Personality Characteristics Associated with Obedience and Defiance toward Authoritative Command

Alan C. Elms

Southern Methodist University


Stanley Milgram

Harvard University

Through variations in situational factors, Milgram (1965) has elicited sharply different amounts of obedience to authoritative command. Subjects thought they were administering electric shocks to a fellow volunteer in a “learning experiment.” As many as 65% and as few as 30% of Ss in different conditions were willing to obey completely an “experimenter’s” commands to deliver increasingly high levels of “shock” to a helpless victim (actually an experimental confederate), depending upon the victim’s proximity.

Milgram has stressed the situational determinants of varying levels of obedience in different experimental conditions. But within any one condition, the situation faced by each subject is quite similar. Some Ss choose to continue obeying the E’s orders, even though they display signs of conflict (Milgram, 1963, 1965). Others, in the same situation and showing similar conflict, refuse at some point to obey further. Additionally, some Ss disobey E’s commands under conditions where the victim cannot be seen or even heard (except for a few raps on the wall); other Ss continue to obey when the victim is sitting near them, or even when they must personally force the resisting victim’s hand down onto a shock plate.

These differences in response suggest strongly that personality variables, as well as situational determinants, influence the degree of willingness to obey authoritative command. The present study was undertaken to gain information on a variety of personality variables relevant to behavior in the obedience experiments, and to indicate areas for more extensive exploration.

[This research was conducted while the authors were at Yale University. It was supported in part by U. S. Public Health Service Pre-Doctoral Research Fellowship Grant #MPM-14,761 to A. C. Elms and by National Science Foundation grants NSF G-17916 and G-251 to S. Milgram.]


Forty Ss were selected from 160 Ss who had participated in Milgram’s four-part “Proximity Series” (1965). In the Remote and Voice Feedback conditions, only aural cues from the “shock victim” were available to S; in the Proximity and Touch Proximity conditions, visual cues were present as well. So that the personality data would be minimally influenced by “borderline” Ss (Ss who might have shifted from obedience to defiance or vice versa with only slight modifications in the experimental procedure), the 20 fully obedient Ss were chosen from the 28 obedient Ss in the latter two conditions, where the pressures for defiance were greatest; while the 20 defiant Ss were chosen from the 29 defiant Ss in the first two conditions, where the pressures for obedience were greatest. The Ss were selected on no other basis, since Milgram had matched experimental groups on age and occupational category; all Ss were males. The Ss were essentially self-selected within each experimental group, by their own obedient or defiant behavior.

The Ss were contacted for the present study in the approximate temporal order of their participation in the original experiments until 20 defiant Ss and 20 obedient Ss were obtained. The subject pools from conditions 1 and 4 wore purposely exhausted first, since they represented the extremes in amount of cues for obedience or defiance. Recruitment was by letter and subsequent telephone call, offering $6.00 for a 2-hour interview dealing with Ss’ “opinions, experiences, and so forth” in connection with the original experiment. Out of 46 persons contacted, five declined to be interviewed, all of them “obedient” Ss. Four of these gave adequate reasons for not participating further (e.g., having moved out of town since the previous study); the fifth gave no reason for declining. The Ss were interviewed individually, and were paid at the beginning of the interview.2


The Ss were first administered the MMPI card form, which had been shortened approximately 25% because of time limitations. Only items not involved in the standard personality or validity scales, or in two more recently developed scales (Do, Re), were omitted. The California F Scale (forms 40-45; Adorno el al., 1950)3 was included with the MMPI, typed on cards shaped and numbered like those of the MMPI cards. The standard MMPI card-form instructions were used for the combined questionnaires, so that the California F items were categorized as “yes,” “no,” or “cannot say,” rather than on the usual Likert scale.

2    These interviews are not to be confused with those conducted by a psychiatrist and discussed in,Milgram (1964), or with the initial interview carried out for each S. The present interviews constituted a supplementary inquiry and were conducted by the first author several months after the S’s participation in the experiment.

3    Twenty-nine of the original items were used, plus a thirtieth topically revised item which replaced the “pre-war authorities” item.

