A meta-analysis of the A-not-B error was conducted using logistic regressio
n on studies conducted before September 1997 (107 data points). An earlier
meta-analysis by Wellman, Cross, and Bartsch revealed that age, delay betwe
en hiding and retrieval, and number of hiding locations were significant pr
edictors of both the proportion of infants who searched correctly on B tria
ls and the proportion of infants who searched perseveratively. The current
analysis replicated these findings with two exceptions: (1) The number of t
rials at the A location was a significant predictor, and (2) the number of
locations was a significant predictor of the proportion of infants who sear
ched perseveratively, but not the proportion of infants who searched correc
tly. Implications of these findings are discussed and a quantitative versio
n of a hierarchical competing-systems model of infant search is proposed.