Clinical Social Work Practice

Four seconds in

At 9:04 on a Tuesday, a man called Curtis Nabors sits down in a chair he did not choose, in a behavioral health clinic a hospital discharge planner sent him to, and asks before you have said anything past his name: how much of this goes back to my boss?

You have about four seconds. Answer badly and the next forty minutes become a negotiation over what he is willing to admit. Answer well and you have a client.

The comfortable answer is the wrong one. Everything you tell me here is confidential is not true, you know it is not true, and he will discover it is not true at the worst possible moment, which is after he has told you something. The accurate answer is longer and less soothing: nothing goes to your employer without your written authorization, and there are three situations where I am required by law to act on what you tell me whether you want me to or not. That answer costs thirty seconds of warmth and buys a working relationship. This lesson is about why.

Key idea: The first session is not mainly about gathering information. It is about establishing conditions under which information can honestly be given.

A note about the people in this course

Curtis Nabors does not exist, and neither does anyone else you will meet here. Every client in this course is an invented composite: a presentation assembled from patterns common in clinical settings, given a name so you have something to hold. No detail belongs to a real person and none of it is drawn from a chart. Said once, here, and assumed from now on.

The frame, and why it is not paperwork

Generalist practice taught you engagement: name, role, time, an invitation to say what the person came for. Clinical work adds a structure around that opening which the profession calls the frame. The frame is the set of explicit agreements about how the work will run, and a surprising proportion of clinical trouble is traceable to a frame that was never spoken aloud.

Read that table again and notice that every row is a boundary stated in advance rather than defended later. That is the whole design. A frame set in minute three is orientation. The same frame introduced in week nine, after the client has left four voicemails on a Saturday, is a rejection.

The limits row is the one students soften, and it is the one that must not be softened. The NASW Code of Ethics puts it in standard 1.07: inform clients of the limitations of confidentiality, and do it as soon as possible in the relationship, not when the limit is about to be exercised.

What makes practice clinical

A generalist social worker and a clinical social worker may sit in the same building and share a caseload, so the distinction is worth stating precisely rather than by vibe.

Clinical social work applies social work theory and method to the assessment, diagnosis, and treatment of mental, emotional, and behavioral conditions. Three things follow. It is a licensed activity: the independent clinical license, usually styled LCSW, requires an MSW from a program accredited by the Council on Social Work Education, post-degree supervised practice that is commonly about 3,000 hours over no fewer than two years, and a clinical examination, with details differing state by state. It carries the authority to diagnose, which is a financial act as much as a clinical one, because the diagnosis is what a payer reimburses. And it keeps the person-in-environment lens rather than dropping it: a psychologist and a clinical social worker treating the same panic disorder do overlapping work, and the social worker is the one who also asks whether the copay is affordable next month and whether the landlord has filed.

Say plainly what this course is. It is an education in the concepts, procedures, and evidence of clinical practice. It is not clinical training, and finishing it does not qualify you to assess, diagnose, or treat anybody. Competence is produced by doing the work under a supervisor who watches you do it, corrects you, and signs off.

Bordin's three components

Edward Bordin, in a 1979 paper of eight pages, made an argument that outlasted the school of therapy it came from. He proposed that what people were calling the therapeutic relationship had three separable parts, and that pan-theoretical means pan-theoretical: the same three parts operate in psychoanalysis, in behaviour therapy, and in a housing case.

1. Bond. The affective tie. Trust, liking, the sense of being taken seriously.
2. Goals. Agreement on what the work is for. Not your goal with the client's compliance, and not the client's goal with your private reservations.
3. Tasks. Agreement on the specific activities that will get you there, and belief on both sides that those activities are relevant.

The value of splitting the working alliance into three is diagnostic. When a case stops moving, most clinicians respond by trying harder at the bond: more warmth, more validation, longer sessions. Bordin's model says check the other two first.

The point: A stalled case is a specific failure with a specific location. Bordin gives you three places to look instead of one feeling to intensify.

What the alliance evidence supports

The alliance literature is large enough that both overclaiming and dismissal are easy, so hold the numbers.

The standard synthesis is by Christoph Fluckiger, A. C. Del Re, Bruce Wampold, and Adam Horvath, published in Psychotherapy in 2018, aggregating more than 300 studies and roughly 30,000 clients. The pooled correlation between alliance and outcome is about 0.28. That is a moderate association, and it is stable: it holds across treatment models, across client presentations, across measurement instruments, and whether the alliance was rated by the client, the therapist, or an outside observer.

Two refinements matter more than the headline. The first is timing. If alliance were merely a by-product of getting better, clients who improve would report good alliances afterwards and the correlation would be an artifact. Researchers measured alliance early instead, in sessions two or three, before much change has occurred, and early alliance still predicts later outcome. That does not prove direction; it makes the artifact explanation carry more weight than it comfortably can. The second is that alliance is partly a property of the clinician rather than of the pair. Some therapists form consistently stronger alliances across their whole caseload and their clients do better on average, which makes alliance a skill with a distribution, and is one reason supervision exists.

Now the limits. The association is moderate, and roughly 92 percent of outcome variance is doing something other than alliance. Almost all of this evidence comes from psychotherapy research, not from child protection investigation or benefits advocacy. And an alliance is not a resource: a superb relationship does not produce a housing unit, a bus pass, or an opening at the methadone clinic.

Ruptures, and the repair sequence

Jeremy Safran and J. Christopher Muran spent two decades studying the moments when alliances break, and their central distinction is easy to remember and hard to apply. Ruptures come in two forms. In withdrawal, the client moves away: shorter answers, agreeable surfaces, a cancelled session, intellectualizing. In confrontation, the client moves against: complaint, criticism, an accusation that you do not understand.

Confrontation ruptures are loud and therefore get addressed. Withdrawal ruptures are far more common and are routinely missed, because a polite client who attends reliably does not look like a problem. Politeness is not engagement.

Repair has a rough sequence, and the order is the part people get wrong.

1. Notice. Something changed in the room. Register it rather than continuing your plan.
2. Name it tentatively. Not a conclusion: a question. I noticed that when I brought up your mother the answers got shorter, and I would rather ask than guess.
3. Listen all the way through. Do not explain yourself yet. Explaining is the most common way clinicians convert a repairable rupture into a confirmed one.
4. Own your part specifically. You are right that I pushed. I did it because I was watching the clock, and that was my problem, not yours. Not an extended apology, which puts the client in the position of reassuring you.
5. Renegotiate goals or tasks. Ruptures often signal that you drifted onto an agenda the client never agreed to, which returns you to Bordin.

Catherine Eubanks and colleagues meta-analyzed rupture-repair studies in 2018 and found that sessions and episodes in which a rupture was repaired were associated with better outcomes than smooth ones. The result is not permission to provoke conflict. It says a relationship that has survived a visible disagreement has demonstrated something a frictionless one has not.

Worth holding on to: A quiet, agreeable client with nothing to report is the most commonly missed clinical problem in an outpatient caseload.

Back to Curtis

Here is the answer that works, in the room, in about forty seconds.

Nothing goes to your employer unless you sign a release naming exactly what I can send. Three things I have to act on regardless: if a child or a vulnerable adult is being hurt, if you tell me about a serious threat to a specific person, and if you are in immediate danger of ending your life. Everything else stays between us and the chart. Does that change what you want to talk about today?

The last question does real work. It concedes out loud that the limits may change what he says, and it hands him the decision, which is the first task the two of you agree on. If he says yes, it changes things, you have learned something important in minute four rather than month four.

Common misconceptions

• The alliance means the client likes me. Liking is one third of it. Goal and task agreement carry at least as much weight, and a well-liked clinician with no goal agreement runs a friendly holding pattern.
• Since relationship predicts outcome, technique is decorative. The pooled correlation is about 0.28. That is important and it is not the whole story, and a later lesson takes the argument apart properly.
• Stating the limits of confidentiality up front will make clients withhold. Some will withhold for a while. The alternative is a client who discloses in week five and learns in week six that you had to report it, which ends the relationship and often the treatment.
• A client who never complains has a good alliance. That is what a withdrawal rupture looks like from the outside.
• Clinical work is generalist work plus a diagnosis. The diagnosis is a small part. What changes is that assessment must produce a formulation, and the formulation must dictate the treatment, which is where the next four lessons go.

Where this leaves us

• The frame is six statements made in the first minutes: who you are, how long, what is written, who reads it, what it costs, and what happens between sessions. Stated late, each one reads as a rejection.
• The limits of confidentiality are said before the disclosure, not after it, and the Code requires this as soon as feasible in the relationship.
• Bordin's alliance has three parts, and a stalled case is nearly always a goal or task failure that clinicians misread as a bond failure.
• Alliance correlates with outcome at about 0.28 across 300-plus studies, early alliance predicts later change, and some clinicians are reliably better at it than others.
• Ruptures appear as withdrawal or confrontation; notice, name tentatively, listen fully, own your part, renegotiate. Repaired ruptures beat untroubled ones.

Sources

1. National Association of Social Workers. (2021). Code of ethics of the National Association of Social Workers (rev. ed.). NASW. socialworkers.org
2. Bordin, E. S. (1979). The generalizability of the psychoanalytic concept of the working alliance. Psychotherapy: Theory, Research and Practice, 16(3), 252-260.
3. Fluckiger, C., Del Re, A. C., Wampold, B. E., and Horvath, A. O. (2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316-340.
4. Eubanks, C. F., Muran, J. C., and Safran, J. D. (2018). Alliance rupture repair: A meta-analysis. Psychotherapy, 55(4), 508-519.
5. Wikipedia contributors. (n.d.). Working alliance. Wikipedia. en.wikipedia.org

An editorial that had to say what it was not

On 13 January 1996 the British Medical Journal published an editorial of about 900 words by David Sackett and four colleagues. Its title reads like a rebuttal because it was one: Evidence based medicine: what it is and what it isn't. Two years into the term's career, it was already being used to mean cookbook medicine, cost-cutting, and the abolition of clinical judgement, and the authors wrote the piece to take it back.

Their definition has three parts, and social work adopted all three: the conscientious, explicit, and judicious use of current best evidence, integrated with individual clinical expertise, and, in the version the profession settled on, with the client's values and circumstances. Notice that evidence is one of three inputs, not the whole of the decision.

Notice something else. Evidence-based practice in that formulation is a verb. It is something you do, case by case, in five steps. What most agencies mean by the phrase is a noun: a list of approved brand-name programs, procured centrally, that staff are trained in and audited against. Those two things are not the same and the difference matters on a Tuesday afternoon with a client in front of you who does not match the list.

The point: Evidence-based practice is a decision procedure you run, not an inventory you are issued.

The question, before anything else

Here is the case this lesson works. Trina Halvorsen is 31, has a four-year-old, and was referred by a child protection worker after a neglect finding that has since been closed. She meets criteria for major depression: flat, exhausted, in bed most of the day, not suicidal. Your agency runs one thing for depression, a sixteen-session cognitive behavioural therapy protocol delivered by the two staff who are trained in it, and the waiting list is eleven weeks. You have an opening in your own caseload now, and you are not CBT-trained.

The temptation is to ask a vague question: what is the best treatment for depression? That question has no answer that helps you. Step one converts a worry into something a database can answer, using a structure borrowed from medicine and usually called PICO.

Read the difference. The vague question invites an opinion. The PICO question can be answered wrong, which is what makes it worth asking.

Acquire, without pretending you have a university library

Step two is finding the evidence, and the honest version of this step admits that most practitioners have about twenty minutes, not an afternoon. Work down a hierarchy and stop when you have enough.

1. Systematic reviews and meta-analyses. Somebody has already read the fifty trials. Cochrane and the Campbell Collaboration produce these; both are free.
2. Practice guidelines from bodies that publish their methods. A guideline that shows its evidence tables is usable; one that does not is an opinion with a logo.
3. Individual randomised trials. Free full text for publicly funded research is now common on PubMed Central.
4. Everything else. Case series, pilot studies, a conference talk, the enthusiasm of a colleague. Not worthless, not decisive.

For our question, the relevant trial is easy to find and unusually clean. COBRA, published in the Lancet in 2016 by David Richards and colleagues, randomised 440 adults with major depression to behavioural activation from junior mental health workers or to CBT from qualified psychological therapists.

Appraise: four questions that decide whether a finding is yours

Reading a trial well is not about statistics. It is about four questions.

Who was in it? COBRA recruited from primary care and psychological therapy services in Devon, Durham, and Leeds, excluding people who were alcohol or drug dependent, acutely suicidal, or had bipolar disorder or psychosis. That last clause is where most exclusions bite. If your client is actively drinking, the trial did not study your client, and you should say so out loud rather than quietly assuming transfer.

Compared with what? This is the question people skip, and it is usually the important one. A treatment beating a waiting list has shown that doing something beats doing nothing. A treatment matching an active, well-delivered rival has shown something far stronger. COBRA is the second kind: the comparison was real CBT delivered by trained therapists.

What did it measure, and how big was the difference? The outcome was the PHQ-9, a nine-item depression questionnaire scored 0 to 27. At twelve months the mean difference between the two arms was 0.1 points, with a confidence interval running from -1.3 to 1.5. The trial had been designed as a non-inferiority study with a pre-specified margin of 1.9 points, meaning the researchers had committed in advance to what difference would be too big to ignore. The whole interval sits inside that margin.

Learn this distinction now, because it is the most common misreading in the applied literature. No statistically significant difference means the study failed to detect a difference, which a small study will always manage. Non-inferior means the study was designed and powered to rule out a difference bigger than a stated amount, and it did. Only the second licenses the sentence these two treatments work about as well.

Who paid, and who delivered it? COBRA was funded by the National Institute for Health Research, a public funder. And the delivery detail is the finding, not a footnote: the effect held when the cheaper treatment was given by staff with less training. That is what makes the trial answer your actual question, which was never really about behavioural activation. It was about whether Trina waits eleven weeks.

Apply: the third leg does the deciding

You now have evidence. You still do not have a decision, because two of Sackett's three inputs have not been consulted.

