-agentive semantic space. Moreover, the fact that cognate lexical items, as, for instance, Ic. lukkast, F. lukkast and G. glücken 'succeed', are found instantiating the construction in all three languages further confirms the validity of the hypothesis that the impersonal construction is a common Germanic inheritance. There are, thus, both semantic and formal/lexical similarities between the categories of the impersonal construction across the three Germanic languages. The compelling question to ponder, at this point, is why the differences exist in the size and distribution of the category of the impersonal construction between Icelandic, on the one hand, and Faroese and German, on the other. As already mentioned in section 2,1 believe that there are two reasons for this. First, many impersonal predicates have changed into nominative subject predicates in Germanic. This change is known to have taken place in all the Germanic languages to a varying degree (ALLEN 1995, 1996, BARBDAL 1998, 2001a: 196-209, BARBDAL & EYTHORSSON 2003b: 465^68, BARNES 1986, EYTHORSSON 2000, 2002, EYTHORSSON & JÖNSSON 2003, FALK 1995, 1997, HALLDÖRSSON 1982, JONAS 2002, PETERSEN 2002, SEEFRANZ-MONTAG 1983, 1984, SMITH 1994, 1996). Together with the loss of the morphological case system the generalization of nominative as a subject case has ousted impersonal predicates in English, Dutch and Mainland Scandinavian. In both German and Faroese, nominative substitution is well documented, whereas in Icelandic it has been found only sporadically. The second reason for the difference in size of the category across Icelandic, Faroese and German is that many impersonal predicates have over time fallen into disuse. OGURA (1986), in his fairly exhaustive overview of impersonal verbs and expressions in Old and Early Middle English, counts 56 simple verbs, of which many have fallen into disuse already in Early Middle English. Similarly, FALK (1997) discusses 39 predicates in Old Swedish, of which several have disappeared from the modern language. SEEFRANZ-MONTAG (1983) also observes that many impersonal predicates have gone lost in the history of German. Icelandic, however, differs from the other Germanic languages in that the vocabulary in general has remained surprisingly stable (cf. KVARAN 1996, RÖGNVALDSSON 1997, BARBDAL 2001a: 167). In an Icelandic Word Frequency Book (PLND et al. 1991), based on a Modern Icelandic corpus of 500,000 running words, a list is provided of the 100 most frequently occurring verbs in the corpus. I have compared this list with the

124

J0HANNA BARDDAL

verbs in two separate 20,000 word corpora of written Icelandic, one containing Old Icelandic texts and the other Modern Icelandic texts (see BARBDAL 2001a). The comparison shows that out of these 100 lexemes a total of 91 occur in both the Old Icelandic and the Modern Icelandic corpora. This suggests a high degree of overlap in the vocabulary from medieval times until the present. Therefore, since the vocabulary items occurring in the impersonal construction have been maintained in Icelandic to a large degree, the construction has also had better chances of surviving (cf. BARBDAL 2001a: 197-209). Another thinkable scenario would be that the high type frequency of impersonal predicates in Icelandic is due to the construction having become productive in the history of that language, as is indeed suggested by SMIRNICKAJA (1972). However, there are certain facts that speak against such an assumption. A comparison between the two corpora, mentioned in the previous paragraph, reveals that the type frequency of the impersonal construction is substantially higher in the Old Icelandic texts than in the comparable Modern Icelandic texts. The Old Icelandic corpus contains 72 types, whereas the equivalent Modern Icelandic corpus counts only 48 types (see BARBDAL 2001a: 180). Since the corpora are well stratified, containing four genres with short text passages from ten different authors representing each genre, the chances of this difference being due to coincidence are low. On the assumption that productivity is intimately related to type frequency (BYBEE 1985, 1995, BARDDAL 2000, 2002, in preparation), this difference indicates, if anything, that the productivity of the construction has decreased in the history of Icelandic. Therefore, it does not seem reasonable to assume that the size of the category of the impersonal construction in the other Old Germanic languages represents the original state of affairs, while the construction would be taken to have expanded in the history of Icelandic. In contrast, these facts suggest that the productivity of the construction has decreased in all the Germanic languages, although at different times and different rate.

4. Subjectification One of the major problems for approaches on non-canonical case marking which take thematic roles as their point of departure is the verb class that I have labeled Performance verbs. The verbal base of these predicates refers to an "actional" event, such as speaking, acting, working, etc., and does not indicate any experience of the subject referent. In this section I will argue that the non-canonical case marking of the (logical) subject of such predicates is not motivated by the semantic relation between the referent denoted by the (logical) subject and the "event" the predicate refers to. On the contrary, I argue that the case marking should be regarded as a result of a process of subjectification, and thus motivated by the emphatic relation between the speaker and his/her attitudes towards the content of the proposition encoded in the utterance. Subjectification is a specific process involving the grammaticalization of pragmatic implicature (cf. TRAUGOTT 1989, 1995, HOPPER & TRAUGOTT 1993, TRAU-

The semantics of the impersonal construction in Icelandic, German and Faroese

125

GOTT & DASHER 2002). Semantic change involving implicature (as opposed to metaphor), such as pragmatic inferences, is understood as a process based on the communicative act of using language in context. That is, change can occur when linguistic objects are used in certain contexts because of the semantically enriching nature of contexts. In other words, a semantic structure is always mapped with a certain form, and a context repeatedly invoking a particular association can cause a new mapping between the form and the newly entailed association, resulting in polysemy, shift in lexical content or profile, and/or semantic change. One such mechanism of language change has been labeled subjectification in various publications of TRAUGOTT and others. In TRAUGOTT (1995: 31) subjectification is defined in the following way: ' [SJubjectification' refers to a pragmatic-semantic process whereby 'meanings become increasingly based on speaker's subjective belief state/attitude toward the proposition', in other words, towards what the speaker is talking about. One example of subjectification often referred to in the literature on semantic change is the development of epistemic modality from deontic modality (cf. TRAUGOTT 1989, SWEETSER 1990, and others). The English modal verb must was originally deontic since it entailed a reference to an external obligation, evident in (15a) below. However, the epistemic usage of must, which has been taken to have emerged later, entails a logical conclusion forced on the speaker by the circumstances, but does not convey any force on the referent denoted by the subject. This epistemic use of must is shown in (15b): (15a) He must be at home by midnight. (15b) He must be at home, the lights are on. TRAUGOTT (1989: 50-51) argues that this change involves the conventionalization of conversational implicature, or, in other words, pragmatic strengthening. Moreover, this kind of semantic development is clearly an example of subjectification since the epistemic meaning of must directly expresses the speaker's perspective, whereas the deontic meaning does not. Many adverbs and adjectives have developed along similar lines (HANSON 1987, TRAUGOTT 1989, PARADIS 2000). The English evidential adverb apparently was originally a manner adverb, meaning 'openly, in appearance', while in the modern language it is a sentential adverb expressing the speaker's evaluation of the sentential proposition (HANSON 1987: 139-40), i.e. a speaker-oriented adverb in the sense of JACKENDOFF (1972). Reinforcing adjectives have also typically developed from content adjectives, inducing a shift from the propositional domain to the epistemic domain (PARADIS 2000). The development of evidential adverbs from manner adverbs and reinforcers from propositional adjectives entails increased prominence of the speaker's perspective, and thus increased subjectivity. Returning to the impersonal construction and the semantics of the nonnominative (logical) subject, it is clear with predicates expressing emotions, attitudes, cognitive abilities, perception, bodily states and even changes in bodily states that the case marking of the dative can be regarded as being motivated by the semantics of these predicates. It is also clear that the meaning of affectedness, as-

126

J0HANNA BARDDAL

sociated with the (logical) subject, coincides in meaning with the meaning of the relevant predicates. Thus, it can be argued that the motivation for the case marking lies in the semantic relation holding between the referent denoted by the (logical) subject and the "event" denoted by the predicate. However, the dative case marking of (logical) subjects of Performance verbs cannot be regarded as being motivated by the semantics of this verb class or by the semantic relation holding between the subject referent and the "event" referred to by the predicate. Rather, I argue that the dative case marking is motivated by the emphatic relation holding between the speaker and his/her attitudes/judgments towards the content of the proposition encoded in the utterance (cf. BARDDAL 2001a: 158— 159, 2003b). Consider the examples in (16) below: (16a) Honum him.DAT

taladist

vel.

spoke-ii.3P.SG

well

'He gave a good speech.' 3 (16b) Leikurunum tokst vel upp actors-the.DAT took-5i.3P.SG well up 'The actors performed well on the stage.'

ά on

sviöinu. stage-the.DAT

In these sentences the referent denoted by the subject is carrying out an "actional" event which is being judged good or bad by the speaker. Therefore, the subject referent is clearly not an Experiencer, not unless we assume a veiy wide and uncon-

3

The more observant reader will notice that the predicates talast 'speak (well/badly)' and takast upp 'perform (well/badly)' are both so-called it-verbs. Therefore, a note on this category of verbs is required: The verbal suffix -st can be attached to both verbal and nominal bases in Icelandic. It has its historical origins in the reflexive pronoun sik 'self, but is by no means restricted to the medio-passive diathesis or middle verbs in Modem Icelandic. OTTÖSSON (1986, 1992) shows that the -st suffix, in both Old Norse-Icelandic and Modern Icelandic, can only be regarded as an inflectional suffix with verbs which stand in a paradigmatic opposition to active and passive verbs. In all other cases, -st is a derivational suffix found with several verb classes which are, more or less, lexically defined. There is no systematic correspondence between dative subject verbs in Icelandic which have -st attached to their base and their hypothetical non-sf-counterparts, since many of these verbs do not have a non-rf-counterpart and for those that have, the variation in meaning is not systematic. Thus, the si-suffix of dative subject verbs in Icelandic is a derivational, but not an inflectional, suffix. Passive formation, on the other hand, is periphrastic in Icelandic, involving the auxiliary vera 'be' and a participle (cf. ZAENEN et al. 1985, SIGURBSSON 1989, BARDDAL & MOLNÄR 2000, among others).

Moreover, ANDERSON (1990) shows, in an overview of the syntactic and semantic properties of it-verbs in Modern Icelandic, that these are not a uniform verb class with respect to either semantic classification or to syntactic complementation. Icelandic Ji-verbs are generally relatively low on the transitivity scale both semantically and syntactically. They are often intransitive, they can select for prepositional objects, subordinate clauses, raising or control infinitives, reflexive objects, and accusative, dative or genitive objects. ANDERSON, therefore, argues that no specific overall semantic or functional content can be attributed to the it-suffix, even though certain lower-level subclass generalizations can be made. Clearly, it is only on a monosemy account that one would expect only one specific meaning to be attributable to the si-suffix. On a polysemy account, on the other hand, it is expected that -st has multiple functions. As a native speaker of Icelandic, I concur with ANDERSON that it is not obvious which semantic or functional content should be assigned to the ii-suffix at a psycholinguistic level. Therefore, I will refrain from glossing the si-suffix as anything but -st in the remainder of this paper.

The semantics of the impersonal construction in Icelandic, German and Faroese

127

strained definition of the Experiencer role, completely independent of the meaning of the predicate or the construction. OTT0SSON ( 1 9 8 6 : 1 0 8 , fh. 6 7 ) takes the dative subject of verbs like talast (and, by implication, takast upp) to be an Agent (presumably on the basis of the lexical content of the verbal stem), since giving a speech and performing in general is eventive, and not emotive, in nature. J0NSSON, however, argues against OTTÖSSON's analysis providing examples like (17) below, which show that subjectoriented adverbs (again in the sense of J A C K E N D O F F 1 9 7 2 ) , expressing purpose or intention on behalf of the referent denoted by the subject, cannot occur with Performance verbs, as would be expected if the subject is an Agent (JÖNSSON 1997— 98: 23):

(17)

*Honum

talaöist vel, fiislega. spoke-j/.3P.SG well willingly Intended meaning: 'He gave a good speech, willingly.'

him.DAT

J0NSSON, thus, concludes that the dative subject cannot be an Agent and must be an Experiencer. There are three problems with JÖNSSON's analysis: First, these predicates do not profile the experience, emotions, or any inner processes of the subject referent, which is a presupposition for analyzing the subject referent is an Experiencer. In fact, the subject referent need not be affected at all by the "event" denoted by the predicate. Secondly, on JÖNSSON's approach the role of the Experiencer has become a dustbin category not independently defined, since the experiencer analysis he proposes is based on lack of evidence for agentive behavior. Thirdly, subjectoriental adverbs cannot always occur with tala 'speak' either, a nominative subject selecting verb cognate to talast. (18)

?Hann taladi vel, fiislega. he.NOM spoke.3P.SG well, willingly Intended meaning: 'He gave a good speech, willingly.'

On JÖNSSON's account, examples like (18) should be perfectly grammatical since the nominative subject of tala is an Agent in his analysis ( 1 9 9 7 - 9 8 : 2 3 ) . It is, of course, true that for predicates which select a dative-experiencer subject, sentences like (17)-(18) above are expected to be ungrammatical, since normally agent-oriented adverbs are not controlled by experiencer subjects.4 However, examples like (17) above need not be infelicitous because of that; their unacceptability may have other reasons. I believe that (17) is bad because the whole proposition encoded in sentences containing Performance predicates of this type has the function of conveying the attitude or judgment of the speaker and not that of the subject referent. On such an account concepts like Agent and Experiencer are both inapplicable and irrelevant to the analysis of these sentences, and hence also to the case assignment of Performance verbs. To compare, consider the following examples:

4

ERNST (2002: 63), however, argues that subject-oriented adverbs can be divided into agentoriented and mental-attitude adverbs, and that the latter subclass can co-occur with experiencer subjects. In fact, he classifies willingly as a mental-attitude adverb.

128 (19a) (19b) (19c) (19d)

J0HANNA BARDDAL

John looks good. John looks good on purpose. *John looks. *John looks on purpose.

Example (19a) expresses the speaker's claim that the subject referent, John, is looking good. When subject-oriented adverbs, like on purpose, are added to the proposition, as in (19b), they take scope over the evaluative subjective complement good, giving the impression that it is the subject referent who judges his own looks as good. In other words, it is at least the subject referent, and perhaps also the speaker of (19b), who evaluates the subject referent's looks as good, whereas in (19a) the evaluation is solely that of the speaker. The verb look must occur together with an evaluative subjective complement of the type good in order to maintain its meaning and its function as a scalar predicate, as evident from a comparison between (19a-b) and (19c-d). The appearance of 'looking good' is a matter of judgment and degree, as evidenced by the fact that the adjective good is inherently scalar (in the sense of PARADIS 2001, 2003). In other words, the evaluative subjective complement is valence bound with predicates referring to scalar "events" of the type in (19a-b). Moreover, as argued by PARADIS (2000), expressions of scalar meanings are always subjective, since they reflect the speaker's stance. Thus, the propositions in (19a-b) are epistemic in nature, conveying the judgment of the, subject referent, the speaker or both. Now, Icelandic is no different from English with respect to evaluative scalar predicates of this type: (20a) Jon John.NOM

litur

'John looks good.' (20b) Jon litur John.NOM

vel

looks.3P.SG well

vel

looks.3P.SG well

ύί. out

ύt

af

äsettu

rädi.

out

of

intentional.DAT

plan.DAT

af

äsettu

rädi.

of

intentional.DAT

plan.DAT

'John looks good on purpose.' (20c) *Jon litur ύί. John.NOM looks.3P.SG out

(20d) *Jon

litur

ύί

John.NOM looks.3P.SG out

The Icelandic examples in (20) are exact counterparts of the English examples in (19), in that without the evaluative subjective complement vel 'well' the predicate liia (vel/illa) ύί 'look (good/bad)' changes its meaning to 'look (out)' and cannot be used to express the judgment of the speaker that the subject referent is looking good, as in (20a) or the judgment of presumably both the speaker and the subject referent, as in (20b). The grammaticality of examples like (19b) and (20b) above shows that with evaluative scalar predicates, subjective complements, like good and vel, and subject(agent)-oriented adverbs, like on purpose and af äsettu rädi can co-occur in the same sentence in both English and Icelandic. When they do, the subject(agent)oriented adverb takes scope over the evaluative subjective complement, enforcing a reading in which it is the subject referent who judges his/her own looks as good. Thus, the stance of the sentence changes from being more subjective to being more

The semantics of the impersonal construction in Icelandic, German and Faroese

129

factual (in the case where the judgment is the subject referent) or more semi-factual (in the case where the judgment can be regarded to be both that of the subject referent and the speaker). Therefore, (19a) and (20a) are subjective, whereas (19b) and (20b) are (semi-)factual. Moreover, if it is a general rule in both English and Icelandic that subjectoriented adverbs take scope over evaluative subjective complements when these two co-occur in the same sentence, it is expected that subject-oriented adverbs will also occur with scalar predicates like talast 'speak (well/badly)' and takast upp 'perform (well/badly)' in Icelandic, yielding a reading in which the judgment expressed is that of the subject referent, exactly as in (20b). Nevertheless, as originally shown by J0NSSON and discussed above, such sentences are ungrammatical in Icelandic. I repeat the relevant example for convenience: (17)

*Honum taladist vel, föslega. him.DAT spoke-5i.3P.SG well willingly Intended meaning: 'He gave a good speech, willingly.'

This is unexpected given the felicitousness of (20b) in Icelandic. However, if there is a grammatical restriction on predicates like talast, such that these predicates are predisposed to only convey the attitude or the judgment of the speaker, then obviously they will be incompatible with a construal in which the judgment conveyed is that of the subject referent. Thus, the ungrammaticality of (17) falls naturally from the pragmatic function of the construction, given the present analysis. Observe that the scalar predicates talast 'speak (well/badly)' and takast upp 'perform (well/badly)' are equivalent to 'look (good)' in that the evaluative complement vel 'well' is valence triggered, and thus obligatory in sentences containing talast and takast upp: (21a) *Honum taladist. him.DAT spoke-5i.3P.SG

(21b) *Leikurunum

tokst

actors-the.DAT took-s/.3P.SG

upp

ά

sviöinu.

up

on

stage-the.ACC

For the nominative subject verbs tala 'speak' and taka upp 'take up', however, evaluative adverbs of this type are not obligatory at all: (22a) Hann he.NOM

taladi. spoke.3P.SG

'He talked.' or: 'He spoke.' (22b) Hann tok upp. they.NOM

took.3P.SG

up

'He did the recording.' This shows that the lexical verbs tala and taka upp are not inherently scalar. The dative subject verb talast 'speak (well/badly)' also differs from the nonsuffixed nominative verb tala 'speak' in Icelandic in that the latter can occur with either evaluative adverbs or subject-oriented adverbs, while the former can only occur with evaluative manner adverbs. Consider the following examples:

130

J0HANNA BARDDAL

(23a) Hann

talaöi

he.NOM spoke.3P.SG

vel. well

'He spoke well.' or: 'He gave a good speech.' (23b) Hann talaöi föslega. he.NOM spoke.3P.SG

willingly

'He was eager to speak.' (24a) Honum him.DAT

talaöist

vel.

spoke-5i.3P.SG

well

'He gave a good speech.' (24b) *Honum talaöist him.DAT

fuslega.

spoke-s/.3P.SG

willingly

The examples in (23) show that tala 'speak' can occur with either subjectoriented adverbs (23b) or evaluative manner adverbs (23a), while the examples in (24) show that talast 'speak (well/badly)' can only occur with evaluative manner adverbs. Subject-oriented adverbs are simply incompatible with talast irrespective of whether the evaluative manner adverb is present (17) or not (24b). The scalar predicate talast 'speak (well/badly)' is thus similar to 'look (good)' in that the evaluative manner adverb is valence bound. Thus, it is an evaluative subjective complement selected by the verb. In contrast, the predicate talast 'speak (well/badly)' differs from 'look (good)' in that it cannot co-occur with subject-oriented adverbs. The verb tala 'speak', however, has no such restrictions. I believe that the scalar predicate talast 'speak (well/badly)' cannot co-occur with subject-oriented adverbs because the speaker's stance has been grammaticalized with such predicates. Consider the two examples below: (25a) Honum him.DAT

liöur

vel.

feels.3P.SG

well

'He feels good.' (25b) Honum talast him.DAT

speaks.3P.SG

vel.

(semi-)factual reading subjective reading

well

'He speaks well.' The evaluative subjective complement vel 'well' with Πόα 'feel' in (25a) and talast 'speak' in (25b) shows that the "events" denoted by these predicates are inherently scalar. Both sentences reflect the speaker's stance, since it is, of course, the speaker who evaluates the emotional state of the subject referent in (25a) and the quality of the speech made by the subject referent in (25b). Therein lies also the motivation for the case marking of these different subclasses of impersonal predicates, i.e. in the emotional state of the subject referent in (25a) and the evaluation of the quality of the speech made by the subject referent in (25b). The verb liöa 'feel' is a verb of Emotion, whereas talast 'speak (well/badly)' is not. The dative case of the syntactic subject of liöa is motivated by the semantic relation holding between the subject referent and the experience-based predicate, since it is the subject referent who is the Experiencer. On the other hand, the dative case of the syntactic subject of talast is motivated by the emphatic relation between the speaker and the content of the proposition, since for this verb it is the speaker who is the

The semantics of the impersonal construction in Icelandic, German and Faroese

131

Experience!*, and not the subject referent. It is the speaker who experiences and judges the quality of the speaking/performing event carried out by the subject referent. The subject referent may or may not have opinions about his/her own performance. Moreover, an evaluation of the quality of the performance, which could be made by the subject referent him/herself, is completely irrelevant for the grammatical explication of the proposition. S/he may evaluate his/her performance quite differently from the speaker, if s/he has an opinion at all. Therefore, the occurrence of dative (logical) subjects with Performance verbs shows that the function of the dative argument has been extended from being subject-oriented, i.e. expressing the judgment of the subject referent, to becoming speaker-oriented, i.e. expressing the judgment of the speaker. To conclude, on the present approach the ungrammaticality of examples like (17) above can be accounted for in a straight-forward fashion; subject-oriented adverbs are semantically/pragmatically, hence also syntactically, incompatible with Performance verbs of the impersonal class in Icelandic, because with Performance verbs the speaker's perspective has been grammaticalized. This subtype of the impersonal constructions has the function of conveying the speaker's judgment and is thus incompatible with a (semi-)factual construal; hence the restrictions on the occurrence of subject-oriented adverbs with this verb class and the obligatoriness of evaluative subjective complements. The present analysis makes the following prediction: Non-canonical case marking of experience-based predicates should be historically prior to non-canonical case marking of Performance verbs, since a process of subjectification entails a change in meaning from factual to subjective. This is not to say that the derived usage cannot arise within the same generation of speakers, and thus be synchronically reflected in language use, it only predicts that the "speaker-oriented (subjective) dative" should not emerge unless the "subject-oriented (experiencer) dative" exists in the language. This prediction is difficult to test in Germanic since Performance verbs like those above are attested with dative (logical) subjects since the earliest documented history of Germanic. Thus, verbs with the meaning 'succeed, do well' exist, to my knowledge, at least in Old Norse-Icelandic, Old Swedish, Old Danish, Old High German and Old English. Since non-canonical case marking of logical subjects of Performance predicates, which express the evaluation of the speaker, is also found, for instance, in Polish (cf. BARBDAL 2003b, based on examples from DABROWSKA 1 9 9 7 : 4 4 - 4 5 ) , this particular subtype of the impersonal construction is presumably an Indo-European inheritance. In a typological perspective, however, it follows that no language should only show non-canonical case marking on the (logical) subject of Performance verbs like 'succeed, do well', and not on the corresponding argument of ordinary experience-based predicates. As far as I know, no such languages exist (cf. the discussion in ONISHI2001).

132

J0HANNA BARDDAL

5. Summary In this paper, I have unfolded two major problems with thematic roles as explanatory concepts in accounting for the non-canonical case marking of the (logical) subject of impersonal predicates in Icelandic, German and Faroese. First, approaches based on thematic roles do not predict that there should be any similarity found between the various verb classes, and, hence, the subconstructions, contra what I have shown in this paper. Second, thematic roles cannot account for the non-canonical case marking of the (logical) subject of Performance predicates, since the case assignment of such predicates is not motivated by the strict lexical semantic content of this particular verb class, and thus not by the semantic relation holding between the subject referent and the "event" expressed by the predicate. Instead, the non-canonical case marking is motivated by the emphatic relation holding between the speaker and his/her attitudes towards the content of the proposition encoded in the utterance. I have put forth a verbal-semantic analysis which holds for the approximately 700 dative selecting impersonal predicates in Icelandic, 80-100 impersonal predicates in German and the 50-60 predicates which can still be used impersonally in Faroese. This investigation has confirmed that the core category of impersonal predicates is experience-based, including Emotive verbs, verbs of Attitude, Cognition verbs, Perception verbs and verbs of (changes in) Bodily States. Other verb classes are related to the central class via semantic/aspectual similarity, metaphorical extensions and extensions based on pragmatic inferences. The similarity relations found between the verb classes are those of family resemblance. The distribution of impersonal predicates across semantic space can be accurately accounted for by conceptual networks and theories of grammar which take lexical, grammatical and constructional semantics as their point of departure. Abbreviations 3P ACC DAT

third person accusative dative

EXPL

expletive

NOM PL SG

nominative plural singular

The semantics of the impersonal construction in Icelandic, German and Faroese

133

References (eds.) ( 2 0 0 1 ) : Non-canonical marking of subjects and objects. Amsterdam: John Benjamins. ALLEN, CYNTHIA L. ( 1 9 9 5 ) : Case marking and reanalysis: Grammatical relations from Old to Modern English. Oxford: Oxford University Press. ALLEN, CYNTHIA L . ( 1 9 9 6 ) : A change in structural case marking in Early Middle English, in: THRÄINSSON, HÖSKULDUR; EPSTEIN, SAMUEL D. & PETER, STEVE (eds.), Studies in comparative Germanic syntax II. Dordrecht: Kluwer, 3-20. ANDERSON, STEPHEN R. (1990): The grammar of Icelandic verbs in -st, in: MALING & ZAENEN (eds.), 235-273. ANDREWS, AVERY D. (1976): The VP complement analysis in Modern Icelandic, in: Proceedings of the North East Linguistic Society 6, 1-21. [Reprinted 1990 (lightly revised) in: MALING & ZAENEN (eds.), 165-185]. BARBDAL, J0HANNA (1998): Argument structure, syntactic structure and morphological case of the impersonal construction in the history of Scandinavian, in Scripta Islandica 49,21-33. BARBDAL, J0HANNA (2000): Case assignment of nonce verbs in Icelandic, in: SKY Journal of Linguistics 13, 7-27. BARBDAL, JÖHANNA (2001A): Case in Icelandic - A synchronic, diachronic and comparative approach. Doctoral dissertation (Lundastudier i Nordisk spräkvetenskap A 57). Lund: Department of Scandinavian Languages. BARBDAL, JÖHANNA (2001b): The perplexity of DAT-NOM verbs in Icelandic, in: Nordic Journal of Linguistics 24,47-70. BARBDAL, JÖHANNA ( 2 0 0 2 ) : "Oblique subjects" in Icelandic and German, in: Working Papers in Scandinavian Syntax 7 0 , 6 1 - 9 9 . BARBDAL, JÖHANNA (2003a): The relation between morphological case, syntactic functions and thematic roles: A frequency analysis, in: HETLAND & MOLNAR (eds.), 149-187. BARBDAL, JÖHANNA (2003b): A review of EWA DABROWSKA: Cognitive semantics and the Polish dative, in: Cognitive Linguistics 14, 379-386. BARBDAL, JÖHANNA (in preparation,).' Productivity: Evidence from case and argument structure in Icelandic, to appear in the Constructional Approaches to Language Series. Amsterdam: John Benjamins. BARBDAL, JÖHANNA & EYTHÖRSSON, THÖRHALLUR (2003a): Icelandic vs. German: Oblique subjects, agreement and expletives, in: Chicago Linguistic Society 39(1), 17-29. BARBDAL, JÖHANNA & EYTHÖRSSON, THÖRHALLUR (2003b): The change that never happened: The stoiy of oblique subjects, in: Journal of Linguistics 39, 439-472. BARBDAL, JÖHANNA & EYTHÖRSSON, THÖRHALLUR (2004): Control infinitives and case in Germanic: 'Performance error' or marginally acceptable constructions, to appear in: KULKOV, L.; MALCHUKOV, A. & SWART, P. DE (eds.), Case, valency and transitivity. Amsterdam: John Benjamins. BARBDAL, JÖHANNA & MOLNÄR, VAL£RIA ( 2 0 0 0 ) : Passive in Icelandic - Compared to Mainland Scandinavian, in: Working Papers in Scandinavian Syntax 6 5 , 1 0 9 - 1 4 6 [also published 2 0 0 3 (lightly revised) in: HETLAND & MOLNÄR (eds.), 2 3 3 - 2 6 2 ] , BARNES, MICHAEL P . ( 1 9 8 6 ) : Subject, nominative and oblique case in Faroese, in: Scripta Islandica AIKHENVALD, ALEXANDRA Y . ; DIXON, R . M . W . & ONISHI, MASAYUKI

37, 13-46. BERNÖDUSSON, HELGI

(1982): Opersonulegar setningar [Impersonal sentences]. Master's thesis, University of Iceland. BYBEE, JOAN L. (1985): Morphology: A study of the relation between meaning andform. Amsterdam: John Benjamins. BYBEE, JOAN ( 1 9 9 5 ) : Regular morphology and the lexicon, in: Language and Cognitive Processes 10 (5), 4 2 5 - 4 5 5 . BRAINE, MARTIN D . S.; BRODY, RUTH E.; FISH, SHALOM M . ; WEISBERGER, MARA J. & BLUM, MONICA

(1990): Can children use a verb without exposure to its argument structure?, in: Journal of Child Language 17, 313-342. CROFT, WILLIAM ( 1 9 9 3 ) : Case marking and the semantics of mental verbs, in: PUSTEJOVSKY, JAMES (ed.), Semantics and the lexicon. Dordrecht: Kluwer Academic, 55-72.

134

J0HANNA BARDDAL

Event structure in argument linking, in: BUTT, MIRIAM & GEUDER, WILHELM (eds.) The projection of arguments: Lexical and compositional factors. Stanford: C S L I Publications, 1-43. CROFT, WILLIAM ( 2 0 0 1 ) : Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press. CROFT, WILLIAM ( 2 0 0 3 ) : Lexical rules vs. constructions: A false dichotomy, in: CUYCKENS, HUBERT; BERG, THOMAS; DIRVEN, REN6 & PANTHER, KLAUS-UWE (eds.), Motivation in language: Studies in honour of GÜNTER RADDEN. Amsterdam: John Benjamins, 49-68. CROFT, WILLIAM (in preparation): Verbs: Aspect and argument linking. Oxford: Oxford University Press. CROFT, WILLIAM & CRUSE, D. ALAN (2004): Cognitive linguistics. Cambridge: Cambridge University Press. DABROWSKA, EWA ( 1 9 9 7 ) : Cognitive semantics and the Polish dative. Berlin: Mouton de Gruyter. ERNST, THOMAS ( 2 0 0 2 ) : The syntax of adjuncts. Cambridge: Cambridge University Press. EYTHÖRSSON, THÖRHALLUR ( 2 0 0 0 ) : Dative vs. nominative: Changes in quirky subjects in Icelandic, in: Leeds Working Papers in Linguistics 8 , 2 7 - 4 4 . EYTHÖRSSON, THÖRHALLUR ( 2 0 0 2 ) : Changes in subject case marking in Icelandic, in: LIGHTFOOT, DAVID (ed.), Syntactic effects of morphological change. Oxford: Oxford University Press, 1 9 6 - 2 1 2 . EYTHÖRSSON, THÖRHALLUR & BARBDAL, JÖHANNA ( 2 0 0 3 ) : Oblique subjects: A Germanic inheritance!, in: Working Papers in Scandinavian Syntax 7 1 , 1 4 5 - 2 0 1 . EYTHÖRSSON, THÖRHALLUR & JÖNSSON, JÖHANNES G . ( 2 0 0 3 ) : The case of subjects in Faroese, in: Working Papers in Scandinavian Syntax 7 2 , 2 0 7 - 2 3 1 . FALK, CECILIA ( 1 9 9 5 ) : Lexikalt kasus i svenska [Lexical case in Swedish], in: Arkiv for nordisk filologi 1 1 0 , 1 9 9 - 2 2 6 . FALK, CECILIA (1997): Fornsvenska upplevarverb [Old Swedish experiencer verbs]. Lund: Lund University Press. FILIP, HANA ( 1 9 9 6 ) : Psychological predicates and the syntax-semantics interface, in: GOLDBERG, ADELE (ed.), Conceptual structure, discourse and language. Stanford: CSLI Publications, 1 3 1 - 1 4 7 . GIBBS, RAYMOND W, JR. (1997): How language reflects the embodied nature of creative cognition, in: WARD, THOMAS B . ; SMITH, STEVEN M . & VAID, JOYTSNA (eds.), Creative thought: An investigation of conceptual structures and processes. Washington, DC: APA Publications, 351-373. GOLDBERG, ADELE Ε . ( 1 9 9 5 ) : Constructions: A construction grammar approach to argument structure. Chicago & London: University of Chicago Press. HALLDÖRSSON, HALLDÖR (1982): Um m£ranir [On dative sickness], in: Islenskt mal og almenn malfrcedi 4, 159-189. HANSON, KRISTIN ( 1 9 8 7 ) : On subjectivity and the history of epistemic expressions in English, in: Chicago Linguistic Society 2 3 , 1 3 2 - 1 4 7 . HETLAND, JORUNN & MOLNAR, VALIRIA (eds.) (2003): Structures of focus and grammatical relations. (Linguistische Arbeiten 477). Tübingen: Niemeyer. HOPPER, PAUL J. & TRAUGOTT, ELIZABETH CLOSS (1993): Grammaticalization. Cambridge: Cambridge University Press. JACKENDOFF, RAY (1972): Semantic interpretation in generative grammar. Cambridge, Mass.: ΜΓΓ Press. JACKENDOFF, RAY ( 1 9 9 0 ) : Semantic structures. Cambridge, Mass.: MIT Press. JONAS, DIANNE ( 2 0 0 2 ) : Recent change in Faroese experiencer constructions. Presented at the LSA 2 0 0 2 Annual Meeting, San Francisco, 3-6 January 2 0 0 2 . JÖNSSON, JÖHANNES GISLI (1996): Clausal architecture and case in Icelandic. Doctoral dissertation. Amherst: University of Massachusetts. JÖNSSON, JÖHANNES GISLI ( 1 9 9 7 - 9 8 ) : Sagnir meö aukafallsfrumlagi [Verbs selecting for oblique subjects], in: Islenskt mal og almenn mälfrceöi 1 9 - 2 0 , 1 1 - 4 3 . JÖNSSON, JÖHANNES GISLI (1998): A list of predicates that take a quiiky subject in Icelandic. Ms. Reykjavik: University of Iceland. KVARAN, GUBRÜN ( 1 9 9 6 ) : E>aettir ΎΓ sögu oröaforöans [On the history of the Icelandic vocabulary], in: Erindi um islenskt mäl. Reykjavik: Islenska mdlfraeöifdlagiö, 35-48. LAKOFF, GEORG ( 1 9 8 7 ) : Women, fire, and dangerous things: What categories reveal about the mind. Chicago: Chicago University Press. LAKOFF, GEORG & JOHNSON, MARK (1980): Metaphors we live by. Chicago: Chicago University Press. CROFT, WILLIAM ( 1 9 9 8 ) :

The semantics of the impersonal construction in Icelandic, German and Faroese

135

Philosophy in the flesh: The embodied mind and its challenge to Western thought. New York: Basic Books. LANGACKER, RONALD W . ( 1 9 8 8 ) : A usage-based model, in: RUDZKA-OSTYN, BRYGIDA (ed.), Topics in cognitive linguistics. Amsterdam/Philadelphia: John Benjamins, 1 2 7 - 1 6 1 . LANGACKER, RONALD W . ( 1 9 9 1 ) : Foundations of cognitive grammar II: Descriptive application. Stanford: Stanford University Press. LAUBSTEIN, ANN STUART ( 1 9 9 9 ) : Word blends as sublexical substitutions, in: Canadian Journal of Linguistics 4 4 ( 2 ) , 1 2 7 - 1 4 8 . MALING, JOAN ( 2 0 0 2 ) : Icelandic verbs with dative objects, in: Working Papers in Scandinavian SynLAKOFF, GEORG & JOHNSON, MARK ( 1 9 9 9 ) :

tax 7 0 , 1 - 6 0 . MALING, JOAN & ANNIE ZAENEN

(eds.) (1990): Modern Icelandic syntax. (Syntax and Semantics 24). San Diego: Academic Press. OGURA, MICHIKO ( 1 9 8 6 ) : Old English 'impersonal' verbs and expressions. Copenhagen: Rosenkilde & Bagger. ONISHI, MASAYUKI (2001): Introduction: Non-canonically marked subjects and objects: Parameters and properties, in: AKHENVALD, ALEXANDRA Y.; DIXON, R . M . W. & ONISHI, MASAYUKI (eds.), 1-51. OTTÖSSON, KJARTAN G. (1986): Mörk orömyndunar og beygingar: Miömynd Ί nütimaislensku [The delimitation of word formation and inflection: The medio-passive in Modern Icelandic], in: Islenskt mil og almenn mälfrceöi 8, 63-119. OTTÖSSON, KJARTAN G . ( 1 9 9 2 ) : The middle voice in Icelandic. Doctoral Dissertation, Lund University. PARADIS, CARITA (2000): Reinforcing adjectives: A cognitive semantic perspective on grammaticalization, in: BERMÜDEZ-OTERO, RICARDO; DENISON, DAVID; HOGG, RICHARD M . & MCCULLY, CHRIS Β. (eds.), Generative theory and corpus studies. Berlin: Mouton de Gruyter, 233-258. PARADIS, CARITA ( 2 0 0 1 ) : Adjectives and boundedness, in: Cognitive Linguistics 12 ( 1 ) , 4 7 - 6 5 . PARADIS, CARITA ( 2 0 0 3 ) : Between epistemic modality and degree: The case of really, in: FACCHINETTI, ROBERTA; PALMER, FRANK & KRUG, MANFRED (eds.), Modality in contemporary English. Berlin: Mouton de Gruyter, 1 9 7 - 2 2 0 . PETERSEN, HJALMAR P . ( 2 0 0 2 ) : Quirky case in Faroese, in: Froöskaparrit 5 0 , 6 3 - 7 6 . PIND, JÖRGEN; MAGNÜSSON, FRIDRK & BRIEM, STEFAN ( 1 9 9 1 ) : Islensk ordtidnibok [Icelandic Word Frequency Book]. [Reykjavik]: Oröabök Häskölans. ROELOFS, ARDI ( 1 9 9 2 ) : A spreading-activation theory of lemma retrieval in speaking, in: Cognition 42,107-142. RÖGNVALDSSON, EIRIKUR

(1983): Mgufallssykin og fallakerfi ί islensku [Dative substitution and the Icelandic case system], in: S/dma 16, 3-6. RÖGNVALDSSON, EIRIKUR ( 1 9 9 7 ) : Oröafar Islendinga sagna [The vocabulary of the Sagas of Icelanders], in: AGNARSDÖTTIR, ANNA; PETURSSON, P£TUR & TULINIUS, TORFI H. (eds.), Milli himins og jardar. Reykjavik: Haskölaütgäfan, 2 7 1 - 2 8 6 . SEEFRANZ-MONTAG, ARIANE VON ( 1 9 8 3 ) : Syntaktische Funktionen und Wortstellungsveränderung: Die Entwicklung „subjektloser" Konstruktionen in einigen Sprachen. Munich: Wilhelm Fink. SEEFRANZ-MONTAG, ARIANE VON ( 1 9 8 4 ) : 'Subjectless' constructions and syntactic change, in: FISIAK, JACEK (ed.), Historical syntax. Berlin: Mouton, 5 2 1 - 5 5 3 . SIGURDSSON, HALLDÖR ÄRMANN ( 1 9 8 9 ) : Verbal syntax and case in Icelandic. Doctoral dissertation, Lund University [reprinted 1992 by the Linguistics Institute, University of Iceland]. SMIRNICKAJA, OLGA ( 1 9 7 2 ) : The impersonal sentence patterns in the Edda and in the Sagas, in: Arkiv fir nordiskfilologi 87, 5 6 - 8 8 . SMITH, HENRY ( 1 9 9 4 ) : Dative sickness in Germanic, in: Natural Language and Linguistic Theory 12, 675-736. SMITH, HENRY (1996): Restrictiveness in case theory. Cambridge: Cambridge University Press. SMITH, MICHAEL B. ( 2 0 0 1 ) : Why quirky case really isn't quiiky (or how to treat dative sickness in Icelandic), in: CUYCKENS, HUBERT & ZAWADA, BRITTA (eds.), Polysemy in cognitive linguistics. Amsterdam: John Benjamins, 1 1 5 - 1 5 9 . SVAVARSDÖTTIR, ÄSTA (1982): "I>igufallssyki" ["Dative sickness"], in: islenskt mäl og almenn mälfrcedi 4,19-62.

136

J0HANNA BARDDAL

Fall er fararheill: Um fallnotkun med öpersönulegum sögnum [Fall is a sign of luck: On case use with impersonal verbs], in: Islenskt mäl ogalmenn mälfrcedi 6 , 3 3 - 5 5 . SWEETSER, EVE E . ( 1 9 9 0 ) : From etymology to pragmatics. Cambridge: Cambridge University Press. TRAUGOTT, ELIZABETH CLOSS ( 1 9 8 9 ) : On the rise of epistemic meanings in English: An example of subjectification in semantic change, in: Language 5 7 , 3 3 - 6 5 . TRAUGOTT, ELIZABETH CLOSS ( 1 9 9 5 ) : Subjectification in grammaticalization, in: STEIN, DIETER & WRIGHT, SUSAN (eds.), Subjectivity and subjectification. Cambridge: CUP, 3 1 - 5 4 . TRAUGOTT, ELIZABETH CLOSS & DASHER, RICHARD B . ( 2 0 0 2 ) : Regularity in semantic change. Cambridge: Cambridge University Press. VIBERG, AKE ( 1 9 8 3 ) : The verbs of perception: A typological study, in: Linguistics 2 1 (1), 1 2 3 - 1 6 2 . WEGENER, HEIDE ( 2 0 0 1 ) : Verbs of affect from a synchronic and a diachronic perspective, in: ϋΕΗέ, NICOLE & WANNER, ANJA (eds.), Structural aspects of semantically complex verbs. Frankfurt: Peter Lang, 2 1 9 - 2 4 8 . ZAENEN, ANNIE; MALING, JOAN & THRÄINSSON, HÖSKULDUR ( 1 9 8 5 ) : Case and grammatical functions: The Icelandic passive, in: Natural Language and Linguistic Theory 3, 4 4 1 - 4 8 3 [Reprinted 1 9 9 0 in: MALING & ZAENEN (eds.), 9 5 - 1 3 6 ] . SVAVARSDÖTTIR, ÄSTA; PÄLSSON, GISLI & I>0RLINDSSON, I>0R0LFUR ( 1 9 8 4 ) :

Appendix: Impersonal predicates in Faroese and German5 Impersonal predicates in Faroese (PETERSEN 2002/EYTH0RSSON & J0NSSON 2003): berast ά 'slip', berast frd 'get a hindrance', berastfyri 'seem', bjööa ί lätur 'burst out laughing', bresta 'burst', däma 'like', detta i hug 'get an idea', vera dätt νίό 'be surprised at', eydnast 'succeed', falla 'like>,falla ί lut ' g e t \ f i n n a s t 'find, fee\',Jysa 'want', gangast 'do well/badly', gremja 'grieve', gruna 'suspect', henda 'happen to', hova 'dislike', huga 'like', hugbita 'want', hugna/hugnast 'like', komafyri 'seem', koma i hug 'get an idea', leiöast viö 'be bored', leingjast 'feel impatience', leingast eftir 'long for', liöa 'make progress', lika 'like', likast 'get', litast ά 'like', lukkast 'succeed', lutast 'get', lysta 'want', mangla 'lack', mislika 'dislike', nita 'shudder at', nytta 'be of use', nytast 'need', mtra 'shudder at', harma 'regret', vanta 'want', droyma 'dream', minnast 'remember', ora 'suspect', renna ί hug 'get an idea', renna tydan til 'have warm feelings', skiljast 'understand', skrida 'tickle', skrima 'like', standast νίό 'detest', synast 'seem', trota 'lack', tylga/tykjast 'seem', vera innari 'have on one's mind'. Impersonal predicates in German (compiled for the purpose of this paper): Verbs: (an)liegen 'be on sb.'s mind', behagen 'please, suit', bekommen 'fit, agree with', dauern 'be sorry, regret', dürsten 'thirst', einfallen 'get an idea', einleuchten 'realize, agree', ekeln 'disgust', fehlen 'lack', frieren 'freeze', frommen 'benefit', frösteln 'freeze', gefallen 'please, like', (gut/schlecht) gehen 'go (well/badly) for sb.', gehören 'belong', gelingen 'succeed', genügen 'satisfy', geziemen 'befit', glücken 'succeed', grauen 'dread', hungern 'hunger', jammern 'feel pity for', jucken 'itch', (eine Idee) kommen 'get an idea', leichtfallen 'be easy for', leid tun 'be sorry', liegen 'be apt at', mangeln 'lack', missfallen 'displease, dislike', missglücken 'fail', misslingen 'fail', munden 'taste good', nutzen 'profit by', nützen 'be of use to', obliegen 'feel obliged', passen 'suit', passieren 'happen to', reichen 'suffice', reuen 'regret', schaudern 'dread', scheinen 'seem', schmecken 'taste', schwanen 'suspect', schwerfallen 'be a burden', schwindeln 'feel dizzy', träumen 'dream', verdrießen 'be sorry about sth.', eilen 'be in a hurry', (vorkommen 'appear, strike as', (vor)schweben 'have in mind', wurmen 'worry', weh tun 'hurt', wundern 'wonder'. Complex predicates with sein 'be': angeboren 'be innate', angst 'be scared', bange 'be scared', bekannt 'be familiar', egal 'be indifferent', eigen 'be typical', einerlei 'be indifferent', geläufig 'be familiar', gemeinsam 'have in common', gleichgültig 'be indifferent', gram 'be sorry/angry', gut 'feel good', heiß 'be warm', kalt 'be cold', komisch 'feel funny', lieb 'like, be glad', recht 'be OK for sb.', schlecht 'feel sick', schwindlig 'feel dizzy', seltsam zu Mute 'feel odd', übel 'feel nauseated', Untertan 'be subservient', unwohl 'feel sick', warm 'be warm', widerlich 'be disgusted', willkommen 'welcome, glad', wohl ums Herz 'feel light at heart', zugänglich 'be clear', zum Heulen 'be painful', zuträglich 'be beneficial', zuwider 'be unpleasant/disgusted'. Recent additions to the group (WEGENER 2001): intressieren 'interest', nerven 'be annoyed', kratzen 'itch', stinken 'dislike', schnuppe sein 'be indifferent', (um)hauen 'be startled', wurst sein 'be indifferent'. 5

My goal, when compiling these lists, has been to represent the linguistic knowledge of a modem speaker. I have therefore tried to include predicates which are still recognized by native speakers as occuning with nonnominative (logical) subjects, but excluded predicates which are obsolete in such usage. However, drawing a distinction between the two is not an easy task. The list may therefore include predicates which cannot occur with non-nominative (logical) subjects for some speakers, and exclude predicates which do occur with such non-canonically case-marked arguments in the language of other speakers. Hopefully, though, the vast majority of the verb class is represented.

Left dislocation in Germanic CEDRIC BOECKX (Harvard) & KLEANTHES K . GROHMANN (Cyprus)

1. The phenomenon of left dislocation While the well-known phenomenon of clitic left dislocation has received considerable attention in the generative literature (see among many others CINQUE 1977, 1990: ch. 2 for Italian and AOUN & BENMAMOUN 1998 for Lebanese Ara-

bic), non-clitic, contrastive left dislocation of the type as it appears in Germanic has not enjoyed the same theoretical success, though there are recent attempts of resuscitation of some of the issues involved. GROHMANN (1997, 2000, 2003), on

whose studies this contribution heavily relies, partly fills this gap by contrasting the properties of two types of left-dislocation constructions in German, also addressing some pertinent concerns in other Germanic dialects.1 The aim of this paper is to account for essentially the following paradigm: (la)

Diesen

Frosch, den hat die Prinzessin gestern geküßt. frog R P . A C C has the princess yesterday kissed 'This frog, the princess kissed (it) yesterday.' (lb) Dieser Frosch, den hat die Prinzessin gestern geküßt. this.NOM frog R P . A C C has the princess yesterday kissed 'This frog, the princess kissed it yesterday.' this.ACC

The construction in (la) is traditionally known as contrastive left dislocation (henceforth, CLD):2 an XP is LDed to the left periphery of the clause and coindexed with a pronominal element; in German, this element is a d(emonstrative)pronoun, which is homomorphous with the determiner. The d-pronoun is standardly taken to be a resumptive pronoun (RP). (la) stands in direct contrast with the construction in (lb), where the LDed element and RP do not match in Case (glancing over apparent preferences for the RPs' position), (lb) is known as nominativus pendens, Freies Thema, or, the term we employ, Hanging Topic Left Dislocation (HTLD). The aim of this paper is to provide an adequate analysis of both CLD and HTLD, with special emphasis on their respective derivational histories. On a theoretical level, we argue for a unified movement analysis which differs in its ancillary operations of Agree and Match. Typologically, this enables us to ac1

2

See GROHMANN (2003: 174-175) for comprehensive references as well as ALTMANN (1981) for

early, FREY (2004) for more recent discussion of the German phenomenon, and the collection of papers old and new in ANAGNOSTOPOULOU et al. (1997), including ample references cited. The term "contrastive (left) dislocation" was arguably first used by THRAINSSON (1979) and has recently come under criticism for German from FREY (2004), who simply calls it "German left dislocation."

140

CEDRIC BOECKX & KLEANTHES K . GROHMANN

count for the phenomenon of left dislocation across the Germanic languages collectively. It should be noted that we focus on German examples for two reasons. First, German is the only Germanic language with a fully functioning paradigm of morphological case (save Icelandic); as such, the crucial difference in case distinctions can be seen most clearly in this representative of the Germania. Second, and with this provocative comment we hope to spark further research, we believe that at this point there is not much of interest to be gained from looking at German's siblings and cousins (but see below); with a bit of luck, we will be proven wrong before long by fruitful research in this domain. Naturally, this does not mean that other dialects don't have left dislocation - they do. Let us illustrate. Noted as early as VAN RlEMSDlJK & Z w ARTS (1997), a seminal paper which initiated some interest in the mid to late Seventies, left dislocation structures do, of course, exist in Dutch. They concentrate on the difference between (2a) and (2b): (2a) De Hollanders, die zijn te flegmatiek. the Dutch they are too phlegmatic 'The Dutch (, they) are too phlegmatic.' (2b) Datportret, ik geloof niet dat hij het that portrait I believe not that he it 'That portrait, I don't think that he still has it.'

nog still

heeft. has

In Dutch, too, we can observe two different pronominal forms resuming the left dislocate: (2a) contains a d-pronoun, (2b) a p(ersonal)-pronoun analogous to German. However, in the absence of case distinctions, we cannot observe the same distributional patterns as in German. As a matter of fact, the d-pronoun is restricted to a "high" occurrence, while the p-pronoun appears in a "low" position. In topological field theory, prevalent in much of the German grammatical tradition, high occurrence equals the Vorfeld, the position in the 'prefield' roughly correlating to CP; it is the constituent that immediately precedes V2, the verb in second position. The low position is some structural position within the Mittelfeld, the 'middlefield' (roughly, between V2 and the sentence-final non-finite verb form, in more modern terms some position between C and V). In fact, one might be tempted to descriptively correlate the low occurrence of a resuming personal pronominal form (p-pronoun) with the above-mentioned hanging topic construction (HTLD) and the high occurrence of a d-pronoun with contrastive dislocation (CLD); this is essentially the Dutch pattern and that found in many other dialects (such as Frisian or West Flemish). Other varieties of Germanic, however, don't seem to show this morphological distinction in RPs. Consider Icelandic, from Z A E N E N (1997): (3a) Pessi

hringur,

Olafur hefur lofad

this.NOM ring.NOM Olaf

(3b) Pessum hring,

honum

has

promised

hefur Olafur

this.DAT ring.DAT it.DAT has Olaf 'This ring, Olaf promised (it) to Mary.'

Mariu

honum.

Mary.DAT it.DAT

lofad Mariu. promised Mary.DAT

141

Left dislocation in Germanic

In this case, the RP is in both cases identical to the p-pronoun; there is no resuming d-form in Icelandic. As the examples so far show, all patterns are found across the Germanic languages: (i) CLD and HTLD with (ii) low and high resumptives of (iii) both the d- and p-pronominal variety as well as (iv) with Case-matching and without.3 As the reader can already deduce from the translations offered above (where we adopt the convention of (3) in the following), English, too, makes a distinction between (HT) left dislocation and a structurally as well as derivationally quite different type of preposing structure, topicalization: when English employs an RP, it resembles in all relevant aspects HTLD of the other Germanic dialects, and when it doesn't use an RP at all, it acts, looks and feels like CLD - or in other words, there is a striking similarity between topicalization and CLD distinct from HTLD, and English simply lacks the formal way of marking CLD. Terminology aside, GROHMANN pursues and refines this descriptive state of affairs extensively in his work. It can easily be observed that only German employs all options in (nearly) all possible ways. This shall serve as ample justification to focus on German in the remainder of our Germanic discussion, with the analysis lending itself naturally to cross-Germanic extension. We will focus on the nearlyproperty of all possible ways and suggest theoretical means to capture the differences and missing patterns.

2. Two kinds of CLLD Because the aforementioned construction known as clitic left dislocation (CLLD), which does not occur at all in the Germanic languages and dialects, has been subject to intensive scrutiny over the years, we find it useful to note some properties of CLLD that will be instrumental for an analysis of CLD. We illustrate with AOUN & BENMAMOUN (1998), who examine in detail patterns of CLLD in Lebanese Arabic. In particular, they focus on how CLLD interacts with other A-bar processes such as topicalization and Wh-movement. (The latter is traditionally considered the operation that subsumes the former.) As in many languages, CLLD is characterized by the fronting of an NP to the beginning of the clause and the presence of an argumental clitic related to (i.e. resuming) the fronted NP. (4) illustrates CLLD in Lebanese Arabic: (4)

Naadya feef-a Kariim Nadia saw.3SG.M-her Karim 'Nadia, Karim saw her yesterday.'

mbeerih. yesterday

For an extension to left dislocation inside nominals in Dutch, West Flemish and (dialects of) Norwegian, see GROHMANN & HAEGEMAN (2003), elaborated and extended to German in GROHMANN (2003: ch. 6). The Dutch construction in (i) illustrates the type of structure we have in mind: (i) Janj, dief ζ 'nt late vertrek Jan that his late departure 'Jan's late departure'

142

Cedric Boeckx & Kleanthes K. Grohmann

As in other contexts of resumption in Lebanese Arabic, the weak RP related to the LDed NP, the dislocate, is insensitive to islands: (5)

Sma?t

7

Naadya

raht

[man duun

ma

heard. 1SG

that

Nadia

lefl2SG.M

without

C

anno

tahke

ma^-a],

talking.2SG

with-her

Ί heard that Nadia, you left without talking to her.' Wh-movement and topicalization (when not accompanied by an RP) behave much as they do in English and are sensitive to islands. Both Wh-movement and topicalization are possible across a left dislocate, as in (6)-(7). (6)

9 fu Naadya (smarte anno) xabbaruw-a? what Nadia heard.2SG.M that told.3PL-her 'What, Nadia, did (you hear that) they tell(/told) her?'

(7)

9 Nakte Naadya (smarte anno) xabbaruw-a. joke Nadia heard.2SG.M that told.3PL-her Ά joke, Nadia, (you heard that) they tell(/told) her.'

