Slide Slam F12
What did they say? Neural evidence for detection of own-name and semantic violations in task-irrelevant speech at a Virtual Café
Adi Brown1, Ksenia Burgart1, Elana Zion Golumbic1; 1Bar Ilan University
Natural environments generate an abundance of sensory inputs and multitude of sounds, which poses an immense challenge for our perceptual system. Dealing with concurrent stimuli is particularly difficult in contexts, such as a noisy café, where many people speak at once. This is because the competition for processing resources is not limited to the perceptual system but extents to the language-system as well. For decades, researchers have tried to understand how the system deals effectively with such competition, and the degree to which listeners can harness attention to ‘filter out’ background speech that is task-irrelevant. Of particular interest is whether semantic aspects of task-irrelevant speech are detected and processed, or whether it is merely represented at the acoustic level. Previous studies point to two specific semantic features that, if present in task-irrelevant speech, may be noticed and potentially capture ones’ attention: (1) personally relevant information, and more specifically hearing ones’ own name, and (2) semantic incongruencies or violations of semantic expectations. Here we tested these hypotheses using a novel experimental approach, simulating the type of naturalistic conditions where this question is ecologically relevant. Using Virtual Reality, participants experienced sitting at a table in a simulated Café across from a conversation partner. Participants listened to their partner’s narrative (target speaker) and subsequently performed a word-recognition task. In the background of the café, a barista character called out a sequence of orders (e.g., “Salad for Dan”; task-irrelevant speaker). Some of the orders were manipulated to include the participant's own name or semantic violations (e.g., “Salad for coffee”). Neural activity was recorded throughout the experiment using electroencephalography (EEG), in addition to continuous eye-tracking and skin conductance measures (GSR). We used the RIDE toolbox (http://cns.hkbu.edu.hk/RIDE.htm) to extract Event-Related Responses (ERPs) from the EEG signal to individual words in the task-irrelevant speech. When comparing responses to own’s own name vs. a control name (name of the previous participant), we found an enlarged negative response around 400ms, followed by an increased positive response around 600ms. This pattern resembles the classic N400 and P600 ERP components, and supports the hypothesis that one’s own name is incidentally detected in task-irrelevant speech. Semantic-incongruencies also elicited an enhanced negative response around 400ms, when compared to congruent sentence-endings. This is in line with the classic effects of semantic congruency on the N400 ERP component, and provides indication that task-irrelevant speech is indeed processed semantically and that violations of semantic expectations are noticed implicitly. Interestingly, neither one’s name nor semantic violations generated systematic shifts in eye-gaze, suggesting that if attention was captured by these items, capture was of a covert nature. This study provides exciting new insights on an old question in cognitive neuroscience. We show that under stimulated ecological conditions, the brain processes semantic aspects of task-irrelevant speech, and actively responds to salient semantic events such as violations or personally-relevant words. As such, our results are in line with proposed ‘late-selection’ models of attention and invite fine-tuning of theoretical accounts of ‘bottlenecks’ for concurrent speech processing.