244 76 6MB
English Pages 343 [344] Year 2014
Michael Adams, Laurel J. Brinton, R.D. Fulk (Eds.) Studies in the History of the English Language VI
Topics in English Linguistics
Edited by Elizabeth Closs Traugott Bernd Kortmann
Volume 85
Studies in the History of the English Language VI Evidence and Method in Histories of English Edited by Michael Adams Laurel J. Brinton R.D. Fulk
ISBN 978-3-11-034591-9 e-ISBN (PDF) 978-3-11-034595-7 e-ISBN (EPUB) 978-3-11-039502-0 ISSN 1434-3452 Library of Congress Cataloging-in-Publication data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliografische Information der Deutschen Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutschen Nationalbibliografie; detailed bibliographic data are available on the internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Munich/Boston Typesetting: Compuscript Ltd., Shannon, Ireland Printing: CPI Books GmbH, Leck ∞ Printed on acid-free paper Printed in Germany www.degruyter.com
Contents Michael Adams Introduction: Evidence and method in the historical study of English
1
Part I Corpora, evidence, method Lauren Fonteyn and Hubert Cuyckens The development of free adjuncts in English and Dutch
15
Joanna Nykiel Semantic dependencies and the history of ellipsis alternation
51
David West Brown and Chris C. Palmer The phrasal verb in American English: Using corpora to track down historical trends in particle distribution, register variation, and noun collocations 71 Reijirou Shibasaki On the grammaticalization of the thing is and related issues in the history of American English 99 Betty S. Phillips Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
123
Part II Reconfiguring History: Overlooked evidence of English Donka Minkova Metrical resolution, spelling, and the reconstruction of Old English syllabification 137 Anatoly Liberman The shortest history of vowel lengthening in English
161
Paul A. Johnston, Jr. Vowel system restructuring in the West Midlands of England
183
Megan E. Hartman Style and politics in The Battle of Brunanburh and The Battle of Maldon Don Chapman Trinitarian terminology in Old English liturgical creeds
219
201
vi
Contents
Part III Emerging paradigms: New methods, new evidence Alexander Bergs and Meike Pentrel Ælc þara þe þas min word gehierþ and þa wyrcþ … : Psycholinguistic perspectives on early Englishes 249 William A. Kretzschmar, Jr. Complex systems and the history of the English language Colette Moore An ideological history of the English term onomatopoeia
Name index Subject index
323 332
277 307
Michael Adams
Introduction: Evidence and method in the historical study of English 1 Introduction How do we know about the history of the English language? Indeed, how do we know about anything at all? In the process of our perceiving the physical world, data cascades over our senses, most of it sheeting away, only a fraction of it retained and interpreted. In A Treatise of Human Nature, David Hume ([1739]1978: 252) noted that our perceptions “succeed each other with an inconceivable rapidity, and are in a perpetual flux and movement” beyond our intellectual management; at the same time, perceptions are neither constant to our minds nor, in themselves, coherent. We believe many things because of the excessively “smooth passage of the imagination” (Hume [1739]1978: 205) over the data we retain, but imagination is not method, and what we imagine is not evidence, nothing on which we can found knowledge of the world around us – which is why Hume is an arch-skeptic. Unwilling to abandon knowledge, Kant proposed the synthesis of apprehension to address part of Hume’s problem: “The synthesis of apprehension (…) is exercised directly on sense-impressions or sense. It involves a ‘taking up’ of given sense impressions into empirical consciousness. Through apprehension of given sense-impressions we acquire sense-perception” (Paton 1970: 1.359; see Kant [1787]1965: 143–146, or A120–123). Categories in the understanding organize our interpretation of data; when impressions become perceptions as a result of those categorical interventions, data (impressions) rise in value to become evidence (perceptions) on which knowledge can depend. Knowledge does not proceed from data, in other words, but only from evidence derived from some intellectual operation. Kant’s interdependent manifolds and categories constitute the mind’s methodology. Obviously, knowing about the history of English is different from knowing about the physical world. First, until recently, there was no cascade of data to flood historical inquiry, certainly no superfluous data; indeed, data for linguistic phenomena of earlier periods are conspicuously scarce. Second, there is no “natural” method for treating those data, no given system of understanding it, and no reason to assume that any method contrived by scholars would be universal in its force and applications. Thus, the relationship between evidence and methodology in English historical linguistics is a perennially vexed one. Does one
2
Michael Adams
enter a set of data with a method already articulated and ready to operate, from which history is then made? Or does a body of data stimulate new method? Or is the interaction of method and data yet more complicated, so that it is difficult to tell which, if either, leads the process? Or perhaps they cooperate in ways that warrant illustration and consideration. Significantly – and here historical linguistics is similar to the physical and other sciences – data are not the same as evidence, and even “the notion ‘data’ is not unproblematic” (Fitzmaurice and Smith 2012: 19). Whatever they are, data become evidence when configured methodologically to answer a certain question or set of questions about the history of English; put another way, data may provoke a methodological intervention, but evidence represents a methodological evaluation of data. Given the paucity of data for early periods or particular features of English – not to mention the textual instability of those data – historical linguistics requires methodologies to locate data as well as to evaluate them. Thus, as Labov (1994: 11) has argued, “Historical linguistics can be thought of as the art of making the best use of bad data”.1 Once the data are assembled, investigation of them requires still further methodologies. The less – or more problematic – the data, the more flexible the methodologies must be, grounded in linguistic theory, certainly, but responsive to the immediate situation – that is, the transformation of data gathered for a purpose into evidence. Examining smaller or isolated sets of data from various methodological perspectives enables us to wring more water from the stone. This flexibility is significantly different from the intersection of method – that is, the mind’s methodology – and psychological (or now we might say cognitive) universality typical of philosophy and the grounding of knowledge-claims about the physical world. Philosophers must assess the relationships between the nature of physical phenomena and the nature of humankind. Hume thought the obstacles on the path from data to knowledge insurmountable, and so he pronounced a universal skepticism. Kant, insisting that we can indeed know things, proposed a universal transcendentalism, but both philosophers were talking about what was true independent of our knowing it; neither believed that we could construct knowledge. Historical linguists, by way of contrast, must assess the relationships between language phenomena of the past and human cultures, a different intellectual burden carried along by different methodological assumptions. Thus, the current volume collects thirteen distinct historical linguistic (re-) constructions, and it thereby demonstrates a range of methodological approaches
1 See also Bergs and Pentrel, this volume.
Evidence and method in the historical study of English
3
over a wide variety of linguistic subjects, from Old English to Present-Day English and from phonology to syntax to lexis. Each article is a valuable, self-contained inquiry, and each is methodologically creative in developing evidence from data – often in some way unexpected data – elicited from the historical record with some thorny problem in mind – but not so creative that it succumbs to smooth imagination. This sort of creativity, heuristic reasoning as opposed to application of an algorithm, as Stefan Dollinger (2008: 358) puts it, is not as prominent in historical linguistics as it is in other disciplines, where it is theoretically grounded in the anti-realist version of methodology (van Fraassen 1980; but see Sarkar 1998, in dissent). As this volume demonstrates, however, its profile in English historical linguistics is rising, especially in response to the bad data problem. Given the contingency of historical data, as well as of the evidence constructed from them, knowledge of the history of English practically requires methodological pluralism, so that we may reassess what we think we already know from fresh perspectives. One might assume that the more advanced a science is, the more univocal its methodology will be, but actually methodological pluralism is typical of the natural and social sciences. For instance, according to Steven Shapin, a leading historian of science, “the” Scientific Method “is, and always has been, subject to diverse construals” (2010: 7). The “the” often mistakenly placed before “Scientific Method” is like the one similarly misplaced before “dictionary” (as in “the dictionary defines word x as y”); while it betrays naïve popular assumptions, professionals with a stake in either science or dictionaries are not wholly immune to such assumptions. Historical linguists, too, must be careful not to assume that one method will solve all problems; indeed, as several contributions to this volume make clear, it may take several methods in some newly proposed relationship to solve just one problem. Even if every scientist of one kind or another – physicists or historical linguists – agreed on the tribe’s methodology, methodological concerns would not necessarily be in the foreground. As the Nobel Prize-winning physicist Percy Bridgman observed, “It seems to me that there is a great deal of ballyhoo about scientific method. (…) Scientific method is what working scientists do, not what other people or even they themselves may say about it. No working scientist, when he plans an experiment in the laboratory, asks himself whether he is being properly scientific, nor is he interested in whatever method he may be using as method. (…) The working scientist is always too much concerned with getting down to brass tacks to be willing to spend his time on generalities” (quoted in Shapin 2010: 37). In the present volume, historical linguists devise and operate historical experiments in order to get down to the brass tacks of the English language, and their methodologies are motived by those tacks, with the methodologies themselves ways of getting down to them.
4
Michael Adams
The implicitness of methodology reflected in Bridgman’s comment leads Shapin to conclude that “all metascientific statements by practicing scientists are best ignored” (2010: 38), because the scientists themselves may not have a well- considered or well-articulated methodological position. The historical linguists who contribute to this volume are of course aware of their own methodologies, and while framing their various experiments, they comment on their methodological orientations. But methodology is not for them a subject in itself (at least, not in this volume); rather, it is the means to explore a discrete problem with the strongest explanatory value compatible with a coherent theoretical orientation. Thus, the editors of this volume, in juxtaposing the various papers included in it, as well as readers of the whole volume, as they contrast the various contributions and consider the methodologies implicit in them, will do the metascientific and metahistorical thinking about historical linguistic inquiry, taking their lead from the contributors. Also significant, and characteristic of most contributions to this volume, is the view that an effective method of studying language change may cross traditional contextual and modular boundaries by relying on language-external, or at least grammar-external, evidence. The theoretical soundness of such crossing and combining has been a matter of dispute in linguistics. As Arnold Zwicky once observed, while mapping the methodological conflicts among generativists and between generativists and everybody else, “The distinction is usually made invid iously – only internal evidence is probative – or defensively – external evidence, or at least some types of external evidence, are relevant and useful” (1980: 598). Thus, while “differences of opinion as to whether external evidence is useful or necessary or neither (…) arise when linguists, even those with similar beliefs about the goals and assumptions of their field, are confronted with a choice among alternative descriptions of some phenomenon in a language”. Synthetic methodologies are sometimes essential, for, in pursuing an explanation for those phenomena, “you can continue to look for lines of internal evidence favoring one alternative, or you can strike out in search of some sort of relevant external evidence, in place of or in addition to internal evidence” (1980: 599). The explanatory value of such methodological synthesis is apparent for anthropological linguistics and sociolinguistics in the case of enregisterment (Adams 2009: 116), for instance, and this volume clearly advances it for English historical linguistics. Richard E. Grandy (1980: 608) noted, in the same symposium from which Zwicky’s comments are culled, that the preoccupation of linguists in the latetwentieth century was to develop theory: “The focus of attention has shifted because pre-Chomskyan grammarians2 emphasized the problem of gathering
2 Grandy points particularly to Leonard Bloomfield.
Evidence and method in the historical study of English
5
data, believing in their Millian ways that carefully gathered data would lead to a unique grammatical theory”. Of course, it has not done so, but “disagreement on theory”, Grandy asserted, “need not mean disagreement on what consti tutes data” (608), and he is surely right. The argument engaged by Zwicky and Grandy has never been resolved, though it has been quite thoroughly elaborated in recent years. The chapters by various hands in Penke and Rosenbach (2007), for instance, explore how data of various kinds can be construed as evidence of innateness and the terms on which linguistic competence operates. In this volume, contributors engage problems of data collection and often the relationship between data and theory, but all are most creative and informative in the reification of data into historical linguistic evidence.
2 The plan of this book The papers in this volume view the relationship between evidence and method from many perspectives – effectively as many perspectives as there are papers – organized into three sections that generalize approaches to evidence and the status of evidence in the historical study of English.
2.1 Corpora, evidence, method If any one thing has altered fundamentally the way we do historical linguistics of any kind in the last two decades, it must be the increasingly sophisticated development as well as the increasingly sophisticated use of corpora, some of which are artisanal, others of which make it possible to search Big Data. Each has its partisans (see Davies 2012 for the latter, Hundt and Leech 2012 for the former), and both have limitations, but a synthetic approach to corpora and their data allows linguists to construct a better sort of evidence (Mair 2012). The section of this volume devoted to the relationships among corpora, evidence, and method includes the following papers: “The development of free adjuncts in English and Dutch” by Lauren Fonteyn and Hubert Cuyckens; “Semantic dependencies and the history of ellipsis alternation” by Joanna Nykiel; “The phrasal verb in American English: Using corpora to track down historical trends in particle distribution, register variation, and noun collocations” by David West Brown and Chris C. Palmer; “On the grammatical ization of (the) thing is and related issues in the history of American English” by Reijirou Shibasaki; and “Alfredian īo ~ ēo, reluctant function words, and Schriftbilder” by Betty S. Phillips.
6
Michael Adams
The various contributions to this section illustrate ways in which methodological impulses push data from corpora in distinct evidential directions. Nearly all the contributions in this section evaluate data comparatively. For instance, Lauren Fonteyn and Hubert Cuyckens focus on syntax and perform an interlinguistic comparison (English and Dutch) in order to analyze the development of free adjuncts in the form of participial phrases, both their semantic diversification and their reanalysis to fully subordinate clauses. As Fonteyn and Cuyckens write, the key factors in that development “do not provide a conclusive answer to the question why the construction has been retained in English but not in another Germanic language such as Dutch. The answer to this question needs to be sought in the increasing discourse-functional and the increasing formal overlap between the English verbal gerund and present participle”. Early Modern English and Early Modern Dutch free adjuncts developed differently because the grammars of the two languages were different in one feature, the presence or absence of a gerund structurally similar to the participle. Fonteyn and Cuyckens challenge conventional methodological assumptions about the interaction of language-internal and -external factors, but they also conclude that “Methodologically, (…) historical linguists need to carefully consider data within its synchronic linguistic context”. Joanna Nykiel likewise brings grammar-external factors to bear when she shows us a syntactic constraint conditioned not by a purely syntactic factor (pres ence or absence of preposition stranding) but by semantic dependencies, in a cross-period comparison. Reflecting on a range of recent studies, including her own, Nykiel notes parallels between corpus and psycholinguistic data, which also hold for her argument about ellipsis in this chapter. Taken together, as she writes, such studies of particular features, now that they are accumulating, “raise the question of whether or not syntactic variation should be accounted for by extra-grammatical factors” more generally than any linguist focusing on one feature or another probably realized. Among contributors to this volume, Nykiel is not alone in relying on psycholinguistics to help establish evidentiary parameters; see also Bergs and Pentrel, and Kretzschmar, below. David West Brown and Chris C. Palmer investigate questions of usage in a study of phrasal verbs in American English and employ a cross-register analysis to show how syntactic, semantic, and sociolinguistic forces can influence the histories of specific phrasal verbs, an unconventional approach rarely invoked in the substantial literature on phrasal verbs. But their paper has ramifications that reach far beyond phrasal verbs. Creating “a full account of the growth or decline of a linguistic form”, they write, requires attention to “large, continuous, multi-register corpora”. As a result of this attention, they “are able to push exist ing quantitative approaches in several underexamined and fruitful directions”,
Evidence and method in the historical study of English
7
including (1) “the effect of register on the history of phrasal verbs and their particles”, but by implication any feature of English; (2) correlated changes in phrasal verbs and their collocates; and (3) the impact of cultural change on language change. Interestingly, they also “bring data from the material world into conversation with linguistic data, and probe their statistical relationships”. As in several other contributions to this volume, data from outside of grammar are important to defining the evidential value of data from within. In his study of (the) thing is, Reijirou Shibasaki shows how a syntactic con struction, specifically a shell construction, when viewed cross-modularly in a pragmatic perspective, can be understood as a multifunctional discourse marker – clearly, variation in use of (the) thing is cannot be explained in syntactic terms alone. Indeed, he writes, “constructional change, expansion and variation” of the feature all “are discourse-based”. He argues, too, that (the) thing is is currently undergoing grammaticalization, though its path is not typical, and some minor variants, if sustained or advanced, would be incompatible with grammaticalization. In order to arrive at these conclusions, Shibasaki rigorously cross-compares half a dozen corpora, representing both historical and contemporary, as well as both written and spoken English, and it is such comparison that raises relatively infrequent data to the status of significant evidence. Finally, Betty S. Phillips draws on corpus data to propose that orthographic variation in Old English – which she tests with Early West Saxon īo ~ ēo – has a phonological and morphological rather than a purely conventional basis, another unexpected conclusion requiring methodological flexibility about the sources, status, and interactions of evidence in historical linguistic inquiry. When considering the “reliability of scribal spellings”, she writes, “one might think that scribes would have a more engrained vision of the spelling of frequent words”, and for function words, this proves to be true, but “for the content words, the more frequent words are the very ones which are more likely to contain the innovative spelling”. Such a conclusion has broad implications for study of the manuscript basis of linguistic data, not only Old English data but data from later periods, too.
2.2 Reconfiguring history: Overlooked evidence of English The section on overlooked evidence in the historical linguistics of English includes “Metrical resolution, spelling, and the reconstruction of Old English syllabification”, by Donka Minkova; “The shortest history of vowel lengthening in English”, by Anatoly Liberman; “Vowel system restructuring in the West Midlands of England”, by Paul A. Johnston, Jr.; “Style and politics in The Battle of Brunanburh
8
Michael Adams
and The Battle of Maldon”, by Megan E. Hartman; and “Trinitarian terminology in Old English liturgical creeds”, by Don Chapman. Operating on conventional methodological assumptions, those studying the history of English can overlook important evidence; the construction of this evidence from available data implies new methodological insight. Leading the section on overlooked evidence, Donka Minkova deals straightforwardly with the ways that a perennial source of interference in historical linguistics, orthographic indeterminacy, injects uncertainty into phonological analysis. In doing so, she projects measurable features of Present-Day English onto Old English to reassess the evidence for ambisyllabicity in the latter. Minkova’s paper illustrates the importance of non-phonological factors in theorizing about Old English phonology, as well as the value of current data in the construction of evidence for Old English. For both of these reasons, it is in the vanguard of a rising analytic trend. Anatoly Liberman’s paper deals with a cluster of phonological problems grouped under the conventional heading “lengthening”. Traditionally, the motivations for some of these lengthenings have not been very apparent. Liberman explains them chiefly as the result of systemic shifts, and he ingeniously posits change to the vowel system of English as a reaction to prosodic constraints and phonological pressure on the consonant system that could not be realized directly, “shadow evidence” of a sort conventionally overlooked. Thus, Liberman points a way forward by explaining phenomena that conventional phonological approaches have been incapable of explaining definitively. He rejects the relevance of data conventionally approved because it works as evidence in those approaches – characterized by formalization, teleology, and circular reasoning – in favor of data that become evidence against those approaches, but also for arguments about lengthening with greater explanatory value. Through a close analysis of the Old English Battle of Brunanburh and The Battle of Maldon, two nationalistic poems probably of the late tenth century, Megan E. Hartman shows how something as external to the grammar as political events can influence syntax and prosody as elements of poetic style. As Hartman puts it, “The verifiably contrasting styles in poems that were composed in roughly the same time period point towards a new type of evidence relevant for poetic analysis: syntactic and metrical features as an indicator of pragmatic aims”. The data in the poems have not changed, but those data are interpreted as evidence for a quite unexpected series of influences, in which nationalism drives rhetoric, rhetoric drives pragmatics, pragmatics drives syntax and prosody, and these last drive poetic meter, the data of which give rise to reconstruction of the whole series. Such an analysis invests poetic meter with greater cultural significance than we might otherwise assume, and ideology with a greater impact on linguistic features of poetry than we might suppose.
Evidence and method in the historical study of English
9
Paul A. Johnston, Jr., demonstrates plainly the value of sociolinguistics, regional history, and geography in solving historical linguistic problems in his study of vowel classes and mergers in late eighteenth- and early nineteenth-century English in Birmingham and the neighboring Black Country. Although conventionally one might investigate a dialect’s phonology on the basis of regularities within the available data, Johnson insists in his contribution to this volume that “The clearest evidence that outside influence has taken place in the history of the dialect’s phonology is the presence of discontinuities: changes that do not occur in a stepwise fashion, that do not look like ‘regular sound change’, and that send the trajectory of the development in a completely different direction than it was otherwise going”. The discontinuities are then explained by recourse to cultural material well beyond linguistics, let alone the regional phonology in question. Don Chapman’s paper shows how the requirements of lexical precision in a religious context – unusual evidence of linguistic motivation – led to the devel opment of a technical vocabulary and ultimately to elaboration of the functional possibilities of the English language. On the basis of collation of Old English creeds, and drawing on their relationship to Latin originals as translations that demanded deliberate choice on the part of their translators, Chapman detects similarities in practice among those translators, including Ælfric, which suggest consolidation of English creedal terminology in the tenth century ce. Interest ingly, data from the collation sometimes converge and sometimes diverge, but both tendencies prove useful evidence of standardization. Chapman’s thesis has implications for understanding the dynamics of standardization in later periods.
2.3 Emerging paradigms: New methods, new evidence The volume ends with three paradigm-shifting papers: “Ælc þara þe þas min word gehierþ and þa wyrcþ … : Psycholinguistic perspectives on early English”, by Alexander Bergs and Meike Pentrel; “Complex systems and the history of the English language”, by William A. Kretzschmar, Jr.; and “An ideological history of the term onomatopoeia”, by Colette Moore. In the first of these, Bergs and Pentrel challenge the unidirectionality of the uniformitarian principle (Labov 1994: 21–23), arguing that historical texts, rather than supporting the supposition that the same conditions and processes should be found in historical linguistic situations as in present-day ones, may require us to revise current theory. A richly layered and imaginative argument, it harnesses the combined power of cognitive linguistics, psycholinguistics, and historical linguistics, and it imports into syntactic analysis considerations external to the sub-discipline of syntax, such as genre, the specific circumstances
10
Michael Adams
of composition, and matters of audience. In the way it induces different types of evidence to cooperate with one another, their approach resembles Angus McIntosh’s influential “fit-technique” of locating dialect features in late Middle English (McIntosh 1963). Bergs and Pentrel express this relationship as a feedback loop balancing historical data – represented by Beowulf and the Paston Letters – and Present-Day English processing strategies, which recursively must account for production strategies embedded in the texts they consider. In The Linguistics of Speech (2009), Kretzschmar proposed that linguistic data can be interpreted newly given the insights of complexity theory, and he has extended that argument in several works since, as have many of his students (see, for instance, contributions to Burkette and Antieau [eds.] 2012). Convention ally, data occurring at high frequencies have a stronger evidential status than less-frequently occurring data. Statistical significance, for instance, is one way to mark the credibility of data as evidence. In his contribution to this volume, in a sort of statistical cat’s-cradle, Kretzschmar shifts our attention from data as distributed on S-curves to that same data redistributed on A-curves; the new pattern, a new distributional model, raises the evidential value of low-frequency tokens, with methodological implications for all sub-disciplines of linguistics. Finally, Moore’s paper brings social history to bear on the developing use of the term onomatopoeia. She develops what she calls an “ideolexical history”, which serves as “a small window into the history of language attitudes and ideol ogies”. As Moore notes, idiolexical history may not be new, but supplying it with systematically collected data certainly is, and Moore “constructed a sample of hits in pre-twentieth-century English texts” from widely available databases. Moore underscores the methodological value of her approach: “The data of our past written uses of the word, then, become evidence for interpretative shifts; to analyze this data is to better understand the cultural use of metalanguage”. She shows how the use and spread of onomatopoeia were influenced by changing ideologies of language, resulting in a metadiscourse on metalanguage.
3 Conclusion The papers included in this volume were originally presented at the seventh Studies in the History of the English Language (SHEL) conference, April 26–28, 2012, on the campus of Indiana University in Bloomington, Indiana. Of course, there were many fine papers presented at the conference in which the nexus of evidence and method was not as conspicuous as it was in the papers collected here; and there are many papers beyond this volume, even within the compass of SHEL, that do focus on evidence and method, for instance, those included in
Evidence and method in the historical study of English
11
Fitzmaurice and Minkova (eds.) (2008). But because historical linguists devoted to some degree to the study of English are busy getting down to brass tacks in the routine of research and not explicitly confronting methodological issues or indulging the metadiscourse, it seems useful to bring together a set of outstanding papers like those included here that bring problems of evidence and method into the foreground of disciplinary conversation and debate. And the papers in this volume do illustrate pellucidly how evidence and methodology in the historical study of English are phases of the same disciplinary process. When we “identify” evidence, we actually construct it out of data on the basis of method, which is nonetheless interactive with the evidence it constructs, so that method shapes evidence and evidence shapes method. This is not the circularity that concerns Liberman (see above and below), but the recursive process by which we query or criticize both evidence and method into the best fit. Understanding this dynamic and then seeing it played out in various guises are essential in considering the nature of historical English linguistics, because the terms on which we generate evidence from data warrant the linguistic and historical knowledge we claim.
References Adams, Michael. 2009. Enregisterment: A special issue. American Speech 84(2). 115–117. Burkette, Allison & Lamont D. Antieau (eds.). 2012. The Linguistic Atlas Projects: Twenty-first century perspectives. [Special issue]. American Speech 87(4). Davies, Mark. 2012. Some methodological issues related to corpus-based investigations of recent syntactic changes in English. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 157–174. Oxford & New York: Oxford University Press. Dollinger, Stefan. 2008. Taking permissible shortcuts? Limited evidence, heuristic reasoning and the modal auxiliaries in early Canadian English. In Susan M. Fitzmaurice & Donka Minkova (eds.), Studies in the history of the English language IV: Empirical and analytical advances in the study of English language change, 357–386. Berlin & New York: De Gruyter Mouton. Fitzmaurice, Susan M. & Donka Minkova (eds.). 2008. Studies in the history of the English language IV: Empirical and analytical advances in the study of English language change (Topics in English Linguistics 61). Berlin & New York: De Gruyter Mouton. Fitzmaurice, Susan M. & Jeremy Smith. 2012. Evidence for the history of English: Introduction. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 19–36. Oxford & New York: Oxford University Press. Grandy, Richard E. 1980. Some thoughts on data and theory in linguistics. PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association: Volume Two: Symposia and Invited Papers. 605–609. Hume, David. [1739] 1978. A treatise of human nature. 2nd edn. L. A. Selby-Bigge & P. H. Nidditch (eds.). Oxford: Clarendon Press.
12
Michael Adams
Hundt, Marianne & Geoffrey Leech. 2012. “Small is beautiful”: On the value of standard reference corpora for observing recent grammatical change. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 175–188. Oxford & New York: Oxford University Press. Kant, Immanuel. [1787] 1965. Immanuel Kant’s critique of pure reason. Norman Kemp Smith (trans.). New York: St. Martins. Kretzschmar, William A., Jr. 2009. The linguistics of speech. Cambridge & New York: Cambridge University Press. Labov, William. 1994. Principles of linguistic change. Vol. 1: Internal factors (Language in Society 20). Oxford & Cambridge, MA: Blackwell. Mair, Christian. 2012. From opportunistic to systematic use of the Web as corpus: Do-support with got (to) in contemporary American English. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 245–255. Oxford & New York: Oxford University Press. McIntosh, Angus. 1963. A new approach to Middle English dialectology. English Studies 44(1). 1–11. Paton, H. J. 1970. Kant’s metaphysic of experience: A commentary on the first half of the Kritik der reinen Vernunft. 2 vols. London: George Allen & Unwin. Penke, Martina & Anette Rosenbach (eds.). 2007. What counts as evidence in linguistics: The case of innateness. Amsterdam & Philadelphia. John Benjamins. Sarkar, Husain. 1998. Anti-realism against methodology. Synthese 116(3). 379–402. Shapin, Steven. 2010. Never pure: Historical studies of science as if it was produced by people with bodies, situated in time, space, culture, and society, and struggling for credibility and authority. Baltimore: The Johns Hopkins University Press. van Fraassen, Bas. 1980. The scientific image. Oxford & New York: Oxford University Press. Zwicky, Arnold. 1980. “Internal” and “external” evidence in linguistics. PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association: Volume Two: Symposia and Invited Papers. 598–604.
Part I Corpora, Evidence, Method
Lauren Fonteyn and Hubert Cuyckens
The development of free adjuncts in English and Dutch 1 Introduction This paper will study the diachronic development of (a particular type of) presentparticipial clauses in English and Dutch, viz. the present-participial free adjunct: (1) a. Having made these necessary provisions, you are to keep in your storehouse every little thing the poor fishermen have occasion for. (1718, PPCMBE1) b. (…) sometimes it acts upon the kidneys, at others on the bowels, producing griping and diarrhoea. (1807, PPCMBE) (2) a. Vatbaar van aard zijnde leerde hij de kunst van het lezen en schrijven, en de gronden der rekenkunst met ongemeene snelheid. (1823, CCHD2) ‘Being of impressionable nature (i.e., being receptive to learning), he learned to read and write and the basics of arithmetic in no time.’ b. [Hij] sloot (…) zyn kaamerdeur, zeggende tegen den Goudmaaker dat hy dadelyk zoude weederkoomen. (1756, CCHD) ‘He closed (…) his door, saying to the goldsmith that he would return immediately.’ Present-participial free adjuncts, as in (1) and (2) (from now on simply “free adjuncts”),3 are non-finite subordinate clauses (or converb clauses) with a present participle as head. They are “detached” or separated from the main clause by means of a pause, and they have a separate intonational contour (Thompson 1983: 43; Kortmann 1991: 1, 6). Further, they lack an explicit subject: their subject is understood and equals that of the matrix clause, i.e., it is controlled by the subject of the matrix clause. In this respect, free adjuncts differ from absolute
1 Penn Parsed Corpus of Modern British English. For more information on this corpus we refer the reader to Section 2. 2 Compilation Corpus of Historical Dutch. For more information on this corpus we refer the reader to Section 2. 3 The term “free adjunct” has been taken from Kortmann’s (1991) monograph, where it covers a much broader group of clauses (e.g., past-participial free adjuncts as in Happily married, they never fight). This study will focus on one particular subtype, namely the present-participial free adjunct. For brevity’s sake, the term “free adjunct” will be adopted here.
16
Lauren Fonteyn and Hubert Cuyckens
constructions (as in (3)), which contain an explicitly expressed subject within the subordinate clause. (3) The coach being crowded, Fred had to stand. (Kortmann 1991: 5) Free adjuncts typically express adverbial meanings, such as time (1a), condition, concession, contrast, instrument, manner, purpose, result, or cause (2a), or adverbial-like meanings such as addition/accompanying circumstance to the event in the matrix clause (add/acc), as in (2b), and exemplification/specification of the event in the matrix clause (ex/spec), as in (1b) (Kortmann 1991; Declerck 1991: 457; Killie and Swan 2009).4 The free adjunct discussed in the present paper is believed to have developed from an adverbial-like present participial use in earlier Germanic. This Germanic participle showed hardly any clausal features and only rarely occurred with an object; it was mostly derived from an intransitive verb and expressed an accompanying circumstance or state of mind (Callaway 1901; Killie 2006), as illustrated in (4). Here, al liueand expresses the circumstance accompanying the subject’s descent into hell: (4) Cum þe deþe vp þe yuel, and descenden hij into helle al liueand. (c1350, PPCME2) ‘Should death come upon the evil, and they descend into hell living.’ Presumably due to Latin influence (Callaway 1901; Sørensen 1957; Killie and Swan 2009), this original Germanic form then gradually acquired more clausal features, resulting in the adverbial present participle clause or free adjunct.5 Interestingly, despite their common Germanic origin, the frequency of the free adjunct in
4 Adverbial present participle clauses have also been termed “converb clauses” (Haspelmath 1995) and “predicative adjuncts” (Huddleston and Pullum 2002); for a lengthy discussion on terminology, see Kortmann (1991). 5 In their (2009) article, Killie and Swan suggest that, while the free adjunct was modeled on Latin participial clauses, it did not fully become part of English grammar until the Middle English period, in the form of add/acc and ex/spec clauses. In their view, such add/acc and ex/ spec free adjuncts were then reanalyzed as fully subordinate clauses expressing temporal and causal relations. Callaway’s (1901) observations, however, suggest a different development: he points out that, even though their number was small and their presence can be attributed to Latin influence, free adjuncts do already occur in Old English. Most of them occur sentencefinally, but some appear in initial position and have a temporal meaning. The question thus arises whether the emergence of the free adjunct in English is merely a result of the reanalysis of add/ acc and ex/spec clauses, or whether the source of the free adjunct was already present as early as Old English. It is very well possible that these are cooperative phenomena; this issue requires an in-depth analysis of Old English data, and it will not be further elaborated in this paper.
The development of free adjuncts in English and Dutch
17
Present-Day Dutch is far lower than in English, where it is used quite extensively. Present-Day Dutch instances such as (2) moreover sound stilted, and the translation of an example such as that in (1) would be problematic. This, in fact, holds true for all other Germanic languages, and English free adjuncts often have to be translated by means of other constructions, such as finite clauses (König and van der Auwera 1990; Killie 2006, 2007). While free adjuncts in Present-Day English have been examined in considerable detail (for instance, in Kortmann’s 1991 seminal monograph), less attention has been paid to their diachronic development. Only a few studies have focused on the history of the free adjunct and have studied the question of why the construction still exists in Present-Day English but was lost in all other Germanic languages. Focusing on Danish and Norwegian, Killie (2006, 2007) proposes several linguistic and sociolinguistic factors that might either have influenced the success of the present participial free adjunct in Present-Day English or stimulated its loss in other Germanic languages. Killie and Swan (2009) point out that free adjuncts in English, which, they assume, “were originally modelled on Latin participial clauses” (2009: 360), saw a dramatic rise from Late Middle English onwards. They argue that the success of these free adjuncts can be traced back to a set of semantic and syntactic changes that occurred exclusively in English: (i) their semantic diversification and (ii) their reanalysis from present-participial clauses with semi-coordinated status to more integrated adverbial clauses, comprising both a positional shift and the acquisition of a higher degree of syntactic integration into the main clause.6 Killie and Swan argue that these changes were triggered and then sustained by a number of related constructions which were unique to English, such as the progressive and the verbal gerund. The existing literature on Dutch present-participial clauses and their diachronic development challenges the idea that the construction expanded exclusively in English (den Hertog 1909; Rinkel 1989; van der Horst 2008; van der Wal and van Bree 2008). Data from comparable corpora in English and Dutch likewise suggest that the Dutch free adjunct underwent a development similar to that of the English free adjunct. The purpose of this paper, then, is to re-examine the development of present-participial free adjuncts in English and Dutch. In doing so, we want to obtain a better insight into the reasons and conditions that led to
6 Although not mentioned in Killie and Swan (2009), the increase of free adjuncts in the Early Modern period is doubtless also to be attributed to the expansion of vernacular writing with Latin models (see, e.g., Blatt 1957; Kohnen 2003; thanks to an anonymous reviewer for pointing this out). We have not developed the influence from Latin syntax on English (and on Dutch) in this paper, because we wanted to focus on what explains the divergent behavior of English and Dutch from the nineteenth century onwards.
18
Lauren Fonteyn and Hubert Cuyckens
the success of the free adjunct in English and consider more closely the specific position of Dutch among the Germanic languages. By comparing the development of free adjuncts in these two languages, it will be shown that their semantic diversification and reanalysis to a fully subordinate clause might have been the initial step towards establishing free adjuncts as a frequent means of clause combining, but these factors do not provide a conclusive answer to the question why the construction has been retained in English but not in another Germanic language such as Dutch. The answer to this question needs to be sought in the increasing discourse-functional and the increas ing formal overlap between the English verbal gerund and present participle. Importantly, these insights can only be gained from a thorough comparison of the evidence in both languages, and not from examining each of the languages individually. More generally, this paper makes clear that language change may be the result of the interaction of language-internal and language-external factors. It further demonstrates that language change is sensitive to the language-internal characteristics and linguistic context in which the change proceeds. Even though the situation of free adjuncts in Early Modern English and Early Modern Dutch was quite similar, they developed differently, among other reasons, because of the different grammatical contexts (presence of formally similar gerund or not). Methodologically, this means that historical linguists need to carefully consider data within its synchronic linguistic context. The paper is structured as follows. Section 2 presents data and methodology, and Section 3 the data analysis. Section 4 highlights the similarities between the historical semantic and syntactic development of the English and the Dutch free adjuncts. Section 5 then zooms in on the gradual decline of the Dutch free adjunct and suggests a number of language-external and language-internal factors underlying the decline of the Dutch free adjunct. Finally, Section 6 offers a summary of the most important findings and presents some conclusions and suggestions for further research.
2 Data and methodology The data for our study were taken from comparable historical corpora in English and Dutch. For English, we made use of the Penn parsed corpora of historical English (3,842,713 words, covering the period from 1150 to 1914); they include the Penn-Helsinki parsed corpus of Middle English, second edition (PPCME2), the Penn-Helsinki parsed corpus of Early Modern English (PPCEME), and the Penn parsed corpus of Modern British English (PPCMBE). For Dutch, we drew on the
The development of free adjuncts in English and Dutch
19
Compilation corpus of historical Dutch (CCHD). As the CCHD does not contain any literary texts for the time span 1150–1500, and lacks non-literary texts for the period 1806–1914, the Dutch data for these periods were respectively supplemented with a selection of literary texts from the Digitale bibliotheek voor de Nederlandse letteren corpus (DBNL) and with legal texts from the 38-miljoenwoorden corpus,7 thus reducing genre-bias as much as possible. The total word count for the CCHD (and supplements) was 1,785,486 words, covering the period 1250–1914.8 Both corpora were divided into six subperiods: M1 (1150–1350), M2 (1351–1500), E1 (1501–1600), E2 (1601–1710), L1 (1711–1805), and L2 (1806–1914). In addition, our data set included four English novels from the nineteenth century, with a total word count of 554,497 tokens, and their Dutch translations. These texts, taken from Project Gutenberg, were selected on the basis of whether a contemporary Dutch translation was available. Free adjuncts were extracted, per subperiod, from the part-of-speech tagged version of the Penn corpora, making use of the following tags as search strings: */HAG, */VAG, */BAG. For the Dutch corpora, which are not annotated, a PERLscript was used to search for words in ‑end(e), the ending of Dutch present partici ples, and not preceded by an article.9 From the total number of hits obtained in this way for each subperiod, a random selection was taken, and manually inspected for spurious hits. The English novels, finally, were aligned with their Dutch translations by means of alignment software; the free adjuncts were then extracted from the orig inal English texts by means of a PERL-script searching for words ending in ‑ing not preceded by an article,10 and their translational equivalents linked up with them. Tables 1 and 2 present the numerical results for the English and Dutch corpora. Present-participial free adjuncts share the ‑ing form with related, yet differ ent constructions; these were, however, excluded from our analysis. One is the ‑ing construction in (5), which has nominal characteristics: it takes the possessive determiner your, has nominal distribution (as object of prevent), and can therefore be classified as a gerund. In a similar vein, the ‑ing construction in (6) can be regarded as a gerund following the preposition on. Example (7) can, strictly speak ing, not be considered a free adjunct either, because after, but also before and
7 For more detailed information on the different corpora, see the section Corpora and primary texts. 8 Actually, the CCHD covers the time span 1250–1999, but in alignment with the Penn corpora, it was excerpted only until 1914. 9 It was anticipated that this search string would filter out most nominal and adjectival uses of the present participle. 10 It was anticipated that this search string would filter out most nominal and adjectival uses of the present participle.
20
Lauren Fonteyn and Hubert Cuyckens
Table 1: Corpus and sample information, English data English
M1
M2
E1
E2
L1
L2
Total word count 352,089 751,272 347,072 281,391 470,757 478,138 Total hits 757 2538 3462 3244 6954 7434 Sample 300 210 285 265 368 420 Hits sample 26 68 57 60 53 48 Per 10,000 1.86 10.89 19.95 26.1 21.27 17.77
Table 2: Corpus and sample information, Dutch data Dutch
M1
M2
E1
E2
L1
L2
Total word count 554,403 166,289 82,575 89,827 64,310 828,082 Total hits 2490 1266 1152 1181 673 5224 Sample 310 200 200 200 331 394 Hits sample 3 19 35 43 66 60 Per 10,000 0.43 7.23 24.41 28.27 21.18 9.61
since, can function as subordinators (followed by a finite clause or a participle) as well as prepositions (followed by a gerund). (5) (...) to prevent your being on board her? (1749, PPCEME) (gerund, object position) (6) On arriving in Oxford, I went to see my old professor. (Kortmann 1991: 13) (gerund following preposition, adverbial time clause) (7) After seeing a second apartment, and what is called the well, we happily made our way back again into the open air (…) (1836, PPCEME) (gerund or augmented free adjunct, adverbial time clause) At the same time, constructions as in (6) and (7) closely resemble augmented free adjuncts, i.e., free adjuncts introduced by a subordinating conjunction: first of all, they do not form an integral part of a clause, but rather constitute a clause on their own, which is clearly indicated by the detachment markers “intonation break” and ‘comma’ in speech and writing respectively. Secondly, (…) [they] serve an adverbial function (Kortmann 1991: 14).
Like free adjuncts, they are moreover controlled by the subject of the matrix clause, and unlike prototypical gerunds, they necessarily imply actualization and cannot designate factivity (Heyvaert 2011). Heyvaert (2011) therefore suggests that, diachronically, these structures may have started out as gerund ives, but synchronically, it has become difficult to classify them unequivocally
The development of free adjuncts in English and Dutch
21
as gerunds or participles. Because of their unclear status, construction types such as those in (6) and (7) will not be included in the analysis. They are, however, likely to have interacted with the free adjunct in its development (see Section 5.2.2). Non-ambiguous augmented free adjuncts introduced by conjunctions that cannot serve as prepositions (such as while, if, when, …) were included in the data. The constructions in (8)–(9) and (10)–(11), representing restrictive and nonrestrictive relative clauses respectively, were not included since this study only focuses on the adverbial use of the free adjunct. The present approach as such deviates from that taken by Rinkel (1989), Kortmann (1991), and van der Horst (2008), who did consider non-restrictive relative clauses to be the same construction as free adjuncts.11 (8) The blonde girl wearing green trousers looks a bit out of place. (Kortmann 1991: 16) (restrictive relative clause) (9) De opperklerck zal de schrifturen fil à fil, soo die ingedient worden, distribueren aen de gone aldaer komende schryven omme daer van copie gemaeckt te worden. (1667, CCHD) (restrictive relative clause) ‘The main clerk will, if they are sent in, distribute the scriptures fil à fil to the ones coming there to make a copy of them.’ (10) The blonde girl, wearing green trousers, looks a bit out of place. (Kortmann 1991: 16) (non-restrictive relative clause or free adjunct in medial position) (11) De comedianten, de lugt gekreegen hebbende van een merkelyke winst, welke Campo op een morgen met speelen gedaan had, besloot hem een kwade trek te speelen. (1756, CCHD) (non-restrictive relative clause or free adjunct in medial position) ‘The comedians, having heard about a reasonable profit that Campo had won one morning with gambling, decided to deceive him.’
11 We are aware that non-restrictive clauses can be ambiguous between a non-restrictive relative clause reading and a free adjunct reading (example [10] could be interpreted as ‘The blonde girl, who is wearing green trousers, looks a bit out of place’ or ‘The blonde girl, because she is wearing green trousers, looks a bit out of place’). In our analysis, we have focused on those instances that were unambiguous free adjuncts. That does not mean, however, that these participial relative clauses cannot be a source of free adjuncts (see Section 4.1); accordingly, it would be worthwhile examining whether the inclusion of non-restrictive participial relatives in our counts would affect our results.
22
Lauren Fonteyn and Hubert Cuyckens
Finally, absolute constructions (as in 12),12 non-clausal uses of the present participle (as in 13),13 ‑ing forms with prepositional or conjunctional value (as in (14)‒(17)), and formulaic expressions (as in (18)) were not included in our data set, either: (12) The resistance of the air having been avoided, (…) all fall exactly in the same time. (1859, PPCMBE) (13) Sir William, laughing, departed. (1608, PPCEME) (14) Seeing that the general tendency of oil is to spread over the surface of water, why does it not do so in this case? (1890, PPCMBE) (15) Notwithstanding the numbers of market growers who now send produce to the London and provincial markets, it is astonishing to see the enormous quantities of fruits, flowers and vegetables that are imported from the Continent and the Colonies. (1913, PPCMBE) (16) Zij waren, niettegenstaande hun solide leven, drie meiden en rijtuig zuinig. (1901, CCHD) ‘They were, notwithstanding their solid life, three maids and a vehicle short.’ (17) Aangaande de Vertaling der Bastaard-Woorden, de Eensvormigheyd der spelding, de Accenten, en wat meer van dat soort is, daarvan spreek ik by den Weg van de (…)(1720, CCHD) ‘Considering the translation of foreign words, the unity of form of spelling, the accents, and what more of that sort is, thereof I speak by means of the poet (…).’ (18) I, God willing, will not come home without doing somewhat. (1660, PPCEME)
12 Pointing out that, especially in Early Modern English, the distinction between the two construction types is far from clear-cut, Río-Rey (2002) suggests that “the unambiguous difference that now exists (at least in written language) between free adjuncts and absolutes must have been reached later than 1700”. However, while even in Present-Day English, “related” free adjuncts, i.e., free adjuncts that are controlled by the matrix clause, are the default case (Kortmann 1991: 8), relatedness is not obligatory: in a small group of free adjuncts the subject is not expressed in the matrix clause. Therefore, Río-Rey suggests looking at the constructions in terms of a relatedness cline. In this paper, we will study unrelated as well as related subjectless converb clauses. In particular, we follow the approaches by Kortmann (1991) and by Rinkel (1989), who, in a detailed study of Dutch participle constructions in the seventeenth century, argues that the absence of the subject in such constructions indicates a tighter bond with the matrix clause and that they are thus still distinct from absolute constructions. 13 The non-clausal uses of the present participle will not be included in our data, because they still occur in all Germanic languages, while the clausal type under investigation here has become marginal in all Germanic languages aside from English (Killie 2006, 2007; Killie and Swan 2009). Including this non-clausal construction type could compromise the frequency results of adverbial or adverbial-like clauses, especially in the later stages of Dutch.
The development of free adjuncts in English and Dutch
23
3 Data analysis After manual inspection, the relevant attestations were coded semantically (in terms of their adverbial[-like] meaning) and syntactically (in terms of word order). This coding was informed by Killie and Swan’s (2009) semantic and syntactic categorization of free adjuncts, thus allowing us to reassess their results (see below). Semantically, our data set was divided into three subtypes. The first group comprises free adjuncts expressing addition/accompanying circumstance (add/ acc) or exemplification/specification (ex/spec). Clauses denoting an addition/ accompanying circumstance, rather than providing details with regard to the circumstances (time, cause, conditions, etc.) of the matrix proposition, add an event/state which is distinct from the event or state in the matrix clause (Kortmann 1991). The resulting complex sentence thus subsumes two situations or events that occur side by side, with no constraints on, or relations between, the two events. The only requirement is that both events take place at approximately the same time and place. This is illustrated in (19), where the situation expressed in the free adjunct is merely added to the event in the matrix clause (the people pass through the town and eat and drink there much and merrily): (19) People, after all, would not be greatly inclined to take up their quarters in a town encompassed by mediaeval walls and towers, when sunny watering places abounded near by, though they pass through it in thousands, eating and drinking much and merrily in its old-fashioned hostelries. (1905, PPCMBE) Free adjuncts of the exemplification/specification type, on the other hand, do not introduce a new event, but simply restate or clarify the information expressed in the matrix clause. Ex/spec clauses can typically be glossed as for example/e.g., that is/i.e., in that, namely/viz., in particular, more exactly, in fact, indeed, etc. (Kortmann 1991: 167; Killie and Swan 2009: 341). In (20), for instance, saying in the free adjunct makes the asking event in the matrix clause more specific: (20) The Jews therefore did not believe concerning him, that he had been blind, and had received his sight, until they called the parents of him that had received his sight, and asked them, saying, Is this your son, who ye say was born blind? (1881, PPCMBE) The second semantic subtype consists of temporal free adjuncts expressing temporal background, as exemplified in (21): (21) So thereupon, standing before him, Chillon John, from whom he had the news, bent to him slightly. (1895, PPCMBE)
24
Lauren Fonteyn and Hubert Cuyckens
The third semantic category comprises all adverbial relations that are related to the notion of cause, such as cause proper, reason, condition, and concession: (22) Thinking it very probable that Lord Hardwicke will come here with his new marry’d son and daughter, I would wish to have my Dear Friends opinion whether I ought or ought not to speak to him. (1763, PPCMBE) Finally, free adjuncts that turned out to be semantically indeterminate were classified as “indeterminate”. It will be argued that these free adjuncts served as bridging contexts in their semantic development. Syntactically, the free adjuncts were coded for position in the sentence: initial, medial, or final, as in (23): (23) a. Walking from Ratley to London, ther was a man bwilding up the bulwark of a dytch with the (...) of ox horns to whom after salutacions, sir, quoth I, wil those quycksets (‘cuttings’) groe think you? (1582, PPCEME) b. (...) after which time, despairing of joining us any more, they proceeded to Van Diemen’s land. (1776, PPCMBE) c. She flounced from him, holding her arms up to the lady. (1895, PPCMBE) The results of the semantic and syntactic coding of the free adjuncts in English and Dutch, across the different periods, is represented in Tables 3‒6. Table 3: Semantic coding of the English free adjuncts for each subperiod: Absolute and normalized frequencies per 10,000 words English
Add/acc
Ex/spec
M1 M2 E1 E2 L1 L2 Total
14 (1) 4 (0.29) 35 (5.63) 11 (1.73) 27 (9.45) 9 (3.15) 20 (9.14) 9 (3.92) 12 (4.82) 4 (1.6) 13 (4.81) 11 (4.07) 112 46
Temporal 2 (0.14) 7 (1.13) 10 (3.5) 11 (4.79) 15 (6.02) 10 (3.7) 54
Causal 0 5 (0.8) 8 (2.8) 15 (6.53) 21 (8.43) 10 (3.7) 57
Indeterminate 6 (0.43) 10 (1.6) 3 (1.05) 5 (2.18) 1 (0.4) 4 (1.48) 29
Total 26 (1.86) 68 (10.89) 57 (19.95) 60 (26.56) 53 (21.27) 48 (17.77) 298
Table 4: Semantic coding of the Dutch free adjuncts for each subperiod: Absolute and normalized frequencies per 10,000 words Dutch
Add/acc
Ex/spec
Temporal
Causal
M1 M2 E1 E2 L1 L2 Total
3 (0.43) 16 (6.09) 18 (12.56) 16 (10.52) 17 (5.37) 18 (2.88) 89
0 1 (0.38) 3 (2.09) 5 (3.29) 4 (1.26) 6 (0.96) 19
0 1 (0.38) 4 (2.79) 7 (4.6) 15 (4.74) 12 (1.92) 39
0 1 (0.38) 6 (4.19) 10 (6.58) 25 (7.9) 18 (2.88) 60
Indeterminate 0 0 5 (3.49) 5 (3.29) 5 (1.59) 6 (0.96) 21
Total 3 (0.43) 19 (7.23) 36 (24.41) 43 (28.27) 66 (21.18) 60 (9.61) 227
The development of free adjuncts in English and Dutch
25
Table 5: Syntactic coding of the English free adjuncts for each subperiod: Absolute and relative frequencies English M1 M2 E1 E2 L1 L2
Initial position
Medial position
Final position
2 (7.7%) 3 (4.4%) 11 (19.3%) 18 (30%) 23 (43.4%) 14 (29.2%)
5 (19.2%) 19 (27.9%) 15 (26.3%) 13 (21.7%) 8 (15.1%) 13 (27.1%)
19 (73.1%) 46 (67.6%) 31 (54.4%) 29 (48.3%) 22 (41.5%) 21 (43.8%)
Table 6: Syntactic coding of the Dutch free adjuncts for each subperiod: Absolute and relative frequencies Dutch M1 M2 E1 E2 L1 L2
Initial position
Medial position
Final position
0 2 (10.5%) 6 (16.7%) 11 (25.6%) 25 (37.9%) 19 (31.7%)
0 1 (5.3%) 2 (5.6%) 3 (7%) 6 (9.1%) 2 (3.3%)
3 (100%) 16 (84.2%) 28 (77.8%) 29 (67.4%) 35 (53%) 39 (65%)
4 Historical development: Strolling along the same path 4.1 Semantic change Let us consider Figures 1 and 2, which represent, in bar chart format, the distribution of free adjuncts in English and Dutch across its various semantic subtypes and across its various subperiods (frequencies are normalized per 10,000 words). As can be seen from Figures 1 and 2, up until the first half of the Late Modern Period (L1), English and Dutch show a parallel distribution of its semantic sub types. The semantic types “add/acc” and “ex/spec” largely make up the category of free adjuncts in Middle English and Middle Dutch (periods M1 and M2), showing similar frequencies. In the periods E1 and E2, as well as in L1, add/acc and ex/ spec take up almost an identical share of the total number of free adjuncts in English and Dutch. Examples of add/acc and ex/spec free adjuncts in English and Dutch are in (24) and (25), respectively.
26
Lauren Fonteyn and Hubert Cuyckens
14 12.62
12.6
12 10
8.88
8
8.03
7.4 6.09
6
6.43
3.7
4 2 0
1.3 0.14 0.43 0
0.97 1.13 1.61
M1
1.75 3.5
M2
6.02
4.79
3.7 2.18
1.05
E1
add/acc & ex/spec temporal causal indeterminate
1.48
0.4
E2
L1
L2
Figure 1: Semantic analysis of English free adjuncts
16
14.65 13.87
14 12 10 8 6.37
5.92
6
0
7.27
3.84
3.49
4 2
6.63
add/acc & ex/spec temporal causal indeterminate 2.56
0.58
0 0 0
M1
0.38 0 0
M2
2.09
E1
3.49
4.6
3.29
4.74 1.58
E2
L1
1.92 0.96
L2
Figure 2: Semantic analysis of Dutch free adjuncts
(24) a. The ermyte partyd fro hir with heuy cher & came hom a-ʒeyn to Lynne, excusing hir to hir confessowr & to oðer frendys, (…) (1592–1603, PPCME2) (addition)
The development of free adjuncts in English and Dutch
27
b. Sche answeryd, seying to hym a-ʒen, “Sir, I hope ʒe Xal be a seynt ʒowrselfe & euery man þat Xal come to Heuyn”. (c1450, PPCME2) (elaboration/ specification) (25) a. Die daar houden dat zy den menschen angheboren werden bouwen zulcke hare meyninghe op henluyder ondervinden, zegghende dat men twerck boven twoord ende tverzoecken boven tvermoeden schuldigh is te geloven. (1585, CCHD) (addition) ‘Those who hold that they are inherent traits of mankind build their opinion on their finding, saying that one is guilty to believe the work above the word and the suggesting above the suspecting.’ b. Dairtegens Jacop Jansz (…) mit synen vrienden ende magen int cort op gheantwort heeft, seggende dat hy (…) vander voirsz saiken all ende geheelic ontschuldich wair (…) (1460, CCHD) (specification) ‘To which Jacob Jansz (…) with his friends and relatives shortly ed, saying that he (…) was all and totally innocent of the answer abovementioned things (...)’ From M2, and especially from E1, the range of semantic types expanded, with the emergence of temporal and causal free adjuncts. With regard to these adjuncts, we observe, firstly, a sharp rise in the frequency between M2 and E1 in both languages: in Dutch, temporal free adjuncts increase from 0.38 to 2.09 (per 10,000 words) and causal ones from 0 to 3.49 (per 10,000 words). This increase is even more pronounced than in the English data, where the temporal free adjuncts rise in frequency from 1.13 to 3.49 and causal ones from 0.97 to 1.75. In both languages, this parallel increase in frequency continues up across the E2 and L1 periods. Significantly, these results go against Killie and Swan’s (2009: 338, 360) contention that only English, not other Germanic languages, saw an increase of temporal and causal clauses in the period under investigation. Our results further suggest that the English and the Dutch free adjuncts start diverging only by the end of the Late Modern period, when there is a general decrease of Dutch free adjuncts, and most markedly in the group of add/acc and ex/spec clauses. How can we account for the semantic expansion of free adjuncts? As was already observed, almost all instances of free adjuncts in periods M1‒M2, up until 1500, have ex/spec or add/acc meaning. From 1500 onwards, however, free adjuncts in both languages expanded, adopting new adverbial meanings such as temporal simultaneity or causality. The main source for this reanalysis was the inherent semantic ambiguity of add/acc and ex/spec clauses (Killie and Swan 2009: 344‒347): because of their unity in time and place with the matrix clause, the additional events in add/acc free adjuncts were presumably interpreted as providing the temporal background for or a causal relationship with the event
28
Lauren Fonteyn and Hubert Cuyckens
expressed in the main clause. Ex/spec clauses, on the other hand, tend to show ambiguity between an ex/spec and an instrumental reading, as in (28): (26) Thare-fore I turnede me to Gode, and with my mynde I said , “A, Jeshu, how precious es thi blude!”, makand ðe crosse with my fyngere in my breste. (c1440, PPCME2) (add/acc or temporal simultaneity) (27) But, for the method of studying the spelling lessons, I am indebted to Dr. Bell, believing it was his peculiar invention. (1806, PPCMBE) (add/acc or causal) (28) Ðe fende trubled me, feʒtand alday oʒayns me. (c1350, PPCME2) (ex/spec, instrumental) ‘The devil made me suffer, fighting all day against me.’ There are many indeterminate instances such as the ones shown in (26) to (28), which can be assumed to have served as “bridging contexts”, i.e., contexts which are semantically and pragmatically ambiguous (Heine 2002; Killie and Swan 2009: 347). In (26), for instance, it is unclear whether the information in the free adjunct is meant to provide additional information or serves as a temporal frame, expressing that the speaker said something while making a cross. In turn, temporal free adjuncts could then be further reanalyzed as having a causal meaning (see, for instance, Traugott and König 1991: 194‒199). In (27), the free adjunct is ambig uous between an add/acc reading, merely adding an accompanying feeling, and a causal reading, expressing the reason why the speaker feels he or she is indebted to Dr. Bell. In (28), the free adjunct further elaborates/specifies the situation of troubling expressed in the matrix clause. Yet, its potential instrumental reading becomes evident, as a paraphrase with by is possible as well (‘The devil makes me suffer by fighting all day against me’). As these same types of implicature have occurred over and over again during the history of English, they “may well have been conventionalized (or ‘semanticized’)”, i.e., “become a part of the semantics of the clause combining relationship” (Killie and Swan 2009: 348‒349). Such bridging contexts are also found in the Dutch data, as is shown in (29) and (30): in (29), the free adjunct denotes additional information as well as temporal simultaneity; in (30), it provides a temporal as well as a causal backdrop for the matrix clause. (29) Hy kon zijn dartelheyd ten lesten niet bedwingen, Of hy begon met vreughd dit liedeken te singen, Al staende by het vyer, (…) (1649, CCHD) ‘Finally he could not suppress his playfulness, and he began to sing this song with joy, standing by the fire, (…).’ (30) (…) en ziende dat mijn vader (…) een weinig begost te bedaren, en zijn arm die hy al opgeheven had, om haar verdiende loon toe te passen, wederom zinken liet, sprak zy Aldus verder tegen hem: Wat vertoeft gij, wreede, myn kuis en onnozel hert (…) te doorboren en aan uw’ bloeddorst op te offeren (…) (1695, CCHD)
The development of free adjuncts in English and Dutch
29
‘(…) and seeing that my father started to relax a little, and again lowered his arm, which he had already raised to give her what she deserved, she thus continued to speak to him: How dare you, cruel one, to pierce my pure and innocent heart and sacrifice it to your thirst for blood (…).’ Another source of the temporal and causal free adjuncts may be the participial relative clause, as there is a great deal of indeterminacy between a non-restrictive relative clause reading and a temporal, or even causal reading (see also note 11). The relative clause in (31), for instance, can be interpreted as ‘who remembered’, ‘when he remembered’, or ‘because he remembered’: (31) The foole, remembering how his head was, strikes it up, and hits the fellowe’s mouth with the pitcht place, so that the haire of his head and the haire of the clowne’s beard were glued together. (1608, PPCEME) Killie and Swan (2009: 342) observe that such indeterminate uses are highly frequent, especially around the sixteenth century. Like the free adjunct, the participial relative clause in general became increasingly used from the fifteenth century onwards, providing a compact means of determining a noun phrase as precisely as possible (Kohnen 2004). The ambiguity of an increasing number of relative clauses in which the relative pronoun is omitted, Killie and Swan (2009: 354‒357) argue, has given rise to new reinterpretations, ultimately contributing to the semantic diversification of the free adjunct. Interestingly, such indeterminate relative clauses are found in Dutch as well, as shown in (32), where the relative clause may also express causal meaning:14 (32) Het steedeken, hierdoor open staande, werd geplondert en tot asch gemaakt. (1642, CCHD) ‘The little city, standing open because of this, was plundered and made to ash.’ Through these various bridging contexts, the free adjunct gradually came to express a wider variety of adverbial meanings in both English and Dutch, as for instance in (33): (33) Doch nu door ondervinding merckende wat hem aan kennis der taalen gelegen was om in de kunst te vorderen, pooghde hy dat gebrek te boeten. (1682, CCHD) (causal)
14 As suggested by van der Horst, these participial relative clauses with additional adverbial meaning in Dutch functioned as the source for the complexe voorbepaling (2008: 1584‒1585). For further discussion see Section 5.3.
30
Lauren Fonteyn and Hubert Cuyckens
‘Though now through experience noticing what knowledge of language he still needed to progress in art, he attempted to compensate for the shortcoming.’ The results of our semantic analysis show that the Dutch free adjunct went through a process of semantic expansion that is very similar to that of the English free adjunct, both qualitatively (in terms of semantic types) and quantitatively (in terms of frequency). While Middle Dutch and English free adjuncts were overwhelmingly of the add/acc or ex/spec type, temporal and causal free adjuncts gradually came into use from the Early Modern period onwards.
4.2 Syntactic change Killie and Swan (2009) suggest that the semantic changes that free adjuncts underwent were accompanied by a number of syntactic changes. While the original add/acc free adjunct clauses were semi-coordinate, free adjuncts, after reinterpretation as temporal and causal adverbial clauses, also underwent a change in syntactic status: their non-finite and semantically underspecified status (due to their lack of conjunctions), Killie and Swan (2009) argue, ultimately led to their reanalysis as fully subordinate clauses. This change in syntactic status is reflected in a positional shift. Once they had acquired subordinate status, “[free adjuncts] came to serve the same types of textual functions as subordinate clauses typically serve. Important functions here are the reader-orienting and text cohesive functions, which are typically associated with initial position” (Killie and Swan 2009: 347). In what follows, we will show that neither the positional shift to initial position nor the presumed increased degree of subordination is exclusive to English temporal and causal free adjuncts. Figures 3 and 4 show the proportion of free adjuncts in initial or preverbal position (see (34)), as opposed to free adjuncts in final or postverbal position (see (35)), for English and Dutch, respectively. (34) Turend naar den strakken horizont begon hij te denken (…) (1904, CCHD) ‘Staring at the fixed horizon, he began to think (…).’ (35) Veertig meters hoog rijzen de slanke zuilen uit den grond, schijnbaar niets anders dragend dan eene smalle kroonlijst. (1879, CCHD) ‘The slender columns rose forty metres from the ground, seemingly carrying nothing but a narrow cornice.’ Figures 3 and 4 show that in period M2 only 3 free adjuncts occur in initial osition in English (4.4%) and only 2 in Dutch (10.5%). Importantly, by L1 (the p first Late Modern English period), both languages have seen a remarkable and
The development of free adjuncts in English and Dutch
31
100% 90% 80% 30
70% 60% 50%
46 24
42
34 Final
65
Inial
40% 30% 23
20% 10% 0%
11 2 M1
3 M2
E1
18
E2
14
L1
L2
Figure 3: Diachronic development of the position of the English free adjunct
100% 90% 80% 70%
41
60% 50%
3
17
30
32
41 Final Inial
40% 30% 20%
25
10% 0%
0 M1
2 M2
6 E1
11
E2
L1
19
L2
Figure 4: Diachronic development of the position of the Dutch Free adjunct
comparable increase of free adjuncts in initial position: in E1: 19% (English) and 18% (Dutch); in E2: 30% (English) and 28% (Dutch), and in L1 43.4% (English) and 37.9% (Dutch). In the final period (L2), both languages show a slight decrease in sentence-initial free adjuncts, but overall it appears that free adjuncts became increasingly acceptable in initial position in both English and Dutch.
32
Lauren Fonteyn and Hubert Cuyckens
Comparable parallels between English and Dutch were also observed with regard to the syntactic integration of the free adjunct into the matrix clause. As mentioned above, Killie and Swan (2009) do not regard add/acc and ex/spec clauses as prototypical instances of subordination, “but rather as exemplars of a clause linkage type which lies in between subordination and coordination” (2009: 344). In light of the fact that clause combining is scalar in nature (Callaway 1901: 268–269; Lehmann 1988; Hopper and Traugott 2003: 176–184; Killie and Swan 2009: 344), and that clause combinations can be positioned on a scale from less to more integrated or subordinate, Killie and Swan (2009) argue that free adjuncts in English – but not in other Germanic languages – have become increasingly inte grated.15 Our analysis of Dutch free adjuncts, however, points to syntactic integration parallel to that in English. Dutch, as König and van der Auwera point out, has “‘Topic-Finite verb-…,’ otherwise known as ‘Verb Second’ or simply ‘V-2’ as a basic word order in independent sentences” (1988: 102). The structure and basic word order of Dutch independent sentences can be represented as follows: Pre-core
Initial
slot,
position
1st pole Middle
(2nd
Final Position
pole) [verb]
, Post-core slot
[verb]
Ze
heeft
een vriend
geholpen met zijn klusjes.
She
has
a friend
helped
with his chores.
Core Sentence Figure 5: Structure and basic word order of Dutch independent sentences
The core sentence consists of two poles: the first pole is the finite verb form, and the second pole is formed by any other verb form(s) (if there are any). Constituents can be placed in Initial position (occupying the slot immediately before the first pole; usually the subject is placed here), in Middle position (in between the two poles; an inverted subject would occur here), or in Final position (following the second pole). Interestingly, the Initial position can only be occupied by one constituent that is part of the core sentence. This particular feature of Dutch provides “a clear signal for the grammatical incorporation (…) of one clause into another” (König and van der Auwera 1988: 102). Indeed, if a clause “is followed
15 While Killie and Swan (2009) do not provide any explicit syntactic evidence for this claim, it is possible to test the degree of syntactic integration by means of Cristofaro’s (2003: 29‒40) syntactic integration tests. If these tests are applied to free adjuncts, it appears that (a subgroup of) causal and temporal free adjuncts do show a higher degree of syntactic integration than add/ acc and ex/spec ones.
The development of free adjuncts in English and Dutch
33
by the finite verb of another, the former can generally be taken to function as a constituent within the other” (König and van der Auwera 1988: 102), i.e., the first clause can be argued to be fully integrated into the core sentence as it occurs in the Initial slot. This is the case in, for instance, example (36): hij ziek is, kan Fred niet meekomen. (36) Omdat Because he sick is, can Fred not come-along. ‘Because he is sick, Fred cannot come along.’ (König and van der Auwera 1988: 102) Some clause types, however, such as the concessive ook al-clause in (37), initiate a sentence but are not followed by the finite verb; hence, they cannot occupy the Initial core position, but only the pre-core slot, and, in turn, they cannot be said to be fully integrated into the core sentence: (37) Ook al help je hem, hij zal het werk niet afkrijgen. Even if help you him, he will the work not finish. ‘Even if you help him, he will not be able to finish the work.’ (Van Belle 2003: 54) Most of the new Dutch free adjuncts with temporal or causal meaning, which were seen to emerge in period E1 and which became increasingly frequent over time (cf. Figure 2), appear sentence-initially. In fact, the sentence-initial free adjuncts in our Dutch data in Figure 4 appear to be exclusively temporal or causal, while add/acc and ex/spec free adjuncts take up the larger part of the sentence-final free adjuncts. Importantly, the sentence-initial free adjuncts do not occur in the pre-core slot, which, like the post-core slot, is located outside the main clause, but take up the initial position within the core sentence, as can be seen from the inversion they cause in the matrix clause (Figure 6). As such, temporal/causal adjuncts differ from the earliest types of Dutch free adjuncts, i.e., add/acc and ex/spec clauses, which can be regarded as intermediate between coordinated and subordinated clauses in that they without exception occur in the post-core slot. Words and clauses in the post-core slot are always positioned sentence-finally, they are separated from the rest of the sentence by a comma, and they do not belong to the PreC, I Wetend dat hij niet
1st
M
hebben ze de vergadering
2nd
F
, PostC
beëindigd.
meer zou komen, Knowing that he would have
they the meeting
ended.
not come anymore, Core Sentence Figure 6: Structure and basic word order of Dutch sentence with sentence-initial free adjunct
34
Lauren Fonteyn and Hubert Cuyckens
PreC, I Ze
1st
M
heeft hem
She has
him
2nd
F
geholpen met zijn huiswerk helped
, PostC , zeurend dat hij niets alleen kan.
with his homework , nagging that he cannot do anything by himself.
Core Sentence
Figure 7: Structure and basic word order of Dutch sentence with sentence-final free adjunct
core sentence (Figure 7). Thus, it appears that, just as in English, free adjuncts in Dutch have become more integrated into the main clause over time.
5 The nineteenth century: Going their separate ways In the previous section, we showed that the semantic and syntactic changes that affected the English free adjunct in Middle and Early Modern English were by no means unique, and that the Dutch free adjunct underwent a strikingly similar development. There is no denying, however, that only English managed to retain the free adjunct construction. A closer look at a set of free adjuncts taken from four nineteenth-century novels and their Dutch translations (published in the late nineteenth or early twentieth century)16 makes clear that in this period, the free adjunct was no longer as common in Dutch as it was in English. Of the 142 English free adjuncts, 38 are attested in Jane Austen’s Sense and Sensibility, 35 in Charles Dickens’ A Christmas Carol, 43 in Harriet Beecher Stowe’s Uncle Tom’s Cabin, and 26 in Mark Twain’s Tom Sawyer. Table 7 lists the frequencies of the Dutch translational equivalents, namely, free adjuncts, coordinated clauses, or finite clauses introduced by a conjunction. In three cases the translator did not use any of these translational alternatives, but merely started a new sentence to add the information contained in the English free adjunct. These instances were added to the coordination group and are placed between brackets. When a free adjunct was omitted in the translation, or when it was translated by means of a non-finite clause introduced by the preposition met ‘with’ or a verbless nonfinite clause, it has been placed in the category “other”. In the following examples, the (a) sentences provide the original, and the (b) sentences the translation: ( 38) Translation by a free adjunct a. As the last stroke ceased to vibrate, he remembered the prediction of old Jacob Marley, and lifting up his eyes, beheld a solemn Phantom, draped
16 The exact publication dates of the translations are: Sense and Sensibility (1922), Uncle Tom’s Cabin (1918), The Adventures of Tom Sawyer (1909), and A Christmas Carol (1905).
The development of free adjuncts in English and Dutch
35
Table 7: Translational equivalents in nineteenth-century novels Austen English Dutch
Free adjunct Free adjunct Coordination Finite adverbial clause Other
38 15 5(2) 11 4
Dickens 35 26 1 (1) 3 3
Beecher Stowe
Twain
Total
43 17 13 7
26 6 3 5
142 (100%) 64 (45.07%) 25 (17.61%) 29 (20.42%)
5
12
24 (16.9%)
and hooded, coming like a mist along the ground, towards him. (1843, A Christmas Carol) b. Toen de laatste slag opgehouden had te trillen, herinnerde hij zich de voorspelling van den ouden Jacob Marley, en zijne oogen opslaand, zag hij een zeer ernstig spook, in een kleed gehuld, met den kap over zijn hoofd geslagen, als een mist langs de grond glijdend, op zich toe komen. (39) Translation by a coordinate structure a. He actually made me feel of his fist, which was like a blacksmith’s hammer, or a nodule of iron, telling me that it was “calloused with knocking down niggers”. (1852, Uncle Tom’s Cabin) b. Hij liet mij werkelijk zijne vuist voelen, die naar een smidshamer of een klomp ijzer gelijk, en zeide mij dat die vereelt was door het neerbeuken van negers. (lit. ‘… and said to me that it was calloused by the knocking down of niggers.’) (40) Translation by a finite clause introduced by a conjunction a. “You may believe how glad we were to see them,” added Mrs. Jennings, leaning forward towards Elinor, (…) (1811, Sense and Sensibility) b. “Je begrijpt, hoe blij we allen waren hen te zien,” voegde mevrouw Jennings erbij, terwijl ze zich naar Elinor vooroverboog, (…) (lit.‘… while she leaned herself over to Elinor.’) (41) Translation by a verbless clause a. A log raft in the river invited him, and he seated himself on its outer edge and contemplated the dreary vastness of the stream, wishing, the while, that he could only be drowned, all at once and unconsciously (…) (1876, The Adventures of Tom Sawyer) b. Op een in de rivier liggend stuk hout van een houtvlot zette hij zich neder en beschouwde den somberen, onafzienbare stroom, met het verlangen van op eens door dezen verzwolgen te worden (…) (lit. ‘with the desire ...’)
36
Lauren Fonteyn and Hubert Cuyckens
(42) Free adjunct omitted a. So they stood, each with a foot placed at an angle as a brace, and both shoving with might and main, and glowering at each other with hate. (1876, The Adventures of Tom Sawyer) b. Zoo stonden zij beiden met één voet vooruit, elkander duwende dat het een aard had. The data show that even though, in the contemporaneous translations of the English novels, free adjuncts are still the most natural translational equivalent, only 45.07% of the English free adjuncts have been translated by Dutch free adjuncts, and this percentage would be even lower if the Dickens translation was excluded.17 In other words, 54.93% of the English free adjuncts have been translated by means of other translational equivalents, mostly by coordinate structures or finite adverbial clauses. In the remainder of this paper, then, we will zoom in on the question why Dutch eventually lost its free adjunct from the nineteenth century, and why the construction was retained in English. 5.1 Language-external factors: Prescriptivism In an attempt to explain why the Norwegian free adjunct was lost while the English one was retained, Killie (2006) suggests that the most important factor contributing to the differential development of the two languages was sociolinguistic in nature, in particular, that it was the result of the actions of purist movements. As a result of the union of Norway with Denmark from 1380 to 1814, Norwegian was substantially influenced by Danish. From the seventeenth century onwards, Danish purists claimed that the only natural use of the present participle was either as a noun or as an adverbium qualitativis, as in He comes running (Killie 2006: 453). In the e ighteenth century, purists attempted to create a “pure” Danish that was freed from Latin influence; more specifically, “they wanted to get rid of the ‘heavy’, hypotactic Latin style and establish a ‘natural’ Danish style instead”, based on parataxis (Killie 2006: 454). Accordingly, as a result of the strong influence of Danish on Norwegian and the continuing, extensive use of Danish in Norway after 1814, Norwegian went “through a long-lasting purification
17 Occasionally, the Dutch translations used free adjuncts where English did not; however, we did not systematically check for these instances.
The development of free adjuncts in English and Dutch
37
process” (Killie 2006: 459). Its speakers thus came to prefer parataxis and finite subordinate clauses instead of participial clauses. The English purist movement, on the other hand, addressed different issues: while the English purists fought overuse of Latin and Greek vocabulary, they “generally seem to have held Latin syntax, rhetoric and style in high esteem, finding that English had a great deal to learn from this language in terms of elegance and logic” (Killie 2006: 459). Hence, the Latinate clause patterns were never really contested. Like the English purists, Dutch purists did not reject the use of the free adjunct construction, as they were mainly concerned with issues such as the influence of Latin and Greek on the lexicon and on case marking (van der Wal and van Bree 2008: 195‒199). There was resistance to some types of present participle free adjuncts, but the free adjunct itself was never really contested. For instance, in the seventeenth century, Joannes Vollenhove (1678) rejected the redundant use of the present participle zijnde ‘being’ in combination with an adjective or a past participle, as in (43): (43) Hier op vernoeght (zynde), aanviert hy de reiz (…) (1574, CCHD) ‘(Being) pleased with this, he accepted the trip (…)’ While Vollenhove (1678) thus opted for a more economical solution, namely the adjectival free adjunct, he did not altogether reject the hypotactic style. Later, in the nineteenth century, two important prescriptivist works pleaded against the use of the absolute construction but not free adjuncts. In Anonieme Syntaxis (likely written by Matthijs Siegenbeek), the author tentatively suggests that the absolute construction be avoided and replaced by a finite clause introduced by a conjunction: Intusschen houden wij, zonder de gemelde spreekwijze [i.e., de oorlog verklaard zijnde] geheel te durven afkeuren, het daarvoor, dat men best doet, met dezelve, zoo veel mogelijk, te vermijden, en zich ten dien van eene omschrijving te bedienen, b.v. nadat de oorlog verklaard was enz. (Anonymous 1810: 33). [In the meanwhile, without entirely rejecting the previously mentioned construction (i.e., de oorlog verklaard zijnde ‘the war being declared’), we suggest that it is best to avoid it as much as possible and use a paraphrase, e.g., nadat de oorlog verklaard was ‘after the war had been declared’ etc.] (translation ours)
David (1853: 163–164), while basing his ideas on Anonieme Syntaxis, takes a slightly more rigid stance towards the absolute construction, calling it not inherent to the language. Strikingly, David (1853) immediately adds that the subjectless
38
Lauren Fonteyn and Hubert Cuyckens
present participle construction, i.e., the free adjunct, should not be avoided, as it is language inherent: Wanneer in dergelyke gezegden het deelwoord betrekking heeft tot het onderwerp van den volgenden zin, alsdan is het verband duidelyk, en de spreekwyze volkomen nederduitsch. B.v. groot geworden zynde, of groot geworden, bleef hy even ziekelyk. (David 1853: 164) [When in such constructions the participle concerns the subject of the following sentence (matrix clause), then the relation is obvious, and the construction is entirely Dutch. For instance, groot geworden zynde, or groot geworden, bleef hy even ziekelijk (‘Having grown up, or grown up, he remained sickly’)] (translation ours)
As the free adjunct was not rejected by Dutch purists and prescriptivists, it seems unlikely that prescriptivism had a direct influence on its decrease. It could, however, be argued that the rejection and subsequent loss of redundant zijnde and the loss of the absolute construction indirectly influenced the frequency of the free adjunct because the total number of clausal V‑(e)nd(e) constructions decreased. As will be suggested in Section 5.3 below, if one type of clausal V‑(e)nd(e) construction disappears from a language, it can no longer strengthen the development of similar clausal V‑(e)nd(e) constructions in their development. Still, when it comes to the decrease of the free adjunct, there seems to be more at stake than just the indirect influence of prescriptivism.
5.2 Language-internal factors influencing the use of the free adjunct In line with Killie’s (2006, 2007) claims on the success of the English free adjunct, we would like to argue that it is linguistic or language-internal rather than language-external (sociolinguistic) factors that were most decisive in leading to the loss of the Dutch free adjunct. In Middle English, the present participle suffix changed from ‑(e)nd(e) to ‑ing, thus coinciding with the marker of the gerund; this made the ‑ing suffix “a highly polyfunctional form, with nominal, adjectival, adverbial, prepositional and verbal uses” (Killie 2006: 462). Moreover, these ‑ing forms were increasingly used with verb-like properties; the progressive, the free adjunct (adverbial and relative clauses) and the gerund all took adverbials and an object. Killie notes that “interestingly, all these categories experienced a huge growth in frequency at [approximately] the same time” (2006: 463) and “it seems likely that they have influenced each other” (2006: 463). The precise relations between these various constructions and the direction of the influence between them are difficult to trace, but Killie (2006) argues that there is no doubt that there has been such influence.
The development of free adjuncts in English and Dutch
39
Killie’s account of the success of the English free adjunct in terms of “mutual influence” becomes particularly plausible when the situation of Norwegian is considered, for in Norwegian the free adjunct never became a frequent means of clause combining. While Old and Middle Norwegian (ca. 1000‒1550) also had particular clausal V‑ende constructions in which the present participle was used in a “progressive-like manner, also with transitive verbs” (Killie 2006: 466), Norweg ian never seems to have had a gerund-like construction. Moreover, the supposed Norwegian V‑(e)nd(e) progressive never grammaticalized in the way the English progressive did. There were, in other words, no Norwegian clausal V‑(e)nd(e) constructions that could have caused participles ending in ‑(e)nd(e) to acquire verbal properties. This relatively unstable status of the Norwegian free adjunct combined with the impact of the purist movements through the nineteenth century may therefore have led to the demise of the construction in Norwegian. While Dutch resembles Norwegian in that it never had a gerund-like construction,18 it used to have a progressive V‑(e)nd(e) construction, whose development shows remarkable resemblances to that of its English counterpart: (44) (…) dat wi midts onse sonden die doot verwachtende sijn in deser zee. (DBNL, ca. 1500) ‘(…) that we with our sins are expecting death in this sea.’ In Section 5.2.1, we will look at how the loss of the Dutch progressive might have triggered the decrease of Dutch free adjuncts. Section 5.2.2 will provide a more accurate view of the role which the absence of a gerund-like V‑(e)nd(e) con struction might have played in this process, and, by extension, when and how the English gerund may have impacted the English free adjunct.
5.2.1 The Dutch progressive In Dutch, the progressive construction experienced a dramatic growth in frequency around the sixteenth century: Belangrijker verandering in de zestiende eeuw is waarschijnlijk echter de sterke toename in het gebruik van tegenwoordige deelwoorden in combinatie met zijn, worden, blijven, enz. ter uitdrukking van het voortduren van een handeling of toestand (…). Het ziet ernaar uit dat zulke combinaties (niet alleen met zijn, maar in zekere mate ook met worden en blijven)
18 Dutch does have nominal uses of the present participle, as for instance the agentive nominal in Terwyl ik dus als eene zoekende (...) myn wegje zo voort ging (van der Horst 2008: 1425), meaning ‘while I thus continued my ways as a seeking person’.
40
Lauren Fonteyn and Hubert Cuyckens
aan het einde van de zestiende eeuw sterk toenemen in frequentie. Dat betekent dat er een opmerkelijke parallel is met de ontwikkelingen in het Engels, waar evenzeer de zg. progressive in de zestiende eeuw sterk toeneemt in frequentie (van der Horst 2008: 863). [A more important change in the sixteenth century is probably the strong increase in the use of present participles combining with zijn ‘to be’, worden ‘to become’, blijven ‘to stay, remain’, etc., to express a continued action or condition (…). It would appear that such combinations (not only with zijn, but to a certain extent with worden and blijven as well) dramatically increase in frequency near the end of the sixteenth century. This means that there is a remarkable parallel with the development in English, where the so-called progressive increases in frequency during the sixteenth century as well.] (translation ours)
As can be inferred from Figure 2, this development coincides with that of the free adjuncts. The growth in frequency in the free adjunct as well as in the progressive is especially remarkable in light of the absence of a real gerund-like construction partaking in the process of mutual influence. The Dutch progressive was, however, arrested in its development, and it gradually disappeared from the language. As van der Horst (2008) points out, the Dutch V‑ende progressive never reached the stage of the passive progressive, which appeared in English in the late eighteenth century. Instead, from the sixteenth century onwards other constructions arose in Dutch to express durative aspect, such as liggen, staan, or zitten + ende + second conjugated verb, as in (45), or aan het + infinitive, as in (46), and later also zitten, staan, or liggen (+ te ) + infinitive, as in (47)‒(48): (45) Hier zit een vroutgen en spint. ‘Here sits a woman and (she is) spinning.’ (lit. ‘Here sits a woman and spins.’) (van der Horst 2008: 879) (46) (…) het oude regiment van de Spaenschen (…) dat op die tijd aan het muyten was. ‘the old regiment of the Spanish (…) that at that time was committing mutiny’ (lit. ‘that was at that time on the rebelling’) (van der Horst 2008: 1161) (47) (…) haer hondeken (…) met het welcke sy al bleef sitten spelen. ‘Her little dog with which she kept playing’ (lit. ‘her little dog with the which she kept sit play’) (van der Horst 2008: 880) (48) Maar ik merk, dat ik zit te babbelen. ‘But I notice that I am talking.’ (lit. ‘But I notice that I sit to chatter.’) (van der Horst 2008: 1472) The exact relation between the emergence of these alternatives and the loss of the V-ende progressive has not yet been studied, but it is likely that as these
The development of free adjuncts in English and Dutch
41
alternatives developed and became more frequent, they began competing with the V‑ende progressive for the same environments, ultimately resulting in its loss in the eighteenth century. This loss of the Dutch progressive could have had an indirect impact on the overall frequency of the free adjunct in that the total number of clausal V‑ende constructions decreased.
