DIALECTS ACROSS BORDERS
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F.K. KOERNER (Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Salt Lake City); Sheila Embleton (Toronto) Brian D. Joseph (Columbus, Ohio); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); E. Wyn Roberts (Vancouver, B.C.) Joseph C. Salmons (Madison, Wis.); Hans-Jürgen Sasse (Köln)
Volume 273
Markku Filppula, Juhani Klemola, Marjatta Palander and Esa Penttilä (eds.) Dialects Across Borders Selected papers from the th International Conference on Methods in Dialectology (Methods XI), Joensuu, Finland, August 2002
DIALECTS ACROSS BORDERS SELECTED PAPERS FROM THE 11TH INTERNATIONAL CONFERENCE ON METHODS IN DIALECTOLOGY (METHODS XI), JOENSUU, AUGUST 2002
Edited by
MARKKU FILPPULA University of Joensuu
JUHANI KLEMOLA University of Tampere
MARJATTA PALANDER University of Joensuu
ESA PENTTILÄ
University of Joensuu
JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA
�
��
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-984.
Library of Congress Cataloging-in-Publication Data Dialects Across Borders : Selected papers from the th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002 / edited by Markku Filppula, Juhani Klemola, Marjatta Palander and Esa Penttilä. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 273) Includes bibliographical references and index. . Dialectology--Methodology--Congresses. P367 .I49 2005 47/.20--dc22 2005055892 ISBN 90 272 4787 0 (Hb; alk. paper) © 2005 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 36224 • 020 ME Amsterdam • The Netherlands John Benjamins North America • P.O.Box 2759 • Philadelphia PA 98-059 • USA
CONTENTS Introduction
vii
PART I: Dialects across political and historical borders
1
The construction of linguistic borders and the linguistic construction of borders Peter Auer
3
Static spatial relations in German and Romance: Towards a cognitive dialectology of posture verbs and locative adverbials Raphael Berthele
31
Ingressive particles across borders: Gender and discourse parallels across the North Atlantic Sandra Clarke and Gunnel Melchers
51
On the development of the consonant system in Mennonite Low German (Plautdietsch) Larissa Naiditch
73
English dialects in the British Isles in cross-variety perspective: A base-line for future research Sali Tagliamonte, Jennifer Smith and Helen Lawrence
87
PART II: Dialects across social and regional borders
119
Dialects across internal frontiers: Some cognitive boundaries Dennis R. Preston
121
On ‘dative sickness’ and other linguistic diseases in modern Icelandic Finnur Friđriksson
157
Can we find more variety in variation? Ronald Macaulay
173
vi
CONTENTS
Pronunciation of /i/ in avant-garde Dutch: A cross-sex acoustic study Vincent J. van Heuven, Renée van Bezooijen and Loulou Edelman
185
A tale of two dialects: Relativization in Newcastle and Sheffield Joan C. Beal and Karen P. Corrigan
211
PART III: Dialects across language boundaries
231
Crossing grammatical borders: Tracing the path of contact-induced linguistic change Ruth King
233
The after-perfect in Irish English Patricia Ronan
253
Dialect history in black and white: Are two colors enough? J. L. Dillard
271
Index of languages and dialects Subject index
287 289
INTRODUCTION
Until fairly recently, the word ‘dialect’ carried the connotation of being something antiquated and having low social status. The situation is now very different: along with the general rise of language awareness in modern societies, nonstandard varieties of languages have become an object of new interest, which in turn is reflected in their generally improved position even in educational systems. A similar change has taken place in scholarly research on dialects. This is partly due to the recent advances in the methods used in dialectological research. The advent of computer-assisted methods has enabled study of significantly larger databases than in traditional dialectological research. Also, methods derived from sociolinguistic and variation studies have greatly added to the general interest of dialect studies and distanced them even further from the ‘butterfly collecting’ mentality often associated with traditional dialectology. New language-theoretical frameworks form yet another source of inspiration for dialect studies today: typology, cognitive linguistics, discourse analysis and pragmatics have provided fresh perspectives on old problems and opened up completely new lines of research such as cognitive dialectology, folk linguistics, and perceptional dialectology. The articles in this volume arise from The Eleventh International Conference on Methods in Dialectology (Methods XI), which was held from 5 to 9 August 2002, at the University of Joensuu, North Karelia, Finland. The special theme for Methods XI was “Dialects across borders”. This theme was chosen because it accords well with the nature of North Karelia as a historical border area between two states and two different linguistic and cultural traditions. As can be seen from the selection of contributions included in this volume, various kinds of borders exert major influence on linguistic behaviour all over the world. The articles have been grouped according to whether they deal primarily with the linguistic outcomes of political and historical borders between states (Part I); various kinds of social and regional boundaries, including borders in a metaphorical sense, i.e. social barriers and mental or cognitive boundaries (Part II); and finally, boundaries between languages (Part III). In the first five articles, grouped under the heading of “Dialects across political and historical borders”, the main concern is the effects of political and historical borders on dialects. This chapter opens with Peter Auer’s article on “The construction of linguistic borders and the linguistic construction of
viii
FILPPULA, KLEMOLA, PALANDER & PENTTILÄ
borders”. Using the German language area as his example, Auer discusses the complex links between the nation-state and geographical space, on the one hand, and the relationship between these two and dialectal variation (or continua), on the other. An important aspect of geographical space is that it is not merely a physical phenomenon but a mental one. This idea, which Auer adopts from the early twentieth-century sociologist Georg Simmel, explains why lay persons’ ‘ethnodialectological’ perceptions about dialect boundaries may be adjusted by the existence of present or past political borders. As an example, Auer cites the dialect differences between Swabian and Low Alemannic: southwest German informants treat these as different dialects because of their past political separation, although this is not supported by dialectological facts. The former political border between East and West Germany had led to similar ‘cognitive adjustments’ in the minds of West German informants. At a more general level, the state borders between Germany, the Netherlands, Luxembourg, France, Switzerland, etc. influence people’ ‘cognitive maps’ and lead to the construction of dialect or language boundaries on the basis of political borders. Auer’s contribution is followed by Raphael Berthele’s article on “Static spatial relations in German and Romance: towards a cognitive dialectology of posture verbs and locative adverbials”, which combines in an interesting way methods used in dialectology and language contact studies with a cognitivelinguistic theoretical framework. Focusing on spatial expressions, he examines the mapping of spatial relational concepts onto syntactic structures in different varieties of German across the German–Swiss border and in the neighbouring Romance languages, including French, Italian, and Romansh. The results suggest that, in the expression of spatial relations, Swiss German and Romansh favour verb phrase constructions consisting of a verb followed by locative prepositional phrase + adverb where the adverb can be said to be semantically redundant. By contrast, in Standard High German, Standard Italian and Standard French this PP+ADV pattern is either rare or non-existent. Instead, these languages use the ‘simple’ prepositional phrase construction. Berthele’s explanation for the distinctive behaviour of Swiss German and Romansh rests on adstratal influences between these languages within the complex contact situation in Switzerland. Third in this group, Sandra Clarke and Gunnel Melchers present an interesting survey of a seldom discussed linguistic feature: pulmonic ingressive articulation. Focusing on ingressive discourse particles, the authors argue that the use of this feature is an areal feature that stretches from the eastern Baltic to the Atlantic seaboard of the United States. Clarke and Melchers suggest that the use of pulmonic ingressive discourse particles has diffused via language
INTRODUCTION
ix
contact over this geographical area, and as such provides evidence for crosslinguistic transmission of socially and pragmatically determined features, a phenomenon seldom discussed in the dialectological or contact-linguistic literature. In her study, Larissa Naiditch investigates the details behind the development of the consonant system of Mennonite Low German, or Plautdietsch, which is an insular dialect of German spoken by the religious minority of Mennonites. The speakers of this dialect can at present be found in various parts of Siberia, Kazakhstan, the USA, Mexico and Germany. The migrant past of the Mennonite community comes out in the richness of their dialect, since its consonant system has traces from a number of dialects spoken in the areas where this minority has resided in the course of history. In the final article in Part I, Sali Tagliamonte, Jennifer Smith and Helen Lawrence compare evidence from six corpora representing relic areas in the North of England, Scotland and Northern Ireland. Their aim is to find suitable diagnostic features for establishing historical relationships between New and Old World varieties of English. They argue that while verbal -s seems to be a suitable diagnostic feature, NEG/AUX contraction, for to infinitives, and zero adverbs are more problematic for testing similarities and differences in the Old and New World varieties of English. Furthermore, they suggest that examining the variable constraints on linguistic features that are shared across all varieties offers a fruitful way forward for tracking trans-Atlantic connections between varieties of English. Part II (“Dialects across social and regional boundaries”) opens with Dennis R. Preston’s article on “Dialects across internal frontiers: some cognitive boundaries”, in which he discusses some of the ongoing vowel changes in the urban dialects of the northern cities of the U.S.A. Also known as the Northern Cities Chain Shift (NCCS), these changes have been investigated by Preston and his research team from the points of view of dialectology, sociolinguistics and ‘folk linguistics’ (this last one being inspired by social psychology). Preston’s uses the term sociophonetics to describe this kind of combination of different methods. His research focuses on the productive aspect of speech and on what individual factors are behind the NCCS: how adoption of this group of changes correlates with age, commitment to residence in a given locality, and how the ethnic background and social network relationships of the immigrant speaker affect his linguistic accommodation process. Also interesting from the folk-linguistic point of view are his findings on how capable an individual is in imitating a dialect which has a sound system different from his own, and on what impact an adoption of a sound change has on his perceiving of the same change in the speech of others. The effect of
x
FILPPULA, KLEMOLA, PALANDER & PENTTILÄ
gender on speakers’ perception of their own dialect area is yet another variable studied by Preston. Next, in his article entitled “On ‘dative sickness’ and other kinds of linguistic diseases in Modern Icelandic”, Finnur Friđriksson writes on a number of changes in the use of some case-inflections (most notably, the dative, accusative and genitive when in subject position) in certain regional and social dialects of Icelandic which have by some researchers been interpreted as a threat to the very stability of the case-inflectional system of the language. Drawing on his own data representing the relevant regional dialects and social groups he, however, seeks to demonstrate that this threat is premature, as the features turn out to be so infrequent that they hardly undermine the stability of the grammatical system of Modern Icelandic. Rather, he sees the whole debate about their alleged spread as something which has originated in the educational system and in the efforts of school teachers to eradicate usages which deviate from the standard. Writing also on social boundaries, Ronald Macaulay explores ways in which the influence of language-external factors upon linguistic variation could be investigated in greater detail than has hitherto been the case in sociolinguistic research. His article, entitled “Can we find more variety in variation?”, is based on data collected from Glasgow English. According to Macaulay, the method of data collection is crucial: special care is needed to ensure that the participants in the communicative situation are on an equal footing; there is no need for an interviewer. Traditional external factors, such as age, gender and social class, should be studied in connection with each other, not as separate factors. Statistical analysis can then be used to discover gender differences within social groups that otherwise do not display significant differences. Furthermore, sociolinguists should look for ‘hidden’ linguistic variables that have not been considered in previous works. Such are, for instance, various discourse features. Fourth in Part II, Vincent J. van Heuven, Renée van Bezooijen and Loulou Edelman present an acoustic analysis of 32 Dutch-speaking guests appearing in a television talk show. They focus on the analysis of the diphthong /i/ in the speech of the speakers representing an emerging ‘avantgarde’ variety of standard Dutch, also known as Polder Dutch. The authors argue that with the help of acoustic measurement procedures they can observe a sound change in progress non-impressionistically and in much more detail than using other methods. From the sociolinguistic point of view, they claim, this new variety of standard Dutch represents yet another instance of the widespread phenomenon of women initiating and leading a linguistic change. Finally in Part II, Joan C. Beal and Karen P. Corrigan discuss regional
INTRODUCTION
xi
variation of English in a paper which is part of an ongoing, extensive project on northern English dialects. They concentrate on analysing the urban dialects of Tyneside and Sheffield from a morphosyntactic perspective. Their tentative conclusions suggest more fine-grained distinctions between northern dialects than have been found in some previous studies, which are based on only phonological criteria. Part III consists of three articles, all concerned with dialect variation or continua across language boundaries. In her article on grammatical borders, Ruth King addresses the question of linguistic constraints on borrowability in a bilingual setting. Using Preposition Stranding data from Prince Edward Island French, King argues against direct syntactic borrowing from English. Instead, she suggests that her Prince Edward Island data support the primacy of lexical borrowing as the source of syntactic effects in the recipient language. Moving on to another kind of linguistic scene which also involves contacts between two languages, Patricia Ronan focuses on the well-known Irish English after-perfect construction. Ronan examines data based on participant observation and on a corpus of Dublin oral history material compiled by the American sociologist K. Kearns. She presents evidence to support the view that the HE after-perfect is not a unified category: for some speakers the after-perfect has grammaticalised to denote ‘hot news’ events, while for others it presents a more general alternative strategy for perfect marking. This volume closes with J. L. Dillard’s article “Dialect history in black and white: are two colours enough?”. This is a critical comment on some recent views on the origins of AAVE. Calling into question the substrate account defended by many linguists, Dillard emphasises the significance of plurilingualism in the historical circumstances surrounding the growth of AAVE. He argues that the West-African slaves of the seventeenth and eighteenth centuries interacted more with indigenous Americans than with the Europeans. Consequently, English was not the only influential language in the contact setting; besides indigenous American languages, the Africans got into contacts with settlers representing different Indo-European languages, e.g. in the West Indian Islands. The plurilingual nature of the contact setting should according to Dillard be taken into account when writing the history of AAVE. Finally, we would like to thank all those people who were involved in the preparation of this volume. Our special thanks go to scholars who acted as referees for the articles submitted for this volume: Angela Bartens, Walter Chichocki, Helena Halmari, Brian Head, Raymond Hickey, Pekka Hirvonen, Heikki Hurtta, Neil G. Jacobs, Kaisu Juusela, Bernhard Kelle, William A. Kretzschmar, Kaija Kuiri, Ulla-Maija Kulonen, Timo Lauttamus, Tapani
xii
FILPPULA, KLEMOLA, PALANDER & PENTTILÄ
Lehtinen, Marja Leinonen, Matti Leiwo, Daniel Long, Kevin McCafferty, Natalie Maynor, Anneli Meurman-Solin, Terttu Nevalainen, Jussi Niemi, Pirkko Nuolijärvi, Patricia Poussa, Heinrich Ramisch, Ilkka Savijärvi, Muusa Savijärvi, Lea Siilin, Marja-Leena Sorjonen, Irma Taavitsainen, Kalaya Tingsabadh, Clive Upton, and Maria Vilkuna. We are also grateful to those who provided us with other kinds of assistance: to Prof. Christopher Hall for his translation of Larissa Naiditch’s article from German to English; to Prof. Stefan Werner for various kinds of editorial and technical assistance; to Mr Mika Saarelainen of the Joensuu University Press for his help with maps; and last but not least, to Ms Minna Korhonen for her invaluable editorial contribution to the preparation of this volume, its layout and the index.
Joensuu and Tampere, October 2005 Markku Filppula Juhani Klemola Marjatta Palander Esa Penttilä
PART I Dialects across political and historical borders
THE CONSTRUCTION OF LINGUISTIC BORDERS AND THE LINGUISTIC CONSTRUCTION OF BORDERS PETER AUER Universität Freiburg 1.
Introduction In the relationship between geographical space, language variation, and the (European) nation-state, space and language appear to be trivially linked; in fact, a whole sub-discipline of linguistics (that of dialectology) construes its professional identity around the assumption that languages vary in geographical space. The link between the nation-state and geographical space appears just as natural — there are no nations without a territory. But unlike dialectologists, political scientists, sociologists and others who have investigated the emergence of the European nation-states have given this link a considerable amount of theoretical reflection (starting with Simmel 1903 [1995]). This suggests that this link — and perhaps that between language and space as well — is not as natural as it may look at first sight. But how are dialectal variation and the nation-state related to each other? Empirically, the question becomes relevant as soon as we look at dialect continua across national borders. I will turn to this in Section 4. However, there may also be a more ideological relationship between the two; after all, the origins of systematic dialect geography go back to the late 19th century, the very same time when nationalist thinking reached a climax in Europe. At first glance, the coincidence seems to be purely coincidental, for dialectology was at the time, and continues to be, interested in (areal, diatopic) diversity; as such, it would hardly seem to be able to contribute to the ideological construction of a geographically bounded nation state. But at a second look, it becomes apparent that early dialectology and the nation-state had some common interests (as will be shown in Section 2). I will try to develop my arguments using examples from German, not because they uniquely apply to this language area (and the German nationstate), but because the German language area is rich in examples of political (nation state) borders cutting across dialect continua. At least the following types can be distinguished:
4 •
•
•
PETER AUER
the German standard language (with small differences) may be one or the only standard language used, and dialects of German are spoken, on both sides of the border (as between Switzerland/Germany, Austria/South Tyrol and Austria/Germany); a Germanic dialect continuum may be ‘roofed’ by a structurally relatively distant (exoglossic) standard variety on one side of the border and by standard German on the other (this is, cum grano salis, the case of the border between Germany and the Alsace/France); or, a Germanic dialect continuum may be roofed by different (endoglossic) Germanic standard varieties, as at the state border between the Netherlands and Germany, but also between Luxembourg and Germany (Letzebuergish is one of the standard languages used in Luxembourg 1 ).
Dialectology and the nation-state: ideological connections The 19th century saw the triumph of the nation-state, on the one hand, and of the establishment of dialect geography, on the other. In Germany, the link between the two was established in the late 19th century with the foundation of a national dialect atlas (the famous Sprachatlas des Deutschen Reichs), which from 1888 on was a state-financed project directly approved of by the government in Berlin. (The literal and metaphorical owner of the atlas was the Prussian ministry of education. With its nationalisation, and following the wish of the ministry, not of its director Georg Wenker, the reach of the atlas was extended to the total territory of the German state.) This raises the question of whether there was a common ideology shared by the nation-state and dialect geography which suggested this kind of cooperation. I would like to suggest that this ideological affinity consisted (among other things) of a (at the time) new and sharpened interest in geographical space in general, and in the boundedness of geographical spaces (i.e. in ‘borders’) in particular. 2 We know, for example, that Georg Wenker, the founder of the Sprachatlas, was determined to detect dialect borders at the start of his project; he thought that dialect areas were set off against each other in a clear-cut way (although this expectation was quickly disappointed when he did his first maps of the Rhineland). 3 To the present day, many dialectologists use maps on which
2.
1
French and German are used as (written) standard languages as well. There were other points of collaboration which moved into the foreground in the 1920s and 30s; in particular, the atlas was able and wanted to prove that the German “language islands” outside the German-speaking territory were ‘rooted’ in the dialects spoken in Germany itself. 3 Contrary to what many textbooks say, it was this geographical interest much more than the theoretical debate around the Neogrammarian idea that phonetic rules have no lexical 2
THE CONSTRUCTION OF LINGUISTIC BORDERS
5
dialect areas are delimited against each other like state territories are delimited against each other on political maps. 4 Thus, the standard map of the German dialects which goes back to the Deutscher Sprachatlas and has been reproduced again and again, is the one in Figure 1 — this version being taken from Niebaum and Macha (1999: 193). Dialect boundaries cut the German language area into sections in a complete and exhaustive way, and without internal differentiations, just as in a political map. In reality, of course, dialect spaces are not organised like this.
Figure 1: The German dialects around 1900 (from Niebaum & Macha 1999: 193).
restrictions (as one may re-formulate their ‘exceptionless sound laws’) which motivated Wenker; cf. Wenker (1886: 190). 4 Transition areas came much later.
6
PETER AUER
But even when dialect areas were not treated as sharply delimited from each other, the keen interest in borders was (and partly continues to be) at the heart of dialectological thinking: isoglosses, i.e. dividing lines between single feature realisations in geographical space, are a good example of this. In the classical books on (German) dialectology (such as Bach 11934, 31969), isoglosses are described as physically real phenomena when they are stable, like contour lines on a geographical map referring to points of equal height. (When they change, they are described like battle lines which recede or advance in a certain territory. While the geographical metaphor suggests perfect isomorphy with some kind of physical reality, the metaphor of the battle lines suggests impermeability and solidity; a battle line which dissolves means defeat.) It is important to note here that the interest in geographical space and the interest in borders are not the same. There are ways to think about geographical space which do not focus on boundaries, such as, for instance, the centre/periphery model. Ethnodialectological (folk) representations of dialect space are usually structured according to this latter model, with more or less prototypical core areas and indeterminate outer limits. Some typical folk maps of the dialects of Germany are reproduced as Figures 2 and 3.
Figures 2 and 3: Folk-dialectological maps of the dialects of Germany, drawn by two informants from Bocholt (from Stegger 2000).
THE CONSTRUCTION OF LINGUISTIC BORDERS
7
The designated dialect areas do not cover the area of Germany exhaustively but rather omit certain areas (often in the middle). Frequently, a large city is taken to be the centre of a dialect, as can be seen in Figures 4 and 5, cumulative representations of the Bavarian ethnodialectology as drawn by informants (students) from Dresden (Saxony), and Detmold (in northwest Germany, close to the Dutch border). For both groups of informants, the radiation centre of Bavarian is Munich. The city is included into the Bavarian dialect space by all informants; the borders of folk Bavarian, however, differ quite considerably among these informants.
Figure 4: Ethnodialectological maps of Bavarian, as drawn by 20 informants from Dresden; shadings refer to degrees of overlap (from Stegger 2000).
8
PETER AUER
Figure 5: Ethnodialectological map of Bavarian, as drawn by 20 informants from Bocholt; shadings refer to degrees of overlap (from Stegger 2000).
Contrary to these folk dialectologists, professional dialect geographers developed a keen interest in spatial boundedness; and the prevalent nationalist discourse all over Europe was equally interested in the borders of the nations (which of course, according to this discourse, ought to be the borders of the states). As an example, one may think of the fixation of Germans and French on the river Rhine as their state border (Wacht am Rhein), or of Hofmann von Fallersleben’s text for the later German national anthem, the first verse of which lists the boundaries of the German nation (Etsch, Belt, Maas, Memel) in order to claim its unity. As Anderson (1983 [21991: 19]) points out, this was a new way of looking at state borders: the pre-national “dynastic model” did not imply a kind of sovereignty which was “fully, flatly, and evenly operative over each square centimetre of a legally demarcated territory”, but rather, these
THE CONSTRUCTION OF LINGUISTIC BORDERS
9
“states were defined by centres, borders were porous and indistinct, and sovereignties faded imperceptibly into one another”. 5 But did the borders of the German nation-state coincide with the language border on the dialectological map? A look at Figure 1 makes clear that the fit was far from perfect: on the one hand, there was (and continues to be) at least one non-Germanic enclave within the German language area, i.e. Sorbian; on the other hand, the German dialects partly transgressed the state borders of Germany at 1900 (for instance, into Czechia), partly they included non-German speaking populations (for instance, in East Prussia or the Lorraine). In this sense, the linguistic borders did not lend support to the existing state borders. However, the argument could be turned around. The misalliance of language and state territory could be used in the framework of nationalist thinking to argue for a ‘correction’ of the state borders, i.e. an extension of the German state. This argument became more and more popular after the first World War, when Germany lost territory both in the east and in the west, and thus excluded many more speakers of German than it included speakers of other languages in the border areas. The argument was not alien to some of the Sprachatlas dialectologists, either. When the ‘correction’ of the borders became reality in the war, Walther Mitzka, then the director of the Sprachatlas and professor at the university of Marburg (and a member of the Nazi party) from 1933 onwards, propagated German settlements in the occupied Slavic territories with the argument that the German dialects would be instrumental in the Germanification of the Slavs who would give up their languages because of the superiority of German (“sprachliche Raumgewinnung”; cf. Mitzka (1941) and (1943/44)). 6 Consequently, the dialect map which Mitzka published in his textbook on dialectology in 1943 is quite different from the traditional one in Figure 1.
5
He refers to the house of Habsburg as the prototypical representative of the dynastic idea of the state. 6 In his yet unpublished PhD thesis on the Sprachatlas during the Nazi period, Wilking (1998: 138) comes to the following conclusion: “(Es) besteht kein Zweifel, daß Mitzka mit seinem Beitrag die Absicht hatte, Hitlers außenpolitisches Programm zu legitimieren” (‘there can be no doubt that Mitzke intended to legitimize Hitler’s foreign policy with his contribution’). But he also points out in a detailed analysis of many sources, that Mitzka’s concept of the Germanification of the Slavic east through language was naive compared to reality: the Nazi government and the SS had no intention to assimilate the occupied territories, they rather intended to reduce them to slavery.
10
PETER AUER
Figure 6: The German dialects according to Mitzka (1943).
THE CONSTRUCTION OF LINGUISTIC BORDERS
11
There are many striking changes: the Dutch language area has now become part of Low German; West Prussia (formerly Slavic: Kashubic) has become incorporated into the German language area; the Slavic areas in the Lausitz (Sorbian) but also in the southern part of East Prussia (Mazurian) have disappeared; language islands are now only found in Hungary but no longer (or at least not in a clear representation) in the Czech Republic and in Slovakia, where an uninterrupted German language area has taken over. The most striking fact on this map is that the German language area is not externally bounded — it seems to be potentially expansive in every direction. Very clearly, this war map reflects and ideologically supports the military expansion of Nazi Germany. By and large, however, the nationalist discourse focused on the national standard language, not on the traditional dialects. In summary, it rested on the equation one (standard) language = one nation = one territory = one state. Anderson (1983) has argued that the “imagined community” of a nation depended to an important degree on the codification of a unifying and unified standard variety and on the spread of its acceptance (which must not be equated with mastery) over the totality of the territory associated with the nation by means of a certain technology, i.e. printing. Anderson referred to the spread of a written standard variety, but it can be argued that the full penetration of modern society by the nation-state is in an important way linked to the spread of a spoken standard as well. This spoken standard was not available as a unifying force in many European nation states before the late 19th century when compulsory schooling was established and brought not only the written but also the spoken language to even the remotest areas of the national territory — and to the speakers of the remotest dialects (cf. Auer, in press). All this is well known, but it needs to be remembered in the context of a discussion of linguistic divergence at political boundaries. Since the nationstate equated nation, language and territory, it located the standard language in geographical space — it territorialised language. 3.
Towards a theory of space in dialectology Space as constitutive of the nationalist discourse of the 19th century has been given a brilliant analysis by Georg Simmel in an essay written in 1903 on the Soziologie des Raums (‘sociology of space’). In many ways, the arguments he develops for the relationship between the nation-state and space can also be applied to the relationship between geographical space and (national standard) languages. The central point in Simmel’s text is that in modern societies (nation-states), the state (and, as we can add, the standard language) becomes
12
PETER AUER
associated with a geographical area in a unique and exclusive way; the area belongs to these institutions, and it is neither possible to think of the institution without a territory 7 nor to imagine two institutions of a kind in the same territory. As a consequence, no more than one nation (and no more than one standard language) can ‘occupy’ a territory, and no nation (and no standard language) is conceivable without a territory. 8 This view, when applied to the standard language, explains why in the nationalist ideology, the Hochsprache ‘roofs’ the totality of the state-territory; it is no longer a medially (written language) nor a regionally bound variety (such as that of the ‘educated people’ in the capital, as the standard language was often defined in previous times), but evenly distributed over the area. The state borders thereby become identical with the boundaries of a standard language. At the heart of Simmel’s theory is the idea that space is not a physical phenomenon, but a mental one: Nicht die Form räumlicher Nähe oder Distanz schafft die besonderen Erscheinungen der Nachbarschaft oder Fremdheit, so unabweislich dies scheinen mag. Vielmehr sind auch dies rein durch seelische Inhalte erzeugte Tatsachen [...]. In dem Erfordernis spezifisch seelischer Funktionen für die einzelnen geschichtlichen Raumgestaltungen spiegelt es sich, daß der Raum überhaupt nur eine Tätigkeit der Seele ist, nur die menschliche Art, an sich unverbundene Sinnesaffektionen zu einheitlichen Anschauungen zu verbinden. (1995: 133, 9 emphasis P.A.)
And, with reference to borders: Wenn dieser Allgemeinbegriff des gegenseitigen Begrenzens von der räumlichen Grenze hergenommen ist, so ist doch, tiefer greifend, dieser letztere nur die Kristallisierung oder Verräumlichung der allein wirklichen seelischen Begrenzungsprozesse. Nicht die Länder [...] begrenzen einander; sondern die Einwohner [...]. Ist sie ((die Grenze, P.A.)) freilich erst zu einem räumlich-sinnlichen Gebilde geworden, das wir unabhängig von seinem soziologisch-praktischen Sinne in die 7 One may think of early 20th-century attempts to establish a Yiddish standard language, which were caught in what was considered to be a dilemma of not having a state territory unique to the Jewish nation and a corresponding national standard language. 8 Of course, the idea never quite tallied with reality; many European nations had or have de facto or de jure more than one standard variety (Switzerland, Belgium, Ireland, Finland). 9 ‘It is not the form of spatial proximity or distance which creates the specific phenomena of being neighbours or strangers, irrefutable as this may appear. Rather these are also facts which are purely caused by mental contents (...). If we require specific mental functions for the individual formations of space in history, this mirrors the fact that space in general is but an activity of the mind, the human way to combine sensual affections into uniform ideas which as such are separate.’
THE CONSTRUCTION OF LINGUISTIC BORDERS
13
Natur einzeichnen, so übt dies starke Rückwirkungen auf das Bewußtsein von dem Verhältnis der Parteien. [...] ((dann)) wird sie zu einer lebendigen Energie [...] (1995: 141). 10
It is the cognitive-mental act of construing those on the other side of the border as being different from those within one’s own social group (nation) that has an impact on language. Note that the application of Simmel’s notion of cognitive space, when applied to language, contrasts with the one traditionally accepted in dialectology. Here, we find a long tradition which tries to explain linguistic divergence in space within an interactional frequency model (cf. Auer & Hinskens in press, for a critical appraisal). According to this model, dialects (or languages) diverge when there is little direct contact (Verkehr, ‘intercourse’) between the speakers on each side of the border. Hence, impermeable borders will lead to more divergence than open borders. Frequency of interaction results in the transfer of linguistic elements from one speaker to the next. Saussure speaks of (aires de) contagion (1974: 283), and Bach, one of the most influential German dialectologists of the 20th century, writes (31969: 80–81): [...] so stellen unsere Mda-Grenzen (‘Linienbündel’) Verkehrsgrenzen dar, die Kernlandschaften aber sind Gebiete verhältnismäßig einheitlichen landschaftlichen Verkehrs [...]. Wenn wir nach den Kräften forschen, die sie ausgeformt haben und zusammenhalten, so finden wir, daß es sich nicht selten um staatliche Organisationsräume des späteren Mittelalters und der frühen Neuzeit handelt, also um die Territorien, die in einem Zeitalter strenger Gebundenheit und schlechter Verkehrsverhältnisse, vor allem aber mangelnder Freizügigkeit, eine außerordentliche Macht über die Verkehrsbeziehungen der in ihnen lebenden Menschen besaßen. 11
There are, however, good arguments against such an explanation. First of all, there is a methodological issue. The dialect boundaries Bach refers to were 10
‘Although the common notion of ‘mutual limitation’ derives from spatial limits, at a deeper level the latter is only the crystallisation or spatialisation of mental processes of limitation which alone are real. It is not the states [...] which limit each other, but the inhabitants [...]. However, as soon as it [the border] has become a spatial-sensual phenomenon which we draw into nature independent of its practical sociological function, it has strong repercussions on the mental representation of the parties’ relationship. [...] It then becomes a living energy [...].’ 11 ‘[...] our dialect boundaries (‘bundles of isoglosses’) represent borders of intercourse, and the core areas are regions of relatively uniform areal intercourse [...]. If we inquire into the forces which formed them and keep them together we find that they are not seldom the state territories of the late middle ages and of early modern times, i.e. the territories which in times of strict boundedness and poor means of transportation, and above all a lack of freedom of movement, had an enormous power over the communicational relationships between the people inhabiting them.’
14
PETER AUER
described empirically around 1900 when the old territories he refers to had long disappeared, i.e. they had survived these territories and the putatively closed networks (communities of discourse) of their inhabitants. If they are to explain these boundaries, one has to answer the question of how linguistic borders can remain intact over centuries despite the fact that the old communication borders (Verkehrsgrenzen) have disappeared, and ‘intercourse’ between people on both sides of the border is no longer impeded. The obvious answer is that historical political borders may leave their traces in cultural memory and the ethnodialectological representations of the population, which in turn may influence their way of speaking. They are not a direct function of communicational frequency. Ethnodialectological representations (cognitive maps) of the language at political borders confirm this view. While they usually follow the centre/periphery model (see above, Figures 2–5), they often switch to the model of bounded, mutually exclusive geographical spaces when it comes to political borders. Evidence for this ideological construction of the political border as a separation line between two linguistic spaces comes from those instances in which a former political border has left its traces in the ethnodialectological map of the speakers although, in fact, there is no linguistic divergence. An example is the dividing line between Low Alemannic (or Lake Constance Alemannic, as this particular area is sometimes called) and Swabian. On ethnodialectological maps drawn by southwest Germans this dividing line usually runs north/south somewhere along the peaks of the Black Forest and turns east south of the ‘Swabian Alps’ towards Lake Constance. This is also where most of the factual isoglosses run. However, although informants do not agree on the exact dividing line between Swabian and Low Alemannic in general, most of them agree on one small geographical detail: their ethnodialectological borderlines all converge to separate the twin towns of Villingen and Schwenningen (which nowadays form one commune). Dialectologists’ maps of the area do not agree with this at all. Only very few isoglosses separate the two towns — most notably, the one between western /ai/ and eastern /oa/ in some MHG ei words (e.g. std. Ei ‘egg’; cf. Klausmann, Kunze & Schrambke 1993, Map 66). The main isoglosses which dialectologists use in order to separate eastern Swabian from (Lake Constance/Low) Alemannic run east (or west) of both Villingen and Schwenningen (cf. Figure 7, and for details, Figure 8).
THE CONSTRUCTION OF LINGUISTIC BORDERS
15
Figure 7: The Swabian/Low Alemannic part of South-West Germany. Thin black lines represent ethnodialectological representations of the dialect border between Swabian and Low Alemannic as drawn by southwest German informants. Thick black lines represent the most important isoglosses. The blue line represents the former state border between Württemberg and Baden. From Hättich (1999).
This means that the ethnodialectological consensus does not reflect the objective facts. An explanation is easy to find: until 50 years ago, when the state of Baden-Württemberg was founded by merging the territories of Württemberg (Swabia) and Baden (which had been independent for centuries), the two towns were separated by a political border. Since the regional identities of Schwaben (Württemberger) and Badener continue to be very strong, the two towns are salient ethnogeographical landmarks of the former political division. The general knowledge southwest German informants have about the dialect differences between Swabian and (Low) Alemannic are mapped onto the political territories: the political border leads to the cognitive adjustment of the objective dialectal isoglosses and their displacement in geographical space.
16
PETER AUER
Figure 8: Isoglosses separating Low Alemannic and Swabian (from Klausmann, Kunze & Schrambke 1993, map 74).
The same holds true for the former political border between East and West Germany. Most West Germans believe that East and West Germans speak differently, and not rarely is the ethnodialectological space of Upper Saxonian, the vernacular spoken in the southern part of the (former) GDR, expanded to the whole of the country. The most salient East German variety is looked upon as typical for the entire (former) East German state territory (see Figure 9).
THE CONSTRUCTION OF LINGUISTIC BORDERS
Figure 9: Ethnodialectological maps of the Upper Saxonian dialect area as drawn by 20 informants from Bocholt (north-west Germany); shadings refer to degrees of overlap (from Stegger 2000).
17
18
PETER AUER
While the intercourse model offers no explanation for this divergence of ethnodialectological and dialectological boundaries, Simmel’s account of space as a cognitive construct does. The German state borders and the dialect continua crosscutting it In the last section, I have argued that political boundaries are not directly (i.e., as a hindrance of communication) responsible for linguistic (dialect) divergence, but can influence people’s cognitive maps to such a degree that an actual or former political border is automatically treated as a dialect boundary as well — even if the dialectological facts contradict this correlation. In this section, I will discuss examples in which there is actual divergence of the dialects spoken on both sides of the border. As outlined in Section 3, the territorialised national standard languages extend their geographical reach exactly to the border of the nation state and roof the dialects up to this spatial limit. From this perspective, it is not the dialects on both sides of the border which cannot surmount it (as the ‘intercourse’ model suggests); rather, the state borders are construed as the insurmountable territorial limits of the national standard varieties because of the ideological equation of one nation = one territory = one standard language. Let us first look at the western border. The speakers of the adjoining Germanic dialects in the Netherlands, Luxembourg, and France (in Lorraine and, above all, Alsace) dispose of linguistic repertoires which include Dutch, French and Letzebuergish, i.e. the standard languages of the respective nationstates. 12 For the Dutch/German border, there is ample evidence collected by Kremer (1990), Niebaum (1990) and others that the dialects have diverged considerably at the national border over the last 60 years. As one example of many, consider the changes in the word for the bird ‘wren’ (std. German Zaunkönig, std. Dutch winterkoninkje) between 1940 and 1975, as shown in Figure 10. It is easy to see on the left-hand map that the usual word, on the German as well as on the Dutch side of the northern part of the border, was nettelkönning before the war, a word which neither corresponds to the German nor to the Dutch standard. On the 1975 map, this word has been replaced on the Dutch side almost entirely by the standard Dutch word winterkoninkje, while on the German side some dialect speakers still use low German nettelkönig, but mostly std. German Zaunkönig or its phonologically adapted Low German form (tuunkönning) have taken over. As a consequence, there is 4.
12 The very small German-speaking population in the westernmost part of Belgium will not be considered here.
THE CONSTRUCTION OF LINGUISTIC BORDERS
19
Figure 10: Word for ‘wren’ on both sides of the national border between Germany and the Netherlands in the area of Enschede/Rheine (En = Enschede, Bo = Bocholt, Bh = Bentheim, Al = Almelo, We = Wesel); left: 1940, right: 1975 (from Kremer 1990).
divergence between one western and several eastern forms exactly at the state border. This is but a small example of the larger trend in this area. It agrees with the ethnodialectological representations; informants were asked to list those of the neighbouring villages in which ‘the same dialect’ was spoken as in their own village (indicated by arrows in the Figure 11). With very few exceptions, the state border turned out to be the ethnodialectological border as well. There can be no doubt that dialectal divergence at the national border between the Netherlands and Germany is but a side-effect of the much more powerful processes of dialect-to-standard advergence which are in full progress, particularly in the Netherlands, but also in Germany. Auer and Hinskens (1996) tried to capture this development as in Figure 12.
20
PETER AUER
Figure 11: The ethnodialectological border between Germany and the Netherlands in the Enschede/Rheine area (from Kremer 1984). Arrows indicate those villages in which according to the informant’s opinion ‘the same’ dialect was spoken as in his/her own village.
THE CONSTRUCTION OF LINGUISTIC BORDERS
standard Dutch
21
standard German
advergence
advergence
national border
convergence
convergence
traditional dialect continuum NL
divergence
D
Figure 12: Dialect divergence at the Dutch/German border as a consequence of dialect/standard advergence; adapted from Auer & Hinskens 1996.
The former dialect continuum across the national border has given way to dialect boundaries coinciding with a political border, because the traditional dialects are strongly influenced by the respective national standard varieties. In terms of the previous discussion, this can be interpreted as a long-term victory of the standard language discourse of the 19th century which, strangely enough at first sight, has had its strongest impact at a time in which the political border between Germany and the Netherlands has lost most its significance (and certainly no longer impedes communication). A similar scenario applies to Luxembourg. However, Letzebuergish is a very young standard variety not yet fully supported by writing (for which mainly French and to a lesser degree German are used); it therefore has a lesser impact on the dialects. Nonetheless, convergence of the eastern dialects towards the central variety of Luxembourg city has progressed considerably, and has dissolved the old dialect continuum with the Franconian dialects spoken in Germany at the German/Luxembourg border (cf. Gilles 1999). The situation at the French/German border is obviously different since French, the standard language almost exclusively used on the French side, is exoglossic to the Alsatian dialects. Therefore, dialect-to-standard advergence can only take place east of the Rhine. On the French side of the border, the strong pressure towards standard French has led to language shift, but also to a high amount of borrowing into Alsatian, not only on the lexical level (cf. Klausmann 2000) but also in grammar and in prosody (cf. Gilles & Schrambke
PETER AUER
22
exoglossic standard (French)
endoglossic standard (German)
borrowing
advergence
NATIONAL BORDER
convergence
convergence
traditional dialect continuum F
divergence
D
Figure 13: Dialect divergence at the French (Alsatian)/German border as a consequence of dialect/standard advergence and borrowing.
2000). The factual divergence of the dialects spoken on both sides of the border is once more accompanied by a strong ethnodialectological divergence. As a recent thesis by Finger (2002) shows, particularly the Alsatians no longer see any resemblance between their language and the Alemannic dialects spoken on the other side of the border. It is mainly the increasing relevance of the standard languages which makes the traditional dialects diverge, together with a strong ethnodialectological feeling of difference (cf. Figure 13). We now turn to the southern border between Germany and Switzerland and Austria. In these countries, German is the/one national standard variety just as in Germany, i.e. the dialects on each side of the border are roofed by (approximately) the same standard. The Swiss/German language border has been described in some detail by Schifferle (1990) and Seidelmann (1989). They agree that there is a rapid process of divergence of these traditionally very similar High Alemannic dialects 13 at the state border. 14 Again, one example may suffice to show the
13
Schifferle and Seidelmann disagree on the age of this divergence; while Seidelmann argues that it goes back to Napoleonic times when the former Habsburg area was divided politically along the river Rhine, Schifferle believes that we are dealing with a 20th century phenomenon. Schifferle stresses the more conservative character of the Swiss dialects and attributes most divergences to northern innovations which were stopped at the border, while Seidelmann points out that the Swiss dialects in the Aargau also undergo changes under the influence of the Zurich dialect. However, these differences do not touch my argument.
THE CONSTRUCTION OF LINGUISTIC BORDERS
23
type of divergence; it is that of the front vowel systems of High Alemannic (cf. Moulton 1961). In the area around Waldshut investigated by Schifferle in the early 80s, there is a strikingly consistent difference between a southern (Swiss German) three-level front vowel systems (/e:/ ~ /ε:/ ~ /æ:/), and the two-level systems (/e:/ ~ /ε/) on the German side of the border. (The difference is salient in words such as northern open [ε]sse vs. southern extra-open [æ]sse ‘to eat’.) Schifferle cites a number of older studies (from the 20s and 30s of the 20th century) in
Figure 14: The front vowel systems north and south of the Swiss/German border (from Schifferle 1990).
14 The possible counterargument that, rather than the state border, the river Rhine or Lake Constance might be responsible for this divergence (the political border runs along the river most of the time) does not only fail to explain why divergence only set in in the last century, it is also falsified by the very strong dialect boundary between the town of Constance in Germany and the adjoining town of Kreuzlingen in Switzerland, which are both located south of the river Rhine/Lake Constance.
24
PETER AUER
which a three-level system is still attested for the southernmost part of Germany as well. 15 Some 70 years ago, the dialects on the German side of the border must have started to diverge with respect to this (as many other) variable(s). How can this divergence be explained? Once more, we have to consider the whole repertoire of the speakers, which, in this case (unlike the Alsatian one), includes a German standard variety and traditional dialects both in Switzerland and in Germany; however, in terms of usage domains and in terms of attitudes towards dialect and standard, the German and the Swiss German situation are very different. The use of Swiss Standard German is more or less restricted to reading and writing (‘medial diglossia’). The Swiss German dialects are much more widely used and have a much higher prestige than the German dialects of Alemannic. With regard to the national discourse of the standard languages in Europe, Switzerland is an interesting exception; it is the only old nation in Europe for which the equation of one nation = one language never held. Instead, Swiss German identity is very much tied to the Swiss German dialects. The linguistic repertoire structure of German-speaking Switzerland therefore stands in sharp contrast with that of southern Germany where the dialects are no symbols of national collective identity, while the standard language is. What ends at the national border between Germany and Switzerland then is not the reach of standard language, but a certain repertoire type (diaglossia on the German side, medial diglossia on the Swiss side; cf. Auer, in press). This leads to divergence because the dialects are no longer considered to be equivalent; the national border is construed as a language repertoire (and prestige) border. As a consequence, new developments do not cross the border. The type of innovations that reach the border area from the south and from the north (i.e., in Switzerland and in Germany) is also different. In Germany, the standard variety, or a regional dialect which is closer to the standard, is more and more accepted as the unmarked way of speaking, while in Switzerland the more prestigious urban dialects (in our case the Zurich dialect) exert an influence on the border region. As a consequence, vertical convergence in Germany contrasts with horizontal convergence in Switzerland. Schematically:
15
The data elicitation techniques of the Südwestdeutscher Sprachatlas, which were able to dig up the most archaic forms as late as the 1970s, also document this older state of the vowel systems, which Schifferle was no longer able to find in Germany; cf. map II–3.02, essen.
THE CONSTRUCTION OF LINGUISTIC BORDERS
25
endoglossic (regionalised) standards (Swiss German standard/southwest German standard)
. southwestern
regiolect
advergence
national border
convergence
convergence
traditional dialect continuum
CH
divergence
D
Figure 15: Dialect divergence at the Swiss/German border as a consequence of dialect/standard advergence and dialect convergence.
Once more, it is not impeded ‘intercourse’ across the border which has led to dialect divergence and the dissolution of a former dialect continuum, but rather the ideological construction of the border as the dividing line between two different repertoire types (which both contribute to national identity). Let us finally turn to the German/Austrian border. Scheuringer (1990) points out that the traditional dialect isoglosses in this area (some of which are shown in Figure 16) run north/south and do not orient themselves to the national border at all. On both sides of the border, very similar dialects of Middle Bavarian used to be spoken. Scheuringer took a closer look at the situation in the border towns of Braunau in Austria and Simbach in Germany. He observed two types of developments. In one group of dialect features traditionally shared by both towns, a gradual process of standardisation has set in on both sides of the border. This process is stronger in Austria than in Bavaria. It applies to basilectal features with a wide geographical distribution such as han instead of (otherwise) Bav. san = std. sind ‘(we/they) are’, or kemm(t)/kimm(t) instead of (otherwise) Bav. kumm(t) = std. kommt/komme
26
PETER AUER
Figure 16: Traditional isoglosses in the border area between Germany (Lower Bavaria) and Austria (Upper Austria); from Scheuringer (1990). The isoglosses separate the western from the eastern area of Middle Bavarian. 1 = western l-vocalization in syllable coda; 2 = western diphthongs for MHG ô; 3 = western han instead of san for 1. & 3.Pl.Pres. of sein ‘to be’.
‘s/he, I come(s)’. The second development holds for dialect features with a more restricted range, not shared by the two towns traditionally. Here, local realisations are given up in favour of more regiolectal ones. Again the process is more advanced in Austria, but in addition, the target of this regionalisation process is different: while Braunau takes over Vienna features, Simbach in Bavaria assimilates to the Munich regiolect. This applies, for instance, to the various vowel changes conditioned by the vocalisation of /l/ in syllable coda, where a Munich system with unrounded vowels such as in /fu:i/ = std. /fi:l/ ‘a lot’ contrasts with a Vienna system with rounded and also monophthongised vowels such as in /fy:/ ‘a lot’, as well as to the realisation of MHG ei as in std. kein ‘no’ as /koa/ in the Munich regiolect, but as /kã:/ in the Vienna regiolect.
THE CONSTRUCTION OF LINGUISTIC BORDERS
27
Compared to the situation in the west or south, divergence is less dramatic. The repertoires on both sides of the border are similarly structured; both are diaglossic (Bellmann 1998; Auer, in press), which means that intermediate forms (regiolects) of speaking between dialect and standard become more and more important. In many ways, they are closer to the standard variety, but they also show features of their own, spreading from the large cities which act as radiation centres for either area. The impact of these cities — Munich and Vienna — stops at the national border. In addition, Braunau has resisted the Vienna regiolect less than Simbach has resisted the Munich regiolect. Schematically:
endoglossic standard (Bavarian, Austrian regional standard variety)
Munich regiolect
Vienna regiolect
advergence
advergence national border
convergence
convergence
traditional dialect continuum
divergence Figure 17: Dialect divergence at the Austrian/German border as a consequence of dialect/standard advergence.
28
PETER AUER
5.
Conclusions On the basis of a cognitive approach to space in the tradition of Georg Simmel, I have argued in this paper that geopolitical (above all, national) borders should be looked upon as cognitive constructs intimately linked to the “imagined communities” (Andersen) they delimit. These “imagined borders” can nonetheless have a strong impact on the dialect continua which they crosscut. Dialect divergence at the national borders of Germany is therefore not due to impeded communication as suggested in traditional dialectology (Verkehr); rather, they are the limits (boundaries) of the reach of the national standard languages or of repertoire types which symbolise, in some way or other, national identities. This divergence can be expected to increase to the degree that (in this order) (a) the national standard languages, (b) the repertoire types (diaglossic/diglossic), or (c) the regional dialects differ on both sides of the border. References Anderson, Benedict. 1983 [21991]. Imagined Communities: Reflections on the Origin and Spread of Nationalism. London: Verso. Auer, Peter. In press. “Europe’s sociolinguistic unity, or: A typology of European dialect/standard constellations”. Perspectives on Variation ed. by Nicole Delbecque et al. Berlin: Mouton de Gruyter. Auer, Peter & Frans Hinskens. 1996. “The convergence and divergence of dialects in Europe. New and not so new developments in an old area”. Sociolinguistica 10.1–30. Auer, Peter & Frans Hinskens. In press. “The role of interpersonal accommodation in a theory of language change”. Dialect Change. The Convergence and Divergence of Dialects in Contemporary Societies ed. by P. Auer, F. Hinskens & P. Kerswill. Cambridge: Cambridge University Press. Bach, Adolf. 1934 [31969]. Deutsche Mundartforschung. Heidelberg: Winter. Bellmann, Günter. 1998. “Between base dialect and standard language”. Folia Linguistica XXXII:1–2.23–34. Finger, Bernd. 2002. “Die Verwendung von Nicht-Standardvarietäten in der transnationalen Kommunikation: Ausgehend von grenzüberschreitendem Dialektgebrauch am Oberrhein”. Sprachbewussheit im schulischen und sozialen Kontext (= forum Angewandte Linguistik Bd. 39) ed. by J.A. Bateman & W.W. Wildgen, 91–116. Frankfurt: Lang. Gilles, Peter. 1999. Dialektausgleich im Lëtzebuergeschen: zur phonetischphonologischen Fokussierung einer Nationalsprache. Tübingen: Niemeyer.
THE CONSTRUCTION OF LINGUISTIC BORDERS
29
Gilles, Peter & Renate Schrambke. 2000. “Divergenzen in den Intonationssystemen rechts und links des Rheins. Die Sprachgrenze zwischen Breisach (Baden) und Neuf-Brisach (Elsass)”. Bausteine zur Sprachgeschichte ed. by E. Funk, W. König & M. Renn, 87–98. Heidelberg: Winter. Hättich, Stefanie. 1999. Erkennen der sog. Schwarzwaldschranke. Unpublished seminar paper. Department of German I, University of Freiburg. Klausmann, Hubert. 2000. “Changes of dialect, code-switching, and new kinds of usage: the divergence of dialects along the border between Germany and France in and around the region of the Oberrhein”. International Journal of the Sociology of Language 145.1009–1130. Klausmann, Hubert, Konrad Kunze & Renate Schrambke. 1993. Kleiner Dialektatlas: Alemannisch und Schwäbisch in Baden-Württemberg. Bühl/Baden: Konkordia. Kremer, Ludger. 1984. “Die niederländisch-deutsche Staatsgrenze als subjektive Dialektgrenze”. Driemaandelijkse Bladen 36.76–83. ----------. 1990. “Kontinuum oder Bruchstelle? Zur Entwicklung der Grenzdialekte zwischen Vechtegebiet und Niederrhein”. Germanistische Linguistik 101:3.85–123. Mitzka, Walther. 1941. “Bauern- und Bürgersprache im Ausbau des deutschen Volksbodens”. In Von deutscher Art in Sprache und Dichtung. Vol. I (Die Sprache, ed. by F. Maurer), 67–97. Stuttgart and Berlin: Kohlhammer. ----------. 1943/44. “Die Ostbewegung der deutschen Sprache”. Zeitschrift für Mundartforschung 19.81–140. ----------. 1943. Deutsche Mundarten. Heidelberg: Winter. Moulton, William G. 1961. “Lautwandel durch innere Kausalität: die ostschweizerische Vokalspaltung”. ZfM 28.227–251. Niebaum, Hermann. 1990. “Staatsgrenze als Bruchstelle? Die Grenzdialekte zwischen Dollart und Vechtegebiet”. Germanistische Linguistik 101:3.49–83. Niebaum Hermann & Jürgen Macha. 1999. Einführung in die Dialektologie des Deutschen. Tübingen: Niemeyer. Saussure, Ferdinand de. 1974. Cours de linguistique générale (ed. Charles Bally etc., critical edition by Tullio de Mauro). Paris: Payot. Scheuringer, Hermann. 1990. Sprachentwicklung in Bayern und Österreich. Hamburg: Buske. Schifferle, Hans-Peter. 1990. “Badisches und Schweizerisches Alemannisch am Hochrhein”. Germanistische Linguistik 101:3.315–340.
30
PETER AUER
Seidelmann, Erich. 1989. “Der Hochrhein als Sprachgrenze”. Dialektgeographie und Dialektologie (= FS Bellmann). Marburg 1989. Simmel, Georg. 1903. “Soziologie des Raums”. Gesamtausgabe ed. by Otthein Rammstedt. Frankfurt am Main: Suhrkamp, 1995. Vol 7: Aufsätze und Abhandlungen 1901–1908, pp 132–184. Südwestdeutscher Sprachatlas. Marburg: Elwert (1989–?). Stegger, Inga. 2000. Ethnodialektologische Untersuchungen zur Wahrnehmung des deutschen Dialektraums durch Sprecher aus Obersachsen und Nordwest-Deutschland. Wissenschaftliche Arbeit im Fach Deutsch, University of Freiburg, Department of German I. Wenker, Georg. 1886. “Vortrag”. Verhandlungen der 38. Versammlung deutscher Philologen und Schulmänner in Gießen vom 30. Sept. bis 3. Okt. 1885 [Proceedings of the 38th Convention of German Philologists and Schoolmen in Gießen, Sept. 30–Oct. 3, 1885], 187–194. Leipzig. Wilking, S. 1998. Der deutsche Sprachatlas im Nationalsozialismus. Studien zur Dialektologie und Sprachwissenschaft zwischen 1933 und 1945. Diss. Heidelberg.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE TOWARDS A COGNITIVE DIALECTOLOGY OF POSTURE VERBS AND LOCATIVE ADVERBIALS
RAPHAEL BERTHELE Universität Bern Traditionally, dialectologists have been interested in the variety of words and expressions used in local vernaculars to refer to physical space. In German dialectology, particularly the spatial adverbs and adverbials have been discussed repeatedly and broadly (cf. Stadelmann 1978; Rowley 1980). At the same time, spatial referencehas been an important issue in other linguistic subdisciplines such as cognitive linguistics (cf. Talmy 2000b), typology (cf. Wälchli 2001), psycholinguistics (Becker 1994; Bowerman 1996), and anthropological linguistics (Levinson 1996; Bloom et al. 1996). This contribution combines methods from variationist linguistics with a framework from cognitive linguistics in order to shed new light on some particularly striking facets of spatial reference in non-standard forms of both German and Romance. An attempt will be made to integrate frameworks from neighboring sub-disciplines into dialectological research, and it will be shown that, ideally, this combination is able to contribute to all sub-disciplines concerned. The data analyzed in this paper are part of a larger research project on variation in the field of spatial reference expressions. The project focuses on German, particularly on differences between the standard variety (Standard High German, SHG) and a group of Swiss German Dialects (SG). Additionally, some of the neighboring Romance languages (French, Italian, Romansh) will be taken into consideration. The project involves both motion in space and static relations. For the present contribution, however, we will only be concerned with linguistic descriptions of static relations of two (or more) objects in space. Two salient phenomena are singled out: firstly, the semantics of the finite verbs and, secondly, the structure and semantics of the verb-sister elements in the locating verb phrase (VP). The variation in these two syntactic slots will be presented and, finally, an attempt will be made to explain the patterns found in the data.
RAPHAEL BERTHELE
32 1.
Data, varieties and informants The data analyzed in this contribution have been collected using a picture book containing 71 pictures. This picture book has been elaborated and used by researchers at the Max-Planck-Institute in Nijmegen in order to collect spatial language data from many different languages (cf. e.g. Bowermann 1996; Bowerman/Choi 2001). The drawings always contain a marked object in a particular spatial situation (cf. the examples below, Sections 3 and 4). The informants are asked to locate the marked object with regard to its environment. The data have been collected by myself and by some of my students. 1 As Table 1 shows, 15 informants are speakers of SG dialects, and 12 informants speak SHG. Since in German-speaking Switzerland all natives are firstly speakers of one of the SG dialects, the Swiss informants speaking SHG do not speak their native dialect in the data collection situation, but rather the Swiss variety of the standard language which they have learnt from media use and in school (for a broader discussion of the Swiss German diglossia, cf. Kolde 1981; Rash 1998). In order to compare the German standard and nonstandard data, Romansh (19 informants), standard Italian (6 informants), 2 and standard French (5 informants) data have been collected as well. Romansh is spoken in the eastern part of Switzerland, the language is highly endangered and virtually all of its speakers are bilinguals (with SG) or trilinguals (with SG and Italian, cf. Carigiet 2000: 235). German
Romansh
Total 27 Inf. Total 19 Inf.
SHG (Germans)
SHG (Swiss)
Bern dialect
Wallis dialect
Muotathal dialect
Sense dialect
5
7
3
3
6
3
Surmiran
Sursilvan
Vallader
7
8
4
French
5 Inf.
(all standard French)
Italian
6 Inf.
(all standard Italian) Table 1: Informants.
1
Acknowledgment: The data presented here have partly been collected by some of the students who participated in a class on spatial language. Many thanks to Michael Ehrler, Martin Klopfenstein, Marlene Schild, Susanne Schneider, Denise Pfammatter, and Michele Werner. I am also very grateful to Paola Gilardi’s help with some additional Italian data, and to Adrian Farrér for collecting additional data from Romansh Surmiran. Finally, I want to thank Georges Darms who allowed me to collect data in his Romansh classes at the University of Fribourg. 2 Unfortunately, it was not possible at this stage to include the still quite vital Italian dialects of Ticino.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
2.
33
Conceptual structure of a spatial expression There are different approaches and competing terminologies in the realm of spatial linguistics (for an overview, see Levinson 1996). However, the basic concepts are quite similar in most of those approaches. The most basic concepts are the objects which are physically related in a particular situation. To give a simple example, the basic parts of a spatial expression are described in Figure 1. By definition, spatial expressions serve to locate objects and events in space. The located entity is the figure (in Langacker’s terminology: trajector, see Langacker 1990: 10), the environment with respect to which the figure is located can be called ground (Langacker: landmark). Since this contribution addresses some issues raised by Talmy (1985, 2000b), Talmy’s terms figure and ground will be used throughout the paper. Figure 2 shows the main conceptual parts of a spatial expression. The figure is the entity which is located in space. In discourse, the figure is the element on which attention is focused. A typical spatial expression involves a — more or less schematic — concept of a process, which Talmy calls ‘activating process’ (Talmy 2000b: 218). The activating process has one of two possible parameter settings, namely static or dynamic. Even if the figure is not moving but simply remains in a static relation to the ground, we can assume that at least some basic interplay of physical forces must take place in order to keep the figure stable at its location. Thus it seems justified to assume a process also for the static relations, although it might be a limiting case.
Figure 1: Example for a picture from the picture book.
34
RAPHAEL BERTHELE
Figure 2: Conceptual structure of a spatial expression.
In the languages and varieties discussed in this paper, the association function is usually expressed by a spatial preposition (P). Some of the Ps are topological — i.e. dimensionally neutral — in nature (e.g. German an, bei, or in), others combine topological relations with axis-related concepts (auf, unter [vertical axis]; hinter, vor [transversal axis], cf. Becker 1994: 18). The ground is the object or the set of objects with relation to which the figure is located. In the languages and varieties falling within the scope of this paper, the ground is most frequently expressed by the NP governed by the spatial preposition. The prepositional phrase (PP), which is in a sister-position to the verb, expresses thus what Talmy calls the core schema, i.e. the association function plus the ground-object. 3 In typical responses to the picture stimuli, the figure is expressed by the Subject-NP and the activating process is mapped onto the verb. In the remainder of this paper, the variable realizations of two conceptual ‘slots’ in Figure 2 are analyzed: firstly, the linguistic and conceptual variation in the domain of the association function (Section 3), and secondly, the conceptual and linguistic construal of the activating process, i.e. the verb-slot (Section 4).
3
Talmy’s terminology has been widely used in order to investigate the typological differences in the domain of motion verbs (Slobin 1991 and 1996; Talmy 1985, 2000a and 2000b). Even though this contribution is concerned with static spatial settings, the debate on motion verbs and motion events is an interesting parallel and will be discussed briefly in Section 5.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
35
3.
Simple vs. complex expression of the association function With only very few exceptions, the association function is mapped onto the P. 4 However, this does not mean that it is exclusively located in the P, as is shown by examples (1) to (4). (1)
d’sigarett isch schräg im mundwinkul dri the cigarette is diagonally in the corner-of-the-mouth insid
WS 5
(2)
dr schluuch isch um d baumwürz ume the hose is around the stomp around
MU
(3)
ds stirnband giit ume chopf um the headband goes around-the head around
BE
(4)
der Hund sitzt in der Hütte drin the dog sits in the kennel inside
SHG
In examples (1) to (4) we discover the same spatial concept both in the semantics of the P and of an additional Adverb (ADV). The syntactic status of this additional ADV is not quite clear. Many scholars assume that the ADV is part of the PP, while for others the ADV is an adverbal modifier or even a part of a (particle-)verb (cf. McIntyre 2001: 63f.; Pittner 1999: 88; Stadelmann 1978: 108; Olsen 1996: 306). For the present purposes, the exact syntactic analysis is not at issue, and it may well be that the constructions listed above are structurally ambiguous, as in examples (5) and (6).
4
(5)
...dass ds stirnband [ume chopf um] [giit]
(6)
...dass ds stirnband [ume chopf] [um giit] ...that the headband around-the head around goes
Exceptions are the cases where a picture is described without using genuinely spatial language, as in diese Tasse hat einen Sprung (‘this cup has a crack’) instead of in dieser Tasse ist ein Sprung. 5 The labels can be read as follows: WS: Wallis dialect; MU: Muotathal dialect; BE: Bern dialect; SHG: Standard High German spoken by Germans; SHG_CH: Standard High German spoken by Swiss (L1 is always a SG dialect); FR: French; IT: Italian; ROM: Romansh; RO_SUS: Romansh Sursilvan; RO_VAL: Romansh Ladin (Vallader); RO_SUM: Romansh Surmiran
RAPHAEL BERTHELE
36
The respective spatial relational concepts which are mapped onto the P and the ADV stand either in a relationship of redundancy, as in examples (1)– (4) above, or in a complementary relation, as in examples (7)–(10): (7)
der henkul hanget uber di täscha embri the handle hangs over the bag down
WS
(8)
il bal ei sut la sutga en the ball is under the chair inside
RO_SUS
(9)
la cozza ei sulla meisa giu the tablecloth is on the table down
RO_SUS
(10) la suga ei sur la cuscha vi the hose is on the stump over
RO_SUS
The English glosses are only approximate, since there is no strict congruence between the application of the German/Romance Ps and ADVs and their English counterparts. It seems appropriate to assume two semantically distinct categories for the verb phrases with P and ADV: the ADV can be either in a complementary or in a redundant semantic relation to the P.
Figure 3: Two picture stimuli for partial containment relations.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
37
The question is whether the semantically redundant PP+ADV constructions are really equivalent to their corresponding simple PP constructions. Examples such as the one in (4) above could evidence a similar distinction as in the English in vs. inside, i.e. partial vs. complete containment of the figure in the ground. The data analyzed in this sample do not point to this direction, since the dog in the picture which stimulates responses such as the one in (4) is not completely inside the kennel (cf. Figure 3). Even the cigarette, which undoubtedly is located mostly outside the mouth, can be located using this construction, as in example (1). None of the varieties investigated here uses the complex PP+ADV constructions categorically for a particular picture stimulus. Since both simple PPs and complex PP+ADV constructions occur for the same pictures and in the same varieties, the data analyzed here do not support any other systematic semantic differentiation than the ‘redundancy-hypothesis’. Other means of data elicitation and/or meta-semantic judgements of informants might shed more light on this issue. 27% of all PP+ADV constructions are uttered in order to describe a stimulus which represents a spatial setting where the figure surrounds the ground: (11) dr haag isch ums huus ume the fence is around-the house around
MU
(12) s mäscheli isch um d cherze ume the ribbon is around the candle around
MU
Almost 9% of all PP+ADV constructions can be attributed to the inverse relation of a ground-element encircling the figure, as shown by example (13): (13) de öpfel isch im ring inne the apple is in(side) the ring in(side)
MU
It seems very clear that the encircling configurations favor the use of PP+ADV, and more precisely redundant PP+ADV constructions. The remaining 64% of the PP+ADV constructions are distributed more or less equally over a multitude of core schema types. In the following paragraphs the occurrences of the complex PP+ADV constructions are analyzed in terms of their frequencies in the respective linguistic varieties. Figure 4 gives an overview of their average frequencies.
38
RAPHAEL BERTHELE
Figure 4: Average frequencies of redundant (left columns) and complementary (middle columns) PP+ADV constructions. The right column represents the sum of all PP+ADV constructions. All bars are relative frequencies in relation to the total number of responses.
