Classroom engagement is important to young children's academic and social development. Accurate methods of capturing this behavior are needed to inform and evaluate intervention efforts. This study compared the accuracy of interval durations (i.e., 5 s, 10 s, 15 s, 20 s, 30 s, and 60 s) of momentary time sampling (MTS) in approximating the duration of classroom engagement as measured by continuous duration recording (CDR). Twenty-four sessions of children in large-group instruction were observed using the "Behavioral Observation of Students in School for Early Education" (BOSS-EE). In general, shorter intervals produced engagement estimates that more highly correlated with CDR data and had less measurement error relative to estimates from longer intervals. Findings suggest that selection of MTS interval duration may affect the accuracy of estimates for children with low levels of engagement to a greater extent relative to children with high levels of engagement. Implications and future directions are discussed.