Content area
Abstract
Behavioral observation is an invaluable assessment strategy that is frequently employed in classrooms and other applied settings. To insure that data obtained via direct observation is appropriate and useful, various recording methods must be carefully compared and evaluated in terms of their accuracy. A great body of research already exists on issues related to the reliability of behavioral observation strategies; however, researchers in this area have focused on accuracy as an even more important consideration when evaluating observational assessment methods. In this study, several methods of observing behavior via time-sampling were compared to determine their relative accuracy in measuring behavior. These time-sampling procedures included momentary time-sampling, partial-interval observing and recording, whole-interval observing and recording, and a newly proposed modified interval sampling procedure (Powell, 1982). Computer simulations producing precise and varying conditions of behavior were generated to serve as an extensive data base for this study. The results of this study combined inferential statistics with a descriptive level of analysis to compare the accuracy of the data produced by each of the four observational systems. These analyses and comparisons were made both overall and within the varying dimensions of frequency and duration. In general, the modified interval procedure produced the most consistently accurate data both when estimating behavioral frequency and duration.
