Caught by surprise: the effect of unpredictable varying talker location on speech intelligibility and listening effort

Background - Having a successful conversation in a crowded environment can be challenging, especially if one has to divide his/her attention over multiple conversation partners. In such complex listening situations, cognition and multitasking skills are of particular importance. Furthermore, head orientations play an important role in improving speech understanding. On the one hand, listeners try to face the speaker to lip read the message. On the other hand, moving the head away from the speaker allows the listener to use interaural differences to improve speech intelligibility (SI), by means of head shadow effects, redundancy and binaural squelch. Research showed that both lip reading and binaural advantages can be combined for head movements up to 30° (1). Nevertheless, not knowing upfront who will speak can hamper appropriate head orientations, possibly impairing both SI and listening effort (LE). In this study, auditory-visual multitalker situations were administered to investigate the effect of unpredictable varying talker locations on both SI and LE.

Methods - 30 normal hearing, Dutch speaking adults (18-30 years) participated. A challenging listening situation was mimicked by means of an audio-visual environment and five virtual human-like characters. In this scenario, participants performed a behavioural speech-in-noise task with sentences either coming from a predefined direction (0°, fixed condition) or from randomly assigned directions between -45° to +45° in the horizontal field (random condition). A uniform noise field was created via a loudspeaker right above the participant’s head. Speech was presented auditory-only or auditory-visually (AV) and free head movements were allowed. In an extra condition, the AV speech-in-noise task was combined with three secondary tasks to investigate LE. It is known that a drop in performance on the secondary tasks, compared to their respective baseline conditions, reflects an increase in LE.

Results - SI was significantly better in the random listening conditions with regard to the fixed conditions. Performance on one of the secondary tasks significantly dropped when combining four tasks together, compared to its baseline condition. However, performance on the secondary tasks did not differ between the random and fixed conditions. First analysis of these results suggest good performance when speech was presented from varying talker locations compared to a fixed location, possibly due to binaural listening benefits. Furthermore, both listening conditions required an equal amount of LE.

References
Grange & Culling (2016) JASA, 140(6), pp. 4061–4072.

Funding — The Oticon Foundation and a TBM-FWO grant from the Research Foundation-Flanders (nr. T002216N).