A meta-analytic review of the literature on infant habituation and recognition memory performance as predictors of later IQ suggests several conclusions: (1) Habituation and recognition memory assessments made on a variety of risk and nonrisk samples in the first year of life predict later IQ assessed between 1 and 8 years of age with a weighted (for N) average of normalized correlations of .36 or a raw median correlation of .45. (2) The size of the predictive correlation is essentially the same for habituation and for recognition memory paradigms. (3) This prediction phenomenon is not obviously associated solely with one laboratory, one particular infant response measure, or a few extremely disordered infants. (4) The level of prediction to childhood IQ is substantial given the reliability of the infant measures. (5) Predictions are somewhat higher for risk than for nonrisk samples. (6) Predictions are consistently higher than for standardized infant tests of general development for nonrisk but not for risk samples, and they are not consistently higher than predicting from parental education and socioeconomic status or a few other infant behaviours for nonrisk samples. (8) Coefficients may be higher when the predicting assessments are made between 2 and 8 months of age than earlier or later, but prediction coefficients are remarkably consistent across the observed outcome age period of 2-8 years.