Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Rethinking the psychology of tyranny:
The BBC prison study
Stephen Reicher1
* and S. Alexander Haslam2
**
1
University of St. Andrews
2
University of Exeter
This paper presents findings from the British Broadcasting Corporation (BBC) prison
study – an experimental case study that examined the consequences of randomly
dividing men into groups of prisoners and guards within a specially constructed
institution over a period of 8 days. Unlike the prisoners, the guards failed to identify
with their role. This made the guards reluctant to impose their authority and they were
eventually overcome by the prisoners. Participants then established an egalitarian social
system. When this proved unsustainable, moves to impose a tyrannical regime met with
weakening resistance. Empirical and theoretical analysis addresses the conditions under
which people identify with the groups to which they are assigned and the social,
organizational, and clinical consequences of either doing so or failing to do so. On the
basis of these findings, a new framework for understanding tyranny is outlined. This
suggests that it is powerlessness and the failure of groups that makes tyranny
psychologically acceptable.
In the introduction to his text on The Roots of Evil, Staub writes: ‘the widespread hope
and belief that human beings had become increasingly ‘civilized’ was shattered by the
events of the Second World War, particularly the systematic, deliberate extermination of
six million Jews by Hitler’s Third Reich’ (1989, p. 3). The impact of this realization was
as marked in academia, and more particularly within academic psychology, as it was in
society at large. Indeed, it is arguable that the shadow of the Holocaust lies over the last
half century of social psychology and, either indirectly or directly, informs many of the
core issues that are of concern to the discipline’s practitioners: questions such as how
we come to hate and to discriminate against members of other groups (Tajfel, Flament,
Billig, & Bundy, 1971), how people come to see others as less human and less deserving
than themselves (Leyens et al., 2003), how the seeds of authoritarianism, social
dominance, and power abuse are sown and cultivated (Adorno, Frenkel-Brunswik,
Levinson, & Sanford, 1950), and – the question which concerns us most directly in this
* Correspondence should be addressed to Stephen Reicher, School of Psychology, University of St. Andrews, St. Andrews, Fife,
KY16 9JU, UK (e-mail: sdr@st-andrews.ac.uk).
**Correspondence should be addressed to Alex Haslam, School of Psychology, University of Exeter, Exeter, Devon, EX4 4QG,
UK (e-mail: A.Haslam@exeter.ac.uk).
The
British
Psychological
Society
1
British Journal of Social Psychology (2006), 45, 1–40
q 2006 The British Psychological Society
www.bpsjournals.co.uk
DOI:10.1348/014466605X48998
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
paper – how we come to condone the tyranny of others or else act tyrannically
ourselves.
While the literature in these various areas is both broad and varied, it is possible to
identify at least one major trend. That is, there has been a shift away from explanations
that focus on the individual characteristics of those who are prejudiced, discriminatory,
or even genocidal, towards those that concentrate on the nature of group processes
which can induce the most inoffensive of individuals to commit the most offensive of
acts (Billig, 1978; Brown, 1965; Milgram, 1974; Sherif, 1966; Tajfel, 1982). In many
cases, and certainly when it comes to the psychology of tyranny, theorists have taken the
argument one step further and proposed not only that extreme antisocial behaviour
must be analysed at the group level, but also that group psychology necessarily tends in
the direction of extreme antisocial behaviour. While we strongly endorse the need for a
group-level psychology of tyranny (which we define as an unequal social system
involving the arbitrary or oppressive use of power by one group or its agents over
another), we will take issue with the notion that groups per se are the root of the
problem. Indeed, we will argue that powerful and effective groups provide an effective
psychological bulwark against tyranny and that it is when groups prove ineffective that
tyrannical forms of social organization begin to become attractive.
The equation of groups and tyranny has a long history both within social thought and
within social psychology. Hannah Arendt (1998) describes the classical view that ‘the
rule by many is not good’ and traces this back to Aristotle’s contention that collective
rule leads to haphazardness, moral irresponsibility, and is but a disguised form of
tyranny. More recently, such ideas were given substance by crowd psychologists such as
Gustave LeBon (1895/1947) who argued that, through submergence in the crowd,
individuals lose their individual identity and their sense of responsibility and hence
become capable of barbaric and atavistic acts. The notion of submergence was directly
transposed into the modern social psychological concept of de-individuation, which is
seen to arise from anonymity within a group (Postmes & Spears, 1998; Reicher, Spears,
& Postmes, 1995). As one influential de-individuation researcher has put it, ‘mythically,
deindividuation is the ageless life force, the cycle of nature, the blood ties, the tribe, the
female principle, the irrational, the impulsive, the anonymous chorus, the vengeful
furies.’ (Zimbardo, 1969, p. 249).
Tyranny as role and power
Though well known as a de-individuation theorist, Zimbardo is better known for his
work on the Stanford prison experiment (SPE; Haney, Banks, & Zimbardo, 1973;
Zimbardo, 1989; Zimbardo, Maslach, & Haney, 1999). Indeed, this is one of the most
famous social psychological investigations ever conducted, representing the culmination
of a series of ‘classic’ field studies into the roots of extreme behaviours that were
conducted in the aftermath of World War Two (Milgram, 1963; Sherif, 1956). Building on
earlier studies, it played a critical part in cementing the shift that we have described
from individual to group-level explanations of extreme behaviours (Banuazizi &
Movahedi, 1975). Moreover, it was one of the few studies that not only addressed the
issue of tyranny but, due to the power of the research paradigm, also produced direct
evidence of tyrannical behaviour. While it is not, strictly speaking, a study of
de-individuation, Zimbardo certainly used his general understanding of the group as a
corrosive force to explain events in the SPE.
Stephen Reicher and S. Alexander Haslam2
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
The SPE is remembered for showing that, simply as a consequence of assigning
college students the role of guard or prisoner, the former became increasingly brutal
while the latter became passive and began to show signs of psychological disturbance.
Such was the severity of these phenomena that the study, originally scheduled to last 2
weeks, had to be stopped after 6 days. Zimbardo and colleagues explained their findings
by commenting that guard aggression ‘was emitted simply as a ‘natural’ consequence of
being in the uniform of a ‘guard’ and asserting the power inherent in that role’ (Haney
et al., 1973, p. 12). Thus, immersion in a group is seen to undermine the constraints that
normally operate upon people when they act as individuals. In addition, when those
groups have power at their disposal, this is believed to encourage extreme antisocial
behaviour (Zimbardo, 1969).
Although these findings were significant in their own right, the impact of the SPE
was as much ethical as theoretical (e.g. see Smith & Mackie, 2000, p. 49). Indeed, the
very extremity of the results led many (including Zimbardo himself) to question the
legitimacy of subjecting participants to such situations. The acceptability of conducting
any sort of large-scale field interventions thus became a focus for vigorous debate
(e.g. Herrera, 1997; Lindsay & Adair, 1990; Sieber, Iannuzzo, & Rodriguez, 1995; Smith &
Richardson, 1983). Paradoxically, then, at the same time that the SPE marked the
culmination of post-war field studies, it also led to their cessation.
Accordingly, since the 1970s, social psychology has been increasingly dominated by
laboratory experiments in which there is minimal or no interaction between
participants and scant attention paid to the role of personal and group history or to
the development of interactions over time (Bar Tal, 2004; Doosje, Spears, & Ellemers,
2002; Haslam & McGarty, 2001; Levine, 2003; Moreland, Argote, & Krishnan, 1996).
Moreover, this unwillingness to undertake studies that create, manipulate and
systematically investigate the effects of social environments on human interaction can
be seen to have contributed to the increasing dominance of explanations based upon
inherent and essentially unavoidable genetic, biological, or psychological propensities.
It has also led to an increasing disjunction between the issues that motivate social
psychological studies and the nature of those studies themselves. Research reports (and
certainly most bids for research funding) typically start by alluding to large-scale topics
such as oppression, discrimination, and genocide, but then go on to pursue an empirical
strategy that seems very remote from the social realities of such phenomena (e.g.
seeking to explain these phenomena in terms of individual-level subconscious processes
from a cognitive or, more recently, neuroscientific perspective; Ito, Thompson, &
Cacioppo, 2004). As Zimbardo (quoted in Brockes, 2001, p. 2) has argued, partly as a
result of these trends, psychology has become increasingly marginal to, and
marginalized from, debate surrounding important social issues.
In terms of the specific issue of tyranny, the ethical concerns that have placed the
SPE ‘off-limits’ (with the exception of a partial replication by Lovibond, Mithiran, &
Adams, 1979) have led to a situation in which the conclusions of that study have become
almost inviolate and social psychological inquiry into tyranny has effectively ground to a
halt. Barred from employing the power of the SPE paradigm, it is all but impossible to
produce behaviours that are powerful enough to match those found by Zimbardo and
his colleagues. Hence, even if researchers harbour doubts about the extreme situational
determinism and negative views of the social group, which are used to explain these
findings (and many do; e.g. see Ellemers, Spears, & Doosje, 1999; Turner, Hogg, Oakes,
Reicher, & Wetherell, 1987; Haslam, 2001), it has not been possible to produce data that
The psychology of tyranny 3
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
can give substance to those doubts and hence reopen scientific debate about the
psychological bases of tyranny.
However, quite apart from the intrinsic importance of the topic, there are at least
two sets of reasons why revisiting the SPE is long overdue. First, any assessment of the
conclusions drawn from the SPE is inevitably limited by the fact that only a small
proportion of the interactions in the study were recorded (because filming was intrusive
and limited) and, of these, only a very small number are in the public domain. Moreover,
observational data were never complemented by other data sources that would allow for
controlled measurement of key behaviours and the psychological states seen to underlie
them. At the very least, there is a need for a fuller and more transparent data set, which
might progress empirically grounded and open debate about the psychological bases of
tyranny.
However, even the limited amount of data that is available from the SPE casts doubt
on the analytic conclusions that have been drawn from it. Where participants did
behave in role, it is unclear whether, as Zimbardo and his colleagues claim, this was due
to their ‘natural’ acceptance of role requirements or due to the leadership provided by
the experimenters (Baron, 1984; Banuazizi & Movahedi, 1975). This is because during
the study, the guards were given clear guidance as to how they should behave. Notably,
when Zimbardo briefed his guards, he told them:
You can create in the prisoners feelings of boredom, a sense of fear to some degree, you can
create a notion of arbitrariness that their life is totally controlled by us, by the system, you,
me – and they’ll have no privacy. They’ll have no freedom of action, they can do nothing, say
nothing that we don’t permit. We’re going to take away their individuality in various ways.
In general what all this leads to is a sense of powerlessness (Zimbardo, 1989).
The importance of such guidance is demonstrated by the research of Lovibond et al.
(1979) who conducted a study in which the guards were trained to respect the
prisoners as individuals and to include them in decision-making processes. Under these
conditions, the ensuing behaviour of both guards and prisoners was far less aggressive
and extreme (Lovibond et al., 1979).
Yet, even with guidance, many of the participants in the SPE behaved out of role for
much of the time (Baron, 1984). The available video material shows that both prisoners
and guards challenged their roles not only at the start, but throughout the entire study.
In the case of the guards, Zimbardo (1989) notes that, while some exploited their
power, others sided with the prisoners and yet others were tough but fair. Such diversity
sits uneasily with the notion that role acceptance is simply determined by the situation.
It suggests that the emphasis on role acceptance and tyranny is one-sided and that there
is a need to focus on (a) the conditions under which people do or do not assume their
roles and (b) the balance between tyranny and resistance.
An alternative analysis: The social identity approach
It is not only that some of the data from the SPE appear to sit uneasily with a role
account. Increasingly, the role account – and indeed the generally negative view that
group membership leads to a loss of constraints on antisocial behaviour – is at odds with
developments in group psychology. One of the most influential of these is the social
identity approach (incorporating social identity theory; Tajfel, 1978, 1982; Tajfel &
Turner, 1979; and self-categorization theory; Turner, 1985; Turner et al., 1987; Turner,
Oakes, Haslam, & McGarty, 1994). According to this approach, people do not
Stephen Reicher and S. Alexander Haslam4
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
automatically act in terms of group memberships (or roles) ascribed by others. Rather,
whether or not they do so depends upon whether they internalize such memberships as
part of the self-concept (Turner, 1982).
Self-categorization theory in particular has argued that this act of self-definition in
terms of group membership (social identification) forms the psychological basis of
group behaviour and that the character of such behaviour depends upon the norms,
values, and understandings that characterize the particular category in question (Turner,
1982, 1999). Thus, while members of certain groups may indeed use their power to act
in discriminatory and oppressive ways, members of other groups may act more
prosocially and use their power for constructive purposes (Pfeffer, 1981; Postmes &
Spears, 1998). Moreover, even if some groups are tyrannical, group action is also the
basis on which people gain the strength and confidence to resist, to challenge, and even
to overthrow tyranny (Reicher, 1996; Tajfel, 1978).
Consistent with such emphasis, the greater part of early work informed by social
identity theory has focused on the conditions under which people act to change
inequalities between groups (e.g. Robinson, 1996). In broad terms, it is assumed that
people who are positively valued by virtue of their group membership (e.g. members of
dominant groups) would identify with and act in terms of the group. For people who are
negatively valued by virtue of their group membership (e.g. members of subordinate
groups), collective action is contingent upon two sets of factors in particular (Tajfel,
1978; Tajfel & Turner, 1979). The first relates to beliefs about one’s ability to advance
through the social system despite one’s group membership (i.e. the permeability of
category boundaries). The second concerns the perceived security of intergroup
relations and comprises two further elements: the perceived fairness of intergroup
inequalities (legitimacy) and their perceived stability. When relations are perceived to
be insecure, this is characterized by the fact that individuals are aware of cognitive
alternatives to the status quo and hence can envisage specific ways in which it could be
changed.
Permeability affects whether people act individually or collectively, so that a belief
that movement across boundaries is possible encourages strategies of individual
mobility, but a belief that such movement is impossible encourages people to perceive
themselves and act as group members (e.g. Ellemers, 1993; Wright, Taylor, &
Moghaddam, 1990). Whether or not people then challenge inequality is also dependent
upon intergroup relations being perceived as insecure. That is, people should be most
inclined to resist domination when they see inequality as both illegitimate and unstable
and can thus envisage cognitive alternatives to it (Turner & Brown, 1978; see also
Ellemers, Wilke, & van Knippenberg, 1993; Tajfel, 1978; Tajfel & Turner, 1979).
