Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Clinical Psychology Review 24 (2004) 513 – 528

Acting on command hallucinations and dangerous behavior:


A critique of the major findings in the last decade
Louise G. Braham *, Peter Trower, Max Birchwood
Department of Clinical Psychology, School of Psychology, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK

Received 22 October 2001; received in revised form 12 February 2002; accepted 20 April 2004

Abstract

Command hallucinations (CH) have recently become the focus of research into positive symptoms of
schizophrenia. The importance of CH has become clear for theoretical and practical reasons, because CH are
regarded as potentially the most dangerous symptoms of schizophrenia. The aim is to critically review research that
has attempted to better understand CH and develop theories that may predict behavior and shape psychological
treatments. This review draws together, presents, and critically discusses the current disparate body of literature
produced (1990 – 2000) considering compliance with CH.
The literature can be grouped broadly into three overlapping classes. (1) Studies testing for a relationship
between CH and compliance. (2) Studies considering factors associated with acting on CH (mediating variables).
(3) Studies that look for a relationship between CH and dangerous behavior.
The body of literature is in its infancy and marred by methodological difficulties. Researchers have asked
differing questions about different samples while hoping to provide similar answers. While acknowledging
difficulties in interpreting the literature, themes do emerge. The weight of the evidence is that some individuals
who hear commands will act on them. However, some studies show no link or only a weak link and although
methodological errors account for some of the inconsistent results, there are also clearly individual differences. The
relationship between command and action is more complex than many researchers have assumed. One conjecture
is that the relationship is established via psychological processes or mediating variables which include beliefs
about the voice and content of instruction. Prediction of compliance with commands is of further interest when
dangerous behavior results. It is postulated that the issue of dangerousness is subject to the same mediating
variables as other commands. It is clear however, that CH are not sufficient to produce action in isolation and that
psychological processes mediate the process.
D 2004 Published by Elsevier Ltd.

Keywords: Command hallucinations; Dangerous behavior; Compliance

* Corresponding author. Department of Clinical Psychology, Rampton Special Hospital, Retford Nottinghamshire, DN22
0PD, UK.

0272-7358/$ - see front matter D 2004 Published by Elsevier Ltd.


doi:10.1016/j.cpr.2004.04.002
514 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

1. Introduction

Considerable progress has been made in recent years into understanding the psychological and
interpersonal characteristics of hallucinations. In this context, one particular class of hallucinations,
namely, command hallucinations (CH) have become a particular focus of research attention. They
are relatively common in voice hearers with between 33% and 74% reporting such activity
(Birchwood & Chadwick, 1997; Rudnick, 1999; Thompson, Stuart, & Holden, 1992). CH are
characterized, as the term suggests, as commanding the hearer to do something. The action
instructed may be dangerous and can result in offending behavior. The predictability of dangerous
behavior depends on a number of biopsychological variables. This is particularly important as CH
are often drug resistant and may present a serious social threat to the individuals themselves or
others. CH are a significant but rather neglected area in schizophrenia. The key feature that
distinguishes them from ordinary hallucinations is that, phenomenologically, the voice is experienced
as commanding rather than commenting, and there is a significant probability that the hearer will
comply with the command. Assuming this is true, we need to understand the processes by which the
individual comes to act or indeed not act on the command. However, the literature that considers
CH and their relation to behavior is disparate and difficult to integrate. Thus, this review paper
summarizes the evidence for the link between command and compliance, considers the processes
involved in complying/not complying with a command, and comments on the issue of dangerous-
ness following commands. The authors reviewed studies produced over the last decade (1990–2000)
according to Medline, PSYCHLIT, and BIDS. The authors searched on combinations of the
following words: command/commanding/imperative; hallucinations/voices; risk/dangerousness; and
acting/compliance.
Because of the small number of papers specifically attending to CH and considerable differences
among reviewed articles in the methodology and parameters studied, a meta-analysis was not performed.
It is acknowledged that a computerized journal search cannot be comprehensive primarily because it
misses work in books; however, work in this field of study is mainly published in journals. Studies that
do not deal directly with CH, although indirectly touching on this issue and methodologically sound, are
not reviewed here. However, some are mentioned to place a theory or suggestion in context where
necessary.
Many people with auditory hallucinations experience instruction or command. Studies also
assert that CH are not always reported to staff and are consequently missed (e.g., Rogers, Gillis,
Turner, & Frise-Smith, 1990; Zisook, Byrd, Kuck, & Jeste, 1995). Thus, actual rates of CH may
be even higher than that reported in studies due to difficulties with underreporting of such
hallucinations. CH are often cited as causal factors for violent behavior, putting voice hearers
and others at risk. Clinical and empirical evidence also suggests that they can be distressing,
with such distress linked with uncontrollability (Close & Garety, 1998). Thus, CH may be
regarded as deserving of special attention in the management of individuals suffering from
psychosis.
The research published in the last decade dealing with these issues can be grouped broadly into three
partially overlapping classes. (1) Studies testing for a relationship between CH and compliance (with
commands). (2) Studies which consider factors which may be associated with acting on CH (mediating
variables/processes). (3) Studies which look for a relationship between CH and dangerous/criminal
behavior (including violent and/or suicidal behaviors).
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 515

2. CH and compliance

Table 1 summarizes the main results of studies considering CH and compliance.

