The effects of presentation and recall trials on organization in multitrial free recall

The effects of presentation and recall trials on organization in multitrial free recall

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 9, 69-74 (1970) The Effects of Presentation and Recall Trials on O r g a n i z a t i o n in M u l t i...

460KB Sizes 42 Downloads 94 Views

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR

9, 69-74 (1970)

The Effects of Presentation and Recall Trials on O r g a n i z a t i o n in M u l t i t r i a l Free Recall z SUE R. ROSNER2 University of Connecticut, Storrs, Connecticut The effects or presentation (P) and recall (R) trials on organization in multitrial free recall were investigated by independently varying the sequences of P-trials and R-trials within eight-trial cycles. An S-unit task, in which Ss indicated their subjective groupings of the recall items, was administered at the end of two successive cycles. The major findings were: (a) P-trials and R-trials produced an equivalent number of two-item units in recall and the same types of S-units; (b) R-trials revealed significantlyhigher sequential ordering of the two-item response units and a greater cohesivenessof the S-unit elements on the recall trials. The results suggested that P-trials and R-trials contributed equally to unit formation while R-trials had a larger effect than P-trials on unit consolidation. were presented at a 1-second rate and test trials were allotted the equivalent period of 36 sec. for item recall. The findings indicated that within a cycle P-trials produced an increase in recall and R-trials led to some decline in item recall. Also, following a common P-trial in each successive cycle, Condition R tended to "catch up" and showed, temporarily, the same amount of item retention as the other conditions before the intracyclical recall divergences recurred. This pattern of intracyclical divergences and intercyclical convergences in item recall was maintained throughout the experiment. The finding that R-trials significantly affected item retention was also shown by Lachman and Laughery (1968). Lachman and Laughery found that recall performance following a standard condition of alternating P R trials was similar or greater than that obtained following a corresponding number of spaced P-trials. The purpose of the present experiment was to study the effects of P-trials and R-trials on organization in MTFR. It involved adapting Tulving's paradigm of three types of trial sequence to the problem of organization by modifying the composition of the cycles. This was done by making the following changes:

Multitrial free recall (MTFR) experiments (Bousfield, Puff, & Cowan, 1964; Tulving, 1962) have shown that over repeated trials there is an increase in both the number of items correctly recalled (item retention) and in the regularity of the pattern of item emission (organization). Since the standard M T F R procedure employs alternating presentation (P) and recall (R) trials, this method cannot effectively be used to differentiate the effects of P-trials and R-trials on item retention and organization. Recently Tulving (1967, Exp. II) compared the effects of P and R on item retention in an M T F R experiment by varying the sequence of P- and R-trials within cycles of four trials each. Three types of trial sequence were used: PRPR, PPPR, and P R R R . These three types of sequences are here designated respectively as Conditions PR, P, and R. For each condition, learning was extended for six cycles of the assigned four-trial sequence. The stimulus list comprised 36 unrelated words which 1 Basedon a dissertation submitted to The University of Connecticut in partial fulfillment of the requirements for the Ph.D. degree. 2 Now at the Institute of Child Behavior and •Development, University of Iowa, Iowa City, Iowa 52240. 69

70

ROSNER

(a) t e r m i n a t i n g a cycle with two successive recall tests to permit the application of a n organizational measure; (b) initiating a cycle with two successive presentation trials so that the initial level of recall would be raised; (c) interposing between the initial PP a n d terminal R R trials of a cycle one of three i n d e p e n d e n t trial sequences, viz., P R P R , PPPP, R R R R . The p h e n o m e n o n of organization was evaluated by applying the I T R measure to the recall output, that is, c o u n t i n g the n u m b e r of obtained intertrial repetitions of sequential word-pairs (Bousfield et al., 1964). Organizational processing was assumed to involve the establishing of S-units, the grouping of individual list items into subjective units or chunks (as suggested by Tulving, 1968). Therefore Ss were asked, after the M T F R task, to indicate their subjective units a n d these " S - u n i t s " were examined a n d c o m p a r e d with the recall records as a further measure of organizational effects.

