Command Hallucinations, Compliance, and Risk Assessment

Keith Hersh, MA, and Randy Borum, PsyD

Command hallucinations are auditory hallucinations that instruct a patient to act in specific ways; these commands can range in seriousness from innocuous to life-threatening. This article summarizes two areas of research regarding command hallucinations: rates of compliance with command hallucinations; and factors associated with compliance. Researchers have reported rates of compliance ranging from 39.2 percent to 88.5 percent. Compliance has not been consistently related to dangerousness of commands. Instead, research suggests that individuals are more likely to comply with commands if they recognize the hallucinated voice and if their hallucinations are related to a delusion. Implications for risk assessment are discussed in light of the research.

Mental health professionals are often actions that command hallucinations or- called upon to assess the risk of violence der the patient to perform range from the posed by people with mental disorders.'-' insignificant, such as making facial gri- In making these assessments, the clini- maces, to those as serious as suicidal or cian must consider and evaluate the rele- homicidal acts. Clinical lore suggests that vant risk factors, including those related people are prone to obey their command to an individual's clinical condition. hallucinations and that "dangerous" com- Command hallucinations are among the mands increase the likelihood that an in- clinical factors that have received individual will engage in violent behavior. creased clinical and empirical attention in Empirical studies have produced more recent years. Command hallucinations are mixed results. auditory hallucinations that instruct the This article examines the empirical ev- patient to act in a certain manner. The idence in two critical areas: (I) the rate of compliance with command hallucinations; and (2) the factors associated with Mr. Hersh is affiliated with the Department of Psychol- increased compliance with command hal- ogy, University of North Carolina at Chapel Hill, NC, lucinations. It then discusses the factors and Mr. Borum with the Department of Psychiatry and Behavioral Sciences, Duke University Medical Center, that a clinician should consider when con- Durham, NC. Address correspondence to: Keith R. Hersh, MA, Davie Hall, CB 3270, University of North ducting a risk assessment involving com- Carolina, Chapel Hill, NC 27599-3270. mand hallucinations.

J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 353 Hersh and Borum

Rates of Compliance with equally low in a different, less structured Command Hallucinations environment. Studies of single subjects or small sam- Researchers report higher rates of com- ples have often indicated that those who pliance with command hallucinations experience command hallucinations are when they have examined compliance likely to comply with the command^.^-^ with all commands (i.e., both lethal and This topic has only recently been exam- nonlethal) and when they have inquired ined using samples large enough to per- about compliance with commands before mit more generalizable statements about the patient was admitted to the hospital. the rate of compliance. Table 1 summa- Indeed, these studies have demonstrated rizes the results of these studies. A cur- relatively high rates of compliance (see sory examination of these data reveals Table l), with estimates ranging from two. apparently discrepant, trends. Some 39.2 percent (~un~inger")to 88.5 percent studies7. have concluded that command (Chadwick and ~irchwood'I).* hallucinations rarely influence the behav- An important issue regarding compli- ior of those who experience them, includ- ance with command hallucinations is ing the dangerous behavior of forensic whether compliance significantly inpatients. These studies, however, do not creases a patient's degree of dangerous- report actual rates of compliance with ness. The current literature is divided on command hallucinations. Instead, they re- this question. Several studies indicate that port that commands were "generally ig- patients experiencing command halluci- nored'' (Goodwin et al., p. 78)' and note nations with dangerous content are likely that patients experiencing command hal- to comply with these orders and are there- lucinations with violent content did not fore more likely to be dangerous them- display higher rates of violent behavior. selves. Junginger, in two separate stud- ~un~in~er,~however, has noted that com- ies,", lo has reported compliance rates of pliance with hallucinations is mediated by 45.8% and 40.0%, respectively, with dangerous command hallucinations. Kasper the hospital environment. Psychiatric in- et n1. l3 reported that 91.7 percent of sub- patients-such as the patients which com- jects experiencing command hallucina- prise the samples for Goodwin et aL7 and tions ordering acts of violence against Hellerstein et a~.~-would be prevented themselves complied with those orders, from complying with most commands to and 66.7 percent of those experiencing injure themselves or others by the rela- command hallucinations ordering acts of tively high degree of observation, struc- ture, and security inherent in a hospital environment. The fact that suicidal or ho- * Study results reported by Taylor" include the highest micidal command hallucinations are not rate of compliance with command hallucinations (100%).We have not included the results from this study associated with a higher rate of suicidal or when discussing the range of con~plianceobserved by homicidal acts in the hospital does not researchers because only two of the participants reported command hallucinations, and conclusions based upon mean that compliance rates would remain such small sample sizes are unlikely to be robust.

