LINGUISTIC VARIATION AND CHANGE

MILROY, J.(1992)

 

  1. INTRODUCTION: LANGUAGE CHANGE AND VARIATION

 

         Language Change

 

One of the most important facts about human languages is that it is continuously changing. Everyone knows that languages have changed in the course of history: it is easy to see from a distance in time that there are differences between Shakespeare’s English and present-day English, but it can also be shown from close at hand that language is continuing to change in the present just as it did in the past. At this very moment changes are being implemented and diffused: old varieties are dying out and new varieties are springing up; pronunciations are changing, new words and constructions are being adopted and old ones adapted to new uses. Sometimes change is rapid, and sometimes it is slow, and at any given time some linguistic structures are changing while others remain stable. Indeed, change seems to be inherent in the nature of language: there is no such thing as a perfectly stable human language.

It is also true that at any given time a language is variable. Languages are never uniform entities: they can be observed to vary geographically and socially, and according to the situational contexts in which they are used. In the study of linguistic change, this heterogeneity of language is of crucial importance, as change in progress can be detected in the study of variation. It is also important to remember this when we look at past states of language: we have to accept that, just as language is variable when observed at the present day, so it must also have been variable in the past. The history of any language is therefore not the history of one “variety”, but is a multidimensional history.

It follows, therefore, that at any time we care to look at a language –or a dialect –it is variable and in a state of change. We may, of course, choose to ignore this and treat language for descriptive purposes as if it were a uniform and unchanging phenomenon, and there are often good practical reasons for adopting this convenient idealization. For instance, we may want to write a grammar of English for the use of foreign learners, and it will be more helpful to our readers if we focus what is constant rather than what is changing. However, the idea that language is static or uniform has also penetrated into the roots of theory, and much of the linguistic theory of this century has been based on the perceived need to treat language as if it were static and uniform. It should, however, be borne in mind that technological advances since the earlier part of this century have made it much more feasible for us to focus on variable states of language than it was for de Saussure, Sapir and Bloomfield.

 

             Uniformity and Variation

 

Early in the century de Saussure emphasized the priority of synchronic descriptions and from that time onwards the dominant trends in language description were synchronic: they focused on states of language at any given time as finite entities. Historical (or diachronic) linguistics was relegated to a subsidiary role and was often conceived of as an exercise in comparing these finite states of language at different times. From this point of view we can liken a synchronic description to a still photograph and a historical description to a comparison of a series of a still photographs taken at different times. In reality, however, the history of language is a continuous process: it is not a series of stills, but a moving picture. If we are to come closer to understanding why and how languages change, we need to bear this in mind.

It has often been pointed out (for example, by Bynon, 1977:104-7) that Saussurean structuralism can lead to a paradox for historical linguistics. According to the Saussurean view, a language is at any given time a system où tout se tient –in which everything holds together in a coherent self-contained structure of interdependent parts. Historical linguistic inquiry then proceeds by comparing different states of language that have been attested from different periods. However, a difficulty arises in the (unattested) intervening stage between one state (state A) and another (state B), as it appears that in the transitional stage that cohesiveness of the state A structure has been to some extent violated. From this point of view, therefore, we may be inclined to think of language as being perfectly structured at some times but flawed at other times. Now, if linguistic change were an abnormal state of affairs, this would not be an unreasonable way to look at language: change could then be seen as something that strikes a language from time to time like a disease. We could talk of healthy languages (where everything holds together) and sick languages (where it does not). But this is not how things are: no real language state is a perfectly balanced and stable structure, linguistic change is always in progress, and all dialects are transitional dialects. Synchronic states, as we observe them at any given time, are therefore changing states, and stable states of language of the kind postulated in Saussurean theory are idealizations

The same, of course, applies to uniform states of language. As Sapir observed long ago (1921:147): “everyone knows that language is variable”. No language is ever uniform. Linguistic theorizing, however, has often proceeded as if languages were uniform entities. Indeed, according to Weinreich, Labov and Herzog (1968), it has gone further than this and has assumed that structuredness is found only in uniformity: thus, variability in language has often been discounted as unstructured. In fact, the equation of uniformity with structuredness or regularity is most evident in popular (non-professional) attitudes to language: one variety –usually a standard language –is considered to be correct and regular, and others –usually “non-standard” dialects –are thought to be incorrect, irregular, ungrammatical and deviant. Furthermore, linguistic changes in progress are commonly perceived as “errors”. Thus, although everyone knows that language is variable, many people believe that invariance is nonetheless to be desired, and professional scholars of the language have not been immune to the consequences of these same beliefs. We have discussed these points elsewhere (J. Milroy & L. Milroy, 1985a), and I shall return to them in later chapters. Here my concern is merely to point out that uniform states of language are idealizations and that variable states are normal; furthermore, variation in language may itself be structured and regular. Languages are not in reality completely stable or uniform, and there is absolutely no reason why they should be.

The discussion above has raised some basic questions about language. Is a language actually a system où tout se tient, in which all the structural parts are interdependent? Is variation in a language also totally structured and systematic? Finally, if languages are actually like this, how can change occur at all? In this book, I shall take the view that these structuralist assumptions are not self-evidently true and that to establish the degree to which language is structured within itself –without reference to outside (for example, social) factors – is an empirical task. Indeed, if language states are not to some extent open-ended –if there are no “leaks” in the system –it is difficult to see how linguistic change can take place.

 

             Foundations for the Social Modelling of Language Change

 

One way of investigating these matters is to focus on the social nature of language change, and this is the purpose of this book. My aim is to examine the extent to which the origins of linguistic change can be shown to be social; to put it in a slightly stronger form, I want to examine the thesis that linguistic change is a product of speaker-activity in social contexts, which cannot be wholly explained from within the properties of language systems themselves. This approach is justified on the grounds that language is a social phenomenon: it is used by speakers to communicate with one another in social and cultural contexts in which the language system (narrowly defined as a “grammar”) is not the sole means of communication and personal interaction. Furthermore, it is commonly observed that languages which have no speakers do not change, therefore, it seems reasonable to inquire into the role of speakers in language change. As socially-based arguments of this kind have not been widely favoured by historical linguists over the last century or so, I shall attempt in Chapter 2 to relate this social theme to the context of intra-linguistic historical argument. First, however, I want to explain more directly why I have felt justified in developing this socially-based approach to historical linguistics. The key point to bear in mind is that language throughout history has been primarily a spoken, and not a written phenomenon –my thesis is largely a matter of following out the implications of that fact.

“The drama of linguistic change”, according to Wyld (1927:21), “is enacted not in manuscripts nor inscriptions, but in the mouths and minds of men”, and historical linguists have generally insisted that the history of language is primarily the history of spoken language. Traditionally, however, it was not possible to follow this out very thoroughly because investigators did not have the technology to study spoken discourse in extensor, and could hardly have imagined how complex the patterning of spoken interaction in situational contexts would actually turn out to be when it did become possible to analyze it. As a result of these limitations, much of the generally accepted body of knowledge on which theories of change are based depends on quite narrows interpretations of written data and decontextualized citation forms (whether written or spoken), rather than on observation of spoken language in context (“situated speech”). However, it is in spoken rather than in written, language that we are able to detect structural and phonetic changes in their early stages; for tis reason and others, our understanding of the nature of linguistic change can be greatly enhanced by observing in a systematic way recurrent patterns of spoken language as it is used in day-to-day contexts by live speakers. I would therefore like to suggest here three general principles, or foundations, for the social modelling of change that arise directly from this emphasis on spoken interaction, and that I have used as guidelines in research into language change.

Speech is a social activity in a sense that writing is not, and the primary locus of speech is conversation. Conversations take place between two or more participants in social and situational contexts, and linguistic change is one type of phenomenon that is passed from person to person in these situations. The first principle for a socially-based model of language change therefore concerns the observation of language in use: it is the principle that speech-exchanges can be observed only within social and situational contexts –they can never be devoid of such a context. To express this more fully:

 

 Principle 1  As language use (outside of literary modes and laboratory experiments) cannot take place except in social and situational contexts and, when observed, is always observed in these contexts, our analysis –if it is to be adequate –must take account of society, situation and the speaker/listener.

 

This first principle carries with a number of implications, the most important of which is that generalizations about language structure depend on a process of abstracting “language” from the situational contexts in which it naturally occurs. We do not actually observe “the language” or “language” in the abstract: we observe people talking. In a social account of language change, therefore, we have to explain how changes get into the abstract structure that we call language (which we cannot observe directly) as a result of the activities of people talking (which we can observe more directly). Furthermore, unstructured observations of very selective phenomena will not be enough here: our descriptions of sociolinguistic patterns will depend on observing recurrent patterns and will have to be systematic and accountable to the data. In chapter 2 I will suggest a distinction in principle between system and speaker, which arises form this discussion; it is a distinction that I think we need to bear in mind when we are analyzing language in use. It also follows form this first principle that close attention to methods of data collection and analysis (and the relation of one to another) is crucial; we regarded this as very important in our work in Belfast, which I describe more fully in chapters 3 and 4.

Whereas Principle 1 concerns the impossibility of observing language independently of society, Principle 2 concerns the impossibility of describing language structures independently of society. This is not as controversial as it may seem.

 

  Principle 2  A full description of the structure of a variety (whether it is “standard” English, or a dialect, or a style or register) can only be successfully made if quite substantial decisions, or judgements of a social kind are taken into account in the description.

 

The word “social” here does not mean social class or prestige –the decisions (or judgements) we are talking about are decisions (or judgements) about the “norms” of the variety concerned, and these norms are social in the sense that they are agreed on socially –they depend on consensus among speakers within the community or communities concerned and will differ form one community to another. The accuracy of the linguist’s description must therefore be judged on how closely it coincides with the socially agreed norm for the relevant community.