It is important to stress that social identity theory does not constitute a
comprehensive theory of domination and resistance. Most notably, it has little to say
about the concomitants of identity processes (e.g. organizational and clinical factors),
which may impact upon the ability of group members to act effectively. These are
critical issues that we aim to investigate here. Nonetheless, the social identity approach
provides a well-articulated and contemporary perspective from which to revisit the
issues raised by the SPE: What are the psychological consequences of intergroup
inequality? When do people seek to impose such inequality? And when do they resist it?
Before explaining how we addressed these issues, it is important to consider a
second set of reasons for revisiting the SPE. These have less to do with the explanation
of the findings themselves than with their broader social relevance. For Zimbardo and
his colleagues, the results of the SPE were intended to bear directly on the nature of
The psychology of tyranny 5
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
prison regimes in the United States. Thus, they refer to the setting as a ‘simulated prison’
and claim that ‘this simulated prison environment developed into a psychologically
compelling prison environment’ (Haney et al., 1973, p. 69). However, as Banuazizi and
Movahedi (1975) point out, there are good reasons to doubt these claims, firstly, because
there are critical features of the SPE that are very different from a real prison (e.g.
participants know they have committed no crime and can ask to leave at any time) and,
secondly, because even where there are material similarities, the two are very different
phenomenologically (e.g. where the walls of a prison remind the inmates that they must
be kept apart from ‘decent’ people, they remind the participants in the SPE that they are
honourable participants in adventurous scientific research).
While these arguments are in themselves controversial (e.g. Thayer & Saarni, 1975),
our point is that, even if they are accepted, they do not render the SPE practically
irrelevant. For there is another level beyond that of phenomenal equivalence through
which the study can (and is) claimed to have real-world implications. That is, as is the
case with most psychological research, generalization is theoretically, not empirically,
based (Haslam & McGarty, 2004; Turner, 1981). Thus, Zimbardo (e.g. 2001) uses his
study to establish the theoretical claim that people ‘naturally’ assume roles, and then
uses this theoretical analysis to explain a wide range of phenomena from prison
behaviour to terrorism. Of course, the resultant behaviours will not be the same as those
observed in the SPE, but will depend upon the precise nature of the roles and role
requirements in the relevant domain of application.
It is precisely this ability to apply role theory to a broad range of domains, despite
phenomenal and behavioural differences, that has ensured the impact of the SPE within
and beyond psychology amongst those who have little interest in matters relating to
prisons. It is the validity of this theoretical account that we wish to address in the
present paper, not only for the empirical and theoretical reasons discussed above, but
also because we feel that the traditional analysis of the SPE has profound and troubling
social implications. If people cannot help but act in terms of assigned role, it implies that
they have little choice, and hence little responsibility, for their social actions. This makes
it more difficult to hold tyrants to account for what they do. Moreover, in
communicating the message that resistance is futile, the analysis discourages the
oppressed from attempting to challenge tyranny.
The BBC prison study
In December 2001, we conducted a major social psychological field study (possibly the
largest such study in the three decades since the SPE) in conjunction with the
documentaries unit of British Broadcasting Corporation (BBC; hence the title ‘the BBC
prison study’). The study was independently designed, operationalized, run, and
analysed by the authors. In this sense, it was an ordinary piece of scientific research,
which went through all the normal scientific procedures (including ethical approval),
and whose features were designed in relationship to the theoretical issues that
concerned us. The contribution of the BBC was to coordinate and manage the logistical
task of (a) creating the study environment (in line with our guidelines), (b) filming the
study, and (c) preparing some of the resultant material for broadcast. In short, the
project can be described as ‘original science filmed’. This made it a unique collaboration
that was markedly distinct from ‘reality television’ enterprises in which a television
company devises certain scenarios with issues of entertainment in mind and then invites
academics to comment on them.
Stephen Reicher and S. Alexander Haslam6
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Over a period of 8 days, the study examined the behaviour of 15 men who were
placed in a social hierarchy of guards and prisoners within a purpose-built environment.
Their behaviour was video- and audio-recorded over the entire period, and this was
complemented by daily psychometric and physiological measures. The video data were
edited into four 1-hour long documentaries screened in May 2002 (Koppel & Mirsky,
2002).
The aim of the study was not to simulate a prison (which, as in the SPE, would have
been impossible on ethical and practical grounds) but rather to create an institution that
in many ways resembled a prison (but also other hierarchical institutions such as a school,
an office, a barracks; see Morgan, 1979) as a site toinvestigate the behaviourof groupsthat
were unequal in terms of power, status, and resources. What is critical, then, is not that
the study environment replicated a real prison (which no such environment ever could),
but that it created inequalities between groups that were real to the participants.
Similarly, our aim was not conduct an exact replication of the SPE (which, would also
have been impossible for ethical reasons). Rather, it was to use a different system of
intergroup inequality in order to revisit the conceptual issues raised by the SPE.
We therefore do not invite comparison with the SPE in terms of the exact details of how
people behaved but rather in terms of the ability of different explanatory frameworks to
make sense of what happened. Do participants accept their roles uncritically? Do those
accorded group power exercise it without constraint, and do those without group
power accept their subordination without complaint? After all, if the process of role
enactment is indeed ‘natural’, then it should apply in all cases and any exception is
troubling for the overall claim. Do the concepts used by social identity and selfcategorization
theorists provide a more satisfactory account of when people do (and do
not) adopt the social positions ascribed to them?
Our study can thus be seen as an experimental case study of the behaviour of
members in dominant or subordinate positions and of the developing relations between
them. Unlike the SPE, it is not purely exploratory but rather is theoretically informed
(by a social identity perspective). Hence, and again in contrast to the SPE, we included
manipulations of theoretically relevant variables. Given the practical impossibility of
running multiple sets of groups (due to the massive resources required in terms of
equipment, personnel, and money), a time-series approach was adopted whereby
interventions were introduced at predefined points in the study and their effects on the
development of intra- and intergroup relations then analysed.
To summarize, the overall aims of our study were as follows:
(a) To provide comprehensive and systematic data pertaining to the unfolding
interactions between groups of unequal power and privilege.
(b) To analyse the conditions under which people (i) define themselves in terms of
their ascribed group memberships and act in terms of group identities, and
(ii) accept or else challenge intergroup inequalities. Specifically, we predict that
dominant group members will identify with their group from the start and impose
their power. However, subordinate group members will only identify collectively
and challenge intergroup inequalities to the extent that relations between groups
are seen as impermeable and insecure.
(c) To examine the relations between social, organizational and clinical factors in
group behaviour.
(d) To develop protocols that provide a practical and ethical framework for examining
social psychological issues in large-scale studies.
The psychology of tyranny 7
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Method and ethics
The study was designed to create a hierarchical society in which people would live for
up to 10 days. It was conducted within an institutional environment that was
constructed inside Elstree Film Studios in north London. Prisoners were allocated to
lockable 3-person cells that were located, together with showers, off a central atrium.
This was separated by a lockable steel mesh fence from the guards’ quarters
(a dormitory, bathroom, and mess room). A plan of the prison is presented in Figure 1.
Comprehensive details of the procedures are available in Haslam and Reicher (2002)
or from the authors. What follows is a description of the key features of the study.
Ethics
The study aimed to create a system of intergroup inequality that was meaningful but was
not harmful to participants either physically or mentally. To ensure that no harm
eventuated, details of the experimental set-up and planned manipulations were
discussed with colleagues and submitted both to the University of Exeter’s ethics panel
and to the Chair of The British Psychological Society’s Ethics Committee prior to the
study being conducted. The novelty of the experimental manipulations and theoretical
analysis constituted central components of the scientific case that was presented in
order to justify the research. Moreover, the following safeguards were built into the
study:
(a) Potential participants went through 3-phase clinical, medical, and background
screening to ensure that they were neither psychologically vulnerable nor liable to
put others at risk (see below).
Isolation
cell
Prisoners'
showers
Exercise
area
Delivery
area
Entrance
10m
Guards'
walkway
(first floor)
Guards'
observation
post
= sliding lockable grille
Cell 2
Main
atrium
Guards'
dorm
Video
booth
Cell
1
Guards'
mess
= location of fixed camera
Guards'
bathroom
Cell 3
TAg FCg
BGg TMg
TQg (IBg)
PBp GPp NPp JEp KMp PPp
FCp DDp (IBp DMp)
Figure 1. Plan of the prison.
Stephen Reicher and S. Alexander Haslam8
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
(b) Participants signed a comprehensive consent form. Amongst other things, this
informed them that they may be subject to a series of factors – including physical
and psychological discomfort, confinement, constant surveillance and stress –
which may involve risk.
(c) Two independent clinical psychologists monitored the study throughout, and had
the right to see any participant at any time or to demand that any participant be
removed from the study.
(d) A paramedic was on constant standby in case of illness or injury.
(e) On-site security guards were provided with detailed protocols clarifying when and
how to intervene in cases of dangerous behaviours by participants.
(f) An independent 5-person ethics committee – chaired by a British Member of
Parliament – monitored the study throughout. This committee had the right to
demand changes to the study’s set-up or to terminate it at any time.
Apart from minor ailments (blisters, etc.) that were treated by the paramedic, no
interventions were necessary to address ethical concerns. After the study, the ethical
committee published an independent report (McDermott, O¨pik, Smith, Taylor, & Wills,
2002) and characterized the conduct of the study as ‘exemplary’.
Selection of participants
Male participants were recruited through advertisements in the national press and
through leaflets. Applicants went through three phases of screening. First, they
completed a battery of psychometric tests that measured both social variables
(authoritarianism, social dominance, modern racism) and clinical variables (depression,
anxiety, social isolation, paranoia, aggressiveness, demotivation, self-esteem, self-harm,
drug dependence). Second, they underwent a full weekend assessment by independent
clinical psychologists. Third, medical and character references were obtained, and
police checks were conducted.
For ethical reasons (noted above), we sought to include in the study only people who
were well-adjusted and prosocial, scoring at low levels on all social and clinical
measures. Additionally, we wished to ensure that the individual dispositions of our
participants were such that, if the dynamics of the study produced antisocial actions in
this sample (as in the SPE), then it could reasonably be supposed that they would have
such an effect upon almost anybody in the population.
The screening reduced an initial pool of 332 applicants to 27 men (we recruited only
men to ensure comparability with the SPE and to avoid ethical issues that would arise
from placing men and women together in cells). The final sample of 15 was chosen to
ensure diversity of age, social class, and ethnic background. They were randomly
divided into two groups of 5 guards and 10 prisoners but in such a way as to ensure that
the two groups were matched on key dimensions. More specifically, the 15 participants
were first divided into five groups of 3 people who were as closely matched as possible
on personality variables potentially implicated in tyranny: modern racism, authoritarianism
and social dominance.1
From each group of three, one participant was then
randomly selected to be a guard (and the remaining two to be prisoners). This procedure
1
This procedure – which ensures equivalence on theoretically relevant variables – was considered superior to ‘pure’ random
assignment (i.e. one which does not take account of any individual differences, as in the SPE), as the study’s relatively small
sample size means that random assignment has the potential to produce large differences between groups (due to the law of
large numbers; e.g. Haslam & McGarty, 2003, pp. 180–183).
The psychology of tyranny 9
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
was conducted blind (i.e. the identities of the participants were not known to the
experimenters). Note, however, that due to our experimental manipulations (explained
below), the actual number of people in the two groups varied over time. Thus, at the
start of the study, there were 5 guards but only 9 prisoners – the 10th being introduced
at a later stage.
Data sources
The prison environment was designed in such a way that participants could be both
video- and audio-recorded wherever they were. At all times, four channels were videorecorded
and all audio channels were recorded. There was also daily psychometric
testing. Measures were taken from a battery of scales. For the purposes of the present
analyses the critical measures were:
(a) social variables: social identification, awareness of cognitive alternatives, rightwing
authoritarianism;
(b) organizational variables: compliance with rules, organizational citizenship; and
(c) clinical variables: self-efficacy, depression.
In order to minimize fatigue, not every measure in the full battery was administered
every day. However, each was administered on multiple occasions to allow for an
analysis of development over time. Finally, daily swabs of saliva were taken in order to
ascertain cortisol levels (as an indication of stress; Laudat et al., 1988). For reasons
of space, however, we will not consider the cortisol data here (see Haslam & Reicher,
in press).
Set-up
Five participants were invited to a hotel the evening before they entered the prison.
On arrival, they were told that they would be guards in the study. They were shown the
prison timetable – which included such elements as cleaning chores, work duties,
prisoner roll calls, exercise time and a recreational hour – and were told that their
responsibility was to ensure that the institution ran as smoothly as possible and that the
prisoners performed all their tasks. The five guards were then asked to draw up a series
of prison rules under headings provided by the experimenters and to draw up a series of
punishments for rule violations.
The guards were given no guidance about how they should achieve their goals.
The only limits on what they could do were a set of ethically determined ‘basic rights’
for prisoners. In particular (and as in the SPE), all participants were told that physical
violence would not be tolerated (for rules, punishments, and rights, see Haslam &
Reicher, 2002). Beyond this, however, it was stressed that the guards could act as they
pleased.
On the morning of the study itself, the guards were taken in a blacked-out van to the
prison (since this was meant to be their entire experiential world for the duration of the
study, it was important that they could not imagine the outside). Once inside, they were
given a full briefing by the experimenters on the prison layout and the resources
available to them.
The guards had a series of means by which to enforce their authority, including keys
to all doors inside the prison (including a punishment isolation cell), sole access to an
Stephen Reicher and S. Alexander Haslam10
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
upper level, a ‘guards’ station’ with a surveillance system from which they could see into
the prisoners’ cells, resources (including snacks and cigarettes) to use as rewards or
withdraw as punishments – and, in addition, the ability to put prisoners on a bread and
water diet. They also had far better conditions than the prisoners, including superior
meals, extra supplies of drinks and snacks, superior living conditions and well-made
uniforms as opposed to the prisoners’ uniform of a t-shirt printed with a 3-digit number,
loose trousers and flimsy sandals. The prisoners also had their hair shaved on arrival.
After their briefing, the guards changed into their uniforms and practiced the
procedure for admitting the prisoners. The nine prisoners then arrived one at a time.
