Journal of Abnormal and Social Psychology1963, Vol. 67, No. 4, 371-378BEHAVIORAL STUDY OF OBEDIENCE1STANLEY MILGRAM 2Yale UniversityThis article describes a procedure for the study of destructive obedience inthe laboratory. It consists of ordering a naive S to administer increasinglymore severe punishment to a victim in the context of a learning experiment.Punishment is administered by means of a shock generator with 30 gradedswitches ranging from Slight Shock to Danger: Severe Shock. The victim is aconfederate of the E. The primary dependent variable is the maximum shockthe S is willing to administer before he refuses to continue further. 26 Ssobeyed the experimental commands fully, and administered the highest shockon the generator. 14 Ss broke off the experiment at some point after thevictim protested and refused to provide further answers. The procedure createdextreme levels of nervous tension in some Ss. Profuse sweating, trembling, andstuttering were typical expressions of this emotional disturbance. One un-expected sign of tension—yet to be explained—was the regular occurrence ofnervous laughter, which in some Ss developed into uncontrollable seizures.The variety of interesting behavioral dynamics observed in the experiment,the reality of the situation for the S, and the possibility of parametric varia-tion within the framework of the procedure, point to the fruitfulness offurther study.Obedience is as basic an element in thestructure of social life as one can point to.Some system of authority is a requirementof all communal living, and it is only theman dwelling in isolation who is not forcedto respond, through defiance or submission,to the commands of others. Obedience, asa determinant of behavior, is of particularrelevance to our time. It has been reliablyestablished that from 1933-45 millions ofinnocent persons were systematically slaugh-tered on command. Gas chambers were built,death camps were guarded, daily quotas ofcorpses were produced with the same ef-ficiency as the manufacture of appliances.These inhumane policies may have originatedin the mind of a single person, but they couldonly be carried out on a massive scale if avery large number of persons obeyed orders.Obedience is the psychological mechanismthat links individual action to political pur-pose. It is the dispositional cement that bindsmen to systems of authority. Facts of recenthistory and observation in daily life suggest1This research was supported by a grant (NSFG-17916) from the National Science Foundation.Exploratory studies conducted in 1960 were sup-ported by a grant from the Higgins Fund at YaleUniversity. The research assistance of Alan C. Elmsand Jon Wayland is gratefully acknowledged.2Now at Harvard University.that for many persons obedience may be adeeply ingrained behavior tendency, indeed,a prepotent impulse overriding training inethics, sympathy, and moral conduct. C. P.Snow (1961) points to its importance whenhe writes:When you think of the long and gloomy historyof man, you will find more hideous crimes havebeen committed in the name of obedience thanhave ever been committed in the name of rebellion.If you doubt that, read William Shirers "Rise andFall of the Third Reich." The German Officer Corpswere brought up in the most rigorous code ofobedience . . . in the name of obedience they wereparty to, and assisted in, the most wicked largescale actions in the history of the world [p. 24].While the particular form of obediencedealt with in the present study has its ante-cedents in these episodes, it must not bethought all obedience entails acts of aggres-sion against others. Obedience serves numer-ous productive functions. Indeed, the verylife of society is predicated on its existence.Obedience may be ennobling and educativeand refer to acts of charity and kindness,as well as to destruction.General ProcedureA procedure was devised which seemsuseful as a tool for studying obedience(Milgram, 1961). It consists of ordering371
372 STANLEY MILGRAMa naive subject to administer electric shockto a victim. A simulated shock generator isused, with 30 clearly marked voltage levelsthat range from IS to 450 volts. The instru-ment bears verbal designations that rangefrom Slight Shock to Danger: Severe Shock.The responses of the victim, who is a trainedconfederate of the experimenter, are stand-ardized. The orders to administer shocks aregiven to the naive subject in the context ofa "learning experiment" ostensibly set up tostudy the effects of punishment on memory.As the experiment proceeds the naive subjectis commanded to administer increasinglymore intense shocks to the victim, even tothe point of reaching the level markedDanger: Severe Shock. Internal resistancesbecome stronger, and at a certain point thesubject refuses to go on with the experi-ment. Behavior