British Journal of Psychiatry (1994), 164, 190—201

The Omnipotence of VoicesA Cognitive Approach to Auditory Hallucinations

PAUL CHADWICK and MAX BIRCHWOOD

We offerprovisionalsupportfora new cognitiveapproachtounderstandingand treatingdrugresistant auditory hallucinations in people with a diagnosis of schizophrenia. Study 1 emphasisesthe relevance of the cognitive model by detailing the behavioural, cognitive and affectiveresponses to persistent voices in 26 patients, demonstrating that highly disparate relationshipswith voices - fear, reassurance, engagement and resistance —¿�reflect vital differences in beliefsabout the voices. All patients viewed their voices as omnipotent and omniscient. However,beliefs about the voice's identity and meaning led to voices being construed as either‘¿�benevolent'or ‘¿�malevolent'.Patients provided cogent reasons (evidence) for these beliefswhich were not always linked to voice content; indeed in 31 % of cases beliefs were incongruouswith content, as would be anticipated by a cognitive model. Without fail, voices believed tobe malevolent provoked fear and were resisted and those perceived as benevolent were courted.However, in the case of imperative voices, the primary influence on whether commands wereobeyed was the severity of the command. Study 2 illustrates how these core beliefs aboutvoices may become a new target for treatment. We describe the application of an adaptedversion of cognitive therapy (CT)to the treatment of four patients' drug-resistant voices. Wherepatients were on medication, this was held constant while beliefs about the voices'omnipotence, identity, and purpose were systematically disputed and tested. Large and stablereductions in conviction in these beliefs were reported, and these were associated with reduceddistress, increased adaptive behaviour, and, unexpectedly, a fall in voice activity. Thesechanges were corroborated by the responsible psychiatrists. Collectively, the cases attestto the promise of CT as a treatment for auditory hallucinations.

Auditory hallucinations have a powerful impact onthe lives of those who experience them (Falloon &Talbot, 1981). However, the experience is alsopersonal: some people experience them as immenselydistressing and frightening whereas others arereassured and amused and many actually seekcontact. Where some people shout and swear atvoices and resist commands, others may harmthemselves or other people at the voices' behest.

This diversity begs the question, how are the contentand form of the voice, and the person's cognitive,affective and behavioural response, connected? Thata link exists is well established. Romme & Escher(1989) used innovatory sampling methods to studyhallucinations in clinical and non-clinical groups, andthey showed how a person's ability to cope withvoicesvariedaccording to his/herappraisalof the

voices. Benjamin (1989) studied 30 hallucinators andfound that all had meaningful, integrated and interpersonally coherent relationships with their voices. Inpeople diagnosed as schizophrenic these relationshipswere orderly and interpersonally ‘¿�normal',but notalways complimentary; these patients might claimthat their voices liked them even though the contentwas hostile and attacking. Benjamin's clearly statedposition (p. 293) is that the content of the voice is

“¿�directlyresponsible― for the person's behaviouraland affective response.

The question arises as to why the individual shouldfeel compelled to behave and respond affectively insuch an apparently congruous manner. Recent workhas suggested that essentially normal thoughtprocesses are involved in the way that delusionalideas are formed (Maher, 1988) and given upduring recovery (Brett-Jones et al, 1987). Indeedpsychological research shows that whether ordinarypeople accept comments and advice, and complywith commands, depends in large part on theirappraisal of the situation. For example, in Milgram'sfamous studies, whether ordinary individuals couldbe persuaded to administer what they believed to be alethal electric shock to other subjects was stronglyinfluenced by their beliefs about the experimenter'sauthority and power, their own degree of control,and the presumed consequences of disobedience(Milgram, 1974).

Likewise, it is possible that the degree of fear,acceptance and compliance shown to voices mightbe mediated by beliefs about the voices' power andauthority, the consequences of disobedience, and soon. For example, the belief that a voice comes froma powerful and vengeful spirit may make the person

190

191OMNIPOTENCE OF VOICES: COGNITIVE APPROACH

terrified of the voice and comply with its commandsto harm others; however, if the same voice wereconstrued as self-generated, the behaviour and affectmight be quite different. This cognitive formulationof voices was inspired by Beck's cognitive model ofdepression(e.g. Becketa!, 1979)whichproposes that thebehavioural and affective symptoms are consequencesof particular negative beliefs (e.g. “¿�Iam worthless―)and not antecedents (e.g. “¿�Shedisagreed withme―). This premise gave rise to cognitive therapy(CT) for unipolar depression (Beck et a!, 1979), aneffective treatment approach that relies heavilyon the disputing and testing of beliefs. Theapplicability of the cognitive model to voicescould have similar important implications fortreatment, in that if beliefs about voices could beweakened, this might reduce associated distress andproblem behaviour.

Useful progress in the cognitive—behaviouralmanagement of schizophrenic symptoms has beenmade in recent years, largely concentrated on positivesymptoms (Birchwood & Tarrier, 1992). There isgrowing evidence that secondary delusions may beweakened using CT (e.g. Fowler & Morley, 1989;Chadwick & Lowe, 1990). Again, Kingdon &Turkington (1991) have described how a number ofestablished cognitive techniques may be used tohelp clients construe their symptoms and experiencein non-psychotic terms. However, research hasyet to investigate the clinical utility of suchtechniques in systematically identifying and categorising the types of belief people hold about theirvoices, and then making these beliefs the targetof treatment. This is the purpose of the presenttwo studies: if successful, the cognitive approachholds promise of a new method of easing distressand problem behaviour associated with drugresistant voices.

Study 1. Applicability of the cognitive modelto voices

The first study tested the prediction that the degreeof distress and problem behaviour are consequencesof beliefs about voices, and not of antecedents suchas voice content. Twenty-six chronic hallucinatorswith a diagnosis of schizophrenia or schizoaffectivepsychosis were interviewed within a cognitiveframework, in the manner in which Hibbert (1984)approached panic disorder. Specifically we assessed:(a) whether there exist beliefs that are held aboutvoices; and, if so (b) the reasons (evidence) for thesebeliefs, including voice content; and (c) to whatdegree these beliefs are tied to individuals' behaviouraland affective responses.

Method

Fourteen men and 12 women were selected who hadheard voices for at least two years; their average age was 35years (range 23—59).All except one were receiving depotneuroleptic medication at All Saints' Hospital, Birmingham;one was in hospital and the remainder were out-patients. Allsatisfied DSM—III—Rcriteria for schizophrenia or schizoaffective diSOrder(American Psychiatric Association, 1987).

All participants volunteered for the study, with norefusals. Each was asked whether he or she would be willingto meet with a psychologist to discuss the experience ofhearing voices. It was made clear that the discussion wouldbe confidential, and that information disclosed wouldneither be entered in the main case notes nor lead to achange in medication.

