The thesis presented here is a study of the acquisition of phonemic
constraints, and of the rates at which the constraints are upheld, as demonstrated
in speech eJTors. These experiments were motivated by Dell et al. (2000), who
tested whether phonotactic constraints could be learnt through recent linguistic
experience (i.e. reading nonsense syllables). The results of these studies showed
that the phonotactics present in the stimuli were followed, and that the speech
production system is sensitive to recent experience. The experiments also
showed that the syllable position constraint was not upheld as strongly as
expected.
The present study began by replicating these experiments. The general
findings of the Dell et al. (2000) study were supported, but other aspects of the
study were not. Weaknesses were found in the statistical analyses provided by
Dell and colleagues, and the syllable position constraint was upheld at a lower
rate than that found in the Dell et al. study. Dell et al. provide no data regarding
either the rate of acquisition, or the durability, of the learned constraints.
The next set of experiments explored the time-course and durability of the
learning. This was done by adding a further section to the experiment that
reversed the previously learned constraints, and analysing the speech errors for
signs of "confusion" (i.e. continuing to use the previous constraints). These
experiments showed that there was a period following the reversal of the
constraints in which participants followed the previous constraints.
The final part of this study modified the paradigm for the auditory
modality. The stimuli were played to the participants through headphones, and
the participants repeated the syllables that they heard. This resulted in an
increase in the en-or corpus, while still producing the same results as the visual
version of the paradigm. The results of these expe1iments are considered in
relation to models of speech production, and the implications for these models
are discussed.