An inaccurate-timing filter for reaction time measurement by JAVA applets implementing Internet-based experiments.

JAVA applets are not accurate enough to measure reaction time unless precautions are taken. A response-filtering technique is introduced that identifies inaccurate measurements by testing whether the client-side JAVA implementation produces false measurements of a time interval of known length during the measured reaction time. Reaction times that indicate such inaccuracy are discarded to increase the reliability of the remaining data. Three studies provide illustrative data on the effectiveness of the new technique, using 11 different computer systems. Study 1 showed differences in the accuracy of different clients and higher variability of the obtained means with applets than with native programs. Study 2 showed sharply increased variability when additional loads were imposed on the clients during reaction time measurement. The response-filtering technique diminishes this variability. Study 3 indicated more accurate measurement, with the new technique being able to reveal even smaller differences in difficult technical conditions.