Examining the influence of interval and observation length on the dependability of data

Direct observation is a tool commonly employed by school psychologists to investigate student behavior, and because these data are used for educational decision-making, accuracy is an important consideration. Procedural aspects of systematic direct observation have been shown to influence reliability, and this study was designed to explore how interval length and observation length influence the dependability of academic engagement data when using a momentary time sampling procedure. Twenty 7th grade students were each observed for two 30-minute sessions during math instruction. A series of generalizability (G) studies were conducted using combinations of common interval and observation lengths to examine how these factors influenced reliability-like coefficient. In general, shorter interval lengths and observation durations produced higher levels of dependability. Implications for practice and research, as well as limitations of the current study, are discussed.