Gustatory cortex neurons perform reliability-dependent integration of multisensory flavor inputs

Flavor is the quintessential multisensory experience, combining gustatory, retronasal olfactory, and texture qualities to inform food perception and consumption behavior. However, the computations that govern multisensory integration of flavor components and their underlying neural mechanisms remain elusive. Here, we use rats as a model system to test the hypothesis that taste and smell components of flavor are integrated in a reliability-dependent manner to inform hedonic judgments and that this computation is performed by neurons in the primary taste cortex. Using a series of two-bottle preference tests, we demonstrate that hedonic judgments of taste + smell mixtures are a weighted average of the component judgments, and that the weight of the components depends on their relative reliability. Using extracellular recordings of single-neuron spiking and local field potential activity in combination with decoding analysis, we reveal a correlate of this computation in gustatory cortex (GC). GC neurons weigh bimodal taste and smell inputs based on their reliability, with more reliable inputs contributing more strongly to taste + smell mixture responses. Input reliability was associated with less variable responses and stronger network-level synchronization in the gamma band. Together, our findings establish a quantitative framework for understanding hedonic multisensory flavor judgments and identify the neural computations that underlie them.