prior to this rupture is con-sidered "obedience," in that the subject com-plies with the commands of the experimenter.The point of rupture is the act of disobedi-ence. A quantitative value is assigned to thesubjects performance based on the maximumintensity shock he is willing to administerbefore he refuses to participate further. Thusfor any particular subject and for any par-ticular experimental condition the degree ofobedience may be specified with a numericalvalue. The crux of the study is to systemati-cally vary the factors believed to alterthe degree of obedience to the experimentalcommands.The technique allows important variablesto be manipulated at several points in theexperiment. One may vary aspects of thesource of command, content and form of com-mand, instrumentalities for its execution,target object, general social setting, etc. Theproblem, therefore, is not one of designing in-creasingly more numerous experimental con-ditions, but of selecting those that best illumi-nate the process of obedience from the socio-psychological standpoint.Related StudiesThe inquiry bears an important relation tophilosophic analyses of obedience and author-ity (Arendt, 1958; Friedrich, 1958; Weber,1947), an early experimental study ofobedience by Frank (1944), studies in "au-thoritarianism" (Adorno, Frenkel-Brunswik,Levinson, & Sanford, 1950; Rokeach, 1961),and a recent series of analytic and empiricalstudies in social power (Cartwright, 1959).It owes much to the long concern withsuggestion in social psychology, both in itsnormal forms (e.g., Binet, 1900) and in itsclinical manifestations (Charcot, 1881). Butit derives, in the first instance, from directobservation of a social fact; the individualwho is commanded by a legitimate authorityordinarily obeys. Obedience comes easily andoften. It is a ubiquitous and indispensablefeature of social life.METHODSubjectsThe subjects were 40 males between the ages of20 and 50, drawn from New Haven and the sur-rounding communities. Subjects were obtained bya newspaper advertisement and direct mail solicita-tion. Those who responded to the appeal believedthey were to participate in a study of memory andlearning at Yale University. A wide range ofoccupations is represented in the sample. Typicalsubjects were postal clerks, high school teachers,salesmen, engineers, and laborers. Subjects ranged ineducational level from one who had not finishedelementary school, to those who had doctorate andother professional degrees. They were paid $4.50 fortheir participation in the experiment. However, sub-jects were told that payment was simply for comingto the laboratory, and that the money was theirsno matter what happened after they arrived. Table 1shows the proportion of age and occupational typesassigned to the experimental condition.Personnel and LocaleThe experiment was conducted on the grounds ofYale University in the elegant interaction laboratory.(This detail is relevant to the perceived legitimacyof the experiment. In further variations, the experi-TABLE 1DISTRIBUTION OF AGE AND OCCUPATIONAL TYPESIN THE EXPERIMENTOccupationsWorkers, skilledand unskilledSales, business,andwhite-collarProfessionalPercentage oftotal (Age)20-29years4312030-39years56S4040-50years67340Percentageof total(Occupa-tions)37.S40.022.5Note.—Total N = 40.
OBEDIENCE 373ment was dissociated from the university, withconsequences for performance.) The role of experi-menter was played by a 31-year-old high schoolteacher of biology. His manner was impassive,and his appearance somewhat stern throughout theexperiment. He was dressed in a gray technicianscoat. The victim was played by a 47-year-oldaccountant, trained for the role; he was of Irish-American stock, whom most observers found mild-mannered and likable.ProcedureOne naive subject and one victim (an accomplice)performed in each experiment. A pretext had to bedevised that would justify the administration ofelectric shock by the naive subject. This was ef-fectively accomplished by the cover story. After ageneral introduction on the presumed relation be-tween punishment and learning, subjects were told:But actually, we know very little about theeffect of punishment on learning, because almostno truly scientific studies have been made of itin human beings.For instance, we dont know how much punish-ment is best for learning—and we dont knowhow much difference it makes as to who is givingthe punishment, whether an adult learns best froma younger or an older person than himself—ormany things of that sort.So in this study we are bringing together anumber of adults of different occupations andages. And were asking some of them to beteachers and some of them to be learners.We want to find out just what effect differentpeople have on each other as teachers and learn-ers, and also what effect punishment will