Present-day Dialectology: Problems and Findings 9783110904765, 9783110167818

Citation preview

Present-day Dialectology

W

Trends in Linguistics Studies and Monographs 137

Editors

Werner Winter Walter Bisang

Mouton de Gruyter Berlin · New York

Present-day Dialectology Problems and Findings

Edited by

Jan Berns Jaap van Marie

Mouton de Gruyter Berlin · New York

2002

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

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

Die Deutsche Bibliothek

— Cataloging-in-Publication

Data

Present day dialectology : problems and findings / ed. by Jan Berns ; Jaap van Marie. - Berlin ; New York : Mouton de Gruyter, 2002 (Trends in linguistics : Studies and monographs ; 137) ISBN 3-11-016781-6

© Copyright 2002 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Contents

Characteristics of spoken vernaculars: parsing strategies? The case of German Werner Abraham

1

Language variation and phonological theory: inflected adjectives in Dutch and related languages Geert Booij

35

Perceptual dialectology: aims, methods, findings Dennis R. Preston

57

Parallels in the phonological systems of Southern Dutch and Northern French dialects Georges de Schutter

105

Britain and the Continent - linguistic interrelations Wolfgang Viereck

123

Aspects of Armenian dialectology Joseph J. S. Weitenberg

141

Typology and dialectology: a programmatic sketch Lieselotte Anderwald and Bernd Kortmann

159

The role of prosodic and verbal aspects of speech in the perceived divergence of Dutch and English language varieties Charlotte Gooskens and Renee van Bezooijen

173

Morphologische Reanalysen bei lokalen Adverbien, Präpositionen und Adjektiven im Thüringischen und Ostfränkischen Rüdiger Harnisch

193

Basque accentuation and dialectology Jose Ignacio Hualde

207

vi

Contents

On the (non-)persistence of dialect features in American Dutch (1): general aspects Jaap van Marie and Caroline Smits

231

On the (non-)persistence of dialect features in American Dutch (2): the case of Iowa Dutch Caroline Smits

243

The impact of internal or contact-induced change on weak preterites in -of in Danish dialects with an outlook to Norway and Sweden Inge Lise Pedersen

269

Meta-constraints and constraint ranking in relation to the representation of nasality and palatality across French dialects Michael D. Picone

283

Probleme bei der Erstellung von Sprachkarten im Untersuchungsgebiet des bayerischen Sprachatlas" am Beispiel von Richtungsadverbien Sibylle Reichel

301

Dialektwandel und Einstellung - Das Beispiel der Aarauer Stadtmundart Dialect change and attitudes Beat Siebenhaar

313

The pronunciation of standard Dutch - An evaluation study Dick Smakman and Renee van Bezooijen

333

On ethnobiological nomenclature in Southern Dutch dialects Jos Swanenberg

353

Characteristics of spoken vernaculars: parsing strategies? The case of German Werner Abraham

Avec le talent, on fait ce qu'on veut. Avec le genie, on fait ce qu'on peut. (Dominique Ingres, 1780-1867)

The following paper will be concerned with the specific properties spoken languages may display in opposition to written counterparts. My point in case is will be German. I will show that, indeed, spoken German, i.e. its dialects, exhibit fundamental differences, and I will attempt to explain this difference. My claim is that if dialects are taken not only to be non-literary or non-normalized, but primarily spoken variants of national vernaculars, it will be easier to characterize their specific modifications of the written counterparts. It will be argued that, if structural considerations of grammatical structure play a major role in the description and explanation of written linguistic phenomena in, let us say, the formal, Cartesian linguistic paradigm, Functionalists, by contrasts, will be the ones who consider analytic inventories which are more apt to grasp facts of spoken linguistic characteristics such as theme and rheme. And, unless the latter discourse functions can be linked directly and uniquely to identifiers in terms of structural properties of the clause, it follows necessarily that the two schools of linguistic thinking deal with different topics of linguistic description: the Cartesian formalists are concerned with deep structural concepts and mechanisms, whereas the Functionalists, motivated by their prime interest in discourse categories such as theme and rheme, will necessarily exploit mechanisms of parsing. Notice that the two methodologies are principally distinct (Pritchett 1991, Bayer (ed.) 1995, Hawkins 1996, Lasnik 1997).

2

Werner Abraham

1. The assumption German,1 just as well as other sov or V-last languages, has one asset which makes the language difficult, even conceptually and without looking at the practice of writing and speaking: this is the fact that the finite predicate in dependent clauses is the last item or that the predicate is split between one part in clause-second and one in clause-final position. This causes aporia, e.g. in the sense of the learned Mark Twain, who denies this language the properties of a natural language given that the clausal item carrying the most important information, i.e. the finite verb, is placed way behind all other material in the clause. This is, as Twain holds, counter-natural.2 In fact, no one could agree more than those engaged in the science of parsing, i.e. the science investigating which linear and cohesive features enable or facilitate the flow of information in clauses. Under such premisses, languages such as German are highly uneconomical and difficult to process because all material preceding the finite V-last in dependent clauses need to be speculated on with respect to its clause-cohesive properties until, in the end, the finite verb would render the required and longed for information were it not that the listener may have forgotten the material preceding the clause-final predicate in the meantime. Parsing researchers of German have to assume that native German speakers and listeners have to learn a particular quality of surface-storing the material that has to wait for clausal identification until the verb appears linearly allowing, not until then, to identify the short-memory stored elements as to their cohesion with the verb and to process the linear chain to arrive at a full understanding in a somewhat timeconsuming succession. At the least, such short-memory storage and reconstruction will separate this type of information processing from a linear steby-step processing as, e.g., in all other Germanic sov languages. This scenario is exaggerated, however. For the writer of German (and any of the Indo-Iranian V-last languages or Japanese etc.) this causes no difficulty, since he has preconceived what he wants to write down, no matter how this fossilization in writing is linearized. Likewise, the reader of German can focus the whole sentence read from back to front if need to. But if German were a language without writing and reading which it is not, this would be indeed a handicap of sorts. Notice that languages such as the Indo-Aryan ones or Japanese have other clues presorting the parts of speech prior to arriving at the valence bearing predicate in clause-final position (the accusative-ergative case paradigms are important clues as are function words such as those for topics and focus constituents). Consider now if the spoken

Characteristics of spoken vernaculars

3

variant of German had no such clues? Wouldn t this variety of German really be in trouble i.e. suffer under the severe handicap that speakers of other languages would always be ahead in timely understanding? Is there such a variant of German? If so we would be interested to see whether, indeed, and how this variety could survive in dignity i.e. with which other means. The scenario that I have just alluded to above does indeed exist in the German speaking areas. And the varieties to investigate with respect to the last issues of identification are spoken variants of German. The first methodical claim is thus: (1) Spoken German must differ from written German unless it succumbs to handicaps in the on-line information flow. But before we engage in an inductive search as to which properties distinguish indeed the written from the spoken varieties, let us speculate how Vlast could be salvaged, or, to phrase this less dramatically, how it could become somewhat easier to transfer information while not violating the grammar of the language. Let us then briefly speculate about the general properties of a true listener-language: no doubt, discourse functions such as topics (themes) should be identified and, consequently, be distinguished from foci (rhemes) (as in Japanese); and sentence information should be cut down to portions which can be surveyed with ease: in short, sentences should become shorter than the average German sentence. Possibly, both goals can be subsumed under the strategy of distinguishing carefully discourse functions such as theme and rheme. Note that, if we succeed in explaining how the spoken variety of German avoids the pitfalls of written German, we might be able to answer other questions. (2) Why is it that other Germanic sov-languages do not share the impediments of their close linguistic brother German? Is it that they have given in, in their respective linguistic histories, to such demands as parsing facilitation? (3) What must a language such as German provide so as to enable its variants to be easier in terms of parsing, but simultaneously retain its fundamental grammatical structure?

4

Werner Abraham

I suggest that the following two types of discourse-oriented mechanisms are important vehicles in arriving at answer to the claims in (l)-(3) above. (4) Discourse functions overwriting grammatical structures and simultaneously retaining the non-expropriatable grammatical structure, must make use of a refined clausal accent structure and of identification of linear positions in and outside the clause. (5) The strategy sketched in (4) may even lead to the emergence of lexical elements and categories that allow grammatical structures to be preserved and secure proper discourse identifications simultaneously. I will demonstrate that both mechanisms are in use in modern spoken variants of German. In fact, I can show that diachronically earlier German indeed heeded these principles something, however, I will pursue only occasionally, i.e. not systematically.

2. Accent and linear position Quite obviously, accent can be marked in written vernacular. The accent diacritic on Dutch vowels are a case in point. German does not provide this means. Where the clausal default accent position and the focussed item coincide this provides no difficulty, as in the ensuing examples. Notice that Dutch does not need to mark accent in these cases. However, in other cases such as those in (7), Dutch is simply facilitating a faster and simpler flow of information avoiding ambiguities. [SMALL CAPS for clausal accent, either default or contrastive]. (6)

a.

GERMAN: Das

ist das EINE

-

DUTCH: Dat

is het

-

DUTCH: Hij heeft

EEN/MN

this is the one b . GERMAN: Er hat das nicht

ANgetastet

(7) a. b. c. d. e.

he has that not on-touched Er is een/een/'n man die nooit vergeet Es gibt einen MANN, der nie VERGISST Es gibt EINEN Mann, der NIE vergißt there is a/one man who never forgets ?? Es gibt einen MANN, der NIE vergißt 77 Es gibt EINEN Mann, der nie VERGISST

AANgetast

dit

niet

Characteristics of spoken vernaculars

5

There are two types of accents: that which marks the position in the sentence which is the typical new information. Let us call this the grammatical accent. This is what prevails in (6). There is also contrastive accent. Contrastive accent marks that which has moved from its grammatical position in order to signal that it requires special attention. There are special grammatical categories which serve to identify the discourse elements previously spoken about or supposed to be identifiable without further pre-mentioning. Such prototypical thematic elements are pronouns and definite determiners. In what follws I discuss briefly strategies where this discourse categorization is pursued with pervasiveness in spoken German.

3. The «/-paradigm Phenomena such as the Bavarian pervasive definite determination ofproper name such as *(der) Peter and the general lack of zw-infinitivals have never been seen to be linked in any reasonable sense. In fact, the two phenomena have never been mentioned anywhere near to one another. Yet, both are (i) clear violations of the Standard written language and (ii) both do in fact have a overt common denominator: the definite article. Witness the illustrations below where the versions in parenthesis stand for what is required in the spoken vernaculars of German. (8) a. *(der) Peter (proper name!) the Peter b. er hat nichts zu*((*de)m) (den ganzen Tag) lesen he has nothing to+DAT (the whole day) read c. von/vo*(m) Tuten und Blasen keine Ahnung haben of/of+DAT sounding and blowing no idea have (W. Friederich Idiomenwörterbuch vs. Bavarian-Austrian spoken version) d. Eine Prinzessin (*zu) küssen, das wollte die Prinzessin immer schon a princess (to) kiss, that wished the princess always already Whereas written Standard German ostracizes the definite article prior to a proper name it is the socially and regionally identifying rule in dialects and regiolects. This simply extends the signal of thematicity (which in the case of proper names is inherent) from common names, or appellatives, to the inher-

6

Werner Abraham

ently theme bearing category of proper names. We may call this a form of paradigm extension. The paradigm is thematicity. Such definiteness extends to what has come to be called 'infinitival prepositions in modern syntax. Note, however, that the inflected preposition in (8b,c) in (a) governs a VP, or certainly more than simply a verbal infinitive), and (b) it does not involve a full determiner in that only the case marking of the determiner is spelled out. We may assume that this signals two things: for one that the preposition has scope over a nominal (including case and aspect, but not tense!) and not a verb, and, second, that despite the case inflection of the verbal nominal, it is not definiteness that is signalled, but, rather, 'dedefinitization a truly derivative, i.e. privative process. Notice that one cannot say that the inflected preposition denotes P+indefinite article+case inflection. Rather, what the inflected Ρ in (8b,c) is something inbetween indefiniteness and definiteness, but closer to the latter. The best one can say is that it stands for despecification of the definiteness marker. Leaving aside the general denominator in terms of definiteness and its extension, what (8b,c) tell us, now co-opting (8d) in this account, is also that there are no 'prepositional infinitivals in the spoken language something that again simplifies the system in terms of category selection: Ρ selects only the categorial characteristics [-verbal, +nominal] an insight that we have known from the standard Indo-European and German grammars all along. The inferences from the stance "prepositional infinitive" are drastic and have led to the interpolation of IP above vp. It pays off to reconsider grammars with rich case inflection against the background sketched here: after all, such an IP is certainly not required, or even plausible, for spoken German. "More case inflection - less clausal syntax".

4. Left dislocation and extraposition as parsing strategies: Austrian/ Bavarian's answer to Mark Twain Phenomena of left dislocation and its specific non-structural linearization principles (Novelty Condition: Wiltschko 1997) are cases in point for a mechanics of facilitating processing long syntactic linear chains, while simultaneously inserting identifying material in terms of discourse functional categories. Notice that the left-dislocated material are topics (themes) as well as contrastive rhemes. See the illustrations in (9).

Characteristics of spoken vernaculars

7

(9) a. Den Frosch, den hat die Prinzessin nicht geküßt the-ACCfrog,that has the princess not kissed b. *Keinen Frosch, den hat sie geküßt no-ACC frog, that has she kissed c. *Nicht den Frosch, den hat sie geküßt not the frog, that has she kissed d. * Sogar den Frosch, den hat sie geküßt even the-ACCfrog,that has she kissed e. * Niemanden, den hat die Prinzessin geküßt no one-ACC, that has the princess kissed Notice the restrictions that hold according to (9b,c,d,e): the left-dislocated material may identify rhematic information. However, it must be referential, (9b,c,e), and the resuming predication must be more rhematic than the dislocated element, (9b). Parentheticals and the Novelty Condition (Wiltschko 1997) serve an identical purpose. See (10) below (where the clause-medial nominal, is to be read with parenthetical status and accent). (10) a. weil die Prinzessin (, dieses hübsche Mädchen,) den Frosch geküßt hat since the princess (this nice girl) the-ACCfroghas kissed 7 b. weil das Mädchen, die Prinzessin, den Frosch geküßt hat since the girl, a/this princess, the-ACC frog has kissed If parentheticals are 'reminders' of what we are talking about and thus referentially 'anaphoric' in kind, then it appears plausible that such 'anaphors', or thematic pro-forms, should be co-referring, without inserting new identifying information. Such later rhematic information would destroy the anaphoric status of the parenthetical. Cf. (10b) where the inserted parenthetical, eine/diese Prinzessin 'a/this princess', is intensionally richer than its left-dislocated binder, das Mädchen 'the girl'. What all this is meant to show is that certain discourse tactics are at work once material is dislocated across the clausal boundaries. No doubt, such dislocation triggers binding mechanisms akin to anaphorization by means of pro-forms (pronominals). Extraposition and right-dislocation are other cases in point. See the few illustrating examples below (from Patocka 1997: 94f.; 112ff.).

8

Werner Abraham

(dialect phonology levelled out to Standard German) (11) a. Da ist der Flakzug drüben gestanden, weil die Bahn, ist ja bei uns gerade so ein paar hundert Meter weit weg, die Bahn .. Eisenbahn.. (...) because the rails are just a few hundred meters away, the rails, the train. b. Also muß das gespritzt werden, weil die. vergrasen ja so, die Äckerr (...) because they grow over wild, those fields. RIGHT DISLOCATION

EXTRAPOSITION

(12) a. ... dann habe ich halt meine Kinder abgerichtet zum Musizieren ... then I coached my kids to make music b. Aber weil man heute so ein wenig eingewöhnt ist mit der Maschine but since one is used to have machines nowadays. Right dislocation as in (1 la,b) shows similarity to left dislocation. However, the linear relation between binder and bindee is reversed. Both the NP in (1 la) and the demonstrative in (lib) need to be co-referred to by intensionally richer, or at least not poorer, binders in dislocation. In the extraposed examples, on the other hand, no presumptive element precedes the extraposed constituent. Thus, as regards their introduction in discourse referential terms, they are different: die in (1 lb) is a discourse cataphor, whose content reference need to be identified by the dislocated postcedent NP, die (Eisen-)Bahn. The extraposed examples, by contrast, occur as new informatioon. But their discourse status is disfocussed in this clause-marginal position. What unites the two types, dislocation and extraposition, however, is their discourse function: they are thematic: by way of cataphoric pre-introduction in (1 la,b), and by way of disfocussing from the rhematic discourse space in (12). The written standard of German does not make use of these discourse-identifying mechanisms.

5. Multiple negation Multiple negation with simple negation readings may touch one as being odd and non-standard, or even backwoodish, certainly judged with the ears of a standard-speaking German. However, this is a conclusion drawn too quickly. If checked across the languages of the world, multiple negation turns out to be very common cross-linguistically (van der Woude 1997). In fact, most of the early Germanic stages displayed multiple negation (as investigated for

Characteristics of spoken vernaculars

9

Old High and Middle High German, Middle Dutch, Old English, and Gothic; see already Bruggmann/Delbrück 1888-1900). The intriguing problem for modern syntax is not only how to tell negation cancellation and multiple negation apart if one single descriptive account is to be chosen. What is even more significant is that we appear to be able to internalize both systems side by side: the standard logical cancellation as speakers of the written standard and multiple, non-cancelling negation as natives in the spoken vernacular. (13a-g) is a case in point. If identified as Standard modern German (and under abstraction from accent distribution, which is a disambiguator!), (13c) would have to denote that 'there for every dog it was the case that he bit the man, whereas in the spoken, substandard vernacular it just means that 'no dog bit him. To answer this apparent aporia, it was argued at some length (Abraham 1999) that negation over definite variables is radically different from negation over indefinites. Furthermore, multiple negation occurs only over indefinites or determinerless clausal arguments and adverbials. See (13a-g) for the distribution between multiple negation and accent as well as linear position in the Austrian-Bavarian substandard of German. Alemannic German would just as well exemplify this phenomenon. Part of the examples are due to Bayer (1990). (13) a. *daß kein MENSCH dem Jäger [yp kein Bier nicht GEZAHLT hat] that no human the-DAT forester no beer not paid has (i) OKdaß kein MENSCH dem Jäger [yp kein BIER nicht gezahlt hat] (ii) *daß kein HUND [yp t nicht einen Postboten GEBISSEN hat] (iii) okdaß kein HJJND [vp t in HUND nicht GEBISSEN hat] b. *daß kein HUND einen Postboten [vp 11 nicht GEBISSEN hat] that no dog a mailman not bitten has ok (iv) daß kein HUND [yp t einen Postboten GEBISSEN hat] c. daß ihn [yp 11 kein Hund nicht GEBISSEN hat] that him no dog not bitten has d. daß er da noch NIE [yp t keinen MENSCHEN gesehn that he there yet never no-ACC man not seen hat ] has not e. *daß er da keinen Menschen noch NIE fyp 11 gesehen hat] that he there no-ACC man yet never seen has f. *daß KEINER in das Auto [yp 11 nicht EiNgestiegen ist] that no one into the car not embarked is

10

Werner Abraham

g. *daß NIEMAND der Sophie [vp 11 nicht WEHtun wollte] that no one the-DAT Sophie no harm wanted (ν) °kdaß der Sophie [

t niemand nicht WEHtun wollte]

We assume, following the stress null hypothesis, that focus accent on nicht can only occur before [+def] NPS, i.e. NPS outside of VP and, therefore, with thematic discourse status. The incorporated constituent negation, on the other hand, is constrained to [-def], inside, rhematic, NPS. There is no doubt that the focal vs. disfocal distinction which we have drawn above plays a crucial role in the distributions in (13a-g) above. How is this distribution to be explained in plain, non-technical linguistic terminology? Notice that, for one, negation may be multiplied outside of VP only once (witness (13e)). Furthermore, what is striking is that the chain of positions extends into the extraposition domain; witness (13d). What can we make of this in generalizing terms? There are two distributional characteristics to be noticed: (14) a. Negation before the indefinite determiner, ein-, always incorporates to kein-.

b. The accent of Neg in multiple distribution is weak; witness (13d) with the extraposed position for the negative element. This leaves us with only one explanation: the weak negation, wNeg, is thematic in the sense that it iterates pronominally down to the last thematic, or anaphoric, position in the clause, the extraposition position. This echoes Delbrücks (1882) generalisation that the negative elements in early Germanic pervades the whole clause without adding any new information. We draw the important conclusion that, again, it is a discourse feature, in our case thematicity in distinction to rhematicity, that allows us to determine the distinctive property of something that appears to be prototypical for the spoken vernacular. Notice that the thematic iteration of negation occurs in the rhematic vp-domain. In other words, the chained negative anaphoricity is implemental in a discourse domain which is reserved for rhematic information in the discourse. This fact makes the discourse status of thematic negation raise to double weight: it has to be in opposition to what is standardly in the status of the opposite discourse category (RHEME); and it has to pervade the whole syntactic rhematic domain, VP, to ensure that the negative operating information is not dropped under the weight of the rhematic weight. Let

Characteristics of spoken vernaculars

11

me point out for syntacticians that this leads to intriguing problems for a unified syntactic description (Abraham 1999). The first three issues above highlight the first type of discourse-strategy of spoken language as claimed in (4). Let us now turn to one that is more intricate, that sketched in general terms in (5) above: the suppletive discourse function of elements for spoken language.

6. TUN 'DO' as parsing-Aux: the grammatical optimalization scenario of German tun as an auxiliary verb3 Let us repeat the two strategies in (4) and (5), which may now be called the prosodic strategy, (15), and the suppletive strategy ensuring the appropriate prosodic structure, (16). (15) The direct prosodic strategy. Discourse functions overwriting grammatical structures and simultaneously retaining the non-expropriatable grammatical structure, must make use of a refined clausal accent structure and of identification of linear positions in and outside the clause. (16) The suppletive strategy ensuring the appropriate prosodic structure: The strategy sketched in (15) will even lead to the emergence of lexical elements and categories that allow grammatical structures to be preserved and secure proper discourse identifications simultaneously. In what follows I intend to suggest a totally novel explanation to the usage of tun 'do' as an auxiliary. It is suffciently well-known (Eroms 1984) that tun is used as an auxiliary in various, and not always very clear, functions in Upper German dialects and also in Low German. Alemannic participates in this non-standard usage, however not without differences from all other German dialects to the extent that Swiss-German knows also forms of the present subjunctive (for the oratio obliquä). Whereas Austrian-Bavarian makes use only of the preterite subjunctive, the tät(-)-, form for the hypothetical function (which in Standard, written German is expressed by a periphrasis with würde or else by a synthetic subjunctive form) High Alemannic has also unambiguous forms and functions of the first (present) subjunctive, for indirect speech. Witness (17)-(18). [HA = High Alemannic BA = Bavarian-Austrian; dialectal phonetic variants are collectively represented by the tät-form].

12

Werner Abraham

Er hat gesagt, (daß) er arbeite/bleibe/ tue (17) 'Rhot geset, (daß) 'r schqffi/bli:bi/ tüaji he has said (that) he works/stays/that does.suBjl (18) a. Ea hot gsogt, ea tät arbatn/bleibm/des *tuan he has said he did work/stay/that do b. Ea hot gsogt, doß ea arbatn/bleibm/des *tuan tät

HA BA BA

Cf. (18a) which supports the conclusion that tät- cannot function as AUX in construction with the full lexical verb tun to all appearances for euphonic reasons. In what follows I claim that Upper German tun-AUX has the function of separating the discourse categories of theme and rheme in the clause or, respectively, to identify the rheme of the clausal predicate. This, on the other hand, would seem to presuppose showing that German highlights in a structural way the distribution between Rheme (RH) and Theme (TH). In Abraham (1995a, b) it was claimed that TH/RH can be fully integrated into the clausal syntax. In other words, the discourse-functional pragmatics of the TH\RHstructure and the distribution of clausal accent can be expressed in terms of syntactic representations and need no access to an extra module (as discussed in detail by Valldüvi 1991; see for an identical claim Chomsky 1995). In agreement with my earlier stance I will demonstrate that the finite tun- in clause-final position is to be identified as the coda position in terms of lexical phonology. This is not a trivial assumption and I admit readily that not all questions following from this position have a satisfactory answer. What makes this assumption even less trivial is the fact that it appears to bear on the phenomenon of the decay of the preterite forms in Upper German ("Oberdeutscher Präteritumsschwund"; Lindgren 1963) in ways that lead to a better understanding of it.

7. Distributional properties and limits 7.1. The tun -support as simple as in English? There is ample agreement about the fact that BA tun and English do share syntactic properties (see, most recently, Eroms 1984: 123). The following examples are meant to illustrate that this is but a very superficial characterization. ['#' marks substandard German forms only; in other wortds, Standard German excludes the subsequent form.]

Characteristics of spoken vernaculars

13

(19) English: a. He does not agree b. *He agrees c. *He does never agree d. * Nobody does agree e. He never agrees f. Nobody agrees g. He never does agree (20) German: a. #Er tut nicht zustimmen/Er stimmt nicht zu BA he does not PREFix.agree/he agrees not PREFIX b. Er stimmt zu Standard German c. #Er tut niemals zustimmen/Er stimmt niemals zu ΒA d. #Niemand tut zustimmen BA no one does PREFIX, agree e. Er stimmt zu Standard German f. Niemand stimmt zu g. #Er tut niemals zustimmen BA Notice that English simply cannot do without the ί/o-support in the constructions under inspection in any of its dialectal or sociolectal varieties much opposed to German. Viz. (20b) above. In addition to that, negation (and, beyond that, also other elements) never intervenes between the finite verb and its separated verbal particle. (21) a. (daß) er nicht zustimmt that he not PREFIX, agrees b. (daß) *er nicht stimmt zu that he not agrees PREFIX See also (18b) above. In total, this does not render a picture which is homogeneous between the fw«/" is more optimal, or has a higher acceptance, than the construction in the position to the right of the angle, with the dependent clause as well as the purely prosodic coda phenomenon occupying the position with the least acceptance among the construction types under inspection. Let us ask what this can mean in terms of optimality theory. The following common properties need to be integrated. Note that, at this point, the assumption of syntactic bracketting has no more status than that of a pretheoretical criterion. In other words, it does not provide an explanation of the facts at this point. (49) a. In the interrogative clause with V-l as well as in the independent declarative clause with V-2 two V-positions need to be ascertained. For both V-l and V-2, the V-last-position is unoccupied in case the left-marginal positions are taken by the full verb.

26

Werner Abraham

b. In the dependent clause both marginal positions forming the verbal bracket are occupied: V-l is taken by the (often inflected) conjunction, whereas V-final by the full verb. This type has the least acceptance with respect to cooperation with the tan-periphrasis. c. The imperative has an insignificantly lesser acceptance than the interrogative and the declarative clause. As in the previous case, the full verb in V-l, i.e. without the periphrasis, has to make sure that the verbal bracket is activated by means of the inflected tun-carrier. d. Beyond doubt, Bavarian reflects complementary distribution between perfectives and imperfectives (Eroms 1984: 131). Since in German ergatives and perfectives are identical in terms of syntactic distributions (Abraham 1994, 1995a, b), the restrictions observed to hold in section 2.2.2.2, example (27), are part of this generalization. This is in need of an explanation, however. Note that safeguarding the verbal bracket cannot explain this. Let us disregard the prosodic criterion of coda formation for the time being. The best evidence for the assumption that the verbal bracket - i.e. lexical insertion in V-l/V-2 (which, according to UG occupy an identical structural position in C°/I°!) and V-last is the determining criterion in the distribution of tun is provided by the ungrammatical perfect periphrases, by the forms of the double perfect as well as the future periphrasis. None of these analytic constructions permits the fan-support, which is restricted in occurrence to the present. No other tense formation is amenable to this support. This conclusion is further corroborated by the occurrence of the double-perfect forms in substandard Upper German, which are no longer restricted to perfective verbs. Nevertheless, they are evidence of an aspectual residual in the organization of the event organization (cf. Litvinov-Rad(enko 1997). If this generalization 'construe a verbal bracket' is correct we shall have to answer two questions: (i) What is the function of the verbal bracket such that it plays its dominant role in Upper German (both BA and HA)? And (ii): What is behind the aspect sensitivity as in (48b) and (49d)?

8.2. Functional explanation: an attempt If the verbal bracket is indeed a crucial criterion in the sentential structure of Upper German (not, however, in the structure of Low German and Dutch, by the same token, since the synthetic preterite remained intact in these lan-

Characteristics of spoken vernaculars

27

guages!) we would want to find an account for this fact on the next accessible level of description: i.e. the discourse functional level, where the discourse functions theme and rheme receive a clause structural representation. I assume that such discourse functions can be represented in a direct way in the sentential syntax (Abraham 1995a, b). What this boils down to is that the grammatical (i.e. the non-contrastive) rhematic accent in German sentence structure lies on the lexical head of the most deeply embedded sentential constituent. This may be the rightmost NP or prepositional object, or V in Vlast itself, in case VP is not filled otherwise (Abraham 1995a, b). What this claim presupposes is the realization of vp in order to identify the grammatical rhematic accent. Note that at least i.e. if vp is empty otherwise the position of V° is occupied. Since all definite (since thematic) NPS and PPS, respectively, move from vp to the left into the thematic domain of the clausal structure it is mainly in categoric (as opposed to thetic) sentences that support by means of an auxiliary verb is useful in discourse-functional terms: in the independent clause just as well as in the verb-left marginal clausal forms (imperative, interrogative). To the extent that another auxiliary is absent such as sein, werden, haben, or a modal verb), tun takes over in this function. It is to be noted that this is in total agreement with the phenomenon of the strict decay of the preterite in Upper German. In Low German and Dutch, on the other hand, where the simple preterite paradigm has been preserved one expects that the mechanism of discourse-functional support is represented to a but negligible extent.4 I leave open at this point whether the prosodic coda criterion is in line with the 'soft' extraposition phenomena disagreeing with the grammar of (written) Standard German. See (50) for extraposition facts. (50) daß er den Kuchen alleine aufgegessen hat heute nachmittag bei seiner Mutter that he the cake alone up-eaten has this afternoon at his mother's It has to be left at this for the time being. The higher acceptance of perfectives as discussed above appears to be part of the discourse functional related preference scenario just sketched. Perfectives are represented without exception as secondary predicates (embedding of small clauses). See Abraham (1994, 1995a, b). The discourse functional auxiliary tun appears to take over the inflective function of the finite matrix verb. See again (15) above as regards the unaccusative verbs identified in German as perfectives without exception.

28

Werner Abraham

What all this, in terms of preferred acceptances as in (32), boils down to is the following representational hierarchy. (51) preferred acceptances: ("a > b" ... 'a is preferred over b') ftw-forms in a. independent interrogative or declarative discourse functional syntax (where perfectives > imperfectives) > discourse functional syntax > imperative > discourse functional syntax > dependent clause > discourse functional syntax > sentence prosodic coda formation sentence prosody

9. Speculation: possible defendenda and refutanda with respect to the decay of the preterite forms in Upper German ("Präteritumsschwund im Oberdeutschen") The speculative order in terms of an optimalization scenario as in (39) above carries over in a direct way to a novel explanation of what has become to be called the "decay of the preterite form in Upper German" ("Oberdeutscher Präteritumschwund") in German philology ever since Lindgren (1963). Quite naturally, the old" arguments (such as the morphological levelling of the paradigms for tense and mood subsequent to, and triggered by, a wideranging apocope of weak vowels in the inflective endings) need be given the methodological status of conditiones sine quibus non (necessary arguments), not, however, those which render support to why, against the background of this inflective decay, exactly those phenomena emerged that indeed have emerged (i.e. answers in terms of satisfying arguments). Such an argument in point would be that the German dialects of Upper German, which avoid subjectivization to standardization processes in Modern (written) German to a large extent, developed their sentential grammar more strongly in terms of distinctions of the discourse functions of theme and rheme something which meets demands of the speech vernacular optimally. Notice that the clausefinal position of the verb in dependent sentences standardized by normative grammarians in the 17th and 18th centuries meets such a "natural" regularity in terms of an oral vernacular (although this has never been consciously admitted as such by those grammarians). After all, the non-dependent clause with the finite predicate in first ("V-l") or second position "V-2") can be processed optimally under truth-functional conditions, i.e. as early as possible in the linear on-line flow of the sentential elements, whereas the depend-

Characteristics of spoken vernaculars

29

ent clause with V-last cannot be processed as readily. On the other hand, what the far-out occurrence of the finite V in the dependent clause ("Vfinal") requires is a mechanism pre-processing the predicate with respect to its valency bound clausal. Such pre-processing will inevitably imply alternatives as to the eventual outcomes - something which is a highly uneconomical impediment to the process of grammatical understanding. If we assume, however, that grammatical processing of the clause came to be but one of the analytic on-line processes of oral speech, next to the discourse-functional analysis co-considering also its contextual embedding, the emergence of the predicate periphrasis becomes a natural goal of the linguistic change. Notice, in the first place, that it had been introduced in the first pklace by way of "stranding" verbal particles and affixoids systematically. Second, and most important, predicate bracketing (or, more analytically, retaining the structural positions of C/I/T as well as V°, in terms of Universal Grammar) implemented distinguishing rhematic vs. thematic material in the sentence: the rheme stays in the right, rear, end of the clause, while the theme moves to the left/front. This meets a context requirement of recognizability: what needs to be taken up in terms of anaphoricity needs to be available sooner; and what is introduced as supposedly new may come later in the on-line order of the lexical and grammatical elements. This is not (ideally) realized in the pure V-2 structures (i.e. structures that constrain the position of the finite predicate to one early position, leaving the late and rheme-identifying period unidentified). This is one explanative option of the facts under inspection, one that, as we have seen, carries over to an traditional set of questions around the "Upper German decay of the preterite forms". The other option we considered one is part of what we called the "prosodic optimalization" - but one which may also have to do with the emerging ubiquitous periphrasis of the finite predicate: the clause-phonological condition of coda formation subject to the following optimality scale: realize the (unstressed) coda directly after the core element. In single-member predicates of dependent clauses, the core would be the constituent directly in front of the finite predicate. This core complex carries the clausal main accent by default. Finite V-last is but an "after-beat" in the clausal melody, in the sense of a coda ("tail"). Nondependent "V-2 only" clauses are not ideal in this respect since they end on what is always the stressed, phonological-lexical core component - unless also non-dependent clauses create the possibility to identify a coda element subsequent to the rhematic core in default stress position. Dependent clauses, with V-last, meet this prosodic principle ideally. This is behind the unrelent-

30

Werner Abraham

ing preservation of V-last even in languages that have given up most of their inflection such as Dutch and Frisian.

10. Conclusion There is at least one more alternative to this "natural" explanative "orality scenario" as discussed above hardly less natural, but arguing, as it were, in reverse: the optimalization of the parsing process. Under the parsing view, the early occurrence of the predicate is highly preferred in that it is the predicate on the basis of which all valency elements are identified (with respect to their theta roles, with respect to morphological case, with respect to subject- and object predicate agreement, to say the least, with an eye on Indo-European languages only). However, the fact that the majority of languages in the world has the subject before the finite predicate ("sv"), and that it is not "vs", shows perhaps that the strategy optimalizing the Theme is preferred over optimalization under parsing conditions. We would like to believe we have shown beyond doubt is that tun as an Aux in substandard German serves a crucial sentence-grammatical purpose deviating, in likewise crucial terms, from the (written) tradition of Standard German. It was assumed that the use of the Aux tun is to be seen against the wider background of creating prominent positions for the identification of discourse functions such as theme and rheme. We have shown that the range of acceptance of the distribution of tun is subject to exactly this discoursefunctional type of grammar organization. We have furthermore placed the occurrence of iwn-support in its appropriate regiolectal distribution. Low German (as well as Dutch and Frisian) have not undergone the decay of the formal (synthetic preterite), despite an even heavier decay of the inflective paradigms than German (see my arguments with respect to "necessary" and "satisfactory" arguments!). It is therefore plausible that it provides no double perfect forms; likewise, there is no preference for the perfect in the formation of the periphrasis of the predicate (as confirmed by Eroms 1984 for Low German). Finally, we had to leave open the question whether the syntax of tun as an Aux is solely a question of discourse functional terms or of clause prosodic conditions (syntax vs. phonology). But even so our findings appear to amount to an adequate progress in dealing with an old phenomenon.

Characteristics of spoken vernaculars

31

Notes 1. I will disregard completely such well-known literature on the topic as Biber (1986), Chafe & Danielewitz (1987), Chafe & Tannen (1987) or Ochs/Schegloff/Thompson (1996) for the simple reason that the relation between the two expressing media, oral and written, are of a specific kind in German not dealt with at all in the literature just mentioned. Furthermore, my methodological approach crucially heeds Ockham's razor principle in the sense that an (unsystematic) pragmatics account is the last instance looked for at the attempt to describe and explain linguistic phenomena, inevitable for the very reason only that all other levels of (more systematic) description such as Syntax, morphology, phonology, prosody and stress yield no satisfactory result. 2. The reader will find my reference to Mark Twain more convincing after the following quote from the Financial Times about the present heated discussion in the media about the acceptance of the new orthographical reform of German (FAZ Aug. 14, 1997, p. 2): "Für ein Land, das stolz auf Methode ist, ist es recht außergewöhnlich, wie häufig Deutschland sich in Wirren verfangt. Das letzte Durcheinander ist dafür ein Modell. Es sieht soaus, als ob das ganze Land in heillose Verwirrung gestürzt ist und das in einem Bereich, in dem Klarheit wesentlich ist: der Sprache. Natürlich hat die deutsche Sprache niemals von großer Präzision gezeugt, überlastet wie sie ist mit zusammengesetzten Wörtern und grammatikalischem Papierkrieg, während die Hälfte der Verben am Ende unendlicher Sätze schmachtet. ...". [my highlighting; W.A.] 3. Chapter 3 has profited greatly by discussions with Annette Fischer, Humboldt-University of Berlin. The bibliographical references in her name reveal beyond doubt how much I owe her in terms of basic data. 4. This is not the result A. Fischer has arrived at on the basis of her own observations (p.c. of 19 Dec. 1997). Notice, however, that she did not include a comparison of the material from Upper and Low German. Based on Modern Dutch, my hypothesis applies without restrictions. What I cannot exclude, however, is that the total picture may be different once Dutch dialects are integrated into the picture.

References Abraham, Wemer 1994 "Ergativa sind Perfektiva." Zeitschrift für Sprachwissenschaft 12: 157-184. 1995a Deutsche Syntax im Sprachenvergleich. Tübingen: G. Narr. 1995b "Structural properties of information packaging in German and in Universal Grammar." Travaux du cercle Linguistique de Prague. NS 1. Amsterdam: J. Benjamins, 125-156. 1997 "The aspectual source of the epistemic-root distinction of modal verbs in German." Sprache in Raum und Zeit. Beiträge zur empirischen Sprachwissenschaft (Gedenkschrift Johannes Bechert). In: Winfried Boeder, Christoph Schroeder, Karl Heinz Wagner, Wolfgang Wildgen (eds.). Vol. 2. Tübingen: G. Narr, 231250.

32

Werner Abraham

1999

"Negation in the diachrony of Germanic and Jespersen's Cycle." In: Gerald F. Carr (ed.), Interdigitations. Essays for Irmengard Rauch. Frankfurt/M.-Bern: P. Lang, 1-9. Bayer, Josef (ed.) 1995 Parsing. [Special volume of Folia Linguistica Europaea]. 1995 Biber, Douglas 1986 "Spoken and written textual dimensions". Language 62: 384-414. 1988 Variations across speech and writing. Cambridge: CUP. Bruggmann, Karl & Berthold Delbrück 1886-1900 Grundriss der vergleichenden Grammatik der indogermanischen Sprachen: kurzgefasste Darstellung der Geschichte des Altindischen, Altiranischen. Strassburg: Trübner Chafe, Wallace & Jane Danielewicz 1987 "Properties of spoken and written language". In: Rosalind Horowitz & S. Jay Samuels (eds.) Comprehending oral and written language. San Diego: Academic Press, 83-113. Chafe, Wallace & Deborah Tannen 1987 "Relation between written and spoken language", Annual Reviews of Anthropology 16: 383-407. Chomsky, Noam 1995 Government and binding. Dordrecht: Foris. Cinque, Gulielmo 1993 "A null theory of phrasal and compound stress", Linguistic Inquiry 24: 239-298. Ellegärd, Alvar 1953 The auxiliary do.· The establishment and regulation of its use in English. [= Gothenburg Studies in English, n]. Göteborg: Wikseil. Erben, Johannes 1969 "tun als Hilfsverb im heutigen Deutsch". In: Festschrift für Hugo Moser, edited by Ulrich Engel, Paul Grebe und Hans Rupp. Düsseldorf: Schwann, 46-52. Eroms, Hans-Werner 1984 "Indikativische periphrastische Formen mit doa im Bairischen als Beispiel für latente und virulente syntaktische Regeln.' In: Peter Wiesinger (ed.) Beiträge zur bairischen und ostfränkischen Dialektologie. Göppingen: H. Kümmerle, 123-135. Fischer, Annette 1996a fon-periphrasis in Early New High German. Paper read at University of Leiden 15 March 1996. Humboldt-University Berlin. 1996b Diachronie und Synchronic von auxiliarem tun im Deutschen. Paper read at Trinity College Dublin Mai 1996. Humboldt-University Berlin. 1996c Zur /««-Periphrase im Deutschen. Handout. Paper read as "Diachronische Adäquatheitsbedingungen an Grammatiktheorien/Diachronic conditions of adequacy in theories of grammar", ZAS Berlin, 12-13.12.1996. Hawkins, John 1995 A performance theory of order and constituency. Cambridge: CUP. Hausmann, Robert 1974 "The origin and development of Modem English periphrastic do". In: John M. Anderson & C. Jones (eds.) Historical linguistics I. Amsterdam: 159-189.

Characteristics

of spoken vernaculars

33

Hoekstra, Eric 1996 "Syntaxis vs. parataxis". ΤΑβυ-dag University Groningen, handout. Lasnik, Howard 1998 "On the locality of movement: Formalist syntax position paper." In: Michael Darnell, Edith Moravcsik, Frederick Newmeyer, Michael Noonan, Kathleen Wheatley (eds.) Functionalism and Formalism in Linguistics. Vol. 1: General Papers. Amsterdam: John Benjamins, 33-54. Leiss, Elisabeth 1985 "Zur Entstehung des neuhochdeutschen analytischen Futurs." Sprachwissenschaft 10 (3/4), 250-273. Lightfoot, David W. 1979 Principles of diachronic syntax. Cambridge: CUP. Lindgren, Kai B. 1963 "Über Präteritum und Konjunktiv im Oberdeutschen". Neuphilologische Mitteilungen 64: 264-283. Litvinov, Viktor P. & Vladimir I. Rad(enko 1997 Doppelte Perfektformen im Deutschen. Tübingen: Stauffenburg Verlag. Ochs, Eli; Mani Schegloff; & Sandy Thompson (eds.) 1997 Interaction and grammar. Cambridge: CUP. Patocka, Franz 1997 Satzgliedstellung in den bairischen Dialekten Österreichs. Frankfurt etc.: Peter Lang. [Schriften zur deutschen Sprache in Österreich; 20] Pritchet, Bradley 1991 "Subjacency in a principle-based pareser". In: Robert C. Berwick et al. (eds.) Principle-based Parsing: Computation and Psycholinguistics. Dordrecht: Kluwer, 301-345. Rohdenburg, Günther 1986 "Phonologisch und morphologisch bedingte Variation in der Verbalsyntax des Nordniederdeutschen". Niederdeutsches Jahrbuch 109, 86-117. Schmeller, Johannes Andreas 1872 Bayerisches Wörterbuch. Band 1. Ahlen 3 1973. Valldüvi, E. 1995 "Structural properties of information packaging in Catalan."In: Katalin έ . Kiss (ed.) Discourse conflgurational languages. Oxford: Oxford University Press. Wiltschko, Martina 1996 "On the syntax and semantics of (relative) pronouns and determiners". The Journal of Comparative Germanic Linguistics 2: 143-181. Zirmunskij, Viktor M. 1962 Deutsche Mundartkunde. Vergleichende Laut- und Formenlehre der deutschen Mundarten. Berlin: Akademie Verlag.

Language variation and phonological theory: inflected adjectives in Dutch and related languages Geert Booij

1. Introduction It is obvious that there is a intimate relation between dialectology and phonological theory. Since each dialect is a language, dialects are as important as standard languages as sources of data for phonologists, and thus dialects form a very extensive testing ground for phonological theory. On the other hand, phonological theory might be of use in achieving our aim of providing accurate and interesting descriptions of dialects. Moreover, the study of dialects is intimately connected to the study of language variation and language change. The issue of how to characterize the phonological differences between related languages or dialects, or between successive historical stages of a language in terms of formal differences between the grammars of the variants involved has always been part of the research tradition of generative phonology.1 In the sra-style of generative phonology, dialects may differ at least with respect to: -

the inventory of underlying segments the underlying forms for related words the set of phonotactic constraints the set of phonological rules the order of shared phonological rules

In the seventies, it was Kiparsky who argued that the differences between two related dialects can sometimes be analysed as differences in rule ordering: the dialects have the same underlying forms, but the phonetic realizations of the relevant words differ because certain phonological rules apply in different orders, thus leading to different phonetic outputs (cf. also Taeldeman 1980). Kiparsky (1968) [1982: 19ff], for instance, characterized the differences between two Swiss German dialects in terms of different orderings for the phonological rules of umlaut and lowering.

36

Geert Booij

This example is not only interesting because it shows how we might formally characterize dialectal differences, but it is also of direct relevance for phonological theory: if this is the correct characterization of these dialectal differences, it implies that extrinsic ordering of rules should be allowed for in grammars of natural languages. Thus, trying to account for language variation may have implications for our theory of the formal properties of grammars. In recent years, the theory of phonology has changed considerably in that the role of output constraints has become much more prominent in phonological analyses. At present, Optimality Theory (McCarthy & Prince 1993, 1994; Prince & Smolensky 1993) is the dominant output-based theory of phonology. In this paper, I will argue that an output-based approach has a number of advantages above a theory which derives phonetic outputs by applying rules to underlying forms. First, I will first demonstrate how this approach can account for prosodically conditioned suffix selection in the realm of Dutch pluralization of nouns (Section 2). In Section 3,1 will give an outline of the systematics in the phonological form of inflected adjectives in Dutch. Then, I will show how the differences in schwa-apocope in inflected adjectives between present-day Standard Dutch and a number of related Germanic languages can be insightfully accounted for in a model that allows for output-based affix selection (Section 4). This will lead to the conclusions that a phonology that makes use of output constraints offers interesting perspectives for the analysis of language variation, and that a rather concrete nonphonological analysis of certain allomorphy patterns is called for (Section 5).

2. Pluralization of nouns in Dutch In Optimality Theory (οτ), the computation of the correct phonetic form of words and phrases is not executed by applying rules to underlying forms, i.e. in terms of a derivation of the phonetic form in a number of steps. Instead, the phonology of a language is conceived of as a language-specific ranking of a set of universal but violable phonological constraints, GEN, the generator, creates a set of possible output candidates for each underlying form, and that candidate is chosen that is most harmonic, i.e. incurs the least severe violations of the set of constraints. A violation is more severe if it violates a higher ranked constraint:

Language variation and phonological theory

/underlying form/ Candidate 1

Condition A

37

Condition Β

*! *

«a- Candidate 2

The asterisk indicates violation, the exclamation mark means that a violation is fatal, and that the candidate involved is the loser. The pointing finger indicates the winning candidate. Let me illustrate this with a very simple example, the phenomenon of syllable-final devoicing of obstruents in Dutch. This phenomenon can be accounted for by the following universal (cf. Stampe 1969), but violable output constraint; (1) Final Devoicing: obstruents in coda position are voiceless This condition competes with the Faithfulness condition that the underlying form of each morpheme is completely realized on the surface. In Dutch, FinalDevoicing is ranked higher than Faithfulness, and thus obstruents in coda position will surface as voiceless, as shown here for the item land [l(nt] 'id.'. In English, on the other hand, the working of the universal constraint FinalDevoicing is made invisible because it is ranked lower than Faithfulness. Thus, in English the final /d/ of land will surface as a [d]. /land/ land

Final Devoicing *!

*

lant

English

/te:nd/ "3s laend laent

Faithfulness

Faithfulness

Final Devoicing

*! *

38

GeertBooij

This example also serves to illustrate that in Optimality Theory differences between languages can be characterized partially in terms of different rankings of the same constraints. This example does not, however, show the superiority of output-based phonology as opposed to the classical derivational model of phonology. Numerous arguments in favor of οτ have been adduced. Here, I will concentrate on one particular advantage, which is crucial for the particular analysis of language variation that I will present in this paper: οτ enables us to give an insightful account of prosodically governed allomorphy and affix selection. Dutch has two plural suffixes for nouns, -s and -en (the e stands for the schwa). The basic selection pattern can be formulated as follows: (2) -s after an unstressed syllable, -en after a stressed syllable. This is illustrated by the following examples (3) kanon 'canon' kanon 'gun' nätie 'nation' genie 'genius'

- kanon-s - kanonn-en - natie-s - genie-en

This selection principle also correctly predicts that monosyllabic nouns have -en as their plural suffix: (4) non 'nun' knie 'knee' bal 'ball'

- nonnen - knieen - ballen

Note that it is obvious that we cannot derive the plural suffixes from a common underlying form by means of general phonological rules. Nevertheless, as is the case in a number of languages (cf. Carstairs 1988), the distribution of competing affixes and allomorphs may be governed by purely phonological principles. The formulation of the selection principle given above is formulated in terms of properties of the input forms for the pluralization process, the nominal stems, which are formally identical to the singular forms. However, such a formulation does not explain why this particular selection principle holds. In terms of complexity of the grammar, it would make no difference if Dutch were just the other way round, that is, if -s occurred after stressed syllables,

Language variation and phonological theory

39

and -en after unstressed ones. The motivation behind this particular suffix selection is that plural nouns will always end in a disyllabic trochee. The disyllabic trochee is the preferred pattern of organization of syllables into higher prosodic units in Germanic languages. We will therefore assume the following violable constraints: (5) Trochee: in a foot, the first syllable is the head FootMax: feet are maximally disyllabic ParseSyllable: syllables are parsed into feet FootMin: feet are minimally disyllabic Let us assume that the ranking of these constraints is as follows: (6) Trochee » FootMax » ParseSyllable » FootMin The conditions ParseSyllable and FootMax imply that a sequence of two syllables both of which are headed by the vowel schwa is not optimal. In Germanic languages, syllables headed by schwa, the default vowel, cannot bear stress, i.e., they cannot head a foot. Therefore, a sequence of two schwasyllables implies that either the second of these is not made part of a foot, or that we have to create a ternary foot. In the first case, ParseSyllable is violated, in the second case FootMax. Since FootMax has the higher ranking in (6), the second of two schwa-syllables will not be parsed into a foot, and hence will be dominated directly by the prosodic word node. However, this particular ranking is not crucial for the claims made in this paper. These constraints also explain why Dutch lexical morphemes never contain a sequence of two schwa-syllables. Such sequences only arise in complex words, due to morphological operations. If we added -s to a word ending in a stressed syllable, we would not create optimal, i.e., disyllabic feet, i.e., FootMin would be violated. On the other hand, if we added -en after an unstressed syllable, the newly created syllable cannot be parsed into a binary foot. Note, however, that if we add /s/ to an unstressed syllable, we create superheavy ( w c or vcc) syllables which form the weak constituent of a trochee, which is also a not-preferred configuration. I will call this constraint Weight: (7) Weight: superheavy syllables do not occur in weak position

40

Geert Booij

The following tableaux show how the selection of the correct plural suffix takes place. Note that we now extend the notion 'candidate' because the set of candidates is not defined exclusively phonologically, but also morphologically: each combination of stem + allomorph is the basis for a set of candidates. kanon+PL

FootMax

ParseSyll

(ka:non)F-en

ka(nons)F ka(nonnen)F

Weight

*

(ka:nons)F

kanon+PL

FootMin

*!

FootMax

ParseSyll

FootMin

*

*!

Weight

*

The constraint FootMin is obviously violated by monosyllabic nouns in their singular forms, and also in plural forms of loanwords which keep the original suffix: (8) English loan: tram - trams 'trams' French loan: paraplü - paraplu's 'umbrellas' Interestingly, Dutch children tend to regularize these patterns, and create plural nouns such as trammen 'trams'. Also, we find the following differences between southern Dutch and northern Dutch: (9) test 'test' expert 'expert'

southern Dutch plural form testen experten

northern Dutch plural form tests experts

The output-based approach is not only superior in that it formally expresses the 'why' behind the selection principle, it also makes the prediction that nouns ending in schwa have both the endings -s and -en because of the effect

Language variation and phonological theory

41

of prevocalic schwa-deletion (a schwa deletes before an adjacent vowel within the same prosodic word), which ensures that the addition of -en after a noun ending in schwa does not create a new syllable. This prediction is correct (the stem-final e stands for schwa): (10) kade 'quay' bode 'messenger' lade 'drawer' methode 'method'

kades, kaden bodes, boden lades, laden methodes, methoden

In some cases, one of them may be preferred, but this is a matter of convention. The kind of variation we meet with here is system-internal variation, i.e., real optionality. Complex words ending in a schwa-final suffix also have the two options, as is to be expected on the basis of the phonological properties, unless the suffix requires a particular suffix: (11) both suffixes: -ade marinades 'marinades', balustraden 'balconies' -se extases / extasen 'extasies', hypotheses / hypothesen 'hyptheses' -de liefdes 'loves', kundes / kunden 'arts' -te gedeeltes / gedeelten 'parts', koeltes / koelten 'coolnessses' -isme modemismes / modernismen 'modernisms', mechanismes/mechanismen 'mechanisms' -ine sonatines / sonatinen 'sonatines' -ide fluorides / fluoriden 'fluorides' only s (words with non-native suffixes borrowed from French preserve the French plural form)'. -aire documentaire-s 'documentaries' -age massage-s 'massages' -asme orgasme-s 'orgasms' -esse secretaresse-s 'secretaries, fem.' -iere cabaretiere-s 'cabaretiers, fem.' -euse masseuse-s 'massager, fem.' -ice directrice-s 'directors, fem.' -ette brunette-s 'brunettes', wasserette-s 'laundries' -ure doublure-s 'doublets'

42

Geert Booij

native suffixes -e acrobate-s 'acrobats, fem.' -(t)je toetje-s 'desserts', kindje-s 'small children' only -(e)n -e blonde 'blond'- blonden 'blonds' (nominalizing deadjectival suffix) In conclusion: morphology may override phonology. The selection of a particular plural suffix may be governed by morphological principles, as shown by the fact that the last suffix of a derived word may determine the choice of plural suffix. That is, the choice of a particular suffix cannot always be reduced to phonological output conditions. The output-based analysis of affix selection makes it also possible to give a proper account of the interaction between stem allomorphy and selection of plural suffix. A number of nouns in -or, -on, -ol have two plural forms: (12) motor /mo:tor/ 'engine'

motors /mo:tors/, motoren /mo:to.T9n/ professor /pro:fesor/ 'professor' professors /pro:fesors/, professoren /pro:feso:rsn/ neutron /noeytron/ 'neutron' neutrons /noeytrons/, neutronen /nceytro:n9n/

The conspiracy of stem allomorphy, stress location (weight-sensitive assignment of primary stress) and choice of plural suffix follows directly from the output constraint. Note that the last syllables of the long stem allomorphs form superheavy syllables, and therefore attract the main stress. A similar generalization concerning the form of plural nouns in terms of the required output can be made with respect to German (leaving aside the special category of plural nouns ending in -s), as pointed out by Wiese (1996: 106): "plural nouns in Modern Standard German must end in a bisyllabic foot, with the second syllable being a schwa syllable" as shown by the following examples: (13) Singular Filter 'filter' Mauer 'wall' Uhr 'watch'

Plural Filter-0 Mauer-n Uhr-en

Language variation and phonological theory

43

The correct prediction of this output condition is that zero-marking of the plural is limited to nouns that end in a schwa-syllable, and that the ending -en only occurs after syllables headed by a full vowel. Now that we have seen some arguments for making use of output constraints, we will now turn to the analysis of the variation in the phonological form of inflected adjectives in Dutch.

