The role of phonological features in the internal representation of American English consonants: Evidence from behavioral, neuroimaging, and computational modeling approaches

The primary aim of this study was to obtain evidence regarding the structure and composition of the internal representations of phonemes through an investigation of feature processing in consonant perception. An interdisciplinary approach combining phonetic and psycholinguistic experimentation, neuroimaging, and computational modeling was used to this end. The experimental results reveal differences in processing within and among voicing, place, and manner features, implying unequal contributions of these features to the phoneme representations. Results from gating and priming experiments indicate that manner is the most prominent feature in the classification of these eight consonants, and results from all three experiments suggest differences between the processing of place and voicing features. Voicing information was less influential than place information in the gating and priming experiments; there were more voicing confusions than place confusions in the gating experiments; and the voicing decision was correlated with greater and more widespread cortical activation compared to the place decision in the fMRI study. Moreover, the gating and priming studies demonstrated processing asymmetries within place and manner features, with alveolars and stops having a greater effect on processing than labials and fricatives, respectively. Modeling the behavioral and imaging results yielded an existence proof illustrating how the various findings fit coherently within a plausible cognitive architecture. Theoretical implications of the results and simulations are discussed in relation to current phonological and psycholinguistic theories of speech processing.