The rôle of history

Internal and external evidence

Any linguist asked to provide candidate items for inclusion in a list of the slipperiest and most variably definable twentieth-century linguistic terms, would probably be able to supply several without much prompting. Often the lists would overlap (simplicity and naturalness would be reason able prospects), but we would each have our own idiosyncratic selection. My own nominees are internal and external evidence.

In twentieth-century linguistics, types of data and of argument have moved around from one of these categories to the other relatively freely: but we can identify a general tendency for more and more types of evidence to be labelled external, a label to be translated `subordinate to internal evidence' or, in many cases, `safe to ignore'. Thus, Labov (1978) quotes KuryƗowicz as arguing that historical linguistics should concern itself only with the linguistic system before and after a change, paying no attention to such peripheral concerns as dialect geography, phonetics, sociolinguistics, and psycholinguistics. Furthermore, in much Standard Generative Phonology, historical evidence finds itself externalized (along with `performance factors' such as speech errors and dialect variation), making distribution and alternation, frequently determined by introspection, the sole constituents of internal evidence, and thus virtually the sole object of enquiry. In sum, `If we study the various restrictions imposed on linguistics since Saussure, we see more and more data being excluded in a passionate concern for what linguistics is not' (Labov 1978: 275±6).

Labov accepts that `recent linguistics has been dominated by the drive for an autonomous discipline based on purely internal argument', but does not consider this a particularly fruitful development, arguing that `the most notorious mysteries of linguistic change remain untouched by such abstract operations and become even more obscure' (1978: 277). He consequently pleads for a rapprochement of synchronic and diachronic study, showing that advances in phonetics and sociolinguistics, which have illuminated many aspects of change in progress, can equally explain completed changes, provided that we accept the uniformitarian principle: `that is, the forces which operated to produce the historical record are the same as those which can be seen operating today' (Labov 1978: 281). An alliance of phonetics, sociolinguistics, dialectology and formal model-building with historical linguistics is, in Labov's view, the most promising way towards understanding the linguistic past. We must first understand the present as fully as possible: `only when we are thoroughly at home in that everyday world, can we expect to be at home in the past' (1978: 308).

Labov is not, of course, alone in his conviction that the present can inform us about the past. His own approach can be traced to Weinreich, Labov and Herzog's (1968: 100) emphasis on `orderly heterogeneity' in language, a reaction to over-idealization of the synchronic system and the exclusion of crucial variation data. However, integration of the synchronic and diachronic approaches was also a desideratum of Prague School linguistics, as expressed notably by Vachek (1966, 1976, 1983). Vachek uses the term `external evidence' (1972) to refer solely to the rôle of language contact and sociocultural factors in language change; this work has informed and influenced both contact linguistics and Labovian sociolinguistics. Although Vachek accepts external causation of certain changes, however, he still regards the strongest explanations as internal, involving the language's own structure. This leads to attempts to limit external explanation, often via circular and ultimately unfalsifiable statements like Vachek's contention (1972: 222) that ‘a language system ... does not submit to such external influence as would be incompatible with its structural needs and wants'. For a critique of the internal/external dichotomy in this context, see Dorian (1993), and Farrar (1996).

More relevant to our discussion here is Vachek's argument that synchrony is never truly static: ‘any language system has, besides its solid central core, its periphery, which need not be in complete accordance with the laws and tendencies governing its central core’ (1966: 27). Peripheral elements are those entering or leaving the system, and it is vital that they should be identified, as they can illuminate trends and changes in the system which would not otherwise be explicable, or even observable. Peripheral phonemes, for instance, might be those perceived as foreign; or have a low functional yield; or be distributionally restricted, like English /h/ or /ŋ/ (Vachek 1976: 178). A dynamic approach is therefore essential: the synchronically peripheral status of certain elements allows us to understand and perhaps predict diachronic developments, while the changes which have produced this peripherality can in turn explain irregularities in the synchronic pattern. This is not to say that Vachek collapses the two; on the contrary, his review of Chomsky and Halle (1968) is particularly critical of ‘the lack of a clear dividing line that should be drawn between synchrony and diachrony' (1976: 307). Vachek considers Chomsky and Halle's extension of the Vowel Shift Rule from peripheral, learned forms like serene ~ serenity, to non-alternating, core forms like meal, an unjustified confusion of synchrony and diachrony: by in effect equating sound changes and synchronic phonological rules, Standard Generative Phonology in practice significantly reduces the useful conclusions which can be drawn about either.