Interviewswereconductedby either one of the authors;in severalcasesit took more than one interviewto collectall relevant information. Information was gathered usinga semi-structured interview (details available from theauthors on request). Interviews assessed the content of thevoices, beliefs about voices, other collateral symptoms thatwere regarded as supporting the beliefs, other confirmatoryevidence, and influence over the voice. ‘¿�Confirmatoryevidence' refers to actual occurrences that are perceived tosupport a belief: for example, a belief that voices give goodadvice would be strengthened if complying with a commandled to a desired outcome. ‘¿�Influence'concerns whether theindividual could determine the onset and offset of the voice,and could direct what was said. The behavioural andaffective responses were also elicited.

Results

A cognitive analysis of each individual's experience ofvoiceswas completed on the basis of the interviewdata.The beliefs elicited fell into distinct categories: those abouta voice's identity, power, and meaning, and those aboutcompliance. While limited space prevents information forall 26 participants being given in detail in Table 1, wedescribe 12 people's experience of hearing voices; informationabout all 26 patients' experience is available from theauthors on request.

Beliefs about voices

Omnipotence. All voices were perceived as being extraordinarily powerful. Nineteen patients (73%) reportedcollateral symptoms (e.g. visual hallucinations) thatcontributed to the sense of omnipotence. In six instancesthe experience of control was attributed to the voices.However, not all the evidence was of this type: 11 people(42°lo)gave examples of how they attributed events to theirvoices, and then cited the events as proof of the voices' greatpower. For example, although S14 and Sl6 both cut theirwrists under their own volition, they subsequently deducedthat the voices had somehow made them do it. Similarly,S17 attributed responsibility for his having sworn out loudin church to his satanic voices. This ‘¿�superhuman'qualitywas reinforced in all cases by ‘¿�hard'evidence that the voiceswereomniscient—¿�that is, knowing(i.e. commentingon the

192 CHADWICK & BIRCH WOOD

person's present thoughts and past history, and predicting

his or her future. Finally, 21 people (8lWo) were unable toinfluence either the onset and offset of their voices or what

was said, once again suggestive of the voices' power.Malevolence and benevolence. Four broad classes of

belief emerged, and three representative examples of eachappearin Table 1. Twelvepeople(Sl—S12)believedthevoices were malevolent. Malevolence, the wish to do evil,took one of two forms: either a belief that the voice wasa punishment for a previous misdemeanour, or anundeserved persecution. For example, Sl believed he wasbeing punished by the Devil for having sinned, and S3believed he was being persecuted without good reason byan ex-employer. Six people (S13—Sl8) believed the voicesto be benevolent. Benevolence, the wish to do good, tooka number of forms: to help the person maintain mental wellbeing; to protect the person, often from malevolent voices;an advisory role; to help the person develop social power;to develop a personal interest in the person (e.g. marriage).For example, Sl4 believed that the voices were from Godand were there to help develop a special power.

The third group (S19—S23)comprised five people whobelieved they heard a mLxture of benevolent and malevolentvoices; a paradigm of this group was S23, who wastormented by a group of evil space-travellers and yetprotected and nurtured by a guardian angel. The final threepeople (S24—S26)were uncertain about their voices becauseof an inconsistency or incongruity in what was said.‘¿�Uncertainty'was defined as having a strong doubt aboutthe voice's identity, meaning or power, where this doubt wasthe result of the person's deduction. For example, S26 wascertain that his voices wanted to help but observed that theyhad got things wrong: they wanted him to kill himself andmove on to the next and better life, yet his religion toldhim that suicide is a sin and those who commit it go to hell.

Connection with voice content. The voice content wasfrequently put forward as evidence for a particular belief.Thus it was commonly said that evil commands wereevidence that the voice was bad, and kind protective wordswere evidence that the voice was good. Also, as we haveseen, the belief in omnipotence was supported in all casesby the apparent omniscience of voices; they all ‘¿�knew'aboutthe person's private thoughts and actions, and manyaccurately foretold the future.

However, the class of beliefwas not always understandablein the light of voice content alone —¿�that is, in eight cases(31%) the beliefs appeared to be at odds with what wassaid. In the case of S2 and Sl2 the voice content was benign,yet these voices were construed as malevolent. The reversewas true of S26, whose voice called him a fool and toldhim to commit suicide and yet was construed as benevolent.Similarly, S24's voice told her to kill her family and herself;she believed the voice was God's and that he was givingher the chance to see her dead daughter by going to heaven.Again, S25 believed his voices to be benevolent in spite ofthem telling him to kill his daughter. S 15's voice told himto commit suicide and yet still was thought to be abenevolent goddess. 5 18's voices insulted her and told herto kill and yet were seen as benevolent. Perhaps moststrikingly, S Il's voice identified itself as God and yet shedisregarded this and believed it to be an evil force.

Behaviouraland affective consequences

The behavioural responses to voices may be organised intodifferent categories. One important criterion is whether theperson willingly engages with the voice. Engagement maybe defined as elective listening, willing compliance, anddoing things to bring on the voices (e.g. watching television,being alone, calling up voices). Resistance was a secondcategory and may be defined as arguing and shouting (overtand covert), non-compliance or reluctant compliance whenpressure is extreme, avoidance of cues that trigger voices,and distraction. A final category, indifference, was definedas ignoring and disregarding the voice.

At times people who habitually engaged with voices triedto ‘¿�shutthem up to get some peace' or shouted at themwhen they became a nuisance. Therefore engagement,resistance and indifference are probably best thought ofas predominant behavioural dispositions that describe theperson's response to voices most of the time.

When these behavioural categories are compared withbeliefs about malevolence and benevolence, the results arestriking. Without fail, when a voice was believed to bebenevolent the person willingly engaged with it, and whena voice was believed to be malevolent it was resisted. Thosepeople who were uncertain about their voices displayed noclear pattern between beliefs and behaviour. Affectiveresponses to voicescorresponded very closely to behaviouralresponses. All 17 malevolent voices habitually provokednegative emotions (anger, fear, depression, anxiety). Tenof the 11 benevolent voices habitually provoked positiveemotions (amusement, reassurance, calm, happiness) whenthey spoke; the one exception was S16 who felt predominantly anxious on hearing a benevolent voice, perhapsbecause the voice issued warnings about possible danger.All three people who were uncertain about their voicesexperienced negative affect when these voices spoke.

Compliance with commands

Although people had clear intentions about whether tocomply with commands, at times these were compromised.Compliance is governed by at least five factors. Firstly,compliance can be total (all commands obeyed in full),partial (onlycertain commands obeyed), or absent. Secondly,there is the general disposition to consider; malevolentvoices are to be resisted and benevolent voices are to becourted. However, and thirdly, all people interviewed whoheard imperative voices had additional beliefs about theconsequences of obedience and disobedience (see Table I).Beliefs about disobedience varied in severity from beingnagged to being killed. A fourth factor is the commandsthemselves —¿�voices may give innocuous commands (“makea cup of tea―,“¿�watchyour step―),or severe commands(“killhim―), or both. Finally compliance would seem tobe influenced by extra factors such as the person's mood,and the pressure and persistenceof the voices —¿�patients whorespond to such factors often present as losing tolerancefor their voices.

