Self-reports are commonly used to ascertain malaria prevalence in epidemiological studies conducted in settings where laboratory diagnosis is impractical. Most studies, however, do not use self-report per se, but indirect report, where one respondent provides responses for all household members. Studies also vary in terms of the time frame used for this ascertainment. The aim of our research was to determine the validity of self-report and indirect report in ascertaining malaria prevalence over six, eighteen and thirty-month time periods. Reports of malaria episodes collected through interviewer-administered questionnaires (193 self-reports, 614 indirect reports) were compared to microscopy-confirmed cases (principally Plasmodium vivax) registered at a government-run health post in the Peruvian Amazon. Test parameters were estimated using a Bayesian latent class approach for imperfect gold standards. Logistic regression analyses were performed to explore determinants associated with accurate responses. Malaria self-report for the thirty-month period prevalence had the highest sensitivity (91.0%). Specificity was maximized when malaria prevalence was measured over the last six months for both self-report (91.6%) and indirect report (96.7%). Accuracy was highest for the six-month period prevalence in both self-report (91.3%) and indirect report (96.4%). Respondents who were female, had more education, or who provided a report on behalf of a child ≤12 years of age, were generally more accurate. Both self-report and indirect report provides accurate estimates of malaria prevalence, especially over shorter periods of time. The choice between self-report and indirect report should ultimately depend on the research question and the target study population. © 2010 Elsevier Ltd.