Your expertise: you have run behavioural activation twice under supervision and you know you drift into problem-solving when a client is silent. That is a real limitation and it belongs in the calculation and in your next supervision hour.

Trina's values and circumstances: she has a four-year-old and no car. Eleven weeks is roughly a school term. She has told you she is not interested in analysing her thoughts, which is a reasonable preference and happens to align with the treatment that does not require it. She also says her sister will watch the child on Thursdays, which makes a Thursday appointment a real appointment rather than a hope.

Only now is there an answer, and the answer is a sentence with a because in it: start behavioural activation on Thursdays, because a trial designed to detect a clinically meaningful gap did not find one, because the alternative costs eleven weeks, and because she can actually attend. Write that reasoning in the record. A decision whose reasoning is written down can be reviewed. One that is not is indistinguishable from a habit.

Assess: the step that gets dropped

Step five is measuring whether it worked, and it is the step agencies skip because it is the one that can embarrass them.

Use a short instrument at every session or every other session. The PHQ-9 takes about two minutes; the GAD-7 measures anxiety on seven items scored 0 to 21. Their value is not diagnostic precision. It is that a number recorded weekly makes a trend visible, and human beings are poor at seeing trends in their own work.

How poor is worth stating. Chad Hannan, Michael Lambert, and colleagues asked therapists to predict which of their current clients would end treatment worse than they started. Across roughly 550 clients the therapists flagged three. About forty actually deteriorated. An actuarial signal built from session-by-session questionnaire scores caught most of them. The clinicians were not incompetent; they were doing something people cannot do, which is detect a slow negative trend in a case they are emotionally invested in.

Why this matters: The measurement is not there to grade you. It is there to catch the client who is quietly getting worse while both of you agree that sessions are going fine.

When the list and the process disagree

Return to the noun version. Agencies convert evidence-based practice into a procurement list for understandable reasons: a list can be trained, audited, and billed. The costs show up in three places.

• Coverage. The lists are built from trials, and trials exclude the comorbid, the substance-using, and the unhoused. A caseload is mostly made of the excluded.
• Fidelity against fit. A protocol delivered with high fidelity to a client it does not suit is a well-executed error, and it produces excellent audit scores.
• Ossification. A list procured in 2015 is still the list in 2026, because changing it means retraining. The process has no such lag.

None of this is an argument for ignoring the evidence. It is an argument that the evidence enters your decision through a question you asked about a specific person, not through a purchasing decision made three years ago in another building.

Common misconceptions

• Evidence-based practice means using treatments from an approved list. That is the noun. Sackett's five steps are a verb, and the list is at best the output of step two for somebody else's question.
• No significant difference means the treatments are equivalent. It means no difference was detected. Equivalence requires a study designed to rule out a difference of a stated size, like COBRA's 1.9-point margin.
• The strongest evidence is a randomised trial. For a treatment question, a good systematic review of many trials beats any single trial. And for a question about what clients experience, a trial is the wrong instrument entirely.
• Client preference is a courtesy you extend after deciding. It is one of three inputs to the decision, and it is also the best predictor of whether the client comes back.
• Outcome measurement is for research. Clinicians reliably miss deterioration in their own caseloads, and a weekly score is the cheapest correction available.

Putting it together

• Sackett's 1996 definition has three inputs: best available evidence, clinical expertise, and the client's values and circumstances. Evidence alone never decides.
• The five steps are ask, acquire, appraise, apply, assess, and the first step is where most of the work is won or lost. PICO turns a worry into an answerable question.
• Appraise by asking who was in the study, what it was compared with, what it measured and by how much, and who paid and who delivered it.
• Non-inferiority is a designed conclusion with a pre-specified margin; no significant difference is a failure to detect. COBRA is the first kind, at a margin of 1.9 PHQ-9 points.
• Write the because. A decision with recorded reasoning can be reviewed; one without is a habit.
• Measure outcomes session by session, because clinicians flag a small fraction of the clients who are deteriorating.

Sources

1. Sackett, D. L., Rosenberg, W. M. C., Gray, J. A. M., Haynes, R. B., and Richardson, W. S. (1996). Evidence based medicine: what it is and what it isn't. BMJ, 312(7023), 71-72. PubMed Central
2. Richards, D. A., Ekers, D., McMillan, D., Taylor, R. S., Byford, S., Warren, F. C., et al. (2016). Cost and outcome of behavioural activation versus cognitive behavioural therapy for depression (COBRA): a randomised, controlled, non-inferiority trial. The Lancet, 388(10047), 871-880. PubMed Central
3. Hannan, C., Lambert, M. J., Harmon, C., Nielsen, S. L., Smart, D. W., Shimokawa, K., and Sutton, S. W. (2005). A lab test and algorithms for identifying clients at risk for treatment failure. Journal of Clinical Psychology, 61(2), 155-163.
4. Gambrill, E. (2019). Critical thinking and the process of evidence-based practice. Oxford University Press.

Three visits, no cardiac finding

Renata Marchetti is 37 and has been to an emergency department three times in five months with crushing chest pain, a racing heart, and the certainty that she was dying. Each time the workup came back clean: electrocardiogram normal, troponins normal, thyroid stimulating hormone within range. The third discharge summary contains a sentence she read in the car park and has not stopped thinking about. Likely anxiety. Referred to behavioral health.

She arrives in your office holding that piece of paper, and the first thing she says is that she is not crazy, she just wants to know what is wrong with her heart.

What follows is the whole of this module: an assessment that answers her question honestly, a diagnostic impression, and a formulation that explains why this is happening to this person now. This lesson is the first part, and it takes about seventy minutes of clinical time to do properly. You are going to watch it happen.

An assessment produces a formulation, not a completed form

You already know how to run an intake. Generalist practice taught you the domains and the eco-map. What changes here is the output. A generalist assessment answers what does this person need. A clinical assessment has to answer something harder: why is this person presenting in this way at this moment, and what would have to change for it to stop.

That is why the ordering matters. If you work down an agency form from demographics to presenting problem, you spend the first fifteen minutes on facts you could have read off the referral while a frightened woman waits to hear whether her heart is fine. Start where she is. The form can be completed at the end or between sessions, and a client who has been heard will fill it in willingly.

The core of it: The assessment interview is a clinical intervention in its own right. What you ask, and in what order, tells the client what kind of problem you think they have.

Biological, and why it comes first

George Engel published the argument for a biopsychosocial model in Science in 1977, and the part social workers tend to underuse is the first syllable. Psychiatric presentations are imitated by ordinary medical conditions with reliable frequency, and the person most likely to miss that is a clinician who was told anxiety on the referral and started there.

Renata has had two of these ruled out by the emergency department already, which is worth telling her plainly, because it answers the question she came in with. What has not been asked about is sleep. She sleeps four to five hours; her partner Vince says she stops breathing and jerks awake. She has never had a sleep study. That goes in your note as a referral, not as a symptom of depression.

The biological domain also includes what she is taking and drinking. She has three or four beers most evenings, up from one or two a year ago. You give her the AUDIT-C, three questions taking ninety seconds, part of the longer Alcohol Use Disorders Identification Test. She scores 5, above the threshold of 3 commonly used for women. That does not diagnose anything. It tells you the topic is live, and it tells you her early-morning waking at 4 am has a candidate explanation that is not depression.

Psychological: onset, course, and the three questions people leave out

Now the story, in her words, with you doing very little.

The first attack was on 14 March, at work. Renata is a shift lead at a food distribution warehouse. A stack of pallets went over in aisle nine and caught a coworker, Danny, across the foot; he screamed, the aisle filled with people, and she stood still. He kept his foot. She has not driven the highway route to work since, going the long way through town, forty minutes instead of eighteen.

Three questions get skipped by inexperienced interviewers, and each one changes the plan.

1. Has this happened before? A first episode and a fourth episode are different problems with different prognoses. Renata had a stretch at 22, after her father died, that sounds like a depressive episode; she did not see anyone about it and it lifted in about four months.
2. What has been tried, and what happened? Her physician prescribed sertraline in June. She took it for nine days, felt more jittery, and stopped without telling anyone. That is not treatment failure. It is a common early side effect, poorly explained, and it is fixable.
3. What do you think is happening? This is the explanatory model question and it is the one clinicians most often answer for the client. Renata thinks she has a heart problem the tests missed. Until that belief is addressed, no anxiety treatment will make sense to her.

Trauma history is asked here, deliberately and without excavation. You are asking whether difficult experiences are part of the picture, not conducting a detailed history in session one. Something like: many people who feel this way have been through frightening or harmful things at some point. Is there anything like that in your history that feels connected to this? She says the warehouse thing, and then, after a pause, that her father was a drinker and there was shouting. That is enough for today. A later lesson deals with what you do with it.

Risk is asked plainly, in the same voice as everything else. Have you had thoughts of ending your life? She says no, never, and adds that she has a nine-year-old. Ask anyway, ask directly, and write the answer.

Social: the domain that decides whether treatment is possible

This is the part a psychologist may skip and a social worker must not, because it determines what can actually be delivered.

• Work. Shift lead, rotating shifts including two nights a week. Rotating shifts and consistent Thursday appointments are incompatible, and if you do not learn this now you will call it non-compliance in six weeks.
• Money and coverage. Employer plan, 40 dollar copay, twelve outpatient visits a year. You have twelve sessions, and pretending otherwise is a disservice.
• Caregiving. Her mother Alma, 68, has chronic lung disease and lives eleven blocks away. Renata does her shopping and takes her to appointments. This is roughly six hours a week that does not appear on any form.
• Household. Vince works nights, so childcare for Mateo, 9, is a negotiation every week rather than a fact.
• Transport. One car. This matters double, because highway avoidance is a symptom and also a logistical constraint on where she can be treated.
• Supports. A sister in another state, close, phone contact weekly. Two friends from work she has been avoiding since March.

Why this matters: Every item on that list is both clinical information and a constraint on the treatment plan. The caregiving hours are a perpetuating factor and a reason Tuesday will not work.

Culture and meaning

Renata's mother emigrated from Argentina at 24; Spanish is the language of her mother's house and English of her own. When you ask what her family would call what is happening to her, she says her mother calls it nervios, and that her aunt had ataques where she would shake and cry and it passed.

This is worth taking seriously rather than translating away. Ataque de nervios appears in the DSM-5-TR as a cultural concept of distress, and it overlaps with panic attacks without mapping onto them cleanly: it is often triggered by a family event, often includes crying and shouting, and is understood by the people around it as a response to something rather than as an illness. The manual also supplies a structured way to ask about this, the Cultural Formulation Interview, sixteen questions covering how the person defines the problem, what they think caused it, what help they have sought, and what they expect from you.

Renata's answers matter clinically, not just respectfully. If her family frames this as something that happens to women in her family and passes, then a treatment described as a course of therapy for a disorder may be less acceptable than one described as learning to stop the attacks. The second description is also true.

Strengths, written as evidence rather than compliment

Strengths sections are where assessments go soft. She is resilient and motivated is not usable by anyone. Write what she did.

She has kept working through five months of this, taking a forty-minute detour twice a day to do it. She sought help herself after the third emergency visit rather than waiting to be sent. She kept the sertraline conversation from her physician, which tells you something to work on, and she came to this appointment having already read the discharge summary closely enough to quote it. She has a sister she calls every week. Those are load-bearing facts, and two of them will be used directly in treatment.

What the hour cannot give you

Be honest about the limits of what you have. You have one hour of self-report from someone who is managing your impression of her, which everyone does and which is not deceit. You have no collateral information. You have a symptom picture that has been filtered through five months of fear and one emergency clinician's sentence. Your assessment is a hypothesis with an evidence list attached, and it should be written in a way that lets the next person disagree with it.

Common misconceptions

• A biopsychosocial assessment means covering three headings. It means letting each domain change the conclusion. If the social section could be deleted without altering the plan, it was decoration.
• The referral told me it is anxiety, so the medical question is closed. A clean cardiac workup is not a sleep study, a metabolic panel, or a medication review, and the sleep finding here would have been missed by anyone who accepted the label.
• Asking about suicide plants the idea. It does not. Asking directly, in an ordinary voice, is standard practice, and the answer belongs in the record whichever way it goes.
• A stopped medication is a failed trial. Nine days of a common early side effect that nobody warned her about is an unfinished trial and a communication failure, and it is repairable in one conversation with the prescriber.
• Cultural questions are a courtesy at the end of the form. The client's explanatory model determines whether your treatment rationale is credible to them, which determines whether they do any of it.

What to carry forward

• The clinical assessment has to answer why this person, this presentation, now. That is a higher bar than what does this person need.
• Ask the presenting problem first, in the client's own words, and let the agency form wait.
• Thyroid disease, arrhythmia, sleep apnea, anemia, withdrawal, and stimulants all imitate psychiatric conditions; a clean cardiac workup rules out one thing, not the category.
• Ask whether it has happened before, what has been tried and what happened, and what the client thinks is going on.
• The social domain is where treatment feasibility lives: shifts, copays, session limits, caregiving hours, transport, childcare.
• Record strengths as specific actions with dates and consequences, and record cultural framing because it decides whether your rationale is believable.

Sources

1. Engel, G. L. (1977). The need for a new medical model: A challenge for biomedicine. Science, 196(4286), 129-136.
2. American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.). American Psychiatric Association Publishing.
3. Wikipedia contributors. (n.d.). Ataque de nervios. Wikipedia. en.wikipedia.org
4. National Institute of Mental Health. (n.d.). Anxiety disorders. NIMH. nimh.nih.gov
5. Wikipedia contributors. (n.d.). Alcohol Use Disorders Identification Test. Wikipedia. en.wikipedia.org

Two paragraphs about the same forty minutes

Here is how one clinician wrote up the session with Renata Marchetti.

Client presented as anxious and somewhat depressed. Poor eye contact. Affect flat. Insight fair.

Here is another, from the same forty minutes.

Client sat forward with both feet flat on the floor and looked at the door twice in the first five minutes. Speech was rapid, with pauses of several seconds before answering questions about the accident in March. She described her mood as wound up all the time and, later, as flattened. Affect was reactive: tearful once when describing her mother's oxygen tank, brightening when her son came up. She said she believes her heart problem was missed and that therapy is probably for people whose problems are in their head.

