Proc. Natl. Acad. Sci. USAVol. 92, pp. 7580-7584, August 1995Psychology

The deferred imitation task as a nonverbal measure ofdeclarative memory

(anterograde amnesia/infantile amnesia/implicit memory/explicit memory)

LARAINE MCDONOUGH*t, JEAN M. MANDLER*, RIcHARD D. MCKEEt, AND LARRY R. SQUIREt§*Department of Cognitive Science, University of California at San Diego, La Jolla, CA 92093-0515; *Department of Psychiatry and Neurosciences, University ofCalifornia at San Diego, La Jolla, CA 92093; and §Veterans Affairs Medical Center, San Diego, CA 92161

Contributed by Larry R Squire, April 28, 1995

ABSTRACT We tested amnesic patients, patients withfrontal lobe lesions, and control subjects with the deferredimitation task, a nonverbal test used to demonstrate memoryabilities in human infants. On day 1, subjects were given setsof objects to obtain a baseline measure of their spontaneousperformance of target actions. Then different event sequenceswere modeled with the object sets. On day 2, the objects weregiven to the subjects again, first without any instructions toimitate the sequences, and then with explicit instructions toimitate the actions exactly as they had been modeled. Controlsubjects and frontal lobe patients reproduced the eventsunder both uninstructed and instructed conditions. In con-trast, performance by the amnesic patients did not signifi-cantly differ from that of a second control group who had thesame opportunities to handle the objects but were not shownthe modeled actions. These findings suggest that deferredimitation is dependent on the brain structures essential fordeclarative memory that are damaged in amnesia, and theysupport the view that infants who imitate actions after longdelays have an early capacity for long-term declarative mem-ory.

Distinctions between different types of memory have beenmade by neuroscientists (1-3) and experimental psychologistsworking with adults (4, 5) and with children and infants (6-8).TIwo divisions are commonly recognized. One refers to thecapacity to acquire specific facts and events that are accessibleto conscious recollection, usually termed explicit or declarative(2, 5), and is typically tested by asking the learner to study alist of words, a story, or an event and relate it after a delay. Incontrast, implicit or nondeclarative memory refers to thecapacity to acquire skills and strategies that are not accessibleto conscious recollection, as well as simple forms of classicalconditioning and the phenomenon of priming.

Neuropsychological evidence for such a distinction has beenobtained from studies of amnesic patients with damage to themedial temporal lobe or midline diencephalon (9). Amnesicpatients can acquire new skills but without the phenomenalawareness that they possess new knowledge. For example,performance on tasks such as mirror reading improves at anormal rate across several sessions, but the patients do notexplicitly remember having previously practiced the task (10).They know how to perform the task but do not remember thatthey ever practiced the task.An important question is how and when declarative memory

develops. Some have proposed that the limbic/diencephalicbrain structures supporting declarative memory develop laterthan nondeclarative memory (11, 12). The late development ofdeclarative memory would provide an account of infantileamnesia-i.e., the inaccessibility of memories dating from thefirst few years of human life. Recently, an alternative view has

been presented: namely, that limbic and diencephalic brainstructures are functional in early life and that infants possessan early capacity for declarative memory (13-15). This work,carried out with various subject populations, has suggested thatthe tasks used to test infant memory in fact require declarativememory. The tasks that have been used typically requirerecognition of a correct choice out of two or more alternativesafter delays not longer than a few hours (15-19).To study infant memory across much longer intervals,

researchers use deferred imitation tasks to show that infantswill imitate novel actions that had been modeled with objectsafter delays from 24 hours to as long as 1 year (7, 17-20).Mandler (6) proposed that deferred imitation tasks requirerecall (i.e., declarative memory) when the events tested areentirely novel at the time of the learning session, the subjectsare not allowed to practice the actions before they are tested,and the actions are not modeled again after the initial "learn-ing" session.There are other possible interpretations of deferred imita-

tion tasks that do not rely on declarative memory. Subjectsmight learn a sensorimotor association between an object andan action by observation alone, such that the later sight of theobject will automatically trigger the action. A variant of thisidea is that young infants represent actions as tightly linked totheir associated objects (21). If this linking takes place withinone or two exposures to an action-object pair, deferredimitation might be similar to skill learning or conditioning,which are tasks of nondeclarative memory that amnesic pa-tients perform as well as normal adults (9).Another possibility is that the deferred imitation task is

