Atlas›Domain II · Behavioural & cognitive›Guide 099
Guide II-099

Cognitive bias modification, attention-bias modification

Colin MacLeod, Andrew Mathews2000Western Australia; Cambridge, UK
A 11 · L 79 · M 1015 interventions in the Trinity explorer
Client cardCompare
[source: …] where a claim comes from[uncertain] not confirmed in a primary source[synthesis] the writer's reconstruction

1. Introduction

Version followed: there is no single treatment manual. This guide follows the paradigms as written in published trial protocols: attention-bias modification (ABM) as the 8-session, twice-weekly face dot-probe protocol (Amir lineage, as specified in ClinicalTrials.gov NCT03932032); interpretation-bias modification (CBM-I) in two forms, Mathews and Mackintosh ambiguous-scenario training as run on the University of Virginia MindTrails platform (NCT05126862) and the Beard and Amir Word-Sentence Association Paradigm (WSAP) as run in the HabitWorks app at Butler Hospital (NCT03509181, protocol of 8 June 2021); positive-imagery CBM as in Blackwell et al. (2015); and alcohol approach-bias modification as in the Wiers and Eberl lineage, specified in NCT03898323. [source: protocols listed in Sources]

Cognitive bias modification (CBM) is a family of computerised training procedures, not a talking therapy. It comes out of experimental psychopathology; the work of Colin MacLeod (Western Australia) and Andrew Mathews (Cambridge, UK) around 2000 is usually cited as its origin. The tasks first measure an automatic processing bias, such as attention drawn to threat, negative readings of ambiguous events, or an automatic pull towards alcohol. They then reverse the measurement contingency so that hundreds of brief repetitions reward the opposite habit. The idea of change is that these biases help cause and maintain anxiety, depression and addiction, rather than just being by-products. Repeated, fast, largely implicit practice is meant to shift the bias without argument or insight, and symptoms should then ease or become easier to treat. [source: Wikipedia CBM; Najmi & Amir 2009; MacLeod & Mathews 2012] Its developers present it as experimental, and as an add-on to established treatment rather than a replacement for it (Najmi & Amir say AMP should be used "as a last resort" for OCD after evidence-based treatment). [source: Najmi & Amir 2009]

For the clinician, the practical job is: choose the right paradigm for the bias that maintains the problem, explain and set it up, make sure the dose gets done, monitor safety, and link the training to the rest of therapy. In the source trials the tasks were delivered by software, with research or non-clinician staff for support. [source: NCT03509181]

2. Framework

CBM has no therapeutic phases in the usual sense. The protocols do share a sequence, set out below. The session counts are those of the protocols opened; any criterion not stated in a protocol is marked [synthesis].

Phase A: Selection and baseline (1 contact).

  • Goal: confirm that a trainable bias plausibly maintains the presenting problem, and rule out reasons not to proceed (see Section 5).
  • Typical content: a symptom baseline (the trials used, for example, PHQ-9 and GAD-7), plus an optional bias measure: a dot-probe assessment, the WSAP or a Scenario Recognition Task, the Brief Body Sensations Interpretation Questionnaire (BBSIQ), or an alcohol AAT pre-test. [source: NCT03509181; NCT05126862; NCT03898323; NCT03932032]
  • Move on when: the client agrees to the rationale and the schedule, and has a device, a quiet place and a time slot. [synthesis]

Phase B: Orientation (part of the first training contact).

  • Goal: explain the rationale, practise the task, and set up the app or web access, reminders and the support plan.
  • In HabitWorks, a trained bachelor's-level coordinator oriented patients, reinforced the app's instructional videos, and offered brief in-person check-ins whenever the patient wanted them. [source: NCT03509181]
  • Move on when: the client completes practice trials correctly and can say what to do. [synthesis]

Phase C: Training block (the "dose"). Protocol doses by paradigm:

