Journal oj Personality and Social Psychology1972, Vol 21, No 2, 204-218

COGNITIVE AND ATTENTIONAL MECHANISMSIN DELAY OF GRATIFICATION1

WALTER MISCHEL," EBBE B. EBBESEN, AND ANTONETTE RASKOFF ZEISS

Stanford University

Three experiments investigated attentional and cognitive mechanisms in delayof gratification In each study preschool children could obtain a less preferredreward immediately or continue waiting indefinitely for a more preferred butdelayed reward Experiment I compared the effects of external and cognitivedistraction from the reward objects on the length of time which preschoolchildren waited for the preferred delayed reward before forfeiting it for thesake of the less preferred immediate one. In accord with predictions from anextension of frustrative nonreward theory, children waited much longer for apreferred reward when they were distracted from the rewards than when theyattended to them directly Experiment II demonstrated that only certaincognitive events (thinking "fun things") served as effective ideational dis-tractors Thinking "sad thoughts" produced short delay times, as did thinkingabout the rewards themselves In Experiment III the delayed rewards werenot physically available for direct attention during the delay period, and thechildren's attention to them cogmtively was manipulated by prior instructionsWhile the children waited, cognitions about the rewards significantly reduced,rather than enhanced, the length of their delay of gratification Overall, atten-tional and cognitive mechanisms which enhanced the salience of the rewardsshortened the length of voluntary delay, while distractions from the rewards,overtly or cogmtively, facilitated delay The results permit a remterpretation ofbasic mechanisms in voluntary delay of gratification and self-control

As early as 1890, William James contendedthat attentional processes are at the very coreof the self-control phenomena usually sub-sumed under the term "will" or, since James'stime, under the concept "ego strength"According to James (1890) "Attention witheffort is all that any case of volition impliesThe essential achievement of will is to attendto a difficult object . [p 549]." In con-trast, psychoanalytic theories of self-controlemphasize unconscious processes and motiva-tional dynamics, as well as internalization ofvalues and intrapsychic conflicts to explainself-control phenomena. In spite of thisshift in emphasis away from attentional topsychodynamic interpretations of self-control,some strands of evidence suggest possible

1 This research was supported by Research GrantMH-6830 from the National Institutes of Health,United States Public Health Service Portions of thispaper were presented by the first author in aninvited address, "Personality and Cognition," at themeeting of the Western Psychological Association,Los Angeles, April 1970

2 Requests for reprints should be sent to WalterMischel, Department of Psychology, Stanford Uni-versity, Stanford, California 94305

links between attentional processes and self-regulatory mechanisms

In particular, beginning with Hartshorneand May (1928), a few correlations havebeen found between indexes of moral behaviorand measures of attention or resistance todistraction on mental tests (e g., Grim, Kohl-berg, & White, 1968). On the basis of suchcorrelations, it has been suggested that theindividual's ability to resist temptation maybe facilitated by how well he attends to atask. In most experimental "resistance totemptation" paradigms, yielding to tempta-tions, such as cheating, depends on the sub-ject's being distracted from the main task towhich he is supposed to be attending In thosesituations a subject's ability to resist distrac-tion may automatically make it easier forhim to refrain from temptations such ascheating (Grim et al, 1968)

Using experimental rather than correla-tional methods, Mischel and Ebbesen (1970)have explored a different link between atten-tion and self-control in the context of thedelay-of-gratification paradigm In that study,preschool children sat waiting for a preferred

204

COGNITIVE AND AITENTIONAL MECHANISMS 205

reward which was available at a later time.At any moment the children could signal toterminate the waiting, thereby obtaining aless preferred reward, while forfeiting themore desired one. The experiment investigatedthe effects of attention to the goal objects onthe length of time that the children actuallywaited Specifically, while the children waited,they were given the opportunity to attend tothe delayed and/or to the immediately avail-able reward or to neither reward

In accord with several previous theoriesdiscussed by Mischel and Ebbesen (1970),it had been expected that attention to thedelayed reward would facilitate delay of grati-fication. In part it was expected that makingthe reward objects salient might facilitate"time binding" by permitting the subject toengage in self-persuasion and anticipatorygratification For example, he might sustainhis delay by imagining how satisfying thepreferred outcome would be (eg., how goodit would taste) when it became available. Infact, the findings obtained by Mischel andEbbesen were exactly opposite to the pre-dictions. It was found that if the child couldattend to either or both of the rewards, hewaited much less than if he could attend toneither reward during the delay period Inaddition, the length of time which the chil-dren waited with only the immediate rewardavailable for attention was similar to thetime they waited with only the delayed rewardavailable Finally, when both rewards wereavailable, children waited a slightly shortertime. Thus, children waited most readilywhen neither the delayed nor the immediatereward was available for attention during thedelay period, and they waited a relativelyshort time when any reward was available

Observation of the children's spontaneousbehavior during delay of gratification sug-gested that the mechanism used by manyyoungsters to sustain their voluntary delayinvolved suppressing rather than enhancingattention to the rewards. The children seemedto reduce the subjective aversiveness of delayof reward by engaging in covert and overtdistracting responses such as staring at themirror, covering their eyes with their hands,and talking to themselves. These responsesseemed to divert their attention away from

the frustration-inducing rewards. Further-more, self-induced distractions seemed to befairly easy to maintain in the condition inwhich no rewards were facing the children,but seemed very difficult to maintain whenany of the rewards were facing the children.Apparently, the children were able to delaylonger when neither reward was available forattention because in this condition it waseasier for them to avoid or suppress cognitionsabout the rewards

Post hoc, the obtained results become mostunderstandable if delay of gratification is seenas a frustration situation Indeed, the essenceof frustration is a delay or interruption inan expected and desired outcome (Mandler,1964). The necessity to delay in order toobtain a more gratifying reward may be deter-mined externally by physical barriers or otherpeople. In self-control and voluntary delay ofreward, this delay is self-imposed. If a persondesires the delayed outcome, he must imposethe frustrative waiting situation upon himself,foregoing the immediately available outcomefor the sake of the more desirable but delayedalternative.

