CBT for psychosis and for command hallucinations

1. Introduction

Version followed: two linked protocols. (a) Generic individual CBT for psychosis (CBTp) as recommended in NICE CG178, following the London–East Anglia manual (Fowler, Garety & Kuipers, 1995) in the relapse-prevention form used in the PRP trial (Garety et al., 2008), with modular detail from the CCI therapist manual (Smith et al., 2003), which draws on that group, Tarrier, Chadwick and Kingdon & Turkington. (b) Cognitive Therapy for Command Hallucinations (CTCH) as in the Byrne, Birchwood, Trower & Meaden (2006) casebook manual, the first RCT (Trower et al., 2004) and the COMMAND trial (Birchwood et al., 2014), whose manual is Meaden et al. (2013). Later relatives (AVATAR therapy, SlowMo, Feeling Safe, Morrison's 2017 protocol) are not covered step by step; their texts are not drawn on [uncertain].

What it is. CBTp is a one-to-one, formulation-driven talking therapy for people with distressing psychotic experiences (voices, unusual or paranoid beliefs) and the distress, low mood and disability around them. It does not aim to remove symptoms by argument. It links the person's experiences, appraisals, emotions and coping into a shared understanding, then tests and loosens the appraisals and coping behaviours that keep distress going, and builds a plan for staying well [source: Garety et al., 2008; Smith et al., 2003].

How change happens. Distress and behaviour depend less on the voice or event itself than on what the person believes about it. Voices are treated as activating events; beliefs about them are what get tested [source: Smith et al., 2003; Birchwood et al., 2018]. For command hallucinations, CTCH draws on social rank theory: the hearer feels subordinate to a powerful, malevolent voice, so they comply or appease to avert punishment. Therapy aims to reduce "the perceived power of voices to harm the individual and to motivate compliance" [source: Trower et al., 2004]. If perceived power falls, the drive to comply or appease falls, and so do harmful compliance and distress [source: Birchwood et al., 2014; 2018].

2. Framework

Format and dose.

• NICE CG178: individual CBTp delivered one-to-one over at least 16 planned sessions, following a treatment manual, offered to everyone with psychosis or schizophrenia (first episode, later episodes and for recovery), alongside family intervention where possible [source: NICE CG178; uncertain: the recommendations page would not load in full, so the wording is not verified here].
• PRP trial version: 12–20 planned sessions over 9 months, mean 14.3 sessions of about 1 hour [source: Garety et al., 2008].
• CTCH: first RCT, 6 months of therapy, median 16 sessions for completers [source: Trower et al., 2004]. COMMAND: up to 25 sessions within a maximum of 9 months, tailored to the person [source: Birchwood et al., 2018].
• Venue and timing are flexible (home, clinic, community), especially early on [source: Smith et al., 2003]. CTCH included in-vivo sessions, with the therapist accompanying the client on outings [source: Birchwood et al., 2018].

Session structure (every session): review progress and homework, agree an agenda, do the main work, summarise and set homework. Re-check mental state and medication regularly [source: Smith et al., 2003]. A structured agenda was what COMMAND participants recalled most [source: Birchwood et al., 2018].

Phases of generic CBTp (PRP version, with CCI modules mapped in)

[source: Garety et al., 2008; Smith et al., 2003. Session numbers per phase are not given in those sources and are my estimate: synthesis.]

Phases of CTCH

The trial papers name four belief targets rather than numbered phases: (1) the voice's absolute power and control; (2) the belief that one must comply or appease or be severely punished; (3) the voice's identity, e.g. the Devil; (4) the meaning of the voice experience, e.g. punishment for past wrongs [source: Trower et al., 2004]. COMMAND made reducing the power gap between voice and hearer central. It tested evidence for the hearer's lack of control over the voice, the voice's apparent omniscience, and its capacity to carry out threats [source: Birchwood et al., 2018]. A workable sequence is as follows. Its stage names follow the casebook's broad outline from recall and are [uncertain], since Byrne et al. (2006) and Meaden et al. (2013) are not drawn on.

