Task switching costs contribute to the apparent speeding of multisensory reaction times

Abstract Faster reaction times for the detection of multisensory compared to unisensory stimuli are considered a hallmark of multisensory integration. While this multisensory redundant signals effect (RSE) has been reproduced many times, it has also been repeatedly criticized as confounding multisensory integration and general task switching effects. When unisensory and multisensory conditions are presented in random order, some trials repeat the same sensory-motor association (e.g. an auditory followed by an auditory trial), while others switch this association (e.g. an auditory followed by a visual trial). This switch may slow down unisensory reaction times and inflate the observed RSE. Following this line of ideas, we used an audio-visual detection task and quantified the RSE for trials derived from pure unisensory blocks and trials from mixed blocks involving a repeat or switch of modalities. The RSE was largest for switch trials and smallest for unisensory trials. In fact, during unisensory blocks the multisensory reaction times did not differ from predictions by the race model, speaking against a genuine multisensory benefit. These results confirm that the observed multisensory RSE can be easily confounded by task switching costs, and suggest that the true benefit of multisensory stimuli for reaction speed may often be overestimated.