As Amsel's (1958, 1962) "frustrative non-reward theory" has suggested, frustration in-volves an actively aversive effect It followsthat any conditions that enhance the aversive-ness of frustration should make it harder towait. Although Amsel's work with frustrationeffects in animals has concentrated on thescheduling of prior reinforcement, it seemsreasonable that any cues that enhance atten-tion to what one wants but cannot haveshould increase the aversiveness of frustra-tion This interpretation of the frustrationeffect suggests that attending to the rewardscognitively, rather than helping to bridge thedelay period, may make it more aversiveto delay gratification, and therefore lead toshorter waiting time The present threeexperiments were designed to test thisproposition

If our speculations regarding the frustrativeeffects of delay are correct, then delay ofgratification should be enhanced when thesubject can readily transform the aversivewaiting period into a more pleasant non-waiting situation This line of reasoning sug-gests that voluntary delay of reward should

206 W. MISCHEL, E EBBESEN, AND A ZEISS

be enhanced by any overt or covert activitiesthat serve as distractors from the rewards.Through self-distraction, the subject shouldbe able to suppress or avoid the aversivenessof wanting gratification but not having it,and thus convert the frustrative delay-of-reward situation into a psychologically lessaversive condition. Therefore, it was expectedthat overt activities and internal cognitionsand fantasy which could help the subject todistract himself from the rewards would in-crease the length of time which he woulddelay gratification

To investigate the role of attention andcognition in delay of gratification, we gaveeach subject either an overt activity, a cogni-tive activity, or no activity to engage induring the delay period. The overt and cogni-tive activities were designed to reduce theprobability that the subjects would be at-tending to the reward during the delay periodWe then assessed how these alternatives af-fected the subjects' voluntary delay time, incomparison to the group of subjects which wasnot supplied with a distractor Consistentwith our extension of frustrative nonrewardtheory, we predicted that voluntary delay ofgratification should be increased by anycovert or overt activities that distract thesubject from the anticipated outcomes and,conversely, should be diminished by attentionto the rewards during the delay period.

EXPERIMENT I

To test our expectations, a study was de-signed with a delay-of-gratification paradigmin which aversive frustration was deliberatelymade high For this purpose, preschool sub-jects were faced with both the immediatelyavailable and the more preferred but delayedreward during the delay period 3 All children

8 This attentional condition was used because itpresumably would provide maximum frustrationcues When the subject attends to the immediatereward and is tempted to take it, he is frustratedby remembering the contingency that attainment ofthe reward now prevents his getting the preferredreward later When the subject attends to the delayedreward, he is frustrated by the fact that he wantsit now but cannot have it yet When he attends toboth objects, both of the above aversive frustrationsoccur, and, hence, delay should be most difficultfor him Indeed, this attentional condition, in which

could signal at any time to terminate thewaiting period, thereby forfeiting the morepreferred gratification but attaining the lessdesirable one immediately.

MethodDesign

The independent variable in this study was amanipulation designed to permit the subject to dis-tract himself from the reward objects for which hewas waiting In all conditions, while the subjectswaited, the immediate and delayed rewards wereboth physically available for direct attention Thedependent variable was the length of time the chil-dren remained alone in the room before they rang abell and thereby ended the delay period

Two methods of self-distraction were used Onetechnique involved an external activity; in the othermethod, instructions were given to generate internalcognitive activity In the external activity, the childwas given the opportunity to play with an attractivetoy while he waited In the cognitive activity group,instructions were given to increase the probabilitythat the child would think pleasant and distractingthoughts while he was waiting A control group wasdesigned to determine how long the children wouldwait for the delayed reward without either theexternal or the internal distractors provided by theexperimenters To control for the effects of playingwith an attractive toy or thinking pleasant thoughtsindependent of waiting for the delayed reward, twoadditional control groups of children were giveneither the toy or the cognitive sets but no delayedreward contingency

Thus, a total of five groups were employed Group1 waiting for delayed reward with external distrac-tor (toy), Group 2 waiting for delayed reward withinternal distractor (ideation) j Group 3 waiting fordelayed reward (no distractor); Group 4 externaldistractor (toy) without delay-of-reward waitingcontingency, Group 5 internal distractor (ideation)without delay-of-reward waiting contingency

SubjectsThe subjects were SO children (25 boys and 25

girls) from the Bmg Nursery School of StanfordUniversity They ranged in age from 3 years 6months to 5 years 6 months, with a mean age of4 years 6 months Six additional children began butdid not complete the experimental procedures be-cause they did not comprehend the instructions Fivemales and 5 females, equated for mean age, wererandomly assigned to each of the five conditionsOne male and one female served as experimentersSpecifically, within each condition, the male experi-menter ran 3 male and 2 female subjects, and thefemale experimenter ran 3 female and 2 male subj'ects

both rewards are present, was the one that hasproduced the shortest waiting time in a previousstudy (Mischel & Ebbesen, 1970)

COGNITIVE AND ATTENTIONAL MECHANISMS 207

ProcedureThe experimental room and setting was similar to

that previously described (Mischel & Ebbesen, 1970)An addition was a barrier, behind which the experi-menter, but not the children, could see and reachBehind this barrier was placed a "Slinky" (a toyspring) and an opaque cake tin Under the cake tinwas a small marshmallow and a stick pretzel A boxof attractive battery- and hand-operated toys wason a second table next to the barrier A table andchair were against one wall, and on the table wasa desk bell

When the experimenter escorted a child to theexperimental room, he first showed the child thebox of toys and explained that they would playwith the toys later on m the session After finishinga brief demonstration of one or two of the toys, theexperimenter escorted the subject to the table withthe desk bell and asked the child to sit in the chairin front of the table The experimenter then intro-duced and piacticed the child's use of the bell signalThey played a "game" m which the experimenterrepeatedly stepped out of the room, closed the door,but returned as soon as the child signaled Everytime the bell was rung, the experimenter thus im-mediately returned to the room The procedure wasthe same as the one previously described (Mischel &Ebbesen, 1970), with the exception that in thepresent study the signal was a bell Thereafter, theexperimenter consulted a concealed slip of paperwhich informed him by a prearranged randomschedule of the condition to which the child wasto be assigned

Delay-of-Grattficatwn ContingencyInstructions

For subjects assigned to the three delay-of-gratification conditions (Groups 1, 2, and 3), theexperimenter next removed the cake tin from behindthe barrier and placed it on the table in front of thechild, giving these instructions

Let's see what's under here I'll bet it's a surpriseOh boy, look at that A marshmallow and a pretzelWhich would you like to eat? You can eat eitherthe marshmallow or the pretzel [At this point thechild chose which one he wanted to eat ] Oh, youknow what? I have to go out of the room now,and if you wait until I come back by myself thenyou can eat this one [pointing to chosen object]right up But, you know, if you don't want towait you can ring the bell and bring me backanytime you want to But if you ring the bell thenyou can't have this one, but you can have thatone (pointing to the unchosen object) So, if youring the bell and bring me back then you can'thave the , but you can havethe

The experimenter then assessed the child's compre-hension by asking three questions "Can you tell me,which do you get to eat if you wait for me to come

back by myself?" "But if you want to, how can youmake me come back?" "If you ring the bell andbring me back, then which do you get?" As pre-viously mentioned, six children were eliminated fromthe study Four failed to pass these questions, andthe other two subjects were lost because one ate thefood objects while the experimenter was out of theroom, and the other one refused to ring the bellduring training

In one of the three delay-of-gratification condi-tions (Group 3), subjects did not receive either thethinking or the toy activity distraction instructions,but they did receive the foregoing waiting contin-gency instructions In this condition, after the sub-jects answered the three comprehension questions cor-rectly, they were simply told "I have to leave theroom now And if you want to you can ring the bellwhenever you want to and bring me back When Icome back, whether you ring the bell or wait for meto come back by myself, we'll play with all my toys."

In Groups 4 and S—the two conditions with nodelay-of-reward contingency—the foregoing instruc-tions of course were not given Instead, the experi-menter, after stating that he would have to leave,merely looked at some papers, while telling the childthat he had to "check something" before he left andthen shuffled through the papers for approximatelyas long as it would take to give the waitingcontingency instructions

Distraction through Overt ActivityInstructions

In the two overt distraction conditions (Groups1 and 4), the child was left alone in the room witha potential distracting activity that involved playingwith a toy (the Slinky) In Group 1 each subjectwas also given the delay-of-gratification contingencyinstructions and thus was waiting with the possibilityof getting the preferred food object if he waited longenough, and the less preferred object if he did notdelay long enough In Group 4 the child was leftalone merely to play as long as he wished In bothof these groups, prior to leaving the room, the experi-menter placed the Slinky on the floor and informedthe child that he could play with the Slinky on thefloor for as long as he wanted, that he could ringthe bell whenever he wanted to bring back theexperimenter, and that, regardless of whether thechild rang the bell or waited, he could play withthe toys when the experimenter came back

Distraction through Cognition-InducingInstructions

In Groups 2 and 5, before leaving the room, theexperimenter gave the subject instructions designedto encourage the child to generate his own thoughtsand covert cognitive activities while waiting Hesaid "Oh, while I'm gone you can think of anythingthat's fun to think of, for as long as you want to,if you want to Can you tell me something to thinkabout that's fun?" (The experimenter paused for the

208 W. MISCHEL, E. EBBESEN, AND A ZEISS

I5r

Ld

P 10

(3

P

s

NODISTRACTION

TOY THINKFUN

WAITING FOR CONTINGENT REWARD

TOY THINKFUN

NONCONTINGENT WAITING

FIG 1 Mean number of minutes of voluntary waiting time for each conditionin Experiment I

child's examples and said "Yes" no matter what thesubject said) The experimenter then added otherexamples "You can also think about singing songs,or think of playing with toys or anything that is funto think o f" His final departing instructions wereidentical to those given in the other groups

Recall that Groups 2 and S differed in that onlythe children in Group 2 expected to get a rewardif they waited long enough, although both groupscould remain alone and "think about fun things" foras long as they wished before terminating the delayby signaling

In all groups, when the experimenter returned heasked the child "What happens now?" No childfailed to respond correctly (either verbally or byeating the proper food reward spontaneously) In allconditions the session ended with the experimenterand child playing with the toys as had beenpromised

Sequence

To review, the order of events were as follows(a) The experimenter demonstrated some toys to thechild (b) The child was taught how to bring theexperimenter back into the room (by ringing thebell) (c) In Groups 1, 2, and 3, the children werepresented with the delay contingency, and compre-hension questions, while m Groups 4 and 5 theexperimenter merely explained that he had to leavethe room and that the children could bring himback whenever they wished to (d) The overt dis-tractors were presented in Groups 1 and 4, and thecovert, in Groups 2 and 5 (e) Finally, all children

were reminded that no matter what they did (ringthe bell or not) when the experimenter returned, theywould play with the toys

Results

Delay of Gratification

The mean length of waiting time was com-puted for each condition (Figure 1) The firstresult to note is the extremely low mean delaytime found in the delay-of-reward conditionin which no distractor was available In thiscondition, since both the chosen and unchosenrewards were present, and no distractor wasavailable, the attention paid to the rewardsshould have been fairly high. The low meandelay time of less than | a minute foundfor this group replicates the low mean waitingtime found in the previous comparable studywhen children also waited with both rewardspresent, and no distractor was available(Mischel & Ebbesen, 1970)

As can be seen in Figure 1, the mean lengthof time which the children waited was muchgreater when they had available either anexternal or a cognitive distractor during thedelay period, the mean delay times in thelatter two conditions being 8.59 and 12 12minutes, respectively

COGNITIVE AND ATTENTIONAL MECHANISMS 209

Table 1 presents a one-way analysis ofvariance for the mean waiting times and fourorthogonal contrasts (Winer, 1962). Theoverall effect of conditions was highly signifi-cant ( />< .001). Orthogonal contrasts werecomputed to determine the exact sources ofthis effect and to test specific hypothesesabout the effect of distractors on the lengthof voluntary delay of reward.

The first orthogonal contrast (Ci in Table1) compared the three delay-of-gratificationcontingency conditions (Groups 1, 2, and 3)with the two conditions in which this con-tingency was absent (Groups 4 and 5). Thiscontrast yielded a highly significant effect(p < .001), indicating that giving the childrena reward for which to wait greatly increasedthe length of time which they spent alone inthe experimental room

It was also predicted that waiting wouldbe long only in those conditions in which achosen reward was contingent on delay and adistractor was available As Figure 1 reveals,the availability of the desired but delayedgratification yielded a mean delay time ofless than | a minute when the subject hadno overt or cognitive distractions available toreduce or avoid frustration while he was at-tending to the rewards. The second contrast(C2 in Table 1) was computed with only thethree contingent delay-of-gratification condi-tions. This contrast compared the two dis-traction conditions (Groups 1 and 2) with theone no-distraction condition (Group 3) TheF for this contrast was highly significant(p < .001) and strongly confirmed the pre-diction that children would wait longer for

TABLE 1ANALYSIS OF VARIANCE AND ORTHOGONAL CON-

TRASTS FOB. MEAN WAITING TIMES As AFUNCTION OF DELAY CONTINGENCY

AND DISTRACTION CONDITIONS

Source

BetweenCi

C3C4

Error

df

41111

45

MS

28914282655162736

151

F

19 21**28 46**43 54**418*<1

TABLE 2

NUMBER OF CHILDREN WAITING OR NOT WAITINGTO THE CRITERION (15 MINUTES) AS A

FUNCTION or DELAY CONTINGENCYAND DISTRACTION CONDITIONS

(EXPERIMENT I)

Waitingto

criterion

NoYes

Waiting for contingentreward a

No dis-traction

100

Toy asdistrac-

tion

64

Thinkingfun

thingsas dis-

traction

46

Noncontingentwaitmgb

Toy asdistrac-

tion

100

Thinkingfun

thingsas dis-

traction

100

bxa= 8 18, df = 2, p < 025

*p < 05**p < 001

rewards if a distractor from the rewards wereavailable during the delay period.

The third contrast (C8 in Table 1) com-pared the two distraction conditions in whichchildren were waiting for a reward. The dif-ference between the mean length of timeswaited in these two conditions reached signifi-cance at the .05 level, a longer mean delaytime was found m the cognitive distractioncondition than in the overt distraction condi-tion. Because this difference was not verygreat, a more stringent statistical test was alsocomputed, using the studentized range statis-tic and the Tukey (a) procedure for a pos-teriori t tests (Winer, 1962). With this morestringent statistic, the difference between thetoy distraction and the cognitive distractionwhen children were waiting for a reward didnot approach significance (t = 2.87).

A fourth contrast (C4 in Table 1) com-pared the two conditions in which reward wasnot contingent on delay (Groups 4 and 5).It was not significant (F < 1).

Another index of waiting behavior is thenumber of children who waited the full ISminutes (i e, until the experimenter returnedby himself) These results (Table 2) were inthe same direction as those reported for themean waiting times. A chi-square comparedthe number of children waiting to criterionwith the number who did not in the threecontingent delay-of-gratification conditions,and yielded a significant effect (x2 = 8.18,df = 2, p < .025). When delay of gratifica-tion was attempted in the presence of re-

210 W. MISCHEL, E EBBESEN, AND A. ZEISS

wards and no distractors were available tosuppress attention from them, not a singlechild waited to criterion. When distractorswere available cogmtively or overtly, half ofthe subjects waited to criterion Moreover,when the distractors were available, but re-wards were not contingent on waiting, not asingle subject waited to criterion

Before discussing the present results, twoadditional experiments will be described.These studies are intended to clarify furtherthe cognitive and attentional mechanisms thatseem most crucial for effective voluntarydelay of gratification.

EXPERIMENT II

The effects found in the first study werestrong and in accord with theoretical expecta-tion The results clearly supported thehypothesis that effective delay behavior isgreatly enhanced by the avoidance or reduc-tion of the frustrative aspects of delay ofgratification Such reduction presumably wasachieved when the subjects shifted attentionaway from the potential gratification andinstead distracted themselves with competingcognitions or with overt activity.

The findings concerning the potency ofthe instruction-induced cognitive distractionsseemed especially provocative It also wouldbe important to determine how the substan-tive content of cognitions (as generated byvarious types of instructions) affects subse-quent delay behavior A second experimentwas designed to explore this topic

It was assumed that the effects of instruc-tion-produced cognitive content on voluntarydelay would parallel those found by manipula-tion of external stimulus objects Therefore,it was predicted that delay of gratificationwould be short when the frustration was madehigh by directing the children to think aboutthe rewards We also expected that thecontent of thoughts could influence theireffectiveness in bridging the delay-of-rewardperiod Thus, we anticipated that aversivecognitions, as in "thinking sad thoughts,"should be relatively ineffective compared tosuch positive cognitions as ideating about"fun things." Such a result could be expectedfor several reasons; first, aversive thoughts

might be avoided and not employed effec-tively as distractors Alternatively, if sadthoughts were generated by the children, theadditional aversiveness might lead them toterminate the already aversive waiting situa-tion. Consequently, it was predicted that delayof gratification would be longer when thecognitions were affectively positive distrac-tors and shorter when the cognitions wereaffectively negative

MethodDesign

A three-condition study was designed which variedthe types of instructions given to the subjects justbefore they began to wait for rewards The instruc-tions were intended to induce in the subject varioustypes of ideation during the delay-of-gratiflcationperiod One condition was a replication of a pre-viously run cell Subjects here were instructed thatthey could think about fun or happy thoughtswhile waiting In a second condition the subjectswere told that they could think unhappy or sadthoughts while they waited In the final conditionthe children were instructed that they could thinkabout the reward objects In all conditions bothrewards were again present during the waiting periodThe dependent variable was the length of time thatthe children waited for their more preferred rewardbefore terminating and settling for the less prefeiredoutcome

SubjectsThirty-two children from the Bing Nursery School

of Stanford University were the subjects in thisstudy Six of them were lost because of incompleteunderstanding of the instructions or because theyate one of the reward objects while waiting for theexperimenter The subjects in the final analysis rangedin age from 3 years 9 months to 5 years 3 months,with a mean age of 4 years 9 months The final datawere based on 10 subjects in each of the two newconditions and 6 subjects in the replicated conditionSex ratios and ages were equated across groups

ProcedureThe initial procedure in the three conditions of this

study was identical to that in Groups 1, 2, and ?of Experiment I, in which the children were waitingfor reward objects The only differences came afterthe child had answered the usual three comprehen-sion questions correctly As in the earlier studies,after the child answered these three questions, theexperimenter was informed as to which one of thefollowing three sets of instructions he was to give,the decision had been determined randomly The"think fun" distraction instructions were identicalto those described for the equivalent group inExperiment I

COGNITIVE AND ATTENTIONAL MECHANISMS 211

In the "thtnk sad" distraction instructions, thekey phrases as the experimenter started to departwere "Oh while I'm gone you can think of anythingthat is sad to think of, for as long as you want to,if you want to. Can you tell me something that issad to think of?" (The experimenter paused forthe child's examples and said "Yes" regardless ofthe answers) The experimenter then added otherexamples. "You can also think of falling down andgetting a bloody knee which hurts a lot, or you canthink of crying with no one to help You can thinkof anything that makes you unhappy"