5.2.2 The role of the gerund Since both the free adjunct and the progressive could develop the way they did in Dutch without any influence of a gerund-like construction, the question arises what the actual contribution of the gerund has been to the development of the free adjunct in English. It has been suggested that the free adjunct and the absolute construction themselves were an important influence in the verbalization of the gerund. Houston (1989: 176‒181) and De Smet (2008: 56‒57), for instance, argue that the earliest attestations of verbal gerunds are found in the syntactic position of prepositional complements. These prepositional complements in Middle and Early Modern English take the form of gerunds following prepositions such as in, by, and for and often serve as adjuncts of manner, means, accompanying circumstance, reason, or cause (see 49). Thus they had (part of) their discourse function in common with the free adjunct (Houston 1989; De Smet 2008): (49) (…) he [i.e., Joshua] obteyned in lyftyng his hondes (…) (De Smet 2008: 264) As result of this shared function, Houston (1989) and Koma (1987) argue, verbal traits of the present participle in the free adjunct could spread to the gerund. Fanego (2004) suggests that the gradual use of (verbal) gerunds in subject position can be explained by the influence of a now obsolete type of absolute con struction, as in (50): (50) (…) and so Vaughan’s Testimonie being credited, Ø may be the material Cause of my Condemnation, as the Jury may be induced by his Depositions to speak their Verdict, (…) (1554, PPCEME) Fanego explains that “[h]ere the participle precedes its superordinate clause and controls its subject, which, as noted by Kortmann (1991: 101), is deleted under identity with the subject of the absolute” (2004: 44), viz. Vaughan’s Testimonie in (50). Some of these structures, then, “were reanalyzed as verbal gerunds in subject position” (Fanego 2004: 44). While examples (49) and (50) make a good case for how the free adjunct might have played a role in the verbalization of the gerund, they do not clarify how the verbal gerund might have influenced the maintenance of the free adjunct
42
Lauren Fonteyn and Hubert Cuyckens
in English. In addressing this issue, Killie admits that it is very hard to prove or pin down such causal influence (2006: 463). She tries to account for the gerund’s influence on the free adjunct in the following way: if one V‑ing + D(irect) O(bject) construction influences another V‑ing + DO construction and boosts its frequency, the latter V‑ing + DO can in its turn again influence and increase the use of the former. This process of mutual influence between V‑ing + DO constructions, i.e., the verbal gerund, the progressive, and the free adjunct, is situated between the sixteenth and seventeenth centuries. Yet, in Middle and Early Modern English the verbalization of the gerund was still in its earliest stages (De Smet 2013: 133‒135), while at that point the free adjunct was already quite frequent. This supports the idea that free adjuncts might have influenced the rise of the verbal gerund, but it raises the question whether, conversely, the gerund impacted the spread of the free adjunct and the progressive. Moreover, the initial development of the Dutch free adjunct was identical to that in English, even though there never was any gerund(-like) construction in ‑ende in Dutch. Our analysis of the Dutch data consequently suggests that the influence of the gerund in the initial stages of the expansion of the English progressive and free adjunct may have to be toned down. The verbal gerund does, however, seem to be of major importance in the maintenance of the free adjunct and of the progressive V‑ing construction in English, as both Norwegian and Dutch never had a gerund-like V‑ende construction and lost their progressive and free adjunct constructions. A crucial point here is that the actual influence of the gerund construction should not be situated in Middle or Early Modern English, as Killie and Swan suggest, but in Late Modern English. In the data presented in Section 3, we have shown that it is only after the Early Modern period that the development of English and Dutch free adjuncts started to diverge. Interestingly, while in Middle and Early Modern English verbal gerunds that functioned as adverbials frequently occurred with prepositions such as in, by, and for, prepositions such as after, at, before, since, and upon do not easily combine with verbal gerunds. These prepositions prefer definite nominal gerunds, i.e., gerunds that are modified by a definite article and encode objects by means of an of-phrase (the eating of the apple).19 Yet, De Smet (2008) points
19 De Smet points out that in Middle and Early Modern English “[d]efinite nominal gerunds combine with these prepositions to express a temporal relation between the state of affairs in the main clause and some specific event that then serves as a temporal reference point vis-à-vis the state of affairs in the main clause” (2008: 273‒274). They lend themselves well to this purpose, as the definite article signals them as “specific and accessible to the reader”, and therefore definite nominal gerunds are ideal to serve as temporal landmarks (De Smet 2008: 274). Verbal gerunds at that point are typically generic or indefinite, and hence they do not favor use as temporal landmarks.
The development of free adjuncts in English and Dutch
43
out, after 1640 verbal gerunds became increasingly frequent in combination with these prepositions: (51) Upon viewing of wich thinges there was devised tow fortes to be made upon the entry of the haven. (De Smet 2008: 83) (52) (…) then, soone after, I tooke my Cotch and went to Linton, wher, I aftor salutinge my mother, praied, and so went to supper. (De Smet 2008: 83) Notice that, when used in these temporal meanings, verbal gerunds are typically controlled by the subject (or another element) in the main clause and therefore resemble free adjuncts, as in (52). Importantly, examples such as (52) show that in Present-Day English, the distinction between gerunds and free adjuncts is not clear-cut, as there is no way of determining whether after in (52) is a preposition or a conjunction. This is also the case with before (53) and since (54): (53) Before {going to bed, I went to bed, my nap}, I turned the lights off. (54) Since {going vegan, I went vegan, my diet change}, I have been a lot hungrier. Throughout history, there are several instances of “-ing-clauses (…) [that] break out of the distributional mold of the phrasal categories, yielding uses that simply cannot be classified as gerund or participle, thereby undermining the historical distinction between the two clause types” (De Smet 2010: 1182). Thus, not only do the free adjunct and gerund by the eighteenth century display a fairly large number of overlapping adverbial meanings, they also show a considerable amount of formal overlap.20 This “blurring of the gerund/participle divide” might have enhanced the communication between both categories, and even led to the possibility of “formal feature exchange” (De Smet 2010: 1173). With the verbal gerund becoming increasingly popular in English (Jespersen 1940: 116‒118) and having enough formal and discourse-functional features in common with the free adjunct, it might be argued to have stimulated the latter’s popularity, while in Dutch no such gerund-like construction was there to boost the Dutch free adjunct.
20 Huddleston and Pullum (2002: 1187–1222) go even as far as saying that in Present-Day English, the present participle and the gerund have become one inflectional category, the gerundparticiple. These can be used as complements and non-complements. De Smet (2010) points out that historical evidence and semantic evidence support this idea, but there is still a considerable amount of formal evidence and evidence from the internal syntax of the constructions suggest ing that gerund and present participle are separate categories. The discussion will not be further elaborated here, but it is important to note that there is in fact a growing amount of overlap between the gerund and the present participle.
44
Lauren Fonteyn and Hubert Cuyckens
5.3 Explaining the divergence of the development of English and Dutch free adjuncts The decline of the Dutch free adjunct in the nineteenth century is likely to have been the result of all the above-mentioned factors coinciding. The position of the free adjunct in a language can be thought of in terms of the related constructions model suggested by König and van der Auwera (1990) (Figure 8). König and van der Auwera suggest that if in a particular language the present participle occurs in one or more of these five uses, the free adjunct will be one of them. From our data, it has become clear that the predicative use (the Dutch V‑ende progressive), the attributive use (e.g., met een lacchend gelaet ‘with a smiling face’), and the absolute use (e.g., Zijnde nu de kerk vol volx, […] zo quam de mis ‘The church being now full of people […], the mass began’) were present in sixteenth-century Dutch.21 What happened then is that, due to a combination of different language-internal and -external developments, nearly all of these uses were lost. Firstly, presumably through competition with new progressive constructions in Dutch, the V‑ende progressive was lost. Approximately at the same time, the attributive use of the present participle (including the participial relative clause, which showed a high degree of formal and functional similarity to the adverbial free adjunct) started moving from its original position following the noun to a position preceding the noun with which it formed a noun phrase, as the pronominal slot started to allow an increasingly large number of words (see van der Horst [2008] on the complexe
object nexus
predicative
“adverbial SS use”* free adjunct
absolute
attributive/apposition *adverbial Same Subject use Figure 8: Free adjunct and related constructions in König and van der Auwera (1990: 347).
21 Whether the object nexus use (e.g., I saw him standing there) occurred in Dutch at the time is unclear, but in Present-Day Dutch the infinitive is used (Ik zag hem lachen ‘I saw him smile’).
The development of free adjuncts in English and Dutch
45
voorbepaling). In addition, the absolute, being rejected by Dutch prescriptivists, became rare and (almost) disappeared. Thus, as its related uses gradually disappeared, the position of the free adjunct weakened and its frequency decreased. By contrast, the English free adjunct and progressive retained the strong position they had acquired by the sixteenth century, most likely due to the presence of the -ing gerund, which became increasingly difficult to distinguish from the present participle. Moreover, English purists did not reject the absolute construction, and no changes affected the attributive use of the English present participle clause.
6 Conclusion This paper has shown that the historical development of the free adjunct is more complex than was suggested by Killie and Swan (2009). Their claim was that, due to a fortunate combination of language-internal (V‑ing progressive, ‑ing gerund) and ‑external factors (language purists did not reject the construction), the free adjunct in English underwent a process of expansion and became a true part of the English grammatical system. This expansion involved a diversification of adverbial meanings, a syntactic reanalysis (from a semi-coordinate to a more integrated, subordinate construction), and an increased preference for sentenceinitial position. Killie and Swan (2009) further claim that, because these favorable circumstances were not present in other Germanic languages, the free adjunct never got established to the same degree; in a language such as Norwegian, for instance, the decline of the free adjunct can be attributed to purists reacting heavily against the Latinate construction as well as the lack of a grammaticalized progressive and of a gerund-like construction. When the development of the Dutch free adjunct is studied in detail and compared to the development of its English counterpart, one finds that, unlike Norwegian, Dutch developed largely in parallel with English, up until the nineteenth century. As in English, the Dutch free adjunct expanded semantically, expressing (progressively more) temporal and causal meanings, and it showed an increasing preference for sentence-initial position. In this position, the Dutch free adjunct caused inversion, thus signifying that it had been reanalyzed as a fully subordinate clause. The construction remained popular in Dutch until the nineteenth century, when the co-occurrence of several other factors (the loss of the progressive in ‑ende, the emergence of the complexe voorbepaling, and the loss of the absolute) weakened its position; this led to a sudden decrease in frequency, which probably continued throughout the twentieth century. The fact that until the nineteenth century the development of the English and the Dutch free adjunct was quite similar is an important finding, which is useful
46
Lauren Fonteyn and Hubert Cuyckens
not only to shed further light on the actual development of the free adjunct in Dutch, but also to refine, or perhaps even adjust, existing views on the development and eventual success of the construction in English. In particular, this paper has questioned whether the semantic diversification and reanalysis to a fully subordinate clause is key to the establishment of the English free adjunct, seeing that the same development has taken place in Dutch, where it has not been retained. We argued that these changes might have been the initial step towards estab lishing the free adjunct as a frequent means of clause combining, but they do not provide a conclusive answer to the question why the free adjunct has remained so successful in English. What does seem to have been crucial is the presence of the verbal gerund. The increasing discourse-functional and – importantly – the increasing formal overlap between the gerund and the present participle could have been key factors in transferring the success of the verbal gerund to the free adjunct in the nineteenth and perhaps the twentieth century. Finally, not only has the analysis of the comparable English and Dutch corpora shed new light on the development of the construction in the two languages, it has also pointed towards a new set of research questions. The findings presented in this paper allow us to speculate on possible deeper causes of the maintenance of the free adjunct, and they raise the question whether the success of the free adjunct in English can be linked to a broader typological shift that (arguably) affected English in the way in which narratives are construed; it has been claimed that while Old English was characterized as a bounded language, presenting a narrative as a sequence of bounded or completed events, it gradually developed to a more unbounded language, using constructions that construe the narrative as a cluster of roughly simultaneous events such as the V‑ing progressive for present tense (Carroll and von Stutterheim 2003; Los 2009, 2012; Petré 2010) and possibly also the free adjunct construction as a backgrounding strategy. Interestingly, while none of the other Germanic languages underwent such a shift, Dutch is argued to take an intermediate position between the unbounded (English) and the bounded type (German), which could explain its initial hospitality to the free adjunct (Los 2012). Still, to provide a more satisfactory conclusion on the deeper causes driving the observed changes, further research on the rise and development of present-participial and related constructions in English, Dutch, and other Germanic languages is needed.
Acknowledgments The research reported in this paper has been funded partly by FWO project GOA 54.12 (“The development of nominal and verbal gerunds from Middle to Late
The development of free adjuncts in English and Dutch
47
Modern English: Towards a semantic and discourse-functional analysis”). We are indebted to Liesbet Heyvaert and Hendrik De Smet for valuable input, and we would especially like to thank Liesbet Heyvaert for careful revision of a previous version.
Corpora and primary texts 38 miljoen woorden corpus (see http://taalunieversum.org/inhoud/corpora/inl-corpus38-mln-woorden). Compilation corpus of historical Dutch (Evie Coussé) (see http://www.diachronie.nl/corpora/ pdf/Cousse_2010_Een_digitaal_compilatiecorpus_historisch_Nederlands.pdf). Digitale bibliotheek van de Nederlande letteren (see http://www.dbnl.org/overdbnl/index.php). Novels & translations (all novels and their Dutch translations are found at http://www. gutenberg.org/): A Christmas carol (Charles Dickens) & translation “Een Kerstlied in proza” by J. Kuylman; Uncle Tom’s cabin (Harriet Beecher Stowe) & translation “De Negerhut” by C. M. Mensing; Sense and sensibility (Jane Austen) & translation “Gevoel en verstand” by Gonne Van Uildriks; The adventures of Tom Sawyer (Mark Twain) & translation “De lotgevallen van Tom Sawyer” by Johan Braakensiek. Penn corpora of historical English (see http://www.ling.upenn.edu/histcorpora/): Penn-Helsinki parsed corpus of Middle English, 2nd edn. (PPCME2); Penn-Helsinki parsed corpus of Early Modern English (PPCEME); Penn parsed corpus of Modern British English (PPCMBE).
References Anonymous. 1810. Syntaxis of woordvoeging der Nederduitsche taal. Leiden & Deventer: de Maatschappij Tot Nut van ’t Algemeen. Blatt, Franz. 1957. Latin influence on European syntax. Travaux du cercle linguistique de Copenhague 11: 33–69. Carroll, Mary & Christiane von Stutterheim. 2003. Typology and information organisation: Perspective taking and language-specific effects in the construal of events. In Anna Giacalone Ramat (ed.), Typology and second language acquisition, 365–402. Berlin: Mouton de Gruyter. Callaway, Morgan. 1901. The appositive participle in Anglo-Saxon. Publications of the Modern Language Association of America 16. 141–360. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. David, Jan Baptist. 1853. Nederduitsche spraekkunst voor middelbare scholen en collegien. Leuven: Vanlinthout en Co. Declerck, Renaat. 1991. A comprehensive descriptive grammar of English. Tokyo: Kaitakusha Co. den Hertog, Cornelis H. 1909. Nederlandse spraakkunst. Eerste stuk: Leer van de enkelvoudige zin. Introduced and edited by H. Hulshof. Amsterdam: W. Versluys, 1972.
48
Lauren Fonteyn and Hubert Cuyckens
De Smet, Hendrik. 2008. Functional motivations in the development of nominal and verbal gerunds in Middle and Early Modern English. English Language and Linguistics 12. 55–102. De Smet, Hendrik. 2010. English -ing-clauses and their problems: The structure of grammatical categories. Linguistics 48(6). 1153–1193. De Smet, Hendrik. 2013. Spreading patterns: Diffusional change in the English system of complementation (Oxford Studies in the History of English). New York: Oxford University Press. Fanego, Teresa. 2004. On reanalysis and actualization in syntactic change: The rise and development of English verbal gerunds. Diachronica 21(1). 5–55. Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Martin Haspelmath & Ekkehard König (eds.), Converbs in cross-linguistic perspective, 1–56. Berlin: Mouton de Gruyter. Heine, Bernd. 2002. On the role of context in grammaticalization. In Ilse Wischer & Gabriele Diewald (eds.), New reflections on grammaticalization, 83–101. Amsterdam & Philadelphia: John Benjamins. Heyvaert, Liesbet. 2011. On the meaning of gerunds. Invited talk delivered at the University of Santiago de Compostela, for the research unit Variation, Linguistic Change and Grammaticalization, 9 June 2011. Hopper, Paul J. & Elizabeth Closs Traugott. 2003. Grammaticalization. 2nd edn. Cambridge: Cambridge University Press. Houston, Ann. 1989. The English gerund: Syntactic change and discourse function. In Ralph W. Fasold & Deborah Schiffrin (eds.), Language change and variation, 173–196. Amsterdam & Philadelphia: John Benjamins. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Jespersen, Otto. 1940. A modern English grammar on historical principles. Part V. Vol. IV: Syntax. Copenhagen: Ejnar Munksgaard. Killie, Kristin. 2006. Internal and external factors in language change: Present participle converbs in English and Norwegian. Neuphilologische Mitteilungen 107. 447–469. Killie, Kristin. 2007. On the history of verbal present participle converbs in English and Norwegian and the concept of change from below. In Stephan Elspass, Nils Langer, Joachim Scharloth & Wim Vandenbussche (eds.), Germanic language histories ‘from below’ (1700–2000), 149–162. Berlin and New York: Walter de Gruyter. Killie, Kristin & Toril Swan. 2009. The grammaticalization and subjectification of adverbial ‑ing clauses (converb clauses) in English. English Language and Linguistics 13. 337–363. Kohnen, Thomas. 2003. The influence of “Latinate” constructions in Early Modern English: Orality and literacy as complementary forces. In Dieter Kastovsky & Arthur Mettinger (eds.), Language contact in the history of English, 171–194. Frankfurt am Main: Peter Lang. Kohnen, Thomas. 2004. Text, Textsorte, Sprachgeschichte: Englische Partizipial- und Gerundialkonstruktionen 1100 bis 1700. Tübingen: Niemeyer. Koma, Osamu. 1987. On the initial locus of syntactic change: Verbal gerund and its historical development. English Linguistics. Journal of the English Linguistic Society of Japan 4. 311–324. König, Ekkehard & Johan van der Auwera. 1988. Clause integration in German and Dutch conditional, concessive conditional, and concessive clauses. In John Haiman & Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 101–133. Amsterdam & Philadelphia: John Benjamins.
The development of free adjuncts in English and Dutch
49
König, Ekkehard & Johan van der Auwera. 1990. Adverbial participles, gerunds and absolute constructions in the languages of Europe. In Johannes Bechert, Giullano Bernini & Claude Buridant (eds.), Toward a typology of European languages, 337–356. Berlin: Mouton de Gruyter. Kortmann, Bernd. 1991. Free adjuncts and absolutes in English: Problems of control and interpretation. London: Routledge. Lehmann, Christian. 1988. Towards a typology of clause linkage. In John Haiman & Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 181–225. Amsterdam & Philadelphia: John Benjamins. Los, Bettelou. 2009. The consequences of the loss of verb-second in English: Information structure and syntax in interaction. English Language and Linguistics 13(1). 97–125. Petré, Peter. 2010. The functions of weorðan and its loss in the past tense in Old and Middle English. English Language and Linguistics 14(3). 457–484. Rinkel, Tineke. 1989. Over zeventiende-eeuwse participiumconstructies: Participiumconstructies bij Hooft, De Laet en De Vries. Voortgang 10. 205–237. Río-Rey, Carmen. 2002. Subject control and coreference in Early Modern English free adjuncts and absolutes. English Language and Linguistics 6(2). 309–323. Sørensen, Knud. 1957. Latin influence on English syntax: A survey with a bibliography. Travaux du cercle linguistique de Copenhague 11. 131–155. Thompson, Sandra A. 1983. Grammar and discourse: The English detached participial clause. In Flora Klein-Andreu (ed.), Discourse perspectives on syntax, 43–64. New York: Academic Press. Traugott, Elizabeth Closs & Ekkehard König. 1991. The semantics-pragmatics of grammaticalization revisited. In Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, Vol. 1, 189–218. Amsterdam & Philadelphia: John Benjamins. Van Belle, William. 2002. Zwijgen is niet altijd toestemmen. De rol van inferenties bij het interpreteren en argumenteren. Leuven: Acco. van der Horst, Johannes. 2008. Geschiedenis van de Nederlandse Syntaxis I & II. Leuven: Universitaire Pers. van der Wal, Marijke & Cor van Bree. 2008. Geschiedenis van het Nederlands. Utrecht: Spectrum.
Joanna Nykiel
Semantic dependencies and the history of ellipsis alternation 1 Introduction Recent research on syntactic variation reveals a complex interaction of multiple factors affecting speakers’ preferences for one variant over another. These factors include features of the context in which a particular variant is used, as well as the difficulty associated with processing the different variants, given the context (Hawkins 2000, 2004; Bresnan 2007; Bresnan and Hay 2008; Nykiel 2013a; Wolk, Bresnan, Rosenbach and Szmrecsanyi 2013). Data used in this research come from both corpus studies and psycholinguistic experimentation, showing remarkable parallels between variants actually found in specific contexts and speakers’ knowledge of the likelihood of finding the same variants in the same contexts (Bresnan 2007; Ford and Bresnan 2013; Nykiel 2014). Such data raise the question of whether or not syntactic variation should be accounted for by extragrammatical factors. In this chapter, my focus is on a single syntactic alternation in English which is associated with two elliptical constructions. Examples (1) and (2) illustrate this alternation. (1) A: What is this cake filled with? B: Whipped cream/With whipped cream. (2) Pat played in one of the houses on the hill, but I’m not sure which/in which. The elliptical construction in (1) is referred to as Bare Argument Ellipsis, and the construction in (2) is referred to as sluicing. The B-sentence in (1) and the second clause in (2) contain stranded phrases – an XP (this may be any phrase, from a noun phrase to an adverbial phrase) in the case of Bare Argument Ellipsis and a wh-phrase in the case of sluicing – which I will henceforth refer to as “remnants”. The prepositions with and in are optionally omitted from the remnants if they are part of the remnants’ correlates, which are the PPs what … with1 and in one of the houses on the hill. In the rest of this paper, I will call a remnant without
1 This correlate is discontinuous compared to the one in example (2) in the sense that its constituents are non-adjacent. In Nykiel (2013b, 2014), I explore the influence that such correlates have on ellipsis alternation, but the difference between the two types of correlates is immaterial to the argument developed here.
52
Joanna Nykiel
a preposition an NP remnant, and one with a preposition a PP remnant. I will refer to the alternation between NP and PP remnants as “ellipsis alternation”, following Nykiel (2013b, 2014). Here, I discuss one constraint on this alternation (for a full analysis, see Nykiel 2013b, 2014), that is, the semantic relationship of the PP correlate and the remnant with the appropriate lexical head. I explore to what extent this constraint is seen operating on remnants in corpus data from the history of English. Building on Claridge’s (2000) work on multi-word verbs, I present evidence that the rate of NP remnants increases in parallel with the rate of multi-word verbs, and that NP remnants co-occur with multi-word verbs more often than do PP remnants. I conclude that what underlies these patterns is a principle of language processing, the same as the one that influences constituent ordering in nonelliptical clauses. Because principles of language processing are normally regarded as part of the performance systems rather than competence grammar, the behavior of NP and PP remnants can be interpreted as having extra-grammatical motivation. Nykiel (2014) demonstrates that a semantic dependency holding between a preposition and another lexical category is a strong predictor of how remnants will be realized in Present-Day English.2 Consider the semantically dependent combination of verb and preposition (come and across) in (3) and the semantically independent one (frolic and in) in (4) below. The notion of a semantic dependency is discussed in detail in section 2; for now, just note that come and across constitute a prepositional verb, while frolic and in do not. (3) A: Pat came across some old books. a. B: ?Across how many? b. B: How many? (4) A: Moose sometimes come and frolic in our pool. a. B: In which one? b. B: Which one? The remnant in (3a) is significantly less likely to appear, if it is at all acceptable, than the remnant in (4a), given the different levels of semantic dependency in these combinations of verbs and prepositions. Levels of semantic dependency give rise to a gradient of combinations of verbs and prepositions, from semantically independent ones (e.g., to walk in (a park)) to semantically dependent ones (e.g., to look after (a neighbor’s cat)) (Huddleston and Pullum 2002: 277; Brinton and Traugott 2005: 128). Multi-word verbs play an important role in this classification in
2 Only 25 percent of all lexical categories to be discussed here are non-verbal, and hence my focus is on combinations of prepositions and verbs.
Semantic dependencies and the history of ellipsis alternation
53
that they instantiate various categories of semantically dependent combinations. One of these categories, prepositional verbs, is of interest to us here, because it promotes the use of NP remnants. The history of English provides a natural testing ground for the issue of whether the effect of semantic dependencies on ellipsis alternation is merely an accidental characteristic of Present-Day English or the outcome of the gradual development of multi-word verbs. I propose that semantic dependencies between prepositions and verbs are linked to the appearance of prepositional verbs, which, according to C laridge (2000), took place in late Old English (450–1150) or early Middle English (1150–1500). Given that prepositional verbs host verbs and prepositions that are semantically dependent on each other, we can expect that if the number of such verbs steadily grows beginning in Middle English, it will influence ellipsis alternation by reducing the frequency of PP remnants and increasing the frequency of NP remnants. This is exactly what we see in the diachronic data collected for this study. The rest of this chapter is structured as follows. The next section overviews the existing empirical support for the role of semantic dependencies in constit uent ordering in nonelliptical clauses and in ellipsis alternation. In section 3, I provide historical background on English prepositional verbs. Section 4 is a corpus investigation of ellipsis alternation in the history of English, contextualiz ing the alternation in light of Claridge’s (2000) study of English multi-word verbs as a means to test the hypothesis that English historical data provide evidence for a processing-based account of this alternation. In section 5, I address the implications of the diachronic distribution of ellipsis remnants with and without prepositions for our understanding of the constraints on ellipsis alternation both in English and in cross-linguistic terms. One possibility suggested by the current data is that languages do not differ in terms of the availability of ellipsis alternation, but rather in terms of how the ease of processing of the two types of remnants influences their distribution in individual languages. Section 6 concludes.
2 Constituent ordering and ellipsis alternation: The effect of semantic dependencies Prepositional phrases (PPs) differ in terms of their status relative to the larger constituents (verb phrases) within which they are embedded. Their status is usually captured by the well-known distinction between adjunct and argument PPs: argument PPs are assumed to be lexically specified by the appropriate verbal heads, but adjunct PPs are not (Kegl and Fellbaum 1988; Grimshaw 1990). It is also a possibility that prepositions alone, rather than whole PPs, are lexically
54
Joanna Nykiel
specified by verbal heads (Boland and Boehm-Jernigan 1998). Thus the distinction between adjunct PPs and argument PPs captures their different relationships with the verbal heads. Hawkins (2000, 2004) offers a processing-based alternative to this view. He suggests that the processing of a verb might require simultaneous access to a preposition, and/or vice versa, and if so, the two form a processing domain. If this is not the case, then the preposition forms a processing domain with its object. The requirement of simultaneous access to two lexical categories is mediated by the semantic dependency between them. As a means to identify these semantic relationships, Hawkins (2000) introduces entailment tests. These tests assign one of two features (dependent or independent) to verbs and prepositions, based on whether they can be processed independently of each other. In Hawkins (2000: 242–243), the tests are defined as in (5) and (6). (5) (6)
Verb Entailment Test If [X V PP] entails [X V], then assign Vi[independent]. If not, assign Vd[dependent]. Pro-Verb Entailment Test If [X V PP] entails [X Pro-V PP] or [something Pro-V PP] for any pro-verb sentence listed below, then assign Pi[independent]. If not, assign Pd[dependent]. Pro-verb sentences: X did something PP; X was PP; something happened PP; something was the case PP; something was done (by X) PP.
For any sequence of X V PP, the Verb Entailment Test removes the PP to see if the verb can be processed without it. The Pro-Verb Entailment test replaces the verb with a general proform (one of the Pro-verb sentences in (6)) to see if the PP’s meaning is independent of that particular verb. Let us now apply these tests to examples (1)–(3), repeated here for convenience as (7)–(9). (7) A: What is this cake filled with? B: Whipped cream/With whipped cream. (8) Pat played in one of the houses on the hill, but I’m not sure which/in which. (9) A: Pat came across some old books. a. B: ?Across how many? b. B: How many? By applying the Verb Entailment Test, we assign Vi to the verb fill in example (7): The cake is filled with whipped cream entails The cake is filled. By applying the same test, we also assign Vi to the verb play in example (8): Pat played in one of the houses on the hill entails Pat played. In example (9), however, Vd will be assigned to the verb come: Pat came across some old books does not entail Pat came. The Pro-Verb Entailment Test outputs Pd for the PP with whipped cream in example (7): The cake is filled with whipped cream does not entail Something is
Semantic dependencies and the history of ellipsis alternation
55
the case with whipped cream. Pi will be assigned to the PP in one of the houses on the hill in (8): Pat played in one of the houses on the hill entails Pat did something in one of the houses on the hill. Finally, Pd will be assigned to the PP across some old books in (9), since Pat came across some old books does not entail Pat did something across some old books. In sum, example (9) shows a two-way dependency (this is typical of prepositional verbs), example (7) a one-way dependency, and example (8) none at all (these dependencies are summarized in Table 1). What these dependencies mean is that in examples (7) and (9) the prepositions form processing domains with the verbs instead of their objects, while in example (8), the preposition forms a processing domain with its object. Hawkins (2004: 31) further argues that the human parser prefers adjacency of constituents that belong in a single processing domain, so that semantic dependencies between them may be processed over as small a distance as possible. This preference is seen, for instance, in the order of postverbal PPs. Their order is normally mediated by the difference in weight between the PPs (Hawkins 2000: 234). Compare the two orders in examples (10) and (11). Because PP1 is three words shorter than PP2, the mother node (the VP and its constituents PP1 and PP2) may be computed sooner in (11) than in (10), which leads to a preference for the order in (11). (10) The astronomer VP[gazed PP2[into the dark but moonlit sky] PP1[through his telescope]] (11) The astronomer VP[gazed PP1[through his telescope] PP2[into the dark but moonlit sky]] When postverbal PPs are of similar weight, however, then it is relations of semantic dependency, not weight, that determine the word order, indicating that the two types of relations are separate factors. Additional evidence for the effect of semantic dependency on word order is offered by Lohse, Hawkins, and Wasow (2004), who discuss the ordering of constituents in verb-particle constructions. Applying the same entailment tests as Hawkins (2000, 2004), they find that the semantic relationship between a verb and a particle affects their position with respect to each other. Presence of a semantic dependency is more likely to lead to the particle occupying the immediately Table 1: Semantic dependencies for examples (7)–(9) Example
Dependency
(7) What is this cake filled with? (8) Pat played in one of the houses on the hill. (9) Pat came across some old books.
One-way: Vi and Pd None: Vi and Pi Two-way: Vd and Pd
56
Joanna Nykiel
postverbal position, as in (12), instead of following the direct object, as in (13). Lohse, Hawkins, and Wasow (2004) also test the independence of the effect of semantic dependency by comparing it with the effect of the length of the direct object NP on the word order, and find that the semantic dependency effect indeed is independent. (12) They carried out a repair. (Lohse, Hawkins, and Wasow 2004: 245) (13) They carried a repair out. (Lohse, Hawkins, and Wasow 2004: 245) The semantics of combinations of verbs and prepositions has previously been addressed in the literature. This line of inquiry has focused on the idea that some combinations are more semantically cohesive than others (Quirk et al. 1985; Cruse 1986; Claridge 2000; Tseng 2000; Brinton and Traugott 2005). Following an early definition, semantically cohesive combinations are assumed to consist of “two or more words forming a semantic unit which is not identical with the combined meanings of its elements” (Kruisinga 1925: II/3, 72). This definition is recast in terms of lexicalization in Brinton and Traugott (2005: 96), where lexicalization is defined as follows: Lexicalization is the change whereby in certain linguistic contexts speakers use a syntactic construction or word formation as a new contentful form with formal and semantic properties that are not completely derivable or predictable from the constituents of the construction or the word formation pattern. Over time there may be further loss of internal constituency and the item may become more lexical.
Brinton and Traugott (2005: 128) argue that combinations of verbs and prepositions represent what they term level L1 on their cline of lexicality. This level refers to “partially fixed phrases, e.g., lose sight of, agree with” as opposed to more idiosyncratic phrases (Brinton and Traugott 2005: 94). However, Brinton and Traugott (2005: 128) also recognize, following Quirk et al. (1985: 1166) and Huddleston and Pullum (2002: 275–277), that combinations of verbs and prepositions have been affected by lexicalization to various degrees, the result of which is a gradient of these combinations, even at L1. The most advanced of these combinations are prepositional verbs (pick on, look after, come across), which are fully semantically noncompositional, but there are also combinations that are only partially noncompositional (invest in, suffer from, rely on). There are also free combinations of verbs and prepositions; these have not been lexicalized and hence are fully compositional (the prepositions are typically part of adjunct PPs). The lexicalization-based analysis of combinations of verbs and prepositions does a good job of recognizing their semantic dependency. And Hawkins’ (2004) processing-based account of why such combinations influence the order of sentential constituents captures the human parser’s preferences with respect
Semantic dependencies and the history of ellipsis alternation
57
to nonelliptical clauses (the insight itself that semantically coherent items tend to be placed adjacent is not new and goes back to Behaghel (1932)). This account also opens up the possibility that adjacency effects may be observed not only in constructions whose constituents are overtly expressed, but also in constructions in which some constituents have been elided. Indeed, as we will see shortly, an analysis of ellipsis alternation from the processing perspective provides support for this possibility, while representing an innovative approach to historical data. Combinations of prepositions and verbs which are semantically dependent on each other strongly favor NP remnants in Present-Day English, as evidenced both in corpus data and in judgments about the naturalness of the same data (Nykiel 2014). The sensitivity of remnants to semantic dependencies points to the involvement of principles of language processing: separation of process ing domains formed by prepositions and verbs is dispreferred also in elliptical clauses. To see this, consider that NP remnants preserve processing domains for pairs of semantically dependent prepositions and verbs by removing prepositions from remnants, which never contain verbs, either. PP remnants, on the other hand, separate processing domains for pairs of the same prepositions and verbs by adding prepositions to remnants, although verbs remain absent. As for pairs of semantically independent prepositions and verbs, where prepositions are more strongly associated with their objects, only PP remnants preserve proc essing domains by adding prepositions to their objects. What is found in the Present-Day English data is that NP remnants are in the majority (they constitute 67.2% of all remnants), unlike what appears to be the case in other languages. Although there is little empirical evidence except for Nykiel (2013), languages other than English are commonly reported to allow NP remnants under limited circumstances (Vicente 2008; Szczegielniak 2008; Rodrigues, Nevins, and Vicente 2009). This suggests that ellipsis alternation possesses a special status in English due to the availability of prepositional verbs, which, as we have seen, represent the highest level of semantic dependency. I next turn to the history of those verbs in English.
3 The history of English prepositional verbs English has gone through a general shift from synthetic to analytic forms, part of which was the loss of Old English prefixed verbs (P-V compounds) in late Old English or early Middle English. Prefixed verbs (e.g., besprecan ‘speak about’) were replaced by verbs followed by prepositions, giving rise to prepositional verbs (Denison 1985, 1993; Fischer 1992; Traugott 1999; Claridge 2000; Goh 2001; Brinton and Traugott 2005). This replacement has the consequence that it enables
58
Joanna Nykiel
a reanalysis of the sort shown in (14), where the verb now forms a unit with the preposition which was formerly part of the PP (Denison 1985). (14) [V + PP[P + NP]] → [[V + P] + NP] Simply put, the mechanism of reanalysis has been proposed as a means to account for increased fusion, and hence, partial or complete loss of composi tionality, for pairs of verbs and prepositions. How exactly this reanalysis should be framed in formal terms remains problematic (see, for example, Hornstein and Weinberg 1981; Baltin and Postal 1996; Huddleston and Pullum 2002). Perhaps it is best seen as operating entirely at the level of semantics, not syntax. Brinton and Traugott (2005: 127–128) propose that the replacement of P-V compounds by reanalyzable pairs of prepositions and verbs led to lexicalization (see the definition above) such that, for some pairs, the two elements have become fused or semantically more unit-like. They also point out that what lends further plaus ibility to this analysis is the fact that some prefixed verbs already had noncompositional meanings, which could have been carried over to the new-formed verbs. Another factor that might have promoted a gradual loss of compositionality in pairs of verbs and prepositions is the availability of preposition stranding in English (Denison 1985). Preposition stranding, illustrated in examples (15) and (16), is known to be sensitive to noncompositional pairs in Present-Day English: environments where it is mandatory are those that host noncompositional pairs of verbs and prepositions, such as (15) (Hawkins 2004; Hoffmann 2011). (15) Whose children does she look after? (cf. *After whose children does she look?) (16) Who did he go to school with? (cf. With whom did he go to school?) It is possible that preposition stranding serves to increase the number of noncompositional pairs by widening the range of environments where verbs and prepositions appear adjacent, so that existing semantic dependencies between them may strengthen and new dependencies may form. At the same time, it would be incorrect to assume that the availability of preposition stranding is the direct cause for the reanalysis in (14) and for the development of prepositional verbs, because preposition stranding has existed in English since the Old English period (Denison 1985). What appears to be a safe assumption at this point is that languages that tolerate preposition stranding are more likely to have stronger semantic dependencies between verbs and prepositions than languages without preposition stranding, and hence prepositional verbs are also more likely to exist in languages with preposition stranding. If the formation of semantically dependent combinations of verbs and prepositions is a gradual process, it is possible to use the history of English as a window into the involvement of such pairs in ellipsis alternation. Given that
Semantic dependencies and the history of ellipsis alternation
59
prepositional verbs first appear in Middle English, we can isolate Old English as a stage where there are no prepositional verbs yet. As we will see in the next section, this stage provides evidence in support of the hypothesis that ellipsis alternation is sensitive to the strength of semantic dependencies between lexical items. The fourteenth century sees the first few instances of prepositional verbs, which gain in frequency in the following century (Denison 1985; Claridge 2000). Claridge (2000: 171) offers data showing that in Early Modern English (1500–1700) prepositional verbs have the frequency of 2,400 per million words, which remains stable over the course of this period.3 Among these verbs, only 17 percent of all tokens carry idiomatic (maximally noncompositional) meanings. These counts may be compared with data for Present-Day English (presumably the mid- to late twentieth century) provided by Biber et al. (1999: 415): the frequency of prepositional verbs is almost 5,000 per million words, more than twice their frequency in Early Modern English. This diachronic rise in the frequency of prepositional verbs predicts that, as in languages with weak semantic dependencies between prepositions and verbs (e.g., the Slavic languages),4 NP remnants in Old and Middle English should be rarer than PP remnants. The following periods, Early Modern English, and, in particular, Late Modern English, should see an increasing rate of NP remnants. By analyzing the contexts in which NP remnants appear, we can probe the involvement of prepositional verbs in ellipsis alternation.
4 The history of ellipsis alternation In this section, I provide data for the distribution of NP and PP remnants from Old English to Late Modern English. My purpose is to relate the distribution of remnants to the list of multi-word verbs identified by Claridge (2000). Old and Middle English data are a strong indication that ellipsis alternation is affected by semantic dependencies. This set of data was extracted from the Dictionary of Old English web corpus, The Helsinki corpus of English texts, and the Corpus of Middle
3 These counts are based on The Lampeter corpus of Early Modern English tracts (1640–1740), whose size is 1,172,102 words. 4 Nykiel (2014) distinguishes three levels of semantic dependency, with the highest level repesented by prepositional verbs, and notes that languages differ in terms of how many of these levels are available. For example, three levels are available in English, but only two in Polish, b ecause Polish does not have prepositional verbs. Nykiel then proposes that such differences underlie the strength of semantic dependencies present in a language. See also Hawkins (1999, 2000), who points out that English has strong semantic dependencies between verbs and prepositons.
60
Joanna Nykiel
English prose and verse. I extracted all wh-phrases, which I then checked for (1) whether they instantiated sluicing and had PP correlates, and (2) whether they were prepositional objects embedded in interrogative clauses and were followed by elliptical responses, as in Bare Argument Ellipsis. This search yielded 37 remnants in total (14 for Old English and 23 for Middle English), all of which were PP remnants. The absence of NP remnants in both periods suggests that such remnants have an association with prepositional verbs. One way of interpreting these data is to speculate that NP remnants were unavailable as long as prepositional verbs were also unavailable or rare. However, a weaker interpretation of the data is that NP remnants were uncommon, though not necessarily unavailable, because prepositional verbs were also uncommon. It could well be an artifact of the data that there is no record of NP remnants in these periods. Instances of ellipsis, including sluicing and Bare Argument Ellipsis, are readily found in dialogue or represented speech, but much less so in prose. Due to the rarity of dialogue in the available Old and Middle English texts, data from these periods may simply be providing unfavorable contexts for ellipsis in general, compared to data from the later periods. A complete absence of NP remnants in Old and Middle English would in fact be rather surprising, given that they are quite robust cross-linguistically, including in languages with weaker semantic dependencies than those seen in English. The subsequent periods provide instances of NP remnants, which show the interactive pattern that is in line with the current hypothesis. The data (including both sluicing and Bare Argument Ellipsis) were extracted from six corpora: the Complete works of William Shakespeare, The Helsinki corpus (for the Early Modern English period), A corpus of English dialogues 1560–1760, Literature online (1500–1900), the Old Bailey proceedings online (only the 1700s), and ARCHER: A representative corpus of historical English registers. For this search, I extracted all wh-phrases with PP correlates which instantiated sluicing. In addition, I extracted all wh-phrases which were part of interrogative clauses (elliptical or nonelliptical) and which acted as prepositional objects, and all responses to these interrogative clauses which were instances of Bare Argument Ellipsis.5 The sample I collected
5 The same data for Old, Middle, Early Modern, and Late Modern English are discussed in Nykiel (2013b), though from a different perspective. In both cases, the data exclude yes/no questions and responses of the kind shown in (i). (i) A: Am I speaking with Sheldon? B: No, (with) Leonard.
Semantic dependencies and the history of ellipsis alternation
61
consists of 351 remnants in total.6 There is, admittedly, a difference between the data collected for the Old and Middle English periods and that collected for the two later periods. Several of the corpora of Early and Late Modern English are rich in dialogue, which creates opportunities for sluicing, and perhaps Bare Argument Ellipsis in particular, to appear. One question to consider in this regard is to what extent the frequency of NP remnants in Early and Late Modern English and their absence before these periods are affected by the selection of texts from which the data come. Although it is clear that genre affects the overall frequency of ellipsis, it remains uncertain that it also affects the distribution of NP and PP remnants. That is, if PP remnants are preferred over NP remnants, we would still expect this preference to be visible both in dialogic and non-dialogic texts, and even if the frequencies of remnants differ between data samples as the consequence of genre. To identify the possible effect of developing semantic dependencies on ellipsis alternation in Early and Late Modern English, I divided the corpora into those representing Early Modern English and those representing Late Modern English. Overall, PP remnants (136; 67%) are favored over NP remnants (67; 33%) in Early Modern English, and PP remnants (82; 55.4%) are also favored over NP remnants (66; 44.6%) in Late Modern English. However, the ratio of NP remnants to PP remnants differs significantly between these periods (χ² = 4.8854, p < 0.05), indicating that Late Modern English has a higher frequency of NP remnants than does Early Modern English. The data are broken down into segments extracted from the individual corpora for Early Modern English (Table 2) and Late Modern English (Table 3). Some examples of both PP and NP remnants from Early and Late Modern English are given in examples (17)–(20) below (the remnants and their correlates are marked in bold). Table 2: Distribution of NP and PP remnants in Early Modern English Corpus Shakespeare Helsinki corpus Corpus of English dialogues Literature online (1500–1700) ARCHER (1600–1700)
PP remnants
NP remnants
30 (55.6%) 4 (66.7%) 77 (71.3%) 23 (82.1%) 2 (28.6%)
24 (44.4%) 2 (33.3%) 31 (28.7%) 5 (17.9%) 5 (71.4%)
Total 54 (100%) 6 (100%) 108 (100%) 28 (100%) 7 (100%)
6 Since a portion of these data (54 instances) comes from the corpus of Shakespeare’s plays, it raises the possibility that the realization of remnants is mediated by metrical considerations. See Nykiel (2013b) for arguments against this possibility.
62
Joanna Nykiel
Table 3: Distribution of NP and PP remnants in Late Modern English Corpus Old Bailey proceedings ARCHER (1701–1900) Literature online (1701–1900)
PP remnants
NP remnants
2 (9.5%) 16 (35.5%) 64 (78%)
19 (90.5%) 29 (64.5%) 18 (22%)
Total 21 (100%) 45 (100%) 82 (100%)
(17)
L. C. J.: Hath he been there? Vincent: Yes, he hath. L. C. J.: In what Company? Vincent: With Praunce. L. C. J.: And with any of the Prisoners? Vincent: Yes, but I can’t tell particularly with whom. (1678/1679 The tryals of Robert Green, Henry Berry & Lawrence Hill; A corpus of English dialogues) (18) Gaymann: The Devil! Show yourself a rascal of parts, Sirrah, and wait on him up with ceremony. Rag: Who, the Devil, Sir? Gaymann: Ay, the Devil, Sir, if you mean to thrive. (1668 Aphra Behn, The lucky chance, or, The alderman’s bargain; ARCHER) (19) Jerome, inquiring who was without, was answered, a herald. [“From whom?”] said he. [“From the knight of the gigantic sabre;”] (1764 Horace Walpole, The castle of Otranto; ARCHER) (20) It was in return, cousin Caroline. In return for what? Your present to me. (1856 John Esten Cooke, The last of the foresters; Literature online) The pattern that consistently reappears in four corpora of Early Modern English is that PP remnants are the more frequent option. One exception is the ARCHER corpus, where the majority of the remnants are NP remnants.7 For two corpora of Late Modern English, the proportions of PP and NP remnants are reversed with respect to Early Modern English, but the data pattern found in Literature online resembles the one familiar from Early Modern English. The Old Bailey proceedings also seem different in the sense that they yield a higher number of NP remnants
7 I found two further instances of ellipsis remnants, of which one is an NP remnant and one a PP remnant, in the Corpus of Early English correspondence sampler. They are not reported in Table 1 due to their paucity.
Semantic dependencies and the history of ellipsis alternation
63
than any of the other corpora. The recalcitrance of the data harvested from the ARCHER corpus, Literature online, and the Old Bailey proceedings may be reflecting some characteristics of these corpora, such as genre or perhaps the level of formality, which introduce imbalance to the dataset and possibly confound the results. However, there is no evidence, to the best of my knowledge, that NP remnants are more strongly associated with oral genres or informal speech than are PP remnants, and I leave it for future research to explore whether they are. What is known from Present-Day English is that two other factors in addition to semantic dependencies impact the distribution of NP and PP remnants. NP remnants are less likely to be used than PP remnants when clarity of expression is at stake (Nykiel 2014). For instance, the sluicing remnant shown in bold in example (21 [from Nykiel 2014: 10]) is asking for clarification of the identity of the individual mentioned in the correlate (Aguilar, who in fact is called Aguirre). A PP remnant is used in order to ensure that speaker A correctly identifies the correlate of the remnant, which is easier to achieve when the remnant provides more cues to what that correlate is (it is assumed in Nykiel 2014 that PP remnants provide more cues to the identity of the correlate than NP remnants do, because they match the syntactic category of the correlate). (21) A: But he couldn’t get along with Aguilar. B: With who? A: I mean Aguirre. B: Oh, Aguirre. What is also known from Nykiel’s (2014) data is that speakers tend to reuse in remnants syntactic structure that they have witnessed in the surrounding discourse. Example (18) also illustrates this point quite well. The first utterance by Gaymann hosts the correlate (the PP on him) for the subsequent remnants. Rag’s question contains the NP remnant Who, which then becomes the immediate correlate (and note that it now is a correlate appearing without the preposition on) for the NP remnant the Devil in Gaymann’s response. The Devil is an NP remnant reusing the syntactic category of its immediate correlate, although it could in principle have been a PP remnant. When one carefully controls for these additional factors, the frequency data collected from ARCHER, Literature online, and the Old Bailey proceedings are not puzzlingly different; rather, they reveal that ellipsis alternation is influenced by several factors at any given time (Nykiel 2013b). It is for these reasons that the frequency data presented in Tables 2 and 3 do not constitute conclusive evidence for the effect of semantic dependencies in ellipsis alternation. This is not to say that it is impossible to probe whether genre confounds these data by showing an increase in the frequency of NP remnants in Late Modern English that is only apparent. Toward this purpose, I developed
64
Joanna Nykiel
a mixed-effects regression model of the data (see Baayen 2008) that predicted the appearance of NP remnants as opposed to PP remnants from a single factor, Period (Early Modern English vs. Late Modern English), and included the corpus from which the data came as a random effect. Inclusion of random effects in a regression model ensures that the results are reliable, even though the data come from varied sources. The simple model I developed output a clear trend, which did not approach statistical significance, for NP remnants to gain in frequency in the Late Modern English period, compared to the Early Modern English period (β = 0.198, SE = 0.162, t = 1.223, p = 0.252).8 The patterns found in Early and Late Modern English are in line with the hypoth esis that a language with few prepositional verbs (and hence, weak semantic dependencies) has a general preference for PP remnants. This in turn suggests that the behavior of ellipsis alternation in Early Modern English is not far removed from that found in several other languages (e.g., Spanish, French, Polish, Czech, and Serbo-Croatian), where NP remnants are reported to be less frequent than PP remnants. But as we move into Late Modern English, we observe a pattern more reminiscent of Present-Day English, where the preference is for NP remnants. The temporal distribution of remnants correlates with the temporal shift in the frequency of prepositional verbs, but it requires further analysis before it becomes compelling evidence for the proposed correlation. Below, I present novel evidence that NP remnants are found together with prepositional verbs more often than PP remnants, in both Early and Late Modern English. In pursuit of this aim, I use the collection of multi-word verbs extracted from The Lampeter corpus (1640–1740) provided by Claridge (2000). Of interest to us here are the list of prepositional verbs (e.g., concur with, fix on, adhere to, belong to) (see Appendix 2 in Claridge 2000), the list of verb-adjective combinations (e.g., fall short of, make light of) (see Appendix 4 in Claridge 2000), and the list of verbo-nominal combinations (e.g., be in love with, take heed of) (see Appendix 5 in Claridge 2000). However, in analyzing ellipsis alternation in Early Modern English, I chose not to draw on data from The Lampeter corpus, even though Claridge’s own data come from it. This choice was forced by the scarcity of instances of ellipsis alternation in that corpus (which may well be due to the nature of the texts found in that corpus). I found as few as five, three of which hosted PP remnants and two hosted NP remnants. While these remnants represent data entirely compatible with the hypothesis tested in this chapter (that is, PP remnants should be more frequent
8 A fuller mixed-effects model of the current data, one that adds more predictors of the distribution of NP and PP remnants to the factor Period, can be found in Nykiel (2013b). In that model, Period is a statistically significant predictor of where NP remnants appear, and the remaining predictors serve to explain the differences among the individual corpora seen in Tables 2 and 3.
Semantic dependencies and the history of ellipsis alternation
65
than NP remnants in Early Modern English), they are not usable as evidence in any statistical sense. In contrast, the data shown in Tables 2 and 3 provide enough instances of ellipsis alternation that the role of multi-word verbs may be evaluated. 208 different combinations of verbs and prepositions (types, not tokens) are found in the data presented there. 150 of these appear with PP remnants and 58 with NP remnants. I analyze these combinations in terms of how many of them are listed in Claridge’s (2000) collection of multi-word verbs; this analysis reveals the following patterns. In the case of NP remnants, a total of eleven combinations are found in Claridge’s (2000) collection: ten are listed as prepositional verbs and one as a verb-adjective combination (to fall short of). In the case of PP remnants, seven combinations are listed: six as prepositional verbs and one as a verbo-nominal combination (to be in love with). There is slight inconsistency in the data (not entirely unexpected, given that different verbs may be more or less lexicalized) in the sense that three of these combinations appear with both NP and PP remnants: care for appears with an NP remnant and a PP remnant in Late Modern English, come to appears with an NP remnant and a PP remnant in Early Modern English, and talk of appears once with a PP remnant in Late Modern English and twice with an NP remnant (once in Early Modern English and once in Late Modern English). More precisely, these numbers indicate that while 11 out of 58 (18.96%) combinations appearing with NP remnants are listed in Claridge (2000), only 7 out of 150 (4.66%) combinations appearing with PP remnants are so listed. Unsurprisingly, the ratio of the total of PP remnants to the 7 combinations listed in Claridge (2000) is significantly different from the ratio of the total of NP remnants to the 11 combinations listed in Claridge (2000) (χ² = 8.6231, p < 0.05). Table 4 lists the combinations found in the current data and in Claridge (2000). Table 4: Verbs and prepositions found in the current data and in Claridge (2000) V-P combination belong to care for come to depend on discourse of dispose of hear of hope for look for talk of think of wait on wait for
66
Joanna Nykiel
This evidence clearly supports the hypothesis that multi-word verbs, and especially prepositional verbs, constitute an environment that promotes the use of NP remnants. However, it is important to note that such verbs do not force the use of NP remnants in Late Modern English, as is clear from their appearance together with PP remnants, as in (22). (22) To whose “war” did he belong? Felix answered, after a little hesitation, to the king’s levy. (1885 Richard Jefferies, After London; Literature online) Nor are prepositional verbs required for NP remnants to appear, as can be seen in examples (23) and (24). (23) A: Who did you walk with? B: Mrs. Ryder; we met her just at Tyburn turnpike; (1779 Charles Atwell, Sexual offences > sodomy, 20th October; Old Bailey proceedings) (24) PORTIUS: Ha! Brutus, take the sword and bravely plunge it! BRUTUS: In whom? PORTIUS: A wretch. (1773 Mercy (Otis) Warren, The adulateur. A tragedy; ARCHER) Notice that the remnants in these examples are NP remnants despite the semantic independence of the verb and preposition in (23): the verb walk can be processed without access to the preposition with and so can this preposition be processed without access to the verb. Equally semantically independent is the verb plunge in (24), although the preposition in requires access to it (I leave it to the reader to apply the entailment tests).
5 Discussion As predicted, PP remnants are preferred over NP remnants in Early Modern English, and much less consistently so in Late Modern English, paralleling the increase in the frequency of prepositional verbs. This result confirms the hypothe sis that semantic dependencies between prepositions and verbs have been developing gradually since Middle English, affecting ellipsis alternation. Also confirmed is the further hypothesis that NP remnants appear in environments that host prepositional verbs more often than PP remnants do. This evidence strongly suggests that what distinguishes Present-Day English from the earlier periods is differences in the strength of semantic dependencies, which are in turn linked to the number of prepositional verbs. Recall that ellipsis alternation exhibits a strong preference for NP remnants over PP remnants in Present-Day English.
Semantic dependencies and the history of ellipsis alternation
67
Semantic dependencies are also what distinguish Present-Day English from other languages (Nykiel 2014). The long-standing account of which remnants are tolerated in English attributes the availability of both NP and PP remnants to the possibility of stranding prepositions in nonelliptical clauses (Merchant 2001). Preposition stranding permits the following derivation of a remnant from a full clause: (25) Pat played in one of the houses on the hill, but I’m not sure which house Pat played in. The remnant is first fronted and then left behind after the rest of the clause deletes, including the stranded preposition. On this proposal, in languages that have a ban on preposition stranding in nonelliptical clauses, there is no possi bility of deriving NP remnants. However, NP remnants are available, though less common than PP remnants, in languages without preposition stranding (Rodrigues, Nevins, and Vicente 2009; Caha 2011; Nykiel 2013a). In addition, the current evidence, as well as the evidence presented in Nykiel (2013b, 2014) strongly suggest that the syntax of nonelliptical clauses may be less relevant to ellipsis alternation than is assumed by Merchant (2001). Rather, languages differ in terms of the range of semantic dependencies available between relevant lexical categories, which leads to a preference, or otherwise, for NP remnants. By hypothesis, without the ability to develop semantic dependencies between the relevant lexical categories, ellipsis alternation in Present-Day English would be subject to the same constraints as those found in other languages, that is, PP remnants would still be preferred over NP remnants. The current data provide compelling evidence in favor of at least one principle of language processing affecting ellipsis alternation. This principle is a conventionalized performance preference, located outside the competence grammar, which is seen in the gradual increase of NP remnants. This increase parallels the spread of prepositional verbs beginning in the Middle English period. The performance preference that becomes conventionalized is the preference for efficiently accessing constituents that form processing domains and hence are semantically dependent on each other. This preference is robust in constituent ordering in nonelliptical sentences (Hawkins 2000, 2004; Wasow 2002), as well as in ellipsis alternation in Present-Day English (Nykiel 2014). Hawkins (2000, 2004) explains the preference’s presence in nonelliptical sentences by proposing that the parser accesses semantically dependent constituents efficiently if they are adjacent; otherwise, recognition of the relevant processing domain is slowed down. For ellipsis, Nykiel (2014) suggests that processing domains are accessed incompletely if one constituent of a domain (preposition) is left in a remnant, while the other is missing (verb), and that the parser avoids such configurations. This preference grows stronger with an increase in the number of prepositional verbs.
68
Joanna Nykiel
At the same time that the current data point to the involvement of semantic dependencies in ellipsis alternation, they reveal that such dependencies are not the only constraint on this alternation. If they were, the frequency data presented in Tables 2 and 3 could be more readily transformed into evidence for the involvement of semantic dependencies in ellipsis alternation.
6 Conclusion This chapter has offered a diachronic perspective on ellipsis alternation seen in English sluicing and Bare Argument Ellipsis. I have explored the hypothesis that the relative freedom to omit prepositions from remnants has not been stable in English. In particular, I have proposed that this freedom is linked to the devel opment of semantic dependencies between prepositions and verbs. As data from Old English through Late Modern English show, NP remnants are first attested in Early Modern English but remain less frequent than PP remnants in this period. They gradually gain in frequency as of Late Modern English, although their frequency does not yet exceed that of PP remnants in this period. Because the frequency data for Early and Late Modern English do not paint a sufficiently clear picture of the involvement of semantic dependencies in ellipsis alternation, I analyze contexts in which NP and PP remnants appear. This analysis reveals evidence that the development of multi-word verbs has played a role in the distribution of remnants, suggesting that the English ellipsis alternation has a special status relative to other languages.
Acknowledgments For helpful comments and suggestions on this manuscript, I thank an anonymous reviewer and the editors of this volume, Laurel Brinton, Michael Adams, and Robert Fulk. I am also grateful to the audience at SHEL-7 in Bloomington, IN.
Sources ARCHER: A representative corpus of historical English registers. 2013. Version 3.2. http://www. alc.manchester.ac.uk/subjects/lel/research/projects/archer/ Complete works of William Shakespeare: His plays and poetry. CD-ROM. 1992 edn. Portland: Creative Multimedia. Corpus of Early English correspondence sampler. 1998. Compiled by Terttu Nevalainen, Helena Raumolin-Brunberg, Jukka Keränen, Minna Nevala, Arja Nurmi & Minna Palander-Collin at the Department of Modern Languages, University of Helsinki.
Semantic dependencies and the history of ellipsis alternation
69
A corpus of English dialogues 1560–1760. 2006. Compiled under the supervision of Merja Kytö (Uppsala University) & Jonathan Culpeper (Lancaster University). http:// www.engelska.uu.se/Research/English_Language/Research_Areas/Electronic_Resource_ Projects/A_Corpus_of_English_Dialogues/ Corpus of Middle English prose and verse. 2006. http://quod.lib.umich.edu/c/cme/ Dictionary of Old English web corpus. 2009. Antonette diPaolo Healey, Editor, with John Price Wilkin & Xin Xiang. http://tapor.library.utoronto.ca/doecorpus/ The Helsinki corpus of English texts. 1991. Department of Modern Languages, University of Helsinki. Compiled by Matti Rissanen, Merja Kytö, Leena Kahlas-Tarkka, Matti Kilpiö, Saara Nevanlinna, Irma Taavitsainen, Terttu Nevalainen & Helena Raumolin-Brunberg. The Lampeter corpus of Early Modern English tracts. 1999. Compiled by Josef Schmied, Claudia Claridge & Rainer Siemund. www.tu-chemnitz.de/phil/english/chairs/linguist/real/ independent/lampeter/lamphome.htm Literature online, third edition. 1996–2014. http://lion.chadwyck.co.uk/marketing/index.jsp Old Bailey proceedings online, 1674–1913. Compiled by Tim Hitchcock, Robert Shoemaker, Clive Emsley, Sharon Howard & Jamie McLaughlin, et al. www.oldbaileyonline.org, version 7.0, 24 March 2012.
References Baayen, R. Harald. 2008. Analyzing linguistic data: A practical introduction to statistics using R. Cambridge: Cambridge University Press. Baltin, Mark & Paul M. Postal. 1996. More on reanalysis hypotheses. Linguistic Inquiry 27(1). 127–145. Behaghel, Otto. 1932. Deutsche Syntax: Eine geschichtliche Darstellung. Band IV: Wortstellung. Periodenbau. Heidelberg: C. Winter. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. Longman grammar of spoken and written English. Harlow: Longman. Boland, Julie E. & Heather Boehm-Jernigan. 1998. Lexical constraints and prepositional phrase attachment. Journal of Memory and Language 39. 684–719. Bresnan, Joan. 2007. Is knowledge of syntax probabilistic? Experiments with the English dative alternation. In Sam Featherston & Wolfgang Sternefeld (eds.), Roots: Linguistics in search of its evidential base, 75–96. Berlin & New York: Mouton de Gruyter. Bresnan, Joan & Jennifer Hay. 2008. Gradient grammar: An effect of animacy on the syntax of give in New Zealand and American English. Lingua 118(2). 245–259. Brinton, Laurel J. & Elizabeth Closs Traugott. 2005. Lexicalization and language change. Cambridge: Cambridge University Press. Caha, Pavel. 2011. Case in adpositional phrases. Manuscript, CASTL, Tromsø. Claridge, Claudia. 2000. Multi-word verbs in Early Modern English: A corpus-based approach. Amsterdam & Atlanta, GA: Rodopi. Cruse, D. Alan. 1986. Lexical semantics. Cambridge: Cambridge University Press. Denison, David. 1985. Why Old English had no prepositional passive. English Studies 3. 189–204. Denison, David. 1993. English historical syntax: Verbal constructions. London: Longman. Fischer, Olga. 1992. Syntax. In Norman Blake (ed.), The Cambridge history of the English language. Vol. 2: 1066–1476, 207–408. Cambridge: Cambridge University Press.
70
Joanna Nykiel
Ford, Marilyn & Joan Bresnan. 2013. Using convergent evidence from psycholinguistics and usage. In Manfred Krug & Julia Schlüter (eds.), Research methods in language variation and change, 295–312. Cambridge: Cambridge University Press. Goh, Gwang-Yoon. 2001. Why prepositional stranding was so restricted in Old English. Korean Journal of English Language and Linguistics 1(1). 1–17. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press. Hawkins, John. 1999. Processing informativity and filler-gap dependencies across grammars. Language 75(2). 244–285. Hawkins, John. 2000. The relative order of prepositional phrases in English: Going beyond manner–place–time. Language Variation and Change 11. 231–266. Hawkins, John. 2004. Efficiency and complexity in grammars. Oxford: Oxford University Press. Hoffmann, Thomas. 2011. Preposition placement in English: A usage-based approach. Cambridge: Cambridge University Press. Hornstein, Norbert & Amy Weinberg. 1981. Case Theory and preposition stranding. Linguistic Inquiry 12(1). 55–91. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Kegl, Judy & Christiana D. Fellbaum. 1988. An analysis of obligatory adjuncts: Evidence from the class of measure verbs. Proceedings from the Eastern States Conference on Linguistics (ESCOL) 5. 275–288. Kruisinga, Etsko. 1925. A handbook of present-day English. 4th edn. Utrecht: Keminh En Zoon. Lohse, Barbara, John Hawkins & Thomas Wasow. 2004. Domain minimization in English verb-particle constructions. Language 80(2). 238–261. Merchant, Jason. 2001. The syntax of silence: Sluicing, islands, and identity in ellipsis. Oxford & New York: Oxford University Press. Nykiel. Joanna. 2013a. Clefts and preposition omission in sluicing. Lingua 123. 74–117. Nykiel, Joanna. 2013b. Constraints on the ellipsis alternation: A view from the history of English. Under revision. Nykiel, Joanna. 2014. The ellipsis alternation: Remnants with and without prepositions. Submitted. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A comprehensive grammar of the English language. London & New York: Longman. Rodrigues, Cilene, Andrew Nevins & Luis Vicente. 2009. Cleaving the interactions between sluicing and preposition stranding. In Danièle Tork and W. Leo Wetzels (eds.), Romance languages and linguistic theory 2006, 175–198. Amsterdam & Philadelphia: John Benjamins. Szczegielniak, Adam. 2008. Islands in sluicing in Polish. In Natasha Abner & Jason Bishop (eds.), Proceedings of the 27th West Coast Conference on Formal Linguistics, 404–412. Somerville, MA: Cascadilla Proceedings Project. Traugott, Elizabeth Closs. 1999. A historical overview of complex predicate types. In Laurel J. Brinton & Minoji Akimoto (eds.), Collocational and idiomatic aspects of composite predicates in the history of English, 239–260. Amsterdam & Philadelphia: John Benjamins. Tseng, Jesse. 2000. The representation and selection of prepositions. Edinburgh: University of Edinburgh dissertation. Vicente, Luis. 2008. Syntactic isomorphism and non-isomorphism under ellipsis. Manuscript. University of California at Santa Clara. Wasow, Thomas. 2002. Postverbal behavior. Stanford, CA: CSLI Publications. Wolk, Christoph, Joan Bresnan, Anette Rosenbach & Benedikt Szmrecsanyi. 2013. Dative and genitive variability in Late Modern English: Exploring cross-constructional variation and change. Diachronica 30(3). 382–419.
David West Brown and Chris C. Palmer
The phrasal verb in American English: Using corpora to track down historical trends in particle distribution, register variation, and noun collocations 1 Introduction Among their varied findings, corpus-based studies of phrasal verbs in English (such as run up in They always run up our electric bill) have pointed to two intriguing patterns. First, they have described phrasal verb distributions across registers in Present-Day English, showing their greater propensity for being real ized in fiction and spoken text-types (e.g., Biber et al. 1999; Liu 2011). Second, scholars have suggested that the use of phrasal verbs has increased over time (e.g., Spasov 1966; Pelli 1976; Martin 1990; Smitterberg 2008). Our study builds from and draws connections between these two lines of research by combining diachronic and register-based approaches to analyze the development of the phrasal verb in American English. In order to carry out the analysis, this study uses a variety of data sources: The corpus of historical American English (COHA) and The corpus of contemporary American English (COCA), which are large, monitor corpora, each containing over 400 million words (Davies 2008–; Davies 2010–); qualitative examples of usage from historical archives; and historical records of the production and consumption of material goods. The investigation comprises two parts. In the first, we present an overview of the development of phrasal verbs in the last two hundred years. Our analysis begins with historical trends in the use of the common adverbial particles and then examines diachronic changes in the frequencies of those particles across registers. In the second part of our investigation, we present case studies of the transitive phrasal verb pick up and selected noun collocates. These case studies explore how the use of phrasal verbs is influenced by the histories of the nouns that function as their objects (how the noun phone impacts the use of pick up, for example). One set of data, therefore, provides a global view of developmental trends focusing on the adverbial particles, the other a close look at some of the grammatical, semantic, and sociolinguistic forces that can influence the histories of specific phrasal verbs. Overall, our study aims to contribute to the wider body of research on phrasal verbs, as well as more specific conversations about the diachronic development
72
David West Brown and Chris C. Palmer
of phrasal verbs in American English. Furthermore, it places such data on phrasal verbs within broader theoretical and methodological contexts. Thus, this article aims not only to address the possibilities and complications of researching multiword units, but also to expand the methodological possibilities of analyzing them via corpus-based studies. Most specifically, our paper argues that, in order to create a full account of the growth or decline of a linguistic form, it is necessary to examine large, continuous, multi-register corpora, rather than to rely solely on an aggregation and comparison of disparate frequency-based studies of that form drawn from corpora of different sizes and registers. By investigating phrasal verbs within large corpora such as COHA and COCA, we are able to push existing quantitative approaches in several underexamined and fruitful directions: (1) the effect of register on the history of phrasal verbs and their particles; (2) the correlations between changes in the use of an individual phrasal verb and changes in nominal objects that frequently collocate with that verb; and (3) the impact of historical change in culture and technology on diachronic changes in phrasal verbs. Moreover, in order to carry out our investigations, we bring data from the material world into conversation with linguistic data, and probe their statistical relationships. These methods lead to several interesting conclusions. One particular register, fiction, contributes to the bulk of phrasal verb growth in nineteenth- and twentieth-century American English. Across multiple registers, verbs accompanied by particles up and out are the most frequently used phrasal verb forms. And the growth of the most frequent phrasal verb with up – i.e., pick up – can be explained, in part, by changes in American fashion, transportation, and communication technologies.
2 Diachronic studies of the growth of phrasal verbs in English Previous quantitative studies of diachronic changes in phrasal verbs, which have focused primarily on frequencies, have shown a general increase in phrasal verb usage from Middle English to Present-Day English. Martin (1990: 101–102) samples letters, from the Paston collection up to private collections in the late twentieth century, illustrating a change from approximately 2000 phrasal verb tokens per million words1 in the period 1440–1479 to approximately 10,900 tokens per million in 1975–1987. Spasov (1966: 19–21) provides a study of phrasal verb
1 Unless otherwise noted, all token frequencies cited from other studies or from our own data represent normalized counts of tokens of phrasal verbs per one million words of total text.
The phrasal verb: particle distribution, register variation, and noun collocations
73
frequencies relative to verb frequencies in British plays from the late medieval period to 1958. Looking only at dialogue, he argues that the percentage of verbs that are phrasal “shows an almost uninterrupted increase”2 from the medieval period to the modern – from 3.80% in Middle English, to 5.25% in Shakespeare, to 6.33% in the early eighteenth century, to 12.86% in the post-World War II era (1966: 18). Pelli (1976: 97–127) turns to American English, also examining phrasal verb use in drama. His study similarly shows an overall increase in phrasal verb usage from the late eighteenth century (e.g., approximately 5853 phrasal verb tokens per million in 1765–1775) to the later twentieth century (e.g., approximately 13,780 tokens per million in 1965–1972) (1976: 98–99). Broadening the range of genres investigated but narrowing the historical focus, Claridge (2000) compares phrasal verb frequencies in The Lampeter corpus, a multi-register corpus covering 1640– 1740, and in LOB, The Lancaster-Oslo/Bergen corpus, covering multiple genres from present-day British English. She discovers 3640 phrasal verb tokens per million in Lampeter, but more than twice that number – 8536 tokens per million – in the LOB material (2000: 126–127). Smitterberg (2008) examines progressive and phrasal verbs in British English within the multi-register CONCE, A corpus of nineteenth-century English. Contrasting the periods 1800–1830 to 1870–1900, he, too, finds an overall increase in phrasal verb use, from approximately 5100 tokens per million at the beginning of the century to approximately 6800 per million at its end (2008: 275). In order to build upon these previous studies, we use COHA, a four-hundredmillion-word corpus covering the period 1810–2009, as our primary source of data. The corpus is tagged for part of speech (including for verbs and the particles of phrasal verbs). It also covers every decade within this period, with ample data from a number of distinct registers: fiction, newspapers, magazines, and non-fiction books. The sheer size, range, and continuity of COHA allow for investigations into phrasal verbs that would be impossible to conduct in studies of small corpora. The first of these investigations occurs in Section 3, in which we examine the development of phrasal verbs within individual registers from decade to decade over the last two centuries of American English. In Section 4, the history of some of the most frequent particles is explored. And Section 5 illustrates the ways in which collocation patterns for the particle up reveal its semantic history and motivations for use. To search for verbs, particles, and phrasal verbs in COHA, we followed the methods outlined in Liu’s (2011) study of phrasal verbs in COCA, which uses the
2 According to Spasov’s charts and graph (1966: 20–21), there is a notable decrease in phrasal verb use to 4.33% of total verbs during the second half of the eighteenth century. By the end of the nineteenth century, he finds a significant increase to 8.88%.
74
David West Brown and Chris C. Palmer
same tagger (CLAWS) as COHA. Verbs were extracted from the corpus by looking for the tag [v*], and particles with the tag [rp*]. Searches for individual phrasal verb constructions considered the tag [v*] [rp*] (e.g., pick up), [v*] * [rp*] (e.g., pick him up), and [v*] * * [rp*] (e.g., pick the guy up). Further separation between the verb and particle was not targeted in searches, since this pattern is infrequent and apt to return false phrasal verbs (Liu 2011: 665). Figure 1 provides the frequencies of particles used in phrasal verb con structions decade by decade in COHA, while Figure 2 provides the total number of verbs (those appearing in both phrasal and non-phrasal constructions). It is immediately clear from these two charts that the COHA data corroborate the general trends observed in previous diachronic studies: namely, that phrasal verb constructions have largely been increasing in use over time, including for most of the last two centuries in American English. To determine whether or not changes in particle use were primarily dependent on overall changes in verb frequencies, we compared the trends in verbs with trends in the particles (which combined with a subset of those verbs to create phrasal verb constructions). The diachronic distributions of particles and verbs are correlated; as measured by Kendall’s-τ (Kendall 1938), the correlation coefficient is 0.786 (p = 0.0055).3 While the Kendall’s-τ suggests a significant overall trend, it does not identify or explain any stages or differences in the underlying distributions. In order to further explore the relationship between particles and verbs, we follow the methods set forth in Hilpert and Gries (2009). Specifically, we use variability-based neighbor clustering (VNC), which partitions diachronic data into statistically meaningful sub-periods (Hilpert and Gries 2012). The results of that clustering are plotted in Figures 3 and 4.
Figure 1 (Left): Token frequencies of particles appearing in phrasal verb constructions, 1810–2009, COHA Figure 2 (Right): Token frequencies of all verbs, 1810–2009, COHA 3 All statistical analysis for this study was carried out using R (R Core Team 2013).
The phrasal verb: particle distribution, register variation, and noun collocations
75
The dendrogram in Figure 3 shows particles separating into three clusters (as highlighted by the numbered, dashed lines): 1810–1859, 1860–1909, and 1910–2009. Alternatively, the dendrogram in Figure 4 shows verbs separating into five clusters: 1810–1829, 1830–1849, 1850–1869, 1870–1899, and 1900–2009. Thus, although there is a general correlation between the diachronic increase in particles and verbs overall, they do not grow entirely in parallel. The nineteenth century is particularly distinct in this regard. It sees changes in verb frequencies clustered into four statistically meaningful sub-periods, and particles into only two, showing their early histories in American English to be characterized by more volatility for verbs than for particles. The differences in the growth of particles and the increases in total verbs are further illustrated in Figure 5, which presents the ratio of frequencies of particles to frequencies of total verbs within each decade of COHA. Between 1810 and 1940, even as the use of verbs generally increased, the use of particles in phrasal verb constructions also increased. In other words, the percentage of verbs combining with particles to create phrasal verbs rose steadily during this period, from less than 2% of verbs in 1810 to over 5% of verbs in 1940. It is also worth noting that this ratio has plateaued since 1940, with a small decrease in the ratio of particles to total verbs between the 1940s and the 1980s and a slight rise within the last two decades. These data somewhat corroborate the findings of Pelli (1976: 98–99), who notes a similar trend in particle frequencies within American drama: after steadily increasing in use from the early nineteenth century to the 1925–1935 sub-period, his next subperiod shows a decrease – from roughly 16,000 tokens in 1925–1935 to about 13,780 tokens in 1965–1972. Unfortunately, Pelli’s study ends there, so further comparisons
Figure 3 (Left): VNC dendrogram for particles appearing in phrasal verb constructions, 1810–2009, COHA Figure 4 (Right): VNC dendrogram for all verbs, 1810–2009, COHA
76
David West Brown and Chris C. Palmer
Figure 5: Ratio of token frequencies of particles in phrasal verb constructions to frequencies of total verbs, normalized to occurrences of particles for every 100 verbs, 1810–2009, COHA
in recent decades cannot be made. Even so, the data from COHA provide a more comprehensive picture of the overall diachronic changes in verbs and particles from the last two centuries, nuancing our understanding of the contours of change. While it is clear that phrasal verbs have been on the rise in American English over the last two centuries, that growth seems to have leveled off in the 1940s.
3 Phrasal verbs and register variation in Present-Day English and the history of American English As the discussion in the previous section makes clear, it is necessary to nuance broad diachronic claims about the growth of phrasal verbs in English. While many studies, including the present one, have generally shown upward trends in the use of phrasal verbs and particles over the centuries, the data from COHA suggest that such growth plateaued in American English beginning in the 1940s. Indeed, Thim (2012: 211–214) cautions that it may not be wise to assume there is consistent growth in phrasal verbs diachronically, especially if frequency studies are used as a sole indicator of growth. Citing a number of frequency studies from different eras (2012: 213), he notes that there is certainly not a steady increase from Middle English to Present-Day English. For example, Castillo (1994: 439–450) finds a frequency of 6500 phrasal verb tokens per million words in Shakespeare; Smitterberg (2008: 275) discovers an average of 5900 tokens in multiple registers of nineteenthcentury English; and Biber et al. (1999: 408–409) reports roughly 1400 phrasal verb tokens in Present-Day English within the multi-register Longman spoken
The phrasal verb: particle distribution, register variation, and noun collocations
77
and written English corpus. These studies alone would suggest a sharp decline in phrasal verb usage, and run counter to the narrative suggested by Spasov (1966), Pelli (1976), and Martin (1990). Thim’s main claim is that frequency largely depends on the type of study conducted on individual registers at particular points in time. He argues that, rather than relying on “general patterns of growth” to explain diachronic change in phrasal verbs, scholars should look deeper into “the development of aspectual and idiomatic constructions over time” and into “changing frequencies in individual text types rather than an overall increase of the construction type” (2012: 214). While Section 5 considers historical changes in idiomatic constructions containing phrasal verbs, the present section digs deeper into the relationship of text-type and diachronic change in phrasal verbs. Studies of the interplay between phrasal verbs and register in Present-Day English indicate that phrasal verb frequency may be a reliable marker of texttype. Dempsey, McCarthy, and McNamara (2007: 221), for example, argue that phrasal verb frequency correlates strongly with the spokenness and informality dimensions of the spoken/written and informal/formal clines used to classify genres in studies such as Biber (1988). In Biber et al.’s corpus study of spoken and written English (1999: 409), phrasal verbs were found to be most frequent in the registers of conversation and fiction, somewhat frequent in newspapers, and least frequent in academic writing. Liu (2011: 674) presents cross-register data on phrasal verbs from COCA, which we re-present in graphic form (see Figure 6). The
Figure 6: Graphic representation of cross-register variation of phrasal verbs in COCA, normalized to occurrences per million words, from Liu (2011)
78
David West Brown and Chris C. Palmer
results in Liu (2011) corroborate the other studies of Present-Day English phrasal verbs: they tend to occur most frequently in the more oral registers of fiction and speech; they have moderate frequencies in magazines and newspapers; and they have the lowest frequencies in academic prose. But can this sort of distribution be observed in earlier historical periods? Smitterberg (2008) provides the only broad historical study of register variation and phrasal verbs within late modern English. He finds that in nineteenthcentury British English, the registers of drama, fiction, and letters show the great est increases in phrasal verb frequencies; debates and history show possible moderate increases; and trials and scientific texts show little to no increases (2008: 276–277). While his analysis follows a line of reasoning similar to that of other studies – namely, that the more colloquial registers are most likely to show increases in phrasal verb usage – it is difficult to compare his results, which concern British English only as late as 1900, to those of Liu (2011) on contemporary American English. COHA, however, provides data on American English up to 2009 for four registers (fiction, magazines, newspapers, and non-fiction), three of which also appear in Liu’s (2011) analysis. Hence, by looking at distributions of phrasal verbs among the registers of COHA, we can compare the past and present of American English, determining whether or not phrasal verbs were distinguishing markers of genre well before the present day. Figure 7 provides a historical distribution of phrasal verbs by register. Broadly speaking, the historical data from COHA reflect register patterns observed in the present day: the most speech-like genre within the
Figure 7: Cross-register variation of phrasal verbs in COHA, normalized to occurrences per million words. Note that newspaper data do not appear until the 1860s in COHA
The phrasal verb: particle distribution, register variation, and noun collocations
79
corpus – fiction – exhibits far higher frequencies of phrasal verbs compared to other registers in every decade of the nineteenth and twentieth centuries. But it is significant to note that these differences have become more marked over time. In the early nineteenth century, fiction’s use of phrasal verbs was already higher than, and yet much closer to, the frequencies observed in magazines and nonfiction compared to what can be observed in the late 1900s. Non-fiction, which includes academic writing, shows a steady decline over time. Newspapers and magazines show general increases during these centuries. Even though Figure 6 shows no significant difference between magazines and newspapers in PresentDay English, the data from COHA in Figure 7 suggest that, historically speaking, magazines have had consistently higher frequencies of phrasal verbs than have newspapers. One explanation for these changes can perhaps be found in Smitterberg (2008: 286), who argues that increases in phrasal verb frequencies within individual text-types are a marker for the “colloquialization” of that text-type. He defines colloquialization as “the process whereby a linguistic feature that occurs more frequently in conversational speech than in writing becomes more common than previously in some written genres” (2008: 271). Smitterberg views colloquialization as a prominent development of nineteenth-century British English, in which fiction, drama, and letters adopt increasingly less formal linguistic features – which he tentatively offers as a reflection of the “informalization and democratization” of British society during the nineteenth and twentieth centuries (2008: 282). However, other genres, such as science, have resisted such colloquialization (2008: 278). If phrasal verbs are colloquial markers, it stands to reason that the data in COHA are showing developments parallel to those observed in the registers of Smitterberg’s study. American fiction, like British fiction, is perhaps being increasingly colloquialized by writers during this period, as are (to a lesser extent) magazines and newspapers. Non-fiction, which includes academic and scientific writing, shows more resistance to colloquialization. Another potential factor affecting phrasal verb frequencies may have been editorial practices, especially those enforcing word limits and prescriptive rules. While every writer in every genre represented in COHA may have had word limits to follow before publication, it is arguably the case that writers of shorter genres – such as newspaper and magazine articles – have had much lower word counts (and thus tighter restrictions) to work with than writers of long fiction. Non-fiction works, which include academic monographs and essay collections, might have faced similar editorial demands on word counts. If so, there may have been more pressure on writers in registers such as magazines, newspapers, and nonfiction to choose one-word verbs (e.g., choose, select) rather than phrasal verbs
80
David West Brown and Chris C. Palmer
(e.g., pick out).4 In addition to word-limits, Thim (2012: 235–238) speculates that emerging prescriptive norms in the eighteenth century – namely, rules against stranded and sentence-final prepositions (many of which were confused with adverbial particles) and against monosyllabic idioms (which would include many phrasal verbs) – might have impacted the use of phrasal verbs well into the twentieth century. Thim cites several grammars that tried to regulate these features, and it seems likely that editors of magazines, newspapers, and academic books would be most influenced by such prescriptivism. Perhaps fiction writers, even when meeting their own editorial demands, were facing less prescriptive pressure on their choice of verbs than were journalists and academics. However, without direct evidence of editorial practices targeting phrasal verbs due to word limits or prescriptive attitudes, these proposed explanations for historical register variation must remain speculative. Overall, the investigation of register variation in COHA confirms that phrasal verbs experienced the largest increase in the most speech-like genre, fiction. There were moderate increases in magazines and newspapers, which mix oral and written features, with a small decrease in the least oral register of COHA, non-fiction. More importantly, these data allow us to further nuance claims about the overall trends observed in Figures 1 and 7: not all registers have contributed equally to the rise in use of phrasal verbs in American English. If anything, the vast majority of increase in use as COHA captures it has been due to fiction, which rises sharply until 1940 and plateaus thereafter.