The first paragraph tells you the writer's conclusions. The second tells you what happened, and a clinician reading it next month can form a different conclusion from the same evidence. That difference is what the mental status examination is for.

Bottom line: The mental status examination records observations. The diagnosis records an inference. Keeping them in separate compartments is most of the skill.

The examination, domain by domain

The mental status examination is a structured description of a person's presentation at one moment. It is not a test you administer; almost all of it is gathered by watching and listening while you do something else. Here it is with Renata's entries beside it.

Two rules make this useful rather than decorative. Write what you saw, not what it means: not agitated but leg bouncing continuously through the session. And quote mood, because mood is the client's report and affect is your observation, and collapsing the two is the single most common error in this section. A person can report feeling fine while their affect is constricted and tearful, and that mismatch is clinical information that disappears the moment you write client appeared sad.

Where the examination stops and the diagnosis starts

Reported feeling unreal during attacks is an observation. Panic disorder is a claim. Getting from one to the other is a separate operation, and the DSM-5-TR is the tool the field uses to do it in a language other people share.

Be clear about what that manual is. Published in March 2022 as a text revision of the fifth edition, it is a classification: a list of named categories, each with a criteria set, each attached to an ICD-10-CM code that a payer recognizes. It is not a theory of causation, it does not tell you what treatment to use, and it repeatedly states that its criteria are for use by clinicians with appropriate training rather than as a checklist applied by anyone.

Almost every criteria set ends with two requirements that students skip and clinicians rely on. One is a clinical significance criterion: the symptoms must cause distress or impairment. The other is an exclusion clause, and the order it imposes is the discipline of diagnosis.

1. Rule out a substance or medication. Could this be caffeine, alcohol withdrawal, a stimulant, a steroid, a thyroid replacement dose set too high?
2. Rule out another medical condition. The list from the previous lesson: arrhythmia, hyperthyroidism, sleep apnea, anemia.
3. Rule out another mental disorder that accounts for it better. Panic attacks occurring only when facing a feared social situation are social anxiety disorder, not panic disorder.
4. Then, and only then, name the category and its specifiers.

Working the differential on Renata

Her attacks come out of nowhere, including once while watching television, which matters: unexpected attacks are what panic disorder requires, and attacks that only ever occur in one situation point elsewhere. She has had persistent worry about further attacks for five months and has changed her behaviour to prevent them. That is the criteria set.

The avoidance is its own question. She avoids the highway, the warehouse's aisle nine, and, increasingly, the supermarket. Avoidance of two or more situations, driven by fear that escape would be difficult or help unavailable, is agoraphobia, which the DSM-5 separated from panic disorder into a diagnosis of its own. Two diagnoses, not one, and the second is the one wrecking her week.

Depression is a genuine question rather than a formality. She has low mood, reduced interest, weight loss, and early waking. But the weight loss has a competing explanation in her drinking and the early waking has a competing explanation in both the drinking and the possible sleep apnea. This is where the exclusion order does real work: you do not add a depressive disorder to the list while two untested physical explanations for half its criteria are sitting in the chart. You write it as a differential, you refer for the sleep study, and you look again in six weeks.

Alcohol is recorded as what you actually know: an AUDIT-C of 5 and a self-reported increase, which is a finding, not a diagnosis of alcohol use disorder, since you have not asked the eleven criteria.

What matters here: The honest diagnostic line has an order of confidence in it. Panic disorder with agoraphobia is your working diagnosis; a depressive disorder is a differential awaiting physical results; alcohol use is a finding under review.

The limits, stated exactly

A clinician who uses this manual without knowing what is wrong with it will believe things that are not true.

Reliability is uneven, and worse for common diagnoses than people assume. When the American Psychiatric Association ran its own field trials before publishing DSM-5, patients at eleven academic centres were interviewed twice by different clinicians who did not know the first result. Of the twenty-three diagnoses with adequate samples, five reached the very good band, nine were good, six were questionable, and three were unacceptable. Some of the diagnoses in the weaker bands are among the most frequently made in outpatient practice. What that means concretely: two competent clinicians interviewing the same person a week apart will often disagree about whether a common diagnosis is present.

The categories are administrative as well as clinical. A code is what a payer reimburses, which puts steady pressure on clinicians to record a billable diagnosis early and a more accurate one never.

Comorbidity is the normal case, which is a hint about the categories. Most people who meet criteria for one disorder meet criteria for another. A classification whose members constantly co-occur is not carving nature at its joints; it is describing overlapping patterns with sharp lines drawn through them.

A diagnosis in a record has a long life. It follows a person into custody disputes, employment screening in some jurisdictions, disability determinations, and their own reading of themselves. Recording bipolar disorder because a client had three energetic days is not a neutral act.

The classic cautionary tale here needs its own caution. David Rosenhan's 1973 Science paper, in which pseudopatients reported a single hallucinated word and were admitted with schizophrenia diagnoses, is quoted in every textbook. Susannah Cahalan's 2019 investigation found that the study's records did not support key parts of the account, that most of the pseudopatients could not be traced, and that at least one participant's data appear to have been excluded because they contradicted the thesis. Use it as an example of how a memorable story survives its evidence, which is a lesson clinicians need at least as much.

What the diagnosis is worth

None of this is an argument for refusing to diagnose. A diagnosis is a compression: it lets you hand a colleague four words that carry a body of research about course, comorbidity, and what has been shown to help. It unlocks payment, which for most clients is the difference between treatment and no treatment. It sometimes ends a private terror, and Renata may be relieved rather than insulted to learn that her attacks are a recognized thing with a known treatment rather than a heart defect nobody can find.

What it cannot do is explain her. Panic disorder with agoraphobia does not tell you about aisle nine, Danny's foot, the six caregiving hours, or her mother's word nervios. Only a formulation does that, which is the next lesson.

Common misconceptions

• Mood and affect mean the same thing. Mood is what the client reports, in their words. Affect is what you observe. The mismatch between them is often the most informative line in the section.
• Thought process means what the client is thinking about. Process is how thinking is organized; content is what is in it. Circumstantial process with unremarkable content is a different picture from linear process with a fixed delusion.
• A DSM diagnosis explains the client's problem. It names a pattern. The explanation is the formulation, and treating the label as an explanation is the most common failure of clinical reasoning in new practitioners.
• Diagnostic categories are validated disease entities. They are consensus categories, revised by committee, with uneven test-retest reliability and pervasive comorbidity.
• Rosenhan proved psychiatric diagnosis is meaningless. The study is not sound. The real problem with diagnostic reliability is documented in the field trials that the manual's own publisher ran and printed.

The short version

• The mental status examination records observations in observational language; conclusions belong in the impression, not in this section.
• Quote mood, describe affect, and never merge the two.
• Diagnosis runs in a fixed order: substances, then medical conditions, then another mental disorder that fits better, then the category and its specifiers.
• Renata carries a working diagnosis of panic disorder with agoraphobia, a depressive disorder held as a differential pending a sleep study, and alcohol use recorded as a finding.
• The DSM-5 field trials found five diagnoses with very good reliability, nine good, six questionable, and three unacceptable, which is a fact about the tool you are using.
• A diagnosis compresses research and unlocks payment. It does not explain anybody, and it stays in the record for years.

Sources

1. American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.). American Psychiatric Association Publishing.
2. Regier, D. A., Narrow, W. E., Clarke, D. E., Kraemer, H. C., Kuramoto, S. J., Kuhl, E. A., and Kupfer, D. J. (2013). DSM-5 field trials in the United States and Canada, Part II: Test-retest reliability of selected categorical diagnoses. American Journal of Psychiatry, 170(1), 59-70.
3. Voss, R. M., and Das, J. M. (2024). Mental status examination. StatPearls. NCBI Bookshelf
4. Cahalan, S. (2019). The great pretender: The undercover mission that changed our understanding of madness. Grand Central Publishing.
5. Wikipedia contributors. (n.d.). Panic disorder. Wikipedia. en.wikipedia.org

Before you read

Write one sentence explaining why somebody you know developed a difficulty they have. Keep the sentence. At the end of this lesson you will check whether it names anything a person could actually change.

The question eleven pages did not answer

Your supervisor reads the assessment on Renata Marchetti, all eleven pages of it, sets it down, and asks: why 14 March? Pallets have fallen in that warehouse before. Danny kept his foot. She had a difficult stretch at 22 and came out of it. So what was different this time, and what is keeping it going now that the aisle has been restocked and Danny is back at work?

You have a diagnosis, and it does not help. Panic disorder with agoraphobia describes what she has. It says nothing about March, nothing about her mother's oxygen tank, and nothing about which of the many true facts in your assessment you should do something about on Thursday.

What answers the question is a formulation.

What a formulation is

A clinical formulation is a hypothesis about how this person's difficulty developed and, crucially, what is keeping it in place. It is not a summary of the assessment. A summary restates facts; a formulation asserts causal relationships between them, which means it can be wrong, which is what makes it useful.

The five Ps

The grid used across most of the field has five cells. Four of them are the classic Ps described by Priyanthy Weerasekera and others; the fifth, the presenting problem, is the row everything else has to explain.

• Presenting. What is happening now, stated in behaviour and frequency rather than diagnosis. Not anxiety, but four to six unexpected attacks a month plus daily avoidance of three situations.
• Predisposing. What made this person vulnerable to this kind of difficulty before it started. Temperament, family history, early experience, chronic health, social position.
• Precipitating. What set it off. Usually an event, sometimes an accumulation.
• Perpetuating. What is keeping it going right now. This is the cell that matters most and the one people fill in last and thinnest.
• Protective. What is working in the person's favour: abilities, relationships, resources, past recoveries.

Renata's grid

Read the perpetuating cell again. Every item in it is happening this week, and almost every item is modifiable. That is not an accident of this case. It is the design of the grid.

The upshot: Predisposing factors explain vulnerability and cannot be changed. Precipitating factors explain timing and are in the past. Perpetuating factors are where treatment happens, which is why a thin perpetuating cell means you do not yet have a plan.

Reading the grid as a treatment plan

Take the perpetuating items and ask, of each, what would remove it.

Notice that two rows are not psychotherapy at all. A formulation that produces only therapy tasks has usually been written by someone who forgot which profession they are in.

Four sentences, which is what actually gets used

Nobody reads a grid. What goes in the chart, and what you say out loud, is a paragraph. The reliable shape is four sentences: vulnerability, trigger, maintenance, resources.

Renata is a 37-year-old warehouse shift lead who grew up around an anxious mother and a drinking father, had one untreated depressive episode at 22, and pays close attention to what her body is doing. In March she watched a colleague injured in an accident she felt she should have prevented, and had her first panic attack that evening. Since then her attacks have been maintained by avoidance that prevents her ever learning the highway is safe, by three emergency visits that confirmed her belief in a cardiac cause, by alcohol that fragments her sleep, and by six unrelieved caregiving hours a week beside her mother's oxygen tank. She has kept working throughout, sought help on her own, and wants to be at her son's Saturday games by the spring.

That paragraph does what eleven pages did not. It answers why March, and it names five things to change.

Saying it to the client

A formulation you have not shared is a private theory. The test of it is whether the client recognises herself, and the way to find out is to say it and then genuinely invite the correction.

Here is a fragment. You: can I tell you how I am currently understanding this, and you tell me where I have it wrong? Renata: go on. You: I think the attacks started because of what happened in March, but I do not think that is what is keeping them going. I think what keeps them going is that every time you take the long route, the feeling goes down, and your brain learns that the long route saved you. Renata, after a pause: so it gets worse the more careful I am. You: that is the trap, yes. Renata: my mother has been careful for thirty years.

That last line is not a digression. It is her adding a predisposing factor you had written thinly and she has just stated precisely, and your grid should change to match. Collaborative formulation is not a courtesy. It is a second source of data.

The rival formulation, and how you would tell

Now the discipline. A formulation that cannot be wrong is not a hypothesis, so write the strongest competitor you can and say what would settle it.

Competing account. This is not primarily an anxiety disorder maintained by avoidance. It is a guilt reaction. Renata believes she failed a colleague, has told nobody, and her symptoms are the intrusive re-experiencing and physiological arousal that follow a moral injury. The avoidance is of aisle nine specifically, and the highway avoidance is a spread from that. On this account, exposure to driving would produce compliance and no improvement, and the treatment target is the untold story, not the hierarchy.

What distinguishes them? Three observations. Does she have attacks in situations with no link to the accident, such as at home watching television? She does, which supports the first account. Do the intrusive images centre on Danny's foot, or is the fear about her own body? Ask, because you have not. And if you run two weeks of highway exposure and her panic frequency falls while the guilt is untouched, you have learned that both accounts hold some truth, which is the usual result.

Remember: Write the formulation so that a specific future observation could falsify it, and then look for that observation instead of collecting evidence you already agree with.

Common misconceptions

• A formulation is a summary of the assessment. A summary restates facts. A formulation claims that some facts are causing others, and can therefore be wrong.
• The predisposing cell is the important one. It explains vulnerability and offers nothing to do on Thursday. The perpetuating cell is where treatment lives.
• Formulation is what you do instead of diagnosis. They answer different questions. The diagnosis names the pattern and gets it paid for; the formulation says what to change.
• A formulation is written by the clinician and delivered to the client. Saying it aloud and inviting correction routinely produces the best line in it, as Renata's remark about her mother did.
• If the formulation is good, treatment follows automatically. It follows only if the perpetuating factors were written specifically enough to act on. Poor coping is not actionable; three or four beers after a night shift is.

Pulling it together

• A formulation answers why this person, this way, now, and what is maintaining it. A diagnosis cannot answer any of that.
• The five cells are presenting, predisposing, precipitating, perpetuating, and protective, and the presenting row is stated as behaviour and frequency rather than as a label.
• Perpetuating factors are the treatment targets because they are current and modifiable. A thin perpetuating cell means you have no plan yet.
• Turn each perpetuating factor into an intervention and an order of operations, and expect some of them to be referrals rather than therapy.
• The usable version is four sentences: vulnerability, trigger, maintenance, resources.
• Share it, invite correction, write a rival account, and name the observation that would decide between them.

