NEW PERSPECTIVES ON ENGLISH HISTORICAL LINGUISTICS I
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENC...
76 downloads
1244 Views
1MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
NEW PERSPECTIVES ON ENGLISH HISTORICAL LINGUISTICS I
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Christchurch, N.Z.); 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 25
Christian Kay, Simon Horobin and Jeremy Smith (eds) New Perspectives on English Historical Linguistics Selected papers from 2 ICEHL, Glasgow, 2–26 August 2002 Volume I: Syntax and Morphology
NEW PERSPECTIVES ON ENGLISH HISTORICAL LINGUISTICS SELECTED PAPERS FROM 12 ICEHL, GLASGOW, 21–26 AUGUST 2002 VOLUME I: SYNTAX AND MORPHOLOGY
Edited by
CHRISTIAN KAY SIMON HOROBIN JEREMY SMITH University of Glasgow
JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA
8
TM
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 International Conference on English Historical Linguistics (2th : 2002 : Glasgow University) New persprectives on English historical linguistics : selected papers from 2 ICEHL, Glasgow, 2–26 August 2002. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 25–252) Includes bibliographical references and index. Contents: v. . Syntax and morphology / edited by Christian Kay, Simon Horobin, Jeremy Smith -- v. 2. Lexis and transmission / edited by Christian Kay, Carole Hough, Irené Wotherspoon. . English language--Grammar, Historical--Congresses. 2. English language--History--Congresses. I. Kay, Christian. II. Horobin, Simon. III. Smith, J. J. (Jeremy J.) IV. Title. V. Series. PE075.I57 2002b 427--dc22 2004047943 ISBN 90 272 4763 3 (Eur.) / 588 54 3 (US) (Hb; Volume : alk. paper) ISBN 90 272 4764 (Eur.) / 588 55 (US) (Hb; Volume 2: alk. paper) ISBN 90 272 4765 X (Eur.) / 588 528 3 (US) (Hb; Set) © 2004 – 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
"ack"> "intro"> "cla"> "den"> "fis"> "lai"> "los"> "mcm"> "meu"> "per"> "smi">
Table of contents
Acknowledgements Introduction Verbal -s reconsidered: The Subject Type Constraint as a diagnostic of historical transatlantic relationship Sandra Clarke
vii ix
1
Do grammars change when they leak? David Denison
15
Grammar change versus language change: Is there a difference? Olga Fischer
31
Indefinite Pronominal Anaphora in English correspondence between 1500 and 1800 Mikko Laitinen
65
From resultative predicate to event-modifier: The case of forth and on Bettelou Los
83
Family values April McMahon and Robert McMahon
103
From inventory to typology in English historical dialectology Anneli Meurman-Solin
125
Consumers of correctness: Men, women, and language in eighteenth-century classified advertisements Carol Percy
153
Accounting for vernacular features in a Scottish dialect: Relic, innovation, analogy and drift Jennifer Smith
177
< /R/TREARGET E FF
"suz"> "tae"> "war"> "wis"> "si"> "toc">
vi
Table of contents
On MV/VM order in Beowulf Hironori Suzuki DARE and NEED in British and American present-day English: 1960s–1990s Martine Taeymans
195
215
What drove DO? Anthony Warner
229
The have-‘perfect’ in Old English Ilse Wischer
243
Subject index
257
DOCINFO AUTHOR ""TITLE "Acknowledgements"SUBJECT "Current Issues in Linguistic Theory, Volume 251"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Acknowledgements
The Editors would like to record their thanks to the contributors to this volume and to those who helped at various stages of the reviewing process: Julie Coleman, Andreas Fischer, Olga Fischer, Margaret Laing, Matti Kilpiö, Bettelou Los, April McMahon, Carol Percy, Jane Roberts, Irma Taavitsainen and Keith Williamson. Our thanks are also due to Konrad Koerner, series editor of CILT, for his support and Anke de Looper at John Benjamins for her enthusiasm, patience and advice. We are also grateful to all those who helped to set up and run the Twelfth International Conference on English Historical Linguistics from which these papers derive, notably colleagues at the Glasgow English Language Department, the STELLA project, the Edinburgh Institute for Historical Dialectology, and all our student helpers. A special thank you goes to Ian Hamilton for his tireless work on conference organisation and the resulting volumes of papers. We would also like to acknowledge with thanks the financial support of the British Academy, the Linguistics Association of Great Britain, Oxford University Press, the University of Glasgow, and Thomas Tunnock Ltd. Glasgow, January 2004 The Editors
Introduction
The papers collected in this volume are a selection of those given at ICEHL 12, which was held at Glasgow University in August 2002. Most are concerned with the historical grammar of English; a second volume, concerned primarily with issues of lexis and transmission, is being published simultaneously. Together the two volumes provide a fair summary of many issues that are currently engaging practitioners in the field of English historical linguistics. As I indicate in the introduction to Volume II of these papers, New Perspectives on English Historical Linguistics is a sober title making a big claim. The study of the history of English grammar has a long pedigree, and it is therefore sometimes hard to see one’s work as much more than simply footnotes to Sweet or Jespersen or Ellegård or Mustanoja or Mitchell or Visser. But the essays presented here all genuinely move the subject forward by pursuing wider theoretical implications; indeed, this consideration guided the authors in making their selection from the papers offered at the conference. This is particularly apparent in the plenary papers by David Denison, Olga Fischer and April and Robert McMahon, all three of which link the engagement with data — an inheritance from the philological tradition — with attention to current linguistic theory. It seems that we have learned (in Michael Halliday’s phrase) that we need not trample on our predecessors in order to make progress — surely a sign of the maturity of our subject. It is perhaps appropriate in a preface of this nature to sum up some of the directions in which the historical study of English grammar is going, as witnessed by the papers presented here. I will choose three issues in particular. First, there is a question of the health of the discipline. It is noticeable that English historical linguistics is in many ways a subject pursued more strongly outwith the Anglophone world than within it, and many of the most exciting contributions to the conference, both in papers and in subsequent discussion, were offered by scholars from continental Europe and Asia (notably Japan). This is in some ways a paradoxical state of affairs, and it certainly challenges
x
Introduction
those of us working in (say) the United Kingdom to develop a greater evangelical zeal for our subject. At the same time, the present volumes remind us that the history of English as well as English itself is now an international possession. Secondly, the welcome rapprochement between linguistics and philology — discussed briefly in the preface to the second volume — has meant a new interest in bringing data into closer articulation with linguistic theory. An encouraging factor here has been the adoption of what may still (just) be called “the new technology”. In another forum, Anneli Meurman-Solin has called for the establishment of “computational philology”, whereby the power of the computer can be harnessed to the analysis of large historical corpora. Of these, the most famous is undoubtedly the Helsinki Corpus, but this pioneering work is now being supplemented by massive new resources with much-enhanced functionality. Some of these new resources are described in these volumes. Finally, an exciting new development is the concern with models. Perhaps “new” is a misnomer; after all, Max Mueller and Auguste Schleicher, responding to Darwin, called for a “biological” linguistics in the 1860s. But historical linguists have always been a somewhat piratical lot, willing to raid other disciplines for useful models and metaphors in order to help them practise their craft more effectively. Our grammars leak and breed, sprout and flow, blend and split; their categories are fuzzy, yet have prototypical cores. (“Perspectives”, of course, is a metaphor.) What all this modelling indicates, it seems to me, is that English historical linguists are always eager to examine the intellectual underpinning of their subject. Such constant self-reflexivity, though hardly healthy for an individual, is an extremely healthy sign for an academic discipline. In sum, it seems that English historical linguists have no shortage of exciting research questions to address. And the essays in this volume indicate that equally exciting answers to these questions — and in turn more questions — are continuing to emerge. Jeremy Smith Glasgow, 2003
Verbal -s reconsidered The Subject Type Constraint as a diagnostic of historical transatlantic relationship Sandra Clarke Memorial University, Canada
1.
Introduction
As observed by Wolfram, Thomas and Green (2000), among others, most variationist studies of English dialects which display non-concord present-tense verbal -s reveal a higher incidence of this inflection when the subject is other than an adjacent personal pronoun; thus examples of non-concord -s involving the NP types in (1) are more likely than the pronominal examples in (2): (1) a. (adjacent simple NP) The children plays together. b. (relative clause) The children/them that plays together… c. (empty subject slot) They laugh and plays together. (2) a. They plays together. b. We plays with them all the time.
This tendency, labelled the subject-type or NP/Pro constraint, is a longstanding one; Poplack and Tagliamonte (2001: 191) note that “… it is clear that the combined effect of subject type and adjacency on verbal -s variability is a legacy of English dialects that has persisted from the tenth century to the present”. As such, it has proven a crucial diagnostic of historical origin, most notably in the debate over the English versus creole origins of African American English (AAE). So apparently pervasive is this tendency that Godfrey and Tagliamonte (1999: 109) have concluded that “… type-of-subject constraint must have been part of the grammatical system of the varieties of English that were transported to North America,” and that transatlantic varieties that do not display this constraint must have lost it subsequent to migration to the New World. While Godfrey and Tagliamonte’s conclusion finds support in recent sociohistorical investigations of transatlantic verbal transmission (e.g. Poplack
2
Sandra Clarke
and Tagliamonte 2001, Wright 2002), this paper challenges the conclusion that a subject-type constraint was a necessary component of the grammars of those varieties of southern British English which figured prominently in seventeenthand eighteenth-century transatlantic migrations. Following the groundwork laid in Clarke (1997), the paper surveys verbal -s patterning in one of the most conservative of the transplanted southern British overseas varieties — the vernacular English spoken on the island of Newfoundland, today the most easterly province of Canada. Along with other conservative southern British regional dialects, Newfoundland English suggests that the historical saga of verbal -s both within Britain and overseas must have involved considerable interdialect variation, resulting in a complex and multi-stranded path of development across space and time. In addition, the value of the subject-type constraint as a diagnostic of historical relationship is undermined by the fact that cross-dialect similarities in subject-related patterning among varieties of English exhibiting non-concord verbal -s may be attributable to factors other than shared historical origin. These include cross-linguistic grammatical, cognitive and psycholinguistic processing tendencies.
2. The Northern Subject Rule The best known statement of the subject-type constraint in historical nonstandard varieties of English has come to be known as the Northern Subject Rule, since it describes -s patterning in the north of England, Scotland, and northern Ireland, where the constraint is most in evidence, and its history the longest (e.g. Meurman-Solin 1993, Montgomery 1997). An early formulation of this rule is found in Murray (1873), who notes for nineteenth-century lowland Scots that — outside of 2nd and 3rd singular forms, which always bear inflection — present-tense verbs are (variably) marked with an -s suffix unless the subject is an immediately adjacent personal pronoun. From a historical perspective, then, it is this last environment which triggered the greatest loss of present-tense inflectional suffixes in northern varieties in the Middle and Early Modern English periods. In vernacular Scottish and Scots-Irish, this trend resulted in categorical or near-categorical loss of the suffix in the context of a non-3rd singular personal pronoun subject, yet its variable retention with other subject types. The same phenomenon also occurs (most obviously in 3rd person plural contexts) in the Scots-Irish varieties transported to America by eighteenthcentury migrants (see Montgomery 1997). There, it has been suggested (e.g.
Verbal -s reconsidered
Montgomery et al. 1993) that in areas of early Scots-Irish settlement the Northern Subject rule was adopted into AAE. However, the earliest interaction between enslaved Africans and British and Irish emigrants in the New World took place in the Caribbean and in the coastal areas of the southern American colonies, a century prior to Scots-Irish migration. There, Africans would have come primarily into contact with speakers of vernacular southern British English; in the Caribbean, much of the indentured population would have originated in southwest England and southern Ireland (see e.g. Campbell 1967, Rickford 1986). Rickford (1997: 331), among others, points to the predominance of slaves from the Caribbean in the early Black population of many American colonies. Therefore, to understand the historical transmission of verbal -s to early African Americans, it is crucial that its status be ascertained in regional southern British varieties of earlier centuries.
3. Present-tense inflection in Early Modern English Recent corpus-based studies which have examined the status of verbal -s in Early Modern English have shed some light on this issue. These include the Early Modern British section (1500–1710) of the Helsinki Corpus (Kytö 1993); the Corpus of Early English Correspondence, covering the period 1417–1681 (Nevalainen & Raumolin-Brunberg 2000); London prisoners’ depositions dating from 1562 to 1623 (Wright 2002); and southern British literary works from much the same period (Schendl 1996). The present-tense inflectional patterning revealed by these studies of what can be broadly described as the emerging southern British standard resembles in large measure that of presentday standard English — apart of course from the range of suffixal variants available in ENE (zero, southern -(e)th, northern -(e)s and midlands -(e)n). The studies, however, do provide evidence of the existence of a low-frequency subjecttype constraint for lexical verbs. Thus, like more northerly varieties, these ENE corpora display a ‘they-constraint’ (cf. Wright 2002), i.e., almost categorical loss of the verbal inflection with adjacent 3pl personal pronoun subjects. In ‘other NP’ 3pl environments, a verbal suffix is variably present, though clearly as a minority variant (e.g. less than 20% overall in Wright’s data). Relatively small numbers of ‘other NP’ 3pl tokens in these ENE corpora mean that further subject-related generalizations must remain tentative. However, taken as a whole, these studies suggest that the southern ENE 3pl pattern was not identical to the northern British pattern of the period, where a
3
4
Sandra Clarke
verbal suffix readily co-occurred with a simple adjacent NP subject. In the ENE corpora, rather, 3pl suffix retention appears most common with two particular subject types. The first of these consists of plural subjects construed as notional or grammatical singulars, such as the conjoined singulars in (3), dummy existential there in (4), and the relative pronoun in (5), which is unmarked for plurality. The second involves an empty subject slot, often following a coordinate conjunction, as in (6): (3) My pride and obstinacy wants humbling and melting (Kytö 1993: 120) (4) …there goeth very many oute of London thither (Wright 2002: 255) (5) They laugh that wins (Shakespeare, Othello 4.1.121; Schendl 1996: 150). (6) And these dread curses… recoil And turns the force of them upon thyself (Shakespeare, 2 Henry 6.3.2.330; Schendl 1996: 151)
4. Evidence from conservative southern British vernacular speech A difficulty with the above ENE corpora, relative to the reconstruction of the history of transatlantic verbal -s, is that — with the possible exception of Wright (2002) — the register they represent approximates more the emerging standard than the actual regional vernaculars spoken by many transatlantic migrants and deportees. In the relative absence of reliable written records or contemporary descriptions of the regional vernaculars of the period of early transatlantic migrations, a chief source of evidence must be found in descriptions of vernacular regional dialects of later centuries. Yet a number of nineteenth- and early twentieth-century regional dialect studies of conservative southern British vernaculars suggest a patterning that was not characterized by a subject-type constraint resembling the Northern Subject Rule. Rather (at least outside of East Anglia) such descriptions indicate a pattern of robust use of verbal -s throughout the paradigm, irrespective of subject type. For rural West Country English, for example, such authors as Elworthy (1877: 51; ex. (7) below) and Wilson (1914: 27–28; ex. (8)) document the ready occurrence of a verbal suffix in the environment of an adjacent personal pronoun: (7) Wee wuurkus, muyn! (“We work, mind (= remember)”!) (8) Dhay treiz hard (“they tries hard”)
A similar finding emerges from the mid twentieth-century Survey of English Dialects (SED), Klemola’s (2000: 333) summary of which attests to the general
Verbal -s reconsidered
resistance of southern British conservative vernaculars to the adoption of a northern-like subject-type constraint: The northern subject rule was alive during the time the SED was conducted in the 1950s: in northern dialects, where the -s form occurs widely with full NP subjects, only scattered examples of -s form are found when the subject is a personal pronoun adjacent to the verb, whereas the type of subject NP (pronoun vs. full NP) does not affect the occurrence of -s in the southern areas.
Descriptive studies of regional varieties, however, are typically qualitative rather than quantitative in nature. Even conservative vernacular southern British varieties which differ from the northern type in their ready use of verbal -s with adjacent pronominal subjects could still be governed by a historically-transmitted subject-type constraint. Though this might not take the typical form of (near-) categorical absence of inflectional suffix in the environment of an adjacent personal pronoun, it might manifest itself via significantly lower frequencies of -s usage in such a context. This is exactly what is reported by Godfrey and Tagliamonte (1999) for a twentieth-century Devon dialect. A similar finding has been documented by Poplack and Tagliamonte (e.g. 1989, 2001) for several New World diaspora enclave varieties of AAE. It is evidence such as this which underlies the conclusion that a subject-type constraint must have been a general feature of southern British vernacular speech at the time of the early transatlantic migrations.
5. Verbal -s in Newfoundland Vernacular English This section presents quantitative evidence which runs counter to the above generalization. This evidence derives from Newfoundland Vernacular English (NVE), a variety that is in an ideal position to shed light on the issue. In terms of such factors as the time-depth of early British colonization and the origins of founder populations, Newfoundland’s settlement history offers close parallels to that of a number of early Caribbean and southern American colonies. Like these, Newfoundland is one of Britain’s earliest overseas colonies, with continuous British settlement from the first decades of the seventeenth century. Well over 90% of Newfoundland’s founder population stems from two highlycontained source areas, the West Country of England and southeastern Ireland. The island’s small population, endocentric community structure, peripheral geographical location and relative isolation from mainland North America have
5
6
Sandra Clarke
led to a situation of source dialect feature preservation to a degree rarely paralleled elsewhere in the New World (see for example Clarke forthcoming). Verbal -s patterning in NVE has been investigated quantitatively in one previous study (Clarke 1997), involving the informal speech of 24 conservative rural Newfoundlanders representing two age groups (over 65 vs. under 35). Outside the 3sg (marked categorically by -s), the sample displayed a high (56%) overall ratio of -s suffixing on lexical verbs. In terms of shared constraints on the presence of non-3sg verbal -s, the NVE corpus exhibited clear similarities to corpora which have played a significant role in the AAE origins debate (cf. Poplack and Tagliamonte 1989, 2001; Godfrey and Tagliamonte 1999). There was however one obvious discrepancy: the NVE corpus yielded no trace of a subject-type constraint. However, the fact that the younger age group made substantially greater use of verbal -s raises the possibility of change in progress, in the form of the loss of a constraint that might have characterized the grammars of the island’s earliest settlers. In the absence of reliable written depictions of the vernacular speech of Newfoundland’s (often non-literate) founder populations, the principal source for reconstructing such patterns must be found in the recorded speech of subsequent generations. This source has yielded a second corpus of presenttense lexical verbal -s usage in NVE, the time depth of which parallels that of the earliest of the AAE diaspora corpora reported on by Poplack and Tagliamonte (1989, 2001). This corpus, drawn principally from Halpert and Widdowson (1996), involves 16 rural working-class speakers born between 1872 and 1904, all but three of them prior to 1900. The corpus consisted of 765 non-3sg present-tense lexical verb tokens (following removal of 555 3sg tokens which, once again, displayed categorical -s suffixing).1 As in the case of the data set analysed in Clarke (1997), verbal -s presence was investigated via the computerized variable rule analysis program Goldvarb 2, in terms of a full range of constraints which have proven of significance in the AAE origins literature. The constraint (or factor group) of concern here is subject type, which opposed adjacent personal pronouns (as in (9) below) to ‘other NP’ subjects, including adjacent simple NPs (in (10)) and non-adjacent pronominals (in (11)):
1.694 of the tokens in this corpus were obtained from the transcribed tape-recorded collection of Newfoundland folktales published in Halpert and Widdowson (1996). The remaining 71 tokens were drawn from two tape-recorded free-speech interviews in the collection of the Memorial University Folklore and Language Archive in St. John’s, Newfoundland.
Verbal -s reconsidered
(9) You can have the best meal they cooks (Halpert & Widdowson 1996: 274) (10) … when the boats comes in (506) (11) [By and by they come (past tense) along] an’ picks un (= him) up an’ carried un out an’ — dumps un off in the river. (794)
As in the case of the Clarke (1997) corpus, this second NVE corpus also yielded no significant results for type of subject. That is, in non-3sg contexts, the tendency for verbs to bear an -s as opposed to a zero inflection was just as strong with adjacent personal pronoun subjects as with other subject types (see Table 1). Table 1.Non-3sg verbal -s, Newfoundland Vernacular English corpus. Subject type
Factor weight [ ] = non-significant
Adjacent personal pronoun [.50] ‘Other NP’ [.48]
% of -s marking (Total) N of tokens 78 85
718 47
Because the corpus contained only 95 3pl lexical verb tokens, no separate 3pl variable rule analysis was warranted. However, there was no evidence of a they-constraint, since the rate of -s inflection for verbs with an adjacent they subject (84%, or 62/74) was very similar to that for ‘other NP’ subjects (86%, or 18/21). The two NVE corpora prove consistent then with respect to absence of a subject-type constraint. Yet can it be assumed that this absence also characterized the speech of many early transatlantic emigrants from southwestern Britain and southern Ireland to Newfoundland, as well as to more southerly New World destinations? The cumulative evidence suggests that such would have been the case. One important component of this evidence is to be found in the quantitative results emerging from both NVE corpora in terms of non-subjectrelated constraints on verbal -s usage (i.e., those involving aspectual, phonological and syntactic conditioning). Though a review of these is beyond the scope of this paper, the NVE results exhibit marked similarities to those which have emerged for other conservative varieties, in particular those that have figured in the AAE origins debate. This renders it less likely that, subsequent to initial settlement, Newfoundland English would have undergone major change with respect to only one of these constraints — subject type. Rather, the NVE evidence suggests that at the time of early New World settlement, a subject-type constraint had not been consistently adopted across southern British vernacular varieties. Indeed, as noted earlier, resistance to the spread of the they-constraint
7
8
Sandra Clarke
was still apparent in at least some non-exported conservative southern British varieties until well into the 20th century, as evidenced by the SED (cf. Klemola 2000, Wright 2002).
6. Similarities in verbal -s patterning: Historical relationship, or alternative explanation? The Newfoundland results suggest that since the Middle and Early Modern English periods the historical trajectory of verbal -s over time has been complex and multi-stranded. While some of the varieties reviewed above are characterized by the near categorical adoption of a they-constraint, others display a lesser degree of inflectional loss in the environment of an adjacent personal pronoun subject. Still others, among them conservative Newfoundland English, have never adopted a subject-type constraint over the course of their history. Since the verbal -s inflection was originally a northern British variant, its appearance in southern British varieties is generally viewed as the result of diffusion from the north, and the spread of a they-constraint is attributed to the adoption of northern-like constraint patterns governing verbal inflection (e.g. Schendl 1996). Likewise, the finding of similarity in verbal -s patterning among English varieties on both sides of the Atlantic is ascribed to historical connection or common ancestry — a crucial factor in the AAE origins issue. A shared linguistic history, however, offers by no means the only interpretation of the observed similarities. In the remainder of this paper, I argue that an alternative explanation lies in cross-linguistic tendencies that are grammatical, cognitive and psycholinguistic in nature. This is best illustrated by means of two small data sets, representing present-tense inflectional usage in standard Englishes separated by almost three centuries. The first (see Table 2) derives from one of the few early eighteenthcentury manuscripts written by a resident Newfoundlander: a 1718 report by William Taverner, a Newfoundland-born member of a Dorset merchant family involved in the Newfoundland fishery. Though Taverner’s present-tense usage differs markedly from that of the NVE analysed earlier, it is clearly representative of the educated written standard of the period. Like the ENE corpora outlined above, it contains low-frequency (in Taverner’s case, under 20%) 3pl verbal -s. The second corpus (see Table 3) represents standard North American English at the turn of the twenty-first century; it contains 75 tokens of 3pl
Verbal -s reconsidered
Table 2.3pl verbal -s, Taverner 1995 [1718]. Subject type
-s
zero
% -s per subject type
Adjacent they 0 ‘Other NP’ subject: Simple adjacent plural noun 0 Heavy NP (= NP + PP with plural local noun) 0 Conjoined singular NPs 2 Relative clause 4 Empty subject slot: Non-overt subject after and/but 1 Inversion 3
10
0
14 6 3 7
0 0 40 36
4 1
20 75
Total
45
18
10
‘errors’ containing the verbal suffix -s, collected by the present author between 1999 and 2002 from academic (faculty and student) writing, as well as North American print, radio and television sources.2 Strikingly, in the two corpora, the environments which trigger presence and absence of 3pl -s are highly similar. In both, verbal -s is categorically absent when the subject is adjacent they. Unlike northern British varieties, these corpora also, with very few exceptions, do not exhibit co-occurrence of verbal -s and an adjacent simple NP bearing an overt plural marker.3 3pl verbal -s presence, however, is readily explicable in terms of factors at least some of which have been commented on by grammarians over the centuries (see e.g. Poplack and
2.Though not part of the NVE corpora discussed earlier, non-existential be was included in both the Taverner (n = 20, or 36% of the total) and the ‘error’ data sets (n = 51, or 68%). In both, 3pl is occurs in the same contexts as does the -s suffix on 3pl lexical verbs. However, existential be examples (e.g. Taverner’s “There is Two Rivers empty Themselves into it…”) were omitted, because of the long-standing and high-frequency tendency towards verbal -s (i.e., is) in this construction, even among standard speakers (cf. Meechan and Foley 1994) — a tendency which may well be linked to some of the factors to be outlined below. 3.Overt plural marking has likewise proven a disfavouring environment for 3pl verbal -s in other non-northern-British corpora, among them the conservative Helsinki Devon (Stigell) corpus of the 1970s (Peitsara 2002). There, verbs whose subject NPs carried plural marking (e.g. farmers, children) were considerably less likely to exhibit -s inflection than those whose subjects were unmarked for grammatical number (e.g. sheep, people). The three exceptions in the current corpora all occur in the more recent ‘error’ data set, and are readily interpretable as grammatical plurals conceived of as notional singulars (e.g. the CBC radio example “The Mokami Players is presenting their original script”; and “Their fifteen minutes is well and truly up”, from The Globe and Mail, Oct. 10/01).
9
10
Sandra Clarke
Table 3.3pl ‘errors’ in standard North American English, 1999–2002. Subject type
-s
% of total corpus
Adjacent they ‘Other NP’ subject: Simple adjacent plural noun Heavy NP (usually NP + PP with singular local noun) Conjoined singular NPs Relative clause Empty subject slot: Non-overt subject after and/but
0
0
3 (see note 3) 28 25 14 5
4 37 33 19 7
Total
75
Tagliamonte 2001: 168–173). In the more recent ‘error’ corpus (though not in Taverner, perhaps because of its less spontaneous style), 3pl -s is obviously triggered by psycholinguistic processing constraints. In approximately onethird of instances (n = 28) the writer or speaker appears to lose track of grammatical agreement as a result of the physical separation of verb and subject head noun. As in the case of the heavy NP subject in (12), verbal agreement is made with the (singular) local noun rather than the plural head noun: (12) ….how differences in speakers’ position in the social hierarchy affects their use of linguistic variants. (University student essay, 1999)
Otherwise, 3pl -s clusters in two well-defined sets of contexts which are identical in both corpora — contexts which also were seen to favour 3pl -s presence in the ENE corpora reviewed above. The first involves cases where a logically plural subject is processed as a singular, thereby triggering standard 3sg verbal -s grammatical agreement. This context may for convenience be divided into two subtypes: on the one hand, notionally plural subjects construed as notional singulars (e.g. the conjoined singulars in (13) and (14)); and, on the other, grammatical subjects which carry no overt plural marking, a typical case of which is the (unmarked) relative pronoun that or which, as in (15) and (16): (13) Spout Cove and East Bay is Tolerably good for Salmon… (Taverner 1995[1718]: 14) (14) I guess Mom and Pop is watching you right now. Where’s Mom and Dad? (CTV interview, Oct 30/01) (15) that harbour has Two Rivers which falls into it… (Taverner:12)
Verbal -s reconsidered
(16) The most important part is interpreting… the weights or probabilities that is associated with each of the factors. (University student exam answer, 1999)
The second general set of contexts which favour 3pl -s involves an empty subject slot, that is, absence of an overt pre-verbal subject. While in both corpora this typically arises in a situation of parataxis, i.e. in a conjoined clause after and/but (see (17) and (18) below), it may also occur in environments involving subject-verb inversion, as in (19): (17) The Indians of Great Britton Freqly. hunt, and takes furrs… (Taverner:17) (18) …where many traditional dialect features…have survived…. and still thrives among the older generation. (University student essay, 1999) (19) Falls Three great Rivers into it.. (Taverner:13)
The conjoined clause environment is one that has figured prominently in the formulation of the Northern Subject Rule. Yet historical transmission offers by no means the only source for the suffixal patterning in the above examples. An alternative explanation can be sought in a general cognitive principle, formulated by Givón (1991:87) as the ‘quantity principle’, namely that “less predictable information will be given more coding material”. Here then, the absence of a grammatical subject from its normal declarative preverbal slot may result in additional marking on its verb, in the form of an overt (and default-like) -s suffix. A common thread across many of the non-standard varieties outlined in this paper is the tendency towards prohibition or reduction of inflectional suffixing when the subject is an adjacent personal pronoun. Here, too, factors other than historical relationship are in all likelihood at play. On the basis of cross-dialect and cross-linguistic evidence, Börjars and Chapman (1998) postulate a highly plausible reason for the trend towards absence of verbal suffix with adjacent they subjects: this is the tendency for adjacent personal pronouns to procliticize to their verbs. As clitic-like elements, non-3sg personal pronouns would thus bear a relationship of complementary distribution to the verbal -s suffix. Indeed, the grammaticalization of free pronominals to inflectional affix status represents a well-known cross-linguistic tendency. In conclusion, a number of factors other than historical relationship or transmission militate in favour of verbal suffix loss in the context of an adjacent overt personal pronoun subject, as well as suffix retention in particular types of ‘other NP’ contexts. The various data sets reviewed in this paper suggest that
11
12
Sandra Clarke
such factors have played a role even in standard-like varieties of English over the course of the entire Modern period. This paper has also demonstrated that, at the same time, at least some conservative southern-British-based varieties display little if any trace of a subject-type constraint, among them the vernacular spoken English of Newfoundland. As a consequence, although it has been employed as an important historical diagnostic in the African American English origins debate, the value of a subject-type constraint in ascertaining close historical connection between conservative southern British and transplanted overseas vernacular varieties must be seriously questioned.
References Börjars, Kersti & Carol Chapman. 1998. “Agreement and Pro-drop in Some Dialects of English”. Linguistics 36.71–98. Campbell, Mildred. 1967. “Social Origins of Some Early Americans”. Essays on American Colonial History ed. by Paul Goodman, 164–180. New York: Holt, Rinehart & Winston. Clarke, Sandra. 1997. “English Verbal -s Revisited: The Evidence from Newfoundland”. American Speech 72.227–259. ———. forthcoming 2004. “The Legacy of British and Irish English in Newfoundland”. Transported Dialects: The Legacy of Non-standard Colonial English ed. by Raymond Hickey. Cambridge: Cambridge University Press. Elworthy, Frederic Thomas. 1877. An Outline of The Grammar of the Dialect of West Somerset (= Transactions of the Philological Society, 1877–79): 143–257. London: Trübner. Givón, Talmy. 1991. “Isomorphism in the Grammatical Code: Cognitive and Biological Considerations”. Studies in Language 15.85–114. Godfrey, Elizabeth & Sali Tagliamonte. 1999. “Another Piece for the Verbal -s Story: Evidence from Devon in Southwest England”. Language Variation and Change 11.87–121. Halpert, Herbert & J. D. A. Widdowson. 1996. Folktales of Newfoundland. 2 vols. St. John’s, Newfoundland: Breakwater. Klemola, Juhani. 2000. “The Origins of the Northern Subject Rule: A Case of Early Contact?” The Celtic Englishes II ed. by Hildegard L. C. Tristram, 329–346. Heidelberg: Universitätsverlag C. Winter. Kytö, Merja. 1993. “Third-Person Present Singular Verb Inflection in Early British and American English”. Language Variation and Change 5.113–139. Meechan, Marjory & Michele Foley. 1994. “On Resolving Disagreement: Linguistic Theory and Variation — There’s Bridges”. Language Variation and Change 6.63–85. Meurman-Solin, Anneli. 1993. Variation and Change in Early Scottish Prose. Helsinki: Suomalainen Tiedeakatemia. Montgomery, Michael. 1997. “Making Transatlantic Connections between Varieties of English”. Journal of English Linguistics 25.122–141.
Verbal -s reconsidered
———, Janet M. Fuller & Sharon DeMarse. 1993. “‘The Black Men Has Wives and Sweet Harts [and Third Person Plural -s] jest like the White Men’: Evidence for Verbal -s from Written Documents on Nineteenth Century African American Speech”. Language Variation and Change 5.335–357. Murray, James A. H. 1873. The Dialect of the Southern Counties of Scotland. Its Pronunciation, Grammar, and Historical Relations. London: Philological Society. Nevalainen, Terttu & Helena Raumolin-Brunberg. 2000. “The Changing Role of London on the Linguistic Map of Tudor and Stuart England”. The History of English in a Social Context ed. by Dieter Kastovsky and Arthur Mettinger, 279–337. Berlin & New York: Mouton de Gruyter. Peitsara, Kirsti. 2002. “Verbal -s in Devonshire — The Helsinki Dialect Corpus Evidence”. Variation Past and Present. VARIENG Studies on English for Terttu Nevalainen ed. by H. Raumolin-Brunberg, M. Nevala, A. Nurmi & M. Rissanen. Helsinki: Société Néophilologique. Poplack, Shana & Sali Tagliamonte. 1989. “There’s No Tense like the Present: Verbal -s Inflection in Early Black English”. Language Variation and Change 1.47–84. ———. 2001. African American English in the Diaspora: Tense and Aspect. Oxford: Blackwell. Rickford, John R. 1986. “Social Contact and Linguistic Diffusion”. Language 62.245–289. ———. 1997. “Prior Creolization of African-American Vernacular English? Sociohistorical and Textual Evidence from the 17th and 18th Centuries”. Journal of Sociolinguistics 1.315–336. Schendl, Herbert. 1996. “The 3rd Plural Present Indicative in Early Modern English — Variation and Linguistic Contrast”. English Historical Linguistics 1994 ed. by Derek Britton, 143–160. Amsterdam: John Benjamins. Taverner, William. 1995. [1718]. “Taverner’s Second Survey”. Newfoundland Quarterly LXXXIX.9–18. Wilson, Sir James. 1914. The Dialect of the New Forest in Hampshire (as Spoken in the Village of Burley). London & New York: Oxford University Press. Wolfram, Walt, Erik R. Thomas & Elaine W. Green. 2000. “The Regional Context of Earlier African American Speech: Evidence for Reconstructing the Development of AAVE”. Language in Society 29.315–355. Wright, Laura. 2002. “Third Person Plural Present-tense Markers in London Prisoners’ Depositions, 1562–1623”. American Speech 77.242–263.
13
Do grammars change when they leak?* David Denison University of Manchester
1.
Introduction
The organisation of this programmatic paper is as follows. Limitations of space mean that the treatment is necessarily both selective and allusive. Some general remarks about language change (§2) are followed by criticism of the conventional view of synchronic grammar (§3) and a consideration of reanalysis in diachrony (§4). I then discuss ‘leakiness’ of the grammar of the English NP in synchronic terms (§5) and suggest graduated change as a plausible alternative to reanalysis for some kinds of diachronic change (§6). The appearance and loss of intermediate forms are considered in §§7–8, and my example of reanalysis reappears in the conclusion (§9).
2. A typology of language change1 All living languages are subject to change. How do they change? Most linguists now distinguish between actuation or innovation on the one hand and propagation or diffusion on the other. Diffusion is the spread of a variant from the point where it has become an option for a number of speakers. Diffusion of change requires the prior existence of variants — alternative ways of saying the same thing. Some variation appears to be stable over long periods (RaumolinBrunberg 2002): variation is a necessary but not a sufficient condition for change to occur.
*I am grateful to Bill Croft and to the audience at 12ICEHL for comments, and to my daughter Rosie’s timely contribution on (screen-)saving face. 1.This section is adapted from draft material prepared for Hogg and Denison (in preparation).
16
David Denison
And why does language change? The answer to this question is, of course, dependent on the theoretical position of the analyst. If the most salient property of language is its grammar, an internalised set of rules unconsciously built up and used by an individual speaker-hearer, then language change may be seen as a consequence of new generations inducing a slightly different grammar from that internalised by their parents’ generation, perhaps because of a slightly changed preponderance of some usage, the cause of which is not really grammatical in origin but some contingent ‘performance’ factor. The new grammar in turn leads to a further change in the output of its speakers, and so things move on. The process of language acquisition in childhood will be critical, and the favoured form of analysis will (usually) be formal and structural. Alternatively, if language is something which crucially belongs to and exists in a speech community, then speaker interaction and relative social status may be the fundamental engines of language change. Speakers may adjust their usage to (or against) community norms throughout adolescence and perhaps beyond, and change is not confined to the acquisition process. The requisite analysis will be sociolinguistic and statistical. Then again, if speakers and hearers are regarded as autonomous individuals, anxious above all to maximise their communicative efficiency, yet other considerations may be identified, typically involving speaker intentions. We can group three main types of force for change under the headings structural, social and functional. Each has its own champions in the literature, though the dividing lines are not always clear-cut. Beyond these three broad categories there are extralinguistic factors to consider too. Here I will mainly be looking at change from the structural and functional points of view. For a fuller discussion see Croft (2000).
3. Synchrony without leakage Many approaches to synchronic language description start from the working hypothesis that grammars don’t leak. That is, there are a certain number of parts of speech or word classes (the number may depend on the theory or the analyst), and every word in a particular utterance of a sentence belongs to just one of them. Constituent structure analysis gives a unique bracketing of these words. (In a theory with movement there may be a series of structures in the derivational history.) Utterances which appear not to fit are either 1. ignored 2. or arbitrarily made to fit
Do grammars change when they leak?
3. or assumed to be in need of a better analysis which remains to be discovered 4. or relegated to ‘mere’ performance (or E-language) 5. or — in extremis — accepted as anomalous but regarded as minor and peripheral parts of the language system. This is unfair to many linguists, but as far as mainstream formal syntax is concerned, not entirely unjustified. What utterances don’t fit? Here is a simple example. If we consider the sentences (1) Jim bought a watch in the market. (2) Jim put his bicycle in the shed. (3) Jim gave money to charity. (4) Jim took pleasure in mathematics. (5) Jim paid attention to his teacher. (6) Jim set fire to the car.
we find that each has different properties despite its superficially similar NP1 V NP2 PP structure. For example, the PP in (1) is an adjunct, whereas the PPs in (2) and (3) are generally regarded as complements; in (3), though, but not (2), the PP commutes with an indirect object. The PP of (4) is syntactically somewhat equivocal — adjunct or complement? — while at the lexical level one might recognise take pleasure in as a transitive group-verb equivalent to enjoy. Similar remarks could be made about (5), though the syntax is not identical: (5) passivises much more readily. So does (6), but now the verbal idiom is closer to being frozen and NP2 is not open to modification; indeed we could ask whether it makes sense to assign fire to the word class Noun at all. We could go on adding sentences with shades of variation and subtly different behaviour until a structuralist model ran out of ways of distinguishing items. The working hypothesis cannot be upheld.
4. Reanalysis Even a watertight system of synchronic analysis still gives many possibilities of change by creation of variants and selective preference among them. One such route is by reanalysis. We need an example. There was (and still is) a prepositional verb run over:
17
18
David Denison
(7) The rabbit ran over the meadow. (8) My car ran over a bottle (lying in the road). (9) [VP ranintr [PP over [NP a bottle] ] ]
Over is a preposition. However, the same linear string, ran over a bottle, is in principle open to reanalysis in English so that over becomes an adverbial particle in a transitive phrasal verb: (10) [VP rantrans [part over] [NP a bottle] ]
(As it happens, no one has come up with a wholly satisfactory formal analysis of either prepositional or phrasal verbs, so the reader’s pet analyses won’t necessarily be the same as mine in (9) and/or (10). All that is needed is agreement that analyses of the two should be different from each other.) Why should this reanalysis take place? It is actually unlikely in the context of (8) but far more likely if the pragmatic situation is: (11) My car ran over a bottle (standing in the road).
This time the bottle is definitely affected by the action: it will be knocked over and in all probability broken. Hence the re-interpretation of over: it need no longer refer to the trajectory of the car across and above the obstacle; now it can refer to the trajectory of the bottle away from the upright position and is resultative. Thus run over is no longer a prepositional verb like run into, but a phrasal verb like knock over. It has been reanalysed, in this instance both semantically and syntactically. Of course, sadly, we take more interest in such things when the victim is human: (12) A drunk driver ran over two pedestrians.
We see the change semantically in that the car need not literally pass on top of its victims. And we see it syntactically in that we will now also begin to attest sentences like (13) A drunk driver ran two pedestrians over.
not previously possible. I chose this rather grisly example because the reanalysis is so intuitively clear. Nevertheless it would be unfair to let this example represent all syntactic reanalyses, (a) because this reanalysis is clearly not purely a matter of syntax, and (b) because even without the change in run over (which is historically attested, incidentally), such everyday patterns as go for and work out are rather embarrassing for a syntax without constructions. On the other hand, whether
Do grammars change when they leak?
reanalysis ever is purely syntactic is questionable (Bill Croft, personal communication, 2 Dec. 2002). I cannot enter into this question here. What is the mechanism of reanalysis? I don’t mean the technical mechanism, which will depend on the particular syntactic model in use; I mean the social mechanism. In diachrony I take it to be the attribution by a younger generation of choices from the watertight grammar for the analysis of some pattern which differ from the analysis of the older generation, whether different word class assignment or different structure or both. Remember: change implies variation. The variants in this case are implicit ones, alternative structures for the same explicit form, and on the individual level the change occurs during acquisition. Statistically the change will enter the language as the new analysis gradually prevails over time. This is essentially the mechanism of Lightfoot (1979) and many succeeding works. Another mechanism allows for more than one grammar in a population at the same time. The idea of competing grammars is found in Pintzuk (1991) and other works. It was devised in order to account for explicit variation in synchronic grammar, but I guess it could be used for the kind of implicit, underlying variation in the string analysed above. One of my problems with that approach is the prospect of a snowballing multiplicity of grammars to account for all the variation encountered in practice, and allowing it to deal with lexical variation like this would exacerbate that problem enormously. Yet another approach has been tried for a specific historical problem by Wim van der Wurff (1992), where he imagines a scenario of contact between dialects (or social networks), in which an innovation in one group has to be interpreted in a particular way by speakers of another group with a different grammar who encounter it, leading to reanalysis by them and starting a chain of developments as one group’s output is reacted to by another. Something similar was recently suggested by Anthony Warner for the history of do (2002). Notice here that reanalysis and change is not confined to the acquisition phase. I am very sympathetic in general to accounts which allow for significant changes in an individual’s language during adolescence at least and perhaps right through adulthood. But the scenarios demand enormous ingenuity and are necessarily highly speculative. Although there are difficulties with all these models of reanalysis, I certainly don’t wish to reject reanalysis outright. But here I want to follow a different tack. Reanalysis of an individual form moves instantaneously from one watertight analysis to another (even if the consequences take time to play out and/or spread). But synchronic grammar just isn’t watertight. The prior assumption is wrong.
19
20
David Denison
5. Synchrony leaks Edward Sapir’s (1921: 38) dictum that “All grammars leak” has long been practically a motto of mine, though I (and probably others too) have unintentionally misappropriated it: Sapir was specifically writing about the mapping between form and meaning. It isn’t entirely clear that he is endorsing the interpretation which I wish to attribute to the saying, namely that the categories and structures imposed by grammarians on a language are never watertight.2 Nevertheless, that is the sense in which I intend my title to be read. Here I will concentrate on simple NPs. The basic structure used by the Cambridge Grammar (Huddleston & Pullum 2002: 331) is (14)
NP Determiner D
the
Head: Nom Modifier A
Head N
old
man
They give reasons for not using the Abney-style DP analysis. And note that they use ‘Determiner’ for the function, as in (14), whereas I will stick to old habits and use it for the category. Something like the three-way distinction of D, A, N is very widely accepted cross-linguistically and in English grammar, in the majority of frameworks. What does it buy us? Determiners are function-words, of general semantics (definiteness, quantification, etc.), and mutually exclusive. Adjectives are lexical, content words, and can be iterated. ‘Lexical’ is an imprecise term which tends to involve specificity of meaning and openness of class. A summary of some key properties is given in Table 1. I have added Prn,3 since pronouns can act as head of NP, like N, but differ
2.Later he discusses the arbitrariness of parts of speech and concludes “how they not merely grade into each other but are to an astonishing degree actually convertible into each other” (Sapir 1921: 118). Once again, though, this is really to do with cross-linguistic comparison rather than assignment of an individual word in one language to a word class. 3.I have shortened pronoun as Prn to distinguish explicit pronouns from the empty category Pro of GB Theory.
Do grammars change when they leak?
Table 1.Distinctive features of some NP categories.
lexical iterate number marking comparison case marking can act as predicate
D
A
N
Prn
− − (−) − − −
+ + − + − +
+ (−) + − − −
− − + − + ?
in many other ways, including general incompatibility with D and A. Almost every cell needs a bit of qualification, as we tell our students: 1. 2. 3. 4. 5.
Prn can co-occur with A in certain circumstances: lucky me, poor you N can iterate, though not as head of NP: pit-head strike ballot not all A show comparison: potential, mere not all A can occur in predicative position: potential, mere some A always occur in predicative position and do not function attributively in NP at all: asleep
So far these subtleties do not vitiate the basic D-A-N analysis of the standard NP. I should emphasise that I think that D, A and N are useful descriptive generalisations, probably with genuine psychological reality. However, in categorial terms it might be interesting to arrange the four categories mentioned so far as in diagram (15): (15)
Prn D
N A
The reason is that every boundary between adjacent categories is leaky in PDE, whereas Prn and A are entirely distinct, and so too (I think) are D and N. Let me briefly exemplify the shading of one category into another. For the A~N boundary I refer to Denison (2001), for example the word key, found in PDE at various points along the continuum from clear N to almost wholly adjectival uses like: (16) There are several reasons. The most key one for many victims is (2001 Women’s Hour, BBC Radio 4 (15 Nov.))
…
21
22
David Denison
(17) There are key fundamental flaws in this method. (2002 Damian Byrne, abstract of article in online Journal of Language and Linguistics 1.1)
For Prn~N, we have the analytical possibility of treating Prn as a subcategory of N, since they are the two categories which can normally be head of NP. Historically there are shifts in the usage of OE mann to man, and arguably of self, body, and so on from the domain of N to that of Prn. The relation between D and Prn is interesting. On one approach, the core of the D category is the articles, since almost their sole function is to determine a noun, and they cannot occur in any pronominal (that is, pro-NP) function. The remainder of the central determiners are then defined, at least in part, by mutual incompatibility: any NP item which cannot co-occur with the articles is itself a determiner too. However, as is well known, apart from the articles, almost every determiner can occur in pronominal function. (18) What do you think of this? (19) Some are born great.
There are good reasons to analyse such items in two different ways: as an NP whose head (and sole constituent) is Prn, bringing out the analogy with (20) What do you think of him? (21) They are born great.
or as headless NPs with a missing or ellipted N, bringing out the analogy with (22) What do you think of this idea? (23) Some people are born great.
If we treat the possessive determiner my and the pronoun mine as variants of the same form, then there is even more overlap between D and Prn. (Historically, of course, that is exactly what they were.) Apparently rather similar is no and none. The status of pronominal determiners is tricky, therefore, reflected in a rather confusing analysis in Quirk et al. (1985: 870–873). Huddleston and Pullum — normally so decisive — actually treat pronominal determiners as a kind of blend, what they call the fused-head construction. I reproduce their diagram of an explicitly partitive fused-head (Huddleston and Pullum 2002: 412):
Do grammars change when they leak?
(24)
NP Head: Nom Det-Head: D
few
Comp: PP
of her friends
This is meant to capture the idea that certain kinds of pronominal determiner (though not personal pronouns, as in (25)–(26)), are simultaneously the grammatical part of the NP, acting as modifier to the lexical part, and the head. Coming at the gradient from the other side, note that even the core of Prn, the personal pronouns, have some determiner-like uses, though only the plural ones (and not 3 plural in standard): (25) We/#us scholars must stick together. (26) You people are all wrong. (27) You idiot(s)! (28) #Them ideas are really stupid.
Only the vocative (27) allows singular as well as plural. The Cambridge Grammar regards these, or at least (25)–(27), as determinatives in category, pointing out that the predeterminer all can precede the plural items as normal (Huddleston & Pullum 2002: 374, 353 n.14). Dick Hudson has an analysis in which Det and Prn are unified as a single category on the same lines as analyses which treat adverbial particles as intransitive prepositions (Hudson 2000a); and of course there is the DP Hypothesis (Abney 1987). As for the last pairing, D~Adj, here we have a veritable minefield. Numerals have long been difficult to classify, but there are subtler difficulties too. Several of us have pointed out the difficulties of classifying such both as pre-determiner and post-determiner (Denison 1998: 117–118, Mackenzie 1997) (29) such a nuisance (30) no such luck
and the recent change from (31) such another Æ another such.
23
24
David Denison
Now there are at least two analyses (Huddleston & Pullum 2002: 435, Spinillo 2003) which make a strong case for analysing such not as any kind of determiner in PDE but as A.
6. Diachrony leaks Let us extend this discussion of the Det~A boundary into diachrony. The Cambridge Grammar (Huddleston & Pullum 2002: 392–393) regards various and certain as marginal determiners in PDE because of their occurrence in a partitive ‘fused-head’ construction (certain of the delegates), general semantics, non-generic semantics, and — for certain only — use with a in (32) This gave her a certain authority.
It is interesting to compare this predominantly syntactic account with OED’s comments on the lexical history of certain: I. 1. a. Determined, fixed, settled; not variable or fluctuating; unfailing. To avoid ambiguity from confusion with sense 7, the adj. is sometimes put after its n., as a certain day, a day certain. […] II. 7. a. Used to define things which the mind definitely individualizes or particularizes from the general mass, but which may be left without further indentification [sic] in description; thus often used to indicate that the speaker does not choose further to identify or specify them: in sing. = a particular, in pl. = some particular, some definite. Different as this seems to be from sense 1, it is hardly separable from it in a large number of examples: thus, in the first which follows, the hour was quite ‘certain’ or ‘fixed’, but it is not communicated to the reader; to him it remains, so far as his knowledge is concerned, quite indefinite; it may have been, as far as he knows, at any hour; though, as a fact, it was at a particular hour. (The absolute uses are in B 4 6.) a1300 Cursor M. 8933 Ilk dai a certain hore| þar lighted dun of heuen ture Angels. [etc.]
In other words, there are two very different lexical senses of certain which can be argued to be categorially different as well, yet there are many early examples which are equivocal. Given that the gradual — or at least graduated — nature of semantic change is reasonably widely accepted, why not allow that syntactic change may proceed by small steps too? I have alluded to many synchronic examples of gradience within the NP, where a word or a construction is neither
Do grammars change when they leak?
one thing nor the other. Many will actually have started off as ‘one thing’ and moved during the history of English towards ‘the other’. Note that this is not necessarily an argument for slowness of change, merely for graduatedness of at least some changes.
7. How intermediate forms arise How and why should this happen? One mechanism would be via what Quirk called serial relationship (1965, 1968). This is a plausible idea, though not part of a fully worked out model of language, in which new intermediate forms are made possible precisely because they share characteristics with, and therefore largely overlap with, already-existing grammatical forms on either side. One of his matrices (1968:172, Table 5) is reproduced here as Table 2. Table 2.An example of serial relationship.
intends wants seems has used is may
1
2
3
4
5
6
7
+ + + + + + +
+ + + + + + −
+ + + + ? − −
+ + + + − − −
+ + ? − − − −
+ + − − − − −
+ − − − − − −
The columns 1–7 represent variant sequences in which the particular verbs shown at the left either are (+), or are not (−), attested in, for example: 4. He would X to come every day 5. He X that 6. He X us to come every day Quirk writes (1968: 172): “it is because of the gradience, in fact, that we […] may find in speech ‘mistaken’ constructions of the form He doesn’t want that anyone should …”. Later he explains (33) He was said to be foolish
in a similar way, as a passive that is fully acceptable despite the non-existence of any corresponding active (1968: 176).
25
26
David Denison
Another mechanism with some promise of forming part of a coherent linguistic theory would be dual inheritance, in a Construction Grammar framework where the properties of a given form may be inherited from more than one more general source construction (Goldberg 1995, Hudson 2000b), but contra (Bresnan 1997). Either way, intermediate forms would arise when they serve some communicative purpose and because speakers — as opposed to linguists — are not confined to producing structures and grammars which are wholly self-consistent and maximally elegant and economical. If sufficiently useful, the innovations would diffuse into the language and might lead to permanent changes in the categories or structures of its grammar. But they might well not. Genuinely intermediate forms are often unstable historically.
8. Why intermediate forms disappear If we assume that humans categorize the world — things in general, that is, and therefore also linguistic ‘things’ — roughly according to the principles identified by Rosch (1978, 1988), then one consequence is that we unconsciously adjust our categories and our categorizations to accommodate in the most satisfying way what we encounter. This was the assumption underlying Anthony Warner’s (1990) account of the development of a modal category in English. Perhaps it is not too much of a stretch to assume that one solution to categorial awkwardness is for language users to find ways of avoiding the troublesome forms. (Again, this was applied by Warner to the loss of old preterite-presents like witan and certain lexical meanings of shall, can, etc. which fitted least comfortably into the emerging modal category.) The psycholinguistic experiment with pictures of cups and bowls is well known (Labov 1973). There the language user typically does make a choice, if sometimes an arbitrary one, as to where to draw the dividing line. We have, then, the outline of a mechanism which disfavours intermediate forms. I have no explanation for why some intermediate forms seem nevertheless to show long-term stability; the best example I know is the P~A form near, though most other former P~A words have moved to one side or other of the boundary.4
4.See now Aarts, Denison, Keizer and Popova (2004: Introduction) for a review of work on mixed and gradient categories.
Do grammars change when they leak?
With two opposing principles in tension we thus allow for both stability and change; it is worth noting, though, that such a diachronic model is rather at risk of post-hoc-ery.
9. Concluding remarks When a lexical item develops a new sense, there are often equivocal ‘bridge’ examples which users do not need to allocate to older or newer senses, since either works fine in the context. It might be compared to a superposition of quantum states whose wave function only collapses when observed. Linguists, however, tend to insist on observing, whereas ordinary speakers and hearers sometimes don’t need to look analytically but only need to have the right chunk of language with the right overall interpretation. And I’m suggesting that this can happen not just in lexical semantics but in morphosyntax too. It might even happen with the example I used earlier to illustrate reanalysis: (34) The car ran over a hedgehog.
Perhaps neither speaker nor hearer will need to decide whether this is the prepositional or the phrasal verb. (And if that doesn’t work, keep substituting different animals — frog, rabbit, sheep, elk — until you find the right size for this to be true.) Much remains to be done at the diachronic level, including the tracing and verification of specific examples, and following enough of them to allow safer generalisations about the relative importance in language change of intermediate forms, and about their markedness and (in)stability. I don’t think this is merely a notational variant of reanalysis. What would such an enterprise buy us? It seems to me that the case for leakiness of grammar in synchrony is difficult to deny. The history of a language is the passage through successive synchronic states — or conversely, the simultaneous existence at any given moment of many historical states in a layered manner. (Notice that this latter re-formulation sounds very much like a description of a grammaticalisation gradient, but I intend it to apply to a wider range of changes than grammaticalisations or lexicalisations.) Either way, a picture of change which both reflects and makes use of a more plausible picture of grammar, is in my opinion a desirable aim.
27
28
David Denison
References Aarts, Bas, David Denison, Evelien Keizer & Gergana Popova, eds. 2004. Fuzzy Grammar: A reader. Oxford: Oxford University Press. Abney, S. P. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, Massachusetts Institute of Technology Bresnan, Joan. 1997. “Mixed Categories as Head Sharing Constructions”. http://wwwlfg.stanford.edu/lfg/ms/ms.html Croft, William. 2000. Explaining Language Change: An evolutionary approach. Harlow: Longman. Denison, David. 1998. “Syntax”. The Cambridge History of the English Language ed. by Suzanne Romaine, vol. 4, 1776–1997, 92–329. Cambridge: Cambridge University Press. ———. 2001. “Gradience and Linguistic Change”. Historical Linguistics 1999: Selected papers from the 14th International Conference on Historical Linguistics (Vancouver, 9–13 August 1999) ed. by Laurel J. Brinton, 119–144. (= Current Issues in Linguistic Theory, 215.) Amsterdam & Philadelphia: John Benjamins. Goldberg, Adele E. 1995. Constructions: A Construction Grammar approach to argument structure. Chicago & London: University of Chicago Press. Hogg, Richard M. & David Denison, eds. In preparation. A History of the English Language. Cambridge: Cambridge University Press. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hudson, Richard. 2000a. “Grammar without Functional Categories”. Syntax and Semantics ed. by Robert D. Borsley, vol. 32, 7–35. New York: Academic Press. ———. 2000b. “*I amn’t”. Language 76.297–323. Labov, William. 1973. “The Boundaries of Words and their Meanings”. New Ways of Analyzing Variation in English ed. by Charles-James N. Bailey & Roger W. Shuy, 340–373. Washington DC: Georgetown University Press. Lightfoot, David W. 1979. Principles of Diachronic Syntax. (= Cambridge Studies in Linguistics, 23.) Cambridge: Cambridge University Press. Mackenzie, J. Lachlan. 1997. “Grammar, Discourse and Knowledge: The use of such”. Studies in English Language and Teaching: In honor of Flor Aarts, ed. by J. Aarts, Inge de Mönnink & H. Chr Wekker, 85–105. Amsterdam & Atlanta: Rodopi. Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and change in Old English word order. PhD dissertation, University of Pennsylvania. Quirk, Randolph. 1965. “Descriptive Statement and Serial Relationship”. Language 41.205–217. ———. 1968. “Descriptive Statement and Serial Relationship”. Essays on the English Language, Medieval and Modern ed. by Randolph Quirk, 167–183. London: Longman. ———, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London & New York: Longman. Raumolin-Brunberg, Helena. 2002. “Stable Variation and Historical Linguistics”. Variation Past and Present: VARIENG studies on English for Terttu Nevalainen ed. by Helena RaumolinBrunberg, Minna Nevala, Arja Nurmi & Matti Rissanen, 101–116. (= Mémoires de la Société Néophilologique de Helsinki 61.) Helsinki: Société Néophilologique.
Do grammars change when they leak?
Rosch, Eleanor. 1978. “Principles of Categorization”. Cognition and Categorization ed. by Eleanor Rosch & B. B. Lloyd, 27–48. Hillsdale NJ: Lawrence Erlbaum Associates. ———. 1988. “Coherences and Categorization: A historical view”. The Development of Language and Language Researchers: Essays in honor of Roger Brown ed. by Frank S. Kessel, 373–392. Hillsdale NJ: Lawrence Erlbaum. Sapir, Edward. 1921. Language: An introduction to the study of speech. New York: Harcourt Brace. Spinillo, Mariangela. 2003. “On Such”. English Language and Linguistics 7. 195–210. Warner, A. R. 1990. “Reworking the History of English Auxiliaries”. Papers from the 5th International Conference on English Historical Linguistics (Cambridge, 6–9 April 1987) ed. by Sylvia Adamson, Vivien A. Law, Nigel Vincent & Susan Wright, 537–558. (= Current Issues in Linguistic Theory, 65.) Amsterdam & Philadelphia: John Benjamins. ———. 2002. “Affirmative and other DOs before 1600”. Paper presented at York-Holland Symposium on the History of English Syntax (York, 21–22 Apr 2002). van der Wurff, Wim. 1992. “Syntactic Variability, Borrowing, and Innovation”. Diachronica 9.61–85.
29
Grammar change versus language change Is there a difference?* Olga Fischer University of Amsterdam The claim, then, is that some aspects of our language capacity are not a result of learning from environmental evidence. Aside from divine intervention, the only other way we know of to get them into the mind is biologically: genetic information determining brain architecture, which in turn determines the form of possible computations. In other words, certain aspects of the structure of language are inherited. (Jackendoff 1987: 87)
1.
Preliminaries
When Ans van Kemenade, Willem Koopman, Wim van der Wurff and I were working on the book Early English Syntax (2000), we had an interesting correspondence with the editor of this new Cambridge series, David Lightfoot, about what the book should be trying to do. Lightfoot wrote to us: There is a fundamental point that has to be stressed for this introduction (and the rest of the book) to make any sense and that is the distinction between language change, the focus of most historical work over the last 200 years, and grammar change. That involves saying what you take a grammar to be … None of us draw this distinction sharply enough but it cannot be overstressed.
*I am deeply grateful to Willem Hollmann, Jet van Dam van Isselt, Niki Ritt and Victor Yngve, who have all read earlier versions of this paper and have given me valuable comments and suggestions, which have definitely improved the argument even if we do not agree on all points, but that would be hard in a matter where so much is still uncertain. I also would like to thank members of the ICEHL audience at Glasgow for lively and pertinent discussion in the coffee-breaks.
32
Olga Fischer
Further down in his letter, referring to a particular chapter I had written for the book, Lightfoot reiterates this point, given the nature of the book … and of the series in which it is going to appear, you have to take off your philological hat and wear your grammatical hat, using philological material to illuminate your grammatical claims and not pursuing philological points for their own sake.
I have been wondering (and worrying) ever since why I found it impossible to take off my “philological hat”. In this paper I will try and work out why I, as a historical linguist, feel that I have to keep it firmly on my head. In order to begin to understand what grammar change is, and how it differs from language change, we must first address the question of what ‘grammar’ is and what we understand by ‘language’. I am afraid that question, which is a hard one, harder than I thought at first when I began to do my reading for this talk, will take up most of the space allotted to me for this paper, but I hope that at the end, after I have taken you along the meandering path of my own reading and other people’s thoughts, I will be able to say a few useful things about the question of linguistic change, and what we, as historical linguists, should investigate in order to understand change, because I believe that that is our ultimate aim, an understanding of linguistic change. But, to set the scene, let us first go back to our very beginning, and consider the ‘divine’ versus the ‘biological’ option mentioned in the quotation heading this article.