Most language description encounters this problem of “norms”, and although it is not always acknowledge, it can be detected in many descriptive accounts of English. Even a statement that Received Pronunciation (RP) of English has a long diphthong with an open first element in such words as tie and tight depends on observing a sample of people who are considered to be speaking this variety and on the linguist’s judgement that this vowel is the majority usage among these persons. But as a more general example of judging the norm, let us consider Palmer’s (1965: 72-7) characterization of the English perfect tense/aspect. Palmer cites sentences in which the adverbs just and already occur with the perfect, but he gives no examples of their occurrence with the simple past tense. Thus, forms like “I just did it” and “I did it already” are not given as possible sequences. A normative judgement is implicit here, and this is a probably a correct judgement for many varieties, chiefly southern English ones. However, it is certainly not correct for all varieties: the [past tense + just/already] collocation is frequently observed in American, Irish and Scottish English. Therefore, the accuracy of Palmer’s characterization has to be assessed, not in terms of some absolute standard of “grammaticality” of the construction, but in terms of the speech community to which is relevant. It is not a matter of accurately describing what is agreed on by speakers in the community concerned as the consensus norm of that community.

The interpenetration of social and linguistic judgements is easily demonstrated in the work of linguists who are ostensibly non-social in their approach. Smith (1989: 111-12), for example, comments that “for most speakers of (British) English” he ate the pie already is “barely acceptable”, whereas he has eaten the pie already is “fine”. This involves the same kind of normative judgement that I have discussed above, and it is more or less correct for English in England and Wales. But much more dubiously, Smith further comments that “for all speakers” (my italics) he has eaten the pie yesterday is “ungrammatical”. However, this construction does occur in SBE (Southern British English). As Trudgill (1984a: 42) points out, “the rules governing the use of the present perfect in Standard English seem to be altering somewhat, and there appears in particular to be an increase in the usage of such form as: I’ve seen him last year; He’s done it two days ago”. Nothing that [perfect tense + yesterday, last year, etc.] is spreading in SBE and is “grammatical” in other languages, such as French, it is advisable to side with sociolinguists and not with Chomskyan linguists. We are not dealing with ungrammaticality, but with a change in the norms of usage for some part of the community. What this demonstrates is how easy it is for the non-social linguist to appear to propose prescriptive judgements. These typically appeal to some idealized superordinate norm which is part of the “standard” or literate language, rather than a consensus community norm, but, although they are not enunciated as social, there are also social judgements.

But what is true of “standard” English norms (as described by Palmer, Trudgill and Smith) is also true of non-standard norms, no matter how violently deviant they appear to be to the prescriptively-inclined observer. For example, if everybody in a social group says we was there, then we was is the consensus norm. To take one of our own examples, it is clear that for many Belfast speakers (and indeed for many speakers of Irish English generally), the pronoun yous (plural) is categorical, contrasting with you (singular):”So I said to our Trish and our Sandra: “You wash the dishes”. Sure, I might as well have said “You wash the dishes”, for our Trish just go up, put her coat on and went out”. The categorical distinction here between you and yous can be said to be a norm for some community of speakers. The difficulty that arises for the descriptive linguist is not so much to determine the extent of what is “grammatical” (on which see especially Labov, 1973), as to determine the extent of the community of speakers within which this particular structure is the consensus norm. It is clear that many people in Belfast have categorical yous, but that many others vary in the plural between you and yous; indeed, there are some who have categorical you (as in Standard English). Therefore, a description which states that yous (plural) is categorical in Belfast English will be valid for some part of the community, but not for all speakers or all styles, and the variability in you/yous usage will certainly exhibit a socially structured pattern.

Thus, although linguists have generally described differences between varieties of language as linguistic facts, these differences are also social facts. The preference for “I did it already” in dialect A as against “I’ve done it already” in dialect B, for example, arises from differences in speaker-agreement within communities and is to that extent a social fact. It follows form this that all language descriptions, no matter how objective they are, must be normative. But although linguists have often equated normative with prescriptive, no such equation is intended here. Language descriptions are normative because to be accurate they have to coincide as closely as possible with the consensus norms of the community concerned. To be normative, the linguist’s account of a variety does not have to be prescriptive; that is, it does not have to prescribe how people in a community should speak. The distinction I am making here can be described as the distinction between observing a norm for descriptive purposes and enforcing a norm prescriptively; but as this is not a familiar distinction, I should perhaps discuss it a little more fully.

Linguistic scholars commonly contrast “descriptive” with “prescriptive”. Daniel Jones, for example, has this to say in the preface to his English Pronouncing Dictionary (1955):”no attempt is made to decide how people ought to pronounce; all that dictionary aims at doing is to give a faithful record of the manner in which certain people do pronounce.” This is as good a statement of the descriptivist position as I can think of, and it seems to me to be irrelevant to point out (as Haas, 1982, does) that people will nevertheless treat the dictionary as prescriptive: they will use it to find out how they ought to pronounce. If they do, this will not be because the presentation is prescriptive, but because the pronunciation that happens to be described here is viewed as one that it is desirably to acquire. If some other dialect had been described, people would not use the description prescriptively. I doubt, for example, if many people will use my own description of Belfast pronunciation (1981) in order to acquire fluency in inner-city Belfast English, although in principle they could. Jones’s English Pronouncing Dictionary is therefore descriptive, and not prescriptive, in exactly the same way as the description of any other dialect is. But it is also normative (as all such descriptions must be) in the sense that it attempts to reflect the socially agreed norms of some particular community of speakers –in this case the community (and social network) of RP speakers.

These first two principles can be described as requirements to acknowledge the importance in language descriptions of: (1) the social and situational context of speech-exchanges, and (2) consensus on linguistic norms of usage within speech communities. There is a third principle, which can be seen as an extension of the idea of consensus norms into the diachronic dimension, and it is based on the notion of language maintenance. If we assume that the natural tendency of language is to diverge, relatively convergent states can be described as arising from language maintenance through agreement on, or acceptance of, particular norms of usage in the community. To the extent that linguistic changes take place in speech communities, however, they take place against a background of language maintenance, and the extent to which they are successful depends on the interplay of these two sets of social influences –those that encourage maintenance (or stability), on the one hand, and those that encourage change (or divergence), on the other. Principle 3 is fundamental to the design of the Belfast research projects which were initiated in order to follow out some arguments about linguistic change. Our various discussions since then, about “language loyalty”, “focusing”, “social identity”, social network and related matters, all grow directly out of the idea of language maintenance in the research design itself.

 

             Language Maintenance and Language Change

 

The third principle can be stated thus:

 

 Principle 3  In order to account for differential patterns of change at particular times and places, we need first to take account of those factors that tend to maintain language states and resist change.

 

This is closely related to the actuation problem, which we discuss more fully in chapter 2 and the emphasis on language maintenance is the most salient difference between the way I have approached historical linguistic change and the approach of most other historical linguists. It also differentiates our sociolinguistic research from other work in that subject, including the large urban projects (such as Labov, 1966) that influenced the Belfast research in the first place. It gives rise to a number of consequential differences in approach. Historical linguists do not generally describe patterns of maintenance: they tend to focus on those things that are known to have changed and ignore those things that have not, and they can often explicate historical changes very elegantly without any reference at all to the social embedding of the changes concerned. What strikes me as important here, however, is the fact that if we focus exclusively on change and ignore maintenance, these non-social procedures can be quite easily justified: we can indeed propose sophisticated descriptions and highly constrained theories of linguistic change, without taking any account of social factors, and this is frequently done. However, if we pose the more basic question why some forms and varieties remain stable while others change, we cannot avoid reference to society. This is one of the justifications for Principle 3 in a socially-based model of change. Let me clarify it briefly.

If we are interested in how language states can remain stable and how speech communities resist change, we have almost no alternative but to take account of social factors. Suppose we notice that the structure of language X has remained stable for a century: it is not very interesting to point this out and then to leave it at that. We naturally want to know why it has remained stable when other states of language have changed, but in order to do this we have to study the social and speaker-based reasons that may account fort the fact that it has not changed. In reality, languages change at given times in some ways and not in others, sometimes they change rapidly and sometimes slowly, some varieties are divergent and some convergent, and so on. Thus, the third principle that I have proposed above is clearly relevant in a range of very diverse language situations, and at widely differing levels of generality, and I shall have much more to say about this in later chapters.

I noted above that our Belfast research projects depended especially on this third principle. It is the idea of language maintenance that is most immediately relevant to the historical interests of this research, and that is what this book is also about. If you look at historical sates of English, it is clear that some characteristics of the language have persisted through time while others have changed, and it is also clear, even in written documents, that early states of English were variable just as present-day English is. Therefore, we want to know how divergent forms and varieties of the language can be maintained across considerable periods of time, and how structured variability can persist through time. These interests influenced the original design of the Belfast research, and there were some other related interests (including dialectological interests, such as cross-dialectal comprehension and the question of access of non-standard speakers to so-called “standard languages”). What these situations all have in common is the maintenance of distinctive norms of language and (very often) persistence of divergent varieties through time, but historical linguistic theorists haven’t shown much interest in matters like this. Therefore, as a historical linguist, I thought that we might get a better understanding of what linguistic change actually is, and how and why it happens, if we could also come closer to specifying the conditions under which it does not happen –the conditions under which “states” and forms of language are maintained and changes resisted.

This mean, amongst other things, that we can see some of the traditional problems of historical explanation in a different light: for example, we might want to ask why apparently “low prestige” varieties of language can persist over centuries, and why dialects of the same language can be maintained for long periods in forms that differ so much form one another that they are mutually incomprehensible. But the most general consequence of an interest in maintenance is the one I have mentioned above: it forces us to ask questions about society and to investigate the structure of the societies in which norms of language are maintained and changes implemented. If we focus on change alone, we can propose explanations that are language-internal without systematic reference to social processes, but we cannot do this if we focus on maintenance; our answers have to be in some way socially based.

