More opaque feeding

(This is kind of a follow-up on this post from last year.)

Speaking of McCarthy’s GLOW talk (related handout here): I got interested in an example of feeding discussed there. Ignoring melodic details, the two rules are:

Vowel Epenthesis: Ø –> V / #__CC
Consonant Epenthesis: Ø –> C / #__V

Interestingly, the generalization expressed by Vowel Epenthesis is rendered non-surface-apparent (opaque) by Consonant Epenthesis: the consonant cluster is no longer word-initial. In other words, Vowel Epenthesis appears to overapply. This is another example of an opaque feeding interaction, except in this case the right interaction is easily statable in OT (which is particularly interesting given that non-surface-apparent opacity is generally problematic for OT).

Is there some other way to state the rules that avoids this opaque feeding interaction? I can think of three basic alternatives.

  1. Stating the context of Vowel Epenthesis as #(C)__CC. This recapitulates the context of Consonant Epenthesis, though this may be independently necessary for underlying #CCC clusters. (Stating the context of Vowel Epenthesis as {#,C}__CC runs into the problem of what to do in the potential underlying #CCC case; assuming C__CC takes priority over #__CC, then we’re just back to #(C)__CC.) At least in the absence of evidence from underlying #CCC, then, this kind of move only serves to make the interaction between the two rules transparent — but then you need feeding anyway, which makes the optional C in the context of Vowel Epenthesis unnecessary!

  2. Relying on syllable structure and persistent syllabification doesn’t really resolve the issue, because Vowel Epenthesis would simply be rendered non-surface-apparent by persistent syllabification. First, you identify that the word-initial C is unsyllabifiable, then you epenthesize the vowel before unsyllabified consonants, then you resyllabify and the C is no longer unsyllabified — the context for epenthesis is gone.

    As McCarthy notes (p.c.):

    “Repair” rules are inherently opaque, since the whole point of a repair is to eliminate the conditions that made the repair necessary!

  3. An Ito-style approach in which epenthesis is a by-product of syllabification, I think, just gets rid of the whole issue of ordering between the two epentheses in the first place: the word-initial C is associated with the coda of the syllabic template, the nucleus and onset of which are (simultaneously) filled by epenthetic segments.

    As McCarthy notes (p.c.):

    But then you’re at the cusp of OT — or, at least, you’re doing phonology on the basis of constraints on the well-formedness of outputs and essentially putting rules aside.

It seems that, like non-surface-true (underapplication) opacity, examples of non-surface-apparent (overapplication) opacity can be divided into two classes (from the perspective of OT): those in which the relevant faithfulness violation is genuinely gratuitous on the surface (counterbleeding, opaque feeding of the Turkish/Tagalog type discussed here) vs. those in which the relevant faithfulness violation(s) is/are motivated by the need to satisfy two undominated markedness constraints (counterfactual derivation, opaque feeding of the repair type above). The latter class is perfectly amenable to a classic OT analysis, but of these only opaque feeding is amenable to a rule-ordering analysis.

Leave a Reply

Your email address will not be published.