2. Introduction: Language, linguistic evolution and ‘design’ Darwin’s theory of evolution has always had its harsh opponents, people who cannot accept that man is just another animal, an “evolutionary accident” as Stephen Jay Gould once put it. Instead they believe that man was carefully designed, and that God is the designer. Recently a number of books have appeared that attempt to show ‘scientifically’ that the complex design that man represents cannot have come about by mere evolution. A new ‘science’ has arisen (especially in the United States), called ‘creation science’. It makes use of many of the findings and methods of evolutionary theory. For instance, these scientists (or ‘creationists’, as they are often called) no longer believe that our planet was created six thousand years ago, and that every creature was fashioned by God in literally six days. By thus giving credit to their evolutionary opponents only in those cases where the data are simply too overwhelming to be denied, they put a good face on their proposal, which purports to show, making
Grammar change versus language change
clever use of still outstanding disagreements among the evolutionary theorists, that the gap between man and the other species is simply too wide to be plausible in evolutionary terms. Creation scientists therefore come up with the notion of ‘irreducible complexity’: “how could one irreducibly complex system”, so they argue, “ever evolve into another”, noting further “that any stepwise mutation that altered the original would have rendered it not just clumsy but useless and thus incapable of survival”. Their conclusion is, in other words, that such a complex design as man is can only be explained with reference to a designer. Frederick Crews, whose words I have just quoted, has reviewed a number of books that advocate this new ‘scientific’ theory of ‘Intelligent Design’ in an article entitled “Saving us from Darwin” in The New York Review of Books (October 4, 2001, pp. 24–27). He strongly disagrees with the creationists and shows that they misrepresent the way in which Evolutionary Theory accounts for the emergence of complex design in nature. No serious evolutionary biologist has ever argued that complex and purposeful organization in nature is brought about by nothing but a lucky sequence of mutations. Instead, the evolution of complex and functional structure is understood as a long and wasteful process of blind trial and frequent error. Only a tiny percentage of innovative variants will come to be retained, or ‘naturally selected’, namely those which happen to be as good as or better at replicating than existing variants. The point about such selections is that they can, and will, accumulate over time to produce structures so complex and apparently purposeful that they strike one as having been designed. It is important, however, that one makes a distinction between what I would call ‘purposeful’ and ‘resultative’ design: Nature does not have design in mind (she has been compared to “a scavenger that makes do with jury-rigged solutions and then improves them as opportunities and emergencies present themselves […]” (ibid., 27)), but evolution itself produces design more or less accidentally on the way. What does this have to do with my topic of today: grammar change versus language change and the difference between them? I think there are a number of links which are relevant here. First of all, Darwin’s ideas on evolution are being used by quite a few linguists coming from various schools (see e.g. Clark and Roberts 1993, MacMahon 1994, Pinker 1994, Ritt 1996, Smith 1996, Lass 1997, Lightfoot 1999, Croft 2000). What I find interesting is that, to some extent (barring the divine intervention!), the two opposing schools of thought existing within evolutionary theory can also be found among linguistic theorists when they write about the development of language (in both an evolutionary
33
34
Olga Fischer
and an acquisition sense). Lass (1997: 314ff.), for instance, using the idea of evolution, writes that in language, too, we see that elements may be discarded because they are no longer needed, and that other elements that happen to be lying around may be picked up and become highly beneficial to the system; he calls this evolutionary process “bricolage”. There is no external design or designer at work, the language user has to make do with the “junk” — the functional as well as the non-functional bits — presented to him, and has to make sense of it. Each language learner or “bricoleur” (Lass 1997: 310) will systematize the input presented to him in his own way, and will reduce it to some systematic structure in order to be able to handle language efficiently (note that language evolution potentially differs from biological evolution in this respect since much of language structure may indeed be adapted to speakers’ needs and thus some of it may be teleologically structured). Language itself thus continuously undergoes a process of evolution whereby accidental mutations occur which may or may not come to be used as part of the overall system, i.e. lexical elements floating around may come to be grammaticalized while already grammaticalized elements may become discarded or reduced to zero. The language that the child learns to understand and produce is complex, and the child will indeed develop a ‘design’ to cope with its complexity but in the evolution or acquisition of language itself, such complexity does not necessarily presuppose an external designer or design. According to this view, design is to a large extent ‘resultative’, i.e. design follows the facts. When we turn to another branch of linguistic science, ‘generative linguistics’, it seems that many linguists do presuppose some independent pre-existing design (or linguistic apparatus) that helps the learner to learn and process language quickly and competently. This design has been called the linguistic blueprint or universal grammar, better known as UG. Why do these linguists presuppose a design? Because, and here I feel there is an analogy with ‘creation science’, they believe that the language (I am substituting ‘language’ for ‘organism’ as Lass [1997: 316] does in his discussion of the similarities between language change and evolution) is too complex for the learner to handle without the help of such a design. A lot, of course, depends on the notion of complexity. The need for a design, for a blueprint, is closely tied up with the ‘poverty-of-the-stimulus concept’. I will return to this concept later at some length. Another aspect of the idea of a linguistic blueprint is that its design is fixed1
1.Even though many generative linguists now see language acquisition as a maturational process, they still accept that there is a fixed initial state which influences and constrains the
Grammar change versus language change
and universal, and indeed part of our genotype. This seems difficult to reconcile with the notion of evolution: language, in order to survive, must needs adapt itself to its environment, and therefore the system behind language, or the machinery that we use to produce it, must adapt itself too. Since the environment does not change drastically, language too will not change drastically, and the grammatical ‘system’ that learners possess (or which they deduce from the data; this is one of the questions that need to be answered) will therefore also have a common core over time. The question is, is this linguistic ‘design’ a developmental accident or a beautiful piece of machinery (a specific grammar module) with which the human brain has been fitted from the start? I think it is important to note in this connection that human beings are predisposed to find design in anything, even in chaos, so that the fact that linguists find ‘design’ (i.e. an innate grammar module) in language does not necessarily mean that it therefore has ontological status.2 Underlying this common grammatical core in individual languages, there is also a common genetic core of cognitive abilities that all human beings share. The latter, of course, is part of our genotype. Crews also writes in the article about creationists that “by denying that natural selection can generate specified complexity … [the creationists] saddle themselves with the task of determining when the divine designer infused that complexity into his creatures” (ibid., 27, italics added). In a similar way, it could be said that linguists who assume the presence of UG or a linguistic blueprint saddle themselves with the problem of when and how it came to be part of our genotype and where it is located inside the brain. Should we assume a separate grammar module, as they do, or is it more likely that the system that we use or deduce in language derives from various bits of machinery, which all developed accidentally and which are located in different parts of the brain. In other words, is the machinery we need to produce language linked to other brainfunctions that had developed long before we acquired language, before we were
maturational development (see also note 3 and Section 4.2.1). 2.Cf. Dawkins (1995: 96; I owe this reference to Lightfoot 1999: 254, but am using it to a rather different purpose), who writes “We humans have purpose on the brain. We find it hard to look at anything without wondering what it is ‘for’, what the motive is for it or the purpose behind it … Show us any object or process and it is hard for us to resist the ‘Why’ question — that ‘What is it for?’ question.” This of course is meant to show our ‘teleological urge’, and that is how Lightfoot uses it, but one could also turn the idea around: we infuse design into things in order to see its purpose, in order to understand it, not necessarily in order to reach some goal.
35
36
Olga Fischer
homo sapiens? In evolutionary terms, this seems to make more sense. But let us move away from evolutionary science at large and take a closer look at language evolution or language change, my topic for today. When I talk about ‘language’ change, I use ‘language’ as a shorthand for linguistic utterances, i.e. spatio-temporally bounded instantiations of ‘language’ as produced by speakers. My main question in connection with language change is: how can we best describe and understand the phenomenon of linguistic change? In order to answer this question, we must turn our attention to a number of further questions, such as, should we study change in language, i.e. linguistic utterances, on a physical historical level, or should we study changes in the grammar of the speaker that produces them? For synchronic linguistics, a similar question might be asked: should we study ‘performance’ (comparable to Saussure’s parole and Chomsky’s ‘E-language’) or should we concentrate on ‘competence’ (langue or ‘I-language’)? I will try dealing with the above questions by devoting some time to the following topics: (1) what is grammar, and more particularly (2) is grammar innate? During the discussion of these two topics, other questions will naturally arise, such as (3) what is the role ‘grammar’ plays in relation to both language acquisition and language change and (4) how are the latter two related to each other, (5) what is explanation in language change, and (6) what is our task as historical linguists?
3. What is grammar and how is grammar related to both language acquisition and change? Quite generally, grammar can be seen as a system that we have or develop in childhood with which we can process and understand linguistic utterances. At this point, however, already a number of fundamental questions arise. Is grammar innate and indeed part of our biological make-up, or is it developmental and the learning of it linked to more general cognitive abilities?3 The majority of
3.This is perhaps too simple a dichotomy, because more recently linguists who accept that grammar is innate also believe that the triggering of the innate rules and principles of grammar may take place in stages; see for instance Clark and Roberts (1993) and Lightfoot’s (1999) cue-based approach. In Lightfoot’s cue-based model children parse utterances, which results in their setting up “mental representations” or abstract structures which they scan against so-called “designated cues” in UG (the genotype). At first some of these representations constitute partial parses because children ignore the more complex parts of the input; only at a later stage do children reach their mature grammar or phenotype (Lightfoot
Grammar change versus language change
generative linguists writing about UG see it as “a biological entity, a finite mental organ” or “module”, “a linguistic genotype that [is] part of our genetic endowment” (cf. e.g. Lightfoot 1999: 52–53, 67). But not all linguists working within the generative framework agree. Wim Klooster, for instance, one of the earliest practising generative linguists in the Netherlands, wrote recently: “This [UG] of course is not a model of what takes place in our heads, but a model of what an idealized language user possesses in terms of implicit linguistic knowledge” (Klooster 2000: 8, my translation). And, I am sure, many more generative linguists are sitting on the fence as far as the physical reality of the model is concerned; indeed even Chomsky did not commit himself for a long time.4 There is one great methodological advantage, of course, in positing a UG that is part of our genetic constitution, our genotype. It makes the enterprise of determining what constitutes UG more scientific in that the contents of UG will ultimately have to be brought into line with the physiological workings of the
1999: 57–58, 148–151). Still, even in this model the grammar or genotype is innate and predetermined. 4.For instance, in Syntactic Structures (1957: 18), he uses the word “device”, or refers to “the theory of grammar”, and he talks about the adequacy of this device only in purely logical terms, no mention is made of a biological base. In Aspects (1965), Chomsky links linguistic theory with language learning, but his idea is that “empiricist theories about language acquisition” are not at all helpful: they “are refutable wherever they are clear, and … further empiricist speculations have been quite empty and uninformative” while “the rationalist approach exemplified by recent work in the theory of transformational grammar seems to have proved fairly productive, to be fully in accord with what is known about language [note that what is known concerns only competence O. F.], and to offer at least some hope of providing a hypothesis about the intrinsic structure of a language acquisition system that will meet the condition of adequacy-in-principle and do so in a sufficiently narrow and interesting way so that the question of feasibility, can, for the first time, be seriously raised” (pp. 54–55). In other words, there is a link, but the language acquisition device can only be productively studied from the top down so to speak, and it is quite clear that logical principles only play a role, i.e. reasoning from competence (which does not constitute empirical data!) is the only productive way forward. The model is therefore not “a psychological model of the way people construct and understand utterances” (Lyons 1970: 85). Only in later work (e.g. Chomsky 1981: 8), do we learn that UG is “an element of shared biological endowment”, but there is still a gap between UG and core grammar, the latter is said to be an “idealization” of “the reality of what a particular person may have inside his head”. More recently, judging from his reaction to John Searle in the New York Review (July 18, 2002, p. 64), Chomsky’s stance has become clearer; he writes: “The long-term goal has been, and remains, to show that contrary to appearances, human languages are basically cast to the same mold, that they are instantiations of the same fixed biological endowment, and that they ‘grow in the mind’ much like other biological systems, triggered and shaped by experience, but only in restricted ways”.
37
38
Olga Fischer
brain and its limitations. It makes the theory of UG falsifiable at another level and links it to other scientific domains, thus avoiding the danger of circularity.5 A model of UG in Klooster’s interpretation will be internally logical, and may indeed show up very interesting relations between parts of the grammar (or between linguistic utterances) that had been hidden before, but these findings are not scientifically or empirically falsifiable, only logically. And in fact this is true also for other linguists of the Chomskyan school who consider UG as part of our genotype. Yngve (1996:324) observes in this connection that “Chomsky’s methods are actually not those of science … but the familiar logical-domain methods of explication of an intuitive concept”.6 Chomsky, in fact, uses only two of the basic assumptions of science, i.e. the assumptions of regularity and rationality; he ignores two others, namely the ontological one (i.e. that there is a ‘real’ physical world out there that can be studied), and the causality assumption, i.e. “that
5.Yngve (1996: 117–118) notes in this respect that “[i]n the physical domain the range of phenomena legitimately taken into account as data is typically much broader than the range of phenomena we wish to study and explain. In grammar, however, the range of data and the range of theoretical predictions are typically identical. This is because of the autonomy of grammar. This widely held position has been put forward explicitly by Hjelmslev and Chomsky [notes omitted]. In the physical domain, on the other hand, the different sciences are not autonomous in this sense … they are all interrelated so that results in one science may well provide important data for a neighboring science” (see also pp. 170, 242). We have therefore a rather paradoxical situation: formalist (generative) grammar claims to be scientific and has a scientific aim (it wishes to find out what competence is, i.e. how the human brain produces language) but it doesn’t use the physical data that the natural sciences use; rather, it starts from assumptions which have not been scientifically established, and builds its arguments (its theory) on the basis of these assumptions, as one does in the logical domain. Yngve (1996: 31) has termed this problem “domain confusion”. Croft (1999: 92) also notes that “much of a student’s learning about these theories is learning how one can construct analyses of particular phenomena using the formal representational language of the theory”, i.e. backgrounding an empirical analysis. 6.Again this may be too simplistic. Croft (2001), referring to Lakatos (1970), notes that although it is true that a theory is “scientific” only when it has an empirical basis, it is also true that “there is no natural (i.e. psychological) demarcation between observational and theoretical propositions”. All empirical facts are in the end our observation of these facts; i.e. the facts themselves are coloured from the start because we describe them. Yngve (1996: 97) works from the assumption that there is a ‘real world’ out there and that it can be distinguished from our observations or description of that world. For him there are three separate “things” in scientific research: (1) the real, physical world, (2) our observations of the real world, (3) our theories about the real world. According to Croft, however, there is no easy division between (1) and (2). It is clear though that the data considered by generative linguists are very much at the end of the continuum between (1) and (2), whereas the data that Yngve uses (people interacting by means of sound waves, light waves etc.) is at the beginning of that continuum.
Grammar change versus language change
observed effects flow from immediate real-world causes” (Yngve 1996:101–102). In other words, the present generative enterprise, although said to be scientific, is in fact far from scientific. To sum up briefly at this point. I think that we can agree that grammar is some system situated in our brain. With it we can understand and produce linguistic utterances. We do not know, however, whether the grammar itself is a generative machine standing at the birth of those utterances, or whether it is the result of our brains’ analysis of utterances heard in context, or whether it is a combination of both. Because we have not yet been able to investigate this ‘machine’ physically, we cannot really know its status. Can we use it as an empirical tool to help us understand the processes of language acquisition and linguistic change, which are both in different ways related to it? That all depends on its status: if the grammar is innate, then of course its role becomes pivotal for the study of both acquisition and change. If it is not, then the study of the linguistic utterances becomes our primary concern. Let us therefore consider the arguments for innateness.
4. Is grammar innate? What pleads for or against innateness of some type of core grammar? The arguments usually put forward by the formalist linguistic school are of a biological and a logical nature; both are, as yet, based on indirect evidence. We will look at the biological facts first. 4.1 Innateness: Biological considerations Evidence from brain-damaged patients is used to show that there is a special part or module in our brains that deals with grammar. Pinker (1994:45–46) and others describe cases of people with Broca’s aphasia, whose grammatical processing is seriously impaired but whose lexical processing is left more or less undisturbed. It was at first thought that this impairment was strictly related to Broca’s area, but further research has shown (see Lieberman 1991: 85; Pinker 1994: 308–310; Slobin 1997[2001]: 281–282, Goldberg 2001: 41) that this is far too simple an idea. Lieberman, for instance, writes, The traditional view of Broca’s aphasia is that damage localized to Broca’s area will result in these [grammatical] deficits, whereas damage to any other part of the brain will not. This belief is reflected in popularized accounts of how the
39
40
Olga Fischer
human brain works, and in the supposition of many linguists that human beings have a specific, localized “language organ” (…). However, that supposition is erroneous… . The damage pattern that produces Broca’s aphasia interrupts the circuits between Broca’s area and other parts of the brain … . In fact subcortical damage that disrupts the connections from Broca’s area but leaves it intact can result in aphasia. (1991: 85, italics in original)
And Slobin (1997: 282) notes that cross-linguistic studies of aphasia have failed “to find any support for a ‘dual-lexicon hypothesis’, which postulates that openand closed-class items are mediated by different mechanisms and/or stored separately”. Rather, this research points to “processing factors alone as distinguishing the two classes”, and he adds that it is more likely that both classes of words are handled within a single lexicon. In other words, a localized grammar module has not (yet?) been found. The autonomy of grammar, too, has been brought seriously into doubt by these findings. A recent book on brain research by Goldberg (2001, esp. Chapter 5) indicates that the idea of modularity (and, by implication, of a grammar module) is finding less and less support. Goldberg’s research has produced a number of experimental results important in this connection. He found that different parts of the lexicon are stored in different places. Thus, the naming of animals activated a different area from the naming of tools (see below). a. ANIMALS
b. TOOLS
Distributed cortical representation of language. The cortical representation of word meanings denoting the objects are highly distributed. Various features of these representations are stored close to those sensory and motor areas that participated in acquiring information about the objects. (a) Area of increased blood flow when subjects name drawings of animals compared to naming tools. (b) Area of increased blood flow when subjects name drawings of tools compared to naming animals (from Goldberg 2001: 66).
Tool-naming “activated the left premotor regions in charge [also] of right hand movements” (Goldberg 2001: 66). From this experiment and others he concludes that “different aspects of word meaning are distributed in close
Grammar change versus language change
relationship to those aspects of physical reality which they denote”.7 This suggests that words may well be learned together with the pragmatic or realworld handling of them (this may constitute proof for the importance of the situational context in learning!) and that for that reason these words get stored in the place which is also in charge of the movements needed to execute whatever the content is of these words. Another important point that Goldberg’s research has shown is that hemispheric specialization is not unique to humans but also occurs in the great apes. This casts serious doubt on the idea that the difference between the right and left hemispheres can be based on language (Goldberg 2001: 41–42) as has been done traditionally. He finds that there is a more fundamental distinction between the functions of the two hemispheres and this has to do with learning: “The brains of higher animals, including humans, are endowed with a powerful capacity for learning. Unlike instinctive behavior, learning, by definition, is change. The organism encounters a situation for which it has no ready-made effective response” (44).8 Goldberg notes that “[a]t an early stage of every learning process, the organism is faced with ‘novelty’, and the end stage of the learning process can be thought of as ‘routinization’ or ‘familiarity’” (44). Goldberg believes that it is the role of learning and learned behaviour at the expense of instinctive behaviour that led to the difference between the two hemispheres. He concludes on the basis of experiments and evidence from brain-damaged patients that the right hemisphere deals with cognitive novelty, with the first stage of any kind of learning, while the left hemisphere deals with learning that has been routinized or automatized (52). These new findings question the idea of distinct and highly language-specific modules. It is more plausible, in other words, that language is dealt with in both hemispheres: the learning of language, by means of the development of a grammar or system, takes place in the right hemisphere, and, once it is learned, the processing of it is relegated to the left. This would explain the hitherto puzzling “lack of adverse effect of left-hemispheric damage in children” (who are still dealing with
7.He adds that this coupling also makes evolutionary sense: “The neural blueprint is both parsimonious and elegant” (Goldberg 2001: 67). It pleads of course against a separate module or place in the brain for language. 8.Goldberg continues: “The length of time, or the number of exposures required for the emergence of effective solutions, is vastly variable”. Note the importance of frequency and the contextual situation in learning, factors that are stressed again and again in the present article.
41
42
Olga Fischer
‘novelties’) as far as language is concerned, and the “particularly severe adverse effect of right-hemispheric damage” (43) on these children. Goldberg also shows that the search since the 1980s for strong dissociations,9 intended to identify the “mysterious modules”, is fallacious in that for “every case of strong dissociations there are hundreds of cases of weak dissociations, where many functions are impaired together, albeit to different degrees” (56). In other words, it is not good practice to single out the few strong cases and ignore the mass of others, as is regularly done in linguistic textbooks trying to convince us of the logical necessity of a grammar module. Before we leave the topic of biological innateness, I must add a rider. Although I have suggested that, on the basis of our present knowledge about the workings of the human brain, there is no necessity to accept a specific grammar module, it does not follow that no part of our language faculty is situated in our genes. On the contrary, it is well-known that the human larynx and vocal tract have evolved differently compared to that of the great apes; they have become adapted to produce sound efficiently, as is necessary for the precise articulation needed in human language. The same is true for other brain mechanisms such as memory, which also has become adapted to the production of language (cf. e.g. Yngve 1996: 90–91). At the same time, however, these organs have also retained their old functions to a greater or lesser extent, and more or less successfully (cf. Lieberman 1991: 14–16, 53–57).10 Such adaptations for language are now part of our genotype. So parts of the language faculty are certainly innate. What is being questioned is the very specific nature (and position) of a grammar module. 4.2 Innateness: Logical considerations The other arguments for innateness come from the so-called poverty-of-thestimulus concept. It should be noted before we address the content of this
9.Dissociation in chemistry involves the direct separation of compound substances into their primary elements, or into less complex compounds; in a similar way it came to be used in neuro-psychology to refer to the decomposition of mental activities into separate centres of consciousness. 10.Lieberman (1991: 54) notes that, because of the changes in the human larynx, its low position compared to that of apes, “we are now liable to choke when we eat or drink [because] solid objects or liquid can fall into the human larynx”. Similarly, “the reduced length of the modern human palate and mandible also crowds our teeth, presenting the possibility of infection from impacted wisdom teeth”.
Grammar change versus language change
notion that the various linguistic paradigms are much divided on this issue: whereas formalist (generative) linguists plead for the poverty of the ‘primary linguistic data’ (PLD), others emphasize the great richness of the data available to the child (e.g. Yngve 1996: 90). The latter point especially to the help given a child by means of conceptual structure (for detailed commentary on the povertyconcept and the nature of the child’s triggering experience, see McCawley 1989, Schlesinger 1989, and Snow & Tomasello 1989). I will briefly review the ‘poverty’ notion here, although I believe the arguments are by now wellknown.11 The concept refers to the child’s ability to learn the parent language within a relatively short period of time in spite of the following obstacles: 1. The PLD contains many incomplete, ill-formed utterances, and yet the child makes the correct choices, it does not overgeneralize. 2. No evidence is provided in the PLD for constructions that do not occur; i.e. the PLD is not rich enough to determine the limits to the generalizations that the child makes. 3. Very little linguistic correction is offered where it could be offered. 4. The child produces novel utterances that it has never heard before. 4.2.1 The poverty-of-the-stimulus concept The poverty-of-the-stimulus concept as an argument for innateness rests on the idea that the PLD is poor (incomplete/ill-formed) and that therefore there must be some linguistic ‘extra’ that helps the child to acquire language. The basis for the concept itself is weak in that there is no precise criterion for establishing ‘poverty’, poverty is merely assumed. No wonder there is such controversy among linguists on this issue, as I noted above. The reason why generative or formalist linguists assume that the PLD is incomplete is because they focus strictly on the production of sentences and on their formal structure. In other words, they fail to take into consideration all contextual variables, both linguistic and situational. To give an example: Lightfoot (1999: 50ff.) discusses how children acquire the correct use of pronouns. Sometimes, it is noted, pronouns refer back to a noun previously mentioned in the sentence (as in (1a–c)), but this is not the case in (1d), where the pronoun him may not refer to Jay,
11.Two clear and fairly recent expositions can be found in Foster 1990: 164 ff. and Lightfoot 1999: Chapter 3.
43
44
Olga Fischer
(1) a. b. c. d.
Jayi hurt hisi/j nose Jayi’s brother hurt himi/j Jayi said hei/j hurt Ray Jayi hurt himj
The question is, how do children acquire the right generalization, and, particularly, how do they acquire knowledge of the exception, i.e. how do they know that him in (1d) does not refer to Jay? Lightfoot’s solution is that “children learn from their environment [from the PLD] that he, his etc. are pronouns, and native principles dictate [in this case those of the so-called Binding Theory, see p. 58] where pronouns may not refer to a preceding noun” (p. 52, italics added). The weakness of this account is that not all the contextually relevant details of the utterances are taken into consideration but only their strictly structural properties (since the grammar is autonomous), as if the child learns to handle these pronouns in an otherwise complete void. It is relevant to ask, when children learn to recognise he, his etc. as pronouns “from their environment”, how they learn this. Do they simply mark them off and then store them up as lexical elements marked ‘pronoun’? That seems highly unlikely. Since pronouns have very little referential content, children can only ‘learn’ these items if they also learn at the same time how they are used in the situation, since extra-linguistic, real-world meaning will not help them here. Pronouns are not apples you can point at or bite into. What I mean is, a word or concept like ‘apple’ is relatively easy to learn for a child because the object in the real world will conjure up the word.12 This is not the case with pronouns, so when the child learns the appropriate use of pronouns, it can only do so in connection with some real-world referent, present in the context of the
12.According to Lightfoot (1999: 63), however, a child cannot “induce the meanings of even the simplest words”: “[c]hildren do not have sufficient evidence to induce the meaning of house, book, or city, or of more complex expressions, even if we grant them everything to advocates of Motherese or those who argue that it’s all data processing of huge corpora”. So children need innate knowledge even for this. I cannot quite see what this innate knowledge might consist of, nor do I see that children would not be able to learn what apple or house means when the objects are there for them to see, and when they occur repeatedly (I take it that learning starts with visible objects; see also Foster (1990: 159), who writes that “[t]he primary input to such a process [i.e. lexical learning] must be exposure to language in situations where the meaning can be deduced; and it is clear that children do get exposure to words under these conditions”). It is strange too that frequency seems to play no role in generative concepts of learning. A recent book by Bybee and Hopper (2001) shows how important a role frequency plays both in the development of language in children and in language change. Bybee has indeed emphasized the importance of frequency in much earlier work.
Grammar change versus language change
situation.13 Even though the real-world referent continually shifts, he soon learns that him refers to some other person or object present in the situation, it cannot refer to himself, and by analogy, the child will understand that him cannot refer to Jay in (1d) because if it did it would refer to the same object. (Note that nose in (1a) is indeed another object, so his can but need not refer to Jay.) The only mistake the child could make is that it interprets him reflexively, but other evidence from the PLD would soon put the child straight in this respect.14 It is interesting to observe that the mistakes that some children make with pronouns concern the pronouns I and you rather than him. These are sometimes turned around, i.e. children using I for you. This is not surprising because here the context is confusing; after all the parent who may be addressing the child calls the child you and himself I, so the child might then think the parent must be addressed as I, and he himself would be you. The UG scenario is problematic also in another respect: the Binding Principles do not account for all the ways in which (a) particular pronoun(s) can refer. Thus, the Principles do not help the child to decide to what noun the pronoun refers in (1a–c). It may refer to the previous noun in the clause, as indicated by the indices; on the other hand, it may also refer to a noun further away, or to an entity that has not been linguistically introduced at all. Yngve (1996: 80–81) gives some examples of the latter. Imagine, he writes, two people “standing in front of a bookshelf of plays by a noted playwright. One of them takes a play off the shelf and says: “This was her most successful. It had a long run.”
Then the other person says: “The second act is particularly riveting”.
Here the pronouns this, her and it refer to the book in hand, the playwright and the staging of the play respectively. These concepts have been “set up in the
13.Itkonen (1994: 46) writes: “It is a well-known fact that, in the beginning, children learn the meanings of only those words whose referents are present when they hear (or see) the corresponding word-forms. He adds in a note, “[b]ecause of its hostility towards associationist learning theory, Chomskyan psycholinguistics is incapable of accommodating this simple fact”. 14.This was indeed a possible interpretation of (1d) in Old English, as Lightfoot (1999: 75, note 4) notes. Anthony Warner (p.c) remarked in this connection that the use of himself in (1d), when the construction is reflexive, would soon make clear to the child that him cannot refer to Jay in Present-day English.
45
46
Olga Fischer
heads of the conversationalists through the visual channel” (p. 81) and through their general knowledge. They are not explicitly mentioned linguistically. In the same way, the reference of the pronouns in (1a–c) can be learned by the child from the situational context. Why then would the child not use the same context and similar learning strategies to understand that him does not refer to Jay in (1d)? There is no need for a complicated innate Binding Principle to forbid the interpretation of coreference between Jay and him in (1d). I would add, moreover, that in the real world it would be as important for the child to know the exact reference of the pronouns in (1a–c) as it is to know the constraints on the pronoun in (1d), but the Binding Principles would only help him with the latter, not with the former. This greatly restricts the use of the Binding Principles.15 The second aspect of the poverty-of-the-stimulus concept concerns the idea that we must explain how the child makes the correct choice in spite of negative (unavailable) evidence. Let me illustrate this point of view again first with an example. UG predicts that children will not produce sentences such as Who do you *wanna take a walk, where want to cannot be contracted because of an intervening trace (i.e. the trace left by who, which is the subject of the infinitive), whereas they may produce clauses such as What do you wanna eat where no trace intervenes (because what has been moved from a different underlying position), which makes contraction possible. Quite an ingenious experiment was set up by Crain (1991: 602–604), which shows that when children between the ages of 2.10 and 5.5 are ‘channelled’ into producing a sentence like the first one (Who do you want to take a walk?), about which they are unlikely to have any positive evidence (see also note 17), the unreduced want to form occurs 67 percent of the time, the reduced wanna form only 4 percent. The second sentence, however, shows a result of 59 percent contracted, 18 percent uncontracted. (In both cases the remaining percentages concern children who did not produce any sentence at all.) This result is then neatly
15.Cf. also Fox (1994: 2), who, like most cognitive linguists, refers to the fact that all language is “indexical”. By this she means that a linguistic expression is underspecified as far as its interpretation is concerned; it needs context in order to make “definite sense”: “the meaning is distributed across the participants, the linguistic expression and other facets of the socially organized environment”. In a different vein, but also important in this context, is the neuro-linguistic research carried out by Jos van Berkum and his associates. Van Berkum et al. (2003) show quite clearly (by means of ERPs — ‘event-related brain potentials’) that the semantic analysis of sentences is “rapidly sensitive to the wider discourse”, i.e. the meaning of every incoming word is immediately related to the hearer’s knowledge of the wider discourse. A holistic type of sentence processing (as envisaged by generative grammar) could not deal with the facts as presented in this research.
Grammar change versus language change
explained as in accordance with the UG rules about movement and traces. But aren’t other explanations equally possible? In his research on children’s language learning,16 Slobin (1985a: 1229) shows that children’s operating principles or strategies avoid “synthetic forms in favor of more analytic expressions” “for purposes of clarity”; i.e. a child would go for the longer, more analytic form (want to in this case) when he realises that a certain notion is complex. There is no doubt that the above sentence (Who do you want to take a walk?) is complex for a child, since it hardly ever (never?) occurs in everyday speech.17 It also seems probable, in the light of the new discoveries in Goldberg (2001), mentioned above, about the different functions of the two hemispheres, that the production of wanna is linked to automatized behaviour stored in the left hemisphere and that the more complex, novel sentence is interpreted by the right hemisphere. In other words, a different mechanism might be involved in the children’s production of the want to and wanna sentences. Moreover, it is important to realise that there may well have been acoustic clues in the experimental situation, which caused the child to differentiate between the who-clause (which requires want to) and the what-clause (which may have wanna). The experiment does not provide any evidence of this (presumably because any situational or not-structural clues are not relevant to the experimenter).18 It is quite clear, however, from child-language experiments
16.Slobin and his associates have developed a model for this, which they term the LMC — the ‘Language Making Capacity’ — a more general learning model than UG. 17.I checked the OED corpus on the structure in question and found no examples of it; the only examples concern structures where the wh-element is the object of the infinitive depending on want to, i.e. What do you want to/wanna eat?. In a large newspaper corpus, again the examples with initial, objective what were numerous, but only three examples were found with initial who. In two of them who was object (Who do you want to punch today (Daily Telegraph 2–9–1997), Who do you want to live with, John? (Daily Mail 26–8–1998). Just one example had who as subject: Who do you want to represent you? (Daily Telegraph 23–4–1997). Note, however, that this last clause is much easier to process than Crain’s example since who, quite clearly, cannot be the object of the infinitive, which already has an object. I am grateful to my colleague Tom van Brederode for providing me with these examples. 18.Indeed, two of the commentators on Crain’s article write: “As part of the task statement, the experimenter says to the child: “One of them gets to take a walk, one gets to take a nap.” It is likely that in each of the ‘gets to’ phrases in the task statement, both ‘gets’ and ‘to’ are clearly enunciated and distinguishable. Since the construction ‘gets to’ is exactly parallel to the expected ‘wants to’ (and not ‘wanna’) response from the child, the child need not previously acquire this construction, but may simply copy the intonation contour, inserting the new word.” With reference to another of Crain’s experiments they write that the children’s response is “likely to be influenced by intonational and pausal differences” (Dodd
47
48
Olga Fischer
that the earliest and most important operating principles handled by children are based on acoustic salience (cf. Peters 1985: 1033–1040), and that indeed children produce (or imitate) melodies or intonational units before they produce words (see also Foster 1990: 46). A related issue is that the rules of UG are unnecessarily specific and complex: it has to allow for the generation of all logically possible sentences, even those that can be shown not to occur in reality (e.g. by means of corpus evidence). Again we come up against the problem that generative linguistics does not study physical data in the real world but abstract linguistic data in our competence, where competence is often associated with the rules and procedures of written language. Within generative theory the characteristics of the spoken language are relegated to the performance level, and to my knowledge, it has never been seriously considered within this framework that the structure of spoken and written language may be essentially different in a number of ways. Since the young child learns language only through the oral/aural channels, it is the system of spoken language we should study. It seems to me that linguists’ intuitions about language are often pre-programmed or modelled by their (grammatical) schooling, and thus reflect the logical notions acquired there rather than the ‘real’ competence of speakers and hearers (whatever that may be).19 This last point has an important historical corollary. Having studied the history of English for a number of years now, I am convinced that certain changes that have taken place in English have very little to do with language (or grammar) change in the narrow sense of the words, but with changes in English as it developed into a written medium. It is not hard to suggest ‘changes’ that may be associated with this oral/written parameter. It would be an interesting topic of research for historical linguists. A large amount of research in this area has been conducted by German scholars, as shown by the references in Feilke et al.
& Fogel 1991: 617). Similarly, a third commentator notes that structural constraints and principles are not the only things a child attends to, rather it has “recourse to a ‘confluence of cues,’ which are perceptual and semantic as well as structure-dependent, and are sensitive to the relative productivity of devices within a given target language” (Berman 1991: 613). The impact of the environment in a broad sense is also mentioned in the comments by Dodd & Fogel 1991: 618, Harris 1991: 623, McCawley 1991: 626 and Slobin 1991: 633, and the importance of frequency is emphasized by Schlesinger 1991: 633. 19.I mention an interesting anecdote in this connection. In a class of first-year linguistics students, we talked about when the students learned their native language, and one of them said quite seriously that she learned it at school. Somehow, knowledge of language is linked in our minds to explicit schooling.
Grammar change versus language change
(2001). This study shows that languages which have developed a written standard undergo a “Verschriftlichung der Sprache” (Feilke et al. 2001: 18–24), i.e. the spoken standard is influenced by the forms of the written standard: “Die Schrift … wird zum Motiv einer weitergehenden Sprachanalyse, sie wird zum Motiv grammatischer Analyse und der professionellen Grammatikschreibung selbst” (18). That is, a written standard influences the language both theoretically (in the way we interpret grammar as linguists) and practically (in the way in which we express ourselves), and the influence of Verschriftlichung is especially strong on syntax and the organization of text (ibid. 19). According to the authors, the changes that occur have in common that they produce text which is maximally decontextualized, i.e. the text is made maximally explicit as to context (ibid. 20, and see also Yngve 1996: 300ff.). Feilke et al. specifically mention the development of complex prepositional constructions, constructions expressing purpose, new conjunctions and complex clauses (20).20 Concerning the third point on the list in Section 4.2, the absence of negative feedback from parents or caretakers, we find many linguists, especially psycholinguists, arguing (in commentaries on articles advocating the poverty-of-thestimulus notion) that children may well have evidence of negative data, for instance by means of frequency of positive data (see also footnotes 12, 23), or that corrections made by parents may have effect (see A. Grimshaw 1989: 340; J. Grimshaw & Pinker 1989: 341–342; Schlesinger 1989: 355–357; Berman 1991: 613; McCawley 1991: 627–628; Sokolov & Snow 1991: 635). Wisely, Sokolov and Snow round off their presentation, in which they show that negative evidence is available to children, as follows: “Rather than argue about its presence, we should be examining its role in development” (635). The fourth point put forward in support of the poverty-of-the-stimulus notion is the fact that children produce novel utterances, phrases or clauses they have never heard before. It seems to me that this aspect has been somewhat exaggerated. The first thing one notices with small children is their canny intuition (or should we call it ‘uncanny’?) for using or imitating phrases they have heard others use, in the right place and at the right time. The novelty, as far as I can see, consists mainly in the use of already encountered and frequent constructions but with different lexical slot-filling, or in the use of the same construction but with a dependent clause stuck onto it or embedded in it.
20.In Fischer (forthcoming), I show how this Verschriftlichung has been of influence on the development of ‘true’ subject clauses in English, i.e. in initial position, which appeared only in the Middle English period.
49
50
Olga Fischer
Lightfoot (1999: 60) calls the latter “iterative devices”, devices which “in principle” may cause “any given sentence [to] be of indefinite length”. Although indeed children could not produce such utterances purely by imitation, very simple operating principles working on the input, such as analogy — a simple awareness of ‘same’ and ‘different’ — could help children to produce such novelties. Holyoak and Thagard (1995) have shown that analogy is one of the prime forces in the learning process of a child, something that is also present in a more primitive form in apes, and in lower mammals too. Analogy is a general principle, an ability that children need, not just to learn language but also as an evolutionary strategy to survive in an ever-changing environment. Itkonen (1994: 45) describes this as follows: The properties of co-occurrence and succession, and in particular the causal properties, of things and events are learned on the basis of analogy. Consider the knowledge that all ravens are black, that the day is always followed by the night, and that (every instance of) fire is hot. This knowledge is acquired in two steps. First we infer from the present case to the next one: raven-1/black-1 = raven-2/X, and X = black-2 … fire-1/hot-1 = fire-2/X, and X = hot-2 Second we perform an analogical (or ‘inductive’) generalization: All ravens observed so far are (have been) black Æ All ravens are black.
If we did not learn soon that fires are hot, that tigers are dangerous, or rather, in our society, that cars move fast, we might not survive for long. It is highly likely that the principle of analogy is innate and that it forms the basis of all types of learning. Most of the operating principles or strategies described by Peters (1985) and Slobin (1985a), which they consider part of the Language Making Capacity (see note 16), are principles based on analogy, on recognising what is same, and, therefore, what is not-same, and drawing conclusions from that.21 Of course, these same/different operations are performed on linguistic utterances in context, on the form as well as the situated meaning of the utterance, and frequency, hardly surprisingly, plays a very important role here (see Slobin 1985a:1165–66). Holyoak and Thagard show that children can do these simple analogies and that they become more and more
21.Note that exactly the same principles operate in science. Yngve (1996: 134ff.), for instance, shows in linguistic science that one moves from very specific properties obtaining in any communicative situation, to more general properties, by observing underlying differences and similarities.
Grammar change versus language change
expert at building up analogies upon analogies (thus making them more complex) in the course of their development. Indeed, Peters and Slobin recognise different stages and different types of operating principles (OPs), beginning with simple ones dealing with extraction and segmentation of the acoustic ‘noise’ children first receive,22 later followed by OPs recognising internal segments (bound morphemes), OPs to distinguish ‘frames’ (syntactic patterns) from ‘slots’ (content words), and OPs monitoring feedback. Such OPs (but of a more flexible type) fit in well with later research on neural networks, which also presuppose simple operations produced by billions of cells, interconnected by pathways, their interaction leading to potentially very complex results. In sum, I think it is important to stress that children’s structural novelties are not simply triggered by the grammatical device, but by the whole environment in which they learn and act, and by what they have already learned and stored at each particular stage. Lightfoot (1999: 149) sees the PLD or the linguistic input children receive as a store of physical, incomplete and often rather degenerate data, from which the child collects cues by means of some sort of partial parsing. These cues are called ‘designated cues’ because they are specified by UG, the idea being that children match their cues against UG. These cues, then, are said to represent the child’s ‘triggering experience’. In other words, the PLD is not directly the trigger in the child’s learning process, but abstract structures or cues derived from it. The developing grammar itself will still be close to the genotype (i.e. UG), but further input and the selection of more specific cues will eventually lead to a mature grammar or phenotype. Lightfoot thus acknowledges the role played by the PLD much more than most generative linguists, who concentrate on competence and not performance. However, he simultaneously reduces the impact of the PLD, by stating that the child’s triggering experience is not the PLD directly, but an abstraction of it, the so-called cues. This rather backstages the role of the PLD, and it also ignores environmental factors, i.e. what Fox (1994) and other cognitive linguists have called the context of utterance: the acoustic, pragmatic, visual etc. signals. The
22.An example of an early extraction-OP is: “EX:EXTRACT. Extract whatever salient chunks of speech you can”; in order to deal with the results two further OPs are necessary at this stage: “EX:COMPARE. Determine whether a newly extracted chunk of speech seems to be the same or different from anything you have already stored’”, and “EX:STORE. If it is different then store it separately; if it is the same, take note of this sameness but do not store it separately” (Peters 1985: 1033). At this early stage, contextual meaning (the child pays attention to chunks that have a readily identifiable meaning, p. 1034) and phonological properties (stress, intonation, rhythm, silence, pp. 1034–1038) are most important.
51
52
Olga Fischer
role of the PLD is thus quite different for Dan Slobin and his colleagues (Slobin 1985a, b, 1997), who have studied children’s language acquisition in great detail and in many different languages. Slobin et al. see the PLD not as a trigger, and certainly not as an indirect abstraction of it, but they see it as a “nutrient” (Slobin 1997[2001]: 310), which gives the PLD a much more primary role than in Lightfoot’s cue-based grammar. Slobin’s LMC model also derives abstract structures from the input by means of, at first simple and general, later more elaborate and specific, Operating Principles, but these are not matched against a pre-existing UG, but rather against more and more input. In this way, novel utterances are perhaps less of a novelty, in that the novelty is closely linked with the world that the child lives in and learns in. It is interesting in this context to record Hopper’s reaction to the ‘novelty’ argument. Hopper (1987: 145) holds the view that novelty has been overstressed for ideological reasons: “novelty is a prized virtue in our society altogether [note omitted], and we have many ways … of censuring perceived repetitions of others’ behavior and an enormous vocabulary dealing with repetition (copying, imitation)”, and yet, he continues, the use of formulaic expressions is the rule, novelty the exception.23 He believes indeed that “[t]here is no room — no need — for mediation by mental structures … speaking is more similar to remembering procedures and things than it is to following rules … grammar is what results when formulas are re-arranged, or dismantled and re-assembled, in different ways”.24 I am not altogether happy with Hopper’s idea of a grammar that is “always in a process but never arriving, and therefore emergent” (141). I feel this may be too strong. It seems to me that the Operating Principles that children apply when they analyse the input do lead to the formation of some kind of abstract system, which may serve as a basis for their later language processing. Evidence
23.This is also the message of the articles in Bybee and Hopper 2001, which study the effect of frequency on pattern formation. 24.For the idea that speakers use procedures rather than rules, see also Fox 1994: 9–10, Yngve 1996: 171ff., 276ff., and Slobin (1985) and Peters (1985) mentioned above. Again it is clear that analogy, or formulas re-arranged, play an important role here rather than ‘rules’. Hampe and Schönefeld (2003) show how new argument structures with verbs of the type: … she wrestled a screaming Dudley into his high chair cannot be the result of the application of a grammatical (or lexical) rule but are formed on analogy of existing argument structures of other verbs, such as in this case force. Yngve (1996: 330, note 91) also writes in this connection that lexical entries are not static but that there is a dynamic interchange between the procedures (the processing system), the lexicon and the conditions of the situational context.
Grammar change versus language change
for this is the errors children make, which suggests that there are rules underlying them. Of course many of the errors disappear on further exposure, but not all, and these types of ‘errors’ or new rules in their system may well lead to or help to explain certain changes that take place in language historically.25 It is also remarkable to see how conservative many speakers are who learn a second language after childhood. The system that they have built up for their first language often oozes through the utterances they produce in their second language.26 It takes a long time for these errors to disappear, and with some learners they do not disappear at all. It seems to me that it is especially the word order of the first-learned language which may influence production in a newly learned language. The importance of word order is indeed emphasized by linguists who believe it is procedures rather than rules that we use in language processing (see also below, Section 4.2.2). Slobin (1985a: 1192), for instance, writes: Everything we know about child language points to the conclusion that children keep track of sequential orders of elements: order tends to be preserved in imitation, amalgams and formulas preserve order. … Indeed order is so essential to human language that an organism unequipped to notice and store sequential information could hardly acquire such systems.
Yngve (1996: 43–44, 48–50) shows that adult language users, too, deal with incoming language units incrementally, in chronological order, not holistically, because the short term memory does not have the capacity to store all the elements necessary in the total product. His ‘depth hypothesis’, already referred
25.I do think, unlike Aitchison (2001: 207ff.), that there is profit in comparing these natural tendencies in children to what happens in language change. Aitchison believes that the resemblances are only superficial, but yet the examples that she gives show that there is a resemblance. She notes, for instance, that both children and adults shorten words but that the methods differ because the position of the lost syllable is not the same. However, it is clear that in both shortening methods it is the salient syllable that is preserved, position of the lost syllable may be different but this is secondary. Similarly, the tendency of children to replace fricatives by stops (according to Aitchison, rare in language change) is also a very clear tendency in change in pidgins and creoles. In Tok Pisin (discussed further on in her book) we find many instances of this: fellow > pela, leaf > lip, brother > brata etc. 26.Cf. Craats-Oosterwold (2000), who investigated the influence of the grammar of the first language on adult, second language learners. Her book bears the title Conservation in the Acquisition of Possessive Constructions: A Study of Second Language Acquisition by Turkish and Moroccan Learners of Dutch, and indeed it is shown that the Turkish and Moroccan secondlanguage learners start their learning of Dutch on the basis of their native grammar, showing that this grammar must be present somewhere as a system.
53
54
Olga Fischer
to above, is closely linked to this. It is perhaps not surprising, therefore, that many of the changes that have been recorded by historical linguists are caused by or related to changes in word order. Almost all of the syntactic changes in English that we discuss in the book I mentioned at the very beginning (Fischer et al. 2000) are linked to word order. I think this is another area where historical linguists and linguists studying language acquisition could cooperate. 4.2.2 Some further points with respect to innateness There are three more points that have to be addressed in connection with the innateness of grammar. One is methodological, a second concerns the ‘formfunction divide’, and a third has to do with product vs. process orientation. With respect to methodology, I agree with Derwing (1977: 79–80), who writes that we should be questioning the value of any linguistic theory that attempts to invoke “innateness” as an explanatory vehicle. For to maintain that some cognitive or behavioral skill is “innate” does not provide any positive insight into either its nature or development, but is rather tantamount to an admission of a failure to explain it. “Innateness” is a purely negative notion; it means that something has not been learned, hence that it can not be explained in terms of any known principles of learning. Explanation does not consist in substituting one unknown for another, but rather in accounting for what puzzles in terms of some general principle which is known and which is understood. And how, in any event, does one ever propose to demonstrate that some particular aspect of human language has, in fact, not been learned? (…) The search for “innate” principles, therefore, strikes me as more of a weakness than an attractive alternative to looking for answers in terms of psychological or physiological capacities that human beings have been shown to possess.
For historical linguists a very similar plea was made by Bybee (1988: 357), who writes, “complete explanations must specify a causal mechanism: thus we cannot explain change with reference to preferred types [she is discussing the use of ‘explanatory’ universal typological principles such as Venneman’s ‘Natural Serialization Principle’], but we must explain common types by referring to the factors that create them”. In other words, one should look further afield — or better still, outside one’s field — in order to explain phenomena. Such cross-domain investigations strengthen one’s own field, and create links with other scientific fields (cf. Yngve 1996: 117). Further on in his article, Derwing (81) adds, “my perspective is that since the ‘language system’
Grammar change versus language change
exists only in the minds of language learners, we must therefore explore these minds in order to ascertain what that system is actually like”. Yngve (1996) also points to the importance of investigating what is physically present, as already noted above: in order to create a scientifically worthwhile linguistics, we must start only with the physical data, with utterances (“sound-waves”), speakers and the situational context; only in this way can we relate linguistics to other sciences and find out how language is really processed. The ‘form-function divide’ is intimately bound up with the ‘product vs. process’ aspect. Formalist linguists see linguistic utterances very much as a product not as a process. The rules of their grammar are designed to represent the formal relations among the constituents of the product, and they do not take into account the psychological states or operations that a language user knows or executes when he processes a linguistic utterance (cf. Derwing 1977: 83).27 In other words, a sentence can be shaped and reshaped in the process of an utterance, and it is not only the speaker who influences the production of a sentence, but also the reaction of the hearer (or overhearer). For a formal, product-oriented linguist it is as if, once the speaker has chosen what to talk about, the structure follows automatically from his grammar (cf. Goodwin 1979). Bolinger (1978), however, stresses that structure itself is meaningful, it is part of a speaker’s choice and not an automatic consequence of his choice. This is where the form-function divide comes in. There is no one-to-one relationship between form and function, or, as Slobin (1997: 280) writes, there is no clear dividing line between “content words” and “functors.” Rather, there is a continuum with clearly lexical items on one end (nouns like computer, couch, zebra, verbs like tackle, broil, sneeze) and grammatical inflections on the other (such as English PROGRESSIVE -ing, Turkish ACCUSATIVE -I, Warlpiri ERGATIVE -ngku). In between, there are lexical items that play more or less specialized roles, sometimes on their way to becoming grammatical morphemes over time. What, then, is a grammatical morpheme? It depends on the purposes of the analysis. In any event, it would be difficult to preprogram the child with an adequate definition.28
27.Fox (1994: 23–24) writes, referring to neuro-psychological studies, that “the appearance of structure rather than process has to do with one’s distance from the phenomenon”; distancing oneself from performance (i.e. by concentrating on competence), therefore, leads to differences in observation of what is intrinsically the same thing. 28.There have been earlier proposals that the lexicon and syntax are closely related or form a continuum, cf. e.g. Langacker 1987.
55
56
Olga Fischer
Children, in other words, will have to discover, while developing their linguistic system, which of the lexical elements are used grammatically, which turn out to be fully lexical, which fully grammatical, and which float somewhere in between; i.e. they have to learn which notions are grammaticizable in their language. This presents a developmental (socio-historical) approach to the structuring of grammaticizable notions, and not a ready-made individual mind source in which content words and functors are pre-wired. The approach is clearly related to the notion of grammaticalization as used in historical linguistics, even though the processes at work are not necessarily or even likely to be the same. But shared features can be discovered: what is important is that both processes, change and acquisition, take place in a communicative context, and, presumably, distribution and frequency of exposure play a crucial role in both. Also in both there will be items that function at the content as well as the functor level, and it would be interesting to investigate whether the ways of identifying or handling them are the same in language acquisition as in adult every-day usage, and whether any relation to tendencies in historical grammaticalization processes can be established. Another interesting aspect is the question of whether the small lexical core of source-concepts from which grammaticalized elements develop is also a central or prototypical core group in child language and is developed early in the process of acquisition.29 I think we can learn more about grammaticalization by studying both processes in combination. Slobin (1997: 282) adds that the developmental notion of grammaticalization also removes “any basis for a neurological definition of the closed class as a linguistic subsystem”, and he refers in this connection to cross-linguistic studies which have failed to find any support for a dual-lexicon hypothesis, for open- and closed-class items being mediated by different mechanisms and/or stored separately (see also Section 4.1 above). It must be clear that one’s point of view in the innateness debate profoundly influences the way in which language is studied. Both cognitive and formalist linguists seem to agree that there is a link between language acquisition and
29.Some evidence for this is given in Hallan (2001), who found that phrasal verbs are acquired before prepositional phrases, presumably because the phrasal particle has stress and is therefore much more salient than a preposition. It is to be noted that, historically, prepositions derive often from locative adverbs, i.e. particles. This may also show that out of a set of homonyms derived from the same source, the most concrete, i.e. the fully lexical homonym, may be learned first and the other may be learned from it, following as it were the historical path of grammaticalization from lexical to grammatical. Another study looking at language acquisition and grammaticalization is Wong 2004.
Grammar change versus language change
language change, and that knowledge about how the one takes place may help us to understand more about the other. The big difference, however, is what exactly the object of research is. Is it the physical language data, the physical context and the physical mind of the speaker, or is it the fictional (?) language system and the idealized, and therefore fictional, competence of the speaker?
5. Grammar change/language change and the historical linguist: Some conclusions The historical linguist has only one firm knowledge base and that is the historical documents. Unlike synchronic linguists, he cannot make use of the intuitions of native speakers, nor has he access to spoken material and visual aids, such as gestures, direction of gaze etc. His prime concern is an accurate description of the data in the documents, in their context, which he investigates in order to understand the regularities underlying the data and the changes that take place. In doing this he may (or rather, must) make use of insights provided by other disciplines, such as language acquisition studies, sociolinguistics, generative grammar, cognitive grammar, discourse analysis, optimality theory etc., and also insights drawn from synchronic variation and typological comparison, but these insights are not primary data to be used in the same way as the written documents themselves. If the historical linguist wishes to contribute to our knowledge of how language works, if he wishes to deepen our knowledge about the system that language users have or develop — the ultimate aim, I take it, of most linguistic subdisciplines — then he should do this from within his own subdiscipline, i.e. use the empirical data that his subdiscipline provides. For me, this means that he must concentrate on physical data, on how language changes and not on how grammar changes. Grammar, at this stage, is a theoretical construct, not an empirically proven fact.30 I therefore cannot go along with linguists who see UG as a biological fact. When Lightfoot (1999: 10), for instance, writes, “For us, a grammar is part of human biology”, the phrase “for us”
30.Not even an empirical fact that can be shown to exist (in the forms or models proposed) by experimentation. Although it is true that in the natural sciences — ever since Francis Bacon and Robert Boyle (cf. Leezenberg & de Vries 2001: 39–42) — empirical facts comprise not just natural phenomena but also non-visible phenomena that can be shown to exist via repeatable and verifiable experiments), psychological or any other tests have not yet shown that the transformations, rules, principles, constraints etc., which are said to exist in formal models of grammar, are actually used in the processing of sentences.
57
58
Olga Fischer
is crucial because it indicates that the idea that UG is part of human biology is a generative assumption, not (yet) a fact. I have tried to show in Section 4 why I believe that this is not a sound assumption. For Lightfoot, indeed, grammar change is what the historical linguist should investigate because grammar is ‘real’, whereas “language … is an epiphenomenon” (ibid. 74).31 He believes that “[h]istorical linguists who limit themselves to phenomena of language change are prone to offer pseudo-explanations which lose touch with reality and which create mysteries where there is nothing mysterious” (ibid. 212). He means, presumably, that such linguists occupy themselves too much with unimportant details, leading to uninteresting explanations, thus getting too far away from the ‘core business’, that is, establishing the language ‘system’. I do not think, however, that a short-cut to the core via UG is possible. As long as UG is a construct, not a fact, such a short-cut may indeed lead to pseudoexplanations. This does not mean that the historical linguist must consider all historical details. He must consider the details scientifically, i.e. observe what is ‘same’ and what is ‘different’ in comparable contextual situations (cf. note 21). Sameness may tell us something about robust patterns in language, while differences may show what patterns are less basic. By looking at the contexts in which the utterances occur, we may learn how the differences come about, and how peripheral patterns are affected. Thus, in my view, the study of physical, written data should provide us with hints from which a theory of how language works should be built up, and this theory should not a priori coincide with any theory set up by other subdisciplines, i.e. they should work independently. Other areas of linguistics may also provide hints on another (more abstract) level, which may feed our imagination as researchers, but these hints should be tested on the historical facts. One of the great achievements of the generative linguistic enterprise has been to show the importance of the study of language acquisition, and to create a link between language acquisition and language change. It has given historical linguistics a new lease of life, and a definite role to play in the search for the (a?) system underlying (a) language. It is now deemed entirely appropriate that
31.In this debate about what is ‘real’, the school of ‘Emergent Grammar’ may be said to represent the other ‘extreme’. According to Hopper (1987), grammar is “epiphenomenal” (142), “an effect”, rather than a “cause”, “always emergent and never present” (148). Lightfoot (1999: 74), on the other hand, sees language as an ‘effect’, and grammar as a ‘cause’ (language is “a derivative concept, the output of people’s grammar”). I would say that grammar, when it develops, is an effect, but once developed, may also act as a cause.
Grammar change versus language change
historical linguists take cognizance of what happens in the subdiscipline of language acquisition. However, I do not think that the generative approach to language acquisition is the most useful theory for historical linguists to use as a basis for their research because it works from competence to performance and not the other way around. The type of operating principles suggested by Slobin and his school seem to me to be much more appropriate and promising since they do not presuppose a ready-made grammar module; rather they allow for developmental changes taking place in children: they take into account the increase in processing capacity, and cognitive developments which influence the way in which children perceive the world around them. Instead of on innate principles, they rely on developmental ‘bootstrapping’. This means that children exploit existing resources or capabilities to raise themselves to a new situation or state; they ‘pull themselves up’ from what they already know, and by doing so, get to know more and thus create more resources by which to pull themselves up even further. Thus, it seems a good idea to investigate whether the operating principles that play an important role in acquisition are also to be found in language change, and whether structures acquired early in acquisition are more stable in periods of change. In that way, we may be able to develop an understanding of a language system which is not predetermined like UG, but which is a direct result of the analysis of the PLD, and consequently likely to be much more language-specific. At the same time, an understanding of such a language processing system might bring us closer to what the brain actually does, and would create a profitable link between linguistics and the other sciences concerned with the working of the brain, since linguistics too would then be based on physical data. It would be part of the physical domain rather than the logical domain. Taking this approach seriously, it would be of interest for historical linguists to study more carefully the types of errors that children make, and compare these to changes that take place in language too. Especially important, according to Peters (1985: 1055ff.), are mistakes in segmentation. We could study, for instance, how children’s developmental segmentation capacity may be aided by typological ‘bootstrapping’ (on the basis of basic word order structures that they have already acquired for their language), and by more general iconic and acoustic bootstrapping (cf. Slobin 1997[2001]: 314ff.). The same segmentation heuristics could then be tried on language change at the historical level. This approach may help to tackle some of the problems noted by Lass (1997: 335ff.). Lass writes that “[n]either the linguist’s nor the speaker’s abductions [or errors] can be guaranteed to be ‘uniform’ in any useful way” (336), i.e. there is very little we can say about language change from the point of view of learning,
59
60
Olga Fischer
because all abductions are in the end personal. If, however, we can establish deeply entrenched heuristic principles, i.e. ones that we develop for language early in life, then it would be interesting to find out whether those same heuristic principles keep playing a primary role in later changes too, and in language use in general. In other words, it might make change less random than Lass believes it to be. Finally I would like to mention some other areas of investigation, in which historical linguistics and language acquisition could cooperate. One is centered on the issues Slobin (1997 [2001]) has written on. It concerns the question of how children acquire their grammar, which resembles the way “in which languages acquire and modify grammatical elements and constructions” (Slobin 1997[2001]: 298), known as grammaticalization. Similar questions or problems play a role in both areas of research, such as the way in which conceptual material is packaged linguistically (ibid. 287 ff.), how we learn about the package, and how this learning may also influence later developments in the language at large. Another aspect that needs more study is the role played by frequency in both learning and change, and connected with this the role played by automatized behaviour. Here again, the results obtained in the two subdisciplines could be compared and prove useful to one another. Finally, I mentioned in my search for areas which need more investigation by historical linguists the way in which the development of a written standard may have influenced the language historically. Here too there may be profitable links with language acquisition and literacy studies, in that the early oral forms are often the only forms used by children, showing us more about the relation between language change, language learning and ease of processing. All in all, I believe it is too early in the day to ascribe grammar, if indeed we can agree that “divine intervention” is not an option (see the quotation from Jackendoff which heads this paper), to a genotype. I thus disagree with Jackendoff that “the only other way” is biology. Similarly, there is not enough evidence that this ‘innate’ grammar, if it exists, would be of the formal, autonomous type as suggested by generative linguists. Therefore, to explain linguistic change first and foremost in terms of such an innate grammar is not a good idea. Moreover, there is still very little agreement about what exactly the rules, principles, constraints and parameters of this innate grammar are. I would advise historical linguists to take the data and their context seriously first, and to work from there using theoretical insights but not taking them as ‘real’. Ascribing such a formal grammar to biology remains an assumption, which may become a fact once we know more about how the mind actually works.
Grammar change versus language change
References Aitchison, Jean. 2001, 3rd ed. Language Change: Progress or Decay. Cambridge: Cambridge University Press. Berkum, Jos J. A. van, Pienie Zwitserlood, Colin M. Brown & Peter Hagoort. 2003. “When and How do Listeners Relate a Sentence to the Wider Discourse? Evidence from the N400 Effect”. Cognitive Brain Research 17.701–718. Berman, Ruth. 1991. “In Defense of Development”. Behavioral and Brain Sciences 14.612–613. Bybee, Joan. 1988. “The Diachronic Dimension in Explanation”. Explaining Language Universals ed. by John Hawkins, 350–379. Oxford: Blackwell. ——— & Paul Hopper, eds. 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam: John Benjamins. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. ———. 1965. Aspects of the Theory of Syntax. Cambridge Mass.: MIT Press. ———. 1981. Lectures on Government and Binding. Dordrecht: Foris. Clark, Robin & Ian Roberts. 1993. “A Computational Model of Language Learning and Language Change”. Linguistic Inquiry 24.299–345. Craats-Oosterwold, Catharina D. van de. 2000. Conservation in the Acquisition of Possessive Constructions: A study of second language acquisition by Turkish and Moroccan learners of Dutch. PhD Thesis, Katholieke Universiteit Brabant. Crain, Stephen. 1991. “Language Acquisition in the Absence of Experience”. Behavioral and Brain Sciences 14.597–612. Croft, William. 1999. “What (Some) Functionalists Can Learn from (Some) Formalists”. Functionalism and Formalism in Linguistics. Vol. I: General Papers ed. by M. Darnell et al., 85–108. Amsterdam: John Benjamins. ———. 2000. Explaining Language Change. An Evolutionary Approach. Harlow: Longman, Pearson Education. ———. 2001. “Some Lessons Drawn from Formalism for Functionalism”. Lecture given at the opening of the Universiteit Leiden Centre for Linguistics, September 20, 2001. Ms. Dawkins, Richard. 1995. River out of Eden: A Darwinian View of Life. New York: Basic Books. Derwing, Bruce, L. 1977. “Is the Child Really a ‘Little Linguist’?”. Language Learning and Thought ed. by John Macnamara, 79–84. New York: Academic Press. Dodd, David & Alan Fogel. 1991. “Nonnativist Alternatives to the Negative Evidence Hypothesis”. Behavioral and Brain Sciences 14.617–618. Feilke, Helmuth, Klaus-Peter Kappest & Clemens Knobloch. 2001. Grammatikalisierung, Spracherwerb und Schriftlichkeit. Tübingen: Niemeyer. Fischer, Olga. Forthcoming. “‘Langue’, ‘Parole’ and the Historical Linguist”. To appear in the proceedings of SELIM ed. by Alicia Rodríguez Álvarez ———, Ans van Kemenade, Willem Koopman & Wim van der Wurff. 2000. The Syntax of Early English. (Cambridge Syntax Guides). Cambridge: Cambridge University Press. Foster, Susan H. 1990. The Communicative Competence of Young Children. London: Longman. Fox, Barbara F. 1994. “Contextualization, Indexicality, and the Distributed Nature of Grammar”. Language Sciences 16.1–37.
