Printer Friendly

Word frequency effect: a test of a processing-based explanation.

The levels of processing theory (Craik & Lockhart, 1972) became one of the most influential theories of human memory during the 1970s. It emphasizes that memory performance depends on depth of processing or elaboration of meaning of individual items at encoding. A number of studies (see Cermak & Craik, 1979) have demonstrated that deeper levels of processing indeed produce better performance than shallow levels of processing or processing of nonsemantic features.

The levels of processing theory, however, ignores the notion of organization developed during the 1960s (Einstein & Hunt, 1980). Many studies (e.g., Bower, 1970; Mandler, 1967; Tulving, 1962) have shown that the organization of materials also improves one's memory. The question was therefore raised how these disparate approaches could be integrated to explain human memory.

Einstein and Hunt (Einstein & Hunt, 1980; Hunt & Einstein, 1981) were able to combine these concepts in their account of recall and recognition. They argued that the processing of individual items or item-specific processing is important because it facilitates recall and recognition by differentiating a given item from the other items in the stimulus list. Organization, by contrast, enhances performance because it aids in the process of generation at the time of retrieval. They called this type of processing relational processing because organization typically takes advantage of the shared properties among to-be-remembered (TBR) materials. Einstein and Hunt concluded that the notions of organization and levels of processing represent two separate processes that cannot be reduced to one or the other (Hunt & Einstein, 1981).

Several studies (e.g., Einstein & Hunt, 1980; Hunt & Einstein, 1981; Hunt & Seta, 1984) have supported their claim. In particular, Einstein and Hunt (1980) and Hunt and Einstein (1981) demonstrated that when combined, relational and item-specific processing produce an additive effect; that is, both types of processing result in higher recall and recognition than either form of processing alone. This additive effect was not an artifact of the amount of processing because performance was higher when subjects performed both types of processing than when subjects studied stimulus words twice using the same processing task (Hunt & Einstein, 1981).

Hunt and Einstein (1981) further demonstrated that characteristics of the stimulus materials also induce relational and item-specific processing. They employed categorized (i.e., items belong to taxonomic categories) and loosely categorized (i.e., items were selected from loosely defined categories such as round, big, sharp) lists. The characteristics of the list interacted with processing strategies such that item-specific processing resulted in higher performance than relational processing when the list was categorized. In contrast, with a loosely categorized list, relational processing produced better performance than item-specific processing. Hunt and Einstein (1981) argued that because categorized lists naturally induce relational processing, item-specific processing on these lists produce an additive effect. Relational processing, in contrast, creates better performance with loosely categorized lists because these lists tend to encourage item-specific processing. These researchers thus concluded that recall and recognition are based on both (a) the strategy that subjects employ and (b) the type of processing that the materials induce. Optimal performance is expected when these two forms of processing are different.

A number of studies have successfully applied Einstein and Hunt's explanations to various memory phenomena, including the difficulty effect (McDaniel, Einstein, & Lollis, 1988), generation effect (Gardiner & Hampton, 1989), and hypermnesia (Klein, Loftus, Kihlstrom, & Aseron, 1989; Otani & Hodge, 1991). The concepts of relational and item-specific processing thus seem to have gained sufficient credibility as explanatory constructs.

These notions, however, have not been extended to explain a phenomenon called the word frequency effect. In this phenomenon, subjects recall high frequency words better than low frequency words, while recognizing low frequency words more accurately than high frequency words (Shepard, 1967). The word frequency effect is difficult to explain because it is counterintuitive to find low frequency words being more memorable than high frequency words. However, to be a general theory of human memory, any theory must be able to account for this paradoxical effect because it is such a robust phenomenon (e.g., Glanzer & Bowles, 1976; Mandler, 1980; Mandler, Goodman, Wilkes-Gibbs, 1982; Mutter & Hashtroudi, 1987). The present study therefore asked whether Einstein and Hunt's notions can be applied to explain the word frequency effect.

The concepts of relational and item-specific processing provide potentially useful explanations of the word frequency effect. Using these concepts, one might argue that high frequency words are easier to recall because high frequency words encourage subjects to process relational information. Because more meanings are associated with high frequency words (cf. Glanzer & Bowles, 1976), subjects should have an easier time elaborating their meanings and relating them together. Thus, given the assumption that recall requires the generation of TBR words at retrieval, it is reasonable to argue that high frequency words should be recalled better than low frequency words. Low frequency words, in contrast, might force subjects to emphasize item-specific information because only a few meanings are associated with them. If recognition involves discrimination of items rather than generation, then the theory correctly predicts that low frequency words should be recognized better than high frequency words.