However, AOUN & BENMAMOUN observe that Wh-movement (and topicalization) across a left dislocate is ungrammatical if the dislocate is separated from the RP by an island. Witness (8), where Wh-movement is involved: (8)

*fu Naadya xabbaro Kariim [9abl ma feef-a] what Nadia told.3PL Karim before C saw.3SG.M-her 9 l-mTallme aalit? the-teacher.F

9

anno that

said.3SG.F

'What, Nadia, did they tell Karim before he saw her that the teacher said?' AOUN & BENMAMOUN show that a similar contrast obtains in the case of long distance Wh-movement/topicalization. If the CLLDed element is not separated from the RP by an island, the long-distance operation is fine (9). If, however, the dislocate is separated from the RP by an island, Wh-movement/topicalization across it gives rise to deviance (10). (9)

9 fu smarte anno Naddya what heard.2SG.M that Nadia 'What did you hear that Nadia, they told her?'

(10)

7 *fu sma?t anno Naadya xabbaro Kariim what heard.2SG.M that Nadia told.3PL Karim 9 9 feef-a] anno l-m^allme aalif>

saw.3SG.M-her

that

the-teacher.F

xabbaruw-al told.3PL-her [9abl before

ma C

said.3SG.F

'What did you hear that Nadia they told Karim before he saw her that the teacher said?' Summing up, an Α-bar dependency can be created across CLLD as long as the dislocate is not separated from the RP it is related to by an island. AOUN & BEN-

143

Left dislocation in Germanic

MAMOUN's interpretation of this generalization is that CLLD may be a case of either base-generation or movement. In island contexts, the movement option is unavailable, hence the patterns of interacting Α-bar dependencies above. AOUN & BENMAMOUN further justify their dual analysis of CLLD by noting that reconstruction effects with resumption are found only in non-island contexts (see also AOUN et al. 2001, AOUN & Li 2003), that is, in contexts where movement can take place, opening the possibility of reconstruction by leaving a copy of the moving element behind. Consider (11) vs. (12): (11)

TalmiiZj-a, l-kssleen

ma

student-her the-bad

NEG want. 1 PL told. 1 PL

nxabbir [wala no

mfallme], teacher

?

ha-l-ma jduubj

za?bar

b-l-fahS.

that

3PL-the-idiot

cheated.3SG.M

in-the-exam

snno

?

baddna

'Her bad student, we didn't tell any teacher that this idiot cheated on the exam.' (12) *Telmiizj-aj l-kasleen student-her

the-bad

ma

hkiina

NEG

talked. 1 PL with

maz [wala m?allme]j no

teacher

?

[ abl-ma ha-l-maljduubj yuuSal\. before the-idiot arrive 'Her bad student, we didn't talk to any teacher before the idiot arrived.' Let's now turn back to German(ic) and revisit the types of left dislocation noted above.

3. Two kinds of Germanic LD GROHMANN (1997, 2000, 2003) essentially entertains the same dual analysis for Germanic as AOUN & BENMAMOUN, with CLD being the movement and HTLD the base-generation structure. Before examining some of the tests GROHMANN brings to bear on the issue, let us note a distinction between CLD and HTLD that will be relevant shortly. Consider the somewhat fuller paradigm in (13)—(15), where the a-examples extend the basic paradigm of (1) above and the b-examples contain the weak pronominal es which is independently known not to occur in the Vorfeld:

(13a) [Diesen Frosch], den/*ihn

hat [π die Prinzessin geküßt]. the princess kissed 'This frog, the princess kissed (it).' (13b) [Dieses Buch], das/ *es sollten [TP . wir Martin geben]. this.ACC book RP.ACC/it should we Martin give "This book, we should give (it) to Martin.' this.ACC frog

RP.ACC/him

has

(14a) [Dieser Frosch], den/ *ihnhat[Tρ this.NOM frog RP.ACC/him has 'This frog, the princess kissed him.'

die Prinzessin the princess

geküßt]. kissed

144

CEDRIC BOECKX & KLEANTHES K . GROHMANN

(14b) [Dieses Buch], das/ *es sollten [TP wir Martin this.NOM book RP.ACC/it should we Martin 'This book, we should give it to Martin.' (15a) [Diese-r/-n (15b) [Dieses

Frosch], Buch],

geben]. give

[TP die Prinzessin hat den/ihn geküßt], [TP wir sollten das/es Martin geben].

As already noted above, in some cases dislocate and RP agree in Case, in others they don't. Of special interest here is that in the structures (13), where both do agree, the RP is outside TP if we take the subject position to demarcate the boundary between 'C-domain' and 'T-domain' (see GROHMANN 2003 for discussion). We can observe that in situations where Case-matching fails, a p-pronoun may be used in addition to a d-pronoun, and that the RP may be either high (above TP) or low (inside TP). The high occurrence of an RP is often equated with a topic position, giving rise to the verb-second structure observed in the clause following the dislocate.4 When the RP is low, Case-matching may be available. For reasons discussed in detail in GROHMANN (2000, 2003), we take both Case-identity and position of RP to be indicators of LD-type. In particular, we claim that low RP signals HTLD. High RPs are compatible with HTLD only in situations of Case-mismatch. In other words, (13) exemplifies CLD, (14) and (15) HTLD. GROHMANN argues further that CLD is the result of movement (like topicalization), whereas HTLD arises from base-generation. Two core facts which he substantiates his claim with are reconstruction asymmetries and island asymmetries.5 3.1. Reconstruction asymmetries The examples in this subsection show that CLD exhibits reconstruction effects for Binding Condition C, whereas HTLD does not. When dislocate and RP agree in Case, coreference of the R-expression inside and a subject pronoun is impossible, but when there is no Case-agreement, coreference is fine. As the translation shows, the same applies to English topicalization. (Intended coreference is indicated with underlining.) (16a) *[Der Freundin, die Martin geholfen the.DAT friend who Martin helped der gab er einen Kuß. RP.DAT gave he a kiss ""The friend who helped Martin, he gave a kiss.'

4

5

hat], has

FREY (2004) makes the case for a TP-internal topic position for RPs that in some instances may appear in an apparently low position, yet behave in all respects like a pronominal resuming a CLDed element. We take FREY'S research as an indicator of well-motivated revived interest in the topic matter and hope for fruitful results in this area, including further discussion of such topic position(s). See GROHMANN'S work for additional tests. The tests illustrated here were chosen to highlight the parallelism between German and Lebanese Arabic presented in section 2.

145

Left dislocation in Germanic

(16b) *[Die Freundin, die Martin geholfen hat], the.NOM friend who Martin helped has der gab er einen Kuß. RP.DAT gave he a kiss 'The friend who helped Martin, he gave her a kiss.' Given that CLD involves movement of the dislocate from a lower position within the clause, there would be a Condition C violation if the CLDed element could reconstruct in (16a). The R-expression in base-generated HTLD may freely be coreferent with the lower pronoun as it would never be in its domain. The same can be witnessed in cases of long-distance LD for Case-agreement or absence thereof as well as high and low RPs (bold). (17a) *[Der

Frau,

die

Martin geholfen hat], der sagte sein has RP.DAT said his Bruder, gab er einen Kuß. brother gave he a kiss '*The woman who helped Martini, her said his; brother he; gave a kiss.' (17b) [Die Frau, die Martin geholfen hat], der sagte sein the.NOM woman who Martin helped has RP.DAT said his Bruder, gab er einen Kuß. brother gave he a kiss (17c) [Die Frau, die Martin geholfen hat], sein Bruder sagte, the.NOM woman who Martin helped has his brother said er gab ihr einen Kuß. he gave her a kiss 'The woman who helped Martini, hisj brother said hej gave her a kiss.' the.DAT woman who Martin helped

These data suggest that both properties play a role, Case-matching between dislocate and RP on the one hand (17a-b) and high vs. low position of the RP on the other (17c). Further connectedness effects are discussed in GROHMANN (2003: ch. 4, sect. 3.2).

3.2. Island asymmetries German LD allows violation of a Wh-island (18), and of the adjunct island (19), but crucially only if the RP is low. (18a) [Diesem Frosch], was hat die Prinzessin dem gegeben? this.DAT frog what has the princess RP.DAT given 'This frog, what did the princess give to (*him)?' (18b) [Diesen Frosch], wer glaubt der Bauer, hat den geküßt? this.ACC frog who believes the farmer has RP.ACC kissed 'This frog, who does the farmer believe kissed (*him)?'

146

CEDRIC BOECKX & KLEANTHES K . GROHMANN

(19a) [Der schöne Mann], Martin haßt die Tatsache, dass den that RP.ACC the.NOM handsome man Martin hates the fact die Frau geküßt hat. the woman kissed had (19b) [Der schöne Mann] Martin haßt die Tatsache, dass the.NOM handsome man Martin hates the fact that die Frau ihn geküßt hat. the woman him kissed had 'The handsome man, Martin hates the fact that the woman kissed * (him).' The data fall out as expected if HTLD (the only LD-type compatible with a low RP) is an instance of base-generation, akin to the English translations indicated (which only has movement-derived topicalization or base-generated HTLD). Contrast this state of affairs with the CLD examples in (20)-(21). (20a) *[Diesem Frosch], was hat die Prinzessin gegeben? '•This frog, what did the princess give to?' (20b) *[Diesen Frosch], wer glaubt der Bauer hat geküßt? ""This frog, who does the farmer believe kissed?' (21) *[Den

schönen Mann] den haßt Martin handsome man RP.ACC hates Martin die Tatsache, dass die Frau geküßt hat. the fact that the woman kissed had '*The handsome man, Martin hates the fact that the woman kissed.' the.ACC

If the hanging topic is base-generated, island effects will certainly fall out immediately.

4. Against base-generation BOECKX (2003a) argues that despite conflicting evidence emerging from locality conditions, a unified approach to resumption is warranted. Although the fact that some RPs, but not others, are sensitive to islands suggests that some RPs relate to their antecedents by movement, while other RPs favor a base-generation analysis, BOECKX catalogs a series of facts that directly argue against a hybrid theory of resumption. In particular, irrespective of their behavior in island contexts, RPs are subject to some solid cross-linguistic generalizations (see BOECKX's work for illustration). For instance, irrespective of island (in)sensitivity, RPs appear to trigger a specific reading on the antecedent and seem to be compatible with discourse-linked interrogatives only (see also BOECKX & GROHMANN 2003). A second important generalization about RPs is that many properties of resumption can be shown to depend on the complementizer system of the language, and not on properties of the pronouns themselves. Third, a vast majority of the languages that make use of RPs

Left dislocation in Germanic

147

isolate the subject position, either by restricting RPs to that position (like Vata, for example) or by .banning them from that position (as e.g. Irish and Hebrew do). On the whole, we think that the importance of the island data has been overestimated. As BoECKX shows in detail, a unified theory of resumption is attainable. Such a theory should be movement-based (for further motivation see HORNSTEIN 2001 or the rigorous pursuit of minimalist desiderata presented in GROHMANN 2003: ch. 2, for example). To make it work, BOECKX takes as its starting point CHOMSKY'S (2000:122) Probe-Goal model of syntactic relations, summarized in (22). (22a) Features Match •=> ex.: φ-features on a subject NP match those on finite T° (22b) (Properties of) Features trigger Agree Ο ex.: the value(s) of the φ-features of the subject NP are transmitted to T° (22c) (Properties of) Features trigger Move Ο ex.: the subject NP raises to SpecTP Following CHOMSKY (2001: 5), we take Match to be a relation holding of two items sharing a feature. To illustrate, there is a [Wh]-feature on C that matches the [Wh]-feature on the Wh-word that ultimately raises (e.g. in a question like What did John buy?). Agree is a potentially long-distance agreement relation holding between two elements (which CHOMSKY calls Probe and Goal) that have matching features. Move is a function of the ill-understood EPP-property of a probe that demands that a goal be remerged into its specifier. Taken together, Match, Agree and Move characterize displacement phenomena in natural languages. The presence of uninterpretable features on an element turns it into a Probe. Match determines what kind of category the Probe seeks (one with a matching feature F). Agree establishes the feature checking relation between Probe and Goal (F on the Goal values F on the Probe). The EPP property determines whether the Probe offers a (specifier) position for Move. In various works (BOECKX 2003a, 2003b, in press, BOECKX & JEONG 2002), BOECKX has argued for the need to depart from CHOMSKY'S assumptions just stated. In particular, he argues that Agree cannot take place in the absence of Match, which seems uncontroversial. He also assumes that Match is a prerequisite for EPP-satisfaction (recall that the EPP is not an independent feature, but a property of a feature). But unlike CHOMSKY, BOECKX does not take Agree to be a prerequisite for Move (see also BOBALJIK & WURMBRAND 2003 for keeping Move and Agree as distinct operations, not parasitic upon one another). In particular, he provides extensive arguments in favor of allowing Move to take place solely under Match. To mention just a few exemplary instances, across languages the distinction between Match and Agree is responsible for the distribution of inflected and uninflected complementizers, Case-(mis)matching between an RP and its antecedent, distinct locality effects (in particular, apparent island-violations), and possibly (anti-)reconstruction effects (BOECKX & HORNSTEIN 2004). We will not detail the differences that result from forming chains via pure Match vs. Match + Agree. Rather, we would like to suggest that HTLD is the result of movement of the LDed antecedent away from its RP-associate (in effect, a case of stranding), as de-

148

CEDRJC BOECKX & KLEANTHES K. GROHMANN

picted in (23), where angled brackets indicate the lower trace/copy of the moved element. (23) NPj... [TP · · · [DP RP []] ... ] Movement of NP in (23) takes place under Match only, i.e. without Agree. As a result (for reasons detailed in BOECKX's work), it evades island effects, and does not license reconstruction (understood as the interpretation of the lower copy of NP). In addition, for the movement to be licit, no agreement between the NP and the RP is allowed to take place - hence, the Case-mismatch between the RP and the hanging topic NP; the NP receives default Case (nominative in German). CLD proceeds differently. In particular, we assume that Agree is involved in such cases, hence the presence of islands effects and reconstruction, but also Caseidentity between dislocate and RP. For present purposes, we follow GROHMANN (1997, 2000, 2003) in taking the RP in such cases to be a (minimal) spell-out of the lower copy of the moving NP, as sketched in (24). (For a slightly different implementation that maintains a stricter parallelism between (23) and (24), see BOECKX 2003a: ch. 4, sect. 4.4.1.) (24) NPj (...) [ Ο RP] ... [TP ... ... ] The Copy Spell Out analysis of CLD captures the fact that for all intents and purposes, the RP acts as a copy of the NP. Because Copy Spell Out is licensed only in cases where movement is too local (see GROHMANN 2000, 2003 for detail), it limits the strategy in (24) to those cases where the RP is a high pronoun (within an articulated Comp or CP). The main point of this section is that we do not need to resort to a dual, basegeneration vs. movement analysis to capture the differences between HTLD and CLD. The independently motivated distinction between Match and Agree suffices to do so, while it allows us to maintain a unified, movement-based analysis of left dislocation in Germanic. As a result of the technical details of this analysis, our analysis of HTLD will also counter FREY'S ( 2 0 0 4 ) quibbles with GREWENDORF's ( 2 0 0 2 ) "Big DP"-approach, of which the current proposal is a refinement (see also KAYNE 2 0 0 2 ) . The parallelism between HTLD and CLD as suggested by GREWENDORF is too simplistic and overgenerates, a problem the present proposals circumvent. We take this conjecture to be a further ingredient of our attempt to revitalize deeper research - empirically as well as theoretically - on the issue of left dislocation in Germanic. To conclude this section we would like to point out that the movement analysis of HTLD accounts for an important, and to our knowledge novel, observation made in GROHMANN ( 2 0 0 3 ) . He notes that in instances of multiple HTLD (with the right intonation certainly possible, unlike CLD; cf. fn. 6 below), the order of RPs mirrors that of full noun phrases in German if they show up in their d-form, and that of pronouns if in p-form. It is well known among Germanicists that pronouns occurring in the Mittelfeld are subject to a strict ordering pattern in form of a pronominal hierarchy (cf. LENERZ 1977, 2001). Thus in the unmarked case, when no additional effects come into play (such as intonation/stress, topic/focus, theme/rheme structure etc.), the order-

149

Left dislocation in Germanic

ing among personal pronouns is s u b j e c t » direct object»indirect object (or NOM » ACC » DAT in the canonical case). We are not interested in pinpointing specific positions in the clause structure. Rather we take the Mittelfeld to correlate with the rough structure of TP in (23)—(24) above, below V2 which sits somewhere within CP. (26) illustrates with an embedded clause, where V2 does not have any effects. This stands in opposition to full (definite argument) phrases, as (25) shows; marked orders are signaled by the hash mark '!'. (25a) ... weil die Mutter dem Alex den Wagen geschenkt hat. since the.NOM mother the.DAT Alex the.ACC car given has '... since the mother gave the car to Alex.' (25b) !... weil die Mutter den Wagen dem Alex geschenkt hat. (25c) !... weil dem Alex die Mutter den Wagen geschenkt hat. (25d) !... weil dem Alex den Wagen die Mutter geschenkt hat. (25e) !... weil den Wagen die Mutter dem Alex geschenkt hat. (25f) !... weil den Wagen dem Alex die Mutter geschenkt hat. (26a) ... weil sie

ihn

ihm

since she.NOM him[it].ACC him.DAT '... since she gave it to him.'

(26b) (26c) (26d) (26e) (26f)

!... !... !... !... !...

weil sie ihm ihn geschenkt weil ihn sie ihm geschenkt weil ihn ihm sie geschenkt weil ihm sie ihn geschenkt weil ihm ihn sie geschenkt

geschenkt hat. given

has

hat. hat. hat. hat. hat.

Observe now that resuming d-pronouns adhere to the unmarked order of full phrases, as (25), regardless of the order of hanging topics if there is more than one:6 (27) [Der Alex]i, [der Wagen\ [seine Mutter]k, the.NOM Alex the.NOM car the.NOM mother (27a) gestern hat diek demj denj geschenkt. yesterday has RP.NOM RP.DAT RP.ACC given

(27b) (27c) (27d) (27e) (27f)

'Alex, the car, the mother, yesterday, she gave it to him.' *gestern hat diek denjdemi geschenkt. *gestern hat demt diek denj geschenkt. *gestern hat demt denj diek geschenkt. *gestern hat demt diek denj geschenkt. *gestern hat demt denj diek geschenkt.

The data in (27) are puzzling if RPs are treated as pronouns linked to their antecedent in a non-movement fashion. However, they fall out straightforwardly from the analysis above, which treats HTLD as a movement process. Recall from (23) that at some point, the antecedent NP and its RP associate form a full DP structure. 6

WERNER ABRAHAM (p.c.) calls such multiple hanging topics themata pendentia in extreme, a fitting term since multiple HTLD calls for special contextual licensing. We address their relevance presently.

150

CEDRJC BOECKX & KLEANTHES K . GROHMANN

It is at that point that the ordering restrictions noted in (27) are determined, prior to movement of the LDed NP. Now, one might hold that what these data show is that hanging topics simply do not properly belong to the clause they occur with, that they are outside the clausal domain. We do not want to specify the position of the dislocates in this article (GROHMANN 2000, 2003 assumes the dislocate to sit in a specifier within the Cdomain in CLD and in an adjoined position in HTLD), but we do want to address the issue of belonging. If the hanging topic did not belong to the clause, one would expect the pronominal ordering without further ado, since we know that dpronouns are heavier and more tonal than the weak p-pronouns. However, as BOECKX's work on resumption shows an integrated account is warranted even in contexts that have traditionally been assumed to be void of any (movement) dependencies. Methodologically, this is a desired result, since we do not have to evoke additional assumptions (such as linking rules) or unwanted entities (such as specific, grammar-internal constructs; see also HORNSTEIN 2001, GROHMANN 2003). In particular, BOECKX (2003c) applies the Match-Agree approach to donkey anaphora and thus demonstrates how apparently clause-external relationships can be integrated into a general theory of grammar conforming to independently motivated operations.

5. Final remarks Types of LD, as first studied by ROSS (1967), are interesting for a variety of reasons. For one, they seem to express the same semantic, i.e. truth-conditional relation that topicalization does, thus are formally similar to the latter construction. But unlike topicalization, LD involves resuming the LDed element somewhere lower in the clause. This strategy presumably accounts for the pragmatic, functional differences observed between the two phenomena (as BLRNER & WARD 1998 show). With regard to resumption, languages either express this with a clitic (such as Romance languages and Arabic) or with a different pronominal element which is specially marked; in German, this is canonically the demonstrative form. Aside from the form of resumption, the position of the resumptive element varies. In German, the language of our main focus, the resumptive is either in a high position or in a low position. We identified the high position with a topic position and the low position with the canonical argument position. However, apart from these descriptive facts, types of LD also pose a real problem: there seem to be different strategies of LDing elements, not only cross-linguistically but also within a given language, so how can we distinguish the two? And if it turns out that we can, what does this show us? We have argued in favor of a unified movement-based analysis of HTLD and CLD in German, and showed how the differences between the two strategies fall out from independently motivated processes, such as movement under pure Match or Copy Spell Out. In so doing, we have incorporated Germanic LD into the rich landscape of resumption detailed in B O E C K X (2003a) and G R O H M A N N (2003).

151

Left dislocation in Germanic

Beyond the rhetoric espoused throughout, we hope to have empirically supported our provocative introductory remarks. In light of the theoretical proposal presented in section 4, it now remains to be seen whether one can revisit the corresponding structures in German's numerous relatives. Just as we relied on a comparison with CLLD found in Lebanese Arabic (but also closer to home, such as the Romance languages or Greek), it might be instrumental to scrutinize the arising patterns not only from a pan-, but also from a per-Germanic comparison. We have in mind such interesting facts such as the following from mainland Scandinavian. As MARIT JULIEN (p.c.) remarks, a subject or a left dislocate can be resumed in the right periphery of the clause. (28) illustrates with Norwegian. (28a) Jeg har sett den I have seen the (28b) *Jeg har sett den I have seen the Ί have seen the film.'

filmen, film.DEF filmen, film.DEF

(29a) Den filmen, den har the film.DEF this have (29b) Den filmen, den har the film.DEF this have 'The film, I have seen.'

jeg I jeg I

sett, seen sett, seen

jeg. I den. the jeg. I den. the

We leave further interpretation as food for thought. It strikes us, however, that here we might have to integrate considerations not only pertaining to left, but also to right dislocation, another phenomenon (more widely) found in Romance languages, for example (see VLLLALBA 2000 for a recent in-depth empirical treatment of right dislocation in Romance and beyond, and a theoretical comparison with left dislocation).

Abbreviations ACC C

CLD CLLD DAT DEF

DP F

HTLD

accusative complementizer contrastive left dislocation clitic left dislocation dative definite Determiner Phrase feminine hanging topic left dislocation

LD Μ NEG NOM

NP PL RP/RP SG

TP

left dislocation masculine negation nominative Noun Phrase plural resumptive pronoun singular Tense Phrase

References Formen der "Herausstellung" im Deutschen: Rechtsversetzung, Linksversetzung, Freies Thema und verwandte Konstruktionen. Tübingen: Max Niemeyer. ANAGNOSTOPOULOU, ELENA; VAN RIEMSDUK, HENK & ZWARTS, FRANS (eds.) ( 1 9 9 7 ) : Materials on left dislocation. Amsterdam: John Benjamins. AOUN, JOSEPH & BENMAMOUN, ELABBAS ( 1 9 9 8 ) : Minimality, reconstruction and PF-movement, in: Linguistic Inquiry 2 9 . 4 , 5 6 9 - 5 9 7 . ALTOANN, HANS ( 1 9 8 1 ) .

152

CEDRIC BOECKX & KLEANTHES K . GROHMANN

(2001): Resumption, movement and derivational economy, in: Linguistic Inquiry 32.3, 371-403. AOUN, JOSEPH & LI, YEN-HUI AUDREY ( 2 0 0 3 ) : Essays on the representational and derivational nature of grammar: The diversity ofWh-constructions. Cambridge, Mass.: MIT Press. BIRNER, BETTY J. & WARD, GREGORY (1998): Information status and noncanonical word order. Amsterdam: John Benjamins. BOBAUIK, JONATHAN DAVID & WURMBRAND, SUSANNE (2003): Relativized phases. Ms., University of Connecticut, Storrs. BOECKX, CEDRIC (2003a): Islands and chains. Amsterdam: John Benjamins. BOECKX, CEDRIC (2003b): Symmetries and asymmetries in multiple checking, in: BOECKX, CEDRIC & GROHMANN, KLEANTHES K . (eds.), Multiple Wh-fronting. Amsterdam: John Benjamins, 17-26. BOECKX, CEDRIC (2003c): (In)direct binding, in: Syntax 6.3,213-236. BOECKX, CEDRIC (in press): Case matters and minimalist concerns, in: HORNSTEIN, NORBERT & URIAGEREKA, JUAN (eds.), The minimalist fact. Cambridge, Mass.: MIT Press. BOECKX, CEDRIC & GROHMANN, KLEANTHES K . (2003): Introduction, in: BOECKX, CEDRIC & GROHMANN, KLEANTHES K . (eds.), Multiple Wh-fronting. Amsterdam: John Benjamins,1-15. BOECKX, CEDRIC & HORNSTEIN, NORBERT (2004): Superiority, reconstruction, and islands. Ms., Harvard University, Cambridge, Mass. & University of Maryland, College Park. BOECKX, CEDRIC & JEONG, YOUNGMI (2002): The fine structure of syntactic intervention. Ms., University of Maryland, College Park. CHOMSKY, NOAM (2000): Minimalist inquiries: The framework, in: MARTIN, ROGER; MICHAELS, DAVID & URIAGEREKA, JUAN (eds.), Step by step. Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, Mass.: MIT Press, 89-155. CHOMSKY, NOAM ( 2 0 0 1 ) : Derivation by phase, in: KENSTOWICZ, MICHAEL (ed.), Ken Hale. A life in language. Cambridge, Mass.: MIT Press, 1-52. CINQUE, GUGLIELMO ( 1 9 7 7 ) : The movement nature of left dislocation, in: Linguistic Inquiry 8 . 2 , 3 9 7 - 4 1 1 . CINQUE, GUGLIELMO (1990): Types ofA'-dependencies. Cambridge, Mass.: MIT Press. FREY, WERNER ( 2 0 0 4 ) : Notes on the syntax and the pragmatics of German left dislocation, in: LOHNSTEIN, HORST & TRISSLER, SUSANNE (eds.), The syntax and semantics of the left periphery. Berlin: Mouton de Gruyter, 2 0 3 - 2 3 3 . GREWENDORF, GÜNTHER (2002): Left dislocation as movement, in: Georgetown University Working Papers in Theoretical Linguistics 2, 31-81. GROHMANN, KLEANTHES Κ . ( 1 9 9 7 ) : On left dislocation, in: Groninger Arbeiten zur germanistischen Linguistik 4 0 , 1 - 3 3 . GROHMANN, KLEANTHES Κ. (2000): Α movement approach to contrastive left dislocation, in: Rivista di Grammatica Generativa 25, 3-65. GROHMANN, KLEANTHES K . ( 2 0 0 3 ) : Prolific domains. On the anti-locality of movement dependencies. Amsterdam: John Benjamins. GROHMANN, KLEANTHES K . & HAEGEMAN, LILIANE ( 2 0 0 3 ) : Resuming reflexives, in: Nordlyd 3 1 . 1 , 4 6 - 6 2 . HORNSTEIN, NORBERT ( 2 0 0 1 ) : Move! A minimalist theory of construal. Oxford: Blackwell. KAYNE, RICHARD S. ( 2 0 0 2 ) : Pronouns and their antecedents, in: EPSTEIN, SAMUEL DAVID & SEELY, T . DANIEL (eds.), Derivation and explanation in the minimalist program. Oxford: Blackwell, 133-166. LENERZ, JÜRGEN ( 1 9 7 7 ) : Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: Narr. LENERZ, JÜRGEN ( 2 0 0 1 ) : Word order variation: Competition or co-operation, in: MÜLLER, GEREON & STERNEFELD, WOLFGANG (eds.), Competition in syntax. Berlin: Mouton de Gruyter, 2 4 6 - 2 8 1 . Ross, JOHN R. (1967): Constraints on variables in syntax. Doctoral dissertation, Massachusetts Institute of Technology, Cambridge. VAN RIEMSDUK, HENK & ZWARTS, FRANS ( 1 9 9 7 ) : Left dislocation in Dutch and status of copying rules, in: ANAGNOSTOPOULOU, VAN RIEMSDUK & ZWARTS (eds.), 1 1 9 - 1 4 8 . [Original manuscript from 1974] THRÄINSSON, HOSKULDUR (1979): On complementation in Icelandic. New York: Garland. VILLALBA, NICOLAS, F. XAVIER (2000): The syntax of sentence periphery. Doctoral dissertation, Universität Autönoma de Barcelona. ZAENEN, ANNE (1997): Contrastive dislocation in Dutch and Icelandic, in: ANAGNOSTOPOULOU, VAN RIEMSDUK & ZWARTS (eds.), 119-148. [Original manuscript from the early 1980s] AOUN, JOSEPH; CHOUEIRI, LINA & HORNSTEIN, NORBERT

Verb sequence and placement: Afrikaans and Dutch compared C. JAC CONRADIE (Johannesburg)

1. Vlast and V2 in Afrikaans and Dutch The process of syntactic divergence which started 350 years ago with the introduction of Dutch as a major language at the Gape of Good Hope has given rise to the following situation: (a) (b) (c)

(d)

Afrikaans still closely resembles Dutch as far as basic constituent order, including verb-first, second and final placement in clauses is concerned. The sequence (or order) of verbs in a multi-verb clause is more fixed in Afrikaans than in Dutch. For reasons of changes in structure, combinatorial restrictions and usage, the placement (or spreading) of verbs in a multi-verb clause is freer in Afrikaans than in Dutch, causing a general shift in verb placement to the left. Thus the verbal "point of gravity", particularly in the spoken language, has shifted much more towards Vl/2 than is the case in Dutch. As Afrikaans retains a strong tendency to end clauses with an (often "frozen") verb string and (where applicable) the repeated negation nie, the structural rift between two systems of verb placement which also characterizes Dutch and German, seems to have widened in Afrikaans.

In sum, though Afrikaans has undergone and is undergoing changes in the direction of the VO pattern of English, a strong verb final tendency, with elements of grammaticalization, has remained. My main aim will be to adduce evidence for the greater leftward spread of Afrikaans verbs in contrast to Dutch, cf. (c) above. This will be based mainly on three corpora of spoken Afrikaans from the late 20th century (KROES 1982) and some recent Afrikaans novels and short stories. First of all, (b) and (d) will be dealt with briefly.

2. Verb sequence The sequence or order of Afrikaans verbs, as they may appear concatenated at the end of a dependent clause - typically followed by the second negation nie to end the rhematic section of the phrase - may be represented as follows:

154

C. JAC CONRADIE

1 modal verbs: e.g. moet 'must' kan 'can'

2 linking verbs1: e.g. laat 'let' probeer 'try'

3 main verb: e.g. sien 'see'

4 auxiliary verbs: word 'become' or wees 'be'

5 auxiliary verb: het 'have'

Table 1: Afrikaans verb sequence This sequence is maintained despite the placement of a verbs (or verbs) in clause initial or second position, which will be referred to as "VI" and "V2", respectively. The first verb to the left not marked as past participle, which may be het, is eligible for movement into Vl/2 - provisos for "fused" verbs to be discussed later on. The following example from VANDEWEGHE (2000: 232), with auxiliary verb + modal verb + linking verbs (including another verb) + main verb suggests a different order for Dutch, particularly as far as the possible placement of hebben (Afrikaans he) is concerned (my gloss, CJC). (1)

Ik (zou) je wel eens I would you really once HEBBEN WILLEN ZIEN DURVEN BLIJVEN STAAN KIJKEN 5 1 2 1 2 2 3 have will see dare-to remain stand look Ί would really have liked to see you daring to keep on looking for once.'

What the two languages have in common is that modal verbs generally precede linking verbs, which in turn precede main verbs, as in Dutch willen blijven staan, Afrikaans wil bly staan. modal verbs, e.g. Dutch willen, Afrikaans wil 'want to' 1

linking verbs, e.g. Dutch blijven, Afrikaans bly 'remain' 2

main verb, e.g. Dutch/Afrikaans staan 'stand' 3

Table 2: Shared verb sequence As is the case in Dutch, Afrikaans has leftward past participle scrambling: The leftmost past participle scrambles to the left, unless bound by an immediately following auxiliary het. Apart from the het constraint, past participle scrambling is very similar to Dutch and does not have functional implications. This type of movement notably involves the only verb type which has retained explicit morphological marking in Afrikaans, viz. the past participle (through the prefix ge-). 1

The term "linking verbs" (Afrikaans skakelwerkwoorde, cf. PONELIS 1979) is used to refer to the main verbs of higher matrix clauses, some of them having aspectual functions vis-a-vis the main verb. The so-called "direct" linking verbs include laat 'let', begin 'begin', probeer 'try', gaan 'go', kom 'come', loop 'walk, go' and by 'get', while the "indirect" linking verbs - linked to the main verb by en 'and' - are restricted to the attitudinal verbs sit 'sit', staan 'stand', le 'lie' and loop 'walk, go'. Apart from en (absent in certain varieties of Afrikaans), there is no syntactic difference between the two types.

155

Verb sequence and placement: Afrikaans and Dutch compared

(2)

vrae

wat

wel...

uitgelaat kon gewees 3 1 4 questions that really omitted could been 'Questions which could in fact have been omitted'

het 5 have [HK]

Dutch has variations in clause-final verb sequence that are completely excluded in Afrikaans, e.g. (i) auxiliary verb + past participle (worden gedaari) and past participle + auxiliary verb (gedaan worden) in Dutch vs. only the latter (gedoen word) in Afrikaans, and (ii) infinitive + finite verb (doen moet) and finite verb + infinitive (moet doen) in Dutch vs. only the latter in Afrikanns. VANDEWEGHE (2000: 237) describes the second order in (i) as the preferred order in Southern Dutch and the second in (ii) as the "normal" one in Standard Dutch. These are also the orders generally preferred in Dutch according to an empirical study by HAESERYN (1990: 340). Note that the Dutch alternative order main verb + modal verb as in doen moet ('do must') forms an exception to the shared order described in Table 2 above. Another variant in Dutch involves the initial position in the sequence of finite hebben 'have' in (4) - similarly in subordinate clauses. (3)

Hij

moet het werk gedaan 1 3 he must the work done 'He must have done the work.'

(4)

Hij

heeft het werk 5 he have the work 'He had to do the work.'

moeten 1 must

hebben 5 have doen 3 do

From one underlying sequence, either a modal verb or an auxiliary verb may be finite or be placed in Vl/2. The latter sequence would either suggest a second underlying verbal sequence for Dutch or the lack of a restriction on which non-past participles can be raised to Vl/2 - both of these being instances of a greater flexibility in verb sequence than is the case in Afrikaans. However, while the constructional type heeft moeten doen/had moeten doen (the latter expressing the irrealis) frequently occurs in Dutch, its Afrikaans reflex as in het moet/moes doen - notably a departure from the Afrikaans verb sequence and Vl/2 selection rule stated above - has become severely restricted. While only expressing the realis (apart from fossilized expressions such as het nooit kon droom/dink 'could never have dreamt/ thought'), it is restricted to main clauses and used infrequently - probably forming part of the competence of a minority of speakers. See (5) and (6): (5)

Pa het brom-brom wou weet waarom ek dan nou sodanig Dad has moan.moan wanted.to know why I then now really Bloekombos toe gesleep moet word Bloekombos to dragged must become 'Dad, moaning all the time, wanted to know why I really had to be dragged off to Bloekombos' [VAN ROOYEN 2001: 25]

156 (6)

C. JAC CONRADIE

Sy't nooit kott dink bunions kan she.has never could think bunions can ' She would never have thought that bunions could

...'

[GREEFF

2001: 107]

In sum, whenever alternative sequences of verbal elements are possible in Dutch, only one of these will be found in Afrikaans and the other will either be non-existent in Afrikaans or severely curtailed in its usage. 3. Verb-final grammaticalization Though past participle + hebben 'have' is a possible sequence in Dutch verbal clusters, the Afrikaans construction deviates in various ways. Afrikaans distinguishes between past participle + he, where the past participle is [+adjectival] and may scramble, and the particle te 'to' is inserted between past participle and he as in Dutch, viz. om die werk gedoen te he 'to have the work done', and past participle [-adj] + het, where the purely verbal past participle does not scramble, the auxiliary het is invariable (i.e. does not have the infinitive form he) and te 'to' precedes the past participle, viz. om die werk te gedoen het 'to have done the work' all this in contrast to Dutch, e.g., in an independent clause: (7)

Hülle

hoef nie soveel moeite te 1 they need not so.much trouble to 'They need not have gone to such trouble.'

gedoen 3 done

het 5 have [BRINK

nie. not 1998: 205]

Het has not developed into a fully fledged suffix in that it is still eligible for Vl/2 placement, as in the case of any other non-past participle verb. The sequence past participle + het, particularly in juxtaposition with the duplicated negation nie is, however, conspicuous as a clause terminator - cf. gedoen het nie above. The verb sequence and nie are only separated by adverbials (including clauses) and through PP-over-V; nie - depending on the scope of negation and certain stylistic factors - is always final. The frequency of het is further enhanced by the fact that Afrikaans no longer employs the auxiliary verb wees 'be' for tensing mutative verbs and widely employs past participle + het to express the past tense, as well as the irrealis and hypothetical modes. Afrikaans verbal clusters thus have certain characteristics of fixture which, in conjunction with the clause final second negative, mark the termination of the rhema portion of the clause.

4. Verb placement The tendency towards earlier placement of verbs in the Afrikaans clause than in the Dutch one relates to (a) a preference for V2 subordinate clauses without conjunctions in general usage; (b) a change in the combinatorial characteristics of conjunctions allowing subordinate order to be replaced by main clause declarative or interrogative order (i.e. with V2 and even VI); (c) far-reaching penetration of final verbal clusters by noun phrases, prepositional phrases and the like, and (d) to

157

Verb sequence and placement: Afrikaans and Dutch compared

the frequent "fusion" of linking verb and main verb in VI and V2 positions. These issues will be discussed in paragraphs 4.1 to 4.4. below. 4.1. Preference for V2 subordinate clauses without conjunctions While also occurring in Dutch and English, V2 objective clauses without conjunctions are extremely frequent in Afrikaans2 as in (8) and (9): (8)

(9)

En ek is bang [hy sal,· my somtner tj and I am afraid he will me just 'And I am afraid he'll shoot me out of hand.' Nou now meer more

laat weet goewerneur let know Governor gestuur tt (moet) word sent must become

skiet] shoot [SK]

Willem Adriaan [daar moetnie Willem Adriaan there must.not nie] [ M a t t h e e 2000: 621] not

'Now Governor Willem Adriaan sent word that no more should be sent.' This clause type is also attested in the case of subjective clauses. Viz. (10). (10) dan sal dit nou veel help [hy soeki daarvoor ] then will this now much help he looks for.it 'It won't be of much help if he looks for it.' (ironical) [BLOEMHOF2001:121] Objective clauses of this kind are often characterized by topicalization, for example by the preposed object haar besluit in (11). (11) maar ek kon sien [haar besluit hett sy klaar geneemt( ] but I could see her decision has she already taken 'But I could see that she had already taken her decision' [BRINK 1998: 314]

4.2. A change in combinatorial characteristics In the spoken language the subordinate clause following dat 'that' - in contrast to Dutch, German and Frisian - very often has main clause order. See (12) and (13). (12) Maar ek glo dat 'η mens moett lank voor die tydti begin. but I believe that one must long before the time begin 'But I believe that one should begin well in advance.' [HK] (13) Dit beteken dat die werkwoord kom{ eerstet, this mean that the verb come first isj die tweede sinsdeel tj. (student in test) is the second constituent

en and

die onderwerp the subject

'This means that the verb comes first and the subject is the second constituent.' In (14) this is the case with the conjunction as ' i f : 2

Cf. FEINAUER (1990) for a description of this type.

158

C . JAC C O N R A D E

(14)

Gewoonlik as sy sou, iewers heen tt gaan usually if she would somewhere go los sy die kind hier by my. leave she the child here with me 'Usually when she would go somewhere, she would leave the child here with me.' [SK]

Interrogative words are increasingly followed by the corresponding main clause, inevitably with V2. This has even become a characteristic of the written language. The canonical order in (15)—(17) below is waar die fout is, wanneer jou geluk kom and hoe my vrou lyk. (15) Laat ons kyk [waar isι die fout ] let us look where is the mistake 'Let us look where the mistake is.'

[SK]

(16) mens weet nooit [wanneer komi jou geluk t, nie] one know never when come your luck not 'One never knows when one is in luck.' [MATTHEE 2000: 468] (17) Hoe weet j'y [hoe lykt my vrou tι ]? how know you how look my wife 'How do you know what my wife looks like?'

[BLOEMHOF 2001: 163]

In the spoken language the order of the corresponding Yes/No question, i.e. with VI, is sometimes attested after the interrogative word o/'whether'. (18) Ek weet nie [of salt I know not whether will Ί don't know whether he'll do it.'

hy he

dit t, it

doen do

nie] not

These changes imply a considerable inroad on the verb final nature of the language from the point of view of general usage. 4.3. Penetration of final verb clusters The clustering of verbs in clause-final position is the result of object raising or other movement, depending on the theoretical point of view adopted. Thus, what is traditionally referred to as the "penetration" (Dutch doorbreking) of final verb clusters, is in fact instances of incomplete raising or partial movement of constituents. Which approach is taken, is immaterial to the present problem. This phenomenon also occurs in Dutch3, but it is likely to occur at least as widely in Afrikaans, if not a lot more. Though a quantitative comparison between the two lan3

According to VAN DER HORST (1997: 300-305) penetration is still common in certain Southern Dutch dialects, but far less common in the North. It seems to have been quite prevalent across the entire speech area in the 16th and 17th centuries, though more consistently in the North than in the South. It therefore might have been part of Afrikaans since its inception, still being maintained in the spoken language.

159

Verb sequence and placement: Afrikaans and Dutch compared

guages is hardly possible at present for lack of sufficient data, an impression can be given of the extent of the phenomenon in present-day Afrikaans. The following are all examples of penetration by the direct object, as well as by a prepositional phrase in (21): (19) toe moes ons weer gaant ander then had.to we again go other 'Then we had to go and buy other glasses.'

glase tt glasses

hoop. buy

(20) Dan sal ek seker maar moet{ then will I probably but have.to 'Then I'll probably have to buy a flat.'

'n woonstel /,· koop. a flat buy

[HK]

[HK]

(21) omdat sy allenig moett 'n ou babatjie in die because she alone must an old baby.DlM into the ' Since she has to give birth to a little baby all by herself

wereld world

bring bring [HK]

Penetration by non-verbal material (or movement of verbs) may result in a stacking of modal verbs towards the beginning of the sentence, as already attested in Middle Dutch (Mariken van Nieumeghen): (22) het it

sal

moeten, int

1

1

will

have.to

eynde i,

wesen

end

be

3

in.the

'It will only be a final resort.'

[JONCKHEERE & CONRADIE 1992: 111 ]

This also occurs in modern Dutch, as indicated by VAN DER HORST ( 1 9 9 7 : 301): (23) Dat zal moeteni. duidelijk 1

1

t{ worden in

de

loop

van

het

3

that will have.to clear become in the course of the volgende jaar. following year 'That will have to become clear in the course of the following year.' In Afrikaans, this usage has become quite common in the spoken language, particularly but not exclusively in the case of sal + moet (or sou + moes*):

4

In Afrikaans several modal verbs may be strung together, the first of which is always sal/sou 'will/ would' (if present), and if the first is preterite, the rest may follow suit, viz. sou moet kan sing 'should by rights be able to sing' > sou moes kon sing. These strengthen the cohesion of the verb sequence. E.g. Die gerusstelling dat sy die donkerte met die geringste roering van haar the consolation that she the darkness with the slightest movement of her hand sou kon verdryf. As sy sou wou. hand would be.able.to dissipate if she would want.to 'The consolation that she would be able to dissipate the darkness with the slightest movement of her hand, should she want to.' [VAN DER VYVER 1999: 403]

160

C. JAC CONRADIE

(24) hulle sal moetj takke onder die wiele /,· pak they will have.to branches under the wheels pack 'They will have to put branches under the wheels.' [VAN ZYL 2001: 47] (25) Jy you

sou moeSj 30 of would have.to 30 or

SO sent op die dollar tt 50 cents on the dollar

vat take

'You would have to accept 30 or 50 cents to the dollar.' [MEYER 2000: 72]

(26) met ander woorde ek moet kant kaartjies tt verkoop, with other words I must be.able.to tickets sell ek moet katij pakkies tj aanneem en wegstuur I must be.able.to parcels accept and away.send 'In other words, I should be able to sell tickets, I should be able to accept and dispatch parcels' [HK] V2, nevertheless, is not filled by two modals. This is indicated by their obligatory interruption by modal particles (if present) such as graag and nog below: (27) maar ek sal graag wilt verder tt gaan daarin but I will gladly want.to further go therein 'But I would very much like to take this further one day' (28) ek sal nog kant stilletjies 'n baba I will still be.able.to quietly a baby Ί would still be able to have a baby on the quiet.'

i,

eendag one.day [HK]

he have [HK]

4.4. Verbal fusion In clauses such as (29) die brandertjies begin komi mos nou tt the wavelets begin come surely now 'The wavelets are surely beginning to come now.'

[SK]

(30) Nee, sy sit [en dink]t alweer tt no she sit and think again 'No, she is in a contemplative mood again.'

[SK]

the verbs appear to be combined in V2 position, not separated by adverbs such as mos and alweer. Sentences (31) to (37) below demonstrate the phenomenon of verbal "fusion" in Afrikaans, which constitutes a salient difference between Dutch and Afrikaans. The term "fusion" is employed here to refer to a stronger syntactic bond than in the case of the clustering described above. Nore, in particular, that no clause-final stranding of the main verb is required in the case of fusion. However, though this "fusion" has morphological implications, the verbs in question always retain the possibility of being used separately, as in (33b) and (36b). Notice the PP in (35b) which is impossible in Dutch and German (infinitivus pro participio, Ersatzinflnitiv).

161

Verb sequence and placement: Afrikaans and Dutch compared

Dutch (31a) Ze bouwen een huis. they build a house 'They are building a house.'

Afrikaans (31b) Hulle bou 'n huis. they build a house 'They are building a house.'

(32a) Ze hebben een huis gebouwd. they have a house built 'They built a house.'

(32b) Hulle het 'n huis gebou. they have a house built 'They built a house.'

(33 a) Ze laten een huis bouwen. they let a house build 'They are having a house built.'

(33b) Hulle laat 'n huis bou. they let a house build 'They are having a house built.'

(34a) Ze hebben een huis laten bouwen. they have a house let build 'They had a house built.'

(34b) Hulle het 'n huis (ge-)laat bou. they have a house let build 'They had a house built.'

(35a) *dateen huis laten bouwen that a house let build moest worden had.to become 'that a house had to be built'

(35b) dat 'n huis (ge-)laat bou that a house let build moes word had.to become 'that a house had to be built'

(36a) Laten ze een huis bouwen? let they a house build 'Are they having a house built?'

(36b) Laat hulle 'n huis bou? let they a house build 'Are they having a house built?'

(37a) *Laten bouwen ze een huis? let build they a house 'Are they having a house built?'

(37b) Laat bou hulle 'n huis? let build they a house 'Are they having a house built?'

In both Dutch and Afrikaans the main verb is realized as a past participle in the presence of the auxiliary verb hebben/het 'have', as in (32). As a result of raising, (33) contains two lexical verbs, the first (i.e. laten/laat) of which may be referred to as a "linking" verb, the second (i.e. bouwen/bou) as the main verb. In the presence of an auxiliary verb, as in (34), the linking verb and main verb are morphologically realized as infinitives in Dutch. As Afrikaans only marks the infinitive in auxiliary verbs, laat and bou are unmarked in regard to finite/infinitive. (For the sake of clarity, what would be morphologically distinct finite verbs, infinitives and imperatives in Dutch, will be glossed as FIN, INF and IMP, respectively, in the Afrikaans examples below.) The prefix ge-, which marks past participles, may be optionally affixed to the linking verb laat. (Ge- would also be optional in the case of a single multisyllabic verb with a rising stress contour.) Moreover, in (35) laat bou scrambles to the left as only a past participle can, suggesting that in Afrikaans the linking verb and main verb may "fuse" into a past participle in the presence of an auxiliary verb. (38) is a corpus example of such scrambling. (38) Mense

wat

laat

weet

moet

people who let.FlN know.lNF must.INF 'People who have to be informed'

word become.INF [BAKER 2001: 232]

162

C. JAC CONRADIE

(36) demonstrates "normal" VI placement in interrogatives, for both languages. (37) demonstrates the placing of both linking and main verb in VI - prohibited in Dutch but common in Afrikaans. In what follows, the phenomenon of fusion will be elucidated with reference to corpus examples. Except for the linking particle en, fused verbs are not separated by lexical elements, be they modal particles, such as dalk 'perhaps', or particles attached to the main verb in the basic sequence, such as oop 'open'. This is demonstrated in (40) and (41). (39) Hy gaan trekt die sluis weer he go.FIN pull.lNF the sluice again 'He goes and opens the sluice again.'

oop tt. open [VAN HEERDEN 2000: 393]

(40a) *Hy gaan dalk trek die sluis weer oop. (40b) Hy gaan trek dalk die sluis weer oop. (41 a) *Hy gaan ooptrek weer die sluis. (41b) Hy gaan trek weer die sluis oop. A verbal particle is stranded clause-finally if fused verbs occur in V2: (42) dan kom gooit ek net so oor 'n halfuur die briewe hier af th thencome.FlN throw.lNF I just so overahalf.hourthe letters here off 'Then I'm dropping off the letters here in about half an hour's time' [SK] Fusion is highly preferred to separation in the case of idiomatic collocations, as in (43)-(45), and even obligatory in (46). (The categorial status of wiel in (46), which could be interpreted as a verb or a nominal, is not clear.) (43) Hy laat staant nie die drank tt nie. he let.FlN stand.lNF not the drink not 'He doesn't stop drinking.' (44) Summier laat weet, jy die prokureurs daardie selfde dag tt straight.away let.FlN know.lNF you the attorneys that same day 'You notify the attorneys straight away, that very same day.' [SK] (45) Reg voor my struikel sy en laat valt die koppies right before me stumble she and let.FlN fall.lNF the cups klaterend flenters ti clattering in.pieces 'She stumbles right in front of me and drops the cups in pieces with a clattering sound.' [VAN ZYL 2001:78] (46) hy laat wielt met daardie twee vingertjies van hom tt he let.FlN wheel.INF/NOM with those two fingers.DiM of him 'He moves speedily with those two little fingers of his.' [SK] A clause may contain more than one linking verb, as in (47) and (48).

163

Verb sequence and placement: Afrikaans and Dutch compared

(47) hy het sy stem soos daai ou in die koffie-advertensie he have.FIN his voice like that guy in the coffee.advertisement probeer laat klink try.PP let.lNF sound.lNF 'He tried to make his voice sound like that guy in the coffee ad.' [MEYER 2 0 0 0 : 3 4 6 ]

(48) wat haar vir die grootste deelvan die rit agter in diemotor laat who her for the greater part of the trip behind in thecar let.PP le en slaap het lie.lNF and sleep.lNF have.FIN 'Who let her lie sleeping in the back of the car for the greater part of the trip' [GREEFF 2001: 56] In sequences containing more than one linking verb, it seems possible that only the linking verbs may fuse, as in (49) (with an extraposed NP). (49) en net daar korn kry Kallie homle die volgende oggend. and just there come.FlN get.iNF Kallie him lie.lNF the next morning 'And Kallie found him lying just there the next morning.' [PHILANDER 2000: 9 2 ]

Yet, the fusion of all verbs involved is the general rule as in (50). (50) Gaan loop vra go.iMP walk.lNF ask.iNF 'Do go and ask her.'

vir for

haar. her [VENTER 2000: 69]

If fusion occurs, the direct (or reflexive) object must follow the main verb, as in (51H53). (51) Die generaal probeer ligt homself nou uit the general try.FlN lift.lNF himself now out 'The general now tried to lift himself out of the car.'

die the

motor tt. car

[VAN HEERDEN 2000: 3 1 0 ]

(52) Die jong meisie begin speelt the young girl begin.FIN play.INF "The young girl begins to play on the piano.' (53) JIP-redakteur gaan maakt nader JlP.editor go.FlN make.lNF closer 'JIP editor makes (their) acquaintance.'

die klavier tt. the piano [NATANIEL 2001: 100] kennist, acquaintance [caption in Beeld, 10/8/02]

The verb gaan 'go' has developed the modal sense of 'intend to'. As a modal verb, it does not take part in fusion, and clause-final stranding of the main verb is required to obtain a modal reading. Thus in (50), (53) and (56) gaan only has the non-modal reading of 'go' and not 'intend to'. Fusion is particularly evident in the slot between a topicalized constituent and the clausal subject:

164

C . JAC CONRADIE

(54) Hier, loop [en dink\ sy met die tuinslang tj, le Kardoes here walk.FlN and think.lNF she with the hose lie.FlN Kardoes '«Here», she thinks as she is walking with the hose, «lies Kardoes».' [GREEFF2001:5] (55) en dan loop roept Felix die and then walk.FlN call.lNF Felix the 'And then Felix goes to call the doctor'

dokter i, doctor [GREEFF 2001: 27]

(56) Agterna gaan sitt sy hulle op verwarmde flesdeksels tt afterwards go.FlN put.lNF she them on heated flask.lids 'Afterwards she goes and puts them on the heated lids of the flasks.' [GREEFF 2001: 60] (57)

Toe kom sleep,· hy mos vlerktj by Sustertjie then come.FlN drag.lNF he actually wing with Sustertjie 'Then, as you may know, he came to court Sustertjie.' [NEL 2001: 76]

(58) bring hulle een Woensdagaand dan kom drinkt bring.lMP them one Wednesday.evening then come.FlN drink.INF julle by ons koffie ti you by us coffee 'Bring them around on a Wednesday evening, then you come and have a cup of coffee at our place.' [SK] In (59) and (60) the NPs following the fused verbs are the underlying subject of the main verb, though not of the linking verb. (59) Hy laat glyt die tekening in die saalsak f, 2 3 he let.FlN slide.lNF the drawing into the saddlebag 'He slides the drawing into the saddlebag.' [VAN HEERDEN 2000: 397] (60) Hy laat gfyj sy tong onthuts oor sy tande tj. he let.FlN slide.lNF his tongue upset over his teeth 'He slides his tongue disconcertedly over his teeth.' [DREYER 2000: 133] The imperative - the context in which verbal fusion may well have arisen5 seems particularly conducive to fusion. If the linking verb is a hortative such as gaan 'go', kom 'come' or loop 'walk, go', fusion is compulsory, as in (61a), unless the subject is overtly stated, as in (62). (61a) Gaan maak ditlnou/koffie in die go.IMP make.lNF it/now/coffee in the '(Now) go and make it/coffee in the house!' (61b) *Gaan dit/nou/koffie in die huis maak\

5

huis! house

PONELIS (1993: 330) surmises that the imperative type "may have served as a launching pad" for this collocation.

165

Verb sequence and placement: Afrikaans and Dutch compared

(62) Gaan jy / julle go.IMP you.SG / you.PL 'You go and make coffee!'

koffie coffee

maak! make.lNF

A great number of imperative examples can be cited, as in (63)-(67). (63) Kom probeert dit sommer nou met my come.lMP try.INF this right now with me 'Come and try it out on me straightaway.' [VERMEULEN 2000: 78] (64) Loop lij jy nou 'n bietjie t{ en probeer go.lMP lie.INF you now a little and try.to.IMP 'Go and lie down a little and try to rest.' (65) Kom haalj my tit ek gaan come.lMP fetch.iNF me I will.FiN 'Come and fetch me; I am sure to forget.'

rus rest [SK]

vergeet forget.lNF [SK]

(66) bring daardie radio en kom sit, horn hier voor tt bring.lMP that radio and come.lMP put.iNF him here in.front 'Bring that radio and put it here, in front.' [SK] (67)

Gaan blyt daarso vir die go.IMP stay.lNF there for the 'Go and stay there for the six years.'

ses six

jaar tt. year [SK]

The main verb in Afrikaans imperatives may attract verbal particles, though not into VI. See (68) and (69). (68)

Gooi (nou) die petrol in! pour.IMP now the petrol in 'Fill it up with petrol now!'

