3. Los Derechos de los niños, niñas y adolescentes 1 La Doctrina de Protección Integral
3.2. Los niños, niñas y adolescentes como sujetos de derechos
Studies that are based on current clinical populations are likely to be more informative in
terms of identifying a relationship between command hallucinations and risk behaviour.
Yet, the findings from early studies into command hallucinations were equivocal at best
and the majority of studies focus on violence.
Several studies have detected a null relationship between command hallucinations and
violence. Three of these have examined violent behaviour in general amongst individuals
with command hallucinations. Hellerstein et al (1987) reported that the assaultiveness of
a group of inpatients with command hallucinations was not significantly different from
individuals without command hallucinations. Zisook et al (1995) also noted that amongst
a group of outpatients with schizophrenia those with command hallucinations did not
have a history of more violent or impulsive acts than a matched group without command
hallucinations. Cheung et al (1997) classified a group of psychiatric inpatients as either
violent or not violent based on staff observations of aggressive incidents and found that
exactly the same number of participants in each group reported command hallucinations
leading them to conclude that there appeared to be no relationship between the presence
colleagues acknowledge that the incidence of violence amongst both groups was very
low, was infrequent and was most often directed towards staff. This null result may be a
reflection of a lack of power or just a reflection of poor classification of experimental
groups.
Whilst these studies may seem reassuring in that they suggest that individuals with
command hallucinations are unlikely to be any more violent or aggressive than
individuals without command hallucinations, they cannot feasibly be used as a reliable
indicator or whether someone will comply with a command to harm others. The major
limitation of all three studies is their failure to control for exposure to violent content
commands. For instance, Hellerstein et al do not draw upon their finding that, based on
the medical records, only 5% of their sample had reported commands to seriously harm
others when they interpreted the non-violent nature of their sample. Cheung et al also
failed to control for content. It may therefore be that the violence that was observed in
their study was totally unrelated to the content of the command (which could have, for
instance, been instructing suicide). Violence and aggression are not uncommon during
the early phases of hospital admission as a result of the early stage of the episode and the
requirement of the individual to adjust to the confines of hospital treatment (McNeil,
Binder and Greenfield, 1988). It is therefore essential that the content of the commands
is carefully considered.
Rogers et al (2002) demonstrated the importance of controlling for the content of the
command in retrospective study of compliance with command hallucinations in a
medium secure unit (n = 54). They attempted to address the much cited confound of not
medical records or legal documents. Individuals with a history of command
hallucinations prior to or at admission were included in the study and their behaviour
followed up throughout their admission by way of an examination of untoward incident
forms detailing any violent or self injurious behaviour. The authors found that there was
a significant relationship between self-harm commands and subsequent self-harming
behaviour (or attempted behaviour that required staff intervention) but that there was no
significant relationship between violent content commands and later violent behaviour.
When the content of the command was ignored, the statistical relationship between
commands and compliance were substantially weakened. For all of its methodological
strengths, the difficulty, of course, with this study is that the individuals themselves were
not interviewed so there is an assumption (which the authors acknowledge) that the
behaviour was driven by command hallucinations when in fact (particularly with the
aggressive treatment and therapy that the authors mention) such symptomatology may
have resolved and the self-harming, violent or aggressive behaviour that was observed
may have been associated with something else entirely, such as the constrains of residing
in a medium-secure unit. There is also the issue of only including a detained sample. It
is likely that in a medium secure unit any attempts to comply with a violent command
would have been thwarted or prevented in a risk management plan thus limiting the
detection of the dependent variable. Nevertheless the study did highlight the importance
of controlling for the content of the voices, a finding that Rogers (2004) has since
replicated in a secondary analysis of the large-scale MacArthur risk assessment database.
An important finding from the Rogers study was that when a three-month, post-discharge
commands had subsequently committed suicide, suggesting that when the safeguarding
provided by the secure facility was not present, individuals may well have complied with
a command to harm themselves. Rogers did not report how many of those reporting
violent content commands went on to commit violent acts or offences.
