DOCUMENT RESUME

ED 094 371 CS 001 282

AUTHOR Bruce, DarrylTITLE Selective and Intc.'tional Forgetting. Final

Report.INSTITUTION Florida State Univ., Tallahassee. Dept. of

Psychology.SPONS AGENCY National Inst. of Education (DHEW), Washington, D.C.

Office of Research.BUREAU NO BE-0-0520-FRPUB DATE Jan 74GRANT O!G -4 -70 -0038 (508)NOTE 43p.

EDRS PRICE MF-$0.75 HC-$1.85 PLUS POSTAGEDESCRIPTORS *Cognitive Processes; *Educational Research;

Learning; Learning Processes; Literature Reviews;*Memory; *Recall (Psychological); Recognition;*Eetention

ABSTRACTThis paper reports on the results of several

experiments concerned with instructions to forget certain informationand to remember other information presented in the context of avariety of laboratory tasks of short-term memory. Assessment of theretention of remember material indicated that it varies directly withthe degree to which clear cognitive differentiation of forget andremember material is permitted and the extent to which one is able todevote all rehearsal activity to remember material. These mechanismsof differential grouping and differential rehearsal were alsoimplicated by the observation that information one is directed toforget is not expunged from memory but is still largely available instorage. The effects of these two processes permit the lateroperation of a third directed forgetting mechanism, selective search(focusing at retrieval time on the to-be-remembered information). Theresults of various studies of non-intentional forgetting are alsoreported. The significant conclusions which they suggest include:tip-of-the-tongue states are relatively transient, the temporaldating of to-be-remembered events may not be important to theirretrieval, and selective rehearsal may be at the root of a number ofsituations in which short-term memory is facilitated. (Author /NR)

US DEPARTMENT OF HEALTHets EDUCATION& WELFARE

NATIONAL INSTITUTE OFEDUCeriON

fit F rr Plr0DFri FirAi'f., i4FCF.10 lgrOY

Final Report e,0 LK', 14E-P4t

NyrirL,r. orD cAr,iir roti Drir noL CV

f`e\

-4-

Cr"CD Project No. 0-0520

CZ) Grant No. 0EG-4-70-0038(508)

Darryl BruceDepartment of PsychologyFlorida State UniversityTallahassee, Florida 32306

SELECTIVE AND INTENTIONAL FORGETTING

January 1974

SCOPE OF INTEREST NOTICE

The ERIC Facility has assignedthis document for processingto

In our Iudderne , this documentis also of interest to the clearinghouses noted to the right. Index-ing should reflect their specialpoints of view.

tS1

M N\k015..

Vo

U.S. DEPARTMENT OF HEALTF, EDUCATION, AND 1,7ELFARE

National Institute of Education

Bureau of Research

Author's Abstract

Instructions to forget certain information andto remember other information were presented in thecontext of a variety of laboratory tasks of short-term memory. Assessment of the retention of remembermaterial indicated that it varies directly with thedegree to which clear cognitive differentiation offorget and remember material is permitted and theextent to which one is able to devote all rehearsalactivity to remember material. These mechanisms ofdifferential grouping and differential rehearsal werealso implicated by the observation that informationone is directed to forget is not expunged from memorybut is still largely available in storage. The effectsof these two processes permit the later oneration ofa third directed forgetting mechanism, selectivesearch (focusing at retrieval time on the to-be-remembered information).

The results of various studies of nonintentionalforgetting are also reported. The significant conclu-sions which they suggest aye to be the following:(1) tip-of-the-tongue states are relatively transient;(2) the temporal dating of to-be-remembered eventsmay not be important to their retrieval; (3) selectiverehearsal may be at the root of a number of situationsin which short-term memory is facilitated, e.g., whereto-be-remembered events are perceptually isolated.

Final Reriort

Project No. 0-0520Grant No. CEG-4-70-0038(508)

SELECTIVE AND INTENTIONAL FORGETTING

Darryl Bruce

Department of PsychologyFlorida State UniversityTallahassee, Florida 32306

January 1974

The research reported herein was performed pursuantto a grant with the Naticnal Institute of Education,U.S. Department of Fealth, Education, and Telfare.Contractors undertaking such projects under Govern-ment sponsorship are encouraged to express freelytheir professional judgment in tl-e conduct of theproject. Points of view or opinions stated do not,therefore, necessarily represent official NationalInstitute of Education position or policy.

U.S. DEPAR=NT OF HEALTH, EDUCATION, AND T,TELFARE

National Institute of EducationBureau of Research

Preface

It is a pleasure to acknowledge the assistanceof the following individuals in the research performedunder the aegis of this grant: DiAnne Bradford,Steve Cary, Paul Freeman, Tim Gaines, Jim Goodwin,Susan Gore, Arthur Graesser, Carl 'Kennedy, DebbieMoehle, and Robert ;7hitney.

TABLE OF CONTENTS

GENERAL INTRODUCTION

INTRODUCTION TO STUDIES OF INTENTIONALFORGETTING

Page

1

2

EXPERIMENTS ON INTENTIONAL FORGETTING 5

CONCLUSIONS OF EXPERIMENTS ON INTENTIONALFORGETTING 26

EXPERIMENTS ON NONINTENTIONAL FORGETTING 28

CONCLUSIONS OF EXPERIMENTS ON NONINTENTIONALFORGETTING 36

BIBLIOGRAPHY 37

LIST OF TABLES

Table 1. Recall of the Eight Lists of

Page

Categorical and Unrelated !,qords 7

Table 2. Recall of the Final List 8

Table 3. Recall of Unrelated and CategoricalRemember Words 10

Table 4. Mean Numbers of Hits (H) and FalseAlarms (FA) for Unrelated andCategorical Words 12

Table 5. Means and Standard Deviations (inParentheses) for Immediate andFinal Retention Tests 14

Table 6. Recall Proportions for CriticalWords 17

Table 7. Proportions of Hits and False Alarmson Immediate Recognition Tests 21

Table 8. Proportions of Hits and False Alarmson Final Recognition Test 22

Table 9. Performance on Immediate RecallTests 24

Table 10 . Mean Word Recall and Proportions ofWords Identified as Remember Items(in Parentheses) on Final RecallTest 24

GEUFRAL INTRODUCTIOV

The general concern of this grant has been withproblems of huran memory. The approach to these pro-blems has sometimes involved simple laboratory tasksin which rather than instructing a subject to remem-ber material, he is instead directed to forget infor-mation intentionally. The underlying belief has beenthat such an emphasis may yield insights into thehuman memory system which would not otherwise havebeen forthcoming through more traditional procedures.In addition, forgetting is such a pervasive phenomenonas to suggest that it has useful or adaptive charac-teristics. Thus questions concerning how one forgetsinformation voluntarily or intentionally were ofespecial interest in this research. The descriptionof our investigations begins with those studies whichhave used a directed forgetting procedure. Thenfollows a second major division covering studies ofnonintentional forgetting.

1

INTRODUCTION TO STUDIES OF INTENTIONAL FORGETTING

A Procedure and Previous Results

The methods and results of experiments in whichsubjects are instructed to forget information deli-berately have been reviewed in monographs by Bjork(1972) and Epstein (1972) and the interested readeris directed to these sources for a complete intro-duction to the topic. For present purposes, it willsuffice to give an example of an intentional forgettingprocedure and to mention the sorts of observationscommonly made. A very simple situation which we haveused involves presenting lists of words sequentiallyafter each of which the subject is tested by themethod of free recall, that is, he is asked to recallas many of the words as possible in any order he wishes.To turn this into an intentional forgetting task, wecan inform the subject that during some of the listsbut not all of them, a signal (e.g., a buzzer) willoccur after one of the list members. The subject isfurther instructed that this means to forget allitems prior to the buzzer (the forget words) and toremember only those subsequent to the buzzer (theremember words). He is not told, however, as to whetherthe particular list about to be presented will con-tain a forget cue.