Of course, none of this implies that historical linguists never appeal to social explanations: sometimes they do, but the appeals tend to be ad hoc appeals to “prestige”, “the standard language” and the like, which assume the existence of speaker-links and power-structures in society, but which do not investigate these systematically. This has various consequences, but the one I need to call attention to here is that, as a result of the superficiality of the social analysis, accounts of the histories of particular languages have often been very heavily coloured by the social attitudes of the investigators themselves. Frequently, the researches have not been able to observe social structures and processes in an impartial way: their subjective social attitudes have often been based on ideological positions which they have simply assumed to be “common sense”, and so not easily open to rational examination. They have then imposed these ideological positions on to the analysis “form above”, as in the following: “Just as fashions in dress are bidding upon all members of a given class and are imitated by all who look up to that class, so fashions in language are bidding upon all people of culture and are followed by other members of the community to the best of their ability”(Sturtevant, 1917: 26). This is a unidimensional imposition of standards form the top; it is basically elitist, but quite mildly expressed. In some cases, as we shall see, the standard language ideology is much more strongly expressed, and in some accounts “non-standard” forms and varieties are rejected as if they were not really “language” at all. In chapter 5 we shall discuss the effects if standard-based attitudes on historical descriptions of English.

What we seem to need, therefore, is a theoretical orientation to the study of language maintenance that takes full account of social processes and therefore of social theories. Such an orientation, in contrast with theories that have focused on change alone, is in the fullest sense of sociolinguistic. In this book I shall explore the social side of our subject more fully than have been usual, and in chapters 6 and 7 I shall attempt to develop an integrated social model for the interpretation of language change. In section 1.4, we first acknowledge the importance of the “empirical foundations” of Weinreich, Labov and Herzog (1968)

 

             The “Empirical Foundations”

 

Whereas the three principles proposed above can be seen as basic in a context of language stability and maintenance, the “empirical foundations” of Weinreich, Labov and Herzog (1968) are quite specifically directed towards a theory of language change. Furthermore, whereas our three principles are very definitely about the social nature of language variation, their principles are in practice more directly focused on locating the linguistic patterns of change. They are devoted to supporting the claim that linguistic innovations move in an orderly manner through space (social, geographical, historical) affecting linguistic structure also in an orderly manner. Thus, although there is no necessary contradiction between their principles and ours, we have tended over time to reinterpret their principles in the light of our own experiences in social and sociolinguistic analysis. The essential difference here can perhaps best be understood by recalling that whereas Labov’s New York City study was specifically directed towards locating linguistic changes, our inner-city study of Belfast was primarily a study of maintenance, and therefore necessarily social for the reasons given above. Although it is assumed that change is in progress at any time in any speech community, we would not have been greatly concerned if, in this limited study, we had not found any evidence for change at all: we would have expected it to appear at some later stage.

According to Weinreich, Labov and Herzog, the task of explaining linguistic change is best divided into five parts –the problems of constraints, embedding, evaluation transition and actuation. The first three of these are interrelated and we have tended to treat them as aspects of the same thing. The problem of the universal constraints on linguistic change would, if it were solved, specify with changes are possible and which are impossible, and predict which changes would happen in particular circumstances: much historical linguistic theorizing has been directed towards a solution to this problem. Here, however, we need notice that although the term “constraints” has often been understood as intra-linguistic, it is obviously possible to speak also of social constraints on change. Let us consider this briefly.

 “Avoidance of homophony” (following Martinet’s (1955) arguments, and see further chapter 2) may be considered to be an intra-linguistic constraint of change. In this spirit, I have pointed out (J. Milroy, 1976a) that the development of [e:] (as in bait, sane) in Belfast vernacular (henceforth BV) to [ei] (as in RP) may be blocked by the fact that the vernacular pronunciation of the word-class of bite, sign is already pronounced with [ei] in BV. Thus, if the RP-like pronunciation were adopted, there would be merger of two distinct lexical sets. However, following Principle 2 (which is concerned with consensus norms), the decisive constraint here and elsewhere is just as likely to be social as linguistic, because despite homonymic clash (of the meat/meet type), mergers do commonly occur in languages. Furthermore, in this case there would have to be some social motivation for moving in the RP direction in the first place. Our evidence suggests, however, that there is a little or no such motivation: throughout most of the Belfast community there is no discernible movement in phonetic realization towards RP. Therefore, we can suggest very plausibly that the so-called “prestige” motivation to adopt Rp forms is overridden here by the solidarity constraint, which requires the speaker to conform to local community norms rather than to norms that are viewed as “external”. It is very striking, after all, that in our inner-city work there were no examples at all of RP-like [ei] in closed syllables of the type bait, sane, and very similar points can be made about a number of other words (for further discussion see chapter 3)

In the Weinreich, Labov and Herzog programme, all aspects are said to be both social and linguistic, but it is the evaluation problem that is most clearly designated as social. This pertains principally to social responses to language change “at levels of awareness, form overt discussion to reactions that are quite inaccessible to introspection” (Labov, 1982: 28). It embraces notions of prestige, attitudes to language (both overt and covert), as well as linguistic stereotyping and notions of correctness. In practice, we have taken a rather wider view of what goes on in speech communities than the Weinreich, Labov and Herzog principles imply, and have tried to look at evaluation within the context of broader structural principles such as power and solidarity (Brown and Gilman, 1960), and interactional factors such as “politeness” (Brown and Jameson, 1987) and “accommodation” (Giles and Smith, 1979, Trudgill, 1986a). We have also been interested in how speech communities can reach consensus on the evaluation of linguistic forms and how this consensus can shift in the course of time.

It seems, however, that the problems of evaluation and constraints can be viewed as constituent, or contributory, parts of the more general problem of explaining the embedding of linguistic changes in the pre-existing states of language and society. Labov’s original contribution here (principally in the New York City study) has been to provide a general model of the social location of a linguistic innovation and of the manner in which it spreads from a central point upwards and downwards through a speech community. But clearly, this overlaps conceptually with the transition problem, in so far as the transition form one state to another must be described here also. Thus, the graphs and diagrams of the quantitative paradigm, when they show stable patterns, can be interpreted a displaying aspects of the linguistic embedding of a variable; when there is a crossover pattern, however, they also display transition.

Transition concerns “the intervening stages which can be observed, or which must be posited, between any two forms of a language defined for a language community at different times” (Weinreich, Labov and Herzog, 1968: 101). Transition is what most historical description has been about, mapping (as it has usually done) the transition between state A at one period and state B at a later one. Quantitative analysis here allows a gain in sophistication: the process by which one form gradually gives way to another can be demonstrated in fine-grained detail in what is generally an orderly progression through social groups and speech styles (for a study of this kind, see Eckert, 1980)

The fifth problem –actuation –is a very different kind of problem from these others, and I shall not consider it here. We shall return to it in Chapter 2 and later chapters, and I shall suggest that a solution to it must be based on behaviour of speakers rather than primarily on the properties of languages. Here, it is appropriate to comment on some general differences that seem to exist between the Weinreich, Labov and Herzog methods of accounting for change in language systems and our own approach to it. The chief difference, as I have suggested above, is that whereas their programme and Labov’s work within it are based on first being able to locate linguistic changes in progress, our work has been more generally based on describing variation in the speech community and accounting for differing patterns (whether or not they exhibit change in progress) in social terms. In other words, our notion of the embedding of change in the speech community is broader, and as a result our idea of what a linguistic change actually is broader and less traditional than Labov’s view appears to be. It also raises the question of what a sound-change actually is: how do we know when we have located one, and how does the pattern of a change differ from other patterns that we might locate?

I shall discuss these matters more fully in later chapters, but I can lay some of the groundwork here. Traditional codifications of sound-change have generally focused on sound-segments as they “change” across time. Thus –to simplify –a linguistic change can be described as a change from A to B in some lexical set, such as that of Old English [a:] in stan, ham, which in the course of time becomes and [o:] –like vowel in PresE (Present English) stone, home. The transitional stages can be postulated in the Middle English period or studied directly, as in Kristensson’s (1967) study of onomastic sources from around 1300, from which figure 1.1 is compiled, showing the northward progress of the “new” rounded vowel in this set at that time (the rounded vowel has penetrated further northward in the west, that is, in Lancashire and the West Riding, than in the east, that is, in Lincolnshire). Labov’s treatment of sound-change seems to be quite similar to this traditional treatment in that a change is generally located by comparative methods (comparing different social groups in real and apparent time) within a single segment or a very limited class of sounds.

We have also used these methods, but our conceptualization of linguistic change is broader and is largely based on the normative principle that I proposed above: linguistic change is to be understood more broadly as changes in consensus on norms of usage in a speech community. During the process there will be some disagreement or conflict on norms at some levels in the community, but if a change is ever “completed”, then it will be possible to say that some community of speakers agrees that what was formerly A is now B. but this can apply at different levels of generality –from a single sound-segment up to a language state as a whole. Thus –to take a much more generalized case than, say post-vocalic /r/ in New York City –if a language state is observed to become more (or less) homogeneous within itself in the course of time, then the trend to greater or lesser homogeneity is itself a pattern of linguistic change that has to be accounted for in terms of consensus or conflict amongst speakers within the speech community.

We might wish to look, for example, at Australian English in this way. It is much more homogeneous than British English, even though the early settlers came from different places. If it has moved from an early heterogeneous state to a more homogeneous one, this is itself a linguistic change relevant to the history of Australian English and aspects of colonial language development in general. Thus, many of the detailed patterns that we are likely to find in sociolinguistic inquiries will not be unidimensional (as more traditional work often suggests), but bidimensional or multidimensional, showing trends towards greater agreement or greater disagreement on norms within the communities (see further J. Milroy, 1982b, and chapters 3 and 4)

 

            Synopsis

The purpose of chapter 2 is to place the general theme of this book in perspective by considering the relationship between historical linguistics explanations of a non-social kind and sociolinguistic explanations of language change. I do not attempt to review all current intra-linguistic work on change, as my main aim is to propose a different (social) way of looking at change, but I discuss some of the main trends. Following out my social arguments and Principle 1, I shall focus on conversational settings as the locus of change, and I shall suggest that language change is made possible to the extent that it is passed form person to person in speaker encounters, in which the apparently dysfunctional nature of language change is counteracted by features of the communicative context (this idea is developed from our early work in Belfast on speech and context: L. Milroy and J. Milroy, 1977)

Chapters 3 and 4 are chiefly about analyzing language in the community and interpreting the patterns revealed, bearing in mind Principles 2 and 3 (on linguistic norms and language maintenance). The main database is the Belfast research, from which I have selected examples. My purpose is to build the general foundations in these chapters, starting with the observation and analysis of language in the community. In chapter 5, I extend the perspective backwards in time, and consider some case-studies of language maintenance over periods of time that are relevant to historical interpretation. In this chapter we have to consider what we mean by a sound-change in history, and we conclude with a brief discussion of Neogrammarian views of how sound-change is implemented.

