Everyday communication typically entails situations in which multiple persons speak simultaneously (“cocktail-party”). Cocktail-party situations can be static (constant target talker) or dynamic (target changes unpredictably). Dynamic situations require listeners to monitor multiple talkers and to switch attention when the target changes, which is why they cause a higher cognitive load and thus lead to a decrease in speech recognition compared to static situations, which is referred to as “costs” (e.g. Lin & Carlile, 2015, doi:10.3389/fnins.2015.00124; Meister et al., 2020, doi:10.1016/j.heares.2020.108020). A common limitation of corresponding studies is, however, the use of matrix sentences that provide hardly any semantic context, as opposed to sentences from everyday life. Semantic context might help listeners to segregate auditory streams and recall words from memory (e.g., Meister et al., 2013, doi:10.1016/j.neuroscience.2012.12.006). We therefore hypothesize that high semantic context can reduce cognitive load and thus allows listeners to use a higher proportion of their cognitive capacity to overcome the challenges of dynamic cocktail-party listening, which in turn lowers costs. Moreover, data from Meister et al. (2013) suggests that this context benefit tends to be smaller in older than in younger listeners, at least when a divided attention task was used.

We simulated cocktail-party situations with three spatially separated competing talkers. Participants had to repeat back the words uttered by the target talker (who had to be identified by her voice). Verbal response times were recorded automatically to serve as a measure of cognitive load. In static situations the target talker’s location remained constant and was announced prior to stimulus presentation, whereas dynamic conditions entailed unpredictable location changes of the target. Low context (“Lisa draws twenty heavy apples.”) as well as high context (“The rocket flies into space.”) sentences were used.

At the previous SPIN workshop we presented the results of a small sample of young normal hearing adults obtained with this setup (Wächtler & Meister, 2024, doi:10.5281/zenodo.10565092). To be able to study the effects of age and hearing loss, our current presentation will also include data from elderly people (> 60 years) with and without hearing loss.

We will examine the impact of semantic context on the costs of dynamic cocktail-party listening and how it is influenced by age and hearing loss. Moreover, verbal response times will be analyzed to shed some light on how semantic context could affect cocktail-party listening beyond recognition performance.

Funding: Deutsche Forschungsgemeinschaft (ME2751/3-2).