Recent theoretical attempts to account for the lengthening of disjunctive reaction time (RT) with increasing numbers of alternatives have assumed the selective use of memory (cf. Welford, 1960). Selectivity has been emphasized, and the operation of processes such as filtering (Broadbent, 1958), simultaneous or successive search (Hick, 1952; Rappaport, 1959), sequential statistical decisions (Stone, 1960; Tanner, 1961 ) , and cognitive sets (Fitts & Switzer, 1962) have been proposed. With all this emphasis on selectiviry, the role of the memory contenc tends to be taken for granted. The importance of the memory contenc, previously stressed in a filter-theory explanation of aesthetic behavior (Alluisi & Adams, 1962), is re-emphasized here with reference to choice reactions. In a normal adult, the long-term store (or memory) may be represented as including in a frequency or probability sense all the information that has previously entered the person's cognitive information-handling channel. The extent to which his aesthetic judgments are affected by this store may be inferred from the multiple-R of 343 found between rank-ordered preferences and three frequency-based rank orderings of English letters (Alluisi & Adams, 1962). There is reason to suspect that a similar relation might exist between RT and the letter frequencies (cf. Fitts & Switzer, 1962), and therefore between RT and preferences also. The present study seeks to determine whether such relations do exist. Specifically, the present study explored the possibilities of predicting, by means of multiple regression, the rank-ordered RT to different letters in the English alphabet on the bases of the rank orderings of the letters for frequency of use in English (fe), for preference of appearance (P), for frequency of use as the initial (fa) and the terminal (ft) letter in English words, for serial position in the alphabet (ao), and for frequency of use as the initial letter in family names (fn). The method and predictor variables have been described fully elsewhere (Alluisi & Adams, 1962); the RT data were adapted from Fitts and Switzer ( 1962, Fig. 4, p. 326). As shown in Table 1, the results indicate that the best predictions are obtained with two predictor variables, fe and P. The least-squares-justified predictive equation, KT (rank) = 0.411 (fe) + 0.356 (P) + 3.147, indicates that the two predictors are weighted and contribute about equally to the multiple