Chapters 6 and 7 are the main theoretical chapters and are devoted to the social modeling of linguistic change. In chapter 6, I outline a social model which is derived from the social network model that we have used to study language maintenance, and I proposed a model of “weak ties” to account for the possibility of language change. As this model does not account for broader social structures and processes, I explore in chapter 7 the links between network and social class in an attempt to build up a more comprehensive social model for the interpretation of language change.

 

        

  1. SOCIAL AND HISTORICAL LINGUISTICS

 

            Explaining Language Change

 

The ultimate aim of historical linguistics is to explain the causation of linguistic change. The question of causation is beset with difficulties but we can focus on it here by stating Weinreich, Labov and Herzog’s formulation of the actuation problem (1968: 102): “Why do changes in a structural feature take place in a particular language at a given time, but not in other languages with the same feature, or in the same language at other times? This actuation problem may be regarded as the very heart of the matter.” In attempting to solve the actuation problem we are concerned with no less than the origin of change: we want to locate its beginning and by any means possible attempt to explain why that particular change was initiated and diffused at some particular time and place. It seems clear that to tackle it, we must take account of how speakers initiate changes, and I shall treat it in later chapters in these terms.

The actuation problem, however, is so challenging that historical linguists do not usually address it directly; this is hardly surprising as, when it is formulated in this way, it is actually insoluble: a solution to it implies the capacity to predict, not only what particular change will happen, but also when and where it will happen. However, the probability of any event in life actually taking place at some particular and specified place and time is close to zero. Weather prediction is a convenient analogy here: we can predict from meteorological observations that it will rain on a particular day with a high probability of being correct, but if we predict that in a particular place it will start raining at one minute past eleven and stop at six minutes past twelve, the probability of the prediction being correct is vanishingly low. Nevertheless, we would be bad meteorologists if we did not try to improve the accuracy includes the ability to specify the conditions under which it will happen. In view of all this, we have no excuse as linguists for not addressing the actuation problem.

The Weinreich, Labov and Herzog formulation has several implications that are important for a theory of language change, and some of these can be understood fairly readily if we cite as an example a kind of sound-change that is frequently observed in languages and is sometimes called “natural”. So let us consider here the palatalization of /k/ before front vowels. Suppose it happens (as it often does) that one particular language (or dialect) undergoes this palatalization, whereas a closely related language (or dialect) of very similar structure does not. Following the Weinreich, Labov and Herzog principles, we have to ask why it happened in one variety but not in the other. We also have to ask why it should have happened at some particular time and not at some other time, when the structure of the relevant language presumably exhibited suitable conditions for the change at time when it did not happen as well as at the time when it did. There are, of course, well-known examples of varying developments of this kind: amongst the continental Scandinavian languages, Swedish and Norwegian have palatalization of Old Norse /k/, whereas Danish now usually has a velar: Old English underwent palatalization before front vowels whereas Old High German and Old Norse did not: hence PresE cheese for German käse and English/Norse doublets in PresE such as shirt/skirt; many Hiberno-English dialects (J. Milroy, 1981, and elsewhere) have [k]-palatalization in words of the type car, cart, whereas most other English dialects do not. What we observe here are conflicting patterns of change and stability in languages and dialects of similar structure. In these examples it seems that the proximity of the velar consonant to a front vowel may be a necessary condition for palatalization, but as it does not happen in every case, it is not a sufficient condition. We need to find out what the other conditions favouring or preventing the change might have been, and it seems that in cases where the change was adopted the social conditions must have been favourable; conversely, when it was not adopted, it may again have been social conditions that prevented the change. This suggests that to make progress in understanding actuation we must take into account the activities of speakers in social contexts in addition to the internal structural properties of a language.

Indeed, although linguistic changes are observed to take place in linguistic systems, they must necessarily come about as a result of the activities of speakers. As we have noted in chapter 1, languages which have no speakers (or –sometimes –writers) do not change, and so these remarks may well seem uncontroversial to any non-specialist who has given a language thought to the matter. After all, there is no point in having a language if it is not used by human beings. It seems to be specialist, rather than non-specialists, who think that language change, can be explained without reference to society. Within orthodox historical linguists, the emphasis has generally been on the properties of linguistic systems, and speaker-roles have been referred to indirectly and sometimes very vaguely. As Lass (1980: 120-2) points out, historical linguists have tended to regard language as an “autonomous formal system” or natural object and have preferred to believe that it is “languages that change and not speakers that change languages”. Thus, historical descriptions and theories of historical change have generally focused on the structural properties of languages and not on its speakers. This tendency is very deeply embedded in historical theorizing, and it is appropriate to look at it a little more closely here

 

            System-based Accounts of Change

 

The orthodox position as stated by Lass (1980) is not entirely a twentieth-century phenomenon: the separation of language from speakers has an ancient and honourable pedigree, and the nineteenth-century emphasis on the independent “life” of language is by present-day standards very striking. According to Trench (1888: 223-4), language has a life “as surely as a man or a tree”, and creativity in language in developing new forms is attributed by Max Müller (1881: 33) not to the creativity of speakers, but to the “marvelous power of language” itself: according to him (1861: 36) “it is not in the power of man either to produce or prevent” linguistic change. Müller’s adoption of the biological metaphor is so strongly started that for him it does not seem to have been a metaphor at all: linguistics, according to Müller, is literally a physical science on a par with geology, botany and biology, and not a historical science, such as art, morals or religion. “Physical science”, including linguistics, “deals with the works of God” whereas “historical” science deals with the works of man” (1861: 22). Language therefore does not have history, it has growth. The metaphor has weakened since Müller wrote, but there have been many publications on language history since then that have been based on the idea of the independent “life” of language. Indeed, the metaphor is by no means dead: this is amply demonstrated by continued references in recent work to “language birth”, “language death” and the “roots” of language.

Of course, it is not true that language is a living thing (any more than swimming, or birdsong, is a living thing): it is a vehicle for communication between living things, namely human beings. Hence, the metaphor has been largely replaced since the nineteenth century by a new metaphor based on the machine age: language is now more often seen as a self-contained system, like a working machine. The acceptance of this metaphor is widespread enough for it to appear in the title of a book on linguistics –The Twitter Machine (Smith, 1989) –and it is clearly greatly encouraged by developments in computer modeling of language. However, whereas the nineteenth-century metaphor could readily incorporate the idea of change (as language was said to have “growth” within it, like a plant), this system-based approach cannot so easily do so. Internal combustion engines, for example, do not initiate structural changes within themselves. From this point of view, therefore, the system-based model may seem to be an unsuitable one to use as the basis for studying language change. What is certainly clear is that within this perspective our attempts at explanation continue to be essentially language-internal. When speakers are referred to, they are decontextualized and asocial abstractions.

Possibly as a result of the emphasis on internal language systems, descriptive accounts (such as histories of English) commonly separate the internal history of a language from its external history (that is, the political, social and attitudinal contexts of a language). Thus, some historical accounts of English, such as Wyld (1927), have been mainly internal (typically focusing on sound-change and morphological change), whereas others (such as McKnight, 1928) have been about the external history of the language, discussing, for example, speaker-attitudes to variation as they were expressed by seventeenth- and eighteenth-century commentators. Both of these approaches can of course yield insights; however, it is commonly believed that the “real” history of a language is its internal system-based history and that the external history is relatively unimportant. The traditional position on internal and external histories has again been clearly stated by Lass (1987: 34-5), who claims quite explicitly that in most respects “external” accounts do not help to explain changes in linguistic structure. According to him, “there was nothing in the 17th century English political or social climate” that could account, for example, the merger of the formerly distinct vowels in words of the type bird, fern, hurt; he further states that “at the structural level there is no connection between language and society” and that “the internal life of language is close to autonomous.” My position, which I shall further develop below, is –on the contrary –that we cannot hope to explain change without inquiring into social factors.

One reason for this is that intra-linguistic arguments with only vague references to speakers, or accounts that explicitly reject speakers, are not in themselves capable of dealing with actuation, as it is speakers who actuate changes. Nevertheless, the causes of change (like the causes of illness) are multiple; therefore, we need to take both speakers and systems into account and, if possible, specify the link between speaker-activity and change in language systems. As for intra-linguistic theorizing, its main contribution has been to specify in a more and more refined way the linguistic constraints on change, not its causes. In order to exemplify the mode of argument used in system-based accounts, we can consider here Kiparsky’s recent review (1988) of progress in the study of phonological change.

This account is system-based and set in the traditional controversy about whether sound-change operates blindly and without exceptions (the Neogrammarian exceptionlessness hypothesis), or whether other approaches over the last century (such as lexical diffusion) have invalidated the hypothesis. The lexical diffusion model (Wang, 1969) holds that sound –changes may be lexically gradual: thus, in a change from /e:/ to /i:/ (such as the EModE (Early Modern English) change in words of the type meat, peace, leave), items are transferred to the new class at differential rates, often leaving a residue of items that do not get transferred (in this case such words as great, break, steak). Neogrammarian theory, however, has generally been interpreted to mean that the relevant class of items all undergo the change at the same time, that is, that sound-change is phonetically gradual and lexical sudden.

