An Evaluation of the Principles and Computational Mechanisms of Redundant Signals Effect in Multisensory Integration

One benefit of multisensory integration is the acceleration of responses in a perceptual decision task, known as the redundant signals effect (RSE). Based on probability summation, two basic principles explaining such benefit were established: First, the “principle of congruent effectiveness” suggests that the benefit is greatest when unisensory performances are similar. Second, the “variability rule” proposes that the benefit is largest when unisensory performances are unreliable. However, it remains unclear whether these principles extend to a complex and dynamic context and whether one principle predominates the other. Furthermore, it remains unknown whether the race model or the coactivation model, optimally explains the RSE and how the principles are reflected in those models. To address these questions, we evaluated RSE in a multisensory context featuring transient temporal dynamics (abrupt onsets and offsets) using a change detection task. Our results showed that both principles predicted the rank of the RSE in this dynamic setting. The principle of congruent effectiveness emerged as the dominant principle, where the similarity in evidence accumulation rate as the key predictor. Model comparison results revealed that relative to coactivation models, two context-variant race models best fitted the observed RSE. In these two models, the key predictor in the dominant principle was well associated with enhanced evidence accumulation rates or lowered decision criterion in each modality. Together, these findings demonstrate the generalization and contribution of the RSE principle in a dynamic environment, as well as its association with the potential computational mechanisms underlying the RSE.