During the past fifteen years or so, we have witnessed something like a revolution in the way that researchers understand and describe the process of reading. The work of psycholinguists like Goodman and Smith (e.g., Goodman 1967, 1970; Smith 1982) is familiar to, and very widely accepted by, reading specialists everywhere, now including researchers, writers of texts, and even teachers of reading in a second language. Goodman's well-known characterization of reading as “a psycholinguistic guessing game” has largely carried the day in our field, as in others.

I take it as almost axiomatic that this “top-down” revolution has resulted in major improvements in both our understanding of what good and many not so good readers do, and in the methods and materials that we now employ. We have come a long way in the right direction since the audiolingual reinforcement-of-oral-language (Fries 1945) days, thanks in large part to the work of these scholars.

But top-down models do have some limitations. They tend to emphasize such higher-level skills as the prediction of meaning by means of context clues or certain kinds of background knowledge at the expense of such lower-level skills as the rapid and accurate identification of lexical and grammatical forms. That is, in making the perfectly valid point that fluent reading is primarily a cognitive process, they tend to deemphasize the perceptual and decoding dimensions of that process.