The contribution of cognition in a variety of speech-perception-in-noise tests in normal hearing listeners

Speech perception is known to be effortful, particularly in adverse conditions. The importance of hearing sensitivity for speech-in-noise (SiN) perception has long been acknowledged, but cannot explain all individual differences. Cognition, and in particular working memory, has emerged as another key factor (Akeroyd, 2008). However, our understanding of cognition for speech perception has been limited by a lack of systematicity and theoretical rigor in their selection of speech stimuli and cognitive tests in many studies.

This study aimed to investigate the contributions of cognition to SiN perception in a systematic and theoretically guided way for a range of speech situations in a cohort of young adults (N = 44, aged 18-30 years) with normal hearing (<20dB HL PTA 0.25-8kHz). Participants were also tested for normal visual acuity (Landolt C test) and were self-reported as native English speakers with no known neurological, psychiatric or language disorders.

In the SiN tests, the speech situation varied concerning the target speech (from high to low predicable semantic content sentences to single words) and background noise (speech-modulated noise or 3-talker babble) with all combinations of fore- and background sounds being tested in the same experiment. Speech target stimuli were presented at 60dB SPL and two signal-to-noise ratios were set dependant on the background noise condition, -2dB SNR for 3-talker babble and -7dB SNR for speech-modulated noise. These different speech conditions were expected to invoke semantic processing and energetic versus informational masking to different extents, which in turn would be associated with the use of different cognitive abilities.
Cognitive abilities were assessed based on Baddeley’s model of working memory (Baddeley, 2000) and included tests for verbal working memory, phonological loop, episodic buffer, visuo-spatial scratchpad, and central executive. Individual differences in cognitive abilities were related to performance on each SiN perception.

We expect different associations between intelligibility and cognition for sentences and words given differences in semantic processing required for intelligibility. We also expect a different association between intelligibility and cognition for babble background compared to speech-modulated noise because babble invokes informational as well as energetic masking while noise invokes only energetic masking. Lastly, by explicitly measuring and associating different cognitive components such as storage, verbal abilities and executive control (mental manipulation) we hope to understand why complex working memory measures such as the reading span are often found to correlate so highly with speech perception.