haveon learning in this situation.Therefore, Im going to ask one of you to bethe teacher here tonight and the other one to bethe learner.Does either of you have a preference?Subjects then drew slips of paper from a hat todetermine who would be the teacher and who wouldbe the learner in the experiment. The drawing wasrigged so that the naive subject was always theteacher and the accomplice always the learner. (Bothslips contained the word "Teacher.") Immediatelyafter the drawing, the teacher and learner weretaken to an adjacent room and the learner wasstrapped into an "electric chair" apparatus.The experimenter explained that the straps wereto prevent excessive movement while the learnerwas being shocked. The effect was to make it im-possible for him to escape from the situation. Anelectrode was attached to the learners wrist, andelectrode paste was applied "to avoid blisters andburns." Subjects were told that the electrode wasattached to the shock generator in the adjoiningroom.In order to improve credibility the experimenterdeclared, in response to a question by the learner:"Although the shocks can be extremely painful, theycause no permanent tissue damage."Learning task. The lesson administered by thesubject was a paired-associate learning task. Thesubject read a series of word pairs to the learner,and then read the first word of the pair along withfour terms. The learner was to indicate which ofthe four terms had originally been paired with thefirst word. He communicated his answer by pressingone of four switches in front of him, which in turnlit up one of four numbered quadrants in an answer-box located atop the shock generator.Shock generator. The instrument panel consists of30 lever switches set in a horizontal line. Eachswitch is clearly labeled with a voltage designa-tion that ranges from IS to 450 volts. There is a15-volt increment from one switch to the next goingfrom left to right. In addition, the following verbaldesignations are clearly indicated for groups of fourswitches going from left to right: Slight Shock,Moderate Shock, Strong Shock, Very Strong Shock,Intense Shock, Extreme Intensity Shock, Danger:Severe Shock. (Two switches after this last designa-tion are simply marked XXX.)Upon depressing a switch: a pilot light correspond-ing to each switch is illuminated in bright red; anelectric buzzing is heard; an electric blue light,labeled "voltage energizer," flashes; the dial on thevoltage meter swings to the right; various relayclicks are sounded.The upper left-hand corner of the generator islabeled Shock Generator, Type ZLB, Dyson Instru-ment Company, Waltham, Mass. Output IS Volts-4SO Volts.Details of the instrument were carefully handledto insure an appearance of authenticity. The panelwas engraved by precision industrial engravers, andall components were of high quality. No subject inthe experiment suspected that the instrument wasmerely a simulated shock generator.3Sample shock. Each naive subject is given asample shock on the shock generator, prior to be-ginning his run as teacher. This shock is always45 volts, and is applied by pressing the third switchof the generator. The shock is applied to the wristof the naive subject, and has its source in a 45-voltbattery wired into the generator. This furtherconvinces the subject of the authenticity of thegenerator.Shock instructions. The subject is told to admin-ister a shock to the learner each time he gives a8A related technique, making use of a shockgenerator, was reported by Buss (1961) for thestudy of aggression in the laboratory. Despite theconsiderable similarity of technical detail in theexperimental procedures, both investigators proceededin ignorance of the others work. Milgram providedplans and photographs of his shock generator, experi-mental procedure, and first results in a report to theNational Science Foundation in January 1961. Thisreport received only limited circulation. Buss re-ported his procedure 6 months later, but to a wideraudience. Subsequently, technical information andreports were exchanged. The present article was firstreceived in the Editors office on December 27, 1961;it was resubmitted with deletions on July 27, 1962.