2.1. Compliance ratings

Compliance is the term used to consider how far an individual carries out the command given by the
voice. Studies indicate that some of those who experience CH are likely to comply with the commands.
All studies reviewed conclude that CH do influence the behavior of those who experience them.
Chadwick and Birchwood (1994) report a compliance rate of 88.5%. In fact, much of the research
suggests that large numbers of individuals comply with their CH. Three such studies (Chadwick &
Birchwood, 1994; Kasper, Rogers, & Adams, 1996; Rogers et al., 1990) assert that 80% or more of their
sample group report compliance with CH. Junginger (1990, 1995) reports slightly more conservative
figures of compliance, 39.2% and 40%, respectively.
The variance in the rate of compliance across studies may be explained as being due to a number of
factors. First, compliance is judged differently in each study. For example, Junginger (1995) rated
compliance on a three-point scale (noncompliance, partial compliance, and full compliance). Kasper et
al. (1996) also include full and partial compliance and Chadwick and Birchwood (1994) differentiate
between occasional and partial compliance. Junginger (1990) does not include partial compliance and
Rogers et al. (1990) do not expand their definition of compliance other than to say that 56% had at least
one experience when they responded to a command with unquestioning obedience. However, all studies
appear to report relatively high rates of compliance. Junginger noted that compliance with CH is
mediated (reduced) by the hospital environment. All the above studies, which report compliance rates,
are predominantly of inpatients. Thus, actual rates of compliance (in community samples) may be even

Table 1
Studies considering compliance with CH from 1990 to 2000
Study Primary research focus Sample characteristics Main data Key findings—compliance
source with CH
Junginger Mediation between N = 51 psychotic inpatients Interview 39.2% compliance with
(1990) command and compliance and outpatients commands
Rogers et al. Outcome following N = 25 forensic inpatients Interview 80% compliance rate
(1990) command
Martell and Compliance in mentally N = 20 referred for Official records 15% of offences reported
Dietz (1992) disordered offenders psychiatric evaluation as due to command
Chadwick and Mediation between N = 26 psychiatric Interview 88.5% compliance with
Birchwood command and compliance inpatients and outpatients commands
(1994)
Junginger Mediating variables N = 93 psychiatric Interview 40% full compliance with
(1995) inpatients commands
Zisook et al. Outcome following N = 46 psychiatric Interview and Did not report
(1995) command outpatients file review
Kasper et al. Violent outcome N = 25 psychotic inpatients Interview 84% recently complied
(1996) with commands
Hersh and CH and risk Review paper N/A N/A
Borum (1998)
516 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

higher than cited in some of these studies. Another concern is the number of individuals who do not
report experiencing CH. Nearly 50% of Rogers et al.’s sample and 27% of the sample considered by
Zisook et al. (1995) had not reported experiencing CH and had thus been overlooked. However, on
further examination, these individuals were found to be experiencing CH. Those individuals who do not
report CH (when they are experiencing them) are given little attention in the literature. Consequently, we
know little about the rates of compliance with commands for this group of people.

2.2. Confounding variables and validity

Other confounding variables still exist in the published research, even with the use of control
groups. For example, gender, age, and beliefs about voices (including familiarity and severity of
command) may all be important mediating variables. Such variables are discussed later in the
review. Thus, when considering compliance rates, if not controlled for, these factors may act as
confounding variables. The validity of the information collected in these studies may also be
questioned. While all reviewed studies, with one exception, utilize standardized measures to collect
the information, they are based on self-report. It is possible, therefore, that information collected
may not be concurrently valid. This is, however, a difficulty inherent in conducting such research.
Zisook et al. (1995) report use of file data to support interview information (which could corroborate
self-reports) but do not report compliance ratings. Another prevailing difficulty with the literature to
date is that all studies are retrospective and thus open to recall bias. Thus, it is possible that those
CH complied with may be those that an individual is most likely to remember, thus pushing
compliance rates up to an artificial figure. There is not one prospective study cited in the literature
searched for in this review.

2.3. Difficulties with sample populations

Difficulties inherent in collecting information from this group of individuals are further apparent when
sample size is considered. All samples are small, with only two with an experimental group larger than
50 (see Junginger, 1990, 1995). This may raise questions about the power of the studies’ ability to detect
an effect, should one exist. Another consideration affecting this concept is the clinical heterogeneity of
the populations from which the data is collected. If the sample population is not representative of the
command hallucinating population as a whole, results may be distorted or provide little useful
information about the wider population (i.e., not generalizable and a threat to external validity). A final
point to note when considering this literature is that all reviewed studies, apart from one (Chadwick &
Birchwood, 1994), were conducted in the USA. In the USA, the mental health services are provided
according to income brackets and determined by the health insurance held by the individual or family.
The likelihood of acquiring a biased sample is therefore high and makes data comparison across
countries and studies difficult. This review also serves to highlight the dearth of current literature in this
field in the UK.

2.4. Comment

While a variety of methodologies and small samples preclude clear conclusion, the common theme
from the literature suggests that some individuals who hear commands will act on them (e.g., Kasper et
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 517

al., 1996; Rogers et al., 1990). Most patients, who experience CH, have harmless commands and
frequently obey them (Junginger, 1995; Kasper et al., 1996; Thompson et al., 1992).
While some of the flaws in the literature have been highlighted, methodological errors alone are
unlikely to account for all the variance in the findings. Researchers have sought to control confounding
variables where possible and some standardized measures have been used. Thus, some of the variance
probably represents true individual differences between people and their situations and the relationship
between voice hearer and voice is likely to be more complex than first assumed. Researchers have begun
to consider why this might be so.

3. Factors associated with acting on CH (mediating variables)

As outlined in the previous section, rate of compliance with CH appears to be variable. One
conjecture to explain this may be that the situation is an extremely complicated process and mediated by
various psychological processes. These psychological processes have been explained by cognitive
theories about the relationship of the voice hearer to their voice (reviewed below). They have been
developed to try to further understand some of the links between delusional beliefs and behavior. Such
processes, or mediating variables, will be discussed in this section. Investigators have begun to explore
the phenomenological nature of the CH. Table 2 shows a summary of the main papers considering
possible mediating variables.

