Several large-scale epidemiological surveys have reported increasing lifetime rates of psychopathology among recently born cohorts. In the case of Major Depressive Disorder (MDD) younger cohorts tend to manifest higher lifetime prevalences of the condition than older cohorts, at any given age. In some studies, cohort differences are so large that the youngest cohort exceeds the lifetime prevalence of the oldest cohort well before passing through their total period of risk. The data on lifetime prevalences that support the existence of secular differences, however, has typically been collected in cross-sectional studies. Thus, individuals are interviewed at a single point in time and asked to recall all prior psychopathology. Due to poor recall, this design may greatly underestimate earlier experiences. In fact, cohort differences are not limited to MDD; similar results have been reported for a wide range of psychopathology, suggesting methodological problems at data collection. We have conducted a simulation study to examine the magnitude of annual rates of forgetting that could produce the secular trends reported for MDD. Small, but constant annual rates produce striking, 'cohort effect-like' curves. The rates needed to reconstruct the reported effect are compatible with published values for test-retest studies of lifetime recall of MDD. This simulation study does not rule out the existence of a cohort effect in some psychiatric disorders, but stresses the possible limitations of using cross-sectional studies to investigate secular trends.