3. Inflected adjectives in Dutch In present-day standard Dutch adjectives in attributive position are inflected: the stem is followed by schwa, except when the adjective forms part of an indefinite NP headed by a singular neuter noun. In the latter case, there is no overt inflection. Traditionally, these forms are called uninfected adjectives. In the examples below, boek 'book' is a neuter noun, and tafel 'table' is a non-neuter noun. (14) een goed-0 boek 'a good book' het goed-e boek 'the good book' (de) goed-e boeken '(the) good books' een goed-e tafel 'a good table' de goed-e tafel 'the good table' (de) goed-e tafels '(the) good tables' When the stem of the adjective ends in -en, however, the schwa does not appear: (15) de open deur 'the open door' de houten tafel 'the wooden table' de gestagen room 'the whipped cream' de vergeten jas 'the forgotten coat' The absence of the schwa in this position does not follow from a purely phonological rule of schwa apocope, because there is only absence of schwa if the schwa is the inflectional morpheme of the adjective. In other cases, the schwa remains, e.g. in the nominalized forms of adjectives:

44

Geert Booij

(16) het geslagen-e 'the beaten' (neuter) de geslagen-e 'the beaten (person)' (non-neuter) Moreover, it is only after -en that the inflectional schwa does not surface; it does surface after other consonants, and after the combination full vowel + /n/: (17) de mager-e vrouw 'the lean woman' de lelijk-e man 'the ugly man' de edel-e mens 'the noble man' That it is the output configuration -ene only that is to be avoided, can be concluded from the fact that past participles used as attributive adjectives do have a final inflectional schwa if they have an exceptional form ending in full vowel + /n/, which is the form for the verbs doen 'do', gaan 'go', staan 'stand', zien 'see' and their derivatives: (18) ge-dan-e zaken 'lit. done business, finished matters' door-stan-e moeite 'lit. suffered effort, efforts' be-gan-e wegen 'trodden paths' een ge-zien-e collega 'lit. a seen colleague, a respected colleague' Furthermore, the same pattern can be observed for infinitives with te used in attributive position. After infinitives in -en, the inflectional schwa does not occur, whereas the five exceptional infinitives for the verbs doen, gaan, slaan and zien, mentioned above, and the verb slaan 'strike' do have the schwa: (19) nog te nemen-0 maatregelen 'lit. still to take measures' versus nog te doen-e zaken 'matters still to be done' nog te gan-e wegen 'ways that must be gone' nog te door-stan-e beproevingen 'trials still to be suffered' niet af te slan-e voorstellen 'proposals not to be rejected' niet te zien-e Organismen 'organisms that cannot be seen' Summarizing: the inflectional schwa of words used as attributive adjectives has a zero-allomorph if the stem ends in -en.

Language variation and phonological theory

45

The avoidance of the sequence -ene can be related to the prosodic structure of Dutch and related Germanic languages: as pointed out above, syllables are organized in disyllabic trochaic feet. A syllable headed by the vowel schwa cannot function as the head of a foot, and therefore, a sequence of two of such schwa-syllables should be avoided if possible, as was also pointed out in Dyk (1996), as required by the constraints Parse and FootMax: the second of the schwa-headed syllables will be left over, and cannot be parsed into a foot. A form like opene 'open' violates this constraint, whereas open is in accordance with the constraint. Therefore, the form open should be selected by the grammar as the proper form of the inflected adjective. In a rule-based analysis of this kind of schwa apocope, we would have to assume the following rule: (20) 9

0 /a η ~] A

Such a description fails to express, however, why it is only in this context, after a preceding schwa, that the final schwa deletes. A rule with the context '[+voc] η --', i.e., a rule that deletes schwa after In/ whatever the preceding vowel, would formally be simpler. Yet, it is obvious that it would not be simpler from the point of view of the overall systematics of the grammar. The constraint ParseSyllable can be violated since we do find words with a sequence of two schwa-syllables, in the following situation: (21) (a) with another consonant than /n/, e.g. edel-e mens 'noble man' (b) in the inflected present participle: lopend-e mensen 'walking people' (c) if the schwa is a nominalizing suffix: het open-e 'the open, neuter', de ge-bor-en-e 'the born person'. Violable constraints are an essential ingredient of οτ. In this case, the constraint Parse is dominated by the principle that morphemes must receive a phonetic realization, i.e. Faithfulness is more important than Parse:

46

Geert Booij

(22) hetopen-e Faithfulness "Sf (ope)pne (open)F

ParseSyllable *

*!

(the part of a string between < > will not be realised). The question thus arises how we account for the difference between the behaviour of the inflectional schwa of adjectives, which alternates with zero, and the other inflectional schwas which always appear on the surface. Such differences in behaviour could be accounted for in terms of different constraint rankings. However, I do not want to assume suffix-specific rankings of constraints, because this would imply that the language learner would have to acquire a substantial number of different grammars for his native language. Therefore, the only reasonable account for the non-overtness of the inflectional schwa of attributive adjectives is that it has a zero-allomorph 0 which is selected by the same grammar:

Faithfulness (open-e)F e

ParseSyllable *!

(open- 0)F The second form with the allomorph 0 does not violate ParseSyllable, and is therefore to be preferred. The inflectional schwa of adjectives must therefore be assumed to have a zero-allomorph which is subcategorized for appearing after /n/. Note that we do not have to subcategorize this zero-allomorph for appearing only after /n/ preceded by a schwa, because this follows from the constraint FootBinarity already discussed in the preceding section. That is, it follows automatically that the zero-allomorph does not appear after /n/ preceded by a full vowel, as is the case for an adjective like groen 'green':

Language variation and phonological theory

47

FootMin «s* (groene)F (groen- 0)F

*!

In other words, if Dutch has the option to create a disyllabic word, it will do so, and not choose the zero-allomorph which results in a less optimal phonetic form. There are also a few adjectives, ending in -er, that also have the zero-form of the inflectional ending: denominal geographical adjectives such as Limburger and Groninger, and the adjectives linker 'left' and rechter 'right'. In other words, the zero-allomorph of the adjectival inflection must also be specified as to appear after such adjectives. This zero-allomorph will then be selected by the constraint system proposed above, because forms like * linkere will be less optimal. However, this zero-allomorph is not allowed for all adjectives in -er. comparatives do have the schwa. This implies that the specification of the context of the zero-allomorph has to refer to specific classes of adjectives. The working of the constraint ParseSyllable is also visible in another case of prosodically conditioned allomorphy: denominal and deverbal nouns exhibit suffix allomorphy, the allomorphs -er, -der, and -aar: (23) denominal nouns geographic noun Amsterdam

Almere Diemen

inhabitant name Amsterdammer / * Amsterdam-der / *Amsterdamm-aar *Bijlmermeer-er / Bijlmermeer-der / *Bijlmermeeraar *Almeer-er / Almeer-der / *Almer-aar *Diemen-er / Diemen-aar / *Diemen-der

deverbal nouns luister 'listen' etter 'nag' leer 'learn'

allomorphs -er, -der, -aar *luisterer / *luisterder / luisteraar *etterer / *etterder / etteraar *lerer / leerder / leraar

Bijlmermeer

48

Geert Booij

The choice of -der instead of -er has to do with the avoidance of the sequence /rar/, which is a manifestation of the general constraint *C.aC., and discussed in detail in Booij (1998). The choice between -aar versus -er or -der is a matter of optimal foot structure. A superheavy syllable ( w c or vcc) in the weak position of a foot is very marked, though not impossible, as in the exceptional word zond-aar 'sinner'. Therefore, -aar will be avoided after a syllable that bears stress, such as Bijlmermeer with the denominal noun Bijlmermeerder. The correct candidate is selected as follows:

FootMax (dieme)Fner

ParseSyllable

FootMin

Weight

*! *

"S" (dieme)F(naar)F

The allomorph -aar is also a special allomorph, subcategorized for appearing after a coronal consonant. Again, the fact that this coronal consonant is always preceded by a schwa, does not have to be stated, but follows from the constraints, in this case Weight discussed above (a weak syllable must be light). Since the suffix allomorph -aar creates a heavy syllable (long vowel followed by a consonant), it will not occur after a stem-final syllable with a full vowel. For instance, Amsterdammaar is out because it would have a superheavy syllable in weak position:

(am

ster)

(dam

ma:r)

(am

ster)

(dam

mer)

Thus, we see again how prosodically conditioned allomorph or affix selection can be accounted for in an insightful way in terms of violable and ranked output constraints.

Language variation and phonological theory

49

4. Language variation As pointed out in the introduction, an adequate theory of phonology should enable us to give an insightful account of differences between related languages or dialects. In this section I will discuss how inter- and intra-dialectal variation in the phonological form of the inflected adjective can be analysed. In Frisian, the second official language of the Netherlands, the omission of a final schwa in the inflected adjective is optional (Dyk 1996), whereas in Dutch the form without schwa is obligatory. In Middle Dutch, the inflectional schwa of the attributive adjective always surfaced. Similarly, in Standard Modern German inflected adjectives always end in a schwa-syllable. How can these differences between related languages be interpreted? The difference between Frisian and Dutch can be seen as a manifestation of the perennial tension between two forces that determine the phonetic realization of a morphologically complex word: on the one hand, each morpheme requires a faithful realization of its underlying form in order to express the relevant information, on the other hand the constraints on phonologically optimal outputs may require that morphemes are not completely realized. In the case of Dutch, it is the phonological output constraint Parse that is complied with, but only in the case of contextual inflection, where a zero-allomorph does no harm. In Middle Dutch, there always was an overt inflection for attributive adjectives. Gradually, the schwa became optional in this context (Raidt 1968: 29-30). Frisian is thus in between Middle Dutch (and German) and Modern Dutch in that it optionally omits the final schwa of inflected adjectives. In the theoretical framework of οτ, some differences between two related languages can be described as ranking differences between constraints. Variation within a language can be interpreted as a situation in which two constraints are not ranked with respect to each other. Hence, there are two possible winning candidates (Löhken 1997). Phonological change can then be depicted as follows: (24) Stage 1 Stage 2, variation Stage 3

Cond A » Cond Β Cond Α, Cond Β Cond Β » Cond A

An example of such an analysis can be found in Golston & Wiese (1996: 155). They argue that the difference between Modern Standard German and the Hessian dialect of German with respect to pluralization of nouns can be

50

Geert Booij

described as a difference in ranking of some constraints. For instance, the form for 'dogs' is Hunde is Standard German, but hon in Hessian. Hessian thus exhibits subtractive morphology: the plural is formed by dropping the final /d/. The crucial difference is that in Standard German the constraint Fill, which forbids epenthesis is ranked below the constraint Parse Segment that forbids deletion of the /d/; hence a schwa is inserted in order to comply with the output condition that a plural noun must end in a sonorant segment. In Hessian, the ranking of these two constraints is the inverse one; hence Hessian prefers deletion of the /d/ to insertion of the default vowel schwa, in order to comply with the output condition. For the phenomenon under discussion here, the form of inflected adjectives, the constraint ranking would be as follows: (25) Middle Dutch,German Frisian Modern Dutch

Faithfulness » ParseSyllable Faithfulness, ParseSyllable ParseSyllable » Faithfulness

This model does not suffice, however, for the case under discussion, because it is only in the case of the inflectional schwa of inflected adjectives that ParseSyllable appears to dominate Faithfulness; in all other cases the schwa remains, as discussed above, i.e., Faitfulness dominates ParseSyllable. Therefore, the differences between these three related languages must be described in terms of allomorphy: Modern Dutch has an obligatory zero-allomorph for the inflectional schwa of adjectives, and Frisian has an optional zeroallomorph for this suffix. In other words, this kind of language variation is not a matter of differences in constraint rankings, but in the availability of specific underlying forms. The Dutch zero-allomorph must be listed as such, and cannot be derived by phonology. Given its possibility of occurrence after /n/, this zero-allomorph will be preferred to the schwa-allomorph, because it violates neither Faithfulness nor ParseSyllable. In the case of material adjectives, Frisian has another means to avoid a sequence of two schwa-syllables (Dyk 1996: 57): instead of the material adjective suffix -en, another suffix is available, namely -s. This is the suffix used after nouns ending in el, en and je: (26) linnen 'linen' duffel 'duffel' flenje 'flanel'

linnen-s duffel-s flenje-s

Language variation and phonological theory

51

The question then arises under which conditions zero-allomorphy is allowed for. The important point to be noted here, and also observed by Dyk (1996) is that the kind of inflection involved is what has been called weak (Kiparsky 1982) or contextual (Booij 1994) inflection: is is inflection that is completely dependent on and governed by the syntactic context, and has no independent information content. Therefore, it is omissble. As observed by Kiparsky (1971) [1982: 67], it is contextual inflection that gets lost first: "Morphological material which is predictable on the surface tends to be more susceptible to loss than morphological material which is not predictable on the surface". A number of illustrations of this principle can be found in Booij (1994). And indeed, as we saw above, schwas that have a non-predictable function, e.g. the function of nominalizer (changing an adjective into a neuter or a non-neuter noun) always surface, and do not have a zero-allomorph. In Frisian, where the schwa-apocope is optional, the zero-allomorph after /n/ is optional, and not an obligatory allomorph. The optionality of the allomorph suggests that phonology has not yet completely won the victory over morphology. In the classical rule-based framework we would say that Frisian has an optional rule of inflectional schwa-deletion, and this is in conformity with the following generalization (Kiparsky 1982: 68): "Grammatically conditioned variability in the application of optional rules favours optimal outputs". It is interesting to see how in the formulation of this generalization the notion 'output' already plays a direct role. What remains something of a mystery is why it is only after the /n/ that there is a zero-allomorph. Why do not we choose for later instead of latere in a phrase such as: (27) de latere ontwikkelingen 'the later developments' Similarly, the allomorph -aar of the suffix -er can only be used afer coronal consonants, so that we have bezem-er 'sweeper' besides reken-aar 'computer': what is wrong with the incorrect bezemaarl We do see tendencies, however, to also avoid sequences of schwa-syllables in Dutch inflected complex adjectives of the following types, with the suffixes -elijk /alak/ and -ig /ay/:

52

Geert Booij

(28) het onvermijdelijk(e) gevolg 'the unavoidable consequence' ons hartelijk(e) weerzien 'our cordial meeting again' gebruikelijker(e) gevallen 'more usual cases' een voorzichtiger(e) formulering 'a more prudent formulation' This optionality implies that the possibility of a schwa-less allomorph for certain adjectives in attributive position must be specified in the lexicon by listing this allomorph. Related languages can thus be put on a scale of output simplification with respect to the constraint ParseSyllable; ParseSyllable is violated less frequently in modern Dutch due to the introduction of a zero-allomorph: (29) Middle Dutch Frisian —> output simplification

Dutch

A parallel development can be seen in the form of inflected infinitives. In Middle Dutch, infinitives could be inflected: the preposition te assigned dative case, expressed by the ending -e: (30) te copene te kerene te vellene

'to buy' 'to turn' 'to fell'

In Modern Dutch, this inflectional ending is no longer found, nor in Frisian.2 Afrikaans introduced the most radical change in the inflection of adjectives (Raidt 1978, 1983) in that the final schwa of adjectives completely lost its morphological status. The schwa became part of the underlying form of the adjective instead, so an Afrikaans adjective may have an attributive form that ends in schwa, but the occurrence of this attributive allomorph does not depend on morphosyntactic properties such as number, gender and definiteness: in this attributive position, the adjective either always ends in schwa, or never. Thus, for a number of adjectives one has to learn that the attributive form is the predicative form plus schwa. This applies first of all to all derived adjectives, i.e. adjectives that end in an adjectivizing suffix (Raidt 1983: 142), for instance:

Language variation and phonological theory

(31) -aal -agtig -erig -end

53

provinsi-ale 'provincial' boek-agtige 'bookish' bang-erige 'somewhat afraid' verskill-ende 'several'

When the schwa becomes part of the underlying form of the adjective, it is protected against deletion by the condition of Faithfulness for lexical morphemes. For underived adjectives, certain tendencies might be formulated, based on the phonological composition of the adjectives. For instance, adjectives ending in a voiceless stop tend not to have the final schwa. In some cases, both forms are possible, as for dubbel(e) 'double' and enkel(e) 'single'. In other words, in Afrikaans, the occurrence of final schwa is no longer a matter of phonology or a matter of morphology, but of massive, syntactically conditioned lexical allomorphy: adjectives that appear with a final schwa in attributive position must be listed with this syntactically determined allomorph in the lexicon. The schwa-less form is the default form, to be used in predicative position, and as the form for word formation (composition and derivation). In the case of complex adjectives, the existence of the schwa-final allomorph is predictable, and this regularity can be expressed by a lexical allomorphy rule. The schwa in these cases no longer has the status of an inflectional morpheme, it is a stem-extension. As may be expected on the basis of this massive allomorphy, there are also cases in which the predicative form differs from the attributive form in unpredictable ways. For instance, the adjective for 'new' is nieuw /niw/ in attributive position, but nuut /nyt/ in predicative position. Moreover, in many cases the two allomorphs exhibit meaning differences (Raidt 1968: 108; 1983: 143ff). For instance, we have the opposition between (32) 'n waar storie 'a true story' 'n ware Afrikaner 'a real Afrikaner' This kind of semantic differentiation supports the claim that the occurrence of final schwa in Afikaans adjectives has to be interpreted as non-phonological allomorphy, to be listed in the lexicon.

54

Geert Booij

5. Conclusions What can we learn from these facts concerning the variation in the form of attributive adjectives in a number of related Germanic languages? There are two main conclusions to be drawn. First, a phonological theory that makes use of output constraints enables us to give a better account of variation and change than a rule-based description, because the latter does not unveil the driving forces behind variation and change, the perennial struggle between optimal phonetic form and the overt realization of linguistic information, whereas the constraint-based approach does. Secondly, we have seen that given an output-constraint-based analysis, the differences between related languages cannot always be expressed as a difference in the ranking of constraints. Instead, the differences had to be stated at the lexical level, namely as a difference between the absence (Middle Dutch) or presence (Dutch and Frisian) of two allomorphs for a certain suffix, which made it possible to create more optimal phonetic forms, at least for attributive adjectives ending in /n/. We also saw that zero-allomorphy in these two languages is constrained by the principle that it is only contextual, i.e. syntactically predictable inflection that allows for zero-allomorphy. The contextual nature of the inflectional schwa in attributive adjectives made it possible that in Afrikaans this schwa completely lost its morphological status, and only remained as the last segment of an allomorph of the adjective that can only appear in attributive position. In sum, phonological variation within languages, and differences between related languages or between different historical stages of a language may be a matter of a kind of allomorphy that cannot be completely reduced to phonology, rather than a matter of differences in the ordering of rules or the ranking of constraints. In that sense, the study of language variation and change supports a rather concrete approach to phonological alternations in the forms of words.

Language variation and phonological theory

55

Notes 1. Cf. also Hinskens et al. (1997) for a recent survey of the issues involved in the relation between phonological theory and the study of language variation. 2. In Frisian, the distinction between the inflected and the non-inflected infinitive forms has been preserved in another way: after to, and in a number of other nominal contexts, the infinitive ends in -en (the residue of -ene), whereas in non-nominal contexts the ending is -e (the residue of-en) (cf. Dyk 1997).

References Booij, Geert 1994 "Against split morphology." In: Geert Booij & Jaap van Marie (eds.) Yearbook of Morphology 1993. Dordrecht: Kluwer, 27-51. Booij, Geert 1998 "Phonological output constraints in morphology." In: Wolfgang Kehrein & Richard Wiese (eds.), Phonology and Morphology of the Germanic Languages. Tübingen: Niemeyer, 141-163 Carstairs, Andrew 1988 "Some implications of phonologically conditioned suppletion." In: Geert Booij & Jaap van Marie (eds.) Yearbook of Morphology 1988. Dordrecht: Foris, 6794. Dyk, Sybren 1996 "From inflected material adjectives to the history of schwa apocope in West Frisian: diverging influences on a sound change." In: A Frisian and Germanic Miscellany. Published in Honour of Nils Arhammar on his Sixty-fifth Birthday. Odense: Odense University Press/ Bredstedt: Nord-Friisk Instituut, 55-67 [= Nowele 28/29] Dyk, Sybren 1997 Noun Incorporation in Frisian. Ljouwert: Fryske Akademy [dissertation, University of Groningen]. Golston, Chris & Richard Wiese 1996 "Zero morphology and constraint interaction: subtraction and epenthesis in German dialects." In: Geert Booij & Jaap van Marie (eds.) Yearbook of Morphology 1995. Dordrecht: Kluwer, 143-159. Hinskens, Frans, Roeland van Hout & W. Leo Wetzels 1997 "Balancing data and theory in the study of phonological variation." In: Frans Hinskens et al. (eds.) Variation, Change, and Phonological Theory. Amsterdam: John Benjamins, 1-33. Kiparsky, Paul 1968 "Linguistic universale and linguistic change." Reprinted in Kiparsky 1982: 1344. 1971 "Historical linguistics." Reprinted in Kiparsky 1982: 57-80 1982 Explanation in Phonology. Dordrecht: Foris.

56

Geert Booij

McCarthy, John J. & Alan Prince 1993 Prosodic Morphology I. Constraint interaction and satisfaction. Ms. Univ. of Mass., Amherst and Rutgers Univ. 1994 "Generalized alignment." In: Geert Booij & Jaap van Marie (eds.) Yearbook of Morphology 1993. Dordrecht: Kluwer, 79-153. Löhken, Sylvia C. 1997 Deutsche Wortprosodie. Abschwächungs- und Tilgungsvorgänge. Tübingen: Stauffenburg Verlag (Studien zur deutschen Grammatik, 56). Prince, Alan & Paul Smolensky 1993 Optimality Theory: constraint interaction in generative grammar. Ms. Rutgers Univ. / Univ. of Colorado. Raidt, Edith H. 1968 Geskiedenis van die byvoeglike verbuiging in Nederlands en Afrikaans. Kaapstad etc.: Nasou Beperk. 1983 Einfiihrung in Geschichte und Struktur des Afrikaans. Darmstadt: Wissenschaftliche Buchgesellschaft. Stampe, David 1969 "The acquisition of phonetic representation." Papers from the 5th Regional Meeting of the Chicago Linguistic Society, 433-444. Taeldeman, Johan 1980 "Inflectional adjectives in the dialects of Dutch." In: Wim Zonnneveld, Frans van Coetsem & Orrin W. Robinson (eds.), Studies in Dutch Phonology. The Hague: Martinus Nijhoff, 223-246. Wiese, Richard 1996 The Phonology of German. Oxford: Clarendon Press.

Perceptual dialectology: aims, methods, findings Dennis R. Preston

In the early 1980's, I began a program of research which has come to be known as 'perceptual dialectology.'1 Unfortunately, when I began this work, I was not aware of similar studies which had been done years earlier. I suppose I might be forgiven in part since the technique I used first (handdrawn maps of regional speech areas by linguistically naive respondents) was, in fact, not a procedure used previously. In this paper I hope to atone somewhat for my ignorance by explicitly connecting those earlier contributions with more recent work in this general area.

1. Similarities and Differences In the earliest work in perceptual dialectology, linguistically naive respondents were asked to evaluate the degree of similarity (or difference) of the speech of surrounding localities, and the earliest systematic technique for determining dialect boundaries based on such data was developed in the Netherlands.2 The following two questions were included in a 1939 Dutch dialect survey: 1) In which place(s) in your area does one speak the same or about the same dialect as you do? 2) In which place(s) in your area does one speak a definitely different dialect than you do? Can you mention any specific differences? (Rensink 1955 [1999]:20) Weijnen (1946) devised an explicit method to represent the information uncovered by the first question. His 'little-arrow' method connects a respondent's home area to another which the respondent says is similar. Groupings of these connected areas, representing the response of a single respond-

58

Dennis R. Preston

ent at each location, are then identified as 'unities' based on the dialect consciousness or 'awareness' of the respondents. The earliest of these maps (for the North Brabant) appeared in Weijnen 1946. Here I show (in Figure 1) only the westernmost portion of that map since it is not so intricately detailed as some other parts but will nevertheless allow an illustration of the method. The thick lines are 'traditional' dialect divisions ('bundles of isoglosses'), and the 'perceptual areas' can be determined by encircling those community labels (letters) which are connected by arrows. For example, in the northwest of this section of the map, the respondent from W (Willemstad) has indicated that no nearby community sounds like W, and, therefore, no arrow is drawn from that site. Similarly, no surrounding communities have identified W as sounding like them, so no arrows are drawn towards it. In contrast, the respondent from D (Dinteloord) believes that the variety in F (Fijnaart) is the same as the local one, and the respondent from F returns the favor; hence, an arrow from D to F and one from F to D. The F respondent also identifies Κ (Klundert) as the same, but, unlike D, this perception is not reciprocal. If there were a perfect match between perception (the arrows) and production (the thick lines), every site within the production boundary (W, D, F, and K) would be connected to every other one with two arrows (W to D, D to W, W to F, F to W, D to F, F to D, etc.). That is obviously not the case. On the other hand, one must be impressed with the perceptual - production match here, for, although not all the sites are connected to one another, none identifies as similar a site outside the production boundary, nor is any identified as similar by a site outside the production boundary. A more complex relationship exists in the area just to the east of this section. There Ζ (Zewenbergen) identifies Μ (Moerdijk, just to the north) as being the same (although reciprocal identification is not given), and Ζ itself is identified as the same by a respondent from one site rather far to its southwest. In both these cases, however, the thick line just to the east is not crossed. The respondent from Z, however, also asserts the similarity to Ζ of both ZH (Zevenbergschen Hoek) and L (Langeweg), both clearly across the production boundary, although the respondent at neither ZH nor L identifies Ζ as similar. In general, however, there are a relatively small number of 'production boundary crossings' in this work. Of course, other interpretations than those which correlate folk perception and actual production boundaries may be offered. On the one hand, one might ask what linguistic facts (or even what sort of fact, i.e., phonological, lexical, grammatical) are most salient to the folk. Weijnen believes they are

Perceptual dialectology

59

Figure 1. The westernmost section of the North Brabant, showing isogloss boundaries (thick lines) and the 'little arrows' of respondent similarity perceptions (enlarged from Weijnen 1946).

phonological ones, since, according to him, they are 'sharper' than syntactic and morphological boundaries and less specific than those which arise as the result of the difference of a single lexical item; they are therefore both more 'locally noticeable'and general (e.g., 1961:5-6; 1966:194-5). As a corollary to that search, one might ask what linguistic facts not uncovered in production dialect studies might play a role in folk perception. Daan (1970 [1999]) reflects on this question, pointing out that intonation, for example, might play just such a role in folk awareness but is rarely studied in traditional dialectology. One might add, of course, vocal quality, speech rate,

60

Dennis R. Preston

and a number of other factors. For example, Goosens (1997) has recently shown how (in Great Britain and the Netherlands and Dutch-speaking Belgium) prosodic and intonational characteristics play a role in respondents' both identifying a variety's 'distance from the standard' and its regional provenience. One might also ask what sorts of socio-historical (or 'nonlinguistic') facts influence perception. For example, in Figure 1 what makes the respondent from Ζ look north and east rather than south and west in identifying similar areas, causing the most significant production 'boundary breaking'? As Daan shows (1970 [1999]), a religious boundary may account for respondents' strong feelings that there is also a linguistic one there (when none exists). We might expect, therefore, that such important social factors will often have dialect repercussions. As I have shown (e.g., Preston 1996b), the culturalhistorical border of the 'South' in the United States has a powerful influence on dialect recognition and evaluation. Rensink (1955 [1999]) provided the first generalized map of Dutch-speaking areas based on those perceptions gathered in the 1939 survey (in which thick lines are drawn around the bundles of interconnected little arrows), but Daan (1970 [1999]) is the most ambitious study of all contiguous Dutchspeaking areas, basing her map of Dutch dialects (Figure 2) on both perception (i.e., 'little-arrow') and production data. In general, although respondents were also asked to mention areas which were different from their own and to cite linguistic features which divided them from their neighbors, the principal motivation in this research seemed to have been a desire to give dialect boundaries greater (or lesser) 'weight' by establishing their folk validity on the basis of perceived similarity at very local levels. This is most directly and thoroughly discussed by Weijnen (e.g., 1966 and 1968 [1999]).3 In the late 1950's and on into the 1960's a series of articles introduced both the study of and controversy over subjective boundaries of dialects in Japan (and in general). Grootaers (1959) notes that the survey of the Itoigawa region (in western Japan) included perceptual questions which were partially inspired by the work in the Netherlands, specifically by a summary of Rensink (1955 [1999]). It is also noted, however, that the Japanese interest in where boundaries are to be drawn (and if folk information should be included) was reflected in the earlier work of Misao Töjö. In the Itoigawa research reported in Sibata (1959 [1999]), respondents were asked to indicate which nearby villages were 1) not different, 2) a little different, 3) quite different, or 4) mostly incomprehensible. The question from which maps for Dutch perceptual dialects were derived (which asked where

Perceptual dialectology

61

Figure 2. Dutch dialect areas, perceptual ('little arrow') and production data combined, with increasingly darker areas showing greater divergence from 'standard' Dutch (adapted from Daan 1970).4

dialects were 'similar,' presumably the equivalent to question 1) was found to be of little or no value in the Japanese research. In fact, Willem Grootaers, a co-worker on this project, notes that for the Japanese research 'the first one "no difference" and the second "slight difference" proved to be superfluous' (1959:356). Therefore, the results of question 1) above were ignored, and the

62

Dennis R. Preston

results of questions 2) and 3) were combined for one mapping effort, while those for question 4) were treated separately. At first glance, therefore, one might suggest that the Dutch area maps are ones of 'similarity' and that the Itoigawa maps are ones of 'difference' (but see below). The Japanese team did not know the little-arrow method developed by Weijnen, and, instead, they indicated, by increasingly thick lines, those areas which formed the 'difference boundaries' for groups of respondents (Figure 3). When respondents performed similarly in stating where such differences were, they were grouped into a subjective 'speech community.' Sibata (1959 [1999]) and Grootaers (1959, 1964 [1999]) claim that the resulting subjective dialect boundaries are of little or no interest to linguists since they do not generally correspond to traditional dialect boundaries. In fact, the Itoigawa and Dutch studies, although they gathered very similar data, produced maps based on radically different facts. In Figure 1, the speakers who say that other areas are 'the same' are assumed to form an interconnected network - a perceptual dialect area (and such areas were

or^'T

Γ. S I B A T A W GROOTAERS MTOKUOAWA

DIALECT MAP

—-1959 Figure 3.

The determination of two 'subjective areas' in Itoigawa (Sibata 1959 [1999]).

Perceptual dialectology

63

explicitly taken into consideration in the preparation of such maps as Figure 2). For example, sites D, F, and Κ form such an area in Figure 1 (and, in fact, site W forms another separate area, an 'isolate'). In Figure 3, however, what makes the sites inside the 'toothed' outlines (one on the left and a second on the right) belong to a perceptual area is not respondent claims that they sound like one another but their agreement about which areas sound different. In short, two different sorts of facts are being dealt with here. One cannot prepare a Dutch-style map for the Itoigawa area since, apparently, the respondents did not provide data about areas which sounded 'the same' or, as the Japanese researchers apparently put the question, 'no different.' One cannot prepare a Japanese-style map for the Dutch-speaking area since the data concerning differences (although sought in the Dutch questionnaire) were not made available. Mase (1964a [1999]), who also asked respondents to indicate surrounding areas which sounded the 'same' or 'different,' provides the first opportunity to look at maps based on both differences and similarities, since, unlike the respondents in Itoigawa, the Alpine Japanese in Mase's survey were willing to name surrounding sites which sounded 'the same.' Mase, however, does not draw maps based only on similarity. He was doubtless influenced by the work in Itoigawa, and he first maps responses to two questions

Figure 4.

The subjective dialect boundaries indicated by a respondent at site #57 (Mase 1964a [1999]).

64

Dennis R. Preston

- which sites sound 'the same' and which sites sound 'a little different.' Figure 4 shows how he combines these results in considerations which he will use in constructing his final map. The respondent at site #57 in Figure 4 has named sites #58 and #59 as 'the same.' He has also named sites #62, #63, #56, #55 and several sites in 'Nagawa' as 'a little different.' Speakers from sites #58 and #59 agree (not only that they are similar to one another and #57 but also that the same sites shown in Figure 4 are 'a little different'). Finally, not shown in Figure 4, respondents from surrounding areas classify sites #57, #58, and #59 together in their evaluations. In short, the 'perceptual dialect area' made up of these three sites is based on reciprocal (not individual, as in the Dutch research) perceptions of similarity, on similar perceptions of the first 'degree' of difference (as in the Itoigawa research), and on the perception by surrounding areas of their similarity to one another. The appearance of areas #57, #58, and #59 as area 'n' in Figure 5 is based, therefore, on three criteria (two, internal and external, of their similarity and one of their agreement about differences). Although this mapping technique seems more sophisticated than either the Dutch or Itoigawa research, it mixes ratings of similarity and difference and does not allow us to see boundaries produced by those different considerations. In fact, Mase's 'calculus' is even more complex. In many cases, there is no such nice agreement among sites as shown in Figure 4. When that occurs, Mase relies exclusively on difference ratings and uses two-thirds and onethird ratios to determine boundaries. In Figure 5, for example, such a complex relationship arose among sites #11 though #26. Mase's procedure was as follows: he counted a 'full point' for each site at which any respondent mentioned a first degree of difference boundaiy ('a little different'). He counted a half-point if the respondent modified that degree downward (e.g., 'a very slight difference'). He then calculated the number of points for all respondents in the region. If they equaled two-thirds or more of the respondents, he considered the boundary a 'major' one; if it equaled more than onethird (but less than two-thirds), he considered it a 'minor' one. In the situation described above, 11.5 points were calculated for the boundary between sites #14 and #15. Since 11.5 is greater than two-thirds of sixteen (the total number of sites, i.e., #11 through #26), sites #11 through #14 are grouped into one major perceptual region ('d') in Figure 5 while sites #15 through #26 are grouped into a second ('e'). Within those regions, however, seven points were given between sites #24 and #25, six between #25 and #26, and 5.5 between both #12 and #13 and #19 and #20. As Figure 5 shows, these

Perceptual dialectology

65

Figure 5. Mase's perceptual dialect areas for a section of Alpine Japan (1964a [1999]).

subdivisions are indicated by dashed lines (since their point totals amount to more than one-third but less than two-thirds of the total respondent judgments from the sites under consideration). This resolution of complex areas with reference to difference ratings, however, reduces our ability to distinguish which regions are identified on the basis of similarities (exclusively or predominantly) and which are identified on the basis of differences. In spite of that flaw, Mase's treatment of boundaries is more quantitatively sophisticated than that of any of his predecessors. It is clear that he developed the first 'quantitative' approach to perceptual dialectology, for, although the Itoigawa research team drew 'thicker' lines to indicate areas which were agreed on as different by a larger number of respondents, it is nowhere clear that a numeric standard was set for the determination of the perceptual areas (as shown in Figure 3). There is no quantitative approach in the 'little-arrow' technique, since only one connection (one similarity judgment) causes a site to be included in a perceptual area.

66

Dennis R. Preston

Mase (1964a [1999], 1964b [1999]) also compares his perceptual boundaries to a number of grammatical, lexical, and phonological isoglosses and, in general, finds a good correspondence between these linguistic boundaries and the ones determined from his perceptual study. On the nonlinguistic side, Mase (1964a [1999], 1964b [1999]) and Nomoto (1963 [1999]) find school districts rather than feudal and other political administrative zones (the areas which dominated the linguistic boundaries of the Itoigawa research, according to Sibata and Grootaers) to be very similar to perceptual boundaries. Weijnen, who devised the 'little-arrow' method for the Dutch perception data, criticizes the Japanese approach to perceptual studies in Itoigawa by noting a fatal flaw - the Sibata-Grootaers team asked people if there were differences (which, according to Weijnen, always exist) rather than asking people where others spoke the same (1968 [1999]). Nor surprisingly, Weijnen praises Mase, who found a greater parallel between perception and production, for his use of this 'more appropriate' question.5 Since Mase nowhere bases maps exclusively on local judgments of similarity, however, I suspect that Weijnen might not have approved if he had had full access to the original Japanese version of Mase 1964a [1999]. The last word in this Dutch-Japanese controversy may have to do with ends rather than means. If one seeks to supplement the details of production dialect maps with 'awareness weights' from local ratings (revealing which dialect boundaries have and do not have greater folk significance), then subjective maps which result in boundaries which do not generally correspond to production boundaries will be of little help. Although that appears to be Weijnen's goal, Daan (1970 [1999]) seems to be interested in what one might learn from the mismatches (as well as the matches). If one seeks a more general approach to dialect mapping, then, like Sibata and Grootaers, there may be disappointment that subjective boundaries do not provide a ready-made (perhaps even guiding) picture of language distribution. If, however, one seeks corroborating and explanatory evidence for dialect distribution (as Grootaers himself concludes [1964 [1999]]), then perhaps the voice of the folk should not be ignored. That this voice would have independent value (and applications) does not seem to be a conclusion reached in any of this early work. Motivated by a desire to explore folk knowledge for its own value, I also asked us respondents to rank regions on a scale of one to four (1 = 'same,' 2 = 'a little different,' 3 = 'different,' 4 = 'unintelligibly different') for the perceived degree of dialect difference from the home area (e.g., Preston

Perceptual dialectology

67

1993a, 1996b). Figure 6 shows the responses of southeastern Michigan respondents to this task; the mean score ratings were divided into four groups as follows: 1.00-1.75, 1.76-2.50, 2.51-3.25,3.26-4.00. I shall not indulge in many interpretive comments here since this review is primarily methodological. Note, however, that Figure 6 shows that when Michigan raters evaluate degrees of difference they perceive a rather large local area of similarity (behaving like Mase's raters rather than like those in Itoigawa). Here, however, the ratings of the South are of greatest interest. A large South emerges as a territory rated '3' (the same given the Northeast). Texas, Arkansas, Oklahoma, and Missouri are rated along with obviously Southern states (e.g., Georgia and South'Carolina). But a 'core' South (Alabama, Mississippi, and Louisiana) earns the '4' rating. These ratings suggest that the Michigan raters are aware of a wide area of influence of Southern speech, emanating from an unintelligibly different core. Even in the generally similarly rated Northeast there is no such 'unintelligible' core. This procedure is unlike those used in Itoigawa, in the Dutch-speaking areas, and in Mase's work. Like most current sociolinguistic work, it focuses

Figure 6. Mean degree of difference ratings for MI respondents (N=147).

68

Dennis R. Preston

on a large number of demographically diverse respondents from one area and tries to generalize (here, by means scores) on their classifications. In contrast, all the studies cited so far, inspired by traditional work in dialectology, surveyed only one (older, usually male) respondent from each (the 'NORMs' - nonmobile, older, rural, males - of Chambers and Trudgill 1980:33). Generalizations from the older studies, therefore, face the difficult task of combining responses from more than one area and use quantitatively and representationally questionable data from only one respondent per site. On the other hand, my work in the us focuses on a broad, non-local assessment of dialect distinctions. This is perhaps justified since us dialects do not (usually) reveal the same finely-tuned local differences one finds in rural Japan and Dutch-speaking areas. Unless there is a significant speech (or culture) boundary nearby, asking about the difference in speech from one town or village to the next in the us (within areas as small as those surveyed in the Japanese and Dutch-speaking research) might reveal only vast areas of 'the same' (as seen in Figure 6 for Michiganders). This is, however, an empirical question, and, unfortunately, I do not have survey data which asks respondents to rate nearby sites, only data which focuses on broad (state) analyses of the entire country. I will consider this question of local detail in the us below when I comment on the use of respondent hand-drawn maps. Evaluations of the degree of difference in the style carried out for southeastern Michigan have been done in other sites in the us: southern Indiana (Preston 1985b, 1988a,b,c, 1989a,b, 1993a,c, 1996b); the us 'South' (Preston 1996b,1997); and Oregon (Hartley 1996, 1999). Sites outside the us include Paris, France (Kuiper 1999); Bursa, Turkey (Demirci and Kleiner 1999), and Germany (Dailey-O'Cain 1997, 1999). In addition to the reporting of means scores (and normal statistical tests of their significance) some of these studies have employed multidimensional scaling and cluster analyses to arrive at generalizations concerning the ratings. Generally speaking, since these more recent studies use pre-set areas (states or other political boundaries), they have seldom been concerned with the correlation between their findings and those of traditional dialectology. That, of course, is a major difference between them and the earlier studies. These more recent efforts seem to be content to analyze the results in sociocultural terms, although that is not necessarily true of all current attempts in perceptual dialectology (e.g., Lance 1999).

Perceptual dialectology

69

2. Hand-Drawn Maps My first interest in perceptual dialectology was realized in a task in which I asked respondents to draw (on a blank map with only state lines and occasionally other prominent geographic features) lines around areas where they believe regional speech zones exist and to label them with names of the area, of the dialect, of typical speakers from them and/or to jot down representative examples of speech for each (Preston 1981). Although respondent handdrawn maps were well-known in cultural geography (e.g., Gould and White 1974), there does not appear to be a tradition for the use of this technique in the study of dialect perceptions. Figure 7 is an example of such a hand-drawn map from a southeastern Michigan respondent. One might first note that, although Michigan respondents include a wide 'Upper North Central' or 'Great Lakes' territory in category number one ('the same') in the degree-of-difference task (Figure 6), this respondent singles out Michigan for the label 'average normal.' Different tasks, therefore, may elicit different responses. There are also a number of interesting cultural stereotypes written on this map ('hillbillies' for Texas, 'British' for New England, 'Eskimo's' for Alaska, and 'Sunny Side' for California). Studies of such labels have

Figure 7. A hand-drawn map from a young, university-enrolled southeastern Michigan respondent.

70

Dennis R. Preston

attracted some attention (e.g., Preston 1981,1993b; Hartley and Preston 1999; Long 1990 [1999]), but I will not summarize those findings here since, methodologically, they are simply classificatory. Since they provide clues to later studies, I will, however, have more to say about them below. As seen in the Dutch and Japanese work outlined above, what dialectologists want to know is the location of the perceptual or subjective dialect boundaries established by such a task. Since in my work maps were collected from a great number of respondents all from one site, it is necessary to prepare a generalization of their perceptions of dialect boundaries. A technique developed by Preston and Howe (1987) allows computerized generalizations of such maps. Each respondent's map was traced onto a digitizing pad which fed the outline information into a program keyed to a us map. For each respondent's identification of an area (e.g., 'South'), the program recorded one 'hit' for each pixel enclosed in or touched by the respondent's boundary. This technique allows automatic compilation of composite maps based on large numbers of respondents and on demographic subdivisions of them (Preston and Howe 1987), although the latter will not be discussed here. The generalizations which emerge from the computer compilations are not automatic. For example, 138 southeastern Michigan respondents drew some representation of the us South, and their maps were subjected to the computer process outlined above. If one asks the computer to display the

Figure 8.

Southeastern Michigan respondents' computer-generalized map, showing where even one respondent outlined a 'South'.

Perceptual dialectology

71

entire territory of the South for which even one respondent included a pixel, such an uninformative map as Figure 8 emerges, obviously an exaggeration and most likely the result of one or two sloppy or idiosyncratic drawings. It is necessary, therefore, to seek other patterns of agreement; the territory outlined by fifty percent of the respondents provides a good generalization, although, to be precise and to provide additional insights (as will be illustrated immediately below), one will want to sample a number of 'percentages of agreement.' Figure 9 shows the area called 'South' by southeastern Michigan and southern Indiana respondents at this fifty percent level of agreement. The Michigan map-drawers obviously have a bigger 'vision' of the South. In fact, I think the explanation of this difference is straightforward. The speech of the us South is prejudiced against, but the territory outlined as 'South' by the Michiganders comes dangerously close to the area inhabited by the southern Indiana respondents. I suspect that the Indiana respondents have 'pushed' the 'South' to the south (away from them) so that they will not be 'contaminated' (or allow others to believe that they are contaminated) by southern speech. This procedure allows questions other than that of the 'best' generalization to be asked. For example, 1) Where is the core of a region (here, the South)? 2) Do different percentages of respondent agreement show concen-

Figure9. Southern Indiana (outlined - 53 of 106) and southeastern Michigan (shaded - 69 of 138) respondents' generalizations of the us South at the 50% level.

72

Dennis R. Preston

Figure 11. Michigan respondents' 91% agreement for the 'South'.

Perceptual dialectology

1. South 2. North J. Northeast 4· Southwest 5- W««t 6 • Inner South 7 · Plains and Mountain· β. Taxa· 9· New England 10. Midweat 11. Florida 12. California 13. West Oo«»"t 14. net Coast

73

It · 1*7 1. 138 (.»») 2. 90 ( .61) 3· 60 {.9*) 9 ϋί.ΟΟ-«.99* HI7.00-7.99 Ηβ.ΟΟ-8.99 * New York City

Figure 17. Mean scores of southern Indiana ratings of 'pleasant' speech.

in the expression of status rather than solidarity matters. Finally, of course, details of stereotype and caricature are more definitively cataloged through such tasks, interpretations which were only hinted at in the earlier work of Daan (e.g., 1970 [1999]) and Mase (e.g., 1964a [1999]). Recent studies of 'pleasant' and 'correct' varieties have been extended to the us West (Hartley 1996, 1999), Turkey (Demirci and Kleiner 1999), Germany (Dailey-O'Cain 1997, 1999), France (Kuiper 1999), Japan (Long 1997b, 1999), and Brazil (Preston 1985a).

4. Putting It All Together This last section is devoted to what one might call the sociolinguistic (or 'modern') trend in perceptual dialectology. It begins with the work of Fumio Inoue (1977/8, 1978/9), who has devised a technique which characterizes speech regions on the basis of what he calls 'dialect image.' His method is clearly related to the semantic differential or matched-guise technique of

82

Dennis R. Preston

many studies of language attitudes, but it uses a different statistical approach and does not rely on the reaction to speech samples. Although Inoue (1995 [1999]) is a recent publication, it illustrates the foundation for much of his work. Inoue elicits evaluative words associated with regions, much in the same way pairs of opposites are elicited for the semantic differential evaluation of matched-guise presentations. He subjects these dialect 'labels' to a Japanese version of multi-dimensional scaling known as Hayashi 3, which allows a researcher to group together both the evaluative labels assigned to varieties and, later, the varieties themselves. The two principal characteristics associated with dialect image in Japan, for example, are 'intellectual' and 'emotional.' These correspond closely to the 'status' versus 'solidarity' factor groups which emerge from most quantitative work done on language attitudes by social psychologists of language (e.g., Ryan, Giles, and Sebastian 1982) and to the 'correct' versus 'pleasant' characteristics of varieties used in the work described above. Inoue also applies this technique in Great Britain, where, interestingly, the components selected in the statistical treatment of labels (given to uni-

-0.51· Figure 18. Distribution of dialects as a result of Hayashi's quantificational theory type 3 for British University students (Inoue 1996 [1999]: 146).

Perceptual dialectology

83

versity students) were not 'emotional' and 'intellectual' (as they were for earlier work in Japan) but 'rural' and 'standard.' Figure 18 shows how a variety of English-speaking regions are evaluated on these same dimensions (renamed 'Accentedness - Standardness' and 'Urbanness - Pastoral (Rural)'). In spite of this statistical sophistication on the evaluative side, Inoue (1996 [1999]), who also has respondents provide him with hand-drawn maps of dialect areas in Great Britain, concludes, like Shibata and Grootaers, that there is little or no correspondence between perception and production boundaries. In fact, he sees a stronger correlation between folk dialect perception and the sorts of maps one encounters in public school education and in such popular media vehicles as weather maps. It is not surprising, therefore, that he has not carried out tasks in which the cognitively real speech areas of respondents (as determined by hand-drawn maps, for example) are used as the basis of evaluation. In recent work I have tried to do just that. That is, I have used a generalization of the cognitively real map of speech areas (as seen from the point of view of southeastern Michigan, Figure 12 above) as the basis for the areas to be rated and have followed the typical procedures used in a language attitude investigation (e.g., Shuy and Fasold 1973). It was assumed that this mental map (a result of the hand-drawn map research described above) would also be typical of the respondents to be investigated in this research. In 'classic' language attitude research style, it was determined that the following labels would be relevant to an investigation of attitudes to those areas. slow - fast educated - uneducated polite - rude nasal - not nasal friendly-unfriendly twang - no twang

formal - casual smart - dumb snobbish - down-to-earth normal - abnormal drawl - no drawl bad English - good English

These descriptors were elicited by showing a large number of respondents (none of whom participated in the subsequent evaluation task) a simplified version of Figure 12 and asking them to mention any characteristics of the speech of those regions which came to mind. The most frequently mentioned descriptors were selected and arranged into these pairs, which were then presented in a six-point 'semantic differential' task as shown below.

84

Dennis R. Preston

The respondent judges (85 young, European-American southern Michigan residents who were undergraduate students at Michigan State University) were shown a simplified version of Figure 12 and given the following instructions: Instructions

This map shows where many people from southern Michigan believe speech differences are in the U.S. We will give you a list of descriptive words which local people have told us could be used to describe the speech of these various regions. Please think about twelve7 of these regions, and check off how each pair of words applies to the speech there. For example, imagine that we gave you the pair 'ugly' and 'beautiful' ugly

beautiful a

b

c

d

e

f

You would use the scale as follows: If you very strongly agree that the speech of a region is 'ugly,' select 'a.' If you strongly agree that the speech of a region is 'ugly,' select 'b.' If you agree that the speech of a region is 'ugly,' select 'c.' If you agree that the speech of a region is 'beautiful,' select'd.' If you strongly agree that the speech of a region is 'beautiful,' select 'e.' If you very strongly agree that the speech of a region is 'beautiful,' select 'f.' Use the op-scan form (and the numbers on it) for all answers. 1) First, please tell us your sex a. female b. male Go on to Region #1 (which begins with question #2 on the next page). Refer back to the map on this page whenever you like. Thank you very much for your cooperation. The first step in classic language attitude work is to determine whether or not the paired items used in evaluating the 'samples' can be reduced. This is normally carried out by means of a factor analysis. The results of such an analysis for all areas rated are shown in Table 1.

Perceptual dialectology

Table 1.

85

The two factor groups from the ratings of all areas. Parenthesized factors indicate items within the .25 to .29 range; prefixes indicate negative loadings and should be interpreted as loadings of the opposite value (given in brackets).8 Factor Group #2

Factor Group #1 Smart Educated Normal Good English No drawl No twang Casual [Formal] Fast Down-to-earth [Snobbish]

.76 .75 .65 .63 .62 .57 -.49 .43 -.32

Polite Friendly Down-to-earth (Normal) (Casual)

.74 .74 .62 (.27) (.27

Two robust factor groups emerge. The first (which I will call 'Standard') shows loadings from those categories which one associates with education and majority norms. Note, however, that the last three factors in this group ('Formal,' 'Fast,' and 'Snobbish') are not necessarily positive traits. Factor Group #2 (which I will call 'Friendly') loads affective factors (including two which are negatively loaded in Factor Group #1 - 'Down-to-earth' and 'Casual'). These groups will not surprise old hands at language attitude research. As Ryan, Giles, and Sebastian (1982) note, 'With regard to the structure of attitudes toward contrasting language varieties, the two major dimensions along which views can vary can be termed social status and group solidarity [italics mine]' (8). A full analysis of these data would, of course, go on to consider the realization of each of these factors (and groups) with regard to each of the areas rated. I have not done this, first, because it would be too space-consuming, and, second, because I believe a sample of two particularly salient areas (for these respondents) will provide a good insight into the mechanisms at work here. I have chosen, therefore, to look at the respondent ratings of areas 1 and 2 from Figure 12 for very straightforward reasons. Region 1 is the us 'South,' and Figure 12 shows that it was outlined by 94% (138) of the 147 respondents who drew maps. For these southeastern Michigan respondents, it is clearly the most salient regional speech area in the us. Although one

86

Dennis R. Preston

might note anecdotal or popular culture characterizations of why that might be so, a look at Figure 14 will remind the reader of the even more dramatic explanation - the salience of southern speech would appear to lie in its distinctiveness along one particular dimension - it is incorrect English. The second most frequently rated region (by 90 out of 147 respondents or 61%) is the local one called 'North' in Figure 12, but more accurately 'North Central' or 'Great Lakes.' At first, one might be tempted to assert that the local area is always salient, but a closer look at Figure 14 will again remind the reader that these southeastern Michigan raters may have something else in mind when they single out their home area. It is only Michigan which scores in the heady 8.00 to 8.99 means score range for language 'correctness.' In short, perception of language correctness (in the positive direction) determines the second most salient area for these respondents. Although investigation of the ratings of other areas will doubtless prove interesting, a careful look at those of the high-prestige local area ('North') and of the most highly stigmatized area ('South') will prove most revealing. Table 2 shows the means scores for the individual attributes for the North and South. Perhaps the most notable fact is that the ranked orders are nearly opposites. 'Casual' is lowest-rated for the North but highest for the South. 'Drawl' is lowest-rated (meaning 'speaks with a drawl') for the South but highest rated (meaning 'speaks without a drawl') for the North. In factor group terms, the scores for Factor Group #2 (and ' - Γ loadings) are the lowest-ranked ones for the North; these same factors (e.g., 'Casual,' 'Friendly,' 'Down-to-earth,' 'Polite') are the highest-ranked for the South. Similarly, Factor Group #1 scores are all low-ranked for the South; the same attributes are all highest-ranked for the North. These scores are not just ordered differently. As indicated by the '@' in Table 2, a series of paired t-tests shows that there is a significant difference (p < 0.05) between the attribute ratings for the North and the South, except for 'Nasal' and 'Polite.' For those attributes which load on Factor Group #1 ('No Drawl,' 'No Twang,' 'Fast,' 'Educated,' 'Good English,' 'Smart,' and 'Normal'), the means scores are all higher for the North. In other words, these Michigan raters consider themselves superior to the South for every attribute of the 'Standard' factor group. This is not very surprising, considering the results from earlier research on 'correct' English shown in Figure 14.

