Search Time as a Function of Context Letter Frequency

Abstract

Neisser (1967) posits the operation of purely feature analytic processes in the scanning of context letters in search lists. This contention was tested by varying the relative frequency of occurrence in English of the context letters in alphanumeric search lists, while holding their feature content constant. It was hypothesized that faster search time on lists of high-frequency context letters would indicate letter-level processing. Equality of search time would indicate the purely feature-level processing required by Neisser's theory. Context letters were segregated into high- and low-frequency sets and their features defined and held constant according to two feature analyses. This yielded a 2 × 2 design with 22 subjects per condition. Attention was given to the control of feature frequency, size of the context-letter set, and approximation of English at the level of bigram, trigram, and word. Results supported a letter level or template model of processing but were shown also to be explainable in terms of some feature-testing models of pattern recognition. Apparatus which allowed for the removal of reaction time in search lists is described.