Compliance, identification, and internalization

1
three processes of attitude change

HERBERT C. KELMAN
Harvard University
A crucial issue in communication research nication produce public conformity without
relates to the nature of changes (if any) primate acceptance, or did it produce public
that are brought about by a particular com- conformity coupled with private acceptance?
munication or type of communication. It is (Cf. 1, 4.) Only if we know something
not enough to know that there has been about the nature and depth of changes can
some measurable change in attitude; usually we make meaningful predictions about the

we would also want to know what kind of way in which attitude changes will be re-
change it is. Is it a superficial change, on a flected in subsequent actions and reactions
verbal level, which disappears after a short to events.
lapse of time? Or is it a more lasting change These questions about the nature of atti-
in attitude and belief, which manifests itself tude changes are highly significant in the
in a wide range of situations and which is study of international attitudes. For exam-
integrated into the person’s value system? ple, we may have observed changes in opin-
Or, to put it in other terms, did the commu- ion toward certain international issues-e.g.,
aspects of foreign policy, international or-
1 An earlier draft of this
paper was written ganization, or disarmament-among the
while the author was with the Laboratory of Psy- population of a given country. The implica-
chology, National Institute of Mental Health, tions that we draw from these changes will
and was read at the annual meeting of the Amer-
Association in Chicago on
depend on their depth and on the psycho-
ican Psychological logical meanings that can be assigned to
August 30, 1956. The experiment reported here
was conducted while the author was at
them. Let us assume that we find an in-
Johns
crease in favorable attitudes toward the
Hopkins University as a Public Health Service
Research Fellow of the National Institute of United Nations among the population of the
Mental Health. Additional financial support was United States at a particular juncture. This
received from the Yale Communication Research
Program, which is under the direction of Carl I.
change in attitude may be due primarily to
Hovland and which is operating under a grant recent pronouncements by high-placed fig-
from the Rockefeller Foundation. The author is ures and may thus represent an aspect of
particularly grateful to James Owings for his &dquo;social conformity.&dquo; On the other hand, the
help in running the experiment; to Ramon J. change may result from a series of interna-
Rhine and Janet Baldwin Barclay for their help tional events which have led large segments
in analysis of the data; and to Roger K. Williams,
Chairman of the Psychology Department at Mor-
of the population to re-evaluate American
gan State College, for the many ways in which foreign policy and to ascribe a more central
he facilitated collection of the data. role to the UN. Depending on which of
52

these motivational processes underlies the of various kinds of communication or ex-

change in attitude, we would make different perience but also for the development of
predictions about the manifestations and propositions about the conditions for change.
consequences of the new attitudes: about In international relations, as in other areas
their durability, about the number of differ- of social behavior, one of our ultimate con-
ent attitudinal areas that will be affected by cerns is the exploration of the conditions

them, and about the ways in which they under which lasting changes occur, changes
will be translated into action and will deter- which are generalized to many situations
mine reactions to international events. Simi- and which represent some degree of value
larly, our predictions about the subsequent reorganization.
history of the new attitudes will depend on In the present paper I should like to
their cognitive links, i.e., the particular atti- describe briefly an experimental study which
tude structure within which the new attitude is concerned with some of the conditions
toward the UN is imbedded. For example, that determine the nature of attitude changes
Americans may have become more favorable produced by communications on social is-
toward the UN because an important reso- sues. The specific content of the attitudes
lution sponsored by the United States dele- that were investigated in this study was in
gate has been accepted. The new attitude the area of race relations rather than inter-
toward the UN is thus an aspect of atti- national relations. The hypotheses refer,
tudes toward one’s own nation and its pres- however, to general processes of attitude
tige and international success. On the other change, irrespective of the specific attitudi-
hand, favorableness toward the UN may nal area. Relationships found should be
have increased because UN action has suc- equally applicable, therefore, to the analy-
cessfully averted war in a very tense conflict sis of international attitudes.
situation. In this case, the new attitude to-
ward the UN is imbedded in an attitude I. Theoretical Framework
structure revolving around the whole ques- The experiment reported here grows out
tion of war and effective means of prevent- of a broader theoretical framework con-
ing its outbreak. Again, we would draw cerned with the analysis of different proc-
different implications from the changed esses of attitude change resulting from social