analogous to priming tasks. When given three-letter stems forwords that have been studied and instructions to complete thestems with the first word that comes to mind, amnesic patientsperform nearly as well as control subjects (22, 23). However,when explicit instructions are given to report the words on thestudied list, amnesic patients are impaired relative to controlsubjects (23). In the case of the deferred imitation test, if theaction modeled with an object is considered to be a single unitof representation (ust as a word can be considered a singleunit), then presentation of part of that unit (e.g., the object)may be analogous to the three-letter cues in word-stem com-pletion priming, and the action may be analogous to producingthe remaining letters of the target word. Thus, infants mightsucceed on deferred imitation tasks in the same manner asamnesic patients succeed when they are asked to respond withthe first thing that comes to mind.To reduce the likelihood that the deferred imitation task is

mediated simply by a stimulus-response association betweenan object and an action, one can require subjects to imitateindividual actions in the same order that they were modeled (7,17). This procedure is similar to serial recall in which, following

Abbreviations: WAIS-R, Wechsler Adult Intelligence Scale-Revised;WMS-R, Wechsler Memory Scale-Revised; E, experimenter.tTo whom reprint requests should be addressed.

7580

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

July

2, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995) 7581

a retrieval cue, subjects recall a sequence of items, such as aprose passage. The sequence of items to be recalled can bedetermined either by causal or arbitrary relationships amongthe items (see Procedure). Infants perform better on deferredimitation tasks ordered by causal relations (7, 17), just as olderchildren and adults do when verbally recalling events andstories (24-26). These findings raise the possibility that per-formance on these tasks is mediated by the same kind ofmemory. Nevertheless, it is still unknown whether or notdeclarative memory is required to perform the deferredimitation task.To determine the kind of memory that supports perfor-

mance in the deferred imitation task, we tested control sub-jects, amnesic patients, and patients with frontal lobe lesionswith the same procedure used to test long-term memory ininfants. Although lesions restricted to the frontal lobe do notproduce amnesia (27, 28), performance on memory tasks byamnesic patients can be influenced by frontal lobe dysfunction,particularly on tasks requiring memory for temporal order(29). Moreover, amnesic patients with Korsakoff syndrometypically have frontal lobe atrophy (30). To evaluate theimportance of the frontal lobes and medial temporal lobe-diencephalic structures for memory of action sequences, wetested patients with damage limited to the frontal lobes,amnesic patients with Korsakoff syndrome, and amnesic pa-tients with hippocampal formation damage. Age- and IQ-matched control subjects were divided into two groups. Onecontrol group (experienced controls) was tested in the samemanner as the patients. The other control group (inexperi-enced controls) was given the tasks in the same manner exceptthis group was never shown the targeted actions. This controlcondition was used to measure the extent to which perfor-mance might change across test sessions simply because thesubjects have repeated opportunities to examine and manip-ulate the objects (18, 19).

METHODSubjects. Amnesic patients. Seven patients (6 men and 1

woman), average age 66.7 years (range = 55-76), with 13.7years of education (range = 9-19) were tested. Three (AB,PH, and JL) had confirmed or suspected damage to thehippocampal formation (31, 32); the remaining patients (RC,VF, PN, and JW) had alcoholic Korsakoff syndrome con-firmed by magnetic resonance imaging (31) or quantitativecomputed tomography (30). Scores on the Wechsler AdultIntelligence Scale-Revised (WAIS-R) averaged 105.9 (range= 98-116). Memory testing indicated moderate-to-severeimpairment in long-term memory with sparing of immediatememory. Scores on the Wechsler Memory Scale-Revised(WMS-R) were attention, 102.7; verbal, 71.0; visual, 77.6;general, 67.0; and delay, 58.7. Additional neuropsychologicaldata for these patients can be found elsewhere (15, 33).

Frontal lobe patients. The three frontal lobe patients (JD,MD, and JV) averaged 73 years of age at the time of testing(range = 71-76), 13.7 years of education (range = 12-16), andscores of 97.3 on the WAIS-R. Scores on the WMS-R averaged86.7, 106.7, 98.7, 104.7, and 100.7 for the five indexes listedabove. For additional neuropsychological data, see ref. 27.