  • ABM face dot-probe: 160 trials a session, 8 sessions, twice weekly for 4 weeks. [source: NCT03932032] Najmi and Amir describe about 20 minutes a session, twice a week, for 8 to 10 sessions. [source: Najmi & Amir 2009]
  • CBM-I ambiguous scenarios (MindTrails): 40 scenarios a session, drawn from a pool of 200; about 20 minutes; 5 weekly sessions. [source: NCT05126862]
  • CBM-I WSAP (HabitWorks), acute phase: daily during partial hospitalisation, about 150 trials or about 10 minutes. [source: NCT03509181]
  • Positive-imagery CBM: 12 sessions over 4 weeks, delivered online, alternating between auditory and visual sessions. [source: Mental Elf summary of Blackwell et al. 2015]
  • Alcohol approach-bias modification: 200 trials a session (about 20 minutes, with a mid-session break); 6 sessions across 3 visits in NCT03898323. Earlier protocols used 4 sessions (Wiers et al. 2011) and 12 sessions (Eberl et al. 2013). [source: NCT03898323; Wiers 2011 and Eberl 2013; the counts are [uncertain]]
  • Move on when: the planned dose is complete. None of the protocols opened switch phases on symptom or bias thresholds. HabitWorks tied the change of phase to discharge. [source: NCT03509181] Treating a fall in the measured bias as a sign the target is engaged is [synthesis].

Phase D: Transition and maintenance (HabitWorks model).

  • Content: at home, at least 3 times a week for 1 month, about 75 trials or about 5 minutes; then 2 months of use as desired.
  • Staff phoned anyone who had done no session in the past week. [source: NCT03509181]
  • Move on when: the month is over. After that, use is client-directed.

Phase E: Review.

  • Content: repeat the symptom and bias measures, and decide whether to stop, add boosters, or move to (or intensify) standard treatment.
  • The MindTrails feasibility trial reassessed at weeks 3 and 5 and two months after the last session. [source: NCT05126862] The decision rules are [synthesis].

Format:

  • Individual and self-administered, on a computer, smartphone or web platform, in clinic or at home.
  • Each session is a single block of rapid trials, with a short break in longer sessions.
  • Clinician or staff contact is brief: orientation, check-ins and safety monitoring.

[source: NCT03509181; NCT05126862; NCT03898323]

3. Guiding principles

  • Target the mechanism, not the content of thoughts. Practical meaning: pick the task by the bias you see (attention to threat, negative interpretation, approach to a substance), not by diagnosis alone. [source: Wikipedia CBM; Najmi & Amir 2009]
  • Reverse the assessment contingency. Practical meaning: training is the measurement task with the contingency switched. For example, the dot-probe target always replaces the neutral face, and only benign readings are scored correct. Check that the software is set to training mode and not to the 50/50 assessment mode. [source: NCT03932032; Wikipedia CBM]
  • Dose through repetition. Practical meaning: many short, frequent sessions. A missed session matters more than a short one. [source: NCT03509181; NCT03932032]
  • Speed and accuracy, little deliberation. Practical meaning: tell clients to respond as quickly and accurately as they can. In the WSAP this is meant to make responding more automatic. Do not turn trials into discussions. [source: NCT03509181]
  • Implicit learning through a task-irrelevant feature (AAT). Practical meaning: in approach-avoidance training the client responds to picture tilt or format, not content, and the content-to-movement link does the training. Do not coach the client to push alcohol pictures "on purpose". [source: NCT03898323; Wikipedia CBM] Whether also telling clients the contingency helps or harms is [uncertain].
  • Corrective feedback shapes learning (CBM-I). Practical meaning: benign interpretations are confirmed and negative ones are marked incorrect. Prepare the client for being told they are "incorrect" (see Section 5). [source: NCT03509181]
  • Imagery strengthens interpretation training. Practical meaning: ask clients to imagine themselves in each scenario, in their own body and as vividly as they can. [source: NCT05126862; Blackwell et al. 2015 via Mental Elf]
  • Gradually raise difficulty. Practical meaning: MindTrails moves from word fragments with 1 letter missing to fragments with 2 letters missing, which keeps the client actively resolving each scenario. [source: NCT05126862]
  • Adjunct, not substitute. Practical meaning: offer CBM alongside, or after, established treatment. Never let it delay exposure and response prevention, CBT or medication review. [source: Najmi & Amir 2009]
  • Low clinician burden, but human support. Practical meaning: delivery can be delegated, but someone should orient the client, watch completion and follow up on lapses. [source: NCT03509181; NCT05126862]
  • Monitor safety even though the task is low-risk. Practical meaning: track mood and suicidality during intensive phases, require sobriety before alcohol training, and stop if the client is under undue strain. [source: NCT03509181; NCT03898323]
  • Personalise and refresh stimuli. Practical meaning: use situations relevant to the client's life and a large, rotating pool, so that training does not become rote recognition. HabitWorks drew on 600 or more word-sentence pairs, mixed anew each session. [source: NCT03509181]
  • Make it engaging. Practical meaning: scheduled prompts, points, levels and progress displays were built into HabitWorks to keep people using it. [source: NCT03509181]
  • Measure the target. Practical meaning: where you can, give an assessment version before and after training. Use different stimuli for assessment and training; NCT03932032 used different faces. [source: NCT03932032]