61
62
Olga Fischer
Goldberg, Elkhonon. 2001. The Executive Brain. Frontal Lobes and the Civilized Mind. Oxford: Oxford University Press. Grimshaw, Allen D. 1989. “Infinitely Nested Chinese ‘Black Boxes’: Linguists and the Search for Universal (Innate) Grammar”. Behavioral and Brain Sciences 12.339–340. Grimshaw, Jane & Steven Pinker. 1989. “Positive and Negative Evidence in Language Acquisition”. Behavioral and Brain Sciences 12.341–342. Hallan, Naomi. 2001. “Paths to Prepositions? A Corpus-Based Study of the Acquisition of a Lexico-Grammatical Category”. Bybee & Hopper 2001.91–120. Hampe, Beate & Doris Schönefeld. 2003. “Creative Syntax: Iconic Principles Within the Symbolic”. From Sign to Signing. Iconicity in Language and Literature 3 ed. by Wolfgang Müller & Olga Fischer, 243–261. Amsterdam: John Benjamins. Harris, Catherine L. 1991. “Alternatives to Linguistic Arbitrariness”. Behavioral and Brain Sciences 14.622–623. Holyoak, Keith J. & Paul Thagard. 1995. Mental Leaps. Analogy in Creative Thought. Cambridge Mass.: MIT Press. Hopper, Paul J. 1987. “Emergent Grammar”. Proceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society 1987 ed. by Jon Aske et al., 139–157. Berkeley Linguistics Society. ———. 1988. “Emergent Grammar and the A Priori Grammar Postulate”. Linguistics in Context. Connecting Observation and Understanding ed. by Deborah Tannen, 117–136. Norwood, N. J.: Ablex. Itkonen, Esa. 1994. “Iconicity, Analogy and Universal Grammar”. Journal of Pragmatics 22.37–53. Jackendoff, Ray. 1987. Consciousness and the Computational Mind. Cambridge Mass.: MIT Press. Klooster, Wim G. 2000. Geen. Over Verplaatsing en Focus [‘No(ne). On Movement and Focus’]. Amsterdam: Vossiuspers AUP. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. 2 Vols. Stanford: Stanford University Press. Lass, Roger. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Leezenberg, Michiel & Gerard de Vries. 2001. Wetenschapsfilosofie voor Geesteswetenschappen [‘Philosophy of Science for the Humanities’]. Amsterdam: Amsterdam University Press. Lieberman, Philip. 1991. Uniquely Human. The Evolution of Speech, Thought, and Selfless Behavior. Cambridge Mass.: Harvard University Press. Lightfoot, David. 1999. The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwell. Lyons, John. 1970. Chomsky. London: Fontana/Collins. MacMahon, April. 1994. Understanding Language Change. Cambridge: Cambridge University Press. McCawley, James D. 1989. “INFL’, Spec, and other Fabulous Beasts”. Behavioral and Brain Sciences 12.350–352. ———. 1991. “‘Negative Evidence’ and the Gratuitous Leap from Principles to Parameters”. Behavioral and Brain Sciences 14.627–628.
Grammar change versus language change
Peters, Ann M. 1985. “Language Segmentation: Operating Principles for the Perception and Analysis of Language”. Slobin 1985b.1029–1067. Pinker, Steven. 1994. The Language Instinct. London: Penguin Books. Ritt, Nikolaus. 1996. “Darwinising Historical Linguistics: Applications of a Dangerous Idea”. Viewz 5.27–46. Schlesinger, I. M. 1989. “Language Acquisition: Dubious Assumptions and a Specious Explanatory Principle”. Behavioral and Brain Sciences 12.356–357. Schlesinger I. M. 1991. “Innate Universals do not Solve the Negative Feedback Problem”. Behavioral and Brain Sciences 14.633. Slobin, Dan. I. 1977. “Language Change in Childhood and in History”. Language Learning and Thought ed. by John MacNamara, 185–214. New York: Academic Press. ———. 1980. “The Repeated Path between Transparency and Opacity in Language”. Signed and Spoken Language: Biological Constraints on Linguistic Form ed. by U. Bellugi & M. Studdert-Kennedy, 229–243. Weinheim: Verlag Chemie. ———. 1985a. “Crosslinguistic Evidence for the Language-Making Capacity”. Slobin 1985b.1158–1256. ———, ed. 1985b. The Crosslinguistic Study of Language Acquisition. Vol. 2. Theoretical Issues. Mahwah, N. J.: Erlbaum Associates. ———. 1991. “Can Crain Constrain the Constraints?”. Behavioral and Brain Sciences 14.633–634. ———. 1997 [2001]. “The Origins of Grammaticizable Notions: Beyond the Individual Mind”. The Crosslinguistic Study of Language Acquisition. Vol. 5. Expanding the Contexts ed. by Dan Slobin, 265–323. Mahwah, N. J.: Erlbaum Associates. (Reprinted in shortened form in M. Bowerman & S. C. Levinson, eds. 2001. Language Acquisition and Conceptual Development. Cambridge: Cambridge University Press, 406–449). Smith, Jeremy. 1996. An Historical Study of English. Function, Form and Change. London: Routledge. Snow, Catherine E. & Michael Tomasello. 1989. “Data on Language Input: Incomprehensible Omission Indeed!”. Behavioral and Brain Sciences 12.357–358. Sokolov, Jeffrey L. & Catherine E. Snow. 1991. “A Premature Retreat to Nativism”. Behavioral and Brain Sciences 14.635–636. Wong, Kwok-shing. 2004. “The Acquisition of Polysemous Forms: The Case of bei2 (‘give’) in Cantonese”. Up and Down the Cline — The Nature of Grammaticalization ed. by Olga Fischer, Harry Perridon & Muriel Norde, 325–344. Amsterdam: John Benjamins. Yngve, Victor H. 1996. From Grammar to Science. New Foundations for General Linguistics. Amsterdam: John Benjamins.
63
Indefinite Pronominal Anaphora in English correspondence between 1500 and 1800 Mikko Laitinen University of Helsinki
1.
Introduction
This study investigates the use of the third person pronouns (3rd person) he and they1 in anaphoric reference to various indefinite elements in Early Modern English correspondence. Following current practice (e.g. Newman 1997; Baron 1986), I will use the term ‘epicene’ for the personal pronouns referring to gender-free NPs or indefinite pronouns. Such pronouns, loosely defined as the lexical items in Modern English that can be used in examples (1) and (2), have been discussed extensively in recent literature. (1) Whenever someone works _____ way up through the ranks… (2) A doctor who does _____ job well will earn a lot of money.
The aim of this paper is first to introduce the topic and present previous research on it. I will then discuss some of the factors that may have influenced pronominal usage in diachrony and establish a preliminary framework within which the anaphoric relations between indefinite animate-referring elements and 3rd person masculine singular and plural personal pronouns could be examined.2 I will then
1.The lexemes he and they are used as generic terms, including the possessive, objective and reflexive forms. 2.It should be noted that the selection of he and they as the pronoun variants for the present study is based on both research economy and the research question set here. All anaphora studies are enormously time-consuming, even with the help of computerised corpora, since determining the textual relations of two or more linguistic elements requires subjective interpretations that can only be achieved by close reading of the texts (see Curzan 2003: 197–203 for an example). The pronouns she, it and disjunctive forms he or she may also function as epicenes, but it should be noted that no such cases have been found while searching the texts for this study. This of course does not prove that they do not occur, but
66
Mikko Laitinen
present the results of a pilot study for a project intended to investigate concord (or to follow the early grammarians, government (rectio) and agreement (convenientia); see Kemp 1972: 38) in Early and Late Modern English. Special attention will be paid to the role of the emerging ideas of grammatical correctness (e.g. Sundby et al. 1991: 102–158). Lastly, conclusions and projections for further approaches to indefinite animate anaphora will be discussed.
2. Previous research Textbooks on English pronouns usually contain a section discussing the selection of ‘singular they’ or ‘generic he’ used as anaphoric pronouns to indefinite elements, either indefinite pronouns or NPs (e.g. Wales 1996: 110–133; Mühlhäusler & Harré 1990: 228–247). This often takes place under heavily stigmatised labels such as ‘sexist usage’, ‘enigma in grammar’, ‘problems in pronouns’, etc. In addition, numerous studies on pronoun selection in Present-day English (PDE) have been published in recent years. Scholars have examined both the spoken and written modes, and indefinite anaphora in various geographical varieties has been studied (e.g. Erdman 1995; Newman 1997; Gerner 2000; Pauwels 2001; Baranowski 2002; Laitinen 2002b). It is fair to claim that in PDE the 3rd person plural pronouns are used much more widely than he as anaphoric pronouns to compound indefinites. The topic under discussion has been approached from numerous angles. Some writers have emphasised feminist approaches to indefinite anaphora and have been interested in the alleged androcentrism in grammar books (Bodine 1990). The 3rd person singular pronouns in PDE retain the linguistic marking of gender, and the choice between the masculine, feminine and neuter forms is primarily based on the sex of the person/animal referred to (notional or natural gender). Thus sexist language debates have often focused on them. Furthermore, the differences between compound indefinite pronoun antecedents in relation to 3rd person pronouns have been discussed, since anaphora always
PDE studies (Gerner 2000; Baranowski 2002) have shown he and they to be the major alternatives. Therefore, to make the manual determination of anaphoric relations feasible, the research question only includes pronoun number as the category to be scrutinised. This also emphasises the fact that various factors may underlie indefinite anaphora, such as structural (i.e. form of the antecedents) and thematic (i.e. their semantic roles) issues that are projected through singularity and plurality. The third factor of future interest is of course the sociolinguistic level.
Indefinite Pronominal Anaphora in English correspondence
involves more than one linguistic marker.3 It has been suggested that, as the 3rd person pronouns distinguish number, it might be possible to examine semantic differences between the two compound indefinite pronoun paradigms of -one and -body, traditionally seen as synonymous, by examining the singular and plural 3rd person pronouns that are co-referential with the indefinites (Svartvik & Lindquist 1997; Laitinen 2002b). It is thus fair to claim that this seemingly simple linguistic phenomenon has turned out to be a complex research issue. Some of the writers concentrating on the synchronic aspects of Modern English, notably Newman (1997: 232), have called for diachronic studies to find out more about the origin and development of indefinite animate anaphora. Epicene pronouns are widely discussed in grammar books and writings on Early and Late Modern English (Jespersen 1914: 137–141; Poutsma 1914: 310–314; Fowler 1926: 648–649; Leonard 1929: 224–225; Curme 1935: 238–239). Some scholars include long lists of examples of both the plural and masculine singular uses in indefinite anaphora (e.g. Visser 1963: 75; McKnight 1925: 12–13). Views on epicene usage have been expressed, as in examples (3) and (4) below, but, until recently, this is where the work has stopped. (3) Singular they has, in fact, been well established in informal usage for centuries; until prescriptive grammarians decreed it was grammatically ‘incorrect’, and so out-lawed it, effectively, from (public) written discourse. (Wales 1996: 126; bold emphasis added) (4) prior to the beginning of the prescriptive grammar movement in English, singular ‘they’ was both accepted and widespread … grammarians’ attack on singular ‘they’ was socially motivated (Bodine 1990: 166)
There are surprisingly few modern corpus-based studies on the diachronic aspects, and in fact Curzan’s seminal study of gender shifts in the history of English (2003) is the first to devote space to the topic. Indefinite animate anaphora and epicene pronouns are touched upon as part of the general shift from grammatical gender to natural gender. The chronological periods in the study cover the earliest varieties of English, from Old English to Early Middle English, and PDE. Curzan’s account of the earliest records of English shows that almost all anaphoric references to the pronouns mann, man, sum, hwa, hwilc, and oðer were carried out using the masculine 3rd person singular (2003:64–70).
3.In its prototypical form, anaphora involves an antecedent and another element, whether a personal or demonstrative pronoun or an ellipsis (see Botley & McEnery 2000: 2).
67
68
Mikko Laitinen
For epicene uses, her results show that both he and they were in use as early as Old English (2003: 70–73). The approach here not only continues Curzan’s study chronologically, but also complements it by emphasising morpho-syntactic stratification of the antecedents as well as discourse semantic aspects of indefinite animate anaphora (as defined in Laitinen 2002b). The present study fills the gap in indefinite anaphora research between late Middle English and PDE not covered in Curzan’s pioneering work. My aim is to take one aspect of indefinite animate anaphora, namely human generic reference, and to examine the proportions of the two 3rd person pronoun variants systematically in corpus material. The work builds heavily on Curzan, whose claim that “he occurs as early as Old English and seems to appear consistently and continuously through later stages in the history of English” will be scrutinised using material from between 1500 and 1800 (Curzan 2003: 70).
3. Background This section discusses the well-attested development of the English indefinite pronoun system and suggests that epicenes must be viewed not purely as a social phenomenon but as linked with more general morpho-syntactic changes in English. Indefinite anaphora and epicene pronoun usage are here seen to be influenced by several inter-related factors, and the aim is to explore whether these changes affected pronominal anaphora to the extent that the phenomena described in (3) and (4) were possible. Four issues underlie this discussion. Firstly, within a few centuries, the English indefinite pronoun system underwent considerable changes. Raumolin-Brunberg (1994) discusses the semantic and morphological properties of the compound indefinites and their development in Early Modern English, emphasising grammaticalisation in this process, since each of the second morphemes (-man, -one, -body) previously had similar meanings, a human being, an individual. She shows that the proportion of -man compounds slowly decreases towards the end of the Early Modern period as a consequence of a semantic drift in the meaning of man; -body forms increase, especially in oral registers. In addition, the compounds in -one slowly increase, particularly in the universal every paradigm. It should be noted that, of these three (i.e. -man, -one, -body), man has continued its existence as a generic term for human beings, as in “Man is a problem-solving animal and his ingenuity is endless”, in contrast to an adult
Indefinite Pronominal Anaphora in English correspondence
male human being. This might be one of the reasons why the masculine he has sometimes been viewed as a more unmarked generic form. Rissanen has shown that man in Middle English underwent semantic and morphological changes in which the indefinite pronoun man (cf. man in present-day Swedish and German) was reduced to me in Early Middle English (1997: 519). Its noun-like use (human being in contrast to God, angels, etc.) ‘degrammaticalised’ and lost its position in the category of pronouns (Rissanen 1997: 517–521; see also Los 2002). Raumolin-Brunberg & Kahlas-Tarkka (1997) trace the development of the indefinite pronouns with singular human reference from Old English to the early eighteenth century in a study based on the Helsinki Corpus of English Texts. Secondly, it would be logical to assume that as a result of a grammaticalisation process (and ‘degrammaticalisation’ in the case of man) singular pronouns would be found more often, at least, with the compound indefinites, since they consist of two morphemes of which the second (-one, -man, and -body) were all syntactically and semantically singular. It could, therefore, be expected that the grammaticalisation process would exhibit some etymological conditioning that would carry over to the co-referential third person pronouns (see e.g. Nevalainen (1991: 256–259) for etymological conditioning). This hypothesis is suggested here since grammaticalisation is normally thought to be a slow process, and the changes in the indefinite pronoun paradigms, as discussed by Raumolin-Brunberg, were said to be remarkably slow (1994: 318). Thirdly, the realisation of indefinite anaphora in epicene pronouns is closely related to morpho-syntactic agreement. One approach to the topic is to examine it as the interface between linguistic performance and the sociocultural emergence of grammatical codification of concord, both subject-verb and pronoun-antecedent. An interesting fact in PDE is that the compound indefinite pronouns are morphologically singular even though they may be used with plural reference, as in “Everybody knows the answer” (Quirk et al. 1985: 377–378). Following the basic tenets of Western structuralism, there is nothing inherently singular in the indefinite pronoun form, and the nonplurality has been brought about by convention. Indefinite anaphora and epicene pronouns, therefore, could be examined in the light of the codification process that began with the emergence of the first English grammars in the sixteenth century and which eventually turned into a means of promoting grammatical correctness and normativism in the eighteenth century (for the relationship between the early grammarians and normativism, see Leonard 1929: 9; Vorlat 1998: 485ff.). Although the first known proscription of epicene
69
70
Mikko Laitinen
they in indefinite anaphora appears as late as 1795 in Lindley Murray (example (5) below), what the early grammarians themselves advocated was the masculine singular. (5) Rule V. Pronouns must always agree with their antecedents, and the nouns for which they stand, in gender, number, and person; …Of this rule there are many violations to be met with; a few of which may be sufficient to put the learner on his guard. “Each of the sexes should keep within its particular bounds, and content themselves with the advantages of their particular districts.” “Can any one, on their entrance into the world, be fully secure that they shall not be deceived?” “on his entrance,” and “that he shall.” “Let each esteem others better than themselves;” “than himself.” (Murray 1795: 95–96)
The advocacy of masculine singular becomes apparent in the agreement rules in the preliminary agreement hierarchies that were devised, although not explicitly concerning indefinite anaphora (6)–(7), as well as in the grammarians’ own usage (examples (8)–(10)). (6) let antecedents be Found-out for eury relatiu: let such too ruls agre: For verbs number and persn must agre untoo his cas: As relatiu, in gender too, with antecedents grac. …Whoo in persn and gender must with most-worthy agre: Whaer first persn is worthiest, the second is the next. The masculin, then feminin gender looketh too be best: (Bullokar 1586 [1980] 57) (7) The Relative agrees with the Antecedent in gender, number, and person …The relative shall agree in gender with the Antecedent of the more worthy gender: as, the King and the Queen whom I honor. The Masculine gender is more worthy than the feminine. (Poole 1646: 21, quoted in Bodine 1990: 172) (8) I ever found in teaching of children accordin to those spelling books; he that does not beleive this, let him make an experiment of the same, and if when he puts another …he findes them (Aickin 1693 [1967] Preface; bold added) (9) The modern, as well as old Grammarians, have given us various definitions, of this very useful Art. That a certain author seems defective, when he says, Grammar is the art of Speaking; (Gildon & Brightland 1711: 1; bold added)
Indefinite Pronominal Anaphora in English correspondence
(10) because a person’s sensations, at the time of perusing it, are of the agreeable kind; and not waiting to make a proper distinction among his ideas, he confounds his ideas of words …and if he read, he is apt to (Priestley 1761: 51–52; bold added)
Lastly, as shown by Schlauch (1958), there must have been something comparable to linguistic insecurity in modern Labovian terms (2001: 275–279) during the Early Modern period, just as there is today (Meyers 1990). Schlauch cites a number of examples, mainly from the fifteenth and sixteenth centuries, in which a plural antecedent is followed by a singular anaphoric personal pronoun. She compares this to a discourse method in which a general situation “is made concrete by a transition from plural to typical singular” (Schlauch 1958: 236). Her many examples also include the contrary cases, with a singular having a co-referent in the plural, as in (11): (11) or els some yong Marchant man or other kynde of Occupier, whose friendes hath geuen them a stock to occupy withal (Awdeley. The Fraternitye of Vagabones. ca. 1560, quoted in Schlauch 1958: 239)
The ambiguity created by these cases, she claims, resembles constructions where a plural personal pronoun is used with a singular indefinite. These four issues form the basis for this work, but since this is a pilot study not all will be treated comprehensively. These issues suggest that there may have been changes going on at the beginning of the Early Modern English period, or (if not changes then) at least some confusion on which the emerging codification process may have had an influence.
4. Material and methods This study is based on material drawn from the Corpus of Early English Correspondence (Nevalainen & Raumolin-Brunberg 1996) and its eighteenth-century Extension, currently being compiled (Laitinen 2002a).4
4.The Corpus of Early English Correspondence (1998 version) and its Extension, currently in compilation, represent “texts that mirror the informal spoken language of past times as closely as possible” (Nevalainen & Raumolin-Brunberg 2003: 28). It is therefore emphasised that the study adopts the hypothesis that the more informal written forms of Early Modern English remained more resistant to coding and the normative practices emerging in the second half of the eighteenth century than, for instance, written records intended for publication. There is no way to verify this assumption, and we have to follow the uniformitarian principle
71
72
Mikko Laitinen
As the purpose was to find as many indefinite and non-specific cases of indefinite pronouns or NPs with co-referential personal pronouns as possible, it was decided to include a wide variety of lexemes in the queries (for a discussion of indefiniteness and (non-)specificity, see Laitinen 2002b). The lexemes have been chosen to account for the changes that took place in the indefinite pronoun paradigms during the Early Modern English period (see Raumolin-Brunberg 1994:314). The lexeme types selected (Table 1 below) include the indefinite pronouns, as well as other lexemes that could be used non-specifically or with nonspecific words, such as the relative pronoun who with its objective and possessive forms, who compounds, such as whoever, whomsoever, etc. other, another, and one and man were also used. As mentioned above, man is somewhat problematic since, as has been shown by Raumolin-Brunberg (1994: 305–306), it is sometimes impossible to know whether the writer is making a generic statement or refers to the male population only when using man. Extract (12) shows an example of the latter. It is excluded from the analysis. (12) Tells her that both of them have their Dalilahs, which shee disclaimes absolutely, and vowes shee will never keepe company with anie man hee shall tell her hee dislikes. (Henry Oxinden, 1651; oxinden,II,168)
In addition to the indefinite elements, a set of nouns was selected (Table 1 below). The purpose of including these was to compensate for the relatively few tokens for each indefinite lexeme. Table 1.Lexeme types used as indefinite antecedents. Every (-body, -one, or any singular, non-specific noun), each Somebody, -one (or some followed by a singular, non-specific noun) Any (-body, -one, or any singular, non-specific noun) No (-body, -one, or any singular, non-specific noun), none Who, whoever, whomsoever Other, another One Man NPs — adult, body, child, cousin, defendant, friend, fellow, individual, member, offender parent, person, patient, relative, victim, witness
in historical linguistics (e.g. Lass 1997: 25). Since the present study aims at modelling the linguistic stratification, whether structural, social or textual, of this particular linguistic construction, it is to be acknowledged that the patterns presented here, based on a particular textual range, remain broad.
Indefinite Pronominal Anaphora in English correspondence
As the study contains several types of pronoun ranging from single-word forms to the emerging compounds and other indefinite lexemes, there is no way to avoid some degree of subjectivity in deciding the borderline cases (see also Raumolin-Brunberg & Kahlas-Tarkka 1997: 20). I would like to emphasise that the quantification here is not intended as the last word on the issue. It is certain, however, that some tendencies will be confirmed by the study. The pronouns searched for had to meet the following criteria: a. They had to be indefinite and non-specific (see Laitinen 2002b). b. They had to have a singular gender-non-specific co-referent. Some of the examples that were included can be found in (13)–(15). The antecedents in these sentences could be paraphrased as anyone, a person, an individual. (13) When one does not know what to say they talk of the weather, the sudden change will give me an opportunity to make my letter a little longer (Peter Wentworth, 1708; wentworth,68) (14) Yndeede, the miserable estate off the worlde ys suche that yff there be anye man fownde to doo his dewtye, espiciallie yn thys hie ministerie (as God knowith I doo hyt not by a great deale), he maye be as (Laurence Saunders, 1551; johnson,1137) (15) I wish amongst the variety of Acquaintance you may find some one to please you, and can’t help the vanity of thinking, should you try them all, you won’t find one that will be so sincere in their treatment, tho’ a thousand more deserving, and every one happier. (Lady Mary Wortley Montagu, c1710; montagu,I,66)
Example (16) shows a case that was discarded on the basis that the antecedent was specific and known, i.e. the speaker referred to (a) specific referent(s) whose identity was/were known to the speaker.5 In (16), the referent is clearly specific and plural, and in such cases the use of a singular would be misleading
5.The distinction between ‘indefinite’ vs. ‘definite’ and ‘non-specific’ vs. ‘specific’ follows closely Haspelmath’s typological survey of indefinite pronoun functions in conceptual space (1997: 31–52). To paraphrase his description, indefinite pronouns with ‘specific known’ referents were discarded from the analysis, and only those with ‘specific unknown’ and ‘irrealis nonspecific’ referents meet the criteria of being capable of use in generic reference. It should, however, be emphasised that since questions of reference, not to mention those of definiteness and specificity, are too broad to be comprehensively discussed here, the study employs the typological classification as a set of selection tools in historical corpus linguistics.
73
74
Mikko Laitinen
and ungrammatical, while in (17) the referent of the indefinite and its coreferential NP must be gender-specific and male, since bishops at the time were all men. (16) Lady Louisa sends her love to everybody and hopes they will not forget the Irish chest of drawers, and begs (Ann Porter, 1792; porter,136) (17) that whosoever shuld be made bishop for england, shuld be bound to residence …though it cost him his life (Thomas Fitzherbert, 1608; fitzherbert,17)
5. Results As the previous systematic corpus data on indefinite anaphora and epicene pronouns in diachrony remain limited, and no clear patterns have been established, the results will be presented on a general level. A more detailed analysis must be left for future studies. The quantification of epicene pronouns is given by century, since there is a great deal of individual variation in the pronoun distributions. The singular and plural variants are divided into five categories according to their antecedents: (a) indefinite pronouns, including the compound forms written both together and separately. This group also includes cases where an indefinite pronoun is followed by a singular noun (every person), or a postmodifier (any of the workers); (b) singular agentive noun phrases, equivalent to a person; (c) man in a generic sense (i.e. synonymous with everyone or a person). A good illustration of the genericness of man comes from William Jones in (18): (18) that men are the principal objects of a man’s study and contemplation, that we cannot see and know too many of them (William Jones, 1778; jones, I,279);
(d) one used alone in its generic sense; (e) the miscellaneous category includes everything else, such as other, another, etc. Non-specific cases such as he who or they who are included here. From these I have excluded the specific cases preceding the letter-ending greeting, as in (19): (19) yr full content from him who must live discontented …to his chast longinge desires. Yrs, in all harty affection. Chr. Hatton. (Christopher Hatton, 1601; hatton,I,2)
Indefinite Pronominal Anaphora in English correspondence
Furthermore, some used alone was discarded, since in Early Modern English, as in Present-day English, it was clearly plural. Table 2 below reveals some overall tendencies. The proportions of the singular uses remain higher throughout the three centuries, with no statistically significant differences between the numbers. The plural forms are, nevertheless, frequent, accounting for nearly half the occurrences in each century. The results here seem not to contradict the intuitive assumptions made by earlier writers, since they were mainly interested in the compound indefinite anaphora as in (20), and my results also include other forms as mentioned earlier: (20) Here is very prodigious news, wch noebody wou’d believe till they saw ye event. (Charles Hatton, 1679; hatton,I,177) Table 2.Total occurrences and proportions of singular and plural variants. Period
N
1500–1599
257
Sg. 131 (51%) Pl. 126 (49%)
1600–1699
670
Sg. 372 (56%) Pl. 298 (44%)
1700–1800
280
Sg. 148 (53%) Pl. 135 (47%)
Varying sample sizes account for the differences in the overall occurrences. The sixteenth century results are based on a sample of about 500,000 words, whereas the other two are somewhat larger, roughly a million words in the seventeenth century, and about 800,000 in the eighteenth. There are, however, no great differences caused by this variation, as the proportions of the singular and plural pronouns remain almost equal. When the antecedents are compared, Table 3 below shows that non-specific indefinite pronouns are indeed used more with plural co-referential 3rd person pronouns. The proportions remain somewhere around 65 to 70 per cent, with no statistically significant differences between the periods, suggesting that the codification process in English had had no impact on indefinite anaphora in correspondence. However, we should bear in mind that prescriptive attitudes had not emerged in full prior to the mid-eighteenth century, and the rough division of the results by centuries may, therefore, not capture much detail. Furthermore, it should be remembered that I have not distinguished here
75
76
Mikko Laitinen
between the indefinite pronoun paradigms to see whether their grammaticalisation process might have affected indefinite anaphora. Sentences (21) and (22) are examples of the plural and (23) the singular: (21) but this later goes of so many, that it is not fitt to father it upon any one in perticular, unlesse they know it better then I doe. (Charles Lyttelton, 1671; hatton,I,64) (22) My father being never desir’d to give the little interest he had, resolved between the two late candidates to stand newter and leave every one to dispose of their votes as they thought most suitable to their inclinations, but it seems the doctrine is this (Henry Liddell, 1712; liddell,69) (23) but such was ye unhappiness of that time as ye Treasury was quite empty, and ye insurrections in severall parts made every one who had money look to himself and none could bee found to lend to ye publick; (Arthur Capel, 1676; essex,57) Table 3.Proportions of singular and plural pronouns according to antecedents. IndefPron
NP
man
one
Misc.
1500–99
Sg. Pl.
37 (36%) 13 (72%) 66 (64%) 5 (28%)
67 (75%) 6 (50%) 22 (25%) 6 (50%)
8 (23%) 27 (77%)
1600–99
Sg. Pl.
69 (30%) 69 (83%) 168 (70%) 14 (17%)
149 (92%) 30 (68%) 13 (8%) 14 (32%)
55 (38%) 89 (62%)
1700–1800 Sg. Pl.
41 (30%) 24 (56%) 94 (70%) 19 (44%)
71 (91%) 8 (57%) 7 (9%) 6 (43%)
4 (40%) 6 (60%)
Anaphoric pronouns with NPs, generic man, and one are mainly singular throughout the three centuries of data examined. There is, however, an increase in the proportions of the plural uses with NPs and one in the 18th century, but the absolute figures in the categories are low. The decrease in plural forms with man compared with the sixteenth century is statistically significant. The development in man suggests a direct relation with the acquisition of semantic weight as suggested in Section 3 above. It is, however, too early to discuss the motivations for this variation, although it seems to be clear that the semantic history of man explains the large proportion of masculine 3rd person singular used with it in the sixteenth century. It seems that the prototypical anaphoric personal pronouns with these antecedents were singular, with occasional plural cases. Sentences (24)–(26) are examples of the singular cases, and (27)–(28) of the plural:
Indefinite Pronominal Anaphora in English correspondence
(24) and as well my selfe as eny other to whom his Highnes thowght the thing to seme other wise, he vsed in his other busynes, abyding (of his abundant goodnes) neuer the lesse graciouse lord vnto eny man, nor neuer was willing to put eny man in ruffle or trowble of his conscience. (Thomas More, 1534; more,496) (25) that unless you will help me to somthing that I may find fault with, I must return… The patient must do so much for his phisitian, he must say where the pain lies, or else the dosis must be given at random. (Brian Duppa, 1650; duppa,18) (26) but I will not quit the sight of the person who does it, lest he should put any deception on me, or commit any mistake. (Ann Porter, 1798; porter,189) (27) and then if every man shall enjoie that whiche they get of our goodes by attachementes, I knowe not how to satisfie your mastership, (John Johnson, 1553; johnson,1664) (28) The troth is if a man do but consider the globe of the world with its many and vast nations to which we seeme so small a spot, ‘twould make one very cautious how they applied the mysteries of the Revelation unto this angle, (Justinian Isham, 1651; duppa,38)
In the Miscellaneous category, the plural cases are favoured. Sentences (29)–(30) are plural examples in this category, while (31) is singular: (29) so genneral a justice for all persons as never to condemn any with out indenyable proufs of theare guiltt but Mrs Scrow is so little a person in my opinion and so sildom in my thoughts that whoever gave her that information I forgive them though I doe not remember (Elizabeth Stanhope, 1664; giffard,13) (30) and who could bear to live with one they despis’d? (Lady Mary Wortley Montagu, c1710; montagu,I,30) (31) Dr. Johnson says (you know) that whoever would entertain another by his Remarks — must make the subject of them Human Life. (Hester Lynch Piozzi, 1784; piozzi,I,113)
Most of the individual informants exhibit a mixed pronoun selection. One of the topics for future studies will be to account for various types of stratification with which indefinite animate anaphora could be modelled in greater detail. The corpus, however, includes a few writers who consistently used one or the other pronoun more. Such outstanding informants do not, however, skew the
77
78
Mikko Laitinen
overall patterns presented above, since the number of all occurrences remains high. Neither can one draw any further conclusions from this, but it may be hypothesised that the consistent usage suggests consciousness of this linguistic construction. For instance, Arabella Stuart in the early seventeenth century has all her indefinite animate anaphoric pronouns in the plural, as in examples (32)–(33), and none in the singular. (32) And whither I may send for whom I thinck good or talke with any that shall voluntarily or upon businesse comm to me, in private if they or I shall so desire (Arabella Stuart, 1603; stuart,135) (33) I trust I have not lost so much of your good opinion as your pleasant postscript would make one that weare suspitious of theyr assured frends (as I never was) beleeve. (Arabella Stuart, 1603; stuart,194)
There are others who mixed the uses but clearly preferred one of the variants, such as Dorothy Osborne, with 44 plural cases and 15 in the singular. Charles Burney, the late eighteenth-century music historian, uses the singular almost exclusively, having only one case in the plural and 17 in the singular. William Clift, the youngest son of a poor Cornish miller and soon-to-be Conservator of the Hunterian Museum, has 9 occurrences in the plural and 13 in the singular, neatly divided so that the plural epicenes occur with the indefinites and in the Miscellaneous category, and the singulars with NPs and man. One of the questions for future studies may be to find out whether the writer’s sex was a significant factor in epicene pronoun usage.
6. Conclusion I have outlined epicene pronoun usage from diachronic research material, suggesting that this topic can be investigated reliably and systematically using computerized corpora. The results also show that the topic merits further investigation. As pointed out, the results should be judged as strong tendencies, since no semantically tagged historical corpora exist to date, and finding indefinite and co-referential pronoun pairs, even with the help of sophisticated corpus tools, requires a lot of hand-editing (cf. Curzan 2003: 197–203). The results of this study indicate that in English correspondence between 1500 and 1800 the use of epicene pronouns in indefinite anaphora can be divided into two groups. Indefinite animate reference is almost equally divided between the singular and plural pronouns (Table 2 above). As far as structural
Indefinite Pronominal Anaphora in English correspondence
stratification of the antecedents is concerned, the indefinites with anaphoric 3rd person pronouns are preferred with the plural and the other categories with the masculine singular (Table 3 above). This division relates to writers as well antecedents. There are those who favour form over meaning, using the singular form of the pronoun with morphologically singular antecedents, and those who prefer meaning over form, using plural pronouns with co-referents whose sense can be interpreted as plural. This topic offers an interesting field of research, especially for studies aimed at investigating the interface between language as a system and language users’ choices and the various factors affecting them (see Rissanen 1999:188–189). This means that not only must linguistic history and the role of grammaticalisation of the antecedents in this process be accounted for, but also that the standardisation/ codification process, along with the growing awareness of language which started during the Early Modern English period, must not be ignored. Another interesting factor in indefinite anaphora is the motivation for various pronominal morphemes in the antecedents. Several recent studies have suggested that in Modern English there seem to be some remains of the former senses of the -one and -body paradigms reflected in the personal pronouns used anaphorically with them (Bolinger 1977; Svartvik & Linquist 1997; Laitinen 2002b). Logically it follows that this must also have been the case in Early Modern English.
References Aickin, Joseph. [1693] 1967. The English Grammar. (= English Linguistics 1500–1800, 21.) Menston: Scholar Press. Baranowski, Maciej. 2002. “Current Usage of the Epicene Pronoun in Written English”. Journal of Sociolinguistics 6:3.378–397. Baron, Dennis E. 1986. Grammar and Gender. New Haven: Yale University Press. Bodine, Ann. 1990. “Androcentrism in Prescriptive Grammar: Singular ‘They’, Sex-indefinite ‘He’ and ‘He or She’”. The Feminist Critique of Language. A Reader ed. by Deborah Cameron, 166–186. London & New York: Routledge. Bolinger, Dwight. 1977. Meaning and Form. London & New York: Longman. Botley, Simon & Anthony Mark McEnery. 2000. “Discourse Anaphora: The Need for Synthesis”. Corpus-Based and Computational Approaches to Discourse Anaphora (= Studies in Corpus Linguistics, 3.) ed. by Simon Botley & Anthony Mark McEnery, 1–41. Amsterdam & Philadelphia: John Benjamins. Bullokar, William. [1586] 1980. Pamphlet for Grammar. Reprinted in The Works of William Bullokar ed. by J. R. Turner, vol. II. (= Leeds Texts and Monographs. New Series I.) Leeds: University of Leeds.
79
80
Mikko Laitinen
Curme, George O. 1935. A Grammar of the English Language. Vol. II Parts of Speech and Accidence. Boston: D. C. Heath and Company. Curzan, Anne. 2003. Gender Shifts in the History of English. Cambridge: Cambridge University Press. Erdman, P. 1995. “Pronoun Agreement with Compound Indefinite Pronouns as Referent”. Arbeiten aus Anglistik und Amerikanistik 20.75–92. Fowler, H. W. 1926. A Dictionary of Modern English Usage. Oxford: Clarendon Press. Gerner, Jürgen. 2000. “Singular and Plural Anaphors of Indefinite Personal Pronouns in Spoken British English”. Corpora Galore. Analyses and Techniques in Describing English. Papers from the Nineteenth International Conference on English Language Research on Computerised Corpora (ICAME 1998) ed. by John M. Kirk, 93–114. Amsterdam & Atlanta: Rodopi. Gildon, Charles & John Brightland. 1711. A Grammar of the English Tongue. Landmarks in English Grammar. The Eighteenth Century (CD-Rom) ed. by Gerald Nelson. London: University College London. Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Oxford University Press. Jespersen, Otto. 1914. A Modern English Grammar on Historical Principles. Part II: Syntax. First Volume. Heidelberg: Carl Winter. Kemp, J. A., ed. 1972. John Wallis. Grammar of the English Language, with an Introductory Grammatico-physical Treatise on Speech. London: Longman. Labov, William. 2001. Principles of Linguistic Change. Social Factors. Oxford: Blackwell. Laitinen, Mikko. 2002a. “Extending the Corpus of Early English Correspondence to the 18th Century”. Helsinki English Studies 2. http://www.eng.helsinki.fi/hes/ ———. 2002b. “Singular HE and Plural THEY in Indefinite Anaphora in Written Presentday English”. International Journal of Corpus Linguistics 7:2.137–164. Lass, Roger. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Leonard, Sterling Andrus. 1929. The Doctrine of Correctness in English Usage. Madison: University of Wisconsin. Los, Bettelou. 2002. “The Loss of the Indefinite Pronoun Man. Syntactic Change and Information Structure”. English Historical Syntax and Morphology. Selected Papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000 ed. by Teresa Fanego, María José LópezCouso & Javier Pérez-Guerra, 181–202. Amsterdam & Philadelphia: John Benjamins. McKnight, George H. 1925. “Conservatism in American Speech”. American Speech 1:1.1–17. Meyers, Miriam Watkins. 1990. “Current Generic Pronoun Usage: An Empirical Study”. American Speech 65:3.228–237. Mühlhäusler, Peter & Rom Harré. 1990. The Linguistic Construction of Social and Personal Identity. Oxford: Blackwell. Murray, Lindley. 1795. English Grammar. Landmarks in English Grammar. The Eighteenth Century (CD-Rom) ed. by Gerald Nelson. London: University College London. Nevalainen, Terttu. 1991. But, Only, Just. Focusing Adverbial Change in Modern English 1500–1900. (= Mémoires de la Société Néophilologique de Helsinki, LI.) Helsinki: Société Néophilologique. ——— & Helena Raumolin-Brunberg, eds. 1996. Sociolinguistics and Language History. (= Language and Computers: Studies in Practical Linguistics, 15.) Amsterdam & Atlanta: Rodopi.
Indefinite Pronominal Anaphora in English correspondence
——— & Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Longman. Newman, Michael. 1997. Epicene Pronouns. The Linguistics of a Prescriptive Problem. New York & London: Garland. Pauwels, Anne. 2001. “Non-sexist Language Reform and Generic Pronouns in Australian English”. English World-Wide 22:1.105–119. Poutsma, H. 1914. A Grammar of Late Modern English. Part II: The Parts of Speech. Gröningen: P. Noordhoff. Priestley, Joseph. 1761. Rudiments of English Grammar. Landmarks in English Grammar. The Eighteenth Century (CD-Rom) ed. by Gerald Nelson. London: University College London. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London & New York: Longman. Raumolin-Brunberg, Helena. 1994. “The Development of the Compound Pronouns in -Body and -One in Early Modern English”. Studies in Early Modern English ed. by Dieter Kastovsky, 301–324. Berlin & New York: Mouton de Gruyter. ——— & Leena Kahlas-Tarkka. 1997. “Indefinite Pronouns with Singular Human Reference”. Grammaticalization at Work. Studies of Long-term Developments in English ed. by Matti Rissanen, Merja Kytö & Kirsi Heikkonen, 17–86. Berlin & New York: Mouton de Gruyter. Rissanen, Matti. 1997. “Whatever Happened to the Middle English Indefinite Pronouns?”. Studies in Middle English Linguistics ed. by Jacek Fisiak, 513–529. Berlin & New York: Mouton de Gruyter. ———. 1999. “Syntax”. The Cambridge History of the English Language. Vol. III Early Modern English 1476–1776 ed. by Roger Lass, 187–331. Cambridge: Cambridge University Press. Schlauch, Margaret. 1958. “The Generic Singular in English: A Supplement”. Kwartalnik Neofilologiczny. Rocznik V. Panstwowe Wydawnictwo Naukowe. 235–239. Sundby, Bertil, Anne Kari Bjørge & Kari E. Haugland. 1991. A Dictionary of English Normative Grammar 1700–1800. (= Amsterdam Studies in the Theory and History of Linguistic Change, 63.) Amsterdam & Philadelphia: John Benjamins. Svartvik, Jan & Hans Lindquist. 1997. “One and Body Language”. From Ælfric to the New York Times. Studies in English Corpus Linguistics ed. by Udo Fries, Viviane Müller & Peter Schneider, 11–20. Amsterdam & Atlanta: Rodopi. Visser, F. Th. 1963. An Historical Syntax of the English Language. Part I: Syntactical Units with One Verb. Leiden: E. J. Brill. Vorlat, Emma. 1998. “Criteria of Grammaticalness in the 16th and 17th Century English Grammar”. A Reader in Early Modern English ed. by Mats Rydén, Ingrid Tieken-Boon van Ostade & Merja Kytö, 485–496. Frankfurt am Main: Peter Lang. Wales, Katie. 1996. Personal Pronouns in Present-day English. Cambridge: Cambridge University Press.
81
From resultative predicate to event-modifier The case of forth and on Bettelou Los Free University of Amsterdam
1.
Introduction
This paper investigates the origin of English particle-verb combinations. These verbs consist of a verb and an adverbial or prepositional element traditionally referred to as ‘particles’, which includes items like away, back, down, forth, in, off, on, out, over, up. Particles are notoriously difficult to define as a class because of their diverse origins — some are homophonous with prepositions, but others represent grammaticalized prepositional phrases. The semantic contribution of the particle is at times unclear, as is its syntactic status: is it a word or a phrase? Then there is the status of the particle-verb combination: do these particle-verbs represent a multi-word verb or do they constitute a phrase? As we will see, their syntactic behaviour is remarkably uniform, so much so that it is possible to speak of a ‘particle syntax’. ‘Particle syntax’ is not restricted to the items traditionally included in the particle inventory, but includes certain adjectives and prepositional phrases. What these items all have in common is that they derive from secondary predicates which have undergone some degree of grammaticalization which has allied them with the verb rather than with the object. I propose to extend the secondary predicate analysis to particles, an analysis that has been around in the literature for some time. Presenting particles as grammaticalized predicates goes some way to account for their dubious syntactic status (they are phrases on their way to becoming heads) but does nothing for the quirks consistently reported in the literature: particles appear to have transitivizing effects, they add telicity to an event, their presence allows the verb to select a different set of objects than the verb would select on its own, they at times appear to be redundant. These quirks can be accounted for by considering the semantics of secondary predicates,
84
Bettelou Los
which we will do in terms of the Lexical Conceptual Structure (LCS) of resultatives. These semantics also have something to say about the sort of verb that enters into a verb-particle combination. Particles have also been said to detransitivize rather than transitivize their verbs. We will show in the second part of this paper that this paradox is resolved by tracing the history of these detransitivizing particles. As it turns out, they too can be argued to derive from resultative constructions, but they have left the resultative LCS to become event-modifiers. They no longer allow objects because the presence of an object would signal resultative semantics. We will present some diachronic data on two of them, forth and on.
2. Particle quirks are due to the resultative semantics of the construction 2.1 Particle quirks The paradoxical behaviour of English phrasal verbs, particularly the particleverbs as in (1a) below, has given rise to a vast body of literature over the years. The syntactic status of the combination cannot be easily determined. Its syntactic separability as in (1b) points to it being a phrase (an ‘XP’), as does the ability of some particles to topicalize, as in (1c). The constraints restricting such preposing include not so much the requirement that the particle must be used literally, as has often been suggested in the past (e.g. Fraser 1976: 58–59), but semantic transparency of both particle and verb; see Capelle’s (2002) corpusbased study. (1) a. He threw away the remains of his dinner. b. He threw the remains of his dinner away. c. Out he went without saying good-bye.
Its semantics, however, are those of a complex verb, pointing to an analysis in which the verb-particle combination is a Head (‘X0’): its meanings are often not recoverable from its constituent parts, again pointing to verb and particle being stored as a unit in the lexicon. Particle-verbs can be input to morphological processes, witness examples like get-at-able, come-at-able and looker-on, all listed in the OED; this, too, is a characteristic of X0s rather than XPs, as phrases cannot normally be input to derivational processes. There are typical characteristics of particle-verbs that crop up again and
From resultative predicate to event-modifier
again in the literature: valency and object selection. The addition of a particle often affects the valency of the verb, as it may transitivize an intransitive verb or intransitivize a transitive one. The particle often adds telicity and converts activities into accomplishments or achievements. The semantic contribution of the particle is often unclear, and the particle sometimes appears to be redundant (print a file versus print out a file). Then there is the matter of the ‘unselected object’: even if simplex and particle-verb are both transitive, they often cannot occur with the same objects, witness (2a–b): (2) a. He bought a house. b. He bought out the shareholders.
Then there is the phenomenon of ‘object transfer’: (3) a.
John poured out the bucket/John poured out the water. (McIntyre 2001) b. Clear out a river (by removing mud)/clear out mud (from a river). (Lipka 1972: 171–173)
The verbs themselves also exhibit certain quirks typical of particle-verbs. Many verbs occur only in combination with a particle, never on their own as a simplex: peter out, jazz up. The non-existence of a simplex verb in many of these cases is due to the conversion properties of particles, witness the nominal conversion of zip up or the adjectival conversion pretty up, but this does not explain why particles possess these conversion properties in the first place. We will see that these characteristics all follow from the resultative semantics of the construction. 2.2 The resultative LCS 2.2.1 The resultative semantics of particle verbs The observation that particles often convey a resultative meaning is not new. Visser employs the term ‘effective adverb’ (1963:I,597), taken from Curme (see Denison 1981: 64); Lipka observes that both German and English particles indicate the result, and often function much like adjectives (Lipka 1972: 115–116). We will show in this section that the typical quirks of particleverbs are due to their resultative semantics. We will adopt for our discussion the semantic representation of resultative predicates in the form of the Lexical Conceptual Structure (LCS, Jackendoff 1990) in the simplified form of (4) from Spencer & Zaretskaya (1998: 6):
85
86
Bettelou Los
(4) [CAUSE[ACT (x)], BECOME[W(y)]], BY[V(x)]]
Example (5a–b) exemplifies how the variables may be filled in: (5) a. He took his coat off. b. [CAUSE[ACT (he), BECOME[off(his coat)]], BY[taking(he)]]
Note first the mismatch between syntactic and semantic embedding: the ‘core predicate’ W, although semantically primary and rendered prominent by stress, is the most deeply embedded constituent syntactically; similarly, the V, prominently encoded syntactically as a verb and therefore in theory expected to play an important role as licenser of arguments, is tucked away in the R-LCS in a peripheral adjunct position. It is this mismatch that is responsible for the quirks presented in the previous section. 2.2.2 The Object We will start with the often-noted transitivizing effect of particle-verbs. Particles appear to create transitive verbs from intransitive ones, or more specifically, unergatives, as unaccusatives do have objects in deep structure. An example of such unergative verbs is play and work in (7)–(8): (7) a.
Handel … asked the organist to permit him to play the people out. (OED 1823) b. They accused the minister of playing down the number of the unemployed.
(8) a. He worked his way up. b. Ten great places to work off weight.1
The relative positions of the verb and the object in the R-LCS show that the V, although prominently encoded syntactically as verb, is in an adjunct position. The object y is not theta-marked by V but by the particle. This accounts for the ease with which secondary predicate constructions build on intransitive verbs: it is not the V that selects the object but the particle. This also accounts for the phenomenon of the ‘unselected object’ in (2) and the ‘object transfer’ in (3): the objects do not fit the selectional restrictions of the higher verb because it is not the verb that does the selecting. All objects in a particle-verb construction are in fact ‘unselected’. There are many statements in the literature to the effect that some particles
1.Taken from http://www.usatoday.com/life/travel/leisure/2001/ltl021.htm
From resultative predicate to event-modifier
appear to detransitivize rather than transitivize (e.g. Kennedy 1920: 26ff, quoted in Lipka 1972: 165; Fraser 1976). Most of the instances cited, however, involve verbs that have joined the causative/unaccusative alternation (also called causative/ergative alternation, Haegeman 1994: 334–336 or causative/inchoative alternation, Lieber & Baayen 1997) that we find in pairs like The enemy sank the ship/The ship sank. Thus we find pairs like (9)–(10): (9) a. The police moved the bystanders on. (causative) b. The bystanders moved on. (unaccusative) (10) a. They cheered him up. (causative) b. He cheered up. (unaccusative)
The alternation may become obscured when the unaccusative of such pairs lexicalizes into a set expression: (11) a. *They stacked up the planes over Kennedy Airport. (causative) b. The planes stacked up over Kennedy Airport. (unaccusative)
There is one class of particles which is really detransitivizing, but they no longer exhibit resultative semantics and have become durative event modifiers. They will be discussed in Section 4. 2.2.3 The particle The ‘W’-variable in the R-LCS in (4) may vary on a scale between extremely specific and extremely abstract. Its only requirement is that it must convey a change-of-state, and this appears to be the function of such bleached particles as up or down in (12), where there is no longer any sense of spatial movement or position. Because W marks the endpoint of the activity, as the activity will stop when the variable y has reached a certain state, up and down are here little more than telicity markers, turning activities, which are (−telic, +durative), into accomplishments (+telic, +durative) or achievements (+telic, −durative): (12) a. Police broke up a silent march of some 500 people. b. Zimbabwe University was ordered to shut down after a day of violent clashes between police and students.
Particles do not have much content to begin with — they typically derive from spatial/directional prepositions or grammaticalized prepositional phrases like away — and are therefore particularly prone to developing idiosyncratic meanings, unlike the more meaningful adjectives of syntactic secondary predicates. Although particles often transitivize verbs, as we saw in Section 2.2.2, there
87
88
Bettelou Los
are cases in which they do not appear to add anything to the simplex meaning of the verb. Examples are print versus print out, phone versus phone up, face versus face up (to), wipe versus wipe down, write versus write up/down. In many of these cases, however, subtle meaning differences do emerge on closer examination (for face – face up to for instance, see Hampe 2000), whereas in others, adding the particle seems to make for greater expressiveness because it activates the resultative LCS and hence draws attention to the effectiveness of the action. 2.2.4 The verb The adjunct-like ‘instrumental’ role of the verb in the LCS template goes some way to account for the often-observed fact that it appears to be the particle that selects the verb rather than the other way around (see Lipka’s 1972 discussion of out and up). Particles combine with four groups of verbs. The first group is the group of monosyllabic ‘light’ verbs as in (13): (13) come, go, keep, let, make, place, put, set, etc.
The second group can be described as ‘means’ verbs as in (14): they denote (a conversion of) the instrument used in causing the object y to reach the state W. (14) boot out, bowl over, branch out, brick up, brush up, buckle up, elbow out, fork out, hammer out, hand over, hem in, knuckle down, pan out, poke about, patch up, peg down/out/away, rake up, rule out, sally forth, seal off, tick off, tide over, top up, trail off, worm out, zip up.
Then there are ‘manner’ verbs as in (15), often verbs describing manner-ofmotion. They describe the specific action required to reach the state W: (15) blot out, bob up, butt in, chew up, chime in, chip in,2 chuck out, crop up, dole out, edge away,3 eke out,4 pare down, peter out,5 point out, polish up, root up/out, rub out, snap up, trot out, veer round, wind up.
Finally, there are the verbs that constitute a conversion of the change-of-state itself, as in (16). It is here that we find the converted adjectives and adverbs, also a much-noted feature of particle-verbs (e.g. Lipka 1972: 98–114):
2.Chip meaning “chop, cut”. 3.Edge meaning “move edgeways”. 4.Eke meaning “increase, lengthen, supplement”. 5.Peter (origin unknown, but first used in mining) meaning “run out and disappear (as a stream, a vein of ore)” (OED).
From resultative predicate to event-modifier
(16) back off/away, brazen out, cheer up, clean up/out/off/away, clear up/out/ off/away, crack up, free up, gloss over, open up/out, parcel out, pretty up, round up/off.
Although one may quibble over the classification of individual items (e.g. is hammer out a conversion of the noun hammer and therefore to be classified as a ‘means’ verb, or was the particle-verb built on an already existing conversion of that noun, with the verb hammer the direct descendant of OE hamerian, hamorian “to hammer” with loss of its derivational morphology, hence a ‘manner’ verb? etc.), the overall tendencies are clear. Because it is the particle that provides the template on which the construction is built rather than the verb, we find many combinations with verbs that are never or rarely used without the particle. It seems that some new verbs in turn derive from the particle-verb rather than the other way around. The OED lists pretty up first (first attestation: 1916), while pretty used as a verb on its own is not attested until 1953. These characteristics of verb, particle and object resulting from the semantics of the R-LCS template activated by the presence of the particle are not restricted to English particle-verbs, but have also been reported of German and Dutch particle-verbs (e.g. Booij 2002; Lipka 1972; Lüdeling 2001; Neeleman & Weerman 1993), including the proliferation of verbs as in (16) that lexicalize the change-of-state. This is all the more remarkable because the morphology of those languages does not allow such zero-conversions as readily as Modern English. Such conversions are made possible by the central role of the changeof-state W, and by the fact that the R-LCS template creates a complex event even before any of the variables have been filled in; ‘CAUSE’, ‘ACT’, ‘BECOME’ are already there by default. The prototypical expression of the R-LCS template is not in fact the particle-verb, but the secondary predicate construction, which we will show in the next section to be the origin of particle-verbs: particles are grammaticalized predicates.
3. Particles as grammaticalized predicates 3.1 The syntax and semantics of secondary predicates Secondary predicates (in traditional grammars referred to as ‘object complements’ (Quirk & Greenbaum 1973) or ‘object attributes’ (Aarts & Aarts 1982))
89
90
Bettelou Los
are syntactically-embedded predicates denoting the result of the action of the verb. They usually contain a ‘light verb’ (one of the group of verbs outlined under (13) above) and, prototypically, an adjective phrase as predicate (as in (17a–b)). PPs (as in (17c)) can also function as predicate.6 (17) a. He made his papers [PREDavailable on the internet]. b. He pushed the door [PREDopen]. c. She put the incriminating documents [PREDinto her briefcase].
The predicate and the preceding NP are in a subject-predicate relation: as a result of his action, his papers are available on the internet (17a); the door is open (17b); the incriminating documents are in her briefcase (17c). The semantics of the construction can be expressed by the R-LCS discussed above; the R-LCS of 17(b), for instance, is (18): (18) [CAUSE[ACT (x)], BECOME[open(door)]], BY[pushing(x)]]
The construction exhibits the same pecularities as the particle-verb: the result is always telic, there are transitivization effects as in (19a), and there is the phenomenon of the ‘unselected object’ as in (19b): (19) a. I’ve worked my fingers to the bone. b. They drank the pub dry.
Again, we find the verbs outlined above in (13)–(16), witness these real-life examples collected by Rappaport Hovav & Levin (2001). (20a) shows a ‘means’ verb and an adjective phrase, (20b) a ‘manner’ verb and a prepositional phrase, and (20c) a ‘lexicalized predicate verb’ with an adjective phrase: (20) a.
Last night, the dog poked me [PREDawake] every hour to go outside. (The Toronto Sun. 27 Nov. 1994, p. 6) b. Sudse cooked them all [PRED into a premature death] with her wild food. (P. Chute. 1987. Castine. Garden City, NY: Doubleday, p. 78) c. She might employ it [her body] as a weapon — fall forward and flatten me [PREDwafer-thin]. (Delia Ephron. 2000. Big City Eyes. New York: Bantam, p. 92)
In spite of the semantic and syntactic similarities between such secondary predicates and particles, there is a clear word order difference. As we saw above
6.Nouns cannot be resultative predicates because such constructions require stage-level predicates, whereas nominal predicates are always individual-level predicates (see Hoekstra 1988).
From resultative predicate to event-modifier
in (1a–b), here repeated as (21a–b), the object of a particle-verb can occur either after or before the particle if it is a full NP; pronominal objects can only occur after the particle if they have heavy stress (21c–e).7 (21) a. b. c. d. e.
He threw away the remains of his dinner. He threw the remains of his dinner away. He threw them away. *He threw away them. ‘If you force your confidence upon me, Mr. Headstone, I’ll give up every word of it. Mind! Take notice. I’ll give it up, and I’ll give up you. I will!’ (Dickens. [1865] 1919. Our Mutual Friend. London: Dent, p. 673)
By contrast, secondary predicate constructions only allow the order as in (21a), with the NP after the particle, with long, complex NPs, as in (22b). The order V-NP-Predicate, then, is the normal order (as in the examples in (20) and in (22a)), whereas the order V-Predicate-NP in (22b) is the derived, special order: (22) a.
She stuffed all the documents containing incriminating evidence [PREDinto her briefcase]. b. She stuffed [PREDinto her briefcase] all the documents containing incriminating evidence.
This special order is actually the normal one for particle-verbs, not just because they allow this order with any non-pronominal NP, regardless of its length or complexity, but also, as we see in Tables 1 and 2 below, because this is also the most frequent order for particle-verbs. These findings are further confirmed by experiments in speech production (Dehé 2002). We conclude that particle-verbs, although still expressing the resultative LCS outlined in the previous section, have a syntax of their own. 3.2 Particles as grammaticalized predicates Resultative predicates may grammaticalize into set phrases in which the predicate has come to be associated very closely with the verb. When they do,
7.The only exceptions are particle-verbs that have become highly idiomatic: (i) Shut up shop (but not *shut shop up). (ii) Let your hair down (let down your hair is only possible with the literal meaning). (iii) He whistled his heart out (but not *he whistled out his heart). He yelled his head off (but not *he yelled off his head). See Bolinger (1971: 121); Den Dikken (1995: 93); Fraser (1976: 19).
91
92
Bettelou Los
Table 1.V-full NP-particle and V-particle-full NP orders in the Helsinki Corpus for on, forth and out. period
V NP part
V part NP
Totals
1150–1250 1250–1350 1350–1420 1420–1500 1500–1570 1570–1640 1640–1710
5 1 1 0 4 7 9
21 38 54 53 70 84 66
27 39 55 53 74 91 75
Table 2.V-full NP-particle and V-particle-full NP in the Microconcord Corpus for out. period
V NP part
V part NP
Totals
20thC
34
349
383
APs (as in (23a)) tend to resist modification by adverbs like very, and PPs (as in (23b)) do not easily allow further modification of the embedded NP-complement (Claridge 2000: 44,76), all pointing to an increasing loss of phrasal status, i.e. an XP on its way to grammaticalize into an X0. (23) a.
break/blow/blast/cut/fling/push/rake/whisk open, cut/stop short, bleach white, blow/keep/make/sift clear, put straight, let/set free, think fit, cast/let/pry/shake/wrestle loose, strip naked etc. b. bring to light, put in execution, take in hand, call to mind, call in question, take into consideration, stand in need, etc.
The items under (24) show that this process may lead to decategorialization: they started out as PP predicates, but the P on has been reduced to a- and many of these items are now analysed as adjectives (A0) rather than as PPs. The diachronic PPs away, from onweg “on (the) way”, and together have made it into the ‘particle’-category (thereby creating difficulties for attempts to subsume the category ‘particle’ under that or preposition, as they lack the typical prepositional property of taking NP complements: *away the door):
From resultative predicate to event-modifier
(24) carry aloft (< on loft), set alight (< on light), take apart (< on part), put awry (< on wry), carry around (< on round), keep asunder (< on sunder), set afoot (< on foot) etc.8
In some cases, the synchronic categorial status is in doubt — particle? adjective? adverb? — particularly if the nouns contained in these PPs (wry “tortuous movement”, loft “sky”) are no longer recoverable. The interesting point here is that items as in (23–24) tend to exhibit the ‘particle’ syntax outlined in the previous section the more grammaticalized they are, witness the corpus-examples presented in Claridge (2000: 67, 76–77): (25) I suspected something, and one day I decided to break open her drawer. (BNC GWH 1486; from Claridge 2000: 67) (26) More ominously its size and capacity also took into consideration future military needs. (BNC AR0 521; from Claridge 2000: 77)
Her drawer in (25) and future military needs in (26) do not constitute particularly long or complex NPs, and would not qualify for extraposition in a genuine secondary predicate construction. The fact that we find the ‘particle’ order with set phrases whose descent from secondary predicates is uncontroversial supports an analysis of particles as grammaticalized predicates. The observation that particles share many characteristics with predicates is not new (e.g. Anthony 1953: 86, Bolinger 1971: 37ff, Fraser 1965: 82ff, and Legum 1968: 55ff). The parallels between the two structures are semantic (resultative meaning), syntactic (same verbs, same word orders) and intonational,9 and have led some scholars to posit a predicate analysis for all particles (e.g. Den Dikken 1995, Grewendorf 1990, von Stechow 1993: 70–71,). The existence of a ‘particle syntax’ and of formations like get-at-able, come-at-able,
8.If we include nautical language we may add: abaft, abeam, aboard, aft, aloft, amidships, aport, ashore, astarboard, astern, overboard, with clear ‘particle syntax’, e.g.: (i) They threw overboard the ballast. (ii) He hauled aloft the pennon. (iii) They set athwart the billets. 9.It has been noticed, for German at least, that the intonation patterns for particles differ from regular adverbs: er versuchte mitzusprechen versus er versuchte laut zu sprechen (Lipka 1972: 20, quoting Hundsnurscher (1968: 6ff) who is in turn quoting Admoni (1966: 51–53, 1970: 49). The same point is made in Winkler 1997: 303, quoted in Lüdeling 1999: 41, who notes the ambiguity of Bill hat den Laden leer gekauft: one reading is Bill bought the shop in an empty condition (depictive) versus Bill bought everything in the shop, so that it was empty (resultative). Winkler finds that particle constructions are stressed like resultatives.
