Testing a Compensatory-Encoding Model.

Performance at complex tasks involves attentiondemanding (controlled) processing activities concurrent with automatic processes that must successfully blend (Anderson, 1983; Schneider, Dumais, & Shiffrin, 1984). It is natural to ask how the efficiencies of relatively automatic processing subcomponents affect comprehension. Do inefficient subcomponents impair performance by drawing away attention or working memory resources from higher level processing activities to make up for the inefficiencies? Perhaps performance subcomponent efficiencies, beyond a minimal level of skill development, are independent of comprehension. Of course, the relation may depend on the task in question (Wickens, 1980, 1984). As yet, there is no clear answer to these questions. Resource notions bear on this question. One prediction of the relation between subcomponent efficiencies and performance outcomes in reading is made by resource theory (LaBerge & Samuels, 1974; Perfetti, 1985). Resource accounts posit the existence of a limited pool of resources: attention and working memory. These resources can be assigned by the executive of the cognitive system to the most pressing need during performance (Wickens, 1984). Normally these resources are used for processing tasks that are inherently attention demanding, such as making high-level inferences. Occasionally, however, they must be reallocated to make up for an inefficient subcomponent. In reading, for instance, if word recognition is not efficient or if an unfamiliar word is encountered, a reader must rely on word attack skills that are inherently resource demanding (Perfetti, 1988). As text difficulty increases, resources must be deployed as a reader reads a text with many unfamiliar words and difficult concepts (see Perfetti, 1985).