instruction, teaching the development of frames for PI, in varying levels of student interaction with the materials. All the groups showed significant improvement from pretest to posttest. The groups performed progressively better as the level of student interaction increased. The result being that this student interaction, be it in the form of overt or covert answer construction, resulted in a 13% better performance on a fill-in-the-blank posttest, and showed a better grasp of the concepts when later applied to constructing PI frames. Tudor pointed out that the differences were comparatively larger than in previous programmed instruction research and may have educational importance. The question raised addresses the functional significance of the behaviors that an instructional program is designed to produce. “Can teachers design frames that actually change behavior? In other words, can students use a washing machine correctly after completing a program?” Tudor recognized a need for future studies to identify “behavior change produced by interactive instruction.” A significantly smaller sample participated in Tudor's (1995) study to isolate the effects of active responding in computer-based instruction. The four students in this experiment worked through a set of programmed instruction that alternated between frames with blanks that required overt answer construction and all-inclusive frames without blanks. Each student produced a higher percentage on posttest questions that corresponded to program segments that called for construction of overt answers. Regardless of the small sample, this study does confirm the importance of active responding in the effectiveness of instructional programs.
The “constructed-response contingency” could be associated to the “generation effect” studied in depth by Rabinowitz and Craik (1986). The generation effect suggests that verbal material that is actively generated (such as the overtly constructed-response) during the presentation of lesson material is later recalled more readily than material that is simply read. Study participants either read or generated target words in the existence of particular “generation” cues. The recall of the