Other studies have also detailed the content of commands (even if they have not later
controlled for them in statistical analyses). Kaspar, Rogers and Adams (1996), conducted
a study of hospitalised patients and detected a high rate of compliance with command
hallucinations. According to case note data, of those reporting command hallucinations,
60% reported complying with harmless commands in the past month, 92% of these
complied with commands to harm self, and 67% with commands to harm others, yet the
authors found that the command hallucinations group were no more likely to engage in
violent or suicidal behaviour than a group of individuals with general auditory
hallucinations. Such high rates of compliance are not necessarily surprising amongst an
inpatient sample as risk behaviour is the most likely prompt for admission to an inpatient
facility. Studies that have incorporated a community sample typically report lower levels
of compliance, for instance Juninger (1990) interviewed 44 individuals from mixed
settings found that 40% admitted complying with dangerous commands and a similar
proportion reported complying with innocuous commands. Unfortunately, Juninger
relied on self-reported violence although does state that the reports were later verified.
It is clear that some of the earlier studies were marred somewhat by their failure to
examine the nature of the commands before correlating the symptom with the subsequent
behaviour. Further, it would appear from the existing research that there is limited and
violence or self-harm). Relying solely on self-report may over or underestimate the
actual occurrence of behaviour depending on the individual’s subjective view of what
constitutes violence or self-harm.
Those studies that have directly interviewed voice hearers, have utilised standardised
measures of violence (McNiel, Eisner, and Binder, 2000) and have also collected
collateral information from informants (Monahan, Steadman, Silver, Appelbaum,
Robbins, Mulvey, Roth, Grisso, and Banks, 2001) have detected a positive relationship
between command hallucinations and violence. McNiel et al (2000) adopted a
retrospective design and interviewed a sample of inpatients in a "civil, non-forensic
context” in the United States asking them whether they had ever heard a command
hallucination to harm others and, if so, how often they had complied with the command.
They found that individuals with command hallucinations were significantly more likely
to report a history of violence in the two months before hospital admission and they were
twice as likely to be violent than a heterogeneous group of inpatients without command
hallucinations. The association between command hallucinations and violence remained
significant even when established predictors of violence (male gender, substance abuse
and low levels of the need for social desirability) were entered as covariates. However,
when the extent of other psychotic symptoms was added into the regression equation,
command hallucinations no longer made a significant contribution. McNiel et al suggest
that although command hallucinations may be associated with violence it is only in the
context of other positive psychotic symptoms. Whilst this study did benefit from
improved methodological rigour, it was retrospective in nature and it did rely wholly on
records. Furthermore, the authors asked participants about the experience of command
hallucinations in the past year yet correlated this with violence in the two months
preceding admission. The two variables may have been totally unrelated given the
different time frames provided. Finally, whilst the initial sample of inpatients was large
(n = 103), less than one third reported command hallucinations in the last year (n = 31).
Of these, twenty individuals reported hearing such commands infrequently (‘sometimes’
to ‘almost never’) and less than half of those reported commands complied more than
‘sometimes’. The findings are therefore based on a very small sample. Given the small
number of individuals reporting hearing commands the generalisability of the findings
must be questioned, particularly when this group are compared to a clinical sample who,
for example, are involved in a randomised controlled trial for command hallucinations
the majority of whom report frequent, distressing and disruptive command hallucinations
(Trower et al, 2004).
The MacArthur risk assessment study, on the other hand was prospective in nature,
employed standardised (albeit purpose designed) measures of violence and obtained
collateral information about violent behaviour from multiple sources. The study
detected a significant association between command hallucinations and violence at 20-
week follow-up and one-year follow-up. As with Rogers et al (2002) this association
was rendered non-significant when the content of the commands was not controlled for.
