Phonetic cue weighting as auditory feature-based selective attention

Listeners use multiple acoustic features, to different extents in different circumstances, in perceiving variable speech signals as streams of phonemes. One class of models conceptualizes this phonetic cue weighting as allocating different amounts of attention to different acoustic features (i.e., attentional theories). These theories posit or assume that similar neural mechanisms underlie phonetic cue weighting and a more general form of auditory feature-based selective attention. We describe multiple attentional theories, consider their relationship to other models of phoneme perception, and review related empirical evidence. In general, attentional theories are able to explain both training-induced and context-dependent cue reweighting. The neuroscientific literature on auditory feature-based selective attention suggests potential neural mechanisms of phonetic cue weighting, including at the single-neuron level. At least three challenges need to be met in order to ascribe all of cue weighting to auditory feature-based selective attention: (1) the complexity of continuous speech, (2) individual differences in cue weighting and selective attention abilities, and (3) any need for specialized neural systems in allocating attention to phonetic cues. To date, using attentional theories to understand speech perception has been fruitful, but more detailed and mechanistic models of feature-based selective attention are needed in order to define the extent to which the attention literature applies in full to cue weighting in speech perception.