Bulletin of the Psychonomic Society 1974, Vol. 3 (4),303-304
Cueing and test time in free recall and recognition * DONALD J. LEHR, BARBARA F. SLOAN, and KATHLEEN J. STILLER State University of New York at Fredonia, Fredonia, New York 14063
The purposes of the present experiment were to further examine the processes underlying recall and recognition performance, and to investigate an implicit assumption of the generation-recognition model of item retrieval. The basic design was a 2 by 2 by 2 factorial, manipulating type of test (free recall, recognition), test time (2 min, unlimited), and cueing (cues, no cues). There were 40 Ss in each of the eight independent groups. Each S learned one of two equivalent lists; each list consisted of 40 words selected from eight conceptual categories. List items were blocked during input with the category names provided. Significant main effects for type of test and for cueing were obtained. Results are interpreted as providing support for a one-process theory of recall and recognition and for the generation-recognition model.
The nature of the process or processes underlying recognition memory is an elusive experimental question. Recognition performance, under most experimental conditions, is consistently higher than free recall. Is this performance discrepancy indicative of differential memory search (Kintsch , 1968, 1970) or not (Jacoby, 1972; Tulving & Thomson, 1971, 1973)? An area of uncertainty exists regarding whether recall involves item retrieval plus recognition processes, while recognition, since the to-be-remembered item is sensorily present, involves only the initial retrieval phase. If the preceding distinction is valid (Kintsch , 1970) , then it seems reasonable to assume that more output time is necessary for recall than for recognition to produce an equivalent number of items. If the generation and recognition phases in free recall occur as successive rather than overlapping stages, then the generation-recognition model (Tulving & Thomson, 1973) suggests that the difference between free recall and recognition performance should decrease under certain conditions, namely, when the recall Ss are given appropriate retrieval cues and an unlimited amount of test time. One of the purposes of the present study was to examine this assumption . If recognition performance is not dependent upon a retrieval phase, then a variable affecting retrieval is irrelevant to recognition ; however, if a variable which affects retrieval has similar effects upon recall and recognition, then support for a one-process theory is implied. A second purpose of the present study was to determine whether a variable affecting retrieval, category cueing, will have differential effects upon recall and recognition performance. The existence of an interaction between cueing and type of performance test would support Kintsch's (1970) two-process theory.
METHOD Design and Subjects The design was a 2 by 2 by 2 by 2 independent groups *This research was supported in part by State University of New York Grant-in-Aid. This paPer is sponsored by Glen D. Jensen, who takes full editorial responsibility for its contents.
303
factorial manipulating type of test (free recall, recognition), test time (2 min, unlimited), cueing (cues, no cues), and lists (two equivalent forms). The Ss were 320 students enrolled in introductory psychology who served for course credit ; 20 Ss were assigned by block randomization and order of volunteering to each of the 16 experimental conditions. The Ss were tested in groups of five; all Ss tested in a given session were in the same experimental condition.
Materials
The stimulus materials were two lists of 40 items each. The 40 items in each list were from eight different conceptual categories selected from the Battig and Montague (1969) norms ; thu s, each list consisted of five items in each of eight conceptual categories. The items in each list consisted of the 1st, 4th, 5th, 8th, and 9th frequency members of each category. Items in the 2nd, 3rd, 6th , 7th, and 10th positions, for the same categories, served as distractor items in the recognition test. To maximize list organization and to insure that the items were encoded by category name, the items were presented in a blocked manner (i.e., all five members of a given category presented successively) with the appropriate category name provided with each item. The order of items within a category and the presentation of categories was randomized for each list for each group of five Ss. For Ss in the cueing conditions, the appropriate category names were presented at the top of a mimeographed sheet in alphabetical order. The category names were followed by 40 blank lines for Ss in the cued recall conditions and by 80 items, 40 old and 40 new, for the cued recognition Ss. Category names were not given Ss in the no-cue conditions. For the recognition test , the 80 items were presented randomly with no restrictions. Procedure All Ss were given free-recall instructions and they were told it was not necessary to remember the category names which were presented in the upper-left corner of the slide. Following the instructions, the words were projected on a screen at a 2-sec rate with a .5-sec interitem interval. Three blank slides preceded the first word to allow the Ss to adjust to the presentation rate. Immediately after presentation of the last item, test instructions were given all Ss. Test instructions did not mention the presence of the category names for Ss in the cued conditions. The test time variable was manipulated by informing half of the Ss they would have 2 min to complete the task and informing the remaining half they would have as much time as desired. Free recall Ss were asked to write as many of the words as they could recall and recognition Ss were asked to circle the words which they recognized as having been included in the list.