4 Variation of particles in the history of American English Few studies have compared historical changes in the use of individual particles in phrasal verb constructions within English. Spasov (1966: 24) provides a table with frequencies of each particle within each sub-period of British English, though no trend lines are analyzed, perhaps due to the small size of his corpus and low frequencies of occurrence for many particles. Even so, up seems to be the most frequent particle in each century from Middle English to Present-Day English, with
4 Anecdotal evidence suggests that word counts were historically relevant: An 1897 edition of The Writer, an American magazine for aspiring writers edited by William H. Hills, states that “articles should be closely condensed (…) [An] ideal length is about 1000 words” (38). In a 1902 edition of The Writer, there is an ad for Comfort magazine which creates an incentive for pieces that are one hundred words or fewer, so that publishing two one-hundred-word anecdotes is worth $2, while one longer two-hundred word anecdote is only $1.50 (176). Garrett (2002: 85) reports that in the early nineteenth century, Mary Shelley felt the pressure of word-limits more on her shorter writing (e.g., book reviews) than on her longer fiction.
The phrasal verb: particle distribution, register variation, and noun collocations
81
perhaps out and down the next most frequent. Martin (1990: 109–111) provides a breakdown of particles within her diachronic corpus of letters, noting that up and out are the most frequent particles in most centuries. She notes some changes as well: in American English, out and up show similar frequencies in the eighteenth century, though out becomes more frequent than up in the twentieth. Pelli (1976: 111–114) provides frequencies for several of the most productive particles within his corpus of American drama, calculating the tokens of each particle type as a percentage of total tokens of phrasal verbs within each sub-period. According to his data, up is the most frequent particle in all sub-periods from 1765 to 1972. The particle out is the second most frequent in all sub-periods, with back always the third most frequent. Little change can be observed in Pelli’s data from sub-period to sub-period, with the exception of back, which steadily increases from the late eighteenth century to the mid-twentieth century. Similar frequencies in PresentDay British English are recorded by Gardner and Davies (2007: 346), who use the British national corpus (BNC) as their source of data. They find up, out, and back to be the three most frequent particles, with down nearly as frequent as back. Furthermore, they find adverbial particle out accounts for a full 97.3% of all occurrences of out, and adverbial particle up 87.4% of all occurrences. To investigate a broader range of registers and particles, we examined the individual histories of the “six common adverbial particles” listed in Biber et al. (1999: 413) – up, out, down, on, in, off – along with back. Figure 8 presents
Figure 8: Percentages of selected particles out of the total of all particles in COHA
82
David West Brown and Chris C. Palmer
the trajectories of the various particles as percentages of total particles. In combination with the data presented in Figures 5 and 7, Figure 8 shows that the overall growth in phrasal verbs has been evenly distributed, for the most part, among the most common particles. The particle up shows a modest increase in the twentieth century, accounting for a little more than 22% of all particles at the beginning of the century and nearly 25% by the end. Other fluctuations are relatively minor. The exception is back, which more than doubles its percentage, going from 5.68% to 11.92%. In fact, back surpasses down at the end of the twentieth century, as down has been declining since about 1940. Within individual registers, these particles tend to follow the broader trends observed for all phrasal verb types in Figure 7. Even so, by examining particle trends, some differences can be observed within individual genres. For example, consider Figure 9, which presents data on particle trends within the genre of newspapers. Like the general trends, the growth of particles in newspapers is not marked by radical changes in distributions. Some of the particles follow similar trajectories in the newspaper text-type. The particle down, for example, decreases as a percentage of the total after 1940 and is overtaken by back near the end of the twentieth century. There are, however, also some register-specific patterns. Compared to Figure 8, in which up is consistently more frequent than out when taking all registers in the aggregate, in newspapers it is only after the 1940s that up begins to lead out with any consistency.
Figure 9: Percentages of selected particles out of the total of particles in the COHA newspaper text-type
The phrasal verb: particle distribution, register variation, and noun collocations
83
An important finding from this investigation is that not all particles c ontribute equally to the growth of phrasal verbs. It is clear, for example, from Pelli’s work on American drama and our own study of multiple registers in American English that up has been the largest contributor to the growth of phrasal verbs in Amer ican English. In other words, individual particle types seem to have their own individual histories, and those histories may differ somewhat within individual text-types. It is useful for frequency studies to account for these differences among particles to better understand larger patterns of growth among phrasal verbs.
5 A case study of pick up and its noun collocates In an effort to shed further light on trends in usage, we finally want to examine some of the history surrounding the use of a single, common phrasal verb: pick up. We choose to focus on this particular combination for a number of reasons. First, as we have noted, up is the most common adverbial particle in COHA. Second, pick both is the most frequent verbal lemma that collocates immediately to the left of up and has the highest association measure as calculated by Mutual Information (see Table 1). Its high MI score suggests not only that the verb pick has the greatest affinity for occurring with up among possible alternatives, but also that when pick appears, it is very likely to co-occur with up, as statistically significant association is usually established when MI > 3 (Church and Hanks 1990). By virtue of both its frequency and its degree of association, therefore, pick up is a notable and interesting candidate for a case study.
Table 1: The ten most frequent lemmatized verbs that collocate to the left of the adverbial particle up in COHA, with their counts, normalized frequencies (per million), percentage of tokens that co-occur with up, and association measure as calculated by Mutual Information Rank
Collocates
Count
1 2 3 4 5 6 7 8 9 10
[PICK] UP [LOOK] UP [COME] UP [MAKE] UP [GET] UP [GIVE] UP [TAKE] UP [SET] UP [GO] UP [STAND] UP
32629 31111 25591 24042 22750 21407 20204 16449 14013 13903
Frequency
Percent
80.32 76.58 63.00 59.18 56.00 52.70 49.74 40.49 34.50 34.22
48.77 4.97 3.03 2.51 3.05 4.05 2.76 9.50 1.31 5.02
MI 8.25 4.96 4.25 3.97 4.26 4.67 4.11 5.89 3.03 4.97
84
David West Brown and Chris C. Palmer
Figure 10: The frequency (per million words) of the lemmatized phrasal verb [pick] up and [pick] up with a personal pronoun [p*] between the verb and the particle (e.g., pick it up) in COHA
The frequency of pick up mirrors trends in the overall growth of phrasal verbs (see Figure 10). Its history is marked by a particularly dramatic increase in the early to middle part of the twentieth century (a period highlighted on the chart). In order to explore some potential reasons motivating this growth, we turn to collocational patterns that emerge during this period. These patterns not only help to explain trends in usage but also shed light on the mutually influential relationship between transitive phrasal verbs and their nominal objects. Research into V + N combinations has been undertaken from a number of different orientations. One prominent focus in the diachronic study of V + N phrases has been the development of composite predicates (verb + (article) + deverbal noun, e.g., give a talk), particularly as such development relates to the processes of lexicalization, grammaticalization, and idiomization (Brinton and Akimoto 1999; Brinton and Traugott 2005; Brinton 2008). V + N collocational patterns have also been the subject of interest in second language learning contexts, where, for example, research has explored the production of such patterns in English learner writing (Nesselhauf 2005; Laufer and Waldman 2007). Taken together, these d ifferent areas of research interest and emphasis highlight several important features of V + N combinations: (1) the relationships among verbs and co-occurring nouns are grammatically and semantically structured; (2) co-occurrence can be described probabilistically; (3) V + N relationships change over time; and (4) as Stubbs (1995: 387) puts it, “Culture is encoded not
The phrasal verb: particle distribution, register variation, and noun collocations
85
just in words which are obviously ideologically loaded, but also in combinations of very common words”. While all four of these points are relevant to our analysis here, it is the last two that we want to foreground in framing our discussion. That is, our analysis suggests the complex interactions between cultural changes and increas ing phrasal verb frequencies. Specifically, we show how changes in American culture can drive the use of nouns, which, in turn, can drive the use of a phrasal verb through its strong associations with those nouns. In doing so, our analysis further elaborates Rodríguez-Puente’s (2012: 83) observation regarding changes in the world and their influence on phrasal verbs. She argues, “Some phrasal verbs have recently been created or have developed new, specific meanings to cope with new technological developments”. In this section, we explore how developments in telephone and automobile technologies in the early twentieth century affect the use of pick up. Additionally, we investigate a counterexample, a word that has an early association with pick up: hat. Its decrease in use and association over time seems to correspond to a cultural change, namely, a change in fashion. Our interest in the connections among pick up, its noun collocates, and changes in the material world emerged from the data illustrated in Figure 11 and the posing of a simple question: Does it matter what is being picked up? In other words, are there any diachronic changes in the nouns in V + N phrases that might help us describe changes in the use of the verb itself? Searching for nouns within four words to the right of the verb pick up in both COHA and COCA yields the results presented
Figure 11: Frequencies (per million) of the lemmatized phrasal verb pick up collocating with phone, hat, and speed in COHA
86
David West Brown and Chris C. Palmer
Table 2: The ten most frequent nouns collocating within four tokens to the right of the lemmatized phrasal verb pick up in both COHA and COCA Rank
COHA Token
1 2 3 4 5 6 7 8 9 10
PHONE BOOK HAT SPEED TELEPHONE PAPER BAG RECEIVER PIECES GLASS
COCA
Count
Frequency
897 363 344 310 308 296 287 260 239 223
2.21 0.89 0.85 0.76 0.76 0.73 0.71 0.64 0.59 0.55
Token PHONE SPEED PIECES TAB BOOK PACE BAG PAPER SLACK RECEIVER
Count
Frequency
1909 483 411 383 359 329 249 215 210 199
4.11 1.04 0.89 0.83 0.77 0.71 0.54 0.46 0.45 0.43
in Table 2. The juxtaposition of the two data sets suggests some interesting points of comparison. The first is that in historical and current American English, phone is the most frequent collocate. Additionally, telephone and receiver are among the ten most frequent collocates in COHA, and receiver is among the ten most frequent in COCA. Thus, the lexeme phone, along with related derivations and meronyms, appears to have a strong connection to the verb pick up. Second, following phone, the word speed is the next most frequent collocate in current American English and the fourth most frequent historically. In the case of phone, the semantic and grammatical relationship between the noun and verb (in a phrase like pick up the phone) is fairly straightforward. In the phrase pick up speed, however, the meaning of pick up has been metaphorized to something like ‘gain’ or ‘increase’, and the whole phrase has been idiomaticized. Thus, the motivating factors for the devel opment of pick up speed would seem to be different from those for something like pick up the phone. Finally, while hat is the third most frequent collocate in COHA, it is the sixty-seventh most frequent in COCA. It, then, appears to be an example of change working against the prevailing trend. As opposed to pick up + phone or pick up + speed, the pattern pick up + hat seems to be a collocation falling out of use. To explain motivations for historical changes in the use of these various combinations with pick up, in the following sections we consider external factors, such as changes in technology and material production in American culture. 5.1 pick up + telephone/phone Although telephone exists in English as the name for a number of different devices throughout the nineteenth century, the technological developments of Alexander
The phrasal verb: particle distribution, register variation, and noun collocations
87
Bell and Thomas Edison in the 1870s reshape its meaning and spur its proliferation in American English (OED Online). This history is confirmed in COHA, where the number of occurrences begins to increase sharply beginning in 1878, one year after Bell and Edison file for patents and the same year the first commercial exchange opens (see, e.g., Fischer 1992). The shortening phone appears nearly concurrently; the two forms exist side by side, for example, in this newspaper feature from 1882, with quotation marks calling attention to the shortened form: (1) A man was ordering some meat from the market by telephone, and, after his order was in, he happened to think that he would like some liver, so he put his mouth to the concern and said, “Say, by the way.” The connection had been broken, and the attendant at the central office had her ear to the “phone” when the man said, “Have you got any liver?” The girl was nearly frightened to death, but she rallied enough to say, “Why, I suppose so; most girls have a liver. But why do you ask?” (“An important inquiry” 2) The lemmatized full form reaches its peak frequency in 1940, when it begins to decline in COHA. The lemmatized shortened form continues to increase in frequency throughout the twentieth century and surpasses telephone in COHA in the 1960s (132.08 per million vs. 116.45 per million).5 Perhaps unsurprisingly, as the technology itself becomes more widespread, so, too, do the words naming that technology. This parallel growth is demonstrated in Figure 12, where the number of telephones in the U.S. (Federal Communications Commission 1946, 1982) is plotted on one y-axis and the frequency of telephone and phone is plotted on a second. The two trends are highly correlated, with Kendall’s-τ = 0.956 (p 3 with a minimum of 5 occurrences (train MI = 7.14, plane MI = 5.88, car MI = 5.43). In COCA, only train and car have scores where MI > 3 (train MI = 6.83, car MI = 4.14). The phrase pick up speed can appear in a variety of contexts, as in the following example from 1916, where it is paired with another idiom (“the ball is rolling”). Its understood subject (something like “plans for American agriculture’s economic stability”) further metaphorizes the semantic content of the V + N phrase: (2) It has been an epoch-making work, for these fundamental steps which have been taken are the enabling acts that will eventually put American agriculture for the first time on a sound economic footing. Not all of the plans that have been devised along this line have been put into actual operation yet, but the ball is rolling and picking up speed from day to day. (COHA: MAG) Although such uses are not unusual, the collocational data show that pick up + speed has a strong association with transportation technologies. That association is attested in examples that predate any data from COHA, like this one from the
90
David West Brown and Chris C. Palmer
Proceedings of the Engineers’ Club (1892: 366), which the minutes attribute to comments made by William S. Aldrich, a professor: (3) In Chicago the overhead steam roads can hardly keep up in speed with the cable surface roads. The great advantage of the cable is that the car, after stopping, very soon picks up speed – in Philadelphia, indeed, almost too quickly. A steam train cannot do this. The cable gives the most rapid surface transit possible. After it comes steam, and then electricity. The storage battery is best on crowded streets, and the trolley on long stretches of open streets and park. It can pick up speed more quickly than the storage battery car. Notably, pick up + speed emerges in the U.S. around the same time as pick up + telephone/phone – in the latter part of the nineteenth century. It has a strong collocational association with rail transportation, which has a different trajectory in American discourse (the use of the noun train, for example, gains in frequency through the nineteenth century, peaks in the 1920s, then declines). This collocational pattern likely motivates some of the early growth of the idiom. However, the idiom’s other significant collocate, car, also influences its history. In Figure 14, the number of automobiles registered in the U.S. (Department of Transportation 2011) is plotted against the frequency of the words car/automobile. Like the wordobject data for telephone/phone, the Kendall’s-τ for the data in Figure 14 suggests a significant correlation: τ = 0.642 (p < 0.0001). The somewhat weaker correlation and differences in the tailing left ends of the charts can probably be attributed to two factors. First, automobile registration was not required early on in most states, so the number of registrations between 1900 and 1910 does not accurately reflect the number of cars. Second, as the quotation above illustrates, car does not necessarily refer to an automobile, but may refer to a trolley, train, or some other passenger compartment. Thus, the frequency is capturing more than references
Figure 14: The number of automobiles registered in the U.S. (in millions) plotted against the frequency of car/automobile (per million words) in COHA
The phrasal verb: particle distribution, register variation, and noun collocations
91
to automobiles. This range of meaning causes greater distortion early (when there are competing forms) than late (when car as automobile is the dominant form). For example, from 1910 to 1919 there are more bigrams of street car (n = 81) than motor car (n = 53). From 2000 to 2009, common bigrams include police car (n = 88) and rental car (n = 83), and there are more instances of rent-a-cop car (n = 28) than subway car (n = 17) or dining car (n = 15). The data, then, seem to indicate a history of pick up + speed that is somewhat more complex than that of pick up + telephone/phone. It is a collocation that develops as an idiom frequently used in association with rail technology, but is also associated with emergent automobile technology. As the latter proliferates, so too does the V + N phrase. But despite the differences in the grammatical and semantic relationships between the phrasal verb and various nouns signifying communication and transportation technologies, in each case the diachronic frequencies of pick up seem to be influenced by changes in the material world. It is worth noting, too, that in the U.S. these changes in telecommunications and transportation are happening concurrently. Fischer (1992), for example, remarks on their mirrored growth in the first half of the twentieth century and dual impact on American culture. These profound cultural changes are partly what lie behind the linguistic changes highlighted in Figure 10.
5.3 pick up + hat Finally, we want to briefly discuss an example that runs counter to the prevailing increase in the use of pick up: the collocation pick up + hat. Unlike pick up + telephone/phone and pick up + speed, we do not tie pick up + hat to changes in technology. However, the rise and fall of hats in American fashion has been interpreted as an indicator of cultural change (e.g., Lieberson 2000; Schudson 1984; Tenner 1989). Tenner (1989: 22), for example, describes images of crowds prior to World War II as “hatscape[s]”, and he connects the decline of hats to a decreasing preference for overt marking of social hierarchy. In his summary of some of the arguments advanced about the decline of hats in American fashion, Lieberson (2000: 82) lists claims such as the emergence of hair as a marker of self-expression; President Kennedy’s personal (often hatless) style; and the rise of cars, which are often difficult to enter and exit while wearing a hat. For his part, Lieberson (2000: 83) argues that the decline in the popularity of hats began earlier than most explanations assume. Their decreasing popularity was one feature of an overall decline in formality, which includes other forms of self-presentation besides clothing. In support of his claims, Lieberson presents data for hats and gloves from the Sears, Roebuck catalogue (see Figure 15). These figures capture the offerings
92
David West Brown and Chris C. Palmer
Figure 15: The popularity of men’s hats and women’s hats in the Sears, Roebuck catalog, 1920– 1990, from Lieberson (2000: 79). The standardized number on the y-axis represents an effort to control for the fluctuating size of the catalogue and is calculated by dividing the number of hats (excluding casual wear, winter hats, ski hats, and baseball caps) by the pages devoted to men’s underwear. Lieberson’s data is plotted against the frequency of hat in COHA.
of only one company (and are thus tied to its varying fortunes) and may not always reflect patterns of consumption. Thus, while there is a positive correlation between the word frequency and the standardized number (Kendall’s-τ = 0.713, p = 0.0141), there are significant limitations to that correlation. Even so, the catalog data intersect with the linguistic data in interesting ways and offer a kind of counterpoint to data presented in Figures 12 and 14. The frequency of the word hat peaks in 1910–1919 (133.56 per million) and remains relatively steady through 1940–1949 (128.39 per million), when it begins a steady decline. The collocational pattern of pick up + hat peaks in 1920–1929, a decade later than hat by itself, but otherwise mirrors the trend in use for the single word (see Figure 11). Thus, both pick up + hat and hat are gaining or maintaining frequency when Lieberson’s graph shows a decline. One possibility is suggested by the data presented in Olds’ (2011) study of the hat industry in Taiwan. Although it is data from outside the U.S., it shows the production of hats that were primarily intended for the U.S. market. His data reveal output rising until 1934, even as the value of that output decreases from the effects of the depression. Lieberson’s chart may be reflecting similarly complicated interactions among the demand for a thing, its availability, and its value during a period of economic decline. In the case of hat, the relationship between the circulation in the material world and circulation in language is certainly difficult to quantify in the early part of the twentieth century. However, the downward trends in both are clear after the 1940s. Additionally, hat decreases its association with pick up over the course of the twentieth century, dropping below the level of statistical significance in 2000
The phrasal verb: particle distribution, register variation, and noun collocations
93
(see Figure 13). This is in contrast to phone, which largely maintains its association with pick up, and speed, which increases its association with pick up over time. In this way, pick up + hat serves as an important caveat. It demonstrates that while some collocational patterns may reinforce prevailing trends in use of a particular constituent, there may be concurrent patterns working as counterweights. Pick up may be riding the coattails of nouns related to telecommunication and transportation technologies, but those are not the sole forces driving its diachronic frequencies. This latter point is key. While we demonstrate here the importance of collocational associations and the sometimes surprising connections between word frequencies and changes in the material world, these represent only some of the circumstances contextualizing the history of the phrasal verb pick up.
6 Conclusion On the basis of our study of phrasal verbs in American English, we offer the following considerations and conclusions: –– Frequency studies tell us some aspects of linguistic history. But that history can be a spotty one. In the case of phrasal verbs, studies of individual registers have painted a picture of mostly unimpeded growth from Middle English to Present-Day English. But such studies have made it difficult to make comparisons between eras and genres to evaluate the contours of such growth in the overall language, especially if only small corpora in randomly selected sub-periods are considered. Such findings can be better understood by comparing them with data drawn from larger, continuous, multi-register corpora such as COHA. Specifically, while we have found that phrasal verbs in the aggregate have continued to increase in use in nineteenth- and twentiethcentury American English, that growth seems to have plateaued around the 1940s. –– It is important to nuance our understanding of “growth” in phrasal verb use by exploring narrower quantitative developments within broader quantitative accounts – whether we mean “broad” in its synchronic sense (“through out the language in any one period”) or its diachronic one (“over a long period of time”). One way to achieve a more nuanced analysis is to consider changes within registers and changes among different particles. Perhaps confirming the increasing colloquialization of some genres, our investigation of register in American English finds that phrasal verbs have grown in use in fiction, newspapers, and magazines – with fiction being the greatest contributor to overall growth by far – while they have decreased in use in nonfiction. And among particles in American English, the particles up and out
94
David West Brown and Chris C. Palmer
continue to be some of the most common and increasingly frequent forms in phrasal verb constructions. –– Quantitative data become a richer form of evidence when put into the context of historical changes in culture. To achieve a richer analysis of diachronic change in phrasal verbs, we first used broad quantitative studies to zero in on individual data points worthy of further investigation – in this case, the most frequent particle historically (up) and its most frequent collocating verb form (pick). We then inspected the most frequent noun collocations of the verb and particle combined. To analyze the history of pick up and its nominal collocates, we juxtaposed the frequency of linguistic forms with data on historical changes within the material production of the real-life referents of those forms. In the case of this particular phrasal verb, it became clear that changes in both fashion (with reference to collocate hat) and technology (in communication, with reference to the collocate phone, and in transportation, with reference to speed) impacted changes in the frequency and semantics of pick up. While the particular impact of these historical devel opments cannot automatically be assumed to be relevant and explanatory for all phrasal verbs, our study demonstrates the value of considering such external factors in any detailed study of motivations for historical change for particular constructions.
Figure 16: Percentages in COCA of phone and cell appearing in phrases with lemmatized answer and pick up of the kind V + (*) + N (e.g., answers her phone vs. picks up the phone)
The phrasal verb: particle distribution, register variation, and noun collocations
95
In the specific case of pick up, it will be interesting to see how, or even if, more recent changes in telecommunications technologies will influence collocational patterns. While we have more phones than ever, we have new words to refer to those phones – notably, in the U.S., cell phone, or cell. It appears that while phone has preference for pick up over answer (e.g., picked up the phone), cell or cell phone has a preference for answer (e.g., answered the cell) (see Figure 16). It is possible that such changes may depress future frequencies of pick up. However, we will still be picking up speed, not to mention picking up the tab and picking up the pieces. It is just as likely, too, that pick up will develop new collocations and new meanings as phrasal verbs are an important, if sometimes obscure, trace of the interaction between the material world and language change.
References Biber, Douglas. 1988. Variation across speech and writing. Cambridge: Cambridge University Press. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. Longman grammar of spoken and written English. London: Longman. Brinton, Laurel J. 2008. “Where grammar and lexis meet”: Composite predicates in English. In Elena Seoane and María José López-Couso (eds.), in collaboration with Teresa Fanego, Theoretical and empirical issues in grammaticalization, 33–53. Amsterdam & Philadelphia: John Benjamins. Brinton, Laurel J. & Minoji Akimoto (eds.). 1999. Collocational and idiomatic aspects of composite predicates in the history of English. Amsterdam & Philadelphia: John Benjamins. Brinton, Laurel J. & Traugott, Elizabeth Closs. 2005. Lexicalization and language change. Cambridge: Cambridge University Press. Castillo, Concha. 1994. Verb-particle combinations in Shakespearean English. Neuphilologische Mitteilungen 95. 439–451. Church, Kenneth Ward & Patrick Hanks. 1990. Word association norms, mutual information, and lexicography. Computational Linguistics 16(1). 22–29. Claridge, Claudia. 2000. Multi-word verbs in Early Modern English. Amsterdam & Atlanta: Rodopi. Davies, Mark. 2008–. The corpus of contemporary American English: 450 million words, 1990– present. http://corpus.byu.edu/coca/ (accessed 1 September 2013). Davies, Mark. 2010–. The corpus of historical American English: 400 million words, 1810–2009. http://corpus.byu.edu/coha/ (accessed 1 September 2013). Dempsey, Kyle B., Philip M. McCarthy & Danielle S. McNamara. 2007. Using phrasal verbs as an index to distinguish text genres. In David Wilson & Geoff Sutcliffe (eds.), Proceedings of the twentieth international Florida artificial intelligence research society (FLAIRS) conference, 217–222. Menlo Park: The AAAI Press. Department of Transportation, United States. 2011. Highway statistics summary to 1995. http:// www.fhwa.dot.gov/ohim/summary95/section2.html (accessed 2 October 2013).
96
David West Brown and Chris C. Palmer
The Engineers’ Club of Philadelphia. 1892. Proceedings of the Engineers’ Club of Philadelphia, Vol. 9. Philadelphia: Published by the Club. Federal Communications Commission, United States. 1946. Statistics of the communications industry in the United States for the year ended December 31, 1944. Washington, D.C.: United States Government Printing Office. Federal Communications Commission, United States. 1982. Statistics of communications common carriers. Washington, D.C.: United States Government Printing Office. Fischer, Claude S. 1992. America calling: A social history of the telephone to 1940. Berkeley, CA: University of California Press. Gardner, Dee & Mark Davies. 2007. Pointing out frequent phrasal verbs: A corpus-based analysis. TESOL Quarterly 41(2). 339–359. Garrett, Martin. 2002. Mary Shelley. New York: Oxford University Press. Hills, William H. 1897. The Writer: A Monthly Magazine for Literary Workers. Volume X. Boston: The Writer Publishing Company. Hills, William H. 1902. The Writer: A Monthly Magazine for Literary Workers. Volume XV. Boston: The Writer Publishing Company. Hilpert, Martin & Stefan Th. Gries. 2009. Assessing frequency changes in multistage diachronic corpora: Applications for historical corpus linguistics and the study of language acquisition. Literary and Linguistic Computing 24(4). 385–401. Hilpert, Martin & Stefan Th. Gries. 2012. Variability-based neighbor clustering: A bottom-up approach to periodization in historical linguistics. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 100–124. Oxford: Oxford University Press. An Important Inquiry. 1882, May 28. Los Angeles Herald. 2. Kendall, M. G. 1938. A new measure of rank correlation. Biometrika 30. 81–93. Laufer, Batia & Tina Waldman. 2011. Verb-noun collocations in second language writing: A corpus analysis of learners’ English. Language Learning 61(2). 647–672. Lieberson Stanley. 2000. A matter of taste: How names, fashions, and culture change. New Haven, CT: Yale University Press. Liu, Dilin. 2011. The most frequently used English phrasal verbs in American and British English: A multi-corpus examination. TESOL Quarterly 45(4). 661–688. Martin, Pamela. 1990. The phrasal verb: Diachronic development in British and American English. New York: Columbia University Ph.D. thesis. Michel, Jean-Baptiste, Yuan Kui Shen, Aviva Presser Aiden, Adrian Veres, Matthew K. Gray, William Brockman, The Google Books Team, Joseph P. Pickett, Dale Hoiberg, Dan Clancy, Peter Norvig, Jon Orwant, Steven Pinker, Martin A. Nowak & Erez Lieberman Aiden. 2011. Quantitative analysis of culture using millions of digitized books. Science 331(6014). 176–182. Nesselhauf, Nadja. 2005. Collocations in a learner corpus. Amsterdam & Philadelphia: John Benjamins. Olds, Kelly B. 2011. The Taiwan hat industry: Pre-war roots of the post-war miracle. Business History 53(7). 1110–1129. Pelli, Mario G. 1976. Verb-particle constructions in American English: A study based on American plays from the end of the 18th century to the present. Bern: Francke. R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/ (accessed 15 February 2014).
The phrasal verb: particle distribution, register variation, and noun collocations
97
Rodríguez-Puente, Paula. 2012. The development of non-compositional meanings in phrasal verbs: A corpus-based study. English Studies 93(1). 71–90. Schudson, Michael. 1984. Advertising, the uneasy persuasion: Its dubious impact on American society. London: Routledge. Shaw, G. M. 1878. Sketch of Thomas Alva Edison. Popular Science Monthly 13(23). 487–491. Smitterberg, Erik. 2008. The progressive and phrasal verbs: Evidence of colloquialization in nineteenth-century English? In Terttu Nevalainen, Irma Taavitsainen, Päivi Pahta & Minna Korhonen (eds.), The dynamics of linguistic variation: Corpus evidence on English past and present, 269–289. Amsterdam & Philadelphia: John Benjamins. Spasov, Dimiter. 1966. English phrasal verbs. Sofia: Naouka i Izkoustvo. Stubbs, Michael. 1995. Collocations and cultural connotations of common words. Linguistics in Education 7(4). 379–390. Tenner, Edward. 1989. Talking through our hats. Harvard Magazine 91. 21–36. “telephone, n.” OED Online. September 2013. Oxford University Press. http://www.oed.com/ view/Entry/198718?rskey=GovGzr&result=1 (accessed 2 October 2013). Thim, Stefan. 2012. Phrasal verbs: The English verb-particle construction and its history. Boston & New York: De Gruyter Mouton.
Reijirou Shibasaki
On the grammaticalization of the thing is and related issues in the history of American English 1 Introduction A set of constructions of clausal origin that function parenthetically to express speaker stance are what Schmid (1999, 2000) labels shell noun constructions. These have received far less attention than they deserve, especially from a diachronic perspective (but see Jespersen 1927: 24–27; Poutsma 1929: 619–620 [both cited in Schmid 2000: 4–6]; cf. Curzan 2012: 219 for a useful overview). Shell nouns are used in the specific constructions shown in Table 1. In the table, elements in boldface stand for shell nouns, while the underlined portions are the shell contents to which the shell nouns refer (Schmid 2000: 7). The types of nouns inserted in the constructions, although they may be informative (e.g., accusation, achievement), are frequently neither specific nor informative (e.g., fact, thing). In both cases, however, they serve “the textual function of linking these nominal concepts with clauses or other pieces of text which contain the actual details of information, thereby instructing the hearer to interpret different sections of a text together” (Schmid 2000: 14, emphasis in original). In Table 1, fact and advantage provide shells for the contents in the following that-clauses that express complex and elaborate chunks of information, while accusation and achievement serve as shells to summarize the preceding content. Some shell noun constructions have been analyzed in a synchronic context (e.g., Aijmer 2007; Imo 2010), but they have not come under close scrutiny in a diachronic context. This study therefore takes a diachronic perspective, focus ing especially on the period from Late Modern English to Present-Day English, based on data from The corpus of historical American English 1810–2009 (COHA) (see Davies 2010–), The corpus of contemporary American English 1990–2012 (COCA) (see Davies 2008–) and The Santa Barbara corpus of spoken American English (SBCSAE) (see Du Bois et al. 2005–2010), and supplemented by data from other corpora as needed. I examine the construction N-be-that in Table 1 with a special focus on the thing is.1 The thing is is not only the the most frequent shell
1 This study deals with the present-tense construction the thing is but excludes the past-tense the thing was, due to space limitation.
100
Reijirou Shibasaki
Table 1: The four lexico-grammatical patterns of shell nouns (based on Schmid 2000: 22) Function
Pattern
Abbreviation
Example
Cataphoric
Shell noun (N) + postnominal clause Variants: that-clause, to infinitive-clause, wh-clause Shell NP + be + complementing clause Variants: that-clause, to infinitive-clause, wh-clause
N + cl
Mr. Bush said Iraq’s leaders had to face the fact that the rest of the world was against them.
Referring item + (modifier) + shell noun
th-N
Referring item as subject + be + shell noun (phrase)
th-be-N
Anaphoric
N-that N-to N-wh N-be-cl N-be-that N-be-to N-be-wh
The advantage is that there is a huge audience that can hear other things you may have to say. (Mr. Ash was in the clearest possible terms labelling my clients as anti-semitic.) I hope it is unnecessary to say that this accusation is also completely unjustified. (I won the freshmen’s cross-country. – Mm.) That was a great achievement, wasn’t it?
construction (as Schmid 2000: 6 found in the Bank of English), but it also shows further development from a shell construction to a pragmatic marker. A diachronic study may shed light on two questions concerning the the thing is construction. Synchronically, the thing is occurs with or without the complementizer that. A historical study should be able to explain how these two forms evolved and what their relation to each other may be. A diachronic study will also reveal whether the range of constructional variants of the thing is (that) is expand ing (e.g., in respect to prenominal modifiers) or becoming more formulaic and fixed. Here, these questions will be addressed in terms of the grammaticalization of the the thing is construction. This study is organized as follows. Section 2 explores the frequency and use of the thing is in American conversational discourse. Section 3 provides a short history of the thing is. Section 4 examines the types of prenominal modifiers for thing is and the newly emergent construction my thing is from the perspective of grammaticalization. Section 5 concludes the study.
On the grammaticalization of the thing is in the history of American English
101
2 Background 2.1 The definition of the thing is in this study The Oxford advanced learner’s dictionary (OALD) (1996) briefly defines the idiomatic expression the thing in (1) and illustrates it with the example in (2): (1) the thing (about sth/sb) is used to draw attention to an important fact or matter. (OALD, s.v. thing, idiom) (2) The thing is, can we actually afford a new car? (ibid.) According to the OALD, the thing is can be thought to serve as a device for introduc ing the speaker’s stance toward what is referred to in the following discourse. In the example given in the OALD, (2), the thing is functions as a pragmatic marker, not attached to the following clause by the complementizer that, while the following clause serves as an independent interrogative clause. A pragmatic marker is here understood as an extra-clausal element, with reduced propositional meaning, serving textual and interpersonal functions. In contrast, in the example in (3), the thing is acts syntactically as a main clause introducing the following complement clause with the complementizer that. (3) “Don’t interrupt. I know what you told me. The thing is that I’m going to do something interesting to my date tonight.” (1973 COHA: FIC, I know what you did last summer) Thus, the thing is can be broadly classified into two syntactic structures, i.e., the main clause type and the pragmatic marker type (but see Section 2.3 for further discussion). In the examples given in this chapter, shell noun constructions are boldfaced and their shell contents are underlined. The thing is form under discussion here, in addition to being a “shell construction”, is related to what has been called the th-pseudo-cleft. For example, Collins (1991: 27; cf. Quirk et al. 1985: 1389) includes among the class of pseudo-clefts those constructions with relative clauses introduced by the “in conjunction with the proform equivalent of the English interrogatives (thing, one, place, time, reason, way)”, e.g., What/the thing the car needs is a new battery (Collins 1991: 30–31) or The thing/ what I like about Joan is her sense of humour (Quirk et al. 1985: 1389n.).2 But note
2 Collins (1991: 31) excludes from the class of th-pseudo-clefts those cases where the th-word (which he restricts to thing(s), one(s), reason, way, place, time) is modified by an evaluative adjective such as only or best, since they do not allow non-cleft versions. He does accept cases where the modifier is a single numeral or quantifier (e.g., first, only).
102
Reijirou Shibasaki
that the relative clause is standardly absent in the cases considered here. I will return to the importance of this classification in Section 2.3. Biber et al. (1999: 448) note that the thing occurs with be over 20 times per million words in the conversational but not the academic part of the corpus underlying the Longman grammar of spoken and written English; they go on to state that lexical bundles such as but the thing is and the only thing is are clause fragments (Biber et al. 1999: 1005). Therefore, it is necessary to see how the thing is and its related expressions function in conversational discourse.
2.2 The frequency of the thing is in conversational discourse My conversational data come from the SBCSAE, which consists of 60 conversations, all among adult native speakers of American English who are friends or family members. Table 2 summarizes the types and tokens of the shell noun constructions in the SBCSAE. At times, the head nouns are modified by adjectives, e.g., the whole point is, while the determiner the can be replaced by others, e.g., my point is. Table 2 includes all of the variant forms of the relevant shell con struction found in the SBCSAE, accompanied by finite clauses with or without the complementizer that. The survey results seem to dovetail with Biber et al.’s (1999) observation because in the SBCSAE, the thing is is by far the most frequent among these constructional variants. Note that nouns such as truth, result, idea, difference, hope, answer, trouble, and danger are not found in this specific construction. Table 2: Types and tokens of the shell noun constructions in the SBCSAE Types (the) thing is (that) (the) point is (that) (the) problem is (that) (the) fact is (that) (the) question is (that) (the) reality is (that) (the) matter is (that) (the) reason is (that)
Tokens 24 4 2 2 2 1 1* 1
*The example is the matter of the fact is.
On the grammaticalization of the thing is in the history of American English
103
2.3 Discourse-syntactic functions of the thing is in conversational discourse The discourse-syntactic pattern in which shell noun constructions are used is predictable, as seen in the following examples. (4) MORRISON: And they said he was still alive, but he’s going to die very soon. Ms. NAVARRO: Yeah. The thing is, they lie. (2010 COCA: SPOK, NBC Dateline) As shown in (4), the thing is (that) has a strong tendency to appear in front of the shell content; what is cataphorically expressed is announced by the construction, which facilitates the speaker-hearer interaction in the immediate discourse. The position of the thing is in this syntactic sequence is discourse-interactive, serving as a prologue to the main (source of) information that immediately follows (as schematized in 5). (5) The discourse-syntactic representation of the thing is [the thing is + the speaker’s statement] structure {prologue} {main (source of) information} discourse function In addition, because the speaker pauses a moment before s/he makes her/his own statement, the thing is can be thought to express a desire to show tact or to be polite: once the addressee hears the construction, s/he can easily get prepared for the speaker’s statement to follow.3 In this respect, the function of the thing is bears a strong resemblence to the “projector” function discussed by Hopper and Thompson (2008) for pseudo-clefts in general. They argue that in conversational interaction most pseudo-clefts are not syntactically biclausal. Rather, the wh-clause “anticipates (or ‘projects’) upcoming talk by the same speaker” (105). The initial “pieces” must be “recognized as playing a key role in the strategic management of the current talk” (114). Hopper (2001) points out a number of functions of the wh-clauses: e.g., to alert the listener that the upcoming utterance is noteworthy, or to make an attitudinal comment about or to state a general theme for the upcoming utterance. Hopper and Thompson (2008) observe that the wh-clauses are “relatively open-slot ‘prefabs’” (116) or “formulaic pieces” (118). Hence, their fixed and formulaic nature, their syntactic
3 Hopper (2011) makes similar comments on the use of the such a/an construction in the S BCSAE. In the use illustrated in (i), “the sequence such a/an anticipates (projects) a second clause which usually begins with that”: (i) (H) And he was such a hit there, that Macy’s gave him his very own charge card. (Hopper 2011: 30) The meaning is resultative: the second clause follows as a result of the first clause (ibid.).
104
Reijirou Shibasaki
independence, and their pragmatic function all point to the status of the wh-clauses as pragmatic markers. The same argument can be made for the thing is. On the other hand, when the speaker wants to summarize the preceding discourse, s/he uses that’s the thing, i.e., another shell noun construction, as shown in (6). (6) Mr. PAGE: There is a runoff. That’s the thing. (2010 COCA: SPOK, The Chris Matthews show, various times, NBC) This has a clearly anaphoric function differing from the cataphoric function of the thing is. At times, the X is (that) and that’s the Y appear together in a stretch of discourse, as in (7); the speaker can change the shell nouns used before and after the shell content if s/he wants the hearer to better understand the shell content (Schmid 1999: 117). (7) Velez-MITCHELL: You know, the thing is, nobody’s ever going to feel sorry for Jennifer Aniston. That’s the problem with the whole premise. (2009 COCA: SPOK, CNN Showbiz) In other words, the thing is (that) serves an introductory or “projector” function (cataphoric), while that’s the problem serves a summary function (anaphoric), each depending on the immediate discourse context. Now look at the following examples from the SBCSAE. All examples in the SBCSAE are transcribed in prosodic segments called Intonation Units (IUs). Accord ing to Chafe (1987: 22), an IU “is a sequence of words combined under a single, coherent intonation contour, usually preceded by a pause” and is taken as the natural unit of discourse in conversation or narrative. Numbers after the SBCSAE examples stand for the file numbers; the key to SBCSAE notation is in note 4.4 (8) FRANK: .. . MELISSA: ... but uh the thing is, I don’t even know what to wear. FRANK: ... Just dress nice. (SBCSAE 019)
4 The transcription conventions of the SBCSAE are as follows. = marcato: each word distinct and emphasized; ‘…’ = medium pause (between 0.3 and 0.6 seconds); ‘..’ = short pause (about 0.2 seconds or less); ‘--’ = truncated intonation unit; ‘-’ = truncated word; (H) = inhalation; % = glottal stop; [ ] = speech overlap. ‘=’ = lengthening. See Du Bois et al. (1993) for more details.
On the grammaticalization of the thing is in the history of American English
105
(9) SHARON: and these – .. The thing is, ... (H) that the teachers that had these third-graders before, and a lot of the fourth-graders, really failed as teachers, (SBCSAE 004) (10) GAIL: Well %, well [the thing is that, PATTY: [Well that’s good to share]. GAIL: it ended up not] being, ... I mean it doesn’t really matter=, (SBCSAE 035) (11) PHIL: what we - and things aren’t that bad. The thing is that we h- do have our attitudes. (SBCSAE 010) The use of the thing is without a that complementizer preceding the complement clause, as shown in (8), is the most frequent form of the construction (see Table 3). Less common are examples such as (9) and (10), where the thing is is accompanied by a finite clause, preceded by that. In examples (8) through (10), the thing is and its finite clause appear in different IUs regardless of whether the complementizer that co-occurs with the finite clause or not; I term this the “stand-alone” type. In contrast, the occurrence of the thing is and its finite clause in the same IU, as in (11), turns out to be infrequent; I term this the “clause merger” type. The determiner the is strongly preferred as a prenominal modifier. Note that in the SBCSAE, the determiner the is always used, i.e., the + (modifier +) thing + is (+ that). Table 3 reveals one important characteristic of the thing is in conversational discourse: the thing is has a strong tendency to appear in an independent IU without the complementizer that, giving rise to a set phrase, i.e., a pragmatic Table 3: The distribution of the types of the thing is in the SBCSAE Types of the thing is
Examples
Prop. Freq. (Raw Freq.)
Stand-alone type without that-comp (the thing is Ø / IU) Stand-alone type with that-comp (the thing is that / IU) Finite clause merger type (the thing is (that) + finite clause / IU)
(8), (9)
79.2% (19 out of 24 tokens) 8.3% (2 out of 24 tokens) 12.5% (3 out of 24 tokens)
(10) (11)
prop. freq. = proportional frequency; raw freq. = raw frequency; comp = complementizer; Ø = complementizer-less; / IU = preceding elements occur in one IU.
106
Reijirou Shibasaki
marker (79.2%). If one includes the stand-alone type with the complementizer that, the independent use of the thing is (that) reaches 87.5% (21 out of 24 tokens). Two other features may be used as evidence that the thing is seems to have been conventionalized as a pragmatic marker – with respect to IUs and the lexical fixedness of the form – in American conversational discourse. First, the overall use of the complementizer that is relatively low (20.8%, or 5 out of 24 tokens); second, both type and token frequencies of prenominal modifiers other than the are limited, at least in this particular database: funny (2 tokens), only (2 tokens), other (2 tokens), main (1 token), second (1 token), and shitty (1 token).
3 A short history of the thing is 3.1 The origin and development of the thing is This particular construction has probably had a long history, although the frequency of recorded use is extremely low, sparsely distributed up until the early twentieth century. The first appearance of this construction recorded in the Oxford English Dictionary (OED) is quite recent (12). (12) I think the thing is that we want to start pushing out the route as fast as possible (…) (1971 OED: C. Bonington, Annapurna south face xiv. 175) The Middle English part (1100–1500) of The Helsinki corpus of English texts (HC) includes one related but not identical example (13); unlike the construction considered here, it does not contain a copula and is consequently paratactic. (13) But oon thyng I dar safly sey in my conceyte, that shee on her parte sithe your departier hath ben vexed (...) (1472 HC: Thomas Mull to William Stonor, The Stonor letters and papers, 1290–1483) ‘But one thing I dare safely say in my opinion (is) that she has been vexed on her part since your departure.’ The th-pseudo-cleft, which Traugott and Trousdale (2014: 139–140) see as the precursor to the wh-pseudo-cleft, likely arises in Early Modern English: (14) The thing which doth amate and most anoy my mind, Is that my hard estate, no remedy can finde. (1580 Gifford, A posie of gilloflowers [LION: EEBO]; quoted from Traugott and Trousdale 2013: 139; bolding and underlining added) ‘The thing that dismays and most toubles my mind is that my difficult condition can find no remedy.’
On the grammaticalization of the thing is in the history of American English
107
There are other early precursors of the thing is, such as as the thing is, in (15): (15) Yea to speak out plainer, as the thing is, I do not remember, that I have found this Phrase Maran-atha, in any sense at all, in any Rabbinical or Tal\mudic Writer; at any time, (1685 EEBO: Lightfoot, The works of the Reverend and Learned John Lightfoot D.D.) But the earliest example of bare the thing is without a relative clause in my corpora is (16), from the Corpus of Late Modern English texts extended version (CLMETEV), dating from the beginning of the nineteenth century. The CLMETEV includes two other examples from the same text, The parent’s assistant: the thing is, I shall never have (…) and but then the thing is, one can never get to hear (…).5 (16) “but then, the thing is, I must go without the uniform, if I have the great-coat.” (1796–1801 CLMETEV: Edgeworth, The parent’s assistant) Interestingly, the earliest three examples of the thing is in The parent’s assistant are all unattached to the following clause and do not contain the complementizer that. The stand-alone usage without a that-complementizer might possibly have been special to Maria Edgeworth (1767–1849). Note that the Early Modern English part (1500–1710) of the HC, the Corpus of early English correspondence sampler (1418–1680) (CEECS), and A corpus of English dialogues 1560–1760 (CED),6 turn out to have no relevant examples, which may confirm actuation of the construction in the late eighteenth century. Uses of stand-alone the thing is, … in corpora in the nineteenth century are infrequent, as well. (17) and (18) are the next two examples attested in my corpora, dating from the beginning and end of the nineteenth century. Given the use of a comma following the thing is and the presence of that, these examples are indeterminate between main clause and pragmatic marker uses. (17) Ignorance alone makes monsters or bugbears: our actual acquaintances are all very commonplace people. The thing is, that as a matter of hearsay or conjecture, we make abstractions of particular vices, and irritate ourselves against some particular quality or action of the person we dislike. (1821–1822 CLMETEV: W. Hazlitt, Table talk)
5 The actual date of the thing is in (16) is 1800. The stories in The parent’s assistant in which the thing is occurs, “Waste not, want not” and “The mimic”, were not included in the work until the six-volume edition of 1800. Because six of the stories had been published in the two earlier editions of 1796 and the compilers of the CLMETEV were probably using the 1801 edition, they give 1796–1801 as the date for The parent’s assistant. (Michael Adams, p.c.) 6 For this paper, Early English books online (EEBO) was not consistently searched.
108
Reijirou Shibasaki
(18) “But the odd thing is, that he always strikes one as practical-minded. Don’t you feel that, Mrs Reardon?” (1891 CLMETEV: G. Gissing, New Grub Street) Despite their rarity, these examples attest to a suggested origin at the turn of the nineteenth century. The corpus of historical American English enables us to delve more deeply into the recent development of the thing is. Example (19) is the third earliest occurrence of the pragmatic marker usage, i.e., the thing is, …, in COHA; the earlier two examples, from the 1830s and the 1860s, respectively, are accompanied by to-infinitive clauses instead of finite clauses and thus are omitted. (20) is the earliest main clause usage in COHA, i.e., the thing is that … (19) “The thing is, Mrs. Scherman, if I can try this anywhere, I can try it here.” (1873 COHA: FIC, The other girls) (20) “The thing is that I’m not game enough to go on and take the punishment. Are you surprised?” (1920 COHA: FIC, The gorgeous girl) The results of the COHA survey are compatible with what was discussed above: namely, that the stand-alone usage (the pragmatic marker usage) is attested earlier than the main clause usage. However, numerical frequencies of the different forms of the the thing is construction allow a more nuanced view. Table 4 summarizes the forms of the the thing is construction found in COHA (last access: March 31, 2014). Numbers indicate raw frequencies; “--” stands for zero. Note that types of adjectival modifiers (ADJ in the table) are analyzed in detail in Section 4.1. Table 4 shows that the main clause type is more frequent than the pragmatic marker type in the earlier periods, although the pragmatic type predominates from the 1970s onwards. This suggests that the earlier appearance of the pragmatic marker type in our data may be an accident of the corpora. The determiner-(adjective) modified construction is attested in both main clause and pragmatic marker types; adjectival modification is much more likely to occur in the main clause type than in the pragmatic marker type. Importantly, the determiner-less construction (Ø (ADJ) thing is) is found only in the pragmatic marker type. This will be discussed in Section 4.2 as a sign of the decategorialization and grammaticalization of the construction. The survey results can be summarized as follows. The bare structure, the thing is (that), seems to have become entrenched around the turn of the nineteenth century, according to the evidence of the OED, CLMETEV and COHA. In COHA, the main clause use (the thing is that) is predominant until the last quarter of the twentieth century. Although the evidence of the corpora used here is too sparse to give a definitive account of the rise of this structure, it is likely that a number of structures may have contributed to its development: the paratactic structure
On the grammaticalization of the thing is in the history of American English
109
Table 4: Occurrences of pragmatic marker the thing is, … and main clause the thing is that in COHA Pragmatic marker
1810s 1820s 1830s 1840s 1850s 1860s 1870s 1880s 1890s 1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s 1990s 2000s Total
the thing is,
the ADJ thing is,
--1 --1 1 -----1 3 5 7 19 7 18 41 104
1 1 --1 1 2 1 3 2 3 6 2 4 10 10 10 15 9 8 89
Main clause
Ø thing Ø ADJ the thing the ADJ is, thing is, is that thing is that ---------1 ----1 1 4 1 4 14 26
-------------2 3 1 2 2 1 10 21
-----------1 -1 3 1 5 --1 12
----1 2 1 7 12 16 18 23 26 17 21 26 30 20 14 16 250
Ø thing is that
Ø ADJ thing is that
--------------------0
--------------------0
shown in (13) (i.e., One thing I can safely say: that …), the th-pseudo-cleft structure shown in (14) (i.e., The thing that I can say is that …), the parenthetical adverbial clause shown in (15) (i.e., as the thing is …), and the main clause use, which predominated in the early period (i.e., the thing is that …). One thing that seems clear from the data is that the presence or absence of that has little effect on the use of the thing is as either a main clause or pragmatic marker (contra Thompson and Mulac 1991). If the paratactic structure represents the origin of the main clause usage, we have a case of clause combining, with evolution from a structure where the that-clause is relatively independent, i.e., neither dependent nor embedded (paratactic), to a structure where the that-clause is interdependent with the initial clause, i.e., dependent but not embedded (hypotactic). This can be seen as a case of grammaticalization (see Hopper and Traugott 2003: 178–179). It is unlikely, however, that this structure ever evolved to a fully subordinate structure, as
110
Reijirou Shibasaki
the that-clause continues to express the main information of the sentence, and the thing is serves as a prologue or “projector” introducing this information, as discussed above in Section 2.3. The semi-independence of the projector element allows its subsequent grammaticalization as a pragmatic marker (see Section 4).
3.2 A note on punctuation In the SBCSAE data, commas delimit IUs, indicating where intonation breaks are in the online deployment of language. They may not correspond to syntactic units, but rather reflect situations in which speakers introduce IU breaks at points of uncertainty, in order to consider the wording of their following utterance. Note the placement of commas in the following examples, the thing is, that … and the thing is that, …., which point to intonation breaks between a verb and its object or following a complementizer (example 9 is repeated as 21): (21) SHARON: and these – .. The thing is, ... (H) that the teachers that had these third-graders before, and a lot of the fourth-graders, really failed as teachers, (SBCSAE 004) (22) DAN: ... (H) U=m, ... the second thing is that, .. people who do not live, .. directly inside, (SBCSAE 026) Punctuation similar to that in (21) and (22) is sporadically attested in COCA as well. (23) LEE: (…) I placed my camera inside the bag and made a hole on the side to secretly film. But the thing is, that the light was being reflected on the camera lens. So I have to be very, very careful. (2006 COCA: SPOK, CNN Live Sat) (24) “The thing is that, within his ideology of African unity, he has created some African units in his Libyan army.” (2011 COCA: NEWS, CS Monitor) Whether the position of the comma in transcribed speech can conclusively reflect spoken discourse cannot, of course, be exactly determined in COHA and COCA. One cannot say with absolute conviction that a comma is placed where a speaker pauses, especially in the remote past, from which no audio data are available. However, given that punctuation patterns in the SBCSAE are congruent with those in COCA, it is likely that the punctuation system used in the COCA examples
On the grammaticalization of the thing is in the history of American English
111
above does not diverge vastly from the intonation system. For instance, in the example in (25), because the following clause is a non-embeddable interrogative, the thing is can no longer be analyzed as serving as a main clause; rather, it is a parenthetical that characterizes what comes next in the discourse: (25) Mr. PEROT: Well, but see, you’re picking at pieces. The whole- the thing is, does this ever balance the budget on a yearly basis, or do we still have deficits? It never brings it down to zero. (1993 COCA: SPOK, CNN Politics) As Tuggy (1996: 725) suggests, the atypical intonation pattern with a stressed copula would be favored, presumably, with a short pause between the thing is and the following interrogative clause, as indicated by a comma (cf. Schmid 2000: 338; Aijmer 2007: 33). Horobin (2012) states that electronic corpora and dictionaries, all of which are based on edited manuscripts, may or may not be faithful to their originals. Although much punctuation has been added by editors to texts, especially prior to the eighteenth century (Parkes 1992), the thing is (that) may be relatively free of such punctuation issues because it can be thought to be in an early stage of con structional development quite late in Late Modern English, as discussed in Section 3.1. It may be safe to say that methodological issues regarding punctuation can be set aside and that the data collected from the above-mentioned corpora can provide evidence for the constructional development and the discourse functions of the thing is (that).
4 Ongoing and emerging changes to the thing is In Present-Day English, the thing is can be understood as undergoing grammaticalization, as it conforms to several of Lehmann’s parameters ([1982] 1995). Section 4.1 argues for the grammaticalization of the thing is on the basis of fixedness, coalescence, and decategorialization, while Section 4.2 investigates the emergence of speaker stance, i.e., subjectification, as further evidence of grammaticalization. Section 4.3 notes a phenonemenon involving the thing is which is suggestive of advanced grammaticalization. Section 4.4 sums up the discussion of grammaticalization.
4.1 Types of modifiers in the pragmatic marker usage of the thing is In order to determine whether certain adjectival modifiers might display a marked preference for occurring in the stand-alone usage of the thing is, the types
112
Reijirou Shibasaki
of modifiers appearing in COHA (last access: Oct. 9, 2013) were tabulated (see Table 5). In all cases, the percentage of forms with no determiner (Ø forms) has been given. Because a substantial number of the prenominal modifiers occur just once in the past 200 years, only those modifiers that are used more than ten times are listed in frequency order. Table 6 summarizes the same modifiers in COCA for comparison (last access: Oct. 9, 2013).
Table 5: Modifiers for the pragmatic marker usage of (the) thing is in COHA: More than 10 tokens the/ Ø
the only/ Ø only
the funny/ Ø funny
1810s 1820s 1830s 1840s 1850s 1860s 1870s 1880s
--2/ -- (0%) -1 2/ -- (0%) 4/ -- (0%) --
------2/ -- (0%) 1/ -- (0%)
---------
1/-- (0%) ----1/-- (0%) ---
-----1/ -- (0%) ---
---------
1890s 1900s
3/ -- (0%) --
---
1/ -- (0%) 3/ -- (0%)
3/ -- (0%) --
---
1910s 1920s 1930s
-2/ 1 (33.3%) --7/ 1 (12.5%)
1/ -- (0%) 3/ -- (0%) 4
-2/ 0 (0%) --
----
-2/ -- (0%) --
1/-- (0%) ---
1940s
9/ -- (0%)
1/ 1 (50%)
1/ -- (0%)
2/ -- (0%)
3/ -- (0%)
1950s
1/ -- (0%)
2/ -- (0%)
4/ --(0%)
1/ -- (0%)
--
1970s 1980s
63/ 5 (7.4%) 27/ 2 (6.9%)
1/ -- (0%) 2/ -- (0%)
---
3/ -- (0%) 2/ -- (0%)
1990s
45/ 15 (25%) 80/ 24 (23.1%) 273/ 51 (15.7%)
1/ 2 (66.7%) 7/ 1 (12.5%) 4/ 1 (20%) 6/ 1 (14.3%) 5/ 1 (16.7%) 5/ 10 (66.7%) 31/ 17 (35.4%)
--
1960s
14/ 2 (12.5%) 17/ 1 (5.6%)
4 /2 (33.3%) 8/5 (38.5%) 2/ -- (0%)
2/ -- (0%)
1/ 1 (50%) 3/ 1 (25%) 14/ 2 (12.5%)
--
2000s Total
1/ 1 (50%) 5/ 2 (28.6%) 2/ 2 (50%) 5/ 3 (37.5%) 41/ 15 (26.8%)
the main/ the strange/ Ø main Ø strange
0/ 2 (100%) 16/ 2 (11.1%)
the important/ Ø important
-11/ 0 (0%)
On the grammaticalization of the thing is in the history of American English
113
Table 6: Modifiers for the pragmatic marker usage of (the) thing is in COCA
1990–1994 1995–1999 2000–2004 2005–2009 2010–2012
the/ Ø
the only/ Ø only
the funny/ Ø funny
the main/ Ø main
the strange/ Ø strange
the important/ Ø important
176/ 11 (5.9%) 190/ 26 (12.0%) 216/ 28 (11.5%) 238 / 35 (12.8%) 132/ 29 (18%)
11/ 3 (21.4%) 17/ 3 (15%) 6/ 2 (25%) 6/ 4 (40%) 5/ 3 (37.5%)
18/ 4 (18.2%) 40/ 11 (21.6%) 28/ 7 (20%) 29/ 9 (23.7%) 13/ 5 (27.8%)
12/ -(0%) 8/ 1 (11.1%) 6/ 1 (14.3%) 7/ -(0%) 1/ -(0%)
2/ -(0%) 3/ -(0%) 8/ -(0%) 2/ 1 (33.3%) 0/ 1 (100%)
10/ -(0%) 8/ -(0%) 14/ -(0%) 13/ -(0%) 7/ -(0%)
In terms of token frequency, the determiner the is the most preferred modifier for thing is; the next most common modifiers in both COHA and COCA – only, funny and main – are found also in the conversational discourse of the SBCSAE (see Section 2.3 for types of prenominal modifiers in SBCSAE). Two observations can be made concerning the occurrence or non-occurrence of the determiner the. First, over time, we see an increase in the frequency of Ø forms, whether or not an adjective modifier is present. (Recall, see Table 4, that the Ø form is not found in the clause type with that.) Second, the higher the frequency of a given adjective + thing is construction is, the greater the likelihood of reduction (loss of the); i.e., the presence or absence of the is subject, to some degree, to frequency effects (e.g., Krug 1998; Bybee 2007: 294–312). Infrequent uses of a particular adjective in the the + adjective + thing is construction are not likely to undergo such a syntactic reduction. For example, real and difficult are rare modifiers in this construction. (One example of the real thing is, … is found in COHA and four in COCA; one example of the difficult thing is, … is found in COHA and four in COCA, two of which are the most difficult thing is … [Aug. 14, 2013].). Importantly, these adjectives always appear with the determiner the, unlike the much more frequent examples with only, funny, main, or strange, all of which have already undergone the loss of the, presumably due to the effects of frequency: (26) Ms. IVEY: (…) Maybe the real thing is, you know, lower income families that are in neighborhoods that don’t have access to high-quality food, they’re the ones more likely to have the higher body mass index. (2011 COCA: SPOK, Tell me more 11:00 PM) (27) Mr. REID: The difficult thing is, in America, nothing defines you. I can’t think of anything that defines you quite like being a professional athlete. (1996 COCA: SPOK, CBS Sun Morn)
114
Reijirou Shibasaki
In addition, COCA includes the following two examples, both of which imply, albeit in written discourse, the gradual process of reducing the determiner the: (28) “Well, but th’ thing is,” said Ernie, lowering his voice, “Junior’s not the sharpest knife in the drawer.” (1999 COCA: FIC, A new song) (29) “Everythin’ just so’s I could get a little piece a’mind. An’ th’ thing is, everybody, every man in th’ South’s, done th’ same as me.” (1996 COCA: FIC, Tell me a tale) In casual speech, function words are often phonetically reduced, and the determiner the may have started undergoing such phonetic shrinkage as in (28) and (29), which gives an account of the ongoing and emerging processes of reduction in Tables 5 and 6. The changes discussed in this section suggest that the thing is is grammaticalizing as a type of construction. A gradual rise in the use of the determinerless Ø thing is in the recent past reflects the gradual loss of the behavioral characteristics of thing as a noun, i.e., its “decategorialization” (Brinton and Traugott 2005: 110). There are also morphosyntactic reduction (of the) and its coalescence with the following noun, further properties of grammaticalization. These changes are likely to be the result of innumerable communicative acts in interaction. When taken together, the thing is and its sub-constructions appear to have developed over the past century into a more-or-less set phrase, subsequently rendering the determiner-less variants more common in the fairly recent past.
4.2 Syntactic reflection of (inter)subjectivity in the recent past Pragmatically, the thing is implies the speaker’s stance toward or responsibil ity for the statement to follow. The speaker’s commitment to the following anchor clause comes to be syntactically reflected in the first person singular possessive my instead of the determiner the or zero. The first person singular possessive construction my thing is emerges in the 1990s; the earliest exam ples for both main clause and pragmatic marker types are taken from COCA. (30) Mr. SEINFELD: Well, my thing is that I don’t think the seeds- (1993 COCA: SPOK, CNN King) (31) SHERYL SWOOPES: You know, my thing is, with this being my first Olympics, I really didn’t know what to expect coming into here. (1996 COCA: SPOK, ABC GMA)
On the grammaticalization of the thing is in the history of American English
115
COCA contains 9 examples of my thing is that … and 19 examples of my thing is, … (last access: Apr. 3, 2014).7 What is being expressed in these examples is the speaker’s epistemic attitude, which is explicitly marked by reference to the speaker, my. Despite the extremely small number of examples of my thing is, it indicates the rise of an epistemic preliminary remark privileging the status of the speaker and provides evidence of subjectification, a process seen as central to grammaticalization. For example, Traugott (1995: 32) defines subjectification as “the development of a grammatically identifiable expression of speaker belief or speaker attitude to what is said” (see also Traugott 2010: 61). Such subjectivity had not been syntactically marked until quite recently. The syntactic or constructional derivation of my thing is from the thing is seems to have been gradual, viewed from the theory of grammaticalization (see Trousdale 2013 and Traugott and Trousdale 2014 for detailed discussions; cf. Company 2006). In fact, the example to follow provides evidence for the process in which the prenominal modifiers the and my are interchangeable in the context of speech. (32) MURPHY: Give me – give me an example of garden variety sexual harassment that you think a guy doesn’t even think twice about? Ms. PAGNOZZI: No because – my – the thing is that I’ve never sued anybody. I like some of my sexual harassers, OK? (1997 COCA: SPOK, NBC Dateline) Here, Ms. Pagnozzi tries to answer Murphy’s question with reservation, first using my (possibly my thing is) and then rephrasing my as the thing is that. This solitary example was recorded in the late 1990s, when my thing is (that) began to appear, albeit sporadically. As a consequence, the rise of my thing is can be regarded as a syntactic reflection of subjectivity in the ongoing process of constructional change. There is also evidence of intersubjectivity in the development of the thing is, indicated by the occurrence of interpersonal hedges either before or after the construction, as in (33) and (34); hedges in focus are in small capitals: (33) “Uh, the thing is, we just remodeled the top floor and wiring it into the old system has been a problem, you know?” (2007 COHA: FIC, Looks to die for) (34) Mr. TREES: … i think the thing is, if he’s uncomfortable with her and that’s why he’s uncomfortable, then that’s a serious issue. (2011 COCA: SPOK, NBC Today) Taking into account the fact that the thing is can be uttered in a strident tone of voice at times like mind you, hedges may serve to soften the speaker’s whole
7 Note that the three examples in COHA are also attested in COCA.
116
Reijirou Shibasaki
utterance, with tact and consideration, depending on the person to whom one is speaking. In this sense, the thing is in its entirety modalizes the shell content clause to follow.
4.3 Signs of further grammaticalization One sign of the advanced grammaticalization of the thing is may be the so-called “double IS” or “reduplicative copula” construction as in (35): (35) See, the thing is, is that guys like me, when I see a buoy, I try to steer around it an [sic] not sit on it (1993 COCA: SPOK, NPR) This construction occurs frequently with the thing is (also the problem is, the fact is, the point is, the issue is, the reality is; see Curzan 2012: 220). The reduplicative copula has been discussed at length in the literature.8 Suffice it to say here that the repetition of is may be a sign that initial the thing is is now processed holistically as a fully grammaticalized unit, thus necessitating the appearance of the second is.9
4.4 Grammaticalization Like many other constructions, the thing is can be considered to be an ongoing or rather emerging phenomenon compared to other well-known cases of grammaticalization (e.g., the modal auxiliaries, be going to). While it shows evidence of grammaticalization in terms of decategorialization, coalescence, and (inter)subjectivity, there are ways in which its development is not typical of grammaticalization. First, the pragmatic marker the thing is, while syntactically independent of the main clause and textual and interpersonal in function, is not entirely fixed in form. It occurs with a (limited) set of adjectival modifiers, with or without the. However, a lack of complete fixedness has been shown to be char acteristic of other pragmatic markers (Van Bogaert 2011: 308–315; see Shibasaki 2014 for the constructional variations and grammaticalization of the point is (that) in a similar fashion). Second, the insertion of my might be seen as an expansion of the construction, a kind of “host-class expansion” in the framework
8 See Bolinger (1987), Tuggy (1996), Massam (1999), Shapiro and Haley (2002), Curzan (2012 and references therein). 9 Holistic processing might also be indicative of lexicalization, though the pragmatic status of the thing is dictates against this interpretation.
On the grammaticalization of the thing is in the history of American English
117
of Himmelmann (2004), and thus incompatible with grammaticalization. In general, therefore, one cannot be quite sure whether the present variations are converging into a couple of more fixed chunks such as the thing is and Ø thing is or whether those variations will continue to render further variations.
5 Concluding remarks This study began with the question stated in Section 1: how did the thing is (that), one of the shell noun constructions labelled by Schmid (2000), develop over time? A diachronic survey of the construction was undertaken, focusing on the history of American English. Building on data collected from a variety of corpora, this study arrived at the following conclusions. Firstly, the two main forms of the thing is – i.e., as a main clause followed by a that complement and as a sentence-initial, extra-syntactic pragmatic marker – emerged at roughly the same time in the late eighteenth century, but were infrequent up until the late nineteenth century; after the turn of the twentieth century, each type began to grow in frequency, with the pragmatic marker type becoming frequent especially in the last third of the century. Secondly, although the data of this study are not conclusive, it seems that a number of related constructions, including th-pseudo-clefts and paratactic structures, led to the rise of the thing is. Thirdly, among the array of prenominal modifiers that are found to occur in this construction, thing has a strong preference for the determiner the over the others. Moreover, adjectival modification is much more likely to occur in the main clause than in the pragmatic marker type. The SBCSAE, for example, which shows a high rate of use of the pragmatic marker type (occurring in a separate intonation unit), has very few types and tokens of adjectival modifiers in this construction. Fourthly, the frequent use of the thing is started to cause a reduction of form, resulting in a determiner-less variant, i.e., Ø thing is, in the pragmatic type only in the fairly recent past. Fifthly, the rise of the first person possessive construction my thing is indicates an increase of subjectivity, i.e., subjectification. These results point to the grammaticalization of the thing is as a pragmatic marker but also underline the fact that this is not a simple linear process: the grammatical ization process is not a solid freezing of one particular construction but rather a constellation of constructions that vary in productivity. Furthermore, all of these changes have been motivated and facilitated in the highly specific discoursesyntactic sequence presented in (5) (Traugott 2003): the constructional change, expansion, and variation are discourse-based. All in all, the history of the thing is provides us with an example of the variability of language that demands attention and creates controversy in the framework of grammaticalization.
118
Reijirou Shibasaki
Taking these survey results into consideration, a diachronic study of the thing is (that) in American English has proved to be a good case study, and the results of the survey of the data collected from a variety of corpora enabled a description with solid evidence, which contributes fruitfully to the volume’s theme, i.e., evidence and method in histories of English. Of course, these survey results do not as yet go beyond the realm of a case study, and the directions of change with respect to the whole paradigm of shell noun constructions will have to be further examined.
Acknowledgments The author thanks Michael Adams, Laurel Brinton, Hurbert Cuyckens, Robert Fulk, Joanna Nykiel, and Lynn Sims for their invaluable comments at SHEL-7, Indiana University, Bloomington, April 27, 2012. My debt of gratitude extends to Joan Beal for providing relevant information to the author and to Elizabeth C. Traugott and Heather Oumounabidji for their input. Note that this study is part of the author’s Grant-in-Aid for Young Scientists (B) research project (No. 22720194) supported by the Japan Society for the Promotion of Science and part of a Grantin-Aid for Scientific Research (C) project (No. 25370569) supported by The Ministry of Education, Culture, Sports and Technology, Japan. Portions of this article have been published previously in Studies in Modern English: The thirtieth anniversary publication of the Modern English Association, edited by Ken Nakagawa (Tokyo: Eihōsha, 2013), and the author is grateful to Ken Nakagawa (Editor-in-Chief) and Hiroshi Yonekura (President) of the Modern English Association for permission to republish them here.
Sources Corpus of early English correspondence sampler (CEECS). Compiled by Terrtu Nevalainen (leader), Jukka Keränen, Minna Nevala (née Aunio), Arja Nurmi, Minna Palander-Collin & Helena Raumolin-Brunberg, ca. .45 million words. http://www.helsinki.fi/varieng/ domains/CEEC.html A corpus of English dialogues 1560–1760 (CED). 2006. Compiled under the supervision of Merja Kytö (Uppsala University) & Jonathan Culpeper (Lancaster University), ca. 1.2 million words. http://www.engelska.uu.se/Research/English_Language/Research_Areas/ Electronic_Resource_Projects/A_Corpus_of_English_Dialogues/ A corpus of Late Modern English texts, extended version (CLMETEV). 2006. Compiled by Hendrik De Smet, ca. 15 million words. https://perswww.kuleuven.be/~u0044428/ clmetev.htm
On the grammaticalization of the thing is in the history of American English
119
Davies, Mark. 2008–. The corpus of contemporary American English: 450 million words, 1990–present (COCA). Available online at http://corpus.byu.edu/coca/. Davies, Mark. 2010 –. The corpus of historical American English: 400 million words, 1810–2009 (COHA). Available online at http://corpus.byu.edu/coha/. Du Bois, John W., Wallace L. Chafe, Charles Meyer, Sandra A. Thompson, Robert Englebretson & Nii Martey. 2000–2005. Santa Barbara corpus of spoken American English, Parts 1–4 (SBCSAE). Philadelphia: Linguistic Data Consortium, ca. 249,000 words. Early English books online (EEBO). http://www.proquest.com/products-services/eebo.html The Helsinki corpus of English texts. 1991. Department of Modern Languages, University of Helsinki. Compiled by Matti Rissanen (Project leader), Merja Kytö (Project secretary) et al. Middle English part 1150–1500/Early Modern English part 1500–1710, ca. 500,000 words for each part. http://www.helsinki.fi/varieng/CoRD/corpora/HelsinkiCorpus/ Simpson, J. A. & E. S. C. Weiner. 1989. Oxford English Dictionary (OED). 2nd edn. on CD-ROM. Version 4.0, 2.5 million words. Oxford: Oxford University Press.
References Aijmer, Karin. 2007. The interface between discourse and grammar: The fact is that. In Agnès Celle & Ruth Huart (eds.), Connectives as discourse landmarks, 31–46. Amsterdam & Philadelphia: John Benjamins. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. Longman grammar of spoken and written English. London: Longman. Bolinger, Dwight L. 1987. The remarkable double is. English Today 9. 39–40. Brinton, Laurel J. & Elizabeth Closs Traugott. 2005. Lexicalization and language change. Cambridge: Cambridge University Press. Bybee, Joan. 2007. Frequency of use and the organization of language. Oxford: Oxford University Press. Chafe, Wallace. 1987. Cognitive constraints on information flow. In Russell Tomlin (ed.), Coherence and grounding in discourse, 21–55. Amsterdam & Philadelphia: John Benjamins. Collins, Peter C. 1991. Cleft and pseudo-cleft constructions in English. London & New York: Routledge. Company, Company C. 2006. Zero in syntax, ten in pragmatics. In Angeliki Athanasiadou, Costas Canakis & Bert Cornille (eds.), Subjectification, 375–397. Berlin & New York: Mouton de Gruyter. Curzan, Anne. 2012. Revisiting the reduplicative copula with corpus-based evidence. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 211–221. Oxford: Oxford University Press. Du Bois, John W., Stephen Schuetze-Coburn, Susanna Cumming & Danae Paolino. 1993. Outline of discourse transcription. In Jane A. Edwards & Martin D. Lampert (eds.), Talking data: Transcription and coding in discourse research, 45–89. Hillsdale, NJ: Lawrence Erlbaum Associates. Himmelmann, Nikolaus P. 2004. Lexicalization and grammaticization: Opposite or orthogonal? In Walter Bisang, Nikolaus P. Himmelmann & Björn Wiemer (eds.), What makes grammaticalization? A look from its fringes and its components, 21–42. Berlin & New York: Mouton de Gruyter.
120
Reijirou Shibasaki
Hopper, Paul J. 2001. Grammatical constructions and their discourse origins: Prototype or family resemblance? In Martin Pütz, Susanne Niemeier and René Dirven (eds.), Applied cognitive linguistics I: Theory and language acquisition, 109–129. Berlin & New York: Mouton de Gruyter. Hopper, Paul J. 2011. Emergent grammar and temporality in interactional linguistics. In Peter Auer & Stefan Pfänder (eds.), Constructions: Emerging and emergent, 22–44. Berlin & New York: De Gruyter Mouton. Hopper, Paul J. & Sandra A. Thompson. 2008. Projectability and clause combining in interaction. In Ritva Laury (ed.), Cross-linguistic studies of clause combining: The multifunctionality of conjunctions, 99–123. Amsterdam & Philadephia: John Benjamins Hopper, Paul J. & Elizabeth Closs Traugott. 2003. Grammaticalization. 2nd edn. Cambridge: Cambridge University Press. Horobin, Simon. 2012. Editing Early English texts. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 53–62. Oxford: Oxford University Press. Imo, Wolfgang. 2010. Mein Problem ist/mein Thema ist (‘My problem is/my topic is’): How syntactic patterns and genres interact. In Heidrun Dorgeloh & Anja Wanner (eds.), Syntactic variation and genre, 141–166. Berlin & New York: De Gruyter Mouton. Jespersen, Otto. 1927. A Modern English grammar on historical principles, Part III: Syntax. Second volume. Heidelberg: Carl Winters Universitätsbuchhandlung. Krug, Manfred. 1998. String frequency: A cognitive motivating factor in coalescence, language processing and linguistic change. Journal of English Linguistics 26. 286–320. Lehmann, Christian. (1982) 1995. Thoughts on grammaticalization. Munich: LINCOM. Massam, Diane. 1999. Thing is constructions. English Language and Linguistics 3(1). 335–352. Oxford Advanced Learner’s Dictionary. 1996. 5th edn. Oxford: Oxford University Press. Parkes, Malcolm B. 1992. Pause and effect: An introduction to the history of punctuation in the west. Berkeley, CA: University of California Press. Poutsma, Hendrik. 1929. A grammar of Late Modern English, Part I: The sentence. Second half: The composite sentence. 2nd edn. Groningen: P. Noordhoff. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A comprehensive grammar of the English language. London: Longman. Schmid, Hans-Jörg. 1999. Cognitive effects of shell nouns. In Karen van Hoek, Andrej A. Kibrik & Leo Noordman (eds.), Discourse studies in cognitive linguistics, 111–132. Amsterdam & Philadelphia: John Benjamins. Schmid, Hans-Jörg. 2000. English abstract nouns as conceptual shells: From corpus to cognition. Berlin & New York: Mouton de Gruyter. Shapiro, Michael & Michael C. Haley. 2002. The reduplicative copula IS IS. American Speech 77(3). 305–312. Shibasaki, Reijirou. 2014. On the development of the point is and related issues in the history of American English. English Linguistics: Journal of the English Linguistic Society of Japan 31(1). 79–113. Thompson, Sandra A. & Antony Mulac. 1991. A quantitative perspective on the grammaticalization of epistemic parentheticals in English. In Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, Vol. 2, 313–330. Amsterdam & Philadelphia: John Benjamins. Traugott, Elizabeth Closs. 1995. Subjectification in grammaticalization. In Dieter Stein & Susan Wright (eds.), Subjectivity and subjectivisation, 31–54. Cambridge: Cambridge University Press.
On the grammaticalization of the thing is in the history of American English
121
Traugott, Elizabeth Closs. 2003. Constructions in grammaticalization. In Brian Joseph & Richard Janda (eds.), The handbook of historical linguistics, 624–647. Oxford: Blackwell. Traugott, Elizabeth Closs. 2010. (Inter)subjectivity and (inter)subjectification. In Kristin Davidse, Lieven Vandelanotte & Hurbert Cuyckens (eds.), Subjectification, intersubjectification and grammaticalization, 29–71. Berlin: De Gruyter Mouton. Traugott, Elizabeth Closs & Graeme Trousdale. 2014. Constructionalization and constructional change. Oxford: Oxford University Press. Trousdale, Graeme. 2013. Gradualness in language change: A constructional approach. In Anna Giacalone Ramat & Piera Molinelli (eds.), Synchrony and diachrony: A dynamic interface, 27–42. Amsterdam & Philadelphia: John Benjamins. Tuggy, David. 1996. The thing is is that people talk that way: The question is is why? In Eugene H. Casad (ed.), Cognitive linguistics in the redwoods: The exploration of a new paradigm in linguistics, 713–752. Berlin & New York: Mouton de Gruyter. Van Bogaert, Julie. 2011. I think and other complement-taking mental predicates. Linguistics 49(2). 295–332.
Betty S. Phillips
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder In this chapter data are gathered from the The dictionary of Old English corpus in electronic form (Healey et al. 2009; henceforth DOE corpus) to act as evidence in the determination of the synchronic status in early Old English of two diphthongs which by later Old English had clearly merged into one. The use of this corpus allows for a methodology that incorporates word frequency into the evidence, a conditioning factor that was missing from earlier considerations and which has been found to be highly relevant to other sound changes (Phillips 2006). Specifically, two views have been expressed about the relationship between the long diphthongs spelled and in early West Saxon. Both /io/ and /eo/ existed in West Germanic, but by the time of most of our West Saxon manuscripts, ca. 1000 a.d., they had merged into /eo/, interpreted by Minkova and Stockwell (2008) as [ew] and by Lass (1994) as [eu]. As for early West Saxon, there is no consensus, however. Hogg (1992: 86), in a passage referring to both long and short diphthongs, states that “Occasionally, eo spellings were substituted for by io- spellings, but such cases are no more than relics of a previous stage without synchronic relevance”. Smith (2007: 163), on the other hand, says of these diphthongs: “This change was still happening in historic times, and the earliest forms of West Saxon often retain -spellings”. But he offers no evidence. This study, therefore, has as its purpose to determine which of these views is most likely to reflect the phonological system of Early West Saxon as evidenced in the spellings of the Hatton 20 manuscript of Alfred’s translation of Gregory’s Pastoral care.
1 Random vs. ordered variation Horgan (1982: 228) probably provides the fullest description of variation between io and eo spellings in Hatton 20, noting that there is not a simple one-way progression from to , for the spelling is used “both for historical io and historical eo” so frequently, that we must regard it as a major, recessive criterion of early West Saxon. In these manuscripts io spellings tend to equal or outnumber eo spellings, of whatever origin. In H[atton 20] first hand, īo appears from historical ēo 18 times, against ēo 6 times: īo from
124
Betty S. Phillips
Gmc. iu 8 times. (…) In the second hand, īo again appears for historical ēo, and as the devel opment of Gmc. iu, much more often than ēo, e.g., seo only 4 times; hio most, ten times as often as heo; also dioful- 24 times, 4 times deoful.
