The influence of methods of selecting concepts of an expository text on different reading representations' predictive ability

Abstract

In this study, we compared two experimental methods of selecting terms in expository text to generate reading representations and tested how well these reading representations predicted reading comprehension. The two experimental methods were the traditional method of using all terms (all keywords) to create participants' representation networks, and the terms categorization (TC) method of using only important terms (core and branch words). Representation networks were assessed using participants' adjacency scores, ratings of relatedness in pairs of terms, and using summary (summary writing) by all turms. An in-subject design was performed in experiments 1 and 2, and an inter-subject design was performed in experiment 3 to test the hypothesis. With the same sample in exp1 and epx2, a different sample in each exp3. Experiment 1 showed that when using only the traditional way of selecting terms, adjacency was better than relatedness in predicting reading comprehension. Reading representations generated based on the summary method could not predict participants' reading comprehension ability, so this method was excluded from subsequent studies. Experiment 2 showed that the terms selected in Experiment 1 were stronger predictors of reading comprehension when the word pairs included a core term (central to understanding of full text) or a branch term (key to understanding paragraph), relative to a detail term (not affect the understanding full text). Experiment 3 found that whereas the two methods were equally effective in generating representations measured by adjacency, TC was superior in generating representations measured by relatedness. These conclusions have important implications for future research and application.