304
LEHR, SLOAN AND STILLER Table 1 Mean Correct Responses for Recall and Mean Difference Scores for Recognition Recall
Limited Test Time Unlimited Test Time
Reco gnition
Cues
No Cues
Cues
No Cue s
23.33 24 .23
22.03 23.18
28 .38 27 .68
25 .68 27.68
RESULTS The dependent measure was number of correct responses for recall Ss and a difference score (D) for Ss in the recognition conditions. A D score is obtained by computing, for each S, the difference between number of hits and number of false recognitions. According to data provided by Jacoby (1972), D is a better correction for response bias than the signal detection measure of recognition (d'). A t test was conducted to compare performance on the two lists. The results of this analysis provided a t(318) = 1.68, p > .05. Since the lists were statistically equivalent, the list variable was not included in the subsequent analysis. A 2 by 2 by 2 independent groups analysis of variance was performed on the test time, cueing, and type of test variables. Main effects due to cueing, F(l ,312) = 4.05, P < .05, and to type of test, F(I ,312) = 46.33, P < .001, were significant. The means were 25.90 and 24.66 items correct for the cueing and no cueing conditions, and 27.38 and 23 .19 for the recognition and free recall conditions, respectively. Neither the main effect of test time nor any of the interactions approached significance. The mean numbers of items correct for each of the eight experimental conditions are presented in Table 1.
DISCUSSION Results of the present experiment are discussed in terms of the generation-recognition model (Tulving & Thomson, 1973) and the two-process theory (Kintsch, 1970). Tulving and Thomson (1973) suggest that their data, which show recall super ior to recognition, are incompatib le with the generation-recognition model. While these authors refer to a free recall task, the task relevant to the genera tion-recognition model, the learning procedure employed approximates a paired-associate rather than a free-recall task. All Ss were required to learn two lists prior to the critical list; the "sole purpose of the first two lists was to induce subjects to encode each target word with respect to, or in the context of, another word [Tulving & Thomson, 1973, p . 361) ." This procedure, followed in the third list, shares considerable overlap with a paired-associate task; the
other word, a weak target word associative cue, may be considered as a stimulus for the target word and the 5 may well be attempting to associate the two . For the first two lists, recall was tested by providing the weak associative cue and asking for the target word, similar to a MFR procedure for an A-B list. On the third and critical list, however, the 5 was provided a new item as the stimulus for recall. Despite the fact that the new item was a strong associate of the target item, it is nonetheless novel to the S in the experimental situation and is, therefore, not within his episodic memory structure. The paradigm for the critical third list is similar to an A-B, CoB paradigm, a negative transfer paradigm (McGovern, 1964), except that the C term has not been paired with B as far as episodic memory is concerned. When Ss free associate to the strong cues, they often produce the target words but fail to recognize them; when Ss are given the weak cues (A terms), they often recall the target words, suggesting that recall is greater than recognition. It is such data, data from a potential negative transfer paradigm , that are supposedly incompatible with the generation-recognition model. Weak support for the generation-recognition model is provided in the present study. Optimal conditions for the generation and recognition of responses existed for those free recall Ss given retrieval cues and an unlimited amount of performance time . Cueing, according to the two-process theory, is irrelevant to recognition since cueing affects retrieval and 2 min was sufficient to complete the recognition task. There was no difference in recall or recognition performance when the recall task with cues and unlimited time was compared with the recognition task performed with no cues and in 2 min , t(78) = 1.19, p > .05. The two-process theory (Kintsch, 1970) clearly predicts an interaction of Type of Test by Cueing such that cueing would facilitate recall and have no effect on recognition. This interaction does not exist in the present data ; cueing improved recall by a mean of 1.18 items and improved recognition by a mean of 1.35 items. Such data suggest that retrieval processes are as critical to recognition as to recall (Tulving & Thomson, 1971, 1973 ; Jacoby, 1972).
REFERENCES Battig, W. F., & Montague, W. E . Category norms for verbal items in 56 categories: A replication and extension of the Connecticut category norms, Journal of Experimental Psychology, 1969,80(3 , Pt. 2) Jacoby , L . L. Effects of organization on recognition memory. Journal of Experimental Psychology, 1972, 92, 325-333. Kintsch, W. Recognition and free recall of organized lists. Journal of Experimental Psychology, 1968, 77, 481-487 . Kintsch , W. Models for free recall and recognition. In D. A. Nonnan (Ed.), Models of human m emory. New York: Academic Press, 1970. McGovern, J. B. Extinction of associations in four transfer paradigms. Psychological Monographs, 1964, 78(16, Whole No. 593). Tulving, E., & Thomson, D. M. Retrieval processes in recognition memory: Effects of associative context. Journal of Experimental Psychology, 1971, 87, 116-124. Tulving, E. , & Thomson, D. M. Encoding specificity and retrieval processes in episodic memory. Psychological Review, 1973, 80, 352-373. (Received for publication February 21, 1974.)