Three experiments investigated the ways memories change over repeated tests. A tradition of previous work started by Ballard (1913) revealed net improvement (or hypermnesia) over tests, while Bartlett (1932) showed intertest forgetting. Differences between the paradigms include the type of material learned (lists of items vs. prose) and the delay interval between recall tests (several days for Bartlett and up to one day for Ballard). Both factors were examined in three experiments. In all experiments, hypermnesia was obtained across repeated memory tests when tests occurred within a single, experimental session. However, intertest forgetting resulted after a one-week delay, replicating Bartlett. Performance on repeated tests was discovered to be a function of the number of prior tests and the delay between tests, while response criterion (lenient or strict) and the nature of the learned material (pictures or prose passages) were not important.