The effectiveness of three basic approaches to sport skills analysis instruction were examined. Ninety-two physical education majors were allocated to four groups and randomly assigned to one of four instructional frameworks: Observational Model (OM), Common Errors (CE), Correct Only (CO), and Control (C). It was hypothesized that the OM group would perform better than the CE, CO, and Control groups on the two analytical tasks of interest: visual retention of overarm throw performance, and knowledge of correct overarm throw technique. It was also anticipated that the CE and CO subjects would perform better on analytical tasks than the control subjects as the result of training. The OM group performance exceeded all other groups on the perceptual task as hypothesized. A majority of the test item response inaccuracies on the part of both the CE and CO subjects were due to a failure to see movement phenomenon that actually occurred in the performances. Following a description of the procedures of the study, these results are analyzed. Copies of the instruments used in the study are appended. (JD)