Neural multiplexing vs place coding: Multiplexing steps in when most needed

Abstract

Sensory scenes contain many different stimuli. Two complementary theories about how the brain segregates signals from different stimuli concern (a) time division multiplexing, such that neurons switch between encoding each item over time; and/or (b) place coding, such that different populations of neurons encode each item. Such time division multiplexing would appear to be required when the population of neurons responsive to each stimulus overlaps, as place coding lacks the granularity to resolve the two stimuli. This predicts that as responses to component stimuli become more similar, and thus less well resolved by place coding, there would be a greater incidence of multiplexing. We tested this hypothesis using single-unit responses in the macaque inferior colliculus to combinations of two sounds of varying center frequencies (given that sound frequency is place coded in this structure). We found that neurons were more likely to multiplex when their responses to each individual sound was more similar, differentiating signals whose neural representations would otherwise be less distinct. This finding supports the theory that neurons multiplex to maintain information about concurrent stimuli when place coding is insufficient to prevent largely overlapping responses in the neural population.