The interactive-compensatory model of reading was developed primarily to explain developmental and individual differences in the use of context to facilitate word recognition during reading. The work leading up to the model is summarized, and more recent empirical studies are described. One major implication derived from these studies and other recent research is that the Goodman-Smith psycholinguistic "guessing game" is an inaccurate conceptualization of individual differences in context use. When a context is adequately instantiated, less-skilled readers utilize context to facilitate word recognition just as much, if not more, than skilled readers. 11-19 (1984) 0741-9325/84/0053-0011$2.00©PRO-ED Inc. 11 going reading, and, more specifically, on how this interaction changes as reading fluency develops. During the early stages of conceptualizing this research problem, Frank Smith's (1971) book Understanding Reading was particularly influential. He imaginatively synthesized some basic research in cognitive psychology and some educational observations into one of the most important "top-down" models of the reading process. Not surprisingly, given this top-down perspective, Smith emphasized the importance of contextual redundancy at all stages of the reading process. One clear developmental prediction deriving from his work is that the performance of the more fluent reader should be more dependent on contextual information.
Levels of ProcessingApproaching this topic from an information processing perspective, we wondered whether Smith's developmental hypothesis was equally tenable at all levels of processing in reading, as Smith had argued. A review of the literature at the time (the mid-1970s) increased our curiosity. There was indeed considerable evidence that skilled readers were more facile at using contextual information at the level of text-integration and comprehension. However, there was a general lack of evidence relevant to the accuracy of the prediction at the level of word recognition. Also, there was considerable looseness and confusion revolving around the use of the term "context effect." Miscitation was rampant.Our background in cognitive psychology may have helped in our reading and organization of the educational literature. One major lesson of modern cognitive psychology (cf. Lachman, Lachman, & Butterfield, 1979) is that human performance can be analyzed, predicted, and explained by specific instantiations of models deriving from the general idea that human information processing results from the exchange of information between processing subroutines that operate at different levels in a partially hierarchical system. Thus, when discussing a particular variable, it is important to distinguish the level, or levels, in the processing system that are affected by the variable.Generally, reading researchers have not been careful in specifying what level in the processing system was being tapped by the particular contextual manipulation employed in a given experiment. The point is that there can be m...