93
94
Bettelou Los
which point at (at least some) particle-verbs having word status at some level, makes an analysis of particles as full-blown secondary predicates unlikely. What is clear from these differences and similarities, however, is the strong suggestion that particles started out as secondary predicates, but grammaticalized into a special category at some point.
4. On, forth and away 4.1 Leaving the resultative LCS Although their syntax may differ, the various semantic effects of particles are surprisingly uniform throughout the Germanic languages. In the light of this uniformity, Dehé et al. (2002: 13–16) propose the following tentative taxonomy of the more productive particle uses in English, Dutch and German, and note that nearly all of these uses are resultative, that is, encode the R-LCS: (27) A. Spatial particles: pull a thread out B. Aspectual particles: i. chew up the meat, think the problem through ii. fight on, sing along, type away C. Other non-spatial particles: i. work off a debt, sleep off a sickness ii. tell/tick someone off (in their ‘castigatory’ senses) iii. Germ. anstarren “stare at”, Du. toespreken “speak to”.
Of this list, A, B (i) and C (i) are resultative; C (ii), however, has become atelic, witness its compatibility with durative adverbials like for ten minutes, and represents an activity rather than an accomplishment or achievement. The focus has apparently shifted from the actual change of state to the state of affairs pertaining after the change of state. This ‘castigatory’ use is recent (the OED’s first attestation of tell off dates from 1919 and that of tick off from 1915). Note that there is no concomitant change in the syntax of such particle-verbs. Type C (iii) is not resultative either. It is no longer found as a particle-verb in English, although it is in OE: he cwæþ him to ‘he spoke to him’, literally ‘he spoke him to’. These ‘postposition’ particle-verbs are the only group of particleverbs for which an analysis as a grammaticalized resultative predicate is not likely. In Dutch and German, the object of these particle-verbs probably represents a reanalysis of the object of the original PP, which might explain why
From resultative predicate to event-modifier
some of them have dative case, as in the OE example or as in German jemandem zulachen “smile at someone (dative)”, whereas the NP in resultative constructions always has accusative case. I do not know of any systematic treatment of the relationship between the case of the P and of the objects of these ‘postposition’ particle-verbs, apart from Stiebels (1996) for German. Goh (2000) for OE is a missed opportunity, as his hypothesis is based on incomplete data.10 The ‘postposition’ particle-verbs have joined fully in the ‘particle syntax’ in those languages that still have them. In Dutch, particles display a special syntax in that they may remain adjacent to the verb in Verb Raising constructions, unlike genuine secondary predicates (for details see Booij 2002:206). For German, ‘particle syntax’ works the other way round: particles insist on strict adjacency to the verb, whereas genuine secondary predicates may, but need not, be adjacent (Dehé et al. 2002: 5). As the analysis such postpositions should receive in OE is still unclear (particles or genuine postpositions? See Mitchell 1985: §1080), and Present-day English has lost them altogether, we will not discuss them here. The third class of particles that do not encode the R-LCS is B(ii), which is also the only class of particles that is truly detransitivizing: the event-modifying ‘argbloc’ particles (Jackendoff 2002; McIntyre 2001, 2002). In English the most notable examples are on, around and away. They will be the subject of my next section. 4.2 Argument-blocking particles Although the Resultative LCS appears to hold for a wide range of particle-verbs in English, not all particle-verb constructions are resultative. There is a particular group of particles which may create durative, not telic, events. When they do, they appear to block the object argument (examples (28a–b) from McIntyre 2001): (28) a. She typed (*her essays) on for hours. b. They played (*a silly game) around. c. She told him ‘to report away’, that she was not afraid. (OED, 1883 Pall Mall G. 27 Sept. 10; cf. ‘report (*the event) away’)
This type of argument blocking is not due to a possible clash between the ingressive/durative semantics of the particles and the ability of direct objects to
10.Goh does not realize that the accusative has been left out of Mitchell’s lists of verbal rection from which he takes his data (Mitchell 1985: §1092), presumably because the accusative case for direct objects is so obvious and universal; see Mitchell’s rider to the lists in §1091).
95
96
Bettelou Los
bring about a telic reading: if the particle is replaced by another expression conveying similar information the result is acceptable, witness she typed essays without stopping (McIntyre 2001: 135). It is the particle that is blocking the object here. McIntyre concludes that one cannot telicise a VP in the absence of a theme, which means that particles cannot add telicity if there is no object11 present (McIntyre 2001: 156–157). In our terms, the addition of an object in sentences (28) would activate the R-LCS and add telicity, which is why it must be omitted, particularly because these particles also have R-LCS uses, as in (29): (29) a. The police moved the bystanders on. b. He really messed us around through his unreliability. (McIntyre 2001) c. She locked the money away.
McIntyre proposes a diachronic link between particles like (28) and those in (29). Although all three particles in (29) encode the R-LCS, they do not indicate any particular direction; McIntyre defines the semantic content of around here in negative terms as “not towards” (2001:147), and the other two are also best described negatively as “not in the same position”. The difference between (28) and (29) is not only that the particles in (28) refer to movement in time rather than in space, but that they have become event-modifiers rather than describing the path followed by an entity. The semantic content of aimlessness and ineffectuality of around in (29) is transferred to the event itself and not to an object. The fact that particles as in (28) have left the R-LCS is also clear from the extreme productivity of these event-modifying particles, and from the fact that they are not restricted to typical R-LCS verbs as in (13)–(16): semantically, the verb is a real verb, and not a manner adjunct, as it is in the R-LCS. The existence of event-modifying particles as in (28), then, does not invalidate our hypothesis that the particle-verb construction ultimately derives from a resultative construction. As this use of around and away appears to be comparatively recent,12 I will concentrate on on and on forth, which appeared to be set on the same road towards event-modification when it became obsolete.
11.Or, of course, a subject that originates as an object in deep structure, as in the case of unaccusative verbs. 12.The OED does not distinguish the event-modifying use of these particles as a special category. Of the two examples that appear to show event-modifying around, the earliest is We generally ‘boarded around’ (1869 S. Bowles Our New West viii. 170). The other one is I shall be just eating around (1927 E. Wallace Hand of Power xliv. 215). For away we find While Pan melodious pipes away (1737 M. Green Spleen) and Scream away if you like it (1821 Scott Kenilw. xxxiii).
From resultative predicate to event-modifier
4.3 From resultative to durative to event-modifying particles 4.3.1 Forth Forth is from earliest times primarily directional (“forwards”), although it did develop a positional sense “out” as well, as in stand forth. As a resultative predicate, its basic meaning was out, as in (30), from the ENE part of the Helsinki Corpus (for which see Kytö 1993): (30) Lettyng you to understand that my Lord Archbishop sent one servant of his unto my son William, chardging him in the Kyngs name to sette in the tenaunts agayne (…). But my sone kepes them forth as yet, and therfor I trow my lord Archbishop will compleane of my son and you (Cepriv1. Stapleton, 1839: 170)
As resultative predicates do, forth evolved a number of specialized meanings. An example is meaning 1b in the OED: “Expressing promptitude or eagerness for action” as in to set someone forth “to urge forward”, or to make oneself forth “to bestir oneself, prepare”, or 5 “Forward, into view” (e.g. bring/come/show/put forth), or 6 “Away or out from a place of origin, residence, or sojourn” as in be forth of town (c1500) “be out of town”. The verbs here conform to the four groups set out in (13)–(16) above, typical for the R-LCS. The human propensity to talk of time in spatial terms (the “time=space” metaphor, see Lakoff & Johnson 1980) gave rise to the durative meaning, OED 3: “of extent in time,” “expressing continuity or progressiveness of action”. The earliest uses have objects (transitive verbs) or subjects that are deep structure objects (unaccusatives): (31) Heald forð tela niwe sibbe. Beowulf ll. 948–949 keep forth truly new friendship “truly keep on this new friendship”
To interpret forth here as a change-of-state predicate, we must make the inference that its sense of “keep something going” negates a presupposition that it might end (see also Lieber & Baayen’s 1997: 804 discussion of verbs like stay). These durative uses are seen to move out of the R-LCS altogether in EME; (32)–(33) are both from the Helsinski Corpus. (32) ah þt ladliche beast leafeð & lest forð & þe of-þunchunge þrof longe þrefter. (Cmhali; Hali Meidenhad, ed. Furnivall 1922: 34) “but that loathsome beast remains and lasts on; and the disgust at it long after”
97
98
Bettelou Los
(33) & he læŠ forð alswa he ær dude (Cmrood; The History of the Holy Rood Tree, ed. Napier 1894: 32) and he lay forth as he before did “he continued to lie there as he had done before”
Crucially, last and lie are unergatives rather than unaccusatives, and unergatives used in resultative constructions invariably acquire objects (see Section 2.2.2). The fact that we do not find objects here is further evidence that this use of the particle forth has left the R-LCS and has become an event-modifying particle like the ones in (28). Forth became archaic in the eighteenth century. Its durative function was taken over by on, its directional function by out. 4.3.2 On The earlier directional particle uses of on appear in the OED as sense 6: “Towards something in the way of approach; approaching in space, time, or condition,” with unaccusatives like go or come, from OE onwards, and sense 9a, “Onward, forward, in space or time”. On as a change-of-state predicate with R-LCS verbs (i.e. the four groups outlined in (13)–(16)) still survives with clearly directional uses as in (34), from the ENE part of the Helsinki Corpus. (34) But the Sea, as before, came rowling on, and without reverence both wet and dash’d him. (Cehist3b; John Milton, The History of Britain (1670), X,281 (ed. Krapp 1932))
Examples of purely durative use, as in (35), can be found in EME in the Katherine-group, but not much outside it until much later: hold on “continue”, fon on “carry on”, on and on. (35) He heold on to herien his heaðne maumez wið mislice lakes. (Katherine 25, l. 1; d’Ardenne 1977) he held on to worship his heathen idols with various offerings “he continued to worship his heathen idols with various offerings”
The object here is not an NP but a to-infinitival clause which makes a resultative interpretation unlikely. Hold on here appears to be a lexical unit, an aspectualizer (in terms of Brinton 1988) synonymous with PE continue. Similar ‘phrasal verb’ combinations with to-infinitival complements appear in the Katherinegroup, with inchoative senses, which probably developed not from directional but from positional uses of on: fon (lit. “take hold of”) on “begin”, bearst (“burst”) on “begin”, take on “begin”. The inchoatives are not found much outside the Katherine-group and have not survived.
From resultative predicate to event-modifier
Crucially, these uses of on, although event-modifying, are clearly restricted to particular combinations, unlike the very productive use of on as an ‘argument-blocking’ particle, and for that reason examples like (35) cannot be taken as evidence that the ‘argument-blocking’ on was already in existence by the early thirteenth century. Better evidence are the more productive uses of durative on exemplified in (36), ME, and (37), ENE, both from the Helsinki Corpus: (36) Telle on þi wronge, quoþ Tholomay. (Cmalisau; Kyng Alisaunder I,407 (ed. Smithers 1952)) (37) They seeing her to fall into merry humors, whetted her on in merriment as much as they could. (Cefict2b; Jack of Newbury p. 79 (ed. Lawlis 1952))
On as a change-of-state predicate is unproblematic in the directional sense of (34), but in the extended-time on of (36)–(37) we have the same dilemma as in the Beowulf example of (31): we have to assume that the assumption is that as a result of the telling and the goading, the situation changes from terminating to continuing. Example (36) approaches the ‘argument-blocking’ sense of PE on closest, as on appears to be event-modifying here rather than in some predication-relationship with the object. McIntyre (2001) argues that the presence of an object in examples such as (28) above would force speakers to try to predicate the particle of the object, and block the intended interpretation of the particle as an event-modifier. The presence of an object in example (36) probably shows that we are seeing here the first beginnings of on as a productive event-modifier. In (38)–(40), from the OED, the object has been dropped: (38) Now telle on, Roger, looke that it be good. (c1386 Chaucer Cook’s Prol. 22) (39) Now say on Diggon. (1579 Spenser Sheph. Cal. Sept. 55) (40) Speculate on! (1795 Burke Regic. Peace iv. Wks. IX. 26)
5. Conclusion We have presented an analysis here of English particle-verbs that assumes these verbs to be grammaticalizations of syntactic secondary predicate constructions. We have seen that the semantics of such secondary predicates, here presented in terms of the Lexical Conceptual Structure of resultatives (the R-LCS), account for the types of verbs that we typically find in such particle-verb combinations, for the transitivizing effect of the particles and for certain properties of the particle itself.
99
100 Bettelou Los
Although resultative constructions proved to be a fruitful starting-point, not all particle-verbs fit the R-LCS profile. We must assume a different origin and analysis for the ‘postposition’ particle-verbs in OE, German and Dutch. The extremely productive event-modifying use of the PE particles on, forth, away and around does not fit the R-LCS profile either, but this must be a later development away from the R-LCS, with these originally directional particles developing first durative, and then event-modifying uses.
References Aarts, F. & J. Aarts. 1982. English Syntactic Structures. Oxford: Pergamon. Admoni, V. G. 1966. Stroj sovremennogovešcˇ nemckogo jazyka/Der deutsche Sprachbau. 2nd edition. Moscow & Leningrad: Prosvescˇenie. Third edition 1970. Munich: Beck. Anthony, E. M. Jr. 1953. Test frames for structures with UP in modern American English. Ann Arbor: University Microfilms. Bolinger, D. 1971. The Phrasal Verb in English. Cambridge, Mass.: Harvard University Press. Booij, G. E. 2002. The Morphology of Dutch. Oxford: Oxford University Press. Brinton, L. J. 1988. The development of English aspectual systems: Aspectualizers and postverbal particles. Cambridge: Cambridge University Press. Cappelle, B. 2002. “And up it rises: Particle preposing in English”. Dehé et al. 2002. 43–66. Claridge, C. 2000. Multi-word verbs in early Modern English: A corpus-based study. (= Language and Computers, 32). Amsterdam & Atlanta: Rodopi. Dehé, N. 2002. Particle verbs in English: Syntax, information structure and intonation (= Linguistik Aktuell/Linguistics Today 59). Amsterdam & Philadelphia: John Benjamins. ———, R. Jackendoff, A. McIntyre & S. Urban, eds. 2002. Verb-Particle Explorations (= Interface Explorations, 1). Berlin & New York: Mouton de Gruyter. Denison, D. 1981. Aspects of the history of English group-verbs, with particular attention to the syntax of the Ormulum. Ph.D. dissertation, Oxford University. Dikken, M. den. 1995. Particles: On the syntax of verb-particle, triadic, and causative constructions. Oxford: Oxford University Press. Fraser, B. 1976. The verb-particle combination in English. New York: Academic Press. Goh, Gwang-Yoon. 2000. “Relative Obliqueness and the contribution of nonheads in the subcategorization of Old English compound verbs”. English Language and Linguistics 4.13–36. Grewendorf, G. 1990. “Verbbewegung und Negation im Deutschen”. Groninger Arbeiten zur Germanischen Linguistik 30.57–125. Haegeman, L. 1994. Introduction to Government and Binding Theory, 2nd edition. Oxford: Blackwell. Hampe, B. 2000. “Facing up to the meaning of ‘Face up to’”. Constructions in Cognitive Linguistics ed. by A. Foolen & F. van der Leek, 81–101. Amsterdam & Philadelphia: John Benjamins. Hoekstra, T. 1988. “Small Clause results”. Lingua 74.101–139.
From resultative predicate to event-modifier
Hundschnurscher, F. 1968. Das System der Partikelverben mit AUS in der Gegenwartsprache. Ph.D. dissertation, University of Tübingen. Jackendoff, R. 1990. Semantic Structures. Cambridge, Mass.: MIT Press. ———. 2002. “English particle constructions, the lexicon, and the autonomy of syntax”. Dehé et al. 2002. 67–93. Kennedy, A. G. 1920. The Modern English verb-adverb combination. (= Language and Literature, I:1). Stanford: Stanford University Publications. Kytö, M., ed. 1993. Manual to the diachronic part of the Helsinki Corpus of English Texts: Coding conventions and lists of source texts, 2nd edition. Helsinki: University of Helsinki, English Department. Lakoff, G. & M. Johnson. 1980. Metaphors we live by. Chicago & London: University of Chicago Press. Legum, St. E. 1968. “The verb-particle constructions in English: basic or derived?” Papers from the Fourth Regional Meeting of the Chicago Linguistic Society ed. by B. J. Darden, Ch.-J. N. Bailey & A. Davidson, 50–62. Chicago: Chicago University Press. Lieber, R. & H. Baayen. 1997. “A semantic principle of auxiliary selection in Dutch”. Natural Language and Linguistic Theory 15.789–845. Lipka, L. 1972. Semantic structure and word formation: Verb-particle constructions in contemporary English. Munich: Wilhelm Fink. Lüdeling, A. 2001. On particle verbs and similar constructions in German. Chicago: University of Chicago Press. McIntyre, A. 2001. “Argument blockages induced by verb particles in English and German: Event modification and secondary predication”. Structural Aspects of Semantically Complex Verbs ed. by N. Dehé & A. Wanner, 131–164. Berlin: Peter Lang. ———. 2002. “Idiosyncracy in particle verbs”. Dehé et al. 2002. 95–118. Mitchell, B. 1985. Old English syntax, 2 vols. Oxford: Clarendon Press. Neeleman, A. & F. Weerman. 1993. “The balance between syntax and morphology: Dutch particles and resultatives”. Natural Language & Linguistic Theory 11.433–447. The Oxford English Dictionary on CD-ROM. 1989. 2nd edition ed. by John A. Simpson & Edmund S. C. Weiner. Oxford: Oxford University Press. Quirk, R. & S. Greenbaum. 1973. A University Grammar of English. London: Longman. Rappaport Hovav, M. & B. Levin. 2001. “An event structure account of English resultatives”. Language 77.766–797. Spencer, A. & Zaretskaya, M. 1998. “Verb prefixation in Russian as lexical subordination”. Linguistics 36.1–39. Stiebels, B. 1996. Lexikalische Argumente und Adjunkte: Zum semantischen Beitrag von verbalen Präfixen und Partikeln. Berlin: Akademie Verlag. Visser, F. Th. 1963–73. An Historical Syntax of the English Language, Vols. 1–3b. Leiden: E. J. Brill. von Stechow, A. 1993. “Die Aufgaben der Syntax”. Syntax. Ein internationales Handbuch zeitgenössischer Forschung ed. by J. Jacobs, A. von Stechow, W. Sternefeld & Th. Vennemann, Vol. 1, 1–88. Berlin: Walter de Gruyter. Winkler, S. 1997. Focus and secondary predication. Berlin & New York: Mouton de Gruyter.
101
102 Bettelou Los
Text editions d’Ardenne, S. R. T. O. 1977. The Katherine Group. Paris: Société d’Edition “Les Belles Lettres”. Furnivall, F. J., ed. [1866] 1922. Hali Meidenhad. London: EETS, Original Series 18. Klaeber, F., ed. [1922] 1950. Beowulf and the fight at Finnsburg, 3rd edition. Lexington, Mass.: D. C. Heath & Co. Krapp, G. V., ed. 1932. The Works of John Milton, Vol. X: the History of Britain, that part especially now call’d England (1670). New York: Columbia University Press. Lawlis, M. E., ed. 1952. The novels of Thomas Deloney: Jack of Newbury (1619). Bloomington: Indiana University Press. Napier, A. S., ed. 1894. History of the Holy Rood Tree. London: EETS, Original Series 103. Smithers, G. V., ed. 1952. Kyng Alisaunder, Vol. 1. London: EETS, Original Series 227. Stapleton, T., ed. 1839. The Plumpton Correspondence. A Series of Letters, Chiefly Domestick, Written In The Reigns Of Edward IV., Richard III., Henry VII. And Henry VIII. London: Camden Society.
Family values April McMahon and Robert McMahon University of Sheffield
1.
Family values
The ex-Prime Minister John Major is strongly associated with ideas of ‘family values Conservatism’; and the connection of ‘family values’ and ‘conservative’ seems an entirely natural one, bringing with it ideas of tradition and a return to a gentler and better age. Our gloss is a rather different one: the family values discussed and developed below centre on the adoption of a quantitative, numerical approach (the values) to language classification and subgrouping (hence the families). Far from being a conservative pursuit, this will involve going against traditional ideas in various respects, since language classification, with a few notable exceptions, has not tended to embrace the quantitative. We shall argue, however, that language classification urgently needs to explore quantification, both for reasons internal to the discipline itself, and because historical linguists are increasingly interacting with colleagues, notably in archaeology and genetics, who place a considerable premium on testable and repeatable hypotheses. This does not, of course, mean that quantification and computational methods are the whole story; but developments elsewhere in linguistics might indicate that they have an essential role to play. Many colleagues would surely agree that one of the biggest steps forward in linguistics over the past 50 years has been the development of sociolinguistics, with the variationist paradigm which crucially does involve quantification (see Labov 1994, 2001). As it turns out, however, not everyone would agree that this is actually relevant to linguists — or at least, to linguistic theory: on the contrary, Smith (1989: 180) argues that: It is obvious that different communities exhibit variation in their speech…it is equally clear that children don’t speak the same way as their grandparents, that males and females are not necessarily identical in their linguistic abilities, and so on. In short, any social parameter whatsoever may be the locus for some
104 April McMahon and Robert McMahon
linguistic difference. Unfortunately nothing of interest to linguistic theory follows from this, so quantifying the difference is irrelevant to linguistics even though it may be of interest to the sociologist if it gives him or her a recognition criterion for some socially relevant variable.
Colleagues interested in the interplay of variation and change may be surprised to find they must be sociologists rather than linguists, and more surprised still by Smith’s continued argument that “In general the numbers game is irrelevant to the linguist…because the nature of his hypotheses do not usually lend themselves to quantitative testing in virtue of being a mental/psychological rather than a physical discipline…” (1989: 185). If being “a mental/psychological rather than a physical discipline” exempts linguistic theory from the usual requirements of the scientific method for repeatability, and statistical testability, and evaluation, then we might wish to ask whether historical linguistics can fit under that theoretical umbrella. For historical linguists, the essential problem is that variation very clearly does relate to change, and the variationist paradigm has been absolutely instrumental in showing how and why the two go together. Quantitative work of this kind has contributed to our theorising about language variation and change in two ways — both by testing and properly establishing correlations which had already been suspected in earlier methodologies, such as traditional dialectology; and by revealing others which had not been suspected at all, but which are no less interesting by being uncovered only by close consideration of the data at a numerical level. It now seems self-evident to say that variation is not just messy and problematic, and that looking at the behaviour and attitudes of whole communities, or at least meaningful cross-sections of them, is important and illuminating for linguists who are interested in change and the motivations for it. The same is arguably true of corpus work: counting and accumulating numbers alone is not the point, but establishing trends over time in corpus data can actually confirm some things we thought we already knew on how change spreads, and might tell us more which, again, we had not suspected. Language classification now faces a choice: it can follow this adoption of quantitative principles and methods which has proved so successful in other areas of historical linguistics, or can set itself aside from these advances in the way Smith seems to advocate. In the next section we shall outline two reasons why we should follow the first route rather than the second.
Family values 105
2. The why and how of quantification There are two main arguments for accepting and exploring quantitative methods in language classification; these can only be outlined here, but are discussed in greater depth and detail in McMahon and McMahon (2003). First, within historical linguistics itself, we are not currently in a position to evaluate hypotheses of relatedness, or the methods used to reach those hypotheses, in a neutral and objective way; and second, the increasing collaboration between historical linguists and colleagues in other, more clearly quantitative, disciplines means we are increasingly out of line with our colleagues. Turning first to arguments from within historical linguistics, many comparativists do of course use the Comparative Method in developing hypotheses of relatedness (Fox 1995, Durie and Ross 1996): but although the Comparative Method might represent the gold standard for comparative work, it is problematic in various respects. First, contact-induced change is a notorious difficulty. Generally, basic vocabulary is prioritised on the grounds that it is less likely to be borrowed; but loans can be particularly hard to detect when they are from related languages, and of course contact can affect the grammar much more generally, as studies of pidginisation, creolisation, convergence and language mixing have demonstrated. Secondly, there is the absolutely central problem that the Comparative Method depends on case law, and is typically taught in association with its results for a particular family. It is true that principles, followed by many careful practitioners of the model, are set out in Hoenigswald (1960): but it is equally true that many historical linguists mistrust the idea of a general theory for comparison and reconstruction, arguing instead for reliance on an individual linguist’s in-depth knowledge of a particular language group. The method also depends on the availability of a great deal of evidence (certainly from the present day, and ideally going back a considerable distance in time). Unfortunately, we are not always in that privileged position, and as language endangerment accelerates, the problem will become more acute — so the question arises as to how we practise comparison on less bountifully attested languages, and how we evaluate hypotheses made on the basis of less complete data. More generally, because the Comparative Method is based on personal knowledge of a language group, it is also inevitably subject to interference from individual linguists’ opinions, so that we cannot guarantee reaching the same conclusions from the same data considered by different linguists. This problem of repeatability might not be such a major problem if we at least all agreed on the method of first choice; but in fact some exceptionally controversial methods
106 April McMahon and Robert McMahon
have been proposed as alternatives to the Comparative Method. The obvious case in point here is Greenberg’s (1987) mass comparison, especially as used in his highly problematic tripartite classification of native American languages. Many historical linguists would argue that mass comparison is simply too lax in its criteria for determining what counts as a match between languages for us to place any confidence in its results (Campbell 1988, Ringe 1999, McMahon and McMahon 1995); but at the moment, we lack any neutral means of evaluation to apply to our assessment of the relative merits of the Comparative Method as opposed to mass comparison, or indeed any potential third method. Moving on to arguments beyond the discipline itself, another very good reason for introducing quantification into comparative linguistics is provided by the so-called ‘new synthesis’, which attempts to unify data and arguments from genetics, archaeology and linguistics in establishing the histories of human populations (Renfrew 1999; Cavalli-Sforza 2000; Sims-Williams 1998). Of course, as Cavalli-Sforza notes “The correlation between genes and languages cannot be perfect…”, because both languages and genes can be replaced independently; but “Nevertheless… the correlation between genes and languages remains positive and statistically significant” (Cavalli-Sforza 2000: 167). This is a commonplace for geneticists; Barbujani (1997: 1011), for instance, notes that: Humans do not easily cross language boundaries when choosing a partner. As a consequence, populations separated by such boundaries are somewhat isolated from each other. The genetic consequences may be substantial. In Europe, for example, … it is well known that several inheritable diseases differ, in their incidence, between geographically close but linguistically distant populations…
If we accept that there is a general statistical correlation between genetic and linguistic features, which reflects interesting and investigable parallelism rather than determinism, then comparative linguistics should be central to the development of the ‘new synthesis’. However, for this to be possible, our data must be interpretable and usable by these neighbouring, quantitative disciplines, and we must therefore deal in probabilities and degrees of relatedness. We cannot expect to dissuade archaeologists and geneticists from adopting the apparently more ambitious hypotheses of relatedness generated using multilateral comparison (see Ruhlen 1991), if we do not seek to demonstrate that these results and methods are unreliable. Moreover, if we do not supply numbers of our choosing, we cannot be surprised if archaeologists and geneticists attempt to invent their own. To give just one example, Poloni et al. (1997: 1017–18) adopt the following methodology:
Family values 107
Linguistic distances between pairs of populations were defined as simple dissimilarity indexes…two populations within the same language family are set to a distance of 3 if they belong to different subfamilies; their distance is decreased by 1 for each shared level of classification — up to three shared levels, where their distance is set to 0…a dissimilarity index of 8 was arbitrarily assigned to any pair of populations belonging to different language families.
What this means is that Poloni and her colleagues, urgently requiring some numbers to feed into their computations, have essentially arbitrarily assigned grades of relatedness of 0, 1, 2 and 3 to pairs of languages, with a score of 8 for pairs generally thought to be unrelated. If we as linguists feel that these are crude overgeneralisations, then the onus is very much on us to provide better reasoned alternatives. Not all colleagues may agree that linguists should feel under any obligation to change the way we do linguistics, just because other disciplines are interested in our results: Smith (1989: 185) takes the more insular view that “linguistic theory is not affected by the fact that its subject matter can also be of interest to others: the hydrologist’s theories are not affected by spitting”. For the most part, however, we might hope and expect that historical linguists will want to pursue a policy of disciplinary integration rather than separation. If we believe this kind of integration is desirable, and see the introduction of quantitative approaches as a possible way of solving some of the problems confronting classificatory linguistics, the next obvious question is how we introduce quantitative ideas and methods into what has tended to be an individual and potentially subjective field. There are two possibilities: to design methods de novo, especially tailored to linguistic problems; or to adapt existing methods which have already been used successfully in other disciplines (and the latter is not, of course, a new direction for linguistics: see Sampson’s (1980) interesting discussion of Schleicher’s “biological” approach to language history). Overlapping with this decision is the question of whether we use data and comparisons which are already employed in historical linguistics, or again, whether we attempt to exploit novel data types which might not traditionally be employed in discussions of relatedness and subgrouping. At present, different research groups are pursuing different possibilities: one approach which involves novel data and a specially designed means of comparison, for instance, would be Heggarty’s (2000a, b) work on quantifying phonetic similarity, at present used mainly in the formalisation of relative degrees of similarity among languages and dialects already known to be related, but potentially generalisable to less clear cases. However, in this paper we shall concentrate on two different
108 April McMahon and Robert McMahon
research groups, each currently attempting to automatise aspects of a preexisting linguistic method. Outlining these approaches, and the differences between them, which partly depend on respective attitudes to contact and borrowing, allows a clearer understanding of what both groups are doing. We shall conclude by showing how these different approaches can also work together to provide greater insight into the history of particular groups, and in this case especially Germanic. The central idea in both cases is that we can start to use computational methods to cast light on some problematic aspects of linguistic history, and to provide some tests of work which has already been done and accepted in the field; potentially, we may also find that in future these methods can be used outside the usual testing ground, which is Indo-European.
3. The Computational Cladistics Project In the Computational Cladistics Project, based at the University of Pennsylvania (http://www.cis.upenn.edu/~histling), Ringe and his collaborators (Warnow, Ringe and Taylor 1996, Ringe, Warnow and Taylor 2002) are essentially attempting to computerise aspects of the Comparative Method. Their methodology is character based: that is, it involves a fixed set of pre-selected features, which may be phonological, morphological or lexical. As (1) shows, characters of all three types are included in the work reported in Ringe, Warnow and Taylor (2002). Although there are many more lexical characters, this bias is purely numerical and reflects the availability of characters of different kinds: in fact, Ringe et al. contend that non-lexical characters provide far better evidence of relatedness, and in fact their program is constructed to ‘fix’ the phonological and some of the morphological characters, automatically rejecting any tree which is incompatible with these (Don Ringe, personal communication). (1) 22 phonological characters 15 morphological characters 333 lexical characters
For each of the characters selected, individual languages may have different values; this can be thought of as a historical equivalent of the familiar sociolinguistic variable, like English (ng) in playing, which may be realised as variants with a velar or alveolar nasal. For each morphological and phonological character (and a selection of lexical ones), Ringe et al. provide (in an extremely helpful appendix to their paper) a list of the different states they have identified
Family values 109
in the set of 24 Indo-European languages they are considering, and a matrix listing the state found in each language. A short example for a small subset of phonological characters can be seen in (2) (character states, for 3 of the 22 phonological characters) and (3) (matrix, for 6 languages only); both are excerpted from Ringe, Warnow and Taylor (2002: 113–115). (2) Subset of phonological characters and their states P1 *p…kw > *kw … kw 1, absent [ancestral]; 2, present; 3, obscured by merger; 4 &c., no evidence P3 ‘ruki’-retraction of *s 1, absent [ancestral]; 2, present; 3, 4, obscured by merger or orthography P7 word-initial *ye- > e1, absent [ancestral]; 2, present (3) Matrix for 3 characters × 6 languages P1 P3 P7 Hittite 4 1 2 Vedic 1 2 1 Old English 1 1 1 Old Irish 2 1 1 Old Persian 1 2 1 Latvian 1 4 1
Morphological characters sometimes have states involving different forms (such as M6, which offers four possibilities for the shape of the thematic optative), or may be more general and systemic (like M1, which has 8 states reflecting the overall organisation of the verb system as having one stem per lexeme, as opposed to a present/preterite contrast, or a contrast of present/subjunctive/ preterite, for instance). As one might expect, there is greater diversity in the number of states per lexical character, and unique states are quite commonly assigned because there is no evidence in the language at issue, or there is no cognate with the same meaning in the database — or less commonly, because the item found is a loan. The main aim of Ringe, Warnow and Taylor (2002) is to assemble evidence for the first-order splitting of Indo-European: they are therefore mainly concerned to identify characters which characterise particular subfamilies, either individually, such as Tocharian, or as groups, such as Italo-Celtic; the character states involved ideally should not be shared by members of any other group.
110
April McMahon and Robert McMahon
The characters selected, of course, are those which have already emerged from prior philological work; hence the claim above that this is effectively a means of computerising aspects of the Comparative Method. As one might expect, therefore, the output of the project is a series of family trees; but the difference is that the software used provides a numerical measure of how good that outcome is, since the tree produced is the one most consistent with the majority of the data. It is true, of course, that the limitations of the family tree model have been appreciated for a considerable time: Bloomfield (1935: 317) notes the “insoluble problem” created for the tree model by conflicting evidence and differential overlap, which led to the development of Schmidt’s (1872) wave model. Nonetheless, the family tree is well established as a tool for modelling and visualising common descent with differentiation in historical linguistics (and, as we shall see below, in biology), and it has the further advantage of being tightly definable in mathematical terms, making it particularly suitable for quantitative methods requiring testability. Technically, Ringe et al. adopt a perfect phylogeny approach — they are looking for a tree with which absolutely all the characters they are examining are compatible. This notion of character compatibility involves the distribution of character states in the tree. Briefly, for a character to be compatible with a particular tree, all the languages sharing a particular state for that character must form a single group or a continuous subgraph. If a single state is shared by languages which must (because of other characters) lie in different, discontinuous regions of the tree, the relevant character is not compatible with that tree. As Ringe, Warnow and Taylor (2002: 76) point out, “A tree with which all characters are compatible is called a perfect phylogeny; … Finding PPs from character data of the leaves of a tree is a known problem in computational cladistics”. The problem is, as it turns out, a notoriously difficult one, but one with an existing algorithmic solution when there is an upper limit on the number of states per character, as is the case here. The Computational Cladistics Project group have produced software implementing the relevant algorithm; in addition, in cases where a perfect phylogeny cannot be found, the software builds a tree which is compatible with as much of the data as possible. Trees can therefore be ranked for consistency, potentially allowing further support for existing, accepted hypotheses, as well as an additional means of evaluating more controversial ones. Ideally, then, Ringe et al. are seeking a perfect phylogeny, a tree compatible with all their 370 characters and 24 languages; but as it turns out, that is rather too much to ask. Where characters are not compatible with a generally acceptable tree,
Family values
Ringe, Warnow and Taylor (2002: 78) suggest two main reasons: either there is a case of undiagnosed parallel development, or a loanword has been missed: …though most words borrowed from foreign languages can be identified as such in a language’s basic vocabulary, there is always the possibility that a few will fail to exhibit the usual diagnostics of loanwords by sheer chance, especially if they were borrowed from closely related languages. That has been known for some time, though there has never been an effective way of dealing with the problem.
In discovering quite where the problem lies, the first step naturally involves identifying the characters which are not compatible with an otherwise good tree; and the main obstacle to the calculation of a perfect phylogeny appears to be Germanic, represented in Ringe et al.’ s database by Old English, Old High German, Gothic and Old Norse. There are two main difficulties with Germanic: first, these languages share states with a wide variety of other groups (including Italic, Celtic, and Baltic). Indeed, of 18 problematic characters, shared by discontinuous sets of languages, Germanic is involved in 16! Furthermore, and partly because of these shared states, Germanic is very difficult indeed to characterise as a unit, and tended to shift around in the tree on different runs of the software. As Ringe, Warnow and Taylor (2002: 88) argue, “The obvious inference is that there is not necessarily anything ‘wrong’ with these characters, but there might be something very peculiar about Germanic”. The next obvious question is what this “something very peculiar” might be; and the answer would appear to involve contact. One possibility is that “… the diversification of the IE family must be modelled at least in part as a network rather than a tree” (Ringe, Warnow and Taylor 2002: 110). On the other hand, Ringe et al. also note that Germanic shares a number of inflectional morphological character states with Balto-Slavic, Indo-Iranian and Greek; “since those are the characters that are the most reliable indicators of genetic descent, it appears that Germanic should be placed in what we are calling the core of the family — the residue after the departure of Anatolian, Tocharian and Italo-Celtic” (2002: 110). Of the relevant lexical characters, Germanic shares states with either this core group, or with Celtic or Italic, suggesting that (Ringe, Warnow and Taylor 2002: 111): …Germanic was originally a near sister of Balto-Slavic and Indo-Iranian…; that at a very early date it lost contact with its more easterly sisters and came into close contact with the languages to the west; and that that contact episode led to extensive vocabulary borrowing at a period before the occurrence in any
111
112
April McMahon and Robert McMahon
of the languages of any distinctive sound changes that would have rendered the borrowings detectable…In sum, it is clear that the development of Germanic exhibits some characteristics which cannot realistically be modelled with a ‘clean’ evolutionary tree, but it is not clear what historical developments have given rise to those anomalies.
Invoking contact as an explanation might initially seem surprising, given Ringe et al.’s robustly restrictive approach to contact-induced change (see McMahon 2002): they regard lexical borrowings as generally detectable, and see nonlexical consequences of contact as indicating discontinuous transmission, which by definition lies outside the normal situation represented in family trees. However, seeing contact as the exception rather than the rule does not mean it cannot justifiably be used as an explanation for the odd behaviour of Germanic: and in fact, the procedure followed by Ringe et al. has the advantage of demonstrating, rather than simply asserting, that Germanic cannot have developed as the straightforward result of tree-like descent, but must have involved contact. Once Germanic is excluded, it is possible to achieve a close to perfect phylogeny which is only inconsistent with 4 characters, all of them lexical: recall that Ringe et al. regard phonological and morphological characters as more reliable than lexical ones in establishing relatedness. This is certainly a good result, though it must be noted that it depends absolutely on the exclusion of languages which can be shown to have substantial borrowing in their histories. The question of borrowing is not the only issue which might be raised in evaluating the Computational Cladistics project. For instance, although the use of non-lexical characters is an extremely promising development, the number of phonological and morphological characters which are actually useful and informative turns out to be very low, and the evidence for first-order branching is sparse and mainly lexical. These findings are also absolutely reliant on prior philological work, because the characters must be identified and coded before involving the software; this means that while the method could be generalised to other language families, there can never be direct comparability (beyond, perhaps, some of the lexical characters) because the diagnostic characters will inevitably be different from family to family. If, as Ringe, Warnow and Taylor (2002: 98) argue, “the higher-order subgrouping of the IE family has remained an unsolved problem for so many generations partly because the evidence is genuinely meagre”, these methods are highly unlikely to resolve outstanding issues at the higher, superfamily level, or for families where historical data is sparser and our confidence in identifying and coding characters is inevitably lower.
Family values
4. Quantitative methods in language classification The approach taken in the Computational Cladistics Project certainly represents one way forward in applying quantitative and computational methods to issues in historical and comparative linguistics; but it is not the only possibility. We turn now to our own ongoing project, Quantitative Methods in Language Classification, based at the University of Sheffield, to illustrate an approach which is rather different, at least in some respects.1 (http://www.shef.ac.uk/ english/language/quantling) First, we take the view that we can learn a great deal from what biologists are already doing. Both population genetics and historical linguistics are working with systems which persist and change over time, and which are susceptible to influence from outside, by admixture from other populations, or other languages. There are even family trees in both disciplines. However, the difference between the methods used to reach conclusions about relatedness could not be more marked. Geneticists incline towards sophisticated computer technology, objective quantitative methodologies, and statistical testing; comparative linguists, on the other hand, have tended to prioritise depth of knowledge of one particular language group on the part of the individual scholar, rather than generalisable techniques which allow the processing of large quantities of data, regardless of region or family. In short, we are therefore investigating the borrowing of biological computer programs, into linguistics. In this respect one can argue for parallels with the work of Ringe’s group, who base their work on algorithms developed elsewhere, this time in computational mathematics. Secondly, and here our assumptions and Ringe’s differ markedly, we believe that the importance of contact-induced change has been vastly under-estimated in at least some traditions of historical linguistics. It is true that in lexicology, and in linguistic geography (Iordan 1970), issues of contact have consistently been recognised and discussed: but in the core areas of comparison and reconstruction, contact has been seen as peripheral at best, and
1.This project (2001–04) is funded by the UK Arts and Humanities Research Board (award number AN6720/APN12536); we gratefully acknowledge their support. We also thank the other members of the project team, Paul Heggarty and Natalia Slaska, for their comments and contributions. April McMahon thanks Christian Kay and the organising team for ICEHL12, and members of the audience at that conference, especially Roger Lass and the ground support team from English Language and Linguistics, Sheffield, for invaluable comments and support. Both authors thank the volume editors, two anonymous reviewers, and Don Ringe, for helpful and informative comments.
113
114
April McMahon and Robert McMahon
the balance is only now being redressed. We believe equally that it is much, much harder to reliably track down borrowings than is sometimes assumed: for instance, Embleton (1986), using basic vocabulary lists which are typically regarded as particularly resistant to borrowing, still found quite significant numbers of loans recorded between, for instance, English and French, or English and the North Germanic languages. It would appear, then, that even where the history is well known and the evidence is plentiful, work in comparative linguistics has been proceeding with borrowings embedded in the supposedly pure and loanproof evidence used (see further McMahon and McMahon 2002). We would argue that borrowing and contact are facts of linguistic life, and that languages which have experienced major borrowing events have just as interesting and legitimate a history as those that have kept themselves to themselves and behaved in a more conventionally tree-like way. Instead of trying to remove all evidence of lexical borrowing, and redefining furtherreaching types of contact-induced change, or indeed arguing that these do not occur, we prefer to search for signals of borrowing, and for ways of using the results of contact in our methods. One point of agreement with Ringe et al. involves the absolute requirement for historical linguists to show why a particular family tree configuration is the right one, or at least the best available. Ringe et al. assess their trees by fit with a pre-selected set of characters; our approach is not character-based, but distance-based. All distance-based methods start by calculating degrees of similarity (or of difference, if you prefer); the resulting matrix of distances between systems is then input to computer programs. On the positive side, distance-based methods allow generalisation across groups, because there is no need to preselect salient features or work with shibboleths; on the other hand, not all aspects of language are suitable for measuring distances and deriving matrices. The method we shall outline here (and which is discussed in more detail in McMahon and McMahon 2003) uses standard Swadesh-type meaning lists; we hope to show that additional information can be derived from already available data by using quantitative approaches. There are undoubtedly difficulties in the use of meaning-lists: as we saw in Section 3, lexical data is generally regarded as inferior to phonological or morphological data in demonstrating relatedness, and is the area most susceptible to contact-induced change. On the other hand, as we also saw in Section 3, lexical data may be the only evidence supporting particular subgroupings, even in wellattested families. Despite problems of translatability across languages, and of potential subjectivity in choosing the ‘unmarked’ translation, meaning lists do
Family values
have the very considerable advantage of being collectable even when available data is sparse; we should therefore at least consider testing the approach. Following Embleton (1986), we use meaning-lists with 200 items. We are currently working with data from Dyen, Kruskal and Black (1992), whose database includes modified Swadesh-type 200-meaning lists for 95 IndoEuropean languages, with their judgements for whether or not any two items were cognate, and a derived distance matrix. We converted this distance matrix into a tree, using three different programs from Felsenstein’s (2001) PHYLIP package, a suite of programs developed to draw and select biological trees. These programs generate all or many of the possible trees, and then select from this set the tree most consistent with the distances in the data matrix; so the process is automatised and the selection criteria are objective. Ringe, Warnow and Taylor (2002: 86) are quite right in noting that the best tree cannot plausibly be identified by exhaustive searching, since the number of possible trees for their 370 characters, or our 200 meanings, will run into many millions. For example, on excluding Germanic from their analyses, Ringe et al. are searching for the best tree consistent with 318 of 322 characters, and calculate the relevant number of sets as 445, 197, 684. However, biologists are quite satisfied with the heuristics included in programs of the PHYLIP type, which exclude at the outset large areas of the tree population which simply cannot be strong contenders; and in any case, one can do further statistical evaluation work in assessing how well a particular tree fits the data, as we shall see below. Although we used three PHYLIP programs (namely Neighbour, Fitch and Kitch), we shall focus here only on the first of these, since its Neighbour Joining approach most accurately reflects the route which would be taken by a linguist drawing a single tree. Each step involves clustering the closest two languages (in terms of percentage similarity), then adding the next closest, and so on. This is a fairly crude but computationally simple procedure, which takes less than 10 minutes to run for the Dyen, Kruskal and Black data on a 700 MHz PC (for comparison, the Ringe et al. program took approximately 8 days with Germanic, and 24 hours after Germanic was excluded). Kitch and Fitch, based on a Maximum Likelihood approach, consider larger populations of trees and might therefore give better results; but at present any benefits seem marginal, since all programs generate several hundred thousand trees, and thus far produce strikingly similar results. Bootstrapping, an additional statistical test involving sequential rerunning of the program with 5% of meanings removed and resampled on each run, shows that all sub-branches are well supported, recurring in all runs. Bootstrapping provides a test of the robustness of the
115
116
April McMahon and Robert McMahon
whole tree by testing all its individual branches, and helps address the criticism that the whole space of possible trees cannot be searched exhaustively. A Neighbour Joining tree, drawn using the Neighbour program (Felsenstein 2001), is shown in Figure 1; although PHYLIP produces star diagrams, or unrooted trees, these can be converted to more conventional family trees by selecting any language as the root, and ‘suspending’ the other subgroups from this. The tree in Figure 1 has been rooted, quite arbitrarily, using Albanian, and has also been redrawn using the program TreeView (Page 1996), which assists display on a personal computer. This use of biological tree selection programs, along with bootstrapping, is helpful in demonstrating the robustness of the agreed IE subfamilies (Romance, Germanic, Celtic, Slavic and so on), and therefore in reinforcing agreed results in the field. However, we can also use these programs to test assumptions about meaning lists. First, it is generally assumed that borrowing will not be a problem for basic vocabulary lists, since loans should have been identified and filtered out; and second, there is a generally unstated assumption that individual items should be contributing equally to the analysis by changing at the same rate. Evidence from initial bootstrap tests argues against both assumptions: although subfamily groups do emerge routinely, certain languages shift within their subgroups, although if all meanings were changing at exactly the same rate, all the reruns should give exactly the same result.2 The question is how we can identify the meanings causing these shifts. Lohr (1999) is particularly concerned with defining the ‘best’ meaning-lists in an objective way; she developed two scales, of relative reconstructibility and retentiveness, on which meanings can vary. Lohr considered reconstructions of four protolanguages — Proto-Indo-European, Proto-Afroasiatic, ProtoAustronesian and Proto-Sino-Tibetan — to establish meanings which could be reconstructed for 2 or more of these protolanguages; she also calculated how often a different form is documented with that same meaning (alternatively, the number of replacements for that meaning), in a range of periods and languages within Indo-European. We tested two extreme samples from the Dyen, Kruskal and Black (1992) database — 30 meanings which scored as high as possible on Lohr’s indices of reconstructability and retentiveness, all being reconstructable for at least 3
2.After this paper was presented, Mieko Ogura kindly sent us a copy of Ogura and Wang (1998), which independently suggested that instabilities in bootstrapped trees may indicate admixture, though this hypothesis is not tested in the paper.
Family values
Figure 1.Neighbour Joining Tree rooted using the Albanian subgroup.
protolanguages, and with no more than 3 replacements. The other set had 23 meanings which were reconstructable for only 2 protolanguages, and which had 8 or more visible replacements in the 31.3 millennium total sample. The numbers in the two classes do not match because 6 meanings in the high
117
118
April McMahon and Robert McMahon
Figure 2.Close-up of the Germanic Branch from the Neighbour Joining tree based on the low reconstructibility, low retentiveness meanings.
Figure 3.Close-up of the Germanic Branch from the Neighbour Joining tree based on the high reconstructibility, high retentiveness meanings.
reconstructability : high retentiveness class were cognate across the whole of Indo-European, and thus totally uninformative. Let us consider just two differences involving these two extreme classes from the meaning lists, both within Germanic. When only the low reconstructability : low retentiveness meanings are used (Figure 2), Frisian appears as a sister of a group containing Afrikaans, Flemish and Dutch, as is also the case with the full 200-meaning list. However, with the high reconstructability: high retentiveness sublist (Figure 3), Frisian is related to these languages only at a deeper level, and the tree indicates an earlier split of Frisian as against the rest of the West Germanic group. In the second case, English, which in the full 200-word list tree and the tree for low : low meanings (Figure 2) appears (along with an English based creole) as a relatively deep, distant sister of the whole Germanic group, ‘migrates’ into the West Germanic group (Figure 3), based on the high reconstructability : high retentiveness meanings.
Family values
Our hypothesis is that these shifts of individual languages arise from borrowing; this seems to be supported by the direction in which languages move. Frisian gravitates towards Dutch, from which it has borrowed heavily, in the tree based on the less stable meanings. Similarly, we know there is considerable early borrowing from Norse into English: although English does not move into the North Germanic group, this may reflect the extensive borrowing from Romance, which is effectively pulling English towards the margins of Germanic altogether. As we have seen already (in Ringe, Warnow and Taylor’s (2002) results on Germanic, for instance), it does not seem possible to guarantee excluding borrowings, even from basic vocabulary lists. Embleton (1986) proposes a formula which corrects for the effects of borrowing; but although she derives a borrowing rate for Germanic, this is highly variable and must be calculated independently for each pair of languages, albeit with an upper limit of 30%. The problem, again, is one of generalisability: in each case we must simply add up the borrowings already diagnosed, then extrapolate that total over a particular time period as the overall rate, and factor this into the calculations. Where this does not help at all is in cases where there may or may not be borrowings; but the comparison of more retentive with less retentive meanings may distinguish such cases. By removing individual meanings to check the effect on the relative position of particular languages, we may in future be able to show exactly which meanings have been affected. There is also considerable potential here for future investigation: for instance, subdivided lists of different kinds may ultimately be helpful in diagnosing the form and degree of contact which has taken place between languages. Furthermore, the current restriction to 100- or 200-meaning basic vocabulary lists may no longer be necessary if we are able to calculate trees from different lists automatically, and to understand the various signals to be found in them. This may be of particular relevance in considering data from endangered languages, where it may not be possible to collect set lists, or when working with meagrely attested earlier stages of languages.
5. Trees and networks One goal for the future, then, is to develop these methods of diagnosing borrowing, improving our reading of these signals of contact-induced change; we are currently designing a series of simulations which should help us understand what signatures to expect in which circumstances. However, there is a
119
120 April McMahon and Robert McMahon
LATV OE
LITH
OHG
OPRU
ON GOTH TA
OCS
TB LUV ALB
OPERS AV
ARM HIT
GRK VEDIC
UMB LYC OSC
WELSH LATIN OIR
Figure 4.Output of Network for Ringe, Taylor and Warnow (2002) morphological and phonological characters.
more pressing question. If borrowing is extensive, or if the background of a language family involves a dialect continuum rather than ‘clean’ speciation events, as Ringe, Warnow and Taylor (2002) suggest for Germanic, then the family tree model will not be wholly appropriate, and we must therefore explore alternative means of displaying our results. We are currently exploring another biological program, Network (Bandelt, Forster, Sykes and Richards 1995, Bandelt, Forster and Röhl 1999, Forster, Torroni, Renfrew and Röhl 2001): again, biology and linguistics share a problem here, since there can be considerable genetic flow into a population from other populations. This network
Family values
model produces a multi-dimensional matrix to show complex relationships within a population; but perhaps the most interesting aspect of network analysis is that, although the program is designed to draw networks, cases where there has been no borrowing will automatically be represented with the most likely tree. That is, the program involved draws a tree when the relationships are clear and tree-like, and a more complex network when the connections are more complex and show more interaction. Network therefore seems to offer an appropriate means of exploring Ringe, Warnow and Taylor’s (2002: 112) concern that “we need to devise appropriate methods for inferring non-treelike networks of linguistic diversification from character data, and a means of deciding whether a tree or a network is appropriate in difficult cases”. It consequently seems reasonable to test the program on data from their paper. Thus far, we have used only Ringe, Warnow and Taylor’s morphological and phonological characters, largely because Ringe et al. are quite clear that non-lexical characters are likely to provide a clearer indication of linguistic relatedness. The coding of these non-lexical characters is taken directly from Ringe, Warnow and Taylor (2002); Network was applied to the resulting coded data, and the outcome is shown in Figure 4. The structure generated by Network is basically tree-like, but there is also very clear evidence of reticulations, where the program draws a box instead of a line; such reticulations can be seen at the root of the tree, and also within Germanic. These reticulations might indicate common parallel development, but could equally be indications of common borrowing from outside the family. Our intention here is not to argue against Ringe, Warnow and Taylor (2002) — they suggest themselves that “the diversification of the IE family must be modelled at least in part as a network rather than a tree” (2002: 110) — but to provide an initial suggestion as to how this modelling can be accomplished. Indeed, we hope to show here that different groups taking apparently radically different approaches (for instance, Ringe et al.’s character-based as opposed to our own distance-based methods) may provide evidence for similar solutions to particular historical problems.
References Bandelt, Hans-Jürgen, Peter Forster, Bryan C. Sykes & Martin B. Richards. 1995. “Mitochondrial portraits of human populations using median networks”. Genetics 141.743–753. ———, Peter Forster & A. Röhl. 1999. “Median-joining networks for inferring intraspecific phylogenies”. Molecular Biology and Evolution 16.37–48.
121
122 April McMahon and Robert McMahon
Barbujani, Guido. 1997. “DNA variation and language affinities”. American Journal of Human Genetics 61.1011–1014. Bloomfield, Leonard. 1935. Language. London: George Allen & Unwin. Campbell, Lyle. 1988. Review of Greenberg (1987). Language 64.591–615. Cavalli-Sforza, Luigi Luca. 2000. Genes, Peoples and Languages. London: Allen Lane/The Penguin Press. Durie, Mark & Malcolm Ross, eds. 1996. The Comparative Method Reviewed. Oxford: Oxford University Press. Dyen, Isidore, Joseph B. Kruskal & Paul Black. 1992. “An Indoeuropean classification: a lexicostatistical experiment”. Transactions of the American Philosophical Society 82, Part 5. Data available at http://www.ldc.upenn.edu. Embleton, Sheila M. 1986. Statistics in Historical Linguistics. Bochum: Brockmeyer. Felsenstein, J. 2001. PHYLIP: Phylogeny Inference Package. Version 3.6. Department of Genetics, University of Washington. Forster, Peter, Antonio Torroni, Colin Renfrew & A. Röhl. 2001. “Phylogenetic star construction applied to Asian and Papuan mtDNA evolution”. Molecular Biology and Evolution 18.1864–1881. Fox, Anthony. 1995. Linguistic Reconstruction: An introduction to theory and method. Oxford: Oxford University Press. Greenberg, Joseph H. 1987. Language in the Americas. Stanford: Stanford University Press. Heggarty, Paul. 2000a. Quantification and Comparison in Language Structure — An exploration of new methodologies. PhD thesis, University of Cambridge. ———. 2000b. “Quantifying change over time in phonetics”. Time Depth in Historical Linguistics, ed. by Colin Renfrew, April McMahon & Larry Trask, vol.2, 531–562. Cambridge: McDonald Institute for Archaeological Research. Hoenigswald, Henry M. 1960. Language Change and Linguistic Reconstruction. Chicago: University of Chicago Press. Iordan, Iorgu. 1970. An Introduction to Romance Linguistics. 2nd edn. translated by John Orr; revised by Rebecca Posner. Oxford: Blackwell. Labov, William. 1994. Principles of Linguistic Change. Volume 1: Internal Factors. Oxford: Blackwell. ———. Principles of Linguistic Change. Volume 2: External Factors. Oxford: Blackwell. Lohr, Marisa. 1999. Methods for the Genetic Classification of Languages. PhD thesis, University of Cambridge. McMahon, April. 2002. “Keeping Contact in the Family: Approaches to language classification and contact-induced change”. Paper presented at the NWCL Conference on Linguistic Areas, Convergence and Language Change, University of Manchester. ——— & Robert McMahon. 1995. “Linguistics, genetics and archaeology: Internal and external evidence in the Amerind controversy”. Transactions of the Philological Society 93.125–225. ——— & Robert McMahon. 2003. “Finding families: quantitative methods in language classification”. Transactions of the Philological Society 101.7–55. McMahon, Robert & April McMahon. 2002. “Lies, damned lies, and cladistics: linguistic classification and genetic correlations”. Paper presented at ARCLING II, Canberra.
Family values
Ogura, Mieko & William Wang. 1998. “Evolution theory and lexical diffusion”. Advances in English Historical Linguistics (1996) ed. by Jacek Fisiak & Marcin Krygier, 315–344. Berlin: Mouton de Gruyter. Page, Roderic D. M. 1996. “TREEVIEW: An application to display phylogenetic trees on personal computers”. Computer Applications in the Biosciences 12.357–358. Poloni, E. S., O. Semino, G. Passarino, A. S. Santachiara-Benerecetti, I. Dupanloup, A. Langaney & L. Excoffier. 1997. “Human genetic affinities for Y-chromosome P49a,f/ TaqI haplotypes show strong correspondence with linguistics”. American Journal of Human Genetics. 61.1015–1035. Renfrew, Colin. 1999. “Reflections on the archaeology of linguistic diversity”. Sykes 1999. 1–32. Ringe, Don. 1999. “How hard is it to match CVC-roots?” Transactions of the Philological Society 97.213–244. ———, Tandy Warnow & Ann Taylor. 2002. “Indo-European and computational cladistics”. Transactions of the Philological Society 100:59–129. Ruhlen, Merritt. 1991. A Guide to the World’s Languages. Volume 1: Classification. London: Edward Arnold. Sampson, Geoffrey. 1980. Schools of Linguistics: Competition and evolution. London: Hutchinson. Schmidt, Johannes. 1872. Die Verwandtschaftsverhältnisse der indogermanischen Sprachen. Weimar. Sims-Williams, Patrick. 1998. “Genetics, linguistics and prehistory: thinking big and thinking straight”. Antiquity 72.505–527. Smith, Neil. 1989. The Twitter Machine. Oxford: Blackwell. Sykes, Bryan, ed., 1999. The Human Inheritance: Genes, Language, and Evolution. Oxford: Oxford University Press. Warnow, Tandy, Don Ringe & Ann Taylor. 1996. “Reconstructing the evolutionary history of natural languages”. Proceedings of ACM-SIAM Symposium on Discrete Algorithms. 314–322.
123
From inventory to typology in English historical dialectology* Anneli Meurman-Solin Helsinki Collegium for Advanced Studies, University of Helsinki
This research is part of a project examining the theoretical and methodological implications of corpus-based inventories in the production of typologically relevant new knowledge. The focus on Scots allows me to highlight the importance of full-scale inventories of local and regional variation in varieties that failed to become national standards, diatopic variation often being downplayed in typological research. The main area of linguistic interest is clause combining systems in Older Scots.
1.
General goals
In the reconstruction of language varieties of the past, the following three general goals seem to me particularly relevant. Firstly, the creation of a wider range of diplomatically transcribed manuscript sources, preferably as part of large electronic databases; secondly, the development of sophisticated tagging systems and software for data retrieval and presentation tailored for diachronic text corpora; and thirdly, the creation of metalanguage and data presentation conventions which are more appropriate for reporting on continued variation and variability attested in corpus-based descriptions of language than those applicable to describing relatively unidirectional processes of change such as language standardization, or construed realities such as language standards. The relevance of this third goal is obvious in my own field of study, that of Scottish historical linguistics. The comparative approach and its metalanguage, frequently
*This research was supported by the Academy of Finland Centre of Excellence funding for the Research Unit for Variation and Change in English at the Department of English, University of Helsinki.
126 Anneli Meurman-Solin
adopted in the field, have not always enhanced our understanding of developments in Scots. Comparability across categories and the typological relevance of findings across languages and language varieties has sometimes been assumed, rather than shown to be valid, as sufficiently detailed corpus-based descriptions have not been available. The risk of assuming direct comparability has become more obvious with the improvement in the quality and quantity of data. In my work I focus on reconstructing variation and variety in the history of English as used in the various regions and localities of Scotland. This reconstruction takes place by producing detailed inventories of idiolect- or textspecific practices. When such things as careful descriptions of micro-level language use based on family letters are available, these micro-level practices are examined to find out the extent to which they converge with or diverge from those attested in other idiolects, texts or groups of texts in databases tailored to be diachronically, diatopically and diastratically representative. Manuscriptbased data representing chiefly non-literary texts are made available in large quantities while trying not to lose any of the quality achieved by applying good philological computing practices in the digitization process. My approach is similar to that adopted at the Institute for Historical Dialectology, University of Edinburgh (Laing & Williamson 2004). The two Edinburgh projects working towards A Linguistic Atlas of Early Middle English (Laing 1993) and A Linguistic Atlas of Older Scots (Williamson 1992/1993) were launched from the tradition of linguistic atlases based on questionnaires, but the methodology has been developed further so that the databases of linguistic material are lexico-grammatically tagged corpora of full texts, diplomatically edited, rather than questionnaire-delimited sets of isolated word-forms. Further, the ‘fit-technique’, a method of interpolating texts of unknown provenance into a dialect continuum, has been computerized (Williamson 2000). To guarantee the validity of linguistic description, the analysis of language variation over time and space is based on microscopically precise text profiles containing each linguistic item in the shape attested in the original manuscript, with no preselection or hierarchization involved. Manuscript-based, lexico-grammatically tagged corpora of this kind and the software developed for extracting and presenting data will give us reliable descriptions, or valid inventories, if you like. The full scale of variation can be detected, and the patterns of change triggered by variety-internal or -external factors can be identified as complex processes rather than as properties of objectified systems. In the case of Scots, the tradition of focusing on the rise of the so-called Scottish English Standard, and especially the anglicization of this standard, often restricting the focus to texts originating from the south-east, has
From inventory to typology in English historical dialectology 127
sometimes imposed a perspective that may prevent the researcher from seeing the multidirectionality of developments over time, place and social milieu. My third general point, the inadequacies of metalanguage, will allow me to contrast our awareness of language variation and variability and the tendency to objectify or reify language varieties and historicize them.
