Listening to naturalistic auditory stimuli elicits changes in brain activity which represent processing of speech from simple acoustic features to complex linguistic processes. Here, we used complex linguistic features, namely the distinction between function and content words from a recording of listening to natural speech, to examine the differences in stimulus processing between clear and degraded conditions.

We recorded fMRI data of 30 healthy, normal-hearing participants listening to the audio description of the movie “Forrest Gump” in German (Hanke et al., 2014, doi:10.1038/sdata.2014.3). The audio movie was presented three times in three different recording sessions in eight segments with different levels of simulated hearing loss (CS – clear stimulus, S2 – mild degradation, N4 – heavy degradation). The degraded stimuli were produced according to Bisgaard et al. (2010, doi:10.1177/1084713810379609). In each session, participants listened to the whole movie but with a randomized sequence of stimulus degradation levels.

We used speech annotations for the audio movie provided by Häusler and Hanke (2021, doi:10.12688/f1000research.27621.1) to categorize each spoken word into word classes – function words, content words, and rest (hard to classify, like interjections). These categories were then used as regressors to predict BOLD time courses. We added a word duration regressor, and orthogonalized the other regressors. This allowed us to obtain word class specific activation estimates unbiased by average word length differences between classes. In addition, we did another analysis where pronouns were split from function words into a separate regressor.

Our results indicate distinctive spatial patterns of BOLD activity in response to function versus content words. Function words elicited significant frontal and temporal activations, whereas content words elicited significant parietal activations. Pronouns displayed the same patterns as the other function words, but with larger effect sizes. Word duration explained a lot of the activity in the temporal lobe presumably related to early auditory processing. Most activations remained across stimulus degradation levels, but with smaller effect sizes. The highest degradation level N4 showed some additional effects – with a slight increase in motor cortex activity.

Our study reveals distinct but overlapping cortical representations of content and function words when listening to continuous speech in natural soundscapes with different simulated hearing capabilities. It also points out the importance of accounting for differences in average word length between word classes.

Funding: This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy—EXC 2177/1—Project ID 390895286.