SHEILA JONESUniversity College London, London, England

Imagery in a short-term memory task

Instructions for two short-term memory spatial tasks involvingpaired-associate (PA) learning were presented under conditions wh ich eithersuppressed or permitted the use of imagery mediation. Differential performanceof the tasks for the two presentation conditions was found to be related to theimagery concreteness of the response members of the PAs when high-imagerystimulus members were used as nonverbal cues in recall. The prediction thatimagery would be used spontaneously and effectively when conditions favoredits use was confirmed. The results also indicate that the mode of presentation ofan instruction sentence (written vs spoken) may be an important variable insubsequent performance of a short-terrn memory task.

Th is experiment examines thehypothesis that imagery will bo usedspontaneously in a short-term memorytask if conditions are maximized tofavor its use.

Imagery, used as a mnemonic ordefined in terms of stimulus attributes,has been shown to be the mosteffective mediator in a paired-associate(PA) learning task (Bower, 1969 ;Paivio, 1969,1971). "Imagery­concre teness" or image-evokingcapacity has been isolated by Paivio asthe most potent stimulus attribute forfacilitating PA learning. Imagery (I)and concreteness (C) are highlycorrelated and defined in terms ofrating scale measures (Paivio, Yuille, &Madigan, 1968). For noun pairsvarying in imagery concreteness, PAlearning correlated highly with the 1value of the stimulus member of thepair, and to a lesser extent with the Ivalue of the response member (Paivio,1968). Paivio interprets this finding asevidence that high-1 stimulus nounsevoke i m ages that function asmediators in much the same way asthey do in mnemonic techniques. Suchtechniques have repeatedly beenshown to improve PA learning ofconcrete noun pairs compared withother learning strategies (Bower &Winzenz, 1970). However, in the caseof abstract nouns (low I), Paivio &Foth (1970) have shown that verbalmnemonics are more effective thanimagery. Moreover, Ss disregardinstructional sets and readily abandonmediation strategies inappropriate forthe particular type of nouns (Paivio &Yuille, 1969).

Bower (1969) presents evidencethat adult Ss have to be instructed touse interactive imagery rather thanimagery per se to obtain significantimprovement over verbal mnemonicsin PA learning. Although Ss are knownto adopt coding devices of varyingcomplexity in PA learning, it appearsthat interactive imagery is notspontaneously used as a mnemonic.This may apply only to long-termmemory tasks such as those used in

76

the experimen ts aiready cited.However, Ss rarely have occasion tolearn long lists of word pairs outsidethe experimental situation and hencetheir learning strategies for such tasksmay not reflect those used inshort-term memory tasks.

The novel feature of the presentexperiment is that only four pairedassociates are used and presented onone occasion only for immediate freerecall. A "realistic" task situation isachieved by presenting the four PAs inthe form of a verbal instruction inwhich Ss are told to perform anonverbal spatial task, viz, to draw orwrite the referent of the first noununder the referent of the second nounof the pair. The task involves cuedrecall in the backward (R-S) direction.It is predicted that (1.) Sc; willspontaneously use imagery mediationin such a task if high-I nouns arepresented nonverbally as stimulus cuesin recall, and (2) the initiation andeffectiveness of imagery will be afunction of the imagery concretenessof the response term, i.e., the firstpresented noun of the pair. To obtaindirect evidence for the use of imagery,the technique developed by Brooks(1967) for suppressing visualization ofverbal messages by varying the modeof presentation (written or spoken)was adopted. This method of selectiveinterference has already been used byAtwood (1971) to demonstrate thatmnemonic imagery actually involvesvisual processes,

METHODA short-term memory task was

devised requiring actual spatial pairingsof the referents of either concretegeometrie shape nouns (C) or abstractnumber nouns (A) with concrete colornouns (C). (Such pairings arefrequently encountered outside theexperimental situation, e.g., in roadsigns, electrical connections.) Two PAlists of four shape-color (CC) and fournumber-color (AC) noun pairs wereconstructed, in which the geometrieshape and number nouns werematched as closely as possible for

frequency, c o n sistent with therequirement that they should also bematched for word and syllable length.The shape and number nouns wererated for concreteness (C) and imagery(I) on a 7-point scale by 32in troductory psychology students,following the procedure andinstructions used by Paivio et al(1968). The nouns were interspersedin a buffer list of 20 other nouns ofvarying abstractness-concreteness, Aperiod of 8 weeks separated the tworatings.

