Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Early neural sensitivity to voice and semantic content shapes auditory selective attention
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Sahil Luthra1, Eric Parker2, Marysia Brown3, Gidey Gezae4, Hee So Kim2, Wusheng Liang2, Abigail Noyce2, Barbara G. Shinn-Cunningham2; 1Stony Brook University, 2Carnegie Mellon University, 3University of Pittsburgh, 4The Pennsylvania State University
Real-world speech comprehension settings require listeners to home in on a source of interest while ignoring simultaneous conversations. Selective attention modulates neural activity across large swaths of cortex, enabling listeners to prioritize key inputs for processing. Behavioral studies have shown that both low-level acoustic differences (e.g., voice pitch) and more abstract linguistic differences (e.g., semantic content) support selective attention. Further, electroencephalography (EEG) studies have demonstrated that pitch differences between sources influence the neural encoding of target stimuli (and specifically, the P1, N1, and P2 components of the auditory event-related response, or ERP). However, it is unclear whether semantic differences between streams similarly support early filtering. On the one hand, voice and linguistic cues appear to influence different aspects of attentional processing and are processed by disparate (but overlapping) networks; as such, they may have distinct influences on neural encoding. On the other hand, extensive feedback circuitry throughout the brain may allow both features to shape early attentional responses. In this EEG study, normal-hearing young adults (N=30) attended to a lateralized auditory stream of target syllables (ba/da/ga) while ignoring a competing stream from the opposite hemifield (streams spatialized to ±30° azimuth). Across conditions, we factorially manipulated two features of distractor stimuli: voice (same or different) and semantic content (same [ba/da/ga] or different [one/two/three]). Critically, distractors were temporally offset from targets, enabling an investigation of whether acoustically matched, always-attended target stimuli are differentially encoded as a function of surrounding context; across trials, we counterbalanced whether targets preceded distractors or lagged behind them. Following each trial, participants recalled the set of target stimuli. In this way, we obtained both a behavioral measure of selective attention (syllable recall) as well as a temporally sensitive measure of neural encoding (ERPs locked to target onset). For ERP analyses, we focused on the P1-N1-P2 complex. Listeners exhibited high accuracy across conditions (mean: 82%). Either voice or linguistic differences were sufficient to support accurate recall; however, accuracy was significantly worse when both streams were in the same voice and had similar semantic content. Although voice and semantic features had interactive effects on behavior, they had independent, additive effects on neural processing. Neural responses to targets (specifically, P1-to-N1 peak-to-peak differences in the event-related EEG response) were strongest when both features supported attention, weaker when either was removed, and weakest when both were absent. The effect was near-perfectly linear, with no interaction between features and no differences in the time course of the modulation. Overall, results clarify how listeners attempt to filter irrelevant, distracting sounds to support selective attention and ultimately language comprehension: When targets and distractors are similar, attention is less effective, leading to diminished brain responses to target sounds. Results imply that regardless of which feature differentiates targets and distractors, early cortical processing of auditory stimuli is impacted similarly by attention. This is consistent with prior findings implying ubiquitous feedback circuitry throughout sensory processing.
Topic Areas: Control, Selection, and Executive Processes, Speech Perception