The present study tested the validity of this explanation by manipulating both normative word frequency and processing strategy. We reasoned that if the differential processing of high and low frequency words is responsible for the word frequency effect, forcing subjects to process either relational or item-specific information should eliminate or even reverse the effect. We predicted that item-specific processing of high frequency words should eliminate or reverse the word frequency effect in recognition because item-specific processing of high frequency words (which encourage relational processing) should produce an additive effect. Relational processing of low frequency words, on balance, would eliminate or reverse the word frequency effect in free recall because relational processing of low frequency words (which naturally induce item-specific processing) should result in another additive effect.

Method

Materials

The stimulus list was constructed by first selecting 20 common English nouns from each of 6 taxonomic categories in the Battig and Montague (1969) norms. These nouns ranged in rank between 1 to over 42. Among the words selected, half were high frequency and the other half were low frequency words. Across the categories, high frequency words had a mean frequency (U-scale) of 64.79 per million, whereas low frequency words had a mean of 2.17 per million in the Carroll, Davies, and Richman (1971) word frequency norms. Ten words from each, category were then randomly selected to form a stimulus list (see the Appendix).

Because of a paucity of materials, ranks within each category of these words in the Batting and Montague norms could not be controlled. Also, both high and low frequency words were selected from the same categories to avoid unnecessary confounding of characteristics that are unique to some categories.

These words were individually typed at the center of 3- x 5-inch index cards in lower-case letters and were presented to subjects in a randomized sequence (the forward order) or its reverse (the backward order). These words were presented to subjects at the rate of either one word per 5 s or 8 s. Two different rates of presentation were used because the previous study (Otani, Ashford, White, & Whiteman, 1990) suggested that 5 s may not be long enough for subjects to perform the relational processing task.

A blank sheet of paper was provided for the free recall test. A yes-no recognition test was constructed by randomizing 120 words (both targets and distractors) and printing them in five columns. Distractors were those words that were not used as stimulus words.

A posterboard (14 x 24 in) with the category labels and a sheet with 5-point pleasantness rating scales (1 - very unpleasant, 5 - very pleasant) were also prepared. A sheet of randomly generated 2-digit numbers was used for the filler task.

Procedure

Subjects were tested in groups of 6 to 12. Relational processing was induced by asking subjects to sort the words into taxonomic categories by placing them face down under the category names on a posterboard. Item-specific processing was established by instructing subjects to rate the pleasantness of each word using a 5-point scale. In both the sorting and pleasantness rating tasks, subjects were not informed about the subsequent memory tests. These procedures were used by Hunt and Einstein (1981) to manipulate relational and item-specific processing. The third group was an intentional learning control group, asked simply to remember the words. The nature of the test was not revealed.

Following the processing task, subjects performed a filler task (a simple division task) for 2 minutes. They then received a free recall test followed by a yes-no recognition test. Subjects were allowed 3 minutes to complete the free recall test. They were asked to write down in any order as many words as they could recall. Based on the previous study (Otani et al., 1990), 3 minutes seemed to be an adequate interval for this test. On the recognition test, subjects were instructed to circle the words that were shown on the cards. No time limit was imposed on the recognition test.

Subjects

One hundred-twenty male and female students at Central Michigan University participated in this experiment to earn extra credit for their introductory psychology courses. An equal number of subjects (n = 10) were randomly assigned to each of 12 conditions created by the factorial of three between-subjects variables: processing strategy (relational processing, item-specific processing, and intentional learning), presentation rate (5 versus 8 s), and presentation sequence (forward versus backward).

Results

Free Recall

Figure 1 shows the mean number correct on free recall tests. As can be seen, high frequency words were recalled better than low frequency words; thus, the traditional word frequency effect in free recall was demonstrated in this experiment. Furthermore, the similar pattern of results across the processing strategies indicates that the word frequency effect was not modified by the processing strategy manipulation.

