Neural responses to target words are modulated by semantic and acoustic similarity of temporally offset distractors

Natural environments require listeners to attend to one target stream while ignoring irrelevant, co-occurring sources. Past studies show voice differences between streams facilitate attention, but typically employ similar semantic content across streams, challenging attentional selection. Here, we examined how two markers of selective attention, content recall accuracy and neural response amplitude, vary with distractor characteristics. We recorded electroencephalography while participants (preliminary N=17) reported the content of a target stream, ignoring a spatially separated, temporally interleaved distractor. The target stream always consisted of syllables from a male voice. A within-subject 2x2 design manipulated the semantic content (syllables same / digits different) and voice (male same / female different) of the non-target stream. Voice differences facilitated target recall only when semantic content was similar across streams; when listeners could leverage semantic differences between streams, voice differences did not facilitate auditory attention. Event-related potentials to (acoustically identical) target syllables differed by distractor condition: Both manipulations affected P1 and N1 amplitudes, resulting in large peak-to-peak differences when distractors were dissimilar to the target and small differences when distractors closely resembled the target. Overall, when selection was easy, voice differences did not influence behavioral recall, but nonetheless influenced neural coding of target syllables.