The card form was followed by an oral questionnaire. Although the interviewer was frequently aware of the group from which S had been selected, the questionnaire was highly structured and was read verbatim for each S. The first question was an open-ended request to “tell me the most important things about yourself,” without further probing questions. The S’s responses, as for the questions asked subsequently, were noted in abbreviated word form and were typed out. in full immediately after the interview. Additional questions involved attitudes as a child toward parents, nature of punishment in childhood, descriptions of personality of father, mother, and self, S’s treatment of own children, experiences in combat duty, descriptions of Experimenter’s and Learner’s (“victim’s”) personalities, and attitudes toward the obedience experiment.

After this questionnaire, S was given a series of concepts to be rated by means of semantic differential scales, with instructions slightly simplified from those suggested by Osgood, Suci, and Tannenbaum (1957). The concepts, each one printed at the top of a separate page, were: Father, Yale University, Conscience, Boss, Myself, Federal Government, Adolf Eichmann, Leader, Justice, Obedience, God, Follower, Defiance, “Learner” in Memory Project, Son, Command, Hate, Experimenter in Memory Project, Law, Mother. The rating scales, chosea from those which seemed to correlate highly and relatively specifically with certain factors in the several factor rotations reported by Osgood el al., were: for the evaluative factor, good-bad, kind-cruel, pleasant-unpleasant; for the power factor, hard-soft, strong-weak, severe-lenient; for the activity factor, active-passive, fast-slow, excitable-calm; and a separate scale which seemed appropriate to the present study, aggressive-defensive. The ten scales were presented in the same order on each page, balanced in direction and ungrouped as to principal factor.

Finally, S was read ten hypothetical situations, each one dealing with problems of obedience or disobedience to authority, punishment or mercy toward others, indulgence in or refraining from cruel behavior.4 The interviewer then assured Sthat his assistance had been valuable, told him a summary of the research project would be mailed to participants, and otherwise attempted to diminish any anxiety or other unfavorable affect which might have been aroused by the interview.

4All differences between obedient and defiant Ss on this measure could easily have occurred through chance variation, and they are not discussed further.


No significant differences were found on the standard MMPI personality and validity scales (Hathaway and McKinley, 1951) when raw scores, with K added where appropriate, were compared for the two groups by means of the Mann-Whitney rank test. Nonsignificant elevations were divided evenly, with obedients ranking higher on F-validity, Sc, D, Hs, Si, and Ma scales, and defiants ranking higher on L, Pt, Pa, Mf, Pd, and Hy scales. Of the two other scales used, the scale for dominance (Do; Gough, McClosky, and Meehl, 1951) yielded no significant difference; defiant Ss were slightly more dominant. The scale for social responsibility (Re; Gough, McClosky, and Meehl, 1952) showed a significant difference (p < .05), with defiant Ss giving answers associated with greater social responsibility.

Obedient Ss ranked significantly higher on the California F Scale (mean number of items agreed to by obedients, 12.39; by defiants, 8.11; p < .003, one-tailed Mann-Whitney rank test).5 Since amount of education has been noted to be negatively correlated with California F scores (Christie, 1954), an analysis of variance of F scores was computed, in which educational level and presence or absence of obedience were used as the two factors. The results appear in Table 1. Since four obedient Ss had had only a grade-school education or less, whereas all defiant Ss had completed high school, those four Ss were not included in the analysis. The remaining Ss were divided into two educational categories, high school vs. college-educated. With education factored out, the main effect of obedience on California F scores remained significant at .06 > p > .05.

5With “cannot say” answers added to “yes” answers, the difference is still significant at p < .003. The first two Ss in each group were not given the California F scale.

The possibility of a greater tendency toward agreement response set among obedient Ss was checked by the simple procedure of taking deviant or minority responses in a sample of MMPI items (those tabulated in the first column of the MMPI card form score-sheet, about 10%

Table 1

Analysis of Variance of California F Scores of Obedient and Defiant Subjects

Source    df square    F

Obedience    1    65.72    4.01*

Education    1    50.00    3.05

Interaction    1    9.17    0.56

Within-condition    28    16.38

* .06 > p > .05.

of the total) and comparing the proportion of deviant yes answers to total deviant answers, for the two groups. A Mann-Whitney rank test showed no significant difference between the two groups (p > .20); obedient Ss gave 27% deviant yes answers, to defiant Ss’ 24% deviant yes answers.