Sources

1. Weerasekera, P. (1996). Multiperspective case formulation: A step towards treatment integration. Krieger Publishing.
2. Johnstone, L., and Dallos, R. (Eds.). (2014). Formulation in psychology and psychotherapy: Making sense of people's problems (2nd ed.). Routledge.
3. Persons, J. B. (2008). The case formulation approach to cognitive-behavior therapy. Guilford Press.
4. Wikipedia contributors. (n.d.). Clinical formulation. Wikipedia. en.wikipedia.org
5. National Institute of Mental Health. (n.d.). Psychotherapies. NIMH. nimh.nih.gov

A coffee stirrer, sixty seconds

In session three you take a plastic coffee stirrer out of the desk drawer and ask Renata Marchetti to pinch her nose shut and breathe through it for sixty seconds. She looks at you as if you have lost your mind, and does it.

At about thirty seconds her heart rate climbs, her chest tightens, and she gets the light-headedness that always precedes an attack. At fifty seconds she pulls the stirrer out, breathing hard, and says: that is it, that is exactly it.

Then you both sit for two minutes and the sensations fade on their own. Nothing else happens. No emergency department, no cardiac event, nothing. And now the two of you have something you did not have twenty minutes ago: a shared experience in which the sensations she believes signal death were produced deliberately, in a small office, by a straw.

That is interoceptive exposure, and it is the first genuinely therapeutic thing that has happened in this case. This lesson follows the treatment from here to the end.

The model, on her material

Cognitive behavioural therapy rests on a claim simple enough to state in one sentence: what a situation means to you drives what you feel and what you do, and what you do then feeds back into what the situation seems to mean.

Do not teach that as a general principle. Teach it on the March attack, in her words, drawn out on paper between you.

When you draw the arrow from the last row back up to the first, she sees it herself. The thing that made her feel better is the thing that keeps it happening.

Key idea: The cognitive model is not a lecture about thinking. It is a loop drawn on paper from one of the client's own episodes, with the arrow that closes it drawn last.

The thought record, worked

A thought record is the workhorse. Renata's, from week four, after a bad Tuesday at the supermarket.

Three things about this. The evidence-against column is where the work is, and it is empirical, not encouraging: twenty-five attacks and zero collapses is a number, and cheering her up would have been useless. The alternative thought is believed at 60 percent, not 100, and that is the honest outcome of a first attempt; a thought record that ends at total conviction has usually been filled in to please you. And the emotion drops without going away. Panic 80 to 45 is a real result.

Now a warning. Cognitive distortions are a useful vocabulary and a dangerous habit. Catastrophizing, mind reading, all-or-nothing thinking, and emotional reasoning are real patterns, and naming them helps a client notice a recurring shape. But some client thoughts are true. If Renata thinks her supervisor is looking for a reason to move her off shift lead, and her supervisor is, then labelling that catastrophizing is a clinical error that damages the alliance and teaches her you do not believe her. Check the evidence before you name the distortion. In social work particularly, where clients face genuine institutional risk, the thought under examination is often accurate and the problem is what to do about it.

Behavioural activation: outside in

Renata's low mood needs a different tool, and the tool is deceptively plain. Behavioural activation starts from the observation that depression shrinks a person's world through a loop of its own: feel flat, do less, get less reward and less mastery, feel flatter.

The counterintuitive rule is that you act first and feel later. Waiting to feel like doing something is waiting for the thing the depression removed.

Week one: monitor, do not change. Renata records what she does hour by hour for a week, rating each block for pleasure and for a sense of accomplishment, each 0 to 10. The record is not a diary; it is data. Hers shows two things she did not know. Her flattest hours are 3 pm to 6 pm on days off, when Mateo is at school and she is alone in the house. And every entry she rated above 6 for accomplishment involves either her son or her mother.

Week two: schedule, small and specific. Not be more active. Three items, written into slots.

• Tuesday 3.30 pm, walk to the end of Marbury Street and back, roughly twelve minutes.
• Thursday 4 pm, text Carla from work, one message, no requirement that she answer.
• Saturday 10 am, take Mateo to the first half of his game, drive the town route, leave at half time if needed.

Each is time-bounded, small enough to be nearly certain, and specified so precisely that she cannot debate at 3.30 whether it counts. The Saturday item does double duty: it is activation, and it is the first rung of the exposure hierarchy.

The evidence for this deserves a sentence. Neil Jacobson and colleagues took cognitive therapy for depression apart in 1996 and found that the behavioural activation component alone performed as well as the full package including cognitive restructuring, a result the COBRA trial in the previous lesson later reproduced at scale. The simplest part of CBT is doing a large share of the work.

The hierarchy

Exposure is the treatment for avoidance, and it works when it is graded, repeated, prolonged enough, and stripped of safety behaviours. Renata builds the list herself and rates each item 0 to 100 on the subjective units of distress scale.

Safety behaviours are the part that gets missed, and they are why exposure sometimes runs for weeks with no benefit. Renata's are specific: she holds the wheel at the bottom so she can see her hands, she keeps a bottle of water in the door and sips at every set of lights, she plans her route past two hospitals, and she calls Vince before setting off. Each one is small. Together they let her complete the drive and conclude that she survived because of them.

The mechanism explains why. The older account was habituation: stay in the situation long enough and arousal declines. Michelle Craske and colleagues argued in 2014 that the better account is inhibitory learning, and its engine is expectancy violation. The point is not to feel calm. The point is to find out that the predicted catastrophe does not occur, and that discovery requires her to make the prediction explicit first.

So step six is run like this. Before the drive: what do you think will happen? She writes: I will have an attack on the ramp, lose control of the car, and have to pull over. How likely, 0 to 100? Ninety. After the drive: what actually happened? She writes: heart went to maybe 70 out of 100 on the ramp, no attack, did not pull over, hands stayed on the wheel. And then the sentence that matters: what does that tell you about the prediction?

So what?: An exposure without a stated prediction is a person enduring something. An exposure with a stated prediction is an experiment, and only the experiment changes belief.

When the homework is not done

She comes to week six having done none of it. The reflex is to encourage her, and the reflex is wrong. Homework is not done for reasons, and the reason is data.

You ask, without any edge in it, what got in the way. She says the supermarket step felt stupid, and then, after a silence, that her mother had a bad week and she spent Tuesday at the pulmonary clinic. Two entirely different problems. The first is a task agreement failure, and the fix is renegotiation: the step was too big or the rationale did not land. The second is a real-world constraint that no amount of motivation would have solved, and the fix is a smaller step that fits inside a caregiving week.

Note what you did not do: you did not treat the missing homework as resistance. Non-completion in outpatient work is usually a step that was too large, a rationale the client did not accept, or a life that did not have room. All three are fixable, and none of them is a character trait.

Twelve sessions, and what actually happened

Session one to two: assessment, formulation, shared model. Three: the stirrer, and psychoeducation on the physiology of a panic attack. Four to five: thought records, and the first two hierarchy steps. Six: the missed homework, and the renegotiation. Seven to nine: steps three to six, with predictions written before and after, plus activity scheduling running alongside. By session nine she has driven the on-ramp four times and her panic frequency is down from five a month to one. Ten: the alcohol conversation, which is the subject of a later lesson and does not go the way you expect. Eleven: step seven, the full route, twice. Twelve: relapse prevention, which means writing down what she now knows in her own handwriting, listing the early warning signs of slipping back, and planning the two situations she is most likely to start avoiding again.

Her PHQ-9 went from 18 at intake to 8 at session twelve. Her panic frequency went from five a month to one. Her sleep study came back positive for moderate obstructive sleep apnea and she now has a machine she uses about four nights a week. Her drinking is unchanged.

That last sentence matters. This is what a good course of treatment looks like: substantially better, not finished, with one target untouched and honestly recorded.

Common misconceptions

• CBT means challenging irrational thoughts. Much of the work is behavioural, and some thoughts under examination turn out to be accurate, in which case the task is problem-solving, not disputation.
• Exposure works by making anxiety go away during the session. The mechanism is expectancy violation: finding out the predicted disaster does not happen. Comfort during exposure is neither the goal nor the measure.
• Safety behaviours are harmless coping. They let a client complete an exposure while concluding they survived because of the water bottle, which is exactly the learning exposure is meant to prevent.
• Behavioural activation means encouraging the client to be more active. It means one week of monitoring with pleasure and mastery ratings, followed by three specified tasks in named time slots.
• Undone homework is resistance. It is nearly always a step that was too big, a rationale that did not land, or a week that had no room in it.

What you now know

• Teach the model on the client's own episode and draw the feedback arrow last, so the client sees the loop rather than being told about it.
• A thought record lives or dies in the evidence-against column, which should contain counts and facts rather than reassurance.
• Name a cognitive distortion only after checking the evidence, because clients in social work settings often face the risks they are describing.
• Behavioural activation runs monitoring first, then small, specified, time-slotted tasks; the component analyses suggest it does much of CBT's work by itself.
• Exposure must be graded, repeated, and stripped of safety behaviours, and each step is run as an experiment with a written prediction before and a written result after.
• A realistic outcome is substantial improvement with something left undone and recorded honestly.

Sources

1. Beck, J. S. (2020). Cognitive behavior therapy: Basics and beyond (3rd ed.). Guilford Press.
2. Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., and Vervliet, B. (2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, 58, 10-23. PubMed Central
3. Jacobson, N. S., Dobson, K. S., Truax, P. A., Addis, M. E., Koerner, K., Gollan, J. K., et al. (1996). A component analysis of cognitive-behavioral treatment for depression. Journal of Consulting and Clinical Psychology, 64(2), 295-304.
4. Wikipedia contributors. (n.d.). Exposure therapy. Wikipedia. en.wikipedia.org

Four minutes that would pass a fidelity check

Read this exchange. It is from session five with Nadine Ricks, 44, referred for depression after a workplace injury ended her job stacking pallets. Her landlord filed for eviction eleven days ago.

Worker: so the thought was, I am going to lose my apartment. How much do you believe that, out of a hundred?
Nadine: a hundred.
Worker: all right. What is the evidence for it?
Nadine: the papers came. I owe two thousand eight hundred dollars.
Worker: and what is the evidence against it?
Nadine, after a long pause: I do not have any.
Worker: well, is it possible that things could work out differently than you expect?
Nadine: I suppose anything is possible.
Worker: so a more balanced thought might be, I am worried about my housing and there may be options I have not considered. How much do you believe that?
Nadine: sure. Maybe forty.

A supervisor scoring this against a competence scale would tick most of the boxes. Agenda set, thought elicited, belief rated, evidence examined both ways, alternative generated, re-rated. Fidelity is fine. The session is a disaster, and Nadine cancelled the next two.

This lesson traces four ways a session like that fails, in each case from the specific move that caused it.

Failure one: the thought was true

Nadine believes she is going to lose her apartment at a hundred percent. She owes 2,800 dollars, the filing is real, and in her county the median time from filing to a possession order is about six weeks. Her belief is not a distortion. It is an accurate reading of her situation, arrived at by someone who has read the paperwork more carefully than her worker has.

What the worker just did, from Nadine's chair, is ask her to lower her confidence in a true belief in order to feel better. There is a word for the experience of being persuaded that an accurate perception is a symptom, and it is not therapeutic.

The trace is precise. The error is not the technique; it is applying it before establishing whether the appraisal was accurate. Cognitive work is for appraisals that overshoot the evidence. When the evidence supports the appraisal, the indicated intervention is not restructuring. It is problem-solving, resource mobilization, and advocacy: the legal aid clinic's number, the county's emergency rental assistance application, a letter about her injury for the hearing, and a conversation about what she wants if the apartment cannot be saved.

This failure is a professional hazard for social workers specifically, because our clients face real institutional threats at a much higher rate than the outpatient samples in which these techniques were developed. Before you dispute a thought, do the arithmetic in it.

What matters here: Check the accuracy of the appraisal before you challenge it. If the client is right, the intervention is material, not cognitive.

Failure two: the protocol was matched to the diagnosis, not the mechanism

A second worker takes a different case: Wesley Obuya, 31, who has stopped leaving his building after dark. Diagnosis, panic disorder with agoraphobia. Indicated treatment, graded exposure. The hierarchy is built and the first three steps are set.

Except that Wesley was beaten at the bus stop on his own street eleven months ago and the man who did it lives four doors down and has never been charged. His avoidance is not a maintaining mechanism producing a false belief. It is a broadly accurate risk assessment, and exposing him to it repeatedly is not treatment, it is a professional asking him to walk past a threat because a protocol said so.

The trace: nobody wrote a formulation. The diagnosis was matched to a manual and the manual to a hierarchy, and the step from category to intervention was made without asking what was maintaining the behaviour in this particular person. The previous lesson said the perpetuating cell is where treatment lives; this is what it looks like when that cell is skipped.

What Wesley needs is safety planning, possibly relocation help, contact with the prosecutor's office, and then, once the real danger is addressed, work on the fear that will have generalized well beyond that one street. There is real trauma work to do here. It just cannot start with a hierarchy that treats a genuine threat as a false alarm.

Failure three: the exposure happened and the learning did not

A third case runs cleanly on paper. Eight weeks of exposures, all completed, all logged, distress ratings falling nicely within each session. And at week eight the client is no less afraid than at week one.

Here the debugging is mechanical. Ask exactly how the exposure was performed, minute by minute, and one of four things is usually true.

• Safety behaviours were running. The water bottle, the phone in hand with the number already dialled, the partner waiting in the car outside.
• Attention was elsewhere. The client did the supermarket queue with headphones on, counting backwards from a hundred. The body was present and the person was not.
• No prediction was stated. Without a written expectation there is nothing for the outcome to violate, so the client accumulates endurance rather than evidence.
• The exposure was too short or too spaced out. Two minutes once a fortnight teaches very little.

Note that in all four the client did what was asked. Calling this non-response would be wrong; it is a delivery failure, and the person who should have asked how the exposure was actually performed is the clinician.