Eight color nouns-rred, green,brown, yellow, white, blue, purple,and black-were used as stimulus cuesand presented in nonverbal form in therecall tests, Two random pairings ofthe CC (shape-color) and AC(number-color ) nouns were used inorder to control for any possibleidiosyncratic effects of specific pairs.The order of presentation of the pairswas systematically counterbalancedwithin each PA list. The two types ofPA lists were presented in the form ofinstructions to carry out specificspatial tasks, e . g., InstructionType CC: Put square under brown,oval under blue, cross under green,circle under red; Instruction Type AC:Put 3 under brown, 7 under bIue, 12under green, 11 under red. Two modesof presentation of the instructionswere used to facilitate or suppressvisualization, following the procedureused by Brooks (1967): listening(L)-S listened to the instruction forthe length of time it took for E tospeak i t (approximately 4 sec);listening and reading (LR)-8 listenedto the sentence spoken by E andconcurrently read a typewrittenversion of it. A constant presentationrate of 1 pair/sec was used, and toprevent 8 from reading the instructionmore than once in Condition LR, thefour noun pairs were typed onseparate cards and presented seriallyby E. Immediately after presentationof each instruction, all of the possibleeight colors were displayed in front ofS in the form of a horizontal array ofcolor squares. 8 then drew the shapesor wrote the digits under theappropriate colors, under free recallconditions and with no time limitimposed. Immediately after S finishedthe first task, the color chart wasremoved and the second instructionpresented, following the sameprocedure.

In addition to the two test trials,each S was given a preliminary test ofnaming all the colors in the array (as acheck for color blindness) and twopractice trials of three PA pairs each,using noun pairs similar to those in theexperiment. This familiarized S withthe procedure and at the same timeindicated that the second (color)

Psychen. ScL, 1972, Vol. 29 (2)

Table 1Mean Correct Responses and StandardDeviations Per Instruction (Out of Four)as a Function of Response Concreteness

and Mode of Presentation

ListeningMean 3.12 2.75SD 1.39 1.11

Listening and ReadingMean 2.62 2.50SD 1.28 1.25

nouns of the pairs were to be used asstimulus cu es in recall and thatr esponses were to be made innonverbal form. To minimize practiceeffects, each S carried out two testtrials only, with one combination ofeach mode of presentation with eachinstruction type. The Ss were assignedaltemately to one of two groups(N = 24 in each group). In the firstgroup, the CC pairs were presentedonce under Condition Land the ACpairs on ce under Condition LR. Forthe second group, the mode ofpresentation was reversed. In bothgroups, the order of presentation ofthe two instructions was altematedbetween Ss. Ss were 48 first-yearstudents at University College London.

RESULTSImagery concreteness of response

nouns: The mean and sn of the Cratings for the shape nouns were 4.1and 1.3 and for the number nounswere 3.1 and 1.6, respectively. Theab stract-concreteness difference forthe two types of nouns was highlysignificant (Wilcoxon, Z 3.75,p< .0001). The mean and sn of the Iratings for shape nouns were 4.6 and1.4 and for the number nouns were3.6 and 1.9, respectively. Shape nounswere rated significantly higher on Ithan were number nouns (Wilcoxon, Z

3.35, p < .0005). These resultsestablish that the imagery-concretenessof the shape (C) response nouns wassignificantly greater than that of thenumber (A) response nouns.

Mean correct responses (with amaximum of four) for each type ofinstruction under each presentationcondition are shown in Table 1,together with standard deviations.Overall, r e gar d l ess of mode ofpresentation, there was no significantdifference in difficulty for AC and CCinstruction (Wilcoxon, Z = 1.24, P =0.22). Comparison of correctresponses for Presentation Mode L vsPr e se n tation Mode LR showed asignificant difference in the case of theCC instruction (median test, Z = 3.0,p< .05), but no significant differencefor the AC instruction (median test, Z= .15, p = .70). Comparison of correct