Kiparsky reconciles these approaches by arguing in terms of lexical phonology: those changes that appear not to fit into the Neogrammarian hypothesis (including instances of lexical diffusion, in which items in a class are not affected simultaneously by a change) are part of the lexical rule component, whereas Neogrammarian exceptionless change is accounted for by post-lexical rules. There is more to his argument than this, but I am not concerned here, of course, with the precise content of the argument –although I have tried not to do any gross injustice to it –but with its intellectual background, and specifically its intra-linguistic nature. In this argument, certain points are evident. First, the argument is system-oriented and not speaker-oriented (specifically, it is about phonological rules), and its goal is a “grammar” of linguistic change; the activities of speakers are not given prominence in the argument. Second, it is set in the traditional controversy about the regularity of sound-change in language systems (and certain other binary distinctions that arise from it, which attempt to specify constraints of change: for example, whether sound-change is lexically gradual or sudden). Third, it is proposed that the two patterns of change can be accounted for by fitting them into a new binary taxonomy based on lexical phonology. Thus, the discussion casts new light on an old controversy by redefining the controversy within a non-social system-based linguistics; it specifies the problem in a more refined and elegant way, and its proposals are then subject to critical discussion, testing and further refinements.

But this account, like many others, does not primarily address the actuation problem, and the question why (and how) speakers initiate changes is not central to the intellectual context in which it is conceived. It is true that there are two pages on “causes of sound change” that acknowledge the work of sociolinguists in this area –particularly Labov, Yaeger and Steiner (1972) and Kroch (1978) –but the discussion of “causes” does not form the backbone of the account. The main aim of historical phonology, as represented here, is the construction of sophisticated system-based grammars of change, and the problem of how speakers actuate changes is not directly addressed. It is interesting (and much to be welcomed), therefore, that Kiparsky includes an extensive and skilful discussion of the Labov-Sankoff variable rule paradigm. The significance of this here is that variable rules are, of course, themselves explicitly system-based (Cerdegren and Sankoff, 1974, and see chapter 6, below, for a fuller discussion of this point): there are grammars of language and not accounts of actuation. Thus, whereas speakers-based studies cannot easily be incorporated into orthodox system-based accounts, variable rules can fit perfectly and add sophistication to grammars of change.

A second quite recent example of intra-linguistic argumentation is Lass’s On Explaining Language Change (1980), to which I have referred above. Lass sates that linguists have proceeded language –internally –and have taken the view that it is languages that change, not speakers that change languages. But he is also critical of sociolinguistic accounts of change and seems to dismiss them. He says that when attempts have been made to introduce the speaker into explanations of language change, these attempts have been “superficial and otiose”, and he considers (p. 121) even Labov’s notion of “speech community” to be a “very tenuous abstraction” (more on this below). Taking his cue from the tradition, he points out that the most fruitful results for theories of language change have come about precisely because historical linguistics has studied “formal objects and their mutations over time, not … their inventors or users”.

The point that great advances in the past have come about through intra-linguistic argumentation cannot be disputed. One thinks especially of the great innovators of the nineteenth century, such as Bopp, Rask, Grimm and Verner, but also of modern advances, such as work on language universals, lexical phonology and many other areas. However, it plainly does not follow from any of this that we should therefore neglect the role of speakers in linguistic change. If we do not know what role speakers play, it seems appropriate that we should investigate it empirically.

Before we look at this more fully, however, we need to notice that it would not be correct to infer from what Lass says that older generations of scholars always neglected the role of speakers. On the contrary, there are several great names who assumed that linguistic change must have social origins (amongst others) and who did not think it beneath them to write extensively on this, while at the same time contributing fully to system-based accounts of language. Among these are Henry Sweet, H. C. Wyld, E. H. Sturtevant and Otto Jespersen. In Sturtevant’s Linguistic Change (1917), we find an emphasis on the idea of social norm-enforcement, childish errors and slips of the tongue, and (incidentally) a plea for the study of universals of language change. All this is presented in a framework that distinguishes primary change (effectively, linguistic change as admitted into grammars of language). Thus, a speaker/system distinction similar to that which we have proposed (J: Milroy and L. Milroy, 1985b) is considered by Sturtevant, and he attempts to integrate the two sides of the question. Jespersen’s Language (1922), which is a better-known book, reads in places like a research proposal for modern sociolinguistics and language acquisition studies: the possible causes of change include features of children’s language, sex-differences (there is a chapter on “The Woman”), taboo and euphemism (unfashionable at the moment, but unquestionably very important), language contact and Pidgin and Creole development.

It seems that if speaker-based arguments have since that time been found unsatisfactory, there may have been contingent reasons for this. The correct generalization seems to be, not that speakers are irrelevant to change in language systems, but simply that it has in the past been extremely difficult to study speaker-behaviour in a systematic and accountable way. As we noticed in chapter 1, early investigators, such as Wyld, did not have the necessary technology –in particular they did not have tape-recorders. So, until about 1960 it was very difficult to explore their ideas further by experiment or systematic observation and impossible to study conversational interaction in a reliable way. Dialect investigators were usually limited to single-word citation forms, and the empirical study of discourse in situational contexts could hardly have been contemplated. This, clearly, is no longer true. Although we are still interested in “formal objects and their mutations over time”, advances in data collection from live speakers, and the analysis of such data, have put us in a position to inquire into the role of speakers in change. If we do not know what role of speakers play in implementing changes –we do not need any longer to dismiss it as impossible to explain.

Before we go on in the next section to amplify this discussion by looking at the functions of change, we can draw together the strands of the argument so far. Historical linguistics in this century has largely depended on the idea that language is a self-contained system, and investigators have generally not systematically addressed the question of how speakers can introduce changes into the structural parts of language. Thus, complaints about the superficiality of “external” explanations of change are in a sense self-fulfilling prophecies. Clearly, if you concentrate exclusively on abstract systems of language and do not develop a coherent and accountable theory of the social embedding of  language change, your comments on the possible social reasons for changes will inevitably be ad soc and superficial.

But although the above remarks are concerned with theory rather than description, they are relevant to descriptive accounts also: for example, to historical descriptions of English. It is impossible to write a reasonably full history of English without making at least some reference to social categories, such as class, and to institutional aspects of language, such as standardization. Indeed, system-based accounts, such as that of Lass (1987) himself, do routinely make reference to “class”, “prestige” and other socially-based “external” categories. What most of these histories of language have done is to refer to these social matters in an ad hoc way, without contemplating an accountable theory of the social embedding and motivation of language change. Indeed, as we noticed in a social matrix, have filtered their social explanations through their own “common sense” views of social class and social prestige, without clearly acknowledging that these are themselves complex theoretical concepts, and certainly not “common sense”.

There seems to be no reason why we should not inquire further into social structures and processes as part of an inquiry into linguistic change, and no reason why we should not also think of linguistic change as being an aspect of social change in general, that is, to think of it an entirely new perspective. But to do these things is not to exclude the possibility of also developing sophisticated internal accounts of language change. Both kinds of approach are needed –and one should contribute to the other –because although linguistic change must be initiated by speakers (and is therefore a social phenomenon) it is manifested as internal to language. Bearing these questions in mind, therefore, I shall move on in section 2.2 to consider in what senses linguistic variation and change can be shown to be functional. It is important to consider this question, because from an intra-linguistic point of view variation and change can actually appear to be dysfunctional.

 

            The Functions and Malfunctions of Change and Variation

 

In the introductory sections of chapter 1, we were mainly concerned with structural aspects of language, exploring the apparent conflict between the structuralist axiom –that a language is a coherent self-contained system of interdependent parts –and the fact that language is continuously changing. In what follows I am concerned with function: I want to consider whether linguistic change is functional, that is, whether it serves a purpose of some kind, and if so in what sense it is functional. Here we need to recall that it is very widely asserted by linguistic scholars that a language system is at any given time equally well adapted to the functions for which it is used: sometimes it is said to be perfectly adapted. Whether or not this is true, it is reasonable to assume that linguistic structure is very sensitive to the changing social and communicative needs of speakers. Furthermore, Weinreich, Labov and Herzog (1968) have claimed that linguistic variation (which can be a symptom of language change) is not only structured, but also functional, and that it is the absence of variation that would be dysfunctional. In order to investigate the functions of language, however, we have to look at language in use, but as the functions of language in use are social and pragmatic, they are not readily accounted for within an exclusively system-based theory of language structure. We therefore have to look at speakers in addition to systems. Traditional speculation as to whether linguistic change is functional or not has, however, been based on the configuration of systems rather than on speaker-use of language.

Speculation about the functions of change has been common in the past: it has, for example, been suggested that there is a theology, or overall purpose, in language change (for a discussion of this, see Lass, 1980), and even that change in language structures may follow a predestined path. This is plainly related to the idea of language as an independent “growth” with a “life” of its own. In this view of language, however, speakers would play little part in this development and could do very little about linguistic change, as the blind force of language in its purposive quest would overrule them (recall again the views of Müller, 1861). Related to these ideas in a general way is Sapir’s (1921) notion of “drift”. It is striking, as Prokosch (1939) points out, that related languages can undergo the same changes apparently independently: all the early Germanic languages, for example (except Gothic, records of which are too early), independently underwent the important change known as front-mutation or i-umlaut. It is as if the “parent” language was programmed in such a way that the conditions were already present in it for the change to take place in all the “daughter” languages.

It has also been suggested that language can make progress and improve in the course of time: for Henry Sweet (Henderson, 1971), the loss of grammatical gender in English constituted progress, as grammatical gender, according to him, is “illogical”; Jespersen (1922) argued that modern languages of “analytic” (weakly inflected) structure, such as English and Danish, have envolved to a higher stage than “synthetic” (heavily inflected) languages, such as classical Latin, and are more efficient instruments of communication, mainly because they are thought to have a more transparent one-to-one relationship between meaning and form. For example, an auxiliary verb phrase, such as English I have said…, is more transparent than the Latin equivalent dixi because person, tense and aspect in English are expressed in separate units, whereas in Latin they are all carried in the second syllable of dixi and cannot be unraveled from its surface form. Jespersen’s judgement here depends, of course, on the assumption that transparency in language structure is a desirable thing. Clearly, however, if some other criterion were used (and such criteria as euphony, economy and elegance have also been used), completely different conclusions might be drawn, and Latin might then be held to be superior to English, as it often has been; therefore, it is safer to assume that differences in overall grammatical structure are neutral. They can all be used equally efficiently or inefficiently by the speakers of the language concerned.

