Performing retrieval practice of previously studied material can enhance subsequent learning of new material. This potentiating effect of interpolated testing has been extensively investigated in multi-list learning paradigms. However, extant studies in this literature have manipulated interpolated testing in a binary manner, such that participants either took a test after every list or they are not tested until the final list. In reality, when students prepare for an exam, they are unlikely to test themselves consistently at regular intervals. Therefore, a critical question is whether interpolated testing would potentiate new learning when interpolated testing stops but encoding continues. In four experiments, we compared memory performance across participants who received an interpolated test after every list, received no interpolated tests, or received interpolated tests during only the early lists. Results showed that retrieval enhanced new learning even when testing was terminated early, although the benefits were weakened substantially. Further, participants’ test expectancy ratings and semantic clustering scores provide support for the idea that retrieval promotes new learning by independently enhancing attentional and organizational processes.