The available evidence suggests that the GHQ is a valid case-finding and screening instrument and a reasonable measure of severity of functional non-psychotic psychopathology. However, recently Henderson et al. (1981) have demonstrated a substantial retest effect for the GHQ, which may affect its usefulness in longitudinal population and outcome studies. In a three-wave longitudinal study among 175 new psychiatric outpatients, we examined whether retest effects can also be found in treatment settings and what its likely causes might be. A substantial retest effect was found, and the social desirability and legitimation hypotheses appeared to offer the most likely explanations. The implications of the findings for longitudinal and treatment evaluation studies are discussed.