3. Guiding principles

• Collaborative empiricism. Beliefs are hypotheses to examine together. In practice you are co-investigators, not the expert correcting errors [source: Trower et al., 2004].
• Socratic dialogue and guided discovery. Ask questions that let the person weigh evidence themselves. In practice you never tell the client the voice is not real or powerful; you help them find out [source: Trower et al., 2004].
• Neither confront nor collude. Listen to psychotic material respectfully, without challenging or endorsing it, in a curious "Colombo" stance. In practice, early on, ask about details and do not argue [source: Smith et al., 2003].
• Target distress and goals, not symptoms as such. Work on what the person wants to change. In practice, a voice can stay if it no longer frightens or commands [source: Garety et al., 2008; Birchwood et al., 2014].
• The relationship with the voice matters more than its content (CTCH). In practice, assess and target power, compliance and appeasement, not voice frequency [source: Birchwood et al., 2018].
• Normalise. Psychotic experiences sit on a continuum with ordinary experience and arise under stress and vulnerability. In practice, use stress-vulnerability and normalising information to cut stigma, catastrophising and hopelessness [source: Smith et al., 2003; Kingdon & Turkington, 1994, uncertain].
• Formulation-driven and individualised. In practice, choose techniques from the shared formulation, not a fixed order. Modules are taken flexibly [source: Smith et al., 2003; Garety et al., 2008].
• Engagement first. In practice, accept slow early sessions, practical help and flexible venues. Rapport is the first task [source: Garety et al., 2008; Smith et al., 2003].
• Transparency. Explain the therapist's role, the team, confidentiality and its limits, time limits, and the person's right not to attend. In practice this lowers suspicion [source: Smith et al., 2003].
• Hope with realism. In practice, address both unrealistically hopeful and hopeless expectations early [source: Smith et al., 2003].
• Pace to insight and distress. In practice, if the person becomes distressed, pause or step back to coping work [source: Smith et al., 2003].
• Measure. In practice, rate conviction, distress and preoccupation, or voice power and compliance, at baseline and repeatedly, and feed results back qualitatively, not as raw scores [source: Smith et al., 2003].
• Behavioural tests over argument. In practice, beliefs shift most when the person gathers disconfirming evidence by acting [source: Trower et al., 2004].
• Respect motivation and ambivalence. Symptoms can bring perceived benefits, and change can threaten support. In practice, weigh costs and benefits openly [source: Smith et al., 2003].
• Work within the team. In practice, keep CBTp alongside medication, care coordination and family intervention, and share relapse plans with consent [source: Garety et al., 2008; synthesis].
• Supervised, competent delivery. In practice, deliver CBTp under specialist supervision with recorded sessions where the client agrees. The trials used tape-rated competence (Cognitive Therapy Scale, CTS-Psy adherence) [source: Garety et al., 2008; Trower et al., 2004].

4. Intervention inventory

Part A covers generic CBTp. Part B covers CTCH-specific work, which sits on top of Part A's engagement, formulation and coping work. Sources: Garety et al. (2008), Smith et al. (2003), Trower et al. (2004), Birchwood et al. (2014; 2018). All example language is illustrative and my own.

Part A: Generic CBTp

A1. Engagement and orientation to therapy

• Purpose and when: Sessions 1–3 and whenever trust dips. The key early task is building and keeping a working relationship, often with someone acutely unwell [source: Garety et al., 2008].
• Steps:
  1. Offer a flexible venue and timing, and keep early sessions shorter if concentration is poor.
  2. Explain your role, the team, confidentiality and its limits, the number of sessions, and the right not to attend.
  3. Ask what they hope for and expect from therapy. Gently correct expectations that are hopeless or magical.
  4. Deal with urgent practical needs first if they block engagement.
  5. Hear psychotic material out with curiosity. Neither challenge it nor agree with it.
  6. Check the agenda every session. [source: Smith et al., 2003]
• Example language (illustrative): "I'm not here to tell you what's real. I'd like to understand what's been happening and what's been hardest, and then see what might help. What would you want to be different in a few months?"
• Watch for: suspicion of you or of note-taking; earlier bad experiences with services. Raise confidentiality early. Signs it is working: they turn up, bring their own topics, and disclose more detail over time.

A2. General and symptom-specific assessment

• Purpose and when: Sessions 1–4, repeated at review points. It maps problems, symptoms, coping and the person's own explanation, and sets baselines [source: Smith et al., 2003].
• Steps:
  1. Spread assessment over several sessions and explain why.
  2. Ask what they understand about their diagnosis and experiences, in their own words.
  3. Choose standard measures to fit the presentation. The CCI manual lists a diagnostic interview, quality-of-life and disability measures, a symptom scale, and depression and anxiety measures. For voices it lists a cognitive assessment of voices interview, a beliefs-about-voices questionnaire and a voice topography scale. Use licensed versions; items are not reproduced here.
  4. For each belief: rate conviction and preoccupation (0–100 or a visual scale), and map what came before it, what keeps it going, and its emotional and behavioural consequences.
  5. For each voice: its physical features, content, beliefs about its identity, power and control, what happens when the person complies, cues and times it is absent, and emotional and behavioural consequences.
  6. Feed back a qualitative summary in A–B–C terms (activating event, belief, consequence), not raw scores. [source: Smith et al., 2003]
• Example language (illustrative): "When the voice says that, how strongly do you believe it, from 0 to 100? … And what do you do straight afterwards?"
• Watch for: someone who will not rate below 100%. Use preoccupation or distress ratings instead. Pause if distress rises. Get the exact wording of voices: content described as threatening sometimes turns out to be neutral, and vice versa [source: Smith et al., 2003].

