Introduction: For decades birth order and gender differences have attracted research attention. Method: Birth order, family size and gender, and the relationship with arithmetic achievement is studied among 1152 elementary school children (540 girls, 612 boys) in Flanders. Children were matched on socioeconomic status of the parents and administered a test on mental arithmetic skills and on number knowledge. Results: Our findings tend to favor boys and suggest that it is possible that only children and last-born children in large families are less competent in arithmetic than first born children. In contrast with the prediction of Zajonc, in families with three and five or more children middle children did best. Moreover, our dataset was not compatible with the Admixture hypothesis. Discussion: The study was partially in line with the Confluence and the Resource Dilution hypothesis. The birth order differences might be explained by the advantage of tutoring younger children by first born or middle children, helping them to process information. However, our data provides evidence that birth order accounts for only two or three percent of the variance in arithmetic scores. Analyses of our dataset demonstrate that last borns have poorer scores in small (family size 2) and large (family size 5 or more) families. However, the last-born disadvantage and first born advantage was slight is not present in families with three or four children. (Contains 6 tables.)