This paper discusses several critical issues concerning feedback in instruction and proposes a text synthesis research paradigm to investigate these variables. The paper is based on the assumption that feedback, a significant adjunct aid in instruction, is currently by and large ineffectual except in some simple instructional contexts. Section 1 deals with general educational issues, including instructional design, behavioral objectives, learning situations, cognitive processing, recall strategies, and memory. It is argued that behavioral management systems have followed Skinner's dictum of small controlled steps to reduce error, and, as a consequence, feedback has been dictated by the assumption that the strategy minimizes significant individual transformations of meaning. Section 2 presents a review of feedback literature, and recent cognitive findings on feedback are discussed. The third section reviews literature relevant to research paradigms. Text analysis is discussed along with the relevance of specific findings to the proposed research model. The paper's fourth section discusses the proposed research; two experiments are desribed. (CJ)