Omer F. Yildiran¹ (), Long Ni^1,2, Michael S. Landy^1,2; ¹Department of Psychology, New York University, ²Center for Neural Science, New York University

How does the brain integrate conflicting temporal information across sensory modalities? Using small conflicts (±50 ms), Hartcher-O’Brien et al. (2014) showed that observers integrate audiovisual duration cues optimally. Does causal inference lead to a breakdown of audiovisual integration when duration conflicts are large? Two levels of auditory signal-to-noise ratio (SNR) were used. We confirmed in a unimodal 2-IFC task that lower auditory SNR increased auditory duration discrimination threshold. Cross-modal duration discrimination revealed a consistent bias: auditory intervals were perceived as longer than physically matched visual intervals. This bias was taken into account in the main bimodal experiment. In the bimodal task, participants compared the auditory durations of two bimodal stimuli. Test stimulus: the two cues were consistent (same apparent duration). Standard stimulus: the two cues differed by seven conflict durations (range: ±250 ms). Auditory duration percepts shifted systematically toward the visual duration, especially when auditory uncertainty was high. The shift was proportional to the amount of cue conflict, inconsistent with the predictions of causal inference. We compared fits of several models. A heuristic model in which the observer probabilistically switches between the visual and auditory cues was preferred for most participants. However, AIC differences across models were small. Simulations revealed why the models were difficult to distinguish: Given the observed sensory noise, competing models can only be discriminated using unreasonably large conflicts. Within the reasonable conflict range used here, all models produced largely overlapping, near-linear patterns of shift as a function of the degree of cue conflict. In conclusion, observers do not rely on causal inference when judging durations under our conditions. The heuristic, cue-switching is favored, but the large uncertainty of duration estimates fundamentally limits our ability to discriminate computational models in this domain.

Acknowledgements: NIH EY08266, NYUAD Center for Brain and Health, funded by Tamkeen under NYU Abu Dhabi Research Institute award CG012