Perceptual dialectology

Table 2.

Means scores of attributes.' *' marks the only significant (p < 0.05) break between two adjacent scores (determined by an analysis-of-variance with a Tukey comparison of means); ' J ' marks values below 3.5 (which indicate the opposite polarity, shown in brackets here and in Table 3); '#' indicates the only scores significantly different for gender (p < 0.05, determined by a series of t-tests); '@' marks the only two attributes ('Nasal' and 'Polite') for which there was no significant difference (p < 0.05 on a series of paired t-tests) between the ratings for North and South.

Means scores (ordered) North Rank 12 11 9.5 9.5 8 7 6 5 4 3 2 1

Factor

Mean

Attribute

-1&2 0 2 2 1 1 2&-1 1 1 1&2 1 1

3.53 3.94@ 4.00 4.00@ 4.09 4.12 4.19 4.41 4.53 4.94 5.07 5.11

Casual Not nasal Friendly Polite Educated Fast Down-to-earth Good English Smart Normal No twang No drawl

Means scores (ordered) South Rank

Factor

Mean

Attribute

1 2 3 4 5

-1&2 2 2&-1 2 0

4.66 4.58 4.54 4.20@ 4.09@ *

Casual Friendly Down-to-earth Polite Not nasal

6 7 8 9 10 11 12

87

1&2 1 1 1 1 1 1

t3.22 »3.04 "12.96 »2.86 *2.72 #t 2.42 »2.22

Normal [Abnormal] Smart [Dumb] No twang [Twang] Good English [Bad Eng.] Educated [Uneducated] Fast [Slow] No drawl [Drawl]

88

Dennis R. Preston

Before considering the scores for the attributes in Factor Group #2 ('Friendly'), recall what Michigan raters have done previously in a direct assessment of the notion 'pleasant.' As Figure 16 above shows, the South fares very badly again. Alabama (actually tied here by New York City) is the worstrated area in the us, and the surrounding southern states are also at the bottom of this ten-point rating scale. One may note, however, that the ratings for the 'pleasantness' of the English of southern states are one degree less harsh than those for 'correctness.' Similarly, there is no 'outstanding' (8.00-8.99) rating as there was for 'correctness,' making Michigan no longer the uniquely bestthought-of area (since it is joined here by Minnesota, Illinois, Colorado, and Washington). As suggested above, I have taken this to indicate that northern speakers have made symbolic use of their variety as a vehicle for 'standardness,' 'education,' and widely-accepted or 'mainstream' values. On the other hand, southern speakers (who are more than a little aware of northern prejudices against their variety) use their regional speech as a marker of 'solidarity,' 'identity,' and local values. For those attributes which load on Factor Group #2 (or -1), the means scores are higher for the South for 'Casual,' 'Friendly,' and 'Down-to-earth.' There is no significant difference for 'Polite' (as noted above), and the North leads the South in Factor Group #2 attributes only for 'Normal,' but it is important to note that 'Normal' loaded (positively) on both groups. This is indeed a new finding for rankings of the prejudiced-against South by linguistically-secure northerners. These data suggest that, for these eightyfive young Michiganders, the 'Friendly' attributes (excepting only 'Polite') are more highly associated with southern speech than with speech from the local area. A few other statistical facts confirm and add to the results reported so far. Sex (the only demographic variable testable in this experiment) played little or no role. No rating of any attribute for the North differed by sex, and only 'Twang' and 'Fast' differed for the South. (Female raters found the South 'twangier' and 'slower.') More importantly, note (in Table 2) that no attribute rating for the North falls below 3.5 (the median value of the six-point scale), while all of the Factor Group #1 ('Standard') attributes are rated below that score for the South. Perhaps even more dramatically, analysis-ofvariance tests of the means scores for North and South independently show that there is no significant break (p. < 0.05) between any two adjacent means scores for ratings of the attributes for the North. On the other hand, there is such a significant difference for the South between the Factor Group #2 (and -1) attributes and the Factor Group #1 attributes, as shown by the '*' in Table

Perceptual dialectology

89

2. In other words, there is a continuum of relatively positive scores for the North and a sharp break between the two factor groups for the South. This break can be even more directly shown in Table 3 which displays the combined means scores for the two factor groups and the two areas focused on here. It is only Factor Group #1 ('Standard') for the South which is very different from any other. Unfortunately, this representation of the results hides the important fact that ratings of individual attributes for North and South are nearly all significantly different (as shown in Table 2 above). What it does reveal, however, is that this model of research found considerably better ratings of the South by northerners along the 'affective' dimension than did previous research on 'pleasantness' ratings (as shown, for example, in Figure 3). Table 3.

Factor group means (all attributes combined)

Factor Group #1 Factor Group #2

North

South

4.44 4.13

3.18 4.24

What does this combined approach contribute to perceptual dialectology (and to language attitude study)? Most importantly, I believe we can be relatively assured that the judges have rated regions which are 'cognitively real' for them; that is, they have rated areas for which the notion 'regional speech' has been shown to have folk linguistic status. Unlike classic matched-guise attitude studies, this research provides respondents with the category name and mapped outline of regions rather than actual voice samples. The obvious benefit of this is that I do not have to use what could only be gross, stereotypical imitations of varieties (if one speaker tried to imitate all the varieties studied here or even the two more carefully looked at). Since some recent language attitude research has shown that there is little or no difference in evaluations when the stimulus is a category name or an actual speech sample (e.g., Coupland, Williams, and Garrett 1994 and Williams, Garrett, and Coupland 1999), I have not considered this manner of presenting the stimulus to be a deficiency. Of course, the question of whether or not respondents can identify varieties is still an open one and requires independent study. Here I chose to investigate the 'stereotypes' respondents have of regional voices (without submitting a sample), and I consider this to

90

Dennis R. Preston

be one approach to the larger question of identification of and attitudes towards regional varieties. It is the case that the respondents in this study were all young, collegeenrolled undergraduates, but the mental map (a simplified version of Figure 12) which was shown to them was derived from a study of a variety of age groups and social classes in southeastern Michigan. In spite of this fact, very few important differences were found in age, gender, and status representations of dialect regions or even in the evaluations of them (e.g., Preston 1988a). I believe, therefore, there are no important differences between the cognitive map of regional speech for the respondents studied here and the generalization shown them. The major finding of this study, however, is that there is a considerable difference in the rankings here of the 'affective' dimension of attitudes of Michiganders to the South and those given by similar respondents in my earlier research. As Figure 16 shows, the ratings for the local area for 'pleasantness' were among the highest, and the ratings for the South along the same dimension were among the lowest. In the present study, however, the South actually did significantly better than the local area in three key characteristics of the affective factor group (#2 - 'Friendly,' 'Casual,' and 'Downto-earth') and was not significantly different on a fourth ('Polite'). As Table 3 shows, there is hardly any difference between the overall ratings of Factor Group #2 for the North and South. That is a very different picture from the one seen in Figure 16. What accounts for this amelioration of attitude towards the South among these raters? I think there are several possibilities. First, one might assume that the global label 'pleasantness' (used in the earlier research) does not as subtly (or perhaps as 'covertly') elicit the attitudes along this dimension. That is a real possibility, but I cannot resolve it here. Second, one might attribute this amelioration to the age of the respondents. Although they are the same age as the youngest group studied in the earlier research, they are certainly not their contemporaries. Those earlier data were collected in 1986-87; the data for the current study were collected in 1996. Since I do not have ratings from older respondents in the current research, however, it is difficult to make this comparison straightforwardly. There is some evidence (although it is confounded by region) that the respondents of the late 1990's may be behaving differently from those of the late 1980's, but the difference would appear to be in the area of evaluations of'correctness' (more closely parallel to Factor Group #1, 'Standard') rather

Perceptual dialectology

91

than along affective lines. In her work in Oregon, Hartley (1996) notes that a number of respondents (citing what sounds very much like a somewhat sophisticated linguistic relativism) refused to evaluate regions of the us for 'correctness' (or rated them all the same).9 She notes, however, no such accompanying reluctance for rating 'pleasantness,' and the student fleldworkers for the present study reported no such reluctance among these young southeastern Michigan judges in rating 'correctness' or 'pleasantness.' Third, one might suspect that some sort of 'covert prestige' attaches itself to southern speech (since it is clearly seen as 'incorrect'). If that were the case, however, one might expect to see a strong gender differentiation (with a male preference for the stigmatized variety), but, as Table 2 shows, there is little gender significance in the ratings. Additionally, high ratings for such attributes as 'Friendly' hardly point to 'tough' characteristics. I believe, however, that this last possibility moves in the right direction, but I also believe that previous definitions of 'covert prestige' are too 'tough' and 'male' oriented to cover the entire territory. Let us consider another possible interpretation. Although many handdrawn maps of us dialect areas by Michigan respondents label the local area 'standard,' 'normal,' 'correct,' and 'good English,' some also call it 'boring.' Since there is obviously no dissatisfaction with the local variety as a representative of 'correct English,' what is the source of the preference for other varieties along affective dimensions? Recall that I have suggested (e.g., Preston 1996b) that a group has a tendency to use up what might be called the 'symbolic linguistic capital' of its variety in one way or another (but not both). Speakers of majority varieties have a tendency to spend the symbolic capital of their variety on a 'Standard' dimension. Speakers of minority varieties usually spend their symbolic capital on the 'Friendly' dimension. I suggest that northerners (here, southeastern Michiganders) have spent all their symbolic linguistic capital on the standardness of local English. As such, it has come to represent the norms of schools, media, and public interaction and has, therefore, become less suitable for interpersonal use. In short, these young Michiganders don't identify other varieties for their 'covert prestige' on the basis of anti-establishment or tough characteristics alone; they also assign covert prestige to a variety which they imagine would have more value than theirs for interpersonal and casual interaction, precisely the sorts of dimensions associated with Factor Group #2. Of course I do not doubt the existence of 'covert prestige' along the traditional 'masculine' or 'tough' lines that Trudgill (1972) points out; I simply suspect that there are other kinds of covert prestige, or at least one in which friendship, solidarity,

92

Dennis R. Preston

trust, informality, strong emotion, and such factors are highlighted. Southern us English would appear to be such a variety for these judges. I will not develop here the popular culture, folkloristic, and qualitative evidence for this interpretation, although I am sure such caricatures (many encoded in the notion 'southern hospitality') are well-known, and northerners indeed comment on the fact that southern speech 'sounds nice.'10 From a language variation point of view, of course, we are ultimately more interested in the general social and linguistic mechanisms which are at work here. Ryan, Giles, and Sebastian (1982:9) outline the following evaluative possibilities for majority (LV1) and minority (LV2) speakers: Judges Type of preference

A. Majority group B. Majority group for Status/ in-group for solidarity C. In-group D. Majority group for status/ minority group for solidarity

LV1 speakers

LV2 speakers

Status Solidarity

Status Solidarity

LV1

LV1

LV1

LV1

LV1 LV1

LV1 LV1

LV1 LV2

LV2 LV2

LV1

LV2

LV1

LV2

In these terms, I wonder if speakers of Inland Northern us English (i.e., the Michiganders studied here) have changed from Type Β to Type D? u Perhaps speakers of some of these varieties have moved in the direction of RP speakers in Britain (the group Ryan, Giles, and Sebastian use to illustrate the LV1 pattern of their Type D). In other words, the inappropriateness of their own ('Standard') variety to interpersonal modes of communication has caused them to evaluate other (nonstandard) varieties higher for the characteristics identified as belonging to the 'Friendly' factor group. Space will not allow a thorough discussion of other interpretive dimensions of this finding, but I will briefly mention some. In Preston (1992) I note that, although young European-American imitations of African-American speech might be regarded as racist, many appear to have other motivations - sounding not only 'tough' and 'cool' but also 'casual' and 'down-to-earth.' This motivation among younger speakers is complex. Although adolescents are often presented with a dichotomous

Perceptual dialectology

93

choice between mainstream ('approved') and nonmainstream ('rebellious') behaviors, a middle ground exists in which there is a desire to succeed along traditional lines but another to display egalitarian principles, ones which require, on the linguistic front, the (at least partial) use of varieties seen as stigmatized. As a result of other associations with both the standard and the perceived nonstandards, these latter varieties also seem to be more appropriate for casual, interpersonal use. I believe the 'in-betweeners' in Eckert's (1989) suburban Detroit study (i.e., those who want to be neither the mainstream 'Jocks' - perceived as 'snooty' - nor the anti-establishment 'Burnouts' - associated with drug culture) display just such an attitude. One of them characterizes this dilemma of such 'neutrals' as follows: They [i.e., the 'neutrals,' neither 'Jocks' nor 'Burnouts'] just don't want to seem to turn to drugs to cope with their problems, and, uh, they want to, they want to have good grades, you know, but not be stuck up where you'll look at someone and say 'Well, you are lower than me,' and stuff like that. (174) In Britain as well, Rampton's work (e.g., 1995) evaluates the occurrence and meaning among adolescents of cross-ethnic language use (i.e., 'code-switching') and concludes that a principal function is its reflection of a desire to do away with ethnic boundaries. In other words, in resolving the adolescent tension between mainstream and nonmainstream behavior, a linguistic option might be the use of 'standard' English in settings which require that variety and a mixing of the speaker's native variety with perceived nonstandards in settings which require 'casual' use. In short, I do not believe that the use of or preference for nonstandard (or stigmatized) varieties by adolescents is uniquely associated with the 'anti-language' interpretation offered by Halliday (e.g., 1976) and apparently embedded in most interpretations of 'covert prestige.' I will not press this favoring of stigmatized varieties into service for general sociolinguistics too much further, but I want to mention the fact that it is one (alternative) option in the search for accounts for the introduction and spread of novel elements (particularly into the mainstream speech community); that is, it is another option in the search for answers to the problems of 'actuation' and 'embedding' (Weinreich, Labov, and Herzog 1968). Rampton's work referred to above also notes that the adolescent use of other varieties provides an opportunity for 'practice' (while making the 'social statement' that such selection implies). If that is true, then all the cases of accommodation, imitation, and acquisition of'socially motivated' non-main-

94

Dennis R. Preston

stream varieties discussed here (and in many other places in the sociolinguistic literature) are opportunities for the introduction and spread of alternative forms. This may be particularly true when those elements are introduced into the wider speech community by adolescents whose own native variety is closer to the mainstream (although their initial motivation in acquiring them was accommodation to nonmainstream varieties as a part of the solution to the 'adolescent dilemma' outlined above). This interpretation is potentially an addendum to the 'weak ties' argument for linguistic change outlined in J. Milroy (1992). In suggesting that persons with low-density networks are likeliest to be the 'early adapters' (e.g., 183) in linguistic change, Milroy appears to overlook the age factor (in which younger, even adolescent, speakers seem to lead). If we use Eckert's terms to characterize the members of adolescent social networks, both 'Jocks' and 'Burnouts' have high-density group relations (corresponding to the high 'solidarity' values of the highest and lowest status speakers shown in Figure 7.1 of Milroy 1992:213). The 'weak ties' group (like the 'lower middle' and 'upper working' status sectors in Milroy's representation) would be the 'Neutrals' of Eckert's suburban Detroit adolescents. They are not among themselves a tightly-knit cohesive group, and they 'borrow' linguistically from the lowerstatus (or 'nonmainstream') burnouts. As such, they are in a privileged position to introduce such elements into the wider speech community.12 The young southeastern Michiganders reported on here have similar privilege, and, although we cannot know their various social status backgrounds, we can assume that their university status will have some effect on their later social position. The fact that they prefer a stigmatized variety to their own for affective characteristics suggests that they are not only changing their attitudinal perception to the 'Type D' outlined above but that they are also (potential) borrowers of norms from stigmatized speech communities in their own attempts to achieve a more casual, interpersonal style. Of course, a great deal more quantitative and qualitative work will need to be done to establish this direct link between perceptions, attitudes, and language change, but I believe such work will be a productive enterprise. Even if this direct link cannot be shown conclusively to exist, the patterns of language regard outlined here form an important part of the study of variation in its social context, one which has implications for both the more broadly-based ethnographic approach to sociolinguistics and linguistic intervention in schools, law, medicine, communications, and other areas of public concern.

Perceptual dialectology

95

5. What Next? I believe the many scholarly perspectives reviewed here on the folk identity of and regard for language varieties is worthy of the attention of dialectologists, sociolinguists, and students of the social psychology of language, and perhaps that is a foregone conclusion. I also believe, however, that there is a more general interest among social and cognitive scientists (including those who would like to apply their knowledge to such public venues as law, medicine, and education) in knowing what the folk believe about this most human of enterprises. If folk linguistics is of any value to such scholars, then this work surely shows that one of the dominating folk concerns in language is variety and pre- (and pro-) scription. Much of this work might continue as it has above, refining the methodologies and applying them to new situations. I also believe, however, that these findings will interest those who more centrally locate their practice in 'linguistics.' To take only one example, I believe that future work in the perception of variety might focus more specifically on the exact linguistic elements which give rise to perception rather than on the global presentation of varieties (or variety or area labels) in eliciting responses. Although eliciting folk imitations is one way of approaching this problem (e.g., Preston 1992, 1996a), the presentation of specific elements (by 'name,' by actual sample, or by computer-modified samples) for identification, placement, and evaluation by respondents is perhaps a better way to grasp even greater details of the 'triggering' mechanisms of language regard among the folk and of the potential influence of such regard on the more general processes of variation and change. One might argue that the presentation of actual speech samples makes some of these techniques more solidly a part of the 'language attitude' tradition, and, of course, the boundaries between that tradition and 'perceptual dialectology' are difficult to draw. In many cases, it is hard to determine whether the principal concern of a piece of research has been with the determination of folk sensitivity to regional speech boundaries or with an assessment of the respondents' 'attitudes' towards regional speech. Admittedly, when attitudes to regional labels rather than actual samples are given, I have personally assigned studies using this technique 'full status' in the perceptual dialectology enterprise. From one point of view, of course, the presentation of labels rather than actual speech samples for judgment is simply an alternative technique in the study of language attitudes. From another point of view, however, since attitudinal factors have been shown to be strong determiners

96

Dennis R. Preston

of the salience of areas themselves, any study of responses to regional speech is an integral part of perceptual dialectology. If we include social dialects as a part of our regular definition of 'dialects' (and I obviously believe we should, e. g., Preston 1993d:2-3), there is an even greater risk of perceptual dialectology's growth. It is clear, for example, that such studies as Labov (1966), which asked respondents in New York City to evaluate the social status of speakers on the basis of the frequency of nonprevocalic /r/ deletion or stop substitution for the interdental fricatives, were, according to the above definition, early examples of 'perceptual social dialectology.' When Graff, Labov, and Harris (1986) instrumentally manipulated the onset of the /aw/ diphthong of an African-American speaker from Philadelphia (in which the /a/ portion was 'fronted' to a position nearer /ae/), they succeeded in showing that this fronting alone (with no other alteration of the speaker's performance) was enough to signal 'white' ethnicity to both African- and European-American respondents. This was, as well, a case of 'perceptual social dialectology' with ethnicity as its principal target. In the long run, I doubt if we will be successful in showing in any ironclad way what 'linguistics' is (at least a 'linguistics' which includes the study of language variation and change) and what the 'social psychology of language' is. Perhaps linguists will be more interested in isolating the specific language features which trigger attitudinal responses and identifications, and social psychologists will be more interested in isolating the sociocultural forces which form and maintain the set of 'predispositions' responsible for 'attitudes,' but I am not at all surprised (nor disheartened) by the prospect of a great deal of interdisciplinary poaching. In conclusion, although I believe dialectologists should also be interested in the general regard speakers have for the speech from various areas, I suspect that folk delimitation of areas is better-suited to attitudinal perception rather than to the preparation of maps based on the occurrence of linguistic elements alone. On the other hand, if we carefully compare maps of features, responses to features, and responses to caricaturistic areas, we will surely have a better overall picture of the interests which engage us all.

Perceptual dialectology

Appendix: us state names

50

1 Washington 2 Oregon 3 California 4 Nevada 5 Arizona 6 Utah 7 Idaho 8 Montana 9 Wyoming 10 Colorado 11 New Mexico 12 Texas 13 Oklahoma 14 Kansas 15 Nebraska 16 South Dakota 17 North Dakota 18 Minnesota

19 20 21 22 23 24 25 26 27 28 29 30 31 32

Iowa Missouri Arkansas Louisiana Mississippi Alabama Tennessee Kentucky Indiana Illinois Wisconsin Michigan Ohio West Virginia

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

>S>

Ο

Virginia North Carolina South Carolina Georgia Florida Maryland Delaware New Jersey Pennsylvania New York Connecticut Rhode Island Massachusetts New Hampshire Maine Vermont Alaska Hawaii

97

98

Dennis R. Preston

Notes 1. The term 'perceptual dialectology' in relation to these studies was first used, so far as I know, in Preston 1981. If it were not for the common and unfortunate misunderstanding of 'folk' as 'false,' I would now prefer 'folk dialectology,' and that use would make it clear that this initiative is but one of any number of subareas of investigation in 'folk linguistics.' 2. According to Goeman (1989 [1999]), the first systematic investigation of dialect perception by nonlinguists was carried out by Willems (1886); although the data are unpublished, Goeman has represented Willems' findings in a 'little arrows' map. 3. The only later specific use of the 'little arrow' method I am aware of is Kremer (1984 [1999]), a study which asks to what degree perceptual dialect areas may or may not cross national boundaries (here the Netherlands - German border), where dialect similarity from a purely linguistic point of view may be great. In fact, the border is seldom crossed by 'little-arrow' identifications. 4. Unfortunately, the color-coded map which results from Daan's work is too detailed to include here. I have re-worked it (Figure 2) to show the main results, but those interested in the details of the characteristics on which it is based should, of course, consult the original. 5. Mase (1964a [ 1999]) goes on to develop similar procedures for drawing separate maps for greater degrees of difference, but they do not contribute methodologically and are not discussed here. 6. A similar technique is suggested by the most outspoken proponent of the need for perceptual data in drawing dialect boundaries. Jernudd (1968) simply contends that folk knowledge is an integral part of the scholarly representation of dialect divisions. He outlines a 'program' for such research in which he recommends eliciting folk responses to actual dialect features, a procedure used in very little of the previous work in Dutch-speaking areas or in Japan except in Nomoto (1963 [1999]) and Mase (196-fa [1999], 1964b [1999]). A similar strategy is used in the Welsh studies reported on here, and Diercks (1988) uses actual voice samples in determining regional speech awareness in a small area of northern Germany. 7. Texas and California (areas 8 and 12 in Figure 12) were excluded from the rating so as to limit the task to one large op-scan form ('electronically scorable answer sheet'). Since 'Texas' and 'Southwest' and 'California' and 'West Coast' overlapped considerably in the generalization of the hand-drawn task, this was not seen as especially detrimental. 8. Although the paired opposites were presented to the respondents with 'negative' and 'positive' sides randomly distributed, the 'positive' poles were all moved to the high (i.e., '6') end of the scale for all the quantitative analyses. I realized after I did this that there might be cultural misunderstandings of what I consider to be the 'positive' ends. They are 'Fast,' 'Polite,' 'Down-to-earth,' 'Educated,' 'Normal,' 'Smart,' 'Casual,' 'Good English,' 'Not nasal,' 'Friendly,' 'Speaks without a drawl,' and 'Speaks without a twang.' I apologize to readers who disagree with my assignments. That should not detract from the contents of the paper. 9. Hartley (1996, 1999) suggests that Ofegonians are aware of their diverse us origins (including southern backgrounds) and are reluctant on those grounds to evaluate other regions, but this interpretation would appear to be confounded by the fact that they were not reluctant to rate the same regions for 'pleasantness.'

Perceptual dialectology

99

10. There are many such comments in Niedzielski and Preston 1999. 11. These northern raters may have been Type D for some time, and the caricaturistically blunt 'pleasant' assessment I asked for in earlier work was simply not sensitive enough to elicit that aspect of their evaluation of southern speech. 12. Labov's 'lames' might also be seen as such 'early adapters.'

References Chambers, J.C. and Peter Trudgill 1980 Dialectology. Cambridge: Cambridge University Press. Coupland, Nikolas, Angie Williams, and Peter Garrett 1994 The social meaning of Welsh English: Teachers' stereotyped judgements. Journal of Multilingual and Multicultural Development 15,6: 471-89. 1999 'Welshness' and 'Englishness' as attitudinal dimensions of English language variation in Wales (A paper presented to the International Conference on Language Contact, Valencia 1995). In Preston (ed.), 1999, 333-43. Daan, Jo 1970 [1999] Dialekten. In Von randstad tot landrand. (Bijdragen en Mededelingen der Dialecten Commissie van de Koninklijke Nederlandse Akademie van Wetenschappen te Amsterdam, xxxvii). Amsterdam: N.V. Noord, Hollandsche Uitgevers Maatschappij. (Translated as 'Dialects,' in Preston (ed.), 1999), 9-30. Dailey-O'Cain, Jennifer 1997 Geographic and socio-political influences on language attitudes in Germany. Unpublished PhD dissertation. Ann Arbor, University of Michigan. 1999 The perception of post-unification German regional speech. In Preston (ed), 1999, 227-42. Demirci, Mahide and Brian Kleiner 1999 The perception of Turkish dialects. In Preston (ed.), 1999, pp, 263-81. Diercks, Willy 1988 Mental maps: linguistisch-geographische Konzepte. Zeitschriftfur Dialektologie und Linguistik 55:280-305. Eckert, Penelope 1989 Jocks and burnouts: Social categories and identity in the high school. New York: Teachers College Press. Goeman, A.C.M. 1989 (1999) Dialectes et jugements subjectifs des locuteurs. Quelques remarques de methode a propos d'une controverse. Espaces Romans (ttudes de dialectologie et de geolinguistique ojfertes ά Gaston Tuaillon, Vol. II). Universite Stendhal Grenoble 3: Ellug, 532-44 (Translated as 'Dialects and the subjective judgments of speakers: Remarks on controversial methods,' in Preston (ed.), 1999, 135-44. Goosens, Charlotte 1997 On the role of prosodic and verbal information in the perception of Dutch and English language varieties. Doctoral dissertation. Catholic University of Nijmegen (privately printed).

100

Dennis R. Preston

Gould, Peter and Rodney White 1974 Mental maps. Harmondsworth, Middlesex: Penguin. Graff, David, William Labov, and Wendell Harris 1986 Testing listeners' reactions to phonological markers of ethnic identity: A new method for sociolinguistic research. In David Sankoff (ed.), Diversity and Diachrony. Amsterdam/Philadelphia: Benjamins, 45-58. Grootaers, Willem A. 1959 Origin and nature of the subjective boundaries of dialects. Orbis 8, 355-84. 1964 (1999) La discussion autor des frontieres dialectales subjectives. Orbis 13:380-98 (Translated as 'The discussion surrounding the subjective boundaries of dialects,' in Preston (ed.), 1999, 115-29. Halliday, M.A.K 1976 Anti-languages. American Anthropologist 78,3:570-83. Hartley, Laura 1996 Oregonian perceptions of American regional speech. MA thesis. Department of Linguistics and Languages, Michigan State University, East Lansing, MI. 1999 A view from the West: Oregonian perceptions of us regional speech. In Preston (ed.), 1999,315-32. Hartley, Laura and Dennis R. Preston 1999 The names of us English: Valley girl, cowboy, Yankee, normal, nasal, and ignorant. In Anthony Bex and Richard C. Watts (eds), Standard English. London: Routledge, 207-38. Inoue, Fumio 1977/8 Högen Imeji no Tahenryo Kaiseki [Multi-variate analysis of dialect image] (part 1). Gengo Seikatsu 311:82-91. 1978/9 Högen Imeji no Tahenryo Kaiseki [Multi-variate analysis of dialect image] (part 2). Gengo Seikatsu 312:82-88. 1995 (1999) Classification of dialects by image - English and Japanese. In Wolfgang Viereck, (ed.), Verhandlungen des Internationalen Dialektologenkongresses, Bamberg, 29.7. - 4.8.1990; Proceedings of the International Congress ofDialectologists, Bamberg, July 29 - August 4, 1990. Zeitschrift fiir Dialektologie und Linguistik 75-77, 4 vols. Stuttgart: Franz Steiner Verlag, 355-68 (Reprinted in Preston (ed.), 1999, 147-59.). 1996 (1999) Subjective dialect division in Great Britain. American Speech 71,2:142-61 (Reprinted in Preston (ed.), 1999, 161-76). Jernudd, Björn Η. 1968 There are no subjective dialects. Kivung 1:38-42. Kremer, Ludger 1984 (1999) Die Niederländisch-Deutsch Staatsgrenze als subjective Dialektgrenze. In Grenzen en grensproblemen (Een bundel studies uitgegeven door het Nedersaksich Instituut van der R. U. Groningen ter gelegenheid van zijn 30-jarig bestaan = Nedersaksiche Studies 7, zugleich: Driemaandelijkse Bladen 36), 76-83. (Translated as 'The Dutch-German national border as a subjective dialect boundary,' in Preston (ed.), 1999, 31-36). Kuiper, Lawrence 1999 Variation and the norm. Parisian perceptions of regional French. In Preston (ed.), 1999, 243-62.

Perceptual dialectology

101

Labov, William 1966 The social stratification of English in New York City. Arlington, VA: Center for Applied Linguistics. 1972 Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Lambert, Wallace E., et al. 1960 Evaluational reactions to spoken language. Journal of Abnormal Social Psychology 60, 44-51. Lance, Donald M. 1994 Variation in American English. In Donald M. Lance and S. Kingsbury (eds), American pronunciation, 12th edition expanded, by John Kenyon. Ann Arbor: George Wahr Publishing Company, 345-74. 1999 Regional variation in subjective dialect divisions in the United States. In Preston (ed.), 1999,283-314. Long, Daniel 1990 (1999) Högen ninchi chizu no kakikata to yomikata [The drawing and reading of perceptual dialect maps]. Proceedings of the 50th meeting of the Dialectological Circle of Japan. Tokyo: Dialectological Circle of Japan, 7-16 (Translated and included as a part of 'Geographical perceptions of Japanese dialect regions,' in Preston (ed.), 1999,177-98). 1993 [1999] The role of linguistic features in perceptual dialect regions. In Andre Crochetiere, Jean-Claude Boulanger and Conrad Ovellion (eds.), Les langues menacees: actes du xve Congres international des linguistes, Quebec, Universite Laval aoüt 1992, Vol. 3. Sainte-Foy, Quebec: Les Presses de LTJniversite Laval, 371-74. (Included as a part of 'Geographical perceptions of Japanese dialect regions,'in Preston (ed.), 1999, 177-98). 1995 Högen ninchi chizu [Perceptual dialect maps]. In D. B. West (ed.), Pasokon kokugo kokubungaku [Japanese linguistics and language on the personal computer]. Tokyo: Keibunsha, 157-71. 1997a Nihon no Högen Ninchi Chizu Senkö Chizushü [A selected atlas of Japanese perceptual dialect maps]. Nihongo Kenkyü Sentä Hökoku [Japanese Language Research Center Reports] 5:45-84. 1997b The perception of 'standard' as the speech variety of a specific region: Computer-produced composite maps of perceptual dialect regions. In Alan Thomas (ed.), Issues and methods in dialectology. Bangor: University of Wales, Department of Linguistics, 256-70. Mapping nonlinguists' evaluations of Japanese language variation. In Preston 1999 (ed.), 1999, 199-226. Mase, Yoshio 1964a (1999) Högen ishiki to högen kukaku. In Nihon hogen kenkyukai, 270-302 (Translated as 'Dialect consciousness and dialect divisions,' in Preston (ed.), 1999, 7199). 1964b (1999) Högen ishiki ni tsuite: Washa no genkyushita högenteki tokuchö. Naganoken Tanki Daigaku Kiyö [Collected Papers of the Nagano Junior College] 18:112 (Translated as 'On dialect consciousness: Dialect characteristics given by speakers,' in Preston (ed.), 1999, 101-13.). Milroy, James 1992 Linguistic variation and change. Oxford: Blackwell.

102

Dennis R. Preston

Milroy, Lesley and P. McClenaghan 1977 Stereotyped reactions to four educated accents in Ulster. Belfast Working Papers in Language and Linguistics 2 (4). Niedzielski, Nancy and Dennis R. Preston 1999 Folk linguistics. Berlin: Mouton de Gruyter. Nomoto, Kikuo 1963 (1999) Kotoba no ishiki no kyokai to jissai no kyokai. Jinruikagaku [Anthropological Sciences] 15:271-81 (Translated as 'Consciousness of linguistic boundaries and actual linguistic boundaries,' in Preston (ed.), 1999, 63-69. Preston, Dennis R. 1981 Perceptual dialectology: mental maps of United States dialects from a Hawaiian perspective. In H. Warkentyne, (ed.), Methods/Mithodes iv [Papers from the 4th International Conference on Methods in Dialectology], Victoria, B. C.: University of Victoria, 192-98. 1985a Mental maps of language distribution in Rio Grande do Sul (Brazil). The Geographical Bulletin 27:46-64. 1985b Southern Indiana perceptions of 'Correct' and 'Pleasant' speech. In Henry Warkentyne (ed.), Methods/Methodes ν (Papers from the Vth International Conference on Methods in Dialectology), University of Victoria, British Columbia, 387-411. 1986 Five visions of America. Language in Society 15, 221-40. 1988a Change in the perception of language varieties. In J. Fisiak (ed.), Historical dialectology: regional and social. Berlin, New York, Amsterdam: Mouton De Gruyter, 475-504. 1988b Methods in the study of dialect perception. In A. Thomas (ed.), Methods in dialectology. Clevedon, Avon and Philadelphia: Multilingual Matters, 373-95. 1988c Sociolinguistic commonplaces in variety perception. In Kathleen Ferrara et al. (eds), Linguistic change and contact: NWAV-XVI. University of Texas, Department of Linguistics 279-92. 1988d The nicest English is in Indiana. Studia Germanica Posnaniensia 14:169-93. 1989a Standard English spoken here: The geographical loci of linguistic norms. In U. Ammon (ed.), Status andfunction of language and language varieties. Berlin & New York: Walter de Gruyter, 324-54. 1989b Perceptual dialectology. Dordrecht: Foris. Talking Black and talking White: a study in variety imitation. In J. Hall, N. 1992 Doane, and D. Ringler (eds), Old English and new. New York: Garland, 327-55. 1993a Folk dialectology. In Preston, 1993d, 333-77. 1993b Folk dialect maps. In Wayne Glowka and Donald M. Lance (eds), Language variation in North American English: Research and teaching. New York: Modern Language Association of America, 105-18. 1993c Two heartland perceptions of language variety. (Revision of 'Standard English spoken here' [1989]). In T. Frazer (ed.), 'Heartland' English. Tuscaloosa: University of Alabama Press, 23-47. 1996a Whaddayaknow: The modes of folk linguistic awareness. Language Awareness 5,1:40-74. 1996b Where the worst English is spoken, in Edgar Schneider (ed.), Focus on the USA. Amsterdam & Philadelphia: Benjamins, 297-360.

Perceptual dialectology

1997

103

The South: The touchstone. In C. Bernstein, Thomas Nunnally, and Robin Sabino (eds), Language variety in the South. Tuscaloosa and London: University of Alabama Press, 311-51. Preston, Dennis R. (ed.) 1993d American dialect research. Amsterdam and Philadelphia: Benjamins. 1999 A Handbook of perceptual dialectology. Amsterdam: Benjamins. Preston, Dennis R. and George M. Howe 1987 Computerized generalizations of mental dialect maps. In Denning, et al. (eds), Variation in language: NWAV-XV at Stanford, Stanford: Department of Linguistics, Stanford University, 361-78. Rampton, Ben 1995 Crossing: Language and ethnicity among adolescents. London: Longman. Rensink, W.G. 1955 [1999] Dialectindeling naar opgaven van medewerkers. Mededelingen der centrale commissie voor onderzoek van het nederlandse volkseigen 1, 20-3. (Translated as 'Informant classification of dialects' in Preston (ed.), 1999, 3-7). Ryan, Ellen B. and Howard Giles 1982 Attitudes towards language variation. London: Arnold. Ryan, Ellen B., Howard Giles and Richard Sebastian 1982 An integrative perspective for the study of attitudes towards language variation. In E.B. Ryan and H. Giles (eds), Attitudes towards language variation. London: Arnold, 1-19. Shuy, Roger W. and Ralph W. Fasold 1973 Language attitudes: Current trends and prospects. Washington, D.C.: Georgetown University Press. Sibata, Takesi 1959 [1999] Högen kyokai no ishiki [Subjective consciousness of dialect boundaries]. Gengo Kenkyu 36:1-30 (Translated as 'Consciousness of dialect boundaries' in Preston (ed.), 1999, 39-62). Trudgill, Peter 1972 Sex, covert prestige and linguistic change in the urban British English of Norwich. Language in Society 1(2): 179-95. Tucker, G. Richard, and Wallace E. Lambert 1969 White and Negro listeners' reaction to various American English dialects. Social Forces 47,463-8. Weijnen, Antonius A. 1946 De grenzen tussen de Oost-Noordbrabantse dialecten onderling [The borders between the dialects of eastern North Brabant]. In Antonius A. Weijnen, J.M. Renders, and Jac. van Ginneken (eds), Oost-Noordbrabantse dialectproblemen [Eastern North Brabant dialect problems]. Bijdragen en Mededelingen der Dialectencommissie van de Koninklijke Nederlandse Akademie van Wetenschappen te Amsterdam 8, 1-15. 1961 Het bewustzijn van dialectverschil [The awareness of dialect differences], ( Voordracten Gehoudes voor de Gelderse Leergangen teArnhem 5). Groningen: J. B. Wolters. 1966 Nederlandse dialectkunde [Dutch dialectology]. Assen: Van Gorcum. 1968 Zum Wert subjektiver Dialektgrenzen. Lingua 21:594-96 (Translated as 'On the value of subjective dialect boundaries' in Preston (ed.), 1999, 131-33).

104

Dennis R. Preston

Weinreich, Uriel, William Labov and Marvin Herzog 1968 Empirical foundations for a theory of language change, in Winfred P. Lehmann and Yakov Malkiel, (eds), Directions for historical linguistics. Austin: University of Texas Press, 95-195. Williams, Angie, Peter Garrett, and Nikolas Coupland 1999 Dialect Recognition. In Preston (ed.), 1999, 345-58.

Parallels in the phonological systems of Southern Dutch and Northern French dialects Georges de Schutter

1. Introduction It is not the first time, not by far, that attention is called to the existence of linguistic parallels across language borders within a geographic area. The issue has been discussed throughout the last century, one of its most prominent instances being the so called Balkan Sprachbund ("language cluster"): scores of phonological, morphological and syntactic phenomena appear to be similar in such languages as Modern Greek, Albanian, Turkish, DacoRomanian, Macedo-Romanian, Serbo-Croatian, Bulgarian and (Slavonic) Macedonian. Most of those similarities cannot be explained by genetic relationships, which are weak or even non-existant between many of the pairs to be formed from this list. This text does not deal with such a well-established and generally acknowledged case: it is about a contact zone that has been off and on in the interest of - almost exclusively - Dutch dialectologists for the last sixty years or so. It was first described by the charismatic, but in no way uncontroversial Dutch linguist J. van Ginneken (Van Ginneken 1934), and its study culminated in a number of publications by one of Van Ginneken's successors at Nijmegen university, A.A. Weijnen (Weijnen 1964,1967), almost as charismatic as a scientist, and much more conscientious in handling the data provided by research, much of this carried out by Weijnen himself, for that matter. More recent studies by Van Hoecke (1988) and Ryckeboer (1991, 1997) rekindled the fire of interest. The contact zone in question is fairly restricted in size, and the scope of phenomena implied seems less spectacular also than in the Balkan case. The area comprises the dialects spoken in and around the Western European geographic entity known (historically) as the Low Countries: the contact zone between Continental West-Germanic and North-Romance dialects, in political terms, the border region between the medieval Kingdom of France and the Holy Roman Empire. In that area northern Langue d'Oi'l French

106

Georges de Schutter

(especially Picardian, Walloon and Lorraine dialects) meet with Germanic (especially Flemish, Brabants, Rhenanian, Palatinatian and Alsatian dialects). Van Ginneken was prone to bed this cluster into a wider scope, and included all other Franconian, western Saxonian and Frisian dialects on one side of the language divide, and Normandy and even Ile-de-France varieties on the other. In this text I will focus mainly on the direct contact zone between southern Dutch dialects in Belgium and the southern part of the Netherlands, and the adjacent French dialect area in southern Belgium and the French departments Nord and Somme. What is so peculiar about this complex dialect area? And is it really possible to characterize the Germanic and Romance dialects currently or formerly spoken there as a Sprachbund? To begin with, it is certainly possible to find not only lexical,1 but also structural correspondences between the regional languages spoken on both sides of the long-standing linguistic border. There are at least two types: one that is probably due to borrowing, surprising sometimes, but fundamentally not so interesting for our purpose; and one that, in my opinion, cannot possibly be due to (rather superficial) borrowing, and is therefore a real challenge to interpretation. I will give some examples of both types of phenomena.

2. Borrowing as a source of cross linguistic parallels In the course of history any dialect is subject to change. Linguists are convinced of the internal motivation of such changes. The dialectologists among them agree with that view, but add external factors, called contact and expansion: languages and dialects are not isolated entities, they affect each other, and this influence may either be occasional (mainly in the lexical field) or, alternatively, it may affect the structure of the language system. It is generally believed that both types of change (both casual and structural) imply expansion: if languages which occupy adjacent territories, or even share a common area, show parallel evolutions over more or less extended parts of those territories, expansion may be in the game. The expansion itself may be explained by the presence of one or more important centers within the geographic domain, centers where the change first originated, and from which it spread. Another question, and a central issue in this context is, whether the first incentive towards the change in one territory may reasonably be attributed to borrowing from the language system used in another, adjacent, area. I will go into this question only in section 4. First I want to give some exam-

Parallels in the phonological systems

107

pies of phenomena that contributed to the development of Van Ginneken's and Weijnen's ideas, and to add some new findings on the subject. Our main interest in this context is the parallel geographic distribution of oppositions on both sides of the language border. Here the geographic dimension enters the discussion as an a priori independent variable: if a choice between plus and minus can be made for any phenomenon, we should take it for granted that each dialect is free to go for either. Each deflection from that independent status, including geographic clustering of dialects with and without the phenomenon, may be an indication of affinity of some kind - possibly simply contact - between dialects, even across linguistic borders. 2.1. The first fact to be presented here was first mentioned by Van Ginneken (1934: 290-292). I will not take over his interpretation of the facts, however, but will join the Belgian Romance scholar Van Hoecke (1988). Consider the following situation: south of the Germanic-Romance dividing line, a pattern originates, by which original [ε], in certain environments, changes into a rising diphtong [je]; in this way the stem of the verb perd(re) "lose" turns into pje(r)d~. Nothing prevents the same phonological change from occurring, independently from the rule in Romance dialects, somewhere to the north of the linguistic border, provided that the same input is available; this happens to be the case, e.g. with the input wordperd (Middle Dutch 'horse'). And, as a matter of fact, there are Germanic dialects in which perd turns into pje(r)d, an exact match of the French stem. Let us be clear, the chance that such parallel evolutions originate independently in the two language systems, is fairly large: breaking of vowels, especially if they precede /r/, is attested in a wide variety of languages of the world. As far as our geographic domain is concerned, however, the change appears to be local in both linguistic parts. That means that it did not affect all varieties, neither of southern Dutch nor of northern French: it is limited geographically to certain areas of both, and more specifically: to adjacent parts of both language areas. The chance that this is due to independent changes is substantially smaller, of course, but coincidence cannot be excluded on the basis of sheer vicinity. The probability of coincidence decreases futher, however, even to the point of disappearing completely, if we have a careful look at the isophones in both areas: the combined areas appear to form one large, compact geographic entity, cf. map 1: it encompasses the medieval County ofNamur, the Romance speaking parts of the Duchy of Brabant, and of the Princedom of Liege, and then also portions of the Germanic-speaking parts of Brabant and

108

Georges de Schutter

Liege. It does not extend further west, to the old Counties of Flanders (which was predominantly Germanic) or Hainaut (almost completely Romance). The geographic distribution suggests that the breaking rule is a very old rule, bound to political entities that have lost their impact centuries ago. Map 1: Breaking of vowel /e/ before Irl in southern Dutch and northern French dialects

Legend: + area of dialects with the breaking rule area of dialects without the breaking rule + / - area of dialects with irregular application of the breaking rule

The distribution of the phenomenon also suggests, contrary to Van Ginneken's view, that the rule originated in the Romance-speaking regions and expanded to the north, across the linguistic border. There its progress was not steady any more: a lot of local dialects, even a relatively very large number situated near the boundary of Romance territory, do not have the

Parallels in the phonological systems

109

breaking rule. They are situated between Brussels, the heart of the Duchy of Brabant, and the western part of the old county of Loon, the Germanic speaking part of Liege (cf. section 4 for a possible explanation of this distribution). This leaves us with two separate core regions, one in the western part of Brabant, and one in Loon (nowadays the Belgian province of Limburg); in between, the Germanic dialects are at variance concerning the rule, which, moreover, shows lexical diffusion. A further indication of "Romance priority" may be that, according to the data mentioned by Van Hoecke (1988: 192-193), the rule has found a reflection much earlier in spellings in Romance texts than it was ever recorded in Dutch ones. And, even more convincing: rising diphtongs play a far more important role in the phonology of virtually all Langue d'O'il dialects (e.g. in words like piece, tiede,fier) than they do in Franconian-based Dutch (where it is mostly restricted to the /r/ + dental consonant type. The evidence, even combined, is not compelling, of course, but all aspects converge in one conclusion: if borrowing has to be considered as a determining factor, which is probable, indeed, it is almost certainly the Dutch area that took over a phenomenon originally bound to Romance territories. 2.2. I want to give one further illustration of a feature that may have been borrowed as such by part of one linguistic community from part of the other one: the apocope, vs. retention of fh/. Apocope of /h/ is common in the whole territory of Gaul where French has become the native language, with the exception of the eastern Walloon dialects in Belgium (province of Liege and the larger part of Romance-speaking Luxemburg). Dutch dialects have mostly kept the /h/ intact; the major area where this is not the case, is the western part of the southern dialects: the whole of Flanders and Zealand, and large parts of North Brabant in the Netherlands, Antwerp and Flemish Brabant in Belgium, have lost fhJ from their phoneme inventary. Now, again, the isophone between the areas with contrasting systems appears to cross the language border at one point, cf. map 2. There is no evidence, this time, that it is the Germanic dialects that have taken over a typically French rule (h-apocope): therefore the Dutch speaking area over which the phenomenon of h-apocope has spread is too vast, and extends to regions (e.g. the northern part of Brabant) that can not readily be associated with any influence from regional (non-literary) French. Apocope of /h/ has, furthermore, been attested from the earliest records of southwestern Dutch (written in Flemish dialect). If borrowing is at stake here, it is most probably the northeastern Romance dialects that have taken over a "Germanic" rule, that of /h/-mainte-

110

Georges de Schutter

Map 2: Distribution of the phoneme /h/ in southern Dutch and northern French dialects

Legend: + area of dialects with phoneme /h/ area of dialects without phoneme /hi nance (or reintroduction). Both the Princedom of Liege and the Duchy of Luxemburg (the latter was predominantly Germanic speaking, but included large, though sparsely populated Romance territories) were very closely affiliated politically, and in matters of religion and culture, to the Holy Roman Empire. Further evidence comes from the fact that /hi in those French dialects is all but restricted to words of Germanic origin (it appears e.g. in hair "hate", haie "hedge", hache "ax", but does not e.g. in Latin-based hier "yesterday", heure "hour", honnete "honest"2). It is most natural, then, to think of direct influence from the Germanic part of the area we consider: /h/ as a phoneme must have been reintroduced in regional French, together with the loans in which it occurred; this has taken place in a region which was in

Parallels in the phonological systems

111

constant interaction with areas to the north, and especially to the east, in which Germanic was the dominant language.

3. Parallel rules due to underlying common tendencies Facts like those mentioned in section 2, may be very interesting in their own right, but they are just individual cases, wide apart from a structural point of view. Structural linguistics takes an interest primarily in the interaction of rules as instances of complexes of phenomena, based on and guided by very distinct tendencies. An undisputably more interesting question is whether such tendencies exist in geographical clusters of originally unrelated dialects. As a matter of fact, Van Ginneken (1934) gives some examples of such complexes: - It has generally been noted that, at least in Continental Germanic, there is a long-standing tendency to turn tense closed (high) monophthongs into diphthongs, and open diphthongs into long monophthongs. These tendencies are restricted, geographically, to parts of the linguistic area, however. The same applies to Romance dialects adjacent to the language border. - The same holds for another very distinct tendency which causes sets of either vowels or consonants to palatalize or depalatalizse. Such (de)palatalization tendencies may occur spontaneously (i.e. without being bound to any specific phonological or morphological environment) or only under certain formal conditions. An instance of the latter process is the general Germanic phenomenon of vowel umlaut. In this paper I want to look somewhat more closely into the latter set of phenomena as they occur in the southern Dutch language area: - Spontaneous palatalization of vowels has a predominantly western distribution, spontaneous palatalization of consonants (especially /d/ => /j/), on the contrary, is weaker in the west than it is in the centre or the east. - Conditioned palatalization of vowels ("umlaut") has its focus in the east, and becomes gradually eaker as we proceed to the west. - Conditioned palatalization of consonants occurs in the whole area, but shows very distinct east-west-oppositions in specific applications (see e.g. the diminutive suffix: fan => {k3(n)/tjd(n)}). This may be interesting in its own right, of course, but what is really amazing, is that a lot of oppositions based on the presence/absence of those - and

112

Georges de Schutter

many other - tendencies recur south of the language border. They surface not necessarily in the same oppositions (i.e., in the presence or absence of any specific rules), not by far; specific rules reflecting the tendency may be different in all possible respects: as to the specific phonological context in which they apply, as to the output, or the conditions required. That is precisely what makes these cases so interesting, direct borrowing being absolutely out in such conditions. Lack of place compels me to make a selection from the relevant data at hand. I will present only some evidence on the palatalization-depalatalization tendencies in consonants. I hope I can make it clear that such cases lead us into a domain where simple borrowing as an explanation of parallel evolutions loses a lot, if not all of its appeal. Palatalization of consonants is a fairly complex domain: palatalization of dental consonants and that of velar consonants appear to be largely independent phenomena, bound, also, to different geographic areas. Both of them are interesting for our case, but I will restrict the discussion to dental consonants in this paper.

3.1. Palatalization of dental consonants in northern Romance dialects It is a well known fact that in Standard French palatalization has affected only the liquid IV and the nasal /n/, yielding /j/ and /ji/ respectively (folia => feuille, vinea => vinia vigne). The expansion of this phenomenon for Late Latin /l/, followed by /j/ in the northern French dialects is represented on map 3: in the words aguille, bouteille, feuille ("needle, bottle, leaf') a fairly massive central and eastern area displays the feature [+ palatalization], a smaller part in the west (the larger part of the Picardian dialect area) retained the old IV sound (cf. ALW, maps 1, 6 and 42). The origin of the forms is quite clear: the western Picardian dialects dropped /j/ at the end of the word, possibly after it had developed to a vowel /i/, which itself could have turned into shwa before disappearing from the phonetic scene. Further to the east IV was palatalized under the influence of /j/, and eventually merged with that sound, yielding simply /j/, as in Standard French [fcej] (feuille), [butej] (bouteille), [egij] (aiguille). Not all types of palatalization in the Romance dialects under consideration are as easy to explain. Let us turn to the forms deriving from Latin scala (French echelle 'ladder'). There is no palatalizing factor after IV in the original form; still, a number of northern dialects of French have forms like

Parallels in the phonological systems

113

Map 3: Distribution of palatalization phenomena in dental consonants in southern Dutch and northern French dialects

Legend: + area with extensive application of palatalization rules area with depalatization phenomena 1 isogloss /V => /j/ in aiguille, bouteille, feuille 2 westernmost application of the rule eskelle => eskyeye 3 easternmost application of the rule eskelle => eskyeye 4 westernmost application of partial palatalization in dental consonant clusters 5 isogloss of complete palatalization in plural pronouns (5a-5b: western-, easternmost distribution ofy'ar-forms) 6 easternmost application of partial palatalization in dental clusters

eskyeye, with palatalization of /]/. But: the geographic distribution of such word forms is far more restricted than in words with such a factor: only the eastern part of the Picardian-speaking province of Hainaut in Belgium has them. It must be noted that the [scala]-forms with palatalization of 71/ have a

114

Georges de Schutter

palatal sound preceding the original IV: eskyeye (note that here again a rising diphthong was created, cf. section 2.1). This makes it very probable that there are two different types of assimilation processes at work: one goes right-to-left ("regressive / cataleptic" assimilation, the folia type); this type is very common in northern French dialects, and also further to the south, for that matter. And then there is the left-to-right type (or "progressive / analeptic" assimilation, the scala type). As far as our geographic area is concerned, the latter is restricted to a tiny portion, near the isophone between the domains with and without left-toright assimilation, but within the area where that type of assimilation applied. We might point at this very small area, actually, as the core of dental consonant palatalization, since it shows both forms of palatalization processes.

