Directional letter-by-letter analysis and the word-superiority effect

Abstract

Manipulations were introduced in three experiments to produce letter-by-letter analysis of orthographic (CVC) and nonorthographic (CCV or VCC) trigrams. Letters in the trigrams were presented in a different form (normal orientation and order, normal orientation but reversed order, reversed orientation but normal order, or reversed orientation and order) to each of four groups in the first experiment, and in each of the two normal orientation forms to different groups in the second experiment. Subjects both detected the presence or absence of a target letter and classified each trigram as a word or nonword. Additional changes were introduced in the third experiment to ensure that letters were being analyzed in the desired order. Performance was consistently better on orthographic trigrams, but only if the letters were oriented normally. This word-superiority effect (WSE) was related to feature testing that may be carried out letter by letter, with more efficient testing on words. Familiar orientation of letter features seems to be necessary; otherwise, testing becomes so difficult that there is no WSE. However, testing apparently is not finished on a given letter before it is begun on the next. 195A letter from a small predesignated set can be identified correctly more often if it appears in a letter string that forms a word than if it appears in a string of randomly chosen letters (Carr, Lehmkuhle, Kottas, AstorStetson, & Arnold, 1976;Chastain, 1981;Paap & Newsome, 1980b;Spector & Purcell, 1977). Questions persist about the nature of the processes that result in this word-superiority effect (WSE). One issue is whether the positions in the string must be simultaneously analyzed for the WSE to appear (Adams, 1979). Johnston (1981) and Johnston and McClelland (1974) found that the advantage produced by the word context is eliminated if the position of the target is known in advance of the appearance of the string. Apparently, the other letters are not analyzed with the target, if at all. Some investigators have attempted to encourage serial analysis of the positions in the string by using wide interletter spacing. This manipulation has been found to eliminate the WSE (Purcell & Stanovich, 1982;Purcell, Stanovich, & Spector, 1978; but see Paap & Newsome, 1980a), but perhaps again because it discourages analysis of all the letters presented. Finally, Travers (1975) found that serial presentation of letters in the string is more detrimental to letter identification in words than in nonwords. However, although no statistical analysis was reported, a weak WSE still seems to be present in those data. A WSE defmitely appeared in a sequential presentation study reported by Krueger (1971), but since noThe author would like to thank two anonymous reviewers for their helpful suggestions on an earlier draft of this article. Requests for reprints should be addressed to Garvin Chastain, Department of Psychology, Boise State University, Boise, ID 83725. masks were used, it is difficult to relate this to a typical WSE, which req...