Table 2
Studies considering factors associated with compliance with CH (mediating variables) from 1990 to 2000
Study Primary research focus Sample characteristics Main data Key findings—mediating
source variables
Junginger (1990) Mediation between N = 51 psychotic inpatients Interview Voice familiarity and
command and compliance and outpatients supportive delusion
Chadwick and Mediation between N = 26 psychiatric Interview Severity of command,
Birchwood command and compliance inpatients and outpatients malevolence/benevolence
(1994)
Junginger (1995) Mediating variables N = 93 psychiatric Interview Voice familiarity and
inpatients dangerousness of commands
Beck-Sander A cognitive approach to N = 35 community sample Interview Beliefs about voice—
et al. (1997) acting on CH malevolence/benevolence and
power. Supportive delusions
Birchwood and Beliefs about voices— N = 62 community sample Interview and Cognitive model supported—
Chadwick a cognitive model questionnaire beliefs about voices
(1997)
Close and Garety Content of CH and N = 30 community sample Interview and Benevolent/malevolent voice
(1998) belief about voices. questionnaire
Cognitive model
Birchwood et al. Power and omnipotence N = 59 community sample Interview and Relationship to voice serves
(2000) of voices questionnaire formations of social rank.
Voice paradigm of social
relationship in general.
Power of voice
518 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

Support for a number of variables thought to influence compliance with CH is derived from this body
of literature. Beliefs about voices, content of command, the hospital environment, and presence of a
delusional belief, concurrent with a command, are all considered mediating variables.

3.1. Beliefs about the voice

3.1.1. Malevolence/benevolence
British researchers have carried out most of the work in this field. A cognitive approach to
understanding CH, proposed by Chadwick and Birchwood (1994), suggests that the beliefs that an
individual holds about his/her voice will influence compliance with commands given by it. They found
that voices perceived as malevolent were resisted while voices perceived as benevolent were courted.
This theme is further supported by three later studies (Beck-Sander, Birchwood, & Chadwick, 1997;
Birchwood & Chadwick, 1997; Close & Garety, 1998). Birchwood and Chadwick’s study showed that
89.3% of the sample resisted malevolent voices. While all four studies lend support to perceived
malevolence/benevolence of the voice-influencing compliance, it is worth noting that these samples may
be clinically and demographically homogeneous as all data is collected from in and around the same city.
Thus, results may only be reflective of a sample and of limited generalizability.

3.1.2. Power
The perceived power of the voice is also considered to be an important factor influencing compliance
(Birchwood, Meaden, Trower, Gilbert, & Plaistow, 2000). The greater the perceived power of the
commanding voice, the more likely an individual will comply with the command (Beck-Sander et al.,
1997). The majority (85%) of the voice hearers in the Birchwood and Chadwick (1997) study perceived
their voices as powerful and omnipotent and perceived themselves as weak and unable to exert control
over their voices. Thus, the behavior arising from the activity of the voice may be understood by
reference to the individual’s relationship with the voice.

3.1.3. Voice recognition


Whether an individual recognizes his/her voice is considered also to have some influence.
Junginger (1990, 1995) reports that familiarity of the voice may influence compliance. It is likely
that individuals who experience CH are more likely to trust those voices they recognize or can assign
an identity to (Rudnick, 1999). Identities that inspire trust consequently are more likely to enhance
compliance, especially if they are considered benevolent. Close and Garety (1998) report that 70% of
their sample were able to identify their hallucinated voice. However, none of the studies comment on
recognition of voices labeled as ‘God’ or ‘Satan’ for which recognition is only through assumption of
what they are believed to sound like. Such entities are bound up in culture, religion, education, and
experience.

3.2. Content of instruction

In addition to beliefs about the voice, the content of the instruction given is reported to influence
compliance. Rudnick (1999), in his review article, argues that compliance with commands cannot be
meaningfully considered without reference to the type of behavior commanded. Content of the
command is assessed according to how severe the behavior instructed is thought (by the researcher) to
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 519

be. Chadwick and Birchwood (1994) report that the content of command was the ‘‘single most
important determinant of compliance’’ (p. 200). None of their 26 participants complied with
commands judged to be dangerous or life threatening (to themselves and/or others) but compliance
with mild commands was reported as commonplace. Percentages of commands classified as mild or
severe and compliance with them were not reported in this study, with the results based on a cognitive
analysis of the voice experience. Junginger (1995) reports rates of compliance with commands of
differing severity with 45.8% compliance found with somewhat dangerous or dangerous commands.
These ratings consisted of a three-point scale: not at all dangerous, somewhat dangerous, and very
dangerous. Although the author reports excellent interrater reliability for the ratings, a definition of
dangerousness is not offered. In an earlier study, Junginger (1990) reports a similar figure (40%) of
compliance with dangerous commands. He thus asserts that the danger of behaviors specified by CH
is not a factor in compliance. He appears however, in this review, to be alone in this assertion. Beck-
Sander et al. (1997) found that benevolent voices were associated with compliance with innocuous
and severe commands. The authors go on to suggest that beliefs held about the instruction
commanded, the power of the voice, and the social acceptability of the action commanded will
influence compliance.
Studies appear to be reporting mixed conclusions about the influence of the content of command in
compliance. However, this may in part be explained by the differences between studies in rating the
command content. Junginger (1995) and Beck-Sander et al. (1997) rate command severity on a three-
point scale, while the other two studies rate command severity on a two-point scale (mild/severe and
dangerous/harmless). Partial and full compliance were not specified in all of the studies. Thus,
compliance with severe commands may be overestimated. Sample sizes range from 26 (Chadwick &
Birchwood, 1994) to 93 (Junginger, 1995), suggesting some degree of generalizability if considered
together. However, Junginger (1990, 1995) asserts that the hospital environment mediates compliance
with commands. Of the four studies, two use community samples, one uses inpatient samples, and one
uses both. Thus, they can only be generalized to their sample groups. Having said that, each study uses
similar methodology (semistructured interview) and goes someway to supporting the influence of the
content of the command on compliance.