Figure 4 shows that the numbers of PP+ADV constructions are negligible in both SHG spoken by Germans as well as by Swiss informants. There is no doubt that the construction is in principle grammatical in SHG, too, 6 but its occurrences in our data are very rare. All dialects, on the other hand, show higher frequencies of the PP+ADV constructions, and within the group of those constructions the use of the redundant variant is by far the most frequent. For the redundant constructions and for the sum-values (but not for the complementary constructions), the difference between the group of SG dialects, on the one hand, and SHG, on the other, is statistically significant (on the p<0.05 level, independent sample t-test). Among the SG dialects, the differences once again are considerable. The Muotathal dialect from central 6
Cf. Olsen 1996 for a syntactic description of very similar phenomena. It is not unlikely that a similar investigation into different dialects, sociolects and registers of German in Germany or Austria could uncover comparable patterns of variation.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
39
Switzerland shows very high values for the redundant constructions, whereas the Bern dialect and the Sense dialect display somewhat lower values. The Wallis dialect, on the other hand, shows a saliently higher use of the complementary constructions of the type given in example (7) above. Additionally, Figure 4 makes it clear that the PP+ADV constructions are not a typical Romance pattern. Neither the French nor the Italian data show even a single token of this type. Only the Romansh informants use comparable constructions. Given the intensive contact situation of Romansh with SG, it seems quite probable that SG as a contact language has left its traces in the Romansh varieties. However, the Romansh examples are not simple copies of the (mostly redundant) German pattern; they rather display a high amount of the complementary constructions, which are only rarely used in SG. The aim of this section is to give an account of the lexical and conceptual variation in the domain of the core schema. Regarding the expression of the association function, the SG dialects are marked by a significantly higher degree of redundancy. The Romance languages, including standard Italian and French, do not seem to share this feature. Both languages simply express the association function by means of the spatial P. The Romansh dialects, most likely due to their sociolinguistic embedding, display similar structural possibilities as the SG dialects. However, they do it with a different semantic ‘stuffing’. The — admittedly expected — convergence in the Romansh dialects has thus to be understood as a self-assured convergence, a linguistic change which takes a structural pattern from the adstratum and uses it in its own way. 4.
Co-event verbs In Section 3, the lexical and syntactic patterns in the domain of the core schema were analyzed. In this section, the lexical choices for the verb will be examined. As we have seen in Section 2 above (cf. Figure 1), the primary role of the verb is to express the activating process in the locating expression. In the case of static spatial relations, a semantically quite simple and schematic concept is fully sufficient to express this activating process. The simple fact that the figure ‘is’ at a particular place can be rendered by the equivalents of sein (‘to be’) in all varieties and languages within the scope of this paper. However, as will be shown in this section, additional concepts (in Talmy’s terminology: ‘events’) can be packed into the verb slot, i.e. there are semantically more complex verbs which can be used in responses to the picture stimuli. Once again, the different languages and varieties show considerable degrees of variation in this area. In order to be able to analyze the frequencypatterns in the verb slot, some additional categories and terms are needed.
40
RAPHAEL BERTHELE
Figure 5: Integration of co-events into a spatial expression (simplified from Talmy 2000b: 212).
Figure 5 enhances the conceptual structure introduced in Figure 1 at the beginning of this contribution. According to Talmy (2000b: 212), speakers tend to integrate more than just a single event into their simple clauses. A ‘macro event’, as displayed in Figure 5, can be seen as a clause that represents a conceptual merger of two events. The syntactic structure of a simple clause mirrors a conceptual integration. Thus, in Figure 5, the verb not only accounts for the (static) activating process, but additionally expresses a subsidiary conceptual entity. This additional content is called a co-event, and in the case of static relations, the co-event most often basically adds information about the dimensional and positional properties of the figure, as expressed by posture verbs such as sit, stand, lie. In addition to that, verbs such as hang, stick, or hide can be used in order to describe the picture stimuli. Some of those coevent verbs thus do not add information about positional properties, but rather about the nature of contact (stick), the physical properties of the ground (swim), or even the figure’s intention (hide; cf. examples [14]–[17]). (14) il pesciolino nuota nell’acqua the fish-little swims in-the water
IT
(15) il flaster tacca vida la comba the bandaid sticks to (at) the leg
ROM_SUS
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
(16) quelqu’un qui se cache derrière un fauteuil somebody who is hiding behind the arm-chair
FR
(17) A däm Mässer chläbt Bluet to (at) this knife sticks blood
BE
41
As shown in examples (14)–(17), the integration of co-events into the verb slot is possible in all varieties. There are two ways of integrating coevents into the predicate: either it is mapped onto the finite verb or onto a past participle. The latter variant, too, is found in all varieties. Examples (18)–(21) are responses to the same stimulus picture — the drawing of a jacket that is hanging on a coat hook. (18) le veston est suspendu à un crochet the jacket is hung-up on a hook
FR
(19) la giacca es pichada vi dal crötsch
RO_VAL
(20) de chittel isch amene haggä ufghänkt
MU
(21) la giacca é appesa al gancio
IT
Just as for the phenomena in the domain of the core schema, the different languages and varieties at issue here vary considerably regarding the extent of co-event integration. In the following paragraphs, the corresponding descriptive statistics will be shown and discussed. In Figure 6, the box-plot conventions are used in order to show the distribution of the co-event verbs. Each informant has a value for the percentage of use of co-event verbs. If an informant uses co-event verbs virtually in every response, his/her value converges towards 1. Conversely, some informants hardly ever use co-event verbs, and their corresponding values are close to 0. Each informant is a case, the hatched boxes represent the middle 50% of all cases. The scattering is expressed by the vertical line, extreme values are marked with asterisks. For every language, the boxes on the left show the amount of co-event verbs (finite), those in the middle the amount of co-event participles, and those on the right the sum of finite verbs and participles. The bold line inside the box is the median, i.e. the value which best represents the whole distribution.
42
RAPHAEL BERTHELE
Figure 6: Percentages of the co-event verb use, grouped according to ‘Abstandsprachen’ (German, French, Italian, Romansh).
The category ‘German’ in Figure 6 includes both SHG and all SG dialects, and, correspondingly, the category Romansh includes all Romansh dialects. Thus, Figure 6 shows clearly that German has a strong tendency to express co-events (in the finite verb). The median line at about 50% suggests that the ‘average’ speaker of German uses co-event verbs in about 50% of all responses. The scattering is very high, from almost 0 to almost 1, which means that some speakers of German hardly ever integrate co-events, whereas others almost always do. The very low values for co-event participles indicate that — although they are possible — co-event participles do not play an important role in this language. If speakers of Italian and French express co-event information at all, they tend to do this using participles. On the whole, as the corresponding boxes in Figure 6 show very clearly, the responses given in these two languages tend to express fewer co-events than those given in German. The differences between German, on the one hand, and Italian or French, on the other, are statistically
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
43
significant (t-test p<0.01). I have no sensible explanation or interpretation for the frequency differences between Italian and French; however, they are not statistically significant. Again, the Romansh data are interesting: on the one hand, Romansh behaves like a ‘decent’ Romance language to the extent that it has much (=significantly) lower values for event-integration than German. This can be seen when we compare the medians of the sum-boxes for French, Italian, and Romansh. However, in the relatively infrequent responses in which the speakers of Romansh do use co-event verbs, they stick, in the majority of the cases, to the ‘German way’. In other words, they use finite verbs and not, as would be the ‘Romance way’, complex predicates with co-event participles. As we have seen in Section 3 above, varieties within the category ‘German’ apparently do not necessarily resemble each other in all aspects that are under investigation in this paper. It seems thus appropriate to take a closer look at the different varieties which make up the German count in Figure 6. In Figure 7, exactly the same tools from descriptive statistics are used as in Figure 6 above. the frequencies here are grouped separately according to the different varieties of German in the sample. At a glance, Figure 7 makes it very clear that the amount of co-event integration varies dramatically across the varieties of German. The speakers of the standard language, both Germans and Swiss, use co-event verbs in about 70% (Swiss) to 80% (Germans) of the responses (cf. the medians for the two categories). What distinguishes the Swiss SHG speakers from the Germans is not so much the average use of co-event verbs, but rather the higher scattering of the cases. This means that some Swiss, when they are speaking SHG, tend to use relatively little co-event verbs. This last finding immediately makes sense if we take into account that all SG dialects display lower co-event values than SHG. If we compare SG (as one group) and SHG (as another group), the differences in the use of finite co-event verbs is very significant (t-test p<0.01). Within the group of SG dialects, there are once more considerable differences. 7 Whereas Bern SG and Sense SG display relatively high co-event values, Muotathal SG scores very low for coevent integration. In other words: in the Muotathal dialect, the most common locating VP is stereotypically headed by the V sein, i.e. be. Although the previously mentioned differences between German SHG and Swiss SHG are not statistically significant, the scattering nevertheless suggests that one of the dialectal features, namely the relatively infrequent use 7
Muotathal dialect is very significantly different from both Bern and Wallis dialects (t-test p<0.01) and significantly different from the Sense dialect (p<0.05).
44
RAPHAEL BERTHELE
of co-event verbs, can interfere with the SHG use of the Swiss (who, as must be noted again, all have a SG dialect as their L1). Even though some Swiss speakers of SHG display relatively low values for this feature, most of them, as a rule, converge towards average SHG choices. To conclude this section, we can state that German on the whole is more inclined to use co-event verbs than the neighboring Romance languages. This tendency, however, applies above all to SHG. The speakers of SG dialects are less likely to use co-event verbs, in some cases co-event verbs average only about 10% of all responses. Romansh behaves, in terms of co-event frequencies, quite like a Romance language but, syntactically, the co-event integration resembles strongly the German pattern. Given the sociolinguistic embedding of the Romansh dialects, it is not very risky to hypothesize a considerable amount of interference with the adstratum, the SG dialects. In the next section, these results will be discussed together with those from the previous section.
Figure 7: Co-event integration frequencies for 6 different varieties of German.
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
45
5.
Discussion In this contribution, several varieties of two typologically different language groups have been compared. Variation in two syntactic-semantic slots has been shown to cross-cut the traditional categories of genetically related languages: if we focus on the feature of co-event integration, some SG dialects seem to look more like Romance languages; in the case of the complexity of the VP, Romansh dialects look more like the SG varieties. This closing discussion of the data presented in Sections 3 and 4 will address the following three points: − Can the variation observed in the data be related to the typological debate about motion events (satellite-framed vs. verb-framed languages)? − Do the observed patterns of variation reflect differences in the cognitive construal of the spatial setting (“thinking for speaking”)? − Are there any explanations for the emergence of the observed differences (particularly SG vs. SHG)? Since the typological debate about the difference between satelliteframed and verb-framed languages primarily addresses issues in the domain of motion verbs, it does not directly concern the data analyzed in this paper. According to Talmy (2000a, 2000b), Slobin (1988, 1991, 1996) and others, satellite-framed languages pack the association function into a particle which, syntactically, is a sister of the verb (such as in English go in/out/up/down), whereas verb-framed languages tend to express the path in the verb (French entrer, sortir, monter, descendre). Since the verb-slot in a satellite-framed language is available for other conceptual content than the association function, it is used very frequently in order to express co-events, whereas the expression of co-event in a verb-framed language is syntactically more complex. As a consequence, verb-framed languages tend to express less co-events than satellite-framed languages (for a cross-linguistic comparison of English and Spanish, see Slobin 1996). The work on satellite- and verb-framed languages has some implications for our data, however, as the issue in this contribution is not exactly the same. The main difference between our static data and motionverb data is that it is — at least in the languages discussed here — very unlikely to find verbs which encode the association function in static spatial relations, and this applies both to the (verb-framed) Romance varieties as well as to the (satellite-framed) German varieties. Undoubtedly, in all responses, the core schema as a whole is located in the PP which is the verb sister. Thus, at least in the case of the static spatial relations, the verb slot is in principle available for additional conceptual content, regardless of the typological
46
RAPHAEL BERTHELE
affiliation of the variety. As has been shown above (Section 4), German varieties on the whole nevertheless tend to integrate more co-events in the verb slot than their Romance counterparts. But, at the same time, some particular German varieties display as little co-event integration as the Romance languages. Given this more complex pattern of resemblance and distinctiveness, it seems that the amount of co-event saliency of a particular variety not only correlates with the typological features of a variety, but with other factors as well. Before we try to hypothesize what kind of factors could govern the lexical and conceptual shape of the verb slot, we need to focus on the link between the variation phenomena observed here and possible corresponding differences in the mental construal of the spatial setting. According to Slobin, satellite- and verb-framed languages do not only have different lexical inventories and different grammars, but those differences favor fundamental differences in construal, i.e. in thinking: I offer this analysis of the constructions associated with motion events as another piece of evidence for the pragmatic functions of grammar. And, further, I suggest that typologies of grammar have consequences for ‘typologies of rhetoric.’ The effects of such typologies on usage may be strong enough to influence speakers’ narrative attention to particular conceptual domains. And, as a consequence, the meanings of verbs of motion in a given language must be considered in the light of both the typological frame of the language and the discourse frame in which such verbs occur. (Slobin 1996: 218)
If we accept the conclusions drawn by Slobin, there must be similar differences linked to the static data discussed in this contribution. As an example, the analyses in Section 4 might suggest that the more elaborate verb slot in the responses from SHG-speakers represents a richer mental construal of the depicted situation. Indeed, speakers of SG dialects and the Romance languages seem to construe the setting using a more schematic representation. The association function, on the other hand, seems to be more profiled in the non-standard varieties of German, but also in the Romansh dialects (Section 5). Clearly, the complementary PP+ADV constructions as found in Romansh and particularly in Wallis SG seem to represent a higher resolution of this association function. The figure is located using two particles which express complementary aspects of the relation between figure and ground. In the case of the redundant constructions, the SG dialects do not seem to be more elaborate, as those constructions are often interpreted as simply emphasizing the spatial concepts (Harnisch 1982: 123; Stadelmann 1978: 216). Even though any thinking-for-speaking hypothesis bears a certain risk of speculation, it
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
47
seems that SG and Romansh focus primarily on the core schema, more precisely the association function, whereas SHG, Standard Italian and Standard French do not profile the association function in the same way. SHG, on the other hand, shows the use of an elaborate repertoire of posture and location verbs, which seems to be a unique characteristic within the sample of varieties treated in this paper. Undoubtedly, more research on the mental representations should be undertaken in order to find out to what extent the differences in form mirror differences in mental construal. The last point to be addressed here is whether there is any explanation or interpretation for the patterns observed in our data. What makes SG dialects simpler in the verb slot and more complex in the PP? Obviously, as discussed before, the typological issue does not help here, since both SG and SHG are undoubtedly satellite-framed languages. Although this type of explanation has to be considered with circumspection, one fundamental difference between SG and SHG could be taken into consideration here: what characterizes our SHG data is lexical variation, i.e. the high diversity of verb-types which occur in the finite verb slot. SG, on the other hand, is characterized by a high amount of redundant expression of the association function. Since SG is overwhelmingly used as an oral variety (speakers of SG use, at least in formal contexts, written SHG), it can be classified as a conceptionally oral variety in the sense of Koch and Oesterreicher (1994–1996). SHG, on the other hand, even in its spoken form, is conceptionally literal. As Reichmann (2002) points out, the historical development of the German standard language must be understood as a gradual process, starting on the basis of a set of conceptionally oral varieties and ending with a dramatically different standardized language in the state of conceptional literacy. According to the framework proposed by Koch and Oesterreicher, a conceptionally literal language is not necessarily written; it may as well be spoken, and vice versa. Since the Swiss diglossia — sometimes labeled a medial diglossia (Kolde 1981: 65ff.) — implies a considerably independent status of the written and the spoken varieties, we can assume that, despite the unquestionable influence of SHG on the SG dialects, the latter still have a genuinely conceptionally oral make-up. Among other features, finally, a conceptionally oral language tends to be more redundant, and it does not show the same degree of lexical differentiation as a conceptionally literal language (Koch & Oesterreicher 1994–1996: 591). There is no doubt that this hypothesis has to be tested with additional data and probably different methods, but in my view the dichotomy of conceptional orality/literacy seems to be a promising path to a better understanding of the phenomena discussed in this paper. In the introductory remarks to this contribution, I stated that the research presented here could contribute both to dialectology and to cognitive and
48
RAPHAEL BERTHELE
typological linguistics. It seems to me that this contribution shows clearly that the latter offer theoretical frameworks which can easily and fruitfully be adapted for dialectological research. Conversely, dialectological and variationist thinking and analyzing involves a genuine questioning of language categories such as ‘German’, ‘Spanish’, etc. As has been shown in this contribution, substantial amounts of variation are neglected if we do not ‘zoom in’ into the different varieties of a particular language. And even within the varieties singled out, rather than simply comparing average uses of a particular feature, spaces of individual variation have to be explored. I am convinced that this is the contribution variationist linguistics in general can make to the fascinating investigation into the interdependency of thinking, speaking, and linguistic form. There is no doubt that this contribution is only a modest beginning for such an interdisciplinary enterprise. References Becker, A. 1994. Lokalisierungsausdrücke im Sprachvergleich: eine lexikalisch-semantische Analyse von Lokalisierungsausdrücken im Deutschen, Englischen, Französischen und Türkischen. Tübingen: Niemeyer. Berthele, R. 2003. “The typology of motion and posture verbs: a variationist account”. Dialectology Meets Typology (= Trends in Linguistics) ed. by B. Kortmann, 93–126. Berlin and New York: Mouton de Gruyter. ----------. In press. “Spatial reference in an endangered Romance language: The case of Romansh”. Proceedings of the Berkeley Linguistics Society (BLS), 2003. Bickel, H. & R. Schläpfer. 2000. Die viersprachige Schweiz. 2., neu bearbeitete Auflage. Aarau: Sauerländer. Bloom, P., M. Peterson, L. Nadel & M.F. Garrett, eds. 1996. Language and Space. Cambridge, Massachusetts and London, England: The MIT Press. Bowerman, M. 1996. “Learning how to structure space for language: A crosslinguistic perspective”. Bloom et al. 1996. 385–436. Bowerman, M. and S. Choi 2001. “Shaping meanings for language: universal and language-specific in the acquisition of spatial semantic categories”. Language Acquisition and Conceptual Development ed. by M. Bowerman & S.C. Levinson, 475–511. Cambridge: Cambridge University Press. Carigiet, W. 2000. “Zur Mehrsprachigkeit der Bündnerromanen”. Bickel & Schläpfer 2000. 235–239. Harnisch, K-R. 1982. “Doppelpartikelverben als Gegenstand der Wortbildungslehre und Richtungsadverbien als Präpositionen”. Tendenzen
STATIC SPATIAL RELATIONS IN GERMAN AND ROMANCE
49
verbaler Wortbildung in der deutschen Gegenwartssprache ed. by L. Eichinger, 107–133. Hamburg: Buske. Kolde, G. 1981. Sprachkontakte in gemischtsprachigen Städten. Wiesbaden: Steiner. Koch, P. & W. Oesterreicher. 1994–1996. “Funktionale Aspekte der Schriftkultur”. Writing and Its Use: An Interdisciplinary Handbook of International Research ed. by H. Günther & O. Ludwig, 587–604. Berlin and New York: de Gruyter. Langacker, R. W. 1990. Concept, Image, and Symbol. The Cognitive Basis of Grammar. Berlin and New York: Mouton de Gruyter. Levinson, S. C. 1996. “Language and Space”. Annu. Rev. Anthropol. 25. 353– 382. ----------. 1996b. “Frames of reference and Molyneux’s question: Crosslinguistic evidence”. Bloom et al. 1996. 109–169. McIntyre, A. 2001. German Double Particles as Preverbs. Morphology and Conceptual Semantics. Tübingen: Stauffenburg Verlag Olsen, S. 1996. “Pleonastische Direktionale”. Wenn die Semantik arbeitet: Klaus Baumgärtner zum 65. Geburtstag ed. by G. Harras & Bierwisch, 303–328. Tübingen: Niemeyer. Pittner, K. 1999. Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpretation. Tübingen: Stauffenburg Verlag. Rash, F. 1998. The German Language in Switzerland: Multilingualism Diglossia - Variation. Bern: Lang. Reichmann, O. 2002. “Die Entstehung der neuhochdeutschen Schriftsprache: Wo bleiben die Regionen?” Die deutsche Schriftsprache und die Regionen. Entstehungsgeschichtliche Fragen in neuer Sicht ed. by R. Berthele et al., 29–56. Berlin and New York: de Gruyter. Rowley, A. 1980. Sprachliche Orientierung: Untersuchungen zur Morphologie und Semantik der Richtungsadverbien in oberdeutschen Mundarten Bayreuth: Sprach- und Literaturwissenschaftliche Fakultät Universitat Bayreuth. Slobin, D. I. 1991. “Learning to think for speaking. Native language, cognition, and rhetorical style”. Pragmatics 1.7–26. ----------. 1996. “Two ways to travel: Verbs of motion in English and Spanish”. Grammatical Constructions: Their Form and Meaning ed. by S. Thompson & M. Shibatani, 195–217. Oxford: Oxford University Press. Slobin, D. I. & A. Bocaz. 1988. “Learning to talk about movement through time and space: The development of narrative abilities in Spanish and English”. Lenguas Modernas 15.5–24.
50
RAPHAEL BERTHELE
Stadelmann, M. 1978. Orts- und Richtungsadverbien bei Siedlungsbezeichnungen. Tübingen: Niemeyer. Talmy, L. 1985. “Lexicalization patterns: semantic structure in lexical forms”. Language Typology and Syntactic Description. Volume III: Grammatical Categories and the Lexicon ed. by Timothy Shopen, 57–149. Cambridge: Cambridge University Press. ----------. 2000a. Toward a Cognitive Semantics. Volume I: Concept Structuring Systems. Cambridge, Massachusetts and London, England: MIT Press. ----------. 2000b. Toward a Cognitive Semantics. Volume II: Typology and Process in Concept Structuring. Cambridge, Massachusetts and London, England: MIT Press. Wälchli, B. 2001. “A typology of displacement (with special reference to Latvian)”. Sprachtypologie und Universalienforschung 54:3.298–323.
INGRESSIVE PARTICLES ACROSS BORDERS GENDER AND DISCOURSE PARALLELS ACROSS THE NORTH ATLANTIC
SANDRA CLARKE & GUNNEL MELCHERS Memorial University of Newfoundland & University of Stockholm Pulmonic ingressive articulation — that is, speech produced when inhaling rather than exhaling air from the lungs — represents an articulatory mode which, because of its very nature, cannot be sustained for speech sequences of more than a few seconds. Its unusual nature as a means of speech production makes it fairly rare in occurrence across the linguistic spectrum, and ensuingly highly marked from a typological perspective. It is also a feature that has come in for little scrutiny in the literature. 1 In this paper we examine one of the several uses of pulmonic ingressive speech: ingressive discourse particles. We establish these forms as an areal feature, and point to the existence of a zone stretching from the eastern Baltic to the Atlantic seaboard of the United States, across which ingressive particles have been transmitted via language contact. More importantly, we show that the cross-language-boundary diffusion of these particles has been accompanied by the diffusion of social and pragmatic constraints on their usage. Though this phenomenon plays an obvious role in language-internal, cross-dialect transmission (as has been amply demonstrated by both social and regional dialectology), it is one that has rarely been noted in the language contact literature. 1.
Background The few publications that deal with pulmonic ingressive articulation suggest that, cross-linguistically, ingressive speech serves a small number of well-defined paralinguistic functions. Two of these functions appear grounded in physical or biological factors. The first involves the gasp-like inhalation of breath that registers strong emotion, including surprise, pain, or fear. When 1
Other types of ingressive articulation (i.e. glottalic or velaric), which are used in a small number of the world’s languages for individual sound production (e.g. clicks), are better known in the literature. Citing Hockett (1955), Shorrocks (2003) also mentions the use of the pulmonic ingressive airstream to articulate b- and d- like sounds in Maidu, a Penutian language spoken in California.
CLARKE & MELCHERS
52
verbalized, this results in such ingressively articulated phrases as English Oh my God or What?. In languages like Swedish, longer emotive sequences may also be articulated ingressively; an example of this, on the topic of Caesarean sections, appears in (1) below (where ingressives are provided in italics, within angled brackets): 2 (1)
nej de skiter man fullständigt i, men just detta att man har så ont så man <å herre gud det var ju läckert> det var därför hon var så stor om skallen “no, you don’t give a shit, but it’s just that you’re in such pain that
that’s why her head was so big”
A second function involves routine tasks to which little attention is devoted, in particular counting ‘under one’s breath’. Though possibly widespread, this has been documented primarily for the Germanic languages (e.g. Crystal 1987, Dieth 1950). As Hill and Zepeda (1999) point out, the use of alternating ingressive and egressive airstreams for this task may constitute a means of avoiding strain on lung capacity. Two other paralinguistic functions of ingressive speech have been noted in the literature, both involving social and pragmatic factors. The first is the use 2
This example has been taken from the Göteborg University Spoken Language Corpus of Swedish (e.g. Allwood et al. 2000), our major source of data for ingressive use in Swedish. For permission to use this corpus we are extremely grateful to Jens Allwood; Leif Grönqvist deserves special thanks for all his assistance with the practical problems of access. We also owe special thanks to Amanda Newhook, Danny Snow and Faith Wilkes, whose tapes constitute our Newfoundland corpus, as well as to Glenn Peters for his transcription of the Newfoundland data. Our paper would not have been possible without input from many sources. We thank Nazuki Kobayashi for providing us with a copy of her thesis on Norwegian ingressives (as well as Lise Opdahl for bringing this thesis to our attention); Robert Eklund for making available a pre-publication version of his 2002 paper on Swedish ingressives; Ragnhild Söderbergh and Åsa Nordqvist for permitting us access to their corpus of Swedish child speech; and Mark Jones for the SED ingressive examples from Patterdale, Cumbria. For information on the occurrence of ingressive variants in their respective regions, we would like to thank Joan Beal (England), Barbara Bird (the Hebrides), Julie Brittain (Scotland), Wladyslaw Cichocki (New Brunswick), Karen Corrigan (Ireland), Peter French (England), Terry Gunnell (Iceland and the Faroes), Caroline Macafee (Scotland), Michael Montgomery (the southern U.S. and Appalachia), and Peter Trudgill (England). We are also grateful to Jim Black and Doug Walker for information on French, and to a number of native speakers for information on Estonian and Russian — not to mention relatives and friends located in various parts of the world, from Hungary to Tennessee to Australia. Finally, we are very thankful to Olle Engstrand for placing a query about ingressives on the (Scandinavian) Phonetics list, and to the various people who responded — in particular, Claes-Christian Elert, who provided us with valuable references.
INGRESSIVE PARTICLES ACROSS BORDERS
53
of ingressive articulation for the purpose of voice disguise and identity concealment, whether on the part of male suitors engaged in courtship (e.g. Conklin 1959 for the Philippine language Hanunóo; Dieth 1950 for Swiss German), or in folk traditions involving disguised participants — notably the ‘julebukk’ in mainland Scandinavia along with the Grýla tradition in the Faroes and possibly Iceland (Gunnell 1995, 2001), ‘skakling’ or ‘skekling’ in Shetland (Gunnell 2001; Johnson 1962), and Christmas ‘mummering’ in Newfoundland (e.g. Kirwin 1971). The second, which constitutes the focus of this paper, is the use of ingressively-articulated discourse particles, which in the relatively sparse literature on the topic occur most frequently as variants of the words for yes and no. In the Scandinavian languages — the language subgroup in which their use is best known — these discourse particles appear as ja ‘yes’ (with such variants as Norwegian [eja:], Norwegian and Swedish jaha, northern Swedish jo); and as nej ‘no’(spelled nei in Norwegian), which also has a number of variants, among them Swedish nää. Ingressive ja — like still another Scandinavian ‘yes’ variant, mm or mmhmm — typically represents agreement with a speaker’s affirmative statement, as in the following example from the Göteborg Spoken Language Corpus (where F = female speaker and M = male speaker): (2)
F: ja(g) tror de M: ja de e dom F: <ja>
“I think so” “Yes, they are” “”
The negative particle nej signifies agreement with a speaker’s negative statement, as in the following Swedish example, collected via participantobservation: (3)
F: Finns det några textilfabriker kvar i Manchester? “Are there any textile factories left in Manchester?” M: Det är inte mycket. “Not many.” F: “<no>”
As a variant of ja, standard Swedish and Norwegian also possess a (variably ingressive) particle jo, typically used in response to a negative query when the conversational interactants share the assumption or knowledge of the truth value of an event. A Swedish participant-observation example is provided in (4), and a Norwegian example in (5):
CLARKE & MELCHERS
54 (4)
Vet du inte att Sverige vann? Jo. “Don’t you know that Sweden won? Yes (I do).”
(5)
Så du ikke din far? Jo. (Peters 1981: 26) “Didn’t you see your father? Yes (I did).”
In the Scandinavian languages, pulmonic ingressive articulation also regularly occurs for a number of evaluative particles (e.g. Swedish bra ‘good’; jorå/jadå, signalling semi-reluctant agreement; cf. Eklund 2002), as well as for such frequently-occurring discourse formulae as tack så mycket (‘thanks a lot’) and ja vet inte (‘I don’t know’). More rarely, it is used to articulate sentential constituents, as in the Swedish example in (6) below, from a story about cats that appears in the Göteborg Corpus: (6)
... ut mot hallen (sighs) ja å <så sprang> ju katterna ut å in här “out towards the landing (sighs) yes and the cats in and out here, you know...”
Whether or not sequences of this last type result from physical considerations (the need to take a breath while continuing to speak) or are governed by other factors (cf. example [1] above) is an issue that merits further investigation. In any case, the Göteborg Corpus suggests that in Swedish such sequences are relatively rare — indeed, in this corpus we uncovered only 217 occurrences (mostly monosyllabic in nature) that were other than discourse particles, out of a total of 1563 ingressive articulations. Though it has been barely documented, pulmonic ingressive articulation is also found in a number of varieties of spoken English, where its use is largely restricted to the particles yeah and no. 3 The following examples from Newfoundland English, obtained via participant-observation, illustrate usage of confirmatory yeah and no (cf. examples [2] and [3] above) after, respectively, affirmative and negative statements: (7)
3
M: The other window has to be painted. F: Well, it’s a good day to do it. M:
While ingressive articulations of yes (in which the glide and vowel are articulated ingressively and the s egressively) do sporadically occur in some varieties of English, it is ingressive yeah which is the focus of this paper.
INGRESSIVE PARTICLES ACROSS BORDERS
(8)
55
F1: Bubble lights [=a Christmas tree decoration] are coming back. F2: They weren’t very popular for a while, were they? F1: <no>
Ingressive discourse particles as an areal feature Our experience suggests that even in areas where their use is fairly common, ingressive discourse particles tend to be favoured by certain segments of the population, and surface in particular registers; they also tend to remain beneath the level of consciousness for their users, at least until pointed out by others. It is therefore quite possible that they may go unreported. That said, our survey of all the available literature on ingressive speech suggests that outside of the well-delimited geographic area to be discussed below, discourse particles articulated via the pulmonic ingressive airstream have been documented only for Tohono O’odham, a Uto-Aztecan language spoken in the southwestern United States and Mexico (Hill and Zepeda 1999). Pulmonic ingressive articulation in Tohono O’odham appears to resemble the situation found in the Scandinavian languages, in that it is not restricted to discourse particles, but also occurs with full clauses. 4 However, the historical and language contact literature provides no reason to suspect that this is other than a coincidental development. Otherwise, the existing literature on ingressive articulation clearly points to Scandinavia, and the Nordic area in general, as the focal area of ingressive discourse particle use. This phenomenon has been described in some detail by Peters (1981) for mainland Scandinavian languages, Stølen (1994, 1995) for Danish, Kobayashi (2001) for Norwegian, and Eklund (2002) for Swedish. Ingressive particles have also been noted for the Scandinavian-settled North Atlantic islands of the Faroes and Iceland (e.g. Peters 1981), and are likewise found in Finland (e.g. Hakulinen 1993). What is less clear from the literature, however, is that there is an obvious pattern of linguistic diffusion of this feature which encompasses a much wider geographical area, the eastern extremity of which is constituted by the eastern Baltic; the current southern extremity, by Switzerland and Austria; and the western extremity, the eastern seaboard of the United States. For want of a better term, we will call this area the ‘North Atlantic/Baltic Zone’. For the eastern edge of this Zone, the eastern Baltic, we are not aware of literature on the topic, apart from a mention in Steinbergs (1993). We have 2.
4
Use of pulmonic ingressive articulation outside of discourse particles has also been noted for a handful of the world’s languages. Japanese, for example, possesses an ingressive whistle-like or hiss-like sound that expresses disapproval (Kobayashi 2001).
CLARKE & MELCHERS
56
however independently confirmed from native speakers the existence of ingressive yes equivalents in Estonian and Latvian, as well as in the Russian spoken in the area of St. Petersburg. Linguistic diffusion from the Nordic area also appears to account for the appearance of ingressive particles further south in Europe: as Pitschmann (1987) notes, ingressive ja is found in northern Germany, occurs to a lesser extent in the south of Germany, and, within perhaps the past thirty or so years, has spread to younger speakers in Austria. Though apparently not overly common, ingressive ja has also been noted in Dutch (Rietveld & van Heuven 1997). And while ingressive oui occurs in French (e.g. Léon 1992), ingressive particles have not been documented for other Romance languages. Ingressive discourse particles have also been recorded in the British Isles, particularly on the coastal and Celtic fringes — that is, in the language varieties spoken in Scotland, Ireland and Wales. Thus Peters (1981) records (infrequent) ingressive articulations of yeah in observational data from Ireland and Wales; and in at least some parts of Scotland (e.g. the areas around Glasgow and Aberdeen), aye may be articulated ingressively. Ingressives also occur in both the Gaelic and English spoken in the Hebrides, where their use is apparently not limited to discourse particles. 5 Within England itself, however, the distribution of ingressive particles appears to be restricted to the northern region. Thus they were not documented in the mid-20th century Survey of English Dialects (SED), apart from a single male speaker from Patterdale in Cumbria, who was recorded in the 1970s. This speaker used ingressive aye and neh, as in the following examples: (9)
I mow the lawns and do fencing, anything what I can do, aye, , that's right
(10) ... no, it wasn't really, wasn't economical, eh, <(n)eh>, aye Shorrocks (2003) also notes reports of ingressive particle use in various parts of northern England (e.g. Durham, Teeside, East Yorkshire). While the distribution of ingressive discourse particles is minimally documented within the British Isles and Ireland, it is even less well-known that this region does not constitute their western geographic boundary. These 5
We have confirmation of ingressive use in Hebrides Gaelic from Barbara Bird (p.c.). However, there are no single particles for yes and no in the Celtic languages: thus yes is typically realized by a verbal form equivalent to ‘it is’. Spoken Gaelic varieties offer an obvious target for future investigation of pulmonic ingressive articulation patterns.
INGRESSIVE PARTICLES ACROSS BORDERS
57
particles also characterize some of the varieties of English spoken along the eastern seaboard or Atlantic coast of North America. Here, the southern boundary of ingressive use appears to be the northern United States, and specifically, New England. Ingressive yeah has been documented by Peters (1981) in a coastal fishing village in Maine (Vinalhaven); Peters also noted its occurrence elsewhere in the state. Ingressive particles were also recorded in five different Massachusetts communities by the Linguistic Atlas of New England (LANE; Kurath et al. 1943). Across the Canadian border, the only reference in print that we could uncover for the three Canadian Maritime provinces was Pratt’s 1988 Dictionary of Prince Edward Island English, which notes ingressive yeah to be a feature of the provinces of Prince Edward Island and New Brunswick. We ourselves have also confirmed its existence in the neighbouring Maritime province of Nova Scotia, and particularly in its insular portion, Cape Breton Island. As has been observed by several writers (Kirwin 1971, Paddock 1981, Steinbergs 1993, and, most recently, Shorrocks 2003), ingressive yeah and no are also common on the island of Newfoundland which, along with mainland Labrador, constitutes Canada’s most easterly province. Though we have occasionally come across users of ingressive yeah in central Canada (e.g. Toronto), the phenomenon appears to be extremely rare outside the Atlantic area. It has however been recorded in one inland Canadian location, namely, the region of Ottawa in the province of Ontario (Woods 1999). The geographical distribution of ingressive particles across the North Atlantic/Baltic Zone clearly suggests a common heritage, and diffusion by language contact across linguistic boundaries. At the northeastern periphery of the Zone, language contact is evident in the presence of ingressive particles in Finnish and Estonian, yet they are absent from a more southerly member of the Finno-Ugric family, Hungarian. The spread of pulmonic ingressive articulation at the western edges of the Zone finds a ready historical explanation, in the form of early westward expansion by the Vikings to the Faroes, Iceland, Scotland, England and Ireland. Indeed, the areas of Britain in which ingressive usage is currently attested are areas which in medieval times came under varying degrees of Scandinavian influence and settlement; the apparent result was the diffusion of ingressive articulation into both the Gaelic and English varieties spoken in these regions. From there, ingressive articulation was in all likelihood carried to the eastern seaboard of North America by Scottish and Irish emigrants, where it has been preserved in peripheral, and largely rural, areas in which these emigrants constituted the numerically dominant founder populations. In Canada, these areas include southeastern Newfoundland, settled heavily by migrants from the southeastern counties of Ireland; Cape Breton
58
CLARKE & MELCHERS
Island, which had a high proportion of highland Scots among its founder population; and the rural Ottawa Valley area of Ontario, which likewise received considerable Irish and Scots-Irish in-migration. As to the eastern seaboard of the United States, in one of the two American states in which ingressive discourse particles have thus far been documented — the fishing community of Vinalhaven, Maine — Peters (1981) attributes their presence to Scandinavian settler input. The above account of transatlantic diffusion notwithstanding, a number of points relating to the transmission of ingressive discourse particles to North America remain unclear. The region of the United States that received the highest proportion of Scandinavian immigrants — the Upper Midwest — does not seem to have preserved ingressive usage: for example, there is no record of this feature in the Linguistic Atlas of the Upper Midwest (LAUM) (Allen 1973, 1975). That the absence of ingressives in this region may not be merely an artifact of the data collection methods used by LAUM receives confirmation from several other sources. As noted by Peters (1981: 22), no ingressive pronunciations were recorded by Einar Haugen in his investigation of the English spoken by Americans of Norwegian ancestry. Likewise, the second author of the present article, Gunnel Melchers, did not come across this feature in her interviews with Swedish immigrants to the U.S. (primarily in Minnesota); nor did she find it among speakers from the Bishop Hill Swedish colony in Illinois, mostly second- or third-generation, whose variant of choice was the American-like ‘uh-huh’. Ingressive particles also do not occur in at least one relatively isolated variety of American English the ancestry of which is primarily Scots-Irish, namely, Appalachian English. Peters’ (1981) proposed Scandinavian origin of ingressives in Vinalhaven, Maine, is also somewhat problematic, in that the founder populations of this community were overwhelmingly British rather than Scandinavian. Similarly, the Massachusetts communities documented by LANE as displaying ingressive particles appear to have been initially settled by emigrants from England. In Newfoundland, ingressive usage is perhaps even more in evidence in rural coastal areas settled almost exclusively by migrants from West-Country England rather than southeastern Ireland. Of course, as a lexically-transmitted feature, ingressive particle articulation could well have diffused throughout the island in much the same way as have a number of Irish lexical items (though not typically phonological and grammatical features; see Clarke 2004). In short, these tantalizing bits of evidence invite speculation as to whether, in previous centuries, ingressive discourse particles might have occurred in a wider geographical portion of England than is suggested by the current situation. Of course, for the moment, this must remain no more than speculation.
INGRESSIVE PARTICLES ACROSS BORDERS
59
Whatever the precise transatlantic connections, the evidence from settlement history and population contact strongly suggests that we are dealing with an areal phenomenon rather than a feature which arose independently in communities on both sides of the Atlantic. The language contact or diffusion position is considerably strengthened by the cross-linguistic parallels in social and pragmatic function which can be observed from one end of the North Atlantic/Baltic Zone to the other. It is to these parallels which we now turn. Our investigation focuses on two linguistic varieties located towards the eastern and western peripheries of the Zone: Swedish and Newfoundland English, respectively. 3.
Ingressive usage in the Swedish and Newfoundland English taped corpora With the exception of several recent studies (Eklund 2002, Kobayashi 2001), documentation of ingressive particle use derives from non-taperecorded observational data. In the present study, however, we rely primarily on tape-recorded corpora. For Swedish, we utilized the Göteborg University Spoken Language Corpus of Swedish. This corpus — the collection of which began in the 1980s and is still ongoing — was designed to include representation of a full range of registers and speech genres, as well as of various regions throughout Sweden. At the time of our study it contained 1,263,408 word tokens. Our Newfoundland English (NE) taped corpus was considerably smaller. It consisted of approximately 50,000 words, or five hours of tape-recorded speech, both same- and cross-sex, representing three separate and largely rural areas of the island: the West-Country-English-settled southwest coast; the central region; and the Irish-settled southeastern area, south of the capital city, St. John’s. Unlike the Göteborg corpus, the NE data consisted entirely of informal small-group interaction, involving an inside interviewer; the groups contained between four and six participants, the majority aged over 40. The recordings which constitute the NE corpus were specifically selected because they contained pulmonic ingressive articulation; a number of NE tapes surveyed — while representing informal conversational style and participants of similar backgrounds — simply did not exhibit any ingressive particles, though many examples of their egressive counterparts. This points to the existence of considerable interspeaker variability in ingressive use, a fact which we have independently confirmed both in Newfoundland and in Sweden. Our analysis has been supplemented with participant-observation data collected over the past number of years in Newfoundland and Sweden by each
CLARKE & MELCHERS
60 Type ja + jo
Egressive tokens Ingressive tokens 39315 (2068 of which = jo)
1082 (9)
Total per type 40397 (2077)
% ingressives (of total) 2.7
mm/mmhmm 12340 114 12454 0.9 nej 7558 150 7708 1.9 Overall total 59213 1346 60559 2.2 Table 1: The Göteborg Spoken Language Corpus of Swedish – Overall Summary.
author, both of whom are native speakers of ingressive varieties. Our records have proven very useful in corroborating the observations that have emerged from the taped corpora, particularly with respect to the range of discursive meanings conveyed by ingressives. We began the investigation of our taped corpora by examining the relative rates of occurrence of ingressive variants of yeah and no equivalents: i.e. ja/jo, mm and nej for Swedish, and yeah, mm and no for NE. This we calculated by dividing the number of ingressive variants by the total number of tokens of each particle, i.e., by the number of ingressive + egressive articulations. Results for Swedish are presented in Table 1, and for NE in Table 2. The overall Swedish rate of only 2.2% ingressive discourse particle use stands in marked contrast to the c. 18% usage in the NE corpus. Yet this discrepancy is undoubtedly the result of the style and genre differences between the corpora. Our NE data consist entirely of informal interactional speech among family members and acquaintances, while the Göteborg corpus contains a wide range of speech genres. In fact, its informal conversational component represents only about 5% of the total, or some 7 ½ hours of the c. 118 hours of total recording; substantially larger portions (just over 29, 15 and 14 hours respectively) are represented by interviews, discussions (including academic, TV, radio), and formal meetings. Speech of an essentially monological nature is also represented in the Göteborg corpus, and the fact that ingressives are very much an interactive device is borne out in that virtually the Type
Egressive tokens
Ingressive Total per type % ingressives tokens (of total) yeah 1926 522 2448 21.3 mm/mmhmm 428 21 449 4.7 no 327 49 376 13.3 Overall total 2681 592 3273 18.1 Table 2: The Newfoundland English Taped Corpus – Overall Summary.
INGRESSIVE PARTICLES ACROSS BORDERS
F, age 72 Particle
F, age 70
F, age c. 45 N Ingr % Ingr Total
F, age 23
Total
N Ingr
yeah
811
325
40.1
570
93
16.3
204
38
18.6
258
7
2.8
mm
105
9
8.6
61
2
3.3
56
10
17.9
85
0
0.0
no
114
31
27.2
99
7
7.1
13
4
30.8
28
0
0.0
1030
365
35.4
730
102
14.0
273
52
19.0
371
7
1.9
Overall total
% Ingr Total
61
N Ingr % Ingr Total
N Ingr % Ingr
Table 3: Interspeaker Variability in Ingressive Use, Newfoundland English (Partial) Corpus. [Key per speaker: col 1 = total n of ingressive + egressive tokens; col 2 = n of ingressive (Ingr) tokens; col 3 = percentage of tokens realized ingressively]
only monologue containing ingressive particles is a lecture on linguistics in which the lecturer discusses ingressives and produces illustrations of them. Nonetheless, there are obvious parallels between the two corpora: in both, the highest ratio of ingressive usage is found for yeah/ja, followed by no/nei, with affirmative mm or mmhmm displaying the lowest ingressive ratio. The NE corpus confirms a point made above, namely the existence of considerable interspeaker variability in ingressive use, even among speakers from virtually identical regional and social backgrounds. Table 3 summarizes ingressive rates in one component of the NE corpus, an all-female group interview in a tiny rural community on the southwest coast of Newfoundland; the total of 2404 discourse particles in this approximately one and a half hour recording displays an overall ingressive rate of almost 22%. The two older women represented in the leftmost columns are almost identical in age and social background, have spent their entire lives in the community, and interact closely. Nonetheless, they differ considerably in the degree to which they favour ingressive as opposed to egressive articulation: the ingressive rate of the 72-year old is two and a half times that of the 70-year old. This same table also suggests a further observation, one which corroborates the intuitions of both authors, although it must remain tentative because of lack of adequate data. This is the fact that in at least some areas of the ‘North Atlantic/Baltic Zone’, ingressive use appears to be declining among younger and more urban or urbanized speakers. Thus in Table 3, the 23-year old — a community member who moved to the city to obtain post-secondary education — uses ingressives at the extremely low rate of just under 2%, even in this extremely informal situation. 6 6
This observation should not be taken to mean that children in Sweden and Newfoundland are no longer acquiring ingressive particles. Though we have no recent child data, we note that there is some evidence that children may acquire these at a very early age: thus the ‘Dollhouse Play Corpus’, assembled in the 1970s by Professor (now Emeritus) Ragnhild Söderbergh of the
62
CLARKE & MELCHERS
4.
Cross-linguistic gender correlates within the North Atlantic/Baltic Zone In both Swedish and English, then, ingressive discourse particles seem to surface primarily, though by no means exclusively, in informal interactional contexts. What is considerably more striking, however, are the social — as well as pragmatic or discourse — parallels that emerge from examination of ingressive usage not only in our corpora, but all across the Zone. In this section, we deal with gender parallels. The existing literature indicates that ingressive particles are used throughout the North Atlantic/Baltic Zone by men as well as women. However, those observation-based ingressive studies that have dealt with gender — among them Peters (1981) for Norwegian (as well as the English of the U.S. state of Maine); Stølen (1994, 1995) for Danish; Pitschmann (1987) for German and Austrian German — have noted an obvious gender difference. Strikingly, this difference is always in the same direction: greater use of ingressives on the part of female speakers. In fact, ingressives seem to be a gender-stereotyped feature: despite the fact that most speakers seem to use ingressive particles unconsciously, there is none the less awareness in at least some regions of their social value as gender markers. Pitschmann (1987), for example, has noted that in some parts of Germany, it would be considered effeminate for men to use ingressive ja. For Newfoundland, Paddock (1981) impressionistically links ingressive speech to women. Steinbergs’ (1993) observational sample of 26 adultspeakers of various ages from the Newfoundland capital, St. John’s, bears this out; in fact, Steinbergs noted that no male members of her sample under the age of 40 made use of ingressive discourse particles. In mainland Canada, Woods’ survey of Ottawa English (Woods 1999) recorded a mere 18 tokens of ingressive speech, from a total of only five speakers in a 100-speaker corpus;
University of Göteborg, involves verbal interactions of Swedish three-year olds and their mothers, and provides clear examples of ingressive ja use among the children. In addition, our observation is not meant to imply that there is no ongoing geographical expansion of ingressive particle usage within the ingressive Zone. In at least one area, Austria, Pitschmann (1987) documents ingressive particle introduction from Germany via the often-observed sociolinguistic mechanism of adoption of a supralocal form by teenage girls and young women. In areas like Atlantic Canada, however, younger speakers are coming into more extensive contact than did their elders with speakers of North American non-ingressive dialect types, a factor which might well clarify the apparent reduction of ingressives among younger generations — particularly when these are linked, as they are for some, with a less prestigious local speech variety.
INGRESSIVE PARTICLES ACROSS BORDERS
Particle ( ) = total n of tokens, egressive + ingressive ja (38320) jo ( 2077) mm (12454) nej ( 7708) Overall total
Total n of ingressive N of ingressives by females tokens
63
Percentage of ingressives produced by females
1073 592 9 6 114 110 150 78 1346 786 Table 4: Results for Gender - Göteborg Corpus of Swedish.
55.2 66.7 96.5 52.0 58.4
of the five, all but one were female, and these females accounted for almost 90% of ingressive tokens. Our corpus data representing both Swedish and Newfoundland English confirm the gender pattern. Table 4 presents results for the Göteborg corpus of Swedish. In this corpus, though they were slightly under-represented relative to men, women not only used more ingressive variants overall (just over 58% of the total), but also consistently produced a higher ratio of ingressive articulations than men for each of the four categories of discourse particles examined. The NE results in Table 5 are even more gender-marked: the ratios of ingressive use in the two all-female interactional segments of the corpus (approx. 22% and 14%) are considerably greater than the ratios in mixed-sex interaction. And while in interaction 3 males accounted for somewhat more ingressive tokens than did women, in interaction 4 they produced not a single ingressive articulation, even though they readily used egressive variants of yeah and no. The Newfoundland results for gender are corroborated by Kobayashi’s recent (2001) study on the use of ingressive ja in Norwegian. This study Group type 1. All-female 2. All-female 3. Mixed
Total n of yeah, mm, no
Total n of ingressive tokens
% ingressives
2404 135 449
526 19 33
21.9% 14.1% 7.3%
% ingressives by gender
— — M 8.0 (26/324) F 5.6 ( 7/125) 4. Mixed 285 14 4.9% M 0.0 ( 0/85) F 7.0 (14/200) Table 5: Results for Gender - Newfoundland English Corpus. [Overall ingressive use in mixed-sex groups: F 6.5% (21/325); M 6.4% (26/409)]
64
CLARKE & MELCHERS
Dyad type
N egressive ja N ingressive ja Total ja tokens % ingressives (of tokens tokens total) Mixed 277 9 (9 F; 0 M) 285 3.2 Mixed 266 48 (45F; 3M) 314 15.3 All-male 617 17 (3; 14) 634 2.7 All-female 348 41 (20; 21) 389 10.5 Overall total 1508 115 1623 7.1 Table 6: Kobayashi’s (2001) Norwegian Results for Same- and Mixed-Sex Dyads. (Numbers within brackets represent n of ingressives per dyad participant.)
involved two and a half hours of taped face-to-face unstructured and informal dyadic conversations between young-adult strangers. As in the case of our NE data, two of these settings involved mixed-sex interaction, and two, same-sex. 7 Of the total of 116 ingressive ja occurrences in Kobayashi’s data, an overwhelming 82.6% (n=95) were produced by women. As Table 6 shows, however, this is not because male participants did not make use of minimal responses in their verbal interaction. In fact, the all-male dyad produced the highest number (634) of ja tokens of any of the dyads, yet at the same time the lowest percentage (under 3%) of ingressive ja variants. In the mixed-sex dyads, the discrepancy between male and female ingressive use was particularly obvious. Women produced from nine to fifteen times as many tokens of ingressives as did male participants; that is, they were responsible for 94% and 100% of the ingressive particles that occurred in the two cross-gender interactions. The above results from Swedish, Norwegian and Newfoundland English offer clear support for the hypothesis that cross-linguistically, ingressive discourse particles carry gender marking. And, as will be shown below, the ingressive variants of discourse particles for yes and no also carry a pragmatic meaning of interactional involvement. Since cross-culturally, as Hill and Zepeda (1999: 38), among many, have pointed out, “a high level of attention to interlocutor needs” is a fundamental part of the construction of female gender identity, it is not surprising that ingressive particles would have been transmitted across linguistic borders not simply as an articulatory type, but also with appropriate gender-related meaning. The results outlined above should however by no means be taken to imply that ingressive discourse particles are the exclusive domain of women. 7
A difference between Kobayashi’s data and ours, however, is that we were unable to include in our NE taped corpus an example of an all-male interaction involving ingressives. Yet ingressives do occur in all-male conversation in Newfoundland.
INGRESSIVE PARTICLES ACROSS BORDERS
65
As several studies have shown (e.g. Eklund 2002, Peters 1981), these particles, like their egressive minimal response counterparts, are readily used by males to provide verbal feedback. For both genders, use of ingressive particles is obviously also role-related: hence Eklund’s (2002) finding of ingressive usage in an all-male setting in which verbal feedback is in order, namely, one involving a relationship of client to service provider. Indeed, in Eklund’s overall results — i.e., including both same- and cross-sex interactions — 82% of ingressives (68 out of 83) were produced by those cast in the role of client rather than service provider (in this case, travel agent). Different interactional roles assumed by males in interactions 3 and 4 of Table 5 undoubtedly account for the slight discrepancies in findings within the Newfoundland data: the males in interaction 3 provided considerable verbal feedback, both to one another and to the females (their wives) who were present, while in interaction 4 the males were obviously cast in the role of information providers, and the females, of interested listeners. None the less, the results outlined above clearly suggest that ingressive usage is gender-stereotyped. Peters’ (1981) investigation of ingressive yeah in the community of Vinalhaven, Maine, provides further support for this interpretation. According to Peters, though in Vinalhaven children of both sexes and women of all ages employ ingressive particles, the use of these forms is substantially reduced in teenage and adult males who opt for traditional employment in the fishery; however, males employed in non-marine pursuits maintain their use of ingressives. Thus avoidance of ingressive particles constitutes one aspect of the construction of working-class male identity in this fishing community. 8 5.
Discourse parallels across the North Atlantic/Baltic Zone Cross-linguistic gender similarities in usage add supporting evidence to the hypothesis that pulmonic ingressive usage across the North Atlantic/Baltic Zone represents an areal phenomenon. The language contact argument is reinforced by the close parallels in discourse or pragmatic meaning that are likewise to be observed within the Zone. The examination of cross-linguistic discourse similarities naturally raises 8
Our interpretation of ingressive usage in terms of a shared social construction of gender across the North Atlantic/Baltic Zone differs from the physiological explanation offered by Hill and Zepeda (1999), who propose that greater use of pulmonic ingressive articulation on the part of women is related to size differences in the vocal tract, i.e., the smaller female pharynx and larynx. However, the fact that these forms are used readily by males in a number of communities suggests that the observed gender difference across the Zone is more social than biological in nature.
CLARKE & MELCHERS
66
the more fundamental issue of whether ingressive particles carry different pragmatic or discourse meanings from their egressive counterparts. According to Pitschmann (1987: 155), German “ja produced by an ingressive air-stream exhibits distinct semantic values which differ from those of the ja produced on the more common egressive air-flow”. While Pitschmann’s (1987) overall assessment also applies to the languages we have surveyed, we do not share his views on the underlying pragmatic function of ingressive articulation, namely, to convey confirmation in a ‘routine’ and ‘fairly passive’ way. 9 Analysis of our Swedish and NE taped corpora as well as participantobservation data — coupled with the remarks of such investigators as Stølen (1994, 1995) and Kobayashi (2001) — points to the conclusion that there is clear semantic and pragmatic overlap between ingressive and egressive variants of discourse particles. However, we suggest that ingressive variants invariably carry special pragmatic meaning which is not a necessary component of their egressive counterparts. This is perhaps best seen by way of the following hypothetical example, which constitutes a request for (new) information: (11) Did you do it? (Egressive) (Ingressive)
Yes/Yeah ?
Native speakers tend to agree that if the ingressive response is judged acceptable, it is only in the sense of confirmation of what the respondent perceives to be the assumptions or expectations of the questioner; in (11), the ingressive implies ‘Yes, I agree with you/ Yes, as you already know/Yes, I admit it...’ That is, an ingressive particle is not generally appropriate in situations in which the particle represents new information. Rather, ingressive variants imply some sharing of contextual or situational knowledge — and opinion — on the part of interactants. Two naturally-occurring participantobservation examples follow, the first from NE and the second from Swedish; the participants in (13) were of the opinion that had the first statement not been articulated, and had the interaction consisted simply of a request for
9
This does not mean that such pragmatic meanings do not surface in certain contexts of usage. For example, repetition at intervals of ingressive ja or yeah may offer a bored listener an easy mechanism for feigning interest in a long-winded speaker to whom he or she is not paying much attention. Even a response consisting of a single ingressive may signal that the speaker’s mind is elsewhere, and that he or she does not wish to be disturbed (cf. the ‘counting under one’s breath’ phenomenon mentioned earlier).
INGRESSIVE PARTICLES ACROSS BORDERS
67
information (‘Have you tried the chair?’), the ingressive would have been unlikely. (12) F: Oh, that’s a different one. You have two, do you? M: (i.e., ‘as you have already concluded...’) (13) F: Den här stolen är väldigt bekväm. “This chair is very comfortable.” M: Har du suttit i den? “Have you tried it?” F: <ja> “” In short, ingressives are characterized by a feature which we will term [+affiliative], that is, a presupposition or presumption of shared or affiliative orientation on the part of discourse interactants. These are particles the most usual pragmatic meaning of which is to establish and maintain interactional solidarity and harmony, particularly in informal conversational settings. While this pragmatic meaning may also be associated with the egressive counterparts of these particles, it is not central to the meaning of the latter. 10 Our investigation of the pragmatic functions carried by ingressive discourse particles across the North Atlantic/Baltic Zone suggests a high degree of commonality of discourse meaning. In their role of maintaining good conversational interaction, these particles are most commonly associated across the Zone with listeners rather than with speakers. Their most frequent use is that of minimal response or backchannel, where they signal not only passive, but in particular, active or supportive listenership — or as Stølen (1994) puts it, interactional alignment. Yet within the languages of the Zone, they are also used by speakers, with the basic meaning of confirmation or agreement, and often with the discourse function of maintaining conversational rapport and harmony. Yet like minimal responses in general, ingressive particles across the Zone carry a second and simultaneous discourse dimension, that of turn regulator. Indeed, in some contexts, this may be their dominant function. As minimal responses on the part of listeners — whether articulated on their own, or in strings of two, three or more, involving both egressive and ingressive 10
Perhaps the most striking confirmation of the ‘shared orientation’ pragmatic meaning conveyed by ingressive discourse particles emerges from Eklund’s (2002) investigation of Swedish ingressives. In this experimental study, when participants were placed in a telephone situation in which they believed they were interacting with a machine voice rather than a human voice, they readily produced egressive discourse particles, but not a single example of an ingressive particle.
68
CLARKE & MELCHERS
articulation — they typically signal a lack of intent to initiate a turn as speaker. They function as what Schegloff (1982:81) has termed ‘continuers’, or verbal indications that the ‘interlocutor orients to a speaker signal that there is a possible turn change point, and uses the backchannel to pass over the opportunity to take the turn’. As turn regulators, they may also be used by speakers as opposed to listeners: here, they carry a full range of functions, as they serve as signals of turn initiation, of intent to continue, or (perhaps most frequently) of end of turn. We illustrate these diverse functions with several examples. (14) below — part of a verbal exchange among the four female participants from our NE taped corpus represented in Table 3 — provides a good illustration of listener backchannelling. Both statements are followed by a chorus of minimal responses from all participants, involving both egressive and ingressive articulation. In addition, the chorus is joined by the two speakers (F2 and F1), both of whose use of ingressive not only indicates verbal support, but concurrently functions as a signal that the speaker does not intend to resume the floor immediately. (In examples (14) and following, ingressives are italicized within < >, as per our usual convention; overlapping speech is indicated by [ ) (14) F2: well they’d have to have their room like, you know, and then we’d have to double up sometimes, [like, you know F3: [yeah F1: [ F4: [yeah F2: [ F1: [yeah, there’d be two or three of us in bed together F2: yeah F4: yeah F1: F2: F3: mm This interaction is also characterized by sequences of backchannels, with ingressives appearing at or near the end of the sequence. An illustration of this occurs in the following example of overlapping speech involving three of the same participants, where the series of minimal responses from F2 and F1 is interwoven with the statement of F3:
INGRESSIVE PARTICLES ACROSS BORDERS
(15) F3:
F2: F1:
69
Was the same thing [with the youngsters you’d take yer youngsters to ‘em or – if there was something wrong or [oh yes, yeah, oh yes, , yeah [oh yeah, yeah, , yeah
In examples (16) and (17), from a mixed sex interaction in the NE taped corpus, speakers use ingressives to reconfirm the truth value of a previous statement. The pattern in (17) — in which the speaker follows an initial egressively-articulated confirmation (often, as here, representing new information) with an ingressive particle when further probed by a conversational partner — is a commonly-occurring one throughout our taped corpora and observational data. (16) F: They says alcohol wasn’t a factor. M: It wasn’t? F: <no>, no (17) M: Chris gone back to ‘im? F: yeah M: Is she? F: Example (18), from the Göteborg corpus, illustrates the same phenomenon in Swedish: (18) B: vilken buse “what a scoundrel” A: ja “yeah” B: <ja> “” Example (19), from the same corpus, provides an illustration of confirmatory ingressive no in turn-final position, a pattern very common across the ingressive Zone: (19) A: nä hon röker inte heller ... gymnastiklärare “no, she doesn’t smoke either ... PE teacher” B: nä de gör hon inte “no she doesn’t, <no>” While in (19) the turn-final ingressive ultimately reconfirms the words of a conversational partner, such is not necessarily the case. Ingressive particles are often found with a pragmatic meaning termed ‘self-directed’ by Stølen
70
CLARKE & MELCHERS
(1994) — that is, confirmation is with the speaker’s own statement rather than that of an interlocutor. Perhaps more importantly in such instances, the ingressive often functions as an end of turn marker, as in the British English examples in (20) (the Cumbrian SED speaker also illustrated in [9]–[10] above), as well as the (overheard) Swedish example in (21): 11 (20) a. No, I've just heard them called Herdwicks, aye, b. Oh aye, aye ...aye, that's right (21) Du menar att du gillar henne inte i den där rollen, “You mean you don't like her in that part, <no>” 6.
Conclusion In our cross-linguistic sketch of pulmonic ingressive discourse particle use, we have delimited a northern European and transatlantic zone through which these particles have diffused via language contact over the course of centuries. The striking parallels evident within this broad geographical region provide evidence of cross-linguistic transmission of social and pragmatic factors, a rarely-noted phenomenon in the language contact literature. Since little has likewise been written on pulmonic ingressive articulation in general, we hope that our findings will stimulate further research in this area. References Allen, H. B. 1973, 1975. The Linguistic Atlas of the Upper Midwest, Vols 1 & 2. Minneapolis: University of Minnesota Press. Allwood, J., M. Björnberg, L. Grönqvist, E. Ahlsén and C. Ottesjö. 2000. “The spoken language corpus at the Department of Linguistics, Göteborg University”. Forum: Qualitative Socialforschung/Social Research 1.3. [For online information on the Göteborg University Spoken Language Corpus of Swedish, see www.ling.gu.se/SLSA/SLcorpus.html] Clarke, S. 2004. “The legacy of British and Irish English in Newfoundland”. Legacies of Colonial English: Transported Dialects ed. by R. Hickey, 242–261. Cambridge: Cambridge University Press.
11 Our participant-observation data suggest that in Sweden (as apparently throughout Scandinavia), ingressive particle articulation may signify little more than turn closure in certain specialized types of interaction. This is most obvious in service interactions involving a sequence of transactions (e.g. a shop assistant providing a number of individual items for a customer) where the phrase ja tack (lit. yes thank-you), articulated ingressively by the serviceprovider, in effect has the pragmatic meaning of ‘What’s next?’. This usage is not parallelled in Newfoundland English.
INGRESSIVE PARTICLES ACROSS BORDERS
71
Conklin, H. C. 1959. “Linguistic play in its cultural context”. Language 35. 631–636. Crystal, D. 1987. The Cambridge Encyclopedia of Language. Cambridge: Cambridge University Press. Dieth, E. 1950. Vademekum der Phonetik. Bern: A. Francke. Eklund, R. 2002. “Ingressive speech as an indication that humans are talking to humans (and not to machines)”. Paper presented at the Seventh International Conference on Spoken Language Processing (ICSLP 7), Denver, Colorado. Gunnell, T. 1995. The Origins of Drama in Scandinavia. Woodbridge UK/Rochester NY: D.S. Brewer. ----------. 2001. “Grýla, grýlur, ‘grøleks’ and skeklers. Medieval disguise traditions in the North Atlantic?” Arv 57 (Nordic Yearbook of Folklore 2001): 33–54. Hakulinen, A. 1993. “Inandningen som kulturellt interakionsfenomen”. Språk och social context ed. by A.M. Ivars et al., 49–67. Helsinki: Institutionen för Nordiska Språk och Nordisk Litteratur, University of Helsinki. Hill, J. H. and O. Zepeda. 1999. “Language, gender, and biology: pulmonic ingressive airstream in women’s speech in Tohono O’odham”. Southwest Journal of Linguistics 18:1.15–40. Hockett, C. 1955. A Manual of Phonology. (= Memoir 11, Indiana University Publications in Anthropology and Linguistics). Reprinted in 1974, 1988 by the University of Chicago Press. Johnson, L. G. 1962. “Laurence Williamson”. Scottish Studies VI.49–59. Kirwin, W. J. 1971. “Ingressive speech reported in Newfoundland ‘mummertalk’”. Regional Language Studies (Newfoundland) 3.24. Kobayashi, N. 2001. Ingressivt ja: Ja på innpust – ikke tegn på overraskelse eller dårlig hjerte. Bergen: M.A. thesis, University of Bergen. Kurath, H., M. L. Hanley, B. Bloch, G. S. Lowman Jr. and M. Hansen. 1943. Linguistic Atlas of New England, vol. 3, part I. Providence, RI: Brown University. Léon, P. 1992. Phonétisme et prononciations du français. Paris: Nathan. Paddock, H. 1981. A Dialect Survey of Carbonear, Newfoundland. (= Publication of the American Dialect Society 68). University of Alabama Press. Peters, F. J. 1981. The Paralinguistic Sympathetic Ingressive Affirmative in English and the Scandinavian Languages. New York, NY: New York University Ph.D. dissertation. Pitschmann, L. 1987. “The linguistic use of the ingressive airstream in German and the Scandinavian languages”. General Linguistics 27.153–161.
72
CLARKE & MELCHERS
Pratt, T. K. 1988. Dictionary of Prince Edward Island English. Toronto: University of Toronto Press. Rietveld, A. C. M and V. J. van Heuven. 1997. Algemene fonetiek. Bussum: Coutinho. Schegloff, E. A. 1982. “Discourse as an interactional achievement: some uses of “uh-huh” and other things that come between sentences”. Analyzing Discourse: Text and Talk ed. by D. Tannen, 71–93. Washington, D.C.: Georgetown University Press. Shorrocks, G. 2003. “Pulmonic ingressive speech in Newfoundland English: a case of Irish-English influence?” The Celtic Englishes III ed. by H.L.C. Tristram, 374–389. Heidelberg: Universitätsverlag C. Winter. Steinbergs, R. 1993. The use of the paralinguistic sympathetic ingressive affirmative in speakers of English in the St. John’s, Newfoundland area. St. John’s, NF: Memorial University (Linguistics), ms. Stølen, M. 1994. “Gender-related use of the ingressive ja in informal conversation among speakers of Danish”. Cultural Performances: Proceedings of the Third Berkeley Women and Language Conference ed. by M. Bucholtz, A.C. Liang, L.A. Sutton and C. Hines, 668–677. University of California at Berkeley: Berkeley Women and Language Group. ----------. 1995. “Multi-functional ja: an analysis of the Danish affirmative ingressive ja as conversational structuring device in informal conversations”. Nordlyd: Tromsø University Working Papers on Language and Linguistics 23.217–229. Woods, H. B. 1999. The Ottawa Survey of Canadian English. Kingston, Ontario: Strathy Language Unit, Queen’s University.
