Obedience in Retrospect
Alan C. Elms
Abstract: Milgram's original paradigm for studying obedience to
authority is briefly described, and the main results are summarized.
Personal observations of the conduct of the initial studies give added
context for interpreting the results. Psychologists' reactions to the
Milgram experiments are discussed in terms of (1) rejecting the
research on ethical grounds, (2) explaining away the results as
expressions of trivial phenomena, (3) subsuming obedience to
destructive authority under other explanatory rubrics, and (4)
endorsing or rejecting the results in terms of their perceived social
relevance or irrelevance.
"The problem of obedience to authority may well be the crucial issue of
our time. The experiments you took part in represent the first efforts
to understand this phenomenon in an objective, scientific manner."
(Stanley Milgram, Report to Memory Project Subjects, 1962b)
Obedience to destructive authority was indeed a crucial social issue in
1962. The Holocaust had ended less than two decades earlier. Adolf
Eichmann recently had been sentenced to death for expediting it,
despite his plea than he had just been "following orders." American
military advisers were being ordered to Vietnam in increasing numbers
to forestall Communist control of Southeast Asia. Whether destructive
obedience could reasonably be described as the crucial issue of the
time is a judgment call; surely other issues offered competition for
that status. But there can be little argument that Stanley Milgram's
experiments were indeed "the first efforts to understand this
phenomenon in an objective, scientific manner."
Milgram was not seeking to develop a grand theory of obedience. His
main concern was with the phenomenon itself. He advised his graduate
students that as they began their own research, "First decide what
questions you want to answer." For him those first questions were
typically substantive, not theoretical. He also told his students he
sought to collect data that would still be of interest 100 years later,
whatever theoretical interpretations might be made of the data. For his
data on obedience, we are a third of the way through that 100 years.
Those data remain of high interest indeed, offering continual
challenges to our theories and to our confidence as psychologists that
we really understand important aspects of human social behavior.
Milgram eventually proposed his own theoretical interpretations. But
what most people still remember are the data themselves, the sheer
numbers of research volunteers who obeyed every order to the very end.
Before Milgram, creative writers had incorporated striking incidents of
obedience into novels, poems, and screenplays. Historians had written
factual accounts of remarkably obedient individuals and groups.
Psychologists had developed F- and other scales to measure inclinations
toward authoritarian tyranny and subservience. Milgram instead
established a realistic laboratory setting where actual obedience and
its circumstances might be closely studied.
The Obedience Paradigm
For those who have forgotten the details, and for the few who have
never read them, here is the basic situation that Milgram devised.
First, he advertised in the New Haven (Connecticut) daily newspaper and
through direct mail for volunteers for a study of memory and learning.
Volunteers were promised $4.00 for an hour of their time, plus 50 cents
carfare. (At the time, $4 was well above minimum wage for an hour of
work; 50 cents would have paid for a round-trip bus ride to and from
most areas of New Haven.) Most of those who volunteered were scheduled
by telephone to come at a given time to a laboratory on the Yale
In the basic experiments, two volunteers arrived at the laboratory at
about the same time. Both were invited into the lab by the
experimenter. The experimenter explained that one volunteer would be
assigned the role of teacher and the other would become the learner.
The teacher would administer an electric shock to the learner whenever
the learner made an error, and each additional shock would be 15 volts
higher than the previous one. By drawing slips of paper from a hat, one
volunteer became the teacher. His first task was to help strap the arms
of the other volunteer to the arms of a chair, so the electrodes from
the shock generator would not fall off accidentally. The teacher was
given a sample 45 volt electric shock from the shock generator, a level
strong enough to be distinctly unpleasant. Then the experimenter asked
the teacher to begin teaching the learner a list of word pairs. The
learner did fairly well at first, then began to make frequent errors.
Soon the teacher found himself administering higher and higher shock
levels, according to the experimenter's instructions. (Male pronouns
are used here because most volunteers were male; in only one
experimental condition out of 24 were female subjects used.)
After a few shocks the learner began to object to the procedure. After
more shocks and more objections, he loudly refused to participate
further in the learning task, and stopped responding. If the teacher
stopped giving him electric shocks at this point, the experimenter
ordered the teacher to continue, and to administer stronger and
stronger shocks for each failure to respond—all the way to the
end of the graded series of levers, whose final labels were "Intense
Shock," "Extreme Intensity Shock," "Danger: Severe Shock," and "XXX,"
along with voltage levels up to 450 volts. In the first experimental
condition, the teacher was separated from the learner by a soundproofed
wall; the learner could communicate his distress only by kicking on the
wall. In subsequent conditions, teachers could hear the learner's voice
through a speaker system, or sat near the learner in the same room
while the learning task proceeded, or sat next to the learner and had
to force his hand down onto a shock grid if he refused to accept the
Teachers were not told several important pieces of information until
their participation in the experiment was finished. Number one, the
experiment was a study of obedience to authority, not a study of memory
and learning. Number two, the volunteer who assumed the role of learner
was actually an experimental confederate. Number three, the only shock
that anyone ever got was the 45 volt sample shock given to each
teacher; the shock generator was not wired to give any shocks to the
learner. Number four, the learner's kicks against the wall, his
screams, his refusals to continue, were all carefully scripted and
rehearsed, as were the experimenter's orders to the teacher. A number
of variables could be (and were) added to the research design in
different conditions (see Miller, Collins, & Brief, this issue),
but these aspects were constant.
Observations from the Inside
The basic series of obedience experiments took place in the summer of
1961. Milgram was at that time a very junior assistant professor, 27
years old, with no professional publications yet in print. I had just
finished my first year of graduate school when he hired me to be his
research assistant for the summer. Stanley sent me a letter on June 27,
a week before I was scheduled to return to New Haven from a brief
"Matters have been proceeding apace on the project. The apparatus is
almost done and looks thoroughly professional; just a few small but
important pieces remain to be built. It may turn out that you will
build them, but that depends on factors at present unknown.
"The advertisement was placed in the New Haven Register and yielded a
disappointingly low response. There is no immediate crisis, however,
since we do have about 300 qualified applicants. But before long, in
your role of Solicitor General, you will have to think of ways to
deliver more people to the laboratory. This is a very important
practical aspect of the research. I will admit it bears some
resemblance to Mr. Eichmann's position, but you at least should have no
misconceptions of what we do with our daily quota. We give them a
chance to resist the commands of malevolent authority and assert their
alliance with morality.
". . . . The goal this summer is to run from 250-300 subjects in nine
or ten experimental conditions. Only if this is accomplished can the
summer be considered a success. Let me know if there is something I
The summer was a success by any reasonable standards, if not fully by
Milgram's. He had not overlooked anything procedural; even at that
early stage in his career, he was already the most well-organized
researcher I have ever encountered. But he had hardly come close to
anticipating the degree to which his subjects would yield to the
commands of malevolent authority, or how readily they would abrogate
their alliance with morality. Milgram knew he would get some obedience;
in a pilot study the previous winter, he had found Yale undergraduates
disturbingly willing to shock their victims. But he recognized that
Yale undergraduates were a special sample in many ways; that the
prototype shock generator was rather crude and perhaps not altogether
convincing; and that the simulated victim's displays of pain were
fairly easy to ignore. For the main experiments, Milgram auditioned and
rehearsed a victim whose cries of agony were truly piercing. He
recruited a larger and diverse sample of nonstudent adults from the New
Haven area, ranging from blue-collar workers to professionals and from
20 to 50 years in age. He constructed a professional-looking shock
generator and purchased other high-quality equipment, including a
20-pen Esterline Angus Event Recorder that registered the duration and
latency of each "shock" administration to the nearest hundredth of a
second. He had decided that his main dependent variable would be the
mean shock level at which subjects refused to go further in each
experimental condition, but he wanted to be able to examine more subtle
differences in their performance as well.
In early August the curtains went up on the first official obedience
experiment. (More accurately, the curtains were drawn aside; Yale's new
Social Interaction Laboratory, on temporary loan from the Sociology
Department, was enclosed by two-way mirrors and heavy soundproofing
curtains.) Would subjects be convinced of the reality of the
learning-and-memory experiment, the shock generator, the victim's
suffering? They were. Would subjects obey the experimenter? They did.
How far would they go? On and on up the sequence of shock levels. Would
any subjects go all the way to the end of the shock board'? Yes indeed.
Behind the two-way mirrors, Stanley Milgram and I (as well as
occasional visitors) watched each early subject with fascination and
with our own share of tension. Stanley had made broad predictions
concerning the relative amounts of obedience in different conditions,
but we paid little attention to the gradual confirmation of those
predictions. Instead we tried to predict the behavior of each new
subject, based on his initial demeanor and the little we knew about his
background. We were gratified when any subject resisted authority.
Sometimes it was quiet resistance, sometimes noisy, but it was exciting
each time it happened. As more and more subjects obeyed every command,
we felt at first dismayed, then cynically confirmed in our bleakest
views of humanity. We were distressed when some volunteers wept,
appalled when others laughed as they administered shock after shock.
The experimenter gave each subject a standard debriefing at the end of
the hour, to minimize any continuing stress and to show that the
"victim" had not been injured by the "shocks." When a subject appeared
especially stressed, Milgram often moved out from behind the curtains
to do an especially thorough job of reassurance and stress reduction.
When a subject did something truly unexpected during the
experiment—an especially resolute show of resistance, for
instance, or a long laughing jag—Milgram would join the
experimenter in giving the subject a detailed cross-examination about
why he had displayed such behavior. For us as well as for the subjects,
the situation quickly became more than an artificially structured
experiment. Instead it presented slice after slice of real life, with
moral decisions made and unmade every evening.
The Most Prominent Results
As matters turned out, Milgram did not need equipment sensitive enough
to measure shock intervals in hundredths of a second. By the end of the
second run of 40 subjects, if not before, his main dependent variable
had become simply the percentage of subjects who obeyed the
experimenter's commands all the way to the end of the shock series,
contrasted with the percentage who disobeyed by quitting at any point
in the whole long sequence of shock levels. In the first condition, a
substantial majority of subjects (26 out of 40, or 65%) obeyed
completely. That was the condition with minimal feedback from the
learner—a few vigorous kicks on the wall. But wouldn't obedience
drop substantially if the teacher could actually hear the learner
screaming and demanding to be set free? It didn't. Twenty-five out of
40 were fully obedient in this second condition. Even when Milgram
tried to encourage disobedience by having the learner claim a
preexisting heart condition ("It's bothering me now!"), obedience
remained at a high level: 26 of 40 subjects again (Milgram, 1974, pp.
56-57). Putting the victim in the same room and near the teacher
reduced obedience somewhat, but 40% still obeyed fully. Indeed, even
when teachers were ordered to press the hand of the screaming victim
down onto a shock plate to complete the electrical circuit, a majority
did so at least twice before quitting, and 30% obeyed in this fashion
to the end of the shock board (Milgram, 1974, p. 35).
Milgram ran more than 700 subjects through various obedience
conditions in less than a year. (The National Science Foundation, which
financed the research, got its money's worth from two grants totaling
about $60,000.) Each subject was run through the procedure
individually, then was subjected to both immediate and follow-up
questionnaires of various kinds. Milgram looked at the effects not only
of the victim's physical proximity to the subject but of the
experimenter's proximity, the amount of group support either for
obedience or for defiance, and the learning experiment's apparent
institutional backing. He made a variety of interesting
findings—enough to fill a book, and more. But the data that
carried the greatest impact, on other psychologists and on the general
public, came from those first few experimental conditions: two-thirds
of a sample of average Americans were willing to shock an innocent
victim until the poor man was screaming for his life, and to go on
shocking him well after he had lapsed into a perhaps unconscious
silence, all at the command of a single experimenter with no apparent
means of enforcing his orders.
Reactions to the Research
Once these data appeared in professional psychological journals (after
initial resistance from editors), they were rather quickly disseminated
through newspaper and magazine stories, editorials, sermons, and other
popular media. With few exceptions, the nonprofessional citations of
the experiments emphasized their social relevance: Milgram had revealed
in ordinary Americans the potential for behavior comparable to that of
the Nazis during the European Holocaust. (According to a TV Guide ad
for a docudrama with William Shatner as a fictionalized Milgram, the
research revealed "A world of evil so terrifying no one dares penetrate
its secret. Until now!" [August 21, 1976, p. A-86.])
Psychologists responded in more diverse ways. Authors eager to enliven
their introductory and social psychology textbooks soon made the
obedience experiments a staple ingredient (see Miller, this issue).
Other psychologists seemed to regard Milgram's results as a challenge
of one sort or another: conceptual, ethical, theoretical, political.
The obedience studies were related, historically and procedurally, to
earlier studies of social influence, but they did not fit readily into
current theoretical models or research trends. Because of their rapidly
achieved visibility inside and outside the field, they were soon
treated as fair game for elucidation or attack by psychologists with a
multitude of orientations.
One type of response to the disturbing results of the obedience studies
was to shift attention from the amounts of obedience Milgram obtained
to the ethics of putting subjects through such a stressful experience.
The first substantial published critique of Milgram's studies focused
on the presumed psychic damage wreaked on his subjects by their ordeal
(Baumrind, 1964). Milgram was not altogether surprised by such
criticism; similar concerns had been expressed by several Yale faculty
members during or soon after the experiments, and ethical questions had
been raised about the research when Milgram first applied for American
Psychological Association membership. But he was disappointed that his
critics did not recognize the care he had put into responding to his
subjects' high stress levels immediately after their participation, as
well as into checking on any lingering effects over time (Milgram,
1964). Milgram was a pioneer in the debriefing procedures that are now
a matter of course in psychological experiments on human
subjects—debriefing in the sense not only of questioning the
subject about his or her perception of the experiment, but of providing
the subject with information and encouragement that will counteract any
reactions to participation that might damage the subject's self-esteem.
As Milgram told me later,
"My membership application to APA was held up for one year while they
investigated the ethicality of the obedience experiment. In the end,
they gave me a clean bill of health and admitted me to membership.
Whenever any group has seriously considered the merits and problems of
the experiment, they have concluded that it was an ethical experiment.
Nonetheless, isolated individuals still feel strongly enough to attack
it." (Personal communication, July 3, 1969)
One consequence of those individual attacks was a set of stringent
federal regulations that made it virtually impossible ever again to
conduct a close replication of the Milgram studies at any U.S.
educational or research institution.
Many social scientists who have considered the ethics of the obedience
studies in print have taken a neutral position or have come down on the
side of Milgram. But outside the field, a similar perception of
appropriate research and debriefing procedures is not widespread. When
I participated in a conference on social science research ethics at the
Kennedy Institute of Ethics 18 years after the obedience research was
completed, several philosophers and professional ethicists devoted a
large part of their energies to what struck me as rather crude Milgram
bashing. The research scientists at the conference were not so
inclined, but they had to work hard to communicate the virtues of a set
of studies that had raised important issues about both the bad and the
good in human nature (Beauchamp, Faden, Wallace, & Walters, 1982).
Questions of Belief
Among other early commentaries on the research, several psychologists
argued that the results were not credible because the subjects did not
believe they were actually harming the victim (e.g., Orne &
Holland, 1968). Milgram's own data, showing that during the experiment
a very high percentage of subjects believed the victim was receiving
extremely painful shocks (1974, pp. 171-174), were ignored or dismissed
as attempts by the subjects to give Milgram the answers he wanted.
Researchers' descriptions of many subjects' visible signs of high
stress were also ignored, or were assumed to be evidence merely of the
subjects' enthusiastic play acting. Even a filmed record of several
actual subjects (Milgram, 1965a), displaying either great stress or
extraordinary improvisational acting ability, did not convince
psychologists who took this dismissive position. Some critics may have
assumed that the four subjects shown at length in the film, plus
several others who appeared more briefly, were the most convincingly
emotional subjects Milgram could find among his thousand participants.
In fact, Milgram chose all of them from the 14 subjects who happened to
be "selected in the normal manner for recruitment" during the two days
he brought movie cameras to the laboratory (Milgram, 1965c, p. 5).
Many social psychologists have accepted the ethical appropriateness of
Milgram's procedures and the believability of the experimental context.
Even they, however, have often redirected attention away from the
specific phenomenon of destructive obedience by subsuming it under a
broader theoretical approach or alternative hypothetical constructs.
Milgram was slow to offer a comprehensive theoretical account of his
own. His definitions of obedience to authority, from his first to his
final writings on the subject, drew upon no theoretical assumptions.
Rather, they were commonsense or dictionary definitions: "Every power
system implies a structure of command and action in response to the
command" (Milgram, 1961, p. 2); "If Y follows the command of X we shall
say that he has obeyed X; if he fails to carry out the command of X, we
shall say that he has disobeyed X" (Milgram, 1965b, p. 58); "[I]t is
only the man dwelling in isolation who is not forced to respond,
through defiance or submission, to the commands of others" (Milgram,
1974, p. 1). In his grant proposals he referred to "internal
restraints" or "internal resistances" that were pitted against the
acceptance of authoritative commands, but he did not specify the nature
of these internal processes (Milgram, 1961, p. 3; Milgram, 1962a, p.
1). He raised the possibility of predispositional factors and of
"highly complex, and possibly, idiosyncratic motive structures" (1962a,
p. 17), but in the research itself he directed his efforts mainly
toward identifying situational factors that increased or decreased
obedience. In his most extensive early discussion of his results
(Milgram, 1965b, largely written in 1962), he cited such midlevel
hypothetical constructs as "empathic cues," "denial and narrowing of
the cognitive field," and a varying "sense of relatedness between his
[the subject's] own actions and the consequences of those actions for
the victim" (pp. 61-63; his italics).
Though it took Milgram less than a year to run all his
subjects and not much longer than that to write several papers on the
results, he worked on his book about obedience for over five years. He
attributed the slowness of the book's writing in part to his becoming
engaged in other sorts of research. But much of his struggle with the
book appears to have centered on the difficulty of developing a general
theory of obedience. The principal theoretical concepts he advanced in
the book, including the agentic state (Milgram, 1974, pp. 133-134) and
the evolution of a potential for obedience in humans (pp. 123-125),
impressed many readers rather less than the results themselves--a
reaction that both frustrated and pleased the data-centric Milgram.
Though he had collected demographic information on all participants and
had supported my collection of personality data from subsamples of
obedient and disobedient subjects (Elms & Milgram, 1965), he gave
short shrift to such data in his book, concluding that "It is hard to
relate performance to personality because we really do not know very
much about how to measure personality" (p. 205).
Others have usefully discussed the interaction of personality and
situational variables in the obedience situations (e.g., Blass, 1991).
A majority of the alternative explanations, however, have stressed
cognitive processes, emphasizing ways in which the subject processed
information about the situation that might have justified his obedience
or strengthened his resistance. Milgram viewed such alternative
explanations with interest, but took steps to rule out certain of them
experimentally. One of the most obvious of these alternatives was the
idea that subjects might be so awed by Yale University and so certain
of its virtue that they would do anything they were told within those
august halls, regardless of any general proclivity toward destructive
obedience. Even before this environment-based explanation of his
subjects' obedience was first offered in print, Milgram had largely
vitiated it by moving the experiments from the awe-inspiring
Interaction Laboratory to a rather less impressive basement facility
and then to the intentionally unimpressive office of a fly-by-night
company in industrial Bridgeport, Connecticut. He got essentially the
same results in all three locations. A number of alternative or
additional explanations of Milgram's results remain as operable
hypotheses, but none has decisively carried the day. Their very
diversity ensures that the larger audience for the research will
continue to be concerned primarily about the subjects' disturbing
behavior rather than about the internal processes that may have
The Question of Relevance
Finally among ways in which psychologists have responded to Milgram's
findings are arguments concerning the social relevance of the
experiments. Many psychologists, at least in their textbooks, have
embraced his findings as being highly relevant to important social
phenomena, including destructive obedience not only in totalitarian
states but among American soldiers, Bosnian combatants, and suicidal
religious cults. But others (including some who also argued that the
research was unethical or experientially unconvincing) have denied any
real social relevance. Even if subjects believed they were really
shocking the victim, these psychologists say, they knew the situation
must not be as bad as it appeared, because somebody would have stopped
them if it was. Or the subjects were in a situation where the
experimenter accepted responsibility for the effects of their behavior,
so their behavior is not really relevant to real-world situations where
blame is less readily transferred to another individual. Or some other
rationale is advanced, presumably peculiar to the Milgram obedience
situation, that somehow does not translate into real-world social
dynamics. Milgram rightly dismissed all such explanations that had been
advanced up to the time of his final writings, and very likely would
have dismissed all subsequent ones, for two simple reasons: Any
effective authority figure in the real world always finds ways to
justify imposing his or her will on underlings. The underlings who obey
authoritative commands in the real world always find rationales for
their obedience. In most prominent real-world cases of destructive
obedience that have been compared (or discompared) to the Milgram
studies, the authorities were able to call upon a social rationale for
their commands that was at least as strong as or stronger than that
available to any psychological experimenter. In addition, they were
often able to promise their followers much greater rewards for
obedience and punishments for disobedience.
Stanley Milgram's research on obedience tapped into psychological
processes that ranked as neither new nor extreme in the history of
human behavior. A "crucial issue of our time," perhaps the crucial
issue, obedience unfortunately remains. Though Milgram was proud that
his studies were "the first efforts to understand this phenomenon in an
objective, scientific manner," he did not want them to be the last.
This issue of the Journal of Social Issues gives strong evidence that
the efforts of other researchers to expand upon his groundbreaking work
will continue unabated.
Baumrind, D. (1964). Some thoughts on ethics of research: After reading
Milgram's "Behavioral Study of Obedience." American Psychologist, 19,
Beauchamp, T. L., Faden, R. R., Wallace, R. J., Jr., & Walters, L.
(Eds.). (1982). Ethical issues in social science research. Baltimore,
MD: Johns Hopkins University Press.
Blass, T. (1991). Understanding behavior in the Milgram obedience
experiment: The role of personality, situations, and their
interactions. Journal of Personality and Social Psychology, 60, 398-413.
Elms, A. C., & Milgram. S. (1965). Personality characteristics
associated with obedience and defiance toward authoritative command.
Journal of Experimental Research in Personality, 1, 282-289.
Milgram. S. (1961). Dynamics of obedience: Experiments in social
psychology. Application for National Science Foundation research grant,
Milgram, S. (1962a). Obedience to authority: Experiments in social
psychology. Application for National Science Foundation grant renewal,
Milgram, S. (1962b). Report to Memory Project subjects. Unpublished manuscript, Yale University.
Milgram, S. (1964). Issues in the study of obedience: A reply to Baumrind. American Psychologist, 19, 848-852.
Milgram, S. (1965a). Obedience [Film]. (Available from the Pennsylvania State University Audiovisual Services.)
Milgram, S. (1965b). Some conditions of obedience and disobedience to authority. Human Relations, 18, 57-76.
Milgram, S. (1965c). Study notes for “Obedience.” (Distributed by the New York University Film Library.)
Milgram, S. (1974). Obedience to authority. New York: Harper & Row.
Orne, M. T., & Holland, C. C. (1968). On the ecological validity of
laboratory deceptions. International Journal of Psychiatry, 6, 282-293.
Originally published in the Journal of Social Issues, Vol. 51, No. 3, 1995, pp. 21-31.
Quotations from unpublished correspondence of Stanley Milgram are used by permission of Alexandra Milgram.
Correspondence regarding this article should be addressed to Alan C.
Elms, Department of Psychology, University of California, Davis,
ALAN C. ELMS, while a graduate student at Yale
University, worked with Stanley Milgram on the first obedience studies
and earned his Ph.D. under the direction of Irving L. Janis. Dr. Elms
did laboratory studies of role-play induced attitude change and
interview studies of right- and left-wing political activists before he
focused his work on psychobiography. He has written Social Psychology
and Social Relevance (1972), Personality in Politics (1976), and
Uncovering Lives: The Uneasy Alliance of Biography and Psychology
(1994). He has taught at Southern Methodist University, has been a
visiting scholar at Trinity College, Dublin, and at Harvard University,
and has been a faculty member at the University of California, Davis,