METHOD Trials and Conditions. On P-trials, a Carousel projector exposed the word items on a screen. Each of the 15 words from Deese's (1959) List 18 having zero normatively indicated interitem associative strength, was exposed for 3 sec. with 0.4 sec. between words. Five different randomized arrangements of the word list were used. The total time for administering a P-trial, from the P-trial signal ("Look") to its end with the exposure of a blank slide, was 1 rain. The time period for an R-trial, from its starting signal ("Write") to its termination signal ("Stop Writing"), was also 1 min. The items of each R-trial were recorded by S on a separate lined page of a booklet. A constant intertrial interval of 10 sec. was used between all trials, both within and between cycles. There were three independent conditions which differed in the sequencing of presentation (P) and recall (R) trials within a cycle of eight trials: (a) Condition PR, PP-PRPR-RR; (b) Condition P, PP-PPPP-RR; (c) Condition R, PP-RRRR-RR. Each condition was imposed for two successive cycles (Cycles I and II) of the same eight-trial sequence so that the relative effects of P-trials and R-trials could be evaluated at two levels of list learning. The complete trial sequence for each condition is shown in Table 1 which appears in the Results section.

The following conventions are here employed in designating the trials for the various conditions. The trials are labeled P or R depending on the nature of the trial. They are numbered consecutively in accord with the temporal period of the trial. Since P- and R-trials were of equal duration, the concurrent recall tests evaluated performance as a function of differences in the preceding trial sequences. Instructions. The same instructions, which described the nature of the MTFR task and the characteristics of the P- and R-trials, served for all conditions. The Ss were told that trial order would vary and that the type of trial (P or R) would be announced at the start of each trial. Subjects. Undergraduate student Ss were tested in small groups. The median number of Ss per group for each condition was 6.5, 6.0, and 5.0 for Conditions PR, P, and R, respectively. A total of 69 Ss participated. Of these, nine were dropped because of such irregularities as failing to follow instructions. This left 20 Ss for each of the three conditions. S-Unit Task. What is here termed the S-unit task was administered at the end of Cycle II after the recall booklets were collected. Individual data sheets presented the instructions for the task and a randomized word list. The Ss were told to write down and to bracket as sets the words from the list that had been remembered together on the recall trials. It was specified that the same items could be included in more than one set and that Ss were to work rapidly. RESULTS Analyses of variance showed that the subgroups within each of the three experimental conditions did n o t significantly differ o n either item recall or on the I T R organizational measure. The data from the subgroups u n d e r each c o n d i t i o n were therefore pooled for subsequent statistical analyses. Item Recall. Table 1 presents the m e a n n u m b e r of items correctly recalled by Ss in all three conditions o n all recall tests. It may first be noted that the overall recall levels were approximately two-thirds of the list in Cycle I a n d tended to be asymptotic in Cycle II. The c o m m o n test recall scores for C o n d i t i o n s PR, P, a n d R on the terminal test trials, Trials 7 a n d 8 in Cycle I a n d Trials 15 a n d 16 in Cycle II, were subjected to a n arcsin transf o r m a t i o n a n d a n analysis of variance. Significant effects were o b t a i n e d for Trials,

71

ORGANIZATION IN FREE RECALL TABLE 1

P- AND R-TRIAL SEQUENCESFOR CONDITIONSAND MEAN NUMBER OF ITEMSCORRECTLY RECALLED AS A FUNCTION OF CONDITION

Cycle I Initial Ps

Cycle II

Intracyclical Ps or Rs

Terminal Rs

Initial Ps

Intracyclical Ps or Rs

Terminal Rs

Condition R P1

P2

R3 10.3

R4 10.1

R5 10.3

R6 10.3

R7 R8 10.6 10.5

PI

P2

P3

R4

P5

- -

1 0 . 8

- -

R6 11.9

R7 R8 12.0 11.8

P3 .

P4

P5 .

