CONTEXT: Measurement of pregnancy intentions typically relies on retrospective reporting, an approach that may misrepresent the extent of unintended pregnancy. However, the degree of possible misreporting is unclear, as little research has compared prospective and retrospective reports of intention for the same pregnancies. METHODS: Longitudinal data collected between 2010 and 2015 on 174 pregnancies were used to analyze the magnitude and direction of changes in intendedness (intended, ambivalent or unintended) between prospective and retrospective measurements of intendedness using versions of the London Measure of Unplanned Pregnancy (LMUP). Changes were assessed both continuously and categorically. Differences in the degree of change-by pregnancy outcome and participant characteristics-were examined using mixed-effects linear and logistic regression models. RESULTS: Over two and one-half years of follow-up, 143 participants reported 174 pregnancies. Approximately half showed changes in intention between the prospective and retrospective assessments, with 38% of participants reporting increased intendedness and 10% decreased intendedness. Reported intendedness increased more among those who gave birth (mean change in continuous LMUP score, 2.2) than among those who obtained an abortion (0.7), as well as among individuals with a college degree (4.1) than among those with a high school diploma (1.2). Participants who reported recent depression or anxiety symptoms showed more stable intentions (0.02) than those who did not (2.1). CONCLUSIONS: Retrospective measurement of pregnancy intentions may underestimate the frequency of unintended pregnancy, with such underestimation being greater among certain subgroups. Estimates based on retrospective reports thus may produce inaccurate impressions of intentionality. Further efforts to refine the measurement of pregnancy preferences are needed.