Process-based therapy

1. Introduction

Version followed: the Hofmann–Hayes process-based therapy (PBT) framework built on the extended evolutionary meta-model (EEMM), as set out in Hayes, Hofmann & Ciarrochi (2020) and Hofmann & Hayes (2019/2024), applied clinically as in the Ong, Hayes & Hofmann (2022) case illustration and Hofmann's workshop materials (Greifswald talk; VGCT 2025 workshop). The menu of change strategies ("kernels") follows the 19 core competencies of Process-Based CBT (Hayes & Hofmann, 2018) as listed on Hofmann's slides. Learning Process-Based Therapy (Hofmann, Hayes & Lorscheid, 2021), the developers' skills manual, is not drawn on here; only a review of it is, so its exact worksheets and step order are [uncertain]. A separate manualised research variant (Stangier, Hofmann et al., 24 sessions, unpublished manual) is noted where it differs.

PBT is a framework, not a single therapy. Its authors call it a "model of models" and say it is not primarily a specific treatment approach [source: Hofmann & Hayes 2024; Ong et al. 2022]. It has no fixed protocol: the clinician builds an individual (idiographic) model of what keeps this person stuck, then picks evidence-based procedures from any tradition that are known to move the processes in that model.

Core idea of change. Problems are treated as maladaptation to context: a network of interacting biopsychosocial processes (thoughts, feelings, attention, sense of self, motives, behaviour, body, social world) that has become self-sustaining in an unhelpful way, "evolution gone awry" [source: Hofmann & Hayes 2024]. Change happens through healthy variation (trying new responses), selection (noticing which responses actually serve the person's goals in this context) and retention (keeping and generalising what works), in a given context. The aim is to disrupt the maladaptive network and grow an adaptive network that sustains itself, with wellbeing (not only symptom reduction) as the outcome [source: Ong et al. 2022; Hofmann Greifswald slides]. The guiding question, adapted from Paul (1969): which core biopsychosocial processes should be targeted with this client, for this goal, in this situation, and how can they be changed most efficiently [source: Hofmann & Hayes 2019, quoted in Hofmann & Hayes 2024 and Praxis article].

2. Framework

The model you assess against: the EEMM grid [source: Hayes, Hofmann & Ciarrochi 2020 abstract; Hofmann Greifswald and VGCT slides].

Rows (psychological dimensions): affect, cognition, attention, self, motivation, overt behaviour.
Extra levels: biophysiological (sleep, diet, exercise, physical health, abilities) and sociocultural (therapeutic relationship, family and friends, social support, culture).
Columns: variation, selection, retention, context. For each cell ask whether the process is healthy or rigid.
Typical maladaptive vs adaptive poles shown on Hofmann's slides (paraphrased):
- Cognition: fusion with thoughts, worry/rumination, jumping to conclusions, rigid negative schemas → defusion, flexible and context-sensitive thinking, reappraisal.
- Affect: suppression and experiential avoidance → acceptance and emotional openness.
- Attention: rigid, threat-biased focus on past or future → flexible, voluntary attention to the present, inside and out.
- Self: rigid conceptualised self, discounting positive self-aspects → perspective-taking self-awareness, self-efficacy, self-compassion.
- Motivation: goal fixation, pleasing others or only looking out for oneself, imposed meaning → chosen values.
- Overt behaviour: impulsivity, inaction, avoidant persistence, short-term gains over long-term competence → committed action, competence.
The slides also list the mediators with best support per row (e.g., cognition: beliefs, defusion, reappraisal; affect: acceptance, anxiety sensitivity, self-compassion; attention: mindfulness, rumination/worry, acting with awareness; self: self-efficacy, self-regulation; motivation: values, intentions, goals; behaviour: coping skills, activation, avoidance; biophysiological: diet, exercise; sociocultural: parenting, social support, alliance).

Phases. PBT does not prescribe a session count. The phases below are drawn from the published case (Amy: 18 sessions, GAD) and the research protocol; treat numbers as examples, not rules [source: Ong et al. 2022; Stangier et al. trial protocol NCT06530888] [synthesis for the phase labels].

