+ All Categories
Home > Documents > Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing...

Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing...

Date post: 13-Sep-2016
Category:
Upload: paul-davidson
View: 216 times
Download: 0 times
Share this document with a friend
13
BEHAVIOR THERAPY 9, 222--234 (1978) Intrinsic Interest and Extrinsic Reward The Effects of a Continuing Token Program on Continuing Nonconstrained Preference PAUL DAVIDSON AND BRADLEY BUCHER The University of Western Ontario It has been hypothesized that token reinforcement programs for children may have detrimental effects on their "intrinsic interest" in the reinforced activities. Studies finding support for this hypothesis have typically used a single short interval of rewarded activity after which a single reward has been delivered. The effects of the intervention have then been assessed in one or two test sessions. The present study assessed the effects of a continuing token reinforcement pro- gram in repeated test sessions. Three preschoolers were given equal numbers of trials on two game-like teaching machines. There were four successive training conditions: two reward conditions alternating with two baseline conditions. Chil- dren earned tokens by responding on one machine in the first reward condition and on the second machine in the second reward condition. No contingent reward was given in either of the two baseline conditions. A test session preceded each training session, in which unconstrained and nonreinforced choices of the machines were permitted. In the test sessions associated with the first reward condition, all subjects chose the reinforced machine more often than the other. They all showed an increased number of choices and increased rate of choices on that machine, as compared to the test sessions in the preceding baseline. A similar preference for the second machine was shown when that machine was associated with reward. The results show no evidence of decreased "'intrinsic interest" in the rewarded activity as measured by unconstrained choices between the activities. Implications of these and other results for token reinforcement programs are discussed. Some psychologists have recently asserted that token economies and other reinforcement systems may adversely affect a child's 'qntrinsic interest" in the reinforced activities, if the child undertakes the activity to obtain an extrinsic reward. This phenomenon has been labelled the over- justification effect (Lepper, Greene, & Nisbett, 1973). The term "extrin- This research was funded by a grant from the Ontario Mental Health Foundation, No. 503-74B. Requests for reprints should be sent to Paul Davidson, Department of Research, Regional Psychiatric Centre, Box 22, Kingston, Ontario, K7L 4U7 Canada. 222 0005-7894/78/0092-0222501.00/0 Copyright © 1978 by Association for Advancement of Behavior Therapy. All rights of reproduction in any form reserved.
Transcript
Page 1: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

BEHAVIOR THERAPY 9, 222--234 (1978)

Intrinsic Interest and Extrinsic Reward The Effects of a Continuing Token Program on Continuing Nonconstrained Preference

PAUL DAVIDSON AND BRADLEY BUCHER

The University of Western Ontario

It has been hypothesized that token reinforcement programs for children may have detrimental effects on their "intrinsic interest" in the reinforced activities. Studies finding support for this hypothesis have typically used a single short interval of rewarded activity after which a single reward has been delivered. The effects of the intervention have then been assessed in one or two test sessions. The present study assessed the effects of a continuing token reinforcement pro- gram in repeated test sessions. Three preschoolers were given equal numbers of trials on two game-like teaching machines. There were four successive training conditions: two reward conditions alternating with two baseline conditions. Chil- dren earned tokens by responding on one machine in the first reward condition and on the second machine in the second reward condition. No contingent reward was given in either of the two baseline conditions. A test session preceded each training session, in which unconstrained and nonreinforced choices of the machines were permitted. In the test sessions associated with the first reward condition, all subjects chose the reinforced machine more often than the other. They all showed an increased number of choices and increased rate of choices on that machine, as compared to the test sessions in the preceding baseline. A similar preference for the second machine was shown when that machine was associated with reward. The results show no evidence of decreased "'intrinsic interest" in the rewarded activity as measured by unconstrained choices between the activities. Implications of these and other results for token reinforcement programs are discussed.