(69) Gooi (nou) m, die petrol tt! This attraction is unacceptable in declaratives and interrogatives, as in (70a) and (71a). (70a) *Hy gooi int die he pour.FlN in the 'He fills it up with petrol.' (70b) Hy gooi die petrol in.

petrol tj. petrol

(71a) *Gooi inj hy die pour.FlN in he the 'Does he fill it up with petrol?' (71b) Gooi hy die petrol in?

petrol t(? petrol

With adjectival complements, attraction is nevertheless possible. See (72) and (73). (72) want sy maak klaar{ die end van die jaar t ; because she make.FlN finish the end of the year 'Because she is finishing (school, etc.) at the end of the year'

[SK]

166

C . JAC CONRADIE

(73) Die meeste ouens in die omgewing maak skoont daai tyd i, the most guys in the vicinity make.FlN clean that time 'Most guys in the vicinity are engaged in cleaning activities at that time.' [SK] Fused verbs attract particles in the same way, as in (74). (74) Kom gooi in, jou een come.IMP throw.INF in your one 'Come and throw in your one bundle!'

bondel bundle

/ (overheard)

As transitive verbs are commonly employed without their direct object in imperative usage, but need to be semantically completed by the particle (e.g. Gooi in!), extraposition of the direct object (as above) seems feasible. The order verb + particle + object is nevertheless a close approximation to the VO order of English. It is also encountered outside the imperative context, as in (75) and (76). (75) Die vroue gaan sluit αα/ι, by SWAPO the women go.FIN join.INF on to SWAPO 'The women go and join SWAPO because ...'

omdat... because... [VAN ZYL 2001: 31]

(76) maar die woorde bly fladder weg/ uit haar dowwe hop tt but the words stay.FlN flutter.lNF away out.of her dull head 'But the words keep on fluttering away from her dull head.' [GREEFF 2001:31] The order in (75) and (76) is explainable as PP-over-V from the similarly acceptable (77). (77) maar die woorde bly fladder uit haar dowwe kop weg, which, without fusion, would have been (78) maar die woorde bly uit haar dowwe kop wegfladder. Thus, processes of fusion and extraposition have combined to create a perfectly natural VO main clause 6 . In sum, verbal fusion, a frequent occurrence not only in spoken but also written Afrikaans, is one of the factors responsible for a general leftward movement of verbs and the weakening of the condition of final verb stranding in the Afrikaans clause. 6

A number of verb + particle expressions, such as hou vas 'hold fast' and gee pad 'give way, get out of the way', have been lexicalized to the extent that they may co-occur in V2 or as past participles in this order, e.g. Ek gee[pad]t nou tt Ί am getting out of the way now', as in (i)-(iii). In (ii) and (iii) verb final stranding would have required padgee and padgegee, respectively. (i) Hy moes relings houvas om af te klim. he had.to hand-rails hold.fast for off to climb 'He had to hold on to hand-rails in order to descend.' [HK] geepad. (ii) oor 'n jaar of twee sal ek seker moet give, way over a year or two will I probably have.to 'In a year or two's time I'll probably have to leave.' [HK] (iii) nadat hulle uit Tanganjika gee pad het after they out. of Tanganjika give way have [BEHR 1993: 2 8 ] 'after they left Tanganjika'

Verb sequence and placement: Afrikaans and Dutch compared

167

5. Conclusion The greater fixation of the sequence of verbal elements in Afrikaans than is the case in Dutch, a corollary of the general deflection of Afrikaans verbs, represents a movement towards a more analytic language type. However, a number of factors, some structural, others related to combinatorial restrictions and usage, contribute to a less restrained spread of verbal elements throughout the Afrikaans clause than would normally be the case in Dutch. Among the factors that deserve special mention are a considerable degree of penetration of clause-final verbal strings (or incomplete verb raising), the frequent usage of VO clauses without conjunction, particularly as object clauses, a growing preference for VO order in otherwise OV subordinate clauses, and the fusion of linking and main verbs into units often appearing in VI or V2 positions without clause-final verb stranding. While the latter phenomena, in particular, are typical of or exclusive to Afrikaans in contrast to Dutch, the language also maintains marked verb-final characteristics with signs of grammaticalization. Thus, in as far as a possible restructuring towards a VO type of language is held in check by the retention of clause-final verbal strings, the rift between an OV basis and VO characteristics, also apparent in Dutch and German, may be said to have widened in Afrikaans. Most of the changes involving leftward verb movement described above are related to increases in the "performative power" of utterances. When the interrogative format of questions is substituted for subordinate clauses as in (15)—(18) above, the sentence as a whole gains directness of expression. Deontic aspects of an utterance are stressed when modal auxiliaries move to the left, as in (24)-(28). If verbal fusion may be derived from directives such as (63)-(67), this may be the clearest link yet between performativity and leftward verb movement. The placement of verbs early in the sentence is iconic in the case of directives as well as questions in that verbs are generally related to "action" and it is a matter of urgency for the speaker to elicit action from the addressee. Abbreviations HK SK DIM FIN IMP INF IRR NOM PL PP SG

Main corpus, see KROES 1982. Spontaneous corpus, see KROES 1982. diminutive finite imperative infinitive irrealis nominative plural past participle singular

168

C . JAC CONRADIE

References BAKER, ELEANOR ( 2 0 0 1 ) : 'n Ou begin. Kaapstad: Human en Rousseau. BEHR, MARK ( 1 9 9 3 ) : Die reuk van appels. Strand: Quellerie. BLOEMHOF, FRANCOIS (2001): 'n Tweede asem vir Jan A. Kaapstad: Human BRINK, ANDRE P. (1998): Duiwelskloof. Kaapstad: Human en Rousseau. DREYER, TOM (2000): Stinkafrikaners. Kaapstad: Tafelberg.

en Rousseau.

Skoon afhanklike sinne in Afrikaanse spreektaal. (Subordinate clauses without complementizers in Afrikaans), in: South African Journal of Linguistics 8(3), 1 1 6 - 1 2 0 . GREEFF, RACHELLE ( 2 0 0 1 ) : Merke van die nag. Kaapstad: Tafelberg. HAESERYN, WALTER J . M . ( 1 9 9 0 ) : Syntactische normen in het Nederlands; een empirisch onderzoek naar volgordevariatie in de werkwoordelijke eindgroep. Doctoral thesis, Nijmegen. JONCKHEERE, WILFREDF. & CONRADIE, C. JAC (eds.) (1992): Mariken vanNieumeghen. Pretoria: HAUM-Literor. KROES, HENK ( 1 9 8 2 ) : Die frekwensiebepaling van die kemwoordeskat en sekere strukture van die Afrikaanse spreektaal. Unpublished research report. Johannesburg: Rand Afrikaans University. MATTHEE, DALENE ( 2 0 0 0 ) : Pieternella van die Kaap. Kaapstad: Tafelberg. MEYER, DEON ( 2 0 0 0 ) : Orion. Kaapstad: Human en Rousseau. NATANIEL ( 2 0 0 1 ) : Tuesday. Cape Town: Human & Roussouw. NEL, ELIAS P. ( 2 0 0 1 ) : Mafoiing. Kaapstad: Tafelberg. PHILANDER, P. J. ( 2 0 0 0 ) : Rebunie. Kaapstad: Human en Rousseau. PONELIS, FRITZ ( 1 9 7 9 ) : Afrikaanse sintaksis. Pretoria: J.L. van Schaik. PONELIS, FRITZ (1993): The development of Afrikaans. (Duisburger Arbeiten zur Sprach- und Kulturwissenschaft 18). Frankfurt am Main: Peter Lang. VAN DER HORST, J. M. (1997): Over en naar aanleiding van Zuid-Nederlandse doorbrekingen, in: VAN FEINAUER, Α . Ε . ( 1 9 9 0 ) :

SANTEN & VAN DER WAL, 2 9 9 - 3 0 7 . VAN DER VYVER, MARITA ( 1 9 9 9 ) : Wegkomkans. Kaapstad: Tafelberg. VANDEWEGHE, WILLY ( 2 0 0 0 ) : Grammatika van de Nederlandse zin. Leuven-Apeldoorn: Garant. VAN HEERDEN, ETIENNE ( 2 0 0 0 ) : Die swye van Mario Salviati. Kaapstad: Tafelberg. VAN ROOYEN, NANETTE ( 2 0 0 1 ) : Om te vlerk. Kaapstad: Human en Roussouw. VAN SANTEN, A. & VAN DER WAL, Μ. (1997): Taal intijden ruimte. Leiden: Stichting Neerlandistiek Leiden. VAN ZYL, DINE (2001): Slagoffers. Kaapstad: Tafelberg. VENTER, EBEN ( 2 0 0 0 ) : Twaalf. Kaapstad: Queillerie. VERMEULEN, JAN ( 2 0 0 0 ) : Geraamtes dra nie klere nie. Kaapstad: Tafelberg.

Reflections on the form and function of passives in English and German HARTMUT CZEPLUCH + (Göttingen)

1. Introduction: sentence and utterance As a starting point it is assumed that any 'sentence' as a grammatical form is at the same time a discourse-functional unit, an 'utterance', that is, a linguistic form determined by contextual factors and/or by the speaker's communicative intention. The 'sentence-utterance' duality implies that linguistic expressions are always subject to two kinds of principles simultaneously: 'grammatical' and 'pragmatic': Grammatical principles would yield, e.g., the SVO order in English; pragmatic principles prefer that 'given information' be arranged before 'new information' called the Theme-Rheme articulation; and furthermore that 'what the utterance is about' be mentioned before 'what is said about it' - called the Topic-Comment articulation. If grammatical and pragmatic ordering requirements run parallel, the double articulation is hardly noticeable. But when the demands at the two layers diverge, the language has to provide for means that resolve or, at least, minimize the conflicting requirements. This is where pragmatic choice among competing linguistic expressions comes in.1 A first demonstration of this effect is provided by the simple active-passive pairs in example (1): (la)

The man kissed a girl Topic < Comment Theme < Rheme (lb) "A man kissed the girl ?Topic < Comment Rheme Theme

(la') "A girl was kissed by the man TTopic < Comment Rheme Theme (lb') The girl was kissed by a man Topic < Comment Theme < Rheme

Simplifying somewhat, definite nominale typically represent discourse-old referents and indefinite nominale typically introduce new discourse referents. This explains why we prefer the active in (la), but the passive in (lb). These forms better comply with discourse-functional preferences. It has often been noticed in the comparative literature that "English tends to evade indefinite subjects" (e.g., KÖNIG Pragmatic choice in the expression of some prepositional content may come in three varieties: via lexical choice (e.g., choosing a periphrastic form over a simplex lexical item); via word order variation (i.e., choosing one grammatically admissible arrangement of sentential elements over another); or via constructional choice (i.e., selecting a particular construction type over the basic simple clause as, e.g., a cleft sentence that splits up one proposition into two informational units).

170

HARTMUT CZEPLUCH

1996; LEGENHAUSEN & ROHDENBURG 1995)2 - which simply means that indefinite subjects, being rhematic normally, do not go well with the principles of Topic be(XP*>

[ w [ v . arrest(AG''PAT,)-edAG the studentsPAT ]]EV [ w [y arrest(AG''PAT,)-edAG the studentsPAT ]] N

1

Why does the ^-complement lose its object Case, while the havecomplement retains it? How comes that the Aave-subject is thematically interpreted as the subject argument of its complement?

The first problem has a simple answer: be is the most neutral verb having no 0nor Case properties and being compatible with almost any complement XP. As it has no SU-0-role, it does not satisfy the EPP; see (16b). Thus, an argument of the 18 19

20

The following is based on ideas developed in CZEPLUCH (1996). Perfective have is most likely related to causative (resultative) have, for which there seem to be diachronic arguments as well (cf. KÖNIG 1995, ABRAHAM 2000). (a) haveCM}S [CAUSER, EVENT ] (b) havepggp [Θ, EVENT ] The situation is similar for German bekommen ('get') that functions as a lexical verb in (a) and as an auxiliary in the recipient passive (b): (a) Hans bekam eine Belohnung 'John got an award' (b) Hans bekam ein Buch geschenkt 'John was given a book' For a G B analysis of the Recipient passive, see, e.g., MOLNÄRFI ( 1 9 9 7 , 1 9 9 8 ) . Causative have comes with active and passive VP complements: (a) TheyCAUS have(CAUS-'EV*> the policeAG [ v . arrest(AG,'PAT') the studentsPAT ]]•* CAUS (CAUS EV,) (b) They have *[ w the students [ v . a r r e s t ^ ' ^ - e d ^ f A T ( b y the police) ] Γ Since the SpecVP position of the complements needs to be lexically filled, causative have is an ECM verb. But perfective have is not (cf. (16a)). Like all auxiliaries, it is not a Case assigner.

Form and fiinction ofpassives in English and German

179

participial VP has to be promoted: This cannot be the SU-argument because it is internally saturated. Therefore an object argument must be promoted to grammatical subject, which is only possible if the participle optionally does not assign Case to it; see (16b'). According to CHOMSKY & LASNIK (1977), rules and principles are optional, i.e., they need not apply whenever possible, but rather apply whenever necessary to guarantee convergence of a derivation. This allows Case assignment by V to be suspended if the NP in question can or must be Case-licensed elsewhere. In this sense, the Case Filter only requires that there must not be an overt NP without Case. This view complies with our assumption that all verbs are (potential) Case assigners (cp. fn. 16 with respect to the Case-assigning properties of true intransitives). In the Aave-structure (16a), the subject position is already θ-indexed, which blocks object movement, and the participle Case-marks the object to satisfy the Case Filter. This brings us to the second question: Let us suppose that an underspecified Θindex, as the unspecific SU-0-role of perfective have, searches for a specific interpretation. That is, it will θ-link with the highest available θ-role of its complement. Since the participle VP does not project the internal SU position SpecVP (otherwise the causative variant would result), this is the V-internally saturated θ-role. In a way then, perfective have is a 'reversed' control verb: It does not 'control' the highest θ-role of its complement, but is 'parasitic' on its specific thematic properties. The θ-chain (Θ1, Agt1) lets one interpret the subject of have as Agent in (16a) and gives structures with perfect aspect a monoclausal character. Comparing the passivization accounts presented here, the major difference seems to be that in standard GB analyses Case absorption plays a central and independent role, even in combination with SU-Θ absorption. The latter approach puts emphasis on SU-Θ absorption, with Case effects as an epiphenomenon. In descriptive terms, whereas the classical analyses treat passivization as a strongly 'object-dependent' process, the latter analysis rather argues for a primarily 'subject-dependent' view of passivization.

5. A comparative view on English and German passives In this section we will consider a couple of differences between German and English passives and evaluate them against the array of functional properties in (4). It should be clear that both German and English passives can realize each of the functions under appropriate contextual conditions. We are interested, though, in the question whether differences in the 'grammatical' properties of passives in the two languages may be indicative of different typical 'functional uses'. 5.1. Passive movement From the heavily English-biased discussion in theoretical linguistics one can easily get the impression that passive formation almost invariably involves move-

180

HARTMUT CZEPLUCH

ment of a post-verbal NP into the grammatical subject position. Exceptions to this quasi obligatory promotion to SpecIP in English are acknowledged only in cases where a passivizable predicate takes a CP complement rather than an NP object. This is almost standardly demonstrated for £>e//e ve-constructions: (17a) They believe [ to have won]

= = = = =

DO-Passive IO-Passive PO-Passive ADV-Passive Acl-Passive

Now, if the passive subject in (24f) is derived from an infinitival subject, this means that it does not need to be in a grammatical relation to the passive participle at 27

Users' grammars often classify Bill in (b) as the Direct Object of believe. But standard users' dictionaries normally characterize believe as a 2-place verb, in the sense of somebody believes something. The latter assumption is the basis for grammatical analysis: With a ίΛαί-clause complement for believe, Bill functions exclusively as the Subject of the complement in (a); and in anaphoric reference, the pronominal it always substitutes for the whole infinitive clause in (b): (a) We believed [that [Bill had won]] (b) We believed [Bill to have won], but Mary didn 't believe it. Recent Minimalist analyses (e.g., LASNK & SAITO 1991, BoScovii: 1997) assume the ECMsubject to raise out of the infinitival clause and be licensed for Case in SpecAgrOP. While there are good empirical reasons for this analysis, it is not clear how such an analysis copes with ex. (b).

184

HARTMUT CZEPLUCH

all. The irrelevance of the GF status of the subjectivized nominal is corroborated if we compare the ΙΟ-Passive (24b) with the help-Passive in (22c): Both objects are diachronically related to former datives that have been syncretized into a common objective Case in English. But only in the latter case the object is also a Direct Object in present-day English. The 'personal' object in a double-object pattern has not (yet?) become a Direct Object in that - except for passivization - it does not partake in grammatical processes typical of Direct Objects; cp., e.g., CZEPLUCH (1982, 1994, 1996) and HUDSON (1992). In particular, the Indirect Object is generally exempt from processes representing Α-bar movements, e.g. (a)-(c): (25 a) *Who did Mary give t the present? = WTi-movement (25b) * It was John that Mary gave t the present = Cleft Formation (25c) *Theysent t the bookyesterday [the girl that had asked for it] = ftavyNRfbcusShift Hence, what is relevant for passivizability in English is not the grammatical relation of the postverbal NP, but that it bears 'objective' Case; and because English has lost relevant morphological Case distinctions, this holds for all V- or Pgoverned nominale. The prepositional passives in (24c-d) fit into this picture: According to KAYNE (1981), a prepositional head in the configuration V [PP Ρ NP] is reanalyzable as a verbal particle [ v V-P] NP, if V and Ρ govern and Case-mark alike in a language. Semantically, V-P reanalysis is licensed if the postverbal NP is able to be conceived of as 'affected by the verb's meaning'; syntactically, V-P reanalysis has the consequence that the postverbal NP becomes susceptible to Case absorption if the complex V is a participle. Thus, 'objecthood' is not quite the right notion to capture the prerequisite for passivization in English. The only thing that seems to be necessary is that there is a nominal immediately following the verb, and this is a purely syntactic matter - in sharp contrast to German, where only the accusative of a Direct Object may be absorbed.

5.4. The subject requirement of passivization English and German also differ with respect to the nature of the SU-0-role absorbed in passives. It has often been noticed that German passivization requires not only that the predicate have an external argument, but that it must, in fact, be an Agent (e.g., ABRAHAM 1995, MOLNÄRFI 1997). Thus, a Non-Agent subject blocks passivization in German, but not in English: (26a) Hans erhielt einen Brief =>*Der Brief wurde von Hans erhalten (= Goal) (26b) John received a letter => The letter was received by John (= Goal) (26c) All the students fear him => He is feared by all the students (= Experiencer) While German seems to have a strict semantic restriction on absorbable subjects in passives, there appears to be a semantic bleaching of the lexical subject properties in English passives28. This is not too surprising because the widening of lexical 28

The English data that JAEGGLI (1986) gives for variable SU-0-roles are not quite clear as his active Goal/Source subjects have an agentive reading also. Even in German there seem to be borderline

Form and function of passives in English and German

185

subjects accessible to absorption correlates with the widening of complement NPs

accessible for promotion to Subject/Topic. All in all, German admits passivization only for activities, whereas English passivizes both activities and events/processes, if other conditions hold. State predicates are excluded in both languages.

6. Which analysis for the passivity differential? The foregoing discussion has revealed several differences between passives in German and English. Although the two relevant properties of SU-0-role absorption and object Case absorption co-occur in both languages, optionally in German, obligatorily in English, they differ as to the triggering role of these properties. If so, we have a twofold distinction between German and English passives; see (27): (27)

-

Passivization:

German

English

level:

lexical

syntactic

trigger:

AGENT

PVNP

German passivization is a lexically triggered process; the relevant lexical property being that the verb selects a SU argument - in fact, an Agent. English passivization, on the other hand, is a syntactically triggered process, the relevant syntactic property is that there is an available postverbal nominal (PVNP).

It is not claimed that passivization in German and English are different formal processes actually. Both languages make use of the same formal features - although to varying degrees of necessity. What we claim, though, is that under a modular approach such cross-linguistically similar grammatical processes may have different grammatical triggers in particular languages. The phenomenon of passivization is a case in point because of its immanent complexity involving lexical (θ-grid & participle formation), morphological (participle formation & Case inflection) and syntactic properties (6-/Case-absorption & movement). It can be expected that such properties have different effects under the modularity of language-specific parameterizations. A second point to be emphasized is that we do not give "functional explanations" for the differences between German and English passives in the sense of explaining formal properties in terms of functional properties.29 It is rather the other way round: Any linguistic system, not being a redundancy free mathematical calculus,

29

cases: Like the English (26c), the Experiencer-verb fürchten 'fear' admits passivization: Er wird von allen Studenten gefürchtet. Do we have to regard it as an agentive verb, then? It makes no sense to say, e.g., that English has lost the V2 property to develop a stricter parallelism between pragmatic Topic-Comment and syntactic Subject-Predicate ordering, yielding a TopicSubject association; or that it has drastically reduced its Case morphology (and other inflection) to develop a syntactisized passivization process.

186

HARTMUT CZEPLUCH

provides grammatical options to some (limited) extent for expressing prepositional content differently. And it is the grammatical variation within and across languages that provides possibilities for associating competing grammatical variants with specific discourse-pragmatic functions. Due to diachronic changes in the grammatical systems of languages, the options of variation may also change and so may the functional load of particular constructions available. On the basis of the widely held opinion that Old English syntax was still quite similar to Modern German syntax, in that Old English was still a V2 and O-V language30 and had a morphological Case system, it is not at all unconceivable that the loss of these two properties in the course of time together with the fact that English has become a V - 0 language31 may well have led to constructions, where the modular interplay of lexical, morphological and syntactic factors is differently balanced. In this sense, passives in German and English may have different grammatical triggers, as the data imply, and we may expect that they serve - or may serve different textual and communicative functions in the two languages. Before addressing this aspect, we should ask how the descriptively motivated differences between German and English passives comply with the formal analyses discussed in section 4. For German 'Subject-dependent' passives, the analysis of section 4.2 seems to be the answer: The participle suffix is assumed only to absorb the SU-8-role of the verb; everything else follows by the interaction of grammatical principles. For English object-dependent passives, the Case-absorption analysis (sect. 4.1) may appear as the better solution. But we are looking for an analysis that can cope with both languages under appropriate parameterization. 30

As against the traditional word-order typology SVO, SOV, VSO etc., languages are more appropriately typologized whether they have V2 characteristics or not and whether they are head-first or head-final: +V2

-V2

V-O

V V

V

O-V

English, Romance languages Japanese, Korean V V Scandinavian languages V V German, Dutch, OE Having lost the V2 property, English is "the odd man out" among Germanic languages. In recent diachronic studies of English, it has been noticed that despite the demise of a morphological Case system in late OE and early ME, it still took at least two centuries for English to lose the Germanic V2 characteristic, hence to become a SVO language, as we know it today. Even in the Early Modern English period (late 15th to early 17th century), English had not completely lost remnants of its former V2 characteristics: Certain lexical verbs still partook in Subject-Verb inversion and could precede the sentential negator not; and certain classes of adverbials, when used sentence-initially, still triggered V movement to second position. The demise of morphological case distinctions, in particular, the loss of the ACC-DAT distinction around 1200 was the time when English changed from an O-V to a V-0 language. It is not clear, though, that Case loss is directly responsible for the word order change, as, e.g., ROBERTS (1997) seems to assume. The loss of case morphology will have an effect on word order 'possibilities'. But it does not necessarily enforce a V-0 order. At least, 'caseless' languages like Dutch and Afrikaans seem to do well with their O-V order (e.g., ABRAHAM 1995; MOLNARFI 1999). Something else must have conspired with the loss of case morphology in English. The reason may be that even as an O-V language, OE allowed extraposition of objects and other complements behind Vfinal fairly freely. So, the V-0 order already existed as an variant. Hence, for the period of 12001400, English appears to have been of the Scandinavian type, combining V2 and V-0 properties. V

Form and function of passives in English and German

187

6.1. Against case licensing of participle suffixes In this respect, one may be skeptical about the implications of the standard Case absorption analysis. First of all, if participle formation is a lexical process, it is not clear that Case absorption should play a role at all. Under word-formation processes syntactic templates may provide an additional Case as in (28) or shift an argument to a vacuous Case position as in (29): (28a) The man kissed the womanACC (28b) The man gave the woman??? a kissACC Der Mann küßte die FrauACC Der Mann gab der FrauDAT einen KußACC (29a) Er legte den SattelACC aufdas Pferd (29b) Er sattelte das PferdACC He put the saddleA0C on the horse He saddled the horseACC And besides intransitives with cognate objects, ECM verbs and resultative small clauses (30) show that Case licensing is not lexically determined but rather a syntactic licensing mechanism: (30a) John ate the dinner mealkCC (30b) John ate [sc the plateACC empty] (30c) *John ate the dinner mealACC [sc the plate^ empty] Thus, it may be preferable not to require the participle suffix to be Case-licensed. Secondly, even if participle formation is regarded as a syntactic process (with the participle suffix and the verb stem having separate projections), Case absorption by the participle suffix is unproblematic only with transitive verbs; the participle suffix of single-Dative verbs (helfen) and intransitives would remain without a licensing Case. 2 6.2. The single-participle analysis for both languages Let us then see how the strong 'syntacticization' and the 'object-dependency' of English passivization may be accounted for under the single-participle analysis (section 4.3). For both languages, we take SU-Θ absorption to be the single independent factor. The differences should fall out as modularity effects. Where the languages differ is in the way they make use of the grammatical Subject-Predicate and the pragmatic Topic-Comment articulations (cf. the syntactic sketches for German and English in sect. 2). Because of its V2 character, German is a Topicprominent language, the Topic-Comment articulation being dissociated from and superseding, the grammatical Subject-Predicate partition. The grammatical Subject-Predicate dichotomy is only weakly expressed morphologically by NOM Case, if at all. The morpho-syntactic SU properties are satisified in situ in SpecVP; cp. schema (31): 32

WANNER ( 1 9 9 1 ) assumes that whenever there is no ACC-feature, it is the NOM-feature of the verb that is absorbed. In fact, DAT-verbs and intransitives do not have a 'grammatical subject' in the passive that would need to be bear ΝΟΜ. But at least in English, the functional projection that licenses NOM under Spec-head agreement may be 'too far away' from the participle or the suffix for Case absorption to work.

188

HARTMUT CZEPLUCH

(31) German:[CT Spec C° Topic V2 C

[ σ Spec

[γρ Spec [ν· ... V° ]] [+Θ] [+nom] NPsu

Topic

If the SU phrase leaves VP, it is for discourse-functional reasons, either bearing a [+th (ematic)] feature, or for being selected as Topic from among the [+th] elements, or in thetic clauses for filling SpecCP. A structural SU position as in English needs not be assumed. English, having lost the Germanic V2 property, has become a Subject-prominent language: it relies on the Subject-Predicate partition for turning prepositional content into a predication. Without Case morphology, English has an abstract NOM-OBJ system in which Case relations are structurally distinguished: a structural 'Subject position' SpecIP, Case-licensed as NOM, and various VP-internal V- or P-govemed positions, uniformly Case-licensed as OBJ. In such a system the Predication Principle (EPP) must be satisfied structurally in SpecIP, which requires that the 'Subject phrase' leave its θ-position in SpecVP and be promoted to SpecIP; cp. (32): (32) Engl.: [α> Spec

C°

[w Spec [+nom] NPsu Topic

1°

...

[yp Spec [v· V° ... ]] [+Θ] isu

If in passives the SU-0-role is absorbed by the participle, there must be another nominal to satisfy the EPP in SpecIP. Such a nominal is not available with intransitives, unaccusatives and weather verbs (s. (22d, f-g)), which cannot satisfy the EPP then. But for the wide range of OBJ-marked NPs in the reach of the participle (s. (24)), the EPP is satisfied if Case-marking is optionally suspended and the nominal is moved to SpecIP. In a sense then, it is the necessity of syntactically satisfying the EPP that gives English passives their appearance of 'object-dependency'.

7. Passivization at work In talking about functional properties of grammatical processes and constructions, one has to cope with one major difficulty: There is rarely - if ever - a neat 1to-1 relationship between the form and the function of linguistic expressions. Thus, we expect that both English and German passives exhibit the range of functions mentioned in ex. (7). But, in view of their language-particular differences, the passive functions should be balanced somewhat differently in the two languages.33 Let's take a look at the 'short passive' first.34 Omission of the active Subject is the common feature in both languages. In German, it is a necessary and sufficient 33

34

Whether this is indeed the case, can be decided only on the basis of text-linguistic or corpus-linguistic investigation. The following remarks are conjectural, therefore. From text counts, we know that passives make up only ca. 15-20% in texts, where the 'short passive' by far outnumbers the 'by-Agent' passive. Numbers may rise, of course, in particular text genres.

Form and function of passives in English and German

189

condition. In English, it is only a necessary condition: It is obligatorily accompanied by preposing a non-subject to Subject position. Thus, in both languages we would expect to find short passives when the Agent of an action should not be mentioned for varying reasons, or for the purpose of presenting the event from an 'impersonal' perspective.35 This use should be found more often in German than in English, though. In English, the obligatory Passive Subject fronting tends to create a TopicComment structure by virtue of the unmarked Subject-Topic association, and this in turn provides a 'discourse link'. The idea, then, is that even short passives serve a textual function in English to a somewhat greater extent than in German, where even the ACC-to-NOM conversion of a Direct Object does not function textually per se (see (19b) and (23b)). The multi-functional character of English passivization may be demonstrated with the short text in (33), the first paragraph in JOHN LYONS' Introduction to Theoretical Linguistics (1968): (33) 1.1.1. Definition of linguistics [1] Linguistics may be defined as the scientific study of language. [2] This definition is hardly sufficient to give the reader any positive indication of the fundamental principles of the subject. [3] It may be made a little more revealing by drawing in greater detail the implications contained in the qualification 'scientific'. [4] For the moment, it will be enough to say that by the scientific study of language is meant its investigation by means of controlled and empirically verifiable observations and with reference to some general theory of language-structure. (LYONS 1968: 1) Clearly, the underlined passives show the 'impersonalization' strategy of scientific texts. At the same time, the text is a nice example of 'topic continuity' hence, of its textual function. In [1], the passive subject takes up the textual theme given in the headline. In [3], the passive pronominal subject continues the topic of [2], which, in turn, is discourse-linked to sentence [1]. And in [4], the passiveinverted fry-phrase in the that-clause links to the qualification 'scientific' in [3].36 For 'short passives', the functions (7.1-3) are relevant, but, I think, in slightly different ways. For 'long passives', we have additionally to consider the functions (7.4—5). Again, these functions will play a role in both languages. But it is likely that they play a more important role in English. If English passivization has the strong syntactic character postulated here, this may mean that passives serve to adapt grammatical ordering to discourse-functional needs - that is the unmarked Theme-Rheme and Topic-Comment orderings (see (1)). And, indeed, examples where English prefers the passive for what is expressed in the active in German are not hard to find. Consider, e.g., the mini-texts in (34), adapted from LEGENHAUSEN

35

36

Apart from this discourse-functional effect, passive is also a means of 'presenting an event from a different perspective' - what is called a 'Figure-Ground' effect according to TALMY (1978) a.o. A comparison of longer passages of German-English texts would show that German passives are not in the same way 'textual' as they are in English; cp., e.g., DOHERTY (1993), OLDENBURG (2000). For the given text, sentence [3] is probably rendered: Das wird etwas deutlicher, wenn ...

190

HARTMUT CZEPLUCH

& ROHDENBURG (1995). Why does English choose the passive, while German prefers the active in this context?

(34a) England beat Germany at the European Football Cup. The winning goal was scored by Alan Shearer. (34b) England schlägt Deutschland bei der Fußballeuropameisterschaft. Das Siegtor erzielte Alan Shearer. As mentioned before, a sentence-initial Subject in English normally functions as the Topic of the utterance, and a Topic normally is a Theme element and discourselinked. According to B L R N E R & W A R D (1998), 'discourse-linking' establishes direct or inferable links to an 'anchor set' in the preceding text: The anchor set is a bounded set of referential options - here with the value football. And it is to this 'anchor set' that the Subject-Topic the winning goal is referentially related.37 The function of the passive in (34a) appears to be twofold, then: it contributes to 'text coherence' by selecting a discourse-linked Topic; and clause-internally, it establishes the pragmatically unmarked 'Theme-Rheme' order by promoting a thematic DO to Subject, i.e. function (7.2), and by postposing a rhematic Subject to the end of the clause; cp. (7.4). Clearly, the same textual and discourse-functional conditions hold in the German example (34b) as well. The difference to English is that German simply does not need a passive in this case to achieve the same discourse-functional effect: It is sufficient to shift the discourse link to the grammaticized Topic position; and the appropriate Theme-Rheme structure follows. It is worth asking why the continuations (35) do not seem quite as felicitous as those in (34): (35a) ... The winning goal [s Alan Shearer scored ] (35b) ... Das Siegtor wurde durch Alan Shearer erzielt. (35a) is the standard Topicalization structure that creates a marked Topic by preposing a constituent to the left of the clause, leaving the clause active otherwise. As to discourse-linking, a pre-sentential Topic seems to be as well as the SubjectTopic in (34a). But the rhematic structure has changed markedly: While in (34a) the focus of assertion is on Alan Shearer, in (35a) it is the verbal predicate that is focused - at least in written language. Hence, the two sentences express quite different assertions. Rhematization of the predicate (i.e. function (7.3)) may be discourse-functionally appropriate in particular situations. But in the given context, asserting of the winning goal that was scored is highly uninformative; whereas it is informative to say who scored it. As to German (35b), its Topic-Comment and Theme-Rheme structure is much the same as that of (34b). So, (35b) is not dispreferred because it is discoursefunctionally inappropriate. But there is also no functional advantage in using the passive variant in this case. And what need not be done should not be done. 37

The passive Subject-Topic in (34a) is thematic qua discourse-linking; and it adds to the given information in that it specifies that England scored exactly one goal more than Germany. But this is not too surprising: According to HALLIDAY (1970), it is the normal situation for Topics - he calls them Themes - to comprise given and new information at the same time - except for simple pronominal links. Generally, the pretext information that a Topic takes up is rarely repeated literally.

Form and function of passives in English and German

191

8. Conclusion It goes without saying that many questions remain open. But perhaps the paper has been successful in showing that among the discourse functions that passives may serve, English and German set different priorities, and that these differences may well be related to different grammatical 'triggers' for the construction. In somewhat greater detail, the major points are: 1. As to the textual and communicative use of passives in English and German, one has to take into account three factors at least: the properties of the process itself, the general syntactic frame of a language, and the discourse-pragmatic functions to be expressed.38 2. The search for the functional range and the functional priorities of grammatical processes and constructions has to be based on the analysis of formal properties. Functional differentiation and specialization arise where the grammars of languages allow for certain variations in the form of linguistic expression. It is these grammatical options that are functionalized for the textual and communicative appropriateness of utterances. 3. Although the form-function relationships of linguistic expressions are manifold, they are by no means arbitrary. Passive is a good example for this. Because of its complexity and its various effects, it can serve a range of functions in languages. But it does not necessarily have the same functional priorities across languages. Different functional priorities correlate with differences in the grammatical properties of processes. For passivization in German and English, these are different triggers on different levels of grammatical analysis. 4. Because of the 'sentence-utterance' duality of linguistic expressions, languages must exhibit regular relationships between the formal properties of sentence structure and the discourse-functional properties of utterances. In this respect, German and English present typologically interesting cases in that they show that languages may differ in the way they relate structural and functional properties: German has 'grammaticized' discourse-pragmatic functions in sentence structure to a high degree; in English sentence and utterance properties are rather correlated as two levels of simultaneous processing. In a way, then, pragmatic predication (Topic-Comment) overrides grammatical predication (Subject-Predicate) in German, whereas in English grammatical predication plays the prominent role. It is this distinction that plays an important role in the formal and functional differences between German and English passives, modulo the language-particular parameterizations of the V2 property, the Head Parameter and the Case system.

38

Other factors not discussed here, but to be considered for a more complete picture, would be: the Focus-Background articulation as a 3rd layer of discourse-functional structuring; the Figure-Ground properties as determined by verb diathesis, contributing to particular 'perspectivizations' of the content of propositions.

192

HARTMUT CZEPLUCH

Abbreviations ACC AG AUX CAUS DAT EV

LBr NOM

accusative agent auxiliary causative dative evidential Left Bracket nominative

NP OBJ PAT PERF

PrF RBr SBJ

V VP

noun phrase object patient perfective Prefield Right Bracket subject verb verbal phrase

References (1995): Deutsche Syntax im Sprachenvergleich. (Studien zur deutschen Grammatik 41).Tübingen: Narr. ABRAHAM, WERNER ( 2 0 0 0 ) : Überlegungen zum Passiv im Deutschen und anderen Sprachen: 'Argumenthypothese' und 'Aspekthypothese', in: ZASPapers in Linguistics 15 (2000), 1-35. ABRAHAM, WERNER & MOLNÄRFI, LÄSZLÖ ( 2 0 0 2 ) : The German clause under discourse functional weight: Focus and Antifocus, in: ABRAHAM, WERNER & ZWART, CORNELIUS JAN-WOUTER (eds.), Issues in formal German(ic) typology. (Linguistik Aktuell/Linguistics Today 45). Amsterdam/ Philadelphia: John Benjamins, 1-44. BAKER, M. (1988): Incorporation. A theory of grammaticalfunction changing. Chicago: University of Chicago Press. BAKER, M.; JOHNSON, K. & ROBERTS, I. (1989): Passive arguments raised, in: Linguistic Inquiry 20, ABRAHAM, WERNER

219-251. BEHAGHEL, OTTO ( 1 9 0 9 ) :

Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern, in: Indogermanische Forschungen 25, 110-142. BEHAGHEL, OTTO ( 1 9 3 0 ) : Von deutscher Wortstellung, in: Zeitschrift fir Deutschkunde 4 4 , 8 1 - 8 9 . BIRNER, BETTY J. & WARD, GREGORY L. ( 1 9 9 8 ) : Information status and non-canonical word order in English. Amsterdam: J.Benjamins. BoScoviC, 2ELJKO (1997): The syntax of nonfinite complementation: an economy approach. (Linguistic Inquiry Monograph 32).Cambridge/MA: MIT Pr. BRÖMSER, BERND ( 1 9 8 1 ) : Funktionale Satzperspektive im Englischen. ( T B L 171). Tübingen: Narr. BÜRING, DANIEL ( 1 9 9 4 ) : Mittelfeldreport V; in: HAFTKA, BRIGITTA (ed.), Was determiniert Wortstellungsvariation? Opladen: Westdeutscher Verlag, 79-96. BURZIO, LUIGI ( 1 9 8 6 ) : Italian syntax. A government-binding approach. Dordrecht: Kluwer. CHOMSKY, NOAM (1981): Lectures on government and binding. Dordrecht: Foris. CHOMSKY, NOAM & LASNIK, HOWARD ( 1 9 7 7 ) : Filters and Control, in: Linguistic Inquiry 8 , 4 2 5 - 5 0 4 . CZEPLUCH, HARTMUT (1982): Case theory and the dative construction, in: The Linguistic Review 2 , 1 - 3 8 . CZEPLUCH, HARTMUT ( 1 9 9 4 ) : Zur Doppel-Objekt-Konstmktion im Englischen und Deutschen, in: HAFTKA, BRIGITTA (ed.), Was determiniert Wortstellungsvariation? Opladen: Westdeutscher Verlag, 97-112. CZEPLUCH, HARTMUT ( 1 9 9 6 ) : Kasus im Deutschen und Englischen. (LA 3 4 9 ) . Tübingen: Niemeyer. CZEPLUCH, HARTMUT ( 2 0 0 0 ) : Deutsch als pragmatisierte Sprache: Zum Verhältnis von grammatischen und funktionalen Eigenschaften, in: HESS-LÜTTICH, ERNEST W . B. & SCHMITZ, H . WAL-

TER (eds.), Botschaften verstehen - Kommunikationstheorie und Zeichenpraxis: Festschrift für Helmut Richter. Frankfurt: Lang, 29-45. DOHERTY, MONIKA ( 1 9 9 3 ) : Parametrisierte Perspektive, in: Zeitschrift für Sprachwissenschaft 1 2 , 3 - 3 8 . EISENBERG, PETER ( 1 9 8 6 ) : Grundriß der deutschen Grammatik Stuttgart: Metzler. ENGEL, ULRICH ( 1 9 8 8 ) : Deutsche Grammatik Heidelberg: J.Groos. EROMS, HANS-WERNER ( 1 9 8 7 ) : Passiv und Passivfunktionen im Rahmen einer Dependenzgrammatik, in: BOBILLON, JEAN MARC et al. (eds.), Das Passiv im Deutschen. Tübingen: Niemeyer, 7 3 - 9 5 . GIVÖN, TALMY ( 1 9 9 0 ) : Syntax. A functional-typological introduction, vol. 2 . Amsterdam: Benjamins.

193

Form and function ofpassives in English and German GIVÖN, TALMY ( 1 9 9 3 ) :

English grammar: a function-based introduction, vol.

2.

Amsterdam: Benja-

mins. et al. (1977): English grammar - a University handbook. Leipzig: Enzyklopädie. GREEN, G . (1974): Semantics and syntactic regularities. Bloomington: Indiana UP. HAFTKA, BRIGITTA ( 1 9 8 8 ) : Linksverschiebungen - Ein Beitrag zur Diskussion um die Konfigurationalität des Deutschen, in: BIERWISCH, MANFRED; MÖTSCH, WOLFGANG & ZIMMERMANN, ILSE (eds.), Syntax, Semantik und Lexikon. (Studia Grammatica X X I X ) . Berlin: Akademie Verlag, GRAUSTEIN, GOTTFRIED

89-145.

HAIDER, H. (1993): Deutsche Syntax - generativ. Tübingen: Narr. HALLIDAY, MICHAEL Α . K . ( 1 9 7 0 ) : Language structure and language function, in: LYONS, JOHN (ed.), New horizons in linguistics. Harmondsworth: Penguin, 1 4 0 - 1 6 3 . HOEKSTRA, TEUN (1984): Transitivity: grammatical relations in government-binding theory. Dordrecht: Foris. HUDSON, R. ( 1 9 9 2 ) : So-called 'Double Objects' and Grammatical Relations, in: Language 6 8 , 2 5 1 - 2 7 6 . JAEGGLI, OSVALDO ( 1 9 8 6 ) : Passive, in: Linguistic Inquiry 17, 5 8 7 - 6 2 2 . JESPERSEN, OTTO ( 1 9 2 7 ) : A modern English grammar on historical principles, Part III. London: Allen & Unwin. JESPERSEN OTTO ( 1 9 3 3 ) : Essentials of English grammar. London: Allen & Unwin. KAYNE, RICHARD S. (1981): On certain differences between French and English, in: Linguistic Inquiry 12, 349-371. KÖNIG, EKKEHARD (1995): On analyzing the tense-aspect system of English: a state-of-the-art report, in: RIEHLE, WOLFGANG & KEIPER, H. (eds.), Anglistentag 1994 Graz. Tübingen: Niemeyer, 153-169. KÖNIG, EKKEHARD ( 1 9 9 6 ) : Kontrastive Grammatik und Typologie, in: LANG, EWALD & ZIFONUN, GISELA (eds.), Deutsch - typologisch. (Jahrbuch Institut für Deutsche Sprache 1 9 9 5 ) . Berlin: de Gruyter, 3 1 - 5 4 . KOOPMANN, H . & SPORTICHE, DOMINIQUE ( 1 9 9 1 ) : The position of subjects, in: Lingua 8 5 , 2 1 1 - 2 5 8 . LASNIK, HOWARD & SAITO, MAMORU ( 1 9 9 1 ) : On the subject of infinitives. Repr. in: LASNIK, HOWARD: Minimalist analysis. Oxford: Blackwell, 5-24. LEGENHAUSEN, LIENHARD & ROHDENBURG, GÜNTER ( 1 9 9 5 ) : Kontrastivierung ausgewählter Strukturen im Englischen und Deutschen, in: AHRENS, RÜDIGER; BALD, WOLF-DIETRICH & HÜLLEN, WERNER (eds.), Handbuch Englisch als Fremdsprache. Berlin: E. Schmidt, 1 3 3 - 1 3 9 . LYONS, JOHN ( 1 9 6 8 ) : Introduction to theoretical linguistics. Cambridge: C U P . MOLNARFI, LÄSZLÖ ( 1 9 9 7 ) : Kasus und Passivierung - Ein Beitrag zur Kasustheorie. Frankfurt/M.: Lang. MOLNÄRFI, LÄSZLÖ ( 1 9 9 8 ) : Kasusstrukturalität und struktureller Kasus - zur Lage des Dativs im heutigen Deutsch, in: Linguistische Berichte 1 7 6 , 5 3 5 - 5 8 0 . MOLNÄRFI, LÄSZLÖ ( 1 9 9 9 ) : Zur Diskurskomponente in der Sprechsprache - vir als Rhemamarkierer im Afrikaans, in: ABRAHAM, WERNER (ed.), Characteristic properties of spoken Vernaculars (Special Issue of Folia Linguistica), 7 5 - 1 0 0 . MOLNÄRFI, LÄSZLÖ ( 2 0 0 2 ) : Focus and antifocus in modern Afrikaans and West Germanic, in: Linguistics AO/6, 1 1 0 7 - 1 1 6 0 . OEHRLE, RICHARD ( 1 9 7 6 ) : The grammatical status of the English dative alternation. PhD Dissertation. Cambridge/MA: MIT (unpublished) OLDENBURG, ANTJE ( 2 0 0 0 ) : Form und Funktion des Passivs in deutschen und englischen Romanen und ihren Ubersetzungen. Frankfurt/M.: Peter Lang. ROBERTS, IANG. (1987): The representation of implicit and dethematized subjects. Dordrecht: Foris. ROBERTS, IAN G . ( 1 9 9 7 ) : Directionality and word order change in the history of English, in: KEMENADE, Ans VAN & VINCENT, N . (eds.), Parameters of morphosyntactic change. Cambridge: CUP, 3 9 7 - 4 2 6 . SCHOENTHAL, GISELA ( 1 9 8 7 ) : Kontextsemantische Analysen zum Passivgebrauch im heutigen Deutsch: Zur Mitteilungsperspektive im Passivsatz, in: BOBILLON, JEAN MARC et al. (eds.), Das Passiv im Deutschen. Tübingen: Niemeyer, 1 6 1 - 1 7 9 . TALMY, LEONARD (1978): Figure and ground in complex sentences, in: GREENBERG, JOSEPH H . ; FERGUSON, CHARLES & MORAVSCIK, EDITH (eds.): Universals of human language, vol. 4. Stanford: Stanford UP, 625-649. WANNER, A. (1991): Modulare Aspekte des englischen Passivs. MA Thesis, Univ. Göttingen.

The upper functional domain in Yiddish MOLLY DIESING (Ithaca)

1. Introduction Since POLLOCK (1989), much work on clausal syntax has taken the approach of splitting functional domains into distinct functional heads for each feature. POLLOCK'S paper focuses on dividing the head INFL(ection) into its tense and agreement components, while RLZZI's (1997) analysis splits the CP domain into topic and focus. CINQUE (1999) develops a theory of adverb placement in which adverbs are located in the specifiers of functional categories, claiming that the resulting hierarchy of functional heads is universal. In recent work, RIZZI (forthcoming) has referred to this overall trend in clause structure as the "cartographic approach", in that its goals are to provide explicit and precise maps of syntactic configurations. On the other hand, there have been those who have considered the possibility that the manner in which languages deploy functional heads is not in fact universal, but may instead be parameterized, with some languages opting (for example) for inflectional heads which host more than one feature. In other words, there are some fundamental questions to be considered: Do apparently distinct syntactic positions (as reflected in linear order) necessarily correspond to distinct projections in the functional domain? Is the structure of the upper functional domain (e.g. the IP and CP "territory") uniform across languages? Indeed, IATRIDOU (1990) formulated an early response to POLLOCK'S original proposal raising these very questions, with the conclusion that Agreement) projections may not in fact be universally motivated. Among others who have investigated this matter are VAN GELDEREN (1993), who takes a diachronic view of functional structure, BOBALJIK & THRAINSSON (1998), who argue for parameterization

of the structure of INFL to account for distinctions between VO and OV Germanic languages, and LAMBOVA (2001) who links the structure of the CP domain in Bulgarian to its discourse-oriented nature. Approaching the issue from yet another angle, ABRAHAM (1997) calls into question the basic assumption underlying the Minimalist Program (CHOMSKY 1995) that derivational functional categories (linked to case checking and agreement) are even necessary in all languages (though CHOMSKY himself questions the necessity of Agr phrases). In this paper I investigate the structure of the "left periphery" in Yiddish in light of these various proposals. I will limit my focus to the phenomena of verb second, wA-movement, and topicalization. Space will not permit me to attempt to resolve the major theoretical questions that arise (and there are many!); I will simply concentrate on situating Yiddish within the larger context of the study of clause structure.

196

MOLLY DIESING

2. Verb second in Yiddish Like the other Continental West Germanic languages, Yiddish exhibits the wellknown verb second property. Unlike Dutch and German, however, Yiddish shows no main/embedded asymmetry with respect to verb second; the verb appears in the second position in clauses embedded below a complementizer as well. The following examples show the verb second effect in main clauses, with the finite verb in second position being preceded by a topicalized constituent - whether it be the direct object, an adverbial, or the subject NP. Note also that unlike Dutch and German, Yiddish is VO, see DIESING (1997). (la) (lb) (lc)

dos bukh hot maks geleyent. the book has Max read 'The book, Max has read.' nekhtn hot maks gezungen a lidl yesterday has Max sung a song 'Yesterday, Max sang a song.' maks vet zingen a lidl Max will sing a song 'Max will sing a song.'

The examples here demonstrate Yiddish is a "symmetric" verb second (V2) language, in that the V2 order also holds for embedded clauses: (2a) ikh veys nit tsi dos bukh hot maks geleyent. I know not whether the book has Max read Ί don't know whether Max has read the book' (2b) es iz a shod vos afile LGB hot maks nit geleyent. It is a shame that even LGB has Max not read 'It is a shame that Max has not even read LGB.' (2c) zi iz gekumen zen ver frier vet kontshen. she is come see who earlier will finish 'She has come to see who would finish earlier.' (2d) ir zolt visn zayn, mayne libe kinderlekh, az vayn you should know be my dear children that wine ken men makhn fun troybn oykh can one make from grapes too. 'You should know, my dear children, that one can make wine from grapes also.' It is clear from the examples that the embedded verb second is not restricted to verbs of saying, as embedded main clauses are in many languages. It even occurs in factive complements (2b) and co-occurs with embedded questions (2c). Furthermore, these examples demonstrate that the "classical" analysis of V2 as developed for German and Dutch (beginning with BACH 1962, BLERWLSCH 1963, KÖSTER 1975, and many others since), in which the complementizer and the finite verb compete for the same slot in the head of CP, is not viable for Yiddish, since the complementarity of distribution of the finite verb and the complementizer which motivates the classic analysis does not hold.