Phase	Goals	Typical length (from published examples)	Move on when
A. Engagement and process assessment	Clarify problems and life goals; functional analysis; build a hypothesised network with the client; start idiographic monitoring	Sessions 1–3 (Amy); sessions 1–5 baseline in the Stangier protocol	Client agrees the network fits (Amy's was confirmed in session 3); a small set of target nodes and monitoring items is chosen
B. Perturbation (variation)	Apply chosen kernels to the target processes; client tries new responses as homework	From session 3–4 on (Amy 4–7); sessions 6–24 in Stangier protocol	Client can produce alternative responses in problem contexts
C. Selection	Track consequences of new responses; keep those that serve goals; probe and redraw the network; swap strategies that do not move the target	Interleaved with B	Adaptive nodes appear and start reinforcing each other (Amy: "empowered" added as a node)
D. Retention and wellbeing	Consolidate, generalise to new contexts, add wellbeing targets (balance, connection, hobbies), test resilience against new stressors	Amy from session 8; sessions 16–18 focused on generalising and sustaining	Adaptive network holds under stress; monitoring stable; client can run the vary–select–retain cycle alone

Format. Individual, weekly in both published examples. Network review with the client is a standing agenda item; homework is tied to functional goals; consequences are tracked [source: Ong et al. 2022]. Monitoring options: brief experience-sampling items (Amy: four times daily about the past three hours, later reduced; Stangier: daily ratings in problem situations via an app until ~100 data points), the Process-Based Assessment Tool (PBAT) at intake, mid-point, end and follow-up, plus progress items [source: Ong et al. 2022; VGCT slides; Stangier protocol; PBAT abstract]. Statistical network estimation (GIMME, graphical VAR) was used in research settings; it is optional in routine practice [synthesis].

Criteria for revising rather than moving on: the empirical or observed pattern disagrees with the hypothesised network; context changes (new stressor, life decision); progress stalls [source: Ong et al. 2022; VGCT slides].

3. Guiding principles

[source: Ong et al. 2022; Hofmann & Hayes 2024; Praxis article by Hofmann & Hayes; Hofmann slides] unless marked.

Processes over syndromes. Diagnosis is a starting point only; plan treatment around what maintains the problem for this person. Practical meaning: ask what role a symptom plays, not which disorder it belongs to.
Function over form (topography). Define nodes by what they do, not what they look like. Meaning: the same behaviour (e.g., checking email) can be a problem at one time and a valued action later; do not count behaviours blindly.
Idiographic first, idionomic generalisation. Group findings are used only if they fit the individual. Meaning: a technique's trial evidence earns it a place on the shortlist; this client's data decide whether it stays.
Techniques are secondary to targeted processes. Meaning: you may use methods from any orientation (CBT, ACT, DBT, humanistic, analytic) if they move the targeted process; no protocol loyalty.
Collaboration and client language. Each node and arrow is agreed with the client and named in their words. Meaning: never add a node the client does not recognise.
As complex as necessary, as simple as possible. Meaning: add a node only if it links back to the network; prune ones that do not change decisions.
Evolutionary logic: vary, select, retain, in context. Meaning: variation alone is not enough; after a new response, always track consequences and decide what to keep.
Any process can help or harm depending on context and history. Meaning: do not label a process (e.g., problem solving, planning) as bad in general; judge it by its effect for this client now.
Multidimensional and multilevel. Meaning: audit all six dimensions plus body and social world; check whether you habitually neglect some.
Hypothesis-driven and iterative. Meaning: the network is a working hypothesis; test it with monitoring and observation and redraw it when data or context change.
Empirical data complement clinical judgement, not replace it. Meaning: treatment success does not prove your model was right; check discrepancies.
Wellbeing as the outcome. Meaning: targets include positive affect, connection, purpose and quality of life, not only less distress.
Aim for a self-sustaining adaptive network. Meaning: you are done when good processes reinforce each other without you.
Bridge schools, avoid divides. Meaning: describe your work in process terms that colleagues of other orientations can share.
PBT is a skill to iterate, not a protocol to execute perfectly [source: Ong et al. 2022]. Meaning: accept rough first networks and improve them.

4. Intervention inventory

Two layers. Part A is PBT's own clinical procedures (assessment, network building, target choice, monitoring, revision). Part B is the menu of evidence-based change strategies ("kernels") that PBT draws on; the 19 listed are those on Hofmann's slides, matching the chapters of Process-Based CBT (2018). PBT does not own these kernels and its sources here do not give step-by-step instructions for them, so Part B steps are written from each method's standard originating procedure and are marked [synthesis] for the step detail. Book chapter authors are named from the review of Process-Based CBT [source: Metapsychology review].