3.2. Palatalization of dental consonants in Dutch dialects Let us now turn to the southern Dutch dialects. Palatalization of dental consonants is found in many Dutch dialects, in a large number of cases; thee intensity and frequency of its occurrence is extremely variable, however. I will discuss four phenomena that may be of interest for our problem. A. The change of intervocalic /d/ to /j/, as in poeder => poeier "powder". This evolution is common in most southern Dutch dialects, and has affected many, though not all words with intervocalic /d/ (there is "lexical diffusion"). The westernmost dialects in West and French Flanders have the rule only in a very limited number of word forms; an explanation might be that originally these dialects did not develop the rule at all, and that whatever palatalized forms are current in the present day language are probably due to borrowing from more eastern dialects.3 The case is interesting: West Flemish dialects are adjacent to the western Picardian dialects without palatalization of the dental consonants Ν and /n/. As things are, this may be another case of "abstract" parallellism across the linguistic border: the phonemes affected are different, but the effects of the rules are parallel. We cannot add any precise geographic delimitation for the Dutch geographic area, however, because the palatalization rule happens to compete with another, independent one, which simply deletes /d/ together with the following shwa in intervocalic position, e.g. voeder => voer "fodder", but also, again: poeder => poer "gun powder". The fact that there are two rules,

Parallels in the phonological systems

115

both of them subject to lexical diffusion in the dialects in which they occur, explains the cooccurrence of forms like poeder, poeier, poer in the Dutch standard language.4 It is, now, the case that forms with syncope of dental (+ shwa) are especially abundant in the area situated east of West Flanders: the Belgian province of East Flanders and the southern Brabant region. So it is simply impossible to match the northern French isophone for the change /V => /j/, with its southern Dutch counterpart for /d/ => /j/: the whole natural contact zone for the isophone has the syncope rule bleeding its palatalization rival. B. We are much luckier, in this respect, with a second case: palatalization applied to the set of compound plural pronouns. A large area, especially in the south of the Dutch language territory has replaced the Middle Dutch simple personal pronouns wi,ji/gi, si 'we, you (pi.), they' with more iconic complex forms, by adding the word for "people" (lieden/luden) to them: wit gi/zi+li/ydQ (cf. Goossens 1994: 127-131); in the west these forms were further suffixed with /r/, originally a genitive marker, yielding wilidsr, etc.; but that does not affect the present discussion. The resulting forms provide a most interesting picture: after complete vowel reduction in unstressed syllables (applied to li/ly-d in the second syllable of the original form), a combination of dental consonants /Id/ was formed, preceded by a palatal vowel: originally simply /i/, in some regions diphthongised to /ei/: wild(or)/wEild3(r), etc. Let us now look at the map of present-day southern Dutch dialects (Map 3): - The extreme west (West Flanders and most of French Flanders) has forms of the widsr-gidar-zidsr-type: /l/ has been assimilated to /d/, the same phenomenon very probably, that we see in the change from uld(o) (itself going back to old(s), "old") to ud(d) in the present-day West Flemish dialects. There is no sign of any form of palatalization. - Adjacent to the east (province of East Flanders) are forms in which /Id/ was maintained, mostly with rounding of the preceding vowel (weldar, geldor, zeldar). Note that there is no palatalization of the consonant cluster in this area, either. - We now skip one small portion near the language border, and proceed towards the east: on Brabants territory we find forms of the welie/web type;4 they must have derived from underlying forms like *wil(i/y)dv, data from the period in which they have come into use (the early Middle Ages) are scarcely available. So we cannot really be sure how they developed;

116

Georges

de

Schutter

the following conjecture (for the 2nd person plural) may be close to what really happened: gilode > gilds > gil 'da > gili (>gih) I have to point out that this derivation is not generally accepted in the small world of Dutch historical linguists; it explains more of the actual forms in the Dutch dialects, however, than does its alternative (gi 'lids > gi'li > 'gili (>gih),5 proposed by i.a. Van Loey (1958b), and taken over by Jan Goossens in his recent discussion of the complex pronouns (1994: 130-131); I cannot go into this issue here. If my proposal is correct, the complex pronouns are treated like other words with dental clusters in a palatalizsing environment in the large area consisting of the former principality of Brabant: 6 palatalization is applied, but it is not carried to the limit (that would have been simply /j/). Here /lj/ is the ultimate palatal stage that may somehow be called "consonantal"; there is no problem with the hypothesis that /lj/ develops further to /li/: that is a very common development in most dialects of Dutch, cf. e.g. Latin folium => f o l ' j s => f o l i e , Ofr. traille => tral 'j9 => tralie, etc., both in Brabants and in Holland dialects. - What then about the little corner near the province border between East Flanders and Brabant? Here Irl appears, as an ending (yielding forms like [wejr] as well as it does in the west; this points to an underlying form weildor etc. How are forms like we.jr etc. to be derived from that base? The most natural development string seems to be the following one: the palatal element of the diphthong /ei/, which developed from /V (cf. above), merged with the dental cluster as a whole; contrary to the Brabants rule, the palatalization was carried to the absolute end, eventually yielding /j/: wE.jr-geJr-ze.jr. This /j/ was dropped before /rΡ in a number of local varieties, which brings us to forms like we.r-ge.r-ze.r (or parallel forms with rounded vowel). In other dialects of the region the cluster /jr/ developed into Ijorl·. we.jr etc. Let us sum up our findings for the complex personal pronouns: no palatalization occurs in the west (the "Flemish" dialects); it does appear in the center of the Dutch-speaking area (the Brabants dialects); it is precisely on the border between both geographic entities that the palatalization rule reaches its height. C. The core status of the small pocket of land near the East Flemish-Brabants border becomes even more apparent if we take other words with dental clusters into account. The same area that has absolute palatalization with the

Parallels in the phonological systems

117

complex pronouns, is the center of a somewhat larger area, which shows another type of palatalization in the following categories of words:8 - First of all, there are Romance loans ending in dental clusters, such as (original)point, turning into [puptj], saint, which survives as [saßtJ], etc. In those words the dental cluster was originally preceded by /if. - A second group consists of words containing a stem ending in dental consonant, followed by a suffix which originally contained /i/. The most salient categories are: third person singular verb forms, and superlatives of adjectives, e.g.: Old Germanic find-ip —> (hij) vceptf, fin-ists —> (het) fceptf'Xhe) finds, (the) finest". - Some stems (or old derivations) originally displayed the same phonological make up; instances are: manisks —» mcepj "human", muj (from mut(i)s, a loan from medieval Latin) "cap". - Finally there are stems in dental clusters, which were followed by a palatalization factor (palatal vowel or glide), at least in some parts of the paradigm; e.g. tand(i)s —» taptf "tooth". This type of palatalization is limited to morpheme-final palatal consonant clusters, including original geminates like, e.g., fat J < [fatid]. Some forms (with intervocalic clusters) seem to be formally akin to the complex pronouns. Common to both types of palatalization is, that there is a palatalizing factor in the environment: though it occupies different positions in relation to the dental cluster (either preceding or following in some underlying forms, or even between the members of the cluster in other ones), it is very plausible that at one time they all were in front of the cluster as a whole (cf. De Schutter 1993). From this stage on, there would be another parallel with the complex- pronouns rule. The output of the rule is different, however: palatalization has not been carried out to the limit (which would have yielded /j/ as with the pronouns,9 but resulted in a combination of the direct palatal counterparts of the original dental consonants, e.g. Ad, jid/ for /ld,nd/. D. The last item in this selection of relevant facts concerns depalatalization, which occurs in a great number of (south)-western dialects of Dutch: the palatal fricatives /IV, /;$/ and ///, whatever their origin, turned into dental /s/ or Izl (e.g. tra.l' => tra.l, garazo => garazo, file => sik, taptf9 =» tantsQ, tfup => tsup ("garage, chic, little tooth, top"). This phenomenon is especially

strong in the western half of the province of East Flanders; this region, adjacent to the core area of the palatalization tendency, consequently presents itself as the stronghold of refutation of the palatalization tendency. At first sight it is most remarkable that in this way the core regions of palatalization

118

Georges de Schutter

and depalatalization are adjacent. At second sight, the amazement may well be overcome by the realization that strong tendencies tend to provoke strong anticurrents. Let us sum up our findings: palatalization of dental consonants is not current in the western part of southern Dutch, where we rather find depalatalization rules, if anything. It is current, however, in the centre and east of the language area. The isophone (or rather the bundle of isophones) between both regions is a perfect match to the isophone between the western and centraleastern parts of the northern Romania. Within the region with palatalization there is a tiny part, both on Germanic and on Romance soil, where palatalization has been carried out further than in the overwhelming majority of dialects concerned. These occupy the extreme west of the area with the palatalization rule. Map 3 suggests that there has been some reduction on Romance territory: the western border of the region with (at least some form of) palatalization is a few kilometres further to the east than its Dutch counterpart. That may well be due to relatively recent expansion from the west: it must be noted that the Picardian speaking region without any palatalization rules contains cities which have been very influential from the Middle Ages until modern times: Lille, Douai, Tournai, Arras, etc. Such economic centres must have had a linguistic influence on their surrounding areas, and the relative weakening of the palatalization tendency at the periphery might be a result of that. What is most interesting in this palatalization case, is that, again, the tendency surfaces in quite different forms in both languages: the input consonants (and/or consonant clusters) affected by the rule are different, and so are the outputs of various rules. Those rules are governed by a wide variety of conditions: there is spontaneous palatalization, but also both cataleptic and analeptic assimilation. Not one single rule to be derived from the data I have, appears to be the same on both sides of the great linguistic divide, but there cannot be any serious doubt that the same fundamental tendency is at work both in Germanic and in Romance dialects. There cannot be doubt either, that the strength of that palatalization tendency is parallel in matching regions on both sides of the linguistic border: The translinguistic continuum we find here is practically perfect. Essential, however, is that the continuum surfaces only if we combine data concerning different specific rules.

Parallels in the phonological systems

119

4. Conclusion The central thesis of this paper was, that parallels on both sides of a linguistic border may be due to at least two causes: borrowing on the one hand, the persistence of long standing tendencies on the other. We have to take into account that for borrowing of phonetic rules to take place, there must have been very intense contact between both groups of native speakers. Has this been the case in the region we have been discussing? It is generally agreed upon that Germanic and Romance speaking groups have mixed throughout the Middle Ages: cities like Brussels and Louvain in the Dutch speaking area of Brabant may have had their administration in the dominant language of the area throughout the late Middle Ages, it is beyond any serious doubt, however, that there has been a steady migration from the southern Romance speaking region, especially in Brussels. Very probably this migration caused bilingualism in large portions of the population, and rules like the vowel breaking rule before /r/ may have passed from one language into the other, first in occasional lexical parallels (among these both native words and loans), then eventually in the whole relevant lexicon. The same may have been the case in the predominantly Romance speaking Princedom of Liege, whose (French language) capital has always been an attraction pole for people from the old County of Loon, the present Belgian province of Limburg. Here may have been the place of diffusion of both the vowel breaking rule in the Germanic east (possibly brought about by temporary migrants after they returned home) and the expansion of /hi, on Romance soil, in words that were akin to Germanic parallels (borrowed from the speech of Germanic-speaking groups in the city and Princedom of Liege and other economical and cultural centers of the Princedom). However intense the contacts between singular groups may have been, it does not seem natural that anything beyond isolated rules should have been passed on in the way suggested, i.e. by borrowing. There must be another explanation for the rather abstract parallels, of which I discussed two in short terms. There should have been a much longer, and an even more intensive contact to account for such a fundamental complex of tendencies to run fundamentally alike. In our case we may think of two such contact types: - Prior to the well recorded phase with occasional language contacts, there must have been a long period of active bilingualism. This period followed the invasion of Germanic tribes in northern Gaul, and it must have extended over geographic entities where either Germanic or Romance dialects are spoken now. Historical findings are in accordance with this

120

Georges de Schutter

hypothesis: the larger part of what is now Belgium, and regions to the south of that must have been actively bilingual for centuries, from the sixth to the tenth, in some areas to the eleventh century, or even beyond. This period may have been long enough to lock a number of common tendencies in the competence of speakers, tendencies which became persistent after populations had, eventually, switched to monolingualism. - Theoretically it is even possible to go further back: both Romance and Germanic are imported languages in this formerly Celtic speaking area. It is possible, in fact even probable, that population groups transferred tendencies existing in their original language(s) into the newly acquired languages. By the way, we may assume that old and new languages have coexisted for some time, much in the same way as what I suggested with respect to Germanic and Romance in the next few centuries. The tendencies eventually surfaced either in new rules, alien to the newly acquired languages until then, or in the refutation of such rules, on the way of development. This might account for at least some of the divergent rules that, eventually, surfaced on both sides of the linguistic border.

Notes 1. Cf. For this aspect, i.a.: Weijnen (1964) and Ryckeboer (1991, 1997). 2. Compare, e.g. the maps for the words hache-haie-herse (maps 49, 50, 51) in the ALW. The former two words conform to the picture of our map, the third one (< Latin harpix) has Ihl only in an extremely limited number of places in the south of Luxembourg. 3. Van Loey (1958: 36) mentions Late-Middle Dutch forms with palatalization of IdJ from Brabant, Limburg and Holland-Utrecht, but not from the southwest of the language area. 4. Poeder is the old, pre-change northern and eastern form, poeier and poer (respectively with palatalization of /d/ and with syncope of /d/) have probably been borrowed from different (?) southern dialects (cf. Van Loey 1958: 37). 5. One very hard problem to solve for this hypothesis is the shift of the word accent from the second syllable (where it is necessary to explain the full form of the vowel Iii occurring in many dialects), to the first syllable, where it is found now, at least in all southern dialects. In my proposal Iii develops from the intermediate /(l)j/ stage. 6. Further east, in the medieval county of Loon, complex pronouns did not generally develop; Goossens (1994:129) points to the existence of a few stray forms in the southwestern quarter of the area. They may be fairly recent import from the Brabant region. 7. Combinations of diphthongs with following Irl are absent in most regional varieties of Dutch, as they are in the Standard language, for that matter. 8. This information is taken from L. Keymeulen (1991) and (1993). It goes without saying that only a (hopefully representative) portion of the rich material in Keymeulen's dissertation can be presented here.

Parallels in the phonological

systems

121

9. The fact that pronouns are treated in a slightly different way is not surprising: language history is full of such accidents, cf. the development of English, German and Standard Dutch pronouns.

References Carton, F. & M. Lebegue 1989 Atlas linguistique et ethnographique du Picard. Paris, Bonneton. Ginneken, J. van 1934 "Waalsche en Picardische klankparallellen". Onze Taaltuin 2, 289-302. Goossens, J. 1994 Sprachatlas des nördlichen Rheinlands und des südöstlichen Niederlands "Fränkischer-Sprachatlas", Zweite Lieferung, Textband. Marburg, Elwert Verlag. Hoecke, W. van 1988 "Paralleles phonetiques des deux cotes de la frontiere linguistique firancon6erlandaise: la diphthongaison de [E] entrave". Distributions spatiales et temporelles, constellations des manuscrits (ed. by Pieter van Reenen and Karin van Reenen-Stein), Amsterdam/Philadelphia, J. Benjamins, 187-202. Keymeulen, L. 1991 De mouillering in de Westbrabantse dialecten - conditionering, implicaties, verspreiding en historische situering. Unpubl. PhD-thesis, University of Ghent. 1994 "De Zuidwestbrabantse mouillering en de umlaut". Handelingen van de Koninklijke Commissie voor Toponymie en Dialectologie 65, 115-142. Loey,A. van 1958a Schönfelds's Historische grammatica van het Nederlands. Zutphen, Thieme. 1958b "Over pronomina van het type "wijlieden". Album Edgard Blancquaert, Tongeren, Michiels, 319-322. Remacle, L. 1953 Atlas Linguistique de la Walonnie. Tableaugeographique des 1953parlers de la Belgique romane d 'apres l'enquete de Jean Haust et des enquetes complementaires. Tome I: Introduction generale. Aspects Phonetiques. Liege. Ryckeboer, H. 1991 "De spontane palatalisatie. Een Nederlands-Picardische parallel?" Handelingen XLV der Kon. Zuidnederlandse Maatschappij voor Taal- en Letterkunde en Geschiedenis, 113-135; also publ. in: H. Ryckeboer: Het Nederlands in NoordFrankrijk - Sociolingu'istische, dialectologische en contactlingui'stische aspecten, 1997, 147-169. 1997 "Nederlandse lexicale dementen in de Noord-Franse dialecten - een dialectlexicografische en dialectgeografische benadering van taalcontact in NoordFrankrijk". Het Nederlands in Noord-Frankrijk - Sociolingu'istische, dialectologische en contactlingui'stische aspecten, 1997, 170-292. Schutter, G. De 1993 "Umlaut en mouillering als concurrentiele verschijnselen". Handelingen van de Kon. Commissie voor Toponymie en Dialectologie 65, 143-158.

122

Georges de Schutter

Weijnen, Α. 1964 "Fonetische en grammatische parallellen aan weerszijden van de FransNederlandse taalgrens". Tijdschrift voor Nederlandse Taal- en Letterkunde 80, 1-25; also publ. in A. Weijnen: Algemene en Vergelijkende Diabetologie, 1975, 151-172. 1967 "Leenwoorden uit de Latinitas, stratigrafisch beschouwd". Verslagen en Mededelingen Kon. Vlaamse Academie voor Taal- en Letterkunde, 1967, 365480; also publ. in A. Weijnen: Algemene en Vergelijkende Diabetologie, 1975, 189-209.

Britain and the Continent - linguistic interrelations Wolfgang Viereck

My time this morning is limited and the problems are many. Consequently I cannot do more than discuss a few of them. The choice of the problems is, of course, subjective. I will start with the prehistory and finish with the present century. When it was noted that many of the features characteristic of present-day African American English were also attested in parts of Britain, the question arose as to how this can be accounted for. Perhaps we have to look back to an Afro-Phoenician pidgin brought to Britain by tin smelters from Africa, and transmitted through Cornish to the dialects of Southern English? After all, the Phoenicians were most interested in Cornish tin, and they also had trading-posts in Africa. Perhaps I should advance the theory without so much tongue in cheek - someone might take it up as a research topic. At about the same time, however, as the Phoenicians set foot on what was much later to be called "English" soil, Celtic-speaking peoples, in several waves, pushed back the inhabitants to more remote areas. One language gradually replaced another. When this process was completed is not certain. Adomnan wrote the Vita Columbae, the Life of Columba, towards the close of the 7th century AD "He relates how Columba, after having come over from the north of Ireland to Britain, went on a mission to Brude, King of the Picts, whose stronghold lay near the mouth of the river Ness. To the king and his court he had no difficulty in making himself understood, but on journeying inland one reads of his preaching to the peasants with the aid of an interpreter. At a later date, on the island of Skye, two young men brought their father to him to be baptized. Here again, he was forced to enlist the services of an interpreter although the convert bore the Celtic name of Artbrannan" (Mourant and Watkin 1952: 30). The conclusion one can draw is that the language was pre-Celtic, whether Indo-European or pre-IndoEuropean is being debated, as, e.g., Wolfgang P. Schmid (1994) and Theo Vennemann (1993) show. Old European hydronomy provides the evidence. Recent research has identified a large number of Old European river names

124

Wolfgang Viereck

in Britain with equivalents or close relatives in Western, Central and Southern Europe. Their durability is probably due to the vital importance of rivers in the life of prehistoric settlers. Examples of identical or related names on both sides of the Channel are: Ale (in the Borders Region of Scotland), Alne (in Northumberland), and Alauna (in Gaul); Ayr (in the Scottish Strathclyde Region), Oare (in Somerset), and Ahr (in Germany), Aar, Are (in Flanders), Aar (in the Netherlands), and Ara (in Spain); Armit (in the Scottish Borders Region) and Ermetz (in Germany), Armence (in Switzerland), Armance (in France); Shiel (in the Highland Region of Scotland) and Seille (in France), Salia (in Spain) and Farrar (again in the Scottish Highland Region) and Vara (in Liguria), Vaire, Veyre (in France). Haematological and archaeological research are of great interest here also. There are three very remote moorlands in Wales where the skeletal features of the population probably indicate a Stone Age foundation. These regions are the Black Mountain area of southeastern Dyfed, the Plynlymon range in Central Wales and the mountain mass of Clwyd termed Mynydd Hiraethog. While, generally, the incidence of blood group Β is higher in the Celtic lands than elsewhere in Western Europe, it is still higher, significantly higher, in these mountain areas. Group Β may be derived in part from the eastern, Slavonic-speaking peoples, the far-western rise in group Β is, however, clearly a different and probably a much older phenomenon. According to Mourant and Watkin, a consideration of some of the details of the distribution of blood group Β in Wales suggests that it may have been introduced at a very early period indeed in the human settlement of the country (1952: 22). The existence of high group Β frequencies among people dwelling on the inhospitable moorlands of Wales - the areas where one would expect the earlier arrivals to have been driven by the later invaders - lead, again according to Mourant and Watkin (1952:23), to the suggestion that a high group Β wave entered the country not merely before the moderately high Germanic group A current, but probably even before the very high Celtic group 0 stream. Are traces of a very ancient stock still detectable further south, such as in Cornwall, where there is also evidence of settlers of various prehistoric cultures and where there is also a Celtic background? Unfortunately, the distribution of blood group data is generally available only for larger areas in England, so that it cannot be tested yet whether there are significantly higher group Β patches in such remote parts as Bodmin Moor in Cornwall, for example, or Dartmoor in Devon where also pre-Bronze Age peoples lived. I shall return to blood-group evidence later.

Britain and the Continent

125

In a number of articles, Mario Alinei contributed to the cultural history of Europe, basing his research mainly on the data elicited for the Atlas Linguarum Europae and interpreting them within the reference framework defined by historians of religions and those of cultural morphology. If this framework of reference is accepted, three layers can be distinguished - one historical layer connected with Christianity and Islam and two pre-historical layers, one characterized by "supernatural", "superhuman" pagan figures and, leaving anthropomorphism, the other by still earlier zoomorphic and kinship representations. The basic structure has remained the same from prehistorical to historical times (see Alinei 1997). Analysing transparent motivations in classes of realia, such as plants, diseases, animals, and natural phenomena including planets, I have shown that Britain participates fully in this cultural set-up. As the Atlas Linguarum Europae only reveals traces of this for Britain, a consultation of additional sources proved necessary. A few examples must suffice here (for a more comprehensive account cf. Viereck 2001).

As Christian motivations belong to the most recent level, they also occur most often in the data. Among animals, designations of the smallest and weakest pig of a litter can be mentioned in this category. In England and Wales Anthony(-pig) was elicited, sometimes as Tanthony, a wrong separation of Saint Anthony. He was the patron saint of swineherds to whom the smallest pig of each litter was usually dedicated. Saint Anthony is frequently represented with a little pig at his side. With regard to diseases, the following Christian motivations were among those I found in the British Isles: St. Anthony's fire or holy fire 'erysipelas', St. Vitus '(s) dance 'Chorea minor and maior' and God's marks or God's tokens for marks of the plague. (For names of diseases see W. Viereck and K. Viereck 1999). For plants the magico-religious motivations are quite numerous. The daffodil (Narcissus Pseudo-Narcissus) is Saint-Peter's bell in Wales, and Saint Peter's herb is an expression for the cowslip (Primula veris) in parts of England. Very many plant-names in the British Isles occur with the motivation 'devil' as Wright's English Dialect Dictionary, i.v. 'devil' π. 2 reveals. The great reedmace, Typhia latifolia, is attested with 'holy', as is the Glastonbury thorn, Crataegus Oxyacantha, whereas the dandelion, Leontodon Taraxacum, and the rose-root, Rhodiola rosea, show a motivation with 'priest' in Britain and the woolly-headed thistle, Carduus eriophorus, one with 'friar' there.

126

Wolfgang Viereck

Natural phenomena, including planets, also testify to a Christianization and Islamization in Europe. The classic example in the Atlas Linguarum Europae is the rainbow - and not only for the most recent level, but for the whole geolexical stratigraphy. Everywhere in Europe we find compounds with, for example, 'belt', 'bow', 'bridge', 'ribbon', 'ring', plus a religious motivation such as 'God's belt', 'Noah's bow', 'St. Bamaby's crown', or 'Allah's bow'. Also the moon once had a religious veneration, still discernible in Hungarian istenkaläcsa ('God's cake'). In the most archaic layer that can be distinguished, i.e., the zoomorphic and totemic layer characteristic of egalitarian societies, the realia investigated appear in the form of either an animal or a kinship name. Starting with supernatural, magico-religious beings, an appropriate example would be the last corn sheaf cut at harvesting into which the vegetation demon, it was believed, retreated. In Ireland we find granny ('grandmother') as a designation for the last corn sheaf. Also animal names are attested for the last corn sheaf: girria ('hare'), hare's bite / sheaf / seat / tail, cow, hog, piardog ('crayfish'), rabbit, and swallow occur in Ireland. Coming to animals, Riegler (1937/2000) had already interpreted wild animals and insects as relics of a totemistic view of the universe in which they would be our closest relatives. This relationship, similar to kinship, is consequently expressed by kinship terms. Propp (1946/1987) noted that the totem animal in its original form is embodied by the 'mother' and by matrilinear kins. As a kinship name for the weasel I found mother in England. For England Wright's English Dialect Dictionary lists many plant names with, for example, the motivation 'pig', 'fox', 'goat', 'toad', 'cat', and 'horse', where, however, the magico-religious belief is not always clear. The middle layer is both pre-Christian/pre-Islamic and post-zoomorphic. This layer is characterized by anthropomorphic representations. Animals provide quite a number of magico-religious names. For the weasel there is 'fairy' in English. The motivation also for the smallest pig of the litter in Ireland is 'fairy' (siog and siabhra). The same motivation is attested in England for the glow-worm, while 'witch' is noted there for the swallow. As to plants, the motivation 'fairy' occurs in England for the Primula veris, 'witch' in English dialects for Pyrus Aucuparia, Leontodon Taraxacum, and Digitalis purpurea. For the supernatural powers such as the corn spirit we also encounter anthropomorphic motivations such as, in Ireland, old maid, (old) hag, and cailleach ('old hag', also meaning 'witch'). In Scotland carlin ('old woman') is attested for the same notion.

Britain and the Continent

127

Among natural phenomena and planets, the rainbow has anthropomorphic representations everywhere in Europe. The moon may be associated in English with the gods Diana, Luna, Phoebe, Cynthia, Selene and with a silver-footed queen, and the sun with the gods Helios, Apollo, Phoebus, and Hyperion (Roget's Pocket Thesaurus 1956, s.v. 'moon' and 'sun'). In the process of the cultural development of Europe we thus find recurrent structural patterns: the same class of realia was first given kinship and zoomorphic names, to be followed by anthropomorphic names and finally by Christian and Islamic names. While dating the last-mentioned layer is unproblematic, Alinei assumes "that the anthropomorphic representations of reality are connected with socially stratified societies, typical of the Metal Age, while zoomorphic and kinship representations are connected with more primitive societies of the Stone Age" (1997: 27). In Britain the Celts were followed by the Romans, then the Angles, Saxons, and Jutes arrived there, to be followed, in turn, by the Vikings. A little later the Norman Conquest took place. Many peoples and their languages came in contact with one another. Even if we restrict our attention to the period before the Norman Conquest, philologists faced and, of course, still face enormous problems. Let me mention only a few. Kenneth Jackson notes in his 1953 book Language and history in Early Britain about the relationships of Britons and Anglo-Saxons that "The whole picture is, at any rate, totally incompatible with the old theory of the complete extermination of the British inhabitants" (p. 246). How can we relate this to the handful of genuine Celtic loans in Old English? Myres (1986) appears to be certain that Anglo-Saxon settlement occurred on a considerable scale in England as early as in the 4th century, about one hundred years prior to the time given in Bede's Historia ecclesiastica gentis Anglorum. Could this affect our views of the history of individual Latin loanwords in Old English? Sawyer (1958) assumes that there was no Scandinavian colonization on a massive scale in England, and that consequently there was only a comparatively small number of Norse speakers. How does this agree with what we know about the impact of Norse on Old and early Middle English? With regard to the first-mentioned problem - the Celtic - Old English contact - there was a recent dissenting voice. Meid (1990) suggests that originally there were far more Celtic loanwords in Anglo-Saxon, especially in northern Britain, but they were given up later under some kind of pressure. Although,

128

Wolfgang Viereck

according to Gneuss (Gneuss and Gillies 1994: 164), this suggestion does not seem to be very convincing, the interaction - also attested in archaeological finds (cf. Capelle 1990: 12) - between Britons and Anglo-Saxons may have been linguistically more fruitful than we have so far assumed. Gillies (loc. cit.) takes as a parallel the case of the Gaelic element in Scots. Here, too, one encounters a persistent view that the Gaelic presence is minimal. Yet McClure (1986) was able to come up with hundreds of Gaelic loans. Many of these loans were very localized or specialized (rural, domestic, trade, etc.), and others were stigmatized. In any case, Andrew Breeze has begun to reexamine the materials and has produced a number of plausible Celtic etymologies for Old and Middle English words. That also the onomastic materials need re-investigation was recently pointed out by John Insley in his review of Laing - Williamson (1994). There are a number of Old English personal names, such as Ceadda and Tüda, that were borrowed from British, were completely integrated into Old English and remained productive as is proved by suffixed hypocoristic formations such as Tuddel and Ceadel(a). The investigation of British onomastic material in Old English has lately concentrated too much on the significance of place names with regard to settlement history. The time has come to see the importance of this material from a linguistic perspective (Insley 1997: 87). Let me mention a final example from this early Celtic - Old English period. As presented in the literature, it also points to a greater interaction between Celts and Anglo-Saxons than can be assumed from the few generally attested Celtic loans. The example concerns the contact theory of iumlaut. This has been with us since 1918 and only recently re-emerged in a 1994 book by Herbert Penzl where he wrote: "Der MJmlaut findet sich in allen germanischen Sprachen mit Ausnahme von Wulfilas Bibelgotisch, ist aber sonst nirgends systematisch durch neue Schriftzeichen bezeichnet, nur im Altenglischen. Ich schreibe das dem Einfluß von romanisierten Kelten zu, die im Kontakt mit den germanischen Siedlern ihre Sprache und sogar ihre Runenschrift lernten. Ihnen mußten aber schon bloße Umlautsallophone wie distinktive Laute, nicht etwa wie Muttersprachlern nur als automatische Varianten, erscheinen. Sie hatten wohl ähnliche gerundete Palatallaute in ihrer eigenen Sprache" (p. 28). As Marcin Krygier remarks "This theory is open to criticism on a number of counts. First of all, it presupposes a dominant Celtic influence on the origin of written Old English; this has been frequently postulated but never conclusively proven, and in all probability the exact circumstances of the introduction of literacy among Anglo-Saxons will never be known. Moreover, it

Britain and the Continent

129

requires an enormous leap of faith in accepting the convoluted path of umlaut grapheme transmission as postulated by Penzl: first Celts interpret AngloSaxon allophones as phonemes and write them accordingly, then AngloSaxons themselves adopt the new spelling, even though it still represents vowels which are only allophonic in their own language. The alternative, i.e., the assumption that earliest Old English up to the point of umlaut phonemicisation was written down exclusively by scribes of foreign (Celtic or otherwise) origin, can be dismissed outright as fairly improbable" (1997:43). Krygier continues: "The next point on which Penzl's theory exhibits a fatal flaw is his interpretation of the state of literacy among early Germanic settlers in England. He believes the art of runic writing to be so popular among Anglo-Saxons that it could be learned by the native Celtic population. Moreover, he presumes Celtic influence was responsible for introducing new runes to represent umlaut vowels. This presupposes a sizeable Celtic contingent of runic writers, otherwise their contribution ... would have passed unnoticed. At this stage it becomes patently obvious that Penzl's reasoning is self-contradictory. Furthermore, the assumption of a high level of runic literacy among the Anglo-Saxons seems a little bit farfetched. Even if one accepts Derolez's figures (Derolez 1990), on which Penzl bases his theory, the evidence of numerous blatant spelling errors (which may suggest mechanic copying of the text without understanding its nature) and unexpected parasitic vowels (which may indicate a painful and slow process of rune writing by a barely literate artisan) demand caution and definitely do not permit drawing such sweeping conclusions" (Joe. cit.). With regard to Latin - Old English contacts, a three-period model had been the basis for nearly all later treatments of the subject for at least one century since Pogatscher postulated it in 1888. He differentiated between continental loanwords, borrowed until ca. 450 AD, early insular loanwords taken over from the spoken, i.e. Vulgar, Latin of Romanized Celts in Britain between ca. 450 and 600 AD and later loanwords, after ca. 600. Pogatscher'S opinion of the survival of spoken Latin in Celtic Britain, weakened but maintained against counter-arguments put forward a little later by Loth (1892), "is among the most controversial issues" (Gneuss 1994: 152). Nevertheless, the model was buried only in 1990 when Wollmann demonstrated that the assumption of a linguistic borderline about the middle of the 5th century is not tenable. This is in line with archaeological research that dates the "adventus Saxonum" much earlier than Bede did (cf. Capelle 1990: 1 Iff.).

130

Wolfgang Viereck

With regard to the Scandinavians, blood-group evidence is again of value. I refer to the area that is called "Little England beyond Wales". This is the name given to the southern part of what used to be called Pembrokeshire, now the south-western part of Dyfed, where the inhabitants have for centuries been non-Welsh-speaking. "The earliest evidence of human habitation in Little England dates back to palaeolithic times. Traces of mesolithic people have also been found whilst the distribution of megalithic monuments shows that the region was familiar to neolithic people. Also Bronze Age invaders have left their mark in the area ..." (Watkin 1960: 148). "That the Scandinavians were responsible for the names of islands such as Gateholm, Grassholm, Skokholm, Caldey and Skomer, of rocks by the name of Emsger, Goscar and Tusker and of a village called Hasguard is extremely probable but whether these names imply settlement is a crucial question. Lloyd [1939] goes as far as to state that 'the salient fact remains that nowhere on Welsh soil was there any permanent Scandinavian settlement'. Collingwood [and Myres 1937], however, express the view that 'the story adds to the evidence of a definite settlement'.... Charles [1934] believes that it was almost inevitable that Norse traders should seek new homes along the coast of South Wales in places as inviting as Milford Haven" (Watkin 1960: 148). More opposing views could be cited. The Royal Commission on the Ancient and Historical Monuments has concluded that the old hundreds of Castlemartin, Rhos and part of Narberth were included in the Norse areas of settlement. Of course, Normans, too, came to the area in the last decade of the 11th century and also Flemings were transferred there from England in the early 12th century. Blood-group evidence revealed - according to Watkin's (1960) investigation - that Little England is inhabited by two rather different groups of people. The north-western part shows close resemblences to Normans from the Seine valley. The existence of a group Β frequency "in harmony with that of Western Wales suggests that traces of an ancient stock, probably akin to the Welsh moorland folk, have not been completely effaced by the later waves of immigrants" (Watkin 1960: 152). For south-east Little England Watkin (1960) notes an exceptionally high group A frequency. There is a corresponding fall in the Celtic group 0. Only in parts of Sweden and Norway have identical group A frequencies been recorded. Thus one can conclude that the Vikings did establish a permanent settlement in south-eastern Little England. Furthermore the high blood-group A frequencies "so akin to those of parts of Scandinavia suggest that the Vikings came to these parts not

Britain and the Continent

131

as warriors who married into native families but as settlers accompanied by their wives. Had the Scandinavians married Welshwomen to any appreciable degree, the resulting population would have been much lower in A. Beddoe [1885] mentions that in some expeditions the Vikings brought not only their wives and children with them but their thralls also" (Watkin 1960: 151). What is quite important in this connection is that haematological evidence correlates closely with linguistic evidence. According to the Survey of Anglo-Welsh Dialects (Parry 1977, 1979) "Little England beyond Wales" is characterized, for instance, by a disproportionately high number of English (versus Welsh) lexemes for the same notions compared with "the rest" of Wales. I should like to mention in passing that in England, too, there are interesting correlations between haematological and linguistic evidence (cf. Viereck 1998). What the contact between Scandinavian and English was like is still a matter of dispute. For a discussion of controversial points such as mutual intelligibility, forms of bilingualism or creolization, I would like to refer to a contribution by Kastovsky (1992). Each of these aspects could easily be the subject of a separate paper. Moreover, I would like to mention Karl Inge Sandred's (1991) survey of Swedish contributions to the study of Scandinavian in England and his own important work in Norfolk as well as Gunnel Melchers' equally important studies on the Norn - Scots contact situation on the Shetlands (e.g., Melchers 1991). Time allows me to mention just a few aspects. Patricia Poussa treated the non-standard relativizers what, at, and as in 1986, 1988, and 1991. In 1986 as was for her "most probably a blend form, made up of Scandinavian and English elements" (p. 115). Two years later she noted: "it seems likely that the English dialectal relativizer AS may be connected with the early Old Norse relative particle es" (p. 459) and in 1991 she wrote: "I have previously ... considered the possibility of influence from the early Old Norse es relative particle, which is possible as minor input... but cannot alone explain the phenomenal spread of this construction" (p. 307f., n. 9). Since "As in Wales is clearly confined to the immediate border region" (p. 300), Poussa now assumes it to be a Brittonic - Old English contact form. All this reminds me in a way of some people's localization of the Indo-Europeans' homeland: Archibald H. Sayce, in 1880, was certain that "This Aryan family of speech was of Asiatic origin", ten years later he was equally sure that it was of European origin, and in 1927 he noted: "So far as my examination of the facts has gone it has led me to the conviction that it

132

Wolfgang Viereck

was in Asia Minor that the Indo-European languages developed" (quoted according to Mallory 1989: 143). As we all know, the search continues. The Linguistic Atlas of Late Mediaeval English that was published in 4 volumes in 1986 represents an enormous achievement in Middle English dialectology. With regard to Scandinavian influence, the atlas provides some interesting maps. Two of them (er(e) 'are' and at as an infinitive marker) are used by Samuels, one of the main editors of the atlas, to support his claim of a Scandinavian focal area in northern England, in what he calls the Scandinavian belt. But here some weaknesses of the atlas also become evident. In 1989 Samuels noted: "Another very striking [case] is the reduction of the definitive [sic] article to t' or th' (e.g. t' man, th' mon), or its complete deletion" (p. 114) in the area of the Scandinavian belt. "Why, we may ask, does this happen only to the definite article and not to other proclitics?" In parentheses it must be said that the Oxford English Dictionary notes the reduction also in other proclitics: t' as a shortened form of to as in tabandon 'to abandon', etc., and 7 as a shortened form of it as in, for instance, 'tis 'it is', 'twere 'it were', and do't 'do it', etc. Samuels answers his own question by saying: "Surely, this must be because the substratum had (except with adjectives) nothing in that position: the Old Norse definite article was a postpositive enclitic ... which in English would come to be treated like any deletable inflectional syllable" (loc. cit.). Samuels does not explain why the absence of the definite article before a noun in Old Norse should have led to t' and th' forms in English. As I tried to show (Viereck 1995), the two morphs of the definite article are rather due to internal developments of English. With regard to the Linguistic Atlas of Late Mediaeval English, it is unfortunate that it documents only the various forms of the definite article but not its absence, the zero morph. In the present context this would have been especially important. Problematic are some deceptively simple equations in textbooks of the type "ditch of native origin" and "dike of Scandinavian origin" or cold with non-palatalized DsJ being due to Scandinavian influence. Here the Anglian antecedents are simply ignored. In that dialect, Dd was retained whereas it became /tJ7 in West Saxon. Thus two sources are likely with cases such as cold and dike supporting each other. That semantic aspects must not be ignored was shown by Ramisch (1997) with reference to the pair ditch!dike. The earliest attestations of dike in the Oxford English Dictionary in the sense of 'sea-wall or embankment against flooding' date back to the 16th century, which was the period when dikes were built in the eastern part of England for reclaiming and draining land. An adoption from Middle Dutch is thus plausi-

Britain and the Continent

133

ble as it is known that Dutch engineers and other specialists were employed in constructing such embankments. Another indication that there has been an influence from Dutch can be seen in the adoption of other expressions in the same word-field. Thus the term dike-grave is also attested in English for the same period. Dike-grave is clearly based on Middle Dutch dijcgrave (see modern Dutch dijkgraaf and modern German Deichgraf). The aforementioned Linguistic Atlas of Late Mediaeval English is regrettably silent as regards penetrations from languages other than Scandinavian, such as Dutch and Low German. In the area of Middle English word geography, much remains to be done. Rolf Kaiser's dissertation Zur Geographie des mittelenglischen Wortschatzes, published in 1937, is still the fundamental work! J.F. Bense's Dictionary of the Low-Dutch Element in the English Vocabulary of 1939 - Bense had been publishing in this area already since the mid1920s - is, although incomplete, the most comprehensive inventory of Dutch, Flemish, and Low German words in English to date. It relies entirely on the Middle Low German dictionary of Schiller and Lübben published between 1875 and 1881. However, the recent dictionary by Lasch, Borchling, Cordes, and Möhn, whose first volume appeared in 1956, shows all too clearly that Schiller and Lübben's work cannot be regarded as an authority any more. Since Bense's time, considerable lexicographic progress has also been made with regard to Middle English and Scots. Also primary sources, such as diaries, including those of Englishmen traveling and/or trading on the Continent, need more intensive consideration as has been the case up to now. A further weakness of Bense's dictionary is that it almost completely lacks a linguistic discussion. The reader has to turn to individual entries to glean information about such aspects as phonetics, phonology, and changes of meaning. Moreover, both the Survey of English Dialects and the Linguistic Atlas of Scotland date from post-World War π days. These surveys are important in that they allow a more intensive study of the impact of Dutch and Low German on English dialects, not just on the standard language. One example must suffice here, namely the distribution of lesk 'groin' in England. Wright's English Dialect Dictionary traces this word back to Low German leeske 'groin' and attests its use in Nottinghamshire and Lincolnshire where it was still used in the 1960s. In other, partly adjoining areas the Survey of English Dialects noted lisk. As no standardization occurred with these dialectal terms, they continued to exist side-by-side and mirror their different, but related origins till the 20th century. Lisk goes back to Old

134

Wolfgang Viereck

Danish and Old Swedish liuske 'groin'. In their Word Geography of England, Orton and Wright lump lisk and lesk together under lisk. While the historical source of English lesk is clear, the genetic source is not. It is likewise with early Dutch - English parallels where the Dutch words could also be of Scandinavian origin. Antonius Weijnen repeatedly examined "Oude Engels - Nederlandse parallelled' (1966,1975) stressing especially the dialectal level on the Dutch side. We need something similar on the English side, a more interdialectal orientation (cf. Viereck 1993 and Berns 1997). Let me return to the geographic area from where I started, namely the south/ southwest of England and comment on the cross-Channel contacts to Brittany or Little/Lower Britain as the Bretagne is also called. Today's Breton is the continuation of the language of the immigrants who left southern/southwestern Britain between the 5th and the 7th century due to the pressure exercized by the Anglo-Saxon invasion. This descent of Breton had been generally accepted until about 1960 when Franpois Falc'hun advocated a different theory. In its extreme form, it says that Breton is the continuation of Gaulish spoken in the Roman province of Aremorica. Thus Gaulish had not died out completely, but had survived in this province. As usual, the truth lies probably here too between the two extremes. Immigration from southern Britain cannot be denied, nor can it be denied that Breton belongs to the Brittonic branch of Insular Celtic. Still, it cannot be excluded that Gaulish survivals have been retained in this isolated province and have influenced the language of the immigrants (cf. Ternes 1993: 106f.). Recently Hildegard Tristram has re-opened our eyes in a number of insightful contributions to linguistic cross-Channel parallels, ranging from about one thousand years ago with the voicing of initial fricatives (1995a), a field as much debated with regard to England as it is with regard to Brittany, over the phonetic realizations of /r/ (1995b) to about 150 years ago with the jargon of the so-called "Johnnies de Roskoff" selling their onions both at home and everywhere in Britain, a jargon called bretanglais (1993). This last-mentioned aspect set aside, the question is how the older cross-Channel linguistic parallels can be explained. Are they a question of polygenesis or are they due to a Celtic substratum in both languages? Tristram favors a third alternative, namely a linguistic federation between neighboring languages bringing about pattern changes through sustained cultural contact (1995a: 293). Waterways, here the English Channel, are not seen as linguistic barriers, but rather as focal centers of innovations shared by the languages on the

Britain and the Continent

135

opposite coasts. Of course, this model is not new, and it has been applied to other parts of Europe and beyond; see, for example, Heinrich Wagner, "Near Eastern and African connections with the Celtic world" (1982) and, earlier, in relation to haematology, Mourant and Watkin (1952). Tristram's suggestion in searching for a West-European Sprachbund, i.e., for features of convergence between Celtic, both earlier Continental and later Insular, Germanic, and Romance languages and dialects is very tempting (see also her papers 1995c and 1997). This approach differs from the traditional one of investigating Britain's languages and dialects in that it is interested in innovative features common to the whole area investigated. Seen in this light the contributions in the conference volume Language contact in the British Isles, edited by Ureland and Broderick and published in 1991, follow the traditional path. Among their list of 34 insular linguistic contact possibilities Breton surprisingly does not show up at all. I would like to conclude with some remarks on English - German linguistic interrelations. The earliest English influence on German began in the Old English period and is linked with the Anglo-Saxon Mission in such places as Fulda and Mainz. Old High German gotspell (< Old English godspell) and der heilago geist (< se hälga gäst) belong here. The question arises whether as a consequence of the clerical connections between England and the Continent, especially in the course of the Benedictine reform, there had also been a lexical impact in the opposite direction, namely from Old High German to Old English. Betz (1949) assumed the latter to have been the case and produced a total of 13 words to prove this. However, Gneuss (1955) questions this list, rejects such an influence in a number of cases, and favors an independent coinage in others. Such problems are still with those concerned with the present-day linguistic impact - even in the face of much improved possibilities of documentation. Is German Arbeitsessen really a translation of English working lunch or working dinner or is German Gelbe Seiten really a translation of English yellow pages? Even after intensive research such cases cannot always be clarified unequivocally. Especially with closely related languages such as English and German, such formations can also have been coined independently; it need not always be interference. Another problem: Teenie ('a young teenager') is usually said to have been introduced into German in the 1970s. Then, accidentally, a first attestation of Teenie is found in German, dating it back to the mid-19th century! What are we to make of such discoveries? Is it a hapax legomenon? Shall we ignore it? Or: A rendering of an English

136

Wolfgang Viereck

expression might change in German, as was the case with English doves and hawks. Until 1966 - at least according to the documentation available - this was Tauben und Habichte', ever since it has been rendered as Tauben und Falken. We will have to consult with ornithologists to find out why. Or: German untertreiben!Untertreibung is usually traced back to English unterstate/understatement with attestations in the 1950s. In a prize competition of the "Gesellschaft für deutsche Sprache" much earlier attestations of German untertreiben!Untertreibung have been unearthed, dating the words back to 1910. Thus, at best, their semantic extension can be attributed to English influence. A final example: Gleitzeit had originally been assumed to be a loan translation from English. In reality it is exactly the other way round. The concept was introduced in Germany in 1967 by that name whereas it aroused quite a number of originally competing expressions in English such as sliding time, flexible time, flexitime, flextime and gliding time. In view of the problems we encounter in contact linguistics when dealing with the present or the recent past, is it any wonder that they are multiplied when periods are investigated where documentation is scarce?

References Alinei, Mario 1997 "Magico-religious motivations in European dialects: A contribution to archaeolinguistics". Dialectologia et Geolinguistica 5: 3-30. Beddoe, John 1885 [1971] The races of Britain. A contribution to the anthropology of Western Europe. London: Trübner. Bense, Johan F. 1939 A dictionary of the Low-Dutch element in the English vocabulary. The Hague: Martinus Nijhoff. Berns, Jan B. 1997 "Niederländisches im englischen Wortatlas, oder die hin- und hergeschobene Etymologie". In: Heinrich Ramisch and Kenneth Wynne (eds.), Language in time and space. Studies in honour of Wolfgang Viereck on the occasion of his 60th birthday. Stuttgart: Steiner, 24-31. Betz, Werner 1949 Deutsch und Lateinisch. Die Lehnbildungen der althochdeutschen Benediktinerregel. Bonn: Bouvier. Capelle, Torsten 1990 Archäologie der Angelsachsen. Darmstadt: Wissenschaftliche Buchgesellschaft. Charles, Berti G. 1934 Old Norse relations with Wales. Cardiff: The University of Wales Press.

Britain and the Continent

137

Collingwood, Robin G. and John N.L. Myres 1937 Roman Britain and the English settlements. Oxford: Clarendon Press. Derolez, Rene 1990 "Runic literacy among the Anglo-Saxons". In: Alfred Bammesberger and Alfred Wollmann (eds.), Britain 400-600: Language and history. Heidelberg: Winter, 397-436. Gneuss, Helmut 1955 Lehnbildungen und Lehnbedeutungen im Altenglischen. Berlin: Schmidt. 1994 "Language contact in early medieval England: Latin and Old English". In: Margaret Laing and Keith Williamson (eds.), Speaking in our tongues. Medieval dialectology and related disciplines. Cambridge: Brewer, 149-157. Gneuss, Helmut and William Gillies 1994 "Panel ra: Languages in contact". In: Margaret Laing and Keith Williamson (eds.), Speaking in our tongues: Medieval dialectology and related disciplines. Cambridge: Brewer, 159-168. Insley, John 1997 Review of Margaret Laing and Keith Williamson (eds.), Speaking in our tongues: Medieval dialectology and related disciplines. Cambridge: Brewer 1994, Zeitschrift fiir Dialektologie und Linguistik 64: 85-87. Jackson, Kenneth Η. 1953 Language and history in Early Britain. Edinburgh: Edinburgh University Press. Kaiser, Rolf 1937 Zur Geographie des mittelenglischen Wortschatzes. Leipzig: Mayer & Müller. Kastovsky, Dieter 1992 "Semantics and vocabulary". In: Richard M. Hogg (ed.), The Cambridge history of the English language. Vol. i: The beginnings to 1066. Cambridge: Cambridge University Press, 290-408. Kiygier, Marcin 1997 From regularity to anomaly. Inflectional \-Umlaut in Middle English. (University of Bamberg Studies in English Linguistics 40.) Frankfurt: Lang. Lasch, Agathe, Conrad Borchling, Gerhard Cordes and Dieter Möhn 1956ff Mittelniederdeutsches Handwörterbuch. Neumünster: Wachholtz. Lloyd, John E. 1939 A history of Wales from the earliest times to the Edwardian conquest. London: Longman, 3rd ed. Loth, Joseph 1892 Les mots latins dans les langues brittaniques. Paris: Bouillon. Mallory, James P. 1989 In search of the Indo-Europeans: Language, archaeology and myth. London: Thames & Hudson. Mather, James and Hans-Henning Speitel 1975, 1977, 1986 The Linguistic Atlas of Scotland. Scots section. 3 vols. London: Croom Helm. Mawson, C.O. Sylvester (ed.) 1956 Roget's Pocket Thesaurus. A treasury of synonyms and antonyms. New York: Cardinal.

138

Wolfgang Viereck

McClure, J. Derrick 1986 "What Scots owes to Gaelic". Scottish Language 5: 85-98. Mcintosh, Angus, Michael L. Samuels, Michael Benskin with the assistance of Margaret Laing and Keith Williamson 1986 A linguistic atlas of Late Mediaeval English, 4 vols. Aberdeen: Aberdeen University Press. Meid, Wolfgang 1990 "Englisch und sein britischer Hintergrund". In: Alfred Bammesberger and Alfred Wollmann (eds.), Britain 400-600: Language and history. Heidelberg: Winter, 97-119. Melchers, Gunnel 1991 "Norn - Scots: A complicated language contact situation in Shetland". In: P. Sture Ureland and George Broderick (eds.), Language contact in the British Isles. Tübingen: Niemeyer, 461-477. Mourant, Arthur Ε. and I. Morgan Watkin 1952 "Blood groups, anthropology and language in Wales and the Western Countries". Heredity 6: 13-36. Myres, John N.L. 1986 The English settlements. Oxford: Oxford University Press. Orton, Harold et al. 1962-71 Survey of English dialects: Basic material. 4 vols., each in 3 parts. Leeds: E.J. Arnold. Orton, Harold and Nathalia Wright 1975 A word geography of England. London: Seminar Press. Parry, David 1977, 1979 The survey of Anglo-Welsh dialects. Vol. 1: The South-East. University College of Swansea. Vol. 2: The South-West. University College of Swansea. Penzl, Herbert 1994 Englisch: Eine Sprachgeschichte nach Texten von 350 bis 1992: Vom NordischWestgermanischen zum Neuenglischen. Bern: Lang. Pogatscher, Alois 1888 Zur Lautlehre der griechischen, lateinischen und romanischen Lehnworte im Altenglischen. Straßburg: Trübner. Poussa, Patricia 1986 "Historical implications of the distribution of the zero-pronoun relative clause in Modern English dialects: Looking backwards towards OE from Map S5 of The Linguistic Atlas of England". In: Sven Jacobson (ed.), Papers from the Third Scandinavian Symposium on Syntactic Variation. Stockholm: Almqvist & Wiksell, 99-117. 1988 "The relative WHAT: Two kinds of evidence". In: Jacek Fisiak (ed.), Historical dialectology: Regional and social. Berlin - New York: Mouton de Gruyter, 443474. 1991 "Origins of the non-standard relativizers WHAT and AS in English". In: P. Sture Ureland and George Broderick (eds.), Language contact in the British Isles. Tübingen: Niemeyer, 295-315.

Britain and the Continent

139

Propp, Vladimir Ja. 1946 [1986] Istoriceskie kortti volsebnoj skazki. Leningrad: Izdat. Leningradskogo Godsud. Univ. German translation: Die historischen Wurzeln des Zaubermärchens. München: Hanser 1987. Ramisch, Heinrich 1997 "Re-examining the influence of Scandinavian on English: The case of ditch/ dike". In: Raymond Hickey and Stanislaw Puppel (eds.), Language history and linguistic modelling. A festschriftfor Jacek Fisiak on his 60th Birthday. Berlin New York: Mouton de Gruyter, 561-569. Riegler, Richard 1937 [2000] "Tiergestalt" and "Tiernamen". In: Hanns Bächtold-Stäubli and Eduard Hoffinann-Krayer (eds.), Handwörterbuch des deutschen Aberglaubens. Vol. 8. Berlin - New York: Mouton de Gruyter, 819-842 and 863-901. Samuels, Michael L. (1985) 1989 "The Great Scandinavian Belt". Repr. in: Margaret Laing (ed.), Middle English dialectology: Essays on some principles and problems. Aberdeen: Aberdeen University Press, 106-115. Sandred, Karl Inge 1991 "The study of Scandinavian in England: A survey of Swedish contributions including ongoing research in East Anglia". In: P. Sture Ureland and George Broderick (eds.), Language contact in the British Isles. Tübingen: Niemeyer, 317-336. Sawyer, Peter H. 1958 "The density of the Danish settlement in England". University of Birmingham Historical Journal 6: 1-17. Schiller, Karl and August Lübben 1875-81 Mittelniederdeutsches Wörterbuch, 6 vols. Bremen: Kühtmann. Schmid, Wolfgang P. 1994 "Alteuropäisch und Indogermanisch". In: Joachim Becker, Jürgen Udolph and Dieter Weber (eds.), Linguisticae scientiae collectanea. Ausgewählte Schriften von Wolfgang P. Schmid anläßlich seines 65. Geburtstages. Berlin - New York: Mouton de Gruyter, 118-133. Simpson, John A. and Eva S.C. Weiner (comp.) 1989 The Oxford English Dictionary. Oxford: Clarendon Press, 2nd ed. Temes, Elmar 1993 "Die bretonische Sprache: Geschichte, Konflikte und soziale Stellung". In: Bernhard Lauer and Bärbel Plötner (eds.), Breizh Bretagne: Zwischen keltischem Erbe und französischer Gegenwart. Die Bretagne und ihre kulturelle Identität. Kassel: Thiele & Schwarz, 105-116. Tristram, Hildegard L.C. 1993 "Zwiebeln und Wörter: Zum Sprachkontakt über den Ärmelkanal". In: Martin Rockel and Stefan Zimmer (eds.), Akten des ersten Symposiums deutschsprachiger Keltologen. Tübingen: Niemeyer, 331-352. 1995a "Zaoz and Zomerzet: Linguistic contacts across the English Channel". In: Wolfgang Viereck (ed.), Verhandlungen des Internationalen Dialektologenkongresses /Proceedings of the International Congress ofDialectologists / Communications du Congres International de Diabetologie. Band 4 / Volume 4. Stuttgart: Steiner, 276-298.

