JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR
11, 356-361 (1972)
Effects of Presentation and Recall Trials on Clustering and Recall1 ROBERT L. HUDSON, MARY L.
SOLOMON, AND
JERRY L. DAVIS
Mlssisstppt State Unt~'ersity, State College, Mlsstsstppi 39762
In Experiment I the pattern of presentation and recall trials was varied for five groups who received 40 related words On common immediate postlnput recall trials there were no differences between any groups on either recall or clustering. Thin included a group given 11 presentation trials followed by a recall trial and a group receiving an initial presentation trial, nine recall trials, a presentation trial and a final recall trial (Group 9R + P). Experiment lI was an attempt to determine ~fa 9R + P group would be equal to a standard group on the final recall when unrelated words were used. The two groups did not differ significantly.
The traditional multltrial free recall (MFTR) experiment consists of alternating presentation (P) and recall (R) periods. Data obtained in standard M T F R experiments (Bousfield, Puff, & Cowan, 1964; Tulving, 1962) have indicated that both item retention and secondary organization are facilitated over repeated trials. The standard M T F R experiment, however, does not offer a reliable paradigm in which to compare and contrast the effects of prior P-trials and prior R-trials on free recall. In order to differentiate the effects of Ptrials and R-trials on subsequent item retention, Tulvmg (1967, Experiment II) departed somewhat from the typical M T F R procedure. The number and sequential combinations of P-trials and R-trials preceding a given R-trial were manipulated within six cycles of four trials each. The three independent levels of trial sequence manipulated were. (a) P-R-P-R, or Condition S; (b) P-P-P-R, or Condition P; and (c) P-R-R-R, or Condition R Tulvlng found that on item retention measures taken within cycles, P-trlals functioned to increase item retention, while R-trials were associated w~th a decrease in retention. However, item
retention measures taken following a common P-trial indicated Ss in Condition R to have achieved the same amount of retention as had Ss in the other groups. Tulving concluded that R-trials functioned not only to allow measurement of the amount of learning, but also appear to facilitate subsequent immediate postinput recall approximately to the same extent as does study of the material in the input phases, that is, on P-trials. The primary objective of the present study was to replicate and extend Tulving's (1967, Experiment II) findings concerning item retention. Experiment I involved the same three conditions used by Tulvlng which differed in regard to the sequential ordering of P-trlals and R-trials within three cycles of four trials: Condition S, Condition P, and Condition R. Two other conditions were also tested. Condition 9R + P called for an initial P-trial, nine consecutive R-trials, and one more P-real preceding the terminal R-trial. Condition 9P + P called for 10 consecutive P-trials following the initial P-trial and terminated in one R-trial. It was expected that item retention scores observed under Condition S and Condition R would be approximately the same on the 1 This research was partially supported by institutional funds granted to Mississippi State University common ~mmediate post presentation Rtrials, Trials 2, 6, and 10. Tulving reasoned by the National Science Foundation. 356 © 1972 by Academic Press, Inc
PRESENTATION AND RECALL TRIALS
conditions in blocks of five, and were tested individually. Each S was instructed that he would receive 12 trials and that some trials would be P-trials and some would be R-trials. Further each S was Informed as to the sequence of P- and R-trials to which he would be subjected. The Ss were told the words could be categorized and the categories were named. Each S was randomly assigned to one of the eight orders of words for his first P-trial The orders used on subsequent P-trials followed the sequence of orders on the memory drum tape. Those Ss in groups which received more than eight P-trials continued to follow the sequence of orders on the tape after the eighth Ptrial. On P-trials the 40 words were exposed for 2 sec. each. When S's condition required a R-trial, E presented him with a pencil and a sheet of paper with 40 blank spaces. The Swas allowed 80 sec. for each recall period. After each R-trial, E collected the recall sheet and signaled the nature of the real to follow.
that R-trials facilitated the overall immediate postpresentation recall approximately s a m e e x t e n t as d o P - t r i a l s .