We feel it is helpful to classify commands as mild orsevere (i.e. life-threatening). Immediately one parametercan then be established. Fourteen voices (eight benevolent,six malevolent) gave only mild commands; all these

Patient no., Voice content Beliefs: identity, (I)Evidence: content(C),AffectiveresponsetoBehaviouralresponsesex,age,meaning (M), andsymptom (S), attri voicestovoicesduration

ofeffect of compliancebution (A), influence(I)illness(C)

OMNIPOTENCE OF VOICES: COGNITIVE APPROACH 193

Table 1A cognitive analysis of voices: a sample of 12 from the 26 patients (all diagnosed schizophrenic)

I. Voice of Devil C. Content is evil and FrightenedM. Beingpunishedfor knows my thoughts Angry

killing someone and pastC.Devilwilldriveme S.Delusionofcontrol

mad if I don't obey I. None

I. First and only C. Readsthoughts Irritatedemployer and knows my past Depressed

M. Controllingme and S.Delusionofcontrolholding me back in A. Voice gets worselife if I make progress

C.My intelligenceis I.Nonebetter than his so Idon't obey

I. Voices from God C. Know thoughtsand have his power and history

M. To protect me and A. Their advice stopsdevelopmy powers conflict

C.IfIobeybadthingsS.Experienceofwill not happen control

I. No influence

I. Powerful witchesA. RecognisevoicesExhaustedwhoused to beC. ReadthoughtsandTormentedneighboursknow

pastScaredM.PunishmentforI.Nonebeing

noisyandstoppingstudyingC.

IfIdisobeytheykeepon atme

I. An ex-girlfnend,who isa goddess

M. She protects me.I hear becauseI'vegreat power

C.She nagsme ifIdon'tcomply

I.A deadfriendM. Iam intouchwith

another dimensionC.IfIresistIwillfall

victim to the other

side

SiM, 49 years>20 years

S2M, 43 years>20 years

S3M, 31 years10 years

S14M, 32 years8 years

Si5M, 24 years3 years

S16F, 23 years2 years

ImperativeTold to rapeand kill

Imperative“¿�Becareful―‘¿�Tryharder―

Imperative“¿�Beuntidy―“¿�Don'twash―

ImperativeVoices give

marvellous advice

Imperative“¿�Killyourself―“¿�Giveup smoking―“¿�Don'tgo to church

today―

ImperativeNot to trust peopleGiveswarnings

C. Know thoughtsand past

A.GrandiosedelusionI.Can callup voiceandinfluencecontent

S.DerealisationDelusionof reference

C. Sounds like friendandidentifieditself

C.Predicteddeathoffriend

C. Voice says soKnows what I'mthinking

A.Feltpinpricksandpassedoutonward—¿�influenceofvoice2

S.Theyappearas

StrongConfidentHappy

InterestedSome irritation

Anxious

Voice1.ReassuredVoice2.Fear

Shoutsbackandswears (covert)

CompelledtolistenNo compliance

Shoutsbackandswears(covert)

Compelledtolistenandcompliesunwillingly

Listensandargueswith voice

No compliance

Electivelisteningandcompliance

Elective listeningSelectivecompliance

Listens carefullyAvoids certain peopleCompliesfully

Voice 1. Listens attentivelyObeyswillingly

Voice 2. Shoutsback inmind. Distracts (TV,talksover).Somereluctant compliance

Si9 Voice1.Advisesand I.Voice1 ismyF, 34 years imperative boyfriend. Voice 210 years “¿�Maketea―,“¿�Youare is Russiansand

the bride in Bible― Germans“¿�That'spossessed― M. I have a missionVoice2.Imperative fromGod tomarry“¿�Killher―“¿�Stabher―my boyfriend

“¿�Eatearth― C. Russianswill nagand kill me if I resist shadows on the wall

194 CHADWICK & BIRCHWOOD

Table 1 lcontinued)

Patient no., Voice contentsex, age.duration ofillness

Beliefs: identity (I),meaning (M), andeffect ofcompliance (C)

Evidence: content (Cl. Affective response tosymptom (SI, attri- voicesbution (A), influence (I)

Behavioural responseto voices

S20Two pairs. ImperativeI. Two male film stars, S. Reference and Stars: feel happy andStars:electiveF,20 yearsand commentand their two visual hallucinations excitedcomplianceand2

yearsPair 1. “¿�Wewant tojealous friends and thought Jealousfriendslisteningknowyou―M. The stars want to stopping Upset and angryFriends: shoutback“She's

beautiful―know me and C. Know my historyand swear.Reluctant“Sitwhere we canperhaps marry me. and my thoughtspartialcompliancesee

you―The friends prevent andfuturePair2. “¿�She'sugly―this and want to kill A. They arepowerful“Doit properly―me and won't goaway“Killher―C. If I comply with the I.Nonestars

I feelbetter.Jealousfriendspunish

disobedienceS21Voice

1. ImperativeI. Devil and God C. 1 good, 1 bad AnnoyedDevil: avoid andresistF,26 years“Hit him―“¿�Killhim―M. Devil punishing for C. Know thoughts FrightenedSometimes give inand10yearsVoice 2. Imperative

Tells to resist Voice 1,and to do good thingscausing

parents' and pastdivorce. God S. Delusion ofprotecting me reference

C. Devil nags if Iresist. God wantsme to resistcomply

ObeyGodS24ImperativeI.

God. but God S. Visual FearAppeases voice;partialF,59 years“You must kill Cathy,wouldn't tell to kill hallucinations (dead Reverencereluctantcompliance35yearsyour family and

yourself―M.Punishment for daughter; seen her

past misdemeanour ‘¿�age'over 30 years)C. If I comply I'll go C. Know what I'm

to heaven and see thinkingmy daughter anddogS25ImperativeI.

God and Devil have S. Delusion of control ScaredListen unwillinglyandM,54 yearsKill daughter, steal,entered my body. C. Know thoughts and Nervoussomeunwilling>

20 yearsread BibleBut God wouldn't past Miserableteach bad things I. None

M. They want to help;they ordain what wemust do

C. If I resist they bitemecompliance

Avoidance ofcuesS26ImperativeI.

Two benevolent They read my mindIrritatedIgnoreM,50 years“Kill yourself―spirits and know my pastNevercomply>20

years“He's a fool―Told to commit

suicideM.

Want me to go to Fantastic visualnext, better world hallucinations.