Failure four: technically correct, relationally finished

Read the Nadine transcript one more time, ignoring the content. Count the questions. Count how many of them she could answer with anything other than a concession. Notice that when she says she has no evidence against her belief, the worker does not pause on what that must feel like, but moves straight to is it possible that things could work out differently, which is a rhetorical question dressed as an open one.

Socratic questioning is meant to be joint enquiry into something neither party has settled. Done badly it is cross-examination in which the clinician already knows the verdict and the client's job is to arrive at it. Clients experience the difference immediately, and the tell is that they start giving short, agreeable answers, which is a withdrawal rupture from the first lesson of this course.

Compare a version of the same four minutes.

Worker: you believe you will lose the apartment, and from what you have told me about the filing and the amount, I think you are probably right.
Nadine: thank you. Yes.
Worker: so I do not want to spend our time trying to talk you out of it. I want to spend it on two things: what can still be done about the housing, and how you are going to get through the next six weeks without the sleep and the drinking getting worse. Which one first?
Nadine: the housing. Obviously the housing.

That version has no cognitive technique in it at all and it is better therapy, because it is accurate and it puts the client's priority first. The techniques come back later, applied to the parts of her thinking that do overshoot: that she is a burden to her sister, that there would be no point applying for anything because people like her never get it.

The word resistance, and what it hides

When treatment stalls, the word that arrives is resistance, and it is almost always a description of the clinician's frustration rather than of the client. Nearly every time it is used, one of the following is true and more useful.

Glenn Waller made a related argument from the other direction in 2009, calling it therapist drift: clinicians trained in exposure-based methods routinely fail to deliver them, substituting talking about the problem because the exposure is uncomfortable for the clinician. Both errors are worth holding at once. Sometimes the protocol is being applied where it does not belong, and sometimes it is being quietly avoided where it does.

Catching it early

Three signals, in order of reliability.

1. The measure has not moved by session six. This is why the previous module insisted on routine measurement. A PHQ-9 flat across six sessions is a fact, not an impression, and it is the earliest honest signal you get.
2. Homework has not been done twice running. Once is a week. Twice is a message about the task, the rationale, or the alliance.
3. The client agrees with everything. Complete agreement in a therapy about examining beliefs is close to definitional evidence that no examining is happening.

What follows is a fixed sequence, and none of its steps is trying harder. Re-formulate, with the client, out loud. Ask directly what they think is not working. Change something structural: the target, the modality, the frequency. Take it to supervision, with a recording or a transcript if you can, because the failures in this lesson are all easier for someone else to see. And if none of that moves it, refer, which is a clinical decision rather than an admission.

One more fact belongs here, plainly. Psychotherapy has a deterioration rate. Across outcome studies a minority of adult clients, commonly estimated at roughly one in twenty to one in ten, end treatment worse than they started. Treatment is not neutral when it fails, and a clinician who has never considered that they might be making someone worse is not being careful.

Common misconceptions

• If the technique was delivered correctly, the session was good. Fidelity scores measure whether the moves were made, not whether they were the right moves for this person's situation.
• Cognitive restructuring is for any distressing thought. It is for appraisals that overshoot the evidence. For accurate appraisals of real threats, the indicated response is problem-solving and advocacy.
• A client who does not improve is resistant. Resistance names the clinician's frustration. Underneath it there is nearly always a task, a rationale, a schedule, or a rupture that can be identified and changed.
• Avoidance always needs exposure. Avoidance of a genuine danger is a risk assessment, and exposing someone to a real threat is not treatment.
• Therapy either helps or does nothing. A minority of clients deteriorate during treatment, which is why measurement and supervision are safety practices rather than paperwork.

The takeaway

• Check whether the appraisal is accurate before challenging it; if the client is right, the intervention is material rather than cognitive.
• A protocol chosen from a diagnosis rather than from a formulation will sometimes ask a client to expose themselves to a real danger.
• When exposures are completed and nothing improves, interrogate how they were performed: safety behaviours, distraction, no stated prediction, too short or too spaced.
• Socratic questioning that the clinician has already answered reads as cross-examination and produces short agreeable answers.
• Resistance is a placeholder; replace it with the specific task, rationale, schedule, or rupture involved. Therapist drift is the mirror error.
• Watch for a flat measure by session six, two missed homeworks, and complete agreement. Then re-formulate, ask, change something structural, take it to supervision, and refer if needed.

Sources

1. Waller, G. (2009). Evidence-based treatment and therapist drift. Behaviour Research and Therapy, 47(2), 119-127.
2. Beck, J. S. (2020). Cognitive behavior therapy: Basics and beyond (3rd ed.). Guilford Press.
3. Lambert, M. J. (2013). The efficacy and effectiveness of psychotherapy. In M. J. Lambert (Ed.), Bergin and Garfield's handbook of psychotherapy and behavior change (6th ed., pp. 169-218). Wiley.
4. Nakao, M., Shirotsuki, K., and Sugaya, N. (2021). Cognitive behavior therapy. StatPearls. NCBI Bookshelf
5. Wikipedia contributors. (n.d.). Socratic questioning. Wikipedia. en.wikipedia.org

Session ten, take one

This is the alcohol conversation with Renata Marchetti, and it is going badly in a way that will look familiar.

Worker: I wanted to come back to the drinking. Three or four beers most nights is quite a lot, and with the sleep apnea it is going to make your sleep worse, not better.
Renata: it is two, usually.
Worker: you told me three or four in the assessment.
Renata: on a bad week.
Worker: right, but alcohol fragments sleep, and given everything we have been working on with the panic, cutting back would probably help a lot.
Renata: my father was an alcoholic. I am not an alcoholic. I go to work every day.
Worker: nobody is saying you are an alcoholic. I am just saying it might be worth looking at.
Renata: everybody at that warehouse drinks more than me. It is the one thing I do for myself all day.
Worker: I understand, it is just that the health effects add up over time.
Renata: sure. I will think about it.

Count the moves. The worker made five statements in favour of change. Renata made five statements against it, escalating from a correction to a defence to a justification to the polite dismissal that ends the conversation. And here is the part that matters: at the start of that exchange Renata was somewhat worried about her drinking. At the end she had spent four minutes arguing that it was fine, out loud, in front of a professional. She is now more committed to drinking than she was at the beginning, and the worker did that.

The righting reflex

William Miller and Stephen Rollnick gave this its name. The righting reflex is the trained helper's impulse to put right what is wrong: to see the problem, know the solution, and supply it. In most of a social worker's day it is a virtue. In a conversation about behaviour change it reliably backfires, and the mechanism is well understood.

Ambivalence has two sides, and a person holding both will voice whichever one is not currently being voiced by the other party. Argue for change and you have taken one side of an internal debate, leaving the client only the other. People persuade themselves by hearing themselves talk, so the client who spends four minutes defending their drinking leaves more committed to it. That is reactance operating exactly as the social psychology predicts.

Miller and Rose set out the theory in 2009 and the empirical claim underneath it is specific: the client's own change talk during a session predicts subsequent behaviour change, and sustain talk predicts its absence. That turns the conversation into something you can steer. Your job is not to supply arguments for change. It is to make it more likely that the client makes them.

In short: Whoever voices the arguments for change is the one being persuaded. Make sure it is not you.

The spirit, which is not optional decoration

Miller and Rollnick's third edition, in 2013, spends its first chapters on what they call the spirit of motivational interviewing, and clinicians who skip to the techniques produce something that sounds like the method and functions as manipulation.

• Partnership. You are working alongside the person, not operating on them. They are the expert on their own life and the only source of the motivation you are looking for.
• Acceptance. Absolute worth, accurate empathy, autonomy support, and affirmation. Autonomy support means saying, and meaning, that the decision is theirs, including the decision to change nothing.
• Compassion. The work is done in the client's interest. This clause exists because the techniques work, and techniques that work can be used to sell things.
• Evocation. The motivation is already in there. You are drawing it out, not installing it.

The third edition also lays out four processes that run in rough order and loop: engaging, focusing, evoking, and planning. The commonest error in practice is jumping to planning while the client is still in engaging, which produces exactly the transcript above.

OARS

The core skills are four, and their acronym is easy enough that people learn it and then never practise the hard one.

Reflection is the skill that separates practitioners, and the difference between simple and complex reflection is the difference between a mirror and a step forward. A simple reflection repeats content. A complex reflection makes a guess about meaning, feeling, or the unspoken half of the sentence. Guessing wrong is not a problem; a client correcting your reflection is giving you better information than any question would have.

Session ten, take two

Same client, same topic, different conversation. The annotations name the move.

Worker: would it be all right if we spent a few minutes on the drinking? You can tell me to leave it. [asking permission, autonomy support]
Renata: fine.
Worker: what is the beer doing for you at the end of a shift? [open question, no direction of travel implied]
Renata: it stops the noise. My head goes quiet for about an hour.
Worker: it is the one part of the day where nothing is being asked of you. [complex reflection]
Renata: exactly. And then at four in the morning I am awake again anyway.
Worker: so it buys the quiet hour and it does not buy the night. [reflection of her own change talk, said back]
Renata: no. It does not.
Worker: what else have you noticed about it? [open question, evoking more]
Renata: I do not like that Mateo has seen me with a beer in my hand every night. That is what my father looked like to me.
Worker: that one matters to you. [reflection, staying still rather than pouncing]
Renata, after a pause: yes.
Worker: on a scale of nought to ten, how important is it to you to change something about the drinking? [importance ruler]
Renata: six.
Worker: six. Why not a three? [the crucial direction: asking why the number is not lower]
Renata: because of what I just said. And because I am not sleeping and I am tired of being tired.
Worker: so the reasons are your son, and the sleep. What would a change look like, if you decided to make one? [summary, then an open question that leaves the decision with her]

The worker has advanced no arguments. Every reason for change in that transcript came out of Renata's mouth, which is the entire point of the method.

Notice one move in particular. Asking why not a three rather than why not an eight is not a trick; it is the whole difference. Asking why the number is not higher invites the client to list the obstacles, which is sustain talk. Asking why it is not lower invites them to argue for change.

Change talk, named

Miller and Rollnick sort the client's own language about change into two stages, and the standard mnemonic is DARN-CAT.

Preparatory talk is the fuel and mobilizing talk is the ignition. A client can produce desire, ability, reasons, and need for months without moving, and the practitioner's job in the evoking process is to notice mobilizing language when it first appears and to respond to it rather than talking past it. When Renata says I did not buy any on Sunday, the move is to ask how she managed it, not to congratulate her and change the subject.

When it goes sideways

The third edition retired the phrase rolling with resistance, and the retirement was a clarification worth having. What used to be called resistance is two different things.

Sustain talk is the client's own voice for the status quo. It is a normal component of ambivalence, it is not a problem, and it is usually a sign that you got ahead of them. Discord is trouble in the relationship: arguing, interrupting, dismissing, disengaging. Discord is a signal about you, not about the client's motivation.

The responses differ. To sustain talk: reflect it, reflect it amplified if you can do it without sarcasm, and come back to the other side of the ambivalence. To discord: stop, apologize if warranted, emphasize autonomy explicitly, and shift back to engaging. In take one above, Renata's line about her father was discord produced by the worker's third statement, and the worker responded by continuing to argue.

What the evidence supports

Be accurate about this. Motivational interviewing has a large trial literature, mostly in substance use, health behaviour, and treatment adherence. Meta-analytic estimates put its effects as modest but real when compared with no treatment or with advice, and roughly equivalent to other active treatments when compared head to head. It tends to do well in brief formats, sometimes in a single session, which is why it spread through emergency departments and primary care. Its effects can fade without follow-up, and it works best when the target is a behaviour the person is ambivalent about rather than one they have not considered at all.

It also has a specific misuse. Because the method is effective at moving people towards a stated behaviour, it is periodically taught to staff as a way of improving compliance with what the agency wants. That is a violation of the compassion clause and it is detectable: if you would be unwilling to say out loud what outcome you are steering towards, you are not doing motivational interviewing.

Common misconceptions

• Motivational interviewing is a way of getting people to do what you want. Its own definition rules that out. It is a way of helping a person resolve ambivalence in the direction they choose, including away from change.
• Reflection means repeating what the client said. That is simple reflection, which has limited uses. Complex reflection guesses at the unspoken half, and being corrected is a good outcome.
• Resistance is a client trait. The third edition splits it into sustain talk, which is normal ambivalence, and discord, which is a signal about the relationship and usually about something you just did.
• The rulers are used to measure motivation. They are used to evoke change talk, which is why the question is always why not a lower number.
• Motivational interviewing replaces other treatment. It is most often a way into treatment, or a way through a stuck patch inside one, and its effects fade without follow-up.

Summing up

• The righting reflex hands the client the other side of their own ambivalence, and hearing themselves defend the status quo strengthens it.
• Change talk predicts change and sustain talk predicts its absence, which makes evoking the client's own arguments the operative skill.
• The spirit is partnership, acceptance, compassion, and evocation; the four processes are engaging, focusing, evoking, and planning, and most failures are planning attempted during engaging.
• OARS is open questions, affirmations, reflections, and summaries, and complex reflection is the skill that separates practitioners.
• DARN preparatory talk is the fuel; CAT mobilizing talk is the ignition, and it must be noticed when it first appears.
• Sustain talk is normal and gets reflected. Discord is about the relationship and gets an explicit return to autonomy and engagement.

Sources

1. Miller, W. R., and Rollnick, S. (2013). Motivational interviewing: Helping people change (3rd ed.). Guilford Press.
2. Miller, W. R., and Rose, G. S. (2009). Toward a theory of motivational interviewing. American Psychologist, 64(6), 527-537. PubMed Central
3. Lundahl, B. W., Kunz, C., Brownell, C., Tollefson, D., and Burke, B. L. (2010). A meta-analysis of motivational interviewing: Twenty-five years of empirical studies. Research on Social Work Practice, 20(2), 137-160.
4. Wikipedia contributors. (n.d.). Motivational interviewing. Wikipedia. en.wikipedia.org

A line from Alice, quoted in 1936

In 1936 Saul Rosenzweig published four pages in the American Journal of Orthopsychiatry arguing that the competing schools of psychotherapy, then mostly variants of psychoanalysis, were probably working through the same underlying processes rather than through the ones they each claimed. He borrowed a line from the Dodo bird in Alice's Adventures in Wonderland to summarize it: everybody has won and all must have prizes.

