In order to separate the effects of experience from other characteristics of word frequency (e.g., orthographic distinctiveness), computer science and psychology students rated their experience with computer science technical items and nontechnical items from a wide range of word frequencies prior to being tested for recognition memory of the rated items. For nontechnical items, there was a curvilinear relationship between recognition accuracy and word frequency for both groups of students. The usual superiority of low-frequency words was demonstrated and high-frequency words were recognized least well. For technical items, a similar curvilinear relationship was evident for the psychology students, but for the computer science students, recognition accuracy was inversely related to word frequency.The ratings data showed that subjective experience rather than background word frequency was the better predictor of recognition accuracy.Gorman (1961) was one of the first to report on the superior recognition oflow-over high-frequency words. The low-frequency advantage in recognition has since been confirmed in a number of studies over a range of conditions and manipulations (see Glanzer & Adams, 1985, for a review). When word frequency effects are studied over a larger range offrequencies (not just low vs. high), the relationship between frequency and accuracy is an inverted U-shaped function (Wixted, 1992;Zechmeister, Curt, & Sebastian, 1978). Recognition is highest for lowfrequency words and decreases as word frequency either increases or decreases. Nonwords are assumed to be poorly recognized either because they have no meaning or because they have never been encountered prior to the experiment. Although poor recognition of very low-frequency words has been linked to low ratings of lexicality (Schulman, 1976) and to low ratings of meaningfulness and familiarity (Zechmeister et al., 1978), the influence of these experiential factors has not been separated from other correlates ofword frequency (e.g., phonemic and graphemic composition).'Allen and Garton (1968) adopted a naturalistic approach in their investigation of the word frequency effect. They compared physics and arts students' recognition memory for two classes of words: physics technical words (expected to be in the ordinary vocabulary of the physics but not the arts students) and common words (Thorndike & Lorge, 1944, rating of 30 or more). Their results showed the usual word frequency effect in that the lower frequency physics words were recognized better