197

The upper functional domain in Yiddish

One possible explanation for this phenomenon of symmetric V2 is that Yiddish exhibits CP-recursion in embedded clauses, allowing for both the co-occurrence of V2 with complementizers and embedded questions (VlKNER 1995). (3)

CP-recursion: [Cpi aζ

[Cp2 XP Vfin... ]]

However, in other languages which allow CP-recursion, these structures display the properties of embedded main clauses, which is not the case for the Yiddish examples (see IATRIDOU & KROCH 1992 for an overview). This fact greatly weakens the CP-recursion proposal, and I will not consider it further here. An alternative, which I proposed in DlESlNG (1990), is to allow the IP-domain to host either a subject or a fronted topic XP. This idea has also been developed for Icelandic by RÖGNVALDSSON & THRÄINSSON (1990) and for Mainland Scandinavian by REINHOLTZ (1989), contra BRANIGAN (1996). (4)

Shared IP: [CP C0 [n- XP/Subj Vfm

[vp···]]]

While in its most basic form this approach advocates a single functional head which can license either a fronted subject NP or a non-subject XP, variants of this strategy exploit the split IP analysis of POLLOCK (1989). For example, the landing site of a sentence-initial subject might be Spec, AgrSP, while a non-subject (A-bar) topic might land in the higher Spec,TP; see THRÄINSSON (1994) for discussion. Taking this tack would allow one to maintain the connection between AgrS and Nominative case, but see ABRAHAM (1997) for an alternative view regarding this derivational assumption. I will simply focus on the IP "zone" as the landing site for XP-fronting in V2 clauses as a phenomenon of the left periphery, without pursuing any possible links to case licensing. A consequence of this approach is that main clauses in Yiddish have a different structure from main clauses in asymmetric V2 languages (such as German), in which all V2 results from movement to CP. Furthermore, it can be shown that within Yiddish, matrix questions and embedded questions have a different structure. Evidence for this comes in part from the fact that they display different word orders. Matrix questions show the expected V2 word order, with the wA-phrase in first position: (5a)

hot [VP has (5b) [IP veri hot [vp ti who has 'Who ate a turnip?'" [.r

VOSj

what

maks Max gegesn eaten

gegesn tj]? eaten a brukve]]? a turnip

Embedded questions have a V3 order, with the wA-phrase preceding both the topic and the finite verb: (6a)

ikh veys nit [Cp vost [iP maks I know not what Max Ί don't know what Max ate.' (6b) ikh veys nit [CP ver [n· es hot I know not who EXPL has Ί don't know who has eaten a turnip.'

hot has

gegesn i,.]] eaten

gegesn eaten

a brukve.]] a turnip

198

MOLLY DIESING

These examples show that in an embedded question, there is no finite verb movement to the position to the immediate right of the wA-phrase, unlike in the matrix clause (embedded direct questions are possible in Yiddish under certain restricted circumstances, but I will not address those here, see SANTORINI 1995: 78-83). In addition, with embedded subject extraction (as in (6b)) a mysterious expletive es 'it' appears (glossed here and elsewhere as EXPL), which cannot appear in matrix questions. These facts can be explained if the embedded wA-phrase moves to Spec,CP (as expected given the effects of selection on embedded interrogatives - the embedding verb selects for a +Wh CP). In essence then, what I have called the shared-IP analysis is a case of "generate only as much structure as you need" (DIESING 1990: 55). (I should mention here that GRIMSHAW 1997 develops an analysis of clause structure which also exploits the idea of minimizing structure, though within a rather different theoretical framework.) A remaining issue is the matter of why there cannot be both a wA-phrase and a topic in a matrix clause, with IP projecting multiple specifiers. One possible answer is that while the EP-level specifier can license either a topic or a wA-phrase, it can only check one such feature. That is, it can host only one feature in any given clause. PPA-phrases and topics can co-occur in embedded clauses, since in this case the wA-feature is selected in CP by the embedding verb, leaving the IP-specifier free to host the topic feature. Thus, here we have our first departure from the strict "cartographic" approach to the functional domain espoused by RlZZl (1997). My claim is that the IP domain can host either a wA-feature or a topic feature in matrix clauses, but not both simultaneously. This gives us the result that we do not see topics in matrix questions. Embedded questions result from wA-movement to the specifier of CP (as a result of selection), and thus are structurally compatible with co-occurring topicalization. Allowing the specifier of IP to host both wA-features and the feature which licenses topics may seem a bit odd, given that wA-phrases are generally considered to be focus, in the sense of being "new information", while topics are traditionally taken to instantiate "old information". Pertinent to this apparent contradiction is the study of the discourse function of XP-fronting (what is commonly referred to as topicalization in V2 languages) in Yiddish undertaken by PRINCE (1999). In a study of OVS orders in Yiddish, PRINCE demonstrates that the XPs in initial position strictly speaking do not function as topics in Yiddish, but in many cases serve to instantiate a form of focus movement (specifically, when the fronted XP receives tonic stress), or establish that the fronted XP is in a salient partially-ordered set ("poset") relation to a previously evoked set of entities in the discourse, and interacts with a focus frame derived from the clause (in this case the tonic stress falls within the clause). In neither of these cases does the fronted XP serve as the topic of the sentence in the classic sense. While space does not allow me to present a full discussion of PRINCE'S (1999) findings, I will take them as providing support for the plausibility of the syntactic analysis I offer here, in that the fronted XPs and fFA-phrases may well have enough featurally in common to target the same landing site (LAMBOVA 2001 makes a similar proposal regarding Bulgarian). Thus, I will assume that IP can check whfeatures and focus features, as well as the topic features traditionally assumed to be

199

The upper functional domain in Yiddish

housed there. An obvious question for further research is that of whether this three-way distinction can be unified in some fashion, either conceptually as with LAMBOVA's ( 2 0 0 1 ) ΔΡ (where Δ is a mnemonic for "discourse-oriented") or through an explicit system relating discourse features and syntax such as that developed by ABRAHAM ( 1 9 9 7 ) or MOLNÄRFI (this volume). Furthermore, movement of a wA-phrase will have to take precedence over XP-fronting, since Yiddish does not allow all wA-phrases to remain in situ, but this problem exists even in the classical analysis of V2, in which "topics" and wA-phrases target the same landing site. More importantly for current considerations, the constraint against having both a PFA-phrase and a fronted XP in the specifier of IP can be reduced to the fact that IP can license only one specifier. A bit more clarification regarding the nature of this focus movement is in order, however. In DIESING ( 1 9 9 7 ) I presented evidence for a preverbal focus position to which a single constituent can move. This is rather different from the sentenceinitial XP-fronting, both syntactically and interpretively. Interestingly, an in-situ wA-phrase must move there (I use here an extended sense of the phrase in situ, meaning 'not fronted'), and in declaratives a single XP can occupy the preverbal slot (with an intonational focus): (7a) ver vet haynt vuhin geyn who will today where go 'Who will go WHERE with you today?' (7b) ver hot nekhtn vos who has yesterday what 'Who bought WHAT yesterday?' (7c) maks hot nekhtn a bukh Max has yesterday a book 'Max read a BOOK yesterday.'

mit aykhl with you? gekoyff! bought geleyent. read

An additional fact is that is not possible for more than one constituent to appear immediately before the verb and to the right of an adverb: (8a)

*nekhtn yesterday (8b) *nekhtn yesterday

hot had hot had

maks Max maks Max

nit not nit not

dem yingl the boy ken yingl no boy

dos bukh the book ken bukh no book

gegebn given gegebn given

Furthermore, the interpretation of the XPs in the preverbal focus position is that of what has been called contrastive focus, as opposed to the informational focus (also called "semantic focus" by GUNDEL 1999) represented by XP-fronting to the sentence-initial position. Both types of focus represent "new information", but the preverbal contrastive focus is distinguished from XP-fronting in that it selects new information from a presupposed set, see ROOTH (1992), rather than simply filling in new information. Thus, there is no inherent conflict (either in syntactic or interpretive terms) between the existence of the preverbal focus position and my following PRINCE (1999) in regarding XP-fronting as also at times instantiating a form of focus movement, nor does the preverbal focus position represent evidence contradicting the Shared-IP proposal. In the next section I take a closer look at the syntax

200

MOLLY DIESING

of wA-questions in Yiddish, which will require examining the next layer of the upper functional domain - that of CP. 3. ^-movement in Yiddish Beyond the word order contrasts between matrix and embedded questions noted above, the most striking feature of question formation in Yiddish arises with multiple questions - questions in which there is more than one wA-phrase. Multiple questions in Yiddish (in particular, the Southeastern dialect - see DLESING (2003) for discussion of dialectal differences with respect to multiple questions) come in two forms: both fronting of a single wA-phrase and multiple wA-fronting are allowed. The following examples illustrate the single fronting option, in which one whphrase fronts and the other remains in situ: (9a) ver vet vuhin geyn mit aykhl who will where go with you? 'Who will go where with you?' (9b) ver hot vos gekoyff who has what bought 'Who bought what?' (9c) ver geyt vuhin? who goes where 'Who is going where?' These sentences contrast with examples demonstrating multiple fronting, where both wA-phrases move to the position preceding the finite verb: (10a) ver vuhin vet geyn mit who where will go with 'Who will go where with you?' (10b) ver vos hot gekoyff. who what has bought 'Who bought what?' (10c) ver vuhin gey ft who where goes 'Who is going where?'

aykhl you [MARK 1978: 380]

[ZARETSKI 1929]

As I noted in DLESING (2003), these two orders for multiple questions are synonymous. In particular, the wh-in-situ in the single-fronting cases need not be D(iscourse)-linked (in the sense of PESETSKY 1987 - the potential answers to these questions are not necessarily limited to sets of entities previously defined in the discourse), as has been claimed to be a condition on wh-in-situ in multiple-fronting languages like Bulgarian and Polish (PESETSKY 2000). A fundamental question in the syntax of multiple fronting in Yiddish is that of where the multiply-fronted wA-phrases move to. Is the landing site CP (as argued for Bulgarian by RUDIN 1988 and others), or some lower functional projection such as IP (as RUDIN has claimed for Polish)? In DlESING (2003) I examined evidence

201

The upper fiinctional domain in Yiddish

from the basic the word order facts in embedded questions, including the expletive that appears with subject extractions. Consider the following examples, which demonstrate that both single fronting (11a) and multiple fronting (1 lb-c) are possible in an embedded question. (11a) hot zi ni(sh)t gekent farshteyn ver es shlogt zikh mit vemen has she not could understand who EXPL hits self with whom (lib) hot zi ni(sh)t gekent farshteyn ver mit vemen es shlogt zikh has she not could understand who with whom EXPL hits self '(So) she couldn't understand who was fighting with whom.' [JACOBS et al. 1 9 9 4 : 4 1 4 ]

(11c) lomir geyn, ver vuhin es gsyt. let's go who where EXPL goes 'Let's go where(ever) who(ever) is going.'

[ZARETSKI 1929]

Each of these examples involves wA-extraction from the subject position, leaving a subject gap. As noted in section 2 , an expletive es 'it' is required (BIRNBAUM 1 9 7 9 , TRAVIS 1 9 8 4 , DIESING 1 9 9 0 ) . The expletive does not (indeed, cannot) appear if a non-subject has been topicalized (for example, in (12), the adverb frier 'earlier', or in (13) haynt 'today'), occupying the [Spec,IP] position (as discussed above in section 2, also see DIESING 1 9 9 0 for arguments justifying this placement of non-subject topics): (12) zi iz gekumen zen ver frier vet she is come see who earlier will ' She has come to see who would finish earlier.'

kontshen. finish [DIESING 1990: 50]

(13) lomir geyn, ver vuhin haynt geyt. let's go who where today goes 'Let's go where(ever) who(ever) is going today.' To summarize the point, the expletive does not appear when there is no gap in [Spec,IP], but when such a gap does exist, the expletive is obligatory. This indicates that the landing site for the wA-phrases is at the CP level, for if the fronted wA-phrases were able to land in [Spec,IP] (as in RUDLN's analysis of Polish), this obligatory appearance of the expletive in the examples in (11) would be totally unexpected, as there would then be no gap. The appearance of the expletive in the multiple fronting cases demonstrates that the multiply fronted wA-phrases are not adjoined to IP, but rather are attached higher up, at the CP level, as in Bulgarian. Furthermore, the fronted wA-phrases form a unit which cannot be broken up (i.e. by a parenthetical), another property which Yiddish shares with Bulgarian (RUDIN 1988): (14)

*ver, who

nokh after

dayn your

meynung, opinion

vuhin where

vet will

geyn? go

This fact too is consistent with the conclusion that Yiddish multiple fronting involves movement to CP. To account for multiple wA-fronting, I will maintain here the proposal I made in DIESING ( 2 0 0 3 ) : Yiddish allows for selection of either [-multiple] or [+multiple] CP. The former allows checking of only one wA-feature, and hence fronting of only

202

MOLLY DIESING

one wA-phrase. The latter allows checking of multiple wA-features, leading to multiple fronting (with resulting multiple specifiers). The [+multiple] CP occurs in both embedded and matrix questions, the [-multiple] CP only in embedded contexts; in matrix single fronting questions the [Spec,IP] is the landing site for whmovement (following the principle of generating "only as much structure as is needed"). Interestingly, comparison of the syntax of multiple fronting with single fronting questions provides additional evidence for the structural difference between matrix and embedded questions proposed in section 2. Although they are semantically identical, the two question strategies do display some important syntactic differences. Most notable of these concerns the phenomenon of superiority. While single-fronting questions allow an object wA-phrase to be fronted over an in-situ subject wA-phrase (thus violating the superiority constraint, which requires that the whphrase that is highest in the tree to move), this is not possible in the multiple-fronting case. As HÖGE (2000) demonstrates in her study of superiority effects in Yiddish and other languages, this immunity to superiority in single fronting is quite general, covering all the possible configurations. Thus, in multiple questions involving both the subject and the object, either wA-phrase can front: (15a) ver, hot U vos who has what 'Who bought what?' (15b) vos, hot ver U what has who 'Who bought what?'

gekoyffl bought gekoyffl bought

Nor are superiority effects seen in interactions between object wA-phrases with wA-adverbials: (16a) vos hot maks vi azoy/farvos what has Max how/why 'How/why did Max play what?' (16b) vi azoy/farvos hot maks vos how/why has Max what 'How/why did Max play what?'

geshpilt! played [HÖGE 2000: 286] geshpiltl played

Multiple questions involving a subject wA-phrase and a wA-adverbial also fail to show any superiority effects: (17a) ver hot vi azoy/farvos geshpilt pyane? who has how/why played piano 'Who played the piano how/why?' (17b) vi azoy/farvos hot ver geshpilt pyane? how/why has who played piano 'Who played the piano how/why?'

[HÖGE 2000: 286]

There are also no "pure" superiority effects, in the sense of involving two whphrases in object positions (making it clear that superiority cannot be reduced to a requirement that traces appear in governed positions).

203

The upper fiinctional domain in Yiddish

(18a) vemen hot der lerer vos geheysn leyenen? whom has the teacher what ordered read 'Who did the teacher tell to read what?' [HÖGE 2000: 287] (18b) vos hot der lerer vemen geheysn leyenen"? what has the teacher whom ordered read 'Who did the teacher tell to read what?' The single fronting examples contrast strikingly with their multiple-fronting counterparts, in which superiority must be respected. Thus, we see that while the subject wA-phrase can precede the direct object w/z-phrase in the initial group of wA-phrases, the reverse order is ungrammatical: (19a) ver vos hot gekoyff! who what has bought 'Who bought what?' (19b) *vos ver hot gekoyff. what who has bought (19c) ikh veys nit ver vos es I know not who what ES Ί don't know who bought what.' (19d) *ikh veys nit vos ver es I know not what who ES

hot has

gekoyft. bought.

hot has

gekoyft. bought.

The same contrast is seen in multiple questions involving a subject and an indirect object. The superiority effect is in evidence regardless of whether the indirect object is marked simply with the dative case (a-b examples), or with a preposition (c-d examples): (20a) ver vemen hot gegebn khanike-geltl who whom has given Hanukkah-money 'Who gave whom Hanukkah money?' [HÖGE 2000: 285] (20b) *vemen ver hot gegebn khanike-gelü whom who has given Hanukkah-money (20c) ver tsu vemen hot geshikt leshonetoyvesl who to whom has sent Rosh Hashanah cards 'Who sent Rosh Hashanah cards to whom?' (20d) *tsu vemen ver hot geshikt leshonetoyvesl to whom who has sent Rosh Hashanah cards An examination of the questioning of double objects reveals that a dative-marked indirect object is superior to the direct object: (21a) vemen vos hot maks whom what has Max 'What did Max give (to) whom?' (21b) *vos vemen hot maks what whom has Max

gegebn? given gegebn? given

The prepositional dative, on the other hand, shows the reverse superiority relation, with the direct object being superior (this contrast between dative-marked and

204

MOLLY DIESING

prepositional indirect objects is consistent with the results of 1986): (22a) vos what 'What (22b) *tsu to

tsu vemen hot maks to whom has Max did Max send to whom?' vemen vos hot maks whom what has Max

BARSS &

LASNHC

geshikt? sent geshikt! sent

I should note here that Yiddish does not seem to allow multiple questions with more than two wA-phrases; speakers tend to conjoin wA-phrases in excess of two. See D I E S I N G ( 2 0 0 3 ) for further discussion of this constraint. It is misleading to characterize this asymmetry with respect to superiority simply as a contrast between single fronting and multiple fronting, however. As I showed in D I E S I N G (2003), superiority effects are also seen with single fronting questions, but only in embedded contexts. Example (23c) below shows a superiority violation in a single fronting embedded question and it is considerably degraded in acceptability. (23a) hot zi ni(sh)t gekent farshteyn ver es shlogt zikh mit vemen. has she not could understand who ES hits self with whom '(So) she couldn't understand who was fighting with whom.' (23b) hot zi ni(sh)t gekent farshteyn ver mit vemen es shlogt zikh. has she not could understand who with whom ES hits self '(So) she couldn't understand who was fighting with whom.' (23c) *? hotzi ni(sh)t gekent farshteyn mit vemen ver shlogt zikh. has she not could understand with whom who hits self (23d) *hot zi ni(sh)t gekent farshteyn mit vemen ver es shlogt zikh. has she not could understand with whom who ES hits self However, (23c) is not nearly as bad as violating superiority in a multiple fronting context, as in (23d). A possible explanation for this contrast is that the superiority violating single fronting sentence is simply too similar to its multiple fronting counterpart (differing only in that the latter has the expletive es 'it' in Spec,IP). Considering examples which do not have a subject wA-phrase allows us to control for this factor: (24a) ikh veys vemen maks hot vos geshikt. I know whom Max has what sent Ί know whom Max sent what.' (24b) *?ikh veys vos maks hot vemen geshikt. I know what Max has whom sent (24c) ikh veys vemen vos maks hot geshikt. I know whom what Max has sent Ί know whom Max sent what' (24d) *ikh veys vos vemen maks hot geshikt. I know what whom Max has sent

205

The upper fiinctional domain in Yiddish

Example (24a) shows a case of embedded single fronting which obeys superiority. As expected, this sentence is grammatical. The superiority-violating variant in (24b), however, is significantly worse, though again it appears that it is not judged to be quite as bad as the superiority-violating multiple fronting case in (24d). A superiority effect is also seen with adjuncts: (25a) ikh veys farvos maks I know why Max Ί know why Max bought what.' (25b) ?*ikh veys vos maks I know what Max

hot has

vos what

gekoyfl. bought

hot has

farvos gekoyfl. why bought

Taken all together, these examples suggest that the determining factor for superiority effects is actually the landing site of wA-movement. JFA-movement to the specifier of CP (whether single or multiple) is subject to superiority, while whmovement to the specifier of IP (as happens with single fronting matrix questions) is not subject to the superiority constraint. This difference in the properties of movement associated with the landing site supports the analysis proposed thus far. (I should note that there are further constraints on multiple fronting which, though they are extremely interesting, do not bear on the questions of functional structure being considered here, see HÖGE 2000 and DlESlNG 2003 for details.) At this point the upper functional domain of Yiddish can be schematically represented as follows: (26) ([cp WhWh C+multjplew/i) [rp XP/Wh ItopK/focwZ-multiplewh [vP ·— [cP C+/.multiplewA [ipXP ^topic/focus ··· ]]]]] The topmost CP layer is only generated "as needed", indicated here by parentheses - it will only be generated in a multiple-fronting question. The matrix IP can license either a wA-phrase or a fronted XP (as a result of some sort of topic/focus movement, the finite verb also moves to the head I), but no multiple specifiers are licensed here. The embedded C is selected by the matrix verb; when a whcomplement is selected, it can be [+/- multiple]. Since the wA-feature in an embedded clause is a function of selection, an embedded IP does not license whmovement. Both CP and IP have the potential to license "focus-related" material whether it is a fronted XP or a wA-phrase. This structure presents an additional departure from RlZZl's (1997, forthcoming) cartographic view in that there seems to be no conflict in Yiddish resulting from having two projections hosting focusrelated material, as RlZZl has claimed to be the case in Italian. Italian does seem to have stricter constraints on focus, in that it seems to disallow multiple questions as well as other cases of multiple foci (CALABRESE 1984), so clearly this is a matter of cross-linguistic variation, and not a universal principle of clause structure.

206

MOLLY DIESING

4. The left edge of the upper functional domain The final construction I will examine involves what I will call discourse topics co-occurring with wA-questions. Specifically, Yiddish allows V3 orders in matrix clauses which result from fronting of a constituent to the left of a wA-phrase: (27a) DU ver bist? YOU who are 'Who do you think YOU are? [ B I R N B A U M 1979: 304 (210d)] (27b) AHIN ver geyt? THERE who goes 'Who's going THERE?' [ZARETSKI 1929:236] (27c) NEKHTN vu bistu geven? YESTERDAY where were-you been 'Where were you YESTERDAY?' (27d) mit di KINDER vos tut men? with the CHILDREN what does one 'What does one do with the CHILDREN?' Both SANTORINI ( 1 9 9 5 ) and HÖGE ( 2 0 0 0 ) refer to these fronted XPs as "focused constituents" without discussion of any criteria for such a classification. However, sentences like these can be paraphrased in terms of "as-phrases" - As for yesterday, where were you?, 'As for the children, what does one do with them?, which is traditionally taken to be an indication of topic-hood (GUNDEL 1 9 7 5 ) . I will assume that these paraphrases indicate that the fronted constituents actually function as discourse topics, in the sense of being foci selected from a set of topics already accessible in the discourse (thanks to WERNER ABRAHAM, p.c., for clarifying this point for me). There are some constraints on the syntax of this topic-fronting. As HÖGE (2000: 281) points out, only one such topic phrase can be placed to the left of the wA-phrase: (28a) *mit di kinder nekhtn vu with the children yesterday where (28b) *shabes bay nakht in vald ver fun Saturday night in forest who of

bistu were-you aykh vet you will

geven? been geyn? go

This restriction contrasts with topicalization in Italian, which is recursive (RlZZl 1997), but corresponds to constraints on a possibly analogous construction in Kashmiri, another symmetric V2 language (BHATT 1999, in particular the discussion on pp. 107-116). Note also that single topic-fronting is compatible with multiple wA-fronting: (29) SHABES BAY NAKHT ver vuhin vet geyn Saturday night who where will go 'Who will go where with you on Saturday night?'

mit aykh? with you [HÖGE 2000: 282]

(Interestingly enough, Kashmiri also allows optional multiple-fronting, which can co-occur with sentence-initial topics.) Topic-fronting is also a distinct process from left-dislocation (which space does not permit me to discuss in this paper) in

207

The upper functional domain in Yiddish

that left-dislocation involves a phrase adjoined to a V2 clause that is anaphoric to an element within the clause (see SANTORINI 1995: 99-100). Non-anaphoric cases of left-dislocation exist, but are limited to certain adverbs that function as discourse connectives (BIRNBAUM 1979: 302, ZARETSKI 1929: 236). A further constraint on topic-fronting demonstrated by HÖGE (2000: 281) is that topic-fronting of this sort is not possible in embedded questions: (30)

*Reyzl

Rose

fregt

zikh

NEKHTN

vu

du

bist

asked

self

YESTERDAY

where

you

are

geven.

been

Bulgarian (LAMBOVA 2001: 356) also allows topic phrases to precede fronted whphrases, but in contrast to Yiddish (and similarly to Italian), multiple topics are allowed. Both LAMBOVA ( 2 0 0 1 ) and BHATT ( 1 9 9 9 ) argue that the topics are moving to the same projection as the fronted wA-phrases, mainly on the basis of the impossibility of intervening material appearing between the topic and the wA-phrase(s). In terms of implementation, the target of wA-movement can host both a topic feature and a wA-feature. In Bulgarian, either or both of the features can be [+multiple], in Kashmiri only the wA-feature can be [+multiple]. This idea can be extended to Yiddish by positing that the CP projection can host both the topic and wA-features (as I argued above in section 2 for IP), but CP, unlike IP, can host both features at the same time. The occurrence of the left-edge topic in an indirect question will be ruled out, however, by locality constraints on selection - an intervening topic feature blocks the required selection configuration between the matrix verb and the embedded wA-feature. This gives us the following structure for the functional domain: (31)

( [ C P X P -multipleTopic WhWH [CP

C. m u itjpi e Topic+multipleii'A) [IP X P / W A Itopic/focus/wA [VP

[IP X P Itopic/focus ] ] ] ] ]

To briefly give a final summary of these structural results: the Yiddish IP corresponds to the "verb second domain". Movement to the specifier of IP is not subject to superiority, perhaps because of the mixed nature of the landing site - it hosts not only both topic and focus movement but also single wA-fronting in matrix clauses. The CP hosts multiple wA-fronting (unlike IP, CP can check multiple features), as well as single fronting in embedded clauses. Movement to the specifier of CP is subject to CP. Finally, a single topic phrase may attach to the left of the whphrase(s), perhaps by means of adjunction, as suggested by LAMBOVA ( 2 0 0 1 ) .

5. Conclusions While this paper represents only a sketch of the structure of the left periphery in Yiddish, with many questions remaining to be answered, it does question the validity of the idea that the functional structure of the clause is universal. While "exploded" functional structure may be well-motivated in some languages, it's not at all clear that the facts of Yiddish word order support this type of analysis. I have

208

MOLLY DIESING

not addressed the issue of whether this indicates that the inventory of functional features themselves (as opposed to the arrangement of the functional projections hosting such features) is or isn't universal. One possibility is that all languages have the same functional features (such as Tense and Agreement), but the syntactic expression of these features is parameterized - they may project separate heads in some languages, and "fused" heads in others (see THRÄINSSON 1996 for a proposal along these lines regarding the structure of IP). Another possibility for typological variation is that languages differ in what kinds of features motivate movement at all, whether it is inflectional features (as in CHOMSKY'S 1995 system), or some set of more discourse-oriented features (as proposed by ABRAHAM 1997). An interesting hybrid view is LAMBOVA's (2001) proposal that the typological property of being discourse-oriented is the factor determining that the CP domain takes a fused form (see also URIAGEREKA 1995), rather than the split projections argued for by RlZZl (1997). At the very least, Yiddish provides evidence for variation of the first kind, but it is also clear that discourse-functional properties play a role in Yiddish word order, not only in XP-fronting, but also in scrambling. Exactly how these properties exercise their influence on clause structure I will leave as a matter for future research. References (1997): The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative ones, in: ABRAHAM, WERNER & VAN GELDEREN, ELLY (eds.), German: syntactic problems-problematic syntax. Tübingen: Max Niemeyer, 11-43. BACH, EMMON (1962): The order of elements in a transformational grammar of German, in: Language 38,263-269. BARSS, ANDREW & LASNIK, HOWARD ( 1 9 8 6 ) : A note on anaphora and double objects, in: Linguistic Inquiry 17, 3 4 7 - 3 5 4 . BHATT, RAKESH MOHAN (1999): Verb movement and the syntax of Kashmiri. Dordrecht: Kluwer Academic Publishers. BIERWISCH, MANFRED (1963): Grammatik des deutschen Verbs. (Studia Grammatica 2). Berlin: Akademie Verlag. BIRNBAUM, SALOMO (1979): Yiddish: a survey and a grammar. Toronto: University of Toronto Press. BOBAUIK, JONATHAN & THRAINSSON, HÖSKULDUR ( 1 9 9 8 ) : Two heads aren't always better than one, in: Syntax 1(1), 3 7 - 7 1 . BRANIGAN, PHIL ( 1 9 9 6 ) : Verb-second and the Α-bar status of subjects, in: Studia Linguistica 5 0 , 5 0 - 7 9 . CALABRESE, ANDREA (1984): Multiple questions and focus in Italian, in: DE GEEST, W. & PUTSEYS, Y. (eds.), Sentential complementation. Dordrecht: Foris, 67-74. CHOMSKY, NOAM ( 1 9 9 5 ) : The minimalist program. Cambridge, MA: The MIT Press. CINQUE, GUGLIELMO (1999): Adverbs and functional heads. Oxford: Oxford University Press. DIESING, MOLLY ( 1 9 9 0 ) : Verb movement and the subject position in Yiddish, in: Natural Language and Linguistic Theory 8(1), 4 1 - 7 9 . DIESING, MOLLY (1997): Yiddish VP order and the typology of object Movement, in: Natural Language and Linguistic Theory 15(2), 369-427. DIESING, MOLLY ( 2 0 0 3 ) : On the nature of multiple fronting in Yiddish, in: BOECKX, CEDRIC & GROHMANN, KLEANTHES (eds.), Multiple fronting. Amsterdam: John Benjamins, 4 7 - 7 0 . GELDEREN, ELLY VAN ( 1 9 9 3 ) : Theriseoffunctional categories. Amsterdam: John Benjamins. GRIMSHAW, JANE ( 1 9 9 7 ) : Projection, heads and optimality, in: Linguistic Inquiry 2 8 ( 3 ) , 3 7 3 - 4 2 2 . GUNDEL, JEANETTE Κ. (1975): The role of topic and comment in linguistic theory. Ph.D. dissertation, University of Texas. ABRAHAM, WERNER

The upper functional domain in Yiddish

209

On different kinds of focus, in: BOSCH, PETER & VAN DER SANDT, ROB (eds.), Focus: linguistic, cognitive, and computational perspectives. Cambridge: Cambridge University Press, 2 9 3 - 3 0 5 . HÖGE, KERSTIN (2000): Superiority. Ph.D. dissertation, University of Oxford. IATRIDOU, SABINE (1990): About AgrP, in: Linguistic Inquiry 21(4), 551-576. IATRIDOU, SABINE & KROCH, ANTHONY (1992): The licensing of CP-recursion and its relevance to the Germanic verb-second phenomenon, in: Working Papers in Scandinavian Syntax 50, 1-24. JACOBS, N.; PRINCE E. & VAN DER AUWERA, J. (1994): Yiddish, in: KÖNIG, EKKEHARD & VAN DER AUWERA, JOHAN (eds.), The Germanic languages. London: Routledge, 388-419. KOSTER, JAN ( 1 9 7 5 ) : Dutch as an S O V language, in: Linguistic Analysis 1 , 1 1 1 - 1 3 6 . LAMBOVA, MARIANA ( 2 0 0 1 ) : On Α-bar movements in Bulgarian and their interaction, in: The Linguistic Review 18, 3 2 7 - 3 7 4 . MARK, YUDEL (1978): Gramatik fun der yidisher shprakh. New York: YIVO Institute for Jewish Research. MOLNÄRFI, LÄSZLÖ (this volume): Some remarks on the formal typology of pronouns in West Germanic. PESETSKY, DAVID ( 1 9 8 7 ) : Wh-in-situ: movement and unselective binding, in: REULAND, E . & TER MEULEN, A . (eds.), The representation of (in)deflniteness. Cambridge, M A : M I T Press, 9 8 - 1 2 9 . PESETSKY, DAVID ( 2 0 0 0 ) : Phrasal movement and its kin. Cambridge, M A : M I T Press. POLLOCK, JEAN-YVES (1989): Verb movement, Universal Grammar, and the structure of I P , in: Linguistic Inquiry 20(3), 365-424. PRINCE, ELLEN (1999): How not to mark topics: 'topicalization' in English and Yiddish, Ms. University of Pennsylvania. REINHOLTZ, CHARLOTTE ( 1 9 8 9 ) : V - 2 in Mainland Scandinavian: Finite verb movement to Agr, in: Working Papers in Scandinavian Syntax 4 4 , 1 0 1 - 1 1 7 . RIZZI, LUIGI ( 1 9 9 7 ) : The fine structure of the left periphery, in: HAEGEMAN, LILIANE (ed.), Elements ofgrammar. Dordrecht: Kluwer Academic Publishers, 2 8 1 - 3 3 7 . RIZZI, LUIGI (forthcoming): Locality and left periphery, in: BELLETTI, ADRIANA (ed.), Structure and beyond: The cartography of syntactic structures. Oxford: Oxford University Press. RÖGNVALDSSON, EIRIKUR & THRÄINSSON, HÖSKULDUR ( 1 9 9 0 ) : On Icelandic word order once more, in: MALING, JOAN & ZAENEN, ANNIE (eds.), Modern Icelandic syntax. San Diego: Academic Press, 3-40. ROOTH, MATS ( 1 9 9 2 ) : A theoiy of focus interpretation, in: Natural Language Semantics 1, 7 5 - 1 1 6 . RUDIN, CATHERINE ( 1 9 8 8 ) : On multiple questions and multiple W H fronting, in: Natural Language and Linguistic Theory 6 , 4 4 5 - 5 0 2 . SANTORINI, BEATRICE ( 1 9 9 5 ) : The syntax of verbs in Yiddish. Ms. Northwestern University. THRÄINSSON, HÖSKULDUR ( 1 9 9 4 ) : Comments on the paper by Vikner, in: LIGHTFOOT, DAVID & HORNSTEIN, NORBERT (eds.), Verb Movement. Cambridge: Cambridge University Press, 1 4 9 - 1 6 2 . THRÄINSSON, HÖSKULDUR ( 1 9 9 6 ) : On the (non)universality of functional categories, in: ABRAHAM, WERNER; EPSTEIN, SAMUEL DAVID; THRÄINSSON HÖSKULDUR & ZWART, C . JAN-WOUTER (eds), Minimal ideas. Amsterdam: John Benjamins, 2 5 3 - 2 8 1 . TRAVIS, LISA DEMENA ( 1 9 8 4 ) : Parameters and the effects of word order variation. PhD. dissertation, GUNDEL, JEANETTE Κ . ( 1 9 9 9 ) :

MIT. URIAGEREKA, JUAN ( 1 9 9 5 ) :

An F position in Western Romance, in: Kiss, KATALIN E . (ed.), Discourse configurational languages. Oxford: Oxford University Press, 1 5 3 - 1 7 5 . VIKNER, STEN (1995): Verb movement and expletive subjects in the Germanic languages. Oxford: Oxford University Press. ZARETSKI, AYZIK (1929): Yidishe gramatik. Vilna: Vilner Farlag.

Präteritumschwund: evidence for areal diffusion* BRIDGET DRINKA (San Antonio)

1. Introduction In their book, Präteritumschwund und Diskursgrammatik (2001), ABRAHAM & CONRADIE begin with an illuminating description of the distribution of the perfect across the European landscape, with special focus on languages which have undergone the "Präteritumschwund" (PS), that is, the disappearance of the synthetic preterite and its replacement in this function by the periphrastic perfect in several "core" languages of Europe. They present, for example, the very informative maps of TERNES (1988: 340) and THIEROFF (2000: 285) (Maps 1 and 2 respectively, see

Appendix), which illustrate the geographical contiguity of languages which have undergone virtually complete preterite replacement: according to TERNES, Northern French, Southern German, Northern Italian, Czech, Slovak, Slovenian, and Northern Serbian and Croatian; according to THIEROFF, Southern German, Northern Italian, and Hungarian.1 ABRAHAM & CONRADIE (2001: 20) go on to make the following rather strong

statement, apparently in support of an areal explanation for the distribution of this innovation: "Die Distributionskriterien teilen Sprachen, die man den Kern eines Sprachbundes nennen kann - etwas, was kaum eine Zufallsverteilung sein kann." However, in spite of this seeming endorsement of an areal explanation, the rest of the book takes a completely different tack, attempting to demonstrate that other factors - especially parsing and discourse factors - are actually responsible for the distribution, and that the languages in question, as well as Yiddish and Afrikaans, have replaced their preterites with perfects independently through the influence of those factors, without the benefit of an external model. In this paper, I will examine in some detail the arguments proposed by ABRAHAM & CONRADIE, and will then

I would like to thank JULES GLIESCHE, BRIGITTE BAUER, CHRISTOPHER WICKHAM, JOSEPH SALMONS, and GILLES SOUVAY for assistance with the interpretation of the medieval texts used in this paper, and for their useful advice and suggestions. Special thanks are also due to SALKOKO MUFWENE, for thought-provoking discussions on theses issues, and for helpful information on the per1

fect in English Creoles. All errors which remain are my own. In earlier work, THIEROFF (1995: 34) also includes Serbian, Croatian, and Upper Sorbian as part of the central group, and he recognizes the process as operating in a number of other Slavic languages (East Slavic Russian, Belorussian, Ukrainian; West Slavic Polish and Lower Sorbian, South Slavic Slovene), as well as in spoken Rumanian, some Albanian dialects, and spoken French.

212

BRIDGET DRINKA

present data which supports an alternative view: that areal factors were, indeed, largely responsible for the reinterpretation of the perfect as a preterite.

2.

Arguments of ABRAHAM

& CONRADIE

ABRAHAM & CONRADIE present a comprehensive, compelling case for the role of parsing and discourse factors in the development of PS. Their most forceful arguments can be summarized as follows (2001: 123):

2.1. Parsing principle In oral communication, the most efficient code is one which establishes identification of the subject early in the string, for example, by means of subjectagreement marking on the auxiliary. This fact encourages the placement of auxiliaries in V-second position. 2.2. Discourse principle Transparency of Theme-Rheme relationships is enhanced with a large "middle field" created by braciation; that is, the new elements of the discourse (the Rheme) can be most coherently presented in a unitary manner in the verbal complement, positioned between the finite element of the verb phrase (the auxiliary) and the non-finite element (e.g., the participle). This fact encourages the retention of Vfinal word order. The authors view this factor, while important, as less essential than the parsing principle (2001: 125). The result of the interaction of these two principles is the creation of SVOV order, which ABRAHAM & CONRADIE claim exists for German in both main and subordinate clauses (2001: 123 et passim). Hence, German is held in a typologically inconsistent state by at least two contesting forces2: speakers find value in conveying information about the subject, by means of the auxiliary, early in the string, yet they continue to find pragmatic, functional value in the archaic V-final tendencies of their language, and see no reason to give them up for a more coherent, thoroughly SVO pattern. According to ABRAHAM & CONRADIE, it was this arrangement, with V-second auxiliary, which allowed for the reanalysis of haben as a true auxiliary in late MHG, rather than as a main verb. Through raising, the auxiliary and participle came to share the subject of the sentence; aspectual implications of the construction were lost, and the perfect became purely temporal, and now capable of replacing the synthetic preterite (ABRAHAM & CONRADIE 2001: 94). Given a periphrastic way to express the preterite, then, speakers unconstrained by the standard would 2

ABRAHAM & CONRADIE ( 2 0 0 1 : 7 5 - 7 6 ; 1 2 4 - 2 5 ) add the important factor of prosodic optimization which occurs in V-final arrangements, best achieved in subordinate clauses in German.

Präteritumschwund: evidence for areal diffusion

213

have naturally chosen that method, since it provided them with more attractive parsing and discourse opportunities, as mentioned above. Essential to this argument is the fact that such developments are predicted to occur most frequently in spoken discourse: vernaculars, unencumbered by the normative pressures of a standard, will offer less resistance to such "natural" developments. Where such innovations - the increased reliance on V-second auxiliaries and the resultant increase in verbal periphrasis - are expected to appear, then, is in historically V-final vernaculars (i.e., where subject agreement is delayed and where standardizing norms are absent). ABRAHAM & CONRADIE provide evidence from three such vernaculars, namely, Afrikaans, Yiddish, and Southern German, and they go on to argue that in each of these non-contiguous varieties, the loss of the preterite was independent, was tied to parsing and discourse factors, and was implemented because the vernacular was not constrained by a standard language. A brief look at the evidence from these three varieties will illustrate the strengths and weaknesses of their arguments.

3. Evidence from the vernaculars 3.1. Afrikaans ABRAHAM & CONRADIE present an impressive array of data to illustrate the changes which Afrikaans underwent in its early stages of development. They argue convincingly that the language replaced its preterites with perfects especially because of two factors: the simplification of the morphological marking of the perfect and a propensity towards the semantic, functional, and pragmatic features of the perfect. With regard to formal simplification, the reduction of the perfect auxiliary to the unvarying het and the simplification of the PPP to ge- + stem (e.g. breek 'break' : gebroken > gebreeken 'broken') virtually eradicated the morphological complexity of the perfect (35-36). The preterites, on the other hand, showed far greater irregularity, especially with the loss of the -t marker: koop : kog (cf. Du kopen: kog) 'buy : bought', dink : dog 'think : thought' (40). With regard to functional and pragmatic suitability, the perfect provided speakers with an opportunity to build solidarity, to promote a sense of participation with co-conversants, and, in general, to enhance the subjectivity of the utterance (37-38). Crucial to the process of replacement in Afrikaans was the presence of nonnative speakers, as ABRAHAM & CONRADIE demonstrate in precise detail (2001: 42-51): early native speakers (pre 1830's) kept the preterite and PPP intact; L-2 speakers, especially after 1830, showed much more frequent replacement of the preterite by the simplified perfect. ABRAHAM & CONRADIE note that imperfect learning of irregular forms by L-2 learners must have played an important role in the development. However, one must still ask the following question: why is it that the perfect was the category chosen to be honed into regularity for use as a preterite? Why did the preterite, for example, not undergo reshaping into more regular form, or an adverb come to be reanalyzed as a temporal particle? One response is, of course, easy to

214

BRIDGET DRINKA

adduce: the preterite had lost its most salient marker, the -t suffix, and so was unsuitable for revitalization; the perfect was close at hand, already replete with (specialized) preterital value, and ready to be reshaped and regularized. Yet, another factor should not be ignored here: a large number of the L-2 speakers analyzed by ABRAHAM & CONRADIE spoke languages which had already undergone the replacement of preterite by perfect (especially German and French).3 ABRAHAM & CONRADIE, in fact, point with approval to the conclusions of PHEIFFER (1980: 123), who claims, "daß das Präteritum in der Sprache der Französisch sprechenden Einwanderer jemals eine bedeutende Rolle spielte - nicht nur da diese die unregelmäßigen Präteritumformen nur unvollständig korrekt erlernten, sondern auch aufgrund des Umstands, daß das im mündlichen Französischen stark verankerte passe indefini ihnen eher den Weg zum zusammengesetzten Perfekt des Afrikaans vorebnete." (paraphrase by ABRAHAM & CONRADIE 2 0 0 1 : 4 3 ) What such observations lead us to understand is that the development of this innovation may not have been truly independent of the European innovation after all, as claimed by ABRAHAM & CONRADIE: this change happened in the shadow of the European trend, even if Dutch, the source language of Afrikaans, did not participate in it to the same extent.4 3.2. Yiddish ABRAHAM & CONRADIE (2001: 1 1 5 - 1 2 2 ) provide convincing evidence that Yiddish should be considered SOV, not SVO, as has been claimed by a number of scholars5. It is not clear, however, why the retention of an archaic syntactic pattern should imply that the introduction of the perfect in place of the preterite should have occurred independently, especially since Yiddish springs originally, incontestable from Southern and Middle German dialects.6 EGGERS ( 1 9 9 8 : 2 8 9 - 2 9 1 ) presents a more complex picture of the development, noting that Western Yiddish dialects tend to retain the synthetic preterite, like the Northern German dialects, while the Eastern Yiddish dialects have wholeheartedly adopted the analytic form, like the southern German dialects. According to EGGERS, an early Bavarian source for the Eastern Yiddish analytic pattern is excluded, even though the innovation 3

4

3

6

VEKEMAN & ECKE (1992: 223) provide the following statistics concerning the national origins of the Cape Colony population between 1650 and 1800: Dutch 36.8%, German 35.0%, French 14.6%, British 0.3%, other Europeans 2.6%, (non-European) Africans and Asians 7.2%, unknown 3.5%. They also note (1992: 233) that the number of loan words and loan translations that were brought into Afrikaans from German in the 19th century was so high that purists were alarmed. The replacement of the preterite by the perfect is claimed by ABRAHAM & CONRADIE (2001: 27) to have occurred only in remote, strongly localized dialects, but a number of Dutch speakers from various locations report that they are able to use past adverbs with the perfect, implying at least a partial adoption of PS in Dutch. SANTORINI (1993), like many others, sees Yiddish as moving towards SVO; WEXLER (1991) goes so far as to argue that Yiddish had calqued its presumably SVO word-order on the model of Slavic. WEXLER (1991) takes the controversial stance that Yiddish was Slavic in origin, but this claim has been generally dismissed (COMRIE 1991; D o w & STOLZ 1991; STANKIEWICZ 1991; EGGERS 1998).

215

Präteritumschwund: evidence for areal diffusion

apparently began in Bavaria (cf. LLNDGREN 1957)7, since the change occurred after Yiddish speakers had already begun moving into Eastern Europe in the 13th and 14th centuries.8 Rather, he argues, Eastern Yiddish must have acquired its usage on the basis of Middle German influences in the 15th and 16th centuries, influences which ceased in these dialects of Yiddish in the 17th century, while Western Yiddish kept more contact with Northern German dialects for a longer time. The Yiddish patterns would thus reflect the corresponding developments in several varieties of German. If this interpretation is correct, the innovations in Yiddish would represent not an independent development of PS, but rather several layers of German-influenced development. It is contact, then, not formal or discourse factors, which would be motivating the innovation in Yiddish in this model, counter to ABRAHAM & CONRADIE'S claim. However, another interpretation can also be proposed. As D E N T L E R (1998: 134— 142) demonstrates, the PS in German can be more accurately characterized as an ongoing expansion of the perfect than as a disappearance of the preterite. D E N T L E R depicts the development in the following way: in 1000 AD, the perfects had present (anterior) function, with little connection to preterite meaning; as the centuries went on, perfects could be found in ambiguous contexts, connected either to anteriority or to past time; gradually, they became more and more capable of appearing with adverbs marking definite past, with atelic verbs, and in other purely past contexts. Three stages in this development are illustrated below, following D E N T L E R (1998:143-145): (1)

(2)

(3)

7 8

Temporal focus ambiguous: ... Vuaz ist mir danne gaskehenl 'What has to-me then happened?' [NOTKER (from Psalm

143)

beginning of the

Increased preterital orientation with past adverb: ...und diu zit, diu da vergangen ist vor tüsentjären... 'and the time that has passed away a thousand years ago' [MEISTER ECKHART (Sermon 3 : 1 6 4 ) end of the

11 TH

century]

th

century]

13

Preterital reference, in a preterite context: zehant er besande/die besten von dem lande/...nu daz si vür in kome sint... 'at once he sent out the best from the land.. .now that they have come before 9 him' [HARTMANN VON A U E (Gregorius 1 9 5 ) end of the 12 TH century]

For an alternative explanation, see §5 below. Even while regarding the shift in German as relatively late, EGGERS still sees a fundamental role for the German construction in the formation of the Yiddish one: " [D]ie Herausbildung der analytischen Vergangenheitsformen [ist] im Jiddischen ein Zeichen dafür, daß das Jiddische sich in der ersten Zeit seines Bestehens im osteuropäischen Raum weiterhin am deutschen Vorbild orientiert e " (EGGERS 1 9 9 8 : 2 9 0 ) .

9

Note that the third example represents a more advanced stage of the shift, even though it appears at an earlier date than the second, less advanced, example; such evidence reflects the variable nature of the change in progress. Also to be noted is the fact that MEISTER ECKHART (C. 1260-1327/8) served as a professor of theology in Paris, and that HARTMANN (d. between 1210 and 1220) wrote many tales based on courtly French models. The importance of such connections with French will be explored in §5.

216

BRIDGET DRINKA

The essential point to grasp here is that the introduction of the semantic shift was gradual and variable; the perfect did not suddenly take over the function of the preterite, but slowly extended its own realm into that of the preterite, overlapping upon it. When we consider how such gradual development might impact the history of Yiddish, it becomes clear that the language that the Jews took with them into Eastern Europe most assuredly possessed perfects already rich in preterital meaning. The trend would have already been in place, in incipient form, in southern German, and would have been carried eastward along with other Northern Bavarian traits, such as special indefinite articles, diminutives, and 2nd person plural pronouns (EGGERS 1998: 2 2 9 - 2 3 0 ) . The fact that Yiddish maintains the BE/HAVE distinction for unaccusative vs. transitive verbs just as German does provides further evidence for the close ties of Yiddish to German. Thus, there is no need to regard the Yiddish adoption of PS as independent from that of German, as ABRAHAM & CONRADIE do.

3.3. Southern

German

The onset of the innovation among southern German dialects is welldocumented, most notably in the work of LLNDGREN (1957), in the chronicles of Augsburg and Nürnberg. These two cities, and the area between them, became prosperous cultural and economic centers in the 15TH century, eventually becoming seats of imperial chanceries (MATTHEIER 2 0 0 0 : 1090; 1098). LlNDGREN's data points to a rapid increase in replacement of the preterite by the perfect in Augsburg: in 1437, early signs of preterite loss were found in documents, but only in quotations of direct speech; by 1500, weak but clear signs appeared in narratives, as well; by 1536, the old preterite system had ceased to operate productively, and the perfect had come to predominate (LlNDGREN 1957: 5 4 - 6 6 ; 9 3 - 9 7 ; 106). ALBERTS ( 1 9 7 7 ) claims that the innovation became pervasive before 1490, and provides evidence for a notable increase in the use of periphrastic perfects in Augsburg over the course of the 15TH century. This fact, along with the findings of DENTLER 1998 (see above, 3.2), may suggest a more gradual development for the innovation, as well as an earlier starting point, than LlNDGREN suggests. The Nürnberg Chronicles only extend to 1500, but provide evidence consistent with the findings in Augsburg: by 1488, the perfect was used more than the preterite, but only in direct speech (LlNDGREN 1957: 66). The fact that the changes appeared first in passages where direct speech was recorded supports A B R A H A M & C 0 N R A D I E ' S claim that PS occurred first in the spoken language of Oberdeutsch. It was here, say A B R A H A M & C O N R A D I E , that "natural" tendencies of the vernacular could take precedence, allowing the perfect to become exclusively temporal and interchangeable with the preterite ( A B R A H A M & CONRADIE 2 0 0 1 :

94).

However, evidence from a distinctly non-vernacular source, the translation of the Bible during the 14TH through 16TH centuries, turns out to deal a severely damaging blow to ABRAHAM & CONRADIE'S argument that braciation played a role in the development of PS in Southern German. SCHILDT ( 1 9 6 8 ) demonstrates that Oberdeutsch translators of the Bible in the years 1 3 5 0 - 1 4 5 0 used minimal or no bracia-

Präteritumschwund:

evidence for areal diffusion

217

tion, as opposed to translators from more northerly regions, who showed strong signs of its use already in 1350: (4)

Vulgate vocavi te nomine tuo (Jes. 43:1) CRANC's translation (1350, Prussia) und habe dich mit dinem namen genant Mentelbibel (1466, Oberdeutsch) und hob dich geruffen mit dem namen 'and I have called you by your name'

Even as late as 1550, Oberdeutsch translators like ECK still used braciation less productively and consistently than LUTHER, who abided by more a northerly East Middle German trend in his frequent use of braciation: (5)

LUTHER (1522-46) ECK (1537) (OD)

ich habe dich bey deinem namen geruffen dä ich hob... dich beruft bei deinem namen

Thus, the use of braciation has a clear geographical distribution, even as early as 1350, but not along the lines predicted by ABRAHAM & CONRADIE: southern, western, and Bohemian translations (e.g., the Wenzelbibel of 1350/1400 from Bohemia, the West Middle German translations of the 14th century) tend toward a more Latin-like lack of braciation; translations from more northern territories (e.g., BEHEIM'S 1343 translation from Saxony-Thuringia, CRANC's 1350 translation from Prussia) already show a clear tendency towards the use of braciation: (6)

Vulgate CRANC (1350, Prussia)

Wenzelbibel (1350/1400, Bohemia) (7)

patres comederunt uvam acerbam (Jer 31:29) dy vetire habin suer winber gezzin vnser veter haben gessen ein herbe trauben 'the fathers have eaten a sour grape'

Vulgate

tradidit enim Dominus inimicos nostros Moabitas in manus nostras West Middle German (14th century) der herre hait gegeben unser gyhende und moab in unser hattde Mentelbibel (OD 1466) wann der herre hat geantwurt unser feinde die moabiter in unser hend LUTHER (1523/24-46) denn der Herr hatt euch die Moabiter ewr feyndeynn ewr hende geben ECK (OD 1537) dan der herr hat unser feind die Moabitergebe in unser hand 'for the Lord has delivered our enemies the Moabites into our hands'

Significantly, the OD Zainerbibel from Augsburg (1475), a modernization of the Mentelbibel from Strasbourg (1466), shows virtually no braciation, even while it does show an increased use of periphrastic verbs (SCHILDT 1968: 181; 189). It is precisely in Augsburg where we should expect to find braciation having its effect on PS, if braciation truly is responsible for the innovation in Southern German, since, as noted above, PS is clearly documented as having occurred here over the course of the 15th and 16th centuries (LLNDGREN 1957; ALBERTS 1977). While one might attempt to argue that the vernacular language operated in an altogether dif-

218

BRIDGET DRINKA

ferent fashion from the formal language of Bible translation, it would seem difficult to prove that a tendency towards braciation in the vernacular was strong enough to generate a substantial semantic shift like PS without leaving a trace in the written record. The virtual lack of braciation in OD translations points to an inevitable conclusion: braciation could not have played the role in the implementation of PS in Southern German which ABRAHAM & CONRADIE claim it did.

4.

Further response to ABRAHAM & CONRADIE

Even if we do not accept ABRAHAM & CONRADLE's explanation for the development of PS in Southern German as being tied to braciation, an examination of their other claims with regard to the innovation will still be useful. The authors themselves pose two crucial questions concerning the innovation in Oberdeutsch (2001:79): 1. If parsing and discourse principles are motivating the replacement process, why didn't other structures participate in the innovation, alongside the perfect? 2. If these principles are fueling the innovation, why didn't other Continental West Germanic languages also undergo this change (e.g. Dutch, W. Frisian, and the Low German dialects)? To these, we can add several other crucial questions: 3. Why do other languages, such as French and Northern Italian, show the same change, even without braciation? 4. How truly essential is the lack of a standard language? Can we expect vernaculars to behave in this way regularly? 5. What is it that really allows the perfect to be construed as a preterite? While some of these questions concern issues which are too large to be broached here, each can at least be addressed in brief. 4.1. Why did other structures not participate? ABRAHAM & CONRADIE respond to this question by noting that the southern German dialects did extend the discourse principle to other structures, namely the tun auxiliary ('do'):

(8)

die theten vnglaublich mit einander zankhen they do unbelievably with each other argue 'They argue with each other to an unbelievable degree' [Abraham a Sancta Clara, 1st Sermon, end of the 16th - beginning of 17th century, quoted in ABRAHAM & CONRADIE 2001: 84)]

The authors' point is convincing: the Southern German dialects have increased their inventory of V-2 auxiliaries, likely in response to the parsing considerations which the authors describe. The /««-periphrastic seems to run parallel to the perfect

Präteritumschwund: evidence for area! diffusion

219

on several levels ( 2 0 0 1 : 8 0 - 9 3 ) , and lends valuable support to ABRAHAM & CONRADlE's argument.

4.2. Why did other continental West Germanic languages not participate? Dutch, W. Frisian, and the Low German dialects have similar syntactic features to Southern German, including braciation; why did they not undergo this innovation, as well? ABRAHAM & CONRADlE's response to this question is less satisfying than to the previous question: they simply point to Afrikaans as a Germanic language which has undergone a parallel shift, presumably unrelated to that of Southern German (2001: 79). The implication is that the development in Afrikaans constitutes independent evidence for the operation of the principles. However, as we have seen above (3.1-2), the evidence from Afrikaans and Yiddish suggests that these vernaculars did not develop their PS independently, but did so on the model of or in conjunction with Western European languages which had undergone the shift themselves. The question remains, why did Dutch not adopt the innovation wholeheartedly?10

4.3. Why do French and Northern Italian show the same change, even without braciation? One of the strongest argument against the explanation given by ABRAHAM & is the fact that, even though French and Northern Italian do not use braciation, they have still undergone a thorough replacement of their preterites by their perfects. In the face of this fact, it seems difficult to argue that parsing and discourse principles are as essential as ABRAHAM & CONRADIE claim that they are. It is, in fact, surprising that, while Northern Italian is listed as having participated in preterite loss (ABRAHAM & CONRADIE 2001: 9), French is not even mentioned by ABRAHAM & CONRADIE; French is, rather, classified as having a marginal preterite, and so is set apart as less involved in the change (cf. ABRAHAM & CONRADIE 2001: 11, Table 2). However, one must ask how large the difference truly is between vernacular French usage and that of Northern Italian. The explanation which I propose is that French is not marginal but rather central to the geographically diffused innovation that the innovation, in fact, moved from French into the other varieties. One can also mention, in this regard, that a number of Slavic languages underwent a somewhat parallel shift of resultative participial construction developing into a synthetic preterite (with or without loss of the BE auxiliary) without the intervention of braciation. CONRADIE

10

It must be admitted, however, that Dutch may represent a problem for the areal approach, as well, since at least some of its dialects underwent heavy French influence. For example, Old French served as a medium of communication between Flemish and French nobles, and was the language of the Burgundian dynasty and its court at Brussels (VEKEMAN & ECKE 1992: 82; see also footnote 19).

220 (9)

BRIDGET DRINKA

Russian my posli utrom 'we left in the morning'

Such innovations can clearly occur even without syntactic/discourse predisposition. To summarize the points made in 4.2 and 4.3, then, the connection of PS to braciation both overpredicts (4.2) and underpredicts (4.3) its implementation. We would expect the change to occur in structurally similar Germanic languages, where it does not, and we would not expect it to occur in contiguous Romance languages, where it does. While we should not underestimate the importance of ABRAHAM & CONRADIE'S two principles, it seems clear that other forces are more responsible for the actual instigation of the change than these are, and the contiguity of these languages on a map makes an areal explanation most appealing: precisely as stated by ABRAHAM & CONRADIE themselves (see §1), this similarity of development among neighboring varieties cannot be due to mere coincidence. 4.4. Is the lack of a standard essential? That is, do vernaculars regularly behave in this way? ABRAHAM & CONRADIE claim that Afrikaans and Yiddish "unnormalisierten, im wesentlichen unkodifizierten, sprechsprachlichen Wurzeln entspringen" (2001: 93). What we are expected to extrapolate from this undeniably true statement is that vernaculars, in general, tend to wipe out distinctions for the sake of regularity or some other optimization. This prediction is, however, not always fulfilled: for example, African-American Vernacular English (AAVE) maintains the perfect category, even if it frequently loses the auxiliary through a deletion/contraction rule.11 It has, indeed, expanded upon the general category, as witnessed, for example, by the development of been as a marker of "remote present perfect" {I been know your name = Ί have known your name for a long time'), the adoption of done as an intensive perfect marker12 {I done told you on that = 'I've already told you that'), and the more frequent use of the future perfect in AAVE than in Standard English: I be

11

12

AAVE tends to delete auxiliaries at precisely the same site where Standard English tends to contract them: I've seen him : I seen him; He's talking : He talkin' (LABOV 1972: 73). LABOV notes that the present perfect is used less frequently in AAVE than the past perfect, and suggests that this distribution may owe its existence to the operation of phonological processes (deletion of [v] and [z]), or to grammatical properties (LABOV 1972: 122). MUFWENE (p.c.) observes that bin has come to signify anterior and don(e) the perfect in many English Creoles. He adds that the past perfects do not present a clear semantic picture: they often refer to the past or, indeed, to the present perfect itself. SCHNEIDER (1993: 214-216) traces back the AAVE done construction to ME have/be + done + PP. The construction, surviving especially in northern England, was transported to America, appearing especially in southern dialects. Evidence for its existence is found in slave narratives {had done helped, to be done, etc.). The first element subsequently underwent weakening, leaving done as the primary marker of the perfect, in which role it has expanded considerably in frequency and functional application. Hence, the marking of the category has shifted, but the semantic value of the perfect has persisted.