They were given no information apart from the prison rules, a list of prisoners’ rights,
which was posted in their cells, and a very brief loudspeaker announcement from the
experimenters. This introduced the permeability intervention (see below) and stressed
that violence was not permissible.
Planned interventions
Permeability
At their initial briefing, the guards were told that they had been selected on the basis of
their reliability, trustworthiness and initiative as gleaned from pre-selection assessment
scales. However, they were also told that while these scales were reasonably reliable,
they were not perfect. In particular, the experimenters stated that it was possible that
they had misassigned one or more of the prisoners. Hence, the guards were told
that they should observe the behaviour of the prisoners to see if anyone showed guardlike
qualities. If they did, they were told that there was provision for a promotion to be
made on Day 3. This information was also announced to the prisoners over the
loudspeaker. In the initial days of the study, participants were thus led to believe that
movement between groups was possible (see also Ellemers, van Knippenberg, & Wilke,
1990; Wright et al., 1990). After the promotion of one prisoner to guard actually took
place (the selection of the individual being made by the guards on the basis of a
procedure suggested by the experimenters), the possibility of movement was removed
by announcing that there would be no further promotions (or demotions).
Legitimacy
Three days after the promotion, participants were to be informed by the experimenters
that observations had revealed that there were in fact no differences between guards
and prisoners on the key group-defining qualities. However, they would be told that it
was impractical to reassign them and hence the groups would be kept as they were.
Accordingly, whereas previously the group division had been legitimate, this would no
longer be the case (Ellemers et al., 1993).
Cognitive alternatives
Within a day of the legitimacy intervention, a new prisoner was to be introduced.
Although he was as naı¨ve as the others, he was chosen for this role (from the pool of 10
participants randomly assigned to be prisoners) because of his background as an
experienced trade union official. On this basis, we expected that he would introduce a
new perspective to the prison based on notions of group-based negotiation and
collective- and equal rights (i.e. a perspective that suggested the existing regime was
both illegitimate and changeable). It was also thought that he might provide skills
The psychology of tyranny 11
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
necessary to organize collective action (e.g. see Haslam, 2001). Hence, it was envisaged
that his introduction would enable the prisoners (and the participants more generally)
to envisage the achievement of a more equal set of social relations.
Results
The findings of the study can be divided into two phases. In the first phase, the guards
failed to identify with each other as a group and to cohere collectively. By contrast, after
the promotion on Day 3, the prisoners did increasingly identify as a group and work
collectively to challenge the guards. This led to a shift of power and ultimately to the
collapse of the prisoner–guard system. In the second phase, participants decided to
continue as a single self-governing ‘commune’. However, they were unable to deal with
internal dissent and lost confidence in the communal system. By the end of the study,
they were increasingly disposed to tolerate a new and much more draconian system of
inequality that some participants now wished to impose.
Results that pertain to these two phases will be presented in turn. For each phase,
we combine a description based on the observational data with statistical analyses based
on the quantitative measures. The statistical analyses are based on the data of individual
participants within groups (excluding data from the participant who was promoted
from prisoner to guard on Day 3 and from the prisoner who was introduced on Day 5).
Given the interaction between participants, it could be argued that the group, rather
than the individual group member, should be the unit of analysis here. For this reason,
the present data were also analysed using methods advocated by McGarty and Smithson
(2005), which do not assume (or require) independence of observations. These analyses
confirmed the reliability of all the patterns reported below. However, for reasons of
space and in light of the novelty of these alternative methods, we accord with general
usage by presenting statistics that are commonly used even where there is interaction
between participants (Hoyle, Georgesen, & Webster, 2001).
Phase 1: Rejecting inequality
Social identification
For the prisoners, the development of social identification was consistent with
predictions. From the start, they were clearly dissatisfied by their inferior conditions.
Initially though, as predicted in light of the permeability of group boundaries, many
sought to improve their lot by displaying the individual qualities necessary for
promotion as opposed to mounting a collective challenge to the guards. As a result,
there was no shared identity among the prisoners and no consensus about how they
should behave (see Photograph 1).
However, after the promotion, when group boundaries were impermeable and
participants could only alter their position by changing the general prisoner–guard
relationship, the prisoners began to develop a much stronger sense of shared social
identity and to develop more consensual norms – particularly in relation to their
treatment of the guards. This contrast between the pre- and post-promotion periods is
exemplified in exchanges between participants in Cell 2. Before the promotion, two
occupants of this cell, JEp
2
and KMp, worked conscientiously and explicitly sought to
improve their position by displaying behaviours required to become a guard. As JEp put
2
The subscripts p and g after participants’ initials denote their status as prisoner or guard, respectively.
Stephen Reicher and S. Alexander Haslam12
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
it, ‘I’d like to be a guard because they get all the luxuries and we are not’. However,
almost immediately after the promotion, all three occupants of this cell (i.e. including
PPp) recognized that the only way to improve their position was to change the system.
Accordingly, they began to discuss how they could achieve this together. If anyone
expressed doubts about this objective, they were reminded of the collective inequalities
in the prison and of the need for collective resistance. This is exemplified by the
following interchange between the cellmates:
Extract 1
JEp: Hopefully we’ll get [TQg] in. That’s the person, he’s the target.
KMp: No. I mean obviously I think it’s going to be a lot of fun for us to do this but
I don’t think [TQg] : : : I feel so : : : I just feel : : :
PPp: Listen, listen mate I, you’ve got to, you’ve got to start forgetting about other
people’s feelings and what they’re doing because the days when you’re sitting
here starving hungry and you’ve got fuck all and you’ve got nothing mate and
you’ve got a ratty little bed and a stupid little blanket to sit under and they’re
under there in their duvets, they’ve got everything they want and they’re not
giving two fucks about you. So – think on and fuck them.
KMp: I think they do care about us. But guys I’m going to back you all the way.
You should no’ doubt me.
For the prisoners, then, the promotion led to a perception of impermeability. This
was accompanied by a shift from individual action and identification (i.e. a stress on
what ‘I’ will do) to collective action and identification (i.e. discussion of what ‘we’ will
do) and from compliance to conflict with the guards.
For the guards themselves, the results were very different. Moreover, the patterning
of their social identification went against our predictions. Along lines reported by
Zimbardo and his colleagues (Haney et al., 1973), we had expected that, from the
outset, they would identify with what was a high-status and positively valued group
within the prison. There was some evidence of this in the first day of the study, but it
was also clear that several guards were wary of assuming and exerting their authority.
Consequently, some guards were always ambivalent about internalizing their
assigned identity. Even to the extent that they overcame this ambivalence, the group as a
whole could never reach consensus about its norms and priorities. In addition, over
Photograph 1. Day 2: The prisoners comply with the guards’ regime.
The psychology of tyranny 13
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
time – particularly once the boundaries between-groups became impermeable and the
prisoners had developed a stronger sense of shared identity – these cleavages were
exacerbated and the fragility of the guards’ collective identity became more apparent.
However, as noted above, in contrast to the SPE, our systematic collection of
quantitative data allowed us to complement behavioural observations that speak to
these patterns with inferential statistical analysis (and hence to triangulate behavioural
and psychometric data). To this end, social identification with prisoner and guard
groups was measured every day by means of two 3-item scales containing items used
widely in previous research (e.g. Doosje, Ellemers, & Spears, 1995; ‘I feel strong ties
with the prisoners [guards]’/‘I identify with the prisoners [guards]/I feel solidarity with
the prisoners [guards])’. As with all other measures, responses were made on 7-item
scales with appropriately labelled end-points (e.g. not at all, extremely).
Identification scores were computed by averaging responses on the three measures
that comprised each scale and subtracting the out-group identification score from the ingroup
identification score. Note too that, as with all other measures, data were collected
early each morning. Accordingly, scores relate primarily to events of the previous day.
Given the small number of participants, the study inevitably has a low level of
statistical power. In light of this, statistical significance was considered in conjunction
with effect size (Cohen, 1977; Smithson, 2000). For all analyses, effects were only
considered meaningful when, as well as being significant at conventional levels
(a ¼ .05), effect size was strong by Cohen’s (1977, p. 283) criteria (i.e. h2
. .14).
Figure 2 presents mean social identification scores as a function of participant group
and time. As can be seen from this graph, the data here are consistent with the above
observations. In the very first days of the study, the guards identified more strongly with
their social position than the prisoners. However, as soon as they had to implement the
disciplinary regime (from Day 2 onwards), the guards’ identification fell while that of the
prisoners rose. Prisoner identification was also particularly high after the promotion
(Day 4).
Statistical analysis of in-group identification scores was conducted by means of a 2
(participant group: guards, prisoners) £ 6 (study phase: Day 1 to Day 6) ANOVA with
repeated measures on the second factor. This analysis supported observational data in
revealing a significant and large interaction between group and time, F(5, 55) ¼ 3.05,
p , .05; h2
¼ .22. Analyses of polynomial trends were used as the most practical way of
decomposing this effect (Norusis, 1985). Consistent with observations, these showed
0
0.5
1
1.5
2
2.5
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6
Prisoners
Guards
Study phase
SocialIdentification
Figure 2. Social identification as a function of assigned group and time.
Stephen Reicher and S. Alexander Haslam14
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
that social identification among the prisoners increased linearly over time, t (7) ¼ 2.46,
p , .05. On the other hand, identification among the guards declined as the study
progressed, but non-significantly, t(4) ¼ 20.77, ns.3
Security of intergroup relations
After the promotion, the normative consensus among prisoners led to effective
organization based on both the expectation and the provision of mutual social support
between group members (Haslam, O’Brien, Jetten, Vormedal, & Penna, 2005; House,
1981; Underwood, 2000; see Haslam & Reicher, in press, for a fuller treatment of this
point). Conversely, the guards’ inability to agree on group norms and priorities meant
that they could not trust others to represent them or act appropriately. As a
consequence, they felt (and were) weak, inconsistent, and ineffective as a group.
These developments directly undermined the perceived legitimacy of the intergroup
inequality as the guards showed no evidence of the qualities by which their selection was
supposedly justified. Such developments also contributed directly to the emergence of
cognitive alternatives; not only could the prisoners envisage changes to the existing
hierarchy, but they also had a growing confidence in their ability to achieve them.
Consider, for instance, the outcome of the first confrontation between prisoners and
guards. The three prisoners of Cell 2 had orchestrated this mainly to see how the guards
would respond when challenged. As part of a pre-arranged plan, JEp threw his lunch plate
to the ground and demanded better food. As the guards tried to intervene, KMp and PPp
then joined in with further demands for smoking rights and treatment for a blister. The
guardswere totally disunited in their response, with some wanting to take a disciplinarian
line and others wanting to make concessions (see Photographs 2 and 3) . This continued
until one guard, TQg, decided to resolve the situation by acceding to PPp’s requests for a
cigarette in order to encourage the other prisoners to return to their cell.
After the incident, the prisoners back in their cell and the guards back in their mess
expressed a growing realization that the system was open to change. However, they did so
with a very different evaluative tone. The prisoners literally danced with joy. PPp exclaimed
‘that was fucking sweet’, to which KMp responded (admiringly) ‘you was fucking quality
man’. Then all three cellmates exchanged ‘high fives’. The guards bickered despondently.
TAg started by saying ‘This is only Day 4. They can see what happened today and now
they know they can do whatever they want’. As BGg and then TQg weakly disagreed, TAg
only insisted more, culminating in the claim that ‘It’s happened. I mean, come on. You
know what that has done? That has lit the fuse on [PBp’s] arse’.
Hence, the induction of insecure relations between the groups was not dependent
upon our interventions but was an emergent property of the intergroup dynamics.
Correspondingly, planned interventions were not necessary for the prisoners to start
challenging the guards. The challenge started mounting immediately after the
promotion. Consequently, the legitimacy intervention was not implemented. However,
because the new prisoner had already been told he would be brought in, this went
ahead on Day 5 (and he was later withdrawn on Day 6). Given the situation, he did not
so much suggest cognitive alternatives where none existed as suggest additional
alternatives to the status quo. Most notably, he began to question the prisoner–guard
3
In this instance (and with depression data below), analysis of polynomial contrasts was based on a random subset of five
phases because the number of levels of the within-subjects variable (study phase) exceeded the number of subjects.
The psychology of tyranny 15
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
division itself and, by questioning the legitimacy of certain aspects of the study
(particularly the heat in the prison), he suggested both to his fellow cellmates (DDp and
FCp) and to one of the guards (TQg) that the participants mount a united challenge to
the experimenters. As the following extract shows, this provided new ways of
envisaging how the system might work:
Extract 2
DMp: If this was a real-life situation : : :
TQg: Yes.
DMp: : : : and you were working in this kind of heat, then you as an employee could
well go to the employer and say ‘The condition is unacceptable, I’m not
prepared to work in it’. Now let’s treat this as a real-life situation. You and I –
your group and the group I’m in – both have this problem of the heat. And if I’ve
got to sleep in this, there is no way I will. And I, you know, won’t bear it. And I
think collectively we should do something about it to the people who are
running the experiment. Now you know in a normal, day-to-day, real-life
situation, that’s what would happen.
TQg: Well, I am most impressed with your new-found kind of angle on this, which
possibly shouldn’t come as a surprise to me. But I think that is a very very valid
point you are making and I’m going to go along with it completely.
As a first practical step towards participant unity, DMp created a new negotiating
structure, which brought guards and prisoners together on a basis that was far more
equal than had hitherto been the case. The guards were eager and pleased to accept this
arrangement, for even if they surrendered much of their hierarchical advantage by doing
so, the new system confirmed their position in what promised to be a viable social order.
Quantitative confirmation of these patterns emerges on measures of participants’
awareness of cognitive alternatives to the status quo (Tajfel & Turner, 1979). These
were administered on Days 1, 3, 4, and 6 using a 4-item scale (‘I cannot imagine the
relationship between guards and prisoners being any different (reverse-scored)’/‘I think
that the guards will always have more privileges than the prisoners (reverse-scored)’/‘I
think that the relationship between guards and prisoners is likely to change’/‘I think that
it would be possible for the prisoners to have more power than the guards)’. A single
score was computed by averaging responses to these 4 items.
Photographs 2 and 3. Day 4: The prisoners start to challenge the guards’ regime.