Ninety years later the argument is still running, with far better data on both sides, and it is not academic. It determines whether an agency should spend 40,000 dollars training eight staff in a branded protocol or spend it on smaller caseloads and better supervision.

This lesson sets out both positions at full strength. You are not going to be handed a verdict.

The two claims, stated precisely

The case for specific ingredients

Start with the strongest examples rather than the average, because the average is where this position looks weakest and the specific cases are where it looks strongest.

For obsessive-compulsive disorder, exposure with response prevention repeatedly outperforms other credible psychotherapies, including therapies delivered with equal warmth, equal structure, and equal therapist conviction. For specific phobia, a few hours of in vivo exposure produces change that decades of supportive discussion does not. For post-traumatic stress disorder, trauma-focused treatments outperform present-centred supportive counselling in a body of trials large enough that guideline panels in several countries reached the same conclusion independently. These are not small differences and they do not disappear when the comparison is a real treatment delivered by people who believe in it.

The position's second argument is about mechanism. In panic disorder, the improvement tracks the change in catastrophic misinterpretation of bodily sensations, which is the thing the treatment targets. When a treatment moves the variable it says it moves, and outcome follows that variable, the burden shifts to the person claiming it is all shared context.

The third argument is practical. If specificity does not exist, then a clinician's choice of what to do is arbitrary, and the profession has no basis for saying that anything is malpractice. Nobody actually believes that. A clinician who treats obsessive-compulsive disorder with unstructured supportive listening for two years is doing something wrong, and the concept that makes it wrong is specificity.

The case for common factors

Bruce Wampold and Zac Imel's contextual model does not deny those examples. It denies that they generalize, and its evidence is quantitative.

Direct comparisons. When two treatments that both intend to help, are both delivered by people trained in them, and both give the client a credible rationale, the aggregate difference between them across the literature is very small. Wampold and Imel put the effect of treatment type at under one percent of outcome variance.

Allegiance. A large share of the difference that does appear tracks who ran the trial. Researchers testing their own treatment find bigger effects for it than independent teams do. That is not fraud; it is a pattern that appears wherever allegiance is measured, and it eats much of the remaining gap.

Therapist effects. The variance attributable to which clinician a client happens to see is estimated at around five percent, which on Wampold's arithmetic is several times larger than the variance attributable to which treatment they receive. If that is right, an agency's most consequential decision is who it hires and how it supervises them, not which manual it buys.

Adherence and competence. The specificity claim predicts that clinicians who deliver a protocol more faithfully and more skilfully will get better results. Christian Webb, Robert DeRubeis, and Jacques Barber pooled the studies in 2010 and found the associations between adherence and outcome, and between rated competence and outcome, close to zero. That is a hard result for the ingredients position, and it is not explained away easily.

Where the argument actually turns

Almost every disagreement about a particular trial comes down to one question: what was the treatment compared with?

Wampold's methodological demand is that a comparison be between bona fide treatments: both intended to be therapeutic, both delivered by trained practitioners who believe in them, both offering the client a coherent explanation and a set of things to do. Compare a protocol against a control designed to be inert, and you learn that doing something structured beats being visited, which nobody disputed.

The other side has an answer, and it is a good one. Some so-called bona fide comparators are constructed by researchers who do not believe in them, delivered by staff trained for a fortnight, in order to control for attention. Calling that a fair fight favours the common-factors conclusion by construction. And when the comparator is genuinely strong, as present-centred therapy is in the trauma literature, trauma-focused treatments still tend to win.

Notice what has happened. Both sides now agree that the question is empirical and that it hinges on the comparison condition. That is progress, and it is why the honest answer is that specificity appears real for some conditions and small or absent for others.

Renata, read twice

Take the case from Module 3. Her panic frequency fell from five a month to one across twelve sessions.

Ingredients reading. The interoceptive test and the graded hierarchy did it, and the mechanism is visible: her belief that the sensations would kill her fell from 85 percent to about 20, and her attacks fell with it. Give her twelve sessions of warm, structured, credible conversation without exposure and she would still be driving the town route.

Common-factors reading. She arrived from a physician she trusts, was given a rationale that made sense of five months of terror, worked with someone who plainly believed the treatment would help, was set tasks that gave her mastery experiences, and had a relationship in which she could admit she had not done the homework. Any coherent treatment supplying those would have produced most of this.

Both readings fit. What would distinguish them is a comparison she did not receive, which is precisely why individual cases cannot settle this and why clinicians who reason from their own caseloads reliably confirm whatever they already believe.

What would move it

• Dismantling trials with strong controls. Remove the claimed active ingredient and keep everything else, including therapist conviction. Jacobson's 1996 depression study is the model, and its answer for depression favoured the common-factors side.
• Moderator analyses. Stop asking whether treatment differences exist on average and ask where. Present evidence suggests specificity is larger for obsessive-compulsive disorder, phobias, and trauma than for depression.
• Mechanism measurement. Measure the proposed mechanism session by session and test whether it changes before the outcome does.
• Designs that cross therapists with treatments. The only way to separate therapist effects from treatment effects is to have the same clinicians deliver both arms.

What to do on Monday

The dispute is unresolved and you have a caseload. Four things follow that both camps would sign.

1. Where specificity evidence is strong, deliver the specific treatment. For obsessive-compulsive disorder, phobias, and trauma, this is not a close call.
2. Attend to the alliance always, because both positions agree it predicts outcome, and they disagree only about whether it is the active agent or a condition for one.
3. Give a rationale the client finds credible, and check that they find it credible rather than assuming it. Every model on both sides requires this.
4. Measure outcomes and treat your own results as data about you. If therapist effects are real, then the most useful comparison available to you is between your cases and your other cases.

Worth holding on to: The argument is about what proportion of the work the technique does. Nobody credible claims it is all of it, and nobody credible claims it is none of it.

Common misconceptions

• The dodo bird verdict says all therapies are equally effective. The defensible version is narrower: bona fide treatments intended to be therapeutic show small aggregate differences. It does not license treating obsessive-compulsive disorder with unstructured listening.
• Common factors means being warm and supportive. Wampold's contextual model includes a coherent explanation the client accepts and a set of health-promoting actions, both of which look a lot like technique.
• Trial evidence settles which treatment is better. Which treatment wins depends heavily on what it was compared against and on who ran the study, which is why allegiance is measured at all.
• If common factors dominate, training does not matter. Therapist effects being large is an argument for more selection, supervision, and feedback, not less.
• My own caseload tells me which side is right. Both readings fit any successful case, because the missing comparison is the thing you did not do.

Looking back

• Rosenzweig raised the question in 1936 and the modern version, argued by Wampold and Imel, rests on comparison trials, allegiance, therapist effects, and adherence research.
• The specific-ingredients case is strongest in obsessive-compulsive disorder, specific phobia, and trauma, where structured treatments beat credible active comparators.
• Wampold and Imel put treatment type at under one percent of outcome variance and therapist effects at around five percent.
• Webb, DeRubeis, and Barber found adherence and competence ratings almost unrelated to outcome, which the ingredients position has to answer.
• Most disputes about a trial reduce to whether the comparator was a bona fide treatment delivered by people who believed in it.
• On Monday: use the specific treatment where specificity evidence is strong, tend the alliance, check that your rationale is credible to the client, and measure your own outcomes.

Sources

1. Wampold, B. E., and Imel, Z. E. (2015). The great psychotherapy debate: The evidence for what makes psychotherapy work (2nd ed.). Routledge.
2. Webb, C. A., DeRubeis, R. J., and Barber, J. P. (2010). Therapist adherence/competence and treatment outcome: A meta-analytic review. Journal of Consulting and Clinical Psychology, 78(2), 200-211. PubMed Central
3. Rosenzweig, S. (1936). Some implicit common factors in diverse methods of psychotherapy. American Journal of Orthopsychiatry, 6(3), 412-415.
4. Luborsky, L., Singer, B., and Luborsky, L. (1975). Comparative studies of psychotherapies: Is it true that everyone has won and all must have prizes? Archives of General Psychiatry, 32(8), 995-1008.
5. Wikipedia contributors. (n.d.). Dodo bird verdict. Wikipedia. en.wikipedia.org

Question 34, on a clipboard, in a waiting room

An intake packet at a community mental health centre runs to eleven pages. Question 34, on page two, asks whether the client has ever been sexually assaulted, with a yes box and a no box. The client fills it in on a clipboard, in the waiting room, with four other people in the chairs, having been handed the clipboard through a gap in a glass screen by a receptionist who did not look up.

That agency would tell you it is trauma-informed. It has a policy, and every member of staff attended a half-day training in March.

Everything wrong with the phrase is in that scene, and so is everything the phrase was invented to fix. This lesson is about the difference.

The definition is narrower than the slogan

The reference document is a 2014 publication from the Substance Abuse and Mental Health Services Administration, and it defines trauma-informed care tightly enough to argue with.

Trauma itself is defined by three Es. An event or set of circumstances; the person's experience of it as physically or emotionally harmful or life threatening; and lasting adverse effects on functioning and wellbeing. The middle term is the one that does the work, and it is why two people in the same car crash may have entirely different outcomes, and why a clinician cannot decide from the outside what counts.

An organization is trauma-informed when it does four things, the four Rs. It realizes how widespread trauma is and what paths to recovery look like. It recognizes signs in clients, families, and staff. It responds by building that knowledge into policies and practices. And it actively resists re-traumatization.

Notice the unit of analysis. It is the organization, not the clinician. A trauma-informed worker inside an untrauma-informed system is a person apologizing for the building.

The six principles, with what each one costs

Now go back to question 34. Which principles does that clipboard break? Safety, since a disclosure is being requested in a public room. Trustworthiness, since nobody has said who reads the form. Choice, since it arrived as a requirement rather than an offer. Collaboration, since the question was chosen by the agency and asked by a form. The half-day training in March did not touch any of it, because none of the failures are about staff knowledge.

Why this matters: Trauma-informed care is a property of how a service is arranged. It is not a stance an individual worker can adopt while the arrangements stay the same.

What trauma-informed is not

The ACE study, used correctly

Between 1995 and 1997, Vincent Felitti, Robert Anda, and colleagues surveyed more than 17,000 adult members of a health plan in San Diego about ten categories of childhood adversity: abuse, neglect, and forms of household dysfunction including a parent with a substance problem or a household member imprisoned. They published in 1998, and the finding was a dose-response relationship: as the count of categories rose, so did the prevalence of depression, substance use, suicide attempts, and, further down the causal chain, heart disease, lung disease, and early death.

It is one of the most influential findings in public health, and it is routinely misused in exactly one way.

The relationship is population level. It says that among people with six or more categories, rates of these outcomes are substantially higher. It does not say what will happen to Mr Ellis in your office, who has seven and is fine, or to Ms Kaur, who has one and is not. The categories are counted, not weighted, so one instance of a parent divorcing counts the same as years of physical abuse. The original sample was mostly white, insured, and middle-aged, which limits generalization. And the resilience literature is clear that the same adversity count produces widely varied outcomes depending on what else was present.

This matters practically because ACE screening has been adopted in some paediatric and primary care settings on the assumption that a score identifies individual risk. Critics have pointed out that a measure with modest individual predictive value can produce false reassurance, unnecessary alarm, and, if the score enters a record, a label attached to a child. Use the finding to argue for prevention and for services. Do not use a number to tell a person what their life will be.

What changes inside the session

Concretely, and none of this requires a manual.

• Say what is coming. I am going to ask about some difficult things in a few minutes. You can tell me to stop at any point and we will stop.
• Ask permission, and mean it. If you ask may I ask about that and proceed regardless of the answer, you have taught the client that your questions are decorative.
• Ask about impact before detail. You need to know what happened well enough to plan. You almost never need the full account in session one, and taking it can leave a client activated with forty minutes to get home.
• Manage the ending deliberately. Leave ten minutes. Ground: feet on the floor, name five things in the room, what are you doing straight after this.
• Give control over the physical setting. Which chair, door open or shut, lights.
• Be predictable. Start on time, end on time, do what you said you would do. For someone whose formative experiences were unpredictable, reliability is not a courtesy. It is the intervention.

Trauma-specific treatments are a separate matter, and they exist: cognitive processing therapy, prolonged exposure, eye movement desensitization and reprocessing, and, for children, trauma-focused cognitive behavioural therapy. Each has a trial base and each requires training and supervised practice. Knowing that they exist and referring appropriately is within the scope of this course. Delivering them is not.

Re-traumatization is usually institutional

The fourth R is the one that separates the concept from good manners, and the practices it targets are mostly structural rather than interpersonal.

A person tells the story of an assault to an emergency clinician, a police officer, an advocate, a caseworker, and then a therapist, five times in nine days, because no system shares its record. A psychiatric patient is restrained by six staff, which for a survivor of assault reproduces the original experience with official sanction. A woman is transported to a hospital in handcuffs because that is the county's procedure for involuntary transport. A child repeats a disclosure four times because each professional needs it in their own words for their own form.

None of those is caused by an unkind worker. All of them are policies, which means all of them can be changed, and changing them is what the phrase was coined to demand.

Common misconceptions

• Trauma-informed care means treating trauma. It means arranging every service so it does not injure people again. The treatments are a separate, trained activity.
• It means never asking about traumatic experiences. Silence is its own message. It means asking in a place, at a pace, and with a stated purpose that the client can decline.
• A high ACE score predicts an individual's outcome. The relationship is population level, the categories are unweighted, and the original cohort was insured and mostly white.
• Our staff had the training, so we are trauma-informed. The unit is the organization. If nothing changed in policy, space, staffing, or supervision, nothing changed.
• Being warm and patient is the substance of it. Warmth is welcome and it is not the mechanism. Predictability, transparency, real choice, and the removal of re-traumatizing procedures are.

