Toward a Model of Strategies and Summary Writing Performance

Abstract

This study explores the construct of a summarization test task by means of single-group and multigroup structural equation modeling (SEM). It examines the interrelationships between strategy use and performance, drawing on data from 298 Taiwanese undergraduates’ summary essays and their self-reported strategy use. Single-group SEM analyses indicated that the task required multiple constructive strategies of comprehension and composition, including planning, evaluating, discourse synthesis, and source use strategies. The planning and evaluating strategies provided administrative control over other strategies used at varying degrees to improve summarization performance. Multigroup SEM analyses showed that the effects of the planning and evaluating strategies on the discourse synthesis and source use strategies appeared to be invariant across the high (n = 127) and low (n = 122) groups. However, the effects of the discourse synthesis and source use strategies on summarization performance varied across groups. The findings offer insight into the nature of summarization tasks and provide useful validity information for the test task. Implications of the findings are discussed in relation to second language writing assessment and pedagogy.

Notes

¹ The English achievement test consisted of two sections: reading comprehension and summarization. Given that the test was developed to measure both reading and writing abilities, the course required students to score 60% or above for both sections for a possibility to get an A in the course and at least 46% to pass. These standards were thus applied for grouping participants.

² After the initial item-level factor analyses, a scale “goal-setting” with four items was not accounted for in the clustering as overall scale reliability was low (α = 0.33) and the scale produced three items that cross-loaded on the planning factor and one that cross-loaded on the discourse synthesis factor with marginal loadings below .40. Thus, the “goal-setting” scale with four items was dropped.

³ An endogenous factor is hypothesized to be affected by other variables in the model (arrows pointing toward it). In contrast, an exogenous factor is not predicted by any other variable in the model (no arrows pointing toward it).