ON THE DEVELOPMENT OF THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN (PLAUTDIETSCH) LARISSA NAIDITCH Hebrew University of Jerusalem 1.
Introduction This article is part of my research on an insular dialect (Sprachinsel) of German, in which I consider both synchronic and diachronic developments. For this research I have been able to consult both primary and secondary sources. The former include tape recordings and notes I was able to make in Kazakhstan (Martuk, Aktyubinsk district) and in Germany among immigrant families from the former Soviet Union (especially from the Orenburg district). In the summer and autumn of 2001 I was able to work for three months in Marburg (Institut für Deutsche Sprache – Deutscher Sprachatlas) studying sources relevant to my project. 1
2.
The historical background The Mennonites are an independent religious community, the oldest evangelical Free Church, which arose out of the Baptist movement at the time of the Reformation. They took their name from their founder, the former Catholic priest from Friesland Menno Simons (1496–1561). Their way of life is characterised above all by the renunciation of violence and absolute truthfulness; the Mennonites do not baptise their children and refuse all oaths. They have frequently been persecuted on account of this way of life and on several occasions have been forced to migrate from one country to another. A considerable number of Mennonites fled to the Weichselland (the Vistula area), which at that time belonged to Poland. This area became their temporary home. According to Walther Mitzka (1930: 7) the Mennonites have ‘ihren stärksten Prozentsatz innerhalb der Bevölkerung […] im Großen und Kleinen Werder, also im ehemaligen Kreise Marienburg und in der westlich der Kreisstadt
1 This study trip was supported by the German Academic Exchange Service, DAAD; I would like to thank the DAAD and the staff in Marburg for their support.
74
LARISSA NAIDITCH
gelegenen Hälfte des alten Kreises Elbing’. 2 After the first partition of Poland (1772), the territory of the Weichselland (but not the Free City of Danzig) became part of Prussia. Soon after this the Mennonites came into conflict with the Prussian government, principally because of their refusal to do military service. On the other hand, the Russian Empress Catherine the Great promised the foreign colonists numerous privileges. The Russian government made suggestions which were particularly advantageous to the Mennonites, including exemption from military service, freedom of worship and large tracts of land (Schirmunski 1928: 22). The migration of the Mennonites to Russia began in 1789 and continued until the 1870s. The first group of Mennonite colonies (Chortitza, also known as Alt-Kolonie) was created in the south of the Russian Empire, not far from Zaporožje on the Dnieper. In 1804 a second group of colonies was founded, this time on the Molochnaya River in the Melitopol’ district. In the 1850s new colonists from Prussia settled on the Volga (the colonies of Am Trakt and Alexandertal). Further Mennonite settlements came into being either as mother colonies as a result of the arrival of new colonist from Prussia or as daughter colonies. Thus there were also Mennonite settlements in the Mariupol’ district, in the Don region, in the northern Caucasus, in the administrative district of Orenburg and in Siberia. The two last-named groups of colonies, especially the settlements in the Kalunda steppe of the Altai region, founded in 1907, have survived up to the present (Klaube 1991). Nowadays the dialects, culture and way of life of these settlements form a valuable material which is utilised by a number of researchers. In the 1880s many Mennonites emigrated to Canada and the USA; the reasons for this emigration were political reforms in Russia, in particular the introduction of general military service. Some of these emigrants moved on from Canada and the USA to Mexico, Honduras, Paraguay, Uruguay, Brazil or Argentina (see Brandt 1992). Emigration to Canada and Paraguay was still possible in the 1920s, and a number of Mennonite colonies were founded at that time (Rotkohl 1993: 17–18). In the 1920s and 30s and during the Second World War Russian Germans were subject to harsh reprisals; finally they were deported from the European parts of the Soviet Union and sent to Siberia and Kazakhstan. In recent times many of them have emigrated to Germany. The dialect of the Russian German Mennonites, usually referred to as ‘Plautdietsch’ or ‘Mennonitenplatt’ (PD, in English ‘Mennonite [Low] German’), was formed in the Weichselland and is in all probability a form of 2
The Mennonites ‘form the majority of the population […] in Gross Werder and in Klein Werder, in other words in the former district of Marienburg and in the part of the old district of Elbing which lies to the west of the district town.’
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
75
the Low German dialect which was spoken in this area. The emigrants who moved there thus gave up their mother tongue in exchange for the Low Prussian dialect of the Weichselland, which according to Mitzka (1930: 9) was ‘gewiss nicht ein schroffer Übergang’. 3 The linguistic basis of Mennonite Low German are the Low-Prussian German dialects of the Weichselland (these were thoroughly researched by Walther Mitzka between the two World Wars; cf. also Pinnow 1998, 2000). Although the places in the Weichselland from which emigration took place are known, a precise determination of the linguistic home of the Mennonites is a complicated problem (Mitzka 1930; Thiessen 1963; Moelleken 1987). Today PD can be heard in far-flung countries such as Siberia and Kazakhstan, the USA, Mexico and Germany. It is used as a home language and is experienced by the religious and cultural community of the Mennonites as an important factor in their identity. Although it is primarily a spoken language, there is also a literature in PD, especially poems, most of which originate from North America, but which are sometimes also know to Mennonites in other countries (Reimer, Reimer & Thiessen 1983). Thus from a sociolinguistic point of view PD is more than a dialect; it can be seen as a daughter language of German and an Ausbausprache (Kanakin & Wall 1994: 1). PD is represented in many variants which differ according to the phonetic and phonological, lexical and — more rarely — grammatical levels. These can often be observed in the same family, for instance where man and wife come from different areas. This variation can be traced to the following factors: 1) the original differences between the dialects of the first two colonies (Chortitza and Molochnaya); 2) different results of autochthonous developments in individual dialects; 3) the results of dialect mixing; 4) the influences of the languages and dialects of the different surroundings. In spite of these differentiations, PD is undoubtedly a single dialect, which can be seen for instance in common trends in its development. 3.
Phonetic and phonological studies on PD: The state of research The first phonetic description of PD was undertaken by Jacob Quiring in his doctoral dissertation (1928), a traditional Neogrammarian sound history treatise on the dialect of Chortitza. In the period between the two World Wars, Walther Mitzka, who was researching the German dialects in the region around Danzig, also studied the Low German of the Russian-German Mennonites. He had the opportunity of making field notes, as a group of Russian-Germans had
3
‘Certainly not a sharp transition.’
LARISSA NAIDITCH
76
to wait months in Germany for their crossing to America. His main purpose was a localisation of this dialect in terms of linguistic geography and typology within the framework of a general study of a linguistic area. An overview of PD is also given by Viktor Schirmunski (1928: 51–53). In the post-war period, the subject of Russian-Germans in the Soviet Union was practically prohibited. It was only in the 1960s, from the beginning of the Thaw, that Russian-German dialects were occasionally discussed in dissertations and publications of Germanists from Siberia and Kazakhstan (Berend & Jedig 1991, bibliography: 269–374). Thus Avdeev (1968) and Jedig (1966) studied the phonetic/phonological system of PD in the Altai region. The latter’s research was published in German. Foremost among the American scholars who contributed to our knowledge of PD phonology were Goerzen (1952) and Moelleken (1972). A number of studies of the dialects and the linguistic integration of Russian-Germans appeared in Germany in the 1990s as a reaction to the new wave of immigration of ethnic Germans from the countries of the former Soviet Union; individual works are devoted to the Mennonites (see e.g. Heidemann & Sawatzky 1996; cf. also works by the Groningen linguists de Graaf & Nieuweboer 2001). The phonology of PD is treated in a number of works published in the last decade (Brandt 1992; Kanakin & Wall 1994; Nieuweboer 1999; Naiditch 2001). In spite of this, the diachronic development of the sound system has not been researched satisfactorily. 4.
The consonant system of PD: Synchronic and diachronic problems According to Niebaum (1985: 1226), the consonant system of Middle Low German had the following structure: /p/ /t/ /k/ /b/ /d/ /[g] /l/ /v/ /[z] [ƒ]/ /j/ /f/ [s]/ /x/ /h/ /m/ /n/ /N/
/r/
In Mennonite Low German we find the following consonants: /p/ /b/ /v/ /f/ /m/
/t/ /t’/ /k/ /d/ /d’/ /[g] /z/ /ž/ /j/ [ƒ]/ /s/ /S/ /C/ /x/ /h/ /n/ /n’/ /N/ /l/ /r/
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
77
The affricates which occur in loan words in PD can be interpreted as biphonemic clusters, and are thus not contained in the table. There are a number of partially overlapping factors which are relevant for the development of the Middle Low German consonant system (which can be seen as the reference system) in PD. The most important of these are changes which affect the system, e.g. by adding new phonemes. These phonemes are 1) the set of palatalised consonants; 2) the phoneme /S/, later also /ž/; 3) the phoneme /ç/. Other features apply to the rules of phoneme use and to the phonological structure of individual lexemes. These are: 1) Auslautverhärtung (final devoicing of obstruents), which was already completed in the Middle Low German period, and the restoration of the consonant oppositions in question in final position; 2) lenition of the plosives intervocalically, which already took place in earlier periods of the language history; 3) the further development of the velar consonants and of /j/; 4) changes in consonant groups with /n/; 5) vocalisation of /r/. In this general list only the first point — the development of /t’, d’, n’/, which is restricted to a small number of the dialects — is not usual. Some of these phenomena are discussed in greater detail below. The phoneme /S/, which is also new for High German dialects, is foreign to many Low German dialects (Žirmunskij 1956: 330–335); in others, including those which can be regarded as linguistic sources for PD (the dialects of the Danzig area), /S/ is present. As is well known, the phoneme /S/ developed as a result of the monophonemisation of the consonant group sk (sch) and as a phonologisation of an allophone of /s/ before consonants in initial position (PD: /S´u/ ‘shoe’, /Slu:p´/ ‘to sleep’), sometimes also in other positions. In PD, as in many dialects, /S/ appears after r in medial position; cf. the examples from PD: /da:rS t / ‘thirst’, where r is retained, and /boÅ t ’/ ‘birch’, where r has been vocalised (Kanakin & Wall 1994: 15). A new phoneme which came into being partly in analogy to the consonant /S/ and has only a weak functional load, is /ž/: /hi´ž/ ‘millet’. Auslautverhärtung, i.e. the neutralisation of the opposition between voiced and voiceless consonants (in other dialects of that between fortis and lenis consonants) in final position, took place in most German dialects. This development was ‘completed by the time of the earliest Middle Low German records’ (Niebaum 1985: 1226). It can also be observed in PD, cf. the sound alternations /ti:t/ – /ti:d´/ ‘time – times’, /bli:v´/ – /bl´if/ ‘to stay – stayed’, /le:z´/ – /lçus/ ‘to read – read (past tense)’. Nowadays this neutralisation rule no longer applies. A new final position, which came about because of apocope, also allows for voiced consonants (cf. Nieuweboer 1999: 82). The phonological character of the sound alternations between voiced and voiceless consonants has also changed: they are no longer automatically determined by their
78
LARISSA NAIDITCH
position. Following the sound changes mentioned, the opposition between voiced and voiceless consonants has gained a new morphonological function: voicing serves as a marker of the plural in one of the morphological noun types: /br´if/ ‘letter’ – /br´iv/ ‘letters’, /frint/ – /frind/ ‘friend’ – ‘friends’. The development of the velar consonants It is well known that the velar consonants have a particular fate in German dialects which is different from that of the other consonants (Frings 1967). The fricative pronunciation of g [ƒ] is characteristic of many High and Low German dialects. This is particularly typical in postvocalic position, in which reflexes of g can appear as [ƒ, C, j, x] (Žirmunskij 1956: 284–293; Lasch 1974: 178–186); there are numerous contractions, which were characteristic of late Old High German, Middle High German and Middle Low German and which can be observed in individual words in the standard language as well as in some dialects (Žirmunskij 1956: 291–293). In PD the plosive [g] occurs in initial position when it is followed by a back vowel in the word stem: /glçus/ ‘glass’, /gçust/ ‘guest’. If the velar consonant is followed by a front vowel (directly or after a consonant) g becomes [j]: /jin’t’/ ‘went’, /je:v´/ ‘to give’, /jle:zÅ/ ‘glasses’, /jri:p´/ ‘to grasp’. After a vowel g is realised as a fricative in many varieties of PD. After a back vowel we have the following picture: [ƒ] intervocalically and in the new final position (as a result of apocope), but voiceless [x] in the old final position — [u:ƒ] ‘eye’, [u:ƒ´] ‘eyes’, [foŃ´l] ‘bird’, [doŃ] ‘days’, [fru´x] ‘asked’. In another variant of PD g occurs as a plosive after back vowels (including intervocalically): /joÅg´/ ‘to hunt’, etc. As a result of the loose connection and the glide before g, which leads to diphthongisation, [ƒ] is only possible after long vowels and diphthongs. After front vowels and sonorant consonants /j/ occurs in this position: /bi:j´/ ‘to bend’, /fçlj´/ ‘to follow’. The fate of gg should also be examined in this context. According to Niebaum (1985: 1225), the original simplex and geminate, generally, merged in the Middle Low German period: the double letter in the spelling indicates a short preceding vowel. But the development of g – gg is more complicated, as the PD material shows. Niebaum writes about the consonant system of Middle Low German: ‘Bei g, gg hat sich etwas kompliziertere Entwicklung ergeben. Als Fortsetzer des alten /g:/ und in Verbindung mit /N/ dürfte der Konsonant, das legen die graphischen Befunde nahe, als Verschlußlaut [g] anzusehen sein, 4.1
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
79
ansonsten als Spirant [ƒ].’ 4 The transition from gg to [d’] in PD, which in all probability took place in a more recent period in the development of the dialect, shows a differentiation between the reflexes of gg and g. At the time of the relevant sound changes, after front vowels the only /g/ remaining was that which had developed from gg, because the simplex had become a fricative. This velar plosive was palatalised to /d’/ (see below). The phonological results of these developments can be summarised as follows: 1) the plosive /g/ in initial position became identified with the reflex of the corresponding geminate in medial and final position following a back vowel, i.e. /g/ in /glçus/ ‘glass’ as in /pog/ ‘frog’ (<pogge). The plosive is identified with the same phoneme in those dialects which did not have spirantisation: /kloÅg´/ ‘to complain’, etc. The voiced velar fricative [ƒ] (in the dialects with spirantisation) is correctly viewed as an allophone of /g/, from which it developed. It occurs only after long vowels and diphthongs — intervocalically or in the new final position: ([foŃ´l] ‘bird’, [doŃ] ‘days’) — and is never in opposition to the plosive /g/ in any position. 2) /j/ ( /a˘/: cf. the alternations /doxtÅ/ – /da:CtÅ/ ‘daughter’ – ‘daughters’ and minimal pairs: /va:C/ ‘way’ – /va:x/ ‘guard’, /a:Ct/ ‘genuine’ – /a:xt/ ‘eight’ (Kanakin & Wall 1994: 15). An additional factor is the development of the new diphthong /oå/ < er: /boÅC/ ‘mountain’ (cf. a rare example of /x/ after this diphthong: /j´froÅxt/ ‘asked’). Finally we should mention some sound alternations in which the phonemes under consideration take part. They have a morphonological function (Kanakin & Wall 1994: 21–28) and at the same time show the connections between the phonemes /g/, /C/ and /j/ in the phonological system.
‘The development of g, gg was somewhat more complicated. As a continuation of the old /g:/ and in conjunction with /N/, the spelling in the records indicates that the consonant is to be viewed as the plosive [g], otherwise as the fricative [ƒ].’
4
80
LARISSA NAIDITCH
/g/ – /j/: /go:n´/ – /jin’t’/ ‘to go – went’, /glçus/ – /jle:zÅ/ ‘glass – glasses’, /g/([ƒ]) – /x/ : [droŃ´ – dru´x] ‘to carry – carried’, [u:ƒ] – /u:xt’´/ ‘eye – little eye’. /j/ – /C/: /boÅC/ – /boaj/ ‘mountain – mountains’, /va:C/ – /veÅj/ ‘way – ways’, /j/ – /x/: /fli´j´/ – /flu´x/ ‘to fly – flew’, /zy:j´/ – /zu´x/ ‘to suck – sucked’, /x/ – /C/: /doxtÅ/ – /da:CtÅ/ ‘daughter – daughters’. The development of the palatalised consonants The most striking peculiarity of PD, which deserves particular attention, is the series of palatalised consonants: /t’, d’, n’/. Diachronically they have the following sources /t’/ < /k/; formerly this was an allophone of /k/ in the positions: 1) before front vowels; 2) before sonorants followed by front vowels; 3) after front vowels and front vowels + sonorant. The phonemic split was triggered by the following factors: 1) e> /a:/ in a closed syllable, e.g. /ra:t’/ ‘skirts’, /dra:t’/ ‘dirt’, /da:t’/ ‘ceiling’, /za:t’/ ‘sacks’, /a:t’/ ‘corner’, /ha:t’/ ‘hedge’. This e was often the result of an umlaut from a or o (> ö, e), in individual cases it was derived from i: /t’a:p/ ‘heads’, /t’a:rtÅ/ ‘shorter’. 2) ir, er >/oå/: /t’oÅt’/ ‘church’, /boÅt’/ ‘birch’. Cf.: /koÅk/ ‘(I) cook’ – /t’oÅt’/ ‘church’, /jeStoÅk´/ ‘stabbed’ – /StoÅt’/ ‘to strengthen’. 3). An additional factor in the phonologisation process is apocope, which causes the ocurrence of minimal pairs: /za:k/ – /za:t’/ ‘sack’ – ‘sacks’. Before the vowel /y:/, which is the result of the so-called spontaneous palatalisation of /u:/, the consonant was not palatalised: /kry:pe/ ‘to crawl’, /kry:t/ ‘cabbage’. (Before /k/, /y:/ does not occur because of the shortening of long /u:/ to /u/: /bruk´/ ‘to need’, /buk/ ‘stomach’, /luk/ ‘hatch’). These data demonstrate that the phonetic/phonological process of consonant palatalisation was completed before the change from /u:/ to /y:/. However, we find consonant palatalisation in some more recent loan words with k after and before e, i; the palatalisatioin process is thus not purely phonetic, but phonetic/phonological. Cf. loans from Russian: /lçuft’ç/ Russ. lawka [lafka]> lafke ‘shop’, /č´snit’/ ‘garlic’ aus Russ. česnok, Wrentjes from Russ.wareniki (a kind of food) , Plemjanit’ Russ. plemjannik ‘nephew’, Respublitje ‘republic’ from Russ. respublika; Kelchosnitj from Russ. kolchosnik ‘kolkhoz farmer’ (Klassen 1993: 63, 50), borrowings from other sources, mainly from Standard German: /tECnit’/ ‘technology’, /muzit’/ ‘music’, /cirt’el/ ‘ compass’. More recent loan words show a loosening of the positional dependence on /k/: cf. a variant plemjannik (Nieuweboer 1999: 194), and also occasional borrowings such as /žulik/ ‘swindler’. 4.2
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
81
Nowadays the phoneme /t’/ is thus found in the following positions: initially before vowels and sonorant consonants – /t’a:lÅ/ ‘cellar’, /t’i:n’Å/ ‘children’, /t’ni:p´/ ‘to pinch’, /t’ri:C/ ‘war’, /t’li:n/ ‘little’; medially following vowels and sonorants – /drint’´/ ‘to drink’, /t’le:t’Å/ ‘wiser’; finally following vowels and sonorants – /rint’/ ‘ring’, /malt’/ ‘milk’, /voÅt’/ ‘work’. The connection between /t’/ and the velar consonants /k/, /x/ is shown by phonemic alternations: /t’´ip´ – koft/ ‘to buy – bought’, /zi´t’´ – zoxt/ ‘to seek - sought’, /za:k/ - /za:t’/ ‘sack – sacks’, /k´u/ - /t’i:/ ‘cow – cows’, /klu´k/ – /t’le:t’Å/ ‘wise – wiser’ (vgl. Kanakin & Wall 1994: 22–26, 28). This corresponds to variations in PD spelling: /t’/ is given as tj in some sources, in others as kj, or in initial position as kj, otherwise as tj (Nieuweboer 1999: 218–225). The development in the palatalised consonants is seen in the emergence of the new phonemes /n’/ und /d’/. The phoneme /d’/ carries a much smaller functional load than /t’/. Its historic source is the geminate gg after /i/, including after /i/ < /y/: /lid’´/ ‘to lie’, /rid’´/ ‘back’, /vrid’l´ / ‘to totter’, /brid’/ ‘bridge’. Thus the changes associated with g and gg can be seen as follows: e:, i: + g + Vowel> ej, ij; e, i + gg + Vowel > eg, ig (+ Vowel) > ed’, id’(+ Vowel). The independence of the phoneme /d’/ in the phoneme system is shown by the following examples: /a:d’/ ‘the edge of material’ - /fla:g/ ‘flag’, /ti:d´/ ‘times’ - /mid’´/ ‘mosquitoes’ (cf. Nieuweboer 1999: 87; Moelleken 1972: 35; Avdeev 1968: 4– 5). The historical source of /n’/ are the phoneme clusters nd, ng after front vowels (in final or intervocalic positions): /vEn’/ ‘walls’, /En’/ ‘end’, /vo:nin’/ ‘apartment’, /rin’/ ‘ring’, /zin’´/ ‘to sing’, /t’in’Å/ ‘children’. Thus we have here a syntagmatic monophonematisation of the two consonant clusters. The fronting of the articulation of /N/ corresponds to that of the reflexes of gg. /n’/ alternates with /N/: /bin’´ – j´buN´/ ‘to bind – bound’. In individual cases nd also became /n’/ after u, causing among other things the opposition /n’/ – /N/: /hun’/ ‘dogs’ - /juN/ ‘young’. The emergence of /n’/, which took place in parallel to the other palatalisations and can thus be explained as a case of analogy, was also promoted by the weakening, assimilation and gutturalisation of d after nasals, which is familiar from the history of German, including the Low German dialects (Niebaum 1985: 1226; Foerste 1957: 1779 f.). The new set of consonants fits in with the old proportional oppositions in the system: /t/:/d/:/n/ = /p/:/b/:/m/ = /k/:/g/:/N/ = /t’/:/d’/:/n’/. Its development is a clear demonstration of the tendency towards symmetry in the system. In spite of this, the old phoneme oppositions retain some functional differences of the new oppositions. Thus the phoneme pair /t’/:/d’/ does not take part in the alternation of the voiceless and voiced consonants (see above).
82
LARISSA NAIDITCH
4.2.1 Hypotheses on the reasons for the development of palatalised consonants in German dialects. There are some spellings which indicate a strongly palatalised pronunciation of k before e in Old Saxon: an i is introduced after k. According to A. Lasch (1974: 178): ‘[Es] ist anzunenehmen, dass die Anlage zu dieser Entwicklung über ein weites Gebiet verbreitet und in verschiedenem Maße vorgeschritten war.’ 5 However, palatalised consonant phonemes are found in only a small number of German dialects: particularly in East Pomeranian and some West Prussian dialects (Teuchert 1913; Koerth 1913; Semrau 1915); also in the Zips dialect islands (Gréb 1921; Valiska 1967). As it is not a typical feature of German dialects, scholars have attempted to explain it by means of language contact, assuming either Frisian or Slavonic influence. As has already been mentioned, the first generation of Mennonites who settled in the Weichselland came from Friesland. The influence of Frisian, which is otherwise apparent only in individual words, can thus be viewed as a kind of substrate. The Frisian palatalisation or assibilation is a striking feature of this language. Even in Old Frisian there are cases where t is written instead of k (a number of spellings are found: t, ty, thi, thy, tti, tty, tthi, tthy, tt); the phonetic realisation of this k is said to have varied in the different dialects (Loopstra 1935: 10). Palatalised /t’/, /d’/ and their development mechanisms in PD can be compared to the phonological developments in the languages of the AngloFrisian group (see e.g. Penzl 1947, who discusses the phonemic split of /k/ to /k/ and /č/ in Old English). On the other hand, north-eastern German dialects were in contact with Polish and Kashubian dialects. It is well known that palatalisations are a typical feature of Slavonic languages; in this respect Kashubian dialects have similarities with the Low German dialects discussed here (Lorentz 1925: 74– 75, 82–83). Walther Mitzka stressed that the dialect speakers in the Slavonic – Low German contact zone were bilingual, and this could have assisted a common development: ‘Am kaschubischen Rande […] spricht der Kaschube seine und die deutsche Mundart, der Deutsche neben seiner das Kaschubische’ 6 (Mitzka 1928: 22). And:
5
‘Presumably the seeds of this development were spread over a wide area and had progressed to different stages.’ 6 ‘In the Kashubian border region […] the Kashubian speaks his and the German dialect, the German speaks Kashubian in addition to his own dialect.’
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
83
‘k wird in deutschen, polnischen und kaschubischen Mundarten Westpreußens zu verschiedenen Formen des tχ-Typs gewandelt [...]. Slawische Ortsnamen zeigen diese Sibilierung in der zweiten Hälfte des 13.Jahrhunderts. [...]. Also hier ist ein geographischer Vergleich der Geltung dieser palatalisierten k in den deutschen und in den slawischen Mundarten derselben Großlandschaft möglich’ (Mitzka 1959: 120f.). 7
The palatalised consonants are said to have been very widespread in the Pomeranian and West Prussian German dialects. The Mennonites may have borrowed this characteristic feature in their dialect. An important argument for the influence of the Slavonic languages is the presence of the same set of palatalised consonants in the dialects which were subject to Slavonic contacts. These include not only the Low German Pomeranian dialects, but also the East Central German dialects in eastern Slovakia, which developed in contact with Polish, Slovakian and Ukranian dialects (Gréb 1921; Valiska 1967). Cf. the following examples of the ‘Chmel’nicaer palatalisation’: [t’ent] ‘child’, [d’i:n] ‘to go’, [diRen’] ‘easy, small’ < geringe (Valiska 1967: 40, 43). Thus we can be dealing with either a contact zone or a development of typical characteristics of certain Germanic languages. These hypotheses are not mutually exclusive if we view them in the framework of the concept of a broader Sprachbund (language union). As we know, Roman Jakobson (1962: 164–190) saw the correlation of palatalisation as one of the most important features of the Eurasian Sprachbund (cf. also Thomason & Kaufman 1988: 247). It is too early for such a general conclusion from our material in the framework of this hypothesis; that requires further typological and contact linguistic research. Finally I would like to mention some examples from Russian dialects which show an obvious typological parallelism with the features of PD examined in this article. It is also significant that the results of the palatalisation in these dialects are phonetically strikingly similar to those in PD. In some modern Central Russian dialects /k’/, /g’/ are replaced by /t’/, /d’/ before front vowels: [ruka] – [rut’i], [rut’e] ‘hand’, [d’ib’el’] < [g’ib’el’] ‘death’ (Kasatkin 1989: 57). Cf. also corresponding forms in the dialects with progressive assimilative softening: [van’t’a] Wan’ka (name), [den’d’am] ‘money, dat. pl.’ (Kasatkin 1968: 4). This similarity can be explained in purely typological terms, but it can also point to common Slavonic-Northwest 7
‘k is changed to different forms of the tχ-type in the German, Polish and Kashubian dialects of West Prussia [...]. Slavonic place names show this sibilation in the second half of the 13th century. [...]. So here it is possible to make a geographical comparison of the appearance of this palatalised k in the German and Slavonic dialects of the same region.’
84
LARISSA NAIDITCH
Germanic developmental tendencies. The question as to which of these two solutions comes closer to reality must be left open here. It can only be resolved by further studies. References Avdeev, I.E. 1968. “O vzaimodejstvii meždu glasnymi i soglasnymi (na materiale nižnenemeckich govorov)“. Novosibirskij Gos. Pedagogičeskij Institut. Naučnye trudy, vypusk 25. Germanskie i romanskie jazyki. Novosibirsk. Berend, N. & H. Jedig. 1991. Deutsche Mundarten in der Sowjetunion. Geschichte der Forschung und Bibliographie. Marburg: N.G. Elwert Verlag. (Schriftenreihe der Komission für ostdeutsche Volkskunde in der deutschen Gesellschaft für Volkskunde e.V. Bd.53). Brandt, C. 1992. Sprache und Sprachgebrauch der Mennoniten in Mexiko. Marburg: N.G. Elwert Verlag. (Schriftenreihe der Komission für ostdeutsche Volkskunde in der deutschen Gesellschaft für Volkskunde e.V. Bd.61). De Graaf, T. & R. Nieuweboer. 2001. “The language of the Siberian Mennonites”. New Insights in Germanic Linguistics ed. by I. Rauch and G.F. Carr, 21–34. N.Y. Berkley Insights in Linguistics and Semiotics, Vol.2. Goerzen J.W. 1952. Low German in Canada, a study of ‘Platdietsch’ as spoken by Mennonite immigrants from Russia. Diss. Foerste, W. 1957. “Geschichte der niederdeutschen Mundarten”. Deutsche Philologie im Aufriß. Hrsg. von W. Stammler. 2.überarbeitete Auflage. Bd.1. Berlin: Erich Schmidt Verlag. Frings, Th. 1967. “Vom g und seinen Lautwerten und von germanischen Sprachlandschaften”. Beiträge zur Geschichte der deutschen Sprache und Literatur 88, 329–357. Halle. Gréb, J. 1921. “Palatalisierung in Zipser Mundart von Hobgarten”. Zeitschrift für deutsche Mundarten 16.67–76. Heidemann, G. & N. Sawatzky. 1996. “Etjch wejtj von Schiller “Der Taucher”. Zur autochtonen Varietät des Niederdeutschen von Aussiedlern aus der ehemaligen Sowjetunion”. Niederdeutsches Wort Bd. 36.75–106. Jakobson, R. 1962. “K charakteristike evrazijskogo jazykovogo sojuza”. Roman Jakobson. Selected Writings. Phonological Studies. V.1 The Hague: Mouton & Co. S’-Gravenhage. Jedig, H. 1966. Laut- und Formenbestand der niederdeutschen Mundart des Altai-Gebiets. Berlin: Akademie-Verlag. (Sitzungsberichte der
THE CONSONANT SYSTEM IN MENNONITE LOW GERMAN
85
Sächsischen Akademie der Wissenschaften zu Leipzig. Philologischhist.Klasse, Bd. 112, H.5). Klassen, H. 1993. Mundart und plautdietsche Jeschichte. Marburg: N.G.Elwert Verlag. (Schriftenreihe der Komission für ostdeutsche Volkskunde in der deutschen Gesellschaft für Volkskunde e.V. Bd. 65). Kanakin, I. & M. Wall. 1994. Das Plautdietsch in Westsibirien. (Lingua Mennonitica. Rijksuniversiteit Groningen). Kasatkin, L.L. 1968. Progressivnoe assimiljativnoe smjagčenie zadnenebnych soglasnych v russkich govorach. Moskau: Nauka. ----------, ed. 1989. Russkaja dialektologija. Izd. 2 pererabotannoe. Moscow: Prosveščenie. Koerth, A. 1913. “Zur niederdeutschen Mundart aus der Gegend von Rogasen in Posen”. Zeitschrift für deutsche Mundarten 6.275–281. Lasch, A. 1974. Mittelniederdeutsche Grammatik. 2nd unaltered edition. Tübingen: Max Niemeyer Verlag. Loopstra J.J. 1935. De assibilatie in de oudfriese oorkonden. Haarlem: H.D. Tjeenk Willink & Zoon N.V. Lorentz, F. 1925. Geschichte der pomoranischen (kaschubischen) Sprache. Berlin und Leipzig: W. de Gruyter. Mitzka, W. 1928. Sprachausgleich in den deutschen Mundarten bei Danzig (= Königsberger Deutsche Forschungen 2). Königsberg. 1928 = Kleine Schriften zur Sprachgeschichte und Sprachgeographie. Hrsg. von L.E. Schmitt, 1968: 211–230. Berlin: Mouton de Gruyter. ----------. 1930. Die Sprache der deutschen Mennoniten. (Heimatblätter des Deutschen Heimatbundes. Danzig. Heft 1). ----------. 1959. Grundzüge nordostdeutscher Sprachgeschichte. 2nd edition. Marburg: Elwert Verlag. Deutsche Dialektgeographie, Bd.59. Moelleken, W. W. 1972. Niederdeutsch der Molotschna- und Chortitzamennoniten in British Columbia / Kanada. Phonai. Deutsche Reihe. Band 10. Monographien 4. Tübingen: Max Niemeyer Verlag. ----------. 1987. “Die linguistische Heimat der rußlanddeutschen Mennoniten in Kanada und Mexiko”. Niederdeutsches Jahrbuch, Jahrbuch des Veriens für niederdeutsche Sprachforschung 110.89–123. Naiditch, L. 2001. “Die Entwicklung des Vokalismus im Mennonitenplatt”. Beiträge zu Linguistik und Phonetik. Festschrift für Joachim Göschel zum 70. Geburtstag. Hrsg. von A. Braun. Stuttgart: Steiner. Zeitschrift für Dialektologie und Linguistik. Beihefte. H.118: 242–256. Niebaum, H. 1985. “Phonetik und Phonologie, Graphetik und Graphemik des Mittelniederdeutschen”. Sprachgeschichte. Ein Handbuch zur Geschichte
86
LARISSA NAIDITCH
der deutschen Sprache und ihrer Erforschung. Hrsg. von W. Besch, O. Reichmann, S. Sonderegger. Berlin, New York. Halbband II: 1220–1227. Nieuweboer, R. 1999. The Altai Dialect of Plautdiitsch. West-Siberian Mennonite Low German. (LINCOM Studies in German Linguistics 07.) München/Newcastle: Lincom Europa. Penzl, H. 1947. “The Phonemic Split of Germanic k in Оld English”. Language XXIII.34–42. Pinnow, J. 2000. Düüsend Worta Schtoothöffa Plaut. Tausend Worte Stutthöfer Platt. Gesammelt von Kurt Gutowski und Harry Grieger. Zusammengestellt, ergänzt, erläutert und herausgegeben von Jürgen Pinnow. Westerland / Sylt. Pinnow, J. & L. Tetzlaff. 1998. Von Labommels on andre spoß’ge Lied von Otto Müller in Danziger Platt mit übertragung in die Mundart der Danziger Niederung und Übersetzung ins Hochdeutsche von Lothar Tetzlaff. Hrsg. von Jürgen Pinnow, Lothar Tetzlaff. Westerland / Sylt. Quiring, J. 1928. Die Mundart von Chortiza in Süd-Rußland. PhD thesis, München. Reimer, A., A. Reimer & J. Thiessen, eds. 1983. A sackfull of Plautdietsch. A Collection of Mennonite Low German Stories and Poems. Winnipeg, Canada: Hyperion Press Limited. Rohkohl, K. 1993. Die Plautdietsche Sprachinsel Fernheim/Chaco (Paraguay). (Schriftenreihe der Komission für ostdeutsche Volkskunde in der deutschen Gesellschaft für Volkskunde e.V. Bd.64). Marburg: Elwert Verlag. Schirmunski, V. 1928. Die deutschen Kolonien in der Ukraine. Moskau: Zentral-Völkerverlag der Sowjet-Union. Semrau, M. 1915. “Die Mundart der Koschneiderei”. Zeitschrift für deutsche Mundarten 8.143–202. Teuchert, H. 1913. “Die niederdeutsche Mundart von Putzig in der Provinz Posen”. Zeitschrift für deutsche Mundarten 6.3–44. Thomason, S.G. & T. Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley, Los Angeles & London: University of California Press. Thiessen, J. 1963. Studien zum Wortschatz der kanadischen Mennoniten. Marburg. (Deutsche Dialektgeographie, 64). Valiska, J. 1967. Die zipserdeutsche Mundart von Chmel’nica (Hopgarten). Bratislava. Žirmunskij, V.M. 1956. Nemeckaja dialektologija. Moskva, Leningrad.
ENGLISH DIALECTS IN THE BRITISH ISLES IN CROSS-VARIETY PERSPECTIVE A BASE-LINE FOR FUTURE RESEARCH ∗
SALI TAGLIAMONTE†§, JENNIFER SMITH†, HELEN LAWRENCE† University of York, UK†, University of Toronto, CDA§ 1.
Introduction The task of establishing historical relationships between new and old world varieties of English is one of the ‘hot topics’ in contemporary language variation and change as researchers attempt to trace the origins and development of linguistic features in North American dialects back to the British Isles (Clarke 1997a, 1997b, 1999, Hickey 2005b, Jones & Tagliamonte 2004, Poplack 2000, Poplack & Tagliamonte 2001, 2005, Rickford 1986, Tagliamonte & Smith 2002). The quest for the origin of dialect differences in North America is not new. It can be traced to the early 1900s when there was a surge of interest in spoken American English and its history (Kurath 1928, 1964). Despite a great deal of research in subsequent decades, however, there is more debate than consensus on precisely how linguistic links can be established between the varieties of English that left the British Isles during the colonization period and those that emerged in the rest of the world. A major conundrum has been to set the standards for comparison (Montgomery 1989, Montgomery 2001, Schneider 2005) and to constitute appropriate research methods (Clarke 1997b, Poplack & Tagliamonte 1991, 2001). Poplack and Tagliamonte (1991: 316) argue that in order to resolve the prior status of a linguistic feature:
∗ The first author gratefully acknowledges the support of the Economic and Social Research Council of the United Kingdom (the ESRC) for research grant #R000239097, Back to the Roots: The Legacy of British Dialects and the Arts and Humanities Research Board (the AHRB) for research grant AN6093/APN11081 Vernacular Roots: A database of British dialects. We would like to thank Annie Johnson, Mercedes Durham, and Rika Ito for their assistance in extracting and coding the data. We dedicate this article to the ‘northerners’ in Cumnock, Cullybackey and Maryport who took the time to share their stories with us and provide this legacy of English dialects of the British Isles for the future.
88
TAGLIAMONTE, SMITH & LAWRENCE
it is not its current existence that is decisive, nor even its rates of occurrence, but its distribution in the language, as determined by the hierarchy of constraints conditioning its appearance.
Bailey and Ross (1992: 520) echo this stipulation with their “roots fallacy” where they argue that simply “identifying the analogs of forms in source languages” cannot adequately explain their history. Given these developments in methodology, the checklists of features which typify some research may not be sufficient, nor are simple proportions. But are the criteria proposed by Poplack and Tagliamonte too rigorous, as argued by Clarke (1997b: 291), or necessary and vital as argued by Jones and Tagliamonte (2004)? Until more linguistic features and more situations have been examined, it is perhaps too early to tell, especially when many old linguistic variants and variables may have already disappeared from transported dialects (Clarke 1997b, Poplack & Tagliamonte 2001: 241). Perhaps the most “crucial long-ignored difficulty” (Montgomery 1997: 123) is the dearth of information from earlier non-standard, regional English dialects of the British Isles 1 (Clarke 1997b: 290, Montgomery 1997: 122). Without the perspective from the source dialects, the enterprise of making trans-Atlantic comparisons suffers from lack of historical and comparative linguistic evidence. Unfortunately, any real facsimile of the original spoken language of the original settlers is beyond reach. Thus, researchers have resorted to studying written sources, particularly old letters written by immigrants to their families back in Britain (e.g. Montgomery 1995). Indeed, Montgomery (1995: 15) argues that such data is “as rich a source as there is to be found”. Others have mined data from the rich oral traditions in enclave communities in North America (Poplack 2000, Poplack & Sankoff 1987, Poplack & Tagliamonte 1989, 1991, 2001, Wolfram & Thomas 2002, Wolfram et al. 1997, Wolfram et al. 1998). However, in this paper we explore the possibility that there is another invaluable source: many of the linguistic features which have figured prominently in the literature involve linguistic phenomena that can still be heard in some dialects in Britain. Can such data provide adequate comparative evidence to track trans-Atlantic connections? An advantage of such materials, if they can be acquired, is that through accountable fieldwork, sample design and corpus construction we can gain access to sufficient amounts of data which are representative of the vernacular norms of dialects at the community level
1
We use “English dialects of the British Isles” as a neutral, geographically based, cover term for the varieties we target for investigation.
ENGLISH DIALECTS IN THE BRITISH ISLES
89
(Labov 1972, see also Poplack & Tagliamonte 1991). If so, we may be able to gain much more in-depth knowledge of the grammar of particular linguistic features. Such information will contribute important linguistic details for establishing which linguistic features are actually diagnostic of the putative source dialects and also which features may distinguish them. One of the major goals of this paper is to introduce new data from a number of communities in Britain whose antecedents have direct historical connections with North America. To date, there are very few accountable quantitative studies of English dialects in the appropriate source regions in the British Isles. In fact, there is almost no evidence from these specific locales. Thus, we target a number of non-standard linguistic features which are differentiated on key characteristics. The perspective afforded by cross-variety comparison will impel us to seek explanations beyond a single community of speakers and outside the boundaries of a single region. A broader goal is to provide new evidence for use in future research contributing to the enterprise of making trans-Atlantic connections. More generally we hope that these results may broaden the understanding of the differential development of English in varying contact situations over the last 300 years (Mufwene 2001). To situate this enterprise, we begin with a historical perspective. 2. 2.1
Historical background British roots: American soil During the colonial period in the 18th century, at least 275,000 people left Britain for North America (Bailyn 1986, Bailyn & DeWolfe 1986, Fischer 1989: 609, Wood 1989). Although these migrants came from many different locales, the vast majority who immigrated between 1717–1775 originated from the lowlands of Scotland, Northern Ireland, and the northern counties of England (Campbell 1921: 51, Fischer 1989: 619, Landsman 1985: 8). The American destinations of these British migrants were somewhat concentrated — southwestern Pennsylvania, western parts of Maryland and Virginia, North and South Carolina, Georgia, Tennessee, Kentucky and the Appalachian mountains (Crozier 1984: 315, Fischer 1989, Leyburn 1962: 184–255). Although certain areas involved British settlers from a wide-range of dialect regions (McDavid 1985), census data reveals that the British “northerners” outnumbered all other population groups in most regions (Fischer 1989, Montgomery 2001). Thus, there are socio-historical links between Northern Ireland, Lowland Scotland and Northern England and the mid to southern United States during the colonial period.
90
TAGLIAMONTE, SMITH & LAWRENCE
2.2
Back to the roots The areas from which most British migrants to North America originated between 1717–1775 were the counties of Derry, Antrim and Down in Northern Ireland; Ayr, Dumfries, and Wigtown in Scotland; and Cumberland and Westmorland in Northern England, (Fischer 1989: 622, Leyburn 1962: 94). The sea itself is said to have united the regions bordering the Irish sea “in a single cultural region” which was “unique in its speech” (Fischer 1989: 622, 786): first, the dialects of Lowland Scotland and Northern England both have their origins in the Northumbrian dialect of Old English (Murray 1873). This, coupled with proximity to Scotland, is said to be “reflected in the linguistic characteristics of the area” (Beal 1993: 187–188), which shares many of its features with Scots. Similarly, Lowland Scotland has strong linguistic links with certain parts of Northern Ireland. The Ulster Plantations of 1610 (e.g. Gregg 1985: 9, Leyburn 1962: 94) led to a dominance of Scottish migrants in certain areas and the development of a unique variety — Ulster Scots (e.g. Barry 1981, Gregg 1985), which is different from other varieties in Ireland, which developed out of contact with speakers from England. The connections between these source linguistic regions as well as the relative proportion of British migrants from these areas in the southern United States in the colonial period suggest that the available linguistic model for English must have been at least somewhat “northern” in nature. According to the Founder Principle (Mufwene 1996: 122–123), the language to emerge in such a situation would be influenced by the frequency and nature of the linguistic features of the varieties spoken by the dominant founder population, as these have ‘selective advantage over competing alternatives’ (Ibid.). Similar processes are invoked for dialects in contact (Trudgill 1986, 1999, Trudgill et al. 2000). Thus, we hypothesize that the variety of English that developed in the American South during the colonial period may have contained features from these areas and some of the constraints on their use. One of the long-term goals of this research program is to test that possibility. Before embarking on such a project, however, we must first have detailed information about the relevant dialects to establish the point of comparison from the right place(s) and further, to assess whether the materials provide data which can be taken to represent the right time. 3.
The data As outlined elsewhere (Poplack & Tagliamonte 2001: 97, Tagliamonte & Smith 2000: 141), one of the most informative constructs which can shed light on the origins and development of languages is the relic area. Such areas, because of their peripheral geographic location and/or isolated social political
ENGLISH DIALECTS IN THE BRITISH ISLES
91
circumstances, tend to preserve older features. In fact, the construct of the relic area has been fundamental to historical linguistics in reconstructing earlier stages in the history of a language (Anttila 1989: 294, Hock 1986: 442). Among the more recent criteria for establishing the status of a variety as “peripheral” include geographic location, historical continuity and group identity (Wolfram & Thomas 2002: 26–36). Our fieldwork and data collection come from just such areas. These are indicated on the Map in Figure 1. 2
Buckie
Cumnock Cullybackey Maryport
Wheatley Hill
Tiverton
Figure 1: The communities from which the data were collected.
2
We refer to these corpora by the names of the communities where the data were collected.
92
TAGLIAMONTE, SMITH & LAWRENCE
Maryport, Cumbria. Maryport is an ex-trading port on the north Cumbrian coast about 27 miles from Carlisle with a population of 11,500. It was settled in the mid 18th century and developed along with the main industries of coal mining, fishing and ship building. In recent years these traditional industries have dramatically declined. Unlike some areas of Cumbria, Maryport was not included in the tourist boom of the last century and still remains off the traditional tourist route. Cumnock, south west Scotland. Cumnock is a small ex-mining town in Ayrshire (population 11,000) in south west Scotland. The nearest substantially sized town is Ayr, sixteen miles to the west. The area has been settled for at least 800 years. In the 18th century there were substantial industrial developments and a railway link established Cumnock as a growing mining community in the 19th century. The collapse of this industry in the 1980’s has resulted in high unemployment and a breakup of the traditional fabric of the community. Cullybackey, Northern Ireland. The village of Cullybackey (population 2,500) is situated 30 miles north-west of Belfast in County Antrim, Northern Ireland. This is a predominantly Protestant area with most of the residents of Scottish descent. The main industries are agricultural, retail, and manufacturing service. The Environmental Improvement Plan for the area states that “despite its picturesque setting, Cullybackey has not developed its potential as a leisure/tourist destination, or stopping off point”. Of these three areas, the connection between Cumnock and Cullybackey is the closest. Ayrshire, where Cumnock is located, was one of the source areas for the Ulster plantations in Northern Ireland. To these corpora we add the rich resources of a number of other datasets, which are comparable across a range of extra-linguistic characteristics. This will allow us to explore further regional comparisons in the use of particular linguistic features. Tiverton. Tiverton is a small town in mid Devon. In this area, the industry has always been primarily agricultural and for many residents, continues to be today (Godfrey & Tagliamonte 1999). Wheatley Hill. Wheatley Hill is a village in County Durham in northeast England. It was once an insular mining community, but in the last few decades many of the mines have closed down, forcing the inhabitants to find work in nearby urban centres (Martin & Tagliamonte 1999).
ENGLISH DIALECTS IN THE BRITISH ISLES
93
Buckie. Buckie is a small town on the far northeast shore of Scotland. Despite the decline in the traditional fishing industry, the community has been able to maintain its cultural cohesiveness due to the maintenance of local employment provided by the oil industry (Smith 2000). The speaker sample and total number of words in each corpus is shown in Table 1. All the speakers are over 60 and were born and raised in the community.
Tiverton [TIV] Maryport [MPT] Wheatley Hill [WHL] Cumnock [CMK] Buckie [BCK] Cullybackey [CLB]
MALE FEMALE TOTAL SPEAKERS 7 2 9 20 23 43 3 3 6 18 23 41 4 5 9 12 5 17 Table 1: Sample design and characteristics.
TOTAL WORDS 96,472 401,376 206,320 349,428 302,894 198,086
In each locale we interviewed the most insular speakers from the oldest generation at the time of the fieldwork. Crucial to the enterprise of collecting representative dialect data, in all the peripheral communities the fieldworker was an in-group community member (Poplack & Tagliamonte 1991). 3 Each corpus comprises tape-recorded conversations representing in many cases hundreds of thousands of words (see Table 1) which include discussions about local traditions, narratives of personal experience, local gossip and informal discussions. While there are undoubtedly formality effects operating within the context of the interview situation, these are within normal parameters of conversational interaction. None of our material contains dramatic styleshifting or the self conscious speech that appears to be present in fieldwork sites where the interviewers were alien to the community (Schilling-Estes 1998). Indeed, the ‘broad’ dialectal quality of these materials along with their generally informal tone makes us confident that the speech faithfully reflects the typical discourse found in each community and is as close as we can get to the vernacular norms of the regional dialects.
3 The only exception is Cumnock where the second author carried out the interviews. Although not native to this particular village, she shared salient cultural characteristics with community members.
94
TAGLIAMONTE, SMITH & LAWRENCE
4. 4.1
Diagnostic features investigated Introduction The most important decision to make in tracking linguistic origins is to find the right linguistic feature to examine (Hickey 2005b, Jones & Tagliamonte 2004, Montgomery 1989, Poplack & Tagliamonte 2001, Wolfram 2000: 47). Ideally, a diagnostic should have a unique association with a source variety. However, this is a very demanding criterion, as we shall see. At the very least the linguistic feature should be attested in the comparison varieties and occur frequently enough in each of them to admit large scale quantitative analysis of variable conditioning (Poplack & Tagliamonte 2001: 100–101). However, many prized features are simply not frequent enough, others may not exist in all the relevant dialects under investigation, some are just not diagnostic (Schneider 2005: 284), but the product of universal tendencies shared by all varieties (e.g. Chambers 1995, 2001). Moreover, any one feature is unlikely to provide all the relevant qualities. As Wolfram and Thomas (2002: 16) argue with respect to African American Vernacular English: An authentic picture of earlier African American speech can emerge only if a wide array of dialect structures is considered … While selective single-structure studies may reveal significant insight into a particular linguistic process, they may obscure or even distort our understanding of the overall relationship of African American speech to other varieties.
In an attempt to approach this level of adequacy, we shall compare and contrast four different variables: verbal -s, (1), NEG/AUX contraction, (2); the for to infinitive, (3) and zero adverbs, (4). Equally as important as having an array of features to study, however, is the fact that each of these features differs in status across varieties. The first two: verbal -s and NEG/AUX contraction, are reported to distinguish northern varieties of British English. Verbal -s is the “show case variable” (Rickford et al. 1991: 104) for making trans-Atlantic connections. NEG/AUX variation presents another putative diagnostic of specific regional dialects in Britain, but much less is known about it. The third variable, the for to infinitive, is a remnant from an earlier stage of English, but may have receded differentially across regions. The fourth — the zero adverb — is more pervasive, it is reported across all vernacular English dialects. Verbal -s (1) a. People has no sense with youngsters. (CLB:j) b. They come up to the moss and they get on. (CLB:j)
ENGLISH DIALECTS IN THE BRITISH ISLES
95
NEG/AUX contraction (2) a. But that isn’t what you wanted was it? (TIV:b) b. And it’s not put on or false or anything like that. (TIV:b) For to (3)
a. There was no buses to take you back and forward then. (MPT:v) b. So the roads were crowded when it was time for to start. (MPT:v)
Zero adverbs (4) a. It was a real lovely. (WHL:c) b. It’s really marvellous when you think back. (WHL:c) Thus, we have data from six different dialects around the British Isles, and four features which should both distinguish them on some counts, and ally them on others. We now subject these data to empirical analysis. We follow methods and criteria previously established in the comparative quantitative sociolinguistic tradition (Poplack & Tagliamonte 1989, 1991, 2001, Tagliamonte & Smith 2000, Tagliamonte 2001). Further, the cross-variety comparative method will serve as an important check in controlling for the possibility of language universals (see also Poplack & Tagliamonte 2001: 101), while the various lines of evidence provided by these techniques will serve to assess the relationship amongst varieties (see Poplack & Tagliamonte 2001: 94). All these will contribute to assessing whether any of the forms under investigation can serve as useful diagnostics for making trans-Atlantic connections. 4.2
Verbal -s Variable use of verbal -s in 3rd person plural subjects, as in (5) is the feature most widely-used for establishing links between North America and Britain. (5)
a. They hate being called Americans. They’re Canadians. The really take offence if they call them American. (BCK:i) b. When they go back, the teachers asks them to write something. (BCK:b) c. No discipline at home, no discipline in the school. And they go on that. (CLB:b)
96
TAGLIAMONTE, SMITH & LAWRENCE
d. Youngsters gets far too much and they’ve no manners some of them at all. (CLB:e) e. They have a student there every year for help with the lambing, a student vet. (TIV:r) f. Small children goes er up there you see, they have super night there, kiddie’s night, very nice really for ’em. (TIV:p) g. I know there’s a lot of young people at the weddings now and they want a disco. (CMK:g) h. Him and my brother goes out drinking. (CMK:G) i. They live in the end bungalow at the bottom of the bridle-path and you go- the farm’s theirs. (WHL:y) j. There ø always one or two starts to fade a bit. (WHL:c) k. I said, ‘Lads at pits has three pound’. (MPT:a) l. They only live two doors off us. (MPT:a) The origin of verbal -s can be traced back to Scottish, Northern Irish and Northern English dialects (Clarke 1997a, Montgomery 1989, 1994, Montgomery & Fuller 1996, Poplack & Tagliamonte 1989, Schneider 1981, 1983). Not only did this feature exist at earlier stages in these regions, but also a number of constraints operated on its application: when the subject is a noun the verb takes an -s suffix, but when the subject is a pronoun, it does not (Murray 1795/1968: 211). 4 This rule is attested as far back as the 13th century (Murray 1795/1968: 212). Montgomery’s (1997) analysis of 16th and 17th century letters confirm that it was present in the written record of Ulster Scots and Bailey and Ross’s study of Ship English confirm that it was present amongst 18th century sailors (1988). Both studies make an explicit transAtlantic link. Indeed, most studies concur that this is a transported constraint par excellence. In fact, the use of plural -s in contemporary Appalachian English in the United States displays “remarkable retention of linguistic patterns and constraints across more than four centuries and two continents in
4 Other constraints relate to the adjacency of the subject (Murray 1795/1968:211); however, these data do not contain sufficient non-adjacent contexts across all communities to make a comparison possible.
ENGLISH DIALECTS IN THE BRITISH ISLES
97
the evolution of Scots English to Scotch-Irish English to Appalachian English ” (Montgomery 1997: 137). What do the data from our six varieties reveal? Table 2 shows the distribution of verbal -s across plural NPs and pronominal they. 5
BCK CMK WHL MPT CLB TIV
NP PRO % N % N 74 23 2 123 25 117 1 555 22 41 0 216 24 192 1 617 71 56 2 259 49 65 33 223 Table 2: Overall distribution of NP vs pro across communities.
All the northern communities maintain the historically renowned NP/PRO distinction. Indeed, we can confirm that it is still intact in these contemporary dialects after 700 years, and this despite its more recent demotion to non-standard status. In Tiverton, on the other hand, we observe something a little different: the pronouns are not categorically unmarked. Nevertheless, the NP over PRO ranking prevails (see also Godfrey & Tagliamonte 1999). The strong convergence of results from all the communities provides further confirmation that the NP/PRO constraint must have been present in the source dialects and made its way via diffusion to North American dialects. This robust data confirms the findings from much smaller data sets gleaned from historical written sources. In addition, these same dialects also exhibit variation in was and were and is and are (Smith & Tagliamonte in preparation), just as we reported for was and were in Buckie in Tagliamonte and Smith (2000). Thus, although “it is difficult to really pin down individual sources and lines of historical transmission” (Schneider 2005: 264), verbal -s appears to be a linguistic feature whose lineage is transparent. These findings substantiate that these present-day dialects do preserve features typical of earlier stages in the history of the same dialects. 6 We submit
5
In these analyses, the NP category included relative pronouns, in line with the historical record (Murray 1795/1968: 211). Subjects such as some of them were first treated separately but since they were found to pattern along with NPs in their patterns of use, they were subsequently collapsed with NPs.
TAGLIAMONTE, SMITH & LAWRENCE
98
therefore that these data provide an informative window on the past, and thus may contribute important evidence for tracking the origins of North American dialects. 4.3
NEG/AUX contraction The second variable is NEG/AUX contraction, which may occur with the verbs be, have and will, as in (6)–(8). 7 (6)
AUX contraction a. But she’s not that daft, though, Geoff. (MPT:%) NEG contraction b. I said “I know it isn’t gonna affect you.” (MPT:%)
(7)
AUX contraction a. And I’ve no been so mobile since. (CMK:A) NEG contraction b. I don’t know her, I haven’t seen her. (CMK:A)
(8)
AUX contraction a. He’ll not be better again Margaret, no.(CLB:e) NEG contraction b. And you won’t have the same interest. (CLB:e)
Although this feature has never been used to establish trans-Atlantic links, there is good reason to assume that it might provide an excellent diagnostic. According to Trudgill (1978: 13) the frequency of AUX contraction increases “the further north one goes”. The forms with n’t are more common in Southern British English, while in Scotland and Northern England, “forms such
6
Data from 18th and 19th century Yorkshire letters and diaries do not exhibit the NP/PRO constraint (Montgomery to appear). Based on this evidence Montgomery concludes that “North Britain does not appear to have been a single linguistic territory for variation between was and were” (Montgomery to appear). This may be true. However, Yorkshire is not amongst the main regions included in the 1717–1775 migrations to North America. Our data, on the other hand, which come from the areas of greatest population migration to North America do in fact appear to represent a single linguistic territory for this particular linguistic feature, i.e. the NP/PRO constraint with was/were, is/are and verbal -s. 7 Some communities differ in the phonological rendition of the overt negator not and the cliticised form n’t. In Cullybackey and Cumnock, the non-cliticised form is no and the cliticised form is nae. In Buckie, the non-cliticised form is nae and the enclitic is na. In this analysis all these forms are referred to as not (non-clitic) or n’t (enclitic).
ENGLISH DIALECTS IN THE BRITISH ISLES
99
as ‘ll not are preferred to forms such as won’t” (Aitken 1984, Beal 1993, see also Haegeman 1981, Miller 1993, Quirk et al. 1985, Swan 1980: 159). Do the data from our six varieties substantiate these claims? Table 3 shows the overall distribution of AUX contraction with be, have and will across the communities. TIV WHL MPT CLB CMK BCK % N % N % N % N % N % N have 1 52 0 37 0 234 7 72 3 99 0 57 would 0 40 0 26 0 151 1 168 0 89 0 54 will 0 24 70 10 6 70 91 64 88 32 0 22 be 43 82 98 48 50 163 97 118 100 162 100 83 Table 3: Overall distribution of auxiliary contraction with have, would, will and be by community.
The propensity for AUX contraction differs dramatically depending on the verb. Have has virtually no AUX contraction, nor does would in any community. This means that examples such as (9) are the predominant forms. (9)
a. They havena been getting the break. (BCK:f) b. No my Daddy wouldn’t let you put artificial manure to the spuds. (CLB:n)
Distributional results for will and be present an entirely complicating perspective. Two communities have very frequent AUX contraction with will: Cullybackey at 91% and Cumnock at 88%. 8 These varieties typically contain examples such as (10). (10) a. Then I’ll no get hurt, I’ll no be choked. (CLB:b) b. It’ll no be a lot smaller but it is smaller, aye. (CMK:A) According to the reports in the literature there should be a consistent pattern for the northern communities. However, Maryport and Buckie in the north pattern with Tiverton in the south, in using forms such as (11). (11) a. I said ‘One thing, when I die, you winna have no debt.’ (BCK:g) b. He won’t talk if ee do that. (TIV:a) c. When I do my talking, thou won’t understand us. (MPT:a)
8
Wheatley Hill has too few contexts of use (N=10) for comment.
100
TAGLIAMONTE, SMITH & LAWRENCE
As far as the verb be is concerned, Buckie, Cumnock, Cullybackey and Wheatley Hill have categorical or near categorical AUX contraction, as in (12). (12) a. The driver’s nae gan naewhere. (BCK:d) b. It’s not every day that your husband goes up in flames. (WHL:a) Maryport and Tiverton are different. Here, AUX contraction occurs, but it is highly variable with NEG contraction, as in (13). (13) a. Children aren’t allowed to buy anything now. (TIV:c) b. There isn’t anybody can tell me that they can dance. (MPT:x) In sum, two main points can be made. First, there is variation amongst the dialects in the propensity of AUX contraction, both for will and for be. But these do not support a broad regional split between north and south. Second, of the three varieties that make up the key input dialects to North America, two (Cullybackey and Cumnock) pattern the same, and the other patterns completely differently (Maryport). The distributional analysis confirms a complex pattern of regional differentiation. As far as we can ascertain a similar situation exists in North America, as Yaeger-Dror (2002) highlights in a recent paper: “The more [a] corpus is expanded to include speakers from different regions, the more diverse the regional pattern becomes.” This is exactly what we have experienced here (see also Tagliamonte & Smith 2002). Moreover, we cannot establish whether the current situation in Britain is a historical remnant or a product of recent developments. Unlike the case of verbal -s, there is no straightforward picture of the historical patterns of use for this feature. For all these reasons, we conclude therefore that this variable is a poor diagnostic for making transatlantic connections. 4.4
A relic feature: for to A third linguistic feature is the for to infinitive, which is variable with the infinitive to in some English dialects in the British Isles (e.g. Beal 1993, Edwards & Weltens 1985, Finlay 1988, Henry 1995, Macafee 1983, Macaulay 1991), as in (14). Tiverton (14) a. He’d light a furnace for to wash the clothes (TIV:a) b. Jobs took five times as long but you had the labour to do it. (TIV:I)
ENGLISH DIALECTS IN THE BRITISH ISLES
101
Wheatley Hill c. They’ve got the money to pay the debts. (WHL:k) d. I was called up for to go and get a- a medical. (WHL:c) Maryport e. I went into a shop to be a shop assistant. (MPT:v) f. I were fortunate enough for to get the higher-national. (MPT: ¢) Cullybackey g. And I went down to the market there last Monday to check the slurry tank. (CLB:b) h. And I washed nine baskets of prittas every day and them was boilt for to feed everything. (CLB:a) Cumnock i. And I brought a pail of coal in to mend the fire and I fell with the pail of coal. (CMK:m) j. He must’ve kent I must’ve got tickets for to get through the gate. (CMK:f) Unlike verbal -s or NEG/AUX contraction, this is not a feature circumscribed to any particular dialect area in Britain. It is mentioned in the inventories provided by Schneider (2005: 264) and Hickey (2005a), though not as a diagnostic. However, certain linguistic constraints operated on this variable in the past (e.g. Warner 1982) and continue to distinguish contemporary dialects (e.g. Henry 1995). Semantic and syntactic constraints distinguish the use of for to. In some areas, it is restricted to clauses of purpose or intention (Macafee 1983: 50–51). In others, a broader range of contexts is found (Finlay 1988, Macaulay 1991: 106). Given that even obsolescing features may still maintain linguistic conditioning (Jones & Tagliamonte 2004) a distributional analysis may uncover important evidence. However, we found that this form was very rare indeed. In Tiverton, there were only 3 tokens, in Wheatley Hill 8, and in Buckie, none. Three of the communities had substantially more tokens — Maryport, Cullybackey and Cumnock. Table 4 shows the overall distribution of for to infinitives across communities.
TAGLIAMONTE, SMITH & LAWRENCE
102 MPT % 2
CLB
CMK
N % N % N 2773 1 1440 1 2423 Table 4: Overall distribution of for to infinitive across varieties.
The rarity of this feature becomes even more dramatic when it is viewed as a proportion of all possible contexts. The for to infinitive represents only 1– 2% of the entire set of infinitives in these three corpora. Closer examination of the data according to these syntactic environments reveals that the for to infinitives are restricted to clauses of purpose in Cumnock and Cullybackey, as in (14g–j). In Maryport, however, for to can appear with a verbal complement as well, as in (14e&f). Thus, like the patterns of use for NEG/AUX variability, the northern communities are actually different amongst each other. Once again Cumnock and Cullybackey pattern in tandem, but not Maryport. However, it is difficult to determine whether these inter-community differences are the result of actual dialect differences or ongoing obsolescence. Thus, in the end the for to variable provides little evidence which can be used for diagnostic purposes. There is too little data here for establishing trans-Atlantic connections. 4.5
Features reported in all English dialects: zero adverbs The last feature we will consider is the unmarked, or what we will refer to as, zero adverbs, as in (15). (15) a. Oh aye, they go regular. You’re lucky, Morag’s daughters go regular. (CMK:g) b. Oh I could’ve had a job quite easy with him. (CMK:C) These are widely reported in English dialects in the British Isles (e.g. Milroy & Milroy 1993, Tagliamonte & Ito 2002, Trudgill 1990: 86) and in North American dialects (Görlach 1991: 103, Mencken 1961), most often for non-standard usage (Quirk et al. 1985). Variation between -ly and zero adverbs is the result of diachronic change. Zero is the earlier form which is gradually replaced by -ly in Early Modern English (Nevalainen 1994a, 1997). Given its productivity both at earlier stages in the history of English as well as more recently, we would expect all our data to evidence this feature. And indeed it does. A full 45% of all the dual form adverbs surface as zero, resulting in robust alternation in our data. This is highlighted by the examples in (16) from the same speakers in the same stretch of discourse.
ENGLISH DIALECTS IN THE BRITISH ISLES
103
(16) a. And he was awful homesick you know my Uncle John. (CMK:j) b. He worked awfully hard. (CMK:j) c. He wanted to get as many finished as he could that night as quick as he could. (WHL:c) d. Be able to think twice as quickly as your customer. (WHL:c) Table 5 shows the overall distribution of this variability across communities. TIV % 27
WHL
MPT
CLB
CMK
BCK
N % N % N % N % N % 55 36 42 19 195 41 247 48 196 92 Table 5: Overall distribution of zero adverbs by individual community.
N 136
Proportions of zero range from a high of 92% in Buckie to a relatively low of 19% in Maryport. If this is a feature that is supposed to be uniform across English dialects, why are the communities so different? An examination of the internal constraints might provide further information. Adverbs which alternate between -ly and zero comprise three main grammatical functions, intensifiers, as in (17), manner adverbials as in (18) and sentence adverbials, as in (19). (17) a. Well I was really scared. (CLB:b) b. If you did nae do what was right you get a real good battering. (CLB:b) (18) a. Aye- I could’ve passed it quite easily, you-know. (CMK:A) b. Oh I could’ve had a job quite easy with him. (CMK:A) (19) a. And I wouldn’t have give tuppence for her lyke, quite honestly. (MPT:P) b. Oh, aye ‘cos they look like that e? Honest. (MPT:P) Unfortunately, sentence adverbs were infrequent in general and erratically distributed across corpora. 9 For this reason, we restrict our examination to manner and intensifier adverbs.