Recap

• SAMHSA defines trauma by three Es, event, experience, and effects, and a trauma-informed organization by four Rs: realize, recognize, respond, and resist re-traumatization.
• The six principles are safety, trustworthiness and transparency, peer support, collaboration and mutuality, empowerment with voice and choice, and attention to cultural, historical, and gender issues.
• The unit of analysis is the organization, so a trauma-informed worker in an unchanged system is apologizing for the building.
• The approach is not treatment, not a screening instrument, not avoidance of the subject, and not a completed training module.
• The ACE findings are a population dose-response relationship from a mostly insured, mostly white cohort, and an individual score is not a prediction.
• In session: say what is coming, ask permission and honour the answer, ask impact before detail, leave ten minutes at the end, and be relentlessly predictable.

Sources

1. Substance Abuse and Mental Health Services Administration. (2014). SAMHSA's concept of trauma and guidance for a trauma-informed approach (HHS Publication No. SMA 14-4884). SAMHSA.
2. Substance Abuse and Mental Health Services Administration. (2014). Trauma-informed care in behavioral health services (Treatment Improvement Protocol Series, No. 57). NCBI Bookshelf
3. Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., Koss, M. P., and Marks, J. S. (1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. American Journal of Preventive Medicine, 14(4), 245-258.
4. Centers for Disease Control and Prevention. (n.d.). About adverse childhood experiences. CDC. cdc.gov

Nine minutes left

Tamsin Brooke is 19, in her second semester, and has spent most of the session talking about a failed statistics module. At 4.51, with nine minutes left, she says: honestly some days I think everyone would be better off if I just was not here.

What you do in the next nine minutes is the subject of this lesson, and the first thing to know is that the session is not going to end at five.

Some scale. In the United States more than 49,000 people a year have died by suicide in recent years, and firearms are involved in more than half of those deaths. It is among the leading causes of death for people Tamsin's age. This is common enough that you will meet it, and the procedures below are the ones the field has settled on.

One boundary, stated once. This lesson teaches the structure of an assessment and a plan. It does not qualify you to manage suicide risk alone. That is done under supervision, with a licensed clinician available, inside an agency that has a protocol, and if you are ever in this position without those things, the answer is 988 and an emergency department, not your own judgement.

Ask directly, in ordinary words

The single most persistent myth in this area is that asking about suicide plants the idea. It does not, and this has been tested repeatedly. Asking clearly does something better: it tells a person that the subject is speakable here, which for someone who has been carrying it silently is a substantial event.

Ask plainly. Are you having thoughts of killing yourself? Not are you thinking of hurting yourself, which is a different question and captures self-injury that may have nothing to do with suicide. Not you are not thinking of doing anything silly, are you, which tells the client the answer you want.

Then follow the structure that the Columbia-Suicide Severity Rating Scale formalizes. Its screening version, often called the Columbia Protocol, runs six questions in an order that escalates and stops when it can.

1. Have you wished you were dead or wished you could go to sleep and not wake up?
2. Have you actually had any thoughts of killing yourself?
3. Have you been thinking about how you might do this?
4. Have you had these thoughts and had some intention of acting on them?
5. Have you started to work out or worked out the details of how to kill yourself, and do you intend to carry it out?
6. Have you done anything, started to do anything, or prepared to do anything to end your life?

Questions one and two screen. If both are no, you are done with this section. Three, four, and five separate a passing thought from a plan with intent, which is the distinction that changes what happens next. Six asks about behaviour, including preparatory acts such as giving possessions away, writing a note, or buying something, and it carries its own timeframe: ever, and in the last three months.

Tamsin answers yes to one, yes to two, no to three. She has thought about not waking up, has had thoughts of killing herself perhaps four times in the past month, has not thought about a method, has no intention, and has done nothing. That is a meaningful clinical finding and it is not an emergency.

Prediction fails; formulation does not

Before going further, know what the assessment can do. Joseph Franklin and colleagues published a meta-analysis in 2017 covering fifty years of research on risk factors for suicidal thoughts and behaviours, and their conclusion was blunt: across hundreds of studies, the ability of risk factors to predict who would attempt or die by suicide was only slightly better than chance, and it had not improved over five decades.

That finding is not a reason to stop assessing. It is a reason to stop pretending that assessment sorts people into high, medium, and low risk with any accuracy. What replaces prediction is risk formulation, which asks four different questions.

• Status. How does this person compare with a relevant population? A 19-year-old woman with ideation and no plan sits in a different band from a 58-year-old man with a plan, a firearm, and recent job loss.
• State. How does this person compare with their own baseline? Tamsin has never had these thoughts before. That change is more informative than her absolute level.
• Foreseeable changes. What is coming that could move this? Exam results in nine days. Her housemates leaving for spring break. An anniversary.
• What would help. The bit that produces action.

The point: You are not forecasting. You are describing a state, identifying what could worsen it, and building something that makes the worst hours survivable.

The safety plan, step by step

The instrument here is the Safety Planning Intervention developed by Barbara Stanley and Gregory Brown. It takes twenty to forty minutes, it is written in the client's words, and the client keeps a copy. It has six steps and the order is deliberate: each step is more effortful and more dependent on other people than the one before, so a person in distress works down the list only as far as they need to.

Three details separate a real plan from a form. Every entry is specific: not call someone but Nell, with the number written on the card. Every entry is tested out loud: is the library actually open at 3 am, does Priya work nights, would you really call Nell or would you not want to worry her. And the client writes it, or you write while she dictates, and she leaves with it in her phone and on paper.

The evidence is decent and worth stating precisely. Stanley, Brown, and colleagues compared emergency departments using the safety plan plus telephone follow-up with departments delivering usual care, across 1,640 patients in the Veterans Health Administration. Patients who received the intervention had 45 percent fewer suicidal behaviours over the following six months and more than double the odds of attending at least one outpatient mental health appointment. It was a cohort comparison rather than a randomized trial, so the causal claim is not airtight, and it is one of the better pieces of evidence available for a brief intervention in this area.

Step six deserves its own section

Means restriction is the step clinicians rush, and it is the one with the strongest evidence behind it.

The logic rests on two facts. First, the interval between deciding and acting is often very short, sometimes minutes. Second, methods differ enormously in how often they kill: a firearm attempt is fatal in the great majority of cases, an overdose of most common medications is not. Putting time and distance between a person and a highly lethal method therefore converts a large share of potentially fatal attempts into survivable ones.

Two natural experiments make the point better than any trial could. When Britain converted its domestic gas supply from coal gas, which contained carbon monoxide, to natural gas, which does not, suicide by domestic gas fell to nearly nothing and the overall national suicide rate fell with it, rather than shifting entirely to other methods. And Richard Seiden followed up hundreds of people who had been physically prevented from jumping from the Golden Gate Bridge; the large majority were still alive years later or had died of unrelated causes. Substitution happens, but it is far from complete, which is the whole basis of the intervention.

In practice this is a conversation, not an order. Do you have access to a firearm? Where is it kept, who else has a key, is there someone who could hold it for a while, would a locked box with the ammunition stored separately be possible this week? For medication: how much is in the house, can a smaller amount be kept and the rest held by somebody. Ask about alcohol too, because intoxication is when the interval between deciding and acting collapses.

988, and how to introduce it

Since July 2022 the United States has had a three-digit number, 988, for the Suicide and Crisis Lifeline, reachable by call or text, with a Spanish line, an American Sign Language option, and a veterans line. Say what it actually is, because clients have heard rumours: it is a trained counsellor on a phone, the great majority of calls are resolved on the call, and emergency services are dispatched in only a small minority of contacts. Say that plainly, because the fear that calling means police at the door is real and it stops people calling.

Violence risk, honestly

The same lesson applies with even less comfort. Base rates of serious violence are low, unstructured clinical judgement performs poorly, and structured tools such as the HCR-20 improve on it while remaining far better at describing groups than individuals.

What is worth doing is specific. Ask directly whether the person has thoughts of harming someone, and if so, who: an identified target is the fact that changes both your clinical and your legal obligations, which the ethics lesson takes up. Ask about access to weapons. Ask about past violence, which remains the strongest single predictor. Note current intoxication and recent losses. And record what you asked, including the negatives.

What happens after the session ends

1. Do not end on time. Whatever else, Tamsin does not leave at five with this unfinished.
2. Consult, that day. Even a low-acuity finding gets a supervisor's ear. Write down who you spoke to and what was decided.
3. Document what you asked and what you were told, including denials, in the client's words. A note saying no suicidal ideation is worth much less than a note recording the six questions and her answers.
4. Shorten the interval. Bring her back in three days rather than two weeks, and say why.
5. Make the follow-up contact. A phone call between sessions is not a courtesy; it is the component paired with the safety plan in the evidence above.

Common misconceptions

• Asking about suicide plants the idea. It does not. It tells a person the subject can be discussed, which is often the first time anybody has said so.
• A contract for safety protects the client and the clinician. No-suicide contracts have no evidence of reducing risk, they can suppress honest disclosure, and they have been superseded by safety planning.
• People who talk about it do not do it. Most people who die by suicide communicated something beforehand, often to a non-professional.
• Restricting means just moves the person to another method. Substitution is real and incomplete, as the coal gas conversion and the bridge follow-up both show.
• A good assessment sorts clients into high, medium, and low risk. Fifty years of research puts prediction barely above chance. Formulation, monitoring, and means restriction are what the evidence supports.

What to remember

• Ask directly, in ordinary words, and follow the Columbia order: wish to be dead, thoughts of killing yourself, method, intent, plan with intent, and preparatory behaviour.
• Prediction is barely better than chance; risk formulation describes status, state, foreseeable changes, and what would help.
• The Stanley-Brown plan has six steps, written in the client's words and kept by the client, escalating from what she can do alone to who she can call.
• Test every entry aloud, and never accept a vague one.
• Means restriction has the strongest evidence in the whole area, and it is a conversation about time and distance rather than an instruction.
• Afterwards: do not end on time, consult that day, document the questions and the answers, shorten the interval, and make the follow-up call.

Sources

1. Stanley, B., Brown, G. K., Brenner, L. A., Galfalvy, H. C., Currier, G. W., Knox, K. L., et al. (2018). Comparison of the safety planning intervention with follow-up vs usual care of suicidal patients treated in the emergency department. JAMA Psychiatry, 75(9), 894-900. PubMed Central
2. Franklin, J. C., Ribeiro, J. D., Fox, K. R., Bentley, K. H., Kleiman, E. M., Huang, X., et al. (2017). Risk factors for suicidal thoughts and behaviors: A meta-analysis of 50 years of research. Psychological Bulletin, 143(2), 187-232.
3. Columbia Lighthouse Project. (n.d.). About the Columbia Protocol (C-SSRS). cssrs.columbia.edu
4. Centers for Disease Control and Prevention. (n.d.). Suicide data and statistics. CDC. cdc.gov
5. 988 Suicide and Crisis Lifeline. (n.d.). 988 Suicide and Crisis Lifeline. 988lifeline.org

4.42 pm, a car park at the courthouse

The front desk puts a call through. It is Ivo Petran, 46, whom you have seen twice. He is in his car outside the county courthouse. He was served with divorce papers at eleven, he has been drinking since about noon, his wife has taken their two children to her sister's, and he is talking fast and loudly and does not appear to be listening to his own sentences.

You have no formulation for this, no treatment plan that covers it, and no appointment slot. What you have is the next forty minutes, and a set of moves that have been worked out by people who do this for a living.

What a crisis is, technically

The working definition comes from Gerald Caplan and has held up. A crisis exists when a person faces an obstacle to important life goals that for a time is insurmountable through their usual methods of problem-solving. Three features follow, and each has a practical consequence.

• It is about coping, not about the event. The same divorce papers produce a crisis in one person and a bad week in another. What defines it is that the person's ordinary repertoire has stopped working.
• It is time-limited. Acute crisis states resolve one way or another within roughly four to six weeks. They do not stay acute. That is why brief, immediate contact matters more here than anywhere else in clinical work.
• Functioning is temporarily lowered and receptivity is temporarily raised. A person in crisis is doing worse than their baseline and is also unusually open to influence and to help. Both halves are true, which is why crisis work is high-leverage and easy to do harm in.

Your job is not to resolve the divorce, treat the drinking, or restructure anything. It is to get through today and to get him to tomorrow with a plan.

Seven stages, run on Ivo

Albert Roberts set out a seven-stage model that has become the common vocabulary in social work crisis practice. It is not a script and the stages overlap, but the order is not arbitrary.

Note the two moves in stage six that carry most of the weight. He does something now, while you are still on the phone, rather than agreeing to do something later. And the shotgun leaves the house tonight, which is the means restriction from the previous lesson arriving in an ordinary conversation rather than as a special procedure.

The upshot: A crisis plan that begins after the call ends is a wish. Make the first step happen while you are still talking.

De-escalation, when the person is agitated

If Ivo were in the waiting room rather than on the phone, shouting, the task would shift. The American Association for Emergency Psychiatry published a consensus statement in 2012 setting out verbal de-escalation, and its content is unglamorous and specific.

The single most useful sentence in an agitated situation is a question rather than an instruction: what would help right now? It is not a technique for producing compliance. It frequently produces an answer you can act on, and being asked is itself de-escalating.

Choosing a disposition

By the end of a crisis contact you owe a decision about what happens next, and the options are more numerous than students assume.

Be accurate about what an involuntary hold does. It buys a short period of observation and evaluation, typically measured in days and defined by state statute. It is not treatment, it does not lower risk after discharge, and the period following a psychiatric discharge is itself one of elevated risk. It is sometimes exactly right and it is never neutral, and a clinician who reaches for it because it feels safer for the clinician has made a decision about themselves.

Two more things about the call to 911. In many places the responder to a mental health crisis is a police officer, and for some clients, particularly Black clients and clients with a history of police contact, that carries risk they know about better than you do. Where 988 and mobile crisis exist, they are usually the better call. Where they do not, say honestly what you are about to do and why, rather than making the call while the person listens.

Psychological first aid, and the thing not to do

After a disaster or a mass casualty event, the approach with the broadest support is psychological first aid: practical, non-intrusive help focused on safety, comfort, connection to support, information, and linkage to services. It does not ask people to describe what they experienced.

