Abstract
Spatiotemporal correlation is important for driving multisensory integration as it indicates the signals derive from a common source. However, it has been shown that sequences of visual flashes and auditory clicks are optimally integrated regardless of the temporal correlation between the sequences in a rate-discrimination task (Raposo et al., J. Neurosci., 2012). Yet, in a localization task with similar stimuli, only synchronous click-flash sequences were integrated optimally (Parise et al., Curr. Biol., 2012). Thus, the conditions for multisensory integration of audiovisual sequences appear to be task-dependent. We examined both the spatial and temporal conditions necessary for optimal integration in rate-discrimination to better understand this phenomenon. Observers judged the rate of a comparison stimulus relative to a multisensory 8 Hz standard. Rate-discrimination thresholds were estimated for visual-only, auditory-only and multisensory conditions in each session using interleaved adaptive procedures. There were four multisensory conditions: Sequences were spatially congruent or not, and temporally synchronous or not. Spatial congruence: The auditory stimulus was either from a centrally placed speaker beneath the visual stimulus or a speaker 53 deg to the right. Temporal sequences were either synchronous or generated independently for each modality. Visual stimuli: Gaussian blurred white disks (2.5 deg SD). Auditory stimuli: bandpass noise (200 Hz – 10 kHz). Event duration: 17 ms. Stimulus duration: 2000 +/- 250 ms. Most observers were indistinguishable from optimal (based on measured single-modality JNDs) when audiovisual stimuli were spatially congruent and temporally synchronous. In all other conditions, most observers were sub-optimal. Therefore, observers relied on temporal cues in the decision to integrate or not, unlike the findings of Raposo et al. (2012), as well as spatial cues (not previously demonstrated). Differences between the two rate-discrimination studies will be discussed (including: sequence generation methods, detectability of stimuli, training). We suggest that multisensory integration for audiovisual sequences is not task-dependent.
Meeting abstract presented at VSS 2016