Concerning the results of this kind of procedure,arranging things such that lists with and without cuesinclude the same number of remember words permits asimple comparison between the mean number of suchitems retrieved from forget-cue and non-forgetcuelists. This gives an idea of the extent to whichforget words interfere with the recall of rememberwords. Very often the interference is negligibleindicating positive or functional forgetting of forgetmaterial. However that such material is still avail-able in storage and is not forgotten is clearlyrevealed by direct tests of forget items. The pro-blems inherent in this approach are perhaps obvious;but in the context of our illustrative procedure, theycan be circumvented in part by administering such atest after a final forget-cue list of a laboratorysession, that is, when contradicting the rules ofthe task can no longer matter. By way of a generalsummary statement concerning the results of intentionalforgetting studies, it would appear that forget items,while still largely available in storage, need nothave any substantial interference effect on theretrieval of remember iters.

2

Mechanisms of Intentional Forgetting

We shall mention five mechanisms of intentionalforgetting which have been proposed. One is an erasureprocess. By this is meant that the human memory systempossesses the capacity to expunge information frommemory upon command.

A second mechanism applies to situations inwhich what are forget materials for one group ofsubjects must be maintained in mind by a comparisongroup while remember materials are being tested. Theidea here is that we have a limited capacity retrievalmechanism and that the presentation of a forget cuereleases this mechanism for attention to only thoseitems which must be remembered thereby fan-Matingperformance. This has also been referred to as arecycling hypothesis by Epstein (1969). By this ismeant that a forget instruction can improve testperformance because it liberates one from having torecycle forget items for later retrieval while oneis searching for remember items.

Three other mechanisms may he discussed simul-taneously, inasmuch as they do not appear to hemutually exclusive of one another. These threemechanisms are differential grouping, differentialrehearsal, and e2lectiVE--seaich. one differentialgrouping notion is a sforage process and means thatforget and remember material are tagged differentiallyor grouped in storage such that any interferencebetween the two kinds of items is substantially reduced.Differential rehearsal is the notion that upon theoccurrence of a forget instruction all rehearsalenergies are oriented toward remember informationand away from forget information. The idea of selec-tive search is an emphasis on the importance of retrievalprocesses. The argument is that a forget signal maypermit a subject to restrict his search to a smallerwell-demarcated set of items thereby improving per-formance. The interrelatedness of these mechanismscan be readily appreciated. For instance, in orderfor differential rehearsal to occur, the forget andremember sets must be differentiated in mind. Differ-ential rehearsal, in turn, would appear to be condu-cive to establishing sets of items upon which a selec-tive search process can operate. Finally, selectivesearch seems to presuppose the existence of sets ofitems which have been differentiated in memory.Speaking more generally, the occurrence of any oneof these processes seems to imply the operation ofat least one of the remaining ones.

3

We turn now to a description of our experimentson directed forgetting. While the research bearsmainly on the notions of differential rehearsal anddifferential grouping, it has irplications for all fiveof the mechanisms just mentioned and these will bepointed out as each experiment is presented.

4

EXPEIMENTS ON INTENTIONAL FOnGETTING

Experiment 1

The main concern of this experiment was with thedirected forgetting mechanism of differential grouping.It has been shown that the interference effects ofan initially presented set of items on the recall ofa second set of items can be largely diminished ifone is instructed to forget the first set of items(e.g., Bjork, 1970). The question raised in thisexperiment was how effective a forget cue would bewhen the discriminability between the two sets ofitems was reduced. To effect such a reduction twovariations were tried. One was to make the rememberand forget items categorically similar; the other wasto institute a delay prior to recall.

Pethod. The general procedure of this experi-ment iTig-Ige alternate presentation and test by themethod of free recall of different lists of words.In some of the lists, but not all of them, a forgetcue (a dollar sign for half the subjects and threeasterisks for the other half) was interpolated betweentwo of the words. This meant that the subject wasnot responsible for recalling any of the words whichpreceded the cue, only those which followed it. And,of course, he did not kncw in advance whether or nota list would contain such a signal.

Each of 128 college students saw eight listsof words presented at a 2.5-second rate. These eightlists represented the factorial combination of threeindependent variables. One such variable was thenature of the list members, either common unrelatedwords or categorically related words chosen from theCohen, Bousfield and Whitmarsh (1957) norms. As asecond variable, the written free recall tests aftereach list were administered either immediately orafter a 2-minute interval of solving number-seriesor arithmetic problems.

The final independent variable was the presenceor absence of a forget cue. For lists of unrelatedwords, there were 15 remember items and 15 forget itemswhich antedated any cue. The comparable figures forlists of categorical words were 25 and 25. For thelatter type of material, the remember items were fiveexemplars from each of five categories. And they wereblocked in the presentation sequence. Any forget

5

items in a list were five different exemplars selectedfrom each of the sane five categories and also presentedin block fashion. Thus, a categorical forget cue listcontained forget and remember words which were drawnfrom the same ccnceptual clzsses. The presentationorder of the lists insured that materials and condi-tions were counterbalanced.

Over and above the lists described to this point,there was also a final list which consisted of 30unrelated words with a forget cue after the 15th item.Half the subjects were tested under i'"forget",condi-tion. They were left to believe in what a forget cuehad always meant up to this juncture. However, atthe end of the list and contrary to the originalinstructions, these subjects were asked to recall onlythose words which had been presented prior to theforget cue. The other half of the subjects weretested under a "divide" condition. Before the lastlist, members of this group were told that a forgetcue would occur but that now they should treat thecue as a point of division of the list into two parts,and that they should try remember each part as ifit were a separate list 'hey were also informedthat they would be teF 1 on both parts although theydid not know in which Like the forget group,of course, they were subsequently asked to recall thefirst segment first. Following recall of this setof items, both groups were then asked to write downas many of the post-cue items as they could.

Results and Discussion. Table 1 shows means andstandard deviations for recall of the remember wordsfor the first eight lists. It is clear that the meannumber of unrelated words recalled whether immediatelyor after a delay was independent of the occurrence ofa forget signal, F (1, 120) 1. By contrast, recollec-tion of categorical items for lists containing a for-get instruction was depressed with an apparentlygreater effect when recall was immediate rather thandelayed, F (1, 120) = 54.24 and 8.87, p < .001 and.01, respectively.

There are at least two problems in drawing defini-tive conclusions from these data. First, if oneconsiders the recall of unrelated words, it cannotbe said that the forget cue eliminated interferenceowing to forget items. Rather it could be simplythat recall from no-forget-cue lists was depressedbecause of the uncertainty as to what was going to betested under such conditions. The point can perhaps be

6

Table 1

Recall of the Lic:ht Lists ofUnrelated and Categorical TIords

Test Forget Cue

Unrelatedwords

Categoricalwords

ean SD Mean SD

Immediate Absent 7.1 2.5 15.8 4.1

Present 6.9 2.5 13.4 3.7

Delayed Absent 5.1 2.7 12.8 4.7

Present 5.2 2.6 11.9 3.7

appreciated by consideration of forget-cue conditionswhere upon the occurrencl of the cue, the subjectknows for certain he is going to be responsible forrecalling each item he sees thereafter. Of course,precisely the same problem exists with respect tocategorical materials. A second point is that eventhough there was interference in the recall ofcategorical lists when forget items were presented,it could still be the case that the forget cues haveled to some diminution of such interference. Butto know this for certain would require a conditionin which two sets of items are presented without thesubject knowing until the end of the list that it isthe second set he is to recall. Experiment 2 wasdesigned to cope with these two difficulties.