A3. Problem list and goals

• Purpose and when: End of assessment. It agrees on a few specific, realistic targets in the person's terms [source: Smith et al., 2003].
• Steps: (1) List the problems they name, using their words, e.g. "my concerns about the neighbours". (2) Turn two or three into concrete goals, e.g. go to the shop twice a week, or feel less frightened when the voice speaks. (3) Rate progress on each goal. (4) Revisit at reviews.
• Example language (illustrative): "If therapy went well, what would you be doing that you're not doing now?"
• Watch for: goals framed as proving the belief, e.g. "get police to arrest the neighbours". Reframe toward the distress or function behind them [synthesis].

A4. Coping strategy enhancement (CSE)

• Purpose and when: Early, as a low-threat way to reduce distress and build a sense of control before any belief work. Derived from Tarrier's CSE [source: Smith et al., 2003; Tarrier et al., 1993, uncertain].
• Steps:
  1. List every current coping strategy without judging it yet: cognitive (e.g. distraction, self-talk), behavioural (e.g. activity, leaving), sensory (e.g. music, headphones), physiological (e.g. relaxation), and self-medicating (e.g. alcohol).
  2. Discuss each one's short- and long-term effects. Flag strategies that ease distress now but keep beliefs going, such as avoidance.
  3. Choose a few to strengthen. Write a plan on a worksheet for when the symptom strikes (CCI calls it a "fire drill"), set specific practice goals, and problem-solve likely blocks.
  4. Practise in session where possible, e.g. trigger a mild version, use the strategy, rate the change.
  5. Review, then keep, modify or drop each strategy. [source: Smith et al., 2003]
• Example language (illustrative): "Let's write down everything you already do when the voices get loud, even things you're not sure help."
• Watch for: strategies that are safety behaviours (A10) or harmful (substances). Signs it is working: the person can name a strategy that reliably takes the edge off, and has a greater sense of control.

A5. Normalising and psychoeducation (stress–vulnerability)

• Purpose and when: Early to middle phase. It reduces hopelessness, stigma and catastrophic meanings ("I'm going mad forever"), and with them low mood, withdrawal and suicidal thinking [source: Smith et al., 2003].
• Steps:
  1. Ask what they already understand.
  2. Go through basic, individualised information about psychosis.
  3. Draw the stress–vulnerability model with them and fill in their own vulnerabilities (e.g. family history, early adversity, low self-esteem), stressors (e.g. bereavement, sleep loss, trauma, drugs, isolation, moves), protective factors (coping, support, help-seeking, medication) and risk factors (substance use, isolation, unsuitable housing).
  4. Link stress to their own episodes, e.g. symptoms rose after leaving home.
  5. Where it fits, give normalising examples: unusual experiences under sleep deprivation, sensory deprivation, bereavement or trauma, and voice-hearing in the general population [source: Smith et al., 2003; Kingdon & Turkington, 1994; 2005, uncertain]. COMMAND used a DVD of other voice-hearers [source: Birchwood et al., 2018].
• Example language (illustrative): "Lots of people hear voices when under enough stress or without sleep. Shall we map what was going on for you before the voices started?"
• Watch for: lecturing. Keep it collaborative and paced to the person's insight. A sceptical person may accept only their personalised model, and that is enough. A diagnosis label is not needed [source: Smith et al., 2003].

A6. Shared formulation (including the relapse formulation)

• Purpose and when: Throughout, made explicit by mid-therapy. It gives a joint understanding that guides which techniques to use [source: Garety et al., 2008].
• Steps:
  1. Start with a maintenance cycle from a recent episode: trigger → experience (voice, odd event) → appraisal → emotion → behaviour (coping or safety behaviour) → consequences that confirm the appraisal.
  2. Add a longitudinal layer as trust grows: vulnerabilities, life events, beliefs about self and others.
  3. Explore their understanding of triggers and relapse risks. Where appropriate, develop with them a new model of the disorder that offers an alternative to the delusional explanation [source: Garety et al., 2008].
  4. Draw it on paper and give them a copy. Diagrams were valued by COMMAND participants [source: Birchwood et al., 2018].
  5. Revise it as new information comes in.
• Example language (illustrative): "So when you hadn't slept, you heard the comment, thought 'they're onto me', felt terrified and stayed in. Staying in meant you never found out what would happen. Does that fit?"
• Watch for: the person hearing the formulation as "it's all in my head". Keep their explanation on the page as one option. Signs it is working: they correct and add to the diagram themselves.

A7. Examining delusional beliefs: evidence and alternatives

• Purpose and when: Middle phase, once there is engagement and a formulation. It reduces conviction and distress linked to a target belief [source: Smith et al., 2003; Garety et al., 2008].
• Steps:
  1. Teach the A–B–C link with a neutral example, then a mildly upsetting one, then the target belief. Show that the same event can be read differently, with different feelings as a result.
  2. Track episodes in a thought diary: event, belief (rated), feeling, action.
  3. Gather evidence for the belief in detail and with respect first. Then explore evidence that does not fit, using questions.
  4. Generate alternative explanations together, including their own earlier views and what others think. Rate each one.
  5. Keep an evidence log between sessions.
  6. Reach a balanced view: a more flexible belief that feels less threatening, even if it is not given up.
  7. Move to a behavioural test (A9). [source: Smith et al., 2003; the CCI module text was only partly retrievable, so details of steps 3–6 are partly my synthesis]
• Example language (illustrative): "What's the strongest reason you have for believing they're following you? … Is there anything that doesn't quite fit? … What other explanations might be worth checking?"
• Watch for: starting with the most central, high-conviction belief. Work at the periphery first, e.g. a less central part of the belief or a less distressing one [synthesis; Kingdon & Turkington describe peripheral questioning, uncertain]. Signs it is working: conviction ratings drop, and the person spontaneously produces alternatives.