140

Wolfgang Viereck

1995b

"Linguistic contacts across the English Channel: The case of the Breton retroflex ". In: Jacek Fisiak (ed.), Linguistic change under contact conditions. Berlin - New York: Mouton de Gruyter, 291-313. 1995c "Aspect in contact". In: Wolfgang Riehle and Hugo Keiper (eds.), Anglistentag 1994 Graz. Tübingen: Niemeyer, 269-294. 1997 "DO-periphrasis in contact?". In: Heinrich Ramisch and Kenneth Wynne (eds.), Language in time and space. Studies in honour of Wolfgang Viereck on the occasion of his 60th birthday. Stuttgart: Steiner, 401-417. Ureland, P. Sture and George Broderick (eds.) 1991 Language contact in the British Isles. Tübingen: Niemeyer. Vennemann, Theo 1993 "Zur Erklärung bayerischer Gewässer- und Siedlungsnamen". Sprachwissenschaft 18: 425-483. Viereck, Wolfgang 1993 "The medieval European Common Market and its impact on Middle English". Neuphilologische Mitteilungen 94: 69-78. 1995 "Realizations of the definite article in dialectal English and how and when they originated". In: Jacek Fisiak (ed.), Medieval dialectology. Berlin - New York: Mouton de Gruyter, 315-326. 1998 "Geolinguistics and haematology: The case of Britain". Links & Letters 5: 167179. 2001 Zur Kulturgeschichte Europas. Pri la kulturhistorio de Eüropo. Acta Sanmarinensia. Vol. v. 3. Viereck, Wolfgang and Karin Viereck 1999 "'Die seltzamen namen all' - Zu einigen Ergebnissen des Forschungsprojekts Atlas Linguarum Europae". In: Eun Kim and Erwin Schädel (eds.), Aktive Gelassenheit. Festschriftfür Heinrich Beck zum 70. Geburtstag. Frankfurt: Lang, 711-723. Wagner, Heinrich 1982 "Near Eastern and African connections with the Celtic world". In: Robert O'Driscoll (ed.), The Celtic consciousness. Portlaoise: The Dolmen Press, 5Ιό?. Watkin, I. Morgan 1960 "A Viking settlement in Little England beyond Wales: ABO blood-group evidence". Man 60: 148-153. Weijnen, Antonius 1966 Nederlandse dialectkunde. Assen: Van Gorcum, 2nd ed. 1975 Algemene en vergelijkende dialectologie. General and comparative dialectology. Een verzameling studies van Antonius Weijnen. Uitgegeven door Antonius Hagen en Joep Kruijsen. Amsterdam: Holland Universiteits Pers. Wollmann, Alfred 1990 Untersuchungen zu den frühen lateinischen Lehnwörtern im Altenglischen: Phonologie und Datierung. München: Fink. Wright, Joseph 1898-1905 The English dialect dictionary, 6 vols. London: Henry Frowde.

Aspects of Armenian dialectology J.J.S. Weitenberg

In 1960 Feydit presented a report on Armenian dialectology to the Premier Congres International de Diabetologie generale. (Feydit 1964). It is amazing to see how the study of Armenian dialectology has gained in scholarly interest since this excellent report was delivered. This is only due in part to a growing accessibility of the work done in Armenia in the middle decades of the twentieth century. Above all it is a witness of the intrinsic interest of the problems of Armenian dialectology which offer challenges both to the linguistic historian and the theoretical linguist. In the following only a short overview of the discussion in some areas of this research field will be presented. The references given do not aim at exhaustiveness, but will also enable those who do not master the Armenian language, an easy access to the materials and to the discussions.

Chronology and geographical distribution The Armenian language is known to us from the fifth century AD onwards in an unbroken oral and literary tradition. The historical development of the Armenian language involves a Classical (500-1000), a Medieval (10001500), and a Modern Period. The modern variant of the Armenian literary language was canonized in the course of the 19th century and consists of two branches: an Eastern one, developed on the basis of the dialects of Tiflis and the Ararat plain, and a Western one, developed in Constantinople, based on the dialects spoken in Western Anatolia. The two literary languages are the focus of a dialect continuum ranging from the Balkan into Iran. The Armenian language currently is spoken by ca. 6 million persons; it is at home in Eastern Anatolia, in the area roughly delimited by the three lakes: Lake Van, Lake Urmia, Lake Sevan. Over the centuries, speakers of Armenian have spread over the Byzantine, later Ottoman Empires, and the Trans-Caucasus

142

J.J.S.

Weitenberg

area. Traders established centers of Armenian economic activity in the Russian Empire, Iran and India. Armenian dialectology is interested in the study of Armenian dialects at large. The current situation of the dialects spoken in the Republic of Armenia, in the Karabagh area and in the adjacent Georgian provinces (Jawakheti) is comparable to the situation in most industrialized countries; dialects are still spoken in rural areas; the influence of the literary language is penetrating everywhere through schools and mass communication. Potentially, therefore these dialects are available for all forms of modern dialect research. By contrast, for the Anatolian dialects research is practically limited to an historical approach of the 19th century situation. This is because of the 1915 extermination of the Armenian population in Anatolia which constituted an amputation of Armenian life, and therefore also an amputation of Armenian dialects. Since then, there is hardly any presence of Armenians, and of Armenian dialect-speakers in Turkey, with the exception of the Hamlin area (east of Trabzon), Istanbul and the area around Antioch (Antakya). This does not mean that all Armenian dialects which were formerly spoken in Anatolia have completely disappeared. Refugees to the East, to what is now the Republic of Armenia, tended and still tend to regroup themselves in separate villages, like Nor ("New") Malatya, Nor Sassun, where the dialect can still be studied. Aged dialect speakers can still be found among the refugees to the West, in particular in France and the United States. Nevertheless, there nowadays seem to be two maps of the same geographical area, two separate realities: the armenologist discusses the dialects of Karin, Sebastia, Evdokia, where the turcologist studies an entirely different set of data from Erzurum, Sivas, Tokat. The nineteenth century linguistic situation was different, however. Armenian was interacting with the surrounding languages: Turkish, Azeri, Kurdish, Zaza, Georgian, and Persian. The study of this interaction is a rapidly developing field thanks to the work of Pisowicz (1989), Bläsing (1992. 1995) and Dankoff (1995) who collected and interpreted lexical and onomastic traces of the Armenian presence in Anatolia. Armenian dialectology, then, might be said to work with a virtual map of Armenian dialects in the 19th century distribution. In that period there lived Armenians in Anatolia, and in Russian Transcaucasia. In Ottoman Anatolia no area ever was uniformly Armenian: in the East, in Van, Erzurum, Mush, Bitlis the Armenians formed a (near-) majority; in the West, in Kayseri, Sivas, Constantinople, Smyrna and many other cities the Armenians formed important minorities; the same is true for the Cilician (Adana, Zeytun) area; also, important concentrations were present at the Black Sea coast (Trabzon,

Aspects of Armenian dialectology

143

Hamshen). In Russian Transcausia, the territory of modem Armenia, including the area of Mountainous Karabagh, possessed a large majority of Armenian speakers; a culturally important minority existed in Tiflis; other minorities were found in modern Azerbaydjan (Shamakhi). An unusual instance of language contact is Gypsy-Armenian, so-called Lomavren, which preserves Rom lexicon with Armenian morphology (Bakker and Cortiade 1991; Boretzky 1995).' Outside of Anatolia and Transcaucasia there existed various Armenian speaking centers, among them the Armenian colony of New Julfa in the city of Isfahan, where in 1605 the Persian shah Abbas II settled deportees from the Ararat plain. Such old or excentric colonies often developed dialects of their own and preserve valuable linguistic features. Dialectologically important communities also existed in Poland, on the Crimea, and in Rumania; their stratification has been studied by Schütz (1980). Some of these (now nearly extinct) dialects had very old roots: it has been proposed that eighth century Armenian settlements in the Balkan left substratum traces in the Bulgarian article (discussion in Pedersen 1905: 11-12= 1982: 7-8). A proper Armenian dialect, the so-called k'aiak'ac'i or "city" dialect, developed in Jerusalem; it has remained undescribed. The current dialectal situation of the most important Armenian centers outside of Asia Minor and the TransCaucasus is unknown. It seems that, with the possible exception of the Armenian community in Lebanon, the larger Armenian population concentrations in the "diaspora" (Egypt, Syria, Rumania, Bulgaria, Greece, France, the USA, Canada, Argentina, Australia) exclusively use Modern Literary Armenian in one of its variants, East or West.

The description and classification of Armenian dialects The systematic study of Armenian dialects developed in the 19th century. Awareness of a dialectal division of the Armenian language, of course, is much older. Already in the 8th century the Armenian scholar Step'annos Siwnec'i explicitly mentions seven regional varieties of Armenian (Acaiyan 1951: 128; text Adontz 1970 [1915]: 187). Awareness of Armenian dialects reached the Western scholarly world in 1711 when in Amsterdam the German scholar Joachim Schroeder printed his Thesaurus Linguae Armenicae. In this work Schroeder, following native Armenian grammatical thought, used a social, rather than a dialect-geographical approach: he pointed out that contemporary Armenian possessed three layers: a literary language, in use

144

J.J.S.

Weitenberg

since the fifth century of our era (to which the bulk of his Grammar is devoted), an educated "common language" (on which the Modern Armenian literary language is built), and a great variety of uncultivated dialects (Parnassian 1985; Weitenberg 1984). To his work, Schroeder appended a conversation guide in the New Julfa variant of "common Armenian". Schroeder had obtained first hand knowledge of this dialect by contacting the Armenian silk-traders and publishers in Amsterdam city, who had come from Iran. The 19th century dialectological studies grew out of linguistic, literary and ethnographic interest. The first full description of any Armenian dialect concerned the Tiflis dialect; it was added casually, as it were, to the editio princeps of the 18th century poet Sayeat' Nova (Axverdean 1852) who used this dialect; it was taken up by the German Orientalist Heinrich Julius Petermann (1867).2 Gradually more dialect materials became available in ethnographic studies, which are connected with the names of important scholars such as Lalayan, and Srvanjteants and with the gradual uncovering of an epic cycle centering around the 10th century hero David of Sasun.3 There followed attempts to come to a more comprehensive overview and classification of Armenian dialects. Geographical considerations were decisive in the classification which was proposed in 1866 by the Mechitarist scholar Arsen Aytanean in the first scientific Critical Grammar of Modern Armenian,4 In this classification, which was to be the most influential in the 19th century, Armenian dialects were divided into four groups: Armenian as spoken in Europe (in particular in the Austro-Hungarian monarchy: Poland, Transsylvania), Armenian as spoken in the Western fringe of Anatolia (Constantinople, Smyrna), the Armenian language of Ottoman Anatolia, Eastern Armenian (as spoken in the Caucasus, Russia, Iran, India, Netherlands Indies). Modern Armenian dialectology starts with Hrac'ya Acaiyan (Adjarian; 1876 Constantinople - 1956 Erevan) who to this day is the most outstanding personality in Armenian linguistic research. He published the first scientific description of Armenian dialects; as a student of Antoine Meillet, he originally wrote this study in French (Adjarian 1909). In a more expanded version it later appeared in Armenian (Acaiyan 1911). As a teacher, traveling around in the Armenian - speaking areas until he settled in Erevan, Acaiyan had a first hand knowledge of many Armenian dialects. In his autobiography he explains that he had developed a uniform descriptive approach enabling him to grasp the essential features of a given dialect in a few days. To him we

Aspects of Armenian dialectology

145

owe lucid descriptions of key dialects (Van, Constantinople, Hamshen, New Julfa etc.), a number of which have died out since (Astrakhan, the Crimea). With Acaiyan the classification of Armenian dialects obtained a firmer linguistic footing. Like Patkanov (1869) before him, Acaiyan used the formation of present (and imperfect) indicative as a classifier. In the Classical Armenian language this form is simple synthetic: sirem "I love", sirei "I loved" (next to an infinitive sirel "to love"). Armenian gradually used this form to express various modal aspects (subjunctive, future) and a new present (and imperfect) indicative was developed in various ways. The development of these characteristic morphological features in the verbal system dates to about the eleventh century; much later than the start of the changes in the consonantal system. Acaiyan distinguished three groups: 1. dialects which form a compound synthetic present (and imperfect) with the element kow, ko in various combinations: ks sirem, sirem kow etc. "I love", ks sirei "I loved"; this type is predominant in the Western part of the Armenian area, from Polish Armenian to the Karin (Erzurum) area. It includes important dialects like Constantinople, Sivas (Sebastia), Mush, Van, Hamshen, Cilicia, Syria. 2. dialects which developed a periphrastic group consisting of a participial form in -owm and the copula: sirowm em "I love", sirowm ei "I loved"; this type is called "Eastern" and occurs in the Armenian dialects within the former Soviet Union (Tiflis, Erevan, Karabagh), Iran (Nor Julfa) etc. In these dialects the type ks sirem often serves as a future or conditional. 3. A smaller group forms a periphrastic group with the help of a participial form in el: sirel em "I love". Members of this group occur around lake Urmia and in the area around Ardwin in North-Eastern Turkey. Acaiyan (1909) distinguished 31 Armenian dialects, seven belonging to the owm- branch, 21 belonging to the ko- group, three to the el- class. His classification is satisfactory in that it combines morphological and geographical aspects in a meaningful way. However, as it is based on only one conspicuous morphological feature, it cannot systematically account for the relationship of dialectal entities at a lower level. From the descriptive point of view his three-fold division soon came under attack. In 1939 fcaribyan (Gharibian) split up Acaiyan's third group and established a fourth group, the -s class: sirelis em "I love"; in his comprehensive Armenian Dialectology fcaribyan (1953: 47-56) assumed the existence of seven main groups by additionally splitting up Acaiyan's ks branch along its allomorphs into a ks,

146

J.J.S. Weitenberg

ka, ha group, and by pointing to the existence of a dialect of the classical type which simply has sirem "I love" (Xotorjowr in the North - East Anatolian £oruh area). Subsequent discussion got hopelessly entangled in what amounts to diachronic considerations on the origin of the various allomorphs. Around the middle of the 20th century a more or less comprehensive overview over the Armenian dialectal material was available. This enabled a classification based on phonetic features. A conspicuous feature of Armenian dialects is the phonemic distribution of the stops (and, to a lesser extent, the fricatives) with respect to the Classical situation. In the traditional perspective the Classical language possesses a three-fold opposition: voiced - voiceless voiceless aspirate b : ρ : p' etc.). Though a few dialects have retained this opposition intact (e.g. Agulis, in the Nakhijevan area), most of the dialects vary widely with respect to the Classical voiced and voiceless stops. Sometimes they possess a two-fold opposition (ρ : ρe.g. Van) or (b : ρ', e.g. Trabzon); others, though retaining a three-fold series, shifted significantly with respect to Classical Armenian and developed so-called voiced aspirate stops (b' : b : ρe.g. Erznka). Starting 1941 Gharibian has developed a classification on the basis of the consonantal system. Independently, Vogt (1958) chose the same classification principle. In Laribyan 1959, seven groups are distinguished in this way. This classification is intimately connected with discussions on the phonetic status of the fifth century Classical Armenian consonant system (which will be briefly presented below), in which both Vogt and Gharibian played an important role. Currently, what has been termed as ' Gharibian's Classification' still is the most viable road towards a general clarification of distribution and genesis of the Armenian dialects. 5 Basing on the dialectal outcomes of Classical Armenian stops in initial, prevocalic position Gharibian's grouping is the following: 6

Aspects of Armenian dialectology

Indo-European Classical

gh g

Group 1 Group 2 Group 3

*

g k

*k k'

g* g* g

g k g

k' k' k'

Group 4 Group 5

k k'

g g

k' k'

Group 6 Group 7

g k

k k

k' k'

Literary: Modem Western Modern Eastern

k'

g k

k' k'

g

147

Sivas, Crimea, Poland Erzurum (Karin), Ararat plain, Tiflis Constantinople, parts of Western Anatolia, Hamshen Cilicia, Antakya, Dersim Malatya, Urfa (Edessa), Diyarbakr (Tigranakert) Agulis, Nakhidjevan Van, Urmia, Karabagh

Classical Armenian shows a sound shift (the "First Armenian Sound Shift'') with respect to the reconstructed Indo-European situation. Modern Literary Western Armenian continues the "civilian language" of Anatolia. Modern Literary Eastern Armenian voiceless stops are glottalic (cf. Job 1977: 100103). This same feature is present in Eastern dialects of group 2 (Ardvin, Tiflis, Erevan) and can be found in other areas, in particular Hamshen, Zeytun (Cilicia), Van (Pisowicz 1976a: 65 - 66; Pisowicz 1997: 217; 229); a systematic descriptive study on the dialectal spread of this feature is lacking. Gharibian distinguished 54 dialects, but in his work the relationship between his seven morphological and the seven phonetic groups has not been worked out systematically. An attempt to incorporate as many data as possible into an overall classification was made by Djahukian's 'multi-feature classification' (Jahowkyan 1972; Djahukian 1986: 19 - 24). Detaching himself from the traditional geographical delineations, Djahukian computed the proximity and distance of 120 individual dialect localities on the basis of 100 isoglosses. The localities cover the complete Armenian speaking area from Poland in the West up to Shamakhi near the Caspian Sea. Djahukian distinguishes two main branches, East and West, eleven dialect-groups and 44 individual dialects; his work is extremely valuable because it systematizes a huge amount of data. The value of Djahukian's classification, of course, is directly dependent on the choice and arrangement of the isoglosses; its explanatory potential has not yet been subjected to a thorough test. 7

148

J.J.S. Weitenberg

The currently most authoritative comprehensive description of Armenian dialects is Jahukyan 1972. A valuable treatment of many of the specific features of Armenian dialect phonology is given by Pisowicz (1976a) and Vaux (1998) who are the first to apply modern phonological theory to Armenian dialect data. These works also contain a detailed bibliography covering all existing descriptions of Armenian dialects. The introductory survey of Greppin - Khachaturian (1986) can only be used with extreme caution because of its unsystematic presentation of data.

The origin and development of the Armenian dialects The study of the genesis of the Armenian dialects has been debated mainly on the basis of the distribution of the various consonant systems. In the debate on how the current systems are to be derived from an original situation, it now is communis opinio that this original situation is the one which we find in Classical Armenian and which still is present in the dialects of group 6 (a.o. Agulis, in Nakhidjevan). It is certain that there existed Armenian dialect isoglosses in the prehistoric period (i.e. before the fifth century AD), but the earliest isoglosses do not reveal genetically different consonant systems. On the other hand, the oppositions of the Armenian consonants are not limited to the traditionally assumed features voiced, voiceless, and voiceless aspirates; they also include glottalic and voiced aspirate. The debate on the development of the Armenian dialects is also a debate on the distribution and origin of all these features in the Armenian language at large. Armenian dialects of the Western type, in particular the Cilician dialects attest a change with respect to the Classical Armenian situation involving the following steps: Class. Arm. voiceless stop —> dialect voiced, and Class. Arm. voiced stop —> dialect voiceless; voiceless aspirates are stable in all dialects. These steps are called the "Second" or "Modern" Armenian Sound Shift. On the basis of external evidence, in particular the study of loanwords, the earliest traces of the Modern Armenian Sound Shift can be dated to the sixth century of our era (Pisowicz 1976a: 92-102). The developments had reached their final distribution about 1200, as is attested in the Cilician Middle Armenian language. Thus, the phonetic splits, leading to the various Armenian dialects, are considerably earlier than the morphological changes in the verbal system. However, the Modern Armenian Sound Shift seems to require an interchange of two features, voiced and voiceless (cf. Classical Tigran resulting

Aspects of Armenian dialectology

149

into Dikran). Such a development is impossible from an relative chronological point of view: at some point the features would have merged. Therefore a third feature must have played a role in this interchange. Two of such features have been adduced in the discussion: voiced aspirate and glottalic. Most attempts have concentrated on the interpretation of the voiced aspirates which are present in what one might call the central dialects, groups 1 (Sivas) and 2 (Erevan). The earliest attempts to solve the problem of the Modern Armenian Sound Shift resulted in a revolutionary proposal: in 1906 Pedersen suggested that the consonantal system of Classical Armenian system knew the opposition voiced aspirate - voiceless - voiceless aspirate (b' ρ ρ *); he eliminated the feature plain voiced from the phonetic system of Classical Armenian. About fifty years later, Benveniste, Vogt, and Gharibian reached similar conclusions. Like Pedersen, Vogt and Benveniste assumed that Classical Armenian did not possess voiced stops, but that the graphemes which always had been considered to represent voiced stops, were in reality voiced aspirates; these voiced aspirates were directly inherited from Indo-European. This means that the traditional formulation of the First Armenian Sound Shift (which among others assumes that PIE voiced aspirates become Armenian plain voiced, PIE *bh —> Class. Arm. b) needed revision. Vogt and Benveniste concluded that the voiced aspirates in Armenian dialects are identical to the Classical Armenian voiced aspirates which in turn are identical to the Indo-European voiced aspirates, and therefore an original feature. The representations in other Armenian dialects are the result of later sound changes involving devoicing and deaspiration. A different position was taken by Gharibian: he, too, assumed that dialects containing a voiced aspirated stop directly continue an Indo-European feature, and in this respect the original situation, but he also accepted the existence of plain voiced stops in Classical Armenian. For Gharibian it followed that such dialects represent an older situation and are closer to IndoEuropean than Classical Armenian itself. On the basis of these viewpoints, a famous discussion was initiated in the 1959-1962 issues of the journal Voprosy Jazykoznanija in which all leading armenologists of the time participated, but which in last resort ended with a non liquet.8 In retrospect, it has become clear that Armenian dialectologists participating in the discussion, like Aghayan and Jahukyan, who argued on the basis of loanwords, have succeeded in showing the secondary, i.e. postClassical, character of the voiced aspirates in the Armenian dialects. Their

150

J.J.S. Weitenberg

arguments (in all dialects the behaviour of Iranian and Greek loanwords is consistent with the traditional assumptions) have been summarized by Pisowicz (1976a: 21-23). The view that Armenian voiced aspirates directly continue an Indo-European voiced aspirate is not accepted any more.9 The reasons for this rejection are twofold: on the one hand, the arguments for assuming that Classical Armenian itself possessed voiced aspirates are not cogent: as will be demonstrated below, there have been proposed models in which the rise of voiced aspirates can be limited to the dialects. On the other hand, and this is the decisive point, our concept of the nature of the Armenian voiced aspirate has been precised. On the basis of the modern interpretation is a study by Khachaturian (1983), although already Allen (1950) had given a correct description of these sounds. By means of oscillographic analysis she showed that Armenian so-called voiced aspirates are rather what we would call now murmured stops. Their most peculiar feature is the "breathy coloring of the following vowel" Khachaturian (1983: 61). They have nothing in common with voiced aspirates as we meet them in the Indie languages, in particular in Hindi. In Hindi a voiced aspirate is perceived as consisting of two articulations: voiced plosive followed by (voiced or voiceless) aspiration. In Armenian dialects "voiced aspirates" are homogeneous. Phonemically they are perceived as variants of voiced phonemes in initial position.10 Therefore, it is improbable that Armenian dialectal voiced aspirates are a continuation of the IndoEuropean voiced aspirate which is rather of the Indie type. This fact eliminates the cogency of projecting the Armenian voiced aspirates directly back into Indo-European. For Pisowicz (1976a: 71-91) the voiced aspirate remains the single key feature which triggers the Armenian dialectal development, which according to him is motivated by two phonetic tendencies, fortition and lenition. In his model all consonantal dialect developments start with a common fortition of Class, d —> *dh, resulting in an intermediate (Common Post-Classical) stage: Class. Common

d *dh

t t

t' t'

On the basis of this common stage, subsequent dialect groups are developed. In 1997 Pisowicz published a modified version of his earlier model in which he incorporated the new insights on the nature of the Armenian voiced aspirates and in which he more explicitly took the glottalic stops into account.

Aspects of A rmenian dialectology

151

Without going into details one may wonder whether the assumption of a common dialectal aspiration of inherited Classical voiced stops (which is also assumed by Vaux 1998: 238-241)11 is able to account for all the features which we meet in Armenian consonantism. In the views discussed so far, the glottalic character of some of the dialectal phonemes is neglected or considered to be of secondary origin and importance. In Pisowicz's modified model, too, the rise of glottalization is secondary, a potential result of devoicing, possibly under Caucasian substratum influence (Pisowicz 1997: 223; 225 226). As Kortlandt (1998) points out, the presence of glottalics in the Southern area (Van, Zeytun) speaks against the latter assumption. Rather, one should treat the presence of glottalic stops as an inherited, Indo-European, feature and account for them in developmental models as being present from the beginning. The model which Kortlandt (1978) proposed meets this requirement. It differs from the models discussed above in two ways: it considers the glottalic character of Armenian dialectal stops as an inherited feature; and, it takes the geographical distribution of the dialects into account. The assumption that the glottalic feature of Armenian dialectal stops is original implies that Classical Armenian t possessed a glottalic feature which it had inherited from an Indo-European glottalic V.12 The starting point for Kortlandt therefore is the situation in Classical Armenian: Voiced d - Voiceless glottalic t - Voiceless Aspirates tThe first step in the ulterior development is the rise of aspiration in the voiced stops, as in the Pisowicz model, but in Kortlandt's model the rise is not general and affects only the central dialect area, creating a first split. From this 'central' group showing voiced aspirates there ultimately arise the dialects of groups 1 (Sivas) and 2 (Ararat). The dialects belonging to groups 7, Van, and 4, Cilicia) find their origin in another change, devoicing of Classical voiced stops d, again affecting only a part of the area. Groups 3 and 5 are later developments; group 6 has remained unchanged. The changes take place in subsequent steps which each time affect only one feature (aspiration, voice).13 The genesis of Armenian dialects has not been studied exclusively on the basis of the consonant system. In recent times a systematic relative chronology of other archaic (i.e. pre-sixth century) phonetic dialect features (Weitenberg 1996; Weitenberg 1999) has been elaborated; these studies revealed the existence of isoglosses which are in agreement with the early isogloss divisions which are assumed by Kortlandt. A study of the oldest isoglosses has shown that from the beginning of the literary tradition there have existed Armenian dialect variants next to the Classical Armenian. At

152

J.J.S. Weitenberg

the current stage of our research it can be shown in one or maybe two instances that there exist dialectal developments which antedate the fifth century AD; this means that - in the Armenian perspective - there are prehistoric dialectal isoglosses. These isoglosses concern the place of the accent in the Eastern dialects. The Eastern dialects show penultimate accentuation, which differs from the final stress in Classical Armenian and Western dialects;14 the feature can be shown to be older than the fifth century AD. What could be its lower chronological limit? Classical Armenian received its final shape through the First Sound Shift and through prehistoric apocope of final vowels (thus märd "man" derives from * mär do-, the genitive mardoy from *mard0syo). The apocope occurred after the proto-Armenian accent had become fixed on the (prehistoric) penultimate syllable; as a consequence of the apocope Classical and Modern Armenian words are stressed on the last syllable. Although the dialects have preserved inherited lexical and phonetic features which are not present in Classical Armenian (Jahukyan 1972: 277-330), everything points to the fact that they have fully participated in these two developments. Therefore, at the current stage of research, the formation of the Armenian dialects cannot be pushed back beyond the date of the apocope. This apocope is relatively young.15 With respect to the Eastern dialectal accentuation we grasp a postapocope period in which there were various tendencies to generalize the accent positions, under conditions which remain to be studied. This seems to be as far as we can go back now.

The Armenian dialect atlas In the 20th century, in particular after World War ii, Armenian dialect studies received a stable organization. Armenian dialect research is now done by the Dialectological Department of the Linguistic Institute in the Armenian capital of Erevan; this institute forms part of the Armenian Academy of Sciences. Thanks to the work of this institute, we now posses a description of most of the surviving Armenian dialects. The institute has prepared an (as yet unpublished) four volume dialect dictionary and has collected a huge mass of data which will serve a the basis for a dialect atlas. These data consist of about 500 uniform descriptions of dialect features which have been collected since 1976 on the basis of a "Questionnaire for Armenian Dialectological atlas materials" the main author of which was H. Muradian. The 800 features collected concern the lexicon, phonology, morphology, and syntax of Arme-

Aspects of Armenian dialectology

153

nian. Currently these data are being digitalized. If funding can be found, the manufacture of a traditional type of map on the basis of these homogeneously organized data will be relatively easy.

Notes 1. The Gypsy Armenian language is still in use. The Eastern Anatolian Armenian gypsies have become sedentary and now live in Armenia and Southern Georgia. 2. The poems and language of Sayeat Nova have now made accessible to wider circles by Dowsett 1997. 3. A detailed bibliography on Armenian ethnography and the epic of David of Sasun now is available with Avakian 1994. In this work entries are provided with translations and short summaries of their contents. An interesting survey of early Armenian dialect studies is given by Patkanov (1869: 1-17); for more details I refer to Djahukian's introduction to Greppin - Khachaturian (1986: 9-13).

4. K'nnakan k'erakanowt'iwn Asxarhabar kam ardi Hayeren lezowi, Vienna 1866. This grammar has been fundamental in the creation of the Western variant of current modern Armenian. Aytanean was a member of an Armenian monastic order, which to this day is established in Venice and in Vienna. The group derives its name from abbot Mechitar who founded the order in 1701. The Mechitarists aim at strengthening Armenian self identity by spreading Armenian scholarship and education. They did this by editing the Classical Armenian texts, by writing dictionaries and by founding schools. Because they lived in Europe, they easily entered into contact with European scholarship. It is for this reason, that 19th century Armenian scholarship, and also Armenian dialectology, has always been perfectly aware of contemporary European scientific and linguistic thought. For the Mechitarists' linguistic orientation see Adalian 1992. 5. Vogt had distinguished 6 groups, because he was unaware of the data of the Sassun (Cilicia) dialect. Gharibian's seven group classification incorporated also the Sassun data; for a comparison see Kortlandt 1978: 10. Important data and additions on Gharibian's grouping have been adduced by Pisowicz 1976. 6. Geographical areas are only given approximately. For a complete picture of the grouping also the development of stops in intervocalic position should be taken into account. Data, literature and listings in Pisowicz 1976; 1976a: 11. A map is a.o. given in Jahowkyan (1972), Vaux (1997: xiv). 7. For an overview of this classification see Djahukian (1984: 21 - 23), where, however, an error occurred (page 22, line 21) resulting in the omission of one of the groups, vi Mush Tigranakert. A complete listing is given in Jahowkyan 1972: 132-136. 8. To avoid overburdening this paper I take the liberty to simply refer the reader to the extensive bibliography and excellent summary of the discussion by Rüdiger Schmitt (1972: 7-8); Job (1977: 94-99). See also Pisowicz (1976a: 20-27), Vaux 1997: 238. The discussions have been published afterwards in one volume by Gharibian (Armjanskaja 1975). 9 See De Lamberterie 1984 evaluating Benveniste's views in this respect. 10. On the nature of the Armenian voiced aspirates also see Khachaturian 1992; Vaux 1997; 1998:214-215. Khachaturian's analysis is consistent with W.S. Allen's (1950) analysis of

154

J.J.S. Weitenberg

the New Julfa data, for which see also Kortlandt 1978:12. The phonemic interpretation of the data as voiced by Khachaturian is challenged by Pisowicz 1997; 1998 (with important remarks on the origin of the term "voiced aspirate" in Armenian linguistics): he remarks that the Armenian "voiced aspirates" are rather perceived as voiceless by speakers, and classifies them as voiceless murmured. 11. Vaux is only concerned with the mechanism of the subsequent development of the alleged common post-classical voiced aspirates 'dh etc. (resulting in dialectal dh, d, t, t' in the various groups). His listing does not claim to be a complete reconstruction of the development of the Armenian dialect groups, as it does not discuss the development of the Classical voiceless stops (Vaux 1997: 239). 12. On the consequences of Indo-European glottalics for Classical Armenian see Kortlandt 1978a; de Lamberterie 1992: 251-255; and (from a more general perspective) de Lamberterie 1998. Objections against a number of statements by in particular Gamkrelidze and Ivanov regarding Armenian dialect features (and rejection of the glottalic theory itself) are given by Pisowicz 1988; see on this also Pisowicz 1997: 217-218. 13. Critics on Kortlandt's model by Pisowicz (1997: 217) concern the following points: glottalics are not inherited, and the change from "voiced aspirate" to plain voiceless involves only one step. A response on this is Kortlandt (1998). 14. For a description of the modalities see Vaux 1997: 141-150. 15 . On the relative date of the apocope see Kortlandt 1980 (stage 16 out of twenty two stages; before the influx of Iranian loans). De Lamberterie 1992: 247.

References Acaiyan (Adjarian), Hrac'ya 1909 Classification des Dialectes Arminiens, Paris. Acaiyan, H. 1911 Hay Barbaragitowt 'iwn, Moskva - Nor Naxijewan. (Armenian Dialectology) Acaiyan, H. 1951 Hayoc' lezvi patmowt 'yown, Vol. n. Erevan (Haypethrat). (History of the Armenian Language) Adalian, Rouben Paul 1992 From Humanism to Rationalism. Armenian Scholarship in the Nineteenth Century, Atlanta (Scholars Press) (= University of Pennsylvania Armenian Texts and Studies No. 10). Adontz, Nicolas 1970 [1915] Denys de Thrace et les commentateurs Arminiens. Traduit du Russe. Louvain: Imprimerie Orientaliste. (Translated by R. Hotterbeex from the original edition St.-Peterburg 1915). Allen, W.S. 1950 "Notes on the Phonetics of an Eastern Armenian Speaker'Transactions of the Philological Society: 180-206. Armjanskja 1975 Armjanskaja Konsonantizm ν Ocenke Mezdunarodnoj Lingvistiki. Armenian Consonant System in the Qualification of International Linguistics. Retrospective information, Erevan. (With a preface by A.S. Gharibyan). (Academy).

Aspects of Armenian dialectology

155

Avakian, Anne M. 1994 Armenian Folklore Bibliography, Berkeley - Los Angeles - London (University of California Press). Axverdean, Georg 1852 Sayeat' - Növay, Moscow 1852. Bakker, Peter and Cortiade, Marcel, edd. 1991 In the Margin of Romani. Gypsy Languages in contact, Amsterdam (Inst. f. General Linguistics) (=Studies in Language Contact i). Bläsing, Uwe 1992 Armenisches Lehngut im Türkeitürkischen am Beispiel von Hemsin, Amsterdam (Rodopi). 1995 Armenisch-Türkisch. Etymologische Betrachtungen ausgehend von Materialien aus dem Hemsingebiet nebst einigen Anmerkungen zum Armenischen insbesondere dem Hemsindialekt, Amsterdam (Rodopi). Boretzky, Norbert 1995 "Armenisches im Zigeunerischen (Romani und Lomavren)", Zeitschrift f . indogermanische und allgemeine Sprachwissenschaft, 100: 137-155. Dankoff, Robert 1995 Armenian Loanwords in Turkish, Wiesbaden (Harrassowitz). De Lamberterie, Charles 1984 "Benveniste et la linguistique armenienne", in Jean Taillardat a.o., edd. E. Benveniste aujourd'hui. Actes du Colloque international du C.N.R.S., Tome n, Paris: 226-238. 1992 "Armenien", LAUES. Actes des sessions de linguistique et de litterature 10: 233289. 1998 "Methode comparative et approche typologique: regards croises sur les deux disciplines", La Linguistique 34 fasc. 1: 19-37. Dowsett, Charles 1997 Sayat' - Nova. An 18th-century troubadour. A Biographical and Literary Study, Leuven (Peeters) (= Corpus Scriptorum Christianorum Orientalium Vol. 561. SubsidiaT. 91). Feydit, Frederic 1964 "Problemes et utilite de la dialectologie armenienne", in Premier Congres International de Dialectologie generale Vol. i, Louvain: 146-153. Greppin, John A.C. - Khachaturian, Amalya A. 1986 A Handbook of Armenian Dialectology with an Introduction by Academician Gevorg B. Djahukian and an Excursus by Dr. H.D. Muradyan f, Delmar NY (Caravan Books). Jahowkyan, G.B. (Djahukian) 1972 Hay Barbaragitowt'yan Neracowt'yown (vicakagrakan barbaragitowt'yown), Erevan (Academy). (Introduction to Armenian Dialectology. Statistic Dialectology)· 1986 "Introduction" in Greppin - Khachaturian 1986: 9-26. Job, Dieter Michael 1977 Probleme eines typologischen Vergleichs iberokaukasischer und indogermanischer Phonemsysteme im Kaukasus, Frankfurt a.M. - Bern (Lang) (= Europäische Hochschulschriften, Reihe xxi. Bd. 2).

156

J.J.S. Weitenberg

Khachaturian, Amalya 1984 "The Nature of Voiced Aspirated Stops and Affricates in Armenian Dialects", Annual of Armenian Linguistics 4: 57-62. 1992 "Voiced Aspirated Consonants in the Nor Bayazet Dialect of Armenian'Proceedings of the Fourth International Conference on Armenian Linguistics, Delmar, New York (Caravan Books), 115-128. Kortlandt, F.H.H. 1978 "Notes on Armenian Historical Phonology n (The Second Consonant Shift)", Studia Caucasica 4: 9-16. 1978a "Proto-Indo-European obstruents", Indogermanische Forschungen 83: 107-118. 1980 "On the Relative Chronology of Armenian Sound Changes'', Proceedings of the First International Conference on Armenian Linguistics, Delmar NY (Caravan Books): 97-106. 1998 "Armenian glottalization revisited". Annual ofArmenian Linguistics 19: 11-14. fcaribyan, A. S. (Gharibian) 1953 Hay Barbaragitowt 'yown, Erevan (Pedagogical Institute). (Armenian Dialectology)· 1959 "Ob armjanskom konsonantizme", Voprosy Jazykoznanija 1959/5: 81 -90. Parnassian, Nevard A. 1985 "On the Formation of Ashkharhabar", Annual of Armenian Linguistics 6: 67-73. Patkanov, K. 1869 Izsledovanie ο dialektax armjanskago jazyka. Filologiceskij opyt, Sanktpeterburg. Pedersen, Holger 1905 Les pronoms demonstratifs de l'ancien armenien, = Det Kongelige Danske Videnskabernes Selskabs Skrifter. Sjette Rcekke. Historisk ogfilosofisk Afdeling vi.3, Copenhagen: 303-353. 1982 Kleine Schriften zum Armenischen, herausgegeben von Rüdiger Schmitt, Hildesheim - New York (Olms). Petermann, Heinrich Julius 1867 "Über den Dialect der Armenier von Tiflis", Berlin (= Abhandlungen der Königlichen Akademie der Wissenschaften zu Berlin, Philosophisch - Historische Klasse 1866 Nr. 1: 57-87). Pisowicz, Andrzej 1976 "Matöriaux pour servir ä la recherche du consonantisme armenien. Continuation dialectale des occlusives et affriquees de la language classique", Folia Orientalia 17: 197-216. 1976a Le Diveloppement du Consonantisme Arminien, Wroclaw etc. (Ossolineum), [= Prace Komisji Jezykoznawstwa Nr.43]. 1988 "Objections d'un arm£nologue contre la theorie glottale", Folia Orientalia 25: 213-225. 1989 "Die armenischen Entlehnungen in den türkischen Dialekten", Folia Orientalia 26: 123-129. 1997 "Consonant Shifts in Armenian Dialects During the Post-Classical Period Revisited", in: N. Awde ed. Armenian Perspectives. 10th Anniversary Conference of the Association Internationale des fctudes Armeniennes, Richmond, Surrey (Curzon - Caucasus World): 215-230.

Aspects of Armenian

1998

dialectology

157

"What Did Hratchia Adjarian Mean By 'Voiced Aspirates' in Armenian Dialects", Annual of Armenian Linguistics 19: 43-55. Schmitt, Rüdiger 1972 "Die Erforschung des Klassisch-Armenischen seit Meillet (1936)", Kratylos 17: 1-68. Schütz, Edmond 1980 "The Stages of the Armenian Settlements in the Crimea", Transcaucasica n (Quaderni del Seminario di Iranistica Uralo-Altaistica e Caucasologia dell' Universita degli Studi di Venezia), 116-135. Vaux, Bert 1997 'The Phonology of Voiced Aspirates in the Armenian Dialect of New J u l f a ' i n : N. Awde ed. Armenian Perspectives. 10th Anniversary Conference of the Association Internationale des Etudes Armeniennes, Richmond, Surrey (Curzon Caucasus World): 231-248. 1998 The Phonology of Armenian, Oxford (Clarendon Press). Vogt, Hans 1958 "Les occlusives de Γ Armenien", Norsk Tidsskrift for Sprogvidenskab 18: 143161. Weitenberg, Joseph J.S. 1984 "Evliya Chelebi on the Armenian Language of Sivas. Some Remarks", Annual of Armenian Linguistics 5: 99-108. 1996 "On the Early Development of the Armenian Dialects", in Dora Sakayan,ed. Proceedings of the Fifth International Conference on Armenian Linguistics, Delmar NY (Caravan Books): 93-118. 1999 "On the Early Development of Armenian Dialects, n. The Monophthongization of fly", Annual of Armenian Linguistics 20 (1999-2000): 1-26.

Typology and dialectology: a programmatic sketch Lieselotte Anderwald and Bernd Kortmann

1. Introduction This is a programmatic paper. Its intention is to invite collaboration on an area of comparative linguistics which has been largely neglected both by dialectologists and typologists: the comparative study of dialect syntax. In tackling this area we consider functional typology as the major source of inspiration and the provider of methods, relevant questions and promising explanations for the observable variation. In the course of the last two decades typological methods and typological thinking have opened up new perspectives both for historical linguistics (notably by putting grammaticalization on the agenda) and for contrastive linguistics (Hawkins 1986, 1992). In a similar way, the typological approach to the study of dialect syntax will add a fascinating new dimension to dialect research. Thereby we would come closer to the ideal of a unified approach to the study of variation in language, be it historical, cross-linguistic or language-internal.

historical linguistics

contrastive




typology

•




vorn

— >

oom

-0

iiüm (dar)an (dar) um (da) von daheim (dar)auf fort dort hier hest

-e(n)

— »

— >

— >

— »

— >

— >

— >

— >

-

»

unt hint uss inn

-en -en -en -e(n)

vor vorn ob om üb *üm

-n -e(n) -en -e(n) -en -e(n)

ann umm vonn heim (dr)uff fort dort hier t hest

-e(n) -e(n) -e(n) -e(n) -en -en -en -en -en

hess

-en

-et

—>

inn äuß hint

-er -er -er

öb vö(r)d ünt

-er -er -er

Als analogische Stütze haben hier sicherlich die semantisch benachbarten Dimensionsadjektive gedient, die bei Suffigierung mit -er im Komparativ auch umlauten: lang/länger, groß/größer, nah/näher usw. Als komparativisch lassen sich auch die Lokaladjektive auffassen, insofern nämlich, als etwa das untere nicht nur das absolut/positiv unten Befindliche bedeuten kann, sondern auch das relational gesehen weiter unten als X Befindliche.

Morphologische Reanalysen

201

3.2.2. Räumliche Präpositionen und Richtungsadverbien Handelte es sich bei der Umlautung der lokalen Adjektive um eine qualitative Änderung des Stammvokals, begegnen in einem weiteren Bereich dialektaler Restrukturierungen funktional eingesetzte quantitative Änderungen im Stammvokalismus.7 Wie bei den Vorgängen, die bisher dargestellt wurden, treiben auch hier Dialekte ein Prinzip der Form-Funktions-Beziehungen bis zur äußersten Konsequenz. Das vorbildgebende Muster bei diesem Phänomen ist lexikonfrequentiell nur einmal vertreten. Allein beim Paar der Basismorpheme in(n)/ein ist im Stammvokalismus der relevante Quantitätsunterschied bereits vorher vorhanden (vor-neuhochdeutsch als in(n)/in). In dieses Schema quantitativer Alternation passen sich jedoch zahlreiche andere Präpositionen und Adverbien ein. Ein langes Grundmorphem wie das der Präposition aus wird gekürzt zu uss. Umgekehrt wird ein kurzes Grundmorphem wie das des Richtungsadverbs ann gelängt zu aan. Damit wird eine durchgehende Polarisierung zwischen den Grundmorphemen von Präposition und Richtungsadverb erreicht. Gekürzte Präposition uss und lang gebliebenes Richtungsadverb aus stehen sich dann ebenso gegenüber wie kurz gebliebene Präposition ann und gelängtes Richtungsadverb aan. Daran sieht man, dass es im System nicht nur zu einer, etwa satzprosodisch erklärbaren, Abschwächung/Kürzung der im Akzentschatten des Vorfelds einer Nominalphrase stehenden Präposition kommt, sondern die formale Polarisierung an sich angestrebt wird. Sonst würden Grundmorpheme wie ann in Funktion als Richtungsadverb nicht gelängt. In Tabelle 7 sind die Fälle dieser Polarisierung im einzelnen aufgeführt.8 Nur das Basismorphem ümm der Richtungsadverbien n-ümm, r-ümm 'hinum, herum' entzieht sich - womöglich zur Homonymenmeidung mit dem langvokalischen Basismorphem üüm in h-iiüm, dr-üüm 'hüben, drüben' - der Längung.

202

Rüdiger Harnisch

Tabelle 7. Alte Kürze und neue Kürzung bei Präpositionen. Alte Länge und neue Längung bei Richtungsadverbien Präpositionen mit Kurzvokal inn uss uff vorr förr zu ze änn äbb vunn mitt

Langvokal

hess-en zu sehen. 6. In der noch lokalen Bedeutung 'zurück'. 7. Die quantitative Opposition besteht hier zwischen kurzen Monophthongen auf der einen und langen Monophthongen sowie Diphthongen auf der andern Seite. Wenn im Folgenden von vokalischer Länge die Rede ist, sind also Diphthonge und lange Monophthonge gleichermaßen gemeint. 8. Kürzungsprodukt des Diphthongs au (in auf und aus) ist u (in uff und uss), die kurze Entsprechung von diphthongischem ei (in ein) ist i (in inn). Umgekehrt sind jedoch Längungsprodukte von u (in vunn) nicht *au (in *vaun) und von i (in mitt und hinn) nicht *ei (in *meit und *hein), sondern die entsprechenden langen Monophthonge uu und ii (in vuun bzw. miit und hiiri). Zu dieser Erläuterung hat mich der Diskussionsbeitrag von Ton Goeman veranlasst, dem ich dafür danke. 9. Ursprünglich ist vor nur der Stamm für statische, also Orts-Ausdrücke, fiir nur der Stamm für dynamische, also Richtungs-Ausdrücke. Vgl. die dialektalen Sätze stell 'sfiirrdie Tür 'stelle es vor die Tür' und es steht vorr der Tür 'es steht vor der Tür' mit Präposition bzw. geh fiiür 'geh vor' und bleib vorr-ne 'bleib vorn' mit Adverb. Ein dynamisches geh vuur 'geh vor!' ist neuer und vermutlich von der Standardsprache beeinflusst, die fiir nicht mehr in räumlicher Bedeutung kennt. 10. Mhd. zuo (Adverb, Präposition). 11. Mhd. ze (Präposition). 12. Die Adverbien hin und her haben keinen präpositionalen Widerpart. 13. Wahrig (1978) zum Beispiel bucht unter hinter kein Adverb. 14. Im Ostfränkischen und Thüringischen kommen sogar die nach Sprecherperspektive unterscheidenden präfigierten Richtungsadverbien n-hinter und r-hinter 'hin-hinter'/'her-hinter' (mit phontaktisch bedingtem Schwund des stammanlautenden h) analog n-unter, runter 'hinunter, herunter' vor. Zu weiteren phonotaktischen Beschränkungen beim Aufeinandertreffen von präfigierten und stamminitialen Lautsegmenten siehe Harnisch (1989: 447-448). 15. Im Hochdeutschen kommen nur die nach Sprecherperspektive unterscheidenden präfigierten Formen h-üben und dr-üben vor. Im Oberostfränkischen gibt es auch unpräfigiertes üben (wie außen).

206

Rüdiger Harnisch

Literaturverzeichnis Eisenberg, Peter 1992 "Suffixreanalyse und Syllabierung. Zum Verhältnis von phonologischer und morphologischer Segmentierung", Folia Linguistica Historica 13: 93-113. 1995 "Grammatik der geschriebenen Sprache als Symbolgrammatik. Ein Versuch an ausgewählten Themen der Morphologie", in: Vilmos Ägel und Rita Brdar-Szabo (eds.), Grammatik und deutsche Grammatiken. Tübingen: Niemeyer, 23-38. Fuhrhop, Nanna 1998 Grenzfälle morphologischer Einheiten. Tübingen: Stauffenburg. Harnisch, Rüdiger 1980 "Die Verknüpfung von Orts- und Richtungsadverbien in der Mundart von Ludwigsstadt (Oberfranken). Morphologische und semantische Strukturen", in: Rowley (ed.), 227-247. 1989 "Lokalpartikeln als Kerne lokalmorphologischer Paradigmen", in: Harald Weydt (ed.), Sprechen mit Partikeln. Berlin/New York: Walter de Gruyter, 441-450. Hinderling, Robert 1980 "Die Richtungsadverbien im Bairischen und Alemannischen", in: Rowley (ed.), 249-296. 1992 "Der Sprachatlas von Nordostbayern und der Gesamtbayerische Sprachatlas", in: Andreas Weiss (ed.), Dialekte im Wandel. Göppingen: Kümmerle, 27-44. Jespersen, Otto 1925 Die Sprache. Ihre Natur, Entwicklung und Entstehung. Heidelberg. Lang, Ewald 1989 "Primärer Orientierungsraum und inhärentes Proportionsschema", in: Christopher Habel et al. (eds.), Raumkonzepte in Verstehensprozessen. Interdisziplinäre Beiträge zu Sprache und Raum. Tübingen: Niemeyer, 150-173. Pfeifer, Wolfgang 1993 Etymologisches Wörterbuch des Deutschen. 2. Aufl. München: Deutscher Taschenbuchverlag. Rowley, Anthony (ed.) 1980 Sprachliche Orientierung I. Untersuchungen zur Morphologie und Semantik der Richtungsadverbien in oberdeutschen Mundarten. Bayreuth: Sprach- und Literaturwissenschaftliche Fakultät, Universität Bayreuth. [Später] Hamburg: Buske. Spangenberg, Karl 1994 Laut- und Formeninventar thüringischer Dialekte. Berlin: Akademie-Verlag. Thüringisches Wörterbuch 1966 ff. Bände 4-6 bearbeitet unter der Leitung von Karl Spangenberg. Band 1 ff. fortgesetzt unter der Leitung von Wolfgang Lösch. Berlin: Akademie-Verlag. Wagner, Eberhard 1970 "'Diesseits' - 'jenseits' im Ostfränkischen", Zeitschrift fiir Dialektologie und Linguistik 37: 158-171. Wahrig, Gerhard (ed.) 1978 dtv- Wörterbuch der deutschen Sprache. München: Deutscher Taschenbuchverlag. Wurzel, Wolfgang U. 1992 "Morphologische Reanalysen in der Geschichte der deutschen Substantivflexion", Folia Linguistica Historica 13: 279-307.

Basque accentuation and dialectology Jose Ignacio Hualde

1. Introduction Our knowledge of Basque accentual systems has increased dramatically in the last few years (Elordieta 1997, Etxebarria Ayesta 1991, Etxebarria 1985, Gaminde 1993, 1994a-d, 1995a-d, 1996,1998, Hualde 1989, 1990,1991a-b, 1992, 1993a-b, 1995a-b, 1996, 1997, 1998, 1999, 2000, Hualde and Bilbao 1992, 1993, Hualde, Elordieta and Elordieta 1993, 1994, Hualde and Sagarzazu 1991, Ibarra 1995, Jansen 1992). The resulting picture is one of great complexity and typological diversity, greater than that found anywhere else in Europe within a single language and perhaps as great as that encompassed by a language family like Slavic. We are forced to revise earlier attempts at classifying Basque accentual types (by Azkue 1931-32 and Michelena 1972) in view of the much more extensive information brought forward by recent fieldwork and analysis. The fit between accentually defined geographical areas and traditional dialects (which were established according to morphological features) must also be reconsidered. Basque is currently spoken in an area of southwestern France and northern Spain along the Bay of Biscay and on both sides the western Pyrenees. Of the 650,000 or so speakers of Basque, most are concentrated in the territory of the Autonomous Community of the Basque Country in Spain (which includes the provinces of Bizkaia, Gipuzkoa and Alava or Araba). Basque is also spoken in the northwestern part of the province of Navarre (which is presently a separate Autonomous Community of Spain) and, in France, in the western half of the Departement des Pyrenees Atlantiques. The area of France where Basque is spoken includes the three historical provinces of Labourd or Lapurdi, Low Navarre (which formerly was part of the Kingdom of Navarre) and Soule or Zuberoa. In all, Basque is thus spoken in parts of seven historical territories or provinces. In all seven provinces Basque historically was the majority language. In the last few centuries, however, the southern boundary of the language has receded considerably, and only small

208

Jose Ignacio Hualde

parts of the two southernmost provinces of Alava and Navarre are still Basque-speaking. In other Basque provinces too inmigration and other factors have resulted in the present situation where Basque is spoken only by a minority of the population and practically all Basque-speakers are bilingual in Spanish or French. Basque is today an official language in the Autonomous Community of the Basque Country (along with Spanish) and also enjoys a more reduced official status in Navarre. In France, it lacks official recognition. Officialization in the Spanish Basque Country, which has taken place in the last twenty years, has had two important consequences. First of all, especially because of schooling in Basque, nowadays one can find (bilingual) Basque-speakers in areas where the language had formerly been lost. Secondly, standard Basque, which was developed only in the 1960s and 70s, is now used by many, especially young, speakers either along with a local variety or as the only kind of Basque that they speak. In central areas, where the local variety may not differ significantly from the standard, the younger generations tend to mix both codes, creating an intermediate variety (cf. Elexpuru 1996). In this paper we will be concerned only with the accentual properties of "traditional" Basque varieties, as have been transmitted through the generations in the areas where the language has been spoken without interruption, and not with the prosodic system employed in Standard Basque in different areas.

