6: Language in Texts

ion, generality and inclusiveness than the latter passage itself" (Ausubel, 1960, p. 268). 3. A cognitive theory as to the organizing process: "New [potentially] mean­ ingful material becomes incorporated into the cognitive structure insofar as it is subsumable under relevant existing concepts'' (Ausubel, 1960, p. 267). 4. An unstated assumption about the reading process; or at any rate an assumption about the way reading affects cognition. The criteria of typog­ raphic priority (that is, potential organizing material comes first) is not strictly entailed by AusubeΓs cognitive theory; it assumes a once-readthrough model, which is not consistent with what is now known about quasi-naturalistic reading (Augstein, 1971; Pugh, 1974; Thomas, Augstein, Deans & Moore, 1970; Waller, 1977a, 1977b). All of these components seem to be either incorrect or vague or both vague and incorrect; this is the main reason for the apparently conflicting results of empirical studies. However, Ausubel deserves admiration for two correct and important decisions: His theory is a cognitive theory, albeit limited by the primitive state of cognitive psychology 20 years ago, and his experiments were conducted with realistic learning material. In both these respects Ausubel was ahead of his time. For all that, his framework is seriously flawed. Most importantly, neither the structure of subject matter nor the process of cognition can be adequately modeled as a hierarchy. Modern theories of cognition vary in detail and in language, but they do agree that a heterarchy is the basic logical structure of cognition. The basic fact is that while on a certain occasion for a particular purpose one procedure may control another (here a concept is considered to be a procedure), on another occasion for another purpose the previously superordinate procedure may be called upon as a lower order subroutine. Thus an appearance of hierarchy results from confusing the set of behaviors appropriate to a particular situation with the structure which generated those behaviors. The same con­ sideration applies to the structure of subject matter (which is, in a special sense, independent of particular cognitions—see Popper, 1972). For recent work on cognitive structure see Johnson-Laird and Wason (1977) and Pask (1975a, 1975b, 1976). From the inadequate models of cognition and the subject matter in Ausubel follows a vague and shallow account of the act of mastering new but related information (i.e., subsumption). Again, the great stimulus given to cognitive theory by artificial intelligence (Boden, 1977) should enable us to reformulate this problem in a more fruitful and more correct manner. Another flaw in AusubeΓs framework is that there is no particular cognitive reason for the typographic primacy of organizers, unless the structure and function of the potential organizers are so tightly defined (by semantic and linguistic methods) as to determine its priority. Otherwise the initial position of the organizer is 254 Review of Research in Education, 6 conflated with an unjustified assumption that effective reading consists of starting at the top, reading everything once, and finishing at the end. Everything we know about effective readers suggests that their strategies are far more sophisticated than this. However, as my colleague Waller points out, it may make sense to put organizers in a consistent place so they can be accessed easily. This is a different reason from the one given by Ausubel and may be more convincing. The accumulated effect of all these problems is devastating (but note AusubeΓs recent defense, 1978). To do Ausubel justice, he does notice and acknowledge many of the consequent subproblems, for example: The pedagogic value of advance organizers obviously depends in part upon how well organized the learning material itself is. If it contains built-in organizers . . . much of the potential benefit derivable from advance organizers will not be actualised (Ausubel, 1963, p. 82). Thus if an organizer can first delineate clearly, precisely, and explicitly the principal similarities and differences between the ideas in a new learning passage, on the one hand, and existing related concepts in cognitive structure on the other, [there would be] fewer ambiguities . . . and fewer misconceptions suggested by the learner's prior knowledge of the related concepts (Ausubel, 1963, p. 83). This is the problem in a nutshell. Although empirical research has done little to solve these problems, there have been great conceptual advances in cognitive psychology during this time. Surely an attempt should be made to reformulate AusubeΓs notions in a modern conceptual framework. The rewards for even partial success would be considerable, for the problems Ausubel addresses are right at the heart of pedagogy. Adjunct Questions and Mathemagenics Work on questions in texts (usually called "adjunct" or "in-text" questions) has been steadily progressing for the