The importance of encoding strategies for associative recall is well established, but there have been no studies of aging and intraindividual variability (IAV) in strategy use during extended practice. We observed strategy use and cued-recall test performance over 101 days of practice in 101 younger adults (M = 25.6 years) and 103 older adults (M = 71.3 years) sandwiched by a pretest and posttest battery including an associative recall test. Each practice session included 2 lists of 12 number-noun paired-associate (PA) items (e.g., 23-DOGS), presented for brief exposures titrated to maintain below-ceiling performance throughout practice. Participants reported strategy use (e.g., rote repetition, imagery) after each test. Substantial IAV in strategy use was detected that was coupled with performance; lists studied with normatively effective strategies (e.g., imagery) generated higher PA recall than lists studied with less effective strategies (e.g., rote repetition). In comparison to younger adults, older adults’ practice (a) relied more on repetition and less on effective strategies, (b) showed lower levels of IAV in effective strategy use, and (c) had lower within-person strategy–recall coupling, especially late in practice. Individual differences in pretest-posttest gains in PA recall were predicted by average level of effective strategy use in young adults but by strategy–recall coupling in older adults. Results are consistent with the hypothesis that experiencing variability in strategic outcomes during practice helps hone the effectiveness of strategic encoding behavior, and that older adults’ reduced degree of pretest-posttest gains is influenced by lower likelihood of using and optimizing effective strategies through practice.