This project attempted to replicate a 1981 study by Munro, Fehling, Blaise, and Towne, which found that instructional feedback for dynamic skill training is most effective when the student chooses when and if feedback is to be received. Using a simpler task, this study focused on an Air Intercept Controller Simulation program. The effectiveness of computer-generated voice output in instruction and simulation in dynamic skill training were also addressed in an experiment in computer-based instruction. Both the intrusiveness and delivery mode (text-voice) factors had statistically significant effects on student errors. Performance was superior in the groups which received feedback in a textual mode and had control over sequence and presence of feedback. The next highest performance group received feedback in a computer voice mode and also had control over feedback. The third highest group received immediate textual-mode feedback to errors, while the poorest performance group received immediate computer voice feedback. Results suggest that instruction in dynamic skills should be non-intrusive, and that current inexpensive voice synthesis technology is inappropriate for dynamic skill training. Six data tables and six figures are included, and a list of crucial and non-crucial errors is appended. (Author/LMM)