354 J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 Command Hallucinations

Table 1 Compliance with Command Hallucinations (CH) and Dangerousness Associated with CH Reported in Empirical Studies Percentage of Compliance Information Related with to Dangerousness Sample Command in Patients Study Characteristics Hallucinations Experiencing CH Among clinical or offender population: Goodwin et a/.' 117 Psychiatric inpatients 0.0% Compliance in three and outpatients; 42 patients experiencing CH patients with CH with suicidal content Hafner and B~ker~~259 Forensic patients 18.5% Ordered by CH to commit crime 203 Forensic patients; 2 One patient complied with patients with CH order to carry knife

Among patients with command hallucinations: Chadwick and 26 Psychiatric inpatients O.OO/oComplied with Birchwood" and outpatients "severe" (i.e., life- threatening) commands Hellerstein et aL8 58 Psychiatric inpatients 51.7% Suicidal content; 5.2% homicidal content; 12.1% lethal injury to self or others contentC Jungingerg 93 Psychiatric inpatients 45.8% Compliance among patients with somewhat dangerous or very dangerous CH Junginger1° 51 Psychiatric inpatients 40.0% Compliance with and outpatients dangerous CH Kasper et aI.l3 25 Psychiatric inpatients; 2 91.7% Compliance with CH patients could not ordering violence toward articulate nature of CH self; 66.7% compliance with CH ordering violence toward others Rogers et 25 Forensic patients 4.5% Criminal content; 59.0% mixed criminal and noncriminal contentc Thompson et a1.14 34 Forensic patients 62.0% Of command (found NGRI) hallucinations related to crime committed Zisook et a/.15 46 Psychiatric outpatients 44% Violent contentc a Rate of compliance not reported. Represents at least occasional or partial compliance. Represents rate of content of commands, not rate of compliance. Represents 'Yull compliance" with command hallucinations; percentage increases to 55.goL when "partial" or "full compliance" is considered. Represents compliance with command hallucinations in recent past; 56.0% of the sample reported at least one instance of unquestioned obedience, and 44.0% reported frequent or very frequent obedience.

J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 Hersh and Borum violence toward others complied with and noncriminal content. They did not, those orders. however, report how many patients com- Not all studies have shown that complied with these commands. mand hallucinations are associated with increased dangerousness. In a retrospec- Factors Associated with tive case review of forensic patients Compliance found not guilty by reason of insanity As indicated above, the dangerousness (NGRI). Thompson et al. l4 reported that of the content of commands is not con- command hallucinations contributed to sistently related to compliance. Several the patient's acquittal offense in 62 per- features of command hallucinations ap- cent of subjects reporting command hal- pear to mediate compliance. Jungin- lucinations during the time of their of- ger9. lo reported that patients are more fense. These patients, however, were less likely to comply with familiar voices than likely to have been acquitted of violent with unfamiliar voices. Similarly, Shore offenses than NGRI patients who had et a1." noted three factors that increase been experiencing noncommand auditory the risk that people with schizophrenia hallucinations and NGRI patients who will engage in self-mutilation: (1) the had not been experiencing auditory hal- presence of command hallucinations, (2) lucinations at all. Chadwick and Birch- calm reactions to the voices, and (3) trust- wood" reported that although 88.5 per- ing the voices. Chadwick and col- cent of their sample complied with leagues 18, 19 have also reported that command hallucinations, none of their voices that were described as "benevo- sample complied with "severe" (i.e., life- lent" by the patient were associated with threatening or dangerous) commands. greater "engagement" (a reaction that in- Similarly, Zisook et a1. l5 reported that the cludes both compliance and seeking out rate of suicide attempts did not differ the voices). It is likely that individuals between patients reporting command hal- experiencing command hallucinations are lucinations with suicidal content and pa- more likely to trust voices they believe tients who did not report command hal- they can identify. This trust may enhance lucinations (although both of the patients compliance, because patients are more who successfully committed suicide in likely to assume that recognizable, trust- this study had been experiencing suicidal worthy voices are ordering behaviors that command hallucinations). The danger- are in the patients' best interests. ousness issue is further complicated when A patient's beliefs-especially his or studies have not reported rates of compli- her beliefs about the command hallucina- ance with command hallucinations but tions themselves-also influence compli- instead have reported only rates of spe- ance. ~unginger" reported that the pres- cific contents of the commands. For ex- ence of a delusional belief consistent with ample, Rogers et al. l6 reported that 59.0 the command hallucination increases the percent of their sample experienced com- likelihood of compliance. In fact, the mand hallucinations with mixed criminal identifiability of hallucinated voices

356 J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 Command Hallucinations

"may be one aspect of a more system- toms. Similarly, Beck-Sander et a1.I9 re- atized disorder of reality in which symp- ported that participants who believed that tom-consistent behavior is likely to fit" they were in control of themselves were (Junginger, p. 96).20 Other research- less likely to comply with command hal- ers19. 21 have also noted that a patient's lucinations than participants who be- interpretation of his or her command hal- lieved that the voices were omnipotent. lucinations are related to compliance. A These findings support and underscore logically consistent relationship between the importance of assessing delusions hallucination and delusion should in- when attempting to estimate the likeli- crease compliance, because such patients hood of compliance with command hal- would be more likely to interpret these lucinations, particularly those with vio- perceptual experiences as congruent with lent content. their understanding of the world, just as familiar voices are more likely to be con- Implications for Risk sistent with patients' experience. Assessment Whatever the underlying mechanism, it The extant literature clearly shows that appears that the presence of a delusional not everyone who experiences a com- belief consistent with the content of the mand hallucination will comply, nor is hallucination increases the likelihood of everyone who experiences such symp- compliance. Concerning the implication toms considered dangerous. However, of this finding for assessing a patient's command hallucinations are certainly risk of violent behavior, it is worth noting within the appropriate scope of inquiry that prior research has shown the pres- when conducting a risk assessment. ence of delusions related to perceived When performing a mental status exami- threat or an overriding of one's internal nation, a clinician will likely ask about controls is associated with increased risk psychotic symptom patterns. If there are for violent behavi~r.~*-~~These symp- indications of psychosis (e.g., hallucina- toms have been referred to in the litera- tions or delusions) or violent ideation, a ture as threat/control override or TCO specific inquiry about command halluci- symptoms. In a large national sample, nations may be indicated (eg, "Have you Swanson and col~eagues~~found that per- ever heard voices that talked to you about sons who reported TCO symptoms were hurting people?"). The necessity of in- about twice as likely to engage in assaul- quiring specifically about command hal- tive behavior as those with other psy- lucinations is highlighted by the results of chotic symptoms and six times more Rogers et ~1.~~These authors found that likely to engage in such behaviors than the clinical staff of an inpatient forensic those with no mental disorder. Those re- assessment unit overlooked the presence porting TCO symptoms in combination of command hallucinations in 47.8 per- with substance abuse were 8 to 10 times cent of their research sample. more likely to engage in violent behavior If command hallucinations are or have than persons with no disorder or symp- been present, the clinician may want to

J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 357 Hersh and Borum explore this phenomenon in more detail. tion concerning base rates of compliance First, it may be useful to note which sit- and factors influencing compliance can uations and which other identifiable pre- be integrated with relevant history and cipitants are associated with the patient's clinical information to help understand command hallucinations. Second, it may the salience of these commands in an be relevant to ask about the content, fre- individual case and to improve the valid- quency, and recency of these commands ity of risk assessments. and to inquire about the individual's past experience with compliance (e.g., "Has helshe ever complied with commands in References the past? What types of commands has he 1. Borum R: Improving the clinical practice of or she complied with in the past? Has violence risk assessment: technology, guide- lines and training. Am Psychol 51:945-56, helshe ever heard a command and not I996 complied? What strategies do they use to 2. Borutn R, Swartz M, Swanson J: Assessing and managing violence risk in clinical prac- resist compliance? When is it hardest to tice. J Practical Psychiatry Behav Health 2(4): resist?"). Third. in some circumstances, it 205-15, 1996 may be appropriate to consider the effects 3. Monahan J: The clinical prediction of violent behavior. Rockville, MD: NIMH. of context and environment on the indi- 4. Hall DC, Lawson BZ, Wilson LG: Command vidual's compliance. For example, people hallucinations and self-amputation of the pe- nis and hand during a first psychotic break. may be less likely to comply with "dan- J Clin Psychiatry 42:322-4, 1981 gerous'' commands while they are in a 5. Jones G, Huckle P, Tanaghow A: Command restrictive treatment setting, or the nature hallucination, schizophrenia, and sexual as- saults. Ir J Psychol Med 9:47-9, 1992 of the commands may change when peo- 6. Martell DA, Dietz PE: Mentally disordered ple make a transition into a new environ- offenders who push or attempt to push victims onto subway tracks in New York City. Arch ment. Fourth, it may be useful to know Gen Psychiatry 49:472-5, 1992 (1) whether the voice in the command 7. Goodwin DW, Alderson P, Rosenthal R: Clin- hallucination is a familiar one and (2) ical significance of hallucinations in psychiatric disorders. Arch Gen Psychiatry 24:76- whether there is a concurrent hallucina- 80, 1971 tion-related delusion, since these two fac- 8. Hellerstein D, Frosch W, Koenigsberg HW: The clinical significance of command hallu- tors have been empirically associated cinations. Am J Psychiatry l44:219-2 1, 1987 with increased compliance. The clinician 9. Junginger J: Command hallucinations and the can then use this information to estimate prediction of dangerousness. Psychiatr Serv 46:911-14, 1995 whether, how, and to what extent the 10. Junginger J: Predicting compliance with com- presence of command hallucinations may mand hallucinations. Am J Psychiatry 147: 245-7, 1990 be a risk factor for violence in a given 11. Chadwick P, Birchwood M: The omnipotence individual. of voices: a cognitive approach to auditory In estimating the degree of risk associ- hallucinations. Br J Psychiatry 164: 190-201, 1994 ated with command hallucinations, clini- 12. Taylor PJ: Motives for offending among vio- cians should rely heavily on empirical lent and psychotic men. Br J Psychiatry 147: 491-8, 1985 knowledge-to the extent that it is avail- 13. Kasper ME, Rogers R, Adams PA: Danger- able-to inform this decision. Informa- ousness and command hallucinations: an in-

358 J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998 Command Hallucinations

vestigation of psychotic inpatients. Bull Am for a content analysis of psychotic experience. Acad Psychiatry Law 24:2 19-24, 1996 Schizophr Bull 22:91-103, 1996 Thompson JS, Stuart GL, Holden CE: Cotn- 2 1. Rogers R: Conducting Insanity Evaluations. mand hallucinations and legal insanity. Foren- New York: Van Nostrand Rheinhold, 1986 sic Rep 5:29-43, 1992 22. Link BG, Andrews H, Cullen FT: The violent Zisook S, Byrd D, Kuck J, Jeste DV: Com- and illegal behavior of mental patients recon- mand hallucinations in outpatients with sidered. Am Sociol Rev 57:275-92, 1992 schizophrenia. J Clin Psychiatry 56:462-5, 23. Link BG, Stueve A: Evidence bearing on 1995 mental illness as a possible cause of violent Rogers R, Nussbaum D, Gillis R: Command behavior. Epidemiol Rev 17: 1-10, 1995 hallucinations and criminality: a clinical 24. Swanson JW, Estroff S, Swartz MS, et al: Violence and severe mental disorder in clini- quandary. Bull Am Acad Psychiatry Law 16: cal and community populations: the effects of 251-8, 1988 psychotic symptoms, comorbidity and lack of Shore D, Anderson DJ, Cutler NR: Prediction treatment. Psychiatry 60: 1-22, 1997 of self-mutilation in hospitalized schizophren- 25. Swanson J, Borum R, Swartz M, Monahan J: ics. Am J Psychiatry 135: 1406-7, 1978 Psychotic symptoms and disorders and the Chadwick P, Birchwood M: The omnipotence risk of violent behaviour in the community. of voices. 11: The Beliefs About Voices Ques- Crim Behav Ment Health 6:317-38, 1996 tionnaire (BAVQ). Br J Psychiatry 26. Rogers R, Gillis JR, Turner RE, Frise-Smith 166:773-6, 1995 T: The clinical presentation of command hal- Beck-Sander A, Birchwood M, Chadwick P: lucinations in a forensic population. Am J Acting on command hallucinations: a cogni- Psychiatry, 147: 1304-7, 1990 tive approach. Br J Clin Psycho1 36:139-48, 27. Hafner H, Boker W: Crimes of Violence by 1997 Mentally Abnormal Offenders. Cambridge: Junginger J: Psychosis and violence: the case Cambridge University Press, 1982

J Am Acad Psychiatry Law, Vol. 26, No. 3, 1998