Although Vachek seems to regard synchronic and diachronic data and analysis as mutually informing, the relationship is seen rather differently in Bailey's time-based or developmental linguistics. Bailey (1982: 154) agrees that `any step towards getting rid of the compartmentalization of linguistics into disparate and incompatible synchronic, diachronic, and comparative or dialectal pursuits must ... be welcomed', and proposes polylectal systems sensitive to diachronic data. He coins the term ‘yroëth' (which is theory spelled backwards) for `something claiming to be a theory which may have a notation and terminology but fails to achieve any deep-level explanation ... All synchronic-idiolectal analysis is yroëthian, since deep explanation and prediction are possible only by investigating and understanding how structures and other phenomena have developed into what they have become' (Bailey 1996: 378). It is therefore scarcely surprising that Bailey regards the influence of diachronic on synchronic analysis as one-way, arguing that historical linguists are fundamentally misguided in adopting synchronic frame works and notions for diachronic work: in doing so, they are guilty of analyzing out the variation and dynamism central to language change by following the ‘nausea principle': `if movement makes the mandarins seasick, tie up the ship and pretend it is part of the pier and is not meant to sail anywhere' (Bailey 1982: 152).

We therefore have four twentieth-century viewpoints. The standard line of argumentation focuses on synchrony; historical evidence here is external, and is usable only as in Chomsky and Halle (1968), where sound changes appear minimally recast as synchronic phonological rules.

Vachek, conversely, argues that synchronic and diachronic phonology are equally valid and equally necessary for explanation. Labov argues that the present can tell us about the past, and Bailey the reverse. My own view is closest to Vachek's: if we are really to integrate synchrony and diachrony, the connection should cut both ways. That is, the linguistic past should be able to help us understand and model the linguistic present: since historical changes have repercussions on systems, an analysis of a synchronic system might sometimes benefit from a knowledge of its development. Perplexing synchronic phenomena might even become transparent in the light of history. But in addition, a framework originally intended for synchronic analysis will be more credible if it can provide enlightening accounts of sound change, and crucially model the transition from sound change to phonological rule without simply collapsing the two categories.

We intend as a contribution to the debate on the types of evidence which are relevant in the formulation and testing of phonological models, and has as one of its aims the discussion and eventual rehabilitation of external evidence. There will be particular emphasis on historical data and arguments; but issues of variation, which recent sociolinguistic work has confirmed as a prerequisite for many changes (Milroy and Milroy 1985; Milroy 1992), will also figure, and some attention will also be devoted to the phonetic motivation for sound changes and phonological rules.

However, although these arguments are of general relevance to phonologists, they are addressed here specifically from the perspective of one phonological model, namely Lexical Phonology. In short, we also constitutes an attempt to constrain the theory of Lexical Phonology, and to demonstrate that the resulting model can provide an illuminating analysis of problematic aspects of the synchronic phonology of Modern English, as well as being consistent with external evidence from a number of areas, including diachronic developments and dialect differences. I shall focus on three areas of the phonology in which the unenviable legacy of Standard Generative Phonology, as enshrined in Chomsky and Halle (1968; henceforth SPE) seriously compromises the validity of its successor, Lexical Phonology: these are the synchronic problem of abstractness; the differentiation of dialects; and the relationship of sound changes and phonological rules. I shall show that a rigorous application of the principles and constraints inherent in Lexical Phonology permits an enlightening account of these areas, and a demonstration that generative models need not necessarily be subject to the failings and infelicities of their predecessor. Finally, just as the data discussed here are drawn from the synchronic and diachronic domains, so the constraints operative in Lexical Phonology will be shown to have both synchronic and diachronic dimensions and consequences.