C. But I don't obey Suicide is a sinbecause they don't I. Nonemake sense

benevolent voices were complied with willingly and in full,and all but one malevolent voice was complied with infull, although reluctantly —¿�the one exception was S3 (seeTable 2). Severe commands were given by 12 voices (onebenevolent, 11 malevolent) and all were currently beingresisted. However, 10 of the 12 voices that gave severecommands also gave mild ones, and in all 10 cases thesewere obeyed at least occasionally. It is as if compliance withmild commands was an attempt to appease the voices'

requirement for sterner actions. Of those people who wereuncertain about the voice, two complied partially and onenot at all.

Comment

The study offers striking support for the cognitivemodel. All the patientsdisplayedmuch “¿�effortafter

PatientBeliefStart oftherapyConclusionoftherapyFollow-upMThe

prophetMasumaistalkingtome10000Thevoicesarehelpingme throughpersonaldifficulties10000I

cannotthinkformyselfwithoutthevoices852025Icannotcontrolthevoices1002015NGod

istalkingtome10055IfIresistIwillbepunished9055I

cannotcontrolmyvoices90350DTheDevilis talkingtome50200I

ambeingusedto transmitmessagesto thePrimeMinisterto ruinthe505050economyI

cannotcontrolthevoice1005085IfI complytheeconomywill bewined1002065TIfIdo notcomplytheywillmake my eyesrollup801515IfI talkwhilehallucinatingit will comeoutgibberish85100I

cannot control my voices952020

195OMNIPOTENCE OF VOICES: COGNITIVE APPROACH

Table2Percentageconviction scores in all beliefs at the start of therapy, at its conclusion, and at follow-up

beliefs that individuals hold about their voices. Theweakening or loss of these beliefs is predicted to easedistress and facilitate a wider range of more adaptivecoping strategies. The intervention briefly comprisesthe following components (a fuller account is inpreparation: Chadwick & Birchwood, 1994).

Opening phase: engagement, education and rapport

It has already been suggested that powerful beliefsand emotions characterise patients' relationships withtheir voices. It follows that patients are likely to findthe prospect of disengaging from their voices riskyand uncomfortable. For this reason, it is essentialthat the therapist comes across as competent,trustworthy, and as understanding about voices. Inour experience it helps if therapists acknowledge thatthey do not hear voices, but emphasise that they havespoken to many people who do and have learnedabout voices in this way. We believe it helps too ifthe therapist anticipates responses and makespredictions: “¿�Usuallythe people I've spoken to aresurprised or shocked by how much the voices knowabout them―,or, “¿�Iexpect the voices are worse atcertain times of the day, and when you are underpressure―. We always tell patients that they maywithdraw from therapy at any point withoutpenalty, and this may also reduce anxiety andfacilitate engagement.

The central beliefs are defined at an early stagetogether with the evidence used to support them. Thecost in terms of distress and disruption associatedwith the voices is specified, and it is made clear howmuch of this is a consequence of the beliefs aboutidentity, power, meaning and compliance. Thepatient is asked to consider the advantages anddisadvantages of these beliefs being false; because the

meaning―(Bruner, 1957)and this meaning frequentlywent well beyond the information given by thevoices. On the basis of beliefs about presumedidentity, omnipotence, and purpose, voices werebelieved to be either malevolent or benevolent, andit is this belief that seems to underlie people'sbehaviour towards voices. This distinction adequatelyexplained important differences in distress andbehaviour (though not compliance), and thereforequestions the prevalent idea that people's copingresponse to voices emerges through trial and erroror serendipity (e.g. Falloon & Talbot, 1981).

These data are encouraging but preliminary, andneed to be replicated on a wider scale. Also, themethodology needs to be operationalised and themeasurements more objective. At present we areconstructing a self-report measure of malevolenceand benevolence, and this should show whether theseconcepts are mutually exclusive or in fact habitualdispositions that sometimes blur.

Study 2. Cognitive therapy (CT) with voices

In Study 1 we concluded that much distress andvoice-driven behaviour is shaped by beliefs about thevoices' power, identity and purpose. When voicesare viewed from this cognitive perspective a newpossible therapeutic approach is evident, namely, thedistress associated with voices might be eased byweakening these beliefs. In Study 2 we describe acognitivetherapywe havedevelopedthatisintendedto undermine the patients' central beliefs aboutauditory hallucinations in a systematic fashion. Fourcase summaries of the therapy in action are includedas provisional evidence of its usefulness.

The cognitive treatment approach to hallucinationsinvolves the elucidation and challenging of the core

196 CHADWICK & BIRCH WOOD

former usually outweigh the latter, this discussionmay be used as an inducement to engage in therapy.However, it is emphasised that one acceptableoutcome of therapy would be that the patientcontinue to hold his or her beliefs. Indeed, the wholetherapy is conducted within an atmosphere of“¿�collaborativeempiricism― (Beck et a!, 1979) inwhich beliefs are considered as possibilities that mayor may not be reasonable.

Very early on in the therapy, the patient isintroduced to other people who hallucinate, andviews videos of clients (at least some of whom havecompleted therapy successfully) discussing theirvoices. Information about voices is provided to backthis up. These measures serve a dual purpose. Firstly,it has been suggested that a common and beneficialtherapeutic process is ‘¿�universality'(Yalom, 1970) —¿�that is, the recognition that many others experiencethe same or similar problems. Secondly, it may preparethe patient for the later suggestion that the voicesmight be self-generated.

Disputing beliefs

Disputing a belief's veracity involves the use of twocognitive techniques: hypothetical contradiction andverbal challenge.

1-lypothelical contradiction (Brett-Jones et a!, 1987).This measure is thought to assess how open peopleare to evidence that contradicts their core beliefs. Thepatient is asked to consider how, if at all, a hypothetical but contradictory occurrence would alter abelief. For example, in the present study one womanheard a voice commanding her to kill, which shebelieved to come from God. She was asked if thisbelief would be altered should a priest inform herthat God would never command anyone to sin.

Verbal challenge (Chadwick & Lowe, 1990). Atfirst, the patient is asked to question the evidencefor his/her beliefs and to generate other plausibleinterpretations. For example, another woman in thepresent study believed that her voices foretold thefuture:as evidenceforthisshe citedhow theywould

predict her husband's return home a few minutesbeforehe actuallyappeared.This woman was asked:

“¿�Let'ssuppose for a moment that the voices cannotforetell the future; how else might they anticipateyour husband's return?―.

Having considered the evidence, the next stage intherapy is to question the beliefsdirectly.This

involves first pointing out examples of inconsistencyand irrationality, and then, offering an alternativeexplanation of events: namely, that the voices mightbe self-generated and that the beliefs are an attempt tomake sense of them. An inconsistency might be discussed

as follows: “¿�Yousay that if you don't comply withcommands you will be punished, maybe even killed;and yet you regularly resist the voices and theseconsequences have not occurred―. Suggesting thatthe voices might be self-generated needs to bedone carefully. It should be put forward as apossibility worth considering, not a certainty, andas a helpful way of making sense of things if thebeliefs fail.