2. Basque dialects and their classification The existence of a high degree of dialectal fragmentation within Basque has often been remarked upon. Some of the first writers to make use of the Basque language in the 16th and 17th centuries already commented on this fact, since they had to deal with it to write their books. For instance Joannes Leisarrague in the foreword to his 1571 translation of the Gospels makes reference to the fact that Basque is spoken in different ways "almost from house to house" (Zuazo 1988a:87). On the other hand, historical linguists such as Luis Michelena have pointed out that a major problem for the reconstruction of early stages of the language is that all the modern dialects of Basque are very much alike. Michelena (1981) concludes that diversification into the modern dialects is a relatively recent phenomenon starting from a more-or-less homogeneous form of speech formed sometime in the Middle Ages.

Basque accentuation and dialectology

209

Opinions have ranged from considering that Basque, more than a language, is a group of related languages to the view that Basque dialects do not differ from each other more than English dialects do (see Zuazo 1988a). As with other languages without a common written tradition and no official status, some local varieties may appear initially very hard to penetrate since they have developed without awareness of the existence of prestige forms beyond the local community. But linguistic scrutiny reveals a high level of homogeneity in lexicon, nominal morphology, and syntax across all dialects. One aspect in which local dialects do differ in important ways is accentuation (there are also other aspects where considerable variation is found, such as verbal morphology). As was mentioned, in the last few decades a written standard has been codified and most younger speakers are now able to write and speak in the standard. This may have an important impact on the current dialectal picture, including accentuation. The groundwork on Basque dialectology was laid at the end of the 19th century by the French prince Louis-Lucien Bonaparte. Bonaparte (1863, 1869) produced a classification of major dialects, subdialects and varieties which has had long-lasting impact and for the most part is still used nowadays. Bonaparte's classification has been the basis for all subsequent work on Basque dialectology. Some changes in this dialectal classification were introduced by R.M. Azkue (1905) in the first comprehensive dictionary of all Basque dialects, and, subsequently, by the great Basque linguist Luis Michelena (1961). Michelena's classification, which can be considered the current standard classification, is shown in map 1 (from Trask 1997). In this classification, the following major dialects are recognized: Bizkaian (B), Gipuzkoan (B), Lapurdian (or Labourdin, L), High Navarrese (HN), LOW Navarrese (LN) and Zuberoan (or Souletin, z). In addition, Michelena gave independent status to three dialects spoken in three valleys on the Spanish side of the Pyrenees: Aezkoan (Aezk), Salazarese (Sal), and Roncalese (R). Bonaparte had considered these last three to be subdialects of dialects spoken on the French side. It should be noted that Azkue's and Michelena's reforms of Bonaparte's classification are relatively minor and do not affect dialectal boundaries. For instance, as can be seen in map 1, the Bizkaian dialect extends over an area of western Gipuzkoa including towns such as Eibar, Bergara and Arrasate, and High Navarrese extends over an eastern area of Gipuzkoa including the town of Irun. These boundaries and others have been left unmodified from Bonaparte's map. Similarly, the boundaries of subdialects and varieties (not shown in the map) fixed by Bonaparte are generally taken

210

Jose Ignacio Hualde

for granted in modern dialectological work. It can be noticed that the major dialectal boundaries in general coincide with those of the historical provinces, although there are some exceptions like those just mentioned. In a way, the success achieved by Bonaparte's dialectal classification is somewhat unexpected, given the fact that it was elaborated according to criteria that seem antiquated to us. Bonaparte had an essentialist view of dialects. For him dialects were compact entities with well-defined boundaries. To establish these boundaries he gave primary importance to certain morphological features (the structure of conjugated verbal forms) and when more than one competing form was found in the same area, he would chose one as the genuine form for that particular subdialect or variety. When chopping up the dialectal continuum into well-defined parts proved difficult, rather than admitting failure, he would declare that the dialect spoken in a particular area was a "mixed" dialect rather than a "pure" one. Obviously the fact is, however, that what one finds is individual isoglosses with different distributions. There are points where large number of isoglosses converge but usually this convergence does not extend over the whole length of Bonaparte's dialect boundaries. For instance, Michelena (1981) remarks that the boundary between Bizkaian and Gipuzkoan is rather sharp in the mountainous south of the area, with a large number of differences between towns on opposite sides of the dialectal border; but as one moves north toward the coast this dialectal boundary tends to fan out and difuminate. The Bonapartian view of Basque dialects such as Bizkaian or Zuberoan as separate entities with sharp boundaries has also recently been challenged by modern researchers, such as Lakarra (1986) and Zuazo (1988b). But the fact is that, in practice, most dialectological work still accepts Bonaparte's classification of dialects, subdialects and varieties as a point of departure (cf., for instance, Yrizar 1981, 1991, 1992, Etxebarria 1983). As mentioned, Bonaparte's classification is based for the most part on morphological differences. The distribution of lexical variants has recently been the object of dialectal work by Jacques Allieres (1995), based on data from the Ethno-Linguistic Atlas of the Basque Country (Arantzadi Zientzia Elkartea 1990). The way lexical isoglosses bundle together tends to support Michelena's observation that the boundary of the Bizkaian dialect is sharper in the south than in the north. Allieres also shows that for other lexical items the isogloss goes more or less along the Spanish/French border. The easternmost dialect, Zuberoan or Souletin is also characterized by a number of particular lexical variants. Other isoglosses which divide the territory into an eastern and a western area fall somewhere between Navarre and Gipuzkoa.

Basque accentuation and dialectology

211

However, although some of these closely follow the Navarrese/Gipuzkoan boundary, here there is a wide transitional area. The one major dialect defined by Bonaparte, and accepted by later researchers, for which Allieres' maps provide no support at all is Lapurdian. The importance of Allieres' work is that it offers us additional criteria for the classification of Basque dialects which are independent from those used by Bonaparte. Here I want to consider another aspect of Basque linguistic structure that Bonaparte did not take into account in his classification: accentuation.

3. Previous work on the classification of Basque accentual systems Aside from brief mentions by earlier writers, the first author to take accentual variation within Basque into account was R.M. Azkue (1923, 1931-32), founder of the Basque Academy. Azkue was mostly concerned with describing what he took to be the "general Basque" accentual type, since one of his goals was the codification of a standard language. However, he became painfully aware that the description that he provided could not be valid for all Basque dialects. In particular, Azkue mentions two areas where his general accentual type is not used: the easternmost Souletin or Zuberoan area and the Bidasoa Valley, which extends over a small area of northern Navarre and Gipuzkoa (see map 2). Azkue was most displeased by this lack of uniformity in accentual matters in the Basque territory: "Aunque con pena, hay que confesar que el acento vasco varia mucho entre sus dialectos. En dialecto suletino y en la variedad altonabarra de Bortzerrieta (Cinco Villas: Bera, Etxalar, Lesaka, Aranatz e Igantzi) tiene la misma intensidad que en espanol, con la diferencia que los suletinos cargan su extrano acento en la penültima silaba por lo general, mientras que los nabarros de la citada comarca tienen lo mismo que en castellano, vocablos agudos, graves y esdnijulos" [Azkue 1931:283] ["Unfortunately, we must confess that Basque accentuation varies very much among dialects. In the Souletin dialect and in the High Navarrese variety of Bortzerrieta (Cinco Villas: Bera, Etxalar, Lesaka, Aranatz and Igantzi), it has the same intensity as in Spanish, with the difference that the Souletin put their strange accent on the penultimate syllable as a general rule, whereas the Navarrese of the abovementioned area have, as in Spanish, oxytonic, paroxytonic and proparoxytonic words"]

212

Jose Ignacio Hualde

As for the "general Basque" accentual type, Azkue describes a pitchaccent system with two main, lexically determined, accentual patterns: "Unos vocablos son monotonos y tienen tantos acentos de intensidad como silabas; otros son ditonos y tienen tambien tantos acentos de intensidad como silabas, menos la ultima que cae en la atonia [Azkue 1931:284]. ["Some words are monotonic and have as many stresses as syllables; other words are ditonic and also have as many stresses as syllables, except for the last one which falls into atony"] Azkue (1923, 1931-32) described this system, which he took to be nearuniversal in Basque (with the two exceptions of Zuberoan and the High Navarrese variety of the Bidasoa Valley, as mentioned), in some detail, providing abundant exemplification of accentual contrasts. Azkue's description of Basque accentuation had some serious unintended consequences, setting the stage for a period of enormous chaos in the study of accentuation in Basque. Azkue's contemporary, and fellow member of the Basque Academy, S. Altube (1934) argued that Azkue's description was fundamentally wrong and that, contrary to Azkue's claims, all Basque words in a given syntactic context are accented in exactly the same manner. Another author and member of the Basque Academy, N. Ormaetxea (1919), provided a description of Basque accentuation which was incompatible with both Azkue's and Altube's. The problem is that the area with pitch-accent systems of the type Azkue describes is considerably smaller than he assumed. Furthermore some of the details he gives for the functioning of this accentual system, far from representing general Basque accentuation, are only valid for the variety spoken in his native town of Lekeitio. Azkue thus described a geographically very restricted accentual pattern but claimed that his description corresponded to Basque in general, leaving aside a couple of marginal areas. Altube and Ormaetxea did exactly the same thing. Each of them described the accentual system of their particular native town under the rubric of "Basque accentuation". These researchers failed to realize how much variation there actually is within the Basque language in accentual matters. The fact is that in spite of a certain apparent uniformity, the details of Basque accentuation vary greatly even within relatively small regions. Michelena (1957, 1972) brought some order back to the study and classification of Basque accentual types (see map 3). Michelena sides with Azkue in recognizing a broad central-western pitch-accent type (his type i), employed in most of Bizkaia and Gipuzkoa and some neighboring areas of

Basque accentuation and dialectology

213

Navarre. He was able to avoid Azkue's problems by being less specific about accentual details. Michelena states that in this central-western system "Noun forms fall, as far as the accentual pattern goes, into two classes: words belonging to the unmarked class... characterized by a sustained accent, and words with a marked accent, signalized by a falling contour" (Michelena 1987[1972]:236). Michelena also follows Azkue in giving a separate status to the accentual systems used in the Souletin or Zuberoan dialect (to which he added Roncalese) and in the variety of the Bidasoa Valley (his types n and iv, respectively). In the Zuberoan-Roncalese accentual type there is regular penultimate accent, as in gizun 'man', gizitna 'the man', alhäba 'daughter', and exceptionally final accent, as in gizumk 'the men, ergative', alhabä 'the daughter'. Final accent often results from an earlier contraction: *gizon+ägJrek> gizonek, alhabd+a > alhabä 'the daughter'. In the system used in the Bidasoa Valley (Michelena's type iv), on the other hand, the unmarked pattern is postinitial accent (on the second syllable) and in the marked case there is initial accent. Michelena is more accurate than Azkue in defining both the characteristics and the geographical extension of these two accentual types. To these three types adopted from Azkue, Michelena adds and defines a further accentual type (his Type hi) employed in the southern High Navarrese area. This type of accentuation "sounds like Spanish", in his words, but without phonological contrasts, the stress usually falling on the penultimate syllable. In recent years a fair amount of fieldwork on Basque accentuation has been undertaken (cf. the references in the first paragraph of this paper) and we now know that the situation is rather more complex than Azkue and Michelena had anticipated. In particular, it now appears that the Bizkaian and Gipuzkoan area, which Azkue and Michelena include under a single accentual type, presents a high degree of internal variation. A new and more complex classification of Basque accentual types is attempted in Hualde (1995a), where a much greater number of types and subtypes are recognized than in previous classifications. In this paper, rather than trying to reduce the existing geographical variation to a classification of major and minor accentual types, I will explore the possibility of defining broad dialectal areas based on those accentual isoglosses that appear to have the greatest importance. Our results will then be compared to those of earlier researchers and consequences will also be drawn for our general view on Basque dialectology.

214

Jose Ignacio Hualde

4. Major accentual isoglosses 4.1. Contrastivevs. noncontrastive accent A first major accentual feature that we may want to consider is the contrastive or noncontrastive nature of the accent; that is, whether or not the position of the accent is used to establish lexical or morphological oppositions. In this respect, Basque dialects vary widely, from some where the position of the accent determines numerous lexical and morphological contrasts to others where accent placement is not phonologically distinctive, all words in a given context being accented in the same manner. In map 4 the areas where accent placement has a contrastive function and to a greater or lesser degree is relevant for meaningful distinctions are separated from those other areas which completely lack accentual contrasts. It can be seen that there are two main large areas where there are accentual contrasts and a large central area without such contrasts. Accent is contrastive in most of the varieties spoken in the western provinces of Bizkaia and Gipuzkoa and a neighboring area of Navarre. In the western area it is only along the western and southern boundary of the language that accent is not contrastive. We can assume that accentual contrasts have been lost in these border regions, perhaps relatively recently. In part of the western area with contrastive accentuation we find a typologically very interesting pitch accent system, first described by Azkue (193132) as we saw above, where a marked class of words with a fall in pitch in some syllable is opposed to a larger unaccented class. In a phrase the pitch raises on the second syllable (unless the first is accented) and remains high up to the accented syllable or up to the final syllable if there are no lexical accents (Basterrechea 1974-75, Hualde 1989). Whereas marked or accented words consistently show a drop from high to low pitch in a given nonfinal syllable, unaccented words in phrase-medial position are realized on a level tone and without prominence on any syllable. Phrase-finally (and in citation form) unaccented words may receive final prominence (penultimate in some varieties). For instance, in the examples in (1) all the words are lexically unaccented, whereas in (2) the words /lengusu/ 'cousin' and /leku/ 'place' have an accent on the first syllable:

Basque accentuation and dialectology

215

(1) Phrase-final accent in northern Bizkaian gi χοηά \ 'the man' gi xon ederrä \ 'the handsome man' guI re gixon ederrari \ 'to our handsome man' g" re herriko gixon ederrari \ 'to the handsome man of our village' (2) Accentually marked words in northern Bizkaian len gusua 'the cousin' Ιέη gusu e derrä \ 'the handsome cousin' gu

re Ιέ kuko gi xon ederrari \ 'to the handsome man of our place'

Pitch-accent systems of this type are nowadays restricted to the northern Bizkaian coastal area, although in the past similar prosodic systems may have been used in most of the Bizkaian and Gipuzkoan dialect area (Hualde 1993a, 1995d). Accent is also contrastive in the easternmost dialects, Zuberoan and Roncalese, in the way described in section 3. We will come back again to the question of what kind of contrasts are expressed by accentual means in different areas. As noted in map 4, there is a third area, or rather two separated areas in Lapurdi and Navarre where there are surface accentual contrasts which are completely dependent on the segmental content of the items. In this area, accentuation is generally penultimate but whereas stem-final mid vowels become glides before a suffix-initial vowel, as in säre 'net', /sare-a/ [säija] 'the net', stem-final high vowels keep their syllabicity in the same context, as in sari 'prize', /sari-a/ [saria] 'the prize', thus giving rise to accentuation in different positions. In all likelihood the two areas with this special phenomenon were geographically connected at some historical point. 4.2. Accent placement A second important aspect of accentual systems, which we must consider, is the position of the accent in the unmarked or general case. As can be seen in

216

Jose Ignacio Hualde

map 5, we find a considerable amount of variation in this respect. There is a large eastern area where in the unmarked case accent is penultimate (= [-2]). The greatest complexity occurs in the west. In the western provinces of Gipuzkoa and Bizkaia and a neighboring region of Navarre (the Bidasoa Valley) the more widespread pattern is accent on the second syllable (= [+2]), but there are smaller areas with other patterns, such as accent on the third syllable (= [+3], provided that this rule does not place the accent on the last syllable) and also final (= [-1]) and penultimate (= [-2]) accent as the unmarked option. These different western patterns are illustrated in (3): (3) Representative western varieties: unmarked accentual Azkoitia Gernika Antzuola 'the head' burui buriia burüe 'the man' gizona gixonä gizona 'to the man' gixonari gizonäi gizonäi 'the mayor' alkatia alkatie alkatie alkatidi 'to the mayor' alkatieri alkatiei [+3] [-2] [-1] (phrasal)

pattern Beasain burue gizona gizonai alkdtea alkäteai [+2]

As was mentioned before, in some of these varieties, those spoken in northern Bizkaia, including Gemika in (3), the prosodic system is of the pitchaccent type. Arguably, this pitch-accent system with unmarked phrase-final prominence represents the most ancient western accentuation, the historical development being [-1] > [-2] > [+3] > [+2] (cf. Hualde 1993a, 1995d, 1997 a-b). Another phenomenon has been the loss in most of the area of the nonaccentual tonal rise on the second syllable and high-toned plateau up to the accented syllable which characterizes phrasal domains in northern Bizkaian. In some places, the rise on the second syllable has been reinterpreted as an accent, resulting in a more direct change into a [+2] system. An independent change has been the loss of the lexical distinction between accented and unaccented words. As the examples in (3) show, varieties such as those of Antzuola, Azkoitia and Beasain agree in the accentuation of short words, which are the most frequent ones, but disagree on the position of the accent in longer words, requiring quite different rules in each of the three dialects for their analysis. In most western local dialects, there is a smaller class of words with a different accentual pattern. This was shown above in (2) for northern Bizkaian pitch-accent systems such as that of Gernika, where the distinction

Basque accentuation and dialectology 217 is between unaccented words, subject only to phrase-final accentuation, and words with lexical accent on a given syllable. On the other hand, in a variety like that of the Beasain, for instance, where words are generally accented on the second syllable, there is a small class of words which has the accent on the initial syllable. An interesting fact is that we find a surprising amount of agreement among the different western varieties regarding which specific lexical and morphological items belong to the accentually marked class. Accentually marked words include borrowings such as /bäso/ 'glass' (< Sp. vaso) or /leku/ 'place' ( < Old Sp. lueco), compounds such as /lengusu/ (= le(e)n 'first' + gusu 'cousin' < Rom. cusinu) as well as a few native words like /beste/ 'other'. In Gaminde and Hualde (1995), the geographical distribution of several of these items is examined in detail. A finding of that study is that there is a high level of agreement among local varieties in which items belong to the accentually marked or irregular class (the agreement is not complete because we also find items which are accentually marked in some varieties but not in others). In all localities for which we have data in the western area, the borrowing /bäso/ 'glass', for instance, presents irregular accent and may contrast with /baso/ 'forest'. The specific realization of the contrast, however, can vary considerably; e.g.: bdsua 'the glass' vs. basita 'the forest', or accented bdsua vs. unaccented basud, or basua vs. basud, etc. In addition, all plural words usually present marked accentuation in the western area. That is, in most of the western area with contrastive accentuation, a singular/plural contrast is made (see map 6). In all these varieties with a singular/plural accentual contrast, the accent is placed farther to the left in the plural than in the singular. For instance, in areas where the general rule is [+2], in the plural we find initial accent: gizona 'the man' vs. gizonak 'the men', eskua 'the hand' vs. eskuak 'the hands'. Similar contrasts obtain in other areas. In the northern Bizkaian pitch accent system, plural forms have lexical accent generally on the last syllable of the stem. We may thus have gizona 'the man' (lexically unaccented, with phrase-final accent, cf. gizona dd 'it is the man') vs. gizonak 'the men' (and gizonak dire 'they are the men'), gizonari 'to the man' vs. gizonari 'to the men', eskua 'the hand' vs. eskuak 'the hands', etc. For the varieties illustrated above in (3), the marked accentual patterns presented by plural forms and by lexical items such as /lengusu/ 'cousin' and /leku/ 'place' in (4), can be compared:

218

Jose Ignacio Hualde

(4) Plural and irregular words Gernika 'the heads' burüek 'the men' gixonak gixonari 'to the men' 'the place' 'the cousin' 'to the cousin'

lekue Ungusue lengusueri

Antzuola biiruak gizonak gizonai

Azkoitia büruek gizonak gizonai

Beasain büruek gizonak gizonai

lekua lengusua lengusuai

lekue lengusue lingusuei

lekue leengusue leengusuei

By comparing the examples in (3) and (4), we see that even though these four varieties have different accentual rules, they agree on which words belong to the general accentual pattern and which are special or marked. This high degree of agreement in the whole of the western area on which items present irregular accentuation is perhaps surprising given the fact that the area includes varieties with very different accentual rules. The agreement is, thus, at a relatively "abstract" level: on which words and morphological forms are grouped together for accentual purposes, rather than on the specific position of the accent. The explanation must be that all these accentual systems are historically related, and have evolved in different directions, giving rise to different accentual rules but keeping the original distinction between accentually marked or exceptional and unmarked or regular items. Line A in map 7 marks the limit of the area where words such as /bäso/ 'glass', /beste/ 'other', /leku/ 'place' and many others receive irregular accentuation. In spite of the multiplicity of accentual rules included in the area to the west of this line, we must take these western accentual systems to be related, at least from a historical point of view. Thus, we agree with Azkue and Michelena in recognizing a western accentual type extending over most of the Bizkaian and Gipuzkoan territory as well as neighboring areas. Our western accentual area also includes the High Navarrese Bidasoa Valley, which Azkue and Michelena classify as possessing a distinct accentual type. It does not seem that the accentuation used in this area is sufficiently distinct to merit its separation from the western area. This is because the Bizkaian-Gipuzkoan area encompasses a much greater diversity of accentual rules than earlier researchers such as Azkue and Michelena seem to have suspected. As we have indicated, a commonality of origin, however, can be discovered in the fact that the same specific lexical items appear with a marked accentual pattern throughout the area. The area where this phenomenon is found extends to the Bidasoa Valley. It seems that one reason why Michelena viewed

Basque accentuation and dialectology

219

the Bidasoa Valley accentual system as substantially different from the Bizkaian-Gipuzkoan one is that singular and plurals receive the same accentuation in this area. But recent dialectological work has shown that the singular/plural distinction is not uniformly observed in the whole of the BizkaianGipuzkoan area either (map 6). The general accentual rule found in the Bidasoa Valley, [+2], is also common in several Bizkaian and Gipuzkoan areas and, on the other hand, is not found to the east of our line A, which defines the boundary of our western accentual type. By taking into account both whether accent is contrastive or not and the position of the accent in the unmarked case, we can also trace a second major accentual boundary. In map 7, line Β separates out the easternmost dialects, Zuberoan and Roncalese. In this eastern area, the accent generally falls on the penultimate but, as was mentioned above, there are lexical and morphological exceptions to this general pattern. Like in the western area, singular and plural forms present different accentual patterns. However, the contrast is rather different and appears to be historically unrelated. Whereas in western varieties where singular and plural forms are accented differently the accent occurs earlier in the word in the plural than in the singular, in the Zuberoan/Roncalese system we find the opposite. In Zuberoan and Roncalese singular and plural forms are accented in the same way in the absolutive case: giziina 'the man, absolutive sg.', gizünak 'the men, absolutive pi.'. But other plural forms have oxytonic accent: gizunek 'the men, ergative pi.', gizuner 'to the men, dative pi.'. As was explained above, it has been argued that plural forms with oxytonic accent are historically contracted forms. Map 6 shows the existence of a third, smaller, area with a singular/plural accentual distinction, in Navarre. In this area the distinction is less robust than in the other two. Singular forms are accented on the first and final syllables, [+1, -1]. In plural words, on the other hand, the accent falls on the syllable containing the plural marker, which can be the final or the penultimate; e.g.: iskuu 'the hand' vs. eskük 'the hands'. This is reminiscent of the Zuberoan/Roncalese situation, since here too the marked accentuation of plural forms seems to be the result of a historical contraction. In the area between lines A and B, we find for the most part prosodic systems lacking contrastive accentuation; that is, all words are accented in the same manner. In the southwestern corner of the Basque-speaking area of Navarre both the initial and the final syllables are accented and it is difficult to determine which one of these accents is primary and which is secondary. But in most of the areas without contrastive accentuation, the accent is generaly penultimate. This applies to the areas mentioned above without

220

Jose Ignacio Hualde

lexical accentual contrasts but with surface contrasts resulting from syllabification.

5. Conclusion The goal of this paper has been to offer an overview of geographical variation regarding accentuation in the Basque-speaking territory and to try to establish criteria for the classification of the varied accentual systems found in this language. We have considered three factors: whether the position of the accent plays a contrastive role in the phonology of the dialect, the rule determining the position of the accent in the unmarked or general case, and, in dialects with contrastive accentuation, which specific lexical items and morphological categories present a marked accentual pattern. We have argued that a combination of these criteria allows us to draw two major accentual boundaries, thus defining three large geographical areas. These conclusions deviate in part from previous classifications of Basque accentual types, but are based on a much more complete knowledge of the facts than was available until recently. If we compare our results with the standard morphologically-based classification of Basque dialects, they coincide in separating out the eastern Zuberoan/Roncalese area. Our western accentual type, on the other hand, covers most of the area of two major dialects (Bizkaian and Gipuzkoa) and part of a third (High Navarrese). The conclusion must be that future work on Basque dialectology must take accentuation into account as an important aspect of dialectal variation.

Basque accentuation and dialectology

Maps Map 1. L. Michelena's classification of main Basque dialects

Β G HN L LN ζ Aezk Sal R

= Bizkaian (Biscayan, vizcaino, bizkaiera) - Gipuzkoan (guipuzcoano, gipuzkera) = High Navarrese (altonavarro, goinafarrera) = Lapurdian (labourdin, labortano, lapurtera) = Low Navarrese (bas navarrais, bajonavarro, behenafarrera) = Zuberera (souletin, suletino, zuberera) = Aezkoan (aezcoano, aezkera) = Salazarese (salacenco, zaraitzuera) = Roncalese (roncales, erronkariera)

221

222

Jose Ignacio Hualde

Map 2. R.M. Azkue's classification of Basque accentual types

Map 3. L. Michelena's classification of Basque accentual types

i. Central-western (Bizkaian and Gipuzkoan); n. Zuberoan (Souletin) and Roncalese; hi. Southern High Navarrese; iv. Bidasoa Valley

Basque accentuation and dialectology

Map 4. Contrastive vs. non-contrastive accent

Phonologically contrastive accent Surface contrasts without lexical marking Position of accent is not contrastive

223

224

Jose Ignacio Hualde

Map 5. Main rule of accent assignment

[-1] Phrase final [-2] Word or phrase penultimate [+1,-1] Initial and final [+2] Postinitial [+3] Postpostinitial — Southern boundary of northern Bizkaian pitch-accent systems

Basque accentuation and dialectology

225

Map 6. Accentuation of singular and plural forms

I. Singular and plural present different accentual patterns (areas 1,2, 3, 4) 1. Zuberoan and Roncalese, e.g.: z. giziina 'the man'vs. gizunek 'the men'. 2. Gipuzkoan, e.g.: gizona, gizonai 'the man, to the man'vs. gizonak, gizonai 'the men, to the men'. 3. Gernika-Getxo (Northern Bizkaian), e.g.: gizona dä '(it) is the man', gizonari 'to the man' vs. gizonak dire '(they) are the men', gizonari 'to the men'. 4. Imotz (Navarre), e.g.: eskuu 'the hand' vs. eskuk 'the hands'. II. Accentual contrasts without singular/plural distinction (dotted area) III. Position of accent is not contrastive (discontinous horizontal lines)

226

Jose Ignacio Hualde

Map 7. Main accentual boundaries

Line A: Boundary of historically-defined western area Line B: Boundary of eastern types with contrastive accent

Basque accentuation and dialectology

227

References Allieres, Jacques 1995 Euskal Herriko atlas etnolinguistikoaz [On the ethno-linguistic atlas of the Basque Country]. In: Ricardo Gömez and Joseba A. Lakarra, (eds.), Euskal Dialektologiako Kongresua (Donostia, 1991ko irailak 2-6) [Basque Dialectology Congress (Donostia, September 2-6, 1991)], 47-58. Gipuzkoako Foru Aldundia: Donostia (Supplements of Anuario del Seminario de Filologia Vasca 'Julio de Urquijo', 28). Altube, Seber 1934 Observaciones al tratado de 'Morfologia vasca' de don R.M. de Azkue. Bermeo: Gaubeka. Repr. with Azkue Morfologia vasca, vol. 3, Bilbao: La Gran Enciclopedia Vasca, 1969. Arantzadi Zientzia Elkartea 1990 Euskalerriko Atlas Etnolinguistikoa [Ethnolinguistic Atlas of the Basque Country]. Munibe: Donostia. Azkue, Resurrecciön Μ. de 1905 Diccionario vasco-espanol-frances. Bilbao. 1923 Morfologia vasca. Bilbao: Euskaltzaindia. Repr., Bilbao: La Gran Enciclopedia Vasca, 1-3, 1969. 1931-32 Del acento tonico vasco en algunos de sus dialectos. Euskera 4: 282-318 and 6: 3-50. Basterrechea, Jose 1974-75 Estudios sobre entonacion vasca segun el habla de Guernica. Fontes Linguae Vasconum 18: 353-393 and 21: 289-338. Bonaparte, Louis-Lucien 1863 Carte des sept provinces basques montrant la delimitation actuelle de l'euscara et sa division en dialectes, sous-dialectes et varietes. London: Standford's Geographical Society. 1869 Le verbe basque en tableaux. London: Strangeways and Waiden. Repr., in Bonaparte, 1991, vol. 1. 1991 Opera Omnia Vasconice, 1-4. Euskaltzaindia: Bilbao. Elexpuru, Juan Martin 1996 Herri-euskararen eremua [The domain of local Basque dialects]. Euskera 41:523529. Elordieta, Gorka 1997 Accent, tone and intonation in Lekeitio Basque. In: Fernando Martinez-Gil and Alfonso Morales-Front (eds.), Issues in the phonology and morphology of the major Iberian languages, 1-78. Washington, DC: Georgetown Univ. Press. Etxebarria, Joxe Maria 1983 Euskal dialektologiarako testo eta ariketak [Texts and exercises on Basque dialectology]. Gasteiz: El Carmen. 1985 Gaintzako euskeraren alderdi fonetiko batzuk [Some phonetic aspects of Gaintza Basque], Eusko-Ikaskuntza. Cuadernos de seccion, Hizkuntza eta literatura, 4: 127-142.

228

Jose Ignacio Hualde

Etxebarria Ayesta, Juan Manuel 1991 Zeberioko euskararen azentuaz [On accentuation in Zeberio Basque]. In: Joseba A. Lakarra, (ed.), Memoriae L. Mitxelena magistri sacrum, 677-708. Donostia: Diputaciön Foral de Gipuzkoa. Gaminde, Inaki 1993 Bizkaiko azentu-motei buruz [On the accentual types of Bizkaia]. Uztaro 7: 111-140. 1994a Elantxobeko azentuaz [On accent in Elantxobe]. Ele 14: 7-41. 1994b Urduliz eta Gatikako azentu ereduaz [On the accentual patterns of Urduliz and Gatika]. Uztaro 11:55-110. 1994c Munitibar eta inguruko azentuaz [On accentuation in Munitibar and surrounding area]. Fontes Linguae Vasconum 65: 81-120. 1994d Bizkaieraren geografia eta bizkaieraren azentua [On the geography of Bizkaian Basque and Bizkaian accentuation], li Jardunaldiak Soziolinguistikaz, 45-56. Bilbao: Bilboko Udala. 1995a Los tipos de acento del dialecto vizcaino del euskara: aproximacion acustica. In: Ana Elejabeitia and Alexander Iribar, (eds.), Phonetica, 11-42. Bilbao: Universidad de Deusto. 1995b Zubereraren azentuaz [On accentuation in Zuberoan]. Uztaro 13: 107-128. 1995c Gipuzkeraren azentuaren azterketa akustikoa [An acoustic analysis of Gipuzkoan accentuation], Fontes Linguae Vasconum 69: 297-321. 1995d Larraungo Alii herriko azentu ereduaz [On the accentual patterns of the town of Alii in Larraun]. Fontes Linguae Vasconum 68: 23-35. 1996 Esteribarko euskararen azentuaz [On accentuation in the Basque of Esteribar], Uztaro 16: 109-123. 1998 Euskaldunen azentuak [The accents of Basque speakers], Bilbao: Labayru. Gaminde, Inaki and Jose I. Hualde 1995 Euskal azentu-ereduen atlaserako: zenbait isoglosa [Towards an atlas of Basque accentual patterns: some isoglosses]. Anuario del Seminario de Filologia Vasca •Julio de Urquijo' 29: 175-197. Gaminde, Inaki and Jasone Salaberria 1997 Ezpeleta, Lekorne eta Makeako azentu ereduez [On the accentual patterns of Ezpeleta, Lekorne and Makea], Uztaro 20: 93-103. Hualde, Jos6 I. 1989 Acentos Vizcainos. Anuario del Seminario de Filologia Vasca 'Julio de Uruijo' 23: 275-325. 1990 Euskal azentuaren inguruan [Regarding Basque Accentuation]. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 24: 699-720. 1991 a Basque phonology. London: Routledge. 1991b Manuel de Larramendi y el acento vasco. Anuario del Seminario de Fillologia Vasca 'Julio de Urquijo' 25: 737-749. 1992 Notas sobre el sistema acentual de Zeberio. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 26: 767-776. 1993a On the historical origin of Basque accentuation. Diachronica 10: 13-50. 1993b Observaciones acerca de los sistemas acentuales de la zona occidental de Gipuzkoa. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 27: 241-263.

Basque accentuation and dialectology

1995a

229

Euskal azentuen sailkapenaz eta historiaz [On the classification and history of Basque accentual patterns]. In: Ricardo Gomez and Joseba Lakarra, (eds.), Euskal dialektologiako kongresua, 189-208 (Supplements of Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 28.) Donostia: Gipuzkoako Foru Aldundia. 1995b Analisis del sistema acentual de Ondarroa. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 29: 319-343. 1995c Sobre el acento roncal6s. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 29: 499-526. 1995d Reconstructing the ancient Basque accentual system: Hypotheses and evidence. In: Jose I. Hualde, Joseba A. Lakarra and Robert L. Trask, (eds.), Towards a history of the Basque language, 171-188. Amsterdam: Benjamins. 1996 Accentuation and empty vowels in Ondarroa Basque. Lingua 99: 197-206. 1997 Euskararen azentuerak (Supplements of Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 42.) Donostia and Bilbao: Gipuzkoako Foru Aldundia and Univ. of the Basque Country. 1998 A gap filled: Postpostinitial accent in Azkoitia Basque. Linguistics 36: 99-117. 1999 Basque accentuation. In Harry van der Hülst, (ed.), Word Prosodic systems in the languages of Europe, 947-993. Berlin: Mouton de Gruyter. 2000 On system-driven sound change: Accent shift in Markina Basque". Lingua 110: 99- 129. Hualde, Jose I. and Xabier Bilbao 1992 A phonological study of the Basque dialect of Getxo". Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 26: 1-117. 1993 The prosodic system of the Basque dialect of Getxo: a metrical analysis. Linguistics 7>\: 59-85. Hualde, Jos61., Gorka Elordieta and Arantzazu Elordieta 1993 Focalizaciön y prosodia en vascuence Vizcaino. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 17: 731-749. 1994 The Basque Dialect of Lekeitio (Supplements of Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 34.) Bilbao and Donostia: Univ. of the Basque Country and Gipuzkoako Foru Aldundia. Hualde, Jose I. and Txomin Sagarzazu 1991 Acentos del Bidasoa: Hondarribia. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 25: 139-152. Ibarra, Orreaga 1995 Ultzamako hizkera: Ingurulco euskalkiekiko harremanak [the dialect of Ultzama: relationships with neighboring dialects]. [Pamplona]: Nafarroako Gobernua. Jansen, Wim 1992 Acento y entonaciön en Elorrio. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 26: 391-440. Lakarra, Joseba A. 1986 Bizkaiera zaharra euskalkien artean [Old Bizkaian among the Basque dialects]. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 20: 639-681. Michelena, Luis 1961 Fonetica historica vasca. San Sebastian: Diputaciön de Guipuzcoa.

230

Jose Ignacio

1972

Hualde

A note on Old Labourdin accentuation. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 6: 110-120. Repr. in Michelena 1987, 235-344. 1981 Lengua comun y dialectos vascos. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 15: 291-313. Repr. in Michelena 1987, 35-55. 1987 Palabrasy textos. Leioa: Universidad del Pais Vasco. Ormaetxea, Nikolas 1919 Acento vasco. Revista Internacional de Estudios Vascos 9: 1-15. Trask, Robert Lawrence 1997 The history of Basque. London: Routledge. Yrizar, Pedro de 1981 Contribuciön a la dialectologia de la lengua vasca, 2 vol. Zarautz: Gipuzkoako Aurrezki Kutxa Probintziala. 1991 Morfologia del verbo auxiliar guipuzcoano. Estudio dialectologico, 1 -2. Zarautz: Euskaltzaindia. 1992 Morfologia del verbo auxiliar Vizcaino. Estudio dialectologico, 1-2. Zarautz: Euskaltzaindia. Zuazo, Koldo 1988a Euskararen batasuna [The unification of the Basque language]. Bilbao: Euskaltzaindia and Univ. of the Basque Country. 1988b Bizkaieraren ezaugarriez gehiago [More on the features of Bizkaian]. Anuario del Seminario de Filologia Vasca 'Julio de Urquijo' 22:367-377.

On the (non-)persistence of dialect features in American Dutch (1): general aspects Jaap van Marie & Caroline Smits

1. Introduction The most frequently posed question that we are confronted with in the context of our investigation of American Dutch is whether the variety of Dutch spoken by the descendants of the 19th century immigrants shows an archaic character. It is not only laymen who are fascinated by this issue - they often ask whether American Dutch still contains "old-fashioned words" -, since linguists appear to be intrigued by this question as well. In the case of the latter, the idea of the potentially archaic character of American Dutch is often made somewhat more precise by asking whether American Dutch has retained dialect features which have already disappeared in the cognate dialects in the Netherlands. These linguists, in other words, focus on the retention of dialect features in this extraterritorial variety of Dutch, i.e., they hardly pay attention to other forces that may have played a part in the development of American Dutch. A prototypical representative of this approach to American Dutch is Mrs. Daan who has raised the issue of the retention of dialect features in American Dutch in a number of papers (cf. most recently Daan 1997a, b). What is more, one of the main motives for Mrs. Daan to formulate the American Dutch project in the mid-sixties was the idea that American Dutch would contribute to our knowledge of the Dutch dialects as they were spoken in the 19th century. American Dutch, to put it differently, was considered a language museum.1 In this paper we intend to discuss the potential occurrence of dialect features in extraterritorial language varieties in a more principled way. We will do this by highlighting the following two issues: (1) We do not claim, of course, that in extraterritorial language varieties dialect features never survive. What we do claim, however, is that the survival of dialect features is a delicate process which only occurs under very specific circumstances.

232

Jaap van Marie & Caroline Smits

(2) Many linguists who particularly stress the relevance of dialect features in language varieties such as American Dutch, appear to have an oversimplified view of the factors playing a part in the coming into existence of such extraterritorial varieties. Before discussing the above two issues we would like to broach another topic: the folk conception of "descent". When asked about their descent, many a third- or fourth-generation immigrant will answer something like: "my greatgrandfather/grandfather came from town A in the Netherlands in 18 hundred and something, when he was χ years old". This actually means that most of these informants hardly have any knowledge of their genealogy, since they do not know anything about the origin of the remaining seven greatgrandparents or the remaining three grandparents, nor do they pay any attention to the in-between generations. Particularly the respective female lines do not feature prominently in the conception of their roots. Evidently, the folk genealogy practiced by the average present-day informant of American Dutch is a little convincing guide-line for linguistic conclusions. Linguists are generally of the opinion that it is particularly the language of the mothers which is most influential in the early years of language acquisition, whereas later in life it is particularly the peer group which is decisive. That is, what is most important in folk-genealogy is linguistically least important. Evidently, the more heterogeneous a settlement is, the more devastating it is to exclusively concentrate on the male line.

2. On the survival of dialect features in an extraterritorial context In discussing the fate of dialect features in an extraterritorial context, we should first of all distinguish between the survival of dialects in toto and the survival of certain properties characteristic of a specific dialect in a generally non-dialectal language variety. In Van Marie (1997) it is stressed that in extraterritorial contexts the survival of a dialect in toto is not impossible, but it presupposes a setting that is linguistically highly homogeneous, meaning that all speakers in question must originate from the same dialect area. However, in case there is a heterogeneous linguistic setting - i.e., a situation in which different dialects are spoken -, accommodation takes place (Trudgill 1986). Put differently, in situations in which different dialects are spoken, speakers tend to accommodate their language to that of others. Eventually, this may result in the structural loss of the characteristic properties of the

On the (non-)persistence of dialect features in American Dutch

233

dialects involved. In sum, in heterogeneous settings dialect levelling will occur (cf. also Van Marie & Smits 1996). The effects of dialect levelling are especially noticeable in the variety of American Dutch spoken in Iowa, that is in Pella, its daughter colony Orange City, and the areas surrounding these two settlements. From historical records we know that the early immigrants in Iowa, all of Protestant religion, came from different parts of the Netherlands, though generally neither from the eastern provinces (in general, immigrants from the eastern parts of the Netherlands went to Michigan), nor from the southern provinces (as far as the Roman Catholics from these parts migrated at all, they went to Wisconsin). We also know that immigrants originating from the western parts of the Netherlands were in the majority (specifically, the Provinces of Gelderland, South Holland, and Utrecht). In Pella, immigrants from specific regions in the old country originally lived together in so-called buurten (neighborhoods). There was, for example, a Friese buurt (inhabited by people from the Province of Friesland), a Herwijnse buurt (inhabited by people originating from the Province of Gelderland), and a Groninger buurt (inhabited by people from the province of Groningen). In the course of time, for instance through "intermarriage", the regional character of the neighborhoods watered down. It has repeatedly been pointed out that the speakers of the mutually rather divergent dialects had great difficulties in understanding each other (e.g. Webber 1988). In sum, the linguistic situation in Iowa was a typical example of linguistic heterogeneity. As a result of accommodation, dialect levelling took place within the Iowa settlements. To appreciate the scale of linguistic heterogeneity, one should keep in mind that in the 19th century standard Dutch hardly functioned as a spoken language. That is, since standard Dutch was hardly spoken at the time of immigration, most immigrants cannot have been but dialect speakers. Surprisingly, the present-day speakers of the variety of American Dutch spoken in Iowa speak a variety of Dutch which is hardly dialectal (cf. also Daan 1987). That is, in many respects it is much more similar to standard Dutch than to any of the Dutch dialects. In Van Marie & Smits (1996) it was pointed out that this change was due to the joint operation of dialect levelling (as a result of accommodation) and direct influence of standard Dutch, particularly through religion (cf. also Smits 1996).

234

Jaap van Marie & Caroline Smits

3. The complexity of extraterritorial language situations Linguists who focus on the importance of dialect features in extraterritorial language varieties appear to take the stand that the deviations in these varieties vis-ä-vis the cognate mother language spoken in the country of origin are generally the result of the following two forces: (1) The survival of dialect features originating from the dialect that the extraterritorial variety is rooted in, and which may have disappeared in the original dialect spoken in the old country. (2) Influence of the socially dominant language of the country in which the extraterritorial variety has come into being. In the case of American Dutch, this is English, of course. Not surprisingly, students of extraterritorial language varieties who stress the role of dialect features in explaining the differences between the variety at issue and its cognate mother language, lay emphasis on force (1). As a matter of fact, they only invoke force (2) if force (1) cannot be made use of at all. As we have pointed out in earlier publications, in our view the number of forces which may become active in extraterritorial language situations is much higher. Specifically, we have stressed the following forces. (1) Given the fact that in many cases the language situation in the country of immigration is heterogeneous, accommodation and dialect levelling are important forces leading to the elimination of dialect features. Moreover, it is not uncommon that dialect levelling results in a language variety that exhibits more resemblance to the standard language than to any of its dialects. In the case of American Dutch (especially in Iowa), the contact between the mutually rather divergent dialects occurred in constant interaction with the (prestigious) standard language, of which the informants already had a passive knowledge through religion, and which became more and more prominent as a spoken language. This caused the various dialects to develop into the direction of standard Dutch. (2) Forces related to language contact, i.e., interference from the socially dominant language of the country of immigration. In the case of American Dutch, this is English. (3) The normal forces of language change which have the structural imbalances of the language as their starting-point, i.e., language-internal changes also occurring in monolingual situations. Cases of language

On the (non-)persistence of dialect features in American Dutch

235

contact, particularly when it is combined with language attrition (cf. 4 below), may create favorable circumstances for such language-internal changes to occur. (4) Forces relating to language attrition, varying from stylistic reduction within the extraterritorial variety to large-scale loss of grammatical rules. Another manifestation of language attrition is the loss of language norms which may lead to large-scale and non-systematic variation. In addition, the tendency to systematize the arisen variation is marginal at best. In Weinreich's words, the trend to "crystallize out" is little prominent (Weinreich 1953: 69-70). Evidently, the potential strength of all these forces is highly variable (whereas, moreover, these forces may also interact). A given language situation may be more or less heterogeneous, meaning that the effects of accommodation may vary considerably. Naturally, this implies that the effects of dialect levelling in an extraterritorial variety such as American Dutch need not be identical in all settlements either. The influence of the socially dominant language is variable as well. Particularly in situations in which there is a question of wide-spread bilingualism - and in which both languages are well-mastered -, the effects of the socially dominant language on the language of the immigrants may be relatively limited. Whether the normal forces of language change may become operative, is also variable. This largely depends on the extent to which the original norms are still in force (cf. 4). The effects of language attrition, finally, are variable, too. Whether the language of the immigrants remains in use as an in-group language or not, is an issue which is completely dependent on the socio-cultural environment in which the immigrant language functions. In the former case, the language may be maintained by many generations to come (either as a first or as a second language), and the effects of attrition may be relatively limited. In the latter case, the language may disappear within a few decades, and attrition may be dramatic.

4. A case study Daan (1971) is a classical example of a study which embraces the languagemuseum concept of American Dutch. On the basis of the language of four

236

Jaap van Marie & Caroline Smits

speakers of American Dutch, she claims that a shift has taken place in the position of an isogloss in the eastern parts of the Netherlands. The isogloss in question relates to verb inflection. To the west of it, 1st, 2nd and 3rd person present-tense plurals end in -n, to the east of this isogloss they end in -t. At present, this isogloss runs to the east of the towns of Nunspeet, Elspeet, Uddel, and Doornspijk. The ancestors of the four speakers discussed by Daan (1971) originated from the towns just mentioned. It concerns speakers 118, 120, 134, and 135 in the American Dutch corpus collected by Daan and Heikens in 1966; the ancestors of these informants went to the United States between 1865 and 1904. Note, by the way, that these informants are no prototypical representatives of American Dutch, neither in Daan's view, nor in ours. As is explicitly stated by Daan, these informants had lived in greater isolation than the average informant in her corpus; therefore, their language was more dialectal than that of most of the other informants, whose language appeared to be far more levelled out.2 Interestingly, these speakers use present-tense plural verb forms ending in -t. From this, Daan concludes that at the time the ancestors of these informants went to the United States, i.e., at the end of the 19th century, the isogloss in question was more to the west, i.e., to the west of Nunspeet, Elspeet, Uddel, and Doornspijk. Crucial to Daan's reasoning is, that she does not doubt that the American Dutch verb forms in -t must be dialectal features that have been preserved in the language of the four speakers in question, and in this way Daan (1971) is an illuminating example of the conception according to which extraterritorial language varieties are considered to be dialect museums. In the remainder of this paper we will show that the actual status of these -t forms in the language of the four informants in question - and the status of potential dialect features in American Dutch in general - is much more complex. Note, by the way, that at this moment we have no external evidence that present-tense plural forms in -t were indeed present at the end of the 19th century in the Dutch region at issue. In fact, the evidence rather seems to point in the opposite direction. In 1904, Van Schothorst wrote a monograph on the dialect of the "Noord West Veluwe", which is exactly the region which is at stake here. From that study we learn that at the turn of the century, the present-tense plural on the Noord West Veluwe ends in -en, meaning that no attestations of present tense plurals in -t were given for any of the Dutch towns in question. In relation to our claim that the status of potential dialect features in American Dutch is very complex, the following points are of interest.

On the (non-)persistence of dialect features in American Dutch

237

(1) Inflectional forms which may, in principle, be considered to have a dialectal origin are generally not used in a systematic way. This, of course, directly relates to the striking degree of variation that American Dutch exhibits. This is even the case with the language of the speakers that Daan's paper is based upon. In the language of informant 134, for instance, there are relatively few present-tense plural forms, which implies that there are not many present-tense plural forms in -t either. The presenttense plural forms in -t that are used, however, are nearly always forms of the highly irregular verb zijn 'to be'. Consider the form bin-t 'are', in a sentence such as de meeste bint Rooms 'most of them are Roman Catholic'. Importantly, the informant in question also uses present-tense plural forms in -e, even in case of the verb zijn. Consider the form binn-e 'are' in zie binne Rooms 'they are Roman Catholic'. Similarly, informant 135 alternately uses present-tense plural forms in -t and -en. Compare verstao-t in wie verstaot dat niet 'we do not understand that' with zegg-en in wiezeggen ook 'skapen' 'we also say 'skapen' (sheep)'. (2) Inflectional forms which may, in principle, be considered dialect features also occur in the language of informants who originate from other regions in the Netherlands, i.e., regions in which such inflectional forms do not occur in the dialects. The occurrence of present-tense plural forms in -t in American Dutch, for instance, is by no means restricted to speakers whose predecessors came from the eastern provinces. That is, forms as wij loop -t (instead of wij lopen 'we walk'), wij herinner-t (instead of wij herinneren 'we remember') and de hinderen slaap-t (instead of de kinderen slapen 'the children sleep') can also be found in the language of present-day speakers of American Dutch whose ancestors came from parts of the Netherlands where the present-tense plural form exclusively ends in -en. (3) Interestingly, in American Dutch "unconventional" present tense forms in -t are not restricted to contexts in which a plural in -en is required. Different from standard Dutch, the present tense 1st person singular may also end in -t. Consider, for instance, forms such as ikgaa-t and ikkom-t (which correspond to respectively ik ga Ί go' and ik kom Ί come' in standard Dutch). Again, such 1st person singular forms can be found in the language of speakers of American Dutch with completely different dialect backgrounds. Put differently, at least for a number of speakers of American Dutch there seems to be no question of the clinging to a dialect feature. This means that there is a general trend in American Dutch to generalize 3rd person singular forms through the paradigm, i.e. to both 1st person singular and plural (cf. also 6 below).

238

Jaap van Marie & Caroline Smits

(4) It is not only the case that forms which may be considered dialectal (such as present-tense plural forms in -t) occur next to rival forms lacking such a potential dialectal origin (cf. 1 above). Even more common is the situation that forms which are "to be expected" in terms of dialect background are completely lacking. Consider, again, the language of the speakers studied by Daan (1971). Obviously, Daan's claim is based on the premise that these speakers of American Dutch have preserved the dialect of the country of origin, meaning that their language has not undergone radical changes. In fact, Daan explicitly states that the language of the informants in question does not deviate from the cognate dialect of the region of origin in the old country as it is described in dialect studies dating from before 1940 (Daan 1971: 77).3 However, if this is indeed the case, we would expect other dialectal forms to be present in the language of these informants as well. As far as inflection is concerned, an interesting case in point concerns the ending -enl-n ("syllabic η") for the 2nd person singular present tense. In Van Schothorst (1904) we find that the 2nd person singular ends in -enl-n, and not in -t as in standard Dutch. Let us consider, once more, informants 134 and 135, two of the informants that Daan (1971) is based upon. In the language of these two informants no single enl-n ending for 2nd person singular can be found. What we do find, however, are 2nd person singular forms in -t (such as draai-t 'turn', kun-t 'can', ken-t 'know', leer-t 'learn', gao-t 'go'), i.e., 2nd person singular forms that are identical to those of the standard language.4 (5) As we have stressed before, the complete or near absence of salient dialect features seems to be an important characteristic of American Dutch. This is not only true for the morphological level (cf. above). Within an approach that claims that particular inflectional forms of a specific dialect of the country of origin have been preserved in the language of later generations of immigrants, it is not immediately clear why other linguistic domains - particularly sounds and syntax - have been levelled out. As a matter of fact, this completely runs counter to the generally accepted idea that it is precisely sounds and certain syntactic patterns that are retained longest. Evidence in favor of the latter is provided by regional varieties of the standard language spoken on the continent. These regional varieties deviate from the supra-regional variety of the standard language precisely as far as sounds and syntax are concerned, not as to their inflection.

On the (non-)persistence of dialect features in American Dutch

239

(6) In the above, we put into perspective the claim that deviations from standard Dutch can simply be attributed to the Dutch dialects underlying American Dutch. In this light, it is also important to realize that most of the deviations attested in American Dutch nicely fit in with our present knowledge on processes becoming active in case of language contact and language attrition. This is also true for features of American Dutch that are claimed to be of dialectal origin by some. Consider, once again, the generalization of 3rd person singular present tense -t forms to the 1st person singular and to the plural (cf. 3 above). From the perspective of language contact/language attrition, this involves the generalization of the present-tense form that is unmarked (and most frequently used) from a pragmatic/communicative point of view. Consequently, it is the form that stands the best chances for survival, i.e., it is the form that will be generalized at the cost of the marked 1st person singular and plural forms. Moreover, we know that the 3rd person singular form survives in other situations of language contact as well; cf., e.g., the survival of the 3rd person singular in "Ontarian French" (Mougeon & Beniak 1981). In addition, we know that the generalization of 3rd person singular forms occurs in other "reduced" varieties of Dutch as well, notably in child language. For instance, at the age of two/two and a half, our son almost consistently used stem + -t forms throughout the present-tense paradigm. (7) Finally, it should be stressed that there are many features of American Dutch that do not occur in any of the Dutch dialects that we know of, meaning that they cannot be considered of dialectal origin whatsoever. Very interesting in relation to our discussion of the status of present tense forms in -t in American Dutch is the fact that we also come across the generalization of the 3rd person singular suppletive form is (of the verb zijn 'to be'). As far as our knowledge goes, there are no dialects in the Netherlands that use is as a present-tense plural form, meaning that the occurrence of is in the plural in American Dutch cannot be of dialectal origin. There are many other features that cannot have originated from the dialects underlying American Dutch. We will only mention some of the most striking examples (for a detailed discussion of these features, cf. Smits 1996): - the generalization of singular nouns at the cost of plural nouns; - the generalization of the declined form of the attributive adjective at the cost of its undeclined counterpart; - the generalization of preterite verb forms at the cost of their present tense counterparts;

240

Jaap van Marie & Caroline Smits

- the generalization of past participles instead of tensed verb forms - the generalization of irregular preterite plural forms at the cost of their singular counterparts. Evidently, we take the line that one should be very careful when claiming that a variety of American Dutch is identical to the cognate variety spoken in the Netherlands, as Daan does, since in this approach the effects of dialect levelling are easily glossed over. In addition, on the basis of the above the conclusion cannot but be that the American Dutch forms in -t that were focused on in Daan (1971) need not have a dialectal origin at all.

5. Conclusion In the above we tried to make clear that one should be very careful to attribute certain features in the immigrant language to the dialect background of the speakers involved. This is all the more true for heterogeneous settlements, where accommodation and dialect levelling are prominent. Dialect features will generally not survive, unless they are supported by most dialects involved in the contact situation (cf. Smits, this volume). By implication, this means that one should hesitate even more to make claims relating to the orginal dialects in the country of origin on the basis of properties of the immigrant language.

