Little is known about temporal patterns of physical intimate partner violence (PIPV) among South African adolescent girls. We sought to identify and describe PIPV risk trajectories and related correlates in this population. METHODS: Our analytical cohort came from the HPTN 068 Cash Transfer Trial in Mpumalanga Province, South Africa. Cohort members were eighth and ninth graders (median age 14 years) who enrolled in 2011 and had three to four annual, self-reported PIPV measurements. We used group-based trajectory models to identify groups of girls with similar longitudinal patterns of PIPV risk over 4 years and potential correlates of group membership. RESULTS: We identified two trajectory groups (n = 907): a higher-risk group (~52.8% of the cohort) with predicted PIPV probabilities of 13.5%-41.1% over time and a lower-risk group (~47.2% of the cohort) with predicted probabilities of 2.3%-10.3%. Baseline correlates of higher-risk group membership were ever having had sex (adjusted odds ratio [aOR]: 4.42, 95% confidence interval [CI]: 1.56-12.57), borrowing money (aOR: 1.95, 95% CI: 1.01-3.79), and older age (aOR per 1-year increase: 1.39, 95% CI: 1.11-1.73), while being in the 068 intervention arm (aOR: .29, 95% CI: .17-.51) and supporting more gender-equitable norms (aOR per 1-unit score increase: .89, 95% CI: .81-.97) were inversely associated. CONCLUSIONS: A high proportion of adolescent girls experience sustained PIPV risk in rural South Africa, suggesting a need for interventions in late primary school that encourage gender-equitable norms, healthy relationships, and safe ways to earn income during adolescence.