READER-TEST INTERACTIONS: AN EXPLANATORY ITEM RESPONSE STUDY ON READING COMPREHENSION

Abstract

According to the RAND model framework, reading comprehension test performance is influenced by readers’ reading skills or reader characteristics, test properties, and their interactions. However, little empirical research has systematically compared the impacts of reader characteristics, test properties, and reader-test interactions across different standardized tests. The present study used the explanatory item response approach to investigate the reader-test interactions in two commonly used standardized tests: Gates-MacGinitie Reading Test-4th edition (GMRT-4; MacGinitie et al., 2000) and Wechsler Individual Achievement Test- 3rd Edition Reading Comprehension (WIAT-III; Wechsler, 2009). Five reader characteristics scores (i.e., decoding, pseudoword reading, vocabulary, fluency, morpho-syntactic knowledge) of 89 fourth graders were obtained. Six test properties (i.e., mean sentence length, mean log word frequency, referential cohesion, deep cohesion, genre, and question type) of the two tests were measured by The Lexile Framework for Reading (Schnick & Knickelbine, 2007) and Coh-Metrix Text Common Core Ease and Readability Assessor (Coh-Metrix-TERA; Graesser et al., 201). Genres were coded as narrative texts and expository texts, and question types were classified into literal questions and inferential questions. Explanatory item response models (EIRM) treated both reader characteristics and test properties as random variables. Results indicated in both GMRT-4 and WIAT-III, fluency and vocabulary were the most crucial reader characteristics over other reading skills. For GMRT-4, lower mean sentence length, higher referential cohesion, and higher deep cohesion made the passage easier to understand, and expository texts were more difficult than narrative texts. For WIAT-III, inferential questions were more challenging than literal questions. Three significant reader-test interactions were found in GMRT-4 between vocabulary and referential cohesion, vocabulary and word frequency, and referential cohesion and question type. Higher-level vocabulary students had better comprehension performance in passages with low referential cohesion and high word frequency in GMRT-4. Literal questions were easier than inferential questions for low referential cohesion texts, but the result was reversed for high referential cohesion texts. Through this study, fourth graders’ reading performance in the two standardized tests was better understood, and limitations and implications for practice and future research were discussed.