to the
A logical extension of this argument qualifies t h e p r e d i c t i o n t h a t t h e Ss o f C o n d i t i o n s S, P, 9 P + P, a n d 9 R + P s h o u l d d e m o n s t r a t e approximately equal retention scores on the c o m m o n R - t r i a l , T r i a l 12, e v e n t h o u g h t h e groups differed considerably sequence of trial-types.
on
357
the prior
A second purpose of Experiment I was to i l l u m i n a t e t h e d i f f e r e n t i a l t y p e - o f - r e a l effects o n c l u s t e r i n g . R o s n e r (1970) h a s r e p o r t e d t h a t R-trials tend to facilitate subjective organizatlon to a greater extent than do P-trials.
EXPERIMENT I
Method
Results
Subjects The Ss were 70 male and female undergraduate students enrolled in general psychology at Mississippi State University. There were 14 Ss per group. All Ss participated as a course requirement. Materials. The stimulus list employed was composed of items selected from the norms of Battig and Montague (1969) The list was comprised of four nouns from each of 10 different categories for a total of 40 words. Eight random orders of the list were typed on tape for presentation by a Stowe memory drum. The Ss were randomly assigned to one of the five treatment
Item retention. T h e d a t a w e r e i n i t i a l l y examined in terms of the number of correctly r e c a l l e d w o r d s o n all R - t r i a l s . T a b l e 1 p r e sents the mean number of words recalled by Ss i n t h e five c o n d i t i o n s . The item recall scores observed on R-trials which coincided temporally for two or more g r o u p s w e r e f u r t h e r e x a m i n e d b y a series o f completely randomized one-way analysis of
TABLE 1 MEAN NUMBER OF ITEMS CORRECTLY RECALLED AS A FUNCTION OF CONDITION FOR EXPERIMENTS I AND II Number of Trial Condition
2
3
4
5
6
7
8
9
10
11
12
P 22.7 P P 16.6
26.6 23.4 26.1 P 16.7
P P P P 17 1
26.8 25.4 P P 17.4
P 25 4 P P P
28.6 26.7 26.9 23.4 25.2
P 10.3
18.8 10.1
P 10.5
20.7 10.3
P P
21.3 19.1
Experiment I S R P 9P + P 9R + P
P P P P P
14 9 15.7 P P 15.1
P 15 6 P P 15.4
18 8 15.6 19 9 P 15 6
S 9R + P
P P
9.9 10 6
P 97
13.9 9.7
P P P P 16.4
23.1 21.1 P P 15 9
Experiment II P 97
16.2 10.0
358
HUDSON, SOLOMON,AND DAVIS
variance (Kirk, 1968). These were followed by Tukey's HSD Test (Kirk, 1968) following a significant F ratio. Trials 6-8, and I0 indicated only that Group 9R + P was inferior to the other groups. Trial 4 indicated that Groups S and P were not different but both significantly exceeded Groups R and 9R + P which were not different. Level of significance used in all tests was .05. Patterns of individual item recall. Tulving (1967) developed a trial-to-trial analysis technique to be employed in the study of recall patterns. The trial-to-trial pattern analysis was applied to the individual items recalled by Ss in Groups S, R, and 9R + P on Trial 2 to Trial 4, Trial 4-6, Trial 6-8, Trial 8-10, and Trial 10-12. Items recalled on any R-trial occurring after the first R-trial may be considered individually in terms of three types of trial-to-trial transitions: (a) the CC transition, referring to items which are recalled on an Rtrial which were also recalled on a previous
R-trial; (b) the N C transition, referring to individual items recalled on an R-trial which were not recalled on a previous R-trial; and (c) the C N transition, referring to items not recalled on a present R-trial which were recalled on a previous R-triaL The mean CC, NC, and CN data observed on the five trial transitions for the three groups are presented in Table 2. The recall data observed in Group S and Group R for each of the three indices of patterning were subjected individually to 2 x 3 split plot factorial analyses of varmnce (Kirk, 1968). Only the 2-4, 6-8, and 10-12 transitions were considered in the analysis because in these transitions a P-trial intervened for Group S and a R-trial intervened for Group R. On both the 4-6 and 8-10 transitions a P-trial intervened for both groups. Group 9R + P was not included because the total number of words recalled was less than the other groups on most trials.