Some psychologists have recently asserted that token economies and other reinforcement systems may adversely affect a child's 'qntrinsic interest" in the reinforced activities, if the child undertakes the activity to obtain an extrinsic reward. This phenomenon has been labelled the over- justification effect (Lepper, Greene, & Nisbett, 1973). The term "extrin-

This research was funded by a grant from the Ontario Mental Health Foundation, No. 503-74B. Requests for reprints should be sent to Paul Davidson, Department of Research, Regional Psychiatric Centre, Box 22, Kingston, Ontario, K7L 4U7 Canada.

222 0005-7894/78/0092-0222501.00/0 Copyright © 1978 by Association for Advancement of Behavior Therapy. All rights of reproduction in any form reserved.

Page 2: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

INTRINSIC INTEREST AND EXTRINSIC REWARD 223

sic" and its counterpart "intrinsic" are defined in the theory in terms of the perceptions of the individual (DeCharms, 1968; Greene & Lepper, Note 1), rather than in terms of objective features of the activities or their consequences (Calder & Staw, 1975). The terms are relative, rather than absolute. Whether an event is intrinsic or extrinsic to a task must depend on the definition of the task.

In a typical study of this phenomenon, Lepper et al. (1973) compared three groups of children who were given a session in which they were asked to draw with felt marker pens. Members of one group were prom- ised and given a certificate as a reward, those in a second group were given the certificate with no prior promise, and subjects in a third group received neither promise nor certificate. Some weeks later, the pens were made available in the children's normal preschool environment, and the children were observed unobtrusively. The group promised the certificate in the prior session played less with the pens. Greene (1975) lists other studies showing similar effects.

On the basis of this body of literature and theoretical statements de- rived from it, Levine and Fasnacht (1974) have cautioned against the "promiscuous" use of token reinforcement and offer advice to practition- ers and teachers on ways to avoid the use of reinforcement. In the same vein, Greene and Lepper (1974, Note 1) have asserted that the use of reinforcement can "turn play into work." Greene (1974) suggests that one of the reasons for the lack of generalization from token systems to un- structured environments (Kazdin & Bootzin, 1972) may be that generali- zation is counteracted by the decrease in intrinsic interest.

According to the proponents of one major variant of this viewpoint (Greene & Lepper, Note 1), several experimental controls are needed to demonstrate the effect of perceived extrinsic rewards on perceived intrin- sic interest. The rewards must be expected and salient. Controls should be exercised for effects of difference in practice, either on preferences among the tasks or in increasing skills that may make new reinforcers available. Initial intrinsic interest should not be negligible. The activity should be one that the subject will sometimes choose when external controls are not present and other activities are also available. The test for the consequences of the manipulation of intrinsic interest should be con- ducted under conditions free of monitoring or coercion or of expected extrinsic incentives that might affect the subject's choice.

Many of the above limiting conditions lack objective specification, so that their translation into experimental operations is unsure. This difficulty complicates clear experimental testing of the theory. Studies cited in support of the theory (Greene, 1975) have used a variety of procedures, not all of which meet all procedural requirements cited above. Thus, the conditions necessary and sufficient for detrimental ef-

Page 3: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

224 DAVIDSON AND BUCHER

fec t s o f e x t r i n s i c r e i n f o r c e m e n t to o p e r a t e a r e no t ye t k n o w n . F u r t h e r , t he

r e l e v a n c e o f t he d a t a s u p p o r t i n g this t h e o r e t i c a l d e v e l o p m e n t to t o k e n

s y s t e m s is v i t i a t e d by the fac t tha t m o s t s u p p o r t i n g r e su l t s h a v e b e e n

o b t a i n e d u n d e r v e r y l imi ted c o n d i t i o n s : in a s ingle t e s t s e s s i o n and a f t e r a

s ingle r e w a r d se s s ion wi th a single r e w a r d w h o s e s ta tus as a r e i n fo r ce r was

no t t e s t ed . D e m o n s t r a t i o n o f t he r e l e v a n c e o f t he p r e d i c t i o n s o f t he t h e o r y

fo r a c o n t i n u i n g t o k e n s y s t e m w o u l d r e q u i r e a d e m o n s t r a t i o n o f t he e f f ec t

u n d e r c o n d i t i o n s m o r e c l o s e l y r e s e m b l i n g t h o s e in e f fec t in an o n g o i n g

t o k e n s y s t e m . T h e p r e s e n t s t udy was d e s i g n e d to p r o v i d e such a t es t .