Part A. PBT core procedures

A1. Process-focused intake: problems, goals and the core question

Purpose / when: First 1–2 sessions; replaces or supplements a diagnosis-led intake [source: Ong et al. 2022; VGCT slides].
Steps:
1. Collect presenting problems in concrete, situational terms (when, where, with whom, what happens next).
2. Ask for life goals and what a better life would look like, including wellbeing goals, not only symptom relief (Amy's goals: clarify values, more physical activity, plan a career change, be more present with others, set healthy boundaries).
3. Record a diagnosis if your setting needs one, but note it as context, not the plan.
4. Keep the core question in view: which processes, for this person, this goal, this situation, changed how most efficiently.
5. Screen for risk, substances, medical issues and anything needing a crisis plan first (see B19) [synthesis].
Example language (illustrative): "Rather than start with a label, I'd like to understand what keeps this going for you, and what you'd want your life to be about if it loosened its grip."
Watch for: goals phrased only as "stop feeling X" (help add approach goals); vague problems that cannot be placed in situations.

A2. Functional analysis interview

Purpose / when: To find what maintains the problem and what would change it; sessions 1–3 and whenever the network is revised [source: VGCT slides; Ong et al. 2022].
Steps:
1. Pick a recent, specific episode of the main problem.
2. Explore antecedents (situation, body state, thoughts, feelings, where attention went) and consequences (short-term relief, long-term costs), using Socratic questions.
3. Ask the workshop's guiding questions: what is the primary problem; what are the obvious mediators; what maintains it; how could it be solved; which dimensions and levels show unhealthy variation, selection or retention.
4. For each candidate factor, ask about its function: what does doing this get you or spare you, right now and over time?
5. Look for repetition across episodes; these become candidate nodes.
Example language (illustrative): "Just before you started checking, what was going on inside you?" / "And right after, what eased? What did it cost you later that week?"
Watch for: mixing up form and function; you know it is working when the client starts predicting their own chain.

A3. Collaborative network building (process-based case conceptualisation)

Purpose / when: The central PBT tool; built in sessions 1–3 and redrawn through treatment [source: Ong et al. 2022; Hofmann slides].
Steps:
1. On paper or a whiteboard, write each candidate factor as a node, using the client's words and functional definitions (Amy's nodes included a self-label of being "selfish", pressure to be responsible, attention biased to what might go wrong, excessive problem solving).
2. Draw arrows for hypothesised influence; agree each node and each arrow with the client before adding it.
3. Use the conventions from the published case: arrow style shows whether an influence increases (excitatory) or decreases (inhibitory) the target; arrowhead size shows hypothesised strength; rectangles mark changeable variables; rounded shapes mark fixed moderators such as past events.
4. Add a new node only if it links back into the network (as complex as necessary, as simple as possible).
5. Mark feedback loops: self-amplifying loops that keep running without outside input, and loops where short-term relief maintains long-term costs (Bill's example: thought of being socially awkward → avoidance → feeling bad → more avoidance).
6. Tag each node with its EEMM dimension/level to check coverage (see A4).
7. Confirm the whole picture with the client; Amy's was confirmed in session 3.
Example language (illustrative): "Does this arrow feel right, that the pressure to be responsible pushes the problem-solving? Or is it the other way round?"
Watch for: networks that grow past what you can use (aim for a handful of key nodes per phase; Amy's monitored network used about five) [source: Ong et al. 2022]; the client correcting your arrows is a good sign of engagement.

A4. EEMM audit (dimension × variation/selection/retention/context)

Purpose / when: After a first network sketch, to catch neglected dimensions or levels and name what is rigid [source: Hofmann slides; Learning PBT review mentions worksheets for auditing one's usual interventions — [uncertain] for format].
Steps:
1. Make a grid: six dimensions plus biophysiological and sociocultural levels down the side; variation, selection, retention, context across the top.
2. For each row, ask: does the client have a range of responses here (variation)? Do they pick responses by what works for their goals or by avoidance, rules or others' approval (selection)? Do helpful responses stick (retention)? Do they read the situation (context sensitivity)?
3. Fill in examples from the functional analysis; leave empty cells empty.
4. Add any missing important process to the network.
5. Separately, audit yourself: which rows do your usual interventions ignore? [source: Learning PBT review]
Example language (illustrative): "When you feel this, how many different ways do you have of responding? Which one usually wins?"
Watch for: filling cells to complete the grid rather than from evidence.