4. Intervention inventory

The inventory has five groups: A, attention; B, interpretation; C, imagery; D, approach-avoidance; E, the clinician procedures that surround any CBM task. The software runs the trials, so the "steps" below cover both what the program does (so you can check it is configured correctly) and what you do. Example language is illustrative unless it is quoted and cited.

A. Attention-bias modification (ABM)

A1. Dot-probe attention training (threat-avoidance training; Attention Modification Program, AMP)

  • Purpose / when to use:
    • Purpose: train attention away from threat cues in anxiety where vigilance for threat is part of the formulation, for example social anxiety (angry or disgusted faces) or generalised anxiety.
    • Najmi and Amir describe it as a game-like task, run twice a week for about 20 minutes, over 8 to 10 sessions. [source: Najmi & Amir 2009]
    • For OCD (contamination), they place it only after evidence-based treatment, possibly to help clients who refuse exposure begin approaching feared situations. [source: Najmi & Amir 2009]
  • Steps (protocol NCT03932032, youth social anxiety, face version):
    1. A fixation cross appears.
    2. Two photographs of the same actor appear one above the other for 500 ms: one angry, one neutral.
    3. Both faces disappear, and a probe (an arrow pointing left or right) appears where the neutral face was, on every training trial. It stays on screen until the client responds.
    4. The client presses the left or right mouse button to show which way the arrow points.
    5. The position of the neutral face, the actor and the probe type are fully counterbalanced.
    6. Dose: 160 trials a session, 8 sessions, twice weekly for 4 weeks.
    7. Comparison condition used in research: neutral-neutral face pairs with no probe, where the client clicks to go on. This is useful to know if a client was previously in a sham arm.
    [source: NCT03932032] Variants:
    • Word-pair versions, threat word versus neutral word, used in MacLeod et al. (2002). [source: MacLeod et al. 2002; details are [uncertain]]
    • Probes that are letters (e.g. E or F) or dots instead of arrows. [source: Wikipedia CBM]
    • Probe contingencies of less than 100%, which some protocols use. [uncertain]
  • Example language (illustrative):
    • "This is a short computer game. Two faces will flash up, then an arrow appears. Click to show which way the arrow points, as fast as you can without making mistakes. You don't need to think about the faces."
    • "Please don't try to work out the trick. Just keep a steady pace. The practice works by repetition."
  • Watch for:
    • Reaction times and accuracy in the session log: accuracy should stay high.
    • Signs the client is guessing the rule and staring at one location. Gently re-instruct them to look at the centre.
    • Fatigue towards the end of the 160 trials.
    • Distress from angry faces. This is usually mild; if it occurs, debrief.
    • Working: the attention bias score falls on a separate assessment block with different faces (A2), and the client reports less automatic scanning for disapproval in real situations. [synthesis for the clinical indicator]

A2. Dot-probe assessment of attention bias (pre/post measure)

  • Purpose / when to use: check that a bias exists before training, and that it has moved afterwards (target engagement).
  • Steps:
    1. Same trial structure as A1, but the probe appears with equal frequency at the threat location (congruent trials) and the neutral location (incongruent trials). Neutral-neutral filler trials are added.
    2. Example set-up in NCT03932032: 240 trials (80 congruent, 80 incongruent, 80 neutral), faces shown for 500 ms, probe shown for 1000 ms, using faces different from the training set. [source: NCT03932032]
    3. Bias index: mean reaction time on incongruent trials minus mean reaction time on congruent trials. A positive value means attention is drawn towards threat. [source: Wikipedia CBM states that bias is inferred from reaction-time differences; the exact formula is the standard convention, [synthesis]]
  • Example language (illustrative): "Before we start training, let's see how your attention works under time pressure. There are no right answers about the faces; just respond to the arrow."
  • Watch for:
    • Dot-probe bias scores are known to be noisy for individuals. Do not refuse or stop training on the basis of one score. [uncertain: psychometric literature]