The interview questionnaire yielded relatively uniform responses from both groups on such categories as source and frequency of punishment reportedly given to S as a child, and punishment given to S’s own children. Blind ratings of responses on the initial open-ended question were compared for a number of quantifiable categories, but no distinct differences were found. The following questions yielded noticeable differences:

“How close were you to your father when you were a child?” Obedient Ss reported being less close than defiants, on a five-point scale from “extremely close” to “extremely distant” (defiant X = 1.95, obedient X = 2.76; p < .05).

“How were you usually punished?” Several obedient Ss reported extremely mild or no punishment at all, and the bulk of the others reported the standard spanking. Defiant Ss more frequently reported physical or emotional deprivation, with several reporting intense physical punishment. The results are not readily quantifiable.

Table 2

Mean Number of Positive and Negative Words in Personality Descriptions by Obedient and Defiant Subjects



Father’s personality, pos… 3.00


Father’s personality, neg… 1.63


Mother’s personality, pos… 3.75


Mother’s personality, neg… 0.75


“Learner’s” personality, pos… 0.35


 ”Learner’s” personality, neg… 3.35


Experimenter’s personality, pos… 3.25


Experimenter’s personality, neg… 0.75



If father is living now, “When you see him, how do you get along with him now?” Both groups report being “closer” to father presently than in the past. Obedients report a greater positive change, so that although defiant Ss still report being closer on the average, there is no longer a significant difference between the two groups (defiant X = 1.30, obedient X = 1.55; p>.20).

“If you had to use five different words to describe your father’s personality, what would the five words be?” The interviewer later did a blind rating of each word, classifying it as a positive, negative, or neutral personality trait description. A chi-square computation for positive and negative word totals for obedient and defiant Ss indicates a significant difference (p < .005), with obedients giving fewer positive words and more negative words than did defiants. A similar question on mother’s personality did not yield a significant difference. (For means, see Table 2.)

Table 3

Mean Semantic Differential Ratings Yielding Significant Differences Between Obedient and Defiant Subjects





Evaluative ratings
Yale University



p < .025




p < .05




.10 > p > .05




p < .005




.10 > p > .05

Power ratings



p < .05




p < .025




.10 > p > .05




p < .05

Activity ratings



p = .025

Yale University



.10 > p > .05




p < .025




.10 > p > .05

Aggressiveness ratings
Yale University



p < .0005




p < .05




p = .10




.10 > p > .05

aComputed by one-tailed Student’s

t test. Ratings were

made on a 7-point



If S had ever been on active military duty, “Did you ever shoot at a man in combat?” Of ten obedient Ss who had been on active duty, eight said they had shot at men (one of these said he had shot at a Forest Ranger stateside). Of eight defiant Ss who had been on active duty, one said he had shot 50-mm guns at distant targets, and one said he had shot a gun in battle, “but not at a person.” Of eight obedient Ss who had shot at a man, two said they had killed a man; three said they did not know whether they had killed anyone or not; two gave tentative denials (e.g., “Not that I know of”). Both defiant Ss who said they had shot a gun in battle gave tentative denials.

“If you had to use five words to describe this man’s (the ‘Learner’s') personality, what would the words be?” Obedient Ss gave fewer positive and more negative words than did defiants (p < .001, chi-square).

“If you had to use five words to describe the Experimenter’s personality, what would the five words be?” Obedient Ss gave more positive and fewer negative words than did defiants (p < .005, chi-square).

The semantic differential data were analyzed in terms of differences in mean judgments assigned to a concept on a specific scale or factor by the two groups. For the evaluative, activity, and power factors, the practice of averaging three contributory scales for each was followed. The single aggressive-defensive scale was analyzed separately. Table 3 presents those differences which approach significance. (For each judgmental factor, the lower average score represents assignment of a more positive rating on that factor to the concept being rated.)


Previous researchers have identified various personality characteristics associated with apparent tendencies toward submission to authority (e.g., Adorno et al., 1950; Rokeach, 1960). But submission to authority and similar characteristics in these studies were usually measured in the same way as the personality characteristics with which they were correlated. As a result, certain forms of response set common to the measures both of authoritarian tendencies and of personality characteristics may have elevated whatever relationships existed, and may have “created” relationships which did not exist. In the present research, the basic measure of submission to authoritative command is an observation of actual behavior in a realistic situation. Thus, although it now seems clear that response sets contribute to scores on the California F Scale, for instance, it is unlikely that the same kind of response set determined Ss’ responses in a situation calling for their actual obedience to commands to administer shocks to an innocent person, ultimately against that other person’s will.