Data concerning recall of the pre- and post-cuesegments of the final list appear in Table 2. Recallof the segment prior to the signal was reliably lowerwhen the cue was a forget instruction, F (1, 126) =6.00, E < .05; the compensatory difference in favorof the forget condition on the second test, however,was not statistically significant, F (1, 126) = 3.49,

7

Table 2

Pecall of the Final List

List segment Cue meaning Mean SD

Pre-cue(test 1)

Forget 3.1 2.7

Divide 4.2 2.6

Post-cue(test 2)

Forget 3.8 3.0

Divide 2.8 2.4

These data accord well with the notion of differ-ential rehearsal. Upon the occurrence of the cue,the assumption is that the divide group would attemptto keep both segments in mind whereas the forgetgroup would rehearse only the second segment, droppingthe first from mind. The obtained pattern of resultsfollows from this analysis. On the other hand, therecycling mechanism discussed earlier does not seemto be a very potent factor. In other words, the dividegroup, even though it had to keep the post-cue segment"alive' while recalling the pre-cue segment, still didbetter on test 1 than did the forget group which wasnot explicitly instructed that it would have to recallthe post-cue segment. In addition, perfect operationof an erasure mechanism is soundly denied by theevidence that forget words can be recalled. Theconclusion, then, is that a forget instruction leadsto a shift toward rehearsal of to-be-remembered itemsand away from rehearsal of those which can be for-gotten. That such differential rehearsal need notaffect overall performance is clear from the fact thatthe divide and forget groups produced virtually thesame number of items from the list considered as awhole. In short, given a fixed amount of study time,a constant amount can be recalled (cf. Reitman, Malin,Bjork, & Higman, 1973; Epstein, 1969).

8

Experiment 2

Method. The main difference between Experiment 1and tHgbresent one, was that rather than simply forgetor non-forget-cue lists, there were five differentconditions. Control lists Cl and. C2 contained only oneset of words. Under the Cl condition, however, theinitial visual label "Set 1" was presented thus leavingopen the possibility that a second set might be pre-sented (it never was). Under the C2 condition, theinitial label was "Set 2" thereby indicating that theset to follow would be the only one to be presented.Three other conditions involved forgetting the firstset of words (F2), probing for the first set (P1),and probing for the second set (P2). Under F2, thefirst set of items was cued after its presentationto be forgotten and thus the subject knew that set2 would be tested. Under P1 and P2, the instructionafter the first set was to retain it with the finalinstruction being to recall set 1 and set 2, respec-tively.

Of the probe conditions, P1 was included toauthenticate the instructions that either the firstor the second set would be tested. However, its dataare of no particular interest for present purposes,and will not be described. The remaining conditions,Cl, C2, F2, and P2, can be conceived of as representingthe orthogonal combination of two two-level factors.One was an uncertainty variable. In the C2 and F2conditions, subjects knew before the end of a listwhich items would be tested but they ,".41 not in theCl and P2 conditions. The other variable'was thepresence of interference items prior to the rememberset. There were no such items in the Cl and C2conditions but there were in the F2 and P2 conditions.

The other main differences from the prior experi-ment were that there was no final list of the kindpresented in Experiment 1, the time-of-testing variablewas a between-subjects factor, and there were 80 sub-jects tested.

Results and Discussion. Immediate and delayedrecall of unrelated and categorical remember wordsunder the critical conditions C2, Cl, F2, and P2are described in Table 3. Considering first recallof unrelated words, it is clear that in general priorknowledge of which items are to be tested improvedrecall, F (1, 60) = 23.34, p <1.001, while the presen-ation of forget items prioF to the set of remember

9

Table 3

Recall of Unrelated and Categorical Remember Words

Test ConditionUnrelated

TordsCategorical

Vords

rean SD Mean SD

Immediate C2 6.7 2.5 15.6 3.5

Cl 5.9 2.3 14.3 4.0

F2 6.7 2.2 13.2 3.9

P2 5.2 2.1 12.2 3.5

Delayed C2 5.2 2.9 11.5 3.6

Cl 4.4 2.1 12.4 4.7

F2 4.2 2.3 10.8 3.3

P2 2.7 2.3 9.1 3.2

items diminished performance, F (1, 60) = 19.91,p < .001. However, the interaction of time of testingand the presence of prior items, F (1, 60) = 6.43,

2. < .05, makes it clear that most of the proactiveinterference occurred under delayed testing. If onetakes 1-((C2-F2)/(Cl-P2)) as a measure of the amountof interference overcome by a forget instruction,then it would appear that a forget cue under imme-diate testing was 100% effective in eliminatinginterference from prior forget items, whereas in thedelayed condition it was approximately only 40%effective.

With categorical words, the debilitating influenceof prior items was evident under both immediate anddelayed testing, F (1, 60) = 43.67, g < .001. Theeffect of prior kriowledge of which items were to betested was somewhat irregular; with delayed testing,performance under the Cl condition was actually betterthan performance under the C2 condition. This reversalwas reflected in a statistically significant inter-action of the three factors of this experiment, F

10

(1, 60) = 4.46, E c .05. As to the question of howeffective a forget cue was in overcoming the inter-ference owing to forget items, use of the formuladescribed above indicates that it was totally ineffec-tive in immediate testing but roughly 80% effectiveunder delayed testing.

Based on this experiment then, we may conclude ethat with unrelated words and immediate testing, aforget cue administered immediately after the pre-sentation of a set of words can neutralize the inter-ference owing to these items. With delayed testingof unrelated words or immediate or delayed testingof categorical words, such cuing is not nearly sopotent. Indeed, with categorical words testedimmediately after their presentation, a forget cuewas totally ineffective in the reduction of forget-word interference. These observations lend consider-able support to the importance of the mental differen-tiation of sets of remember and forget words; makingtheir differentiation more difficult can result inthe absence of any evidence of positive forgetting.

Experiment 3

If set differentiation is a necessary conditionfor positive forgetting, then one should be able toobserve evidence of difficulties in actually dis-criminating between forget and remember items,particularly where our recall tests have shown inter-ference effects owing to forget items. Accordingly,the objective of Experiment 3 was to observe theaccuracy of discrimination under conditions likethose of Experiments 1 and 2.

Method. Unlike the free-recall procedure of theprevious two experiments, Experiment 3 used a "Yes-No"recognition test. For this purpose, remember and forgetwords were scrambled together and laid out in a columnon a test sheet. In the case of a list which hadcontained a forget cue (a dollar sign in this study),the subject was asked to circle the remember words butnot the forget words. For lists which did not containa ;forget cue, that is, consisted only of remember1:,.c.ds, the subject was asked to circle only thosewords which had been presented to him and not thosewhich had not.

As in Experiments 1 and 2, half the lists con-sisted of unrelated words and half of categoricalwords. Lists of unrelated words contained 15 remember

11

items and if a forget signal occurred, 15 forget items.The comparable figures for the categorical lists wereagain 25 and 25. Testing was either immediate ordelayed for a period of 2 min. The latter variablewas a between-subjects factor; the materials andforget-cue variations were within-subjects factors.Thus, each subject saw four lists, either categoricalor unrelated and with or without a forget instruction.Once again, list presentation orders observed a counter-balancing of materials and conditions. Presentationwas visual at a 2.5 second rate. There were 58 sub-jects in this experiment, 29 in each of the immediateand delayed testing conditions.

Results and Discussion. Table 4 summarizes dis-crimination performance as indexed by the Yes-No recog-nition procedure. Hits and false alarms refer respec-

Table 4

;lean Numbers of Hits (H) and False Alarms (FA)for Unrelated and Categorical Words

Forget Unrelated CategoricalTest Cue *lords Prords

H FA H FA

Immediate Absent 11.1 .3 21.4 1.1

Present 8.7 3.2 18.8 5.2

Delayed Absent 10.2 1.3 20.3 2.9

Present 7.6 3.5 17.1 6.2

tively to items circled which were and were notremember words. It is quite clear that discrimina-tion performance was uniformly superior for listswhich did not contain a forget instruction. In allinstances forget cue conditions on the averageresulted in fewer hits and more false alarms thanthe appropriate comparison conditions, F (1, 56)

33.32, R < .001 in all cases. Thus,-It wouldappear that experimental arrangements in which for-get items can interfere with the recall of rememberitems can also produce difficulties in discriminatingbetween forget and remember words. That such impaired

12

discrimination is a necessary although not a sufficientcondition for interference from forget items is indi-cated by the fact that whereas discrimination diffi-culties were observer', in the recognition of unrelatedwords under immediate testing, such conditions inExperiments 1 and 2 produced no evidence of forget-word interference.

