A computational model of phonotactic acquisition
Predictability of exceptional patterns in Hungarian
This paper presents a model that connects phonotactic exceptionality to perceptibility, more specifically to functional load and acoustic detail. I identify two patterns in exceptionality: lexical exceptions and phonotactic vacillation, where the former is restricted to specific lexical items, while the latter affects two contrastive sound categories as a whole. Through the example of Hungarian word-final phonotactics, the Model of Perceptual Categorization associates these two patterns with different functional load and acoustic properties of contrasts, that lead to two categorizational malfunctions. On the one hand, phonotactic vacillation is a result of a frequent failure to categorize ambiguous tokens: low functional load coinciding with little acoustic difference. On the other hand, lexical exceptions are systematic categorizational mistakes brought about by salient categories – in this case distributional generalizations are hindered by interference from mislabeled tokens.