A3. Other attention-training formats (named in the literature, procedures not verified)

  • Purpose / when to use: alternatives for clients who disengage from the dot-probe.
  • Steps:
    • Spatial-cueing and visual-search training (for example, repeatedly finding the one positive or neutral face in a grid of threatening faces) are named as ABM paradigms. [source: Wikipedia CBM]
    • Eye-tracking (gaze-contingent) variants, which reward looking at neutral rather than threatening faces, have been developed. [uncertain: primary papers; do not deliver without the developers' software and instructions]
    • "Attention control" or balanced-probe training, where the probe follows threat and neutral cues equally often, has been tested as an active comparison. [uncertain]
  • Example language (illustrative): "In this version you search for the one friendly face as quickly as you can."
  • Watch for: use only versions that come with published software and parameters. Do not build your own. [synthesis]

B. Interpretation-bias modification (CBM-I)

B1. Ambiguous-scenario training (Mathews and Mackintosh design; MindTrails version)

  • Purpose / when to use: anxiety (including health anxiety and anxiety in people with medical conditions), worry and low mood, where the client habitually reads ambiguous events as threatening. MindTrails offers this free on the web. [source: NCT05126862]
  • Steps:
    1. The client reads a short ambiguous scenario written in the second person and is asked to imagine being in it.
    2. The last word appears as a fragment with letters missing. The client must type the missing letters, and the completed word resolves the ambiguity, benignly in about 90% of scenarios.
    3. Early sessions remove 1 letter; later sessions remove 2.
    4. After completing the fragment correctly, the client answers a comprehension question about the scenario. The question checks reading and reinforces the benign meaning.
    5. Dose: 40 scenarios a session from a pool of 200, about 20 minutes, weekly for 5 weeks.
    6. Support: staff phone or email clients who fall behind.
    [source: NCT05126862; Wikipedia CBM, which describes the fragment-plus-question format and the typically benign resolution; Mathews & Mackintosh 2000, details [uncertain]]
    • The proportion and handling of the remaining non-benign scenarios in MindTrails is [uncertain].
  • Example language (illustrative):
    • "Read each little story and really picture yourself in it, as if it's happening to you now. Then fill in the missing letters. The word will tell you how the situation turns out."
    • "Some will seem too easy. That's fine. The point is practising the habit of landing on a non-threatening reading."
  • Watch for:
    • Rushing without imagining. Ask after a session: "Which one could you picture most clearly?" (illustrative).
    • Complaints that the scenarios are irrelevant or unrealistic (see Section 5).
    • Working: on review, the client spontaneously offers a benign alternative for ambiguous everyday events, and interpretation questionnaires (such as the BBSIQ) shift. [synthesis; BBSIQ use from NCT05126862]

B2. Word-Sentence Association Paradigm training (WSAP; Beard and Amir; HabitWorks version)

  • Purpose / when to use: a fast, repeatable interpretation training suitable for a smartphone. It was used with adults in a psychiatric partial hospital programme with mixed depression and anxiety, and was designed to support and extend acute treatment. [source: NCT03509181]
  • Steps:
    1. A fixation display appears.
    2. A word (benign or negative) is shown for about 500 ms, followed by an ambiguous sentence.
    3. The client presses a button to say whether the word is related or unrelated to the sentence.
    4. Feedback follows. Endorsing the benign interpretation, or rejecting the negative one, earns "You are correct!"; the reverse is marked incorrect. [source: NCT03509181, quoted under 15 words]
    5. Instructions stress responding "as quickly and accurately as possible". [source: NCT03509181]
    6. Dose:
      • Acute phase (hospital): daily, about 150 trials or about 10 minutes.
      • Transition month (home): at least 3 times a week, about 75 trials or about 5 minutes.
      • Long-term (2 months): as desired.
    7. The stimulus pool held 600 or more pairs, mixed anew each session. A personalised set based on the client's life situations was planned. [source: NCT03509181; whether personalisation was used in the trials is [uncertain]]
    8. Engagement features: scheduled prompts, points, levels and progress tracking.
  • Example language (illustrative):
    • "A word flashes up, then a sentence. Decide quickly whether the word fits the sentence. Don't deliberate; go with your first response. The app will tell you which reading we're practising."
    • "When it says 'incorrect', it isn't saying your feelings are wrong. It's nudging you to practise the less threatening reading."
  • Watch for:
    • The client's reaction to "incorrect" feedback, especially in depression or self-criticism.
    • Completion falling after discharge. In the protocol, staff called after a week with no sessions.
    • Working: endorsement of benign interpretations rises and negative endorsement falls on the WSAP assessment or the Scenario Recognition Task. [source: NCT03509181 lists these measures]