A8. Reasoning-style work (jumping to conclusions, belief inflexibility)

• Purpose and when: When the formulation shows hasty conclusions or rigid thinking keeping beliefs going. The PRP trial named reasoning styles such as jumping to conclusions as a target [source: Garety et al., 2008].
• Steps (synthesis): (1) Explain, in a non-blaming way, that everyone jumps to conclusions under threat. (2) Use their own recent examples to slow decisions down: "what else could explain it?", "what would I need to see to be sure?" (3) Practise gathering more information before deciding, on neutral and then personal material. (4) Link this to the experiments in A9.
• Example language (illustrative): "When you're scared, how quickly do you decide what's going on? What happens if you give it an extra hour before deciding?"
• Watch for: the person feeling called stupid. Frame it as a universal tendency. Detailed procedures, e.g. from Freeman and Garety's later work, are not drawn on [uncertain].

A9. Behavioural experiments and reality testing

• Purpose and when: After some belief flexibility has been established. It tests predictions directly [source: Smith et al., 2003; Trower et al., 2004].
• Steps:
  1. State the specific prediction and rate conviction, e.g. "If I walk to the shop, the men will attack me" at 90%.
  2. Agree an alternative prediction.
  3. Design a safe, small, graded test with clear observations. Agree what result would support each prediction.
  4. Check risk and consent. Agree an exit plan and a coping plan, and accompany the person in vivo if needed.
  5. Run it, record what happened, re-rate conviction, and draw conclusions together.
  6. Plan the next step up.
• Example language (illustrative): "What do you predict will happen, and how would we know? Shall we set it up so it's small enough that you'd be willing to try?"
• Watch for: experiments that could create real danger or confirm the belief by chance. Do not stage tests that involve confronting alleged persecutors. Stop if distress becomes unmanageable. Contraindications: acute crisis, active risk to self or others linked to the test, intoxication [synthesis].

A10. Identifying and dropping safety behaviours

• Purpose and when: When avoidance or protective rituals stop disconfirmation (e.g. staying in, hiding, vigilance, complying). Avoidance can ease distress now while keeping beliefs going [source: Smith et al., 2003].
• Steps (synthesis built on the sources): (1) List what the person does to stay safe from the feared outcome. (2) Ask what would happen if they did not do it. (3) Explain that safety behaviours stop them finding out whether the threat is real. (4) Drop them in graded steps inside behavioural experiments (A9), rating the outcome each time.
• Example language (illustrative): "You've been safe each time you kept the curtains shut. How would you ever know if you'd have been safe without that?"
• Watch for: substitute safety behaviours, and rising anxiety (expected and temporary). Signs it is working: they take part in more activities and conviction falls.

A11. Motivational balance sheet

• Purpose and when: When motivation fluctuates, or symptoms bring perceived gains (company, specialness) or change threatens support [source: Smith et al., 2003].
• Steps: (1) Draw four boxes: costs and benefits of changing, and costs and benefits of not changing. (2) Fill them in with the person, including the benefits of the symptoms and the fear of losing services or support. (3) Summarise the balance and link it to goals. (4) Come back to it when stress rises.
• Example language (illustrative): "What would you miss if the voices were quieter? What might you lose if things got better?"
• Watch for: making changes that disrupt their routine too much. Motivation shifts with stress [source: Smith et al., 2003].

A12. Cognitive work with voices (generic CBTp)

• Purpose and when: For distressing voices that are not mainly commands, or as groundwork for Part B. Voices are treated as activating events, and the beliefs about them as the target [source: Smith et al., 2003; Chadwick, Birchwood & Trower, 1996, uncertain].
• Steps:
  1. Keep a voice diary: situation, what the voice said, what they made of it, feeling, action.
  2. Find cues for onset and offset, and times the voice is absent.
  3. Identify beliefs about the voice's identity, power, intent (malevolent or benevolent) and control, and rate each one.
  4. Use A7 and A9 methods on those beliefs.
  5. Add coping strategies (A4).
• Example language (illustrative): "When the voice comes, who do you think it is, and what does it want?"
• Watch for: the voice commenting on therapy. Plan for this (see 5). Signs it is working: distress and preoccupation fall even when frequency does not.

A13. Mood, self-esteem and negative beliefs about self and others

• Purpose and when: When the formulation shows persistent negative beliefs about self or others, depression or social anxiety driving vulnerability. This is a PRP target [source: Garety et al., 2008]. The CCI manual has a module for secondary depression and anxiety [source: Smith et al., 2003].
• Steps (standard CBT applied; details are synthesis): (1) Identify core beliefs from the formulation, e.g. "I'm worthless" or "others are dangerous". (2) Link them to life history and to how they shape psychotic appraisals. (3) Use thought records, evidence logs, a positive-data log and behavioural tests to build alternative self-beliefs. (4) Treat depression with activity scheduling and cognitive restructuring, and social anxiety with graded exposure and dropping safety behaviours.
• Example language (illustrative): "The voice calls you worthless. Is that something you'd already been afraid was true about yourself?"
• Watch for: hopelessness and suicidal thinking, especially as insight grows. Assess risk regularly (see 5).

A14. Behavioural skills

• Purpose and when: Throughout, for low activity, withdrawal, anxiety and practical problems. The CCI module lists relaxation, graded exposure, activity scheduling, distraction and problem solving [source: Smith et al., 2003; steps not retrieved, standard CBT procedure given as synthesis].
• Steps (synthesis):
  • Activity scheduling: (1) Record current activity for a week. (2) Rate pleasure and mastery. (3) Plan small, valued activities into the week. (4) Review.
  • Relaxation: teach slow breathing or progressive muscle relaxation in session, then practise daily, then use when cues for voices or anxiety appear.
  • Graded exposure: build a hierarchy of avoided situations and climb it step by step, staying until anxiety drops. Get consent, rule out real danger, and do not use it during acute crisis.
  • Distraction: plan activities that absorb attention, e.g. humming, reading aloud, conversation, music.
  • Problem solving: define the problem, brainstorm options, weigh pros and cons, choose one, plan it, try it, review.
• Watch for: sedation or negative symptoms limiting uptake. Use very small steps.

A15. Early warning signs and relapse management plan

• Purpose and when: The final phase, and earlier if relapse risk is high. This was the PRP trial's main aim [source: Garety et al., 2008; Smith et al., 2003; Birchwood's earlier relapse-signature work, uncertain].
• Steps:
  1. Reconstruct the timeline of past episodes: what changed first (sleep, withdrawal, irritability, suspiciousness, odd experiences), then what followed.
  2. Identify their personal relapse signature and triggers (stressors from A5).
  3. Write staged actions: early signs → self-help, e.g. sleep, reduce stress, use coping and talk to someone. Later signs → contact the team or review medication. Crisis → crisis numbers.
  4. Name who else holds the plan (family, care coordinator) and get consent to share it.
  5. Rehearse the plan using a likely scenario. [source: Garety et al., 2008; Smith et al., 2003]
• Example language (illustrative): "Looking back at last time, what was the very first thing that was different, maybe weeks before?"
• Watch for: plans that are too generic. Signs it is working: the person can recite their first two signs and first two actions.

A16. Ending and post-treatment review

• Purpose and when: The last sessions. It consolidates gains [source: Smith et al., 2003].
• Steps: (1) Repeat the baseline measures and compare them with the person qualitatively. (2) Write a therapy summary or blueprint covering the formulation, what helped, the relapse plan and future goals [synthesis]. (3) Space out the final sessions and offer boosters where the service allows [synthesis]. (4) Hand over to the care team.
• Watch for: loss reactions or symptom flare-ups as therapy ends. Plan for them in advance.

Part B: CTCH (command hallucinations)

CTCH is offered to people whose command voices have led to harm or serious risk. COMMAND required at least 6 months of commands and recent major episodes of harmful compliance [source: Birchwood et al., 2014]. Use Part A's engagement, coping and formulation tools throughout. Keep the focus on beliefs about power, compliance and appeasement, not on getting rid of the voice [source: Trower et al., 2004].

B1. Assessing the voice relationship, commands and compliance

• Purpose and when: The first sessions. It maps the power differential and the compliance cycle [source: Trower et al., 2004; Birchwood et al., 2018].
• Steps:
  1. List every command, from innocuous to severe (harm to self or others), with how often each comes.
  2. For each, record what the person did: full compliance, partial compliance, appeasement (doing a lesser act or a gesture to placate the voice or "buy time"), or resistance. Note what they believed would happen if they did not comply [source: Trower et al., 2004].
  3. Rate beliefs about the voice: its identity, power to harm, knowledge of the person, control over them, intent, and the person's own power relative to the voice.
  4. Note emotional consequences (fear, shame) and safety behaviours.
  5. Use standard instruments. The trials used voice-belief, voice-power and compliance measures linked to Birchwood's group. Exact instruments in COMMAND were not verified, and items are not reproduced here [uncertain].
• Example language (illustrative): "When the voice tells you to do that and you don't, what do you believe it will do? How sure are you, 0 to 100?"
• Watch for: under-reporting of serious commands through shame or fear that the voice will punish disclosure. Ask directly and calmly. Watch also for the voice forbidding therapy (see 5).

B2. Risk assessment and safety plan for harmful commands

• Purpose and when: From session 1, and whenever command severity or compliance shifts. This is [synthesis] following standard risk practice; the trial papers do not give a protocol.
• Steps:
  1. Assess current commands to harm self or others: frequency, intensity, history of compliance, intent, plans, protective factors, substance use, and access to means.
  2. Write a collaborative safety plan: warning signs, coping strategies (A4), people to contact, team and crisis numbers, and lethal-means counselling (secure storage or removal of medication stockpiles, weapons and similar items, with family help).
  3. Agree with the team on thresholds for urgent review and information-sharing, and explain confidentiality limits at the outset.
  4. Follow local safeguarding and duty-to-protect procedures where commands target identifiable others.
• Example language (illustrative): "Because the voice sometimes tells you to hurt yourself, I'd like us to make a plan together for the times it's strongest."
• Watch for: risk rising mid-therapy when the person starts to resist and the voice escalates threats. Monitor more closely during B5 and B6.

B3. Promoting control

• Purpose and when: Early. It weakens the belief that the hearer has no control over the voice's activity, which COMMAND tested as one strand of power [source: Birchwood et al., 2018].
• Steps:
  1. From diaries, find cues that bring the voice on or make it worse (e.g. stress, isolation, certain places) and times it is quiet.
  2. In session, run a small test: an activity that tends to quiet the voice (e.g. subvocal humming or conversation), and where tolerable, a deliberate cue. Observe the change.
  3. Draw the conclusion with the person: "I can influence when it's loud."
  4. Build this into the coping plan. [source: Birchwood et al., 2018 for the target; the steps are partly my synthesis from CSE, uncertain]
• Example language (illustrative): "When you were talking to your sister on Tuesday, what was the voice doing? What does that tell us about who has some control here?"
• Watch for: the voice claiming credit for quiet periods. Treat that as a belief to test, not a fact.

B4. Testing the voice's claimed knowledge and predictions (omniscience)

• Purpose and when: After B3. It disputes the belief that the voice knows everything or foresees the future [source: Birchwood et al., 2018].
• Steps:
  1. List what the voice has claimed to know or predicted.
  2. Ask whether it ever says things the person did not already know.
  3. Log the voice's predictions in advance with the person, then check outcomes together. COMMAND participants recalled testing whether predictions came true [source: Birchwood et al., 2018].
  4. Consider alternatives: the voice may draw on the person's own memories and fears.
• Example language (illustrative): "Let's write down what it says will happen this week and check together next time."
• Watch for: vague, unfalsifiable predictions. Help the person pin down specifics in advance.

B5. Disputing the voice's power to carry out threats

• Purpose and when: The middle phase. This is the core of CTCH: it reduces the perceived power to harm that drives compliance [source: Trower et al., 2004; Birchwood et al., 2014].
• Steps:
  1. Collect the threats and the feared consequences of not complying.
  2. Review history Socratically: times the person did not comply, or only partly complied, and what actually followed.
  3. Examine the evidence the voice can act in the world, as opposed to only speak.
  4. Generate alternatives, e.g. the voice's power comes from frightening words, not deeds.
  5. Rate conviction in its power before and after.
  6. Move to behavioural tests (B6). [source: Trower et al., 2004; Birchwood et al., 2018]
• Example language (illustrative): "It said something terrible would happen if you didn't go to the bridge last March. You didn't go. What happened next?"
• Watch for: the person attributing bad luck to the voice after the fact. Agree in advance what would count as the voice carrying out a threat.

B6. Reducing compliance, appeasement and safety behaviours (behavioural tests)

• Purpose and when: After some loosening in B5. It gathers disconfirming evidence through action [source: Trower et al., 2004].
• Steps:
  1. Make a graded list of commands and appeasing acts, from the most harmless to the most serious.
  2. Start by not complying with, or delaying, a harmless command, or dropping a small appeasing ritual. Rate the feared outcome beforehand.
  3. Get explicit consent, plan coping for any escalation of the voice, set a review contact, and accompany the person in vivo if helpful [source: Birchwood et al., 2018; synthesis].
  4. Record what happened and re-rate the voice's power.
  5. Climb the list step by step. Never ask the client to test harmful commands by doing anything risky. The test is always to not do it, under a safety plan [synthesis].
• Example language (illustrative): "It tells you to tap the door five times or something bad will happen. Would you be willing to try three times this week and we'll see what happens?"
• Watch for: louder, angrier voices after resistance. Prepare the person for this as a sign the voice has words, not powers [synthesis]. Signs it is working: compliance falls, and so do power and distress ratings.

B7. Raising the client's own power and status

• Purpose and when: Alongside B5 and B6. It builds alternative beliefs about the person's own power and status, both in the social world and relative to the voice [source: Trower et al., 2004; grounded in social rank theory].
• Steps (synthesis built on the sources): (1) Gather evidence of the person's strengths, achievements and moments of standing up to people or the voice. (2) Re-rate their own power against the voice's after each successful test. (3) Work on social rank beliefs, e.g. "I'm inferior to everyone", with A13 methods. (4) Encourage valued roles and relationships. COMMAND participants also valued peer support [source: Birchwood et al., 2018].
• Example language (illustrative): "You've ignored it four times this month. What does that say about who's stronger now?"
• Watch for: discounting of successes. Use a written log.

B8. Responding to the voice differently (including talking back)

• Purpose and when: Once power beliefs have loosened. COMMAND participants reported practising new coping responses, including talking back to voices, and some found this hard [source: Birchwood et al., 2018].
• Steps (synthesis): (1) Draft calm, firm replies or boundary statements with the person, e.g. "I'll decide what I do". (2) Role-play them in session, with you voicing the command only if the person agrees and finds it tolerable. (3) Practise at low-distress moments first, quietly or with a phone held to the ear in public. (4) Review the effect on fear and on the voice.
• Watch for: the voice escalating, or the person feeling overwhelmed. Fall back on A4. AVATAR therapy develops this dialogue work, but it is not drawn on [uncertain].

B9. Re-evaluating the voice's identity

• Purpose and when: When identity beliefs (e.g. the Devil, God, a persecutor) underpin power [source: Trower et al., 2004].
• Steps: (1) Explore the evidence for the identity: what the voice says, how it sounds, when it comes. (2) Look for inconsistencies with that identity, e.g. a God who only repeats one's own fears. (3) Generate alternatives, e.g. a memory of an abuser's voice, one's own critical thoughts, or a stress-related experience (link to A5). (4) Rate each. [source: Trower et al., 2004 lists the target; steps are synthesis]
• Example language (illustrative): "If this were really the Devil, what would you expect it to be able to do? Does it fit?"
• Watch for: religious beliefs. Respect the person's faith and work with the specific attribution, not their religion. Consider consulting their faith leader with consent [synthesis].

B10. Meaning of the voice and origins of the belief schema

• Purpose and when: Later phase, when the voice is taken as a sign of something about the self (e.g. punishment for past wrongs, being bad). Where appropriate, explore where the beliefs came from so the person can understand why they developed [source: Trower et al., 2004].
• Steps (synthesis): (1) Identify the meaning, e.g. "I deserve this". (2) Draw a longitudinal formulation linking early experiences of subordination, shame, abuse or trauma to beliefs about self and the voice. (3) Test the meaning with A7 and A13 methods. (4) Build a compassionate alternative account.
• Watch for: trauma disclosures. Pace the work, check dissociation and distress, and refer to trauma-focused work if indicated.

B11. CTCH relapse prevention

• Purpose and when: The ending phase. It prepares the person for future escalation of commands [synthesis, extending A15].
• Steps: Add to the A15 plan: the commands most likely to return, the evidence log of non-compliance without consequence, rehearsed responses (B8), and early contact points if harmful commands intensify.
• Watch for: relapse during life stressors. Rehearse with a likely scenario.

5. Handling problems

The sources mention some difficulties: suspicion, fluctuating motivation, distress during assessment, ratings stuck at 100%, unrealistic expectations, urgent practical needs [source: Smith et al., 2003], and some participants finding talking back hard [source: Birchwood et al., 2018]. Everything else below is [synthesis] unless tagged.

• Suspicion of the therapist or of being recorded. Explain confidentiality and its limits early [source: Smith et al., 2003]. Make recording optional, show where notes go, and offer to write the notes together [synthesis]. If the person brings you into the delusion ("you're working for them"), treat it as a belief to understand, not to deny defensively: "What have I done that made that seem likely? What would help you check?" (illustrative) [synthesis].
• The voice forbids therapy or threatens punishment for disclosure (CTCH). Expect this. Ask directly what the voice says about the sessions. Make attending a first, low-level test of non-compliance (B6), with a coping plan for afterwards. Shorten sessions or meet more often if fear is high [synthesis].
• Stuck at 100% conviction. Stop disputing. Switch to distress, preoccupation and goal ratings [source: Smith et al., 2003]. Work on coping, mood and function, or on a peripheral part of the belief. Use "even if" reasoning: "Even if they are watching, how could you have a better week?" (illustrative) [synthesis].
• Resistance to the formulation or normalising. Keep the person's own explanation as one hypothesis alongside the stress–vulnerability one [source: Smith et al., 2003 on pacing to insight]. Agree to differ and work on shared goals [synthesis].
• Overwhelm or rising distress in session. Pause, ground, and return to coping strategies (A4) [source: Smith et al., 2003]. Recalibrate: smaller steps, less time on symptom content, more on problem-solving. Check for emerging trauma memories and dissociation [synthesis].
• Escalating commands or risk mid-therapy (especially during B5–B6). Re-assess risk and update the safety plan and lethal-means steps. Slow the graded tests and inform the team according to the agreed thresholds. Never set a test that involves acting on a command [synthesis].
• Suicidal thinking as insight grows (post-psychotic depression). Assess risk at every session where low mood appears. Work on hopelessness with normalising and recovery information (A5), mood work (A13) and the safety plan (B2). Involve the team [synthesis].
• Non-adherence to homework. Give simpler tasks: verbal homework, phone reminders, or the therapist and client starting the task together in session. Ask what got in the way and problem-solve [synthesis]. Cognitive difficulties are common, so use written summaries and diagrams [synthesis; COMMAND participants valued diagrams, Birchwood et al., 2018].
• Missed sessions or disengagement. Reach out by phone or letter and offer home visits or flexible venues [source: Smith et al., 2003]. Liaise with the care coordinator. Keep the door open rather than discharging quickly [synthesis].
• Ruptures, e.g. the client feels disbelieved or pushed. Name it and own any part you played (too much disputing, too fast). Return to collaboration and the client's goals [synthesis].
• Collusion worry. Exploring a belief respectfully is not agreeing with it. Reflect the person's view as their view ("so you're convinced that…") and hold alternatives open [synthesis; consistent with Smith et al., 2003].
• Negative symptoms and sedation limiting engagement. Shorter, more frequent or active sessions (walk-and-talk), very small activity goals (A14), and a medication review via the team [synthesis].
• Substance use undermining progress. Include it in the formulation as a stressor and a self-medicating coping strategy [source: Smith et al., 2003]. Address it using motivational methods (A11) and refer on if needed [synthesis].
• Acute relapse during therapy. Switch to engagement and coping only. Pause belief work, activate the relapse plan, and resume once stable. Garety et al. (2008) began therapy with people who were often acutely unwell, with engagement as the main task [source: Garety et al., 2008; synthesis].
• Talking back is too hard (CTCH). Some COMMAND participants struggled with it [source: Birchwood et al., 2018]. Use quieter forms (internal reply, writing), role-play first, or rely on non-compliance and coping alone [synthesis].
• Voice improvement leaves a void (loneliness, loss of identity). Plan social and valued activity, use peer support [source: Birchwood et al., 2018 for peer support] and do self-esteem work (A13, B7) [synthesis].
• Therapist anxiety about risk. Use specialist supervision with taped sessions, as in the trials [source: Garety et al., 2008; Trower et al., 2004], and share risk decisions with the team [synthesis].

Sources

Sources

• COMMAND trial record, Lancet Psychiatry 2014 (WRAP, University of Warwick)
• Birchwood et al. (2018) participants' experiences of COMMAND, BMJ Open (University of Birmingham repository)
• Trower et al. (2004) Cognitive therapy for command hallucinations: randomised controlled trial, British Journal of Psychiatry
• Garety et al. (2008) CBT and family intervention for relapse prevention and symptom reduction in psychosis, British Journal of Psychiatry
• Smith, Nathan, Juniper, Kingsep & Lim (2003) CBT for Psychotic Symptoms: A Therapist's Manual, Centre for Clinical Interventions (partly retrieved)
• A Casebook of CBT for Command Hallucinations, Routledge publisher page (contents only)
• NICE CG178 Psychosis and schizophrenia in adults
• Morrison & Barratt (2010) What are the components of CBT for psychosis? A Delphi study, Schizophrenia Bulletin (abstract only)

Further reading (not drawn on) [uncertain where relied on]

• Birchwood, M., Michail, M., Meaden, A., Tarrier, N., Lewis, S., Wykes, T., Davies, L., Dunn, G., & Peters, E. (2014). Cognitive behaviour therapy to prevent harmful compliance with command hallucinations (COMMAND): a randomised controlled trial. Lancet Psychiatry, 1(1), 23–33 (abstract read via WRAP).
• Byrne, S., Birchwood, M., Trower, P., & Meaden, A. (2006). A Casebook of Cognitive Behaviour Therapy for Command Hallucinations: A Social Rank Theory Approach. Routledge.
• Meaden, A., Keen, N., Aston, R., Barton, K., & Bucci, S. (2013). Cognitive Therapy for Command Hallucinations: An Advanced Practical Companion. Routledge.
• Chadwick, P., Birchwood, M., & Trower, P. (1996). Cognitive Therapy for Delusions, Voices and Paranoia. Wiley.
• Fowler, D., Garety, P., & Kuipers, E. (1995). Cognitive Behaviour Therapy for Psychosis: Theory and Practice. Wiley.
• Kingdon, D., & Turkington, D. (1994). Cognitive-Behavioral Therapy of Schizophrenia. Guilford; and (2005) Cognitive Therapy of Schizophrenia. Guilford.
• Tarrier, N., et al. (1993). A trial of two cognitive-behavioural methods of treating drug-resistant residual psychotic symptoms in schizophrenic patients. British Journal of Psychiatry, 162, 524–532.
• Morrison, A. P. (2017). A manualised treatment protocol to guide delivery of evidence-based cognitive therapy for people with distressing psychosis. Psychosis, 9(3), 271–281.