English /r/
Introduction
In The rôle of history, I quoted Labov (1978) in defence of my intention to reintegrate synchronic and diachronic evidence. Labov's (1978: 281) view is that, provided we adopt the uniformitarian principle, and therefore accept that `the forces which operated to produce the historical record are the same as those which can be seen operating today', we can use the linguistic present to explain the linguistic past. However, if we are serious about the reintegration of synchrony and diachrony, the connection should work both ways: that is, the linguistic past should ideally also help us understand and model the present.

The first part of the equation has already been proved: in Synchrony, diachrony and Lexical Phonology: the Scottish Vowel Length Rule, I showed that a possible life-cycle for sound changes and phonological rules can be formulated in Lexical Phonology. The default case was represented by æ-Tensing; and a variant pathway, involving two cycles of alteration of the underlying representations, was required for processes like the Scottish Vowel Length Rule, which involve historical rule inversion. A model designed primarily for synchronic phonological description therefore provides insights into change. I hope to show that we can indeed also use the past to explain the phonological present, with special reference to English /r/, which is of particular relevance because it has been discussed in a variety of phonological frameworks (see Broadbent 1991, McCarthy 1991, 1993, Scobbie 1992, Donegan 1993, Harris 1994, Giegerich in press); is characterized by interesting interactions between /r/ itself and preceding vowels; and arguably again involves rule inversion. We shall see that the ostensibly arbitrary synchronic process of [r]-Insertion in varieties with both linking and intrusive [r] is in fact historically principled, and that the synchronic situation has been produced by a series of historical steps, each conditioning the next. Moreover, each of these steps seems still to be preserved in some variety of Present-Day English. This further notion of the interaction of historical change and synchronic variation is familiar from creolistics: Bickerton, for instance, in describing the creole continuum in Guyana, claims that `a synchronic cut across the Guyanese community is indistinguishable from a diachronic cut across a century and a half of linguistic development' (quoted by Romaine 1988: 165). We shall see that this also holds, mutatis mutandis, for the English speech community in its widest sense.

However, it is not enough to explain synchronic patterns with reference to the changes which have created them: we must also account for the contributory changes themselves. To do so, I suggest we require further reference to phonetic parameters, and indeed a wholesale revision of the feature system: to this end, I shall tentatively propose the incorporation of the gestural system of representation used in Articulatory Phonology (Browman and Goldstein 1986, 1989, 1991, 1992; McMahon, Foulkes and Tollfree 1994, McMahon and Foulkes 1995, McMahon1996) into Lexical Phonology.

But this reliance on historical and phonetic evidence does not mean the analysis arrived at is phonology-free: on the contrary, a highly con strained phonological model will again force us to draw certain conclusions, not only about the current status of /r/, but also about its history. Thus, some arguments below will be historical, but others will be theory-internal, depending on particular properties of the model assumed here. The emphasis on formal models distances this approach a little from Labov (1978): while he defines the linguistic present largely in terms of social influences on speakers, and the quantitative analysis of inter- and intra-personal variation, I believe that part of a synchronic phonological analysis must also involve idealization from these data and consequent model building and evaluation. Where such models cast light on the interaction between synchronic phonology, dialect variation and sound change, this may in itself be evidence for that view.