Präteritumschwund: evidence for areal diffusion

221

done did this lil' spot a hair fo' you know it = Ί will have done this little spot of hair before you know it' ( L A B O V 1972: 5 3 - 5 5 ; S M I T H E R M A N 1977: 2 5 - 2 6 ) . Ver-

nacular status may be useful in engendering regularization and loss of category, but this result is not a given. An illuminating example of perfect development in a vernacular language is provided by TAGLIAMONTE 2 0 0 0 in her description of Samanä English, an offshoot of African American Vernacular English which has been spoken in the Dominican Republic since the mid-nineteenth century. Like Yiddish, this vernacular has been separated from its source language, and thus retains a number of elements from an earlier stage of English. However, it is important to note that the direction that the perfect took in this variety of English does not parallel that found in Yiddish, and does not, indeed, follow the predictions of ABRAHAM & CONRADIE. Rather than the perfect taking over the value of the preterite, the perfect has remained viable as a semantic category, with a profusion of forms developing to express it: besides the frequent have + past participle {Many of 'em have died), speakers of Samanä also use been/ain 't/be + past participle (I'm forgot all them things), a lone past participle {She gone to St. Martin), three verb clusters (I'm done been over there plenty), etc. It is not even necessarily the case that only periphrastic constructions are chosen to fulfill this function: simple pasts or suppletive forms and simple verb stems also frequently appear in the function of a perfect (They all died out already, I never like the city) (TAGLIAMONTE 2 0 0 0 : 3 3 0 - 3 3 2 ) . While many auxiliary structures are thus available for marking the perfect, these structures do not come to be interpreted as preterites; on the contrary, some preterites are being used to mark perfect function (TAGLIAMONTE 2 0 0 0 : 3 4 1 - 3 4 3 ) . Clearly, the vernacular status of Samanä did not foster replacement of the preterite by the perfect. Rather than introducing radical changes, the vernacular, as an enclave, tended to preserve the rich variability that the category must have had in the mid-nineteenth century.13 Finally, we might also mention the case of Berbice Dutch, "the only Caribbean Creole language for which the substrate input can be isolated with reasonable specificity" (ROBERTSON 1993: 310). The language has maintained a substantial number of morphological elements from its West African substrate, Ijo, among them, a perfect marker -te, (10) Berbice Dutch oorii gruii-te wiirunii. she grow (te) Wiruni 'She has grown up on the Wiruni.' corresponding precisely to Nembe -te.

13

These conclusions are also supported in TAGLIAMONTE 1996: the have-perfect in Ex-Slave narratives and Samanä English are shown to have the same distribution patterns and co-occurrence restrictions as those of Standard English. For example, HAVE + past participle occurs more often with the conjunction since, and almost never with when or after, just as in Standard English (1996: 384-389).

222 (11) Nembe togu bei duba-te boy the big has become 'The boy has become big.'

BRIDGET DRINKA

[ROBERTSON 1993: 302-303]

Like Afrikaans, this variety is a Dutch derivative, creolized, quintessentially vernacular, yet, unlike Afrikaans, the perfect category persists and thrives. It is, of course, not just the vernacular status of a given variety which is of concern to ABRAHAM & CONRADIE, but, in fact, the ensemble of factors - parsing and discourse pressures coupled with vernacular status - which is assumed to engender PS. Still, the question can be asked: if vernacular status were a true prerequisite to the change, why don't other non-standard varieties with similar parsing and discourse configurations undergo it, e.g., non-prestigious Plattdeutsch? Thus, while the disassociation of the vernacular from the Standard is a very reasonable contributory factor, such a detachment is clearly not sufficient independent motivation for such a change (nor do ABRAHAM & CONRADIE claim that it is such).14 4.5. Why is the perfect construed as a preterite? What is it that actually allows the perfect to be interpreted as a preterite? ABRAHAM & CONRADIE (2001: 94) reply that it was ultimately the urge, within these vernaculars, for subject marking to be accomplished early in the utterance which led to the reanalysis of main verb haben as a true auxiliary and to the reinterpretation of the construction as temporal and semantically equivalent to the preterite. What ABRAHAM & CONRADIE stress is that periphrastics succeed in meeting parsing and discourse challenges better than synthetic forms in OV languages, and so when a periphrastic solution becomes available, it will naturally be preferred. But this explanation still does not address the essential question: why did the perfect construction become semantically equivalent to the preterite? In attempting to respond to this question, we should first of all note that the shift of perfect to preterite should not be regarded as inevitable (pace BYBEE et al. 1994): languages like Greek have succeeded in maintaining a perfect category for several millennia (DRINKA 1998: 123-125), and English, despite its drastic morphological reduction, preserves the old anterior meaning for its perfects to this day. In addition, granted that the parsing and discourse pressures described here are credible factors in shaping the direction of the change, there is still nothing automatic about the way that these factors were implemented. It would have been a perfectly reasonable and workable solution, for example, for German speakers to give up their commitment to V-final word order, and simply become VO, as English did, if the discourse pressures had been strong enough.

14

In defense of the argument that vernacular status is essential, if not capable of motivating the change on its own, we should note that preterite loss and its replacement by the perfect was first attested not in the formal layers of Parisian French, but in the vernacular (FOULET 1920).

Präleritumschwund: evidence for areal diffusion

223

It is, in fact, a comparison of the situation in English with that of Southern German which will allow us to understand why the former did not change and the latter did. Like Middle High German, Middle English developed a new trend towards auxiliation, but it did not undergo the semantic shift and realignment of its major temporal-aspectual categories from three (present, perfect, preterite) into two (present, perfect > preterite, preterite > 0 ) , as Southern German did. ABRAHAM & CONRADIE's explanation for this fact is that the present perfect of Middle English kept its aspectual function: "Im Englischen konnte sich kein Präteritumschwund entwickeln, weil das (present) perfect wegen seiner aspektuellen Funktion kein Konkurrent für das Präteritum war und bis heute geblieben ist." (2001: 94) This explanation is circular and non-informative: the fact that English did not undergo a semantic re-evaluation of its perfect does not tell us why this did not occur. The question remains: why did Southern German and English behave so differently in their handling of the perfect? The answer I would like to propose is the following: English was not a member of the nuclear group which implemented this change; it was out of the realm of influence by the time the innovation occurred.15 Southern German was, as we shall see in §5, well within this realm. If increased reliance on periphrasis is to be regarded as part of the motivation for PS, English should represent a likely candidate for such a change: it has adopted auxiliation enthusiastically, filling its verb system with periphrastic forms, through the reinterpretation of its modals as auxiliaries in ME, its development and expansion of the progressive category, its whole-hearted adoption of ch /_e,i) ches meesmes journex a chele meesme iglise, but word-final , rare in Picardy in the 13TH century (EINHORN 1974: 137), does appear variably: devant diz 'aforementioned' is followed two lines later by the more expected devant dis. It is puzzling, however, that the document which contains the highest level of perfect use of these four, the document written in 1263 (5.4.1.3), also has the highest number of Picard traits. The explanation for this intriguing fact will require further investigation, and is beyond the scope of the current paper. Interestingly, two later documents concerning the same participants and the same issue appeared in Latin in 1242.

Präteritumschwund: evidence for areal diffusion

229

(15) Et cestes couvenenches a tenir ont fianchii les parties sus paine de chiunquante livres. (1, 16) 'The parties have promised/promised to keep this agreement on pain of 50 livres (i.e., at a penalty of 50 Fr. pounds).' 5.4.1.2. The second document, written in 1258, concerns an amiable agreement between the same abbey and Guillaume of Beaussault on the use of forest land given to the abbey (CAROLUS-BARRE 1964: 33-34). In referring to past contexts, the passage contains 3 present perfects, 1 preterite, and 1 (passive) imperfect, with a resultant ratio of 60% perfects vs. preterites and imperfects. Pluperfects are important for describing previous arrangements (5x); 4 passives appear, but no modal constructions. An example of apparent preterital use of the perfect can be found in the following: (16)

...je...les devant diz .XIII. journeux...leur ai escangies por autres .XIII. journex de bos que je leur en ai donne.. .(29, 5-7) Ί have exchanged/exchanged with them the aforementioned 13 pieces of land for another 13 pieces of forested land that I have given/gave them.'

5.4.1.3. The third French document, written in 1264, finalizes the sale of a piece of land and an alder grove by Lady Eustachie of Genvry to Saint-Eloi of Noyon (CAROLUS-BARRE 1964: 8 8 - 9 1 ) . It contains a remarkably high number of present perfects, 29, as opposed to 1 imperfect and no preterites, so that the percentage of perfects reaches 96.7%. There are also 7 passives and 8 modal constructions - altogether a strong showing of periphrastic forms. Here, clearly, the perfect is being used for preterite function, as signaled by the adverb, and especially by the fact that virtually no other tense is being relied upon to make past reference:22 (17) et ainsi lor aijou livre chel aunoy devise et esbonni de bounes certaines qi i sont mises. 'And then I gave them that alder grove, divided and separated by certain boundary markers which were set there.' (74, 6) 5.4.1.4. The fourth document, written in 1285, documents an exchange of vineyards between Νένείοη of Ronquerolles and the abbey of Lannoy (CAROLUS-BARRE 1964: 273-275). The text contains 9 present perfects, 11 imperfects, and no preterites, along with 2 passives and 2 modal constructions. Present perfects, though less per22

It should not be assumed that the increased number of perfects listed here operates along a simple trajectory: variability is to be found throughout these documents. For example, in a document written 20 years after this one, in 1284 (an agreement between Renaud of Equennes and the abbey of Notre-Dame of Breteuil concerning rent payment), there are no active present perfects. The writer rather relies on preterites (2x) and especially on imperfects (8x) to convey past meaning.

230

BRIDGET DRINKA

vasive than in the previous document, still represent almost half (45%) of verbs with past reference. (18)

...Ii devant dit religieus m 'ont baillie bien et souffisaument et delivre les vignes devant dites qui estoient leur.. .(189,15-16) 'the aforementioned monks have protected/protected me well and adequately and have given over/gave over the aforesaid vineyards which were theirs'

We can grasp from this small sampling that the perfect was not used exclusively in a preterital sense, but that it often did allow for this interpretation; this fact is especially clear when the perfect is virtually the only past tense used in a document. 5.4.2. German examples Turning to the parallel German documents, we see some similar patterns: 5.4.2.1. The first example is one of the earliest official documents to appear in German, written in 1264 on behalf of four arbitrators23 in a case concerning a final atonement between Archbishop Engelbert of Cologne and the city of Cologne ( B O H N & RAPP 1995b: 223-228). In this document, modal constructions predominate (36 instances) because of the nature of the decrees that the arbitrators are handing down, but in contexts where past tense is being referred to, periphrastic perfects predominate over synthetic preterites 1 8 : 3 . Alongside these, 22 stative passives also occur, which are formally linked to the perfects (copula + transitive PPP). In some cases of perfect use, an anterior meaning is still clearly intended: (19) Nu hain wir gehöret clage vnses herre(n) van koine inde antwurde der bürgere) inde derstede van koine... (2,4-2,5) 'Now have we heard complaints of our lord from Cologne and the reply of the citizens and the city of Cologne...' but in other examples, a preterital meaning seems preferable. (20)

23

Vort va(n) der hail sunen des unse herre ziet die burge(re)/dat vordere unse herre mit der Scheffene(n) vrdeile van kolne/alse id van aldirs her kumen w...(2, 15-2,16) 'Based upon the secret atonements of our lord, the citizenry sees that earlier our lord arrived from Cologne with the judgment of the lay assessors, that is from the elders (i.e. city fathers) .'

The arbitrators involved were Heinrich, Bishop of Lüttich, Gerhard, Bishop of Münster, Count Otto of Gelder, and Count Wilhelm of Jülich (1995b: 223); the scribe was Gottfried Hagen (BOHN & RAPP 1995b: 254).

Präteritumschwund: evidence for areal diffusion

231

Most cases are quite ambiguous; there are few temporal adverbs to provide us with a clue as to the intended reference. But the sheer percentages (86% perfects : 14% preterites) indicate a noteworthy tendency towards the use of the perfect in past context, and the document as a whole demonstrates a clear preference for periphrastic forms (modal constructions, stative passives, and periphrastic perfects) over synthetic ones (preterites). 5.4.2.2. The second document, Regulations for Cloth-cutters of Cologne (BOHN & RAPP 1995b: 228-231), was written in 1270 to deal with such concerns as guild jurisdiction, regulation of competition, certification rules for journeymen, etc. In this document, the perfects and preterites are equally represented with 4 instances each, making the percentage of perfects as opposed to preterites 50%. Alongside these are 5 stative passives and 3 modal constructions. The contrast of the ratio of perfects to preterites with that of the previous document is rather remarkable, all the more so since both documents were written by the same scribe, Gottfried Hägen. BOHN & RAPP (1995b: 255; 278) note, however, that, while Gottfried's learned hand is evident in particular spelling conventions and certain calligraphic features, there is clear evidence for a second hand in the transcription, with strong dialectal features emerging (e.g., mat and last for mit and list, as occurs in the dialect of Luxemburg and along the Mosel). It is difficult to say whether the participation of two scribes in the production of the document affected the distribution of perfects. But it was clearly Gottfried's hand which produced the following example, and also his hand which made the subsequent emendations (noted here by my underlining): (21) Jn vrkunde inde in vestinninge aller dieser vorsprochginre dinge / inde vp dat si stede ewelichge bliuen / so haint unse zwene meistere her Diederich van me hirze / inde her Cunrait ranke / inde der gezide diese vorworde bescrieven sint /van werscheffe des Raitz vnd(er) den seddemen iere ingesiegele an diesen briefgehangin. (3,32-3,35) '...thus (the following people) have appended their seals to this letter: our two guildmasters Herr Dietrich of Hirz and Kondrad Ranke, and the assembly who are described in the forward (of this document), (and) from the defenders of the city council and the city accountants.' What is so remarkable about this example is the expanse of material which appears between auxiliary and participle, with a phrase even added to the braciated segment. Whatever the reason, it must be admitted that, while this document shows substantial use of the perfect, it does not provide the striking evidence for preference for the perfect that the others do. 5.4.2.3. The last will and testament of Mechthild, Countess of Sayn, written in 1283, was one of the first testaments to be written in German (BOHN & RAPP 1995b: 241-

232

BRIDGET DRINKA

245; 258-259).24 In this document, modal constructions are again preferred, since the testament concerns, above all, the future disposition of Mechthild's possessions. Nevertheless, 25 examples of HAVE + PPP appear, next to 6 preterites, making an impressive ratio of 81% perfects : 19% preterites. (22) .. .alle mine scholt gelden die ich selvegemacht haven.. .(7.2) 'all my owed money that I myself have made' Significantly, most of the preterites are the verb be, and two of these occur in the formulaic first and last lines. In addition, as we saw in the earlier document, a large number of stative passives (13) also occur; again, there is a clear preference for periphrastic forms over synthetic ones in this document. 5.4.2.4.

The fourth document, written in 1293 (BOHN & RAPP 1995b: 246-249), is virtually a copy of the 1270 document (5.4.2.2 above), with an additional section adding 10 more regulations to the earlier 13. This new section is especially useful for purposes of comparison, since it allows us some insight into the changes which had occurred since the first document was composed.2 In this 18-line section, modal constructions predominate (14 instances), but in preterital contexts, only perfects appear (3x), never preterites. While it would be unwise to place too much weight on the evidence from such a short passage, it is still worth noting that periphrastic forms are markedly preferred over synthetic ones. Also fascinating to note are the emendations made to the old document in this newer version: the passage quoted above with an enormous braciated segment has been revamped so that the auxiliary and the participle occur next to each other: (23) ...so hauint anegehangin vnse mest(er)e / her Simon van me Numarte / inde herJohan van Schildergassin. (8,45-6) '...thus have Herr Simon van me Numarte and Herr Johan van Schildergassin appended.' This example represents an isolated but rather remarkable counterexample to claim: V-final order has been given up, yet preterital meaning is apparently being implied. To summarize what can be learned from these documents, while we would not wish to make sweeping generalizations based on this limited data, it is clearly the case that, in the second half of the 13th century, both in Northern French and in Western German, the perfect is a viable, productive category, well-represented in preterital contexts. Indisputable evidence for a preterital interpretation of these forms is not available, but the percentage of perfects in comparison to other past ABRAHAM & CONRADIE'S

24

25

An interesting, though perhaps inconsequential, relation to French is noted by BOHN & RAPP (1995b: 259, footnote 291): French noblewomen had been producing such testaments since the 1240's. They also note that Mechthild's scribe, "Lambert", used remarkably consistent spelling, and was clearly well-educated in the writing of German, both for official documents, but probably also for literary manuscripts (1995b: 280-281). The new section was probably not written by the same scribe as the 1270 document (BOHN & RAPP 1995b: 255).

Präteritumschwund: evidence for areal diffusion

233

forms in both locations is uniformly high, usually above 50% of possible cases, and so demonstrates a propensity towards the use of perfect in past contexts. The similar pattern of preference in these two areas is surely not coincidental; it represents, I would claim, suggestive evidence that both locations derived their patterns from a single source: the cultural and intellectual center of Paris.

6. Conclusion A synopsis of the arguments presented above will clarify the claims made here: 1. The Parisian vernacular was the earliest variety in Europe to have periphrastic perfects replacing synthetic preterites as markers of the past. 2. Western European languages which have undergone this shift are contiguous (or, in the case of Yiddish and Afrikaans, developed PS on the model of a nuclear language). 3. As part of this cultural and linguistic diffusion, French influence reached the Western German territory in the 12-14th centuries. 4. This influence soon spread to the southeastern German area, but did not extend to the Northern German area. 5. Perfects in preterital contexts appeared in Western German dialects even earlier than in southeastern ones. We can therefore assume with some confidence that it was French influence which was ultimately responsible for the new morphosyntactic patterns which developed in the dialects of Cologne, Trier, and Strasbourg, before spreading to the more easterly dialects of Augsburg and Nürnberg. On a more general note, I would like to suggest that there is a point of nexus between the model proposed here and that of ABRAHAM & CONRADIE, in that the adoption of auxiliaries in verb-second position in German has clearly had the effect that ABRAHAM & CONRADIE claim it has (2001: 123), in making subject congruence more immediate, and in easing and speeding up comprehension, but that this tendency towards partial SVO word order has developed with the help of an external model, through the adoption of Latinate/Romance style auxiliation. Thus, I suggest that we view the developments described in this paper as part of an ongoing series of innovations which have flowed from Latin and Romance into the Germanic languages, beginning with the full adoption of auxiliation, including the development of the perfect as a tense, followed by the later reinterpretation of the perfect as a preterite in a limited, geographically defined area, according to the model of French.

234

BRIDGET DRINKA

References Präteritumschwund und Diskursgrammatik Präteritumschwund in gesamteuropäischen Bezügen: areale Ausbreitung, heterogene Entstehung, Parsing sowie diskursgrammatische Grundlagen und Zusammenhänge. Amsterdam & Philadelphia: Benjamins. ALBERTS, WERNER ( 1 9 7 7 ) : Einfache Verbformen und verbale Gefüge in zwei Augsburger Chroniken des 15. Jahrhunderts. Göttingen: Vandenhoeck & Ruprecht. ANDERSEN, HENNING ( 1 9 8 7 ) : From auxiliary to desinence, in: HARRIS, MARTIN & RAMAT, PAOLO (eds.), Historical development of auxiliaries. Berlin, New York & Amsterdam: Mouton de GruyABRAHAM, WERNER & CONRADIE, C . JAC ( 2 0 0 1 ) :

ter, 2 1 - 5 1 . BAUM, RICHARD ( 2 0 0 0 ) : 1107-1117.

Französisch als dominante Sprache Europas, in:

BERTELOOT, AMAND ( 1 9 9 5 ) : Das Mittelniederländische als GÄRTNER, KURT & HOLTUS, GÜNTER (eds.), 1 7 3 - 1 9 6 .

BESCH, WERNER

Urkundensprache im

13.

et al. (eds.),

Jahrhundert, in:

(eds.) (2000): Sprachgeschichte: Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. 2nd ed. Part II. Berlin & New York: Walter de Gruyter. BOHN, THOMAS & RAPP, ANDREA (1995a): Zur Untersuchung westmitteldeutscher Urkunden des 13. Jahrhunderts, in: GÄRTNER, KURT & HOLTUS, GÜNTER (eds.), 40-59. BOHN, THOMAS & RAPP, ANDREA (1995b): Nachträge zum 'Corpus der altdeutschen Originalurkunden' mit Editionen und Untersuchungen, in: GÄRTNER, KURT & HOLTUS, GÜNTER (eds.), 215-283. BYBEE, JOAN; PERKINS, REVERE & PAGLIUCA, WILLIAM ( 1 9 9 4 ) : The evolution of grammar: tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press. CAROLUS-BARR£, LOUIS ( 1 9 6 4 ) : Les plus anciennes chartes en langue frangaise. Paris: Klincksieck. COMRIE, BERNARD (1991): Comment: Yiddish is Slavic?, in: International Journal of the Sociology of Language 91, 151-156. DAHL, OSTEN (ed.) ( 2 0 0 0 ) : Tense and aspect in the languages of Europe. Berlin & New York: Mouton de Gruyter. DENTLER, SIGRID (1998): Gab es den Präteritumschwund?, in: ASKEDAL, JOHN OLE (with assistance of ZICKFELDT, A . W . ) (ed.), Historische germanische und deutsche Syntax. (Osloer Beiträge zur Germanistik 21). Frankfurt a. M.: Lang, 133-148 Dow, JAMES R. & STOLZ, THOMAS (1991): Comment: The Sorbian origins of Yiddish: Linguistic theory in search of historical documentation, in: International Journal of the Sociology of Language 91, 157-165. DRINKA, BRIDGET (1998): The evolution of grammar: Evidence from Indo-European perfects, in: SCHMID, MONIKA; AUSTIN, JENNIFER & STEIN, DIETER (eds.), Historical Linguistics 1997. Amsterdam & Philadelphia: Benjamins, 117-133. DRINKA, BRIDGET ( 2 0 0 3 ) : Areal factors in the development of the European Periphrastic Perfect, in: Word, April 2 0 0 3 . EGGERS, ECKHARD (1998): Sprachwandel und Sprachmischung im Jiddischen. Frankfurt a. M.: Lang. EINHORN, Ε . ( 1 9 7 4 ) : Old French: A concise handbook. Cambridge: Cambridge University Press. FISHER, JOHN (1996): The emergence of standard English. Lexington, KY: University of Kentucky Press. FOULET, LUCIEN (1920): La disparition du preterite, in: Romania 46,271-313. GÄRTNER, KURT ( 1 9 9 5 ) : Zur Erforschung der westmitteldeutschen Urkundensprachen im 13. Jahrhundert, in: LERCHNER, GOTTHARD; SCHRÖDER, MARIANNE & Fix, ULLA (eds.), Chronologische, areale und situative Varietäten des Deutschen in der Sprachhistoriographie. Frankfurt a. M.: Peter Lang. GÄRTNER, KURT & HOLTUS, GÜNTER (1995): Einführung in das Projekt 'Westmitteldeutsche und ostfranzösische Urkunden- und Literatursprachen im 13. und 14. Jahrhundert', in: GÄRTNER, KURT & HOLTUS, GÜNTER (eds.), 1-39. GÄRTNER, KURT & HOLTUS, GÜNTER (eds.) ( 1 9 9 5 ) : Beiträge zum Sprachkontakt und zu den Urkundensprachen zwischen Maas und Rhein. Trier: Verlag Trierer Historische Forschungen. GVOZDANOVI0, JADRANKA ( 1 9 9 5 ) : Western South Slavic tenses in a typological perspective, in: THIEROFF, ROLF (ed.), 1 8 1 - 1 9 4 . BESCH, WERNER; BETTEN, ANNE; REICHMANN, OSKAR & SONDEREGGER, STEFAN

Präteritumschwund: evidence for areal diffusion

235

L'enseignement de l'ecriture aux universites medievales. (2nd expanded ed.) Budapest: Maison d'Edition de l'Acadimie des Sciences de Hongrie. HARRIS, MARTIN (1978): The evolution of French syntax: A comparative approach. London & New York: Longman. HARRIS, MARTIN ( 1 9 8 2 ) : The 'past simple' and the 'present perfect' in Romance, in: VINCENT, NIGEL & HARRIS, MARTIN (eds.), Studies in the Romance verb. London and Canberra: Croom Helm, HAJNAL, ISTVÄN & MEZEY, L Ä Z L 0 ( 1 9 5 9 ) :

42-70. HERRMANN, HANS-WALTER ( 1 9 9 5 ) :

Volkssprache und Verwaltung in Oberlothringen im Spätmittelalter und der frühen Neuzeit, in: GÄRTNER, KURT & HOLTUS, GÜNTER (eds.), 1 2 9 - 1 7 1 . LABOV, WILLIAM (1972): Language in the inner city: Studies in the black English vernacular. Philadelphia: University of Philadelphia Press. LINDGREN, KAJ Β . (1957): Über den oberdeutschen Präteritumschwund. Helsinki: Suomalaisen Tiedeakatemian Toimituksia. Ser. B, Tom 112,1. LÖTZSCH, RONALD (1995): Das sorbische Tempussystem, in: THIEROFF, ROLF (ed.), 167-179. MATTHEIER, KLAUS ( 2 0 0 0 ) : Die Herausbildung neuzeitlicher Schriftsprachen, in: BESCH, WERNER et al. (eds.), 1 0 8 5 - 1 1 0 7 . MUFWENE, SALIKOKO S. (ed.) ( 1 9 9 3 ) : Africanisms in Afro-American language varieties. Athens & London: University of Georgia Press. ÖHMANN, EMIL ( 1 9 3 1 ) : Der französische Einfluss auf die deutsche Sprache im Mittelalter, in: Neuphilologische Mitteilungen 32,195-220. ÖHMANN, EMIL (1951): Die mittelhochdeutsche Lehnprägung nach altfranzösischem Vorbild. Helsinki: Suomalaisen Tiedeakatemian Toimituksia. Ser. B, Tom 68,3. PHEIFFER, R . H . ( 1 9 8 0 ) : Die gebroke Nederlands van Franssprekendes aan die Kaap in die eerste helfte van die agtiende eeu. Kaapstad: Academica. ROBERTSON, IAN ( 1 9 9 3 ) : The Ijo dement in Berbice Dutch and the Pidginization/Creolization process, in: MUFWENE, SALIKOKO S. (ed.), 2 9 6 - 3 1 6 . SANTORINI, BEATRICE ( 1 9 9 3 ) : Jiddisch als gemischte OV/VO-Sprache, in: ABRAHAM, WERNER & BAYER, JOSEF (eds.), Dialektsyntax. Opladen: Westdeutscher Verlag, 2 3 0 - 2 4 5 . SCHILDT, JOACHIM ( 1 9 6 8 ) : Zur Ausbildung des Satzrahmens in Aussagesätzen der Bibelsprache ( 1 3 5 0 - 1 5 5 0 ) , in: Beiträge zur Geschichte der deutschen Sprache und Literatur 9 0 , 1 7 4 - 1 9 7 . SCHNEIDER, EDGAR W. (1993): Africanisms in the grammar of Afro-American English: weighing the evidence, in: MUFWENE, SALIKOKO S. (ed.), 209-221. SMITHERMAN, GENEVA ( 1 9 7 7 ) : Talkin and testifyin: the language of Black America. Boston: Houghton Mifflin. SQUARTINI, MARIO, & BERTINETTO, PIER MARCO ( 2 0 0 0 ) : The simple and compound past in Romance languages, in: DAHL, OSTEN (ed.), 4 0 3 - 4 3 9 . STANKIEWICZ, EDWARD ( 1 9 9 1 ) : Comment, in: International Journal of the Sociology of Language 9 1 , 205-213.

A. (1996): Has it ever been 'perfect'? Uncovering the grammar of early Black English, in: York Papers in Linguistics 17, 351-396. TAGLIAMONTE, SALI A. (2000): The grammaticalization of the present perfect in English: Tracks of change and continuity in a linguistic enclave, in: FISCHER, OLGA; ROSENBACH, ANETTE & STEIN, DIETER (eds.), Pathways of change: grammaticalization in English. Amsterdam & Philadelphia: Benjamins, 329-354. TERNES, ELMAR (1988): Zur Typologie der Vergangenheitstempora in den Sprachen Europas (synthetische vs. analytische Bildungsweise), in: Zeitschrift für Dialektologie und Linguistik 55, 332341. THIEROFF, ROLF (1995): More on inherent verb categories in European languages, in: THIEROFF, ROLF (ed.), 1-35. THIEROFF, ROLF (ed.) ( 1 9 9 5 ) : Tense systems in European languages II. Tübingen: Niemeyer. THIEROFF, ROLF ( 2 0 0 0 ) : On the areal distribution of tense-aspect categories in Europe, in: DAHL, OSTEN (ed.), 2 6 4 - 3 0 5 . VEKEMAN, HERMAN & ECKE, ANDREAS (1992): Geschichte der niederländischen Sprache. Bern: Pete· Lang. WEXLER, PAUL ( 1 9 9 1 ) : Yiddish - The fifteenth Slavic language. A study of partial language shift from Judeo-Sorbian to German, in: International Journal of the Sociology ofLanguage 9 1 , 9 - 1 5 0 . TAGLIAMONTE, SALI

Appendix

Karte 2. »Die Vergangenheitstempora in den Sprachen Europas« (nach Ternes 1988:340) Map 1: (from TERNES 1988: 3 4 0 )

238

BRIDGET DRINKA

LSrb

3 t It

ANT:PRET ^

^

ANT:AOR:IMPF

present anterior > past development, stages 1 - 3 present anterior > past development, stage 3

other languages

G P S T (stage 4 )

Map 2: (from THIEROFF 2000: 285)

239

Appendix

Ice Nor Swd

Dan Eng

Frs Dut NGrm

(Yidd USrb

Glc Leon Rum Prt Cast

Ctl

Stlt [Sic] [Cal]

Bold

Italics Underline [Brackets]

|b /

Med Grk

Bulg

= preterite E.Bsq: preterite; S. Bsq, including StBsq: anterior; N.Alb (Gheg): preterite; S.Alb (Tosk): anterior; = anterior = durative or iterative in near past = present state resulting from past action, as in Vulgar Latin

Map 3: Semantic value of periphrastic perfect

240

French

1241 1258 1264 1285 Table 1:

BRIDGET DRINKA

Percentage Perfect Preterite Imperfect Passive Modal ofPerf. vs. (passe (ρβ$$έ Pret + Imperf. compose) simple) 37.5% 3 1 4 8 9 60% 1 1 4 0 3 1 96.7% 29 0 7 8 11 2 2 45% 9 0 Distribution of temporal-aspectual categories in 4 representative French documents from the archives of Oise Modal Percentage of (Stative) Perfects vs. Pret Passive 22 36 85.7% 18 3 4 4 5 3 50.0% 22 81.0% 25 6 13 3 0 14 100.0% 0 Distribution of temporal-aspectual categories in 4 representative German documents from Cologne and nearby locations

German 1264 1270 1283 1293 Table 2:

Perfect Preterite

The European demise of the simple past and the emergence of the periphrastic perfect: areal diffusion or natural, autonomous evolution under parsing facilitation? WERNER ABRAHAM (Wien)

0. Introduction There is a strong tendency in recent linguistic typological research to see several common or very similar, phenomena in the European languages as having come about under areal diffusion of one sort or another. Thus, the specific formation of discontinuous and postverbal negatives appears to have diffused out from a core area of Germanic to fringe areas of Romance and Celtic (RAMAT & BERNINI 1990); according to HEINE (1994), the comparative construction seems to have much in common throughout European languages; likewise, the formation of tenses and aspects has proved to be similarly marked (DAHL 1996, 2000); HASPELMATH (1999) has recognized similar convergences of the categories of the article, the external possessive, and the passive construction; and, finally, KUTEVA (2001) concludes that the auxilial verbal constructions show convergence in the main European area. The interesting factor about these attempts is that they rest on areal contingency in Central Europe (see THIEROFF 1995, 2000; DAHL 2000). Yet, no explanation in terms of the real historical national-linguistic influences has ever been provided. Of all European languages, the Germanic representatives have been seen to form the core of diffusing developments: French, German, Dutch, and Northern Italian (the latter is only a dialectal area!); a less central area, comprising English, Swedish, Standard Italian, Czech, which nevertheless shares with the core many of the features subsumed under the SAE characteristics; and, finally, a peripheral area adopting the fewest of these common features (Russian, Hungarian, Basque). While any areal Sprachbund diffusion is a necessary prerequisite for any such assumption, it is in equal need of showing in historical, longitudinal reality the linguistic transfer from the centers of diffusion to the receiving languages. Attempts relating ie-languages (i.e., languages which use etre/essere/be/sein/byt', alone or alongside avoir/avere/have, for the periphrastic perfect), to their underlying ergative nature, are clearly mistaken (ABRAHAM 1995,1988, 2001, 2002). Of all such endeavours, DRINKA (2003) is a notable exception. This is what she espouses. First, she highlights the fact (noted by DAHL and THIEROFF, among others) that only the west of the European language community developed have as an auxiliary, while be is dominant in the east. Second, the preference for have in the west is explained through inheritance from Late Latin and its model from Ancient Greek, in particular its use of the past participle, which became an integrated part

242

WERNER ABRAHAM

of the newly developing periphrastic perfect - something that the older Germanic languages had not known or used only as secondary, resultative, predications in the first place (ABRAHAM 1990). Finally, DRINKA claims, the more recent shifts from resultative secondary predications to simple past functions emerged from early Modem Parisian French and diffused from there. What this boils down to is another claim, albeit somewhat better motivated, to the status of an SAE convergence. The present paper will contest part of that by showing that the spread of the analytic past has heterogeneous bases. Simple pasts have disappeared in widely adjoining parts of Europe, in particular the southern and eastern parts (South Germanic, Romance, Slavic), and periphrastic preterits ('present anterior') have taken over their function. Yet, the consequences in their respective languages are different. One can distinguish 4 types of preterit-loss languages and dialectal areas within these. The classifying criteria are: ((0) present anterior different from simple past: Scandinavian, English, a.o.s); (1) present anterior has the same function as the simple preterit: Standard German and Italian; (2) the simple preterit has become very marginal: Serbo-Croatian, Rumanian, French; (3) no simple preterit exists any longer; its function is taken over fully by the present anterior: South German, Yiddish, Slavic. The latter group of languages can be subdivided according to whether or not they have a paradigmatic representation of the periphrastic future perfect and/or a modal usage linked to the present anterior: South German vs. (4) Slavic. All of this is suggestive of areal spreading, the source being those languages which have advanced the farthest on this cline of preterit demise and newly acquired properties of the periphrastic perfect (which is Russian).

1. The recent SAE-fashion There is a strong tendency in recent linguistic typological research to see several common or very similar phenomena in the European languages as having come about under areal diffusion of one sort or another (HASPELMATH 1 9 9 9 ) . Thus, the specific formation of discontinuous and postverbal negatives appears to have diffused out from a core area of Germanic to fringe areas of Romance and Celtic (RAMAT & BERNINI 1990). According to HEINE ( 1 9 9 4 ) , the comparative construction has much in common throughout European languages. Likewise, the formation of tenses and aspects has proved to be similarly marked (DAHL 1 9 9 6 , 2 0 0 0 ) . HASPELMATH ( 1 9 9 9 ) has recognized similar convergences of the categories of the article, the external possessive, and the passive construction; and, finally, KUTEVA ( 2 0 0 1 ) concludes that the auxilial verbal constructions show convergence in the main European area.1 The interesting factor about all these attempts is that they rest on areal contingency in Central Europe (see THIEROFF 2 0 0 0 , THIEROFF & BALL1

There are numerous attempts at defining sub-European areas of convergence. See, among the recent ones, RAMAT & STOLZ'S (2002) collection of a variety of articles on possible convergences in the Mediterranean countries including the non-Indo-European, North African, vernaculars.

The European demise of the simple past

243

WEG 1994, and DAHL 2001 listed in the references below). Yet, no realistic scenario in terms of the real historical national-linguistic influences and exchanges has ever been unfolded (see, for a principled discussion, of the notion of Sprachbund in terms of "mature" vs. "less mature" languages: DAHL 2001, STOLZ 2002). Of all European languages, the Germanic representatives have been seen to form the core of diffusing developments. One such group is constituted by French, German, Dutch, and Northern Italian (the latter only as a dialectal area!). Another areal group is constituted by a less central area comprising English, Swedish, Standard Italian, and Czech, which nevertheless shares with the core many of the features subsumed under the SAE characteristics. Finally, a peripheral area is characterized by adopting the fewest of these common features (Russian, Hungarian, and Basque). While any areal Sprachbund diffusion is a necessary prerequisite for any such assumption, it is in equal need of showing in historical, longitudinal reality the linguistic transfer from the centres of diffusion to the receiving languages. Attempts relating öe-languages (i.e., languages which use etre/essere/be/sein/byt', alone or alongside avoir/avere/have, for the periphrastic perfect), to their underlying ergative nature (SHANNON 1990, among others), are clearly mistaken (ABRAHAM 1990, 1995, 2001); see below for additional arguments. See STOLZ (2002) for a principled attempt at clarification of the notion of Sprachbund. Of all such endeavours, DRINKA (2003) is a notable exception. This is what she espouses. First, she highlights the fact (noted earlier by DAHL and THIEROFF, among others) that only the west of the European language community developed have as an auxiliary, while be is dominant in the east. Second, the preference for have in the west is explained through inheritance from Late Latin and its model from Ancient Greek, in particular its use of the past participle. The latter form became an integrated part of the newly developing periphrastic perfect, something that the older Germanic languages had not known or had used only as secondary, resultative predications in the first place (ABRAHAM 1990). Finally, DRINKA claims, the more recent shifts from resultative secondary predications to simple past functions emerged from early Modern Parisian French and diffused further from there. It is easy to see that what this boils down to is another claim, albeit one somewhat better motivated, to the status of an SAE convergence. The present paper will contest that, not only by casting doubt on details of DRINKA's claims. The prime direction of my argument will be that there are natural processes of parsing that will overwrite principles of logical grammatical constituency (i.e., speaker-hearer merging grammatical principles), primarily in the spoken/aural codes of languages. One such principle that will be invoked is that of early-on identification of the clausal parts (subjects, objects) by verbal agreement and the valency selection of the predicate (in terms of HEIM's staple mechanism). According to this type of parsing-by-stapling, an optimal language is SVO - in other words, in SVO the agreement features are processed sentence initially to distinguish from objects and adjuncts. For SVOV-languages (such as German, Dutch, and West Frisian), however, it is advantageous to split the predicate into a finiteness-realizing auxiliary (to secure early subject identification by agreement features) and a lexical predicate to fill the last clausal position. (See Table 1 in the Appendix).

244

WERNER ABRAHAM

Only the latter verb slot distribution, Vlast (2nd VB), in union with AUX/V2 (1st VB), allows for scrambling to order the elements of the clause in discoursemeaningful ways. German, in particular South German (because of its dialectal, i.e. strongly aural/spoken character), is thus a language which has split its structural grammatical devices between two strategies of optimization. The parts-of-clause identification is organized by strict AUX/V2, while discourse-optimization is achieved in terms of the structural space between AUX/V2 and Vlast. Only this specific structural space allows for scrambling, which in turn makes discourse prominence virulent. It is obvious that this online-identification and discourse prominence has nothing to do with any type of convergence with the properties of any other language surrounding German, Dutch, and West Frisian.2 In other words, if Modern Standard English (MStE) is co-classified with the languages identified under SAE, then the fundamental differences with the continental West Germanic languages are unproportionally specified. If, however, for MStE exception is taken with respect to SAE, then the equally fundamental and pervasive structural similarities with the Scandinavian languages, no doubt truly SAE, remain outside of consideration. Certainly, this makes a reassessment of SAE at least with respect to MStE and the rest of the Germanic languages necessary, in terms of an autonomous development of many common phenomena. Strong support for such an autonomous development is provided by North Italian, Yiddish, and Afrikaans. All these languages and their vernaculars have been, or were, respectively, oral codes (dialects, regiolects, non-standards (non-written)). Their written codes unmistakably bear different structures (Standard Italian, Standard German, Hebrew, Dutch). We will investigate the requirements of oral codes in greater detail. But, first, let us examine more precisely what DRLNKA's Sprachbund-position is about.

2. Focussing the European periphrastic perfect: the auxiliation tripartition Map 1 (see Appendix) shows the distribution between have and be auxiliation in periphrastic perfects in Central Europe (taken from D R I N K A 2003: 3). The modern situation is that have-only auxiliates in the west, be as well as have in the middle, and be-only in the east. The distribution shows also that there is an east-west cline of be vs. have use for the periphrastic perfect. Striking, as this is, this polar areal distribution is added to by a transient area, which features both be and have. This transient area comprises almost all languages of the Germanic branch (with Yiddish as an exclave nowadays), almost all of the Romance languages, Basque and, furthermore, Albanian, Macedonian, and Bulgarian. What is striking, beyond the East-West cline of the have vs. be distribution, is the centre 2

If Old English had not become a substratum to Norman French, and strongly creolized subsequently by it, it would have greatly preserved the features that Modem German shows nowadays.

245

The European demise of the simple past

area of transient have-as-wel\-as-be representations comprising almost all Germanic languages - including Yiddish as an exclave nowadays - almost all Romance languages, furthermore Albanian, Macedonian and Bulgarian, as well as Basque. See, in (la-c)-{2), German, Dutch, and Italian, as representatives of this latter AUX-converging group, as opposed to Swedish and English representing the only-Aave-group. (la)

Intransitive situatives Wer ist da gelegen?

German

who is there lain

(lb)

Wie is daargaan liggen?

Dutch

who is there gone lie

(lc)

Eandato

aViaczenza.

Italian

is.3SG gone to V.

(Id) Han er gäet i byen

Danish

he is gone to buy

(2a) Han hargätt

Swedish

he has gone

(2b) He has traveled to London

English

Agentive transitives Wer hat dich das gelehrt? who has you this taught Wie heeft je dat geleerd? who has you that taught Ho visto Roma has.3SG seen Rome Jeg har kept kagen I have bought the cake Vem har lärt dig det? who has taught you that Who has taught you that?

3

Yiddish - but not Afrikaans (see ABRAHAM & CONRADIE 2 0 0 1 : ch. 2, in particular p. 28) - has mostly preserved the limited class of zayn-verbs, also in its diaspora inside the Slavic and Balkan linguistic environments: er iz gegangen heis-walked/gone 'went', iz geforn is-driven/gone by car 'drove/went', iz geblibn he is-stayed 'stayed', etc., vs. er hot geshribn he-has-written 'wrote', hot farkoyft hehas-sold 'sold', etc. Note also: er iz gekumen he-is-come 'he came' vs. er hot bakumen he-has-become 'he got', but er iz bagangen aleyn-mord he-has-committedalone-murder 'he committed suicide'. 4 Not unexpectedly, however, two major geographic areas for Yiddish where have-AUK has tended towards generalization as the sole past-tense AUX are Northeastern Yiddish er hot gekumen he-has-come 'he came', etc., as well as among some language-attriting speakers in America (see LEVINE 2 0 0 0 ) . DRINKA (2003: 3f.) spends some space and time to point out Slavic and Baltic languages, which lost their be-AUX in the course of time, but had been part of an older fee-only area together with the Germanic languages (the latter conclusion does not come forward explicitly in DRINKA'S paper, but must be assumed nevertheless, as will presently be seen). What DRINKA (2003: 4ff.) assumes as an historical socio-cultural tenet is that the spread of have as an AUX started from the Isle de France and went east to intrude in Germanic (and, to a much lesser extent, into Russian; ISACENKO 1974: 50) in terms of grammatical caiques and, from German further east, into Slavic (Slovak,

3 4

Thanks go to NEILL JACOBS, OSU. Such zayn-periphrastics as have been observed in Eastern Yiddish as boydekzayn 'to check', maskem zayn 'to agree' are really adjectival phrases; see the present tense: er iz boydek, er iz maskem, while the past tense is: er hot boydek geven, er hot maskem gevert (geven is past participle of 'be').

246

WERNER ABRAHAM

Czech, Upper Sorbian: ISACENKO 1974: 5 7 1 - 5 7 3 ) , while Macedonian appears to have developed its Aave-perfect under the influence of (first Byzantine, later Modern) Greek (VASILEV 1968: 2 2 4 ) . The process was a grammatical caique. The borrowing language did not simply take over the auxilial lexical from the borrower, but reanalyzed the semantic and functional properties of lexicals in their own language in accordance with the pattern of the borrower. The grammatical function was borrowed, not the foreign form itself. The most persuasive cases of such calquing come from Northern Russian dialects, Breton Celtic, and Basque. Northern Russian dialects do not share the use of have as an AUX with Standard Russian and appear to have borrowed from Finnish or from Low German in the course of Hanse influence (the centre of emanation being Hanse-Novgorod; cf. VASILEV 1968: 227; PANZER 1984); no other Celtic language shares the Aave-periphrasis with Breton, which appears to have calqued the form from surrounding French; and Basque, most unrelated to Romance, has developed the Aave-periphrasis with transitives (but retained its original be-AUX in the possessive construction with intransitives). The questions to be asked and answered in this context of assumptions are the following ( D R I N K A is aware of questions 1 - 3 , but has not touched upon 4 - 6 ) : 1. Why did the languages in question (above all, Germanic) not retain the beperfect/periphrastics in the first place? What made the Aave-perfect superior to the fee-perfect such that it was the Aave-perfect that spread east, but the feeperfect spread west and took over Romance? 2. How and why did the Aave-perfect arise in the first place? 3. Where, and in which historical layers, did the Aave-perfect come from originally? 4. Why has Afrikaans, exclave to its former source language in European context, Dutch, developed only the Aave-periphrasis, while its mother tongue, Dutch, has retained, and even optimized grammatically, the old division of be- and haveperiphrasis, along with German and all of its dialects? 5. How come Yiddish retained strictly the German be- and Aave-division in the diaspora and in a pure fee-environment (Russian)? Would we not expect, under the force of DRLNKA's convergence theory that East Yiddish would give way first? But it did not. Given 5 above: Is there a good reason for the fact that Afrikaans, in the diaspora from her mother language just as Eastern Yiddish, gave up the have/be AUX-distinction in the verbal lexicon completely, while Yiddish retained it - the latter, as appears, even extending it according to a strict aspectual parameter (much like Dutch, by the way - thus, this is not even to be taken as an idiosyncratic development of either language)? 6. Why did South German lose its simple preterit, while Standard German as well as Low German and Dutch (and West Frisian) retained it? Would we not expect all of German, after the encompassing force of LUTHER'S bible translation and its wide publicity after GUTENBERG'S introduction of printing, to caique its periphrasis as a whole? D R I N K A (2003: 7ff.) has an answer to part of the last question in 3: Greek played a fundamental role in the development of Late Latin, Romance, and other

The European demise of the simple past

247

European perfects; and Parisian French ultimately triggered the emergence of preterit function of periphrastic perfects in a number of European languages. What used to be a bipartition originally, i.e., a division between be-only languages in the east and Aave-representatives in the west, developed to be a tripartition in the sense described above. In what follows the view espoused by DRINKA and the others on whom she bases her conclusions will be confronted by a different view - the view that the periphrastic perfect is an autonomous development triggered by non-written, oral/aural codes. The crucial triggering criterion is that the Aave-periphrasis has psychological advantages in spoken, co- and context-bound, online discourse. This is an autonomous criterion, which may - but need not, much in dependence on the degree with which a language is codified according to grammarians' standards or not - lead to the demise of the simple preterit and the emergence of a periphrastic substitute. We expect, on the basis of this, that any periphrasis will be the preferred preterit expression in non-written, oral/spoken codes. However, this poses difficulties in languages with V-last positions, like German, Dutch, and Frisian, simply because there the periphrasis severely impedes the parsing possibilities of the linear clausal chain. It has been argued in ABRAHAM & CONRADIE (2001; see also ABRAHAM 1999a, b, c) that in languages with SVOV-order the wide middle field between the two complementary V-positions allows for scrambling and, as a consequence, for considerable discourse-functional reordering between objects and subjects as well as adjuncts. This is a tremendous advantage for aural decoding, since, first, other than in a written code, agreement identification cannot rely on reappearance of any single code, and, second, the identification and distinction of topics (themata) vs. comments (rhemata) is of utmost importance for speakers and hearers involved in the discourse. Standard German is such a discourse-prominent language. Its dialects and spoken varieties, that share this coding property, however have gone one step further by they adding a strategy to facilitate the identification of subject-predicate agreement better than Standard German (which relies heavily on the written code). I will be more specific about that presently (see, ABRAHAM 1999a, b, c; ABRAHAM & CONRADIE 2001). The remainder of this paper will be devoted to two main sections. First, in Section 3, a focused discussion of DRINKA's tenets with respect to her claim that Greek and French were the giving languages, and Germanic was among the receiving languages for have as a 'periphrastic AUX'. Second, in Section 4 are the main arguments in favor of my claim that, at least in German and Afrikaans, the emergence of have as a 'periphrastic AUX' was an autonomous process. In doing so, a view quite different from the Sprachbund thesis about SAE will be unfolded that in a way undermines its fashionable convergence component.

3. /favi-developing arguments revisited: why was be not retained? It is common opinion by today's research standards that the Aave-periphrasis in German(ic) was calqued from Late Latin, but that, in the first phase, it was used, in agreement with the grammatical possibilities of the receiving language, only with

248

WERNER ABRAHAM

perfective verbs (ABRAHAM 1990: case agreement obtaining with the object) and extended to the temporal function not until aspectual marking by morphological means was lost in Early New High German. Extending this view to areal convergence of the Aave-periphrasis in the European languages, we expect different incorporations across languages depending, essentially, on the temporal-aspectual paradigms of the receiving language. Specifically, if the language in question provides aspectual (perfective vs. imperfective) distinctions in morphological terms, the process of incorporation of the Aave-as-AUX caique might run in ways different from a language where the aspect distinctions are not represented paradigmatically. See Table 2 in the Appendix for what appears to be a multi-varied reaction to the emergence of the have-as-AUX across the European languages (gleaned from THIEROFF 2000). Notice the parallel between Standard Italian in the 1st stage (level 1) of development and the historically more advanced, regional North Italian in stage, on the one hand, and the situation in German (level 3) on the other: South Italian is like North German in that it is congruent with the Standard (written) code, while the Italian north coincides with the German South in deviating from the standard. The parallel is not complete, though, since in Italian, different than in German, reference to the past is split between an aorist and an imperfect. Note, however, that the aorist paradigm is missing in the North Italian regiolect, whereas the imperfect is retained. The common feature, as is easy to see, is the fact that both North Italian and South German are non-written, regiolectal varieties of the standards and, as such, open for online parsing processes (the single linguistic signal occurs only once and cannot be retrieved other than by short term memory). We shall argue that this oral/aural status of the code, and the subsequent parsing requirements, are the determining factor for the rise of any periphrastic (not only the periphrastic perfect) construction in SVOV languages such as the Germanic ones. This conclusion is not in line with the SAE ('European Sprachbund') convergence assumption. A picture as that in Table 2 radically invalidates any simple convergence explanation along several distinct criteria. First, with respect to the depth of evolution of the periphrasis, the convergence should have started from Slavic to Yiddish, Afrikaans, Pennsylvanian German, South German - but not Standard German etc., which is clearly implausible. Second, on the criterion of areal distribution, the result is equally implausible. Third, it is dubious that the single development can be co-classified. After all, the Slavic languages recycled their former past participles, after dropping the auxiliary copula, as a new simple preterit. This is a process that does not appear to be impending in any of the other, non-Slavic languages. Fourth, not all languages that developed stage 3 in Table 2 were consistent enough to drop stage 2 (see labels 1 and 2) and, fifth, the Slavic languages break the consistent picture in not providing a fiiture perfect, which is there in all other languages presented. No doubt this is due to the profound morphological aspect paradigm in Slavic. In short, the longitudinal and latitudinal distributions shatter profoundly the very allusion at an areal convergence, a Sprachbund scenario of the analytic perfect at the codt of the simple past. What else is there as a common explanans to the Europe-wide emergence of the composite past? And, above all, can we imagine development autonomous in each

The European demise of the simple past

249

single language and yet converging to an identical result? What makes any such assumption still more bizarre is the fact that such an autonomous trigger cuts in the middle the national languages: Standard German vs. South German; North Italian vs. Standard Italian; spoken ('CAMUS'-)French vs. Standard French etc. There is an interesting link that suggests itself between the choice of AUX that a language provides and the appearance of ergativity. Ergative languages are usually 6e-only languages (BENVENISTE 1952/1960, TRASK 1979: 399f., KAYNE 1993 and a number of others; Basque being one of the rare exception, see ABRAHAM 2001: 159-163). Since, as is brought to evidence by languages with split ergativity (ABRAHAM 2001), the ergative syntax correlates with the perfect or perfectivity and since a resultative component is indispensable for perfectivity, the status of be as the perfectivity AUX becomes more than plausible. Furthermore, languages that have disposal of be as an AUX are usually aspect, or at least aktionsart, languages (French-Italian, Slavic, German-Dutch-Yiddish; not, however, Scandinavian, English, Afrikaans). The crucial property is that the resultative component of the event structure can be predicated of only by means of be (or a perfective variant of it like German werden 'become'), not, however, by have. See (3). (3) The event structure of perfective resultatives/terminatives: (3a) Terminative (in)transitive verbs: cf. German einschlafen iV 'fall asleep' vs. einschläfern tV 'put to sleep' BECOME phase BE phase I >>>>>>>>> I 1 ti Ei tm E2 t„ (3b) Non-terminative (intransitive verbs: cf. German schlafen iV '(be a)sleep' vs. langweilen tV 'bore' GOING-ON phase GOING-ON phase I (I) 1 ti Ei tm E2 tn Key: t h tm, tn are points on the temporal axis representing the event. The event structure of terminative verbs is biphasic containing an approach phase as well as a resultative phase (Vtx:x,a.n) (ί,-0(Ει) Ρ (tm-tn) (E2)); U is a referential point belonging to both event phases simultaneously, Ei as well as E2. On the other hand, the event structure of non-terminative verbs is mono-phasic, (Vtx:xi(1_n) (ti-tn,)(Ei) = (tm-OiEz)))· (3a) describes terminative, or resultative, events. The time-axial structure is biphasic. (3b), on the other hand, describes non-terminatives, duratives as well as states (mono-phasicness as genus proprium). (4) and (5) show how the verbal arguments, iA and eA, are to be mapped on the event phases.5

5

Abbreviations: tV transitive verbs; eV "ergative" (=terminative-intransitive) verbs; iV intransitive (non-terminative) verbs; EvP event passive; StP Stative (adjectival) passive; PPP perfect participle passive; PPA perfect participle active; iA internal argument (direct object); eA external argument subject)].

250

WERNER ABRAHAM

(4a) Terminative tV: viz. umbringen 'kill' (two-place verb) BECOMING-DEAD phase BE-DEAD phase !>»>>>>>>»>! I Ε, φ E2 t, E, tm E2 t„ eA: β Α ^ η . , . η ε Α ^ . , ) iA: ί Α ^ Ο η . , . η ί Α ^ η . , . η ϊ Α ί Ο (4b) Terminative iV (=eV): viz. sterben 'die' (one-place verb) GOING-DEAD phase BE-DEAD phase I >>>>>>>>>>>>> I I Εί Φ E2 t, E, U E2 t„ iA:iA(t 1 n..^iA(tJv^iA(0 (5)

Non-terminative (i)tV: viz. schlafen 'sleep' (one-place verb) SLEEPING phase SLEEPING phase I (I) E, = E2 ti E, tm E2 t„ eA: eAOOrv.neACO (iA: Ϊ Α α θ η , . η ΐ Α ( υ )

It has been established beyond doubt that what is regarded to be "ergative" (or "unaccusative") predicates (both simple and affixed verbs) in German and Dutch are in fact perfective (resultative) intransitives. All distributional evidence for this claim rests on the AUX selection of be (ABRAHAM 1990, 2 0 0 1 : 1 3 7 - 1 4 2 ) . This correlates closely with the event structures described in some detail in ( 3 ) - ( 4 ) ; viz. the respective resultative fee-components.6 Notice that the perfective-resultative event component is an adjective, which extends internal argument (direct object) status for its subject. This is in line with the absolutive-ergative relation in the diathesis of ergative languages. This also explains straightforwardly that Agentive predicates cannot be ergative, since (resultative) adjectives, as event derived categories, can only be states (i.e., second components in ( 3 ) - { 4 ) ) . Have, it appears, prevails as AUX where aorist, passe simple, and passe composi are distinguished in the morphological tense-aspect paradigm - in other words, where aspect and tense are interlinked. On the other hand, where aspect more or less replaces tense as a paradigmatic form, be prevails to secure the perfective function. Why is that? Notice that have is less easily accommodated to express aspect than be. Be, for example, goes syntactically together with every participle, because it selects A, N, verb participles, and P, whereas have, as tV, is more constrained as regards the selection of categories. With respect to past participles, be is independent of aspect or aktionsart of the predicate it combines with as AUX, whereas have is not. In dialects in Northern Russian (VASILEV 1968: 220; quote taken over from DRINKA 2 0 0 3 : 5), just like in northeastern Macedonian (DRINKA 2 0 0 3 : 5), what has the meaning of a Aave-construction is in fact a possessive feeconstruction. See (6a,b) below. Standard Russian, on the other hand, uses a former 6

I have claimed in ABRAHAM (2001, and earlier) that in true aspectual languages such as Russian, the "modern" ergative, or unaccusative, discussion plays a very minor role due to the fact that the attentive linguist identifies the alleged "ergative" behavior as an integral component of the behavior of perfectives in the verbal paradigm. Only fte-languages can identify perfectivity in morphological terms. This typological property carries over to the alleged "syntactic ergativity" in the Indo-European languages.