This second hand, according to Ker (1957: 386), accounts for 682 lines of Sweet’s edition, out of approximately 5592 total lines, or approximately 12%. Horgan (1982: 228) adds that a third hand also frequently has io, but has an increased number of eo spellings; this divergent third hand, however, is that of Alfred’s preface, which is not included in the present investigation. Given Horgan’s description of the alternations in the manuscript, and given that the manuscript must reflect a late stage of the ultimate lowering of [iw] to merge with [ew], the spellings in long and here are analyzed as they occur in the synchronic system exhibited by Hatton 20. If the spellings are random, then the merger has probably been completed but the orthographic symbol for the resulting sound has not been settled upon. If the spellings are not random, and especially if they pattern similarly to other, known sound changes, then the merger is probably still in progress. It is clear that eventually these sounds were merged in West Saxon, even if, as Smith (1996: 19) observes, “the relationship between Early West Saxon and Late West Saxon is distorted if the material remains which make up the latter are taken as descended straight forwardly from the former”. Investigations into modern-day mergers have found that as vocalic sounds approach merger, a number of things are possible. Speakers, for instance, may hypercorrect if one pronunciation is the perceived norm (Pitts 1986: 136; DiPaolo 1992; Hall-Lew 2011), they may distinguish the two sounds in production but not in perception (Labov, Karen, and Miller 1991; Pierrehumbert 2003: 211), or they may even flip-flop the pronunciations, as Faber (1992) found for pool and pull in Utah. Speakers also may take a lexically diffused change and change it from lexically conditioned to phonetically conditioned, as Labov (2007: 362 n. 20) notes for the development of earlier tense vs. lax /æ/ in Cincinnati, which now “follows the general shift of Midland short-a toward the nasal system, in which tensing takes place before all nasals and only before nasals”. Campbell (1959: 8) suggested that early West Saxon spellings were influenced by Mercian scribal habits, and we will return to this possibility in Section 3.3 below. However, given new insights from exemplar theory and evidence of speak ers’ encoding of details in the phonetic signal (Pierrehumbert 2002; Port and Leary 2005), variation in a manuscript may also have other sources. For instance, it is possible that scribal spellings do not necessarily reflect classical phonemes, and that the scribes of early English are in general using the closest orthographic
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
125
units they have in their repertoire to indicate the sounds they wish to convey. As Hogg (1988: 191) expresses it: “It only makes sense to assume that scribal habits must have remained constant, at least in the earliest period, in that they always attempted, insofar as circumstances allowed, to represent the dialect of the writer accurately”. Colman (1983: 264) went so far as to suggest, “it is possible that OE spelling may fail to register all phonemic distinctions in (even one variety of) the language, and further, that it may register distinctions which are not phonemic, but allophonic”. The latter might especially apply to the change of [iw] to [ew], since the scribes distinguished the simple vowels [i] and [e] and had at their disposal orthographic symbols to represent those sounds. The methods used in the next section are designed to reveal any phonological or morphological patterns in the variation between long and .
2 Methodology and results The version of Hatton 20 used in the current study is Sweet’s (1871–1872) edition as it appears in the DOE corpus, entered into the Wordsmith program to create a wordlist, plus frequency information, for words containing the desired diphthongs. Tallied were only those words for which the diphthong received primary stress; inflected forms were tallied together, and derivatives were tallied separately, except for adjectives and adverbs, which were tallied together because all of the adverbs were transparent adjective + -lice formations. Verb participles used adjectivally were excluded because of the arbitrary choice of including them under verbs or adjectives and the difference in prosodic prominence of verbs and adjectives, which might affect vowel choice. Words beginning with the diphthong, such as eower and eowað, were omitted, since it was apparent that all but one instance of such words were spelled , most likely because word-initial could be mistaken for [j]. (In fact, the adverb geo ‘earlier’ appears as io five times in the manuscript.) Given those constraints, the results were as follows, starting with the percentage of spellings by word class, given in Table 1. It is clear from this table that word class matters. At 67.51%, nouns are the most likely to contain the spelling, whereas the three function words on average are spelled with only 7.85 per cent of the time.1
1 Brown and Raymond (2012) demonstrate that word class effects are not independent of word frequency for reductive processes in Spanish. See, however, Phillips (2006: 104–109) for cases of sound changes in which word class and word frequency behave independently.
126
Betty S. Phillips
Table 1: Percentage of spellings by word class Number of tokens
Number of headwords
Average % of per headword
254 56 399 386
50 18 23 3
67.51 47.76 41.56 7.85
Nouns Adjectives and adverbs Verbs Function words
Within word classes, there is also some difference, although it is less striking. The frequency figures for nouns are given in Table 2, where it is shown that the least frequent words have the smallest percentage of spellings and the most frequent words the highest percentage. An interesting pair to compare are stiora(n) (masc.) ‘steersman’ (5×) – as well as stiorere (masc.) ‘steersman’ (1×) and stiorroðer ‘rudder’ (4×) – versus steor(e) (fem.) ‘restraint, correction’ (3×). These may represent near-minimal pairs.2 The frequency range of verbs is displayed in Table 3. Here, too, the more frequent words are more likely to be spelled with , but by a very small margin. Table 2: Nouns: Word frequency effects Frequency range
Number of tokens
Number of headwords
Average % per headword
1–10 11–20 21–33
126 73 55
43 5 2
65.06 80.18 88.64
1–10: (1) deorum, fiounga, freondscipe, frioum, geðiode, hleorslægeas, hreofl, preostas, preosthades, stiorere ‘steersman’, ðeo, ðweorscipe, ðiofas, weop; (2) diobul, elðiode/ ælðiodigum, feowung, leohtmodnesse, plio, steopmodur, ðeodscipe, ðiowdomes, ðweoran, ungetreownesse, weopun; (3) steor ‘restraint, correction’, tweobleom; (4) ælðeodignesse, diofulgielde, gebeorscipum, sueor, stiorroðer; (5) bleo, feondscipe, freodom, stiora ‘steersman’, weod; (6) feos; (7) freond; (8) hreosunga, leohtfæt; (10) gestreon. 11–20: (11) leoht, weobud/weofud; (12) hreod; (19) breost; (20) deofle. 21–33: (22) feond; (33) geðeode.
2 The words cneowu, þeowote, treow, and þeow were excluded because of uncertainty over the length of the vowel at the time the manuscript was written. (See Fulk 1992: 146–152.)
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
127
Table 3: Verbs: Word frequency effects Frequency range
Number of tokens
Number of headwords
Average % per headword
1–10 11–100 250
65 87 250
18 4 1
32.18 35.30 35.60
1–10: (1) æteowde, ateon, bewreon, dreogað, atiowan, besio, gecniowon, gefiolle, plion; (3) giotan, ðiondan; (4) adriogan; (5) tiolað; (6) ðiowian, oftion; (8) afioll, bebeod; (10) underðiodan 11–100: (13) ðurhtion; (14) (ge)teon; (15) forsion; (45) geseon. 250: (250) beon.
Adjectives and adverbs were combined because of the transparency of the adverbs, which were all formed with -lice. They do not exhibit as wide a range of frequency variation, only 1–10, so their frequency distribution is not displayed in a table. The details, however, are as follows, with the frequency number in parentheses: (1) biorhte, diogolran, dioplices, freondlice, leoftælra, triowleas ena, tweolice, untreowsige; (2) deofullic, diorra/undiorestan; (3) diorwyrðe; (4) ðeonde; (5) leofre/leofusta; (7) freo(lice); (8) (un)deogollice; (10) elðiodig. Their average percentage per headword was 47.76. As for function words, there were only four that fit into the study, and eow(er) was omitted because all its spellings were in , as explained above. Table 4 shows that the least frequent, ðeos, has the highest percentage of spellings, namely 13.60; the next most frequent, heo, has a lower percentage of spellings, namely 8.54; and the most frequent, seo, has the lowest percentage of spellings, namely 1.42. Why the less frequent function words should show the innovative vowel contra the distribution observed in nouns and verbs will be addressed below. That this pattern is not peculiar to this manuscript is confirmed by similar behavior in the early West Saxon Orosius Lauderdale (or Tollemache) manuscript (siglum L). Despite her statement that “the graphs eo and io had become Table 4: Function word frequencies Token frequency 22 82 282
Word
%
ðeos heo seo
13.60 8.54 1.42
128
Betty S. Phillips
Table 5: Function word frequencies, Hatton 20 vs. Old English Orosius Word
% Hatton 20
seo heo ðeos beon þreo
1.42 8.54 24.00 35.60 75.00
% Orosius 65.29 39.17 58.33 90.00 82.35
equivalent in force by the time of the oldest manuscripts” and that “io-spellings occur only in a limited number of contexts in L and far less frequently than in [the] C[ura] P[astoralis]”, Bately’s (1980: xliii) introduction to her Early English Text Society edition adds that exceptions to “original” īo appearing as are “bion (3×), hio (40×), (…) sio (25×), (…) and þrio (3×).” (Ellipses in the quotation contain short diphthongs.) Most noticeably, these also are all function words, if one considers the number þrio as a function word, and a search of the DOE corpus for L reveals that ðeos also occurs with both and (7×). Since the numbers pattern differently from other words in other studies, ðreo was not included among the adjectives calculated above for Hatton 20, where it occurs once with and three times with : ðrio (1×), ðreora (3×) = 75% , 25% . Using the DOE corpus to determine the percentages of vs. in the Orosius for these words, it becomes clear in Table 5 that Hatton 20 is far more c onservative, i.e., has a much lower percentage of spellings for these function words than does the Orosius manuscript.
3 Discussion 3.1 Pattern of the merger The preceding data have shown that the long ~ spellings in the Hatton 20 manuscript are not random, but our test for whether the merger was still in progress included a requirement that the words affected pattern similarly to other, known sound changes. Our finding that function words are the most resistant to the change, and that for content words, the most frequent words are more likely to change first, indeed does mimic the pattern found in other strengthen ing processes, a category which includes vowel shifts (Kiparsky 1988: 378). It also fits some of the determinants of strengthenings identified by Zuraw (2009: 13–15), one of which is “Acoustic amplitude”, that “Lower vowels are (…) predicted to be stronger than higher vowels – the more-open vocal tract of lower vowels
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
129
gives them higher amplitude, all else being equal” (14). Another is “Complexity”, which according to Harris’s (1994: 103) autosegmental treatment, finds mid vowels stronger than the peripheral vowels [i, a, u]: “Viewed in terms of its elemental effects, mid-vowel raising is a reduction process”. Thus high-vowel lower ing would of necessity be a strengthening process. And seeing the strengthening property of this shift also helps to explain why, according to Table 4, within function words frequency patterns differently from that within content word categor ies. That is, sēo ‘the/that’, being an attributive demonstrative, would likely have a lighter stress than hēo ‘she’ and ðēos ‘this’. At least three vowel shifts behave similarly. Two occur in the Mercian gloss to the Vespatian Psalter, where, according to Hogg (1992: 139–140), the general raising of */æ/ > /e/ “in positions of weak stress (…was) not always carried through, hence æt ‘at’, cwæð ‘he said’, ðæs , ðæt (…) wæs are found alongside et, etc.”; in addition, earlier /a/ was also fronted to /æ/ in the Vespasian Psalter “with great regularity” except in “positions of weak stress” (140). Also, the modern shift of the Dutch vowels [ει] and [αι] in North-Hollandic dialects is documented by Gerritsen and Jansen (1980), who convincingly argue for these changes as strengthenings, including the evidence that the new variants occur preferentially in stressed syllables (36). In that shift, as in our study, the vowels of function words did not shift as readily. In fact, Gerritsen and Jansen, who saw this change as a spread of the Amsterdam variants, were able to set up an implicational scale whereby the dialects which exhibit this change in their pronouns also exhibit this change in the rest of their lexicons, and “All dialects with infrequent words changing to the Amsterdam variants have Amsterdam variants in their frequent words, too” (44). Strengthening consonant changes have also been found to affect function words last. Barrack (1976: 181) found that the shift from /d/ to /t/ in the Old High German Isidor affected only very frequent content words, as in fater, muoter, and got. And in another part of the shift, Schirmunski (1956: 217) found that the change of /t/ > /s/ in Middle Franconian noticeably affected all words except such function words as dat ‘that’, wat ‘what’, et ‘it’, dit ‘this’, alt ‘as’, at ‘all’, mōt (Ripuarian) ‘must’, and lēt (Ripuarian) ‘let’. Thus, the patterning of the /io/ > /eo/ shift is what one would expect of a strengthening sound change.
3.2 Schriftbilder Another interesting facet of the Hatton 20 data on and is the light it provides for the question of the reliability of scribal spellings, given that one might think that scribes would have a more engrained vision of the spelling of frequent words. For instance, Murray (2009: 489) expresses the concern that “an
130
Betty S. Phillips
apparent diffusion pattern from less to more frequently used words gleaned from manuscripts might have nothing to do with the properties of a particular sound change, but rather with the tendency to conservatively maintain the Schriftbild [written image – BSP] of frequently used words in spite of (a completed) sound change”. Thus, the tendency of the Hatton 20 scribes to maintain the spelling for pronouns might be accounted for by their strong Schriftbild, since they are also very frequent. Yet in our investigation, for the content words, the more frequent words are the very ones which are more likely to contain the innovative spelling. Another indication that a strong Schriftbild is not responsible for spellings in the manuscript is the treatment of function words for another sound change in the same manuscript. In the raising of /a/ before a nasal in Hatton 20, in contrast to the /io/ > /eo/ merger, function words were the category most likely to undergo change. The left half of Table 6 is taken from Phillips (1983: 489). On the right half is the data from Table 1 rounded to the closest whole number for comparison. The percentage change in the content words, especially adjectives and adverbs, is quite similar in the two sound changes. But the percentage change in function words is dramatically different: 84% for a-raising versus 8% for io/ eo merger. Because the assimilatory raising of /a/ before a nasal was a weaken ing process, function words were more likely to exhibit the change than were content words – with 84% of the function words exhibiting the innovative spelling (adverbs being combined with function words because of the difficulty of separating usages for words such as ðonne adv./conj., which occurs 1400 times in the manuscript (Phillips 2006: 200). In contrast, the strengthening of /io/, as shown above, affected words in low stress positions last. Therefore, one has to be careful when appealing to the influence of a Schriftbild. In a situation of a strong standardized spelling, certainly it has to be taken into account. But for many manuscripts that was not the case, and it is clear that the scribes of Hatton 20 were not limited by a Schriftbild, and certainly not for frequent words. Table 6: A comparison of two changes in progress in Hatton 20 a > o /__ Nasal Word class Nouns Adjectives Verbs Adverbs & Function words
io > eo
% of spellings Word class 44 44 32 84
Nouns Adjectives & -lice Adverbs Verbs Function words
% of spellings 68 48 42 08
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
131
3.3 Borrowing Separate from the issue of influence from a Schriftbild but related to it is the possible influence of Mercian spelling. Campbell (1959: 8), for instance, remarks that Early “West Saxon is fighting against the strong influence of Mercian spelling”. Again, the differences in spelling displayed in Table 6 are pertinent. The spellings in have also been attributed to Mercian influence (Toon 1983: 90–118; Nielsen 1992: 640–641), but, as indicated above, function words appear more often with the innovative spelling. The difference between it and the spellings in makes sense only if the scribes are recording their own pronunciation, since as a phonetically conditioned assimilation, one would expect function words to change first for the pre-nasal raising of [a], whereas the strengthening change of [io] to [eo] naturally affects the function words last. But if both changes are imitations of Mercian, they should behave alike regarding word class.
4 Conclusion A close investigation of the ~ spellings in the Hatton 20 manuscript of Alfred’s translation of Gregory’s Pastoral care has revealed that the spellings do most likely reflect the pronunciation of the scribes during the period before /io/ fully merged with /eo/, given that the spellings in Hatton 20 pattern similarly to the distribution of other strengthening sound changes. That is, the spelling is favored in content words over function words and, within content words, in more frequent over less frequent words. Other possible confounding factors such as borrowing from another dialect or the influence of the Schriftbild or visual image that a scribe might have for a word, which should cause frequent words to be more conservative, were found incapable of accounting for the patterns of variation displayed in the spellings in ~ in this early Old English manuscript. Thus, the evidence supports Smith’s (2007: 163) position that “[t]his change was still happening in historic times”.
References Barrack, Charles. 1976. Lexical diffusion and the High German consonant shift. Lingua 40. 151–175. Bately, Janet M. (ed.). 1980. The Old English Orosius (Early English Text Society s.s. 6). London: Oxford University Press. Brown, Esther & William D. Raymond. 2012. How discourse context shapes the lexicon: Explaining the distribution of Spanish f-/h- words. Diachronica 29. 139–161.
132
Betty S. Phillips
Campbell, Alistair. 1959. Old English grammar. Oxford: Clarendon. Colman, Fran. 1983. Old English /a/ ≠ /æ/ or [a] ~ [æ]? Folia Linguistica Historica 4. 271–291. DiPaolo, Marianne. 1992. Hypercorrection in response to the apparent merger of [a] and [ɔ] in Utah English. Language and Communication 12. 267–294. Faber, Alice. 1992. Articulatory variability, categorical perception, and the inevitability of sound change. In Garry W. Davis & Gregory Iverson (eds.), Explanation in historical linguistics, 59–75. Amsterdam & Philadelphia: John Benjamins. Fulk, R. D. 1992. A history of Old English meter. Philadelphia: University of Pennsylvania Press. Gerritsen, Marinel & Frank Jansen. 1980. Word frequency and lexical diffusion in dialect borrowing and phonological change. In Wim Zonneveld, Frans van Coetsem & Orrin Warner Robinson (eds.), Dutch studies IV: Studies in phonology, 31–54. The Hague: Nijhoff. Hall-Lew, Lauren. 2011. The completion of a sound change in California English. Online proceedings of the 17th international conference of the phonetic sciences (ICPhS XVII), Hong Kong, 17–21 August 2011, 807–810. http://www.icphs2011.hk/resources/OnlineProceedings/RegularSession/Hall-Lew/Hall-Lew.pdf (accessed 10 October 2013). Harris, John. 1994. English sound structure. Oxford: Blackwell. Healey, Antonette diPaolo, Joan Holland, David McDougall, Ian McDougall & Xin Xiang. 2009. The dictionary of Old English corpus in electronic form, TEI-P5 conformant version, 2009 release. Toronto: DOE Project 2009, on CD-ROM. Hogg, Richard. 1988. On the impossibility of Old English dialectology. In Dieter Kastovsky & Gero Bauer (eds.), Luick revisited: Papers read at the Luick-Symposium at Schloss Liechtenstein, 15.–18.9.1985, 183–204. Tübingen: Gunter Narr. Hogg, Richard M. 1992. A grammar of Old English. Vol. 1: Phonology. Oxford: Blackwell. Horgan, Dorothy M. 1982. The distribution of West Saxon dialect criteria in the extant manuscripts of the Pastoral Care. Studia Neophilologica 54. 217–235. Ker, N. R. 1957. Catalogue of manuscripts containing Anglo-Saxon. Oxford: Clarendon. Kiparsky, Paul. 1988. Phonological change. In F. J. Newmeyer (ed.), Linguistics: The Cambridge survey. Vol. I: Linguistic theory: Foundations, 363–415. Cambridge: Cambridge University Press. Labov, William. 2007. Transmission and diffusion. Language 83. 344–387. Labov, William, Mark Karen & Corey Miller. 1991. Near-mergers and the suspension of phonemic contrast. Language Variation and Change 3. 33–74. Lass, Roger. 1994. Old English: A historical linguistic companion. Cambridge: Cambridge University Press. Minkova, Donka & Robert Stockwell. 2008. Phonology: Segmental histories. In Haruko Momma & Michael Matto (eds.), A companion to the history of the English language, 29–42. Malden, MA: Wiley-Blackwell. Murray, Robert. 2009. Review of Word frequency and lexical diffusion by Betty S. Phillips. Language 85. 487–490. Nielsen, Hans F. 1992. Variability in Old English and the continental Germanic languages. In Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen & Irma Taavitsainen (eds.), History of Englishes: New methods and interpretations in historical linguistics, 640–646. Berlin: Mouton de Gruyter. Phillips, Betty S. 1983. Lexical diffusion and function words. Linguistics 21. 487–499. Phillips, Betty S. 2006. Word frequency and lexical diffusion. New York & Houndmills, Basingstoke, Hampshire: Palgrave Macmillan.
Alfredian īo ~ ēo, reluctant function words, and Schriftbilder
133
Pierrehumbert, Janet. 2002. Word-specific phonetics. In Carlos Gussenhoven & Natasha Warner (eds.), Laboratory Phonology 7, 101–140. Berlin: Mouton de Gruyter. Pierrehumbert, Janet. 2003. Probabilistic phonology. In Rens Bod, Jennifer Hay & Stephanie Jannedy (eds.), Probabilistic linguistics, 177–228. Cambridge, MA: MIT Press. Pitts, Ann. 1986. Flip-flop prestige in American tune, duke, news. American Speech 61. 130–138. Port, Robert F. & Adam P. Leary. 2005. Against formal phonology. Language 81. 927–964. Schirmunski, Viktor. 1956. Schwachbetonte Wortformen in den deutschen Mundarten. In E. Karg-Gasterstädt (ed.), Fragen und Forschungen im Bereich und Umkreis der germanistischen Philologie, 204–220. Berlin: Akademie Verlag. Smith, Jeremy. 1996. An historical study of English: Function, form and change. London: Routledge. Smith, Jeremy. 2007. Sound change and the history of English. Oxford: Oxford University Press. Sweet, Henry. 1871–1872 [1958]. King Alfred’s West-Saxon version of Gregory’s Pastoral Care (Early English Text Society o.s. 45, 50). London: Oxford University Press. Toon, Thomas. 1992. Old English dialects. In Richard M. Hogg (ed.), The Cambridge history of the English language. Vol. 1: The beginnings to 1066, 409–451. Cambridge: Cambridge University Press. Wordsmith, Version 5. 2007. Lexical analysis software, Ltd. and Oxford University Press. Zuraw, Kie. 2009. Treatments of weakness in phonological theory. In Donka Minkova (ed.), Phonological weakness in English: From Old to Present-Day English, 9–28. New York: Palgrave Macmillan.
Part II Reconfiguring History: Overlooked Evidence of English
Donka Minkova
Metrical resolution, spelling, and the reconstruction of Old English syllabification 1 Some preliminaries The basics of syllable structure are well known: the most sonorous part of the syllable is its peak, or nucleus, the segment(s) preceding the peak form the onset, and the segment(s) following the peak form the coda. The peak and the coda together form the syllabic rhyme. The peak is the only obligatory part of the sylla ble, while onsets and codas are optional. Vowels are always peaks, and the sonor ants /r, l, m, n/ can also be peaks in English. Syllable structure is characterized by one constant: the peak has to be filled. The optionality of onsets and codas allows consonants in VCV strings to be differently associated with the peak. The check-list of how word-medial syllable division works, as provided for example by Cruttenden (2008: 50, 258), involves three basic criteria:1 (1) Criteria on word-medial syllable division: –– Presence or absence of transparent morphemic boundaries (morphemic): glee.ful, slow.ness, non.ethical, mono.lith –– Compatibility of onsets and codas with the distribution of word-initial and word-final singletons and clusters (phonotactic): Sa.hara, Per.sia, ath.lete –– Syllable division as a rationale for specific consonantal realizations (allophonic): US á{ɾ}om, cápi{ɾ}al, informal UK á{ʔ}om, cápi{ʔ}al, aspirated [th] in a[t h]íre, unaspirated [t] in pla.stic. A widely attested syllable structure preference cross-linguistically is for intervo calic singleton consonants to be placed in the onset. The principle of filling the
1 Slashes enclose phonemic units, square brackets enclose phonetic realizations, angled brack ets enclose spelling forms, and curly brackets indicate ambisyllabicity. Syllable division is marked by a period.
138
Donka Minkova
onset in preference to the coda is known as onset maximality.2 An alternative way of syllabification in Present-Day English (PDE) is to assume that at least some intervocalic singletons are ambisyllabic, associated simultaneously with the onset and the coda. In addition, the consonant in VCV strings can be variably placed in the coda, or it can be analyzed as being both in the coda and in the onset. These options for English are summarized in (2): (2) Syllabification of happy in English: [hæ.pi] [hæ{p}i] [hæp.i]~[hæ{p}i] [hæp.pi]
Hayes (1995), Gussmann (2002) Kahn (1976), Giegerich (1992), Kreidler (2004), Hayes (2009b) Hammond (1999) Burzio (1994)
[p] onset maximal {p} ambisyllabic [p] ambiguous [p] functioning like a geminate
In Present-Day English ambisyllabicity is one way of accounting for the realiza tion of the dental stops /t/ and /d/ as alveolar approximant taps [ɾ] before an unstressed vowel in American English, as in ladder [ˈlæ{ɾ}ɚ], waiter [ˈweɪ{ɾ}ɚ]. In Received Pronunciation, the phoneme /ɹ/ is realized as a tap [ɾ] in ambisyllabic environments (Rubach 1996). While alternative accounts of these and other allophonic patterns have been proposed, ambisyllabicity remains a widely used analytical strategy for addressing both phonetic alternations and stress assignment.3 Clarifying the empirical base and the type of argumentation in favor of or against ambisyllabicity in earlier English is therefore central to the
2 The principle of onset maximality is commonly extended to clusters which are also possible word-initial clusters, the phonotactic criterion in (1), so that pilfer syllabifies as [-ɪl.fə-], while petrified syllabifies as [-ɛ{t}r-], see Hooper (1972), Anderson and Jones (1974) for an early treat ment, and Jones (1989: 183–195) for applications of this principle to cases of historical epenthesis and metathesis in the history of English. Some problems with this approach are discussed in Ní Chiosán, Welby and Espesser (2012), see also their references. This paper concentrates on -VCV- sequences; the syllabification of intervocalic consonant clusters will not be discussed here. A related issue is the co-extensiveness of word structure and syllable structure in Germanic: a minimal syllable should equal a minimal word (Liberman 1990a, b). For Old English, this as sumption is questioned in Fulk (1997); his objection is based on the treatment of some items in verse, e.g., disyllabic scansion of dōn, gān. Non-isomorphy between syllable and word is also assumed in Russom (2002: 309), see also Gordon (2006) for cross-linguistic findings allowing non-isomorphy. 3 For the history of the research on ambisyllabicity in Present-Day English, the phonetic diag nostics and the rules of ambisyllabicity, see the overview in Hayes (2009b), Minkova and Zuraw (forthcoming).
Metrical resolution, spelling, and Old English syllabification
139
reconstruction of the structural and prosodic roots of Present-Day English, and to the study of the universal principles of speech segmentation. Following these preliminary remarks, Section 2 surveys the positions on ambisyllabicity in Old English (OE). Section 3 discusses the relevance and reliabil ity of Old English verse evidence for syllabification. Section 4 is a reappraisal of scribal practice with respect to word-division and the incidence of orthographic gemination in Old English manuscripts. Section 5 offers a summary and some suggestions for further research.
2 Was there ambisyllabicity in Old English? The question of possible ambisyllabicity was not discussed in the earlier canon ical descriptions: Campbell (1959), Luick (1914–40), and Jespersen (1909). CHEL 1 (Hogg 1992b: 96–97) leaves the options open, syllabifying OE stānas ‘stones’ as staa.nas (p. 96) and staa{n}as (p. 97) without arguing for either option. Ambisyl labicity restricted to V̆CV- for Old English, for example sċi{p}u ‘ships’, wo{r}orld ‘world’, is assumed in Hogg (1992a: 44–45) and Suzuki (1994, 1995). Jones (1989), Lass (1992: 69, 74) and Ritt (1994: 52–64) assume ambisyllabicity for Middle English (ME), while onset-maximal syllabification for Old English and Middle English has been defended by Fulk in his entire work, most prominently in Fulk (1997).4 Suzuki (1994) treats all singletons and possible word-initial clusters following any stressed vowel as ambisyllabic. His arguments are based on syllabification of the Proto-Germanic (PrG) trisyllabic sequence VVCR̩(R)V in pre-Old English, Old English Breaking, voicing of fricatives, and /h/-deletion. Identifying empirical problems in the cases discussed by Suzuki, and appealing to high-vowel dele tion, resolution in the meter, Old English word-division at line ends, and opensyllable lengthening, Fulk’s vigorously “contrary” view (1997) is that none of the evidence for ambisyllabicity in Old English is compelling, and some of it is phil ologically unsustainable. Fulk’s reaction to Suzuki’s assumption that ambisyl labic consonants are not full-fledged contributors to syllable weight, unlike real coda consonants, is that it is an untestable phonetic speculation (1997: 29), and pursuing it would be fruitless for a dead language. His objections are structural and philological, and he does, indeed, identify a number of problematic points
4 Rejecting ambisyllabicity as too restrictive, Bermúdez-Otero (2007: §20) includes a diachronic perspective limited to phrase-level resyllabification in Middle English, referencing onset maxi mality in the realignment of, e.g., an uncle > nuncle, an ewt > newt. This is not conclusive because the re-association can also go in the other direction: a napron > an apron, a nadder > an adder.
140
Donka Minkova
in Suzuki’s choice of examples providing empirical support for ambisyllabicity, especially with regard to the behavior of intervocalic clusters. He also points out, correctly, that in some cases alternative accounts may predict the same behavior, as is the case with the intervocalic voicing of the fricatives in, e.g., knife ~ knives, grass ~ graze, bath ~ bathe. This paper will not review all arguments for treating intervocalic consonants in Old English as ambisyllabic or not. Instead, the principles of syllabification in Old English will be addressed with reference to two areas: the behavior of VCV strings in verse, and the orthographic choices on division of words at line-ends and consonant gemination in the Old English manuscripts.
3 Resolution in Old English meter Prosodic rules (…) constitute a paralinguistic system that specifies the poetic language as a derivative of the system (not necessarily of the surface representations!) of ordinary lan guage. From this point of view, poetic language is a potentially abstract entity, in that the metrically relevant representations need only have a virtual existence as stages in a gram matical derivation, with no physical realization at all (Kiparsky 1977: 241).
One of the most important tests for syllable weight in Old English is the metrical phenomenon known as Resolution, illustrated in (3): (3) Old English Resolution:5 a. mihtig meredēor ‘mighty sea-beast’ b. snellic sǣrinc ‘sturdy sea-man’
S w / S ͡ w s Beo 558a S w / S s Beo 690a
Resolution is based on the metrical equivalence of a sequence of a stressed L(ight) syllable plus the following unstressed syllable, as in me.re ‘sea’ in (3a) and a H(eavy) stressed syllable, as in sǣ ‘sea’ in (3b). In (3a), a “normal”, i.e., four-position, scansion is impossible unless both syllables of mere jointly fill the first strong position of the second foot; mere is thus metrically equivalent with sǣ in (3b). It is this equivalence that seems to be an insurmountable problem for the ambisyllabicity hypothesis: since syllables with filled codas are heavy in
5 Capital S indicates a strong position filled by a syllable with an alliterating onset. Small s is a strong position filled by a non-alliterating syllable. Small w stands for a weak metrical position; weak metrical positions can be filled by more than a single unstressed syllable. The tie symbol in S ͡ w indicates a “resolved” strong position. The slashes divide metrical feet. Further informa tion on the conventions of Old English scansion and resolution can be found in Stockwell and Minkova (1997a, b).
Metrical resolution, spelling, and Old English syllabification
141
the meter, as in mih- in mihtig ‘mighty’ in Beo 558a, if the intervocalic consonant of mere is also in the coda of the first syllable, that syllable is already heavy and resolution ought to be blocked.
3.1 Resolution and the variable weight contribution of intervocalic consonants The argumentation for and against ambisyllabicity rests on the shared assump tion that syllable boundary placement is a factor determining the weight of the syllable: in V́.CV the first syllable is always light, while in V́C.V the first syllable can be heavy – the assumption relates to the cross-linguistic propensity of heavy syllables to attract stress. Functionally, V́C syllables in Old English align with heavy syllables of the type V́V, V́VC, V́CC, V́VCC; all five types can be full lexical items as in fær ‘journey’, sǣ ‘sea’, blōd ‘blood’, benc ‘bench’, brēost ‘breast’. In the verse VC syllables also behave like the other heavy syllables, as illustrated in (3). Suzuki (1994, 1995) confronts the metrically based objection to ambisyl labicity and sets it aside with the argument that “the syllable that is closed on account of ambisyllabicity (-V́CV) consistently counts as lighter than the genu inely closed syllable (-V́CCV- or -V́VCV-), as most persuasively demonstrated by meter” (1994: 79). I believe that skepticism about metrical evidence as an irrefut able argument against amisyllabicity is the right approach, but not for the reason given by Suzuki. The position needs to be explored and clarified further. To avoid the conundrum of a V́{C} syllable as light or heavy, Suzuki’s formal representation of the relevant syllabic associations (1994: 80) assigns units of weight (moras) to intervocalic consonants derivatively, whereby the initially onset-maximal consonant is adjoined to the coda of the syllable to the left. The claim is that ambisyllabic consonants are never moraic, while full codas count as moras. However, there is good evidence that “intrinsic” mora count is an unstable notion (Hammond 1999: 137), and extensive cross-linguistic research has shown that the reference to discrete and strictly binary timing units of syllable weight is problematic. Gordon (2006) and Ryan (2011) argue convincingly that in natural living languages syllable weight is gradient; for gradient syllable weight in the Old English verbal system, see Minkova (2012). The specific contradictions that emerge from the all-or-nothing approach to weight adopted in Suzuki have to do with the nature of the consonant and the conflict between weight and stress. First, labeling all singletons following a stressed vowel as non-weight-contributing faces the problem of compensatory lengthening following intervocalic voiceless velar fricative loss in Old English, as in PrG *teuh-an, OE tēo(ha)n ‘to tug’; PrG *sehw-, OE sēon ‘to see’; Gothic
142
Donka Minkova
swaihra, OE swehor (early) ~ swēor ‘father-in-law’, PrG *sleah-an > OE slēan ‘slay’, PrG *þleu.han > OE flēon ‘flee’. Of significance in this context is that in the verse words like sēon ‘to see’, flēon ‘to flee’ scan preferentially as disyllabic in the early poems, as was thoroughly documented and argued in Fulk (1992: 92–104). This is illustrated in (4a, 4b): (4) Absence of resolution in Old English VhV sequences: a. feorhsēoc flēon ‘life-sick flee’ S s / S w b. dēaþwīc sēon ‘death-place see’ S s / s w c. ǣfre flēon ‘ever flee’ S w / S w
Beo 820a Beo 1275b Instructions 40a6
It has to be admitted that the negative evidence of absence of resolution at the right edge of the verse is not conclusive because of the special requirements that constrain the material placed in that position. But the disyllabic scansion of the verbs in (4) has to be evaluated also in the context of later changes: loss of /h/ and lengthening of the vowel. The usual explanation of the lengthened vowel is onset [h]-loss and vowel contraction, but an alternative scenario would be lengthening as a consequence of the weakening and loss of ambisyllabic [-h-] which yields its weight to the preceding vowel. The loss of the already weak consonant in the coda is compensated by increased vowel length; coalescence with the previously unstressed vowel follows. The two pathways yield the same result, and both rely crucially on the inherent properties of the segment, but attributing /h/-loss to the association of the consonant with the coda has the (limited) advantage of avoid ing the problem of stability of OE /h-/ word-initially, not a good fit for an account linking [h] exclusively with the onset position.7 Second, in Suzuki’s interpretation, the coda of an unstressed -VC syllable counts as weight-contributing, as in swutol ‘clear’, which is a sequence of a
6 The full line is ǣ fre flēon / unrihte gestrēon ‘ever flee / unright gains’. The relevance of the ex ample is the pairing of flēon : gestrēon, the latter with an etymologically long diphthong, which is strongly suggestive of the completion of the lengthening in flēon. 7 As Robert Fulk has pointed out to me, this is not the only possible scenario for vowel con traction involving /h-/ loss in Old English. The end-points are not in dispute: disyllabic –VhVstrings emerge as monosyllabic –VV- strings, but an intermediate disyllabic string with the first vowel preserving its original length in hiatus, -V. V- , could have been used by the poets, since uncontracted disyllabic forms do not testably appear in positions where an initial heavy syllable is required in the meter. A parallel of this is the treatment of sie ‘be, subj. sg.’, dōn ‘to do’ and similar forms, where the contraction is non-compensatory (Hogg and Fulk 2011: §6.148, §6.154). A full re-examination of the metrical and spelling evidence might help us decide which pathway is more probable; for now I prefer to think of /h-/-loss as a trigger of compensatory lengthening, a factor which overrides the tendency to overall length stability.
Metrical resolution, spelling, and Old English syllabification
143
L(ight) + a (H)eavy syllable, but the shared coda of a stressed syllable, the [‑t-] of swu{t}ol, is non-moraic, as is supposedly the coda of ǣ r ‘before’. Such moraic assignments are based on an ideal of a bimoric stressed syllable or foot, but run against the phonological realities of conservative Old English, where “super heavy” syllables, such as the first syllables of nǣddre ‘adder’, ðēostre ‘dark’ are unexceptional. Assigning a full mora to all codas in unstressed syllables is also problematic, especially in view of the possibility that stress-assignment is blind to such codas, i.e., they are considered non-weight-bearing, or extrametrical – for this analysis in Present-Day English, see Hayes (1982). One inference from this specific instantiation of V́CV is that for the period when [h] was still present intervocalically, it was possibly ambisyllabic, enhancing the weight of the first syllable. The point to be taken away and studied further is that historically the behavior of ambisyllabic consonants may have been variable depending on the nature of the segment, as is indeed the case in Present-Day English.
3.2 Syllabification and the paraphonological component of meter Another argument related to considering the option of ambisyllabicity in Old English in spite of resolution comes from the general principle of matching metrical templates to surface realizations. In Kahn’s (1976) classic analysis of Present-Day English, the basic rule for syllabification is onset-maximal: VCV is syllabified V.CV. In slow and careful speech syllabification stops at that point. The allophonic variation observable in, e.g., a./t/om ~ á{ɾ}om ~ a.[th]ómic is attri buted to s ubsequent association of the onset to the coda of the stressed syllable to the left. This process is optional and characterizes more informal and rapid speech styles. Put in terms of analytic levels, underlyingly all VCV sequences are V.CV, but the surface realizations can be either V.CV or V{C}V.8 Projecting Kahn’s analysis to earlier English: if we assume that like PresentDay English, Old English syllabification starts out as onset-maximal, the s election of V́CV structures for the metrical rule of resolution is predictable: V́.CV meets the general structural description of the environment for resolution. Potential ambi syllabic associations in speech which may affect the weight of the first syllable are ignored in evaluating whether a line is metrical or not. At first blush this may look like a heretical decoupling of the verse evidence from the forms of the spoken
8 It should be noted for comparison that in the framework of interleaved morphological and phonological operations (Bermúdez-Otero 2011), the re-association of onsets to codas occurs in a different cycle of the derivation.
144
Donka Minkova
language, yet any theoretical framework that distinguishes between meter and speech will resolve the apparent conflict. As Kiparsky (1977) argues, meter is com ponential. One component is the pattern generator, specifying the size and the internal structure of the verse design. A second component is the paraphonology, or the “metrically relevant range of phonological representations” of phonological units (1977: 90).9 A third component, the “comparator”, matches the paraphono logical representations to the meter to ensure compliance with the verse design. Since paraphonological options are available to all poets at all times, the argument in the context of Old English meter is that the metrically relevant phonological representation of a V́CV string is V́.CV in those positions where resolution applies. In some well-studied cases of paraphonological choices such as postvocalic syncope in Present-Day English (Kiparsky 1977), Middle English pre-vocalic elision of , vowel apocope, syncope, and contraction (Minkova 2009), this is the “phonology of opportunity”, where optional representations are used selectively in the verse. Taking just one example, the treatment of unstressed final in an early, southern Middle English text, The owl and the nightingale, we find that the poet treats stem-final variably, as illustrated in (5): (5) Treatment of , OE lufu ‘love’ in The owl and the nightingale:10 a. Disyllabic: ne last his luue no leng more 517 al mai þe luue gan awai 1510 hit nis for luue noþeles 510 Swiche luue ich itache & lere 1347 b. Monosyllabic before vowels an habbe boþe luue & þonc 461 & soþ hit is of luue ich singe 1339 c. Monosyllabic before consonants an luue ne deþ noȝt bute rest 1452 þat dusi luue ne last noȝt longe 1466 In non-northern Middle English in the first half of the thirteenth century the final unstressed vowels in lexical disyllables were probably still part of such words’
9 For a discussion of the componential organization of meter based on Kiparsky (1977) and an application of the componential approach to metrical analysis within Optimality Theory, see Hayes (2009a). 10 The lines are cited from Stanleyʼs edition of The owl and the nightingale. The edition is based on the older manuscript, London, British Library, Cotton Caligula A. ix, dated c1275(?a1216) (MED), Central Worcester (LAEME).
Metrical resolution, spelling, and Old English syllabification
145
underlying representation. The poet uses this “regular” phonological option in (5a), where the nom. acc. sg. of the originally disyllabic stem OE lufu and its inflected forms scan as disyllabic both before consonants and before vowels. The more common pattern before vowels for both inflected and uninflected forms, however, is shown in (5b); elision is a contextually conditioned paraphonologi cal option which must have been present in the poet’s rapid speech phonology. The elision option is extended in (5c) to a pre-clitic position, irrespective of the segmental environment. The text represents a diachronic stage in which the para phonology is morphologically constrained: in the absence of a prevocalic context dative forms of luue in The owl and the nightingale are all disyllabic. Jumping a century and a half forward, this is no longer the case in Chaucer: a spot-check of the phrase for loue verse-medially in the first four Fragments of The Canterbury tales shows love as monosyllabic three times before consonants and three times before vowels. In historical phonology such evidence is valued highly as one of the quantifiable indicators of change in progress; in the case shown here an argument can be made that by the second half of the fourteenth century London English had reanalyzed love as monosyllabic, with disyllabic love as a paraphon ological option, especially favored at line ends. For another illustration of how meter selects phonological representations opportunistically, we can turn to the variable realization of post-consonantal /r, l, m, n/ as syllabic or non-syllabic in the Old English poetic corpus. Examples of this behavior for tācn ‘token’, wuld(o)r ‘glory’, symb(e)l ‘feast’, drawn from the data cited in Fulk (1992: 66–91), are shown in (6): (6) Variable treatment of -C + /r, l, m, n/ in Old English verse: a. Monosyllabic: Sete sigores tacn S ͡ w / S ͡ w w s GenA 2313a wuldorspēdige weras S s w w / S ͡ w And 428a symbelwynne drēoh S s w / s Beo 1782b b. Disyllabic tacen sette Sw/sw GenA 1044b wuldorþrymmes S w / s w And 702b symbel þicgan S w / s w Beo 1010b Focusing on etymologically monosyllabic stems, Fulk shows that “parasiting”, the disyllabic metrical use of such stems, is not randomly distributed in the poems. He tabulates the proportion of the occurrences of mono- vs. di-syllabic forms (1992: 83–84) and finds a positive correlation between monosyllabicity and early date of composition of a poem, proposed on independent grounds. Poems presumed to belong to the ninth century or later invert the earlier proportions in favor of disyllabic use. Fulk argues convincingly that this is a good heuristic
146
Donka Minkova
procedure for testing the dating of Old English verse. Since West Germanic “para siting” is a pre-Old English change, the monosyllabic use of the relevant stems in verse means that the poet has access to archaic forms of the language. This analy sis does not preclude disyllabic realization in speech. In terms of paraphonolo gical options, we can interpret the chronology of the poetic use as a switch from original underlyingly non-syllabic sonorants in the tācn, symb(e)l type words to fully syllabic [-ən / -n̩], [-əl / -l̩], etc., in later verse. In both early and late verse the primary choice is with the underlying realization. Although couched in a different theoretical framework, the potential of meter to select different representations for post-consonantal sonorants can be illus trated for Present-Day English, too. A well-known example of meter referring to underlying form rather than surface realization is the account of the behavior of in Present-Day English meter and speech. The first step of the analy sis posits an input non-syllabic /m/ in words like abysm, baptism, chasm, prism, schism; /m/ becomes syllabic on the surface by the rule in (7): (7) m → [+syllabic] / C − # (from Hayes 1988: 228) The rationale for positing underlying non-syllabic analysis as set out in Hayes (1988) is that /m/ is realized as non-syllabic before vowel-initial suffixes (spasmodic, orgasmic, etc.; that -sm# words are surface exceptions to a general rule placing the primary stress on the rightmost non-final stressed syllable of a stem, e.g., exceptional enthúsiàsm vs. regular enthùsiástic, suggesting that stress-place ment treats /m/ as non-syllabic. The “lateness” of /m/ becoming syllabic either as [m̩] or as [əm] is arguably also attested by the fact that if ‑sm# was syllabic in baptism, it would violate a rule of Post-Stress Destressing, which removes weak stress from nonfinal syllables when it immediately follows strong stress, as in sénsory from /sénsòry/ (cf. áuditòry). On the other hand, the realization of as disyllabic in speech is confirmed by the intuitions behind dictionary transcrip tions for such forms, which are always with a syllabic : [-zəm ~ -zm̩]. The evidence of English verse shows that poets treat commonly as non-syllabic, especially, but not exclusively, before vowel-initial function words where the sonorant can be associated with the onset of the vowel-initial sylla ble within the clitic group, e.g., abysm of [‑ɪz{m}əv], chasm of [‑æz{m}əv].11 This lends support to the theory that poetic language can draw on representations that differ from the surface forms. The option of disyllabic exists, too, especially before consonant-initial words, as the examples in (8b) show.
11 “Down verge of an abysm of stagnant air”, Walter De la Mare (1873–1956), Winged chariot (1951 l. 688) and “A sudden chasm of ghastly light”, Emily Brontë (1818–1848), The complete poems (25:1).
Metrical resolution, spelling, and Old English syllabification
147
(8) Variable treatment of in English verse: a. Monosyllabic : (There floweth daily forth a stream of joy) Into a chasm whose depth we know not of (Henry Alford, The school of the heart 1868, lines 123–124)
Her own fond image in this prism survey’d, (The farmer-lady sees a grace display’d) (Samuel Jackson Pratt, Cottage-pictures 1803, lines 571–572)
b.
(Since here thy word hath shown wherein) The deadly guilt of schism consists. (Bernard Barton, The true schismaticks 1826, lines 3–4) Disyllabic : And from the chasm flung between (Comes up the roar of tides unseen.) (Arthur Henry Adams, London streets, interlude – Eurydice 1906, lines 9–10)
With force endowed it prism-wise, whereby (All motives to themselves men justify) (Philip James Bailey, Festus XL 1877, lines 28551–28552)
(Become the arms and ammunition) To muster schism and sedition. (Jacob Bailey, Narrative verse satire in Maritime Canada 1779–1814, lines 156–157) The selective use of underlying representations is “widely observed in other met rical systems: the phonological representation scanned is one in which some or all of the phonological rules are ‘undone’” (Hayes 1988: 228). Going back to the issue at hand: Old English resolution and the syllabification of V́CV sequences. It has been shown clearly that resolution applies selectively in the meter depending on the position of the resolvable sequence; for details see Fulk (1992, 2002), Suzuki (1995), Hutcheson (1995), Russom (1995, 2002) and references there. Resolution applies systematically only to stressed syllables. Resolution is much more robustly attested in the on-verse than in the off-verse.12
12 “A resolvable sequence occupies one metrical position of the verse when resolved, but may stand unresolved toward the end of the verse, occupying two metrical positions” (Russom 2002). Hutcheson (1995: 95) points out that in some contexts, e.g., S s / S ͡ w w : dryhtcwēn duguþe (Wid 98a), or S ͡ w s / S w: merestrēam mōdig (Ex 469a), resolution is effectively restricted to the a-verse.
148
Donka Minkova
These are metrically circumscribed restrictions, not unlike the choices regarding the syllabicity of in (5) or the treatments of the sonorants in (6) and (8). The actual performance of verse in Present-Day English is an open-ended matter, quite possibly varying from individual to individual; it is at best p eripheral to the analysis of the correspondences between the metrical template and the spoken language. Nevertheless, some commonly assumed pragmatic aspects of the composition, consumption, and transmission of Old English verse do seem relevant in the context of resolution and the modes of syllabification. One such consideration is the oral and often formulaic nature of the compositions inten ded not for private reading, but for public recitation. The delivery of verse, there fore, would have been associated with slow and careful speech, increasing the likelihood of onset-maximal syllabification of V́CV strings. Also, the scribes who recorded Old English verse were familiar with Latin verse, where onset-maximal syllabification is the rule not just word-internally, but also across word-bound aries, see also 4.1. below.13 The “sloppiness” and optionality of ambisyllabicity (since the resulting lenitions and deletions are less effortful) would be inadmis sible in the context of strict adherence to syllabic measures. As Ælfric writes on meter in his Grammar: (9)
Ælfric (ca. 955 – ca. 1010) on meter: Se cræft is swa ameten, þæt ðær ne mot beon furðon an stæf ofer getel ac beoð ealle þa fers geemnytte be anum getele, gif hit aht beon sceal. ‘This art is so measured out that there may not be one letter beyond the number and all the verses should be equal, if it should be anything good’.
Thus, seen from this practical angle, too, the maintenance of V́.CV syllabification in Old English verse seems logical. The interpretation of why resolution survives with any regularity in Old English verse in this section is compatible, but not identical, with Fulk’s descrip tion of Old English verse as diaphanous: A characteristic of Old English meter (…) is its embodiment of metrical archaisms. In part this is a consequence of poets’ knowledge of verse traditions (…) A meter of this sort (…) is thus (…) a diaphane, inasmuch as in the course of scansion one must look through the surface forms and take into account a historical dimension that underlies the recorded text (Fulk 2005: 151).
13 In Latin, word boundaries can be ignored in syllabification: ab oris is metrically a.bo.ris. In Old English, compounds retain pre-compound word boundaries: Beo 78a: healærna mǣst ‘of hall-rooms largest’ is S s w / s, i.e., heal.ær.na not *hea.lær.na, in conformity with the morphemic criterion in (1).
Metrical resolution, spelling, and Old English syllabification
149
This is obviously not the first time that a parallel has been drawn between syn chronic modeling and diachronic change. Old English resolution can thus repre sent both an earlier and an underlying type of syllabification, which is phonolo gically also onset-maximal syllabification.14 Summing up the discussion so far: like Present-Day English, Old English syl labification was underlyingly onset-maximal; this is the principle behind the use of resolution in the verse, while ambisyllabicity is a “surface” phonetic event, also possibly a chronologically later event. Positing exact moraic content for a syllable is unreliable without reference to the nature of the segments, a point illustrated by the behavior of intervocalic [h] in early Old English. An appeal to the concept of paraphonology is important because it allows us to identify phonological repre sentations that are specific to a particular type of verse. On the assumption that the paraphonology of Old English provided access to both onset-maximal V́CV syllabification and ambisyllabification, the selection of V́.CV structures in speci fic verse positions does not constitute evidence against the possibility of realiza tion of VCV structures as V́{C}V.
4 Orthographic tests for Old English syllable structure Spelling has always provided support for the reconstruction of past phonological states. Scribal omission of final and its unetymological insertion in late Old English through Middle English is a cornerstone in the reconstruction of final vowel loss in English (Minkova 1991: 45–62). One type of argument used in Fulk
14 One of the compelling initial arguments in favor of positing ambisyllabicity in Present-Day English is the absence of (C)V́# monosyllables (Kahn 1976). Fully stressed light monosyllables do not exist in Old English, either, setting aside the possibility of ambivalent weight for some pronominal forms; the same situation obtains in Proto-Germanic. From that point of view an ambisyllabic realization in pre-Old English is an appealing option historically. However, the co extensiveness of word structure and stressed syllable structure can be challenged (see note 2 above), as in Singapore, gingham in Present-Day English, where [ŋ] is ambisyllabic (Hayes 2009). There is also the issue of pre-Old English High Vowel Deletion, one of the most highly valued tests of the ‘lightness’ of the stressed syllable in V́CV forms, accounting for the historical differ ence between nom. acc. plural sci.pu ‘ships’ vs. word ‘words’ < *wor.du. It is therefore probable that pre-Old English syllabification was more pervasively onset-maximal, with ambisyllabic re presentations emerging and gradually becoming stronger in Old English – this scenario fits the demise of resolution as a metrical device in late Old English – and early Middle English, but it does not exclude the strikingly archaic use of resolution as late as the Middle English Poema morale, as argued in Fulk (2002).
150
Donka Minkova
(1992: 90–91) to support his analysis of “parasiting” noted in 3.2 above, is that in some instances the orthographic evidence indicates unambiguously that mono syllabic forms of, e.g., ēþel ‘country’ could not have arisen by analogy, because irrespective of the use of the form in early verse, the spelling shows a vowel before overwhelmingly, over 200 times, against a single instance of , while with the originally disyllabic sāwol, spellings with and without a vowel in the second syllable are common. Examples of reliance on orthographic evidence in historical phonology are easily multiplied; what I want to focus on in the following section, however, is a more specific variety of orthographic evidence, the treatment of words at line-ends in the Old English manuscripts.
4.1 Word-division in Old English manuscripts The principles and the data on word-division in the Old English manuscripts were extensively documented and analyzed by Wetzel (1981); a subset of the same data was also discussed in Lutz (1986). Wetzel collected a corpus of 125,000 attes tations of divided words in 168 Old English manuscripts. The most significant single group of attestations, 52,000, are derived words. Unsurprisingly, the sylla bification principle followed for this set is morphemic, as in (1) above; in 98.9% of the cases (Wetzel 1981: 45–46) this principle governs the word-division. The subset of divided words with a single intervocalic consonant, graphic sequences, in Wetzel’s study, a total of 30,442 instances, shows a division in 99.4% of the cases (1981: 110). This subset bundles together inflected and uninflected words and words derived with -ing, -ung, -ende, etc. A closer look at the remaining 0.6% of exceptions (188×) where the division is , reveals that these are almost exclusively divisions in which the single consonant belongs to the stem, followed by a vowel-initial suffix: gebyr-eð, cwæd-on, muð-es, min-um, dys-ige, etc., raising the level of predictability even higher, to 99.9%. Indeed, only 48 divi sions out of 30,442 total are of monomorphemic words, e.g., of-er ‘over’, wor-uld. This vanishingly small number of exceptions should make anyone suspi cious, and Wetzel recognizes, of course, that the level of consistency is surpris ing in such a large corpus, but he dismisses the possibility of scribal convention. His reasoning is that if the scribes were taught to divide orthographic into , this should apply to the letter representing [ks], but in fact this letter is placed to the right only about 15% of the time. But this is not an argument against the learned practice of word-division: the inconsistency of division has to do with the fact that it is a single letter representing a cluster, while all other single-consonant letters correspond to a single segment. There were no
Metrical resolution, spelling, and Old English syllabification
151
[ks-]-initial words in the language, while [-ks], spelled was common, so we can expect the scribes to be reluctant to place to the right. The unprecedented pedantic adherence to division could be doubly motivated. On the one hand, one can imagine that the Old English scribes, writing under conditions of slow and careful self-dictation, would syllabify onset-maxi mally, without regard to further changes in speech, see 3.2 above. Reading would also be slow and deliberate. Addressing specifically the treatment of written language when read aloud, Ælfric Bata, a disciple of Ælfric’s, admonishes the student: “legite distincte et aperte atque verbatim sed et syllabatim ac sensatim” (Gwara 1997: 182) [‘read distinctly and clearly and word for word, but also syllable by syllable and according to sense’]. On the other hand, a 99.9% consistency is so unusual for any set of forms in a living language that it suggests more than a mere linguistic basis in the practice of word-division. Additional consideration should be given to the scribes’ back ground and the setting of Old English manuscript production. As far as we know, in Anglo-Saxon England lay literacy was practically non-existent. The scribes who produced vernacular manuscripts were trained in Latin; knowledge of Latin was equivalent to education. In Alfred’s age the main attention of the men of learning was on translating from Latin, and even the distinction between the two scripts used for writing Latin, the Caroline minuscule, and the Anglo-Saxon minuscule, became “blurred (…) and most of the letter-forms distinctive of AngloSaxon minuscule dropped out of use in writing English” (Roberts 2005: 2–3). The only linguistic instruction available to the scribes was Latin orthography and grammar. And this, I believe, is the key to the striking regularity of the pattern of word-division found in the corpus. From the Greeks15 to the Romans, in Priscian’s Institutiones grammaticae, and Donatus’ Ars grammatica, both of them widely used textbooks for the study of Latin, the “rule” is onset-maximal worddivision. It is commonly assumed that the Romans borrowed this practice from the Greeks and it was an easily acquired spelling convention: [T]he particular statements of the Roman grammarians on which the received doctrine [ syllabifies as , DM] is founded represented neither the facts of Greek pro nunciation nor the facts of Roman pronunciation, but had their origin in a mere practical rule – admirably simple and easy of application – devised by some Greek grammarian for the division of words in writing, when one was near the end of a line and had room
15 The Greek roots of the word-division practice are widely recognized, though not attributed to a particular source. William Goodwin’s highly influential Greek grammar (1892: 24) states: “The following rules, based on ancient tradition, are (…) observed in dividing syllables at the end of a line: Single consonants (…) are placed at the beginning of a syllable”.
152
Donka Minkova
for part of a word only; (…) this rule was adopted bodily by the Roman grammarians (Hale 1896: 251–252).
The extent to which this “admirably simple and easy of application” rule over laps the “natural” syllabification in Latin cannot be discussed here, nor is this the place to explore the issues posed by the syllabification of hetero-organic consonants clusters.16 What is of interest in the context of evaluating the evi dence from the manuscript practice of word-division in Old English is that the original “doctrine” of adopted by the scribes referred to the written word: “in scribendo”, “in scriptura” are phrases used repeatedly by the grammarians in reference to the rule. Syllable division is then not about phonetics, but about writing conventions. Hale (1896: 269) concludes that the rules “provide a clear method to follow ‘in scribendo’, and nothing could be more natural than that the scribes should accept gratefully what was so happily laid down for them”. An appeal to a simple Latin model, therefore, seems a logical explanation for the uniformity of the word-division practice at line ends in Old English. Writing was a hard-earned and valuable skill in the scriptoria: the instruction must have been rigorous, and replicating the Latin prescriptive pattern was not dif ficult. It is also noteworthy that the minute number of “exceptions” occur in only 7 out of 168 examined manuscripts; indeed, in one single manuscript, Ker 334 (Junius XI in the Bodleian Library, The Cædmon manuscript of Anglo-Saxon Biblical poetry), 36.7% of the VCV divisions are (Wetzel 1981: 117–118). Clearly, and reassuringly, there were some “incompetent” and careless scribes whose training was not sufficient to suppress their insecurity about syllabifica tion in speech.
4.2 Orthographic gemination in Old English Another aspect of the Old English orthographic records relevant to the metho dology of reconstructing syllabification is the incidence of what can be labeled
16 See note 13 above on onset-maximal syllabification in Latin across word boundaries. Note that the division of words containing muta cum liquida in Old English manuscripts recorded in Lutz (1986) and discussed in linguistic terms is also in line with the practice recommended by the Roman gram marians. Although for the purposes of reconstructing the history of English the vernacular docu ments are crucial, it has to be acknowledged that at any point in Old English and “even in the century before the Conquest far more manuscripts were written in Latin than in English” (Roberts 2005: 3).
Metrical resolution, spelling, and Old English syllabification
153
‘orthographic’ gemination. The existence of phonological geminates in Old English is part of all descriptions of the Old English consonantal system. The minimal pairs in (10) show that singletons and geminates were contrastive word-medially: (10) Singletons and geminates in Old English: bitela ‘beetle’ bitter ‘bitter’ bite ‘bit, morsel, cut’ bitte ‘bucket’ cyle ‘chill, fever’ cyll(e), cyllan ‘wineskin’ hopian ‘to hope’ hoppian ‘to hop’ manu ‘mane’ mann(a) ‘man’ sæp(e) ‘sap’ sæppe ‘spruce, fir’ With the exception of the approximants /w/ and /j/, all Old English non-initial consonants could appear either as singletons or as geminates. The Old English geminates can be traced back to various sources. Some arose already in ProtoGermanic from assimilation: PrG *wullō > OE wull ‘wool’; PrG *sterron- > OE steorra ‘star’. Geminates due to assimilation could arise after both short and long vowels, as in the verb paradigm of cȳþan ‘to make known’ > past sg. cȳþ-de ~ cȳdde; after syncope of the inflexional vowel in lǣdan ‘to lead’, > 3rd sg. lǣd-þ > lǣdt > lǣt(t) ~ lǣd; mētan ‘to meet’ > 3rd sg. mēt-þ > mēt(t). By far the most important source of consonant gemination in Old English, however, is West Germanic (Consonant) Gemination, which was under way, but not neces sarily completed, by the early fifth century. The gemination is in evidence in the earliest Old English written records, around the first quarter of the eighth century, compare Gothic sibja ‘amity’ to early OE megsibbi ‘affection among relatives’. Since West Germanic (Consonant) Gemination occurred only after short vowels, the most common distribution pattern of geminates was in the environment . The syllabification of true intervocalic geminates is uncontroversial: they are heterosyllabic: /VC1.C1V/. In the verse geminates after short stressed vowels block resolution, e.g., wæs se grimma gæst / Grendel haten (Beo 102), where grimma ‘grim’ has to fill both a strong and a weak metrical position – the a-verse scans w w S / w S. In the course of Old English the salience of the geminatesingleton contrast varied according to the position of the geminate in the word and the length of the vowel to the left. Word-final geminates were particularly vulnerable: ‘bed’ alternates in spelling with in the nom. acc. sg, but in the inflected forms only spellings are attested, similarly ~ ‘bull’, but inflected ; ~ ‘grim’, but inflected
154
Donka Minkova
. Another environment where the contrast was apparently unstable is when a double consonant of any origin was followed by another consonant: eallne ~ ealne, lȳttle ~ lȳtle, æddre ~ ǣdre ‘vein’, blæddre ~ blǣdre ‘bladder’ etc. The vowel length in the latter forms can be questioned, see Campbell (1959: 182–183). There is no metrical test for either the authenticity of a geminate or the length of the vowel of such forms in Old English. Neither /VC1C2V/ nor /VC1C1C2V/ can be resolved, and /VVCCC/ sequences are shortened to /VCCC/ as in the obscured compound gōd-spell > godspell ‘gospel’, but both before and after the shortening the initial syllable is treated as heavy in the meter. The geminate’s position with respect to stress is also of consequence (Luick 1914–1940: §672.2). By the tenth century, geminates became unstable in unstressed medial syllables, as in (11): (11)
Variable consonant-doubling in medial syllables in Old English: gyldene ~ gyldenne ‘golden’ dropude ~ droppode ‘dropped’ atelice ~ atollice ‘horribly’ to brucane ~ brucenne ‘for using’ forene ~ forenne (1×)‘before’ goretende ~ gorettende ‘roving’ candele ~ candelle ‘candle’ ēowere ~ ēowerre ‘your’
The interatonic alternations of are especially common in Old English, but the variability is attested in other forms, too. Such spelling instabi lity indicates blurring of the difference between heterosyllabicity and ambisyl labicity; the etymological length of the consonants was no longer a determining factor for orthographic doubling in positions other than after a primarily stressed vowel.17 This is an initial step in the long-term loss of singleton-geminate steminternal contrast in English: although the full simplification of such geminates is a Middle English development, the instances of sporadic inverse orthographic doubling of consonants in Old English are arguably an indication of perceived similarity between VC.CV and V{C}V. Of special interest in this context is a pattern originally identified by Bül bring (1902: §546) as “Northumbrian geminations” found in The Lindisfarne Gospels (Li), The Rushworth Gospels (Ru), The Durham Ritual (DurRitGl), etc. This orthographic gemination occurs specifically after short vowels, but it cannot be
17 Even in those positions an unsystematic search of the Dictionary of Old English (DOE) (Cameron et al. 2007) reveals some variation: begyten ~ begyttan ‘acquire’ (Nic MsB), hetend ~ hettend ‘enemies, nom.pl.’ (Brb 10), cited in Orton (2000: 37), droppende ~ dropende ‘dropping’; forgytol ~ forgyttol ‘forgetful’, āginað ~ āginnað ‘begin’, and after a long vowel hlūtor ~ hlūttor. Occasional variants such as Brb B, C inwitta ~ Brb A inwidda ‘evil one’ (cited in Orton 2000: 168) also inwitum, inwuda (DOE); bite ~ bide (PsCaJ) may well be the precursors of the first more solid evidence of tapping of the kind attested in ME potage (1230) podech (1528) ~ porage (1533) ‘porridge’ (Minkova 2014: 147).
Metrical resolution, spelling, and Old English syllabification
155
attributed to West Germanic (Consonant) Gemination. It is attested most robustly, but not exclusively, if the consonant is a voiceless stop: (12) Northumbrian geminations: ‘break, part’ (Li, Ru) ‘eat’ (Li) ‘seize’ (Li, DurRitGl)
‘torn’ (Li, HomS 42) ‘speak’ (Li) ‘written’ (Li, DurRitGl)
One interpretation of these spellings (Luick 1914–1940: §§391, 670) is that the doubling of the consonants is intended to mark the shortness of the preceding vowel. Kortlandt (1997) rejects this hypothesis and proposes instead that the spellings are best analyzed as a case of preglottalization – in his account they are simply a continuation of the preglottalized stops which have been reconstructed for Proto-Germanic. Kortlandt references Liberman as the original proponent of the idea that PrG had glottal stops and that the glottalized consonants of some Present-Day English dialects are a replication of an earlier state. He supports that reconstruction, but uses different arguments. According to Kortlandt (2003: 9) in Old Northumbian “the double consonant cannot denote either a preceding short vowel (because the attested form is earlier than the lengthening of short vowels in northern English) or a true geminate (because the short vowel is regularly lengthened at a later stage in these dialects)”. Glottalization, the overlap of glottal and oral closure in the production of a single consonant, is typical word- or syllable-finally: shop, stop me; shot, atlas; shock, chocolate show ‘glottal reinforcement’ of [ˀp], [ˀt], [ˀk] in many types of British English (Cruttenden 2008: 166–167, 180). If the phonetic mechanism of glottalization in the older language is assumed to have been the same as the cur rently observable instances of glottalization, then we can posit a link between orthographic doubling and the doubly associated original singletons in (12) – an intervocalic consonant behaving both like a word-final consonant and an onset represented in the spelling by doubling.18 There is a typological comparison of relevance to the reconstruction of ambisyllabicity. Recall that the overall distribution of phonological geminates in the Old English system favors geminates after short vowels. Similarly, the
18 Kortlandt overlooks the fact that Northumbrian gemination is also found with /m/: cymmes ‘comes’ (Li), summum ‘some, dat. pl.’ (Li). While these will not fit his account of orthographic doubling indicating pre-glottalization of oral stops, it is possible to interpret as perceived syllable-final phonetically lengthened /m/, parallel to the phonetic lengthening observed in Present-Day English. Ladefoged (2005: 71) illustrates the difference between initial and final /m/ in Present-Day English with the word mum, where ‘the final m is much longer than the initial m’.
156
Donka Minkova
Northumbrian orthographic geminations are most common intervocalically after short vowels. Inverse spellings, as in the Northumbian set in (12), limited as they may be, suggest that orthographic doubling could signal insecurity about the exact association of the consonant. Analytically, this corresponds to some accounts of intervocalic consonants following a short vowel in Present-Day English, where “certain single consonants behave like biconsonantal sequen ces” (Burzio 1994: 53), see the reference in (2) in Section 1. Also, as in the case of ~ ‘grim’, but inflected , noted above, or æcer ~ æccer ‘field’, (Old Frisian ekker, Old Saxon akkar), where the singleton is attributed to the uninflected form (Campbell 1959: 168), insecurity could also have been trig gered by paradigmatic alternations. One should also consider the historic connection between doubling of conso nants and the notion that the stressed syllable of a word should be structurally the same as a stressed lexical monosyllable. In Present-Day English this was one of the motivations behind positing ambisyllabicity in the first place (see notes 2, 14). There were no #(C)V# lexical words in Old English. With reference to spelling in Present-Day English: 66% of disyllabic English words with one medial consonant that contain a stressed lax vowel in the first syllable are written with a geminate (e.g., rabbit, grammar), see Treiman, Bowey, and Bourassa (2002), Eddington, Treiman, and Elzinga (2013). The experimental results they cite support the crite ria cited in (1), but they also underline the significance of orthotactic constraints (positional constraints on the sequences of graphemes that are used in words), which include the way in which are syllabified: sequences are unac ceptable word-initially, and therefore are never syllabified *V. C1C1V, thus bitter cannot be bi.tter. The same principle would apply to the way in which Old English scribes employed orthographic geminates for etymological singletons: the doubling of the consonant would guarantee that a short vowel in a stressed sylla ble was closed by a consonant and was therefore orthographically and phonolo gically ‘legal’. On a final note, orthographic gemination in older English has a long and inter esting history. Without rehearsing the voluminous literature on the orthography and phonology of the late twelfth-century autograph manuscript The Ormulum,19 it should be pointed out that the innovative spelling system of that manuscript is fully compatible with a potentially ambisyllabic status of intervocalic singletons. Orm’s system involves doubling of consonants after short vowels except in VCV structures ( ‘god’ nom., but ‘god’s,). Put differently, an unambig uous coda (godd ‘god’, serrfenn ‘serve’) is doubled, while potentially ambisyllabic codas are not doubled.
19 See Murray (1995), Fulk (1996), Anderson and Britton (1999) and references there.
Metrical resolution, spelling, and Old English syllabification
157
5 Summary Syllables and their exact composition cannot be accessed directly. Even in PresentDay English, where scholars can conduct experiments with native speakers, the debate about the principles of syllabification and the methodologies of finding the right solutions continues. This study focuses on two potential paths of inquiry into the history of syllabification in earlier English. The first topic is the onsetmaximal association of singletons between a short stressed vowel and another vowel in the verse. The broader discussion of the special phonological properties of meter in relation to speech led to the proposal that the behavior of VCV sylla bles with respect to resolution in Old English verse reflects only initial, deep-level syllabification. Any further re-syllabification is optional and linked to informal registers, which makes it logical that it would be ignored by the well-trained AngloSaxon scribes. Thus the verse-specific application of the syllabification procedure remains uninformative as to the possible ambisyllabic association of intervocalic consonants in speech. The second line of inquiry was the evidence from spelling collected from worddivision at line-ends and from orthographic gemination. The astonishing regularity of onset-maximal division of strings as at line-ends can be attributed to formal register syllabification, as in the verse, and to the rigorous schooling of the scribes: the simple rule of word-division at line-ends was handed down from the Greeks to the Romans and was easily adopted by the Old English scribes. Another orthographic source, consonant gemination, was found to be strongly suggestive of ambisyllabicity, though admittedly the proposed interpretation which draws on phonetic parallels with Present-Day English must remain conjectural. Throughout the study, I have tried to make inferences on what can be taken as evidence and what can be discarded as deviance in the light of general lin guistic probabilities. A full discussion of the role of ambisyllabic consonants in calibrating the weight of the preceding syllable in earlier English is still needed, and so is a detailed study of the behavior of different consonant types and their dependence of the nature of the flanking vowels. The extent to which Old English spelling reflected categorical distinctions is debatable. Luick (1914–1940: §27) believed that the Old English tradition of faith ful and regulated scribal practices was relaxed in the twelfth century, and thereaf ter man schrieb wie man sprach ‘one wrote as one spoke’, a dictum which Stanley (1988: 328) pronounced completely wrong, insisting instead that man schrieb nie wie man sprach ‘one never wrote as one spoke’. The single conclusion that one can take away from the material in this paper is that in historical reconstruction both points of view have some validity.
158
Donka Minkova
Acknowledgments I am grateful to the SHEL-7 organizers for inviting me to present a talk, of which this paper is a much-revised portion. I am also grateful to an anonymous reviewer and this volume’s editors for catching infelicities and some blunders in the initial submission. The paper has benefitted from research on a larger project on the evidence and interpretation of syllabification in English conducted jointly with Kie Ross Zuraw, Department of Linguistics, UCLA.
References Anderson, John & Charles Jones. 1974. Three theses concerning phonological representations. Journal of Linguistics 10. 1–26. Anderson, John & Derek Britton. 1999. The orthography and phonology of the Ormulum. English Language and Linguistics 3(2). 299–334. Bermúdez-Otero, Ricardo. 2007. Word-final prevocalic consonants in English: Representation vs. derivation. www.bermudez-otero.com/OCP4.pdf Bermúdez-Otero, Ricardo. 2011. Cyclicity. In Marc van Oostendorp, Colin Ewen, Elizabeth Hume & Keren Rice (eds.), The Blackwell companion to phonology, 2019–2048. Chichester: Wiley-Blackwell. Bülbring Karl Daniel. 1902. Altenglisches Elementarbuch. Heidelberg: C. Winter. Burzio, Luigi. 1994. Principles of English stress. Cambridge: Cambridge University Press. Cameron, Angus, Ashley Crandell Amos, Antonette diPaolo Healey et al. (eds.). 2007. Dictionary of Old English: A to G online. http://tapor.library.utoronto.ca/doe/ Campbell, Alistair. 1959. Old English grammar. Oxford & New York: Clarendon Press. Cruttenden, Alan. 2008. Gimson’s Pronunciation of English. London: Hodder Education. Eddington, David, Rebecca Treiman & Dirk Elzinga. 2013. Syllabification of American English: Evidence from a large-scale experiment. Part I. Journal of Quantitative Linguistics 20(1). 45–67. Fulk, Robert D. 1992. A history of Old English meter. Philadelphia: University of Pennsylvania Press. Fulk, Robert D. 1996. Consonant doubling and open-syllable lengthening in the Ormulum. Anglia 114. 481–513. Fulk, Robert D. 1997. Ambisyllabicity in Old English: A contrary view. In Irmengard Rauch & Gerald F. Carr (eds.), Insights in Germanic linguistics II: Classic and contemporary (Trends in Linguistics, Studies and Monographs 94), 29–45. Berlin: Mouton de Gruyter. Fulk, Robert D. 2002. Early Middle English evidence for Old English metrics: Resolution in Poema morale. Journal of Germanic Linguistics 14. 331–355. Fulk, Robert D. 2005. Review of The metre of Old Saxon poetry: The remaking of alliterative tradition, by Seiichi Suzuki (Cambridge: D. S. Brewer [2004]). Journal of Germanic Linguistics 17(2). 149–153. Giegerich, Heinz, 1992. English phonology. Cambridge: Cambridge University Press. Goodwin, William Watson. 1892. A Greek grammar. Boston: Ginn & Co. Gordon, Matthew. 2006. Syllable weight: Phonetics, phonology, typology. New York & London: Routledge. Gussmann, Edmund. 2002. Phonology: Analysis and theory. Cambridge: Cambridge University Press.
Metrical resolution, spelling, and Old English syllabification
159
Gwara, Scott (ed.). 1997. Anglo-Saxon conversations: The colloquies of Aelfric Bata. Woodbridge, Suffolk: Boydell & Brewer. Hale, William Gardner. 1896. Syllabification in Roman speech. Harvard Studies in Classical Philology 7. 249–271 Hammond, Michael. 1999. The phonology of English: A prosodic Optimality-Theoretic approach. Oxford: Oxford University Press. Hayes, Bruce. 1982. Extrametricality and English stress. Linguistic Inquiry 13(2). 227–276. Hayes, Bruce. 1988. Metrics and phonological theory. In Frederick J. Newmeyer (ed.), Linguistics: The Cambridge survey II, 220–250. Cambridge: Cambridge University Press. Hayes, Bruce. 1995. Metrical stress theory: Principles and case studies. Chicago: University of Chicago Press. Hayes, Bruce. 2009a. Faithfulness and componentiality in metrics. In Sharon Inkelas & Kristin Hanson (eds.), The nature of the word: Studies in honor of Paul Kiparsky, 113–148. Cambridge, MA: MIT Press. Hayes, Bruce. 2009b. Syllabification in English. UCLA manuscript. Hogg, Richard M. 1992a. A grammar of Old English, I: Phonology. Oxford: Blackwell. Hogg, Richard M. 1992b. Phonology and morphology. In Richard M. Hogg (ed.), The Cambridge history of the English language. Vol. 1: The beginnings to 1066, 67–168. Cambridge: Cambridge University Press. Hogg, Richard M. & R. D. Fulk. 2011. A grammar of Old English. Vol. 2: Morphology. Oxford: Wiley-Blackwell. Hooper, Joan B. 1972. The syllable in phonological theory. Language 48(3). 525–540. Hutcheson, Bellenden Rand. 1995. Old English poetic metre. Cambridge: D.S. Brewer. Jespersen, Otto. 1909. A Modern English grammar on historical principles, Part 1: Sounds and spellings. Heidelberg: C. Winter. Jones, Charles. 1989. A history of English phonology. London: Longman. Kahn, Daniel. 1976. Syllable-based generalizations in English phonology. New York: Garland. Kiparsky, Paul. 1977. The rhythmic structure of English verse. Linguistic Inquiry 8. 189–247. Kortlandt, Frederik. 1997. How old is the English glottal stop? North-Western European Language Evolution: NOWELE 31–32. 175–179. Kortlandt Frederik. 2003. Glottalization, preaspiration and gemination in English and Scandinavian. Amsterdamer Beiträge zur älteren Germanistik 58. 5–10. Kreidler, Charles. 2004. The pronunciation of English: A coursebook. Oxford & Malden, MA: Blackwell Publishing. Ladefoged, Peter. 2005. Vowels and consonants. 2nd edn. Oxford: Blackwell Publishing. Lass, Roger. 1992. Phonology and morphology. In Norman Blake (ed.), The Cambridge history of the English language. Vol. 2: 1066–1476, 23–155. Cambridge: Cambridge University Press. Liberman, Anatoly. 1990a. The phonetic organization of Early Germanic. American Journal of Germanic Linguistics and Literatures 2. 1–22. Liberman, Anatoly. 1990b. Some debatable questions of Germanic prosody. American Journal of Germanic Linguistics and Literatures 2. 149–158. Luick, Karl. 1914–1940. Historische Grammatik der englischen Sprache. 2 vols. Leipzig: Tauchnitz. Reprint 1964. Lutz, Angelika. 1986. The syllabic basis of word division in Old English manuscripts. English Studies 87. 193–210. Minkova, Donka. 1991. The history of final vowels in English. Berlin: Mouton de Gruyter. Minkova, Donka. 2009. Diagnostics of metricality in Middle English alliterative verse. In Judith Jefferson & Ad Putter (eds.), Approaches to the metres of alliterative verse (Leeds Texts and Monographs New Series 17), 77–115. Leeds: Leeds Studies in English.
160
Donka Minkova
Minkova, Donka. 2012. Syllable weight and the weak-verb paradigms in Old English. In David Denison, Ricardo Bermúdez-Otero, Chris McCully & Emma Moore (eds.), Analysing older English, 402–440. Cambridge: Cambridge University Press. Minkova, Donka. 2014. A historical phonology of English. Edinburgh: Edinburgh University Press. Minkova, Donka & Kie Zuraw. Forthcoming. Ambisyllabicity in English: Present and past. In Merja Kytö & Päivi Pahta (eds.), Handbook of English historical linguistics, Chapter 23. Cambridge: Cambridge University Press. Murray, Robert W. 1995. Orm’s phonological-orthographic interface and quantity in early Middle English. Amsterdamer Beiträge zur älteren Germanistik 42. 125–147. Ní Chiosán, Máire, Pauline Welby & Robert Espesser. 2012. Is the syllabification of Irish a typological exception? An experimental study. Speech Communication 54(1). 68–91. Orton, Peter. 2000. The transmission of Old English poetry (Westfield Publications in Medieval and Renaissance Studies 12). Turnhout, Belgium: Brepols Publishers. Ritt, Nikolaus. 1994. Quantity adjustment: Vowel lengthening and shortening in Early Middle English. Cambridge: Cambridge University Press. Roberts, Jane. 2005. Guide to scripts used in English writing up to 1500. London: The British Library. Rubach, Jerzy. 1996. Shortening and ambisyllabicity in English. Phonology 13. 197–238. Russom, Geoffrey. 1995. Constraints on resolution in Beowulf. In M. J. Toswell (ed.), Prosody and poetics in the early Middle Ages: Essays in honour of C. B. Hieatt, 147–163. Toronto: University of Toronto Press. Russom, Geoffrey. 2002. A bard’s-eye view of the Germanic syllable. The Journal of English and Germanic Philology 101(3). 305–328. Ryan, Kevin M. 2011. Gradient syllable weight and weight universals in quantitative metrics. Phonology 28(3). 413–454. Stanley, Eric G. (ed.). 1972. The owl and the nightingale. Manchester: Manchester University Press. Stanley, Eric G. 1988. Karl Luick’s “Man schrieb wie man sprach” and English historical philology. In Dieter Kastovsky & Gero Bauer (eds.), in collaboration with Jacek Fisiak, Luick revisited: Papers read at the Luick-Symposium at Schloss Liechtenstein, 15–18.9, 1985, 311–333. Tübingen: Narr Verlag. Stockwell, Robert & Donka Minkova. 1997a. Old English metrics and the phonology of resolution. North-Western European Language Evolution: NOWELE 31–32. 389–406. Stockwell, Robert & Donka Minkova. 1997b. Prosody. In Robert Bjork & John D. Niles (eds.), A ‘Beowulf’ handbook, 55–85. Lincoln: University of Nebraska Press. Suzuki, Seiichi. 1994. Breaking, ambisyllabicity, and the sonority hierarchy in Old English. Diachronica 9. 65–93. Suzuki, Seiichi. 1995. Resolution and mora-counting in Old English. American Journal of Germanic Linguistics and Literatures 7(1). 1–29. Treiman, Rebecca, Judith A. Bowey & Derrick Bourassa. 2002. Segmentation of spoken words into syllables by English-speaking children. Journal of Experimental Child Psychology 83. 213–238. Wetzel, Claus-Dieter. 1981. Die Worttrennung am Zeilenende in altenglischen Handschriften. Frankfurt am Main: Peter Lang Verlag.
Anatoly Liberman
The shortest history of vowel lengthening in English Why is it that nothing about historical linguistics is ever settled once and for all? We still argue about the most fundamental issues. (...) Hardly a year and certainly not a decade goes by that there is not a brilliant new interpretation of such well known sound changes as Grimm’s Law in Germanic, the Great Vowel Shift in English, Grassmann’s Law in Sanskrit and Greek, the Old High German Consonant Shift. And, inevitably, what was brilliant when it first appeared looks faded and unstylish ten years down the road. (King 1988: 22)
1 Introduction This paper may as well have had the subtitle “50 ans après”. In 1962, I, a fearful extramural graduate student at St. Petersburg University, met the great Professor M. I. Steblin-Kamenskij, who agreed to direct my dissertation on some problem of English historical phonology. “Draw up a list of the major changes in Old and Middle English”, he said, “and we will choose something”. I did as I was bid, and we decided that open syllable lengthening would be a good subject. In 1965 I defended my Ph.D. dissertation and could not imagine that in my sunset years I would still be interested in a topic chosen so long ago and almost at random. At that time, apart from the perennial classics, there was a single article (Dobson 1962) on open syllable lengthening (see Liberman 1992b). Now this topic has become the mainstay of a busy industry and is recognized by the initialism OSL. Since the mid-sixties I have published articles and books on every aspect of length and accentuation in Germanic and in what follows will give references to some of them, to avoid repeating what I have already said in print. However, it may be useful to emphasize a few principles that underlie the exposition below. 1) The recent approaches to historical phonology tend to substitute formalization for cognition. This development replaces linguistics with algebra and can hardly be equated with progress. 2) Not only formalization but also reference to teleology poses a danger to the science of reconstruction. Today it is widely believed that the motor of sound change can be discovered in a striving for certain preferred structures or complying with certain general laws. Thus, the result of the change acquires the status of its cause. Like every version of circular reasoning, it should be abandoned. A sound change has been explained only if we discover why it occurred when it did. No general principle can perform this task.
162
Anatoly Liberman
3) An opinion is often heard that some sound changes are “non-NeoGrammarian”. It refuses to accept the fact that every systemic sound change has a beginning and an end and occasionally meets with resistance. The cause of such a change is usually another change. The process begins once “the command” is given and stops only after the impulse (command) loses its force. That is why some changes continue for centuries and are implemented even after the speakers resettle far from their original home. Historical linguists have to admit that every exception should be accounted for and that every change is “Neo-Grammarian”. 4) As regards vowel length, certain popular approaches also seem unproductive. Since Trubetzkoy’s days it has been known that length is a metaphor. In Germanic, we deal with mora-counting (the old period), the correlation of syllable cut (Modern West Germanic and Danish), and the correlation of syllable quantity (/V:C/ versus /VC:/; all the modern Scandinavian languages except Danish). Today no Germanic Standard language counts moras. Consequently, encoding modern systems in terms of moras disguises the difference between the structure of Germanic before and after the period of apocope. 5) The difference between the prosody of old and modern Germanic languages explains why the quantitative changes in their history had different causes. There is no need to look for a unifying factor in this area. Spontan eous lengthening, like the one posited for pre-historic monosyllables of Old English, is unthinkable. Besides, in studying those changes, attention should often be paid to the distinctive features, rather than phonetic properties, of postvocalic consonants.