Experiment 4

This experiment was further concerned with theinfluence of forget words on the recollection ofremember words. Up to this point we have observedthat with immediate testing of unrelated words, theadministration of a forget cue can in some way over-come interference owing to forget items. The ques-tion asked in this experiment was whether turningthe forget items into a somewhat more active storageload would produce debilitating effects on recall.By this we meant getting our subjects to expend someeffort to keep the forget material "alive" in memory.

Method. Each of 134 subjects saw 10 lists ofunrelated common words presented at a 2.5 second rate.Five of the lists contained forget cues and fivedid not with no more than two consecutive occurrencesof either kind of list. The number of items prior toa forget instruction varied. There were either 4, 6,8, 10, or 12 forget words with each value being usedonce. After a forget signal and in lists without sucha signal, there occurred 12 remember words and thesewere tested by the method of free recall.

The nature of the memory load due to the forgetitems was varied in the following manner. For halfthe subjects, the signal to forget conveyed the samemeaning that it did in the experiments which havealready been described. The cue for this purposewas the words "Now forget" against a background ofslanting lines. Since these items could be dismissedfrom any further active consideration, they weredeemed to represent a passive storage load. The com-parable instruction for the remaining half of thesubjects was the words "Test later" again displayedagainst a background of slanting lines. This meantthat the subject was to try to keep the forget itemsin mind, because while they would not be tested atthe end of the list in which they occurred, theywould be tested at the end of the experiment. Becauseof the responsibility for having to keep the forgetitems alive, they were judged to represent an,activestorage load.

13

At the conclusion of the experiment, two testswere given to all subjects. First, there was a free-recall test of all forget words. After 3 minutes ofworking at this, a 40-item, 4-alternative, forced-choice recognition test of the forget items wasadministered.

Results and Discussion. Table 5 presents meansand, in parentheses, standard deviations for perfor-mance on immediate and final retention tests. Our

Table 5

Means and Standard Deviations (in Parentheses) forImmediate and Final Retention Tests

Nature ofInstruction Cue

ImmediateTest

Final Test

CorrectRecall

Recall Recognition

Forget Absent 6.4(1.2)

Present 6.4 2.6 26.3(1.3) (2.3) (6.0)

Test later Absent 6.5(1.3)

Present 5.9 4.4 2(.5(1.6) (3.3) (6.8)

initial analysis indicated that the effect of thenumber of forget words in the list - 4, 6, 8, 10, or12 - was neither systematic nor consistent. Accordingly,the five cue lists were considered as one as werethe five no-cue lists thus yielding two mean recallscores for each subject. What is shown in the imme-diate test column of Table 5 are means and standarddeviations of these means for the four main conditionsof the experiment. The immediate recall data indicatedan interaction between the nature of the forgetinstruction and whether or not any instruction occurred,F (1, 132) = 11.71, p <:.001. That is, only in theEest-later condition, where presumably the forget items

14

served as an active memory load, was there an adverseinfluence on the immediate recall of remember items.The difference in performance between conditions inwhich a forget cue and a test-later cue was present wason the order of half an item and was statisticallyreliable, F (1, 132) = 16.42, p < .001.

The interpretation of the immediate recall datapreferred here is that the active storage load ofitems to be tested later has its effect by reducingthe amount of rehearsal devoted to remember items inlists where such a cue occurs. The possibility thata more general phenomenon was involved, that is,that items to be tested later were being recycledduring noncue lists seems nrecluded because perfor-mance in the condition where the test-later cue wasabsent was not depressed relative to performance inthe forget group.

The final recall test indicated that subjectsacted in accordance with our instruction about themeanings of the forget cues. Specifically, perfor-mance when subjects knew they were going to be testedlater on forget items was significantly superior toperformance when such a test was not expected, F(1, 132) = 13.99, p .001. By contrast, the per-formance of the two groups cn the subsequent recog-nition test was not reliably different, F (1, 132).< 1. These results are consonant with Bjork's (1972)analysis of the influence of differential rehearsalon directed forgetting. Bjork's position is thatsuch rehearsal can be expected to influence retrievaland thus performance on a recall test, which involvesretrieval. By contrast, the recognition of forgetitems depends only upon their initial registration andis independent of the amount of rehearsal devoted tothem. Thus, the absence of any substantial differencein recognition between the forget and test-latergroups is to have been expected.

Experiments 5 and 6

The main thrust of these experiments concerns thedirected forgetting mechanism of differential rehearsal.Secondarily, they have implications for two of theother mechanisms described earlier, namely, differen-tial grouping and erasure. Unlike the previous experi-ments where the interest was in the retention ofremember words, the aim of these experiments was toassess retention of forget material.

15

Method. The task was single-trial free recallof common unrelated words presented visually at a1.5-sec rate. A list container. 32 words with 6lists presented in Experiment 5 and e in experiment6. There were 98 and 57 subjects in these two experi-ments, respectively.

A basic independent variable was whether or nota list contained a forget cue. In these experimentsa dollar sign after the 12th word was the signal toforget. Thus in a forget-cue list there were 12forget words and 20 remember words and in a no-fotgetcue list, 32 remember words. In each experiment,half the lists contained a forget cue.

The other major independent variable was theoccurrence or nonoccurrence of repeated words withinlists. In Experiment 5, either no words were repeated,the 18th word was a repetition of the 6th, or the 24thword was a repetition of the 12th. In Experiment 6,both the 18th and 24th words of a list had been pre-sented previously in positions 6 and 12, respectively,or they had not Thus in Experiment 5, a list couldcontain one repeated word and in Experiment 6, tworepeated words. In any event, wnether repeated ornot repeated, we shall hereafter refer to wordsoccupying the 18th and 24th serial positions ascritical items.

The major point of these experiments that mustbe appreciated is the positioning of the first andsecond occurrences of repeated words vis-a-vis the pre-sentation of any forget cue. Note first, that the cueseparated the occurrences of a repeated item. Second,note that the initial position of an item occurringboth before and after a forget cue was either imme-diately prior to the cue or six inputs prior to it.The rationale behind this variation with respect tothe mechanism of differential rehearsal was as follows:According to this mechanism, rehearsal of forget itemsshould terminate upon the presentation of g forget in-struction. At this point, it would be expected that mostof the rehearsal of an item six positions in advanceof the cue would already have been accomplished,whereas the rehearsal accorded to an item presentedimmediately before the cue would be substantiallycurtailed. Assuming that the extent of long-termstorage is an increasing function of rehearsal, thenthe long-term strength of the former kind of itemshould not be appreciably diminished whereas it shouldbe for the latter kind of item. Subsequent repeti:-tions of these two kinds of words were assumed tolead to eventual recall performance which mirrored

16

the effects on long-term trace strengths due to theoccurrence of a forget cue. In particular, on thebasis of differential rehearsal, one would expectthat the benefits of two presentations of an itemseparated by a forget signal would be substantiallyreduced only when the initial presentation of therepeated item immediately precedes the signal.

Results and Discussion. Table 6 presents recallproportions for critiEar7g6rds in the two experiments.The pattern of results is the same in both cases

Table 6

Recall Proportions for Critical Words

Words presented twice

Serial position of the second

Words

presentation. Lists with afoa7get cue are denoted by an F.

