| dc.description.abstract | The relationship between phonetic and phonological information within phonological
theory is investigated with specific reference to the phonological feature [voice]. Given
the often algebraic and therefore arbitrary nature of constructs employed by phonological
theory, important generalisations are missed. This thesis proposes a version of
constraint-based phonology that integrates aspects of the representation devices of
cognitive grammar with the notion of a constraint as conceived in optimality theory. It is
shown that an adequate account of Afrikaans voicing phenomena can be formulated on
this basis.
The phonological feature [voice], alongside the related notions [spread] and [tense]. is
defined in phonetic terms. In chapter 2, two kinds of models for representing the
relationship between these features are identified in the literature: one-way models that
only recognise one of the three features, either [voice] or (tense); and two-way models
that recognise [voice], together with either [spread] or (tense]. Two further models are
proposed as hypotheses about feature representation: a three-way model that assigns
superordinate status to (tense] and basic level status to [voice] and [spread]; and a model
with two independent systems, a (tense]-system and a [voice]-[spread] system. These
models are subjected to evaluation in the subsequent chapters of the thesis. It is
concluded that the three-way model accounts most adequately for the voicing phenomena
in Afrikaans, as well as various other languages, particularly Dutch.
A few voicing alternations are known to affect the feature [voice]: regressive, voicing
assimilation, final devoicing and progressive devoicing. At least three types of accounts
are available for most of them: phonetic, generative phonological and optimality theoretic
accounts. These three types of accounts are examined separately in chapters 3-5.
identifying the insights they contribute, as well as their shortcomings.
On the basis of these comparisons it is concluded in chapter 6 that a unified account for
all the relevant insights is not available. Although optimality performs better than
generativism in terms of t11e comprehensiveness of its account, a number of obvious
phonetic findings cannot be accommodated within optimality theory. The notion of
constraints in optimality is identified as an important contribution, but constraint ranking
is found to be inadequate because of its essentially arbitrary character. Various phonetic
findings indicate that constraint ranking should be predictable on external grounds.
As an alternative. it is proposed in chapter 7 that schematic networks, which serve as
licensing schemas for the distribution of distinctive features, should be adopted. These
networks employ the notions of prototype and category extension, as developed within
cognitive grammar, alongside constraints. The constraints are regarded as the primary
phonetic devices for category extension. The distance between a par1icular extension and
its prototype is identified as a predictor of the degree of gradient behaviour of
phonological units. This model is applied to aspects of the Dutch voicing phenomena
discussed in chapters 2-6.
Such an account makes it possible to incorporate phonetic findings that voicing
phenomena are both optional and incomplete. It makes possible a statement of the kinds
of optional behaviour, as well as the reasons why incompleteness effects occur. In
chapter 8, a schematic network that licenses the occurrence of the feature [voice) in the
phonology of Afrikaans is proposed, together with a network that explicates the effects of
devoicing in Afrikaans, in syllable-final and morpheme-internal environments, as well as
across morpheme boundaries that are not particularly salient. While most of the networks
operate with the feature [voice], the impact of the superordinate category [tense] on the
existence of incompleteness effects is identified. A separate network for the feature
[tense] is proposed for Afrikaans. This network is employed with the function of
disambiguation in the case of phonetic incompleteness effects during final devoicing, and
also lies at the basis of morphophonological alternations and distributions among
Afrikaans fricatives. | |