2 Some basic concepts 2.1 Length and mora counting Trubetzkoy must have been right: length is always a metaphor. At least in Germanic it is. Vowels cannot be short and long as they are open or close and front or back. Length is not a distinctive phonematic. In dealing with Germanic quantity, we invariably deal with prosody. It follows that lengthening is a misnomer; the term conceals an upheaval in the word’s “suprasegmental” structure. Like all the Old Indo-European languages, Old Germanic was a mora-counting language. The concept of the mora has been much abused in contemporary scholarship. In Classical Greek and Latin, a bimoric vowel consisted of two semi-independent parts. It follows even from the school example of Latin legĕre versus delēre as légere versus deléere. Hence the rule: stress falls on the third mora from the end. The rule
The shortest history of vowel lengthening in English
163
of Greek stress has also been formulated in terms of moras. Therefore, the mora is not only a unit of length but also the locus of accent. Outside this double function it does not exist. Modern Germanic languages (except some Scandinavian dialects) are not mora counting. (Contrary to Trubetzkoy, whose opinion I once shared, stød does not make Danish a mora-counting language, but in the past it probably served as a signal of bimoric complexes.) An interplay of stressed and unstressed moras (Akzentumsprung) under the influence of sentence stress is the cause of multiple monophthongizations and diphthongizations in Old Germanic. For instance, Proto-English *ai with sentence stress on a resulted in “smoothing” (*ai > ā), while the phonologization of stress on the second mora turned Germanic *ō into Old High German (southern) ua, possibly pronounced like oi in Modern French moi [mwa]. Likewise, the later split of ME lēosan into leese and lose (the whole of the strong second class behaved like lēosan) is thinkable only in terms of shifting sentence stress from mora to mora (léosan versus leósan).
2.2 Positional and compensatory change In older works, one can read that certain changes are the product of the position in which a sound happens to occur (see discussion in Liberman 1992a: 172–174). According to most manuals, in the Old English group ld, l was velar, and so resulted in breaking *a > ea, etc. An important link is missing in this reasoning. (I will ignore its circularity: the velar pronunciation of l is usually deduced from breaking.) Apparently, l had always been velar. Why then did it suddenly cause breaking? Or did it exercise its influence gradually until, to use Charles Hockett’s phrase, the harm was done? Position in such reconstructions emerges as a force rather than as an arena, a battlefield. Compensation is not a mechanical process reducible to arithmetic. Rather, it is a weapon devised to offset change. The word loses something, but the defensive mechanism produces the illusion of stability. Later such makeshift devices may disappear: for example, lengthened vowels may undergo shortening, which signifies the ultimate triumph of change, mastery of the historical arena.
2.3 Accent, stress, and syllabification Older scholars often connected quantitative changes with accents. Here the main figure was Sievers, who discovered the syllable cut (Silbenschnitt). Unfortunately, he called loose, or free, contact and close, or tight, contact schwach
164
Anatoly Liberman
geschnittener Akzent (gravis) and starkgeschnittener Akzent (akut) respectively. Sievers’s approach became dogma immediately after the appearance of Grundzüge der Lautphysiologie, as his Grundzüge der Phonetik was first called. Complications arose when Luick and Morsbach applied Sievers’s idea to history (for details see Liberman 1979 and 1992b: 175). Following Sievers, Luick taught that lengthening could occur only under gravis. Morsbach misunderstood Sievers and believed that lengthening caused the change of acute to gravis. Luick pointed to his error. However, both were mistaken (or at least neither could prove his point), for how did they know that open syllables invariably had gravis? Only from the lengthening. And why did lengthening occur in them? Because they had gravis. It is almost unbelievable that in the twenty-first edition of Paul’s Middle High German Grammar (Paul 1975: 52), Sievers’s formulation was still repeated almost verbatim. Luick, who may have been aware of this circularity, thought that in Old Germanic gravis was the only accent. Another popular idea was the ever-increasing strong stress in Germanic. Endings were said to disappear, vowels to diphthongize, monophthongize, shorten, and lengthen under something called dynamic stress, whose presence was inferred from those changes. Becker (2002: 35) begins his otherwise informative paper by stating that open syllable lengthening happened bei starker Betonung. Sievers’s accents found a worthy place in Trubetzkoy’s conception of quantity, but, predictably, only in relation to synchrony. There they stayed and were subjected to detailed investigation by instrumental phoneticians (who showed that the picture was far less clear than Sievers, Trubetzkoy, and later Martinet thought) until Theo Vennemann rediscovered them with such success that his predecessors faded from view and he acquired the status of the trailblazer who once again showed the way, to quote his most ardent follower. With regard to English, Ormulum was brought into prominence to support Vennemann’s idea (see especially Murray 2000), but Ormulum is irrelevant in this context, regardless of how we interpret Orm’s spelling reform. I subjected Orm’s geminates to a detailed analysis in my earlier publications. However, since they are in Russian, no one outside the former Soviet Union and Donka Minkova has read them. Despite the prevailing negative attitude toward Trautmann’s conclusions (an attitude everybody seems to share today), I think Trautmann was right, even though no single formula can explain all the idiosyncrasies of Orm’s system. From the point of view of common sense, Trautmann’s suggestion is the most plausible: Orm must have written double letters when he heard long consonants, and he lived in an area in which the Scandinavian influence was strong, which explains the abundance of geminates. All the rest (vowel length, diacritics, apparent exceptions, etc.) is a matter of interpretation. Orm was not a linguist but a recorder of his pronunciation, though his practice of scanning every syllable
The shortest history of vowel lengthening in English
165
(as Trautmann was probably the first to point out) resulted in all of them being stressed in his recitation. As far as Orm’s so-called open syllables are concerned, vowels in them remained short (see especially Fulk 1996). The correlation of syllable cut means only that in some words of the Modern West Germanic languages the syllable boundary is allowed after certain vowels, while after others it is not. Thus, Engl. /leitǝ(r)/ (later) is /lei-tǝ(r)/ (two syllables and a point of syllable division), whereas /letǝ(r)/ (letter) is disyllabic but indivis ible (two partly overlapping waves). Monosyllables defy comparison. We observe that pick and peak are different words, but whether the difference depends on vowel length, the quality of the vowels, the type of contact between the vowel and the consonant, or on the final consonants cannot be decided. It is even debat able that they constitute a minimal pair. Phoneticians will tell us where the locus of the contrast appears to be the strongest, but we need procedures rather than indications of the most sophisticated apparatuses. Only if we succeed in turning our forms into disyllables and produce picking and peaking or picker and peaker and assume (for proof is unavailable) that pick ~ picking (picker) and peak ~ peaking (peaker) share their prosody pairwise, can we state that pick has close contact (picking and picker are indivisible), while the contact in peak (cf. pea-ker, pea-king) is loose. Then we may perhaps be allowed to say that brick, insofar as it rhymes with pick, also has close contact, in contradistinction to geek, which rhymes with peak. No phonology is possible without some prephonological procedures. Dividing a word into syllables is among them. Two hundred years of attempts to determine the principles of syllable division in Greek, Latin, Gothic, and other languages have made it clear that we lack the evidence to arrive at indisputable results. Even in modern languages with complicated prosody, agreement sometimes evades us, but at least the later – letter opposition, the backbone of the correlation of syllable cut, is beyond doubt. Given our ignorance of syllable division in Middle English (unless we accept Kuryłowicz’s rule: see below), all talk about the Silbenschnitt in the thirteenth century and earlier should be abandoned as futile. We may also ask: How did the posited correlation emerge in Middle English? If it existed in Old English or even earlier, its presence will remain a riddle, for it has no roots in reconstructed Proto-Indo-European or Proto-Germanic. By contrast, the development of the correlation can be traced to apocope and quantitative changes. Murray (2002: 110) wrote: “I assume that syllable cut had already been phonologized in the mid-eleventh century. This dating is important because it strongly suggests, as proposed by Vennemann 2000, that the phonologization of syllable cut is inextricably linked to the major quantity changes of the period including those traditionally known as Closed Syllable Shortening, Homorganic Cluster Lengthening and Open Syllable Lengthening”.
166
Anatoly Liberman
It is precisely this chronology that I find unacceptable, for all of the reasons just given and more. Another pseudocause of quantitative change is resyllabification. It would occasionally be declared that in a certain century German or English underwent a shift of syllable boundaries, with lengthening or shortening being the result. Yet, resyllabification can only be the consequence but never the cause of other prosodic changes. 2.4 Living speech and diachrony It is now common practice to refer in works on historical phonetics to William Labov and John Ohala. Since both have studied living speech, they, not unexpectedly, recorded variation and gradual change, for a close look at synchrony cannot yield any other results. Those who turn for support to Labov’s and Ohala’s works face a dilemma: on the one hand, all is fluid and gradual; on the other, the phonemic principle requires that change occur per salto. No one doubts that sound change spreads from a certain epicenter. Phonemes clash and merge for the most various reasons, and demographic perturbations take their toll on people’s speech. Those who use Labov’s and Ohala’s results and remain true to the phonemic principle hide behind the idea of a stable underlying form versus its surface realizations. This is the same idea that allowed the descriptivists of Bloomfield’s era to acknowledge the phoneme and resort to allophonic change preparing for the phonemic “leap”. Their phonology did not differ from the teachings of the previous generations (“sounds change gradually, and at a certain moment the threshold is crossed”). To get to the root of the matter, one needs a model. Neither Labov nor Ohala has offered a new theory of sound change; they merely trace its bewildering and often unpredictable routes. Quite untenable is the idea of “sound change” versus “prosodic change” (as advocated in Page 2007: 340). Both are instantaneous in theory and gradual in practice. This paradox underlies them in equal measure. If we compare the implementation of open syllable lengthening and the initial voicing of fricatives in English dialects, we will see that the problems confronting language historians are no less paradoxical. 2.5 Theory and practice Those who have followed the development of phonetic sciences since roughly the sixties of the past century witnessed a rapid proliferation of “schools”: generative
The shortest history of vowel lengthening in English
167
phonology (with its “classical” and “post-classical” periods), natural phonology, autosegmental phonology, and optimality theory (to name a few). From Murray (2000) alone, I have culled the following labels: moraic phonology (this trend has especially many supporters), dependency phonology, and nuclear phonology. However different they may be, they share the conviction that the clue to language study lies in formalization: all the facts are known and have been known for more than a century. Our task allegedly consists in representing them as cleverly as possible. For veritable feasts of formalization see Auer (1989) and especially Kaminashi (1993). From a huge collection of relevant statements at my disposal, I will quote only two. First, Lahiri and Dresher (1999) believe that the raison d’être of the main quantitative changes in Middle English was a striving to maintain and maximize the Germanic foot (to put it differently: the changes will be explained if we represent them in terms of metrical phonology). The striving is taken for granted. Second, Bermúdez-Otero and McCully (1997) wrote an unenthusiastic review of Ritt (1994). My own appraisal (Liberman 1994) was much harsher, but we criticized different things. Here is the two authors’ final statement: “Although the volume under review focusses on an area of considerable topical interest, probabilistic laws such as Ritt envisages are not viable tools for historical lin guists. They do not advance our understanding of optimalization in sound change. It is consequently ironic that his book should have seen the light shortly after Prince and Smolensky’s 1993 theoretical breakthrough” (Bermúdez-Otero and McCully 1997: 624–625). Prince and Smolensky (1993), still a manuscript at that time, but widely circulated, was the first book-length introduction to optimality theory. The reviewers felt confident that if Ritt had had a chance to read the manuscript, he would have produced a different book. In this review optimaliz ation is not questioned, but the basis of rule writing is different from Lahiri and Dresher’s. Ritt continued to defend his approach in later years and probably familiarized himself with optimality theory; yet the encounter with the “breakthrough” did not affect his thinking in the least. Never before in the history of linguistics have its practitioners been so divided and so pleased to belong to different schools as they are today. There are “trailblazers” who have adherents, but no consensus is visible among them. Sievers, Luick, Trubetzkoy, and a few others remain our teachers, whereas post-World War II schools come and go, though some last longer than the others. Consider the epigraph to this paper: it does not seem that historical linguists of English have produced much work of permanent value, and in this respect we are like the rest of modernism: noisy, often ingenious, even talented, and usually transitory.
168
Anatoly Liberman
3 Lengthening before consonant groups My contribution to vowel lengthening before consonant groups (Liberman 1992a) contained a critical survey of the literature (at that time I missed only Hubmayer [1983], who had also offered a condensed summary of earlier works, and Markus [1987: 278 and 283]) and addressed the questions briefly touched on above: lengthening and mora counting, lengthening and accents (a central problem), lengthening and context, a comparison of the change in Old English with its obvious and seeming analogs elsewhere in Germanic and in Balto-Slavic, and the connection between lengthening and epenthetic vowels (svarabhakti). I am not aware of any critique of my hypothesis and still believe that Luick (1914–1940: 245–246), who after much vacillation ascribed lengthening before consonant groups to prehistoric apocope and syncope, was right, but his circumflexes are fictitious and should be disregarded. In mora-counting languages, systemic lengthenings are always the product of compensation. It may not be inappropriate to quote Skeat, who, most probably, erred in his rejection of the traditional approach, but an opinion of such a scholar is always useful to know. His remarks are hidden in a report (Anonymous, 1899): The phrase “compensatory vowel-lengthening” is unsatisfactory and misleading. When we read that “short ο is lengthened in Attic as compensation for the loss of a nasal in the Attic τόυς for Cretan τούς”, it seems to be implied that the short vowel ο was altered to ου because the nasal had been lost, in order that substantial justice might be done. But phonetic laws care nothing for substantial justice, and a lost nasal would arouse no sentimental regret nor obtain any hearing if it clamoured for redress.
The horse should be put before the cart. The rule should be thus expressed: “When a short vowel occurs before a combination of consonants, it is sometimes lengthened, chiefly when the former of the consonants happens to be a liquid. After the vowel has become long or has passed into a diphthong, the former of the consonants frequently (but not invariably) disappears”. Take, by way of example, the Old Mercian form ald, with short a, as in G[erman] alt. It is now old with long o. (...) The English word balsam furnishes a good example of vowel lengthen ing unaccompanied by loss.
Evidently, Skeat considered the phonetic environment (context) to be the main motor of change. Apocope and syncope must have left a compensatory trace in postvocalic consonants, which emerged as distinctively velar or palatalized (Liberman 1992a: 185–194). Now I know more about West Germanic palatalization thanks to my
The shortest history of vowel lengthening in English
169
work on umlaut (Liberman 2007) and to van den Hoek’s dissertation (2010). The reality of the opposition palatalized : velar in Germanic, with markedness depending on the situation, can no longer be called into question. The idea of two OE l’s and r’s (not just “sounds”, but phonemes) is not new; see, for example, Reszkiewicz (1953). Every attempt to understand the nature of Old English breaking and velar umlaut begins and ends with defining the char acteristics of the postvocalic consonants or the vowels in the next syllable. The agreement is confined to the triviality that l and r, for example, transferred their velar quality to the preceding monophthongs. But the phonemic aspect of the transfer remains undisclosed. The solutions tend to follow the familiar path (the context triggers the change): lengthening is ascribed to the environment, and we do not learn why the old environment became active when it did. Hubmayer (1983), Markus (1987), and Minkova and Stockwell (1992) remained in this familiar mold. Following in Reszkiewicz’s footsteps, Fisiak (1967) examined the graphemic clusters and in Old English and suggested that the letter designated velarity in l and r. Scandinavian linguists reconstructed two sets of resonants as far back as the nineteenth century (for details see Liberman 1992a: 188–192). It appears that the correlation of palatalization developed gradually, from several sources (prehistoric syncope/apocope and West Germanic gemination among them), which could be expected, because palatalization is always a secondary (acquired) feature. In the Germanic area, palatalization was unstable and did not make it into any Standard language (though in dialects it occurs from northern Scandinavia to Alemannic), and for this reason it has attracted insufficient attention outside dialectology. The collapse of palatalization in Germanic (its causes cannot be explained here) had predictable consequences: the feature (a distinctive feature, not just an articulation) was indeed transferred to the preceding vowels. I find some parallel ism to this process in Irish (see Ó Dochartaigh 1981: 226). In Old English, umlauts, breaking, and lengthening before consonant groups owe their origin to this process. An often-discussed matter is the exact composition of the lengthening groups. In recent years, Wełna (1998; 1999; 2000) paid special attention to it, but, viewed against a broader background, this detail is of relatively little importance. Not only the onslaught but also the retreat of palatalization took a long time. This explains the difficulties connected with the dating of lengthening before consonant groups and the great disparity between languages (English: opinions go all the way from the proto-period to the tenth century; German: the Middle period). Hart (1906: 120) wrote: This question of vowel-lengthening bristles with difficulties as we approach it from the M[oder]n E[nglish] side, difficulties which can scarcely be brushed aside with an assumption
170
Anatoly Liberman
that lengthened vowels have been re-shortened. (...) I am unable to convince myself that vowel-lengthening was as early as the beginning of the ninth century, or that it was universal. To be safe, we ought to regard lengthening before consonants, like palatalization of the g, as a tendency rather than as a rule. It works here and there, but not everywhere; its operation can not be gauged accurately by the century, can not even be fixed rigorously to any one dialect. That particular Mercian-Midland dialect from which has come standard Mn. E., although lavish with palatalization, has certainly been chary of lengthening before consonant groups.
Naturally: a dialect “lavish with palatalization” was “chary of lengthening before consonant groups”. The same correlation exists between palatalization and i-umlaut in some Dutch dialects: either the one or the other. But Hart had no need to refer to a tendency. The lengthening rule did not work “here and there”; it worked everywhere, depending on the surrender of distinctive palatalization : velarization. In the recent literature, I encountered only one attempt to refer lengthening before consonant groups to compensation. “Surprising as these lengthenings may seem in view of the quantity facts, lengthening by compensation appears to be the most attractive solution, at least within a moraic analysis. That is, for some reason the homorganic consonants would be syllabified as onset to the second syllable (or as extrametrical if word-final) and the vacated mora would be filled by a vowel” (Riad 1995: 162, note 5). This explanation looks like an exercise in paper phonetics. Two considerations should be emphasized here. 1) The lengthening group did not have to be homorganic, and 2) the plosive following the resonant did not have to be voiced (in some cases the second consonant was not even a plosive). Old Icelandic makes both such conclusions indubitable. The lengthening groups there included lf, lp, lk, more rarely lm and lg, and later ng and nk, ld, nd, and rd are conspicuous by their absence. All attempts to derive lengthening from the fact that ld, nd, rd, and their likes were homorganic and ended in a voiced plosive are doomed to failure. An important question is why i and u, when lengthened before consonant groups, yielded ī and ū, rather than close ē and ō, as happened in Middle English, and why the new long ī and ū retained their length when shortening before consonant groups set in (hence Modern English bind and bound versus bend and band). It will be more convenient to deal with this question in the next section.
4 Vowel lengthening in open syllables (so called) This change has in recent years attracted more attention than perhaps any other in the history of English sounds. The current literature on open syllables is so
The shortest history of vowel lengthening in English
171
vast that a good deal of valuable early research has been forgotten. Today the authors of the most visible publications often give an unwelcome twist to truth. Thus, Theo Vennemann emerged as the discoverer of the Silbenschnitt (see above). Donka Minkova (as in Minkova 1982) could hardly suspect that Nikolaus Ritt, Peter Auer, and Roger Lass would ascribe to her Gregor Sarrazin’s central thesis. She, of course, merely referred to Sarrazin. The groundbreaking article on the nature of so-called open syllable lengthening appeared in the middle of the nineteenth century (Nerger 1846). Nerger anticipated and answered nearly all the controversial questions that still occupy us. He dealt with Low German and suggested that lengthening is connected with the weakening of endings. The situation in Low German and English is similar. In 1993, Kim offered the same idea with regard to Middle English as an important novelty. Sarrazin (1898) reasoned like Nerger, but, carried away by Streitberg’s explanation of the Dehnstufe, overstated his case, for, outside English, vowels lengthened in words that preserved their reduced endings. Dobson (1962) reformulated Sarrazin in terms of phonemes and allophones. Below, I will touch only on a few pivotal problems. Without solving them no progress in this area is possible.
4.1 Lengthening and syllabification The name of the change under consideration presupposes that ME mane ( [ɛɪ] [i:] [oə~uə~u:] [ɔu] [o:] [ɪu~ɘu] [aə~a:] [ɑɪ~ɒɪ]
Vowel class
SE Shropshire
Severn type
Majority
green beat name rain stone old grow Goose Mouth Price
[i:] > [ɪi] [e:] [e:] > [ɛɪ] [ɛɪ~e:] [o:] > [ʌu] [ʌu~ɔʊ] [o:] [u:] > [ɘu] [əu] [əɪ~aɪ]
[e:~i:] [e:] [ɛə] [ɛɪ~aɪ] [oə~ o:] [o:~ɔʊ] [ɔʊ] [ɪu] [ɛʊ] [aɪ]
[ɛɪ] [e: ~ i:] [ɛə] [ɛɪ] [o:] [ɔʊ] [o:] [ɪu] [aʊ] [ɑɪ~ɒɪ]
194
Paul A. Johnston, Jr.
informant’s minority forms, one wonders whether a more typical Birmingham pattern was more like the Black Country one, which matches the “majority rule solution” in this respect (and can be heard today; see below). Otherwise, the merger could have come in through Standard influence, if it is not a hyperdialectism. In the back vowel series, Wells’s GOAT class is arranged into three subclasses, since ME /ɔu/ underwent a split in West Midland dialects, monophthongizing early on to [ɔ:], but only in final position. Often, but not in the Second Vowel Shift districts, this leads to a merger with the vowel in words like stone and goat. At the same time, an epenthetic /u/ crept in between the main vowel and the /l/ in words like old, cold and so on, as is usual in the North. This split pattern /ɔu/ char acterizes South Lancastrian (Orton and Halliday 1962) but, without total merger of stone and grow, is also found in West Central Midland dialects. Selly Oak fits best with Eastern Warwickshire here, and except for the variable ingliding diphthongs for stone, with the majority rule outcomes, too. The Black Country is also split, with Staffordshire-like forms for stone, but both [o:] and Severn Valley-like [ɔu] for old/grow. The remaining three nuclei at Selly Oak fit in well with what the majority rule principle would predict, and again, the Black Country is more Staffordshire in type. While our analysis does not totally prove that Brummie results from an eighteenth-century koine, these data, combined with what we know about the demographics and principles of koineization, strongly suggest it did, and mark the area as, in a sense, still transitional between the Severn Valley and Shropshire dialects on one hand and the Staffordshire and East Warwickshire ones on the other.
4 The Industrial Revolution and the second koineization 4.1 Birmingham in the nineteenth century If the eighteenth century marked the point when Birmingham turned into a major English city in England, the nineteenth carried its growth and influence even further. A number of factors made cities grow in the early part of the century; the Industrial Revolution became general, with cheap, coal-driven steam energy (fostered by the engineering advances made in the West Midlands) powering all kinds of machines and engines, and creating a demand for labor throughout. The process of enclosure forced people out of the countryside, and many poured into the growing cities, coming in such numbers that they swamped the labor market and led to slum growth taking hold in some neighborhoods in all conurbations. Birmingham itself grew nearly tenfold over the course of the century, and became England’s second city by 1880, a position it has never relinquished. Its industries
Vowel system restructuring in the West Midlands of England
195
diversified further, adding electrical engineering, carriage, train, bicycle and, eventually, car manufacturing to its industrial repertoire. By the end of the nineteenth century, with so many diverse influences affecting them, new urban varieties, including that of Birmingham, underwent another process of mixture, changing noticeably from what they had been before. In London, the dialect immortalized by Dickens’s Sam Weller metamorphosed into Shaw’s modern-sounding Cockney (Ellis 1869[1889]: 226–236). In addition, the expansion of schooling gradually introduced Standard English and the new RP accent to a wider public, even though only a minority spoke it on a daily basis; nevertheless, its prestige and impact led to the influence from social as well as geographical dialects on the development of urban vernaculars, and continua between vernacular and Standard/RP began to form. Still, the details of the new pronunciation systems remained in part local, and vernacular-to-vernacular contact, especially that involving London English, stayed important.
4.2 The second koineization and the “fine-tuning” The informants in Ellis (1968[1889]) were already absorbing this second wave of koineization on top of the first, a process that even the older Selly Oak informant was not totally immune to. It is incipient in the forms that Ellis’s younger speakers begin to use, shown in Table 4 above after the “>” signs, and is carried ever further in the data of The survey of English dialects (SED) (Orton and Halliday 1962; Orton and Barry 1969; Orton and Tilling 1969), elicited from speakers born in the 1880s and 1890s. Though there is a lot of variability in the process, the ultimate result of this round of koineization is the drawing of the lexical incidence of the Birmingham bimoric vowel system closer to that of London by implementing processes like the three Mid Vowel Mergers, and then, fine tuning the results to make the phonetic values closer still. The name=rain merger is the first one to be affected here, occurring by the substitution of upgliding [ɛɪ] realizations for the earlier ingliding diphthongs in name. This is the only one of the Mid Vowel Mergers to go to completion in Ellis’s Birmingham data, but it becomes increasingly common throughout the area by the time the SED informants were born, toward the end of the nineteenth century. Table 5 shows this merger to be solid in six out of the eight SED survey points around Birmingham, and variably present in the other two – and bear in mind that these are country places, though Himley and Romsley could be considered Black Country localities and Nether Whitacre now has a Birmingham postcode, so Birmingham is presumably further along in this merger.
196
Paul A. Johnston, Jr.
Table 5: Birmingham area reflexes of selected vowel classes, 1950s (Source: Orton and) Vowel class
Nether Whitacre WA 1
Lapley ST 8
Himley ST 11
Hockley Heath WA 2
green beat name rain stone old grow Goose Mouth Price
[i:~ei:] [i:~ei:] [ɛ,ɪ] [ɛ,ɪ] [o,ʊ] [ɔ,ʊ] [o,ʊ] [u:] [aʊ] [ɒɪ]
[i:] [i:] [ɛɪ] [ɛɪ] [oʊ] [oʊ] [oʊ] [u:] [aʊ] [ɒɪ]
[i:] [i:] [ɛɪ] [ɛɪ] [oʊ] [oʊ] [oʊ] [u:] [æʊ] [a:ɪ]
[i:] [i:~ɛɪ] [eɪ~ɛɪ] [ɛɪ~æɪ] [ɔʊ] [ɔʊ] [ɔʊ~ɑʊ] [u:] [ɛʊ] [ɔɪ]
Vowel class
Hilton SA 8
Kinlet SA 11
Romsley WO 1
Hartlebury WO 2
green beat name rain stone old grow Goose Mouth Price
[i:] [i:~e:] [e:] [e:~e:ɪ~ɛɪ] [o:] [aʊ] [o:] [u:] [aʊ] [ɒɪ]
[i:] [i:~e:] [e:] [e:] [o:] [ɒʊ] [o:] [u:] [au:] [ai:]
[i:] [i:~ɛɪ] [ɛɪ] [ɛɪ] [oʊ~ɒʊ] [oʊ~ɒʊ] [oʊ~ɒʊ] [u:] [aʊ] [ɒɪ]
[i:] [i:] [e:ɪ] [e:ɪ] [oʊ] [oʊ] [oʊ] [u:] [ɛʊ] [əɪ]
We can show from an East Midland dialect area what may have driven this merger; Ellis (1968[1889]: 213–220) describes the dialect of Northamptonshire, an area undergoing similar change at that time. In his Variety 16iii, Lower Benefield has the name=rain merger, but Hannington and Harrington do not, and instead have an ingliding diphthong for name but a rising one for rain, like older West Midland dialects. Lower Benefield is right next to Oundle, which had a railway station, while the other two towns were not near train lines (Butler 2006; Collins Barthol omew Map Archive 2013). This suggests that the existence of a railway and easy access to travel is a factor in spreading this feature, presumably from London, since this merger is old there, and the SED majority forms for these classes are identical (Orton and Tilling 1969). However, this is where the Black Country gets off the koineization train, so to speak. The broadest vernacular speakers there not only resist the mergers in the nineteenth century, but the cities, which develop the so-called “Yam-Yam” dialect, still do not have them consistently. Name and rain still can contrast,
Vowel system restructuring in the West Midlands of England
197
with [æ:] found in some name items (and rain can equal beat, as traditionally in Staffordshire), though both Mathisen (1999: 109) and Clark and Asprey (2013: 48–49) report wide upgliding diphthongs in a unified FACE class, as in Birmingham, as the usual outcome. The GOAT class is parallel to this; upgliding diphthongs showing merger prevail, but froze has a triphthong that could have come from an earlier ingliding diphthong (Clark and Asprey 2013: 61). Green and beat are separate, although beat may have an upgliding diphthong (Mathisen 1999: 109; Clark and Asprey 2013: 44–45) and the SED material from Himley (ST11) or Romsley (WO1) distinguishes these subclasses only variably. There does seem to be a stronger influence from conservative Shropshire next door, which one would expect, given its proximity, as well as from Northern Staffordshire. But a stronger impetus towards keeping these forms, as they are stereotypes (Clark and Asprey 2013: 17–22), must be the strong local identity of this district, which takes pride in being separate from Birmingham, despite both areas being in the West Midlands conurbation. The formation of unified FLEECE and GOAT classes is still very much in progress in the SED data, though some dialects have completed the requisite mergers. The prevalence of intermediate forms like [oʊ] and [ei:] suggests that merger by approximation may have happened in some cases; these data, however, may be a trifle on the conservative side, and the diphthongs in Birmingham are wider, and more like London forms (see Table 1). The London and Birmingham variants of PRICE already overlapped and have not changed much, though [ɑɪ] predominates in the former, and [ɔɪ] in the latter. The vowel in MOUTH, however, did undergo change during this time. Originally, Birmingham had [aʊ~æʊ] with even higher first elements found to the south and east, connecting with a large area in the Midlands and South that originally included London.7 London, Birmingham, and the Black Country all tend to turn this into an ingliding diphthong [æə] or a monophthong [a:], developing in parallel, but probably with London in the lead.8
7 It is likely that a good deal of this area already had fronted V2s in this diphthong even at the time of SED. They seem to be the rule now in much of Midland and South Central England. I have personally heard MOUTH as [ɛɪ] in Northampton and North Buckinghamshire as early as the 1970s. 8 A nucleus that seemingly also followed London’s lead, if a little earlier, is the [ɑ:] of START. This is already in the SED data for Nether Whitacre, if not in Ellis, where [a:] predominates. It also appears in the Black Country, and in the non-rhotic parts of Warwickshire and Worcestershire when rhoticity disappears. Staffordshire appears to retain fronter nuclei, at least until the early twentieth century.
198
Paul A. Johnston, Jr.
5 Conclusion Brummie seems to have arisen from a three-stage process of dialect mixture, with little influence from London in the early stages, as the contact due to population influx came overwhelmingly from the immediate surrounding area in the late seventeenth and eighteenth centuries. Since the city was in a transition zone anyway, koineization due to the input coming from both West Midland and Southwest Midland vernaculars took place, and the output, partially visible in the forms of Ellis’s (1968[1889]) Selly Oak informant, can be derived via a “ majority rules” paradigm, just as settlers’ dialects overseas can (Kerswill and Williams 2000: 85; Trudgill 2004). Connections with London via road, canal, and, finally, railway routes improve during the time of the Britain-wide Industrial Revolution, during the nineteenth century, when the population increased tenfold. At this point, the Mid Vowel Mergers reshape the lexical incidence of the original long vowel and diphthong system, with the name/rain merger leading, followed by that of stone/old/grow, and finally, green/beat. London influence is implicated in the early implementation of the merger, as it seems to be accelerated by the pres ence of a railway connection, but merger by approximation may have played a role, as well. This stage does not affect the Black Country strongly, though that area grows as Birmingham does, and probably undergoes the first round of koineization, if with a bigger influence of Shropshire and more northerly dialects. Finally, developments in the early twentieth century “fine tune” the system, drawing the phonetic values of several bimoric vowel classes closer to London norms, or, at least, making them change in parallel with London English, resulting in the forms described in Table 1, and most common at present. As this work shows, it is essential for historical linguists to correlate linguistic changes with the history of the locality studied and, particularly, any population movements that might have occurred, before assuming that a given variety today underwent a straight-line development from Old or Middle English times. This is certainly true with urban varieties, since cities take in more people over the decades. But there are other factors that might produce migration even in medieval times, such as wars, epidemics, or chronic poverty. We have no problem in postulating the great effects of Viking or Norman invasions in the 800s or 1100s in reshaping the dialect map of England, even if the results are attested only later; we should bear that in mind for all periods. Punctuated equilibrium characterizes linguistic changes in general better than steady-state change, it seems.
Vowel system restructuring in the West Midlands of England
199
References Butler, Peter E. 2006. A history of the railways of Northamptonshire. Kettering, Northamptonshire: Silver Link. Clark, Urszula & Esther Asprey. 2013. Dialects of English, West Midlands English: Birmingham and the Black Country. Edinburgh: Edinburgh University Press. Collins Bartholomew Map Archive. 2013. Mapseeker: Railway Map of England and Wales 1852. http://www.mapseeker.co.uk/collins-bartholomew/Collins-Railway%20Maps/?tcID=&id= 1261&mapcode=BC0211&areaA=a2&areaB=b1b6#phead. Glasgow: Collins Bartholomew (accessed 22 December 2013). Domesday Book online. 1999–2013. http://www.domesdaybook.co.uk/ (accessed 19 November 2013). Dyer, Alan. 2000. Midlands. In Peter Clark (ed.), The Cambridge urban history of Britain, Vol. 2, 93–110. Cambridge: Cambridge University Press. Ellis, Alexander J. 1968[1889]. On early English pronunciation, Vol. 5. New York: Greenwood. George, Paul S. 1996. Miami: One hundred years of history. South Florida History 24 (2). 22–36. http://www.historymiami.org/research-miami/topics/history-of-miami/ (accessed 17 November 2013). Hill, David. 1981. An atlas of Anglo-Saxon England. Toronto: University of Toronto Press. Hopkins, Eric. 1989. Birmingham: The first manufacturing town in the world, 1760–1840. London: Weidenfeld & Nicolson. Hutton, William. 2004[1783]. An history of Birmingham. Champaign, IL: Project Gutenberg. http://www.gutenberg.org/files/13926/13926-h/13926-h.htm (accessed 15–17 November and 23 December 2013). Johnston, Paul A. 1979. A synchronic and historical view of Border Area bimoric vowel systems. University of Edinburgh Ph.D. thesis. Johnston, Paul A. 2014. The anatomy of a Middle English transition zone: Middle Brummie and its relatives. Paper delivered at the 49th Annual Kalamazoo Medieval Congress, 8 May 2014. Kerswill, Paul. 1984. Social and linguistic aspects of Durham (e:). Journal of the International Phonetic Association 14. 13–34. Kerswill, Paul & Ann Williams. 2000. Creating a new town koine: Children and language change in Milton Keynes. Language in Society 29. 65–115. Kristensson, Gillis. 1987. A survey of Middle English dialects, 1290–1350: The West Midland counties. Lund: Gleerup. Langton, John. 2000. Urban growth and change: From the late seventeenth century to 1841. In Peter Clark (ed.), The Cambridge urban history of Britain, Vol. 2, 453–491. Cambridge: Cambridge University Press. Mathisen, Anne Grethe. 1999. Sandwell, West Midlands: Ambiguous perspectives on gender patterns and models of change. In Paul Foulkes & Gerard J. Docherty (eds.), Urban voices: Accent studies in the British Isles, 107–123. London: Arnold. McIntosh, Angus, Michael L. Samuels & Michael Benskin. 1986. A linguistic atlas of Late Mediaeval English. 4 vols. Aberdeen: Aberdeen University Press. McIntosh, Angus, Michael L. Samuels, Michael Benskin & Margaret Laing. 2013. An electronic version of the linguistic atlas of Late Middle English. http://www.lel.ed.ac.uk/ihd/elalme/ elalme.html (accessed 18 November to 15 December, 2013).
200
Paul A. Johnston, Jr.
Moore, Robert. 2011. Heart of England. http://heartengland.blogspot.com/2011/09/forest-ofarden.html (accessed 19 November 2013). Moore, Samuel, Sanford B. Meech & Harold Whitehall. 1935. Middle English dialect characteristics and dialect boundaries. Essays and Studies in English and Comparative Literature 13. 1–60. Ann Arbor: University of Michigan Press. Office of National Statistics. 2013. Table KS101UK: 2011 Census: Usual resident population, local authorities in the United Kingdom. http://www.ons.gov.uk/ons/datasets-and-tables/ (accessed 26 November 2013). Orton, Harold & Michael F. Barry. 1969. The survey of English dialects, Vol. 2: The West Midland counties: Basic material. Parts 1–3. Leeds: Arnold. Orton, Harold & Wilfred J. Halliday. 1962. The survey of English dialects, Vol. 1: The six Northern counties and the Isle of Man: Basic material. Parts 1–3. Leeds: Arnold. Orton, Harold & Philip M. Tilling. 1969. The survey of English dialects, Vol. 3: The East Midland counties and East Anglia: Basic material. Parts 1–3. Leeds: Arnold. Pollner, Clausdirk. 1985. Englisch in Livingston (Bamberger Beitrage zur englischen Sprach wissenschaft 16). Bern: Peter Lang. Schofield, R. E. 1963. The Lunar Society at Birmingham: A social history of provincial science and industry in eighteenth-century England. Oxford: Clarendon Press. Serjeantson, Mary S. 1927. The dialects of the West Midlands in Middle English. III. Tentative assignments of texts to the West Midland dialect area. The Review of English Studies 3 (11). 319–331. Stephens, W. B. 1964. Economic and social history: Industry and trade, 1500–1880. In W. B. Stephens (ed.), A history of the county of Warwick, Vol. 7: The city of Birmingham, 81–139. http://www.british-history.ac.uk/report.aspx?compid=22964 (accessed 5 December 2013; 23 December 2013). Tillott, P. M. 1964. Economic and social history: Medieval industry and trade. In W. B. Stephens (ed.), A history of the county of Warwick, Vol. 7: The city of Birmingham, 73–80. http:// www.british-history.ac.uk/report.aspx?compid=22963 (accessed 23 December 2013). Trudgill, Peter. 2004. The inevitability of colonial Englishes. Edinburgh: Edinburgh University Press. Trudgill, Peter & Tina Foxcroft. 1978. On the sociolinguistics of vocalic mergers: Transfer and approximation in East Anglia. In Peter Trudgill (ed.), Sociolinguistic patterns in British English, 69–79. London: Arnold. Wells, John C. 1982. Accents of English 2: The British Isles. Cambridge: Cambridge University Press.
Megan E. Hartman
Style and politics in The Battle of Brunanburh and The Battle of Maldon 1 Introduction The Battle of Maldon and The Battle of Brunanburh have been described as poems that are “at the periphery of heroic literature, simultaneously evoking and alter ing heroic conventions” (Fulk and Cain 2013: 320). Their liminal status comes largely from their subject material: while adhering to many conventions of the heroic tradition, these poems depict contemporary historical figures rather than legendary heroes from out of the past. This unusual choice of topic lends itself to a particularly nationalistic agenda, which both of these poems take up. The combination of subject matter and political preoccupation has led many scholars to discuss these two poems together and illustrate the many similarities between them. In spite of the comparable subject matter, however, the poems are quite different when it comes to their style of composition, as exhibited by the length of the lines, the distribution of the different verse types, and the use of traditional poetic features. The verifiably contrasting styles in poems that were composed in roughly the same time period point towards a new type of evidence relevant for poetic analysis: syntactic and metrical features as an indicator of pragmatic aims. Poetic style is something that seems to have been shifting near the end of the Old English period. Cable (1991: 41–65) and Bredehoft (2005: 70–98) in particular note the profound differences between traditional poems such as Beowulf and eleventh-century poems such as Durham, The death of Edward, and The grave (although Bredehoft also argues that some tenth-century poems, such as The Battle of Maldon, exhibit a major stylistic shift as well). They argue that these poems no longer follow the same rules as classical Old English verse and should be analyzed with a different system that allows for more flexibility and does not include features such as resolution or secondary stress. I argue that some tenthcentury poets compose with a relaxed style that anticipates these later poems while still maintaining the traditional poetic features. Fulk (1992: 251–268) demonstrates that many poems that can be externally dated to the tenth century have some general features that set them apart from more tradi tional poems such as Beowulf. One change he notes is that stress seems to have been reduced on the secondary level. For evidence, he shows that tenth-century scribes confused vowels in medial syllables, syncopated medial syllables in longer words, and simplified geminates after medial vowels; all of these changes suggest that the
202
Megan E. Hartman
syllables were not particularly distinct and therefore received little stress. A second change, which occurs perhaps as a result of the first, is that poetic compounds were not used as frequently in later poetry; because the second element would no longer be as strongly stressed, it would have been more difficult to distinguish nonlexicalized compounds that require that both elements of the word be understood. Without these frequent compounds, the distribution of the verse types also changes, with a decrease in the incidence of type D and E, since these are the metrical types that require the secondary stress characteristic of compounds.1 Finally, some of the poems show a freer treatment of the unstressed syllables. While not all tenth-cen tury poems show all of these features, they are quite common among demonstrably late poetry and create an overall poetic style that sets late poems apart. However, while a looser style overall characterizes late Old English poems as a group, the individual poets demonstrate it to different degrees. Some poets use lines that are as concise and regulated as those in Beowulf, with relatively short drops, a variety of stress patterns, and careful particle placement that matches the tendencies that Hans Kuhn describes in his laws. Others adapt their line to a greater extent – frequently expanding the length of the drops, using anacrusis often, and not con straining the placement of particles in the line to a great degree – to accommodate the contemporary language more easily into each line. And, of course, some compose with a style between these two extremes. I have previously illustrated these overall differences, demonstrating that syntactic change motivated the increasingly flexible metrical patterns but that some poets made stylistic choices, particularly using an unusually large number of short verses in variation with a single noun, in order to keep their lines short (Hartman 2014). Here, I will look more closely at The Battle of Maldon and The Battle of Brunanburh in particular to demonstrate how political factors and rhetorical aims can determine those stylistic choices. Maldon and Brunanburh quite clearly illustrate opposing stylistic choices. A comparison of their individual styles to Beowulf, which I use as an example of a particularly conservative poem that is representative of traditional heroic poetry, demonstrates that the Brunanburh poet uses a very conservative poetic style while the Maldon poet expands his lines substantially. The contrasting styles in these poems are important not only because they illustrate the diversity of the late Old English poetic tradition, but also because the poets’ stylistic choices
1 The five verse types, as first defined by Eduard Sievers (1893), are type A: P x P x; type B: x P x P; type C: x P P x; type D: P P S x; and type E: P S x P (where P stands for primary stress on a heavy syllable, S stands for secondary stress on a heavy syllable, these letters in lowercase stand for the corresponding stress on a light syllable, and x stands for an unstressed syllable). Unstressed positions, or drops, can be formed from more than one syllable, though they tend to stay rela tively short in classical verse.
Style and politics in The Battle of Brunanburh and The Battle of Maldon
203
are directly linked to the rhetorical purpose of each poem. In the case of these two poems, each poet’s relationship to the overall poetic tradition and how he uses it to develop his nationalistic agenda determines his style. The Brunanburh poet’s conservative style allows him to employ the poetic tradition in a relatively straightforward way as a means of exultant praise of the English royal line: he echoes the short verses and terse diction of earlier poets to implicitly position his heroes, King Æthelstan and his brother Edmund, in the place of the legen dary figures in traditional poems. While also ascribing the heroic values of the past to the contemporary figures of his poem, the Maldon poet uses a consciously contemporary style, perhaps trying to bring those values into the present and to suggest that the English nobles should remember them in the face of the current Viking invasions. Thus, not only is the style of late Old English poetry influenced by overarching linguistic change and personal choice, but it also reflects, and can therefore be used as evidence to analyze, the aims of the poems themselves.
2 The Battle of Brunanburh 2.1 Pragmatic aims In its rhetorical purpose, Brunanburh seems relatively straightforward: scholars agree that it is a nationalistic poem that celebrates the English victory at Brunan burh to the point of gloating, thereby elevating the royal brothers who are the heroes of the poem. In fact, close attention to the text illustrates numerous ways in which the poet subverts the tone and thematic material of traditional heroic poetry in order to reinforce his nationalistic ideology. What stands out, then, is his slavish adherence to stylistic norms: in contrast to the innovative content, the poetic style remains quite traditional, with relatively short verses and very noun-heavy diction. Significantly, the terse diction used to form these short verses does not echo the style of the Anglo-Saxon Chronicle, where the poem is found, but instead illustrates that the poet shifts away from contemporary syntax to compose the short lines.2 This combination may represent an attempt to revise
2 Although the syntax of the Chronicle is often simple and straightforward, it still contains quite a few function words. For example, the entry for 933 opens with Her for Æϸelstan cyning in on Scotland, ægϸer ge mid landhere, ge mid scyphere ‘Here, King Æthelstan went into Scotland, both with his land army and with his ship army’ (Plummer and Earle 1892: 1.106). This sentence uses two prepositions to indicate a single direction (in on), contains a correlative conjunction that consists of three function words (ægϸer ge … ge), and repeats the preposition mid to express ac companiment. These are the types of function words that are, if not avoided, at least limited in Brunanburh.
204
Megan E. Hartman
the poetic tradition by bringing historical and overtly nationalistic material to the nation’s native literary form. By doing so, the poet is attempting to insert contem porary historical heroes into the role of figures from the legendary past, thereby elevating the glory of the current English regime. In spite of arguments by some scholars of later periods that true nationalism did not exist until after the printing press, Hall (2006: 4–17) illustrates that the Anglo-Saxon Chronicle in general and Brunanburh in particular have some clearly nationalistic properties. The Chronicle is part of King Alfred’s translation project and therefore uses a standard state language, speaks of a united Angelcynn, and makes reference to features of a well-established state such as an organized governing body and a standing army. Brunanburh emphasizes the idea of unity by erasing any sort of ambiguities the battle might have presented and creating a clear picture of good versus evil: the poem presents an alliance of Scots and Scandinavians that is stronger than it probably was in reality, and it suppres ses any indication that some Scandinavians fought on the side of the English (as Egils saga Skallagrímssonar suggests might have been the case). Thus, the poem presents a picture of a unified English people fighting and defeating a powerful alliance of their enemies against the backdrop of a larger narrative of a nation coming into being. Both Irvine (1991: 202) and Thormann (1991: 7–8; 1997: 64–65) show how the poetic form itself is particularly important to this endeavor. As a traditional form from a shared past, the alliterative long line can emphasize the unity of the kingdom, especially when the poet uses it, as he does here, to create a narrative of shared victory that leads directly to a new form of heroism illus trated by the current king. Significantly, as Hill (2000: 109–122) shows, the poem focuses on the virtues of the past rather than its heroes.3 Therefore, he argues, these virtues serve to unite England behind a set of leaders who share the heroic traits that are common to the entire culture. To create this nationalistic pride, the poet shifts away from the typical tone and subject material of heroic poetry. Typically, heroic poetry focuses on tragic subjects, thereby allowing the poet to measure the character of men who are
3 The poem does mention two past figures, Hengest and Horsa, but they are from what an AngloSaxon might have considered the immediate historical past rather than the long-ago past of leg ends. Hill suggests that because the royal line traces their lineage back to Hengest and Horsa, the reference evokes the right to rule that the Anglo-Saxons won in battle and also gives further legitimacy to the current kings, thereby reinforcing the unity and justification of the victorious English army. He argues that the poet avoids the legendary references in order to create a new model of a hero and, with it, a new model of lordship: an “inherently legislative lordship that is both God-Fearing, mythologically sanctioned, and boundary setting as firmly as such things can be done in this world” (Hill 2000: 110).
Style and politics in The Battle of Brunanburh and The Battle of Maldon
205
facing hopeless odds.4 Brunanburh, in contrast, depicts a battle which appears not to have challenged the English warriors at all, creating a nationalistic propa ganda that suggests English rule is uncontestable. The poem does not focus on the actual battle, but rather the aftermath, when victory was assured. That focus in itself prevents the poet from exploring the character of the warriors when facing adversity, since the adversity has already passed. Moreover, the battle seems to have been rather one-sided. Every time the poet speaks of death, he connects the death only to the Northmen, to the point that even the beasts of battle feast upon the Northmen alone.5 In contrast, the English are depicted only as actively killing and hunting down their enemy or else reveling in their glory after the battle. This focus on triumph creates a related shift in tone. As Fulk and Cain (2003: 223) put it, “The tone of exultation is uncharacteristic of early Germanic verse, which celebrates valor – on whichever side of a conflict – rather than the humiliation of the defeated”.
2.2 Method and syntactic style The style of the poem seems to stand in contrast with this tone because it is very traditional in nature. The feature that most clearly illustrates this traditional style is the heavy use of variation. Arthur Brodeur (1959: 40) defines variation in Old English poetry as “a double or multiple statement of the same concept or idea in different words, with a more or less perceptible shift in stress: (…) the second member, while restating essentially the same concept or idea, may do so in a manner which emphasizes a somewhat different aspect of it”. He studies the variation of Beowulf in depth, noting that it can take multiple forms, such as a few words or short phrases with parallel structure to an initial word, entire clauses in variation with each other, or even clauses that stand in variation with
4 For discussion of the tragic subject of heroic poetry, see Phillpotts (1992[1928]: 4–5), Tolkien (1992[1936]: 23–26), and Fulk and Cain (2003: 223–224). 5 As Niles (1987: 358) points out, while the three beasts of battle typically enter as the battle is about to commence and feed on the dead of both sides, in Brunanburh they appear only after the battle and feast solely on the enemy dead. In addition to the placement that Niles focuses on, the diction used to illustrate their action also gives the English royal brothers more power. The poet states that they “Lēton him behindan | hrǣw bryttian / sealwig-pādan” ‘they let the dark-coated one enjoy the corpses behind them’ (Brunanburh 60–61a). By using the verb lǣtan, the poet im plies that Æthelstan and Edmund are somehow in charge of the beasts of battle, suggesting that the brothers exercise some control over the fate that they bring on the enemy soldiers.
206
Megan E. Hartman
individual words. Brodeur illustrates the various types of variation in one partic ularly complex passage (1): (1)
Ðā wæs gylden hilt gamelum rince, hārum hildfruman, on hand gyfen, enta ǣrgeweorc; hit on ǣht gehwearf æfter dēofla hryre Denigea frēan, wundorsmiþa geweorc; (Beowulf 1677–1681a)6 ‘Then was the golden hilt, the ancient work of giants, given into the hand of the ancient warrior, the hoary war-chief; it, the work of wonder-smiths, passed into the possession of the lord of the Danes after the fall of the demon.’
Here we can see the variation of multiple noun phrases, such as gylden hilt ‘golden hilt’ with enta ǣrgeweorc ‘ancient work of giants’ or gamelum rince ‘ancient warrior’ and hārum hildfruman ‘hoary war-chief’. At the same time, though, the passage has variation of longer phrases, such as wæs on hand gyfen ‘was given into the hand’ and on ǣht gehwearf ‘passed into the possession’. Also, even when a single word appears more than once in variation – the sword, for instance, is also called wundorsmiϸa geweorc ‘the work of wonder-smiths’ – the variation occurs in different clauses, breaking it up. According to Brodeur, this variety of variation is what makes Beowulf such a sophisticated and aesthetically pleasing poem. In contrast to that in Beowulf, the variation in Brunanburh consists largely of single noun phrases in variation, which can come in groups of three or more within the same clause, putting some instances of variation up to four verses away from the original word. For example, when the poet describes the flight of Constantine, the king of the united Picts and Scots, he writes as in (2): (2)
Swelce þǣr ēac sē frōda mid flēame cōm on his cȳþþe norþ, Constantīnus, hār hilde-rinc. (Brunanburh 37–39a) ‘Likewise there also the old one came with flight to his kin in the north, Con stantine, the hoary battle-warrior.’
After the initial clause in this passage, two noun phrases, Constantīnus and hār hilderinc ‘hoary battle-warrior’, stand in variation with the subject, sē frōda ‘the old one’. Such repeated instances of variation are common throughout the poem. The poet’s choice to rely on that feature and make it so prominent shows clear knowledge of and
6 All references to Beowulf refer to Fulk, Bjork, and Niles (2008), though most of the textual diacritics have been omitted. All other references to Old English poems refer to Pope and Fulk (2001), in which the spelling is normalized. I chose this text because it marks vowel quantity, which is necessary for metrical studies.
Style and politics in The Battle of Brunanburh and The Battle of Maldon
207
investment in traditional poetic features. Nevertheless, scholars have debated whether or not this style of variation is effective. Brodeur (1959: 69) characterizes the varia tion in Brunanburh as “almost constant and mechanical”, while Pearsall (1977: 65) goes further to say that the poet “has more enthusiasm than judgement, and strings together ‘poetic’ compounds with a wild abandon”, and Scragg (2003: 115) argues that the repetitive syntax shows a certain limitation on the poet’s range of lexical variety. In contrast, other scholars argue that the poet uses variation with careful precision that highlights the poet’s rhetorical goals (see especially Bolton 1968: 363–372; Lawler 1974: 52–67). I do not wish to comment on the aesthetic value of the variation; instead, my focus is on why the poet would choose this style. In addition to the mere prominence of variation, the style also allows him to compose a large number of short, noun-heavy verses that help him to maintain traditional metrical patterns. The length of the variation is the first feature that aids the poet. Poets can put any number of grammatical units in variation, from single words all the way up to entire clauses. As one might expect, the longer phrases and clauses can create complex instances of variation, which may span multiple verses or include numerous unstressed function words. The Brunanburh poet rarely uses these longer instances of variation and instead prefers single nouns or short noun phrases con sisting of a noun plus a modifier. This preference allows him to compose several verses such as Constantīnus and hār hilde-rinc that have exactly four syllables. Indeed, over half the verses in the poem similarly contain only stressed words and nearly 60% have exactly four syllables. Such verses are relatively common in par ticularly conservative poems such as Beowulf, which similarly uses four-syllable verses approximately half the time, but as poets began to treat the unstressed posi tions more freely, they tended to write longer verses more often. Secondly, the variation on noun phrases also creates an especially nounheavy diction. A predominance of nouns, and particularly nouns that are part of the specific poetic koine, is another defining feature of Old English poetry. In the alliterative long line, poets add poetic terms to the text largely through variation. In the passage above, the poet uses not only the poetic term frōda, but also a poetic compound, hilderinc. As a result, a poetic simplex occurs in 18.49% of the verses in Brunanburh and a compound in 10.96% of the verses, as compared to an average of 19.05% and 14.28% in Beowulf.7 While the Brunanburh poet does not use poetic terms quite as often as the Beowulf poet does, he comes close. Finally, the style also allows the poet to include a greater variety of metri cal patterns than some of his contemporaries. In contrast to other late poems,
7 The statistics from Beowulf come from three different fifty-verse passages: lines 1–25, 301–325, and 2444–2468. I average three different passages in order to account for the stylistic differences that might result from narrating different types of scenes.
208
Megan E. Hartman
Brunanburh contains an incidence of verses with secondary stress (those of type D and E) that is comparable to, and even slightly higher than, Beowulf ’s: in Brunanburh, 21.23% of the verses are of type D or E, as compared to 19.4% in Beowulf. The variation helps to facilitate the variety of verses because poets frequently use compounds to express an idea in variation, and the compounds contain secon dary stress. Thus the poetic compound hilderinc above is used to create a type D4 (P P x S), adding to the variety of verse types in the passage. Significantly, though, the Brunanburh poet uses type D much more often. Only 4.11% of the verses in the poem are a type E, as compared to 8.8% in Beowulf. However, he uses type D so frequently, in 17.12% of the verses, that he still has a greater combined incidence of the verse types that contain secondary stress. It would seem, then, that he is less interested in a maximum variety of verses and more interested in verses with secondary stress, perhaps because he felt that the secondary stress alone was enough to evoke the traditional sound of conservative verse. Although the style seems to stand in contrast with the tone of the poem, it in fact makes sense that the poet would compose with such formal conservatism, because doing so helps him create a connection to the past. All of these traditional features together give the poem a distinctive sound that would have been recognizable. The terse verses would stand in contrast to other poems composed at that time, such as Judgment Day II or The Menologium, and presumably to normal spoken prose, which would likewise exhibit the weakening of stress on final syllables that leads to a decline in compounding and an increased use in function words. The disting uishing sound, together with the traditional lexical and syntactic features, would evoke the poetic tradition in spite of the change in tone. If anything, the shift in tone, together with the lack of traditional heroic figures, seems appropriate to the nationalized political context of the poem. By combining the traditional style with the new material, the poet is replacing the legendary figures that are the more usual focus of heroic poetry with the heroic figures he wishes to glorify.
3 The Battle of Maldon 3.1 Interpreting the poet’s pragmatic aims The Battle of Maldon has frequently been compared to The Battle of Brunanburh, partly because they are both relatively late poems8 and largely because they both
8 Maldon cannot be dated with absolute certainty, but a date close to the date of the actual battle seems probable for two main reasons: first, the poem depicts King Æthelred positively,
Style and politics in The Battle of Brunanburh and The Battle of Maldon
209
make the unusual move of glorifying a contemporary battle in the heroic mode. However, scholars have generally been less certain about what to make of the combination of heroic and contemporary elements in Maldon. For a long time, scholars viewed the warriors’ actions as an anachronistic but heroic picture of the ideal Germanic warrior as seen in Tacitus’s Germania (see Scragg 1981: 32; 1991: 1–36). Woolf (1986[1976]: 189–190), however, argues that the warriors exceed the loyalty of legendary retainers because they cannot effect the vengeance they vow to achieve and are therefore essentially vowing to die. Further scholars have built on this assessment, arguing that this new version of retainership may have more ties to Christian ideals.9 One major consideration in these arguments is the concept of men dying for their lord and what it might signify that these warriors choose to die on the battlefield rather than live to fight on. Niles (1991: 207–208) offers a different per spective, that the soldiers do not vow to die for their lord but intend in fact to avenge him. He argues that there is nothing in the poem as we have it to show that these men would inevitably die, and they could legitimately expect to live.10 This explanation demonstrates that the warriors’ vows would not have been startlingly new or out of place, but it does not remove the anachronistic elements from the situation. Not only do the vows mirror old heroic poetry, but the nature of the
in contrast to later versions of the Chronicle (see Scragg 1991: 26–27; Niles 1991: 90–91); second, the poem depicts the people involved with too much detail and accuracy for it to be likely to be the work of a man a generation or more removed from the battle (see Locherbie-Cameron 1988: 159–171; Fulk 1992: 417–418). The main argument for a late date is that the term eorl is generally not used to refer to an English nobleman until after Cnut came to power (see McKinnell 1975: 121–136). However, Fulk (1992: 415–416) demonstrates that eorl can more plausibly be explained as a dialectal form from Essex, which was at one time part of the Danelaw. This date would put the poem not exactly contemporary with Brunanburh, but still within the same century. 9 For various theories on what this new version of a retainer might consist of, see Anderson (1986: 247–272), Frank (1991[1976]: 95–106), Hill (1991: 67–88), and Ryner (2006: 266–276). 10 His main evidence is lines 207–208, hī woldon ϸā ealle | ōðer twēga | līf forlǣtan | oððe lēofne gewrecan, which he translates as ‘They all wished one [that is, the second] of two alternatives: to lose their lives or to avenge their beloved lord’. According to this argument, the men are only acknowledging that death will await them if they fail; they neither wish to die nor believe that they are likely to do so. While that reading of ōðer would be unique and is therefore unlikely, the warriors could be legitimately presenting the two options, saying that either of the two could happen, but that they no longer want any other possible outcome. In this, they would be speak ing in a way similar to Beowulf when, preparing to fight Grendel, he acknowledges two possible outcomes for the battle: victory or death. Certainly Beowulf believes that he can win – and in fact he elsewhere seems to fully expect that he will – but he nevertheless acknowledges the other possibility and ultimately states that God alone knows what the outcome will be.
210
Megan E. Hartman
relationship between lord and retainer appears to be a very personal one that emphasizes individual loyalty more than a larger, nationalistic impulse.11 The question then becomes, how should this nostalgic mode of heroism be interpreted? Certainly the soldiers who stay on the field are to be admired, specif ically because they have a choice and choose the more difficult option.12 Many also argue that this admiration leads to a kind of nation-building that is similar to that seen in Brunanburh; the poet both establishes unity by showing the solidarity of the soldiers and leads the audience to admire them.13 Other scholars argue that, because the soldiers ultimately lose the battle, the poem would have inspired lamentation rather than action. The poet may have admired such actions, but he knows that they are no longer politically viable – both because the rise of the central monarchy discourages personal relationships and because the strength of the Viking invasions makes such heroics impractical – and so the poem stands as an elegy to a heroic spirit that has been lost.14 Though we may expect this kind of elegy in light of the political climate in the late tenth century, the elegiac reading does not seem entirely viable because the poet does not solely dwell on the past; while the warriors’ ideologies reflect past values, the poet mixes these values with more contemporary ideas. One of these ideas is Christianity, which is highlighted in this poem through the devout attitudes of the warriors, particularly Byrhtnoth. Davis (1999: 151–169) argues that the poet places devout Christian warriors in these traditional heroic roles to use the influential poetic tradition to illustrate how heroic values can merge with Christian ideals. Similarly, the poet transports the individual loyalty of the warri ors in the meadhall to a more modern political organization when he puts vows in the mouths of soldiers of various backgrounds, from prominent nobleman to
11 The men demonstrate the personal relationship to their lord mostly strongly in their vows, where they refer to him by titles such as mīnne wine-dryhten ‘my lord and friend’ (Maldon 248b) or lēofum men ‘dear man’ (Maldon 319a). In addition, they call attention to their own personal vows given on the mead-bench before the battle, which illustrate their loyalty as individuals. In this way, the poet seems to be praising past attitudes of loyalty, thereby calling attention to these older traditions and giving them precedence over the existing system. 12 Studies that focus specifically on the idea of choice are Irving (1961), Frese (1986), Anderson (1986), and Morgan (1999). Cf. Harris (2010), who argues that the oaths highlight the importance of binding a community together rather than individual choice. 13 For arguments on how Maldon presents a positive view of Anglo-Saxon culture, see especially Hill (1991) and Campbell (1993). Some go so far as to say that the poem should be considered a corrective, praising this type of bravery in the face of King Æthelred’s policy of buying the Vi kings off; see in particular Scattergood (1984) and Keynes (1991), although Keynes does acknowl edge that the situation might have been seen as one with no good choice, and Knightly (2010). 14 For arguments on the elegiac nature of the poem, see Jones (1989) and Niles (1991).
Style and politics in The Battle of Brunanburh and The Battle of Maldon
211
freemen and even a hostage.15 By touching on the various strata of society, the poet is not quite creating the tightly knit group of the comitatus; the reciprocal relationship appears the same, but it extends more broadly across the men of the army, depicting an organization relevant to the English nation at large. Rather than looking backwards, then, the poet seems to be drawing past values into his own society.
3.2 Syntactic and prosodic correlates of style The style of the poem reflects and reinforces this move to modernize the older heroic values. Unlike the style of Brunanburh, Maldon’s style is not archaizing: the lines can be very long, sometimes with multiple polysyllabic drops or drops with multiple unstressed words; the poetic effects, such as the use of variation or the poetic koiné, are less prominent; and the lines tend toward the alternating stress patterns of type A and type B at the expense of verses with clashing or falling stress. The poet does use traditional poetic features to a degree, showing that he was in fact capable of doing so and not just an incompetent poet, but he seems to prefer a line that is closer to his contemporary speech than to traditional poetic diction. Several aspects of Maldon’s style point to a relatively contemporary syntax and sound. Most prominent are the changes imposed by the language’s shift towards a more analytic syntax.16 Due to the increasingly analytic nature of English, the poet employs many more function words, such as prepositions and pronouns, than poets of poems such as Beowulf and Brunanburh. For example, in the first fifty lines of all three poems, Maldon contains 18 pronouns and 15 prepositions, where Brunanburh has 3 pronouns and 10 prepositions and Beowulf has 8 pronouns and 5 prepositions. The Maldon poet’s startling use of function words can be seen in clauses such as Hē hæfde gōd geϸanc | ϸā hwīle ϸe hē mid handum | healdan meahte | bord and brād sweord ‘he had an unflinching spirit the time which he with hands could hold a shield and broad sword’ (Maldon 13b–15a). In this statement, the poet uses the pronoun hē twice, even though the subject does not change, so it would be possible to imply the third-person singular subject through the verb conjugation. He also
15 For further discussion of the significance of these different ranks, see Blake (1965: 338), Scragg (1991: 34), Williams (1992: 43), and Hill (1991: 75–76). 16 The shift in syntax occurred primarily during the Middle English period and was mostly completed by 1250. However, early signs of the shift can be seen in the late Old English period, starting in the tenth century (Traugott 1992: 185–186).
212
Megan E. Hartman
uses the preposition mid to show that hands are the instrument with which the subject, Byrhtnoth, holds his sword. Brunanburh can use similar features, but more commonly has clauses such as Bord-weall clufon | hēowon heaðu-linda | hamora lāfum ‘They cleaved the shield wall, hewed the battle shield with the leavings of hammers’ (Brunanburh 5b–6), in which the poet does not repeat the subject for two separate verbs and uses the instrumental/dative case ending to show that hamora lāfum is the instrument with which the subjects, King Æthelstan and his brother Edmund, are doing the cleaving. As the example shows, the Maldon poet seems to employ the incipient language changes much more fully. The Maldon poet’s choice to integrate these language changes to a greater degree than the Brunanburh poet results in relatively long verses that can have a large number of independent words in the drops, as can be seen in the example above. Table 1 shows the contrast between the conservative and contemporary styles. The Brunanburh poet’s preference for particularly short verses results in an even smaller proportion of polysyllabic verses than can be found in Beowulf. In contrast, Maldon has both a greater mean incidence of polysyllabic verses overall and a proportionately larger number of verses with four- and five-syllable drops. The same general pattern holds for the number of independent words per drop in the three poems, as shown in Table 2. In this case, the Maldon poet seems even more likely to include longer versions of the verses, since he uses more verses with two independent words in the drop than just one, and even three words are relatively common. Table 1: Percentage of verses with polysyllabic drops in Beowulf, Brunanburh, and Maldon
Beowulf (sample) Brunanburh Maldon
2 syllables
3 syllables
4 syllables
5 syllables
6 syllables
7 syllables
Total
33.33%
12.0%
1.33%
1.33%
1.33%
0%
49.32%
30.13% 40.22%
6.16% 13.71%
2.05% 7.24%
1.37% 4.01%
0% 0.62%
0% 0.15%
39.71% 65.95%
Table 2: Percentage of verses with one or more independent words in a drop in Beowulf, Brunanburh, and Maldon
Beowulf (sample) Brunanburh Maldon
1 word
2 words
3 words
4 words
5 words
Total
27.33% 28.08% 17.26%
14.67% 10.96% 29.28%
7.33% 4.11% 12.94%
1.33% 0.68% 5.24%
0.67% 0% 0.31%
51.33% 43.84% 65.02%
Style and politics in The Battle of Brunanburh and The Battle of Maldon
213
Other ways in which the poet integrates language change into his verse to a degree are evident in the lexical choices which seem to be a direct result of the phonological changes in process. In addition to an increase in function words, the poem also demonstrates a decrease in words with secondary stress. This change affects the incidence of compounds in particular, as well as other words with secondary stress. Where Brunanburh contains, proportionately, even more true compounds than Beowulf, with a compound in 23.97% of the verses as com pared to 19.08% of my corpus in Beowulf, Maldon has compounds in only 9.24% of the verses.17 The phonological shift can also be observed through the Maldon poet’s treat ment of some trisyllabic words, because he apparently disregards tertiary stress on some words that would have received it in more conservative poems. He uses several verses such as stīðlīce clipode ‘sternly called out’ (Maldon 25b), which cannot fit into one of Sievers’s five types if the long medial syllable of a trisyllabic word receives stress. Normally, the medial syllable of a word can be unstressed only if it is light and occurs before the second lift of the verse (see Cable 1991: 16–26; Fulk 1992: 201–202). In this case, putting tertiary stress on the medial syl lable of stīðlīce would result in a disallowed verse pattern: P S x p x x.18 Simi larly, the verse ǣr him Wīghelmes bearn (Maldon 300a) should also probably not receive tertiary stress on Wīghelmes. In this case, the verse could be fit into an existing verse type, an aE (x x P S x P), but this verse would be quite unusual, both because ǣr him would be particularly heavy anacrusis and because type E nor mally does not take anacrusis. A better scansion would be a type B2 (x x P x x P).19 As a result of the phonological effects on the verse, the distribution of the verse types has changed as well. Type A dominates in Maldon even more than it does in more conservative poems, and type B is also more prominent. Conse quently, the incidence of type D, the most prominent of the verses that generally
17 Interestingly, the breakdown of poetic versus non-poetic compounds is quite different in Brunanburh as compared to Beowulf. Poetic compounds are compounds that never appear in prose, often because one of the elements is from the poetic koiné, in contrast to non-poetic com pounds, which can appear regularly in prose. Although Brunanburh has proportionately more compounds overall, fewer than half of them are poetic compounds, while almost three quarters of the compounds in the selection from Beowulf are poetic. It would seem, then, that the Brunanburh poet is making an effort to choose compound words to vary his poetic style, but not coining them the same way that the Beowulf poet did. 18 Like this verse are Maldon 28a, 234b, 261b, 265b, and 282a. 19 Like this verse is Maldon 320a. Brunanburh has just one verse in which tertiary stress must be disregarded on a long syllable before the second lift: grǣdigne gūϸ-hafoc (D*2: P x x P s x) (Brunanburh 64a).
214
Megan E. Hartman
Table 3: Distribution by percentage of the five verse types in Beowulf, Brunanburh, and Maldon
Beowulf20 Brunanburh Maldon
Type A
Type B
Type C
Type D
Type E
Irregular
45.6% 47.26% 51.93%
15.2% 16.44% 21.57%
18.3% 15.07% 15.25%
10.6% 17.12% 6.16%
8.8% 4.11% 5.08%
1.5% 0% 0%
take secondary stress, decreases sharply. The verse distribution in Maldon as compared to the other two poems is as in Table 3. The table shows that the com bined incidence of type A and type B in Maldon far exceeds the incidence of those verse types in the other two poems, while the incidence of verses with secondary stress is considerably lower. This distribution gives the poem a different cadence than traditional poems, since the poet does not change verse types from verse to verse as frequently, making the rhythms more regular. Because the poet turns specifically to two forms of alternating stress patterns in particular, the resulting changes would resemble normal, spoken prose to a greater degree. While linguistic change certainly motivated these features in Maldon, the dif ference is more a stylistic choice than a linguistic necessity. As a near contem porary of the Maldon poet, the Brunanburh poet would have used a similar spoken syntax, yet the poems do not correspond. Where the Brunanburh poet uses fre quent variation to maintain some of the traditional features, the Maldon poet uses variation more sparingly and with a different style. Variation certainly occurs, but not with the same frequency, and the poet often puts clauses and longer phrases in variation rather than relying on short noun phrases with poetic compounds. For example, when Byrhtwold promises to remain by his lord near the end of the poem as we have it, saying ac ic mē be healfe | mīnum hlāford | be swā lēofum menn | licgan ϸence ‘but I for my part by my lord’s side, by so dear a man, intend to lie’ (Maldon 318–19), the poet puts be swā lēofum menn ‘by so dear a man’ in variation with the first line. Yet this second verse is also relatively long, with two unstressed words in the first drop. This use of variation serves to elevate the tone of this par ticular moment as well as to heighten the praise of Byrhtnoth, who receives two complimentary epithets. But it does not bring in poetic vocabulary or create a variety of verse types by adding a verse with falling stress; instead it creates an additional type B with a long drop that contains multiple unstressed words. 20 Unlike my other statistics from Beowulf, which use a sampling of 150 verses, these percen tages are calculated from the complete list of verse distribution complied by Bliss (1967). Bliss’s system varies from Sievers’s in that he introduces the idea of a type d set of verses with only one full stress in analogy with Sievers’s type A3 (which he calls a type a). In Sievers’s system, these verses would be scanned as types B or C, so I have counted them as such when calculating the total percentages of each verse.
Style and politics in The Battle of Brunanburh and The Battle of Maldon
215
By allowing contemporary syntax to influence the shape of the line to the degree that he does, the Maldon poet composes a poem that has a very different sound when recited than either Beowulf or Brunanburh: the unstressed positions are longer, the lines are more uniform, and many of the typical poetic effects are minimized. This is not to say that the poem is virtually prosaic; it maintains con sistent poetic features and even occasionally brings in rhyme. But in performance, the poem would have echoed the cadence and features of contemporary language to a degree rather than avoiding them in preference for something more archaic. Since poets traditionally distinguish their syntax from that of contemporary speech and make it sound more archaic, the Maldon poet’s shift toward everyday speech must have stood out. By juxtaposing this new sound with the old ideal, the poet may be trying to bring heroic ideals into the present and make them relevant. He is not lamenting their loss as a thing of the past, for he does not represent them as a thing of the past. Instead, he uses the poetry itself to represent them as ideas that can still be integrated into modern sensibilities.
4 Conclusion Close, systematic analysis of the contrasting styles in Maldon and Brunanburh demonstrates how style can be used as a new kind of evidence to analyze the rhe torical aims of Old English poetry, in this case specifically distinguishing between the two poets’ particular nationalistic goals. The Brunanburh poet promotes a new nationalistic agenda that supports the legitimacy of King Æthelstan and his line: although he is concerned with tradition and uses it to his advantage in his poem, he ultimately focuses more on the goals and values of the present. By recasting heroic poetry with a new tone and a new focus, he tries to rewrite the tradition to support a new type of hero in the ruling king. He taps into the power of the poetic form itself, specifically through the short, even terse, lines that bring to mind a particularly conservative poetic form, and then uses the poetry’s associations to support the new ideology he is setting forth For the Maldon poet, the heroic ideology carries the most importance, specif ically the ideology that he sees as making his nation great. By maintaining the tone and mindset of old poems, he does not suggest that his warriors’ actions would allow them to take the place of the heroes of old, but rather that they are great because their actions reflect those heroes’. Yet in spite of their adherence to past traditions, these heroes' actions do not come across as anachronistic and out of place, because the poet shifts both the context of their actions and the style of his writing in order to translate the Anglo-Saxon heroic ideal in the present. Rather than rewriting the heroic tradition, then, the Maldon poet uses his style in an attempt to reclaim it for the men of his day.
216
Megan E. Hartman
References Anderson, Earl R. 1986. The Battle of Maldon: Reappraisal of possible sources, date, and theme. In Phyllis Rugg Brown, Georgia Ronan Crampton & Fred C. Robinson (eds.), Modes of interpretation in Old English literature, 247–272. Toronto: University of Toronto Press. Blake, N. F. 1965. The Battle of Maldon. Neophilologus 49. 332–345. Bliss, A. J. 1967. The meter of ‘Beowulf’, revised edn. Oxford: Blackwell. Bolton, W. F. 1968. Variation in The Battle of Brunanburh. The Review of English Studies 19 n.s. 363–372. Bredehoft, Thomas A. 2005. Early English metre. Toronto: University of Toronto Press. Brodeur, Arthur Gilchrist. 1959. The art of ‘Beowulf’. Berkeley, CA: University of California Press. Cable, Thomas. 1991. The English alliterative tradition. Philadelphia: University of Pennsylvania Press. Campbell, James. 1993. England c. 991. In Janet Cooper (ed.), The Battle of Maldon: Fiction and fact, 1–17. London: Hambledon Press. Davis, Craig R. 1999. Cultural historicity in ‘The Battle of Maldon’. Philological Quarterly 78. 151–169. Frank, Roberta. 1991. The ideal of men dying with their lord in The Battle of Maldon: Anachronism or nouvelle vague? In Ian N. Wood & Niels Lund (eds.), People and places in northern Europe 500–1600: Studies presented to Peter Hayes Sawyer, 95–106. Rochester: Boydell & Brewer. Frese, Dolores Warwick. 1986. Poetic prowess in Brunanburh and Maldon: Winning, losing, and the literary outcome. In Phyllis Rugg Brown, Georgia Ronan Crampton & Fred C. Robinson (eds.), Modes of interpretation in Old English literature, 83–99. Toronto: University of Toronto Press. Fulk, R. D. 1992. A history of Old English meter. Philadelphia: University of Pennsylvania Press. Fulk, R. D. & Christopher M. Cain. 2013. A history of Old English literature. 2nd edn. Chichester: Wiley-Blackwell. Fulk, R. D., Robert E. Bjork & John D. Niles (eds.). 2008. Klaeber’s Beowulf and the Fight at Finnsburg. 4th edn. Toronto: University of Toronto Press. Hall, Stefan Thomas. 2006. Did the Anglo-Saxons have a sense of national identity? The Anglo-Saxon Chronicle and The Battle of Brunanburh. In Bruce E. Brandt & Michael S. Nagy (eds.), Proceedings of the 14th Northern Plains Conference on Earlier British Literature, 1–26. Brookings, SD: South Dakota State University. Harris, Stephen J. 2010. Oaths in The Battle of Maldon. In Robin Waugh & James Weldon (eds.), The hero recovered: Essays on medieval heroism in honor of George Clark, 85–109. Kalamazoo: Medieval Institute Publications. Hartman, Megan E. 2014. Poetic attitudes and adaptations in late Old English verse. Leeds Studies in English 91. 43–73. Hill, John M. 1991. Transcendental loyalty in The Battle of Maldon. Mediaevalia 17. 67–88. Hill, John M. 2000. The Anglo-Saxon warrior ethic. Gainesville, FL: University of Florida Press. Irvine, Martin. 1991. Medieval textuality and the archaeology of textual culture. In Allen J. Frantzen (ed.), Speaking two languages: Traditional disciplines and contemporary theory in medieval studies, 181–210, 276–284. Albany: State University of New York Press. Irving, Edward B., Jr. 1961. The heroic style in ‘The Battle of Maldon’. Studies in Philology 58. 457–467. Jones, Alex I. 1989. Lexical structure in The Battle of Maldon. In Geraldine Barnes, John Gunn, Sonya Jensen & Lee Jobling (eds.), Words and wordsmiths: A volume for H. L. Rogers, 63–69. Sydney: Department of English, University of Sydney.
Style and politics in The Battle of Brunanburh and The Battle of Maldon
217
Keynes, Simon. 1991. The historical context of the Battle of Maldon. In Donald Scragg (ed.), The Battle of Maldon, AD 991, 81–113. Oxford: Basil Blackwell. Knightly, Michael R. 2010. Communal interdependence in The Battle of Maldon. Studia Neophilologica 82. 58–68. Lawler, Traugott. 1974. Brunanburh: Craft and art. In John H. Dorenkamp (ed.), Literary studies: Essays in memory of Francis A. Drumm, 52–67. Worcester, MA: College of the Holy Cross. Locherbie-Cameron, M. A. L. 1988. Byrhtnoth, his noble companion, and his sister’s son. Medium Ævum 57. 159–171. McKinnell, John. 1975. The date of The Battle of Maldon. Medium Ævum 44. 121–136. Morgan, Gerald. 1999. Spectamur agendo: The universality of The Battle of Maldon. In Helen Conrad O’Briain, Anne Marie D’Arcy & John Scattergood (eds.), Text and gloss: Studies in insular learning and literature presented to Joseph Donovan Pheifer, 182–206. Dublin: Four Courts Press. Niles, John D. 1987. Skaldic technique in Brunanburh. Scandinavian Studies 59. 356–366. Niles, John D. 1991. Maldon and mythopoesis. Mediaevalia 17. 89–121. Pearsall, Derek. 1977. Old English and Middle English poetry. London: Routledge & Kegan Paul. Phillpotts, Bertha. 1992 [1928]. Wyrd and providence in Anglo-Saxon thought. In R. D. Fulk (ed.), Interpretations of ‘Beowulf’: A critical anthology’, 1–13. Bloomington, IN: Indiana University Press. Plummer, Charles & John Earle (eds.). 1892. Two of the Saxon Chronicles parallel, with supplemental extracts from the others. A revised text. 2 vols. Oxford: Clarendon Press. Pope, John C. & R. D. Fulk (eds.). 2000. Eight Old English Poems. New York: W. W. Norton. Ryner, Bradley D. 2006. Exchanging battle: Subjective and objective conflicts in The Battle of Maldon. English Studies 87(3). 266–276. Scattergood, John. 1984. The Battle of Maldon and history. In John Scattergood (ed.), Literature and learning in medieval and Renaissance England, 11–24. Dublin: Irish Academic Press. Scragg, Donald G. (ed.). 1981. The Battle of Maldon. Manchester: Manchester University Press. Scragg, Donald G. 1991. The Battle of Maldon. In Donald G. Scragg (ed.), The Battle of Maldon, AD 991, 1–36. Oxford: Basil Blackwell. Scragg, Donald G. 2003. A reading of Brunanburh. In Mark C. Amodio & Katherine O’Brien O’Keeffe (eds.), Unlocking the wordhord: Anglo-Saxon studies in memory of Edward B. Irving, Jr., 109–122. Toronto: University of Toronto Press. Sievers, Eduard. 1893. Altgermanische Metrik. Halle: Max Niemeyer. Thormann, Janet. 1991. The Battle of Brunanburh and the matter of history. Mediaevalia 17. 5–13. Thormann, Janet. 1997. The Anglo-Saxon Chronicle poems and the making of the English nation. In Allen J. Frantzen & John D. Niles (eds.), Anglo-Saxonism and the construction of social identity, 60–85. Gainesville, FL: University Press of Florida. Tolkien, J. R. R. 1992 [1936]. Beowulf: The monsters and the critics. In R. D. Fulk (ed.), Interpretations of ‘Beowulf’: A critical anthology, 14–44. Bloomington, IN: Indiana University Press. Traugott, Elizabeth Closs. 1992. Syntax. In Richard M. Hogg (ed.), The Cambridge history of the English language, Vol. 1: The beginnings to 1066, 168–289. Cambridge: Cambridge University Press. Williams, Ann. 1992. The Battle of Maldon and ‘The Battle of Maldon’: History, poetry, and propaganda. Medieval History 2(2). 25–44. Woolf, Rosemary. 1986 [1976]. The ideal of men dying with their lord in Germania and The Battle of Maldon. In Heather O’Donoghue (ed.), Art and doctrine: Essays on medieval literature, 175–196. London: Hambledon Press.