Notes 1, Note, however, that Daan's conception of the factors that have played a crucial role in the coming into existence of American Dutch is not always clear, and neither is the way in which she motivated the 1966 project. The fact is, that in Daan (1987) all kinds of remarks can be found which imply that the occurrence of dialect features in American Dutch is little prominent. Cf. also Daan's remark in the Pella Chronicle of October 18, 1966: "We expected to find here in isolation, an older form of the dialects than could be found in the Netherlands (...). But this has not been true, because most people have learned Dutch through association with other dialects. The result has been very near the standard Dutch language, but more mixed up than in the Netherlands". 2. In contrast to Daan's claim, not all four informants are third-generation immigrants. The two immigrants mentioned first (118 and 120) are siblings, and their parents (father as an adult, mother at the age of 15) immigrated to the USA, which makes them secondgeneration immigrants.

On the (non-)persistence of dialect features in American Dutch

241

3. Note, however, that Daan's claim is not in conformity with the classic description of this dialect, i.e., Schothorst (1904). 4. One of many other examples relating to the fact that many dialect forms are completely lacking concerns, again, the 2nd person singular ending of the present tense. Apart from dialects with a 2nd person singular ending -enl-n, there are also eastern Dutch dialects having a 2nd person singular ending -s or -st (Van den Berg 1949). Irrespective of the fact that in the American Dutch material that we have at our disposal there are quite some informants whose ancestors came from dialect areas with this -s or -st ending for 2nd person singular, 2nd person singular verb forms in -s or -st are nearly (if not completely) absent. Again, then, a salient dialect feature has disappeared in the levelled variety spoken in the New World.

References Berg, Berend van den 1949 "De conjugatie van het praesens in de Noordnederlandse dialecten", Taal & Tongval 1:6-13. Daan, Jo 1971 "Verschuiven van isoglossen", Taal & Tongval 23: 77-79. 1987 Ik was te bissie. Nederlanders en hun taal in de Verenigde Staten. Zutphen: Walburg pers. 1989 "Review of Philip E. Webber, Pella Dutch. The portrait of a language and its use in one of Iowa's ethnic communities", Taal & Tongval 41:179-180. 1990 "Code-Wechsel unter psychosozialem Druck", Zentrum fiir Niederlande Studien, Jahrbuch 1: 87-96. 1997a "Review of Caroline Smits, Disintegration of inflection: the case of Iowa Dutch", Nederlandse Taalkunde 2 (1): 70-75. 1997b "Dutch in the United States of America", in: J. Klatter-Folmer & S. Kroon (eds), Dutch overseas. Studies in maintenance and loss of Dutch as an immigrant language. Tilburg: University Press, 237-253. (Studies in Multilingualism 9). Marie, Jaap van 1997 "Waarom er geen Amerikaans Zeeuws bestaat", Nehalennia 114: 47-56. Marie, Jaap van & Caroline Smits 1996 "American Dutch: general trends in its development", in: P. Sture Ureland (ed.), Language contact across the North Atlantic. Tübingen: Niemeyer, 427- 442. Mougeon, R. & E. Beniak 1981 "Leveling of the 3sg/pl verb distinctions in Ontarian French", in: J.P. Lantolf & G.B. Stone (eds.), Current research in Romance languages. Bloomington: Indiana University Linguistic Club, 126-144. Schothorst, W. van 1904 Het dialect der Noord- West- Veluwe. Utrecht: Kemink. Smits, Caroline 1996 Disintegration of inflection. The case of Iowa Dutch. Dissertation Free University Amsterdam. (HIL-dissertations 22) Trudgill, Peter 1986 Dialects in Contact. Oxford: Blackwell.

242

Jaap van Marie & Caroline Smits

Webber, Philip.E. 1988 Pella Dutch. The portrait of a language and its use in one of Iowa's ethnic communities. Ames: Iowa State University Press. Weinreich, Uriel 1953 Languages in contact. Findings and problems. The Hague: Mouton.

On the (non-)persistence of dialect features in American Dutch (2): the case of Iowa Dutch1 Caroline Smits

1. Introduction Smits (1996) discusses the inflectional deviations from standard Dutch in two corpora of Iowa Dutch, the first corpus gathered in 1966 and the second in 1989. Although she acknowledges that dialect features are present in Iowa Dutch inflection, Smits (1996) claims that the effects of language contact and language attrition - and the interaction between these processes - account for most of the deviations. Some linguists have raised the question whether this is justified (cf. most recently Daan 1997; Van Reenen p.c.). They state that a number of inflectional deviations in Iowa Dutch that Smits (1996) attributes to the effects of language contact and attrition are actually dialect features that were already present in the varieties of Dutch transplanted to Iowa from the second half of the 19th century onwards. In this paper, I will discuss further data supporting my claim that most of the inflectional deviations should be attributed to the effects of language contact and language attrition, indeed. Focus is on data from Iowa Dutch 1966, since it is particularly in relation to this corpus that the effects of language contact and attrition are questioned by some critics. The fact is that the language spoken in 1966 is considered most closely connected to continental Dutch, meaning that dialect features are expected to occur most frequently in this corpus. The data supporting my claim relate to the distributional patterns of deviations in the three subcorpora that can be distinguished in the 1966 corpus. It will become clear that the deviations that, according to Smits (1996), are not of dialectal origin are differently distributed over the three subcorpora than the deviations that, according to Smits (1996), are of dialectal origin indeed. In the sections below, the following topics will come up for discussion: general characteristics of Iowa Dutch (section 2); the three subcorpora that can be distinguished in the Iowa Dutch corpus gathered in 1966 (section 3); inflectional forms in Iowa Dutch 1966 and the distribution

244

Caroline Smits

of deviations from the standard language over the three subcorpora (section 4); a further discussion of these data (section 5); conclusion (section 6).

2. General characteristics of Iowa Dutch Iowa Dutch is the language spoken by Dutch Protestant immigrants - and their descendants - who settled on the prairie in Iowa (mainly in Pella and its daughter colony Orange City and the areas surrounding these places) during the second half of the 19th and the beginning of the 20th century. For many of its speakers Dutch was the language of communication, at least within the family, until about 1950. It was particularly during and after the second World War that the number of domains in which Dutch was used rapidly decreased and a definite shift from Dutch to English took place. At the moment, English is the language commonly used, and the linguistically dominant language (cf. Van Coetsem 1988, 1995 on linguistic dominance), while the use of Dutch is highly restricted. However, even nowadays there are quite a number of people who master at least some Dutch, albeit sometimes rather imperfectly. Since the children of the present-day speakers of Iowa Dutch have not acquired the language, Iowa Dutch is on the verge of extinction. As said, Smits (1996) claims that Iowa Dutch is generally not strongly colored by dialect features, notwithstanding the fact that the Dutch immigrants who went to Iowa in the 19th century were mainly speakers of dialects and not the standard language. Instead, it resembles standard Dutch more closely than any of the Dutch dialects spoken on the continent (cf. also Daan 1987: e.g. 103). Particularly on the grammatical level dialect features are largely absent.2 At the same time, however, the language is certainly not identical with standard Dutch. Put differently, Iowa Dutch bears more resemblance to standard Dutch than to any of its dialects, although it obviously deviates from the standard language too. In this light, the following two stages in the development of Iowa Dutch should be distinguished (cf. also Van Marie & Smits 1996; Smits 1996): (i) Stage I: convergence Convergence towards standard Dutch particularly occurred during the period in which Dutch was still regularly used as a means for communication (and was still the linguistically dominant language), meaning that it primarily occurred in the second half of the 19th and the first decades of the 20th century. Convergence involves both a) dialect levelling between the mutu-

On the (non-)persistence of dialect features in American Dutch (2)

245

ally rather divergent dialects that came into contact in the Iowa settlements, and b) the influence of standard Dutch, particularly through religion. The process of dialect levelling, then, occurred in constant interaction with the standard language, which caused the various dialects in the Iowa settlements to develop in the direction of standard Dutch.3 Note that this process of convergence may have been enhanced by the fact that the majority of the immigrants who settled in Iowa originated from the western part of the Netherlands. That is, they came from the part of the Netherlands in which dialect varieties are spoken which are relatively close to standard Dutch. Clearly, the prominence of these western - though mutually rather divergent - dialects may also have enhanced convergence with the standard language. (ii) Stage II: divergence Divergence from standard Dutch especially took place from the time onwards that English gained in importance and became the commonly used and, subsequently, the linguistically dominant - language. The process of divergence became especially prominent after World War II, and it involves both a) the effects of language contact, i.e., interference from English, and b) the effects of language attrition, i.e., internally-induced changes and reduction. Divergence, then, is the combined effect of language contact and language attrition.

3. Iowa Dutch 1966 In the 1966 corpus, convergence towards standard Dutch has largely been completed, meaning that this corpus bears clear marks of stage I (convergence). However, it also shows the first signs of stage II (divergence). Smits (1996) claims that Iowa Dutch as it was gathered in 1966 approximates standard Dutch quite closely, and that dialect features have largely been levelled out, whereas the effects of language contact and language attrition are only just becoming noticeable. The 1966 corpus was collected by Jo Daan and Henk Heikens, and it consists of free conversations. The participants in the conversations are Jo Daan, who speaks the standard language, and usually two or more informants. Often, informants are urged to use a dialect instead of standard Dutch, one of the goals of the 1966 enterprise being the collection of older - i.e., 19th century - varieties of Dutch dialects in these isolated settlements. How-

246

Caroline Smits

ever, generally this did not produce a noteworthy change in the language used by these informants.4 As can be inferred from the interviews made in 1966, at that time the use of Dutch was largely restricted to the domain of the family. It was mainly used in communication with parents and (particularly) grandparents, and often on an incidental basis only. The informants had generally not regularly used the language for at least one and a half decade. English, on the other hand, was the language commonly used. Undoubtedly, the latter language was socially dominant, and, most likely, linguistically dominant as well. The 1966 corpus can be divided into three subcorpora on the basis of the circumstances in which the informants acquired Dutch (Smits 1996: 20-23). These subcorpora are of equal size as far as the amount of word tokens is concerned (20,000 for each subcorpus): (i) Group 1 consists of 10 informants who were born and raised in the Netherlands, which means that they acquired Dutch in an exclusively Dutch context. (ii) Group 2 consists of 10 informants whose parents were born in the Netherlands.5 These informants acquired Dutch in a primarily Dutch context, i.e., in the United States, but from parents who acquired their Dutch in the Netherlands. (iii) Group 3 consists of 14 informants who had grandparents born in the Netherlands. These informants acquired Dutch in a primarily English context, i.e., in the United States, and from parents who had spent a considerable period of time in the United States as well. In comparing the informants in the three subcorpora distinguished above, group 1 informants acquired Dutch in the most favorable circumstances, group 2 informants acquired Dutch in less favorable circumstances, and group 3 informants acquired Dutch in the least favorable circumstances. Given these differences in acquisitional history, the language of the informants in group 1 is expected to be most closely connected to continental Dutch, whereas the language of the informants in group 3 is expected to be most dramatically affected by the effects of language contact and attrition.

4. Inflectional forms in Iowa Dutch 1966; the data In general, there is an overwhelming majority of standard Dutch inflectional forms in all three subcorpora and in all inflectional categories, meaning that the effects of language contact and attrition are only limited. Given the large

On the (non-)persistence of dialect features in American Dutch (2)

247

number of standard Dutch forms it can also be deduced that inflectional forms typical to dialectal varieties of Dutch are marginal at the most. Cf. table 1 below, taken from Smits (1996: 156). The categories in which the amounts of standard Dutch forms have dropped below 90% in at least one of the three subcorpora are printed in bold: 6 Table 1.

Standard Dutch inflectional forms in the three subcorpora of Iowa Dutch 1966 Group 1

Nominal plurals Regular plurals: monosyll. base polysyll. base + final stress polysyll. base - final stress Irregular plurals: monosyll. base + -eren Attributive adjectives adjectives in -en/en sg. definite het nouns sg. indefinite het nouns sg. de nouns/pl. nouns Present tense verbs regular verbs infinitive in -n irregular zijn irregular hebben Preterite verbs regular verbs irregular main verbs irregular zijn irregular hebben irregular modal verbs

Group 2

Group 3

Total

SD

Total

SD

Total

SD

(n)

(%)

(n)

(%)

(n)

(%)

173 18 75

96.5 94.4 97.3

192 19 62

97.4 100 98.4

168 18 48

100 100 100

58

60.3

34

55.9

39

84.6

25 23 33 177

92.0

14 20 40 192

100

11 15 40 202

100

73.9

97.0 97.7

416 70 528 227

91.6 95.7 86.6 75.3

447 101 488 233

113 323 329 144 169

92.0 95.0 97.0 96.5 97.0

98 344 328 115 188

(SD = standard Dutch; sg. = singular; pi. = plural.)

85.0

97.5 94.3 90.8 91.1 80.9 75.1 81.6

94.8 94.5 96.5 96.3

427 86 442 235 72 304 380 91 164

80.0 65.0

96.0 94.6 72.1 84.6 70.6 83.3

93.4 93.7 72.5 89.0

248

Caroline Smits

In the remainder of this paper, focus is on those categories in which the percentage of standard Dutch forms has dropped more or less dramatically, i.e., below 90% (cf. the figures printed in bold in Table 1). In the remaining categories, the percentage of standard Dutch forms is too high to distinguish any significant differences in terms of the distribution of deviations over the three subcorpora. The categories that will come up for discussion are also given below: (For an outline of inflection in continental Dutch, the reader is referred to the Appendix.) Nominal plurals: (i) monosyllabic nouns with an irregular plural in -eren in standard Dutch Attributive adjectives: (ii) adjectives modifying singular definite neutral (or het) nouns (iii) adjectives modifying singular indefinite neutral (or het) nouns Present tense verb forms: (iv) verbs with an infinitive in -n (v) the irregular verb zijn 'to be' (vi) the irregular verb hebben 'to have' Preterite verb forms: (vii) verbs with a regular preterite in standard Dutch (viii) the irregular verb hebben (ix) irregular modal verbs. In relation to each of these categories, the following two questions will be answered. First, which of the inflectional deviations from standard Dutch can be considered dialect features - i.e., features that were already present in the varieties of Dutch transplanted to Iowa that have survived the process of levelling -, and which of the inflectional deviations are the result of the (combined) effects of language contact and language attrition? Put differently, which deviations are of dialectal origin, and which deviations are not of dialectal origin? In answering this question, we will follow the classification made in Smits (1996). (For a detailed discussion of these deviation types, the reader is referred to the latter study.) Second, how are both the deviations that are probably of dialectal origin and the deviations that are most likely not of dialectal origin distributed over the three subcorpora? In order to answer this question the distributional patterns will be presented in tables drawn from Smits (1996).

On the (non-)persistence of dialect features in American Dutch (2)

249

(i) Monosyllabic nouns with an irregular plural in -eren in standard Dutch The deviations from standard Dutch in Iowa Dutch 1966 are the following: a) Plurals in -erl-ers, e.g. kind-ers (instead of standard Dutch kind-eren 'children'). These are unmistakably plurals of dialectal origin (cf. note 5)·7 b) Plurals in -en, e.g. ei-en (instead of standard Dutch ei-eren 'eggs'). These plural are not of dialectal origin. They are the result of the generalization of the regular standard Dutch plural -en at the cost of irregular -eren, which involves an internally-induced change, i.e., it is based on the standard Dutch system for pluralization of nouns (Smits 1996: 77). Consider the distribution of plurals in -erl-ers and plurals in -en in tables 2a and 2b. Table 2a. Plurals in -erl-ers in Iowa Dutch 1966 among nouns with an irregular plural in -eren in standard Dutch

group 1: group 2: group 3:

Total

-erl-ers (+dialect)

(n)

(%)

58 34 39

39.7 32.4 12.8

Table 2b. Plurals in -en in Iowa Dutch 1966 among nouns with an irregular plural in -eren in standard Dutch

group 1: group 2: group 3:

Total

-en (-dialect)

(n)

(%)

58 34 39

-

11.8 2.6

The dialectal plurals in -erl-ers (table 2a) are most prominent in group 1. The plurals in -en (table 2b) are absent in group 1, most prominent in group 2, and highly infrequent in group 3.

250

Caroline Smits

(ii) Adjectives modifying singular definite neutral (het) nouns The deviation from standard Dutch consists of the undeclined form of the adjective being used instead of the standard Dutch declined form, e.g., dat droog weer (instead of standard Dutch dat drog-e weer 'that dry weather'). This deviation is not of dialectal origin. It represents an internally-induced change, based on the standard Dutch system for adjectival declension itself. In standard Dutch the use of undeclined adjectives is restricted to adjectives modifying singular indefinite het nouns. In Iowa Dutch, however, the use of undeclined adjectives is extended to the context of singular definite het nouns. Note that this tendency also occurs in continental Dutch. In addition, the generalization of the undeclined adjective may have been enhanced by interference from English, in which attributive adjectives are never declined (Smits 1996: 80-81). Consider table 3 for the distribution of undeclined adjectives in contexts in which standard Dutch has declined adjectives: Table 3.

group 1: group 2: group 3:

Undeclined forms in Iowa Dutch 1966 among adjectives modifying singular definite neutral nouns, declined in standard Dutch Total (n)

Undeclined (-dialect)

23 20 15

26.1 15.0 20.0

(%)

The use of undeclined adjectives in the context of singular definite het nouns is most frequent in group 1. (iii) Adjectives modifying singular indefinite neutral (het) nouns The deviation involves the fact that the declined form of the adjective is used instead of the standard Dutch undeclined form, e.g. een Hollands-e versje (instead of standard Dutch een Hollands versje 'a Dutch rhyme'). Again, this deviation is not of dialectal origin. It involves an internally induced change: in standard Dutch attributive adjectives are normally declined in -e, with the exception of the adjectives modifying singular indefinite het nouns. In Iowa Dutch this exception is eliminated by generalizing the declined form of the adjective to the context of singular indefinite neutral nouns (Smits 1996: 8182).

On the (non-)persistence of dialect features in American Dutch (2)

251

Consider table 4 for the distribution of declined adjectives in contexts in which standard Dutch has the undeclined form of the adjective: Table 4.

group 1: group 2: group 3:

Declined forms in Iowa Dutch 1966 among adjectives modifying singular indefinite neutral nouns, undeclined in standard Dutch

Total (n)

Declined in -e (-dialect)

33 40 40

3.0 2.5 35.0

(%)

The generalization of declined adjectives in the context of singular indefinite het nouns is far more prominent in group 3 than in groups 1 and 2. The difference between groups 1 and 2 on the one hand, and group 3 on the other is significant (Chi-square = 19.552, ρ < .001, η = 113). (iv) Present-tense forms of verbs with an infinitive in -n The most prominent deviations are the following: a) Generalization of 3rd singular stem + -t forms to the 1st singular, e.g. ik doe-t (standard Dutch ik doe Ί do'; cf. standard Dutch 3rd singular hij doe-t 'he does'). This generalization is an internally-induced attempt to adapt the 1st singular of verbs with an infinitive in -n to the canonical shape of 1st singular forms in standard Dutch (i.e., of verbs with an infinitive in -en; cf. Appendix). Unlike the 1st singular of -n verbs (which is vowel-final), the 1st singular of -en verbs is consonant-final (cf., e.g., ik zwem Ί swim') (Smits 1996: 85-86). b) Addition of -en [a] to plural forms, e.g. we doen-e (standard Dutch we doen 'we do'). The use of plural forms in -en is also considered an internally induced attempt to adapt plural forms of -n verbs to the canonical shape of plural forms of -en verbs in standard Dutch, which are polysyllabic and end in -en [9] (cf., e.g., we zwemm-en 'we swim') (Smits 1996: 85).8 c) The use of stem forms in the 1st plural, e.g., we ga (standard Dutch we gaa-n 'we go'). This is considered an externally induced change, i.e., it should be attributed to interference from English, where stem forms are

252

Caroline Smits

used throughout the present-tense paradigm (with the exception of the 3rd singular) (Smits 1996: 86-87). This means that none of the above discussed deviations attested among verbs with an infinitive in -n is considered to be of dialectal origin. The distribution of these three types of deviation over the three subcorpora of Iowa Dutch 1966 is given in tables 5a-5c: Table 5a. Present tense; -n verbs: stem + -t forms in the 1 st singular in Iowa Dutch 1966 instead of standard Dutch stems

group 1: group 2: group 3:

Total

Stem + -t (-dialect)

(n)

(%)

4 8 10

-

25 70

Table 5b. Present tense; -n verbs: -en forms in the plural in Iowa Dutch 1966 instead of standard Dutch stem + -n

group 1: group 2: group 3:

Total

-en forms (-dialect)

(n)

(%)

31 32 36

3.2 3.1 41.7

On the (non-)persistence of dialect features in American Dutch (2)

253

Table 5c. Present tense; -n verbs: stems in the 1st plural in Iowa Dutch 1966 instead of standard Dutch stem + -n

group 1: group 2: group 3:

Total (n)

Stems (-dialect)

12 12 16

8.3 25.0 6.3

(%)

The generalization of stem + -t in the 1st singular (table 5a) is absent in group 1, and becomes particularly prominent in group 3. The use of plural + -en in the plural (table 5b) is highly marginal in groups 1 and 2, and only becomes prominent in group 3; the difference between groups 1 and 2 on the one hand, and group 3 on the other is significant (Chi-square = 21.235, ρ < .001, η = 99). Stem forms in the plural (table 5c) are particularly prominent in group 2. (v) Present tense forms of the irregular verb zijn 'to be' The most prominent deviations are the following: a) The use of benne in the plural, e.g., we benne (standard Dutch we zijn 'we are'). The plural form benne is formed on the basis of the (irregular) singular form ben. This analogically formed plural also occurs in various dialects of Dutch (cf. Appendix), which means that we may be dealing with a dialect feature here (Smits 1996: 71, 128-129). b) The generalization of the 1 st singular form ben to the 2nd singular, e.g. Je ben in Hospers geboren 'you (singular) were born in Hospers' (standard Dutch je bent 'you are'). The use of ben in the 2nd singular is also known in dialectal varieties of Dutch, meaning that we may be dealing with a dialect feature here, too (Smits 1996: 87, 128-129). c) The generalization of the dialectal singular form ben to the plural, e.g., de kerkdiensten ben ook allemaal Engels 'the church services are all in English, too' (standard Dutch ze zijn 'they are'). This generalization is not the direct result of dialect influence. The generalization of the zero form ben to the plural is attributed to interference from English, since in that language zero forms (i.e. forms not marked for person and number) are prominent throughout the present tense paradigm (with the exception of the 3rd singular) (Smits 1996: 129).

254

Caroline Smits

Consider tables 6a-6c for the distribution of these deviations over the three subcorpora: Table 6a. Present tense; irregular zijn: benne in the plural in Iowa Dutch 1966 instead of standard Dutch zijn

group 1: group 2: group 3:

Total (n)

benne (+dialect)

121 125 72

28.9 24.8 38.9

(%)

Table 6b. Present tense; irregular zijn: ben in the 2nd singular in Iowa Dutch 1966 instead of standard Dutch bent

group 1: group 2: group 3:

Total (n)

ben (+dialect)

11 9 2

45.5 66.7 50.0

(%)

Table 6c. Present tense; irregular zijn: 1st singular ben in the plural in Iowa Dutch 1966 instead of standard Dutch zijn

group 1: group 2: group 3:

Total

ben (-dialect)

(n)

(%)

121 125 72

19 36.8 33.3

On the (non-)persistence of dialect features in American Dutch (2)

255

Dialect forms - both benne in the plural (table 6a) and ben in the 2nd singular (table 6b) - are already prominent in group 1.9 The use of ben in the plural (table 6c) is relatively marginal in group 1, but it becomes more prominent in groups 2 and 3; the difference between group 1 on the one hand, and groups 2 and 3 on the other is significant (Chi-square = 9.824, ρ < .005, η = 318). (vi) Present-tense forms of the irregular verb hebben 'to have' The most prominent deviations are: a) The use of het in the 3rd singular (instead of standard Dutch hij heeft 'he has'). This is unmistakably a deviation from the standard language that has its roots in dialectal varieties of Dutch, where this 3rd singular form is very popular (Smits 1996: 71, 131). b) The use of heb in the 2nd singular (instead of standard Dutch je hebt 'you (singular) have'). This is most likely of dialectal origin, too (Smits 1996: 161). However, the use of this stem form may have been enhanced by interference from English, where verb stems (zero forms) are used throughout the paradigm (with the exception of the 3rd singular). c) The generalization of het to the 2nd singular (instead of standard Dutch je hebt 'you (singular) have'). It does not result from direct influence from the dialects. Instead, it involves an externally-induced generalization of dialectal 3rd singular het (a present tense zero form) through the paradigm (Smits 1996: 161). The distribution of these deviations is given in tables 7a-7c: Table 7a. Present tense; irregular hebben: het in the 3rd singular in Iowa Dutch 1966 instead of standard Dutch heeft

group 1: group 2: group 3:

Total (n)

het (+dialect)

42 49 42

85.7 71.4 71.4

(%)

256

Caroline Smits

Table 7b. Present tense; irregular hebben: heb in the 2nd singular in Iowa Dutch 1966 instead of standard Dutch hebt

group 1: group 2: group 3:

Total (n)

heb (+dialect)

15 10 11

40.0 40.0

(%)

-

Table 7c. Present tense; irregular hebben: het in the 2nd singular in Iowa Dutch 1966 instead of standard Dutch hebt

group 1: group 2: group 3:

Total (n)

het (-dialect)

15 10 11

13.3 40.0 54.4

(%)

Both dialectal het in the 3rd singular (table 7a) and dialectal heb in the 2nd singular (table 7b) are already prominent in group 1. The use of het in the 2nd singular (table 7c) is relatively marginal in group 1, and it becomes more prominent in groups 2 and 3. (vii) Preterite forms of verbs with a regular preterite in standard Dutch The most prominent deviations are: a) The use of preterite forms in -tenl-den in the singular (instead of standard Dutch -te!-de). Most likely, the use of-«-final preterites has a phonological basis rather than a dialectal or morphological basis. It appears that there is large-scale confusion among the speakers of Iowa Dutch as to the pronunciation of word-final en. As a result, the preterite plurals in -tenlden in Iowa Dutch are pronounced as either [te/ds], [tan/dsn] or [tn/dn], apparently in free variation. The variable realization of preterite plurals seems to have lead to confusion as to the pronunciation of preterite singulars in -te!-de as well. Speakers of Iowa Dutch have inferred that

On the (non-)persistence of dialect features in American Dutch (2)

257

final -tel-de may be pronounced as either [ta/do], [ton/dan] or [tn/dn] as well (Smits 1996: 137-138). b) The use of present-tense forms instead of preterite forms. The use of present-tense forms is not the direct result of dialect influence. It involves the non-conventional use of existing inflectional forms leading to reduction; the present tense is no longer distinguished from the preterite. It is a manifestation of language attrition, i.e., of the fact that speakers of Iowa Dutch do not fully master the standard Dutch tense system any longer. As a result, (unmarked) present-tense forms are generalized at the cost of (marked) preterite forms (Smits 1996: 98-99). Consider tables 8a and 8b for the distribution of these deviations over the three subcoipora: Table 8a. Preterite; regular preterites: the use of preterite forms in -ten!-den in the singular in Iowa Dutch 1966 instead of standard Dutch -tel-de

group 1: group 2: group 3:

Total (n)

-tenl-den (-dialect)

70 63 42

7.1 17.5 9.5

(%)

Table 8b. Preterite; regular preterites: the use of present tense forms in the preterite in Iowa Dutch1966 instead of standard Dutch preterite forms Total (n) group 1: group 2: group 3:

113 98 72

Pres. tense forms (-dialect)

(%) 0.9 5.1 9.7

The generalization of regular preterites in -ten!-den to the singular (table 8a) is most prominent in group 2. The generalization of present tense forms at the cost of preterite forms (table 8b) is highly marginal in group 1, it grows in prominence in group 2, and it is most prominent in group 3.

258

Caroline Smits

(viii) Preterite forms of the irregular verb hebben 'to have' The most prominent deviations are: a) The use of the singular form had in the plural (instead of standard Dutch hadden 'had'). It is attributed to interference from English, in which zero forms are used throughout the preterite paradigm (Smits 1996: 97). b) The use of present-tense forms instead of preterite forms. As in the case of the regular preterites (cf.(vii) above), the generalization of present tense forms involves the non-conventional use of existing inflectional forms leading to reduction. Table 9a. Preterite; irregular hebben: the use of singular had in the plural in Iowa Dutch 1966 instead of standard Dutch hadden

group 1: group 2: group 3:

Total (n)

had (-dialect)

62 52 64

4.8 7.7 12.5

(%)

Table 9b. Preterite; irregular hebben: the use of present tense forms in Iowa Dutch 1966 instead of standard Dutch preterite forms Total (n) group 1: group 2: group 3:

144 115 91

Pres. tense forms (-dialect)

(%) 1.4 -

13.2

The use of singular had at the cost of plural hadden (table 9a) is most marginal in group 1, it grows in prominence in group 2, and it is most prominent in group 3. The use of present tense forms instead of preterite forms (9b) is highly marginal in groups 1 and 2, and relatively frequent in group 3. The difference between groups 1 and 2 on the one hand, and group 3 on the other is significant (Chi-square = 23.891, ρ < .001, η = 350).

On the (non-)persistence of dialect features in American Dutch (2)

259

(ix) Preterite forms of irregular modal verbs The most prominent deviations are: a) The use of preterite singular forms in the plural. As in the case of hebben (cf. (viii) above), the use of such forms instead of plural preterite forms is the result of interference from English. b) The use of present tense forms. As in the case of hebben and the regular preterites, the use of present tense forms instead of preterite forms in the case of modal verbs involves the non-conventional use of existing inflectional forms leading to reduction. Consider tables 10a and 10b: Table 10a. Preterite; irregular modal verbs: the use of singular forms in Iowa Dutch 1966 instead of standard Dutch plural forms

group 1: group 2: group 3:

Total (n)

Sg. forms (-dialect)

56 44 68

8.9 15.9 8.8

(%)

Table 10b. Preterite; irregular modal verbs: the use of present tense forms in Iowa Dutch 1966 instead of standard Dutch preterite forms Total (n) group 1: group 2: group 3:

169 188 164

Pres. tense forms (-dialect)

(%) _ -

7.3

The use of preterite singular forms instead of preterite plural forms (table 10a) is most prominent in group 2. The use of present tense forms instead of preterite forms (table 10b) only occurs in group 3.

260

Caroline Smits

5. Inflectional forms in Iowa Dutch 1966; further discussion The above discussion of the inflectional deviations from standard Dutch in Iowa Dutch 1966 (19 in sum), and their distribution over the three subcorpora is summarized in table 11. This table summarizes the percentages that were presented in tables 2a-10b: Table 11. Inflectional deviations in Iowa Dutch 1966; distribution over the three subcorpora Dial, origins? Group 1 Group 2 Group3

Significant differences?

NOMINAL PLURALS

monosyll. base + -eren 1. -ert-ers 2. gen.-en

yes no

39.7

32.4 11.8

12.8 2.6

no no

no

26.1

15.0

20.0

no

no

3.0

2.5

35.0

yes: 1/2 vs 3

70.0 41.7 6.3

no yes: 1/2 vs 3 no

-

ATTRIBUTIVE ADJECTIVES

sg. definite het nouns 3. gen. 0 sg. indefinite het nouns 4. gen. -e PRESENT TENSE VERBS

infinitive in -n 5. -t forms in 1st sg. 6. -en forms in pi. 7. stems in 1st pi. irregular zijn 8. benne in pi. 9. ben in 2nd sg. 10. ben in pi. irregular hebben 11. Ae/in3rdsg. 12. heb in 2nd sg. 13. het in 2nd sg.

no no no

3.2 8.3

25.0 3.1 25.0

yes yes no

28.9 45.5 19.0

24.8 66.7 36.8

38.9 50.0 33.3

no no yes: 1 vs 2/3

yes yes no

85.7 40.0 13.3

71.4 40.0 40.0

71.4

no no no

-

-

54.4

On the (non-)persistence of dialect features in American Dutch (2)

Dial, origins? Group 1 Group 2 Group3

261

Significant differences?

PRETERITE VERBS

regular verbs 14. -ten!-den in sg. 15. present tense forms irregular hebben 16. had in pi. 17. present tense forms irregular modal 18. sg. forms in pi. 19. present tense forms

no no

7.1 0.9

17.5 5.1

9.5 9.7

no no

4.8 1.4

7.7

12.5 13.2

no no

8.9

15.9

-

-

-

8.8 7.3

no no no yes: 1/2 vs 3 no no

We will now further discuss the differences between the deviation types that are most likely of dialectal origin and those that are probably not (according to Smits 1996), as far as their pattern of distribution over the three subcorpora is concerned: (i) In group 1 (i.e. the informants who acquired Dutch in the Netherlands) the deviations of dialectal origin (5 in sum) are always more prominent than the deviations that are the result of language contact/attrition (the remaining 14): the percentage of dialect deviations (i.e. 1, 8, 9, 11, 12 in table 11) ranges between 28.9 and 85.7 in group 1; the percentage of nondialectal deviations (i.e. 2-7,10, 13-19 in table 11) ranges between 0 and 26.1 in group 1. This difference in group 1 is significant (Chi-square = 259.916, ρ kastade is a regular sound change that stopped at this point, although he stresses that tendencies towards loss of the

278

Inge Lise Pedersen

dental are neutralized by the grammatical function of the inflectional suffix (cf. the quotation above). I am not convinced of the old age of kastade. The very region that is dominated by kastade has, at the same time, very few verbs belonging to this conjugation type, the majority otherwise belonging to this type having changed to the 3rd conjugation (cf. Western Norway and Jutland). We are told that the weakening t> d had a smaller extension after unstressed vowel than after stressed one. However, it is important to bear in mind that most f s after unstressed vowels do not belong to stems but to conjugational suffixes, i.e., the system of conjugation or of declension could have a strong impact on the development. Both Wessen and Benson speak about preserved dentals in the preterite. It sounds strange that the final consonant should not only have been preserved, but strengthened in exactly those areas where the unstressed vowel i, u, and ο have merged into schwa. The dental d or t is not necessarily a result of regular sound change but of restoration.

4. A discussion of different explanations In general the weakening of postvocalic consonants has gone much further in Danish than in Norwegian and Swedish: k has been lost after certain vowels, for example, and a number of the ö's in modern standard Danish were reintroduced in the 19th century, even in Sealand where the ö's are most numerous, some of them have been restored. These extreme weakenings are not reflected in the spelling, however. The modern spelling system dates in principle back to the 16th century, and the spelling of the weak preterite forms was also fixed at that time: for example, -ede is the normal spelling in the 1550 Bible. The Swedish - ade spelling also dates back to the 16th century, but many d's following stressed vowels are absent until the beginning of the 18th century. This might indicate a certain parallel between the Danish and the Swedish developments. In late medieval Danish manuscripts, with their greater variation in spelling, we find the preterite ending written as -ed(h)e, -ed(h), -et, and also in some cases as -e. Since ί-endings are also found in words like met 'with', where the only known pronunciations are mce and moed, they are likely to indicate a general confusion of ö, d, and t and not a "final hardening". In view of such facts I would postulate that all -ot's (perhaps also - ad's) are restored forms, as can be seen from map 4, showing dialect correspond-

The impact

279

Map 4

Old Danish -t

β

jj zero (loss)

i /

β 1θ mm i ;] d. voiced

M gillM|IIfI| I d \ d voiced ] J\r\R tsaRtlSfS! t, d, R Willi 1 ί

ences to medieval -t: if the preterite suffix followed postvocalic -t (which is the general tendency), the present and preterite tenses would merge throughout the whole of southern Jutland. We find many examples of -et adjectives following this route of change, i.e., ending in -9 in south-west Jutland; this points in the same direction, a functionally conditioned restoration. A merger of the preterite and the present tense seems to be rather unfunctional. As far as we know, this merger did not take place in Funen,

280

Inge Lise Pedersen

since -r is kept in the present tense, and the preterite ending was -a; for this reason, the re-establishment of the dental preterite is not functionally conditioned in this area, but is connected with the dialect levelling in general. In Jutland the restoration is much older, but was also much more "necessary" (although the north-western part has managed until now with the presentpreterite merger). It is interesting to note that Norwegian linguists choose different explanations for this development than their (Swedish and) Danish colleagues do. It is tempting to see this in relation to their different linguistic tradition. In Denmark and Sweden there has been an uninterrupted literary tradition from the 13th century to the present day, and the history of the language is mostly based upon this literary tradition. Most researchers into the history of language have worked within a (modified) neo-grammarian paradigm, in which regular sound change has played a far greater role than morphology. Words have been treated as combinations of sounds, not as parts of conjugational categories: we have no history of conjugations in Danish. In Norway the situation is quite different, since the literary tradition was interrupted around 1500: there was no Norwegian central government from 1536, i.e. no Norwegian chancellery, and Danish was the only public and written language. For this reason, the spoken dialects of the 19th and 20th century provided the main sources for the history of Norwegian: they were the link between Old Norse and modern Norwegian. But which modern Norwegian do we mean? To Seip it was the Dano-Norwegian Bokmäl; he therefore looked for a continuity from the medieval manuscripts to Bokmäl. It was quite otherwise for supporters of Nynorsk, like Kjell Venäs. Since Nynorsk is based upon spoken dialects, these researchers have looked for continuity in the dialects themselves,2 whereas forms from Bokmäl are regarded as secondary and foreign (i.e., Danish) influenced. In quite another context Inger Ejskjoer has asked the very guarded question: why should hyper-restitutions not be possible in Jutland? (Ejskjcer 1971:89). I would like to ask the same question, and add another: why do Jutland linguists, for example Peter Skautrup, insist on the regularity of sound changes in Jutlandish far more than their colleagues within insular dialectology do? The answer is, I think, for historical reasons. Danish dialectology was first studied in the 1850's, as part of the construction of a Danish national identity, and the study of (southern) Jutland dialects played a crucial role. The job of the dialectologists, first and foremost K.J.Lyngby, was to prove that these dialects were Danish and not German, as Jacob Grimm had

The impact

281

claimed (Grimm 1848, cf. Adriansen 1990:523). Dialectology in general was anxious to portray the dialects as part of a pre-modern peasant culture based on family, local community, religion, and tradition. If we add to this the politically tense situation (which resulted in the Danish-German War of 1864), it is not surprising that deep-rooted forms and words, continuity and authenticity were given prominence. The description of insular dialects has never had this national agenda. In addition, we encounter an endeavour to construct a Jutish language history as a regional corrective to mainstream language history, that is the history of the standard language and of the standardization process based upon the insular and Copenhagen dialects. Peter Skautrup's five-volume history of the Danish language Det danske sprogs historie was originally perceived as a historical description of his mother's West Jutlandish dialect (Skautrup 1944:VIII). Seen in this context, it is understandable that Jutland dialectologists are inclined to explain changes as examples of regular sound change, i.e., to stress continuity, whereas an explanation of -et as restoration would imply a less continuous linguistic history of Jutland. Dialectology is part of the construction of the past. Although Danish dialectologists are traditionally wary of justifying themselves publicly, the -ef-story demonstrates that there is an interaction between the way in which dialectology is constructed and the way in which changes are described and explained.

Notes 1. Map 50 in Bennike & Kristensen does not distinguish clearly between preterite and past participle, but it is evident from the background material in IDD that the forms quoted from West Funen are past participle only. 2. Jahr claims that this explains why structuralism had so little impact on Norwegian dialectology: it was due to the fact that the connection between Old Norse and the modern Norwegian dialects was a central issue for New Norwegian linguists (Jahr 1996:89). 3. Grimm's claims made a sensation in Denmark, since he drew political consequences from his linguistic postulates.

282

Inge Lise Pedersen

References: Aaserl, Ivar 1899 Det norske Folkesprogs Grammatik. Kristiania. Adriansen, Inge 1990 Fcedrelandet, folkeminderne og modersmälet, Senderborg: Museumssrädet for Sendeijyllands Amt. Bennike, Valdemar & Marius Kristensen 1898-1912 Kort over de danske Folkemäl. Kjabenhavn: Gyldendal. Benson, Sven 1965-70 Südschwedischer Sprachatlas 1-4. vol. 4. Lund: Gleerup. Ejskjoer, Inger 1971 "Nogle kortdiftongdannelser". In: Studier i dialektologi og sproghistorie tilegnede Poul Andersen. K0benhavn:Akademisk forlag, pp. 75-90. Hellevik, Alf 1953 Omstridde sprakspersmäl. Oslo: Det Norske Samlaget. Jahr, Ernst Häkon. 1996 "Nynorsk spräkforskning - en historisk oversikt". In: Carol Henriksen et al.(eds.). Studies in the development of linguistics in Denmark, Finland, Iceland, Norway and Sweden. Oslo:Novus, pp. 84-101. Levander, Lars 1928 Dalmalet. Beskrivning och historia 2. Uppsala. Pedersen, Inge Lise 1994 "Dialektkontakt og regionalisme". In: Ulla-Britt Kotsinas & John Helgander (eds..). Dialektkontakt, spräkkontakt och spräkförändring i Norden. MINS 40. Stockholm, pp. 242-251. Petterson, Egil 1975 Spräkbrytning i Vest-Norge 1450-1550, vol. 1. Bergen: Universitetsforlaget. Seip, Didrik Arup 1934 Studier i norsk spräkhistorie. Oslo: Aschehoug. Seip, Didrik Arup 1952 Omstridde sparsmäl i norsk spräkutvikling. Oslo: Fabritius. Skautrup, Peter 1944-1970 Det danske sprogs historie. Kebenhavn: Gyldendal. Venäs, Kjell 1974 Linne verb i norske malfere. Oslo: Universitetsforlaget. Wessen, Elias 1969 Vara Folkmäl. Lund: Fritzes.

Meta-constraints and constraint ranking in relation to the representation of nasality and palatality across French dialects1 Michael D. Picone

1. Introduction Inherent phonemic instability in relation to the posterior area of the oral cavity has been cited as an important contributor to historical linguistic change (Martinet 1955, recapitulated in Labov 1994: 117-118). The Great Vowel Shift in English is one of the best-known examples. Martinet claims that such shifts are to be attributed to two antagonistic pressures. On the one hand, there is a horizontal pressure to achieve symmetrical oppositions between front and back vowel inventories in order to arrive at parallel gradations of height. This is part of a larger, functional-perceptual theory citing the tendency towards achieving maximal phonemic contrastivity among the members of the phonemic inventory. On the other hand, the posterior region of the vocal cavity, compared to the anterior region, is small and liable to overcrowding, which results in a similar kind of vertical pressure - unintentionally antagonistic to the first one in its consequences - to alleviate overcrowding by shifting members of the phonemic inventory away from the posterior region. This took the form of vowel shifting and diphthongization in the English language as vowels jockeyed for space while simultaneously seeking to maintain contrastivity. Whether or not posterior overcrowding is the real motor for this kind of change (Labov is not convinced, 1994: 220221) will not be crucial for what follows, even though this study will be dealing with consonants in the posterior region. What is crucial, however, is the evidence for the general tendency to promote equilibrium through symmetry in the overall phonetic inventory.2 This tendency is certainly ubiquitous across languages, and any phonological theory that ignores its import does so at its own explanatory peril. In this vein, let me point to the evidence from French dialects suggesting that symmetry-enhancing pressures can exert influence on the phonemic inventory of consonants in ways that account not only for various shifts, but also consonant loss and consonant acquisition.

284

Michael D. Picone

In this paper, then, I assert that, for a variety of French dialects, the historical and contemporary instability of the palatal nasal /ji/, as well as the newly prominent velar nasal /η/, are best understood within a framework incorporating a symmetry-sensitive constraint and perhaps other such "metaconstraints" (see below for a full explanation of the term and of the notions underlying it). I further assert that inclusion of such constraints, which are arguably of a universal nature, in the Optimality-Theoretic framework of phonology will appropriately outfit that theory in a way that helps expand its explanatory power and its relevancy to the domain of language change.

2. Background: historic changes in the French consonantal inventory Historic instability in relation to French consonants has manifested itself in the posterior region of the oral cavity. A tripartite opposition for obstruents and nasals, inherited from Latin and consisting of a labial, coronal, and velar member for each class, became enriched with the addition of palatals. These palatals were first introduced in the second century due to the migration and palatalization of the respective velar counterparts in certain environments (mainly preceding [ - back] vowels) due to place assimilation, as shown in (1). (1) k > c, g > j, η >ji or + consonantal + high + back

>

[-back]

- consonantal + back

/.

By the 13th century (Bourciez [1967] 1982: 128), in these same environments, the palatal obstruents had been subjected to further frontward migration while simultaneously retaining palatalization by resorting to a strategy of disarticulation, which, in turn, eventually permitted obstruent loss for the dominant dialect, resulting in the type of sequence displayed in (2). (2) Latin /kEntum/ k

>

c

>

French

/sä/

>

ts

s

>

Meta-constraints and constraint ranking

285

However, various regional and popular dialects retained disarticulated palatalization in some environments. Though these dialectal features are now disappearing in France, they have been preserved in the speech of descendants of the original French colonists in Quebec (Gendron 1966: 111, 115), accounting for the fact that palatalized coronal obstruents are still frequent there among some speakers, as in (3)a-b. (3) a. On vapa[i*]iner. 'We're gonna skate.'

b. On va [dz]iner. 'We're gonna eat lunch.'

This trait was not retained, however, by the Acadian French, whose population and dialectal traits are traceable to the Poitou region of France more than any other. This detail will take on importance when we look at the status of palatality in conjunction with nasality as manifested in the Francophone population of Louisiana whose dialect is rooted in Acadian French.3 What is important to take note of at this juncture is that the original velar obstruents, /k/ and /g/, did not palatalize in all environments whereas as the velar nasal /η/ did and was, therefore, replaced de facto by the new phoneme /ji/. Hence, after the disappearance of the palatal obstruents in various dialects - including those of the Ile-de-France and the Touraine, which converged to become the prestige dialect that is now codified as "French" - a disparity in posterior symmetry was the result, as displayed in Table 1. Table 1. Inventory of stops and nasals for "Metropolitan" French labial

coronal

stops [ - voice] [+ voice]

Ρ b

t d

nasals

m

n

palatal

velar

k g Ji

The foregoing is the backdrop against which we will now consider the disarticulation tendencies associated with the palatal nasal /ji/ for the following dialect areas: France, Quebec and Louisiana. A brief description of the observed phenomena (Section 3) will be followed by an analysis (Section 4) of the various constraints that may be involved.

286

Michael D. Picone

3. Disarticulation of nasality and palatality across dialects 3.1. In France, there is a strong tenancy toward complete disarticulation of nasality and palatality intervocalically, as in (4): (4) ρ > V nj V

[konje] cogner 'to knock, to bump'

The palatal nasal tends to be retained intact only word-finally (e.g. montagne [mötaji]) and preconsonantally (e.g. saignement [sejimü]), but, crucially, this is precisely where it is not contrastive with /nj/ because the latter is disallowed in such environments. 3.2. In Quebec, there is a greater tendency to retain the palatal nasal, even intervocalically, but, in keeping with a general property of Quebec French, there appears to be a stronger tendency for nasalization of the preceding vowel, as in (5): (5) ji > Vji V

[kojie]

Hence, there is only partial disarticulation of nasality and palatality based on the observation that they are not coextensive in their spread. In fact, the opposing phenomenon is often observed whereby fortuitously contiguous palatality and nasality are combined and co-articulated, as in example (6), possibly as a means of reinforcing the integrity of the palatal nasal: (6) /panje/panier 'basket' > [päjie] 3.3. In Louisiana, in what may constitute more of an historic shift than an ongoing, synchronic one, the palatal nasal may have been entirely replaced in the intervocalic position (and perhaps in other environments) with nasalization of the preceding vowel in contiguity with a palatal glide, as in (7): (7) ji > ν j V

[koje]

This amounts to complete disarticulation of nasality and palatality, excepting those instances where nasality is co-articulated with the glide, which occurs especially word-finally and preconsonantally but may also occur intervocalically for some speakers (cf. Papen & Rottet 1997: 76; in such cases, there is still partial disarticulation since palatality is not coextensive with nasality).

Meta-constraints and constraint ranking

287

This tendency has generalized such that an underlying coronal nasal [n] will undergo a similar treatment when juxtaposed to a following palatal glide. This appears to be categorical for some lexical items, such as (8), but variable for others, such as (9) (cf. Conwell and Juilland 1963: 123). This tendency forms an interesting contrast with the tendency of Quebec French exemplified in (6). (8) [pakän] pacane 'pecan'

>

[pakOje] pacanier 'pecan tree'

(9) [manjar]/[mäjaer] maniere 'manner, way' This might point to different levels of processing, ä la Kiparsky's model of lexical phonology, with initial processing taking place at the lexical level, such as in example (7), and subsequent processing at the post-lexical level whenever the appropriate structural description is met, such as in examples (8) and (9). However, the connection I want to pursue in this present essay is not one with lexical phonology, as this last remark might suggest, but rather with more recent notions as embodied in Optimality Theory.

4. A symmetry-sensitive Optimality-Theoretic approach

4.1. To quote from Itö, Mester and Padgett (1995: 573), "Optimality Theory [...] holds that all phonological constraints are uniformly present in all grammars and at all levels, but are in principle violable [...]." Hence, output is determined not by a series of inviolable rules applied to the input (or even by repair and rescue rules compensating for some violations) but simply by relative ranking of the constraints. Though the constraints are deemed universal, their relative ranking is particular to a language, thereby accounting for different outputs. Archangeli (1997: 11) sums this up by saying, "Optimality Theory views Universal Grammar as a set of violable constraints [and] the grammars of specific languages as the language-particular ranking of those constraints." Pulleybank (1997:101) echoes this by remarking that "the range of possible rankings gives us the rich diversity that we observe in the phonological systems of the world's languages". It is selfevident that whatever is true for languages must also be true for dialects if indeed the claim to universality for the Optimality-Theoretic model is to be upheld. The following is an experiment in the application of Optimality Theory to the explanation of aspects of dialectal diversity. However, I will be

288

Michael D. Picone

introducing an important modification in the theory, allowing for the possibility of overarching, system-sensitive constraints, which might be conveniently referred to as "meta-constraints". 4.2. Essential to the working of Optimality Theory are the proposed "faithfulness constraints" (Prince & Smolensky 1993), such as PARSE and FILL. PARSE stipulates that all input features actually appear in the output and, likewise, (incorporating elements rooted in autosegmental phonology) that all association lines in the input are kept in the output. Conversely, FILL stipulates that all output features are part of the input and that all association lines in the output are part of the input. In other words, the faithfulness constraints, taken together, require that the input and the output be identical.4 PARSE and FILL appear to be redundant at first, but this is not really the case. For example, the inclusion of an additional feature in the output would violate FILL but not PARSE. In some languages (or dialects), all faithfulness constraints might operate as a bundle and be ranked together in the constraint hierarchy. In other cases, they might have different ranking relative to each other and/or relative to other constraints. 4.3. Let us consider data from three dialects of French using an OptimalityTheoretic approach as represented in the tableaux below. As a starting point, I have assumed the same underlying input for the three dialects for the lexical item cogner. In an attempt to account for its divergent realizations in the three dialects, I will invoke four constraints. Two of them, PARSE and FILL, come from preestablished models (Prince & Smolensky 1 9 9 3 , Itö, Mester and Padgett 1 9 9 5 ) . Concerning the other two, PRENAS (= "prenasalization") is introduced based on the rationale that prenasalization of vowels preceding nasal consonants must surely rank as a universal tendency, if indeed universal constraints exist. SYM, which stands for "symmetry sensitivity" and which is also of my invention, introduces a whole new property into the system. The nature of that property and the rationale for the introduction of SYM will be explicated in the following. For the time being, let us disregard SYM and examine the interrelation of the other three constraints as demonstrated in Tableau 1.

Meta-constraints and constraint ranking

289

Tableau 1. Quebec Input: /kojie/ cogner 'to knock, to bump' Constraints

—>

(SYM)

PRENAS

PARSE

FILL

Output Candidates I I

*| '

I I

I

*!

|

**

1

*

kojie konje

(*)

köjie köje

1 Ο

,

1

_i . §§j , * - ρ I iiiiliSilHÄiSj .v.

In Tableau 1, as in all the subsequent tableaux, the set of output candidates is made up of all the variant forms of cogner that appear in the three dialects plus a perfectly faithful rendering of underlying /kojie/. The winning candidate in Tableau 1 (indicated by a pointed finger) is the third from the top, [kojie]. It doesn't violate PARSE because everything in the underlying input was parsed, but it does violate FILL (indicated by an asterisk) because a new association line has been added to the preexistent nasal feature, spreading nasality to the left, which of course is in conformity with PRENAS. Though it violates FILL, [köjie] is still the winning candidate because of the lower ranking of FILL relative to the other two constraints PRENAS and PARSE (disregarding SYM for the moment). Though the ranking of PRENAS relative to PARSE is not yet crucial (indicated by the broken line), PRENAS must be ranked higher than FILL (still disregarding SYM) in order to result in the disqualification of the first two output candidates [kojie] and [konje] which do not conform to the requirements of PRENAS ([konje] also violates FILL, twice, because of the addition of [+coronal] and [+continuant] features). Shading indicates that a constraint's relevancy has been preempted due to a prior fatal violation (symbolized by *!) of a higher-ranked constraint (see Archangelil997: 12 for the conventions of tableau design). Notice that even without benefit of SYM, the fourth candidate [köje] is disqualified by virtue of the higher ranking of PARSE. If FILL had been ranked higher than PARSE, however, then SYM would have been necessary to disqualify the fourth candidate. Otherwise, the third and fourth candidates would have tied as winners (the fourth candidate also violates FILL by virtue of an added [+continuant]

290

Michael D. Picone

feature in the output; it simultaneously violates the PARSE constraint owing to the presence of a [ - continuant] feature in the input that was not parsed). Though potentially useful, SYM is not indispensable in this first example. Its greater usefulness will become apparent in subsequent tableaux. For example, in Tableau 2, if we were limited to the three constraints PARSE, FILL and PRENAS, the correct candidate [konje], which is the second candidate, cannot win, regardless of the relative ranking of those three constraints: if PRENAS were ranked highest of the three and PARSE were second highest (as in the ranking established for Quebec: PRENAS » PARSE » FILL), then the third candidate would be the incorrect winner; if FILL were next highest after PRENAS, then the third and fourth candidates would tie as winner; if the ranking were PARSE » FILL » PRENAS, the first candidate would win; finally, if the ranking were FILL » PARSE » PRENAS, the first candidate would become the incorrect winner. Only the addition of the constraint SYM in the order shown in Tableau 2 (and in alternate orders where PARSE and SYM reverse ranking relative to each other and/or where FILL and PRENAS reverse ranking relative to each other) allows for the second candidate [konje] to be the winner.5 Tableau 2. France Input: /kojie/ cogner 'to knock, to bump' Constraints

—>

SYM

PARSE

FILL

PRENAS

Output Candidates I kojie es"

konje

*I

i

1

|1 1

:

> •*

köjie köje

**

*!

