Post-Saussurean Linguistics Guy specific. In recent work, I have given an autosegmental account of -t,d deletion that explains certain aspects of the process in terms of resyllabification of the final stops as onsets of the following syllables (as has transparently occurred in phrases like band-aid). This analysis made the prediction that following /11 and /r/ would have different effects on the deletion rate, because resyllabified /tr-, dr-1 onsets are possible in English, but language specifically, /tl-, dl-/ onsets are not. This prediction has been strongly confirmed in a series of recent studies; Table 6 shows the figures from Guy (1991). Hence, once again, we would conclude that the process is constrained by a specific feature of English phonology, not some articulatory universal. -------------- Following Segment ---------------- Obstruent !II Glide /r/ Vowel Rate of deletion .66 .80 .57 .42 .19 (Varbrul2 factor weights) Table 6. Following segment effect on -t,d deletion contrasting /11 and /r/ (from Guy 1991). Performance grammar. Thus, the attempts to remove systematic variability from the grammar by attributing it to nebulous production universals has failed. But other approaches have risen in its stead. A second possible argument that has sometimes been advanced in order to keep the grammar invariant is to attribute the regularities of variability to a hypothesized separate (but so-far, uninvestigated) system governing production, a kind of grammar of performance. This could have its own language-specific characteristics, thus circumventing the above failings of universalist explanations. However, this approach is itself suspect because the units and explanatory principles required in such a performance grammar keep looking eerily like the units and principles required to account for categorical phenomena. We have already seen an example that illustrates this point. If the avoidance of *tl- and *dl- sequences that is evidenced for a variable process in Table 6 is due to a separate performance constraint on English and is NOT related to the categorical constraint against such onsets that appears in the competence grammar of English, this would be a very surprising coincidence indeed. Another example appears in the analysis of preceding context effects on -t,d deletion. Deletion rates have long been known to depend in part on place and manner features of the preceding context: for example, English speakers always delete more after Is! (e.g. mist, west) than after /11 (e.g. melt, old). Recent work by Charles Boberg and myself shows that this effect can be reduced to a simple similarity function between the preceding segment and the target coronal stop, involving the features [cor], [cont], [son], (and also [voice]). Figures from Guy and Boberg (1994) are given in Table 7. When the target shares like values of these features with the trigger, deletion is favored, but when target and trigger disagree for some feature value, deletion is disfavored. Thus higher rates of deletion in a word like west arise because the preceding Is/ shares the feature values [+cor, -son] with the target It!, while lower rates of deletion in melt arise because the /IJ shares only the feature [+cor]. This, any phonologist will note, looks like an OCP effect. The Obligatory Contour Principle in current tht;.ory is postulated to prohibit same-tier sequences of adjacent identical autosegments, segments, or even features. It is the constraint that explains facts like the English prohibition of geminates, which is why final -t,d are never preceded by another -t or -d (when this would arise thru affixation, for example, epenthesis occurs to break up the
Posted on: Wed, 30 Oct 2013 14:07:48 +0000
Trending Topics
Recently Viewed Topics
© 2015