Previous research suggests that when participants engage in generative study activities, the processing of text is enhanced and improvements in memory and metacomprehension result. However, few studies have investigated the influence of processes required by the testing situation or the interaction between encoding and retrieval processes on metacomprehension accuracy. The present experiments examine whether the congruency of processes generated during study and required at retrieval affect memory, metacomprehension, and control processes. Study orientation and test type were congruent (i.e., letter-reinsertion: detailed test), incongruent (i.e., letter-reinsertion: conceptual test), or neutral (i.e., read: conceptual test). After generative study, but before testing, participants made metacomprehension predictions for previously studied texts. Controlled strategy selection was measured in Experiment 2. When processes at study and test were congruent, cued recall performance and metacomprehension predictions were more accurate than when study and test were incongruent. For incongruent conditions, metacomprehension predictions were no better than chance; thus, controlled strategy selection was based on inaccurate metacomprehension, thereby further penalizing memory performance relative to congruent conditions. These findings extend a transfer-appropriate processing framework to metacomprehension.