TABLE 2 MEANPATTERNINGOFITEMRECALLOVERFIVETRIAL-TO-TRIALTRANSITIONS ASA FUNCTIONOF CONDITIONFOREXPERIMENTSI ANDII Measure
Transition Condition
CC
CN
NC
CC CN NC
2-4
4-6
S R 9R + P S R 9R + P S R 9R + P
8.8 13.1 13.3 61 2.5 1.8 10.0 2.5 2.4
Experiment I 12.3 10.7 •4.6 7.2 4.9 .9 11.0 10.4 1.2
S 9R + P S 9R + P S 9R + P
5.7 8.6 4.3
Experiment II 7.7 9.0 6.1
1.8
.7
8.2
8.4
.9
1.0
6-8
8-10
10-12
17.1 16 9 15.4 5.8 4.2 .4 9.6 6.5 1.3
18.5 17.1 16.5 8.1 6.1 .2 9.2 8.2 .9
~.8 ~.9 12.9 5.9 4.6 4.4 7.9 5.9 12.3
10.1 9.5 6.1 .5 8.7 .7
12.5 9.7 6.3 .4 8.2 .6
13.9 8.9 6.8 1,4 7.3 10.3
359
PRESENTATION AND RECALL TRIALS TABLE 3 MEAN CLUSTERINGINDEXES AS A FUNCTION OF CONDITIONS FOR EXPERIMENT I N u m b e r of Trial Condmon S R P 9P + P 9R + P
1
2
3
4
5
6
7
8
9
10
11
12
P P P P P
4.94 5.45 P P 4.33
P 5.74 P P 4.87
7.11 5.99 7.93 P 6.02
P P P P 5.68
8.87 7.47 P P 6.37
P 8.98 P P 6.31
10.42 9.50 9.79 P 6.07
P P P P 6.02
9.77 10 44 P P 5.69
P 10.27 P P P
11.16 10.35 11.35 9.49 9.58
The analysis of CC data indicated an insigni- index used. When the A R C index (Roenker, ficant main effect for treatment conditions Thompson, & Brown, 1971) as applied the (C), F(1, 26) = .76, p > .05. The main effect for mean score on Trial 10 is .65 and the mean temporal position of the trial-to-trial transi- score on Trial 12 is .74. tion (TP) was significant F(2, 5 2 ) = 25.97, p < . 0 1 . The C x TP interaction was not EXPERIMENT I I significant, F(2, 52) = 1.69, p > .05. Analysis of C N data indicated a significant Experiment II was carried out to determine C effect, F(1, 26) = 8.57,p < .01. Both TP and if postlnput recall on Trial 12 for Group 9R + P the C × TP interaction effects were insignifi- would be equal to Group S when the words cant, F(2, 5 2 ) = .91, p > .05, and F(2, 5 2 ) = were not related. 1.36, p > .05, respectively. Analaysis of N C data indicated that the C Method and C × TP effects were significant, F(1, 26) Subjects. The Ss were 30 male and female under= 37.90, p < .01 and F(2, 52) = 7.61, respect- graduate students enrolled m general psychology at ively. The TP effect was short of significance, Mississippi State University. All Ss participated in order to partially fulfill a course requirement. F(2, 52) = 2.95, p > .05. Materials. The stimulus list employedwas composed Clustering. The clustering index used was of items selected from the norms of Battlg and one presented by Dunn (1969) which is a Montague (1969). The list was comprised of one noun standard score indicating in SD units the ex- from 40 dxfferentcategories for a total of 40 words Eight randomizations of the sequentxal order of the tent to which S departs from a chance number of category repetitions. The mean trial by trial hst of items were generated and typed on tape for presentation by a Stowe memory drum. clustering indexes are presented in Table 3. Procedure The Ss were randomly assigned to one of The clustering indexes observed on all R- the two treatment groups and treated exactly as trials which two or more groups shared described in Experiment l, except, of course, they were temporally were subjected individually to not told the words could be categorized. completely randomized analyses of variance. Since the clustering index is not totally inde- Results pendent of the number recalled, the Group Table 1 presents the mean number of items 9R + P was excluded from these analyses recalled by Ss in the two groups on each recall except for Trial 12. The analyses indicated no trial. Trial 2 and Trial 12 were examined by significant differences on any trial. completely randomized one-way analysis o f It should be noted that the apparent sharp variance. For Trial 2, F(1, 28) = .37, p > .05, increase in clustering from Trial 10 to Trial 12 and for Trial 12, F(1, 28) = 2.19, p > .05. for Group 9R + P is partially an artifact Table 2 presents CC, NC, and C N data resulting from a limitation of the clustering observed on the 2-4, 4-6, 6-8, 8-10, and 10-12
360
HUDSON, SOLOMON, AND DAVIS
trial transitions. These are presented for relative comparison only as the absolute number of words recalled differed for the two groups on most recall trials.