Testing beliefs empirically

In all the cognitive therapies, beliefs are subject toempirical test. We use two approaches to testing. Onthe one hand, we have a set procedure for testingthe ubiquitous belief: “¿�Icannot control my voices―.First, it is reframed as: “¿�Icannot turn my voices onand off―.The therapist then engineers situations toincrease and then decrease the probability of hearingvoices.An initialthorough cognitive assessmentshould identify the cues that provoke voices, and onetechnique with a high likelihood of eliminatingvoices for its duration is concurrent verbalisation(Birchwood, 1986). For example, with one womanin the present study the voices were elicited bydiscussing a provocative topic, and then stopped byconcurrent verbalisation on a neutral topic. Thepatient rouses and quells the voices several times toprovide a complete test.

With all other beliefs the empirical test wasnegotiated by the patient and the therapist: forexample, the therapist might ask: “¿�Yousay that thevoices take all your decisions for you; can you thinkof how we might test if this is actually the case?―.It is essential to examine beforehand the implicationsof the test not bearing out the belief; if the beliefwill be modified or adapted, or whether the patienthas a ready explanation for the outcome that leavesthe belief untouched.

Case reports

Three out-patients and one in-patient were referred bypsychiatrists because of drug-resistant and troublesomeauditory hallucinations (all had participated in Study 1).One was not on medication, one had medication withdrawnshortly after the intervention started and reinstated at alowerdosejustbeforeitclosed,and two wereon establishedand stableregimesthatwere notalteredatallduringthestudy. All satisfied DSM—IlI—Rcriteria for schizophreniaor schizoaffective disorder. Cases 1, 2 and 4 were treatedby MB and case 3 jointly by MB and PC.

Case I. M

M is a 34-year-old married woman of Iranian origin witha three-yearhistoryof schizoaffectivedisorder.Onset

OMNIPOTENCE OF VOICES: COGNITIVE APPROACH 197

followedthedissolutionof herfirstmarriageandwasmarkedby visualhallucinationsof the IslamicprophetMasumaandaccompanying auditory hallucinations in Arabic. At thetime M experienced a significant depression. The visionsand depression resolved, but the voice remained and resistedneuroleptic treatment. For several months before the study,and throughout it, M was not on medication.

The voice was ever present and offered advice on herfunctioning as housewife (e.g. telling her to try new recipes),mother (e.g. telling her when to change a nappy) and wife(e.g. advising how to please her husband). M felt greatreverence for the voice and it directed the majority of herbehaviour. To most observers, includingher husband, Mwas increasingly undermined by her voices and was losingall confidence in her own judgement.

Four beliefs were identified concerning the voice'sidentity, purpose and power (see Table 2). Engagement wasinitially a problem because the voice was superficially sosupportive. Also, M disclosed that previous attempts tobe independent of the voice had dissolved once the voicesaid: “¿�Rememberyour faith―. Dependence was alsoreinforced by fear that if left alone she might fail in herdifferent functions as wife and parent.

Belief 3, that she could not think without the voice, wasweakest and therefore tackled first. The main piece ofevidence was that the voice was involved in all day-to-daydecisions. However, a detailed diary revealed that over aweek 5807o of her actions were voice-driven, 32% were selfgoverned, and lO°lorepresented a rejection of the voice'sadvice. M was impressed by this result. The belief aboutcontrolling the voice was refuted in the standard mannerof switching it on and off.

Belief 2, that the voice helped her through personaldifficulties, rested on two points: that it gave good adviceand that it predicted the future. However, discussionclarified that the predictions had high probability —¿�perhapsthat the husband would be home at the usual time. Themost potent advice was novel recipes. However, questioningrevealed that none of the ingredients was actually new anda subsequent test involving the husband confirmed that therecipes were not novel but rather variations on one theme.The main evidence that the voice was a prophet was thatit spoke in Arabic, directed her to pages of the Koran, andknew her thoughts. M acknowledged that this evidencewas equally consistent with the view that the voice was selfgenerated. As therapy progressed, M practised planningeach day's activity the preceding night and therebyincreasing her autonomy.

Therapy lasted 18 weekly sessions, some involving herhusband, and conviction in all four beliefs weakenedconsiderably over this time (Table 2). This improvementwas confirmed by her psychiatrist,who also noted asignificantliftin mood, confidenceand self-initiatedbehaviour.M volunteeredthatthevoicewas lessintrusiveand regarded it more as an irritation. These benefits ledto her returning to work.

Case 2. N

20years.Shefirstpresentedwithdisorientation,depressionand hallucinations; there ensued numerous suicide attemptsand bouts of depression, which were woven around thehallucinations. Her unusual presentation drew inconsistentdiagnosis. She was referred during a five-month admissionwhich had been precipitated by a worsening of thehallucinations and her beginning to act on them withpossible danger to herself and her family. During her stayin hospital she was treated with neuroleptics, electroconvulsivetherapy (ECT) and antidepressants, with no effect. Therewere no negative symptoms but N was moderately depressedas a result of the prolonged admission. Neuroleptics werestopped three weeks into the study and not resumed until amaintenance dose was initiated towards the close of therapy.

N heard one voice through her ears that announced itselfas God. The voice commanded her to kill herself, membersof her familyand a work colleague(“Doyour work, killher―, “¿�Goto the canal―). Also, the voice sometimesaccused persons who were not family members of beingevil and advised her against talking to them. On averagethe voice spoke in ten-minute episodes ten times aday. These episodes were linked to visual hallucinations ofher dead daughter and her dog, who “¿�cameto her―at nightand beckoned her to join them in heaven by obeying God'scommands. In this vein, at times N carried a knife withher, although she susbsequently suggested that this was anact of appeasement.

N believed the voice to be God, although she was puzzledwhy God should compel her to murder and suicide. Shealso believed that she could not control the voice and thatif sheresistedpunishmentwouldensue. Consequentlyshestopped talking and listened carefully when it spoke, andappeased it by shadowing commands and going a small waytowards compliance (e.g. carrying a knife).

CT spanned three months, with short meetings heldregularly. A central consideration that engaged N incollaborative empiricism was the question: “¿�Justsupposethat this is not the voice of God, what then would be theconsequence of compliance?―. N was already puzzled bywhy God should compel her to murder and suicide as thiswent against her Catholic faith (she was classified asuncertain in Study 1). The doubt that the voice was Godwas increased when she was invited to consider the wisdomof the voice's plan of how she was to complete the murders —¿�essentially, kill one person in a public place and then takea 12-mile bus journey to complete the mission. The beliefthat disobedience brought punishment was disputed bypointingoutthatshehad resistedthecommands forthreeyears, and sometimes even ignored the voice, usingdistraction.

There was much additional support for the belief thatthe voice was self-generated. It worsened when she thoughtabout her daughter and dog and eased when she usedconcurrent verbalisation, so displaying that she couldcontrol the voice. The visions were illogical and went againsther faith —¿�the dog spoke, whereas her faith stated thatanimals were without souls and could not enter heaven; herdaughter ‘¿�aged'over the years when ageing should not occurin heaven.

The outcome of treatment was very good. As is shown inTable 2, her conviction in the three beliefs fell dramatically.

N is a 59-year-old woman with a 16-year psychiatric history,and visual and auditory hallucinations dating back

198 CHADWICK & BIRCH WOOD

Also, she came to doubt strongly the identity of the visions,and on those nights when these troubled her she was ableto turn over and ignore them. These changes werecorroborated by her psychiatrist and she was grantedincreasing home leave which hitherto had been thought too

dangerous. At the time of writing there had been nohallucinations in either modality for 2 months.

Case 3. D

I) is a 41-year-old single unemployed economics graduatewith a ten-year psychiatric history. For the last three yearsshe had heard voices daily, in half-hour bursts, usually in themorning and at bedtime. The content was invariably to dowith economics, such as: ‘¿�‘¿�Infinitelypower the rise ininflation― , “¿�Negativelypower productivity a million,trillion times' ‘¿�. These and similar statements were usuallyperceived as commands and occasionally as predictions.D also held a delusional belief that she could transmit herthoughts using telepathy.

D believed the voice to come from the Devil, and thathe was using her telepathic power to destroy the Britisheconomy. Specifically, the Devil would give a command thatin economic terms was disastrous, D would be compelledto repeat this command and in so doing would unwittinglytransmit it telepathically to the Prime Minister, who wouldact upon it. She believed that if she resisted, the economy

would be saved but the voice would continue to torment her.These beliefs had a profound impact. Each time the voice

began she would resist by saying exactly the opposite ofthe command, until she finally weakened and repeated theDevil's command, when the voice would stop. Shemonitored the economy religiously and felt guilt, anger anddepression when it dipped.

Therapy lasted for 13 sessions spaced over six months. Eachsession lasted one to two hours. CT raised a number of doubtsin D's mind about her beliefs. For example, how was thePrime Minister to know that he was to act on the commands,and even should he know this, what does “¿�infinitelypower―require? Again, on the one hand she believed that repeatingthe Devil's commands would destroy the economy and yetshe reluctantly did repeat the commands on many occasions.The belief about compliance was tested rigorously. Themethod used was that a command be chosen and D repeatit up to 100 times and wait to see if disaster ensued. This wasdone sequentially with commands demanding enormousincreases in the cost of bus fares, the cost of milk, theinflation rate, the interest rates, and the level of personaltaxation. In all these the beliefs were refuted.

Initially CT worked well for D. Conviction in all fourbeliefs about the voice fell and she lost the certainty that shehad telepathic power (a fall from 100% to 50°7o).Also, shereported feeling less guilty and depressed, and a reduction invoice activity. Follow-up indicated that the extraordinaryeconomic events in Britain in 1992, including a 5°7orise and

fall in interest rates in one day, appeared to have undonesome of this good work. D was agitated and conviction

in three beliefs had risen sharply. On the positive side, conviction in one belief had continued to fall during the followup period, and, perhaps most importantly, D was no longer

resisting the voice by saying the opposite to the commands.

Case 4. T

T was a 29-year-old woman who had experienced voicessince the age of 17. When she was 19 she experienced aflorid episode that involved a worsening of the voices andrelated paranoid ideas about schoolfriends and her mother.This pattern wasrepeated threeyearslater. On neitheroccasionwas admission required as neuroleptics quelled the floridsymptoms, although the hallucinations persisted. Shehad been maintained on 15 mg trifluoperazine since 1981and this was maintained during the present study. AlthoughT was demoralised by her illness, there were no negativesymptoms and she held down a factory job.

She heard three voicestalking in the second or third personinside her head, and ‘¿�recognised'these voices as old schoolfriends. The content was hostile and threatening (“Wewon'tbe so lenient next time, we're going to make your eyes rollup―, “¿�Lookat her eyes―, “¿�Youcan't get rid of us―).T'seyes did frequently roll up in what appeared to be an ocularspasm. The voices also taunted her about a peccadillo fromher youth. She had full insight, but felt caught in the voices'power. Accordingly, the beliefs selected were that if shedid not do what the voices said they would make her eyesroll up, that if she tried to talk when hallucinating it wouldbe gibberish, and that she could not control the voices

CT took place fortnightly for about 18 months and thenevery two months for a year. In the case of T theunderstandable concern about questioning the voices'authority was confounded by concern that work colleaguesmight discover her problem. However, she engaged wellin collaborative discussion of alternative viewpoints andgained direct disconfirmation of all beliefs during CT. Itwas discussed how the eye rolling might be a spasmprovoked jointly by tension and her coping strategy of fixedstaring when hallucinating. She explored this by not leavingthe workplace when the voices instructed. She found thateye rolling occurred regardless of whether she complied withthe voices and that relaxation methods shortened the spasm.Similarly she discovered not only that talking over the voiceswas possible (initially she simply named objects in the room)but that it actually silenced them, and that her speech,although restricted, was far from gibberish. Finally, usingthis strategy in combination with resisting the voices allowedT to experience how she could indeed control them.

Conviction in all beliefs fell dramatically (see Table 2).She described a sense of liberation from the voices and aconsequent increase in social activity. These improvementswere corroborated by the referring psychiatrist. Also,frequency of voice activity was monitored throughout thestudy (see Fig. 1). Over a lengthy baseline period it fellslightly and then settled at 70—75minutes a day for four months;it fell abruptly during CT to around 25 minutes a day, whereit remained at follow-up.

Comment

Study 2 describes how the cognitive frameworkdescribed earlier has spawned a new treatment fordrug-resistant auditory hallucinations. The therapyinvolves disputing and testing the central beliefs

199OMNIPOTENCE OF VOICES: COGNITIVEAPPROACH

these voices are maintained by reinforcing affectiveand behavioural consequences, and that as theseweaken the voice activity is gradually extinguished.

The possibility that malevolent and benevolentvoices are maintained through different psychologicalprocesses begs the question of whether challengingbeliefs about voices is always beneficial. Given thatmalevolent voices appear to provoke mainly fear,distress and resistance, it might be presumed thatreduced voice activity was desired by all. In contrast,benevolent voices are usually courted and producepositive emotional responses; because of this,reduction or loss of the voices may be less desirableboth to the individual and to carers, who may perhapsfear a concurrent reduction in self-esteem. In thepresent study two women saw their voices as malevolent, one as benevolent, and one was uncertain, and yettherapy seemed to work well for each with no obviousuntoward consequence. At present we are runninga larger trial to examine whether the success ofcognitive therapy depends in part on these categories.

The cognitive model embodied in the present paperhas some important implications for treatment. Firstly,it implies that simply advising patients to use a different coping strategy is unlikely to be helpful because ofthe probable conflict with central beliefs. For example,if a person believes that the voices are malevolent andintolerant of disobedience, he or she may be reluctantto employ distraction techniques. A second implication is that a lot of effort needs to go intoestablishing rapport, trust, and confidence, becausethe voices have a strong emotional and cognitive holdover the patients. A fmal implication is that if therapyis successful the person will inevitably come tosee the voices as self-generated. In this sense,cognitive therapy for voices may be said to contributeto an individual's level of insight.

Although the methodology and measurement usedwere not sophisticated, the data and the reports fromthe responsible psychiatrists give the cognitiveapproach clinical validity. In no way could these casesbe proof of the cognitive model of voices, nor is itsuggested that they prove the efficacy of the cognitivetherapy. Rather, the purpose is to present provisionalsupport for a new approach to the understanding andtreatment of voices, one that emphasises the necessityof appreciating the patient's efforts to understand and structure the experience of hearingvoices.

100@‘¿�90

8070

@ 60C 50

‘¿�0e 40@ •¿�@ 30

@ 20(a-c 10

Baseline I Intervention I Follow-up

o 8 1624@ 404696 6472809696104112I2)1@1361441961601661?6Weeks

Fig. 1 Number of minutes patient T (case 4) spent hallucinatingeach day during baseline, cognitive therapy, and follow-up.

patients hold about their voices: namely, those ofidentity, meaning, power and compliance. Clinicallysignificant and stable reductions in the strength ofthe beliefs were reported in three cases described.These improvements were corroborated by thepsychiatrists concerned, and led to other changes inobjective burden. Thus, M returned to work, Tbecame more active socially, and N was dischargedfrom hospital —¿�an action which hitherto had beenthought too dangerous for her and her family.The fourth patient, D, also responded well to CT,but an extraordinary set of external economiccircumstances undid much of the earlier goodwork - this woman's experience stresses how important environmental factors are to the maintenance ofcore delusional ideas.

All four patients reported a reduction in thefrequency and duration of voice activity, and in twocases the fall was documented. This was unexpectedbecause easing the hallucinatory experience itself hadnot been an objective. How, then, might such achange have come about? There is now considerableempirical backing for the ‘¿�stress-vulnerability'model,which asserts that acute or chronic stress can precipitate or exacerbate episodes of disorder (Clements& Turpin, 1992). For malevolent voices it mightbe hypothesised that the distress and resistanceoccasioned increases the likelihood of voice activity,which leads to further distress, and so on in a viciouscircle: thus, successful therapy influences notonly the beliefs and their affective and behavioural consequences, but also the link between theseconsequences and the experience of hearing voices.However, benevolent voices habitually evoke positiveemotions and yet reportedly voice activity stilldiminished (see M). One possible explanation is that

Discussion

The first experiment bore out the prediction thatresistance, engagement and distress are consequences

200 CHADWICK& BIRCHWOOD

of people's attempts to make sense of the experienceof hearing voices. This cognitive perspective is incontrast to the ‘¿�copingstrategy' approach (Falloon& Talbot, 1981; Brier & Strauss, 1983; Tarrier, 1987)which rather assumed that behavioural and affectiveresponses (‘copingstrategies') are as it were randomlyassigned to hallucinators. The first experiment offersprovisional evidence that different ‘¿�copingstrategies'appear to be driven by differences in underlying beliefsabout voices, which in our sample was meaningfullycaptured by the malevolence/benevolence distinction.

Without exception, voices were seen as omnipotent -indeed, the presumed identity of a voice was almostalways one traditionally associated with omnipotence.For many patients this attribution was supportedby an experience of control, by fantastic visualhallucinations, and by the patient having noinfluence over the voice. Also, all voices were seenas omniscient, again emphasising their superhumanquality. These indications of omnipotence combineto imbue voices with a ‘¿�terrifyingand compellingquality' so that people feel caught in their power(Bauer, 1969, p. 199). A critical part of CT withvoices is to refute the belief in omnipotence and totest the possibility that the patient may learn toregulate the activity of the voices. Given the degreeof fear and reverence involved, we believe that anopen and collaborative atmosphere is vital.

The patients' beliefs displayed considerable effortafter meaning (Bruner, 1957) and went well beyondthe information given (i.e. content). Indeed, manytimes beliefs were at odds with content. There wasalso evidence of verbal regulation of behaviour, anormal process whereby strongly held beliefs candrive the way in which we behave, feel, and interpretevents (Vygotsky, 1962). For example, one manreported how he swore out loud in church, and howhe subsequently deduced that his powerful andmalevolent voice had been responsible for this action.In this regard, the voices' perceived omnipotence andomniscience are pertinent because people are morelikely to apportion blame to those with authority,knowledge and ability (Tennen & Affleck, 1990).What this points to is the fact that people's subjectiveexperience of voices is not an irrelevant by-productbut an active and potent influence. Strauss puts hisfinger on this point precisely: “¿�Whencloser attentionis paid to patients' reports of their experiences, onekey phenomenon suggested is the importance of theinteraction between the person and the disorder. Thisinteraction evolves over time and has implications forunderstanding, studying, and treating schizophreniaand related disorders―. (Strauss, 1989, p. 179).

The explanatory power of the cognitive model wasweakest in relation to compliance. In our group the

severity of the command, and not beliefs, was thesingle most important determinant of compliance —¿�there was no compliance with life-threateningcommands, and compliance with mild commandswas commonplace. This might be because therelationship with a voice is regulated by widerconsiderations such as protecting self-esteem, structureand involvement with the world (Strauss, 1989), orperhaps people ask themselves the question: “¿�Whatif my beliefs about the voices are wrong?―. Bothinterpretations would explain why severe commandswere resisted even when issued by benevolent voices.At present we are interviewing people who have actedon serious commands, to investigate whether suchcompliance is associated with factors specific to thehallucinatory experience (e.g. total certainty in thebeliefs) or more general predictors of violence (e.g.previous history).

The precise importance of beliefs within acognitive formulation is currently under review. Inits early form, the cognitive position took a simpleone-way causal model that gave cognition primacyover affect. The fmding that affect may also determinecognition has meant that new reciprocal models nowpredominate (e.g. Gilbert, 1984). In relation tovoices, our position is that beliefs are vital to themaintenance of affective and behavioural responsesand render them understandable. Reciprocally,behaviour and affect strengthen or weaken beliefs.It is probably futile to assert primacy for eithercognition or emotion, because starting points indynamic reciprocal sequences are arbitrarily defined,and because both responses are forever evolving (seeParkinson & Manstead, 1992).