A mixed-design analysis of variance (ANOVA) with the presentation rate, presentation sequence, and processing strategy as between-subjects variables and word frequency as a within-subjects variable was performed. It revealed that the effects of processing strategy, F(2, 108) = 4.26, [MS.sub.e] = 16.64, word frequency, F(1, 108) = 110.23, [MS.sub.e] = 4.88, and their interaction, F(2, 108) = 8.16, [MS.sub.e] = 4.88, were significant. The presentation rate, F(1, 108) < 1, [MS.sub.e] = 16.64, and presentation sequence, F(1, 108) < 1, [MS.sub.e] = 16.64, did not reach significance. (The significance level was set at .05 throughout this paper.)

Newman-Keuls tests demonstrated that high frequency words produced significantly better recall than low frequency words in every strategy condition. The tests also indicated that the relational processing group recalled significantly more low frequency words than did the item-specific processing or intentional learning group. The latter two did not differ from one another. High frequency words were recalled equally well by the relational and item-specific processing conditions. The intentional learning group, however, did not recall high frequency words as proficiently as the other processing groups.

To maximize the effect of word frequency, an additional analysis was performed by selecting the three most and least frequent words from each category. However, the pattern of results did not differ from that of the previous analysis.

Recognition

The mean corrected scores (hits - false alarms) are shown in Figure 2. The mean hit and false alarm rates are displayed in Table 1. As can be seen, low frequency words were recognized more accurately than high frequency words in the intentional learning condition, demonstrating the traditional word frequency effect in recognition. Furthermore, there was a reversal in the word frequency effect. That is, subjects in the item-specific processing condition recognized more high frequency words than low frequency words. In contrast, low frequency words were recognized more accurately than high frequency words in the relational processing condition.

[TABULAR DATA 1 OMITTED]

A mixed-design ANOVA confirmed these observations in that the processing strategy, F(2, 108) = 24.44, [MS.sub.e] = 31.13, and Strategy x Word Frequency interaction, F(2, 108) = 10.47, [MS.sub.e] = 8.56, were significant. The effect of word frequency was marginal, F(1, 108) = 3.60, [MS.sub.e] = 8.56, p < .06. The effects of the presentation rate, F(1, 108) = 1.84, [MS.sub.e] = 31.13, and presentation sequence, F(1, 108) < 1, [MS.sub.e] = 31.13, were not significant.

Newman-Keuls tests demonstrated that high frequency words were recognized significantly better than low frequency words in the item-specific processing group. However, low frequency words produced better performance than high frequency words in the relational processing and intentional learning conditions. Newman-Keuls tests further indicated that the item-specific group recognized more low frequency words than the relational group, whereas the relational group recognized more low frequency words than the intentional learning group. The same rank order was found for high frequency words.

As in free recall, three most and three least frequent words were selected from each category for an additional analysis. However, the pattern of the results was the same as that of the previous analysis.

Discussion

Recognition results support an explanation based on the notions of relational and item-specific processing. As predicted, high frequency words were recognized more accurately than low frequency words when subjects were instructed to process item-specific information. In contrast, subjects recognized low frequency words better than high frequency words in the relational processing and intentional learning conditions. Thus, it appears that the word frequency effect in recognition is based on the subjects' tendency to emphasize item-specific information when they try to process low frequency words. However, when they are directed to pay attention to item-specific information, subjects recognize high frequency words better than low frequency words.

Free recall data did not show clear-cut evidence that processing strategy is responsible for the word frequency effect. As predicted, subjects recalled more high frequency words than low frequency words in both the item-specific processing and intentional learning conditions. Contrary to the processing expectation, however, the instruction to process relational information did not eliminate or reverse the word frequency effect. Thus, in the relational processing condition, high frequency words were recalled better than low frequency words, even though recall performance of low frequency words was also enhanced by the relational instructions, as compared with the item-specific or intentional instructions.

There are at least two ways of interpreting the recall results. One is to assume that the word frequency effect was produced by the differential processing of high and low frequency words, as Einstein and Hunt's views suggest. However, because subjects have such a strong tendency to process relational information when they are presented with high frequency words (especially in a categorized list), the performance of low frequency words could not exceed that of the high frequency words.