B3. Homograph and recognition-task procedures (assessment and early training formats)

  • Purpose / when to use: the recognition task measures the interpretations a person spontaneously makes.
  • Steps: the client reads scenarios that remain ambiguous, then rates how similar four kinds of sentence are to each scenario:
    • positive and negative interpretations that capture the scenario's emotional meaning, and
    • positive and negative foils that do not.
    Higher ratings for the positive interpretations indicate a more benign style. [source: Wikipedia CBM; Salemink & van den Hout 2010 cited in NCT03509181]
    • Homograph training, in which a word with a threatening and a neutral meaning is followed by a fragment of a neutral associate, was an early CBM-I format. [uncertain: Grey & Mathews 2000]
  • Example language (illustrative): "Read each story, then tell me how close in meaning each of these sentences is to what you read."
  • Watch for: use different scenarios for assessment and training. [synthesis]

C. Imagery-based CBM

C1. Positive-imagery CBM (Holmes and Blackwell, Oxford)

  • Purpose / when to use: depressed mood, where negative interpretations and a lack of vivid positive future imagery are part of the picture. It was tested as a low-intensity, web-based tool for adults in a current major depressive episode. [source: Mental Elf summary of Blackwell et al. 2015]
  • Steps:
    1. Training happens online at home: 12 sessions over 4 weeks.
    2. Half the sessions are auditory: the client listens to descriptions of everyday situations that begin ambiguously and resolve positively, and imagines each one as vividly as possible, as if happening to them.
    3. Half are visual: a picture is paired with a word, and the client combines them into one imagined positive scene.
    4. In the comparison condition, by contrast, people attended to the meaning of the words or built sentences from the picture-word pairs without imagining. That contrast is the key instruction to get right: imagine, do not just analyse the words.
    5. At least 6 sessions were treated as an adequate dose.
    [source: Mental Elf summary of Blackwell et al. 2015; trial NCT01443234]
    • The content of the preliminary imagery-practice session, the number of items per session and the vividness-rating format were not verified. [uncertain: the full text]
    • A related line of work tests whether clients generating their own positive resolutions works better than imagining given ones. [uncertain: "Optimizing the ingredients for imagery-based interpretation bias modification for depressed mood", title only]
  • Example language (illustrative):
    • "For each one, see it through your own eyes, as if you're there now. Notice what you'd see, hear and feel as it turns out well."
    • "If a picture won't come, start with one small detail, like the colour of the door or the sound of the room."
  • Watch for:
    • Clients describing themselves from the outside, as an observer, rather than from inside the scene. Re-instruct them to imagine from their own point of view. [synthesis]
    • Clients who cannot form images at all (see Section 5).
    • Working: imagined scenes feel more vivid, and the client shows more anticipatory pleasure and behavioural activation. [synthesis]

D. Approach-avoidance training

D1. Alcohol approach-bias modification (alcohol avoidance training, AAT)

  • Purpose / when to use: alcohol use disorder, as an add-on during or after detoxification or inpatient treatment, to weaken the automatic tendency to approach alcohol cues. [source: NCT03898323; Wikipedia CBM; Wiers et al. 2011, cited]
  • Steps (NCT03898323, Batki, San Francisco VA Medical Center, 2019):
    1. Check that breath alcohol is 0.00 and assess withdrawal before each session.
    2. The client uses a joystick. Pictures appear: 20 alcoholic drinks and 20 non-alcoholic drinks (soft drinks).
    3. The client is told to respond only to a feature that has nothing to do with content, the tilt of the picture: push for one tilt direction, pull for the other, whatever the picture shows.
    4. Pushing shrinks the picture (a zoom-out effect) and pulling enlarges it, so the movement feels like avoiding or approaching.
    5. In training, alcohol pictures appear only in the "push" format and soft drinks only in the "pull" format (100% contingency). Wiers et al. 2010 used 90%.
    6. Dose: 200 trials (about 20 minutes) with a short break halfway. In this protocol, 6 sessions were given across 3 visits.
    7. Stop immediately if the client appears under undue strain.
    [source: NCT03898323]
    • Earlier protocols used picture format (landscape versus portrait) as the irrelevant feature. [source: Wikipedia CBM describes a "distinguishing feature"; the format variant is [uncertain]]
  • Example language (illustrative):
    • "Ignore what's in the picture. Only notice which way it's tilted. Tilted one way, push the joystick away; the other way, pull it towards you. Go as fast and accurately as you can."
    • "You may notice you are pushing away a lot of drinks. That's part of the training; just keep responding to the tilt."
  • Watch for:
    • Craving or distress triggered by the alcohol pictures. Debrief, and link to the relapse-prevention plan.
    • Fatigue.
    • Intoxication or withdrawal (do not train).
    • Working: the approach-bias d-score on a separate AAT assessment falls (in NCT03898323, negative values mean less approach bias, from 80 test trials after 10 practice trials), and relapse-prevention goals hold. [source: NCT03898323; the clinical indicator is [synthesis]]

