Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Individual differences in the impact of semantic ambiguity on speech-in-noise perception
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster D79 in Poster Session D, Wednesday, October 25, 4:45 - 6:30 pm CEST, Espace Vieux-Port
Jaimy Hannah1, Stephen van Hedger2, Jenni Rodd3, Ingrid Johnsrude1; 1Western University, 2Huron University College, 3University College London
The cognitive demands imposed by a speech-listening situation can vary dramatically and include both perceptual demands such as the presence of masking background noise, and linguistic demands, such as homophony, which is ubiquitous in English. Whereas speech in noise requires the listener to disambiguate phonological form (e.g., distinguish /kæt/[cat] from /kæp/[cap]) homophony provides a clear form (e.g., /bɑːk/ [bark]) but listeners must select the appropriate meaning. Although listeners can use meaningful context to select the appropriate word form (for masked sentences) or word meaning (for sentences with homophones), brain imaging data from our lab revealed that the brain networks recruited to understand acoustically degraded sentences and sentences containing homophones were largely disjoint, overlapping only in the cingulo-opercular network. Furthermore, when the two challenges were combined, intelligibility of masked sentences without homophones was substantially higher than that of identically masked sentences with homophones, even when intelligibility overall was high. The reduction in intelligibility for sentences with, compared to without, homophones varied dramatically across participants and may be related to individual differences in cognitive ability. Given how common homophony is in everyday speech and the size of the effect, this phenomenon probably has substantial real-world impact on speech understanding. In the current study, we further investigate how these two challenges (homophony in sentences, and background noise) interact. We examine the intelligibility of sentences with and without homophones, masked with different levels of multi-talker babble (a more naturalistic masker compared to the signal-correlated noise masker used in our previous work). We also take measures of matrix reasoning (related to fluid intelligence) and working memory (reading span, n-back) to examine whether these relate to individual differences in the degree to which homophony reduces intelligibility of sentences. In an online study, 121 participants transcribed 112 sentences, one at a time, half containing homophones (“high-ambiguity”) and half without (“low-ambiguity”). Sentences were masked with 12-talker babble, at +2 dB, +4 dB, and +6dB signal-to-noise ratio (SNR), and were also presented clearly (no masker). In a second session, participants completed the cognitive tests. Word-report performance decreased as masking increased. High-ambiguity sentences were less intelligible than low-ambiguity sentences at all SNRs. The difference in intelligibility between low and high-ambiguity sentences increased as SNR decreased. Individual variability in this difference also increased as SNR decreased: some individuals appear to be less affected than others by semantic ambiguity when background noise is also present. We will investigate individual variability in the combined effects of semantic ambiguity and noise on intelligibility by relating scores on the cognitive tasks to the difference in word-report accuracy between low and high-ambiguity sentences. Higher scores on all tasks should lead to higher intelligibility, regardless of the presence of ambiguity. We also predict that fluid intelligence, indexed by matrix reasoning, will help to mitigate the effects of both noise and ambiguity on intelligibility, such that those who score higher will be less affected by background noise and will show less of an effect of ambiguity as the noise level increases.
Topic Areas: Speech Perception,