Five reaction time experiments explored word-frequency effects in wordnonword decision tasks and in pronunciation and memory tasks. Highfrequency words were recognized substantially faster than low-frequency words in the word-nonword decision tasks. However, there was little effect of word frequency in the pronunciation and old-weiv memory tasks. Further, in the word-nonword lexical decision task, prior presentations of words produced substantial and apparently long-lasting reductions in the basic frequency effect. The occurrence of natural language frequency effects only in the word-nonword decision task supports the use of this task to study the organization of, and retrieval from, the subjective lexicon. The modification of frequency effects by repetition suggests that natural language frequency effects may be attributed partly to the recency with which words have occurred. Analysis of the response latencies using Sternberg's additivefactors approach indicates that frequency effects consist of both effects in encoding and in retrieval from memory.In the corpus of 1 million words ana-Word frequency has been used as an exlyzed by Kucera and Francis (1967), recur-planatory variable in psychological research, rences of the most common 133 words ac-especially in word recognition and memory count for about 50% of the entire sample, experiments. Word frequency is a consistent Even when there are alternative ways of predictor of response time differences in communicating an idea, we typically use lexical decision tasks (Landauer & Freedcommon high-frequency words to express man, 1968; Rubenstein, Garfield, & Milourselves. One explanation is that words likan, 1970). On a typical trial in a lexical that have been used frequently in the past decision task, subjects are shown a letter are the ones most likely to "come to mind" string and are instructed to indicate as in language production. quickly as possible whether the string forms a word. In previous research we have found several variables that predict response times This research was supported by a Faculty Re-(e.g., orthography, grammatical class, search Award to the first author from the City j ma geability) ; 0116 of the Strongest predic-University of New York. Experiments 1 and 2 , & , .. J J , , j • r were reported at the meeting of the Eastern Psy-tors of differences between words is frechological Association in Philadelphia, April 1974. quency. Consistently, words that differ by a We wish to thank Thomas Landauer for many factor of 10 in frequency differ by about 50 helpful comments on an earlier version of the msec m response time, resulting in about ^Req'uests for reprints should be sent to Don L. 150-msec differences between relatively rare Scarborough, Department of Psychology, Brook-(1 occurrence per million) and frequent lyn College,