METHOD

Subjects Four cooperative and available children were selected from a day care center. Subjects 1

and 2 were boys, whose ages were 5 years 6 months and 4 years 4 months, respectively, at the beginning of the study. Subjects 3 and 4 were girls, both aged 4 years 10 months. Subject 4 refused to come to sessions after the third day, so no usable data were obtained.

Apparatus and Setting The apparatus consisted of two child-operated "teaching" machines designed for this

study and a marble dropping apparatus described by Gewirtz and B aer (1958). One teaching machine (Clown Task) operated as follows. The child selected a large red card displaying drawings inside three squares and placed it in a slide in the apparatus. The drawings presented a numerical matching-to-sample problem: a sample and two comparison stimuli (e.g., the numeral 3 in the sample square, three butterflies in one comparison square, and two butterflies in the other). When placed in the apparatus the card coded a switch that designated which of two keys was correct. The keys were adjacent to the card and were illuminated when the card was positioned correctly. The task was to press one of the keys. An error caused the keys to be darkened, ending the trial. A correct response led to the flashing of lights in the face of a clown (dubbed Ronald McDonald) painted on the apparatus, the ringing of a bell, and the playing of taped ragtime piano music for 2 sec, followed by the end of trial. If the two keys were pressed within 0.5 sec of one another, the trial was treated as an error,

The House Task teaching machine (Sesame Street) was similar to the Clown Task. The subject selected a small yellow card containing three stimuli and placed it in a multicolored machine, shaped like a house. The cards contained alphabet problems (e.g., letter C with the comparisons of a cow and a bird picture). Feedback for correct responses consisted of the ringing of a bell and the illumination of lights in various windows in the house. The windows lit in a haphazard order, allowing subjects to see pictures of characters from Sesame Street. After a response the subject removed the card and placed it in a box. This response activated the machine for a new trial.

The marble drop machine assessed reinforcer effectiveness. It consisted of a box with two holes. Marbles dropped in either hole were returned to the subject. In any session, one hole was designated S+, and when the subject dropped a marble into that hole, he was given a token by the experimenter.

Procedure Sessions were divided into two parts: a testing session, followed by a training session. In

the testing sessions subjects played with the teaching machines by themselves, while the

Page 4: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

INTRINSIC INTEREST AND EXTRINSIC REWARD 225

training sessions were for the purpose of experimenter intervention. Both were carried out in one room of a mobile laboratory, equipped with a through-the-wall observation system to allow for unobtrusive observation. The room contained a long table for the two machines, separated by the common receptacle for used cards. Two unordered piles of unused cards were located on a chair at the other side of the room.

Training sessions were given in four phases. In the Baseline phase subjects were individu- ally trained to use the teaching machines. In the Reinforce Clown phase a token was delivered contingent upon a response to the Clown Task during training. In the third training phase no rewards were given for the machines, but reinforcer effectiveness was assessed using the Marble Drop Machine. In the final phase (Reinforce House) the House Task was reinforced during training. In all training sessions subjects were required to make an equal number of responses to both teaching machines. Subjects were under no such constraints in testing.

Testing. For all four phases, just prior to every training session (except Session I in Baseline 1), the child's free play choice of the machines was assessed. After the child was brought to the trailer, the experimenter said: "Today I want you to play by yourself with the machines for a few minutes while I make the orange juice. After that we can play together." The experimenter retired to an adjoining room, locked the door, watched surreptitiously for 5 min, and then returned. The subject was asked if he had played with the machines. If he reported that he had not, he was asked why not and was encouraged to do so the following day. If he reported playing with only one machine, he was informed ' 'You can play with both machines if you want to, you know." A report of playing with both machines was reinforced by praise from the experimenter. The experimenter did not inquire which machine or machines the child had used, and no comments were made about the subjects' performance other than those above. Subjects always aoswered the questions truthfully.

