Presentation: Hsiang-Yu
Summary: Sarah
This study testified the functional load hypothesis
with corpus data from nine language varieties. It is suggested by the
hypothesis that phoneme pairs are less likely to undergo merging if they have a
heavier function load, i.e., they are used to distinguish many words in the
language. In this study, specific predicting factors identified were the number
of minimal pairs that a particular phoneme pair distinguishes, phoneme
probability, and the entropy changes at the phonemic and word levels. Logistic
regression analyses revealed that the number of minimal pairs is a powerful
predictor, in which the probability of merger occurrence decreases as the
phoneme pair distinguishes more words in the language. Phoneme probability is
also effective, with more frequent phonemes being more prone to merging
phenomena. The effect, nonetheless, is limited to phoneme pairs that do not have
any minimal pairs. With respect to the two entropy measures, the effects are
not very robust, since they are highly correlated with the former two factors. Taken
together, the results of this study serve as strong evidence for supporting the
functional load hypothesis.