251

The European demise of the simple past

past participle (vypil, uechaf) to render a modern simple past indeclinable for person and number. Notice that the past participles, vypil and uechal originally denoted the resultative state and implied (but did not denote directly) the event component leading to this result. (6) Northern Russian (6a) u nego vypito with him drunk-up 'he has drunk up' (6b) u nego uechano with him departed 'he has departed'

instead of Standard Russian originally: instead of Standard Russian originally:

on vypil he drunk up 'he (was) drunk up' on uechal he departed 'he (was) departed'

This relates to BENVENlSTE's claim that have is be + an empty adposition, i.e. something like be with or LOC/DAT + be (reminiscent of Latin; see (7)). (7)

Possession in Latin mihi est I-DAT

COP-3SG

SUBJECTj NOM(-SG)

Ί have OBJECTi' (8)

viz.

Aave(x,y) =.P(y,x)A0e(x)

Lexical decomposition of possessive have have be

Pj

[T] [AGR] [CASE] LOQ

[from GUERON

1995]

Latin mihi '(to) me' in (7) is taken up in (8) by both [Case] and LOC, daughters of P, no doubt a constituent highly marked with respect to case morphology and syntactic mobility. Now consider the observation that ergative languages do not appear to select the auxiliary have for periphrastic constructions. This was an early observation in typological research and has led linguists such as KLIMOV (reported by TRASK 1 9 7 9 : 3 9 9 ) to suggest that there may be a correlation between ergativity and oblique subject constructions (in the DAT, GEN, or LOC) to give form to a possession predicate (see the similar discussion in DIXON 1 9 9 4 : 1 9 1 ) . The latter generalization (if indeed correct; see, however, Basque, which is ergative and nevertheless possesses the auxiliary distinction) and explanation hinges crucially on the claim that have is be + an (possibly empty) adposition, i.e. something like be with or LOC/DAT + be as shown in ( 7 ) - ( 8 ) above. Note the case affix for the ergative in Hindi (MAHAJAN 2 0 0 1 : 6) in ( 9 ) below. Thus, if be is a category representing tense and agreement, then have incorporates the properties of be and configurates simultaneously both a grammatical case (as in the case of some ergative languages) and the semantics of some locality. See again ( 7 ) - ( 8 ) (following KAYNE 1 9 9 3 ) . Now consider the following Hindi example. The person-GOAL dative of Latin in (7) corresponds to what is the subject in the (equally marked) ergative case in Hindi, raam-ne. Notice that ergative case is a marked morphological feature, while the case of the object, kitaabe, is the pure nominal stem (unmarked; in many of the terminologies also called "nominative" or "direct case").

252 (9)

WERNER ABRAHAM

ΡαΓ"

hz

read-PERF-F-PL

be-PRES-PL

those-books-have-been-read (are-read)-to/for/by-Ram 'Ram has read those books'

[MAHAJAN

raam-ne

vah

kitaabe

Ram-ERG(M)

t h o s e books(F.PL)-0

1994: 6]

In V-medial languages (SVO, like all European Romance languages as well as the Germanic ones with the exception of German, Dutch, and Frisian), this adposition is V-incorporated and therefore does not surface as a separate adposition morpheme or as a case marker due to the direct vicinity of finite V in second position. In V-marginal languages, however, this adposition surfaces either as a separate morpheme or as case morphology, thereby making superfluous the separate have/be distinction. Consider that Hindi is SOV. MAHAJAN's idea is perhaps best illustrated by the change of agreement in French. As is well-known, agreement for gender and number on the participle is expressed with preverbal pronominal clitic objects, not, however, with postverbal full NPs. (10a) Jean lesj a repeintf-esj object - object agreement (10b) Jean a repeintfzOJ les maisonsi object-NP - no (object) agreement (10c) Jean lesj a cuitf-sj (lOd) Jean a cuitfzOJ les pommes de terrei

proclitic postverbal

Following MAHAJAN one may assume that the subject in French, like Jean in (10), originates as a PP, the head of which, P, is incorporated into AUX in the course of derivation from the underlying structure to the surface. This pertains to those positions of the finite predicate (AUX or full verb) immediately adjacent to the subject, Jean. Where a preverbal clitic object interrupts such adjacency, such incorporation is blocked and object-agreement applies, as in Hindi. The clitic constructions in French (10b,d), in a way then, can be regarded as ergative-like orders. Notice that this runs structurally parallel to DRINKA'S alleged "periphrastic have perfect [...] in the North Russian dialects" (DRINKA 2003: 5f.). We argued above that there is no reason to consider this as a true Aave-construction and, much less, as a /zave-auxiliary construction. Consider another example from DRINKA (2003: 6). (11) u volkov jideno with wolves.GEN eaten '(The) wolves have eaten (the) cow'

korovu cow.ACC

The accusative on korovu need not mark transitivity of a functional, but covert, have, as DRINKA seems to assume (along with PANZER 1984). It can be taken to be lexically licensed - see (12) for a deep structural description - as in the German examples in (13) below. (12)

[ ip LVP

(13) Es

[V·

eaf

cow.ACC][ PASTPART -eri][PP

wird jetzt anständig is now regularly 'Tea is to be drunk as always'

EXPL

at wolves][v (den) the.ACC

COPULA]]

Tee tea

getrunken drunk

253

The European demise of the simple past

Now what follows from this longish discussion of interlinked be, have, and ergativity or perfectivity? The following generalizations appear to be legitimate: (14) Licensing lexical case such as in (11) and (12) appears to be possible only under intransitive ie-predication, never under transitive Aave-predication. (15) Moreover, it seems that Aave-languages can be caseless languages - languages that license case through position in the SVO-frame; fe-languages, on the other hand, need to be languages with richer case morphology simply because the possessive construction, as in (6)-{7), cannot be expressed otherwise. (16) Languages with be as well as have will distinguish, in one form or another, aspectual or aktionsart perfectivity. (17) 5e-only languages will provide a simple preterit (as is shown by Russian) to at least express imperfectivity. Let us substantiate these assumptions. One of the crucial arguments developed in the line of bringing the spread of the periphrastic AUX-use of have to evidence is that the Ancient Greek be +perfect participle possession construction served as a model for the emerging Aave-construction. DRINKA (2003: 8; her example (18)) refers to the following illustration. (18) Luke XII,2 ouden de sugkekalumminon estin ho oiik apokaluphthesetai nothing however covered up is which not should be covered up 'nothing is concealed which should not be concealed' However, her conclusion is hard to follow because there is simply no auxiliary have involved, nor is there in (19). (19) Plutarch Pelopidas 35,4 toils men adelphous the

aph' hemeras eikheplesion

en

however brothers.ACC.PL from morning kept neighborhood in

oiköi

tini

kekrummenous

house.DAT

certain.DAT

hidden.ACC.PL

'From morning on, she kept her brothers hidden in a house in the neighborhood' The crucial criterion is in either of the last examples that the verbs involved are perfectives, "conceal" in (18) as well as "hide" in (19). The past participles denoting the resultative components, E2, as in (3a, b) and (4a) imply Ej, the prior event component in both cases, i.e., with estin 'is' as well as the transitive verb eikhe 'kept, held' the second construction, (19), being an object predication presupposing a perfective predicate in the first place. This is light-years away from an auxiliary construction comparable to the modern periphrastic temporal construction.7 That 7

In order for DRINKA'S example (DRINKA 2 0 0 3 : 9 ) , from Sophocles' Oedipus Tyrrannus 7 3 1 , oude ρδ lexant'ekhei but-not yet-ceased holds '(the story) goes on not yet ceased.AOR.ACT (being told)', to be a convincing subject predication with an active aorist we would have to see agreement by person and number and gender on the participle, lexant'. But there is no agreement proof. This

254

WERNER ABRAHAM

"have was already being used as a full-fledged auxiliary in Greek by the time the Romans came in contact with them" (DRINKA 2003: 9) is not substantiated. The same holds for Classical Latin, which supposedly took Greek as its model for the development of the have-AUX construction. Notice that there are usually only esse+possessive dative constructions - viz. (20) - but no constructions with habere. The latter did not turn up until the late phase of Latin. However, usually habere then occurs only with object predication. See (21)-(22) below. (20) ex libris Punicis [...] interpretatum from books Punic translated 'For us has been translated from Punic books'

nobis est for us is

Despite interpretari being a deponential lexical (passive form, active meaning), the translation appears to work with the meaning of the passive participle. The verb is perfective as in the following examples with the lexical habere. (21) (Flamines)[...] caput cinctum priests head(ACC) girdled ACC 'The priests had their heads girdled with a fillet'

habebant had

filo fillet. ABL

(22) dicam de istis Graecis suo loco, Marce fili, quid Athenis exquisitum habeam Ί will report about those Greeks in its proper place, my son Marcus, what I might have found out in Athens' (23)

episcopum [...]

invitatum

habes [GREGORY OF TOURS, 6th century A.D.]

bishop.ACC

invited.ACC

have.2SG

Both caput cinctum and quid exquisitum agree on case, number, and gender. Thus, the constructions are object predications. In such cases, the main verb habere (in our case, more precisely: habebant, habeam) is to be taken as a full verb, not as an AUX. They are not temporal perfects. Their relation to the later periphrastic perfect is not yet anticipated on the ground that habere does not agree, if in construction with a past or passive participle, with the subject. (23) does not mean 'you have invited the bishop', but 'you got a bishop as an invitee', just like Greek ton episkopon kalesas ekheis does. Generally speaking there is no reason to assume that either Greek or Latin had developed the auxiliary function of have already. The Germanic branch, with Gothic at its beginning pole, did not in any case (see, for Gothic and Old High German, ABRAHAM 1992). The best proof for this conclusion is the fact that Old High German used eigan next to haben for the object predication construction: eigan is a cognate of Modern German verb eignen (cf. adjectival eigen 'own'), which cannot enter into an auxiliary construction on account of its full lexical meaning. Let us not weigh too heavily the fact that the etymological source of OHG have cannot be Latin habere. That it has to be capio 'to rent, hire, seize, arrest; take, choose, attack, injure, comprehend' is no doubt relevant for someone who assumes means that ekhei is, to all appearances, nothing but a full lexical verb the literal translation being '(the story) holds out unceasingly'.

The European demise of the simple past

255

calquing ("drawing analogues from functions to similar forms") in the first place. When, and under which conditions, would the Aave-calque be expected to work? In other words, when can one expect auxiliation of German haben to complete in the history of the language? This is what appears to be required. (24) (24a) (24b) (24c) (24d) (24e) (24f)

Conversion of have as a lexically full verb to an auxiliary in the perfect periphrasis takes place under the following conditions: extended complement selection to include imperfective verbs; functional distinction of the simple past and the periphrastic past; and/or the separate orientation of the periphrasis under conditions of perception (oral code); developing and consolidating the sequence of tenses (which is complementary to aspectual systems); have loses its semantic roles, while preserving transitivity (accusative government); possession no longer expressed by have in the selection of past participles; i.e., selection, under this function, is no longer restricted to {N} as a complement; conversion of the semantic role selection: have changes from a full verb with a complete semantic role grid to a raising predicate (i.e., it takes over the external semantic role of the embedded predicate).

In conclusion one may say that, despite the somewhat thin layer of evidence for the rise of the Aave-periphrasis in Greek and Latin through the ancient period and the Middle Ages, it cannot be excluded that Greek laid the foundation to the emergence of the Aave-periphrasis in Late Latin and subsequently in the Romance languages. But this is not really the point. What was raised as a real issue is that, first, there must be good reasons why have replaced be in auxiliary, non-possessive constructions; second, why, in quite many languages, be remained, side by side with the new have, with auxiliary status in a clearly defined class of mutative predicates (not only verbs, but also VPs); and, third, why a large body of languages could do without the AUX have. We have shown that the emergence of have was dependent upon a number of intralinguistic factors. This in itself makes improbable the assumption of areal convergence across language borders, because borrowing by calquing is at least favored, and made possible only by, identical paradigmatic and functional systems, just as much as it is impeded by non-identical ones. This was a factor that was not considered between Ancient Greek and Latin, in the first place: Greek was a clear aspect language, Latin, at least during the Classical period, was not (no aorist); and Greek had a diathetic paradigmatic setup that was more complex than that of Latin (no middle diathesis). Similar differences hold between the Romance languages and the Germanic as well as the Slavic ones. While the process of the emergence of the periphrastic perfect connoting the past may very well have started, and gone furthest, in Parisian French, or the French of the Midi, as early as in the 12th and 13th centuries and spread from there in the Romance languages, there remains the question why it did not spread in all of Italian and in all of the remaining Romance languages alike. As for the Germanic languages, however, the spread thesis is even tougher to maintain. First, if English got the periphrasis from Normannic French, why did English develop a periphrastic function so very different from the Romance periphrasis? Second, why did North/

256

WERNER ABRAHAM

Low German, along with the Standard, not caique the French periphrastic model, while the South apparently did; and why North Italian (only a regiolectal area were dialectal speakers more open to calqued borrowing from Standard French than Standard Italian?); third, how come Yiddish went along with Upper (=South) German (Austrian, Bavarian, Alemannic); and why did Afrikaans develop the periphrasis so radically, i.e. without exception - even more so than Upper German and Yiddish - while its source language before the diaspora, Dutch, i d not? How come Afrikaans was under the French influence more so than Dutch (where, beyond doubt, French had so much cultural and political-administrative influence in the Napoleonic period, but nevertheless has retained its of-old have/be split according to semantic and syntactic factors)? All these questions do not weigh that much for one who does not believe in the SAE convergence process in the first place. In what follows I specify only the main tenets of the complementary position of an autonomous, but nevertheless codebased emergence of the Aave-periphrasis in German and its dialects as well as in Yiddish, and in Afrikaans (see, for a book-length discussion, ABRAHAM & CONRADIE 2001; see already ABRAHAM 1999a,b,c). Convergence phenomena may well be the rule rather than the exception in grammaticalizing processes in that they can create novel gram families (DAHL 1996: 363). But grammatical borrowing (calquing) does not emerge just by throwing a functional model across the border, with the syntactic system on the receiving side being indiscriminately receptive like a waste bag. Notice that in making a disclaimer to the convergence stance at least for German, we have not addressed the lacking evidence of a longitudinal documentation of the spread (cf. STOLZ 20028). This is the most urgent question that the convergence position will have to answer. No doubt, this is a sizable task yet to be begun.

4. The autonomous roots of the demise of the simple preterit in German, Afrikaans, and Yiddish Clearly, what is at the core of the present discussion is not so much the question why and how the Aave-periphrasis spread, but, rather, how come it was the only form remaining to function as a preterit. In other words, we want to focus on those languages or their varieties where the periphrasis is the only form to denote the past. Now, we saw (Table 2) that simple pasts have disappeared in widely adjoining parts of Europe, in particular the southern and eastern parts (South Germanic, Romance, Slavic), and periphrastic preterits (present anteriore) have taken over their function. Yet, the consequences in their respective languages are different. 8

This constitutes a defining criterion which is to be added to STOLZ'S list, which comprises: genealogical diversity; extension beyond everyday lexical material; multilevel phenomena; (no) symmetrical exchange; logical connection of the features shared; idiosyncratic status for Sprachbund area excluding other languages; homogeneity and discreteness; solid extralinguistic and sociohistorical explanation for the emergence of similarities (STOLZ 2002: 263f.)

The European

demise of the simple

past

257

One can distinguish 4 types of preterit-loss languages and dialectal areas within these. The classifying criteria are: ((level 0) present anterior different from simple past: Scandinavian, English, and others) (level 1) present anterior has the same function as the simple preterit: Standard German and Italian; (level 2) the simple preterit has become very marginal: Serbo-Croatian, Rumanian, French; (level 3) no simple preterit exists any longer; its function is taken over fully by the present anterior: South German, Yiddish, Slavic. The last group of languages can be subdivided according to whether or not they have a paradigmatic representation of the periphrastic future perfect and/or a modal usage linked to the present anterior: South German vs. (level 4) Slavic. All of this is suggestive of areal spreading, the source being those languages which have advanced the farthest on this cline of preterit demise and which acquired new properties of the periphrastic perfect (i.e. Russian). We have argued that this assumption is misguided for several reasons. What is foremost, there is simply no diachronic record of such an overall influence on the temporal and/or aspectual paradigms and functions (except, possibly, for the noteworthy grammatical loan of Slavic byt' for German werden as a future and passive auxiliary; exotic, since it is not shared by any other Germanic language). Second, and what is even more, in the languages in the diaspora from their Germanic ancestry, Afrikaans, partly also Eastern Yiddish, identical phenomena have evolved without areal contact with their original linguistic environments. Dutch has not dropped the simple preterit. What is common to the three, non-neighboring areas with the loss of the simple preterit and a complementary rise of Aave+anterior is this: (25) The essential preterit loss in the European neighboring and non-neighboring, enclave, languages is restricted to spoken codes. (26) In spoken codes, a fundamentally different grammatical mechanism is employed: i.e., parsing; a hearer's on-line computation is different from a speaker's/writer's/ reader's (or a speaker-hearer's) computation which does not need to be on-line. (27) Languages are to be grouped on the basis of "(subject-predicate) agreementobserving" and "discourse-prominent" strategies. German and Dutch are discourse-oriented languages. The grammatical device for this functional classification is the wide middle field (SVOOV), which allows for ample scrambling in terms of thematic (given) and rhematic (new) nominal reference neither English nor the Romance languages provide a similar scrambling device and, consequently, are not discourse-prominent in the same syntax and word-order based fashion as German. Scandinavian does so only in a very restricted sense (with Swedish even less so than the rest; cf. A B R A H A M 1 9 9 0 , 1995).

These three factors have been discussed, and exemplified in detail, for (South) German and Afrikaans (see A B R A H A M & C O N R A D I E 2001). The main conclusion was that South German (partly also Yiddish and Afrikaans, but not Standard (and North-)German nor Dutch) uses structural devices emptying V-2 from lexical material inserting AUX in V2 (first verbal bracket) and putting it preferably in V-last

258

WERNER ABRAHAM

(second verbal bracket). This ensures that V-2 realizes the parsing of subject agreement and, at the same time, retains V-last (and the ensuing wide scrambling possibilities) in terms of the principally (spoken) discourse-prominent code. (28) Parsing filter with respect to the spoken/aural coding of German: *[CP [c' Vfinjte ··· [vr t]]] at the cost of [CP [c' AuXfinite ... [vp V non-finite]]] In terms of the clause field theory (28) boils down to this: preferably, the first verb bracket contains no full finite verbs, but only auxiliaries (among them tun and, naturally, the modal verbs). In the tradition of German philology one particular phonological development has been invoked as a trigger for the loss of the simple preterit. This is what makes this pervasive reasoning (apocope of schwa-e) weak as a sufficient cause: 3RD SG is homophonous in the past tense (cf. for MHG ABRAHAM & CONRADIE 2001: 4); no doubt, the context solves time reference to a great

(29a) Only the

deal; adverbs help - in general, there is much tense opacity in German (no future tense necessaiy; adverbs do the job; historical present tense! Only in very few cases - of a basic aspectual implication - is there a functional difference between the simple (aorist) past and the analytic perfect tense: 2001: 17f.). (29b) The analytic preterit spread in many other languages in Europe (see ABRAHAM & CONRADIE 2001: 9; 11). This cannot be due to -e-apocope alone! (See Table 3 in the Appendix.) Recall our argument against the assumption that areal influence (basically a Sprachbund phenomenon) took place with respect to the loss of the simple preterit in South German? The answer was: no! If that were indeed the case we would have better clues as to what the source area would be. But we have none! Yet, there is a diachronic cline in the areally distributed emergence of the analytic preterit - some develop the analytic perfect next to the synthetic preterit, others replace it, and again others function in balance with aspectual systems that German has lost ever since MHG (such as the Romance and the Slavic languages!). However, there are two new clues: first, there is preterit loss also in Afrikaans and Yiddish, which are areally independent from the European range of phenomena; second, there is the general South German avoidance of Comp/V2 (first verbal bracket) for lexically full verbs, i.e., enforcing auxiliaries instead (including tun as AUX in V2; see Table 1). This drives the full verb to Vlast in declaratives; ( A B R A H A M & C O N R A D I E 2001: chapter 5.5, 92ff.). This could be taken as the explanation: the lexical verb (as opposed to Unitizing AUX) carrying the main rhematic information is placed at the end of the sentence (cf. (28) above). This ensures that V+its objects are always in rhematic position, where they belong in the first place. This pictures German as a discourse prominent language, much in distinction from English with its strict S(X)VO and without scrambling and refocusing (using clefts instead). Bear in mind that refocusing can be perceived in the acoustic code only, not in the written.

The European demise of the simple past

259

The conclusion from these observations is thus: all three areas, South German, Afrikaans and Yiddish are characterized by oral/aural L-representation which is fundamentally distinct from the written code. What can we expect to be the main differences between an oral (acoustically perceived/heard) vs. a written (visual/read) code? The oral (acoustically perceived/ heard) code processing step-by-step, one lexical each from 1st to last (according to HEIM's staple and proceed model! See ABRAHAM & CONRADIE 2001, ch. 3.3. ). The written/read code, on the other hand, due to its substantive recycleability is outside of any such parsing exposure (after all, one can always reread to ensure full semantic identification). Clearly, the written code appears to have advantages over the acoustic code. But this is only partly so, as will turn out. Now, VI and V2 optimally support the acoustic/linear perception/recognition because of two factors: (30a) agreement for subject identification is ascertained early on (as in SVO language generally); (30b) valency identification is ascertained early on, too. There is one major problem, though, for the oral-code: South German does not insert lexical V in V2/C0 - not in main clauses with AUX in sentence-second position (Comp/first verbal bracket) and lexical Vlast (V° in VP), and not in dependent clauses either. How can this be aligned in a reasonable way with the claim that oral/acoustic syntax recognition is in line with the loss of the simple past and emergence of the new perfect tense and its motivation in terms of oral/acoustic coding? Clearly, Vlast in the perfect tense predication in German (both in main and in dependent clauses) does not appear to be the logical solution for aural decoding, in the first place. Are the two, then, not in outright contradiction? The oral/aural/acoustic code is not restricted to early-agreement computation. It pursues two strategies: the discourse identification strategy (the question being: what is the place of the clausal units in larger context?), which is paramount for oral/aural communication, and an agreement securing strategy (restricted to each individual clause) equally important for oral/aural communication. The claim is that either requires a different linearity strategy: the agreement securing strategy needs to identify the subject ASAP and thereby keep it distinct from objects. This is optimally achieved by VI or V2 (thus, English generally and main clauses in German). Notice that oral South German has VI-declaratives, not acceptable in Standard German at all (ÖNNERFORS 1997, SIMON 2001). But notice what English and German do differently in independent clauses: (31a) English co-establishes agreement (by way of AUX: do, have, be, finite V) and valency (by immediately inserting the lexical V: He does not give a damn!) (31b) German, however, always splits AUX and V, sometimes across long linear distance (i.e. from V2 to Vlast): Er tut/kann darüber nicht lachen 'He does/cannot laugh about that'. The discourse-unit identifying strategy targets the identification not on the level of the clause (clausal parts), but on the level of larger text portions - with thema and rhema being the only interesting goals for identification and distinction. (31a)

260

WERNER ABRAHAM

and (31b) are categorically different tasks of recognition. Clearly, however, discourse prominence cannot neglect the clause identificatory task. The German clause bracketing (wide 'Middle field' split between V2 and Vlast), appears to provide a concrete disadvantage with respect to the linear steps of identification: in the main clause because valency cannot be determined (which suspends object identification) until Vlast at the end of the linear chain; and in the subordinate clause even more so, since both subject agreement and object identification by means of valency is suspended until the last word in the clause! (Imagine a main clause with subject and object in homophonous case representation, which disallows the identification of subject vs. object.) However, South German dialects, as extreme representatives of oral/aural codes, have two ways that help balance the aporia described above: (32a) They (partially) mark agreement on the first word in the dependent clause, Comp ('conjunction inflection': see ABRAHAM & CONRADIE 2001: 64; 107). Cf. (vi/vii) in Table 4 as well as the following illustration: Ob-s (es) wirkli kummt-s? whether .2PL you. 2PL really come.2PL 'Whether or not you really will come?' (32b) They have VI-declaratives, something that Standard German does not possess. Cf. (ix) in Table 4: Derfn-s nachher zahln heit may-you.CL later pay today 'You may pay later today' (32c) They allow for scrambling, which triggers refocusing (i.e. accents distinguished with respect to base (=presuppositionless) vs. contrastive (presupposing in different grammatical degrees) sentential accent. Cf. the distinction between grammatical accent (GA) and contrastive accent (CA) in Table 1. (32d) Full lexical verbs are avoided in CVl5* verbal bracket; instead, auxiliaries (including tun 'do') are inserted carrying inflectioa Cf. (v) in Table 4 (see Appendix): SiePL tutsa 's//es ja gar nicht gerne in Schwierigkeiten bringen them does it not with pleasure in trouble bring so ohne Anlass. without reason 'It is them that she does not want to get into trouble with without good reason.' Notice that, under such a fundamental discourse prominence (clear distinction of regrouping of the basic linear order of the clausal parts), one expects much of the overt case requirements to be superfluous. This bears out in the South German dialects: the discourse-determining order of elements, distinct in terms of thema and rhema, suspends much of the case identification requirement (since valency with theta role identification is present anyway). Case redundancy is economized upon. There remains the question why North German (and Dutch) does not follow the demise of the simple preterit to the same extent and exclusiveness as the South. NLEBAUM & Μ ACH A (1999) provide a perfunctory, yet telling reason: North German dialects have been leveled almost to non-existence by industrialization and the

261

The European demise of the simple past

subsequent urbanization and migration. Low German (Piatt) has lost its grip on the population ever since the bloom of the Hanse League (early 16th century). Furthermore, its case morphology, just like that in Dutch, has been levelled to near nothing. This results in a rather inflexible word order disallowing scrambling for discourse-functional purposes as in High and Southern German. Where dialect characteristics in Northern German are found nowadays they were reintroduced and differ from Modern Standard German only phonetically, not syntactically. It is important to keep in mind the structural prerequisites for the claim that spoken German is so different from written German. The dividing line between oral coding/perception/recognition and written/reading coding is nowhere as drastic as in an SVOV-language. In other words, in such a language (exotic, as it is, with SVOV in the independent clause, but strictly SOV in the dependent clause) the tendency toward the analytic preterit, with its psychological advantages, is a direct reflex of its discourse prominence echoed by the wide reshuffling possibilities of scrambling and refocusing. The "exercise" of discussing Afrikaans as another, but non-identical representative of the preterit demise was necessary in order to show to what extent analytic perfect in oral coding responds to intra-linguistic, non-psychological factors. For Afrikaans, the tendency toward simplification of the whole tense system is a logical consequence of the general and radical loss of morphological distinctions of forms (-e as well as -^-apocope; total loss of the irregular past participles). For Afrikaans and its new and sudden need for linguistic and social integration, simplification for L-planning played a much more profound role, quite different from South German. In (South) German, a similar process at the level of linguistic contact to yield a lingua franca never took place. Thus, for German, which generally retains the irregular past participles, the grammaticalized discourse strategy worked out on different motivation and much more radically than for morphologyimpoverished Afrikaans. The example of Yiddish is a welcome confirmation of the factorial argument claimed to hold for spoken German: it had only oral coding (the written and canonized code was Hebrew); it had no exposure to normalizing literature, to schoolmasters normalizing language, and administrative linguistic prescriptions. It had to cope predominantly with oral/aural discourse situations. No alignment took place with respect to the two aspectually distinct AUX-classes. Yiddish has indeed preserved the limited class of intransitive perfectives (in the generative jargon: "ergative" or "unaccusative") zayn-verbs. See (33): (33a) Intransitive perfectives: er iz gegangen, iz geforn, izgeblibn, he is gone is ridden is stayed (33b) Intransitive imperfectives: er hot geshwummen hot gelegn he has swum has lain

izgekumen, is come

izgesturbm is died

Yiddish, much like Modern Dutch, even appears to overwrite the the transitivity condition by the perfective criterion. See (34):

262 (34) erjz bagangen he is committed 'He committed suicide'

WERNER ABRAHAM

aleyn-mord alone-murder

As observed before, in two major geographic areas for Yiddish the have-AUX has tended towards generalization as the sole past-tense AUX. These areas are Northeastern Yiddish. (35) er hot gekumen 'He has come' Here the Aave-emergence is possibly motivated by the massive linguistic contact with Russian SVO, a situation similar to Afrikaans. The second group are language attriting speakers of Yiddish in America (Levine 2 0 0 0 ) . Quite probably, in the latter case, the influence of surrounding SVO-English has lead to this convergence in American Yiddish. That the analytic past is the result of a development no earlier than in Early New High German is proven beyond doubt by the fact that in Middle High German the analytic past would have been restricted to perfective verbs. It could never have played the pervasive tensing function that it has in modern South German.

5. Wrap-up On discussing heavy attrition and the subsequent functional losses and syntactic simplification in the development of English, McWHORTER (2002) concludes that most had already been present when the Scandinavian tribes settled in North England and Scotland. But he also raises the question why similar processes like those between Old English and Modern Standard English did not occur elsewhere in Germanic. What he invokes are substitution processes of exaptation. Let us assume that there is a tendency for a grammar to fill in "open spaces" in syntax as well as in phonetic inventories as V L S S E R surmises (VlSSER 1963: 135; M c W H O R T E R 2002: 224). In this spirit we might ask why English did not submit its pronouns to "exaptational" usage in L A S S ' (1990) conception borrowed from evolutionary biology. In other words, why is it that English did not recruit reflexive pronouns as markers for derived valency, voice, and mood just as dozens of other languages were concurrently doing across Europe (McWHORTER 2002)? More generally we may ask which of the phenomena, possibly relevant within some notion of SAE, might the notion of exaptation be applicable to? Notice that "exaptation" presupposes a scenario of evolutionary formal (morpho-syntactic) emergence. I focus on three key phenomena in McWHORTER's discussion of the particular exbraciation of English from the rest of West Germanic in the course of its history (McWHORTER 2002: 224). (36a) Applied to the wide or narrow usage of reflexivity (from lexical reflexivity (sich schämen 'to be ashamed') to diathetic-syntactic reflexivity (s-passives in Slavic and Scandinavian; middle construction in German: Das verkauft

The European demise of the simple past

263

sich leicht 'this is selling rapidly'), exaptation does not appear appropriate if syntactic processes are incorporated below the word level as much as outside, as in syntax proper. (36b) As little as under 1 above, the notion appears applicable to the threefold gender distinction morphologically expressed on the definite article as well as the attributive adjective (strong, weak, and mixed inflection distinctions) in German: this particular exaptation, if applicable, served a higher purpose, namely that of discourse structuring. (36c) Indefinite pronominal man, extant in German, but no longer in English ('one'), not only reduces dependency on the context, but, what is more, makes the online-interpretation of German proportionally easier in relation to English and other non-Germanic languages, which allow for homonymies unknown in German. There is no reason to assume that all similarities among the European languages are rooted in contact induced convergences. Any such assumption lays a heavy methodological load on the shoulders of the assumer. I have argued that the rise of the periphrastic perfect is a case in point. What appears to be such a contact induced spread, at superficial sight, is in fact an autonomous, and by no means exceptionsless, spread of a phenomenon that facilitates oral and aural coding and is, in its emergence, restricted to this specific type of coding. The most convincing arguments come from Afrikaans, Yiddish, and South German, all of which are in fact solely spoken variants either in their original emergence or in their present state. Since online parsing as well as discourse processing strategies play a determining role in such oral codings, these three languages and language varieties are particularly characteristic of a development under such factors. Online recognition of the clausal parts supports V-early or V2 to the extent that the finite predicate of the clause early on identifies the distinction of subject as opposed to objects as well as the distribution of objects by way of lexical valency. Yiddish, despite its historical OV-base, is a case in point since it has become SVO. South German has gone beyond a mere V2 in that the partial conjunction inflection fulfils the function of early on agreement identification, on the one hand, and in that Vlast opens the wide middle field that allows for extensive manipulation of NP-material for discoursefunctional purposes. For either function, i.e. for agreement as well as for discoursescrambling in the middle field, the periphrastic perfect, as opposed to the simple preterit, serves a clear structural (=logical-grammatical) as well as psychological (sentence-processing) goal. At least on the level of this detail, this represents the deeper and concrete reason for what has come to be considered part of a pervasive convergence phenomenon among (some) European languages, the SAE. I argued that this general stance deserves closer inspection guided by sound methodological criteria and pursued on diachronic as well as synchronic levels as for the emergence of the periphrastic perfects.

264

WERNER ABRAHAM

Abbreviations ablative accusative actional agreement aorist clitic(pronoun) copula dative ergative expletive feminine

ABL ACC ACT AGR AOR CL COP DAT ERG EXPL F

GEN LOC

Μ NOM PART PAST PERF PL PRES SG

genitive locative masculine nominative partitive past tense perfective plural present singular

References (1988): Vorbemerkungen zur Modalpartikelsyntax im Deutschen, in: Linguistische Berichte 118,443-466.

ABRAHAM, WERNER

ABRAHAM, WERNER ( 1 9 9 0 ) : A n o t e o n t h e a s p e c t - s y n t a x i n t e r f a c e , in: MASCARÖ, Μ . & NESPOR, MARTINA (eds.), Progress in linguistics. Dordrecht: Reidel, 1-12.

(1992): The emergence of the periphrastic passive in Gothic, in: Leuvense Bijdragen 81/1-3,1-15. ABRAHAM, WERNER ( 1 9 9 5 ) : Deutsche Syntax im Sprachenvergleich. Grundlegung einer typologischen Syntax des Deutschen. Tübingen: Narr. ABRAHAM, WERNER (1999a): Preterite decay as a European areal phenomenon, in: Folia Linguistica Europaea XXXIII/1,11-18. ABRAHAM, WERNER (1999b): SVOV im Deutschen: Schnittstelle zwischen Diskursprominenz und ABRAHAM, WERNER

S u b j e k t i d e n t i f k a t i o n als P a r s i n g b e d i n g u n g , in: KATNY, ANDRZEJ & SCHATTE, CHRISTOPH (eds.),

Das Deutsche von innen und von außen. Ulrich Engel zum 70. Geburtstag. Poznaü: Wydawnictwo Naukowe, 3-7. ABRAHAM, WERNER (1999c): Präteritumschwund in German: the parsing trigger, in: Folia Linguistica Europaea XXXIII/1 [special thematic volume: Characteristic properties of spoken vernaculars], 39-58. ABRAHAM, WERNER (2001): The aspect-case typology correlation, in: REULAND, ERIC J. (ed.), Arguments and case. Explaining Burzio's generalization. Amsterdam-Philadelphia: Benjamins, 131-194. ABRAHAM, WERNER (2002): The European demise of the simple past and the emergence of the analytic perfect: areal spread or autonomous evolution? Plenary paper given at the Annual SLE Meeting July 2002 at Potsdam : "Kognitive und soziale Aspekte des Sprachwandels". ABRAHAM, WERNER & CONRADIE, JAC (2001): Präteritumschwund und Diskursgrammatik. Präteritumschwund in gesamteuropäischen Bezügen: areale Ausbreitung, heterogene Entstehung Parsing sowie diskursgrammatische Grundlagen und Zusammenhänge. Amsterdam-Philadelphia: Benjamins. BENVENISTE, EMILE ( 1 9 5 2 / 1 9 6 0 ) : Etre et avoir dans leurs fonctions linguistiques, in: Bulletin de la Sociite de Linguistique de Paris 55,113-134. COMRIE, BERNARD ( 1 9 7 6 ) : Review of: KLIMOV, GEORGU A . ( 1 9 7 3 ) : Ocerk obSöej teorii ergativnosti. Moscow: Nauka, in: Lingua 39,252-260. DAHL, OSTEN (1996): D a s T e m p u s s y s t e m d e s D e u t s c h e n i m t y p o l o g i s c h e n V e r g l e i c h , in: LANG, EWALD & ZIFONUN, GISELA (eds.), Deutsch - typologisch. Berlin, New York: de Gruyter, 3 5 9 - 3 6 8 . DAHL, OSTEN (ed.) (2000): Tense and aspect in the languages of Europe. Berlin, New York: de Gruyter. DAHL, OSTEN ( 2 0 0 1 ) : P r i n c i p l e s o f a r e a l t y p o l o g y , in: HASPELMATH, MARTIN; KÖNIG, EKKEHARD; ÖSTERREICHER, WULF & RAIBLE, WOLFGANG (eds.), Language typology and language univer-

salis. Berlin: de Gruyter, 1456-1470. DAHL, OSTEN (2003): The growth and maintenance of linguistic complexity. Ms. Univ. of Stockholm. DIXON, ROBERT M.W. (1994): Ergativity. Cambridge: CUP. DRINKA, BRIDGET (2003): Areal factors in the development of the European periphrastic perfect, in: Word 54.1 (page quotes from her ms.). GUIRON, JACQUELINE ( 1 9 9 3 ) : Sur la syntaxe du temps, in: Langue frangaise 100, 1 0 2 - 1 2 4 .

The European demise of the simple past

265

GUERON, JACQUELINE (1995): On have and be. Paper Univ. Paris, Nanterre (read at Groningen April 26,1995; handout). HASPELMATH , MARTIN (1999): How young is Standard Average European?, in: Language Sciences 2 0 , 2 7 1 - 2 8 7 . HASPELMATH, MARTIN ( 2 0 0 1 ) : The European linguistic area: Standard Average European, in: HASPELMATH, MARTIN; KÖNIG, EKKEHARD; OESTERREICHER, WULF & RAIBLE, WOLFGANG (eds.) Language typology and language universals. (Handbücher zur Sprach- und Kommunikationswissenschaft). Berlin: de Gruyter, 1492-1510. HEINE, BERND (1994): Areal influence on grammaticalization, in: PUTZ, MARTIN (ed.),

Language

contact and language conflict. Amsterdam-Philadelphia: Benjamins, 55-68. ISACENKO, ALEKSANDR (1974): On have- and fee-languages. (A typological sketch.), in: FLIER, MI-

CHAEL S. (ed.), Slavic Forum. The Hague-Paris: Mouton, 43-77. KAYNE, RICHARD S. (1993): Towards a modular theory of auxiliary selection, in: Studia Unguistica 47,3-31. KLIMOV, GEORGIJ A. (1973): Ocerk obscej teorii ergativnosti. Moscow: Nauka. KUTEVA, TANJA (2001): Auxiliation. An enquiry into the nature of grammaticalization. Oxford: OUP. LASS, ROGER (1990): How to do things with junk: exaptation in language evolution, in: Journal of Linguistics

26,79-102.

LEVINE, GLENN S. (2000): Incomplete LI acquisition in the immigrant situation: Yiddish in the United States. Tübingen: Niemeyer. MAHAJAN, ANOOP (1994): The ergativity parameter: have-be alternation, word order, and split ergativity, in: NELS 23, 317-331. MAHAJAN, ANOOP (2001): Oblique subjects and Burzio's generalization, in: REULAND, ERIC J. (ed.), Argu-

ments and case. Explaining Burzio's generalization. Amsterdam-Philadelphia: Benjamins, 79-102. MCWHORTER, JOHN H. (2002): What happened to English?, in: Diachronica XIX, 2,217-272. NIEBAUM, HERMANN & MACHA, JÜRGEN ( 1 9 9 9 ) : Einführung

in die Dialektologie

des

Deutschen.

(Germanistische Arbeitshefte 37). Tübingen: Niemeyer. ÖNNERFORS, OLAF (1997): Verb-erst-Deklarativsätze. Grammatik und Pragmatik (Lunder germanistische Forschungen 60). Stockholm: Almqvist & Wikseil International. PANZER, BALDUR (1984): Parallel internal development or interference? Post-positive article and possessive perfect in North Germanic and northern Russian, in: URELAND, P. STURE & CLARKSON, IAIN (eds.), Scandinavian language contacts. Cambridge: CUP, 111-132. RAMAT, PAOLO & BERNINI, GIULIANO (1990): Area influence versus typological drift in Western E u r o p e : t h e c a s e o f n e g a t i o n , in: BECHERT, JOHANNES; BERNINI, GIULIANO & BURIDANT, CLAUDE

(eds.), Toward a typology of European languages. Berlin, New York: W. de Gruyter, 25-46. RAMAT, PAOLO & STOLZ, THOMAS (eds.) (2002): Mediterranean

languages.

Papers from

the

MED-

TYP workshop; Tirrenia, June 2000. Bochum: Ν. Brockmeyer. SHANNON, TOM (1990): The Unaccusative Hypothesis and the history of the perfect auxiliary in Germanic and Romance, in: ANDERSEN, HENNING & KOERNER, KONRAD (eds.), Papers from

the

International Conference on Historical Linguistics. Amsterdam-Philadelphia: Benjamins, 461-487. SIMON, HORST J. (2001): KinnanS Eahna fei heid gfrein. - Über einen Typ von Verb-ErstAussagesätzen im Bairischen, in: DONHAUSER, KARIN & EICHINGER, LUDWIG M . (eds.),

Deutsche

Grammatik - Thema in Variationen. Festschrift für Hans-Werner Eroms zum 60. Geburtstag. (Germanistische Bibliothek 1). Heidelberg: Winter, 137-153. STOLZ, THOMAS (2002): No Sprachbund beyond this line! On the age-old discussion of how to define a linguistic area, in: RAMAT, PAOLO & STOLZ, THOMAS (eds.), Mediterranean languages. Papers from the MEDTYP workshop; Tirrenia, June 2000. Bochum: Ν. Brockmeyer, 259-282. THIEROFF, ROLF (ed.) (1995): Tense systems in European languages II. Tübingen: Niemeyer. THIEROFF, ROLF (2000): On the areal distribution of tense-aspect categories in Europe, in: DAHL, OSTEN ( e d . ) , 2 6 4 - 3 0 5 .

THIEROFF, ROLF & BALLWEG, JOACHIM (eds.) (1994): Tense systems

in European

languages.

(Lin-

guistische Arbeiten 308). Tübingen: Niemeyer. TRASK, ROBERT L. (1979): On the origins of ergativity, in: PLANK, FRANS (ed.), Ergativity:

toward a

theory of grammatical relations. London-New York: Academic Press, 385-404. VASILEV, CHRISTO ( 1 9 6 8 ) : D e r r o m a n i s c h e P e r f e k t t y p i m S l a v i s c h e n , in: KOSCHMIEDER, ERWIN &

BRAUN, MAXIMILIAN (eds.), Slavistische Studien. Munich: Trofenik, 121-143. VISSER, FREDERIKUSTH. (1963): An historical syntax of the English language, Vol. 1. Leiden: E.J. Brill.

Appendix

Map 1: /wve/fce-Auxiliation in Perfects be + have be only have only (excluding archaic Gk. Perf. in have/be in periphery) historically be, with some examples of have (esp. W.Slav., N.Rus dialects etc.) preterit greatly preferred over Perfect (Sic, Cal) or Periphrastic Perfect not found (Rus, Blr, Ukr, Pol, LSrb, Slve): old be Perfs > Past (be usually lost); likewise Hng., Trk: no Periphrastic Perfect per se, but categories which are closely related. Gag: no Perf, no AUX's. = tener used as AUX. [Brackets] {Curly Brackets} = haver remains as AUX., but main vb. > tener CAPS = be (+ 'after') + verbal noun (Ir, Wis, ScGl)] [Bold Underline Italics (Parentheses) Crossed out

= = = = =

Coord

I I

I

I

:s 'm 's 'him it'

I

wieso 'why'

es ihm 'it him'

\

I

Aber 'but'

6 wirst 'will' I

Wieso 'why'

ihn/'n 'him/him'

I

Du 'you'

pünktlich 'on time' pünktlich 'on time' sofort 'at once'

denn 'then' denn 'then'

'show'

ZEIGEN

'turn it'

ABGEBEN?

'turn it'

ABGEBEN!

heute? 'today'

heute! 'today'

auch 'also'

'see'

SEHEN

pünktlich 'on time'

heute 'today'

31 ευ β AgrOP

Extraposition

0. >

's'm 'it him'

Pronouns/ Clitics

£

SpecCP/IP

Thema

Rhema

j»α.

Λ Ü c ao (Λ u ΐο - 8I .2 -de c « η .ε 8 2

Ξ υ Ε Ο >

J tol

>

> r«

421 , cc tsε α tu si m f *

Tt·

κ Comp

main clause

main clause

1

i

Sie,L she/her/they /them 'them'

WP I /them 'she'

zahlen 'pay'

angesehen ha-st

nachher 'later'

earn 'him.DAT'

denn 'then'

abgeben! 'turn it'

gekommen sei-ds 'come be.2PL'

in Schwierigkeiten 'in trouble'

ja gar nicht gerne 'not with pleasure'

heute gleich 'today quickly'

in Schwierigkeiten 'in trouble'

ja gar nicht gerne 'not with pleasure'

TP/AgrP

sieshe/her/they SG

Middle Field

•s; -

I/Aux

Η I

is-qo

υ «5 %

viii

Ι S ο.3

SpecCP/IP

Ο

Front Field

C

Rhema

C/3 ε ρ Middle Field

-g

Middle Field

!

ΤΗ

ο α

TH/RH

> V+Aux

heute 'today'

so ohne Anlass 'without reason'

Extraposition

Post Field

ΤΗ

....

Appendix 271

Ed vorgestellt hat as himclitjc the HansNOM the new bossACC introduced has

307

Formal typology of pronouns in West Germanic

other hand, move only to SpecAFPi in order to license their destressing. As this license is violable, full nouns and XP-pronouns may stay to the right of the sentence adverb as well, if supported by appropriate prosodic patterns. The relevant data are given in (55a-c): (55a) da Jan Marie [AFPI dat [Vp gisteren gegeven eet]] that John Mary that yesterday given has (55b) da Jan Marie [Vp gisteren DAT gegeven eet] (55c) da Jan Marie [AFPI het boek [vp gisteren gegeven eet]] that John Mary the book yesterday given has (55d) da Jan Marie [Vp gisteren het BOEK gegeven eet]

(SA) (CA) (SA) (CA)

7.3.3. XP-pronouns in Afrikaans In German and Afrikaans, there is no morphological paradigm to support the deficient/strong distinction so that violation effects manifest themselves either purely prosodically (via assignment of contrastive stress in (46b) and (47b))19 or morphologically (via the insertion of the particle vir in (46c) and (47c)). In fact, we are now able to say more about the syntactic status of the discourse particle vir in Afrikaans. It is the morphological spell-out of an unchecked AF-feature, contained in the complex DP-layer of pronouns (cf. the next section for analysis). Analysing vir as the spell-out of some functional head corresponds to our empirical observations in section 5. The data have indicated that vir if used in two-place predicates is not a preposition, but should be analyzed below the X°-level. Vir is thus part of the functional nominal layer, just as C A R D I N A L E T T I & S T A R K E (1999) assume for the case marker a in Italian. However, its function is not to make a deficient element "strong". Recall that pronouns marked by vir do not necessarily correlate with strong properties in Afrikaans (cf. section 5). Rather, the insertion of vir indicates the marked, rhematic status of a prototypically thematic element. The proposed discourse functional value of vir can be supported by two observations. On the one hand, as rhemes proper, indefinites lack the destressing license, and as such, can never be associated with vir in Afrikaans. This holds irrespective of their surface position or semantic status (cf. MOLNÄRFI2002: 1130): (56) dat that

ek I

(*vir) meisies yesterday for girls

yesterday

gesoen het. kissed have

On the other hand, referential full definite nouns, just as pronouns, have to be supported morphologically in the relevant positions. This holds at least for spoken varieties of Afrikaans, where the on-line identification of grammatical and dis19

Just as Dutch, German also allows the shading of the stress-attracting position of SpecCP by prosodic intensification of some other element in the sentence: (i) Sie habe ich GESTERN geküsst. (CA) Interestingly, such prosodic shading seems to be more marked in the case of in-situ pronouns: (ii) lieh habe GESTERN sie geküsst.

(CA)

This may relate to the observation that the linearization of discourse functions in the sentence has to be thema precedes rhema (cf. LENERZ 1977). While in (i) the shaded (thematic) element precedes the main rhema of the sentence, in (ii) such thematic element follows the focalized sentence adverb.

308

LÄSZLÖ MOLNÄRFI

course functions (parsing) is facilitated by the existence of such morphological markers (cf. MOLNÄRFI 1999): (57a) dat ek gister *(yir) Tanya gesoen het. that I yesterday for Tanya kissed have (57b) *(Vir) Tanya het ekgister gesoen. (57c) dat ek Tanya gister gesoen het. As a comparison between (46c)/(47c) and (57a)/(57b) shows, VP-internal or topicalized nominal and pronominal definite NPs carry an identical morphological marking in Afrikaans. As this marking-strategy ensures a morphologically unique identification in discourse flow, prosodic distinctions are secondary in Afrikaans. Rhematic pronominal and nominal definites in situ or at the left-periphery may, but do not have be marked by contrastive stress. Our analysis indicates that, contrary to ABRAHAM (1997: 29), pronouns never change their categorial status if scrambled. Just as the insertion of the particle vir in (46c) does not trigger a categorial shift in Afrikaans (recall that pronouns marked by vir do not correlate necessarily with "strong" properties), contrastive stress cannot affect the syntactic status of an in-situ element. This is very clearly the case with definite full nouns. Nobody would argue for an analysis that takes the in-situ direct object in (42a) to belong to another syntactic class than the shifted definite object in (42b) on account of their different stressability properties. Rather, in the case of pronouns too, the structural conditions may change, under which referential definites have to satisfy their destressing license. Such conditions may yield anaphoric destressing, if the pronoun has moved out of VP to its appropriate checking site or assignment of contrastive stress, if the pronoun is outside its discourse functional chain. Effects of pronominal "strength" are thus violation effects, arising in derivations, where the destressing license cannot be satisfied. In some languages, they may be partly grammaticalized, excluding phonologically reduced forms from stressed positions. This is the reason for the more limited distribution of weak pronouns with respect to strong pronouns and the marked prosodic, discourse functional or morphological status of strong pronouns across languages.20 Weak and strong pronouns are thus two sides of the same coin that is stressed and destressed forms of the same XP-category. One could object that such classification does not account for the more restricted referential properties of strong pronouns as opposed to deficient forms. In fact, there is no deep reason for the [+human]-restriction, which, as indicated by empirical data from Afrikaans (cf. 24a and b), might be too strong to hold universally in the first place. As noted earlier, there exists a different paradigm for denoting stressed, coordinated or modified inanimate objects across languages: demonstrative pronouns. This simply renders the use of weak personal pronouns dysfunctional for the same purpose. Recall also the puzzling fact that (most of the) Dutch clitic pronouns share the [+human]restriction with strong forms (cf. section 6.5.2). This strongly indicates that referen20

Note that our proposal can also avoid the theoretically suspect rightward movement of weak pronouns to the VP for purposes of rhematic stressing (cf. ABRAHAM 1999: 46) and, more generally, the assumption that lexical material (i.e. weak pronouns) is base-generated outside VP (see ROSENGREN 1994, VAN DER DOES & DE Hoop 1998: 408, among others).

Formal typology of pronouns in West Germanic

309

tiality properties cannot provide direct clues about the internal structure of pronouns. If they did, clitics and strong pronouns should form one natural syntactic class in Dutch. Similarly, the state of phonological or morphological reduction does not allow direct conclusions about the complexity of the internal structure of pronouns. Indeed, the direction of phonological reduction may as well be derived from the discourse-linked status of deficient pronouns. As proven by phonological experiments (cf. NOOTEBOOM & KRUYT 1 9 8 7 and TERKEN & NOOTEBOOM 1988), speakers readily interpret deaccented elements as discourse-linked. This derives the correlation between the referentiality properties and the phonologically weak status of pronouns without recourse to a theory of structural deficiency. Deficient elements are not lighter because they contain less structure in terms of a morphological deficiency. Rather, discourse linked, thematic elements are very likely to be destressed, which may lead to phonological reduction.

7.4. The structure of pronouns Similarly to CARDINALETTI & STARKE ( 1 9 9 9 ) , I also assume a complex functional layer for pronouns and nouns in Afrikaans and other West Germanic languages. The nominal functional layers, however, do not mimic the architecture of the sentential level. Rather, they are a complex domain, licensing referentiality properties of (pro)nouns. Such NP-complex consists of a higher DP, embedding a lower DP, which, in turn, contains the lexical core. Such double DP-layer has been independently motivated by the properties of the possessive marker se in Afrikaans (cf. OOSTHUIZEN & WAHER 1994): (58)

DP,

In (58) the higher DP is taken to encode referentiality in form of a destressing license, while the lower D-head may be specified for [±definiteness]. As pronouns are category-inherently definite, pronominal NPs, but not full NPs, will raise to

310

LÄSZLÖ MOLNÄRFI

SpecDP2 to check their definiteness feature. The position of pronouns is thus a derived one within the DP-complex, while their status is phrasal, irrespective of strength or weakness. The higher D-head is related to the referential properties of the (pro)noun. If the lower D-head is positively specified for definiteness, the higher D-head will carry a destressing license in form of the AF-feature. However, as there is no AF-feature on the (pro)noun itself, the whole DP-complex must be moved to Spec-AF to satisfy the checking requirements of AF. Checking of the destressing licence will eliminate the triggering feature, with the discussed prosodic and discourse effects obtained. If, however, AF-licensing is violated by topicalization (which is movement along an A'-chain to SpecCP) or by leaving the (pro)noun in situ, the AF-feature will survive to PF. The presence of AF after Spell-out may trigger morphological and/or prosodic effects. In Afrikaans, the surviving feature is spelled out morphologically, lexicalizing the higher D-head. The result of this spell-out is the particle vir. In other West Germanic languages, the violation of the AF-licensing triggers purely prosodic effects. Clitics are taken to be base generated in the lower D-head as X°-elements. While this entails that the complex DP-containing a clitic pronoun must raise to Spec-AF, the clitic, on account of its head status, must raise further, incorporating into a syntactic head for phonological support. The complex DP-layer, so to speak, saves the clitic from violating head movement (cf. ABRAHAM & WlEGEL 1992). Such a clitic is taken along as a free-rider (CHOMSKY 1993), to be incorporated before spell-out. This explains why clitics have mixed properties of a head and a maximal projection. Until the stage of AF-checking, clitics are part of a larger nominal structure. However, after AF-checking the clitic head moves out, attaching itself to a appropriate host, such as C or T. This is the reason why clitics cannot stay in argument positions within VP or be topicalized: While the AF-licensing may be violated, the head-license must be respected. Movement of pronouns of a phrasal status, on the other hand, is only triggered by a destressing license, which is violable under certain conditions.

