Measures of multisensory integration based on dependent probability summation: from spike counts to reaction times

TL;DR

This paper introduces a new index for assessing multisensory integration in neurons based on probability summation, which accounts for variability and stochastic dependence, providing a more accurate measure than traditional methods.

Contribution

The paper proposes an alternative, distribution-free index for multisensory integration that considers stochastic dependence, improving upon traditional measures based solely on mean responses.

Findings

The new index often reclassifies neurons previously labeled as multisensory.

It is sensitive to data variability, unlike traditional mean-based indices.

The index is easy to compute and does not assume specific response distributions.

Abstract

A single neuron is categorized as"multisensory" if there is a statistically significant difference between the response evoked by an audio-visual stimulus combination and that evoked by the most effective of its components individually. Crossmodal enhancement is commonly expressed as a proportion of the strongest unisensory response. However, being responsive to multiple sensory modalities does not guarantee that a neuron has actually engaged in integrating its multiple sensory inputs, rather than simply responding to the most salient stimulus. Here, we propose an alternative index measuring by how much the crossmodal response surpasses the level obtainable by optimally combining the unisensory responses. Optimality is defined by probability summation combining the unisensory responses under maximal negative stochastic dependence. The new index is analogous to measuring crossmodal enhancement by the amount of violation of the "race model inequality", which is widely used in reaction time studies of multisensory integration. Neurons previously labeled as "multisensory" may lose that property since the new index tends to be smaller than the traditional one. This is exemplified with a data set collected from single SC neurons. The new easy-to-compute index does not require any specific distributional assumption. It is sensitive to the variability in the data, in contrast to the traditional index which, by definition, only depends on the means of the uni- and crossmodal response distributions.