ON THE REVERSIBILITY OF MERGERS: /W/, /V/ AND EVIDENCE FROM LESSER-KNOWN ENGLISHES

Conventional linguistic wisdom has it that mergers cannot be reversed: "it is generally agreed that mergers are irreversible: once a merger, always a merger" (Labov 1994: 311). The reason for this is clear: once two phonemes have converged, Speakers have no way of knowing which one of the two original units belongs in which one of the two original lexical sets, and restoration is impossible. As is well known, however, there are a number of reports in the historical linguistics literature of phonological mergers which have been reversed. One often quoted example is that of the merger in English of the lexical sets of MATE and MEAT, which is well-attested from earlier periods of the language, but which is not found in any modern variety of English. This "once a merger, always a merger" maxim has quite naturally led historical linguists to consider how to explain these reports of mergers which have been reversed. In earlier work on this topic, historical linguists (e.g. Kökeritz 1953) typically employed explanations for this puzzling phenomenon which were based on dialect contact. They agreed that mergers could not be reversed äs such, but their thesis was that while, say, MATE and MEAT were indeed genuinely merged in some dialects, the merger was later undone äs a result of contact between Speakers of these dialects and Speakers of other dialects where it had not occurred. That is, Speakers were able to accurately repair the merger by Consulting the distribution of vowels over lexical sets in the speech of Speakers of the non-merging dialects. Wyld (1956: 210) writes that we have to assume that the MATE and MEAT part of the English vowel System was "differentiated among different classes of Speakers whether in a Regional or a Class dialect I am unable at present to say into two types", and that the unmerger was not a sound change äs such but "merely the result of the abandonment of one type of pronunciation and the adoption of another" (1956: 211). More recently, a brilliant and pioneering alternative explanation has been advanced by Labov. This is that these mergers were never actually mergers at all but rather "near-mergers". That is, they may have been perceived and speit and reported äs mergers because of a very close phonetic proximity between the two phonemes concerned. Labov (1994: 349-370) discusses this issue at considerable length. He cites several instances of Speakers being able to produce a very small phonetic distinction