This study manipulated the timing and control of error feedback in problem solving and examined their effects on skill acquisition by 40 undergraduate students learning to program in the computer language Lisp under four error feedback conditions. These four conditions included two types of symbol-by-symbol feedback that vary in content, a condition in which feedback is presented only upon request as the student works, and a condition in which no feedback is available until the student has completed an exercise. Questions of interest in the study were whether: (1) the feedback manipulation affects posttest performance and the timecourse of learning; (2) the feedback manipulation affects students' perceptions of how well they have learned the material; and (3) students have preferences among feedback conditions. Results indicated that students who received feedback learned better than students in the no feedback condition, and that students in the two immediate feedback conditions finished the exercises most quickly. (8 references) (DB)