Reading theorists agree that the outcome of reading comprehension is a text representation (Gernsbacher, Varner, & Faust, 1990; Kintsch, 1998). To measure reading comprehension, however, many use testing formats such as multiple-choice and short answer, that have been shown to provide very little information about the text representations created during reading (Kintsch & Kintsch, 2005). A different type of format, proximity measures, is a promising measure for text based representations, although few studies have examined the validity of this type of comprehension measure. The current dissertation addressed this issue by creating a proximity measure named the Passage Comprehension for Structured Text (PCST) and by examining the validity of the PCST through experimental manipulations on the text.

This investigation tested the comprehension of 236 ninth-grade students with the PCST. Students were asked to read a short text followed by a computer task where the students rated the similarity of 11 concepts. Two components were extracted from the PCST, the textbase component and the situation model component. Text manipulations included changes in text coherence and familiarity. It was hypothesized that a coherence manipulation should have an effect on the textbase component and a familiarity manipulation should have an effect on the situation model component. Further, both manipulations should influence the strength of the factor that determines the scores on these components.

A multivariate analysis of variance was used to compare the conditions. Results confirmed that students with coherent text outperformed students with incoherent text on the textbase component and students with familiar text outperformed students with unfamiliar text on the situation model component. Confirmatory factor analysis was used to further explore the effect of text manipulations on the structure of the PCST components. Results indicated that there was a stronger factor for the situation model component when the text was familiar compared to when the text was unfamiliar. Limited evidence suggests that there was also a stronger factor for the textbase component when the text included macrosignals compared to when the text did not include macrosignals.