Don Chapman
Trinitarian terminology in Old English liturgical creeds 1 Introduction What evidence do we have of Anglo-Saxons exercising control over the lexicon to create a technical vocabulary? This paper will examine attempts to fashion a technical vocabulary for liturgical creeds in Old English. To a degree, this study aligns with others that examine vocabulary for specific domains, such as books (Gneuss 1985), plants (Sauer 1992), birds (Kitson 1997, 1998), animals (Sauer 1999), people (Sauer 2007), and grammar (Chapman 2010). Such studies are largely onomasiological, emphasizing the resources and methods Anglo-Saxons used to fashion a vocabulary for particular domains. One of their main purposes is to provide an inventory of terms used for a specific domain, and like those studies, this paper will identify an inventory of Old English terms, in this instance for the propositions about the nature of God found in liturgical creeds. But this paper has a further purpose: to investigate that inventory of terms for evidence of conscious crafting of technical vocabulary. It is one thing for a translator or glossator to come up with an English equivalent for a Latin term; it is quite another to standardize and propagate that term. Conscious control of vocabulary goes beyond the exigencies of translating a text to fostering mechanisms for language planning. We can at least infer some kinds of social infrastructure like education or scriptorium standards when we see such conscious control. Evidence for conscious control of these terms will come, paradoxically, from both uniformity and departures from uniformity among the terms. The importance of uniformity for establishing a technical vocabulary should be clear. Ideally, we expect one concept to be consistently represented by one term (Temmerman 2000: 125–154). Without such uniformity, it is difficult to even recognize technical terms. When we see an English term (like þriness) consistently used as a translation equivalent for a specialized Latin term (like trinitas), we can be confident that it was used as a technical term and that somehow a tradition had established the suitability of that correspondence. Yet uniformity does not always reveal the choices of individual writers. Some uniformity in the record may have owed to no more than close copying of manuscripts, since in the present case, many manuscripts containing English versions of the creeds are known to have been copied from earlier exemplars, with only minor alterations. Additionally, uniformity can result from mechanically choosing the best-known translation
220
Don Chapman
equivalent for a Latin term, regardless of whether the English term would have been fitting for the specialized context. In such cases, contrasts will sometimes offer better evidence of control. When we see a choice of terms (like English hlaford) contrasting with the predominant choice (like English drihten), we have some evidence that the writer was not merely copying the text or using a traditional English translation equivalent. And if that writer consistently employed a minority choice, we have greater evidence of that writer’s control. In short, we get important evidence of linguistic control from uniform use, but paradoxically, we also get important (sometimes better) evidence of linguistic control when we see departures from that uniformity. Working out the value of both kinds of evidence is central to the current argument. This paper will mainly examine English terms relating to the nature of the Trinity. While there are a few other kinds of technical terms within the creeds, like catholica and crucifixus, the theological doctrine of the Trinity calls for the largest number of technical terms in the creeds and requires the most complex semantic treatment. I have chosen to examine the Trinitarian terms in creeds, because creedal vocabulary is a domain where we should expect to see writers exercising conscious control of the lexicon. One purpose of Anglo-Saxon creeds was establishing and propagating orthodox belief, and, in a culture that took orthodox belief seriously, formulating that belief in acceptable and precise statements was serious work. Indeed, Ælfric and Wulfstan both noted the importance of learning “the Creed” in English, and both emphasized correctness. In a sermon entitled De fide catholica, Ælfric wrote, ælc cristen man sceal æfter rihte cunnan ægþer ge his pater noster ge his credan. Mid þam pater nostre he sceal hine gebiddan. mid þam credan he sceal his geleafan getrymman (ÆCHom I 20, 335.1) ‘Every Christian man should by right [or ‘correctly’] know both his Pater noster and his Creed. With the Pater noster he should pray, with the Creed he should confirm his faith’ (Thorpe 1844: 275).1
Wulfstan gave a similar injunction: Leofan menn, understandað georne, eallswa eow mycel ðearf is, þæt ge eowres cristendomes gescead witan, & ge eac geleornian þæt ge cunnan þæt ælc cristen man mid rihte cunnon sceall; þæt is, pater noster & credo in Deum. (WHom 7a 3) ‘Beloved men, understand fully, just as there is great need, so that you will know the basis of Christendom, and also
1 Unless otherwise noted, Old English texts have been taken from the Dictionary of Old English Corpus (Healey, Wilkin, and Xiang 2009). Short titles are taken from Dictionary of Old English short titles (Cameron et al. 2007). Translations are mine, unless otherwise noted. Bold-faced emphasis in any citation has been added.
Trinitarian terminology in Old English liturgical creeds
221
so that you will learn so that you will know what each Chistian man should correctly know, that is, the Pater noster and the Creed.’
Both Ælfric and Wulfstan reinforced their belief in the importance of learning the creeds in English by providing vernacular versions of the creeds. We may expect that the emphasis on correctness noted here would have extended to establishing a standard vocabulary for the creeds.
2 Sources for investigation 2.1 Creed translations and glosses The main sources of Old English creedal vocabulary are the English translations and glosses of the principal Latin creeds in the medieval liturgy: the Apostles’ Creed, the Nicene Creed, and the Athanasian Creed. Of these, the Athanasian Creed provides the best evidence of conscious control in developing a suitable vocabulary. First, the terms in the Athanasian Creed are both semantically and morphologically more complex than those in the other two. The Athanasian Creed is much longer than both the Apostles’ Creed and the Nicene Creed and explores in more depth the complexity of the Trinity (Kelly 1964); thus it uses more philosophical and theological terms. Second, we have continuous glosses of the Athanasian Creed from several different scribes, who sometimes differed in their glossarial choices, allowing greater opportunity to contrast different perspectives on the creed and the terminological choices that follow from them. Together, these two factors make the Athanasian Creed the most important of the three for exam ination here. The Apostles’ Creed is also well attested in our extant Old English corpus: it was translated or adapted in three manuscripts, fully glossed in six manuscripts, and sparsely glossed in two more.2 But the Apostles’ Creed is not as important as the Athanasian Creed for this investigation, because its vocabulary is less complex. In contrast, the Nicene Creed contains more complex terms than the Apostles’ Creed, but our only English version of it comes from one translation by
2 See Appendix A1 for a list of the manuscripts that contain versions of the Apostles’ Creed, where I have adapted the sigla used in Banks (1968). Close translations of the Apostles’ Creed are found in ApJ, ApK, ApM. These have been collated and published as Lit 3.1. Full, continuous glosses are in ApZ, Apa, Apb, Apc, Apd, and Ape. Scattered glosses are found in ApDR1 and ApDR2. Two additional versions of the Apostles’ Creed exist – one by Wulfstan (WHom 7a) and a metrical adaptation (Creed) – but these are fairly loose translations, and have not, therefore, been included in this investigation.
222
Don Chapman
Ælfric.3 So while this paper includes the English vocabulary from all three creeds in the inventory, it mostly focuses on vocabulary from the Athanasian Creed. The Athanasian Creed shows up as an additional canticle added to (or in some cases included in) a number of Anglo-Saxon psalters in the tenth and eleventh centuries. Three manuscripts (known by their sigla as A, D, and I) are important enough to require comment here. According to Gretsch (1999: 273), the oldest of the Athanasian Creed glosses is in the Royal Psalter (D). The creed gloss was done by the same person who glossed the Psalms (Gretsch argues that the glossator was Athelwold). Most of the other creed glosses follow D fairly closely, including the creed gloss in the Vespasian Psalter (A). Of course the Vespasian Psalter is famous for its early specimens of the Mercian dialect in the interlinear glosses to the Psalms, but the gloss of the Athanasian Creed in that manuscript was done much later (Gretsch 1999: 278). The gloss that stands most apart from all the others is the Lambeth Psalter (I) (Gretsch 1999: 26–27), in which the creed gloss was done by the same person who glossed the Psalms. Like the Psalm glosses, the creed gloss shows at once care and independence from the other glossing traditions (Sisam and Sisam 1959: 72).
2.2 Trinitarian passages in prose Another source for Old English Trinitarian vocabulary comes from writings that do not necessarily translate or gloss a particular liturgical creed, but instead discuss the theological doctrines in the creeds. Of the writers who discussed Trinitarian concepts in the creedal formulae, Ælfric predominates (Grundy 1991: 26). He devoted an entire homily to De fide catholica (ÆCHom I, 20), in which he stated and explained the doctrines of the Trinity found in the creeds. This is the fullest treatment of the Trinity in Old English and contains many phrases that are close renditions of the Latin phrases of the creeds. In addition to this homily, Ælfric discussed the Trinity on numerous other occasions, sometimes using English formulations that are very close to the Latin creedal formulations, and sometimes using phrases that are not as close, but still recognizable as technical Trinitarian vocabulary. Apart from Ælfric, only a few writers occasionally discussed the Trinity, none in similar depth. Throughout my discussion of terms, I distinguish the use of Trinitarian terms in Ælfric from other Anglo-Saxon writers. A list of passages using Trinitarian language, from Ælfric and others, is given in Appendix A2.
3 This translation has survived in three manuscripts listed in Appendix A1.
Trinitarian terminology in Old English liturgical creeds
223
2.3 Other glosses Several texts provide useful points of comparison for seeing how those English terms used for Trinitarian concepts were used in non-Trinitarian contexts. Among the most important of these are the English glosses to the Psalms, especially for those manuscripts in which both the Psalms and the creeds were glossed by the same writers, as was the case with the Royal (D) and Lambeth (I) manuscripts. Since the same Latin terms in the Psalms usually have different senses from their use in the creeds, it is instructive to see the extent to which an English glossator chose the same English term or opted for a different one. Other glossed texts are also useful for evaluating the English terminology, even though they would not have been made by the same glossators who glossed the creeds. They allow us to see how a given Latin word would have been translated in various contexts.
2.4 Other prose The non-Trinitarian passages in Ælfric’s corpus are important as a comparison to Trinitarian passages, mainly because most Trinitarian passages were written by Ælfric. It is useful to see how he used particular English words for discussing both Trinitarian and non-Trinitarian concepts. Needless to say, the rest of the Old English corpus will also be useful for similar comparisons. In short, there are seven sources referred to throughout the paper, as shown in Table 1. From these sources, I have identified 34 Latin terms that could be considered technical terms. A collation of these terms for the seven sources listed in Table 1 can be found at http://linguistics.byu.edu/faculty/chapmand/oetrinvocab.pdf. For most of these terms, we see a fair amount of uniformity among Old English equivalents, though variation persists for a few. The terms that Table 1: Sources consulted Sources of Trinitarian terms 1 Creed translations and glosses 2 Prose treatment of Trinitarian concepts: Ælfric 3 Prose treatment of Trinitarian concepts: Others Points of comparison Psalm glosses
4 5 6 7
Other glosses Non-Trinitarian prose: Ælfric Non-Trinitarian prose: Others
224
Don Chapman
show uniformity can be distributed into three groups: two sub-groups of terms with high uniformity and low variation, and another group with less uniformity overall, but a patterned variation that reveals considerable uniformity among some of the variants. All three groups provide evidence of the degree that AngloSaxons controlled the vocabulary in translations of liturgical creeds.
3 Degrees of uniformity in the English terms 3.1 High uniformity in the English terms: Terms with mainly specialized senses This group consists of terms that are mainly, if not exclusively, used for Trinitarian concepts more or less uniformly in the manuscripts under scrutiny. These terms each have one main sense confined to one main Trinitarian concept. Such terms are shown in Table 2. Terms in Table 2 tend to be morphologically complex in both Latin and English, with easily recognizable affixes, or, in the case of Halig Gast, two words in a fixed phrase. Most of them look like loan-formations that could well have been created, though not necessarily by the creed glossators, to render the specific Trinitarian sense. The terms are usually fairly abstract as well, and the combination of abstractness and morphological complexity helps the terms remain specialized, since they have easily recognized morphology dedicated to a specialized meaning that is not likely to become more general. Thus, these terms persist with considerable stability. Their uniformity attests to the control writers used in creating a technical vocabulary.
3.2 High uniformity in the English terms: Terms with both specialized and general senses This group of terms also shows high uniformity among the English terms, but in this case, the terms are used in general language and do have other meanings Table 2: Terms of high uniformity with mainly specialized senses Latin term
Old English term
Modern English meaning
Trinitas unitas coaeternus divinitas incarnatus Spiritus Sanctus
Þrinnesse annesse efenece godcundnes geflæschamod Halig Gast
‘Trinity’ ‘unity’ ‘co-eternal’ ‘divinity’ ‘incarnate’ ‘Holy Ghost’
Trinitarian terminology in Old English liturgical creeds
225
besides the Trinitarian ones. Such terms are shown in Table 3. The correspondences between the English and Latin terms in Table 3 had already been well established for their more general senses before those same English terms were pressed into service for the more specialized Trinitarian senses. Such an adoption of a specialized sense where a correspondence between more general senses already existed is the essence of semantic loans. Semantic loans provide only weak evidence of control, mainly because they would presumably be default options when a translator or glossator would encounter a technical term that also had general use. A translator or glossator might have simply chosen the usual word for translating the Latin term without considering how apt the English term would be for a Trinitarian sense. That the glossators and translators would have chosen the terms consistently points to uniformity, but not so much to control. Yet the desirability of semantic loans as a translation strategy should not be dismissed, and little evidence of control is not the same as evidence of little control. Semantic loans do not have to be unthinking, mechanical applications; they can serve the purposes of translation, perhaps better, in some cases, than other strategies. Sometimes, the general use of a term may well give it a core meaning applicable or even desirable for discussing the nature of the Trinity. The core meanings of pater and filius, for example, are essential for discussing the persons of the Trinity, even though in that context, notions of father and son are highly specialized and unlike those operating in everyday speech. In one passage, for example, Ælfric explicitly points to where our experience breaks down, since we know of no fathers who did not precede their sons, yet the doctrine of the Trinity posits just such a relationship between the Father and the Son (ÆCHom I 20, 337.57–60). In that doctrine of the Trinity, the core meanings of pater/fæder and filius/sunu are salient, so the terms are retained. These terms seem to be instances of concepts with a prototype structure for which polysemy can be useful (Temmerman 2000: 149–154). So even though semantic loans give poor evidence of control of terminology, they should not be dismissed as showing no control. We see considerable uniformity for the terms in Tables 2 and 3, but in some ways, these would have been the easiest terms to standardize. For the first group, Table 3: Terms of high uniformity with both specialized and general senses Latin term
Old English term
Modern English meaning
Pater Filius eternus perfectus humana
Fæder Sunu ece fullfremed mennisc
‘Father’ ‘Son’ ‘eternal’ ‘perfect, complete’ ‘human’
226
Don Chapman
there is only one (highly specialized) sense for the Latin term, so matching one specialized sense to an English term would have been a relatively straightforward matter. For the second group, the well-established English equivalents of partic ular Latin words will still work well for the more specialized use in Trinitarian language. There is little to be gained by choosing another term. Harder terms to standardize, however, would be those with more than one meaning from a variety of domains, and whose precise meaning in the context of the creeds is not clearcut. For these, drawing upon established Latin-English correspondences may not be as effective a strategy. Such terms constitute the next group, which in many cases demonstrates the strongest evidence for linguistic control.
3.3 Less uniformity, patterned variation of English terms For this group of words, the degree of overall uniformity is not as high, but within the variation, we see notable uniformity in the uses of certain writers. For that reason, I refer to it as patterned variation. For these terms, several English equivalents are usually used in the glosses and translations. The contrasts among them make the process of standardization and cultivation more apparent, and thus provide some of the best evidence of conscious and conscientious attempts to formulate and control a technical vocabulary. In this group of terms, we see a common pattern. There is usually a predominant choice for the Latin term in the various glosses, but whatever it is, the Lambeth gloss (I) usually disagrees with it. It is this pattern that makes examining the differences between I and other glosses the most fruitful. Thus, while other glosses occasionally depart from the predominant reading, their patterns of departure are not sufficiently regular to warrant further examination. What makes the departures of the Lambeth glosses from the rest of the tradition more noticeable is that Ælfric, in his discussions of the Trinity, often agrees with Lambeth against the other glosses. In this case, it appears that both the Lambeth glossator and Ælfric were not satisfied with the term the others used, and adopted another term. Such terms are discussed in the following section.
4 Degrees of agreement between Ælfric and the Lambeth glossator What follows is a discussion of some of the words that best illustrate the agreement or non-agreement of Ælfric and the Lambeth glossator. For each word, I first give the Latin context and the distribution of the majority glosses and of those in the Lambeth Psalter (I). Then I give the distribution in Ælfric, followed
Trinitarian terminology in Old English liturgical creeds
227
by the other non-Ælfrician texts, and further speculate as to why Ælfric and the Lambeth glossator chose different terms.
4.1 Substantia The Latin term that best illustrates the choices of the Lambeth glossator and Ælfric in forming a technical vocabulary is easily substantia. Substantia occurs four times in the Athanasian Creed, where it always means something like ‘essence’ (Kelly 1964: 81): 4. Neque confundentes personas, neque substantiam separantes 31. Deus est ex substantia Patris ante saecula genitus: et homo est ex substantia matris in saeculo natus. 36. Unus omnino, non confusione substantiae, sed unitate personae.4 In nearly every instance in the Athanasian Creed, substantia is glossed as sped by all the glossators except the Lambeth glossator, who used edwist in all four instances.5 In the Psalm glosses, by way of comparison, this same division between the Lambeth glossator and the rest of the glossators also holds, even though substantia as used in the Psalms usually means something closer to ‘fortune, property, abundance’, which was another major sense of substantia (Lewis 1879: s.v. substantia). The five occurrences of substantia in the Psalms are glossed by sped in all the glosses except Lambeth. Much of this uniformity owes to a tradition of copying psalter glosses, not necessarily to independent glossators choosing the same form. Yet the Lambeth Psalter did not follow the tradition for this gloss (and many others). In the Lambeth Psalter, sped is never used by itself for substantia; instead, either edwist or the pairing of sped and edwist is used in every instance. To modern eyes, the choice of sped for both the ‘abundance’ sense in the Psalms and the ‘essence’ sense in the Athanasian Creed seems surprising. We can still detect the ‘abundance’ or ‘success’ sense in modern (albeit obsoles cent) reflexes like good speed, but we have no sense of speed that comes close to ‘essence’. Because of this discrepancy, we might be tempted to assume that the
4 (4) ‘Neither confounding the Persons; nor dividing the Essence (…) (31) God, of the Essence of the Father; begotten before the worlds; and Man, of the Essence of his Mother, born in the world. (…) (36) One altogether; not by confusion of Essence; but by unity of Person.’ English translations of all three creeds are from Schaff (1877). 5 J and Har each use edwist one time.
228
Don Chapman
use of sped for substantia represents unthinking matching of one English term to its customary Latin equivalent, regardless of the specialized context. While such rote matching is possible, we should not assume too quickly that the AngloSaxons would have seen sped as a poor choice for the ‘essence’ sense of substantia. Perhaps the Anglo-Saxon writers who used sped for both senses saw nothing strange about its having the same kind of polysemy that substantia had. It could be that sped acquired the ‘essence’ sense as a semantic loan: writers would have considered it acceptable to borrow the ‘essence’ sense, since there were already strong correspondences between sped and substantia. At any rate, sped is also used for ‘essence’ in prose, where there was less opportunity for mechanically mapping sped to substantia: Hu mæg ic þæt ongitan be ðære halgan þrynnesse, þæt syndon þreo hadas & hwæðre an God is & an godcund sped? (HomS 2 [ScraggVerc 16], 142) ‘How may I understand concerning the Holy Trinity, that there are three persons, but there is nevertheless one God and one divine substance?’ nu swa þonne seo halige þrynis, fæder & sunu & se haliga gast, þæt is an God & an godcund sped & an miht & an þrym & an wuldor & an willa & efenece (HomS 2 [ScraggVerc 16], 159). ‘Now thus, then, [is] the Holy Trinity, Father, Son, and the Holy Ghost; that is one God and one divine substance and one might and one power and one glory and one will and coeternal.’ us gedafenaþ ærest þæt we gemunen & gereccen be Gode ælmihtigum, þe geworhte heofonas & eorðan & ealle gesceafta, þone we sculon gelyfan þrynlicne on hadum & anlicne on spede (HomS 34 [ScraggVerc 19], 1). ‘It is appropriate for us first to remember and confess concerning God Almighty, who created the heavens and the earth and all creation, whom we must believe as triple in person and single in substance.’
These early uses of sped in prose suggest that we should not view the choice of sped in the Athanasian Creed glosses as a result of unreflective glossing that simply applies one gloss for one sense of a word to that same word when it is used in a markedly different sense. Sped could be used for both senses. Yet Ælfric never used sped in that ‘essence’ sense, but only in the ‘abundance’ sense. For the ‘essence’ sense in general and for the Trinitarian use of substantia in particular, Ælfric always used edwist, as in these two passages: ðæs ælmihtigan godes sunu becume on minum innoðe. & mennisce edwiste of me genime (ÆCHom I 13, 286.148) ‘[Let] the Son of Almighty God enter my womb, and receive human substance from me’ (Thorpe 1844: 201). ælc ðæra þreora is god þeahhwæðere hi ealle an god for þan ðe hi ealle habbað an gecynd & ane godcundnysse. & ane edwiste. & an geþeaht. & an weorc & ane mægenþrynnysse. & gelic
Trinitarian terminology in Old English liturgical creeds
229
wuldor. & efenece rice (ÆCHom I 20, 339.134). ‘Each of the three is God, yet they are all one God; for they all have one nature, and one Godhead, and one substance, and one counsel, and one work, and one majesty, and like glory and coeternal rule’ (Thorpe 1844: 284).
Ælfric was as consistent as the Lambeth glossator in his preference for edwist to render the ‘essence’ sense. Both seem to have isolated the ‘essence’ sense of substantia and assigned it to edwist. Even in the glosses of the Psalms, where the Lambeth glossator used the pair “sped vel edwist”,6 the division in senses might hold, since those passages are arguably ambiguous – they can be interpreted with either sense. Perhaps the two terms (sped and edwist) were not meant to be appositive choices, as much as choices between two different readings of the passage. In using edwist, Ælfric and the Lambeth glossator seem to have chosen a fairly unusual word. Edwist does not show up in very many texts outside Ælfric, and most of those are glosses, not prose. Furthermore, they occur mostly in late texts – at least as late as Ælfric’s writings. Why Ælfric and the Lambeth glossator made this choice is a matter of speculation. Perhaps the polysemy of sped was too bothersome. In general, polysemy will work against the one-form-one-meaning “univocity” principle of technical vocabulary, and when the range of meanings was as wide as it was with sped, the polysemy may have been that much more objectionable. The use of sped and edwist to gloss substantia provides a snapshot view into the creation of a technical term. First we have a word sped showing up relatively early in prose with the meaning needed for rendering the sense of substantia used in the Athanasian Creed. Then we see that same word picked up and adopted as a gloss of the creed. Yet one of the glosses of the creed does not use that word, but looks for a different word instead. That other word is the same one preferred by Ælfric, the writer who wrote most about the Trinity in English. The contrasts between Ælfric and the Lambeth glossator on the one hand and practically all other writers on the other hand allow us to see the conscious choice that these writers exerted in cultivating this theological term in English.
6 Double and triple glosses are characteristic of the Lambeth glossator (Sisam and Sisam 1959: 72; Gretsch 1999: 27).
230
Don Chapman
4.2 Conversio The glossing of conversio likewise highlights the differences between Ælfric and the Lambeth glossator versus other writers. There is only one occurrence of conversio in the Athanasian Creed: 35. Unus autem non conversione divinitatis in carnem, sed assumptione humanitatis in Deum.7 Nearly all the glosses of the creed used gecyrrednes, but the Lambeth glossator used awendednes instead. Lambeth was joined by L, who used onwendednes. The majority form gecyrrednes is also the typical gloss of conversio elsewhere, namely, all eight instances of conversio outside the Athanasian Creed. Once again, Ælfric agrees with the Lambeth glossator, but this time the comparison is slightly more removed, since Ælfric never used either nominalization – gecyrrednes or awendednes – to talk about the Trinity, though he used both for other meanings. Nonetheless, Ælfric did use the verb form awend to convey a similar meaning. In two passages explicating the notion of the Word being made flesh, Ælfric wrote, Næs þæt word to flæsce awend: ac hit wæs mid menniscum flæsce befangen (ÆCHom I 2, 196.174). ‘The Word was not turned to flesh, but it was invested with human flesh’ (Thorpe 1844: 41). Nis þæt halige Word awend to flæsce, ac se heofonlica æþeling her on þas woruld com & þa menniscnysse genam of Marian innoðe (ÆHom 1, 404). ‘That holy Word is not changed to flesh, but the heavenly prince came here into this world and assumed human nature from Mary’s womb.’
So we have some evidence of Ælfric’s favoring the form also used by the Lambeth glossator. In the non-Ælfrician corpus, we again find use of an English word to capture this same notion of the Trinity. In this instance, it is similar to the Lambeth glossator’s choice, namely onwendednes: Þonne marþon þære godcundnesse nænig onwendnesse on carcerne wæs of þære menniscan gecynde, na las of þære godcundan miht (HomS 8 [BlHom 2], 89–91). ‘But, moreover, there was no change either of the divine nature or of the divine power in its imprisonment in the human nature’ (Morris 1967: 18).
7 ‘One altogether; not by confusion of Essence; but by unity of Person.’
Trinitarian terminology in Old English liturgical creeds
231
This passage from one of the Blickling Homilies would have been written earlier than the glosses of the Athanasian Creed, and if it represents any kind of wider pparently preference among Anglo-Saxons for discussing the incarnation, Ælfric a would have shared that preference with his choice of verb, and certainly the Lambeth glossator did with his choice of nominalization. The Royal glossator, on the other hand, chose the most typical term for conversio – gecyrrednes – and most of the other glossators followed him. Though the evidence is not as clear as it is for substantia, it still points in the same direction, namely that Ælfric and the Lambeth glossator show some conscious care in their choice of terms for the Trinity. Again we can speculate as to why they would have chosen awendednes instead of gecyrrednes, even though gecyrrednes seems to be the more common English gloss for conversio. All those meanings of conversio / gecyrrednes in the Old English corpus are different from the one in the creed. Elsewhere, conversio is usually taken to mean ‘conversion to some form of religious life’, whether it is from paganism to Christianity, secular to monastic life, or simply from sinfulness to righteousness. The Dictionary of Old English (DOE; see Healey 1986–2004) lists the definitions as variously ‘a turning to God (from sinfulness), conversion’, ‘conversion to Christianity’, and ‘change from secular life to religious life’ (s.v. gecyrrednes). These are the senses that Ælfric used gecyrrednes for, as well. The Lambeth glossator’s choice, awendedness, more typically glosses Latin words meaning ‘change’, such as commutatio, and ‘change, alteration’ is practically the only definition that the DOE gives for awendednes (s.v. awendednes). When awendednes is used in prose texts, it usually refers to the mutability of this life compared with the immutability of the next, as in Ælfric’s use: Hwæt is godra manna deað buton awendednys. and færr fram deaðe to ðam ecan life? Se lichama awent to eorðan. and anbidað æristes (ÆCHom II, 13, 132.155). ‘What is the death of good men, except change, and a journey from death to eternal life? The body changes to earth and awaits the resurrection.’
It could be that Ælfric and the Lambeth glossator recognized that awendednes was closer than gecyrredness to this sense of conversio, because it speaks of change of nature, not a change of affiliation or behavior. It could also be that Ælfric and the Lambeth glossator were aware that gecyrredness already had an established sense in religious vocabulary, and they did not want to make it more polysemous, and thus less precise as a technical term. Additionally, gecyrrednes would have positive connotations from a religious point of view, but awendednes would not necessarily. We can imagine that one would not want to talk about God converting from divinity to humanity (better to worse), even to deny such change happened, in the same way as a person converting from paganism to Christianity
232
Don Chapman
(worse to better). There is some evidence, then, that again Ælfric and the Lambeth glossator chose the forms that resulted in less polysemy rather than more.
4.3 Assumptio The term that balances conversio in the Athanasian Creed is assumptio, and it too pairs the Lambeth glossator with Ælfric against the rest of the glossators. The term occurs in the second part of the same verse: 35. Unus autem non conversione divinitatis in carnem, sed assumptione human itatis in Deum. Nearly all the glosses of the Athanasian Creed have a form of onfangenness, while Lambeth (I) alone has underfangenness. (The only other departure is K, which has genamon.) Ælfric did not use either nominalization in discussing the Trinity,8 but he regularly used the verb underfon, not onfon, to refer to the same concept, as seen in these examples: Swa eac crist ana underfeng þa menniscnysse and na se fæder ne se halga gast (ÆCHom I 20, 340.154). ‘In like manner Christ alone assumed human nature, and not the Father, nor the Holy Ghost’ (Thorpe 1844: 287). and se sunu ana underfæng menniscnysse and on þisum dege wearð to menn geboren (ÆLS [Christmas]: 77). ‘and the Son alone took human nature, and on this day was born as man’ (Skeat 1881: 15). and he ana underfeng þa menniscnysse. and þrowade deað. (ÆCHom I 15, 306.189). ‘And he alone assumed human nature, and suffered death’ (Thorpe 1844: 229).
Ælfric was not as uniform in this choice, however, as he was with the other terms discussed so far. Ælfric also wrote of this same concept with two other formula tions, namely niman and befon. He used niman four times; two examples are given here: ðæs ælmihtigan godes sunu becume on minum innoðe. and mennisce edwiste of me genime (ÆCHom I 13 286.148). ‘[Let] the Son of Almighty God enter my womb, and receive human substance from me’ (Thorpe 1844: 201).
8 He used anfangnes three times for accepting or receiving other things. He never used underfangnes.
Trinitarian terminology in Old English liturgical creeds
233
se þe us alisde of urum þeowte syððan mid þære menniscnisse, þe he of Marian genam; (AELet 4, 36). ‘who afterwards saved us from bondage with that human nature, which he assumed from Mary’
Ælfric used befon three times: Næs þæt word to flæsce awend: ac hit wæs mid menniscum flæsce befangen (ÆCHom I 2, 196.174). ‘The Word was not turned to flesh, but it was invested with human flesh’ (Thorpe 1844: 41). Ne wearð se fæder mid menniscnysse befangen ac hwæðere he asende his sunu to ure alysednysse (ÆCHom I 20, 340.140). ‘The Father was not invested with human nature, but yet he sent his Son for our redemption’ (Thorpe 1844: 285). Nis na se fæder mid þære menniscnysse befangen. ne se halga gast. ac se sunu ana (ÆCHom II 3, 23.130). ‘The Father has never been invested with human nature, nor the Holy Ghost, but the Son alone.’
Of these three possibilities, underfon is Ælfric’s predominant choice, as he used it some 20 times for this concept. Tellingly, he never used onfon. Overall, Ælfric agrees with the Lambeth glossator on this choice. For prose writers besides Ælfric, we have three writers and five instances of the formula menniscnys underfeng. Three of those instances come from Wulfstan, like this one: Eac we gelyfað & georne witan þæt Crist Godes sunu is ægðer ge soð Godd on godcundnesse ge eac soð man þurh ða menniscnesse þe he underfeng þurh his modor, Sancta Marian (WHom 7, 33). ‘We also believe, and earnestly know that Christ, the Son of God, is both true God in divinity and true man by the human nature which he assumed from his mother Saint Mary.’
I have not been able to find any instances of the combination menniscnes and onfeng. There is not a great deal of difference between onfon and underfon; indeed, they have the same root fon meaning ‘take’ or ‘receive’. For the nominalization itself, onfangenness is more common in Old English generally, occurring some 22 times compared to the 3 occurrences of underfangenness. Yet, outside the creed glosses, neither onfangenness nor underfangenness is ever used for the sense of God assuming human nature. Instead the non-Trinitarian uses are closer to the DOE definition of onfangenness: “act of receiving, acceptance (of something / someone gen.)” (DOE s.v. anfangnes). A typical use of onfangenness is to speak of God not showing acceptance of persons or people receiving the Eucharist.
234
Don Chapman
It appears that we have two tendencies: onfangenness is more common than underfangenness, but underfon menniscnes is more common than any other formulation of this concept. It may be that the Royal glossator (D) and those who followed him chose the most common nominalization for assumptio, while the Lambeth glossator used a nominalization of underfon, since underfon is the typical term for this concept throughout the Old English corpus. If underfon menniscnes was already established, Ælfric and the Lambeth glossator were not so much promoting a new term as preserving an old one. As a further matter of speculation, perhaps the term underfon has the added benefit of suggesting a downward direction, thus showing the condescension involved in the incarnation of God.
4.4 Separantes Separantes occurs once in the Athanasian Creed: 4. Neque confundentes personas, neque substantiam separantes.9 The glosses are more diverse for separantes, with three or four different possibil ities. Most glosses are some form of syndriende (A, D, E, G, and Har); L might be lumped in with these, since that glossator used asyndrien. On the other hand, C used ascyrgynde, and K used synderlice. Not surprisingly, Lambeth differs from all of these, using the term totwæmende. None of these present participle forms is common outside the creed, but forms of (a)syndrian are the most popular glosses of forms of Latin separo, and totwæman is the least, though none of those outside glosses treats the same concept given in the Athanasian Creed. Ælfric did not use any of these participial forms to discuss the inseparability of the Trinity, but he did use other forms of totwæman in every instance that he discussed the notion. Fæder & sunu & halig gast. ne magon beon togædere genamode. ac hi ne beoð swa ðeah nahwar totwæmede (ÆCHom I 20, 339.110). ‘Father, Son, and Holy Ghost, may not be named together, but yet they are nowhere separated’ (Thorpe 1844: 283). Hi ne magon beon togædere genemnede fæder & sunu. & halig gast. ac hi ne beoð mid ænigum fæce. fram him sylfum ahwar totwæmede (ÆCHom I 33, 463.139). ‘They may not be named together, Father, and Son, and Holy Ghost, but they are not by any space anywhere separated from themselves’ (Thorpe 1844: 501).
9 ‘Neither confounding the Persons; nor dividing the Essence.’
Trinitarian terminology in Old English liturgical creeds
235
I have not been able to find any writers beside Ælfric discussing the inseparability of the Trinity, so we do not have a term to compare with Ælfric’s. Once again, we have a case of the Lambeth glossator and Ælfric agreeing against the other glossators, and once again, we have the most common form in the creed glosses (syndriende) aligning with the most common form in all glosses (syndrian). Whether Ælfric and the Lambeth glossator were promoting a new term or preserving an old one is hard to tell.
4.5 Procedens Procedens occurs once in the Athanasian Creed: 23.
Spiritus Sanctus a Patre et Filio: non factus, nec creatus, nec genitus, sed procedens.10
This is a crucial term in the creed, as it distinguishes the third person of the Trinity from the other two. In the creed glosses of procedens, the Lambeth gloss disagrees with Royal (D), but this time it finds more company than usual: as many agree with I as agree with D. The gloss for D is forðgewitende, and those agreeing with D are A, E, and F. The gloss for I, on the other hand, is forðstæppende, and those agreeing with I are C, J, and G.11 K used forðgangende, and L used forðcumende. In their choices, both I and D were consistent with their choices to gloss forms of procedo in the Psalms, where the Lambeth glossator consistently used forðstæppende and the Royal glossator used forðgewitende. In a way, the creed gloss is a continuation of the choices for the psalter gloss. For the most part, Ælfric agrees with the Lambeth glossator in choosing forðstæppende. In eleven passages Ælfric used this term in a fairly close rendition of the creed formulation: Se Halga Gast wæs æfre of ðam Fæder and of ðam Suna na acenned, ac forðstæppende (ÆAdmon 2). ‘The Holy Ghost was never begotten from the Father or from the Son, but proceeding.’ and se halga gast is angin æfre of þam fæder and of þam sunu, na acenned ac forðstæppende (ÆLS [Christmas]: 76). ‘and the Holy Ghost is the Beginning, eternally of the Father and of the Son, not begotten, but proceeding’ (Skeat 1881: 15).
10 ‘The Holy Ghost is of the Father and of the Son; neither made, nor created, nor begotten; but proceeding.’ 11 The DOE says that forðstæp in G is a clipping for forðstæppende (s.v. forþ-stæppan).
236
Don Chapman
Yet Ælfric did not use this term exclusively. Almost as often he used the word gæð, as in these examples (there are at least 8 occurrences): Nis se halga gast na sunu. for þan ðe he nis na acenned. ac he gæð of þam fæder. and of þam suna gelice (ÆCHom I 20, 338.81). ‘The Holy Ghost is not a son, for he is not begotten, but he proceeds from the Father and the Son’ (Thorpe 1844: 281). And eac se rihta geleafa us tæcð þæt we sceolon gelyfan on þone halgan gast. he is se liffæstenda god se gæð of þam fæder and of þam suna (ÆCHom I 20, 338.86). ‘And the right faith also teaches us, that we should believe in the Holy Ghost: he is the quickening God, who proceeds from the Father and the Son’ (Thorpe 1844: 23).
Half of these eight occurrences come from ÆCHom I 20, which is the homily on the Trinity, yet in that homily, Ælfric also used forðstæppende: Se halga gast is god forðstæppende of þam fæder. and of þam suna (ÆCHom I 20, 337.49). ‘The Holy Ghost is God proceeding from the Father and the Son’ (Thorpe 1844: 279).
But there is a hint that the word might not be very clear, since Ælfric soon explains it with a different term: Nis he geworht ne gesceapen ne acenned. ac he is forðstæppende. þæt is ofgangende. of þam fæder. and of þam suna þam he is gelic. and efenece (ÆCHom I 20, 337.79). ‘He is not made, nor created, nor begotten, but he is proceeding, that is going from, the Father and the Son, with whom he is equal and coeternal’ (Thorpe 1844: 281).
Here, the explanatory phrase “þæt is ofgangende” calls attention to forðstæppende as a more specialized term. Apparently forðstæppende would have been unusual enough to require further explanation. After this explanation, Ælfric used gæð for the rest of the homily. This is similar to the technique that Ælfric used to introduce grammatical terminology in his grammar, where he always gave an English equivalent at the first instance of the term, and if that English term was familiar, he would usually re-use it. Otherwise, he would re-use the Latin term (Chapman 2010). In this instance regarding procedens, it appears that Ælfric used forðstæppende as a more technical formulation, but he did not hesitate to use the common term gæð when explaining the concepts. More telling, Ælfric even used gæð in his translation of the Nicene Creed. There the Latin formulation is “Qui ex Patre Filioque procedit”, which Ælfric translated as “se gæð of ðam Fæder and of ðam Suna”. It appears that Ælfric sometimes preferred the more technical formulation and sometimes the more general. Perhaps when he wanted to emphasize precision, he used the technical term, and when he wanted to emphasize clarity he used the general one.
Trinitarian terminology in Old English liturgical creeds
237
Outside the Ælfrician corpus, I have found two writers treating this concept. Byrhtferth of Ramsey used a form of forðsteppan: se þridda of þam Fæder & þam Suna forðstop (ByrM 1, 198.5). ‘The third proceeded from the Father and the Son.’
Byrhtferth flourished slightly later than Ælfric and seems to have been influenced by the Winchester schools in other ways (Gretsch 1999: 216–217), so perhaps he was following Ælfric or was at least participating in the same tradition. The other, an early text,12 used forðleorendne, a word found in none of the creed glosses: & Haligne Gast forðleorendne of Fæder & Suna (Bede 4). ‘And the Holy Ghost proceeding from the Father and the Son’
This is a rare English equivalent for procedens, but not unique.13 Still, two eleventh-century manuscripts changed it to forðferendne, and in one of those, stæppende is written above the word (DOE s.v. forðleorend). This could consitute slight evidence that forðstæppende was becoming increasingly favored as a technical term for the concept. At any rate, it moved in the same direction as the other formulations. For procedens, we find once again that Ælfric and the Lambeth glossator agree in their choice of terms. Their choices were shared by other creed glossators – more than with most words where there is any disagreement, and in a few instances outside Ælfric, glossators also used forðstæppende as the term to render procedens. We have some evidence, then, that writers were cultivating forðstæppende as the preferred technical term. At the same time, there is some indication that forðstæppende was an unusual word, which may make it a good choice for a technical theological concept, but a less optimal choice for explaining the concept to non-specialists. Ælfric often resorted to a more common word, gæð, presumably for simpler explanations. It might be worth noting that the other repeated choice in the creed glosses – forðgewitan – had many meanings that would have made it less than desirable as a gloss. In particular, it was a common euphemism for dying (cf. Present-Day English pass away), and it was also frequently used to
12 The Old English translation of Bede’s Historia ecclesiastica was made in the late ninth century; this passage renders the confession of faith made at the Council of Hatfield in 679. 13 Leoran is an Anglian word which would have been replaced in later texts. See Wenisch (1979: 175–178) and Campbell (1951).
238
Don Chapman
refer to the past (see DOE s.v. forðgewitan). Neither association would be welcome in a discussion of the nature of the Holy Ghost. 4.6 Genitus So far, we have seen several instances in which the Lambeth gloss has a different English term than D, and in most instances, it is a minority choice. In all these, Ælfric agrees with Lambeth against D. Genitus provides a noticeable instance in which Ælfric actually agrees with the majority against I. Genitus occurs 4 times in the Athanasian Creed and once in the Nicene creed: 21. Pater a nullo est factus: nec creatus, nec genitus 22. Filius a Patre solo est: non factus, nec creatus, sed genitus. 23. Spiritus Sanctus a Patre et Filio: non factus, nec creatus, nec genitus, sed procedens. 31. Deus est ex substantia Patris ante saecula genitus: et homo est ex substantia matris in saeculo natus. Nic. 9. Genitum, non factum, consubstantialem Patri.14 The overwhelming choice in the glosses and translations of these lines is acenned. Of the 36 glosses of genitus in the Athanasian Creed, 29 are acenned and 3 more are gecenned. Both translations of the Nicene Creed used acenned. The Lambeth glossator is completely alone, but consistent in choosing gestryned, which is used in all 4 instances of genitus in I. In this case, Ælfric does not match Lambeth. Instead, Ælfric overwhelmingly chose acenned. Ælfric used acenned some fifty times to refer to the relationships mentioned in the creeds. At least 18 of these instances are in close renditions of the Latin, such as the following: Nis se sunu na geworht ne gesceapen ac he is acenned (ÆCHom I 20, 337.55). ‘The Son is neither made nor created, but he is begotten’ (Thorpe 1844: 279). Nis he geworht ne gesceapen ne acenned. ac he is forðstæppende. þæt is ofgangende. of þam fæder. and of þam suna þam he is gelic. and efenece (ÆCHom I 20, 337.79). ‘He is not
14 Athanasian: (21) ‘The Father is made of none; neither created, nor begotten. (22) The Son is of the Father alone; not made, nor created; but begotten. (23) The Holy Ghost is of the Father and of the Son; neither made, nor created, nor begotten; but proceeding. (…) (31) God, of the Essence of the Father; begotten before the worlds; and Man, of the Essence of his Mother, born in the world.’ Nicene: (9) ‘Begotten, not made, being of one substance with the Father.’
Trinitarian terminology in Old English liturgical creeds
239
made, nor created, nor begotten, but he is proceeding, that is going from, the Father and the Son, with whom he is equal and coeternal’ (Thorpe 1844: 281). Ðu eart unacenned fæder: he is sunu of þe æfre acenned. and se halga gast is æfre forðstæppende of þe and of þinum bearne (ÆCHom I 31, 445.167). ‘Thou art the unbegotten Father, he is the Son ever of thee begotten, and the Holy Ghost is ever proceeding from thee and thy Son’ (Thorpe 1844: 465, 467).
In only one case did Ælfric use gestryned: Se ælmihtiga fæder nis of nanum oðrum gestrynde (ÆIntSig 69.514). ‘The Almighty Father is not begotten of any other.’
Acenned is so clearly the preferred term in Old English that it is surprising the Lambeth glossator would choose anything else. It is possible that the Lambeth glossator was aiming for more precision: most glossators use one English term (acenned) to render two different Latin terms (genitus and natus), so it could be that the Lambeth glossator was trying to reduce the polysemy of acenned by using gestryned for genitus and accened for natus, as in the gloss of verse 31: God is of edwiste fæder ær woruldum gestryned & man is of edwiste moder on worulde acenned.
But if the Lambeth glossator was trying to create less polysemous terms, Ælfric was content to go along with the traditional terms. This constitutes one of the few times that Ælfric did not opt for the more specialized term. In fact, if the Lambeth glossator had not broken from the tradition, we might well be classifying acenned as the same kind of gloss as pater and filius – a term with both a general meaning and specialized sense within the creeds.
4.7 Dominus We see a similar split with the gloss of Dominus. Dominus occurs 12 times: 9 times in the Athanasian Creed, twice in the Nicene Creed, and once in the Apostles’ Creed: 17. Ita Dominus Pater, Dominus Filius, Dominus Spiritus Sanctus. 18. Et tamen non tres Domini, sed unus est Dominus. 19. Quia, sicut singillatim unamquamque personam Deum ac Dominum confiteri christiana veritate compellimur. 20. Ita tres Deos aut Dominos dicere catholica religione prohibemur.
240
Don Chapman
29. Sed necessarium est ad aeternam salutem, ut incarnationem quoque Domini nostri Iesu Christi fideliter credat. 30. Est ergo fides recta ut credamus et confiteamur, quia Dominus noster Iesus Christus, Dei Filius, Deus et homo est. Nic. Et in unum Dominum Iesum Christum Nic. Et in Spiritum Sanctum, Dominum et vivificantem Ap. filium eius unicum dominum nostrum15 The choice of gloss is overwhelmingly drihten: of the 79 glosses or translations of dominus, 70 of them use drihten. The remaining nine glosses used hlaford: six are found in the Lambeth creed glosses, and three are in K. In his use of hlaford, the Lambeth glossator shows a surprising degree of uniformity, but even he used drihten 3 times. This preference is another one of the few that Ælfric did not share. In reference to the Trinity, Ælfric used drihten 8 times and hlaford 2 times. In the corpus of Trinitarian passages, writers besides Ælfric use drihten 9 times and never use hlaford. This strong preference for drihten reflects the language as a whole, where drihten predominates. There are over 5000 instances of drihten in the DOE corpus, not even counting spellings with , but only 101 instances of hlaford. It is no surprise that Ælfric is following the strong preference already established in the language. Yet because of such a strong preference for drihten generally, perhaps the two instances of hlaford leaking into Ælfric’s writings is significant – at least it goes in the same direction as the Lambeth glossator’s choices.
5 Discussion and conclusions This paper has examined three groups of Trinitarian concepts that exhibit enough uniformity in word choice to suggest some degree of conscious and conscientious cultivation of a technical Trinitarian vocabulary. Two of those groups have complete or nearly complete uniformity across a variety of writers. Of those (nearly) completely uniform groups, we can discern one subgroup that consists of highly
15 Athanasian: (17) ‘So likewise the Father is Lord; the Son Lord; and the Holy Ghost Lord. (18) And yet not three Lords; but one Lord. (19) For like as we are compelled by the Christian verity; to acknowledge every Person by himself to be God and Lord; (20) So are we forbidden by the catholic religion; to say, There are three Gods, or three Lords (…). (29) Furthermore it is necessary to everlasting salvation; that he also believe faithfully in the Incarnation of our Lord Jesus Christ. (3) For the right Faith is, that we believe and confess; that our Lord Jesus Christ, the Son of God, is God and Man.’ Nicene: ‘And in one Lord Jesus Christ. (…) And in the Holy Ghost, the Lord and Giver of life.’ Apostles’: ‘And in Jesus Christ his only Son our Lord.’
Trinitarian terminology in Old English liturgical creeds
241
specialized concepts whose only use is for the Trinitarian doctrine. These are words like trinitas and divinitas. There is a low degree of polysemy in these Latin terms, and the English words chosen to gloss or translate them show a similarly low degree of polysemy. The low polysemy of these terms suits them for technical terminology and helps them to remain uniform. These words also tend to be morphologically complex, making their forms more distinct and less likely to be picked up for general use. The other subgroup of highly uniform terms seems to be just the opposite, as it contains words that are well established in general use and tend to be morphologically simple, like pater and filius. The explanation for their uniformity likely has something to do with the strong prototypical core sense of these words, so that whatever new use they are put to, they maintain that strong core sense. Thus, in any figurative use of pater, the core sense will always be salient. Because that salience is desired in even the specialized use of pater and filius in discussions of the Trinity, the terms are well suited and remain uniform. The third group of terms exhibiting uniformity consists of those that actually show variation in the choice of an English term, but uniformity in the choices of the Lambeth glossator and Ælfric. Their persistent agreement with each other against the choices of the other glossators and writers stands out and gives us a stronger suggestion of conscious control in the development of technical theological vocabulary. Somehow these two writers were able to maintain a high degree of uniformity in their own writings, and in the cases where their choices differ from the predominant choice (most notably their choice of edwist for substantia), that uniformity most likely resulted from conscious word choice. Such uniformity is more striking when we see how often Ælfric and the Lambeth glossator’s choices correspond to each other. It has been recognized for some time that Ælfric and the Lambeth glossator share lexical affinities, and their writings have served as important texts for the identification of the Winchester vocabulary (Hofstetter 1988: 157–158; Gretsch 1999: 259–260). At the very least, some of their correspondences in Trinitarian vocabulary perhaps should be listed among the “Winchester words” redominantly in Winchester texts, but not in non-Winchester texts. that are used p Again, edwist vs. sped for Latin substantia would be the strongest candidate. It remains a matter of speculation how Ælfric and the Lambeth glossator could have achieved such consistent agreement. Perhaps it resulted from the cultivation of a uniform vocabulary in the instruction at Winchester, as several have suggested. Perhaps writers were able to influence each other on the lexical level, even without explicit dictionaries (though glossaries and glossed texts may have served a similar function [Lendinara 1999: 61]). Whatever the mechanism, there are some hints that the choices of Ælfric and the Lambeth glossator spread to other writers. The use of edwist outside these writers, for example, shows up mainly in later texts, while the use of sped predominates in early texts. We
242
Don Chapman
have also seen similar hints in the use of forðstæppende for procedens. Byrhtferth, who used forðstop, came slightly later than Ælfric, and has been noted to share other lexical affinities with him. The term that does not agree with Ælfric and the Lambeth glossator comes from an early text, and as we have seen, later manuscripts of that text changed the word. Perhaps it is not merely coincidence that one of those changes was toward Ælfric’s and the Lambeth glossator’s choice of forðstæppende, which they may have been cultivating as a technical term. Ælfric and the Lambeth glossator share lexical tendencies we would expect to see if they were indeed creating and establishing technical terms. One tendency is to reduce polysemy. When a word has high polysemy and may be overloaded, Ælfric and the Lambeth glossator assign the technical meaning of the Trinitarian concept to a more specialized word, thus reducing the polysemy of the original English term. This can be seen readily in the case of substantia, which had several fairly different meanings, such as ‘wealth’, ‘success’, and ‘abundance’, as well as meanings closer to our own notion of substance, such as ‘essence’ or the stuff that something is made of. The Old English word sped seems to have had similar polysemy, as it was used for all those same senses of substantia. Perhaps the polysemy of sped made it less suitable as a technical term; at any rate, Ælfric and the Lambeth glossator chose a much less common and much less polysemous term to translate substantia as it is used in the Athanasian Creed. Similarly, it appears that Ælfric and the Lambeth glossator chose a less common forðstæppende as the technical term for procedens than other choices, especially forðgewitan, the choice of the Royal glossator. Forðgewitan had a variety of meanings compared to the few for forðstæppende. Ælfric and the Lambeth glossator also tended to choose a term that had more felicitous associations with the theological concept they were translating or glossing than did the glosses in other manuscripts. Besides the degree of polysemy that may have made sped a less desirable choice than edwist for substantia, it could be that the particular polysemous meanings proved unsuitable as well. An important core sense of sped is ‘wealth, success, abundance’. Its use for the more ethereal notion of essence – especially with reference to the Trinity – may have rung a little off-key for these writers. Similarly, other meanings of forðgewitan – ‘to pass away’ and ‘the past’ – may have made it less suitable as a term describing the relationship between the Holy Ghost and the other persons of the Trinity. Suitability may have been a factor in Ælfric’s and the Lambeth glossator’s choice of awendednes instead of gecyrrednes for conversio, since gecyrrednes often refers to a religious conversion made by humans, which runs opposite in associations to the change of nature spoken of in the Athanasian Creed. In Ælfric’s and the Lambeth glossator’s choice of underfangnes, we see not so much an avoidance of unsuitable associations as a felicitous choice in its semantic rhyming, as it refers to the condescenscion of God. One should be cautious about declaring particular choices felicitous, but the tendencies are at least pronounced enough to see in a
Trinitarian terminology in Old English liturgical creeds
243
number of words, giving us at least some hints that Ælfric and the Lambeth glossator were consciously seeking suitable vocabulary for technical terms. Ælfric and the Lambeth glossator thus attempted to reduce polysemy and enhance suitability along lines familiar from the specialization we see in some other Winchester words, particularly terms for ecclesia and corona. Apparently, when ecclesia refers to a building, the preferred term in the Winchester vocab ulary is ciric, but when referring to a more abstract sense of ecclesia as body of believers or an institution, the preferred word is gelaðung. Similarly, when a corona refers to a concrete crown, it is rendered as helm or cynehelm, but when it refers to a metaphorical crown, it is rendered as wuldorbeag (Hofstetter 1988: 146, 149–150; Gretsch 1999: 98–113). In both these instances, the polysemy of an English term (like ciric) is reduced by assigning a more spiritual or metaphorical meaning to another, less common word (like gelaðung). These same tendencies are easy to see with edwist / sped, and they may be present to a lesser degree with forðstæppende / forðgewitan or awendednes / gecyrredness. This study began by asking whether the terminology of the vernacular versions of the creeds in Anglo-Saxon England could give us evidence of writers’ ability to develop a technical vocabulary. I believe that the uniformity we see in these three groups, particularly the last group, indeed gives us evidence of such crafting. More research will likely uncover finer-grained correspondences and tendencies. More lexicographical treatment of all the terms, especially those from the Nicene Creed, may turn up more evidence and give us more insight into the relationships between Anglo-Saxon writers. But for now, the tendencies uncov ered in this study give us some confidence that at least some Anglo-Saxons were exercising control over the vocabulary to create a technical vocabulary.
Appendices: Sources of Old English Trinitarian vocabulary A1 Manuscripts containing glosses or translations of creeds Apostles’ Creed in Old English. Glosses ApDR1 DurRitGl 1 (Thomp–Lind) ApDR2 DurRitGl 2 (Thomp–Lind, C21.2) ApZ London, British Library, MS Royal 2.A.xx Apa PsCaJ (C11.4) London, British Library, MS. Arundel 60 Apb PsCaE (C11.2) Cambridge, Trinity College MS. R.17.1 Apc PsCaG (C11.7) London, British Library, Cotton Vitellius E.xviii Apd PsCaI (C11.11) London, Lambeth Palace 427 PsCaK (C11.12) Salisbury, Cathedral MS 150 Ape
244
Don Chapman
Close Translations ApJ Cambridge, University Library ms. Gg.3.28 ApK London, British Library, Cleopatra B.xiii ApM London, British Library, Cotton Tiberius C.i Nicene Creed in Old English. All these texts are translations by Ælfric. NI Lit 3.2 (B12.3.2) Oxford, Bodleian, Hatton 114 NJ Lit 3.2 (B12.3.2) Cambridge, University Library Gg.3.28 NK Lit 3.3 (B12.3.3) Oxford, Bodleian, Junius 121 Athanasian Creed in Old English. All these texts are glosses. A PsCaA (C11.6.1): London, BL Cotton Vespasian A.i; Canterbury xi1-2 C PsCaC (C11.1): Cambridge, University Library Ff.1 23 D PsCaD (C11.9): London, BL Royal 2.B.v; Glastonbury? Abingdon? xmed E PsCaE (C11.2): Cambridge, Trinity College R.17.1 G PsCaG (C11.7): London, BL Cotton Vitellius E.xviii I PsCaI (C11.11): London, Lambeth Palace Library 427; Winchester? xi1 J PsCaJ (C11.4): London, BL Arundel 60 K PsCaK (C11.12): Salisbury, Cathedral Library 150; Shaftesbury? x2 L PsCaL (C11.3): London, BL Add. 37517 Har PsCaHar (C11.8): London, BL Harley 863; Exeter xi3 A2 Prose passages containing Trinitarian language Text ÆAdmon 2 ÆCHom I, 1 ÆCHom I, 2 ÆCHom I, 4 ÆCHom I, 9 ÆCHom I, 10 ÆCHom I, 13 ÆCHom I, 15 ÆCHom I, 18 ÆCHom I, 20 ÆCHom I, 31 ÆCHom I, 33 ÆCHom II, 1 ÆCHom II, 3 ÆCHom II, 12.1 ÆCHom II, 25 ÆCHom II, 28
Lines 48–109 178.6–180.51 190.1–192.59 212.174–212.189 254.163–254.169 261.87–261.93 284.106–285.122 306.178–306.189 319.61–320.78 335.1–344.275 444.163–445.167 463.127–464.158 3.1–3.20 22.117–23.155 117.268–117.271 207.31–209.109 224.95–224.113 (Continued)
Trinitarian terminology in Old English liturgical creeds
245
A2: (Continued) Text ÆCHom II, 40 ÆHom 1 ÆHom 6 ÆHom 12 ÆHom 19 ÆHom 22 ÆHomM 1 (Bel 9) ÆHomM 8 (Ass 3) ÆIntSig ÆLet 2 ÆLet 4 ÆLS (Christmas) ÆLS (Denis) ÆLS (Memory of Saints) Bede 4 ByrM 1 (Crawford) CP GD 1 HomS 2 (ScraggVerc 16) HomS 8 (BlHom 2) HomS 10 (BlHom 3) HomS 34 (ScraggVerc 19) HomS 37 (Baz–Cr) HomU 34 (Nap 42) HyGl 2 (Stevenson) HyGl 3 (Gneuss) OccGl 54 (Zupitza)
Lines 301.65–301.84 375–393 111–122 1–20 206–217 12–28 1–29 20–77 69.504–69.541 54–63 36–36 13–16 30–46 94–119 18.308.34–20.314.1 2.9–2.20 15.95.15–15.97.5 9.59.31–9.60.34 129–171 85–99 85–109 1–12 1–34 11–12 62.1–62.18 38.1–38.9 12.60–12.90
Acknowledgments I am very appreciative of the help I have received from Miranda Wilcox, Michael Adams, Robert Fulk, Laurel Brinton, and Megan Hartmann.
References Banks, Ronald Alfred. 1968. A study of the Old English versions of the Lord’s Prayer, the Creeds, the Gloria, and some prayers found in British Museum MS. Cotton Galba A. XIV, together with a new examination of the place of liturgy in the literature of Anglo-Saxon magic and medicine. Ph.D. diss., University of London. Cameron, Angus, Ashley Crandell Amos, Antonette diPaolo Healey et al. (eds.). 2007. Dictionary of Old English: A to G online. http://tapor.library.utoronto.ca/doe/dict/bibl/index.html (accessed 25 November 2013).
246
Don Chapman
Campbell, Jackson J. 1951. The dialect vocabulary of the Old English Bede. Journal of English and Germanic Philology 50. 349–72. Chapman, Don. 2010. Uterque Lingua / Ægðer Gereord: Ælfric’s grammatical vocabulary and the Winchester tradition. Journal of English and Germanic Philology 109. 421–445. Gneuss, Helmut. 1985. Linguistic borrowing and Old English lexicography: Old English terms for the books of the liturgy. In Alfred Bammesberger (ed.), Problems of Old English lexicography, 109–129. Regensburg: Pustet. Gretsch, Mechthild. 1999. The intellectual foundations of the English Benedictine reform. Cambridge: Cambridge University Press. Grundy, Lynne. 1991. Books and grace: Aelfric’s theology. London: King’s College Centre for Late Antique and Medieval Studies. Healey, Antonette diPaolo (ed.). 1986–2004. Dictionary of Old English: A-G on CD-ROM. Toronto: Centre for Medieval Studies. Healey, Antonette diPaolo, John Price Wilkin & Xin Xiang (eds.). 2009. Dictionary of Old English Web Corpus. http://www.doe.utoronto.ca/pages/pub/web-corpus.html (accessed 1–25 November 2013). Hofstetter, Walter. 1988. Winchester and the standardization of Old English vocabulary. Anglo-Saxon England 17. 139–161. Kelly, J. N. D. 1964. The Athanasian Creed. New York: Harper & Row. Kitson, Peter R. 1997.Old English bird-names. English Studies 78. 481–505. Kitson, Peter R. 1998. Old English bird-names. English Studies 79. 2–22. Lendinara, Patrizia. 1999. Anglo-Saxon glosses and glossaries. Aldershot: Ashgate. Lewis, Charlton T. 1879. A Latin dictionary. Oxford: Clarendon. Morris, R. (ed.). 1967 [1874–1880]. The Blickling Homilies of the tenth century (Early English Text Society o.s. 58, 63, and 73). London: Oxford University Press, 1874–1880. Reprinted as one volume in 1967. Sauer, Hans. 1992. Towards a linguistic description and classification of the Old English plant names. In Michael Korhammer (ed.), Words, texts and manuscripts: Studies in Anglo-Saxon culture presented to Helmut Gneuss on the occasion of his sixty-fifth birthday, 381–407. Cambridge: Brewer. Sauer, Hans. 1999. Animal names in the Épinal-Erfurt Glossary. In Helen Conrad O’Briain, Anne Marie D’Arcy & John Scattergood (eds.), Text and gloss: Studies in insular learning and literature presented to Joseph Donovan Pheifer, 128–158. Dublin: Four Courts Press. Sauer, Hans. 2007. Old English words for people in the Épinal-Erfurt Glossary. In Hans Sauer & Renate Bauer (eds.), ‘Beowulf’ and beyond, 119–182 Frankfurt: Peter Lang. Schaff, Philip. 1877. Creeds of Christendom, with a history and critical notes. 3 vols. New York: Harper Brothers. Sisam, Celia & Kenneth Sisam (eds.). 1959. The Salisbury Psalter (Early English Text Society o.s. 242). Oxford: Oxford University Press. Skeat, Walter W. (ed.). 1881–1900. Ælfric’s Lives of Saints (Early English Text Society. o.s. 76, 82, 94, 114). London: Trübner. Reprinted in 2 vols. 1966. Temmerman, Rita. 2000. Towards new ways of terminology description: The sociocognitive approach. Amsterdam & Philadelphia: John Benjamins. Thorpe, Benjamin (ed.). 1844. The homilies of the Anglo-Saxon church. Vol. 1. London: Aelfric Society. Wenisch, Franz. 1979. Spezifisch anglisches Wortgut in den nordhumbrischen Interlinearglossierungen des Lukasevangeliums (Anglistische Forschungen 132). Heidelberg: C. Winter.
Part III Emerging Paradigms: New Methods, New Evidence
Alexander Bergs and Meike Pentrel
Ælc þara þe þas min word gehierþ and þa wyrcþ ... : Psycholinguistic perspectives on early Englishes 1 Introduction What do we know about speakers and hearers, or writers and readers in language history? In most cases, very little. There is no recorded spoken data before 1877, when Edison invented the phonograph. Even the amount of written data for certain types of speakers and styles is often very small and in any case limited. The further we go back in time, the worse this situation gets. There are virtually no written records composed by female English speakers before the fourteenth century, and certainly no records from the uneducated poor. This situation – described as the “bad-data problem” – is the result of several causes, which, according to Labov, include the fact that “historical documents survive by chance, not by design” and that they “are riddled with the effects of hypercorrection, dialect mixture, and scribal error” (Labov 1994: 11). As we have no chance of conducting any experiments, we also “know nothing about what was understood” (Labov 1994: 11). Therefore, any study of historical language stages needs to make “the best use of bad data” (Labov 1994: 11). This, however, need not always be the case. Romaine (1982) laid the foundations for a new field of research, historical sociolin guistics, which puts actual language users at center stage and focuses on the embeddedness of language in social context. One of its aims is to “crossfertilize historical linguistics with sociolinguistics in order to ‘use the past to explain the present’” (Romaine 1982: x). The present study sheds light on yet another hitherto neglected aspect of language users in history: what can we say about how actual speakers and hearers, writers and readers processed language? Can we assume that what we know about cognitive and psycholinguistic factors in language processing equally applied in the past? And what are the consequences if we find that there are differences between the present and the past? In order to answer these questions, this paper demon strates that cross-fertilizing cognitive linguistics, psycholinguistics and historical linguistics is not only possible, but a highly rewarding enterprise that can also be used to explain both the past and the present.
250
Alexander Bergs and Meike Pentrel
2 A theory of historical psycholinguistics 2.1 What is historical psycholinguistics? In this paper we present an approach which is based on the overlap of three different linguistic fields: cognitive linguistics, psycholinguistics and historical linguistics (see Figure 1 below). Some attempts to combine diachronic linguistics and cognitive approaches have already been made. Semantic change, metaphor and metonymy (e.g., Dirven 1985; Fabiszak 2001; Traugott and Dasher 2002), and grammaticalization (Langacker 2011; Bybee 2003) have recently been discussed from a cognitive point of view, within the framework of historical cognitive linguistics (HCL), as summarized, e.g., in Winters, Tissari, and Allan (2010). Labov (2010) also aims at combining cognitive and historical linguistics. With the approach presented here we try to show that psycholinguistic aspects can also be usefully integrated into historical investigations – despite obvious methodological differences. Psycholinguistics, with its roots in nineteenth-century experimental psychology, usually requires experiments with and data from live speakers (cf. Krause 2013: 67; McEnery and Hardie 2012: 193). By definition historical studies cannot be psycholinguistic in the strict sense, as there are no living subjects available. Yet it is clear that there is a fruitful and vital exchange between psycholinguistics and cognitive linguistics, as the two share fundamental assumptions: While experimental psycholinguistics is not usually considered a branch of functionalcognitive linguistics, its fundamental methodological assumption – that the nature of language in the brain or mind can be investigated in much the same way that experimental psychology in general looks at other aspects of the nature of thought – is in accordance with the general tenet of functionalism that there is no absolute divide between form and function, between language and non-linguistic cognition (McEnery and Hardie 2012: 193).
PsyLing
HPL CogLing
HistLing HCL
Figure 1: Historical psycholinguistics (HPL) as an overlap of psycholinguistics, cognitive linguistics and historical linguistics
Psycholinguistic perspectives on early Englishes
251
McEnery and Hardie (2012: 195) in particular point out the usefulness of corpus studies for both fields: [I]t may be argued that while such an experiment [eye-tracking of ambiguous words, with a literal and a metaphorical meaning] may indeed tell us something interesting about the processing of ambiguously metaphorical words, it cannot tell us about the normal processing of language in use. We can see, then, that a corpus-derived awareness of how words (and other linguistic items) are actually used can serve as a useful anchoring-point for psycholinguistic experimentation.
As historical linguistics is the corpus-based field of linguistics par excellence, we suggest that there is indeed an interesting overlap among all three fields, i.e., cognitive linguistics, psycholinguistics, and historical linguistics. This overlap is visualized in Figure 1. Here we see a methodological and theoretical overlap between cognitive linguistics and psycholinguistics and between cognitive lin guistics and historical linguistics, which forms historical cognitive linguistics. The overlap of all three disciplines constitutes what we term historical psycholinguistics (HPL). Note that a mixture solely of psycholinguistics and historical linguistics as such is excluded due to their incompatible methodologies (see above). In historical psycholinguistics, by contrast, corpus-based approaches form the common methodological basis for all three fields and hence allow for a productive new approach. Just as McEnery and Hardie suggest that corpus-based studies of actual language use can be a helpful anchoring point for psycholinguistics, we also assume that historical studies can give us new input for the development of our current psycho- and cognitive linguistic theories and methodologies. This creates a potential “feedback loop” between modern and historical studies. Thus, the aim of the present paper is twofold. First, we will discuss the inter esting feedback loop that develops when we not only apply modern theories to historical data but also critically reflect on our findings: both confirmation and disconfirmation can give us valuable insights into historical text production and into the validity of current theories. To that end, we will approach historical data with concepts and ideas from current psycholinguistic studies on sentence processing as our background. More precisely, we will look at cognitive strategies determining the structure of relative clauses in Old and Middle English.
2.2 Relative clauses in Middle and Old English: A case study Prideaux and Baker (1986) study the production and comprehension of PresentDay English (PDE) relative clauses in great detail and isolate several cognitive
252
Alexander Bergs and Meike Pentrel
processing strategies, e.g., CLOSURE (constituents should be closed as soon as possible) or ADJACENCY (antecedent should immediately precede relativizer). We will take these strategies, which make distinct predictions about the frequency and acceptability of different relative clause types, and compare them to relative clause structures found in Middle English letters and the Old English heroic poem Beowulf. We begin with the assumption that the processing principles guiding the production of certain grammatical structures in Present-Day English are also valid for the processing (i.e., production and comprehension) of such structures in historical texts. This assumption is based on the idea of the uniformitarian principle, discussed subsequently, which allows us to compare our findings with present-day studies, in order to determine the degree to which the data we find in these historical texts fit in with current theories. A confirmation of the hypothesis that the same processing strategies have been operative then and now means that despite several changes in media, i.e., conceptual orality versus literacy (cf. Goody 1987; Chafe 1994; Ong 2002; Clanchy 2009), historical text production/reception was no different from modern text production/reception. In other words, writing and reading are more or less the same neurocognitive activities then and now. At the same time, historical evidence confirming the proposed processing strategies underlines and strengthens the psycholinguistic/cognitive enterprise, as modern data are supplemented, supported and strengthened by historical evidence. This in turn means further support of the diachronic persistence of the processing principles in question and would thus open up a window on the mind of both modern and historical language users. Assuming that we find evidence for similar processing strategies in historical texts, we confront the “bad data problem” itself. Current processing studies may offer us new perspectives on how historical texts were actually produced and perceived. Almost every historical document brings with it a number of questions: Was it written by one scribe or several? Was it written during one sitting or several? Was it dictated? Was it translated? Is the text conceptually oral or written? Present-Day English findings are otherwise suggestive of whether the relativization strategies are more likely to be found in oral and informal texts or whether they belong to a more elaborate, written style. Studies on relative clauses in PDE texts (e.g., Bever 1970; Slobin 1973; Sheldon 1977) suggest different factors for the processing of these structures. While some studies (e.g., Slobin 1973) concentrate on syntactic constraints based on cognitive principles, such as NORMAL FORM (e.g., the preference for canonical word order), others focus on discourse-functional factors, highlighting for instance the role of Information Structure (e.g., Clark and Haviland 1977). These strategies make different, sometimes even opposing, predictions for the processing and realization of relative clause structures. The question which of the factors overrides the others, or rather is more influential,
Psycholinguistic perspectives on early Englishes
253
often depends on the text type and situational characteristics of it. We find, for example, different explanatory factors for oral and written text production and comprehension (more details below). Starting with the relative clause structure dominant in a historical text and knowing which structure relates to which processing/discourse strategy and which PDE text type, we may thus be able to deduce for instance the historical text’s mode of production. This can help us to shed light on how texts may actually have been designed (e.g., conceptually oral versus written). Testing hypotheses of relativization in Present-Day English on historical data and linking the results to what (little) we know about the historical text as well as to their cognitive underpinnings, we might be able to enrich both fields, historical linguistics and cognitive linguistics. Although there are no native speakers who can be subjects for experiments, we should be able to transfer the cognitive principles based on PDE data onto historical texts and textuality. Just as “sociolin guistic findings provide testable hypotheses for language historians” (Nevalainen 1999: 499), cognitive linguistics and psycholinguistics should provide equally testable hypotheses, assuming that cognitive abilities have hardly changed during the past centuries. This way, we have the new and interesting feedback loop in which modern theory instantiates historical data as evidence of certain linguistic and textual practices, and historical data in turn becomes evidence for modern theory (see Figure 2).
3 The basis: The uniformitarian principle Before discussing in more detail how we define processing strategies – or cognitive principles – we must address the theoretical basis for our research. We take Nevalainen’s (1999: 499) view “that human languages in the past were not in any principled way different from those spoken today” further by assuming that cognitive principles guiding the production and comprehension of certain
predict Historical Data
PDE Processing Strategies Literary Producon Theory
support or revise
Figure 2: Theoretical feedback loop
254
Alexander Bergs and Meike Pentrel
grammatical structures must have been equally valid for the production and comprehension of similar structures in historical texts. This assumption is based on the “uniformitarian principle” (cf. Lass 1980, 1997; Nerlich 1990; Bergs 2012). Originating in the natural sciences, this principle is often connected to Charles Lyell,1 who, according to the subtitle of his three-volume work Principles of geology (1830–1833), set out (…) to explain the former changes of the earth’s surface by reference to causes now in operation. W. D. Whitney, an admirer of Lyell (cf. Nerlich 1990: 31), was among the first lin guists2 to discuss uniformitarianism as a principle offering an approach to language change, stating that “the forces which are at work in it [our common speech] are the same now that they have always been, and the effects they are producing are of the essential character: both are inherent in the nature of language, and indispensable from its use” (1867: 24). Whitney further specifies that [t]he nature and uses of speech, and the forces which act upon it and produce its changes, cannot but have been essentially the same during all the periods of its history, amid all its changing circumstances, in all its varying phases; and there is no way in which its unknown past can be investigated, except by the careful study of its living present and its recorded past, and the extension and application to remote conditions of laws and principles deduced by that study (1867: 184).
Lass (1980: 53) sees uniformitarianism as a “preliminary answer to the question of where the foundation of a theory [about history] comes from: it comes from the present”, and the uniformitarian principle is one of the axioms that guide us when we come up with such theories. Lass (1997:28) sums this up as the General Uniformity Principle: No linguistic state of affairs (structure, inventory, process, etc.) can have been the case only in the past.
The problem, of course, is that there are only a limited number of entirely universal properties holding for these “linguistic states of affairs”. Lass (1980: 57) identifies one such universal: “It is, in our present state of knowledge, part of the definition of natural language that one of its properties is the presence of at least one high vowel in a system”. That is, any reconstruction of a language lacking
1 The geologist James Hutton, however, proposed this idea in one of his studies before Lyell (cf., e.g., Bergs 2012), and the term “uniformitarianism” was coined later by William Whewell in 1840 to oppose “catastrophism” (cf. Nerlich 1990). 2 Lyell, being a geologist, also discussed the development of languages in comparison to the development of species (see Nerlich 1990: 61–65).
Psycholinguistic perspectives on early Englishes
255
high vowels would have to be rejected (but cf. Fischer and Rosenbach 2000: 22 for a critical discussion of the notion of “synchronic uniformitarianism”). As there are only a few universals like this, we rather have to draw upon “tendencies” or “widely distributed properties” (Lass 1980: 57). A second principle (or weaker form of the General Uniformity Principle), the Uniform Probabilities Principle, guides us neither to expect nor accept any miracles and at the same time leaves a way open for unlikely, i.e., less probable, findings: The (global, cross-linguistic) likelihood of any linguistic state of affairs (structure, inventory, process, etc.) has always been roughly the same as it is now (Lass 1997: 29).
The fact that we have not encountered any living language without high vowels yet does not mean that it is a priori impossible that such a language existed or exists; however, it is very unlikely. In other words, “in cases of indeterminacy, the probable happened” (Lass 1980: 57). It is thus always “the best of our knowledge” which plays a crucial role in the assumptions we make: This is the sense in which we ought to take all claims about “universal” properties of human languages, preferred or most likely scenarios for change, etc. as guides to historical discourse. With the added proviso (which goes for all scholars) that “the best of our knowledge” usually means “the best of my knowledge”: in the hopes that I’ve managed to read widely enough so that the two will (largely anyhow) coincide (Lass 1997: 29).
In our approach, the uniformitarian idea suggests that the (universal) cognitive principles determining speech production and comprehension must be the same for all language stages – at least, that is the most probable scenario – from Old English to the present day. If cognitive principles are universal they should be valid for all speakers at all times.3 Hence, “the processes which we observe in the present can help us to gain knowledge about processes in the past” (Bergs 2012: 80). Winters clearly points out that [m]ost of the conclusions which have been drawn about cognitive function are completely synchronic, in part because all experimentation with speakers is carried out necessarily on contemporary forms of language and in part because the relationship between language and the brain is investigated in the great majority of instances by researchers who are not historians. Probably more important, however, is the fact that most if not all brain structure has been invariant over even the most extensive span of time considered in diachronic linguistics (to the extent that time depth can be accurately established), in fact since the evolution of other related species into homo sapiens. As a result of this pervasive cognitive invariance,
3 With the exceptions imposed by, e.g., text type, discourse situation, or interplay of opposing strategies making contradictory predictions for the same grammatical structure.
256
Alexander Bergs and Meike Pentrel
we must proceed with caution in exploring how to characterize language change in its relationship to cognitive processing. As far as we can tell through what direct evidence exists and through the judicious use of the uniformitarian hypothesis, humanity categorizes today as we have always categorized, in both linguistic and non-linguistic matters (Winters 2010: 16; emphasis added).
Following from this we begin with the null hypothesis that time has no effect on cognitive or psycholinguistic processing strategies. If strategies such as CLOSURE are found to be valid for certain structures today and grounded in the “cognitive make-up of the language user” (Prideaux and Baker 1986: 27), they should also have been operative in earlier English texts. It is thus improbable that different principles should have guided language processing given the same cognitive pre-conditions.
4 What are processing strategies? In a nutshell, producing a linguistic message involves several problems. Vague content must be turned into an appropriate semantic representation, which must then be arranged morphosyntactically – by taking into account the order of phrases or given and new information (based on assumptions about the listener’s knowledge). Finally, phonological processing results in some concrete output. The task of the receiver of a linguistic message, on the other hand, is to comprehend the message, which involves parsing and interpreting the concrete output in order to arrive at an identification of the intended message (cf., e.g., Clark and Clark 1977; Prideaux and Baker 1986; Levelt 1995; Garman 1996). Both production and comprehension can involve unconscious as well as conscious mental processes. Linguistic, or rather psycholinguistic, processing can be defined as “[t]he analysis, classification and interpretation of a stimulus (…) particularly used for the cognitive operations underlying (a) the four language skills (speaking, listening, reading, writing); (b) the retrieval of lexical items; and (c) the con struction of meaning representations” (Field 2004: 224; emphasis original). Processing can be seen either as a serial procedure in which one operation has to be completed before the other or as parallel processing. In the latter model “the user operates at several different levels of representation (sounds, words, semantics, syntax) at the same time” (Field 2004: 201). For Clark and Clark (1977), a process is the linking element between structure and function of language. Whereas structure is seen as “[t]he grammar of language” (8), and the function is “[a] description of how sentences communicate what they are meant to communicate”, a process is the “description of the mental tools, materials,
Psycholinguistic perspectives on early Englishes
257
and procedures people use in producing or comprehending” sentences (9). Formulating processes always means formulating hypotheses, as we can only deduce the strategies used to process or comprehend a linguistic structure; as Clark and Clark put it, “the structure and function of a sentence do not specify the process by which it was produced or understood, but they probably do give hints about it” (9). Our definition of what a processing strategy or principle constitutes draws in large part on Prideaux and Baker’s understanding of such a strategy. Mainly in accordance with the general definition above, their “strategy approach to language processing assumes the existence of a set of general strategies and it views production and comprehension as involving the interaction of languageindependent strategies with language-specific structural factors” (Prideaux 1984: 62). A processing strategy is seen as a procedure which is usually subconsciously and automatically employed by the speaker (in production) or the hearer (in comprehension) during language processing in order “to construct a meaning representation” (Prideaux and Baker 1986: 26). Strategies in this sense have a cognitive grounding and can often be assumed to be universal. Hence, a strategy is generally a mental operation independent from any specific language structures. In actual language processing, however, the knowledge of specific language structures has to interact with these mental operations and thus needs to be (subconsciously) known by the speaker or hearer (cf. Prideaux and Baker 1986: 27). Highlighting this aspect, Prideaux and Baker refer to their approach to processing as “the ‘cognitive strategies’ framework” (Prideaux and Baker 1986: 5). A particular strategy may predict a certain sentence structure to be more likely, i.e., more frequent, than an alternative structure with functional equivalence (cf. Lavandera 1978): If two or more alternative structures representing the same basic meaning are available and if a given principle predicts that one is easier to process (comprehend or produce), then, every thing else being equal, the easier structure should be more frequent (Prideaux 1989: 28).
For the present study this means that the processing strategies identified by Prideaux and Baker (see below) make predictions about the frequency and acceptability of different relative clause structures with regard to ease of production and comprehension. Prideaux and Baker distinguish among three different levels of strategies: cognitive (i.e., language-independent) strategies, language-specific strategies, and discourse strategies. For each of these levels, a number of differ ent strategies are developed (and tested) on the basis of previous research (e.g., Bever 1970; Slobin 1973; Kimball 1973; Frazier and Fodor 1978; Frazier 1979; Slobin
258
Alexander Bergs and Meike Pentrel
and Bever 1982). Language-independent, sentence-level strategies include (Prideaux and Baker 1986: 32): –– CLOSURE. “In processing a particular linguistic unit (phrase, clause, etc.), the language user (speaker or hearer) attempts to obtain closure on that unit as early as possible”. –– NORMAL FORM. “The language user assumes that the unit being processed is in its ‘normal’ or ‘canonical’ form unless the unit is overtly marked to the contrary”. –– NON-AMBIGUITY. “The language user assumes that the unit being processed is not ambiguous”. –– BRACKETING. “The language user assumes that when a new unit for processing is encountered or initiated, it will be marked as such”. Language-specific strategies on the sentence level which directly involve relative clauses include (Prideaux and Baker 1986: 40–41): –– PARALLEL FUNCTION (based on Sheldon 1974). “Comprehension for sentences containing relative clauses is facilitated if the relative pronoun plays the same grammatical role (subject or object) as is played by the modified noun”. –– ADJACENCY (based on Sheldon 1977). “In parsing a noncompound sentence, start from the left and group together as constituents of the same clause two adjacent NPs (i.e., those not separated by another NP) and an adjacent verb not already assigned to a clause. Interpret the first NP as the subject and the second NP as the object of the verb”. Finally, as a discourse strategy Prideaux and Baker mention (1986: 43, based on Chafe 1970): –– GIVEN-NEW. “Given information is normally separated from New information in a systematic way. Typically, Given information precedes New information, although New can precede Given when special linguistic devices are employed to signal the atypical ordering of Given-New”.
