In most large‐scale programs of tests that aid in making high‐stakes decisions, such as the TOEIC® family of products and service, it is not unusual for a significant portion of test takers to retake the test at multiple times. The study reported here used multilevel growth modeling to explore the score change patterns of nearly 20,000 TOEIC Listening and Reading test takers who repeated the test six times during a 4‐year period. The study revealed that (a) on average, repeaters' scores increased with each subsequent testing; (b) repeaters' score increases were larger for initial retests than for later ones; (c) test takers' educational backgrounds were related to their initial scores but not to their score increases; and (d) test takers' gender was related both to initial scores and to score increases. The results suggest that multilevel growth modeling analysis has potential for evaluating and monitoring test performance across administrations by exploring repeaters' score change patterns over time. The study also provided empirical evidence for the reliability and validity of TOEIC scores.