Another way is to adopt Mandler's (1979, 1980) theory. He assumes that human memory is based on two processes, familiarity and retrieval. Familiarity is based on perceptual integration of stimulus words. Well-integrated materials give rise to a feeling of familiarity because even a partial activation of this memory trace reactivates the whole representation (see also Graf & Mandler, 1984; Mandler et al., 1982). The process of retrieval, in contrast, depends on the amount of contextual information associated with the stimulus materials. According to Mandler, recall is based on the process of retrieval, whereas recognition relies on both familiarity and retrieval.

Using this dual-process model, Mandler (1980) explains that high frequency words are easier to recall than low frequency words because the greater amount of contextual information associated with high frequency words facilitates the process of retrieval. In contrast, low frequency words are recognized better because the shift in the amount of familiarity that occurs between preexposure and postexposure of the stimulus items is more dramatic for low frequency than for high frequency words.

Because more contextual information is associated with high frequency words, Mandler would predict that high frequency words should be recalled better than low frequency words no matter what kind of processing strategy subjects would employ. Thus, better recall of high frequency words in this experiment appears to support Mandler's view.

Mandler's theory, however, cannot fully account for the recall results. If the amount of contextual information associated with the words is the only factor that determines the level of recall, then why did relational processing produce better recall of low frequency words than item-specific processing? Although it is possible that relational processing elaborates contextual features better than item-specific processing, Mandler does not distinguish these two forms of processing.

Mandler's theory is also less successful in predicting the recognition data. He assumes that low frequency words are easier to recognize because an increment in item familiarity is greater for low frequency than for high frequency words following their presentation. However, in the present study, the word frequency effect was reversed in the item-specific processing condition. Mandler must accommodate this finding by assuming that item-specific processing improved recognition of high frequency words by aiding the retrieval process. The problem with this explanation, however, is that comparable improvement in performance was not observed in free recall. In other words, if item-specific processing did increase recognition via a retrieval process, a similar increment should have occurred in free recall because, after all, recall is based solely on the process of retrieval. However, subjects did not recall any more words in the item-specific processing condition than in the relational processing condition, indicating that the difference between these conditions in recognition cannot be attributed to a difference in the amount of retrieval.

To conclude, the recognition results seem to support the relational and item-specific processing account of the word frequency effect. The free recall data, however, did not clearly indicate that the differential processing of high and low frequency words is responsible for the word frequency effect. It appears that both Einstein and Hunt and Mandler have difficulty explaining the recall results.

One note of caution is that typicality of the stimulus words was not controlled in this experiment. It is therefore possible that the reversal of the word frequency effect observed in recognition may have been attributed to typicality rather than frequency of the items. We are currently conducting a study that separates these two factors. Also, in future studies, a within- and between-subjects manipulation of word frequency should be compared. It is possible that the reversal was specific to a mixed-list of high and low frequency words.

