Final consonant devoicing

  Final consonant devoicing applies to the final tail of roots which appear either uninflected or with inflectional suffixes which do not begin with a vowel. Wiese (1996a, 201-3) has suggested that this amounts to devoicing of all syllable-final consonants, since the addition of a vowel-initial suffix results in the tail of the root becoming an onset of the following syllable (as required by the maximal onset principle). In the case of root-internal syllable-final consonants, these are all unvoiced, but they are invariant, since there cannot be any situation in which they become syllable initial. Therefore the only actual alternation between voiced and voiceless consonants appears root-finally. Like Kloeke (1982, 30-32), we take the view that the lexical representation of root-internal syllable-final consonants always specifies the voiceless segment, since this is the only one which can ever appear. Kloeke actually argues that the term ``final consonant devoicing'' is inaccurate because the feature in question is tenseness, rather than voice, but this is immaterial in an account which is formulated segmentally rather than featurally. The lexical representation of roots where there is an alternation specifies the voiced counterpart, thereby distinguishing such roots from those which have invariant voiceless consonants root-finally (e.g., to distinguish between Rat and Rad). The devoiced variants are then determined by checking the suffix of the form in question. If there is a suffix which begins with a vowel, then no devoicing takes place. If there is no suffix, or the suffix begins with a consonant then devoicing does take place. This is defined by the following equation (which forms part of our definition of the Word node):

Word:
    <phn syl1 tail> == IF:   <VOWEL:<"<mor suffix>">
                       THEN          "<phn syl1 tail->"
                       ELSE Devoice:<"<phn syl1 tail->">>.

Here the path-initial attribute phn in <phn syl1 tail> contrasts with the path-initial mor in <mor suffix> (see the section on morphology). These path-initial attributes (along with syn and sem) serve to partition the feature space into phonological and morphological domains, respectively. Note also the distinction between <tail-> and <tail> attributes where the former corresponds to the `underlying' segment and the latter to its surface realization.

The Devoice function itself simply maps voiced stops and fricatives to their voiceless counterparts and maps all other consonants to themselves:

Devoice:
    <b> == p
    <d> == t
    <g> == k
    <v> == f
    <z> == s
    <>  == IDEM.

Note that final consonant devoicing is simply irrelevant to adjective, determiner and third person pronoun inflection (since all the suffixes begin with a vowel) and we will thus ignore it in our discussion of these (likewise the tail/tail- distinction).