The "think food reward" instructions directed at-tention to the reward objects Therefore, the experi-menter simply mentioned both reward objects asthings the child could think about while the experi-menter was gone "Oh, while I'm gone you canthink of the marshmallow and the pretzel for aslong as you want to, if you want to " This thoughtwas repeated several times in reversed order andrephrased form, with the experimenter noting thatthe child could think anything he wanted to aboutthe pretzel and the marshmallow for as long as hewanted to

In all groups the final departing instructions wereidentical and the same as described in Experiment I,thus, they emphasized that the child could ring thebell or wait, and in either event the experimenterand child would play with all the toys at the end

Results

The mean number of minutes waited in thethree conditions ("think fun," "think sad,"and "think food") is summarized in Figure 2A one-way analysis of variance of these datais summarized in Table 3 It can be seen thatmost of the between variance is accountedfor by the comparison between the think funcondition and the other two conditions, thinksad and think food This finding is exactlyin accord with the prediction The directionof the difference was as expected1 namely,children waited longer for a reward when they

TABLE 3ANALYSIS OF VARIANCE AND ORTHOGONAL CON-

TRASTS FOR MEAN LENGTH OF DELAY OFGRATIFICATION AS A FUNCTION OFVARIOUS THOUGHT-INDUCING IN-

STRUCTIONS (EXPERIMENT II)

I5r-

Source

BetweenG!C2

Error

if

2\1

23

MS

17613489

33322

F

547*10 84**<1

*p < 05**p < 01

$

THINKFUN

THINKSAD

THINKREWARDS

FIG 2 Mean number of minutes of voluntary wait-ing time for each condition in Experiment II

presumably were distracted by thinking "funthings" than when they were thinking aboutthe food rewards or than when they werethinking "sad things "

As Figure 2 indicates, instructions to thinkabout the food rewards and instructions tothink about sad things produced similar delaytimes It is also interesting to compare thedelay times in these two conditions with thedelay time in a similar condition but withoutany ideation instructions—namely, Group 3from Experiment I

To compare these three conditions, thedistribution of waiting times for subjects wasdichotomized at the median A chi-squarecomparison among conditions on this dichot-omy did not approach significance (x2 = 3.2,dj — 2), suggesting that delay times in thethree conditions were essentially similarThus, instructions to think about the rewards,or to think sad, did not significantly facilitatedelay of gratification when compared to ano-ideation condition Instructions to thinkabout fun things, however, produced waitingtimes well above either of the two otherthink groups and also greater than the compa-rable but uninstructed delay-of-gratificationcondition (Group 3 from Experiment I). Inthis regard, note that the "think fun" groupof Experiment II closely replicated the longmean waiting time found in the comparablecondition of Experiment I.

212 W. MISCHEL, E. EBBESEN, AND A. ZEISS

EXPERIMENT IIIExperiments I and II involved distraction

manipulations when the subjects were waitingfor delayed rewards that were always visuallypresent and directly available to attention. Itis conceivable that under conditions in whichdelayed gratifications are not physically avail-able for direct visual attention, their mentalrepresentation in the form of images or ideasdoes have a "time-binding" function andwould facilitate voluntary delay Indeed, thatexpectation is one that might follow fromFreud's (orig. publ. 1911) formulation of pri-mary process and the development of delayingability. Freud implied that delay capacitybegins to develop when the child provideshimself with mental representations or imagesof the delayed object, but only when he can-not see it externally. That is, some dischargeof tension may be achieved by cathecting animage of the reward objects, but only whenthey are physically unavailable for direct at-tention For example, the hungry child mayachieve some gratification by hallucinating themother's breast when it is absent If onetested that hypothesis in the present para-digm, one would have to predict long delaytimes when subjects were instructed to ideateabout the rewards but the rewards werephysically obscured from view On the otherhand, a focus on the frustrativeness of non-reward predicts that such instructions wouldproduce short delay times Experiment IIIwas designed to test which expectation wascorrect.

METHODDesign

The previous experiments were undertaken todetermine the effects of distraction from the rewardobjects on delay of gratification This experimentwas designed to direct the subjects' attention to thereward objects cognitively when the objects them-selves were physically obscured during the waitingperiod One experimental condition and two controlgroups were run In all three groups both the im-mediate and the delayed rewards were obscured fromthe sight of the children during the waiting periodThe groups differed, however, in the instructionsgiven to the subjects In the experimental condition,the subjects were instructed that they could thinkabout the reward objects while waiting In the twocontrol conditions, the subjects were either toldnothing or they were instructed that they could

think about "fun things" while waiting. The de-pendent variable was the length of time that thechildren waited for the delayed reward beforeterminating by settling for the less preferred, butimmediately available, outcome.

SubjectsSixteen subjects were run in this study The sub-

jects ranged m age from 3 years 5 months to 5 years6 months, and their mean age was 4 years 6 monthsThere were 11 males and 5 females Half of thesubjects were assigned to the experimental condition(which directed the children's attention to the rewardobjects) The remaining 8 subjects were equallydivided between the two control conditions ("thinkfun" and no ideation) The sex ratios and the meanages of the subjects were similar across conditionsOne male experimenter tested all of the subjects

ProcedureAll procedural aspects of this experiment were

identical to those in the previous two experimentsexcept for the following modifications In the pre-vious experiments both of the reward objects weredirectly available for attention during the delayperiod In this study it was necessary to removethe reward objects from the child's visual field Ifthe experimenter took the rewards with him whenhe left the child waiting in the room, he might affectthe child's trust that the promised rewards wouldultimately be returned Therefore, the reward objectswere placed under an opaque cake tin and put underthe table at which the child sat so that they couldnot be seen by him during the waiting periodThe children weie told that this operation wouldkeep the food objects fresh while the experimenterwas out of the room The rewards were obscuredm this fashion after the child had passed the compre-hension questions Thereafter, the experimenter wasinformed of the condition in which the subject wasto be run

In the experimental condition the experimentergave the subjects the "think food rewards" instruc-tions used in Experiment II These instructions weredesigned to direct the children's attention to thereward objects during the delay interval In onecontrol condition, the experimenter merely left theroom, in the other control condition, the experi-menter gave the identical "think fun" instructionsthat were used m Experiments I and II Aftercompleting these instructions, the experimenter leftthe room

RESULTS

All of the means are depicted in Figure 3The "no-ideation" group mean was 12 86minutes, and the "think fun" condition meanwas 14 48 minutes It is interesting to notethat these means are very close to the "thinkfun" means found in Experiments I and II,

COGNITIVE AND ATTENTIONAL MECHANISMS 213

which were 12.12 and 13.33 minutes, respec-tively In the experimental condition ("thinkrewards") in which the children were directedto think about the physically obscured rewardobjects, the mean waiting time was only .78minute. These results utterly contradict thebelief that ideating about the reward objectsin their absence enhances voluntary delay ofgratification for them.

