264TOBIAS SCHEER & PHILIPPE SEGERALsonority slope or phonetic properties tells us nothing at all about the syllabic identity of TR clusters. Muta cum liquida is a plastic syllabic object, and a reasonable means to uncover the identity of individual TR clusters is to look at them through the positional prism: the interpretation of their behaviour in terms of a theory of lenition and fortition may be enlightening. Therefore, we argue, phonological theory must not feature any predetermining device that favours some particular syllabification of TR clusters. Even if it is true (and it may well not if one puts down the Indo-European glasses) that there are more TR clusters in this world which instantiate branching Onsets than there are muta cum liquida with Coda-Onset or affricate status, this statistical information is irrelevant. We follow Newmeyer (1998) on this issue, who recalls that linguists committed to the concept of competence define what a possible, not what a frequent grammar is. Therefore, the traditional devices Onset Maximisation, ambisyllabicity and Coda Capture, in whatever modern coat, are ill-inspired: they produce a grammar with in-wired branching Onsets. Turning now to the general picture of Gallo-Romance diachronics, we believe that the foregoing discussion can reintegrate a number of scattered phenomena that are usually thought of as unrelated and marginal into an overall scenario where they appear as an instance of the central regularity that governs Gallo-Romance: the action of positional (syllabic) forces that cause lenition and fortition. Beyond what could be treated in the present article, the Gallo-Romance situation described raises an interesting question for phonological theory: if it is true that muta cum liquida can appear in different coats in the same synchronic state of a language (a hypothesis that is fed by our analysis), is their distribution random, or does it obey certain rules We are reluctant to accept that the syllabic status of TR clusters is an idiosyncratic and unpredictable property of each lexical item. If, then, the various types of TR clusters like to appear in certain positions, the data reviewed give quite clear indication as to how they line up: all cases of affricate muta cum liquida discussed occur in the Coda Mirror, i.e. word-initially or after Codas. Hence they seem to be uneasy in weak positions (i.e. intervocalically and in Codas). It may be speculated, then, that homosyllabic TR clusters (1a) rather elect home in intervocalic position. And, of course, heterosyllabic muta cum liquida (or rather, its first member) is the natural candidate for Codas. Further study must run these hypotheses against a greater empirical record. ReferencesAli, Fatima. 2003. es positions fortes et faibles des consonnes en malgache� DEA thesis, Universit de Nice. Allires, Jacques. 2001. Manuel de linguistique romane. Paris: Champion. Bertinetto, Pier Marco. 1981. Strutture prosodiche delltaliano. Firenze: Accademia della Crusca. Bourciez, Edouard. 1930. Elments de linguistique romane. Paris: Klincksieck. Bourciez, Edouard & J. Bourciez. 1967. Phontique franaise. 9e edition. Paris: Klincksieck.
