Massed Practice: Does It Change the Statistical Properties of Performance Tests

摘要

Repeated trials on a task are frequently required for assessing training procedures or experimental treatments. Limited time, money, or availability of research subjects often result in the need to give a substantial number of trials on a task within a short period of time. However, in many laboratories repeated measures are traditionally separated by 24 hours or more to reduce the chances of fatigue, interference, or other factors introducing undesirable error variance. Massing practice is an obvious alternative to distributing it, particularly when time constraints exist. However, massed practice is only a desirable alternative if the resulting test scores maintain the statistical properties required for repeated measures analysis. Paper-and-pencil and computerized versions of traditional human performance tests were examined under massed practice conditions. Many of the tests had been shown to have high reliabilities and to meet the statistical requirements for repeated measures applications under distributed practice conditions in earlier studies at our laboratory. It is recommended that distributed practice with trials separated by 24 hours or more be used whenever feasible. If massed practice is required tasks should be chosen which have been shown to have high reliability and which meet the statistical requirements for repeated measures experimentation. It is expected that once computer tasks are refined they too will lend themselves to massed practice administration when required.