Hysteresis in auditory decisions

Coordinator: Dorothee/Psyche/James??

Project idea

a short depiction of the experimental design

The goal of this project is to investigate electrophysiological correlates of hysteresis in speech perception with EEG.

The hypothesis is the following: a subject's perceptual decision depends not only on the physical stimulus, but also on prior expectations. For example, a subject that expects to hear a speech signal in a given sound stimulus is more likely to identify the signal as containing speech.

To investigate this, I propose to use a hyteresis paradigm (depicted on panel B of the figure). In "descending" trials, the first sound is clean speech, which is gradually degraded (in a way that still has to be defined) such that the last few trials do not contain speech at all. The subject has to report the number of stimuli containing speech that he has heard. In "ascending" trials, the opposite happens, that is the first sound does not contain speech, and the last one contains clean speech. An intuitive prediction is that subjects will display a form of hysteresis: some stimuli will be perceived as containing speech in ascending trials, but not in descending trials. Those stimuli are therefore ambiguous, as they can be perceived either way depending on the preceding context (as schematized in panel A of the figure). This ambiguity arises because in descending trials the subject is strongly primed towards hearing speech, this does not happen in ascending trials in which the first few sounds do not contain speech.

We first have to check that this hysteresis effect does indeed happen in the task that we choose to design. This will most likely depend on the "degradation dimension" that we use. A classical choice would be SNR: we gradually sink a word in white/pink/babble noise. Other options include using vocoded speech stimuli, and use a decreasing number of channels in the vocoder.

If we record the EEG signals of subjects while doing this task, we may be able to quantify the change in evoked potentials to the ambiguous stimuli. Those should reflect the effect of the subject's expectation.

Related work (and papers)

References? good question