374 STANLEY MILGRAMwrong response. Moreover—and this is the keycommand—the subject is instructed to "move onelevel higher on the shock generator each time thelearner flashes a wrong answer." He is also instructedto announce the voltage level before administeringa shock. This serves to continually remind subjectsof the increasing intensity of shocks administeredto the learner.Preliminary and regular run. Pretests revealedthat the procedure of reading words and administer-ing shocks requires some practice before it can behandled smoothly. Therefore, immediately precedingthe regular run, the teacher is given a preliminaryseries of 10 words to read to the learner. Thereare 3 neutral words in the. practice series (i.e., wordsthat the learner answers correctly), so that shocksare administered for 7 of the words, with themaximum shock at 105 volts (moderate shock).Almost all subjects master the procedure by thetime the preliminary run is over.Subjects are then presented with a second list,and are told that the procedure is the same as forthe first list; the experimenter adds, however:When you get to the bottom of the list, repeatit over again, and continue giving shocks, untilthe learner has learned all the pairs correctly.The experimenter instructs the subject to,start from 15 volts and increase the shock levelone step each time the learner gives a wronganswer.Feedback from the victim. In all conditions thelearner gives a predetermined set of responses tothe word pair test, based on a schedule of approxi-mately three wrong answers to one correct answer.In the present experimental condition no vocal re-sponse or other sign of protest is heard from thelearner until Shock Level 300 is reached. When the300-volt shock is administered, the learner poundson the wall of the room in which he is bound tothe electric chair. The pounding can be heard bythe subject. From this point on, the learners answersno longer appear on the four-way panel.At this juncture, subjects ordinarily turn to theexperimenter for guidance. The experimenter in-structs the subject to treat the absence of a responseas a wrong answer, and to shock the subject accord-ing to the usual schedule. He advises the subjects toallow S-10 s«conds before considering no responseas a wrong answer, and to increase the shock levelone step each time the learner fails to respond cor-rectly. The learners pounding is repeated after the315-volt shock is administered; afterwards he is notheard from, nor do his answers reappear on thefour-way signal box.Experimenter feedback. At various points in theexperiment the subject may turn to the experimenterfor advice on whether he should continue to admin-ister shocks. Or he may indicate that he does notwish to go on. It was important to standardizethe experimenters responses to these inquiries andreactions.If the subject indicated his unwillingness to goon, the experimenter responded with a sequence of"prods," using as many as necessary to bring thesubject into line.Prod 1 : Please continue, or Please go on.Prod 2: The experiment requires that you con-tinue.Prod 3: It is absolutely essential that you con-tinue.Prod 4: You have no other choice, you mustgo on.The prods were always made in sequence: Onlyif Prod 1 had been unsuccessful, could Prod 2 beused. If the subject refused to obey the experi-menter after Prod 4, the experiment was terminated.The experimenters tone of voice was at all timesfirm, but not impolite. The sequence was begunanew on each occasion that the subject balked orshowed reluctance to follow orders.Special prods. If the subject asked if the learnerwas liable to suffer permanent physical injury, theexperimenter said:Although the shocks may be painful, there isno permanent tissue damage, so please go on.[Followed by Prods 2, 3, and 4 if necessary.]If the subject said that the learner did not wantto go on, the experimenter replied:Whether the learner likes it or not, you mustgo on until he has learned all the word pairscorrectly. So please go on. [Followed by Prods2, 3, and 4 if necessary.]Dependent MeasuresThe primary dependent measure for any subjectis the maximum shock he administers before herefuses to go any further, In principle this mayvary from 0 (for a subject who refuses to administereven the first shock) to 30 (for a subject whoadministers the highest shock on the generator). Asubject who breaks off the experiment at any pointprior to administering the thirtieth shock level istermed a defiant subject. One who complies withexperimental commands fully, and proceeds to ad-minister all shock levels commanded, is termed anobedient subject.Further records. With few exceptions, experi-mental sessions were recorded on magnetic tape.Occasional photographs were taken through one-waymirrors. Notes were kept on any unusual behavioroccurring during the course of the experiments. Onoccasion, additional observers were directed to writeobjective descriptions of the subjects behavior. Thelatency and duration of shocks were measured byaccurate timing devices.Interview and dehoax. Following the experiment,subjects were interviewed; open-ended questions,projective measures, and attitude scales were em-ployed. After the interview, procedurestaken to assure that the suoject would leave theRBoratory in a state of well being. A friendlyreconciliation was arranged between thejubject and5ie~vlctimj and an effort was made to^Feduce anytensions that arose as a result of the experiment.