9
Moreover, most of these were the expression “funnily enough”.
104
TAGLIAMONTE, SMITH & LAWRENCE
Table 6 shows the results for the communities across these different adverb types. TIV WHL MPT CLB CMK BCK % N % N % N % N % N % N manner 56 16 62 13 35 49 83 35 36 39 77 22 intensifier 15 39 27 26 12 121 34 194 50 141 95 114 Table 6: Overall distribution of zero adverbs by function and community.
We now observe a number of similarities and differences. Some communities exhibit a higher percentage of zero marking on manner adverbs: Tiverton, Wheatley Hill, Maryport, Cullybackey. Others have a higher percentage of zero marking on intensifiers: Cumnock and Buckie. This reveals that there is no pan-community hierarchy here either, therefore, we need to delve deeper into the data. Further examination of the lexical distributions within these categories may provide the key. Intensifiers. The main intensifiers used in these data are shown in Table 7. In line with Quirk et al. (1985) we also include the adverb nearly (which is an approximator) in this category. TIV WHL MPT CLB CMK BCK % N % N % N % N % N % N nearly 0 12 0 5 0 46 15 130 10 20 89 18 awfully 0 0 0 0 0 0 100 2 97 69 100 90 really 15 20 33 21 20 51 44 32 3 31 20 5 terribly 0 0 0 0 33 3 100 30 0 2 0 0 absolutely 0 3 0 0 7 14 0 1 0 16 0 0 other 75 4 0 0 38 8 100 4 0 3 100 1 Table 7: Use of zero adverbs by individual intensifier and community.
This table shows that there are community specific uses of particular intensifers: Cumnock and Buckie use awful, Tiverton, Wheatley Hill and Maryport use reall(ly) and Cullybackey uses terrible and real(ly). Moreover, within these intensifiers, the propensity for zero differs dramatically. Cullybackey, Cumnock and Buckie have categorical zero marking on awful and terrible, as in (20): Tiverton, Wheatley Hill, Cullybackey and Maryport have variable use of zero on really, as in (21):
ENGLISH DIALECTS IN THE BRITISH ISLES
105
(20) a. Aye and his mother was aie awful cheery ken. (CMK:E) b. She’s an awful nice quinie. (BCK:f) c. That’s terrible good for your blood. (CLB:l) (21) a. b. c. d.
It was real dark and the handlebars broke. (CLB:f) He was a real quiet Scotchman. (MPT:r) Makes me really wicked. (TIV:h) Your nana used to have a really hard hand you know. (WHL:d)
Nearly is also differentiated across the six communities: variable zero marking in Buckie, Cullybackey and Cumnock, but no zero marking in Tiverton, Maryport and Wheatley Hill, as in (22). (22) a. It’s near all pensioners that they depend on for the post-office. (BCK:a) b. In fact it nearly fetched country till its knees. (MPT:j) Thus, we find dramatic inter-variety differences on two counts: first, community specific selection of adverbs used to intensify and second, propensity for zero marking on these. Manner. The main manner adverbs used in these data are shown in Table 8. TIV
WHL MPT CLB CMK BCK % N % N % N % N % N % N quickly 75 4 80 5 60 5 100 7 71 7 100 3 regularly 0 1 0 0 29 7 100 1 60 10 0 0 easily 0 0 50 2 40 5 89 9 33 3 100 3 other 55 11 50 6 31 36 74 23 11 19 22 18 Table 8: Use of zero adverbs by individual lexical item and community.
In contrast to the results for intensifiers, here all communities use -ly and zero variably across all manner adverbs where there is data to tell. In other words, manner adverbs behave consistently across the communities. The differential status of intensifiers and manner adverbs in Buckie, Cullybackey and Cumnock — zero marking on intensifiers, but variable marking on manner adverbs — may be a remnant of a constraint from the historical record. Examples of the following type, (23) are found easily in written sources:
TAGLIAMONTE, SMITH & LAWRENCE
106
(23) ‘tis terrible cold … it has snowed terribly all night. (Swift, J., quoted in Jespersen 1961a: 371–372) If this historical constraint is still operational in some communities, then we might also expect the varieties to maintain another. In the trajectory of change from zero marking to -ly, concrete adverbs, as in (24), were said to occur more with the zero form than abstract adverbs, as in (25) (Donner 1991, Jespersen 1961b, Nevalainen 1994a, 1994b, 1997, Schibsbye 1965). (24) a. I’m missing a lot of conversation because I can nae hear properly. (CMK:t) b. It was all carpetted beautiful. (WHL:d) c. They did used to come and clean the drains out very, very regularly. (MPT:t) (25) a. But they’d be sore needing it. (BCK:c) b. I was never loved properly. (WHL:l) c. They were all very closely connected. (CMK:A) Table 9 separates the data according to this factor. TIV
WHL MPT CLB CMK BCK % N % N % N % N % N % N concrete 60 15 64 11 36 39 84 37 50 26 88 16 abstract 0 1 50 2 30 10 100 2 8 13 50 5 Table 9: Overall distribution of zero adverbs by CONCRETE vs ABSTRACT and community.
Remarkably, in the communities where there are sufficient data to tell — Maryport, Cumnock and Buckie — all retain this older constraint. In sum, examining the overall distribution of zero adverbs across communities revealed relatively little. From this perspective one would have reasonably concluded that zero adverbs were simply part of “general nonstandard English grammar” (Schneider 2005: 264). However, once the syntactic function of forms was distinguished, we discovered that adverbs behave quite differently, not only amongst each other, but also across communities. Moreover, once we took into account lexical effects, we discovered that intensifiers are idiosyncratic to the individual community: each selects its own. On the other hand, manner adverbs are remarkably the same across the board. Thus, as far as making trans-Atlantic connections is concerned, only one highly circumscribed area of the adverbial system
ENGLISH DIALECTS IN THE BRITISH ISLES
107
provides a potential diagnostic. This highlights the importance of underlying constraint ranking effects in a distributional analysis (see also Poplack & Tagliamonte 1991, and passim). 5.
Conclusion We have now compared four linguistic features across six English dialects in the British Isles. We have attempted to fulfill as many of the rigorous standards for comparison laid out in earlier research as possible, including historical connections, the use of distributional data, constraint hierarchies (Poplack & Tagliamonte 1991) and “the joint weight of several sets of data” (Schneider 2005). What conclusions can be drawn from these results? As we might have predicted given the tremendous consensus in the literature, verbal -s, which can be traced back to 13th century northern British regional dialects is still evident today: the five northern varieties showed identical patterns for verbal -s, along with the foremost constraint on its use. The ‘legacy’ of English dialects in the British Isles seems straightforward in this case. Such evidence corroborates the notion, following Mufwene’s (1996) Founder Principle, that this feature was transported, as well as selected into the emergent varieties in North America. Perhaps a more vexing finding for the purposes of pinpointing particular regions for trans-Atlantic connections is that the patterns and constraints for this feature apply for all the communities, not just the northern ones. Verbal -s does not appear to be a unique feature of a particular source variety in Britain. Thus, it does not present a sure diagnostic for northern origins (Godfrey & Tagliamonte 1999, Tagliamonte 2001). Interestingly, however, the regional split between a categorical pattern in the north and a variable pattern in the south may distinguish North American dialects. If so, this may provide a new angle for tracking the origins of dialects which employ verbal -s. The promising diagnostic of NEG/AUX contraction does not differentiate dialects in the way the literature predicts — higher rates of AUX contraction in the north. Nor is there any clear regional distribution of forms. However, in the midst of varying patterns across communities, two varieties stand out for their parallelism — Cumnock and Cullybackey. This isn’t surprising, of course, given the 17th century population movements from the Lowlands of Scotland to Northern Ireland during the Ulster Plantations. Indeed, these results provide quantitative linguistic evidence strongly aligned with the socio-historical links between these two communities. This provokes an interesting question — given the parallels between Appalachian English and Ulster Scots already established by Montgomery (1989, 1997, 1992), were the NEG/AUX patterns visible in these communities transported to North America
108
TAGLIAMONTE, SMITH & LAWRENCE
as well? And if not, why not? Indeed, if the current literature on NEG/AUX contraction is any indication (Yaeger-Dror et al. 2001), it does not appear that AUX contraction is robust in any dialect in North America. This may be due to the fact that unlike the case of verbal -s there is and was no uniform parallelism across all the northern dialects with this linguistic feature. The northern English community, Maryport, differs from Cullybackey and Cumnock. This means that competing variants must have been transported to North America. On the other hand, there does seem to be some evidence for distinct regional patterns in the United States, e.g. there is more AUX contraction with be in the southern United States (Yaeger-Dror et al. 2001). Thus, further research on this variable may prove to be quite informative. The obsolescing form for to is on the brink of extinction. There was simply too little data here for investigation. Unfortunately, for this feature we are too late to establish the basis for any trans-Atlantic link, at least with these varieties. This highlights the worst problem for tracking the origins of dialects. Many of the older features are long gone. With zero adverbs, on the hand, no particular regional differentiation was expected: all the varieties had robust variation between -ly and zero. But once the different adverbial functions are separated, we find that some patterns are consistent across the dialects, but others are highly community specific. Manner adverbs have all the characteristics of a universal non-standard English form. All the varieties had the same patterns of use. In the case of adverbial intensifiers, however, the dialects present a highly differentiated and multilayered system. First, the selection of the particular intensifier differs from community to community. Second, the propensity for zero or -ly marking on intensifiers differs as well. Thus, the use of intensifiers appears to be a salient marker of particular communities. Can such features be used for establishing trans-Atlantic links? This remains a highly interesting, and empirical, question for future research. To return to our original goal, the impetus for this analysis was to provide a preliminary base line perspective on a number of linguistic features in some of the non-standard regional dialects of English in the British Isles. The varieties we studied are situated in the regions of the largest migrations out of Britain and Northern Ireland and into North America during the early colonization period. Thus, in future comparative research we hope these results may shed important light on the origins and development of North American dialects. Building from on Mufwene’s (1996) and Trudgill et al.’s (2000) theories about the formation of new dialects, the findings from our four linguistic variables in six British varieties permit us to posit the following hypotheses. Where source varieties share the same linguistic feature, as well as
ENGLISH DIALECTS IN THE BRITISH ISLES
109
the constraints on its use, the feature stands an excellent candidate for transportation as well as selection intact into emergent dialects. The show case variable, verbal -s in 3rd person plural, is precisely this type. No wonder it has engendered so much research in the study of trans-Atlantic dialect connections. Not only does the feature endure in contemporary peripheral dialects, but so do constraints that were likely in operation from the time period of colonisation. In contrast, where source dialects are differentiated, the features that are transported are mixed from the outset, both in form and patterning. This dilution may have disastrous consequences for the selection process. Features which are marginal or marked when viewed against the total cohort of linguistic features appear less likely to find their way into the developing dialects. Variable NEG/AUX contraction is of this type. Two out of the three the source dialects use the form ‘ll not; the other uses won’t. This means that ‘ll not might be expected to have selective advantage. Yet it is the form won’t that is actually selected into North American varieties (Yaeger-Dror et al. 2001). Why? Here, we may appeal to the rest of the structured set in the language. Most of the other forms in the input dialects are NEG contraction, e.g. wouldn’t, haven’t, hasn’t, hadn’t. Therefore, in this case, it may be that what gets selected is the form which matches the rest of the system, pushing out the marked variant ‘ll not. The evidence from adverbs permits further elaboration of these general processes. First, given the high frequency of zero adverbs in these data we might predict that they would have had selective advantage in North American dialects. Indeed, research suggests that the zero adverb is far more frequent in North America and Britain (Opdahl 2000a, 2000b). The reason for this may now be plausibly reconstructed. The transported dialects likely had more zero than mainstream English dialects in the British Isles do today pre-disposing the emergent dialects to use it (see also Tagliamonte & Ito 2002). Given the findings for intensifiers however, we might expect that peripheral dialects in southern United States and other locales relevant to these founding dialect areas might retain their ancestor’s intensifiers as well — terrible and/or awful with varying rates of real vs. really and near vs. nearly. However, aside from Labov (1985) and Ito & Tagliamonte (2003), the study of intensifier use, layering and conditioning in peripheral North American dialects (to our knowledge) is a virtually untapped area of study, which could be fruitfully investigated. In sum, these findings offer the type of evidence which can now be used in future research to test similarities and differences in the comparison of old and new world varieties of English on a comprehensive scale. The results confirm earlier admonitions that the frequency of a given feature in one community cannot unambiguously establish if that form or its frequency is
110
TAGLIAMONTE, SMITH & LAWRENCE
unique to that community or where it would be positioned amongst others. Moreover, we suggest that unless underlying linguistic factors and lexical distributions are taken into account, the conclusions may be questionable. This highlights the advantages to be gained from taking a cross-variety quantitative, constraints-based perspective. Perhaps the most compelling finding of this study is the fact that linguistic features which are shared across all varieties, but whose variable constraints, relative strength and significance differ offer previously untapped evidence for tracking trans-Atlantic connections. Regionalisms are often difficult, if not impossible to pin down precisely (Montgomery 2001). However, given the results we have reported here, in conjunction with earlier findings on was/were variation (e.g. Tagliamonte and Smith 2000) it may be the case that linguistic features which have previously been relegated to the area of “generalized vernacular features” may actually provide very useful evidence for disentangling the origins of different dialects (see also Smith 2001, Tagliamonte 2001). Such variables may be differentiated across dialects, reflecting their progress along the cline of grammaticization in ongoing linguistic change (e.g. Poplack and Tagliamonte 2000). Thus, although communities may share the same linguistic forms, the varying strength and conditioning of factors can be correlated with the relative degree of separation of the communities from mainstream varieties (Poplack & Tagliamonte 2000, Tagliamonte & Smith 2000, Tagliamonte 2003). In this way, the imprint of dialect origins and development may be subtly etched in the contrasting patterns of alliance or variance in linguistic weights and constraints (Labov 1982: 76). Studies putting such comparative work into action promise to keep us busy for years to come. References Aitken, A. J. 1984. “Scottish accents and dialects”. Language in the British Isles ed. by P. Trudgill, 94–114. Cambridge: Cambridge University Press. Anttila, R. 1989. An Introduction to Historical and Comparative Linguistics. Amsterdam & Philadelphia: John Benjamins. Bailey, G. & G. Ross. 1988. “The shape of the superstrate: Morphosyntactic features of Ship English”. English World-Wide 9:2.193–212. Bailey, G. & G. Ross. 1992. “The evolution of a vernacular”. History of Englishes: New Methods and Interpretations in Historical Linguistics ed. by M. Rissanen, O. Ihalainen, T. Nevalainen & I. Taavitsainen, 519–531. Berlin and New York: Mouton de Gruyter. Bailyn, B. 1986. The Peopling of British North America: An Introduction. New York: Alfred A. Knopf.
ENGLISH DIALECTS IN THE BRITISH ISLES
111
Bailyn, B. & B. DeWolfe. 1986. Voyagers to the West: A Passage in the Peopling of America on the Eve of the Revolution. New York: Knopf; Distributed by Random House. Barry, M. V. 1981. “The southern boundaries of northern Hiberno-English speech”. Aspects of English Dialects in Ireland. Vol 1: Papers Arising from the Tape-Recorded Survey of Hiberno-English Speech ed. by M. V. Barry, 52–95. Belfast: Institute of Irish Studies, The Queen’s University of Belfast. Beal, J. 1993. “The grammar of Tyneside and Northumbrian English”. Milroy & Milroy 1993. 187–213. New York: Longman. Campbell, J.C. 1921. The Southern Highlander and His Homeland. New York. Chambers, J.K. 1995. Sociolinguistic Theory: Linguistic Variation and Its Social Significance. Oxford: Blackwell Publishers. ----------. 2001. “Vernacular universals”. ICLaVE 1: Proceedings of the First International Conference on Language Variation in Europe ed. by J.M. Fontana, L. McNally & M.T. Turell, 52–60. Barcelona: Universitat Pompeu Fabra. Clarke, S. 1997a. “English verbal -s revisited: The evidence from Newfoundland”. American Speech 72:3.227–259. ----------. 1997b. “On establishing historical relationships between New World and Old World varieties: Habitual aspect and Newfoundland Vernacular English”. Englishes Around the World I ed. by E. W. Schneider, 277– 293. Amsterdam and Philadelphia: John Benjamins. ----------. 1999. “The search for origins: Habitual aspect and Newfoundland Vernacular English”. Journal of English Linguistics 27:4.328–340. Crozier, A. 1984. “The Scots-Irish Influence on American English”. American Speech 59.310–331. Donner, M. 1991. “Adverb form in Middle English”. English Studies 72.1–11. Edwards, V. & B. Weltens. 1985. “Research on non-standard dialects of British English: Progress and prospects”. Focus on: England and Wales ed. by W. Viereck, 97–139. Amsterdam and Philadelphia: John Benjamins. Finlay, C. 1988. Syntactic Variation in the Speech of Belfast Schoolchildren. PhD Dissertation. University of Ulster at Jordanstown. Fischer, D. H. 1989. Albion's Seed: Four British Folkways in America. Volume 1. New York and Oxford: Oxford University Press. Godfrey, E. & S. Tagliamonte. 1999. “Another piece for the verbal -s story: Evidence from Devon in Southwest England”. Language Variation and Change 11:1.87–121. Görlach, M. 1991. “Colonial Lag? The alleged conservative character of American English and other ‘colonial’ varieties”. Englishes: Studies in
112
TAGLIAMONTE, SMITH & LAWRENCE
Varieties of English 1984–1988 ed. by M. Görlach, 90–107. Amsterdam and Philadelphia: John Benjamins. Gregg, R. J. 1985. The Scotch-Irish Dialect Boundary in the Province of Ulster. Ottawa: Canadian Federation for the Humanities. Haegeman, L. 1981. The Use of Will and the Expression of Futurity in Presentday British English. Part 1: Future Time Expression in a Corpus of Standard British English. D.Phil. Dissertation. University College, London. Henry, A. 1992. “Infinitives in a for-to dialect”. Natural Language and Linguistic Theory 10:2.279–301. ----------. 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. New York and Oxford: Oxford University Press. Hickey, R. 2005a. “Introduction”. Transported Dialects: The Legacy of Nonstandard Colonial English ed. by R. Hickey, 8–50. Cambridge: Cambridge University Press. ----------, ed. 2005b. Legacies of Colonial English: Studies in Transported Dialects. Cambridge: Cambridge University Press. Hock, H. H. 1986. Principles of Historical Linguistics. Amsterdam: Mouton de Gruyter. Ito, R. & S. A. Tagliamonte. 2003. “Well weird, right dodgy, really strange: Layering and recycling in English intensifiers”. Language in Society 32:2.257–279. Jespersen, O. H. 1961a. A Modern English Grammar on Historical Principles: Part II: Syntax. VII. London: George Allen & Unwin. ----------. 1961b. A Modern English Grammar on Historical Principles: Part VII: Syntax. London: George Allen & Unwin. Jones, M. & S. Tagliamonte. 2004. “From Somerset to Samaná: Pre-verbal did in the voyage of English”. Language Variation and Change 16:2.93–126. Kurath, H. 1928. “The origin of the dialectal differences in spoken American English”. Modern Philology 25.385–395. ----------. 1964. “British sources of selected features of American pronunciation: Problems and methods”. In Honour of Daniel Jones: Papers Contributed on the Occasion of His Eightieth Birthday 12 September 1961 ed. by D. Abercrombie, D. B. Fry, P. A. D. MacCarthy, N. C. Scott & J. L. M. Trim, 146–155. London: Longman. Labov, W. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. ----------. 1982. “Building on empirical foundations”. Perspectives on Historical Linguistics ed. by W. P. Lehmann & Y. Malkiel, 79–92. Amsterdam and Philadelphia: John Benjamins.
ENGLISH DIALECTS IN THE BRITISH ISLES
113
----------. 1985. “Intensity”. Meaning, Form and Use in Context: Linguistic Applications ed. by D. Schiffrin, 43–70. Washington, D.C.: Georgetown University Press. Landsman, N. C. 1985. Scotland and Its First American Colony 1683–1765. Princeton: Princeton University Press. Leyburn, J. 1962. The Scotch-Irish: A Social History. Chapel Hill: University of North Carolina Press. Macafee, C. 1983. Varieties of English Around the World: Glasgow. Amsterdam & Philadelphia: John Benjamins. Macaulay, R. K. S. 1991. Locating Dialect in Discourse: The Language of Honest Men and Bonnie Lasses in Ayr. New York and Oxford: Oxford University Press. Martin, D. & S. Tagliamonte. 1999. “Oh, it beautiful” Copula variability in Britain. Paper presented at NWAV (New Ways of Analyzing Variation) Conference. Toronto, Canada. October 1999. McDavid, R. I. 1985. “Dialect areas of the Atlantic seaboard”. American Speech: 1600 to the Present ed. by P. Benes, 15–26. Boston: Boston University. Mencken, H. L. 1961. The American Language: Supplement II. New York: Alfred A. Knopf. Miller, J. 1993. “The grammar of Scottish English”. Milroy & Milroy 1993. 99–138. New York: Longman. Milroy, J. & L. Milroy. 1993. Real English: The Grammar of English Dialects in the British Isles. New York: Longman. Montgomery, M. B. 1989. “Exploring the roots of Appalachian English”. English World-Wide 10:2.227–278. ----------. 1994. “The evolution of verbal concord in Scots”. Studies in Scots and Gaelic: Proceedings of the 3rd International Conference on the Languages of Scotland ed. by A. Fenton & D. A. MacDonald, 81–95. Edinburgh: Canongate Academic Press. ----------. 1995. “The linguistic value of Ulster emigrant letters”. Ulster Folklife 41.1–16. ----------. 1997. “Making transatlantic connections between varieties of English”. Journal of English Linguistics 25:2.122–141. ----------. 2001. “The British and Irish antecedents”. Cambridge History of the English Language. Volume 6: English in North America ed. by J. Algeo, 86–153. Cambridge: Cambridge University Press. ----------. To appear. “Making trans-Atlantic connections for variable grammatical features”. Selected Papers from NWAV 29: Penn Working
114
TAGLIAMONTE, SMITH & LAWRENCE
Papers in Linguistics. Volume 7.3. Philadelphia: University of Pennsylvania. Montgomery, M. B. & C. Chapman. 1992. “The pace of change in Appalachian English”. History of Englishes: New Methods and Interpretations in Historical Linguistics ed. by M. Rissanen, O. Ihalainen, T. Nevalainen & I. Taavitsainen, 624–639. Berlin: Mouton de Gruyter. Montgomery, M. B. & J. M. Fuller. 1996. “What was verbal -s in 19th-century African American English?” Focus on the USA ed. by E. W. Schneider, 211–230. Amsterdam and Philadelphia: John Benjamins. Mufwene, S. S. 1996. “The founder principle in creole genesis”. Diachronica 13:1.83–134. ----------. 2001. The Ecology of Language Evolution. Cambridge: Cambridge University Press. Murray, J. A. H. 1873. The Dialect of the Southern Counties of Scotland: Its Pronunciation, Grammar and Historical Relations. London: Philological Society. Murray, L. 1795/1968. English Grammar. Menston, England: Scolar Press. Nevalainen, T. 1994a. “Aspects of adverbial change in Early Modern English”. Studies in Early Modern English ed. by D. Kastovsky, 243–259. Berlin and New York: Mouton de Gruyter. ----------. 1994b. “Diachronic issues in English adverb derivation”. Creating and Using English Language Corpora: Papers from the Fourteenth International Conference on English Language on Computerized Corpora, Zürich 1993 ed. by U. Fries, G. Tottie & P. Schneider, 139– 147. Amsterdam and Atlanta: Rodopi. ----------. 1997. “The processes of adverb derivation in Late Middle and Early Modern English”. Grammaticalization at Work: Studies of Long Term Developments in English ed. by M. Rissanen, M. Kytö & K. Heikkonen, 145–189. Berlin and New York: Mouton de Gruyter. Opdahl, L. 2000a. LY or Zero Suffix? A Study in Variation of Dual-form Adverbs in Present-day English. Volume 1. Frankfurt am Main: Peter Lang. ----------. 2000b. LY or Zero Suffix? A Study in Variation of Dual-form Adverbs in Present-day English. Volume 2. Frankfurt am Main: Peter Lang. Poplack, S., ed. 2000. The English History of African American English. Malden: Blackwell Publishers. Poplack, S. & D. Sankoff. 1987. “The Philadelphia story in the Spanish Caribbean”. American Speech 62:4.291–314.
ENGLISH DIALECTS IN THE BRITISH ISLES
115
Poplack, S. & S. Tagliamonte. 1989. “There’s no tense like the present: Verbal -s inflection in Early Black English”. Language Variation and Change 1:1.47–84. Poplack, S. & S. Tagliamonte. 1991. “African American English in the diaspora: The case of old-line Nova Scotians”. Language Variation and Change 3:3.301–339. Poplack, S. & S. Tagliamonte. 2000. “The grammaticization of going to in (African American) English”. Language Variation and Change 11:3.315–342. Poplack, S. & S. Tagliamonte. 2001. African American English in the Diaspora: Tense and Aspect. Malden: Blackwell Publishers. Poplack, S. & S. Tagliamonte. 2005. “Back to the present: Verbal -s in the (African American) diaspora”. Hickey 2005. 203–223. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik. 1985. A Comprehensive Grammar of the English Language. New York: Longman. Rickford, J. R. 1986. “Social contact and linguistic diffusion: Hiberno-English and New world black English”. Language 62:2.245–289. Rickford, J. R., A. Ball, R. Blake, R. Jackson, & N. Martin. 1991. “Rappin on the copula coffin: Theoretical and methodological issues in the analysis of copula variation in African American Vernacular English”. Language Variation and Change 3:1.103–132. Schibsbye, K. 1965. A Modern English Grammar. London: Oxford University Press. Schilling-Estes, N. 1998. “Investigating ‘self-conscious’ speech: The performance register in Ocracoke English”. Language in Society 27.53– 83. Schneider, E. W. 1981. Morphologische und syntaktische Variablen im amerikanischen Early Black English. Frankfurt am Main and Bern: Peter Lang. ----------. 1983. “The diachronic development of the Black English perfective auxiliary phrase”. Journal of English Linguistics 16.55–64. ----------. 2005. “The English dialect heritage of the southern United States”. Hickey 2005. 262–309. Smith, J. 2000. Synchrony and Diachrony in the Evolution of English: Evidence from Scotland. D.Phil. dissertation. University of York. ----------. 2001. “Negative concord in the old and new world: Evidence from Scotland”. Language Variation and Change 13:2.109–134. Smith, J. & S. Tagliamonte. In preparation. Agreement Patterns Prevails: The Legacy of -s in Northern Britain. Manuscript. University of York; University of Toronto.
116
TAGLIAMONTE, SMITH & LAWRENCE
Swan, M. 1980. Practical English Usage. Oxford: Oxford University Press. Tagliamonte, S. A. 2001. “Comparative sociolinguistics”. The Handbook of Language Variation and Change ed. by J. Chambers, P. Trudgill & N. Schilling-Estes. Malden and Oxford: Blackwell Publishers. ----------. 2003. “‘Every place has a different toll’: Determinants of grammatical variation in cross-variety perspective”. Determinants of Grammatical Variation in English ed. by G. Rohdenburg & B. Mondorf, 531–554. Berlin and New York: Mouton de Gruyter. Tagliamonte, S. A. & R. Ito. 2002. “Think really different: Continuity and specialization in the English adverbs”. Journal of Sociolinguistics 6:2.236–266. Tagliamonte, S. & J. Smith. 2000. “Old was; new ecology: Viewing English through the sociolinguistic filter”. The English History of African American English ed. by S. Poplack, 141–171. Oxford and Malden: Blackwell Publishers. Tagliamonte, S. A. & J. Smith. 2002. “Either it isn’t or it’s not: NEG/AUX contraction in British dialects”. English Word Wide 23:2.251–281. Trudgill, P., ed. 1978. Sociolinguistic Patterns in British English. London: Edward Arnold. ----------. 1986. Dialects in Contact. Oxford: Blackwell Publishers. ----------. 1990. The Dialects of England. Oxford: Blackwell Publishers. ----------. 1999. “New-dialect formation and dedialectalisation: Embryonic and vestigial variants”. Journal of English Linguistics 27:4.319–327. Trudgill, P., E. Gordon, G. Lewis and M. Maclagan. 2000. “Determinism in new-dialect formation and the genesis of New Zealand English”. Journal of Linguistics 36.299–318. Warner, A. 1982. Complementation in Middle English and the Methdology of Historical Syntax. London and Canberra: Croom Helm. Wolfram, W. 2000. “Issues in reconstructing earlier African-American English”. World Englishes 19:1.39–58. Wolfram, W. & E. Thomas. 2002. The Development of African American English: Evidence from an Isolated Community. Malden: Blackwell Publishers. Wolfram, W., E. Thomas & E. Green. 1997. Reconsidering the Development of AAVE: Insights from Isolated African American Speakers. Paper presented at NWAVE 26. Québec City, Canada. Wolfram, W., E. R. Thomas & E. Green. 1998. The Regional Context of Earlier African-American Speech: Evidence for Reconstructing the Development of AAVE. Manuscript. North Carolina State University. August 1998.
ENGLISH DIALECTS IN THE BRITISH ISLES
117
Wood, P. H. 1989. “The changing population of the Colonial South: An overview by race and region, 1685–1790”. Powhatan’s Mantle: Indians of the Colonial Southeast ed. by P. H. Wood, G. A. Waselkov & T. M. Hatley, 25–103. Lincoln and London: University of Nebraska Press. Yaeger-Dror, M., L. Hall-Lew & S. Deckert. 2002. “It’s not or isn’t it? Using large corpora to determinate the influence on contraction strategies”. Language Variation and Change 14:1.79–118.
PART II Dialects across social and regional borders
DIALECTS ACROSS INTERNAL FRONTIERS SOME COGNITIVE BOUNDARIES
DENNIS R. PRESTON Michigan State University 1.
Introduction Over the last several years a number of previously compartmentalized activities have been coming together, or are at least sharing some secrets with one another. First, dialectology, including its social as well as regional aspects, has become increasingly sensitive to acoustic analysis (e.g., Thomas 2001; Labov, Ash, & Boberg forthcoming). The use of such techniques has reached down even to individuals, so that variable speaker identity, related specifically to the differential use of repertoire alternatives, has also been investigated (e.g., Bucholtz 2001). Second, the social psychology of language (or language attitude research) has also been influenced by detailed phonetic studies. The ratings of different varieties, pioneered by Lambert (e.g., Tucker & Lambert 1969) and most fully exploited in the work of Giles (e.g., Ryan & Giles 1982), were studies in which the more or less influential elements of the stimulus signal were not determined. More recently, synthesized or carefully selected stimulus signals have allowed researchers to pinpoint the specific linguistic elements that trigger attitudinal responses, and a growing number of such studies have appeared in the literature (e.g., Milroy & Preston 1999). Finally, the discoursal turn in the social sciences has had an impact on both language variation and social psychological studies. This turn should not be confused with increased interest in the structure or organization of discourse itself (e.g., Schiffrin 1994). The influence I refer to here is the one that extracts meaning, attitude, or ideology from conversational evidence. It is essentially ostensive, i.e., it points to what a respondent has said and then engages the analyst’s skills in interpretation. Such approaches have been particularly exploited (along with other more quantitatively -oriented techniques) in the general area of folk linguistics, an area of enquiry that attempts to discover, analyze, and account for nonlinguists’ beliefs about language (e.g., Niedzielski & Preston 1999).
122
DENNIS R. PRESTON
Acoustic sophistication has appeared to help free dialectology from errors of perception and language attitude studies from those of production. Dialectologists are now more assured that what they hear is what was produced, and dialectology has always had a production bias, making such accuracy of primary importance. On the other hand, language attitude studies, in spite of the dominance of perception in that enterprise, have always had difficulty with production. How could this or that response to a regional voice be a pure response to region, one not cluttered by other factors — speech rate, pitch, and the notorious vocal quality, among others? New acoustic techniques allow attitude studies to go beyond the matched guise approach, which many believe was inappropriately used in studies of multiple dialects when only one speaker provided all the stimuli (e.g., Garrett, Coupland, & Williams 2003: 59). Dialectologists, however, have not been uniquely concerned with production. For example, the evaluation and/or identification and the comprehension of variety differences have played an important role in more recent work (e.g., Graff, Labov, & Harris 1986; Labov & Ash 1997). Similarly, the development of accommodation theory in the social psychological tradition has emphasized the production repercussions of perception (i.e., that hearers adjust their variety towards that of their interlocutors, e.g., Giles & Powesland 1975). We have now reached a point where a sociolinguistic-cum-social psychological approach to speech production and perception is being articulated, perhaps under the label of “sociophonetics,” although that label might for some include much more or much less than I intend to discuss here. By sociophonetics, I mean to cover at least the following territory: 1. Studies of the correlation between pronunciation and the demographic, personal, social identity, and interactional facts about speakers. 2. Studies of the perceptual abilities of hearers, not only to comprehend varieties in general and to discriminate among specific elements within them but also to identify regional and social varieties. 3. Studies of the entire range of attitudes and ideologies factors that influence both the correlations suggested in 1) and the abilities suggested in 2). One might argue, and I certainly would, that a meaningful folk linguistic (or attitudinal and ideological) component could be added to any linguistic enterprise, not just that of sociophonetics. I hope it is clear that the enterprise I outline here, including 3) just as much as 1) and 2) above, has as one of its goals the account of language variation and change in speech communities. In
DIALECTS ACROSS INTERNAL FRONTIERS
123
short, one of the goals of 3) is to shed further light on why certain choices are made in 1) and why some perceptions in 2) are promoted while others are sidetracked (e.g., Preston 1999: xxiv; Niedzielski & Preston 1999). In previous justifications of folk linguistics, perhaps the emphasis on the necessity of folk linguistics to a complete ethnographic record of a speech community and on the usefulness of such findings to applied linguists has overshadowed the importance of the role of language ideologies and folk beliefs as important considerations in the general study of language variation and change. I hope here to show that dialect boundaries, in the very broad social as well as geographical sense, are products of much more than Bloomfieldian speech-similarity networks (1933) or the long-term accommodation principles suggested in Trudgill (1986). I will review recent work on the influence of ideologies and attitudes on language perception, and, therefore, on the development of new production standards. It is only an accident that sociophonetics is the key element here, for, in one sense, sociophonetics is simply a sub-branch of sociolinguistics in general. One might also investigate recent findings in morphology, syntax, semantics, pragmatics, and conversation or discourse analysis, but the availability of multiple tokens for investigation in the phonological world has always made this level the darling of sociolinguists. I will survey several examples of such sociophonetic work carried out by me, my colleagues, and students at Michigan State University over the last decade or so. I do so not to ignore or belittle other important contributions; I simply know these best. 2.
Production I begin in the familiar territory of the sociolinguistic influence on production, a well-established area, at least since Labov’s seminal work on Martha’s Vineyard, in which island-oriented residents were shown to prefer a more local variety of the /aw/ and /ay/ diphthongs. 2.1
Local loyalty The well-known Northern Cities Chain Shift (NCCS) is a vowel rotation in the urban north of the United States that has come later to rural areas and minorities (if at all). Figure 1 shows 1) the position of these vowels in so-called General American English (the base of each arrow, and, as acoustic research on recordings of older respondents has shown, the likely jumping off place for many Michigan speakers before the influence of the NCCS) and 2) the direction and eventual site after the NCCS rotation (the point of each arrow).
124
DENNIS R. PRESTON
Figure 1: The Northern Cities Chain Shift (see Labov 1994).
In several investigations of the acquisition of this shift, particularly the tensing or raising and fronting of (æ), step #1 in Figure 1, the role of ideology has been shown to be important. In some cases it outweighs the demographic factors of class, age, sex, and network; in others it supplements one or more of them. Ito (1999) and Ito and Preston (1998) have looked at the difference in the acquisition of NCCS (æ) in rural mid-Michigan. Three middle class women (in their 40s and 50s) and their high school age children are reported on here. The respondents are all European-Americans who were brought up and have lived in the area most of their lives. Data collection included a recorded informal interview and the reading of a word list. The means scores of F1 and F2 (determined by LPC extraction of formants) for all tokens of each vowel are shown in Figures 2–7. All these respondents are participating in the NCCS, although the degree of participation differs considerably from one to another. The low-front vowel /æ/ has begun the typical NCCS upward and forward movement for all six speakers, but the remaining four vowels (of the five more which could be involved) appear to have been little affected by the NCCS. The differences in advancement, however, are strongly correlated to individual and social identity.
DIALECTS ACROSS INTERNAL FRONTIERS
125
Figures 2–7: The mean scores of F1 and F2 for all tokens of each vowel of six rural midMichigan informants.
Mary’s vowel system (Figure 3) shows an /æ/ that is raised and fronted: her mean score value for it is much higher than that for her // and very close to //. Cathy’s system (Figure 2) is more conservative, despite the fact that there is only a six-year age difference between her and Mary. Her /æ/ is fronted, but the degree of raising is minimal, with a mean score lower than that for her //. Candy’s system (Figure 4) is more similar to Mary’s than to Cathy’s; her /æ/ is fronted and raised to a position much higher than //. In the older generation, then, the first step of the NCCS (/æ/ fronting and raising) is in progress, and, not surprisingly, Cathy, the oldest, has the most conservative system. Ron (Figure 5) is Mary’s sixteen-year old son. The comparison between his system and his mother’s shows that his /æ/ is raised about as much as hers,
126
DENNIS R. PRESTON
but it is not unusual for a male speaker to be no more or even less advanced than his mother in on-going phonological change. Tammy (Figure 6) is Cathy’s seventeen year-old daughter. Her /æ/ is higher than her // although lower than her //; she is more advanced than her mother, Cathy, and than Ron. The last vowel system is Sherry’s (Figure 7), Candy’s seventeen year-old daughter. Hers is clearly the most advanced of the six, with an /æ/ extremely fronted and raised. The second step of the shift involves the fronting of //. If // has not fronted, then it should be much further from // than it is from // on the F2 dimension. Since // is expected to move toward the empty position vacated by /æ/ in the NCCS, however, the more it is fronted, the closer it will be to //. Using this criterion, Ron’s // appears not to have moved; it is considerably back of //, much closer to his //, and the same relation can be observed in Cathy’s and Candy’s systems; however, // fronting seems to be in progress in Mary’s system since her // is equidistant between // and //. // fronting is even more advanced in Tammy’s system, since her // is much closer to // than it is to //. Finally, Sherry’s // is also fronted even though it is closer to // than //, but this may be due to the fact that Sherry appears to be involved in the third step of the chain shift, i.e., // lowering and fronting. As far as // fronting is concerned, only three people seem to be participating — Mary, Tammy, and Sherry, although Mary seems to be least advanced. The third step of the NCCS is the lowering and fronting of //, toward the position of //. The degree of this shift also can be assessed by comparing the relative positions of //, //, and //. If // has been lowered and fronted, then it should be moving to the height of // and closer to //. This is exactly what is found in Sherry’s system. Her // and // are very close, and their height is about the same. Additionally, since Sherry’s // is only slightly higher than her // and //, one might even claim that the fourth step, i.e., lowering and/or backing of //, has started in her system. These data clearly show that Northern, non-urban Michiganders are participating in the NCCS. The general results are as follows: 1. Everyone seems to be involved in the first stage, /æ/ raising, although Ron and Cathy’s participation is minimal. 2. In the second stage, // fronting, Sherry, Tammy, and Mary seem to be involved, but Mary’s participation is less obvious than the other two. 3. Only Sherry seems to be participating in the third stage, // lowering and fronting. 4. Only Sherry shows any sign of // lowering and/or backing, an advanced step in the NCCS not focused on in this study.
DIALECTS ACROSS INTERNAL FRONTIERS
127
These results may be correlated with several characteristics of linguistic change common to many variationist studies, which show that the leaders of the NCCS are likely to be young, well-educated, upper working or lower middle class, European-American women (e.g., Labov 1994). Since all these respondents are well-educated middle class European-Americans, apparently only age and sex remain as relevant variables. The oldest respondent, Cathy, is the least influenced by the NCCS among the six; her vowel system is dramatically different from the two other mothers, who are in their forties. In addition, the vowel systems of the teenage children are more advanced than those of their mothers, except for Ron, the only male studied here. In addition to these predictable patterns of age and sex, the NCCS is also characterized as an urban phenomenon. Labov (1994: 178) notes that a modified wave model accounts for less advanced NCCS change in smaller places, such as Chili, New York (outside Rochester). Thus, the model predicts for Michigan that 1) the farther north, and 2) the smaller the place, the less advanced the change. However, these hypotheses were not confirmed in this study. Distance from major cities did not uniformly have the expected influence on the data. The most advanced speaker among the six was Sherry, who lives in the smallest and the most northern place. How can this apparently deviant result be explained? During the interviews it was revealed that both Tammy and Sherry are involved in various kinds of extra-curricular activities and both appeared to be very popular at school. In general, however, Tammy appeared to enjoy living in the country. The following excerpts illustrate her loyalty to and preference for local, rural or small-town culture. For example, when Tammy talks about hobbies, she notes that TV is “the big thing” but that during the summer, she likes outdoor activities such as biking (“R” is the interviewer): R: T: R: T: R: T: R:
So y- you sounds y- yo- you really like an outdoor - activities. Uh huh. ((pause)) I don’t really like the winter as much as the summer though, but, But at least you can do skii:ng and snowmobile, and stuff. [ [ Uh huh. That’s why I like Michigan. It’s got a variety. That’s true. So - u- u- when uh the summer comes, do you usually go: uh wh- what kind of thing do you usually do. In the summertime, hm. Ummm [ Go to the- the lakes or
DENNIS R. PRESTON
128
T:
[ Uh huh. I go to the beaches, we go camping: ... Me and Jessie - used to ride our bikes everywhere. ... Just as long as we’re outside, we have a good time.
Although many rural and small-town Michigan young people elect to go to the larger more urban institutions of higher education in the state (e.g., The University of Michigan in Ann Arbor or Michigan State University in East Lansing), Tammy notes that she is going to attend Central Michigan University (in Mt. Pleasant, where she lives and where her mother works as a secretary) to become a local kindergarten teacher. Her loyalty to the local culture is most distinctively expressed when she was asked the ideal place to live and advantages and disadvantages to living in that area. R: T:
R:
T:
R: T: R:
T:
Have you ever wished to live somewhere ELSE? Um:. I’ve thought about it, but - I don’t really want to go anywhere else, cause if - if I go to Florida or somewhere, it’s summer all the time. Which I don’t mind, but but I miss the - Huh [ [ (It’s too) humid. Too humid over there. [ Yeah, and I’d miss the winter and if I were to go somewhere where its snow (-y, -ing) all the time I’d miss the summer. And I like the fall too and it’s [ Yeah It’s beautiful. [ Uh huh. And that’s why I like Michigan. I don’t want to - go anywhere. I see. Uh. So::. Well you mentioned a little bit, but what is the best thing or the advantage or the worst thing or disadvantage to: live somewhere around here? To living somewhere around here? (pause) Um::. I don’t know as there’s disadvantages. Some of the advantages - I look at is - iwhen I- ever I have a family I would rather be up here because - I would be less apt to run into the crime and the gangs and all that stuff. And if you’re out in=
DIALECTS ACROSS INTERNAL FRONTIERS
R: T:
R:
129
[ Uh huh. =the country it doesn’t happen as often, and - then I wouldn’t have to be= [ (whispered) That’s true. =so worried about when my kids go out at night and things like that. So I’d rather live here. Where I know my territory more or less.
Her preference to remain in her home area is direct: “I don’t want to go anywhere.” The repetition of this comment and “That’s why I like Michigan” demonstrate her attachment to the local area. There are no “disadvantages” to living there as far as she is concerned. She prefers to live in small places because “there are no gangs and crimes.” Sherry, on the other hand, displays very different interests. She likes traveling because she likes seeing new places and wants to move to the West coast. R: S: R: S:
Do you like to travel around? Yeah, I LOVE to travel. Yeah. Oh yeah, you just mentioned um going to uh Mexico. Yeah. That’s probably one of my favorite things to do aside from running. We’ve- every summer my family and I take a vacation. And we’ve- my favorite has been California. We’ve went there a couple of times. And I hope to live out there when I get out of college. But yeah, I love to travel. I hope to ... travel around the world, by the time I’m, done with life. I- IR&S: (laugh) S: Yeah, I’m like excited. I just- ... Yeah.
In contrast to her preference for the West coast, Sherry does not have any attachment to the place where she grew up; she simply wants to get out of there. R: S:
Have you ever wished to live somewhere else— Oh: I wouldn’t mind living on the la:ke. Higgins Lake, it’s like real close to here, and I usually waitress at restaurants out by the lake. So, ... I’ve often thought that it would be neat to live on the lake,
DENNIS R. PRESTON
130
but ... I’d just as soon get out of Roscommon. So it doesn’t really, I haven’t really been (laughs) something I’ve given much thought to. Her preference for life elsewhere can also be observed in the following: R: S:
I see. Um. So, what is the advantage to live ... here? Well, I don’t know. Living in California I just- I liked the climate, I liked the place, it’s something new. It’s just seems to suit to my personality. I like seeing new places. So, ... I guess it would be more ... close to things, than — I don’t know, it just clicks with me, so — R: So what would be the ... worst thing or the disadvantage living here? S: Living here? Ah- well, being far from everything. R&S: (laugh) S: Living kind of ... sheltered away from, uh —
Sherry was easily able to list the things she likes in California (“climate,” “the place,” and “something new”), and she feels that it suits her personality. Contrary to this, one phrase was enough for her to describe her town: “far from everything.” This discoursal evidence confirms that Sherry is not interested in local, small-town or rural life, and the same preference was mentioned by her mother, Candy. She commented that she feels “isolated.” Other respondents, such as Cathy and Ron, indicated, however, that they enjoy their small-town or country life. Cathy is planning to remain in Mt. Pleasant after her retirement. Ron likes small towns because they are “quiet,” and “everyone knows everyone.” Cathy is the oldest and most loyal to the local area, and she participates least in the NCCS. Ron is male and similarly locally oriented, and he is the next strongest non-participant, despite his youth. Mary and Candy are older respondents, but they are somewhat less oriented to their home area and are more advanced in the NCCS than Cathy and Ron. Of the two teenage girls, Sherry is more advanced than Tammy. This crucial difference is accounted for by their difference in orientation: Sherry is much less loyal to the local area than Tammy, and her advanced position in the NCCS symbolizes her wish to dissociate herself from the local rural environment. Individual positions in the NCCS strongly correlate not only with such traditional independent variables as age, sex, size of city, and distance from major metropolitan area but also with individual orientation, i.e., one’s preference for the local small town or rural lifestyle.
DIALECTS ACROSS INTERNAL FRONTIERS
131
Including personal orientation as an important variable is not ad hoc, at least not in the history of the field. The finding in this study, that an unconscious choice of vowels reflects the identity or orientation of the individual, is similar to many previous studies. Perhaps more importantly for general sociolinguistic theory, since the two young women most advanced in the NCCS (Tammy and Sherry) are popular in their schools and well-integrated into their local communities, one might focus on their roles as current (and potential) leaders of local linguistic change rather than on Sherry’s attraction to non-local norms. They may be precisely the sorts of respondents whose open networks make them not only most susceptible to change but also more likely to be capable of spreading it among their peers (e.g., Milroy 1992). 2.2
Ethnicity In work on the acquisition of the NCCS, we have also focused on local African-American (in Lansing, MI) and South Midland immigrant (in Ypsilanti, MI) populations (Evans, 2001; Evans, Ito, Jones & Preston, 2000; Evans, Ito, Jones & Preston, to appear; Jones 2003; Preston & Jones, in progress). I survey here the unusual finding of how network score was not a predictor for one of these groups. Age
Sex
Status
Network correlation to F1/F2 .02 .01 n.s.
Mean Index Score
F2 Mid-Michigan X Female Working 2.64 Appalachian Young Female Middle 1.82 African-American X X X 2.33 F1 Mid-Michigan X X X n.s. 2.14 Appalachian Young Female Middle .01 1.32 African-American Young Female Middle n.s. 1.52 Table 1: F1, F2 and various demographic characteristics among three Michigan groups for the raising and fronting of (æ). (NB: “X” indicates no significant difference among the groups in that category.)
Look first in Table 1 at the Column labeled “Mean Index Score.” The higher the score (on a 5-point scale), the more advanced (æ) is in fronting (F2) and raising (F1). The Mid-Michigan speakers (of the sorts just discussed above) are clearly the most advanced, but I focus here on the Appalachian and African-American groups. We also determined for each speaker a network score based on both a neighborhood evaluation (e.g., Milroy 1980) as well as a friends and associates preference scale. One would expect a correlation
132
DENNIS R. PRESTON
between a low network score, in which a respondent group reports fewer associations with friends, co-workers, and family in a local area and fewer associations with persons ethnically similar, and a higher “Mean Index Score” if a less locally-oriented group is more likely to acquire the more general norm of the wider speech community. As Table 1 shows, in “Network correlation to F1/F2,” however, that does not occur in every case. Network is not correlated to African-American respondents’ F2 and F1 scores, but age, gender, and status are important to F1 (raising) in the African-American speech community. This is all the more puzzling when we note that the Appalachian F1 and F2 scores, important (although not always statistically significant) in every demographic category for both fronting and raising, are correlated with network position. The Appalachian pattern of change shows the middle rather than working class in the lead, with women and younger speakers also out in front. These immigrants live in urban, southeastern Michigan, the area where the NCCS is most advanced. It appears to them, therefore, to be the local norm, and their age, class, gender, and network scores reflect the rather rapid adoption of that prestige form, a case of change involving a “marker” rather than an “indicator.” Labov notes the following concerning such a variable at the stage the Appalachian group has begun to acquire it. As the sound change with its associated values reached the limits of its expansion, the linguistic variable became one of the norms of the speech community, and all members of the speech community reacted in a uniform manner to its use (without necessarily being aware of it). The variable is now a “marker” and begins to show stylistic variation. (1972: 179)
In spite of the fact that this feature of the NCCS is different from the traditional Appalachian pattern the older speakers in this group would have brought with them, there is no indication that they are aware of its emerging status in the local area, and it is not, therefore, a case of “change from above” nor a candidate for a “stereotype.” Putting aside the lack of any change in F2 for the African-American group (since fronting is perhaps an original feature of their dialect), they would appear to parallel the Appalachian group, since young, female, and middle class speakers lead in raising (F1). For Lansing African-Americans, therefore, the pattern of acquisition of raising also indicates the adoption of a local, prestige norm, but the lack of any network correlation is in need of further explanation. If a lack of Appalachian network density predicts participation in the NCCS for Appalachians, why doesn’t it for African-Americans? The answer almost certainly lies in the racial difference between the two groups. The Appalachian speakers, particularly those in the locally-raised
DIALECTS ACROSS INTERNAL FRONTIERS
133
second generation, are allowed full participation in the new majority speech community; there is no visual barrier to that participation, and they may fully integrate; the less dense network scores of those who make that adjustment make that clear. African-Americans, however, who have encountered racial barriers to full participation in the larger society, have apparently chosen to adopt parts of the NCCS (and other norms of the wider speech community) for instrumental rather than integrative purposes. That is, younger, female, middleclass speakers in particular have found a need to copy the norms of the majority speech community for educational and economic advancement. Those who have made that choice, however, apparently have not done so on the basis of weaker network ties to their home speech community. Experience with these (and many other African-American respondents) suggests that, unlike their Appalachian peers, they are not linguistically assimilating to the majority, surrounding speech norm. Instead, they borrow parts of it as they are seen necessary to social advancement and perhaps to a regionally local (but not white) identity. This borrowing is made a part of a larger linguistic repertoire that contains an ability to perform within the narrower, network dense community. In short, even the younger, female, middle class group of AfricanAmerican speakers can code-switch, and, although that identity in itself may have some future influence over their ability to control their more vernacular system, their continuing identity as African-Americans (as regards language as well as other cultural facts) seems intact, a claim I would not like to make for the younger Appalachian respondents. In summary, the NCCS is establishing itself very slowly in mid-Michigan nonurban areas. It is being led by working-class groups and women with unusually low-density network relations, although they come from a background which would have supported such networks only one generation past. Although they are acquiring this new system very slowly, I do not believe there is a strong caricature of the NCCS system (an anti-urban mentality) as might have been suggested in Ito and Preston (1998); instead, there may be considerable variation in local loyalty and urban favor or disfavor. For example, although young women often focus on the fact that urban areas are more fashionable, they also note how they are bad places to raise families due to crime, drugs, and other big-city woes. Young men, although they may have been displaced from local agricultural work, are still very much a part of the local hunting or outdoors culture (whether they are participants or not) and deride the urban weekend hunters, who are fancy dressers, don’t spend enough time in the woods, and shoot small deer (Ito 1999). Appalachians in Ypsilanti and African-Americans in Lansing are both originally speakers of stigmatized dialects. The Appalachian group is adopting
134
DENNIS R. PRESTON
local pronunciation and, due to racial similarity with the surrounding majority, the degree to which that adoption is taking place is reflected in lower density network scores with the home Appalachian population. African-Americans, on the other hand, have, at least in certain groups, found it important to acquire some aspects of the surrounding majority speech norm, but they do not do it in connection with a loss of network density. Their acquisition of the NCCS reflects an instrumental rather than integrative motivation, and they retain ability in the home variety, which allows them to maintain local network strength. 2.3
Imitation Next, I provide evidence from a study by Evans (2002) that finely tuned characteristics of pronunciation are, in fact, replicable by nonnatives. Linguists have generally assumed that it is not possible for a speaker to modify speech systematically (e.g., Labov 1972: 215): Although one can achieve a certain amount of insight working with bilingual informants, it is doubtful if as much can be said for bidialectal informants, if indeed such speakers exist. We have not encountered any nonstandard speakers who gained good control of a standard language, and still retained control of the nonstandard vernacular. Dialect differences depend upon low-level rules which appear as minor adjustments and extensions of contextual conditions, etc. It appears that such conditions inevitably interact, and although the speaker may indeed appear to be speaking the vernacular, close examination of his speech shows that his grammar has been heavily influenced by the standard. He may succeed in convincing his listeners that he is speaking the vernacular, but this impression seems to depend upon a number of unsystematic and heavily marked signals.
More recent work suggests the same: “Ash...did a test survey of individuals who were asked to disguise their voices over the telephone. The subjects modified tempo, voice quality, and intonation, but none modified the segmental features specific to their geographical dialect” (Labov 1994: 111). Imitation would seem to be ineffective even when it is long-term and unconscious. When children with even one nonnative parent in the Philadelphia area were tested for their acquisition of the local complex short-a raising rule, they were found to be lacking (Payne 1980). The failure of imitation in second and foreign language learning (especially for pronunciation by adults) is well known. There is, however, recent contrary evidence. In Schilling-Estes’ (1998) study of the Okracoke “brogue” (the traditional speech of Okracoke Island, NC), she found regular patterning in the performance and normal speech of
DIALECTS ACROSS INTERNAL FRONTIERS
135
several respondents. She examined the first and second formants of the diphthong /ay/ in these two styles and found systematic patterning in various phonological environments in relation to both the height and backness of /ay/. Her findings “suggest that the patterns of linguistic variation observed in selfconscious speech are not necessarily different from, or less regular than, those observed in non-self-conscious speech” (1998: 64). Are imitations the use of only a few stereotypical features, either easily adapted from lexical or morphological caricatures, or, if at a phonological level, are they simply inaccurate (at least at the level of acoustic realization and/or detailed contextual specification)? In this study, imitation refers to the conscious use of a variety not the speaker’s vernacular. This study samples only a single respondent, “Noah,” a 29-year-old male born in Morgantown, West Virginia who lived there until age 23. His parents are natives of Detroit, speakers of northern US English. Because the respondent has had adequate input of both relatively more southern (native Morgantown) and nonsouthern (or northern) speech, there is good reason to believe that he could have control over both varieties. However, his usual vernacular reflects a conservative northern system, which is expected, given the stigmatized status of southern speech in Morgantown, the home of West Virginia University. The question addressed here is how accurate his command of those elements of southern speech that do not appear in his usual speech is. The respondent read a word list and a reading passage with items deliberately selected to reflect differences between northern and southern speech. Later, he was asked to re-read the word list and reading passage in the West Virginia style of the people from his hometown. Noah’s performances were compared first to a conservative northern vowel system, presumably not unlike his parents’ (and not unlike one he would have learned as a university-oriented child in Morgantown, in spite of the local variety). His Detroit parents might even have shown some tendency towards the Northern Cities Chain Shift (Figure 1). Noah’s vowels were then compared to the configuration of the so-called “Southern Vowel Shift” (Figure 8). If the vowels in both his normal and imitation systems correspond to the conservative system, there is little low-level skill in imitation (apparently the claim in most earlier research). If, however, his normal performance conforms to the conservative system and the imitation system conforms (at least in part) to the Southern Vowel Shift system in acoustic detail, his imitative or performance ability is rather more precise than expected.
136
DENNIS R. PRESTON
Figure 8: The Southern Vowel Shift (Labov, Ash, and Boberg 1997).
A comparison of various subsystems of Noah’s normal and imitation speech reveals systematic differences between them. Figures 9 and 10 display the mean scores for the front vowels of Noah’s two systems. The highest and most front vowel in the normal system is /iy/; /ey/ is more front than //, and // is the lowest, most back of the high and mid-front vowels. This is the system we would expect in a conservative, northern US speaker. The imitation front system (Figure 10) is quite different. The highest, most front vowel in this system is //; /iy/ is lower than //, and /ey/ is the lowest most back high-front vowel. The // tokens are split into two groups. The nasalized tokens are quite high, but the others aren’t. This is almost certainly due to Noah’s imitation of the typical southern / merger before nasals; he has grouped pen with the // tokens (which, as indicated above, are raised), although that would not explain the position of neck. Nevertheless, even the non-nasalized /e/ tokens are higher than the /ey/ tokens. The positions of // and /ey/ and /e/ and /ey/ have flipflopped when compared to his normal system, exactly the characteristics one finds in the Southern Vowel Shift system seen in Figure 8. I will not detail the rest of Noah’s vowel system, but it shows fronting of nearly all back vowels, fronting and monophthongization of /ay/, and monophthongization of /w/, all southern characteristics. Briefly, Noah’s imitative ability was acoustically very accurate. To test the perceptual as well as acoustic accuracy of Noah’s imitations, a matched guise tape was created with Noah’s normal and West Virginian speech and five other male speakers; three from West Virginia, one from a
DIALECTS ACROSS INTERNAL FRONTIERS
Figure 9: Noah’s normal front vowel system.
Figure 10: Noah’s imitation front vowel system.
137
DENNIS R. PRESTON
138
South Midlands area, and one from the Inland North. Each speech sample was the same excerpt from a reading passage. 69 West Virginia University student respondents listened to the tape and were asked to indicate whether they believed each speaker was from West Virginia or not, using a five-point scale. The voices were played again so that they could indicate why they rated each speaker the way they did. Respondent differences of region, age, and sex were important for only one or two sample voices and so are conflated for all the results reported here. Mean scores for respondent’s ratings of speakers are based on the following five-point scale: 1 you are sure that the speaker grew up in West Virginia 2 you think the speaker grew up in West Virginia 3 you don’t know if the speaker grew up in West Virginia 4 you think that the speaker did not grow up in West Virginia 5 you are sure that the speaker did not grow up in West Virginia
4
3.5
3
2.5
2
1.5
1
0.5
0 Charleston
Imitation
Parkersburg
Noah's normal
South midlands
Keyser
Inland north
Figure 11: Ratings of 6 speakers and Noah’s imitation of the likelihood that the speaker is from West Virginia (1 = sure the speaker is from West Virginia; 5 = sure the speaker is not from West Virginia).
DIALECTS ACROSS INTERNAL FRONTIERS
139
Figure 11 shows the mean scores for all speakers. The mean score for the Charleston speaker (1.67) indicates that respondents felt surest that he grew up in West Virginia, but the mean score for Noah’s imitation (1.93) shows that the respondents also believed that the imitation was also the speech of a native West Virginian. The difference between the mean scores for the imitation and Charleston speaker are not significant (.08). Thus it seems that the respondents perceived the imitation just as likely to be from West Virginia as the Charleston speaker. The mean score even for the West Virginia speaker from Parkersburg (2.69) shows that respondents were unsure about his West Virginian status. Mean scores for the other speakers on the tape also show that the respondents weren’t sure or perceived those speakers to be from elsewhere, even the Keyser speaker, who is also from West Virginia, but from the eastern panhandle of the state. Comments from respondents who rated the imitation as native show that their judgments were based on general impressions (“southern twang,” “I know people who talk like this,” “sounds like Uncle Jesse”) as well as specific details (“‘u’ sound,” “sounds like its mixed with the ‘a’ in ‘cat’,” “the pronunciation of ‘hand and spilled’”). “Has a definite southern twang of a West Virginian” “I know people who talk like this from WV: accent, speed” “Man sounds like Uncle Jesse. Likely to be from this region. Yee haw.” “strong drawl, especially his ‘u’ sound” “A heavy southern dialect. The pronunciation of ‘hand & spilled’” “The way he says slices. His words are drawn out and his speech is not rapid” Data from those who rated the imitation a 3, 4, or 5 (14/69) reveal that many felt the imitation was from another southern region, but only two perceived it as inauthentic: “drawl is tricky-maybe more southern” “has a southern accent but I don’t recognize it”
140
DENNIS R. PRESTON
“Southern drawl maybe from Texas” “he has a very strong southern accent-too strong to be from WV” Given the acoustic accuracy of the imitation, it was not expected that many respondents would perceive it as false. The two respondents who did said: “sounds too much like a hick to be one” (native West Virginia respondent) “Tries to sound southern but does not keep tone consistent” (non-native West Virginia respondent) Finally, an acoustic analysis of the Charleston and Parkersburg speakers was conducted in order to compare the imitation system to other native West Virginia speakers (no acoustic analysis was conducted for the Keyser speaker as respondents overwhelmingly found him to not be from West Virginia). Mean scores of F1 and F2 for the Charleston speaker show some characteristics of the Southern Shift such as fronting of /uw/, /ow/ and /aw/ but no reversal of /iy/ and /i/ nor of /ey/ and /e/. Mean scores for F1 and F2 for the Parkersburg speaker also show some characteristics of the Southern Shift such as fronting of /uw/ and /ow/ but again no reversal of /iy/ and /i/ nor of /ey/ and /e/. Noah’s imitation demonstrates more features of the southern shift than the native speakers and can be said to be more ‘southern’ than the West Virginia natives. This could explain the difference in ratings for the imitation and the Charleston speaker. Twelve of the fourteen respondents who gave the imitation a 3, 4, or 5 felt that it was too southern to be West Virginian. Perhaps these respondents reacted to very minute details of the imitation (perhaps even the reversal of /iy/ and /i/ and /ey/ and /e/, the major difference between Noah’s imitation and the Charleston and Parkersburg speakers). Yet the majority of the respondents were convinced by Noah’s imitation, and it must be that their perception of West Virginia speech is that it is very southern, perhaps even more southern than it really is. This perception must have been the same one that Noah drew from when he did the imitation and perhaps lead him to create speech with perhaps too many southern features to be authentically local. Acoustic analysis of the vowel formants of Noah’s normal and imitation speech styles shows consistent patterning in both. The analysis of Noah’s normal and imitation systems, like Schilling-Estes’ Okracoke study of Rex, demonstrates that at least some speakers have considerable imitation ability.
DIALECTS ACROSS INTERNAL FRONTIERS
141
Perhaps the motivation for Noah’s ability was his childhood and adolescent awareness of some covert prestige value of the variety. Additional research must be conducted in order to more clearly understand both the levels of accuracy of such imitations and the social contexts and motivations for both acquisition and performance. This study should make it clear, however, that imitation speech is not just the ability to evidence a few gross caricatures or stereotypes, for at least some speakers have the ability to modify their speech, at fairly detailed levels, even when the modification is done fairly selfconsciously. It is also obvious that speakers adjust their speech not only in the direction of emerging norms and practices but also according to their own ideologies and identities when those factors can be correlated to speech practices, even when the adjustment is in the direction of stigmatized speech. I turn now, however, more directly to the question of perception, and I begin in the area of the possible dialect influence of both speaker and hearer on phoneme perception. 3. 3.1
Perception Region Plichta and Rakerd (2002) investigate the degree to which respondent dialects may influence phoneme perception. If respondents are involved in a shift themselves, they might have greater tolerance for a shifted vowel’s identity; if they are not, they might have a greater tendency to identify a shifted vowel as belonging to another phoneme class. The NCCS provides just such an opportunity with // fronting. As // moves forward into the territory vacated by /æ/, which has already fronted and raised (see Figure 1), participants in the shift may tolerate a more dramatically fronted vowel and yet retain an // interpretation. That is, NCCS affected speakers might continue to hear words such as hot and sock in positions which non-shifted speakers might identify as hat and sack. To test this hypothesis, Plichta and Rakerd first determined mean F2 values for // for a group of shifted speakers (southeastern lower Michigan, hereafter LM) and a group of nonshifted speakers (Upper Peninsula of Michigan, hereafter UP, a group even further from the newly established urban norm of the NCCS in southeastern Michigan than the mid-Michiganders discussed in 2.1 above). These differences were considerable; the mean F2 score for LM speakers was 1525 Hz while the UP speakers showed a conservative (backer) mean of 1159 Hz. They next used the HLsyn parametric synthesizer to produce 11 hVt and sVk tokens at 33hz intervals along the trajectory from 1210 to 1540, the lower arrow in Figure 12. (F0, F1, and F3 were kept relatively stable with the exception of some formant transition
142
DENNIS R. PRESTON
adjustments in F3 and a 120 to 100 Hz fall in F0, matching the UP and LM performances). The hVt tokens could be interpreted as “hot” or “hat,” and the sVk tokens could be interpreted as “sock” or “sack.” After a pilot study revealed no variation in identification of the vowels at the extremes of the continuum (everything back of 1243 was heard as // and everything front of 1441 was heard as /æ/), the continuum was reduced to a seven-step scale with those edges, the upper arrow in Figure 12.
Figure 12: 11-step (lower arrow) and 7-step (upper arrow) continua of synthesized // tokens (adapted from Plichta and Rakerd 2002).
