Abstract
One hundred twenty-seven 7-, 9-, and 11-year-old children were presented large or small samples of of own-gender enhancing or other-gender enhancing observations. Children read arguments based on the observations, rated argument intelligence, judged the number of other children to whom the observations could he generalized, and provided verbal justifications for their judgments. Own-gender reasoning biases declined with age: these declines were, however, partially accounted for by declines in the strength of self-reported gender affiliations. Reasoning biases-demonstrated by problem-to-problem shifts in reasoning quality-were constrained by sample size, indicating a modest degree of rationality even among 7-year-olds. Specifically, biases co-existed with reasonably limited generalizations from small samples of own-gender evidence and with reasonably extensive generalizations front large;samples of other-gender evidence. Children were thus able to satisfy motivations for own-gender favoritism and reason in accord with the law of large numbers. Several explanations of the findings-based on changes in the salience of gender, multiple classification skills, and the ability to reason independently from beliefs-are offered.