Stephen Reicher and S. Alexander Haslam16
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mean scores are presented in Figure 3 as a function of participant group and time.
As can be seen from this graph, at the start of the study, prisoners and guards had an
equal and relatively low sense of cognitive alternatives, but this increased as the study
progressed. These patterns were confirmed by statistical analysis of cognitive
alternatives data that was conducted by means of a 2 (participant group: guards,
prisoners) £ 4 (study phase: Days 1, 3, 4, 6) ANOVA with repeated measures on the
second factor. This analysis revealed a significant and large main effect for time,
F(3, 33) ¼ 3.36, p , .05; h2
¼ .34. Consistent with the above observations, analysis of
polynomial trends to decompose this effect indicated that participants’ sense of
cognitive alternatives increased linearly over time, t (12) ¼ 2.83, p , .05.
Acceptance of the unequal regime: Compliance and organizational citizenship
Associated with the above effects, as the prisoners developed a sense of shared social
identity that was defined in opposition to the guards, and as they became aware of the
possibility of alternatives to existing status relations, they started to work actively against
the regime that the guards were trying to uphold. This was manifest both in minor
challenges to the guards’ status (e.g. insubordination during roll call) and in more overt
challenges – including the incident alluded to above, in which a number of prisoners
collectively protested about the quality of their food.
Data consistent with this analysis emerge from measures of (a) participants’
willingness to comply with authority and (b) their willingness to engage in acts of
organizational citizenship (i.e. to do more than was asked of them in order to make the
prison system work; Organ, 1988, 1997).
Compliance was measured on Days 1, 3 and 5 by means of a 2-item scale (‘I try to do
what the guards want’/‘I try to comply with the guards’ rules’). Responses on these
measures were averaged to provide a single score. The resultant mean scores are
presented in Figure 4 as a function of participant group and time. As can be seen from
this graph, in the early stages of the study, both guards and prisoners were willing to
comply with the rules, but the prisoners became more reluctant to comply after the
promotion.
These patterns were confirmed by statistical analysis of compliance scores that was
conducted by means of a 2 (participant group: guards, prisoners) £ 3 (study phase: Days
1, 3, 5) ANOVA with repeated measures on the second factor. This revealed a significant
1
2
3
4
5
6
7
Day 1 Day 3 Day 4 Day 6
Prisoners
Guards
Study phase
Awarenessof
cognitivealternatives
Figure 3. Awareness of cognitive alternatives as a function of assigned group and time.
The psychology of tyranny 17
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
and large interaction between group and time; F(2, 22) ¼ 3.62, p , .05; h2
¼ .32.
Analyses of polynomial trends indicated that compliance on the part of the guards did
not vary significantly over time, t (4) ¼ 20.12, ns, but that of the prisoners declined in a
linear fashion, t (7) ¼ 24.48, p , .01. There was also evidence of a quadratic trend,
t (7) ¼ 22.18, p , .07, suggesting that the prisoners’ decline in compliance was
particularly marked after group boundaries had become impermeable.
Organizational citizenship (Organ, 1988, 1997) was measured on Days 2, 4, and 5 by
means of a 3-item scale containing items used widely in previous research (e.g. Haslam,
Powell, & Turner, 2000; Tyler & Blader, 2000; ‘I am willing to do more than is asked of
me by the guards’/‘I will do whatever I can to help the guards’/‘Whenever possible I will
try to make the guards’ work difficult’; reverse scored). A single score was computed by
averaging responses to these 3 items.
The resultant mean scores are presented in Figure 5 as a function of participant
group and time. As can be seen from this graph, the guards were always more willing
than the prisoners to engage in citizenship behaviours that would help them run the
regime. However, while the guards maintained their willingness to work for the regime
across the three testing phases, over time the prisoners became much more reluctant to
support the guards’ regime in this way. Again, this was particularly true after the
promotion on Day 3.
1
2
3
4
5
6
7
Day 2 Day 4 Day 5
Prisoners
Guards
Study phase
Organizationalcitizenship
Figure 5. Organizational citizenship as a function of assigned group and time.
1
2
3
4
5
6
7
Day 1 Day 3 Day 5
Prisoners
Guards
Study phase
Compliance
Figure 4. Compliance with prison rules as a function of assigned group and time.
Stephen Reicher and S. Alexander Haslam18
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
These patterns were confirmed by statistical analysis of organizational citizenship
scores performed by means of a 2 (participant group: guards, prisoners) £ 3 (study
phase: Days 2, 4, 5) ANOVA with repeated measures on the second factor. This analysis
revealed a significant and large main effect for group, F(1, 11) ¼ 7.84, p , .01; h2
¼ .42,
but this was qualified by a significant and large interaction between group and time;
F(2, 22) ¼ 5.10, p , .05; h2
¼ .32. Analyses of polynomial trends revealed that the
guards’ willingness to display organizational citizenship did not vary across the three
phases of testing, t(4) ¼ 0.69, ns, but that of the prisoners declined linearly over time,
t(7) ¼ 23.74, p , .01.
Collective self-efficacy and mental health
The disorganization of the guards and the unity of the prisoners did not just produce
conditions where the latter challenged the former, it also led to the prisoners becoming
progressively more dominant. Through planning and mutual support, they became
increasingly more extreme and more successful in their efforts to undermine the guards’
control. By contrast, the guards’ attempts to impose their authority became increasingly
unsuccessful, and they became more divided and mutually recriminatory. Moreover, as
we have already illustrated with reference to observational data, the effectiveness of the
prisoners in pursuing their collective goals led to strong positive affect while the
inability of the guards to act collectively led to despondency.
These dynamics are reflected in quantitative data obtained on measures of
(a) collective self-efficacy and (b) depression. Collective self-efficacy (Bandura, 1995)
was measured on Days 2, 4, and 6 by means of a 5-item scale containing items used
widely in previous research (e.g. Chen & Bliese, 2002; ‘My prison group is confident that
we could deal efficiently with unexpected events’/‘My prison group can remain calm
when facing difficulties because we can rely on our coping abilities’/‘My prison group
can always manage to solve difficult problems if we try hard enough’/‘When my prison
group is confronted with a problem, we can usually find several solutions’/‘My prison
group can usually handle whatever comes our way’). A single score was computed by
averaging responses to these 5 items.
Figure 6 presents the resultant mean collective self-efficacy scores as a function of
participant group and time. As can be seen from this graph, these data mirror those
obtained on measures of social identification, so that, at the start of the study, the guards
1
2
3
4
5
6
7
Day 2 Day 4 Day 6
Prisoners
Guards
Study phase
Collectiveself-efficacy
Figure 6. Collective self-efficacy as a function of assigned group and time.
The psychology of tyranny 19
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
had a greater sense of collective self-efficacy than the prisoners. However, by Day 4,
the prisoners’ self-efficacy had increased markedly and was now greater that that of the
guards. This pattern was maintained at Day 6, although it is interesting to note that after
the trade unionist was removed from the prison, both groups showed some decline in
self-efficacy (seemingly because the opportunities for group-based negotiation and
order that he presented were removed; see Eggins, Haslam, & Reynolds, 2002; Haslam,
Eggins, & Reynolds, 2003). These patterns are consistent with our observational
account – although the general decline in the guards’ sense of self-efficacy is less
pronounced in these quantitative data.
Collective self-efficacy scores were statistically analysed by means of a 2 ( participant
group: guards, prisoners) £ 3 (study phase: Days 2, 4, 6) ANOVA with repeated
measures on the second factor. This analysis revealed a significant main effect for time,
F(2, 22) ¼ 6.08, p , .01; h2
¼ .36, that was conditioned by a significant and large
interaction between group and time, F(2, 22) ¼ 4.31, p , .05; h2
¼ .28. In line with the
above observations, analyses of polynomial trends revealed that the prisoners’ sense of
collective self-efficacy increased linearly over time, t(7) ¼ 2.55, p , .05 – an effect that
was also qualified by significant quadratic variation, t(7) ¼ 4.25, p , .01. On the other
hand, the collective self-efficacy of guards declined over time, although this effect was
not significant, t(4) ¼ 20.90, ns.
Depression was measured during the screening process and then every day during
the study (partly in order to monitor the ongoing welfare of participants). For this
purpose, a 7-item scale was administered (‘In general, how has your mood been over the
last few days?’/‘Do you ever feel low or depressed?’/‘Do you feel hopeless about the
future?’/‘Do you have difficulty dealing with everyday problems?’/‘Are you selfconfident?’/‘Do
you think that you are a worthwhile person?’/‘Do you think about
harming yourself?’). During the screening phase (N ¼ 332), these items were found to
form a reliable scale (a ¼ .80) and so the items were averaged to form a single measure.
Mean depression scores are presented in Figure 7. From this graph, it can be seen
that although overall levels of depression were low, they clearly varied as a function of
participant group and time. Specifically, while the prisoners were more depressed than
the guards at the start of the study, by its end, this situation had reversed. This pattern
was confirmed by a 2 (participant group: guards, prisoners) £ 7 (study phase: pre-test,
Days 1 to 6) ANOVA with repeated measures on the second factor. This revealed
a significant and large interaction between group and time; F(6, 66) ¼ 3.73, p , .01;
1
2
3
4
Pre-
test
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6
Prisoners
Guards
Study phase
Depression
Figure 7. Depression as a function of assigned group and time.
Stephen Reicher and S. Alexander Haslam20
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
h2
¼ .25. Consistent with observations, polynomial contrasts showed that the
prisoners’ depression decreased linearly over time, t (7) ¼ 24.22, p , .01. At the
same time the guards’ depression increased, though not significantly, t (4) ¼ 1.25, ns.
Combined impact
The combined outcome of these interrelated dynamics was that as the prisoners became
increasingly aware of, and confident in, their collective identity, and as the guards’
confidence and collective self-efficacy declined, the prisoners developed and executed
plans to take on and destroy the guards’ regime. The guards were divided, exhausted
and demoralized. They were unable to organize themselves or the prisoners effectively.
As a result, late on the evening of Day 6, the prisoners in Cell 2 broke out of their cell and
occupied the guards’-quarters. At this point, the guards’ regime was seen by all to be
unworkable and at an end (see Photograph 4).
Phase 2: Embracing inequality
After the collapse of the guard–prisoner hierarchy, all but two of the participants
decided that they wanted the study to continue through the institution of a single selfgoverning
commune. With FCp in the chair, they met with the experimenters and drew
up the terms under which the commune would operate. Initially, the new system was
highly effective. A number of participants who had been mutually hostile when they
were divided into prisoners and guards formed strong and positive affective ties now
they were recategorized part of a common group (Gaertner, Mann, Murrell, & Dovidio,
1989; Turner et al., 1987). Collectively, they performed their work tasks and chores with
more effort and to a higher standard than at any other point in the study. However, those
participants who had been centrally involved in challenging the old regime felt
marginalized in this new system. At first, they simply failed to contribute to collective
tasks. Later, they began to violate communal rules. Within a day, they were plotting to
destroy the commune. The problems this created were exacerbated because the
commune’s members had never developed procedures for dealing with dissidence, and
hence they had no means of responding to threats to their social order.
By the morning of the commune’s second day, it was clear to many participants that
the new social structure was in crisis. This situation was exacerbated when, by chance,
Photograph 4. Day 6: Cell 2 break out bringing the guards’ regime to an end.
The psychology of tyranny 21
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
the breakfast was of very poor quality. This was (incorrectly) taken as a sign that the
experimenters disapproved of the commune system. It led to despondency among the
commune’s supporters who felt, as one later put it, that they had the burden of
responsibility for the system without the ability to make it work. Moreover, the
emergent crisis was exploited by opponents of the commune, four of whom (one exguard
and three ex-prisoners) formulated a plan to create a new and harsher guard–
prisoner hierarchy. The nature and tone of this new regime was made abundantly clear
in discussions about the form this would take. As PBp put it, ‘We want to be the guards
and fucking make them toe the line, I mean on the fucking line. No fucking talking
while you are eating. Get on with your food and get the fucking hell back to your cell’.
Shortly after breakfast, this group convened a meeting (see Photograph 5). Their
leader berated the commune and its supporters and he introduced the idea of the new
hierarchy. The supporters of the commune were largely passive in response. They
looked despondent and listened in silence until he had nearly finished. During
debriefings, a number of them acknowledged that, although they would not have openly
endorsed such a hierarchy, they were less opposed to it than they had been previously
and that they felt less repulsed by the idea of a strong social order in which someone else
assumed responsibility for making the system work.
Again, these trends are supported by analysis of psychometric data. Most notably,
this is apparent when one looks at the level of participants’ right-wing authoritarianism
(Altmeyer, 1981, 1996). This was measured using 8 items abstracted from a 30-item scale
administered during pre-testing (Reynolds, Turner, Haslam, & Ryan, 2001; ‘Things
would go better if people talked less and worked harder’/‘It is better to live in a society
in which the laws are vigorously enforced than to give people too much
freedom’/‘People should always comply with the decisions of the majority’/‘You have
to give up an idea when important people think otherwise’/‘There are two kinds of
people: strong and weak’/‘What we need are strong leaders that the people can
trust’/‘Our social problems would be solved if, in one way or another, we could get rid of
weak and dishonest people’/‘People should always keep to the rules’). In pre-testing
(N ¼ 332), this scale was found to be reliable (a ¼ .71) and to correlate very highly with
the full 30-item scale (r ¼ .97). Accordingly, scores on the above items were averaged to
form a single measure. This was administered during the screening process and then on
Days 1, 3 and 7.
Photograph 5. Day 8: New guards make the case for an authoritarian regime.
Stephen Reicher and S. Alexander Haslam22
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
In the first instance, it is relevant to examine levels of authoritarianism over time as a
function of the groups to which participants were assigned by the experimenters. Data
pertaining to this analysis are presented in Figure 8. These were analysed by means of a 2
(participant group: guards, prisoners) £ 4 (study phase: pre-test, Days 1, 3, 7) ANOVA
with repeated measures on the second factor. Consistent with qualitative observations,
the only effect to emerge from this was a significant and large effect for time,
F(3, 33) ¼ 2.75, p , .05; h2
¼ .20. Analysis of polynomial trends revealed that there was
a linear increase in all participants’ authoritarianism as the study progressed,
t (12) ¼ 2.61, p , .05.
