Levy (1983)demonstrated that more spelling errors were detected, within a limited time period, when familiar passages were proofread than when unfamiliar passages were proofread, In the present series, Experiment 1 eliminated a possible confound in the Levy (1983)studies and showed that errors were detected both faster and more accurately in familiar texts. Experiment 2 demonstrated higher order involvement in the proofreading transfer effect, suggesting that a strictly word-level account was insufficient. Experiment 3 explored the proofreader's sensitivity to the semantic properties of the proofreading passage, showing that the familiarity effect resulted from more efficient processing, not from lack of either visual or semantic analyses. The results are more consistent with a resource-allocation explanation than with either a visual-scanning or a skilled-Visual-processing account.The theoretical issue addressed in the present experiments was the relationship between familiarity with a text and the processing of the written words, with their constituent letters, within that text. Familiarity may lead to faster, more fluent rereading of a passage, but the question is whether a gain in rereading speed is through less thorough analysis of the printed display. That is, reading speed may be gained because less visual processing is needed to recognize familiar word patterns, and thus the printed text is less completely analyzed during rereading. Or, rapid rereading may include just as thorough, but more efficient, analysis of the visual display. Thus, speed may be gained either through attenuated or through more fluent visual reprocessing of a familiar text. Our interest, then, was in the relationship between familiarity and the extensiveness of visual word and letter analyses.Many earlier studies have used constituent-letterdetection tasks to investigate the relationship between word familiarity and constituent-letter recognition. These studies frequently have reported a word superiority effect (Reicher, 1969;Wheeler, 1970), whereby the perception of a target letter is facilitated because the letter is embedded in a familiar word, rather than in a nonword. These studies suggest that familiarity with a word unit This research was supported by Grant A7657 from the Natural Sciences and Engineering Research Council of Canada. The manuscript was prepared while the first author was on sabbatical leave visiting the School of Behavioural Sciences, Macquarie University, Sydney, Australia, and the Department of Psychology, Monash University, Melbourne, Australia. We thank colleagues there for helpful discussions of the work reported here, and particularly Adrienne Bennett, Steven Hausfeld, and Donald Thomson for their comments on drafts of the manuscript. We also thank Meredyth Daneman for providing most of the passages used in Experiment 3 and for her critical comments on the Levy (1983) paper that motivated some of the issues studied here. The authors' mailing address is: Department of Psychology, McMaster University, Hami...