.

*

1 j

I .

•BNHHHMHNM

4.4. Having demonstrated its useful potential in this preliminary example, it is now appropriate to make specific statements about the nature of SYM, which stands for "symmetry sensitivity". SYM stipulates that the output form must be sensitive to the symmetry requirements of the rest of the phonological system. Hence, this constraint corresponds explicitly to the observable

Meta-constraints and constraint ranking

291

universal tendency towards symmetry in the inventory of speech sounds in any given language or dialect and turns it into a formal part of the grammatical apparatus. Like any other constraint in the Optimality Theoretic framework, SYM is violable if it is outranked by some other constraint. SYM is similar to the faithfulness constraints (and different from other constraints) in that it is overarching in its application. That is, it transcends the realm of phonological particularities (as exemplified here by PRENAS) and becomes concerned with the system as a whole. One might reasonably allude to it as a "meta-constraint". What is special about SYM and what sets it apart even from the faithfulness constraints, however, is the fact that it will allow for different output across languages and dialects even where the candidate set is the same and even if SYM has the same ranking relative to other constraints. This is because SYM is sensitive, in each case, to the symmetry requirements of a different overall phonetic inventory that is external to both the candidate set and the constraint hierarchy. Consequently, SYM will not disqualify the same candidates for every language or dialect. Notice the different placement of the disqualifying asterisk under SYM for Quebec and France in Tableaux 1 and 2, even though the candidate set is identical in both cases and even though SYM is given first place in the relative ranking in both cases. Thus Sym would introduce an important new property to Optimality Theory by converting it into a three-variable model: Output will vary not only as a function of input and constraint rankings but also as a function of systemsensitive meta-constraints (I will return to this point and expand upon it after completing discussion of the tableaux). For the Metropolitan French of France in Tableau 2, SYM manifests itself in the elimination of the palatal nasal to help redress the imbalance displayed in Table 1. Conversely, in Quebec, where palatalized obstruents are present in the phonetic inventory, SYM manifests itself by maintaining the integrity of the palatal nasal (cf. Gendron 1966:115). The incorporation of SYM in Tableau 2 succeeds in disqualifying the first and third candidates so that the correct candidate, [konje], is the winner. The addition of SYM to Tableau 1 for Quebec French in no way compromises the earlier correct result obtained for that tableau without taking SYM into consideration, since, in the context of that dialect, SYM will disqualify different candidates than it did in Tableau 2. The ranking of SYM over the faithfulness constraints in Quebec French, moreover, is justified as an explanation for the conversion of underlying /nj/ to [p], as in example (6) and Tableau 3. That is, co-articulation of nasality and palatality in [pQjie] represents the efficient exploitation of an opportunity for symmetry reinforcement that is seized upon by SYM. This being the

292

Michael D. Picone

case, the partial disarticulation of nasality and palatality in examples such as (5) is put in its proper perspective as simply being part of a general process stemming from the application of PRENAS in Quebec French rather than from some other disarticulatory tendency. (Notice that the ranking of FILL at the bottom of the hierarchy was established in Tableau 1 for Quebec. In Tableau 3 , taken alone, the ranking of FILL relative to PARSE would not be evident. Nothing in either Tableau 1 or 3 would preclude the reversal of SYM and PRENAS relative to each other as long as both rank higher than the faithfulness constraints. Having arrived at the latter determination, based on the example in Tableau 3, it would now be possible to go back and revise Tableau 1 by inserting a solid line between the constraints PRENAS and PARSE and then adjusting the shading accordingly.) Tableau 3. Quebec Input: /panje/panier 'basket' Constraints

—>

SYM

PRENAS

PARSE

FILL

Output Candidates -l panje

*|

1 I

* *

pönje pajie BS"

pope

.

•

•

*!

1 1

FFSYP

**

*

Consider now Tableau 4 for Louisiana ("Cajun") French. Since prenasalization is also fairly ubiquitous for Louisiana French (Conwell & Juilland 1963: 116, Papen & Rottet 1997: 75), it is reasonable to begin by positing the same order of constraints, PRENAS » PARSE » FELL, that obtained for Quebec French. Since the high ranking of SYM for France and Quebec proved felicitous, I will begin by giving it the top ranking for Louisiana as well. For some speakers of Louisiana French, it is possible that the palatal nasal is no longer a part of the phonemic inventory. Those who do retain it, usually have it in word-final position (cf. Conwell and Juilland 1963: 68). Since Louisiana French does not have palatalized obstruents, any reinforcement of the palatal nasal would work contrary to SYM and would contribute to an imbalance

Meta-constraints and constraint ranking

293

similar to the one for the French of France. Using the same four constraints as before, the ranking shown in Tableau 4 succeeds in disqualifying all but the intended winner. (Notice that the constraints PARSE and FELL could be made to exchange places in the ranking shown in Tableau 4 and it would no make no difference in the correct selection of the winning candidate. As in the preceding Tableaux 1 and 3 for Quebec, the order of SYM and PARSE relative to each could be reversed in Tableau 4 for Louisiana without disturbing the outcome.) Tableau 4. Louisiana Input: /kojie/ cogner 'to knock, to bump' Constraints

—»

SYM

PRENAS

PARSE

FELL

Output Candidates i kojie konje

*Ι

1

*

I1 ι

*!

•

1

• , . -ί: i

köjie es3

köje

1

*

t 1

**

Ι

. *

1

4.5. In my introductory remarks, I made the claim that the modified model being proposed here - by virtue of the inclusion of SYM - can help explain consonant addition as well as elimination. While SYM works to reduce the incidence of [ji] in the French of France and Louisiana, presumably due to the absence of corresponding obstruents, it may be simultaneously favoring the incorporation of the velar nasal [η] in order to complement the presence of the velar obstruents. The presence of the velar nasal in place of the palatal nasal has been remarked upon by Conwell and Juilland (1963: 68) for some bilingual Louisiana French speakers. Conwell and Juilland attribute it to English influence. The accessibility of the English velar nasal does indeed play a likely role, but it does so, I contend, in concert with the pressure exerted by the symmetry constraint. This is perfectly in keeping with most language change due to language contact: Such change does not happen fortuitously but rather interacts with some other ongoing dynamic (or complex of dynamics) affecting the recipient language (Aitchison 1991, Picone

294

Michael D. Picone

1996). The involvement of some other, internal dynamic is even more obvious in the case of the French of France, where the incidence of the velar nasal has dramatically increased due partly to English lexical borrowings yet without the aid of widespread (or even small-scale) bilingualism. In Tableaux 5, for France, the candidate set consists of virtually all the variants that have actually been noted in the literature as phonological renderings for the borrowed term parking 'parking lot/garage/place'. Tableau 5. France Input: /paRkiq/ parking 'parking lot/garage/place' Constraints

—»

SYM

Output Candidates ψ paRkiji PARKIJIG paRkin

1 I 1 He | Ψ

ι

*1* *!

FILL

PARSE

I * IBllgiiiiiiiiM ' · ·/· | * ' • • J ·"· - , '· · Hi .. . IHPWP·· * in i : : • • : : •

!

*

PRENAS

Τ

|

PARKIJIG Ι®"

PARKIQ

1 1 1 1

*!

*

*

Quantitative studies, such as the one conducted by Walter (1976), agree that the velar nasal obtains in approximately 77% of the observed cases (Walter 1983). The SYM constraint helps explain why the non-velar candidates are disqualified in most cases: Only the inclusion of velar forms help redress the imbalance displayed in Table 1. A double asterisk appears in the SYM column for the first two candidates in Tableau 5 because not only do they retain the asymmetrical palatal nasal, but they fail to exploit an appropriate opportunity to include the symmetrical velar nasal. The third candidate has only a solitary asterisk because, though it also fails to exploit the opportunity to include the velar nasal, the replacement segment arrived at, the dental nasal, is symmetrical in the overall phonological system. At first glance, it would appear that SYM is not needed in Tableau 5 to arrive at the correct result: In the absence of SYM, two other constraints, PARSE and FILL, would have still disqualified all but the correct result. However, Tableau 5 does not exist in a vacuum: It was previously determined in another case, as shown in Tableau

Meta-constraints and constraint ranking

295

2, that SYM is a necessary constraint in the French of France regardless of the relative rankings of PARSE, FILL and PRENAS and that SYM held a rank superior to FILL and PRENAS, regardless of the order of the latter two constraints. Moreover, the obligatory (based on Tableau 2) inclusion of SYM in Tableau 5 proves to be quite natural to the result and helps explain the adoption of a new consonant into the phonological inventory. Notice, nevertheless, that Tableau 5 now allows us to determine with certainty the relative ranking of FILL and PRENAS, which was elusive in Tableau 2. FILL must be ranked higher than PRENAS in Tableau 5 or else the last two candidates will tie. (Notice also that, in Tableau 5, taken alone, SYM and PARSE could exchange rankings with each other and with FILL without disturbing the result; but it was previously determined in Tableau 2 that SYM and PARSE, while free-floating relative to each other, must both be ranked more highly than FILL.) At this juncture, let it be pointed out that it is perfectly consistent that the disarticulation of the features of the palatal nasal (as displayed in Tableau 2) and the adoption of the velar nasal (as displayed in Tableau 5) are happening simultaneously in the French of France, for both phenomena are being driven by SYM in an effort to redress the same phonological imbalance.

5.

Proposing a family of equilibrium constraints: SYM and

MAXCON

Should it be argued that SYM favors or disfavors a shift from the palatal nasal to the velar nasal in the case of Quebec French (cf. Gendron 1966:115)? Since velar obstruents exist in Quebec French, one is tempted to invoke SYM to justify the inclusion of the velar nasal, as in the case of the French of France (Tableau 5). To do so, however, by virtue of the categorical conversion of the palatal nasal would simultaneously go counter to SYM if palatal obstruents are also present in the inventory, which is in fact the case for Quebec French. But SYM can still favor inclusion of the velar nasal in certain positions and/or from some other source such as English borrowings. In fact, for Quebec French, not only are such borrowings present but Walker (1984: 115) reports a strong tendency to convert underlying /ji/ to [η] preconsonantally and in the word-final position.6 For example the intervocalic palatal nasal in ligner is realized as a velar nasal word-finally in ligne and preconsonantally in enlignement. As such, the velar nasal arguably constitutes an allophone of the palatal nasal. It is interesting to consider, at this juncture, what might militate for and against the ascension of the velar nasal and the palatal nasal to co-phonemic

296

Michael D. Picone

status. If the simultaneous presence of the palatal and velar nasals in a given inventory is demonstrated to be universally unstable in the world's languages, perhaps because of insufficient phonological space in the posterior region (see Section 1 of this essay), then another meta-constraint such as MAXCON (for "maximal contrastivity") might be invoked to explain this phenomenon. SYM and MAXCON would work in tandem, much as the family of faithfulness constraints do, accounting for the types of language change that Martinet has delineated (1955; see Section 1 of this essay) but within the Optimality-Theoretic framework. In many instances SYM and M A X C O N requirements would align to strongly enforce certain configurations in phonological systems: for example, in maintaining the primacy of the near-universa! vocalic triangle, shown in Figure 1,

Figure 1. The vocalic triangle

In other cases, however, such as the one presently under discussion, different rankings for SYM and MAXCON would determine whether or not posterior palatal obstruents and/or nasals can be maintained as phonemes simultaneous to velars - the two sets being in close proximity - in a given language or dialect. SYM and M A X C O N , and perhaps others, could be conceived of as a family of "equilibrium constraints" related to cognitive requirements associated with the production and perception of human language. This can be compared to the presumed cognitive value of the faithfulness constraints for the promotion of transparency in relation to efficient language acquisition.7

6. Concluding remarks In the foregoing, it has been shown that a modified Optimality-Theoretic approach can be useful in understanding dialectal diversity with regard to the disarticulation of nasality and palatality across French dialects, and also with regard to the resurgence of the velar nasal. To achieve this, the most impor-

Meta-constraints and constraint ranking

297

tant modification proposed in this essay consists of the introduction of metaconstraints which are sensitive to the entire phonological system, thereby having the potential for converting the overall apparatus to a three-variable model. In addition to the two previously established variables, which are input and constraint ranking, the overall state of the phonological system is itself empowered de facto to determine aspects of the output via systemsensitive meta-constraints which will sort output candidates differently for different languages and dialects even when input and constraint rankings are identical. The three-variable model carries with it an explanatory advantage since it confers upon the overall theory the kind of flexibility that is necessary to account for certain types of language change. This was demonstrated in the foregoing for cases where language change (or conservatism) revolves around phonological equilibrium. It was suggested that this equilibrium is not fortuitous but is related to the exigencies of human cognition and perception. The explanatory advantage of the three-variable model may extend even further if, in addition to allowing for the existence of system-sensitive, equilibrium-related meta-constraints such as SYM (for "symmetry sensitive") and M A X C O N (for "maximal contrast"), the three-variable model can also be used to incorporate sociolinguistically and pragmatically-sensitive constraints and possibly others as well. For example, by introducing constraints that make the sorting apparatus sensitive to socially weighted factors, the model can be made to account for variation between speakers and registers. Some scholars, who hope to maintain a sharp distinction between the operations of an hypothesized Universal Grammar and various overarching cognitive and socio-pragmatic influences, may object to the expansion of Optimality Theory to embrace these latter elements. Rejection of such elements, however, is not entirely in keeping with the prior acceptance of the faithfulness constraints, which share some properties with the equilibrium constraints herein proposed. Furthermore, the inclusion of such elements does not challenge per se the integrity of Universal Grammar, it simply provides for a plausible interface with other elements and, by virtue of this, makes possible a more powerful conceptual model that connects directly with language variation and change across dialects.

298

Michael D. Picone

Notes 1. Grateful recognition is extended to the Research Grants Committee of the University of Alabama for research funding which made possible field work in Louisiana that contributed to this study. I am also indebted to Yolanda Rivera-Castillo, Pieter van Reenen, Jean Leo Leonard, Norval Smith, and Helga Humbert for comments, criticisms, and encouragement following earlier presentations of this paper. The usual disclaimers apply concerning any errors or inaccuracies. 2. Symmetry is a ubiquitous human trait at a number of levels. It is an innate property of countless biological systems, including human perceptual systems related to cognition (see Note 7). Consequently, its presence in linguistic systems is by no means anomalous, even if Martinet's explanation of it is not retained and its linguistic function remains unclear. Furthermore, humankind in virtually every culture finds symmetry utilitarian and aesthetically pleasing, which is demonstrated systematically in the production of symmetry-constrained artistic, cultural, and technical artifacts. 3. The various subdialects of French now spoken in Louisiana are often referred to collectively as "Cajun French" {Acadian > Cajun). Because of other important non-Acadian dialectal sources and also because of historic non-Acadian ethnic populations using French in Louisiana (namely, African American, American Indian, Quebecois and Metropolitan French; see Picone 1997), some scholars prefer the more neutral term "Louisiana French." This discrepancy in terminology is readily evident among those contributing chapters to the recent collection edited by Valdman (1997). 4. Alternative representations of the faithfulness constraints are possible (see for example Archangeli 1997 and Pulleybank 1997), but their function remains unchanged. 5. Other constraints not shown here would have to come into play to account for the preservation of [p] and the avoidance of [nj] in word-final and preconsonantal positions. Restrictions on permissible consonant clusters are obvious and likely possibilities. 6. Curiously, Gendron (1966: 115) reports no such tendency twenty years earlier. In fact, he sees little evidence of velarization of the palatal nasal in popular French Canadian pronunciation. 7. Consider, for example, how symmetry and contrastivity work in tandem in other perceptual systems such as sight and hearing: These make constant use of perceptual parallax by virtue of, respectively, stereoscopic and stereophonic reception of identical signals that are contrasted in order to make essential judgments about distance and directionality in relation to the source of the signal (see Note 2).

References Aitchison, Jean 1991 Language Change: Progress or Decay? [2nd ed.] Cambridge: Cambridge University Press. Archangeli, Diana 1997 Optimality theory: An introduction to linguistics in the 1990s. In: Diana Archangeli and D. Terence Langendoen (eds.), Optimality Theory: An Overview, 1-32. Oxford: Blackwell.

Meta-constraints and constraint ranking

299

Archangeli, Diana and D. Terence Langendoen (eds.) 1997 Optimality Theory: An Overview. Oxford: Blackwell Bourciez, Edouard and Jean Bourciez 1982 Phonetique franqaise, Etude historique. Paris: Klincksieck. Bream, Carol 1968 La nasalisation des voyelles orales suivies de consonnes nasales dans le franijais et l'anglais paries au Canada. In: Pierre R. Leon (ed.), Recherches sur la structure phonique du franqais canadien, 100-118. Montr6al: Marcel Didier Conwell, Marilyn and Alphonse Juilland 1963 Louisiana French Grammar, Volume 1: Phonology, Morphology and Syntax. The Hague: Mouton. Gendom, Jean-Denis 1966 Tendancesphonetiques du franqaisparle au Canada. Paris: Klincksieck. Itö, Junko, Armin Mester, and Jaye Padgett 1995 Licensing and underspecification in optimality theory. Linguistic Inquiry 26: 571-613. Labov, William 1994 Principles of Linguistic Change, Volume I: Internal Factors (Language in Society 20.). Oxford: Blackwell. Martinet, Andre 1955 Economie des changements phonetiques. Traite de phonologie diachronique. Bern: A. Francke. Papen, Robert A. and Kevin J. Rottet 1997 A structural sketch of the Cajun French spoken in Lafourche and Terrebonne parishes. In: Albert Valdman (ed.), French and Creole in Louisiana, 71-108. New York: Plenum Press. Picone, Michael D. 1991 The phonology of the velar nasal in contemporary metropolitan French. The SECOL Review 15: 121-145. 1996 Anglicisms, Neologisms and Dynamic French. (Lingvisticae Investigationes Supplementa 18.) Amsterdam: John Benjamins. 1997 Enclave dialect contraction: An external overview of Louisiana French. American Speech 72: 117-153. Prince, Alan S. and Paul Smolensky 1993 Optimality theory: Constraint interaction in generative grammar. (Rutgers University Center for Cognitive Science Technical Reports 2.) New Brunswick, N.J.: Rutgers University. Pulleybank, Douglas 1997 Optimality theory and features. In: Diana Archangeli and D. Terence Langendoen (eds.), Optimality Theory: An Overview, 59-101. Oxford: Blackwell. Valdman, Albert (ed.) 1997 French and Creole in Louisiana. New York: Plenum Press Walker, Douglas C. 1984 The Pronunciation of Canadian French. Ottawa: University of Ottawa Press.

300

Michael D. Picone

Walter, Henriette 1976 La dynamique des phonemes dans le lexique frangais contemporain. Paris: France Expansion. 1983 La nasale v61aire /η/, un phonSme du franfais? Languefranqaise 60: 14-29.

Probleme bei der Erstellung von Sprachkarten im Untersuchungsgebiet des "Bayerischen Sprachatlas" am Beispiel von Richtungsadverbien Sibylle Reichel

Gegenstand dieses Beitrags sollen erste Auswertungsversuche des Materials zu den Richtungs- und Lageadverbien im derzeit entstehenden Bayerischen Sprachatlas (BSA) sein. Dabei möchte ich zum einen auf die Erhebungs- und Auswertungsmethode der einzelnen Teilprojekte eingehen, zum anderen möchte ich auffällige Erscheinungen bei Lokaladverbien ansprechen. Der bayerische Sprachatlas besteht aus sechs Teilprojekten, die zusammen die Fläche des Freistaates Bayern abdecken: -

SBS: Sprachatlas von Bayerisch-Schwaben in Augsburg SNOB: Sprachatlas von Nordostbayern (Oberfranken und Oberpfalz) in Bayreuth - SMF: Sprachatlas von Mittelfranken in Erlangen - SUF: Sprachatlas von Unterfranken in Würzburg SNIB: Sprachatlas von Niederbayern in Passau - SOB: Sprachatlas von Oberbayern in Passau Diese sechs Teilprojekte stehen miteinander in Verbindung, verwenden die gleichen Erhebungsmethoden und zum Teil die gleichen Auswertungswerkzeuge. Über das Untersuchungsgebiet wurde ein Planquadratraster mit 7 km Kantenlänge gelegt. In jedem der dadurch entstehenden etwa 1500 Planquadrate wird eine Dialektbefragung nach dem Vorbild des Sprachatlas der deutschen Schweiz (SDS) durchgeführt. Ein Fragebuch umfaßt etwa 2500 Fragen zu Phonologie, Lexik, Morphologie und Syntax und wird in direkter Befragungsmethode von Exploratoren vor Ort aufgenommen. Als Lautschrift dient die Teuthonista. Die in den einzelnen Teilprojekten verwendeten Fragebücher sind nahezu identisch. Die Daten werden in den meisten Projekten durch die Eingabe in ein Textverarbeitungssystem (ζ. B. WordPerfect) mittels eines detaillierten Kodierungsplanes, der die Diakritika und Grundzeichen in alphanumerische

302

Sibylle Reichel

Zeichen "übersetzt", weiterverarbeitet. Hierbei bleibt zunächst die Struktur der Fragebücher erhalten. Jede Ortserhebung ergibt also eine eigene, später in Äsen1 konvertierte - und damit systemunabhängig lesbare - Datei. Der nächste Schritt besteht im Einlesen der Daten in eine Datenbank. Um die Textdaten in ein entsprechendes Datenbankformat zu bringen, stehen verschiedene Programme zur Verfügung, die in Augsburg, und teilweise schon in Würzburg und Erlangen verwendet werden. Nachdem die Kodate in die Datenbank eingelesen sind, können sie nach verschiedenen Gesichtspunkten sortiert werden. So kann beispielsweise eine Belegliste zu einer beliebigen Frage für alle eingelesenen Orte schnell ausgegeben werden. Es besteht auch die Möglichkeit, mit weiteren Programmen, die in Augsburg bereits zum Einsatz kommen, noch im Datenbankformat die einzelnen Belege mit Symbolen zu verknüpfen und auf einer Grundkarte in Form von Sprachkarten ausdrucken zu lassen.2 Da jedoch die unterschiedlichen Teilprojekte des bsa unterschiedlich weit fortgeschritten sind und demnach auch die Daten in den verschiedensten Formaten vorliegen, konnte ich die ersten Auswertungsversuche nicht auf die fur einen späteren Zeitpunkt geplante, komfortablere Weise vornehmen. Die Fragen zu den Ortsadverbien sind in den Fragebüchern auf jeweils vier bis fünf Seiten zu finden. Es liegen mir bisher etwa 1200 kopierte Fragebuchseiten aus ca. 380 Erhebungen des Sprachatlas von Nordostbayern, 350 kopierte Fragebuchseiten aus ca. 100 Erhebungen des Sprachatlas von Niederbayern, 80 kodierte Dateien des Sprachatlas von Unterfranken, 100 kodierte Dateien des Sprachatlas von Oberbayern, 35 Beleglisten im Datenbankformat zu den 360 Orten des Sprachatlas von Bayerisch-Schwaben und sowohl kodierte als auch kopierte Daten des Sprachatlas von Mittelfranken aus 140 Belegorten vor. Insgesamt also Belege in divergierenden Formaten aus etwa 1100 Erhebungsorten. Diese unterschiedlichen Dateiformate mussten zunächst auf einen Nenner gebracht werden. Dazu erstellte ich handschriftliche Beleglisten, übertrug diese mit Symbolen in ein Planquadratraster einer Tabellenkalkulation und erhielt so einfache Karten für einen ersten Überblick über das Untersuchungsgebiet. Bei der Untersuchung der Ortsadverbien bin ich zunächst von dem Teilprojekt ausgegangen, an dem ich selbst mitarbeite, dem Sprachatlas von Mittelfranken. Dort konnten die Erhebungen im September 1997 zum Abschluß gebracht werden. Mir stehen aus diesem Gebiet die Daten vollständig als handschriftliche Fragebücher und zum Großteil zusätzlich in maschinenlesbarer Form zur Verfügung.

Probleme bei der Erstellung von Sprachkarten

303

Im Untersuchungsgebiet des Sprachatlas von Mittelfranken ist die zentrale Mundart das Oberostfränkische. Daneben existieren noch weitere Hauptmundarten und Übergangsgebiete. Insgesamt vier größere Dialektgruppen sind hier vertreten. Im Nordosten ist der bairische Sprachraum zu nennen, der mit dem Nürnberger Übergangsstreifen in manchen Bereichen bis Nürnberg und darüber hinaus reicht; im Nordwesten befindet sich der Würzburger Übergangsraum, mit der Frankenwald-Steigerwaldschranke im Westen und der Höchstädter Schranke im Osten, der das Oberostfränkische vom Unterostfränkischen abgrenzt. Die Hauptmundartlinie zum schwäbisch-alemannischen Dialektgebiet verläuft im Südwesten des Untersuchungsgebiets (vgl. Steger 1968: 398; Kranzmeyer 1956). Karte 1

Dialekte in Mittel franken

Inwieweit diese Dialektgebiete mit den unterschiedlichen Bildungsmustern von Lokaladverbien zusammenhängen, möchte ich untersuchen. Zur Kategorisierung des Feldes der Ortsadverbien liefert die Theorie der Deixis (nach Diewald 1991), insbesondere die Betrachtung der lokalen Dimension, einen guten Ansatzpunkt, um ein differenziertes Konzept der Systeme dieser "Zeigewörter" zu entwickeln.

304

Sibylle Reichel

Simplizische Lokaladverbien sind entweder starke Deiktika ohne zusätzliche nennende Seme {da, dort) oder Nennwörter ohne deiktische Elemente {außen, innen), die eine Entität in der Umwelt benennen und keinen Bezug zur Origo haben. Die Semantik der Nennwörter bezieht sich auf den Anfangspunkt (bei ab), die Verbindung von Anfangs- und Zielpunkt (bei über) oder den Zielpunkt (bei ein) einer Relation. Werden diese intrinsischen (vgl. Diewald 1991: 90), in reiner Nennfunktion gebrauchten Wörter mit einem reflexiven Bezug ausgestattet, sind das zweigliedrige Deiktika. Richtungs- und Lageadverbien bestehen demnach aus einerzeigenden (deiktischen) und einer nennenden (präpositionalen) Komponente. Der reflexive Bezug entsteht durch die Verbindung der Semantik einer präpositionalen Relation mit einem Morphem, was den deiktischen Prozeß bewirkt (z.B. /dr-/ + /innen/, /h-/ + / innen/). Je nach Art des Morphems kann primär zwischen einer origoinklusiven und einer origoexklusiven Variante der Relation unterschieden werden. Dabei kann dieses Wortbildungsmorphem im Dialekt nicht nur als Präfix, sondern auch als Suffix auftreten (ζ. B. aeni für, hinein').

Probleme bei der Erstellung von Sprachkarten 305 Bei der folgenden Betrachtung der Richtungs- und Lageadverbien im Untersuchungsgebiet greife ich hier exemplarisch die Gruppe um den präpositionalen Gehalt von ein/in auf. Dabei differenziere ich hier nicht die feinen Lautunterschiede, sondern lediglich die Morphologie der Adverbien. Bei den Richtungsadverbien, also den dynamischen deiktisch verwendeten Lokaladverbien erhalten wir so im Untersuchungsgebiet des Sprachatlas von Mittelfranken eine deutliche Grenze zwischen dem präfigierten Typ und dem suffigierten Typ. Beim Betranchten von Karte 1 kann auch vom bairischen (suffigierten) und nicht-bairischen (präfigierten) Typ gesprochen werden. Die anderen Dialekträume weichen hierbei nicht voneinander ab. Das gleiche Bild zeigt die Karte des entgegengesetzten Richtungsadverbs hinein. Entgegengesetzt meint hier, daß der Sprecher im ersten Fall den Adressaten zu sich her holt, während im zweiten Fall der Adressat vom Sprecher zum Ziel hingeschickt wird, beide Male ist der Endpunkt in einem umgebenden Hindernis. Das Ziel der Bewegung ist also durch die Relation innen bestimmt. Karte i :

„dort innen" SMF 15.VI. SBS 308/5

306

Sibylle Reichel

Bei der Betrachtung der Lageadverbien zu dieser Relation, also den statischen, deiktisch verwendeten Lokaladverbien zu innen, fallt im Untersuchungsgebiet bei der origoexklusiven Form dort innen eine deutliche NordSüd-Teilung auf. Während im nordöstlichen Teil Mittelfrankens die origoexklusive Position durch den Typ ausgedrückt wird, herrscht im südwestlichen Teil der Typ (ohne r) vor. Kombiniert man diese Karte mit der Karte zur entgegengesetzten Perspektive, also der origoinklusiven Variante hier innen, zeigen sich Räume in denen die Sprecherperspektive unterschiedlich realisiert wird. Karte 4: „hier innen - dort innen" als Kombinationskarte der Einzelknrlen SMF 155/1, SBS 308:5 undSMF 154.13, SBS 308/4

Wir sehen nun weitestgehend die bekannten Dialekträume. An den Grenzen zum bairischen Dialektgebiet gibt es allerdings mehrere Varianten (da innen / drinnen und rin / drin). Daher sollte das Betrachtungsgebiet ausgeweitet werden, um diese Erscheinung genauer zu untersuchen. Die folgende Karte zeigt die Mundartgrenzen im gesamtbayrischen Untersuchungsgebiet, wie sie nach den Daten des Deutschen Sprachatlas (DSA) verlaufen:

Probleme bei der Erstellung von Sprachkarten

307

Kane 5:

Dialekte in Bayern gemüß

Deutscher Sprachatlas

Die Auswertung des Materials zu den Richtungsadverbien aus den Fragebüchern aller Teilprojekte des Bayrischen Sprachatlas gibt weiteren Aufschluß über den Zusammenhang mit den Dialektgrenzen. Zunächst zeigt sich, daß sich die bei den Richtungsadverbien in Mittelfranken ermittelte Isoglosse entlang der Grenze des bairischen Sprachraumes zum fränkischen, bzw. alemannisch-schwäbischen Raum fortsetzt. Auch im größeren Untersuchungsgebiet sind die anderen Sprachräume nicht durch die Morphologie der Richtungsadverbien voneinander abzugrenzen.

308

Sibylle Reichel

Kart« 6: „herein" und „hinein" SMF 153/1. SBS 304/1 undSMF 153/7, SBS 304/7

Die Lageadverbien erscheinen auch hier komplexer. Bei der origoexklusiven Variante kann eine deutliche Unterscheidung in zwei Wortbildungstypen festgestellt werden. Im gesamten bairischen Gebiet und einem Teil des Unterostfränkischen, bzw. dem Nürnberger Übergangsraum und darüber hinaus herrscht der Typ 3 vor, in den übrigen Gebieten dagegen .4 Setzt man diese Karte in Bezug zu der entgegengesetzten hier innenKarte, entstehen fünf Gebiete, in denen die Opposition der Sprecherperspektive unterschiedlich realisiert wird. Dabei fällt auf, daß die Unterscheidung der Perspektive immer vorhanden ist. In der Standardsprache kann drinnen beide Perspektiven ausdrücken (ζ. B. ein Sprecher im Garten fragt: "Bist du im Haus drinnen?" - ein zweiter Sprecher im Haus antwortet: "Ja,

Probleme bei der Erstellung von Sprachkarten 309

Karte 7:

„dort innen"

SMF 155/1, SBS 308/5

ich bin im Haus drinnen"). Im Dialekt wird differenziert. Auch wenn drinnen beide Seiten ausdrücken kann (beim Sprecher im hina - drina Gebiet und vom Sprecher abgewandt im bairischen Gebiet), so doch nicht im selben Dialektgebiet. Es gibt immer einen, bezogen auf die Origo entgegengesetzten Partner, der auch eine andere Morphologie hat. (1)

Hochdeutsch: drinnen im Zimmer (Duden 1989: 366) (ohne Unterscheidung der Sprecherperspektive)

310

Sibylle Reichel

(2) Bairisches Gebiet: drina im Zimmer (Sprecher ist nicht innen) herina im Zimmer (Sprecher ist innen) (aus den Daten des Bayerischen Sprachatlas) Karte 8:

„hier innen" - „dort innen" als Kombinationxkartc der Ii in/elkurtoi SMP 155/1, SBS .508/5 und S.V1F 154/13, SBS 308/4

Typ: hina - dina (Unterfranken und ein Teil Oberfrankens) Typ: herinnen - drinnen (bairisches Gebiet: Oberiranken, Oberpfalz, Nieder- und Oberbayern) Typ: drina - dina (südliches Mittelfranken und nördliches Schwaben) Typ: da ina - drina (nördliches Mittelfranken)5 Typ: hina - drina (südliches Schwaben)

Probleme bei der Erstellung von Sprachkarten 311 Der "Typ": rin - drin läßt sich nun auch als Lautvariante zu (he)rinnendrinnen erklären, der sich wiederum dem bairischen Typ unterordnen läßt. Da eine ähnliche lautliche Variante im Zentrum des drina - dina -Gebietes auftaucht, muß hier wohl ein Ausfall des anlautenden d- zu der Form rin gefuhrt haben. Da diese Formen nur ausdrucksseitig identisch sind, sind sie nicht dem gleichen Lokaladverbientyp zuzurechnen. Man kann festhalten, daß die Lokaladverbien nicht isoliert betrachtet werden sollten, sondern immer paarweise mit ihrem bezüglich der Sprecherperspektive zu interpretierenden Pendant. Die Sprachräume, die sich dadurch ergeben, sind differenzierter, als dies Einzelwortkarten erkennen lassen. Es wird deutlich, daß in der Mundart die Opposition origoinklusiv-origoexklusiv ausdrucksseitig getrennt ist. Da im Dialekt, einer (nur) gesprochenen Sprache, Deiktika eine wichtigere Rolle spielen als in der Standardsprache, sind diese Differenzierungen stärker ausgeprägt und im Sprachbewußtsein der Sprecher auch vorhanden.6 Dieser Aufsatz behandelt nur einen kleinen Ausschnitt des gesamten Systems von Richtungsadverbien, die idealerweise in ihrem präpositionalen Gehalt mit allen möglichen Relationen untersucht werden müssen. Auch das Datenmaterial war zum Zeitpunkt der Auswertung noch unvollständig. Die Pilotauswertung hat jedoch gezeigt, daß es nicht nur im lautlichen Bereich möglich ist, Sprachgrenzen sichtbar zu machen, sondern daß auch die unterschiedlichen Systeme der Bildung von Lokaladverbien zur Typisierung von Dialekten geeignet sind.

Anmerkungen 1.

2. 3. 4. 5. 6.

American Standard Code for Information Interchange ist eine Programmsammlung, die zur Kartenproduktion in den kürzlich erschienenen ersten Bänden des Bayerischen Sprachatlas verwendet wurde. wohl aus: wohl aus: Hierbei könnte es sich um eine Grenzerscheinung, bzw. eine Verschiebung zu einer hochdeutschen oder nicht-deiktischen Variante durch Interferenz handeln. Die Gewährsperson 1 in EbersdorfTMittelfranken sagte beispielsweise bei der Frage nach dem Satz: Ersitzt hier innen (154/13) "Man kann sagen drin oder din\ Das r ist dann drin, wenn man selbst dabei ist". ASCII:

AUGUSTAPLOT

312

Sibylle

Reichel

Literatur Diewald, Gabriele 1991 Deixis und Textsorten im Deutschen: (Reihe germanistische Linguistik 118) Tübingen. Drosdowski, Günther (ed.) 19892 Duden, Deutsches Universalwörterbuch. Mannheim, Wien, Zürich. Feik, Christiane und Werner König (eds.) 1996 Sprachatlas von Bayerisch-Schwaben, Band ii: Wortgeographiel, Der menschliche Körper, körperliche und seelische Äußerungen, die menschliche Gemeinschaft, Kleidung. (Bayerischer Sprachatlas, Regionalteil 1) Heidelberg. Ihle, Sabine und Werner König (eds.) 1997 Sprachatlas von Bayerisch-Schwaben, Band i: Einführung. (Bayerischer Sprachatlas, Regionalteil 1) Heidelberg. Kranzmayer, Eberhard 1956 Historische Lautgeographie des gesamtbayerischen Dialektraumes. Wien. Renn, Manfred und Werner König (eds.) 1997 Sprachatlas von Bayerisch-Schwaben, Band ra: Lautgeographie i: Ehemalige Kurzvokale. (Bayerischer Sprachatlas, Regionalteil 1) Heidelberg. Steger, Hugo 1968 Sprachraumbildung und Landesgeschichte im östlichen Franken. Neustadt/ Aisch.

Abbildungsverzeichnis Karte 1 und 5: nachgezeichnet in Anlehnung an Steger 1968, sowie Kranzmayer 1956. Karte 2-4 und 6-8: erstellt nach den Daten der Teilprojekte des Bayerischen Sprachatlas.

Dialektwandel und Einstellung Das Beispiel der Aarauer Stadtmundart Dialect change and attitudes

Beat Siebenhaar

Abstract Sociolinguistic research has shown that attitudes towards linguistic variants can distinguish different speech communities. The importance of attitudes for an explanation of linguistic change was examined and compared to traditional explanations by sociolinguistic and dialectologic variables. Therefore the dialect of Aarau was investigated, a small town situated between the two cities of Bern (80 km in the west) and Zürich (50 km in the east) in the German speaking part of Switzerland. Bern and Zürich both are centres of a larger dialect region, Aarau lies in the contact zone of these two dialects. Phonetic variables of the idiolect of 55 speakers were compared to historical data and related to their attitudes towards the neighbouring dialects. The findings so far show no significant correlation of attitudes and language change, but further research including morphology will refine the results. The inclusion of attitudes to explain linguistic change can complement the understanding of linguistic change, but it can not explain it.

0. Einleitung In der Sprachwandeltheorie von der "unsichtbaren Hand in der Sprache" hat Keller (1994) dargelegt, dass die Sprache sich durch deren Verwendung ändert. Die Sprachteilnehmer wollen sich erfolgreich mitteilen, und weil das die meisten infolge sozialer Erwartungen und der Vorwegnahme der sozialen Erwartungen anderer auf dieselbe Art tun, ändert sich die Sprache, ohne dass die Sprachteilnehmer sie bewusst ändern. Diesem Ansatz entsprechen implizit auch soziolinguistische Erklärungen zum Sprachwandel, die davon ausgehen, dass sich eine Sprache dadurch ändert, dass die Sprecherinnen mit sprachlichen Zeichen nicht nur Inhalte senden und empfangen, sondern auch - und zwar meist unbewusst - Hinweise auf außersprachliche Fakten, wie

314

Beat Siebenhaar

zum Beispiel ihre soziale Stellung.1 Sprache hat aber nicht nur eine vertikale, soziale Komponente, sondern auch eine horizontale, sprachgeographische. Auch die Sprachgeographie bringt Erklärungen für den Sprachwandel, denen zufolge sich eine Varietät dadurch ändert, dass Elemente eines benachbarten Sprachsystems aufgenommen werden.2 In diesem Punkt decken sich traditionell soziolinguistische und sprachgeographische Erklärungen: in beiden Modellen werden Elemente eines partiell verschiedenen Sprachsystems in das eigene Sprachsystem aufgenommen. Damit das geschehen kann, braucht es einerseits den Kontakt der beiden Varietäten, bzw. deren Sprecherinnen, und andererseits eine Ursache für die Übernahme. Soziolinguistische und dialektologische Erklärungen gehen dabei beide meist von einem höheren Prestige der gebenden Varietät aus, was dazu führt, dass deren Varianten die entsprechenden Varianten der prestigeärmeren empfangenden Varietät verdrängen. An dieser Stelle setzt meine Untersuchung an: Die soziolinguistische Forschung hat gezeigt, dass die Einstellung zu einzelnen Varietäten Sprechergruppen voneinander unterscheidet.3 Die Bedeutung der Einstellung zur Erklärung von Sprachwandel soll geprüft, das Konzept des höheren Prestiges hinterfragt und differenziert werden. Dabei sollen die traditionell soziolinguistischen und dialektologischen Fragen keineswegs außer Acht gelassen werden, sondern ergänzt werden durch die Untersuchung der Einstellung.

1. Forschungsdesign Um diesen Aspekten von Sprachwandel nachzugehen wird eine Ortsmundart in ihren Varietäten und in deren Abweichen von historischen Daten untersucht. Exemplarisch wurde die Mundart der Kleinstadt Aarau im deutschschweizerischen Mittelland ausgewählt. Aarau liegt, mit öffentlichen und privaten Verkehrsmitteln gut erschlossen, an der Hauptverkehrsachse zwischen Zürich und Bern. Hotzenköcherle (1984: 79) hat diese Region im breiten Grenzbereich der größeren Mundarträume des Bemdeutschen im Westen und des Zürichdeutschen im Osten als Labilitätszone bezeichnet. Diese sprachgeographische Einbettung lässt zu, dass die Frage nach Übernahmen aus benachbarten Varietäten reduziert werden kann auf den Einfluss aus östlichen oder westlichen Mundarten. Die Varianten in der Mundart von Aarau werden in Bezug gesetzt zu denjenigen der Nachbarvarietäten Berndeutsch und Zürichdeutsch. Als zweites wird die Varianz mit extralinguistischen Faktoren korreliert, und damit werden die Umstände bestimmt,

Dialektwandel und Einstellung

315

welche die Varianz und die Übernahme aus den Nachbarvarietäten beeinflussen. Dann wird die Einstellung der Gewährspersonen zu den Nachbardialekten mit ihrem eigenen Idiolekt in Verbindung gebracht. Ziel ist es, Einstellungsdifferenzen zu zeigen zwischen Gewährsleuten, die sich eher an den Berner Großraumdialekt anlehnen, und solchen, die sich eher dem Zürcher Großraumdialekt angenähert haben. Damit soll geprüft werden, ob ein Zusammenhang zwischen Einstellung und der Sprachverwendung der Gewährsleute besteht, und somit, ob ein Zusammenhang zwischen Einstellung und Sprachwandel besteht.

2. Datenerhebung und Sample 2.1. Die Erhebungsdaten Aufgrund der wichtigsten lautlichen Ost-West-Unterschiede, die sich aus der Durchsicht der ersten drei Bände des Sprachatlas der deutschen Schweiz (SDS 1962 ff.) ergaben, wurde ein zweiteiliger Fragebogen erstellt. Einerseits wurden einige Bilder gezeichnet, die zu kommentieren waren, andererseits wurde ein standardsprachlicher Text zur Übersetzung zusammengestellt.4 Für die Einstellung zu den Nachbardialekten wurde ein semantisches Differential mit 16 Variablen erstellt. Diese ergaben sich aus der Durchsicht der Literatur zur Einstellungsuntersuchung und aus einem Vortest, aus dem stark korrelierende Variablen herausgenommen wurden. Der Fragebogen findet sich im Anhang. Da auch außersprachliche Daten in die Analyse einbezogen werden sollten, wurden die biosozialen Daten der Gewährsleute erhoben.

2.2. Gewährspersonen und Erhebungsmerkmale Befragt wurden 55 Gewährspersonen, die seit mindestens 10 Jahren in Aarau wohnhaft sind; unterschieden wurde nach folgenden Merkmalen (in Klammer: Merkmalskriterien und die in den Tabellen und Abbildungen verwendeten Abkürzungen). 1. Eingesessenheit (alteingesessen / eingesessen / zugezogen - AE / Ε / Z) 5 2. Generation (ältere / mittlere / jüngere Generation - äG / mG / jG) 3. Kontakt zu Nachbardialektgebieten (nach Osten / nach Aarau / nach Westen-O/A/W)6

316

Beat Siebenhaar

4. Bildung (obere / untere Bildungskategorie - ο / u)7 5. Geschlecht (männlich / weiblich - m / w)

3. Analyse der Sprachdaten An dieser Stelle soll nur das Analyseverfahren präsentiert werden, ohne auf einzelne Lautveränderungen einzugehen. Um einen Überblick über die Vertretung eines mittelhochdeutschen Lautes in allen Belegwörtern zu bekommen wurde jeweils ein Index erstellt: fur jede östliche Realisierung wurde eine 1 gesetzt und für jede westliche eine 0; Abweichungen und Sonderlautungen wurden nicht berücksichtigt. Der Mittelwert ergab dann für jeden untersuchten Aspekt einen Index, der die Positionierung der Gewährsperson angibt.8 In einem weiteren Schritt wurde dieser Index mit den verschiedenen extralinguistischen Faktoren in Zusammenhang gebracht. Dies geschah mittels einer statistischen Varianzanalyse. Die Varianzanalyse bestimmt die Signifikanz von einzelnen Faktoren in einem Modell, indem errechnet wird, wie stark die Variabilität der Sprachdaten erklärt werden kann durch die Ausprägung dieser außersprachlichen Faktoren. Die signifikanten Unterschiede werden dann weiter kommentiert. Die Darstellung der Elemente im Wandel kann überblicksmäßig anhand der Tabelle 1 erfolgen. Sie zeigt, dass sich Varianz vor allem im Bereich des Vokalismus findet, während der Konsonantismus mit Ausnahme der /-Vokalisierung relativ stabil ist. Aus der Tabelle ist auch ersichtlich, welche außersprachlichen Faktoren den Sprachgebrauch signifikant beeinflussen. Die dargestellten Unterschiede sind jedoch in allen Fällen nur quantitativ, nicht qualitativ.9 Der Faktor Geschlecht begründet nie sprachliche Unterschiede, er ist deshalb in Tabelle 1 auch nicht aufgeführt. Ebenfalls wenig von Bedeutung ist die Schulbildung. In den Fällen, wo sie als Erklärung für die Variation beigezogen werden kann, lässt sich feststellen, dass die untere Bildungsschicht meist den konservativeren Lautstand vertritt. Als bestimmend für die Variation haben sich die Faktoren Eingesessenheit, Alter und Kontakt zu den Nachbargebieten herausgestellt: Wie zu erwarten, repräsentieren die Alteingesessenen den konservativsten Lautstand, die Eingesessenen nehmen mit einer leicht höheren Streuung eine modernere Position ein, während die Zugezogenen die größte Varianz erkennen lassen. Die ältere Generation hält größtenteils an den tradierten Formen fest, während die jüngere Generation in vielen Fällen einen höheren Anteil Abweichungen zeigt. Die mittlere

Dialektwandel

und Einstellung

317

Generation nimmt meist eine Mittelposition ein. Bei der Qualität der Kurzvokale und der Diphthonge ist das Verhältnis erstaunlicherweise umgekehrt, was vermutlich auf standardsprachlichen Einfluss zurückzufuhren ist. In Ausnahmefallen zeigt die mittlere Generation Neuerungen, die sich bei der jüngeren nicht finden. Chambers and Trudgill (1980) erklären diese Verteilung damit, dass "mittelalterliche" Sprecher mehr Prestigeformen verwenden, um ihre Karriere zu fördern, während junge und ältere noch nicht oder nicht mehr so stark auf die sprachlichen Erwartungen der Sprachgemeinschaft eingehen. Der Kontakt zu den Nachbargebieten schlägt sich im lautlichen Bereich meist so nieder, dass die entsprechende Variante bevorzugt wird beziehungsweise die Lautung aus dem andern Dialektgebiet im Gegensatz zu den übrigen Gewährspersonen nicht übernommen wird. Tab. 1.

Übersicht zur wichtigsten im Aarauer Lautsystem festgestellten Varianz Historische Vergleichslautung; im SDS belegte Lautung westlich und östlich von und dazwischen in Aarau. In den folgenden Spalten sind die außersprachlichen Faktoren und signifikante Unterschiede mit deren Realisierung aufgezeigt, Belege in Klammern zeigen Werte unter 25% an.

Eingesessen

SDS

Generation

Kontakt

Bildung

ν ergicicnsiaiu W

A

0

0 oe

ο/ο oe

AE

Ε

Ζ

äG

mG

}G

0 0

0

0

0

(-)

(+)

mhd. ο mhd. ö Umlautung von mhd. u

(+)

m h d . ae

ae:

ε:

ε:

ε:

m h d . ae (in Verba pura) mhd. ΐ

ae:

ae:

ε:

ε:

i:

ii

i:

m h d . ei

ει

ΕΙ/

SSI

aei

aei

ει

ει

mhd. ou mhd. öu

ου

sei aey

aey

aey

aeu

ου

ου

01

01/

(EI

01

01/

aobd. iu

oei/ 13

Y:

oei/ 13

oei oei/ y-

oei oei

(-)

E'J ae: ae: E'J E'J ae: 3SI

W

A

0

oe

oe

0

u

i:

Ε

cei

01

(+) (+) (-) ae:

ε:

ε:

ae:

ε:

Ε

as'J ε: Ε

aei

ει

aei

aei

aey oei

ου

aey

aey

Y:

ο

oei

oei/ Y:

i:

Y:

318

Beat Siebenhaar

SDS

Eingesessen

Generation

Kontakt

Bildung

1

ν ergieicnsiaui

\ / a c m α ί Λ Η f i l m i l

W

A

0

AE

Ε

Ζ

oei oei/ y* aobd. iu (germ. oei (EI y: eu + w) y: oei/ oei/ oei/ mhd. iu im Auslaut 01 01 01 mhd. ä + w o:y ae:y 3KU ae:u aeiy/ α:υ α:υ Dehnung offener Silbe Dehnung vor r-Verbindung Schwund von η vor Reibelaut mhd. 1 Geminate bei 11 mhd. d /t im Anlaut mhd. t in Endsilben

+

+/-

+

(+)/

äG mG jG

W

A

0

oei

oei/

Y>

oei

oei/

7'

01

Y> (El/

oei

_

01 a:u aiu/ ae:u ae:y o:y ae:y

-

-/

+/-

+/

+/

(-) 1/i + t

(-) 1/i/u

1/ (d)

t/d

(+)

(-) υ + d

1/i/u + t

1/i t

1/i + t

t/d

t

t

t

-

+/-

t/d

u

y-

(-) +/

0

+/-

+/

+/

+

+

+/-

+

+

+

(-)

+/ (-)

(+)

y

Vi

1/i

t/d

t/d

t

+/-

-

1

i/y

Vereinzelt findet sich auch eine Variation, die sich mit keinem der untersuchten außersprachlichen Faktoren in Verbindung bringen lässt. Meist sind dabei Systemausgleichstendenzen festzustellen: Der Rückgang der Senkung der kurzen Hochzungenvokale bringt einen größeren phonetischen Abstand zu den kurzen Mittelzungenvokalen, welche parallel dazu vereinzelt gehoben werden. Die in Einzelfallen belegte Senkung des Diphthongs aus mhd. 1, ΰ im Auslaut reduziert die Anzahl der Diphthonge von 6 auf 3, analog dem standardsprachlichen System. Die konsequente Weiterführung des Staubschen Gesetzes (n-Schwund vor Reibelaut) verdrängt lokale Sonderformen zugunsten der regionalen und überregionalen Form (hier gerade aber gegen die Standardsprache). Um einen Überblick zu bekommen, wurden die bis dahin untersuchten Elemente - Vokalqualität und -quantität sowie der Konsonantismus - in einem neuen Index zusammengefasst, der den Mittelwert aller Indizes umfasst, die einen Ost-West-Gegensatz bewerten. Das Histogramm (Abb. 1) zeigt diese Verteilung und die Normalverteilung. Der Mittelwert liegt bei .647, das bedeutet, dass insgesamt mehr östliche Lautungen realisiert werden als westliche.

Dialektwandel und Einstellung

319

ι ι ι L

3

.3

.4

.5

.6

.7

.8

Anteil östlicher Formen

Abb. 1.

Histogramm der mittleren Anteile östlicher Elemente und Normalverteilung

Dieser Index wurde varianzanalytisch untersucht. In zwei Fällen ergeben sich signifikante Ergebnisse. Für die Kategorie Kontakt zu den Nachbargebieten (siehe Abb. 2 und Tab. 2 bis Tab. 3) zeigen diejenigen mit Westkontakten signifikant weniger östliche Lautungen als die übrigen; diejenigen mit Ostkontakten und diejenigen mit Bezug zu Aarau unterscheiden sich nicht signifikant.

.725 .675 "

S

.575 -

A

.525 .475 W

A

O

Kontaktkategorie Abb. 2.

Varianzanalyse: Anteil östlicher Laute, Mittelwert und 95%-Konfidenzintervall, Kontakt zu Nachbardialektgebieten

320

Beat Siebenhaar

Tab. 2.

Mittelwerttabelle: Anteil östlicher Laute, Kontaktkategorie

Kontaktkategorie

Anzahl

Mittelwert

Standardabw.

W A 0

11 18 26

.565 .646 .683

.109 .053 .061

Tab. 3.

Signifikanztest, Fisher's Protected Least Significant Difference: Anteil östlicher Laute, Kontaktkategorie

Vergleich W,A

w,o A,0

Mittelwertdifferenz Kritische Differenz

-.081 -.118 -.037

.054 .051 .044

p-Wert

.0043

2 .64 a behind with the colour white > someone with a white behind > a certain species of bird with a white behind A feature of the behind is transferred by metonymy to the possessor of the behind and after that the meaning is specialized to "house martin", since not every animal with a white behind is a house martin (cf. Geeraerts and Bakema (1993: 215-6)). Ethnobiological nomenclature is an important entry into the insights in categorization of the natural world. The dialects of Western European languages are just as valid for research into ethnobiological nomenclature as languages of non-Western traditional societies. Folk names provide evidence for the features of concepts. The names of animals in dialects also provide insights in the manner of assigning lexical items to concepts. There are a number of principles available to us with which we form ethnobiological terms. The most common member of a category will get a primary name, and the peripheral members will get secondary names. But as we have seen, some other principles are at work in this process. What their exact appearance is, remains to be solved.

References Berlin, Brent 1992 Ethnobiological classification. Principles of categorization of plants and animals in traditional societies. Princeton: Princeton University Press. Brök, Har 1995 Lexicale differentiatie bij sexe- en leeftijdsgebonden diernamen in de Nederlandse dialecten. Taal en Tongval 47: 3-30.

On ethnobiological nomenclature in Southern Dutch dialects

365

Geeraerts, Dirk 1983 Prototype Theory and Diachronic Semantics. A Case Study. Indogermanische Forschungen 88: 1-32. Geeraerts, Dirk and Peter Bakema 1993 De Prismatische Semantiek van Idiomen en Composita. Leuvense Bijdragen 82: 185-226. Suolahti, Hugo 1909 Die Deutschen Vogelnamen. Eine wortgeschichtliche Untersuchung. Strassburg: Trübner. Swanenberg, Jos 1997 Een onderzoek naar de regels van volksnaamgeving, getoetst aan enkele vogelnamen in Zuidnederlandse dialecten. Taal en Tongval 49: 61-84. Wierzbicka, Anna 1996 Semantics. Primes and universals. Oxford: Oxford University Press.