E45 Multimodal sensory integration

Topic

In general, this insect simultaneously integrates the visual signal (which defines another model in this set), the mechanical signal (defining a distinct model in the set), a specific chemical signal, the electrical signal (defining another model in the set), and the thermal signal (defining a distinct model in the set), precisely to construct a robust final representation of its environment.

This optimal multimodal integration—described by Bayes' rule—consists of weighting each sensory channel inversely proportional to its specific noise variance: a significantly more reliable sensory channel is consequently assigned a significantly higher weight within the integration. The greater the reliability of the insect's sensory channel, the greater the final weight assigned to that channel within the multimodal integration.

The criterion of superadditivity establishes that the insect's final response to a specific bimodal stimulus is greater than the best response to any single unimodal stimulus—exactly the phenomenon observed in an equivalent insect neuron.

The reduction of variance through integration predicts that the final precision of the insect's integrated estimate exceeds the precision of any single, isolated sensory channel.

Shannon's information theory—which establishes another model in this set—quantifies the amount of information regarding the stimulus transmitted by each of the insect's neuronal channels and how that information is combined across the different channels. Consequently, this model—incorporating the simultaneous visual, mechanical, chemical, electrical, and thermal signals that initiated the process—results in the insect’s final processing occurring within the specific circuits of its optic lobe, antennal system, and mushroom body. This process entails a final latency on the millisecond scale—matching the latency established by another model in this set regarding a receptor's temporal response. The fact that the insect's circuitry integrates these same initiating signals—visual, mechanical, chemical, electrical, and thermal—with even greater speed means that the resulting robust representation of its environment is, in turn, even more precise.

ID:2532