Auditory Word Comprehension is Less Incremental in Isolated Words

Abstract

Speech input is often understood to trigger rapid and automatic activation of successively higher-level representations for comprehension of words. Here we show evidence from magnetoencephalography that incremental processing of speech input is limited when words are heard in isolation as compared to continuous speech. This suggests a less unified and automatic process than is often assumed. We present evidence that neural effects of phoneme-by-phoneme lexical uncertainty, quantified by cohort entropy, occur in connected speech but not isolated words. In contrast, we find robust effects of phoneme probability, quantified by phoneme surprisal, during perception of both connected speech and isolated words. This dissociation rules out models of word recognition in which phoneme surprisal and cohort entropy are common indicators of a uniform process, even though these closely related information-theoretic measures both arise from the probability distribution of wordforms consistent with the input. We propose that phoneme surprisal effects reflect automatic access of a lower level of representation of the auditory input (e.g., wordforms) while cohort entropy effects are task-sensitive, driven by a competition process or a higher-level representation that is engaged late (or not at all) during the processing of single words.

Cortical tracking of continuous speech under bimodal divided attention

“…This effect was statistically only seen after excluding the auditory single-task condition and should thus be interpreted with care, but it is consistent with several extant findings. The dissociation between entropy and surprisal is consistent with recent evidence that these two processes may reflect different neural processes (Gaston et al, 2022). Neural responses associated with surprisal may reflect prediction errors that signal the difference between predicted and observed phonemes.…”

“…Behavioral studies with simple speech stimuli indicate that reduced acoustic processing under bimodal divided attention may lead to compensatory changes manifested by increased reliance on higher-order linguistic knowledge during auditory lexical perception (Mattys et al, 2009). However, to date, there is a lack of a systematic and holistic analysis of divided attention-related changes across different levels (acoustic-to-linguistic) of natural continuous speech processing, which is distinctly different from processing simple speech stimuli (Gaston et al, 2022; Hamilton & Huth, 2020).…”

Recurrent neural networks as neuro-computational models of human speech recognition

“…MEG data were drawn from a public dataset (Gaston et al, 2022). Preprocessing and source localization are summarized here, and are described in more detail in a previous publication (Gaston et al, 2023).…”