The ability of humans, particularly children, to rapidly appreciate correlations and patterns in the information they perceive visually. A salient non-visual example is the ability of infants as young as 8-months-of age to detect the statistical properties of syllable co-occurrence to segment novel words. Seemingly, they do not just detect how frequently syllable pairs occur, but rather the probabilities with which one syllable predicts another. It may be the case then that infants discover word boundaries by selecting syllable pairs with low transitional probabilities. The same ability seems also to apply to non-linguistic sounds in that infants can detect the probabilities with which musical tones predict one another, thus suggesting that statistical learning abilities used for word segmentation apply to their perception of music. What is even more striking is the demonstration that infants, but not adults, can detect the statistical of sequences of absolute pitch in a tone sequence task.
See Bayesian learning, Co-occurrence learning, Learning, Moderate-discrepancy hypothesis