Generally speaking, multisensory integration is more likely to occur when the stimuli are synchronous (Stein and Meredith, 1993). Repeated exposure to temporally offset multisensory stimuli can change the perceived delay between the stimuli so that synchrony is perceived closer to the adapted delay rather than physical synchrony (Fujisaki et al., 2004). If the perception of synchrony is adaptable, might the point (or delay) of maximal integration also be altered after adaptation? Temporal adaptation might be achieved by changing the processing times of the component stimuli (Harrar and Harris, 2008; Navarra et al., 2009), or changing the integration mechanism. In the present study, each participant underwent daily adaptation to either synchronous or asynchronous (auditory lagging by 200 ms, or visual lagging by 60 ms) stimulus pairs. To assess unimodal processing time changes, we measured reactions times (RTs) to audio and visual stimuli after adaptation. In order to assess the effects of adaptation on multisensory integration, we measured RTs to synchronously presented AV stimuli and compared these with the RTs predicted from the Miller’s race model (Miller, 1982) for each participant (Molholm et al., 2004). The results comparing RTs following synchronous and asynchronous adaptation conditions are discussed in the context of perception versus action and current models of multisensory integration. The RTs changed considerably over a period of a week; these patterns are discussed in the context of learning to perceive synchrony.

We investigate the time constant of recovery from adaptation to temporal asynchrony. Subjects adapted to a 4-min movie of a nail being hammered, a naturalistic animation with strong audiovisual temporal cues. The movie soundtrack had a constant asynchrony of either ±200 ms. Following adaptation, we sampled synchrony judgments every 2 seconds for 2 min using a brief audiovisual synchrony task (a flash and beep). The flash and beep varied over several ±SOAs and subjects indicated whether they were synchronous. Binning the synchrony responses within a short, rolling time window we obtained estimates of PSS over the course of recovery from temporal adaptation. The rolling average PSS showed a significant recalibration initially followed by a clear recovery function, with PSSs steadily returning to baseline after ∼60 s. We also analysed short-time scale recalibration by testing for adaptation effects between successive synchrony probes. Although these probes were brief (60 ms), we found that a given synchrony judgment in the post adaptation period was strongly influenced by the sign of the previous synchrony probe, showing an adaptation effect in the direction of the preceding synchrony probe’s SOA. In sum, these results show long- and short-scale temporal recalibration effects, with the short-scale inter-probe effects superimposed on the long-scale recalibration. In a second experiment, we delayed the synchrony probes until 60 s post-adaptation and observed no long-scale temporal recalibration, showing there is no storage of long-scale temporal adaptation.