tmue to add small increments with each presentation until they recalled all words. Consideration of the patterning of item recall data indicates that the processes operating are different for the various groups. Inspection of Table 2 makes it clear that Group S both DISCUSSION acquired (NC) and forgot (CN) considerably The first thing to note regarding word recall more than Group R and, in the first cycle, is that Groups S and R do not differ on im- Group R retained (CC) considerably more than mediate postinput recall trials (Trials 2, 6, Group S. When Group 9R + P is considered and 10). This is consistent with the findings of it as clear that on the 2-4 and 6-8 transitions both Tulvmg (1967) and Rosner (1970) using this group forgot little and acquired little. unrelated words. Thus on immediate post- Thus almost all of their recall was retention input recall whether using related or unrelated of previously recalled material. This suggests words the effects of P- and R-reals are that Tulving (1967) may have been incorrect essentially equivalent on total recall for the in his assertion that input and output events are sequences of P- and R-trials studied. When the approximately equally responsible for forgetlevel of orgamzatlon is high as m Experiment I tmg of individual items from one output phase all comparisons between Groups S, R, and P to another. In fact the data of Experiment I except Trial 4 are insignificant whether would suggest the hypothesis that the intraimmediate postinput or not. In addition, on cycle forgetting of Group R was a residual Trial 12 there are no slgmficant differences effect of the P-trial starting the cycle since between any of the five groups even though Group 9R + P forgot essentially nothing on Group 9R + P has received only two presen- the 6-8 transition but forgot an amount equltations and Group 9P + P has received 11 valent to the S and R groups on the 10-12 presentations. The same thing was found transition when a P-trial intervened. The data would also suggest that storage when unrelated words were used (Experiment II); on Trial 12 Group S and Group 9R + P processes have not been given enough attention. It does not seem reasonable that the did not differ significantly. It should be pointed out, however, that one dramatic increase shown by Group 9R + P could not expect to add an indefinite number from Trial 10 to Trial 12 is primarily the result of recall trials for a 9R + P-type group and of a change in retrieval processes. Rather it would seem that the presentation has added have post-input recall be equal to a S group In Experiment II, for example, the mean to the store. It must be remembered that the Ss of Experiment I were clustering at a high number of words recalled by the 9R + P group on Trial 2 was 10.6. The mean number rate. Yet on Trial 2 Group 9R + P recalled a of words added (NC) from Trial 10 to Trial 12 mean of 2.1 words per category recalled, on was 10.3. Thus the Ss of Group 9R + P can Trial 10 they recalled a mean of 2.3 words per retain about 10 words from one presentation category recalled and on Trial 12 they recalled o f the list. These same 10 words are then a mean of 3.1 words per category recalled.Thus recalled for several trials so that they are highly the increase in words per category from Trial overlearned. The Ss are then able, following a 10 to Trial 12 accounted for more than half second presentation, to recall approximately the total increased recall from Trial 10 to 10 new words plus the 10 overlearned words. Trial 12. Thus, given the high degree of clusThis would hkely be the case no matter how tering it seems reasonable to assume that the many recall trials intervened (past some cue for retrieval was present on Trial 10 but the minimum number) but Group S should con- TBR word was not in storage. High organiza-
361
PRESENTATION AND RECALL TRIALS