P6

R7 R8 11.7 11.5

P9 P10

Rll 13.5

R12 13.6

R13 13.6

R14 14.1

R15 R l 6 13.9 14.0

Pll --

R12 13.6

P13

R14 14.0

R15 R16 13.9 14.0

14

R15 R16 13.8 13.6

Condition PR P9 P10

- -

Condition P P1

P2

.

.

F(3,171) = 165.97, p < .001, and for the Condition x Trial interaction, F(6,171) = 8.81, p < .01, but not for the condition variable of type-of-trial, F = . 3 3 . Scheff6 comparisons verified the impression from Table 1 that the source of the variance in the interaction term was the lower recall for Condition R as compared with that of Conditions PR and P on Trials 7 and 8 in Cycle I (p < .01). Neither the Cycle I difference between PR and P nor the Cycle II differences among conditions approached statistical significance. Intracyclical retention comparisons between Conditions PR and R on the concurrent tests in Cycle I also indicated that P-trials increased item recall. There was no item recall difference between conditions on Trial 4, t = . 8 9 . Trial 6, which followed an intervening P-trial for Condition PR and an intervening R-trial for the R Condition, showed significantly more recall for Condition PR, t(38) = 2.36, p < .05. A trial-to-trial analysis as suggested by Tulving (1964, 1967) was applied to the individual items recalled by Ss in Conditions P R and R on Trial 4 and on Trial 6. The amount of common item recall, i.e., the mean number of individual items correctly recalled in common on both trials, C4C6, w a s the same

P9 P10

Pll .

.

P12P .

P13 .

(9.4 items) for the two conditions. Both item f o r g e t t i n g , C 4 N 6 (the mean number of items recalled on Trial 4 that were not recalled on Trial 6), and item acquisition, N4C 6 (the mean number of items recalled on Trial 6 that were not recalled on Trial 4) were significantly higher for Condition PR than for Condition R, with t's(38) of 2.68 ( p < . 0 2 ) and 4.01 (p < .001), respectively. The intracyclical recall levels of Conditions PR and R were obviously equivalent on Trials 12 and 14 in Cycle II (see Table 1). Condition R showed a relatively greater intercyclical recall increment than Condition PR since the recall level of Condition R was significantly lower than that of Condition PR on the terminal tests in Cycle I but identical for the two conditions on the first concurrent postinput trial, Trial 12. Thus, the item recall for Condition R, as compared with that of Condition PR, benefited more from the two initial P-trials in Cycle II. It is apparent from Table 1 that Condition R maintained a stable recall level over successive R-trials within both cycles. Condition R also showed relative stability in the recall of individual items over trials. Of the total number of individual items that were recalled

72

ROSNER

at least once in the first three recall trials, 79 % were correctly recalled on all three trials. On Trials 6, 7, 8, the percentage of the total item recall that comprised the common recall of individual items was 89 %. The comparisons between conditions on the terminal and on the intracyclical tests suggest the following relationships between trial-type and item recall during initial learning: P-trials provided additional items for recall; P- and R-trials resulted in equal common item recall and in equal recall on postinput tests; R-trials decreased the forgetting of individual items which had been previously emitted and produced a higher intercyclical increment in recall.

Organization. The number of obtained ITR units was calculated according to the method of Bousfield et al. (1964). An ITR unit was indicated whenever an item "B" followed another item " A " on two successive recall trials. The mean number o f l T R units obtained on successive recall trials for each condition is presented in Figure 1. All of the obtained ITR scores plotted in Figure 1 may be assumed to CYCLE TT

8

7

/

26 Z

~s Z

/ R

'-'-'71~.p

R .p

CYCLE 3"

"R

~4 i

,

i

!

i

i

I

I

3"4 4-5 5-6 6-7 7-8 11-1212-1313-1414-1515-16 TEMPORAL PERIODS OF RECALL TRIALS

FIG. 1. Mean number of I T R units on successive recall trials for all conditions. The trial numbers on the abscissa refer to the temporal period of the recall trials involved in each successive pair of recall tests.