Two young male speakers, one from the UP the other from LM, matched for age, vocal tract size, and voice quality then recorded phrases into which these synthesized words were placed. These carrier phrases offered no cue to the identity of the word (e.g., “The key to winning the game of Boggle, is to know lots of short words like _____” and “Bob was certain that he heard his wife Shannon say _____”). The respondents were five young men and women from Detroit (ages 19–30, all heavily influenced by the NCCS) and five men and four women (ages 19–34) from Ishpeming, MI (in the UP), none influenced by the NCCS. Each respondent first heard the single synthesized words (one hVt and one sVk) repeated four times in each of the seven positions on the F2 continuum for a total of 56 trials. Each respondent then heard each of the two synthesized words in four carrier phrases from the two different speakers (one LM, one MP), each repeated four times, again in each of the seven steps of the F2 continuum for a total of 448 trials. This sophisticated acoustic approach allows us to begin answering classic questions at the interface of sociolinguistics and speech perception. As long ago as 1957, Ladefoged and Broadbent, in conducting an experiment on phoneme perception, speculated that there was “...tentative evidence that
DIALECTS ACROSS INTERNAL FRONTIERS
143
subjects belonging to different sociolinguistic groups gave different responses to some of the test materials.” On the other hand, there is also evidence for uniformity in perception, based on vowel-intrinsic features (e.g., Miller 1989, Syrdal and Gopal 1986). Plichta and Rakerd’s results suggest that both these interpretations may be plausible, the first in the carrier-phrase context, the second in the singleword context. I examine the latter first. Figure 13 shows that there is very little difference in the UP and LM perceptions of these tokens as /æ/ or //, although the LM respondents continue to hear // more robustly at step 5 (only). Predictably, the backest (i.e., lowest F2) tokens (step 1) are nearly all recognized as // (i.e., as “hot” or “sock”), and the frontest tokens (step 7) are nearly all recognized as /æ/ (i.e., as “hat” or “sack”). It is worth noting, however, that unlike classic phoneme perception tests of consonants, in which gradient acoustic characteristics are modified, a sharp (or categorical) boundary does not exist between one phoneme and the other. The similarity of identifications regardless of dialect background of the hearer suggests that some intrinsic factor of vowel quality or perhaps a stereotype of typical vowels is the primary input into perception. 100 90 80 70 60 LM-SW UP-SW
50 40 30 20 10 0 7
6
5
4
3
2
1
Figure 13: UP and LM respondent classifications of single-word hVt and sVk tones along a seven-step continuum (shown as percentage of // identifications).
DENNIS R. PRESTON
144
100 90 80 70 60 LM-UPP UP-UPP
50 40 30 20 10 0 7
6
5
4
3
2
1
Figure 14: UP and LM respondent classifications of hVt and sVk tokens along a seven-step continuum in UP carrier phrases (shown as percentage of // identifications).
100 90 80 70 60 LM-LMP UP-LMP
50 40 30 20 10 0 7
6
5
4
3
2
1
Figure 15: UP and LM respondent classifications of hVt and sVk tokens along a seven-step continuum in LM carrier phrases (shown as percentage of // identifications).
DIALECTS ACROSS INTERNAL FRONTIERS
145
The submissions of the same sample with carrier phrases, however, show significantly different results. The respondents from the two dialect areas again are in substantial agreement when the carrier phrases are from the UP, the area not influenced by the NCCS (Figure 14), but when the words are embedded in LM (i.e., NCCS-influenced) carrier phrases (Figure 15), the LM respondents show a decided tendency to interpret much fronter tokens as //. Figure 15 shows no reduction at all from step 1 to step 3 for LM respondents (who show the characteristic drop in // for these same tokens in both single-word and UP carrier-phrase environments). Even more dramatically, by step 4, LM respondents still hear // nearly eighty percent of the time; at the same step 4 in UP carrier phrases, they hear // only fifty percent of the time. It is clear, therefore, that, whatever intrinsic characteristic vowels have, those features may be overcome by sociolinguistic identity of the hearer in connection with additional information about the system of the speaker. UP respondents are not familiar enough with the LM (i.e., NCCS-influenced) system for carrier phrases to have any influence on their recalibrating the phonological space of a speaker; they calculate phonemes on the basis of their own experiences, ones not significantly different from the evidence given by individual word tokens. When LM hearers are given carrier phrase evidence of a modified vowel system which is very familiar to them, they adjust their hearing to match the divergent phonetic facts of that system. Some may be disappointed, however, to see so much attention paid to regional differences alone in a sociolinguistic setting. After all, the LM and UP speakers are, at least as far as the NCCS is concerned, speakers of different varieties and have different linguistic systems, at least as regards the phonetic values of the similar phonemic systems. Isn’t it possible to find perceptual strategies which are dramatically influenced by such traditional sociolinguistic identities as age, gender, or status? I believe such different strategies can be detected, and I believe they run in the direction of the perception of the linguistic reality as well as in the direction of person or group caricature, the latter a fact established at least as long ago as the work of Frederic Williams and his colleagues, in which standard English voices were rated lower on various scales when respondents believed the speakers were African American or Mexican American (Williams 1976: 103–107). Gender In Plichta and Preston (2003), the southern US speech stereotype of monophthongized /ay/ is investigated to determine the following: 3.2
146
DENNIS R. PRESTON
a. What is the smallest modification in the monophthongization of /ay/ which can be perceived? b. Will gradient versions of monophthongized /ay/ be perceived gradiently or will there be some (or several) categorical break-points? c. Will speaker characteristics (age, sex, etc…) affect perception? d. Will hearer characteristics (age, sex, region, etc…) affect perception? There is no doubt that southern US English is the most salient dialect of the US. When I asked southeastern Michigan and southern Indiana respondents to draw boundaries around significantly different US speech areas (on a blank map), ninety-four percent of the Michigan respondents and eighty-six percent of the Indiana respondents identified an area which could be characterized as southern. The Michigan respondents identified their local area (a Great Lakes region) at only the sixty-one percent level, and Indiana respondents identified a northeastern US region focused on New York City at only the fifty-one percent level, the second most frequently outlined areas for each group, respectively (Preston 1996).
Figure 16: The fifty states, New York City and Washington, D.C rated for language correctness on a scale of 1 to 10 (1 =- least correct) by southeastern Michigan respondents (Preston 1996).
DIALECTS ACROSS INTERNAL FRONTIERS
147
It is also clear that southern distinctiveness is not related to purely linguistic differences. When I asked Michigan respondents to rate the fifty states and New York City and Washington, D.C. for language correctness, the results were as shown in Figure 16. Except for New Jersey and New York City, the other locus of perceived nonstandard English in the US, the south is the only region which consistently gets ratings in the 3.00 to 4.00 range. Ethnographic characterizations also reveal negative caricatures of southern speech. Figure 17 shows a hand-drawn map of US regional speech by a young, male Chicago resident. For him, the south is where there is “Southern talk the worst English in American [sic].” (His own Chicago speech, by the way, is “Normal talk for the average person.”) Speakers of US English are also aware that /ay/ monophthongization is a — perhaps the — principal phonetic caricature of southern speech. In American Tongues, a documentary dealing with variation in US English, a southerner tells a story in which he tolerates northerners’ imitations of his monophthongal pronunciation of “ice” until he exasperatedly calls his tormentors “ice holes.”
Figure 17: A young, male Chicago respondent’s hand-drawn map of US dialect areas (Preston 1996).
148
DENNIS R. PRESTON
Figure 18: Home sites of the nine voices played in the identification task (Preston 1996).
DIALECTS ACROSS INTERNAL FRONTIERS
149
In earlier research on the sensitivity to the north-south dimension in US English, I asked respondents to listen to nine voices reaching from Saginaw, Michigan in the north to Dothan, Alabama in the south (Figure 18). The speech samples from male, European American, middle age, middleclass speakers from these nine sites were devoid of all dialect-specific features except pronunciation. The respondents, from southernmost Indiana and southeastern Michigan, were asked to match these voices (played in scrambled order) with their sites. A simple account of their guesses was tallied; if respondents identified the voice from site #1 as being from site #1, a “1” was tallied for voice #1; if they identified the voice from site #3 as being from site #1, a “1” was counted for voice #3, and so on. The scores were averaged and arranged in order. The respondents did very well on the basis of means scores rankings. The Michigan respondents misplaced only the Dothan, AL voice, putting it between Bowling Green, KY (site #6) and Nashville, TN (site #7), rather than identifying it as the southernmost one (Figure 19).
Figure 19: Southern Indiana (left) and southeastern Michigan (right) voice identification placements (Preston 1996).
150
DENNIS R. PRESTON
Groupings of these scores by a statistical analysis (Wilcoxon) are shown in the circles around the sites in Figure 19. For the southern Indiana speakers, once south of the home site (New Albany, IN, site #5), southern areas are distinct, with the exception of their being no statistically significant difference between Florence, AL (site #8) and Dothan, AL (site #9). To the north, however, they link themselves and all other areas together in a chain of combinations — the home site is not different from Muncie, IN (site #4), but Muncie is also not different from the next two sites to the north, and the three northernmost sites are also indistinct. For the Michiganders, all areas are more distinct except for the short chain of northern pairs formed by linking Saginaw, MI (site #1) and Coldwater, MI (site #2) then Coldwater and South Bend, IN (site #3) and the southern linkage of Nashville, TN (site #7) and Florence AL (site #8). Although this test reveals that US respondents hear a continuum of northern and southern voices, we do not know, since each sample was about a 30-second stretch of speech, what phonetic characteristic(s) was/were most responsible for the relatively northern or southern selection of a site. Assured that /ay/ was one such clue (and that it occurred at least once in all the samples played), we proceeded to test increasingly monophthongal variants of /ay/ to test the questions a. – d. asked above. In a web-based experiment, we presented male and female seven-step resynthesized pronunciations of the word “guide” and asked respondents to indicate which of the nine sites in Figure 18 they thought each token was from. We determined the gender, age, profession, ethnicity, and region of the respondents, each of whom heard each resynthesized token of “guide” three times (42 tokens). The LPC analysis/resynthesis was done from samples in which target F1 (maximum change 150 Hz) and F2 (maximum change 550 Hz) values at each 20ms frame were calculated to obtain the seven-step continuum of monophthongization and fronting. Fronting of the onset was added since, at least in the part of the US studied here, fronting as well as reduction of the glide is common. Regional, professional (i.e., status), and ethnic distributions among the respondents were not large enough for statistical tests to be run, but both age and sex were calculated (61 female, 35 male; 19 under 20, 15 from 21 to 30, 23 from 31 to 40, 19 from 41 to 50, 16 from 51 to 60, and 4 over 60); however, neither sex nor age of the respondents showed any difference in the pattern of responses. Respondents were, however, very outspoken concerning the appropriateness of the task, most complaining that they could not hear the fine
DIALECTS ACROSS INTERNAL FRONTIERS
151
distinctions and/or that they did not have experience in hearing the speech of the nine sites. The following are typical responses: I don’t do surveys ordinarily. … Unfortunately, after doing the sample. [sic] I realized that I am clueless. My ignorance could only compromise your findings. Had I a printed version of the word before it was sounded, I might have had more of a snowball’s chance. I’ve just done the web linguistic survey and I kept going even though I was quite clueless. How should anyone other than travelling salesmen have any familiarity with these regional dialects? Who’s a "good respondent" as opposed to one like me who’s mostly guessed in a tripartite way: north, middle, south? I had one stop in Indianapolis once during which I was so struck by a sixteen year waitress’s accent that I’ve always remembered the difficulty of understanding her, but I sure haven’t remembered her voice. I hope your survey produces something of value, but I’m curious how you might get there if other respondents are as clueless as I am! In spite of such claims (similar to ones made even by linguists who have heard the seven-step continuum), the respondents did very well in placing the seven tokens along the nine-step north-south continuum (Figure 20). In fact, only steps one and two (the most diphthongal) were not significantly different from one another; an ANOVA with a Tukey post-hoc test shows every other step, whether for male or female voices, to be significantly different from the adjacent (and all other) values. The complaint by respondents, therefore, that they could not discriminate among these small differences in degree of monophthongization was realized only for steps 1 and 2. I conclude, therefore, that at least one element which guided respondents in placing the nine voices in the older test was degree of monophthongization. More importantly, although not discussed here at any great length, is the fact that minimal degrees of phonetic difference can be perceived in such tasks and that they are perceived gradiently, not categorically. What is most important here, however, is the fact that a sociolinguistic element, namely, sex of speaker, resulted in a significant difference, as determined by independent t-tests, in the ranking for each step between the male and female voice. As Figure 20 shows, the male voice was always rated
DENNIS R. PRESTON
152
Male & Female voice ratings 9 7.61
Mean area assignments
8
7.16 6.59
7 5.77
6 5 4 3
6.43 5.99
5.4
4.5 3.31 2.39
Female Male
4.01
3.69 3.25 2.65
2 1 0 Step 1
Step 2
Step 3
Step 4
Step 5
Step 6
Step 7
Monophthongization steps
Figure 20: Assignment of seven-step monophthongized male and female samples of “guide” to the nine sites of Figure 18 (Plichta and Preston 2003).
as more southern (i.e., assigned a site farther south among the nine choices given in Figure 18). Since the resynthesized signals offered no phonetic clue to this, I conclude that two well-established caricatures have come together to force this perception. First, as I have shown above and elsewhere (e.g., Preston 1996), southern speech in the US is not only salient to US English speakers but is salient because of its perceived nonstandardness. Second, perhaps beginning with Fischer (1958) but elaborated on in Trudgill (1972) and confirmed in considerable further sociolinguistic work, women are much more likely to be users of more standard or overtly prestigious forms than men. It follows that if female usage is generally more standard that both men and women perceive female speakers as more standard and males as less so and that respondents would be much less inclined to rate a female voice, regardless of the phonetic reality, as one from farther south, a region caricatured as nonstandard. Of course we hope that as we gather more data, we may be able to show that characteristics of the hearer (age, sex, region, ethnicity, etc…) may also be important to the perception of degree of monophthongization and placement on the north-south dimension. What we have shown already, however, is that the perceived gender of the speaker
DIALECTS ACROSS INTERNAL FRONTIERS
153
causes a radical alteration in the assignment of monophthongized steps to a regional site. 4.
Conclusion In conclusion, I hope to have shown that newer trends in sociophonetics are important to perceptual as well as production matters and that even production itself may be a more delicately tuned matter than some have previously assumed. Most importantly, however, I hope to have shown that, whatever the approach to variable data, ideological, attitudinal, or folk linguistic factors play a central role in the study of language variation and change. References Bloomfield, Leonard. 1933. Language. New York: Henry Holt. Bucholtz, Mary. 2001. “From ‘sex difference’ to gender variation in sociolinguistics”. A paper presented as part of Symposium III: Variation studies in context at NWAV 30. North Carolina State University. Evans, Betsy E. 2001. Dialect Contact and the Northern Cities Shift in Ypsilanti, Michigan. Doctoral dissertation. Michigan State University, East Lansing, MI. ----------. 2002. “An acoustic and perceptual analysis of imitation”. Handbook of Perceptual Dialectology, Volume 2 ed. by Daniel Long and Dennis R. Preston, 95–112. Amsterdam: John Benjamins. Evans, Betsy E., Rika Ito, Jamila Jones, and Dennis R. Preston. 2000. “Change on top of change: Social and regional accommodation to the Northern Cities Chain Shift”. De Toekomst van de Variatielinguitiek (a special issue of Taal en Tongval to honor Dr. Jo Daan on her ninetieth birthday) ed. by H. Bennis, H. Ryckeboer, & J. Stroop, 61–86. Evans, Betsy E., Rika Ito, Jamila Jones, and Dennis R. Preston. To appear. “Accommodation to the Northern Cities Chain Shift”. Language Variation and Change in the American Midland ed. by T. Murray and B. L. Simon. Amsterdam: John Benjamins. Fischer, John L. 1958. “Social influences on the choice of a linguistic variant”. Word 14.47–56. Garrett, Peter, Nikolas Coupland, and Angie Williams. 2003. Investigating Language Attitudes. Cardiff: University of Wales Press. Giles, Howard and Peter F. Powesland. 1975. Speech Style and Social Evaluation. London: Academic Press. Graff, David, William Labov, and Wendell A. Harris. 1986. “Testing listeners’ reactions to phonological markers of ethnic identity: A new method for
154
DENNIS R. PRESTON
sociolinguistic research”. Diversity and Diachrony (= Current Issues in Linguistic Theory, 53) ed. by David Sankoff, 45–58. Amsterdam: John Benjamins. Ito, Rika. 1999. Diffusion of Urban Sound Change in Rural Michigan. Doctoral dissertation. Michigan State University, East Lansing, MI. Ito, Rika and Dennis R. Preston. 1998. “Identity, discourse, and language variation”. Journal of Language and Social Psychology 17:4.465–483. Jones, Jamila. 2003. African Americans in Lansing and the Northern Cities Vowel Shift: Language Contact and Accommodation. Doctoral dissertation. Michigan State University: East Lansing, MI. Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. ----------. 1994. Principles of Linguistic Change: Internal Factors. Oxford: Blackwell. Labov, William and Sharon Ash. 1997. “Understanding Birmingham”. Language and Variety in the South Revisited ed. by Cynthia Bernstein, Thomas Nunnally, and Robin Sabino, 508–573. Tuscaloosa and London: University of Alabama Press. Labov, William, Sharon Ash, and Charles Boberg. 1997. A National Map of the Regional Dialects of American English. Retrieved from University of Pennsylvania Linguistics Website: http://www.ling.upenn.edu/phono_atlas/NationalMap/NationalMap.html. Labov, William, Sharon Ash, and Charles Boberg. Forthcoming. Atlas of North American English. Berlin: Mouton de Gruyter. Ladefoged P. and Broadbent D.E. 1957. “Information conveyed by vowels”. Journal of the Acoustical Society of America 29.98–104. Miller. J. D. 1989. “Auditory-perceptual interpretation of the vowel”. Journal of the Acoustical Society of America 85:2.114–134. Milroy, James. 1992. Linguistic Variation and Change. Oxford: Blackwell. Milroy, Lesley. 1980. Language and Social Networks. Oxford: Blackwell. Milroy, Lesley and Dennis R. Preston, eds. 1999. Journal of Language and Social Psychology 18:1 (Special Issue: Attitudes, perception, and linguistic features). Niedzielski, Nancy and Dennis R. Preston. 1999. Folk Linguistics. Berlin: Mouton de Gruyter. Payne, Arvilla. 1980. “Factors controlling the acquisition of the Philadelphia dialect by out-of-state children”. Locating Language in Time and Space ed. by William Labov, 143–178. New York: Academic Press.
DIALECTS ACROSS INTERNAL FRONTIERS
155
Plichta, Bartek and Dennis R. Preston. 2003. “The /ay/s have it. Stereotype, perception, and region”. A paper presented to NWAV 32. Philadelphia, October. Plichta, Bartek and Brad Rakerd. 2002. “Perceptions of /a/-fronting across two Michigan dialects”. A paper presented to NWAV 31. Stanford University, October. Preston, Dennis. 1996. “Where the worst English is spoken”. Focus on the USA ed. by Edgar Schneider, 297–360. Amsterdam: John Benjamins. ----------, ed. 1999. Handbook of Perceptual Dialectology, Volume I. Amsterdam: John Benjamins. Preston, Dennis R. and Jones, Jamila. In progress. The language varieties of African-Americans in Lansing, MI. Ryan, Ellen Bouchard and Howard Giles, eds. 1982. Attitudes Towards Language Variation. London: Edward Arnold. Schiffrin, Deborah. 1994. Approaches to Discourse. Oxford: Blackwell. Schilling-Estes, Natalie. 1998. “Investigating ‘self-conscious’ speech: The performance register in Ocracoke English”. Language in Society 27.53– 83. Syrdal, A. K. and H. S. Gopal. 1986. “A perceptual model of vowel recognition based auditory representation of American English vowels”. Journal of the Acoustical Society of America 79.1086–1100. Thomas, Erik. 2001. An Acoustic Analysis of Vowel Variation in New World Englishes. (= Publication of the American Dialect Society, 85). Durham, NC: Duke University Press. Trudgill, Peter. 1972. “Sex, covert prestige and linguistic change in the urban British English of Norwich”. Language in Society 1:2.179–195. ----------. 1986. Dialects in Contact. Oxford: Blackwell. Tucker, G. Richard and Wallace E. Lambert. 1969. “White and Negro listeners’ reactions to various American-English dialects”. Social Forces 47.463–468. Williams, Frederick. 1976. Explorations of the Linguistic Attitudes of Teachers. Rowley, MA: Newbury House.
ON ‘DATIVE SICKNESS’ AND OTHER LINGUISTIC DISEASES IN MODERN ICELANDIC FINNUR FRIÐRIKSSON University of Akureyri 1.
Diagnosis of illness Judging by the public debate in Iceland, the Icelandic language is currently undergoing a number of changes which are commonly feared to disrupt the image of stability that the language has both in the eyes of the Icelanders themselves (see, e.g. Gíslason et al. 1988) as well as in international sociolinguistics circles (see, e.g. Milroy & Milroy 1985). Furthermore it is often claimed that these changes might disrupt or even break the direct links there are believed to exist between the Iceland of today and its historical past, which is preserved primarily in the sagas and other medieval literature. Unsurprisingly, English influences are amongst the main culprits in this respect, especially as far as the Icelandic vocabulary is concerned, as various loanwords are now quite well established at the same time as much of the everyday slang stems from English. However, there are also a few types of changes which are normally not “blamed” on English and affect the morphology and syntax of Icelandic rather than its vocabulary. These changes are the focus of this paper. Three types of change not attributable to English have attracted most of the attention: a) ‘dative sickness’; b) ‘new passive’; and c) ‘genitive avoidance’. As for the first of these, i.e. ‘dative sickness’, there are a number of so-called impersonal verbs in Icelandic which allow their subjects to be in oblique case. 1 For most of these verbs the case in question is dative, as in mér finnst (‘I-DAT think/feel’ [when expressing an opinion]), but there is also a smaller number of verbs that take the accusative, as in mig langar (‘I-ACC want’), and a handful of verbs that take the genitive, as in mín nýtur við (‘IGEN am present’). Lately, a tendency for the verbs that normally take accusative to take dative instead is believed to have emerged, resulting in constructions such as mér langar and the term ‘dative sickness’, or 1
The existence of this phenomenon in Icelandic was first established by Andrews (1976), and his analysis appears to be generally accepted.
158
FINNUR FRIÐRIKSSON
þágufallssýki, as it is called in Icelandic. Also, some personal verbs, particularly hlakka (= ‘look forward to’) and kvíða (= ‘be anxious’) apparently tend to appear in impersonal forms with either accusative or dative subjects instead of nominative. The second example of this type of change, i.e. ‘new passive’ is, as the term implies, a change in how passive sentences are constructed in modern Icelandic. The standard passive can be divided into three types (Sigurjónsdóttir & Maling 1997): a) Nominative Passive – Ég var laminn (‘I-NOM was hit’); b) Oblique Passive – Honum var hrint (‘He-DAT was pushed’); and c) Impersonal Passive – Það var sungið (‘It/there was sung’). For a) and b) new variants have begun to emerge and thus Ég var laminn becomes Það var lamið mig (‘There/It was hit me-ACC’), whereas Honum var hrint becomes Það var hrint honum (‘There/It was pushed him-DAT’). As can be seen, the ‘new passive’ is thus constructed by inserting a dummy subject rather than moving the object of the corresponding active sentence into subject position. Furthermore, the object from the active sentence keeps its case in these passive constructions. The third type to be described here, i.e. ‘genitive avoidance’ gets its label from a tendency that many Icelanders think they see in modern Icelandic, namely, that the position of the genitive case is weakening. What appears to be happening is either that some genitive case endings are disappearing to be replaced by accusative/dative endings or simply that the accusative/dative case is used where genitive is the standard case (Kjartansson 1979; Svavarsdóttir 1994). Suggestions have also been made that this phenomenon is restricted to certain nouns, especially feminine ones which in the nominative case end in -ing, such as breyting (‘change’). The accusative and the dative ending is -u, and the standard genitive ending is -ar, which some now believe to be under threat from the -u-ending. It might also be the case that ‘genitive avoidance’ is an instance of a more major change or problem, rather, as many Icelanders would probably have it, as many — especially teachers of Icelandic in elementary schools and high schools — claim that the entire case-inflectional system of Icelandic is under threat, or at least showing signs of general instability. According to these teachers this emerges either in their students’ using an incorrect case, e.g. accusative where there should be dative, or in a tendency not to inflect at all, leaving nominative as the only productive case. 2 Of the alleged changes described above, ‘dative sickness’ is probably the most strongly stigmatized and actively fought against in Icelandic schools. 2
Personal communication with Kristín S. Árnadóttir, a teacher of Icelandic at Verkmenntaskólinn á Akureyri (a high school).
ON ‘DATIVE SICKNESS’
159
However, the stigmatization attached to the ‘new passive’ is also quite strong, and deviant case inflections are generally frowned upon at the same time as the correct usage of case inflections is by tradition heavily emphasized in the teaching of Icelandic. Despite the claims of ongoing change and the public debate, very little work has been carried out on these alleged new features. Thus no work exists on the supposed general instability of the inflectional system, and ‘genitive avoidance’ has only been identified and described without any real attempt at finding out how firmly established it is in the language (Kjartansson 1979, 1999; Svavarsdóttir 1994). As for the remaining two features, i.e. ‘dative sickness’ and the ‘new passive’, some work has been carried out which indicates that the claims made about their progress may have some footing. Thus, in a study on ‘dative sickness’, Svavarsdóttir (1982) found, e.g. that more than 30% of her subjects — all of whom were 11-year-old children — used the dative form of the pronoun hún (‘she’), i.e. henni, rather than the ‘correct’ accusative form hana, when asked to fill in blanks for the subjects of the verbs langa (‘want’) and vanta (‘need/lack’). These are probably the two most commonly used verbs that in standard Icelandic take accusative subjects and also the two verbs most often pointed out as problematic as regards ‘dative sickness’. Interestingly, Svavarsdóttir also found signs of a reverse ‘dative sickness’ or even ‘accusative sickness’, as her study included a handful of verbs which in the standard language take dative subjects, and are generally not regarded as problematic in any way, but here frequently showed up with accusative subjects instead. This, however, Svavarsdóttir interprets primarily as a sign of hypercorrection. As for the ‘new passive’, recent research (Sigurjónsdóttir & Maling 2001) has been taken as a sign of this feature having become quite frequent amongst Icelandic teenagers. For this research, which was carried out in 1999– 2000, 1731 tenth-grade pupils (born in 1984; 45% of the total cohort in Iceland) and 205 adults from all over Iceland were presented with a number of test sentences and asked to indicate whether or not they were acceptable. The general results were that the majority, or 51–75%, of the younger subjects living outside Reykjavík and in its suburbs accepted the ‘new passive’ constructions, while the acceptance rates amongst teenagers in inner Reykjavík dropped to about 30% and did not exceed 9% amongst adults anywhere in the country. Furthermore, there was a strong correlation between the teenagers’ acceptance rates and their parents’ education; the higher the educational level the lower the acceptance rates. Gender, however, did not seem to be a significant factor. To this it should be added that the pattern seen in these results resembles that in Svavarsdóttir’s (1982) research mentioned above and the pattern seen in a follow-up study aimed at using Svavarsdóttir’s original
160
FINNUR FRIÐRIKSSON
data to look more closely at the connection between ‘dative sickness’ and a number of social factors (Svavarsdóttir et al. 1984). Here it emerged that ‘dative sickness’ was less frequent in Reykjavík than in other parts of the country and that children from the lower classes tended to be afflicted by this ‘disease’ more often than children of a higher social standing. While the work just described undoubtedly makes valuable contributions to the description of the features in question and can probably be said to confirm their existence, it is in many ways less than satisfactory from a sociolinguistic point of view. The major problem with this work is that it is based on different kinds of tests rather than any kind of authentic data, written or spoken, which means that it is less than clear how widely spread these features have actually become in everyday language, even though the results discussed earlier have been used to draw conclusions on this matter. Also, the greater part of this work deals only with either written or spoken language, but when discussing their results researchers tend not to make any distinction between the two and simply apply their findings to Icelandic in general. Finally, there appears to exist a strong tendency in Iceland to regard the new features in question as restricted to children and teenagers, and thus only Sigurjónsdóttir’s and Maling’s (2001) research includes adult subjects. To get a clear picture of how common these features have become all age groups need to be included, not the least since the clear difference between adults and teenagers as regards the acceptance rates for the ‘new passive’ might just be seen as a sign of children and teenagers not having mastered the rather complex construction of passives rather than as an indication of an emerging change. As we only have results for children on ‘dative sickness’ and no results at all on the other features mentioned above, similar questions arise here about just how far these so-called changes really have proceeded. 2. 2.1
The present study Methods As we can see, the picture regarding the occurrences of ‘dative sickness’, ‘new passive’, and ‘genitive avoidance” and other deviant case inflections in everyday language is quite unclear. In order to clarify this picture somewhat I have embarked on a project which aims at collecting authentic written and spoken data from selected locations in each of the eight regions generally recognized in Iceland for various official and semi-official purposes. The choice of locations is intended to reflect the general societal structures of each region and to include as representative a sample as possible of the various kinds of urban centres and rural areas there can be said to exist in Iceland. What is to be presented here are the results from the first steps taken within this framework. So far data have been collected from a total of 18
ON ‘DATIVE SICKNESS’
161
subjects; 6 from each of the three towns Neskaupstaður in the Eastern Fjords, Akureyri in the northern part and Patreksfjörður in the Western Fjords. Neskaupstaður and Patreksfjörður can both be described as typical Icelandic fishing villages, with about 1,400 and 750 inhabitants respectively. Akureyri, however, is, with its nearly 16,000 inhabitants, the largest town outside the Reykjavík area, and its economical foundations are more mixed than those of Neskaupstaður and Patreksfjörður, as besides the local fishing industry one can in Akureyri find several other industrial branches as well as a small university and a reasonably large hospital. The characteristics of the three towns are reflected in the choice of subjects, all of whom had lived most — if not all — of their lives there. In each of the three locations there are two subjects from each of the three age groups included; teenagers (aged 17–20), adults/middle-aged (aged 21–64) and seniors (aged 65 and above). Each age group in each location consists of one male and one female subject. The results presented below are based on a total of about 5 ½ hours of free conversation, recorded without the presence of the researcher at the participants’ homes, or — in the case of two of the teenagers — at school; and a total of 116 pages (ca. 6,5/subject) of written material of a very varied nature (school papers, personal letters/e-mails, meeting notes, personal homepages etc.), none of which were written specifically for the purpose of this project. 2.2
Results The most striking general finding so far is that, contrary to what could be expected based on previous research and regardless of whether we are talking about written or spoken language, none of the new features occur often enough to justify any claims about them being established in any way. Obviously, the limited data presently available do not justify any strong claims to the contrary either, but it can nonetheless be seen as an indication of what might be expected when all the data have been collected. This is especially surprising, as the results reported here are based on fieldwork outside of Reykjavík, where the new features were expected to be more prevalent than in the capital district. Another striking finding, which also contradicts what was expected, is that there are no clear differences between spoken and written language as regards the frequency of any of the new features; they are just about as rare in spoken language as they are in written language. Some differences emerge, however, in how non-standard forms appear, especially as regards case inflections. We will come to this shortly.
2.2.1 ‘Dative sickness’. If we now look more closely at each of the different features in question we can see (Table 1) that verbs normally taking accusative-
162
FINNUR FRIÐRIKSSON
case subjects are quite rarely used in general, both in written and spoken language. Thus only 5 such verbs occurred a total of 21 times in the written material and 29 times in the spoken material. As for non-standard forms of the subjects of these verbs, only 1 occurred in written language and 2 in spoken language, which gives us an ‘error’ percentage of 4.8 and 6.9 respectively. Spoken language Written language Standard forms 27 20 Non-standard; ‘dative sickness’ 1 0 Non-standard; nominative 1 1 % Non-standard forms 6.9 4.76 Table 1: Frequencies of standard and non-standard cases used for subjects of impersonal verbs standardly taking accusative subjects.
More interestingly, however, only one of these three non-standard forms is an example of ‘dative sickness’; on the other two occasions nominative was used rather than the standard accusative, thus, in a sense, making the verb personal rather than impersonal. On both these occasions the verb in question was dreyma (‘dream’). This verb was used five times in all, which means that it was used with a non-standard subject in 40% of these instances, and even though this general number of usages is too low to draw any conclusions at this stage, it will be interesting to see if this tendency is repeated in the data as a whole. As for the one instance of ‘dative sickness’, the verb involved was vanta (‘need/lack’), which, as has been mentioned earlier, is one of the two verbs most often linked to ‘dative sickness’ in the public debate. However, this verb was used with a subject in the accusative case on 18 other occasions, indicating that it may not be showing as clear signs of ‘dative sickness’ as many believe it to do. It can also be added that the other verb most strongly linked to ‘dative sickness’, i.e. langa (‘want’) was used with a standard accusative subject on each of the 15 occasions it appeared both in speech and writing. As was discussed earlier, Svavarsdóttir (1982) found signs of a reverse ‘dative sickness’, or ‘accusative sickness’, in her study and for that reason impersonal verbs normally taking dative-case subjects were also looked at in the present study. As has been mentioned earlier, these verbs are generally not regarded as problematic and there also seems to be little reason for fear since, as Table 2 shows, only 3 non-standard forms were found in a total of 131 occurences of such verbs in the written data. The corresponding figures for the spoken data were 1 out of 113. The percentage of non-standard forms is thus in this case 2.29 and 0.88, respectively. Interestingly, however, the case used in all four instances of non-standard usage was nominative and there are three
ON ‘DATIVE SICKNESS’
163
different verbs involved. When this is added to the results for the verbs taking accusative-case subjects in the standard language, the only possible pattern to be detected here is a tendency for impersonal verbs in general to use nominative subjects. This is interesting, as this possibility appears only to have been hinted at earlier (see, e.g. Pálsson 1979) without it having aroused any special interest. Instead, ‘dative sickness’ and the tendency — so far undocumented in my study — to use some personal verbs in an impersonal way have been in the spotlight. However, it is doubtful whether an overall frequency of little more than 2% for the usage of nominative subjects can be called a tendency, even though this is a higher figure than is reached by any of the other non-standard features that were looked at in this study. Spoken language Written language Standard forms 112 128 Non-standard; ‘accusative sickness’ 0 0 Non-standard; nominative 1 3 % Non-standard forms 0.88 2.29 Table 2: Frequencies of standard and non-standard cases used for subjects of impersonal verbs standardly taking dative subjects.
2.2.2 ‘New passive’. The ‘new passive’ hardly occurs at all. Thus, as can be seen in Table 3, there are only 3 instances (0.71%) of it amongst a total of 422 written nominative and oblique passives, and 1 instance (1.27%) out of a total of 79 such passives in the spoken data. One of the three ‘new passives’ in the written data emerged in an oblique passive while the other three instances, in spoken and written language, were found in nominative passives. However, the total number of ‘new passives’ is in all probability far too low for any pattern to emerge from this. Spoken language Written language Standard 78 419 ‘New passive’ 1 3 % ‘New passive’ 1.27 0.71 Table 3: Frequencies of standard and ‘new passive’ forms.
2.2.3 ‘Genitive avoidance’. As far as ‘genitive avoidance’ is concerned, the results can once again be summed up by saying that hardly any traces of it were found. Table 4 shows that non-standard forms were used in only 7, or 0.55%, of the total number of 1267 written constructions requiring genitive, while the corresponding figure for spoken language is 2, or 0.83%, of 242.
164
FINNUR FRIÐRIKSSON
Spoken language Written language Standard 240 1260 Non-standard 2 7 % Non-standard 0.83 0.55 Table 4: Genitive: Frequencies of standard and non-standard usage. 3
Not only do non-standard genitive forms appear very infrequently in the data this study is based on, it is also doubtful whether any of the few ‘errors’ that do appear form examples of ‘genitive avoidance’ as it has been described. Thus six of the total nine non-standard forms occur in a long stretch of caseinflected words, only one of which would be expected to appear in the genitive in the standard language. The ‘errors’ made here should probably be interpreted as a sign of the effect of the surrounding context rather than as a sign of ‘genitive avoidance’ or lack of knowledge of how to inflect for genitive. Also, two of the remaining three non-standard forms seem to have more to do with an insecurity about the gender and number of the word in question than the case-inflection as such. As for the last one, which occurred in spoken language, we might there have a special instance of ‘genitive avoidance’. Here the speaker breaks off his speech after the first syllable of a word which the context requires to appear in genitive. However, he starts over immediately and then uses the nominative case, thus apparently avoiding genitive altogether rather than risking making an error in inflecting for it. This last feature might also be a reflection of a general avoidance of genitive in spoken language, as it is much less used there than in written language; genitives make up 15% of all the oblique cases in the written data, but only 4.85% of the oblique cases in the spoken data. Of course, certain genitive constructions occur fairly frequently in spoken language, e.g. phrases such as pabbi Jóns (‘Jón’s father’) and a number of phrases referring to time or quantity, such as hann er tveggja mánaða (‘he is two-GEN months-GEN old’). However, several other constructions which are seen from time to time in written language hardly occur at all in spoken language. These include, e.g. prepositional phrases with the prepositions vegna (‘due to’) and auk (‘besides’), both of which govern the genitive case, and a number of constructions which in syntactic terms hardly differ at all from the example of pabbi Jóns given above. The difference here emerges on the level of formality, as superfluous usage of genitive in Icelandic is felt to result in a very formal and ‘heavy’ style, which is normally avoided in spoken language. Thus while pabbi Jóns can hardly be said to have a formal ring to it, most speakers of 3
Note that Tables 4, 5 and 6 show the total numbers of non-standard usage of the cases in question, i.e. of non-standard usage of morphology as well of non-standard usage of syntax.
ON ‘DATIVE SICKNESS’
165
Icelandic would probably feel that the following construction, taken from the written material provided by one of my subjects, would be somewhat out of place in spoken everyday language: Ástæðan fyrir falli Hannesar Hafsteins og flokks hans... (‘the reason for Hannes-GEN Hafstein’s-GEN and his-GEN party’s-GEN fall...’). In the spoken language, this would probably be paraphrased without any genitive construction, more often than not with the aid of prepositions governing other cases. In addition to this, several genitive case endings can be difficult to use, not the least in the context of spoken language, as they diverge quite sharply from the other case endings. This can be seen in the example given in Section 1 above and is further witnessed in the last of the nine non-standard forms discussed above. The word used there is síldarvinnslan (‘the herring factory’), the genitive form of which is síldarvinnslunnar. All in all, therefore, if we want to talk about ‘genitive avoidance’ altogether, it might be more appropriate to do so in terms of genitive constructions being avoided in spoken language rather than in terms of the genitive case merging with the accusative/dative case. 2.2.4 Other case inflections. As Tables 5 and 6 show, much the same pattern emerges for the accusative and dative case inflections as for the genitive. Firstly, we see far too low frequencies of non-standard forms to justify any claims about an instability creeping into the inflectional system. Thus, as regards inflections for the accusative, 19 non-standard forms appear in the written language, which make for 0.53% of a total of 3607. In the spoken language, 13 non-standard forms appear, constituting 0.46% of a total of 2846. As for the dative case inflections, the 30 ‘errors’ that occur in written language make up 0.85% of a total of 3544 constructions where the dative is required. Spoken language Written language Standard 2833 3588 Non-standard 13 19 % Non-standard 0.46 0.53 Table 5: Accusative: Frequencies of standard and non-standard usage.
Spoken language Written language Standard 1873 3514 Non-standard 29 30 % Non-standard 1.52 0.85 Table 6: Dative: Frequencies of standard and non-standard usage.
FINNUR FRIÐRIKSSON
166
The corresponding figures for the spoken language are 29 non-standard forms out of a total of 1902, resulting in an ‘error’ percentage of 1.52. Non-standard forms thus appear slightly more frequently in dative constructions than in accusative constructions, but this is not a major difference. Secondly, even though about a third of all the non-standard forms, written and spoken in both accusative and dative constructions, appear to be the result of a lack of knowledge of what prepositions or verbs govern what case or of how to inflect for the correct case, the remaining two thirds can probably best be explained by simply looking at the context in which they appear. Thus, in the written language the majority of the non-standard forms occur in a long stretch of inflected words and/or when there is a gap between the case-governing verb or preposition and the inflected word. This seems to have the effect that somewhere along the way the writer loses track of what case should be used. A typical example of an ‘error’ of this kind can be seen in (1) below: (1)
þar má nefna (verb-ACC) minnisleysi, munaðarleysi, ‘there can be mentioned amnesia, orphanhood, barnsrán, föðurlaust barn, peningavandamál eða the kidnapping of a child, a fatherless child, financial problems or kannski vond tengdamamma (adj.+noun-NOM) maybe an evil mother-in-law’
What is especially noteworthy in this case is that, even though all the words in the string minnisleysi, munaðarleysi, barnsrán, föðurlaust barn, peningaleysi appear, strictly speaking, in the accusative as assigned by the verb nefna, none of them take any inflectional ending for that purpose. Thus they appear as identical with their nominative form. To form the accusative of vond tengdamamma, however, inflectional endings are needed, but the long string of seemingly non-inflected, or nominative, nouns between the verb and this last noun phrase seems to have almost a priming effect which results, as it were, in the annulment of the case-governing effect of the verb. In the spoken data there are, unsurprisingly, very few long stretches of inflected words, but gaps between the case-governing item and the inflected element are all the more common, in this case either due to pauses in a single speaker’s speech or to interruptions by the conversational partner or partners. Just as in the written example above, however, these gaps appear to weaken the case-governing effects of the preceding elements, as the non-standard forms which should be in the accusative or dative case nearly always occur in the
ON ‘DATIVE SICKNESS’
167
nominative in these cases. Thus we see in (2a) below that, after having correctly used the dative form einhverju, ‘something’, governed by the preposition í, the speaker makes a short pause to then restart with the nominative form of einhverju, i.e. eitthvað, even though the dative form is still — at least strictly speaking — required by the preposition í. In (2b), on the other hand, it can be seen how D’s entrance into the conversation in turn 5 breaks the pattern established by A using the verb kaupa in turn 1. Kaupa governs the accusative case, and both B and C adhere to this in turns 2 and 3. D, on the other hand, has joined in with a nominative form, even though there is nothing in the conversation that directly indicates the discontinuation of the case-governing effect of kaupa. However, the contribution of D may also be interpreted as the beginning of a new contribution and explanation, or a specification of the content of B’s and C’s previous utterances. What these examples do, then, is raise the question as to whether non-standard forms of the kind presented here really should be analysed as such in spoken language, or whether we should accept two different standards for spoken and written language where case governing across gaps of this kind in spoken language is seen as optional. 4 (2a) Pause in single speaker’s speech A:
hann er í (prep. – DAT) einhverju (DAT) svona / eitthvað (NOM) ‘he is in something like / something í menntageiranum in the educational branch’
(2b) Interruption by conversational partner(s) 1 2 3 4
B:
ég ætla að kaupa (verb-ACC) kaptein morgan ‘I am going to buy captain morgan’ C: pela (ACC) flösku (ACC) ‘a flask a bottle’ B: flösku (ACC) ‘a flask’ C: e+ já ‘e+ yes’
4 For a further discussion of the need to distinguish between spoken and written language, see, e.g. Linell (1982).
FINNUR FRIÐRIKSSON
168 5 6 7 8
D:
ein (NOM) ‘one’ B: mhm stór (NOM) ‘mhm big’ D: hm ‘hm’ B: stór (NOM) kannski lítil (NOM) ‘big maybe small’
2.2.5 Tendencies amongst the subject groups. The low frequencies of any of the non-standard forms discussed here prohibit any statistical comparisons between the different groups involved, at least at this stage. However, a few tendencies can be detected: first, that the subjects from Patreksfjörður in the Western Fjords have the highest frequency rates for nearly all the non-standard forms; second, that all the non-standard forms with impersonal verbs are used — somewhat surprisingly — by women; and third, that teenagers generally show a higher frequency of non-standard case inflections than do the other two age groups. However, adults and seniors are by no means completely free from non-standard case inflections, while the non-standard forms occurring with impersonal verbs and with passive constructions are quite evenly spread amongst the three age groups. As mentioned above, very little can be read into the results as regards possible sociolinguistic differences at this stage. Rather, these findings remind us of the need for including all age groups in sociolinguistic research instead of taking for granted that stigmatized features are restricted to children and teenagers. 3.
Discussion and conclusion It seems clear that, although the present results are based on a limited amount of data, they indicate that ‘dative sickness’, ‘new passive’, ‘genitive avoidance’ and a general instability in the case inflectional system have not advanced as far as many Icelanders believe them to have done. Furthermore, there is very little to suggest that the non-standard features are spreading innovations, as they, e.g. occur fairly evenly amongst all the age groups. If they were examples of innovations, they would probably be most common among younger speakers, a claim which has been made about at least ‘dative sickness’ and the ‘new passive’. One possible explanation for the findings is the pervasive, uniform message of the educational system, which, as mentioned above, tries to combat ‘dative sickness’ and the ‘new passive’, and emphasizes careful use of case inflections quite strongly. This receives some support from the fact that several of the informants claimed that the correct usage of the features in question, especially as regards impersonal verbs, had been more or
ON ‘DATIVE SICKNESS’
169
less ‘hammered in’ during their school days. However, this is unlikely to be the only explanation, as can be seen, e.g by the fact that the non-standard forms appear to be used only slightly more in spoken language than in written language. In fact, the main difference between spoken and written language seems to be in terms of how the non-standard forms appear — especially as regards the case-inflectional system — rather than in terms of their frequency. If the educational system was having major effects on this variation, these effects would be most likely to emerge primarily in written language. Thus, explanations for the apparent stability of the spoken language need to be found elsewhere. Just where these explanations can be found is unclear at the moment, but hopefully as yet unanalyzed interviews which were carried out with the informants about, e.g. their linguistic attitudes and social networks, will give me some direction in searching for them. What this leaves us with are questions as to why the belief has emerged that all the new features discussed here have become quite common in Icelandic, when what we might actually be witnessing at the present time is low-level variation, which may or may not indicate that large-scale changes may take place in the future. There is no single, simple explanation for the belief in the pervasiveness of this variation, and much further research is needed before any final answers can be given. Nonetheless, a few possible factors should be mentioned at this stage. First of all, at the same time as the educational system can probably be given credit (if that is the right word to use) for its part in keeping the occurrence of the new features at a low level, it is a likely source of beliefs of the kind mentioned above. The emphasis put on case inflections and the stigmatization attached to ‘dative sickness’ and the ‘new passive’ puts pressure on teachers to pay special attention to these features in their pupils’ speech and writing. This turns the features into a few trees which might conceal a much larger linguistic forest of variation and change behind them. Secondly, this development is further aided by the fact that, with the possible exception of accusative and dative case inflections, the features in question are easy to pick out, as they can in many ways be said to be atypical constructions. Thus, impersonal verbs are much rarer than personal verbs, genitive is the case least used in Icelandic, and passive constructions are not used nearly as frequently as active ones. Furthermore, the use of non-nominative subjects makes the impersonal verbs stick out, and the endings used to inflect for genitive case often deviate sharply from the other case endings, making them — and nonstandard forms of them — easily distinguishable. This of course also means that it is relatively easy even for the general public to pick these features out and even though very little research exists on the matter, it is probably safe to say that the general linguistic atmosphere in Iceland is highly conservative.
170
FINNUR FRIÐRIKSSON
This atmosphere implies that most signs of variation or change, not least those which are easily and quickly detected as is the case here, are soon noticed. The frequency of their occurrence is then exaggerated, and stigmatization ensues in the effort to stop their further spread. Finally, it is possible that the earlier research discussed above plays its part in establishing this belief, as its results have been interpreted as a sign of at least ‘dative sickness’ and the ‘new passive’ being used quite frequently in modern Icelandic, even though the grounds on which these conclusions are drawn are less than adequate. However, it is doubtful that these results are well known to the general public in any great detail. Be that as it may, it nonetheless seems clear that when all the factors just mentioned are combined, they create an image of instability which — somewhat paradoxically — is likely to play a crucial part in maintaining linguistic stability: if you do not feel that there is anything to fight against, there is little reason to keep on fighting. References Andrews, A. D. 1976. “The VP Complement Analysis in Modern Icelandic”. Proceedings of the North East Linguistic Society 6.1–21. Gíslason, I., B. Jónsson, G. B. Kristmundsson & H. Þráinsson. 1988. Mál og samfélag: Um málnotkun og málstefnu. Reykjavík: Iðunn. Kjartansson, H. S. 1979. “Eignarfallsflótti: Uppástunga um nýja málvillu”. Íslenskt mál og almenn málfræði 1.88–95. ----------. 1999. “Orð í belg um eignarfallsflótta”. Íslenskt mál og almenn málfræði 21.151–160. Linell, P. 1982. The Written Language Bias in Linguistics. Linköping: Tema Kommunikation. Milroy, J. & L. Milroy. 1985. “Linguistic change, social network and speaker innovation”. Journal of Linguistics 21.339–384. Pálsson, G. 1979. “Vont mál og vond málfræði: Um málveirufræði”. Skírnir 153.175–201. Sigurjónsdóttir, S. & J. Maling. 1997. “The ‘New Passive’ in Icelandic”. Proceedings of the 21st Boston University Conference on Language Development ed. by E. Hughes, M. Hughes and A. Greenhill, 378–389. Somerville: Cascadilla Press. Sigurjónsdóttir, S. & J. Maling. 2001. “Það var hrint mér á leiðinni í skólann: þolmynd eða ekki þolmynd?”. Íslenskt mál og almenn málfræði 23.123– 180. Svavarsdóttir, Á. 1982. “‘Þágufallssýki’: Breytingar á fallnotkun í frumlagssæti ópersónulegra sagna”. Íslenskt mál og almenn málfræði 4.19–62. ----------. 1994. “Enn um eignarfallsflótta”. Jónína hans Jóns G. Friðjónssonar fimmtugs, 7–13. Reykjavík.
ON ‘DATIVE SICKNESS’
171
Svavarsdóttir, Á., G. Pálsson & Þ. Þórlindsson. 1984. “Fall er fararheill: Um fallnoktun með ópersónulegum sögnum”. Íslenskt mál og almenn málfræði 6.33–55.
CAN WE FIND MORE VARIETY IN VARIATION? RONALD MACAULAY Pitzer College Claremont, CA 1.
Introduction Three points are central to the present paper. The first is that quantitative methods are useful for the examination of linguistic variation. The second is that dialects are multi-layered and not restricted to any stratum of the community. The third is that there are more features of language variation to be investigated by quantitative methods than has happened so far. Table 1 lists a number of studies in which language variation has been examined by quantitative methods and also shows the extralinguistic factors examined in each. In Table 1 the extralinguistic factors have been cited as yes where they have been correlated with the linguistic variables. Thus, since Milroy 1980 deals only with working-class speakers, there is no social class comparison; similarly, Cheshire 1982 deals with adolescents but does not contrast their Social class Gender Age Other factors Labov 1963 no no yes ethnicity, ambition Labov 1966 yes yes yes ethnicity Wolfram 1969 yes yes yes racial isolation Fasold 1972 yes yes yes race of interviewer Trudgill 1974 yes yes yes rurality Macaulay 1977 yes yes yes religion Feagin 1979 yes yes yes ethnicity, locale Milroy 1980 no yes yes religion, social network Cheshire 1982 no yes no peer group status Coupland 1988 yes yes no education Macaulay 1991 yes no no none Eckert 2000 yes (?) yes no peer group status McCafferty 2001 yes yes yes ethnicity, religion Labov 2001 yes yes yes social network Table 1: Extralinguistic factors in quantitative studies of language variaion.
RONALD MACAULAY
174
behaviour with that of other age groups, so age is not an extralinguistic variable. Table 1 shows that sociolinguists employing a quantitative approach have generally been interested in a wide range of social factors, so my first two points have been well covered. What about my third point regarding the kind of variation studied? Table 2 shows the kind of variables that have been examined in these studies. The most salient differences are phonological and morphological since they are often diagnostic of social differentiation in communities (Coupland 2001: 189), so it is hardly surprising that they have received the most attention. Nevertheless, a sceptical observer might wonder whether these studies have fully investigated what J. Milroy (1979: 91) calls ‘the sociolinguistic complexity’ of the communities. The most prominent example of going beyond phonological and morphological features (‘dialect features’ in Coupland’s 2001: 189 sense) was the work of Basil Bernstein (1962, 1971), and the reaction to his views was so extreme that it probably discouraged sociolinguists from exploring the kind of questions Bernstein raised (Edwards 1987). As Bernstein himself rather ruefully remarked on recalling his contribution to the study of language differences, his distinction between an elaborated and a restricted code became a means of bestowing ideological purity on those who denounced it. As part of my research into variation in discourse features I have recently been looking at some of the features Bernstein examined and have found some interesting comparisons. Labov 1963 Labov 1966 Wolfram 1969 Fasold 1972 Trudgill 1974 Macaulay 1977 Feagin 1979 Milroy 1980 Cheshire 1982 Coupland 1988 Macaulay 1991
two phonological variables five phonological variables four phonological variables, four grammatical variables two phonological variables, two grammatical variables sixteen phonological variables, one grammatical variable five phonological variables six grammatical variables nine phonological variables sixteen grammatical variables six phonological variables eight phonological variables, four morphological variables, eight syntactic and three discourse features Eckert 2000 six phonological variables, one grammatical variable McCafferty 2001 five phonological variables Labov 2001 nine phonological variables, one grammatical variable Table 2: Sociolinguistic variables in quantitative studies of language variation.
CAN WE FIND MORE VARIETY IN VARIATION?
175
In order to move beyond the examination of phonological and morphological features, it is necessary to have extended samples of talk-inaction with speakers interacting with each other. It is impossible to investigate discourse features through the use of word lists and reading passages. The materials examined in this paper come from a set of recordings made in 1997. Jane Stuart-Smith recorded thirty-three speakers for her study of language change in Glasgow (Stuart-Smith 1999). She recorded both adults and adolescents using the methodology developed for the Newcastle/Derby study (Docherty, Foulkes, Milroy, Milroy & Walshaw 1997). The sample was balanced by age (adolescents 13–14 and adults 40+), social class (middle-class and working-class), and gender (Stuart-Smith 1999: 204). Participants were asked to choose a friend or acquaintance with whom they would be willing to talk for half an hour in the presence of a tape-recorder, without the investigator being present. The conversations were transcribed in their entirety and the contribution of each speaker isolated. The resulting transcripts provide comparable samples that allow the investigation of age, gender, and social class differences in the use of various discourse features. The working hypotheses for the quantitative analysis of discourse features are as follows: 1. All speakers have the same opportunity to use certain discourse features in the recording sessions. 2. Variation in the frequency of use of any of these feature reflects a different discourse style. 3. Differences in using a discourse feature that correlate with membership of a social category such as age, gender, or social class show that such variation is not simply idiosyncratic. What to count and how There are now sophisticated computer programmes that can extract items from texts (e.g., some of the methods illustrated in Baker, Francis, and Tognini-Bonelli 1993) but there are still important decisions to make. Since even the most sophisticated programme will not be capable of the kind of judgments that human beings make, the value of these programmes is limited. While the identification of tokens should as far as possible be carried out without recourse to the kind of subjective criteria that require human interpretation for some features the latter approach is essential. Thus, in selecting discourse features to count there are several distinctions worth making, and the following classification may be useful. 2.
RONALD MACAULAY
176 I.
II. III.
Unambiguous forms 1. Invariant forms a. Consistent use i. Function ii. Meaning b. Variable use 2. Minimally variant forms Ambiguous forms Complex forms Table 3: Types of discourse features.
I. 1. a. i. Unambiguous forms are those that can be found by a simple word search through the transcript. An example of an unambiguous form that is consistent in function and meaning is very. It is possible to make a simple frequency count of the use of very without taking into account the linguistic context in which it occurs. It might not seem that this would be a rewarding item to investigate but in fact the middle-class speakers in the Glasgow conversations use very thirteen times more frequently than the working-class speakers, as can be seen in Table 4. Working-class Middle-class # Freq. # Freq. very 16 0.32 147 4.28 quite 60 1.19 125 3.64 Table 4: Relative frequency of very and quite in Glasgow adults (freq. = per 1,000 words).
(p.< .001) (p.< .002)
This social class difference in the use of very in the Glasgow sample is remarkable and is highly significant (p.< .001). Half of the working-class Glasgow adults do not use very even once. There is also a significant age difference. The adolescents use very (p.< .001) significantly less than the adults (both groups of adolescents barely use very). There is no significant gender difference, though the middle-class women use very slightly more frequently (4.86) than the middle-class men (3.66). I. 1. a. ii. An example of an unambiguous form that is consistent in function but varies in meaning is quite. The social class difference among the Glasgow adults in the use of quite is also significant (p.< .002), as can also be seen in Table 4. Adolescents also use less quite (p.< .05). Quite can be used as a Maximizer (Quirk, Greenbaum, Leech & Svartvik 1985: 590), for example, ‘but I think clothes-wise we're quite different,’ where quite has an emphatic
CAN WE FIND MORE VARIETY IN VARIATION?
177
function. It can also be used as a Downtoner (Quirk et al. 1985: 597–578), for example, ‘it is actually quite nice,’ where the function of quite is as a hedge. Identifying tokens of the uses of quite therefore requires taking into consideration the context in which it occurs. The middle-class speakers appear to use quite more frequently in its emphatic function (67%) than in its hedging function (33%). For the working-class speakers the difference is less: 56% emphatic, 44% hedging. However, the middle-class speakers use quite with an overall frequency of 3.64 per thousand words compared with the working-class frequency of 1.19. The frequency with which the middle-class speakers use quite in its emphatic function is 2.42 per thousand words compared with the working-class frequency of 0.66. In the hedging function the frequencies are: middle-class 1.2, working-class 0.52. The middle-class thus use quite twice as often as the working-class speakers in a hedging function and almost four times as often in the emphatic function. The figures are shown in Table 5. Working-class Middle-class Freq. Freq. Emphatic function 0.66 2.42 (p.< .01) Hedging function 0.52 1.2 (p.< .05) Table 5: Relative frequency of two functions of quite in Glasgow adults (freq. = per 1,000 words).
I. 1. b. An example of an unambiguous form that varies in both function and meaning is the particle oh. Oh can be used by itself as an acknowledgement marker or as part of an agreement marker, such as oh right, oh yeah. It can also be used as part of an exclamation of emotion, usually dismay, as in oh no, oh God, or oh shit. It can also be used to introduce a question that is often a kind of other repair, e.g., oh were you there? oh it goes on without you even touching it? Oh can also be used to introduce statements, often expressions of personal feeling, oh I hate him, oh I’m so tired. Finally, oh may occur in quoted dialogue to mark a change of speaker: (1)
(Glasgow working-class woman) a. I says to them “Your new phone number’s in” b. “Oh! Was it the day?”
The speaker is reporting an exchange. Her remark is quoted in (1a) and the response in (1b). This is part of a narrative concerning her aunt and uncle, the referent of them in (1a). One of them is reported as having given the response in (1b). This is an example of a zero quotative, where the change of speaker is indicated by intonation and the use of the discourse marker oh. The
178
RONALD MACAULAY
Working-class Middle-class Freq. Freq. Adults 5.61 7.23 (n.s.) Adolescents 3.18 6.61 (p.< .05) Table 6: Relative frequency of oh in Glasgow (freq. = per 1,000 words).
tabulation of oh forms consequently requires an examination of the context in which the token occurs to distinguish the different uses. Table 6 shows that there are social class differences. The middle-class adults use oh with a frequency of 7.23 per thousand words, the working-class adults with a frequency of 5.61, but this difference is not statistically significant. The difference is greater in the adolescents with the middle-class adolescents using oh with a frequency of 6.61, compared with a frequency of 3.18 among the working-class adolescents, and this is significant (p.< .05). The adults use oh with a higher frequency (6.26) than the adolescents (4.92) but this is not significant. More importantly, the females use oh with an overall frequency of 7.31 compared with 3.62 for males, and that this is true of both adults (women 7.62 vs. men 4.12; girls 6.64 vs. boys 2.74). This difference is statistically significant (p.< .005) as can be seen in Table 7. Males Females Freq. Freq. Adults 4.12 7.62 (p.< .01) Adolescents 2.74 6.64 (p.< .005) Table 7: Relative frequency of oh Glasgow (freq. = per 1,000 words).
As regards different uses of oh, all the adults use oh most frequently when signalling agreement, with the exception of the working-class women who use oh most often in quoted dialogue. Of the other differences in the use of oh the most interesting is the frequency with which middle-class women (22%) and middle-class girls (31%) use it for the expression of emotion. I. 2. Examples of minimally variant forms include the personal pronouns. It is reasonable to treat as one lexical item the forms I, me, mine, my, and myself. Since the number of variant forms is small it is quite easy to identify all the forms of the pronoun used by a speaker. There is one personal pronoun that is more complex and that is it. This is because it has other functions than that of an anaphoric pronoun: Identifying only the anaphoric uses of it requires more detailed study of the linguistic context than for the other personal pronouns.
CAN WE FIND MORE VARIETY IN VARIATION?
179
Table 8 shows that the adolescents use more pronouns that the adults, and the females more than the males. The working-class use more pronouns than the middle-class speakers, but this difference just fails to meet statistical significance. Freq. Freq. Adults 156 Adolescents 178 Middle-class 159 Working-class 173 Males 148 Females 178 Table 8: Relative frequency of pronouns in Glasgow (freq. = per 1,000 words).
(p.< .01) (n.s.) (p.< .001)
Table 9 shows that the women and girls use the pronoun she significantly more than the men and boys. There are no significant differences in the use of other pronouns, except that the females use the pronoun I more frequently (p.<.05), but this effect is the result of more frequent use by the adolescent girls. Freq. Freq. Men 5.2 Women 20.2 (p.< .001) Boys 6.6 Girls 30.6 (p.< .001) Table 9: Relative frequency of the pronoun she in Glasgow (freq. = per 1,000 words).
There are other features that correlate with this gender difference in pronoun use (Macaulay, in prep.). II. Many discourse markers are ambiguous. For example, the discourse use of well has been studied by Schiffrin (1987), Schourup (1985), and Svartvik (1980). However, the form well can also be an adverb (‘he did it well’) or a noun (‘he went to the well too often’). To investigate the use of the discourse marker well it is necessary to separate out the other uses. A more complex example of an ambiguous discourse marker is you know (Östman 1981; Schourup 1985; Holmes 1986; Schiffrin 1987; Macaulay 2002). It is complex because it is necessary to distinguish the use of you know as a discourse marker (as in ‘I could see you know the hunted look on his face’) from its basic meaning (as in ‘not what you know who you knew’). In the first case, you know is not an integral part of the syntax and it is in the second. It is usually fairly easy to distinguish these uses but some investigators have not always done so consistently (Macaulay 2002). So in order to study variation in the use of a discourse marker such as you know it is necessary to look at each
180
RONALD MACAULAY
Freq. Freq. Middle-class adults 4.46 Working-class adults 5.19 (n.s.) Women 4.65 Men 5.28 (n.s.) Adolescents 1.93 Adults 4.89 (p.< .003) Table 10: Relative frequency of the discourse marker well in Glasgow (freq. = per 1,000 words).
example in context. The figures for the distribution of you know in the Glasgow conversations can be seen in Macaulay (2002). Here the category of ambiguous features is illustrated by the figures for well. Table 10 shows that there are no significant social class or gender differences in the use of well, though there is a highly significant age difference. At this point it may be worth emphasizing that negative results, i.e., those that show no difference between groups in their use of a discourse feature, may be just as interesting and important as those showing differences. There is a natural bias in favour of reporting differences (Brenneis & Macaulay 1996: 75) so it is important to draw attention also to those areas in which speakers do not differ. III. Complex variables, such as the use of passive voice, questions, or quoted dialogue, obviously require careful examination of the context and thus extraction becomes even more time-consuming. Identifying narrative sections also requires human judgment. Tables 11 and 12 show the proportion of narrative in the Glasgow conversations. The figures, however, mask some significant differences. There is no overall social class difference but the females have a significantly higher proportion of narratives (p.< .01). The social class picture is confounded by a gender difference. The working-class women show a proportion of narrative Middle-class Working-class Adolescents 27% 40% Adults 33% 33% Table 11: Proportion of narrative in Glasgow conversations.
Middle-class Working-class Girls 28% 60% Boys 27% 5% Women 32% 44% Men 35% 13% Table 12: Proportion of narrative in Glasgow conversation.
CAN WE FIND MORE VARIETY IN VARIATION?
181
that is three times higher than that of the working-class men (p.<.01), and among the adolescents the working-class girls devote twelve times the amount that the working-class boys do to presenting narratives (p.< .001). However, the social class difference among the females just fails to reach significance (p.= .055) although the working-class women and girls have a higher proportion of narratives. Among the males the situation is reversed with the middle-class men and boys producing more narratives (p.< .001). It is the women who include the most dialogue in their stories (p.< .05), at roughly a quarter of the narratives, while the men’s stories only contain less than a tenth of dialogue. 3.
Conclusion The examples I have given are only a sample of the variables that I have investigated in the Glasgow conversations, and I have chosen them to illustrate the method rather than for the reason that they are the most important. I am completing a full-length study of all the variables and there I discuss the implications of the variation in the use of a range of discourse features (Macaulay, in prep.). Here I simply want to make a plea for more research into variation in the use of discourse features. One obvious reason is that we do not know which features are likely to vary consistently until we look for them. For example, as far as I know, nobody had predicted the very significant social class difference in the use of very before I discovered it in an earlier study. 1 You do not need an undergraduate degree to learn how to use the word very and nobody who deplores the shortcomings of working-class speech ever cites their failure to use the word very. On the other hand, there are those who believe that working-class speakers are more likely to use the often-stigmatized discourse marker you know, although the evidence may not be there. Another plea I wish to make is for more research into discourse variation that takes into account more than a single extra-linguistic factor. In the investigation so far, a total of 42 discourse features in the Glasgow conversations have been analyzed using the Mann-Whitney nonparametric test. These tests produced 46 statistically significant differences. Of these 46, 10 refer to social class differences, 16 to gender differences, and 20 to age differences. In other words, the statistically significant variation in the Glasgow conversations can be ranked as follows:
1
Kroch (1995), in his study of upper class Philadelphia speech, found that upper class men were more likely than upper class women or upper middle-class men to use ‘intensifying adverbs’ (e.g., very, extremely) though the results were not statistically significant.