With this in mind, the significant difference between obedient and defiant Ss on the California F Scale remains meaningful even when educational level is not factored out. (As noted, the difference between obedient and defiant Ss’ scores still approaches significance even with educational level controlled statistically.) Less-educated Ss may score higher on the F Scale because they are more willing to agree to blanket statements than are well-educated Ss, or because cliches and commonplaces appeal to them; but behavior in the experimental obedience situation requires more than acceptance of cliches or broad generalizations. Low education may be associated with authoritarian verbalizations or authoritarian behavior not only because of flaws in the measuring instrument, but because degree of education is itself in some way related to incidence of authoritarian personality characteristics. Whether the educative process itself diminishes authoritarianism in some way, or whether some other independent variable influences the two dependent variables of educational achievement and authoritarianism, is not evident from the present data.

Other similarities to findings reported in The Authoritarian Personality are found in obedient Ss’ feelings of the father’s lack of closeness when S was a child, and the relative glorification of the E and downgrading of the Learner. The authoritarian tendency toward stereotyped glorification of the father is not observed in obedients— quite the contrary—, but this may be partly the result of the interviewer’s negative ratings of “stern” and other authoritarian adjectives describing the father, in the blind categorization of descriptions. Additionally, the generally older Ss in the present study may find it easier to express critical attitudes toward their fathers than the young adults who composed a large proportion of Adorno et al.’s samples.

The reporting of less severe childhood punishment by obedients is also atypical of the “authoritarian personality.” Lack of differences between obedients and defiants in attitudes toward mother, toward Adolph Eichmann, and in several other categories where differences might be expected if a strictly authoritarian-unauthoritarian dichotomy were present, may have resulted in part from extreme elevations or depressions of judgments by all Ss, which did not allow for real differentiation on a 5-or 7-point scale.

Obedient Ss’ praising of the power figure (E) and denigration of the weak figure (Learner) may derive more directly from features of the experiment itself than do other characteristics of obedients. Self-justification for one’s experimental behavior may come ex post facto from recalling the E as more benevolent and the Learner as less worthy: the Good Scientist deserved to be followed, while the stupid, excitable. weak Learner deserved to be given a lesson. It is impossible to derive from the present data an indication of the direction of causality—whether the subject behaved as he did because he entered the experiment with potential stereotyped responses to power and weakness, or whether he later gave these stereotyped responses to power or weakness because of what he had done in the experiment, or both. Whichever pattern of causation is more accurate, obedients and defiants did respond differentially at the time of the original experiment, and such differences indicate probable initial differences in the two groups.

The response to the question “Did you ever shoot at a man in combat?” indicates a difference in attitude toward infliction of pain or harm. Obedient Ss may or may not have shot at men more frequently than defiant Ss have; if so, they may have felt less conflict over merely shocking a man. The reliability of their self-reports is unknown, but we can at least assume that obedient Ss more easily accept the idea of injuring others under certain circumstances. Eighty percent of obedient Ss with combat experience admit without qualification to having shot at people, while even the 25% of defiant combat-experienced Ss who had shot a gun in battle deny having shot at an individual person within seeing distance. The ability or lack of ability to accept such behavior could have (and may well have had) important consequences in an experimental situation in which an authority figure demands aggressive action against another person. The similarity to the authoritarian personality is again apparent; for instance, Adorno et al. (p. 386) report that “we often find in our high-scoring subjects both overconformity and underlying destructiveness toward established authority, customs, and institutions. A person possessed by such ambivalence may easily be kept in check and may even behave in an exemplary fashion in following those external authorities who take over the function of the superego—and partly even those of the ego. On the other hand, if permitted to do so by outside authority, the same person may be induced very easily to uncontrolled release of his instinctual tendencies, especially those of destructiveness.”

