When samples of psychiatric patients are assessed on 2 occasions on state and trait measures, without any formal intervention having yet taken place, a mean change in scores towards less psychopathology is often observed. This re-test effect, which is a potential threat to the validity of longitudinal studies, has not been given serious attention by researchers using repeated administrations. The present study addressed this issue by examining re-test effects in 2 independent clinical samples with the Symptom Check List-90-Revised and other widely-used state and trait measures. Time intervals between intake and re-take were 3 months and 11-350 days (M=93 days, median=81 days). Results indicated that the re-test effect occurred for most of the measures. Where significant testing effects were observed, 54-72% of the patients had time 2 scores that were lower than the average time 1 scores. The importance and practical and scientific implications of the findings are discussed. Eleven hypotheses concerning the nature of the re-test effect are offered, including mood-congruent associative processing, natural coping mechanisms, self-monitoring hypothesis and response-shift. (C) 2001 Elsevier Science Ltd. All rights reserved.