Letters in briefly presented masked letter strings were detected more accurately when the strings were three-consonant acronyms than when they were nonwords. In the absence of orthographic regularity, this word superiority effect (WSE) could not have depended on visual units corresponding to familiar bigrams. Since rendering the acronyms visually unfamiliar by alternating the case of their constituent letters did not introduce the left-right scanning effects observed for nonwords, it is concluded that the processing of the acronyms did not depend on the formation of whole-word visual units. It is argued instead that the WSE resulted from the postlexical activation of associatively connected single-letter codes. Finally, the results of case-alternation, sizealternation, and mixed-type-font experiments are interpreted in conjunction with the view that lexical access is based on both lowercase and uppercase letter recognition units for words and only uppercase letter recognition units for acronyms.In Reicher's (1969) tachistoscopic recognition paradigm, the brief presentation of a test string (e.g., WORK) was preceded and followed by masking characters that interfered with the processing of the string. Subjects were then provided with two response alternatives (e.g., WORK-WORD). Their choice between the alternatives, which in the above example would indicate whether they had detected the presence of a K in the last letter of WORK, was better for strings that were words than for strings that were nonwords. This result, which has been termed the word superiority effect (WSE), indicates that tachistoscopic letter detection is facilitated by the activation of lexical entries for the test word. Carr and Pollatsek (1985) have recently proposed that the WSE results from the formation of nonvisual, wholeword unitizing codes that "protect" the information in the briefly presented word from the effects of masking and memory loss. Evidence consistent with this hypothesis has been reported by Hawkins, Reicher, Rogers, and Peterson (1976), who showed that letter detection was more accurate for standard response alternatives (e.g., WORD-WORK) than for phonologically identical response alternatives (e.g., SITE-CITE). Another possibility is that the WSE could result from the formation of unitized whole-word codes that are visual rather than phonological. Experiments in which consecutive letters in a test string are presented in different cases (e.g., wOrK) are consistent with the hypothesis that higher order visual codes could mediate the WSE. Pollatsek, Well, andThe research reported in this paper was supported by Grant MDA903-82-C-Q317from the Army Research Institute. We thank Mary LaLomia for preparing the computer programs used to conduct the experiments and analyze the data, and Edward O'Brien for his careful reading of the manuscript. Helga Noice is currently at Rutgers University. H. S. Hock's mailing address is Department of Psychology, Florida Atlantic University, Boca Raton, FL 33431. Schindler (1975) found that case alt...