I have described these kinds of argument as speculative because it is impossible to test them empirically: they are based on value judgements, which are often ethnocentric or class-based, and in the wrong hands they can sometimes lead to quite damaging opinions. For example, if it is believed that the structure of one language is superior to that of another, it can then be suggested that this is due to the cognitive superiority of its speakers. In fact, there seems to be nothing inherent in the structural properties of a language to suggest that change has a positive function within language structure. Within these structural parts –phonology, grammar, lexical and semantic structure –it is quite impossible to demonstrate empirically that language systems have in-built tendencies towards progress or decay, that one language is more or less “efficient” than another, or that there is a teleology in linguistic change. If change is functional, it must be speaker-functions that are involved.

In fact, if we focus exclusively on the internal properties of language, it is much easier to make a prima facie case for the argument that variation and change in language are dysfunctional, rather than functional. This is because dialect divergence and language change lead to the difficulties in communication between speakers –a fact that is obvious to any fieldworker who studies a “divergent” dialect, and to any analyst who transcribes the tape-recordings (they can be very difficult to understand); it is also familiar to anyone who tries to read an early English text. Therefore, if linguistic structure exists for the purpose of successful communication, why does language change? Why do languages diverge from one another in the course of time, and why are some dialects of a language partly or wholly incomprehensible to speakers of other dialects? The apparent paradox has of course been noticed by many, and it has been quite recently commented on by Francis (1983: 15-16):

 

If the purpose of language is communication, it would seem that the more homogeneous the language, the more efficient the communication. Why, then, does the increasing variation resulting from differential change not make communication difficult, unreliable, and eventually impossible? If so, the propensity for language to change would ultimately lead to break down in the principal function for which language exists.

It is interesting that Francis –as the dialectologist –seems to presuppose here that increasing divergence does not lead to breakdown of communication. One only has to look at the histories of related languages that were at one time dialects of the same language to realize, of course, that it does lead to breakdown. There is no need to disagree with Wyld (1936: 7): “the process of differentiation is almost infinite, and the tendency of language is not, as it has sometimes been wrongly said, in the direction of uniformity, but of variety.” Language change can, and does, lead to breakdown of communication.

The belief that language change is dysfunctional is most clearly expressed in popular attitudes to language. These commonly conceive of languages as ideal and perfect structures, and of speakers as awkward creatures who violate these perfect structures by misusing and corrupting “language”: this is essentially a belief in the rape of language by speakers. These attitudes are strongly expressed and highly resistant to rational examination. So strong is this intolerance of speaker-variation and change that in many countries academies have been set up to enforce a uniform “correct” usage and to prevent uncontrolled divergence; indeed, in the biblical story of the Tower of Babel, language diversity is attributed to original sin. There are, of course, socio-political and economic reasons for these attitudes, which we have discussed elsewhere (J. Milroy and L. Milroy, 1985a), but they are powerful and deep-seated and they cannot be ignored.

In so far as these prescriptive agencies have a rational purpose, this purpose seems to be maintenance of communicative efficiency in carrying information-bearing messages over long distances and periods of time. For conveying information in these ways, uniformity and standardization of language are highly valued, and it is usually in the written channel that the highest level of this kind of efficiency is achieved. Spoken language, as far as we can tell, continues to vary and change, and it is in spoken language, and not in writing, that structural (for example, phonological) change is implemented. Writing systems promote uniformity and suppress variation and change it is typically in the day-to-day situational contexts of speaker-interaction that structural changes take place, and it is in these contexts that they have to be investigated. That is why Principle 1, which I put forward in chapter 1, should be borne in mind.

Popularly expressed attitudes to correctness and uniformity, therefore, do not square with what human beings actually do in conversational contexts. On the one hand, they believe in uniformity, while on the other they promote diversity. Clearly, if speakers consistently carried their expressed beliefs into practice, the result would be a uniform and stable state of language –the world of the idealized native speaker in a perfectly homogeneous speech community –and not the diversity that exists in the real world. It seems, therefore, that in expressing adverse judgements on variation, speakers are subscribing to the notion that the main function of language should be the successful communication of information-bearing messages (as in the writing system), and there is plenty of evidence that from this point of view variation and change can indeed be dysfunctional. But there is also plenty of evidence that the successful communication of information-bearing messages is not the only function (or necessarily the main function) of language in use. To clarify the argument, I shall now consider some examples of communication difficulties that arise from language diversity.

 

        Malfunctions of Language Diversity

 

Cross-dialectal miscomprehension was one of the interests that led to the setting-up of the Belfast research programme. We were interested in the first place in how commonly differential linguistic structures could lead to miscomprehension, and more specifically whether “non-standard” speakers actually have the easy access to standard English that is so often assumed. This has further theoretical implications –especially for the idea of the “polylectal” or “pan-lectal” grammar of English, which was a current interest in the 1970s. within the projects, we collected many examples of cross-dialectal divergence and miscomprehension, of which the following is an example. Speaker A is a speaker of a Hiberno-English dialect, and B and C are “standard” speakers:

(1)   A How long are yous here?

       B Oh, we’re staying till next week. (Silence of about 2 seconds)

       C We’ve been here since Tuesday.

       A Ah well, yous are here a while then

      

The miscomprehension here is indicated by the period of silence after B’s reply, and it arises from the class between different linguistic systems. In certain constructions Hiberno-English dialects consistently use the present tense where standard English uses the present perfect. Thus, “How long are yous here?” (In HibE) means How long have you been going here? Whereas in SE (Standard English) the present tense form means (or at least implies) How long are you going to stay here? Or What is the total length of your stay? Speaker B construes the HibE utterance according to the SE rules. An “appropriate” response, however (one which would be immediately perceived as relevant to the question in this conversation), would in HibE refer to the past (for example, “since Tuesday”) and not to the future. While it is true that breakdowns occur for many reasons other than this, there is no doubt that breakdowns arising from the different structures of divergent dialects are quite common (see also Berdan, 1977; Trudgill, 1981), and that they are naturally perceived as inconvenient when they are noticed.

Sometimes, however, there is no period of silence or hesitation, as in (1), and the conversation proceeds apparently normally, even though there actually has been a misunderstanding. The hearer continue to believe that the first speaker has said something quite different from what was intended, or the hearer may wait for the miscommunication to be “repaired”, or it may simply not seem to matter very much at the time. But we do have attested cases where the miscomprehension could be quite serious. In some Hiberno-English (and Scottish) dialects, the conjunction whenever is used almost equivalent to when. Thus, if someone says “whenever my husband came in he beat me”, we know from observing Ulster usage that the reference is to one occasion only the speaker is not suggesting that her husband beat her every time that he came in. an outsider, however, is likely to interpret this to mean that the husband beat her many times. Consider also the following example:

 

(2)   A Do you think he’s going to die?

B I doubt so

 

B, who is a speaker of Hiberno-English, does not mean that he believes that the person will not die; his meaning is the opposite. He is saying that he is afraid that the person will die (the utterance means approximately “I’m afraid so”). In this case, I was Speaker A, and the usage was known to me as widely distributed one in Scots and Irish dialects. If the interlocutors do not have access to the two different systems, however, there will be a miscomprehension. Thus, if a newscaster says “Mr. Major doubts that the economy will deteriorate”, speakers of these dialects may interpret this to means that he is afraid the economy will deteriorate, when the speaker actually means that he thinks that it won’t deteriorate.

These cases of breakdown arise from deep-lying differences in the syntactic structures of the different dialects, which make the pandialectal grammar (Bailey, 1973) of English a dubious proposition. The semantic distinctions carried in the Hiberno-English tense/aspect system are structured totally different from Standard English (Harris, 1984; J. Milroy and L. Milroy, 1985a); thus, the miscommunication in example (1) is due to major differences in the abstract rule-systems of these dialects. Furthermore, all those involved in the miscommunications reported above are native speakers of English, but it appears that their “native speaker intuitions” (or competences) do not extend to the rules of all relevant varieties. These examples also show, therefore, that our native speaker competence does not necessarily guarantee comprehension of varieties that are removed from us in time, space or social space. To put it simply –we do not have total comprehension of dialects divergent from our own.

The assumption that speakers of a language are (or should be) mutually comprehensible is rather basic in popular attitudes to language, and it also seems to underlie some professional linguistic approaches (for example, Smith and Wilson, 1981). Clearly, if there can be miscomprehensions arising from structural differences in the dimensions of spatial and social variation, it is likely that in the chronological dimension of change there will also be miscomprehensions. A change entering the language of a younger generation, for example, may well be miscomprehended by older generation speakers. However, the idea that mutual comprehensibility between generations is a constraint on possible changes is quite deep-rooted amongst historical linguists (see Lightfoot, 1979: 376, for a justification), and is often taken for granted. But it is dangerous to assume this too lightly: we cannot demonstrate that mutual comprehensibility between generations (or between groups of other kinds) is always a necessary factor in determining which linguistic changes are possible and which are impossible, and sociolinguistic investigations (from which most of the data reported in this book are derived) strongly suggest that it is not always primary. It may be that some innovative groups do not particularly wish to be comprehensible to others, or that rapid social change (for example, in the genesis of Pidgin/Creole languages) overrides the principle of mutual comprehensibility. The constraint proposed by Lightfoot is, it seems, a variant of the idea that heterogeneity is necessarily dysfunctional, and that mutual intelligibility is functional. It seems to me, however, that the question we should be asking is why linguistic changes that lead to miscomprehension (and divergence into mutually incomprehensible varieties) can happen at all. What is the motivation for such changes, and what is their function?

