The present meta-analysis of the predictive validity of visual habituation and visual dishabituation shows that the weighted and normalized average correlation between infant habituation/dishabituation and childhood cognitive performance is 0.37. In contrast to the findings of earlier reviews, for risk samples, dishabituation is apparently superior to habituation in predicting later intelligence, whereas for nonclinical samples, there is evidence to suggest that the opposite pattern holds true; that is, predictive validity is higher for habituation than for dishabituation. Prediction from habituation/dishabituation is not affected by when these variables are assessed in infancy. Furthermore, the predictive validity of habituation/dishabituation is substantial and stable up until adolescence. These and further conclusions are elaborated upon and discussed.