30.3 Some History

Research on mathemagenic activities has become confused with the flimsy issue of whether adjunct questions are a useful instructional method. The confusion was probably caused by misreading of early research and an insidious preoccupation with method in the educational research community. Early work on the modification of mathemagenic activities relied on the experimental use of questions. However, the matter of the educational value of questions is on the periphery of research on mathemagenic activities and does not generate many interesting practical questions. No report from our laboratory has ever concluded that questions are necessarily an effective instructional maneuver.

The use of adjunct questions in early research on mathemagenic activities grew out of the analysis of active responding in programmed instruction (Rothkopf, 1963). It was proposed by some, at that time, that coffective feedback (i.e., knowledge of results) for student responses reinforced substantive knowledge gained from self-instructional material (e.g., McDonald, 1965, pp. 90-91). The analytic metaphor was taken from studies of instrumental learning in birds and small mammals. Contingencies between key pecking by pigeons and food reinforcement was known to increase the likelihood of the future occurrence of key pecking. By analogy, it was thought that contingencies between responses to a program frame and evaluative feedback strengthened the subject-matter competence underlying that response (McDonald, 1965). The equation of the effect of knowledge of results on substantive learning with reinforcement of substantive knowledge seemed faulty. In instrumental conditioning in animals, withdrawal of reinforcement leads to a weakening of the learned response. Similar effects in human substantive learning have not been observed, i.e., elicitation of subject matter knowledge without feedback did not result in the deterioration of this knowledge. This suggested that one important effect of active student responses during the use of programmed material was on reading and studying activities rather than on substantive knowledge. Active responses and corrective feedback might shape an ' d maintain study activities. This concept broadly equated the use of specially programmed materials calling for student responses with any reading activity that was accompanied by intermittent questions.

Two reports in the literature of reading supported our working hypothesis. Hoffman (1946), studying visual fatigue, observed college students during a 4-hour reading period. He recorded their eye movements using electrooculographic methods for 5 minutes every half hour. Hoffman reported sharp deterioration of reading activities (e.g., fewer fixations, fewer lines read, increases in regressions and in blinking) after only 30 minutes of reading, and this deterioration progressed throughout the entire 4-hour period. By contrast, Carmichael and Dearborn (1947) observed readers during a 6-hour period, using similar procedures and subjects. As can be seen in Figure 30- 1, the change in eye movements due to prolonged reading was much less marked for Carmichael and Dearborn's subjects than for Hoffman's. The Carmichael and Dearborn experiment differed in one important respect from the Hoffman study: Carmichael and Dearborn administered a comprehension test every 25 pages. This test included about one question for each page in the preceding 25-page reading section. The implications were fairly clear. Systematic use of questions, asked by an authoritative source, maintained effective reading patterns that would otherwise deteriorate. These results supported the conjecture that periodic questions could exert some influence on reading activities. The Hoffman (1946) and the Carmichael and Dearborn (1947) reports are the result of very substantial experimentation, but their work has not received the attention that it deserves. They clearly stimulated our early work.

Figure 30-1. Persistence of close inspection during extended reading as a function of the use of intermittent questions. The plot shows the number of lines read during 5-minute samples. (After Rothkopf, 1982.).