More on the way we 'see' letters from words within memory

The present experiment is a replication and an extension of an earlier study (Johnson, 1986, Journal of Memory and Language, 25, 558-570) which demonstrated that when subjects are asked to scan a word in order to detect a predesignated target letter, the strategy they adopt is quite reminiscent of a memory scan, rather a visual scan, in that they appear to engage in an exhausive search. On the other hand, when the displays to be scanned visually were consonant arrays, the resulting pattern of data did appear to reflect the type of self-terminating search that is typical of a visual scan. The present experiment replicated that differential effect and demonstrated that whether or not subjects knew the nature of the display in advance was irrelevant. In fact, the small differences that did occur suggested en enhanced processing efficiency when subjects did not have such advance knowledge. The data were interpreted as indicating that: (1) words are processed holistically; (2) such processing is not under the strategic control of the perceiver; and (3) component-level (letter) information can be obtained only by deriving it within memory from the word-level representation. Finally, it was suggested that the usual difference in scan rate obtained between words and nonwords may not reflect any direct influence of that factor on the scan rate itself, but rather it may be a derivative effect that stems from a difference in the type of scan.