Experimentpresented

once18 18F 24 24F

5 .31 .57 .52 .65 .49

6 .23 .50 .49 .54 .33

despite the fact that the critical words appeared to besomewhat easier to recall in Experiment 5. The firstthing to note is that recall of repeated words evenwhen separated by a forget cue was always superior torecall of nonrepeated words. Apparently, then, theoccurrence of a forget cue does not lead to the com-plete erasure of information stored in conjunctionwith forget items. Nevertheless, it does seem tomake a difference as to when a forget cue is presentedrelative to the input of items which are instructedto be forgotten. Whereas recall of repeated itemsrecurring in position 18 was independent of whetherCr not a forget cue separated the two presentations,F (1, 10) < 1 and F (1, 18) < 1, for Experiments 5

17

and 6, respectively, there were substantial differenceswith respect to repeated items recurring in position24. Specifically, recall of a repeated word originallypresented immediately prior to a forget cue was de-pressed relative to recall in the absence of a cue,F (1, 10) = 4.66, E < .10 for Experiment 5 and F 18)= 6.74, e -- .05 for Experiment 6. The tenor of theseresults is thus quite consistent with the expectationsderived from the mechanism of differential rehearsalwhich were outlined previously.

Another implication of these data concerns themechanism of differential grouping. To begin with,it is to be noted that to some extent, differentialrehearsal implies cognitive differentiation of forgetand remember items. Does such differentiation in-volve storage in different locations in memory, however,as Bjork (1972) has recently implied? If such isthe case, then it would seem to follow that in thisexperiment, re-presentation of a forget word as a mem-ber of the remember set should not make contact withthe trace of its original presentation with the effectthat repetition benefits should not accrue to a forgetword. Since the results of these experiments werequite to the contrary, the conclusion to be drawn isthat 3ifferential grouping of remember and forgetitems is not literal, that is, does not involvestorage in different memory locations.

Experiments 7 and 8

These experiments bear principally on the processesof differential grouping and differential rehearsal.The differential grouping factor was investigated byeither blocking or mixing remember and forget itemsin the presentation sequence. The thinking was thata mixed order of forget and remember items on a studytrial would make set differentiation somewhat moredifficult thereby leading to poorer retention ofremember items. The factor of differential rehearsalwas examined at three levels: conditions were arrangedsuch that all rehearsal we's devoted to remember items,more rehearsal was de,roted to remember items thanforget items, or rehearsal was approximately equallydivided between remember and forget items. Theanticipation was, of course, that the amount ofinterference from forget items should vary inverselywith the degree of differential rehearsal favoringthe remember materials. Retention in these experi-ments was assessed by a recognition procedure. Themain difference between the studies was that in

18

Experiment 8 a test involving the forget items andsome of the remember items was administered afterall lists had been presentee'.

Method. In contrast with the previous experiments,an item-B -item cuing procedure was adopted. In parti-cular, each word was followed by either an "X" or an"0". Half the time the X meant remember the item andthe 0 meant forget it, and half the time, the meaningsof these signals were reversed. The study material --common unrelated words -- and the cues were visuallypresented at a 2-sec rate, and the subjects said eachword aloud as it was presented. Each of six listsconsisted of 12 remember words and depending on condi-tions, either 12 forget words or 12 three-digit numbers.At the end of a list attention to the recognition testwas delayed for 30 sec by the administration of a setof number series problems (Experiment 7) or a task ofrecognizing nonword CVC trigrams as new or old (Experi-ment 8).

In each experiment, there were six independentgroups of 12 subjects. These groups reflected thefactorial combination of two variables. The firstconcerned the integration of forget and remember itemsin the presentation sequence. For half the subjects,the forget and remember words were blocked in thestudy sequence; that is, half the time the rememberitems occurred first and then the forget items andhalf the time the forget items preceded the rememberitems. For the other half of the subjects, the remem-ber and forget words were integrated in the presen-tation sequence. That is, they were presented randomlysubject to the restriction that no more than two ofeither type occur in succession. Hereafter, we shallrefer to the former condition as the block conditionand the latter, as the mixed condition.

The other main independent variable concerneddifferential rehearsal and its levels may be mostconveniently described according to how much rehearsalwas devoted to the forget items, namely, none, aminimum amount, or a maximum amount. In the none condi-tion, the forget items were simply not presented. Theywere replaced by three-diqit numbers which the subjectknew in advance he did not have to recall. Each ofthese nuirbers was followed by a forget cue. Thus,in the none condition, all the words were remerberitems and each was succeeded by a remember signal.In the minimum condition, both forget and rememberwords were presented and cued appropriately. In this

case, the subject was told at the beginning of eachlist what an X and what an 0 meant. Since the sub-jects knew immediately after a word was presentedwhether or not it could bs forgotten, it was feltthat only a minimum amount of rehearsal would bedevoted to each forget item, namely, that involvedin registering the wore. The maximum conditiondiffered from the minimum condition in that thesubjects were not preinstructed as to the meaningof the Xs and Os. Rather, an instruction immediatelyafter the terminal input indicated which were theremember items. The thinking here was that since thesubject did not know until the end of the list whichwere the remember items he would have to devotecllearsal activity to both sets of words.

In Experiment 7, each recognition test consistedof 12 targets and 12 nontargets. In the conditionswhere forget words were presented, 6 of the 12 non-targets were forget words and 6 were new words in thesense that they had not been seen previously in thecontext of the experiment. For groups which saw noforget words, the 12 nontarget items were simply newitems. In Experiment 8, eazh recognition test con-sisted of 10 target items and 12 nontarget items andthe nontarget items were divided in the same way asin Experiment 7. The subject made a confidence judg-ment to each test item ranging from 6 (positive theword was a member of the remember set) to 1 (positivethe word was not a member of the remember set). InExperiment 8 there was also a final recognition testadministered approximately 2 minutes after the recog-nition test on the last list. This final test con-sisted of the 12 remember or target words not pre-viously tested, the 36 forget words not previouslytested, and 12 entirely new words. Confidence judg-ments as previously described were required for thistest.

Results and Discussion. The proportions of hitsand false alarms on immediate recognition tests inExperiments 7 and 2 are presented in Table 7. Themore potent of the independent variables was thedegree of rehearsal accorded to forget words (oralternatively, the degree of differential rehearsalin favor of the remember words). In Experiment 7,the rehearsal variable was statistically reliablein the case of hits, false alarms on forget words,and false alarms on new words, F (2, 66) 9.46,a < .001. In Experiment 8, thig was true for hits

20

Table 7

Proportions of Hits and False Alarms on Immediate Recognition Tests

=Rehearsal

Experiment

Blocked

Mixed

False Alarms

False Alarms

Hits

Hits

Forget Words

New Words

Forget Words

New Words

None

7.80

.03

.03

.81

.03

.03

8.89

.03

.04

.87

.05

.06

Minimum

7.74

.25

.06

.63

.34

.09

8.72

.20

.06

.63

.30

.08

Maximum

7.61

.31

.14

.63

.31

.15

8.64

.24

.13

.61

.27

.09

and false alarms on forget words, F (2, 66) > 34.20,E < .001, but not in the case of false alarms onnew words. In general, the less the rehearsal accordedto forget items, the better the performance. Theonly statistically significant effect of block versusmixed presentation was in Experiment 8 where falsealarms on forget words in the mixed condition weregreater than under blocked presentation, F (1, 66)= 4.55, 2. < .05.

Looking at the total picture presented by bothexperiments and all dependent variables, it wouldappear that recognition performance in the six majorconditions can be classified into three categories.The no-rehearsal blocked and mixed conditions areequivalent and best; the minimum rehearsal blockedcondition is intermediate; and the minimum rehearsalmixed and maximum rehearsal blocked and mixed condi-tions are worst but not much different among them-selves. What appears to have happened then is thatperformance deteriorated when forget items were simplypresented in block fashion and it deteriorated evenmore when they were either mixed in the presentationsequence with remember items or were accorded a con-siderable degree of rehearsal regardless of whetherthey were blocked or mixed in the presentation seauence.