The examples discussed above concern variations in language structure that lead to comprehension difficulties, but they have also raised, in a much more general way, the question of what actually happens in conversational exchanges between speakers. In section 2.4, I shall be concerned with this latter point: I shall attempt to show that is the multiple speaker-functions of language in use that make linguistic change possible and suggest that we must look at these speaker-functions if we are to make progress in understanding the nature of the language change. This implies a modeling of the locus of linguistic change that differs from system-based models of the kind that we have discussed. The language-internal pressuppotions of these models have been projected on to the mental capacities of human beings, and linguistic change has thus been seen as consisting primarily of changes in the mental representations of the speaker in the form of rule-addition, rule-loss, and other rule-changes. But as we cannot directly observe mental representations (whether they are described as rules or in some other way, for example, as parameter-settings), suggestions of this kind are somewhat more speculative than matters that can be verified by observing speaking-interaction. In order to propose a more socially realistic account of change, therefore, Labov has argued that the locus of change is not in the individual speaker, but in the group, or at least that we have to look for it in group behaviour. What is implied here is more specific than that: it is that linguistic change is located in speaker-interaction and it is negotiated between speakers in the course of interaction, much as other aspects of discourse are negotiated between them. Bearing in mind this speaker/system distinction, therefore, I shall introduce the discussion of speaker-function in section 2.4 by first considering the way in which functional change has been handled in system-based historical linguistics.

 

        The System-oriented Approach to Function

 

The system-oriented (as opposed to speaker-oriented) approach to functional change originates with Martinet (1955) whose arguments depend on the information-bearing function of the language and the presume need to preserve mutual intelligibility. I have commented briefly on this chapter 1. Martinet proposes that in phonetic/phonological change a phonetic opposition that is useful in maintaining meaning-bearing distinctions will, other things being equal, resist neutralization and loss of distinctiveness. Thus, for example, if the speakers of English had found it useful in communicating information to maintain the EModE phonetic distinction between words of the meet, beet class of words of the meat, beat class, they would, according to the theory, have been inclined to maintain it. The fact that they did not (in some dialects) requires a functional explanation, which in effect proposes that the distinction between the two phoneme classes was no longer useful in maintaining meaning-bearing distinctions, and this is generally argued in terms of functional load (see specially Samuels, 1972). It is suggested that one of the two categories had low functional load in distinguishing between words in ordinary usage; for example, the words in that class might not have been used very often, and when they were used, they were perhaps used in contexts where confusion with the other class was unlikely to occur (thus, as meet is a verb, whereas meat is a noun, it is unlikely that you will misunderstand I meet him as * I meat him).

This type of functional argument can also been used to explain exceptions to “regular” changes. For example, the idea of homonymic clash may be used to explain what certain items in early English did not undergo regular development: if they had, they would have become phonetically identical with other forms. Lass (1980) cites the item shut, which if it had developed regularly would have become identical with shit: according to the theory this was prevented by the functional need to keep the items distinct. However, it is easy to find counter-examples, where the need to prevent homonymy did not operate: for example, homonymy of rush (‘hasten’) with rush (a plant) has not been prevented. In some circumstances quite dramatic loss of distinctions can take place, and there are examples in the literature of wholesale merger of previously distinct classes of items: Labov, Yaeger and Steiner (1972) cite the reduction of five distinct vowel phonemes of ancient Greek to one (/i/) in modern Greek. Because of apparent counter-examples like these, Lass (1980: 75-80) points out that these functional arguments as applied to given cases are unsatisfactory. Social dialectologists are in little doubt that speakers will happily tolerate a great deal of phonemic merger, allophonic overlap and approximation, and homonymic clash.

None of this, however, detracts from Martinet’s original insight, which recognizes the importance of speaker-function –in this case the information-bearing function: the apparent circularity in arguing about given instances is probably very largely due to the limitations of the historical database (which is incomplete and deprived of situational context), and to some reluctance to investigate what happens in load can be more fully tested. Historical linguists know that language is used to convey information, but they cannot specify very easily what additional social and pragmatic functions might have been involved in particular changes (which after all took place long ago in circumstances that we cannot fully investigate). Yet, although the information-bearing function is the one that comes most easily to mind, it is only one of the functions of language in use: other functions interact with it. Thus, in historical change, the need for mutual intelligibility over distances and the maintenance of meaning-bearing distinctions can be overridden by these other functions –by the identity function, for example –This is suggested by the cross-dialectal miscomprehensions discussed above, by the pattern of multiple merger in Greek cited by Labov, Yaeger and Steiner (1972), and by many other instances.

In this connection, it is useful to recall the metaphor that I appealed to in chapter 1 –the idea that in viewing language as a system où tout se tient we are implicitly comparing it to a machine. The function of a machine is known beforehand: the function of an internal combustion engine is to propel a vehicle, and if it does not succeed in propelling the vehicle, we know that it has malfunctioned. If we believe that we know beforehand that the principal function of a language in use is to communicate decontextualized information-bearing messages explicitly and unambiguously, then –pursuing the system-based metaphor –we shall have to view communicative breakdowns as malfunctions. Of course, in given situations, they may be perceived in this way, and if we put the argument on a more abstract and general plane, we shall then have to conclude that language change and variation are in general dysfunctional. But if this is so, why then do human languages vary and change?

We can draw only one conclusion, which is that variation and change must also be functional for speakers of languages: if this were not so, languages would be uniform and they would not change. Indeed, there is a case for claiming that if linguistic change were impossible, speakers could not function adequately in speech communities. But the functions of language in speech communities are multiple and are not limited to the information-bearing function. In social dialectology we attempt to explore the question of function by observing and analyzing the language of ordinary speakers in conversational contexts. To return to the machine metaphor: we can take for granted the function of a machine, but we cannot take for granted the functions of language. Thus, as we do not know beforehand what all the functions of language in the speech community might be, we have to find out what they are and how they interact with one another by exploring the speech community. This is by far the single strongest justification for research in social dialectology of the kind that we and others have carried out.

 

            Speaker Functions in Discourse and Conversation

 

What I intend to do here is to tackle the question of discourse functions in an introductory way, without venturing too far into the enormous literature on discourse and conversational analysis. We can start by noticing some of the binary distinctions between types of discourse that have been suggested. Ochs (1979) proposes that discourse can be divided into two broad types: planned and unplanned. Brown’s (1982) distinction between message-oriented and listener-oriented speech is based explicitly on the functions of discourse in social settings. Message-oriented speech is characterized by explicitness and independence of situational context or shared knowledge between participants. According to Brown (p. 77), message-oriented speech is “goal-directed. It matters… that the listener understands… and that he understands… correctly. The point of the utterance is… communication of a propositional or cognitive (information-bearing) message to the listener.” Listener-oriented speech, on the other hand, is characterized by inexplicitness and vagueness, with primary attention to the feelings and attitudes of conversational partners. As Brown comments:” it is often the case that speakers in primarily listener-oriented dialogue don’t seem to be talking about anything very much… we may judge it to be successful if the participants succeed in maintaining friendly relationships.”

Historical linguistic views of language function tend to assume without comment that the function of language is something like the message-oriented function –which is often described as communicative –and not like the listener-oriented function (in which the meanings conveyed are social or context-depended); they assume that communication between speakers is, or ought to be, explicit and context-independent in the interests of conveying cognitive propositions efficiently from speaker A to speaker B, in a context where the aim is to convey new information to someone who does not already know it. Let us consider, however, what this would actually mean in terms of the use of language in social contexts.

Clearly, it would allow the written medium, as the function of writing is to communicate messages outside the immediate interpersonal context, and, to be effective, it must be explicit. It would also allow certain kinds of speech event, such as reports or lectures, for much the same reasons. In conversational exchange, however, such a view of language use would account for only those parts of conversations in which “new” and explicit information is given and received; for example, the information that is passed in question-answer adjacency pairs of the kind.:

 

(3)   A Where were you on the night of 15 August?

B I was looking after Mother at the motel

 

Such exchanges are, however, especially characteristic of, and frequent in, formal settings, such as classrooms, courtrooms and interviews of various kinds (including some sociolinguistic ones) and may indeed be required in such settings: they are not especially characteristic of conversation. The conversations that we collected and analyzed in the inner-city Belfast projects (L. Milroy and J. Milroy, 1977), for example, could not have been adequately described in these terms. Speakers in most conversational contexts are not solely (or even mainly) concerned with passing decontextualized new information to one another, and the casual conversations that sociolinguists record are not mainly made up of question-answer adjacency pairs or elicitation-response sequences. On the contrary, much of our discourse is unplanned and listener-oriented, in which speakers “don’t seem to be talking about anything very much” and in which the primary goal seems to be maintenance of “friendly relationships” (Brown, 1982: 77). It can be assumed, therefore, that speakers in casual social contexts are not usually concerned with avoiding homonymic clash or with being especially clear and explicit: they are satisfied if the conversation progresses successfully, and the success of the conversation is judged in social terms. If misunderstandings occur because of homonymic clash or for any other reason, they can be repaired if necessary: speakers appear to accept the results of vagueness and ambiguity on the assumption that “intended” meanings will be clarified if necessary as the conversation proceeds.

Underlying the distinction between discourse functions there is another more general distinction. This is the stark contrast between what is desirable in the written medium (or context-independent speech- styles such as lectures) and what is desirable in, and characteristic of, speech-exchanges in social settings. It is very clear that much of the historical linguistic tradition has been based on assumptions derived from the functions of writing, rather than speech. However, such features as redundancy, vagueness and ambiguity, which are disfavoured in writing, are wholly characteristic of everyday speech. Furthermore, many of the features that are positively dysfunctional in context-independent language are actually functional and necessary in the conduct of successful conversation: lack of explicitness, hesitation, ambiguity, incompleteness and repetition are themselves very important aspects of how conversation is organized.

This has been clearly demonstrated by conversational analysts such as Schegloff (1979). Far from being random and disorganized, ambiguity and the other characteristics we have mentioned are systematic strategies of conversational interaction. They are used for monitoring the reactions of conversational partners and for clarifying and repairing the mistakes or misunderstandings that might have occurred in the interaction. In fact, one of the most important aspects of conversation is the very high value that is placed on indirectness: this is in obvious contrast to message-oriented styles, in which directness is valued and indirectness disvalued. Direct imperatives (demanding actions), for example, and direct interrogatives (demanding relevant responses, as in (3)) are quite rare in conversation. This is because exchanges in speech are social and personal: the high value placed on indirectness is a result of the speakers’ concern for their own “face” and that of their partners (Brown and Levison, 1987): they are often much more concerned with being polite, that is, avoiding threats to “face”, than with passing information efficiently and economically. In spoken contexts, the directness that is so highly valued in information-bearing styles is perceived as threatening.

