Who are you listening to? Towards a dynamic measure of auditory attention to speech-on-speech
Listening to a speech signal is a perceptual task that is ecologically relevant and used in many experimental studies in psychology and neurosciences. During the last 20 years, in particular, a growing number of psychoacoustic or brain imaging studies have used a task of listening to concurrent speech signals in order to investigate, for example, different phenomena of auditory masking, or, brain correlates of selective auditory attention.
In general, in these speech-on-speech tasks, the listener is asked to selectively direct his attention on one of the voices. However, these tasks assume that the listener keeps a constant focus on the target voice, while in real-life situations, listeners' attention can quickly vary from one voice to another. To approach more realistic situations, it is therefore important to have new methods to infer retrospectively if, and at what times, the attention of the listener has turned away from the target voice.
We have therefore developed a paradigm and a set of materials to infer the dynamics of auditory attention over time. After listening to two simultaneous stories — a target and a masker — the participants have to find, among a set of words, those present in the target story. The words were selected according to several criteria such as repetition, their frequency in the language, the recency effect and the primary effect.
In the present study, the masker were uttered by the same talker as the target stories, but the voice parameters (F0 and vocal tract length) were manipulated to parametrically control similarity of the two voices, hence controlling the difficulty of the task. The difficulty of the task was also manipulated by alternating the modes of the presentation of the stimuli, namely, dichotically and diotically presentations. As the target and masker stimuli are more similar, we expect to observe more responses corresponding to the masker voice, which would reflect more spurious switches of attention to the wrong talker. Preliminary results of this study will be presented.