A subtype of ultrasonic vocalizations during highly palatable food consumption in rats identified by machine learning-assisted classification

Abstract

Identifying behavioral and physiological responses to rewarding stimuli is essential for understanding positive emotional states in animals and for investigating their neural basis. Ultrasonic vocalizations (USVs) in rats are modulated by various behavioral contexts and are thought to represent affective-like states. However, their diversity during palatable food consumption remains underexplored. In this study, we investigated acoustic features of USVs in male rats consuming chocolate, a highly palatable food. Using a machine-learning-assisted logistic regression model trained on spectrogram features, we identified a distinct USV subtype--the 40-kHz inverted-U type--that was selectively emitted during chocolate consumption. The emission of this subtype was tightly time-locked to chocolate feeding behavior. Systemic administration of naloxone, an opioid receptor antagonist, significantly reduced the emission of the 40-kHz inverted-U USVs during chocolate intake. These findings suggest that these USVs may reflect internal states associated with palatable feeding and are under modulation by the endogenous opioid system. Moreover, our data demonstrate the utility of machine learning for high-throughput, objective classification of USV subtypes. This framework provides a promising approach for decoding emotion-related vocal expressions in rats and highlights the potential of specific USVs as behavioral readouts for studying the neurobiology of positive emotions. Significance StatementUltrasonic vocalizations (USVs) are high-frequency calls produced by rats known to correlated with emotional and motivational states, yet their diversity and functional relevance remain underexplored. In this study, we identified a novel 40-kHz inverted-U USV subtype associated with chocolate consumption. These calls were closely tied to feeding behavior and were selectively suppressed by an opioid receptor antagonist, suggesting modulation by the brains reward system. Using machine learning-assisted classification, we established a robust method to detect these vocalizations. Our findings highlight the potential of specific USVs as non-invasive, real-time behavioral indicators of internal states, offering new tools for studying affect, reward, and communication in animal models.