Control subjects. Two groups of healthy subjects were alsotested, 8 in the experienced control group (3 women and 5men) and 7 in the inexperienced control group (4 women and3 men). These 15 subjects were selected to match the amnesicpatients with respect to age (mean, M, = 64.9 years, range =54-76 years) and two WAIS-R scores: information (controls,M = 21.9; amnesic patients, M = 21.4) and vocabulary(controls, M = 54.0; amnesic patients, M = 55.3).

Procedure. Eight deferred imitation tasks were used. Eachtask consisted of three actions, which were modeled in thesame order for each of the subjects (except in the inexperi-

enced control group, who were not shown the actions). Forfour tasks (causal), the order in which the actions weremodeled was based on enabling or causal relations. Forexample, in one taskwe demonstrated the Bernoulli effect. Weturned on a blow dryer, placed a balloon in the air stream, andthen tilted the blow dryer to show that the balloon wouldremain captured in the airstream. This effect cannot besuccessfully demonstrated in any other order. For the fourother tasks (arbitrary), the three actions were arbitrarilyordered. For example, we modeled the construction of apicture by folding and unfolding a piece of paper, cutting offits corners, and drawing a star. The same result can be obtainedregardless of the order in which the actions are performed.The test sessions were structured as tasks of learning and

recalling words. A list of 15 simple English words with four toeight letters was presented as a word-learning task. Thedeferred imitation tasks were presented as distractor tasks thatwere interposed between word presentation and efforts atword recall. In this way, we attempted to administer thedeferred imitation tasks as incidental events in the same waythat they are typically given to nonverbal infants.The study was conducted in four phases, which were video-

taped for later scoring. In phase 1 the experimenter (E) readthe words aloud to the subject with the instructions to attendto the words. Then E told the subject: "Now I am going to haveyou do some things to keep you from thinking of those words.I will give you different sets of objects which I want you tohandle, manipulate, or do whatever you like with them. It isimportant that you pay close attention to the objects, so pleasedo not talk while you are handling them." Then E brought out,one at a time, each of the object sets. The subjects were given1 min to manipulate each object set (baseline condition). Afterthe eighth set of objects was removed, the subject was asked torepeat aloud as many of the words on the list as could beremembered. E then instructed the subject that recall of thewords would be tested again later.

Phase 2 followed immediately. For subjects in the amnesic,frontal lobe, and experienced control groups, E said, "Now Iwant you to observe some objects while I manipulate them.Please pay close attention, do not talk, and try not to think ofthe words that I read to you earlier." E then modeled the targetactions for each of the eight sets of objects. For the inexpe-rienced control group, E gave similar instructions with theexception that the subjects were told to look at but not touchthe objects when theywere placed on the table in front of them.Finally, E asked the subjects in all four groups to recall thewords. After the recall test, E read the list again, telling thesubjects that recall would be tested again on the next day.On the next day, phase 3 began when the subjects were asked

to repeat the words that they had been read the day before. Ethen read the words again, instructing the subjects to remem-ber them. Then, each object set was presented for 1 min inexactly the same manner as during phase 1, and the subjectswere told they could handle them as they wished (uninstructedcondition).

In phase 4, which followed immediately, memory for theactions associated with each set of objects was tested explicitly(instructed condition). For the amnesic, frontal lobe, andexperienced control subjects, E said, "I want you to show meas accurately as you can remember the actions I performedyesterday and the order in which I performed them." For theinexperienced control subjects, E said, "For some participantsin this study, I demonstrated particular kinds of actions that Imodeled in a particular order with each set of objects. I wouldlike you to guess what those actions might have been and theorder in which I modeled them." After a subject was clearlyfinished with an object set, it was removed and the next set waspresented until the subject had an opportunity to manipulateall eight sets.

