Distinct patterns of multisensory integration underly the expression of flavor preferences and aversions in rats

Flavor is an inherently multisensory experience, combining gustatory and retronasal odor signals during consumption. Previous work has shown that rats weight taste and smell components of flavor mixture solutions to express flavor preferences. However, a comprehensive quantitative understanding of taste-smell integration is lacking. Here, we evaluated the validity of the maximum likelihood estimation (MLE) framework for explaining multisensory integration of taste and smell signals toward hedonic flavor judgments. MLE is a dominant framework for explaining multisensory integration in a wide variety of contexts, and states that components of a mixture are weighted according to their relative reliability, resulting in more reliable multisensory judgments. In our experiment, rats expressed preferences for taste, odor and taste+odor solutions in a series of two-bottle preference tests featuring an unpalatable odor component (amyl acetate), and either a palatable taste component (sucrose) or an unpalatable taste component (citric acid). Taste reliability was experimentally varied by presenting taste components at two different concentrations. Reliability of animals' preference judgements was quantified as the variance in preference across repetitions of the same conditions in the same animals. Our results confirm that preferences for mixtures featuring a palatable taste can be described as a weighted average of the component preferences where the weights depend on the relative reliability of the components-a finding that is in line with the MLE framework. However-contrary to MLE-component weighting does not increase the reliability of multisensory preference judgments themselves. In addition, preferences for taste+smell mixtures featuring an unpalatable taste follow a qualitatively different pattern: taste and smell components are combined to enhance flavor rejection judgements. These results demonstrate that the multisensory integrative principles underlying flavor preference differ from other multisensory domains and depend on the valence of the flavor components.