Response

as the results indicated.A further factor which might favor

recall of the concrete-concrete relativeto abstract-concrete pairs was the useof backward rather than forwardrecall, i.e., the use of the second (orright-hand) members of the pairs asstimulus cues for recall. According tot h e a ssocia t i ve d irecti onali tyhypothesis proposed by Paivio (1971 ),visual imagery is specialized forparallel processing in the spatial senseand, to the extent that associationsinvolve imagery, either member of thepair will be equally effective as aretrieval cue. Verbal symbolicprocessing, on the other hand, isspecialized for sequential processing,which clearly implies directionality ofassociations between the pairs. To theextent that noun pairs are encodedverbally, the associations will bestrenger from left to right andbackward recall will be more difficult.Research findings on retrievalasymmetry related to word imageryconcreteness, cited by Paivio (1971),are generally consistent with thehypothesis that the more available theimagery code, the more likely it is thatassociative recall will be symmetrical,i.e., independent of which member ofthe pairs is used as stimulus cue inrecall. This hypothesis is supported bythe finding in the experiment that,when imagery was permitted, therewas a significantly greater number ofcorrect p a i r i n g s for theconcrete-concrete pairs than for theabstract-concrete pairs.

The extent to which the variables ofimagery-concreteness and mode ofpresentation of a verbal instructionaffect the efficiency of performance ofa short-term memory task, as indicatedby the results of this experiment, mayhave more general implications.Previous research (Jones, 1966, 1968,1970) has shown that verbal factors ininstruction senten ces affectperformance of nonverbal tasks, Theextent to which the conclusions fromthe present study can be generalized toother instruction sentences remains acrucial question,

REFERENCESATWOOD, G. An experimental study of

visual Imagination and memorv. CognitivePsvchologv , 1971, 2, 290-299.

BOWER, G. H. Mental imagery anda ss o c i a t iv e Ie ar n i ng , Fifth AnnualSymposium in Cognition,

. Camegie-Mellon University, Pittsburgh,Pa., 1969. To be published in L. Gregg(Ed.), Cogniiion in learning and memory.New Volk: Wiley.

BOWER, G. H. Imagery as a relationalorganizer in associative learning, Journalof Verbal Learning & Verbal Behavior,1970, 9, 529-533.

BOWER, G. H., & WINZENZ, D.Comparison of a ssociative learningstrategies, Psvchonornic Science, 1970,20, 119·120.

BROOKS, L. R. The suppression ofv i su a l i z a t i o n b v reading, Quarlerly

responses under Condition L showedsignificantly better performance underCC compared with AC instruction (x 2

= 5.34, p < .025). The interaction ofmode of presentation withconcreteness was calculated bycomputing the error difference scoresfor each S for each type of instructionand comparing these difference scoresbetween the two groups. A significantinteraction was found (Mann-Whitney,Z = 1.70, P = .045), showing that theeffect of allowing Ss to visualize(under Condition L) was to produce asignificant improvement inperformance of the CC instructionrelative to the AC instruction.

nISCUSSIONThe results support the hypothesis

that imagery will be u se dspontaneously to mediate recall ofnoun pairs when the task situation isrealistic and conditions are appropriatefor imaginal mediation, viz, if the taskentails recall of verbal material in anonverbal spatial relationship and ifhigh-imagery stimulus members areused as cues in recall. Initiation andeffectiveness of imagery, under suchconditions, is shown to be related tothe imagery-concreteness of theresponse members of the pairs.

These results accord with the dualcoding theory proposed by Bower(1969) and Paivio (1969), whichhypothesizes that nonverbal imageryand verbal symbolic processes are thetwo major components in theencoding and retrieval of verbalmaterial and that imagery is thepreferred and most effective mediatorfor concrete material, whereas verbalmediation is better suited to abstractmaterial. In terms of the concrete(shape-color) noun pairs used in thepresen t experiment, the theorypredicts that recall would be expeditedby the encoding of the pairs intocompound Images easily retrievable bythe presentation of a nonverbal(high-I) stimulus cue. Such"chunking" of the noun pairs wouldclearly benefit recall in a short-termmemory task. Suppression of imageryfor such pairs would thus be expectedto impair performance as indicated bythe data.

In the case of abstract-concrete(number-color ) noun pairs, however,the first member of the pairs wouldtend to be encoded in a predominantlyverbal form and there would be lesslikelihood of encoding the pairs ascompound images. This corresponds tothe encoding of separate rather thanrelated items, as discussed by Bower(1970). Presentation of a nonverbalstimulus cue, in this case, would beless advantageous for the retrieval ofthe abstract members of the pairs. Ifthis analysis is correct, suppression ofimagery should make little difference,

AbstractConcreteMode of

Presentation

Psychen. ScL, 1972, Vol. 29 (2) 77

Journal of Experimental Psychology,1967,19,289-299.