5 Strategies for the processing of relative clauses in Present-Day English In the following, we will limit our analysis to two of the more central languageindependent strategies discussed by Prideaux and Baker, i.e., CLOSURE and NORMAL FORM, and the two language-specific strategies, i.e., ADJACENCY and PARALLEL FUNCTION. NON-AMBIGUITY, BRACKETING and the discourse-level strategy GIVEN-NEW were not evaluated. BRACKETING becomes relevant only
Psycholinguistic perspectives on early Englishes
259
when clauses with and without relativizers are compared (and the latter were not considered in the present study). NON-AMBIGUITY and GIVEN-NEW would require not only a discussion and definition of GIVEN and NEW, but also a detailed analysis of the individual occurrences, which would require a qualitative rather than a quantitative study. Among the sentence-level strategies, we find one of the best-known, best-understood strategies: CLOSURE. CLOSURE is a language-independent strategy expressing the tendency that listeners (as well as speakers) expect a sentence, clause or phrase to be “closed” as soon as possible. All elements needed for the unit to be complete should thus be encountered without delay. This strategy is closely connected to constraints imposed by short-term memory, particularly in spoken language. Keeping a proposition “open” and unresolved for a long time can be very strenuous and brings with it the danger of losing the thread of the message. Knowing the English SVO (subject-verbobject) sentence structure, a hearer will expect a verb to follow upon hearing the first subject NP (noun phrase, as in 1a below); an interruption of this structure by, for instance, a relative clause on the subject NP is unexpected and perhaps even delays closure unnecessarily, as in (1b) (Prideaux and Baker 1986: 32–34). Hence, the CLOSURE strategy predicts that an interrupted sentence structure would be harder to process than a non-interrupted one (Prideaux and Baker 1986: 53). Example (1a) is thus easier to process than (1b), and should therefore also be more frequent: (1) a. He is playing the role of a man [who just can’t keep his homophobia to himself]. (COCA, ABC Primetime 2012 [120127]) b. And I think everyone [who watched the debate] was surprised by that. (COCA, ABC This week 2012 [120108]) The second relevant strategy, NORMAL-FORM, operating on the sentence level, states that listeners expect that the linguistic unit they are processing will be in its canonical (most common and least marked) form (Prideaux and Baker 1986: 34). For a non-canonical from, then, overt marking facilitates processing. The NORMAL-FORM strategy would predict again that the first example is easier to process than the second, as (2a) has an unmarked, canonical SVO order in both matrix and relative clause, whereas (2b) has SVO in the matrix clause, but noncanonical OSV (object-subject-verb) in the relative clause. You’ve got people [who are very excited about Rick Santorum]. (COCA, ABC (2) a. This week 2012 [120108]) b. Now I am finally taking a stand on behalf of two women [who I consider to be national treasures]. (COCA, NBC Today 2012 [120319])
260
Alexander Bergs and Meike Pentrel
One of the language-specific strategies discussed by Prideaux and Baker is PARALLEL FUNCTION (originally proposed by Sheldon 1974), which predicts that a sentence containing a relative clause is easier to process if the grammatical function of the relative pronoun and the antecedent are identical, i.e., when both have the role of subject or object in the two clauses. In (3a) both the antecedent, teens, and who in the relative clause each function as the subject in their respective clauses. In (3b), however, the antecedent somebody functions as the object of the matrix clause, whereas who functions as the subject: (3) a. (…) teens [who talk back] may have a brighter future. (COCA, NBC Today 2012 [120125]) b. They want to see somebody [who looks more like a commander-in-chief than a candidate], (…). (COCA, ABC This week 2012 [120122]) The parallel function strategy has been heavily criticized, as Sheldon’s original results could not be replicated, and its explanatory power is in doubt, “since it follows from no general syntactic or cognitive principles” (Prideaux and Baker 1986: 41). Sheldon therefore suggested another strategy which could explain her results. The ADJACENCY strategy states that the parser during processing, proceeding from left to right, interprets the first of two adjacent NPs as the subject and the second as the object of an adjacent verb (relative pronouns are ignored deliberately). Therefore, in (4b) the strategy would correctly assign the subject role of the verb caught to the man, and the object role to the thief. Simultaneously, the thief is also (wrongly) assigned the subject role of the verb receive and the reward the object role. The higher the number of such misalignments, the higher the processing difficulty. In (4a) the assignment of subject and object for both verbs does not result in any misalignment of syntactic functions (I is the subject of love, the man its object; the man is the (semantic) subject of caught, the thief its object), which means that this structure should be easier to process: (4) a. I love the man [who caught the thief ]. b. The man [who caught the thief ] received the reward. (Prideaux and Baker 1986: 41) Table 1 summarizes the proposed strategies and the predictions they make for the structure of the relative clause. Note that some of the strategies make contradictory predictions about the optimal sequence of subject and object. In particular, language-universal sentence-level strategies and language-specific strategies seem to be at odds here.
Psycholinguistic perspectives on early Englishes
261
Table 1: Summary of relative clause strategies Strategy
Description
Preferred structure(s)
CLOSURE NORMAL FORM
obtain closure as early as possible the unit is in its “normal” or “canonical” form
OS, OO4 SS, OS
PARALLEL FORM
same grammatical role of relative pronoun and the modified noun first NP = subject and second NP = object of the verb
SS, OO
ADJACENCY
OS > SS, OO > SO
OO = Object (O) Antecedent before Subject Relativizer (S) OS = Object Antecedent (O) before Object Relativizer (O) SS = Subject Antecedent (S) before Subject Relativizer (S)
Prideaux and Baker (1986: 111–126) investigated these and the other processing strategies on the basis of two different lines of empirical research. On the one hand, they devised five traditional subject-based psycholinguistic experiments (two of which worked with written language stimuli and no time pressure, whereas three were more concerned with actual on-line processing). On the other, they conducted corpus-based “text count studies” with a small corpus of written material, i.e., five different written text samples from three different genres (novel, non-fiction entertainment, and academic text). In two out of these three subject-based experiments CLOSURE did not appear to be operative. Prideaux and Baker claim that this may have to do with the fact that the stimuli were presented in written form. CLOSURE as a strategy is related to short-term memory. When subjects can think about the (written) sentences as long as they wish, there is no need for CLOSURE to be operative. In those cases, other strategies such as NONAMBIGUITY and NORMAL FORM seem more important. CLOSURE, however, was found to be operative both in spoken language tests and in cases of complex multiple embeddings (even in writing). NORMAL FORM, on the other hand, was found to be important in relative clauses with reversible structures (e.g., The man watched the horse which followed the cow). PARALLEL FUNCTION and ADJACENCY could not be shown to be operative in any of the experiments. In the corpus-based, written (“text count”) studies Prideaux and Baker found that CLOSURE and NORMAL FORM as strategies were operative in all samples. At first sight, this may sound contradictory to their findings from the experiments. However, upon close inspection it turned out that CLOSURE and NORMAL FORM were most visible in fictional texts associated with, mirroring, or even representing spoken language. Prideaux and Baker suggest that this, again, may have to do with the reliance on short-term memory, which is important in spoken rather than written language
262
Alexander Bergs and Meike Pentrel
(1986: 121). And again, ADJACENCY and PARALLEL FUNCTION were found to be irrelevant across the board (Prideaux and Baker 1986: 126, 146, and 152). In sum, Prideaux and Baker stipulate that many of the results from the text samples are due to the dyadic relationship of author and reader, and that we see something like audience design in a broad sense. In other words, authors probably design their texts (consciously or unconsciously) in ways that can and should be processed and understood on particular terms by individual readers.
6 Middle English In the following, we will supplement Prideaux and Baker’s study with data from the late Middle English Paston letters and papers, a group of 421 documents (with approximately 245,000 words in total) written between 1421 and 1503 by various members of the Paston family in Norfolk (see Bergs 2005 for details) and edited by Davis (1971). Some of the texts in this collection are more formal, others more informal. We also find a few indentures, wills, recipes and memoranda. While we do have quite a lot of information on the social background of the family and other interesting extra-linguistic factors, we know comparatively little about how the letters were actually written, or how letters were written generally in that period (cf. Bergs 2013, forthcoming). Were they dictated? Autographed on the spot? Drafted beforehand? How much planning went into their production? Present-day personal letters, emails and short messages often are rather conceptually oral. Does this also hold for late Middle English personal letters? Middle English syntax does not differ significantly from Present-Day English syntax. Of course, do-support, modal auxiliaries and certain tense/aspect structures were still not (fully) established by 1500. We find remnants of Old English grammatical morphology and some variability in word order. Nevertheless, late Middle English already shows some kind of basic or canonical word order, i.e., SVO (e.g., Fries 1940). All relative clauses in the Paston letters and papers were manually searched and tagged for the syntactic functions of the antecedent and the relativizer. Examples (5) through (8) illustrate the four different structures, i.e., subject antecedent – subject relativizer (SS), subject antecedent – object relativizer (SO), object antecedent – subject relativizer (OS), and object antecedent – object relativizer (OO): (5) SS: (…) and som [ þat are greete klerkys and famous doctorys of hys] goo now ageyn to Cambrygge to scoolle. (1472, John Paston II, no. 268) (6) SO: Item, that the horse [ þat Purdy hathe of myne] be put to some good gresse in hast. (1471, John Paston III, no. 346)
Psycholinguistic perspectives on early Englishes
263
(7) OS: Also that he delyuer the remenant off the mony [that is fonde in hys hande] (…). (1472, John Paston II, no. 268) (8) OO: (…) I see nothere hym ne the letters [þat he schuld haue brought] (…). (1473, Margaret Paston, no. 220) In order to arrive at the best compatibility and comparability between the results here and those of Prideaux and Baker, we replicated Prideaux and Baker’s method and counted only subject and object (including direct and indirect) functions in restrictive relative clauses. Non-restrictive clauses and all other functions were not considered. Figure 3 contains data on restrictive relative clause constructions in the Paston letters and papers, with the syntactic functions of antecedent and relativizer. Out of more than 1,200 restrictive relative clauses, about 40% had an OO, followed by ca. 37% OS, 16% SS and only about 7% SO. This means that the ranking in the letters is OO and OS more or less on a par, followed by SS, followed by SO. How do these results compare with those of Prideaux and Baker? Obviously, the data from the late Middle English Paston letters and papers do not correspond completely with any of the predictions made by Prideaux and Baker (1986) (see Table 1 above). They come closest, though, to what is predicted by CLOSURE (OS and OO as preferred strategies that allow early completion of constituents). The interesting point now is that CLOSURE is associated with short-term memory, as has been discussed above. The longer constituents are left open, the more information has to be stored in short-term memory until finally closure has been reached. This can be shown quite easily. Consider an example sentence such as John and his team of fifteen of the best chocolate cake experts in the world, who worked for our company for more than fifty years and have now decided to move on and concentrate on cheesecakes,
Figure 3: Restrictive relative clause constructions in the Paston letters and papers, with the syntactic functions of antecedent and relativizer (S [Subject], O [Object])
264
Alexander Bergs and Meike Pentrel
will host a farewell party next Monday. Here we see an SS structure; the matrix clause is interrupted by the relative clause. The speaker/hearer has to keep the proposition open until the matrix clause is completed – in this case for 21 words and several clauses, which can cause trouble for any on-line processor. CLOSURE is a strategy that reduces the cognitive workload. As a consequence, Prideaux and Baker argue, we find structures that adhere to CLOSURE most frequently in comprehension and production in spoken on-line tests, with time limits, and also in writing when multiple embeddings occur. It may also be speculated that CLOSURE can be used on purpose by speakers and writers (as a kind of audience design strategy) in order to make processing easier for their addressees, since CLOSURE also allows hearers to construct propositions rapidly and reliably. NORMAL FORM (SS, OS) accounts only for parts of the data, as does PARALLEL FUNCTION (SS, OO). Both fail to account for the high number of OO and OS strategies, respectively. ADJACENCY (OS > SS, OO > SO) obviously also does not play any key role in the distribution, as it explains only parts of the overall picture. It should be noted, though, that the Paston letters and papers, like our modern data, show a very low frequency for SO, which can be considered the most complex construction, as it usually requires an interrupted matrix clause and reversed constituent ordering in the relative clause. In fact, Romaine (1984: 456) argues that SO was impossible in Old English. Thus, the very low frequency of this order in late Middle English is unsurprising. While results from the Paston letters and papers do not fit perfectly with any of the predictions for Present-Day English discussed above, they correspond interestingly with what Baker and Prideaux have found in their study of spoken and written texts. They found CLOSURE to be the most important strategy, operative in all texts. However, it was most often used in fiction, in representations of spoken language. There is some reason to assume that most of the Paston letters and papers were actually dictated (cf. Bergs 2013, forthcoming), and therefore are char acterized by memory constraints that affect spoken, not written language. In order to explore this effect, we can compare letters from the collection which we know on independent grounds were autographed (and hence conceptually written) with those that were dictated (and hence conceptually spoken). Out of the 421 documents, 231 (ca. 55%) with ca. 145,000 words (= ca. 59%) were written by various scribes, 190 (ca. 45%) with ca. 100,000 words (= ca. 41%) were autographed. The results for autographed versus dictated letters are as in Figures 4 and 5. It turns out that the general distribution of the different structures is in both cases roughly the same as in Figure 3 above, i.e., OO is most common, followed by OS and SS, and finally SO. However, we also see a slight tendency for OO and OS to be more common in dictated letters. Taken together, they account for
OS 34%
Psycholinguistic perspectives on early Englishes
265
SS 18%
SO 9%
OO 39% Figure 4: Restrictive relative clause constructions in autographed texts
SS 15% SO 6%
OS 39%
OO 40% Figure 5: Restrictive relative clause constructions in dictated texts
ca. 79% of the restrictive relative clauses in dictated letters versus ca. 73% in autographed letters. This is in line with CLOSURE as a strategy and might indicate that letters were often (though probably not always) dictated “on-line”, i.e., with the author dictating the text and the scribe actually writing the text while the author spoke (rather than, for example, giving the scribe a rough sketch or keywords for a letter which the scribe completed without the author; cf. Bergs 2013, forthcoming). This kind of “on-line” dictation puts them closer to conceptually oral language. Hence, they were subject to memory constraints on the part of the author and the scribe and resemble what we find in spoken language. Also, dictated letters show the lowest frequency for the complex SO structure. This structure not only has an interrupted matrix clause, but the relative clause also exhibits non-canonical word order. The use of this structure is something
266
Alexander Bergs and Meike Pentrel
we expect in more formal, carefully planned complex writing styles. The avoid ance of this complex structure in the Paston letters and papers may have to do with audience design, i.e., with the use of language with an addressee in mind. In other words, these letters seem to be composed in a reader-friendly style with low complexity. Consequently, the high use of CLOSURE and the low use of SO structures in both dictated and autographed letters can lead us to the conclusion that (personal) letters in late Middle English were essentially an oral text type, and were written in a reader-friendly way according to Gricean principles (e.g., be orderly and clear, be relevant, only say as much as is required). Their aim usually was to convey information quickly and reliably, and this, despite their sometimes clumsy and formulaic style (see below), is what we see in the linguistic structures that were used. This analysis is supported by the fact that the results seem to shift when we differentiate between personal letters and other (formal) text types, such as testaments, indentures, petitions (as in Figures 6 and 7). If we exclude “letters” as one text type from our considerations and focus on the more formal genres found in the Paston documents (wills, memoranda, indentures, petitions, etc.) we see that in the latter more than 50% of the relative clause structures are OS, 26% are OO, 16% are SS and only ca. 6% are SO. Again, we see CLOSURE as very important, but NORMAL FORM (which also favors OS) seems to be strong as well. At the same time, SS structures appear to be more common in more technical, formal genres. In contrast, letters show 42% of OO, followed by 35% of OS, 16% of SS and only 7% of SO. The battle seems to be raging between OO (letters) and OS (more formal genres). The results discussed above need to be taken with a grain of salt, however. Davis (1965, 1967) has already shown that formulaic language played a great
OS 35%
SS 16% SO 7%
OO 42% Figure 6: Restrictive relative clause constructions in personal letters
Psycholinguistic perspectives on early Englishes
267
SS 16% SO 6%
OS 52%
OO 26% Figure 7: Restrictive relative clause constructions in formal texts (excluding letters)
role in the composition of the Paston letters and papers. In other words, many phrases and sequences in the letters were not necessarily composed word by word, but were available as more or less fixed chunks, not unlike modern letter writing phrases such as “best wishes” or “we are writing to inform you …”. This means that we will have to discuss the extent to which the results obtained above are actually based on word-by-word processing. Rather, they could be artifacts of fixed formulae that were employed in a number of writing styles. However, rather than contrasting these two aspects – free composition versus fixed formulae – as binary opposites, we would like to suggest that both can be captured with a more recent approach to language processing. A few decades ago, language processing was essentially seen as an “assembly line”: different elements are put together serially and in real time to eventually form a whole. Today, some alternative language models, such as lexico-grammar, construction grammar, and emergent grammar share a different perspective on language processing: “Linguists seem to underestimate the great capacity of the human mind to remember things while overestimating the extent to which humans process information by complex processes of calculation rather than by simply using prefabricated units from memory” (Lamb 1998: 16), and “much of our entirely regular input and output is not processed analytically, even though it could be” (Wray 2002: 10). While the present paper cannot explore such models in any detail (cf. Hoffmann and Trousdale 2013; Wray 2002, for details), we would still like to suggest that some part of processing deals with formulaic or fixed expressions, which are processed as chunks and not as sequences of individual elements and therefore are not subject to the same constraints and strategies. For example, one of the most frequent letter-closing formulae in the Paston letters and papers is by þe grace of God who haue yow in hys kepyng (Margaret to John Paston 1462, 11, 01). Here we find an oblique (prepositional) object followed by
268
Alexander Bergs and Meike Pentrel
the relative clause who haue yow in hys kepyng. This part of the complex formula is a very common phrase in the letters and is used in a number of different contexts, such as God bryng vs oute of it, who haue yow in his kepyng (Margaret to John Paston II 1470, 07, 15), where we do not have a prepositional object, but rather a subject (with following complement structure) and the relative clause following this. We argue that this is not necessarily an example of on-line processing as discussed above, but rather of the formulaic nature of certain aspects of language. In other words, the interrupted SS structure in the second example might not be a counterexample to ease of processing as a factor, as it might be the result of formulaic rather than “micro-sequential” (word by word) processing. This, however, is a question that cannot be dealt with in detail here, but which warrants future research. Again, rich historical data from individual speakers and their personal language use can help us to shed more light on more general issues of psycholinguistic language processing. Current debates on holistic versus piecemeal processing could be complemented not only by present-day experimental data, but also by detailed studies on formulaic language use in historical contexts. Eventually, we expect the kind of feedback loop discussed at the beginning of this paper: while current theoretical debates on holistic versus individualistic processing provide interesting hypotheses for historical linguistic research, actual findings (and problems!) in historical studies can be used for refining current theoretical work.
7 Old English In this section we will go back further in time and look at Old English data in order to explore whether the kind of approach presented here can also be useful at earlier language periods with even more dramatic gaps in the data. The database is the Old English heroic poem Beowulf. A comparison of Beowulf and the Paston Letters and Papers is interesting for at least two reasons. We know very little about the exact nature of Beowulf with respect to conceptual orality and literacy. In other words, there is an on-going debate about whether Beowulf was composed in the written or spoken mode (cf. Orchard 2003: 85–86, and references therein). An analysis from the viewpoint of language processing might shed some light on this question. Second, the Paston letters and papers apparently belong to the more informal, spoken registers. These are not readily available for earlier periods of English. Old English prose texts are usually conceptually written. Epics, such as Beowulf, however, might be a little bit closer to being conceptually oral, as they border at least on what is known today as “oral poetics”.
Psycholinguistic perspectives on early Englishes
269
Unfortunately, very little is known about the provenance of Beowulf as a text – both author and year of composition are unknown. The only extant manuscript dates from about the millennium (the late tenth or early eleventh century), but scholars usually agree that much of the text must have been written earlier, maybe in the eighth or ninth century, probably somewhere in the Midlands (see the extensive and illuminating discussion provided in Fulk, Bjork, and Niles 2008: clxii–clxxxvii). The poem contains about 17,000 words in total. If we want to apply Prideaux and Baker’s strategies to Old English data, we have to keep a few things in mind. Linguistically, Old English is very different from both Middle and Present-Day English. One important difference, for example, is the set of relativizers. In Old English we usually find three different types of relativizers: an indeclinable particle þe, which was very common in subject and object function, the declinable demonstrative se, and a combination of both, i.e., declinable se plus indeclinable þe, se þe. Moreover, Old English had more relaxed constraints on the position of the verb (Fries 1940; but cf. Molencki 2012: 295–297; Mitchell 1985: Vol. II, 957–962 for a more detailed discussion), which means that it lacked canonical word order in the modern sense, at least in poetry. So simply tagging for Subject and Object does not necessarily mean that we can draw any conclusions about the ordering of these elements, as we can usually do in PresentDay and even late Middle English. As a consequence, a language-specific strategy such as NORMAL FORM, for example, does not necessarily apply, since there is no canonical word order. Also, ADJACENCY, a language-specific strategy from Present-Day English, cannot apply in Old English, quite simply because Subject and Object function are not assigned by default through left-to-right ordering. Thus, the only factors we can reliably check in Old English at this point are CLOSURE and PARALLEL FUNCTION, the first of which is said to be a universal, cognitively motivated strategy, the second one language specific, at least for Present-Day English. Just as in Prideaux and Baker’s study and in our analysis of the Paston letters and papers we will restrict our analysis to restrictive relative clauses. In the ca. 17,000 words of Beowulf, we find about 141 relativizers in restrictive clauses. These were tagged as in Figure 8 (note that double tagging as both CLOSURE (OS, OO) and PARALLEL (SS, OO) may be possible: see below). In the PARALLEL FUNCTION group we find 73 SS structures, and 22 OO, which means there is some overlap with CLOSURE, which also favors OO. There were only about 7 relative clause structures which followed neither CLOSURE nor PARALLEL FUNCTION (e.g., SO). 28 relative clauses have not been considered, as they showed other syntactic structures and functions, such as possessives. The most obvious and interesting difference between the two results is that in Beowulf, both CLOSURE and PARALLEL FUNCTION seem to have been operative more or less to the same degree. Prideaux and Baker (1996: 122) found PARALLEL
270
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Alexander Bergs and Meike Pentrel
27
33
27
23
se Þe se Þe 32
39
Closure
Parallel
Figure 8: Types of relativizers in Beowulf
FUNCTION to be almost completely absent from their fiction set of texts (and practically all of their experiments), where CLOSURE and NORMAL FORM dominated. This is actually supported by the fact that few other studies, apart from Sheldon (1974, 1977), who focused on language acquisition, found PARALLEL FUNCTION to be important. CLOSURE, as has been discussed above, on the other hand, is generally considered to be a very important strategy associated with memory effects in on-line language processing. This in turn supports the idea that Beowulf could essentially have been produced within an oral tradition. What is interesting here, however, is the surprisingly large number of instances of PARALLEL FUNCTION. On one hand, this may be an artifact of the data. Since word order in Old English poetry was less constrained, we can get sentences with both PARALLEL FUNCTION and CLOSURE operating at the same time, both with subjects and objects: (9) (…) Sægenga bad age[n]dfrean, se þe on ancre (…) Sea-goer (ship) [S] waited for owner and lord, which [S] on anchor. rad. (Beowulf 1882–1883) rode. ‘Riding at anchor, the ship waited for her master.’ (10) Gemunde ða ða are þe he him ær forgeaf. (Beowulf 2606) Recalled then the properties [O] that [O] he him formerly gave. ‘He recalled the favors that he had formerly given to him.’ In both these examples you see CLOSURE and PARALLEL FUNCTION at work. In (9) we find both the antecedent and the relativizer in subject function, and the matrix clause is actually closed before the relative clause even begins. In (10) we find antecedent and relativizer in the object function and, again, the matrix
Psycholinguistic perspectives on early Englishes
271
clause is completed before the relative clause begins. On the other hand, there are also cases where PARALLEL FUNCTION operates independently:
se þe ær feorran com, (11) Hæfde þa gefælsod Had then cleansed he [S] who[S] before from afar came snotor ond swyðferhð, sele Hroðgares, genered shrewd and strong-minded the hall of Hrothgar rescued wið niðe (…) (Beowulf 825–827) from ruin (…) ‘He had then cleansed, he who came before from afar, shrewd and strongminded, the hall of Hrothgar, rescued from ruin …’ This issue, which cannot be pursued further here, might inspire further research: does the apparently higher frequency of PARALLEL FUNCTION in early literary texts like Beowulf stem solely from language-internal factors, or does it correlate at least to some extent with stylistic factors? Do, for example, some formulaic structures of oral poetry lend support to seemingly dispreferred cognitive strategies, just as some formulae in the Paston letters and papers have obscured our picture? Should they, therefore, simply be discounted in our considerations, or should they be integrated into our modern theorizing in psycholinguistics and cognitive linguistics? None of the studies available so far – not even from a typological perspective – seem to suggest that PARALLEL FUNCTION is a central strat egy today. Also, how do we deal with the fact that there are far more occurrences of SS in PARALLEL FUNCTION than OO, as in the Paston letters and papers? This pattern might suggest some language-internal or information-structuring principle at work here, but this needs further investigation. In any case, the obvious presence of a strategy such as PARALLEL FUNCTION in historical texts – but not in contemporary studies – might imply either that the uniformitarian principle needs to be taken with a grain of salt, or that modern processing theories need be reconsidered.
8 Conclusion In this paper, we presented some relative clause processing strategies suggested in Prideaux and Baker (1986). These were, on the general, cognitive level, CLOSURE and PARALLEL FUNCTION, and on the language-specific level NORMAL FORM and ADJACENCY. We investigated two historical data sets in which all relative clauses were manually searched and tagged appropriately. These were the late Middle English Paston letters and papers on the one hand, and the Old English poem Beowulf on the other. It was shown that the two differ not only in the type
272
Alexander Bergs and Meike Pentrel
of data as such, but also in the details of the linguistic material they present. Late Middle English, as is found in the Paston letters and papers, is similar enough to be equated with Modern English, at least with regard to syntax, so that all four strategies could be considered. Old English, on the other hand, is linguistically so different that the language-specific constraints had to be ignored. In the Paston letters and papers a rather diffuse picture regarding the differ ent strategies emerged, which suggested that, since CLOSURE was the predominant strategy for processing relative clauses, those letters – and perhaps letters of the period generally – were essentially of an oral nature and were composed with a certain readership and reader behavior in mind. At the same time, the high frequency of CLOSURE also suggests that Beowulf, like the Paston letters and papers, was not a highly complex, conceptually “written” text, but rather something associated with an oral tradition, as CLOSURE is mostly associated with factors such as memory constraints on the speaker and hearer (which does not preclude, of course, that Beowulf is characterized by certain features of formal poetic art and corresponded with literary patterns; cf. Kendall 2006). Digging further we might ask ourselves what this means for the production and interpretation of literary texts. It might be speculated that the linguistic form of a literary text can give us some very interesting clues regarding its production and its possible interpretations. Beowulf, just like the Paston letters and papers, was probably a text which was meant to be read aloud and heard instead of being read (and re-read) in private. This kind of literature cannot employ more complex structures, as it needs to be understood, at least in parts, spontaneously, without a chance of rereading or repeating any difficult passages. So its aesthetic, literary effects apparently are associated with an oral, on-line style, and this needs to be taken into account. Needless to say, the same applies to socio-historical interpretations of the Paston letters and papers. Eventually, what might become visible here is the fact that linguistic forms and constructions can be symptoms of (literary) text production and modes of (literary) text reception – often mediated in particular genres. Pending such further interpretation and investigation of the data, we can, for the time being, establish preliminary versions of the feedback loops in which historical data challenges the processing strategies established for Present-Day English. Figures 9 and 10 illustrate the compatibility of the historical data and the relevant processing strategy CLOSURE. The results support this psycholinguistic strategy, which again backs up the assumption that such a strategy is universal and thus should, in the light of the uniformitarian principle, not be restricted to a specific language stage. In Beowulf, however, CLOSURE and PARALLEL FUNCTION were remarkably balanced processing strategies. In PARALLEL FUNCTION, SS Structures were
Psycholinguistic perspectives on early Englishes
273
on-line dictaon Paston Leers
CLOSURE Strategy
Historical Data
Literary Producon Theory
support
Figure 9: Feedback loop for the production theory of Paston letters and CLOSURE strategy
origin: oral tradion Beowulf
CLOSURE Strategy
Historical Data
Literary Producon Theory
support Figure 10: Feedback loop for the production theory of Beowulf and CLOSURE strategy
particularly common; while this might be an artifact of data – e.g., the effect of interferences from formulaic language use, or gaps in the textual coverage – it seems nevertheless worthwhile to investigate this issue further. Figure 11 illus trates the feedback loop we predict between our Beowulf data and the PARALLEL FUNCTION principle. Future research on Old English will have to investigate whether PARALLEL FUNCTION turns out to be a robust strategy for this period, irrespective of its use in formulaic constructions (which obviously also deserve more detailed investigations in the future). If this is the case, then another issue will obviously have to be discussed, namely why PARALLEL FUNCTION effects have not been observed in Present-Day English. Potentially, these findings might also require a revision of
origin: oral tradion Beowulf
PARALLEL FUNCTION
Historical Data
Literary Producon Theory
revision? Figure 11: Feedback loop for Beowulf and PARALLEL FUNCTION strategy
274
Alexander Bergs and Meike Pentrel
modern theories on the persistent nature of this cognitive processing principle. It is tempting to assume that human cognition has been stable for many centuries, or even millennia. This only follows from the uniform probabilities principle discussed above. However, if we should find that historical data do not match our predictions made on the basis of modern data and experiments, we may have to rethink this assumption. In other words, while the uniformitarian principle is an indispensable and useful tool in historical research, it also seems important and interesting to check every once in a while whether our trust in uniformitarianism in linguistics and the humanities generally is actually always justified.
References Bergs, Alexander. 2005. Social networks and historical sociolinguistics: Studies in morphosyntactic variation in the Paston Letters (1421–1503). Berlin & New York: De Gruyter Mouton Bergs, Alexander. 2012. The uniformitarian principle and the risk of anachronisms in language and social history. In Juan Manuel Hernández-Campoy & Juan Camilo Conde-Silvestre (eds.), Blackwell handbook of historical sociolinguistics, 80–99. Oxford: Blackwell. Bergs, Alexander. 2013. Writing, reading, language change – A sociohistorical perspective on scribes, readers, and networks in medieval Britain. In Esther-Miriam Wagner, Ben Outhwaite & Bettina Beinhoff (eds.), Scribes as agents of language change, 241–260. Berlin & Boston: De Gruyter Mouton. Bergs, Alexander. Forthcoming. The linguistic footprints of authors and scribes: A medieval whodunnit. In Richard Watts, Daniel Schreier & Anita Auer (eds.), Heterogeneity vs. homogeneity in language: Searching for a ‘standard’ in letters. Cambridge: Cambridge University Press. Bever, Thomas G. 1970. The cognitive basis for linguistic structures. In John R. Hayes (ed.), Cognition and the development of language, 279–362. New York: John Wiley & Sons. Bybee, Joan. 2003. Cognitive processes in grammaticalization. In Michael Tomasello (ed.), The new psychology of language, Vol. 2, 145–167. Mahwah, NJ: Lawrence Erlbaum. Chafe, Wallace. 1970. Meaning and structure of language. Chicago: University of Chicago Press. Chafe, Wallace. 1994. Discourse, consciousness, and time: The flow and displacement of conscious experience in speaking and writing. Chicago: University of Chicago Press. Clark, Herbert H. & Eve V. Clark. 1977. Psychology and language: An introduction to psycholinguistics. New York, Chicago, San Francisco & Atlanta: Harcourt Brace Jovanovich. Clark, Herbert H. & Susan E. Haviland. 1977. Comprehension and the given new contract. In Roy O. Freedle (ed.), Discourse production and comprehension, 1–40. Hillsdale, NJ: Lawrence Erlbaum. Clanchy, Michael T. 2009. From memory to written record: England 1066–1307. Oxford: Blackwell. Davies, Mark. 2008–. The corpus of contemporary American English: 450 million words, 1990– present. Available online at http://corpus.byu.edu/coca/. Davis, Norman. 1965. The “Littera Troili” and English letters. Review of English Studies n.s. 16. 233–244.
Psycholinguistic perspectives on early Englishes
275
Davis, Norman. 1967. Style and stereotype in early English letters. Leeds Studies in English n.s. 1. 7–17. Davis, Norman (ed.). 1971. The Paston letters and papers of the fifteenth century, Part I. Oxford: Clarendon Press. Dirven, René. 1985. Metaphor as a basic means for extending the lexicon. In Wolf Paprotté & René Dirven (eds.), The ubiquity of metaphor in language and thought, 85–120. Amsterdam & Philadelphia: John Benjamins. Fabiszak, Malgorzata. 2001. The concept of “joy” in Old and Middle English: A semantic analysis. Piła: Wyzsza Szkola Biznesu. Field, John. 2004. Psycholinguistics: The key concepts. London & New York: Routledge. Fischer, Olga & Anette Rosenbach. 2000. Introduction. In Olga Fischer, Anette Rosenbach & Dieter Stein (eds.), Pathways of change: Grammaticalization in English, 1–38. Amsterdam & Philadelphia: John Benjamins. Frazier, Lyn. 1979. On comprehending sentences: Syntactic parsing strategies. Bloomington, IN: Indiana University Linguistics Club. Frazier, Lyn & Janet Dean Fodor. 1978. The sausage machine: A new two-stage parsing model. Cognition 6. 291–325. Fries, Charles C. 1940. On the development of structural use of word order in Modern English. Language 16. 199–208. Fulk, R. D., Robert E. Bjork & John D. Niles (eds.). 2008. Klaeber’s Beowulf and the Fight at Finnsburg. 4th edn. Toronto: University of Toronto Press. Garman, Michael. 1996. Psycholinguistics. Cambridge: Cambridge University Press. Goody, Jack. 1987. The interface between the written and the oral. Cambridge: Cambridge University Press. Hoffmann, Thomas & Graeme Trousdale (eds.). 2013. The Oxford handbook of construction grammar. Oxford: Oxford University Press. Kendall, Calvin. 2006. The metrical grammar of Beowulf. Cambridge: Cambridge University Press. Kimball, John. 1973. Seven principles of surface structure parsing in natural language. Cognition 2. 15–47. Krause, Carina Denise. 2013. Psycholinguistik und Neurolinguistik. In Achim Stephan & Sven Walter (eds.), Handbuch Kognitionswissenschaft, 66–70. Stuttgart & Weimar: Metzler. Labov, William. 1994. Principles of linguistic change. Vol. 1: Internal factors. Oxford: Blackwell. Labov, William. 2010. Principles of linguistic change. Vol. 3: Cognitive and cultural factors. Chichester, West Sussex: Wiley-Blackwell. Lamb, Sidney M. 1998. Pathways of the brain. Amsterdam & Philadelphia: John Benjamins. Langacker, Ronald W. 2011. Grammaticalization and cognitive grammar. In Heiko Narrog & Bernd Heine (eds.), The Oxford handbook of grammaticalization, 79–91. Oxford: Oxford University Press. Lass, Roger. 1980. On explaining language change. Cambridge: Cambridge University Press. Lass, Roger. 1997. Historical linguistics and language change. Cambridge: Cambridge University Press. Lavandera, Beatriz R. 1978. Where does the sociolinguistic variable stop? Language in Society 7. 171–182. Levelt, Willem J. M. 1995. Speaking: From intention to articulation. Cambridge, MA: MIT Press. Lyell, Charles. 1830–1833. Principles of geology, being an attempt to explain the former changes of the earth’s surface, by reference to causes now in operation. 3 vols. London: John Murray.
276
Alexander Bergs and Meike Pentrel
McEnery, Tony & Andrew Hardie. 2012. Corpus linguistics. Cambridge: Cambridge University Press. Mitchell, Bruce. 1985. Old English syntax. 2 vols. Oxford: Clarendon Press. Molencki, Rafal. 2012. Old English: Syntax. In Alexander Bergs & Laurel J. Brinton (eds.), English historical linguistics: An international handbook, Vol. 1, 294–312. Berlin & Boston: De Gruyter Mouton. Nerlich, Brigitte. 1990. Change in language: Whitney, Bréal, and Wegener. London & New York: Routledge. Nevalainen, Terttu. 1999. Making the best use of “bad” data: Evidence for sociolinguistic variation in Early Modern English. Neuphilologische Mitteilungen 100(4). 499–533. Ong, Walter J. 2002. Orality and literacy: The technologizing of the word. London: Routledge. Orchard, Andy. 2003. A critical companion to ‘Beowulf’. London: D.S. Brewer. Prideaux, Gary D. 1984. The role of closure in language processing. Working Papers of the Linguistic Circle (WPLC) 4(1). 58–74. Prideaux, Gary D. 1989. Text data as evidence for language processing principles: The grammar of ordered events. Language Science 11(1). 27–42. Prideaux, Gary D. & William J. Baker. 1986. Strategies and structures: The processing of relative clauses. Amsterdam & Philadelphia: John Benjamins. Romaine, Suzanne. 1982. Socio-historical linguistics. Cambridge: Cambridge University Press. Romaine, Suzanne. 1984. Towards a typology of relative clause formation strategies in Germanic. In Jacek Fisiak (ed.), Historical syntax, 437–70. The Hague: Mouton. Sheldon, Amy L. 1974. The acquisition of relative clauses in English. Bloomington, IN: Indiana University Linguistics Club. Sheldon, Amy L. 1977. The role of parallel function in the acquisition of relative clauses in English. Journal of verbal learning and verbal behaviour 13. 272–281. Slobin, Dan I. 1973. Cognitive prerequisites for the development of language. In Charles A. Ferguson & Dan I. Slobin (eds.), Studies of child language development, 175–208. New York: Holt, Rinehart & Winston. Slobin, Dan I. & Thomas G. Bever. 1982. Children use canonical sentence schemas: A crosslinguistic study of word order and inflections. Cognition 12. 229–265. Traugott, Elizabeth Closs & Richard B. Dasher. 2002. Regularity in semantic change. Cambridge: Cambridge University Press. Whitney, William Dwight. 1867. Language and the study of language: Twelve lectures on the principles of linguistic science. 5th edn. New York: Charles Scribner’s Sons. Winters, Margaret E. 2010. Introduction: On the emergence of diachronic cognitive linguistics. In Margaret E. Winters, Heli Tissari & Kathryn Allan (eds.), Historical cognitive linguistics, 3–27. Berlin & New York: De Gruyter Mouton. Winters, Margaret E., Heli Tissari & Kathryn Allan (eds.), Historical cognitive linguistics. Berlin & New York: De Gruyter Mouton. Wray, Alison. 2002. Formulaic language and the lexicon. Cambridge: Cambridge University Press.
William A. Kretzschmar, Jr.
Complex systems and the history of the English language 1 Introduction When I last visited Cambridge University, I found this in the King’s Parade (Figure 1). Cambridge is one of those old places that seem never to change, and yet here was the new Corpus Clock right in the middle of things, on an outside corner of the
Figure 1: The Corpus Clock (http://commons.wikimedia.org/wiki/Corpus_Clock#mediaviewer/File:Corpus_clock_pol.jpg)1
1 By Rror (Own work) [CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0) or GFDL (http://www.gnu.org/copyleft/fdl.html)], via Wikimedia Commons.
278
William A. Kretzschmar, Jr.
staid Corpus Christi College (founded in 1352) and right across from the sixteenthcentury King’s College Chapel, famous for its Christmas Eve choral broadcasts. Not long after my previous visit to the city in 2008, Stephen Hawking, who wrote A brief history of time (1988), had unveiled the clock and thus added a bit to time’s narrative. The rather disturbing “chronophage”, or ‘time eater’, atop the clock eats the hours and the seconds as they pass. Its insect-like form was perhaps suggested by the name for the workings of the clock, a “grasshopper escapement” originally invented in the eighteenth century and here enhanced by an electric motor that winds a spring to keep the clock in motion. The other disconcerting thing about the clock, after its striking appearance and its equally amazing appearance at all in old Cambridge, is that it is accurate only every five minutes: time slows down for a while, even stops, and then speeds to catch up. Of course this trick is on purpose, something that, like the chronophage, is meant to unsettle us from our comfortable assumptions about time and its regular passage. Thus Stephen Hawking was the right person to unveil it, owing to his work in contemporary physics that challenges our assumptions about time and the universe around us, and thus too the idea to present it here, as a fitting emblem of the new ideas about our field of historical linguistics that I want to bring to you in this essay.
2 Complex systems The new science of complex systems (Mitchell 2009; Kretzschmar 2009) is something that historical linguists not only can use but should use in order to improve the relationship between the speech we observe from historical settings and the generalizations we make from it. After a brief introduction to complex systems, I take up the topic of language evolution from the point of view of the Corpus Clock, that the evolutionary process can speed up and slow down in time, rather than proceed in regular steps. I then suggest some practical applications of complex systems to historical linguistics, things that we can do now to improve how we think about our work. My examples come from the language whose history I know best, English, but at the same time it should be clear that the principles I describe are not limited to English and can be applied to any linguistic interest.
2.1 The nature of complex systems Complex systems, as described in physics, evolutionary biology, and many other sciences, are made up of massive numbers of components interacting with one
Complex systems and the history of the English language
279
another, and this results in self-organization and emergent order. Let us begin with a non-linguistic example and consider the line of ants in Figure 2 (the ant section is derived from Mitchell 2009). Ants are not smart. They can do only a few things, like exploit food sources, build nests, and defend themselves against intrusions, but no commissar of ants tells them to do any one of them. Instead, ants just happen to be doing one of these tasks at any given time. In searching for food, for example, ants wander around randomly; if they find some food, they leave chemical traces along their path to bring it back to the colony. Other ants can follow the traces, and leave more traces on their way back, till the path becomes a line like this one with lots of ants on it to exploit the food resource. However, not all of the ants in the nest follow the path; some keep foraging and some stay home on nest and defense duty, again not because they are told to do so by the queen (whose role is just for reproduction) or some ant general. When ants come boiling out of a nest when it is disturbed, they are not reacting to a chemical trace but to touching the antennae of other ants, and then changing their behavior from food or nest duty to defense. However, not all of the ants leave to gather food or rush to defense given the stimulus to do so. Some stay on nest duty during a provocation, and some ants look randomly for food when a great many have joined the
Figure 2: Ants in a line (http://en.wikipedia.org/wiki/File:Safari_ants.jpg)2
2 By Mehmet Karatay (Own work). Permission to use under the terms of the GNU Free D ocumentation License.
280
William A. Kretzschmar, Jr.
line to a known food source. What the ant does at any given time is influenced, but not determined, by what happens near it; when it detects chemical traces or touches antennae, an ant becomes more likely to enact one of the three behav iors. Individual ants are subject to feedback from other ants; in a primitive way, they exchange information. What looks to us like highly organized behavior is not controlled by any leader, and it is not absolutely determined by instinct or particular stimuli, but instead patterns of activity emerge from the instinctual behaviors of ants as they are conditioned by circumstances. These patterns of activity that emerge from the complex system make the whole more than just the sum of a few instinctive behaviors. We can sum up the process of all complex systems in just a few principles: (1) random interaction of large numbers of components, (2) continuing activity in the system, (3) exchange of information with feedback, (4) reinforcement of behaviors, and (5) emergence of stable patterns without central control.
2.2 Speech as a complex system Complex systems like this were originally described in the physical and biological sciences, and the definite procedures of information exchange have been explored in computer science (e.g., Holland 1998). Stephen Hawking includes complex systems in his latest popular book about physics, The grand design (Hawking and Mlodinow 2010); and Stephen Jay Gould (2003) and many others have pursued complex systems in evolutionary biology and genetics. Complex systems also occur everywhere in speech, as I have described from first principles in The linguistics of speech (2009). For speech, the randomly interacting “components” are all of the different variant realizations of linguistic features as they are deployed by human agents, speakers. These “features” might be different pronunciations, or different words for the same thing, or different ways of saying or writing the same thing – really, any aspect of speech that is recognizable for itself and therefore countable. The activity in the system consists of all of our conversation and writing. The exchange of information is not the same as sharing the meaningful content of what we say and write (which is exchange in a different sense), but instead our implicit comparison of the use of different components by different speakers and writers, as they use them in different kinds of conversations and writings. Feedback from exchange of information causes reinforcement, in that speakers and writers are more likely to employ particular components in future occurrences of particular circumstances for conversations and writing. Human
Complex systems and the history of the English language
281
agents, unlike ants, can think about and choose how to deploy linguistic variants, but that does not change the basic operation of feedback and reinforcement; we all make choices inside of the complex system of speech, in relation to current circumstances. The order that emerges in speech is simply the configuration of components, whether particular words, pronunciations, or constructions, that comes to occur in the local communities, regional and social, and in the occasions for speech and writing, text types and registers, in which we actually communicate. The process at work in complex systems just explains better what we already knew: we tend to talk like the people nearby, either physically near or socially near, or both, and we tend to use the same linguistic tools that others do when we are writing or saying the same kind of thing.
2.3 The emergence of complex systems of speech It is easy to see the patterns emerging from ant behavior, but it is much harder to watch emergence in speech over long stretches of time. Fortunately, the complex system of speech exhibits two technical characteristics that we can look for as signs
Figure 3: The A-curve, thunderstorm responses, LAMSAS survey
282
William A. Kretzschmar, Jr.
that the system has operated: nonlinear distribution and scaling. When the variant types of any linguistic feature are graphed according to their token frequency, as in Figure 3, the chart exhibits a nonlinear asymptotic hyperbolic curve (henceforth, A-curve), characterized by a small number of highly frequent responses and a much larger number of less-frequently-occurring responses (the long tail). The curve is the result of reinforcement from feedback in the continuing exchange of component variants for, here, the 109 different words and short phrases that 1162 Americans in a survey of speech in the Eastern states, the Linguistic atlas of the Middle and South Atlantic States (LAMSAS; see Kretzschmar et al. 1993 for a complete description) call a thunderstorm (e.g., common ones like thunderstorm or electrical storm, and uncommon ones like bluster or squall gale). The concept of the A-curve will be familiar to those who know Zipf’s Law (1949), which states that the frequency of words in a text is inversely proportional to their rank at any scale. Jean Séguy noted the same nonlinear distribution of linguistic distance as a function of geography (1971), called Séguy’s Law. Until now these “laws” have been viewed as curiosities by most linguists, but in complex systems the nonlinear distributional pattern occurs literally everywhere, for every feature in every survey I know of, in whatever language is being studied (English, Finnish, French, German, Japanese, Polish, Thai, and Vietnamese, from those my own students have studied so far). The word “law”, however, is really an overstatement. The distributions in linguistic atlas data from LAMSAS and other regional projects are always nonlinear, but quite variable in proportions, and so not subject to exact description with a precise formula, as we might have in physics for the Law of Gravity or of Thermodynamics. Thus the so-called “80/20 Rule” is more a practical rule of thumb (Figure 4), in which about 20% of the types account for 80% of the tokens, and 80% of the types in the long tail account for only 20% of the tokens, when the actual proportions may well be 70/30 or, as in Figure 4 for thunderstorm words, about 90/10. The top-ranked 15 variants for thunderstorm (13%) account for 1747 of the 2005 tokens (87%). The A-curve is a perfect example of the nonlinear distributional pattern characteristic of complex systems. Another hallmark of complex systems is the property of scaling, or “scalefree networks”. Scaling in speech takes the form of repeating nonlinear distributions of variants for the data overall and for every subsample. In Figure 5 we see the distribution of variant realizations of the [ɪ] vowel in the word six graphically, for the entire linguistic atlas survey in the Eastern States. Table 1 particularizes the token count represented on the graph in Figure 5. In Figure 6(a–c), we see the distribution of the same data for our three “types” of speakers. Table 2 pro vides the token counts by type of speaker. Our three “type” classifications describe levels of education and social involvement, where Type I is the lowest and Type III is the highest. Each of these charts of the [ɪ] vowel have about 80% of their tokens in the three top-ranked types.
Complex systems and the history of the English language
283
Figure 4: The 80/20 Rule, thunderstorm responses
Table 1: Token distribution of [ɪ] in six, LAMSAS survey sɪks sɪ˄ks sɪ·ks sɪ˄{ə}ks sɪ˃ks sɪ˗ks sɪ˄·ks si˅ks sɪ{ə}ks sɪ˄˃ks
606 213 130 69 24 17 16 16 11 9
sɪ˅˃ks sɪ˅{ə}ks sɪ˄{ɪ˗}ks sɪ˗˂ks sɪ·{ə}ks sɪ ̆ks sɪ{ə˃}ks sɪ˄˃{ə}ks si˅{ɪ˗}ks sɪ·{ə˃}ks
5 4 4 4 3 2 1 1 1 1
sɪ˄·{ə}ks sɪ·{ɪ˗}ks sɪ˅·ks sɪ˄˃·ks sɪ˂ɪks sɪ˂ks sɪ˗˄ks sɪ˗˅ks sɪ˄˂ks (top 3 = 83%)
1 1 1 1 1 1 1 1
The curves have subtle differences, but the nonlinear shape appears clearly in each graph, overall and in both subsamples – and throughout all of my Atlas survey data. Complexity science accordingly tells us about the process by which order emerges from massive numbers of random interactions among the components in the complex system of speech, rather than from a simple cause or a set of rules
284
William A. Kretzschmar, Jr.
Figure 5: Realizations of [ɪ] in six, LAMSAS survey 350
300
250
200
150
100
50
0
Figure 6a: Realizations of [ɪ] in six, divided by “type” of speaker: Type I
Complex systems and the history of the English language
300
250
200
150
100
50
0
Figure 6b: Realizations of [ɪ] in six, divided by “type”of speaker: Type II
70
60
50
40
30
20
10
0
Figure 6c: Realizations of [ɪ] in six, divided by “type” of speaker: Type III
285
286
William A. Kretzschmar, Jr.
Table 2: Token counts for [ɪ] in six, LAMSAS survey, arranged by “type” of speaker Type I (564) sɪks sɪ˄ks sɪ·ks sɪ˄{ə}ks sɪ˄·ks sɪ˃ks si˅ks sɪ˗ks sɪ{ə}ks sɪ·{ə}ks sɪ˄˃ks sɪ˅{ə}ks sɪ˄{ɪ˗}ks sɪ˅˃ks sɪ{ə˃}ks si˅{ɪ˗}ks sɪ˄·{ə}ks sɪ˄˃·ks sɪ ̆ks sɪ˂ks sɪ-˂ks
Type II (442) 295 94 81 28 12 10 10 8 5 3 3 2 2 2 1 1 1 1 1 1 1
sɪks sɪ˄ks sɪ·ks sɪ˄{ə}ks sɪ˃ks sɪ˗ks sɪ{ə}ks si˅ks sɪ˄·ks sɪ˄˃ks sɪ˅{ə}ks sɪ˗˂ks sɪ˄˃{ə}ks sɪ˄{ɪ˗}ks sɪ·{ə˃}ks sɪ˅·ks
Type III (140) 246 79 43 29 12 8 5 5 4 3 2 2 1 1 1 1
sɪks sɪ˄ks sɪ˄{ə}ks sɪ·ks sɪ˄˃ks sɪ˅˃ks sɪ˃ks sɪ{ə}ks sɪ˄{ɪ˗}ks sɪ·{ɪ˗}ks sɪ˂ɪks sɪ ̆ks sɪ˗ks si˅ks sɪ˗˂ks sɪ˄˂ks
65 40 12 6 3 3 2 1 1 1 1 1 1 1 1 1
or any controlling agent. As I have shown here, whether we are talking about ants, physics, or speech, a few simple principles can give rise to frequency-based patterns. These patterns are not the same as what we usually call grammar (more on that below). Instead, complexity science defines the relationship between language in use and any generalizations we may wish to make from it. It allows us, really for the first time, to get away from reliance on our perceptions of the language around us, and instead make effective use of the actual patterns in speech as people use it in order to resolve practical problems in language study.
3 Corpus evidence To illustrate these points, let us now return to a linguistic analogue of the Corpus Clock: a brief historical analysis of the word clockwork in Mark Davies’ American English corpora, The corpus of contemporary American English (COCA) and The
Complex systems and the history of the English language
287
corpus of historical American English (COHA) (see Davies 2008–, 2010–). In COCA (Figure 7, at this writing in April 2012, 425 million words from 1990 to 2011), the word clockwork occurs 468 times, and it is accompanied by the collocates shown in Figure 7 within a span of four words before and after. The list clearly follows the nonlinear curve we expect of a complex system, though the long tail of single occurrences is cut off in Figure 7. We can set aside for the moment the collocation “clockwork orange” as the creative title of a popular book and movie. The largest use of clockwork comes with like, and with words such as every with a time unit (as in “every month like clockwork”), or everything (as in “everything ran like clockwork”). Similar to such usages, we have regular (as in “regular as clockwork”). These are all similes, marked metaphorical uses that represent clockwork as the epitome of control. Perhaps the title “clockwork orange” plays creatively with this meaning as well. The same might be said for the unmarked metaphorical use “clockwork universe”, which understands the universe to operate by Newton’s laws as regularly as a clock runs. As against these metaphorical uses, we have mechanism, which refers to the literal wheels and gears of a clock or of a machine like a clock. The point here is to show that our contemporary understanding of clockwork is not so much the literal sense, although we do have it still, but instead the metaphorical sense of regularity, whether in daily events or in the universe itself. Time passes at a constant rate, as measured more or less precisely by clockwork. If we compare these results to those derivable from COHA, which at this writing in April 2012 comprises 400 million words of American English, collected by decade over two centuries (Figure 8), we see that clockwork occurred 353 times. This must mean that its rate of occurrence has increased in recent decades. To shift for a moment to the Oxford English dictionary (OED), while mechanical clocks were invented in the Middle Ages, the first citation of clockwork in the OED is from 1628, in its metaphorical sense in the phrase “this curious clocke-worke of religion”. The literal sense is documented in the OED only from 1662. The rise in frequency of clockwork in American English as shown in COHA corresponds in
Figure 7: Collocates of clockwork in COCA (http://corpus.byu.edu/coca)
288
William A. Kretzschmar, Jr.
Figure 8: Collocates of clockwork in COHA (http://corpus.byu.edu/coha)
time to the mass production of mechanical clocks in the nineteenth century (it was one of the initial such industries in America). The unusual facts of the time lag in the appearance of clockwork, and of the appearance of the metaphorical sense before the literal sense in the OED, and its relatively sudden expansion in use in American English, are topics to which I will return shortly. In the COHA list we again see the nonlinear curve we expect from complex systems in the distribution of collocates, with many of the same words. However, we do see the operation of historical change. In the nineteenth century, piece, wound, and mechanism are much more common than the metaphorical uses with regularity and precision. The typical meaning of clockwork changed during the twentieth century, so that the metaphorical sense became more common than the literal sense. Both of these senses were possible in each century; what changed was their frequency of use. If we look a little closer at clockwork universe, however, between its emergence in COHA in 1940 and its greater prevalence in the last two decades in COCA, we see that almost all of the examples are used in a negative sense: they refer to the demise of seventeenth-century Newtonian ideas of a regular, controlled universe in favor of the relativism of Einstein and Hawking. So, for example, people are writing things in reference to quantum mechanics like “That new theory replaced the perfectly predictable, clockwork universe of Isaac Newton with a mysterious world” (COHA 1980s token). So, in another historical development, it appears that in the most recent time period, we have come to reject the clockwork metaphor when it refers to our universe. In a particular domain, in usage at one small scale, we find that the meaning of the word does not follow the frequency
Complex systems and the history of the English language
289
pattern that it does more generally at the time. Again as we would expect of a complex system, we need to take scale into account in our observations, to understand that linguistic behavior may be different in different subsets of our data, in different situations of use. Furthermore, if we do see different linguistic behavior in different places or in different text types, nothing has gone wrong; such differ ences are not evidence of some error by the writers or of some error in our analysis. Not only does clockwork have different frequencies of its meanings over time, but those meanings occur at different frequencies in different domains at the same moment in time. We now see that, because of the operation of speech as a complex system, and in line with our modern view of a “clockwork universe”, this semantic aspect of the history of the language does not operate like clockwork. It took too long, we might think, for clockwork to appear in American English, and the word has had a complicated and changing semantic history. Ironically, the history of clockwork shows us how time appears to speed up and slow down, on occasion even to stop (in the delayed appearance of the word), when we inspect the continuous history of the word’s semantic development.
4 Complex systems and change 4.1 Grammaticalization and change Now let us consider time over the long term not for semantics but as regards the topic of grammaticalization in language change. The origin of the idea is Paul Hopper’s influential (1987) paper, “Emergent Grammar”. Hopper did not derive his term “emergent” from complexity science, but his position accords exactly with complex systems as described here: This is, then, roughly the context in which the term Emergent Grammar is being proposed. The term “emergent” itself I take from an essay by the cultural anthropologist James Clifford, but I have transferred it from its original context of “culture” to that of “grammar”. Clifford remarks that “Culture is temporal, emergent, and disputed” (Clifford 1986: 19). I believe the same is true of grammar, which like speech itself must be viewed as a real-time, social phenomenon, and therefore is temporal; its structure is always deferred, always in a process but never arriving, and therefore emergent (…) (1987: 141).
Hopper took the term emergence from cultural anthropology, from an article published about the same time that the Santa Fe Institute was established to study complex systems, an example of how similar ideas can be “in the air” at a
290
William A. Kretzschmar, Jr.
moment in time.3 The key point here is that Hopper maintains an idea of grammar as structure, but one that is never instantiated in the speech that generates it. If the structure of grammar is “always in a process but never arriving”, a wonderful, much-cited phrase, this means that we can conceive of grammar as a rational object, but one that is always indirectly related to speech. In another expressive piece of prose, Hopper explains further: The notion of emergence is a pregnant one. It is not intended to be a standard sense of origins or genealogy, not a historical question of “how” the grammar came to be the way it “is”, but instead it takes the adjective emergent seriously as a continual movement towards structure, a postponement or “deferral” of structure, a view of structure as always provi sional, always negotiable, and in fact as epiphenomenal, that is at least as much an effect as a cause (1987: 142).
Grammaticalization, then, is not an explanation for current structure in language, but instead should be understood as “continual movement”. Structure is
3 Keller’s On language change (1994) might also be mentioned as an early discursive treatment in linguistics of “the invisible hand”, Adam Smith’s name for emergence, although his book does not directly address complex systems. One of the best current treatments of complex systems and language has a central concern with this topic. The Five Graces Group, which includes Joan Bybee, Nick Ellis, John Holland, and Diane Larsen-Freeman, among others, has produced a position paper called “Language is a complex adaptive system” (Beckner et al. 2009). They begin with a basic tenet of grammaticalization: istorical changes in language point toward a model in which patterns of co-occurrence H must be taken into account. In sum, “items that are used together fuse together” (Bybee, 2002). For example, the English contracted forms (I’m, they’ll) originate from the fusion of co-occurring forms (Krug, 1998). Auxiliaries become bound to their more frequent collocate, namely the preceding pronoun, even though such developments run counter to a traditional, syntactic constituent analysis (6). They continue to specify the goal of their position: In the usage-based framework, we are interested in emergent generalizations across languages, specific patterns of use as contributors to change and as indicators of linguistic representations, and the cognitive underpinnings of language processing and change (7). rammaticalization, then, is a universal process, one which describes the operation of the comG plex adaptive system of speech. It is taken to operate at the level of single languages. Since the Five Graces propose that grammar is “a network built up from the categorized instances of language use” (2009: 5), then grammaticalization is the mechanism by which they propose that such a network arises and changes.
Complex systems and the history of the English language
291
“epiphenomenal” not because it is unimportant, but because it is always contingent and never directly observable. Finally, Hopper affirms that [b]ecause grammar is always emergent but never present, it could be said that it never exists as such, but is always coming into being. There is, in other words, no “grammar” but only “grammaticization” – movements toward structure which are often characterizable in typical ways (1987: 148).
This statement might be modified only to clarify that, while grammar never exists as such in language in use, it can well exist as a description of regularities indirectly derived from speech performance by perceptual means. In later years, the idea of grammaticalization became much more aligned with formal linguistics, and less with Hopper’s idea of emergence that connects so well with complex systems. Hopper and co-author Elizabeth Traugott (1993) define grammaticalization as a process, but now one characterized by how the properties of sentences “come into being” (Hopper and Traugott 1993: xv) or have synchronic organization. Grammar has become an object, and individual con structions are “selected” as grammatical parts of it with reference to a framework, a set of categories which are non-discrete.4 What may have begun as emergence in a complex system has now become highly formal, reified as a structure, more over one which is thought to participate in the continuing process. So, in Grammaticalization, what had started in “Emergent Grammar” the article has become radically different, and in so doing the most brilliant insight of Hopper’s article, the idea of grammar as continual movement, has been returned in the book to a much more mainstream linguistic discussion.5
4 Items may be “more” or “less” grammatical, i.e., more or less a part of the framework. Grammar is a framework which is more or less contingent (because the categories are non-discrete, not naturally given as discrete), but which is nonetheless objectified, reified, and above all char acterized by constraints of syntactic, morphosyntactic, and morphological structure. The role of most twentieth-century linguists is to select grammatical constructions in an attempt to make useful idealizations from speech data. The second edition of Hopper and Traugott (2003) c hanges the definitions and does mention emergence, if briefly. Discussion here of the first edition of the book emphasizes the important point that Hopper’s initial insight about emergence did not gain traction in linguistics, but was subordinated to the prevailing notion of grammar. 5 The Five Graces also describe grammar as a kind of object, a “network”, that converts Hopper’s process into a state, and identifies the state of the complex system as its grammar. In the language of complex systems, a network is a set of nodes, the elements in the complex system, without regard to the condition of each node. A “state” is the condition of all of the elements in the s ystem at one moment in time, or of a single element. For a single traffic light (to borrow an example from Larsen-Freeman and Cameron 2008: 32), there are three possible conditions, red, yellow,
292
William A. Kretzschmar, Jr.
For historians of the language, the key question is how to accommodate Hopper’s sense of continual movement, and still be able to describe the grammar of the language at any moment in time. The best answer from complexity science, again, is to observe the nonlinear distribution and scaling properties of complex systems, and to put the 80/20 Rule to good practical use, and these notions lead me to suggest some improvements in our analytical practices for historical studies. The idea of selection should be reserved for rule-based generative grammars that need to “select” the most frequent forms of constructions in order to preserve the elegance of their logical systems. Generative grammars can afford to, and therefore should, ignore the 80% of possible constructions that occur rarely in order to concentrate on the 20% of the construction types that account for 80% of the tokens. To do otherwise results in “rule creep”, the inelegant addition of more and more rules to account for infrequent cases, and thus also the loss of what best distinguishes generativism from structuralism. On the other hand, in structural grammars that collect paradigmatic lists of possible constructions, there is no good linguistic reason to privilege the most common variants as having been “selected” and therefore having status as being “grammatical”, and to relegate less-common variants to “noise” in the system.
or green, that the light could have at any moment in time. In the language of complex systems, the “state space”, the set of possible states in the system, of the single traffic light consists of those three states. If we had two traffic lights, the state space would have nine states (RR, RY, RG, YY, YR, YG, GG, GR, RY). For any single linguistic feature, which is not so simple as a traffic light, the Graces refer to the process of “selection” as the emergence of a preference for one state of the feature, one variant, over other possible variants. They associate change with long-term alteration in social practices, “which in the extreme case leads to the fixation of [new forms] and extinction of [old forms]”: [C]hanges in lifestyles lead to the rise and fall of words and constructions associated with those lifestyles (e.g., the rise of cell [phone] and the fall of harquebus). In the latter case, the social identity and the social contexts of interaction lead to the rise and fall of linguistic forms that are associated with various social values by speakers (Larsen-Freeman and Cameron 2008: 32). gain, the preference for, or “fixation”, of a state takes the dynamic movement of the complex A system and freezes it, so that one variant of a feature becomes “grammatical” in the sense of having been selected. The Five Graces Group, in my view, has been too eager to identify grammar directly with one aspect of complexity science, states and state space. In so doing, again in my view, they lose Hopper’s sense of continual movement, and also lose the benefit of understanding speech as a complex system. Of course there is nothing at all wrong with making grammars. This kind of formal analysis has been highly productive. But there is an essential conflict be tween making a grammar as a static hierarchy for a language at one moment in time, and the process of change in language that is best described as frequency change within a complex system.
Complex systems and the history of the English language
293
It will be enough to note whether any particular variant is in the 80% core or the 20% periphery of constructions. All of the variants on the A-curve are actually just as relevant for inclusion in the structural system. The notion of language-internal selection, then, cannot operate within the A-curve because nothing is really chosen or preferred. For the historian of the language, this means that “change” can no longer be defined as a different variant having been “selected” at a later moment in time. If we observe the distributional pattern of linguistic features at any moment in time, we can get more historical information than we might have thought. Paul Hopper raised just this issue in “Emergent Grammar” (1987: 142–143) using the example of the development of the English indefinite article a/an: To take just these three functions of the predecessor of a/an in Old English, we find in modern English not a uniform, over-all weakening of the meaning, but rather a situation in which the weakened meanings and the older stronger meanings exist side-by-side.
In other words, English has preserved the historical functions of a/an, as well as developing new ones. Hopper provided as examples “They introduced me to a young woman [whose name was Ethel]; My husband and I went to a showroom to pick up a new car we had ordered. After we had test-driven it, the salesman asked us ...” versus “Birds of a feather flock together; They are all of a kind; A stitch in time saves nine; A penny saved is a penny earned; How much is that pictureframe? – A dollar; Linguists of a theory attend the same conferences; What was left of the woods after they built the parking-lot? – A tree”. As Hopper admitted, there are retentions of archaisms in proverbial language, which is where Hopper was able to illustrate modern usage of the older functions. Still, whether in proverbial language or in special domains like law or religion, we do still employ such old constructions, which gives them a low frequency on the modern A-curve. The indefinite article is not an exception but a good example of a general capacity of speech to retain old feature variants at low frequencies. One of the points made by C. S. Lewis in his Studies in words (1990) is that the meanings of key words in our culture have “ramifications”. That is, words do retain their older meanings even while they gather new ones. This fact may be obscured for incautious readers by what Lewis called the “dangerous sense”, the meaning that is so frequent in modern usage that we automatically think of it. Lewis could have had no idea of A-curves, but the dangerous sense is of course the top-ranked meaning on the modern nonlinear curve of meanings for a word. Our corpus analysis of clockwork showed us this process in operation in contemporary time: we can still use the word to refer to gears and wheels, and the dangerous sense of the word is regularity and precision, but we need to be aware that now we often speak of clockwork in a negative sense, as with universe and orange, that may give the word
294
William A. Kretzschmar, Jr.
an unsavory or ironic semantic prosody, a semantic layer that reflects an overall feeling about a word rather than a specific definition. Thus, for grammatical constructions, the lexicon, and semantics, all of the evidence suggests that historical forms tend to be retained as low-frequency variants in the tail of the nonlinear distribution of contemporary usage, and that such usage continues actively to change.
4.2 The S-curve and the A-curve For historians of the language, this means that “change” in the complex system of speech or in structural grammars will consist of an alteration in frequency of any particular feature, instead of the selection of one form over another. One way to track feature frequency, the S-curve, has already been described for the progress of linguistic change (notably in Kroch 1989; Labov 1994: 65–67), and the A-curve distribution is in no way at odds with the S-curve. The two curves are actually different expressions of the same basic distributional facts. The S-curve just describes the successive frequencies of a single variant at different moments in time. In Figure 9 we see two different A-curves that correspond to different moments in time for the same variant, with the position of the variant located on each curve. We can recall, for instance, the changing frequencies of the meanings of clockwork over the past two centuries. If we then draw A-curve charts that correspond to different times on a given S-curve chart, we can see that the characteristically sudden positive movement of variants on the S-curve is mirrored by the shape of the A-curve (Figure 10). When a particular variant “climbs” the A-curve by moving up in frequency rank, the distributional patterns of both
900
900
800
800
700
700
600
600
500
500
400
400
300
300
200
200
100
100
Moment A in me
0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
0
Moment (A + n) in me
Figure 9: A-curves at two moments in time (Kretzschmar 2009: Fig. 8.2)
Complex systems and the history of the English language
295
the A-curve and the S-curve predict that there is a larger relative change in the middle of the curves. As a variant moves from one rank to the next in the middle of the curve, each step in rank describes an increasingly large number of occurrences. This means that the small number of occurrences when a variant begins to become more frequent will be expressed as a slowly growing proportion, and
Figure 10: A-curve and S-curve (Kretzschmar 2009: Fig. 8.3)
296
William A. Kretzschmar, Jr.
the larger number of occurrences in the middle of the curve will be expressed as a rapidly growing proportion. The A-curve and the S-curve are thus complementary descriptions of the distributional facts of variant linguistic forms at different moments in time, new and improved ways to document historical change. And this focus on frequency distributions, rather than qualitative change, also allows explicitly for what Laura Wright (2000: 6) has waggishly called “W curves” that describe increases and decreases in the frequency of forms over time.
4.3 The 80/20 Rule and estimates of change The A-curve offers historians of the language a way to judge their sparse data. Historical studies often cannot acquire enough data for a good survey, and the data that is available does not constitute anything like a valid sample of the speech or texts in use at the time of the study. Still, the 80/20 Rule offers a metric to estimate the status of forms from the limited evidence available. This means that it is highly likely that a single occurrence of a form will come from the topranked variants on the curve (if we could make one), but there is also a decent chance, 20%, that it is actually an uncommon use somewhere in the long tail of the curve. Similarly, if we have a small number of occurrences of variant forms, it is likely, but not certain, that the distribution will begin to show a nonlinear curve. Therefore, when the amount of evidence is small, we should especially avoid the temptation to make categorical interpretations, i.e., that the form we found necessarily was the most common one at the time, or to assume that the results are necessarily typical of the domain that we searched. The best interpretation will be one that assumes the nonlinear pattern and attempts to fit the actual returns to that pattern as well as possible. Results from small data sets will not show us the whole picture, but the 80/20 Rule gives a chance to make a reasonable estimate. And we should always recognize that, because of scaling, findings from any particular domain will not yield a result that is always generalizable to larger or smaller domains. Merja Stenroos and I have offered a good example of this kind of reasoning (Figure 11), with application also to aggregations of features in geographical areas for Middle English dialects (Kretzschmar and Stenroos 2012), the first consonantal element of the verbs shall and should. In The Middle English grammar corpus (Stenroos, Mäkinen, Horobin, and Smith, 2011), which consists of 405 text samples from the fourteenth and fifteenth centuries, eleven different spellings are found. Of these forms, only three or four occur with a reasonable frequency in any given area and period; however, the dominant forms vary greatly between areas and periods. The frequency distribution in the texts localized in the Eastern part of the country and dated to the fifteenth century shows that the spelling is clearly the dominant
Complex systems and the history of the English language
297
1600 1400 1200 1000 800 600 400 200 0
sch-
sh-
x-
s-
sc-
ch-
xh-
Figure 11: Initial consonant in shall, fifteenth-century Eastern texts (from Kretzschmar and Stenroos 2012)
one, while we know that eventually moves up to become the most frequent variant at the larger national scale in England. The most common spellings in the thirteenth century, and , have, on the other hand, moved down the curve, surviving in the long tail of occasional forms found in any corpus of reasonable size. Evidence like this of the operation of complex systems does not accord well with our traditional means of describing geographical varieties. Traditional dialect surveys in England and America have both posited the existence of dialects and used survey evidence selectively to show where they might be (Kretzschmar 2009: 66–74). Similarly, the “fit-technique” developed for Middle English dialects by McIntosh (McIntosh et al. 1986: 23) localizes texts based on the assumption that Middle English linguistic variation forms a regular continuum, into which any dialectally consistent text may be placed. Both for Middle English and for modern English, these traditional approaches apply a formal assumption of the regularity of a dialect to evidence whose variation is anything but regular. On the other hand, when we apply our knowledge of complex systems, Merja Stenroos and I would prefer to map texts according to their provenance, as far as retrievable, arriving at a messier picture that simply answers the question “who wrote what where”. The study of texts associated with a particular location allows us to reconstruct something of the sociolinguistic reality within the community, which may have been complex (see Stenroos and Thengs 2011). At the same time, while some scribes travelled and some reproduced the forms of non-local exemplars, we expect
298
William A. Kretzschmar, Jr.
that the local variants will turn out to be most frequent overall in the A-curve of variant frequencies – like the one for the shall spellings in Eastern texts. Thus, the fit-technique in Middle English dialectology, like traditional modern dialectology, is best thought of as a method dominated by its formal assumptions. This does not make it bad in itself, but it does mean, as was the case for the construction of grammars, that we need to understand the degree of abstraction of the fit-technique as it defines one dialect, one grammar, to fit all the texts of a place – because complex systems tells us that that is never actually the case in the evidence itself.
5 Language evolution For my final point, I would like to pick up the geographical thread from our discussion of Middle English dialectology, and combine it with what I promised to say about the Corpus Clock. The clock, I said, is only accurate every five minutes: time slows down for a while and then speeds to catch up. It turns out that this seemingly perverse aspect of the clock is actually a good model for what we do with long-term change in historical linguistics, because it matches a model for biological evolution, called “punctuated equilibrium”, that capitalizes on how complex systems work. In linguistics, many others have discussed potential parallels between biological evolution and language change, most often to reject them (e.g., Lass 1990; Labov 2001). Salikoko Mufwene has offered the latest influential parallel (2001, 2008), in which he makes the analogy of language to a parasitic species. I would like here to say, not that evolutionary biology in the form of punctuated equilibrium provides an apt metaphor for language change, but that the workings of a complex system actually account for exactly the same process in both biology and language over time.
5.1 Punctuated equilibrium and the evolution of language Gould and Eldredge proposed the idea of punctuated equilibrium in 1972 (Eldredge and Gould 1972). As they reported in a later review article (Gould and Eldredge 1993), their theory has now received wide acceptance after an initial period of misunderstanding.6 They proposed that evolution was not a slow, gradual process that operated like clockwork, but instead that in the process of evolution species enjoyed long periods of stasis during which sub-populations of
6 They did not propose, as some claimed, that species suddenly leaped to new genetic configurations (the “saltational theory”), and they opposed attempts to coopt their theory by creationists.
Complex systems and the history of the English language
299
the species came to coexist, whether through random genetic drift or because of differences in environmental conditions, so that change might appear to be quite sudden. Our modern view of this process is affected by what they call “the notor ious imperfection of the fossil record” (1993: 222), that is, the fact that we find fossils at widely separated locations representing significantly different times, rather than collecting fossils that represent a valid sample over time of the central population of the species and of any sub-populations that may have developed. As Gould and Eldredge put it in paleontological terms, “small populations spec iating away from a central mass in tens or hundreds of thousands of years, will translate in almost every geological circumstance as a punctuation on a bedding plane, not gradual change up a hill of sediment, whereas stasis should characterize the long and recoverable history of successful central populations” (1993: 222). To be clear, Gould and Eldredge are pointing out that different species will be present in the same stratum of fossils, because of the long static periods that each species can exist and because of coexistent sub-populations of the species. Before punctuated equilibrium, it was thought that up to 90% of the differences in fossils came from the gradual evolutionary process, but now, they report, “all substantial evolutionary change must be reconceived as higher-level sorting based on differential success of certain kinds of stable species, rather than as progressive transformation within lineages” (Gould and Eldredge 1993: 223). And, to get back to the Corpus Clock, the accidents of what fossils we find can make it look as if time speeds up and slows down, that change can appear to happen quickly
Figure 12: Gould and Eldredge (1993) evolutionary map
300
William A. Kretzschmar, Jr.
sometimes and at other times can seem to stop. Punctuated equilibrium looks like syncopated time. The Gould and Eldredge map of the evolutionary process (Figure 12) contrasts the stability of the population at left (which represents gradual phylogenetic change), with the greater speciation of populations at right (which represents the simultaneous development of sub-populations). As I just suggested, the appear ance of sudden jumps in evolution comes about because imperfections in the geological fossil record may lead us to miscategorize the simultaneous existence of different populations as sequential populations. If we were to draw a straight line across the graph from any point on the vertical Time access (actually if we draw a plane, since this is meant to be a 3D chart), it would strike multiple populations from the sub-populations at right. Thus, from observations at any given moment in time, we are nearly certain to retrieve evidence that belongs to differ ent populations, and the trick for us is to make an appropriate analysis of that evidence. That is, we need to deal with the appearance of slow change and fast change that emerges from such an array of facts. What must be clear, however, is that the evidence we collect from any fossil record that corresponds to the right side of the chart cannot come from a single, continuous, gradual phylogenetic process. To try to make the right side of the chart look like the left side would make us discount some of the evidence we have, or prefer some parts of the evidence over other parts, in order to force the evidence to make the neat generalization of a gradual, clockwork evolutionary process. What Gould and Eldredge could not know in 1972, but what Gould had realized by the time he wrote his last book (2003), is that the right side of the chart corresponds exactly to what we would expect of a complex system. The scale-free property of complex systems predicts that there will always be sub-populations at different scales. We have already seen this in the distribution of individual features from the linguistic data; now we are just observing the same property at a higher level of scale of species, or in linguistic terms, in the aggregations of features we might call a dialect or a language. Gould and Eldredge originally matched their sub-populations to different geographic populations, known as allopatric, parapatric, and sympatric modes of spec iation (Figure 13). In each of these cases, a sub-population occurs in a geographic niche, whether created by a barrier, an extension, or an internal division. What we now understand from complex systems is that the pattern of sympatric speciation is not limited to geographic boundaries, although these remain an important consid eration, but that speciation can occur for other reasons as well: the scale-free property of the complex system predicts that there will always be many sub-populations, each one defined by its own nonlinear distribution of morphological characteristics. Complex systems also tells us that it will always also be possible to consider
Complex systems and the history of the English language
301
Figure 13: Geographic speciation modes (Drawn in Inkscape by Ilmari Karonen. Adapted from Spring 2006, Lecture Notes for EVOLUTION AND BIODIVERSITY class [BIL 160 Section HJ] by Dr. Dana Krempels. Permission to use under the terms of the GNU Free Documentation License).
populations at higher levels of scale, to consider any combination of sub-populations we want right up to the population as a whole, and that at every level of scale the different morphology of the sub-populations will be expressed in the nonlinear distribution of characteristics. We see this represented at the bottom of the chart, “after equilibriation”, when individuals from a particular sub-population are mixed in with the population as a whole. Thus, classic examples of geographical speciation such as Darwin observed in the finches of the Galapagos can involve geography, but our new knowledge of complex systems tells us that geography is not the whole story.
5.2 The Great Vowel Shift as a complex system An excellent example of the process of punctuated equilibrium in the history of English is the Great Vowel Shift, and Jeremy Smith has shown it to us (1996: 48–49 and 86–111).7 The general outline of the Great Vowel Shift should be clear,
7 Smith was already interested in chaos theory and complexity in 1996, but did not yet have the means to describe his evidence in those terms. Of course, now we can understand from complex systems that there will be an A-curve distribution of variants for every vowel following the 80/20 Rule, and that we can assemble the individual vowel frequency patterns to make a generalization at a larger level of scale for a regional dialect.
302
William A. Kretzschmar, Jr.
including the raising of both the long front and long back vowels. Smith, however, points out that in the North, the Great Vowel Shift did not proceed according to the usual generalization: an extra front vowel was raised, and the back vowels remained unraised (Figure 14). Smith offers us complications of the Great Vowel Shift in the South, too (Figure 15). He describes two competing systems in fifteenth-century London, the first among those whom he calls “descendants of Chaucer”, from higher social circles, and the second system prevailing among the remaining speakers in the Midlands and South. We see here geographical subpopulations, but also a pair of social sub-populations within a region. Somewhat later in his argument, Smith also introduces a third system which originated in East Anglia but whose speakers also went to London, where their speech pattern
Figure 14: Great Vowel Shift, Northern region (adapted from Smith 1996: 100)
Figure 15: Great Vowel Shift, London (I) and the Midlands/South region (II) (adapted from Smith 1996: 102)
Complex systems and the history of the English language
303
Table 3: Three systems involved in the Great Vowel Shift, ca. 1650 (adapted from Smith 1996: 108–110) System I: Middle English
System II: Middle English System III: Middle English
/e:/ > /i:/, e.g., meed /ɛ:/ > /e:/, e.g., mead /a:/, /ai/ > /ɛ:/, e.g., made, maid /e:/ > /i:/, e.g., meed /ɛ:/, /a:/, /ai/ > /e:/, e.g., mead, made, maid /e:/, /ɛ:/ > /i:/, e.g., meed, mead /a:/, /ai/ > /e:/, e.g., made, maid
competed with that of the descendants of Chaucer; speakers retained System II longer in the countryside (Table 3). This amounts to a kind of linguistic equilibriation, as described earlier for modes of speciation. I recommend the orthoepic evidence and details of Smith’s argument to you from his book. My point here, of course, is that complex systems predicts that both the regional and social sub-populations could develop different characteristics, and that these sub-populations could coexist at the same time, and that, just as in punctuated equilibrium, we can observe equilibriation as members from one sub-population become mixed with others. So, the linguistic evidence is not just parallel to the fossil record. It is the same procedure to use the fossil record to describe species as it is to use fragmentary linguistic evidence to describe competing dialects in English, or in the longest historical perspective, compet ing languages in the Indo-European family. The same underlying process is at work for both paleontology and historical linguistics. To rephrase Gould and Eldredge, while we used to worry about dialect contamination in the gradual evolution of a language, now all substantial evolutionary linguistic change ought to be reconceived as higher-level sorting based on differential success of certain kinds of stable language varieties.
5.2 The Great Vowel Shift and scale Given Smith’s evidence and my argument, then, what should we make of the Great Vowel Shift? The complex systems approach would suggest that we should not abandon it: the Great Vowel Shift remains a useful generalization at the top
304
William A. Kretzschmar, Jr.
level of scale. But we cannot just work at the top level of scale, because we are then subject to the dangers of misinterpretation of our imperfect evidence that come from punctuated equilibrium. Just as for the fit-technique, we cannot apply the top-level generalization of the Great Vowel Shift back down onto primary evidence, because it is based on the formal abstraction that we can define one grammar to fit all the texts of a place – and complex systems tells us that that is never actually the case. Just as we must do with the Cambridge Corpus Clock, we have to expect that time in some places will appear to go faster and in some places slower in historical linguistics. We now know that linguistic change does not just run like clockwork. There were good reasons why the Neogrammarians created a mechanical process for linguistic change, chiefly to adopt the best science then available to govern their work. We, too, need to do this: we should not just accept traditional methods because they are traditional. We should adopt the best science avail able now to govern our work. I am pleased to offer you complex systems for that purpose. If we choose to make formal generalizations, and in historical linguistics a great many of us will do so, then any formal statements we make should reflect what we now know about the emergence in our evidence of scaling, nonlinear frequency patterns from the complex system of speech.
References Beckner, Clay, Richard Blythe, Joan Bybee, Morton H. Christiansen, William Croft, Nick C. Ellis, et al. 2009. Language is a complex adaptive system: Position paper. Language Learning 59, Supplement. 1–26. Davies, Mark. 2008–. The corpus of contemporary American English: 450 million words, 1990– present. Available online at http://corpus.byu.edu/coca/ (accessed April 2012). Davies, Mark. 2010–. The corpus of historical American English: 400 million words, 1810–2009. Available online at http://corpus.byu.edu/coha/ (accessed April 2012). Eldredge, Niles & Stephen J. Gould. 1972. Punctuated equilibria: An alternative to phyletic gradualism. In Thomas J. M. Schopf (ed.), Models in paleobiology, 82–115. San Francisco: Freeman, Cooper. Gould, Stephen Jay. 2003. The hedgehog, the fox, and the magister’s pox: Mending the gap between science and the humanities. New York: Three Rivers Press. Gould, Stephen J. & Niles Eldredge. 1993. Punctuated equilibrium comes of age. Nature 366 (18 Nov). 222–227. Hawking, Stephen. 1988. A brief history of time. London: Bantam. Hawking, Stephen & Leonard Mlodinow. 2010. The grand design. New York: Bantam. Holland, John. 1998. Emergence: From chaos to order. New York: Basic. Hopper, Paul J. 1987. Emergent grammar. Berkeley Linguistics Society 13. 139–157 (Viewed April 2012 at http://home.eserver.org/hopper/emergence.html).
Complex systems and the history of the English language
305
Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. 1st edn. Cambridge: Cambridge University Press. [2nd edn., 2003]. Keller, Rudi. 1994. On language change: The invisible hand in language. Trans. by Brigitte Nerlich. London: Routledge. Kretzschmar, William A., Jr. 2009. The linguistics of speech. Cambridge: Cambridge University Press. Kretzschmar, William A., Jr. & Merja Stenroos. 2012. Evidence from surveys and atlases in the history of the English language. In Terttu Nevalainen & Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English, 111–122. Oxford: Oxford University Press. Kretzschmar, William A., Jr., Virginia G. McDavid, Theodore K. Lerud & Ellen Johnson. 1993. Handbook of the linguistic atlas of the Middle and South Atlantic States. Chicago: University of Chicago Press [=LAMSAS]. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1. 199–244. Labov, William. 1994. Principles of linguistic change. Vol. 1: Internal factors (Language in Society 20). Oxford: Basil Blackwell. Labov, William. 2001. Principles of linguistic change. Vol. 2: Social factors (Language in Society 29). Oxford: Basil Blackwell. Larsen-Freeman, Diane & Lynne Cameron. 2008. Complex systems and applied linguistics. Oxford: Oxford University Press. Lass, Roger. 1990. How to do things with junk: Exaptation in language evolution. Journal of Linguistics 26. 79–102. Lewis, C. S. 1990. Studies in words. 2nd edn. Cambridge: Cambridge University Press. McIntosh, Angus, Michael L. Samuels & Michael Benskin. 1986. A Linguistic atlas of late mediaeval English. Aberdeen: Aberdeen University Press. Mitchell, Melanie. 2009. Complexity: A guided tour. Oxford: Oxford University Press. Mufwene, Salikoko. 2001. The ecology of language evolution. Cambridge: Cambridge University Press. Mufwene, Salikoko. 2008. Language evolution: Contact, competition, and change. London: Continuum Press. Oxford English dictionary online (OED). Michael Proffitt (ed.). 3rd edn. http://www.oed.com Séguy, J. 1971. La relation entre la distance spatiale et la distance lexicale. Revue de Linguistique Romane 35. 335–357. Smith, Jeremy. 1996. An historical study of English. London: Routledge. Stenroos, Merja, Martti Mäkinen, Simon Horobin & Jeremy J. Smith (compilers). 2011. The Middle English grammar corpus, version 2011.1. January 2011, University of Stavanger. http://www.uis.no/research-and-phd-studies/research-areas/history-languages-andliterature/the-middle-english-scribal-texts-programme/meg-c/ Stenroos, Merja & Kjetil Thengs. 2011. Two Staffordshires: Real and typological space in the study of Late Middle English linguistic variation. Paper presented at Helsinki Corpus Festival, Helsinki. Wright, Laura (ed.). 2000. The development of Standard English 1300–1800. Cambridge: Cambridge University Press. Zipf, George. 1949. Human behavior and the principle of least effort. Cambridge, MA: Addison Wesley.