These remarks may be sufficient here to draw our attention to some of the functions of language in use apart from the message-oriented function, and they have some consequences for our ideas about how linguistic changes are implemented. Theories based exclusively on the message-oriented function of language must plainly be insufficient: it seems very unlikely that linguistic changes over time could have been implemented mainly in this function, that is, in formal styles such as lectures, or in formal settings such as courtrooms or classrooms, especially since this function values stability and resist change; for the most part, changes have been initiated in countless millions of casual (mainly unplanned, listener-oriented and context-dependent) encounters between speakers. It is in these casual exchanges (and not primarily in formal settings) that the sociolinguist looks for the evidence of change in progress. It is the use of speech in these context-tied situations that actually allows linguistic change to be negotiated between speakers, and we need to emphasize this here, because linguistic innovations plainly belong to the class of phenomena that may be miscomprehended in context.

If we can accept that is characteristic of conversation that some utterances will be miscomprehended, we can also presumably accept that is the principles of conversational organization that permit repair of such miscomprehensions. Clearly, it is possible that in the course of a conversation innovations may be introduced, and, as they may be unfamiliar to some participants, they may be miscomprehended; however, as we have seen, the principles of conversation allow for clarification of these miscomprehensions. For this reason, and for other contextual reasons, the principles of conversation permit linguistic changes to be negotiated between speakers and thus admitted into the language system: it is the conversational context that provides the conditions for change to be accommodated. Decontextualized discourse, on the other hand, does not in principle cater for these misunderstandings: it therefore resists ambiguity and vagueness, and in the present perspective it is hardly surprising that in its written forms especially it has also been observed to resist structural change.

 

            Speaker-Functions: Marking Social Roles

 

The discussion so far has focused on the functions of discourse in conversational settings, and I shall continue to refer to these pragmatic aspects of language from time to time. However, there is another relevant aspect which is largely independent of the idea of discourse types, and which has been prominent in social dialectology since Labov’s (1966) New York City study. This proposes that variation within the structural parts of language (such as phonological and morphological variation) is used by speakers to mark varying social roles. Social meanings –in the words of Blom and Gumperz (1972) –are carried in linguistic structures. Style-shifting and code-switching (switching from one language or dialect to another) are socially functional: they are related to changes in the situational contexts of speech events, to the social characteristics of the participants, and to the varying purposes of exchanges in speech. Furthermore, speakers normally attach great importance to this kind of variation and assign strong social values to what are essentially arbitrary differences. To exemplify this kind of function, I shall now briefly consider the question of style-shifting (which functions in essentially the same way as code-switching)

The fact that style-shifting is functional is rather neatly demonstrated in cases where the expected variation is, apparently, absent. We have discussed such a case in some detail elsewhere (J. Milroy and L. Milroy, 1985a: 123-5), and this example is a useful demonstration of Weinreich, Labov and Herzog’s (1968) point –that the absence of variation is dysfunctional. According to Lavandera (1978), members of the Argentinian Italian community, who are bilingual speakers of Italian and the cocoliche dialect of Spanish, are perceived by monolinguals as deficient in their Spanish. However, there is actually nothing deficient about the structure of their Spanish (as Lavandera shows): what is lacking is the stylistic variation, sensitive to occasion of use (and other factors), that monolinguals observe. There is a lack of stylistic variability, not a deficiency in command of the “core” structure of the language. The perceived absence of variation in this case demonstrates, of course, that stylistic variation is –for the monolinguals –functional; that is why we notice it. For cocoliche speakers, on the other hand, it is their bilingualism that is functional: their communicative competence is exhibited in their command of two languages, rather than in observing the stylistic variation inherent in one of them.

But it is also clear that speakers may have very strong feelings about particular regional or social dialects, even though in linguistic terms the differences between dialects are arbitrary. One of the first things I noticed in my early descriptions of Belfast vernacular was the strong “stigma” associated with certain non-standard pronunciations. In a dynamic account of the phonology (J. Milroy, 1976a), I found it most convenient to think in terms of the avoidance of stigma rather than convergence towards a higher-class or standard form, but it was not clear why some non-standard forms were avoided and others favoured. Amongst younger inner-city speakers, it was clear that stigma was attached to certain forms of rural origin (such as palatalized /k/ and dental /t/), but not to the others (the raising of /a/ before velars showed no recessive tendencies). Thus, it seems that communities can disfavour pronunciations that were formerly favoured; and despite the fact that sound-segments do not carry meaning in the ususal sense, this phenomenon seems to have something in common with the operation of taboo, which is so well attested in the vocabulary. It is also associated with the “face” and politeness.

These brief discussions of social dialect, style-shifting and conversational functions are sufficient, I think, to indicate that there is more to language in use than the communication of decontextualized information of a purely cognitive kind. It now seems appropriate to summarize the main points, as I have not been trying to argue merely that language in use has multiple functions –this is a view that will be readily accepted by any sociolinguist. I have been concerned with the limitations of what I have called system-based historical linguistics. By this I mean virtually any approach to historical linguistics (traditional or current) that is centred entirely, or almost entirely, on the properties of language as an abstract object, and that excludes the systematically observed behaviour of speakers of languages. These language-internal approaches have made immense progress in producing sophisticated “grammars” and models of linguistic change, but they have not come very close to the actuation problem and the causes of change. Furthermore, as we have noticed above, when speaker-roles are referred to in system-based arguments, they tend to be referred to in a rather ad hoc and unsystematic way, usually on the assumption that the message-oriented function of the discourse is the one that matters. We therefore need a theory of the embedding of language change in society, but we do not so far have a social theory of this kind that can rival the sophistication of system-based linguistic theory. Yet, it seems that we cannot develop such a theory if we remain wholly within the constraints of orthodox historical linguistics. In section 2.7, therefore, I shall conclude this chapter by commenting on a state of affairs that underlies some of the matters raised in this introduction, that is, the limited nature of the database of historical linguistics. By discussing this we can focus on the methodological interface between historical and social linguistics and go in later chapters to suggest what we involved in a socially-based theory of language change.

 

        Limitations of Historical Inquiry

 

It is obvious that data preserved from the past are likely to be more limited in certain ways than data collected at the present day. Here we notice two major limitations. The first is that past states of language are attested in writing, rather than in speech. This has many consequences, of which the most general ones depend on the fact that written language tends to be message-oriented and is deprived of the social and situational contexts in which speech events occur. This is relevant, of course, to our discussion of function, above, and to other matters that will arise in later chapters: for example, interpreting written texts as evidence for pronunciation.

The second limitation is that historical data have been accidentally preserved and are therefore not equally representative of all aspects of the language of past states. Thus, whereas research into present-day states proceeds in a controlled way by collecting and analyzing data for the specific purpose of drawing generalizations about language and about specified aspects of language, the researcher into past states must use materials which were not in the first place collected for this purpose. Some styles and varieties may therefore be over-represented in the data, while others are under-represented. For some periods of time there may be a great deal of surviving information: for other periods there may be very little or none at all. It is reasonable to say that the database of historical linguistics, as compared with that of sociolinguistics, is impoverished.

 To the extent that historical linguistics is subject to these limitations, it is what Diaconis (1985) has called an “uncomfortable” science. In this respect it is similar to some aspects of other sciences such as geophysics, macro-economics or astronomy, in which the scientist has relatively little control over the database. The astronomer, for example, does not have experimental control over the visits of Halley’s Comet: thus, just as the astronomer does not have control over space, so the historical linguist does not have control over time; to be more specific, historical linguistics does not have experimental control of its database, and so it cannot always isolate the variables that may be involved in an explanation. It is quite appropriate here to mention the analogy of the blind men and the elephant, which has so often been mentioned before, or to use metaphors of the “tip-of-the-iceberg” type, because it is very much a matter of proceeding from a base of very limited knowledge. Thus, whereas social dialectology can plausibly claim to be some to some extent an experimental science (because it is possible to control some variables in the frame of the investigation), historical linguistics cannot. Sometimes the data may be so impoverished that decisions cannot be made as to what is the best description amongst a set of possible descriptions, or what is the best explanation amongst a set of possible explanations. The result is that interpretations of the surviving evidence are often strongly dependent of current theoretical assumptions (which may of course be dubious) and, more widely, on current ideological positions (which are even more dubious). Many examples of this difficulty can be cited: here I shall refer one briefly to a rather general one.

One example of a difficult area for the historical investigator is the chronology of sound-changes in the history of a language: this can often be uncertain and controversial. As a result of the limitations we have noticed, and the imposition of certain theoretical orthodoxies on what is or is not possible in sound-change, there has been a strong tendency in historical descriptions to assign a date to a sound-change at what seems to be a time of its completion, and (until recently) for relatively little interest to be shown in the earlier stages of change. Indeed, sometimes it seems as if change in a whole phoneme class is believed to have taken place all at once with simultaneous actuation and completion –perhaps overnight at some date in the earlier seventeenth century. However, if we are to understand the nature of change, we want to known as much as possible about its actuation, implementation and diffusion; therefore, we want to explore the early stages if possible (I shall discuss this further in later chapters). But the limitations of historical databases often make this difficult. Because of these limitations, therefore, it seems that our understanding of the nature of linguistic change will ultimately depend not mainly on historical data, but (recalling Principle 1) on our ability to observe it systematically at the present day in social contexts of use, because that is where we can most readily locate change in progress in a specifiable social context.

This in turn makes it possible to project backwards. By using the insights we are able to derive from the much richer data of present-day researches we should be able to understand more fully what happened in the history of the language, and I shall return to this in later chapters. However, the main point here is that in order to observe in a detailed way the contexts in which linguistic change takes place we need to focus on present-day data. Accordingly, in the following chapters, my aim is to build up a theoretical approach to the social origins of linguistic change by focusing on present-day data. In chapters 3 and 4, we focus on analyzing and interpreting patterns of variation in the speech community.