A5. Choosing targets

Purpose / when: Once the network is agreed [source: Ong et al. 2022; Stangier protocol].
Steps:
1. Look for the most connected (central) node, strong arrows, self-loops and positive feedback loops.
2. Prefer changeable nodes (rectangles) whose change would weaken a loop or strengthen an adaptive one.
3. Also name the adaptive counterpart of each target (e.g., flexibility about standards as the counterpart of rigid responsibility standards); in the Stangier variant each dimension is rated on a bipolar maladaptive–adaptive scale.
4. Agree with the client one or two targets for the next phase.
Example language (illustrative): "If we could loosen just one link in this circle, which one do you think would make the others ease?"
Watch for: choosing the node that is most distressing rather than most influential.

A6. Selecting kernels (hypothesis-driven intervention choice)

Purpose / when: After targets are set, and again whenever a strategy fails [source: Ong et al. 2022; VGCT slides; Praxis article].
Steps:
1. List kernels with evidence for moving this process (see Part B and the mediator list in Section 2).
2. Choose by the client's goals and context, not by diagnosis (Bill's options for performance anxiety: exposure, social-skills training, attention training, relaxation, depending on whether distress tolerance, skills, self-focused attention or arousal was the target).
3. State the hypothesis: "If we use X on node Y, we expect Y and its downstream nodes to change."
4. Pick a measure to test it (A8).
Example language (illustrative): "My guess is that if your attention can come off the threat a bit, the worry loop loses fuel. Let's test that for two weeks."
Watch for: eclectic drift (techniques without a target).

A7. Perturb – probe – redraw cycle (variation, selection, retention in practice)

Purpose / when: The treatment engine in phases B–D [source: VGCT slides; Ong et al. 2022].
Steps:
1. Perturb: deliver the chosen kernel in session and set homework that asks for new responses in real contexts (Amy: list 5–10 personal standards and track how they steer her; then practise behaviours that serve wellbeing).
2. Select: have the client track consequences of the new responses; variation alone is not enough.
3. Probe: review monitoring and the week's episodes; which links weakened, which new nodes appeared?
4. Redraw the network with the client; add emerging adaptive nodes (Amy's sense of being "empowered").
5. If the target did not move, choose an alternative strategy and repeat.
6. Retain: once a response works, plan where else it can be used and how to keep it going.
Example language (illustrative): "You tried saying no twice this week. What happened next, and is that something worth keeping?"
Watch for: adding variation without tracking outcomes; good sign is the client spontaneously reporting effects of new responses.

A8. Idiographic process monitoring

Purpose / when: To test the network and track change [source: Ong et al. 2022; VGCT slides; Stangier protocol; PBAT abstract].
Steps:
1. Write a few short items for target nodes, adaptive counterparts, behavioural goals and progress (Amy's later items covered empowerment, acting for wellbeing, flexibility about standards, awareness of feelings, connection).
2. Choose a schedule matched to how fast the process changes and to the client's tolerance (daily ratings or a few prompts a day; pilot first).
3. Optionally add the PBAT (an EEMM-based process questionnaire; describe and use under its licence, do not reproduce items) at start, mid-point, end and follow-up.
4. Review graphs in session; compare with the hypothesised network. Statistical network modelling needs roughly 60–100 data points per person and is mainly a research tool [source: Ong et al. 2022].
5. Revise items when the network changes.
Example language (illustrative): "These three questions on your phone will tell us whether what we're trying is moving the things we think matter."
Watch for: burden and missing data (see Section 5); single items can be unreliable.

A9. Network revision

Purpose / when: When data contradict the model, progress stalls, or context changes [source: Ong et al. 2022; VGCT slides].
Steps: review discrepancies (Amy's model missed that feeling at peace dampened responsibility pressure, and had one arrow reversed); ask the client about new stressors or decisions; redraw; re-select targets and kernels; update monitoring items.
Example language (illustrative): "The ratings suggest something we missed: on calmer days the pressure is lower. Shall we add that?"
Watch for: defending your first model.

