Observer Accuracy under Two Different Methods of Data Collection: The Effect of Behavior Frequency and Predictability

Abstract

The present study compared the accuracy of observers employing momentary time sampling vs. continuous data collection systems when scoring behaviors that differed along the dimensions of frequency and predictability. Twelve trained observers were randomly assigned to one of the two methods of data collection when scoring a series of four videotapes of teacher-student interaction. The results indicated that there was no overall difference in observer accuracy between the two methods of data collection. Interesting disordinal interactions, however, were identified. Continuous data collection significantly enhanced observer accuracy when scoring high frequency behaviors and behaviors with high transitional probabilities to other behaviors. Continuous data collection, however, impeded accuracy when behaviors displayed low levels of predictability (low transitional probabilities). Momentary time sampling produced the most stable levels of observer accuracy. Some implications of these findings for the selection of recording procedures and observer training are presented.