The effects of observational session length, method of recording, and frequency of teacher behavior on reliability and accuracy of observational data

Abstract

An increased concern for the accuracy of observational data has stimulated research assessing the effects of various parameters on the accuracy and reliability of observational data. Some research has investigated the effects of time sampling procedures on the accuracy of obtained data (Powell, Martindale, & Kulp, 1975; Powell, Martindale, Kulp, Martindale, & Bauman, 1977). This research indicated that different methods of scoring behavior in various lengths of intervals produced discrepancies in the obtained rates. The partial interval method of scoring behavior was found to overestimate the rate of behavior when compared with a continuous measure while the whole interval method underestimated behavior across interval length (Powell et al., 1977). The implications for observational data collection were that experimental effects could possibly be masked by increased levels of measurement error. Reliability and accuracy are two related but possibly separate issues (Kazdin, 1977). In addition to the necessity for empirical research on the effects of different features of observational instruments on accuracy of data, observer reliability should be investigated as a concurrent issue. The possibility of differential effects on the observers' ability to use the observation code efficiently would have important implications for the collection of behavioral data. Additionally, the relationship between accuracy of data and observer reliability could be more closely explored.