No rewards were given, or present, during testing. After testing, the rewards were brought in by the experimenter, who solicited the subject's assistance in this activity. Then the training session followed.

Pretraining. One pretraining session was given to check that the subject could identify and match the numbers, letters, and pictures to be used in the experiment and to train the subject to relate tokens to rewards. The subject learned to fill circles on a card with tokens to receive a reward of his choice. The machines were not present.

Baseline training. In each session, the child was given the pretraining task. Correct answers earned tokens. Twelve tokens were delivered and exchanged. The subject was then taken to the machines. In the first training session the experimenter described the salient features of the machines and cards, demonstrated their use, and then asked the subject to play with them. The experimenter gave the subject a card and then gave another card each time one was used in the machine. Each machine was operated 12 times in each session, in blocks of six cards on each machine, The starting machine was determined by a coin flip. In the first session (only) prompts were given (both verbal and physical) to aid in the operation of the apparatus. After each correct response the experimenter said "very good," and '~oops, that's a mistake" after an error. Nothing else was said for at least 5 sec after the response. After that time, brief responses would be given to questions, if any, or the next card would be handed out. After 24 trials, the child was returned to the day school. In sessions where no tokens were given for teaching machine performance, the rewards were not present in the session room until the end of the session, when tokens earned earlier could be exchanged.

Reinforce Clown. The pretraining task was dropped. The 24 machine trials were given as before, in blocks of six. Tokens were delivered contingent on playing with the Clown. To equalize token reinforcement, and to ensure that the tokens did not add feedback about the subjects" skill, tokens were given for either choice response. At the beginning of sessions subjects were asked if they would like to play with Ronald McDonald for tokens that day.

Page 5: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

226 DAVIDSON AND BUCHER

They were told that they would get no tokens for Sesame Street. When they said that they would (no subject ever refused) they were asked which machine they were playing with for tokens, and a correct response was elicited. Token delivery was always accompanied by the experimenter saying "here is a token.'" (After a correct response to the reinforced apparatus he said: "Very good, here is a token." After an error he said: "Oops, that's a mistake, here's a token.") Experimenter comment on performance was given as in Baseline for trials at the unreinforced House task. The experimenter made no other comments for at least 5 sec after the child's response. Tokens were traded at the end of each session. Each day each subject was asked, at the end of the session, which machine was his favorite (if he didn't respond he was asked which of the machines he would prefer to take home with him) and which machine he was getting tokens for. For both questions, when the subject named a machine, his response was met with a noncommittal "Thank you."

Baseline 2. Contingencies for machine use returned to those of the Baseline. However, instead of gaining tokens for the simple identification task, subjects were given tokens for playing with the marble drop machine, to give an assessment of the tokens' effectiveness as reinforcers.

The marble drop machine was a box with two holes in it, into which subjects dropped marbles. One hole was selected as S+ by a coin flip each session, and the subject received a token for a marble dropped into that hole. Subjects were instructed: "'Today we are going to earn tokens by playing with a new machine. When I say so, drop a marble into one of the holes. Sometimes I will give you a token." Each trial was then cued: "Drop a marble into one of the holes." If the hole selected was S+ the experimenter said: "Here is a token." If subjects selected the other hole, the experimenter made no comment and no token was given. This procedure continued until after 12 tokens had been delivered.

Reinforce House. This condition mirrored the previous reinforcement condition. Tokens were provided for playing with the House instead of the Clown, Use of the marble drop machine was discontinued. In the final session, after testing, reinforcer effectiveness was again measured, and no other training was done.

RESULTS

Test Sess ions

Responses were r ecorded by the e x p e r i m e n t e r in the adjacent room and

by counte rs a t t ached to each machine . The two sets of data were invari-

ably in agreement . Tes t sess ions are indexed accord ing to the training

session that p receded them, that is, in the p rev ious day ' s session.

The f r e q u e n c y of choice and the p ropor t ion o f cho ices of the C lown or

H o u s e act ivi ty will be used to measure in teres t , as de te rmined f rom the

test sess ions in each phase. Table 1 shows the resul ts on which these

ca lcula t ions are based.

In the test sess ions for the Reinforce C lown phase (1) all subjects chose

Clown more of ten than H o u s e , (2) all subjects made more choices of the

Clown task than they had done during the p reced ing Basel ine (per ses-

sion), and (3) the p ropor t ion of choices o f the C lown act iv i ty increased , as

c o m p a r e d to the p ropor t ion of C lown cho ices in the p reced ing Basel ine . These resul ts run coun te r to predic t ions o f the over jus t i f ica t ion

hypothes is .

Page 6: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

I N T R I N S I C I N T E R E S T A N D EXTRINSIC R E W A R D 2 2 7

15

(3-

9~

E Z

12

o

15

12

15

12

BASELIN E 1

I I

" " ~ / r

i ; I I t I • I I

i s j I I I

A / /', ,;

/\ e I I i I |

REINFORCE CLOWN

t ,

i t - ;

• L •

I l

i

i I o~

~'~'~.__.

s "*~

/

i J | • I

BASELINE 2

I i

w

/

/

REINFORCE HOUSE

51

t ~ •• / •

/ •

/

\ \ /

$2

Sessions

$3

FIG. I. Number of responses by each subject for each task during each test session.

- - H O U S E

I I i i

. . . . . . . . CLOWN

Page 7: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

228

BASELINE 1

DAVIDSON AND BUCHER

REINFORCE CLOWN BASELINE 2 REINFORCE HOUSE

11

1o !

9

i / )

~ 8 c-

O

o. 7 t/'l @

n- 6

o 5 l-,-

c '

~,4

3

J

I I I I I i i I I I

X=7-7

l m i

~ 7 tel C"

~e

n- 5 "6

~4 E ~ 3 z

C

~ 2 q)

1

0

f ~=2-3 /

/ /

/ 4

/A~x

= 7. 2 / / x

J J

f I

/

I I

,/

I I I I I

FIG. 2.

S e s s i o n s

T ', /

%1 V

I I I

, X = 4 . 7

XVX 4 4

t I ¢ i

Clown House

Average number of responses for each task for each test session and average

Page 8: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

INTRINSIC INT E R E S T A N D EXTRINSIC R E W A R D 229

T A B L E I AVERAGE RESPONSES TO EACH ACTIVITY IN TEST SESSIONS

FOR EACH CONDITION, FOR EACH SUBJECT

Baseline Reinforce Baseline Reinforce 1 Clown 2 House

Subject I: Clown 3.4 6.6 3.7 3.25 House 2.4 1.4 1.3 7.5

Subject 2: Clown 3.1 7.8 5.33 7,0 House 2.0 0.6 1.67 2,5

Subject 3: Clown 0.2 7.0 6.67 3,75 House 6.8 5.0 4.33 4.25

In test sessions for the Reinforce House phase (1) two of the three subjects (S~ and $3) chose House more than Clown, (2) two of the three subjects (SI and $2) made more choices of House than they had during the preceding Baseline (per session), and (3) all subjects showed increased proportion of choices of House as compared to the proportion in the preceding Baseline. The preponderance of these results runs counter to predictions of the overjustification hypothesis.

Comparison of test sessions in the two Baseline conditions gives another measure of the persisting effect of the Reinforce Clown phase. The (per session) number of choices of Clown, in test sessions for Baseline 2, is greater than for the number for Baseline 1, for all subjects. The proportion of choices of Clown is greater in Baseline 2 than in Baseline 1 for all subjects.

Training Sessions

Accuracy and speed of responding were examined to test for possible differential effects of reinforcement on the skill with which subjects per- formed the two activities. Reinforcers were delivered for responding, independent of accuracy, and the same number were given in each rein- forced session, independent' of the child's speed of responding. Accuracy exceeded 85% in every session. Most errors were for the Clown task. Errors for House were rare. Subject 1 made one and Subject 2 made four, scattered over the first three Baseline sessions. Subject 3 made none. Thus, reinforcement showed no differential effect on errors. Speed of response might also reflect the subject's skill. Response time was calcu- lated as the interval between the subject's initial contact with a card and the time the subject pressed a button on the machine. If reinforcement for one task had a differential effect on response speed, the difference be- tween mean response times for the two tasks might differ in the Reinforce House and the Reinforce Clown phases. This hypothesis was tested for

Page 9: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

230 DAVIDSON AND BUCHER

each subject. The difference in mean response times for each activity was calculated for each training session. The differences for the Reinforce Clown and Reinforce House conditions were compared, for each subject, using a t test for independent means. The sets did not differ significantly for any of the three subjects. Thus, no differential effect of token delivery on response speed is evident. This test assumes independent errors for difference scores from different sessions, which may not be justified. If the degrees of freedom were reduced to take account of possible depen- dencies, the p values would be even less nearly significant. The apparent absence of an effect here may be related to the fact that performance speed on either task could affect the rate at which tokens were delivered, though only slightly, and could not affect the total tokens given.

Self-Report Data

Subjects were asked, at the end of sessions, to indicate which machine they liked. If the subject made no response, or appeared not to under- stand, this was expanded: " I f you could take one of the machines home with you, which one would you take.'?" On 75% of the reinforced sessions, subjects chose the machine currently being reinforced in training ses- sions. In Baseline, subjects typically named various performance charac- teristics (e.g., color, bell, music, etc.) as the reasons for their preferences. However , Subject I attributed his preference to the fact that he got tokens for it, in three of nine reinforced sessions, and Subject 2 did so in six of nine reinforced sessions. Subject 3 never verbalized that his preference was due to tokens. After the first training session of Reinforce Clown, subjects were also asked which machine they were earning tokens for. Subjects made no errors in identifying which machine was associated with tokens.

Reinforcer Effectiveness

The index of effectiveness used was the percentage of responses made to the reinforced hole (S+) after the first reinforced response. The mean of this value across all sessions for all subjects was 77% (range 70-90% for the three subjects).

DISCUSSION

Greene and Lepper (Note 1) state that intrinsic interest can be de- creased if a child engages in an activity (a) of initial intrinsic interest and (b) under conditions which make salient the instrumentality of that en- gagement toward the attainment of some extrinsic goal. In the studies cited by Greene and Lepper (Note 1) and Greene (1975), on which much of the evidence for the extrinsic reward effect rests, the conditions under which rewards are given and effects are tested are quite unlike those that

Page 10: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

INTRINSIC INTEREST AND EXTRINSIC REWARD 231

operate in a token reinforcement program, in two major respects. First, a single session with one or a few rewards is given to create the extrinsic reinforcement effect. This contrasts with a token reward system where repeated rewards are given. Second, the test of the extrinsic reinforce- ment effect is made typically in one or perhaps two sessions. If the extrinsic reward effect is to be of much interest, a continuing postreward effect should be present. Two studies cited in Greene and Lepper (Note l) and Greene (1975) meet these two conditions. Greene (1974) and Feingold and Mahoney (1975) studied the effects of a continuing token reinforce- ment program. Neither of these studies controlled for the effects of differential exposure to the reinforced activity. Such a control is needed to ensure that experiences with the activity during the experimental manipulation do not in themselves bring about changes in the relative preference or tendency to engage in that activity, through changes in skill, familiarity, etc.

The present study represents an effort to measure the effects of a continuing extrinsic reward system on concurrent choices among activi- ties, where extraneous variables affecting choices are controlled, and the effects of extrinsic rewards on intrinsic interest should be observable. The overjustification hypothesis would predict that children reinforced for playing with the House or Clown activity in training sessions would show decreased choices of that reinforced activity in concurrent and sub- sequent test sessions. Most of the results show the opposite effect.

An opposing hypothesis is that reinforcement for an activity in one situation would show some generalization to another situation, perhaps due to secondary reinforcement, even when the difference in reinforce- ment conditions could be readily discriminated. This hypothesis would predict that the proportion of choices of the reinforced activity would increase in the nonreinforced test sessions associated with the reinforce- ment phases as compared to those in the preceding baselines. This hypothesis generates several individual subject predictions for compari- sons between phases: comparing choices in Reinforce Clown phase to those in Baseline l, comparing Baseline 2 and Baseline l, and comparing Reinforce House and Baseline 2. All nine individual subject comparisons show proportions of choices in the direction predicted by this hypothesis. However, since some scores are used more than once, these predictions are not independent. The hypothesis that "intrinsic interest" will be decreased by the reinforcement procedures followed here obtains no support from these results.

Previous research has produced a number of procedural safeguards that should be present in an adequate test of the overjustification effect (Greene & Lepper, Note 1). These safeguards are not typically present in actual token reinforcement systems. "~Intrinsic interest" in the reinforced

Page 11: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

232 DAVIDSON AND BUCHER

activities in reinforcement programs may often be very low, and effects on interest may be masked by other changes: in skill, expectation of reward, the nature of the reinforced activities, etc. Thus, a test of the theory for a reinforcement system must be conducted under conditions where the overjustification effect has adequate opportunity to occur and to be observed.

Rewards must be perceived by the children as extrinsic or intrinsic in accordance with the experimenter's labeling. This assurance is typically obtained by varying the degree to which reward occurs through socially mediated actions. Socially mediated rewards must be expected and sa- lient. This is typically assured through instructions to the subject and by providing other clear cues for the announced reward. The present exper- iment followed these conventions. In sessions with reinforced perfor- mance, token availability was announced and the child himself brought the box of rewards into the experimental room. Tokens were given to the child repeatedly throughout the rewarded activity.

A further requirement is that the tasks should have some intrinsic reinforcement properties. This condition is infrequently tested; instead, tasks are chosen that children generally choose in unconstrained settings. In the present study the two experimental activities involved novel de- vices designed to appeal to preschool children, but their appeal was not directly assessed. Suggestive evidence that the activities were attractive come from the frequency with which they were chosen in test sessions. Frequency of choice rose from the first to the second Baseline condition and was high in both reinforcement conditions, but with a very slight decline in the second. However, use of the machines was encouraged by the experimenter's request for the subject to play with the machines.

In a token system rewards are typically used as reinforcers, that is, to increase the rate of some specified behavior. It has not been established whether the effects of rewards on intrinsic interest depend on this charac- teristic, since the relevant studies have rarely tested whether the extrinsic rewards used actually showed a reinforcing effect. In the present study the effect of tokens as reinforcers was demonstrated in the marble drop task. Their effect as a reinforcer for the House and Clown games was not tested. Neither speed nor accuracy of responding affected the number of tokens awarded; the same number were given in each reinforced session.

If intrinsic interest is to be assessed by some measure of rate or choice, extraneous influences must be controlled. One frequently noted potential source of extraneous influence is the child's perceived expectation of social coercion or reward for responding in the test. Lepper and Greene (1975) recommend testing in a familiar environment, different from that used in training, with no experimenters visible. This recommendation does not appear to be crucial. Ross, Karniol, and Rothstein (1976) found

Page 12: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

INTRINSIC INTEREST AND EXTRINSIC REWARD 233

support for the overjustification hypothesis when test sessions were con- ducted in the training room with the experimenter present. In the present study, the test sessions were conducted in the same room as the training sessions, but the experimenter was absent and did not comment or ask questions about choice of the machines. Expectation of social coercion or reward for choices in testing should have been reduced by the continuing indifference of the experimenter to the child's test choices.