References

BATTIG, W. F., & MONTAGUE, W. E. (1969). Category norms for verbal items in 56 categories: A replication and extension of the Connecticut category norms. Journal of Experimental Psychology Monograph, 80. BOWER, G. H. (1970). Organizational factors in memory. Cognitive Psychology, 1, 18-46. CARROLL, J. B., DAVIES, P., & RICHMAN, B. (1971). Word frequency book. New York: American Heritage. CERMAK, L. S., & CRAIK, F. I. M. (1979). Levels of processing in human memory. Hilisdale, NJ: Lawrence Erlbaum. CRAIK, F. I. M., & LOCKHART, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 1 1, 671-684. EINSTEIN, G. O., & HUNT, R. R. (1980). Levels of processing and organization: Additive effects of individual-item and relational processing. Journal of Experimental Psychology: Human Learning and Memory, 6, 588-598. GARDINER, J. M., & HAMPTON, J. A. (1988). Item-specific processing and the generation effect: Support for a distinctiveness account. American Journal of Psychology, 101, 495-504. GLANZER, M., & BOWLES, N., (1976). Analysis of the word-frequency effect in recognition memory. Journal of Experimental Psychology: Human Learning and Memory, 2, 21-31. GRAF, P., & MANDLER, G. (1984). Activation makes words more accessible, but not necessarily more retrievable. Journal of Verbal Learning and Verbal Behavior, 23, 553-568. HUNT, R. R., & EINSTEIN, G. O. (1981). Relational and item-specific information in memory. Journal of Verbal Learning and Verbal Behavior, 20, 497-514. HUNT, R. R., & SETA, C. E. (1984). Category size effects in recall: The roles of relational and individual item information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 454-464. KLEIN, S. B., LOFTUS, J., KIHLSTROM, J. F., & ASERON, R. (1989). Effects of item-specific and relational information on hypermnesic recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 1192-1197. MANDLER, G. (1967). Organization and memory. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation (Vol. 1). New York: Academic Press. MANDLER, G. (1979). Organization and repetition: Organizational principles with special reference to rote learning. In L. Nilsson (Ed.), Perspectives on memory research: Essays in honor of Uppsala University's 500th anniversary (pp. 293-327). Hillsdale, NJ: Lawrence Erlbaum. MANDLER, G. (1980). Recognizing: The judgment of previous occurrence. Psychological Review, 87, 252-271. MANDLER, G., GOODMAN, G. O., WILKES-GIBBS, D. L. (1982). The word frequency paradox in recognition. Memory & Cognition, 10, 101-109. MCDANIEL, M. A., EINSTEIN, G. O., & LOLLIS, T. (1988). Qualitative and quantitative considerations in encoding difficulty effects. Memory & Cognition, 16, 8-14. MUTTER, S. A., & HASHTROUDI, S. (1987). Cognitive effort and the word frequency effect in recognition and lexical decision. American Journal of Psychology, 100, 93-116. OTANI, H., ASHFORD, V. L., WHITE, J. M., & WHITEMAN, H. L. (1990, April). The role of relational and item-specific processing in word frequency effect. Paper presented at the Southern Society for Philosophy and Psychology, Louisville, KY. OTANI, H., & HODGE, M. H. (1991). Does hypermnesia occur in recognition and cued recall? American Journal of Psychology. 104, 101-116. SHEPARD, R. N. (1967). Recognition memory for words, sentences, and pictures. Journal of Verbal Learning and Verbal Behavior, 6, 156-163. TULVING, E. (1962). Subjective organization in free recall of "unrelated" words. Psychological Review, 69, 344-354.

Appendix

Stimulus Words

Animal
High Low
cat jaguar
lion gazelle
rabbit aardvark
cow badger
bear iguana


Fruit
High Low
orange apricot
cherry cantaloupe
lime raspberry
coconut mango
olive persimmon


Weapon
High Low
arrow bayonet
missile slingshot
sword saber
cannon laser
rifle boomerang


Earth Formation
High Low
mountain crevice
lake ravine
cave gorge
island delta
harbor dike


Tool
High Low
hammer wrench
saw awl
square sanders
bolt vise
ladder spade


Weather
High Low
rain cyclone
wind typhoon
lightning squall
fog haze
ice aurora


The results of this experiment were presented at the 31st annual meeting of the Psychonomic Society at New Orleans, LA. The authors thank Dr. Milton H. Hodge and anonymous reviewers for helpful readings of the manuscript. Reprint requests should be sent to Hajime Otani, Department of Psychology, Central Michigan University, Mount Pleasant, Ml, 48859 or 3Lxu6oe[at]CMUVM.BITNET.
COPYRIGHT 1993 The Psychological Record
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Otani, Hajime; Whiteman, Howard L.
Publication:The Psychological Record
Date:Mar 22, 1993
Words:3711
Previous Article:Responding changes across introductory psychology tests.
Next Article:Michotte's Experimental Phenomenology of Perception.
Topics:


Related Articles
Neuron selectivity: down memory lane.
Understanding strain sensitivity effects in vibration isolators.
A further attempt to demonstrate hypermnesia in recognition.
INTER- AND INTRAMODAL ENCODING OF AUDITORY AND VISUAL PRESENTATION OF MATERIAL: EFFECTS ON MEMORY PERFORMANCE.
Computer Technology Training for Prospective Teachers: Computer Attitudes and Perceived Self-Efficacy.
DIRECTED FORGETTING IN EXPLICIT AND IMPLICIT MEMORY: THE ROLE OF ENCODING AND RETRIEVAL MECHANISMS.
Acclimatization in wide dynamic range multichannel compression and linear amplification hearing aids.
Prominence and gemination in Ingrian/[TEXT NOT REPRODUCIBLE IN ASCII].

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters