This programme is a treatment manual written for mental health professionals. It assumes clinical training, the ability to make a diagnosis, and access to supervision. It replaces neither your clinical judgement nor your professional responsibility. The diagnostic criteria here are reformulated in our own words and never reproduced: consult the original manuals for the letter of the text.
1. The programme at a glance
Indication. Social anxiety disorder (social phobia) in adults, generalised or performance-only subtype.
Reference model. Individual cognitive therapy based on the Clark and Wells model (1995), as evaluated by Clark and colleagues (2003, 2006) and recommended as a first-line treatment by NICE (2013).
Format. Fourteen individual sessions of 90 minutes, weekly, over roughly four months, followed by two booster sessions at one month and three months. The 90 minutes are not negotiable in the early sessions: behavioural experiments and video feedback do not fit into 50.
Target mechanism. Not habituation to fear, but change in the three processes that keep the disorder going: self-focused attention, safety behaviours, and the distorted image of how one appears from outside.
| Session |
Focus |
What the session produces |
| 1 |
Assessment, measures, alliance |
List of situations, baseline measures |
| 2 |
Individual formulation |
Vicious-circle diagram, written with the patient |
| 3 |
Manipulation experiment |
Lived evidence that safety makes things worse |
| 4 |
Video feedback |
The gap between the felt image and the seen image |
| 5 |
Attention training |
External attention that holds up in situation |
| 6 |
First experiments outside session |
Written predictions, recorded outcomes |
| 7 |
Surveys |
Real data on other people's standards |
| 8 |
Anticipatory processing |
Spotting and dropping the mental rehearsal |
| 9 |
Post-event processing |
Less mental debriefing |
| 10 |
Images and early memories |
Imagery rescripting |
| 11 |
Rules and assumptions |
Assumptions made explicit and tested |
| 12 |
Generalisation |
Experiments in the avoided domains |
| 13 |
Residual avoidance |
Final inventory of safety behaviours |
| 14 |
Personal plan, relapse prevention |
Summary sheet written by the patient |
What the patient takes away. Seven printable worksheets, listed in section 30 and downloadable from this page: situation record, predictions and outcomes, survey, attention training, anticipation and rumination diary, exposure hierarchy, personal plan.
What sets this programme apart from graded exposure. Here exposure is a means of testing a prediction, not an end in itself. The patient does not enter the situation in order to "get used to it" but in order to gather information they cannot obtain any other way. This difference is not cosmetic: it changes what you say, what you write down, and what counts as success.
2. Before you start
Who this programme is for
This text is addressed to psychologists, psychiatrists, psychotherapists and mental health nurses trained in cognitive behavioural therapy. It assumes that you can conduct a diagnostic interview, recognise suicide risk, and recognise when a patient needs a different kind of care.
It is not written for patients. The text contains guidance on what must not be said, on the mistakes that make a session fail, and on what you observe when a treatment is not working. Handed directly to someone affected, it would produce more worry than help.
What this programme is not
It is not a research protocol. The sessions described follow the logic of the Clark and Wells model and the order used in the trials that validated it, but they do not reproduce a clinical trial manual word for word. You keep the freedom to spend two sessions on a point that will not shift.
It is not a treatment for social anxiety in children or adolescents. The model and the instruments differ, and the evidence on parental involvement does not transfer.
It is not a substitute for supervision. Three moments in this programme call for a steady hand: video feedback, dropping safety behaviours, and imagery work on early memories. If you have never run video feedback, have it supervised first.
How to use it
Read the whole thing before the first session. Sections 3 to 11 lay down the conceptual frame without which the sessions become a string of exercises; sections 13 to 26 are worth rereading the day before each session.
Every session is described on the same frame: the objective, a step-by-step run-through, what you say, the common mistakes, the materials, and the criterion for moving on. That last point deserves attention: this programme advances by acquisition, not by calendar. A session 4 run before external attention holds up produces nothing.
3. The clinical picture
What the patient describes
Patients rarely talk about social anxiety. They talk about shyness, about a lack of confidence, about a personality problem, sometimes about a physical defect. They often consult for something else: depression, burnout at work, drinking, difficulties in a relationship. The presenting request is usually about the consequence, not the disorder.
What organises the picture is a fear of negative evaluation in situations where one can be observed or assessed. The list of situations is long: speaking in a meeting, eating in front of others, making a phone call within earshot, writing while watched, entering a room where you are expected, passing a neighbour in a corridor, holding someone's gaze, voicing disagreement, receiving a compliment.
Three things separate the disorder from ordinary shyness. First, intensity: the fear is out of proportion to the real risk, and the patient knows it, which changes nothing. Second, reach: the avoided situations end up drawing the shape of a life — the choice of job, of where to live, the absence of a relationship. Third, suffering: the patient is not merely uncomfortable, they suffer from it and blame themselves for suffering.