The semantic differential ratings lend additional detail to this picture of the obedient S as authoritarian personality. For instance, obedient Ss attribute more “goodness” to the two authority figures in the experimental situation, Yale University and the E himself; less goodness to the weak Learner; and apparently view the command-obedience situation more positively even in the abstract. At the same time, they see the E and Yale University as more aggressive, like themselves; and, consonant with their own actions, they impute more aggressiveness to obedience itself.

The absence of differences between groups on the MMPI scales is not altogether surprising. The scales may be able to detect certain pathological or “normal” personality patterns, but none of these patterns fits, in a clear majority of particulars, the qualities which might be expected of an obedient or defiant S. Even the description of the reference groups for the social responsibility scale (Gough et al., 1952), which showed the only significant difference, contains conflicting characteristics from the viewpoint of the present study. The socially responsible person “shows a ready willingness to accept the consequences of his own behavior,” a “sense of obligation to the [peer] group,” and “greater concern for social and moral issues,” as one might expect of the higher-scoring defiant Ss; but he also shows “dependability” and “trustworthiness” and is “more compliant and acquiescent,” “less rebellious and recalcitrant,” as one might expect of obedient Ss.

Various other differences appear between the two groups on the several parts of the interview, but among the multiplicity of questions it might be misleading to try to gather together the differences which do not approach statistically significant levels. Even significant differences may sometimes be deceptive. For instance, although in a number of instances obedient Ss displayed characteristics similar to those of high scorers in The Authoritarian Personality, several obedient Ss appeared to have warm relationships with family and with associates. One of these, otherwise apparently kind and sensitive to an extreme, seemed to have taken as a model for public behavior a grandfather who he said “believed one should take and carry out an order whether one believed it was right or wrong, as long as the person giving it was in authority to give it.” Likewise, defiant Ss did not consistently show themselves in the interviews to be warmly humanitarian; one, a liberal and humanist by principle, displayed considerable bitterness toward his fellow man, including “a generally low opinion of the intellectual level of mankind.”

The experimental obedience situation allows the S to express himself in binary fashion. He may continue to obey the E or he may break off the experiment. Behind this simple behavioral possibility may lie highly complex and possibly idiosyncratic motive structures. There may be only a functional equivalence of motive structures, in that different combinations of motives may lead to the same behavioral outcome. For example, a highly compliant S who is low in aggressive needs may obey the E to the very end. The same effect could be produced in a person whose submissive needs are low, but who possesses a strong need to release aggressive tensions. Because of the complexity of motives which subjects bring to the experimental situation, it is not possible to reduce twenty individuals to a statistically average “obedient subject” or “defiant subject.” The results of this study suggest certain broad personality differences which relate to obedience or defiance in the experimental obedience situation; but they do not reveal a single personality pattern which is inevitably expressed in one behavior or the other.


Adorno, T. W., Frenkel-Brunswik, E., Levinson, D., and Sanford, R. N. The authoritarian personality. New York: Harper, 1950.

Christie, R. Authoritarianism re-examined. In R. Christie and M. Jahoda (eds.), Studies in the scope and method of “The authoritarian personality.” Glencoe, Illinois: The Free Press, 1954. Pp. 123-196.

Gough, H. G., McClosky, H., and Meehl, P. E. A personality scale for dominance. Journal of Abnormal and Social Psychology, 1951, 46, 360-66.

Gough, H. G., McClosky, H., and Meehl, P. E. A personality scale for social responsibility. Journal of Abnormal and Social Psychology, 1952, 47, 73-80.

Hathaway, S. R., and McKinley, J. C. The Minnesota Multiphasic Personality Inventory manual. Revised. New York: The Psychological Corporation, 1951.

Milgram, S. Behavioral study of obedience. Journal oj Abnormal and Social Psychology, 1963,    67, 371-78.

Milgram, S. Issues in the study of obedience: A reply to Baumrind. American Psychologist, 1964,    19, 848-52.

Milgram, S. Some conditions of obedience and disobedience to authority. Human Relations, 1965, 18, 57-76.

Osgood, C. E., Suci, G. J., and Tannenbaum, P. H. The measurement of meaning. Urbana: Univ. of Illinois Press, 1957. 
Rokeach, M. The open and closed mind. New York: Basic Books, 1960.

[Originally published in Journal of Experimental Research in Personality, Vol. 1, No. 4, Dec.1966, pp. 282-289.]

Leave a Reply

The Website of Alan C. Elms