2. Metalanguage: Language variability and the objectification and historicization of varieties 2.1 Reification The concepts of ‘objectification’, ‘hierarchization’, and ‘historicization’ summarize the essence of trends that, in my view, have perhaps slowed down the process of increasing our knowledge of varieties of English, both those varieties which developed into the so-called standardized varieties and those that have remained local or regional. I would like to claim that these three trends have greatly influenced both the theoretical and methodological approach widely adopted in the study of varieties of English and the metalanguage used in the description of linguistic findings. In his chapter entitled “The consequences of standardization in descriptive linguistics” James Milroy (1999: 17) discusses “the interaction between scholarly linguistic attitudes to language and the publicly expressed attitudes of nonlinguists and critics of linguistics” and claims that “the consequences of standardisation are discernible in the attitudes of linguistic scholars themselves: their judgements as to what the object of description consists of have been influenced by their knowledge that a standard form exists in some abstract dimension and by some consequences of the ideology of standardisation.” Crowley (1990) and Benson (2001) have also drawn the conclusion that despite our awareness of the inherent variability of language we tend to provide descriptions suggesting that we have in fact identified a language variety. With reference to Crowley (1990), Benson (ibid. 20) points out that “[i]n the Saussurean scheme … language variation can only be explained if each nonstandard form is treated as a system in its own right” and, as a consequence, “[t]he inherent variability of language, which justifies the distinction between langue and parole, is thus reduced to a proliferation of distinct language ‘varieties’, each having its own invariant systematicity”. Benson (ibid. 21) has also identified the tendency to objectify in the
128 Anneli Meurman-Solin
lexicographical practice of using national and regional labels, which is his main concern: Anthropological and sociological approaches to language variation both tend to reduce the inherent variability of language to language variety. Variety imposes order on variability and produces innumerable independent objects for scientific investigation within a single language. This process is especially apparent in the literature on World Englishes, which has tended to emphasise the identification, description and classification of regional and national varieties of English over the description of the fluidity of English as a language functioning across geographical and political boundaries. It is also apparent in the broad approach of twentieth-century English dictionaries to the international dimensions of English, which tends to adopt the concept of language variety uncritically in the form of national and regional labels. While language variation is now generally seen as ‘good’, there is at the same time a tendency to systematically reduce it to objectified and describable forms. Twentiethcentury dictionaries are implicated in the objectification and homogenisation of language varieties in much the same way as their predecessors were implicated in the objectification and homogenisation of the English language as a whole. (Benson 2001: 21)
Objectification or reification may be an innate characteristic reflecting the functioning of the human brain, but mental frames or schemata of this kind are problematic as they may regulate processes of analysis in which linguists are trying to identify some order in heterogeneity, that is, some relatively consistently preferred practices in data that otherwise chiefly give evidence of heterogeneity and continued variation. Demarcating areas as the territories of specific varieties may divert our attention from examining ordered heterogeneity recordable only by crossing such imposed boundaries. The following questions seem relevant. To what extent do the methods applied to the study of varieties of English reflect this tendency to insist on the existence of language entities, or systems, even at a stage when there may not yet be sufficient, i.e. statistically significant, evidence of uniformity or systematicity within such an entity? How has the tradition of giving legitimacy to a description of language use in a specific place by identifying it as a distinctive entity affected our ability to develop appropriate discourses for discussing linguistic variability? How has reification affected the way in which we conceptualize Scots or Scottish English? One of the main points of my paper is that the negative influence of reification can be avoided by improving both the quantity and quality of text corpora containing digitized and preferably tagged data produced by as wide a
From inventory to typology in English historical dialectology 129
range of text and discourse communities as possible. A particularly important consequence of de-reification in describing Scots is that variation and variety resulting from varieties and languages having come into contact on Scottish soil will be given due attention. 2.2 Hierarchization A further problem arises from this tendency to reify varieties of English: the entities, or objectified systems, tend to be hierarchically ordered. This seems to lead to the widespread practice of adopting a comparative viewpoint especially when examining the non-standardized or otherwise less prestigious varieties. From the centrist perspective, regional and local varieties are satisfactorily described in terms of how they relate to standardized varieties. Moreover, descriptions of this kind are often expected to focus on how regional or local varieties diverge from standardized varieties, so that the often considerable proportion of shared features between them will not be discussed with sufficient thoroughness. The tendency to adopt the comparative approach to the study of regional and local varieties can be seen as a reflection of the views in Milroy (1999) quoted above. Judging by earlier research, it seems to have been difficult to resist the temptation of approaching the study of the Scots language from this comparative angle. For a number of political and socio-cultural reasons, the tradition of objectifying the Scottish English variety may be especially powerful, and in the non-corpus-based studies in particular the tendency to highlight divergence and downplay convergence has not taken us closer to understanding the history of Scots. No comprehensive account of the Scots variety has thus yet been made available. The most important achievements are in the field of lexicography — the Dictionary of the Older Scottish Tongue and The Scottish National Dictionary, in particular (Dareau 2004). 2.3 Historicization The third trend, ‘historicization’, has had similar consequences in the reconstruction of text languages as reification. Communities using regional varieties may have a tendency to historicize the variety for it to gain legitimacy (cf. Milroy 1999: 28). As Milroy points out, “[m]uch of the legitimisation of the standard variety is in fact achieved through creating a history for the standard language — a process that we can reasonably call historicisation (involving the
130 Anneli Meurman-Solin
creation of a legitimate historical canon)”. While the legitimacy of the standard variety has influenced the ways in which we conceptualize regional and local varieties, the historicization of the Scots language, for instance, has maintained attitudes that have justified the use of an allegedly independent language variety in the pursuit of nationalistic goals. Having been educated in a canon-saturated tradition makes it difficult not to be influenced by the historical canon. More work at the record offices browsing through bundles of documents is needed to improve the range of data (Meurman-Solin 2001a). When a text community is defined in terms of what written texts verifiably had a social and communicative function among the literate members of that community, it may be possible to deobjectify and dehistoricize.
3. Tagged corpora Four corpus projects are relevant to providing better inventories of Scots. The Helsinki Corpus of Older Scots (HCOS) is available on the ICAME CDRom, but a new version is being produced by checking the texts against manuscript and annotating them using the Institute for Historical Dialectology software developed for lexico-grammatical tagging. Cooperation between research units in Edinburgh and Helsinki will allow us to create a diatopically, diachronically and diastratically representative corpus of Scots which covers the time-span from the earliest vernacular texts around the mid-fourteenth century up to the end of the eighteenth. In addition to the Edinburgh-Helsinki corpus (E/HCOS), I am compiling a Corpus of Scottish Correspondence (CSC), based on a careful search of the family archives, in order to ensure diversity as regards dialect, the writers’ social and geographical mobility and linguistic and stylistic competence. This corpus will also be tagged. My fourth database (CESWW) will consist of early Scottish women’s writings. In all of these, the texts will be diplomatically transcribed and then keyed in using principles of philological computing (Meurman-Solin 2001a: 18–22). No normalization or emendation is allowed. At this point some comments on tagging will shed light on tools that facilitate the use of detailed inventories in typology-oriented work. Not all tagging is useful. It is possible to distort evidence by applying overly rigid rules for category membership or ignoring the innate fuzziness between categories, and by simplifying complex patterns of variation or using tags that fail to reflect processes of change over a long time-span, for instance. Tagging is not just a tool that facilitates searching. To be of general relevance it cannot apply
From inventory to typology in English historical dialectology
definitions created by a specific theoretical approach. We focus on giving information about structural, i.e. formal, and semantic features rather than category-defined or syntactic properties. In cases in which it has not been possible to avoid theory-specific practices, these must be made as transparent as possible. Since tagging must allow a sufficient degree of flexibility, our tagging system allows different degrees of refinement in the choice of type of tag by introducing a hierarchical way of ordering information in the tags. The lexeme consider usefully illustrates the practice of refining tags to indicate categorical fuzziness or polyfunctionality, so that the grammaticalization of the present participle can be traced: $consider/vpsp_CONSIDER+ING $/vpsp_+ING $consider/vpsp-cj_CONSIDER+ING $/vpsp-cj_+ING $consider/vpsp-pr_CONSIDER+ING $/vpsp-pr_+ING $consider/vn_CONSIDER+ING $/vn_+ING The practice is similar when tagging conjunctive phrases incorporating nouns such as on the consideration that: $on/pr_ON $/T_THE $consideration/n-cj_CONSIDERATION $that/cj<_THAT Even though the tag also provides a basis for the analysis of syntactic properties by indicating the connective function, the main criterion in the identification of candidates for a full inventory of realisations is semantic. The tagging principles applied to clause-combining devices at this experimental stage have been influenced by the discussion of notional or conceptual properties of wordclasses (Anderson 1997, Jackendoff 2002). This can be illustrated by the following set of conjunctive phrases implying purpose: (1) $for/pr_FOR $fear/n-cj_FEAR $that/cj<_THAT $to/pr_TO $/T_THE $intent/n-cj_INTENT $that/cj<_THAT $to/pr_TO $/T_THE $end/n-cj_END $that/cj<_THAT $to/pr_TO $/T_THE $effect/n-cj_EFFECT $that/cj<_THAT $in/pr_IN $intent/n-cj>vi_INTENT $to/im+C_TO $live/vi_LIVE $long/ av_LONG $in/pr_IN $intent/n-cj>pr_INTENT $of/pr_OF $live/vn_LIVING $long/av_LONG
We will find that in the case of some candidates a grammaticalization process may not actualize; if it does, the spread may take place at a varying rate or only
131
132
Anneli Meurman-Solin
affect restricted domains (on the grammaticalization of seeing and considering, see Meurman-Solin and Pahta forthcoming). These practices allow us to mark topic areas and specific features for a later more detailed analysis, and are therefore appropriate for studying language change over a long time-span. In other words, this method ensures that category changes, polyfunctionality or fuzziness requiring closer analysis can be examined with all the necessary rigour, and items that may play a role in a pattern or taxonomy can be identified and defined. The order of information in the tag is from general to particular, so that searches can be modified according to the kind of information considered relevant in a particular piece of research. Arrows are used to indicate that an item should be examined as being possibly part of a multi-unit structure or collocate. Thus I have indicated the presence of an adverb in the immediate context whenever that adverb can be viewed as having semantic properties enriching those of the linking device in some systematic way (e.g. providing always, if ever, or ever, ay quhill (cf. ay and quhill), as soon as ever, how soon that ever, how soon and incontinent, how soon… incontinent). An interesting cluster of a conjunctive participle semantically enriched or refined by the frequently co-occurring adverb is providing always, as illustrated in example (1): (1) The inquest ordains to ansuer Robert Atzin, and ilk ane of the officaris, of ane ferlot of meill in this storme to help thair wiffis and barnis, providing allwayis that thai clame na possessioun thairof in tyme cuming. (HCOS 1555 Peebles records, 225; ed. W. Chambers 1872)
The tagging is as follows: $provide/vpsp-cj_PROVYD+ING $/vpsp-cj_+ING $always/av<_ALWAYS $that/cj<_THAT This cluster occurs in 62 per cent of the occurrences of providing as a subordinator in the HCOS data, the proportion in statutory and administrative texts being as high as 73 per cent. This feature is related to the dominance of generic time reference in such texts. Corpus compilers are improving the quantitative and qualitative validity and relevance of electronic databases but, for reasons of research economy they may make compromises in the tagging by relying on pre-corpus-linguistics descriptions, resorting to automatic (i.e. non-interactive) tagging, or imposing neat category labels on the data. Yet the high degree of scholarly rigour in ensuring the authenticity of data should also be applied in developing methods
From inventory to typology in English historical dialectology
for annotating texts. In our view, tags suggesting membership of predefined categories based on theory-bound syntactic properties must be avoided in the study of regional and local varieties of English in particular, since the grammar of such varieties may significantly differ from that of standardized varieties. In other words, great caution is necessary as decision-making about individual tags may reflect an implicit application of grammatical rules written on the basis of data not valid or relevant in the context.
4. Inventories and typologies: Clause-combining The typology of clause combining-devices in English as well as other European languages has been extensively discussed in recent literature, diachronic developments of adverbial subordinators in Kortmann (1997) being particularly relevant as regards points discussed in this section; for synchronic descriptions, see, for example, Devriendt et al. 1996 and van der Auwera 1998. As the ongoing corpus-based work has not yet produced detailed inventories that draw on data extracted from qualitatively and quantitatively representative text corpora, typologies have been construed chiefly by using secondary sources such as dictionaries and grammars (Kortmann 1997: 53). Regional or dialectal variation tends to be ignored or marginalized, which may lead to an excessive degree of streamlining and simplification in the selection and presentation of information to ensure the production of valid typologies. Problems related to insisting on typological relevance even though inventories are insufficient are multiplied in the comparative approach since the lack of consistency and coherence in earlier descriptions may flaw comparisons seriously. The streamlining effect can be illustrated by examining how the noncorpus-based inventory of adverbial subordinators in Present-day English (Kortmann 1997:292–294) differs from a corpus-based inventory of Present-day Scots (Häcker 1999:211). Beside information about variant forms (whenever/ finiver/whaneir, frae/fae/from/faem, once/yince/eence), the following subordinators in Häcker’s list are absent in Kortmann’s inventory: as gin, barrin, da wye at, even for as, for aa, gin, the minute/da meenit, onietime, say and the doubtful okay. Häcker’s study is based on both written and spoken texts dating from the period 1980–1994, but, as regards genres, only narrative prose is represented (ibid. 18–19). As numerous written genres in present-day Scotland are not assumed to contain diatopically relevant data as a result of the widespread preference for Standard English, the differences can be assumed to be more salient earlier in
133
134 Anneli Meurman-Solin
the history of Scots. Section 5 will highlight changes in the inventories of some adverbial subordinators in Scots in the 1450–1700 period, focusing on the chronology of changes, the identification of periods when the most dramatic changes take place, and the developments in the general frequency of items in specific semantic roles. I will also illustrate developments in the morphological complexity of adverbial subordinators and their preferred categorial sources. As the present study is chiefly based on the HCOS, a corpus of not more than 850,000 words, a full inventory cannot yet be provided even though the wide range of genres in this database, fifteen altogether, increases the significance of the findings. Comments on the typological relevance of the inventory must thus also be preliminary. Beside the expansion of databases, my earlier research has suggested a number of areas where further research is necessary. Meurman-Solin (2001a: 20–22) highlights the importance of using diplomatically transcribed manuscript sources as data because patterns of complex sentence structures in Present-day English may have been imposed as part of the editorial practice of introducing punctuation. In assessing the representativeness of corpora it also seems important to use stylistic competence as a criterion. Meurman-Solin (2001b: 42–45) contrasts letter-writers whose repertoire of connectives only includes items such as and and for with those who manage to indicate logical relationships between propositions more explicitly. Leaving aside regional and dialectal variation, the question of what can be claimed to provide a representative range of data will have to be considered in the selection of genres and discourses as well as the varying stylistic competence of the informants. A thorough knowledge of the evolution of discourse- and genre-specific practices over time and space (see Brinton 2000, for instance) will permit the assessment of the performance of individual informants in a wider context. Moreover, Meurman-Solin (2002: 189–193) discusses the problems related to categorization, overlapping and intersection between categories such as conjunctions, prepositions and adverbs, suggesting that in order to define the concepts of ‘subordination’ and ‘subordinating clausal link’ the whole repertoire of connectives will have to be analysed. Degrees of fuzziness have important typological implications. Instead of seeing fuzziness as ambiguity, it is useful to apply the concept of polyfunctionality here (cf. Kortmann 1997:89–94). In addition, in considering the hierarchical downgrading of subordinate syntagms from parataxis to adjoined clauses, correlative diptychs, clause chaining (including co-subordination), and embedding (Lehmann 1988: 183–186), the identification of subordinating conjunctions that should be included in
From inventory to typology in English historical dialectology
inventories of the kind illustrated above complicates matters further.1 As suggested in Section 3, identification can be facilitated by a tagging system which is protean in the sense that it can flexibly record information about structural features — single features and collocates — as well as arrange that information in various ways without preconceiving their syntactic properties. Within the scope of the present study, the focus has been restricted to what Lehmann calls ‘adjoined clauses’, where “[o]ne of the two clauses constituting the complex sentence contains a subordinative conjunction and may thus be identified as the subordinate clause” (ibid. 185). It has also been necessary to exclude non-finite and verbless adverbial clauses, as their inclusion may also have required including a wide range of structural realisations on the continuum from non-finite clauses to nominalizations.
5. Adverbial subordinators in English Kortmann offers a valuable description of the typology and history of adverbial subordinators in European languages as well as “a macro-analysis of the history of adverbial subordinators in English” (1997: 346–348). He draws our attention to a number of topics where further research is necessary (ibid. 291). Fluctuations over time in the composition of the inventories and the relative proportions of adverbial subordinators signalling temporal relations compared with causal, conditional, concessive and related interclausal relations should be traced. Secondly, he stresses the need to examine changes in “the morphological make-up of adverbial subordinators” and to assess the varying importance of the origins of subordinators in English. Other interesting questions are whether adverbial subordinators in English developed “towards an ever higher degree of semantic differentiation” or whether “a decrease in (semantic and syntactic) polyfunctionality” can be attested. In the summary of developments from Old English to Present-day English, Kortmann (ibid. 291–335) compares the donor rates of the different periods (ibid. 301), and lists the major results as regards diachronic developments (ibid. 345–348), also formulating further specific research questions (ibid. 348–351).
1.In ‘correlative diptychs’… “the relative clauses … are subordinate, but not embedded”; some types of ‘clause chaining’ are referred to by Foley & Van Valin (1984: 238–263) as ‘cosubordination’ and involve the use of adverbs defining the semantic link (Meurman-Solin 2002: 188, 206, fn 4).
135
136 Anneli Meurman-Solin
According to Kortmann, Old English “provided the majority of those adverbial subordinators in Present-Day English with the highest frequency of use”, including after, as, as long as, as soon as, if, since, so, that, though, and while, so that “about 70% of those Old English subordinators which made their way into the 20th century belong to this set of high-frequency items”. He writes that “[t]he adverbial subordinators with Old English ancestors can therefore justly be viewed as forming the backbone of the subordinator inventory in PresentDay English, both with respect to their frequency of use and the range and nature (i.e. basicness) of the interclausal relations they cover” (ibid. 296). The innovativeness of Middle English is obvious since “roughly 75% (!) of its adverbial subordinators had no predecessors or were not used in this function in Old English” (ibid. 299). Kortmann (ibid. 302) explains the fact that “Middle English, especially Late Middle English… crucially shaped the inventory of adverbial subordinators” referring to the emergence of “a fairly stabilized written language in the period from roughly 1350 onwards”, both “as a literary language” and “the language of official use”. He considers the low donor rate and high drop-out rate (ibid. 301, Tables 10 and 11), and the sheer size of the Early Modern English inventory of adverbial subordinators (“larger by a third compared with Middle English and by some 50% compared with Present-Day English”), and draws the conclusion that “Early Modern English was essentially a period of experiment and transition” (ibid. 302). Section 6 provides some answers to Kortmann’s questions, illustrating the quite dramatic changes in the inventories based on Scots data.
6. Changes over time in the inventories of some adverbial subordinators in Scots The data in the HCOS have been structured by the variable of time into the following four sub-periods: 1450–1500, 1500–1570, 1570–1640 and 1640–1700. Information in tables and graphics in this section has been arranged accordingly. In the first sub-period there are fewer texts (a third of those dating from the third period) as the range of prose genres was smaller at that time. The selection of topics concentrates on the chronology of changes, that is, the identification of periods when the most dramatic changes take place. I will examine developments in the general frequency of semantically more or less closely related subordinators that compete with each other as well as illustrate developments in the morphological complexity of adverbial subordinators and their preferred categorial sources.
From inventory to typology in English historical dialectology
6.1 Relative frequencies of semantically related subordinators Can we identify varying degrees of dynamism in processes of change in the history of Scots? Are there periods of transition reflected in patterns of covariation, variants being used interchangeably, or can specific types of intermediary variant be attested? I would like to illustrate some of the major changes in the relative frequencies of semantically-related linking devices in Scots by first examining the semantic fields of “before a point of time” and “up to a point of time”. All variant forms are included in the statistics. In inventories of the kind listed here syntactic, semantic and pragmatic differences between the items have been ignored. We can see the subordinator before taking the status of a prevailing subordinator in the third sub-period, thus replacing or as an abrupt change in the percentage and the mean frequency takes place: before from 14 (0.03/1,000) to 72 per cent (0.19/1,000) and or from 70 (0.17/1,000) to 23 per cent (0.06/1,000). It is significant that the other two semantically closely related subordinators are shown to occur only in a few texts (afore) or quite late (ere), both remaining quite infrequent (0.04 and 0.01–0.04/1,000 respectively) (Table 1 and Figure 1). As regards subordinators sharing the sense “up to a point of time”, the most significant change in relative frequencies takes place later than in the cases of before and or (Table 2). A change in the relative frequencies of the two main subordinators takes Table 1.Percentages (absolute numbers in brackets) and mean frequencies (M per 1,000) of or, before, afore and ere in the HCOS. TM = mean frequency of the total. Period
or
1450–1500 1500–1570 1570–1640 1640–1700
94 (17) 0.18 70 (35) 0.17 23 (18) 0.06 5 (5) 0.02
M
before M
afore
6 (1) 0.01 14 (7) 0.03 72 (56) 0.19 84 (79) 0.32
(0) 16 (8) 0.04 (0) (0)
M
ere
M
(0) (0) 5 (4) 0.01 11 (10) 0.04
Total
TM
(18) (50) (78) (94)
0.19 0.25 0.26 0.38
Table 2.Percentages (absolute numbers in brackets) and mean frequencies (M per 1,000) of quhill/while, till and until in the HCOS. TM = mean frequency of the total. Period
quhill
M
till
M
until
1450–1500 1500–1570 1570–1640 1640–1700
97 (38) 94 (103) 65 (107) 18 (28)
0.40 0.51 0.35 0.11
3 (1) 6 (7) 25 (41) 61 (92)
0.01 0.03 0.14 0.37
(0) (0) 10 (16) 21 (32)
M
Total
TM
0.05 0.13
(39) (110) (164) (152)
0.41 0.55 0.54 0.62
137
138
Anneli Meurman-Solin
100 90 80 70 60
or before afore ere
50 40 30 20 10 0
1450–1500
1500–1570
1570–1640
1640–1700
Figure 1.Percentages of or, before, afore and ere in the HCOS.
place in the latter half of the seventeenth century, the proportion of till increasing from 25 to 61 per cent and, as a contemporary development, that of quhill/ while decreasing from 65 to 18 per cent. These two processes are accompanied by a significantly smaller increase in the frequency of until. Figure 2 shows that until is relatively late and only slightly more frequent than the receding variant quhill, its proportion increasing from 10 to 21 per cent in the latter half of the seventeenth century. Statistical information about other sets of semantically related subordinators identifies the third sub-period as one of dramatic change. An increase in the relative proportion of unless from 6 to 80 per cent in the third sub-period has been recorded, while but if is very rare in the fourth subperiod (Table 3 and Figure 3). The shape of the curve in Figure 4 is somewhat different from the previous ones, the change towards the preference for as soon as being less abrupt (Table 4). As in the case of unless, the proportion of the spreading variant, as soon as, is very high in the last sub-period, a development also reflected in the mean frequencies. Inevitably, graphs of this kind do not give the full picture, but they do show that we are dealing with major changes in a relatively short time. In Figure 5, Figures 1–4 have been superimposed. Figure 5 shows that the difference between the prevailing and quite infrequent variants in the 1450–1500 period is quite striking. The proportions do
From inventory to typology in English historical dialectology 139
100 90 80 70 60 quhill till until
50 40 30 20 10 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 2.Percentages of quhill/while, till and until. Table 3.Percentages (absolute numbers in brackets) and mean frequencies (M per 1,000) of but if and unless in the HCOS. TM = mean frequency of the total. Period
but if
Mean
unless
Mean
Total
TM
1450–1500 1500–1570 1570–1640 1640–1700
100 (23) 94 (17) 20 (3) 3 (1)
0.24 0.08 0.01 0.00
(0) 6 (1) 80 (12) 97 (38)
0.00 0.04 0.15
(23) (18) (15) (39)
0.24 0.09 0.05 0.16
Table 4.Percentages (absolute numbers in brackets) and mean frequencies (M per 1,000) of how soon as and as soon as in the HCOS. TM = mean frequency of the total. Period
how soon as
Mean
as soon as
Mean
Total
Mean
1450–1500 1500–1570 1570–1640 1640–1700
(0) 65 (17) 45 (9) 12 (5)
0.08 0.03 0.02
(0) 35 (9) 55 (11) 88 (35)
0.04 0.04 0.14
(0) (26) (20) (40)
0.13 0.07 0.16
not change significantly in the 1500–1570 period, the decrease in or being somewhat greater because of the competing before and afore, which together represent 30 per cent. The previously less frequent item becomes the majority variant in the 1570–1640 period. Interestingly, quhill is a more resistant feature, till beginning to gain more ground in the post-1640 texts. Inventories based on purely quantitative data can sometimes be misleading
140 Anneli Meurman-Solin
100 90 80 70 60 but if unless
50 40 30 20 10 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 3. Percentages of but if and unless in the HCOS. 100 90 80 70 60 how soon as as soon as
50 40 30 20 10 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 4.Percentages of how soon and as soon in the HCOS.
as examination of the subordinators though and although in Older Scots illustrates. The peak in the use of although cannot be related to general trends in Older Scots.2 The high proportion of this subordinator in the 1570–1640 period can be explained by the idiolectal preference of James VI, since almost 70 per cent
2.The subordinator although also remains much less frequent than though in Present-day written Scots (Häcker 1999: 140).
From inventory to typology in English historical dialectology
100 90 80
or before afore ere quhill till until but if unless how soon as as soon as
70 60 50 40 30 20 10 0
1450–1500
1500–1570
1570–1640
1640–1700
Figure 5.Change over time in the relative frequencies of some adverbial subordinators in the HCOS. Table 5.Percentages (absolute numbers in brackets) and mean frequencies (M per 1,000) of though and although in the HCOS. TM = mean frequency of the total. Period
though
Mean
although
1450–1500 1500–1570 1570–1640 1640–1700
100 (7) 100 (9) 39 (27) 93 (154)
0.07 0.04 0.09 0.62
(0) (0) 61 (42) 7 (11)
Mean
Total
TM
0.14 0.04
(7) (9) (69) (165)
0.07 0.04 0.23 0.67
of the occurrences of although in the period 1570–1640 have been attested in his writings (Table 5 and Figure 6). The total mean frequencies in the tables show a significant difference between the various subordinators. Listed in the order of mean frequency in the period 1640–1700, though is by far the most frequent, till and before being the second and third most frequent subordinator among those analysed here (Table 6). 6.2 Changes in the morphological complexity of adverbial subordinators Since a full account of changes in the morphological complexity of adverbial subordinators in Older Scots cannot be provided within the scope of the present study, the data are restricted to subordinators of concession. The number of types varies between nine in the first sub-period and thirteen in the third. The prevailing concessive subordinator is whatever or whatsoever in the first sub-period (28 per cent), while howbeit takes this position in the second (43 per
141
142 Anneli Meurman-Solin
100 90 80 70 60 though although
50 40 30 20 10 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 6.Percentages of though and although in the HCOS. Table 6.Change over time (1450–1700) in the mean frequencies of some adverbial subordinators in Older Scots. M/10,000. Subordinator
1450–1700
though till before as soon as unless quhill until although or how soon as but if
0.7–6.2 0.1–3.7 0.1–3.2 0.0–1.4 0.0–1.5 4.0–1.1 0.0–1.3 0.0–0.4 1.8–0.2 0.0–0.2 2.4–0.0
cent), albeit in the third (30 per cent), and though in the fourth (51 per cent). In addition to being used for condition, suppose also implies concession in Scots. We notice, however, that, after the first sub-period, the concessive suppose becomes one of the many relatively infrequent subordinating devices in this semantic group. I would also like to draw attention to the very late use of while as a subordinator of concession in Scots. There are only eight examples in the HCOS, all dating from the seventeenth century. Overall, the difference between high- and low-frequency items increases over time (Figures 7a–d).
From inventory to typology in English historical dialectology 143
1450-1500 100 90 80 70 60 50
%
40 30 20 10
ho w( so )e ve wh r at (s o) ev er wh o( so )e ve wh r en (s o) ev wh er er e( so no )e tw ve r ith sta nd in g wh et he r su pp os e
th ou gh
0
1500-1570 100 90 80 70 60 50 40
%
30 20
eit ho w( so )e wh ve r at (s o) wh ever o( so )e ve wh r en (s o) wh ev er er e( so no )e tw ve ith r sta nd in g wh et he r su pp os e
ho wb
eit alb
th ou gh
10 0
When we look at the developments in the use of the -ever type (what(so)ever, when(so)ever, how(so)ever, etc.) and the syntactic phrase type (albeit and
144 Anneli Meurman-Solin
1570-1640 100 90 80 70 60 50 40
%
30 20
th ou g alt h ho ug h alb ei ho t w ho b w( eit so wh )ev er at (s o) wh ev o( er s wh o)e en ver ( wh so) er eve r no e(s t w o)e ve ith sta r nd in wh g et he r wh il su e pp os e
10 0
1640-1700 100 90 80 70 60 %
50 40 30 20 10 0 though
how(so)ever
when(so)ever
whether
Figures 7a–d.Percentages of subordinators of concession in the HCOS.
howbeit), we see an overall decrease in their use over time. I find it interesting that the syntactic phrase type should have as high a proportion as 56 per cent in the second sub-period, the percentage (38) also being quite high in the next sub-period (Table 7).
From inventory to typology in English historical dialectology
Table 7.Percentages of the -ever type and the syntactic phrase type for concession in the HCOS. Period
-ever type
syntactic phrase
1450–1500 1500–1570 1570–1640 1640–1700
48 24 22 17
– 56 38 15
6.3 Changes in the preferred categorial sources of adverbial subordinators To illustrate changes in the preferred categorial sources of adverbial subordinators in Scots, we might consider the following charts in which nouns and verbs have been compared as sources of subordinators. Figure 8a tells us about changes over time in the mean frequencies: firstly, of subordinators incorporating nouns such as by reason (that), in case (that), to the end (that) and, secondly, those that have been derived from verb forms, in Scots mostly present participles, considering (that), for instance. 10 9 8 7 6 5 4 3 2 1 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 8a.Number of subordinator types incorporating nouns and derived from verbs in the HCOS.
We see a significant increase in both types, which is earlier in the group having verbs as a source. We also notice that the mean frequency of connective
145
146 Anneli Meurman-Solin
6 5 4 3 2 1 0 1450–1500
1500–1570
1570–1640
1640–1700
Figure 8b.Mean frequencies of subordinator types incorporating nouns and derived from verbs in the HCOS. Table 8.Occurrences of some conjunctive phrases incorporating nouns in the HCOS. TM = mean frequency of the total. Conjunctive phrase
1450–1500 1500–1570 1570–1640 1640–1700 Total
TM
to the intent (that) to the end (that) to the effect (that) for fear (that) by reason (that) in case (that) upon condition (that) in regard (that) in respect (that) on the consideration (that)
1 2 – – – 2 – – – –
– – 27 – 18 9 – – 3 –
1 2 42 5 13 17 1 3 19 –
– 8 31 – 4 21 1 8 7 1
2 12 100 5 35 49 2 11 29 1
0.00 0.01 0.12 0.01 0.04 0.06 0.00 0.01 0.03 0.00
Total TM
5 0.05
57 0.28
103 0.34
81 0.33
246 0.29
0.29
phrases incorporating nouns remains relatively stable after the increase in the first half of the sixteenth century, while the mean frequency of those derived from verbs increases considerably. Thus, assessed by number of types, the use of nouns as a source becomes particularly productive towards the end of the sixteenth century, whereas the number of verbs which have adopted the connective function reaches its peak in the 1500–1570 period. However, the
From inventory to typology in English historical dialectology 147
mean frequencies of connectives incorporating nouns increase considerably as early as this second sub-period, remaining relatively stable after that. Verbs as a source develop differently, as reflected in the continuously increasing mean frequencies, which are almost 2.5 times more frequent in the last sub-period than in pre-1500 texts (Figure 8b). The data reflect an increase in conjunctive phrases incorporating nouns in the sixteenth and seventeenth centuries, their frequency being very low in pre1500 texts. Moreover, the number of types is smaller in the second sub-period than the considerably enriched inventory in post-1570 texts. Some of the conjunctive phrases remain relatively rare (to the end (that) 0.02–0–0.02–0.03), while some show an increase over time (in case (that) 0.02–0.04–0.06–0.09), but there are also phrases that decrease in frequency (by reason (that) 0–0.09–0.04– 0.02) after a peak in the first half of the sixteenth century. It is not possible for me to go into details here, but I would like to stress that the spread of items of this type does not suggest that their use is conditioned by genre, so that discoursal preferences may be related to a formal communicative setting. In the Corpus of Scottish Correspondence in respect (that) is used in the following way: (2) I think he was asshamed to enter with you in respect \ he did not creue your opunione nor your \ housbands in his vp coming (CSC 1616 Lady Anna Livingston, Countess of Eglinton, NLS Adv. MS 33.1.1., vol. 10:120) (3) I doutt not bot our brother \ Jhon cann gei? {torn} zour lo gritter assurance thairof, in respect he hes seinne, whatt hes \ beine my estaitte thir dayes bygaine (CSC 1613 George, 3rd Earl of Winton, NAS GD 3/5/50) (4) I am sorie \ sister that ze soulde geif me thanks whar ther is \ non deseruid be me Inrespek that I deid nothing \ bot that vhilk vas my bound deutie to do (CSC 1617 Lady Isabella Seton, Countess of Perth, NAS GD 3/5/133) (5) pleis zour \ la that this zung gentilman, my sister the Lady carnebeis \ sone being desyrus to be placit in service vt sum nobilman, and \ hes villit me to mak my moyen for the saming, in respect his \ father hes sumquhat outschot him self in misgoverning his \ rent, and I accounting zour la and {torn} husband \ as the most speciall freinds that I heaue to imploy vill most \ affectuiesly requeyst zour la that ze vill place this zung man \ ather in my lord zour husbands chamber to put on or af his \ cloths {ins} and vait on his lo {ins} or in qt vther office zour la thinks meit for such a \ ane, and I vilbe ansuerable to zour la that he sal {damaged} be faith\ful and obediant to my lord or zour la in qtsoeuer ze im ploy \ him (CSC 1612 Dame Grissall Roos, Lady Keir, NAS GD 3/5/32)
148 Anneli Meurman-Solin
Table 9 provides a more detailed account of verbs as a source of adverbial subordinators and information about the chronology of their spread in Older Scots. Table 9.Occurrences of some conjunctive phrases using verbs as source in the HCOS. TM = mean frequency of the total. Conjunctive phrase
1450–1500 1500–1570 1570–1640 1640–1700 Total
TM
notwithstanding (that) suppose (that) conditioned (that) giving (that) given (that) providing (that) provided (that) seeing (that) considering (that)
1 18 – – – – – – 3
13 9 1 3 1 31 – 3 14
14 18 – – – 22 – 80 20
12 5 – – – 24 5 50 47
40 50 1 3 1 77 5 133 84
0.05 0.06 0.00 0.00 0.00 0.09 0.01 0.16 0.10
Total TM
22 0.23
75 0.37
154 0.51
143 0.58
394 0.47
0.47
While the general frequency of subordinators derived from verbs is greater than that of incorporating nouns, we can see a similar chronology in their spread. This is even more obvious if we exclude the non-participial suppose (conjunctive participles: 0.04–0.33–0.45–0.56). The table also shows that past participles are very rare. There are no occurrences of granted (that), outtaken (that), during (that), excepting (that) or supposing (that), listed in Kortmann’s inventory of Early Modern English adverbial subordinators (1997: 293). The high-frequency items providing and considering rarely occur in letters, seeing (that) being used instead. In fact, in Scots seeing (that) was introduced into mid-sixteenth-century official letters and has not been attested in other texts in this period in the HCOS. In the 1570–1640 period its mean frequency is 1.2/1,000 in letters, that of all texts representing this period being as low as 0.3 (for further information, see Meurman-Solin and Pahta forthcoming).
7. Concluding remarks I would like to conclude by relating this brief summary to the three general points I made at the beginning. I have only briefly referred to problems related to categories in the analysis of connectives in complex sentence structures,
From inventory to typology in English historical dialectology 149
pointing out that a full-scale manuscript-based electronic corpus is required to deal with such problems. At the moment, my corpus based on diplomatically transcribed manuscripts is too small to allow such an endeavour. When the tagged databases being compiled are available we will have better ways of analysing developments over time in the system of clause linkage. The introduction of modern punctuation into text editions has of course fatally distorted evidence in edition-based corpora, however representative these may be by other criteria. It is also necessary to examine texts in manuscript to understand how each writer structures the text, whether the clause boundaries have been left unmarked or devices such as capitals or spaces have been used to mark them (for illustrations see Meurman-Solin 2004). The complexities of clause linkage systems may be difficult to deal with appropriately without annotating the manuscript-based texts with tags providing information about the lexicogrammatical properties of linguistic items. We may not succeed in identifying all the items entering patterns of syntactic variation in any other way. The reanalysis of category membership may necessitate redefining conjunction plus adverb collocates, and positioning them in our subordinator inventories in a semantically appropriate way. Polyfunctionality will have to be paid due attention. I have referred to Kortmann’s typology of adverbial subordinators in European languages to arouse our interest in questions that remain unanswered, not to suggest that I find the comparative perspective timely. We cannot compare before we have managed to describe what we want to compare fully. The relevance of a typology depends on the quality of inventories available for creating it.
References Anderson, John M. 1997. A Notional Theory of Syntactic Categories. Cambridge: Cambridge University Press. Auwera, Johan van der, ed. 1998. Adverbial Constructions in the Languages of Europe. Berlin & New York: Mouton de Gruyter. Benson, Phil. 2001. Ethnocentrism and the English Dictionary. London & New York: Routledge. Brinton, Laurel J. 2000. “The Importance of Discourse Types in Grammaticalization: The Case of Anon”. Textual Parameters in Older Languages ed. by Susan C. Herring, Pieter van Reenen & Lene Schøsler, 139–162. Amsterdam & Philadelphia: John Benjamins. Croft, William. 1990. Typology and universals. Cambridge: Cambridge University Press. Crowley, T. 1990. “That obscure object of desire: a science of language”. Ideologies of Language ed. by J. E. Joseph & T. J. Taylor, 27–50. London: Routledge.
150 Anneli Meurman-Solin
Dareau, Marace. 2004. “DOST: a significant instance of historical lexicography”. New Perspectives on English Historical Linguistics (2): Lexis and Transmission. Selected Papers from 12ICEHL (Glasgow, 21–26 August, 2002) ed. by Christian J. Kay, Carole A. Hough & Irené A. W. Wotherspoon, 49–64. Amsterdam & Philadelphia: John Benjamins. Devriendt, Betty, Louis Goossens & Johan van der Auwera, eds. 1996. Complex Structures. A Functionalist Perspective. Berlin & New York: Mouton de Gruyter. DOST = A Dictionary of the Older Scottish Tongue, ed. by William A. Craigie, A. J. Aitken et al. Aberdeen: Aberdeen University Press & Oxford: Oxford University Press. Foley, William A. & Robert D. Van Valin Jr. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Haiman, John & Sandra A. Thomson, eds. 1988. Clause combining in grammar and discourse. Amsterdam & Philadelphia: John Benjamins. Haumann, Dagmar. 1997. The Syntax of Subordination. Tübingen: Max Niemeyer. Häcker, Martina. 1999. Adverbial Clauses in Scots. A Semantic-Syntactic Study. Berlin & New York: Mouton de Gruyter. Jackendoff, Ray. 2002. Foundations of language: brain, meaning, grammar, evolution. Oxford: Oxford University Press. Jones, Charles, ed. 1997. The Edinburgh History of the Scots Language. Edinburgh: Edinburgh University Press. Kortmann, Bernd. 1997. Adverbial Subordination: A Typology and History of Adverbial Subordinators Based on European Languages. Berlin & New York: Mouton de Gruyter. Laing, Margaret. 1993. Catalogue of Sources for a Linguistic Atlas of Early Medieval English. Cambridge: D. S. Brewer. ———. 2000. “Never the twain shall meet. Early Middle English — the East-West divide”. Taavitsainen et al. 2000. 97–124. ——— & Keith Williamson. 2004. “The Archaeology of Medieval Texts”. Categorization in the History of English ed. by Christian Kay & Jeremy Smith. Amsterdam: John Benjamins. Lehmann, Christian. 1988. “Towards a typology of clause linkage”. Haiman & Thomson 1988. 181–225. McIntosh, A., M. L. Samuels & Michael Benskin. 1986. A Linguistic Atlas of Late Mediaeval English. Aberdeen: Aberdeen University Press/Edinburgh: Mercat Press. Matthiessen, Christian & Sandra A. Thompson. 1988. “The structure of discourse and ‘subordination’”. Haiman & Thomson 1988. 275–329. Meurman-Solin, Anneli. 1995a. The Helsinki Corpus of Older Scots (HCOS). Distributors: Oxford Text Archive, Norwegian Computing Centre for the Humanities. http:// www.hd.uib.no/corpora.html ———. 2001a. “Structured Text Corpora in the Study of Language Variation and Change”. Literary and Linguistic Computing 16:1.5–27. ———. 2001b. “Women as informants in the reconstruction of geographically and socioculturally conditioned language variation and change in 16th and 17th century Scots”. Scottish Language 20.20–46. ———. 2002. “Simple and complex grammars: The Case of Temporal Subordinators in the History of Scots”. Variation Past and Present. VARIENG Studies on English for Terttu Nevalainen (= Mémoires de la Société Néophilologique de Helsinki, 61) ed. by Helena
From inventory to typology in English historical dialectology
Raumolin-Brunberg, Minna Nevala, Arja Nurmi & Matti Rissanen, 187–210. Helsinki: Société Néophilologique. ———. 2004. “Towards a Variationist Typology of Clausal Connectives. Methodological Considerations Based on the Corpus of Scottish Correspondence”. Methods and Data in English Historical Dialectology ed. by Marina Dossena & Roger Lass. Bern: Peter Lang. ———. forthcoming a. Corpus of Scottish Correspondence (CSC), 1500–1800. ———. forthcoming b. Corpus of Early Scottish Women’s Writings (CESWW). ——— & Päivi Pahta. 2004. “Circumstantial adverbials in discourse: A synchronic and a diachronic perspective”. Language and Computers: Proceedings of the ICAME conference, Guernsey 2003 ed. by Antoinette Renouf. Amsterdam: Rodopi. Milroy, James. 1999. “The consequences of standardization in descriptive linguistics”. Standard English. The Widening Debate ed. by Tony Bex & Richard J. Watts, 16–39. London: Routledge. Moessner, Lilo. 1997. “The Syntax of Older Scots”. Jones 1997. 112–155. Rissanen, Matti. 1999. “Syntax”. The Cambridge History of the English Language, Vol. 3, Early Modern English 1476–1776 ed. by Roger Lass, 187–331. Cambridge: Cambridge University Press. Taavitsainen, Irma, Terttu Nevalainen, Päivi Pahta & Matti Rissanen, eds. 2000. Placing Middle English in Context. Berlin & New York: Mouton de Gruyter. Thompson, Sandra A. & Robert E. Longacre. 1985. “Adverbial Clauses”. Language and typology and syntactic description, vol. II: Complex constructions ed. by Timothy Shopen, 171–234. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs & Bernd Heine, eds. 1991. Approaches to Grammaticalization, vol. 1: Focus on theoretical and methodological issues. Amsterdam & Philadelpia: John Benjamins. Williamson, Keith. 1992/93. “A Computer-aided Method for Making a Linguistic Atlas of Older Scots”. Scottish Language 11/12.138–73. ———. 2000. “Changing spaces: Linguistic relationships and the dialect continuum”. Taavitsainen et al. 2000. 141–179. ———. 2001. “Spatio-Temporal Aspects of Older Scots Texts”. Scottish Language 20.1–19.
151
Consumers of correctness Men, women, and language in eighteenth-century classified advertisements Carol Percy University of Toronto
1.
Introduction
The rising social value of English and of correct English in eighteenth-century Britain is sometimes statistically illustrated: the number of new grammar books increased in the third quarter of the century, decade by decade (see Table 1). Some grammars prescriptively labelled linguistic variants as correct or incorrect and correlated them with such social attributes as politeness and vulgarity (see, for example, Sundby, Bjørge & Haugland 1991: 38–53; Fitzmaurice 1998: 320; Watts 2002: 156–159). In a period epitomized by new wealth and social mobility (Holmes & Szechi 1993: 153–154; McKendrick 1982: 23–24), good English has thus been seen as a symbol of social advancement especially for those to be known as “the middle classes” (Raven 1992: 152–153; Crowley 1996: 73–81; Fitzmaurice 1998: 325–327). While not contradicting this picture, my paper animates it by drawing on contemporary classified advertisements. A well-established, common, and respectable medium for seeking employment (Hecht 1956: 32–33), the eighteenth-century newspaper advertisement provides some evidence for what perceived value English had relative to other languages, and for whom. Table 1.Publication rate of grammar books 1740–1780 (see also Tompson 1977: 83). Source
Decade
1740s
1750s
1760s
1770s
Michael (1987: 12) (adapted from)
6
10
20
30
Raven (1992: 153)
–
–
At least 22
At least 28
154
Carol Percy
The audience of a grammar could be invited by a title page or an advertisement. For instance, an advertisement for The British Grammar, written anonymously by the Scot James Buchanan and published by his fellow Scot Andrew Millar, added to the already broad audience mentioned on the title page (“the Schools of Great Britain and Ireland … private young Gentlemen and Ladies”) “grown Persons … without a Master” represented as motivated to teach themselves grammar (Public Advertiser [PA] 1762: Nov 8, 1, 1). In 1775 Ellin Devis anonymously published her Accidence … for Young Ladies. Can we learn more about these consumers of correctness? The importance of grammar to well known individuals is well documented — grammar preoccupied Scots like the novelist and Critical reviewer Tobias Smollett and ambitious autodidacts like the future grammarian William Cobbett. It is essential to remember that advertisements reflect desires rather than reality, and that (with the exception of advertisements posted by named educational entrepreneurs like Thomas Sheridan or the lesser-known Reverend John Trusler) we can never know anything about the background or the future of the advertisers. Nevertheless, the advertisements surveyed in this paper suggest some functions and value of good English for less extraordinary individuals, especially but not exclusively in the field of education. 2. Data and methodology: The Daily Advertiser in 1750, 1762, and 1775 Newspaper advertisements vividly reflect and construct contemporary desires and practices in a nascent consumer society (McKendrick, Brewer & Plumb 1982). For Plumb, advertisements for schools witness the importance, “especially for the commercial classes”, of education as an instrument of their children’s advancement and a symbol of their own (1982: 292). Although the number and nature of advertisements for schools cannot reflect actual practice, their content identifies trends. For instance, surveying school advertisements in Jackson’s Oxford Journal Skedd observes how many female educational entrepreneurs derived (or hoped to derive) some “prospect of independence” from “the expansion of girls’ schooling” after 1760 (Skedd 1997: 102). Advertisements appeared in the many London newspapers, some “most at the west end of town”, others “most in the city” (Anon. 1984 [Trusler 1790]: 37). “Certain newspapers concentrated on particular types of advertising: … in the late eighteenth century, the Public Ledger had more shipping adverts, the Morning Post focussed on auctioneers and the Morning Chronicle on books” (Barker 2000: 97; see also Barker 1998: 33).
Consumers of correctness
An outstanding single source of information for tracking such trends in the London employment market is the Daily Advertiser (DA). Established in 1730 (Morison 1932: 123–125) and described in 1790 as among “those most in circulation” and as the “best calculated for advertisements” (Anon. 1984 [Trusler 1790]: 37), the DA is a source that has already been tapped by modern historians charting changes in servants’ wages (Hecht 1956: 141–157) or the careers of peripatetic educators and scientific lecturers like James Ferguson (Millburn 1988: 299 n8). Advertisements from the DA predominate in Andrew’s “overview of 60 years of” published “appeals for assistance in the eighteenthcentury London press” (1998), a genre that (as I shall illustrate) overlaps slightly with that of the classified advertisement. Determining the demographics of newspaper readership is difficult (e.g. Harris 1987: 191–195; Barker 1998: 22–42). Published in the City and commercial in content, the DA (“never … a first-class newspaper”) might reflect and appeal to the desires of “the commercial classes” (Morison 1932: 125,148) who, for Raven, number among the consumers of taste and correctness (1992: 151–154). Yet a wide range of eighteenth-century readers (like modern historians) must have recognized the DA’s usefulness as a source and read it when necessary; as a respectable and efficient medium for matching domestic servants with employers, it would have attracted readers of many ranks (cf. Hecht 1956: 32, Barker 1998: 33). For this paper I have collected advertisements that feature English (spoken as well as written) that is in some way superior: for example, teach or write English “grammatically”, speaks “correctly”, “with Propriety”, “no hopeless Brogue”, “spells well”. I do not consider writing per se: many clerks for instance need to “write well”. I have sampled classified advertisements at three equallyspaced (twelve and thirteen years) points in time: 1750, 1762, and 1775. For these years, I have read through four months of the DA: the winter months of December, January, February and March furnish a particularly high number of advertisements placed by booksellers exploiting the holiday season and schools and teachers anticipating the beginning of term (cf. Skedd 1997: 107). My expectations have been shaped by trends in bookselling and education: for instance, Skedd’s survey of advertisements for girls’ schools indicates that the “basic school curriculum of reading and needlework” “expanded … to include lessons in grammar … both English and French” (1997: 121). The paper principally focuses on classified advertisements; most feature teaching in private families and schools. Because of the relative infrequency of relevant advertisements, the paper illustrates broad trends with detailed analysis.
155
156 Carol Percy
While assessing the DA as a source, I surveyed other papers for the same winter months in 1762 and 1775, and other years from the DA. Advertisements from these furnish quotations but do not contribute to the tables of trends in the DA. More upscale papers are represented by The Public Advertiser and the St. James’ Chronicle (Morison 1932: 148,151–153; Barker 1998: 25–27). In 1762 the Gazetteer was more commercial than the former two, but by 1775 itself reflected the commercial tendency to emulate higher-class culture: theatrical advertisements appear after 1769 (Haig 1960: 65–67; Barker 1998: 25–27). I chose the dates 1762 and 1775 deliberately: those years saw the public consumption of key books and lectures about English. Its market value by 1762 is demonstrated overwhelmingly, not only by the publication of such grammars as Robert Lowth’s and Buchanan’s, but also by the recent appearance of a grammar by Joseph Priestley (1761) and the ongoing elocutionary activities of Thomas Sheridan. In 1775, Ellin Devis wrote and published the first grammar exclusively for “young ladies”. Twenty-five years earlier, another woman grammarian, Ann Fisher, had published a very influential grammar containing exercises of bad English to be corrected (see Michael 1987: 325–326). The activities of authors like Fisher, the commissioning of texts like Samuel Johnson’s dictionary by booksellers (Tieken-Boon van Ostade 2000), and the assiduous correction by Samuel Richardson of the grammar of the second (1741) and subsequent editions of his 1740 novel Pamela (Eaves & Kimpel 1971: 108–110) all confirm that some individuals were highly self-conscious about correctness well before 1750. To what extent are authors’ and booksellers’ activities reflected in contemporary classified advertisements? And what trends are evident in snapshots taken in 1750, 1762, and 1775?
3. 1750: English in a cosmopolitan capital Despite the importance of grammatical English to some authors and the prominence of “English” in many school advertisements in the DA, in advertisements for employment English per se is not a prominent commodity for the DA’s readers in 1750. One advertisement that for modern linguists may signal a fear of provincial accents might for contemporaries have reflected a fear of Jacobites not long after the rebellion of 1745: ANY Person of a sober Life and Conversation, that can teach Latin, and write a good Hand, properly qualified to tutor young Gentlemen; if a Clergyman,
Consumers of correctness
may stand a good Chance for Preferment; may hear of an Employ by applying to Mr. Simpson, Distiller, at the Ship and Still on Snow-Hill. Note, ’Tis to go abroad, and the Person to be sent must be neither Irish nor Scotch. (DA 1750: Jan 2, 3, 2)
The well-documented prejudice against “the swarm of Scotsmen” and Irishmen in London (e.g. Langford 1989: 323–329) should remind us of the city’s cosmopolitanism. And although English was stereotypically rising in status through the eighteenth century, its value on the employment market needs to be assessed in a broader context. English was not yet a lingua franca. In my small sample, it is foreign languages rather than English that are described as being taught or spoken “with Propriety” or “Purity”. The few explicit advertisements for good English reflect not the superior status of certain varieties of the vernacular, but the competence of a bilingual or multilingual speaker and the importance of French and other modern languages for commerce and culture. Although French was sometimes attacked as a useless accomplishment, it was used in international commerce by merchants and clerks and on the Grand Tour by employers and by their servants. Essential to the education of elite eighteenth-century boys and girls (Cohen 2003:99–100), French was also taught at numerous private academies to children of the middling classes (Cohen 2003: 102) and also to aspiring apprentices or upper servants (Hecht 1956: 61–62). While an upper servant’s literacy in English signaled his or her ability to supervise other servants and run a household (e.g. Hecht 1956: 38–43; DA 1750: Feb 5, 3, 1; Mar 13, 3, 1), advertisements confirm the importance of French for persons ambitious to become upper servants like valets and ladies’ maids. “A Sober young Person” “who can speak French” hopes thereby to avoid “too much hard Labour” by her ability to speak it (DA 1750: Dec 12, 3, 1). Some advertisers admitted to speaking only “a little of the French tongue” (DA 1750: Jan 1, 3, 2; cf. Hecht 1956: 46). For the more ambitious, there was evidently a market for linguistic instruction (Hecht 1956: 61). One teacher who explicitly solicited a market of would-be “Lady’s Maid[s]” was Mrs Le Bright, “a Complete Mistress of the French Tongue (DA 1750: Dec 12, 3, 1). With the French language in such demand, it is unsurprising to find advertisements attempting to discourage un(der)qualified teachers of it. Some prospective employers specified that the French teacher speak “the French Tongue in its (greatest) Purity” (DA 1750: Jan 26, 2, 3; Feb 6, 3, 1; see also Feb 14, 3, 1). A reliable and common indicator of a French teacher’s education was a classical education:
157
158
Carol Percy
WANTED, A Native of France to teach in a Boarding-School near Town. To prevent Trouble, ’tis expected that no Person apply who does not speak English currently, and is not a Classical Scholar … (DA 1750: Dec 29, 3, 1; see also e.g. Jan 15, 3, 1; Mar 22, 2, 3; Mar 23, 3, 1)
A person who could “speak English currently” could certainly communicate with English-speaking students (see also DA 1750: Feb 2, 3, 1). And a person who could teach English was likely to be educated: another school solicits in addition to a classics teacher a “French Master (a Native of France) capable of teaching that Language with Propriety, and English if required” (DA 1750: Jan 4, 3, 2). Prospective teachers duly emphasized their educational qualifications: multilingualism demonstrated present usefulness as well as past schooling. Men featured their competence in the classics (e.g. DA 1750: Mar 12, 3, 1; Dec 15, 3, 3). One “Young Gentlewoman” seeks employment either as a governess or at the top of the female domestic service hierarchy, as a waiting woman (Hecht 1956: 60–70). Her professed status as “Gentlewoman” is epitomized by her “genteel Education” and specifically by her ability to “talk French and English extremely well”. That French is listed first signals its superior status; that English is mentioned suggests the advertiser’s genteel awareness of the importance of good English, or her ambitious self-presentation as a particularly cultivated native speaker of French. The claim to be able to speak “French and English” is a better indicator of education than of nationality or rank. A Young Gentlewoman of a genteel Education, that talks French and English extremely well, and works at her Needle, would be very willing to wait on a Lady as her Woman, or be Governess to young Ladies. If wanted, please to enquire or direct to Mrs. Callais, Sackmaker, in Vine-Street, Piccadilly. (DA 1750: Feb 23, 3, 1)
The same advertisement appears three days later. While other details remain unchanged, the adverbial phrase “extremely well” has been omitted (DA 1750: Feb 26, 3, 1). The retraction, however motivated, confirms that linguistic skills were susceptible to inflation and therefore marketable in mid-century London. In my sample, though, explicit references to good (and adequate) English generally involve multilingual, non-native speakers.
Consumers of correctness 159
4. 1762: English proper, properly By 1762, classified advertisements suggest that superior English is more explicitly demanded of native speakers, although the value of a speaker’s vernacular is sometimes correlated with his or her command of Latin or French (see Table 3). In newspapers, more prominent indicators of a demand for grammar and correctness are advertisements placed by booksellers and educational entrepreneurs, who were vigorously selling superior varieties of the vernacular — written and spoken — to a wide spectrum of British society. While the audience of Lowth’s anonymous Short Introduction to English Grammar was constructed as able to read the Latin epigraph that aggrandized the grammar’s title page and the English it codified, the “Compendious English Grammar” “prefix’d” to Dyche and Pardon’s New General English Dictionary claimed to enable “such as understand English only” to “be able to write as correctly and elegantly as those who have been some Years conversant in the Latin, Greek, &c. Languages” (General Evening Post 1762: Jan 5–7, 4, 3). The author and private teacher Daniel Farro taught even “such who can read but indifferently, to divert themselves in the Use of his Royal English Grammar in one Quarter of a Year” (DA 1762: Jan 7, 3, 3). The latter advertiser’s emphasis on speed and easiness suggests that grammar’s value derived in part from its difficulty. Merely “to read and spell English well” was less impressive — or at least one prospective employer soliciting a “middle aged Man” to “teach the Children” at “a large Boarding-School” “to read and spell English well” rationalizes that his “Salary will not be large” because “no great Abilities are here required” (DA 1762: Dec 24, 3, 2). A few school advertisements explicitly offer grammar to an apparently commercial audience (cf. Raven 1992: 138–156; Fitzmaurice 1998: 325–326). The academy keeper William Wood taught both “all useful Branches of the Mathematicks, and the English Grammar”, the latter taught “to Natives and Foreigners” “in a very familiar easy Way” when Wood was not “expeditiously qualif[ying]” “young Gentlemen … for the Compting House” (DA 1762: Dec 10, 1, 3). The importance of grammatical correctness to girls, perhaps of the commercial classes, is suggested by an advertisement for the services of the Reverend Mr. Porter. To his regular curriculum of applied mathematics for boys he briefly (see DA 1764: Jan 2, 3, 3) added English grammar for girls: THE Rev. Mr. PORTER, at his House in Stanhope-Street, Covent-Garden, continues to instruct Gentlemen in Geometry, Trigonometry, Algebra, Navigation, Fortification, Gunnery, Geography, the Use of the Globes, and
160 Carol Percy
other Branches of the Mathematicks. He also instructs Ladies in reading Pope, Milton, Shakespear, and other English Poets, and teaches them to write correctly, by explaining to them the Principles of Grammar in English… (DA 1762: Dec 2, 3, 1)
By extending his clientele to “Ladies” who want to read literature aloud and write correctly, Porter may have been attempting to emulate the success of Thomas Sheridan, who since the late 1750s had lectured on elocution and education around Britain (Benzie 1972: 20–34) and who advertised frequently and widely in newspapers. Ladies could attend Sheridan’s lecture series for half price: the subscription fee of a guinea would “admit one Gentlemen, or two Ladies” (PA 1762: Jan 22, 1, 3). Sheridan’s lower rate for ladies may have acknowledged their lower incomes, or the impropriety of their attending alone. He may also have encouraged women’s presence, which signaled (at least in theory) social refinement and intellectual accessibility. One index of Sheridan’s popularity is his imitation not only by entrepreneurs like the Reverend John Trusler (cf. DA 1762: Feb 16, 2, 3; PA 1762: Mar 15, 4, 1) but also by the satirical dramatist Samuel Foote (PA 1762: Apr 29, 1). In his satire The Orators, Foote epitomizes the audience for Sheridan’s lectures with Mr Suds, a City soap-boiler with political aspirations and a social-climbing, newspaper-reading wife (Foote 1762(1788):2.7–8). What practical applications might English have had for the real people to whom these entrepreneurs appealed? The value of the vernacular for some Londoners is increasingly confirmed by contemporary classified advertisements. Properly accented and grammatical spoken English might have been important both to would-be politicians like Suds and also to upper servants who might have to read aloud to and write for their employers (Barker c1770: 12; Hecht 1956: 61). My sample yields a few relevant advertisements. An aspiring young Companion would find it “extremely agreeable to her to serve an elderly Gentlewoman” and “read to her, write for her, and do such other Business as is suitable to her Capacity” (DA 1762: Feb 1, 3, 1). One employer valued the ability to “read well” in a male upper servant: WANTED, a Person who can Write and read well and understands BookKeeping; he is to serve in the Capacity of a Butler, as far as regards the Care of the Cellar and Plate, and Attendance at the Sideboard, but must spend much of his Time in writing for, and reading to his Master. It is expected that he has a decent Figure. He would be the more agreeable if he speaks French, and can dress hair, but this is not insisted on … (DA 1762: Dec 1, 2, 3)
Consumers of correctness
Asserting his own taste, the advertiser expects the man to have a “decent Figure” and hopes that he can speak French (Hecht 1956: 53–55). A few hopeful employees include English among their commercial qualifications. Outside my sample, a boy seeking “an Apprenticeship to a Business by which he may get a genteel Livelihood” with a “reputable Tradesman” specifies that he “has been taught Grammar” in addition to the more typical commercial accomplishment of “writ[ing] a good hand” and “know[ing] Figures well”; here, grammar may symbolize the boy’s thorough education (DA 1764: Jan 4, 3, 3). English more often features in advertisements placed by multilingual men seeking commercial work; as in 1750, references to good English most likely promise competence rather than superiority. Advertisers characterize their competence with terms like “understands” or “well versed in” or “Master of”. One “Gentleman” “bred to Business” for “Part of his Time” (“writes a good Hand, and understands Merchants Accounts”) “has lived some Years in France and Spain” and is “well versed in the English, French, and Spanish languages” (DA 1762: Dec 21, 2, 3); we can infer that he is a native speaker of English. Another advertisement, a variation on advertisements for multilingual clerks, similarly features the advertiser’s character and his business and linguistic skills. It is unusual in that it emphasizes his specific experience in the weaving industry and distinguishes his ability to “understand Book-Keeping and the Italian language” from his ability to “speak and write both French and English correctly”: TO the Master Weavers. A Person whose Character can bear the strictest Enquiry, who can speak and write both French and English correctly, understands Book-Keeping and the Italian Language, and has served a regular Apprenticeship in a noted Silk Weaving Manufactory, where he continued several Years Head Clerk and Foreman, would be glad to be employ’d by some eminent Weaver, to oversee the Workmen, keep the Books, &c. Whoever this may suit, either now, or some Time hence, please to direct a Line for A. B. at Mr. Tournier’s, in Spital-Square, Spittlefields, and it shall be answered in a few Days. (DA 1762: Mar 19, 3, 1)
The advertiser may have flattered his intended audience not only by characterizing them as “eminent” or “Master” but also by constructing them as attentive to correctness, both French and English. The currency of English can be contextualized in the teaching of modern languages. Good English continues to indicate good French teaching (DA 1762: Feb 26, 3, 1; Mar 30, 3, 1). English, unlike French, is in my sample not yet advertised as taught “grammatically” — in the case of French by “rule” rather than by “rote” memorization of passages, as Cohen shows. The adverb promised
161
162 Carol Percy
a methodical and thus efficient initiation to learning: grammar was marketed by pedagogues like Peyton (1757) as a “golden Key” to other learning (Cohen 2003: 107–109). As a gateway to learning, grammar must also have been fashionable. “Polite” and “grammatical” are collocated twice in my 1762 sample: e.g., “to teach that language grammatically, and to speak it politely”; “grammatically, in the politest manner” (DA 1762: Jan 20, 3, 1; Dec 11, 3, 1). That this invitation to initiation might appeal particularly to women, as Cohen notes, is suggested in my 1762 sample by the absence of references to French being taught “grammatically” by or to females (see Table 2).1 In an advertisement from 1760, however, one “French Boarding School” asserts that it has been “brought into a regular and concise Method [my emphasis], where young Ladies are boarded, and educated in the English and French Languages, and all Sorts of Needlework…” (DA 1760: Jan 11, 1, 3). This “regular and concise Method” encompasses both “English and French”. In the advertisement below, the value of English to parents who would prize both Accomplishments (“Dancing (and Drawing, &c. if required”)) and economy (“taught cheaper than at any other School”) is apparent in the way that English twice precedes French: THE French Boarding School in Pond-Street, Hampstead, now there established, is brought into a regular and concise Method, where young Ladies are boarded, and educated in the English and French Languages, and all Sorts of Needlework, at the respective annual Sums hereafter mention’d, to wit, from the Age of ten Years and upwards at Fourteen Guineas and Two Guineas Entrance, from the Age of seven Years to ten Years Twelve Guineas and Two Guineas Entrance, from the Age of three Years to seven Years at Twelve Guineas and One Guinea Entrance. Writing and Dancing (and Drawing, &c. if required) taught cheaper than at any other School in or near London. There are proper Teachers for the English, and the Governess of the School and the French Teacher, are Natives of France. …(DA 1760: Jan 11, 1, 3)
The assertion that “[t]here are proper Teachers for the English” implies that this cannot be taken for granted. In my 1762 sample, teachers are twice described as capable of teaching English “properly” or “with Propriety”. Some schools do not identify qualifications for “proper Teachers” (above and DA 1760: Jan 5, 3, 1). On other occasions, prospective teachers’ English derived value by association with Latin
1.My earliest reference from the DA to a woman “capable of teaching [French] grammatically” appears in 1763 (DA 1763: Mar 23, 2, 3; see also Dec 27, 2, 3).
