Studies of Verbal Problem Solving. II. Prediction of Performance from Sentence-Processing Scores.

摘要

This study explored the extent to which separate scores on these processes could predict performance on difficult problems. Scores on the sentence-logic items correlated well(r = .68, N = 34) with number of reasoning problems solved, as did the ordering scores (r = .75). These scores, then, presumably are 'closer' to the actual performance than are verbal scores such as McGraw-Hill Reading Rates (r = .40 to .50). Individual timing of inference responses showed that subjects often had long pauses during inter-sentence processing, whereas intra-sentence responding was relatively fast and regular. The inter-sentence portions of the performance appeared to be key discriminators between success and failure. A small training experiment was carried out with seven new subjects, who were matched on reading scores with the previous group. These subjects were given six hours of intensive, individual practice on the four processes; a standard matrix format was used, and five rules and heuristics were taught which were supposed to facilitate inter-sentence reasoning. The trained people did show improved sentence-logic scores (median about 40% over the comparison group); and if a large reasoning problem contained only straightforward sentences, then the training was very effective. In fact, all seven subjects solved correctly a 4-dimension, 5-variable, negative-disjunction problem within a few minutes. For those problems which hinged upon appreciation of verbal subtleties, though, the training did not help at all.