be beyond the level of ordering expected by chance since the lowest ITR score exceeded the 99 % confidence limits. It is evident from Figure 1 that increasing the number of R-trials per cycle increased the level of ITR. An

analysis of variance on the ITR scores for the common terminal Trial Pair 7-8 in Cycle I and Trial Pair 15-16 in Cycle II indicated significant effects for Condition, F(2, 57)= 3.69, p < .05, and for Trials, F(1, 57) = 42.18, p < .01, while Condition x Trial interaction did not reach significance, F(2, 57)= 1.60, p > .05. The significant source of the condition effect, based on Scheff6 comparisons of the mean number of ITR units for the three conditions over cycles, was due to the difference between Condition R and Condition P (p < .01). The ITR analyses thus indicated that although the number of sequentially ordered response pairs increased over trials for all conditions, Condition R showed a higher ITR level than Condition P in both cycles. S-Unit Task. The S-units reported by Ss in the three conditions were essentially the same in terms of: the mean number of S-units formed (4.70, 5.20, 5.15 for Conditions PR, P, and R, respectively); the frequencies with which various sizes of S-units were reported (X2=4.01, df=6, . 7 0 > p > . 5 0 ) ; and the mean number of times the same item was repeated in more than one S-unit (KruskalWallis H = 3 . 2 5 , df=2, . 2 5 > p > . 1 0 ) . It is evident that there was no differential effect for the characteristics of the S-units emitted in the S-unit task. The Ss' recall protocols of the terminal trials in Cycle II were examined to determine whether items in recall were chunked according to the item groupings indicated on the S-unit task. The congruence between the item composition of the S-units and item emission order in recall will be referred to here as "S-Unit Items in Recall" or"SIRs". Individual SIR scores were tabulated by counting for each S-unit the proportion of S-unit items that occurred as a continuous run in recall, summing these proportions over S-units and then dividing by the total number of S-units formed by that particular S. The SIR scores varied as a function of condition. For Trial 16 the average SIR scores of Conditions PR, R, and P were .87, .88, and .65, respectively. The

73

ORGANIZATION IN FREE RECALL

Conditions' SIR scores on Trial 15 were almost identical to those of Trial 16. The SIR scores for Condition R on Trial 16 were significantly higher (p < .02) than for Condition P as indicated by the Mann-Whitney test, Z(38) = 2.41. This indicates that Ss in Condition R showed a higher congruence between their defined S-units and recall output than did Ss in Condition P. For all conditions there was a statistically significant, though small, relation between I T R units on Trial Pair 15-16 and SIR scores on Trials 15 and 16. The Kendall Taus for Conditions PR, P, and R were: .41, p < .04; .48, p < .02; .53, p < .01 (all df's were 38).

Commutative ITR. The commutative intertrial repetition (C-ITR) counts item pairs whose members occur sequentially ordered on two successive trials, i.e., the traditional ITR, and, in addition, item pairs whose members appear in reverse order on two adjacent trials. A tabulation was made of the number of C - I T R units in Cycle I as a function of condition. The basis for this posthoc analysis was the finding that the various conditions had formed the same number of S-units on the S-unit task. The purpose of the C - I T R analysis was to determine whether the conditions had similarly produced the same number of two-item units in recall. It was assumed that two-item recall units would be representative of S-unit size only in initial learning. The C - I T R analysis was therefore limited to recall tests in Cycle I. Neither the C - I T R difference between Conditions R and P on the terminal tests (Trial Pair 7-8, t = .60), nor the C - I T R difference between Conditions PR and R on the concurrent intracyclical tests (Trial Pair 4-6, t = .94), approached significance. It should be noted that R-trials had produced a significantly higher level of traditional I T R than P-trials on the terminal Cycle I test and also on the intracyclical test, Trial Pair 4-6, t(38) = 2.25, p < .05. The combined results of tile C - I T R and the I T R analyses thus indicated that P- and R-trials led, in initial learning, to

an equivalent number of two-item units while R-trials tended to produce a larger number of unidirectionally ordered or serialized two-item units than P-trials. DIscussION The present findings on the differential effects of type of trial on organization have shown: (a) R-trials produced significantly greater sequential ordering than P-trials as indicated by the ITR. (b) P-trials and R-trials did not differ in the characteristics of the S-units as measured by the S-unit task. (c) R-trials led to a significantly greater degree of correspondence between the reported S-units and the pattern of item emission in recall, as assessed by the SIR scores, than P-trials. The results of the various measures applied to evaluate organization did not show concordance. For the dependent variables of I T R units and SIR scores, R-trials had a greater organizational effect than P-trials. Analyses of S-unit performance and of the commutative or C - I T R results revealed that P-trials and Rtrials operated similarly on subjective organization of the list items. It is conjectured that different aspects of organizational processing are indicated by the S-unit and C - I T R measures, on the one hand, than by the I T R and SIR scores, on the other hand. The S-unit and C - I T R measures seem to refer to the establishing of item groupings, or unit formation, while the I T R and SIR measures seem to denote the unit properties of stability and cohesiveness, or unit consolidation. The organizational results may thus be interpreted as showing that P-trials and R-trials exert similar influence on unit formation while R-trials have a greater effect than P-trials on unit consolidation. This interpretation also serves as a basis for explaining the obtained effects of type of trial on item recall. It seems reasonable to assume that the greater intercyclical increment shown by R-trials is produced by the occurrence o f higher intracyclical unit consolidation on

74

ROSNER

R-trials. It also appears plausible to attribute suggested by Tulving (1967) and Lachman the similarity in common item recall and in and Laughery (1968), but also operate to postinput recall shown by P-trials and R-trials facilitate organization. The organizational to an equivalence in unit formation on P-trials effects of R-trials appear to be partly similar and R-trials. to that of P-trials and partly of a specialized Unfortunately, these interpretations cannot nature. By specifying organizational effects be evaluated by reference to other experiments as unit formation and unit consolidation, dealing with organizational trial-type effects. interpretations of type-of-trial effects on recall However, the typicality of the differential and organization are feasible. Moreover, the retention effects obtained in this experiment hypothesis offered here suggests explicitly that may be examined by comparing them with S-units, like items, show oscillation, instability, those of Tulving's (1967) study. Although the and differential strength during recall. present experiment's design was an adaptation of Tulving's, the two studies differed in the REFERENCES number of trials per cycle, the sequencing of trials within cycles, the number of cycles, the BOUSFIELD,W., PUFF,C., & COWAN,T. The development of constancies in sequential organization length of the word list, the duration of item during repeated free recall. Journal of Verbal exposure, and the time allotted for item recall. Learning and Verbal Behavior, 1964, 3, 449-459. Nevertheless, all of the type-of-trial effects DEESE,J. Influence of inter-item associative strength upon immediate free recall. Psychological Reports, revealed for the retention phenomena of com1959, 5, 305-312. mon item recall, terminal test recall, postinput recall, and intercyclical increments in recall LACHMAN, R., & LAUGHERY, K. R. Is a test trial a training trial in free recall learning? Journal of had previously been demonstrated by Tulving. Experimental Psychology, 1968, 76, 40-50. The one divergence in results--the R-trial TULVIN%E. Subjective organization in free recall of "unrelated" words. Psychological Review, 1962, Condition in this study showed more stability 69, 344-355. in item recall level and in the individual items TULVING, E. Intratrial and intertrial retention: Notes recalled on successive tests--was probably due towards a theory of free recall verbal learning. to the longer processing time available in the Psychological Review, 1964, 71, 219-237. present experiment. The overall comparability TULVING,E. The effects of presentation and recall of material in free-recall learning. Journal of Verbal of the retention results increases the probaLearning and VerbalBehavior, 1967,6, 175-184. bility that the organizational effects shown in this study may be generalizable to other TULVING, E. Theoretical issues in free recall. In T. Dixon and D. Horton (Eds.), Verbal Behavior and experimental conditions. general behavior theory. Englewood Cliffs, N.J. : In conclusion, the present experiment Prentice-Hall, 1968. indicates that test trials in the M T F R experiment function not only to facilitate recall, as (Received September 3, 1969)