182
RONALD MACAULAY
AGE DIFFERENCES > GENDER DIFFERENCES > SOCIAL CLASS DIFFERENCES Note, please note, this does NOT mean that the age or gender differences are necessarily more important than the social class differences. This pattern underlines the necessity of looking at more than one extralinguistic dimension and the danger of taking one subset (e.g. adolescent boys) and making generalizations about social class, as Bernstein did, for example. Similarly, drawing conclusions about gender differences without taking age and social class into consideration may give a misleading picture. This might appear self-evident but there are many studies of variation in discourse features that do not make this clear. For example, the London-Lund corpus has provided scholars with a useful data set of transcribed speech, but the speakers are educated, middle-class adults with a predominance of males. It is possible to look at gender differences in the LLC transcripts, but it is tedious to separate out the contributions and few have done so. My third plea is for replication. Because investigation of discourse variation must examine samples of talk in action, the use of a specific feature is locally determined, and thus any conclusion from a specific data set may not generalize to other situations. For this reason, any conclusions drawn from a single study may give an unreliable indication of a more widespread difference. The American psychologist, Donald Campbell points out the need for replication: Because we social scientists have less ability [than physical scientists] to achieve experimental isolation, because we have good reason to expect our treatment effects to interact significantly with a wide variety of social factors many of which we have not yet mapped, we have much greater needs for replication experiments than do the physical sciences. [Campbell 1969: 427–428]
In this paper I have shown one way to investigate a wide range variation in language. Go thou and do likewise. Please. References Baker, M., G. Francis, and E. Tognini-Bognelli, eds. 1993. Text and Technology: In Honour of John Sinclair. Amsterdam: John Benjamins. Bernstein, B. 1962. “Social class, linguistic codes, and grammatical elements”. Language and Speech 5.31–46. (Reprinted in Class, codes and control, vol.1, 95–117. London: Routledge and Kegan Paul, 1971.) Brenneis, D. and R. Macaulay, eds. 1996. The Matrix of Language: Contemporary Linguistic Anthropology. Boulder, CO: Westview Press.
CAN WE FIND MORE VARIETY IN VARIATION?
183
Campbell, D. T. 1969. “Reforms as experiments”. American Psychologist 25.409–429. Cheshire, J. 1982. Variation in an English Dialect. Cambridge: Cambridge University Press. Coupland, N. 1988. Dialect in Use: Sociolinguistic Variation in Cardiff English. Cardiff: University of Wales Press. ----------. 2001. “Language, context and the relational self: Retheorizing dialect style in sociolinguistics”. Style and Sociolinguistic Variation ed. by P. Eckert and J. Rickford, 185–203. Cambridge: Cambridge University Press. Docherty, G. J, P. Foulkes, J. Milroy, L. Milroy, and D. Walshaw. 1997. “Descriptive adequacy in phonology: A variationist perspective”. Journal of Linguistics 33.275–310. Eckert, P. 2000. Linguistic Variation as Social Practice: The Linguistic Construction of Identity in Belten High. Oxford: Blackwell. Edwards, J. 1987. “Elaborated and restricted codes”. Sociolinguistics, vol. 1, ed. by U. Ammon, N. Dittmar, and K. J. Mattheier, 374–378. Berlin: de Gruyter. Fasold, R. 1972. Tense Marking in Black English. Washington, D.C.: Center for Applied Linguistics. Feagin, C. 1979. Variation and Change in Alabama English: A Sociolinguistic Study of the White Community. Washington: Georgetown University Press. Foulkes, P. and G. Docherty, eds. 1999. Urban Voices: Variation and Change in British Accents. London: Arnold. Holmes, J. 1986. “Functions of you know in women’s and men’s speech”. Language in Society 15.1–22. Kroch, A. 1995. “Dialect and style in the speech of upper class Philadelphia”. Towards a Social Science of Language: Papers in Honor of William Labov, vol. 1, ed. by G. R. Guy, C. Feagin, D. Schiffrin, and J. Baugh, 23–45. Amsterdam & Philadelphia: John Benjamins. Labov, W. 1963. “The social motivation of a sound change”. Word 19.273– 309. ----------. 1966. The Social Stratification of English in New York City. Washington, D.C.: Center for Applied Linguistics. ----------. 2001. Principles of Linguistic Change: Social Factors. Oxford: Blackwell. Macaulay, R. 1977. Language, Social Class, and Education: A Glasgow Study. Edinburgh: Edinburgh University Press. ----------. 1991. Locating Dialect in Discourse: The Language of Honest Men and Bonnie Lasses in Ayr. New York: Oxford University Press.
184
RONALD MACAULAY
----------. 2002. “You know, it depends”. Journal of Pragmatics 34.749–767. ----------. In prep. Discourse sociolinguistics. McCafferty, K. 2001. Ethnicity and Language Change: English in (London-) Derry, Northern Ireland. Amsterdam & Philadelphia: John Benjamins. Milroy, J. 1979. Review of Macaulay 1977. Language in Society 8.88–06. Milroy, L. 1980. Language and Social Networks. Oxford: Blackwell. Östman, J.-O. 1981. You Know: A Discourse-functional Approach. (= Pragmatics and Beyond II: 7.) Amsterdam: John Benjamins. Quirk, R., S. Greenbaum, G. Leech, and J. Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Schiffrin, D. 1987. Discourse Markers. Cambridge: Cambridge University Press. Schourup, L. C. 1985. Common Discourse Particles in English Conversation: Like, Well, Y’know. New York: Garland. Stuart-Smith, J. 1999. “Glasgow”. Urban Voices: Variation and Change in British Accents ed. by P. Foulkes and G. Docherty, 203–222. London: Arnold. Svartvik, J. 1980. “Well in conversation”. Studies in English Linguistics for Randolph Quirk ed. by S. Greenbaum, G. Leech, and J. Svartvik, 167– 177. London: Longman. Trudgill, P. 1974. The Social Differentiation of English in Norwich. Cambridge: Cambridge University Press. Wolfram, W. 1969. A Sociolinguistic Description of Detroit Negro Speech. Washington, D.C.: Center for Applied Linguistics.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH A CROSS-SEX ACOUSTIC STUDY †
VINCENT J. VAN HEUVEN , RENÉE VAN BEZOOIJEN§ & LOULOU EDELMAN§ † ULCL, Universiteit Leiden & CLS, Nijmegen University§ 1.
Introduction In the last decade a new variety of Modern Dutch has emerged, which was first noted and commented on by the dialectologist Reker in 1993 in a press release by the regional broadcasting station Radio-Noord (for details see Stroop 1998: 13, 107). Taking a cue from Reker’s observation, the emerging variety has been closely monitored by Stroop, who christened it Polder Dutch. The variety was elegantly described and placed in a wider sociolinguistic perspective in a popular-scientific brochure by Stroop (1998). More recently, the variety got an entire website devoted to it (www.hum.uva.nl/poldernederlands), with very thorough documentation — both in Dutch and in English — on the phenomenon. The most elaborate description in English is Stroop’s (1999) lecture presented at the New Methods in Dialectology Conference X in Saint John’s (Canada), the full version of which has been published on the above website. 1.1
The phonetics of the sound change The new variety differs from the Standard language only in its phonetics. Stroop presents the change as a chain shift, whereby the low-mid diphthongs /i, œy, u/ are lowered. As a result, the onset of the low-mid diphthongs assumes a position very close to open /a/, so that the three diphthongs are no longer clearly differentiated in their onsets. However, the end points of the diphthongs — which may possibly be lowered as well — still differentiate adequately between the front unrounded /i/, the front rounded /œy/, and the back rounded /u/. As a consequence of the lowering of the low-mid diphthongs the tense high-mid vowels /e:, ø:, o:/, which have slight diphthongization in the standard language (see Nooteboom and Cohen 1976: 62–63), are also somewhat lowered and more noticeably diph-
186
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
thongized. 1 The entire chain is illustrated in Figure 1, which has been copied from Stroop (1999). In the figure it is shown how the onsets of both the highmid vowels and of the low-mid diphthongs have shifted to more open positions. 2 It is unclear at this stage whether the shift should be analyzed as a push chain or a drag chain. From the informal description provided here (and by Stroop) it transpires that we conceive of the shift as a drag chain, i.e., the low-mid diphthongs move their onsets towards /a/, after which the high-mid vowels are drawn into the void that is being created in the lower mid area of the vowel space. However, the alternative is also feasible: the chain may have started off with the high-mid monophthongs getting diphthongized, thereby assuming more open onsets, which in turn push the low-mid diphthongs further towards open /a/. Historically, of course, it would seem more reasonable to analyze the change as a drag shift, since the change of long /i/ and /u/ all the way to /ai/ and /u/, respectively, has taken place in English and German, as well.
e:
ø: i
o:
œy
ai
ay
u
au
Figure 1: Shift of long vowels and diphthongs in avant-garde Dutch (after Stroop 1999: Figure 4).
1
The Dutch tense mid vowels are phonologically analyzed as monophthongs. As has often been remarked, however, these vowels feature slight diphthongization towards a closer end point in all contexts except when followed by tautosyllabic /r/, in which latter case the vowel offset is slightly raised and the offglide is towards the centre of the vowel space (Collins and Mees 1984: 115–116). Although the closing diphthongization is not essential for their perceptual identification, it does contribute to the naturalness of these vowels (Cohen, Slis and ‘t Hart 1963; Nooteboom and Cohen 1976: 63) 2 There are also subtle shifts within the set of short (lax) vowels, most noticeably in // which is raised and fronted towards [], and // which is raised towards [] (especially before /l/). Whereas the general drift is towards more open articulations for the tense (non-high and nonlow) vowels and diphthongs, the opposite direction is found for the lax vowels.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
187
It remains unclear from the descriptions provided whether the degree of diphthongization is affected by the sound change. If it is only the onset of the diphthongs that is more open, and the end point remains stable, then the strength of diphthongization (the size of the diphthong trajectory) should have increased. However, if the onset and the endpoint have been lowered together, then the strength of diphthongization should have remained the same. Interestingly, Stroop (1998) transcribes the new variant of the diphthong /i/ as aai, which trigraph is the conventional orthographic representation of the long open vowel /a:/ followed by a consonant /j/, as in haai /ha:j/ ‘shark’ or maait /ma:jt/ ‘mows’. The rhymes of these words are appreciably (ca. 50 to 90 ms) longer than their counterparts in hei /hi/ ‘heather’ and meid /mit/ ‘maid’. 3 Also, the vowel in haai and maait remains stationary for a relatively long time, and then abruptly glides off toward /i/, whilst the vowel quality in hei and meid changes from the beginning onwards (Nooteboom and Cohen 1976, Collier and ‘t Hart 1983, Peeters 1991, Rietveld and van Heuven 2001). Referring to Labov (1994) Stroop claims that the lowering of /i/ — and the lowering of diphthongs in general — reflects a natural tendency. In fact, low-mid diphthongs are rare in the world’s languages. Cognates of Dutch /i/ in the neighboring languages English and German are fully open diphthongs, e.g., English wine /wain/, German Wein /wain/. 4 The Dutch mid-open diphthongs, especially /i/ and /œy/ are notoriously difficult for foreign learners of the language. The fully open alternative would be easier to produce on the strength of the argument that the speaker just has to open his mouth as wide as he can to get the vowel right, while more intricate articulatory control would be required for the low-mid variety. If it is true that the more open variety of /i/ would also have a larger change in vowel quality from onset to offset, there is the added advantage that the diphthongal nature of the vowel would stand out more clearly, reducing potential perceptual confusion with tense /e:/, which is also slightly diphthongal (see note 1).
3
No full-scale studies are available on these durations. The differences given here are visible in spectrograms provided by Cohen and Nooteboom (1976: 62) and by Rietveld and van Heuven (2001: 153-154). 4 Indeed, Stroop (1999) — in jest — claims that the Great Vowel Shift, which took place in English at the end of the Middle Ages, is currently taking place in Dutch. Probably it would be more accurate to say that a large but incomplete vowel shift did take place in Dutch, changing the original West Germanic long high vowels /i/ and /u/ to low-mid diphthongs. The sound change that we are now observing, is merely the delayed final stage of the shift.
188
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
In this paper we concentrate on the phonetic details of the diphthong /i/ in Polder Dutch. This sound has been advanced as the exponent of the new variety, as is evidenced by the following quotation from Stroop (1998: 25, our translation) 5 : “The most conspicuous feature of Polder Dutch is the pronunciation of the diphthong /i/, which is spelled as either ei or ij.” The first aim of this study, then, is to clarify the phonetics of the sound change in so far as it relates to the pronunciation of the diphthong /i/. Moreover, rather than using the traditional impressionistic-phonetic approach, i.e. listening and transcription, we wish to investigate the issue using acoustic measures of greater mouth opening and/or stronger diphthongization and lengthening. 1.2
The sociolinguistics of the change Stroop (1998) claims that the change is typical of (relatively) young, highly educated, progressive Dutch women, who wish to make a statement through speech that they are unconventional and emancipated. The variety is often found among women with high-prestige social positions such as authors, actors, film producers, artists, left-wing politicians (either local or national), high-ranking academics, and pop-singers. It is for this reason that we prefer the use of the term ‘avant-garde’ Dutch for the new variety, rather than Polder Dutch, which in hindsight seems a misnomer. It should be pointed out here that the avant-garde variety is found throughout the country. According to Stroop it is not based on any existing dialect of Dutch, and it has no documented geographic epicenter. If this claim is true, avant-garde Dutch would truly qualify as a sociolect rather than a dialect or regiolect. Recently, experimental evidence has come available corroborating Stroop’s claims. Van Bezooijen and van den Berg (2001) presented excerpts of recordings made by twelve female Dutch speakers. Three speakers spoke the avant-garde variety, three spoke Dutch with a typically Amsterdam accent, three more spoke some different regional accent associated with the West of the Netherlands, and three women spoke the traditional (ABN) version of
5
Indeed, whenever Stroop presents illustrations of the new variety on the website or during lectures, it is the diphthong /i/ that is used. In fact, the website even invites the public at large to submit a recording of their own production of /i/ — rather than some other vowel or diphthong in the language — in order to test whether the particular speaker has already fallen victim to the Polder Dutch sound change.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
189
Standard Dutch. 6 All women were comparable in terms of age (between 25 and 40 years of age) and socio-economic status (highly educated). The materials were presented to eight groups of 20 listeners, stratified by sex (male vs. female), age (young/early twenties vs. old/late forties) and regional origin (Western vs. Eastern part of the Netherlands). Listeners rated the speech samples along several 7-point scales including broad vs. standard, modern vs. old fashioned, sloppy vs. polished, ugly vs. beautiful, and ‘my cup of tea’ vs. ‘not my cup of tea’. The avant-garde variety was judged to be more modern than the other varieties by all groups of listeners, indicating that the novelty of the avant-garde type is appreciated by all listeners, irrespective of sex and age. The (high-prestige) traditional Standard variety was generally rated most favorably by listener groups on all remaining evaluation scales whereas the regional Western and Amsterdam accents were rated least favorably. The ratings for the avant-garde variety, however, differed widely depending on the listener group. Crucially, the young female listeners considered the avant-garde variety (almost) as standard, normal, polished, and beautiful as the traditional Standard language. Finally, the young females indicated much more strongly than the other three groups that the avant-garde variety was something they identified with (‘my cup of tea’) while at the same time they identified less than the other groups with the traditional Standard variety. These results indicate that the avant-garde variety does indeed belong to the social class of young, highly-educated, independent women. Although the change is led by women, Stroop predicts that men will follow suit. 7 It has often been observed, in this connection, that women tend to adhere more strictly to established (pronunciation and other) norms than men. The prima facie contradiction with the leading role of women in the avant-garde Dutch vowel shift is resolved if we make a distinction between what Labov calls ‘changes from above’ and ‘changes from below’ (see Labov 2001: 366–382 for an elaborate treatment of the issue). According to Labov …women deviate less than men from linguistic norms when the deviations are overtly proscribed, but more than men when the deviations are not proscribed. (2001: 367)
6 The traditional version of Standard Dutch is commonly referred to as ABN (Algemeen Beschaafd Nederlands, literally ‘General Civilized Dutch’). Its definition and social status is roughly comparable to that of RP (Received Pronunciation) in British English. 7 In fact, Stroop produces some examples of male pop-singers who feature the open diphthongs, specifically /i/, even today.
190
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
In the case of the avant-garde variety the deviations from the traditional standard are not overtly proscribed, i.e., they are not imposed ‘from above’ but develop spontaneously and surreptitiously ‘from below’. The speakers of the new variety are not aware of the fact that their pronunciation of the language differs in any linguistically relevant way from the standard variety. When accused of using the avant-garde pronunciation their first reaction is denial. The sound change, therefore, has all the characteristics of a change from below (see also van Bezooijen, Kroezen and van den Berg 2002). Such sound changes are commonly initiated by women, and the avant-garde Dutch variety is no exception to the general rule. The second aim of the present study is to test the claim that the avantgarde variety of Dutch is indeed more widespread among women than among men of similar social status and age. 2. 2.1
Experimental approach Objective measurement of vowel quality There is agreement among experimental phoneticians and sociolinguists that vowel quality, and change of vowel quality in diphthongs, can be quantified with adequate precision and validity by measuring the center frequencies of the lower resonances in the acoustic signal. Specifically the center frequency of the lowest resonance of the vocal tract, called first formant frequency or F1, corresponds closely to the articulatory and/or perceptual dimension of vowel height (high vs. low vowels, or close vs. open vowels). For an average male voice, the F1 values would range between 200 hertz (Hz) for a high vowel /i/ to some 800 Hz for a low vowel /a/. The second formant frequency (or F2) reflects the place of maximal constriction during the production of the vowel, i.e., the front vs. back dimension, such that the F2 values range from roughly 2200 Hz for front /i/ down to some 800 Hz for back /u/. The relationship between the formant frequencies and the corresponding perceived vowel quality is not linear. For instance, a change in F1 from 200 to 300 Hz brings about a much larger change in vowel quality (height) than a numerically equal change from 700 to 800 Hz. Over the past decades experimental phoneticians and psycho-physicists have developed an empirical formula that adequately maps the differences in hertz-values onto the perceptual vowel quality (or timbre) domain, using the so-called Bark transformation (for a summary of positions, see Hayward 2000). Using this transformation, the perceptual distance between any two vowel qualities can be computed from acoustic measurements.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
191
We used the Bark formula as advocated by Traunmüller (1990): Bark = [(26.81 × F) / (1960 + F)] – 0.53 where F represents the measured formant frequency in hertz. The acoustic characterization of diphthongs is relatively easy, once adequate measures are available to capture vowel quality in a perceptually realistic fashion. All that is needed, then, is a comparison of the formant values computed at the onset and at the offset of the diphthong. Given that the first and last portions of any vowel, monophthongs and diphthongs alike, are strongly influenced by (the articulation place of) the neighboring consonants, it is customary to sample the formant values for the starting point of the diphthong at one-quarter of the time-course of the diphthong, and to measure the formants for the endpoint of the diphthong at 75% of its duration. The degree of diphthongization is then expressed in a straightforward fashion by computing the distance between the onset (25%) and the offset (75%) vowel quality. In terms of the traditional vowel diagrams used by impressionistic linguists and phoneticians, this procedure is the equivalent of measuring the length of the arrow that represents the diphthong. 2.2
The problem of vowel normalization Unfortunately, formant values measured for the same vowel differ when the vowels are produced by different individuals. The larger the differences between two speakers in shape and size of the cavities in their vocal tracts, the larger the differences in formant values of perceptually identical vowel tokens. Given that the vocal tracts of women are some 15 percent smaller than those of men, comparison of formant values is especially hazardous across speakers of the opposed sex. Numerous attempts have been made, therefore, to factor out the speaker-individual component from the raw formant values such that phonetically identical vowels spoken by different individuals would come out with the same values. None of these vowel normalization procedures have proven fully satisfactory (Adank, van Heuven and van Hout 1999; Labov 2001: 157–164; Rietveld and van Heuven 2001). Broadly, two approaches to the normalization problem have been taken in the literature (for a detailed discussion of the issue of vowel normalization, see also Neary 1989). The first approach, called intrinsic normalization, tries to solve the problem by considering only information that is contained in the single vowel token under consideration, typically by computing ratios
192
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
between pairs of formant values such as F1/F0, F2/F1. 8 The alternative, extrinsic normalization, looks at tokens of all the vowels in the phoneme inventory of a speaker and expresses the position of one vowel token relative to the other tokens within the individual speaker’s vowel space. Obviously, extrinsic normalization procedures will be more successful but they achieve their goal by drawing on much more information than the intrinsic procedures. There is ample evidence that human listeners apply some sort of intrinsic normalization, and do not require (much) extrinsic normalization. For the purpose of the present study, however, we will adopt a hybrid normalization procedure, which combines virtues of both intrinsic and extrinsic normalization. The intrinsic part of our normalization does not involve anything more than just transforming the measured formant values from hertz into Bark. The extrinsic part of the procedure is basically what has been called ‘end-point normalization’ in the literature. A vowel token is scaled according to its relative position between the extreme (lowest and highest) values for F1 and F2 found for the individual speaker. 9 In the present problem, we will not need to include the full vowel system of the speakers in the analysis. Since the study is limited to the sound change in /i/ — a front, unrounded vowel — we only require reference vowels that allow us to determine the individual implementation of the front region of the speaker’s vowel space. All that is required, therefore, is a reliable estimation of the speaker’s /i/ (maximally high front vowel) and /a/ (maximally open front vowel). 10 We make the explicit assumption that the point vowels /i/ and /a/ do not participate in the sound change in progress that affects the Dutch mid vowels. We are probably correct in making this assumption as Stroop’s vowel diagram does not indicate any involvement of the point vowels /i, a, u/ (1998: 28, see also our Figure 1). Our operationalization of openness of a vowel (across speakers and across sex) will be to express the onset of a vowel relatively to the entire close-open dimension of the vowel space, or — in impressionistic-phonetic 8 When formant values are rescaled to Bark, the numerical difference (F1–F2; F2–F1, etc.) is preferred over the ratio. 9 Gerstman (1968) by convention gives the extremes (lowest and highest) for any speaker, whether male or female, arbitrary values of 250 and 750 Hz for F1, and 850 and 2250 Hz for F2. These extreme formant values are realistic averages for male speakers; female values are rescaled so as to fit the male formant ranges. 10 Dutch tense /a/ is quite close to cardinal vowel 4, which phonetically qualifies as a front vowel. Interestingly, tense /a/ patterns with the back vowels in the phonology of Dutch, and is normally given the feature specification [+back], cf. Booij (1995). This analysis does not invalidate the adoption of tense /a/ as the optimal reference vowel for our study.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
193
terms — as the relative position along the front edge of the cardinal vowel diagram. Given that we can measure the degree of vowel height and vowel backness, we will express the position of a vowel token as a percentage of the full length of the front edge of the diagram, i.e. as a percentage of the distance between the /i/ (maximally close vowel) and the /a/ (maximally open vowel) of the speaker. Accordingly, we will not attempt full-scale vowel normalization, but limit the problem to translating the formant measurements into an index (a percentage). 2.3
Speakers and tasks The second basic point to consider is the sampling of the speakers. Remember that we wish to test the hypothesis that women lead the sound change. It seems imperative, therefore, that we should compare groups of male and female speakers that are equivalent in all sociolinguistically relevant aspects, such as socio-economic status and age. The speakers should not be aware of the fact that their speech production is being recorded for linguistic analysis, and their speech should be non-scripted, i.e., unpremeditated and spontaneously produced rather than rehearsed or — even worse — read out from paper. To aggravate matters, the type of speaker we were targeting is not easily accessible. These are typically well-known public figures, celebrities who will not be persuaded to participate in a scientific study. As a feasible alternative we decided to record a televised series of weekly talk shows featuring precisely the type of speakers that we were looking for. The particular talk show, Het Blauwe Licht (‘The Blue Light’), was produced by the ‘high-brow’ VPRO television network in the Netherlands. In each show two guests discussed recent television programs, press photos and newspaper articles. 11 3. 3.1
Method Materials From the winter season of 1998/99 onwards, the first 16 male and 16 female Dutch-speaking guests who appeared in the television talk show Het Blauwe Licht, were recorded on video and audio tape. Recordings were made as the programs were broadcast by the national public television network. 12 11
Ms. Lies Kulsdom, producer of Het Blauwe Licht, characterized the guests as ‘people with considerable cultural payload such as film producers, authors and intellectuals, who are able to uncover the deeper layers of meaning in film footage and who are not afraid to voice their opinions’. 12 In fact a seventeenth female speaker had to be recorded in order to replace opera singer Roberta Alexander. Although Ms. Alexander has lived in The Netherlands for a long time
194
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
Only the audio recordings (Sony TC D5M cassette recorder) were used for the present study. The mean ages of the men and women were the same (47) but the women spanned a somewhat wider range than the men (ranges 28–64 and 32– 52, respectively). Per speaker some 6 minutes of spontaneous, non-rehearsed speech were recorded (see § 2.3 above). For each speaker 10 tokens of the target diphthong /i/ were selected from the recordings, along with 5 tokens of /i/ and 5 tokens of /a/ (see § 2.2). Tokens to be selected into the database preferably occurred before obstruents in stressed syllables of content words (multiple tokens of the same word by the same speaker were avoided). 3.2
Acoustic processing The audio recordings were digitally sampled (16 kHz, 16 bits) and transferred to computer disk. Using the Praat speech processing software (Boersma and Weenink 1996, Boersma and van Heuven 2001) the beginnings and end points of the target vowels were located in oscillographic and/or spectrographic displays. Formant tracks for the lowest four formants (F1 through F4) were then automatically computed using the Burg LPC algorithm implemented in Praat, and visually checked by superimposing the tracks on a wideband spectrogram. Whenever a mismatch between the tracks and the formant band in the spectrogram was detected, the model order of the LPCanalysis was changed ad hoc until a proper match was obtained between track and spectrogram. Once a satisfactory match was obtained, the values for F1 and F2 were automatically extracted at 25, 50, and 75% of the duration of the target vowel, as well as the vowel duration as such, and stored for off-line statistical processing. Figure 2 shows the measurement procedure adopted. The figure presents the wide-band spectrogram of a short utterance containing a target diphthong /i/ with the formant tracks drawn through the formants as white lines. The duration of the diphthong was measured to be 160 ms. Segmentation lines have been drawn at the onset and offset of the target diphthong (0 and 100% duration, respectively), as well as at one, two and three quarters of the diphthong’s duration (25, 50, and 75% duration, respectively). A read-out of the center frequencies of F1 and F2 as extracted at the three temporal measurement points are listed below the figure. The figure and speaks Dutch fluently and effortlessly, a (slight) American accent is still audible in her spoken Dutch. We felt that this warranted her exclusion, and subsequent replacement by a native speaker.
195
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
Frequency (KHz)
Amplitude
shows that the /i/ token is a true diphthong: F1 and F2 never remain stable but diverge immediately from the onset. Observe that F1 starts at a rather high frequency (over 800 Hz) and — except for the very initial portion of the vowel — steadily drops to lower values as time progresses. This, obviously, reflects the closing gesture made during the articulation of the diphthong. At the same time the F2 value steadily rises, thereby widening the gap between the first and second formant, a clear sign of a tongue shift toward a more fronted articulation.
5 4 3 2 1 0 0
25
50
75
100 0.16
0 Time (s)
F2 F1
25% 1950 793
50% 2051 829
75% 2185 723
Figure 2: Oscillogram (top panel), wide-band spectrogram (middle panel) and superimposed formant tracks for F1 and F2 of an utterance containing /i/, as spoken by a female speaker. Formant values (in hertz) extracted at 25, 50, and 75% of the duration of the diphthong (segmentation points indicated in bottom panel) are listed below the figure.
196 4. 4.1
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
Results General orientation In order to give the reader a feel of the type of data that we are dealing with, we will begin by presenting a few overviews of vowel data for individual speakers. Figure 3 represents the acoustic vowel space for one male and one female speaker, whose vowel tokens are dispersed in a perfectly regular fashion. The male and the female vowel spaces have been plotted in the same coordinate system so as to facilitate cross-sex comparison. Observe that vowel height ranges between 2 and 8 Bark for the male speaker and between 3 and 9 for the female. Similarly, the front-back values range between 10 and 14 Bark for the male as opposed to 12 and 16 Bark for the female. This is a direct consequence of the cross-sex difference in the size of the oral and pharyngeal cavities. Still, by virtue of the Bark transformation, equal distances across the male and the female vowel spaces are perceptually the same. The data typically show that the reference point vowel tokens are tightly clustered in the left-hand top corner for /i/ and the open-central area for /a/. There is more variability for the target diphthong /i/ for both speakers, possibly indicating within-speaker instability for this diphthong. Also, visual inspection reveals that the cloud of /i/ onsets for the male speaker finds itself roughly halfway between the /i/ and /a/ clusters. For the female speaker, however, the cloud of /i/ onsets seems to have dropped to a relatively lower position between the /i/ and /a/ reference clusters, leaving a wide gap between the /i/ and /i/ onset clusters. More problematic cases arise in Figure 4, where two more female speakers have been plotted as in Figure 3. Figure 4 illustrates two problems. Female speaker A.M.’s /i/ tokens (left-hand panel) are far from tightly clustered; in fact, one token seems completely off target, which, incidentally, is not a measurement error but due to extreme rounding and centralization of this /i/ token (which sounds more like /y/). Centralization and/or vowel reduction is also manifest in the /a/ reference tokens of female speaker A.L. (right-hand panel). Although we would expect the female /a/ tokens to have F1 values around 9 Bark — as we did indeed find for the female speakers D.L. (Figure 1) and A.M. (Figure 2 left) — A.L.’s /a/ tokens never extend below 8 Bark.
197
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
Male P.W. -2
-4
-4
-6 i
-8
F1 (Bark)
F1 (Bark)
Female D.L. -2
-6 i
-8
ei
-10 -16
ei
-10 -16
a
-14
-12
-10
-8
a
-14
F2 (Bark)
-12
-10
-8
F2 (Bark)
Figure 3: Acoustic vowel diagrams plotting five tokens of /i/, five tokens of /a/ and ten tokens of diphthong /i/ in the Bark-transformed F1 versus F2 plane, for one male (left) and one female (right) speaker. Formants for /i/ and /a/ tokens were measured at the temporal midpoint of the vowel; formants for /i/ onsets were extracted at 25% of the vowel duration.
Female A.L. -2
-4
-4
F1 (Bark)
F1 (Bark)
Female A.M. -2
-6 i
-8
-6 i
-8
ei
ei
-10 -16
a
-14
-12
F2 (Bark)
-10
-8
-10 -16
a
-14
-12
-10
-8
F2 (Bark)
Figure 4: Acoustic vowel diagrams for two more female speakers (cf. Figure 3).
4.2
A normalization procedure In view of the susceptibility of the reference point vowels to reduction (centralization) it seems unwise to adopt the centroid (center of gravity) of the /i/ and /a/ clusters as the reference values when defining the speakerindividual vowel height dimension. Rather we decided to select the single most extreme (i.e. front-most) token within the speaker’s /i/ cluster as the high-front endpoint of the dimension, and the most extreme (i.e. most open) /a/ token as the other endpoint. Consequently, the speaker’s /i/ token with the
198
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
highest F2 value and the /a/ token with the highest F1 value were adopted as the extremes of the speaker-individual vowel height dimension. This procedure allows us to express vowel height speaker-individually as a relative measure. The spectral distance between the extreme /i/ token and the extreme /a/ token is set at 100%, such that /i/ has 100% vowel height and /a/ 0%. When some /i/ onset finds itself exactly midway between the extreme /i/ and /a/ tokens, its relative height will come out as 50%. This measure is implemented by computing (a) the Euclidian distance between the reference endpoints, (b) the Euclidian distance between the /i/ onset and the /a/ reference value, and (c) taking the percentage of b relative to a. The smaller the percentage c, the lower the relative starting point of the diphthong. The distances (a) and (b) are computed as follows: (a) sqrt [ (Bk1i – Bk1a)2 + (Bk2i – Bk2a)2 ] (b) sqrt [ (Bk1 – Bk1a)2 + (Bk2 – Bk2a)2 ] where Bk1 and Bk2 are the Bark-transformed center frequencies of formants F1 and F2; the vowel subscripts identify the reference vowels /i/ and /a/ measured at their temporal midpoint, and the onset value of the diphthong /i/ measured at 25% of its duration. 13 By the same reasoning we define a relative spectral change measure so as to express the speaker-individual degree of diphthongization for the /i/ tokens. First we compute the Euclidian distance in the Bark-transformed F1 by F2 plane between onset (formant measurements at the 25% temporal point) and offset (measurements at the 75% point) and then take this distance as a percentage of the total distance between extreme /i/ and /a/ (i.e. the distance computed under (a) above) of the speaker. A relative glide measure of 25% would then indicate that the /i/ glide extends along one quarter of the entire front edge of the speaker’s vowel diagram. The smaller the percentage, the shorter the length of the arrow representing the diphthongal glide in the traditional impressionistic vowel diagram. 4.3
Effect of sex of speaker These speaker-normalized measures of (relative) vowel height of the /i/ onset and of the magnitude of the diphthongization are shown in Figures 5 and 6, respectively. In these figures the values have been plotted separately for the male and female speakers, such that the speakers are ordered from left In the formulae for (a) and (b) ‘sqrt [...]’ is an instruction to take the square root of whatever is listed between the brackets.
13
199
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
to right in ascending order of conservatism in both figures, i.e., speakers with the most open onsets (in Figure 5) and the longest diphthongal glides (in Figure 6), indicative of the avant-garde speaking style appear to the left, and those with closer onsets and smaller diphthongal glides appear towards the right of the figures. It is obvious from Figure 5 that the female speakers, on the whole, have lower /i/ onsets than the males. There are one man and one woman with an extremely open /i/ onset of 20% vowel height. It seems that the change from [i] to [ai] has been completed for these two speakers. At the conservative end of the scale, there is one woman with a higher (i.e. more conservative) /i/ onset than the most conservative of the male speakers. For the 2 × 14 remaining speakers the women consistently lead in the change from [i] to [ai]. The effect of sex is significant by a paired t-test, t(15) = 5.46 (p < .001, one-tail).
Rel. onset height of /ei/ (%)
50
40
30 fem.
20
male
Speakers Figure 5: Relative vowel height of /i/ onset for 16 male and 16 female speakers. Individuals are ordered within sexes from left to right in ascending order of conservatism (as determined by the relative height of their /i/ onsets).
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
200
Rel. size of /ei/ trajectory (%)
40
30
20
10 fem.
0
male
Speakers Figure 6: Relative magnitude of diphthongization of /i/ for 16 male and 16 female speakers. Individuals are ordered within sexes from left to right in ascending order of conservatism (as determined by the magnitude of diphthongization).
Rel. size of /ei/ trajectory
40
30
20
10 male 0 10
fem. 20
30
40
50
Rel. onset height of /ei/ Figure 7: Relationship between normalized /i/ onset and magnitude of spectral change from onset to offset of /i/ for 16 male and 16 female speakers. Separate regression lines for male (open symbols) and female (closed symbols) speakers.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
201
Figure 6 reveals the same state of affairs with respect to the (normalized) magnitude of the spectral change in the diphthongs. Clearly, the women generally have a larger difference between onset and offset of the diphthongs than the men, t(15) = 2.93 (p = .005, one-tail). Figures 5 and 6 together indicate that the phonetics of the sound change in progress are best characterized as a combined lowering and magnification of the low-mid diphthong: the onset changes from low-mid to fully low but the offset remains more or less stationary, such that a larger spectral distance has to be covered between onset to offset, which would perceptually enhance the diphthongal nature of the vowel. Figure 7 plots the relationship between onset lowering and magnitude of spectral change in the /i/ diphthongs of the 16 men (in gray) and 16 women (in black). The figure, and subsequent statistical analysis, reveals that there is a moderate but significant correlation between onset lowering and strength of diphthongization for the female speakers, r = .481 (p = .030, onetail) but not for the males, r = .168 (ins.). This finding strengthens the claim that the sound change in progress is predominantly found with female speakers. 4.4
Perceptual validation One could still argue, of course, that the acoustic differences between the male and female diphthongs /i/ presented so far do not reflect a difference in phonetic vowel quality but are merely due to non-uniform scaling differences between the dimensions of the vocal tracts of men versus women. In order to bear out that the acoustic measures adopted truly reflect differences in auditory vowel quality, the first two authors independently judged the vowel height of the onsets of the 320 /i/ tokens in our dataset. Each token was played to the rater over loudspeakers in a quiet room as often as he or she deemed necessary in order to evaluate the token along a scale from 0 (maximally high onset) to 10 (maximally low onset). Tokens were made audible with a minimal acoustic context of 500 ms both preceding and following the target diphthong token. The scores for perceived onset height were averaged per speaker, for each rater separately, so that 32 pairs of onset height scores were obtained. Scoring onset height of diphthongs excerpted from spontaneous, continuous speech was done with substantial — although far from perfect — between-
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
202
Perceived openness of /ei/ (Z)
rater consistency, as is evidenced by the correlation coefficient that was found between the 32 pairs of scores, r = 0.795 (p < 0.001). 14 Figure 8 presents the perceptual ratings collapsed over both raters (after Z-normalization per rater) for the 16 male and 16 female speakers, ordered pair-wise from left to right in ascending order of conservatism (as was done in Figures 5 and 6). 15 4 3 2 1 0 -1
fem.
-2
male
Speakers Figure 8: Perceived openness of /i/ onset (mean judgments obtained from two raters, after Z-nominalization per rater) for 16 male and 16 female speakers. Individuals are ordered within sexes from left to right in ascending order of conservatism (as determined by their openness score).
14
As a precaution the first author judged the onset height of the 320 tokens a second time a few days later. His within-rater consistency was found to be considerable, with r = 0.820 (N = 32, p < 0.001). 15 The two raters interpreted the perceptual vowel height scale somewhat differently. The scores given by VH ranged between 3 and 9, whilst RB’s judgments ranged between 0 and 6. For the sake of visual comparability the ratings were standardized by Z-normalization before being plotted in Figures 8 and 9.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
203
It is quite clear from Figure 8 that the female speakers lead the male counterparts in the lowering of the /i/ onsets. When the individuals are ordered within the sexes from left right in ascending order of conservatism, i.e. in descending order of perceived openness of the /i/ onset, every female member of the pair has a more open onset than the male counterpart, t(15) = 6.61 (p < 0.001, one-tail). The perceived difference between the sexes is (even) larger, then, than appeared from the (normalized, relative) acoustic measures of onset height and size of the diphthongal glide. If it is true that the /i/ onset lowering is insipient only in the female part of the Dutch avant-garde, it makes sense that there should be a greater range of /i/ openness values among the female speakers than among the corresponding group of male speakers. This prediction is clearly borne out by the data in Figure 8. The female scores range between + 3.5 Z and – 1 Z, whereas the male scores are roughly between + 1 and – 1 Z. Figure 9 shows the relationship between perceived and measured /i/ onsets for rater VH (panel A) and for RB (panel B). In each panel the male and female speakers are represented by grey and black squares, respectively. Separate regression lines have been drawn through the male and the female data points.
Figure 9: Perceived /i/ onset height as a function of measured acoustic height of /i/ onset for raters VH (panel A) and RB (panel B). Regression lines have been drawn through the scatter clouds for the 16 male and 16 female speakers separately.
204
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
Both raters obtained moderate, but highly significant, correlations between their perceptual openness scores for the /i/ onsets and the normalised height index computed from the acoustic formant measurements, with r = – 0.686 (N= 32, p < 0.001) for RB and r = – 0.617 for VH (N = 32, p < 0.001). The correlation coefficients are negative since 0 % vowel height corresponds to the maximum openness score on the perceptual scale from 0 to 10. The correlation coefficients improve substantially, however, if only the ratings and acoustic indices obtained for the 16 female speakers are taken into account, with r = – 0.798 (N = 16, p < 0.001) for VH and r = – 0.786 (N = 32, p < 0.001) for RB. 16 Figure 9 illustrates clearly that the relationship between perception and acoustics is quite strong for the female speakers but much poorer for the males. Although the male speakers are concentrated towards the non-low extreme of the acoustic onset height dimension, there are a few men who have acoustically low /i/ onsets, but who are not perceived as having particularly open onsets. Closer inspection of the acoustic data (see also Figure 7) reveals that these men do not feature the stronger diphthongal glide that would be expected to accompany the low /i/ onsets. In fact, the male speaker with the perceptually most open /i/ onset has virtually no diphthongal glide at all. The women with extremely low /i/ onsets, however, do have the corresponding stronger glides. It would appear, therefore, that our perception of onset height in the diphthong /i/ is co-determined by the size of the closing gesture. In view of the above findings it is difficult to claim that our relative acoustic measure (even after normalization) corresponds in a straightforward fashion with the perceived vowel openness. It would appear that the acoustic measure either underestimates the perceived openness of vowels in female voices, or — more likely — that it overestimates the perceived openness of vowel produces by men. It is important, in this context, to note that our perception of vowel onset in closing diphthongs seems to have been influenced by the magnitude of the diphthongal glide. More specifically, it 16
In an earlier report (van Heuven, van Bezooijen and Edelman 2002) we also regressed the acoustically defined relative /i/ onset values against the perceptual measure for vowel opening that was reported earlier in Edelman (1999). Edelman reported a perceptual index (based on her narrow phonetic transcriptions of diphthongs) for the strength of the Polder Dutch impression that was made by (a random selection of) 13 out of the 32 speakers in the present study. The correlation between normalised acoustic height of the vowel onset and the perceptual strength of the avant-garde quality of these 13 speakers was considerable (r = .742). The correlation coefficient was not computed separately for the male and female data points, however.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
205
seems to be the case that a rather open /i/ onset is perceived as relatively close when the diphthongal glide (from [] to [] is (almost) absent. It is unclear at this stage how this effect — if it were found to be reproducible with other phonetically trained listeners — should be explained. One possibility would be that this is a low-level psychophysical effect that would apply whenever the openness of a vowel-like sound with an offglide has to be judged. It might be the case that the listener extrapolates a lower (i.e. more open) vowel onset from the off-sweep towards []. A similar effect has been reported by ‘t Hart (1969) for the /r/-colored tense high-mid vowels (see also note 1); ‘t Hart’s study was based on analytic listening to gated portions of vowel sounds. He observed that the /r/-colored vowels feature an offglide from the periphery of the vowel space towards the center (i.e. a centralizing diphthongal trajectory) together with a higher (i.e. closer) onset. The closer onsets for the tense high-mid vowels was contradicted by acoustic measurements by Koopmans-van Beinum (1969). We surmise — on the strength of our present findings — that both ‘t Hart and Koopmans-van Beinum may have been correct in their observations, each in their own domain: the closer onsets do not exist in acoustic reality but are perceived through an auditory illusion whereby the onset of a glide is traced back to a starting point that lies before the acoustic onset. The psychophysics of this phenomenon, if indeed it exists, is clearly in need of further research. As long as the results of such research are not available, obviously, acoustic measures of vowel quality and quality change in dynamic sound categories such as diphthongs should be approached with extreme caution. For the time being, then, we must conclude that the women lead the men in the /i/ to [i] change. The effect can be demonstrated acoustically, but for all the post-processing that was performed on the raw acoustic measurements, the leading role of the women seems to have been underestimated relative to the size of the effect as it is perceived by phonetic expert listeners. At best, then, the acoustic measurement procedure as outlined in this article may serve as a conservative measure. It will probably provide an adequate reflection of perceived vowel differences for monophthongs, whether produced by men and women, but it does not compensate for perceptual exaggeration of diphthongal trajectories, which effect we now believe may indeed exist. 4.5
Duration effects Let us, finally, consider the duration issue of the [i] to [a(:)i] change. Figure 10 presents the duration of the target /i/ as well as that of the reference point vowels /i/, which is a phonetically short vowel when
206
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
immediately followed by an obstruent (Nooteboom 1972), and /a/, which is a long vowel. Figure 10 shows that the /i/ of both male and female speakers has the same duration as the long, tense reference vowel /a/, and that both /a/ and /i/ are some 50 ms longer than the short reference vowel /i/. We computed a speaker-normalized duration measure for /i/ by dividing the duration difference between /i/ and /i/ into the difference between /a/ and /i/. A paired t-test on the normalized /i/ durations revealed no effect of sex of speaker. 5. Conclusions and discussion The results that were obtained from the acoustic analysis of the 320 targets diphthongs (10 tokens of /i/ for each of 16 male and 16 female speakers) allow us to answer the phonetic issues raised in the introduction. The phonetic characterization of /i/ in the emerging avant-garde variety of standard Dutch is that it has a lowered onset. The offset, or end-point of the diphthong, tends to keep its original vowel height, so that the quality change between the onset and offset of the diphthong has increased accordingly. The duration of the new variant of /i/ has not changed; as a result the speed of the spectral
Figure 10: Duration (ms) of short /i/, long /a/ and diphthong /i/ for 16 male and 16 female speakers.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
207
change (rate of formant change as visible in a spectrogram) must have increased as well. The analysis bears out that the onset of the new /i/ variety [ai] has the phonetic quality of a low front vowel, close to or even identical to the Dutch tense monophthong /a/ that was used as a reference vowel in the present study. The phonetic quality was therefore judged correctly by Stroop (1998). The lowered /i/ should not be equated with the loan diphthong /i/ that occurs in some Dutch words that were borrowed from French, such as detail ‘id.’, email ‘enamel’, canaille ‘riff-raff’. The onset of the imported /i/ would qualify as a back vowel, pronounced close to the lax Dutch vowel //. The analysis also shows that the timing properties of the lowered /i/ have remained the same before and after the change. The new variant [ai] cannot be equated to the long vowel plus glide combination /a:j/ as in haai ‘shark’, which combination should be some 50 to 90 ms longer than /i/ (see introduction). Consequently, we predict that native speakers will continue to observe a phonetic contrast in pairs such as hei ~ haai [hai ~ ha:i] and mei ~ maai [mai ~ ma:i]. Sociolinguistically, the data bear out that the avant-garde variant of /i/ is more strongly present in the female speaker group than in the male counterparts. Although extremely progressive and conservative speakers are found among both sexes, the women lead the change quite noticeably, especially in the middle portion of the range. This conclusion supports Stroop’s (1998) observation that the avant-garde variety of standard Dutch was initiated by women in precisely the socio-economic group that we targeted in this study. Methodologically, our study has the added advantage that the sound change in progress could be studied in more detail, and in a nonimpressionistic fashion, through the use of acoustic measurement procedures. Moreover, although cross-speaker and cross-sex comparison of acoustic measures of vowel quality are hazardous in principle, the procedure that we applied in our study, i.e., recording reference vowels and performing partial extrinsic speaker normalization on Bark-transformed formant measurements, affords useful comparison across vowels produced by male and female speakers. As far as we have been able to ascertain, we are the first researchers to have adopted this specific normalization procedure, which is really a mixture of extrinsic and intrinsic normalization. It bears a resemblance to Gerstman’s (1968) so-called end-point normalization, but differs from it in two details: (i) our procedure specifically looks for the front-most /i/ and the open-most /a/ in the front vowel continuum only, while the Gerstman procedure
208
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
indiscriminately adopts the lowest and highest F1 and F2 values in an entire vowel set as the end-points, and (ii) our procedure is applied after Barktransformation (which is a form of intrinsic normalization). It should be reiterated, finally, that our acoustic procedure should only be used with extreme caution when making comparisons between steady-state vowels and dynamic vowels such as the diphthongs in our study. We have a strong suspicion that the onset of a closing diphthong is heard with a more open vowel quality as the diphthongal gesture is larger. This effect, if indeed it can be shown to exist in a full-scale psychophysical experiment with static and dynamic vowel sounds, should be modeled in future vowel normalization procedures. Only then can acoustic measurements be used as a fully adequate substitute for (expert) human perception of vowel quality and quality change in diphthongs. References Adank, P., V.J. van Heuven & R. van Hout. 1999. “Speaker normalization preserving regional accent differences in vowel quality”. Proceedings of the 14th International Congress of Phonetic Sciences, San Francisco, 1593–1596. Bezooijen, R. van and R. van den Berg. 2001. “Who power Polder Dutch? A perceptual-sociolinguistic study of a new variety of Dutch”. Linguistics in the Netherlands 2001 ed. by T. van der Wouden and H. Broekhuis, 1–12. Amsterdam: John Benjamins. Bezooijen, R. van, S. Kroezen and R. van den Berg. 2002. “Approximant r: a new and vigorous change in Dutch”. Linguistics in the Netherlands 2002 ed. by H. Broekhuis and P. Fikkert, 1–11. Amsterdam: John Benjamins. Boersma, P. and V.J. van Heuven. 2001. “Speak and unSpeak with Praat”. Glot International 5.341–347. Boersma, P. and D. Weenink. 1996. “Praat, a system for doing phonetics by computer”. Report of the Institute of Phonetic Sciences Amsterdam, 132. Booij, G.E. 1995. The Phonology of Dutch. Oxford: Oxford University Press. Cohen, A., I.H. Slis and J.’t Hart. 1963. “Perceptual tolerances of isolated Dutch vowels”. Phonetica 9.65–78. Collier. R. and J. ’t Hart. 1983. “The perceptual relevance of the formant trajectories in Dutch diphthongs”. Sound Structures. Studies for Antonie Cohen ed. by M. van den Broecke, V. van Heuven and W. Zonneveld, 31–45. Dordrecht: Foris.
PRONUNCIATION OF /i/ IN AVANT-GARDE DUTCH
209
Collins, B. and I. Mees. 1981. The Sounds of English and Dutch. The Hague: Leiden University Press. Edelman, L. 1999. Het Poldernederlands: Een Vrouwentaal? Een Sociolinguistisch Onderzoek [Polder Dutch: A Woman’s Language? A Sociolinguistic Study]. Unpublished report, Linguistics Program, Nijmegen University. Gerstman, L.J. 1968. “Classification of self-normalized vowels”. IEEE Transactions Audio Electro-acoustics AU-16.78–80. Hart, J. ‘t 1969. “Fonetische steunpunten [phonetic stepping stones]”. De Nieuwe Taalgids 62.168–174. Hayward, K. 2000. Experimental Phonetics. Harlow: Pearson Education. Heuven, V.J. van, R. Bezooijen and L. Edelman. 2002. “The pronunciation of /i/ by male and female speakers of avant-garde Dutch”. Linguistics in the Netherlands 2002 ed. by H. Broekhuis & P. Fikkert, 61–72. Amsterdam and Philadelphia: John Benjamins. Koopmans-van Beinum, F.J. 1969. “Nog meer fonetische zekerheden [Even more phonetic certainties]”. De Nieuwe Taalgids 62.245–250. Labov, W. 1994. Principles of Linguistic Change. Internal Factors. Oxford: Blackwell. ----------. 2001. Principles of Linguistic Change. Social Factors. Oxford: Blackwell. Neary, T.M. 1989. “Static, dynamic, and relational properties in vowel perception”. Journal of the Acoustical Society of America 85.2088– 2113. Nooteboom, S.G. 1972. Production and Perception of Vowel Duration, a Study of Durational Properties of Vowels in Dutch. Dissertation, RU Utrecht. Nooteboom, S.G. and A. Cohen. 1976. Spreken en verstaan. Een inleiding tot de experimentele fonetiek [Speaking and Understanding. An Introduction to Experimental Phonetics]. Assen: van Gorcum. Peeters, W.J.M. 1991. Diphthong Dynamics: A Cross-linguistic Perceptual Analysis of Temporal Patterns in Dutch, English, and German. Dissertation, RU Utrecht. Rietveld, A.C.M. and V.J. van Heuven. 2001. Algemene fonetiek [General Phonetics]. Bussum: Coutinho. Stroop, J. 1998. Poldernederlands. Waardoor het ABN verdwijnt [Polder Dutch. What Makes Standard Dutch Disappear]. Amsterdam: Bert Bakker.
210
VAN HEUVEN, VAN BEZOOIJEN & EDELMAN
----------. 1999. “Young women’s farewell to Standard Dutch”. Paper presented at the New Methods in Dialectology Conference, X, 1–6 August 1999, St. John’s, Newfoundland, Canada. (www.hum.uva.nl/poldernederlands/english/stjohnspaper_engels.htm) Traunmüller, H. 1990. “Analytical expressions for the tonotopic sensory scale”. Journal of the Acoustical Society of America 88.97–100.
A TALE OF TWO DIALECTS RELATIVIZATION IN NEWCASTLE AND SHEFFIELD
JOAN C. BEAL & KAREN P. CORRIGAN University of Sheffield & University of Newcastle Introduction: Northern Englishes This paper represents the results of preliminary research for a much more extensive project on ‘Northern Englishes’, which will be conducted by scholars in the Universities of Sheffield, Newcastle and Edinburgh. The primary aim of this enterprise will be to collect and analyse linguistic data from a number of cities within and just beyond the boundaries of the linguistic north of England: Newcastle, Liverpool, Manchester, Sheffield and Nottingham in England, and Edinburgh in Scotland. Not surprisingly, given the paucity of research on dialectal morphosyntax, 1 typologies of British English vernaculars, such as those presented in Trudgill (1990), Hughes and Trudgill (1996) and Wells (1982), tend to divide the British Isles into dialect regions according to phonological rather than morphosyntactic criteria. In particular, the definition of ‘Northern’ dialects in England is usually based on the distribution of those two highly salient variables, BATH and STRUT, both of which provide isoglosses neatly bisecting the country at approximately the latitude of Birmingham. These linguistic markers of the ‘North-South divide’ in England are remarkably stable. Indeed, as recently as 2001, David Britain in published work on the Fenland region reported that the ‘mixed’ and ‘fudged’ lects of this transition zone are developing more clearly ‘Northern’ and ‘Southern’ variants in the northern and southern parts of this dialect area, respectively. Finding morphosyntactic equivalents of these northern and southern phonological markers is much more difficult. More often than not, features considered to be stereotypically ‘northern’ are frequently restricted to only part of northern England. Examples of this would be the existence of second person thee/thou and reduced or deleted definite articles, neither of which is found in the far north of England. The distribution of these latter features shows England north 1.
1
See Cornips and Corrigan (2005) for an outline of the reasons for this discrepancy.
212
BEAL & CORRIGAN
of the Tees differing radically from what Wells (1982) terms ‘the middle North’. In fact, the ‘North’ of England is a region whose boundaries appear to differ according to the perspective of the judge (Wales 2000). To a Northumbrian, Sheffield is a long way South, whilst to a Londoner, it is a long way North (in fact, Sheffield is equidistant from London and Newcastle). Historically, we might think of the North as the area covered by the AngloSaxon kingdom of Northumbria, stretching from the Humber to the Firth. By this definition, Newcastle is clearly in the Northern part of Northumbria (Bernicia), but Sheffield is a border city, marking the southern limit of Northumbria and the border with Mercia. Sheffield and Newcastle are both on the fringes of what was historically the Danelaw, but we might expect to find more evidence of Scandinavian influence in Sheffield than in Newcastle, because settlement, and therefore, language contact, was much less extensive in Bernicia. The same can also be said of these communities as late as the Middle Ages. Poussa (2002), for instance, provides evidence that Tyneside, as a coastal region, was considerably more exposed to the influence of Dutch and Flemish dialects via trade links than Sheffield will have been, on account of its rather different geographical location and history. This too may have important consequences for synchronic distinctions between these vernaculars. 2 2.
Relativization in Northern Dialects In Romaine’s (1982) discussion of the evolution of relativization strategies in English (summarized in Table 1 below), she argues that the changes were incremental and that relative marker omission was the norm in the most complex positions until WH-forms acquired their relativizing and subordinating functions. Gradually, subject contact relatives, in which the marker was entirely omitted, became disfavoured to the extent that in the Modern Standard, WH-relative and TH-markers have become categorical in this function, though omission continues with objects and indefinites in certain registers (cf. Biber 1986). Hence, Quirk et al. (1985: 865) declares the use of unmarked subject contact relatives, as in (1) below, to be ungrammatical, though it would be perfectly acceptable in a number of non-standard dialects, including Tyneside, 3 as (2)–(5) show: 2
It is likely too that the divergent language and dialect backgrounds of nineteenth and twentieth century migrants to these two cities will also have had an impact, as argued recently in Allen, Beal, Corrigan and Maguire (2004), Beal (2004) and Beal and Corrigan (2004). This issue will likewise be explored further in the Northern Englishes project mentioned above. 3 See also Tottie (1995) for a discussion of the persistence of the zero form in written varieties of British and American Englishes.
A TALE OF TWO DIALECTS
213
(1)
*The table [Ø stands in the corner has a broken leg]
(2)
Leck is a young boy [Ø was coming home from school] (Beal 1993: 208)
(3)
There’s a shortcut [Ø takes you to the shops] (Henry 1995: 125)
(4)
He's got one boy [Ø is twelve, and the other is nine] (Mesthrie 1991: 466)
(5)
The magistrate [Ø was to try him was an old soldier too] (Corrigan 1997: 88)
Genitive of which and oblique object to which are thought to have appeared first and to have become established in the period 1400–1500. The pronouns whom and whose, which distinguish animacy of the antecedent as well as object and genitive functions, surface in the fifteenth century, while the nominative who only ousts the subject-contact unmarked type in the sixteenth. Moreover, the adoption of the WH-strategy appears to have been sensitive to both stylistic stratification in the sense of Labov (1966) and Eckert and Rickford (2001) and restrictive (R) versus non-restrictive (NR) relative function. Hence, it occurs first in more formal (particularly Latinate) styles and, again, the nominative type is confined to formal usage for longer than the object or genitive types. Similarly, forms with co-referential WH-pronouns predominate in NR post-modification and the TH-forms favour R contexts. Romaine (1982: 60) concludes, therefore, that: WH-marked relative clauses entered the language and spread from more formal to less formal styles and from less frequently relativized syntactic positions (e.g. genitive) to more frequently relativized ones (e.g. subject) gradually displacing that as a relative clause marker. While the modern written language shows WHforms in nearly all styles and all syntactic positions, there remains a significant residue of that forms in the spoken language.
What is even more interesting from the perspective of this research is her subsequent assertion that ‘infiltration of WH into the relative system...has not really affected the spoken language’ (1982: 212).
BEAL & CORRIGAN
214 Marker type Demonstratives
OE se, seo, thæt etc.
ME that
ModE ----> that
Indeclinable Relative Particle Interrogatives
that
the the hwa, hwilc, hwær who, which, etc. where etc. Ø Ø
who, which, where etc. Omission Ø (non-subject relatives only) Table 1: The expansion of relative markers in the history of English: A reconstruction. (After Romaine 1982: 53ff.)
Recent investigations into the relative formation strategies employed in contemporary spoken non-standard British and extra-territorial Englishes, suggests that, in many of these varieties, the typical ratio of WH- to TH- and Ø relatives does, indeed, lag behind that of present-day Standard (English) English. 4 These vernaculars appear instead to be at various points along the reconstructed continuum of Table 1. For instance, in the traditional dialects of Northumberland and Yorkshire, as exemplified in the Survey of English Dialects (SED), WH- is not used where the antecedent is subject. The SED elicited forms of the relative with subject antecedent by asking informants: ‘HOW WOULD YOU COMPLETE A SENTENCE LIKE’: The woman next door says: The work in this garden is getting me down. You say: Well, get some help in. I know a man….will do it for you. In Northumberland, in five locations, the zero strategy was offered, i.e. ‘a chap would do it’; in three at was used; and in one location that was given. By contrast, informants from Sheffield, at locations 32 and 34, gave the responses Ø and at respectively and the that option was not included. (Orton & Halliday 1963: 1083–1084). Considering the distribution of responses to this question throughout England, Poussa (1986: 101) writes: we might argue that the development from the OE se…the relative to the modern system in the spoken language has generally passed through a ZERO stage, and that these areas [the extreme north and south] are relicts of that development.
On the other hand, as might be predicted from Romaine’s account, WHis used where the antecedent is possessive (genitive). In response to the 4
See, for instance, Beal and Corrigan (2002); Cheshire (1991); Edwards et al. (1984); Kendrick (2001); Mesthrie (1991); Miller (1988), (1993); Milroy and Milroy (1993); Policansky (1982); Poussa (2002), Romaine (1981), (1982), (1984); Tagliamonte (2002) and Tottie and Rey (1997) inter alia.
A TALE OF TWO DIALECTS
215
question ‘HOW WOULD YOU COMPLETE A SENTENCE LIKE’: That man’s uncle was drowned last week. In other words, you might say, that’s the chap…, WH- in the form of /hwe:z/ or /wi:z/ was given in seven locations in Northumberland, ‘at his uncle was …’ in one location and ‘as his uncle was…’ in the remaining location. In the Sheffield area, informant 32 from Ecclesfield, then a village just outside Sheffield, avoided the relativization strategy altogether by producing (That’s the chap thou knows, his uncle drowned hissen), whilst the informant from the city of Sheffield gives the WH-marker whose. Evidence from the SED thus suggests, that, in the traditional dialects of Northumberland and Sheffield, in the mid 20th century, WH- had infiltrated in the genitive but had not yet done so in subject position. Although this contradicts Romaine’s assertion in her 1982 monograph that WH- has not infiltrated the spoken language, the fact that WH- is used with possessive, but not with subject antecedents, would be predicted by Table 1. If the presence of whose, albeit in forms which are phonetically non-standard, represents the vanguard of a linguistic change in progress, we might predict that this change would be still more advanced in the urban dialects of Newcastle and Sheffield in the later 20th century. 2.1
The Northern Englishes Database Two major surveys of Tyneside English and one of Sheffield English have been carried out in the second half of the 20th century. The Tyneside Linguistic Survey (TLS), for which fieldwork was conducted in the late 1960’s, and the Phonological Variation and Change (PVC) project, undertaken in 1994, were carried out in Gateshead and Newcastle, respectively. The data from both these surveys has now been incorporated into the web-based corpus known as The Newcastle Electronic Corpus of Tyneside English (NECTE) 5 (see www.ncl.ac.uk/necte/). TLS Informants were selected by means of a random sample stratified by ‘rateable value per dwelling by polling district’ (Pellowe et al. 1972) and free-form interviews were recorded in subjects’ homes by a single interviewer, who had lived in Gateshead all his life and spoke with a local accent. The average length of interview was 30 minutes. Of these interviews, 83 are extant, though 150 were planned. The PVC data was collected in the Chapel House and Newbiggin Hall areas in the west of 5
The authors and their associates are grateful to the Arts and Humanities Research Board, the British Academy, the Catherine Cookson Foundation and the Small Grants Committee of the University of Newcastle upon Tyne for funding various phases of this resource enhancement project.
216
BEAL & CORRIGAN
Newcastle (on the north bank of the Tyne). Informants were selected using a social network model (Milroy & Milroy 1985), ‘divided to sample the community along parameters of age, gender and broadly-defined socioeconomic class’ (Watt & Milroy 1999: 27). They were recorded in dyadic pairs (friends or relatives), with the fieldworker deliberately keeping out of the conversation and as inconspicuous as possible. Although only 32 speakers make up the PVC corpus, a total of 20 recordings were made involving 39 speakers (one was recorded twice with different partners) and all of these have been incorporated into NECTE. The Survey of Sheffield Usage (SSU) was conducted in 1981 under the direction of Graham Nixon, one of the initial team of researchers for the TLS. The sampling methodology and interview structure were identical to those of the TLS, but, instead of using one interviewer, the SSU had several students conducting about 10 interviews each. The average length of SSU interviews, like the TLS ones, is 30 minutes. 100 interviews were recorded, and the original tapes are catalogued and stored in the Archives of Cultural Tradition, University of Sheffield. 6 In 2000–1, 52 of these interviews were digitised (those whose subjects were Sheffield-born) and in 2002, a small grant from the British Academy was secured in order to have some of these orthographically transcribed, following the same orthographic transcription protocol as that used in the NECTE project. At the time of writing, there were 9 complete transcriptions and it is this electronic corpus that forms the database of Sheffield material analysed in this paper. 2.2
Relativization in Newcastle and Sheffield Beal and Corrigan (2002) offered an analysis of relativization strategies in Tyneside English, based on a sub-sample of 9 informants from each of the two Tyneside corpora (TLS and PVC), which comprise NECTE. The findings of this investigation are compared with those for the SSU below with a view to addressing the research questions raised in §§1.0 and 2.0 above. 7 Figure 1 shows a broad-based comparison of relative types in all clauses for the three sub-corpora.
6
http://www.shef.ac.uk/english/natcect/archives/archives.html. Whilst the authors fully accept that the results from such small samples should not be viewed as definitive, they may well provide pointers for further research into the exogenous and endogenous factors that have influenced these distinctive varieties once the CSU and the electronic corpus arising from the Northern Englishes project are complete. Indeed, it is hoped longer term to offer the level of analysis provided by Tagliamonte et al., this volume, whose more persuasive results are based on a significantly larger data-set.
7
A TALE OF TWO DIALECTS
217
60
50
40
% Relative Type 30
20
10
0
TLS
PVC
SSU
wh
57
37
27
that
32
37
38
zero
11
26
18
what
1
0
17
Corpus
Figure 1: Relative Clause Marking in TLS, PVC and SSU.
Overall, the TLS sub-sample has the highest proportion of WH-relatives, whilst the SSU has the lowest. In the PVC corpus, that and WH-relatives are evenly distributed, whilst speakers in the SSU sub-sample employ TH-relatives slightly more frequently than they do the WH-type. The use of what as a relative pronoun (as in My play what I wrote) is almost nonexistent in the Tyneside sub-corpora, but accounts for almost as many relative clauses as the Ø variant does in the SSU database. Whilst this initial analysis shows some points of contrast between the Sheffield and Tyneside dialects, especially with regard to what, previous research in this area suggests that a comparison of the distribution of relative types according to clause and antecedent type is required in order to arrive at a more meaningful picture. 2.2.1 Restrictive vs. non-restrictive relative clauses. According to Quirk and Greenbaum (1973: 383), in Standard English non-restrictive relative clauses: ‘the repertoire of pronouns is limited to the wh-items’. This categorical use of WH- with non-restrictives is also found in certain other dialects of British English (see Cheshire 1982), though the evidence presented in (6)–(8) below clearly demonstrates that this is not the case for the Northern Englishes which are the focus of this research. In the Tyneside sub-sample, non-restrictive
BEAL & CORRIGAN
218
relative clauses accounted for just 11% of all relatives in the TLS sample (16 out of 147), and 8% (17 out of 222) in the PVC sample. However, WH-markers were not categorical in these since 3 (19%) had that in the TLS sample, and 1 (6%) had that in the PVC sample. In the Sheffield sub-sample, non-restrictive clauses accounted for 13% of all relative clauses (15 out of 115) and, of these, 2 (13%) were marked by that. Whilst the raw scores are too small to show statistically significant differences, this does indicate that, in both Tyneside and Sheffield English, WH- is not categorical in non-restrictive relative clauses as previously claimed for northern dialects more generally by Jones (1972: 140), Nevalainen and Raumolin-Brunberg (2002: 112) and Romaine (1982: passim). (6)
The old grammar school on Durham Road, that was a coeducational school. (TLS G219)
(7)
This is Louise, that was meant to come. (PVC V7) 8
(8)
You know Mr. Hill, that you got down there. (SSU 080)
2.2.2 Sentential relatives. In the above discussion of non-restrictive relative clauses, we excluded those clauses that have whole sentences as their antecedents, such as (9), (10) and (11) below. According to Macafee (1983: 52) ‘Scottish speakers use which almost exclusively in non-restrictive relative clauses with sentential antecedents’. This may well be a ‘pan-northern’ phenomenon since which is used exclusively in the sentential relative clauses given in (9), (10) and (11), all of which are drawn from the NECTE and SSU corpora. These uses of which, however, account for only a minority of WHrelatives in the databases. In the TLS sample, for instance, there are 13 sentential relatives out of 83 tokens of WH-relatives (16%), and in the PVC sample, 16 of the 82 (20%) WH-relatives counted were of the sentential type. In the Sheffield sub-sample, there are 8 sentential relatives out of 31 WHrelative tokens (26%). Thus, in neither dialect is which confined to sentential relatives, but the use of which in these contexts appears to be more common in the Sheffield data than in that for Tyneside.
