EXPERIMENTER CALCULATION ERRORS: A POTENTIAL FACTOR AFFECTING INTERPRETATION OF RESULTS1

Abstract

Following a study of socially withdrawn children, it was noted that checks of experimenter calculations indicated discrepancies. The present study investigated the magnitude and direction of experimenter errors in data computations. Two full‐time research assistants, who served as experimenters, processed the behavioral observation data. These experimenters were referred to as the subjects in this study. Standard data values, against which the subjects' computations were compared, were established by two other research assistants and a computer staff. The results indicated that experimenters' original data values disagreed with the standard data values. The daily errors showed considerable variability for both experimenters. The specific factors accounting for this variability could not be identified, although frequency of experimenter responses required to obtain rate and duration data, time spent calculating, and complexity of task were suggested as possible contributing variables. While disagreements occurred across all experimental days, the magnitude of differences was generally small. An examination of direction of differences in both experimenters' data showed no systematic trends in favor of the experimental hypothesis. Rather, there appeared to be a trend in the other direction. This study suggests that the same research decisions would have been made during the original investigation using either set of data.