A10. Retention, generalisation and wellbeing building

Purpose / when: Final phase (Amy from session 8, generalising in 16–18) [source: Ong et al. 2022].
Steps:
1. Shift targets to wellbeing processes (detachment from thoughts, balance, hobbies, connection).
2. Review which strategies worked and in which contexts; plan other contexts to use them.
3. Teach the cycle explicitly so the client can run it: try options, keep what works, repeat when context changes [source: VGCT "adaptation algorithm"].
4. Use real stressors during this phase as tests of the network's resilience.
5. Agree what signs would mean booster sessions [synthesis].
Example language (illustrative): "When life shifts again, what's your own version of try, notice, keep?"
Watch for: new over-commitment patterns (Amy began over-committing to values; attention and selective value choice were revisited).

Part B. Kernels (evidence-based change strategies)

List and process mapping: [source: Hofmann Greifswald and VGCT slides; Metapsychology review of Process-Based CBT 2018 for chapter authors]. Step detail in each entry: [synthesis] from the standard procedure of the method's originating tradition; the 2018 chapters themselves are not drawn on [uncertain for chapter-specific variants]. In PBT, each kernel is used only after A5–A6 link it to a target node, and its effect is checked with A8.

B1. Contingency management

Purpose / when: Overt behaviour, motivation. When a target behaviour is maintained or blocked by its consequences (e.g., substance use, attendance, avoidance) [chapter: Higgins, Kurti & Keith].
Steps: define the target behaviour objectively; set a verification method; agree a reinforcer the client values and a clear schedule (reinforce promptly and every time at first); escalate for continuous success, reset after lapses only as agreed; thin the schedule and shift to natural reinforcers as the behaviour stabilises.
Example language (illustrative): "Each day you log the walk, you earn a point toward the thing you picked. Let's make the rule simple enough that there's no argument about it."
Watch for: reinforcers that lose value; rules perceived as control; working when target frequency rises and the client links it to natural payoffs.

B2. Stimulus control

Purpose / when: Overt behaviour, context sensitivity. When cues in the environment trigger or fail to trigger behaviour (e.g., sleep, eating patterns, checking, procrastination) [chapter: McIlvane].
Steps: map cues that precede the target behaviour (from A2); remove or limit triggering cues; build cues for the desired behaviour (place, time, prompts); restrict the problem behaviour to narrow contexts where useful (e.g., bed only for sleep); review and adjust.
Example language (illustrative): "Where and when does it start? Let's change the setting so the old habit has fewer doors in."
Watch for: over-restriction that becomes avoidance; working when cue–behaviour links weaken in monitoring.

B3. Shaping

Purpose / when: Overt behaviour. When the full target is too large to do now [chapter: Miltenberger et al.].
Steps: define the end goal and the current starting level; break it into small successive approximations; reinforce each step (praise, natural reward, self-reinforcement) until reliable; move up one step at a time; drop back a step after failure.
Example language (illustrative): "We'll start one notch above what you can already do, and only move up when it feels routine."
Watch for: steps too big (repeated failure) or too small (boredom).

B4. Self-management

Purpose / when: Self, overt behaviour, retention. To transfer control to the client between sessions.
Steps: client sets a specific goal; self-monitors the behaviour; arranges cues and consequences for themselves; self-evaluates against the goal; self-reinforces; reviews and revises weekly with you, then alone.
Example language (illustrative): "You're becoming your own coach: track it, set the stage, and give yourself credit when you follow through."
Watch for: harsh self-evaluation (consider B14/B17 for self-compassion and acceptance); working when the client runs the loop without prompts.

B5. Arousal reduction

Purpose / when: Affect, biophysiological level. For high physiological arousal that drives other nodes (Bill's performance arousal) [chapter: McKay — breathing, relaxation, mindfulness, visualisation, autogenics].
Steps: rate arousal; teach one method at a time (slow diaphragmatic breathing; progressive muscle relaxation, tense–release then release only; calming imagery; autogenic phrases of warmth and heaviness); practise daily when calm, then in mild stress, then in target situations.
Example language (illustrative): "Let's practise this when you're already fairly calm so it's ready when you need it."
Watch for: relaxation used as avoidance or safety behaviour during exposure; relaxation-induced anxiety in some clients; working when arousal ratings fall and other nodes follow.

B6. Coping and emotion regulation

Purpose / when: Affect. When emotions are managed by suppression, avoidance or rumination [chapter: Aldao & Plate — reappraisal and acceptance].
Steps: map the emotion episode (situation, attention, appraisal, response); widen the repertoire of strategies (situation selection/modification, attention shifting, reappraisal, acceptance, problem solving, expressive behaviour); help the client match strategy to context; track which works where (selection).
Example language (illustrative): "No strategy is good everywhere. Let's find out which ones fit which moments for you."
Watch for: rigid use of one strategy (the PBT problem is inflexibility, not a missing technique).

