A large number of studies in the last decade (e.g. Roe, Case, and Roe, 1962; Payne, Krathwohl, and Gordon, 1967; Niedermeyer, Brown, and Sulzen, 1969) attempted to assess the effect that order of information presentation has on learning. The usual method of these studies has been to present to one group of subjects some unit of programmed instruction in its logical order, while another group studied the same material but with the frames presented in random or "scrambled" order. The most significant finding of these studies was that scrambling did not adversely affect terminal performance. The only case in which scrambling has had that effect, to the authors' knowledge, was in a study by Brown (1970). The result has annoyed and puzzled researchers, because it implies that our intuitions about cumulative learning processes, and some of the more important learning theories, are seriously in error. The failure of scrambling to significantly reduce performance seems to imply that human learners can usually reconstruct their own knowledge structures, given sufficient time and information to do so. It is then worthwhile to speculate about the nature of this reconstruction process. Its most obvious feature is a strong dependence on memory while in the process of learning. If a learner studies any type of sequential knowledge structure, he cannot comprehend later elements if earlier ones are forgotten or not learned. If the presentation order is different from the logical order, as in the scrambled version of a program, the dependence on memory should be greatly increased, since a certain amount of meaningless information would have to be retained before restructuring could occur. Should the learner forget one important item before it can be used, the entire process would terminate. Thus, it would be predicted that scrambling would interfere