From conceptual frameworks to testable and simplified models of language learning strategy assessment

In the study of how students learn a second language, inventories have provided a readily available methodological instrument used to assess strategies. Items describing ways of learning a new language have been written, added to similar items, and organised into categories of types and subtypes of strategies. In the construction of one inventory for learning Russian and Spanish (Chamot, O'Malley, Kupper, & Impink-Hernandez, 1987), items were linked to conceptually defined types of strategies (specifically, metacognitive, cognitive, and socialaffective approaches to learning foreign languages). A 63-item version of this inventory has been adapted to assess strategies of learning English as a Second Language (ESL), and strategic preferences of ESL high school students from Sri Lanka and Japan have been linked to learner variables such as ethnicity, mother tongue, religion, gender, and personality type (Liyanage, 2004, 2005). More recent scrutiny of participant responses to this language learning strategies inventory (LLSI), has exposed limitations in its notional classification of strategies and basic design. Factorial analysis of the responses of large heterogeneous samples did not support preconceived links between items, strategies, and types of strategies. Moreover, the different numbers of items for the three types of strategies affected measurement properties. Re-analysis involving content analysis of inventory items, together with exploratory and confirmatory factor analysis utilising a large data set, revealed a much simplified model of language learning. Clear differences between what LLSI scales were supposed to test and outcomes based on analyses of (a) item content and (b) responses elicited by these items indicate the need for caution when using inventories to measure strategies for learning a second language