B7. Problem solving

Purpose / when: Cognition, overt behaviour. For real-world problems that maintain distress [chapter: Nezu et al., problem-solving therapy].
Steps: build a hopeful, approach orientation to problems; define the problem concretely with realistic goals; brainstorm many options without judging; weigh pros and cons and choose; plan and carry out; evaluate and revise.
Example language (illustrative): "Let's list every option, even silly ones, before we pick."
Watch for: in PBT, problem solving can be a maladaptive node (Amy's excessive problem solving fed worry); check its function before prescribing more of it.

B8. Exposure

Purpose / when: Affect, overt behaviour, attention. When avoidance of feared situations, sensations, memories or thoughts maintains the network [chapter: Davies & Craske].
Consent and safety first: explain the rationale and that anxiety will rise; obtain informed consent and agree the client can pause; screen for medical contraindications to interoceptive or intense exercises (cardiac or respiratory conditions, pregnancy, seizure disorders — check with physician), current crisis, active suicidality, unstable dissociation or substance intoxication; for trauma memories use a trauma-specific protocol and training [synthesis].
Steps: list avoided situations or sensations and safety behaviours; build a flexible hierarchy; state what the client predicts will happen; enter the situation and stay long enough to test the prediction, dropping safety behaviours; afterwards compare prediction with outcome; vary contexts and stimuli to strengthen generalisation; repeat between sessions.
Example language (illustrative): "What do you expect will happen if you stay? Let's find out, and you can stop if you need to."
Watch for: subtle avoidance and safety behaviours, overwhelm (step back, do not end abruptly unless safety requires), dissociation; working when predictions shift and approach widens.

B9. Behavioural activation

Purpose / when: Overt behaviour, motivation, affect. For withdrawal, inactivity and low reward [chapter: Martell — values, reinforcement, activity monitoring].
Steps: monitor activity and mood; identify avoidance patterns and their function; schedule activities linked to values and mastery/pleasure; grade them; problem-solve barriers; review outcome versus mood.
Example language (illustrative): "Let's act from the plan, not from the mood, and see what mood does afterwards."
Watch for: activity chosen to please others (maladaptive selection); working when activity and reward rise together.

B10. Interpersonal skills training

Purpose / when: Sociocultural level, overt behaviour. When skill deficits or rigid interpersonal patterns maintain the network (Amy's asking for help and saying no; Bill's social skills) [chapter: Mueser; Ong et al. 2022 cites DBT interpersonal skills as an example kernel].
Steps: pick a specific situation; identify the goal (objective, relationship, self-respect); model the skill; client role-plays; give specific feedback; repeat; set real-life practice; review.
Example language (illustrative): "Let's try the conversation here first. I'll play your manager."
Watch for: assertiveness rules applied rigidly; cultural norms about directness [synthesis].

B11. Cognitive reappraisal

Purpose / when: Cognition. When specific appraisals drive affect and behaviour and the client can step back enough to examine them [chapter: Wenzel — thought records, coping cards].
Steps: catch the thought in a specific situation; rate belief and emotion; examine evidence for and against and alternative views; form a balanced alternative; re-rate; write a coping card for recurring situations; test predictions behaviourally.
Example language (illustrative): "If a friend had this thought, what else might you point out to them?"
Watch for: reappraisal turning into reassurance-seeking or arguing with worry; if thoughts are sticky and repetitive, defusion (B13) may fit better.

B12. Modifying core beliefs

Purpose / when: Cognition, self. When broad negative schemas about self, others or the world anchor the network (Amy's "selfish" self-label) [chapter: Arntz — schemas; methods listed are standard schema-focused methods, chapter contents [uncertain]].
Steps: identify the core belief through recurring themes (e.g., downward arrow); trace its origins; collect contrary evidence over time (positive data log); use continuum ratings to soften all-or-nothing beliefs; for emotionally held beliefs consider experiential methods such as imagery rescripting (requires specific training); build and rehearse a new, more balanced belief.
Example language (illustrative): "Where on a line from 'utterly selfish' to 'saintly' would you put yourself, and the people you admire?"
Watch for: intellectual agreement without emotional shift; distress when revisiting childhood material (pace, ground).

