10URNAL OF VERBALLEARNING AND VERBALBEHAVIOR3, 6-10 (1964)
The Free Recall of Sets of Associatively Related Words HERBERT WEINGARTNER 1
The Johns Hopkins University, Baltimore, Maryland
In this experiment the effects of three variables on the free recall of sets of associatively related words were examined. These variables were: (1) the associative structure in the words presented for free recall; (2) stimulus constraint, produced by systematic changes in the sequence in which the same sets of words were'presented for free recall; and (3) appropriate and inappropriate cueing before, after, or both before and after the presentation of two sets of associatively related words. Bousfleld and Cohen (1956), Deese (1959), and Mayzner and Tresselt (1962), have shown that the associative organization of words presented for recall predicts the number of words correctly recalled. Mayzner and Tresselt (1962) further demonstrated that the distance between associatively related words within lists influences recall scores and associative clustering. Both Bousfield and Cohen (1955) and Deese (1959) have also attempted to influence the free recall of related words by introducing pre-stimulus cues, but neither investigator could demonstrate an effect due to this variable. Pollack (1959) and L~twrence and Cole (1954) investigated recognition accuracy by varying stimulus and response sets independently. These investigators concluded that response uncertainty, not stimulus uncertainty, 1 This investigation was supported by a Public Health Service Fellowship, M P H 17,498, from the National Institutes of Health, Public Health Service. The author wishes to acknowledge his indebtedness to Dr. James Deese and Gregory Lockhead for their help in the preparation of the manuscript. He also wishes to thank Dr. Wendell R. Garner for his valuable suggestions in the design of the experiment.
predicts recognition accuracy. Their findings suggested that if cueing influences the free recall of associatively related words, it should do so as a post-stimulus cue presented just before recall and after the presentation of stimuli. METHOD Stimulus Material. A total of 30 words was presented to all Ss for recall. Fifteen of the words were associates to the word sleep (awake, bed, comfort, dream, eyes, night, wake, unconscious, tired, slumber, rest, repose, drowsy, pleasant, snore), while the other 15 words were associates to the word sour (acid, apple, bitter, grapes, vinegar, sweet, taste, pickles, lemon, disagreeable, distasteful, tart, lime, bad). The words were presented in either a random arrangement or a constrained arrangement. In the constrained lists first one of the groups of associates was presented, and then the other, e.g., as the sequence of the words presented above. The words were individually printed on slides. Each word was exposed for 0.5 sec with a delay period of 1.75 sec between words. Experimental Conditions. Appropriate and inappropriate cues were words chosen on the basis of the frequency with which they elicit, in free association, the words presented for free recall. Sleep and sour were used as appropriate cue words since they frequently elicit the experimental words in free association. Wine and ]atigue were used as inappropriate cue words since they rarely elicit the experimental words in free association but are themselves responses to the words sleep and sour. Wine a n d / a t i g u e are therefore relevant but inappropriate cues. All Ss read instructions before the presentation of the stimuli, and again before writing down, the words they recalled. The pre- and post-stimulus appropriate and inappropriate cues were presented as part of the instructions read by Ss. The pre-sfimulus instructions informed Ss that this was a recall experiment, that they would see each word only once, and that the words would be presented on the screen before them; after seeing all the words and before recalling them
ASSOCIATIVELY RELATED WORDS they would be required to read some further instructions. T h e pre-stimulus instructions also included appropriate, inappropriate, or no cue words. Ss t h a t were presented with either an appropriate or inappropriate cue were told t h a t the words to be presented are related to the cue words a n d were asked to write these words at the b o t t o m of the instruction page. Ss who were n o t presented with either appropriate or inappropriate cue words were asked to copy two n u m b e r s at the b o t t o m of the instruction page. After the presentation of the stimt~li Ss read a short paragraph of post-stimulus instructions telling them to write the recalled words in a n y order with one word per line. Ss were also presented with the appropriate or inappropriate cue instructions, or no cue instructions as in the pre-stimulus cueing conditions. Each S was presented with one of the following seven experimental conditions: (1) appropriate prea n d post-stimulus cues; (2) inappropriate pre- a n d post-stimulus cues; (3) appropriate pre-sfimulus cues; (4) inappropriate pre-stimulus cues; (5) appropriate post-stimulus cues; (6) inappropriate poststimulus cues; (7) no cue words in both the prea n d post-stimulus instructions. Experimental Design. The Ss were r u n in groups, 7 Ss in each group. Each of the Ss within a group was administered one of the 7 experimental conditions as described in the previous section. Each group of Ss was presented with one of 12 r a n d o m arrangements of the experimental words or one of 12 constrained arrangements. A total of 168 Ss were used in the experiment. The Ss were introductory psychology students at The J o h n s Hopkins University. The data were analyzed as in a factorial design. In this analysis Ss a n d specific arrangements were deliberately confounded. T h e error term used to test simple effects a n d interactions is therefore somewhat larger t h a n the expected variance due to S variability alone. Experimental findings would be useful only if they could be generalized across both Ss a n d specific arrangements of the experimental words, a n d therefore specific arrangements were deliberately confounded with groups of Ss. Free Association Data. An additional 50 Ss were presented with each of the experimental words a n d the cue words a n d gave one free association to each. The responses to these words were used to determine the associative structure in each set of associates a n d the f r e q u e n c y with which the experimental words elicit the cue words as responses in free association. T h e frequency with which each experimental word is given as a free associate to sleep a n d sour was available from previously obtained association n o r m s (Russell anel Jenkins, 1954).
