The Monster in the Classroom: Assessing Middle School Language via Monster, P.I. to Inform Instruction

• In Event: Mapping Morphology and Its Multidimensionality, Assessment, Relations to Outcomes, and Relation to Covariates

Abstract

Purpose:
Language is powerful. Across orthographies, by middle school, language skills are the stronger predictor of reading comprehension compared to word reading skills (Catts, Adlof, & Weismer, 2006; Florit & Cain, 2011; Gough, Hoover, & Peterson, 1996). Hence, language plays a foundational role in reading comprehension. Language has been theoretically conceptualized and assessed in many ways (see Language and Reading Research Consortium[LARRC], 2015), yet language is rarely included in a school-based yearly assessment battery (Adlof & Hogan, 2019). Building on Foorman et al. (2015), the current study explores a model of language as consisting of morphology, vocabulary, and syntax. In particular, a multidimensional model of morphology is considered. This model is then operationalized into a gamified, computer-adaptive assessment with links to standardized reading and vocabulary explored.

Methods:
Participants were 3,214 fifth through eighth graders (N=1,026 fifth, 742 sixth, 715 seventh, and 731 eighth graders). Seven morphology, four vocabulary, and one syntax measure were administered within the Monster, P.I. platform (see www.worddetectives.com; Goodwin et al., 2020). As part of Monster, P.I., students solve word puzzles (i.e., items) to earn clues to identify and catch a monster, saving the city! Also, standardized vocabulary was assessed via the Gates-MacGinitie Reading Vocabulary assessment (MacGinitie et al., 2002) and standardized reading was assessed via the Measures of Academic Progress [MAP] (see Northwest Evaluation Association, 2014). To explore our language model, a series of multiple-group unidimensional and multidimensional item response theory models were fit to the item-level data within each of the global domains of measurement (i.e., morphology, vocabulary, and syntax), taking into account vertical equating and scaling across the grade-levels. To explore links to reading, we used hierarchical multiple regression modeling.

Results:
Results indicate language and specifically morphology are multidimensional with morphology and vocabulary best fit by bifactor models that identify performance overall and on specific skills. Results also suggest reliability and validity of operationalization of our language model via Monster, P.I. as marginal reliability yielded strong results for most Morphology skills (Skill 1= .90, Skill 2= .93, Skill 3=.70, Skill 4=.92) and acceptable results for Vocabulary (.80) and Syntax (.70). Also, strong links to reading and vocabulary were shown as the combination of morphology skill ability scores and vocabulary and syntax resulted in 56%-75% of the variance explained in MAP scores (standardized reading) and 62%-78% of the variance explained in Gates vocabulary scores. Importantly, each additional morphological skill explained significant additional variance in reading comprehension.

Discussion:
This study emphasizes considering language as multidimensional consisting of morphology, vocabulary, and syntax and also considering the components of language, particularly morphology, as multidimensional. It also emphasizes considering multiple skills to acknowledge the broader role of morphology in reading. In terms of practice, findings suggest support from students and teachers in using Monster, PI to assess language in a reliable, valid, and instructionally-relevant manner that supports design of instruction to capitalize on students’ strengths and compensate for weaknesses. These understandings have important implications for theory, research, and practice.

Authors

• Amanda P. Goodwin, Vanderbilt University

• Yaacov Petscher, The Florida Center for Reading Research