Recent emphasis on the interactive nature of processing during reading has focused attention on how higher level syntactic-semantic processes might constrain or alter the proesseing of letters and words during reading. The present studies addressed this question by examining the effect of prior knowledge about a passage on the subsequent ability to see visual errors when rereading the same text. Experiment 1 demonstrated that prior knowledge of a passage leads to better proofreading of that passage. Experiment 2 showed that this facilitation is at the level of visual letter and word analyses, not through higher level constraints on rereading. The data are discussed in terms of skilled visual pattern analyses and in terms of the redistribution of processing resources.Since recent models of reading have focused on the that "set into striking relief any discrepancy or 'correcinteractive (e.g., Rumelhart, 1977), rather than the tion' present among its parts" (Kreuger, 1975, p. 969). serial, nature of processing during reading, considerableThat is, does prior knowledge act to conceal or reveal research has addressed the question of whether context a discrepancy from the expected item? or prior knowledge might constrain the processing of the Recent studies of visual word recognition, using letters and words being read. Interest is in whether manipulations of contextual constraint or of past higher level processes (e.g., word, syntactic, or semantic familiarity with the linguistic units, have provided processes) might change or constrain the analyses of some evidence on this question. Since the early dernonIetters and words within a text. The experiments to be strations of the word superiority effect (Reicher, 1969; described were motivated by an interest in this basic Wheeler, 1970), evidence has accumulated to suggest question.that the perception of a constituent letter is faciliAlthough most experiments have addressed this issue tated by a familiar word context (see Kreuger, 1975, by studying on-line contextual constraints during read-and Levy, 1981, for reviews of the familiarity and coning, the present experiments looked at the influence of textual effects). Recently, Rumelhart and McClelland prior knowledge by examining the transfer of inforrna-(1982) reported an impressive series of studies showing tion from one reading of a text to the subsequent read-that the presentation conditions of contextual letters ing of the same text. Having once read a text, most affected the perception of a target letter. Again, confluent readers will reread that passage faster and more textual facilitation was observed in identification of a easily on subsequent occasions. That is, there appears constituent letter. Thus, processing of letters in a word to be some transfer of text-specific knowledge from context facilitated perception not only of the whole one reading session to another. The theoretical issue pattern, but also of its constituent elements. addressed, then, was whether this faster rereading is due But does prior knowle...