B13. Cognitive defusion

Purpose / when: Cognition. When the client is fused with thoughts, rules or self-labels (Amy used defusion with her standards) [chapter: Blackledge; Ong et al. 2022].
Steps: notice and name the thought as a thought; use distancing methods (prefacing with "I'm having the thought that…", saying it slowly or in a silly voice, repeating a word until it loses meaning, labelling recurring stories); ask whether following the thought is workable for the client's goals; practise in triggering situations.
Example language (illustrative): "Notice the mind saying 'I must be responsible for everything'. Is following that helping you right now?"
Watch for: defusion used to get rid of thoughts (that is avoidance again); mockery that feels invalidating.

B14. Psychological (experiential) acceptance

Purpose / when: Affect. When suppression and experiential avoidance maintain the network [chapter: Forsyth & Ritzert].
Steps: review the costs of control attempts (creative hopelessness); invite willingness as an alternative; practise making room for a feeling (locate it in the body, breathe around it, describe it, allow it); link willingness to valued action.
Example language (illustrative): "What if, just for this minute, the task isn't to make it go away but to let it be here while you do what matters?"
Watch for: acceptance heard as resignation or tolerating abuse; working when avoidance falls while distress is present.

B15. Attentional training

Purpose / when: Attention. For rigid, threat-biased or self-focused attention (Amy's focus on what might go wrong; Bill's self-focus) [source: listed on Hofmann slides; Amy's case reviewed attentional control in session 12; standard method: Wells' attention training technique, [synthesis] for steps].
Steps: explain the goal as flexible control of attention, not distraction; practise selective attention to several external sounds or objects, then rapid switching, then divided attention, in short daily sessions; transfer by redirecting attention outward in target situations.
Example language (illustrative): "We're training the muscle of choosing where attention goes, not trying to stop thoughts."
Watch for: use as distraction or safety behaviour.

B16. Values choice and clarification

Purpose / when: Motivation. When behaviour is selected by avoidance, others' approval or rigid goals rather than chosen values (Amy's goal of clarifying values) [chapter: Lundgren & Larsson].
Steps: explore what matters in life domains (relationships, work, health, growth, leisure); separate values (ongoing directions) from goals (achievable); rate importance and current consistency; turn values into specific next actions; revisit regularly.
Example language (illustrative): "If no one would ever know your choice, what would you want this part of your life to be about?"
Watch for: values written as "shoulds"; over-commitment to too many values (Amy, session 12: choose selectively) [source: Ong et al. 2022].

B17. Mindfulness practice

Purpose / when: Attention, affect, self. To build present-moment, non-judging awareness (Amy reviewed mindfulness in session 12) [chapter: Baer].
Steps: start with brief guided practices (breath, body scan, sounds); add informal practice in daily activities; debrief by asking what was noticed; link to noticing early signs of loops in the network; build a regular routine.
Example language (illustrative): "Each time you notice the mind has wandered, that noticing is the practice."
Watch for: trauma-related distress or dissociation during inward focus (shorten, keep eyes open, use external anchors) [synthesis].

B18. Enhancing motivation

Purpose / when: Motivation. For ambivalence about change or about the treatment plan [chapter: MacKillop et al.; standard method: motivational interviewing, [synthesis] for steps].
Steps: ask open questions; reflect; explore both sides of ambivalence; elicit the client's own reasons for change; explore discrepancy between values and current behaviour; support autonomy; ask for commitment to a small step.
Example language (illustrative): "Part of you wants to cut down the checking, and part of you worries what happens if you do. Tell me about both."
Watch for: pushing (arguing for change provokes argument against it).

B19. Crisis management and suicidality

Purpose / when: Any time risk rises; takes priority over other targets [chapter: Comtois & Landes].
Steps (current standard practice, [synthesis]): ask directly about suicidal thoughts, intent, plans, preparation, past attempts and protective factors; assess risk collaboratively; build a written safety plan (warning signs, own coping, people and places for distraction, people to ask for help, professional and crisis contacts); lethal-means counselling focused on secure storage or removal of means, involving family where possible; agree follow-up and contact; involve team or supervisor and document; treat suicidal drivers as nodes in the network once safe.
Example language (illustrative): "I'd like to ask you directly: have you had thoughts of ending your life?"
Watch for: escalating risk (arrange urgent care); a team approach is recommended because no formula predicts individual risk [source: Metapsychology review].

