Effects of three text variables on the comprehension of a norm-referenced test

It is felt that the assessment of reading comprehension on norm-referenced tests does not reflect new knowledge about the processes of comprehension. Comprehension, viewed from a process perspective, is an interaction between the text, the reader, and the writer of the text. This study was an attempt to investigate the effects of three text variables—number of propositions, hierarchical position of propositions, and imagery level—on the literal and interpretive comprehension on a norm-referenced test. The theoretical model of comprehension of van Dijk and Kintsch gave guidance. The model utilizes the proposition as a means of cognitive representation and acknowledges both a microstructure (literal), and a macrostructure (gist) of text. A role for imagery is accounted for by elaborative inferences. Passages from a standardized test were propositionally analyzed to derive number of propositions, and propositionally analyzed questions were matched to propositionally analyzed test passages to derive a hierarchical ordering of questions and texts. An a priori computer analysis program was used to identify imagery level of passages. Data collection was carried out in August and September 1989 and January 1990, at The University of Tennessee, Knoxville. College students were administered a multiple choice, standardized test and, based on test results, two groups were formed—one good readers and one poor readers. The literal and interpretive responses of the good and poor readers were determined. Then, the literal responses of good and poor readers were determined for passages ranked high and low in number of propositions, for the questions at the top and at the bottom of the question/text hierarchies, and for passages ranked high and low in imagery level. Further, the interpretive responses of good and poor readers were determined for text variables under investigation. Data were subjected to chi-square analysis to determine significant differences. The results of the study indicated that the literal and interpretive responses of good and poor readers was significantly different. However, effects for number of propositions, hierarchical position of propositions, and imagery level were nonsignificant. Thus, it appears that when the measurement task is multiple choice questions, effects for text variables is diminished.