According to the time on-task hypothesis, the amount of time an individual devotes to an instructional task determines the extent to which learning occurs. Therefore, time off-task hampers learning by limiting learning opportunities. Prior research has generally found a positive relationship between time on-task and achievement; however, the correlation strength is highly variable across studies. Differences in the ways in which time has been operationalised may be one factor contributing to the divergent results. We utilise an existing data set of twenty classrooms (K-4) to investigate whether operationalising time in a consistent manner will yield a stable association between on-task behaviour and learning. Overall, on-task behaviour was positively correlated with learning, controlling for gender, school type, and grade-level. However, this correlation was weak. Importantly, considerable variability in the correlation strength was observed, indicating variations in the prior literature cannot be attributed solely to issues of measurement.