B20. Biophysiological- and sociocultural-level interventions

Purpose / when: The EEMM extends targets beyond psychological dimensions: sleep, diet, exercise, physical health, social support, family and culture, therapeutic alliance [source: Hofmann slides; Learning PBT review, chapter 6]. Specific procedures are not given in the sources [uncertain]; steps below are [synthesis].
Steps: include body and social nodes in the network; target them with kernels already listed (e.g., stimulus control for sleep, activation for movement, interpersonal skills for support) or refer (GP, dietitian, sleep service, community resources); involve significant others when relationships are hubs; treat the alliance itself as a node to monitor.
Example language (illustrative): "Your sleep sits right in the middle of this map. Shall we work on it directly?"
Watch for: no specific weight, calorie or exercise targets; refer for medical issues.

5. Handling problems

The PBT sources discuss a few practical difficulties (measurement burden, function changes, model errors, new stressors, over-commitment). Everything else here is [synthesis], applying PBT's own logic: treat the difficulty as a process to place in the network.

Stuck client / progress stalls. Revise the plan; do not just repeat the kernel [source: Ong et al. 2022]. Response: (1) check the target was the most connected node, not the most distressing; (2) check function vs form (Amy's email checking stayed level but had changed function to running a business); (3) look at unaudited dimensions or levels (A4); (4) try an alternative kernel for the same process (perturb – probe – redraw) [source: VGCT slides]; (5) remember change may be slow then suddenly rapid [source: Learning PBT review; VGCT slides on sudden gains].
Hypothesised network is wrong. Expected; Amy's model missed an inhibitory link and reversed an arrow [source: Ong et al. 2022]. Response: show the client the discrepancy, redraw together, update items. Good outcomes do not prove the model was right.
Monitoring burden and missing data. Four daily prompts became stressful for Amy and data went missing [source: Ong et al. 2022]. Response: pilot early; cut to the fewest items that matter; reduce frequency; match timescale to expected change; accept that statistical network models need many data points and are optional; consider passive data where available.
New stressors or big life decisions reactivating old loops. Amy's job change and surgery temporarily revived worry [source: Ong et al. 2022]. Response: name it as a context change, check which old links reactivated, use it as a retention test rather than a failure.
Over-correction (e.g., over-committing to values, too much variation). Amy began over-committing to values and lost presence [source: Ong et al. 2022]. Response: return to attention and mindfulness; choose values selectively; remind that the goal is flexibility, not more of the new behaviour everywhere.
Resistance or disagreement with the model [synthesis]. Response: treat it as data; ask which node or arrow feels wrong and redraw; if ambivalence about change, use B18; never keep a node the client rejects.
Overwhelm by the network or the session [synthesis]. Response: show only the two or three nodes in today's focus; use the client's words; if emotional overwhelm, ground (external attention, slow breathing), lower exposure intensity, and return to the plan when regulated.
Homework non-adherence [synthesis]. Response: run a mini functional analysis of the missed task (what got in the way, what did not doing it give relief from); shrink the task (B3), add cues (B2), link it to a value (B16); check whether homework was set for variation without a way to notice consequences.
Alliance ruptures [synthesis; the alliance is listed as a sociocultural-level process on Hofmann's slides]. Response: name the strain openly, invite the client's view, own your part, and adjust; add the alliance as a node if ruptures recur; repair is itself a chance for the client to vary an interpersonal pattern.
Client wants a diagnosis or a fixed protocol [synthesis]. Response: give a diagnosis where needed and explain that the plan targets what maintains the problem; offer a clear session structure (network review, kernel practice, homework, monitoring) so the flexibility does not feel like drift.
Therapist drift into unprincipled eclecticism [synthesis]. Response: before each new technique, state the target node and expected effect, and choose a way to check it. The Stangier trial uses a decision-making questionnaire to check process-based decisions [source: Stangier protocol].
Kernel from outside your training (e.g., imagery rescripting, trauma exposure, contingency management schemes) [synthesis]. Response: get training or supervision, or refer; PBT does not license untrained use.
Rising risk or crisis. Pause other targets and use B19; in the research protocol, rising suicidality led to review of continuation [source: Stangier protocol]. Note that the trial excluded high suicidality, substance dependence, psychosis, bipolar and eating disorders, so experience of PBT in those groups is thin [uncertain].
Data-literacy or tool limits [synthesis]. Response: a hand-drawn network and a few weekly ratings are enough to run A1–A10; statistical modelling is optional.

Sources

Sources: