Abstract
Multisensory neurons integrate concordant visual-auditory cues in order to enhance the detection and reaction to such cross-modal events. As shown in studies of superior colliculus (SC) neurons, the magnitude of multisensory enhancement obeys a principle of inverse effectiveness: proportionately larger products are achieved when the effectiveness of the component cues is reduced; whether that reduction is caused by changes in their physical properties or the complexity of the scene in which they appear. The question posed here was whether these products were truly sensitive to the effectiveness of the cues or, as we hypothesize, to the underlying reliability of the sensory estimates. These two features generally co-vary in laboratory circumstances. In the present study they were dissociated in the alert cat by repeated presentations of visual, auditory, or visual-auditory cues. Temporal redundancy, like spatial complexity, reduces stimulus efficacy (often described as "habituation") but, unlike spatial complexity, does not decrease the reliability of the sensory estimates. The results revealed that the unisensory and multisensory responses of SC neurons are reduced in equal proportion so that, contrary to the stated principle of inverse effectiveness, the products of multisensory integration are maintained despite changes in response efficacy. Thus, while the principle of inverse effectiveness remains a good “rule of thumb” because efficacy generally scales with informational content, it is the latter rather than the former that is crucial for the adaptive scaling of multisensory integration. In this way familiar and non-salient cross-modal events can be appropriately discounted in favor of novel and salient events of any modality or modality combination. Supported by NIH grant (EY016716) and the Tab Williams Family Foundation.
Meeting abstract presented at VSS 2015