Cognitive therapy for social anxiety disorder

1. Introduction

Version followed: Oxford Centre for Anxiety Disorders and Trauma (OxCADAT) individual cognitive therapy for social anxiety disorder (CT-SAD), Clark and Wells (1995) model, as described in the NICE CG159 specification (up to 14 sessions of 90 minutes) and the Oxford group's practice papers (Warnock-Parkes et al. 2017 on video feedback; Warnock-Parkes et al. 2020 on full-protocol remote delivery). Adolescent variant (CT-SAD-A) noted where relevant. The full treatment manual (supplied at OxCADAT workshops) and the original 1995 chapter are not drawn on; content attributed to them is taken from the papers and marked accordingly.

CT-SAD is an individual, present-focused cognitive therapy. Its premise is that social anxiety persists, despite repeated exposure to social situations, because of a small set of maintaining processes: when a feared situation is entered, the person turns attention inward, builds a distorted, usually observer-perspective image of how they appear (a negative "felt sense" taken as what others see), and uses safety behaviours to prevent feared outcomes. Before and after events, anticipatory worry and post-event "post-mortems" add further negative material. These processes stop the person noticing disconfirming evidence, can make feared outcomes more likely, and let them credit any good outcome to the safety behaviour. Change happens experientially: the client discovers in vivo, through experiments, video feedback and externally focused attention, that their self-image is inaccurate, that their feelings are not a reliable guide to how they come across, and that dropping safety behaviours and even showing feared signs does not lead to the feared social catastrophe. [source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020; Oxford neuroscience overview; Clark & Wells 1995 as summarised in James 2005]

The three classes of belief the model targets [source: Clark & Wells 1995 as summarised in James 2005]:

Excessively high standards for social performance (e.g. a rule that one must always be amusing).
Conditional beliefs about evaluation (e.g. if I look shy, people will think I am an idiot).
Unconditional negative beliefs about the self (e.g. I am uninteresting, I am unlikeable).

2. Framework

Format [source: NICE CG159 rec 1.3.13; Warnock-Parkes et al. 2020; Leigh et al. 2021]:

Individual, weekly, up to 14 sessions over about 3 to 4 months; sessions up to 90 minutes, because most sessions contain live behavioural experiments. Longer sessions are also recommended remotely when experiments are run.
Booster sessions after the main course are part of the Oxford programme (7 of 12 adolescents in the CT-SAD-A case series had at least one). [uncertain] The exact booster schedule (commonly cited as up to 3 monthly boosters in the Clark et al. 2006 trial) is not confirmed.
Adolescents (CT-SAD-A): 14 weekly 90-minute individual sessions, same core components, plus optional work on parental beliefs and behaviours and on bullying. [source: Leigh et al. 2021] Note that NICE's generic recommendation for children and young people (8 to 12 sessions of 45 minutes, with social skills training and parent sessions) is a different, non-CT-SAD specification. [source: NICE CG159 rec 1.5.4]
Remote: all components can be delivered by video call; telephone-only delivery is discouraged, especially early, because it can feed avoidance of being seen. [source: Warnock-Parkes et al. 2020]

CT-SAD is not formally phased; the sources give a typical order of components. The stage labels below are an organising device [synthesis]; the component order and session numbers are from the sources [source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020].

Stage A: Assessment and set-up (pre-treatment).

Goals: confirm SAD, collect baseline symptom and process measures, obtain written consent for routine recording of sessions.
Contents: SPIN or LSAS for severity; Social Cognitions Questionnaire (SCQ) and Social Phobia Weekly Summary Scale (SPWSS) as process measures; Social Behaviours Questionnaire (SBQ) and Social Attitudes Questionnaire (SAQ) at baseline, mid and end. Remote: test call and privacy set-up.
Move on when: diagnosis and consent are in place.

Stage B: Engagement and formulation (session 1).

Goals: specific therapy goals; individualised version of the Clark and Wells model built from a recent episode.
Move on when: the client can see how their own self-focus, self-image and safety behaviours link to their anxiety and feared predictions.

Stage C: Experiential demonstration (sessions 2 to about 4).

Session 2: self-focused attention and safety behaviours experiment (two recorded conversations).
Session 3: video (and still photo) feedback of the session 2 conversations.
Next session: externally focused attention training.
Move on when: the client has seen the gap between how they felt they looked and how they appeared, and can shift attention outward well enough to run experiments properly.

Stage D: Behavioural experiments (most sessions from about session 4).

Goals: test key predictions in and out of the office, first by dropping safety behaviours and focusing outward, later by deliberately showing feared signs (decatastrophising); surveys, virtual audiences and online experiments as needed; homework experiments linked to each in-session one.
Move on when: SCQ belief ratings and SPWSS scores fall and the client can design and run experiments themselves. [synthesis for the hand-over criterion]

Stage E: Targeted modules as indicated (usually mid to late therapy).

Anticipatory worry and post-event rumination: added if SPWSS rumination items are not improving.
Memory work (discrimination training, imagery rescripting): when negative self-images persist and link to earlier social trauma, usually later in therapy.
Unconditional beliefs and self-criticism: targeted from session 2 within experiments; extra methods if beliefs persist.

Stage F: Ending (final sessions) and boosters.

Therapy blueprint written collaboratively: key learning, plan to build on progress, setback plan. Boosters review the blueprint and continuing experiments.

3. Guiding principles

Sources: Warnock-Parkes et al. 2017 and 2020, NICE CG159; one-line practical meanings are the guide's wording.

Treat the maintaining processes, not only the situations. Each session should target self-focus, self-imagery, safety behaviours, or pre- and post-event processing, not just situation exposure.
The individualised model drives every decision. Choose each experiment from the model and the latest SCQ and SPWSS: which belief, which safety behaviour, which image?
Learning is experiential, in vivo and in session. Most sessions contain a live experiment, with the therapist present; long sessions make that possible. [source: NICE CG159; Leigh et al. 2021]
Drop safety behaviours and focus outward every time. An experiment run with safety behaviours or self-focus intact tests nothing; build both instructions into every task.
Feelings are not a reliable guide to appearance. Repeatedly separate "how I felt" from "how I came across" using video, photos and others' feedback.
Make predictions specific and measurable. Every prediction gets an observable form and a 0 to 100 rating, so it can be checked.
Generalise after every experiment. Always ask what the result means for the belief, the self-image and the unconditional self-view.
Decatastrophise, not just disconfirm. Later experiments let feared signs show, to learn that visible anxiety and imperfection are acceptable and human.
Make recording ordinary. Consent to routine recording at assessment, so video is part of therapy rather than a special ordeal.
The therapist may be a "phobic object". [source: Warnock-Parkes et al. 2020, short quote] Moderate eye contact early, sit or position the camera to reduce scrutiny, and ask for feedback often.
Use process measures every session. SPWSS and SCQ shape the agenda; lack of change on specific items triggers specific modules.
Homework continues the in-session experiment. Each in-session experiment has a linked homework experiment. [source: NICE CG159]
Compassionate, non-scrutinising stance toward the self. Invite the client to watch and judge themselves as they would a friend.
Collaborative guided discovery. The therapist asks; the client draws conclusions from data they generated. [source: Warnock-Parkes et al. 2017 on Socratic discussion of video]
Present focus, with the past only where it maintains the present. Memory work is used when early social trauma keeps a current self-image alive. [source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020]

4. Intervention inventory

All example therapist language below is illustrative wording written for this guide unless shown in quotation marks with a citation. The OxCADAT forms, measures and record sheets are licensed materials and are described, not reproduced.

4.1 Assessment, routine recording consent and process measures

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020]

Purpose / when: Before session 1 and throughout. Establishes severity, identifies the client's idiosyncratic fears and safety behaviours, and normalises recording.
Steps:
1. Give a severity measure (SPIN or LSAS) at assessment and regularly during therapy (the remote paper lists every session).
2. Give the SCQ (frequency and belief ratings for fearful thoughts) and the SPWSS (weekly ratings of avoidance, self-focus, anticipatory worry and post-event rumination) every session; give the SBQ (safety behaviours) and SAQ (standards, conditional and unconditional beliefs) at baseline, mid-point and end.
3. Obtain written consent at assessment for all sessions to be recorded; explain that recordings let the client see how they come across and that they will receive audio copies to review.
4. Review measures at the start of each session and pick the session's target from the highest-rated beliefs and the SPWSS items that are not moving.
Example language (illustrative): "We record every session here as a matter of routine. Most people find it awkward for the first few minutes and then forget the camera. The recordings become some of the most useful evidence we have."
Watch for: Refusal of recording (see section 5). Working well: the client completes measures without prompting, and the measures visibly shape the session agenda.

4.2 Goal setting

[source: Warnock-Parkes et al. 2020]

Purpose / when: Session 1. Gives direction and later provides material for experiments and for the blueprint.
Steps:
1. Ask what the client would be doing differently if social anxiety were no longer a problem.
2. Turn each answer into a specific, observable goal (e.g. speak at least once in team meetings; eat lunch in the staff room).
3. Record the goals and revisit them at mid-point and in the blueprint.
Example language (illustrative): "If you woke up and this fear had gone, what would I see you doing next week that you are not doing now?"
Watch for: Goals framed as feelings ("never feel anxious") rather than actions; reframe toward behaviour. [synthesis]

4.3 Individualised cognitive model (formulation)

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020; NICE CG159; Clark & Wells 1995 as summarised in James 2005]

Purpose / when: Session 1, then updated throughout. Shows the client how their own processes keep the anxiety going and sets up the session 2 experiment. This also delivers the "education about social anxiety" element in the NICE specification.
Steps:
1. Choose a recent, typical anxiety-provoking social situation.
2. Elicit the negative automatic thoughts in it (what was the worst that could happen; what did you think people would think of you?).
3. Elicit the self-image: "When you felt most anxious, did you have an image or impression of how you looked? What did you look like in it?" Ask whether it was seen from inside or as if from outside.
4. Elicit the somatic and cognitive symptoms of anxiety (blushing, shaking, sweating, mind going blank).
5. Elicit safety behaviours in detail, overt (hiding hands, avoiding eye contact, gripping a cup) and covert (rehearsing sentences, monitoring what one says, suppressing). The SBQ helps.
6. Elicit where attention was (on self, monitoring) and what the client noticed about others.
7. Draw the links on paper or on screen: thoughts lead to self-focus and the self-image, which feeds the anxiety symptoms and safety behaviours, which in turn feed back into the thoughts and image. Add anticipatory worry before and post-event processing after if present.
8. Ask Socratically what effect each safety behaviour and the self-focus might have (on how they feel, on how they come across, on what they can learn).
9. Give the client a copy. Remote: type it live in a shared document, or use the blank OxCADAT template.
Example language (illustrative): "So in the meeting your attention went inward, you got this picture of yourself bright red and stammering, and you started rehearsing every sentence. If you are busy rehearsing, how much of the other people can you actually see?"
Watch for: Too abstract a formulation; keep it to one concrete situation. Working: the client starts to say things like "so the rehearsing might make me come across as distant".