8. Conclusions Deficiency is not a core property of pronouns. Rather, deficiency denotes a bundle of epiphenomenal properties rooted in the inherent thematic discourse value of personal pronouns. The reason why pronouns distribute the way they do is that they share some common grammatical property to be licensed and not because they have to "recover" missing parts of their structure. In this paper, I have argued that a unified trigger for leftward object shift of referential füll nouns and pronouns is the need to evacuate the focus domain and license discourse linking in terms of destressing. I have proposed the following typological classification of pronouns in West Germanic:

311

Formal typology of pronouns in West Germanic

Personal Pronouns

(59)

XP(weak/strong) +def +AF

X° (clitic) +def +AF AF-checking

AF-checking

»Head-incorporation Spell-out of the base position

Spell-out

of the head position

XP [+AF]

XP [-AF]

[-AF]

-reduced +stressed

±reduced -stressed

+reduced -stressed

Pronouns are category-inherently definite elements carrying a formal destressing license in form of an antifocus-feature. Contrary to CARDINALETTI & STARKE (1996), I have only assumed one unified class of XP-pronouns and another class of clitic pronouns with head-status in West Germanic. Within the class of XP-pronouns, the old distinction between weak and strong has no syntactic relevance, holding only after Spell-out, at PF. Weak and strong are properties which directly relate to the stressability conditions of pronouns. I have assumed that antifocus constitutes a soft trigger, which is violable under certain circumstances. That is, while certain syntactic contexts (topicalization or the in-situ status of pronouns) do not allow satisfying the destressing license, the resulting derivations do not crash, still converging at PF. The presence of an unchecked antifocus feature at PF [+AF] triggers prosodic or morphological effects (cf. Dutch and Afrikaans respectively) as well as the selection of a phonologically non-reduced pronominal form (if the given language paradigmatically supports such distinctions). The absence of the antifocus feature at PF ([-AF]) correlates with the absence of stress only. In other words, there is no one-to-one relation between strong (phonologically non-reduced) forms and strong pronominal properties, not even in languages, in which the stressability conditions (partially) correspond to a morphologically distinct paradigm. The source of cross-linguistic variation is the parametrization of the [±reduced]property with respect to the formal destressing position of pronouns. Weakness and strength are distinctions which directly relate to the distribution of pronouns, i.e., they can be read off of their respective surface position. Some West Germanic languages (but by no means all of them!) have phonologically reduced pronouns, which have a status of a syntactic head. Such clitic pro-

312

LÄSZLÖ MOLNÄRFI

nouns are also subject to the licensing conditions of destressing. Additionally, on account of their non-phrasal status, they have to incorporate into a syntactic host before spell-out. As such incorporation process is (a) not violable (b) presupposes AF-checking, only those clitics will survive that have been formally destressed in the appropriate AF-position. This explains why the distribution of clitics versus phonologically reduced XP-pronouns is more limited, while their stressability properties are similar.

References (1995a): De discourse-functionele motivatie voor de plaatsing van voomaamwoord-clitics, in: TABUlSß, 107-111. ABRAHAM, WERNER (1995b): Morphological case: no need for functional projections in German, in: JONKERS, ROEL; KAAN, EDITH & WIEGEL, ANKO (eds.), Language and cognition 5. Groningen, 1 - 1 2 . ABRAHAM, WERNER ( 1 9 9 7 ) : The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative ones, in: ABRAHAM, WERNER & VAN GELDEREN, ELLY (eds.), German: syntactic problems-problematic syntax. Tübingen: Niemeyer, 1 1 - 4 3 . ABRAHAM, WERNER ( 1 9 9 9 ) : Präteritumschwund in German: the Parsing Trigger, in: ABRAHAM, WERNER (ed.), Characteristic properties of spoken vernaculars. (Folia Linguistica 3 3 ( 1 ) , special issue). Berlin: de Gruyter, 3 9 - 5 8 . ABRAHAM, WERNER ( 2 0 0 2 ) : Pronomina im Diskurs: deutsche Personal- und Demonstrativpronomina unter 'Zentrierungsperspektive'. Grammatische Überlegungen zu einer Teiltheorie der Textkohärenz, in: Sprachwissenschaft 2 7 . 4 , 4 4 7 - 4 9 1 . ABRAHAM, WERNER (this volume): The European demise of the simple past and the emergence of the periphrastic perfect. ABRAHAM, WERNER & WIEGEL, ANKO (1992): Reduktionsformen und Kasussynkretismus bei deutschen und niederländischen Pronomina, in: Groninger Arbeiten zur germanistischen Linguistik 3 5 , 1 - 3 6 . ABRAHAM, WERNER & CONRADIE, C . JAC ( 2 0 0 1 ) : Präteritumschwund und Diskursgrammatik. Amsterdam: Benjamins. ABRAHAM, WERNER & MOLNÄRFI, LÄSZLÖ ( 2 0 0 2 ) : The German clause under discourse functional weight: focus and antifocus, in: ABRAHAM, WERNER & ZWART, C . JAN-WOUTER (eds.), Issues in comparative German(ic) typology. (Linguistik Aktuell/Linguistics Today 45). Amsterdam: Benjamins, 1-43. BAYER, JOSEF (1999): Comment on Cardinaletti and Starke 'The Typology of structural deficiency', in: VAN RIEMSDIJK, HENK (ed.), Clitics in the languages ofEurope. Berlin: de Gruyter, 235-241. BOSKOVIC, 2 E U K O (2002): On multiple Wfc-fronting, in: Linguistic Inquiry 33/3, 351-383. CARDINALETTI, ANNA & STARKE, MICHAEL (1996): Deficient pronouns: a view for Germanic, in: THRÄINSSON, HOSKULDUR; EPSTEIN, SAMUAL D. & PETER, S. (eds.), Studies in comparative Germanic syntax II. Amsterdam: Kluwer, 21-65. CARDINALETTI, ANNA & STARKE, MICHAEL (1999): The typology of structural deficiency: a case study of the three classes of pronouns, in: VAN RIEMSDIJK, HENK (ed.), Clitics in the languages of Europe. Berlin: de Gruyter, 145-232. CHOMSKY, NOAM (1993): A minimalist program for linguistic theory, in: HALE, KENNETH & KEYSER, SAMUEL J. (eds.), The view from building 20. Essays in honor of Sylvian Bromberger. Cambridge: MIT Press, 1-52. CHOMSKY, NOAM (1995): The minimalist program. Cambridge: MIT Press. CHOMSKY, NOAM ( 1 9 9 8 ) : Minimalist inquiries: the framework. Ms. MIT, 1 - 5 6 . (published as CHOMSKY, NOAM ( 2 0 0 0 ) , in: MARTIN, R . ; MICHAELS, D . & URIAGEREKA, J. (eds.), Step by step. Cambridge: MIT-Press, 8 S M 5 5 . ) CINQUE, GUGLIELMO ( 1 9 9 3 ) : A null theory of phrase and compound stress, in: Linguistic Inquiry ABRAHAM, WERNER

24(2), 2 3 9 - 2 9 8 . CORVER, NORBERT & VAN RIEMSDIJK, HENK

tive Grammar 41). Berlin: de Gruyter.

(eds.)

(1994):

Studies on scrambling. (Studies in Genera-

313

Formal typology of pronouns in West Germanic CZEPLUCH, HARTMUT

(this volume): Reflections on the form and function of passives in English and German.

DIESING, MOLLY (1992): Indefinites. Cambridge: MIT Press. DIESING, MOLLY (1999): Comments on Cardinaletti and Starke: 'The typology of structural deficiency', in: VAN RIEMSDUK, HENK(ed.), Clitics in the languages of Europe. Berlin: de Gniyter, 243-247. DOBROVIE-SORIN, CARMEN (1999): The typology of pronouns and the distinction between syntax and morphophonology, in: VAN RIEMSDUK, HENK (ed.), Clitics in the languages of Europe. Berlin: de Gruyter, 2 4 9 - 2 5 5 . VAN DER DOES, JAAP & DE HOOP, HELEN (1998): Type-shifting and scrambled defmites, in: Journal of Semantics 15, 393-416. FANSELOW, GISBERT (1987): Konfigurationalität - Untersuchungen zur Universalgrammatik am

Beispiel des Deutschen. (Studien zur deutschen Grammatik 29). Tübingen: Narr. FIRBAS, JAN (1992): Functional sentence perspective in written and spoken communication. bridge: Cambridge University Press.

Cam-

GEERTS, G.; HAESERYN, W . ; DE ROOIJ, J. & VAN DEN TOORN, M . G . (eds.) ( 1 9 8 4 ) : Algemene

Neder-

landse Spraakkunst (ANS): Groningen: Wolters-Noordhoff. GREWENDORF, GÜNTHER (2002): Minimalistische Syntax. Tübingen: A. Francke. GROAT, E . & O ' N E I L , J. ( 1 9 9 6 ) : S p e l l - o u t at t h e L F - i n t e r f a c e , in: ABRAHAM, WERNER; EPSTEIN, SAMUAL D . ; THRÄINSSON, HOSKULDUR & ZWART, C . JAN-WOUTER (eds.), Minimal ideas. A m -

sterdam: Benjamins, 113-139. HAEGEMAN, LILIANE (1991): On the relevance of clitic placement for the analysis of verb second in West Flemish, in: The Linguistic Review 7, 333-363. HAIDER, HUBERT (1993): Deutsche Syntax - generativ. Vorstudien zur Theorie einer projektiven Grammatik. Tübingen: Narr. HAIDER, HUBERT & ROSENGREN, INGER ( 1 9 9 8 ) : S c r a m b l i n g , in: Sprache

und Pragmatik

49,1-104.

HÖHLE, TILLMAN, Ν. (1982): Explikationen für 'normale Betonung' und 'normale Wortstellung', in: ABRAHAM, WERNER (ed.), Satzglieder im Deutschen. Tübingen: Narr, 75-153. HOLMBERG, ANDERS (1999): Comments on Cardinaletti and Starke: 'The typlogy of structural deficiency', in: VAN RIEMSDIJK, HENK (ed.), Clitics in the languages of Europe. Berlin: de Gruyter, 263-266. DE HOOP, HELEN (2000): O p t i o n a l s c r a m b l i n g a n d interpretation, in: BENNIS, HANS; EVERAERT, MARTIN

& REULAND, ERIC (eds.), Interface strategies. Amsterdam: Royal Netherlands Academy, 153-168. DE HOOP, HELEN & KOSMEUER, WIM (1995): Case and scrambling: D-structure vs. S-structure, in: HAIDER, HUBERT; OLSEN, SUSAN & VIKNER, STEN ( e d s . ) : Studies

in comparative

Germanic

syn-

tax. Dordrecht: Kluwer, 139-159. Kiss, KATALIN EVA (1998): Identificational focus vs. information focus, in: Language 74,245-273. LAMBRECHT, KNUD (1994): Information structure and sentence form. Cambridge: Cambridge University Press. LENERZ, JÜRGEN (1977): Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: Narr. LENERZ, JÜRGEN (1993): Zu Syntax und Semantik deutscher Personalpronomina, in: REIS, MARGA

(ed.), Wortstellung und Informationsstruktur.

Tübingen: Niemeyer, 117-153.

LENERZ, JÜRGEN (1994): Pronomenprobleme, in: HAFTKA, BRIGITTA (ed.), Was determiniert

lungsvariationl

Wortstel-

Opladen: Westdeutscher Verlag, 161-173.

LENERZ, JÜRGEN (2001): Word order variation: competition or co-operation?, in: MÜLLER, GEREON &

STERNEFELD, WOLFGANG (eds.), Competition in syntax. (Studies in Generative Grammar 49). Berlin: de Gruyter, 249-281. MEINUNGER, ANDRE (2000): Syntactic aspects of topic and comment. (Linguistics Today 38). Amsterdam: Benjamins. MOLNÄRFI, LÄSZLÖ (1997): Preposisie of kasusmarkeerder? Oor die status van vir in Afrikaans, in: Tydskrifvir Geesteswetenskappe 37/2, 89-106. MOLNÄRFI, LÄSZLÖ (1999): Zur Diskurskomponente in der Sprechsprache: vir als Rhemamarkierer im

Afrikaans, in: ABRAHAM, WERNER (ed.), Characteristic properties in spoken vernaculars. (Folia Linguistica 33(1), special issue). Berlin: de Gruyter, 75-100. MOLNÄRFI, LÄSZLÖ (2002): Focus and antifocus in modern Afrikaans and West Germanic, in: Linguistics 40/6, 1107-1160.

314 MOLNÄRFI, LASZLÖ (2003): On optional movement and feature ABRAHAM, WERNER & MOLNÄRFI, LASZLÖ (eds.), Optionality

LASZLÖ MOLNÄRFI

checking in West Germanic, in: in syntax and discourse structure. Aspects of word order variation in (West) Germanic and other languages. Folia Linguistica 37/1-2,129-162. MÜLLER, GEREON (1999): Optimality, markedness and word order in German, in: Linguistics 37/5, 777-818. NEELEMAN, A D (1994): Complex predicates. Doctoral Dissertation. Utrecht: O T S . NEELEMAN, A D & REINHART, TANYA ( 1 9 9 8 ) : Scrambling and the P F interface, in: BUTT, MIRIAM & GLIEDER, WILHELM (eds.), The projection of arguments: lexical and compositional factors. Stanford: CSLI-publications, 3 0 9 - 3 5 3 . NOOTEBOOM, SIEB G . & KRUYT, JOHANNA G . (1987): Accents, focus distributions and the perceived distribution of given and new information, in: Journal of the Acoustical Society of America 82(5), 1512-1524. OOSTHUIZEN, JOHAN & WAHER, HEESTER ( 1 9 9 4 ) : On the syntax of the se-construction in Afrikaans, in: Stellenbosch Papers in Linguistics 2 8 , 2 1 - 4 3 . POLLOCK, JEAN-YVES (1989): Verb movement, universal grammar and the structure of I P , in: Linguistic Inquiry 20, 365-424. PONELIS, FRITZ (1979): Afrikaanse Sintaksis. Pretoria: Van Schaik. PONELIS, FRITZ ( 1 9 9 3 ) : The development of Afrikaans. Frankfurt am Main: Peter Lang. RAIDT, EDITH ( 1 9 7 6 ) : Die herkoms van objekkonstruksies met vir, in: 1875-1975 Studies oor die Afrikaanse taal. Feesbundel van die SA Akademie vir Wetenskap en Kuns. Johannesburg: Lex patria, 7 2 - 1 0 1 . ROSENGREN, INGER (1994): Scrambling - was ist das?, in: HAFTKA, BRIGITTA (ed.), Was determiniert Wortstellungsvariation? Opladen: Westdeutscher Verlag, 175-195. SELKIRK, ELISABETH O . ( 1 9 8 4 ) : Phonology and syntax: the relation between sound and structure. Cambridge: MIT Press. TERKEN, JACQUES & NOOTEBOOM, SIEB G . (1988): Opposite effects of accentuation and deaccentuation on verification latencies for given and new information, in: Language and Cognitive Processes2(3/4), 145-163. THRÄINSSON, HÖSKULDUR ( 2 0 0 1 ) : Object shift and scrambling, in: BALTIN, MARK & COLLINS, CHRIS (eds.), The handbook of contemporary syntactic theory. Oxford: Blackwell, 1 4 8 - 2 0 2 . VIKNER, STEN ( 1 9 9 4 ) : Scandinavian object shift and West Germanic scrambling, in: CORVER & VAN RIEMSDUK (eds.), 4 8 7 - 5 1 7 . WEBELHUTH, GERT ( 1 9 9 2 ) : Principles and parameters of syntactic saturation. Oxford: Oxford University Press. ZUBIZARETTA, MARIA LUISA (1998): Prosody, focus and word order. (Linguistic Inquiry Monographs 33). Massachusetts: MIT Press. ZWART, C . JAN-WOUTER ( 1 9 9 7 ) : Morphosyntax of verb movement. A minimalist approach to the syntax of Dutch. Dordrecht: Kluwer.

The subjunctive mood in German and in the Germanic languages ROLF THIEROFF (Osnabrück)

0. Introduction Together with Icelandic, German is in many respects the most conservative of the Germanic languages. This holds true especially for the morphology. In the noun phrase, the Scandinavian languages, Dutch and English have abandoned case marking almost completely (with several exceptions concerning the personal pronouns), whereas German has preserved the marking of four cases, with case markings on the noun, the pronoun, the article and the adjective. With regard to the morphology of the verb, the Mainland Scandinavian languages have abandoned person and number marking altogether, English has preserved only the marking of the 3rd person singular in the present, and in Dutch in the preterite only number is marked and in the present person is only marked in the singular, whereas German has preserved person and number marking in both present and past. And finally, in the Scandinavian languages, in Dutch and in English the morphological mood subjunctive has almost completely vanished, whereas in Icelandic and, to a somewhat lesser degree, in German the subjunctive is vivid and alive. In German, the functions of subjunctive forms with the finite form in the present (subjunctive 1) are well distinguished from those forms with the finite form in the preterite (subjunctive 2). This same difference can also be observed for the few remnants of the subjunctives in the other Germanic languages. Where the subjunctive is lost, the former subjunctive 1 is replaced by other means of expression than the former subjunctive 2. In Icelandic and German the subjunctive has developed a special function not found in the remaining Germanic languages, namely the function of indicating reported speech.

1. Morphology of the subjunctive in German 1.1. Present subjunctive In present day German, the forms of the subjunctive do not differ in all tenses, persons, and numbers from the forms of the indicative. As in the other Germanic languages, with respect to the formation of the preterite, weak verbs have to be distinguished from strong verbs.

316

ROLF THIEROFF

In the present the subjunctive is marked by schwa (written e) which is inserted between the verb stem and the person-number suffix. In the third person singular, the person-number suffix is omitted in the subjunctive. If the person-number suffix already contains e, a second e (to mark the subjunctive) cannot be added. As a result, in the first person singular and in the first and third person plural, the forms of the subjunctive do not differ from the forms of the indicative. This holds for the weak as well as for the strong verbs (Table 1). weak verb (leben 'live') subjunctive indicative ich leb-e = leb-e du leb-st leb-e-st er/sie/es leb-t leb-e-_ wir leb-en = leb-en ihr leb-t leb-e-t leb-en sie leb-en =

strong verb (sehen 'see') indicative subjunctive seh-e = seh-e sieh-st seh-e-st sieh-t seh-e-_ seh-en = seh-en seh-t seh-e-t seh-en = seh-en

Table 1: Present indicative and present subjunctive of weak and strong verbs The future and the future perfect are built with the auxiliary werden 'become' in the finite form. The auxiliary werden is itself a strong verb, conjugated like sehen. Consequently, in the future and in the future perfect too, the indicative and the subjunctive are homonymous in the first singular and in the first and third plural (ich werde leben/sehen Ί will live/see' etc., ich werde gelebt/gesehen haben Ί will lived/seen have' i.e. Ί will have lived/seen' etc.). The perfect is built with the auxiliary haben 'have' (transitive verbs and intransitive atelic verbs) or with the auxiliary sein 'be' (telic intransitive verbs). The auxiliary haben is again conjugated like sehen, with the result that the perfect of the verbs forming their perfect with haben have again homonymous forms in the indicative and subjunctive of the first singular and the first and third plural (ich habe gelebt/gesehen etc. Ί have lived/seen' etc.). The only verb whose present subjunctive forms differ in both numbers in all persons from the indicative forms is the verb sein 'be', whose present subjunctive forms are built with the stem sei. As a result, there are no indicative/subjunctive homonymies in the perfect of telic intransitive verbs (ich bin gekommen vs. ich sei gekommen Ί have [lit.: am] come' etc.; Table 2). ich du er/sie/es wir ihr sie

indicative bin bist ist sind seid sind

subjunctive sei seist sei seien seiet seien

Table 2: Present indicative and present subjunctive of the verb sein 'be'

The subjunctive mood in German and in the Germanic languages

317

1.2. Preterite subjunctive With weak verbs, in the preterite the indicative and the subjunctive are not morphologically differentiated. The answer to the question whether there are homonymous forms of the indicative and the subjunctive in all persons and numbers or if there is simply no mood distinction here is not straightforward. I will come back to this problem in section 2.2. In Table 3, the traditional view of the German grammars is presented. leben 'live' ich du er/sie/es wir ihr sie

indicative lebte lebtest lebte lebten lebtet lebten

= = = =

subjunctive lebte lebtest lebte lebten lebtet lebten

Table 3: Preterite indicative and preterite subjunctive of weak verbs (traditional view) One reason for the analysis according to Table 3 is the fact that in strong verbs, the indicative and the subjunctive are well distinguished in the preterite. As in the present, the subjunctive is marked by -e, inserted between the stem and the person/number suffix. Only if the vowel of the preterite stem is a front vowel is there a homonymy of the first and third person singular (Table 4). fallen 'fall' ich du er/sie/es wir ihr sie

indicative fiel fielst fiel fielen fielt fielen

=

subjunctive fiel-e fiel-e-st fiel-e fielen fiel-e-t fiel-e-n

Table 4: Preterite indicative and preterite subjunctive of strong verbs with front vowel If the vowel of the preterite stem is a back vowel, the subjunctive is always umlauted. Thus with back vowel preterite stems, all six forms are distinguished between indicative and subjunctive (Table 5).1

1

Since all subjunctive forms are distinguished by the umlaut, -e can be omitted in the second person singular and plural (but not in the first and third singular - at least not in the written standard).

318

ROLF THIEROFF

kommen 'come' indicative ich kam du kamst er/sie/es kam wir kamen ihr kamt kamen sie

subjunctive käme käm(e)st käme kämen käm(e)t kämen

Table 5: Preterite indicative and preterite subjunctive of strong verbs with back vowel Finally, it is noteworthy that both the auxiliary haben 'have' and the auxiliary sein 'be' have a preterite with the back vowel [a], with the consequence that the subjunctive of these forms is umlauted, which in turn means that in the past perfect of all verbs (weak and strong) indicative and subjunctive in all six positions of the paradigm are morphologically distinguished (Table 6).

ich du er/sie/es wir ihr sie

leben 'live' indicative hatte gelebt hattest gelebt hatte gelebt hatten gelebt hattet gelebt hatten gelebt

subjunctive hätte gelebt hättest gelebt hätte gelebt hätten gelebt hättet gelebt hätte gelebt

kommen 'come' indicative war gekommen warst gekommen war gekommen waren gekommen wart gekommen waren gekommen

subjunctive wäre gekommen wär(e)st gekommen wäre gekommen wären gekommen wär(e)t gekommen wären gekommen

Table 6: Pluperfect indicative and pluperfect subjunctive of verbs with haben and sein

1.3. Subjunctive 1 and subjunctive 2 In German grammatical tradition, a difference is being made between what is called the Konjunktiv I (subjunctive 1) and the Konjunktiv II (subjunctive 2) in a way that all subjunctive forms either belong to the subjunctive 1 or to the subjunctive 2. One criterion for belonging to either of these two groups is a morphological criterion: each form whose finite part is in the present is a subjunctive 1, each form whose finite part is in the preterite is a subjunctive 2. This renders the picture in Table 7.

The subjunctive mood in German and in the Germanic

319

languages

singen 'sing' subjunctive 1 (er) singe (er) habe gesungen (er) werde singen (er) werde gesungen haben

present: present perfect: future: future perfect: preterite: pluperfect:

subjunctive 2

(er) sänge (er) hätte gesungen

Table 7: Forms of subjunctive 1 and subjunctive 2 A second reason for making this distinction is the fact that the subjunctive 1 and the subjunctive 2 do not differ with regard to time reference. For example, both er singe and er sänge have non-past time reference and differ only with regard to their modal meaning. In contrast, in the indicative the present form er singt has non-past time reference, whereas the preterite form er sang has past time reference. Since this same behaviour holds for the subjunctive forms in all Germanic (and in most other European) languages, I shall adopt the terms subjunctive 1 and subjunctive 2 for the subjunctive forms in the other Germanic languages, too. Thus, in what follows, the term subjunctive 1 is used for any subjunctive form whose finite part is in the present, the term subjunctive 2 is used for any subjunctive whose finite part is in the preterite. 1.4. würde + infinitive There are two more forms which are missing in Table 7: the form würde singen and the form würde gesungen haben. Würde is a form of the auxiliary werden, which is conjugated like any ordinary strong verb. Together with the perfect participle, the auxiliary werden serves to build the passive. In the passive, the auxiliary can appear in both present and preterite and in both indicative and subjunctive. In (1), werden has the function of a passive auxiliary. (la)

Der Brief wird the letter become-PRES:lND:3SG 'The letter is being written.' (lb) Der Brief werde the letter become-PRES:SUBJ:3SG 'The letter shall be written.' (lc) Der Brief wurde the letter become-PRET: IND: 3 SG 'The letter was being written.' (Id) Der Brief würde the

letter

become-PRET: SUBJ: 3 SG

geschrieben. write-PART.-PERF

geschrieben. write-PART.PERF

geschrieben. write-PART:PERF

geschrieben,

wenn.

write-PART:PERF

if...

'The letter would be written, if...' The auxiliary werden can also be combined with the infinitive to yield the future. However, the combination of the preterite indicative with the infinitive is not possible. Consider (2):

320 (2a)

R O L F THIEROFF

Er

wird

he

become-PRES:lND:3SG

he

become-PRES:SUBJ:3SG

'He will write a letter.' werde (2b) Er (2c) (2d)

'He will write a letter.'2 *Er wurde

he become-PRET: IND: 3 SG [not interpretable]

Er he

würde

become-PRET: SUB J: 3 SG

einen

Brief

schreiben.

einen

Brief

schreiben.

einen

Brief

schreiben.

einen

Brief

schreiben.

a a a a

letter

letter letter

letter

write-iNF

write-iNF write-iNF write-INF

'He would write a letter.' From a morphological point of view, the form würde schreiben is unproblematic: it is a future-in-the-past (or future preterite) subjunctive. But: a corresponding future-in-the-past indicative form does not exist and this lacuna implies several problems. The construction würde + infinitive has been analyzed in many different ways. According to J0RGENSEN'S analysis the missing *wurde singen has been replaced by würde singen, i.e. there is an indicative form würde singen and a second, homonymous subjunctive form würde singen (J0RGENSEN 1966: 28 and passim).3 ZLFONUN et al. (1997: 1735) stress that the VVWNFE-construction is systematically outside the paradigm of both the indicative and the subjunctive forms. The authors acknowledge that the form can be used like an indicative and like a subjunctive, but it does not become clear what the construction really is supposed to be. In a thorough discussion of all functions of the wm/e-construction, FABRICIUS-HANSEN (2000) comes to the conclusion that it is the future preterite subjunctive and nothing else. ABRAHAM (1999: 382) claims that we have to deal with a mood in its own right, a third mood which is neither indicative nor subjunctive but a Potentialis or Irrealis. What is certain and beyond any doubt is that from a purely morphological point of view würde singen is in any case future (qua auxiliary werden) and preterite (qua high rounded vowel [be it back or fronted]). For the moment, I leave the decision open whether it is also a subjunctive or whether it is ambiguous between indicative and subjunctive, or neither of the two. Instead, constructions like würde singen will be indicated as FUT PRET and constructions like würde gesungen haben a s FUT PRET PERF.

I will come back to this problem in section 3.3.3., where I shall argue that neither of the above analyses is correct. Instead, I shall propose that the würdeconstruction is a future preterite which is not associated with any mood category.

2 3

(2b) differs from (2a) in that it is reported speech. See 2.3., 4.3. This is also the view taken in THIEROFF (1992).

321

The subjunctive mood in German and in the Germanic languages

2. Functions of the subjunctive in German There are three domains of the subjunctive in present day German. The forms of the subjunctive 1 are predominantly used with an optative meaning in the largest sense of the word. The forms of the subjunctive 2 can be used with a counterfactual meaning. As we will see below, these two functions are uses which all Germanic languages have in common. In addition, in Icelandic and German both the subjunctive 1 and the subjunctive 2 are used to indicate reported speech. This third function is not found in the other Germanic languages (see section 4.).

2.1. Subjunctive 1 Apart from reported speech, the subjunctive 1 is archaic in all its uses. This is true for the optative use proper, which occurs almost exclusively in the third person singular (where the subjunctive 1 always differs formally from the indicative; cf. Table 1) and which could be regarded as a form substituting the missing imperative of the third person. The example in (3), from a grammar of German, sounds very archaic and has little chance to be uttered by a German speaker living today. (3)

Wer who der he 'Who

schadhaftes Gerät empfangen hat, defective equipment receive-PART:PERF have-PRES:iND:3SG gebe es zurück. [FLÄMIG1991:411] give-PRES:SUBJ:3SG it back has got defective equipment shall give it back.'

In many grammars it is maintained that this use of the subjunctive 1 occurs frequently in cookery books; however, ENGEL ( 1 9 8 8 : 4 2 0 ) is right in observing that even in recipes the subjunctive 1 is extremely rare today. An example for this use is given in (4). (4)

Man nehme zwei Eier, ein Pfund Mehl one take-PRES:SUBJ:3SG two eggs one pound flour ein Pfund Zucker. one pound sugar 'Take two eggs, one pound of flour and one pound of sugar.'

und and

Only in mathematical texts as in (5) the optative use of the subjunctive 1 still seems to be less infrequent today: (5)

ABC sei ABC be-PRES:SUBJ:3SG 'Let ABC be a triangle.'

ein a

Dreieck. triangle

In everyday speech, however, the optative use of the subjunctive 1 is largely restricted to certain formulae such as

322

ROLF THIEROFF

(6a) Er lebe he live-PRES: SUB J: 3 SG 'Cheers to him!' (6b) Es lebe it live-PRES:SUBJ:3SG 'Long live the first of May!' (6c) Gott sei God

be-PRES:SUBJ:3SG

hoch! high der 1. Mail the first May Dank! thanks

'Thank God!' In a larger sense, the subjunctive 1 has also optative meaning in final subordinate clauses, as in (7). This use is archaic too. In contemporary German, the indicative is used instead. (7)

Er

sprach

sehr

langsam

und

deutlich,

he

speak-PRET:lND:3SG

very

slow

and

articulate

damit

sie

ihn

verstehe.

that

she

him

understand-PRES:SUBJ:3SG

'He spoke very slowly and articulated in order that she understood him.' Finally, the subjunctive 1 is used in two more cases, where it does not have an optative meaning. The first is in certain concessive clauses, where it is today restricted to certain formulae such as (8a) Komme, was cqme-PRES: SUB J: 3 SG what 'Come what may, ...' (8b) Sei dem,

da there wie

be-PRES:SUBJ:3sG that-DAT:SG how ' B e that as it m a y , . . . '

(8c)

wolle,... will-PRES:SUBJ:3SG

ihm

wolle,...

it-DAT:SG Will-PRES:SUBJ:3SG

Wie

dem

auch

sei, ...

how

it-DAT: SG

PTL

be-PRES:SUBJ:3SG

'However it may be,...' The second non-optative use is in unreal comparative clauses. These clauses are the only cases (besides indirect speech) where the subjunctive 1 and the subjunctive 2 can both be used with the same meaning. (9)

Es

sieht

it

look-PRES:lND:3SGout

aus, als habe!

hätte have-PRET:SUBJ:3SG

it

'It looks as if it has rained.'

as

have-PRES:SUBJ:3SG

es

geregnet.

rain-P ART: PERF

The subjunctive mood in German and in the Germanic

languages

323

2.2. Subjunctive 2 The main function of the subjunctive 2 is the indication of counterfactuality. It is important to note that the subjunctive 2 forms do not have the same time reference as their indicative counterparts. Whereas the preterite indicative has past time reference, the preterite subjunctive only occurs with non-past time reference. And contrary to the pluperfect indicative, which indicates that an event took place before a reference time which in turn is before speech time, the pluperfect subjunctive merely indicates past time reference. In other words, the preterite subjunctive has the same time reference as the present indicative (and the present subjunctive) and the pluperfect subjunctive has the same time reference as have present perfect and preterite indicative (and present perfect subjunctive). Compare: (10a) Jetzt / morgen scheint die Sonne. now tomorrow shine-PRES:lND:3SG the sun 'Now the sun is shining/tomorrow the sun will be shining.' (10b) Gestern schien die Sonne / the sun yesterday shine-PRET:iND:3SG hat die Sonne geschienen. have-PRES:iND:3SG

the

sun

shine-PART: perf

'Yesterday the sun was shining.' (IIa) Wenn doch jetztI morgen die Sonne schienel if only now tomorrow the sun shine-PRET:SUBJ:3SG 'If only the sun were shining now/tomorrow!' geschienen (lib) Wenn doch gestern die Sonne shine-PART:PERF if only yesterday the sun hätte! have-PRET:SUBJ:3SG

'If only the sun had been shining yesterday!' The examples in (10) and (11) are so-called contrary-to-fact wishes. Most counterfactual subjunctives occur, however, in conditional sentences. In counterfactual conditional sentences, both the main clause and the «^clause can be in the subjunctive, as in (12).

(12) Wenn jetzt if ginge

now

die the

gO-PRET:SUBJ:lSG

Sonne

schiene,

sun

shine-PRET:SUBJ:3SG

ich

spazieren.

I

walk-lNF

'If the sun were shining now, I would go for a walk.' Note that (12) is somewhat exceptional in as far as both the protasis and the apodosis contain a strong verb in the 3rd person singular, i.e. a form which is unambiguously subjunctive. However, as in the other Germanic languages, the number of strong verbs is by far smaller than the number of weak verbs. If we now replace the strong verb by a weak verb, either in the protasis or in the apodosis, this verb form is not morphologically different from the indicative. Yet, the whole

324

ROLF THIEROFF

sentence is still understood as a counterfactual conditional sentence, the preterite of the weak verb is interpreted as a subjunctive form: (13) Wenn es if

it

regnete, rain-PRET:iND/SUBJ:3SG

bliebe

ich

zu Hause.

stay-PRET: SUB J: 1SG

I

at home-DAT: SG

'If it were raining, I would stay at home.' (14) Wenn die Sonne schiene, if

the sun

shine-PRET:SUBJ:3SG

arbeitete

ich

im

Garten.

workPRET:iND/suBJ:lSG

I

in-the

garden

'If the sun were shining, I would work in the garden.' However, if both verbs (the verb in the apodosis and the verb in the protasis) are weak verbs, the respective sentence is normally not understood as a conditional sentence. Since the conjunction wenn does not only mean 'if but equally 'when', in sentences with two weak verb forms and the conjunction wenn the verb forms are interpreted as indicative and the vvenw-clause is interpreted not as an i/-clause but as a when-clause. Hence: (15) Wenn

es

iffwhen it

regnete,

arbeitete

rain-PRET:IND/SUBJ:3SG

workPRET:lND/SUBJ:lSG

ich nicht im Garten. I not in-the garden 'Whenever it rained, I did not work in the garden.' The fact that (15) can hardly be understood as a conditional sentence could be interpreted in such a way that forms like regnete or arbeitete are not ambiguous with regard to indicative/subjunctive, but that they are indeed indicative forms. Note, however, that such an analysis would imply that German would be a language with one group of verbs which have a preterite subjunctive form and another group of verbs which lack this category. In any case, in order to be able to understand a sentence like (15) as a conditional sentence, it is necessary that at least one of the two verb forms (either the one in the protasis or the one in the apodosis) provides a form which is formally distinct from the preterite indicative. The form being used in such cases is the würdeconstruction. In (16) three variants are represented which are equally possible. (16a) Wenn es if

regnete,

würde

rain-PRET:iND/SUBJ:3SG

become-PRET:lSG

it

ich nicht I not es (16b) Wenn

im

Garten

arbeiten.

in-the

garden

work-INF

regnen

würde,

arbeitete

rain-lNF

become-PRET:3SG

workPRET:lND/SUBJ:lSG

if

it

ich I

nicht

im

Garten.

not

in-the

garden

325

The subjunctive mood in German and in the Germanic languages

es

regnen

würde,

würde

if

it

rain-INF

become-PRET:3SG

become-PRET:lSG

ich I

nicht

im

not

in-the

(16c) Wenn

Garten

arbeiten.

garden work-lNF 'If it were raining, I would not be working in the garden.'

Starting from cases like (16), the würde-form is expanding and is being used more and more in both the apodosis and the protasis, with weak as well as strong verbs. In everyday speech, sentences like the ones in (16c) and (17) are preferred. (17a) Wenn die if

the

Sonne sun

scheinen

würde,

shine-INF

become-PRET:3SG

würde

ich

Spazierengehen.

become-PRET:lSG

I

walk-go-lNF

'If the sun were shining [now], I would go for a walk.' (17b) Wenn es regnen würde, if

it

rain-lNF

würde

become-PRET:3SG

ich

zu

Hause

bleiben.

become-PRET:lSG I at home stay-lNF 'If it were raining [now], I would stay at home.'

2.3. Reported speech In German, in contrast to the other Germanic languages with the exception of Icelandic, reported speech can be marked by the subjunctive. In written German, especially in newspapers and magazines, the subjunctive 1 is the first choice. The sentences in (18) contain the four tenses available in the subjunctive 1. (18) Er he

sagte,

sie

habe

say-PRET:lND:3SG

she

have-PRES:SUBJ:3SG a

einen

Brief letter

geschrieben

und jetzt

lese

sie

die Zeitung.

write-PART:PERF

and

read-PRES:SUBJ:3SG

she

the newspaper.

now

Wenn sie die Zeitung

gelesen

when she the newspaper

read-PART:PERF have-lNF

haben

werde, becom&PRES:SUBJ:3SG

werde

sie

im

Garten

arbeiten.

become-PRES:SUBJ:3SG

she

in-the

garden

work-INF

'He said that she had written a letter and that she was reading the paper now. When she would have finished the paper, she would work in the garden.' Note that in (18) all verb forms are distinct from the indicative, since they are in the 3rd person singular. As has been shown in 1.1., in the 1st person singular and in the 1st and 3rd person plural the forms of the indicative and of the subjunctive 1 are always identical (with the exception of the verb sein 'be'). In these persons, in order to make clear that indirect speech (and hence the subjunctive) is meant, the subjunctive 1 can be replaced by the subjunctive 2. Future and future perfect sub-

326

ROLF THIEROFF

junctive can be replaced by the vvwri/e-construction which here obviously has future time reference. Consider (19). (19) Er sagte, sie hätten einen Brief he say-PRET:lND:3SG they have-PRET:SUBJ:3PL a letter geschrieben und jetzt läsen sie die Zeitung. write-PART:PERF and now read-PRET:SUBJ:3PL they the newspaper Wenn sie die Zeitung gelesen haben würden, when they the newspaper read-PART:PERF have-INF become-PRET 3PL würden sie im Garten arbeiten. become-PRET:SUBJ:3PL they in-the garden work-INF 'He said that they had written a letter and that they were reading the paper now. When they would have finished the paper, they would work in the garden.' With weak verbs, however, the preterite forms are not unambiguously subjunctive forms either. Compare the sentences in (20a)-(20c). (20a) Er sagte, sie arbeite gerade. he say-PRET:lND:3SG she work-PRES:SUBJ:3SG just 'He said (that) she was working at the moment.' (20b) Er sagte, sie arbeiten gerade. he say-PRET:lND:3SG they work-PRES:lND/SUBJ:3PL just 'He said (that) they were working at the moment.' (20c) Er sagte, sie arbeiteten gerade. he say-PRET:lND:3SG they work-PRET:lND/SUBJ:3PL just 'He said (that) they were working at the moment.' In such cases, the ambiguous forms can again be replaced by the würde-construction, as in (20d). (20d) Er sagte, sie würden gerade he say-PRET:lND:3SG theybecome-PRET:SUBJ:3PL just same as (20 b/c)

arbeiten. work-INF

Note that, in contrast to (19), in (20d) the wwrife-construction does not have future time reference, but the same time reference as a present or a preterite subjunctive. Finally, in everyday speech, the subjunctive 1 of reported speech is often replaced by the subjunctive 2 even if the corresponding subjunctive 1-forms are distinct from the indicative. And, what is more, even the wwn/e-construction is used instead of unambiguous subjunctive 1 forms. So, the sentences in (21) sound quite natural. (21a) Er he

sagte, sie läse jetzt die say-PRET:lND:3SG she read-PRET:SUBJ:3SG now the

Zeitung. newspaper

327

The subjunctive mood in German and in the Germanic languages

(21b) Er sagte, sie würde he say-PRET:lND:3SG she become-PRET:SUBJ:3SG die Zeitung lesen. the newspaper read-INF 'He said that she was reading the paper now.'

jetzt now

2.4. würde + infinitive As we have seen in 2.2. and 2.3., the würc/e-construction is used to replace several tenses in the subjunctive. In the written language the vvwnfe-construction mainly replaces such forms of the subjunctive which are homonymous with the indicative, but in the spoken language the construction is also used in the place of forms which are unambiguously subjunctive. The uses of the wwnfe-construction treated so far are the following: 1.

Hwnfe-construction instead of preterite subjunctive with counter-factual meaning:

(16c) Wenn es regnen würde, if it rain-lNF become-PRET:SUBJ:3SG würde ich nicht im Garten become-PRET:SUBJ:lSG I not in-the garden 'If it were raining, I would not be working in the garden.' 2.

arbeiten. work-INF

würife-construction instead of future subjunctive in reported speech:

(22) Er sagte, wenn sie die Zeitung gelesen he said when they the newspaper read-PART:PERF haben würden, würden have-INF become-PRET:SUBJ:3PL become-PRET:SUBJ:3PL sie im Garten arbeiten. they in-the garden work-INF 'He said that, when they would have finished the paper, they would work in the garden.' 3.

würc/e-construction instead of present/preterite subjunctive in reported speech:

(20d) Er sagte, sie würden gerade he say-PRET:lND:3SG they become-PRET:SUBJ:3PL just 'He said (that) they were working at the moment.'

arbeiten. work-INF

Until this point it appears as if the wwrcfe-construction always had the value of a subjunctive and the assumption that the construction is indeed a future preterite subjunctive seems to be correct. However, there are other uses of the würde-construction where it is questionable whether we are dealing with a subjunctive. One such use is in certain subordinate clauses with future time reference. In (23a) the future tense in the subordinate clause is obligatory to convey future time reference, i.e. the future tense cannot be replaced by the present (as is often the case in German).

328

R O L F THIEROFF

(23a) Er

weiß,

was

he know-PRES:lND:3SG what 'He knows what will happen.'

geschehen

wird.

happen

become-PRES:lND:3SG

If (23a) is transformed into the past, the future has to be replaced by the würdeconstruction: (23b) Er wusste, was he know-PRET:lND:3SG what 'He knew what would happen.'

geschehen happen

würde. become-PRET:3SG

Note that in (23a) the subjunctive is not possible in the subordinate clause (*Er weiß, was geschehen werde is ungrammatical). Since in (23) only the time reference of the sentence is changed (not the modal contents), one would not expect that in (23b) the two verbs are in another mood than in (23a), and indeed the verb of the matrix sentence is in the indicative, as it is in (23a). Consequently it is reasonable to conclude that geschehen würde in (23b) is an indicative exactly as is geschehen wird in (23a) and all other verb forms in (23a) and (23b). And indeed this is precisely the reason for J0RGENSEN (1966) to analyze the Hwrafe-construction in (23b) as an indicative. In other words, if we followed J0RGENSEN, the future preterite forms in (16c), (22) and (20d) would be subjunctives (J0RGENSEN calls them "Conditionals"), whereas the wwri/e-construction in (23b) would be a future preterite indicative. In addition to cases like (23b) with the würrfe-construction in subordinate clauses, würde + infinitive is used to refer to the future in free indirect speech (FID). (24) is quoted by J0RGENSEN (1966: 41). (24) Der Rückzug the retreat

von

hier

from here

war

nicht

schwer;

be-PRET:lND:3SG

not

difficult

er

würde

die Treppe

wieder

hinuntergehen

he

become-PRET:3SG

the staircase

again

down-go-lNF

und

verschwinden; die

Deckung

war

ausgezeichnet;

and

disappear-lNF

cover

be-PRET:lND:3SG

excellent

the

niemand

würde

sich

nobody

become-PRET:3SG himself

fiir for

ihn

interessieren.

him

interest-lNF

'The retreat from here was not difficult; he would go downstairs again and disappear; the cover was excellent; nobody would be interested in him.' In (24) too, all verb forms are in the indicative mood and there is no reason why the clauses referring to the future should be in another mood. In other words, here again würde + infinitive obviously has the function of an indicative. ABRAHAM (1999: 376), who only deals with the wwnie-construction in FED, compares the behaviour of the copula werden with that of the future auxiliary werden in (25). (25a) Er wird klug / Arzt / schlafen / einschlafen he become-PRES:lND:3SG clever physician sleep-lNF fall-asleep-lNF 'He becomes clever/a physician/he will sleep/fall asleep'

329

The subjunctive mood in German and in the Germanic languages

(25b) Er wurde klug / Arzt / *schlafen / *einschlafen he become-PRET: IND: 3 SG clever physician sleep-lNF fall-asleep-lNF 'He became clever/a physician' geworden / (25c) Er ist klug become-PART:PERF he be-PRES:lND:3SG clever *schlafen geworden / Arzt geworden / physician

sleep-lNF

become-PART:PERF

* einschlafen

geworden

fall-asleep-lNF

become-PART:PERF

become-P ART: PERF

'He has become clever/a physician' From the fact that with an adjective or a noun werden is possible in the preterite (or "imperfect", as ABRAHAM has it), whereas with a verb in the infinitive it is not, ABRAHAM concludes: "Während präsentisches werd- den Blick direkt (prädikativ) auf die Entstehensphase lenkt wie in [(25a)], wird durch imperfektisches wurd- die Entstehensphase bloß impliziert; sie bezieht sich auf direkte, das heißt nicht bloß implikative Weise ausschließlich auf den resultierenden Zustand; man vergleiche [(25b)]. [(25b)] bedeutet so viel wie [(25c)]." (ABRAHAM 1 9 9 9 : 3 7 7 ) While it is true that in normal, direct speech Er wurde Arzt and Er ist Arzt geworden have indeed the same meaning (both sentences mean that there is a timespan after the event Arzt werden where "he" actually is [or was] a physician), in FID the sentence Er wurde Arzt does precisely not have this meaning. Incidentally, this can be shown by the very example of FID ABRAHAM quotes at the beginning of his article. However, ABRAHAM could not notice that, because there is one decisive mistake in the text he quotes. The text is the following, from THOMAS MANN'S Tonio Kröger, where Tonio reflects (not "talks", as ABRAHAM says) about the future: (26) Es it

kam der Tag, wo er come-PRET:lND:3SG the day wherehe

berühmt war, famous be-PRET:lND:3SG

wo

alles

gedruckt

where

everything

print-PART:PERF become-PRET:iND:3SG what

wurde,

was

er he

schrieb, und write-PRET:lND:3SG and

dann würde man sehen, then become-PRET:3SG one see-INF

ob

es

nicht

Eindruck

auf

Inge Holm

machen

if

it

not

impression

on

I. H.

make-lNF

würde ...

Es

würde

become-PRET: 3 SG

it

become-PRET: 3 SG no

keinen Eindruck

nein, das

war

es

ja.

no

be-PRET:iND:3SG

it

PTL

that

impression

machen, make-lNF

'The day would come when everything that he was writing would be printed, and then one would see if it would impress Inge Holm ... It would not impress her, no, that was the problem.'

330

ROLF THIEROFF

The mistake in ABRAHAM'S quote is that he has gedruckt war where the original version has gedruckt wurde. And this form has - because it appears in FED - not the same meaning as the present perfect (wo alles gedruckt worden ist): in (26) the preterite cannot be replaced by the present perfect, and the preterite does not only not imply that the day where everything is published has in fact appeared at speech time - on the contrary: it is very well possible that Tonio is erroneous and that none of his works ever will be published. This is precisely the value of preterites with future-in-the-past time reference and future preterites in FDD: they have exactly the same function with respect to the speech time of the protagonist as have present with future time reference and simple future with respect to speech time. Thus, ABRAHAM'S claim that würde + infinitive is a mood in its own right ("selbständiger Modus"; p.382), namely a Potentialis or Irrealis (ibid.), has to be rejected - for the use as an indicative as in FID, but also for the use as subjunctive, as we will see later. In sum, does all that allow the conclusion that there are really two forms würde singen, one indicative and one subjunctive? Before we make a final decision, let us first have a look at the other Germanic languages.

3. The subjunctive in the other Germanic languages In all Germanic languages, the difference between a present or present perfect subjunctive, on the one hand, and the preterite or pluperfect subjunctive on the other hand is not a difference of time reference but a modal difference. Consequently, the terms subjunctive 1 and subjunctive 2 are also appropriate for the other Germanic languages. Most grammars of the respective languages only speak of a present subjunctive and a preterite (or past) subjunctive. This is due to the fact that (with the notable exception of Icelandic) in the other Germanic languages the category subjunctive has almost completely disappeared and only some remnants occur, in general only in the present and in the preterite. Since there is not very much left of the subjunctive in the majority of the Germanic languages, in what follows I will concentrate on the question of how the functions fulfilled by the subjunctive in Icelandic and, to a lesser degree, in German, are rendered in the other Germanic languages. 3.1. Subjunctive 1 3.1.1. Optative meaning Whereas in German the use of the subjunctive 1 in all cases mentioned in section 2.1. is more or less archaic and the subjunctive 1 is practically not used in the spoken register, in Icelandic the subjunctive 1 is a regularly used category. The Icelandic subjunctive 1 has the same functions as in German; it is used as an optative as in (27) and in final subordinate clauses as in (28).

The subjunctive mood in German and in the Germanic

331

languages

(27) Icelandic

[SIGURDSSON1990:322]

Gud blessi

pig\

god

you

bless-PRES:SUBJ:3SG

'God bless you!' (28) Icelandic Svo ad

[KRESS 1982: 248] eg komist

so that I come-PRES:SUBJ:lSG 'To come to the p o i n t . . . '

ad

eftiinu...

to

matter-the

For further functions of the subjunctive in Icelandic see section 4. In modern English, the subjunctive "is generally an optional and stylistically somewhat marked variant of other constructions", as QUIRK et al. (1985: 155) observe, although "it is not so unimportant as is sometimes suggested" (ibid.). That the subjunctive cannot be more than a "marked variant of other constructions" follows already from its morphological make-up. A specific subjunctive form with a morph indicating the subjunctive does not exist any longer in English. Instead of the subjunctive 1, in a few cases the basic form, i.e. a form homonymous with the infinitive is used (cf. KÖNIG 1994: 540). However, in grammars of English this form is generally called "present subjunctive". This "present subjunctive" is morphologically distinguished from the present indicative only in the third person singular and in the verb be. Of the different uses of the subjunctive with an optative meaning, only the subjunctive in that-clauses "introduced by an expression of demand, recommendation, proposal, resolution, intention, etc." (QUIRK et al. 1985: 156) has remained. This use of the subjunctive is called "mandative subjunctive" in grammars of English. The mandative subjunctive can always be replaced either by the indicative or by should + infinitive, or by both, as in (29). (29) English (29a) The employees have demanded that the manager (29b) (29c)

[QUIRK et al. 1985: 157] resign. resigns. should resign.

A second use of the subjunctive 1 is called "formulaic subjunctive" by QUIRK et al. (1985: 157f). This use of the subjunctive 1 is restricted to "certain set expressions chiefly in independent clauses" (ibid.). This use is common to all Germanic languages including German. In the Scandinavian languages and in Dutch, the subjunctive 1 is altogether restricted to such set expressions, i.e. the subjunctive 1 only occurs in examples like (30)-(37). Many of these expressions are identical in several languages. In the following, German examples from section 2.1. are repeated, with their equivalents in some other Germanic languages.

332

(4)

(30)

(31)

(32)

ROLF THIEROFF

Subjunctive 1 in recipes: German Man nehme zwei Eier, ein Pfund Mehl one take-PRES:SUBJ:3SG two eggs one pound flour und ein Pfund Zucker. and one pound sugar 'Take two eggs, one pound of flour and one pound of sugar.' Dutch Men neme one take-PRES:SUBJ:3SG 'Take two eggs.'

[KLOOSTER 2 0 0 1 : 4 4 ]

twee two

eieren. eggs

Danish Man tage tre skefold one take-PRES:SUBJ:3SG three spoonful 'Take three teaspoonfuls of potassium chloride.'

[HABERLAND 1994: 3 3 4 ]

hjortetaksalt. potassium chloride

Subjunctive 1 in formulae with optative meaning: German Es lebe der König! it live-PRES: SUB J: 3 SG the king [KLOOSTER 2 0 0 1 : 4 3 ] Dutch Leve de koningl live-PRES:SUBJ:3SG the king Norwegian [ASKEDAL 1994: 238] Leve Köngen!/ Köngen level king-the live-PRES: SUB J: 3 SG live-PRES: SUB J: 3 SG king-the Swedish [DAHL 1995: 6 4 ] kungen\ Leve live-PRES: SUB J: 3 SG king-the '(Long) live the king!'

(33) Danish Lcenge leve long live-PRES:SUBJ:3SG English

[HABERLAND 1 9 9 4 : 3 3 4 )

dronningen\ queen-the

Long live (PRES SUBJ) the Queen\ (34) German Gott schütze die Königin! god save-PRES:SUBJ:3SG the queen English God save (PRES SUBJ) the Queenl The very fact that almost all authors choose the same example (with variation only with regard to the sex of the sovereign) proves the extent to which this use of

333

The subjunctive mood in German and in the Germanic languages

the subjunctive 1 has become formulaic. Even more formulaic is perhaps the expression in (6) which, again, is the same in German and Dutch and in the Scandinavian languages. In Dutch, this formula can even be written in one word (godzijdank; KLOOSTER 2001: 115) (6c) German Gott sei

Dankl

god

thanks

be-PRES:SUBJ:3SG

(35) Dutch God zij god

dank

be-PRES:SUBJ:3SG

thanks

Danish, Swedish Gud ske lov! god happen-PRES:SUBJ:3SG thanks 'Thank God!' Subjunctive 1 in other formulae: (36) German Komme, was da come-PRES:SUBJ:3SG

what

wolle,

there

will-PRES:SUBJ:3SG

wir

werden

mit

unserem

Plan

weitermachen.

we

become-PRES:lND:lPL

with

our

plan

further-do-lNF

English [QUIRK et al. 1985: 157] Come (PRES SUBJ) what may, we will go ahead with our plan.

(37) German Sei be-PRES:SUBJ:3SG

dem,

wie

ihm

wolle,

that-DAT:SG

how

it-DAT:SG

will-PRES:SUBJ:3SG

wir

haben

nichts

zu

verlieren.

we

have-PRES:iND:lPL

nothing

to

lose-lNF

English

[QUIRK et al. 1985: 158]

Be (PRES SUBJ) that as it may, we have nothing to lose. 3.1.2. Constructions replacing the subjunctive 1 Since the subjunctive 1 is no longer a productive means of expression in the languages mentioned, it had to be replaced by other means of expression. These are, again, the same in all Germanic languages. For most uses of the subjunctive 1 the modal verb shall and its equivalents in the other Germanic languages are used. An example from English was given in (29), where the subjunctive 1 resign can be replaced by the past tense of the modal shall, i.e. should + the infinitive. In German it is the modal sollen, either in the present or in the preterite, which is used instead of the subjunctive 1. For the example in (3), gleaned from FLÄMIG (1991: 411), the author himself "translates" his example with the more natural sollen + infinitive construction.

334 (3)

R O L F THIEROFF

German

[FLÄMIG 1991: 411 ]

Wer schadhaftes

Gerät

empfangen

hat,

who

equipment

receive-PART:PERF

have-PRES:lND:3SG

defective

(3a) der he (3b) der

gebe give-PRES:SUBJ:3SG soll

he

shall-PRES:lND:3SG

es it es

zurück. back zurückgeben.

it

back-give-INF

'Who has got defective equipment should give it back' In the English translation the same construction as the German construction with sollen 'shall' is used, however with shall in the future preterite. The modal shall!sollen and its equivalents also replace the subjunctive 1 in many other constructions, as for example in the wish that the king or queen shall live long: (32') German Der König soll leben\ the king

shall-PRES:lND:3SG

Iive-INF

English The King shall live! A second modal which is used instead of the subjunctive 1 is English may, German mögen, as in the following examples: (36') German Es mag kommen, was da kommen mag, it may-PRES:IND:3SG

wir

come-INF

werden

mit

what there come-INF

unserem

we become-PRES:lND:lPL with our English

may-PRES:lND:3SG

Plan

weitermachen.

plan

further-do-lNF [QUIRK et al. 1985: 157]

What may come may come, we will go ahead with our plan. (37') German Es mag

sein,

wie

es

will,

be-lNF

how

it

will-PRES:lND:3SG

wir haben

nichts

zu

we

nothing to

it

may-PRES:lND:3SG have-PRES:lND:lPL

verlieren. lose-lNF

English

[QUIRK et al. 1985: 158]

That may be as it may, we have nothing to lose. The subjunctive 1 used in mathematical texts as in (5) can be replaced by sollen + infinitive in German, too. In other languages, as in English, let and its equivalents are used instead. (5) (5')

German ABC sei

ein Dreieck.

ABC

a

be-PRES:SUBJ:3SG

triangle

German ABC soll

ein Dreieck

sein.

