The reliability of multisensory integration in enhancing behavioral performance.

Although it is widely recognized that multisensory integration is a critical neural process underlying normal perception and behavior, estimates of its impact have varied among studies involving different contexts, tasks, and response measures. To address the reliability and magnitude of its effect, analysis was conducted on data from a benchmark visual-auditory detection and localization task used in seven studies using cat as the animal model, spanning 35 years. This extensive dataset enabled a robust evaluation of the magnitude, consistency, and optimality of multisensory integration while including common sources of experimental variability. The results revealed that the multisensory behavioral performance in this paradigm was consistently above the most stringent referent criteria, the sum of the unisensory performance levels. This "superadditive" multisensory effect was consistent with the predictions of an optimal model in which visual and auditory input signals are summed before being nonlinearly transformed into observable behaviors by a decision-making process modeled on signal detection theory. In operational terms, the multisensory effect reflected a consistent proportional enhancement of ∼50 % in behavioral performance and the analysis revealed that there was minimal variance across testing sessions and minimal sensitivity to factors such as animal sex, stimulus location, and trial history. Due to its reliability, significant multisensory effects can be detected with as few as three subjects and 15 trials per test condition (visual, auditory, visual-auditory), highlighting the utility of this paradigm for evaluating the efficiency of this neural process in enhancing functions essential for survival.