Colette Moore
An ideological history of the English term onomatopoeia Some of the earliest words that we learn are mimic-words, imitative of the sounds of animals and things. Animal sounds, for instance, are generally considered some of the canonical facts of young child knowledge; they are part of the culture of children’s English. The typical name for these sound words, onomatopoeia, is also part of the general knowledge bank – in a way that is quite surprising for a six-syllable Greek word that virtually no one can spell. Informal inquiry suggests that, when it is remembered, the usual category term for these is onomatopoeia, and not echoism, or imitative lexical construction or other variants. Surprisingly, as old as the phenomenon of language imitating sound is, and as basic as it seems to be in our process of language learning, having an English word for it is, in fact, modern: the first attested uses of onomatopoeia in English occur in the mid-sixteenth century, in the English rhetorics that proliferated during the Early Modern period (Oxford English dictionary [OED], s.v. onomatopoeia, n., def. 1a; see also Lynch 2009: 7).1 In what follows, I am going to examine the shifts in pragmatic needs and attitudes that affected the recorded use of the term in Early Modern English, and our complex relationship with the term in the ensuing centuries. This chapter is not about onomatopoeia in English, then (for this, see Bredin 1996; Liberman 2005: 16–44), but about “onomatopoeia” in English: what we talk about when we talk about onomatopoeia. In Institutio oratoria (ca. 95), the foundational work for classifying and theorizing figurative language, Quintilian comments that the Greeks regarded onomatopoeia (which for him is word creation) as a virtue, whereas the Romans were unaccepting of it (8.6.31). Whether or not Greeks and Romans actually differed in their admiration for the concept, Quintilian’s myth-making demonstrates how the purported difference might index differing cultural perspectives between Greek and Roman societies, and how such ideas about metalanguage have long been taken to be culturally emblematic and to index cultural values. Attitudes about the word onomatopoeia reflect cultural perspectives on the connection between words and sounds, and the story of the
1 Thanks to Philip Lynch, whose class paper, “Some Thoughts on Onomatopoeia”, pointed out to me how late the word entered English. This observation became the starting point of this investigation.
308
Colette Moore
response to the word in English is also the story of our cultural attraction and aversion to the notion that words are created through mimicry of non-linguistic sounds. Lexical change by itself, especially change in the meaning and reception of a single word, is insufficient evidence of broader cultural change, of course, but the shifts in meaning and usage of a single word can be emblematic of broader ideological repositioning (see, for example, C. S. Lewis’s 1960 Studies in Words). This “ideolexical history” – as we might call it – can be a small window into the history of language attitudes and ideologies. To investigate the shifting usage of onomatopoeia and attitudes about it, I constructed a sample of hits in pretwentieth-century English texts using Early English books online, Google books, and the Chadwyck-Healey Literature online database. The data of our past written uses of the word, then, become evidence for interpretative shifts; to analyze this data is to better understand the cultural use of metalanguage. Qualitative analysis of this sample pursues the history of the word in its various contexts and s ituates it in the evolving cultural landscape – a methodology that blends English historical linguistics and the history of English linguistics. The history of onomatopoeia, I find, its usage and its reception, tells a story about attitudes in English and reveals how the word serves as a screen for speakers to project our anxieties about language and language study.
1 The prehistory of onomatopoeia If, as seems evident, sound words are so basic and fundamental, then why is there no English language evidence for a dedicated word to describe them until the Early Modern period? This question was my initial point of entry into the topic, and I do not yet have an altogether satisfactory answer to the question, though its investigation has yielded other results. Premodern English certainly employs a range of sound words of many sorts, as any reader of all of the medi eval tales featuring birds could testify (see also Taavitsainen 1997). The Dictionary of Old English provides some early English examples of sound words: clatrung (a clattering), crachelung (a rattling, used to describe the sound made by storks).2 The Middle English dictionary also provides some of the interjective kind of onomatopoetic words, as well: cok, chuk (the clucking sound of a chicken), bus! (bang!). Neither dictionary or any of the databases, however, provides any evidence for a category term to describe these imitative words. I expected to find “animal sounds” at the very least – though perhaps it makes sense that if there
2 Examples courtesy of Peter Grund and the University of Kansas libraries.
An ideological history of the English term onomatopoeia
309
had been an existing English hypernym, a Greek monstrosity like onomatopoeia would have had a hard time displacing it. It is hard to project how much speakers needed a category term – maybe they did not feel the need for one, or perhaps it was used in spoken language but not used in the surviving English texts, or possibly it is just that the kinds of texts in which it was likely to be used were still being written in Latin. Bede’s Latin work on figures and tropes, De schematibus et tropis (ca. 710?), for example, includes onomatopoeia; this is the first use of the term that I find in a Latin text written in England. Onomatopoeia was in use in classical texts like Quintilian’s handbook that were in circulation (albeit limited circulation) in medieval England; it is unclear how much the Latin term was a known category term by English speakers. Some evidence that it was part of the active Latin word stock is provided by Sir Thomas Elyot’s Latin-English dictionary The Dictionary of Sir Thomas Elyot knight (1538), which includes the term. Elyot’s dictionary is an incomplete lexicon of terms, with a selection of the words most useful to English speakers; it functions almost as a proto-hard words dictionary, anticipating the lists of English specialized words that begin to circulate in the early seventeenth century. Elyot defines onomatopeia as: Onomatopeia, a worde made of sowne or pro|nouncynge, as bletynge of a shepe, lowynge of cowe, hissynge of an adder (Elyot image #96).
The fact that Elyot includes onomatopoeia in his volume suggests that it was perceived as a helpful word to know in Latin. And this restricted use of onomatopoeia in Latin texts seems to be the primary appearance of a metalinguistic term for sound words in premodern English. The rhetorical and grammatical works that might have had most use for such a term are typically written in Latin, and the early English reference works do not seem to contain it. Ælfric’s Grammar, a vernacular grammar of Latin, even contains a list of animal sounds in Latin, paired with English equivalents, but only as examples of verbs that appear exclusively in the third person rather than the first or second person (his list includes, for example: ouis balat, scêp blæ̂t, sus grunnit swîn grunad [129]). He does not include a category term to classify them, and it stands to reason that if such a word were in general English use he would employ it here. If there was a preexisting English hypernym for imitative words, therefore, it does not appear to have been in such widespread use that it can easily be found in our reference works of Old and Middle English.
2 Early modern introduction The OED gives the first attribution of onomatopoeia in English to Thomas Wilson’s The arte of rhetorique (1553), one of the English rhetorics that proliferated in the
310
Colette Moore
sixteenth and seventeenth centuries. Only the first sentence of Wilson’s definition is cited in the dictionary, however: “A worde makinge called of the Grecians Onomatopeia is when we make wordes of oure owne mynde, suche as be deriued from the nature of thinges”. This seems straightforward enough when we consider the OED definition of the term, “the formation of a word from a sound associated with the thing or action being named; the formation of words imitative of sounds”. But the citation in Wilson’s own text is a bit more complex – even confusing: A worde makinge called of the Grecians Onomatopeia is when we make wordes of oure owne mynde, suche as be deriued from the nature of thinges. As to call one Patche or Cowlson, whom we see to do a thinge folyshelye, because these two in their time were notable foles. Or when one is lustye to saye Taratauntara, declaringe therby that he is as lustye, as a Trumpette is delitefull, and styrringe: or when one woulde seme galaunte, to crye hoyghe, whereby also is declared courage. Boyes beynge greued will saye some one to an other, Sir I wyll cappe you, if you vse me thus, and withholde that frome me whyche is myne owne: meanynge that he will take his cappe from him. Againe, when we see one gaye and galaunte, we vse to saye, he courtes it. Quod one that reasoned in diuinitie wyth his felowe, I like well to reason, but I can not chappe these textes in scripture, if I shoulde dye for it: meaning that he coulde not tell in what chapiter thinges were conteyned, althoughe he knewe full well that there were suche sayinges (1553 Wilson, The arte of rhetorique f 92 v).
Wilson’s definition seems to include kinds of word formation and stylistic coinage that are not at all what we would presently consider onomatopoeia. His examples contain all manner of creative lexical usage: nouns coming from names, interjections, idioms, category shifting. Although this expansiveness seems surprising to us (and the OED sets aside the broader purview of his usage), the etymological sense of onomatopoeia as word creation comes directly from Quintilian, who describes the term as creating words ex novo – although, interestingly, after defining the term this way, he goes on to illustrate it with a list of examples that are all sound words (8.6.31). We also see vestiges of this broader sense in the use of onomatopoeia in George Puttenham’s Arte of English poesie (1589); his anglicization also includes all new coinages. Puttenham’s rhetorical text is distinguished not only for its list of figures, tropes, and schemes (which all of the rhetorics have), but by his attempt to anglicize all of the terms in order to “make English” these classical methods for ordering and classifying poetic language. Puttenham’s text provides an interesting perspective on word borrowing when read down the center margins; the headings provide a list of how English might repackage all of the classical terms for figures, schemes, and tropes. (Hyperbole, for example, is renamed the Overreacher [3.18.276], and micterismus is the Fleering Frump [3.18.275].) To anglicize onomatopoeia, he rechristens it the New namer:
An ideological history of the English term onomatopoeia
311
Onomatopeia or the New namer Then also is the sence figurative when we devise a new name to any thing consonant, as neere as we can to the nature thereof, as to say: flashing of lightening, clashing of blades, clinking of fetters, chinking of mony; & as the poet Vergil said of the sounding a trumpet, ta-ra-tant, taratantara, or we give special names to the voices of dombe beasts, as to say, a horse neigheth, a lyo[n] brayes, a swine grunts, a hen cackleth, a dogge howles, and a hundreth mo such new names as any man hath libertie to devise, so it be fittie for the thing which he covets to expresse (1589 Puttenham, Arte of English poesie 251–252).
Puttenham’s anglicization of onomatopoeia as “the new namer” invokes the broader sense of its role in word formation. But his discussion of the concept goes on to describe it as a process of word formation creating new words linked to the nature of the thing described, and his examples are all sound words. He seems to be pulling the word in the narrower direction of sound words, but it is the notion of word creation that anchors his approach. The sense of onomatopoeia as word creation overlaps in interesting ways with ideologies of word borrowing and word origins in the period. Borrowing onomatopoeia, for example, taps into the famous “inkhorn terms” conundrum about whether large numbers of specialized foreign terms should be imported into English. The treatises and rhetorical texts of the Early Modern period reveal somewhat contradictory impulses in the debate: on the one hand, they intro duced these words in part to promote English as a language fit for scholarship, but on the other, they engaged in large-scale borrowing to accomplish this. Tradi tional accounts of the history of the English language describe the early modern cultural chafing at the dominance of classical and European languages (see, for example, Millward and Hayes 2012: 225). One of the central works quoted in the textbooks, in fact, is Thomas Wilson’s The arte of rhetorique (1553), the same work that provides OED’s first cited usage of onomatopoeia. The linguistic insecurity of the period stems in part from the widespread belief in the deriv ative nature of many English words, and the feeling that English represents a latter-day corruption of earlier languages. Onomatopoeia, in this light, describes a force for reinvention, for freshness in language – a blow for natural over artificial reference, a blow on the side of the inventiveness of English speakers and against the wooden acceptance of terms from ancestral languages. The ambiguity in the rhetorics, then, over whether onomatopoeia refers narrowly to sound words or broadly to any new word is significant – and it is especially notable that even the name for such a phenomenon is, ironically, derived from classical languages. Fascinating, too, is the fact that the first recorded use of onomatopoeia should be attributed to these rhetorics: that onomatopoeia came into English hand in hand
312
Colette Moore
with scores of classical terms for the figures, tropes, and schemes used to analyze literary and poetic language: anadiplosis, epizeuxis, etc. The educated reader of the Early Modern period appears to have embraced the figure-hunting mindset indicated in (or encouraged by) these rhetorics – we have existing texts of, for example, Sidney’s Arcadia in which all of the figures and tropes are marked in the margin (see frontispiece to Adamson, Alexander, and Ettenhuber [eds.] 2007). The creation of these rhetorics, therefore, became a cottage industry in the language classification of the Early Modern period, providing tools for English literary readers to evaluate figures, schemes, and tropes. (Henry Peacham in the Garden of eloquence [1577], for instance, listed nearly 200 of these.) The Early Modern drive to improve and organize English, to set it out as a language worthy for literary art and for literary analysis, was a powerful motivation for the importing of metalanguage, therefore, and this is the ideological wave that onomatopoeia rode in on. Only these first rhetorics, however, employ onomatopoeia in the broad sense that includes all kinds of word creation; the treatises and dictionaries that employ it in the sixteenth and seventeenth centuries invoke its more current meaning of sound-imitative words. The hard-word dictionaries tend to give variations of a similar formulation: Onamatopeious, pertain|ing to the Figure Onomatopoeia, which is a faining a name from any kind of sound (1661 Blount, Glossographia). Onomatopy, -paeia, g. the feigning of a name from some kind of sound (1677 Coles, An English dictionary explaining…). Onomatopoea, (Greek) the feigning of a name from any kind of sound, as Bombarda, i.e. a Gun, from the sounding of Bom. (1678 Phillips, The new world of words)
The consensus in the adoption of the word, then, is the narrower sense of the creation of new words to mimic sounds. We can see from the inclusion of onomatopoeia in these texts, too, how the word was conceived: a “hard” word, part of the “new world of words” being imported into English. But it shows, too, that these new analytic specialized terms are often fancy classical dressing for old phenomena. Onomatopoeia takes on a fascinating double role with respect to sixteenthand seventeenth-century language ideas. On the one hand, its borrowing is consonant with a “new antiquity” in metalinguistic terms: the importation of large numbers of classical rhetorical terms. (These are imported by the same rhetorics that decry the over-importation of “inkhorn terms”.) On the other hand, the word itself describes word creation through a native process (or “native-seeming process”, more accurately). Either way, the introduction of onomatopoeia to high-register classificatory texts shows that (despite Puttenham’s best efforts at
An ideological history of the English term onomatopoeia
313
anglicization) metalinguistic discussion in English is still accomplished in terms that are markedly non-English.
3 Eighteenth- and nineteenth-century linguistics In the eighteenth century, uses of onomatopoeia begin to appear in different kinds of texts, texts with names like The logographic-emblematic English spelling book; or a method of teaching children to read (Lenoir 1800) and Of the origin and progress of language (Monboddo 1774). Usage shifts, then, as texts that are classificatory in nature and intended for use in literary analysis begin to share the ground with texts that are theoretical, especially texts dealing with the beginnings of language or systems of pedagogy. Perhaps it is not surprising that onomatopoeia gets conscripted into the debates on the origins of language and on word etymology: it is a category term which groups words together based on their shared relationship to sound (and, thus, on a seeming independence from ancestor languages). It is only one more logical step to imagine that every word was derived from sound at an earlier stage of language development. The increased number and variety of hits in the databases in the late seventeenth and eighteenth centuries also suggests that the word is moving towards part of an assumed base of knowledge – at least among educated readers. Unlike its peers in the rhetorics (figures like anadiplosis and epizeuxis), onomatopoeia has expanded to a broader usage, and we start to see more general instances. Not only do the eighteenth-century citations demonstrate the diffusion of onomatopoeia into the general lexicon of educated writers, but they also show us the entry of onomatopoeia into conversations in the early field of linguistics, the systematizing of thinking about comparative language that gains momentum in the nineteenth century. The philologist Max Müller in 1851 credits Johann Gottfried Herder and Jean-Jacques Rousseau with suggesting onomatopoeia as the origination of language, and this notion is taken up in nineteenth-century texts such as Hensleigh Wedgwood’s “On onomatopoeia” (1845) and On the origin of language (1866), William Dwight Whitney’s Language and the study of language (1870), and by others such as philologist Heymann Steinthal, who asserted that it is inconceivable that anyone could deny that onomatopoeia was the primeval tendency of language that has furnished us with all elementary words (Steinthal 1855: 309, cited in Farrar 1865: 163). The mid-nineteenth century saw active debate over the question, but Müller mentioned the notion of onomatopoeia in language origins in order to challenge it; he called it the “bow wow theory” of language origin (1861: 344), and sniffily remarked: “I doubt whether it deserves the name of language” (1861: 346). Müller’s dismissal seems to have
314
Colette Moore
ultimately prevailed in the discourse (Liberman calls him “at one time the most popular linguist in the world” [2005: 42]), convincing many of his readers that scientifically-minded thinkers about language did not take seriously an originating hypothesis for language built on sound words. And yet I found more than one discussion that cited Müller as the source that convinced them of the bow wow theory; a review in the Educational Times, and Journal of the College of Preceptors (Anon. 1861), for example, reads, “We confess that we are better satisfied with what he says against them [theories of language origins], than with what he gives us in their place”. The importance of Müller’s text and the investigation of language origins also shows another fault line for nineteenth-century cultural anxiety: the cultural place of the Bible. One reviewer for the Westminster Review in 1865 points out that such theories imply that language originates in primitive pre-linguistic impulses, rather than emerging all at once for Adam and Eve as a fully-formed and sophisticated system: The popular theology of Christendom asserts that man started into being in the full perfection of his mental powers and with the privilege of an immediate intercourse with his Maker. The science of language has shown that man may at the first have been mute, and certainly that he was unable, during a long period, to express more than the merest bodily sensations (Anon. 1865: 62).
The reviewer goes on to suggest that philology troubles a naïve, literal interpretation of Genesis, and in doing so it points to the need for science to assist in developing a more sophisticated rendering of Biblical narratives. Many thinkers of the period were clearly struggling to reconcile the prevalence of Biblical metaphysical models with considerations of language origins. F. W. Farrar writes: If we consider any number of names for animals in any modern language, we shall find that they fall into various classes, viz: 1. Those for which no certain derivation can be suggested; 2. Those derived from some analogy, or characteristic, or combination of characteristics which the animal presents; 3. Those which are distinctly onomatopoetic in origin or in form. The first class of words cannot of course furnish us with any linguistic inferences, and may here be left out of the question; under the second and third classes fall all names of recent origin; and if, as the Bible asserts, and as has been shown to be independently probable, animals were the first objects to receive names, they MUST have received names belonging to the third class (viz.: onomatopoeias), because no previous words would have existed wherewith to designate or combine their observed qualities (Farrar 1865: 23–24).
Farrar argues that since animals were the first things to receive names, their names must have been onomatopoeic, because there were no other words to work with. We see onomatopoeia, therefore, also touching upon Victorian anxiety
An ideological history of the English term onomatopoeia
315
about how literally the Bible is to be read and how to integrate Genesis accounts of creation with models emerging in geologic, biological, and linguistic science. Onomatopoeia seems to have its finger firmly on the intellectual pulse of the Victorian period, just as it did in earlier centuries. Another influence of Max Müller’s expressed hesitations about onomatopoeia as a theory of linguistic origins, too (exceeding Müller’s own rather gentle critique), was the emerging set of connotations that onomatopoeia developed in the late nineteenth century of being a pseudo-scientific perspective for the armchair scholar. An author in the New York publication the Continental Monthly (1864) evidences the scholarly dismissiveness that was clearly beginning to characterize attitudes towards onomatopoeia: This mere imitation is what Lingual Scholars have dignified by the high-sounding and rather repulsive technicality, Onomatopoieia. In the early and simple period of Lingual Science much has been made, in striving to account for the Origin of Language, of this faculty of imitation, and of the fact that there are undoubtedly certain words in every language consisting of such imitations. It is against this simple and superficial theory that Professor Müller has argued so well (Anon. 1864: 467).
The sneering tone here (“high-sounding and rather repulsive technicality”, “simple and superficial theory”) indicates how the connotations of onomatopoeia are developing in the later nineteenth century: the term begins to sound at once overly elevated (a pretense to technical discourse) and overly naïve (a shibboleth for the simple and unsophisticated approaches to early theories of language). We see in these connotations the beginning of the avoidance of the term. Perhaps an aversion to onomatopoeia is not surprising: the term is 11–13 letters long (depend ing on whom you ask), hard to spell, hard to derive forms from (onomatopoeic? onomatopoetic?), and now, additionally, burdened with assumptions that it is incompatible with serious scholarship. Writers and scholars responded to these connotative shifts in different ways. I found one attempt to make the word more palatable to English speakers by shortening it: an 1874 text by Callistus Augustus Goddes-Liancourt and Frederic Pincott clips it to onomatop. The authors provide the following introduction as they set out their project of creating a dictionary of onomatops (perhaps an alternate spelling of “onomatopes”?): The inconvenient length of the old term, on the one hand, and the desire to avoid the affectation of coining an altogether new word, on the other hand, have induced us to cut off boldly the latter portion of the word Onomatopoiëia, and to reduce it to the more wieldly proportions of Onomatop. (…) Onomatops have escaped the convulsions which have agitated the globe, and the revolutions which have again and again remodelled society, because they are fundamental and eternal principles. (…) The special object of writing this first Dictionary of Onomatops is to show, that we must look to nature only for the bonds
316
Colette Moore
uniting all languages together; and in adverting to the numerous affinities or analogies connecting languages, it is hoped that the proof of their true origin will be demonstrated. (Goddes-Liancourt and Pincott 1874: 54–55).
This text – Primitive and universal laws for the formation and development of language: A rational and inductive system founded on the natural basis of onomatops – typifies many of the characteristic issues in later nineteenthcentury studies. The second sentence holds up the “onomatop” as a beacon of stability in a mid-nineteenth-century world of revolution (“the convulsions which have agitated the globe”). And the third sentence grounds the project of the dictionary of onomatops in the attempt to search for a universal language, to find bonds of nature between languages. The link between onomatopoeia and the origins of language ultimately weakens the scholarly reputation of onomatopoeia, since the project of searching for the origins of language became suspect and associated with amateurism – the Linguistic Society of Paris, for example, banned debate about the origins of language in 1866, and this prohibition influenced other scholarly perspectives for a century (Stam 1976: 255). Nineteenth-century lexicographical practice was also shaped by these developing attitudes. James Murray, the editor of the OED, attempted to avoid the term onomatopoeia; in the etymologies of the OED, he used echoic or echoism. The source of the first citation for the word echoic (OED, s.v. echoic, adj.), in fact, is given in the dictionary to Murray himself, making the term a sort of a lexicographic embedded loop. The definition reads: “Of the nature of an echo: a term proposed by J. A. H. Murray and used in this Dictionary to describe formations which echo the sound which they are intended to denote or symbolize”. Other scholarly dictionaries follow Murray’s lead and avoid onomatopoeia in etymologies. Even William Dwight Whitney’s own The Century dictionary (1891) included in its entry for onomatopoeia that the dictionary’s etymological practice is to use the terms imitation, imitative, or imitative variation; the entry even went so far as to quote Murray addressing the Philological Society, saying, “onomatopoeia [as a word], in addition to its awkwardness, has neither associative nor etymological application to words imitating sounds” (The Century dictionary 1891: s.v., onomatopoeia, n). The contemporaneous reviews and the practices employed by nineteenth-century lexicographers, then, gesture towards the recalibration in register that onomatopoeia underwent in the latter half of the nineteenth century; they reinforce the perception that onomatopoeia declined in scholarly prestige, even as it developed general popularity.
An ideological history of the English term onomatopoeia
317
4 Analysis: Register shifts The shifting usage of the word onomatopoeia indexes the shifts in English metalanguage. The word came into English through Early Modern rhetorics, which borrowed it from classical texts for use in the development of an English metalanguage to analyze literary language. Then, as language analysis moved from classification to theorization of language in the eighteenth century, onomatopoeia assumed an important role in conversations about language origin. Later in the nineteenth century, the word assumed a notoriety among philologists for this alleged role in language origins, and language scholars (including lexicographers) began to shy away from it. The lexical history of onomatopoeia, then, shows a shifting relationship between registers of text. The word (which I had naïvely assumed to be broadly useful across spoken and written registers) came into the language through the elevated written register of Renaissance rhetoric, was taken up by the elevated written register of language scholarship, and then began to make its way across a broader spectrum of spoken and written registers, eventually becoming part of common knowledge. Meanwhile, its stock began to fall in language scholarship – to the point where some began to avoid the term as connoting a lack of rigor (as we saw in the reviewer quoted in text 9), and others began to favor different terms that avoided the ambiguity between its imitative sense and its earlier etymological sense of “new name”. As a result, although onomatopoeia never truly lost its presence in academic discourse (especially in the analysis of poetry), it became no longer a markedly academic word, despite its fancy pedigree. Of the 37 hits for the word in The corpus of contemporary American English (Davies 2008–), for example, only 13 are from academic sources. The usage of onomatopoeia indicates the shifting currents of language analysis as the field became institutionalized and professionalized in the eighteenth and nineteenth centuries. The word changed as our aims for language analysis changed – from classification to theorization. And its connotations changed differently in different registers as scholarly vocabulary began to be increasingly separate from broader discourse.
5 Conclusion This study concludes in the year 1900. Following onomatopoeia into the twentieth-century conversations on language, however, would suggest that links to these past perspectives often influence its use. Its place in traditional English
318
Colette Moore
textbooks of basic linguistics (for example, Fromkin, Rodman, and Hyams 2003: 7), is typically as background; the word appears in an early section mentioning onomatopoeia as a seeming exception to the principle of the arbitrariness of the sign – “seeming” because while the impulse may be non-arbitrarily imitative, the manifestations are linguistically and culturally conditioned. Perhaps the concept of onomatopoeia engenders some discomfort in a field built on Ferdinand de Saussure’s model of the arbitrariness of the linguistic sign. Certainly critics of Saussure have lit upon onomatopoeia; it became one of the sticks that Jacques Derrida used to beat Saussure in Glas (1974). Derrida took as a starting point Saussure’s example of French glas (‘knell’) as a term that is reputed to be onomatopoeic, but is actually derived from Latin classicum. Derrida suggested, among other things, that Saussure’s assumptions about onomatopoeia are flawed, that word origins are constructed, and that onomatopoeia is a function that can be grafted onto a word – and this discussion has been understood to show the limitations of structural linguistics (Joseph, Love and Taylor: 2001). Otto Jespersen, of course, had earlier criticized Saussure along partly similar lines about his antiquarian attachment to linguistic tenets hinged on historical etymology (Jespersen 1922: ch. 20 §12, see also Jakobson and Waugh 1987 [1979]: 186). Even as the word onomatopoeia became more a part of common parlance, then, it was less the choice for new linguistic research, which has opted more recently for other terms – sound symbolism, phonoaesthetics, ideophonology, and phonosemantics are twentieth-century keywords to investigating the connection of sound and language. Perhaps the lingering nineteenth-century disenchantment with onomatopoeia inhibited the use of the word in upperregister linguistic research in favor of new scholarly word formations. The word onomatopoeia acts as a kind of weather vane, showing us which way the winds of English language analysis are blowing. The path of onomatopoeia, then, is an interesting journey from a markedly high-register word to a common word abandoned by the high register – and the travel narrative of this journey shows us a little bit about our shifting cultural ideologies, about the place of metalanguage, and about our changing linguistic goals and changing linguistic prejudices.
Dictionaries and databases The Century dictionary. 1891. William Dwight Whitney (ed.). New York: The Century Co. http://www.micmap.org/dicfro/introduction/century-dictionary (accessed 7 May 2014). Davies, Mark. 2008–. The corpus of contemporary American English: 450 million words, 1990–present. http://corpus.byu.edu/coca/ (accessed 17 January 2012).
An ideological history of the English term onomatopoeia
319
Dictionary of Old English: A to G online. 2007. Antonette diPaolo Healey (ed.). Toronto: Pontifical Institute of Mediaeval Studies. http://www.doe.utoronto.ca/ (accessed 9 April 2012). Early English books online. http://eebo.chadwyck.com/home (accessed 17 January 2012). Eighteenth-century collections online. http://gdc.gale.com/products/eighteenth-centurycollections-online/ (accessed 17 January 2012). Google books. http://books.google.com/ (accessed 17 January 2012). Literature online. Chadwyck-Healey, through ProQuest LLC. http://literature.proquest.com (accessed 17 January 2012). Middle English dictionary. University of Michigan Press. quod.lib.umich.edu/m/med (accessed 17 January 2012). Oxford English dictionary online. Michael Proffitt (ed.). 3rd edn. Oxford: Oxford University Press. www.oed.com (accessed 17 January 2012).
References Adamson, Sylvia, Gavin Alexander & Katrin Ettenhuber (eds.). 2007. Renaissance figures of speech. Cambridge: Cambridge University Press. Ælfric. 1880 [ca. 995]. Ælfrics Grammatik und glossar. Julius Zupitza (ed.). Berlin: Weidmannsche Buchhandlung. Cited from Open Library, https://openlibrary.org/books/ OL14010200M/Aelfrics_Grammatik_und_Glossar (accessed 5 May 2014). Anonymous. 1861. Review: Lectures on the science of language delivered at the Royal Institution of Great Britain in April, May, and June 1861, by Max Müller. The Educational Times, and Journal of the College of Preceptors 2. London: F. Hodgson. Anonymous. 1864. The scientific universal language: Its character and relation to other languages. Continental Monthly 6(4). 456–468. Anonymous. 1865. Art II – the science of language. Review: Lectures on the science of language, delivered at the Royal Institution of Great Britain, by Max Müller. Westminster Review 27(1). 35–64. Bede. 1991 [ca. 710?]. Libri II De arta metrica et De schematibus et tropis (The art of poetry and rhetoric). Calvin B. Kendall (ed). Saarbrücken: AQ-Verlag. Blount, Thomas. 1661. Glossographia, or, A dictionary interpreting all such hard words of whatsoever language now used in our refined English tongue with etymologies, definitions and historical observations on the same: Also the terms of divinity, law, physick, mathematicks and other arts and sciences explicated. Cited from Early English books online (accessed 17 January 2012). Bredin, Hugh. 1996. Onomatopoeia as a figure and a linguistic principle. New Literary History 27(3). 555–569. Coles, Elisha. 1677. An English dictionary explaining the difficult terms that are used in divinity, husbandry, physick, phylosophy, law, navigation, mathematicks, and other arts and sciences, containing many thousands of hard words, and proper names of places, more than are in any other English dictionary or expositor: Together with the etymological derivation of them from their proper fountains, whether Hebrew, Greek, Latin, French, or any other language: In a method more comprehensive than any that is extant. Copy from Bodleian Library. Cited from Early English books online (accessed 17 January 2012).
320
Colette Moore
Derrida, Jacques. 1986 [1974]. Glas. John P. Leavey, Jr. & Richard Rand (trans.). Lincoln: University of Nebraska Press. Elyot, Thomas. 1538. The dictionary of Syr Thomas Eliot knyght. British Library. Cited from Early English books online. Farrar, Frederic William. 1865. Chapters on language. London: Longmans, Green and Co. Fromkin, Victoria, Robert Rodman & Nina Hyams. 2003. An introduction to language. 7th edn. Boston: Wadsworth, Thomson. Goddes-Liancourt, Callistus Augustus (comte de), and Frederic Pincott. 1874. Primitive and universal laws for the formation and development of language: A rational and inductive system founded on the natural basis of onomatops. London: W. H. Allen. Jakobson, Roman & Linda Waugh. 1987 [1979]. The sound shape of language. 2nd edn. Bloomington, IN: Indiana University Press. Jespersen, Otto. 1922. Language: Its nature, development and origin. New York: H. Holt. Joseph, John E., Nigel Love & Talbot J. Taylor. 2001. Landmarks in linguistic thought II. New York: Routledge. Lenoir, P. V. 1800. The logographic-emblematical English spelling book; or, a method of teaching children to read. Being founded upon an entirely new principle (...). London. Eighteenth-century collections online (accessed 17 January 2012). Lewis, C. S. 1960. Studies in words. Cambridge: Cambridge University Press. Liberman, Anatoly. 2005. Word origins ... And how we know them: Etymology for everyone. Oxford: Oxford University Press. Lynch, Philip. 2009. Some thoughts on onomatopoeia. Final course paper for English 569, University of Washington. Millward, Celia M. & Mary Hayes. 2012. A biography of the English language. 3rd edn. Boston: Wadsworth, Cengage Learning. Monboddo, James Burnet. 1774. Of the origin and progress of language. Edinburgh and London: Printed for J. Balfour and T. Cadell. Müller, Friedrich Max. 1861. Lectures on the science of language delivered at the Royal Institute in Britain in April, May, and June, 1861. London: Longman, Green, Longman, and Roberts. Peacham, Henry. 1971 [1577]. The garden of eloquence. Menston: Scolar Press. Phillips, Edward. 1678. The nevv vvorld of vvords. Or a general English dictionary containing the proper significations, and etymologies of all words derived from other languages, viz. Hebrew, Arabick, Syriack, Greek, Latin, Italian, French, Spanish, Brittish, Dutch, Saxon, &c. useful for the adornment of our English tongue. Together with the definitions of all those terms that conduce to the understanding of any of the arts or sciences, viz. theology, philosophy, logick, rhetorick, grammar, ethicks, law, magick, physick, chirurgery, anatomy, chymistry, botanicks, arithmetick, geometry, (...) & c. To which are added, the significations of proper names, derived from the ancient or modern tongues; as also the very sum of all mythology and ancient history, deduced from the names of persons eminent in either; and likewise the geographical discriptions of the chief countries and cities in the world; (...) The fourth edition. Containing besides an addition of several thousand words in the abovementioned terms, a brief view of the most eminent persons of the ancients (...). London: printed by W.R. for Robert Harford at the Angel in Cornhill. Cited from Early English books online (accessed 17 January 2012). Puttenham, George. 1589. Arte of English poesie. Cited from Early English books online. Quintilian. 1922 [ca. 95]. Institutio oratoria, with an English translation. Bk. 8. Harold Edgeworth Butler (ed.). Cambridge, MA: Harvard University Press.
An ideological history of the English term onomatopoeia
321
Stam, James H. 1976. Inquiries into the origins of language: The fate of a question. New York: Harper and Row, Ltd. Steinthal, Heymann. 1855. Grammatik, Logik, und Psychologie, ihre Principien und ihr Verhältniss zu einander. Berlin: Ferd. Dümmler’s Verlagsbuchhandlung. Taavitsainen, Irma. 1997. Exclamations in Late Middle English. In Jacek Fisiak (ed.), Studies in Middle English linguistics, 573–607. Berlin: Mouton de Gruyter. Wedgwood, Hensleigh. 1845. On onomatopoeia. Philological Society 2. 34. Wedgwood, Hensleigh. 1866. On the origin of language. London: N. Trübner & Co. Whitney, William Dwight. 1870. Language and the study of language. Twelve lectures on the principles of linguistic science. London: N. Trübner & Co. Wilson, Thomas. 1553. The arte of rhetorique for the vse of all suche as are studious of eloquence, sette forth in English. London: Richardus Graftonus. Copy from Huntington Library. Cited from Early English books online (accessed 17 January 2012).
Name index Abner, Natasha 70 Adams, Michael 1–12, 68, 107, 118, 245 Adamson, Sylvia 312, 319 Aertsen, Henk 179 Aiden, Aviva Presser 96 Aiden, Erez Lieberman 96 Aijmer, Karin 99, 111, 119 Akimoto, Minoji 70, 84, 95 Aldrich, William S. 90 Alexander, Gavin 312, 319 Allan, Kathryn 250, 276 Amodio, Mark C. 217 Amos, Ashley Crandall 154, 158, 220, 245 Anderson, Earl R. 209–210, 216 Anderson, John 138, 156, 158 Antieau, Lamont D. 10–11 Asprey, Esther 184, 186, 191–192, 197, 199 Athanasiadou, Angeliki 119 Auer, Anita 274 Auer, Peter 120, 167, 171, 179 Baayen, R. Harald 64, 69 Baker, William J. 251, 256–264, 269, 271, 276 Baltin, Mark 58, 69 Bammesberger, Alfred 246 Banks, Ronald Alfred 221, 245 Barnes, Geraldine 216 Barrack, Charles 129, 131 Barry, Michael F. 183, 195, 200 Barton, Bernard 147 Bately, Janet M. 128, 131 Bauer, Gero 132, 160 Bauer, Renate 246 Beal, Joan 118 Bechert, Johannes 49 Becker, Thomas 164, 179 Beckner, Clay 290, 304 Bede 237, 309, 319 Behaghel, Otto 57, 69 Beinhoff, Bettina 274 Benskin, Michael 187, 199, 297, 305 Bergs, Alexander 2, 6, 9–10, 249–276
Bermúdez-Otero, Ricardo 139, 143, 158, 160, 167, 179 Bernini, Giulano 49 Bever, Thomas G. 252, 257–258, 274, 276 Biber, Douglas 59, 69, 71, 76–77, 81, 95, 102, 119 Bisang, Walter 119 Bishop, Jason 70 Bjork, Robert E. 160, 206, 216, 269, 275 Blake, N. F. 69, 159, 211, 216 Blatt, Franz 17, 47 Bliss, A. J. 214, 216 Blount, Thomas 312, 319 Blythe, Richard 304 Bod, Rens 133 Boehm-Jernigan, Heather 54, 69 Boland, Julie E. 54, 69 Bolinger, Dwight L. 116, 119 Bolton, W. F. 207, 216 Bourassa, Derrick 156, 160 Bowey, Judith A. 156, 160 Braakensiek, Johan 47 Brandt, Bruce E. 216 Bredehoft, Thomas A. 201, 216 Bredin, Hugh 307, 319 Bresnan, Joan 51, 69–70 Bridgman, Percy 3–4 Brinton, Laurel J. 52, 56–58, 68, 69–70, 84, 95, 114, 118–119, 245, 276 Britton, Derek 156, 158 Brockman, William 96 Brodeur, Arthur Glichrist 205–207, 216 Brown, David West 5–6, 71–97 Brown, Esther 125, 131 Brown, Phyllis Rugg 216 Bülbring, Karl Daniel 154, 158, 179 Burghauser, Gustav 171, 179 Buridant, Claude 49 Burkette, Allison 10–11 Burzio, Luigi 138, 156, 158 Butler, Harold Edgworth 320 Butler, Peter E. 196, 199 Bybee, Joan 113, 119, 250, 274, 290, 304
324
Name index
Cable, Thomas 201, 213, 216 Caha, Pavel 67, 69 Cain, Christopher M. 180, 201, 205, 216 Callaway, Morgan 16, 32, 47 Cameron, Angus 154, 158, 220, 245 Cameron, Lynne 291–292, 305 Campbell, Alistair 124, 131, 132, 139, 154, 156, 158 Campbell, Jackson J. 237, 246 Campbell, James 210, 216 Canakis, Costas 119 Carr, Gerald F. 158, 180 Carroll, Mary 46–47 Casad, Eugene H. 121 Castillo, Concha 76, 95 Celle, Agnès 119 Chafe, Wallace 104, 119, 252, 258, 274 Chapman, Don 8–9, 219–246 Christiansen, Morton H. 290, 304 Church, Kenneth Ward 83, 95 Clanchy, Michael T. 252, 274 Clancy, Dan 96 Claridge, Claudia 52–53, 56–57, 59, 64–65, 69, 73, 95 Clark, Eve V. 256–257, 274 Clark, Herbert H. 252, 256–257, 274 Clark, Peter 199 Clark, Urszula 184, 186, 191–192, 197, 199 Collins, Peter C. 101, 119 Coles, Elisha 312, 319 Colman, Fran 125, 132 Conde-Silvestre, Juan Camilo 274 Conrad, Susan 59, 69, 71, 76–77, 81, 95, 102, 119 Cooper, Janet 216 Cornille, Bert 119 Crampton, Georgia Ronan 216 Cristofaro, Sonia 32, 47 Croft, William 290, 304 Cruse, D. Alan 56, 69 Cruttenden, Alan 137, 155, 158 Culpeper, Jonathan 118 Cumming, Susanna 99, 104, 119 Curzan, Anne 99, 116, 119 Cuyckens, Hubert 5–6, 15–49, 118, 121
Davidse, Kristin 121 Davies, Mark 5, 11, 71, 81, 95–96, 99, 119, 274, 286–287, 304, 317 Davis, Craig R. 210, 216 Davis, Garry W. 132 Davis, Norman 262, 266, 274–275 Davis, Roland C. 173, 179, 181 Dasher, Richard B. 250, 276 Declerck, Renaat 16, 47 Dempsey, Kyle B. 77, 95 den Hertog, Cornelis H. 17, 47 Denison, David 57–59, 69, 160 Derrida, Jacques 318, 320 De Smet, Hendrik 41–43, 47–48, 118 Devriendt, Betty 182 Diensberg, Bernhard 173, 179 Diewald, Gabriele 48 DiPaolo, Marianne 124, 132 Dirven, René 120, 250, 275 Dobson, E. J. 161, 171, 179 Docherty, Gerard J. 199 Dollinger, Stefan 3, 11 Dorenkamp, John H. 217 Dorgeloh, Heidrun 120 Dresher, B. Elan 167, 177, 180 Dubenion-Smith, Shannon 181 Dyer, Alan 191–192, 199
D’Arcy, Anne Marie 217, 246 David, Jan Baptist 37–38, 47
Faber, Alice 124, 132 Fabiszak, Malgorzata 250, 275
Earle, John 203, 217 Eddington, David 156, 158 Edwards, Jane A. 119 Einstein, Albert 288 Eldredge, Niles 298–300, 303, 304 Eliason, Norman E. 173, 179, 181 Ellis, Alexander J. 183, 192–193, 195–199 Ellis, Nick C. 290, 304 Elspass, Stephan 48 Elyot, Thomas 309, 320 Elzinga Dirk 156, 158 Emsley, Clive 69 Englebretson, Robert 119 Espesser, Robert 138, 160 Ettenhuber, Katrin 312, 319 Ewen, Colin 158
Fanego, Teresa 41, 48, 95 Farrar, Frederic William 313–314, 320 Fasold, Ralph W. 48 Featherston, Sam 69 Fellbaum, Christiana D. 53, 70 Ferguson, Charles A. 276 Field, John 256, 275 Finegan, Edward 59, 69, 71, 76–77, 81, 95, 102, 119 Fischer, Claude S. 87, 91, 96 Fischer, Olga 57, 69, 255, 275 Fisiak, Jacek 160, 169, 179, 182, 276, 321 Fitzmaurice, Susan M. 2, 11 Fodor, Janet Dean 257, 275 Fonteyn, Lauren 5–6, 15–49 Ford, Marilyn 51, 70 Foulkes, Paul 199 Foxcroft, Tina 183, 200 Frank, Roberta 209, 216 Frantzen, Allen J. 216–217 Frazier, Lyn 257, 275 Freedle, Roy O. 274 Frese, Dolores Warwick 210, 216 Fried, Vilém 182 Fries, Charles C. 262, 269, 275 Fromkin, Victoria 318, 320 Fulcher, Roseanne 171, 180 Fulk, Robert D. 68, 118, 126, 132, 138–139, 142, 145, 147–150, 156, 158–159, 165, 179, 201, 205–206, 209, 213, 216–217, 245, 269, 275 Gardner, Dee 81, 96 Garman, Michael 256, 275 Garrett, Martin 80, 96 George, Paul S. 183, 199 Gerritsen, Marinel 129, 132 Giegerich, Heinz 138, 158 Gneuss, Helmut 219, 246 Goddes-Liancourt, C. A. 315–316, 320 Goh, Gwang-Yoon 57, 70 Goodwin, William 151, 158 Goody, Jack 252, 275 Gordon, Matthew 138, 141, 158 Gould, Stephen J. 280, 298–300, 303–304 Grandy, Richard E. 4–5, 11 Gray, Matthew K. 96
Name index
325
Greenbaum, Sidney 56, 70, 101, 120 Gretsch, Mechthild 222, 229, 237, 241, 243, 246 Gries, Stefan Th. 74, 96 Grimshaw, Jane 53, 70 Grund, Peter 308 Grundy, Lynne 222, 246 Gunn, John 216 Gussenhoven, Carlos 133 Gussman, Edmund 138, 158 Gwara, Scott 151, 159 Haiman, John 48–49 Hale, William Gardner 152, 158 Haley, Michael C. 116, 120 Hall, Stefan Thomas 204, 216 Halliday, Wilfrid J. 183, 194–195, 200 Hall-Lew, Lauren 124, 132 Hammond, Michael 138, 141, 159 Hanks, Patrick 83, 95 Hanson, Kristin 159 Hardie, Andrew 250–251, 276 Harris, John 129, 132 Harris, Stephen J. 210, 216 Hart, J. M. 169–170, 179 Hartman, Megan E. 8, 201–217, 245 Haspelmath, Martin 16, 48 Haviland, Susan E. 252, 274 Hawking, Stephen 278, 280, 288, 304 Hawkins, John 51, 54–56, 58–59, 67, 70 Hay, Jennifer 51, 69, 133 Hayes, Bruce 138, 143–144, 146–147, 149, 159 Hayes, John R. 274 Hayes, Mary 311, 320 Healey, Antonette diPaolo 69, 123, 125, 132, 158, 220, 231, 245–246, 319 Hebda, Anna 171, 179 Heine, Bernd 28, 48–49, 120, 275 Hernández-Campoy, Manuel 274 Hesselman, Bengt 176–177 Heyvaert, Liesbet 20, 47–48 Hill, David 186–187, 199 Hill, John M. 204, 209–211, 216 Hills, William H. 80, 96 Hilpert, Martin 74, 96 Himmelmann, Nikolaus P. 117, 119 Hitchcock, Tim 69
326
Name index
Hockett, Charles 163 Hoffmann, Thomas 58, 70, 267, 275 Hofstetter, Walter 241, 243, 246 Hogg, Richard M. 123, 125, 129, 132–133, 139, 142, 159, 173–174, 179, 217 Hoiberg, Dale 96 Holland, Joan 132 Holland, John 280, 290, 304 Hollien, Harry 180 Hollien, Patricia 180 Hooper, Joan B. 138, 159 Hopkins, Eric 191, 199 Hopper, Paul J. 32, 48, 103, 109, 120, 289–293, 304–305 Horgan, Dorothy M. 123–124, 132 Hornstein, Norbert 58, 70 Horobin, Simon 111, 120, 296, 305 Houston, Ann 41, 48 Howard, Sharon 69 Huart, Ruth 119 Hubmayer, Karl 168–169, 173, 179 Huddleston, Rodney 16, 43, 48, 52, 56, 58, 70 Hulshof, H. 47 Hume, David 1–2, 11 Hume, Elizabeth 158 Hundt, Marianne 5, 12 Hutcheson, Bellenden Rand 147, 159 Hutton, James 254 Hutton, William 191–192, 199 Hyams, Nina 318, 320 Ihalainen, Ossi 132, 180 Imo, Wolfgang 99, 120 Inkelas, Sharon 159 Irvine, Martin 204, 216 Irving, Edward B. 210, 216 Iverson, Gregory 132 Jacobsson, Ulf 174, 179 Jakobson, Roman 178, 318, 320 Janda, Richard 121 Jannedy, Stephanie 133 Jansen, Frank 129, 132 Jeffers, Robert J. 179 Jefferson, Judith 159 Jenkens, Steven 182
Jensen, Sonya 216 Jespersen, Otto 43, 48, 99, 120, 139, 159, 318, 320 Jobling, Lee 216 Johansson, Stig 59, 69, 71, 76–77, 81, 95, 102, 119 Johnson, Ellen 305 Johnston, Paul A. 7, 9, 183–200 Jones, Alex I. 210, 216 Jones, Charles 138–139, 158–159 Joseph, Brian 121 Joseph, John E. 318, 320 Kahlas-Tarkka, Leena 68 Kahn, Daniel 138, 143, 149, 159 Kaminashi, Keiko 167, 179 Kant, Immanuel 1–2, 12 Karen, Mark 124, 132 Karg-Gasterstädt, E. 133 Kastovsky, Dieter 48, 132, 160 Kegl, Judy 53, 70 Keller, Rudi 290, 305 Kelly, J. N. D. 221, 227, 246 Kendall, Calvin 272, 275, 319 Kendall, M. G. 74, 96 Kennedy, John F. 91 Ker, N. R. 124, 132 Keränen, Jukka 69, 118 Kerswill, Paul 183, 192, 198–199 Keynes, Simon 210, 217 Kibrik, Andrej A. 120 Killie, Kristin 16–17, 22–23, 27–30, 32, 36–39, 42, 45, 48 Kilpiö, Matti 68 Kim, Myngsook 171, 180 Kimball, John 257, 275 King, Robert D. 161, 180 Kiparsky, Paul 128, 132, 140, 144, 159 Kitson, Peter R. 219, 246 Klein-Andreu, Flora 49 Knightly, Michael R. 210, 217 Kohnen, Thomas 17, 29, 48 Koma, Osamu 41, 48 König, Ekkehard 17, 28, 32–33, 44, 48–49 Korhammer, Michael 246 Korhonen, Minna 97 Kortlandt, Frederik 155, 159
Kortmann, Bernd 15–17, 20–23, 41, 49 Krause, Carina Denise 250, 275 Kreidler, Charles 138, 159 Kretzschmar, William A. 6, 9–10, 12, 277–305 Kristensson, Gillis 187, 199 Kroch, Anthony 294, 305 Krug, Manfred 70, 113, 120 Kruisinga, Etsko 56, 70 Krygier, Marcin 182 Kuhn, Hans 202 Kuryłowicz, Jerzy 172, 177 Kusmenko, Jurij 177, 180 Kuylman, J. 47 Kyes, Robert L. 180 Kytö, Merja 68, 118–119, 160 Labov, William 2, 9, 12, 124, 132, 166, 249–250, 275, 294, 298, 305 Ladefoged, Peter 155, 159 Lahiri, Aditi 167, 177, 180 Laing, Margaret 187, 199 Lamb, Sidney M. 267, 275 Lampert, Martin D. 119 Langacker, Ronald W. 250, 275 Langer, Nils 48 Langton, John 191, 199 Larsen-Freeman, Diane 290–291, 305 Lass, Roger 123, 132, 139, 159, 171, 174, 180, 182, 254–255, 275, 298, 305 Laufer, Batia 84, 96 Laury, Ritva 120 Lavandera, Beatriz R. 257, 275 Lawler, Traugott 207, 217 Leary, Adam P. 124, 133 Leavey, John P. 320 Leech, Geoffrey 5, 12, 56, 59, 69, 70, 71, 76–77, 81, 95, 101–102, 119, 120 Lehmann, Christian 32, 49, 111, 120 Lendinara, Patrizia 241, 246 Lenoir, P. V. 313, 320 Lerud, Theodore K. 305 Levelt, Willem J. M. 256, 275 Lewis, Charlton T. 227, 246 Lewis, C. S. 293, 305, 308, 320 Liberman, Anatoly 7–8, 11, 138, 155, 159, 161–182, 307, 314, 320 Lieberson, Stanley 91–92, 96
Name index
327
Liu, Dilin 71, 73–74, 77–78, 96 Locherbie-Cameron, M. A. L. 209, 217 Lohse, Barbara 55–56, 70 López-Couso, María José 95 Los, Bettelou 46, 49 Love, Nigel 318, 320 Luick, Karl 139, 154–155, 157, 159, 164, 167–168, 173, 180 Lund, Niels 216 Lutz, Angelika 150, 152, 159 Lyell, Charles 254, 276 Lynch, Philip 307, 320 Mair, Christian 5, 12 Mäkinen, Martti 296, 305 Malsch, Derry L. 171, 180 Markus, Manfred 168–169, 179–180 Martey, Nii 119 Martin, Pamela 71–72, 77, 81, 96 Martinet, André 164, 178 Massam, Diane 116, 120 Mathisen, Anne Grethe 184, 197, 199 Matto, Michael 132 McCarthy, Philip M. 77, 95 McCully, Chris B. 160, 167, 179 McDavid, Virginia G. 305 McDougall, David 123, 132 McDougall, Ian 123, 132 McEnery, Tony 250–251, 276 McIntosh, Angus 10, 12, 187, 199, 297, 305 McKinnell, John 209, 217 McLaughlin, Jamie 69 McNamara, Danielle S. 77, 95 Meech, Sanford B. 188, 200 Meid, Wolfgang 179 Mensing, C. M. 47 Merchant, Jason 67, 70 Mettinger, Arthur 48 Meyer, Charles 119 Michel, Jean-Baptiste 96 Miller, Corey 124, 132 Millward, Celia 311, 320 Minkova, Donka 7–8, 11, 123, 132–133, 137–160, 164, 169, 171, 180 Mitchell, Bruce 269 Mitchell, Melanie 278–279, 305 Mlodinow, Leonard 280, 304
328
Name index
Molencki, Rafał 269, 276 Molinelli, Piera 121 Momma, Haruko 132 Monboddo, James Burnet 313, 320 Moore, Colette 9–10, 307–321 Moore, Emma 160 Moore, Robert 186, 200 Moore, Samuel 188, 200 Morgan, Gerald 210, 217 Morris, Richard 230, 246 Morsbach, Lorenz 164, 173 Moser, Hugo 181 Mufwene, Salikoko 298, 305 Mulac, Antony 109, 120 Müller, Max 313–315, 319–320 Murray, J. A. H. 316 Murray, Robert W. 129, 132, 156, 160, 164–165, 167, 181 Nagy, Michael S. 216 Nakagawa, Ken 118 Narrog, Heiko 275 Nerger, K. 171, 181 Nerlich, Brigitte 254, 276, 305 Nesselhauf, Nadja 84, 96 Nevala, Minna 69, 118 Nevalainen, Terttu 11–12, 68–69, 96–97, 118–120, 132, 180, 182, 253, 276, 305 Nevanlinna, Saara 68 Nevins, Andrew 57, 67, 70 Newmeyer, F. J. 132, 159 Ní Chiosán, Máire 138, 160 Nidditch, P. H. 11 Nielsen, Hans F. 131, 132, 180 Niemeier, Susanne 120 Niles, John D. 160, 205–206, 209–210, 216–217, 269, 275 Noordman, Leo 120 Norvig, Peter 96 Nowak, Martin A. 96 Nurmi, Arja 69, 118 Nykiel, Joanna 5–6, 51–70, 118 O’Briain, Helen Conrad 217, 246 Ó Dochartaigh, Cathair 169, 181 O’Donoghue, Heather 217 Ogura, Mieko 173, 181
Ohala, John 166 O’Keeffe, Katherine O’Brien 217 Olds, Kelly B. 92, 96 Ong, Walter J. 252, 276 Orchard, Andy 268, 276 Orton, Harold 174, 181, 183, 194–196, 200 Orton, Peter 154, 160 Orwant, Jon 96 Oumounabidji, Heather 118 Outhwaite, Ben 274 Page, Richard B. 166, 177, 181 Pahta, Päivi 97, 160, 182 Palander-Collin, Minna 69, 118 Palmer, Chris C. 5–6, 71–97 Paolino, Danae 99, 104, 119 Paprotté, Wolf 275 Parkes, Malcom B. 111, 120 Paton, H. J. 1, 12 Paul, Hermann 164, 181 Peacham, Henry 312, 320 Pearsall, Derek 207, 217 Pelli, Mario G. 71, 73, 75, 77, 81, 83, 96 Penke, Martina 5, 12 Pentrel, Meike 6, 9–10, 249–276 Penzl, Herbert 173, 181 Petré, Peter 46, 49 Pfänder, Stefan 120 Phillips, Betty S. 5, 7, 123–133 Phillips, Edward 312, 320 Phillpotts, Bertha 205, 217 Pickett, Joseph P. 96 Pierrehumbert, Janet 124, 133 Pincott, Frederic 315–316, 320 Pinker, Steven 96 Pitts, Ann 124, 133 Plummer, Charles 203, 217 Pollner, Clausdirk 183, 200 Pope, John C. 206, 217 Port, Robert F. 124, 132 Postal, Paul M. 58, 69 Poutsma, Hendrik 99, 120 Prideaux, Gary D. 251, 256–264, 269, 271, 276 Prince, Alan 167, 181 Pullum, Geoffrey K. 16, 43, 48, 52, 56, 58, 70
Puttenham, George 310–312, 320 Putter, Ad 159 Pütz, Martin 120 Quak, Arend 177, 181 Quintilian 307, 309–310, 320 Quirk, Randolph 56, 70, 101, 120 Ramat, Anna Giacalone 47, 121 Rand, Richard 320 Rauch, Irmengard 158, 180 Raumolin-Brunberg, Helena 68–69, 118 Raymond, William D. 125, 131 Reszkiewicz, A. 169, 181 Restle, David 179, 181 Riad, Tomas 170, 181 Rice, Keren 158 Rinkel, Tineke 17, 21–22, 49 Río-Rey, Carmen 22, 49 Rissanen, Matti 68, 119, 132, 180, 182 Ritt, Nikolaus 139, 160, 167, 171, 179–181 Roberts, Jane 151–152, 160 Robinson, Fred C. 216 Robinson, Orrin Warner 132 Rodrigues, Cilene 57, 67, 70 Rodriguez-Puente, Paula 85, 97 Rodman, Robert 318, 320 Romaine, Suzanne 249, 264, 276 Rosenbach, Anette 5, 12, 51, 70, 255, 275 Rubach, Jerzy 138, 160 Russom, Geoffrey 138, 147, 160, 180 Ryan, Kevin M. 141, 160 Ryner, Bradley D. 209, 217 Salmons, Joseph C. 181 Samuels, Michael L. 187, 199, 297, 305 Sarkar, Husain 3, 12 Sarrazin, Gregor 171, 181 Sauer, Hans 219, 246 Saussure, Ferdinand de 318 Scattergood, John 210, 217, 246 Schaff, Philip 227, 246 Scharloth, Joachim 48 Schiffrin, Deborah 48 Schirmunski, Viktor 129, 133 Schlüter, Julia 70 Schmeja, Hans 179
Name index
329
Schmid, Hans-Jörg 99–100, 104, 111, 117, 120 Schmied, Josef 69 Schöbler, Ingeborg 181 Schofield, R. E. 191, 200 Schopf, Thomas J. M. 304 Schreier, Daniel 274 Schudson, Michael 91, 97 Schuetze-Coburn, Stephen 99, 104, 119 Scragg, Donald G. 207, 209, 211, 217 Séguy, Jean 282, 305 Selby-Bigge, L. A. 11 Seoane, Elena 95 Serjeantson, Mary S. 187–188, 200 Seynnaeve, Johan 173, 181 Shapin, Stephen 3–4, 12 Shapiro, Michael 116, 120 Shaw, G. M. 88, 97 Sheldon, Amy L. 252, 258, 260, 270, 276 Shen, Yuan Kui 96 Shibasaki, Reijirou 5, 7, 99–121 Shoemaker, Robert 69 Siegenbeek, Matthijs 37 Siemund, Rainer 69 Sievers, Eduard 163–164, 167, 202, 213–214, 217 Simpson, John 119 Sims, Lynn 118 Sisam, Celia 222, 229, 246 Sisam, Kenneth 222, 229, 246 Skeat, W. W. 168, 232, 235, 246 Slobin, Dan I. 252, 257, 276 Smith, Adam 290 Smith, Jeremy J. 2, 11, 123–124, 131, 133, 174, 181, 296, 301–303, 305 Smith, Norman Kemp 12 Smitterberg, Erik 71, 73, 76, 78–79, 97 Smolensky, Paul 167, 181 Sørensen, Knud 16, 49 Spasov, Dimiter 71–73, 77, 80, 97 Stam, James H. 316, 321 Stanley, Eric G. 144, 157, 160 Steblin-Kamenskij, M. I. 161, 178 Stein, Dieter 120, 275 Steinthal, Heymann 313, 321 Stenroos, Merja 296–297, 305 Stephan, Achim 275 Stephens, W. B. 191–192, 200
330
Name index
Sternefeld, Wolfgang 69 Stockwell, Robert 123, 132, 140, 160, 169, 174, 180, 182 Stubbs, Michael 84, 97 Stubkjær, Flemming T. 180 Sutcliffe, Geoff 95 Suzuki, Seiichi 139–142, 147, 158, 160 Svartvik, Jan 56, 70, 101, 120 Swan, Toril 16–17, 22–23, 27–30, 32, 42, 45, 48 Sweet, Henry 124–125, 133 Szczegielniak, Adam 57, 70 Szmrecsanyi, Benedikt 51, 70 Taavitasainen, Irma 68, 97, 132, 180, 182, 308, 321 Taylor, Talbot J. 318, 320 Temmerman, Rita 219, 225, 246 Tenner, Edward 91, 97 Thengs, Kjetil 297, 305 Thim, Stefan 76–80, 97 Thompson, Sandra A. 15, 48–49, 103, 109, 119–120 Thormann, Janet 204, 217 Thorpe, Benjamin 220, 228–230, 232–234, 236, 238–239, 246 Tilling, Phillip M. 183, 195–196, 200 Tillott, P. M. 191, 200 Tissari, Heli 250, 276 Tolkien, J. R. R. 205, 217 Tomasello, Michael 274 Tomlin, Russell 119 Toon, Thomas 131, 133 Tops, Guy A. J. 182 Tork, Danièle 70 Toswell, M. J. 160 Traugott, Elizabeth Closs 11–12, 28, 32, 48–49, 52, 56–58, 69–70, 84, 95–96, 106, 109, 114–115, 117–121, 211, 217, 250, 276, 291, 305 Trautmann, Reinhold 164–165 Treiman, Rebecca 156, 158, 160 Trnka, Bohumil 175, 178, 182 Trost, Pavel 177 Trousdale, Graeme 106, 115, 121, 267, 275 Trubetzkoy, Nikolai 162–164, 167 Trudgill, Peter 183, 198, 200
Tseng, Jesse 56, 70 Tuggy, David 111, 116, 121 Van Belle, William 33, 49 Van Bogaert, Julie 116, 121 van Bree, Cor 17, 37, 49 van Coetsem, Frans 132 Vandelanotte, Lieven 121 Vandenbussche, Wim 48 van der Auwera, Johan 17, 32–33, 44, 48–49 Van der Hoek, Michel 169, 182 van der Horst, Johannes 17, 21, 29, 39–40, 44, 49 van der Wal, Marijke 17, 37, 49 van Fraassen, Bas 3, 12 van Hoek, Karen 120 van Oostendorp, Marc 158 van Uildriks, Gonne 47 Vennemann, Theo 164–165, 171, 182 Veres, Adrian 96 Vicente, Luis 57, 67, 70 Vollenhove, Joannes 37 von Strasburg, Gottfried 173 von Stutterheim, Christiane 46–47 Wagner, Ester-Miriam 274 Waldman, Tina 84, 96 Walter, Sven 275 Wanner, Anja 120 Warner, Natasha 133 Wasow, Thomas 55–56, 67, 70 Watts, Richard 274 Waugh, Linda 318, 320 Waugh, Robin 216 Wedgwood, Hensleigh 313, 321 Weinberg, Amy 58, 70 Weiner, E. S. C. 119 Welby, Pauline 138, 160 Weldon, James 216 Wells, John C. 184–185, 194, 200 Wełna, Jerzy 169, 182 Wenisch, Franz 237, 246 Wetzel, Claus-Dieter 150, 152, 160 Wetzels, W. Leo 70 Whewell, William 254 Whitehall, Harold 188, 200 Whitney, William Dwight 254, 276, 313, 316, 321 Wiemer, Björn 119
Wilcox, Miranda 245 Wilkin, John Price 69, 220, 246 Williams, Ann 183, 192, 198, 199, 211, 217 Wilson, David 95 Wilson, Thomas 309, 311, 321 Winters, Margaret E. 250, 256, 276 Wischer, Ilse 48 Wittig, Kurt 173, 182 Wolk, Christoph 51, 70 Wood, Ian N. 216 Woolf, Rosemary 209, 217 Wray, Alison 267, 276
Name index
Wright, Laura 296, 305 Wright, Susan 120 Xiang, Xin 69, 123, 132, 220, 246 Yonekura, Hiroshi 118 Zaefferer, Dietmar 179, 181 Zipf, George 282, 305 Zonneveld, Wim 132 Zuraw, Kie 128, 133, 138, 158, 160 Zwicky, Arnold 4–5, 12
331
Subject index 80/20 rule 282–283, 292, 296, 301 A-curve 10, 281–282, 293–296, 298, 301, 304 Adams, Arthur Henry 147 Ælfric 9, 148, 151, 220–223, 225–244, 309, 319 Ælfric Bata 151 Æthelred 208, 210 Æthelstan 203, 205, 212, 215 Alford, Henry 147 Alfred 123–124, 131, 204 algorithmic reasoning 3 ambisyllabicity 8, 137–141, 143, 148–149, 154–156, 172 American English 6, 71–97 anti-realism 3 Arte of English poesie 310–311 Arte of rhetorique 309–311 Athelwold 222 Atwell, Charles 66 Austen, Jane 34, 35, 47 bad data problem 3, 249 Bailey, Jacob 147 Bailey, Philip James 147 Balto-Slavic 168 Bare Argument Ellipsis 51, 60–61, 68 Barton, Bernard 147 Battle of Brunanburh 8, 201–215 Battle of Maldon 8, 201–215 Bede 237, 309, 319 Behn, Aphra 62 Bell, Alexander 86–87 Beowulf 209 Beowulf 10, 201–202, 205–209, 211–215, 252, 268–273 Berry, Henry 62 Birmingham 9, 184–187, 189–198 Black Country 9, 184, 186–187, 189–191, 193–198 Bonington, Chris 106 Brontë, Emily 146 Brunanburh poet 202–203, 205–208, 212–213, 215 Byrhtferth of Ramsey 237, 242
Canterbury Tales 145 Chaucer, Geoffrey 145, 173, 302, 303 clause 15–46 – adverbial 17, 30, 36, 38, 109 – adverbial-like 22 – converb 15–16, 22 – coordinate 32, 33–36 – finite 17, 20, 34–37, 102, 105, 108 – non-elliptical 52–53, 57, 60, 67 – participial 15–17, 29, 37, 44 – relative 21, 29, 38, 44, 101–102, 107, 251–253, 257–272 – semi-coordinate 17, 30, 45 – subordinate 6, 15–16, 18, 30, 32–33, 37, 45–46, 109 – wh- 100, 103–104 coalescence 111, 114, 116, 142 Cockney 185, 195 cognitive linguistics 9, 249–253, 255–257, 260, 267, 269, 271, 274 colloquialization 79, 93 complementizer 100–102, 105–107, 110 complex systems 10, 277–304 compositionality 56, 58–59 consonant 8, 129, 137–146, 150–157, 161–165, 168–172, 174–176, 178, 296–297, 311 – fricative 139–141, 166, 178 – glottalization 155 – lenition 148, 178 – liquid 168 – palatal 174, 185 – palatalization 168–170 – resonant 169–170, 172, 174 – stop 104, 138, 155, 172, 178 – velar 141, 163, 168, 169, 174 – velarization 170 – voicing 139–141, 155, 166, 170, 178 Constantine (king of the Picts and Scots) 206 contact 177, 183–185, 192, 195, 198 conversation 77, 79, 100, 102–106, 113, 280 Cooke, John Esten 62
copula 106, 111, 116 corpora 5–7, 52, 111 – 38-miljoenwoordencorpus 19 – A corpus of English dialogues 60–61 – A corpus of Late Modern English texts extended version 107–108 – A corpus of nineteenth-century English 73 – ARCHER: A representative corpus of historical English registers 60–63 – The corpus of contemporary American English 71–73, 77, 85–87, 89, 94, 99, 103–104, 110–116, 259–260, 286–288, 317 – The corpus of historical American English 71–76, 78–90, 92–93, 99, 101, 108–110, 112–113, 115, 287–288 – collocation in 287–289 – Compilation corpus of historical Dutch 15, 19, 21–22, 27, 29 – Complete works of William Shakespeare 60–61 – Corpus of Early English correspondence sampler 62 – Corpus of Middle English prose and verse 59 – data from 17, 52, 57, 62–65, 72–92 – Dictionary of Old English web corpus 59 – Digitale bibliotheek voor de Nederlandse letteren 19 – Early English books online 106–107, 308 – frequency in 10, 22, 27, 30, 38–42, 59, 61, 63–64, 66, 68, 72, 76–77, 83–84, 86–88, 90, 92–94, 100, 102, 105–106, 112–113, 117, 123, 125–127, 129, 252, 257, 264–265, 271–272, 282, 286–288, 292–294, 296, 301, 304 – The Helsinki corpus of English texts 59–61 – The Lampeter corpus of Early Modern English 59–73 – The Lancaster-Oslo/Bergen corpus 73 – Literature online 60–63, 308 – multi-register 6, 72–73, 76, 93 – Old Bailey proceedings online 60, 62–63 – Penn-Helsinki parsed corpus of Early Modern English 18, 20, 22, 29 – Penn-Helsinki parsed corpus of Middle English 18, 26–28
Subject index
333
– Penn parsed corpora of historical English 18 – Penn parsed corpus of Modern British English 15, 18, 22–23, 28 – Santa Barbara corpus of spoken American English 99, 102–105, 110, 113, 117 creed 9, 219–243 Czech 64 Danish 17, 36, 162–163 data 1–11, 19 decategorialization 108, 111, 114, 116 De la Mare, Walter 146 determiner 19, 102, 105, 108, 112–14, 117 Dickens, Charles 34, 25, 36, 47, 195 dictation 151, 252, 262–266, 273 dictionary 3, 111, 146, 241, 312 – Century dictionary 315 – Dictionary of Old English 154, 231, 308 – Dictionary of Old English short titles 220 – “Dictionary of onomatops” 315–316 – Dictionary of Sir Thomas Elyot knight 309 – hard-word 312 – Middle English dictionary 144, 308 – Oxford advanced learner’s dictionary 101 – Oxford English dictionary 87, 106, 108, 287–288, 307, 309–311, 316 discourse 63, 88, 90, 100, 102–106, 110, 117, 253, 255, 257–258, 317 – function 6, 18, 41, 43, 46–47, 103, 111, 252 – marker 7 Durham Ritual 154 Dutch 6, 15, 17–18, 24–46, 129, 170–176 Early Modern English 6, 17–18, 22, 30, 34, 41–42, 59–62, 64–66, 68, 106–107, 307–309, 311–312, 317 Edison, Thomas 87–88, 249 Edgeworth, Maria 107 Edmund (brother of Æthelstan) 203, 205, 212 ellipsis 6, 61 – alternation 51–68 emergent grammar 289–292, 304 enregisterment 4 evidence 1–11 – grammar-external 4, 6–7, 9 – shadow 8
334
Subject index
extra-grammatical factors in change 18, 38, 45, 51–52, 72, 85–95, 183–185, 198, 201–215, 219–243 fashion, as factor in change 72, 85–86, 91–95 feedback loop 10, 251, 253, 268, 272–273 Finnish 282 First Consonant Shift 178 fit-technique 10 fixedness 106, 111, 116 free adjunct 6, 15–46 – evidence from novels 34–36 – in initial position 30–32 – semantic categorization of 23–27, 33, 34 – semantic diversification of 17, 27, 34, 45–46 – syntactic categorization of 17, 24–25, 34, 45 French 64, 282 function words 7, 114, 125–131, 146, 203, 207–208, 211, 213 Garden of eloquence 312 geminate 138, 153–155, 164, 172, 176, 201 gemination 139–140, 152–157, 169 generative linguistics 4 genre 9, 19, 61, 63, 73, 77–80, 82, 93, 261, 266, 272 gerund 17–21, 38–43, 45–46 German 171–176, 282 Germanic 138, 161–163, 167–169, 171–172, 176–178 – Middle 176 – Old 162–164, 205 – Proto- 16, 139, 149, 153, 155, 165 – West 123, 146, 153, 155, 162, 165, 168–169, 171–172, 177 Germanic languages 16–18, 22, 27, 32, 45–46, 163, 165, 175–176 Gifford, Humphrey Gissing, George 108 glossing 129, 219–243 Google books 308 Gothic 141, 153, 165, 175 grammaticalization 7, 39, 45, 84, 100, 108–111, 114–117, 250, 289–291
Grassmann’s law 161 Greek 37, 151, 157, 161–163, 165, 307, 309, 312 Green, Robert 62 Gregory (Pope) 123, 131 Grendel 209 Grimm’s Law 161 Hatton 20 123–125, 128–131 Hazlitt, William 107 Hengest 204 Hesselman’s Law 176–177 heuristic reasoning 3 Hill, Lawrence 62 historical linguistics 1–11, 18, 161, 249–251, 253, 278, 298, 303–304, 308 historical psycholinguistics 250–251 Horsa 204 hypotaxis 36–37, 109 idiolexical history 10 idiomatization 84, 86 inkhorn term 311–312 Japanese 282 Jefferies, Richard 66 Kendall’s-τ 74, 87, 90, 92 koineization 184, 187, 191–192, 194–196 Kuryłowicz’s rule 172, 177 Lambeth glossator 226–227, 229–232, 234–235, 237–243 language evolution 278, 298–303 language ideology 9–10, 307–318 Late Modern English 30, 42, 59–66, 68, 78, 99, 107, 111 Latin 9, 16–17, 36–37, 45, 148, 151–152, 165, 219–228, 231, 234, 236, 238–239, 241, 309, 318 lemmatizaion 83–87, 94 lexicalization 56, 58, 65, 84, 116, 202 lexis 9–10, 52–54, 58, 67, 102, 106, 124, 141, 146, 156, 184–185, 195, 198, 207–208, 213, 241–242, 256, 308, 310, 317 Lightfoot, John 107
Linguistic atlas of the Middle and South Atlantic States 282–286 Logographic-emblematic English spelling book 313 Londonization 183–184 Maldon poet 202–203, 210–215 Mercian 124, 129, 131, 168, 170, 187, 222 merger 9, 105, 124, 128, 130, 175, 185, 193–198 metaphorization 86 meter 8, 61, 139–149, 153–154, 157, 167, 170, 201–202, 206–207, 221 methodology 1–11, 18, 72–74, 111, 118, 123, 125, 152, 157, 250–251, 263, 298, 304, 308 metrical resolution 139–144, 147–149, 153, 157, 201 Middle English 10, 16–18, 25–28, 38, 42, 53, 57, 59–61, 66–67, 72–73, 76, 80, 93, 106, 139, 144, 149, 154, 161, 165, 167, 170–172, 174, 176, 178, 183, 185, 187, 190, 198, 211, 251–252, 262–264, 266, 269, 271–272, 296–298, 303, 309 Middle Franconian 129 Middle High German 171, 177 migration, as factor in change 183–185, 198 mora 141, 143, 149, 162–163, 167, 170, 172–174, 177–178, 184, 195, 198 Mull, Thomas 106 Mutual Information 83–84 nominalization 230–234 Northumbrian 154–156, 183, 185 Old English 3, 7–10, 16, 46, 53, 57–61, 68, 123–131, 137–157, 162–163, 165, 168–169, 174–175, 186–187, 201–215, 219–245, 251–252, 255, 262, 264, 268–273, 293 Old Frisian 156 Old High German 129, 161, 163 – Consonant Shift 161 Old Saxon 156 onomatopoeia 307–318 – alternative terms for 307, 315–316, 318
Subject index
335
onset maximality 138–139, 141, 143, 148–149, 151–152, 157 Orosius 127–128 The Owl and the Nightingale 144–145 nationalism 8, 204–205, 208 Norwegian 17, 36, 39, 45 optimality theory 144, 167 Orm 156, 164–165 orthography see spelling parasiting 145–146, 150 parataxis 36–37, 106, 108–109, 117 participle 6, 18, 20, 22, 38–39, 46 particle 55, 71–76, 80–84, 93–94, 202, 269 Paston, John 262–263, 267–268 Paston, Margaret 263, 267–268 Paston letters 10, 72, 262–264, 266–269, 271–273 phonology 7–9 phrase 51–68 – adverbial 51 – noun 51–53, 56–57, 59–68, 100 – prepositional 51–68 – remnants 51–53, 57, 59–68 – stranded 51 – verb 55 – wh- 51, 60 phrasal verb 6–7 Poema morale 149 Polish 64, 282 politics 8, 201–202, 208, 210 polysemy 225, 228–229, 232, 239, 241–243 pragmatic marker 100–101, 104, 106–114, 116–117 pragmatics 7–8, 28, 114, 116, 148, 201, 203–205, 208–211, 307 Pratt, Samuel Jackson 147 preposition stranding 6, 58, 67, 80 prescriptivism 36–39, 45, 79–80, 152 Present-Day English 3, 8, 10, 17, 22, 43, 52–53, 57–59, 63–64, 66–67, 71–72, 76–80, 93, 99, 111, 138–139, 143–144, 146, 148–149, 155–157, 251–253, 258, 262, 264, 269, 272–273
336
Subject index
processing strategies 10, 52, 54–56, 116, 249–274 – adjacency 55, 57–58, 252, 258, 260–262, 264, 269, 271 – bracketing 258 – closure 252, 256, 258–261, 263–266, 269–273 – given-new 258–259 – non-ambiguity 258–259 – normal form 252, 258–261, 264, 266, 269–271 – parallel function 258, 260–262, 264, 269–273 progressive 39, 40, 41–42, 45 Project Gutenburg 19 prosody 8, 104, 125, 139–140, 162, 165–166, 171, 174, 177–178, 211–215, 294 pseudo-cleft 101, 103, 106, 109, 117 psycholinguistics 6, 9, 51, 249–253, 255–257, 259, 261, 268, 271–272 punctuated equilibrium 198, 298–304 punctuation 110–111 reanalysis 6, 16–18, 27–28, 30, 41, 45–46, 58, 145 register 6–7, 71–73, 75–83, 93, 157, 268, 281, 312, 316–318 relativizer 252, 259, 261–263, 269–270 rhetoric 8, 37, 202–203, 207, 215, 307, 309–313, 317 Royal glossator 231, 234–235, 242 Rushworth Gospels 154 Sanskrit 161 scale 282, 288–289, 292, 296–297, 300–301, 303–304 Schriftbild 129–131 scientific method 3 Scots 185 Scots Gaelic 185 scribal practice 7, 124–125, 129–131, 139, 148–152, 156–157, 201, 221, 252, 264–265, 297 S-curve 10, 294–296
semantics 6, 17–18, 23–30, 34, 43, 45–47, 70, 73, 84, 86, 89, 91, 94, 220–221, 225, 228, 242, 250, 256, 260, 289, 294 semantic dependency 5–6, 51–61, 63–64, 66–68 sense-impressions 1 Serbo-Croatian 64 shell constructions 7, 99–104, 116–118 Shelley, Mary 80, 96 Sidney, Philip 312 Sievers types 201–215 Sinclair, Upton 87 skepticism 1–2 sluicing 51, 60–61, 63, 68 sociolinguistics 9 Spanish 64 spelling 7, 123–131, 137, 142, 149–151, 153–157, 164, 174–175, 190, 206, 240, 296–298 stance 99, 101, 111, 114 standardization 9, 219–243 Stonor, William 106 Stowe, Harriet Beecher 34, 35, 47 style 8, 36–37, 201–215, 249, 252, 266–267, 271–272, 310 subjectification 111, 115, 117 subordinate clauses 6 syllabification 138–140, 143, 147–153, 156–158, 166, 170–172, 177 syllable 137–138, 149, 150–152, 154, 201–202, 207–208, 212–213 – closed 141, 165 – cut (Silbenschnitt) 162–163, 165, 176 – di- 144, 165, 176–178 – division 137, 152, 165, 176 – heavy 140–143, 154, 202 – light 140, 143, 149, 202 – mono- 149, 156, 162, 165, 175–178 – open 139, 161, 164–166, 170–171, 173–174 – stress 125, 129, 140–143, 146–147, 149, 154–156, 172, 201–202, 208, 211–214 – weight 139–143, 157
syntax 3, 6–9, 15–46, 51–70, 100–118, 201–203, 205, 207–208, 211, 214–215, 252, 256, 260, 262–263, 269, 272, 290–291 Tacitus 209 telecommunications, as factor in change 72, 85–88, 91, 93–95 terminology 9, 219–243, 307–318 text-type 71, 77, 79, 82–83, 253, 255, 266, 281, 289 Thai 282 translation 9, 17, 19, 34–36, 219–244, 252 transportation, as factor in change 72, 85–86, 88–91, 93–95, 183–184 Twain, Mark 34, 35, 36, 57, 195 uniformitarian principle 9, 252–256, 271–272, 274 variation 127, 166 – allophonic 143 – lexical 223–224, 226, 241 – Old English poetic 202, 205–208, 211, 214 – orthographic 7, 123–131 – regional 297 – register 5, 76–78, 80 – syntactic 6, 51, 116–117 verb 52–56 – collocate 7, 71–72, 83, 85–86, 89–90, 94 – entailment 54–55, 66 – multi-word 52–53, 59, 64–66, 68 – phrasal 71–97 – prepositional 52–53, 55–60, 64–67 verbal gerunds 17–18, 20 verbalization 41–42 Verner’s Law 178 Vespasian Psalter 129, 222, 244 Vietnamese 282 von Strasburg, Gottfried 173 vowel – apocope 144, 162, 165, 168–169, 172–173, 175–178 – assimilation 130–131, 153, 185 – breaking 139, 163, 169, 172–173 – classes 9, 185, 193–194, 197–198
Subject index
337
– contraction 142, 144 – diphthong 123, 125, 128, 142, 173–175, 185, 194–198 – diphthongization 163–164, 174, 185 – fronting 129, 184, 197 – lengthening 8, 139, 141–142, 155, 161–162, 164–166, 168–178 – lowering 124, 129 – monophthongization 163–164, 194 – quantity 162, 164–165, 170, 174, 206 – raising 129–131, 185, 302 – shift 8, 124, 128–129 – Great Vowel Shift 161, 175, 301–304 – Second Vowel Shift 194 – short 153–156, 168, 173–174, 177, 184 – shortening 154, 163, 165–166, 170 – smoothing 163 – strengthening 128–131 – syncope 144, 153, 168–169, 202 – system restructuring 183–198 Walpole, Horace 62 Warren, Mercy Otis 66 Weller, Sam 195 Welsh 185 West Saxon 7, 123–124, 127, 131 writing 20, 79, 80, 84, 151–152, 229–243, 252, 256, 261, 264, 266, 280–281 – academic 77, 79 – drama 60–61, 73, 75–76, 78–79, 81, 83–84 – edited 79–80 – fiction 19, 34–36, 71–73, 77–80, 87, 93, 261, 264, 270 – letter 10, 19, 72, 78–79, 81, 106, 192, 252, 262–269, 271–273 – magazine 73, 78–80, 93 – newspaper 73, 77–80, 82, 87, 93 – non-fiction 73, 78–80, 261 – poetry 8, 60–61, 76, 138–150, 152–153, 157, 201–215, 252, 268–271, 317 – scientific 79 – vernacular 17 Wulfstan 220–221, 233 Zipf’s Law 282