ABC

a

be-INF

shall-PRES:lND:3SG

triangle

335

The subjunctive mood in German and in the Germanic languages

English Let ABC be a triangle. Finally, instead of the subjunctive in recipes the imperative form is normally used. Another choice is the infinitive. Thus, (4) can be expressed as (4') or as (4") in German. (4') German Nehmen Sie zwei Eier, ein Pfund Mehl take-IMP

und

and

you: POLITE

ein

one

Pfund

two

eggs

pound

pound

flour

sugar

(4") German Zwei Eier, ein Pfund Mehl und two

one

Zucker.

eggs one pound flour and

ein

one

Pfund Zucker pound sugar

nehmen.

take-INF

3.2. Subjunctive 2 3.2.1. Icelandic and German In section 2.2. it was shown that in German counterfactuality normally has to be indicated by subjunctive 2 forms. In Icelandic, the subjunctive 2 is used in exactly the same way. Compare the following examples and their German translations: (38)

Icelandic

[KRESS 1982: 237]

Vieri

eg

kominn

til

tunglsinsl

be-PRET:SUBJ:lSG German

I

come-PART:PERF:ΜASC

to

moon-the:GEN:SG

Wäre

ich doch auf

dem

Mond\

be-PRET:SUBJ:LSG

I

the-DAT:SG

moon

PTL

on

'If I only were on the moon!' (39) Icelandic [KRESS 1982: 237] Hefdir pit bara komid fyrr\ have-PRET:SUBJ:2SG you only come-PART:PERF earlier German Wärst du bloß früher gekommen\ be-PRET:SUBJ:2SG you only earlier 'If you only had come earlier!'

come-PART:PERF

In counterfactual conditional sentences, in both Icelandic and German the subjunctive 2 is possible in both the apodosis and the protasis. Compare: (40a) Icelandic Ef ig if I fieri

[SlGURBSSON 1990: 322] Vieri be-PRET:SUBJ:lSG ig til

go-PRET:SUBJ:lSG

I

to

rikur rich tunglsins.

moon-the:GEN:SG

336

ROLF THIEROFF

German Wenn ich reich wäre, if I rich be-PRET:SUBJ:lSG flöge ich zum Mond. fly-PRET: SUBJ: 1SGI to-the-DAT: SG moon 'If I were rich, I would go to the moon.' Instead of the subjunctive 2, in both languages in the apodosis the future preterite is equally possible: (40b) Icelandic Ef eg vteri rikur if I be-PRET:SUBJ:lSG rich skuldi eg fara til shall-PRET:lSG I go-lNF to German Wenn ich reich wäre, if I rich be-PRET:SUBJ:lSG würde ich zum become-PRET:lSG I to-the-DAT: SG 'If I were rich, I would go to moon.'

tunglsins. moon-the:GEN:SG

Mond moon

fliegen. fly-lNF

Finally, in German, but not in Icelandic, the future preterite is also possible in the protasis, though normally not with the verb sein 'be': (40c) German Wenn mir jemand Geld geben würde, if me someone money give:lNF become-PRET:3SG würde ich zum Mond fliegen. become-PRET:LSG I to-the-DAT:SG moon fly-lNF 'If someone gave me money, I would go to the moon.' 3.2.2. Swedish The only other Germanic language which still has distinct preterite subjunctive forms of other verbs than 'be' is Swedish4, where the preterite subjunctive "differs from the Preterit Indicative only for strong and (some) irregular verbs" in that it "takes the ending -e, together with idiosyncratic ablaut alternations for the stem vowel, as in drucke, the Preterit Subjunctive of dricka 'drink'" (DAHL 1995: 65; cf. also A N D E R S S O N 1994: 284). However, the subjunctive forms of other verbs than 'be' are very rarely used in modern Swedish and "sound a bit stilted even in the written language" (DAHL 1995: 65). Thus, the only subjunctive 2 form which "is still relatively frequent in daily speech" (ibid.) is the preterite subjunctive of

4

I am grateful to OSTEN DAHL and LARS JOHANSON for helping me to understand the Swedish system and patiently answering the numerous questions I had concerning Swedish counterfactuals.

337

The subjunctive mood in German and in the Germanic languages

the verb vara 'be', vore, as in (41), although even this use is now going back and the subjunctive is being replaced by the indicative. (41)

Swedish

[DAHL 1995: 65]

Om han vore if

he

hemma, skulle

be-PRET:SUBJ home

ljuset

vara

tänt.

shall-PRET:lND light-the be:lNF

'If he were at home, the light would be on.' Om han var hemma, skulle if he be-PRET:lND home the same

ljuset

vara

lit

tänt.

shall-PRET:lND light-the be:lNF

lit

With other verbs than vara 'be', the preterite indicative is the normal choice to express counterfactual statements. The preterite indicative is possible in contraryto-fact wishes as in (42), in counterfactual conditionals as in (43), and in unreal comparative sentences as in (44). (42)

Swedish

Om

[WESSEN 1968: 100]

han

ändä

gick

snartl

if he PTL go-PRET:iND 'If he would only leave soon!' (43)

soon

Swedish

[THORELL 1973: 132]

Om han nu

stod

if

stand-PRET: IND still

he

now

stilla,

fotografera

honom.

photograph-lNF

him

skulle

jag

shall-PRET:lND

I

'If he were standing still now, I would photograph him.' (44)

Swedish

[RITTE 1986: 68]

Han

nickade

he

nod-PRET:iND

vad

det

what it

som

om

as

if

han

begrep

he

understand-PRET:iND

handlade

om.

deal-PRET:lND

about

'He nodded as if he understood what it was about.' As in German, but in contradistinction to all other Germanic languages, in Swedish the future preterite is also possible in /^clauses. There are however restrictions to this use unknown to German, since skulle + infinitive can also mean 'in case it should V'. Compare the Swedish sentences in (45) with their German equivalents and the translations into English. (45) Swedish Om det regnade, skulle jag stanna hemma. if

it

rain-PRET:lND

shall-PRET:lND

I

stay:lNF

home

German Wenn es

regnete,

würde

ich

if

it

rain-PRET:rND/suBJ:3SG

become-PRET:lSG

I

zu

Hause

bleiben.

at

home

stay:lNF

'If it were raining, I would stay at home.'

338

ROLF THIEROFF

Swedish Om det skulle regna, skulle jag stanna hemma. if it shall-PRET:lND rain-INF shall-PRET:lND I stay:lNF home 'If it were raining, I would stay at home'/'In case it should rain, I would stay at home' German Wenn es regnen würde, würde ich if it rain-INF become-PRET:3SG become-PRET:lSG I zu Hause bleiben. at home stay:lNF 'If it were raining, I would stay at home.' In addition to the preterite subjunctive (virtually only with vara 'be'), the preterite indicative and the future preterite, counterfactuality can also be expressed by the pluperfect indicative (see section 5.). 3.2.3. English, Dutch, Danish, and Norwegian In English, only the verb be has retained a preterite subjunctive form which is distinct from the indicative only in the first and third person singular. But even this subjunctive form, were, is not obligatorily used and the "indicative form was is substituted in less formal style" (QUIRK et al. 1985: 158). Hence: (46) English If I were (PRET SUBJ) / was (PRETIND)rich,I would buy you anything you wanted. With other verbs than be, counterfactuality is expressed by the preterite indicative (present time reference) or by the pluperfect indicative (past time reference): (47) English If he came (PRET IND), I would stay at home. (48) English If he had come (PLPF IND), I would have stayed at home. In Dutch, the preterite subjunctive of the verb zijn 'be' has only survived in fixed expressions like (49) and in unreal comparative clauses introduced by als 'as i f , whereas in unreal comparative clauses introduced by alsofias if the preterite indicative is used. (49) Dutch als het ware as-if it be: PRET: SUBJ 'As if it were so'

[DE SCHUTTER 1994: 455]

(50) Dutch als ware het een verplichting as-if be: PRET:SUBJ it a commitment alsof het een verplichting was as-if it a commitment be-PRET:lND 'As if it were a commitment'

[KLOOSTER 2001: 115]

339

The subjunctive mood in German and in the Germanic languages

In all other cases and with other verbs than zijn 'be' preterite and pluperfect indicative are the only choice to express counterfactuality: (51)

Dutch

[JANSSEN 1994: 111 ]

Als

je

echt

if

you really

zou

van

mij

hield,

of

me

hold-PRET:lND

je

dat

niet

zeggen.

shall-PRET:iND you that not say-iNF 'If you really loved me, you would not say that.'

Finally, in Danish and Norwegian, no traces of a preterite subjunctive have remained, not even of the verb 'be'. In these languages counterfactuality is expressed by the preterite (present time reference) and the pluperfect (past time reference). (52) Danish Hvis jeg if

I

[HABERLAND 1994: 333]

vandt

i

lotto,

win-PRET:LND

in

pools

ville

jeg

kebe

dig

en

Mazda.

will-PRET:IND

I

buy-INF

you

a

M.

'If I won in the pools, I would buy you a Mazda.' (53)

Norwegian

[ASKEDAL 1994: 245]

Hadde

jeg vinger,

skulle

jeg

fly.

have-PRET:lND

I

shall-PRET:lND

I

fly-INF

wings

'If I had wings, I would fly.' (54)

Norwegian

[ASKEDAL 1994: 245]

Hvis jeg hadde if I have-PRET:IND skulle jeg

hatt ha

vinger, wings fleyet.

shall-PRET:LND

have-LNF

fly-PART:PERF

I

have-PART: PERF

'If I had had wings, I would have flown.' In addition, in both languages the pluperfect is also used with non-past time reference (see section 5). To summarize, there is a development from the use of the preterite (or pluperfect) subjunctive to the use of the preterite (or pluperfect) indicative in counterfactuals in almost all Germanic languages in different degrees. In Icelandic, the subjunctive is the only choice. In German, the subjunctive is the preferred mood. In Swedish, subjunctive forms of strong verbs still exist, but in non-formal registers the subjunctive is virtually restricted to the verb 'be'. In English, only be has a preterite subjunctive, but the indicative of be is used too. In Dutch, the preterite subjunctive of 'be' is restricted to certain fixed expressions. Finally, in Danish and Norwegian, no past subjunctive forms have survived at all. This leads to the picture of decreasing use of the preterite subjunctive in Germanic languages in Figure 1.

340

Full-fledged subjunctive 2 Icelandic > German > Swedish > English > Dutch

R O L F THIEROFF

No subjunctive 2 > Danish Norwegian

Fig. 1: Degrees of loss of subjunctive 2

3.3. Future preterite or "conditional"? 3.3.1. Preterite, future, and future preterite In descriptions of the use of the past indicative in counterfactuals, it is very often maintained that the preterite is used instead of the subjunctive. A typical statement is the following, from KÖNIG ( 1 9 9 4 : 5 4 0 ) : "Where other Germanic languages use a specific subjunctive form, English uses the past tense". Though KÖNIG does not explicitly say so, he gives the impression that English uses the past tense in a context where other Germanic languages do not use it, but use the subjunctive instead. However, it must be stressed that in reality the Germanic (and not only the Germanic) languages do not differ in whether they use the past tense in counterfactuals or not. They do differ in whether they use the subjunctive or the indicative, but irrespective of which mood is used, counterfactuals are always in the past tense. This is not a specific feature of the Germanic languages, but it holds for all or almost all European languages.5 At least in all Germanic, all Romance and all Slavic languages, with counterfactual statements the finite verb form must be in a past tense. The only difference we find in the languages mentioned is whether the verb form is in the past indicative or in the past subjunctive. In addition to counterfactual conditionals, we find past tenses with non-past time reference in contrary-to-fact-wishes {If only they were coming!), in "attenuative uses" (/ would like to talk to you), in polite advices (Dutch Maar ik vertrok (PRET IND) morgen [als ikjou was]! 'but I would leave tomorrow [if I were you]), "affective uses" (German Da wären (PRET SUBJ) wir! 'here we are', "preludic uses" (Dutch Dan was (PRET SUBJ) jij de vader en ik was de moeder, KLOOSTER 2 0 0 1 : 4 3 , 'then you are the father and I am the mother') and others (see THIEROFF 1 9 9 9 : 1 4 7 - 1 5 4 for further examples in Germanic and other European languages). Thus, in the languages of Europe past tenses are quite regularly used with non-past time reference in so many contexts that it is obviously not correct to maintain that a past tense necessarily has past time reference. This is generally not true for past subjunctives, but it is not true for past indicatives either. Thus past tenses in general have either past time reference or non-past time reference with a modal meaning, but in some languages (such as Icelandic and German) the subjunctive mood

5

It is debatable for Finnish, where in the traditional grammar a mood with the name "Conditional" rather than a preterite subjunctive is assumed (see THIEROFF 1 9 9 4 for a detailed discussion). - The Finnish so called Conditional has (at least morphologically) nothing to do with the "Conditionals" of the Germanic languages discussed in 3.3.2.

The subjunctive mood in German and in the Germanic

languages

341

is necessary to indicate that a given past tense has non-past time reference (and hence a modal meaning). The fact that past tenses are regularly used with non-past time reference and modal meaning in virtually all European languages does not seem to be very well known. It is interesting to see that in describing individual languages authors again and again note this fact about the specific language they are dealing with. But there seems to be little knowledge of the fact that this is not a feature of some specific language only. It is also important to note that even if such non-past time uses of a past tense are realized, no consequences as to the meaning of the respective past tense are drawn. Even if non-past time uses of a past tense are mentioned, it is in general maintained that this past tense has past time meaning, although past time reference is only one of the uses of the respective tense. With respect to future tenses, the situation is a bit different. Again, the modal uses of future tenses are in most grammars described as if they were a specific feature of a specific language, whereas in reality future tenses, like past tenses, have modal uses in all European languages too. However, in contrast to the treatment of past tenses, with future tenses many authors draw the conclusion that future time reference is not the basic meaning of the future tense, but that it has a less concrete meaning covering future time reference and, for example, supposition or prediction (for examples in the Germanic and in other European languages see THIEROFF (forthcoming)). What all that boils down to is that the different traditions in describing past and future tenses are by no means justified. On the contrary: for both past and future tenses it is true 1) that they have both a temporal and at least one (in most languages several) modal functions and 2) that certain modal functions of both tenses are common to virtually all European languages in very much the same way as are their temporal functions. If a verb form containing a past (for the Germanic languages: a preterite) morpheme can have either a temporal or a modal meaning and a verb form containing a future morpheme can have either a temporal or a modal meaning, too, then one would expect that a verb form containing a past and a future morpheme should have temporal and modal meanings too. This is exactly what we find. Consider: Preterite temporal: (55) He came yesterday. modal: (56) If he came now, (I would stay at home). Future temporal: (57) He will come tomorrow. modal: (58) That will be the postman.

342

ROLF THIEROFF

Future Preterite temporal: (59) (/ did not expect) that he would win the race. modal: (60) (If he came now,) I would stay at home. In most grammars, in describing forms like sang, will sing and would sing, for the preterite only instances like (55) are considered; for the future, both uses (in (57) and (58)) are considered; and for constructions like would sing, exclusively instances like (60) are taken into account with the result that the preterite is regarded as a "good" tense, as for the future there is much debate on whether it is a tense or not, and the future preterite is regarded as a non-indicative mood, called "Conditional". 3.3.2. Temporal and modal functions of the future preterite Only occasionally has it been noticed that would sing does not necessarily have conditional meaning. To my knowledge the first scholar to remark that this construction can function as a tense has been REICHENBACH. With respect to (59), which is from REICHENBACH (1947: 297), the author remarks that "grammar does not officially recognize [the form 'he would do'] as a tense" (1947: 298) and he adds in a footnote: "[The form 'he would do'] is sometimes classified as a tense of the conditional mood, corresponding to the French conditional. In the examples considered [...] however, it is not a conditional but a tense in the indicative mood" (1947: 298; boldface mine). What REICHENBACH seems to be suggesting here is that in examples like (59) we are dealing with a tense in the indicative mood, whereas in an example like (60) we are dealing with the "Conditional mood" - an analysis which comes close to the analysis of German würde + infinitive by J0RGENSEN (1966; cf. sections 1.4. and 2.4.). However, the examples in (55) to (60) clearly show that the postulation of two homonymous forms would do is neither necessary nor consistent. For past tense forms, to the best of my knowledge nobody ever has suggested to analyze them as two homonymous forms, one being a tense and the other being an extra mood, although, as we have seen, past tenses have temporal and modal uses in exactly the same way as have future preterites; and likewise, for futures such a homonymy is normally not assumed either.6 Consequently, there is no need to assume it for the future preterite either. Morphologically, would sing is a combination of the future marker (will) with the preterite marker (would = preterite of will), i.e. a combination of two tense markers which should yield another tense (in exactly the same sense as for example the combination of a perfect marker with a preterite marker yields a past perfect and not a category of a third kind).

6

In fact, for German it has been suggested by EISENBERG (1989: 124f.) and also in THIEROFF (1992: 125). It follows however from the above that this view cannot be maintained.

The subjunctive mood in German and in the Germanic languages

343

With respect to the semantics it is true that if a morph A with a meaning X and a morph Β with a meaning Y are combined to a unit AB, this unit does not necessarily have the meanings X+Y, but that it is possible that AB has a meaning Ζ instead (cf. THIEROFF 2000: 270-274). However, for the future preterite this is obviously not true. In (59), would win has indeed nothing but the temporal meanings of the future marker plus that of the past marker and in (60) the modal meaning of the past marker is predominant. All these considerations allow only one conclusion: In the Germanic languages there is no such thing as a conditional mood. With the exception of Icelandic and German, the Germanic languages considered in the present article have lost their only non-indicative mood, the subjunctive, and all functions of the past subjunctive have been taken over by the past (indicative) and the future past (indicative). Since there is no (or virtually no) subjunctive left, it is not even necessary to speak of a past or future in the indicative mood, given that these forms are not in opposition to another mood. What has been shown here with the example of English is equally true for the other Germanic languages without subjunctive. Especially in conditional sentences, as a rule, the apodosis is in the preterite indicative, as in the following sentences from section 3.2. The only difference between the respective languages is that the auxiliary is 'shall' in Dutch and Swedish7, both 'shall' and 'will' in Norwegian, and only 'will' in English and Danish. (43) Swedish [THORELL 1973: 132] Om han nu stod stilla, skulle jag fotografera honom. if he now stand-PRET:lND still shall-PRETl photograph-lNF him 'If he were standing still now, I would photograph him.' (46) English If I were (PRET SUBJ) / was (PRET IND) rich, I would buy (FUT PRET) you anything you wanted. (51) Dutch [JANSSEN 1994: 111 ] Als je echt van mij hield, zou je dat nietzeggen. if you really of me hold-PRET:lND shall-PRET you that not say-lNF 'If you really loved me, you would not say that.' (52)

Danish

[HABERLAND 1994: 333]

Hvis jeg vandt i lotto, if I win-PRET:lND in pools ville jeg kebe dig en Mazda. will-PRET I buy-lNF you a M. 'If I won in the pools, I would buy you a Mazda'

7

In contradistinction to the other languages, in Swedish the bare preterite is also (marginally) possible in the main sentence, i.e. besides (i) Om det regnade skulle jag stanna hemma it is possible to say (ii) Om det regnade stannadejag hemma. (i) sounds more natural though (OSTEN DAHL, p.c.).

344 (53)

ROLF THIEROFF

Norwegian

[ASKEDAL 1994: 245]

Hadde

jeg vinger,

have-PRET:lND I wings ' I f I had wings, I would fly.'

skulle

jeg

fly.

shall-PRET

I

fly-INF

Likewise, in Icelandic and German a future preterite, built of the future auxiliary in the past and the infinitive, exists. However, in both languages this future auxiliary is not only in the past but, in addition, in the subjunctive. An equivalent in the indicative is lacking in both languages. The explanation for this lacuna is not straightforward. 3.3.3. The future preterite in German and Icelandic First of all, in Icelandic and in German the structure of a counterfactual conditional sentence with the preterite in the protasis and the future preterite in the apodosis may be the same as in the other languages8. There is, however, one important difference: in contradistinction to English, Dutch and the Scandinavian languages, in both Icelandic and German the preterite of the protasis is in the subjunctive. Compare once again: (40b) Icelandic Ef eg voeri if

I

rikur

be-PRET:SUBJ:lSG

rich

fara

til

tunglsins.

go-lNF

to

moon-the:GEN:SG

skuldi

ig

shall-PRET: lSG

I

German Wenn ich reich

wäre,

würde

if

I

be-PRET:SUBJ:lSG

become-PRET:lSG

ich

zum

Mond

fliegen.

I

to-the-DAT:SG

moon

fly-INF

rich

'If I were rich, I would go to moon.' In both languages, the future preterite is, as will be recalled, morphologically a subjunctive, but in both languages an analogous indicative does not exist, even though we are able to tell what it would look like if it existed (*wurde fliegen 'become-PRET:IND fly-INF' in German). If we only regard instances like (40b), it is highly plausible that the future preterite is in the subjunctive. Just as in the other languages, where the preterite is in the indicative and hence the future preterite is in the indicative too, in Icelandic and German, too, both tenses should be in the same morphological mood. However, as we have seen in section 2.4. with regard to German (on which language I will concentrate in what follows), the same form is also used in the context of indicatives, as in, recall, (24).

8

In addition, in these two languages the apodosis can also be in the subjunctive 2 (see 3.2.1.).

The subjunctive mood in German and in the Germanic

languages

345

(24) Der Rückzug von hier war nicht schwer; the retreat from here be-PRET:lND:3SG not difficult er würde die Treppe wieder hinuntergehen und he become-PRET:3SG the staircase again down-go-iNF and verschwinden; die Deckung war ausgezeichnet; disappear-iNF the cover be-PRET:iND:3SG excellent niemand würde sich für ihn interessieren. nobody become-PRET:3SG himself for him interest-lNF 'The retreat from here was not difficult; he would go downstairs again and disappear; the cover was excellent; nobody would be interested in him.' In 2.4. we have seen that in (24) the würcfe-construction has the function of an indicative and the question remains why then it is not expressed by an indicative form. It has been a well known fact for a long time that languages tend to avoid too many marked categories in one single item. Well known examples from the noun phrase are that in many languages the third person is marked for gender in the unmarked number, the singular (cf. English he/she/it), whereas in the marked number, the plural, gender is not marked (English they, without gender distinction). In languages inflecting for case, the number of formal case distinctions is very often greater in the singular than in the plural, and it is greater in the unmarked gender than in the marked gender(s). For example, in German the definite article distinguishes all four cases formally in the unmarked masculine gender {der, den, dem, des), but only three cases in the more marked neuter {das, dem, des; with nominative and accusative having the same form) and no more than two cases in the most marked feminine {die, der; with, in addition, only one form for dative and genitive). In other words, marking of one kind of (inflectional) category very often implies non-marking of another kind of (inflectional) category. This same principle also holds for the inflection of the verb. In the introduction to this article, with respect to the morphological marking of person, number and tense we have seen that in Dutch in the marked tense, the past, only number is marked, but not person; and in the unmarked present, person is marked only in the unmarked number, the singular, not in the marked number, the plural. Similarly, in English, marking of the third person is restricted to the unmarked tense and the unmarked number at the same time (the present singular). With respect to tense and mood, this principle can be observed to hold already in Latin. In this language, in the unmarked mood, the indicative, six different tenses can be distinguished, but in the marked mood, the subjunctive, there are only four tenses since there is no future and future perfect of the subjunctive. In Icelandic and German, two restrictions of the Latin system do not hold: in contrast to Latin, the future can combine with the past and it can combine with the subjunctive. What is however not possible is a combination of the future with both the past and the subjunctive - that would be too many marked categories in one single verb form. That forms like would sing have been interpreted - irrespective of their morphological make-up - as a conditional mood can probably be explained by the fact

346

ROLF THIEROFF

that their use in conditionals is more frequent than their use in free indirect speech, which is a relatively recent literary form. This means that in Icelandic and German the future preterite with the function of a subjunctive has no doubt been by far more frequent than its use with the function of an indicative, and this is certainly one of the reasons why the subjunctive form won over the indicative form. Nevertheless it remains true that we are dealing with one form and one form only, which, in contrast to all other tense-mood combinations, has indeed both functions of the indicative and functions of the subjunctive. However, since the marking of subjunctive and future and preterite would be too heavy a functional load, this form is not marked for mood, i.e. it does not belong to one of the two mood categories and that is why it can be used with the function of both.9

4. Reported speech in Icelandic and German The uses of the subjunctive forms (subjunctive 1 as well as subjunctive 2) treated in section 3 are roughly the same in all Germanic languages, with major differences only with regard to the use of the subjunctive in certain cases in Icelandic and German and with some differences concerning the degree of the decline of the subjunctive. Similarly, the means which replace the subjunctive forms are more or less the same in all languages treated in this paper. However, there are two features which are unique to only a few languages. These are the use of the pluperfect to indicate counterfactuality in Norwegian, Danish and Swedish (section 5) and the use of the subjunctive in reported speech in Icelandic and German.

4.1. Subjunctive with (non-)factive verbs There are two Germanic languages where the subjunctive has a special function unknown to the remaining Germanic languages, namely Icelandic and German, which use the subjunctive to signal reported speech. In both Icelandic and German this use of the subjunctive is correlated with (non-)factivity of the matrix verb; however there are some important differences between the two languages. In Icelandic, the subjunctive is the only choice after non-factive verbs, as is exemplified in (61) and (62). In these sentences, the indicative at the place of the subjunctive would be ungrammatical. (61) Icelandic [THRAINSSON 1990: 291] Jon segir ad tveir plus tveir sett fiorir. John say-PRES:lND:3SG that two plus two be-PRES:SUBJ:3PL four 'John says that two plus two is four.'

9

Another more modern fashion to say the same would be to say that the future preterite is underspecified with regard to mood.

347

The subjunctive mood in German and in the Germanic languages (62)

Icelandic

[SlGURBSSON 1990: 323]

Pad er

mögulegt

αό

tveir

plus

tveir

it

possible

that

two

plus

two

be-PRES:lND:3SG

siu

fimm.

be-PRES:SUBJ:3PL

five

'It is possible that two plus two is five.' In contrast, with factive verbs, only the indicative is allowed in the subordinate sentence. In (63) and (64), the subjunctive is ungrammatical: (63)

Icelandic

[THRÄINSSON 1990: 290]

Jon veit ad tveir plus tveir eru jjorir. John know-PRES:iND:3SG that two plus two be-PRES:iND:3PL four 'John knows that two plus two is four.' (64)

Icelandic

[SlGURBSSON 1990: 323]

Pad

er

augljost

ad

tveir plus

tveir

it

be-PRES:lND:3SG

obvious

that

two

two

eru

fimm.

be-PRES:lND:3PL

five

plus

'It is obvious that two plus two is five.' Besides factive verbs, which take obligatorily the indicative in the subordinate clause and non-factive verbs, which take obligatorily the subjunctive in the subordinate clause, there is a third class of verbs which have a factive as well as a nonfactive reading. With these verbs, in Icelandic, the indicative delivers a factive reading, whereas the subjunctive delivers a non-factive reading. Compare the sentences in (65) and (66). (65) Icelandic

[THRÄINSSON 1990: 292]

Jon las pad ί John read-PRET:lND:3SG that in hafdi komid have-PRET:iND:3SG

bladinu ad newspaper-the that heim.

come-P ART: PERF

Maria Mary

home

'John read in the newspaper that Mary had come home' (66) Icelandic [THRAINSSON 1990: 292] Jon las pad i bladinu ad John read-PRET:lND:3SG that in newspaper-the that Maria hefdi komid heim. Mary have-PRET:SUBJ:3SG come-P ART: PERF home 'John read in the newspaper that Mary had come home.' Whereas in (65) the speaker presents the contents of the subordinate clause as a fact (i.e. the speaker has no doubt that Mary has come home), the subjunctive in (66) "reports what John read and the speaker does not commit himself or herself to the truth of that material" (THRAINSSON 1990: 292). Consequently, adding 'but she did not come home' to the indicative version in (65) produces a contradiction, as in (67), whereas it does not in the case of the subjunctive version in (66), as in (68).

348 (67)

ROLF THIEROFF

Icelandic

[THRÄINSSON 1990: 292]

*Jon las pad i bladinu aö Maria John read-PRET: IND: 3 SG that in newspaper-the that Mary hafdi komid heim en hm have-PRET:lND:3SG

come-PART:PERF

kom

ekki

heim.

come-PRET:lND:3SG

not

home

home

but she

'John read in the newspaper that Mary had come home, but she hadn't come home.' (68)

Icelandic

[THRÄINSSON 1990: 292]

Jon las pad John read-PRET:IND:3SG that hefdi komid have-PRET:SUBJ:3SG

i bladinu aö in newspaper-the that heim en

come-PART:PERF

kom

ekki

heim.

come-PRET:lND:3SG

not

home

home

Maria Mary Mn

but

she

'John read in the newspaper that Mary had come home, but she hadn't come home' Thus, whereas with factive and non-factive verbs the speaker is not free to choose between the indicative and the subjunctive in the complement clause, with ambiguous verbs the indicative delivers the factive, the subjunctive the non-factive reading of such verbs. In German, like in Icelandic, the three groups of factive, non-factive and ambiguous verbs can be distinguished. However, the use of the two moods is organized less clearly. First of all, as in Icelandic, the subjunctive is not possible in subordinate sentences governed by a factive verb. Hence, the subjunctive in (69) is ungrammatical. (69)

German

Karl

[EISENBERG 1999: 117]

versteht/

vergisst/

weiß,

Charles understand-PRES:lND:3SG forget-PRES:lND:3SG know-PRES:lND:3SG

dass

Egon

bleiben

will /

*wolle

that

Egon

stay-lNF

want-PRES:lND:3SG

want-PRES:SUBJ:3SG

'Charles understands/forgets/knows that Egon wants to stay.' However, contrary to Icelandic, the indicative is also possible with non-factive verbs, as in (70): (70)

German

[ElSENBERG 1999: 117]

Karl

meint /

hofft /

glaubt,

Charles

think-PRES:lND:3SG

hope-PRES:lND:3SG

believe-PRES:lND:3SG

dass Egon bleiben will that Egon stay-lNF want-PRES:lND:3SG 'Charles thinks/hopes/believes that Egon wants to stay'

349

The subjunctive mood in German and in the Germanic languages

Finally, the indicative is equally possible with ambiguous verbs. In contrast to Icelandic, however, in German the indicative cannot disambiguate between factive and non-factive reading, i.e. with the verbs in (71) the statements remain ambiguous between the two readings: (71)

German

[ElSENBERG 1999: 117]

Karl

berichtet!

erzählt I

sagt,

Charles

report-PRES:lND:3SG

tell-PRES:lND:3SG

say-PRES:lND:3SG

dass Egon bleiben will that Egon stay-iNF want-PRES:iND:3SG 'Charles reports/tells [us]/says that Egon wants to stay' However, if with ambiguous verbs the subjunctive is chosen, then only the nonfactive reading is possible, as in (72): (72)

German

[ElSENBERG 1999: 117]

Karl

berichtet /

erzählt /

sagt,

Charles

report-PRES:lND:3SG

tell-PRES:lND:3SG

say-PRES:lND:3SG

dass Egon bleiben wolle that Egon stay-iNF want-PRES:SUBJ:3SG 'Charles reports/tells [us]/says that Egon wants to stay' Thus, as in Icelandic, the subjunctive in German serves to signal non-factivity, but, in contradistinction to Icelandic, in German the indicative can be used with nonfactive verbs too. 4.2. Consecution temporum in reported speech Another difference between Icelandic and German concerns the use of the tenses of the subjunctive mood in reported speech. In Icelandic, there are certain rules of consecutio temporum which are not valid for German. SlGURBSSON explains the situation in Icelandic as follows: "[T]he temporal reference point of embedded subjunctives is basically different from that of indicatives: it is the time of saying, thinking, believing, and so on that is defined in an indicative matrix clause of the embedded subjunctive. If the indicative matrix clause is in the present, the temporal reference point of the embedded subjunctive is normally also (derived) now, but if the matrix clause is in the preterite, it is (derived) then." (SlGURBSSON 1990: 329)

This is exemplified with the sentences in (73) and (74), with both the matrix and the subordinate clause in the present or in the preterite respectively: (73)

Icelandic

Maria segir

[SlGURBSSON 1990: 329]

ad

Mary say-PRES:lND:3SG that 'Mary says that John will come.'

Jon

komL

John come-PRES:SUBJ:3SG

350

ROLF THIEROFF

(73') Icelandic

[SlGURBSSON 1990: 329]

Maria sagöi ad Jon kcemL Mary say-PRET:lND:3SG that John come-PRET:SUBJ:3SG 'Mary said that John would come.' From data like the ones presented in (73) to (74), SlGURBSSON concludes: "[S]ubjunctives and indicatives are present and past in a very different sense. While present and past denote opposite values of a relation, [±Past], to a temporal reference point (now) in the indicative, they must be thought of as standing for the temporal reference point itself (derived now versus then) in the subjunctive. In fact, both present and past subjunctives substitute for the present indicative, that is, they express much the same temporal relation (basically, not-past) to their respective temporal reference points as the present indicative does with respect to the moment of utterance." (SLGURBSSON 1 9 9 0 : 3 3 0 )

As a result, if the situation described in the subordinate sentence is anterior to that of the matrix sentence, the present perfect has to be chosen if the matrix sentence is in the present and the pluperfect if the matrix sentence is in the past. Compare (74) and (74'): (74) Icelandic [SlGURBSSON 1990: 330] Maria hilt ad hün seei Jon. Mary think-PRET:iND:3SG that she see-PRET:SUBJ:3SG John 'Mary thought that she saw John.' (74') Icelandic [SlGURBSSON 1990: 330] him hefdi Maria helt ad have-PRET:SUBJ:3SG Mary think-PRET:lND:3SG that she sed Jon. see-PART:PERF

John

'Mary thought that she had seen John.' In contrast, in German there is no such difference between the forms of the subjunctive 1 and 2. Irrespective of whether the matrix verb is in the present or in the preterite, in the subordinate sentence both the subjunctive 1 and the subjunctive 2 are possible. This is true for simultaneity and for anteriority, as in (75)-{78): (75) German Maria denkt, Mary

think-PRES:lND:3SG

dass Hans that John

kothme / come-PRES:SUBJ:3SG

käme. come-PRET: SUB J: 3 SG

(76) German Maria dachte, Mary

think-PRET:lND:3SG

käme. come-PRET: SUBJ: 3 SG

dass Hans

komme /

that

come-PRES:SUBJ:3SG

John

The subjunctive mood in German and in the Germanic

351

languages

(77) German denkt, dass Maria think-PRES:lND:3SG that Mary sei / wäre. be-PRES:SUBJ:3SG be-PRET:SUBJ:3SG (78) German Maria dachte, dass Hans Mary think-PRET:IND:3SG that John sei / wäre. be-PRES:SUBJ:3SG be-PRET:SUBJ:3SG

Hans John

gekommen come-PRES:PERF

gekommen come-PRES:PERF

4.3. Reported speech not introduced by a verb of saying In section 4.1. we saw that in both Icelandic and German (and in contradistinction to the remaining Germanic languages) the subjunctive in subordinate clauses serves to indicate the non-factivity of the matrix verb. Since non-factive matrix verbs are in most cases verba dicendi, this use is generally being described as the use of the subjunctive in reported speech. While this analysis is probably not entirely true, it is true that in both Icelandic and German the subjunctive can have the function to indicate reported speech all by itself, i.e. without any matrix verb indicating that someone is speaking or who is speaking. An example for this use of the subjunctive is given in (79): (79) Icelandic [SlGURBSSON 1990: 316] Formadurinn vard oskaplega reidur. Tillagan chairman-the become-PRET:lND:3SG furiously angry proposal-the vceri svivirdileg og vceri henn be-PRET:SUBJ:3SG outrageous and be-PRET:SUBJ:3SG aim-PARTPERF beint gegn sir personulega directly against himself personally 'The chairman became furiously angry. The proposal was outrageous, and it was aimed against himself personally.' Note that the second sentence in (79) reproduces the words of the chairman. That this sentence is reported speech can only be understood by means of the subjunctive verb forms in this sentence. In the first sentence, there is no verb of saying signalling the reported speech following in the second sentence. The very same situation holds for German, as in the quotation from MAX FRlSCH's Homo faber in (80). (80) German Er lachte: he Iaugh-PRET:IND:3SG

"Was what

[KAUFMANN 1976: 93] zeichnen Sie draw-PRES:lND:2POLlTE you:POLlTE

352

ROLF THIEROFF

denn, Faber?" Ich zeichnete PTL

F.

I

auf das Marmor-Tischlein [...],

draw-PRET:iND: 1SG on

sein Lachen störte

his laughter annoy-PRET:lND:3SG me

zu

sagen

wusste.

nothing

to

say-lNF

know-PRET:lND:lSG

I

ja

be-PRES:SUBJ:lSG PTL

dass ich einfach

in-a-way

nichts Ich sei

the marble-table-DlM

mich derart,

that

so

schweigsam.

so

taciturn

I

simply

'He laughed: "What are you drawing, Faber?" I drew on the little table made of marble [...], his laughter annoyed me in a way that I did not know what to say. I was so taciturn, he said.' In (80) the subjunctive of the last sentence indicates that the sentence "You are so taciturn" is uttered by someone, and from the context the reader knows that it is "he" from the first sentence who utters this sentence. A verb of saying or the like is not necessary. This is not possible in any other Germanic language - in the English translation a verb of saying or the like must be added.10 Only in Dutch a similar possibility to introduce reported speech without a verb of saying seems to be possible. At least this is what is suggested by the example quoted by DE HAAN ( 2 0 0 1 : 2 1 3 ) in (81). In Dutch this is done by means of the past tense of the verb zullen 'shall', zou. (81)

Dutch De Congolese

president Laurent Kabila

zou

the Congolese

president

shall-PRET:LND:3SG

dinsdag Tuesday gedood.

[DE HAAN

tijdens een during an Dat meldt

L. K.

2001: 213]

poging attempt

tot to het

staatsgreep zijn state-grasp be-INF persbureau Reuters [...]

kill-PART:PERF that report-PRES:lND:3SG

the

press-office R.

'The President of the Congo Laurent Kabila was reportedly killed Tuesday during a coup attempt. This is reported by Reuters [...]'

5. Pluperfect with non-past time reference in Scandinavian languages In section 3.2. it has been stressed that in all Germanic languages (and also in virtually all other European languages) it is the past form which serves to indicate counteifactuality and it was said that languages only differ as to whether they use the past in the indicative mood or the past together with the subjunctive mood. This, it was argued, can be interpreted as follows: in one group of languages (Icelandic and to certain degree German, marginally also Swedish) a subjunctive is 10

For a detailed description of the use of the subjunctive in reported speech in German and the difficulties to translate instances like the one in (80) see PÜTZ (1989).

353

The subjunctive mood in German and in the Germanic languages

necessary to give a non-past time (and hence, modal) reading to a given past form

(i.e. without the subjunctive the past can only be understood as referring to past time); in the other group of languages (the remaining languages treated in this article) the past can have past time reading or non-past (modal) reading without the help of an additional category such as the subjunctive. While English and Dutch indeed do not use any morphological means to indicate that a past tense has non-past time reference (modal meaning), in Norwegian, Danish and Swedish this use can remain unmarked too, but at the same time these languages have invented a new means to indicate that a given past tense is to be understood modally, with non-past time reference. This means is, astonishingly enough, the perfect marker, which in those same languages (and in English and Dutch, but also in Icelandic and German exclusively) serves precisely to indicate past time reference. The use of the perfect to indicate non-past time reference of a past tense is especially frequent in Norwegian. According to KVIFTE & GUDEHUSKEN (1997: 95) the two sentences in (82), one with the preterite and one with the pluperfect, have the same meaning. (82) Norwegian [KVIFTE& GUDE-HUSKEN 1997: 95] Hvis jeg vant i Lotto, ville jeg kjepe et hus. if I win-PRET:iND in pools will-PRET:iND I buy-INF a house Hvis jeg hadde vunnet i Lotto, hadde if I have-PRET:lND win-PART:PERF in pools have-PRET:iND kjept. et hus. J'eg I buy-P ART: PERF a house 'If I won in the pools, I would buy a house.' Whilst HABERLAND ( 1 9 9 4 ) does not mention HANSEN ( 1 9 9 4 : 5 6 ) claims that in Danish too the

this use for Danish, FABRICIUSpluperfect can be used with non-

past time reference to indicate counterfactuality.11 As for Swedish this use of the pluperfect is mentioned by who gives the example in (83). (83)

Swedish Det hade it have-PRET:LND hade

THORELL ( 1 9 7 3 : 1 3 0 )

[THORELL 1 9 7 3 : 1 3 0 ]

varit

roligt, om barnen be-PART:PERF fine if children-the varit här nu. have-PRET:LND be-PART:PERF here now 'It would be fine if the children were here now.'

Finally, in Swedish (but not in Norwegian or Danish) a combination of both the pluperfect and the s£u//e-construction, in other words, a future past perfect is possible in the //-clause (ÖSTEN DAHL, p.c.). This means that in Swedish with verbs

11

FABRICIUS-HANSEN (1994: 57) also mentions a difference in meaning between the preterite and the pluperfect in sentences like the ones in (82), but space forbids going into this.

354

ROLF THIEROFF

which can still be used in the subjunctive there are no less than five different possibilities to express counterfactuality in the //"-clause (cf. (41) in section 3.2.):12 (84) Swedish Om han vore if

he

hemma, skulle

be-PRET:SUBJ h o m e

Om han var if

he

hemma, skulle

be-PRET:lND h o m e

Om han skulle if

he

vara

vara

tänt. lit

Om han hade if

he

hemma, skulle

varit

have-PRET:lND be-PART:PERF

varit

tänt.

be-PART-PERF

lit

ljuset

shall-PRET:lND light-the

shall-PRET:iND be-lNF h o m e

be-lNF

ljuset

shall-PRET:lND light-the

vara

tönt.

be-lNF

lit

vara

tänt.

be-lNF

lit

ljuset

shall-PRET:iND light-the

hemma, hade home

ljuset

have-PRET:lND light-the

Om han skulle

ha

varit

hemma,

if

have-lNF

be-PART-PERF

home

he

shall-PRET:lND

hade

ljuset

varit

tänt.

have-PRET:lND

light-the

be-PART-PERF

lit

'If he were at home, the light would be on.' This use of the pluperfect is restricted to Norwegian (where it is quite common), Danish and Swedish. It is true that in German there are some very specific contexts where a pluperfect subjunctive can have non-past time reference too (for details see LEIRBUKT ( 1 9 9 1 ) and LEIRBUKT (forthcoming)), but in instances like the one in (84) the pluperfect is excluded with non-past time reference. The only possible translation of (84) is (85)13, whereas (86) can only have past time reference. This also holds for the remaining languages, Icelandic, Dutch and English (compare the English translations). (85) German Wenn er zu Hause wäre, wäre das Licht an. if he at home be-PRET:SUBJ:3SG be-PRET:SUBJ:3SG the light on 'If he were at home, the light would be on.'

12

13

Note that with each of the five different (/^clauses in (85) different forms in the main clause are possible (preterite, future preterite, pluperfect, future past perfect, according to the form of the ifclause). This yields at least ten, if not more different possibilities to express the sentence If he were at home, the light would be on. With the verb sein 'be' the vvwn/e-construction is normally not used in German.

The subjunctive mood in German and in the Germanic

(86) German Wenn er if

he

languages

zu

Hause gewesen

wäre,

at

home

be-PRET:SUBJ:3SG

be-PART:PERF

wäre

das

Licht

angewesen.

be-PRET:SUBJ:3SG

the

light

on-be-PART:PERF

355

'If he had been at home, the light would have been on.'

6. Conclusion In the Germanic languages, the subjunctive forms can be divided into the subjunctive 1, with the finite verb form in the present tense, and the subjunctive 2, with the finite verb form in the past tense. The main function of the subjunctive 1 is that of an optative, the main function of the subjunctive 2 is the indication of counterfactuality or "irreality" in counterfactual conditionals, contrary-to-fact-wishes, unreal comparative sentences and the like. With the exception of Icelandic, the subjunctive 1 is well on its way to being abolished in all Germanic languages. In German it can still be used with optative meaning in several contexts, but all these uses are becoming more and more archaic. In English, the optative meaning is narrowed down to a "mandative" meaning and restricted to the third person singular in the present. In the other Germanic languages, the subjunctive 1 only remains in a small number of fixed expressions of the type 'Long live the king', the example quoted by most grammars. In all Germanic languages, the subjunctive 1 is replaced by the two modal verbs 'may' and 'shall' and their equivalents. Additionally, in some languages 'let' + infinitive is used to convey some of the optative meanings. As a productive category, the subjunctive 2 also remains only in Icelandic. In all other Germanic languages, the subjunctive 2 tends to be replaced by the indicative (with the preterite tense remaining as it is). In German the subjunctive is still the preferred mood in counterfactuals. In Swedish some subjunctive forms of strong verbs still exist, but in non-formal registers the subjunctive is virtually restricted to the verb vara 'be'. In English, only be has a preterite subjunctive, but the indicative of be is used too. In Dutch the preterite subjunctive of zijn 'be' is restricted to certain fixed expressions. In Danish and Norwegian no past subjunctive forms have survived at all. In the apodoses of counterfactual conditional sentences (and in some other cases), the past indicative of the future auxiliary with the infinitive of the main verb is used. This construction has temporal and modal uses in the same fashion as have the preterite and the future. Therefore the would-construction has to be analyzed as a tense (the future preterite) with temporal and modal uses, not in any other way than past and future. Thus the wowW-construction is not a third morphological mood. There is no such thing as a "Conditional" in the Germanic languages. Whereas the future preterite has indicative morphology in the Scandinavian languages, in Dutch and in English (i.e. in those languages which virtually lack a subjunctive 2), it has subjunctive morphology in Icelandic and German. The in-

356

ROLF THIEROFF

dicative counterpart is lacking in both languages. While this can be explained at first look with the fact that in counterfactuals in these languages the subjunctive is the rule, the fact that in free indirect discourse (FID), which by definition is in the indicative, the subjunctive forms are used too, remains problematic. One reason for the lack of the opposition indicative vs. subjunctive in the future preterite could be the fact that verb forms with too many marked categories would result. The fact that it is the subjunctive form (rather than the indicative) which exists could possibly be explained with the greater frequency of the future preterite in counterfactuals and the youth of FID. Whereas the use of the subjunctive forms as well as the means chosen to replace the vanishing subjunctives are very similar in all Germanic languages, there are two features which are unique in individual Germanic languages. These are the function of the subjunctive to indicate reported speech in Icelandic and German (with some differences as to the use of the tenses in the two languages) and the use of the counterfactual pluperfect with non-past time reference in the Scandinavian languages (and of the counterfactual future past perfect in Swedish). Abbreviations 1st person 2nd person 3"1 person dative diminutive future genitive imperative indicative infinitive

1 2 3 DAT DIM FUT GEN IMP IND INF

MASC PART PRES PRET PERF PL PLPF PTL SG SUBJ

masculine participle present preterite perfect plural pluperfect particle singular subjunctive

References ABRAHAM, WERNER (1999): Über allgemeine Zusammenhänge zwischen Präpositionsinfinitiv, Futur, deontischer Modalität und Perfektivitäi: lokalistische Allativität, in Sprachwissenschaft 24,361-405. ANDERSSON, ERIK ( 1 9 9 4 ) : S w e d i s h , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 2 7 1 - 3 1 2 . ASKEDAL, JON OLE ( 1 9 9 4 ) : N o r w e g i a n , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 2 1 9 - 2 7 0 .

DAHL, OSTEN (1995): The tense system of Swedish, in: THIEROFF, ROLF (ed.), Tense systems in European languages II. Tübingen: Niemeyer, 59-68. DE HAAN, FERDINAND (2001): The relation between modality and evidentiality, in: MÜLLER, REIMAR

& REIS, MARGA (eds.), Modalität und Modalverben im Deutschen. (Linguistische Berichte. Sonderheft 9). Hamburg: Buske, 201-216. D E SCHUTTER, GEORGES ( 1 9 9 4 ) : D u t c h , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 4 3 9 - 4 7 7 .

EISENBERG, PETER (1989): Grundriß der deutschen Grammatik. 2nd edition. Stuttgart: Metzler. EISENBERG, PETER (1999): Grundriß der deutschen Grammatik. Vol. 2. Der Satz. Stuttgart/Weimar: Metzler. ENGEL, ULRICH (1988): Deutsche Grammatik. Heidelberg: Groos. FABRICIUS-HANSEN, CATHRINE (1994): Das dänische und norwegische Tempussystem im Vergleich mit d e m deutschen, in: THIEROFF & BALLWEG (eds.), 4 9 - 6 8 . FABRICIUS-HANSEN, CATHRINE (2000): Die Geheimnisse der deutschen WÜNFE-Konstruktion, in: THIEROFF, ROLF; TAMRAT, MATTHIAS; FUHRHOP, NANNA & TEUBER, OLIVER ( e d s . ) ,

Grammatik in Theorie und Praxis. Tübingen: Niemeyer.

Deutsche

357

The subjunctive mood in German and in the Germanic languages FLÄMIG, WALTER (1991): Grammatik des Deutschen. zusammenhänge. Berlin: Akademie Verlag.

Einföhrung

in Struktur und

Wirkungs-

HABERLAND, HARTMUT ( 1 9 9 4 ) : D a n i s h , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 3 1 3 - 3 4 8 .

JANSSEN, THEO A.J.M. (1994) Tense in Dutch: eight "tenses" or two tenses?, in: THIEROFF & BALLWEG ( e d s . ) , 9 3 - 1 1 8 .

JORGENSEN, PETER (1966): German grammar III. New York: New York University Press. KAUFMANN, GERHARD (1976): Die indirekte Rede und mit ihr konkurrierende Formen der Redeerwähnung. München: Hueber. KLOOSTER, WIM (2001): Grammatica van het hedendaags Nederlands. Een volledig overzicht. Den Haag: Sdu Uitgevers. KÖNIG, EKKEHARD ( 1 9 9 4 ) : E n g l i s h , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 5 3 2 - 5 6 5 .

KÖNIG, EKKEHARD & VAN DER AUWERA, JOHAN (eds.) (1994): The Germanic

Languages.

London/

der norwegischen

Sprache.

New York: Routledge. KRESS, BRUNO (1982): Isländische Grammatik. Leipzig: Enzyklopädie. KVIFTE, BJÖRN & GUDE-HUSKEN, VERENA (1997): Praktische

Grammatik

Wilhelmsfeld: Egert. LEIRBUKT, ODDLEIF (1991): "Nächstes Jahr wäre er 200 Jahre alt geworden." Über den Konjunktiv Plusquamperfekt in hypothetischen Bedingungsgefügen mit Zukunftsbezug, in: Zeitschrift fiir germanistische Linguistik 19,158-193. LEIRBUKT, ODDLEIF (forthcoming): Über Konjunktiv Plusquamperfekt und würde + Infinitiv II als Ausdruck von Potentialität oder Irrealität in Konstruktionen mit Gegenwarts- oder Zukunftsbezug, in: LEIRBUKT (ed.).

LEIRBUKT, ODDLEIF (ed.) (forthcoming): Tempus/Temporalität und Modus/Modalität im Deutschen. Tübingen: Stauffenburg. MALING, JOAN & ZAENEN, ANNIE (eds.) (1990): Modern Icelandic syntax. Syntax and semantics Vol. 24. San Diego etc.: Academic Press. PÜTZ, HERBERT (1989): Referat - vor allem Berichtete Rede - im Deutschen und Norwegischen, in: ABRAHAM, WERNER & JANSSEN, THEO (eds.): Tempus - Aspekt - Modus. Die lexikalischen

und

grammatischen Formen in den germanischen Sprachen. Tübingen: Niemeyer, 183-223. QUIRK, RANDOLPH; GREENBAUM, SIDNEY; LEECH, GEOFFREY & SVARTVK, JAN ( 1 9 8 5 ) : A

comprehen-

sive grammar of the English language. London/New York: Longman. REICHENBACH, HANS (1947): Elements of symbolic logic. New York. RITTE, HANS (1986): Schwedische Grammatik. München: Hueber. SIGURDSSON, HALLDÖR ÄRMANN (1990): Long-distance reflexives and m o o d s in Icelandic, in: MALING & ZAENEN ( e d s . ) , 3 0 9 - 3 4 6 .

THIEROFF, ROLF (1992): Das finite Verb im Deutschen. Tempus - Modus - Distanz. Tübingen: Narr. THIEROFF, ROLF (1994): Inherent verb categories and categorizations in European languages, in: THIEROFF & BALLWEG ( e d s . ) , 3 - 4 5 .

THIEROFF, ROLF (1999): Preterites and imperfects in the languages of Europe, in: ABRAHAM, WERNER

& KULIKOV, LEONID (eds.), Tense-aspect, transitivity and causativity. Essays in honour of Vladimir Nedjalkov. (Studies in language companion series [SLCS] vol. 50). Amsterdam/Philadelphia: Benjamins, 141-161. THIEROFF, ROLF (2000): O n the areal distribution of tense-aspect categories in Europe, in: DAHL,

OSTEN (ed.), Tense and aspect in the languages of Europe. (Empirical Approaches to Languages Typology [EALT]/EUROTYP 20-6). Berlin, New York: Mouton de Gruyter, 265-305. THIEROFF, ROLF (forthcoming): Modale Tempora - non-modale Modi. Zu Bedeutung und Gebrauch inhärenter Verbkategorien in verschiedenen europäischen Sprachen, in: LEIRBUKT (ed.), 63-85. THIEROFF, ROLF & BALLWEG, JOACHIM (eds.) (1994): Tense systems in European languages. Tübingen: Niemeyer. THORELL, OLOF (1973): Svensk grammatik. Stockholm: Esselte Studium. THRÄINSSON, HÖSKULDUR (1990): A semantic reflexive in Icelandic, in: MALING & ZAENEN (eds.), 289-307 THRÄINSSON, HÖSKULDUR ( 1 9 9 4 ) : I c e l a n d i c , in: KÖNIG & VAN DER AUWERA ( e d s . ) , 1 4 2 - 1 8 9 .

WESSEN, ELIAS (1968): Värtsvenska spräk. Stockholm: Almqvist & Wiksell. ZIFONUN, GISELA; HOFFMANN, LUDGER & STRECKER, BRUNO ( 1 9 9 7 ) : Grammatik

che. 3 volumes. Berlin/New York: de Gruyter.

der deutschen

Spra-

List of authors

WERNER ABRAHAM

MOLLY DLESING

Institut für Allgemeine Sprachwissenschaft Universität Wien Berggasse 11 A-1090 Wien Home: Maitschern 128 A-8942 Wörschach AUSTRIA [email protected]

Department of Linguistics Morrill Hall Cornell University Ithaca, NY 14853 USA [email protected] [email protected]

J0HANNA BARDDAL

Department of Scandinavian Languages and Literature Section of Scandinavian Linguistics University of Bergen Sydnesplassen 7 NO-5007 Bergen NORWAY [email protected] CEDRIC BOECKX

Dept of Linguistics Harvard University Boylston Hall Cambridge, MA 02138 USA [email protected] C. JAC CONRADIE

Randse Afrikaanse Universiteit/ Rand Afrikaans University Auckland Park Johannesburg Posbus 524 Auckland Park 2006 ZUID AFRICA [email protected]

BRIDGET DRINKA

Department of English, Classics, and Philosophy University of Texas at San Antonio San Antonio, Texas 78249 USA [email protected] KLEANTHES K. GROHMANN

University of Cyprus Dept. of Foreign Languages & Literatures P.O. Box 20537 1678 Nicosia CYPRUS [email protected] JOHN MCWHORTER

Linguistics University of California Berkeley, CA 94720-3243 USA [email protected]

360 LÄSZL0 MOLNÄRFI

University of Tilburg Grammaticamodellen Postbus 90153 NL-5000 LE Tilburg THE NETHERLANDS [email protected] HALLDOR SIGURDSSON

Institute og Scandinavian Linguistics Lund University Helgonabacken 14 SE 223 62 Lund SWEDEN [email protected] ROLF THIEROFF

FB Sprach- und Literaturwissenschaft Universität Osnabrück Neuer Graben 40 D-49069 Osnabrück GERMANY [email protected]

List of authors