What happens during the situation
Understanding what happens during fifteen minutes of a meeting is the core of the work. The Clark and Wells model (1995) describes a sequence that closes on itself.
The situation activates assumptions about oneself and about others. Attention then turns inward: the patient stops processing the situation and starts monitoring themselves. They no longer see faces, they feel their cheeks. This self-focused attention produces an image of oneself seen from outside, built from internal sensations rather than external information. The patient feels their hands shaking and concludes that their hands are visibly shaking, on the scale of what they feel.
That image feeds the fear, which feeds the sensations, which confirm the image. In parallel, the patient deploys safety behaviours: rehearsing sentences, speaking quietly, avoiding silences, holding a glass with both hands, keeping their eyes on their notes, sitting at the back of the room, having a drink beforehand. These behaviours have three effects, all bad. They consume attentional resources and genuinely degrade performance. They prevent disconfirmation, since the patient credits the absence of catastrophe to the precaution. And some directly produce what they are meant to hide: gripping a glass makes the hand shake, holding your breath makes you blush.
What happens before and after
Two processes bracket the situation and make it worse.
Before, anticipatory processing. The patient runs the situation through in their mind, pictures what could go wrong, revisits memories of failure. This pre-processing means they enter the situation already on alert, with an already negative self-image, and sometimes leads them to cancel.
After, post-event processing. The patient replays the scene, holds on to the awkward moments, judges themselves. This mental debriefing turns a neutral situation into evidence of failure, and then feeds the anticipation of the next one. It is through post-event processing that the disorder survives even when situations objectively go well.
Epidemiology, in two useful figures
Lifetime prevalence is in the region of 4 to 13 % depending on the study and the definition, with higher estimates in North America than in East Asia (Stein & Stein, 2008; Kessler et al., 2005). Onset is early: the median falls in adolescence, often before fifteen, which explains why many patients cannot remember a life without the disorder and take it for a character trait.
The delay before a first request for help is long, frequently more than ten years. That figure has a direct clinical consequence: the patient in front of you has usually built an entire life around avoidance, and the treatment will require undoing long-standing arrangements, not merely facing a fear.
4. The model that guides this programme
Why this one
Two cognitive models dominate the literature: Clark and Wells (1995) and Rapee and Heimberg (1997). They overlap substantially. The first has been chosen here because the individual protocol derived from it shows the largest effect sizes in direct comparisons, including against exposure with applied relaxation and against fluoxetine (Clark et al., 2003), and against group therapy (Stangier et al., 2003; Stangier et al., 2011).
NICE (2013) recommends, as a first-line treatment, individual cognitive behavioural therapy based either on the Clark and Wells model or on Heimberg's. It specifies a dose: fourteen sessions of 90 minutes over roughly four months. That is not an arbitrary figure: it is the dose that was evaluated.
The five levers
The model identifies five points to act on. The whole programme follows from them, and every session should be traceable to one.
Self-focused attention. The patient has to learn to direct attention outwards, onto the task and onto the other people. This is not distraction: it is the condition for disconfirming information to become available at all.
Safety behaviours. They have to be identified one by one, then dropped within experiments, so that the patient can see that the catastrophe does not happen and that their performance improves.
The self-image. It has to be corrected by external information: video feedback is the most effective tool for this, provided it is prepared according to a precise procedure.
Anticipatory and post-event processing. They have to be spotted, named, and deliberately interrupted.
Assumptions and rules. These are worked on late in the programme, once the patient has new data to set against them.
What the model implies you should not do
The model makes certain common practices useless or counterproductive.
Social skills training is not routinely indicated. Most patients have the skills; what degrades their performance is self-focused attention and safety behaviours. Offering skills training at the outset endorses the idea of a deficit and reinforces the disorder. It is justified only in the face of a genuine deficit, one you have observed and not merely been told about.
Relaxation is not indicated as a central technique. It easily becomes one more safety behaviour, and it turns attention towards internal sensations — that is, precisely where it should not go.
Long, repeated exposure without cognitive work produces inferior results. It leaves the self-image intact, along with the attributions to safety.
5. The DSM-5-TR criteria, reformulated
The criteria below are a reformulation in our own words, as an aide-memoire. They do not replace the manual: consult the DSM-5-TR (American Psychiatric Association, 2022) for the letter of the text and the application notes.
Social anxiety disorder requires the following elements together.
Marked fear of one or more social situations in which the person is exposed to the scrutiny of others, with a fear of being evaluated negatively — of being judged anxious, ridiculous, incompetent, or of giving offence.
Fear that is almost invariable in those situations: they provoke fear or anxiety nearly every time.
Avoidance, or endurance at the cost of marked distress.
Disproportion between the fear and the actual social risk, assessed with the cultural context in mind.