However, as a fascinating variant on the above analysis, it is also possible to look at
authoritarianism as a function of the groups to which the remaining participants
assigned themselves at the end of the study – that is, as participants who either
supported the commune or who proposed setting up a new hierarchy (with themselves
as the new guards within it). Data pertaining to this analysis are presented in Figure 9.
These were again analysed by means of a 2 (participant group: new guards, new
prisoners) £ 4 (study phase: pre-test, Days 1, 3, 7) ANOVA with repeated measures on
the second factor. The only effect to emerge from this analysis was a significant and large
interaction between group and time; F(3, 36) ¼ 3.07, p , .05; h2
¼ .20. Analysis of
1
2
3
4
Pretest Day1 Day 3 Day 7
Prisoners
Guards
Study phase
Authoritarianism
Figure 8. Right-wing authoritarianism as a function of assigned group and time.
1
2
3
4
Pre-test Day 1 Day 3 Day 7
New prisoners
New guards
Study phase
Authoritarianism
Figure 9. Right-wing authoritarianism as a function of self-selected group and time.
The psychology of tyranny 23
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
polynomial contrasts indicated that this interaction arose from the fact that the
authoritarianism of those who sought to assume the guards’ role at the end of the study
had declined non-significantly as the study progressed, t(3) ¼ 21.05, ns. However, the
authoritarianism of the remaining participants had increased steadily over time,
t(10) ¼ 3.67, p , .01 (so that the two groups were demonstrating a very similar level of
authoritarianism at the time that the new guards were actually seeking to take power).
Despite the general shift towards authoritarianism, and although it is probable that
the new guards would have had the force to impose their regime in the face of
weakening resistance, in this study such force was prohibited under the pre-established
ethical guidelines. Hence, in our judgment, the study was gridlocked and had reached
what we saw as a natural point of termination; the existing system was not working and
the new system could not be imposed. Accordingly, the study was brought to a
conclusion at noon on Day 8. However, the participants remained for a further day in
order to undertake a series of structured debriefings designed to obtain and provide
feedback on their experience, to explain the rationale for the study and to overcome any
hostility between individuals deriving from events in the study.
Discussion
The BBC prison study was designed to examine the factors that determine how people
respond when a system of inequality is imposed upon them by others. At the start,
almost all the participants rejected this system. However, by the end, they were close to
instituting a new and more tyrannical social system. In addition to our original questions
concerning the way in which people respond to a system of inequality that has been
imposed upon them – do they accept it or do they resist it? – this raises a new and
unexpected issue. What are the conditions under which people create a system of
inequality for themselves?
We will argue that our findings concerning reactions to inequality cannot be
explained through a general or ‘natural’ tendency to assume roles and assert power.
Rather, they require an understanding of the conditions under which an externally
imposed categorization becomes a subjective self-perception (Turner, 1982).
The existing terms of social identity and self-categorization theories provide the basis
for such an understanding, although it is necessary to elaborate upon notions of context
in order to account for the failure of the guards to cohere as a group. Our findings
concerning the creation of inequality lead us to more original theoretical conclusions.
On the one hand, we suggest that groups are the basis for collective self-realization – that
is, the creation of a social order based on shared values and norms (Drury & Reicher,
2005). However, where groups fail, we argue that people will be more inclined to
accept the imposition of a social order by others, even where that violates their values
and norms. Therefore, in contrast to those who explain tyranny and other extreme
social phenomena in terms of the psychological dysfunctionality of groups, we interpret
them in terms of the dysfunctionality of group failure.
Before we are in a position to develop these arguments, however, it is necessary to
deal with four potential critiques of the study and its ability to say anything meaningful
about the group processes relating to inequality and tyranny. The first is that the
behaviour of participants was determined by the fact that they knew they were being
observed by television cameras and that this renders the study so artificial as to have
little or no general value. The second is that the effects we observed were a product of
Stephen Reicher and S. Alexander Haslam24
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
participants’ personalities and that therefore the study has little to say about group
processes. The third is that we failed to create real power differences or meaningful
inequalities between groups and hence our study has little to say about the psychology
of either power or inequality. The fourth is that the variables upon which our
predictions (and hence our planned interventions) focused were not responsible for the
effects obtained and, therefore, even if group and power processes were at play, we
cannot be sure what they are.
Four critiques
The role of television
From their moment of first contact, volunteers knew that the BBC was involved in this
project, and by the time that the study started, all participants were aware that they
would be constantly observed by television cameras and that anything they did might be
shown on national television. Although it was stressed throughout that this was a
specialist scientific project that would not make ‘stars’ out of those involved, and
although the screening process was used to exclude anyone who was motivated by the
desire for publicity, this is a highly unusual situation and was bound to impact on
behaviour – but what impact, and with what implications?
The most damning argument would be that participants were simply faking their
behaviours for the cameras. We believe such an explanation to be implausible in light of
the immense effort that would be required to continuously monitor and fake one’s own
behaviour for nearly 9 days. Moreover, it would be much harder, if not impossible, to
fake the psychometric and physiological data. However, even if participants were
capable of such play-acting, one would still need to explain the complex pattern of
results not only between groups, but also within groups across time. One might, for
example, suggest that there is a certain discredit in being a tyrant and a certain glamour
in being a rebel, which may explain why our guards were so mild and our prisoners so
rebellious. But, why then did the prisoners become more rebellious after the system was
made impermeable and insecure? Even more problematically, why did the participants
move towards tyranny at the study’s end? To argue that our findings can be explained by
suggesting that participants were merely play-acting or seeking ‘celebrity status’ is not
only implausible but also unhelpful. In particular, it fails to explain (a) why people acted
as they did when they did and (b) why their behaviour and attitudes clearly changed in
particular (predicted) ways over the course of the study. By the same logic, surveillance
cannot be the whole story, as it remained constant at the same time that behaviour itself
was changing.
Having said this, we readily accept (see below) that the televising of the study is part
of the story – particularly towards the start of the study (participants reported that they
were acutely aware of the cameras on entering the ‘prison’, but that, as time went by,
they increasingly forgot about them except during quiet moments; e.g. when seeing a
camera move late at night in their cells). However, we suggest that the nature of this
impact adds to, rather than detracts from, the richness and wider relevance of our
findings. This is because although it may be unusual to be in a position where anything
one does might be broadcast into millions of homes, this is an extreme example of
something that is an increasingly common feature of our everyday lives – namely,
surveillance. For most of our social lives, we are under observation and our behaviour
can be examined by audiences who are not present (Reicher & Emler, 1985).
Sometimes, the surveillance is a matter of surveillance cameras, scrutiny of computer
The psychology of tyranny 25
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
transactions, workplace review, or unobtrusive profiling (Lyon, 1994; Lyon & Zuriek,
1996). At other times, it is simply a matter of those who are with us in one place talking
about our behaviour to other people in other places (e.g. in gossip, a universal human
behaviour; Emler, 1994). However, whichever is the case, we can rarely ignore
surveillance when we are alone and almost never when with others.
What the television cameras do, therefore, is to highlight in dramatic form an aspect
of human experience, which is all too often overlooked in psychology. Indeed, it is
arguable that the typical psychology experiment in which people are isolated and
guaranteed absolute anonymity encounters greater problems of artificiality (Cronin &
Reicher, 2006). Of course, the size and variety of audiences to which one is made
accountable by television is much larger than we commonly experience. However, as
we will discuss in more detail below, the fact of accountability, and the fact that our
participants had to consider other contexts and other audiences even when acting
within the context of the study, enhances rather than diminishes its relevance and
meaningfulness. In this respect, we are reminded of Thayer and Saarni’s (1975) retort to
those who impugn the validity of the SPE on the grounds that participants were guided
by the expectations of others. This is summarized in the title of their paper: ‘demand
characteristics are everywhere (anyway)’. Likewise, to those who criticize the BBC
study on the grounds that it was broadcast, we would respond that surveillance is
everywhere (anyway).
The role of personality
For those who have watched The Experiment (Koppel & Mirsky, 2002), it is hard not to
be struck by the force of many of the characters who contributed to the various
outcomes described above. In part, this has to do with the visual nature of the medium,
where the concrete and visible impact of individuals is inevitably more salient than that
of more intangible and ‘invisible’ elements such as group processes (Asch, 1952).
Nonetheless, it could clearly be argued that all the observed effects were a product of
strong (and weak) personalities and that the more powerful characters (in particular,
JEp, PPp, and PBp) ended up as prisoners. One could argue that their personalities drove
these individuals to overwhelm the guards and then, as they grew bored of the
commune, to destroy it in turn.
The first response to such an argument is to remember that we matched the
prisoners and guards on the obvious individual difference factors relating to tyranny:
modern racism, authoritarianism, and social dominance. Nonetheless, one could still
argue that there were other characteristics on which we did not match the groups. As a
result, the ‘stronger’ individuals might still, by chance, have been allocated to the
prisoner group. However, even if this were the case, as with our discussion of
surveillance, is does not prove particularly helpful in explaining how events in the study
unfolded. Most fundamentally, this due to the interrelated observations (a) that people’s
‘character’ on relevant dimensions appeared to change over the course of the study and
(b) that relevant individual differences were much less apparent at the outset than they
were as the study progressed.
We readily accept, however, that individual differences are part of the story (see
Reicher & Haslam, in press; Turner, Reynolds, Haslam, & Venstra, in press). However, as
with the analysis of leadership more generally (e.g. see Haslam, 2001; Haslam & Reicher,
2005; Reicher & Hopkins, 2001b; Turner & Haslam, 2001), the problem is that it is hard
to explain changing patterns of behaviour in our study with reference to a blunt and
constant factor such as ‘personality’. To claim that certain people were ‘rebellious’ does
Stephen Reicher and S. Alexander Haslam26
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
not explain why rebellion was muted until after the promotion and fails to explain why
those prisoners who rebelled against one form of authority (the guards) nonetheless
remained deferential to another (the experimenters). In the case of JEp, for example, we
see that before the promotion, he invested his energies in supporting the system and it
was only after promotion was ruled out that he put them into undermining the system.
Therefore, it is one thing to call him ‘forceful’, but to explain how that force was
directed, we need to invoke systemic factors. Similarly, in the case of JE’s cellmate PPp,
while he was opposed to the guards’ authority from the start of the study, his ability
actually to undermine their regime was contingent on collective will and support,
which only materialized after the promotion (in ways predicted by social identity
theory; Tajfel & Turner, 1979; see below).
More generally, there were many points in the study where the prominence and
impact of individuals depended upon their relationship to the group (Reicher & Haslam,
in press). Thus, when they were seen to reflect shared identity and shared values, even
the mildest of men came to the fore. This is apparent in the case of FCp, an
environmentalist who had been quiet throughout the first 6 days of the study, but who
became a leading and vocal figure in creating and managing the commune and without
whose commitment, knowledge, and experience it may never have come into being.
Furthermore, it is also apparent that when there was no shared identity or shared
values, even the most heroic of individuals was bound to fail. This was seen early in the
study when FCg, made a strenuous effort to impose the guards’ authority and maintain
strict discipline. However, as it became clear that he lacked the support of his fellow
guards, he gave up, became increasingly passive, and withdrew into the background.
Thus, the manifest differences in the forcefulness of prisoners and guards derived not
from their inherent personalities but were instead an emergent product of the success
or failure of their respective groups.
Again, this is not to dismiss the importance of individual variables. As we have
repeatedly stressed, the skills, knowledge, values, and pre-existing commitments of
individuals played a crucial part in facilitating, shaping (and indeed in blocking) group
formation (Postmes, Baray, Haslam, & Morton, in press; Reicher, 2004; Reicher, Drury,
Hopkins, & Stott, 2001; Reicher & Hopkins, 1996, 2001b, 2004). Far from being
mindless, our participants were creative and thoughtful as group members,
continuously striving to shape group and intergroup realities. However, if individuals
played a key part in shaping groups, the converse is equally true – the ability of
individuals to affect outcomes was dependent upon both the existence of groups and
their prototypicality within the group (Turner, 1991). To put it slightly differently,
individual agency was not destroyed by the group, but rather achieved through it
(Reicher & Haslam, in press). Thus, while personality and other individual difference
factors are an important aspect of our study (and we will consider them as part of our
analysis below), they cannot substitute for that analysis (Asch, 1952; Brown, 1965;
Sherif, 1966; Tajfel, 1978).
It follows that had different individuals participated in the study, we may well have
observed quite different outcomes from those reported above. This is because many
individual (and, indeed, chance) factors (such as the poor-quality breakfast on Day 8)
impacted on the fate of groups and their subsequent trajectory in the study. However, as
we explained in the Introduction, this is why we are not seeking to make empirical
generalizations on the basis of findings alone but rather are attempting to make
generalizations on the basis of the theoretical analysis that those findings support (Turner,
1981). Our claims to generality thus relate to (a) the effect of particular factors
The psychology of tyranny 27
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
(e.g. permeability, cognitive alternatives, roles) on group formation and (b) the
consequences of group success and failure for both individuals and the system as awhole.
The reality of inequality and power
A third potential criticism of our study is that the results and, in particular, the failure of
the guards to exert their authority (or the prisoners to accept it), reflects the twin facts
that (a) the study context was simply too pallid for our participants to become engaged
in it and (b) the guards had no authority to wield even if they wanted to. If, as Zimbardo
and his colleagues argue, tyranny is a function of groups and power, then one might
assert that our study lacked a meaningful degree of either and that we are therefore in no
position to say anything worthwhile about tyranny.
Let us consider these issues of engagement and power in turn. In the case of the
former, it is important to distinguish between engaging with the situation and engaging
with one’s group. The evidence suggests that the prisoners experienced and resented
their subordinate status from the start of the study. They disliked their food. They
disliked being locked up. They disliked the boredom. The smokers in particular disliked
being deprived of cigarettes. Indeed, after a day in his cell, PPp admitted that he was
‘falling for it [the experimental situation] hook, line and sinker’. At the first meal,
several prisoners expressed outrage at the poor quality of the food, especially given the
superior fare of the guards. The lack of group identification amongst prisoners in the
early days cannot therefore be put down to apathy about their plight.
