Background: Speech is a highly redundant signal: any given feature is communicated across multiple acoustic cues. Initial consonant voicing, for example, is conveyed by voice onset time and the F0 of the subsequent vowel (among other cues), while word emphasis is conveyed by increases in pitch, amplitude, and duration. This redundancy makes speech robust, with comprehension possible even when the signal is distorted. However, information being spread across different cues poses a problem for listeners: how to make sure that the most important source of information is prioritized. Attention-to-dimension models suggest that selective attention is directed towards highly weighted cues and away from less useful cues. One way to test this hypothesis is to investigate whether relative cue weighting changes depending on whether target speech is presented in quiet or during informational masking. We tested this hypothesis by examining categorization of consonant voicing and word emphasis in quiet and in the presence of a single distracting talker.

Methods: Participants were presented with target speech to be categorized in one ear while either silence or a distracting talker was presented to the opposite ear. Stimuli were sampled from two-dimensional stimulus grids in which pitch and duration varied independently in the extent to which they conveyed whether a word was emphasized (Experiment 1) or whether a consonant was voiced or unvoiced (Experiment 2). Mixed-effects logistic regression was used to measure the extent to which pitch versus duration cues were weighted across individuals and conditions.

Results: Across both voicing and word emphasis categorization, individuals who tended to weight a single cue highly in quiet switched strategies under informational masking, moving towards integration across cues. Individuals who already used an integrative strategy did not, however, change their weighting because of distraction.

Conclusions: We find that perceptual strategies can, in some individuals, differ between quiet and distracting listening environments. Listeners seem to have difficulty solely relying upon a single primary cue during distraction, which supports the hypothesis that cue weighting draws upon selective attention. Moreover, these results suggest that perceptual strategies measured in quiet are not necessarily indicative of the ways in which listeners sample information across acoustic channels in more ecologically valid, complex listening situations.