D2. Approach training for phobic avoidance

  • Purpose / when to use: the reverse use of the same joystick task, training approach to a feared stimulus. For example, in fear of spiders, spider pictures are pulled and comparison pictures pushed. [source: Wikipedia CBM; Rinck & Becker 2007, cited]
  • Steps:
    • The same mechanics as D1, with the contingency reversed so that the feared category is always pulled.
    • Session dose for clinical use was not found. [uncertain]
  • Example language (illustrative): "Respond to the tilt only. Some pictures will be of spiders; just keep going."
  • Watch for: fear and avoidance of the task itself. Treat this as preparation for, not a replacement for, in-vivo exposure. [synthesis]

D3. Memory bias modification

  • Purpose / when to use: memory is listed as a domain that CBM can target. [source: Wikipedia CBM]
  • Steps: no CBM memory-training procedure is given in the sources. [uncertain] Do not improvise one; refer to a published protocol.

E. Clinician procedures surrounding any CBM task

E1. Selection and matching

  • Purpose / when to use: at intake to CBM.
  • Steps:
    1. Write down the bias that maintains the problem in the formulation: threat vigilance (go to A1), negative interpretation (B1 or B2), lack of positive imagery (C1), or automatic approach to alcohol (D1).
    2. Confirm that established treatment is in place or has been offered first. [source: Najmi & Amir 2009]
    3. Optionally run the matching bias assessment (A2, B3, or an AAT pre-test).
    4. Check reading level, sensory and motor ability (joystick, timed responses), device access and language. [synthesis]
  • Example language (illustrative): "One thing that keeps your anxiety going seems to be that your mind jumps to the threatening reading before you've had a chance to think. There's a computer training that practises the other habit. It's an add-on to our work, not instead of it."
  • Watch for: clients who expect CBM to replace therapy (see Section 5).

E2. Rationale and orientation

  • Purpose / when to use: first training contact.
  • Steps:
    1. Explain the rationale in plain terms: anxious people tend to focus on negative information, and the task trains the opposite habit through repetition. [source: Najmi & Amir 2009]
    2. Show the instructional video or demo and supervise the first practice block.
    3. Set reminders and agree session days and times.
    4. Name a contact person for check-ins.
    [source: NCT03509181] Explicitly predicting that the task will feel boring and repetitive is [synthesis].
  • Example language (illustrative): "Think of it like physiotherapy for attention. Each session is short and repetitive, and the effect comes from doing it regularly."
  • Watch for: low credibility. In the HabitWorks trial, credibility and expectancy were measured (Credibility/Expectancy Questionnaire). [source: NCT03509181] Address doubts openly.

E3. Adherence monitoring and check-ins

  • Purpose / when to use: throughout the training block and the maintenance phase.
  • Steps:
    1. Review completion logs, daily during intensive phases.
    2. Contact the client if a week passes with no session (by phone in HabitWorks; by phone or email in MindTrails).
    3. Offer brief in-person check-ins whenever the client wants them.
    [source: NCT03509181; NCT05126862]
  • Example language (illustrative): "I noticed the app's been quiet this week. What got in the way? Shall we find a five-minute slot that'll work?"
  • Watch for: a pattern of drop-off after discharge or after the novelty wears off.

E4. Safety monitoring

  • Purpose / when to use: whenever CBM is delivered to clinical populations, especially people with depression, suicidality or substance use.
  • Steps (HabitWorks model):
    1. Monitor daily PHQ-9 scores during intensive care.
    2. Alert the treating team if the PHQ-9 rises by 5 or more points, or if item 9 rises at all.
    3. Positive answers to C-SSRS items 4 or 5 trigger an emergency evaluation, with risk assessed the same day.
    4. After discharge, the app directs the user to contact their provider, emergency services or a crisis text line when PHQ-9 item 9 is 1 or above.
    5. Every client has a safety plan and crisis resources.
    [source: NCT03509181] For alcohol training: sobriety check, withdrawal assessment and adverse-event review at every visit. [source: NCT03898323]
  • Example language (illustrative): "The app will ask how you're doing. If it picks up thoughts of not wanting to be here, it'll point you to help, and I'd want to hear about it too."
  • Watch for: treat any rise in risk as a matter for standard risk assessment and management, not as something about the app.

E5. Integration with therapy

  • Purpose / when to use: to turn implicit training into explicit therapeutic gains.
  • Steps:
    • Najmi and Amir suggest that ABM may make approach behaviour easier for clients who resist exposure. [source: Najmi & Amir 2009]
    • Practical sequence [synthesis]:
      1. Schedule CBM sessions before exposure or behavioural-experiment sessions.
      2. In therapy, ask for real-life ambiguous situations the client noticed and the benign readings that came to mind.
      3. Use those readings as starting hypotheses in cognitive restructuring.
      4. After the block, review symptom and bias measures and decide on boosters or step-up care.
  • Example language (illustrative): "This week, when your colleague didn't reply, did any of the other readings from the training come to mind? Let's test which one fits best."
  • Watch for: the client crediting the app and dropping homework, or the reverse. Keep both going. [synthesis]

5. Handling problems

The primary sources say little about clinical problem-solving. Except where a source is given, everything below is [synthesis].

  • Non-adherence or drop-off (the commonest problem).
    • Sourced responses: scheduled prompts; points, levels and progress displays; daily log review; a phone call after a week without sessions; in-person check-ins on request. [source: NCT03509181; NCT05126862]
    • Add [synthesis]:
      • Tie sessions to an existing routine, such as after breakfast.
      • Agree a minimum viable dose: the 5-minute, 75-trial transition dose in HabitWorks shows that short sessions are acceptable.
      • Review completion at the start of every therapy session, as you would homework.
  • Boredom and "this is pointless" (low credibility).
    • Acknowledge that the task is repetitive by design.
    • Use the physiotherapy analogy and show the client their own pre-training bias score.
    • Agree a time-limited trial, such as the 4-week block, with a review date.
    • Measuring credibility and expectancy early, as HabitWorks did, flags this risk. [source: NCT03509181 for the measurement]
  • Client works out the contingency and "games" it.
    • In ABM and AAT this is expected. Re-instruct the client to respond to the probe or tilt only, at speed.
    • Learning still depends on repeated responding. Whether awareness reduces benefit is [uncertain].
  • Feeling invalidated by "incorrect" feedback (CBM-I).
    • Explain before the first session that the feedback marks the habit being practised, not whether the client's feelings are true.
    • Explore in session any situations where the negative reading was accurate, and do cognitive work on those directly.
    • If shame or self-criticism rises, consider scenario training without explicit error messages, or stop. [synthesis]
  • Distress from threat or alcohol stimuli.
    • Sessions should stop if the client is under undue strain, and adverse events should be reviewed. [source: NCT03898323]
    • Add [synthesis]:
      • Debrief after the session.
      • Normalise mild arousal.
      • For alcohol cues, run a craving-management plan straight after training.
  • Overwhelm or worsening mood or risk during training.
    • Follow the safety-monitoring procedure (E4): alert on a PHQ-9 rise of 5 or more points or any rise in item 9; emergency evaluation on C-SSRS items 4 or 5; safety plan and crisis resources. [source: NCT03509181]
    • Pause CBM while the risk is managed. Restart only once the client is stable. [synthesis]
  • Stuck: dose completed, no change.
    1. Check fidelity: right mode (training, not assessment), right contingency, enough sessions completed (for example, at least 6 of 12 in the imagery trial; [source: Mental Elf summary of Blackwell 2015]), and imagery instructions actually followed.
    2. Check that the bias was present at baseline. A client with no measurable bias may have little to train.
    3. If both are fine, stop and move to, or intensify, established treatment. Developers describe CBM as experimental, and the codebook's known risk is that it delays effective treatment. [source: Najmi & Amir 2009]
  • Resistance: "I want real therapy, not a computer game."
    • Agree with the client. CBM is offered as an add-on.
    • Make human contact around it explicit, and let the client decline without losing access to therapy.
  • Rupture: the client feels palmed off onto an app.
    • Name it and own the referral decision.
    • Return to joint formulation, and offer to drop CBM.
    • Keep therapy sessions focused on the client, not on reviewing the app.
  • Imagery difficulties (C1).
    • For clients who report no images or only observer-perspective images, coach them to imagine from their own point of view with brief, simple practice items.
    • If imagery stays absent, switch to a verbal paradigm (B1 or B2).
  • Scenarios that do not fit the client's life, culture or circumstances.
    • HabitWorks planned personalised stimulus sets, and an open trial examined acceptability with Black and Latinx adults. [source: NCT03509181; search-result title only for the open trial, so its findings are [uncertain]]
    • Choose platforms with large, relevant pools. Ask clients to flag scenarios that do not fit, and discuss those in session.
  • Reading, cognitive or motor barriers.
    • Fragment completion needs literacy. Timed joystick or mouse tasks need motor speed.
    • Use the auditory imagery format, or adapt with the developer's software settings rather than self-made versions.
  • Substance-specific problems (D1).
    • Do not train anyone who is intoxicated or in withdrawal. Require a breath alcohol reading of 0.00 and a withdrawal check first. [source: NCT03898323]
    • If craving rises after sessions, reduce the session frequency and strengthen relapse-prevention supports.
  • OCD and contamination.
    • Use ABM only after exposure and response prevention has been offered. At most, use it as a bridge for clients who refuse exposure. [source: Najmi & Amir 2009]
  • Using unvalidated consumer "brain-training" apps.
    • Advise against them. Use research-platform versions with published parameters (MindTrails, HabitWorks, or trial software).

Sources

Sources:

Further reading (not drawn on): the Blackwell et al. 2015 full text (Southampton eprints, Oxford ORA, Europe PMC); the BMJ Open 2013 imagery CBM trial protocol; Springer Medizin (Williams et al. imagery CBM); the ClinicalTrials.gov record NCT03855488 (imagery-enhanced CBM for worry).

Books and papers cited (not drawn on unless listed above):

  • Amir, N., Beard, C., Taylor, C. T., et al. (2009). Attention training in individuals with generalized social phobia: A randomized controlled trial. Journal of Consulting and Clinical Psychology.
  • Beard, C., & Amir, N. (2008). A multi-session interpretation modification program. Behaviour Research and Therapy.
  • Blackwell, S. E., Browning, M., Mathews, A., Pictet, A., Welch, J., Davies, J., Watson, P., Geddes, J. R., & Holmes, E. A. (2015). Positive imagery-based cognitive bias modification as a web-based treatment tool for depressed adults. Clinical Psychological Science, 3, 91–111.
  • Eberl, C., Wiers, R. W., Pawelczack, S., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2013). Approach bias modification in alcohol dependence. Developmental Cognitive Neuroscience.
  • Grey, S., & Mathews, A. (2000). Effects of training on interpretation of emotional ambiguity. Quarterly Journal of Experimental Psychology.
  • MacLeod, C., Rutherford, E., Campbell, L., Ebsworthy, G., & Holker, L. (2002). Selective attention and emotional vulnerability. Journal of Abnormal Psychology.
  • MacLeod, C., & Mathews, A. (2012). Cognitive bias modification approaches to anxiety. Annual Review of Clinical Psychology.
  • Mathews, A., & Mackintosh, B. (2000). Induced emotional interpretation bias and anxiety. Journal of Abnormal Psychology.
  • Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear of spiders. Journal of Behavior Therapy and Experimental Psychiatry.
  • Salemink, E., & van den Hout, M. (2010). Validation of the "recognition task" used in the training of interpretation biases. Journal of Behavior Therapy and Experimental Psychiatry.
  • Wiers, R. W., Eberl, C., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2011). Retraining automatic action tendencies changes alcoholic patients' approach bias for alcohol and improves treatment outcome. Psychological Science.
← Previous · II-098Competitive memory trainingNext · II-100 →Neuroscience-informed CBT