attitudes, depending on which of these atti- influence. It is impossible to present this

tude areas was primarily involved in the framework in detail in the present paper, but
occurrence of change. I should like to outline its main features.2
The same considerations apply when we The starting point of the theoretical analy-
interpret the effects of international commu- sis is the observation discussed in the pre-
nications. For example, if we find changes ceding paragraphs, i.e., that changes in
in the way in which nationals of different attitudes and actions produced by social
countries perceive one another, it would be influence may occur at different &dquo;levels.&dquo; It
is proposed that these differences in the
important to know at what level these
nature or level of changes that take place
changes have occurred and to what motiva-
tional and cognitive systems they are linked. correspond to differences in the process
These questions are important not only for whereby the individual accepts influence
the analysis of changes in attitude toward
----------------------------------------------------------

2A detailed description of the theoretical

various international issues, objects, or
framework and of the experiment reported here
events which may have occurred as a result will be published early in 1959 (5).
 53

In other words, the under-

(or &dquo;conforms&dquo;). the solution of a problem or find it congenial
lying processes in which an individual en- to his needs. Behavior adopted in this fash-
gages when he adopts induced behavior may ion tends to be integrated with the individu-
be different, even though the resulting overt al’s existing values. Thus the satisfaction
behavior may appear the same. derived from internalization is due to the
Three different processes of influence can content of the new behavior.
be distinguished: compliance, identification, The three processes represent three quali-
and internalization.33 tatively different ways of accepting influ-
Compliance can be said to occur when an ence. A
systematic treatment of the processes
individual accepts influence because he might, therefore, begin with an analysis of
hopes to achieve a favorable reaction from the determinants of influence in general.
another person or group. He adopts the in- These determinants can be summarized by
duced behavior not because he believes in the following proposition: The probability
its content but because he expects to gain of accepting influence is a combined func-
specific rewards or approval and avoid spe- tion of (a) the relative importance of the
cific punishments or disapproval by con- anticipated effect, ( b ) the relative power of
forming. Thus the satisfaction derived from the influencing agent, and (c) the prepo-
compliance is due to the social effect of ac- tency of the induced response. A variety of
cepting influence. experimental findings can be cited in sup-
Identification can be said to occur when port of this proposition.
an individual accepts influence because he
Compliance, identification, and internali-
wants to establish or maintain a satisfying zation can each be represented as a function
self-defining relationship to another person of these three determinants. For each proc-
or a group. This relationship may take the ess, however, these determinants take a
form of classical identification, in which the qualitatively different form. Thus the deter-
individual takes over the role of the other, minants of the three processes can be dis-
or it may take the form of a reciprocal role
tinguished from one another in terms of the
relationship. The individual actually believes nature of the anticipated effect, the source
in the responses which he adopts through of the influencing agent’s power, and the
identification, but their specific content is manner in which the induced response has
more or less irrelevant. He adopts the in- become prepotent.
duced behavior because it is associated with In other words, each process is character-
the desired relationship. Thus the satisfac- ized by a distinctive set of antecedent condi-
tion derived from identification is due to the tions, involving a particular qualitative vari-
act of conforming as such. ation of a more general set of determinants.
Internalization can be said to occur when Given the proper set of antecedents, then,
an individual accepts influence because the
influence will take the form of compliance,
content of the induced behavior-the ideas
identification, or internalization, respectively.
and actions of which it is composed-is in-
Each of these corresponds to a characteristic
trinsically rewarding. He adopts the induced pattern of internal responses (thoughts and
behavior because it is congruent with his
value system. He may consider it useful for feelings) in which the individual engages
while adopting the induced behavior.
3 A similar distinction, between four processes
Similarly, each process is characterized
of conformity, was recently presented by Marie by a distinctive set of consequent conditions,
Jahoda (3). involving a particular qualitative variation
54

in the subsequent history of the induced re- 1. When an individual adopts an induced

sponse. Responses adopted through different response through compliance, he tends to

processes will be performed under different perform it only under conditions of surveil-
conditions, will be changed and extinguished lance by the influencing agent.
under different conditions, and will have 2. When an individual adopts an induced
different properties. response through identification, he tends to
Since each of the three processes medi- perform it only under conditions of salience
ates between a distinct set of antecedents of his relationship to the agent.
and a distinct set of consequents, the pro- 3. When an individual adopts an induced
posed distinctions between the three proc- response through internalization, he tends
esses can be tested by experiments which to perform it under conditions of relevance

attempt to relate the antecedents postulated of the issue, regardless of surveillance or

for a given process to the consequents postu- salience.
lated for that process. The present experi-
ment was designed to vary one of the ante- II. Procedure
cedents-the source of the influencing agent’s The
subjects in this experiment were

power-and to observe the effects of this Negro college Freshmen in a border state.
variation on one of the consequents-the The experiment was conducted in the spring
conditions of performance of the induced of 1954, just prior to the announcement of
response. the Supreme Court decision on desegrega-
Power is defined as the extent to which tion in the public schools. The social influ-
the influencing agent is perceived as instru- ence situation to which the students were

mental to the achievement of the subject’s exposed consisted of a fixed communication

goals. The sources of the agent’s power may designed to change their attitudes on an is-
sue related to the impending Court decision.
vary (cf. 2). The following hypotheses are
offered regarding the variations in source Specifically, each of the communications
of power: employed in the study presented essentially
1. To the extent to which the power of the following message: If the Supreme Court
the influencing agent is based on means- rules that segregation is unconstitutional, it
control, conformity will tend to take the would still be desirable to maintain some of
form of compliance. the private Negro colleges as all-Negro insti-
2. To the extent to which the power of tutions, in order to preserve Negro culture,
the influencing agent is based on attractive- history, and tradition. Preliminary testing
ness, conformity will tend to take the form indicated that a large majority of the sub-
of identification. jects would initially oppose the message pre-
3. To the extent to which the power of sented in the communication.
the influencing agent is based on credibility, The communications were tape-recorded
conformity will tend to take the form of interviews between a moderator and a guest
internalization. (the communicator). They were presented
Now let us look at the consequent side. to the subjects as of radio pro-
recordings
One of the ways in which behaviors adopted grams which we were interested in evalu-
through different processes can be distin- ating. By varying the nature of these com-
guished is in terms of the conditions under munications, it was possible to manipulate
which the behavior is performed. The fol- experimentally the
and degree of the
source

lowing hypotheses are offered regarding the communicator’s power, while keeping the
conditions of performance: message of the communication constant.
 55

Four different communications were used, same message as the first communicator, but
as can be seen from Table 1, which outlines he made it clear that he was presenting
the basic design of the experiment (see left- not simply his own opinions but the over-
hand column). whelming consensus of opinion of the col-
In one communication the attempt was lege students represented in the polls. He
made to present the communicator in such was portrayed as a representative of one of

a way that he would be perceived as pos- the subjects’ reference groups and as a per-
son who was in a position to supply valid
sessing high means-control. He was intro-
duced as the president of the National Foun- information on the group norms.
dation for Negro Colleges. In the course of In the third communication the commu-
the interview it became evident that his nicator was presented in such a way that
foundation had been supporting the college he would be perceived as possessing high
in which the study was being conducted; credibility. He was introduced as a professor
that he had almost complete control over of history in one of the country’s leading
the funds expended by the foundation; and’ universities. In the course of the interview,

TABLE 1*
DESIGN OF THE EXPERIMENT AND PREDICTIONS

* H =
high probability that attitude will be expressed; L =
low probability that attitude will be expressed.

that he was the kind of person who would it became evident that he was one of the

not hesitate to use his control in order to top experts the

on history and problems of
achieve conformity. He made it clear that minority groups; that he was highly respect-
he would withdraw foundation grants from ed both by his colleagues and by members
any college in which the students took a of minority groups; and that he had a pro-
position on the issue in question which was found concern for the welfare of the Ameri-
at variance with his own position. can Negro community. He presented the

In the second communication the com- same position as the other speakers, but he

municator was presented in such a way that made it clear that this position was based
he would be perceived as possessing high on his research and on the evidence of
attractiveness. He was introduced as a Senior history.
and president of the student council in a For purposes of comparison, a fourth
leading Negro university. He was also chair- communication was used in which the com-
man of his university’s chapter of an organi- municator was presented in such a way that
zation called Student Poll, which recently he would be perceived as possessing low
did a study on the attitudes of Negro col- power, that is, as being low in means-con-
lege Seniors on issues relating to the Su- trol, attractiveness, and credibility. He was
preme Court decision. He presented the introduced as an &dquo;ordinary citizen,&dquo; who
56

had no control the subjects and no spe-

over Questionnaire I was completed under
cial knowledge about the topic. His attrac- conditions of salience and surveillance. In
tiveness and credibility were further under- order to maximize the salience of the com-
mined by the fact that he was portrayed as municator and the communication, the
a white man with a southern accent, who questionnaire was administered immediately
had recently come from Mississippi and who after the communication. In order to maxi-
did not clearly dissociate himself from a mize surveillance by the communicator, it
was made clear-both by the appearance of
pro-segregation point of view.
Each of the four communications was the questionnaire and by the instructions-
presented to a different experimental group. that this questionnaire was being adminis-
There was also a control group, which was tered at the communicator’s request and
not exposed to any communication at all. that he would see each subject’s answers.
This group provided a base line for post- Moreover, the subjects were asked to sign
communication attitudes, since we did not their names and to give other identifying
obtain before-scores from the experimental information.
subjects. Questionnaire II was completed under
After exposure to the communication, the conditions of salience and non-surveillance.
subjects in each experimental group filled In order to maximize salience, the question-
out attitude questionnaires, designed to naire was administered immediately after
measure the extent of their agreement with Questionnaire I, so that the relationship be-
the communicator. The questionnaire con- tween the questionnaire and the communi-
sisted of eighteen statements which were cation was still clearly apparent. In order to
directly or indirectly related to the central minimize surveillance, Questionnaire II was

message of the communication. Subjects anonymous, and it was made clear to the
were asked to respond to each item on a subjects that their responses would not be
seen by the communicator or by anyone
six-point scale. Scores for each item ranged
from 1 for extreme disagreement with the other than the research staff.
communicator’s position to 7 for extreme Questionnaire III was completed under
agreement (the score of 4 was not used, conditions of non-salience and non-surveil-
since the scale did not contain a neutral po- lance. In order to minimize the salience of
sition). A total score for each subject was the communicator and the communication,
computed by adding up the individual the questionnaire was administered from
scores on the 18 items. Thus the range of one to two weeks after the communication

possible total scores was from 18 (complete session, in a different place, under different
disagreement) to 126 (complete agree- auspices, and by a different experimenter.
ment). The questions were imbedded in a large
To vary the conditions of performance, number of other attitudinal items; and no
these eighteen statements to mention of the communication was made.
we presented
each subject on three separate question- In order to minimize surveillance, Question-
naires. As can be seen from Table 1, it is naire III was anonymous and was in no way
assumed that issue-relevance was present on tied in to the communication.
all three questionnaires, since the attitude
items dealt with the issues involved in the III. Hypotheses
communication. The questionnaires differed, The that were tested
specific hypotheses
however, with respect to surveillance and were as follows:
salience. 1. Attitudes adopted from a communi-
 57

cator whose power is based on means-con- really perceive each of the variations in
trol will tend to be expressed only under communicator power in the way in which
conditions of surveillance by the communi- we intended it? To
provide an answer to
cator. The mediating process postulated this question, Questionnaire II included a
here is compliance. series of statements about the speaker and
The specific prediction can be noted in the communication to which the subjects
Table 1. In the means-control group there were asked to react. An analysis of these

will be a greater amount of acceptance of data indicated that, by and large, the ex-

the communicator’s position on Question- perimental manipulations succeeded in pro-

naire I than on Questionnaires II and III, ducing the conditions they were intended to
since Questionnaire I is the only one com- produce, thus making possible an adequate
pleted under conditions of surveillance. test of the hypotheses.
There should be no significant difference be- The findings which are directly relevant
tween Questionnaires II and III. to the hypotheses are summarized in Tables
2. Attitudes adopted from a communica- 2 and 3. Table 2 presents the mean attitude
tor whose power is based on attractiveness scores for the four experimental groups on
will tend to be expressed only under condi- each of the three questionnaires. All sub-
tions of salience of the subject’s relation- jects who had completed the three ques-
ship to the communicator. The mediating tionnaires were used in this analysis.
process postulated here is identification. It can be seen from the summary of the
Specifically, it is predicted that in the at- significance tests that all the experimental
tractiveness group there will be a smaller predictions were confirmed. In the means-
amount of acceptance of the communica- control group, the mean score on Question-
tor’s position on Questionnaire III than on naire I is significantly higher than the mean
Questionnaires I and II, since Questionnaire scores on Questionnaires II and III; and

III is the only one completed under condi- there is no significant difference between
tions of non-salience. There should be no the scores on Questionnaires II and III. In
significant difference between Question- the attractiveness group, the mean score on
naires I and II. Questionnaire III is significantly lower than
3. Attitudes adopted from a communica- the mean scores on Questionnaires I and II;
tor whose power is based on credibility will and there is no significant difference be-
tend to be expressed under conditions of tween the scores on Questionnaires I and II.
relevance of the issue, regardless of surveil- In the credibility group, there are no signif-
lance or salience. The mediating process icant differences between the three ques-
postulated here is internalization. tionnaires.
The specific prediction for the credibility While these results are all in line with the

group is that there will be no significant dif- hypotheses, examination of the means in
ferences between the three questionnaires, Table 2 reveals that the findings are not so
since they were all completed under condi- clear-cut as they might be. Specifically, we
tions of issue-relevance. should expect a relatively large drop in
mean score for the means-control
group
IV. Results from Questionnaire I to Questionnaire II.
Before proceeding to examine the data In actual fact, however, the drop is only
which bear directly on the hypotheses, it slightly higher than that for the credibility
was necessary to check on the success of the
group. This might be due to the fact that
experimental variations. Did the subjects the analysis is based on all subjects, includ-
58

ing those who were not influenced by the those subjects who had scores of 60 or above
communication at all. The hypotheses, how- on Questionnaire I. Examination of the
ever, refer only to changes from question- means reveals a pattern completely consist-
naire to questionaire for those people who ent with the hypotheses. In the means-con-
were initially influenced. trol group, agreement with the communica-
It was not possible to identify the sub- tor is relatively high on Questionnaire I and

jects who were initially influenced, since declines on Questionnaires II and III. In the
there before-scores available for the
were no attractiveness group, agreement is high on

experimental groups. It was possible, how- Questionnaires I and II and declines on

ever, to approximate these conditions by Questionnaire III. In the

credibility group,
using only those subjects who had a score changes from questionnaire to questionnaire
TABLE Q
EFFECTS OF VARIATIONS IN COMMUNICATOR POWER ON ACCEPTANCE OF
INDUCED ATTITUDES UNDER THREE CONDITIONS OF MEASUREMENT

of 60 or above on Questionnaire I. If we are minimal. Analyses of variance clearly

