Item

Abstract

Despite continual temporal discrepancies between sensory inputs, signals arising from the same event are bound together into a coherent percept. It has been suggested that multiple timekeepers monitor the different sensory streams, producing differences in perceived duration of events. Given this, what is the integration strategy adopted for combining sensory information in the time domain? Specifically, if the brain has information about the duration of an event from more than one source, can the uncertainty of the duration estimate decrease, and can the Maximum Likelihood Estimate (MLE) model predict such a change? Using a 2AFC procedure, participants had to judge which interval was longer (1st or 2nd) for auditory, visual and audiovisual stimuli. Each trial contained 2 intervals: a standard stimulus (sampled from one of three durations), and a comparison interval whose duration changed randomly in relation to standard stimulus duration. The reliability of the auditory stimulus was manipulated to produce the unimodal weighting scheme. Data was fit with a cumulative Gaussian psychometric function from which PSE and JND were extracted.
Results for unimodal trials showed JND changes that depended upon the duration of the standard, according to Weber's law. JND values also decreased with decreases in signal noise. Comparison of the present bimodal results with MLE predictions revealed optimal integration of auditory and visual duration cues. Additionally the results show that the integration of uncertain visual and auditory duration signals is a weighted average of these signals. That is, PSE shifts in perceived duration tended to reflect MLE predictions with shifts following the more reliable unimodal signal. These results are the first to demonstrate ‘optimal’ integration of sensory information in the time domain and contradict other studies applying MLE to this stimulus feature.