4.4 Self-focused attention and safety behaviours experiment

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020; NICE CG159]

Purpose / when: Usually session 2. Demonstrates experientially that self-focus and safety behaviours increase anxiety and worsen how one comes across. Produces video for session 3.
Steps:
1. Pick a task that activates the key fears without overwhelming the client: usually a brief one-to-one conversation with a stranger (a colleague the client does not know), or a small group if less threatening. For public-speaking fears, two short talks to two people.
2. Brief the conversation partner to talk naturally; they should not be told the client's specific fears.
3. Set up recording (see 4.5 for camera set-up), with both people in a wide shot.
4. Condition A (a few minutes): the client deliberately focuses on themselves, monitors their performance and uses all their usual safety behaviours (e.g. prepares what to say, grips a cup tightly).
5. Immediately rate 0 to 100: attention on self, safety behaviour use, anxiety, self-consciousness, how far each feared outcome occurred (e.g. how much they blushed, how boring they seemed), and overall how the conversation went.
6. Condition B: same partner, the client drops the safety behaviours (e.g. holds the cup loosely, says whatever comes to mind) and focuses outward, getting absorbed in the other person and the conversation.
7. Re-rate the same items.
8. Compare the two columns on a table and draw out the conclusion Socratically.
9. Ask the partner for written feedback (general impressions plus ratings of the client's specific predictions) to use later.
Example language (illustrative): "For the first conversation, I would like you to do everything you normally do to stay safe: plan what to say, keep checking how you are coming across. For the second one, do the opposite: drop the planning and let yourself get interested in her."
Watch for: Clients who do not truly drop covert safety behaviours (check by asking what was going on in their head). Order is usually A then B; if anxiety simply habituated, B will also look better, which the client may point out. [uncertain] Handling of order effects is not described in the sources; a pragmatic response is to repeat the comparison in later experiments [synthesis]. Working: anxiety and self-consciousness ratings drop in condition B, and the client reports feeling they came across better.

4.5 Video feedback (with cognitive preparation)

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020; NICE CG159]

Purpose / when: First in session 3 (the session 2 recordings), then repeatedly with recorded experiments and unplanned in-session moments. Corrects distorted self-images by comparing them with what is actually visible.
Recording set-up:
1. Elicit the main fears first, so you know what must be visible.
2. Camera small and out of the direct line of sight (e.g. on a bookshelf).
3. Wide shot of both people, not a close-up of the client; close-ups exaggerate fidgeting and invite self-scrutiny, while wide shots show the other person moving similarly.
4. Calibrate for the feared sign, e.g. for blushing, have a colour chart or coloured objects in the frame without drawing attention to them; confirm before viewing that the feared feature would be visible if present (this pre-empts "the camera did not pick it up").
Cognitive preparation (before viewing):
1. Ask the client to predict on 0 to 100 scales how anxious they will look, each idiosyncratic feared sign, and (optionally) overall performance.
2. Operationalise vague predictions: what exactly will we see, and for how long?
3. Have the client demonstrate the predictions physically: shake a hand as much as they think it shook, pick the blush shade on the chart, show the size of a sweat patch, recreate the length of a pause.
4. Ask them to close their eyes and run an internal video of how they think they looked; optionally note it down.
5. Explain that how we feel and how we look are different, and that the only way to tell them apart is to set feelings aside while watching.
6. Instruct an observer stance: watch as if watching two strangers, "like you are watching a television show" [source: Warnock-Parkes et al. 2020, short quote]; look at everyone, like a group conversation in a cafe; optionally call yourself "that person" or another name.
Viewing:
1. Watch the whole clip through first without pausing, unless the client is highly self-critical (then pause early, see below).
2. Rewind to the moments the client feared most (the felt panic, the long pause, the felt blush) and compare what is seen with the demonstration given before.
3. Ask: did the other person seem to notice? Are they reacting as if a big mistake happened? Do the two people look very different? Is the other person also fidgeting? Would an outsider pick out one of them as odd?
4. Look at how safety behaviours appear to others (e.g. turning away to sip a drink can look like disinterest; rehearsed topics can come across as lecturing), and note that they are choices that can be dropped.
5. Re-rate each prediction after viewing each clip and put before and after ratings side by side (a four-column table if two conditions were filmed).
6. Draw the conclusion Socratically: what stands out; how visible is your anxiety; are your feelings a good guide to how you come across?
7. Ask the client to bring the actual video image to mind in future situations when they feel they are coming across badly.
Adjustments for highly self-critical clients: watch the first 30 seconds with the sound off; cover the client's image for about the first minute so they watch the other person's reactions; ask how they listen to a close friend and invite them to watch themselves the same way ("go with the flow" [source: Warnock-Parkes et al. 2017, short quote]); pause early to ask whether they are watching as they would anyone else or as their own harshest critic.
Unplanned moments: when a client says in session that they just blushed, stumbled or talked nonsense, ask for a prediction of how it looked and check the recording. Therapists' own small dysfluencies, which clients usually do not notice, can also be pointed out.
Example language (illustrative): "Before we watch, show me with your hand how much it was shaking when you lifted the cup." / "Pause there. Are you watching her the way you would watch anyone, or the way your harshest critic would?"
Watch for: Selective searching for new flaws not predicted beforehand (hold the client to their specific predictions); "I felt anxious so I must have looked it"; discounting the camera. Working: post-viewing ratings well below predictions (in the Oxford series 98% of patients rated themselves more favourably after viewing) and a spontaneous statement that feelings are not what others see.

4.6 Conversation partner (confederate) feedback

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020]

Purpose / when: After the session 2 experiment and other experiments with a partner. Adds another person's perspective, especially when the client discounts their own viewing.
Steps:
1. Use a two-sided form: side one asks for general impressions; side two asks the partner to rate each of the client's specific predictions on 0 to 100 scales.
2. Read the feedback yourself first, then decide when to share it. For highly self-critical clients, sharing it before viewing the video can set a different frame.
3. Discuss: partners typically comment on what interested them, not on the feared signs, and their ratings match or are more positive than the client's post-viewing ratings.
Example language (illustrative): "Notice what she wrote about first. What does it tell you about where her attention was?"
Watch for: Very positive feedback that the client cannot absorb; share it with care and focus on the specific ratings.

4.7 Still photographs and flashcards

[source: Warnock-Parkes et al. 2017; Warnock-Parkes et al. 2020]

Purpose / when: After video feedback and experiments. Freezes moments of disconfirmation, which are easily lost from moving images and from memory.
Steps:
1. Capture a still from the video (or take a discreet photo during an outdoor experiment) at the moment the feared outcome did not happen, e.g. a stranger responding warmly.
2. Where useful, place images side by side: the predicted blush shade beside the actual face; the felt-worst moment beside a calm, externally focused moment.
3. Add a short learning statement in the client's own words, such as a version of "my feelings are not as visible as I think".
4. Keep it on the phone or a pinboard; review before difficult situations and call up the image when the old self-image intrudes.
Example language (illustrative): "Let's grab this frame. What would you want to write under it so that next Tuesday, before the meeting, you remember what we saw?"
Watch for: Flashcards used as reassurance rituals during situations rather than as reminders beforehand [synthesis]. Working: the client reports recalling the real image when anxious.

4.8 Externally focused attention training

[source: Warnock-Parkes et al. 2020; NICE CG159; Leigh et al. 2021]

Purpose / when: Usually the session after the first video feedback, then practised as homework. Builds the skill of moving attention from self to the environment and the other person, so experiments can be done properly. CT-SAD-A describes this as externally focused, non-evaluative attention.
Steps (graded sequence):
1. Sounds in the room and outside.
2. Colours.
3. Shadows, reflections or textures.
4. Music played aloud.
5. The therapist reading a passage aloud, the closest analogue to listening in conversation.
6. Then apply it in conversation: practise moving attention to the other person's face, words and reactions, and notice what happens to anxiety and self-consciousness.
7. Homework: daily practice in low-stakes situations, then in social ones. OxCADAT attention-training videos can be used.
Reducing self-consciousness while training: the therapist sits or positions the camera so they are not looking directly at the client; both can close their eyes during listening tasks; watch the training videos together.
Example language (illustrative): "Just notice all the sounds you can hear, near and far. When you find your attention has drifted back to yourself, that's fine; gently move it out again."
Watch for: Clients turning external focus into a new form of monitoring (e.g. scanning faces for disapproval). Emphasise being absorbed and non-evaluative. Working: lower self-focus ratings on the SPWSS and better-run experiments.

4.9 Behavioural experiments (dropping safety behaviours, external focus)

[source: Warnock-Parkes et al. 2020; Warnock-Parkes et al. 2017; NICE CG159]

Purpose / when: Most sessions from about session 4, each with linked homework. The main engine of belief change. Early experiments test what happens when safety behaviours are dropped and attention is external; later ones decatastrophise (4.10).
Consent and safety (exposure-type work): Agree each experiment collaboratively; explain the rationale and that anxiety will rise; choose a step that activates the fear without overwhelming the client; the client can stop at any time. Public experiments must respect local law and venue rules and should not involve deception that could harm or embarrass third parties [synthesis]. Contraindications or reasons to defer: acute suicidality, intoxication, or a current crisis that makes the client unable to process the experiment; manage these first [synthesis].
Steps (six-step sequence described by the Oxford group):
1. Identify the key belief to test, guided by the model and the SCQ (e.g. "if I don't prepare what to say, I will dry up and people will think I'm stupid").
2. Loosen it beforehand: review evidence, consider what other people would think, consider a survey.
3. Design the situation and make specific, observable predictions with percentages: what exactly will happen, how will we know, how will we measure it?
4. Run it with safety behaviours dropped and attention externally focused (prompt both just before starting).
5. Discuss the outcome against the predictions; generalise to the belief and update the self-image (what does this say about how you come across?). Re-rate belief.
6. Set further experiments as homework, recorded by video, photos or partner feedback where possible, on a record sheet (situation, prediction and rating, safety behaviours to drop, what happened, what I learned, belief re-rating).
Settings: in the office with a colleague or stranger; phone calls in session (e.g. to a shop, without rehearsing); outside with the therapist (e.g. asking directions, counting how many people actually look); remotely with a colleague joining the call or the therapist on the phone while the client is in public.
Example language (illustrative): "You predict 80% that if you stop rehearsing you will go blank and she will look confused. How will we know if that happened? What would she do?" / "Before you go in: drop the script, and put your attention on what she is saying."
Watch for: Hidden covert safety behaviours (ask afterwards what was going on in your head); the client attributing success to luck or to a remaining safety behaviour (find it and drop it next time); predictions too vague to disconfirm. Working: belief ratings on the SCQ fall across experiments and the client starts proposing their own experiments.

4.10 Decatastrophising experiments (deliberately showing feared signs)

[source: Warnock-Parkes et al. 2020; Warnock-Parkes et al. 2017]

Purpose / when: Mid to late therapy, after the client has tested dropping safety behaviours. Tests the assumption that visible anxiety or imperfection leads to rejection, and targets high standards and unconditional beliefs. Shows that visible "flaws" are part of being human.
Steps:
1. Identify the feared sign or behaviour (blushing, sweating, shaking, pausing, saying something boring or wrong, having nothing to say, being stared at).
2. Agree how to display it deliberately and safely, e.g. dampen the underarms with water to mimic sweat patches, let a hand tremble while holding a cup, add a deliberate pause during a talk, make a mundane comment, ask a simple question.
3. Predict specifically how others will react, with ratings.
4. Carry it out with external focus, ideally recorded or with the therapist observing others' reactions.
5. Compare reactions with predictions; ask what this means for the rule that one must never show anxiety, and for what kind of person they are.
Example language (illustrative): "What do you predict will happen if you let your hand shake while you drink your coffee in front of the barista? How many people will notice, and what will they do?"
Watch for: The client minimising the display (a disguised safety behaviour); a genuinely negative reaction from someone (see section 5). Working: the client concludes others are far less focused on them than they assumed.

4.11 Surveys

[source: Warnock-Parkes et al. 2020; Warnock-Parkes et al. 2017]

Purpose / when: As needed, often to loosen a belief before an experiment, and to normalise physical symptoms. Gives data on how other people actually view the feared sign.
Steps:
1. Write a few short questions with the client, e.g. for a blushing fear: do you blush? have you noticed others blushing? what do you think when you see someone blush? how odd would it seem, 0 to 100?
2. The therapist sends the survey to colleagues or acquaintances (or the client sends it if able).
3. The therapist reads responses first, then shares them in session or by screen share.
4. Discuss what the answers say about the client's prediction.
5. Optional: OxCADAT supplies images of public figures visibly sweating or blushing; ask whether those people look anxious or incompetent.
Example language (illustrative): "Your rule is that blushing makes people think you are weak. Let's ask twenty people what they actually think when they see someone blush."
Watch for: Responses dismissed as politeness; ask whether anonymity changes that, and include an anonymous option [synthesis].

4.12 Virtual audiences and public-speaking experiments

[source: Warnock-Parkes et al. 2020]

Purpose / when: As needed for presentation and performance fears, especially remotely.
Steps:
1. Choose one of the OxCADAT online audience videos (nine audiences of graded difficulty; harder ones include distracted members).
2. The client stands, sets the audience to full screen, hides their own self-view, drops safety behaviours and focuses on the audience.
3. Record the talk.
4. For video feedback, keep the audience image beside the recording, or replay the audience, so the client can judge the talk from the audience's viewpoint.
5. Decatastrophise: e.g. add a deliberate pause and then check on video how noticeable it was.
Example language (illustrative): "Let's watch it the way they would have: here is the audience, and here is you. Where is the shaking you predicted at 70%?"
Watch for: Rehearsed scripts creeping back in.

4.13 Online and social-media experiments

[source: Warnock-Parkes et al. 2020]

Purpose / when: When clients have parallel fears and safety behaviours online (censoring posts, repeatedly rehearsing messages, checking reactions).
Steps: Identify the online safety behaviour; predict the outcome of dropping it (e.g. sending a message without rewriting it); run it as homework; record the outcome. Because responses can be delayed, set a reminder (flashcard on the computer, phone screensaver) to shift attention to another task after posting rather than monitoring.
Example language (illustrative): "Write the message once, send it, and then go and do something absorbing for an hour. Note what you predict will happen."
Watch for: Repeated checking of reactions after posting, which works like post-event processing.

4.14 Anticipatory worry and post-event processing

[source: Warnock-Parkes et al. 2020; NICE CG159; Clark & Wells 1995 as summarised in James 2005]

Purpose / when: Added if SPWSS ratings for anticipatory worry or rumination are not improving. Pre-event worry primes self-focus and expected failure; the post-mortem reconstructs events from anxious feelings and the negative self-image, so they are remembered as worse than they were.
Steps:
1. Elicit a recent example of each: what did you go over before and after, for how long, and what did it make you conclude?
2. Examine the perceived advantages (preparing, avoiding mistakes, learning) and disadvantages (more anxiety, entering self-focused, a distorted memory), Socratically.
3. Compare the post-mortem's account with objective data where available (video, partner feedback).
4. Switch from worrying to testing: when a worry arises, the client asks how it could be tested by an experiment, and writes it on the experiment record sheet.
5. Agree a plan for dropping or postponing the worry and moving attention to an absorbing task; use reminders.
Example language (illustrative): "After the party you spent three hours replaying it. Did the replay give you better information about how you came across, or mainly information about how you felt?" Example from the source: a client predicted 80% visible shaking and saw roughly 5 to 10% on video.
Watch for: Rumination disguised as "reflecting" or problem solving. Working: reduced SPWSS worry and rumination ratings.

4.15 Updating the self-image (ongoing)

[source: Warnock-Parkes et al. 2020; Warnock-Parkes et al. 2017]

Purpose / when: Throughout therapy, built into every experiment and video review.
Steps: After each experiment ask what it says about the old self-image; how did you actually come across on the video; could a still be your reminder? Rate belief 0 to 100 in the self-image (how much do you believe you come across as X?) and track it. Encourage deliberate recall of the updated image before and in situations.
Example language (illustrative): "When you walked into the meeting, the picture was of a red, sweating face. Which picture fits what we saw today?"
Watch for: An image that does not shift despite repeated disconfirmation; this is the cue to look for an early memory (4.16, 4.17).

4.16 Discrimination training for memory-linked images (THEN vs NOW)

[source: Warnock-Parkes et al. 2020; NICE CG159]

Purpose / when: When a self-image or felt sense is triggered by reminders of past social trauma (bullying, ridicule), usually later in therapy; also during video feedback when the client re-experiences old humiliation.
Steps:
1. Identify the memory that the image or feeling belongs to.
2. Identify triggers in current situations that resemble the past (a laugh, a group, a look).
3. Write a two-column table: THEN (age, setting, who was there, what they did) vs NOW (current age, setting, people, how they actually respond).
4. Practise noticing the differences in the moment, especially when the old feeling arises; make a flashcard to keep visible (e.g. by the screen for video calls).
Example language (illustrative): "That laugh took you straight back to the classroom. Let's look at what is different: then you were ten, with those boys; now you are thirty-four, in a meeting with colleagues who were laughing at the joke."
Watch for: Intense distress that suggests a PTSD-level memory; consider a fuller trauma focus [synthesis].

4.17 Imagery rescripting of early social memories

[source: Warnock-Parkes et al. 2020, describing the Wild & Clark 2011 procedure; NICE CG159. Wild & Clark 2011 itself is not drawn on.]

Purpose / when: When a negative self-image persists and is linked to an early socially traumatic memory, usually later in therapy. Updates the meaning encapsulated in the memory.
Steps (four-stage outline):
1. Identify the recurrent negative self-image and the early memory associated with it (ask when the client first remembers feeling like this or seeing themselves this way).
2. Identify the present-day meaning of the memory (the encapsulated belief, e.g. I am weird and unacceptable) and rate belief 0 to 100.
3. Cognitive restructuring first: from an adult perspective, use Socratic questions to find alternative evidence and understanding (what was going on for the bullies, what would you think if this happened to a child you know now?).
4. Rescript the memory in the present tense in three passes: (a) the younger self recalls and re-experiences the event; (b) the client re-runs it as the adult self, watching and intervening, bringing in the new understanding; (c) the client, as the younger self, experiences the adult self's intervention and receives compassion from them. Re-rate belief and distress after.
Example language (illustrative): "Now go back into the playground as the adult you are today. What do you want to do or say to those boys, and to the girl standing there?" / "Now be ten again. The adult you is beside you. What does she say to you, and how does that feel?"
Remote adaptation: more verbal warmth and encouragement; check distress verbally ("how distressing is it to talk through the memory in this way?" [source: Warnock-Parkes et al. 2020]); use photos of the client as an adult as a present-day anchor.
Watch for: Overwhelm or dissociation (slow down, ground in the present, return to restructuring); the adult self unable to intervene (the therapist can suggest options or join in imagery) [synthesis]. [uncertain] Detailed instructions on session timing, how much of a session to allot, and handling of stuck rescripts are in Wild & Clark 2011.

4.18 Unconditional beliefs, low self-esteem and self-criticism

[source: Warnock-Parkes et al. 2020; Leigh et al. 2021; NICE CG159 ("examination and modification of core beliefs")]

Purpose / when: From session 2 onward within experiments; extra methods when beliefs (e.g. I am unlikeable, I am boring) persist.
Steps:
1. Include the unconditional belief as one of the rated predictions in the session 2 experiment, video feedback and later experiments.
2. After each experiment ask explicitly what it says about the broader belief; use decatastrophising to show that visible imperfection is acceptable.
3. If the belief persists: look for linked early social trauma (4.16, 4.17).
4. Positive data log (method from Padesky 1994): the client records daily any evidence, however small, consistent with an alternative self-belief.
5. Label the self-critic: help the client see the critical voice as an "internal bully" [source: Warnock-Parkes et al. 2020] rather than fact, and practise responding as they would to a friend.
6. Write a flashcard with the alternative belief and evidence, kept somewhere visible (e.g. the fridge).
Example language (illustrative): "If a friend told you they had said something a bit dull at lunch, would you conclude they were a boring person? What would you say to them?"
Watch for: Beliefs typically improve after experiments show the client is more acceptable than they feel and as goals are met; if not, check for comorbid depression [synthesis].

4.19 Therapy blueprint (relapse prevention)

[source: Warnock-Parkes et al. 2020; NICE CG159]

Purpose / when: Final sessions. Consolidates learning and plans for setbacks.
Steps:
1. Collaboratively write a document covering: how the problem developed and was maintained (the model); the key things learned (e.g. feelings are not what others see; safety behaviours make things worse); what helped most; further experiments and goals to build on progress; likely future high-risk situations; what to do in a setback.
2. Give it as homework to draft, then review together (remote: email it and review by screen share).
3. The client saves it somewhere easily accessible.
Example language (illustrative): "Imagine it is a year from now and the old fear flares before a big interview. What will you need to remind yourself of, and what will you do?"
Watch for: A blueprint written by the therapist rather than the client. Working: the client can state the model and the plan in their own words.

4.20 Booster sessions

[source: Leigh et al. 2021; Clark et al. 2006 trial design [uncertain]]

Purpose / when: After the main course. Review the blueprint, continuing experiments and any returning safety behaviours or avoidance.
Steps: Review measures; identify any situations still avoided; set new experiments; update the blueprint. [synthesis for detailed content]

4.21 Remote delivery set-up

[source: Warnock-Parkes et al. 2020]

Purpose / when: Before the first remote session and ongoing.
Steps:
1. Test video call beforehand; agree to phone if the connection drops.
2. Agree a private space, headphones, lighting that avoids a bright window behind the client, and minimising interruptions.
3. Ask the client to hide their self-view (it promotes self-focus).
4. Record with consent, check both people are captured, give audio copies. If clients record themselves for experiments, they should not watch the video alone before joint feedback.
5. Share blank forms by email or screen share; clients keep a file of completed work.
6. Avoid staring into the camera early; check in often because non-verbal cues are harder to read.
Watch for: Telephone-only sessions becoming an avoidance; move to video as soon as possible.

4.22 Adolescent adaptations (CT-SAD-A)

[source: Leigh et al. 2021; Leigh & Clark 2016 [uncertain]]

Purpose / when: Young people (CAMHS settings).
Steps: Deliver the same core components (individualised model, self-focus and safety behaviours experiment, video feedback, external non-evaluative attention training, behavioural experiments), with age-adapted measures. Add, when indicated: work on socially traumatic memories, anticipatory and post-event processing, low self-esteem, parental beliefs and behaviours that maintain the anxiety, and bullying and its consequences. [uncertain] Specific procedures for parent involvement and school-based experiments are described in Leigh & Clark 2016; plan parent sessions to share the model and to reduce parental accommodation (e.g. speaking for the young person) [synthesis].
Watch for: Parents or school staff inadvertently maintaining safety behaviours [synthesis].

5. Handling problems

Items tagged [source] come from the Oxford papers; items tagged [synthesis] are this guide's own clinical guidance, consistent with the model but not stated in the sources.

Problems during video feedback [source: Warnock-Parkes et al. 2017]

Still sees themselves negatively after viewing: redirect attention to the other person's reactions and the whole scene; bring in partner feedback (sometimes before viewing, to set a frame); use stills of disconfirming moments.
Searches for new flaws: hold the client to the specific predictions made beforehand; restore the whole-scene, stranger-watching stance.
"I felt anxious, so I must have looked anxious": separate feeling from appearance explicitly, guide the client to the conclusion that their feelings are private and not a reliable judge, and prompt recall of the video image in future.
Discounts the video (poor colour, not zoomed in): prevent with calibration and a wide shot; confirm before viewing that the feared feature would be visible.
Mistakes safety behaviours for personal deficits: ask whether they were deliberately holding back or monitoring; once seen as a choice, it can be dropped.
Ratings still driven by feelings: if partner ratings are more positive, ask whether private feelings are still shaping the client's own ratings.
Physical appearance concerns: isolate the feared body part from the full image, remove identifying features and compare with the calibration image. [synthesis] If appearance preoccupation dominates, assess for body dysmorphic disorder.
Re-experiencing past humiliation while watching: contrast present people and setting with the original event (THEN vs NOW).
Harsh self-critical commentary: sound off for the first 30 seconds, cover the image initially, adopt the friend stance, pause to check the viewing mode.

Stuck or slow progress

SPWSS rumination items not improving: add the anticipatory and post-event processing module (4.14). [source: Warnock-Parkes et al. 2020]
Self-image unchanged despite disconfirmation: look for a linked early memory; use discrimination training and rescripting. [source: Warnock-Parkes et al. 2020]
Unconditional beliefs persist: memory work, positive data log, internal-bully labelling, self-compassion, flashcards. [source: Warnock-Parkes et al. 2020]
Experiments not producing change: check for covert safety behaviours (rehearsing, monitoring, mental scripting), check attention was really external, make predictions more specific, and move to decatastrophising. [synthesis drawing on 4.9 and 4.10]
Short sessions or too little in-vivo work: in the CT-SAD-A case series, poorer outcomes were tentatively linked to shorter sessions that limited in-session experiments and less therapist preparation time; protect 90-minute slots. [source: Leigh et al. 2021]

Avoidance, resistance and non-adherence [synthesis unless stated]

Avoids sessions, video or being seen: treat it as a safety behaviour to formulate, not a failure; start with less exposing tasks (e.g. a group rather than a one-to-one conversation, cover the image during viewing). Remotely, avoid telephone-only work and hide the self-view. [source in part: Warnock-Parkes et al. 2017 and 2020]
Refuses recording: explore the predicted catastrophe about being filmed and treat that as the first experiment; offer audio first or delete-after-viewing; do not coerce. Consent must remain voluntary.
Homework experiments not done: review what was predicted and what got in the way; make the next step smaller; run a version together in session first; schedule a specific time and place; use the phone to record it.
Does experiments but keeps a safety behaviour "just in case": name it, predict what would happen without it, and test that in the next experiment.
Alcohol or substances used before social events: formulate as a safety behaviour; assess for misuse; plan reduction alongside experiments, referring if dependence is present. [uncertain] Not addressed in the sources.

Overwhelm and distress

Anxiety too high to engage in an experiment: choose a less threatening step that still tests the belief; keep the therapist present; distress during experiments is the known risk of the method (codebook note). [synthesis]
Distress during rescripting: check distress verbally, slow down, ground in the present (current photo, current setting), return to the restructuring stage before re-entering imagery. [source in part: Warnock-Parkes et al. 2020]
Very negative partner feedback or a genuinely rude reaction from a stranger: treat it as data: how typical was it, what did others do, what does it say about that person rather than the client, and what did the client do that was acceptable? Rerun with other people. [synthesis]
Comorbid depression or suicidal ideation: assess risk at assessment and when mood drops, following current standard practice (ask directly, collaborative safety plan, lethal-means counselling about safe storage or removal, crisis contacts); treat severe depression first or alongside. [synthesis]

Ruptures and the relationship [synthesis unless stated]

The therapist as feared evaluator: expect it; reduce direct gaze early, sit at an angle, ask for feedback often, and use the client's reactions to the therapist as a live example of the model. [source in part: Warnock-Parkes et al. 2020]
Client feels pushed or humiliated by an experiment: acknowledge it, invite the client to say so (this is itself an experiment against the fear of disapproval), re-agree pace and choice, and revise the plan collaboratively.
Reassurance seeking from the therapist ("did I look OK?"): redirect to data the client can gather themselves (video, partner feedback, surveys) rather than giving reassurance.

Modality-specific difficulties

Cannot find conversation partners or strangers: colleagues from other teams, trainees, admin staff; phone calls to shops or helplines; public tasks with the therapist nearby or on the phone; virtual audiences. [source: Warnock-Parkes et al. 2020]
External focus turns into scanning for disapproval: reframe attention as absorbed and non-evaluative interest in what the other person is saying. [synthesis; CT-SAD-A wording "non-evaluative" from Leigh et al. 2021]
Flashcards and stills become reassurance rituals: use them before situations, not as checking during. [synthesis]
Adolescents with accommodating parents or ongoing bullying: work on parental beliefs and behaviours and on bullying is part of CT-SAD-A when indicated; current bullying needs safeguarding action with the school, not cognitive work alone. [source in part: Leigh et al. 2021; safeguarding point is synthesis]

Sources

Sources:

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

Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope & F. R. Schneier (Eds.), Social phobia: Diagnosis, assessment, and treatment. Guilford Press.
Clark, D. M. (2001). A cognitive perspective on social phobia. In W. R. Crozier & L. E. Alden (Eds.), International handbook of social anxiety. Wiley.
Clark, D. M., Ehlers, A., Hackmann, A., McManus, F., Fennell, M., Grey, N., Waddington, L., & Wild, J. (2006). Cognitive therapy versus exposure and applied relaxation in social phobia: a randomized controlled trial. Journal of Consulting and Clinical Psychology, 74, 568-578.
Wild, J., & Clark, D. M. (2011). Imagery rescripting of early traumatic memories in social phobia. Cognitive and Behavioral Practice, 18, 433-443.
Leigh, E., & Clark, D. M. (2016). Cognitive therapy for social anxiety disorder in adolescents: a development case series. Behavioural and Cognitive Psychotherapy, 44, 1-17.
Padesky, C. A. (1994). Schema change processes in cognitive therapy. Clinical Psychology and Psychotherapy, 1, 267-278.
OxCADAT resources (oxcadatresources.com): forms, measures (SCQ, SPWSS, SBQ, SAQ), attention-training and virtual-audience videos.

← Previous · II-101Cognitive therapy for panic Next · II-103 →Intolerance-of-uncertainty therapy for GAD