3.3. Presence of a congruent delusion

As well as beliefs about the voice, severity of command, and the environment in which the voice
was heard, presence of a delusion congruent with the CH is also considered to influence compliance
(Junginger, 1990). This may further support the individual’s hallucination. If the voice is familiar and
there is consistency between these two experiences, chances of compliance would thus increase, as
the individual is more likely to interpret them as congruent with his/her understanding of the world.
Such delusions may be secondary to CH (as a consequence of such) or independent of CH, but
theoretically related. Research to date does not appear to separate individuals who experience CH
with a congruent delusion and those who experience delusions that are not congruent with the CH.
Future investigations that would help clarify whether individuals who hear CH and experience
incongruent delusions would be less likely to act on commands given as the perceptual experience is
thus less consistent. It is also possible that presence of a consistent delusion may indicate more severe
psychotic disturbance than just CH alone. Thus, these individuals may be more likely to comply with
commands.
520 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

3.4. Emotion

Another component identified as linking CH and behavior is emotion. Taking a cognitive view,
cognitions and behavior evolve together with affect. Rogers et al. (1990) noted that individuals with CH
experienced a more self-punishing and aggressive content than those with non-CH. This could be due to
the perceived lack of power/control experienced over the presentation of the voice and the effect it exerts
on the self. The voice experience itself is thought to be enough to provoke negative affect (Close &
Garety, 1998). Close and Garety (1998) also postulate that the voice may activate core beliefs about the
self which give rise to the affective response. The affective response to voices is also recognized by
Chadwick and Birchwood (1994) who noted that malevolent voices provoked negative reactions and
benevolent voices provoked positive reactions in voice hearers.

3.5. Other cognitive theories

In exploring this link between CH and behavior, literature suggests it is helpful to consider four
cognitive theories illustrating reasoning processes leading individuals to action. These theories highlight
importance of beliefs, values, and norms in mediating action. They are not exclusive of each other and
may all impact on the reasoning process leading to action.

3.5.1. Threat control override


One such theory is that of threat control override (TCO), suggested by Link and Steuve (1994). They
argue that psychotic symptoms are likely to lead to violence/antisocial behavior if they cause a person to
perceive others as out to harm them or intrude in such a way as to override proscriptions against
violence (Link, Steuve, & Phelan, 1998). The authors postulate that despite the irrational content of the
belief, if it is accepted that the individual experiences them as real or holds the belief that they are real,
then it would be possible to understand why a person who holds them would act in an antisocial manner.
This assertion can be considered in terms of the CH. If the seemingly irrational belief that a voice may
be able to do the hearer harm (or those the hearer wishes to protect) is accepted as experienced or
believed to be real, then it is possible to understand why a voice hearer may act on commands given.
Thus, if a voice is believed to be powerful and threatening, an individual may carry out the commands
given to remove the fear of the threat (even if that means acting in an unacceptable fashion). This further
lends support to the theory that beliefs the individual holds about the voice are an important factor when
considering compliance.

3.5.2. Effect of transgression on compliance


The second theory to consider is the effect of transgression on compliance. Transgression refers to
the breach of the command. The effect of transgressing (or resisting) the command may arouse
feelings of guilt (Carlsmith & Gross, 1969), leading an individual to engage in behaviors to improve
his/her self-esteem. This theory suggests that transgression exerts an effect on compliance whereby
feelings of guilt and shame are aroused should the individual not comply, leading to engagement in
behaviors that raise self esteem (Brewin, 1988). Such actions may also serve to protect against
lowering of self-esteem. This behavior may be appeasement behavior, although the resulting behavior
is not directly related to the command. For example, a command to harm a friend may be
transgressed. The voice hearer may believe that harming himself/herself will appease the voice.
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 521

Consequently, a behavior is performed which is not related in content to the command given. In a
more recent paper, Beck-Sander et al. (1997) support this view that some individuals will try to
appease their voices, having transgressed other commands. This suggests that self-esteem may also be
a mediating variable. Behaviors may be performed therefore, which, although not directly related in
content to the CH, are believed by the voice hearer to appease his/her voice (Beck-Sander et al.,
1997).

3.5.3. Effects of obedience


The effects of obedience may also be useful when considering action. The effects of obedience to
authority were demonstrated by Milgram (1974). He showed that beliefs held by the participant about
the authority of the experimenter predicted consequences of his/her behavior and beliefs about his/her
own degree of control mediated his/her compliance with commands given by the experimenter. Thus,
some participants could be persuaded to administer apparently painful electric shocks to others.
Applying this to CH, Chadwick and Birchwood (1994) report that 100% of their sample believed their
voices to be omnipotent. In a later study, Birchwood et al. (2000) suggest that much appraisal of the
voice appears to be made in relation to the self. They found this relationship to be characterized by the
subordination of the voice hearer to his/her voice. They go on to suggest that individuals with less
perceived power than their voice may cope by appeasement/submission or compliance. Thus, stimuli
perceived as powerful as well as threatening (a malevolent voice) might prompt the individual to act
upon it.
These three cognitive theories suggest that individuals who experience CH make a decision to act or
not act upon them in a manner similar to the processing of those without CH. That is, that they base a
decision to act/not act on a number of logical stages considering their beliefs about the possible
consequences of action and nonaction. These theories help explain why the variables discussed may be
important in mediating whether the action given in the command is carried out.

3.5.4. Social rank theory


While the theories discussed above may help in understanding processes that lead an individual to
action following a command, they do not explain the phenomenological difference between command
and intrusion and the coping responses of compliance, resistance, or appeasement. A theory which could
account for these phenomena is ‘social rank theory’ (Gilbert, 1992). The theory focuses on the role of
dominance hierarchies as coordinators of social behavior and affect and attempts to integrate social and
cognitive models (Champion & Power, 1995). Birchwood et al. (2000) applied this theory to understand
the nature of the relationship between the individual and persecutor—in this case, the voice. The
relationship between voice hearer and voice is characterized as one of involuntary subordination to a
powerful other (Birchwood et al., 2000). In social rank theory (Gilbert & Allen, 1998), stimuli perceived
as powerful or threatening (such as a malevolent voice) activate defensive responses, including
submissive and escape behavior. Resistance to a malevolent voice is an example of such a response.
Stimuli perceived as safe (such as a benevolent voice) illicit cooperation. The appraisal of another (or in
this case, the voice) comes from a process of social comparison serving the function of social ranks. In
their recent study, Birchwood et al. confirmed the subordination of voice hearer to voice. Thus, this
theory helps to further explain the relationship between command and action. It is not at odds with the
other theories discussed but rather fits with them, incorporating further the importance of the beliefs
about the voice.
522 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

3.6. Comment

There appear to be a number of mediating variables that influence the action arising from a
command. A cognitive approach suggests that beliefs about the voice are an important variable,
especially in relation to perceived power and malevolence. The theories discussed above all support
this idea, most notably perhaps the idea of social ranks applied to voice by voice hearer. The affect
response prompted by the voice also has support as a mediating variable, but it is probable that affect
is directly related to the beliefs held about the voice. The content of the instruction or the severity of
command is thought to effect compliance. The less severe the command, the more likely an individual
is to act on it. Finally, presence of a delusion congruent with the beliefs about the voice may increase
the likelihood of the voice hearer carrying out the command. Consistency in the voice hearer’s world
may prompt trust and thus action. Given the influence of these mediating variables, research should
further consider CH within a framework of beliefs about the commanding voice. CH and resulting
behaviors must be considered in accordance with the voice hearer’s appraisal of his/her own
experience.

4. CH and dangerousness (including suicidal and/or violent behavior)

Much of the literature considering mediating variables has included forensic populations. The
implication of hallucinated commands for subsequent action is particularly critical in the context of
criminal or antisocial behaviors. The main papers considering such behavior and CH are summarized in
Table 3.
Compliance with CH may entail dangerous, violent, or illegal behavior, the prediction and
modification of which is of clinical, social, and political interest. Mental health clinicians are often
required to assess dangerousness of individuals experiencing psychotic symptoms. In conducting such
assessments, the clinician must consider the relevant risk factors of psychotic symptoms, including the
CH. The clinical approach is based on an individual functional analysis and usually informed by
actuarial findings. However, findings in the literature focus on generalizations and do not offer
explanations for variation between individuals. Recently, there has been a move to assess the risk
posed by an individual in a specific context with specific symptoms. Thus, the nature of the variables
must be considered in relationship to risk.
The literature considering CH and dangerousness has not been researched in the same way as the
other two areas discussed. That is, CH and dangerousness have been explored in terms of risk of
acting on a CH and resulting antisocial or dangerous behavior, rather than in terms of a psychological
theory relating to causation and mediation of CH. The focus has been in considering a relationship
between CH and dangerous behavior. As discussed in the previous section, risk of acting on a
command is mediated by a variety of variables, including beliefs about the voice, content of
instruction (severity), presence of a congruent delusion, and emotion. Cognitive theories are also
highlighted as a way of understanding mediating variables. Such variables influence whether an
individual will act on a command given. It is possible to consider that dangerous compliance is a more
extreme version of these compliance phenomena discussed earlier. Therefore, the same psychological
processes could be expected to be at work and applied here. Such variables and phenomena will not
be discussed again here but their relevance should be borne in mind when considering the antisocial
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 523

Table 3
Studies considering CH and criminal or antisocial behaviors from 1990 to 2000
Study Primary research focus Sample characteristics Main data Key findings in relation to
source dangerous behavior
Junginger (1990) Mediation between N = 51 psychotic inpatients Interview Voice familiarity and
command and compliance and outpatients supportive delusion. 40%
compliance with dangerous
commands
Rogers et al. Violent outcome N = 25 forensic inpatients Interview and Supportive of some
(1990) questionnaire relationship between CH and
violent behavior
Jones et al. (1992) CH and sexual assault N = 4 forensic inpatients Interview 100% complied with
commands to sexually assault
Martell and Dietz Outcome N = 20 offenders Official 15% offending due to CH
(1992) records
Thompson et al. Violent outcome N = 34 offenders found not Official 62% of CH related to crime
(1992) guilty by reason of insanity reports committed. CH
overrepresented in forensic
populations
Chadwick and Mediation between N = 26 psychiatric Interview 88.5% compliance with
Birchwood command and compliance inpatients and outpatients commands, but none with
(1994) dangerous commands
Junginger (1995) Mediating variables N = 93 psychiatric Interview 46% compliance with
between CH and inpatients dangerous commands.
dangerousness Patients with CH at risk of
dangerous behavior
Zisook et al. Outcome N = 46 psychiatric Interview and 44% violent commands.
(1995) outpatients file review Suicidal commands should be
taken seriously
Kasper et al. Violent outcome N = 25 psychotic inpatients Interview CH with violent content signal
(1996) risk of aggressive behavior
and self-harm
Cheung et al. CH and violent behavior N = 31 inpatients Interview and No relationship between CH
(1997) questionnaire and violent behavior
Beck-Sander and Risk assessment N/A Discussion
Clark (1998) paper
Hersh and Borum CH and risk N/A Review paper
(1998)
Rudnick (1999) CH and dangerousness N/A Review paper

behaviors that may result. Research needs to further consider different variables and their specific or
direct relationships to risk.

4.1. Compulsive power of CH

Individuals suffering from mental illness who experience CH are often considered particularly
dangerous to themselves and others via the belief that they cannot, or do not, resist their voice’s
commands. Thompson et al. (1992) report that this popular conception stems from the work of Bleuler
(1924), who suggested that CH had a compulsive power, making them difficult to ignore. Thus, CH may
524 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

carry a heightened risk of dangerousness to the self or others. It is not so clear, however, what evidence
supports such common clinical knowledge. A literature has emerged to investigate this assumption, but
its coherency is poor, as researchers have tried to answer different questions.

4.2. Lack of coherency in the literature

Some consider symptoms of psychosis and violent behavior (e.g., Swanson, Borum, Swartz, &
Monahan, 1996); others consider specific types of violent behavior, for example, sexual violence (e.g.,
Jones, Huckle, & Tanaghow, 1992), suicide or self-harm (e.g., Kasper et al., 1996; Zisook et al., 1995),
or violence to others (Cheung, Schweitzer, Crowley, & Tuckwell, 1997; Kasper et al., 1996; Swanson et
al., 1996). Some do not attempt to differentiate between types of behavior. Still, others consider CH in
terms of approaches to risk assessment (Beck-Sander & Clark, 1998) or in terms of dangerousness of the
command (Chadwick & Birchwood, 1994; Junginger, 1995) and resulting behavior.
As well as these questions being only loosely (but critically) linked, other methodological problems
arise when trying to consider the literature as a coherent body. As discussed in previous sections, little is
known about the actual frequency with which CH are heard, obeyed, or result in antisocial behavior. Our
current understanding of CH is limited by the small number of studies in diverse settings and by the
absence of standardized measures for evaluating symptoms and classifying severity of command. While
it may be speculated that studies in forensic institutions may be somewhat biased due to possible
malingering of participants, there also is evidence that CH are very much underreported. Rogers et al.
(1990) note that 48% of their sample in a forensic assessment unit were overlooked as having CH, when
in reality, they were experiencing such phenomena. Similarly, Zisook et al. (1995) note that 27% of their
psychiatric outpatients did not report experiencing CH to staff. Another confusion is with the definition
of violent behavior. Studies have utilized a variety of definitions making comparison all the more
difficult. Some studies rely on criminal or official records (Jones et al., 1992; Thompson et al., 1992)
which are unlikely to be meticulously recorded for research purposes. It also invites bias due to different
individuals’ recording behaviors over unspecified periods of time. It also must be borne in mind that
studies utilize only participants who have been referred to services. Thus, such research may have
missed those members of the community who are not currently receiving service support, but who do
experience CH. Those individuals who were referred may be more severely distressed or unwell than
other members of the community. While these and other difficulties within the literature remain, there is
bound to be a variety of conclusions reported.

4.3. Symptoms of psychosis and dangerousness

Studies that consider symptoms of psychosis report that there is a relationship between such
symptoms and dangerousness or antisocial behavior (Beck-Sander & Clark, 1998; Swanson et al.,
1996). CH, therefore, as a symptom of psychosis, may be considered especially important. Some
researchers report a link between CH and violent or antisocial behavior. Rogers et al. (1990) report that
presence of CH in a forensic population raises a high level of concern (and possible risk) with respect to
violent or antisocial behavior (N = 25). They go on to conclude that obedience to CH may have a
substantial influence on antisocial behavior. ‘‘Perhaps most disturbing’’ they state, ‘‘is the capacity of
these hallucinatory commands to exact unquestioning obedience, often on a frequent basis’’ (Rogers et
al., 1990, p. 1306). With this argument, they support some earlier work that there is obvious importance
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 525

of CH to forensic evaluations (Rogers, Nussbaum, & Gillis, 1988). Junginger (1995) also supports this
assertion, reporting that psychiatric patients who experience CH are at risk for dangerous behavior. In an
earlier study (Junginger, 1990), 16% of the sample (N = 51) reported compliance with dangerous
commands.
In their small sample (N = 25), Kasper et al. (1996) found 92% reported compliance with CH ordering
violence to themselves, while 67% reported compliance with CH ordering violence toward others. Fifty-
six percent of the population studied reported having engaged in violent or suicidal behavior in the last
12 months. From their results, the authors conclude that CH with violent content signal a substantial risk
of aggressive behavior. However, they go on to say that risk of violence is no greater for psychotic
patients with CH than without them. This seems somewhat contradictory to the figures reported above
and suggests that other symptoms of psychosis may be influencing dangerousness. While considering
this point, it is notable that the literature does not differentiate between individuals who have been
violent to themselves or others before and after onset of psychosis. Perhaps some of those who act in an
antisocial manner after onset of psychosis may have the predisposition to act in such a manner prior to
experiencing any psychosis. Jones et al. (1992) report another study of four cases, supportive of the
influence of CH on dangerous behavior. While not perhaps methodologically sound (no description of
methods, data collection, or time frame), this paper offers an interesting insight into sexual assaults in
response to CH. All four cases attempted to sexually assault women as a direct response to CH. They
conclude that some patients do as their voice commands.

4.4. No direct relationship between CH and dangerousness

As discussed above, there appears to be support for the clinical lore that CH influence antisocial or
dangerous behavior. However, there is also support for the nonexistence of a relationship between CH
and dangerous behavior. Zisook et al. (1995), in their study of 46 stable outpatients, found no direct
relationship between CH and violence. It is possible that such stability may be indicative of less distress
and a greater ability to cope appropriately with CH. Having found no direct relationship, the authors note
that the two patients who committed suicide during the study had been experiencing suicidal CH. They
also report that 27% of their sample did not report experiencing CH to staff. Cheung et al. (1997) also
report no relationship between CH and violent behavior (N = 31). It is worth noting that in this study,
staff observations of violent incidents were used to split the two groups of patients diagnosed with
schizophrenia (rather than CH/non-CH). The only difference found between the two groups was that the
violent group was more likely to experience negative emotions toward their voices. (This supports
further the work of Chadwick and Birchwood (1994) and Close and Garety (1998) and in considering
the importance of affect as a mediating variable).
Thompson et al. (1992), in their study of offenders (found not guilty by reason of insanity), also found
no direct relationship between CH and dangerous behavior. These researchers report that command
hallucinators were significantly less likely to commit violent crimes than other psychiatric forensic
patients. However, they also report that 62% of the sample experienced at least some commands related
to the crime committed (N = 34). They go on to say that comparison between groups whose CH were or
were not related to their offense found no significant differences on any variable (demographic, clinical,
and offense characteristics). They do note, however, that CH appear to be overrepresented in the forensic
population. Such overrepresentation of CH in forensic populations is also suggested by Rogers et al.
(1990). From this, we may speculate that a relationship exists between CH and dangerous behavior.
526 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

However, it is also possible that persons who do act on their commands are more likely to commit
dangerous acts resulting in forensic referrals. We must also consider the patient’s possible motivation to
minimize consequences of his/her antisocial actions and avoid responsibility (Thompson et al., 1992).
Thus, the possibility of malingering is real with individuals attributing their behavior to their CH.
A paper by Rudnick (1999) reviewing papers from 1967 to 1997 concluded that most agreed on the
nonexistence of an immediate relationship between CH and dangerous behavior. The more recent papers
reviewed here do not appear to support this assertion. Results reported in this paper show a picture of CH
and dangerous behavior lacking in clarity. However, it is suggestive of a relationship between CH and
behavior perhaps mediated by variables discussed in the previous section.

4.5. Comment

The methods used are disparate and not rigorous enough to offer definitive answers. Defining types of
dangerousness would help clarify matters considering different types of assault and whether it is directed
at the self or others. Other variables, as discussed earlier, appear to influence the behavior resulting from
CH. These are also apparent when considering dangerous behavior specifically: most notably perhaps,
familiarity of the voice (Junginger, 1990, 1995), beliefs about the voice that the individuals holds
(Birchwood & Chadwick, 1997; Chadwick & Birchwood, 1994), and severity of commands given by the
voice (Beck-Sander et al., 1997; Chadwick & Birchwood, 1994; Junginger, 1995). All studies reported
suffer from less than optimum sample sizes. The stability or stage of illness that participants are in is
rarely reported (e.g., Zisook et al., 1995), which also may influence the results reported. There also is the
added confusion of legal classifications, which are often different in North America as compared to the
UK.

5. Conclusions

This body of literature is still in its infancy, and as discussed throughout the review, marred by
methodological difficulties. Researchers have asked differing questions about different samples from
different populations while hoping to answer similar questions. Sample sizes are small and often
clinically heterogeneous. It is recognized, however, that engaging suitable participants in such studies is
extremely difficult. Data are collected retrospectively and possibly subject to recall bias. Although some
standardized measures have been used, there is no one measure currently available to comprehensively
assess CH and resulting behavior. Stability of patients is often not reported, and there are often
differences in the definitions of ‘‘dangerousness,’’ ‘‘compliance,’’ and ‘‘severity.’’ Perhaps discrepancies
would be lessened if definitions were clearer and both the researcher and participant definitions were
considered. Little or no account of participant’s appraisals of the voice is provided and few studies offer
details of the process of recruitment. Data used in some studies have been recorded for clinical purposes
and thus may not be scientifically rigorous. It may also be open to bias due to differences in who
observed and or recorded the behavior. Thus, in considering the literature as a body of knowledge, it
must be remembered that the validity and reliability may be flawed.
While there are difficulties interpreting this body of literature, it does seem to suggest a number of
conclusions. Many people with auditory hallucinations experience commands. Many experience
harmless commands and frequently obey them while others do not choose to report them at all. The
L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528 527

relationship between the two is a complex one and mediated by a number of different variables. There is
most support for beliefs held about the voice (including familiarity), the content or severity of command,
and the presence of a consistent delusion. Cognitive theory provides some explanation of how this
phenomenon may occur. It is postulated that the issue of dangerousness is subject to the same mediating
variables as other commands. However, further research needs to be carried out to clarify this. The
support for a direct relationship between CH and dangerous behavior is variable, but CH are notably
overrepresented in clinical forensic populations. What does seem clear is that each individual’s
experience of his/her voice is subjective and must be considered in terms of his/her own subjective
reality. A command may be neither necessary nor sufficient for action. It is not necessary as action may
result from hallucinations without command or attempts to appease the voice. Having discussed the
many variables thought to mediate between action and command, it is clear that CH are not sufficient to
produce action in isolation.
There is undoubtedly more work to be done in this field for the purposes of meeting clinical need and
risk assessment. The body of knowledge is still someway from predicting and controlling the behavior of
those who hear commands. Future research may be best directed at considering specific aspects of the
command in specific populations, as well as carrying out longitudinal studies that could further clarify
the course and clinical correlates of CH. Methodology should leave the reader without question as to the
sample considered or his/her recruitment to the study as well as providing clear definitions about the
phenomenon under investigation.

References

Beck-Sander, A., Birchwood, M., & Chadwick, P. (1997). Acting on command hallucinations: A cognitive approach. British
Journal of Clinical Psychology, 36, 139 – 148.
Beck-Sander, A., & Clark, A. (1998). Psychological models of psychosis: Implications for risk assessment. Journal of Forensic
Psychiatry, 9(3), 659 – 671.
Birchwood, M., & Chadwick, P. (1997). The Omnipotence of voices: Testing the validity of a cognitive model. Psychological
Medicine, 27, 1345 – 1353.
Birchwood, M., Meaden, A., Trower, P., Gilbert, P., & Plaistow, J. (2000). The power and omnipotence of voices. Subordi-
nation and entrapment by voices and significant others. Psychological Medicine, 30(2), 337 – 344.
Bleuler, E. (1924). Textbook of psychiatry. New York: Macmillan.
Brewin, C. R. (1988). Cognitive foundations of clinical psychology. Hove: LEA.
Carlsmith, J. M., & Gross, A. E. (1969). Some effects of guilt on compliance. Journal of Personality and Social Psychology, 11,
232 – 239.
Chadwick, P., & Birchwood, M. (1994). The omnipotence of voices: A cognitive approach to auditory hallucinations. British
Journal of Psychiatry, 164, 190 – 201.
Champion, A., & Power, M. J. (1995). Social and cognitive approaches to depression: Towards a new synthesis. British Journal
of Clinical Psychology, 34, 485 – 503.
Cheung, P., Schweitzer, I., Crowley, K., & Tuckwell, V. (1997). Violence in schizophrenia: Role of hallucinations and delusions.
Schizophrenia Research, 26, 181 – 190.
Close, H., & Garety, P. (1998). Cognitive assessment of voices. British Journal of Clinical Psychology, 37, 173 – 188.
Gilbert, P. (1992). Depression. The evolution of powerlessness. Hove: Lawrence Erlbaum Associates.
Gilbert, P., & Allen, S. (1998). The role of defeat and entrapment (arrested flight) in depression: An exploration of an
evolutionary view. Psychological Medicine, 28, 585 – 598.
Hersh, K., & Borum, R. (1998). Command hallucinations, compliance and risk assessment. Journal of American Academy of
Psychiatry and the Law, 26(3), 353 – 359.
528 L.G. Braham et al. / Clinical Psychology Review 24 (2004) 513–528

Jones, G., Huckle, P., & Tanaghow, A. (1992). Command hallucinations, schizophrenia and sexual assaults. Irish Journal of
Psychological Medicine, 9, 47 – 49.
Junginger, J. (1990). Predicting compliance with command hallucinations. American Journal of Psychiatry, 147, 245 – 247.
Junginger, J. (1995). Command hallucinations and the prediction of dangerousness. Psychiatric Services, 46(9), 911 – 914.
Kasper, M. E., Rogers, R., & Adams, P. A. (1996). Dangerousness and command hallucinations: An investigation of
psychotic inpatients. Bulletin of the American Academy of Psychiatry and the Law, 24(2), 219 – 224.
Link, B., & Steuve, A. (1994). Psychotic symptoms and the violent/illegal behavior of mentally ill patients compared to
community controls. In J. Monahan, & H. J. Steadman (Eds.), Violence and mental disorder ( pp. 137 – 59). Chicago:
University of Chicago Press.
Link, B., Steuve, A., & Phelan, J. (1998). Psychotic symptoms and violent behaviors: Probing the components of ‘‘threat/
control-override symptoms. Social Psychiatry and Psychiatric Epidemiology, 33, S55 – S60.
Martell, D. A., & Dietz, P. E. (1992). Mentally disordered offenders who push or attempt to push victims onto subway tracks in
New York City. Archives of General Psychiatry, 49, 472 – 475.
Milgram, S. (1974). Obedience to authority. New York: Harper and Row.
Rogers, R., Gillis, J. R., Turner, R. E., & Frise-Smith, T. (1990). The clinical presentation of command hallucinations in a
forensic population. American Journal of Psychiatry, 147, 1304 – 1307.
Rogers, R., Nussbaum, D., & Gillis, R. (1988). Command hallucinations and criminality. Bulletin of the American Academy of
Psychiatry and the Law, 16, 251 – 258.
Rudnick, A. (1999). Relation between command hallucinations and dangerous behavior. Journal of American Academy of
Psychiatry and the Law, 27(2), 253 – 258.
Swanson, J. W., Borum, R., Swartz, M. S., & Monahan, J. (1996). Psychotic symptoms and disorders and the risk of violent
behavior in the community. Criminal Behavior and Mental Health, 6, 309 – 329.
Thompson, J. S., Stuart, G. L., & Holden, C. E. (1992). Command hallucinations and legal insanity. Forensic Report, 5, 29 – 43.
Zisook, S., Byrd, D., Kuck, J., & Jeste, D. V. (1995). Command hallucinations in outpatients with schizophrenia. Journal of
Clinical Psychiatry, 56, 462 – 465.

You might also like