Second-grade children were given general training in monitoring the utility of strategies, the affective consequences of strategy use, or no strategy-monitoring training. They then performed an associative learning task, first without strategy instructions and then with instructions to use either an effective or ineffective strategy. All training conditions produced short-term maintenance of the effective strategy, but only the strategy-utility training resulted in long-term maintenance. Subjects given strategy-utility training abandoned the ineffective strategy at a higher rate than children given strategy-affect or no training. Responses to metamemory questions indicated that only in the strategy-utility condition was strategy efficacy a prime consideration in strategy-use decisions. This experimental evidence bolsters the case for including monitoring instruction in multicomponent training packages aimed at producing durable strategy use. (Author)