Duration of at least six months.
Clinically significant impact: distress, or impairment of social, occupational or personal functioning.
Exclusion of other causes: the effect of a substance or a medical condition, another mental disorder that accounts better for the picture, or a link with a condition or visible difference — a stammer, a scar, Parkinson's disease — in which case the fear must not be solely related to it, or must be excessive.
One specifier exists: "performance only", where the fear concerns only situations of public speaking or performance. That subtype calls for partly different management, noted in section 28.
What these criteria do not say, and what you must assess
The criteria do not rank situations. Build the full list, situation by situation, with the degree of avoidance: that is the raw material of the programme.
They say nothing about attention or safety behaviours, which are nonetheless the targets of treatment. An assessment that stops at the diagnosis gives you nothing to work with.
Frequently asked questions
Is filming really necessary?
Yes, and it is probably the least substitutable element of the programme. The gap between the felt image and the seen image cannot be demonstrated in words. If the patient refuses, do not give up immediately: the refusal is almost always about the idea of seeing oneself, not about the procedure. Explain the run-through, offer to start without sound, and offer to watch only thirty seconds. The refusal rarely persists beyond session 4.
What if I only have 50 minutes per session?
The programme loses effectiveness, and that has to be said. Two adaptations are possible: double up sessions 3, 4 and 6 into two consecutive appointments, or extend the total to eighteen sessions of 50 minutes. What does not work is keeping fourteen sessions of 50 minutes with the same content: the behavioural experiments are the first to go, and they are what acts.
The patient wants medication. What do I say?
That serotonin reuptake inhibitors are effective, that therapy is at least as effective and that its effects hold up better after stopping, and that the combination has not shown a clear advantage. If they are already taking something, they continue. If they want to start, that contraindicates nothing, with the exception of benzodiazepines.
How long before something shifts?
The session 3 experiment often produces an immediate change of conviction, and the session 4 video feedback produces another. The measures move later: wait until session 7 to judge.
Can this programme be run in a group?
Yes, and it remains effective, but less so than individual therapy in direct comparisons (Stangier et al., 2003). Two elements suffer in a group: the individual formulation and the video feedback. If a group is all you have, keep at least two individual sessions for those two moments.
And if the patient really does have a social skills deficit?
It happens, but it is rare, and it is established by direct observation, not by self-report. In that case targeted training is justified — but after the work on attention and precautions, never before. Offered at the outset, it confirms to the patient that their problem is the one they believed it to be.
The patient really does blush a great deal. Does that change anything?
The programme stays the same, and the surveys in session 7 are particularly useful: what people think of someone who blushes is almost always kinder than what the patient imagines. Avoid treating blushing as a symptom to be suppressed; it is the meaning that is treated.
Should the partner or the family be involved?
The protocol does not provide for it and the evidence is lacking. In practice, an information session with a partner is sometimes useful, especially if they are facilitating the avoidance in good faith — speaking for the patient, making their calls. In that case the message is simple: stop helping.
How many patients improve?
In the trials of the Clark and Wells model, the majority of treated patients no longer meet diagnostic criteria at the end of treatment, and gains hold at follow-up. Exact figures vary with the study and the thresholds used; tell the patient that marked improvement is the most likely outcome, without promising that the disorder will disappear.
The patient works remotely and has almost no social situations left. What now?
This has become common, and it is a trap: the absence of situations looks like improvement. Two practical consequences. First, the situations have to be built rather than waited for: an evening class, a club, an activity that imposes interactions one has not chosen. Second, remote situations have their own safety behaviours, to be treated like the others: camera off, microphone muted by default, written messages instead of calls, sentences drafted and reread ten times. Video calls are in fact excellent ground for video feedback, since recording there is trivial — subject to the consent described in section 12.
Should social anxiety be treated before or after an addiction?
It depends on severity and on function. Light alcohol use, clearly in the service of social avoidance, is treated within the programme: it becomes one of the safety behaviours to be dropped, and that often works well. An established dependence is treated first or alongside, with the specialist service appropriate to it. The simple rule: if the patient cannot get through a social situation without drinking, the programme cannot start, because no experiment will be interpretable.
The patient has a genuinely visible sign: a stammer, a tremor, heavy sweating.
DSM-5-TR provides for this: the fear must not be solely related to that condition, or must be excessive in relation to it. In practice the programme applies, with two adjustments. The surveys concern what people actually think about a stammer or about sweating, and the answers are often far less harsh than the patient anticipates. And the experiments do not aim to hide the sign, but to check what it produces when it is not hidden. Medical treatment of the sign, where it exists, proceeds in parallel without waiting.
Can I give this programme to my patient?
No. It is written for you, and it contains what must not be said. Give them the seven worksheets in section 30: that is what they are for.