(9)
8
He said that…er…Anthony Eden was going the wrong way, which to me was ridiculous (TLS G216)
In all cases analysed as non-restrictive that, the audio tapes have been checked to ensure that that was pronounced with schwa, and therefore could not be a demonstrative.
A TALE OF TWO DIALECTS
219
(10) They’re busy wondering where their next meal’s going to come in and stuff, which I think is really sad (PVC T10) (11) if it’s say like at Doncaster or wherever he goes which he’s got (-T) car so it’s no problem (SSU 015) 2.2.3 Restrictive relatives by antecedent type. 9 Corpus-based studies of relativization in Standard English and other English dialects (cf. Ball 1996, Beal 1993, Corrigan 1997, Miller 1988 and 1993, and Tagliamonte 2002) have demonstrated the importance of distinguishing between different antecedent types, with regard to both the grammatical category and the animacy of the antecedent. Since non-restrictive relative clauses favour WH-, even in the, otherwise, highly non-standard dialects of Tyneside and Sheffield, and which is categorical in sentential relative clauses in both vernaculars, it is only inrestrictive relatives that variation between WH-, TH- and Ø may be constrained by antecedent type. Similarly we noted in §2.0 that, even in the more archaic dialect represented by Northumbrian speakers in the SED, whose was used in the genitive in 7 out of 9 locations. Although there were only 2 instances of relatives with genitive antecedents in the entire Tyneside subsample, both of these used whose. Whilst the numbers here are too small to allow us to claim categorical status for whose as a genitive relative in Tyneside English (we have no examples here of genitives with non-human or inanimate antecedents), this certainly confirms the infiltration of whose into the relative system of urban Tyneside English. Unfortunately, there were no genitive relatives at all in the Sheffield sub-sample, so it is not possible at this stage to confirm that the use of whose by the Sheffield informant in the SED is replicated in the modern dialect. (12) You know my cousin whose husband died (TLS G216) (13) Some kids whose parents were on low income (PVC D23)
9
The account of restrictive relative markers that follows is not intended to be exhaustive. We have omitted the use of where as a relative marker in clauses like Apart from that it’s, youknow, the cases where you’re washing the car, or gardening or something (TLS G224) and will return to these in future research.
BEAL & CORRIGAN
220
70
60
50
40 % Relative Type 30
20
10
0
SA
OA
SI
WH
63
35
34
OI 8
THAT
28
22
59
48
ZERO
9
44
7
44
Antecedent Type
Figure 2: Restrictive Relatives by Antecedent Type in the NECTE Corpus.
Figures 2 and 3 below show the results of the analysis of restrictive relative clauses in the Tyneside and Sheffield samples, respectively, according to antecedent type (S = Subject, O = Object, A = Animate, I = Inanimate). 10 A comparison of Figures 2 and 3 would seem to confirm the point mooted earlier, namely, that the WH-strategy is more prevalent overall in Tyneside than it is in Sheffield. They also show the interesting distribution of the what relative in these dialects, which we will discuss, in further detail below. Of particular note in this context, is that, whilst never a majority choice, the what relative marker is found with all antecedent types in the SSU corpus. 2.2.4 What as a relative pronoun. Beal and Corrigan (2002) raised the absence of what in relative clauses found in NECTE as important in the light of the findings related in Cheshire, Edwards and Whittle (1993: 68), i.e.: the schools taking part in the survey reported what far more frequently than any of the other non-standard relative pronoun forms, and that relative what was reported just as frequently…in the North of England as in the South….What, then, appears to be the preferred relative pronoun in the urban centres of Britain today. 10
In Figure 1, the results from the TLS and PVC sub-samples have been conflated.
A TALE OF TWO DIALECTS
221
70
60
50
40 % Relative Type 30
20
10
0
SA
SI
OA
WH
26
10
10
OI 2
THAT
50
60
70
18
ZERO
9
10
0
62
WHAT
15
20
20
18
Antecedent Type
Figure 3: Restrictive Relatives by Antecedent Type in the SSU Corpus.
It is worth noting here that Cheshire et al.’s (1993) survey was based on elicitation questionnaires conducted by schoolteachers, rather than on the analysis of recorded usage. The TLS interviews, likewise, included elicitation questions towards the end of each session. The interviewer would ask the subjects whether they would say certain sentences, one of which was: (14) There was all these bottles of beer what we had brought Of those asked this question, 23 (49%) gave an unequivocal ‘yes’, 16 or (34%) said ‘no’, 4 gave a qualified ‘yes’ (e.g. they’d heard it but wouldn’t say it), 1 said ‘don’t know’ and 3 didn’t understand the question. 11 The relatively high numbers of speakers who gave a negative acceptability judgment for what as a relative marker appears to indicate that this feature is not particularly wellintegrated in the grammars of Tyneside speakers. The results from the NECTE sub-samples support this in that there is only one example in this sample, and 11 TLS subjects whose speaker codes are in the higher numbers (presumably because they were interviewed towards the end of the survey) weren’t asked this question at all. This seems to suggest that the TLS researchers may have lost confidence in the elicitation technique as the survey progressed, which is something the NECTE team intend to investigate further.
222
BEAL & CORRIGAN
that occurs in the 1960’s TLS database where it is produced by one of the oldest speakers in the survey: (15) Bairns don’t play the games what we did. (TLS G59) In the Sheffield sub-sample, however, what seems well established and is attested amongst younger as well as older speakers. Examples of its use are: (16) We listen to Radio-Sheffield to all (-T) news what’s going off youknow (SSU 008) (17) It’s double (-T) money what you’re getting at home (SSU 014) The pattern of what relative usage across all antecedent types in Figure 3 of §2.2.3 above is similar to that found in Cheshire’s study of Reading English reported in her (1982) monograph and reproduced as Figure 4 below. The level of overall employment of this strategy is higher in Reading than in Sheffield, but the proportion of use between antecedent types is similar. In this respect, then, the Sheffield dialect is patterning more like a southern variety than the northern Tyneside vernacular appears to be. The SED shows what appearing in three Eastern areas: the area around London, the Wash, and around the Humber estuary. There is more recent evidence that what is found in urban dialects neighbouring that of Sheffield. In particular, Petyt’s (1985: 238) study of the dialect of West Yorkshire (Bradford, Halifax and Huddersfield) gives evidence of 8 informants (out of 106) producing examples of what relatives. The latter could have spread to Sheffield and West Yorkshire from either the Humber or the Wash well before 1981, or the phenomenon could be more recent, and be due to the influence of London. Clearly, this is an area that warrants further investigation. 2.2.5 WH-relatives in NECTE and the SSU. Beal and Corrigan (2002) compared the use of WH-relatives in non-standard dialects such as Tyneside and Reading with the account of Standard English given in Quirk (1957) noting that in all three varieties this marker was preferred with animate antecedents in subject position. A comparison of figures 2 and 4 above shows that Tyneside and Reading had WH- in, respectively, 63% and 58% of relative clauses with animate antecedents in subject position, but in Standard English (Quirk), the WH-marker was almost categorical at 91%. The results from the Sheffield subsample in Figure 3 of §2.2.3 above show a decidedly different pattern. In
A TALE OF TWO DIALECTS
223
60
50
40
% Relative Type 30
20
10
0
SA
OA
SI
OI
WH
58
20
8
12
THAT
17
20
54
16
ZERO
14
40
0
44
WHAT
11
20
38
28
Antecedent Type
Figure 4: Restrictive Relative Clause Markers by Antecedent Type, Reading Corpus.
this case, that (cf. [18] and [19] below) is the preferred relative marker even with animate antecedents in subject position (50%), while the WH-form is used in only 26% of cases leading to a pattern which is more or less the reverse of that which obtains in the NECTE corpus. (18) there were a school teacher that lived in here in this house (SSU 008) (19) I’ve got two other sisters that are both working (SSU 015) We noted earlier (and in Beal & Corrigan 2002) that the results from the NECTE corpus seemed to contradict Romaine’s (1982) assertion that spoken vernacular Englishes remained relatively devoid of WH-marking. 12 In this respect, the dialect of Sheffield seems much closer to the received view of what non-standard dialects of British English are like. There is certainly no evidence of a steady northward drift of WH-infiltration, so perhaps another explanation 12 This is a controversial area as can be seen from the contradictory findings of Tottie (1995) as well as Tagliamonte (2002) on the one hand and Ball (1996) on the other.
224
BEAL & CORRIGAN
for the presence of WH-marking in this context in Tyneside English is called for — though we are not in a position to suggest any hypotheses at this juncture. Moreover, any explanation will have to take account of the fact that previous investigations of the usage of this variable in subject relative contexts have also produced rather unexpected results as is clearly demonstrated, on the basis of evidence from the variety represented in the York corpus, in Tagliamonte (2002). 2.2.6 That and at in NECTE and the SSU. Perhaps the most striking point to emerge from the findings presented in Figure 4 is the predominance in the SSU of that with all antecedent types (except for inanimate antecedents in object position, where zero is favoured.) In the traditional dialect of Sheffield, as exemplified in the SED, the preferred relative marker in contexts where an animate antecedent occurs in subject position (a chap ---- do it for you) is at. However, in the SSU sub-sample analysed for this investigation, only one token of at was found and it may be no accident that it occurred in the speech sample of one of the oldest and most ‘traditional’ speakers in the entire corpus. (20) Kelvin at my first husband came out of (SSU 008) Although at clearly looks and sounds like a reduced form of that, Wright (1892: 91) argues that it is an independent form of Norse origin. Petyt (1985: 201) notes that in his corpus of West Yorkshire speech (collected 10 years earlier and within a 50-mile radius of the SSU survey) the TH-relative marker is more frequent and seems to be gaining ground on the more traditional at form. Although we concur with Wright’s argument that historically at is probably not a reduced form of that, the current predominance of that in the SSU, in particular, needs further explanation. One hypothesis might be that the TH-marker is a ‘restored’ form of at which would once have been prevalent in this former Danelaw area as a result of linguistic contact. Northumberland was a Ø rather than an at area in the SED, which is to be expected, considering that it is outside the main area of Danish influence. Perhaps the lesser dominance of that in present-day Tyneside English can, likewise, be attributed to the lack of at? Although this is only a tentative suggestion, resting as it does on negative evidence at this point, it may prove to be interesting from a linguistic contact perspective. 2.2.7 The zero relative marker in NECTE and the SSU. By comparing Figures 2, 3 and 4 above, we see that, in Sheffield English, the zero relative is most prevalent with inanimate antecedents in object position (62%, contrasting
A TALE OF TWO DIALECTS
225
with 44% in Reading and 48% in Tyneside). Like Reading, but unlike Tyneside, Sheffield does not have any instances of zero relative with inanimate antecedents in subject position, but all 3 non-standard varieties have zero relative for between 8 and 14% of sentences with animate antecedents in subject position. Examples from the Tyneside and Sheffield sub-samples are: (21) I’ve a mother [RP Ø’s still living] she’s a widow (SSU 017) (22) I’ve a sister [RP Ø’s over there] she loves stotties (TLS G52) This evidence (coupled with that from diverse vernaculars such as the examples of the feature from Irish-English and South African Indian English given in §1.0 above) would suggest that the use of zero relatives in sentences with animate antecedents in subject position is a minority variant in a wide range of non-standard dialects of English and does not appear to be a particularly useful marker of regional distinctiveness. 3.
Conclusion On the basis that the samples analysed here are far too small to be subjected to rigorous tests of statistical significance of the sort that are key to the findings of Tagliamonte et al., this volume, any conclusions that we have reached at this stage as regards morphosyntactic dialect boundaries in this investigation of relative clause markers must be tentative. Nevertheless, it would appear, for example, that the what relative is not as widespread in usage across the British Isles as Cheshire et al. (1993), for instance, have previously suggested. Possibly, this is one feature that divides the ‘Middle North’, including Sheffield and West Yorkshire, from the ‘Far North’ of Tyneside and Lowland Scotland. Similarly, the prevalence of that in the SSU corpus contrasts sharply with the greater reliance on both WH- and Ø relative marking strategies in NECTE. Perhaps this phenomenon can be attributed to the influence of earlier Norse at in the former Danelaw with ‘lightly Scandinavianised’ regions showing minimal traces of the feature synchronically. Finally, there would appear to be little mileage in pursuing a morphosyntactic ‘North/South’ isogloss based on the evidence of zero relatives as markers in sentences in which animate antecedents occur in subject position. The infrequency of the variable (possibly as a result of computational processing and/or pragmatic factors [cf. Fodor 1998 and Cornips & Corrigan 2005]) and the prospect that these structures are not true relative clauses in structural terms (cf. Doherty 1993 and Henry 1995) both mitigate against the treatment of this phenomenon in any straightforward manner.
226
BEAL & CORRIGAN
References Allen, W., J.C Beal, K.P. Corrigan & W. Maguire. 2004. “‘Paddy’ and ‘Jock’ meet ‘Geordie’: A prolegomenon to investigating the reflexes of nineteenth century linguistic contact in the North East”. Paper presented to The Influence of the Languages of Ireland and Scotland on Linguistic Varieties in Northern England, seminar, University of Aberdeen, June 28th, 2004. (http://www.abdn.ac.uk/langling/resources/sympprog.html) Ball, C.N. 1996. “A diachronic study of relative markers in spoken and written English”. Linguistic Variation and Change 8.227–258. Beal, J.C. 1993. “The grammar of Tyneside and Northumbrian English”. Milroy & Milroy 1993. 187–213. ----------. 2004. “Sheffield: Life on the edge in the largest village in the world”. Paper presented to The Influence of the Languages of Ireland and Scotland on Linguistic Varieties in Northern England, seminar, University of Aberdeen, June 28th, 2004. (http://www.abdn.ac.uk/langling/resources/sympprog.html) Beal, J.C. & K.P. Corrigan. 2000a. “Comparing the present with the past to predict the future for Tyneside (British) English”. Newcastle & Durham Working Papers in Linguistics 6.13–30. Beal, J.C. & K.P. Corrigan. 2000b. “New ways of capturing the 'Kodak moment': Real-time vs. apparent time analyses of Syntactic variation in Tyneside English, 1969–1994”. Paper presented to VIEW, University of Essex, September 2000. Beal, J.C. & K.P. Corrigan. 2002. “Relativization in Tyneside and Northumbrian English”. Poussa 2002. 125–134. Beal, J.C. & K.P. Corrigan. 2004. “The impact of nineteenth century Celtic English migrations on contemporary Northern Englishes: Tyneside and Sheffield compared”. Paper presented to Celtic Englishes IV, University of Potsdam, 22nd–26th September 2004. (http://www.uni-potsdam.de/u/CE/col4/colloquium4.htm) Biber, D. 1986. “Spoken and written textual dimensions in English”. Language 62.384–414. Cheshire, J. 1982. Variation in an English Dialect. Cambridge: Cambridge University Press. ----------, ed. 1991. English around the World. Cambridge: Cambridge University Press. Cheshire, J., V. Edwards & P. Whittle. 1993. “Non-Standard English and dialect levelling”. Milroy & Milroy 1993. 53–96. Cornips, L. & K.P. Corrigan. 2005. “Convergence and divergence in grammar”. Dialect Convergence and Divergence ed. by P. Auer, F.
A TALE OF TWO DIALECTS
227
Hinskens and P. Kerswill, 106–134. Cambridge: Cambridge University Press. Corrigan, K.P. 1997. “The acquisition and properties of a contact vernacular grammar”. Dán do Oide ed. by A. Ahlqvist & V. Capková, 75–94. Dublin: ITÉ. ----------. To appear. Parametric Variation within a Socially Realistic Linguistics: Syntactic Variation and Change in South Armagh English. TPS, Oxford: Blackwell. Doherty, C. 1993. The Syntax of Subject Contact Relatives. Unpublished ms., University of Southern California. Eckert, P. & J. Rickford, eds. 2001. Style and Sociolinguistic Variation. Cambridge: Cambridge University Press. Edwards, V., P. Trudgill & B. Weltens. 1984. The Grammar of English Dialect: A Survey of Research. London: ESRC. Fodor, J.D. 1998. “Learning to parse?” Journal of Psycholinguistic Research 27:2.285–319. Henry, A. 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Hughes, A. and P. Trudgill. 1996. English Accents and Dialects. London: Edward Arnold. Jones, C. 1972. An Introduction to Middle English. London: Holt, Rinehard and Winston. Jones, V. 1985. “Tyneside syntax: A presentation of some data from the Tyneside Linguistic Survey”. Focus on England and Wales ed. by W. Viereck, 163–177. Amsterdam: John Benjamins. Kendrick, A. 2001. The Use of Relative Clauses in Tyneside English. Unpublished undergraduate dissertation, University of Newcastle. Labov, W. 1966. The Social Stratification of English in New York City. Washington D.C.: Center for Applied Linguistics. Macafee, C. 1983. Glasgow. Amsterdam & Philadelphia: John Benjamins. Macaulay, R.K.R. 1991. Locating Dialect in Discourse: The Language of Honest Men and Bonny Lasses in Ayr. Oxford: Oxford University Press. Mesthrie, R. 1991. “Syntactic variation in South African Indian English: The relative clause”. Cheshire 1991. 462–490. Miller, J. 1988. “That a relative pronoun? Sociolinguistics and syntactic analysis”. Edinburgh Studies in the English Language ed. by J.M. Anderson & N. MacLeod, 113–119. Edinburgh: John Donald. ----------. 1993. “The grammar of Scottish English”. Milroy & Milroy 1993. 99–138.
228
BEAL & CORRIGAN
Milroy, J. L. Milroy & G. Docherty. 1997. Phonological Variation and Change in Contemporary Spoken British English. ESRC, Unpublished Final Report, Dept. of Speech, University of Newcastle-Upon-Tyne. Milroy, J. and L. Milroy. 1985. “Linguistic change, social network and speaker innovation”. Journal of Linguistics 21.339–384. Milroy, J. and L. Milroy, eds. 1993. Real English: The Grammar of English Dialects in the British Isles. London: Longman. Nevalainen, T. & H. Raumolin-Brunberg. 2002. “The rise of relative who in Early Modern English”. Poussa 2002. 109–121. Orton, H. & W. Halliday, eds. 1963. Survey of English Dialects. Vol. 1, part 3: The Six Northern Counties and the Isle of Man. Leeds: E. J. Arnold & Son. Pellowe, J., G. Nixon, B. Strang & V. McNeany. 1972. “A dynamic modelling of linguistic variation: The urban (Tyneside) linguistic survey”. Lingua 30.1–30. Petyt, K.M. 1985. Dialect and Accent in Industrial West Yorkshire. Amsterdam: John Benjamins. Policansky, L. 1982. “Grammatical variation in Belfast English”. Belfast Working Papers in Language and Linguistics 6.37–66. Poussa, P. 1986. “Historical implications of the distribution of the zeropronoun relative in Modern English dialects: Looking backwards towards OE from Map S5 of the Linguistic Atlas of England”. Papers from the Third Scandinavian Symposium on Syntactic Variation, Stockholm, May 11–12, 1985 ed. by S. Jacobson, 99–117. Stockholm: Almqvist & Wiksell. ----------, ed. 2002. Relativization on the North Sea Littoral. Munich: Lincom Europa. Quirk, R. 1957. “Relative clauses in educated spoken English”. English Studies 38.97–109. Quirk, R. & S. Greenbaum. 1973. A University Grammar of English. London: Longman. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik, eds. 1985. A Grammar of Contemporary English. London: Longman. Romaine, S. 1980. “The relative clause marker in Scots English: Diffusion, complexity and style as dimensions of syntactic change”. Language in Society 9.221–249. ----------. 1981. “Syntactic complexity, relativization and stylistic levels in Middle Scots”. Folia Linguistica Historica 2.56–77. ----------. 1982. Socio-Historical Linguistics: Its Status and Methodology. Cambridge: Cambridge University Press.
A TALE OF TWO DIALECTS
229
----------. 1984. “Relative clauses in child language, pidgins and creoles”. Australian Journal of Linguistics 4.257–281. Tagliamonte, S. 2002. “Variation and change in the British relative marker system”. Poussa 2002. 147–165. Tottie, G. 1995. “The man Ø I love: An analysis of factors favouring zero relatives in written British and American English”. Studies in Anglistics ed. by G. Melchers and B. Warren, 201–215. Stockholm: Almqvist & Wiksell. Tottie, G. and M. Rey. 1997. “Relativization strategies in earlier African American Vernacular English”. Language Variation and Change 9:2. 219–247. Trudgill, P. 1990. The Dialects of England. Oxford: Blackwell. Wales, K. 2000. “North and South: An English Linguistic Divide?” English Today 16:1.4–15. Watt, D.J.L. and L. Milroy. 1999. “Phonetic variation in three Tyneside vowels: Is this dialect levelling?” Urban Voices ed. by P. Foulkes & G. Docherty, 25–46. London: Arnold. Wells, J.C. 1982. Accents of English. 3 vols. Cambridge: Cambridge University Press. Wright, J. 1892. A Grammar of the Dialect of Windhill in the West Riding of Yorkshire. London: English Dialect Society.
PART III Dialects across language boundaries
CROSSING GRAMMATICAL BORDERS TRACING THE PATH OF CONTACT-INDUCED LINGUISTIC CHANGE *
RUTH KING York University 1.
Introduction This paper deals with grammatical borders, more specifically, with the linguistic mechanisms of contact-induced language change. While all linguists would probably acknowledge that social factors come into play in determining the linguistic outcomes of contact situations, there is considerable disagreement as to whether linguistic factors are important or not. In addressing this issue, it is impossible to avoid reference to easily the most influential work in the field over the last two or three decades, Sarah Thomason and Terrence Kaufman’s 1988 Language Contact, Creolization and Genetic Linguistics. For cases of language maintenance, Thomason and Kaufman (1988: 74–76) establish a very general borrowing scale whereby they predict type and degree of borrowing on the basis of degree of contact, ranging from the borrowing of content words in cases of what they call ‘casual contact’ to heavy structural borrowing in cases of what they call ‘very strong cultural pressure’. For cases of language shift, they propose a scale which ranges from a small group shifting, with few structural effects, through to large scale shifting with extreme unavailability of the target language, with only vocabulary successfully acquired (Thomason & Kaufman 1988: 50). Thomason and Kaufman (1988: 14–15) have an ‘anything goes’ perspective according to which elements from any linguistic subsystem may be borrowed, depending on the particular social factors at play. They argue that ‘linguistic constraints on linguistic interference...are based ultimately on the premise that the structure of a language determines what can happen as a result ∗
I thank John Benjamins for permission to publish here material which appeared in The Lexical Basis of Grammatical Borrowing (2000). I thank as well Sandra Clarke, David Heap, Michol Hoffman and several members of the audience for questions and comments concerning the conference version of this paper. I especially thank Yves Roberge for collaborating on the initial analysis of Prince Edward Island French prepositional usage, published in King and Roberge (1990). The research reported on here was supported by standard research grants from the Social Sciences and Humanities Research Council of Canada.
234
RUTH KING
of outside influence. And they all fail’. Therefore, they suggest that ‘[a]ny linguistic feature may be transferred from any language to any other language’. 1 However, as Gillian Sankoff (2002) concludes with regard to this position, the outright rejection of linguistic constraints amounts to throwing out the baby with the bathwater. There are problems with evaluating the “no-linguistic-constraints” perspective, problems which are due to a great deal of the language contact literature itself, including the literature Thomason and Kaufman survey. In the first section of this paper I discuss four general problems, with examples. In subsequent sections I propose a theory of the process by which contact-induced grammatical change comes about, a theory which involves the idea of linguistic constraints, exemplified with data from Atlantic Canada Acadian French. 2.
Problems with the language contact literature The first problem in evaluating the language contact literature involves classification of contact phenomena, a problem which arises because the literature tends to focus on outcomes, not processes. For example, quite a number of works make statements of the type ‘The word order of language X has influenced the word order of language Y’ or ‘Language X borrowed its system of negation from language Y’ without the author(s) providing sufficient evidence for the claim. And often the claim gets repeated as fact, a point to which I return below. Related to this is the tendency to cite examples of ‘phonological’ or ‘morphological’ or ‘syntactic’ borrowing, while disregarding the fact that at a nonsuperficial level the change in question is actually neither phonological nor morphological nor syntactic. To choose an example from languages whose histories are well known, the introduction of Latin -able (as in laughable) is often cited as an example of structural borrowing into English, since a derivational affix is involved. However, the suffix entered the English language attached to adjectives, such as visible. It then began to attach to other Romance stems (e.g. palatable) and to native Germanic stems (e.g. readable). 2 But while the result does involve the addition of a new affix to English, the process is lexical: whole words were borrowed, not affixes on their own. While this is readily discernible in cases such as Latin influence on English, in cases of lesser known languages the reader is left to take the researcher’s characterization of ‘phonological’ or ‘morphological’ or ‘syntactic’ borrowing at face value. 1
In more recent work, specifically Thomason (2001), there is some retrenchment, in that social factors are viewed as probablistic rather than deterministic. 2 This example comes from April McMahon’s Understanding Language Change (1994).
CROSSING GRAMMATICAL BORDERS
235
A second problem with the contact literature concerns what forms the basis for comparisons. Many studies make comparisons with the standard variety of a contact language and/or do not consider the history of the languages in contact. For example, Irish English has five different constructions which would be expressed by the present perfect in standard English. However, Harris (1991: 206) notes with regard to contact-based explanations for such usage: Virtually all of those who have claimed that the peculiarities of Irish English aspectual usage can be traced to a substratal source have based their conclusions on a straightforward comparison with present-day Standard British English. They make no reference to the fact that, in some cases, very similar patterns of usage are to be found in other regional varieties of English as well as in earlier forms of the standard language.
A second point with respect to the comparison problem is that it is not always made clear whether the basis for analysis is the contact variety which provides the data, or the standard version of the language. A third problem involves a tendency for claims that may be based on sketchy data and/or secondary sources. For example, Ma’a, spoken by the Mbugu people of the Usambara Mountains of Tanzania, is typically taken to be a mixed language with Bantu grammar and southern Cushitic lexical roots. Numerous proposals as to the origins of Ma’a are found in the literature based on limited data collected by a number of researchers over the last century. However, on the basis of extensive fieldwork conducted in the early 1990s, Martin Mous (1996, 1997, 2001) has cast doubt on these proposals. Mous argues that what has been called a mixed language is in fact one language, Mbugu, with two registers. There is an ethnoregister, Inner Mbugu, which consists of a parallel lexicon to a regular Bantu language that all Mbugu speak, what they call Normal Mbugu. Inner Mbugu (or Ma’a) is a register used to signal group identity. Crucially, speakers of Inner Mbugu all speak Normal Mbugu. Further, Mous shows that Ma’a lexical items come from a variety of sources, including two different Cushitic sources, along with Maasai and several Bantu languages. While it is essential in developing a theory of contact-induced change to begin by systematizing the linguistic facts, which Thomason and Kaufman cite as the aim of their work, it is important to keep in mind that such theories will only be as good as the synchronic descriptions on which they are based. A final problem involves the relationship between theory and data. There is an intuition, supported by numerous case studies, that morphology and syntax are less susceptible to linguistic interference than are phonology or the
236
RUTH KING
lexicon. Proposed universals based on this intuition have had a rocky history, with opponents saying that a particular proposal is false because of the existence of data from some particular language. However, isolated counterexamples should not invalidate the generalization, especially when these counterexamples are based on fairly superficial analysis. With some notable exceptions, claims concerning structural borrowing tend not to be couched in terms of any particular framework of grammatical change, nor in terms of any particular theory of grammar. For instance, John J. Gumperz and Robert Wilson’s 1971 study of the linguistic consequences of intensive language contact in the village of Kupwar in India figures prominently in the contact literature. Gumperz and Wilson (1971: 155) claim Marathi and Urdu, both Indic languages, along with Kannada, a Dravidian language, have fallen together syntactically. The argument is made on the following grounds: [Since it is possible] to translate one sentence into the other by simple morph by morph substitution...the codes used in code-switching situations in Kupwar have a single syntactic surface structure.
Reliance on surface syntax is problematic, however, since it is a fairly simple matter to come up with a sentence one can translate from one language to another by simple morph by morph substitution for any number of languages which are generally regarded as distinct. In their article, Gumperz and Wilson posit sixteen convergent morphosyntactic changes in the space of fewer than eight pages of text, usually with just one supporting example for each (in the shape of a sample sentence from each relevant language). While it may be the case that the grammars of these particular languages have indeed converged, the evidence given in this article is hardly convincing. What are needed are more sophisticated analyses of the languages spoken in villages such as Kupwar, cases where it is argued that convergence is ongoing. 3.
Towards a theory of grammatical borrowing In recent work (King 2000), I have argued that an account of the grammatical consequences of language contact must begin with a theory of grammar. I propose a way of capturing the intuition that structural borrowing is ‘different’ by turning to the role of the lexicon in building grammatical structure, from the perspective of the Principles and Parameters approach embodied by Government and Binding theory and the more recent Minimalist Program. Within this framework the basic idea is that syntactic structure is largely determined by lexical information, or, more precisely, by the feature specifications of lexical items. So for the sentence Kelly will kick the ball the
CROSSING GRAMMATICAL BORDERS
237
lexical entry for the verb kick specifies that it takes two arguments, a subject bearing the thematic role of agent and an object bearing the role of patient; this will project a certain structural configuration, a different one than that associated with the verb put, for example, which has a different argument structure. Within the Principles and Parameters framework, syntactic variation between languages is viewed as explicable in terms of differences in the morphosyntactic properties of different lexical items (Chomsky 1993). I apply this basic approach to grammatical borrowing: borrowing a lexical item, which is in essence a bundle of features, including morphosyntactic features, may trigger change(s) in the borrowing language. Thus when structural change occurs, it is mediated by the lexicon. In theory, members of any category may be borrowed, given sufficient intensity of contact. Note that this is a substantially different claim than the notion that anything can be borrowed: Thomason and Kaufman argue that structural borrowing may take place independently of the borrowing of lexical items; I argue that grammatical borrowing has a lexical basis. This approach is compatible with calquing as a process of contact-induced change, given that calquing involves change in the properties of lexical items, rather than the direct importation of grammatical structure (cf. King 2000 for details). I will illustrate the idea of the lexical basis of grammatical borrowing through an analysis of the syntax of prepositions in Prince Edward Island French, a variety of Atlantic Canada Acadian French. While the analysis is couched in the technical apparatus of Principles and Parameters, my general point is that language contact research should be responsible to some theory of grammar. 4.
Varieties of French in Canada Canadian French provides a nice testing ground for theories of language contact since we are able to compare the structure of varieties spoken in more and less intense language contact situations. There are large sociolinguistic corpora for many Canadian varieties, including the French of Montreal (the Cedergren-Sankoff corpus and more recent Montreal corpora constructed by Pierrette Thibault, Diane Vincent and their associates), Ottawa, the nation’s capital and nearby Hull (Shana Poplack), the Eastern Townships of Quebec (Normand Beauchemin) and all four Atlantic Provinces, i.e. New Brunswick (Louise Beaulieu; Karin Flikeid), Nova Scotia (Karin Flikeid), Prince Edward Island (Ruth King) and Newfoundland (Ruth King). There is also ongoing construction of diachronic corpora for early twentieth and nineteenth century Quebec French varieties (cf. Poplack & St-Amand 2002), and for varieties
238
RUTH KING
from the time of New World settlement (cf. Forget & Martineau 2002; Martineau 2002). Structurally, there are two main varieties of Canadian French, i.e. Quebec French and Acadian French. Quebec French is spoken in the province of Quebec, of course, but the French of Ontario and western Canada may also be included under this rubric, as the francophone populations of central Canada and the west largely owe their origin to 19th century migration from Quebec. Acadian French is spoken in the four Atlantic provinces (see Map 1), in small pockets of Quebec and has a close relative in Louisiana, in the United States. The most important factor for structurally distinct varieties has been the relative isolation of the Acadian people from contact with other francophones and from normative influences of a French education system, from the early 17th century until the early 20th century in some cases, until the late 20th in others, and continuing to the present day in still others. Until fairly recently Acadians have had little contact with other francophones, since they have tended to live in rural, isolated areas. What has happened is that often we get far better preservation of archaic features of the language in Atlantic Canada than in France or in Quebec; in some cases we also get post-settlement innovations. The most conservative Acadian varieties turn out to be important for the study of French more generally, as they provide in many respects a mirror on the past (cf. King and Nadasdi 1997). As for the influence of English on Canadian French, our knowledge of the history of French and of modern French varieties spoken in France by monolinguals should prevent specialists at least from making erroneous claims. But since English and French are fairly similar typologically, it might appear that Canadian data would not be particularly revealing in the study of linguistic effects of language contact. However, there are striking differences between linguistic borrowing in Acadian and in other varieties of Canadian French. For instance, Acadian French and Quebec French both borrow verbs, but, as we shall see, only certain Acadian varieties and some western Canadian ‘Quebec’ varieties have borrowed prepositions. In the following sections I turn to a detailed look at the French of Prince Edward Island (see Map 2). In Prince Edward Island, my work has focussed on the French of two small villages. One community from the Évangéline region, Abram-Village, and one from Tignish region, Saint-Louis, were chosen for study. There are striking contrasts between the two communities: in AbramVillage the vast majority are of French origin and report speaking French, or both French and English, in the home. The village boasts a French-medium school which serves the surrounding area, and most essential services are provided in French. On the other hand, Saint-Louis is in a state of language
CROSSING GRAMMATICAL BORDERS
Map 1: The Atlantic Provinces.
Map 2: Prince Edward Island.
239
RUTH KING
240
loss: there is little institutional support for French, and over the last few decades French has rapidly lost ground, even as a home language. In 1987, twenty-four Abram-Village residents and nineteen Saint-Louis residents were interviewed by native Acadian French speakers who were born and raised in the community, using conversation modules designed to elicit approximately two hours’ worth of conversation per individual. In the following year, the same individuals were re-interviewed, this time by a speaker of European French, a community outsider, in order to tap a more formal style. The interview data were supplemented by participant observation and by data elicitation with native speakers. The English-origin data found in these corpora were divided into two groups. One group consisted of what I consider unambiguous codeswitches, i.e. English multi-word sequences not integrated morphologically or syntactically into French, along with single-word tokens involving translation or metalinguistic commentary. The second group consisted of the remaining tokens of English origin, just over 8200 words for each community for the main corpora, which I broke down according to parts of speech by community and by type vs token. Twelve different English-origin prepositions (types) occurred in the Abram-Village corpus, with a total of 257 occurrences (tokens). Sixteen different prepositions, with a total of 405 occurrences, were found in the corpus for the other village under study, Saint-Louis. However, while the social circumstances of the two communities differ and English-origin prepositions occur more frequently in Saint-Louis French, the actual patterning of preposition usage was found to be the same for the two communities; likewise there are no significant differences according to age, sex or linguistic marketplace ranking. Therefore, the data presented below are representative of Prince Edward Island French in general. 5.
Prepositions Stranding in Prince Edward Island French The data in (1)–(4) show prepositions occurring without an adjacent lexical complement, in English, French, Italian and Spanish. (1)
Who did you talk to?
(2)
*Qui as-tu parlé à?
(3)
*Cui hai parlato a?
(4)
*Quien has hablado a?
CROSSING GRAMMATICAL BORDERS
241
The phenomenon is called Preposition Stranding because movement, in these cases wh-movement, leaves behind a “stranded” preposition. 3 Preposition Stranding is quite rare among the world’s languages, and is widely accepted not to occur in French, nor indeed in any Romance language (cf. van Riemsdijk 1978; Zribi-Hertz 1984). The data in (5)–(7), then, taken from the Prince Edward Island Main Interview Corpora, are unexpected. (5) involves a whquestion, (6) a pseudo-passive and (7) a relative clause. In none of these examples is there an overt complement of the preposition: (5)
Quoi ce-que tu travailles dessus? 4 what that you work on “What are you working on?”
(6)
Le ciment a été marché dedans avant d’être sec. the concrete has been walked in before to-be dry “The concrete was walked on before being dry.”
(7)
Lui, c’est le gars que je travaille pour. him he is the guy that I work for “Him, he’s the guy I work for.”
If we apply the ‘morph-by-morph translation’ diagnostic used by Gumperz and Wilson in their Kupwar study, this is a prima facie case of syntactic borrowing. And, indeed, there are linguists who have made that claim for other varieties of Acadian French (cf. Gérin 1984). However, I argue that the picture is more complicated than that, and, ultimately, that the argument for strictly syntactic borrowing doesn’t hold up. We begin by taking a look at what appear at first glance to be similar data for other varieties of French. There are some discourse contexts in standard French in which an orphan preposition may occur: some examples, taken from Le bon usage (Grevisse 1986, cited by Roberge and Rosen 1999), are presented in (8). They have been given a variety of explanations in the 3 I use the term orphan preposition to refer to any case in which the preposition does not have an adjacent lexical complement and reserve the term Preposition Stranding for cases of movement, where the empty object position is argued to be filled by a trace. 4 Quoi ce-que (lit. ‘what that’) is an example of Acadian French’s so-called Doubly-filled COMP. Wh-movement occurs across an overt complementizer, ce-que, which I analyze elsewhere as an allomorph of the complementizer que. Ce-que occurs under agreement with French-origin wh-words when Wh-movement takes place; lone wh-words occur in situ. (cf. King 1991).
RUTH KING
242
literature: as recoverable through discourse linking (Martinet et al. 1979), as resulting from deletion (Arrivé et al. 1986), or as pro (Tuller 1986; Zribi-Hertz 1984). (8)
a. Et vous coulez avec. and you sink with (it) b. Tu n’es pas fait pour. you are not made for (it) c. Il a écrit des poèmes avec rimes et des poèmes sans. he has written poems with rhymes and poems without (them)
Both standard French and colloquial French allow an orphan preposition in topicalized structures such as in the example in (9), taken from Zribi-Hertz (1984). (9)
Cette valise, je voyage toujours avec. this suitcase I travel always with “This suitcase, I always travel with it.”
In addition to allowing the structures in (8) and (9), many colloquial varieties of French, including Quebec French, allow orphan prepositions in relative clauses, as in (10), taken from Bouchard (1982). (10) la fille que je sors avec the girl that I go-out with “the girl that I go out with” While any number of writers have considered such data as resulting from English influence, Bouchard (1982) has pointed out that this assertion is suspect, given that the structure existed in French in the 14th century and still exists in some European varieties not in contact with English. Both Bouchard and Zribi-Hertz have presented arguments that while an example such as (10) looks superficially like the English equivalent, it is in fact something quite different. For the English equivalent the standard analysis is a movement analysis with a null relative operator, a silent counterpart to ‘who’, which moves to spec-CP and leaves behind a trace. However, if that were the case for the French in (10), we would not be able to explain the contrast in grammaticality between (11a) and (b).
CROSSING GRAMMATICAL BORDERS
243
(11) a. la fille CP[que je connais très bien DP[le gars CP[qui sort avec]]] b. *the girl that I know very well the guy who went out with (11b) is a ‘morph by morph’ translation of (11a); yet the French is grammatical while the English is not. The idea is that wh-movement displays subjacency effects. Both (11a) and (b) contain complex relative clauses. If movement is involved the extracted element would cross two bounding nodes, CP and TP, which would be too ‘long’ a movement in both English and French. This is the case for the English sentence in (11b), where a movement analysis is the standard analysis and the ungrammatical result completely predictable. However (11a) is perfectly grammatical so we assume Preposition Stranding is not involved. Bouchard and Zribi-Hertz present convincing arguments that the empty complement position in (10) is filled by pro, not a wh-trace, i.e. movement is not involved in the colloquial varieties that allow such structures. Thus Quebec French and English relatives are different with respect to orphan prepositions. Another argument showing that the resemblance is only superficial is provided by Marie-Thérèse Vinet (1984). She notes that in (12)–(15) the French examples are grammatical but their English equivalents are not, which Vinet takes as evidence of the superficial nature of the French/English similarity with respect to orphan prepositions. (12) j’ai voté pour *I voted for (13) j’ai invité Marie pour danser avec *I invited Mary to dance with (14) parler avec a toujours été difficile *talking with has always been difficult (15) c’est pas facile de parler avec *it is not easy to talk with 6.
Preposition Stranding in Prince Edward Island French Prince Edward Island French allows all of the structures in (1)–(10); it does not, however, allow (11a), a point to which we shall return. It also allows structures such as in (16) and (17), with à (to) and de (of) occurring as orphan prepositions in relative clauses.
244
RUTH KING
(16) le gars que je te parle de the guy that I you talk of “the guy I am talking of” (17) le gars que j’ai donné la job à the guy that I have given the job to “the guy I’ve given the job to” The data in (16)–(17) are surprising. Zribi-Hertz (1984) states that those prepositions which are acceptable as orphan prepositions are most of the locative prepositions, such as contre ‘against’ and devant ‘in front of’, time prepositions such as avant ‘before’ and pendant ‘during’, and just a few others, such as avec ‘with’ and sans ‘without’. She states categorically that à and de never occur as orphan prepositions in any variety of French. Vinet, for her part, says that she has heard examples such as (17) in Montreal French but states that this is very advanced, quite limited usage which people would find decidedly odd. However, examples such as (16) and (17) are common in Prince Edward Island French. Crucially, the Prince Edward Island data show evidence of extraction. Relative clauses, pseudo-passives and wh-questions all display subjacency effects. 5 For example, while (18) involves simply movement of the DP to spec-CP, the unaccepable (19) involves movement from a complex DP. Recall as well the unacceptability of (11a) in Prince Edward Island French. Whenever the possible linking of an empty position and its antecedent can be seen as subject to subjacency we can conclude that movement is involved, in these cases Preposition Stranding. (18) Quii ce-que tu connais ti? who that you know “Who do you know?” (19) *Quii ce-que TP[tu connais DP[le projet à ti]]? who that you know the project of *”Who do you know the project of?” (20) and (21) show stranded de and à in wh-questions while (22) Prince Edward Island French allows the stranding of English-origin prepositions.
5
This analysis first appeared in King & Roberge (1990).
CROSSING GRAMMATICAL BORDERS
245
(20) Où ce-qu’elle vient de? where that she comes from “Where does she come from?” (21) Quelle heure qu’il a arrivé à? what time that he has arrived at “What time did he arrive?” (22) Quoi ce-qu’ils parlont about? what that they are-talking about “What are they talking about?” Recall that the Prince Edward Island corpora contains many instances of English-origin prepositions. In addition to cases like (22) with about in on its own, the Abram-Village and Saint-Louis main corpora show 67 different English verb + preposition combinations, comprising 17% of the Englishorigin verbs in the two corpora. In the tables which follow, frequentlyoccurring combinations are in bold. bailer out ganger up piler up slipper out bosser around getter along plugger up slower down builder up getter over puller out smoker up chickener out giver up puller through sporter up clearer up grower up putter up starter off cusser down hanger around reflecter off stepper out dier down hanger up runner around straightener out se dresser up kicker out runner about tier down dropper out layer off runner out turner down ender up looker up setter up turner on figurer out maker up shipper out turner out filler out mixer up shower off turner over finder out passer out shower up walker out fooler around picker up shutter out walker up freaker out picker from signer in washer out fusser pour picker out singler out Table 1: Verb + Preposition Combinations in the Abram-Village Main Corpus (-er is the French infinitival marker).
246
RUTH KING
backer up getter over puller through slower down bosser around giver up putter up stepper in clearer up grower up reflecter off stepper out cusser down hanger around runner around tier down se dresser up layer off runner out turner down figurer out looker up setter up turner on filler out maker up shower up turner out finder out mixer up shutter off walker out fooler around plugger up signer in walker up getter along puller out slipper out Table 2: Verb + Preposition Combinations in the Saint-Louis Main Corpus.
We also find English-origin verb + preposition (23)–(24), French-origin verb + English-origin preposition (25)–(26), and English-origin verb + Frenchorigin preposition combinations (27)–(28), all of which allow Preposition Stranding: (23) Ils avont layé off du monde à la factorie. they have laid off some people at the factory “They have laid off people at the factory.” (24) Qui ce-qu’a été layé off? who that has been laid off “Who were laid off?” (25) Il a parlé about le lien fixe. he has talked about the link fixed “He talked about the fixed link.” (26) Quoi ce-qu’il a parlé about? what that he has talked about “What did he talk about?” (27) L’avion a crashé dans la grange. the airplane has crashed into the barn “The airplane has crashed into the barn.” (28) Quoi ce-que l’avion a crashé dedans? what that the plane has crashed into “What did the plane crash into?”
CROSSING GRAMMATICAL BORDERS
247
Such data, I suggest, are key to understanding Preposition Stranding in Prince Edward Island French. The presence of borrowed prepositions always coincides with the presence of Preposition Stranding. We have seen that Prince Edward Island French borrows prepositions and has Preposition Stranding. Quebec French, on the other hand, does not borrow prepositions and does not have Preposition Stranding. Moncton, New Brunswick French borrows prepositions and appears to have Preposition Stranding, according to data presented in Roy (1979). Another Acadian variety, the highly conservative Newfoundland French, like Quebec French does not borrow prepositions and does not have Preposition Stranding. These are suggestive correlations but they do not show clearly the relationship between borrowing lexical items and borrowing grammatical structure. In what follows I seek to answer the question of whether or not Prince Edward Island French Preposition Stranding is identical to English Preposition Stranding and the question of whether or not there is evidence for direct borrowing. Syntacticians have long noted that not just any preposition, in any context, can be stranded. To take an example from Hornstein and Weinberg (1981), all native speakers whom I have polled find (29), starred by these authors, to be at least odd, if not outright ungrammatical: (29) *Who did Pugsley give a book yesterday to? Simply put, the positioning of the preposition following the adverb yesterday prohibits the close syntactic relationship needed for the reanalysis of the verb plus preposition as a complex verb. However, the Prince Edward Island corpora contain examples which resemble (29). Consider the data in (30)–(35). (30) Tu as parlé à Jean de ça hier. You have spoken to John about that yesterday “You spoke to John about that yesterday.” (31) Quoi ce-que tu as parlé hier à Jean de? “What did you speak yesterday to John about?” (32) Quoi ce-que tu as parlé à Jean hier de? “What did you speak to John yesterday about?”
248
RUTH KING
(33) Quoi ce-que tu as parlé à Jean de hier? “What did you speak to John about yesterday?” (34) Quoi ce-que tu as parlé de à Jean hier? “What did you speak about to John yesterday?” (35) Quoi ce-que tu as parlé hier de à Jean? “What did you speak yesterday about to John?” In Prince Edward Island French there is considerable leeway in terms of where stranding can take place: (31)–(35) are all acceptable whereas native speakers of English find the translations to vary in grammaticality. The extreme freedom of Preposition Stranding indicates that the structural relationship between the verb and the preposition found in English is not relevant in Prince Edward Island French, understandable since French does not have the strong adjacency requirements found in English in a variety of constructions. These differences between Preposition Stranding in English and in Prince Edward Island French may be explained by the borrowing of prepositions: the direct borrowing of English-origin prepositions has triggered in Prince Edward Island French an extension of a property of English prepositions, i.e., the ability to be stranded (technically the ability of the preposition to govern the trace) to the whole set of Prince Edward Island French prepositions. However, the syntactic mechanisms associated with the structure in English, involving reanalysis of the verb plus preposition when they appear in particular configurations, have not been borrowed. An objection which might be raised at this point is that there is another explanation of the data in (31)–(35), i.e., one might argue that there has been overgeneralization of an English rule, to the point that ‘anything goes’. However, there are clear limits on grammaticality in Prince Edward Island French, since we saw that Preposition Stranding in this variety shows clear Subjacency effects. Secondly, an analysis in terms of overgeneralization would not be able to account for crossdialectal variation in North American French varieties in terms of what prepositions may be stranded. Roberge and Rosen (1999) present evidence that while the French of the Canadian province of Alberta and the French of Louisiana both allow Preposition Stranding they do not allow the stranding of the prepositions à or de. Lexical variation across varieties is a finding at odds with an argument in terms of overgeneralization of an English rule.
CROSSING GRAMMATICAL BORDERS
249
7.
Conclusion While at first glance Prince Edward Island French looks like it provides good evidence for direct syntactic borrowing, on closer inspection, data from this variety support a view of lexical borrowing having syntactic effects in the recipient language. In the discussion above lexical aspects of Preposition Stranding were separated from other aspects of the phenomenon, allowing an explicitly characterization of the changes which have taken place in the variety. In the theoretical perspective I adopt, the lexicon plays a central role in the production and comprehension of sentences: not surprisingly, then, we see that lexical borrowing may have grammatical repercussions. References Arrivé, M. et al. 1986. La grammaire d’aujourd’hui. Flammarion: Paris. Bouchard, D. 1982. “Les constructions relatives en français vernaculaire et en français standard: étude d’un paramètre”. La syntaxe comparée du français standard et populaire: approches formelles et fonctionnelles ed. by C. Lefebvre, 103–134. Québec: Office de la langue française. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. ----------. 1993. “A Minimalist Program for Linguistic Theory”. The View from Building 20 ed. by K. Hale & J. Keyser, 1–52. Cambridge, Mass.: MIT Press. Forget, D. & F. Martineau. 2002. Des identités en mutation dans l’ancien et le nouveau monde. Ottawa: Éditions David. Gérin, P. 1984. “Emplois aberrants de quelques prépositions dans le français des Acadiens”. Papers from the Seventh Annual Meeting of the Atlantic Provinces Linguistic Association ed. by H. Zobel, 39–53. Moncton: Université de Moncton. Gérin, P.M. 1984. “La création d’un troisième code comme mode d’adaptation à une situation où deux langues sont en contact, le chiac”. Papers from the Seventh Annual Meeting of the Atlantic Provinces Linguistic Association ed. by H. Zobel, 31–38. Moncton: Université de Moncton. Grevisse, M. 1986 [12th ed.]. Le bon usage. Duculot: Paris. Gumperz, J.J. & R. Wilson. 1971. “Convergence and Creolization: A Case Study from the Aryan/Dravidian Border in India”. Pidginization and Creolization of Languages ed. by D. Hymes, 151–167. Cambridge: Cambridge University Press. Harris, J. 1991. “Conservatism versus Substratal Transfer in Irish English”. Dialects of English ed. by P. Trudgill & J. K. Chambers, 191–214. London: Longman. Hornstein, N. & A. Weinberg. 1981. “Case Theory and Preposition Stranding”.
250
RUTH KING
Linguistic Inquiry 12.55–91. Kayne, R. 1980. “De certaines différences entre le français et l’anglais”. Langages 60.47–64. King, R. 1991. “Wh-words, Wh-questions, and Relative Clauses in Prince Edward Island French”. Canadian Journal of Linguistics 36:1.65–85. ----------. 2000. The Lexical Basis of Grammatical Borrowing: A Prince Edward Island French Case Study. Amsterdam and Philadelphia: John Benjamins. King, R. & T. Nadasdi 1997. “Left Dislocation, Number Marking and Canadian French”. Probus 9.267–284. King, R. & Y. Roberge. 1990. “Preposition Stranding in Prince Edward Island French”. Probus 2.351–369. Martineau, France. 2002. “Un corpus de textes français pour l’analyse de la variation diachronique et dialectale”. Paper presented at the Congrès de l’ACFAS. Martinet, A. et al. 1979. Grammaire fonctionnelle du français. Didier: Paris. McMahon, A. 1994. Understanding Language Change. Cambridge: Cambridge University Press. Mous, M. 1996. “Was there ever a Southern Cushitic Language (Pre-) Ma’a?” Cushitic and Omotic Languages. Proceedings of the Third International Symposium, Berlin ed. by C. Griefenow-Mewis & R.M. Voigt, 201–211. Köln: Rüdiger Köppe Verlag. ----------. 1997. “The e/a Alternation in Mbugu: The Limits of Allomorphy”. Linguistics in the Netherlands 1997 ed. by J. Coerts & H. de Hoop, 123– 134. Amsterdam and Philadelphia: John Benjamins. ----------. 2001. “Ma’a as an Ethno-Register of Mbugu”. Sprache und Geschichte in Afrika 16/17.293–320. Poplack, S. & A. St-Amand. 2002. “The Emergence of Quebec French: Constituting a Corpus of 19th Century Speech”. Paper presented at NWAVE-31, Stanford University. Riemsdijk, H. van. 1978. A Case Study in Syntactic Markedness. Lisse: Peter de Ridder Press. Roberge, Y. & N. Rosen. 1999. “Preposition Stranding and que-Deletion in Varieties of North American French”. Linguistica Atlantica 21.153–168. Roy, M.M. 1979. Les conjonctions anglaises ‘but’ et ‘so’ dans le français de Monction. Une étude sociolinguistique de changements linguistiques provoqués par une situation de contact. M.A. thesis, Université du Québec à Montréal. Sankoff, G. 2002. “Linguistic Outcomes of Language Contact”. The Handbook of Language Variation and Change ed. by J.K. Chambers, P. Trudgill &
CROSSING GRAMMATICAL BORDERS
251
N. Schilling-Estes, 638–668. Oxford: Blackwell. Thomason, S. 2001. Language Contact: An Introduction. Washington, D.C.: Georgetown University Press. Thomason, S. & T. Kaufman. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley and Los Angeles: University of California Press. Tuller, L. A. 1986. Bijective Relations in Universal Grammar and the Syntax of Husa. Doctoral dissertation, University of California at Los Angeles. Vinet, M.T. 1984. “La syntaxe du québécois et les emprunts à l’anglais”. Revue de l’Association québécoise de linguistique 3.221–242. Weinreich, U. 1953. Languages in Contact. Mouton: The Hague. Zribi-Hertz, A. 1984. “Prépositions orphelines et pronoms nuls”. Recherches linguistiques 12.46–91.
THE AFTER-PERFECT IN IRISH ENGLISH PATRICIA RONAN Universität Bonn 1.
Introduction In this paper a well-known and salient aspect of Irish English, also termed Hiberno-English, will be described and evaluated, namely the afterperfect, as in A man is after getting shot in the street. This construction is generally taken to denote ‘hot news’ events, and not intentionality as in Standard English. This category will be examined against the background of not only other intra-dialectal means to describe situations in the perfect, but also as compared to the use of perfect marking by the ‘have-perfect’. In order to examine the category in question, a brief survey of meanings of perfects is given and the Hiberno-English dialectal perfect variants are introduced. Against this background we will first discuss the after-perfect and its development in Irish. In the following the rise of the structure in HibernoEnglish will be re-evaluated; we will find that it does not originally denote hot news. Secondly, the after-perfect in modern Hiberno-English will be examined as to its distribution and function. It can be observed that significant differences manifest themselves when distribution with respect to that of the standard variety is taken into account. 1 2.
The perfect in English In English, as in other languages, the category of the perfect can have different shades of meaning. To assess its meaning it is important to note that perfect and perfectivity are distinct: perfectivity is an aspect category opposed to imperfectivity and refers to an action or situation which is not described as having a specific duration but is viewed in its entirety without reference to its internal structure (cf. Comrie 1976: 16). The perfect in English, on the other hand, is used to express the temporal location of an event prior to the moment of utterance, which may coincide with the ‘speaker-now’, but does not necessarily have to. The event is perceived as 1 Research funded by an Irish Research Council of the Humanities and Social Sciences scholarship.
PATRICIA RONAN
254
having relevance for the present situation. In English, as in other languages, this relevance can have different manifestations, which will be briefly introduced here, following Comrie (1976). The perfect in English is periphrastically constructed, with the lexical verb and the auxiliary BE or HAVE, and in this it follows a cross-linguistic tendency (cf. Dahl 1985: 129). First of all, the ‘perfect of persistent situation’ can be observed: duration of a situation or action from a point of time in the past until the present moment is expressed: (1)
I have known Max since 1960. (McCawley 1976: 263)
Furthermore, it may be expressed that situations in the past may have an impact on the present moment. This is indicated by the ‘perfect of result’: (2)
I can’t come to your party tonight – I have caught the flu. (McCawley 1976: 263)
Thirdly, the ‘experiential perfect’ describes that experiences made at some stage between past and present still have relevance for the present moment: (3)
Have you ever studied French?
A further group, which however is rather marginal in Standard English (StE), is the BE-perfect. This is a state perfect formed with the auxiliary BE and a restricted number of verbs, particularly GO. (4)
He is gone now.
Occasionally the perfect is also employed in order to indicate that something happened recently. The event has just taken place and is still present in the mind of the speaker: (5)
I’ve just seen a shooting star.
This category we will meet again in connection with the Hiberno-English (HE) after-perfect. What all types of perfect have in common is the connection between an event in the past and the ‘speaker-now’.
THE AFTER-PERFECT IN IRISH ENGLISH
3.
255
Types of perfects in Hiberno-English Compared to the situation in StE there are more possibilities to express different manifestations of the perfect in HE. While in StE only the HAVEperfect, and vestigially the BE-perfect, are used, Irish English offers the possibility to employ both StE and additionally various dialectal types. Both Kallen (1989, 1990, 1991) and Harris (1984), in studies of Dublin and Ulster dialects respectively, have pointed out that both standard and dialectal variants are employed for perfect marking in Irish English. While differences in the use of the standard English perfect were observed in Ulster between rural and urban speakers, this cannot be the case in the urban Dublin context. The data considered here stem from two sources: data have been collected in participant observation from speakers of different varieties of Irish English. Insights and data collected from this source are supplemented by material from interviews, led in Dublin, which have been collected by the American sociologist K. Kearns. The interviewees were elderly people living in Dublin, who were mostly, but not exclusively, born and raised locally. The informants were predominantly working-class. Some class-conscious speakers like school-teachers or Sunday-school teachers did feature and their speech patterns were generally observed to be more standardised. While the ultimate authenticity of the latter data could not be checked, the features observed in that corpus are corroborated by similar observations in the former source of data. Material taken from published texts has been included because, in contrast to participant observation, the texts offered the possibility to quantify the data in question with regard to the unmarked variants, which would otherwise go unnoticed. In order not to blur the numbers of the different categories observed here, in quantifications only the numbers from the source texts will be given and examined in the tables in this paper. The approach used here is to treat not only constructions which are overtly perfect-marked with HAVE or BE auxiliaries, but also those which are functionally equivalent to the StE perfect as perfect tokens. In the HE texts examined, out of the 311 tokens of perfect use, 58.5% are StE HAVE-perfects while the remaining 41.5% are distributed among five dialectal variants. Thus in 35 cases (11.3% of all perfect tokens) in the environment of a StE perfect, preterite forms are used with reference to an indefinite moment in the past leading up to the present; temporal reference is underlined by an indefinite adverbial: (6)
It was one of the most interesting places I ever worked.
PATRICIA RONAN
256 (7)
All me life, from me grandfather and father, pigs were always part an’ parcel of the family.
This category is termed ‘Indefinite Anterior’ by Harris (1984) and Filppula (1999). Its reference equals that of the StE ‘experiential perfect’. A further category that does not contain overt perfect marking denotes states which began in the past and extend until the present moment; the verb is in the present tense, and temporal reference is likewise expressed by an adverbial. Kallen (1989) calls this category ‘extended present’-perfect, Filppula (1999) and Harris (1984) term it ‘extended-now’ perfect. Examples are the following: (8)
There is not a neighbour on the road I don’t know all my life.
(9)
Now I’m fifty-five years drinking in this pub.
This category is a manifestation of the ‘perfect of persistent situation’, which in StE is expressed by an overtly marked perfect. In the present corpus only 14 tokens (i.e. 4.5%) are of this type. Furthermore, the BE-perfect, only marginally used in StE, is employed with a distinct range in Hiberno-English to denote a state resulting from a prior action. Its token-frequency is notable with 24 tokens or 9% of all employed perfects: (10) I am finished doing it. (11) Most of those areas, like Henrietta Street and Dominick Street are all gone now. In addition to higher frequency than in StE, a greater variety of verbs is used in this type of perfect in HE, including some intransitive verbs, particularly those with “mutative” semantics (Harris 1984: 308). Rather infrequent in this corpus is a construction which is used to indicate a resultative state, consisting of HAVE, followed by its direct object and the past participle. This construction is usually termed ‘Accomplished Perfect’ or ‘Medial Object Perfect’. (12) When she died, we already had the grave bought. (13) But the woman, when she had her shopping done, she was never brought into the pub.
THE AFTER-PERFECT IN IRISH ENGLISH
257
There were 15 examples (i.e. 4.8%) found in this corpus, a figure which is significantly lower than that found by Filppula (1999) for the south-west of Ireland. However, Filppula himself observed a marked difference in its distribution in different parts of the country, noting that in the South-West the use of the Medial Object Perfect (MOP) was highest, whereas in the Dublin area it was significantly lower, with the counties Clare and Wicklow providing intermediate figures. It seems important to note that significant geographical variation is displayed here, and this feature will be further interpreted in the following section. The overall distribution of these categories is shown in Table 1. StE Perfect 182 tokens 58.5%
Indefinite ExtendedBe-Perfect Medial-Object After-Perfect Anterior Now Perfect Perfect 35 tokens 14 tokens 28 tokens 15 tokens 37 tokens 11.3% 4.5% 9% 4.8% 11.8% Table 1: Distribution of the perfect types in the present corpus.
While the categories mentioned above all have British English parallels, particularly in Early Modern English, as shown for instance by Harris (1984), the fifth variety, namely the after-perfect, does not have any comparable structures in any other variety of English. 4.
The after-perfect and its development The after-perfect is a type of perfect that is particular to Hiberno-English, and it is named after its distinct marker. It consists of a form of the copula BE, the preposition after and a present participle: (14) That was a good while ago, the place was only after opening then. In this type of example after has no intentional meaning, as it has in StE to be after something, i.e. to be intent on doing or want something, but it is normally analysed as a marker of ‘hot news’ perfect senses. It is thus described by scholars such as van Hamel (1912), Harris (1984), and Filppula (1999). Kallen, for his Dublin corpus, however, finds a broader range of uses. He notes that it is employed for further shades of perfect meaning, namely ‘perfect of persistent situation’, ‘existential perfect’, and ‘perfect of result’, thus largely spanning the ranges of the StE perfect. In the following we will have a short look at a parallel construction in the Irish language, at the development of the HE construction, and finally, examples collected from the modern language will be presented and analysed.
258
PATRICIA RONAN
For the after-perfect, in contrast to the controversial discussions surrounding the other types of HE dialectal perfects, it appears to be common consensus that this category was created by transfer of a parallel Irish language construction: (15) Tá sé tréis leitir a scríobh (Greene 1979: 122) BE, 3.sg. he after letter to write “He is after writing a letter.” Earlier English varieties offer no comparable structures which could have contributed to the creation of this salient feature of Irish English. The earliest examples in the Irish language can be found from about the Early Modern Irish period, which started about 1200 and lasted till the mid-seventeenth century (cf. Greene 1979: 122 ff.). At that time a different preposition meaning ‘after’ could be used, namely íar, which had a by-form ar. This can be observed in the following examples: 2 (16) A-taoi ar n-am chrádh BE, 2sg. pres. after my tormenting ‘You have tormented me.’ (Ddána §64, 3, cf. Greene 1979: 125) (17) […] mar adeir Iob san .30. caibidi., Comparatus sum luto, as says Iob in the 30th chapter ‘atáim arn-am choimmeas ré lathaigh BE, 1sg pres after my comparing against mud “As Iob says in the 30th chapter Comparatus sum luto, ‘I have been compared to mud.’” (Three Shafts, 328) (18) […] an choguais bhíos ar n-a uile-losgadh The conscience BE, pres.rel.sg. after its all-burning lé teine gradha Dé ar altoir na hanma. By fire of love of God on altar of the soul. “The conscience which has been burned completely by the fire of the love of God on the altar of the soul.” (Three Shafts, 387) 2
As an example of non-passive usage Greene (1979:125) also cites the following, without noting the source: A-táid mo thrí chéile -se ar ndul uaim. BE, 3rd pl MY THREE LORDS emph. AFTER GOING FROM, 1sg. “My three lords have gone from me.”
THE AFTER-PERFECT IN IRISH ENGLISH
259
We can see in these examples from Early Modern Irish that originally it is not recency of an event which is denoted, but rather the effect of an action. Greene points out that these constructions have a very low yield in Early Modern Irish and, where appearing, are most common in passives (1979: 125). He observes that in contrast to the Modern Irish constructions (as in [15]) they do not denote recent completion (1979: 128). As the perfect cross-linguistically usually expands its functional range, he proposes to treat the early examples as not fully grammaticalized but as collocations at that linguistic stage (1979: 129). Yet this type of construction would have been used in Irish at the formative stages of Hiberno-English, as illustrated by the examples form Three Shafts, and its predominant function in the Irish language could have had an impact on the developing HE dialect. For the synchronic level Greene points out that outside the Irish-speaking area, the Gaeltacht, this type of perfect is extending its functional range towards the simple past. Hiberno-English on the other hand, he argues, does not show expansion beyond the feature of recency since alternative structures are available due to the availability of the StE perfect. It will be seen in the following that this claim has to be modified. Looking at early examples of Hiberno-English it appears that a narrowing down of the meaning of the after-perfect construction can also be observed here. In his 1979 volume Spoken English in Ireland, Bliss published a collection of texts from the period 1600 to 1740. If the texts are taken at face value, even at this early stage a number of constructions involving the preposition after and a present participle can be found. Most of these have modal characteristics, however, and are formed with the future auxiliary will. (19) You vill [sic] be after being damn’d (Bliss 1979: 300, from Shadwell, Lancashire Witches 1681/2) This type of construction has recently been investigated by McCafferty (2003), who has argued that it originated in the Irish–English language contact situation and represents a combination of the Irish syntactic structure after + verbal noun and the English intentional semantics of that preposition. He argues that these intentional readings, together with spatial reference and movement towards a goal, lead to future meanings. This structure is shown to be prevalent in earlier Hiberno-English texts. On the other hand, the ability of after to also denote anteriority enables it to be used for past-time reference. In the data found in Bliss’s collection, only one example appears to have perfect reference:
260
PATRICIA RONAN
(20) You shee here de cause dat is after bringing you to dis plaace. (Bliss 1979: 300, from Dunton, Report of a Sermon, 1698) McCafferty argues that the perfect came to be denoted in HibernoEnglish only when the influence of the Irish language on the English spoken in the country increased in the late eighteenth century due to the large-scale shift from Irish to English. As a further possible scenario we could assume negative transfer by Irish-language learners who were confronted with the intentional structure in Standard English and, familiar with the structure in Early Modern Irish, wrongly analysed it as equivalent to the Irish after-perfect as described by Greene (1979) and Ó Sé (1992). Negative transfer can happen in a language-acquisition situation as presented by that in Ireland during the formative stages of HE where there was little contact with native or proficient speakers of English (cf. Odlin 1991: 186 and Bliss 1972: 63). In one example a combination of HAVE and after marking can be found: (21) [...] the Irish Brogue, fait Joy, he has been after wearing dem himself. (Bliss 1979: 146, from Michelburne, Ireland Preserved, 1705) This is the only example which displays this kind of double marking and the question arises whether any significance at all can be attached to it, and if so, what exactly the significance would be. Is it possible to argue on the strength of one example that this duplicity is in fact a sign that the two markers do not mark the same, but are distinct? Comparing Bliss’s earlier examples, resultative interpretation seems to be a possibility for all of them. Resultativity is also a possible outcome of a transfer of Irish tar éis or tréis to Hiberno-English. From the 19th century onwards, the after construction is increasingly grammaticalized as a perfect marker and is used to denote ‘hot news’ (Filppula 1999). Specification of the perfect-marker after to marking ‘hot news’ could have been triggered by competition with other resultative markers in Irish English. Considering now the use of the after-perfect in modern-day Ireland, it is noteworthy that this construction is attested less frequently in south-western dialects of HE than in more easterly dialects as visible in Table 2, compiled from Filppula (1999).