Final recognition performance in Experiment 8 ispresented in Table 8. Once again, the effective variableis the degree of differential rehearsal favoring the

Table 8

Proportions of Hits and False Alarmson Final Recognition Test

ItemOrder

Rehearsal Hits False AlarmsForget Words New Words

Blocked None .67 .12 .11Minimum .51 .31 .11Maximum .39 .44 .15

fixed None .65 .16 .14Einimum .44 .27 .08Maximum .43 .46 .17

22

remember items. Both for hits and for false alarmson forget words although not on new words, perfor-mance declined the more the forget items were rehearsed,F (2, 66) 9.13, p (r.001. The differences betweenblocked and mixed presentation orders were not signi-ficant. Performance continued to be most adverselyaffected when forget items were presented withoutregard to whether they were mixed or blocked in thepresentation sequence with some indication that thedeterioration was perhaps not so great when they wereaccorded only minimum rehearsal.

These data point to the significant role thatdifferential rehearsal can play in the process of in-tentional forgetting. Our attempt to hinder differen-tial grouping by mixing remember and forget itemsin the presentation sequence was on balance notterribly successful. The possibility exists thatthis may have been due to the use of a recognitiontest and accordingly Experiment 9 was carried out invirtually the same way as these experiments exceptthat a free recall procedure was used.

Experiment 9

Method. The prccedura, materials, and designof Experiment 9 were the same as in Experiments 7 and8 save for the fact that a free-recall test was com-pleted at the end of each list. There was also afinal recall test which asked subjects to write downas many of the words as they could that they hadseen earlier regardless of whether they wereremember or forget words. Further, the subjectswere asked to circle those words which they thoughthad been earlier cued to be remembered.

Results and Discussion. Means and standarddeviations for performance on the immediate recalltest are shown in Table 9. Unlike the two previousexperiments, both independent variables proved tobe statistically significant sources of variation.Thus, performance C?.clined the more that rehearsalwas devoted to forget words, F (2, 66) = 49.02,p < .001, and was reduced when remember and forgetitems were mixed as opposed to blocked in the pre-sentation sequence, F (1, 66) = 5.20, E < .05. Des-pite the absence of a significant interaction ofthese variables, it is apparent that the brunt ofthe blocked-mixed effect occurred only when forgetitems were presented and also that the sheer presen-tation of forget items was not overly interfering in

23

Table 9

Performance on Iwmediate Recall Tests

Blocked Mixed

RehearsalMean SD Mean SD

None 6.9 1.0 7.1 1.5Minimum 6.6 1.6 5.7 1.4Maximum 4.2 1.1 2.7 1.0

the absence of their maximum rehearsal or their beingintermingled with remember items in the presentationsequence.

The mean numbers of remember and forget wordsrecalled on the final test are set forth in Table 10.The differential rehearsal variable continued to be

Table 10

Mean Word Recall and Proportions of WordsIdentified as Remember IteDs (in Parsntheses)

on Final Recall Test

Rehearsal

Blocked Mixed

RememberWords

ForgetWords

RememberWords

ForgetWords

None 24.3 25.8 rMinimum 21.8 3.3 17.3 4.0

(.95) (.17) (.92) (.52)

Maximum 13.8 8.6 11.3 6.8(.76) (.30) (.62) (.28)

24

effective for both dependent variables, F (2, 66)= 17.54, E .001 for remember words and F (1, 66)= 56.71, E < .001 for forget words. By the time ofthe final recall test, however, the blocked-mixedvariation was not statistically significant. Afinal note concerns the accuracy with which subjectsidentified the words they recz-.11ed as remember orforget items. The proportions of items circled asbeing remember words are shown in parentheses. Withthe excepticn of the fact that forget words are mis-identified more under minimum than under maximumrehearsal, these statistics by and large were res-ponsive in the same way as was recall to variationsin rehearsal and in the presentation of rememberand forget words.

25

CONCLUSIONS OF EXPERIMENTS ON INTENTIONAL FORGETTING

What do the studies just reported have to sayabout each of the five intentional forgetting mech-anisms described earlier? Concerning erasure, itwould appear that if there is such a mechanism, itsoperation is far from perfect. In all experimentswhich we have conducted wherein the retention ofmaterial which one has been instructed to forget hasbeen assessed, we have been able to find evidencethat traces of forget material are still availablein storage. Nor do our studies, in our opinion,give much support to the recycling hypothesis. InExperiment 1, a group of subjects recalled a set ofwords in the face of having to recall a second setwhen it had finished with the first set. Neverthe-less, this group still achieved a higher level ofrecall than a group simply tested on the first setof words after having been told earlier to forgetthem. Additionally, Experiment 4, wherein some sub-jects were cued to be tested later on certain items,indicated that the adverse effects of keeping suchitems in mind were restricted to the lists in whichthey were presented and, as one might have expectedon a recycling notion, did not extend to other listswhich did not contain forget items. Of course, thepossibility that subjects in this experiment recycleditems to be tested later only during the retentiontest following the list in which they were presentedcannot be ruled out.

As for the three non-mutually exclusive mechanisms-- differential rehearsal, differential grouping, andselective search -- the experiments reported bearmainly on the first two of these mechanisms. Never-theless, because selective search is undoubtedlyfacilitated by differential rehearsal and is pro-bably dependent upon differential grouping of rememberand forget items, the data of the present studies dosuggest some grounds upon which the selective searchmechanism may operate. Insofar as set differentia-tion is concerned, Experiments 1, 2, and 9 indicatedthat such differentiation is eroded or even eradi-cated by semantic similarity between sets, by alack of temporal differentiation of the sets, and bya delay prior to recall of remember items. Further-more, Experiment 3 suggested that discriminativedifficulties between sets are a necessary althoughnot a sufficient condition for interference fromforget items effected by manipulations in such variables.

26

Operation of the mechanism of differential rehear-sal in these studies is most strongly implicated bythe results of Experiments 5 through 9. By means ofan item repetition procecluxe, Experiments 5 and 6revealed that the less rehearsal a forget item hasbeen accorded the poorer will be its retention. Experi-ments 7, 8, and 9 demonstrated that the less rehearsalaccorded to forget items, the less will they decimatethe retrieval, either recognition or recall, of remem-ber words.

27

EXPERIMENTS ON NONINTENTIONAL FORGETTING

Several lines of research have been conducted underthe aegis of this grant which do not bear directly onthe problem of intentional forgetting. Nevertheless,like the research on directed forgetting, they do bear onthe general problem of human forgetting. The purpose,general method, and summary outcome of each of theseprojects will be taken up in turn.

Explorations of the Tip-of-the-Tongue Phenomenon

The concern of this experiment was with thatstate of mind in which recall of information from long-term memory while unsuccessful is agonizingly imminent.Such experiences have occurred to most of us andusually Concern retrieval of such things as words,people's names, and miscellaneous facts of informa-tion. This experiment attempted two things; first, tocapture the course of the resolution of such tip-of-the-tongue (TOT) states over time and second, to refinethe relationship between ultimate recognition andprior feelings of knowing about the contents of memory.

Method. Forty male and female undergraduate stu-dents were tested on 24 general information questions.To each one they gave the following: a feeling-of-knowing judgment of the answer ranging from 1 (definitelyknow) to 4 (definitely don't know); an indication ofwhether or not the answer was on the tip-of-the-tongue;an answer, even if it was a guess; and a confidencejudgment of the answer ranging from 1 (certain correct)to 4 (certain incorrect). Thence ensued an interval ofapproximately 48 hours during which the subjects areasked not to think of any of the questions. However,if the answers to any missed questions came sponaneously. to mind, he was to record them on a preparedresponse sheet he had been given. The second sessionconsisted of another presentation of the 24 questionswith the nature of responding as just described. Athird presentation of the questions in the form of afour-alternative forced-choice recognition test followedimmediately with a confidence judgment -- 1 (certaincorrect) to 3 (a guess) -- required for each response.

Results and Discussion. Of the session 1 TOTsmee+ing the numerous definitional criteria, 36% wereresulved by the end of the first questioning, 61% bythe end of the 48hour interval, 69% by the end of thesecond questioning, and 100% by the end of the recogni-tion test. Cumulative percentages for a comparable

28

group of non-TOT questions Imre 2%, 18%, 43%, and 100%,respectively. Such observations give testimony to thestrong liability to recall associated with TOT states.

Ultimate recognition gas nositively related to thestrength of feeling-of-knowing jmdgments tendered duringthe second testing session. 1owever, the range of thefunction could be increased over what is generallyobserved by treating questions evoking TOT states asbelonging to a category stronger than the strongestfeeling of knowing. Thus, the percentage of correctrecognitions for TOT states was 82% and this declinedto 46% for the weakest feeling of knowing. By contrast,Hart (1967) found that subsequent correct recognitionsfor feeling-of-knowing states were 56% and for non-feeling-of-knowing states, 42%. Clearly, the relationshipbetween subsequent recognition and feelings of knowingcan be refined by considering as a separate categorythose feeling-of-knowing states where recall is onthe tip of the tongue.

Proactive Inhibition and Practice Effects in Single-Trial Free Recall

This research represents the doctoral dissertationof James Goodwin. He had observed that when subjectsare repeatedly presented and tested by the method offree recall on lists of common unrelated words, somereliable changes occur over the course of testing. Oneis a change in the function relating percent correctand input position. On list 1, performance on firstpresented items is superior to recall of last presenteditems. In other words, there is more primacy thanrecency. As the experimental session wears on, thelevel of recall of first presented items decreases andthat for last presented items increases. That is, thereis a shift from primacy to recency. Another observa-tion, and one which is correlated with this shift fromprimacy to recency, is a change in output statisticswith additional lists. Early in an experimentalsession, recall tends to begin with first presenteditems. On later lists, however, it is the terminalitems which are recalled first.

Goodwin's particular interest was in the declinein the primacy effect (i.e., level of recall of firstpresented items) with practice. Five studies wereconducted and they succeeded in showing rather convin-cingly that this effect is only minimally if at alldue to the correlated change in output strategy whichoccurs with practice. The general approach which Goodwin

29

adopted was to show that output statistics could bevaried considerably without any corresronding varia-tion in the decline in primacy with practice. Ratherthan describe each of the five experiments in detail,a general flavor of the investigation will be conveyedby considering only one of the experiments, one whichemployed, in fact, a variation of a directed forgettingprocedure.

Method. Three lists of 15 common unrelated nounswere visually presented at a 2.5-sec rate to two groups.In addition to standard free recall instructions, thesubjects were told that they might not be tested afterevery list, and that if they were not tested, theywould spend time equivalent to what they otherwise devoteto recalling words to working on number-series problems.At the end of each list, subjects were cued either torecall the words or to work on the number-seriesproblems. In one of the two groups, the recall group,subjects were tested on all three lists. In the other,the number series group, only the last list was tested.

Results and Discussion. If one looks at serialposition curves for list 1 end list 3 of the recallgroup, the usual drop in primacy with practice is clearlyevident. !lore importantly, the primacy portion of theonly list on which the number series group was testedwas virtually identical to the third list on which therecall group was tested. Apparently then, a decline inprimacy can occur even though one has had no priorexperience in recalling lists of words. In addition,the output statistics indicated that the number seriesgroup showed no disposition toward recalling eitherprimacy or recency items first. This contrasted withthe recall group wherein a definite shift in outputstrategy occurred, that is, from recalling first pre-sented items first on list 1 to retrieving last pre-sented items first on list 3.

Since an output strategy interpretation of thedecline in primacy with practice did not seem very com-pelling, Goodwin concluded by suggesting that thedecline is due to a build-up of proactive inhibitionas more and more lists are studied and tested. Whilesuch an interpretation possesses face validity andalso makes contact with explanations advanced for asimilar phenomenon in another context (cf. Wickens,1970), it suffers from being rather general and cer-tainly requires more severe testing than Goodwin's experi-ments subjected it to.

30

Temporal Dating and Single-Trial Free Recall

It is clearly the case that memory has a temporaldimension; more often than not we remember which eventsoccurred before, simultaneous with, or after whichother events. T.lhether such information is functionalin the retrieval of information from memory, however,is another question. And it was to this questionthat James Goodwin addressed himself in this researchwhich served as his Master's thesis and which isreported in the American Journal of Psychology, 1972,85, 597-604.

As in the research just described, the empiricalfocus was once more on the decline in primacy whichoccurs with practice in the task of single-trial freerecall. Is it possible that this decline is paralleledby a decline in memory for temporal information acquiredduring the presentation of a lict which is to be testedby free recall? If so, then one could continue toentertain the notion that temporal information is impor-tant to the retrieval cf at least initial list membersin free recall. But if not, then one possibility couldbe that temporal tags are of little functional valuein free recall. To assess. the retention of temporalinformation, a reconstruction task in which one reassemblesa scrambled order of list members into its proper orderwas used.

Method. Each list contained 15 unrelated concretenouns presented visually at the rate of 1 word every1.5 sec. There were two independent variables ofinterest whcse factorial combination yielded four inde-pendent groups with 3G subjects being assigned to eachone. Half the subjects received 1 list and half thesubjects received 6 lists. The critical lists in thesetwo conditions are the 1st and 6th lists, respectively.These critical lists were tested half the time by freerecall and half the time by reconstruction. However,they were studied under the impression that they wereto be tested by free recall.

Results and Discussion. The critical observationsin this experiment were e-TEE primacy effacts on thecritical lists under the two methods of testing. Forfree recall, there was the conventional decline inprimacy with practice, that is, from list 1 to list 6;but with reconstruction, no such wall-defined declinein primacy occurred. This result is certainly suscep-tible to a number of interpretations. But one ofthem, and the one preferred by Goodwin, was that the

31

results are not consistent with the point of view whichstresses the efficacy of temporal tags as cues for theretrieval of early list members during free recall.

Retroactive Facilitation 3n Short-Term retention ofManimally Learned Paired Assoaates

Three experiments conducted in collaboration witha faculty colleague, George E. Weaver, and reported inthe Journal of Experimental Psychology, 1973, 100, 9-17,were-a-jgl-Ted to extend to the realm of short-termmemory some of the basic transfer paradigms used in thestudy of long-term memory. The initial concern was totest the generality of the interference theory offorgetting and to determine whether or not short- andlong-term memory are mediated by a single type ofstorage mechanism. As the experiments progressed,however, it became apparent that the results wouldhave little to say about such issues. Nevertheless,they did generate observations which were of interestin their own right.

Method. The procedure, which was common to allthree experiments, involved the presentation of shortlists of paired associates. Following each one, reten-tion was tested for two or three of the pairs by pre-senting their left-hand members and asking the subjects togive what they thought were the corresponding right-handmembers. The composition of these tested or criticalpairs was one major independent variable. We mayillustrate by reference to the first of the three experi-ments. The pairs were entirely dissimilar or they con-tained identical left-hand members. A second majorindependent variable concerned the number of occurrences(either 1 or 2) of the second .presented critical pair.The other two experiments were mainly variations onthe theme of manipulating the nature of the criticalpairs. Common to all three experiments was a thirdindependent variable, namely, where the critical pairsoccurred in the list. They were presented either atthe beginning, in the middle, or at the end of a list.And, of course, the subjects did not know in advancejust where they would he placed.

The pairs in each list consisted of either trigramsor words as left-hand members and words as right-handmembers. Each subject was tested 2-4 times under allconditions of an experiment. Finally, pair presenta-tion was visual and at the rate of one pair every 2 or3 sec.

