A method for assessing approach and avoidance behavior across multiple olfactory stimuli in mice including multivariate hypothesis comparisons

Abstract Approach and avoidance behavior towards sensory stimuli serve as powerful behavioral readouts of our mental representations of the external world and our expectations and motivations in navigating it. In the olfactory system, approach or avoidance of odors statistically associated with people, places, and things relate to ecologically critical functions like feeding, fear, and reproduction. However, experimental methods for quantifying approach/avoidance behavior in relative terms across odors have been limited. Here we present a novel method for quantifying mouse approach/avoidance in an open field arena scented with up to four odors simultaneously. In lieu of traditional inferential statistics (which greatly limit the information that can be learned in this multivariate experiment), we demonstrate the a priori definition of quantitative hypotheses for the distribution of time among scented corners and the use of information theory-derived statistical metrics to quantify the relative likelihood of each competing hypothesis given the data collected. Finally, we use data from a fear conditioning experiment to demonstrate the application of this method to conclude that fear conditioned mice exhibit a fear generalization gradient that decreases as odorants become more different from the threat-predictive odorant, as opposed to competing hypotheses that mice are specifically avoiding the threat-predictive odorant or have overgeneralized their fear and avoid all test odors regardless of similarity. Critically, this method takes only a few minutes per animal with no prior behavioral training required, and it can be performed easily without automated apparatus.