THE AFTER-PERFECT IN IRISH ENGLISH
261
Region MOP-tokens MOP-‰ AFP-tokens AFP-‰ West: Clare 14 0.47 3 0.1 Kerry 15 0.34 1 0.02 East: Wicklow 8 0.19 9 0.21 Dublin 3 0.07 12 0.29 Table 2: Frequencies of the AFTER-Perfect versus ‘Medial Object’ Perfect (compiled from Filppula 1999: 101, 110).
It can be seen from the data in this table that the ‘Medial Object Perfect’ and the after-perfect are used in almost complementary distribution. In the West the after-perfect appears as a marginal construction only from the numbers given here, and the attested examples of the after-perfect are confined to ‘hot news’ senses. The resultative ‘Medial Object Perfect’, on the other hand, is remarkably more frequent. Filppula (1999) points out that the differences in attestation may be due to the fact that the after-perfect has only marginal status in the Irish language in the South-West. In the eastern dialects of HE the opposite can be observed. The ‘Medial Object Perfect’ has only marginal status, according to Filppula’s data, and the after-perfect has a much higher frequency. This observation suggests the possibility that an element of complementary distribution may interfere in its usage. How the use of the after-perfect in a Dublin context can be described will be illustrated by the following collection of quantifiable data. The after-perfect in the present corpus Of all the dialectal perfect variants in the present corpus, the after-perfect is the most frequently attested with 37 tokens (12% of all tokens). A varied distribution was found in the data, representative of categories of universally observed perfect usage. In 12 cases (32% of all after-perfect tokens) these were examples of ‘hot news’-perfect. This can appear both as a temporally or conceptually close event: 5.
(22) Going back to my fried Billy Fox now, he was only after starting his business and Guinness’s gave him no credit. (23) Many a woman, she’d only be six or seven weeks after having the baby and she’d be expecting again! Yet in 25 cases (68% of all after-perfect tokens) the employment of this marker does not, or not exclusively, seem to be due to temporal or conceptual
262
PATRICIA RONAN
proximity, but further meaning components are clearly visible. Thus it is said about prostitutes that they were (24) Mostly girls from the country who were after being in [domestic] service and went wrong [i.e. became pregnant]. Here the main emphasis is on the result of an event. This also holds true for the following example: (25) So I went in and when I went in she was laid out in a habit – she was after dying. In these cases resultativity is more prominent than the ‘hot-news’ meanings. And in some cases both features, resultativity and ‘hot-news’, seem to appear, rather than only one meaning component: (26) If he was after doing a play or writing a story and it had been accepted [for money] he’d then go on a binge [...]. (27) But I thought somebody was after seeing me going in and they went for my mother. These events could be analysed either as recently happened or also as denoting the reason for an event, thus receiving a resultative interpretation. This ambiguity can be observed in 7 cases (19 % of the after- tokens). Furthermore, obvious resultativity without ‘hot-news’ senses is observable in additional examples collected from a speaker from the south-eastern part of Co. Wexford: (28) I’m after doing it so many times now, I just get a dirty look every time. (29) If you look at editions [of manuscripts], they [editors]’ll be after putting the glide-vowel in for you. Explaining the absence of certain syllables in 8th-century Irish the following sentence was uttered: (30) Whenever I write a cross that means they [syllables] are after being syncopated.
THE AFTER-PERFECT IN IRISH ENGLISH
263
These examples were noted in an in-class teaching situation. They are probably the clearest examples of the result of an event and, though not stemming from the quantifiable written corpus, provide further evidence of the diverse uses of the construction. And as already observed by Kallen (e.g. Kallen 1989), further shades of perfect meaning can be observed. Thus in the present corpus six examples (16%) seem manifestations of the experiential perfect, e.g. (31) I’m after paying 12 pounds for a pram for Tony forty-seven years ago that wasn’t worth one pound. One example may be either ‘experiential perfect’ or ‘perfect of persistent situation’: (32) Eventually you rambled off home about nine or ten in the morning and there’d be one or two [who] would be after sleeping all night. It is not surprising that after-perfects are only seldom observed in ‘perfect of persistent situation’-senses, as neither ‘hot-news’-perfects nor resultatives seem well suited to mark persistent situations. Nevertheless, it is clearly visible that the after-perfect is used not only to mark ‘hot news’ but overall its distribution is parallel to that of the StE perfect. The distribution in the present corpus can be observed in Table 3. Total ‘hot news’ ‘resultative’ ‘resultative’ or ‘experimental’ ‘experiental’ or (of 311 perfect ‘persistent’ ‘hot news’ tokens) 37 examples 12 11 7 6 1 11.8% of 311 32% 30% 19% 16% 3% Table 3: Distribution of meanings of the AFTER-perfect.
We can see that of the total of 311 perfect tokens, 182 of which were Standard English have-perfects, 37 were instances of the after-perfect. Of those, the biggest group is ‘hot-news’, but this is closely followed by resultatives. Categories that have no elements of recency or result are less prominent. Yet they still feature in low numbers. These results showing a width of different meanings are not unexpected in the light of Kallen’s research. They could be explained, as suggested by Hickey (2000), in terms of a prototype approach: Hickey analyses the after-perfect as having the core meaning of immediacy as opposed to result. Yet a transition from recency to result could
264
PATRICIA RONAN
conceptually be possible, as recent events would most typically have the results we are concerned with. ‘Hot news’senses are the single most frequently observed category in this corpus and may be analysed as forming part of the core meaning of perfect constructions marked by after. The data presented suggest that resultative meanings account for another central meaning component. But the data become even more instructive when we put it into the context of the overall use of perfect markers by the language users. Depending on whether or not the speakers use ‘have’-perfects in their speech, significant differences can be shown in the employment and meaning of the after-perfects. Some speakers use StE perfects as well as well as after-perfects while others do not use StE perfects at all, or they use ‘have-perfects’ only in special contexts, following another auxiliary: in the interviews examined, a total of 28 speakers used afterperfect constructions. Of these, 8 speakers used it in ‘hot-news’senses exclusively. All these speakers also used ‘have-perfects’ to indicate other than ‘hot-news’-perfect senses. A different group of 20 speakers used the afterperfect in various contexts as a general perfect marker. Of these a clear majority, 14 speakers, only employed dialectal perfect variants, and no ‘haveperfects’ at all. 6 speakers used no “ordinary” ‘have-perfects’, but used them in special morpho-syntactic contexts, such as pluperfects (33) or modal perfects (34): (33) He was after doing a play or writing a story and it had been accepted, he’d then go on a binge [...]. (34) [...] you kept your mouth shut and you done things that you shouldn’t have done. In both these instances ‘have-perfects’ are used in these special contexts, which might be influenced by StE pressure on special formations, but the ‘have-perfect’ is avoided in other possible contexts as illustrated by example 33). These differences are illustrated in Figure 1. The conclusion to be drawn from these observations is that for speakers observed in this corpus, those who used StE ‘have-perfects’ use the afterperfect only in the special context of a ‘hot-news’ perfect. For a second group of speakers, who do not use StE ‘have-perfects’, the after-perfect has grammaticalized to such a degree that it takes over multiple functions of the StE ‘have-perfect’, with some perfect senses optionally expressed by other dialectal perfect variants. This tendency was observed for speakers of Dublin
THE AFTER-PERFECT IN IRISH ENGLISH
265
28 Speakers’ use of ‘after perfect’
20 Speakers General Perfect-marker
14 Speakers No StE Perfects
6 Speakers StE Perfect only for Pluperf. and modal Perfects
8 Speakers HOT NEWS Perfect only
8 Speakers also StE Perfects
Figure 1: Use of AFTER -perfects by speakers who employ or do not employ ‘HAVE-perfects’.
and Wexford dialects in particular. Yet a similar situation seems to hold for a different variety of Irish influenced English, namely in the Newfoundland dialect. 3 6.
Conclusions Considering the data in hand it seems plausible to suggest that the afterperfect in HE is not one unified category but needs to be considered in distribution with other “marked” dialectal perfect variants, and also in its distribution as opposed to the usage of unmarked, StE perfect markers. While for speakers who use StE ‘have-perfects’, ‘after’-perfects appear to have grammaticalized to denote ‘hot-news’ senses only, for other speakers the afterperfect has grammaticalized as an alternative strategy for perfect marking. In the latter case the marking could be based on resultative marking, which is a central element of perfect meanings. Though further research is needed into the development of the category in the Irish language, it appears that the Irish construction has experienced a similar shift from an original resultative marker to one marking recency. Thus the influence of the Irish language on the HE construction may originally not have been one of recency, but of broader perfect meaning. Even though perfect markers are cross-linguistically observed
3
I am grateful to Dr. Sandra Clarke of the Memorial University of NFLD for pointing this out to me.
266
PATRICIA RONAN
to rather broaden their meaning than to restrict it (cf. Comrie 1976:60), it appears to be the case that the after-perfect has done just this in Irish and Irish English, and that this restriction is due to competition with other perfect markers where used. Where the construction is used in a non-‘hot news’ sense in Irish English, this might therefore be due either to a retention of older dialect features or else a later broadening of the range of the after-marker, ousting the have-perfect in non-stative senses for some speakers or dialects. 4 7. Abbreviations Ddána: Dioghluim Dána. Ed. Mac Cionnaith. HE: Hiberno-English Pres.: Present tense Rel.: Relative Form MOP: Medial Object Perfect Sg.: Singular StE: Standard British English Three Shafts: Trí Bior-Ghaoithe an Bháis. Keating, G. Ed. By Bergin, O. References Anderson, L. B. 1979. “The ‘Perfect’ as a Universal and as a Languagespecific Category”. Tense-Aspect: Between Semantics & Pragmatics ed. by P.J. Hopper, 227–264. Amsterdam and Philadelphia: John Benjamins. Bliss, A. 1972. “Languages in contact. Some problems of Hiberno-English”. Proceedings of the Royal Irish Academy 72, Section C, 63–82. ----------. 1979. Spoken English in Ireland: 1600–1740. Dublin: Dolmen Press. Bybee, J., R. Perkins & W. Pagliuca. 1994. The Evolution of Grammar. Tense, Aspect and Modality in the Languages of the World. Chicago and London: The University of Chicago Press. Christian Brothers, The. 1997. New Irish Grammar. Dublin: Fallon. Comrie, B. 1976. Aspect. Cambridge: Cambridge University Press. ----------. 1985. Tense. Cambridge: Cambridge University Press. Dahl, Ö. 1985. Tense and Aspect Systems. Oxford and New York: Blackwell. Denison, D. 1993. English Historical Syntax: Verbal Constructions. London and New York: Longman. Filppula, M. 1996. “Investigating the Origins of Hiberno-English Perfects: The Case of ‘PII’”. Speech Past and Present: Studies in English Dialectology 4
Ó Sé’s article ‘The after perfect and related constructions in Gaelic dialects’ in Ériu 54, 2004 only appeared after this article was written and unfortunately could not be used for this discussion.
THE AFTER-PERFECT IN IRISH ENGLISH
267
in Memory of Ossi Ihalainen ed. by J. Klemola, Merja Kytö & Matti Rissanen, 33–55. Frankfurt am Main: Peter Lang. ----------. 1997. “The Influence of Irish on Perfect Marking in Hiberno-English. The Case of the ‘Extended-now’ Perfect”. Focus on Ireland ed. by J. Kallen, 51–69. Amsterdam and Philadelphia: John Benjamins. -----------. 1999. The Grammar of Irish English. Language in Hibernian Style. London: Routledge. Greene, D. 1979. “Perfects and Perfectives in Modern Irish”. Ériu 30.122–141. Harris, J. 1984. “Syntactic variation and dialect divergence”. Journal of Linguistics 20.303–327. ----------. 1993. “The grammar of Irish English”. Real English: The Grammar of English Dialects in the British Isles ed. by J. Milroy and L. Milroy, 139–186. London and New York: Longman. Henry, P.L. 1957. An Anglo-Irish Dialect of North Roscommon. Dublin: Department of English, University College Dublin. Hickey, R. 2000. “Models for describing aspect in Irish English”. Celtic Englishes II ed. by H.L.C. Tristram, 97–116. Heidelberg: Universitätsverlag C. Winter. Huddleston, R. 1984. Introduction to the Grammar of English. Cambridge: Cambridge University Press. Kallen, J.L. 1989. “Tense and Aspect Categories in Irish English”. English World Wide 10.1–39. ----------. 1990. “The Hiberno-English Perfect: Grammaticalisation Revisited”. Irish University Review 20.120–136. ----------. 1991. “Sociolinguistic Variation and Methodology: After as a Dublin Variable”. English around the World: Sociolinguistic Perspectives ed. by Jenny Cheshire, 61–74. Cambridge: Cambridge University Press. Kearns, K.. 1989. Stoneybatter. Dublin’s Inner Urban Village. Reprinted 1996. Dublin: Gill & Macmillan. ----------. 1994. Dublin Tenement Life. Repr. 1996. Dublin: Gill & Macmillan. ----------. 1996. Dublin Pub Life & Lore. Reprinted 1997. Dublin: Gill & Macmillan. Keating, G. 1992. Trí Bior-Ghaoithe an Bháis ed. by O. Bergin. Dublin: Dublin Institute for Advanced Studies. Leech, Geoffrey N. 1987 [2nd ed.]. Meaning and the English Verb. London: Longman. McCafferty, K. 2003. “‘I’ll be after telling dee de raison […]’. Be after Ving as a future gram in Irish English, 1601–1750”. Celtic Englishes III ed. by H.L.C. Tristram, 298–317. Heidelberg: Universitätsverlag C. Winter.
268
PATRICIA RONAN
McCawley, J. 1971. “Tense and time reference in English”. Reprinted in J. McCawley. 1976. Grammar and Meaning: Papers on Syntactic and Semantic Topics. London: Academic Press, 257–272. MacCionnaith, L., ed. 1938. Dioghluim Dána. Dublin: Government Publications Office. Odlin, T. 1989. Language Transfer. Cross-linguistic Influence in Language Learning. Reprinted 1996. Cambridge: Cambridge University Press. Ó Sé, D. 1992. “The Perfect in Modern Irish”. Ériu 43.39–67. Quirk, R., S. Greenbaum, G. Leech, & J. Svartvik. 1985. A Comprehensive Grammar of the English Language. London and New York: Longman. Reichenbach, H. 1947. Elements of Symbolic Logic. Reprinted 1966. New York: The Free Press. Thurneysen, R. 1980. A Grammar of Old Irish. Dublin: Dublin Institute for Advanced Studies. Traugott, E. C. 1972. A history of English syntax. New York: Holt, Rinehard & Winston. van Hamel, A.G. 1912. “On Anglo-Irish syntax”. Englische Studien 45.272– 292. Visser, F. Th. 1963–1973. An Historical Syntax of the English language. 4 vols. Leiden: Brill. Wagner, H. 1959. Das Verbum in den Sprachen der Britischen Inseln. Tübingen: Niemeyer. Appendix: After-Perfect 1. ‘Hot news’ sense VI.1.1. Many a woman, she’d only be the six or seven weeks after having the baby and she’d be expecting again! (STB 165, f 67) VI.1.2. Going back to my friend Billy Fox now, he was only after starting up his business and Guinness’s gave him no credit [...]. (DPL 145, m 81) VI.1.3. But what he didn’t bloody know [...] was that what he was after sampling was a glass of McCardle’s stout! (DPL 146, m 81) VI.1.4. I remember one time she said, ‘I’m after leaving a glass of stout behind me and don’t think I’m letting you away with that [...]. (DPL 186, f 74) VI.1.5.And when the husband come in she said to him, ‘I’m after getting a new mattress. (DTL 81, f 80) VI.1.6. And the day they came my father was after getting a boat [to work] [...]. (DTL 86, f 78) VI.1.7. Now there was one woman, Biddy was her name, and she woke up one morning and she was only after having a baby [...]. (DTL 109, m 75) VI.1.8. See, they always followed the mother when the mother’d be after giving birth. (DTL 109, m 75) VI.1.9. Oh, and they’d be in great voice after half two when they were after being in the pubs. (DTL 142, m 70) VI.1.10. That’s as true as I’m after walking from Holy Communion. (DTL 185, f 75)
THE AFTER-PERFECT IN IRISH ENGLISH
269
VI.1.11. Well, here one day me mother was up there and here she sees the door opening and a big horn, and him carrying the thing. He was after winning £3-10-0! And bought the grammophone, [...]. (DTL 187-8, f 75) VI.1.12. Jesus, I pawned his old suit and he’s after coming in to put in on and he hasn’t a bit of clothes. (DTL 196, m 61) 2. ‘Hot news’ or Resultative Reading VI.2.1.They were all good men, though men, after doing some damage to the British. 5 VI.2.2. And it was amazing about him that he varied his pubs. If he was after doing a play or writing a story and it had been accepted [for money] he’d then go on a binge [...]. (DPL 109, m 60) VI.2.3. And this man in the pub went stealing his turkeys. We knew the fella that was after stealing the turkeys but you cold do nothing about it [...]. (DPL 148, m 76) VI.2.4. He was after sacking the barmaid and she had been with him for about five or six years [...]. (DPL 176; m 68) VI.2.5. She was looking for a taxi driver, [...]. But I thought somebody was after seeing me going in and they went for me mother. (DPL 198, m 60) VI.2.6. Well, a couple of fellas used to go behind him and take up the pint and take a mouthful out of it. Well, he knew that someone was after drinking that pint [...]. (DPL 214, m 75) VI.2.7. Well, when this argument started in Maguire’s pub the fella was a big fella that was after hitting his friend, so Mooch went home and got one of the swords down [...]. (DPL 227, m 80) VI.2.8. A fight could start over a simple little thing, maybe someone [was] after stealing your pint. (DPL 237, m 68) VI.2.9. I was after being fishing for salmon that day and I got £1 and that was the first £ 1 I ever earned. (DPL 246, m 83) VI.2.10. [...] and she used to go into the air raid shelters and hit her head off the wall and come out split and go down to the police station and say her husband was after doing it. (DTL 110, m 75) VI.2.11. Mostly girls from the country who were after being in [domestic] service and went wrong [...]. (DTL 130, f 66) 3. Experiential Reading VI.3.1. Then Mike Horan’s pub was at the far side of Smithfield, but Mike Horan was after dirtying his bib [i.e. his reputation]. (DPL 176, m 68) VI.3.2. My mother‘d sneak her and the baby in at night-time and sneak her back out in the morning. And her mother didn’t know that she was after being in our house. (DTL 108, f 72) VI.3.3. Oh, they wouldn’t use it [cane] for anything trivial now, but for talking in class or something that you didn’t answer or if you were after being mitching. (DTL 139, f 72)
5
In this example there may be an ellipsis of they were before after due to its presence in the matrix clause. The intended meaning may alternatively be that ‘the men were tough after (through) their deeds’.
270
PATRICIA RONAN
VI.3.4. Now once I took this big top coat up and very, very dirty, like coal. [...] And I said to meself, ‘I’m not going to put me hands into that with all the dirt on it, God knows who’s after being wearing it. (DTL 147, f 91) VI.3.5. Then I lost a girl at eight months, of meningitis. I don’t know. They said I left it too late, but I was after going to all the hospitals and they said there was nothing wrong with her,[...]. (DTL 159, f 76) VI.3.6. I’m after paying 12 pounds for a pram for Tony forty-seven years ago that wasn’t worth 1 pound. (DTL 190, f 75) VI.3.7. Eventually you rambled off home abut nine or ten in the morning and there’d be one or two would be after sleeping all night [...]. (DTL 207, f 72) 4. Resultative Reading VI.4.1. They were after breaking into a local public house down the street and they were full of whiskey [...]. (STB 177, m 73) VI.4.2. It was a young army and we were only after getting over the Civil War end of it and life in the army was tough at that time. (STB 244, m 81) VI.4.3. See, in the thirties you were only after coming out of the twenties, you know. (DPL 136, m 72) VI.4.4. [...] they’d be in the pub selling hot stuff. [...] Oh, it could be anything ... parts of cars or boxes of oranges they’d be after stealing. (DPL 148, m 76) VI.4.5. They were after coming back from a race meeting and they were in the pub divinding out the money [...]. (DPL 212, m 75) VI.4.6. So I went in and when I went in she was laid out in a habit-she was after dying. (DTL 74, f 66) VI.4.7. I brought her to this doctor and she was very sick and he says, ‘You areafter killing your own child. This child has meningitis.’ (DTL 159, f 76) 5. Data from participant observation with Wexford speakers VI.5.1.1. It’s after being drunk by me [translating Welsh: Y mae ef yn ei yfed ef gen i IS IT IN IT’S DRINKING IT WITH ME It is drunken by me.]. VI.5.1.2.Whenever I write a cross that means they are after being syncopated [7th cent. grammatical phenomenon]. VI.5.1.3. I’m after putting a fair bit of work into that [i.e. doctoral dissertation]. VI.5.1.4. Don’t ask me what it is for I’m after forgetting. VI.5.1.5. If you look at editions [of manuscripts], they [editors]’ll be after putting the glide vowel in for you. VI.5.1.6. I’m after doing it so many times now, I just get a dirty look every time. VI.5.1.7. Do you see how I am after coming to this? [i.e. a reconstructed verb form]. VI.5.1.8. They call it uru [‘eclipsis’], because they think the ‘S’ is after being eclipsed. VI.5.1.9. We are after moving on a fair bit in our Welsh [commenting on course progress].
DIALECT HISTORY IN BLACK AND WHITE ARE TWO COLORS ENOUGH?
J. L. DILLARD
In a purely speculative history of African-American Vernacular English (AAVE), Mufwene (2000) touches on only three linguistic forms — those in cursory fashion — and cites no documents. It is apparently assumed that bilingual contact, with the familiar process of interference, will be the only factor to deal with besides “colonial English”. Particularly, it is seen that colonial English, here seen as primarily or exclusively British regional dialect, was the dominant factor, with additional input only from West African languages, which can be reduced to minimal influence by familiar processes. Although the great diversity of the West African languages involved is known, a typological similarity is assumed to be enough to reduce those languages to “substratum” (Singler 1988). Equally familiar processes have reduced substratal influences on American English to minor status (Romaine 2001). The concept of substratum, in its simplest form, is much too simple for the complex early American situation. Yet anyone dealing with AAVE history has been assumed to be dealing in substratum. For example, Reinecke (1975) attributed a substratum orientation to Dillard (1970), a paper dealing with maritime contact and not containing the word “substratum”. The bilingual contact model, even allowing an extreme influence for typology, is oversimple for the early American situation — whether limited to the African slaves or not. Contemporary reports stress how African slaves interacted with Native Americans — often more extensively than with their European masters and almost always more willingly. Brinton (1887) pointed out how the Nantucket tribe had so assimilated with escaped slaves that a number system once assumed to be Native American was actually African. Even in Rhode Island, Native American women were reported to prefer African-American men to their tribesmen as husbands (Cottrol 1982). Many other sources document close early relations between African slaves, whether escaped or not, and Native Americans, later partially disrupted by actions of slave owners. For example, the limited number of slaves — largely house servants — who escaped via the much-touted Underground Railroad was far
272
J.L. DILLARD
overmatched by those who took refuge in the “Indian nations”, especially in the early period before treaties were concluded between slaveholding colonies and the Native American tribes (Franklin and Schweninger 1999). So far as American English was concerned, Native American influence was quickly swept under the rug by language historians, especially those in the Linguistic Atlas tradition (Marckwardt 1958). Except for place names, a fairly small number of lexical items found their way into general American English. (Special varieties can tell a different story.) With the Indians disposed of in this fashion, dialect geographers could go about what they considered to be their major function, tracing British regional forms into American regional dialects (McDavid 1983). A recent movement (Poplack, Tagliamonte, all references) has seen a revived interest in the influence of “colonial English”, although not in the narrowly regional format favored by McDavid (1967). Contact varieties, although natural and well attested, are somehow assumed to be the opposite of “colonial”, and demonstration of “colonial” origin is held to disprove the existence of contact varieties. Mufwene (2000: 238), singled out for quotation in Poplack’s “Introduction” (2000: 25), identifies “56 years [1619–75] of founder population that presumably spoke the same kind of English as the Europeans with whom they interacted and worked on a regular basis”. This “founder principle”, according to Mufwene, presumably [sic] formed the language model for the much greater numbers who arrived later. Chambers (1996: 110) estimates the Virginia slave population — the one Mufwene chooses to deal with — as 3,000 in 1680 and 60,000 in 1740. 1675 is suggested as a terminus a quo for the formation of an extended pidgin in Dillard (1993). Thereafter, West Africans, not all of them slaves (Atkins 1735), had available an excellent language of wider communication for use by members of different tribes and speakers of different languages. Use of Pidgin English as the first expedient in contact between Africans who do not know each others’ native language was observed in the Cameroun in 1963–4, on Fernando Poo in 1964, and has been reported for Nigeria and other parts of West Africa. Tribal and language mixing in the slave trade was reported by Atkins (1735), Moore (1738), Smith (1744), and many others. These are obviously late compared to Mufwene’s date (1675) for the Founders’ Principle. Much greater numbers and interactions with groups other than Englishmen might however, “swamp”, in a familiar metaphor, the 56-year group — in much the same way that AAVE dialects associated with the South swamped more conservative Black dialects in the inner city after WWII. Where use of Pidgin English is concerned, one straw man which has hindered work has been the artificial requirement that in situ pidginization be
DIALECT HISTORY IN BLACK AND WHITE
273
proved. It cannot be proved, of course; but such proof is unnecessary. Development of contact languages by the Portuguese, which strongly influenced other varieties if not to the point of “relexification”, French, and English is well documented. The Basques provide a highly significant addition (Bakker 1987). In West Africa, the slave trade was not a one-nation, monolingual activity; Moore (1738: 27–28) advised English slavers to learn “Creole Portuguese, a bastard form of Portuguese, scarce understood in Lisbon”. Romaine’s (2001) compartmentalization of these various attempts to solve the complex maritime contact problem is not justified by the historical records. That Pidgin English used British dialect forms has been well established (Hancock 1972) before Poplack and Tagliamonte began their demonstrations. (Hancock’s work has generally been ignored or misunderstood.) Even if regional dialects did not survive into the continental colonial picture, they could easily have been represented — especially in unmarked or “zero” forms — in the maritime pidgin. Of course, the obligatory nature of deletion rules might now require that the dialect-speaking sailors acquire final -s and -ed and then delete them. P. A. Luelsdorff long ago (pc, 1967) pointed out to me that such deletion rules violated parsimony by requiring that the speaker first learn the endings and then delete them. The coastal colonies of the sixteenth, seventeenth, and eighteenth centuries contained a highly diverse population — many European groups besides Englishmen, Native Americans, and Africans from many tribes. Englishmen soon gained dominance politically and economically — not without some resistance in New York and New Jersey from the Dutch. When Dr. Alexander Hamilton (1744/1948) went with his AAVE-speaking slave Dromo into upstate New York in the 1740s, he heard such “a medley of Dutch and English as would have tired a horse”. Where the African slaves are concerned, ownership was primarily by Englishmen — although again Dutch are a significant exception. (Prince 1910 recorded a kind of Negro Dutch in New Jersey.) Except for South Carolina, the spread of slavery to the South — especially the inland South — was far in the future. According to Berlin, Favreau, and Miller (1998: xxiv) “In 1810, more than 80 percent of the slave population resided between the Delaware and Savannah rivers in Maryland, the Virginias, and the Carolinas.” Like the freer residents of the thirteen colonies, the slaves lived in a coastal environment, close to and dependent upon the sea as well as upon river transportation. Douglass (1855/1970) tells how field hands from Col. Lloyd’s plantation supplemented their meager rations by fishing in Chesapeake Bay, and one of the earliest of the successful ex-slave entrepreneurs, Paul Cuffee,
274
J.L. DILLARD
launched himself into the shipping business. Runaways in the early period, whether African or European in origin and whether running from slavery or subtler forms of injustice often escaped to a sailor’s life (Cottrol 1982; Bolster 1997: 4,178). Importations from Africa, which were commonplace in Frederick Douglass’s youth, were referred by terms like “salt water” (Johnson 1934). The virtually instantaneous overland transmission often assumed by the reconstructive tradition — most likely based upon the Indo-European overland migrations — was an impossibility. Of the various linguists who have dealt with European languages in the Americas, only Yakov Malkiel (1976) has taken the maritime factor into account. In my opinion, however, Malkiel was a kind of one-man majority. In the early years of the eighteenth century, Newport, Rhode Island, was second only to Boston as a center of the slave trade (Cottrol 1982: 14). From those ports, slaves were regularly transshipped to and from the West Indies and a highly significant trade was carried on with those islands, where no one doubts that a creolized English was used. In the early period, Native Americans were shipped as slaves to the West Indian islands (Franklin and Schweninger 1999). Slaves landing in Rhode Island met a friendly welcome from Native Americans, particularly of the Narraganset tribe (Cotroll 1982: 33). Frequent intermarriage between African men and Narraganset women, who were said to prefer African husbands to “Indians” (Cottrol 1982: 11–13) took place. Even in a place so underpublicized in historical treatments In Rhode Island and the rest of New England, slaveholding patterns and the cultures of the Puritan slave masters permitted the development of a semiautonomous slave culture. The records of New England reveal that slaves had a large amount of freedom of association, often mingling with lower-class whites and Indians. (Cottrol 1982: 5)
Mufwene (2000: 241) is in direct contradiction: “The Africans were not able to form separate communities of their own.” West African and Native American interaction on a large scale (not to mention groups like “Palatine Germans . . . Dutch, Scots, French, and all stripes of Englishmen” [Balmer 1989: 64] and even Scandinavians, including Finns [Jordan and Kaups 1989]) forces the consideration of a different picture. The overall picture of the continental colonies is extreme multilingualism, not the dominance of English observable in the twentieth and twenty-first centuries and perhaps by the middle of the nineteenth. It is well known that a language of wider communication usually exists in such situations, and that a pidgin is a paragon of such languages. There are widespread attestations of American
DIALECT HISTORY IN BLACK AND WHITE
275
Indian use of pidgin English in the colonies and the early United States (Massachusetts attestations variously and in Kittredge [1904] on Waban and other “praying Indians”; Leechman and Hall 1955; Miller 1968; Dillard 1972; Goddard 1977, 1978; Brandt and MacCrate 1979; Forbes 1976), in addition to the much-debated West African attestations. Contact language and pidgins are more usual in maritime contexts and on islands and in coastal areas. In fact American Indian Pidgin English appears to be a major exception, so much that Hancock (1987) would not include it in his survey. The British colonies of the “founding” period were coastal settlements, formed and supplied by sea, with concentrations of population in Boston, New Haven, New York, Philadelphia, Newark, Newport, and Charleston. Reconstructions from the inland United States in the twentieth century, even from inner city dialects, are likely to be misleading. Mufwene (2000: 241) excludes the northern colonies, but coastal was a more viable geographic distinction than north/south in 1675–1725 and a fortiori earlier. Bloomfield (1933: 474) characterized “Negro dialect” as in “the last stage of decreolization”. Krapp (1925) had provided enough documentary evidence for the conclusion but had persisted in the absurd belief that all the attestations, from authors of widely varying backgrounds, were somehow a literary movement — even a kind of conspiracy. Cotton Mather (1721), Benjamin Franklin (n.d), Frederick Douglass (1855/1970) — not to mention Crevecoeur and numerous others — were said to be participants in the same conspiracy to make West African-derived slaves appear to be ridiculous through the use of “bad” (i.e., pidgin) English. Bloomfield obviously brushed that one aside as not worthy of consideration. Resistance to the idea of creolization (of a prior pidgin, probably an extended pidgin) with its concomitant effects on American English, has taken three forms. Loss of Bloomfield’s insight even by those of the same behavioristic (physicalistic) theoretical orientation took place in the environment of a desire to trace all possible American forms to regional British dialects (McDavid 1967, 1983). The only competitor was seen as the substratum. Romaine (2001) basically follows the established patterns of considering only the lexical borrowings. The possibility of an African substratum (actually suggested, in a modified form, by Hall 1950) was disposed of by the same method (Mackwardt 1958, inter alia). Settlement patterns — exclusively British in nature — were for a time the reigning dogma, supported by administrative and editorial authority. Loflin, Sobin, and Dillard (1973) contained a reference to those patterns in the proofs
276
J.L. DILLARD
which had not been in the typescript submitted. It is my fault not to have withheld the article, but I had promised Loflin to get it published. Internal reconstruction was, obviously, an important contribution to historical linguistics. It may not have been, however, the miracle tool that it was sometimes held to be. In particular, it did not help to reconstruct the coastal and maritime colonies of the seventeenth and eighteenth centuries from the midwest of the twentieth century, or to assume that a limited set of migration patterns could be adopted to the exclusion of all other considerations. The work of linguists who concerned themselves with Caribbean creoles (Bailey 1965; Stewart 1967, 1968; Dillard 1962, 1970, 1972) and had carefully considered maritime transmission and insular and coastal distribution — as well as the ethnic factors — provided a basis for reconsideration of the Bloomfieldian thesis. Unfortunately, Bailey 1965 and Dillard 1962 depended too heavily on the “zero copula”; our basic idea was that it was frequent in occurrence and easily identified, not that it was in any way the unique feature. Others are presented in Loflin (all references) and summarized in Rickford (1999: 4–14) and Green (2002). The second major challenge to Bloomfield’s thesis was the contraction and deletion paradigm (Labov 1969). Labov’s justly famous but not entirely perfect “Social Stratification” (1966) had paved the way for consideration of socioeconomic factors, and Labov did not consider ethnic factors until much later (Labov et al. 1968, Labov 1972). Contraction and deletion of the zero copula was soon balanced against the contrasting “invariant” be (which, by the way, had a variant be’s), the contrast having been emphasized in the works and public speeches of Stewart. Be could apparently be considered as a kind of underlying base form which need not surface; am/is/are, all of which contained items often deleted in allegro English speech, could apparently be considered more or less as inflections of the underlying be, obviating any need for an additional syntactic category. That “zero” negated by ain’t and be by don’t, as well as there being different question patterns was not to my knowledge ever considered in that framework. (The Linguistic Atlas had ignored any syntactic differences, on the theory that “the same language” could not involve any such differences. That idea was often advanced in early transformational-generative work.) Loflin (all references) provided evidence to the contrary, including cooccurrence with past time adverbials. Much of the same material had been presented less formally by Stewart (1967, 1968). The battle came to be between those who saw a separate AAVE syntactic system and those who saw phonological (and a few lexical) differences only. Apparently no one now denies the uniqueness at some level of AAVE. Poplack and Tagliamonte (all references) have recently been conducting
DIALECT HISTORY IN BLACK AND WHITE
277
experiments in which they purport to show that the nefarious zero copula, zero relativization, nonredundant pluralization, and unmarked third-person, present tense verbs are the result of (contraction and) deletion rules. Unmarked plurals are said to occur only where the same words would have been unmarked in earlier English and could have been brought in with British regional dialects. (They have given little attention, to my knowledge, to unmarked possessives and to hypercorrections in plurals and “past tenses”.) They have centered their investigations on Samana and Nova Scotia, with an occasional reference to Jamaica as though it were alone in the English-speaking Caribbean and completely ignoring (as has DARE) the extensive work of Holm on Bahamian English. (Holm and Shilling 1982.) The propensity of Afro-Americans for deletions proves to be absolutely astounding. Such claims are not unknown to Caribbean varieties (Allsopp 1996: xlvii). In fact, the West Indian forms were what originally motivated researchers to look for contact unmarking rather than to wholesale deletion. Unlike the Linguistic Atlas researchers, Mufwene, Poplack, and Tagliamonte seem to agree in having no specific regional variety of British English in mind — only “colonial”. The latter term, however, seems to exclude — arbitrarily — a contact variety like pidgin. However, they resemble the Linguistic Atlas people in supporting a general theory. Poplack and Tagliamonte seem to go no further than Labovian contraction and deletion. Mufwene, on the other hand, apparently bases his theory on the newly popular genetic factors (Mufwene 2002). Like contraction and deletion rules — or direct importation of British regional forms, for that matter — the genetic theory does not rest entirely on the non-pidgin/creole nature of early AAVE. As in the case of inherent variability, opposition to the specific approach to AAVE does not constitute opposition to the general theory. The Founders’ Principle has real merit and cannot be overlooked. On the other hand, the details of Mufwene’s and his associates’ (in Poplack 2000) treatment of those founders can be questioned. The dialects of those British colonists of “promiscuous” origin give evidence of the koiné formation which is commonplace for dialects (more than two, I would say) in contact. Puritans, who had a history of residence in Holland behind them, and other less publicized groups ran into (less publicized) contact with Native Americans, who were by no means an unappreciable group, speaking many languages. West African slaves, introduced as early as 1619 but in significant numbers only later, also spoke numerous languages, perhaps typologically related but not mutually intelligible. Furthermore,
278
J.L. DILLARD
During nearly two centuries of settlement along the seaboard, African and African-American slaves had created complex communities, linked by ties of kinship and friendship and resting upon a foundation of shared values and beliefs. (Berlin, Favreau, and Miller, 1998: xxv)
The possibility of such cultural ties is important, since D’Eloia (1973) has established that large plantations like Col. Lloyd’s in Maryland (Douglass 1855/1970) were the exception rather than the norm. Chambers (1996: 118) argues, however, that “nearly three-quarters of enslaved people [in Virginia] were owned in groups of eleven or more and most of the region’s slaves lived on plantations of sufficient size to support a measure of community life”. Even these formulations leave out the persistence of slaves on different farms and plantations in maintaining, despite many obstacles, relationships with other slaves. Until the colonies made treaties with the Native American tribes to return runaway slaves, maroons regularly were found among the Native American tribes. As late as 1850 Frederick Douglass (1855/1970: 436) reported, with some emotional overtones, on how the slaves had chosen to escape to “the savages”. Native American tribes were fairly often slaveholders (Cottrol 1982; Berlin, Favreau, and Miller 1998). Wood (1974) documents the extensive, if ambiguous, relationships between West African slaves and the Native American tribes in the South Carolina colony, including Indian enslavement of escaped slaves at some periods. Barr (1973: 28–29) documents the association of escaped slaves with Native Americans in Texas, dating back in some cases to the Spanish period (3). Blacks were important in dealing with the Indian tribes as scouts and interpreters in the westward expansion. In Clifton, Louisiana, a tribe who called themselves Choctaw but who were phenotypically African with a dialect to match were rejected by the Intertribal Council until the 1980s on the grounds that they were African, not Indian. Significant differences between the artifacts of the Clifton and other Choctaw are easy to discern. Slaves, especially field hands, who fled to sea and to the wet with Indians were, however, numerically inferior to those who went south. (See below.) American Indian Pidgin English was reported by artist George Catlin (1973), who provided some significant attestions, from what he called the “blasting frontier”, not the areas where tribal “purity” was maintained. The obstacle here has been that, with obligatory in situ origin, it has been assumed that AIPE was entirely separate from WAPE. (There has also been the factor that Indian Pride has preferred Leap’s [1976] Interlanguage at all stages.) Considerable evidence exists, however, that a maritime pidgin English (based
DIALECT HISTORY IN BLACK AND WHITE
279
on general maritime contact factors in which the slave trade was a prominent but not the exclusive factor) was being differentiated in 1704 when, in Sarah Kemble Knight’s Journal (1714), Justice Junr told a Justice Senr, who was addressing an Indian in the Boston area, Brother, you speak Negro [that is, WAPE] to him and then proceeded, somewhat officiously, to speak “Indian” (AIPE-like), to which the Indian replied, Now me stomany that. It is, in my opinion, noteworthy that this exchange took place on the coast, considering the time. Bickerton (1974) has ridiculed the widespread transmission “by sailors”, but I think he is thinking of sailors in the twentieth and twenty-first centuries rather than in the seventeenth and eighteenth. Bickerton’s bioprogram (1982) has close affinities, obviously, with the newly developing genetic theories and is herein considered as a variant of them. Consider the Puritans under Bradford (1647/1901). They had encountered Samoset who spoke an English which they “could well understand but much marveled at” and his very helpful friend Squanto (Tisquantum, who presumably spoke to them in the same variety). Samoset had learned that English in contact with fishing vessels on the Massachusetts coast. Heavily dependent upon the maritime trade, the thirteen colonies were perfect places for that pidgin to spread, to be adopted by those in multilingual contact although ignored by those who stayed at home in the villages (Cooper’s Natty Bumpo, anyone?) and to diversify in a range of contacts, especially among slaves from different tribes and Native Americans from equally diversified tribes. It required a special campaign of Andy Jackson’s heroes to eliminate the most persistent and costly (to the slave owners) contact, that of the detribalized Creeks known as Seminoles with slaves from the South Carolina, Georgia, and Florida farms and plantations. Seminole Gullah, differing only slightly from Low Country Gullah (Hancock 1980) persists in Bracketville, Texas, to which the predominantly Black “Indians” were removed. In spite of Romaine’s (2001: 160) assertion that it was spoken “until recently”, Hancock (pc) assures me that it is still a lively variety. Back to linguistic forms — remember them? If an alternative to consideration of a complete syntactic system must be undertaken — if Poplack and Tagliamonte are to be met on their own grounds of divide and destroy — I
J.L. DILLARD
280
choose preverbal “been”, for which I have written examples from AfricanAmerican college students in Louisiana and Ohio. Despite my reluctance to isolate individual features and insistence upon treating the interrelationships of the entire system (Smith 1969), I would be willing to advance it under the traditional dialectological rubric of “relic form”. DARE recorded it (I: 178) and even considered forms like “been V-ing” as “decreolization while the creolist Cassidy was editor-in-chief. His successors have chosen to eliminate it from the otherwise exhaustive index of Volumes I and II (DARE 1993). Why? Doesn’t it fit into the theory they wish to support? Mufwene (2001: 301) lists I bin home all day and We bin talkin about this forever now but not You bin know that I bin had it a long time which are virtually fixed phrases in AAVE. Mufwene (1983: 6, 16–20) had accepted just such forms. Occurrence with modals (You oughta been did that) is also quite frequent. Mufwene (2001, not 1983) offers his examples to support his point that it is impossible to account for AAVE with a “creole” (presumably Gullah) grammar. True — but hardly the historical point. Summarizing: If colors are to be the dominant metaphor, “red” will have to be figured in with “black” and “white” (not to mention “yellow” farther west). “White” is by no means all English, and “English” does not necessarily entail regionally distributed dialects. A founders’ group may have been formed 1619–75 and have formed the basis for northern Black dialects, but even in Massachusetts there is evidence of quite early occurrence of AAVE features. (Here there is similarity to the Africans who were sent to England during the early period of slaving to learn to serve as interpreters.) Moreover, this group was swamped by the huge numbers of later imports who spoke already or learned Pidgin and spoke it with Native Americans as well as among themselves. (People who speak Pidgin English as well as another variety of English are commonplace; see the works of Chinua Achebe.) Colonial English
DIALECT HISTORY IN BLACK AND WHITE
281
forms, some of which had been or were leveled out in the koiné, may well have come to American English by way of the maritime contact variety which it is convenient to call Pidgin. Pesky relic forms like preverbal anterior marker “been” remain. One version of the founders’ principle may ultimately prove to be multiethnic and multilingual maritime contact.
References Algeo, John, ed. 2001. English in North America, Vol. VI of the Cambridge History of the English Language. Cambridge: Cambridge University Press. Allsopp, S. R. R. 1996. Dictionary of Caribbean English Usage. New York: Oxford University Press. Atkins, John. 1735. A voyage to Guinea, Brasil, and West Indies. London: C. Ward and R. Chandler. Bailey, Beryl. 1965. ‘Toward a new perspective on American Negro dialectology’. American Speech 11. 171–177. Bakker, Peter. 1987. “A Basque nautical pidgin”. Journal of Pidgin and Creole Languages 2:1.3–30. Balmer, Randall H. 1989. A Perfect Babel of Confusion. New York: Oxford University Press. Barr, Alwayn. 1973. Black Texans, A History of Negroes in Texas 1528–1971. Austin: Pemberton Press. Berlin, Ira, Marc Favreau & Steven F. Miller, eds. 1998. Remembering Slavery: African Americans Talk about Their Personal Experiences of Slavery and Emancipation. New York: The New Press. Bickerton, Derek. 1974. “Creolization, linguistic universals, natural semantics and the brain”. Working Papers in Linguistics, University of Hawaii, 6:3.124–141. ----------. 1982. Roots of Language. Ann Arbor: Karoma Press. Bloomfield, Leonard. 1933. Language. New York: Henry Holt. Bolster, W. Jeffrey. 1997. Black Jacks: African American Seamen in the Age of Sail. Harvard: Harvard University Press. Bradford, William. 1647/1901. Of Plimouth Plantation, 1620–1647. Boston: Wright and Potter. Brandt, Elizabeth and Christopher MacCrate. 1979. The multilingual variation of contact vernaculars in the Southwest. Vancouver, B.C., XIII Congress of Americanists. Brinton, D. G. 1887. “On certain supposed Nantucket words, shown to be of African origin”. American Antiquarian and Oriental Journal 9.350–354.
282
J.L. DILLARD
Cassidy, Frederic G. (editor-in-chief). 1985, 1991, 1996. Dictionary of American Regional English, Vols. I, II, and III. Cambridge, Mass.: Belknap Press. Catlin, George. 1973. Letters and Notes on the Manners, Customs, and Conditions of the North American Indians Written During Eight Years’ Travel (1832–9). Two Vols. New York: Dover Publications. Chambers, Douglas B. 1996. “Historical creolization in eighteenth-century Virginia”. The African Diaspora ed. by Alusine Jalloh and Steven E.Miller, 100–133. College Station, Texas: Texas A&M University Press. Cottrol, Robert J. 1982. The Afro-Yankees, Providence’s Black Community in the Antebellum Era. London: Greenwood Press. D’Eloia, Sarah G. 1973. Review of Dillard, Black English. Journal of English Linguistics 7.87–106. DARE. 1993. An Index by Region, Usage, and Etymology to the Dictionary of American Regional English. Vols. I and II. [Publication of the American Dialect Society 77.] Tuscaloosa: Published for the Society by the University of Alabama Press. Dillard, J. L. 1962. “Toward a bibliography of works dealing with the creole languages of the Caribbean area, Louisiana, and the Guianas”. Caribbean Studies 3, No.1.84–95. ----------. 1970. “Optional ordering rules and genetic relationships in pidginrelated languages”. Paper presented at AAA meeting, San Diego, unpublished. ----------. 1972. Black English: Its History and Usage in the United States. New York: Random House. ----------. 1993. “The transitivizer and pidgin chronology”. Atlantic Meets Pacific ed. by Frank Byrne and John Holm, 333–340. Amsterdam: John Benjamins. Douglass, Frederick. 1855/1970. My Bondage and My Freedom. Chicago: Johnson Publishing Co. Forbes, Jack D. 1976. Nevada Indians Speak. Reno: University of Nevada Press. Franklin, Benjamin. n.d. Advice to those who would remove to America, in Writings (ed. Smythe), VIII.606. Franklin, John Hope and Loren Schweninger. 1999. Runaway Slaves, Rebels on the Plantation. New York: Oxford University Press. Green, Lisa J. 2002. African American English, A Linguistic Introduction. Cambridge: Cambridge University Press.
DIALECT HISTORY IN BLACK AND WHITE
283
Goddard, Yves. 1977. “Some early examples of American Indian Pidgin English”. IJAL 43:1.137–141. ----------. 1978. “A further note on pidgin English”. IJAL 44.73. Hall Robert A., Jr. 1950. “The African subsatum in American English”. American Speech 25.51–54. Hamilton, Dr. Alexander. 1744/1948. Gentleman’s Progress: The Itinerarium of Alexander Hamilton 1744 ed. by Carl Bridenbaugh. Chapel Hill, N.C.: University of North Carolina Press. Hancock, Ian F. 1972. “A domestic origin for the English derived Atlantic creoles”. Florida FL Reporter X.49–54 [often revised and reprinted]. ----------. 1980. “Texas Gullah: The Creole English of the Brackettville AfroSeminoles”. Perspectives on American English ed. by J.L. Dillard, 305– 333. The Hague: Mouton. ----------. 1987. “A preliminary classification of the Anglophone Atlantic creoles”. Pidgin and Creole Languages: Essays in Memory of John E. Reinecke, 264–333. Honolulu: University of Hawaii Press. Holm, John and Alison Shilling, eds. 1982. Dictionary of Bahamian English. Cold Spring, N.Y.: Lexik House. Johnson, Charles Spurgeon. 1934. Shadow of the Plantation. Chicago: University of Chicago Press. Jordan, Terry and Matti Kaups. 1989. The American Backwoods Frontier. Baltimore: Johns Hopkins Press. Kittredge, George Lyman. 1904. The Old Farmer and his Almanac. Boston: W. Ware and Co. Knight, Sarah Kemble. 1714/1935. The Journal of Madam Knight. New York: Peter Smith. Krapp, George Philip. 1925. The English Language in America. New York: Frederick Publishing Co. Labov, William. 1966. The Social Stratification of English in New York City. Washington: Center for Applied Linguistics. ----------. 1969. “Contraction, Deletion, and Inherent Variability of the English Copula”. Language 45.715–762. ----------. 1972. Language in the Inner City. Philadelphia: University of Pennsylvania Press. Labov, William, Paul Cohen, Clarence Robins, and John Lewis. 1968. A Preliminary Study of the Nonstandard English of Black and Puerto Rican Speakers. New York City Office of Education Cooperative Research Report No. 388. Leap, William. 1976. Studies in Southwestern Indian English. San Antonio: Trinity University Press.
284
J.L. DILLARD
Leechman, Douglas & Robert A. Hall, Jr. 1955. “American Indian Pidgin English: Attestations and grammatical peculiarities”. American Speech 30.163–171. Loflin, Marvin D. 1969. “Negro nonstandard and standard English: Same or different deep structure”. Orbis XVIII.74–91. ----------. 1970. “On the structure of the verb in a dialect of American Negro English”. Linguistics 59.14–28. Loflin, Marvin D., Nicholas Sobin & J. L. Dillard. 1973. “Auxiliary structures and time adverbs in Black American English”. American Speech 48.22– 36. McDavid, Raven I., Jr. 1967. “Historical, regional, and social variation”. Journal of English Linguistics 1.25–40. ----------. 1983. “Retrospect”. Journal of English Linguistics 16.47–54. Malkiel, Yakov. 1976. “Changes in the European languages under a new set of sociolinguistic circumstances”. First Images of America, II ed. by Fredi Chiapelli, 581–593. Berkeley: University of California Press. Marckward, Albert H. 1958. American English. New York: Oxford University Press. Mather, Cotton. 1721. The Christian Philosopher. Miller, Mary Rita. 1968. “Attestations of American Indian Pidgin English in fiction and nonfiction”. American Speech 41.142–147. Moore, Francis. 1738. Travels into the Inland Parts of Africa. London: Edward Cave, St. John’s Gate. Mufwene, Salikoko. 1983. Some Observations on the Verb in Black English Vernacular. African and Afro-American Studies and Research Center, The University of Texas at Austin. ----------. 2000. “Some sociolinguistic inferences about the development of African American English”. The English History of African American English ed. by Poplack. London: Blackwell. ----------. 2001. “African-American English”. Algeo 2001. 291–324. ----------. 2002. “Creolistics and genetic linguistics: Rethinking linguistic evolution”. Plenary presentation given at LSA annual meeting, 2002. Poplack, S. 2000. The English History of African American English. Malden, MA: Blackwell. Poplack, Shana and David Sankoff. 1987. “The Philadelphia story in the Caribbean”. American Speech 62.291–314. Prince, J. Dyneley. 1910. “Jersey Dutch”. Dialect Notes III.459–460. Reinecke, John, ed. 1975. A Bibliography of Pidgin and Creole Languages. Honolulu: University Of Hawaii Press, Linguistics Publication No. 14.
DIALECT HISTORY IN BLACK AND WHITE
285
Rickford, John. 1999. African American Vernacular English. Malden, Mass.: Blackwell. Romaine, Suzanne. 2001. “Contact with other languages”. Algeo 2001. 154– 183. Singler, John. 1988. “The homogeneity of the substrate as a factor in pidgin/creole genesis”. Language 64:1.27–51. Smith, William. 1744. A Voyage to Guinea. London: J. Nourse. Smith, Riley B. 1969. “Interrelatedness of certain deviant grammatical structures in Negro nonstandard dialects”. Journal of English Linguistics 3.82–88. Stewart, William A. 1967. “Sociolinguistic factors in the history of American Negro dialects”. Florida Foreign Language Reporter 5:2.11–29. ----------. 1968. “Continuity and change in American Negro dialects”. Florida Foreign Language Reporter 6:2.3–14. Tagliamonte, Sali & Shana Poplack. 1988. “How Black English past got to be present: Evidence from Samana”. Language in Society 17:4.513–534. Wood, Peter H. 1974. Black Majority, Negroes in Colonial South Carolina from 1670 through the Stoner Rebellion. New York: Alfred A. Knopf.
INDEX OF LANGUAGES AND DIALECTS
Acadian French 234, 237, 238, 240, 241 African-American Vernacular English (AAVE) 94, 271, 272, 276, 277, 280 American English 57, 58, 123, 212, 271, 272, 275, 281 American Indian Pidgin English (AIPE) 278, 279 Appalachian English (US) 131–134 Baltic-area languages (Latvian; Estonian; Baltic Russian) 52, 56, 57 British English 70, 211, 212, 217, 223, 257 Canadian English (e.g. Atlantic Canada; Ottawa Valley) 57, 62 Canadian French (incl. Montreal, New Brunswick, Newfoundland) 237, 238, 244, 247 Celtic English (Ireland; Scotland, incl. Hebrides; Wales) 56, 57 Celtic languages 56 Colonial English 271, 272, 280 Creole Portuguese 273 Danish 55, 62, 224 Dutch 11, 18, 21, 56, 185, 187, 188, 192, 193, 207, 212 ABN Standard Dutch 188, 189 Amsterdam-accented Dutch 188, 189 Avant-Garde Dutch 185–209 Polder Dutch 185, 188 Western Dutch accent 188, 189 English 214, 238–244, 248, 257–260, 265 Finnish 57 Flemish 212 French 18, 21, 22, 31–33, 40, 43–44, 48, 52, 56, 207 Frisian 82 Gaelic 56, 57 German 4, 5, 9, 11, 18, 21, 22, 24, 62, 66, 75, 77, 78, 80–82 Austrian German 62 Letzebuergish 4, 18, 21 Low German 75–78, 81, 82 Low German Pomeranian dialects 83 Mennonite Low German (Plautdietsch) 73–86 Standard High German (SHG) 31–32, 35, 39, 43–46, 48–49
Swiss German (SG) 23, 24, 31–32, 39– 40, 43–46, 48–49, 53 Gullah 279, 280 Hiberno-English 253, 254, 256, 257, 259, 260 see also Irish English Icelandic 157–171 Irish, Early Modern 259, 260 Irish English 225, 235, 253–270 see also Hiberno-English Italian 31–33, 40, 43–44, 48 Kannada 236 Kashubian 82 Latin 234 Ma’a 235 Marathi 236 Maritime Pidgin English 273, 278, 281 Mbugu 235 ‘Negro’ Dutch 273 ‘Negro’ dialect 275 see also AAVE Newfoundland English 54, 59, 60, 62–64 Nordic languages see Scandinavian languages Norse 224, 225 ‘Northern Englishes’ 211, 215, 217 Norwegian 53, 55, 62, 64 Polish 82 Prince Edward Island French 237, 238, 240, 243, 244, 247–249 Quebec French 237, 238, 242, 243, 247 Romansh 31–33, 40, 43–46, 48 Russian dialects 83 Scandinavian languages 53–55 Scots 90 South African Indian English 225 Swedish (incl. American Swedish) 52–55, 58–60, 62–64, 66, 69, 70 Tohono O’odham (Uto-Aztecan family) 55 Urdu 236 Ulster Scots 90, 96, 108 West African Pidgin English (WAPE) 278
SUBJECT INDEX A. accusative sickness 159, 163 acoustic analysis/phonetics 121–123, 135, 136, 140, 142, 143 adverbs 31, 35–39, 46, 189 zero adverb 94, 95, 102–104, 106–108, 110 age differences 127, 130, 132, 138, 145, 150, 175, 176, 180–182 alternations of consonants 77, 79 antecedent 213–215, 217–225 areal features 51, 55 association function 34–35, 40, 47–49 attestations 274, 275 B. backchannels 67, 68 bark transformation 190, 196, 197, 198, 208 between-rater consistency 202 borrowing 21, 22, 80, 81, 133 see also constraints on borrowing see also grammatical borrowing see also lexical borrowing see also morphological borrowing see also syntactic borrowing C. case inflections 159, 161, 162, 165, 166, 169, 170 chain shift 126, 185 change from above 132, 189, 190 change from below 189, 190 cognitive linguistics 31 consonant system 76–78 consonants 76–78, 81, 82, 143 palatalised consonants 77, 80–83 velar consonants 77, 78, 81 constraints on borrowing 233, 234 see also borrowing contact variety 277, 281 contact with Polish and Kashubian 82 conversational interaction 67 cross-sex differences 196, 207 see also gender differences
D. decreolization 275, 280 dialect atlas 4 dialect continuum 4, 21, 25, 142, 150, 151, 214 dialect identification 143–145 discourse 33, 93, 103, 121, 123 discourse features 174–176, 181, 182 discourse functions/meaning 54, 66, 67 discourse particles 51, 53–56, 58, 61–67 E. endpoint normalization 192, 207 ethnicity 131, 150, 173 ethnodialectology 6, 7, 14–16, 18–20, 22 see also folk linguistics event integration 41–42, 44–47 extended pidgin 272 F. figure 33–35, 37–38, 40–41, 48 see also ground folk linguistics 121, 123 see also ethnodialectology formant 124, 135, 140, 190–195, 198, 204, 208 for to infinitive 94, 95, 100, 102 founder principle 272 G. gender 132, 145, 150, 152, 159 gender differences 176, 179, 180–182 see also cross-sex differences gender effects on language use 62–65 genitive avoidance 157, 162–164, 168–170 Göteborg University Spoken Language Corpus of Swedish 59 grammatical borrowing 236, 237 see also borrowing grammaticalization 110, 259, 260, 265 ground 33–34, 37, 41, 48 see also figure
290
SUBJECT INDEX
I. identity 24, 25, 75, 91, 121, 122, 124, 131, 133, 145 impersonal verbs 157, 162, 163, 168–170 influence of Frisian 82 ingressive articulation (pulmonic) 51, 53– 57, 59, 63, 66, 68, 70 internal reconstruction 276 isogloss 6, 14, 15, 25, 26, 211, 225 K. koiné 277, 281 L. language language attitudes 121–123, 169 language contact 39, 40, 51, 57, 59, 65, 70, 82, 83, 212, 224, 279 language ideology 121–124, 153 language variation 122, 123, 173, 174, 182 non-standard language 31, 32, 46, 47, 152, 214, 215, 219, 222, 223, 225 standard language 4, 11, 12, 18, 21, 22, 24, 28, 31, 32, 43, 47, 78, 152, 159, 163, 164, 253 lenition 77 lexical borrowing 235, 249 see borrowing linguistic stability 157–159, 166, 169, 171 literacy 49 M. minimal responses 64, 65, 67, 68 morphological borrowing 234, 235 see also borrowing morphosyntactic 211, 225, 236, 237, 264 multilingual contact 279 N. narratives 180, 181 nation state 3, 4, 11, 18 Neg/Aux contraction 94, 95, 98, 101, 108, 109 network 123, 124, 131–134, 169, 173 The Newcastle Electronic Corpus of Tyneside English (NECTE) 215, 216 non-redundant (zero) pluralization 277
Northern Cities Chain Shift (US) 123–127, 130–135, 141, 142, 145 NP/PRO constraint 97, 98 O. orality 49 P. passive 158, 160, 163, 164, 169, 170, 180, 259 new passive (Icelandic) 157–161, 163, 164, 169, 170 pseudo-passives 244 perception 122, 123, 140–143, 145, 146, 152 perceptual dialectology 122, 123, 140–143, 145, 146, 152 perfect 253–261, 263–266 physicalism (behaviourism) 275 pragmatic pragmatic functions/meaning 52, 59, 62, 64–66, 67 pragmatic constraints on usage 51 prepositions 34–35, 164–167, 237, 238, 240, 244, 245, 247, 248 orphan prepositions 242–244 preposition stranding 240, 241, 243, 244, 246–249 pronouns 97, 98, 178, 179, 213, 217, 220 R. recency 259, 263, 265 relative clauses 242–244 non-restrictive 213, 217–219 restrictive 213, 219–223 sentential relative 218, 219 relexification 273 replication 134, 182 resultativity 260, 262 rural versus urban 123, 127, 128, 130 S. settlement patterns 275 social social class differences 175, 176, 178, 180–182 social psychology of language 121 sociophonetics 122, 123, 153
SUBJECT INDEX
Southern Vowel Shift (US) 135, 136 spatial reference 31 Sprachbund 83 stigmatization 133, 135, 141, 158, 159, 168–170 stylistic stratification 213 subjacency effects 248 Survey of English Dialects 52, 56, 214, 215 Survey of Sheffield Usage 216 swamping 272, 280 syntactic borrowing 234, 235, 241, 249 see also borrowing T. transfer 13, 258, 260 transition zone 211 turn regulators 67, 68
291
Tyneside Linguistic Survey 215 V. verbal feedback 65 verbal -s 94–97, 100, 101, 107–109 Verkehr 13, 28 vernacular 31, 93, 94, 133, 135, 211, 212, 214, 219,223 vocalisation 77 vowel normalization 191, 192, 208 W. WH-questions 244 Z. zero copula 276, 277 zero relativization 277
CURRENT ISSUES IN LINGUISTIC THEORY
E. F. K. Koerner, Editor
Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from scholars who have significant proposals to make towards the advancement of our understanding of language, its structure, functioning and development. CILT has been established in order to provide a forum for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the prevailing mode of thought in linguistic science. It offers an outlet for meaningful contributions to the current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have. A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 273 FILPPULA, Markku, Juhani KLEMOLA, Marjatta PALANDER and Esa PENTTILÄ (eds.): Dialects Across Borders. Selected papers from the 11th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002. 2005. xii, 291 pp. 272 GESS, Randall S. and Edward J. RUBIN (eds.): Theoretical and Experimental Approaches to Romance Linguistics. Selected papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake City, March 2004. 2005. viii, 367 pp. 271 BRANNER, David Prager (ed.): The Chinese Rime Tables. Linguistic philosophy and historical-comparative phonology. 352 pp. Expected December 2005 270 GEERTS, Twan, Ivo van GINNEKEN and Haike JACOBS (eds.): Romance Languages and Linguistic Theory 2003. Selected papers from ‘Going Romance’ 2003, Nijmegen, 20–22 November. 2005. viii, 369 pp. 269 HARGUS, Sharon and Keren RICE (eds.): Athabaskan Prosody. 2005. xii, 432 pp. 268 CRAVENS, Thomas D.: Variation and Reconstruction. viii, 222 pp + index. Expected November 2005 267 ALHAWARY, Mohammad T. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics XVII– XVIII. Papers from the seventeenth and eighteenth annual symposia on Arabic linguistics. Volume XVII–XVIII: Alexandria, 2003 and Norman, Oklahoma 2004. 2005. xvi, 315 pp. 266 BOUDELAA, Sami (ed.): Perspectives on Arabic Linguistics XVI. Papers from the sixteenth annual symposium on Arabic linguistics, Cambridge, March 2002. ca. 250 pp. Expected December 2005 265 CORNIPS, Leonie and Karen P. CORRIGAN (eds.): Syntax and Variation. Reconciling the Biological and the Social. 2005. vi, 312 pp. 264 DRESSLER, Wolfgang U., Dieter KASTOVSKY, Oskar E. PFEIFFER and Franz RAINER (eds.): Morphology and its demarcations. Selected papers from the 11th Morphology meeting, Vienna, February 2004. With the assistance of Francesco Gardani and Markus A. Pöchtrager. 2005. xiv, 320 pp. 263 BRANCO, António, Tony McENERY and Ruslan MITKOV (eds.): Anaphora Processing. Linguistic, cognitive and computational modelling. 2005. x, 449 pp. 262 VAJDA, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp. 261 KAY, Christian J. and Jeremy J. SMITH (eds.): Categorization in the History of English. 2004. viii, 268 pp. 260 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp. 259 CARR, Philip, Jacques DURAND and Colin J. EWEN (eds.): Headhood, Elements, Specification and Contrastivity. Phonological papers in honour of John Anderson. 2005. xxviii, 405 pp. 258 AUGER, Julie, J. Clancy CLEMENTS and Barbara VANCE (eds.): Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp. 257 FORTESCUE, Michael, Eva Skafte JENSEN, Jens Erik MOGENSEN and Lene SCHØSLER (eds.): Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp. 256 BOK-BENNEMA, Reineke, Bart HOLLEBRANDSE, Brigitte KAMPERS-MANHE and Petra SLEEMAN (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30 November 2002. 2004. viii, 273 pp. 255 MEULEN, Alice ter and Werner ABRAHAM (eds.): The Composition of Meaning. From lexeme to discourse. 2004. vi, 232 pp. 254 BALDI, Philip and Pietro U. DINI (eds.): Studies in Baltic and Indo-European Linguistics. In honor of William R. Schmalstieg. 2004. xlvi, 302 pp. 253 CAFFAREL, Alice, J.R. MARTIN and Christian M.I.M. MATTHIESSEN (eds.): Language Typology. A functional perspective. 2004. xiv, 702 pp.