past 15 years. This research was also started in the days of programmed learning, though gradually the behaviorist framework has been shed, and the problems now addressed by researchers have to do with the effects of questions on cognitive processing. There have been several excel­ lent reviews of this work (the most recent being Rickards & Denner, 1978), so again I will discuss the research mainly from the viewpoint of curriculum development. The notion of adjunct questions is difficult to separate from an important concept of Rothkopfs, namely, mαthemαgenic behavior. In his early papers Rothkopf conceived of this idea in extremely broad terms. Any activities of the subject in the learning situation are called mathemagenic activities; they might include both behaviors relevant to learning and nonrelevant behaviors: Mathemagenic activities refer to the activities of the student in the instructional situa­ tion . . . such activities as reading, asking questions . . . and mentally reviewing a reMacdonald-Ross: Language ¡n Texts 255 cently seen motion picture. Mathemagenic activities also include looking out of the classroom window, yawning, turning the pages of a textbook without reading . . . and sleeping, either in class or in a library carrel. (Rothkopf, 1968, p. 116) Mathemagenic behaviors include gross postural adjustment of the head and body toward the printed page and the movement of the eyes over the page. It is assumed that there are other mathemagenic activities that cannot be observed directly and which must be inferred. These inferred activities are in many respects more interesting than the directly observable aspects of mathemagenic behavior. (Rothkopf, 1965, p. 199) These are extremely broad definitions, but in practice the operational definition used by Rothkopf in his experiments was comparatively strict; it depended upon an important distinction between the direct effect of questions and their indirect effect upon unquestioned material: Basically, the adjunct question experiments at the Bell Telephone Laboratories were incidental learning studies. At the simplest level they involved measurement of performance changes on reading material that was not directly related to the text components on which the experimental questions were based. Hence any measured behavioral changes could not be attributed to the direct-instructive-effects of questions' content. The changes were ascribed instead to changes in presumed mathemagenic activity during inspection of the text. (Rothkopf, 1970, p. 333) Early results (summarized in Rothkopf, 1968) suggested a small but definite mathemagenic effect for questions placed after relevant text segments. This effect was much smaller than the direct instructive effect, and a difference of opinion arose as to whether these results were sufficient to permit the conclusions Rothkopf drew from them (Carver, 1972; Ladas, 1973; Rothkopf, 1974). As guides to curriculum development, these experiments'have some weaknesses. The size of the mathemagenic effect was of marginal statistical significance, whereas the curriculum developer looks for large effects which are robust enough to withstand the extraordinary variety of field learning situations. Certainly these results do not lend weight to Rothkopf's opinion that mathemagenic variables were more significant than structural variables, for, as discussed above, we know that structural variables can produce quite striking effects. A second, and perhaps more decisive, query concerns the nature of the text material and the nature of the adjunct questions. The material had a "relatively high factual content" (Rothkopf, 1968, p. 120). As Frase (1969) remarks: "Many of the studies . . . employed only factual questions, which is unfortunate if the researcher's interest is in a variety of behaviors that can produce learning." Finally, the experimental method constrains the subjects not to look back at the material after seeing the adjunct questions. This is "ecologically suspect;" in other words, it does not correspond to the normal reading process. Later research has to some extent tackled these problems and has produced results which are more applicable to real curricular texts. In particular, re256 Review of Research in Education, 6 searchers have made determined efforts to use adjunct questions which elicit higher order cognitive strategies. Frase (1969, 1970, 1971) found that readers given inferential prequestions retained more passage material than those given verbatim prequestions. Watts and Anderson (1971) found that "application of principle" postquestions induced a more thorough processing of the material than did other kinds of questions. Similar results were obtained by Rickards and DiVesta (1974) and Mayer (1975); Rickards and Denner (1978) conclude on this evidence: It has been repeatedly demonstrated that high level questions provide more and better organized recall of passage information [but] a consistent position effect for the co