Mean delay times in the two control groups("no ideation" and "think fun") obviouslydid not differ from each other, and, conse-quently, they were combined and comparedwith mean waiting time in the experimentalcondition. The resulting t test was highlysignificant (#=11.93, df = 14, p<.00l).Moreover, of the eight children in the experi-mental group, not a single one waited morethan 2 minutes; in contrast, five of the eightcontrol children waited the full IS minutes(p = .05, two-tailed by Fisher's exact test).Thus, even when the rewards are not visuallypresent, ideating about them and attending tothem cognitively serves to substantially de-crease, rather than to enhance, the durationof delay of gratification for the sake ofattaining the preferred reward

DISCUSSION: EXPERIMENTS I, II, AND III

Considering all three experiments together,it seems remarkable how well our briefcognition-inducing instructions seemed towork with our young subjects The presentseemingly simple techniques may provide afruitful methodology for studying cognitionand attention experimentally in young chil-dren. By manipulating cognition-inducing in-structions and visual presence of rewards,extremely powerful effects on delay time wereobtained.

To provide an overview of all the results,the mam findings from the present threeexperiments are summarized in Figure 4. AsFigure 4 shows clearly, effective delay ofgratification depended on cognitive avoidanceor suppression of the reward objects during thewaiting period This conclusion is based onseveral sets of data. First, when the subjectswere waiting for the preferred but delayedreward with the reward objects in their atten-tional field, delay of gratification was minimal

15

UJ

I 10i-13Z.

uj 5S

NOIDEATION

THINKFUN

THINKREWARDS

FIG 3 Mean number of minutes of voluntary wait-ing time for each condition m Experiment III.

In contrast, delay was dramatically facili-tated when the subjects engaged in affectivelypleasant cognitive distractions ("think fun"conditions) during the delay period. Thatresult was replicated in the second experiment

Second, and completely consistent withthese findings, are the results from Experi-ment III, in which the rewards were notexternally available for attention during thedelay period. As Figure 4 shows, under theseconditions the "think fun" group againyielded extremely long periods of delay ofgratification In contrast, cognitions about therewards (induced by the "think rewards"instructions) resulted in an average delaytime of less than 1 minute. Thus, when thechildren thought about the absent rewards,it was as difficult for them to delay gratifica-tion as when the rewards were directly intheir attentional field (the no-ideation condi-tion of Experiment I). Note also that whenthe rewards were not available for directattention, uninstructed subjects (no ideation)found it relatively easy to delay gratification,waiting no less than the "think fun" children.These findings in the two no-ideation condi-tions essentially replicate those from compa-rable conditions in a previous study (Mischel& Ebbesen, 1970), In that study it was foundthat delay of gratification was exceedinglydifficult when the youngsters faced the rewardobjects (either the delayed one, the immediate

214 W MISCHEL, E EBBESEN, AND A ZEISS

EXPERIMENT I

EXPERIMENT 2

EXPERIMENT 3

NOIDEATION

THINKFUN

THINKREWARDS

REWARDS AVAILABLE FOR ATTENTION REWARDS NOT AVAILABLE FOR ATTENTION

FIG 4 Mean number of minutes of voluntary waiting time for treatmentconditions in Experiments I, II, and III, comparing different ideation instruc-tions with controls

one, or both, with no differences betweenthese conditions). However, just as in thepresent Experiment III, when the childrenwaited with no reward objects in their atten-tional field, they were able to substantiallydelay gratification in order to attain thepreferred but delayed reward

The present data may be relevant topsychodynamic theorizing regarding delayingcapacity Predictions from psychodynamictheory concerning the development of delaycapacity are complex since the process is seenas one that involves transitions from primaryprocess to secondary process thinking Ac-cording to Rapaport's (1967) elaboration andclarification of Freudian theory, delay ca-pacity begins with "the emergence of a hal-lucinatory image of the need-satisfying objectwhen tension rises to the point where dis-charge should take place but the need-satisfying object is not present [pp 315-316] " This part of the theory seems tosuggest that instructions designed to helpsubjects imagine or ideate about absent butdesired delayed gratifications should facili-tate voluntary delay time Clearly, the present

data do not support such a view and there-fore might be interpreted as undermining theFreudian position

But Rapaport (1967) goes on to state thatthis image of the need-satisfying object doesnot provide "more than a minute opportunityfor discharge [p. 316]." He therefore arguesthat delay is further developed by transitionfrom primary process hallucinatory imagesto secondary process reality testing. In thisphase of the development of delay capacity,discharge is postponed "until external realityconditions have been found suitable [p318] " Psychodynamic theory (as interpretedby Rapaport) thus suggests that effectivedelay begins to occur when the ego can divertenergy away from images of delayed rewardsand toward reality consideration and instru-mental activity Rapaport noted that internalimpulse control requires "countercathexes" asin repression This part of the psychodynamicformulation of delay seems to imply thatremoving attention from the delayed rewardsmight enhance effective impulse controlRapaport correctly comments that "Little isknown about the nature of the process by

COGNITIVE AND ATTENTIONAL MECHANISMS 215

which these countercathexes arise [p. 219],"Future research needs to explore more closelythe exact conditions that moderate the rela-tions between specific cognitive activity anddelay behavior.