JONES, S. The effect of a negative qualifierin an instruction. Journal of VerbalLearning & Verbal Behavior, 1966, 5,495-491.

JONES, S. Instructions, self Instructions andperformance. Quarterly Journal ofExperimental Psychology , 1968, 20,74-78.

JONES, S. Design of instruction, TrainingInformation Paper 1. London: HerMaiestv's Stationery Office, 1970.

Dr. Harley A. Bernbach, forrnerly ofCornell University, has accepted theposition of Associate Professor,Department of Psychological Sciences,Purdue Unicersit y,

Dr , William H. Calhoun hascompleted a one-year postdoctoralf e l l o w s h ip under NSF's ScienceFaculty Fellowship Program at IndianaUniversity, where he worked with Dr.George Heise on problems inpsychopharmacology. He has nowreturned to the Universit y ofTennessee where he is Professor ofPsychology.

Dr , Stephen F. Davis has recentlygone to Austin Peay State University,Clarksville, Tennessee. from KingCollege. He is Assistant Professor ofPsychology, teaching bothundergraduate and graduate courses,and developing the animal laboratory.

Dr. Nancy Frost, has moved frornthe University of Washington toPrinc eton Uniuersit y ; where she isAssistant Professor of Psychology.

78

PAIVIO, A. A factor analvtic study of wordattributes and verbal Ieazning, Journal ofVerbal Learning & Verbal Behavior,1968,7,41-49.

PAIVIO, A. Mental imagerv in associativelearning and memorv. PsvchologicalReview, 1969, 76, 241-263.

PAIVIO, A. Irnagerv and verbal processes.New York: Holt, Rinehart & Winston,1971.

PAIVIO, A., & FOTH, D. Irnaginal andverbal mediators and noun concretenessin paired-associate learning: The elusive

XOTES & r\EWS

Dr. Gilbert J. Harris has been , as ofSeptember 1, Assistant Professor ofPsychology at the Herbert J. LehmallCollege of the City Universitv of Neu:York.

Dr. Kenneth R. Laughery has leftthe State Universitv of New York atBuffalo to beco~e Professor andChairman in the Department ofPsychology at the Uriiversity ofHouston.

Dr. Geoffrey R. Loftus, previouslyon a postdoctoral fellowship at NewYork University, is currently AssistantPr 0 f essor of Psychology at theUnicersitv of Washington.

Dr. Lester A. Lefton, recently ofthe University of Rochesters Centerfor Visual Seiences. is now AssistantProfessor of Psychology at theUniversity of South Carotine.

Dr. Stuart J. McKelvie. who hascompleted his PhD at McGillUniversit y, is now Visiting Assistant

interaction. Journal o f Verbal Learning &Verbal Behavior, 1970,9,384-390.

PAIVIO, A., & YUILLE, J. C. Change inassociative strategies and paired-associatelearning over trials as a function of wordimagerv and type of Iearning set , Journalof Experimental Psvchologv, 1969, 78,458-463.

PAIVIO, A., YUILLE, J. C., & MADIGAN,S. Concreteness, i m a g e r y andmeaningtulness values for 925 nouns.Journal of Experimental Psychologv ,1968, 76, 1-25.

Professor of Psychology at Bishop 'sUniiersity Lennoxville, Quebec.

Dr . Curtis W. McIntyre has recentlycompleted a two-year postdoctoralfellowship at the Center for Researchin Human Learning and at theInstitute of Child Development,University of Minneseta. He is nowAssistant Professor of Psychology atthe Cniversitv of Virginia.

Dr. Lawrence C. Perlmuter , recentlyof Bowdoin College, is AssociateProfessor of Psychology, 'VirginiaPo l yt e c h n ic Institute and Stat cUniversit y.

Dr. Robert M. Schwartz, who hasrecently completed his PhD at theUniversity of British Columbia, is nowa postdoctoral fellow (NIMH) atStanford Unicersity, wor king underDr. R. C. Atkinson.

Dr. Harry Strub is taking asabbatical leave from the University ofWinnipeg to work with David R,Thomas at the Universit y of Colorado.

Psychen. Sei., 1972, Vol. 29 (2)