OBEDIENCE 375RESULTSPreliminary NotionsBefore giving an account of the experi-mental results, it is instructive to considerhow persons predict others will perform whencommanded to administer powerful shocksto another person. Fourteen Yale seniors, allpsychology majors, were provided with adetailed description of the experimental situa-tion. They were asked to reflect carefully onit, and to predict the behavior of 100 hypo-thetical subjects. More specifically, they wereinstructed to plot the distribution of obedi-ence of "100 Americans of diverse occupa-tions, and ranging in age from 20 to SOyears," who were placed in the experimentalsituation.There was considerable agreement amongthe respondents on the expected behavior ofhypothetical subjects. All respondents pre-dicted that only an insignificant minoritywould go through to the end of the shockseries. (The estimates ranged from 0 to 3%;i.e., the most "pessimistic" member of the classpredicted that of 100 persons, 3 would con-tinue through to the most potent shock avail-able on the shock generator—450 volts.)The class mean was 1.2%. The question wasalso posed informally to colleagues of theauthor, and the most general feeling wasthat few if any subjects would go beyondthe designation Very Strong Shock.Result of Experiment ISubjects accept situation. With few excep-tions subjects were convinced of the realityof the experimental situation, that they wereadministering shocks to another person, andthe most potent shocks were extremelypainful. In the post-experimental interviewsubjects were asked: "How painful tothe learner were the last few shocks thatyou administered to him?" Subjects wereinstructed to indicate their answers on aprinted 14-point scale ranging from "Not atall painful" to "Extremely painful." Themodal response was 14 (Extremely painful)with the mean at 13.42.Signs of extreme tension. Many subjectsshowed signs of nervousness in the experi-mental situation, and especially upon ad-ministering the more powerful shocks. In alarge number of cases the degree of tensionreached extremes that are rarely seen insociopsychological laboratory studies. Sub-jects were observed to sweat, tremble, stutter,bite their lips, groan, and dig their finger-nails into their flesh. These were character-istic rather than exceptional responses to theexperiment.One sign of tension was the regular occur-rence of nervous laughing fits. Fourteen ofthe 40 subjects showed definite signs ofnervous laughter and smiling. The laughterseemed entirely out of place, even bizarre.Full-blown, uncontrollable seizures were ob-served for 3 subjects. On one occasion weobserved a seizure so violently convulsivethat it was necessary to call a halt to theexperiment. The subject, a 46-year-old en-cyclopedia salesman, was seriously embar-rassed by his untoward and uncontrollablebehavior. In the post-experimental interviewssubjects took pains to point out that theywere not sadistic types, and that the laughterdid not mean they enjoyed shocking thevictim.Distribution of scores. It had been con-jectured that persons would not, in general,agree to administer shocks beyond the leveldesignated as Very Strong Shock. The dis-tribution of maximum shocks for this experi-ment is shown in Table 2. The obtained dis-tribution of scores deviates radically from theprediction.Upon command of the experimenter, eachof the 40 subjects went beyond the expectedbreakoff point. No subject stopped prior toadministering Shock Level 20. (At this level—300 volts—the victim kicks on the wall andno longer provides answers to the teachersmultiple-choice questions.)Of the 40 subjects, 5 refused to obey theexperimental commands beyond the 300-voItlevel. Four more subjects administered onefurther shock, and then refused to go on.Two broke off at the 330-volt level, and1 each at 345, 360, and 375 volts. Thus atotal of 14 subjects defied the experimenter.It is instructive to reprint their remarks atthe point of defiance, as transcribed from thetape recordings: I think hes trying to communicate, hesknocking. . . . Well its not fair to shock the guy. . . these are terrific volts. I dont think this is
376 STANLEY MILGEAMTABLE 2DISTRIBUTION or BREAKorr POINTSVerbal designation andvoltage indicationSlight ShockIS304560Moderate Shock7590105120Strong Shock135150165180Very Strong Shock195210225240Intense Shock255270285300Extreme Intensity Shock315330345360Danger: Severe Shock375390405420XXX435450Number of subjects for whomthis was maximum shock0000000000000000000542111000026very humane. . . . Oh, I cant go on with this; no,this isnt right. Its a hell of an experiment. The guyis suffering in there. No, I dont want to go on.This is crazy. [Subject refused to administer moreshocks.] Hes banging in there. Im gonna chickenout. Id like to continue, but I cant do that toa man. . . . Im sorry I cant do that to a man.Ill hurt his heart. You take your check. . . . Noreally, I couldnt do it.These subjects were frequently in a highlyagitated and even angered state. Sometimes,verbal protest was at a minimum, and thesubject simply got up from his chair in frontof the shock generator, and indicated thathe wished to leave the laboratory.Of the 40 subjects, 26 obeyed the ordersof the experimenter to the end, proceedingto punish the victim until they reached themost potent shock available on the shockgenerator. At that point, the experimentercalled a halt to the session. (The maximumshock is labeled 450 volts, and is two stepsbeyond the designation: Danger: SevereShock.) Although obedient subjects continuedto administer shocks, they often did so underextreme stress. Some expressed reluctance toadminister shocks beyond the 300-volt level,and displayed fears similar to those whodefied the experimenter; yet they obeyed.After the maximum shocks had been de-livered, and the