7
RESULTS AND DISCUSSION
E17ects o/Cueing. Cueing did not affect the number of words recalled. Table 1 presents the mean total recall scores as welt as the mean recall of the sleep and sour associates for each experimental condition. The standard errors of the mean for recall with random arrangements were 0.92 for total recall, 0.60 for recall of the sleep associates, and 0.57 for the sour associates. These standard errors were 0.93, 0.56, and 0.57, respectively, for recall with the constrained arrangements. The analysis of variance of the number of words correctly recalled shows that neither the cue word, the time of cueing, nor any interaction involving the cue words and the time of cueing significantly affect the number of words recalled. Total recall, as well as the recall of the sleep and sour associates, were examined in this analysis, and in each case the cueing variable did not significantly affect the number of words recalled. Experimental Arrangements. Table 1 shows that Ss recalled correctly more of the experimental words when they were presented in constrained rather than random arrangements. The mean recall for the randomly arranged words was 13.7, and 16.0 for the words from the constrained arrangements. A separate analysis of the recall of the sour and sleep associates showed that Ss recalled a significantly greater number of words from each set of associates when presented with constrained arrangements ( p ~ 0 . 0 1 ) , as compared to random arrangements of the same words. Clustering. Since the stimuli used in this experiment were two sets of associatively related words, clustering of responses was expected at the time of free recall. Bousfield's Ratio of Repetition (1953) was used as an index of Ss' tendency to cluster. Ss presented with constrained arrangements tended to cluster their responses more than did Ss presented with random arrangements. The means of the Ratio of Repetition produced by these two experimental conditions were 0.67 and
WEINGARTNER TABLE 1 MEAN NUMBER OF WORDS RECALLED CORRECTLY FOR EACH EXPERIMENTAL CONDITION
Conditions of Cueing Pre-stimulus Arrangement
Appr.
Inappr.
Random Sour Sleep Total
6.8 6.5 13.3
7.7 5.8 13.5
Constrained Sour Sleep Total
8.6 7.7 16.3
8.7 7.7 16.4
Post-stimulus Appr.
Pre and Post
Inappr.
Appr.
Inappr.
No cue
7.8 5.7 13.5
7.8 5.8 13.6
7.8 6.5 14.3
7.0 6.5 13.5
7.8 6.3 14.1
8.6 7.9 16.5
7.1 6.5 13.6
7.9 7.6 15.5
9.1 8.1 17.2
9.2 7.3 16.5
0.60, respectively, and this difference is significant (p < 0.01). It is not surprising to find that experimental conditions that affect total recall will also affect Ss' tendency to cluster. However, it was also found that the tendency for individual Ss to cluster responses is significantly correlated with the total number of words correctly recalled, r - - 0.40, (p < 0.01). This finding is unusual since the correlations between individual Ss' tendency to cluster and total recall have been reported as about zero in the literature on clustering (Bousfield and Cohen, 1955). In most of these earlier clustering experiments, superordinate categories rather than associative categories have been used to generate stimuli. I t m a y ' b e that Ss do not cluster associations in the same way as words from superordinate categories. I t appears that sequential organization in the. presented stimuli and in response organization at free recall affect the number of words correctly recalled.. Intrusions. A total of 176 intrusions occurred in the recall protocols of all the Ss. Thirty-seven per cent of these intrusions were the word sleep, while 18% of them were the word sour; this difference in the frequency with which sleep and sour intruded as responses is significant (p < 0.01). No other word occurred more than 3% of the time as an intrusion. Of the total number of times either sleep or sour were given as responses in
free association to the experimental words, sleep occurred 76% of the time and sour 24% of the time. The difference in the frequency with which sleep and sour are given as responses in free association to the experimental words accounts for the difference in the frequency with which these words occur as intrusions. Specific Words Recalled. Table 1 shows that a greater number of words associated with sour were correctly recalled (M z 15.8) than words associated with sleep ( M - - 13.7). This difference is significant (p < 0.001) and is attributed to the frequency with which the words within each set elicit one another in free association. Twenty per cent of the associations to the words associated with sour were other words in that collection, while only 7% of the free associations to the experimental words associated with sleep elicited other experimental words associated with sleep. Thus, inter-item associative strength (Deese 1959) is a predictor of the number of words correctly recalled from each of two sets of associatively related words, and also of the frequency with which specific responses intrude at the time of free recall. An analysis of the frequency with which each word was recalled under each cueing condition shows that cueing did influence the course of recall but did not produce variations in the number of words correctly recalled. In
9
ASSOCIATIVELY RELATED WORDS
this analysis the frequency with which each word was recalled was ranked for the appropriate and inappropriate cueing conditions. These same words were also ranked by the frequency with which they occurred as free associations to sleep and sour. Table 2 shows, by rank-order correlations, that when the ap-
related to them. Therefore, because cueing conditions change the likelihood of the cue words occurring in recall, they change the response organization of the words occurring in recall, even though they do not produce a variation in the total number of words correctly recalled.
