|Title:||Yoruba vowel patterns: deriving asymmetries by the tension between opposing constraints|
|Abstract:||Yoruba vowel patterns:
deriving asymmetries by the tension between opposing constraints
University of British Columbia
This paper re-examines a class of pervasive asymmetric properties in the
vowel system of Yoruba. Pulleyblank (1988) argues that high vowels in
general and [i] in particular exhibit a special array of properties. For
example, certain rules of assimilation and deletion are triggered by all
vowels except high vowels. Other rules of assimilation target only high
vowels. In loan vocabulary, [i] is the epenthetic vowel in non-
assimilatory contexts. In Yoruba, as in numerous other cases in the
literature, the recurrence of a particular feature as a conditioning or
nonconditioning factor in various processes can be interpreted as an
argument in favour of underspecified representations (Pulleyblank 1988).
If [-high], [+back], etc. are structurally present in Yoruba vowels, but
[+high] and [-back] are not, then the cited facts receive an explanation.
When assimilation fails to be triggered by [i], this is due to the
absence of features specified for that vowel. When only [i] is targeted
by a rule, this is because the rule is structure-building, not
structure-changing. The epenthesis of [i] is analysed as syllabically-
motivated insertion of a featureless prosodic vowel.
While resolving a number of problems, an underspecified approach raises
others. Clements & Sonaiya (1990) argue that a rule of L Nasalisation
makes crucial reference to [+high], as does a Morpheme Structure
Constraint; Akinlabi (1993) makes a similar argument for a rule of
r-deletion. Such patterns that reference [+high] are incompatible with
an underspecified analysis of the Yoruba vowel system. Note in
particular that certain rules requiring reference to [+high] apply very
early (e.g. as a morpheme structure condition) while certain rules
behaving asymmetrically apply very late (e.g. a rule of vowel deletion
that crosses word boundaries). The apparent conflict in behaviour cannot
be resolved, therefore, by assuming an early stage of underspecification
followed by a subsequent stage of more complete specification.
Of particular interest to the treatment of asymmetric phonological
behaviour, both the paper by Clements & Sonaiya and the paper by
Akinlabi reinforce the asymmetric nature of high vowels, each presenting
additional data of a special asymmetric nature. Hence even though the
underspecification approach appears to fail, if we accept their
arguments then it becomes even more important therefore to find a
theoretical account that can satisfactorily account for the asymmetries
in the Yoruba vowel system.
This paper discusses three possible analyses for Yoruba. First, it lays
out the possibility that there are no special asymmetries at all, that
the particular properties of any individual construction are due to
idiosyncratic properties of individual rules. Second it examines the
possibility that asymmetries are significant and due to a structural
property such as underspecification. Third, it explores the possibility
that asymmetric behaviour results from the relative ranking of
constraints. That is, the special properties of high vowels result not
from structural encoding, but from the ranking of constraints on high
vowels above and below constraints implicated in particular
constructions. It is shown that both rule-governed and underspecified
accounts require special stipulations of an ad hoc nature as additional
constructions are considered, where the constraint-based approach does
not. The general conclusion is that a constraint-based approach resolves
the problems raised by the analyses of Clements & Sonaiya and Akinlabi,
as well as other problems intrinsic to the proposal of underspecification.
The paper therefore constitutes an argument in favour of the encoding of
asymmetries in a constraint grammar rather than in elaborated structures.
In developing the treatment of asymmetric behaviour, harmonically ranked
sonority constraints play a central role. The paper argues that such
reference to sonority must be encoded directly into the expression of
the faithfulness constraints, rather than through an independent set of
well-formedness conditions. That is, the paper argues for the encoding
of "sonority as faithfulness".