<p>Behavioral researchers typically measure academic engagement by observing “on-task” behavior, or what students appear to be doing (e.g., they have a pencil in their hand, are oriented toward work materials). Sometimes, increasing time spent on task is an intervention goal, but on-task behavior is often used as a feasible correlate for more nuanced outcomes (e.g., academic productivity, attending to instruction). The scant literature on this topic suggests that correspondence between observed on-task behavior and other metrics of academic engagement may be highly variable and sensitive to environmental variables and individual differences. Thus, the purpose of the current study was to examine correspondence between observed on-task behavior and product measures of academic productivity (i.e., work completion and accuracy of math problems) under various conditions. Six children aged 8–13, some diagnosed with disabilities, experienced five conditions presented in a multielement format in a laboratory setting: reinforcement for observed on-task behavior, work completion, and work accuracy, plus no reinforcement and noncontingent reinforcement control conditions. We saw strong measure correspondence across conditions for three participants, but the other three had at least one condition in which we saw moderate to high levels of on-task behavior and no work completion. These findings indicate a need for additional research on correspondence between measures of academic engagement.</p>