Psychology: McDonough et aL

Dow

nloa

ded

by g

uest

on

July

2, 2

020

7582 Psychology: McDonough et al.

Subjects were questioned on two occasions to assess theirunderstanding of the purpose of the study. After the unin-structed condition on the second day, E asked, "Did you thinkthat there was any connection between what you were doingand what I did the day before? When you manipulated theobjects today, were you influenced by the actions you saw meperform with them yesterday?" After the instructed conditionon the second day, E further inquired, "What do you think isthe purpose of this experiment? At any point in the experimentdid you expect to be asked to demonstrate the actions Iperformed for you on the first day?" Subjects who expressedany understanding of the purpose of the experiment wereasked when it occurred to them.The videotaped sessions were scored by assistants skilled in

coding deferred imitation tasks. Intercoder reliability ex-ceeded 95%. For each of the eight tasks, the total number ofdifferent target actions performed was calculated. To evaluatememory for the sequences in which the actions had beenmodeled, the number of different pairs of actions produced inthe correct target order was also calculated. For the lattermeasure, only the first occurrence of each target action wasincluded to reduce the likelihood of assigning credit for asequence being produced accidentally as the result of repeat-ing some of its elements. If the subject produced all threeactions in the target order, credit was given for two sequencepairs.

RESULTSActions. The number of actions performed by each subject

during each phase of the experiment was submitted to a 4 x2 x 3 mixed-design analysis of variance. The between-subjectsfactor was group (amnesic patients, frontal lobe patients,experienced control subjects, inexperienced control subjects).The within-subjects factors were task type (causal, arbitrary)and assessment condition (baseline, uninstructed, and in-structed).The mean number of actions (averaged across the four

causal and four arbitrary tasks) performed by each groupduring each of the three phases of the experiment (baseline,uninstructed, and instructed) is shown in Fig. 1. Main effectswere found for group [F(3, 22) = 15.02; P < 0.001] andassessment period [F(2, 44) = 51.80;P < 0.001]. There was alsoa group x assessment interaction [F(6, 44) = 17.28; P < 0.001]reflecting the fact that the four groups exhibited differentpatterns of performance across the three assessment condi-tions. Follow-up one-way analyses of variance were conductedto examine the locus of the interaction. The assessmentcondition was significant for all but the inexperienced controlsubjects. Post-hoc Tukey tests (P = 0.05) revealed that signif-icantly more actions were performed in the uninstructed and

12 * Baselineo Uninstructed

10 Instructed

X 8

E E4

instructed memory conditions than in the baseline conditionfor the experienced control subjects (baseline, M = 3.8;uninstructed, M = 8.6; instructed, M = 9.6) and frontal lobepatients (baseline,M = 2.5; uninstructed,M = 8.5; instructed,M = 8.2) but not for the amnesic patients (baseline, M = 2.4;uninstructed, M = 4.4; instructed, M = 4.0). Significantdifferences were also found among the groups in both theuninstructed and instructed memory conditions, but not in thebaseline condition. Specifically, Tukey tests (P = 0.05) re-vealed that the experienced control subjects and frontal lobepatients performed significantly more actions in both memoryconditions than either the inexperienced control subjects oramnesic patients, whose scores did not differ from each other.Finally, a main effect was also found for task type [F (1, 22) =10.89; P < 0.01], indicating that significantly more actions fromthe causal sequences (M = 5.7) were performed than from thearbitrary sequences (M = 4.1).

Sequences. The number of sequences (averaged across thefour causal and four arbitrary tasks) performed by subjects inthe order in which they were modeled is shown in Fig. 2. Themaximum number of sequences that can be performed on eachof the three-step tasks is two (a score of 1 for each ordered pairof actions); thus, a maximum score of 8 can be awarded foreach task type. A 4 x 2 x 3 mixed-design analysis of variance(group, task type, and assessment) revealed main effects forgroup [F(3, 21) = 15.83; P < 0.01] and assessment [F(2, 20) =26.55; P < 0.001], which were qualified by a significantinteraction between them, F(6, 40) = 5.66; P < 0.001. One-wayanalyses of variance for simple effects of assessment weresignificant for the experienced control subjects [F(2, 20) =35.25; P < 0.001] and frontal lobe patients [F(2, 20) = 11.69;P < 0.001] but not for the inexperienced control subjects or theamnesic patients. Tukey tests (P = 0.05) showed that both theexperienced control subjects and frontal lobe patients per-formed significantly more ordered sequences in the memoryconditions (uninstructed: experienced controlsM = 3.9; fron-tal lobe patientsM = 3.5; instructed: experienced controlsM= 4.5; frontal lobe patients M = 3.7) than in the baselinecondition (experienced controls M = 0.88; frontal lobe pa-tientsM = 0.33). Performance in the two memory conditionsdid not significantly differ for either group. A comparisonamong all four groups also showed significant differences inthe uninstructed [F(3, 21) = 11.19; P < 0.001] and instructedmemory conditions [F(3, 21) = 19.58; P < 0.001] but not in thebaseline condition. Tukey tests (P = 0.05) indicated that boththe experienced controls and frontal lobe patients scoredsignificantly higher than either the amnesic patients (unin-structed M = 1.4; instructed M = 1.3) or the inexperiencedcontrol subjects (uninstructedM = 0.57; instructedM = 0.57),whose scores did not differ from each other. Finally, a maineffect for task type [F(1, 21) = 10.82; P < 0.01] indicated that

6 * Baselineo Uninstructed

X X* Instructeda -3

Q0

35E T

xperiencecontrols controls patients patients

FIG. 1. Mean number of actions (maximum = 12 actions averagedacross the four causal and four arbitrary tasks) performed by eachgroup at each of the three assessment periods. Standard errors areshown for each score.

Experienced Inexperienced Amnesic Frontal lobecontrols controls patients patients

FIG. 2. Mean number of correctly ordered sequences (maximum =eight sequences averaged across the four causal and four arbitrarytasks) performed by each group at each of the three assessmentperiods. Standard errors are shown for each score.

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

July

2, 2

020

Proc. Natl. Acad. Sci. USA 92 (1995) 7583

more causal sequences were performed in the correct orderthan arbitrary sequences (2.2 vs. 1.3).Words Recalled. As expected, the amnesic patients recalled

fewer words (M = 0.15) than the other groups (frontal lobepatients M = 4.5; experienced control group M = 4.7; andinexperienced control group M = 6.2).

Questions About the Purpose of the Study. In response tothe questions asked after the uninstructed condition on day 2,all of the experienced control subjects and frontal lobe patientsand five of the amnesic patients believed that they wereinfluenced by the actions E had modeled. Note that a subjectwho believes that she or he was influenced by earlier events isnot necessarily stating that she or he remembers the events.Indeed, three of these five amnesic patients added that theyhad difficulty remembering what had occurred on the previousday. In response to the questions asked after the instructedcondition on day 2, one experienced control subject and noamnesic or frontal lobe patient described correctly that theyknew the purpose of the study on the first day. But by thesecond day, four of the amnesic patients thought the study wasto evaluate their memory problem in general. The remainingamnesic patients (n = 3), experienced control subjects (n = 7),and frontal lobe patients (n = 3) stated that they believed theywere to remember the actions modeled.

DISCUSSIONThe results indicate that amnesic patients are impaired indeferred imitation tasks, tasks that are commonly used to testrecall in infants. First, the experienced control subjects repro-duced more actions and action sequences than the amnesicpatients in both the uninstructed and instructed memoryconditions. Second, performance by the amnesic patients didnot significantly differ at any assessment period from that ofthe inexperienced controls, who never saw the sequences.Third, the pattern of results exhibited by the experiencedcontrol subjects resembled the patterns found in verbal tasksof recall. Specifically, more of the causal tasks were performedthan the arbitrary tasks, just as in previous studies using verbalmeasures of recall with older children and adults (24, 25). Thissame result has also been reported in other studies usingdeferred imitation with infants (7, 17).The results also showed that the patients with frontal lobe

lesions were able to recall the actions and also the order inwhich they were modeled. This finding might seem surprising,given that some tests of temporal order are sensitive to frontallobe dysfunction. However, deficits in temporal ordering aretypically found when tasks require frontal lobe patients toremember a sequence of unrelated episodes or unrelated wordlists (29, 34) or to judge the recency of presented stimuli (35).The deferred imitation tasks in the present study involvedrelated rather than unrelated events. Accordingly, the patientswith frontal lobe lesions may have been able to perform wellbecause the information to be remembered was coherent.The amnesic patients were not able to perform the se-

quences they had seen in either the instructed or uninstructedconditions. Thus, the uninstructed condition did not automat-ically or nonconsciously prime them to perform the sequences.Although the number of actions and sequences performed bythe amnesic patients did increase from baseline to the twomemory conditions (marginally in the case of the actions andnot significantly in the case of the sequences), this smallimprovement in performance relative to baseline likely reflectssome residual declarative memory capacity in the amnesicpatients. Note that, even in the verbal memory recall task, theamnesic patients obtained an above-zero score. We concludethat the deferred imitation task depends largely on declarativememory rather than on skill learning or priming.

Because infants can do deferred imitation, they must have anearly capacity for declarative memory that can support rela-

tively long-lasting recall memories. What, then, accounts forinfantile amnesia? Although recent research in monkeys andhumans indicates that the primate hippocampus matures rel-atively early (36) and probably earlier than the neocortex (13,37), the hippocampus is only one part of a large neural systemimportant for declarative memory (38). It is not yet clear whenthis neural system becomes functionally mature. One impor-tant feature of early neural development is that there is initiallyan excess of synaptic connections and that the final adultpattern of connectivity is achieved by regressive events thatinclude cell death, synapse elimination, and collateral elimi-nation (39). For example, the density of synapses, corrected forvolume change, reaches a maximum of about 150% of the adultlevel at 8 months of age in the visual cortex (40). An earlierstudy of frontal cortex suggested that in this region maximumsynaptic density is reached more slowly, at about 2 years of age(41). Moreover, by 3 to 4 years of age, metabolic activity in thecortex, as measured by positron emission tomography, reachestwice the adult level (42). These stages of development arefollowed by gradual synapse elimination (40) and the declineof metabolic activity (42) during a period of several years.Although the functional implications of these regressive eventsare still unknown, it is apparent that the human brain reachesits final adult pattern of connectivity only after a protractedperiod of development. This gradual development, reflected inboth progressive and regressive events, provides a way tounderstand why declarative memory may not be sufficientlydeveloped to provide rich and enduring recollections, eventhough some capacity for declarative memory is available inearly life.The present study adds to a growing body of literature

suggesting that some of the tasks used to test memory in humaninfants are tasks of declarative memory, tasks that are failed byamnesic patients and by adult and infant monkeys with limbic/diencephalic lesions (13-15). This literature provides compel-ling evidence that human infants have an early capacity fordeclarative memory. Previous studies of infant memory haveused relatively short delays (e.g., delayed nonmatching tosample tasks and visual paired-comparison tests) and havedepended primarily on recognition tests (13, 15). Deferredimitation provides a reliable and valid method to test the earlycapacity for recall memory and to explore the basis of infantileamnesia. Deferred imitation should be particularly useful instudies spanning long delays-i.e., delays that span the severalyears over which infantile amnesia is observed (19).

We thank Joyce Zouzounis, Kamilla Willoughby, and JenniferSwerdlow for research assistance. This work was supported in part bya postdoctoral fellowship from the Center for Human InformationProcessing at the University of California, San Diego (to L.M.),National Science Foundation Research Grant BNS89-19035 (toJ.M.M.), the Medical Research Service of the Department of VeteransAffairs, National Institute for Mental Health Grant MH24600, theOffice of Naval Research, and the McKnight Foundation (to L.R.S.).

1. Mishkin, M., Malamut, B. L. & Bachevalier, J. (1984) in Neuro-biology ofLearning and Memory, eds. Lynch, G., McGaugh, L. &Weinberger, N. M. (Guilford, New York), pp. 65-77.

2. Squire, L. R. (1987) Memory and Brain (Oxford Univ. Press, NewYork).

3. Weiskrantz, L. (1990) Philos. Trans. R. Soc. London 329, 99-108.4. Mandler, G. (1985) Cognitive Psychology: An Essay in Cognitive

Science (Erlbaum, Hillsdale, NJ).5. Schacter, D. L. (1985) in Memory Systems of the Brain: Animal

and Human Cognitive Processes, eds. Weinberger, N., Lynch, G.& McGaugh, J. (Guilford, New York), pp. 351-379.

6. Mandler, J. M. (1990) in The Development and Neural Bases ofHigher Cognitive Functions, ed. Diamond, A. (N.Y. Acad. Sci.,New York), pp. 485-516.

7. Mandler, J. M. & McDonough, L. (1995) J. Exp. Child Psychol.,in press.

Psychology: McDonough et al.

Dow

nloa

ded

by g

uest

on

July

2, 2

020

7584 Psychology: McDonough et al.

8. Nelson, K. & Brown, A. L. (1978) in Memory Development inChildren, ed. Ornstein, P. A. (Erlbaum, Hillsdale, NJ), pp. 233-241.

9. Squire, L. R., Knowlton, B. & Musen, G. (1993) Annu. Rev.Psychol. 4, 453-495.

10. Cohen, N. J. & Squire, L. R. (1980) Science 21, 207-209.11. Bachevalier, J. & Mishkin, M. (1984) Behav. Neurosci. 98,

770-778.12. Schacter, D. L. & Moscovitch, M. (1984) in Infant Memory, ed.

Moscovitch, M. (Plenum, New York), pp. 173-216.13. Diamond, A. (1990) in The Development and Neural Bases of

Higher Cognitive Functions, ed. Diamond, A. (N.Y. Acad. Sci.,New York), pp. 267-317.

14. Bachevalier, J., Brickson, M. & Haggar, C. (1993) NeuroReport 4,77-80.

15. McKee, R. D. & Squire, L. R. (1993)J. Exp. PsychoL Learn. Mem.Cognit. 19, 397-404.

16. Diamond, A. (1988) Child Dev. 59, 523-527.17. Bauer, P. J. & Mandler, J. M. (1989) Dev. Psychol. 25, 197-206.18. Meltzoff, A. N. (1988) Child Dev. 56, 217-225.19. McDonough, L. & Mandler, J. M. (1994) Memory 2, 339-352.20. Corkin, S. (1968) Neuropsychologia 6, 255-265.21. Bates, E. (1979) The Emergence of Symbols: Cognition and

Communication in Infancy (Academic, New York).22. Warrington, E. K. & Weiskrantz, L. (1974) Neuropsychologia 12,

419-428.23. Graf, P., Squire, L. R. & Mandler, G. (1984) J. Erp. Psychol.

Learn. Mem. Cognit. 10, 164-178.24. O'Brien, E. J. & Myers, J. L. (1989) Mem. Cognit. 15, 419-427.25. Mandler, J. M. (1986) Lang. Cognit. Proc. 1, 309-320.26. Trabasso, T., Secco, T. & Van der Broek, P. (1984) in Learning

and Comprehension of Text, eds. Mandl, H., Stein, N. L. &Trabasso, T. (Erlbaum, Hillsdale, NJ), pp. 83-111.

27. Janowsky, J. S., Shimamura, A. P., Kritchevsky, M. & Squire,L. R. (1989) Behav. Neurosci. 103, 548-560.

28. Stuss, D. T. & Benson, D. F. (1986) The Frontal Lobes (Raven,New York).

29. Shimamura, A. P., Janowsky, J. S. & Squire, L. R. (1991) inFrontal Lobe Functioning and Dysfunction, eds. Levin, H. D.,Eisenberg, H. M. & Benton, A. L. (Oxford Univ. Press, NewYork), pp. 173-195.

30. Shimamura, A. P., Jernigan, T. L. & Squire, L. R. (1988) J.Neurosci. 8, 4400-4410.

31. Squire, L. R., Amaral, D. G. & Press, G. A. (1990)J. Neurosci. 10,3106-3117.

32. Polich, J. & Squire, L. R. (1993) Electroencephalogr. Clin. Neu-rophysiol. 86, 408-417.

33. McKee, R. D. & Squire, L. R. (1992) J. Neurosci. 12, 3765-3772.34. Schacter, D. L. (1987) Psychobiology 15, 21-36.35. McAndrews, M. P. & Milner, B. (1991) Neuropsychologia 29,

849-859.36. Kretschmann, H., Kammradt, G., Krauthausen, I., Sauer, B. &

Wingert, F. (1986) Bibl. Anat. 28, 27-52.37. Bachevalier, J. (1990) in The Development and Neural Bases of

Higher Cognitive Functions, ed. Diamond, A. (N.Y. Acad. Sci.,New York), pp. 457-484.

38. Squire, L. R. & Zola-Morgan, M. (1991) Science 253, 1380-1386.39. Cowan, W. M., Fawcett, J. W., O'Leary, D. M. & Stanfield, B. B.

(1984) Science 225, 1258-1265.40. Huttenlocher, P. R. & deCourten, C. (1987) Hum. Neurobiol. 6,

1-9.41. Huttenlocher, P. R. (1990) Neuropsychologia 28, 517-527.42. Chugani, H. T., Phelps, M. E. & Mazziotta, J. C. (1987) Ann.

Neurol. 22, 487-497.

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

July

2, 2

020