tion was present even on the first recall real. The fact that G r o u p S slowly increased the mean number of words per category from 2.2 on Trial 2 to 3.1 on Trial 12 suggests that retrieval is limited by content as well as the number of accessible memory units. It seems quite reasonable to assume that the S in a multitrlal F R L task will be forming new units as well as "filling out" existing units. One cannot, of course, necessarily assume a one-toone correspondence between subjective units and E-defined units. Perhaps a modification of Experiment I using cued recall (Tulving & Pearlstone, 1966) could more clearly examine whether the difference noted here is the result of changes in retrieval or storage. The present data for G r o u p 9R + P for both Experiments I and II show a great deal of stability. The S essentially recalls the same words for nine trials. The recall for Trial 12 consists of most of the overlearned words plus an increment essentially equivalent to the recall following the first presentation trial. Thus whereas the pattern of item recall for G r o u p S suggests a high "trade off" of forgotten for acqu,,red words, G r o u p 9R + P repeatedly recalls the same words. The situation for G r o u p R is intermediate between G r o u p S and G r o u p 9R + P. Thus it is clear that even though the net effect of P- and Rtrials is the same on immediate post-input recall, P- and R-trials do not function identically. It should be noted that the recall of G r o u p R did not decrease across the three intracycle recall trials of any phase and, in fact, showed a shght increase in the second and third phases. This is different from the results of Tulving (1967) and Rosner (1970) who used unrelated words. This was probably due to the high degree of organization exhibited by the Ss of the present study. This high degree of organization was probably also responsible for the high degree of item retention (CC) noted in Experiment I as compared with Experiment II. The clustering data indicated that for taxo13
nomic hsts there is essentially no difference in the degree of E-defined organization as a result of numbers or pattern of R-trials and P-trials as they were mampulated in the present study. This would appear to be at variance with the findings of Rosner (1970) who found that Rtrials facilitated subjective organization more than P-trials. Visual inspection of the present data would indicate that m a n y of the Ss in G r o u p R, after a P-trial, would recall "new" words first and follow this with "old" words rather than fitting "new" words into existing categories. The effect of this would be a more rapid increase in subjective organizatmn than E-defined organization REFERENCES BATTIG, W. F., & MONTAGUE,W. E. Categoly norms for verbal items in 56 categories: A replication and extenslon of the Connecticut Category norms. Journal of Experimental Psychology, 1969, 80, (3, Pt 2). BOUSFIELD,W. A., PUFF, C. R., & COWAN, T. M. The development of constanczes an sequential orgamzation during repeated free recall. Journal of Verbal Learning and Verbal Behavior, 1964, 3,
489-495 DUNN, J. E. A compound multiple runs dlstnbutmn Journal of the American Statistical Association,
1969, 64, 1415-1423. KIRK, R. E. Experimental Destgn" Procedures for the behavtoral sciences. Belmont, Califorma: BrooksCole Publishing Company, 1968. ROENKER,D L, THomPSON,C. P., & BROWN, S. C. Comparison of measures for the estimation of clustering m free recall. Psychological Btdletm, 1971, 76, 45-48. ROSNER,S. The effects of presentation and recall reals on orgamzatlon m multitrial free recall. Journal of Verbal Learning and Verbal Behavior, 1970, 9, 69-74. TUL'nN~, E. Subjectwe orgamzation in free recall of "unrelated" words. Psychological Review, 1962, 69, 344-354. TCLVING,E. The effects of presentation and recall of materml an free-recall learning Jota'nal of Verbal Learning and Verbal Behavtor, 1967, 6, 175-184. TULVING, E., ~; PEARLSTONE, Z. Availability versus accessibility of informatmn an memory for words.
Journal of Verbal Learning and Verbal Behavior, 1966, 5, 381-391. (Received November 19, 1971)