Abstract

This paper offers a framework that has been used for both developing the tasks used to measure literacy and for understanding the meaning of what has been reported with respect to the comparative literacy proficiencies of adults in participating countries. The framework consists of six parts that represent a logical sequence of steps, from needing to define and represent a particular domain of interest, to identifying and operationalizing characteristics used to construct items, to providing an empirical basis for interpreting results. The various parts of the framework are seen as important in that they help to provide a deeper understanding of the construct of literacy and the various processes associated with it. A processing model is proposed and variables associated with performance on the literacy tasks are identified and verified through regression analyses. These variables are shown to account for between 79% and 89% of the variance in task difficulty. Collectively, these process variables provide a means for moving away from interpreting performance on large-scale surveys in terms of discrete tasks or a single number, toward identifying levels of performance that have generalizability across pools of tasks and toward what Messick (1989) has called a higher level of measurement.

Key words: assessment design, construct validity, evidence centered assessment design, framework, large-scale assessment, literacy/reading, test interpretation