32

Results and Discussion. The most consistent andoutstanding Fa of these experiments concerned recallof the first presented critical item. hen first andsecond pairs shared identical stimuli, performance onthe first pair was facilitated relative to the casewhere the two pairs did not share identical stimuli.Such a finding cctrasts sharply with the data fromstudies of long-term retention wherein recall of afirst learned set of paired-associates is depressedif a list learned second shares the same left-handmembers. The explanation tnat we prefer for ourresults is simply that pairs with identical stimuliare rehearsed together more than arc items from aparadigm comprising different stimuli.

ape of Rehearsal and Delay in Single -Trial ee Recall

Rundus and Atkinson (1970) have described an overtrehearsal procedure which they have used in the contextof a free recall task. As used in their experiment,each word in a list was presented visually for a 5-secinterval. During this time, the subject rehearbesaloud any items from the list which enter his mind.There are no restrictions on the order of rehearsalsand the only requirement is that subject must maintaina constant flow of rehearsal activity for the entire5-sec interval. Such a procedure has proved useful ingenerating theoretical notions about human memory ingeneral and the task of free recall in particular.Nevertheless, the possibility exists that this proce-dure yields a somewhat distorted picture of what sub-jects are ordinarily doing during list presentation,and it was this possibility which was investigated inthe current experiment.

The procedure was siiple encash. Two groups ofsubjects'were compared, one of which rehearsed overtlyand the other which rehearsed covertly. A within-subjects variable was then manipulated in each of thetwo groups; specifically, the free recall test followingeach list was administered either immediately or aftera delay of 20 sec. The question was whether or notperformance would be comparable in the two groups whichrehearsed differently.

Method. Six lists of 15 unrelated nouns were pre-sented at the rate of one word every 5 sec and eachwas tested by the method of free recall. Half thelists were tested immediately and half were testedafter a 20-second delay during which subjects readdigit arrays aloud as fast as they could. No more

33

than two consecutive lists were of the same type. Onegroup of 36 subjects was given standard free recallinstructions and hence rehearsed silently or covertly.Another group of 36 subjects rehearsed overtly followingthe procedure described by Rundus and Atkinson (1970).That is, subjects in this group were free to rehearselist items in any order provided that they filled eachitem presentation interval with rehearsal activity.

Results and Discussion. Inspection of the func-tions relating percent correct to input position re-vealed that the delay in testing resulted in similarreductions in the level of recall of those items pre-sented last for both overtly and covertly rehearsingsubjects. Nevertheless, there were differences betweenthe groups. In particular, overt rehearsal led tobetter recall of initial items whereas differences infavor of covert rehearsal were evident over nearlyall of the subsequent serial positions. Taking allinput positions together, the group of subjects rehearsingcovertly recalled an average of 9 words per list whereasovertly rehearsing subjects recalled an average of only7.9 words per list. These differences lead one to wonderabout the extent to which one may safely draw conclu-sions about the operation of the human memory systemunder ordinary circumstances from data collected withthe overt rehearsal procedure of Rundus and Atkinson(1970). At the very least, what is indicated is thatsubjects left to their own devices do not adopt pro-cesses for remembering information which correspond toovert rehearsal activity.

Retention of Isolated Words

Why is a word which is made perceptually distinc-tive in a list remembered better than a comparable wordwhich is not perceptually distinctive? One possibilityis that when a particular item in a list is associatedwith a distinctive stimulus attribute, the attributeserves as a cue which directly facilitates subsequentretrieval of the item. Another possibility is a selec-tive attention-rehearsal hypothesis: an isolated itemis more likely to be remembered than a comparable non-isolated one because it tends to be held in mind longer.These two possibilities were examined in this research.

Method. Three experiments were conducted whichdiffered only slightly from ona another. In all ofthem, lists of 20 common words were visually presentedat either a 1- or 2-second rate and tested by eitherfree recall or recognition. Each list contained four

34

critical words. These words were either (1) semanti-cally related (2) semantically unrelaed but enclosedin rectangles (isolated items), (3) both semanticallyrelated and enclosed in rectangles, or (4) neithersemantically related nor eAclosed in rectangles (con-trol items). The critical words either occurred con-secutively, that is, were massed during presentation,or were distributed throughout the list.

Results and Discussion. The results of all thi..eeexperiments appeared to support the conclusion thatthe isolation effect is the result of special attentionor rehearsal given to perceptually distinct stimuli,and that it is not the result of the distinctiveattribute per se serving as a retrieval cue. Theresults which supported this conclusion were: (1) thatenhanced recall of isolated items was at the expenseof other items in the list; (2) that the superiorityin the recall of isolated over control items wasgreatest for the first isolated words in a list;(3) that the isolation effect persisted in recognitionwhere problems of accessibility are presumably eliminatedwhereas the effect of the semantic cue (i.e., semanticrelatedness) was cut sharply; and (4) that the isolationeffect was not reduced after distributed presentationwhereas the effect of semantic relatedness was.

35

CONCLUSIONS OF EXPERIMENTS ON NONINTENTIONAL FORGETTING

General conclusions based on all the data reportedin this section are virtually impossible in view ofthe widely varying nature of the researches. However,we see three points of contact among them and it isthese which will be emphasized here.

The first of these concerns rehearsal as a controlprocess. Two of our lines of research, namely, thatinvolving the short-term retention of paired associatesand the retention os isolated items have appealed torehearsal in explaining observed phenomena. Further-more, the experiment in which controlled and uncontrolledrehearsal were compared confidently concluded by sayingthat covert rehearsal activity is unlike the simplerepetition of items. But hidden behind these appealsto rehearsal and assertions about its nature is a basicignorance of just what rehearsal does. Thus, to attri-bute some phenomenon to rehearsal only raises the fur-ther question of how rehearsal produces that phenomenon.Moreover, to say what rehearsal is not is quite adifferent matter from saying what it is.

Two of the researches which were described, namely,the Master's thesis and dissertation of James Goodwin,dealt with the decline in primacy which occurs withpractice in single-trial free recall. Mile this is arather minor empirical phenomenon, it neverthelessappears to implicate and exclude explanatory factorsof rather considerable importance. In particular, itappears that temporal information ordinarily stored inthe course of rememberina is not an important basisfor the retrieval of information from memory. Andsecondly, the importance of prevously studied materialas a powerful interfering force in the retention ofinformation is once again implicated.

The final point we should like to make concernsthe importance of retrieval. The tip-of-the-tonguephenomenon is an excellent example of a retrievalfailure; what is sought after is clearly in memorybut there is a failure to access fully the requisiteinformation. One wonders whether most of our memorylosses are due not so much to the unavailability ofinformation but to its inaccessibility. If this is true,then the real problem becomes one of finding the appro-priate retrieval cues. And it may be that this is theessential nature of rehearsal, that is, identifyingappropriate retrieval information by which to locateand extract at some later time information from memory.

36

Bil'liography

Bjork, R. A. Positive forsetti.ng: The noninterferenceof items intentionally for3otten. Journal of VerbalLearning and Verbal Behavior, 1970,4, 253.4-68.

Bjork, R. A. Theoretical implications of directedforgetting. In A. W. Nelton & E. Martin (Eds.),Coding processes in human memory. Washington:Winston, 19 Pp. 217=YI5.

Cohen, B. H., Bousfield, W. A., & Whitmarsh, G. A. Cul-tural norms for verbal items in 43 categories.Technical Report No. 22, 1957, University ofConnecticut, Contract Nonr 631(00), Office ofNaval Research.

Epstein, W. Poststimulus output specification anddifferential retrieval from short-term memory.Journal of Experimental Psychology, 1969, 82, 168-174.

Epstein, W. Mechanisms of directed forgetting. In G. H.Bower (Ed.), The psychology of learning and motiva-tion: Advances in research iiiii-theory. Vo17-77--NewYorE: AcadeMic-ffiess, 1g77. Pp?. 147-191.

Hart, J. T. Memory and the memory-monitoring process.Journal of Verbal Learning and Verbal Behavior,1967, 6, 685-691.

Reitman, W., Malin, J. T., 8jork, R. A., & Higman, B.Strategy control and directed forgetting. Journal,of Verbal LearRin aid Verbal Behavior, 1973, 12,140-149.

Rundus, D., & Atkinson, R. C. Rehearsal processesin free recall: A procedure for direct observa-tion. Journal of Verbal Learning and Verbal Behavior,1970, 9, 99-105.

Wickens, D. Encoding attributes of words: An empiricalapproach to meaning. Psychological Review, 1970,77, 1-15.

37