Familiarity, habituation, skill differences, etc., may also affect choices among activities. In the present study the rate of interaction with each task was partially controlled by requiring the subject to engage in each one equally often in training; but choices in test sessions were not con- strained. A reversal procedure was used to control for familiarity. The effect of reward for one task was first tested, then for the second.

The theoretical development underlying the hypothesis that extrinsic reward may undermine intrinsic interest, as most recently formulated by Greene and Lepper (Note i), is incomplete and depends on a number of variables whose values are inferred or observable only indirectly. It cannot be predicted with certainty whether any given set of conditions will be a sufficient test of the hypothesis. Neither is it possible to deter- mine with certainty whether any specific exposure to an expected extrin- sic reward will bring about a change in intrinsic interest. Thus, the present study is not intended as a test of the theory, and these negative results cannot be taken as a refutation of the theory in its present form.

The application to token systems of the concepts proposed by Deci (1972), Greene and Lepper (Note 1), etc., represents a theoretical ex- trapolation into an area far removed from any in which adequate experi- mental data have been obtained. The present study, a realistic test of this extension, gives contrary evidence.

REFERENCE NOTE 1. Greene, D., & Lepper, M. R. An information-processing approach to intrinsic and

extrinsic motivation. Paper presented at the Annual Meetings of the American Psy- chological Association, Chicago, August, 1975.

REFERENCES Calder, B. J., & Staw, B.M. Self-perception of intrinsic and extrinsic motivation. Journal

of Personality and Social Psychology, 1975. 31, 599-605. DeCharms, R. Personal causation. New York: Academic Press, 1968. Deci, E. L, Intrinsic motivation, extrinsic reinforcement, and inequity. Journal of Per-

sonality and Social Psychology, 1972, 22, 113-120. Feingold, B. D., & Mahoney, M.M. Reinforcement effects on intrinsic interest: Under-

mining the overjustification hypothesis. Behavior Therapy, 1975, 6, 367-377.

Page 13: Intrinsic interest and extrinsic reward: The effects of a continuing token program on continuing nonconstrained preference

234 DAVIDSON AND BUCHER

Gewirtz, J. L., & Baer, D. M. The effect of brief social deprivation on behaviors for a social reinforcer. Journal of Abnormal and Social Psychology. 1958, 56, 49-56.

Greene, D. Comment upon Feingold and Mahoney's "Reinforcement effects on intrinsic interest: Undermining the overjustification hypotheses." Behavior Therapy, 1975, 6, 712-713.

Greene, D. Immediate and subsequent effects of differential reward systems on intrinsic motivation in public school classrooms. Unpublished doctoral dissertation, Stanford University, 1974.

Greene, D., & Lepper, M. R. How to turn play into work. Psychology Today, 1974, 8, 49-54.

Kazdin, A. E., & Bootzin, R. R. The token economy: An evaluative review. Journal of Applied Behavior Analysis, 1972, 5, 343-372.

Lepper, M. R., & Greene, D. Turning play into work: Effects of adult surveillance and extrinsic rewards on children's intrinsic motivation. Journal of Personality and Social Psychology, 1975, 31, 479-486.

Lepper, M. R., Greene, D., & Nisbett, R.E. Undermining children's intrinsic interest with extrinsic reward: A test of the "overjustification'" hypothesis. Journal of Personality and Social Psychology, 1973, 28, 129-137.

Levine, F. M., & Fasnacht, G. Token rewards may lead to token learning. American Psychologist, 1974, 29, 816-820.

Ross, M., Karniol, R., & Rothstein, M. Reward contingency and intrinsic motivation in children: A test of the delay of gratification hypothesis. Journal of Personality and Social Psychology. 1976, 33, 442-447.

RECEIVED: June 24, 1976; REVISED: December 31, 1976 FINAL ACCEPTANCE: January 18, 1977


Recommended