Consumers of correctness 163
(Watts 1999, 2002) or French (see Table 3), signaling both learning and versatility: “WANTED, in a Boarding-School pleasantly situated near Town, a Gentleman well qualified to teach English, and occasionally to instruct the lower Class either in Latin or French.” (DA 1762: Feb 20, 3, 1). Elsewhere a “Young Man” juxtaposes his abilities to “teach the Classicks” and “to read and write English properly”: A Young Man, of a liberal Education, who has devoted his whole Attention to instruct young Gentlemen, and whose Morals will bear Examination, would gladly engage himself in some reputable School to teach the Classicks, to read and write English properly; he has made some Proficiency in the Mathematicks and useful Sciences, which he will readily communicate when permitted. Any Gentleman who will please to honour him with his Commands, directed to A. B. at the Kensington Coffee-House, or Old Slaughter’s in St. Martin’s Lane, shall be waited on immediately. His Time being entirely vacant, he would be glad to employ Part of it in teaching two or three Boys at home or abroad. (DA 1762: Jan 15, 3, 1)
A “young Ladies Boarding School” seeks both a “well-bred French Teacher” and a “Person … capable of teaching the English Tongue with Propriety”. An implicit indicator of propriety is her ability to “speak French”. WANTED in a young Ladies Boarding School near London, a prudent middleaged well-bred accom[plished] Person, one that can speak French, and understands Needle-[work] &c. and is capable of teaching the English Tongue with Propriety. A genteel well-bred French Teacher is likewise wanted. None but Persons of the best Character and Qualifications need give themselves the Trouble to apply. (DA 1762: Dec 31, 3, 1)
By 1762, six years into the Seven Years’ War, some advertisers in the DA question the value of French. Conflicts about the content and purpose of girls’ education are much more explicit in (for instance) essays in contemporary newspapers (e.g. Gazetteer and London Daily Advertiser 1762: Aug 23, 1, 3–4; London Chronicle 1762: Jan 21–23, 76, 1–3; Jun 26–29, 612, 1–3; St James Chronicle 1762: Jan 14–16, 1, 4), but some advertisements for governesses also debate the education of girls destined for domesticity. One “Person who is not thought improper for the Care and Instruction of young Ladies” capitalizes on her monolingualism when she asserts her “ignoran[ce] of any Language but English” and appeals to “any Lady … who does not think a French Woman more proper for the educating her Children” (DA 1762: Jan 14,3,1). Such employers evidently existed. One advertisement for a “French Teacher” who is
164 Carol Percy
“not a fine Lady, but a solid, good-natured, useful Person, very capable” (DA 1762: Jan 5, 3, 1) suggests the difficulty of reconciling competence in French with general capability. Indeed, another prospective employer seeks a governess “perfectly Mistress of [the English] Language” who need not “understand French or French Fashions”: WANTED, a Governess for some young Ladies, a considerable Distance from London; no young Person from a Boarding-School need apply; she must be of a middle Age, have seen and been conversant with the polite Part of the World, but is better qualified to instruct young Ladies in the moral and essential Parts of Education, hath herself been a good English Housekeeper, and perfectly Mistress of that Language; the understanding French or French Fashions, not necessary …(DA 1762: Dec 29, 3, 1)
Idiosyncratic syntax links what is “moral and essential”, domestic (“English Housekeeper”), and vernacular (“that Language”) and separates them from what is “polite” and “French”. The assertively pro-English tone of these two advertisements typifies “the linkage of language and nation” described by Crowley as characteristic of “mid- to late eighteenth-century Britain” (1996: 68). French certainly retained its status for families and its function for those who could speak it. One family sought to have their daughter taught “both the French and English Languages, together with a polite Behaviour” by a “genteel, well-bred” “Protestant” “Woman”. Including English has several implications: if anglophones, the employers could be emphasizing the importance of English, or excluding uneducated francophones. The association of “Languages” with upward mobility is suggested by the employers’ economic status: the woman must both “instruct” and also “dress and wash” the “young Lady” (DA 1762: Feb 12, 2, 3). A “Middle aged” prospective employee of such a family features both her experience and her ability to instruct children in “both French and English”; she explains that “her Reason for taking this publick Method is, that she has but few Acquaintance” (DA 1765: Dec 19, 2, 3). For literate, economically vulnerable women, employment as an upper servant or elementary teacher had long been among the few respectable and accessible sources of subsistence (Hecht 1956: 16–20; Hill 2001: 63). For monolinguals the rising status of the mother tongue promised to add value to what in some cases may have remained limited skills.
Consumers of correctness 165
5. 1775: Widows, mathematicians, and “English grammatically” Well before 1775 — by or before the beginning of 1764 — English is advertised in the DA as being taught “grammatically”.2 Knowing “English grammatically” evidently has value, especially though not exclusively on the educational market. Although some advertisers solicit classically-educated English teachers, others expect that English can be taught “grammatically” by less learned instructors. A contemporary satire called The Academy Keeper (1770) implies that such male instructors had very little learning. The opportunities offered to some females by the rise in status of the mother tongue are suggested by a cluster of advertisements placed by widows in the winter of 1775. A few advertisements confirm both the value of vernacular accuracy in the commercial world and the importance of that commercial world to those with a classical education. One “young man … who has had a classical Education” and “is well versed in the English and French languages … would be glad to be employed in a reputable Counting-House” (DA 1775: Feb 20, 2, 3). Several hopeful clerks feature their correctness, their ability to “write a fast Hand, and spell well” (DA 1775: Feb 21, 2, 2), “write and … engross very neatly and correctly” (DA 1775: Feb 20, 2, 3), or “with Accuracy and Expedition” (DA 1775: Mar 13, 2, 2). “[A] Person who can write a quick, pretty, Law hand, grammatically” “may apply to Mr. Millidge, at the Printing-Office, Russell-Street, CoventGarden” (DA 1775: Feb 28, 2, 2). Although it is unclear whether Mr. Millidge is the prospective employer or merely a contact, another advertisement confirms the unsurprising importance in the printing business of correct English: To the Gentlemen engaged in the Printing Business. A Person (who has had a classical Education; and served an Apprenticeship in London as a Compositor) wishes to engage as Reader and Corrector of English Works with any Gentleman who may want such an Assistant. He flatters himself that his Diligence, Sobriety, and Abilities, can be sufficiently authenticated to any one willing to treat with him… (St James Chronicle 1777: Sept 4–6, 2, 2)
2.There were no references to English taught “grammatically” in my 1762 sample (JanuaryMarch, December). Kate Stevenson kindly checked the DA for January 1763 and January 1764. Kate found no examples from January 1763, and I found none from February, March, or December. In January 1764 “the Academy in Angel-Court, Westminster” announces that “young Gentlemen and Ladies are taught, and expeditiously instructed, in English grammatically, Latin and French in its Purity…” (DA 1764: Jan 6, 3, 2). Referring to the teaching of English, the adverb “grammatically” had appeared in English grammar books well before this date (see e.g. Stirling 1735: preface).
166 Carol Percy
Another advertiser links his classical education with his ability to “correct the Stile of Pamphlets”. That he does so “on moderate Terms, and with Secresy” [sic] suggests that in 1775 money-conscious clients might feel some shame about writing incorrectly: TUITION. English grammatically, Geography, French, Latin, Greek, and the Elements of Hebrew, taught abroad by a Gentleman, who, after having quitted a University, employed near 16 Years in intense Study, during which Time he published some well received Treatises on literary Subjects. His Terms may be known by applying or writing to T. H. No. 6, Catherine-Street, Strand. He also writes Letters, Advertisements, &c. corrects the Stile of Pamphlets, or other Performances intended for the Press, on moderate Terms, and with Secresy. (DA 1775: Dec 21, 3, 1)
By beginning his list with “English grammatically”, the university-educated gentleman seems to anticipate most demand for it and other “useful” subjects from a clientele who might nevertheless be impressed by learning. Most of the explicit demand for good English is unsurprisingly in the field of education. The importance of proper expression for boys across a socioeconomic range is suggested by advertisements for both private tutors and for teachers in academies. One advertisement solicits a “genteel” Gentleman “who can express himself with Propriety and Purity of Language” to teach “a young Master” bound for a public school “the first Rudiments of the English Tongue” (DA 1775: Jan 4, 2, 3; see also Mar 2, 2, 2). The importance of pronunciation at boys’ academies is suggested by advertisements seeking classics masters who “speak good English” (DA 1775: Mar 21, 2, 2), “[Latin, French, and English] with Fluency and Propriety” (DA 1775: Mar 16, 2, 2). Clustered advertisements specifically discourage Irish or Scottish accents and applicants: “No one who has the Scotch or Irish Accent” (DA 1775: Jan 28, 2, 3), “those who have not the true English accent need apply” (DA 1775: Feb 24, 2, 3). WANTED immediately, at an Academy near London, an Usher, who is Master of the English and Latin Languages, and exceptionally recommended for his Diligence, Sobriety, Integrity, and Temper. No Irish Gentleman can be accepted, nor any whose Dialect has a hopeless Brogue; as the Choice must be delicate, the Encouragement will be proportioned. (DA 1775: Jan 22, 2, 2)
Although the status of grammar is here suggested by the word “understands”, which in this context often connotes difficulty, not all advertisers looked for a “learned” applicant:
Consumers of correctness 167
WANTED an Usher to a School, who understands the English Grammar, is a very good Accountant, and who writes a very fine Hand; he must be well recommended for Sobriety and Diligence. (DA 1775: Feb 9, 2, 2; Feb 7,2,2)
We might conclude that grammar has become more widely understood by 1775. Some debasement related to the demand for male teachers (see Table 3) is implied in a contemporary satire. The Academy Keeper (1770) is provided with “approved copy” for advertisements: Wanted at an academy near London three domestics: A compleat penman, accomptant, and mathematician, with an undeniable character: A steady careful person capable of teaching the English language grammatically, and willing to attend the children to bed: A cleanly sober wench to look after the children’s linnen, and do other occasional work: Enquire of Mr. Twitch, broom-maker, in Kent-Street. (quoted in Best 1968: 391)
The debasement of grammar teaching is implied by the grammar teacher’s other duty, the domestic one of putting the children to bed, and is further confirmed by another hiring strategy in the event that a grammarian cannot be found: “Ask the mathematician if he understands English” (quoted in Best 1968: 391). That the academy keeper is himself a “broken attorney, or exciseman … disbanded Frenchman, or super-annuated clerk” or at best “an emigrant from some northern university, or a tuftless child of one of our own” (quoted in Best 1968: 390) confirms both that the commodification of education was perceived to have produced incompetent teachers and also that in an economically and socially volatile society teaching the socially mobile was one way for others to avert downward mobility. Advertisements suggest that correct English was important not only for ambitious parents of young children but also for economically vulnerable educated women. Although a widow might take over her husband’s business in order to “live in the Manner she used to do” (DA 1775 Mar 4, 2, 3), this was not always possible — for instance, for a clergyman’s widow. Middle-aged women lacked the youth and perhaps the capital for apprenticeship (e.g. to a milliner or a mantua-maker), among the few other genteel occupations (Hecht 1956:19–20). Some matrons and widows seeking dignified domestic work exploited their experience as household managers, especially for a single or widowed man; some advertised their ability to raise and teach young children — by no means a new profession for such women (Hill 2001: 55–56). One widow can “educate
168 Carol Percy
young Children as soon as they are capable of speaking” (DA 1775: Feb 17,1,3); others can teach children “to read” and “speak” “with … Propriety” (DA 1775: Feb 16,2,3; Mar 4, 2, 3). In a particularly rich advertisement, a clergyman’s widow markets herself as housekeeper or governess while vigorously asserting the social status of her husband’s position, her status as “Gentlewoman”, her education, and her “Cloaths and Address”: A Well-qualified, middle-aged Clergyman’s Widow, wants to be Companion and entire House-Manager to a single Gentleman or Tradesman, has been in that State, is a good Oeconomist, tender Nurse, and of chearful Temper; knows common Accounts, and a little French; none more proper to instruct Youth, having a strict Eye to their Morals and Behaviour; can read, teach, and spell English much better than most of her Sex; has Cloaths and Address fit to see and receive the best Company; Salary would leave entirely to the Choice of those she has the Honour to serve; is the Gentlewoman, and hopes to be used as such. It is begged none will enquire out of idle cruel Curiosity, at Mrs. Roe’s, Pastrycook, Great Tower-Street. (DA 1775: Jan 10, 2, 3)
Her insistence that she “can read, teach, and spell English much better than most of her Sex” highlights her perception of the importance of linguistic skills to potential employers — and perhaps her anxiety about the job prospects for most middle-aged women, however valuable maturity may have been in a household (cf. Hecht 1956: 64). Table 3 implies that there were more women than positions. That the clergyman’s widow mentions French quite early in her advertisement but admits to only “a little” confirms the continuing status of French. Like the clergyman’s widow, “A Middle-aged” former “Teacher” mentions French although she only “understands something of” it: A Middle-aged Person, who has been Teacher in several Boarding-Schools near London, is capable of teaching English grammatically, Mistress of Shading and all Sorts of Needlework, and understands something of French, is desirous of being employed in the above Capacity, or as Governess to young Ladies; has a great Affection for Children, having had several of her own. (DA 1775: Jan 20, 2, 3)
Another advertiser, a Jewish widow, “speaks French very fluently”: A Person of extraordinary good Character would be extremely glad to serve as Inspector of a Nursery (in some Jewish Family) and educating young Children as soon as they are capable of speaking, being perfect Mistress of the English Language, speaks French very fluently, understands all Sorts of Needleworks;
Consumers of correctness 169
she has but few Acquaintance, and her Husband dead, Town or Country will be agreeable. (DA 1775: Feb 17, 1, 3)
An apothecary’s widow asserts her gentility first of all by advertising her ability “to talk and teach the French Language grammatically”, and secondly by claiming (albeit ambiguously) that she would do so “chiefly for the Sake of Employment” (DA 1775: Mar 2, 2, 3). Similarly, some advertisements soliciting governesses mention French, and not English (e.g. DA 1775: Jan 23, 2, 3). The status of French for the younger generation is confirmed by the following young advertiser, who asserts her competence to teach both French and English: A Lady, 13 Years of Age, who has had a liberal Education, and resided some Time in France, understands both French and English grammatically, and speaks them correctly, would be glad to go into any genteel Family where there are some young Ladies, to whom she may probably be useful. (DA 1775: Mar 17, 2, 3)
As a recent product of the educational system, this young lady might be expected to be particularly sensitive to social and pedagogical trends: she puts French first. While attesting to the prestige of French, these advertisements also emphasize the importance of good English and English grammar in English culture. The French-speaking thirteen-year-old emphasizes her grammatical understanding of and spoken competence in English. Moreover, by 1775 both a young and a middle-aged woman claim to be able to teach English “grammatically”; my 1762 sample had no examples of English being taught “grammatically” and no examples of women claiming to teach French “grammatically” (see Table 2). The Jewish widow precedes her claim to speak French fluently with a description of herself as “perfect Mistress of the English Language”, another phrase common in advertisements for multilingual teachers. The middle-aged mother who “understands [only] something of French” advertises herself as “capable of teaching English grammatically”. Two other advertisers who cannot claim any competence in French exploit the status of correctly spoken English by appealing to parents who want their children taught both to “read” and “speak” English “with … Propriety”: A Gentlewoman who has lately lost her Husband, and in Consequence thereof unable to live in the Manner she used to do, would be happy to superintend the Family of a single elderly Gentleman, or a Widower with Children (she being particularly fond of them, and capable of teaching them to read and speak English with Propriety) no Objection to the Country, nor are Wages her principal Object, not entirely destitute; can have an undeniable Character from Persons of Credit. (DA 1775: Mar 4, 2, 3)
170 Carol Percy
In addition to this, “a sober steady Person” can also “make [Children’s] Cloaths” in addition to being “remarkably qualified to teach them English, as well to read as speak with the utmost Propriety” (DA 1775: Feb 16, 2, 3). These advertisements placed by women seeking such domestic service remind us that education (Hecht 1956: 16–23), and the correct English that epitomized it, as much prevented downward mobility as it symbolized upward mobility. There were positions for such women: WANTED a middle-aged Woman, as Governess, with an undeniable Character, to have the entire Care and Management of three young Ladies; she must be qualified to teach them English, French, Writing, and all Sorts of Needlework. Enquire at No. 12, in Devonshire-Square, Bishopgate-Street. (DA 1775: Feb 15, 2, 3)
But the competitiveness and desperation of genteel women seeking to keep “liv[ing] in the Manner [they] used to do” is suggested by the relatively high number of women placing advertisements (Table 3), by the claim of the “Clergyman’s Widow” to “read, teach, and spell English better than most of her Sex” (DA 1775: Jan 10, 2, 3), and also by the appearance ten days later of an advertisement almost identically worded, with a different contact address and with a few additions. If she cannot be “Companion and sole House-Manager” and instructor of “Youth”, she “in Spring should like going abroad; [and] would not dislike an Assistant’s Place in an eminent Boarding-School” (DA 1775: Jan 20, 2, 3). The importance of English is intensified by the tone of a few longer advertisements. In the age of sensibility some advertisers of both sexes deployed its language and tropes to emphasize their personal and economic reversals and thus to move readers keen to see themselves as benevolent (Andrew 1998: 100–101). We have seen that vernacular correctness could be claimed by women in distress. It is unsurprising that her ability to “read, write, and speak English correctly” is the first attribute featured by “[a] Person of genteel Family, and good Education” and “undeniable character” who has “met with misfortunes”: PRIVATE GOVERNESS. A Person of genteel Family, and good Education, having met with misfortunes, would be glad to under take the instruction of one or two young ladies. She reads, writes and speaks English correctly; understands French, the tambour and every sort of needlework, and will make it her study to advance the improvement of those committed to her care, and pay a strict attention to their whole deportment; her character is undeniable… (Gazetteer 1775: Mar 9, 3)
Consumers of correctness
Men could also appeal to the sympathy of readers, and could also exploit their readers’ evident desire for correct English. The “Gentleman” “confined by Lameness” with a “numerous Family in great Distress” appealed explicitly to those “humane Gentlemen of the Law, &c” who could appreciate his ability to “write” and “engross very neatly and correctly” (DA 1775: Feb 20, 2, 3). More men had more options: a clergyman who “cannot conscientiously officiate longer in the established church” can teach “the learned languages” and “arithmetic and merchants accompts” as well as the “mother tongue”, and indeed can deploy his knowledge of accounts by serving as a steward: A Clergyman Thirty One Years of Age, a single man, who has had a University Education, but cannot conscientiously officiate longer in the established church; would be glad to be taken into the family of any lady or gentleman, to whom he offers his service to instruct the children in the learned languages, or to teach them their mother tongue, arithmetic and merchants accompts. He has no objection against serving any person of what sect or persuasion soever, provided he can get into a regular and creditable family. If from a good knowledge in arithmetic, he can be thought useful in any other shape, in keeping merchants accompts, in acting as a steward, &c., he has no objection to a laborious employment. Any person sending proposals to Will’s Coffeehouse, Lincoln’s-inn-fields, directed to A. B. shall have an answer from the person who makes this application, and thereby the means of enquiring into his character and abilities. (Gazetteer 1775: Mar 18, 3, 1)
This advertisement confirms both the apparent appeal of a university education to potential parent employers, but also the educated advertiser’s awareness of his merchant readership, for whom his arithmetic and mother tongue may be as or more useful than the “learned languages”.
6. Discussion The increasing importance of English for “the middling sort” has sometimes been correlated with the publication of grammar books. This survey of classified advertisements from 1750 through 1775 has confirmed that English was of use beyond and within the educational realm, where it became advertised as taught “grammatically” to young children by nurses and tutors and schoolteachers. These developments confirm the value of vernacular grammar for London parents — and its importance to instructors, among whom we must remember to enumerate the economically vulnerable, especially women, whose educational
171
172 Carol Percy
Table 2.Individuals wanted (W) and available (A)a with explicit knowledge of grammar,b 1762 & 1775 (January-March, December). French and other modern languages
French and English
English
Male
Female ?
Male
Female
Male
Female
1762
W A
4 4
0 0
1 0
0 0
0 0
0 3
0 0
1775
W A
1 5
4 2
1 0
0 0
0 1
3 8
0 1
a
includes tuition offered by individuals at their schools includes “a Grammarian”, teaching “grammatically”, “Grammar”
b
Table 3.Qualifications of individuals wanted (W) and available (A; [at their] S[chool]) to teach English and Elocution, 1762 & 1775 (January-March, December). Multilingual
1762
1775
Monolingual Experience
Total
Classics
Classics French &/or French
W:8 A:1 S:5
4 1 0
1 0 2
0 0 0
3 0 3
0 1 0
Women W:3 A:4 S:1
0 0 0
0 0 0
2 3 0
1 1 1
1 1 1
Men
W:13 A:2 S:8
4 1 1
3 1 3
0 0 1
6 0 3
1 0 2
Women W:7 A:17 S:2
0 0 0
0 0 0
4 9 1
3 8 1
3 8 0
Men
experience and career prospects were particularly restricted. The promise offered to such women by the rise in status of the mother tongue might moderate Crowley’s picture of women’s language as “disciplined” and “police[d]” (1996: 87–92). The number of women teachers available (Table 3) suggests that their opportunities were limited. While this survey has illustrated the importance of commercial employment for the classically-educated, it has also reminded us of the continuing status of
Consumers of correctness
both French and Latin for some readers of the DA, slightly qualifying Crowley’s contextualization of the “pronounced” “shift towards the vernacular” in this period in an oppositional “association of Latin with the learned and leisured, and the vernacular with the mercantile class” (1996: 75–78). Although English could be taught “grammatically” by non-classically educated instructors, a teacher’s versatility and vernacular competence could be appraised by his or her command of other languages — French for women, the classics for men (see Table 3), confirming Watts’ construction of “polite English” as “educated English” (2002: 167–8), and of the most educated variety of English as that associated with Latin (1999: 51). To a great extent foreign languages like French had more symbolic value than practical use: some of the desperate widows who claimed to be able to teach children correct English admitted to only a little French. However, modern languages were evidently used not only abroad on the Grand Tour by employers and servants but at home in a commercial world where English was not yet the lingua franca. Any sociolinguistic picture painted of eighteenth-century London needs to capture its cosmopolitanism. The readership of DA is difficult to sketch specifically because the paper was a widely recognized source of advertisements and because the advertisements seem quite formulaic in comparison, for instance, to some of the longer ones in the Gazetteer. More specific social profiles might be produced by more rigorously classifying further classified advertisements, perhaps from a wider range of newspapers. In advertisements for schools and academies, what other subjects (classical, commercial, etc.) are correlated with “English grammar” as opposed to “English” or “Writing”, for instance? Did more potential clerks than employers specify an ability to “spell well” or “write … correctly”? How representative is the thirteen-year-old female grammarian? Is there more of a demand for young male than for old female teachers of grammar? In this qualitative pilot study I have tried to identify some broad trends. A more sophisticated picture requires both a larger sample and a denominator of all the advertisements which might, for instance, have advertised “English grammatically” as opposed to, say, “English”. The sample for any one “point” in time should also be broad enough to be representative. I have found for instance that the most memorable advertisements tend to cluster: clusters in the 1775 sample comprise the attacks on Celtic classicists and the appeals by widows. Yet, however ‘representative’ the sample, advertisements of all media remind us of the gap between representation and reality. As The Academy Keeper warns us, advertisements cannot tell us much about what was really happening in a classroom.
173
174 Carol Percy
7. Conclusion This qualitative pilot study3 has affirmed that eighteenth-century newspaper advertisements provide a valuable source of information about the social and economic impact of prescriptivism in London. As evident as the growing importance of an increasingly “grammatical” English in the home, school, and commercial workplace is the continuing importance of the classics and especially of French — not only as an indicator of the value of an individual’s vernacular or of his or her social status, but in the case of French as a medium of communication in a world where English was not yet the lingua franca. For men needing to earn a living, teaching English was one of many options. With the relative rise in status of English, increasingly evident in advertisements for children’s nurses, came the promise of opportunities for the more economically vulnerable — especially educated women, who had few other opportunities to avert downward social mobility in a world of new wealth.
References The Daily Advertiser [DA] The Gazetteer and London Daily Advertiser The General Evening Post The London Chronicle The Public Advertiser [PA] The St. James’s Chronicle, or, British Evening-Post Andrew, Donna T. 1998. “‘To the Charitable and Humane’: Appeals for Assistance in the Eighteenth-Century London Press”. Charity, Philanthropy, and Reform. From the 1690s to 1850 ed. by Hugh Cunningham & Joanna Innes, 87–107. New York & Houndmills, Basingstoke, Hampshire: St. Martin’s & Macmillan.
3.I would like to thank my Work-Study research assistants Kate Stevenson and Rachel Caballero, who read through a number of the newspapers cited in this study, and Kate Stevenson for checking references and for her characteristically thorough and resourceful bibliographical work on servants’ education and women’s teaching. Thanks also to the staff of the Microtext Collection at the University of Toronto Library, and to Jeri McIntosh and Paul Cobb for their hospitality in the UK. For travel and research support I gratefully acknowledge the Association of Commonwealth Universities, International Collaboration Grant CADF-2001-20. I am grateful for the insights and interest of Donna Andrew, Joan Beal, Juliane Brown, Michèle Cohen, John Percy, and an anonymous reviewer. All remaining errors and infelicities are mine.
Consumers of correctness
Anon. 1984 [Trusler 1790]. “Sources for Newspaper & Periodical History [Trusler’s London Adviser and Guide]”. Journal of Newspaper and Periodical History 1:1.37–38. Barker, Hannah. 1998. Newspapers, Politics, and Public Opinion in Late Eighteenth-Century England. Oxford: Clarendon Press. ———. 2000. Newspapers, Politics and English Society, 1695–1855. Harlow: Longman. Barker, Mrs Anne. c1770. The Complete Servant Maid: Or Young Woman’s Best Companion. Containing Full, Plain, and Easy Directions for Qualifying Them for Service in General, But More Especially for the Places of Lady’s Woman, Housekeeper, Chambermaid, Nursery Maid, Housemaid, &c. London: J. Cooke. Benzie, W. 1972. The Dublin Orator. Thomas Sheridan’s Influence on Eighteenth-Century Rhetoric and Belles Lettres. Leeds: University of Leeds School of English. Best, John Hardin. 1968. “The Academy Keeper: A Dimension in the History of English Education”. History of Education Quarterly 8:3.386–398. Cohen, Michèle. 2003. “French Conversation or ‘Glittering Gibberish’?: Learning French in Eighteenth-Century England”. Didactic Literature in England, 1500–1800: Expertise Constructed ed. by Natasha Glaisyer & Sara Pennell, 99–117. Aldershot: Ashgate. Crowley, Tony. 1996. Language in History. Theories and Texts. London & New York: Routledge. Eaves, T. C. Duncan & Ben D. Kimpel. 1971. Samuel Richardson: a Biography. Oxford: Clarendon Press. Fitzmaurice, Susan. 1998. “The Commerce of Language in the Pursuit of Politeness in Eighteenth-Century England”. English Studies 79:4.300–328. Foote, Samuel. 1788. The Dramatic Works of Samuel Foote, Esq; to which is Prefixed a Life of the Author. In Four Volumes. Vol. II. Containing The Orators, The Minor, The Lyar, and The Patron. 1762. London: J. F. and C. Rivington et al. Haig, Robert L. 1960. The Gazetteer 1735–1797. A Study in the Eighteenth-Century English Newspaper. Carbondale: Southern Illinois University Press. Harris, Michael. 1987. London Newspapers in the Age of Walpole. A Study of the Origins of the Modern English Press. Rutherford, New Jersey: Fairleigh Dickinson University Press; & London: Associated University Presses. Hecht, J. Jean. 1956. The Domestic Servant Class in Eighteenth-Century England. London: Routledge & Kegan Paul. Hill, Bridget. 2001. Women Alone. Spinsters in England 1660–1850. New Haven & London: Yale University Press. Holmes, Geoffrey & Daniel Szechi. 1993. The Age of Oligarchy. Pre-Industrial Britain 1722–1783. London & New York: Longman. Langford, Paul. 1989. A Polite and Commercial People. England 1727–1783. Oxford: Oxford University Press. McKendrick, Neil. 1982. “The Consumer Revolution of Eighteenth-Century England”. McKendrick, Brewer & Plumb, 9–33. ———, John Brewer & J. H. Plumb. 1982. The Birth of a Consumer Society. The Commercialization of Eighteenth-Century England. Bloomington: Indiana University Press. Michael, Ian. 1987. The Teaching of English. From the Sixteenth Century to 1870. Cambridge: Cambridge University Press. Millburn. John R. 1988. Wheelwright of the Heavens: The Life and Work of James Ferguson. London: Vade-Mecum Press.
175
176 Carol Percy
Morison, Stanley. 1932. The English Newspaper. Some Account of the Physical Development of Journals Printed in London Between 1622 & the Present Day. Cambridge: Cambridge University Press. Plumb, J. H. 1982. “The New World of Children”. McKendrick, Brewer & Plumb, 286–315. Raven, James. 1992. Judging New Wealth: Popular Publishing and Responses to Commerce in England, 1750–1800. Oxford: Clarendon Press. Skedd, Susan. 1997. “Women Teachers and the Expansion of Girls’ Schooling in England, c.1760–1820”. Gender in Eighteenth-Century England: Roles, Representations and Responsibilities ed. by Hannah Barker & Elaine Chalus, 101–125. London: Longman. Stirling, John. 1735. A Short View of English Grammar. London: for the author. Sundby, Bertil, Anne Kari Bjorge & Kari E. Haugland. 1991. Dictionary of English Normative Grammar 1700–1800. Amsterdam & Philadelphia: John Benjamins. Tieken-Boon van Ostade, Ingrid. 2000. “Robert Dodsley and the Genesis of Lowth’s Short Introduction to English Grammar”. Historiographia Linguistica 27:1.21–36. Tompson, Richard S. 1977. “English And English Education In The Eighteenth Century”. Facets Of Education In The Eighteenth Century ed. by James A. Leith, 65–83. (= Studies On Voltaire And The Eighteenth Century 167). Oxford: Voltaire Foundation At The Taylorian Institution. Watts, Richard J. 1999. “The Social Construction of Standard English: Grammar Writers As a ‘Discourse Community’”. Standard English. The Widening Debate ed. by Tony Bex & Richard J. Watts, 40–68. London & New York: Routledge. ———. 2002. “From Polite Language to Educated Language. The Re-Emergence of an Ideology”. Alternative Histories of English ed. by Richard J. Watts & Peter Trudgill, 155–172. London & New York: Routledge.
Accounting for vernacular features in a Scottish dialect Relic, innovation, analogy and drift Jennifer Smith University of York
1.
Introduction
The origins and subsequent development of vernacular features in dialects of English, as in (1)–(3), have been at the forefront of diachronic and synchronic research over the past few decades (e.g. Chambers 2001; Christian et al. 1988; Ellegård 1953; Montgomery 1994; Nevalainen 2002; Smith 2000; Smith 2001; Smith & Tagliamonte 1998; Tagliamonte & Smith 2002; Trudgill & Chambers 1991). (1) a. There was one nicht we were lyin’ at anchor. (b:875.32)1 b. We played on ’at beach ’til we was tired, sailin’ boaties, bilin’ welks… (b:254.15) (2) a.
I went oot for a walk and saw thon little birdies dashin’ aboot the bushes. (l:600.9) b. You seen bits floatin’ in it. (l:528.39)
(3) a. She’s in the huff if I do na let her. (g:659.13) b. God, I Ø na ken far my ain face is here. (a:654.18)
A number of hypotheses are advanced to explain the use of such vernacular forms including relic, innovation, analogy and drift (e.g. Andersen 1988; Antilla 1989; Chambers 1995; Hock 1986; Malkiel 1981; Manczak 1980). How can these different hypotheses be disentangled? The mere presence of a form in synchronic data tells us little about its evolution across time and space. For
1.The letters and numbers in parentheses represent speaker code and place in transcription.
178 Jennifer Smith
example, is it used by all community members; is it in the process of obsolescing; is it new or does it have historical precursors; is it undergoing change and if so, in which direction; what are the constraints on its use; do other dialects exhibit the same feature? In this paper, I conduct a large-scale quantitative analysis of the use of the variable vernacular features in (1–3) in a community of speakers from the north east of Scotland. Through distributional patterns of use across linguistic structure and three generations of speakers, I aim to shed light on the different mechanisms at work which lead to the use of these vernacular forms. I first turn to the community under investigation.
2. The community Buckie is a small fishing town situated on the coast 60 miles north of Aberdeen, shown in Figure 1.
Buckie
Figure 1.Map showing location of Buckie.
It is quite isolated on both geographic and economic grounds and therefore remains relatively immune to more mainstream developments. This impacts on the dialect, as “the relative isolation and geographic semi-independence from the rest of Scots has assured a strong linguistic identity” (Johnston, 1997: 445). Indeed, Buckie is a classic example of a traditional dialect which is “found only in long settled and especially remote and peripheral rural areas” (Trudgill and Chambers 1991:3). Communities such as these are “of particular interest because
Accounting for vernacular features in a Scottish dialect 179
they do diverge most markedly at the grammatical level from the already relatively well-known standard and other mainstream varieties of English” (ibid.: 3). Crucially, Buckie provides a rich data source for research into the origins of dialect features relatively uninfluenced by prescriptive dictates. The data were collected using standard sociolinguistic methodology (Labov 1984). I was able to obtain a representative sample of vernacular Buckie due to my in-group status, as I am from this community. The data amount to approximately forty hours of tape-recorded spontaneous speech. The corpus has been fully transcribed and consists of over 300,000 words. The speakers in the sample had been born and raised in the community, were working class and exhibited networks that were generally confined to the community in question. Three age ranges were selected (22–31, 50–60 and 80+) to represent three different generations in the community. The speaker sample is shown in Table 1 (see detail in Smith, 2000). Table 1.Speaker sample. age
male
female
22–31 50–60 80+
8 7 4
8 7 5
3. Vernacular features in Buckie I now turn to an examination of the three linguistic variables in Buckie: was/were alternation, irregular verbs, and do absence in negative declaratives. 3.1 Use of was in standard were Use of was in contexts of standard were is widespread in vernacular varieties of English (e.g. Cheshire 1982; Christian et al. 1988; Tagliamonte 1998), therefore it is not surprising to see it used in Buckie, as in (4): (4) a.
We were a’ thegither. I think we was a’ thegither. (f:231.8)
A number of hypotheses have been put forward to explain this use: Chambers (1995: 242) maintains that was in were or “default singulars” is a “primitive of
180 Jennifer Smith
vernacular dialects in the sense that [it] recurs all over the world” (ibid:247). Alternatively, the verb be is said to be undergoing a process of analogical levelling (Christian 1988; Fries 1940; Wolfram 1969), with the distinction between preterite singular-plural disappearing in line with other verbs where this change has gone to completion (Fries 1940; Wolfram 1976; Feagin 1979). Whether use of was in were is the product of analogical levelling or a primitive, the outcome would be the same: was appearing in all contexts of use. I now turn to an analysis of its use in Buckie, starting with the overall distribution of use (Table 2). Table 2.Overall distribution of was in were.
N %
was
were
Total
628 46
723 54
1351
The high percentage of use of the vernacular form (46%) may indeed suggest that use of was is a primitive, or subject to analogical levelling. However, such overall distributions can only give a very broad view of variability and tell us very little about how use of was in were is distributed across linguistic structure, or indeed across the three generations of speakers. Thus, the data were divided by subject type in order to uncover possible internal conditioning: second person singular you as in (5); first person plural we, as in (6); third person plural pronoun they, as in (7); full plural NP, as in (8); plural existentials, as in (9); and by age.2 (5) a. He says ‘I thocht you were a diver or somethin’.(7:262.41) b. ‘Aye, I thocht ye was a scuba diver’. (7:259.21) (6) a. There was one nicht we were lyin’ at anchor. (b:875.32) b. We played on ‘at beach ‘til we was tired, sailin’ boaties, bilin’ welks… (b:254.15) (7) a. They were aie sort o’ pickin’ on me, like. (j:504.3) b. They were still like partyin’ hard. (j:635.28) (8) a. Buckie boats were a’ bonny graint. (g:1066.0) b. The mothers was roaring at ye comin’ in. (g:256.34)
2.Second person plural you was excluded in the following analysis because of the small number of contexts of use (n = 10).
Accounting for vernacular features in a Scottish dialect
(9) a. There were a puckle thatched hooses like that. (c:335.17) b. Oh, there was a lot of coopers ’at time. (c:13.45)
Figure 2 presents the results.
Figure 2. Distribution of was in were by grammatical person and age.
Figure 2 reveals a remarkable finding on the use of was in contexts of standard were: high rates of use of was in all contexts, except for third person plural pronoun they, which is categorically standard. Moreover, the figure demonstrates that these patterns of use are not idiosyncratic but represent community norms for all age groups, as they are replicated across the three generations of speakers.3 While a theory of analogical levelling or primitive status may be called upon to explain was in were in some dialects, the specific patterns in Figure 2 argue against such an explanation for Buckie. Instead, the historical record may provide the key to explaining these particular constraints across linguistic structure. A number of external and internal constraints affecting the distribution of was and were are attested as far back as the Middle English period (e.g. Jespersen 1954; Pyles & Algeo 1993; Visser 1970. An important geographical difference documented is the type of subject and adjacency of the subject and verb constraint, which operated in northern regions of the country from the thirteenth century:
3.In fact, the only pattern that distinguishes itself from the rest is the young speakers’ relatively high rates of was in first person plural we.
181
182 Jennifer Smith
When the subject is a noun, adjective, interrogative or relative pronoun, or when the verb and subject are separated by a clause, the verb takes the termination -s in all persons. (Murray 1873: 211)
Thus, sentences such as They cut them contrast with The men cuts them. Moreover, expressions “such as the men syts are not vulgar corruptions, but strictly grammatical in the Northern dialect” (Murray 1873: 212). Although this rule applies to the present morpheme -s, “in the same way was, wes intruded upon were, war in the past tense” (Murray 1873: 213). This constraint on third person plural contexts is clearly instantiated in these synchronic data from Buckie, with high rates of use of was in full NPs vs. categorical use of were with pronominal they. This pattern can be seen most clearly with the older speakers who have 81% use of was with plural NPs in contrast to 0% with they.4 A second geographical constraint arises from the historical record on use of was in were: in Northern dialects, second person singular appeared with was (Brunner 1963; Forsström 1948; Mossé 1952) as in (10): (10) a.
Lete punysch for the, when that thou was thrall. (c1350: Mirror of Lewed Men, 608) b. He wreitt to me that ye ves in Edinburgh for sick occasiounes. (1609: Letters of Duntreath)
Again, we see this pattern reflected in the Buckie data: high rates of was with second person singular you. In fact, with the exception of existentials, these have the highest frequencies of use in the middle-aged and older speakers.5 The opposing pressures at work in the use of was in were are summed up by Ferguson (1996: 191), who states that [the] spread of the -s to … 2nd persons and the plural [is] evident in many [non-standard] varieties of English around the world. In some instances they apparently continue local features that have never become standardised, and in other instances they are new emergences of ‘natural’ tendencies that harmonise with the drift of morphological simplification. (italics added)
4.With the middle-aged and younger speakers, the categorical vs. variable split is still evident, but with a gradual erosion of was in NPs, presumably due to influence from the standard. 5.The high rates of use of was with plural existentials is said to be due to its unique syntactic structure: there in subject position.
Accounting for vernacular features in a Scottish dialect 183
The specific distributional patterns in the Buckie dialect demonstrate that this vernacular feature exhibits a remarkable continuation of local features or retention of forms passed on from generation to generation. I now turn to the second vernacular feature, use of irregular verbs. 3.2 Irregular verbs Jespersen (1954: 23) states that “on no other point, perhaps, has the Old English grammatical system been revolutionised to the same extent as in the formation of tenses of the verbs”. This “revolution” continues today, as many nonstandard varieties, including Buckie, continue to exhibit a great deal of variation in the past tense verbal paradigm, as in (11)–(19): (11) a.
I taen three of them and the other lad took the rest of them. (n:210.40)
(12) a. They’ve just broke up. (t:228.52) b. They’ve broken up, so I think…(t:242:0) (13) a. So that’s what she did. (u:785.39) b. I canna mine if she done anythin’ aifter the factory. (u:704.3) (14) a.
I went oot for a walk and saw thon little birdies dashin’ aboot the bushes. (l:600.9) b. You seen bits floatin’ in it. (l:528.39)
(15) a. We gied across atween Christmas and New Year. (t:312.23) b. We picked her up in Glasgow and went across and bade there. (t:297.39) (16) a. And eh, she’d put hame a lovely velvet frock. (r:813.25) b. And we was putten into the black hole. (r:846.22) (17) a. You could’ve had a lovely museum. (b:1008.37) b. And it couldna been better laid even though you’d haen a right ain. (b:573.53) (18) a.
He would raither’ve got settled in and gotten a job or somethin’. (x:48.37)
(19) a. So Doctor Paterson told her it was multiple sclerosis. (a:206.43) b. Doctor Paterson telt him right up, right out. (a:196.29)
Table 3 shows that of a total of 5403 contexts of use of irregular verbs, nonstandard forms are used 21% of the time in the Buckie data. Further analysis of the data reveals that frequently occurring verbs account
184 Jennifer Smith
Table 3.Overall distribution of forms in strong verb paradigm. standard
non-standard
N
%
N
%
4285
79
1118
21
for 89% of the data. From these, a total of six verbs (say, think, make, buy, meet, give) were categorically standard. These, along with less frequently occurring verbs, were excluded from the following analysis. From the remaining variable verbs, examination reveals specific variable patterns: use of past participle forms in preterite contexts with the verbs seen, done and taen (taken), as in (20); -en retention with past participles, as in (21); standard irregular forms which are regularised in Scottish dialects, as in (22); use of come in preterite contexts, as in (23); use of preterites in past participle contexts, as in (24). (20) a. And I seen his death in the paper. (w:465.41) b. Aye, well, I done it in Aberdeen, like. (l:388.20) c. Well, it taen a long time. He aie says it taen him a long time. (v:565.16) (21) a. We’ve haen your dad across. ($:16.0) b. So, they were putten bane the house. (a:677.38) c. She’s gotten a mixer but she winna use it. (1:294.16) (22) a. It was your granny that telt me on Sunday. (a:27.12) b. The more you selt, the bigger the commission. (c:652.14) c. And hees bones gied rattlin’ up against the wa’. (g:1015.16) (23) a.
But he come to me aie day and he says ‘Was ever ee in Uigg, John?’. (c:293.7)
(24) a. Since they’ve went to a trawler they’ve haen problems. (t:33.15) b. And I mean, I had drove home fae Elgin heaps of times. (j:424.32)
The verbs were grouped according to these patterns of use and stratified by age. Figure 3 shows the results. In sharp contrast to was/were, which exhibited continuity of variable patterns through three generations, the irregular verb system shows severe disruption from one generation to the next. Moreover, the changes appear to be rather disparate. There are preterites as past participles (forgot, bade etc.), past participles as preterites (seen/done/taen). There is loss of regularised forms (gied) and stability of regularised forms (selt/telt). There is movement towards
Accounting for vernacular features in a Scottish dialect
Figure 3.Distribution of non-standard irregular verb forms by age.
the standard (e.g. came, went) and movement away from the standard (seen, forgot, etc.). There is innovation in apparent time (preterites in past participles, past participles in preterites) and retention of older forms (gotten, haen). How can these seemingly disparate processes be reconciled and what can explain the rapid change over time? The different systems in places when learning past temporal verb forms may provide a clue. Regular verbs are created anew each time by rule, but irregular forms are learned by rote, stored in the lexicon and accessed when needed (e.g. Bybee 1985; Pinker & Prince 1988; Kiparsky 1982). In other words, the irregular forms are “roughly where the grammar leaves off and memory begins” (Pinker & Prince 1988: 126). I suggest that this learning system has an impact on the forms used: the younger speakers are in the process of restructuring the existing paradigm in order that it consists of the same preterite and past participle forms, thus decreasing the cognitive burden on memory. The resulting paradigm is shown on Table 4. Here we have a case of change by analogy where an irregular form is made to conform to a regular pattern (Hock 1986; Kurylowicz 1965; Manzak 1958): the same preterite and past participle forms, as is the case with all regular verbs.
185
186 Jennifer Smith
Table 4.Strong verb paradigm for younger speakers. stem
preterite
past participle
see do take come go get have sell tell forget/fall etc.
seen done taen came went got had selt telt forgot/fell
seen done taen came went got had selt telt forgot/fell
Hence, the use of seen/done/taen in preterite contexts and the use of preterites such as forgot in past participles and came in both preterite and past participle contexts. The verbs sell and tell continue to be regularised as they already have identical forms for both preterite and past participle, therefore fit this schema. Obviously, there is some sociolinguistic patterning to be accounted for, such as the rapid rejection of gied in favour of went in the younger speakers, but the overriding motivation is simplification through analogical change. But Buckie is not unique in this change as there is “a fairly strong tendency to reduce the number of form distinctions for a given irregular verb to two” (Christian et al. 1988: 108) in many dialects of English worldwide (cf. Christian 1988; Eisikovits 1991; Feagin 1979). I suggest that this ‘communal’ change across a diverse range of dialects may be the result of drift, which “is constituted by the unconscious selection on the part of the speakers of those individual variations that are cumulative in some special direction” (Sapir 1921: 155). Thus, analogical change, which results in drift towards one form: two functions in the past tense verbal paradigm may provide an explanation of the use of irregular verbs in Buckie and indeed in other vernacular dialects also. I now turn to the last vernacular feature: do absence in negative declaratives. 3.3 Do absence One of the many changes that has taken place in the English language over the past few centuries is in the formation of negatives. In Middle English and Early Modern English, the negative particle appeared in post-verbal position, as in (25):
Accounting for vernacular features in a Scottish dialect 187
(25) a. It perteyneth not to hym of the scheep. (c1380: Wyclif John, 10, 13) b. But looked not on the poison of their hearts. (c1594: Shakespeare, Richard III: III, i, 12)
But by the end of the fourteenth century, negation formed with periphrastic do and the negative particle before the verb, as in (26), came into use (e.g. Denison 1993: 265, Ellegård 1953), and is the form used in present day English. (26) a.
I have grete mervayelle … that they do not attaine an accion ayenst Sir Thomas. (c1450: Fastolf, Paston Letters No 162, line 198) b. Ye do not speke as ye thynke. (1548: John Bale, Kynge Johan 317)
In Buckie, however, do in present tense negative declaratives is variably realised, as in (27–29) (27) a. I dinna mine fa taen it. (a:391.34) b. I Ø na mine fa come in. (a:972.36) (28) a. You dinna ken fit tae dee wi’ quines. (t:881.39) b. You Ø na ken a-thing aboot me. (u:65.27) (29) a. They dinna ken they’re gan to wear a kilt. (i:31.3) b. They Ø na seem to bide in the Beacons lang. (3:208.13)
While present tense negative declaratives are variable, past tense negatives are always marked with did.6 From a total of 756 contexts of present tense negative declaratives, Table 5 shows that the overall distribution of do absence, the non-standard feature, is 40%. Table 5.Overall distribution of do.
N %
with do
without do
total
451 60
305 40
756
We saw that distribution across grammatical person was vital in explaining use of was in were. Initial observations and my own grammaticality judgements suggested a person and number constraint with do absence also. Therefore, a distinction was made between first singular, as in (30), second singular, as in
6.In cases where do is absent, stress falls on the pronoun, and the negative particle is unstressed, i.e. appears as the cliticised form na.
188 Jennifer Smith
(31), first plural, as in (32). Third person singular and plural pronouns, as in (33) and (34), were categorised separately from full NPs, as in (35) and (36): (30) God, I Ø na ken far my ain face is here. (a:654.18) (31) Ye Ø na hear o’ him onywye, ken. (u:54.86) (32) We Ø na hae raffles. (%:32.30) (33) It doesna cost nothin’ to walk ower the hill. (1:604.21) (34) They Ø na lose trade. ($:44.32) (35) No, Willy doesna play much golf. (3:455.56) (36) My crowd divna like barley. (q:27.45)
Figure 4 shows the distribution of do absence by grammatical person and age.
Figure 4.Distribution of do absence by grammatical person and age.
Note the dramatic split in absence/presence of do by grammatical person across all ages. Third person singular pronouns, and both singular and plural full NPs, show categorical use of do. In other words, they are categorically standard. The remaining contexts, on the other hand, are variable. This categorical versus variable use of do demonstrates that grammatical person is an important effect on the use of do. Moreover, with the exception of slightly lower rates of use across first person singular in the middle-aged and younger speakers, the patterns of use are consistent across the three generations, indicating minimal change across time. How can these distributional patterns be explained?
Accounting for vernacular features in a Scottish dialect 189
It has been suggested that do absence is the result of “absorption of the auxiliary” (Scottish National Dictionary (SND) s.v. no), as examples such as (37) are found in Scots dialects from 100 years ago. (37) a.
I no mind o’ over hearin her saying onythin o’ the sort. (1894: SND s.v. no) b. I no want to see the man that put ma Wullie in prison. (1906: SND s.v. no) c. I no want onything, I said. (1924:SND s.v. no)
In other words, do absence is the product of phonological deletion. However, the patterning of the Buckie data suggest otherwise: the variation is highly constrained to particular grammatical persons despite the fact that all phonological environments are the same in each pronominal person and number of the subject in which do occurs — preceding vowel and following nasal. Similarly, negative declaratives in the past tense have the same phonological environment, but are not variable. The highly specific distributional facts that this analysis reveals make it unlikely that phonological deletion provides the answer to do absence. In the case of was/were, the differential patterning across grammatical person was explained in terms of the historical record. Could the same be true for do absence? In its surface syntactic form, the construction employed in this dialect is identical to the Old English pre-verbal negator ne/na, a use which continued in Older Scots, as in the examples from the Dictionary of the Older Scottish Tongue (DOST) in (38): (38) a. That thai na will…flei. (DOST s.v. na adv. 2) b. That we na gang forth. (DOST s.v. na adv. 2) c. He na dyd it bot in saufte of the schyp. (DOST s.v. na adv.2)
However, the patterns of do absence in Buckie differ significantly from the Old English/Older Scots preverbal negator ne/na. In (38c), for example, the construction is third person singular and past tense. These are the two contexts which do not allow do absence in the Buckie data; therefore, the distributional facts are not compatible with a ‘relic’ argument. Instead of appealing to phonological processes or relic forms, I suggest that the categorical versus variable absence of do in Buckie looks much more like a syntactic process. Recall that in the Middle English and Early Modern English period, negatives were formed by placing the negative particle not immediately after the verb, as in (25). In these constructions, the verb carries tense and
190 Jennifer Smith
agreement information: third person singular (perteyn-eth) and past tense marking (look-ed). In Modern English, however, as in (26), negatives are formed with preverbal negation but with no tense and agreement marking on the verb (e.g. attaine, speke). Instead these appear on do (e.g. does, did). In fact, according to current syntactic frameworks, do is inserted as a support for these inflections (e.g. Kroch 1989). Note that in Buckie, do is categorical only in contexts which carry overt tense and agreement marking — the -s suffix in present indicative and past tense contexts with did.7 In the remaining contexts, do is ‘redundant’, as it is not strictly needed to support third person singular or past tense marking.8 This explains its variable appearance in these contexts. A syntactic account explains the categorical vs. variable distinction, but we are still left with the question of the evolution of this feature. Indeed, one of the most intriguing aspects of the variable use of do in Buckie is that it only occurs in Buckie and proximate rural dialects.9 Andersen (1988: 39) states that “the observation that central and peripheral parts of a speech area typically develop differently is one of the most durable insights in historical dialectology”. In peripheral speech areas, “the same linguistic system is transmitted from generation to generation in communities which by reason of their location in space are more or less closed and which, as a consequence, present different conditions for the maintenance and elaboration of complex norms” (ibid.: 78). I suggest that do absence is an example of this: elaboration of an existing system which results in an independent innovation (Andersen 1988) with patterns of use passed on from generation to generation, largely intact.
4. Conclusion I have now analysed in detail three vernacular features in the Buckie dialect. Quantitative analysis of the distributional patterns of these forms revealed that
7.Note that third person plural NP is also categorically standard. This may be the result of small Ns (n = 12). Alternatively, and perhaps more convincingly, it may be due to the NP/pro rule discussed in Section 3.1, where plural NPs also take the -s morpheme. 8.See Smith (2000) for further analysis of the variable contexts of use. 9.There are instances where do is absent in imperative constructions in some Scottish dialects, as in Away and no haiver, but these appear to be formulaic and limited to a few utterances.
Accounting for vernacular features in a Scottish dialect
they appear to be the result of entirely different processes: retention of relic forms, analogy and drift, and independent innovation. The analysis of was in contexts of standard were was strongly conditioned by grammatical person. I claim that these differential rates across subject type could not be due to primitive status or analogical levelling. Instead, the specific patterns of use are explicable against the backdrop of the historical record, as these constraints are attested for Northern dialects as far back as the thirteenth century. In other words, the patterns in this contemporary dialect reflect those attested in diachrony and are the product of retention of forms over centuries. “Conjugation regularisation” (Chambers 1995:242), which is “the tendency towards identical past tense and past participle forms” (Trudgill and Chambers 1991: 216), characterised the use of irregular verbs in Buckie. I suggest that this is due to change by analogy, where irregular verbs are conforming to a more regular pattern of use. Moreover, the restructured paradigm evidenced in Buckie is a widespread phenomenon in other dialects: this may be the result of drift, where languages of common origin follow certain pathways of change (Sapir 1921). The last variable examined was the use of do in negative declaratives in the present tense. Do presence is categorical when there is overt third person -s or past tense morphology. In all other contexts, it is variable. These distributional patterns can be found neither in the diachronic record nor other contemporary dialects. This leads to the conclusion that the Buckie dialect had further elaborated on the existing system of do support, resulting in an independent innovation. In sum, this research has demonstrated that the study of three non-standard features in one small community can shed light on the origins and subsequent development of vernacular norms. Moreover, analysis across linguistic structure and apparent time may provide a window into the mechanisms underlying language variation and change in language more generally.
References Andersen, H. 1988. “Center and periphery: Adoption, diffusion, and spread”. Historical Dialectology: Regional and Social ed. by J. Fisiak, 39–83. Berlin: Mouton de Gruyter. Anttila, R. 1989. An introduction to historical and comparative linguistics. Amsterdam & New York: John Benjamins. Brunner, K. 1963. An outline of Middle English grammar. Oxford: Blackwell.
191
192 Jennifer Smith
Bybee, J. L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam & Philadelphia: John Benjamins. Chambers, J. K. 1995. Sociolinguistic theory: Linguistic variation and its social significance. Oxford: Blackwell. ———. 2001. “Vernacular universals”. ICLaVE 1: Proceedings of the First International Conference on Language Variation in Europe ed. by Joseph M. Fontana, Louise McNally, M. Teresa Turell & Enric Vallduvi, 52–60. Barcelona: Universitat Pompeu Fabra. Cheshire, J. 1982. Variation in an English dialect. A sociolinguistic study. Cambridge: Cambridge University Press. Christian, D., W. Wolfram & N. Dube, 1988. Variation and change in geographically isolated communities: Appalachian English and Ozark English. Tuscaloosa, Alabama: American Dialect Society. Denison, D. 1993. English Historical Syntax. Harlow, Essex: Longman. Eisikovits, E. 1987. “Variation in the lexical verb in Inner-Sydney English”. Australian Journal of Linguistics 7.1–24 Ellegård, A. 1953. The auxiliary do: the establishment and regulation of its use in English. Stockholm: Almqvist & Wiksell. Feagin, C. 1979. Variation and change in Alabama English: A sociolinguistic study of the White community. Washington, D. C.: Georgetown University Press. Ferguson, C. A. 1996. “Variation and drift: Loss of agreement in Germanic”. Towards a social science of language: Papers in honor of William Labov ed. by G. R. Guy, C. Feagin, D. Schiffren & J. Baugh, 173–99. Amsterdam & Philadelphia: John Benjamins. Forsström, G. 1948. The verb ‘to be’ in Middle English: A survey of the forms. Lund: C. W. K. Gleerup. Fries, C. C. 1940. American English grammar. New York: Appleton, Century, Crofts. Hock, H. H. 1986. Principles of historical linguistics. Amsterdam: Mouton de Gruyter. Jespersen, O. H. 1954. A modern English grammar on historical principles. Part VI: Morphology. London: George Allen & Unwin. Johnston, P. 1997. “Regional variation”. The Edinburgh history of the Scots language ed. by C. Jones, 433–513. Edinburgh: Edinburgh University Press. Kiparsky, P. 1982. “Lexical phonology and morphology”. Linguistics in the Morning Calm ed. by I. S. Yang, 3–91. Seoul: Hanshin. Kroch, A. 1989. “Function and grammar in the history of English: periphrastic do”. Language change and variation ed. by R. Fasold and D. Schiffren, 133–172. Washington, D. C.: Georgetown University Press. Kurylowicz, J. 1965. The evolution of grammatical categories. Diogenes. 51: 55–71 Labov, W. 1984. “Field methods of the project on linguistic change and variation”. Language in Use: Readings in Sociolinguistics ed. by J. Baugh & J. Sherzer, 28–54. Englewood cliffs, NJ: Prentice Hall. Malkiel, Y. 1981. “Drift, slope and slant: background of, and variations upon, a Sapirian theme”. Language 57:535–70. Manczak, W. 1980. “Laws of analogy”. Historical morphology ed. by J. Fisiak, 283–288. The Hague: Mouton de Gruyter.
Accounting for vernacular features in a Scottish dialect 193
Montgomery, M. B. 1994. “The evolution of verbal concord in Scots”. Proceedings of the Third International Conference on the Languages of Scotland ed. by A. Fenton & D. A. MacDonald 81–95. Edinburgh: Canongate Academic Press. Mossé, F. 1952. A handbook of Middle English. Baltimore: Johns Hopkins University Press. Murray, J. A. H. 1873. The dialect of the Southern Counties of Scotland: Its pronunciation, grammar and historical relations. London: Philological Society. Nevalainen, T. 2002. “For there are many that is faln from the faith: the use of is/was in the plural in Late Middle and Early Modern English”. Paper presented at ICEHL 12, Glasgow, August 2002. Pinker, S. & A. Prince. 1988. “On language and connectionism: analysis of a parallel distributed processing model of language acquisition”. Cognition 28.73–193. Pyles, T. & J. Algeo. 1993. The origins and development of the English language. Orlando: Harcourt Brace & Company. Sapir, E. 1921. Language. New York: Harcourt Brace. Smith, J. 2000. Synchrony and diachrony in the evolution of English: Evidence from Scotland. Ph.D. dissertation, University of York, England. ———. 2001. “Ye Ø na hear that kind o’ things: Negative do in Buckie Scots”. English World Wide, 21.2 231–259. ——— & S. Tagliamonte. 1998. “‘We was all thegither, I think we were all thegither’: Was regularisation in Buckie English”. World Englishes. 17:2.105–126. Tagliamonte, S. 1998. “Was/were variation across the generations: View from the city of York”. Language Variation and Change. 10:2.153–191. ——— & J. Smith. 2000. “Old was; new ecology: Viewing English through the sociolinguistic filter”. The English history of African American English ed. by S. Poplack, 141–174. Oxford: Blackwell. Trudgill, P. & J. Chambers, eds. 1991. Dialects of English: Studies in grammatical variation. London & New York: Longman. Visser, F. 1963–73. An historical syntax of the English language. Leiden: E. J. Brill. Wolfram, W. 1969. A sociolinguistic description of Detroit Negro speech. Washington, DC: Center for Applied Linguistics.
On MV/VM order in Beowulf * Hironori Suzuki Fukushima National College of Technology, Japan
1.
Introduction
The principles of Old English prose syntax are on the whole well established, but there is little agreement concerning verse syntax. Regarding the word order of the modal auxiliary (M) and its infinitive complement, a non-finite verb (V), in the subordinate clauses of Old English, Ohkado (2000) argues that the choice between the MV and VM orders depends largely on the presence of an extra element, such as an adverb or a prepositional phrase (i.e. besides S, M, V). Thus, the MV order tends to be prevalent in subordinate clauses with an extra element, as in (1a), while the VM order predominates in those without an extra element, as in (1b). In the quotations, modal auxiliaries are in boldface; nonfinite verbs are underlined; extra elements are double-underlined. (1) a.
Modal-Verb order (with one extra element) swa sceal æghwylc mon alætan lændagas as each one must give up transitory days (Beo 2590b–1a) b. Verb-Modal order (with no extra element) gif he wealdan mot, if he may prevail (Beo 442b)
*An earlier version of this paper was presented at the 12th International Conference on English Historical Linguistics held at the University of Glasgow (August 23, 2002) and originally at the 17th National Conference of the Japan Society for Medieval English Studies held at Keio University (December 8, 2001). I am grateful to Keiko Ikegami, Masa Ikegami, Leena Kahlas-Tarkka, Eichi Kobayashi, Willem Koopman, Martina Noteboom, Hiroshi Ogawa, Michiko Ogura, Masayuki Ohkado, Susan Pintzuk, John Scahill, Masatomo Ukaji, Hideki Watanabe, and Kazuyoshi Yamanouchi for their invaluable comments. I am also grateful to Dylan Jones for his advice on style.
196 Hironori Suzuki
This is a useful theory, but how widely can it be applied to Old English verse? The purpose of the current survey is to examine whether the influence of the extra element on the word order of M and V can also be observed in subordinate clauses of Old English alliterative verse, specifically Beowulf, and what influence certain metrical features have exercised on the word order of the two elements, M and V. In order to explore the possible influences, the following factors will be examined in terms of their relationship to the MV and VM orders: the presence and absence of any extra element(s), the ‘heaviness’ of M and V, and the distribution of the alliteration in relation to M and V. Although previous theories, pointing to the extra element or ‘heaviness’ as dictating the word order, seem valid when applied to Old English prose, this survey indicates that alliteration is in fact the crucial factor in determining word order in Old English verse.
2. Text The Early Old English alliterative poem Beowulf (Klaeber 1950) was chosen as the database for the current survey for several reasons. Firstly, since Beowulf is generally regarded as quintessentially representative of an early stage of Old English (Klaeber 1950), it will be of help in comparing other texts belonging to different stages of the English language. Secondly, the analysis of a poetic text will help to illuminate the differences among the genres in Old English. Finally, the analysis of this poetic text should demonstrate how the metrical features affect the syntactic structure. In this paper, word order in the clauses in Beowulf that contain a single modal auxiliary and a single infinitive complement will be examined.1 Sometimes a single modal auxiliary is accompanied by several subordinate infinitives, and in such cases, the clauses refuse to fall into clear categories of the MV order and the VM order, as illustrated in (2).
1.Following Ohkado and Mitchell, the modal verbs included in this study are agan, cunnan, durran, magan, motan, sculan, þurfan, and willan. See Ohkado (2001: 1) and Mitchell (1985: Sections 990–1), where Mitchell lists these eight verbs and explains, “I call these ‘“modal” auxiliaries’ for want of a better name”.
On MV/VM order in Beowulf 197
(2) Verb-Modal-Verb order oð þæt him æghwylc ymbsittendra ofer hronrade hyran scolde, gomban gyldan; (9–11a) until each one of his neighbours over the whale-road had to obey him and pay tribute
Therefore, such cases have had to be excluded. Subject to this limitation, the total number of clauses in Beowulf containing a single modal auxiliary and a single infinitive complement is 248.
3. Clause types and word order frequencies In some cases there is some difficulty in determining the clause type. For example, in sentence (3) the word ða may be interpreted either as an adverb ‘then’, introducing a main clause, or a subordinate conjunction ‘when’. Mitchell (1985: Section 2445) contends that “both may be right”, and the choice depends ultimately on personal interpretation. The other ambiguous clauses begin with þær, which makes sense either as ‘there’, introducing a main clause, or as ‘where’, entailing subordination. (3)
ða hine Wedera cyn for herebrogan habban ne mihte. (461b–2) When the Weder people for fear of war might not hold him / Then …
The figures in (4) show the clause types and word order frequencies. As has often been observed, the MV order is typically found in main clauses, while both the MV and VM orders are generally found in subordinate clauses. The figures in (4) confirm this tendency. In main clauses, which are almost exclusively declarative, the MV order typically occurs (83 out of 112 instances, or 74%), while in subordinate clauses there is a slightly more even occurrence (although the VM order predominates: 82 out of 123 instances, or 67%). In ond/ac clauses, the MV order outnumbers the VM order by seven instances to three, which is similar to the tendency in main clauses. In ambiguous clauses, all three instances take the VM order.
198 Hironori Suzuki
(4) Clause types and word order frequencies
MV VM
Main Clause
Ambiguous Clause
Subordinate Ond/Ac Clause Clause
83 (74%) 29 (26%)
0 (0%) 3 (100%)
41 (33%) 82 (67%)
7 (70%) 131 (53%) 3 (30%) 117 (47%)
3
123
10
total 112
Total
248
4. Extra elements and the order of modal and non-finite verbs The figures in (5) illustrate the relationship between extra elements and the order of modal and non-finite verbs. In counting the number of extra elements, appositive expressions are counted as one. (5) Number of extra elements and order of Modal and Non-Finite Verbs a. Extra Element = 0 MV VM total 2 (40%) 3 (60%) 5 b. Extra Element = 1 MV VM total 4 (9%) 43 (91%) 47 c. Extra Elements = 2 or more MV VM total 35 (49%) 36 (51%) 71
If the same tendency as Ohkado (2000) has discovered in Old English prose also applies to Old English verse, subordinate clauses with the extra element should tend to take the MV order. However, the figures in (5) show that the majority of all the clauses take the VM order, irrespective of the presence or absence of any extra element(s). In particular, of all the instances with a single extra element, forty-three out of forty-seven, or ninety-one percent, an overwhelming majority, take the VM order, which apparently contradicts the expectation. Some examples of the MV and VM orders without the extra element are given in (6). Examples of MV and VM orders with one extra element are given in (7), where the extra element is an object noun phrase. Illustrated in (8) are cases showing two extra elements containing prepositional phrases and object noun phrases.
On MV/VM order in Beowulf 199
(6) Examples with no extra element a. MV swa sceal man don as a man should do b. VM þæt se byrnwiga bugan sceolde that the mailed warrior had to bow (7) Examples with one extra element a. MV se ðe wyle soð specan he who will speak truth b. VM gyf þu healdan wylt maga rice If you wish to hold your kinsman’s kingdom
(1172b) (2918)
(2864b)
(1852b–3a)
(8) Examples with two extra elements a. MV þeah þe ne meahte on mere drifan hringedstefnan (1130–1a) though he might not drive his ring-prowed ship on the sea b. VM ðonne healgamen Hroþgares scop æfter medobence mænan scolde (1066–7) when Hrothgar’s scop should relate hall-pastime among the mead-benches
5. Weight of non-finite verbs This section, to apply further the analysis of Ohkado (2000), will also examine the influence of the weight of non-finite verbs as a possible factor affecting the word order of the modal and non-finite verbs. Ohkado refers to the similar theories of Kohonen (1987), Suzuki (1994), and Davis (1997) that the ‘heaviness’, or the length of the object, is an important factor in determining the position of objects in Old English prose. Thus, clauses with light or short objects tend to manifest the OV order, while those with heavy or long objects tend to take the VO order, as schematized in (9). (9) light object fi OV order heavy object fi VO order
If a similar theory can also be applied to the position of non-finite verbs in Beowulf, then short or light non-finite verbs should tend to precede modal auxiliaries, while long or heavy ones will tend to follow them, as schematized in (10).
200 Hironori Suzuki
(10) light non-finite verb fi VM order heavy non-finite verb fi MV order
The figures in (11) present the relationship between the number of syllables in, and the position of, non-finite verbs (here the ‘heaviness’ of non-finite verbs is measured simply by the number of syllables). (11) Number of syllables and position of Non-Finite Verbs MV VM total 1 syllable 3 (100%) 0 (0%) 3 2 syllables 24 (26%) 67 (74%) 91 3 syllables 14 (48%) 15 (52%) 29 total 41 (33%) 82 (67%) 123
The first column indicates the number of syllables in the non-finite verbs. The figures do not suggest that there is any significant correlation between the number of syllables and the position of non-finite verbs. In clauses with onesyllable non-finite verbs, where the VM order might be expected, all three instances take the MV order, as illustrated in (12a). Similarly, contrary to expectation, in clauses with heavier two-syllable non-finite verbs, a majority (67 out of 91, or 74%) take the VM order, as illustrated in (12b). Even in the instances where the non-finite verb is heaviest, with three syllables, the anticipated MV order does not occur any more frequently than the VM order. (12) a.
MV
b. VM
swa sceal man don as a man should do (wene ic) þæt he mid gode gyldan wille uncran eaferan that he will repay our sons with good
(1172b)
(1184–5a)
However, this notion of ‘heaviness’ is not the only theory that does not seem to be viable. An alternative analysis, which takes into consideration the ‘heaviness’ of both modal auxiliaries and non-finite verbs, also seems flawed. If a weightordering principle operates, the lighter or shorter of the two elements should precede the other heavier or longer one, as illustrated schematically in (13). (13) light modal auxiliary, heavy non-finite verb fi MV order heavy modal auxiliary, light non-finite-verb fi VM order
The figures in (14) show the relationship between the number of syllables of modal and non-finite verbs and their order.
On MV/VM order in Beowulf 201
(14) Number of syllables of Modal and Non-Finite Verbs, and their order Modal Verb MV VM total 1 1 2 (100%) 0 (0%) 2 1 2 2 (29%) 5 (71%) 7 1 3 5 (56%) 4 (44%) 9 2 1 1 (100%) 0 (0%) 1 2 2 22 (27%) 62 (73%) 84 2 3 9 (45%) 11 (55%) 20 total 41 (33%) 82 (67%) 123
Table (15) is derived from (14). Table (15) shows the relative weight of modal and non-finite verbs. The first line in (15) gives figures concerning the cases where modal verbs are heavier than non-finite verbs. The figures in the second line are concerned with the cases where modal and non-finite verbs are equal in syllable length. The figures in the last line are concerned with the cases where modal verbs are lighter than non-finite verbs. (15) Relative weight of Modal and Non-Finite Verbs and their order MV VM total Modal>Verb 1 (100%) 0 (0%) 1 Modal=Verb 24 (28%) 62 (72%) 86 Modal
For clauses with a heavy modal and light non-finite verb, the VM order might be predicted; however, the only instance shows the MV order. Similarly, the MV order might be anticipated to be prevalent for clauses with a light modal and heavy non-finite verb; however, the VM order occurs more often (20 out of 36, or 56%). Where the modal and non-finite verbs are of equal weight, the occurrence of each order might be expected to be equal, but a majority takes the VM order (62 out of 86, or 72%). These results demonstrate clearly that the relative weight of the modal and the non-finite verb is not an important factor affecting the order of the two elements. Similar approaches analysing ‘heaviness’ were tried by Bliss (1981) and Donoghue (1987), who examined the metrical value of the auxiliary and its position in the clauses in Beowulf. However, their treatment of ‘heaviness’ differs from mine in the following way: auxiliaries are divided into two categories according to whether they are metrically monosyllabic or metrically disyllabic. Metrically disyllabic are those which consist of two syllables of which the first is long, such as cuðe, dorste, wolde and a great many others. Metrically
202 Hironori Suzuki
monosyllabic auxiliaries include not only those which are genuinely monosyllabic, such as mag, sceal and a great many others, but also those which consist of two syllables of which the first is short so that they might be subject to ‘resolution’ such as mæge, mægen, sculon, scyle, wile and (ge)witon. Although Bliss and Donoghue examine all kinds of auxiliaries, the present survey will focus on the modal auxiliary specifically. In order to avoid confusion, following Donoghue (1987), it seems preferable to use the term ‘light auxiliaries’ for Bliss’s ‘monosyllables’ and ‘heavy auxiliaries’ for his ‘disyllables’. The examination of the metrical ‘heaviness’ and the position of the modal auxiliary is presented in (16). (16) Metrical weight and position of Modal Auxiliaries MV VM Light auxiliaries 14 (56%) 11 (44%) Heavy auxiliaries 27 (28%) 71 (72%) total 41 (33%) 82 (67%)
total 25 98 123
The chi-square value computed from this table (χ2 = 7.25; df = 1, p < .01) shows a significant dependence at the 0.01 level between the metrical ‘heaviness’ and the position of the modal auxiliaries in subordinate clauses. There is a slight tendency for ‘light auxiliaries’ to take the MV order (14 out of 25, or 56%, such as example (17a)), and a relatively stronger tendency for ‘heavy auxiliaries’ to take the VM order (71 out of 98, or 72%, such as the example in (17b)). Thus, the metrical ‘heaviness’ of auxiliaries does seem to be a factor in determining the word order, but not other types of ‘heaviness’ at all. (17) a.
MV
b. VM
hwæðer sel mæge æfter wælræse wunde gedygan uncer twega (2530b–2a) which of us two may better bear wounds after one bloody meeting þæt se beodcyning ðafinan sceolde Eafores anne dom (2963–4b) so that the people’s king had to submit to the judgment of Eofor alone
Next, the same metrical ‘heaviness’ theory will be applied regarding the metrical weight of non-finite verbs. The figures in (18) show the metrical weight and the position of non-finite verbs.
On MV/VM order in Beowulf 203
(18) Metrical weight and position of Non-Finite Verbs MV VM total Light verbs 10 (29%) 24 (71%) 34 Heavy verbs 31 (35%) 58 (65%) 89 total 41 (33%) 82 (67%) 123
For both clauses with light non-finite verbs and those with heavy non-finite verbs, there is a high percentage of the VM order (24 out of 34, or 71%; and 58 out of 89, or 65%). These results do not verify the weight-ordering principle hypothesis. Furthermore, the weight-ordering principle hypothesis holds that if nonfinite verbs are metrically heavier than modal verbs, there should be a higher percentage of the MV order, and if non-finite verbs are lighter than modal verbs, then there should be a higher percentage of the VM order. Such is the expectation; however, it is not substantiated by the results. The result of the investigation into this question is shown in (19). (19) Relative metrical weight of Modal and Non-Finite Verbs and their order MV VM total Modal>Verb 3 (11%) 24 (89%) 27 Modal=Verb 31 (40%) 47 (60%) 78 Modal
Significantly, the VM order is more likely to occur in all three possible permutations: one, where the modal is metrically heavier than the non-finite verb (24 out of 27, or 89%, as expected); two, where the modal and non-finite verbs are of equal metrical weight (47 out of 78, or 60%, contrary to expectation); three, where the modal is metrically lighter than the non-finite verb (11 out of 18, or 61%, again contrary to expectation). Clearly, the order of the modal and nonfinite verbs is not affected by their relative metrical ‘heaviness’. Thus, as with the extra element theory, the ‘heaviness’ theory would appear to have only limited applicability to Old English verse.
6. Alliteration and order of modal and non-finite verbs As the theories of ‘heaviness’ and the extra element seem to be confounded when applied to the word order of modal auxiliaries and non-finite verbs in Beowulf, the influence of metrical features on this word order might be examined more
204 Hironori Suzuki
profitably instead. Metricists debate keenly the intricacies of the metrical system, especially regarding the patterns of stressed and unstressed syllables of the verse or the half-line. Sievers (1893) observed five main rhythmical types of verse, and Bliss (1967) enlarges the list to fifteen types. Russom (1987) rejects Sievers’s assignment of subtypes within his five types, instead listing twenty-five verse patterns. However, there is broad consensus as to the function of alliteration. Sievers’s widely-accepted definition (1893) might be summarised as follows: the Old English verse line is composed of two half-lines, which are bound into pairs by alliteration. The first accented syllable of the second half-line in the pair alliterates with one or both of the accented syllables in the first half-line of the pair. In the scansion of Old English poetry alliteration is essential to determining the metrical stress in the verse line since the alliterating syllable is usually assumed to bear the metrical stress. Accordingly, the crucial effect of alliteration on the order of modal and non-finite verbs bears closer analysis. For the purpose of this analysis, a set of symbols will be adopted to represent the modal auxiliary and the non-finite verb with the presence or absence of alliteration. In the following tables an ‘M’ or a ‘V’ serve for the modal auxiliary and its infinitive complement respectively. Alliteration will be indicated by the symbol of a boxed M or V and the line boundary by the slash mark, while, in quotations, modal auxiliaries continue to be in boldface, and non-finite verbs underlined. The following tables exhibit the pattern of alliteration and the order of the modal and non-finite verbs by the boundary type: ‘a’-verses, ‘b’-verses, within the same lines, and in different lines (since the modal and non-finite verbs may appear in different units). The figures in (20) show the relation between the pattern of alliteration and the order of modal and non-finite verbs by the boundary type for subordinate clauses. The first column shows the distribution of alliteration, providing four possible permutations: alliterating M and non-alliterating V; non-alliterating M and alliterating V; both alliterating; and neither alliterating.
On MV/VM order in Beowulf 205
(20) Distribution of alliteration and order of Modal and Non-Finite Verbs (Subordinate Clauses) a. ‘A’-verses MV
VM
total
M , Va M, V M , V M, V
0 (0%) 0 (0%) 0 (0%) 0 (0%)
0 (0%) 9 (100%) 0 (0%) 0 (0%)
0 9 0 0
total
0 (0%)
9 (100%)
9
a
Note: boxed letters in this column indicate just the alliterating element, not their order.
b. ‘B’-verses MV
VM
total
M , V M, V M , V M, V
0 (0%) 0 (0%) 0 (0%) 3 (100%)
0 (0%) 73 (100%) 0 (0%) 0 (0%)
0 73 0 3
total
3 (4%)
73 (96%)
76
MV
VM
total
M , V M, V M , V M, V
2 (100%) 0 (0%) 0 (0%) 0 (0%)
0 (0%) 0 (0%) 0 (0%) 0 (0%)
2 0 0 0
total
2 (100%)
0 (0%)
2
MV
VM
total
M , V M, V M , V M, V
3 (100%) 13 (100%) 3 (100%) 17 (100%)
0 (0%) 0 (0%) 0 (0%) 0 (0%)
3 13 3 7
total
36 (100%)
0 (0%)
36
c.
The same lines
d. Different lines
Overall Total (a–d) 41 (33%)
82 (67%)
123
206 Hironori Suzuki
As shown in (20a), there are nine instances of the modal and non-finite verb construction in the same ‘a’-verse where the non-finite verb carries the alliteration and the modal auxiliary does not alliterate, resulting exclusively in the VM order, as illustrated in (21a). This VM order in half-lines seems to support the claim of Donoghue (1987: 17) that “Verbal-auxiliary half-lines appear so often…that they must be considered formulaic”. The figures in (20b) are concerned with cases where the modal and nonfinite verbs occur in the same ‘b’-verse. Of all seventy-six cases, seventy-three take the typical VM order where the non-finite verb alliterates and the modal auxiliary does not alliterate in the same ‘b’-verse, as illustrated in (21b). On the other hand, there are three instances of the MV order where neither the modal nor non-finite verbs alliterates, as illustrated in (21c). Thus, within the bounds of each half-line, the VM order occurs only when the non-finite verb alliterates rather than the modal auxiliary. However, in instances beyond the scope of each half-line boundary, the VM order never occurs. The figures in (20c) are concerned with cases where the modal and non-finite verbs appear in the same line; one of the two elements appears in the ‘a’-verse and the other in the ‘b’-verse. There are two instances where the modal auxiliary in the ‘a’-verse alliterates and the non-finite verb in the ‘b’-verse does not share the alliteration, resulting in the MV order, as illustrated in (21d). (20d) gives data on cases where the modal and non-finite verbs occur in different verse lines. Although the kind of alliteration alters from line to line, of the possible permutations of the MV and VM orders, only the MV order appears, irrespective of alliteration patterns, as illustrated in (21e–h). Thus, the results of (20c) and (20d) indicate that the MV order occurs exclusively when the modal and non-finite verbs are separated by a half-line boundary. (21) a.
V M
b. V M c.
MV
d. M V
þæt he lytel fær longgestreona brucan moste. (Beorh eallgearo) (2240–1a) that he might a little while enjoy the long-held treasure þæs ðe ic ðe gesundne geseon moste (1998) because I may see you sound (mildum wordum) swa sceal man don (1172b) as a man should do ða ne dorston ær dareðum lacan on hyra mandryhtnes miclan þearfe (2848–9) who had not dared fight with their spears in their liege lord’s great need
On MV/VM order in Beowulf 207
e.
M /V
f.
M/V
g. M / V
h. M/V
þæt hie ne moste (þa Metod nolde) se s[c]ynscaþa under sceadu bregdan (706–7) that (when the Ruler did not wish it) the evil-doer might not drag them down into the shadows (torht getæhte) þæt hie him to mihton gegnum gangan (guðbeorna sum) (313b–4a) so that they might go straight to it þæt he ma moste manna cynnes ðicgean ofer þa niht (Þryðswyð beheold) (735–6a) that he should feed on more of mankind beyond that night gyf him edwendan æfre scolde bealuwa bisigu bot eft cuman (280–1) if change, relief from evil’s distress, ever again come to him
The figures in (22) indicate the relationship between the pattern of alliteration and the order of modal and non-finite verbs for ambiguous clauses. (22) Pattern of alliteration and order of Modal and Non-Finite Verbs (ambiguous clauses) a. ‘A’-verses
M, V
MV
VM
total
0 (0%)
3 (100%)
3
There are three instances in ‘b’-verses, all showing the typical VM order where the modal does not alliterate and the non-finite verb alliterates within the same half-line boundary, as illustrated in (23). Thus, ambiguous clauses show the same tendency as subordinate clauses. (23) V M
(balwon bendum) ðær abidan sceal maga mane fah miclan domes (977b–8) …deadly bonds. There, like a man outlawed for guilt, he shall await the great judgment / …deadly bonds where…
Example (24) gives data on the relationship between the pattern of alliteration and the order of modal and non-finite verbs for main clauses.
208 Hironori Suzuki
(24) Distribution of alliteration and order of Modal and Non-Finite Verbs (Main Clauses) a. ‘A’-verses MV
VM
total
M , V M, V M , V M, V
0 (0%) 7 (64%) 0 (0%) 0 (0%)
0 (0%) 4 (36%) 0 (0%) 0 (0%)
0 11 0 0
total
7 (64%)
4 (36%)
11
MV
VM
total
M , V M, V M , V M, V
0 (0%) 0 (0%) 0 (0%) 12 (100%)
0 (0%) 25 (100%) 0 (0%) 0 (0%)
0 25 0 12
total
12 (32%)
25 (68%)
37
MV
VM
total
M , V M, V M , V M, V
2 (100%) 0 (0%) 0 (0%) 12 (100%)
0 (0%) 0 (0%) 0 (0%) 0 (0%)
2 0 0 12
total
14 (100%)
0 (0%)
14
MV
VM
total
M , V M, V M , V M, V
0 (0%) 22 (100%) 0 (0%) 28 (100%)
0 (0%) 0 (0%) 0 (0%) 0 (0%)
0 22 0 28
total
50 (100%)
0 (0%)
50
b. ‘B’-verses
c.
The same lines
d. Different lines
Overall Total (a–d) 83 (74%)
29 (26%)
112
On MV/VM order in Beowulf 209
The figures in (24a) are concerned with cases where the modal and non-finite verbs appear within the same ‘a’-verse. For these cases, where the non-finite verb alliterates and the modal auxiliary does not alliterate within the same halfline, the typical VM order might be expected, as in the subordinate and ambiguous clauses. However, unexpectedly, a high percentage of the MV order exists (7 out of 11, or 64%, as shown in (24a) and exemplified in (25a) and (25b)). It should be noted that this clearly different picture for ‘a’-verses in main clauses cannot be explained by Bliss’s theory of metrical ‘heaviness’ of the modal auxiliary since the majority of the MV order (4 out of 7) contains the ‘heavy auxiliary’, as illustrated in (25a). One might ascribe the variation in word order in the ‘a’-verse to the possibility of ‘double alliteration’ in the ‘a’-verse, where two accented syllables alliterate. However, the majority (6 out of 7) of instances of the MV order in (24a) manifest only ‘single alliteration’. For clauses where both the modal and non-finite verbs appear in the ‘b’-verse; as presented in (24b), again a tendency similar to that of the subordinate clauses exists in that the VM order occurs exclusively where only the nonfinite verb alliterates (all 25 instances), and the MV order dominates completely where neither of them alliterates (all 12 instances). Examples of the VM and MV orders are illustrated in (25c) and (25d) respectively, The figures in (24c) are concerned with cases where the modal and nonfinite verbs appear in the same line. All the instances show the MV order with the modal auxiliary in the ‘a’-verse and the non-finite verb in the ‘b’-verse. There are two instances with the modal verb alliterating and the non-finite verb not alliterating, as illustrated in (25e), and there are twelve instances with neither of them alliterating in the same line, such as example (25f). The figures in (24d) show cases where the modal and non-finite verbs occur in different lines. As also found in the subordinate clauses, all fifty instances take the MV order. There are twenty-two instances where the modal does not carry the alliteration in the line and the non-finite verb alliterates in a different line, as illustrated in (25g), and twenty-eight instances where neither the modal nor non-finite verb alliterates in the different lines, as illustrated in (25h). Thus, the results of (24c) and (24d) again show that only the MV order occurs across the half-line boundary, as is the case with subordinate clauses.
210 Hironori Suzuki
(25) a.
MV
b. V M
c.
V M
d. MV
e.
M V
f.
MV
g. M/V
h. M/V
Wille ic asecgan sunu Healfdenes mærum þeodne min ærende aldre þinum (344–6a) I will tell my errand to Healfdene’s son, the great prince your lord no on wealle læ[n]g bidan wolde (ac mid bæle for) (2307b–8a) It would not stay longer within the walls (snyttrum besyrwan.) Hwæt þæt secgan mæg efne swa hwylc mægþa swa ðone magan cende æfter gumcynnum, gyf heo gyt lyfað, þæt hyre ealdmetod este wære bearngebyrdo. (942b–6a) Yes, she may say, whatever woman brought forth this son among mankind — if she still lives — that God of Old was kind to her child-bearing (godum togenes) Nu sceal gled fretan weaxan wonna leg wigena strengel þone ðe oft gebad isernscure (3114b–6) Now shall flame eat the chief of warriors — the fire shall grow dark — who often endured the iron-shower nænig þæt dorste deor geneþan swæsra gesiða nefne sinfrea þæt hire an dæges eagum starede (1933–5) no bold one among her retainers dared venture — except her great lord — to set his eyes on her in daylight sceolde [ofer] willan wic eardian elles hwergen (2589–90a) Against his will he must take up a dwelling place elsewhere (searwum gesæled) Sinc eaðe mæg gold on grunde gumcynnes gehwone oferhigian (hyde se ðe wylle) (2764b–6a) Easily may treasure, gold in the ground, betray each one of the race of men (siða gesunde) Ic to sæ wille wið wrað werod wearde healdan (318b–9) I will return to the sea to keep watch against hostile bands
Table (26) gives figures concerning ond/ac clauses. These figures clearly show a similar tendency to that in subordinate clauses.
On MV/VM order in Beowulf
(26) Distribution of alliteration and order of Modal and Non-Finite Verbs (Ond/Ac Clauses) a. ‘A’-verses
M, V
MV
VM
total
0 (0%)
1 (100%)
1
MV
VM
total
0 (0%)
2 (100%)
2
MV
VM
total
3 (100%) 4 (100%)
0 (0%) 0 (0%)
3 4
b. ‘B’-verses
M, V c.
Different lines
M, V M, V
Overall Total (a–c) 7 (70%)
3 (30%)
10
In (26a) and (26b), where only the non-finite verb alliterates within the halfline boundary, the typical VM order occurs. There are three instances: one in the ‘a’-verse and the other two in the ‘b’-verse, as in (27a). For clauses where the modal and non-finite verbs appear in different lines, as is common with main and subordinate clauses, all the seven instances take the MV order, as illustrated in (27b). Thus, across the half-line boundary, again only the MV order is observed. (27) a.
VM
b. M/V
(hafalan hydan) ac he me habban wile d[r]eore fahne (gif mec deað nimeð) (446b–7a) but he will have me blood-stained (forgrand gramum) ond nu wið Grendel sceal wið þam aglæcan, ana gehegan ðing wið þyrse (424b–6a) and now, alone, I must settle the affair with that giant, the monster Grendel
211
212 Hironori Suzuki
7. Conclusions The following conclusions can be drawn regarding the factors affecting the word order of the modal auxiliary and its non-finite verb complement. First of all, unlike the tendency in Old English prose, the influence of the extra element cannot be observed in the subordinate clauses in Beowulf. Secondly, the influence of ‘heaviness’ cannot be observed either, with the single exception of the metrical ‘heaviness’ of the modal auxiliary. Instead, the far more important factor in choosing between the MV and VM orders in subordinate clauses is the distribution of alliteration. Thus, the VM order is always observed when only the non-finite verb alliterates and both verbs appear within the same half-line boundary. On the other hand, the MV order is always observed either when the alliteration pattern differs from the above, or when the modal and non-finite verbs are separated by a half-line boundary. The same is true of ambiguous clauses and ond/ac clauses. Furthermore, this phenomenon can also generally be observed in main clauses in Beowulf. However, the unexpected prevalence of the MV order, with the non-finite verb alliterating, in main clauses of ‘a’-verses presents a different picture from that of ‘b’-verses. This fact demonstrates that a clear distinction between not only the clause types but also between the main clauses of ‘a’-verses and ‘b’-verses should be made in the investigation of verse syntax in Old English. The present study has established the nature of the tendencies that govern the word order of the modal and its non-finite verb complement in Beowulf, with particular stress on the importance of alliteration as an organizing factor, and it is hoped that this will encourage the examination of many more texts, leading to a deeper understanding of verse syntax in Old English.
References Bliss, Alan J. 1967. The Metre of Beowulf. Oxford: Blackwell. ———. 1981. “Auxiliary and Verbal in Beowulf”. Anglo-Saxon England 9.157–182. Blockley, Mary. 2001. Aspects of Old English Poetic Syntax. Urbana & Chicago: University of Illinois Press. Butler, Christopher. 1985. Statistics in Linguistics. Oxford: Blackwell. Davis, Graeme. 1997. The Word-Order of Ælfric. Lewiston, Queenston, & Lampeter: The Edwin Mellen Press. Donoghue, Daniel. 1987. Style in Old English Poetry: The Test of the Auxiliary. New Haven: Yale University Press.
On MV/VM order in Beowulf
——— & Bruce Mitchell. 1992. “Parataxis and Hypotaxis: A Review of Some Terms Used for Old English Syntax”. Neuphilologische Mitteilungen 93.163–183. Klaeber, Fr., ed. 1950. Beowulf and the Fight at Finnsburg. Boston: Heath. Kobayashi, Eichi, ed. 1991. The Story of Apollonius of Tyre in Old and Middle English. Tokyo: Sanshusha. Kohonen, Viljo. 1978. On the Development of English Word Order in Religious Prose around 1000 and 1200 AD: A Quantitative Study of Word Order in Context. Åbo: Åbo Akademi. Mitchell, Bruce. 1985. Old English Syntax. 2 vols. Oxford: Clarendon Press. ——— & Fred C. Robinson, eds. 1998. Beowulf: An Edition. Oxford: Blackwell. Ogura, Michiko. 2004. “Variations and Diachronic Changes of Alliterative Patterns and Alliterative Elements”. Report on the Research Projects 80 ed. by Michiko Ogura, 23–49. Chiba: Graduate School of Social Sciences and Humanities, Chiba University. Ohkado, Masayuki. 2000. “On MV/VM Order in Old English”. Folia Linguistica Historica XX:1–2.79–106. ———. 2001. Old English Constructions with Multiple Predicates. Tokyo: Hituzi Syobo. Pintzuk, Suzan & Anthony S. Kroch. 1989. “The Rightward Movement of Complements and Adjuncts in the Old English of Beowulf”. Language Variation and Change 1.115–143. Pope, John C. 1966. The Rhythm of Beowulf. New Haven: Yale University Press. Russom, Geoffrey. 1987. Old English Meter and Linguistic Theory. Cambridge: Cambridge University Press. Sievers, Edward. 1893. Altgermanische Metrik. Halle: Max Niemeyer. Suzuki, Hironori. 1993. “Two Determinants of Word Order in the Parker Chronicle”. Aspects of Modern English: The Tenth Anniversary Issue of Modern English Association ed. by Modern English Association, 376–392. Tokyo: Eichosha. ———. 1994. “Two Determinants of the Position of Objects in Apollonius of Tyre”. Bulletin of the Faculty of Humanities, Ohu University 6.89–105. Suzuki, Seiichi. 1996. “On the Syllable Weight of — VC# in Old English: A Metrical Perspective”. English Historical Linguistics 1994: Papers from the 8th International Conference on English Historical Linguistics (8. ICEHL, Edinburgh, 19–23 September 1994) ed. by Derek Britton, 39–55. Amsterdam & Philadelphia: John Benjamins.
213
DARE and NEED in British and American present-day English 1960s–1990s Martine Taeymans University of Antwerp, Limburg University Centre
1.
Introduction
This study will throw light on recent changes in the behaviour of the marginal modals dare and need on the one hand, and the semi-modals dare to and need to on the other, in British and American PDE writings. Through a diachronic comparison of four corpora of British and American English from the 1960s and 1990s, an attempt will be made to reveal whether there have been any significant shifts in frequency of the respective variants of both verbs. Changes in frequency may tell us more about how DARE and NEED have fared with respect to (de)grammaticalization.1 1.1 Synchronic variation with DARE and NEED: Formal features In PDE DARE and NEED both show variable morphology and syntax. When used as modals, they meet most of the criteria identified by Quirk et al. (1985: 121–128) to define (modal) auxiliaries. The most striking ones are that they act as operators in negative and interrogative clauses as in (1); the third person singular inflection for the simple present is absent (cf. (2)); there are no imperative, finite or participial forms; and they take bare infinitival complements as in (3):
1.DARE and NEED in capitals refer to lexemes. Lower case is used for the modal and semimodal variants, dare/need and dare to/need to.
216 Martine Taeymans
(1) Colmore was ready to leave, but delayed his departure, as one who dare not go to bed early for fear of missing some wholly unanticipated but remotely possible event of absorbing interest. (LOB, K01 15) (2) It was not said as a new discovery, but as a moral platitude, something so obvious that it need only be mentioned to be accepted. (FLOB, G52 9) (3) “How dare you wear fur?” (FLOB, R05 212)
The semi-modals dare to and need to are different in that they are semantically related to the central modals, but pattern entirely like main verbs in showing regular verb morphology (e.g. 3sg -s (cf. (4)) and past forms), taking do-support in negative and interrogative constructions (cf. (5)), having non-finite forms (cf. (6)), co-occurring with central modals (cf. (7)) and taking a to-infinitival complement (cf. (8)): (4) The college teacher needs the stimulus of communication with other faculty members but he also needs to feel that such communication, even informal debates over the luncheon table, are a contribution to the total good of the institution. (BROWN, A35 159) (5) He must not mince his words and he does not need to. (FLOB, B04 90) (6) A Resident of Mere was moaning because a squirrel had dared to eat nuts from a tree in her garden. (LOB, B27 47) (7) We would only need to have a dozen clubs to make the League worthwhile. (LOB, A33 142) (8) He did not, however, dare to attack most of the graft and licences. (FLOB, B12 85)
For DARE a third variant needs to be discerned, i.e. the blend constructions with DARE. These display a mixed use of both main verb syntax (e.g. 3sg inflection or do-support) and auxiliary syntax (e.g. a bare infinitival complement). Blend uses are illustrated in examples (9)–(10): (9) We are left helpless to cope with it because we do not dare speak of it as anything real of fear that to do so would imply a commitment to that which has already been discredited and proved false. (BROWN, D01 103) (10) The professor in turn dares not tolerate the influence in his classes of an organization in the policies and standards of which he has no voice. (BROWN, E29 109)
DARE and NEED in British and American present-day English 217
1.2 Synchronic variation reflects diachronic change The synchronic variation of marginal and semi-modals and blend constructions noted in 1.1 reflects diachronic change. Until the end of the ME period DARE was a preterite present verb taking a bare infinitive or directional phrase as its complement.2 In ENE the forms and constructions typical of a modal verb continued to exist, but “[…] [dare] simultaneously develop[ed] main-verb characteristics that it had not displayed before” (Beths 1999:1069), so that semimodal dare to came to compete with modal dare. Since this implies a shift from modal to main verb morphosyntax, the unidirectionality claim central to grammaticalization seems to be countered. Therefore, DARE’s development has been considered an example of degrammaticalization (Beths 1999). NEED’s history is different in many respects. In ME it occurred in a range of personal and impersonal constructions, but in the sixteenth century it started to show modal features, perhaps in response to the loss of the verb tharf ‘need’, and “[…] moved somewhat toward modal morphology and away from contiguous to [−infinitives] in negatives and questions” (Nagle 1989: 104). This shift towards a more modal syntax implies that NEED grammaticalized, by analogy with true central modals such as e.g. can, may and must. Although their history is different, the result is the same: variable usage with DARE and NEED persists until today and is the natural outcome of change in progress. However, their diachronic development and predominant status in PDE remain a controversial issue. If one accepts the common view that grammaticalization proceeds along a cline from lexical to more functional (Traugott & Heine 1991; Hopper & Traugott 1993), one would expect that diachronically there has been an increase of grammatical, i.e. modal, usage for NEED. Information on DARE is scarcer. But since scholars generally seem to agree on the fact that the verb DARE developed main verb characteristics from the sixteenth century (Beths 1999) or seventeenth century onwards (Visser 1970; Nagle 1989), one might expect an increase of the main verb at the cost of modal dare. However, Barber claims that both “[…] dare and need are ceasing to be auxiliaries, and are coming more and more to be used as ordinary lexical verbs” (1993: 275f). And Traugott (2001), in a talk given at the University of Freiburg,
2.Preterite present verbs are a group of originally strong Old English verbs whose original preterite forms came to be used as present tense forms, while new weak forms and past participles were introduced. The most important verbs in this category are witan, agan, *dugan, unnan, *durran, *sulan, *motan and magan. Most of them eventually turned into central modal verbs.
218 Martine Taeymans
states that “[c]hanges do not have to occur. They also do not have to go to completion, in other words they do not have to move along a cline, or even continue down it once they start out on it”. Traugott’s view implies that a grammaticalization or degrammaticalization process may stop before it is fully implemented and seriously questions the assumptions made above. Since only empirical evidence can reveal what directions DARE and NEED have taken in the course of time and what their status in PDE really is, I decided to conduct a corpus-based study, taking present-day English as a starting point.
2. Aims of the study Through a short-term diachronic comparison of corpora of British and American English, an attempt will be made to reveal whether there have been any significant shifts in frequency of the respective variants of DARE and NEED. Quantitative data will be presented that bear on the following general questions: a. What is the predominant status of DARE and NEED in American and British PDE? To what extent do they have functional as opposed to main verb status? To what extent does DARE find itself in between the auxiliary and main verb categories by appearing in blend constructions? b. When we look at their frequency and distribution over a 30-year period, do we find evidence for the claim that both DARE and NEED are ceasing to be auxiliaries (Barber 1993: 275f)? Or are they indeed moving in opposite directions (cf. Nagle 1989 on NEED; Beths 1999 on DARE)?
3. Materials The corpora consulted are the parallel corpora LOB and FLOB for British English and BROWN and FROWN for American English. All four corpora contain around one million tagged words of written English, with texts published in 1961 for LOB and BROWN, and texts published in 1991 for FLOB and FROWN. The text categories and total number of words within them match as closely as possible across LOB and FLOB, and BROWN and FROWN. For the present study no selection was made as to text types or register. NEED, and especially DARE, are rather infrequent items; therefore all available occurrences were taken into consideration. One of the drawbacks of this study is that the present corpora only com-
DARE and NEED in British and American present-day English 219
prise written texts, while spoken language is more suitable for spotting trends in language change. Unfortunately no parallel corpora of spoken American and British English are currently available. Therefore, I will regularly refer to an earlier study where I looked at the behaviour of DARE and NEED in the written and spoken parts of the British National Corpus (BNC). All BNC figures quoted in this paper are based on random samples of approximately 2000 occurrences for each verb.
4. DARE in American and British PDE: 1960s and 1990s 4.1 The status and behaviour of DARE in PDE In Table 1 we get the overall frequencies of DARE + (to) infinitive in the complete corpora of LOB, FLOB, BROWN and FROWN. Direct comparisons among all four corpora are possible because they are the same size. Considering the small number of occurrences found for DARE in both British and American writings, tendencies should be regarded with great caution. Any apparent signs of change will be checked against evidence from the spoken BNC. A comparison of the total number of occurrences of DARE in the 1960s and 1990s reveals that DARE is becoming increasingly rare in PDE, with 38 attestations of DARE + (to) infinitive in LOB versus only 28 occurrences in FLOB, and 36 versus only 30 occurrences in BROWN and FROWN respectively. As for DARE in British PDE, we notice a clear preference for modal dare, accounting for 50% and 53.6% of the total number of occurrences in LOB and FLOB respectively, while blend and semi-modal constructions occur in more or less similar proportions, with a slight preference for the semi-modal dare to. A completely different picture emerges from the American PDE data: here, the semi-modal construction dare to clearly is the predominant one, accounting for 55.5% and 43.3% of the total occurrences in BROWN and FROWN respectively. Blend constructions account for some 30% in both corpora, while modal dare is rare. It appears that the use of modal dare has slightly increased at the cost of the semi-modal dare to in recent American English (compare only 13.5% in BROWN versus 26.7% in FROWN). However, the total number of occurrences looked at is probably too small to draw any definite conclusions. We can conclude from Table 1 that DARE is still very modal-like in British PDE, while it mostly occurs as a semi-modal in American PDE. The claim that DARE is moving more and more towards main verb syntax thus only seems to
220 Martine Taeymans
Table 1.DARE + (to) infinitive in the complete corpora of LOB, FLOB, BROWN and FROWN.
LOB FLOB BROWN FROWN
modal dare
% of total blend dare
19 15 5 8
50% 53.6% 13.9% 26.7%
8 5 11 9
% of total semi-modal % of total total dare to 21.1% 17.8% 30.6% 30%
11 8 20 13
28.9% 28.6% 55.5% 43.3%
38 28 36 30
hold for DARE in American PDE. The modal predominance of DARE in British PDE is confirmed by (more innovative) data from the spoken BNC, with 192 out of 297 occurrences of DARE + (to) infinitive being modal. This trend was also noted by Hundt (1998: 65) and Krug (2000: 201), whose research on DARE “[…] yielded no trend towards main verb usage on the diachronic axis”, either. 4.1.1 DARE in frozen constructions Although overwhelming evidence points to the fact that DARE in British PDE is very modal indeed, the situation remains somewhat puzzling, given the fact that DARE has been adopting unequivocal main verb characteristics at least from the seventeenth century onwards. A possible explanation for the seemingly modal predominance of DARE in British PDE may lie in the relatively frequently occurring frozen constructions I dare say, dare I say (it) and the idiomatic expression How dare you (+ bare infinitive), exemplified in (11) and (12): (11) How dare you refuse to acknowledge the divinity of the one Lord above? (FROWN, K21 147) (12) “I dare say he was milder than he looked,” he said carelessly. (LOB, P27 130)
In LOB, 6 out of 19 occurrences with modal dare are either of the I dare say or How dare you + bare infinitive type and in FLOB as many as 8 out of 15 occurrences with modal dare are of the same types. Interestingly, I dare say is not attested in BROWN or FROWN, which may partially account for the fact that modal dare is considerably less frequent in the American than in the British corpora. The idiomatic expression How dare you (+ bare infinitive), however, does occur in BROWN and FROWN; in fact, most of the occurrences of modal dare in American English are of this type. The above-mentioned frozen or idiomatic expressions are clearly modal in structure, but may obscure our view of DARE’s actual development: I dare say
DARE and NEED in British and American present-day English 221
and how dare you are relatively frequent, entrenched phrases which tend to resist any change and may well assure the conservation of older structures. 4.1.2 DARE in (implicitly) negative clauses If we look at the modal occurrences with DARE in greater detail, it becomes clear that — apart from the frozen expressions I dare say and How dare you — modal dare is exclusively retained in non-assertive contexts (i.e. mostly negative, but also interrogative, conditional and comparative clauses, clauses containing semi-negative adverbials such as hardly and scarcely, etc.). The majority of modal occurrences are directly negated by not. Example (13) shows modal dare under direct negation, while (14) is an example of modal dare used in a non-assertive context. Beths (1999: 1099f) notes that modal dare has been a negative polarity verb since the earliest records, preferring non-assertive over assertive contexts. (13) He dared not ask her, for if it had been an isolated incident, she would be curious and wonder what had provoked his question. (FLOB, L02 20) (14) She hesitated, wondering if she dared approach him, when he suddenly appeared beside her. (FLOB, P15 24)
A closer inspection of the more recent blend constructions, showing a mixture of modal and main verb syntax, reveals that their use is also restricted to nonassertive contexts, the overwhelming majority being negated by do (15) or another modal (16). (15) If I did not dare have sex with my own girlfriend, I could scarcely imagine it with anyone else. (FROWN, G01 154) (16) Certainly, she wouldn’t dare ask her father afterward. (BROWN, N10 38)
This is remarkable because the semi-modal construction dare to, also taking do-support under negation, is unrestricted in its use and freely appears in affirmative and declarative clauses (cf. (4) and (6)). Table 2 provides an overview of the semi-modal occurrences with DARE appearing in truly assertive (or positive) contexts on the one hand, and in non-assertive contexts on the other, showing that the exclusive relation between DARE and negation does not hold for semi-modal dare to. Both the frozen constructions discussed in 4.1.1 and the restriction of modal dare to non-assertive contexts (this section) throw new light on the
222 Martine Taeymans
Table 2.Semi-modal dare to in affirmative and non-assertive contexts.
LOB FLOB BROWN FROWN
dare to in assertive clauses
% of total
dare to in non-assertive contexts
% of total
dare to total
6 2 12 9
54.5% 25% 60% 69.2%
5 6 8 4
45.5% 75% 40% 30.8%
11 8 20 13
modal predominance noted for DARE in the British corpora LOB and FLOB. We have seen that the frozen constructions are responsible for ‘The Conserving Effect’: “[…] the idea that high frequency sequences take on a life of their own, and resist change on the basis of newer productive patterns for juxtaposing words or morphemes” (Bybee & Thompson 2000: 381). Further, DARE has only retained its conservative modal features in negative and non-assertive environments. The result is a relatively high token frequency for modal dare in British PDE.3 However, Bybee & Thompson note that a different type of frequency plays a role in the determination of the productivity of a lexical item, “[…] where productivity is defined as the likelihood that a pattern will apply to new forms” (2000: 384). The kind of frequency determining the degree of productivity of a pattern is called type frequency and “[…] counts how many different lexical items a certain pattern or construction is applicable to” (Bybee & Thompson 2000: 378). While many of the modal constructions with dare have some of the features of an archaic construction preserved through high frequency, e.g. by occurring in frozen expressions and by being limited to negative contexts, the semi-modal construction is perfectly general, occurring with many different verbs and in many different contexts. So, although semi-modal dare to in British PDE has a lower token frequency than modal dare, it has a much higher type frequency, allowing it to be used productively. The fact that dare to, having meanings very close to modal and blend dare, is free to appear in all possible contexts is an asset: it does not seem unlikely that modal dare will further decline and that the semi-modal construction will become the only truly productive one with DARE.
3.It was pointed out by Hundt (1998: 64f) that — on closer inspection — there are no true examples of blend constructions with NEED, as they appear in fictional dialogues and represent foreigner talk.
DARE and NEED in British and American present-day English 223
5. NEED in American and British PDE: 1960s and 1990s 5.1 The status of NEED Table 3 provides overall frequencies of the verb NEED + (to) infinitive in British and American writings from the 1960s and 1990s. Whereas the use of DARE is gradually decreasing in PDE, we see that the use of NEED is rising spectacularly in British and American PDE (notice 244 occurrences in FLOB and 172 in FROWN vs. only 132 and 107 in LOB and BROWN respectively). This rise is caused by a sharp increase in the use of semi-modal need to (56 versus 200 in LOB and FLOB; 71 versus 141 in BROWN and FROWN). At the same time, we see that modal need is becoming increasingly rare in both varieties of English (76 versus 44 in LOB and FLOB; 36 versus 31 in BROWN and FROWN). NEED thus shows the same tendencies in British and American PDE, although the rise of semi-modal need to is less outstanding in American than in British PDE. This is probably due to the fact that modal need in American English was already quite rare in the 1960s. It seems that British English has done some serious catching up with American English over the past 30 years in favouring the use of semi-modal need to over modal need. The above findings raise questions about the grammaticalization of NEED. Some historical grammars and descriptions indicate first signs of modal characteristics with NEED somewhere between the end of the fourteenth century (Visser 1970) and the early sixteenth century (Warner 1993), but in an investigation of the use of NEED in Shakespeare, Krug found that, by the end of the sixteenth century, “[…] modal constructions by far outnumber main verb constructions” (2000:202). So, although we have evidence for the grammaticalization of NEED towards the end of the sixteenth century, PDE data show that the process must have been reversed since. Modal need has now become a rare form on both sides of the Atlantic. Table 3.NEED + (to) infinitive in the complete corpora of LOB, FLOB, BROWN and FROWN.
LOB FLOB BROWN FROWN
modal need
% of total
semi-modal need to
% of total
total
76 44 36 31
57.6% 18% 33.6% 18%
56 200 71 141
42.4% 82% 66.4% 82%
132 244 107 172
224 Martine Taeymans
5.1.1 NEED in (implicitly) negative contexts Like DARE, NEED and negation are closely related. The intimate relation between NEED and negation is also most striking in its modal use. Over 70% of the modal occurrences attested in all four corpora are directly followed by not (17), while the remaining 30% is restricted to other non-assertive contexts such as questions, comparative and conditional clauses, and clauses containing seminegative adverbials such as hardly, scarcely and only (cf. (2)): (17) The historian need not be concerned with the philosophical problems suggested by religion. (BROWN, G65 135)
NEED was pressed into the role of negative auxiliary in the course of the late ME/beginning of the ENE period. This may have happened to fill the gap left by must, which expressed notions of necessity but could not express absence of necessity, i.e. “it is not necessary that”. The fact that modal need is now restricted to non-assertive contexts need not surprise us: it was noted by Heine (1995: 46) for German modals that their most conservative behaviour is best retained in negative and interrogative rather than affirmative and declarative sentences. Like dare to, the semi-modal construction need to is unrestricted in its use and thus free to occur in assertive contexts (cf. (4)). Table 4 provides an overview of the semi-modal occurrences with NEED appearing in assertive contexts on the one hand, and in non-assertive contexts on the other, showing that the exclusive relation between NEED and negation does not hold for the semi-modal need to. Table 4.Semi-modal need to in affirmative and non-assertive contexts.
LOB FLOB BROWN FROWN
need to % of total need to in assertive clauses in non-assertive contexts
% of total need to total
46 170 53 116
17.8% 15% 25% 17.7%
82.2% 85% 75% 82.3%
10 30 18 25
56 200 71 141
Section 5.1 has shown that modal need is becoming increasingly rare in favour of semi-modal need to, which expresses more or less the same notion of necessity in negative sentences, but is free to appear in positive affirmative constructions as well. The overwhelming evidence here points to the conclusion that NEED has mainly become a semi-modal verb.
DARE and NEED in British and American present-day English 225
6. Conclusion This study has shed light on the different variants of DARE and NEED as they are used in British and American PDE. By looking at their token frequencies over time, recent changes as well as their predominant status in British and American PDE have become apparent. For NEED, corpus data reveal unequivocal evidence for the fact that it has largely become a semi-modal auxiliary. Although the more conservative modal construction is still found in nonassertive contexts, a short-term diachronic comparison predicts that it will further decline in both varieties of English. As for DARE, research findings seemed arbitrary at first, with DARE being very modal-like in British PDE on the one hand, but showing mainly semi-modal behaviour in American PDE on the other. On closer inspection, it appears that the modal predominance in British PDE is mainly caused by a number of highly entrenched modal constructions which are relatively frequent and resist any change. This ‘Conserving Effect’ is responsible for a fairly high token frequency for modal dare in British PDE. It is the semi-modal construction, however, which is unrestricted in its use and is the most productive variant. Although the corpora used are limited in scope and only cover a short time span, it seems reasonably safe to conclude that both DARE and NEED are moving from modal to main verb morphosyntax. While the unidirectionality hypothesis central to grammaticalization is said to apply on all levels, i.e. the semantic, syntactic and phonological levels (Fischer et al. 2000: 19f), the evidence presented here suggests that it is breached on a structural level: while NEED had become very modal-like by the end of the sixteenth century, its modal variant is now obsolescent in favour of semi-modal need to. As for DARE, the more recent semi-modal variant is becoming the only truly productive one in PDE, featuring main verb morphology and syntax. In order to explain the motivation of the changes noted in this paper, it will be necessary to look at older corpora and outline a detailed historical development for both verbs. Since data on DARE and NEED are limited in BROWN and FROWN it will also be interesting to study their behaviour in the American National Corpus once it has been released, as it will comprise ten million words of spoken and ninety million words of written American PDE.
226 Martine Taeymans
References Barber, Charles. 1993. The English Language: A Historical Introduction. Cambridge: Cambridge University Press. Beths, Frank. 1999. “The History of Dare and the Status of Unidirectionality”. Linguistics 37.1069–1110. Bybee, Joan & Sandra Thompson. 2000. “Three Frequency Effects in Syntax”. Berkeley Linguistics Society, 23.378–388. Fischer, Olga, Annette Rosenbach & Dieter Stein. 2000. Pathways of Change. Grammaticalization in English. Amsterdam & Philadelphia: John Benjamins. Heine, Bernd. 1995. “Agent-Oriented vs. Epistemic Modality: Some Observations on German Modals”. Modality in Grammar and Discourse ed. by Joan L. Bybee & Suzanne Fleischman. 19–53. Amsterdam & Philadelphia: John Benjamins. Hopper, Paul J. & Elizabeth C. Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Hundt, M. 1998. New Zealand English Grammar — Fact or Fiction? A Corpus-Based Study in Morphosyntactic Variation. Amsterdam & Philadelphia: John Benjamins. Krug, Manfred. 2000. Emerging English Modals. A Corpus-Based Study of Grammaticalization. Berlin & New York: Mouton de Gruyter. Nagle, Stephen J. 1989. Inferential Change and Syntactic Modality in English. Frankfurt, Main: Lang. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Traugott, Elizabeth C. 2001. “Legitimate Counterexamples to Unidirectionality”. Paper presented at Freiburg University, Germany, October 2001. http://www.stanford.edu/ ~traugott/papers/Freiburg.Unidirect.pdf ——— & Bernd Heine. 1991. Approaches to Grammaticalization. (= Typological Studies in Language, 19.) Amsterdam & Philadelphia: John Benjamins. Visser, Frederikus Th. 1970. An Historical Syntax of the English Language (3 vols.). Leiden: Brill. Warner, Anthony R. 1993. English Auxiliaries: Structure and History (= Cambridge Studies in Linguistics, 66.) Cambridge: Cambridge University Press.
Sources The British National Corpus (BNC), current British English, comprising 10m words of spoken and 90m words of written British PDE. Brown Corpus of Standard American English (BROWN), texts from 1961, 1m words of written American English. Freiburg Version of the Lancaster-Oslo-Bergen Corpus (FLOB), texts from 1991, 1m words of written British English.
DARE and NEED in British and American present-day English 227
Freiburg Version of the Brown Corpus (FROWN), texts from 1991, 1m words of written American English. Lancaster-Oslo-Bergen Corpus (LOB), texts from 1961, 1m words of written British English.
What drove DO?* Anthony Warner University of York
In late Middle and Early Modern English there were several centuries of variation in the use of periphrastic DO as an alternative to the simple finite verb (see Denison 1993, Rissanen 1999: 239ff, Nurmi 1999, and references). In this variation the proportion of DO increased over time in the categories in which it is used today, in particular in questions, negative declaratives, and negative imperatives. Variation in these categories is illustrated in (1). (1) Examples from Thomas Otway Cheats of Scapin; Friendship in Fashion 1677–8 (Works 1812), Ellegård’s (1953) text 377; references by page and line. a. affirmative direct yes-no questions: Lord, madam, do you know what you do? (324–8) Alas, sir, think you the captain has so little wit as to accept of such a poor rascally fellow as I am instead of your son? (229–37) b. negative declarative sentences (with not): she does not deserve it, … (317–4) I question not your friendship in the matter, … (291–23) c. negative imperatives: don’t run yourself into danger thus rashly. (320–7) get into the sack, and stir not, whatever happens; …(237–31)
We might think that forms with DO had some advantage for learners or users of the language, or that there was some aspect of the acquisition process or a grammatical bias at acquisition which tended to favour it. There have been
*I am very grateful to Tony Kroch, who gave me the initial database used here, to my colleague Ann Taylor, who had compiled it, and to Celeste Tereszczuk, who completed it by entering text for the seventeenth century. I am also very grateful to the British Academy for the Research Readership which gave me time to pursue this matter, and wish to thank my colleagues at York for comments received after a departmental research seminar.
230 Anthony Warner
various suggestions here (see Denison 1993, Rissanen 1999 for some discussion), but in this paper I want to consider one in particular because it seems initially plausible, and because it was proposed on the basis of quantitative data, and can be further investigated on the same, precise basis. The proposal was simply that processing considerations favoured DO in transitive questions where the object followed the verb, and that this was responsible for the general increase in the use of DO (Kroch 1989b, developing Kroch, Myhill and Pintzuk 1982). In investigating this proposal I shall use an electronic version of Ellegård’s (1953) database in which the individual examples are coded for Ellegård’s factors and for a series of further relevant factors, and I shall restrict my attention to yes-no questions and wh-adverbial questions (negative and affirmative), omitting questions where the wh-word is the object of the verb.
1.
The ordering of question contexts and processing complexity
Ellegård noted that the transitivity of the verb was associated with higher rates of DO in inverted questions where the object followed the verb (1953: 203). In yes-no and adverbial wh-questions the identity of the subject and object is also relevant, at least for 1490–1600 (as established by Kroch, Myhill and Pintzuk 1982). At this period transitives with a full NP subject (as in 2a, b) consistently show higher rates of DO than transitives with personal pronoun subject (as in 2c, d), as do transitives where the object is a personal pronoun (as in 2a, c) rather than a full NP (as in 2b, d).1 (2) a.
Transitive verb, NP subject, personal pronoun object: Alas wherfore lyghteth me the sonne, (Valentine and Orson. 304. 25–23) howh mykil avayleth the the feith / (St Bartholemew’s Church. 190. 21–24) b. Transitive verb, NP subject, NP object: Why moved and stered Phaleryus the kynge Ptholome, oft and dilygently to rede bokes? (Bourchier Froissart. 309. 5–22) why breke thy dyscyples the tradicions of our olde faders: (Fitz James Sermo Die Lunae. 299. 74–1)
1.The number immediately after the title of the text in these examples is the number assigned to the text in Ellegård’s (1953) Part 3; it is followed by Ellegård’s reference.
What drove DO?
c.
Transitive verb, personal pronoun subject, personal pronoun object: Why folowyste thou me, kychyn knave? (Malory Works. 243. 308–15) Peter, lovest thou me? (Hugh Latimer, Letters. 323. 348–7–11) d. Transitive verb, personal pronoun subject, NP object: Byleuist thou this? (J Fisher English Works 305. 308–13) Why secheste thou reste, whan thou arte ordeyned in this lyfe to laboure. (Atkinson De Imitatione Christi. 302. 188–35) e. Intransitive verb: Why eateth youre master with publicans and synners? (Tyndale Four Gospels. 310. Matthew 9–11) Why wepe you lyke a childe? (Bourchier Guevara’s Golden Book. 319. 4518)
Figures for the period 1465–1575 are given in Table 1. The percentages simply refer to the period as a whole. The most illuminating statistic is the probabilistic coefficient, which is derived from a run in the logistic regression package GoldVarb (Sankoff and Rand 1990) in which other relevant factor groups were present (including subdivisions of period of time).2 This coefficient is directly related to the model’s prediction of the comparative rates of incidence of DO in the different categories. These results are parallel to those obtained by Kroch, Myhill and Pintzuk Table 1.Incidence of DO in Yes-no and Adverbial Wh-questions 1465–1575.
a b c d e f
subject
object
DO%
ProbCoeff
n
NP NP NP pers prn pers prn pers prn
pers prn NP intrans pers prn NP intrans
96% 81% 47% 50% 43% 30%
0.978 0.868 0.563 0.582 0.493 0.366
25 57 96 131 260 475
Total
1044
2.Other significant factors in this run: period of text; question type; affirmative vs negative. In Table 2 (for 1575–1710) the first two of these other factors were significant; in Table 3 (for 1465–1500) none were. The data throughout omit parenthetical questions and main verb DO in line with Ellegård’s practice.
231
232 Anthony Warner
(1982) for a sample of Ellegård’s data taken from the period 1490–1600. Kroch (1989b) and Kroch, Myhill and Pintzuk suggested that the results were best interpreted in terms of relative processing complexity. The justification is roughly as follows. With the decline of V2 order, English increasingly prefers adjacency of verb and object, and this increases the processing complexity of inversion of the full verb when the object also follows. The use of DO avoids this complexity. Hence the major difference between the transitive group and the intransitives in incidence of DO (so in Table 1 there is more DO in (a), (b) than (c); more in (d), (e) than (f)). The complexity is increased if the element intervening between verb and object is salient: that is, if it is a full NP rather than a personal pronoun (which is typically a clitic). So transitives with NP subject show more DO than those with a personal pronoun subject (more DO in Table 1 in (a), (b) than in (d), (e)). Examples (under (a) in Table 1) in which a clitic object pronoun precedes the subject have object-subject order. This is highly unusual, hence problematic. They also suggest that speakers prefer to avoid a sequence of clitics, as in the examples of (d) in Table 1, so that (d) shows more DO than (e). This justification seems broadly appropriate, and it is clear that the claim that these differences of distribution involve processing, being related in some way to parsing complexity (or associated production strategies), is a rather plausible one. This does not however extend to a general difference in the incidence of DO between intransitives with NP and pronoun subjects; the relative incidence of DO in these categories is not distinct in the seventeenth century. There is a range of possible models of processing complexity (see the discussion in Kroch 1989b). For the sake of concreteness I will here outline one particular interpretation. It involves a rather natural way of thinking about the possible relevance of parsing, supposing that the progress of the change might depend on accommodation between language users. If individuals alter their level of use of DO to accommodate to their interlocutors, and if this can lead to a permanent change within the individual (cf. Trudgill 1986), then we can sketch a mechanism to account for the progress of the change. It is one of the models discussed in Kroch (1989b), and further in Sprouse and Vance (1999). Applied to the present case the proposal would be that parsing difficulties interfered with a hearer’s perception of the relative frequency of the two forms, DO and the simple finite. More instances of transitive questions with the simple finite would have been misperceived, or interpreted without undergoing a complete syntactic parse. So hearers would have perceived a higher proportion of DO, and a lower proportion of the simple verb than was actually produced.
What drove DO? 233
Hearers thus accommodated not to the actual relative frequencies of DO and the simple finite, but to their misperception of these frequencies. Once a hearer had adopted the new relative frequencies, they fed into production and became in turn the input to other hearers, again subject to misperception. Hence the change made cumulative progress with successive accommodations. This interpretation involves some assumptions. Change could progress through a differential loss of data in parsing which took place at acquisition rather than in adult use (as noted in Pintzuk, Tsoulas, and Warner 2001: 13–14). But we will find reason to reject this as a sole mechanism below. More seriously there is the point made by Kroch concerning an implication of the Constant Rate Effect. This says that where a single underlying change is manifested in different surface contexts, the s-curves of change in those different contexts are parallel. The implication is that “the pattern of favoring and disfavoring contexts does not reflect the forces pushing the change forward. Rather it reflects functional effects, discourse and processing, on the choices speakers make among the alternatives available to them in the language as they know it; and the strength of these effects remains constant as the change proceeds”. (1989a: 238). Kroch notes here that his discovery of the Constant Rate Effect “cut[s] against the assumption” (1989a: 238) which underlay the discussion of Kroch, Myhill and Pintzuk (1982). But the Constant Rate Effect shows us that speakers are exercising a single choice, and do not alter relationships between the contexts of a change across time. This need not imply that they are insensitive to differences between the frequencies they produce and those they perceive, or that they do not accommodate to what they perceive. Clearly, though, they must do this globally; a subliminally perceived difference in transitive questions must be interpreted as part of an overall difference affecting all contexts. So I shall assume that it is possible in principle for functional effects to play some role in change, though the existence of the Constant Rate Effect means that it is not possible to argue with confidence from the ordering of contexts to the strength of corresponding effects (Kroch 1989a: 240). Returning now to the plausibility of the proposal developed above, two questions immediately arise. The first is: is a similar pattern found elsewhere in Ellegård’s data? If differential parsing loss motivates change, we might reasonably expect the pattern of Table 1 to be stable. The second question can be put simply, but involves a complex of assumptions. Simply put it is: is there any evidence that change involved the use of the language, rather than its acquisition?
234 Anthony Warner
2. Is a similar pattern found elsewhere in Ellegård’s data? Is the hierarchy of Table 1 stable across all of Ellegård’s data: both before 1465 and after 1575? and is it stable for subperiods? Figures for the earlier and later periods are given in Tables 2 and 3. These show the same rank ordering of the frequencies of DO as in Table 1; with subject NP, complements are ordered: personal pronoun > NP > intransitive.3 The same ordering is found with personal pronoun subject. And within each category of object (personal pronoun, NP) the ranking of subjects is NP > personal pronoun. Moreover, when the two major periods are each divided into three subperiods, the same pattern of relationships is to be seen. So this is a stable pattern which holds across all the data from 1400 to 1710. And for each of the periods it is statistically highly significant. This is clearly in accordance with the suggestion that this aspect of parsing is responsible for the progress made by DO — though the effect could, of course, follow for quite other reasons: there is no argument for the relevance of parsing, only the failure of a possible argument against. At each period yes-no and wh-adverbial transitive questions are in advance of other categories. And the highest transitive question category goes to completion early in the sixteenth century as far as Ellegård’s data are concerned. We might suppose that
Table 2.Incidence of DO in Yes-no and Adverbial Wh-questions 1575–1710.
a b c d e f
subject
object
DO%
ProbCoeff
n
NP NP NP pers prn pers prn pers prn
pers prn NP intrans pers prn NP intrans
100% 95% 58% 91% 75% 63%
– 0.898 0.366 0.784 0.518 0.347
38 82 141 251 410 721
Total
1643
3.I have divided the data at 1575 because this is (roughly) the point of a major dislocation in Ellegård’s data (see graph p 163; Kroch 1989a) and again at 1465 since a substantial proportion of the earlier data is western, and it is not clear that the later data provide a simple continuation of it. I have re-dated some of Ellegård’s texts in the light of more recent scholarship. There is no coefficient for (a) in Table 2 because this is a knock-out factor which was necessarily omitted from the regression.
What drove DO? 235
Table 3.Incidence of DO in Yes-no and Adverbial Wh-questions 1400–1465.
a b c d e f
subject
object
DO%
ProbCoeff
n
NP NP NP pers prn pers prn pers prn
pers prn NP intrans pers prn NP intrans
50% 43% 7% 15% 9% 1%
0.972 0.964 0.716 0.862 0.779 0.226
2 7 15 20 55 122
Total
221
transitive questions supply the essential dynamic, with the other categories being carried along by virtue of the fact that a single grammatical choice is involved. Here I adopt the insight of Kroch’s Constant Rate Effect: relative levels of use may vary in different contexts, but where they involve a single underlying choice, the grammar will progress as a single entity. Hence change in negative declaratives could in effect be a response to the development of DO in questions. The mechanism which leads to a higher rate of DO in the more strongly motivating context needs to be spelled out. But whether relative levels are learned as such, or are a consequence of some internal feedback mechanism in producers, the suggestion that it is appropriate for contexts with the highest rates of differential loss to have the highest rates of DO is hardly implausible, although (as noted above) it is not a necessary assumption.
3. Could change depend on adult use? Evidence for communal change If individuals tend to accommodate to their interlocutors by altering their level of use of DO across time, the resulting pattern of use at a single time will be that of Labov’s “communal change” (1994: 83ff). Labov makes a broad distinction between two types of change. One he calls “generational”. Successive generations acquire different values and maintain them, so that at any particular point in time there will be differences between the usage of different generations, with older people maintaining an earlier level of usage. That is, there is an “apparent time” distribution in the community. He suggests that this is typical of phonological and morphological change. In the other type of change, the whole community moves forwards together. At any particular point in time there will not be an “apparent time” effect between the usage of different generations. Labov suggests that syntactic change is typically communal. In addition to this there
236 Anthony Warner
may be socially maintained patterns of age-grading which recur across generations, so the presence of differences between generations is in fact consistent with communal change, within a more complex situation (Labov 1994: 94ff). One question we can put to Ellegård’s data is whether or not the community moved forwards together, or whether the data show the “apparent time” profile typical of generational change. The answer to this question can be approached in several ways. First I can look at authors who produced texts at more than one time, to see whether and how their usage changed. There are difficulties here though. First, there are not many instances of authors recurring in Ellegård’s database. Second, they do not always write the same kind of text. Third, numbers are often low. Finally, there is also considerable variation between texts written by the same author within a short period (Deloney 33%, 54%; Otway 94%, 74%). For the fifteenth and sixteenth centuries, Ellegård’s evidence is difficult to interpret. Caxton shows a clear increase; Fisher and Gardner show declines; Bourchier with three texts shows both decline and level development. In the seventeenth century, however, Jonson has plentiful evidence, and he shows a modest increase over time. The difference is not statistically significant, but it shows the same rate of increase as holds for questions generally in the seventeenth century, which is suggestive. It is also clear from Stein’s figures that Shakespeare increased his usage of DO in questions with subject you across time (1990: 215–217). Stein gives figures for thirty-six plays. If we compare the totals for the earlier eighteen and the later eighteen we see a statistically significant increase (measured on chi-square tests) in both wh- and yes-no questions. As possible evidence for communal change this is, however, pretty unsatisfactory. But given my database there are three other ways of approaching this problem. In each case only texts with an author whose date of birth is known are considered, and each text is coded for the author’s age at the date of the text. We can then ask whether the age of authors is a significant variable when considered alongside time (i.e. the overall rate of change based on the dates of texts). If we have communal change, it will not be, since there will be no overall distinction between older and younger authors: the model will treat them as following the same s-curve. But this will not be the case with generational change. Here, there will be an overall distinction between older and younger authors: the model will treat younger authors as following a more advanced s-curve. The coefficient for age will be negative, and in the (idealized) case where individuals show no change after acquisition, it will simply be the negative of the value of the overall rate of change across time, so that
What drove DO? 237
coefficients for time and for age sum to zero. So I have tested this for yes-no and wh-questions both before and after 1575 within a regression programme (DataDesk, Velleman 1995) which allows time and age to be treated as continuous variables (as GoldVarb does not). Texts whose authors had no known date of birth were omitted, as noted above. The finding is that for neither period was there any significance whatever for age as a factor.4 Thus there is no evidence here for generational change, but the evidence is consistent with communal change, under which individuals change their usage as they get older. The second approach investigates the profile of the change. Instead of entering age as a continuous variable, as in the analysis above, it was divided up into five-year blocks, and these were analysed as a single discrete factor group. This means that the regression calculates a probabilistic coefficient for each five-year period of the lifespan of the authors. If this appears in a regression with time as a factor then we should be able to see if there is any patterning in the coefficients. The coefficients for each period are given in Tables 4 and 5. Table 4.Coefficients for Age of Authors 1465–1575. Age of author
Probabilistic Coefficient
n
24–30 31–35 36–40 41–45 46–50 51–55 56–60 61–76
−0.1578 −0.2385 −0.0890 −0.3065 −1.246 −1.192 −0.5355 −0.2752
119 150 21 40 32 209 152 104
Total
827
Major authors included
Tindale
Malory Palsgrave
Yes-no and wh-adverbial questions. Age in five year periods except for the last three: one of these is a knockout, one has small n. Other factors in run: date of text; affirmative vs negative; transitivity and type of object; question type.
4.In regressions for 1465–1575, with a range of other significant factors, age had p > 0.75 (χ2 = 0.022, 1 df); with only date of composition and constant, p > 0.99 (χ2 = 0.00006, 1 df). In 1575–1710, with a range of other significant factors, age had p > 0.90 (χ2 = 0.0023, 1 df); with only date of composition and constant, p > 0.25 (χ2 = 1.07, 1 df. but with a positive coefficient, where generational change predicts a negative one).
238 Anthony Warner
Table 5.Coefficients for Age of Authors 1575–1710. Age of author
Probabilistic Coefficient
n
Major authors included
21–25 26–30 31–35 36–40 41–45 46–50 51–55 56–60
−0.4248 −0.2846 −0.1337 −0.3601 −0.0102 −0.4409 −0.5367 −0.2809
92 269 397 113 256 92 39 106
Otway Jonson Lyly, Jonson, Dekker
Total
1564
Bunyan, Swift
Queen Elizabeth
Yes-no and wh-adverbial questions. Age in five year periods. Other factors as in Table 4.
For 1465–1575 there is no relevant pattern. If we had to deal with generational change we would expect a steady decline in values with rising age. But what we see is a random series of increases and decreases in value. This bears out the suggestion that this does not show generational change. In Table 5 we see the same absence of pattern in the coefficients, and can conclude that there is no support for generational change in 1575–1710. Finally, it is possible to isolate the age distribution of DO by adopting a different strategy which shows us more of the relationship between individual texts. Since the rate of change across time for yes-no and wh-adverbial questions in different periods (1465–1575; 1575–1710) can be calculated, we can estimate what the corresponding proportion of DO would have been for any text had that text been written at some other date in the period, using the rate of change in the appropriate formula. The point of the conversion is to mimic a situation in which the texts of each period were written within a single year, to make it easier to see interrelationships (such as an “apparent time” distribution). Here I calculated for each text first what the percentage value of DO was in yes-no and wh-adverbial questions. Then I used the general coefficient of change for the period in these question types to work out the corresponding proportion for the target date. This was 1525 in the first period; 1625 in the second. So (for example) Malory has 5% DO in 1468. When I apply the relevant rate of increase (3.88 logit units per century for 57 years), I estimate that this
What drove DO? 239
corresponds to a value of 32% in 1525.5 In effect this says: if Malory had been writing in 1525, he would (other considerations being equal) have used 32% DO. Similarly, Bourchier, translating Guevara’s Golden Book in 1535, uses 54% DO, but this corresponds to a figure of 44% DO a decade earlier in 1525. This conversion is based on the assumption that change is broadly stable within each period (as seems to have been the case), and that any “apparent time” effects are also broadly stable. The results of plotting authors’ ages against the predicted rate of DO is given in Figures 1 and 2. The distributions look pretty random. Certainly the overall decline which would indicate generational change is absent (a point confirmed by other indicators: correlation and linear regression).6 The reasonable conclusion is that there is no order-preserving relationship between age and incidence of DO, and, in particular, there is no sign of the type of “apparent time” distribution we would expect in generational change. These three ways of looking at the data have provided no evidence for generational change, but the evidence is throughout consistent with communal change, under which individuals change their usage in step as they get older.7 This is in turn consistent with the presence of accommodation between language users. And the presence of accommodation is, in its turn, consistent with the proposal that the differential difficulties of parsing questions with DO and with the simple finite form of the verb is responsible for the progress of the change. These points are distinct; there is no simple passage from the first to the last without assumptions which may prove inappropriate in the future. But the straightforward question: “Is Ellegård’s question data consistent with an interpretation in which differential parsing loss between accommodating adults was what drove change forwards?” can be answered with a simple “yes”. This particular story seems a coherent and plausible one. Moreover, if change was essentially located at acquisition we would expect to see some evidence of lag across age groups, even if accommodation across the generations blurred the sharp
5.The formula for 1465–1575 is: (exp(ln((do/total) / (1 − do/total)) − (date − 1525) × rate) / (1 + exp(ln((do/total) / (1 − do / total)) − (date − 1525) × rate))) 6.For 1464–1575, a Spearman Rank Correlation (which tests for a monotone relationship) yields a coefficient of +0.082 (where 0 is random; +1.0 or -1.0 fully systematic). For 1575–1710, the Spearman Rank Correlation is +0.267. Linear regressions for each period have small positive coefficients which are not significant; generational change should show up as a negative coefficient. 7.There are however some possible signs of age-grading, which need further investigation.
240 Anthony Warner
0.8
0.6 D O
0.4
0.2
25.0
37.5
50.0
62.5
Age
Figure 1.Estimated Proportion of DO for 1525 in Yes-no and Adverbial Wh-questions versus Age of Author.
apparent-time distribution typical of phonological change. Finally, the potential relevance of parsing is underlined by the facts (noted above) that the categories involved (transitive yes-no and wh-questions) are the categories with the highest rates of DO, and that the highest of these categories goes to completion early. This interpretation also, of course, implies the possible relevance of other functional factors. One alternative within the literature is Stein’s claim that phonotactics, in particular the potential combination of a verb which had a consonant final stem with -st, provided a “trigger or motivation” so that the change was “initiated by a phonotactically motivated syntactic bypass function of do” (1986: 129), that is, DO is used to avoid a final consonant group on the main verb. If functional considerations are potentially relevant, as I have argued, then this may indeed be among the motivations of DO’s progress. But it is notable that in Ellegård’s data the incidence of DO with thou in questions is not in advance of usage with you where such considerations do not hold. This sheds doubt on the idea that this was an initiating motivation. But a range of factors may have been involved, of which I would argue that the parsing differential just discussed was the most crucial.
What drove DO? 241
0.875
0.750
D O
0.625
0.500
30.0
37.5
45.0
52.5
Age Figure 2.Estimated Proportion of DO for 1625 in Yes-no and Adverbial Wh-questions versus Age of Author.
4. Conclusions These conclusions hold for yes-no and wh-adverbial questions in Ellegård’s database. The first two are factual, the third is speculative. 1. There is a stable ordering of the incidence of DO across particular subtypes of question throughout the period 1400 to 1710. 2. There is no evidence for generational change in these types of question; instead the evidence points to communal change in which the community moves together as a whole. 3. Since it has been suggested that the ordering of 1 is motivated by the relative difficulty of processing, the facts of 1 and 2 allow the following interpretation: the continued increase of periphrastic DO in competition with the simple finite verb 1400 to 1710 is (at least partly) motivated by processing considerations in transitive questions. One possibility is a mechanism whereby hearers accommodate their usage to their (subliminal) misperception of the level of DO used by speakers.
242 Anthony Warner
References Denison, David. 1993. English Historical Syntax: Verbal Constructions. London & New York: Longman. Ellegård, A. 1953. The Auxiliary ‘Do’: The Establishment and Regulation of its Use in English. Stockholm: Almqvist & Wiksell. Kroch, A. 1989a. “Reflexes of grammar in patterns of language change”. Language Variation and Change 1.199–244. ———. 1989b. “Function and grammar in the history of English: Periphrastic Do”. Language Change and Variation ed. by R. W. Fasold & D. Schriffin. 132–172. Amsterdam & Philadelphia: John Benjamins. Kroch, A., S. Pintzuk, J. Myhill. 1982. “Understanding DO”. Papers from the 18th Regional Meeting of the Chicago Linguistics Society ed. by K. Tuite et al. 282–294. Chicago: Chicago Linguistics Society. Labov, W. 1994. Principles of Linguistic Change: Internal Factors. Oxford: Blackwell. Nurmi, Arja. 1999. A Social History of Periphrastic DO. Mémoires de la Société Néophilologique de Helsinki LVI. Pintzuk, Susan, George Tsoulas & Anthony Warner. 2001. “Syntactic Change: Theory and Method”. Diachronic Syntax: Models and Mechanisms ed. by Susan Pintzuk, George Tsoulas & Anthony Warner. 1–22. Oxford: Oxford University Press. Rissanen, M. 1999. “Syntax”. The Cambridge History of the English Language vol III 1476–1776 ed. by Roger Lass. 187–331. Cambridge: Cambridge University Press. Sankoff, David & David Rand. 1990. GoldVarb: Variable Rule Analysis for the Macintosh. http://www.crm.umontreal.ca/~sankoff/GoldVarb_Eng.html http://www.crm.umontreal.ca/~rand/GoldVarb_Eng.html Sprouse, Rex A. & Barbara Vance. 1999. “An Explanation for the Decline of Null Pronouns in Certain Germanic and Romance Languages”. Language Creation and Language Change: Creolization, Diachrony and Development ed. by Michel DeGraff. 257–283. Cambridge, Mass.: MIT Press. Stein, Dieter. 1986. “Syntactic Variation and Change: the Case of Do in Questions in Early Modern English”. Folia Linguistica Historica 7.121–149. ———. 1990. The Semantics of Syntactic Change: Aspects of the Evolution of ‘Do’ in English. Trends in Linguistics 47. Berlin, New York: Mouton de Gruyter. Trudgill, Peter. 1986. Dialects in Contact. Oxford: Blackwell. Velleman, Paul et al. 1995. DataDesk. (Software for Macintosh) Ithaca: Data Description.
The have-‘perfect’ in Old English Ilse Wischer University of Potsdam
1.
Introduction
The development of the have-participle construction has been extensively studied and described by numerous authors (cf. Visser 1963–73, Mitchell 1985, Brinton 1988, Denison 1993, etc.). Its status in Old English, however, is still a controversial topic. Because the participle occasionally carries nominal inflection agreeing with the object that it modifies, many scholars (Kisbye 1971, Berndt 1982, Mitchell 1985, etc.) consider it a precursor of the modern periphrastic verb construction, containing the lexical verb have and a participial adjective: (1) … oðrum þe hiora dæl getynedne hæbben,… others who their part enclosed.acc have …others who have their part enclosed,… (ox/2 sta law ine 106)1
However, these occurrences were rare. As Brunner (1965) points out, participles mostly lacked an ending when used predicatively. Other authors therefore acknowledge the verbal character of the construction and describe it as ‘perfect’ (Caro 1896, Brinton 1988, Strang 1970:311), ‘perfective aspect’ (Hoffmann 1934, Quirk & Wrenn 1993), or as a ‘resultative construction’ (Carey 1994, 1995). My studies focus on the characterization of have-participle constructions in Old English. I analysed all occurrences in the Old English part of the Helsinki Corpus to identify the semantic and syntactic contexts in which they are used and the grammatical functions that the periphrastic constructions may have fulfilled.
1.The examples are taken from the historical part of the Helsinki Corpus.
244 Ilse Wischer
2. Syntactic and semantic contexts If the periphrasis originated in a possessive construction, as is generally assumed, a syntactic reanalysis as demonstrated in (2a) and (2b) must have taken place at some time: (2) a.
Principal Clauses b. Subordinate Clauses have LexV + [np+partMod]Acc [np+partMod]Acc + have LexV Reanalysis Reanalysis have Aux + np + partLexV np + [partLexV + have Aux] Modern English: [have Aux+partLexV] (+np)
It is important to note that — before the reanalysis — it was have (as the lexical verb) that assigned case to its object. After the reanalysis, the participle took over the function of case assignment, and so the object may occur in the genitive, dative, as prepositional object, clausal object, or it may be missing altogether, depending on the valency features of the participle. And this was already the case in Old English as Table 1 shows. Only three-quarters of the examples have an accusative object, and the vast majority of these examples clearly show that the carrier of semantic valency is the participle, and not have. So we can see that the syntactic reanalysis must have taken place as early as the beginning of the Old English period. Later the syntactic reanalysis was followed by a transposition of the constituents, so that in Modern English we have the sequence have + part (+obj). As Mitchell (1985: 282) states, in OE we can find seven different patterns according to the sequence of constituents. These findings are supported by my studies, as shown in Table 2. Table 2 also shows that more than 50% of the constructions still have the typical OE word order, but 23% (patterns (3) and (6)) already have the Modern English constituent structure. This is another indicator that the grammaticalization of the have+participle construction must have been fairly well advanced in OE. But what about the inflected participle constructions that are always provided as evidence against an advanced stage of grammaticalization? First, it must be emphasized that the uninflected forms clearly predominate (cf. Mitchell 1985: 284, Hoffmann 1934: 52, Traugott 1992: 190, Harrison 1887: 134–5, Caro 1896: 406, Sprockel 1973: 206–7). In my corpus, out of 232 occurrences of the construction, only 23 participles were inflected, i.e. about 10%. Secondly, the occurrence of a case-inflected participle need not be evidence
The have-‘perfect’ in Old English 245
Table 1.Types of ‘Objects’ in have-participle constructions in the OE part of the Helsinki Corpus (232 o.). Acc-Obj
O-Clause
no Obj
Gen-Obj
Prep Obj
Subj Comp
180 / 76%
18 / 8%
12 / 5%
11 / 5%
10 / 4%
1 / 0%
Table 2.Patterns according to the sequence of constituents in the OE part of the Helsinki Corpus (232 o.). (1) (2) (3) (4) (5) (6) (7)
obj + part + have (common in OE subordinate clauses) have + obj + part (common in OE principal clauses) have + part + obj obj + have + part part + have + obj have + part part + have
75 / 32% 54 / 23% 46 / 20% 35 / 15% 10 / 4% 8 / 3% 4 / 2%
of a less grammaticalized stage of the construction. It may just be a meaningless residue of a former state. How could one otherwise explain the fact that inflection is used so variably in OE? Sometimes the same object is accompanied by two participles, one of them inflected, the other not (cf. also Mitchell 1985: 285f., Brinton 1988): (3) …hafað butu gedon, ealle gemanode and eac getogen,… …has both done all.pl warned.pl and also led …he has done both, he has warned them and also led them,… (o2/3 xx xx mbo 167)
Sometimes the accusative plural masculine inflection is analogically extended to instances where, for example, a neuter plural like hus would typically trigger the endingless participle berypt: (4) & we habbað Godes hus inne & ute clæne & we have God’s house.acc-pl-neut inside & outside fully berypte. spoiled.acc-pl-masc & have fully spoiled God’s houses inside and outside (o3 ir hom wulf 20 268)
Sometimes an adjectival case marker occurs, where a stative, possessive interpretation, based on have as lexical verb, would contradict the actual dynamic reading inherent in the participle as lexical verb:
246 Ilse Wischer
(5) Loca nu; þin agen geleafa þe hæfþ gehæledne. look now your own belief you.acc-sg-masc has healed.acc-sg-masc Look now, your own belief has healed you. (o2/3 ir hom blick2 15)
Otherwise, sometimes clear adjectival forms that have no inflection are used: (6) …gyf he ænigne gylt ungebet hæfð, … …if he any.acc guilt unatoned-for has “… if he has any guilt unatoned for, …” (o3/4 ir relt lsigew 56)
From these examples we can conclude that inflection is no indicator of a less grammaticalized stage of the have-participle construction in OE. In recent studies on grammaticalization (Heine 2002, Diewald 2002), particular contexts have been characterized that give rise to new grammatical meanings, and that may help to determine the stage of grammaticalization of a particular item. In what follows I will show how these theories can be applied to the grammaticalization scenario of the have-participle construction in Old English. Diewald (2002) distinguishes three types of syntactic context in the grammaticalization of specific German modal constructions: 1. the untypical context 2. the critical context 3. the isolating context. According to her scenario, a lexical unit is first used in contexts that are untypical. Here the lexical unit can be interpreted in a new way, based on a conversational implicature. This unusual interpretation allows its use in what she calls the critical context, where multiple structural and semantic ambiguities occur, which allow several alternative interpretations. This is also the context in which grammaticalization takes place. As a consequence, the new grammatical meaning, when it is established, is isolated from the old lexical meaning, and both constructions occur in different, so-called isolating, contexts. This is also what Hopper (1991) calls layering. Diewald’s theory works very well in the case of the grammaticalization of the German modal constructions, because the three different contexts can be observed in Old and Middle High German. With respect to the English haveparticiple construction, however, it is not very useful. It is difficult to imagine what could have been an untypical context of the lexical have-construction, since, according to the OED, from the earliest attestations in the Germanic
The have-‘perfect’ in Old English 247
languages, have shows a broad range of meanings. It was not restricted to the meaning possess, and so could occur in a wide range of structural and semantic contexts. Maybe an untypical context was that with a modifying participle lacking the case marker, but this must have been in pre-OE times. So, are we in the phase of the critical context in Old English? The problem is that only very few examples, if any at all, allow alternative interpretations: (7) þu, fæder Agustinus, hie hæfst on þinum bocum sweotole gesæd; you father Agustinus them have in your books clearly said you, father Agustinus, possess them (words) in your books, clearly said; / you, father Agustinus, have clearly said them in your books; (o2 nn hist oros 102)
The critical context presupposes that the referent of the subject may fulfil a double role, namely that of the carrier of possession and that of the agent who affects the possessed object (as in (7): You have these words in your books. You have clearly said them.). However, as I mentioned before, real possessive interpretations are definitely the minority in Old English have-participle constructions. Most examples contain a dynamic sense: (8) Gif he ær hæfþ attor gedruncen… if he before has poison drunc If he has drunk poison before… (o2 is handm laeceb 336)
If he has drunk it, he cannot possess it any more. (9) Gif he hwæt yfla gedon hæbbe ðær he ær wæs,… if he what evil done have there he before was If he had done something evil where he was before,… (o2 sta law alflaw 70)
In (9), the present possessive interpretation is in contradiction not only with the semantics of the object (‘to possess some evil done’), but also with the subordinate adverbial clause, which describes a past circumstance. Such uses can only have been the result of analogical extension, after the have-construction had been established in other contexts. The have-participle construction seems to be in the beginning of the isolating context phase in Old English. A particular word order develops to distinguish the periphrastic construction from the lexical (possessive) verb construction. But it does not become fixed before the end of the Early Modern period. And in Modern English there is still some dialectal variation with regard
248 Ilse Wischer
to the treatment of stative have as lexical verb in relation to auxiliary have (cf. Trudgill, Nevalainen & Wischer 2002). Heine’s approach differs from Diewald’s insofar as he concentrates more on the lexical contexts leading to new grammatical meanings. He also applies his scenario to a number of grammaticalization phenomena in various genetically unrelated languages. Heine distinguishes four different contexts: 1. 2. 3. 4.
Initial Stage Bridging Context Switch Context Conventionalization.
At the Initial Stage we find the source meaning of the item. With respect to English have, it would be ·possessÒ. The Bridging Context gives rise to an inference to the effect that another meaning offers a more plausible interpretation. In our case it would be a resultative or perfective interpretation. In the Switch Context, the construction is incompatible with the source meaning. And if a construction is conventionalized, the target meaning no longer needs to be supported by the context that gave rise to it. It may be used in new contexts. According to the scenario developed by Heine, English have may have acquired its new grammatical meaning through the following steps: 1. 2. 3. 4.
possessive resultative/perfective foregrounded possessive backgrounded resultative/perfective meaning only.
My data suggest that the have-participle construction had reached stage III, the Switch Context, in Old English, since most constructions are incompatible with the source meaning, ·possessÒ. Abraham (2002) lists a number of conditions that must be fulfilled for the auxiliation of have in the perfect periphrasis: 1. 2. 3. 4. 5. 6.
complementation also with imperfective verbs functional distinction between simple past and perfect developing/consolidating the sequence of tenses have loses its semantic roles possession no longer expressed by have conversion of the semantic role selection.
The first condition is important if the have-participle construction grammaticalizes as a tense marker. If it becomes an aspect marker, complementation with
The have-‘perfect’ in Old English 249
all semantic types of verbs is not required. Compare the progressive aspect in Modern English, which is not applicable to meanings that cannot be divided into phases. It is often claimed (Hoffmann 1934, Carey 1994, 1995) that, in Old English, have only combines with participles of telic verbs. My studies support the claim that telic verbs predominate as participles, but they are not used exclusively, as is shown in (10): (10) a.
…æfterðæmðe hie gesyngod habbað. …after-that they sinned have …after they have sinned. (o2 ir relt cp 405) b. …& hira mægeðhad habbað gehealdenne; …& their virginity have kept …& have kept their virginity; (o2 ir relt cp 409) c. & ðonne hi hi gesewene hæbben, & then they them seen have & then they have seen them (o2 ir relt cp 413)
So the first condition has been fulfilled: have has extended its complementation patterns to other types of verbs. The second and third conditions concern the have-construction as a tense marker. If it, however, turns into an aspect marker, the two conditions become irrelevant. Condition (4) and (6) belong together and are closely connected to (5), which refers to semantic bleaching. If have is no longer understood in the sense of ·possessÒ, it must lose its semantic participants (roles). The subject and object are no longer analysed as participant roles in a possessive state, but as participants in the event expressed by the participle, as in (11) and (12): (11) & þæt we habbað ealle ægþer ge mid worde ge mid wedde gefæstnod, & that we have all either with word or with pledge confirmed, þæt we under anum cynedome ænne Cristendom healdan willað. that we under one royal-law one Christianity hold will & we have all confirmed either with words or with pledges that we will hold one Christianity under one royal law. (o3 sra law law11c 236) (12) mid þam godum, þe he us geunnen hæfð with the goods that he us granted has with the goods that he has granted to us *with the goods that he possesses granted to us (o3 ir rule benedoe 1)
250 Ilse Wischer
In (11), the complement clause is an argument of the participle, ‘gefæstnod’, and not an object of the verb ‘habbað’ in the meaning of ·possessÒ. In (12), ‘he’ is the agent of ‘grant’, ‘us’ the recipient, and the relative particle ‘þe’ the affected participant. ‘hæfð’ is auxiliary. All arguments put forward so far support the idea that have had been auxiliarised in Old English and the periphrastic construction had been grammaticalized to a considerable extent. But what was its grammatical function? 3. Grammatical function It is difficult to say what the grammatical function of the Modern English perfect construction is. Some authors list it under tense, others under aspect. I consider it a secondary tense in the sense of Jespersen (1909–49) or Schopf (1984), or a form of correlation, as in Graustein et al. (1980), who describe it as anteriority. Bybee et al. (1994: 54) define anteriors as grams that “signal that the situation occurs prior to reference time and is relevant to the situation at reference time”. Using a crosslinguistic study, they have come to the conclusion that anteriors typically develop out of resultatives, if they are formed with stative auxiliaries. Resultatives are constructions that signal that a state exists as a result of a past action, which is not a temporal, but an aspectual dimension. They are compatible with ‘still’, as for instance in he is gone — he is still gone; and they only combine with telic verbs. Resultatives can be compared to stative passives, as in the door is closed. So, according to Bybee et al. (1994), resultatives can develop into anteriors, and these in turn can further develop into perfectives or pasts. Perfectives are defined as constructions that signal that something is completed within a particular time. The situation is viewed as bounded temporally, which is an aspectual dimension. Pasts are temporal grams that signal that something happened before the moment of speaking. If we now assume a chain of developments as in Figure 1, we might conclude that the Old English have-participle construction functioned as a resultative or already as an anterior. The first view is taken by Traugott (1972: 94f.) on the basis of constructions such as those in (1), here repeated as (13): Inference from results Æ Indirect evidence ‘be/have’ Æ Resultative Anterior Æ Perfective/simple past
Figure 1.Simplified from Bybee et al. (1994: 105).
The have-‘perfect’ in Old English
(13) … oðrum þe hiora dæl getynedne hæbben, others who have their part in an enclosed state, i.e., who have enclosed it before, (ox/2 sta law ine 106)
where the construction was used in a possessive context. Similarly, Hoffmann (1934) thinks that it is only in Early Middle English that we can speak of a true perfect and pluperfect; he considers the perfect a temporal gradation, and as such he does not consider it part of Old English grammar. He cites Ælfric, who understood the Latin tenses very well, but did not give English equivalents with the periphrastic constructions. According to Hoffmann, the ‘Perfektum’ described the completion of an action in the present. Carey (1994; 1995) also denies the existence of a perfect in Old English and characterizes the have-participle construction “as a type of resultative construction in which the subject is in a relation of possession/control with the final state of a completed process”. She bases her argument on the fact that resultatives are incompatible with stative verbs, and she claims that she could not find any stative verbs in combination with have in her data. (However, in her texts from the early OE period, almost 40% of the participles combining with have are mental state verbs!). Another argument is that the resultative constructions combine with temporal adverbs that refer to a final state, like nu (‘now’), whereas perfects, i.e. anteriors, would combine with adverbials which refer to an anterior period, like siþþan (‘since’). Caro (1896: 402), on the other hand, states that “Die in den ältesten Denkmälern vorkommenden Fälle, gleichviel, ob mit flektiertem oder unflektiertem Partizipium, sind aber schon vollentwickelte einheitliche Tempora”.2 Brinton similarly argues (1988: 102): “The earliest examples of have + past participle constructions in Germanic thus appear to be semantically well-formed as perfects”. Her argument is supported by the following observations: 1. 2. 3. 4.
the vast majority of the participles appears without case endings constructions with possessive meaning are at best rare have can be combined with participles formed from intransitive verbs perfect constructions with actional meaning appear from the beginning.
It is interesting that she concludes that no clear chronology of the development from stative, resultative meaning to actional, perfect meaning can be established.
2.The occurrences in the oldest documents, whether with inflected or uninflected participles, are fully developed uniform tenses.
251
252 Ilse Wischer
She claims that “Finally, the last step in the postulated semantic development of the perfect, from resultant state to antecedent completed action, has never been taken in English” (102). The perfect in Modern English still has a dual nature, combining past action and present state. Indeed, the Modern English anterior present has two different uses: 1. a continuative/inclusive use 2. a resultative/exclusive use The first is used with durative or atelic verbs and describes a continuation from a point in the past, leading up to, and often including, or even going beyond, the moment of speaking (he has lived in Glasgow for 20 years); the second is used with telic verbs and describes an event that took place in the past, but has some current relevance. It is this second use, the resultative use of anteriority, that is found with most Old English examples of have + past participle (cf. also Mitchell 1985: 297): (14) Eala, cniht, þy þu eart gesælig þæt þu hit swa ongiten hæfst; oh boy therefore you are blessed that you it so understood have Oh, boy, you are blessed for having it understood that way; (o2 xx philo boethal 78)
Even most of the constructions with inflected participles cannot be interpreted as pure resultatives, since they can hardly be analysed as stative possessive constructions, as in (15): (15) Ac ic ongite ðæt ic þe hæbbe aðretne nu mid þy langan spelle,… but I understand that I you have wearied now with the long talk But I understand that I have wearied you now with the long talk,… (o2 xx philo boethal 135)
However, a few examples can already be found in Old English that are used with atelic verbs that express continuative anteriority, as in (16): (16) … & hira mægeðhad habbað gehealdenne; & their virginity have kept … & have kept their virginity; (o2 ir relt cp 409)
Apart from the function of marking anteriority, we can also find the have-participle construction expressing a certain aspectual meaning. I would call it perfective, as opposed to imperfective. It becomes especially obvious in clauses that refer to futurity or that contain some modal features, as in for instance (17) and (18):
The have-‘perfect’ in Old English 253
(17) Hafa þe ær geworht clam of beor … & of … have.imp-sg you.dat-sg previously made paste of beer & of Previously, have yourself made a paste of beer … & of … (o2 is handm laeceb 330) (18) ælflæd gæswytelaþ on þis gewrite hu hæo wile habban gefadad hiræ Ælflæd reveals in this writing how she will have disposed her æhta for Gode & for worldæ. properties for God & for world. Ælflæd reveals in this writing how she wants to have her properties disposed for God and for the world. (o3 xx doc whit15 38)
In (17) the have-construction emphasizes the completion of an action in a sequence of events. Example (18) is also perfective, but rather in a causative sense, a kind of precursor to constructions like She has her hair cut. A last function of the have-participle construction in Old English is a merely temporal function, namely that of referring to pastness. It occurs with temporal adverbs like ær (‘before’) or þonne (‘then’), as in (19), or denotes a contrast to present and future events, as in (20). (19) Gif he hwæt yfla gedon hæbbe ðær he ær wæs,… if he what evil done have there he before was If he had done something evil where he was before,… (o2 sta law alflaw 70) (20) þa ðing ealle þe he us gedon hæfð & dæghwamlice deð & gyt don wile, the things all that he us done has & daily does & still do will all the things that he did for us & that he does every day & that he will do in future, (o3 ir hom wulf13 228)
4. Conclusion I hope to have shown that there is sufficient evidence to state that have had acquired the status of an auxiliary as early as in Old English (cf. also Traugott 1992: 191). In combination with the second participle of a verb it was in an advanced state of grammaticalization as a marker of tense and aspect distinctions. As was pointed out earlier by Mitchell (1985: 291), the periphrasis had no single fixed function. Furthermore, our data confirm Brinton’s conclusion that no clear chronology of the development from stative, resultative meaning to actional, perfect meaning can be established. Its use as a marker of resultative
254 Ilse Wischer
anteriority predominates, but perfectives, as well as mere past markers, can also be attested. What becomes obvious is that grammaticalization is not driven by functional needs, but rather by syntactic ambiguities — by particular syntactic contexts that open up specific grammaticalization paths. A comparative study of the situation in Old High German might prove a useful way of shedding more light on the development of perfect markers in the Germanic languages, and maybe also on their later divergent behaviour in German and English.
References Abraham, Werner. 2002. “The European demise of the simple past and the emergence of the periphrastic perfect: areal diffusion or natural, autonomous evolution under parsing facilitation?”. Paper presented at the 35th Annual Meeting of the Societas Linguistica Europaea, Potsdam, July 2002. Berndt, Rolf. 1982. A History of the English Language. Leipzig: Enzyklopädie. Brinton, Laurel J. 1988. The Development of English Aspectual Systems: Aspectualizers and Post-verbal Particles. (= Cambridge Studies in Linguistics, 49) Cambridge: Cambridge University Press. Brunner, Karl. 1965. Altenglische Grammatik. Tübingen: Niemeyer. Bybee, Joan L., R. Perkins & W. Pagliuca. 1994. The Evolution of Grammar: Tense, aspect and mood in the languages of the world. Chicago: University of Chicago Press. Caro, George. 1896. “Zur Lehre vom altenglischen Perfectum”. Anglia XVIII.386–449. Carey, Kathleen. 1994. “The grammaticalization of the Perfect in Old English: An Account Based on Pragmatics and Metaphor”. Perspectives on Grammaticalization (= Current Issues in Linguistic Theory, 109) ed. by William Pagliuca, 103–117. Amsterdam & Philadelphia: John Benjamins. ———. 1995. “Subjectification and the Development of the English Perfect”. Subjectivity and Subjectivisation: Linguistic Perspectives ed. by Dieter Stein & Susan Wright, 83–102. Cambridge: Cambridge University Press. Denison, David. 1993. English Historical Syntax: Verbal Construction. London: Longman. Diewald, Gabriele. 2002. “A model for relevant types of contexts in grammaticalization”. New Reflections on Grammaticalization (= Typological Studies in Language, 49) ed. by Ilse Wischer & Gabriele Diewald, 103–120. Amsterdam & Philadelphia: John Benjamins. Graustein, Gottfried et al. 1980. English Grammar — a University Handbook. Leipzig: Enzyklopädie. Harrison, J. A. 1887. “The Anglo-Saxon Perfect Participle with ‘habban’”. Modern Language Notes 2.134–135. Heine, Bernd. 2002. “On the role of context in grammaticalization”. New Reflections on Grammaticalization (= Typological Studies in Language, 49) ed. by Ilse Wischer & Gabriele Diewald, 83–101. Amsterdam & Philadelphia: John Benjamins. Hoffmann, G. 1934. Die Entwicklung des umschriebenen Perfektums im Altenglischen und Frühmittelenglischen. Dissertation, Breslau.
The have-‘perfect’ in Old English 255
Hopper, Paul J. 1991. “On some Principles of Grammaticalization”. Approaches to Grammaticalization ed. by Elizabeth Closs Traugott & Bernd Heine, vol. I, 17–35. Amsterdam & Philadelphia: John Benjamins. Jespersen, Otto. 1909–49. A Modern English Grammar on Historical Principles, vol. IV. Heidelberg: Carl Winter. Kisbye, Torben. 1971. An Historical Outline of English Syntax. Tryk: Akademisk Boghandel. Lussky, G. F. 1922. “The Verb Forms Circumscribed with the Perfect Participle in the Beowulf”. Journal of English and Germanic Philology 21.32–69. Mitchell, Bruce. 1985. Old English Syntax, vol. I. Oxford: Clarendon Press. Quirk, Randolph & C. L. Wrenn. 1993. An Old English Grammar. London: Routledge. Schopf, A. 1984. Das Verzeitungssystem des Englischen. Tübingen: Niemeyer. Sprockel, C. 1973. The Language of the Parker Chronicle II: Word-formation and syntax. Dordrecht: Martinus Nijhoff. Strang, Barbara. 1970. A History of English. London & New York: Routledge. Traugott, Elizabeth Closs. 1972. The History of English Syntax: a transformational approach to the history of English sentence structure. New York: Holt, Rinehart & Winston. ———. 1992. “Syntax”. The Cambridge History of the English Language I: The beginnings to 1066 ed. by Richard Hogg, 168–289. Cambridge: Cambridge University Press. Trudgill, Peter, Terttu Nevalainen & Ilse Wischer. 2002. “Dynamic have in North American and British Isles English”. English Language and Linguistics 6:1.1–15. Visser, F. Th. 1963–73. An Historical Syntax of the English Language, 3 parts in 5 volumes. Leiden: E. J. Brill.
Subject index
A AAE see African American English Abney-style DP analysis 20 accommodation 232, 233, 239 acquisition 229 acquisition of language see language acquisition actuation see innovation Adjective 20, 83, 85, 87, 88, 92 Adverbial subordinator 133–136, 141–142, 148–149 in Scots 134, 136, 145 of concession 141–142 African American English (AAE) 1, 3, 5, 8, 12 agreement 10, 66, 69, 70, 190 see also concord alliteration 196, 203–209, 211, 212 analogical extension 247 analogical levelling 180–181, 191 analogy 45, 50, 52, 177, 185, 191, 217 anaphora 65–70, 74–79 androcentrism 66, 79 anteriors 250–251 aphasia 39–40 apparent time 235–239 archaeology 103, 106 argument-blocking particles 95 auxiliary 189, 216, 218, 224, 248, 250, 253 see also modal auxiliation 248 B Beowulf 97, 99, 195–197, 199, 201, 203, 212
Binding Principles 45, 46 Binding Theory 44 biological evolution 34 biological tree 115–116 biology 57, 60, 110, 120–121 bleaching 249 blend constructions 216–219, 221, 222 Bloomfield, Leonard 110 BNC see British National Corpus bootstrapping 59, 115–116 borrowing 108, 111–114, 116, 119–121 brain-damaged patients 39, 41 British English 2, 3, 218, 219, 223 southern 2–3 British National Corpus (BNC) 93, 219–220 BROWN see Brown Corpus Brown Corpus of Standard American English (BROWN) 174, 216, 218–225 C Cambridge Grammar (Huddleston & Pullum) 20, 23, 24 capacity for learning 41 Caribbean 3, 5 categorization 26, 134 Celtic 109, 111, 116, 173 Chomsky, Noam 36–38 classification of language 103–105, 113 classificatory linguistics 107 cognitive abilities 35, 36 see also linguistics communal change 235–237, 239, 241
258 Subject index
communicative context 56 comparative approach 125, 129, 133 comparative method 105, 106, 108, 110 competence 36–38, 48, 51, 55, 57, 59 complex design 32–33 Computational Cladistics Project 108, 110, 112, 113 computational methods 103, 108, 113 concord 1–2, 66, 69 see also agreement connective 134, 147, 148 Constant Rate Effect 233, 235 constituent structure analysis 16 content words 20, 51, 55, 56 convergence 105, 129 Corpus of Early English Correspondence (CEEC) 3, 71 Corpus of Early Scottish Women’s Writings (CESWW) 130 Corpus of Scottish Correspondence (CSC) 130 correctness 66, 69, 153–156, 161, 165, 170 grammatical 66, 69, 159 creation science 32, 34 creolisation 105 D Daily Advertiser (DA) 154, 155, 163 decategorialization 92 degrees of relatedness 106 (de)grammaticalization 215 derivational morphology 89 derivational processes 84 design 32–35 determiner 20, 22, 24 dialect 1, 5, 19, 107, 177–178, 180–182, 184, 186, 189–191 continuum 120, 126 dialectology 104, 190, 191 Dictionary of the Older Scottish Tongue (DOST) 129, 189 diffusion 8, 15 discourse analysis 57 DOST see Dictionary of the Older Scottish Tongue drift 177, 182, 186, 191
semantic 68 Dutch 89, 94, 95, 100, 118, 119 E Early Modern English (ENE) 2–4, 8, 10, 65, 68, 71–72, 75, 79, 97–99, 136, 148, 186, 189, 217, 224, 229, 247 ease of processing 60 E-language 17, 36 see also performance Ellegård, A ix, 177, 187, 229–234, 236, 239–241 EME see Middle English ENE see Early Modern English epicene 65, 68, 78 see also pronoun errors 9, 10, 53, 59 event-modifiers 84, 96 Evolutionary Theory 32, 33 F family tree 110, 112, 113, 116 feminist approaches 66 fit-technique 126 FLOB (Freiburg-LOB Corpus) 216, 218–224 form-function divide 54, 55 French 114, 155, 157–166, 168–170, 172–174 Frisian 118, 119 FROWN (Freiburg-Brown Corpus) 218–225, 227 fuzziness 130–132, 134 G gender 65–67, 70, 73–74 generational change 236–239, 241 generic term 68 genetics 103, 106, 113, 121 genotype 35–38, 42, 51, 60 German 69, 85, 89, 93–95, 100, 111, 254 Old High 111, 246, 254 modals 224, 246 Germanic 108, 111–112, 114–116, 118–121 languages 94, 246, 251, 254 North 114, 119
Subject index 259
Goldberg, Elkhonon 26, 28, 39–42, 47 GoldVarb 6, 231, 237 Gothic 111 grammar books 66, 67, 153, 165, 171 change 31–33, 57, 58 cognitive 57 construction 26, 28 core 37, 39 cue-based 52 generative 46, 57 localized module 40 module 35, 40, 42, 59 synchronic 15, 19 transformational 37 grammatical device 51 grammaticalization 11, 27, 56, 60, 68, 69, 76, 79, 83, 131–132, 149, 215, 217–218, 223, 225, 244, 246, 248, 253–254 grammaticalized 34, 87, 89, 91, 93, 94, 245, 246, 250 H HCOS see Helsinki Corpus of Older Scots Heine, Berndt 217, 224, 246, 248 Helsinki Corpus (HC) x, 3, 69, 92, 97–99, 130, 243, 245 Helsinki Corpus of Older Scots (HCOS) 130–148 hemispheric specialization 41 hierarchization 126, 127, 129 historicization 127, 129, 130 Hittite 109 Hopper, Paul 44, 52, 58, 217, 246 I I-language 36 see also competence imitation 50, 52, 53 impersonal constructions 217 indefinite pronoun 65–69, 72–75, 76 Indo-European 108, 109, 115, 116, 118 infinitive complement 195–197, 204 inflected participle constructions 244
innateness 35–39, 42–46, 50, 54, 56, 59–60, 128 innovation 15, 19, 177, 185, 190, 191 intermediate forms 15, 25–27 Ireland 2, 3, 5, 7, 154 irregular verbs 179, 183, 186, 191 Italic 111 J Jespersen, Otto ix, 67, 181, 183, 250 K Klooster, Wim 37–38 Kortmann, Bernd 133–136, 148–149 Kroch, A. 229–235 Kroch, Myhill & Pintzuk 230–233 L Labov, William 26, 103, 179, 235, 236 language acquisition 16, 19, 34, 36, 37, 39, 52–54, 56–60, 229 language acquisition device 37 language change 15–16, 27, 31–34, 36, 44, 53, 56–60, 132, 219 contact-induced 105, 112–114, 119 language faculty 42 language learner 34, 53–54 Language Making Capacity 47, 50 language mixing 105 language-specific modules 41 see also grammar Lass, Roger 33–34, 59, 60, 72, 113 Latvian 109 LCS see Lexical Conceptual Structure learning process 41, 50–51 lexeme 65, 72–73, 109, 131, 215 Lexical Conceptual Structure (LCS) 84–91, 94–100 lexicon 40, 52, 55, 56, 84, 101, 185 Lieberman, Philip 39, 42 Lightfoot, David 19, 31–33, 35–37, 43–45, 49, 51–52, 57–58 Linguistic Atlas of Early Middle English 126 Linguistic Atlas of Older Scots 126
260 Subject index
linguistic blueprint 34–35 linguistic evolution 32, 34, 36 linguistic geography 113 linguistics cognitive 46, 51 formalist 43, 55, 56 generative 34, 37, 38, 48, 51, 60 loanword 111 LOB (Lancaster-Oslo/Bergen Corpus) 216, 218–220, 222–224 M manuscript sources 125, 134 meaning-lists 114–116 memory 42, 53, 185 mental frames 128 metalanguage 125, 127 Middle English (ME) 69, 97–98, 136, 186, 229 Early (EME) 67, 251 Late (LME) 68, 136 period 2, 8, 49, 181, 189 Milroy, James 127, 129 modal 195–197, 200–212, 216 see also semi-modal German 224, 246 marginal 215 Modern English 67, 89, 190, 244, 247, 249–252 see also Early Modern English, Present-day English Late 66 modularity 40 morpho-syntactic 68, 69 morphosyntax 27, 217, 225 Motherese 44 Murray, James 2, 182 N Neighbour Joining approach 115 network 111, 119–121 analysis 121 neural 51 social 19, 179 new synthesis 106 Newfoundland English 2, 5–8, 12
normativism 69 Northern Subject Rule 2–5, 11, 12 novel utterances 43, 49, 52 numerals 23 O object transfer 85, 86 objectification 127, 128 OE see Old English OED (Oxford English Dictionary) 24, 47, 84, 88, 89, 94–99, 246 Old English (OE) 22, 67–69, 94, 95, 98, 109, 111, 135–136, 183, 189, 195–199, 203, 204, 212, 217, 243, 254 prose syntax 195 Old Irish 109 Old Norse 111 Old Persian 109 operating principles 47, 48, 50–52, 59 optimality theory 57 parole 36, 127 parsing 51, 232–234, 239, 240 particle 56, 83–100 adverbial 18, 23 grammaticalized predicates 83, 89, 91, 93 negative 186–187, 189 particle-verbs 84–86, 88–89, 91, 94–95, 99, 100 see also prepositional verbs PDE see Present-day English perfect phylogeny 110–112 performance 16, 17, 36, 48, 51, 55, 59, 69 periphrastic do 187, 229, 241 phenotype 36, 51 philological ix, 32, 110, 112 computing 126, 130 phrasal verb 18, 27, 56, 98 see also particle verb, prepositional verb PHYLIP programs 115 pidginisation 105 polyfunctionality 131–132, 134, 135, 149 preposition 18, 23, 83, 87, 92, 134 prepositional phrase 56, 83, 87, 90, 195, 198
Subject index 261
prepositional verb 17–18, 27 see also phrasal verb, particle verb Present-day English (PDE) 21, 24, 45, 66–69, 75, 95, 133–136, 215–225 see also Modern English primitive 179–181, 191 status 181, 191 processing complexity 230, 232 processing constraints 10 pronoun 1–8, 11, 20, 23, 43–46, 65–69, 70–79, 180–182, 187–188, 230, 232, 234 indefinite (epicene) 65–69, 72–76 possessive 22 relative 10 protolanguage 116, 117 psycholinguistics 45 punctuation 134, 149 Q quantitative analysis 178, 190 quantitative approaches 107, 114 quantitative methods 105, 110, 113 Quirk, Randolph 22, 25, 215 R Raumolin-Brunberg, Helena 15, 68–69, 71–73 real-world meaning 44 reanalysis 15, 17–19, 27, 94, 149, 244 reification 127–129 resultative 84, 93, 99, 250–252 Ringe, Don 106, 108–115, 119–121 S Sapir, Edward 20, 186, 191 schemata 128 scientific method 104 Scots 2, 125, 128–130, 134, 136, 137, 142, 145, 148, 178 accent 166 Lowland 2 Older 125, 140–142, 148, 189 people 154 Present-day 133
Scots-Irish 2–3 Scottish English 126, 128, 129 Scottish National Dictionary (SND) 129, 189 second language acquisition 53 SED see Survey of English Dialects semi-modals 215–217, 225 see also modal sexist language debates 66 Shakespeare, William 4, 187, 223, 236 situational context 41, 46, 52, 55 Slobin, Dan 39–40, 47–48, 50–53, 55–56, 59–60 SND see Scottish National Dictionary sociolinguistics 57, 103, 179 spoken language 48, 219 standard English 3, 23, 133, 179, 185 see Scottish Standard English emerging 4 written 8, 49, 60 stress 51, 86, 91, 187 metrical 204 structuralism 69 stylistic competence 130, 134 subfamilies 107, 109, 116 Survey of English Dialects (SED) 4, 5, 8 Swadesh-type meaning lists 114 syntactic change 24, 54, 68, 235 T tagging 125, 130–132, 135 Taylor, Ann see Ringe telic verbs 249, 250, 252 telicity 83, 85, 87, 96 trace 46 tree model 110, 120 see also biological tree, family tree TreeView 116 universal grammar 34, 54 V valency 85, 244 variable rule analysis 6–7 variationist paradigm 103–104 Vedic 109 vulgarity 153
262 Subject index
W Warnow, Tandy see Ringe wave model 110 weight-ordering 200, 203 word order 53, 54, 59, 90 in OE 195–199, 202, 203, 209, 244, 247 written language 48, 136
Y Yngve, Victor 31, 38–39, 42–45, 49–50, 52–55
CURRENT ISSUES IN LINGUISTIC THEORY E. F. K. Koerner Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin The Current Issues in Linguistic Theory (CILT) series 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 alternative 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/jbp 210 DWORKIN, Steven N. and Dieter WANNER (eds.): New Approaches to Old Problems. Issues in Romance historical linguistics. 2000. xiv, 235 pp. 211 ELŠÍK, Viktor and Yaron MATRAS (eds.): Grammatical Relations in Romani. The Noun Phrase. with a Foreword by Frans Plank (Universität Konstanz). 2000. x, 244 pp. 212 REPETTI, Lori (ed.): Phonological Theory and the Dialects of Italy. 2000. x, 301 pp. 213 SORNICOLA, Rosanna, Erich POPPE and Ariel SHISHA-HALEVY (eds.): Stability, Variation and Change of Word-Order Patterns over Time. with the assistance of Paola Como. 2000. xxxii, 323 pp. 214 WEIGAND, Edda and Marcelo DASCAL (eds.): Negotiation and Power in Dialogic Interaction. 2001. viii, 303 pp. 215 BRINTON, Laurel J. (ed.): Historical Linguistics 1999. Selected papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999. 2001. xii, 398 pp. 216 CAMPS, Joaquim and Caroline R. WILTSHIRE (eds.): Romance Syntax, Semantics and L2 Acquisition. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2001. xii, 246 pp. 217 WILTSHIRE, Caroline R. and Joaquim CAMPS (eds.): Romance Phonology and Variation. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2002. xii, 238 pp. 218 BENDJABALLAH, Sabrina, Wolfgang U. DRESSLER, Oskar E. PFEIFFER and Maria D. VOEIKOVA (eds.): Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000. 2002. viii, 317 pp. 219 ANDERSEN, Henning (ed.): Actualization. Linguistic Change in Progress. Papers from a workshop held at the 14th International Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999. 2001. vii, 250 pp. 220 SATTERFIELD, Teresa, Christina M. TORTORA and Diana CRESTI (eds.): Current Issues in Romance Languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann Arbor, 8–11 April 1999. 2002. viii, 412 pp. 221 D'HULST, Yves, Johan ROORYCK and Jan SCHROTEN (eds.): Romance Languages and Linguistic Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp. 222 HERSCHENSOHN, Julia, Enrique MALLÉN and Karen ZAGONA (eds.): Features and Interfaces in Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp. 223 FANEGO, Teresa, Javier PÉREZ-GUERRA and María José LÓPEZ-COUZO (eds.): English Historical Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 1. 2002. x, 306 pp. 224 FANEGO, Teresa, Belén MÉNDEZ-NAYA and Elena SEOANE (eds.): Sounds, Words, Texts and Change. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp. 225 SHAHIN, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp. 226 LEVIN, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics. 2002. xviii, 592 pp. 227 FAVA, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002. xxiv, 353 pp. 228 NEVIN, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp.
229 NEVIN, Bruce E. and Stephen B. JOHNSON (eds.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp. 230 PARKINSON, Dilworth B. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley, California 2000. 2002. xiv, 250 pp. 231 CRAVENS, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance sound change. 2002. xii, 163 pp. 232 BEYSSADE, Claire, Reineke BOK-BENNEMA, Frank DRIJKONINGEN and Paola MONACHESI (eds.): Romance Languages and Linguistic Theory 2000. Selected papers from 'Going Romance' 2000, Utrecht, 30 November–2 December. 2002. viii, 354 pp. 233 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume I: Segmental structure. 2003. x, 308 pp. 234 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp. 235 LINN, Andrew R. and Nicola McLELLAND (eds.): Standardization. Studies from the Germanic languages. 2002. xii, 258 pp. 236 SIMON-VANDENBERGEN, Anne-Marie, Miriam TAVERNIERS and Louise J. RAVELLI (eds.): Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp. 237 BLAKE, Barry J. and Kate BURRIDGE (eds.): Historical Linguistics 2001. Selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial Assistant: Jo Taylor. 2003. x, 444 pp. 238 NÚÑEZ-CEDEÑO, Rafael, Luis LÓPEZ and Richard CAMERON (eds.): A Romance Perspective on Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages (LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp. 239 ANDERSEN, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp. 240 JANSE, Mark and Sijmen TOL (eds.): Language Death and Language Maintenance. Theoretical, practical and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp. 241 LECARME, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp. 242 SEUREN, Pieter A.M. and Gerard KEMPEN (eds.): Verb Constructions in German and Dutch. 2003. vi, 316 pp. 243 CUYCKENS, Hubert, Thomas BERG, René DIRVEN and Klaus-Uwe PANTHER (eds.): Motivation in Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp. 244 PÉREZ-LEROUX, Ana Teresa and Yves ROBERGE (eds.): Romance Linguistics. Theory and Acquisition. Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002. 2003. viii, 388 pp. 245 QUER, Josep, Jan SCHROTEN, Mauro SCORRETTI, Petra SLEEMAN and Els VERHEUGD (eds.): Romance Languages and Linguistic Theory 2001. Selected papers from 'Going Romance', Amsterdam, 6–8 December 2001. 2003. viii, 355 pp. 246 HOLISKY, Dee Ann and Kevin TUITE (eds.): Current Trends in Caucasian, East European and Inner Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp. 247 PARKINSON, Dilworth B. and Samira FARWANEH (eds.): Perspectives on Arabic Linguistics XV. Papers from the Fifteenth Annual Symposium on Arabic Linguistics, Salt Lake City 2001. 2003. x, 214 pp. 248 WEIGAND, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004 vi, 280 pp. + index. 249 BOWERN, Claire and Harold KOCH (eds.): Australian Languages. Classification and the comparative method. 2004. xii, 377 pp. (incl. CD-Rom). 251 KAY, Christian, Simon HOROBIN and Jeremy SMITH (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and Morphology. 2004. x, 255 pp. + index. 252 KAY, Christian, Carole HOUGH and Irené WOTHERSPOON (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and Transmission. 2004. xii, 265 pp. + index.