Access to sub-lexical information affects degraded speech processing: insights from fMRI

Natural speech perception suffers from uncertainty and ambiguity due to environmental noise. One factor that greatly improves the perception of degraded speech is contextual information. Previous studies on the modulatory effect of contextual information mostly focused on the beneficial role of semantic-lexical knowledge. Several studies manipulated the semantic expectancy of the last word within a sentence (Obleser et al. 2010) or, the semantic relationship between words (Golestani et al. 2013) and whole sentences (Guediche et al. 2016). Others investigated semantic priming of the degraded stimulus material through prior exposure to undistorted written or spoken presentation of the same stimulus (Sohoglu et al. 2012, Clos et al. 2014). However, the modulatory effect of pure sub-lexical speech context on speech intelligibility and the underlying neural responses are not well understood. We conducted a functional magnetic resonance imaging (fMRI) study (N=20) to determine the underlying neuronal activation pattern during perceptual adaptation to degraded sub-lexical speech. To isolate language specific sub-lexical characteristics from semantics our stimulus material consisted of pseudowords, phonetically and phonologically balanced to the German language (Wendt et al. 2007). Subjects listened to three successively presented pseudowords in each trial. The first and third pseudowords were identical and physically degraded. The intermediate pseudoword was intelligible and either matched or did not match the degraded pseudoword. Subjects were instructed to repeat the last presented degraded word. For the assessment of perceptual clarity, we analyzed success or failure of repetition. In addition, we analyzed the blood oxygen level dependent (BOLD) activation changes of the second degraded word in dependency of the intermediate clear word. Matching clear pseudowords compared to non-matching clear pseudowords significantly improved the listener’s perception of degraded speech stimuli (from 3.44% to 68.44% correct responses). Speech intelligibility was associated with activation increases in a frontal-temporo-parietal network, including bilateral posterior superior temporal cortex, angular gyrus, right supramarginal gyrus, right middle temporal cortex and left somato-motor cortices (p<0.001, uncorrected). To further investigate the adaptational changes in primary auditory areas we conducted a bilateral region of interest (ROI) analysis on the cytoarchitectonic-maps of the primary auditory cortex (Wild et al. 2012). All ROI’s showed increased activation when prior exposure to identical clear pseudowords enabled successful speech perception. Our results indicate that the exposure to intact sub-lexical speech characteristics, independent of semantic-lexical context, supports the perception of severely degraded speech and enhances activations along a broad cortical speech perception network.