Neurocomputational models of cognition have framed aesthetic appreciation within the domain of knowledge acquisition and learning, suggesting that aesthetic appreciation might be considered as a hedonic feedback on successful perceptual learning dynamics. Such hypothesis, however, has never been empirically demonstrated yet. In order to investigate the relationship between aesthetic appreciation and learning, we measured the EEG mismatch negativity (MMN) response to more or less appreciated musical intervals, which is considered as a reliable index of perceptual learning. To this end, we measured the MMN to frequency (Hz) standard and frequency deviant musical intervals (Experiment 1) while participants were asked to judge their beauty. For each single stimulus, we also computed an information-theoretic index of perceptual learning (Bayesian surprise). We found that more appreciated musical intervals were associated with a larger MMN responses, which, in turn, correlated with trial-by-trial fluctuations in Bayesian surprise (Experiment 1). Coherently with previous results, Bayesian surprise was also found to correlate with slower RTs in a detection task of the same stimuli, evidencing that motor behavior is inhibited in presence of surprising sensory states triggering perceptual learning (Experiment 2). Our results provide empirical evidence of the existence of a positive correlation between aesthetic appreciation and EEG indexes of perceptual learning. We argue that the sense of beauty might have evolved to signal the nervous system new sensory knowledge acquisition and motivate the individual to search for informationally profitable stimuli.