Within-Category Mismatch Responses to Parametrically Varying Speech Sounds

Mismatch negativity arises from the construction of a memory trace encoding regularities extracted from auditory stimuli in an oddball paradigm. A previous study argued that when phonetic standards are varied within the limits of a phoneme category (i.e., the mental representation of speech sounds), the auditory cortex retrieves a discrete phoneme representation from long-term memory and uses it as the memory trace for deviance detection. This claim has motivated a body of work using varying-standards paradigms to probe phonological underspecification. A strong version of this claim predicts that varying standards should eliminate mismatch negativity when both standards and deviants belong to the same phoneme category, because the memory trace is purely phonemic and lacks acoustic detail. We tested this prediction in three electroencephalography experiments using English /t/ varying in voice onset time. In Experiment 1, a within-category deviant with a long voice onset time elicited a robust mismatch negativity even when standards varied across shorter values, contrary to the prediction of an exclusively symbolic phoneme-based memory trace. Experiments 2 and 3 further examined whether the within-category mismatch response reflects long-term memory representations of phonetic realizations associated with the evoked phoneme category or listeners' ad hoc representation of the stimulus distribution. The weight of the evidence suggests that the within-category mismatch negativity in the varying-standards paradigm reflects sensitivity to the statistical structure of the stimulus physical properties, rather than exclusive access to abstract phonological categories. Findings invite reinterpretation of studies using varying-standards paradigm in light of encoding of stimulus statistics.