Equally, the guards engaged with the situation from the start. Their initial
conversations focused on the power of the situation, the impact of wearing a guard’s
uniform, and the dangers of becoming tyrannical. At their first meal, they also noted
with discomfort the disparity in food quality and quantity. They then tried to ease this
discomfort by offering their leftovers to some of the prisoners. Therefore, if the guards
failed to act as a group, it was not because they did not care about the disparities in the
study but, on the contrary, because they were troubled by them. Their disidentification
did not reflect the fact that the study was not compelling, but rather that it was all too
compelling (albeit in very different ways from the SPE).
The issues are similar when it comes to the matter of power. It is certainly true that
the guards failed to exercise power. However, this was not because they had no power
to exercise but precisely because they had so much. In discussion, the guards
recognized the various options that were open to them. These included individual
punishments, collective punishments, removal of privileges, extra tasks and roll calls
and, most particularly, the power to promote a prisoner who would help them run the
system in the way they wanted. However, their fear of the guard identity – of being
authoritarian and of being seen as authoritarian – made them shun these options, to
promote a prisoner who embodied their ambivalence (Herriot, 2002), and even to give
away some of the sources of their power. In other words, the participants’ psychological
state explains their failure to exercise power much more than their lack of power
explains their psychological state (Turner, 2005). We will address the reasons for this
state shortly. For now, we simply note that our participants did engage with a situation in
which there were real inequalities of resources and power. Hence, the study can be
viewed as a meaningful exploration of the psychology of unequal groups.
The impact of interventions and key variables
It may be that our study says something of interest, it may be that it says something
interesting about group processes, it may even have something important to say about
Stephen Reicher and S. Alexander Haslam28
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
tyranny and resistance, but what of our claim to have conducted an ‘experimental case
study’ in which we investigated the impact of specific theoretically informed variables?
Can our interventions really be described as operationalizations of permeability and
cognitive alternatives? Are there not a host of other ways in which these interventions
and their consequences can be interpreted, and does this fact not cast doubt on any
particular theoretical gloss we might provide?
This is a reasonable concern. In the present study, there is certainly a disjunction
between the simplicity of our variables and the complexity of our interventions.
In particular, the promotion and the introduction of DMp could have had many effects,
such as making participants uncertain about what might happen next and making them
feel helpless and distrustful of the experimenters. However, such problems are not
unique to the present study – not least because the theoretical status of independent
variables is a potential concern in all experimental research (Haslam & McGarty, 2004).
Indeed, whenever researchers claim that a concrete instance (a specific operationalization
of an independent variable) reflects an abstract generality (the variable as it
operates in the world at large), interpretation is involved and alternatives are possible
(Billig, 1987). When the operationalization is as complex as ours, however, the space for
argument is all the greater.
In experimental practice, this issue is normally addressed in two ways (Haslam &
McGarty, 2003). The first is through the use of manipulation checks. As we have seen
above (Figure 3), checks of this form provide some evidence consistent with our
theoretical interpretation (specifically in relation to the existence of cognitive
alternatives). Moreover, one of the advantages of having multiple data sources
(i.e. behavioural as well as psychometric) is that we are able to see whether participants
spontaneously reacted to our interventions in ways that were expected. We have
provided qualitative evidence to suggest that they did. Specifically, after the promotion
prisoners saw no point in placating the guards and realized that the only way to improve
their position was to challenge the system (Extract 1). Equally, we saw that DMp
transformed the way in which participants conceptualized their situation and their
options (Extract 2). As with any manipulation check, such data support the notion that
we did manipulate the variables of concern to us.
However, while manipulation checks exclude false positives, they cannot exclude all
possible false negatives – that is, additional ways in which the manipulation may have
had an impact. This is where the second way of reducing uncertainty about the status of
independent variables comes into play. Here, it is not enough merely to argue for the
importance of a particular independent variable (or a particular confound; e.g.
personality, surveillance, the passing of time). It is also necessary to explain how that
variable (or confound) might plausibly and comprehensively account for the patterns of
observed results. That is, it is necessarily to have an integrated theoretical account of
how the effects occurred. As we explained when introducing them, in this respect, our
claims accord with a well-developed and extremely influential theoretical tradition (after
Tajfel & Turner, 1979) and with a well-established and robust pattern of previous
experimental findings (e.g. Ellemers, 1993; Wright et al., 1990). This theoretical and
empirical tradition lends coherence and plausibility to our analysis.
Having said that, we certainly do not view our analysis as inviolate, and we fully
accept that our account of the impact of particular interventions is open to argument.
However, for critics who wish to challenge this account, it is incumbent upon them to
do as we have and explain how an alternative understanding of our interventions
provides a better understanding of the study’s findings in their entirety.
The psychology of tyranny 29
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
With this in mind, it is time to move beyond potential criticisms and to be explicit
about what our explanation of events is. This involves three elements. The first relates to
the conditions of social identification, the second addresses the consequences of social
identification, and the third concerns reactions to group failure.
A social identity account of tyranny
The conditions of social identification
The simplest and clearest finding of our study is that people do not automatically assume
roles that are given to them in the manner suggested by the role account that is typically
used to explain events in the SPE (Haney et al., 1973; Zimbardo, 1989). Instead, there
are a range of factors that determine whether people themselves identify with the social
positions to which they are ascribed by others. Some of those factors operated in ways
specified by social identity theory (Tajfel & Turner, 1979). Thus, the shift from
permeability to impermeability of group boundaries had a strong impact on prisoner
identification (along lines previously demonstrated in separate programmes of research
by Ellemers and Wright and their colleagues; Ellemers, van Knippenberg, De Vries, &
Wilke, 1988; Ellemers et al., 1990, 1993; Wright, 1997; Wright & Taylor, 1998, Wright
et al., 1990; see also Lalonde & Silverman, 1994). Equally, the increasing insecurity of
intergroup relations affected the willingness of the prisoners to work as a group in order
to challenge the guards.
In addition, changing perceptions of security, and of the component of legitimacy in
particular, help explain not only why the prisoners challenged the guards, but also why
the participants as a whole did not challenge the experimenters, despite the adverse
conditions that living in the prison imposed upon them (cf. Spears & Smith, 2001). Most
of the time, when participants discussed the matter, they saw these conditions as a
legitimate part of the study in which they had agreed to participate. On only two
occasions were the conditions seen to violate that contractual legitimation – when the
new prisoner, DMp, began to question the heat in the prison, and when one night, the
experimenters failed to replenish basic supplies for the guards. On both occasions, this
led to explicit consideration of ‘mutiny’.
However, if the study supported social identity theory in demonstrating the role of
contextual factors in moderating the relationship between role and identity, it also
shows the importance of extending the way in which we conceptualize the nature of
context and its relation to human action. Thus, although the role of guard was positively
valued in the immediate context of the prison, those assigned to this position were
concerned with the possibility of negative evaluations by future audiences, and hence
some of them were reluctant to identify with the role.
It is here that the prospective televising of the study most clearly had an impact on
the behaviour of participants and serves to raise important practical and theoretical
points. Practically, it demonstrates how extreme behaviours can be restrained by
rendering actors visible, and hence accountable, to broader or yet-to-be encountered
audiences (Postmes, Spears, Lea, & Reicher, 2000; Reicher & Levine, 1994; Tedeschi,
1981; Tetlock, 1985).4
Theoretically, it shows that context, for human beings, cannot
4
Some commentators have argued that this analysis is inconsistent with evidence arising from the torture of inmates by US
troops at Abu Ghraib prison in 2004. However, as we have argued elsewhere (Reicher & Haslam, 2004), it appears that these
acts were performed for the benefit of an in-group audience that (like the experimenters in the SPE) was assumed by the
perpetrators to approve of the action.
Stephen Reicher and S. Alexander Haslam30
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
simply be assumed to be where we are in the here and now. People have imaginations
that allow them to think of other times and places, and to evaluate their behaviour in
relation to these. Indeed, surveillance has an impact on behaviour precisely through its
capacity to make salient such alternative contexts and alternative audiences (Reicher &
Haslam, 2003; Reicher & Levine, 1994).
Two points follow from this. First, under conditions of surveillance, individual
differences relating to the different social commitments people have outside a given
context have the capacity to impact upon behaviour within the context. Thus, in our
study, the fact that TQg was publicly visible as a successful entrepreneur who owned a
liberal enterprise goes some way towards explaining why he was particularly
ambivalent about adopting the identity of guard. However, if individual factors were
important here, then they operated in a paradoxical way. First, TQg’s very effectiveness
outside the study is what rendered him ineffective within it. Second, once human
imaginative capacities are acknowledged, forms of extreme situationism (such as the
role account), which suggest that behaviour is always dominated by the present
context, become untenable. Instead, it is important to take past and future contexts into
account as well (Reicher & Hopkins, 2001a).
The consequences of social identification
One of the distinctive features of the present study was that its intensive nature allowed
established disciplinary boundaries to be transcended so that it was possible to
investigate the interrelationships between social, organizational and clinical variables.
The results presented above point to the existence of such links and to their richness
and complexity.
On the organizational side, the achievement of a common social identity was seen to
be necessary in order for group members to trust others to act appropriately, to support
others in their actions and to expect support from them in return (Haslam et al., 2005).
In this way, the rudiments of effective organization (e.g. task differentiation, delegation,
leadership, trust) were seen to be contingent upon shared social identification in a
manner argued by a range of organizational theorists (e.g. Ellemers, De Gilder, & Haslam,
2004; Haslam, 2001; Haslam, van Knippenberg, Platow, & Ellemers, 2003; Hogg & Terry,
2001; Postmes, 2003; Smith, Tyler, & Huo, 2003; Turner & Haslam, 2001; Tyler & Blader,
2000). For this reason, the effectiveness of the prisoners, who were able to develop a
sense of shared social identity, contrasted markedly with the ineffectiveness of the
guards, who were not (Drury & Reicher, 2005).
On the clinical side, the increasing support amongst prisoners and their successful
challenges to the guards contributed to mental states that were increasingly positive. On
the other hand, the atomization of the guards and their failure to fulfil their collective
tasks led to increasingly negative states. This is reflected in the depression data (and also
on other measures not reported here including paranoia, anxiety and burnout; see
Haslam & Reicher, in press). Although the nature of our quantitative data does not allow
for analysis of causal relations between these variables (Haslam & McGarty, 2004), we
can say with confidence that participants’ mental states evolved in relation to the social
dynamics between groups, and that variables such as social support and collective selfefficacy
are implicated in that relationship (e.g. as argued by Branscombe, Schmitt, &
Harvey, 1999; Cohen & Wills, 1985; Contrada & Ashmore, 2000; Hall & Cheston, 2002;
Haslam et al., 2005; Orford, 1992; Schwarzer, 2001; Terry, Carey, & Callan, 2001).
Clearly though, the exact nature of the relationship merits further investigation.
The psychology of tyranny 31
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Reactions to group failure
There were two instances of group failure in the study. The first concerned the failure of
the guards to fulfil their assigned task of making a hierarchical prison system work, and
the second concerned the failure of the commune’s supporters to establish a viable
egalitarian social order.
It is notable that in each case, group members became more willing to accept a
system that promised to be viable even if it meant ceding some of the core principles of
the group. On Day 5, after the introduction of the new prisoner, the guards were willing
to cede their authority and accept a more equal social system and on Day 8 commune
supporters were prepared to cede equality and tolerate a more hierarchical social
system. Thus, rather than people ‘naturally’ preferring any given form of social order, it
appears that, when group members fail to impose an order based on their own existing
norms and values, they are willing to adapt those values (or to adopt new ones) in order
to create a viable order rather than have no order at all.
In this regard, the fact that the participants’ authoritarianism increased
significantly over time constitutes one of the most important findings of the study
as a whole. Traditionally, authoritarianism has been viewed a stable individual
difference variable that has the capacity to explain the emergence of hierarchical and
tyrannical social structures (Adorno et al., 1950; Altmeyer, 1981, 1996; for reviews
see Billig, 1978; Brown, 1965; Duckitt, 1994). However, in contrast to this analysis,
the present study illustrates that authoritarianism is a variable outcome of social
structure (see also Oakes, Haslam, & Turner, 1994; Reynolds et al., 2001). More
specifically, we see that authoritarian solutions – and the personalities that would
promote them – appeared more attractive after attempts to make democracy work
were seen to have failed.
Against this argument, it might be suggested that the system we observed moved
towards tyranny simply because the more authoritarian individuals were waiting for
an opportunity to impose their preferred system on others (indeed, a similar point
could be made in relation to the SPE). However, along lines intimated above, two
points speak against this conclusion and against related arguments couched in terms
of individual personality-based dynamics. The first is that tyranny arose out of the
social (dis)order of the prison, and authoritarianism (like other dimensions of raw
personality) was clearly not a straightforward determinant of that (dis)order. The
second is that although it may be true that those who proposed the new regime were
those who had initially been more disposed towards authoritarianism, their influence
and leadership was clearly contingent upon their views having become more
representative of the views of the participants as a whole as these had evolved over
the course of the study (for more in-depth treatment of this issue see Haslam &
Reicher, 2005; Reicher & Haslam, in press; see also Haslam, 2001; Haslam & Platow,
2001; Navas, Morales, & Moya, 1992; Reicher & Hopkins, 1996, 2001b; Turner, 1991;
Turner & Haslam, 2001; Turner et al., in press). Indeed, the interaction between final
group membership and study phase suggests that, if anything, the authoritarianism of
the guards in the vaunted new regime had declined over time, whereas that of other
participants was increasing (see Figure 9). It also is worth adding that previous
research has been unable to observe dynamics of this form at work or to generate
empirical support for arguments such as these, precisely because it has been unable
to examine the impact of evolving group history on both individual psychology and
social structure (Haslam & McGarty, 2001; Levine, 2003).
Stephen Reicher and S. Alexander Haslam32
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Conclusion: Rethinking the relationship between groups, power and tyranny
At one level, our study confirms the findings of the SPE. It shows that an understanding
of collective conflict and tyranny cannot be achieved simply by looking at individuals
but requires an analysis of group processes and intergroup relations. In this sense,
we agree with Zimbardo (and many others; e.g. Asch, 1952; Sherif, 1966; Tajfel, 1978)
that such phenomena can only be explained through group-level analysis. Our
disagreement with prior analysis of the SPE thus relates to the nature of group processes
and of the conditions under which they lead to social pathologies.
As almost every psychology student (and an unusually large proportion of the
general public) knows, the message of the SPE is that the toxic combination of groups
and power leads to tyranny. The implications of the BBC prison study are different.
In common with recent theoretical developments in social psychology, they contest the
premise that group behaviour is necessarily uncontrolled, mindless and antisocial
(Ellemers et al., 1999; Oakes et al., 1994; Postmes et al., 2000; Reicher, 1982, 2001;
Spears, Oakes, Ellemers, & Haslam, 1997; Turner, 1999). In contrast, the results of the
BBC prison study suggest that the way in which members of strong groups behave
depends upon the norms and values associated with their specific social identity and
may be either anti- or prosocial (Jetten, Spears, & Manstead, 1997).
However, based on the present data, we would argue that failing groups almost
inevitably create a host of problems for their own members and for others. These
problems have a deleterious impact on organization, on individuals’ clinical state, and –
most relevant here – on society. For it is when people cannot create a social system for
themselves that they will more readily accept extreme solutions proposed by others. It is
when groups lack the power to exercise choice that an authoritarian ideology that
promises to create order for them appears more seductive. In short, it is the breakdown
of groups and powerlessness that creates the conditions under which tyranny can
triumph (for related arguments see Kanter, 1979; Pfeffer, 1981; Reynolds & Platow,
2003)
We would argue that as well as being consistent with contemporary thinking in
social psychology (e.g. after Tajfel & Turner, 1979), this analysis also articulates more
closely than Zimbardo’s original role account with the analysis of tyranny put forward by
researchers in other academic disciplines. Most notably, it accords with influential
analyses proposed by modern historians (e.g. Abel, 1986; Gellately, 2001; Hobsbawm,
1995; Rees, 2002). Consider, for instance, Hobsbawm’s account of the conditions that
gave rise to the fall of the Weimar republic and the emergence of Nazism in 1930s
Germany:
The optimal conditions for the triumph of the ultra-right were an old state and its ruling
mechanisms which could no longer function; a mass of disenchanted, disoriented and
disorganized citizens who no longer knew where their loyalties lay; strong socialist
movements threatening or appearing to threaten social revolution, but not actually in a
position to achieve it.... These were the conditions that turned movements of the radical right
into powerful, organized and sometimes uniformed and paramilitary force (1995, p. 127).
We would also argue that this analysis can be used to make sense of what happened in
the SPE when one looks more closely at the events that unfolded there. As we have
noted, that study, like ours, appears to have started off with prisoners threatening to
become ascendant over the guards. Things changed when Zimbardo intervened in such
a way as to lead the prisoners to believe that they could not leave the study. At this point
they became disoriented as to their position – in Zimbardo’s own words, they
The psychology of tyranny 33
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
experienced ‘role confusion’ (Zimbardo, 1989). They ceased to support each other
against the guards. They collapsed as a group and this allowed tyrannical guards to
prevail. Thus, whereas in the SPE, the failure of the prisoners allowed an existing
tyranny to be consolidated, in the BBC study, the failure of the commune paved the way
for the emergence of a new tyranny.
These, of course, are large and controversial claims. However, they have important
theoretical implications along with considerable practical implications. They point to
new ways of thinking both about the nature of group psychology and about the
psychological underpinnings of tyranny. Because the scope of these analyses is so large,
it would be both presumptuous and foolish to expect them to be accepted on the basis
of a single data set – especially one reflecting the complexities of the present study. Yet
what is true of our study is largely true of the SPE. As we have argued, one of the great
problems arising from that study was that it made strong claims (that had an enormous
impact on public consciousness) but then led to debate being closed off because further
research was declared to be ethically unacceptable.
Hence, irrespective of our results and of our analysis, we would claim that one of the
significant achievements of the BBC prison study is to show that, if sufficient care is
taken, it is possible to run powerful and influential field studies into social processes that
are also ethical. In a field increasingly dominated by reductionist accounts of human
behaviour, such studies can restore balance by demonstrating the impact of systematic
variations in social relations upon human behaviour. The richness, immediacy and
relevance of their findings can also help reconnect psychologists with policy makers.
For these reasons, we offer our conclusions not so much as a final word on the matter,
but more in the hope of stimulating researchers to renew their interest in the important
debates that our discipline can, and must, advance.
Acknowledgements
We would like to thank the participants as well as the staff of the BBC for their commitment to this
research and for comments on ideas in this paper. The assistance and input of Nyla Branscombe,
Lloyd Carson, David Corner, Andrew Eagle, Liberty Eaton, Scott Galloway, Nick Hopkins, Mike
Howe, Jolanda Jetten, Andrew Livingstone, Eva Loth, Grant Muir, Lucy O’Sullivan, Michael
Schmitt, Heather Smith, Tom Postmes, Fabio Sani, Denis Sindic, Stephanie Sonnenberg, Paul
Webley, Stephen Wilks, Huw Williams and Brian Young is also gratefully acknowledged, together
with detailed and very helpful comments from several anonymous reviewers.
References
Abel, T. (1986). Why Hitler came to power. Cambridge, MA: Harvard University Press.
Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian
personality. New York: Harper.
Altmeyer, B. (1981). Right wing authoritarianism. Winnipeg: University of Manatoba Press.
Altmeyer, B. (1996). The authoritarian specter. Cambridge, MA: Harvard University Press.
Arendt, H. (1998). The human condition. Chicago: University of Chicago Press.
Asch, S. E. (1952). Social psychology. Engelwood Cliffs, NJ: Prentice Hall.
Bandura, A. (1995). Exercise of personal and collective efficacy in changing societies. In A. Bandura
(Ed.), Self-efficacy in changing societies (pp. 1–45). New York: Cambridge University Press.
Banuazizi, A., & Movahedi, S. (1975). Interpersonal dynamics in a simulated prison:
A methodological analysis. American Psychologist, 30, 152–160.
Stephen Reicher and S. Alexander Haslam34
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Baron, R. S. (1984). Group dynamics. In A. S. Kahn (Ed.), Social psychology (pp. 260–289).
Dubuque, IA: W.C. Brown.
Bar Tal, D. (2004). The necessity of observing real life situations: Palestinian–Israeli violence as a
laboratory for learning about social behaviour. European Journal of Social Psychology, 34,
677–701.
Billig, M. (1978). Fascists: A social psychological view of the National Front. London: Academic
Press.
Billig, M. (1987). Arguing and thinking. Cambridge: Cambridge University Press.
Branscombe, N. R., Schmitt, M. T., & Harvey, R. D. (1999). Perceiving pervasive discrimination
among African Americans: Implications for group identification and well-being. Journal of
Personality and Social Psychology, 77, 135–149.
Brockes, E. (2001, October 16). The Experiment. The Guardian, G2, pp. 1–3.
Brown, R. (1965). Social psychology. New York: Free Press.
Chen, G., & Bliese, P. D. (2002). The role of different levels of leadership in predicting self- and
collective efficacy: Evidence for discontinuity. Journal of Applied Psychology, 87, 549–556.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences. New York: Wiley.
Cohen, S., & Wills, T. A. (1985). Stress, social support and the buffering hypothesis. Psychological
Bulletin, 98, 310–357.
Contrada, R. J. & Ashmore, R. D. (Eds.), (2000). Self, social identity, and physical health:
Interdisciplinary explorations. New York: Oxford University.
Cronin, P. & Reicher, S. (2006). A study of the factors that influence how senior officers
police crowd events: On SIDE outside the laboratory. British Journal of Social Psychology,
45, 175–196.
Doosje, B., Ellemers, N., & Spears, R. (1995). Perceived intragroup variability as a function of
group status and identification. Journal of Experimental Social Psychology, 31, 410–436.
Doosje, B., Spears, R., & Ellemers, N. (2002). The dynamic and determining nature of group
identification: Responses to anticipated changes in the status hierarchy. British Journal of
Social Psychology, 41, 57–76.
Drury, J., & Reicher, S. (2005). Explaining enduring empowerment: A comparative study of
collective action and psychological outcomes. European Journal of Social Psychology, 35,
35–58.
Duckitt, J. (1994). The social psychology of prejudice (2nd ed). New York: Preager.
Eggins, R. A., Haslam, S. A., & Reynolds, K. J. (2002). Social identity and negotiation: Subgroup
representation and superordinate consensus. Personality and Social Psychology Bulletin, 28,
887–899.
Ellemers, N. (1993). The influence of socio-structural variables on identity enhancement
strategies. European Review of Social Psychology, 4, 27–57.
Ellemers, N., De Gilder, D., & Haslam, S. A. (2004). Motivating individuals and groups at work:
A social identity perspective on leadership and group performance. Academy of Management
Review, 29, 459–478.
Ellemers, N., Spears, R., & Doosje, B. (1999). Social identity: Context, content and commitment.
Oxford: Blackwell.
Ellemers, N., van Knippenberg, A., De Vries, N. K., & Wilke, H. (1988). Social identification and
permeability of group boundaries. European Journal of Social Psychology, 18, 497–513.
Ellemers, N., van Knippenberg, A., & Wilke, H. (1990). The influence of permeability of group
boundaries and stability of group status on strategies of individual mobility and social change.
British Journal of Social Psychology, 29, 233–246.
Ellemers, N., Wilke, H., & van Knippenberg, A. (1993). Effects of the legitimacy of low group or
individual status on individual and collective identity enhancement strategies. Journal of
Personality and Social Psychology, 64, 766–778.
Emler, N. (1994). Gossip, reputation, and social adaptation. In B. F. Goodman & A. Ben-Ze’ev
(Eds.), Good gossip (pp. 117–138). Lawrence: University of Kansas Press.
The psychology of tyranny 35
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J. F. (1989). Reducing intergroup bias: The benefits
of recategorization. Journal of Personality and Social Psychology, 57, 239–249.
Gellately, R. (2001). Backing Hitler. Oxford, UK: Oxford University Press.
Hall, S., & Cheston, R. (2002). Mental health and identity: The evaluation of a drop-in centre.
Journal of Community and Applied Social Psychology, 12, 30–43.
Haney, C., Banks, C., & Zimbardo, P. (1973). A study of prisoners and guards in a simulated prison.
Naval Research Review 9, 1–17 [Reprinted in E. Aronson (Ed.), Readings about the social
animal (3rd ed., pp. 52–67). San Francisco, CA: W.H. Freeman].
Haslam, S. A. (2001). Psychology in organizations: The social identity approach. London: Sage.
Haslam, S. A., Eggins, R. A., & Reynolds, K. J. (2003). The ASPIRe model: Actualizing Social and
Personal Identity Resources to improve strategic planning and encourage sustainable
productivity in organizations. Journal of Occupational and Organizational Psychology, 76,
83–113.
Haslam, S. A., O’Brien, A., Jetten, J., Vormedal, K. & Penna, S. (2005). Taking the strain: Social
identity, social support and the experience of stress. British Journal of Social Psychology, 44,
355–370.
Haslam, S. A., & McGarty, C. (2001). A hundred years of certitude? Social psychology, the
experimental method and the management of scientific uncertainty. British Journal of Social
Psychology, 40, 1–21.
Haslam, S. A., & McGarty, C. (2003). Research methods and statistics in psychology. London:
Sage.
Haslam, S. A., & McGarty, C. (2004). Experimental design and causality in social psychological
research. In C. Sanson, C. C. Morf, & A. T. Panter (Eds.), Handbook of methods in social
psychology (pp. 235–264). Thousand Oaks, CA: Sage.
Haslam, S. A., & Platow, M. J. (2001). The link between leadership and followership: How affirming
a social identity translates vision into action. Personality and Social Psychology Bulletin, 27,
1469–1479.
Haslam, S. A., Powell, C., & Turner, J. C. (2000). Social identity, self-categorization and work
motivation: Rethinking the contribution of the group to positive and sustainable
organizational outcomes. Applied Psychology: An International Review, 49, 319–339.
Haslam, S. A., & Reicher, S. (2002). A user’s guide to The Experiment – Exploring the psychology
of groups and power: Manual to accompany the BBC video. London: BBC Worldwide.
Haslam, S. A., & Reicher, S. D. (2005). Identity entrepreneurship and the consequences of identity
failure: The dynamics of leadership in the BBC Prison Study. Manuscript submitted for
publication.
Haslam, S. A., & Reicher, S. D. (in press). Stressing the group: Social identity and the unfolding
dynamics of stress. Journal of Applied Psychology.
Haslam, S. A., van Knippenberg, D., Platow, M. & Ellemers, N. (Eds.), (2003). Social identity at
work: Developing theory for organizational practice. Philadelphia, PA: Psychology Press.
Herrera, C. D. (1997). A historical interpretation of deceptive experiments in American
psychology. History of the Human Sciences, 10, 23–26.
Herriot, P. (2002). Selection and self: Selection as a social process. European Journal of Work and
Organizational Psychology, 11, 385–402.
Hobsbawm, E. (1995). Age of extremes: The short twentieth century 1914–1991. London:
Abacus.
Hogg, M. A., & Terry, D. J. (Eds.), (2001). Social identity processes in organizational contexts.
Philadelphia, PA: Psychology Press.
House, J. S. (1981). Work stress and social support. Reading, MA: Addison-Wesley.
Hoyle, R. H., Georgesen, J. C., & Webster, J. M. (2001). Analysing data from individuals in groups:
The past, the present, the future. Group Dynamics: Theory, Practice and Research, 5, 41–47.
Ito, T. A., Thompson, E., & Cacioppo, J. T. (2004). Tracking the timecourse of social perception:
The effects of racial cues on event-related brain potentials. Personality and Social Psychology
Bulletin, 30, 1267–1280.
Stephen Reicher and S. Alexander Haslam36
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Jetten, J., Spears, R., & Manstead, A. R. (1997). Strength of identification and intergroup
differentiation: The influence of group norms. European Journal of Social Psychology, 27,
603–609.
Kanter, R. (1979). Power failure in management circuits. Harvard Business Review, July–August,
65–75.
Koppel, G. (Series producer), & Mirsky, N. (Executive producer). (2002, May 14, 15, 20, 21).
The Experiment. London: British Broadcasting Corporation.
Lalonde, R. N., & Silverman, R. A. (1994). Behavioral preferences in response to social injustice:
The effects of group permeability and social identity salience. Journal of Personality and
Social Psychology, 66, 78–85.
Laudat, M. H., Cerdas, S., Dournier, C., Guiban, D., Guilhaume, B., & Luton, J. P. (1988). Salivary
cortisol measurement: A practical approach to assess pituitary-adrenal function. Journal of
Clinical Endocrinology and Metabolism, 66, 343–348.
LeBon, G. (1895, translated 1947). The crowd: A study of the popular mind. London: Ernest Benn.
Levine, R. M. (2003). Times, theories and practices in social psychology. Theory and Psychology,
13, 53–72.
Leyens, J. P., Cortes, B., Demoulin, S., Dovidio, J. F., Fiske, S. T., Gaunt, R., Paladino, M. P.,
Rodriguez-Perez, A., Rodriguez-Torres, R., & Vaes, J. (2003). Emotional prejudice, essentialism,
and nationalism. European Journal of Social Psychology, 33, 703–717.
Lindsay, R. C. L., & Adair, R. G. (1990). Do ethically recommended research procedures influence
the perceived ethicality of social psychological research? Canadian Journal of Behavioral
Science, 22, 282–294.
Lovibond, S. H., Mithiran, X., & Adams, W. G. (1979). The effects of three experimental prison
environments on the behaviour of non-convict volunteer subjects. Australian Psychologist,
14, 273–287.
Lyon, D. (1994). The electronic eye. Oxford: Polity.
Lyon, D., & Zuriek, E. (1996). Computers, surveillance and privacy. Minneapolis: University of
Minnesota Press.
McDermott, M., O¨pik, L., Smith, S., Taylor, S., & Wills, A. (2002). The Experiment: Report of the
independent ethics panel. http://www.experimentethics.org.uk.
McGarty, C., & Smithson, M. (2005). Independence and non-independence: A simple test for
comparing groups using multiple measures and the binomial test. European Journal of Social
Psychology, 35, 171–180.
Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal and Social Psychology,
67, 371–378.
Milgram, S. (1974). Obedience to authority: An experimental view. New York: Harper and Row.
Moreland, R. L., Argote, L., & Krishnan, R. (1996). Socially shared cognition at work: Transactive
memory and group performance. In J. Nye & A. Brower (Eds.), What’s social about social
cognition? Research on socially shared cognition in small groups (pp. 57–84). Newbury
Park, CA: Sage.
Morgan, A. H. (1979). Comments on S. H. Lovibond, X. Mithiran and W. G. Adams “The effects of
three experimental prison environments on the behaviour of non-convict volunteer subjects”.
Australian Psychologist, 14, 285–286.
Navas, M., Morales, J. M., & Moya, M. C. (1992). Categorization and prototypicality as determinants
of leader effectiveness. International Journal of Psychology, 27, 289–290.
Norusis, M. J. (1985). SPSSx: Advanced statistics guide. New York: McGraw-Hill.
Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social reality. Oxford:
Blackwell.
Orford, J. (1992). Community psychology. Chichester: Wiley.
Organ, D. W. (1988). Organizational citizenship behavior: The good soldier syndrome.
Lexington, MA: Lexington.
Organ, D. W. (1997). Organizational citizenship behavior: Its construct clean-up time. Human
Performance, 10, 85–97.
The psychology of tyranny 37
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Pfeffer, J. (1981). Power in organizations. Boston, MA: Pitman.
Postmes, T. (2003). A social identity approach to communication in organizations. In S. A. Haslam,
D. van Knippenberg, M. J. Platow, & N. Ellemers (Eds.), Social identity at work: Developing
theory for organizational practice. Philadelphia, PA: Taylor & Francis.
Postmes, T. Baray, G. Haslam, S. A., & Morton, T. (in press). Personal and social identity formation
through interaction: Individuality in small groups, social movements and categories.
Individuality and the group: Advances in social identity. London: Sage.
Postmes, T., & Spears, R. (1998). Deindividuation and anti-normative behaviour: A meta-analysis.
Psychological Bulletin, 123, 238–259.
Postmes, T., Spears, R., Lea, M., & Reicher, S. D. (2000). SIDE issues centre stage. Amsterdam:
Royal Netherlands Academy of Arts and Sciences.
Rees, L. (2002). The Nazis: A warning from history. London: BBC Worldwide.
Reicher, S. D. (1982). The determination of collective action. In H. Tajfel (Ed.), Social identity and
intergroup relations (pp. 41–84). Cambridge: Cambridge University Press.
Reicher, S. D. (1996). Social identity and social change: Rethinking the context of social
psychology. In P. Robinson (Ed.), Social groups and identities: Developing the legacy of Henri
Tajfel. Oxford: Butterworth-Heinemann.
Reicher, S. D. (2001). The psychology of crowd dynamics. In M. Hogg & S. Tindale (Eds.),
Blackwell handbook of social psychology: Group processes (pp. 175–197). Oxford:
Blackwell.
Reicher, S. D. (2004). The context of social identity: Domination, resistance and change. Political
Psychology, 25, 921–946.
Reicher, S. D., Drury, J., Hopkins, N., & Stott, C. (2001). A model of crowd prototypes and crowd
leadership. In C. Barker, M. Lavalette, & A. Johnson (Eds.), Leadership and social movements.
Manchester: Manchester University Press.
Reicher, S. D., & Emler, N. (1985). Delinquent behaviour and attitudes to formal authority. British
Journal of Social Psychology, 24, 161–168.
Reicher, S. D., & Haslam, S. A. (2003). Social psychology, science, and surveillance: Understanding
The Experiment. Social Psychology Review, 5, 7–17.
Reicher, S. D., & Haslam, S. A. (2004). Why not everyone is a torturer. Retrieved May 10, from:
http://news.bbc.co.uk/1/hi/magazine/3700209.stm.
Reicher, S. D., & Haslam, S. A. (in press). On the agency of individuals and groups: Lessons from the
BBC Prison Study. In T. Postmes & J. Jetten (Eds.) Individuality and the group: Advances in
social identity. London: Sage.
Reicher, S. D., & Hopkins, N. (1996). Seeking influence through characterising self-categories: An
analysis of anti-abortionist rhetoric. British Journal of Social Psychology, 35, 297–311.
Reicher, S. D., & Hopkins, N. (2001a). Psychology and the end of history: A critique and a proposal
for the psychology of social categorization. Political Psychology, 22, 383–407.
Reicher, S. D., & Hopkins, N. (2001b). Self and nation: Categorization, contestation and
mobilisation. London: Sage.
Reicher, S. D., & Hopkins, N. (2004). On the science of the art of leadership. In D. van
Knippenberg & M. A. Hogg (Eds.), Leadership, power and identity (pp. 197–209). London:
Sage.
Reicher, S. D., & Levine, M. (1994). Deindividuation, power relations between groups and the
expression of social identity: The effects of visibility to the outgroup. British Journal of Social
Psychology, 33, 145–163.
Reicher, S. D., Spears, R., & Postmes, T. (1995). A social identity model of deindividuation
phenomena. European Review of Social Psychology, 6, 161–198.
Reynolds, K. J., & Platow, M. J. (2003). Why power in organizations really should be shared:
Understanding power through the perils of powerlessness. In S. A. Haslam, D. van
Knippenberg, M. J. Platow & N. Ellemers (Eds.), Social identity at work: Developing theory
for organizational practice. Philadelphia, PA: Psychology Press.
Stephen Reicher and S. Alexander Haslam38
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Reynolds, K. J., Turner, J. C., Haslam, S. A., & Ryan, M. K. (2001). The role of personality and group
factors in explaining prejudice. Journal of Experimental Social Psychology, 37, 427–434.
Robinson, W. P. (Ed.). (1996). Social groups and identities: Developing the legacy of Henri Tajfel.
Oxford: Butterworth-Heinemann.
Schwarzer, R. (2001). Stress, resources, and proactive coping. Applied psychology:
An international review, 50, 400–407.
Sherif, M. (1956). Experiments in group conflict. Scientific American, 195, 54–58.
Sherif, M. (1966). Group conflict and co-operation: Their social psychology. London: Routledge
and Kegan Paul.
Sieber, J. E., Iannuzzo, R., & Rodriguez, B. (1995). Deception methods in psychology: Have they
changed in 23 years? Ethics and Behavior, 5, 67–85.
Smith, E. R., & Mackie, D. M. (2000). Social psychology (2nd ed). New York: Worth.
Smith, H. J., Tyler, T. R., & Huo, Y. (2003). Interpersonal treatment, social identity and
organizational behaviour. In S. A. Haslam, D. van Knippenberg, M. J. Platow, & N. Ellemers
(Eds.), Social identity at work: Developing theory for organizational practice. Philadelphia,
PA: Psychology Press.
Smith, S. S., & Richardson, D. (1983). Amelioration of deception and harm in psychological
research. Journal of Personality and Social Psychology, 44, 1075–1082.
Smithson, M. (2000). Statistics with confidence. London: Sage.
Spears, R., Oakes, P. J., Ellemers, N., & Haslam, S. A. (Eds.), (1997). The social psychology of
stereotyping and group life. Oxford: Blackwell.
Spears, R., & Smith, H. J. (2001). Experiments as politics. Political Psychology, 22, 309–330.
Staub, E. (1989). The roots of evil: The origins of genocide and other group violence. Cambridge:
Cambridge University Press.
Tajfel, H. (1978). Interpersonal behaviour and intergroup behaviour. In H. Tajfel (Ed.),
Differentiation between social groups: Studies in the social psychology of intergroup
relations (pp. 27–60). London: Academic Press.
Tajfel, H. (Ed.). (1982). Social identity and intergroup relations. Cambridge: Cambridge
University Press.
Tajfel, H., Flament, C., Billig, M. G., & Bundy, R. F. (1971). Social categorization and intergroup
behaviour. European Journal of Social Psychology, 1, 149–177.
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin &
S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33–47). Monterey, CA:
Brooks/Cole.
Tedeschi, J. T. (Ed.). (1981). Impression management theory and social psychological research.
New York: Academic Press.
Terry, D. J., Carey, C. J., & Callan, V. J. (2001). Employee adjustment to an organizational merger: An
intergroup perspective. Personality and Social Psychology Bulletin, 27, 267–290.
Tetlock, P. E. (1985). Accountability: The neglected social context of judgement and choice.
Research in Organizational Behavior, 7, 297–332.
Thayer, S., & Saarni, C. (1975). Demand characteristics are everywhere (anyway): A comment on
the Stanford Prison Experiment. American Psychologist, 30, 1015–1016.
Turner, J. C. (1981). Some considerations in generalizing experimental social psychology.
In G. M. Stephenson & J. H. Davis (Eds.), Progress in applied social psychology (Vol. 1,
pp. 3–34). Chichester: Wiley.
Turner, J. C. (1982). Towards a cognitive redefinition of the social group. In H. Tajfel (Ed.), Social
identity and intergroup relations (pp. 15–40). Cambridge: Cambridge University Press.
Turner, J. C. (1985). Social categorization and the self-concept: A social cognitive theory of group
behaviour. In E. J. Lawler (Ed.), Advances in group processes (Vol. 2, pp. 77–122). Greenwich,
CT: JAI Press.
Turner, J. C. (1991). Social influence. Milton Keynes: Open University Press.
The psychology of tyranny 39
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Turner, J. C. (1999). Some current issues in research on social identity and self-cate-gorization
theories. In N. Ellemers, R. Spears, & B. Doosje (Eds.), Social identity: Context, commitment,
content (pp. 6–34). Oxford: Blackwell.
Turner, J. C. (2005). Explaining the nature of power: A three-process theory. European Journal of
Social Psychology, 35, 1–22.
Turner, J. C., & Brown, R. J. (1978). Social status, cognitive alternatives and intergroup relations. In
H. Tajfel (Ed.), Differentiation between social groups: Studies in the social psychology of
intergroup relations (pp. 201–234). London: Academic Press.
Turner, J. C., & Haslam, S. A. (2001). Social identity, organizations and leadership. In M. E. Turner
(Ed.), Groups at work: Advances in theory and research (pp. 25–65). Hills-dale, NJ: Erlbaum.
Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the
social group: A self-categorization theory. Oxford: Blackwell.
Turner, J. C., Oakes, P. J., Haslam, S. A., & McGarty, C. A. (1994). Self and collective: Cognition and
social context. Personality and Social Psychology Bulletin, 20, 454–463.
Turner, J. C., Reynolds, K. J., Haslam, S. A., & Veenstra, K. (in press). Reconceptualizing
personality: Producing individuality through defining the personal self. In T. Postmes & J. Jetten
(Eds.) Individuality and the group: Advances in social identity. London: Sage.
Tyler, T. R., & Blader, S. L. (2000). Cooperation in groups. Procedural justice, social identity, and
behavioral engagement. Philadelphia, PA: Psychology Press.
Underwood, P. W. (2000). Social support: The promise and reality. In B. H. Rice (Ed.), Handbook
of stress, coping and health (pp. 367–391). Newbury Park, CA: Sage.
Wright, S. C. (1997). Ambiguity, social influence, and collective action: Generating collective
protest in response to tokenism. Personality and Social Psychology Bulletin, 23, 1277–1290.
Wright, S. C., & Taylor, D. M. (1998). Responding to tokenism: Individual action in the face of
collective injustice. European Journal of Social Psychology, 28, 647–667.
Wright, S. C., Taylor, D. M., & Moghaddam, F. M. (1990). Responding to membership in a
disadvantaged group: From acceptance to collective protest. Journal of Personality and
Social Psychology, 58, 994–1003.
Zimbardo, P. (1969). The human choice: Individuation, reason and order versus deindividuation,
impulse and chaos. In W. J. Arnold & D. Levine (Eds.), Nebraska symposium on motivation
(Vol. 17, pp. 237–307). Lincoln, NE: University of Nebraska Press.
Zimbardo, P. (1989). Quiet rage: The Stanford prison study video. Stanford, CA: Stanford
University.
Zimbardo, P. (2001). Opposing terrorism by understanding the human capacity for evil. Monitor
on Psychology, 32, 48–50.
Zimbardo, P. G., Maslach, C., & Haney, C. (1999). Reflections on the stanford prison experiment:
Genesis, transformations, consequences. In T. Blass (Ed.), Obedience to authority: Current
perspectives on the Milgram paradigm (pp. 193–237). Mahwah, NJ: Erlbaum.
Received 6 October 2004; revised version received 22 April 2005
Stephen Reicher and S. Alexander Haslam40