It has been argued that patrilineal joint family systems tend to bias family planning decisions in favour of sons. A simple model suggests that in such societies, any given son will be more highly valued by his parents (1) the fewer his brothers and (2) the earlier his birth is in the brother series. A daughter's value will be greater (1) the fewer brothers she has and (2) the earlier her birth is relative to other sisters. This study first addresses the extent of son preference as inferred from family composition data for 772 Taiwanese first-graders born in the mid-1970s in two socioeconomically distinct communities in Taipei, Taiwan. It then uses linear regression to consider whether the model criteria help account for statural variation among children in each study area when controlling for differences in measurement age, parental education and housing. With respect to family composition and gender preference, available evidence was consistent with previous surveys. While better-educated parents in the more affluent study area had significantly fewer children (p<0·0005) and were more willing to stop without a son, girls there, as in the less affluent area, were still significantly more likely than boys to belong to large sibships (p≤0·005). Evidence from mean height of males and females partially accords with hypothetical predictions. In the less affluent area, the interaction effect of male birth order and the presence of younger siblings was significantly associated with mean stature (p=0·002). Males without brothers were 2·0 cm taller than males with either an older or a younger brother (116·3 ± 0·5 cm vs 114·3 ± 0·4 cm). Males who had both younger and older brothers, but often no sisters, were about as tall, however, as those without brothers. A similar, but less pronounced, pattern was found amoung males in the more affluent area, but only among those who had sisters. These boys were also consistently shorter than boys without sisters (115·6 ± 0·6 cm vs 117·7 ± 0·6 cm; p=0·001). Patterns of mean female stature did not clearly support the hypothesis. Girls in the more affluent area were relatively tall and did not show significant variation. Results among less affluent girls showed significant contrasts, but not necessarily in the predicted direction.