Rogers (2004) urges caution when interpreting these results suggesting that the use of a
chi-square analysis does not make the best use of the available data (as 5 follow-up time
periods were available and only two were utilised) and does not permit a detailed analysis
as gender, substance abuse etc). Whilst the MacArthur study benefited from the largest
sample of command hallucinators to date (n = 239 of whom 44% reported violent content
commands), the group consisted of individuals with a lifetime history of commands and
no distinction was made between those that may not have heard such voices for two years
and those that had heard them the day before the assessment. Like the McNiel study this
limits the findings to saying that individuals with a lifetime history of command
hallucinations appear significantly more prone to violence rather than command
hallucinations per se cause or significantly contribute towards future violence. The
findings could therefore be attributable to diagnosis or treatment effects rather than the
phenomenology per se.
Rogers (2004) addressed some of the limitations of the original analysis of the
MacArthur Risk Assessment database by controlling for a number of important
confounders. He found that even when pertinent sociodemographic variables (age,
gender, marital status and ethnicity), drug and alcohol abuse, severity of
psychopathology, presence of delusions and psychopathy were controlled for the
relationship between command hallucinations and violence remains significant. Indeed
the results from this secondary analysis suggest that individuals with violent content
command hallucinations are 1.39 times more likely to engage in a violent act in the
following year than individuals without violent content command hallucinations. Again,
however, the results were limited to individuals with a lifetime history of violent content
commands rather than current command hallucinations.
It would appear from this review, that if the results of any future examination of the
must pay close attention to the content of commands, must recruit command hallucinators
from multiple settings (inpatient and outpatient, community and forensic), and must be
prospective in nature. Further, any assessment of the outcome variable must be verified.
The design for the current study of the factors underpinning compliance with command
hallucinations seeks to incorporate all of these elements.
Regarding decisions about risk, it is difficult to conclude from the above literature
whether an individual hearing commands will comply or not. Whilst the issue of
controlling for the nature and content of the voice has been identified and addressed in
some of the later studies, there remain a number of methodological issues that decrease
the generalisability of the findings. There is a good deal of inconsistency amongst the
studies in terms of what constitutes compliance and how compliance behaviour is
assessed. Whilst some have conceptualised compliance as a dichotomous variable
(Rogers et al, 2002), others have noted an additional category of partial compliance
(Juninger, 1996). Conceptualising those that do not obey the exact instructions of the
voice as non-compliers may underestimate or misguide the assessment of risk and fail to
capture the complexity of the individual’s behavioural response to the command.
Furthermore, it would appear that the preponderance of the existing research has focussed
on violence. It is clear from those studies that have been conducted that whilst a minority
of individuals obey commands to commit dangerous and violent acts to harm others, a
significant proportion obey commands to harm themselves (Beck-Sander et al, 1997; Fox
et al, 2004). If we are to begin to fully understand compliance behaviour, it is essential
that we understand voices in the full context. For instance, an individual may be
explains why two individuals in the Zisook et al (1995) study committed suicide during
the process of the research (see Beck-Sander et al, 1997). Categorising one such
individual as a non-complier is overly simplistic and almost overlooks the risk that they
pose to themselves and perhaps the risk they may pose to others in the future since they
are clearly not able to fully resist the voice. Moreover, it has been suggested that
commands to self-harm have a different nature and course from violent content
commands (Buccheri et al, 2007) and the factors that underpin compliance with self-
harm commands may differ from those that underpin compliance with commands to harm
others (Fox et al, 2004) such that the findings from one group of compliers cannot be
generalised to another group.
The above review of the available evidence at a phenomenological level demonstrates
that study at this level alone is not sufficient in identifying those factors that influence the
risk of complying with dangerous commands and supports the need for more detailed
study. Whilst it is clear that, when the content of the command is taken into account,
there is a significant association between hearing a dangerous command and complying,
it is also evident that a significant proportion of individuals do not comply with the
voice’s instructions. It would seem that hearing a command hallucination (the
phenomenon) does not automatically result in obedience on behalf of the voice hearer
and that although command hallucinations may be pre-determinants of potential risk,
other extra-phenomenological variables may play a mediating role (e.g. belief about the
voice). Specifically, some individuals hear command hallucinations instructing them to
hear such commands and readily resist or ignore them. It seems timely now to begin to
consider what distinguishes those that comply from those that resist.