The influence of word-knowledge on the word-frequency effect in recognition memory

Abstract

The influence of word-knowledge on the word-frequency effect in recognition memoryThe signal detection theory technique of using ope rating characteristics (OCs) in the study of recognition memory is derived from S's confidence ratings made to a test sequence of words containing both initial inspection and additional items (see, for example, Green & swets, 1966, for details). This method provides a graphic estimate, ds , of the degree of difficulty in discriminating new from old words. As part of one such application of this technique, Schulman (1967) has demonstrated that rare words (Thorndike-Lorge frequency 1) were easier than very common words (Thorndike-Lorge frequency of A or AA) to identify as old or new; but, as Schulman has noted, the rare words used in his experiment were by no means unfamiliar ones. The experiment reported here uses OCs in an attempt to separate the effects of rarity defined in terms of the S's inability to define the meaning of a word. This is achieved by asseSSing the recognition performance of two groups of Ss, physics and arts students, on a sequence of test words containing both technical physics words and common words. In such a situation a physics word, e.g., "tetrode," is rare (low word count) for both physics and arts Ss, yet its meaning is likely to be rare (word knowledge) only for arts Ss.In order to demonstrate a word-frequency effect as reported by Schulman, the OC curves of both physics and arts Ss would be expected to indicate that physics words, rather than common words, were easier to differentiate as being either old or new. In order to demonstrate a word-knowledge effect, significant differences would be expected in the OC curves of physics and arts Ss for physiCS words but not for common words. The superior discrimination of physics words by physics Ss would suggest that a knowledge of word meaning facilitates recognition while a superior performance by arts Ss on physics words would suggest that their rarity aided discrimination as in the wordfrequency effect. The Ss, tested in small groups, were 50 university students, 25 having completed at least one year of an honors physics course and 25 honors arts Ss having little if any science training at high school.An inspection sequence of 60 words contained 30 common words (Thorndike-Lorge rating of300rmore), and 30 physics words. The latter were chosen in order that they would be very familiar to second year university students enrolled for a major in physics while at the same time they would not be a part of the arts Ss ordinary vocabulary. Each word was projected onto a screen for 3 sec using a Kodak Carousel automatic projector (Model 550R) with a 0.5 sec interval between words. The Ss were instructed to try to remember the words appearing on the screen. There was a 3 min interval between the completion of the inspection task and the start of the recognition test. A recognition booklet contained 120 words in random order: 60 physics and 60 common words including all the inspection items.! The Ss marked these words...

A naturalistic study of the word frequency effect in episodic recognition

“…3 These data show the same pattern of results as that reported by Allen and Garton (1968). First, recognition ofzero-frequency CS-tech items was significantly'[ higher for computer science students than for psychology students [F(l,70) = 14.43,MS e = .32],similartothesuperi or recognition of physics words by physics students in Allen and Garton's study.…”

“…The present study is based on the naturalistic approach used by Allen and Garton (1968). The advantage of this approach is that between-group differences in preexperimental familiarity occur naturally, and because the same materials are used for both groups, there is no confounding ofstructural word characteristics and word frequency.…”

“…One difficulty in evaluating Allen and Garton's (1968) results is that normative word frequency and word type were confounded in their study. All rare words were physics words; all common words were nontechnical words.…”

Recognition memory for words presented at a slow or rapid rate

“…It is dear, then, that rare words, although harder to recall, are easier to recognize than common words. What is also apparent is that their greater distinctiveness cannot depend entirely on their more specialized meanings (but see Allen & Garton, 1968), since so many rare words in the present study were totally unfamiliar ones, e.g., "pagrus" and "zabeta." Now, it can be shown that rare words differ significantly in structure from common words, for instance, in their greater incidence of u n usual digrams.…”

Determinants of the word-frequency effect in recognition memory

“…One of the most interesting paradoxes in memory research is the finding that if retention is tested by means of a recall test, high-frequency (HF) words are generally recalled more accurately than low-frequency (LF) words (e.g., Deese, 1960;Sumby, 1963), whereas on a recognition test, it is LF words that are most accurately designated as having been seen before (e.g., Allen & Garton, 1968;Glanzer & Bowles, 1976;Gorman, 1961;Underwood & Freund, 1970). It is puzzling that such different results are evident when retention of the same material is assessed via different tests.…”