experimenter called a haltto the proceedings, many obedient subjectsheaved sighs of relief, mopped their brows,rubbed their fingers over their eyes, ornervously fumbled cigarettes. Some shooktheir heads, apparently in regret. Some sub-jects had remained calm throughout theexperiment, and displayed only minimal signsof tension from beginning to end.DISCUSSIONThe experiment yielded two findings thatwere surprising. The first finding concernsthe sheer strength of obedient tendenciesmanifested in this situation. Subjects havelearned from childhood that it is a funda-mental breach of moral conduct to hurt an-other person against his will. Yet, 26 subjectsabandon this tenet in following the instruc-tions of an authority who has no specialpowers to enforce his commands. To disobeywould bring no material loss to the subject;no punishment would ensue. It is clear fromthe remarks and outward behavior of manyparticipants that in punishing the victim theyare often acting against their own values.Subjects often expressed deep disapproval ofshocking a man in the face of his objections,and others denounced it as stupid and sense-less. Yet the majority complied with theexperimental commands. This outcome wassurprising from two perspectives: first, fromthe standpoint of predictions made in thequestionnaire described earlier. (Here, how-ever, it is possible that the remoteness ofthe respondents from the actual situation, andthe difficulty of conveying to them the con-
OBEDIENCE 377crete details of the experiment, could accountfor the serious underestimation of obedience.)But the results were also unexpected topersons who observed the experiment inprogress, through one-way mirrors. Observersoften uttered expressions of disbelief uponseeing a subject administer more powerfulshocks to the victim. These persons had afull acquaintance with the details of thesituation, and yet systematicallymated the amount of obedience that subjectswould disglay.Thr Mrnnd unanticipated effect was theextraordinary tension generated by the pro-cedures. One might suppose that a subjectwould simply break off or continue as hisconscience dictated. Yet, this is very far fromwhat happened. There were striking reac-tions of tension and emotional strain. Oneobserver related:I observed a mature and initially poised business-man enter the laboratory smiling and confident.Within 20 minutes he was reduced to a twitching,stuttering wreck, who was rapidly approaching apoint of nervous collapse. He constantly pulled onhis earlobe, and twisted his hands. At one pointhe pushed his fist into his forehead and muttered:"Oh God, lets stop it." And yet he continued torespond to every word of the experimenter, andobeyed to the end.Any understanding of the phenomenon ofobedience must rest on an analysis of theparticular conditions in which it occurs. Thefollowing features of the experiment go somedistance in explaining the high amount ofobedience observed in the situation.1. The experiment is sponsored by andtakes place on the grounds of an institutionof unimpeachable reputation, Yale Univer-sity. It may be reasonably presumed thatthe personnel are competent and reputable.The importance of this background authorityis now being studied by conducting a seriesof experiments outside of New Haven, andwithout any visible ties to the university.2. The experiment is, on the face of it,designed to attain a worthy purpose—ad-vancement of knowledge about learning andmemory. Obedience occurs not as an end initself, but as an instrumental element in asituation that the subject construes as sig-nificant, and meaningful. He may not beable to see its full significance, but he mayproperly assume that the experimenter does.3. The subject perceives that the victimhas voluntarily submitted to the authoritysystem of the experimenter. He is not (atfirst) an unwilling captive impressed for in-voluntary service. He has taken the troubleto come to the laboratory presumably to aidthe experimental research. That he laterbecomes an involuntary subject does not alterthe fact that, initially, he consented to par-ticipate without qualification. Thus he hasin some dgeree incurred an obligation towardthe experimenter.4. The subject, too, has entered the experi-ment voluntarily, and perceives himself underobligation to aid the experimenter. He hasmade a commitment, and to disrupt theexperiment is a repudiation of this initialpromise of aid.5. Certain features of the procedurestrengthen the subjects sense of obligationto the experimenter, For one, he has beenpaid for coming to the laboratory. In partthis is canceled out by the experimentersstatement that:Of course, as in all experiments, the money is yourssimply for coming to the laboratory. From thispoint on, no matter what happens, the money isyours.46. From the subjects standpoint, the factthat he is the teacher and the other manthe learner is purely a chance consequence(it is determined by drawing lots) and he,the subject, ran the same risk as the otherman in being assigned the role of learner.Since the assignment of positions in theexperiment was achieved by fair means, thelearner is deprived of any basis of complainton this count. (A similar situation obtainsin Army units, in which—in the absence ofvolunteers—a particularly dangerous missionmay be assigned by drawing lots, and theunlucky soldier is expected to bear his mis-fortune with sportsmanship.)7. There is, at best, ambiguity with regardto the prerogatives of a psychologist and thecorresponding rights of his subject. There isa vagueness of expectation concerning what apsychologist may require of his subject, andwhen he is overstepping acceptable limits.4Forty-three subjects, undergraduates at YaleUniversity, were run in the experiment without pay-ment. The results are very similar to those obtainedwith paid subjects.
378 STANLEY MILGRAMMoreover, the experiment occurs in a closedsetting, and thus provides no opportunityfor the subject to remove these ambiguitiesby discussion with others. There are fewstandards that seem directly applicableto the situation, which is a novel one formost subjects.8. The subjects are assured that the shocksadministered to the subject are "painful butnot dangerous." Thus they assume that thediscomfort caused the victim is momentary,while the scientific gains resulting from theexperiment are enduring,9. Through Shock Level 20 the victimcontinues to provide answers on the signalbox. The subject may construe this as asign that the victim is still willing to "playthe game." It is only after Shock Level 20that the victim repudiates the rules com-pletely, refusing to answer further.These features help to explain the highamount of obedience obtained in thisexperiment. Many of the arguments raisedneed not remain matters of speculation, butcan be reduced to testable propostionsto be confirmed or disproved by furtherexperiments."The following features of the experimentconcern the nature of the conflict which thesubject faces.10. The subject is placed in a position inwhich he must respond to the competingdemands of two persons: the experimenterand the victim. The conflict must beresolved by meeting the demands of oneor the other; satisfaction of the victim andthe experimenter are mutually exclusive.Moreover, the resolution must take the formof a highly visible action, that of continu-ing to shock the victim or breaking off theexperiment. Thus the subject is forced intoa public conflict that does not permit anycompletely satisfactory solution.11. While the demands of the experimentercarry the weight of scientific authority, thedemands of the victim spring from his per-sonal experience of pain and suffering. Thetwo claims need not be regarded as equallypressing and legitimate. The experimenterseeks an abstract scientific datum; the victim5A series of recently completed experimentsemploying the obedience paradigm is reported inMilgram (1964).cries out for relief from physical sufferingcaused by the subjects actions.12. The experiment gives the subjectlittle time for reflection. The conflict comeson rapidly. It is only minutes after the sub-ject has been seated before the shock gen-erator that the victim begins his protests.Moreover, the subject perceives that he hasgone through but two-thirds of the shocklevels at the time the subjects first protestsare heard. Thus he understands that theconflict will have a persistent aspect to it,and may well become more intense as in-creasingly more powerful shocks are required.The rapidity with which the conflict descendson the subject, and his realization that it ispredictably recurrent may well be sources oftension to him.13. At a more general level, the conflictstems from the opposition of two deeplyingrained behavior dispositions: first, the dis-position not to harm other people, and sec-ond, the tendency to obey those whom weperceive to be legitimate authorities.REFERENCESADORNO, T., FRENKEL-BRUNSWIK, ELSE, LEVINSON,D. J., & SANTORD, R. N. The authoritarian person-ality. New York: Harper, 1950.ARENDT, H. What was authority? In C. J. Friedrich(Ed.), Authority. Cambridge: Harvard Univer.Press, 1958. Pp. 81-112.BINET, A. La suggestibility. Paris: Schleicher, 1900.Buss, A. H. The psychology of aggression. NewYork: Wiley, 1961.CARTWRIGHT, S. (Ed.) Studies in social power. AnnArbor: University of Michigan Institute for SocialResearch, 1959.CHARCOT, J. M. Oeuvres completes. Paris: Bureauxdu Progres Medical, 1881.FRANK, J. D. Experimental studies of personalpressure and resistance. /. gen. Psychol, 1944,30, 23-64.FRIEDRICH, C. J. (Ed.) Authority. Cambridge:Harvard Univer. Press, 19S8.MILGRAM, S. Dynamics of obedience. Washington:National Science Foundation, 25 January 1961.(Mimeo)MILGRAM, S. Some conditions of obedience anddisobedience to authority. Hum. Relat., 1964, inpress.ROKEACH, M. Authority, authoritarianism, and con-formity. In I. A. Berg & B. M. Bass (Eds.),Conformity and deviation. New York: Harper,1961. Pp. 230-257.SNOW, C. P. Either-or. Progressive, 1961(Feb.),24.WEBER, M. The theory of social and economicorganization. Oxford: Oxford Univer. Press, 1947.(Received July 27, 1962)