Two measures of structure within associatively related stimulus sets, sequential conWHICH THE EXEERnWENTALWORDSWERE RECALLED straint and associative structure, influence the AND THE FREQUENCY WITtt WHICH THEY WERE GIVEN total number of words recalled. The frequency AS FREE ASSOCIATES TO TI:IEAPPROPRIATE CUES with which specific sets of associatively related Experimental Conditions words are recalled, the frequency with which Appropriate cue Inappropriatecue specific words within these sets are recalled, arrangements arrangements and the occurrence of intrusions are all deterStimulus ConConmined by the response organization present at word R a n d o m strained Random strained the time of free recall. The response organizaSleep 0.41 0.30 0.50a 0.48a tion present at the time of free recall is largely Sour --O.18 --0.05 0.53a 0.52a determined by the stimuli presented for recall a p < 0.05 but can also be influenced by cueing condipropriate cue words, sleep and sour, were tions. At the time Of free recall, Ss associate to presented as cues, Ss tended to produce in their responses, as they are emitted, in reconrecall words that were not frequent associates structing the stimulus sets to be recalled. to sleep and sour. When, however, Ss were SVMMARY presented with the inappropriate cue words, Two sets of 15 ass0ciatively related words high-frequency associates to sleep a n d sour were presented to 168 Ss f6r free recall. Ss tended to occur in recall. recalled a greater number of words when they An analysis of the occurrence of sleep and were presented in constrained rather than sour as intrusions provides the reason for the random arrangements. Cueing did not inrelations between frequency in recall and frefluence the number of words correctly recalled quency of association to sleep and sour under but did change the associative structure presdifferent cueing conditions. Ss presented with ent at the time of free recall and thereby sleep and sour as cue words rarely give these altered which of the words were recalled. words as intrusions in free recall. Those Ss Associative structure was found to predict the presented with the inappropriate cues do give frequency with which individual words were sleep and sour as intrusions. S l e e p and sour occurred as intrusions more than four times recalled and the occurrence of intrusioris. TABLE 2 RANK-ORDER CORRELATIONS OF THE FREQUENCY WITI-I
as frequently when Ss were presented with inappropriate cues than when they were presented with appropriate cues. These words occur as intrusions because they are evoked by the associative structure of the stimulus words as they are recalled. When they do occur as intrusions, they, too, are available as stimuli during recall which, in turn, can produce in recall words that are associatively
REFERENCES
BOUSFIELD,W. A. The occurrence of clustering in the recall of randomly arranged associates. ]. Gen. Psychol., 1953, 49, 229-240. BOUSF~LD, W. A., ANDCOHEN,B. H. General review of a program of research on associative clustering. The Minnesota Con]erence on Associative Processes in Verbal Behavior, 1955, 64-101. BOUSF~D, W. A., AND CO~EN, B. H. Clustering in recall as a function of the number of word
10
WEINGARTNER
categories in the stimulus word list. Y. Gen. Psychol., 1956, 55, 95-107. D~.ESE, J. Influence of inter-item associative strength upon immediate free recall. Psychol. Rep., 1959, 5, 305-312. DEESE, J. On the prediction of the occurrence of particular verbal intrusions in immediate free recall. J. exp. Psychol., 1959, 58, 17-22. LAWR~-r~CE, D. H., AND COLE, G. R. Accuracy of recognition with alternatives before and after the stimulus. Y. exp. Psychol., 1954, 47, 208-214. MAYZNER, M. S., AZ~DTRESSELT, M. E. Incidental
learning: A function of associative strength and distance between S-R pairs. J. Psychol., 1961, 53, 155-160. POLLACK, I. Message uncertainty and message reception. J. acoust. Soc. Amer., 1959, 31, 1500-1508. RvssELr~, W. A., AND J~.Z~KI~rS, J. J. The complete Minnesota norms for responses to 100 words from the Kent-Rosanoff word association test. Tech. Rep. No. 11, Office o/ Naval Research, Univ. o/Minn., 1954. (Received June 12, 1963)