This investigation applies two approaches for representing and comparing text structures as undirected network graphs to describe the influence of narrative and expository lesson texts on readers’ knowledge structure elicited as free recall. Narrative and expository lesson texts and undergraduate participants’ free recall essays (n = 90) from a study by Wolfe and Mienko (Br J Educ Psychol 77, 541–564, 2007) were reanalyzed for lexical proximity as sequential occurrence of selected important terms in the text and as actual minimum distances between these terms. The proximity data were then rendered as Pathfinder networks for analysis. Compared to human-rater benchmark measures, the convergent validity of the sequential approach (range of r = .53 to.83, median r = .70) was a little better than that of the minimum distance approach (.51 to.80, median r = .67). Further, we anticipated that the lesson text structure would be reflected in the text structure of the free recall essays, but this was not observed. On average, the essays in all three lesson conditions tended to converge on a sequential expository structure. Further, compared to the expository lesson texts, the narrative lesson text had a distinctly different influence on posttest recall essay text structures. Overall then, the sequential occurrence approach appears to provide a reasonably good, automatically derived method for representing and comparing lesson texts and participants’ essays as network graphs. If further confirmed and fully automated, there is a wide range of application of such measurement approaches for learning and research.
All Science Journal Classification (ASJC) codes