To determine the magnitude of Plasmodium vivax relapsing malaria in rural Amazonia, we carried out a study in four sites in northeastern Peru. Polymerase chain reaction-restriction fragment length polymorphism of PvMSP-3α and tandem repeat (TR) markers were compared for their ability to distinguish relapse versus reinfection. Of 1,507 subjects with P. vivax malaria, 354 developed > 1 episode during the study; 97 of 354 (27.5%) were defined as relapse using Pvmsp-3α alone. The addition of TR polymorphism analysis significantly reduced the number of definitively defined relapses to 26 of 354 (7.4%) (P < 0.05). Multivariate logistic regression modeling showed that the probability of having > 1 infection was associated with the following: subjects in Mazan (odds ratio [OR] = 2.56; 95% confidence interval [CI] 1.87, 3.51), 15-44 years of age (OR = 1.49; 95% CI 1.03, 2.15), traveling for job purposes (OR = 1.45; 95%CI 1.03, 2.06), and travel within past month (OR = 1.46; 95% CI 1.0, 2.14). The high discriminatory capacity of the molecular tools shown here is useful for understanding the micro-geography of malaria transmission.