What can be asserted is that the affective, cognitiveand behavioural responses evolve together and arealways meaningfully related. In the present study,behavioural and affective responses to voices werealways understandable in the light of the beliefs;without recourse to the beliefs many responses wouldhave seemed incongruous. Indeed, one advantage ofa cognitive formulation is that it draws out thestructure and meaning that exist within a person'ssubjective experience of illness —¿�something which isoften disregarded or overlooked within psychiatry(Strauss, 1989).

The two studies reported are preliminary experiments that require replicating. The critical conceptsof malevolence, benevolence and omnipotence needclear operational definitions and the measurementand methodology need to be tightened. Should theresults prove to be typical, we believe they havemajor implications for the treatment of drug-resistantpsychotic auditory hallucinations. An assumption ofprevious approaches to voice management has been

OMNIPOTENCE OF VOICES: COGNITIVE APPROACH 201

that alleviating distress is contingent upon eliminatingthe experience, and this has led to therapies such asmonaural occlusion and distraction (Birchwood,1986) or indeed pharmacotherapy. Against this, thepresent research strongly suggests that degree ofdistress is inextricably bound to subjective meaning,and that weakening critical beliefs about the voicesmight alleviate much of the associated distressand difficulty.

Acknowledgements

We should like to thank Joanne Mclllraith for her assistancein the preparation of this manuscript and our colleagues DrsEl-Gohary and Mistry of All Saints' Hospital, Birmingham, andDr Swatkins of Burton Road Hospital, Dudley, for their helpand cooperation.

References

AMERICAN PSYCHIATRIC ASSOCIATION (1987) Diagnostic and Statistical

ManualofMentalDisorders(3rdedn,revised)(DSM—III—R).Washington, DC: APA.

BAUER,S. (1970) The function of hallucinations: an inquiry intothe relationshipof hallucinatoryexperience to creativethought.In Origin and Mechanisms of Hallucinations (ed. W. Keup).New York: Plenum.

Baac, A. T., Russo, A. J., Sso@w,B. F., et al (1979) CognitiveTherapy of Depression. New York: Guilford.

BENJAMIN, L. S. (1989) Is chronicity a function of the relationship

between the person and the auditory hallucination? SchizophreniaBulletin, 15, 291—310.

BIRCHWO0D,M. J. (1986) Control of auditory hallucinations usingearplugs through occlusion of monaural auditory input. BritishJournal of Psychiatry, 149, 104-107.

—¿� & TARRIER, N. (1992) Innovations in the PsychologicalManagementofSchizophrenia.Chichester:Wiley.

BERrr-Jop4es, J., GARm, P. A. & HEMSLEY,D. R. (1987)Measuring delusional experiences: a method and its application.British Journal of Clinical Psychology, 26, 257-265.

BRIaR, A. & SrRAuss, J. S. (1983) Self control in psychiatricdisorders. Archives of General Psychiatry, 40, 1141-1145.

BRUNER, J. (1957) Going beyond the information given. In

Contemporary Approaches to Cognition (eds H. E. Gruver,

K. R. Hammond & R. Jessor). Cambridge, MA: HarvardUniversity Press.

CHADWICK,P. D. J. & LowE, C. F. (1990) Measurement andmodification of delusional beliefs. Journal of Consulting andClinical Psychology, 58, 225—232.

—¿� & BIRCHWOOD, M. J. (1994) Cognitive Therapy for

Delusions and Hallucinations. Chichester: Wiley. (In press.)CLEMENTS, K. & TURPIN, 0. (1992) Vulnerability models and

schizophrenia: the assessment and prediction of relapse. InInnovations in the Psychological Management of Schizophrenia(eds M. Birchwood & N. Tarrier). Chichester: Wiley.

FALLOON, I. R. H. & TALBOT, R. E. (1981) Persistent auditory

hallucinations: coping mechanisms and implications for management. Psychological Medicine, 11, 329—339.

FOWLER, D. & MORLEY, S. (1989) The cognitive behavioural

treatment of hallucinations and delusions: a preliminarystudy.Behavioural Psychotherapy, 17, 267-282.

GILBERT,P. (1984) Depression: From Psychology to Brain State.London: Lawrence Erlbaum Associates.

HIBBERT, 0. A. (1984) Ideational components of anxiety: their

origins and content. British Journal of Psychiatry, 144, 618-624.KINODON, D. 0. & TURKINCTON, D. (1991) The use of cognitive

behaviour therapy with a normalising rationale in schizophrenia.JournalofNervousDisease,179,207-211.

MAHER, B. A. (1988) Anomalous experience and delusionalthinking: the logic of explanations. In Delusional Beliefs (edsT. E. Oltmanns & B. A. Maher). New York: Wiley.

MARKOVA,I. S. & BERRI0S, 0. E. (1992) The meaning of insight inclinical psychology. British Journal of Psychiatry, 160, 850-860.

MILIGRAM, S. (1974) Obedience to Authority. New York: Harper& Row.

PARKINSON,B. & MANSTEAD,A. S. R. (1992) Appraisal as a causeof emotion. In Emotion (ed. M. Clarke). London: Sage.

R0soME,M. A. J. & EsCHER,A D. M. A. C. (1989) Hearing voices.Schizophrenia Bulletin, 209—217.

Suu)E, P. & BENTALL,R. P. (1988)SensoryDeception: A Scient(flcAnalysisofAuditoryHallucinations.Chichester:Wiley.

STRAuss, J. S. (1989) Subjective experience of schizophrenia.SchizophreniaBulletin,15,179—188.

TARRIER, N. (1987) An investigation of residual positive symptoms

in discharged schizophrenic patients. British Journal of ClinicalPsychology, 26, 141-143.

TENNEN,H. & AFFLECK,0. (1990) Blaming others for threateningevents. Psychological Bulletin, 108, 209—232.

VYGOTSKY,L. S. (1962) Thought and Language. Cambridge, MA:Harvard University Press.

YAI.oM,I.(1970)TheTheoryandPnzcticeofGroupPsychotherapy.New York: Basic Books.

Paul Chadwick, BA, MSc,PhD,CPsychol,Lecturer in Clinical Psychology, University of Birmingham, andClinical Psychologist, The Archer Centre, All Saints' Hospital, Lodge Road, Birmingham B18 SSD; MaxBirchwood, BSc,PhD,CPsychol,FBPS,Consultant Clinical Psychologist,All Saints' Hospital, Birmingham,and Professor of Psychology, University of Birmingham

6Correspondence

(First receivedAugust 1992,final revision April 1993,acceptedMay 1993)