make certain limited assumptions (which I confirmed all the experimental predictions.
cannot spell out in this brief report), it can
be shown that the use of a cutoff point of 60 V. Conclusions
&dquo;purifies&dquo; the experimental groups to some It would be premature to accept the hy-
degree. That is, the subsamples selected by potheses tested in this experiment as gen-
this criterion should have a higher ratio of eral principles that have been proved. The
influenced to uninfluenced subjects than the experiment does, however, lend consider-
total groups from which they were selected. able support to them. To the extent to
It was anticipated that an analysis based on which the hypotheses were substantiated,
these subsamples would provide a better the experiment also gives support to the
test of the hypotheses and would yield more theoretical framework from which these hy-
clear-cut results. This did, in fact, happen, potheses were derived. The mediating con-
as can be seen from Table 3. cepts of compliance, identification, and in-
Table 3 presents the mean attitude scores ternalization seem to provide a unified and
for the three high-power groups, using only meaningful way of organizing the present
 59

experimental findings and of relating them government may be based largely on com-

to a more general conceptual framework. pliance : they may go along with the ac-
The framework presented here can be ap- cepted norms in order to avoid social ostra-
plied directly to the analysis of the effects cism or perhaps even persecution. For oth-
of various communications and other forms ers, attitudes toward their government may
of social influence on attitudes and actions be largely identification-based: their rela-
in the international sphere. In the study of tionship to their own nation and its major
public opinion, for example, it should help institutions may represent an essential as-
us identify some of the conditions which pect of their identity, and acceptance of
are likely to produce one or another of these certain political attitudes and beliefs may
processes and predict the subsequent his- serve to maintain this relationship and their

tories and action implications of attitudes self-definition which is anchored in it. For

TABLE 3
EFFECTS OF VARIATIONS IN COMMUNICATOR POWER ON ACCEPTANCE OF
INDUCED ATTITUDES UNDER THREE CONDITIONS OF MEASUREMENT*

* Data based on a selected sample, containing a higher proportion of influenced subjects. Criterion. for selec-
tion was a score of 00 or above on Questionnaire I.

adopted under these sets of conditions. This a third group of individuals, belief in the
framework may also be helpful in the study country’s system of government may be in-
of the social influences which affect deci- ternalized: they may see this political form
sion-making processes and negotiations on as fully congruent and integrated with their
the part of various elites. value systems and likely to lead to a maximi-
Some of the concepts presented here zation of their own values. Our evaluation
might be useful not only for the study of of the meaning of &dquo;favorable attitudes&dquo; on
change but also for the analysis of existing the part of a particular individual or group
attitudes and their motivational bases. Let or subpopulation and our prediction of the

us take, for example, people’s attitudes to- consequences of these attitudes would cer-
ward their own country’s system of govern- tainly vary with the motivational processes
ment. Even if we look only at those individ- that underlie them. The conditions under
uals who have favorable attitudes, various which these attitudes are likely to be
distinctions suggest themselves. For some changed, the kinds of actions to which they
individuals, acceptance of their system of are likely to lead, and the ways in which
60

they are likely to affect reactions to particu- Social Power," Psychological Review, LXIII
lar events will be different, depending on (1956),181-94.
whether these attitudes are based on com- 3. JAHODA, MARIE. "Psychological Issues in Civil
Liberties," American Psychologist, XI ( 1956),
pliance, identification, or internalization. 234-40.
4. KELMAN, H. C. "Attitude Change as a Func-
REFERENCES tion of Response Restriction," Human Rela-
1. FESTINGER, L. "An Analysis of Compliant Be- tions, VI (1953), 185-214.
havior." In M. SHERIF and M. O. WILSON 5. ---. Social Influence and Personal Belief:
(eds.), Group Relations at the Crossroads. A Theoretical and Experimental Approach to
New York: Harper & Bros., 1953. the Study ofBehavior Change. New York:
2. FRENCH, J. R. P., JR. "A Formal Theory of John Wiley & Sons (to be published in 1959).