The research thus far demonstrates thateffective delay, rather than being mediatedby consummatory fantasies, probably dependson suppressive and avoidance mechanismsthat reduce frustration This interpretation iscongruent with the "Satan get thee behind

•me" approach to self-control recognized bySkinner (1948) in Walden Two In his novel,Skinner described children learning to controlthemselves by learning to physically andmentally remove temptations This approachemphasizes that one may effectively resisttemptations by engaging in activities (overtlyor covertly) that prevent one from attendingto them Obviously, not just any cognitionserves as an effective distractor from aversivefrustration or temptation As the second ex-periment indicated, "think fun" is betterthan "think sad," whereas "think rewards"increases frustration most, hence makingcontinued delay most difficult

Observations of the children, in thereward-absent delay conditions, lend furthercredence to these interpretations. When thedistress of waiting seemed to become espe-cially acute, children tended to reach for thetermination signal, but in many cases seemedto stop themselves from signaling by abruptlycreating external and internal distractionsfor themselves They made up quiet songs("Oh this is your land in RedwoodCity"), hid their heads in their arms,pounded the floor with their feet, fiddledplayfully and teasingly with the signal bell,verbalized the contingency ("If I stop nowI get , but if I wait I get

)," prayed to the ceiling,and so on. In one dramatically effective self-distraction technique, after obviously experi-encing much agitation, a little girl restedher head, sat limply, relaxed herself, andproceeded to fall sound asleep.

These observations and the results fromthree experiments suggested that the ease withwhich subjects can cope with the frustrationdepends on the overt and covert responsealternatives available to them during the im-

posed delay. The manipulations of the presentexperiments may be construed as having pro-vided subjects with various types of plannedalternative responses for coping with thatfrustration. The more the available responsedirected the child's attention away from thefrustration, the better he was able to con-tinue the delay and substitute a new adaptiveactivity during the frustration, in the presentparadigm that activity could have been any-thing that kept him waiting without ideatingabout the goal objects.

It would be interesting to know if pre-school subjects such as those in the presentexperiments were aware of the principle of"Satan, get thee behind me" before theyactually began waiting. That is, given a choiceof waiting for the preferred reward, with therewards obscured or available for attention,would subjects make the right choice? Toanswer this question, 29 preschool childrenwere administered the previously describedinstructions for the delay-of-gratificationparadigm. Just before the point at which theexperimenter would usually leave the childwaiting alone in the room, she picked up anopaque cake cover and gave the subject thechoice of covering with it either the rewards,or another set of comparable but irrelevantobjects also lying on the table, or simplyplacing the cover over another fixed spot onthe table. The results showed that the chil-dren chose the place to cover quite randomlybefore commencing their voluntary delayThat is, they did not seem aware that ob-scuring the relevant rewards would facilitatetheir ability to wait for them. Thus, youngchildren do not seem to have insight into therole of cognition and attention in self-control,at least prior to actually waiting

The data from the main experiments seemto contradict James's (1890) belief that "theessential achievement of the will" requiresone to bear up and force oneself to maintaindirected attention to the difficult or boringRather than trying to maintain aversiveactivities such as delay of reward through"acts of will" and focused attention, effectiveself-control may hinge on transforming thedifficult into the easy, the aversive into thepleasant, the boring into the interesting, whilestill maintaining the task-required (reward-

216 W. MISCHEL, E. EBBESEN, AND A. ZEISS

contingent) activity. Such transformationsmay occur either by engaging in the appropri-ate overt distracting activity (e g., the "toy"condition) or changing one's own mental con-tent and ideation so that it functions as acovert distractor.

It is important to recognize, however, thatthe mental transformations and distractionswhich occur during delay do not erase orundo the role of the reward contingencies inthe waiting situation This is evident in thedata from Experiment I, which show howlittle persistence there was in "thinking fun"or playing with a toy when there was nowaiting contingency. The distracting activityitself, while pleasant and distracting enoughto maintain waiting for a contingent reward,did not in itself keep the children in the roomfor more than a minute Additional evidencethat the contingency was available mentallythroughout the waiting period is that thechildren easily reproduced, verbally or byappropriate action, the contingency at the endof the waiting period Children who had beenbusily distracting themselves for the full ISminutes, playing with a toy or singing songs,immediately and spontaneously ate the appro-priate food reward when the experimenterreturned Obviously then, the transformationof the aversive waiting into a pleasant playperiod does not efface the task-oriented pur-pose of the behavior, and presumably thetwo processes somehow coexist Subjectswere guided by their goals, even when seem-ingly absorbed in distractions designed toobscure them Just how the contingency wasoperating is an interesting point for specula-tion The contingency may have been avail-able but never reproduced mentally until theend of waiting, even more likely, subjectsmay have reminded themselves of the con-tingency episodically throughout the waitingperiod As mentioned previously, verbaliza-tions of the contingency often occurred whenthe subjects momentarily left their distractingplay and seemed about to terminate the wait-ing period It is as if the subject periodicallyreminds himself of the goal for which he iswaiting, distracts himself from it to makedelay less frustrative, and then repeats theprocess

Extending these speculations further, a

good way to master the difficult or aversivemay be to think or do something that ispleasant, while still performing the necessarytask-relevant response (e.g., waiting, work-ing) Rather than "willing" oneself to heroicbravery, one needs to perform the necessary"difficult" response while engaging in anotherone cognitively The principle involved hereseems similar to the one underlying "counter-conditioning" of aversive emotional reactionsin behavior therapy To master a snakephobia, for example, the subject needs to •deal with the problematic stimulus whileengaged in a fear-incompatible positive inter-nal response (eg, relaxation); whereas indelay of gratification one must perform adifficult or problematic response while en-gaged in ideation of stimuli that are positiveor distracting. In either case, it is easier todo something difficult if one also does some-thing easy or pleasant at the same time.

The findings from the present studies seemextremely reliable, being based on severalreplications and diverse convergent data.However, one obviously cannot generalizefrom them to the role of cognition in formsof self-control other than the delay-of-gratification paradigm For example, it mightbe adaptive to ideate about desired or neededbut currently unavailable goal objects, butonly in situations m which the subject'sactions can be potentially instrumental inproducing the desired outcome Thus, whenattainment of a positive outcome is contin-gent on the subject's own problem-solvingbehavior, it might help him to think aboutthe goal object while seeking means forachieving or reaching it in reality. In contrast,in the present delay-of-gratification paradigm,attainment of the preferred goal required onlypassive waiting beyond delaying there wasabsolutely nothing the subject could do toinfluence the occurrence of the desired out-come Moreover, even his delay behavior(while a necessary condition for attainmentof the preferred outcome) could in no wayinfluence the time at which gratification wouldultimately occur

Data relevant to the dilemma, in whichsubjects cannot do anything to attain a de-sired but unavailable outcome, may be foundin studies of "defensive perception." Our

COGNITIVE AND ATTENTIONAL MECHANISMS 217

studies on attention in delay employ a design,which, in a way, seems the reverse of theseclassic "new look" studies. In the latter thesubject was, first, frustrated or pained andwas then asked what he thought or perceivedmost readily (on projective measures). Incontrast, in the present methodology, the sub-ject is given the "thoughts," and then theireffect on what he does is noted; specifically,his ability to sustain goal-directed delaybehavior and to cope with frustration aremeasured. Consistent with our data, findingsfrom perceptual defense studies indicate atendency to avoid painful stimuli cognitivelyand perceptually when nothing can be doneby the subject to cope with them instrumen-tally (e.g., Reece, 1954). It also has beenreported that in response to projective ma-terial, sleep-deprived subjects showed fewersleep-related ideas and themes than did con-trols (Murray, 1959) Likewise, food ideationis less when subjects are severely food de-prived than when they are not hungry(Lazarus, Yousem, & Arenberg, 1953) Simi-larly, Clark's (1952) pioneering experimentson Thematic Apperception Test (TAT)sexual imagery compared the amount ofsexual responses in TAT stories written bysexually aroused and nonaroused males. Hefound less sexual imagery and also less "sexualguilt" in the stories of sexually aroused sub-jects. Clark attributed his results to a simul-taneous, predominating increase in sexualguilt which he suggests was evoked by thesexual arousal and is "sufficient to inhibit theexpression of sex with a consequent loweringof guilt [p. 398] " But a more parsimoniousinterpretation in terms of the present experi-ments is that the Clark results could reflectcognitive avoidance of all sexual thoughts(including those scored as "guilt") underconditions that made sexual thoughts highlyfrustrative, that is, sexual arousal with noopportunity for satisfaction.

Thus, when subjects cannot cope with aver-sive stimuli instrumentally (eg , by goingto sleep when sleepy, by finding food whenhungry, by avoiding painful shock, and byobtaining sexual release when aroused, theymay engage in cognitive avoidance of thosestimuli. To decrease frustration, subjects maygenerate their own distractors and avoid the

aversive stimuli cognitively, if possible, as weobserved repeatedly in our studies. This con-clusion—that aversive stimuli are avoidedcognitively—may be restricted, however, toparadigms in which the subject believes thatthinking about the aversive stimulus cannotchange the contingencies in the situation.

The overall results of the presentexperiments help to clarify some widespreadbasic theoretical misconceptions regardingself-control. In particular, following dynamicformulations, it has been customary to con-strue voluntary delay of reward as involvingthe ability to defer immediate gratification.This ability has been viewed as an enduringtrait of "ego strength" on which individualsdiffered stably and consistently in manysituations In fact, as the present data indi-cate, under appropriate motivational andattentional-cognitive conditions, virtually allsubjects, even young children, could manageto delay for lengthy time periods

Taken collectively, research on delay ofgratification permits us now to speculateabout a two-part process in delay of gratifica-tion. First, one must consider the determi-nants of the choice to delay for the sake ofmore preferred delayed outcomes. This choiceis influenced mainly by the subject's expecta-tions concerning the probable consequencesof his choice These consequences include therelative subjective values of the immediateand delayed outcomes themselves as well asother probable reinforcing outcomes associ-ated with each alternative. As previous re-search has shown, expectancies relevant tothese outcomes depend on the subject's directand vicarious past experiences and trust rela-tionships, modeling cues, the specific contin-gencies in the choice, and so on (eg , Mischel,1966).

Second, once the choice to delay gratifica-tion has been made, effective delay dependson cognitive and overt self-distractions toreduce the aversiveness of the self-imposedfrustration. For this purpose, the subjectneeds to "tune out" on the goal objectsand generate his own distractions whilemaintaining the contingent behavior for goalattainment.

Thus, the subject can wait most stoicallyif he expects that he really will get the

218 W MISCHEL, E EBBESEN, AND A. ZEISS

deferred larger outcome in the waiting para-digm, and wants it very much, but shifts hisattention elsewhere and occupies himself in-ternally with cognitive distractions Any con-ditions that shift attention from the delayedobjects appear to facilitate voluntary waitingtimes appreciably In order to bridge thedelay effectively, it is as if the subject mustmake an internal notation of what he is wait-ing for, perhaps remind himself of it periodi-cally, but he must spend the remaining timeattending to other less frustrative internaland external stimuli

REFERENCES

AMSEL, A The role of frustrative nonreward innoncontmuous reward situations PsychologicalBulletin, 1958, 55, 102-119

AMSEL, A Frustrative nonreward in partial reinforce-ment and discrimination learning PsychologicalReview, 1962, 69, 306-328

CLARK, R A The projective measurement of experi-mentally induced levels of sexual motivationJournal of Experimental Psychology, 1952, 44,391-399

FREUD, S Formulations regarding the two principlesin mental functioning 1911 In Collected PapersVol 4 New York Basic Books, 1959

GRIM, P F, KOHLBERG, L, & WHITE, S H Somerelationships between conscience and attentional

processes Journal of Personality and Social Psy-chology, 1968, 8, 239-252

HARTSHORNE, H , & MAY, M A Studies in the natureof character Vol 1 Studies in deceit New YorkMacmillan, 1928

JAMES, W Principles of psychology New YorkHolt, Rmehart & Winston, 1890

LAZARUS, R S , YOUSEM, H , & ARENBERG, D Hungerand perception Journal of Personality, 1953, 21,312-328

HANDLER, G The interruption of behavior NebraskaSymposium on Motivation, 1964, 12, 163-219

MISCHEL, W Theory and research on the antecedentsof self-imposed delay of reward In B A Maher(Ed), Progress in experimental personality re-search Vol 3 New York Academic Press, 1966

MISCHEL, W, & EBBESEN, E B Attention in delayof gratification Journal of Personality and SocialPsychology, 1970, 16, 329-337

MURRAY, E J Conflict and repression during sleepdeprivation Journal of Abnormal and SocialPsychology, 1959, 59, 95-101

RAPAPORT, D On the psychoanalytic theory of think-ing In M M Gill (Ed), The collected papers ofDavid Rapaport New York Basic Books, 1967

REECE, M M The effect of shock on recognitionthresholds Journal of Abnormal and Social Psy-chology, 1954, 49, 165-172

SKINNER, B F Walden two New York Macmillan,1948

WINER, B J Statistical principles in experimentaldesign New York McGraw-Hill, 1962

(Received November 3, 1970)