Gender constancy judgments in children referred for problems in their gender identity development (N = 206) and controls (N = 95) were compared. On Slaby and Frey's (1975) gender constancy interview, the gender-referred children performed more poorly than the controls at three stage levels: gender identity, gender stability, and gender consistency. On the Boy-Girl Identity Task, a second measure of gender constancy (Emmerich et al., 1977), the gender-referred children also performed more poorly. Gender-referred children who had not attained gender consistency engaged in significantly less same-sex-typed play on a free-play task than the gender-referred children who had, but there were no gender consistency effects for the controls. Two other measures of sex-typed behavior were unrelated to gender consistency. In the gender-referred group alone, children who "failed" the gender identity or gender stability stages were more likely to draw an opposite-sex person first on the Draw-a-Person test and to evince more affective gender confusion on the Gender Identity Interview (Zucker et al., 1993) than children who had "passed." It is concluded that children referred for problems in their gender identity development have a developmental lag in gender constancy acquisition. Possible reasons for the lag are discussed.