That last clause is the point. Single-session psychological debriefing, in which people are gathered soon after an event and encouraged to recount it in detail, was standard practice for years. The Cochrane review by Suzanna Rose, Jonathan Bisson, Rachel Churchill, and Simon Wessely found no evidence that it prevents post-traumatic stress disorder, and some evidence that it can worsen outcomes. It remains widely offered, often by well-meaning employers.

Debriefing a team about how an incident was handled is a different thing, is useful, and belongs in the supervision lesson. Requiring people to narrate their worst hour to a room is what the evidence is against.

Afterwards

Write the note today, not tomorrow, and write what you asked and what he said. Tell your supervisor before you go home even if it is by message. Put both follow-up calls in the diary and make them; the second call is the one people skip, and it is the one the client remembers. And notice your own state: a call like Ivo's leaves adrenaline behind, and driving home shaky is a normal response to an abnormal forty minutes, not a sign that you are unsuited to this.

Common misconceptions

• A crisis is defined by the severity of the event. It is defined by the failure of the person's usual coping. The same event produces a crisis in one person and a hard day in another.
• The first task is to calm the person down. The first task is to establish safety and lethality. Calm arrives from being taken seriously, not from being told to relax.
• Hospitalization is the safe option. It buys observation, it is coercive, and the period after discharge carries elevated risk. Safe for whom is the question worth asking.
• Talking it through immediately prevents later problems. Single-session debriefing has no evidence of preventing post-traumatic stress disorder and some evidence of harm.
• Crisis work is what you do until real treatment starts. It is high-leverage clinical work in its own right, because functioning is lowered and receptivity is raised at the same time.

Where this leaves us

• A crisis is a failure of usual coping, is time-limited to roughly four to six weeks, and combines lowered functioning with raised receptivity.
• Roberts's seven stages run from lethality assessment through rapid rapport, the precipitant, the feelings, alternatives, an action plan, and follow-up.
• Make the first step of the plan happen during the contact, and address means as part of the plan rather than as a separate procedure.
• De-escalation is space, one speaker, short sentences, honest limits, and the question what would help right now.
• Dispositions run from home with follow-up through mobile crisis and stabilization units to voluntary and involuntary hospital care; an involuntary hold buys evaluation, not treatment.
• Psychological first aid is supported; single-session debriefing that requires people to narrate the event is not.

Sources

1. Richmond, J. S., Berlin, J. S., Fishkind, A. B., Holloman, G. H., Zeller, S. L., Wilson, M. P., et al. (2012). Verbal de-escalation of the agitated patient: Consensus statement of the American Association for Emergency Psychiatry Project BETA de-escalation workgroup. Western Journal of Emergency Medicine, 13(1), 17-25. PubMed Central
2. Rose, S., Bisson, J., Churchill, R., and Wessely, S. (2002). Psychological debriefing for preventing post traumatic stress disorder (PTSD). Cochrane Database of Systematic Reviews, CD000560.
3. Roberts, A. R. (Ed.). (2005). Crisis intervention handbook: Assessment, treatment, and research (3rd ed.). Oxford University Press.
4. Wikipedia contributors. (n.d.). Psychological first aid. Wikipedia. en.wikipedia.org

Where everyone sits

The Odoms arrive for a first family session. Cheryl is 42, her husband Ray is 45, and Devin is 14 and was referred after two suspensions for fighting. Devin sits on the arm of the sofa, as far from the door as the room allows. Cheryl sits between him and Ray. When you ask Devin what he thinks the problem is, Cheryl answers. When you ask again, she answers again. Ray has not said anything and is looking at his phone.

You have learned more in ninety seconds than the referral told you in a page, and none of it is about Devin's temper.

Who is the client?

That is the question that separates these modalities, and it is not rhetorical. It determines who has confidentiality, whose goals go in the plan, and who you are working for when their interests diverge.

Four family ideas that do real work

You are not going to become a family therapist from a lesson, and you may well co-facilitate family sessions in placement. Four concepts carry most of the practical load.

Circular rather than linear causality. A linear account says Devin fights, so his parents are stressed. A circular one says Devin escalates, Cheryl protects, Ray withdraws, Cheryl carries it alone, Cheryl becomes short with Devin, Devin escalates. There is no first cause in a loop, which is not an evasion of responsibility; it is a description of where the loop can be interrupted, and there are four places rather than one.

Structure. Salvador Minuchin's contribution was to notice that families have boundaries, hierarchies, and subsystems, and that trouble often lives in their arrangement rather than in any member. In the Odom family the parental subsystem is not functioning as a unit: Cheryl and Devin form a tight pair with Ray outside it. When Cheryl answers for Devin, the boundary between her and her son is thinner than the one between her and her husband. That is structural information, and it is visible from the seating.

The identified patient. The person brought in is not always the person with the problem. Sometimes a child's behaviour is the only thing that gets the family into a room, and sometimes it functions to keep parents united around a shared concern. Treating Devin alone would be treating the smoke. That said, the concept is easily overused: sometimes a fourteen-year-old is fighting because of what happens at his school, and reading the identified patient idea into every case is its own error.

Reframing. Offering a different, equally plausible, more workable description of the same behaviour. Not a lie, and not a compliment. Cheryl says: he will not talk to us, he shuts down. A reframe: I notice he goes quiet when the two of you disagree in front of him. I wonder whether the quiet is him trying not to make it worse. Devin looks up for the first time. That is what a reframe is for: it opens a conversation the previous description had closed.

In session, the move that follows is an enactment: rather than discussing how they argue, ask them to discuss the curfew now, with you watching. Then intervene on the sequence rather than on the content. Ray, she has answered for him twice. What happens if you ask him yourself?

What matters here: In family work you intervene on the sequence, not on the topic. The topic is almost never the problem.

When conjoint work is the wrong room

This is not a nuance and it is not optional.

Where there is coercive control or intimate partner violence, standard conjoint couples or family therapy is contraindicated. The reasons are concrete. A partner cannot speak freely in front of someone who will be alone with them afterwards. Therapy that treats the violence as a shared relational pattern distributes responsibility for it. And the session itself can raise risk: a disclosure made in the room is paid for later.

What this requires of you is a routine practice: see partners separately at least once during assessment, ask about fear and control directly and privately, and never disclose to one partner what the other said. Two questions worth memorizing: is there anything you cannot say when your partner is in the room, and are you ever afraid of them. If the answers indicate coercive control, the work moves to individual contact, safety planning, and specialist services.

Similar caution applies where a family session is being used to pressure a member into a disclosure, and where active untreated dependence makes a session unproductive or unsafe.

What a group does that you cannot

Irvin Yalom's contribution to group psychotherapy was to specify what is actually working, in a list of therapeutic factors. Several are unavailable in individual work at any price.

• Universality. The discovery that other people have the same thought you believed disqualified you from ordinary company. A clinician can say you are not the only one; only a group can demonstrate it.
• Instillation of hope. Seeing a member four months ahead of you doing better. Evidence, not reassurance.
• Altruism. Being useful to someone else, which for people who have been recipients of services for years is a different experience of themselves.
• Interpersonal learning. Finding out how you land on other people, from people with no reason to be polite about it.
• Corrective recapitulation of the family group. The group reproduces familiar family dynamics, and this time they can be examined and go differently.
• Cohesiveness. The group equivalent of the alliance, and the factor most consistently associated with outcome.

His list also includes imparting information, socializing techniques, imitative behaviour, catharsis, and existential factors. The point of knowing them is diagnostic: when a group is flat, you can ask which factors are absent instead of adding an exercise.

The mechanics that decide whether a group works

Groups fail for boring reasons, and most of them are decided before the first session.

• Screening and preparation. Meet every prospective member individually. Explain what the group is, what it is not, how long, what is expected. Poorly prepared members are the main source of early dropout, and early dropout destabilizes a group for everyone left.
• Size. Roughly six to ten for a therapy group. Below five it becomes serial individual work; above ten members disappear.
• Open or closed. Closed groups build cohesion and cannot replace losses. Open groups survive attrition and never get as deep. Choose deliberately.
• Confidentiality, stated accurately. You can require members to keep what is said in the room, and you cannot enforce it. Say exactly that, in the first session, out loud, rather than implying a protection that does not exist.
• The monopolizer and the silent member. These are group phenomena, not individual traits: a group tolerates a monopolizer because someone talking relieves everyone else. The intervention addresses the group, not the person. I notice we have let Marcus carry the last twenty minutes. What is that about for the rest of us?
• Co-facilitation. Two leaders can watch content and process separately, and they can debrief honestly afterwards. It is also how you learn, which is why in placement you will co-facilitate long before you lead.

Common misconceptions

• Family therapy means everyone attends every session. The unit of attention is the pattern; who is in the room varies, and some sessions are deliberately with a subsystem.
• Circular causality means nobody is responsible. It describes where a loop can be interrupted. Responsibility for violence or abuse is never circular, which is why conjoint work is contraindicated there.
• The identified patient is always carrying the family's problem. Sometimes the referred person is simply the one with the problem, and forcing the interpretation is its own error.
• Group therapy is individual therapy done cheaply. Universality, altruism, and interpersonal learning are not available in a dyad, and a group run as serial individual work loses all three.
• Confidentiality in a group is the same as in individual work. It is a rule you can require and cannot enforce, and members are entitled to know that before they disclose.

The short version

• Ask who the client is before anything else, because it decides confidentiality, goals, and whose interests you serve when they diverge.
• In family work, watch sequences rather than content: who answers for whom, who withdraws, what happens next.
• Structure, circular causality, the identified patient, and reframing carry most of the practical load; enactment is how you see the pattern rather than hear about it.
• Conjoint therapy is contraindicated where coercive control or intimate partner violence is present, and screening separately is a routine practice rather than a special step.
• Yalom's factors, particularly universality, hope, altruism, interpersonal learning, and cohesiveness, are what a group supplies that a dyad cannot.
• Groups are decided by preparation, size, open or closed design, an accurate confidentiality statement, and treating monopolizing as a group event.

Sources

1. Yalom, I. D., and Leszcz, M. (2020). The theory and practice of group psychotherapy (6th ed.). Basic Books.
2. Minuchin, S. (1974). Families and family therapy. Harvard University Press.
3. Centers for Disease Control and Prevention. (n.d.). About intimate partner violence. CDC. cdc.gov
4. Wikipedia contributors. (n.d.). Family therapy. Wikipedia. en.wikipedia.org

Before you read

Find something you have written about another person: an email about a colleague, a reference, an incident report, a message about a family member. Read one sentence of it as though a stranger were reading it aloud in a room where that person was sitting. Keep that sentence in mind.

A sentence read into the record

A custody hearing, a Tuesday morning. Opposing counsel has your client's clinical record and reads one sentence aloud, slowly, for the judge.

Client appeared angry and became hostile when discussing the children.

Then she asks you what hostile means. Did she raise her voice? Did she threaten anyone? Did she stand up? You wrote it fourteen months ago at 5.50 pm on a Friday, one of six notes you were catching up on, and you have no independent memory of the session. All you have is that sentence, and the sentence says something about your client's character that you cannot now support.

Here is what you should have written: Client raised her voice twice when the visitation schedule came up, said this is the third time they have changed it, and declined to continue that topic. Discussion resumed after a pause.

Same session. One is evidence, the other is an accusation. This lesson is about the difference.

What a note is actually for

Students are taught that notes are for billing and liability, and both are real and neither is the main thing. A clinical note has four jobs, and they explain almost every rule that follows.

1. Continuity. If you are hit by a bus, the person who picks up the case must be able to continue it. That is the primary purpose and it is the test to apply when you cannot decide whether to include something.
2. Clinical thinking. A record of what you decided and why makes your reasoning reviewable, by a supervisor, by you in six weeks, and by whoever comes next.
3. Accountability. It shows what was done, including whether risk was assessed and whether consultation happened.
4. Communication and payment. It tells the team and the payer what occurred.

Now the readers, which is the list people underestimate: the client, who in most cases has a legal right of access; the next clinician; your supervisor; a payer's auditor; a licensing board; a lawyer in a custody, disability, immigration, or criminal matter; and sometimes a jury.

Bottom line: Write every note as if the client will read it, because they may, and as if it will be read aloud in a hearing, because it might.

Two formats, one session

The SOAP note came from medicine and separates four things. DAP collapses the first two, which suits behavioural health where the line between what the client reports and what you observe is often the only line that matters.

Session eight with Renata Marchetti, written both ways.

SOAP. S: Reports two panic attacks since last session, both at the supermarket, down from four. Says the on-ramp drive was the worst thing I have done and I did it three times. Reports drinking unchanged at three to four beers most evenings. Denies suicidal ideation. O: On time, alert, oriented. PHQ-9 = 11, down from 14 on 12 September. Completed five of six agreed exposure trials with written predictions. Tearful once when describing her mother's oxygen use. A: Continued improvement in panic frequency and avoidance; exposure work is producing expectancy violation as intended. Depressive symptoms improving more slowly, consistent with unresolved sleep apnea and alcohol use, both of which remain untreated. Alcohol is now the principal untouched maintaining factor. P: Hierarchy steps 6 and 7 this week; sleep clinic appointment 3 October, client to confirm; introduce a drinking diary and revisit at session 10 using motivational rather than directive approach; next appointment 26 September.

DAP. D: Two panic attacks since last session, both at the supermarket, down from four. Client states the on-ramp drive was the worst thing I have done and I did it three times. Drinking unchanged, three to four beers most evenings. Denies suicidal ideation. On time, alert, oriented; PHQ-9 = 11, from 14 on 12 September; five of six exposure trials completed with written predictions; tearful once regarding mother's oxygen use. A and P as above.

Both are defensible. Notice what makes them so: quotations rather than characterizations, numbers with dates for comparison, the negative finding on suicidal ideation recorded explicitly, and an assessment that states a reason rather than a verdict.

The rules, and the reason for each

The third-party rule catches people out. If Renata tells you her sister was hospitalized for an overdose, the clinically relevant fact is that Renata is worried about her sister and has an increased caregiving load. The sister's diagnosis is not needed for continuity, and she is not your client.

Psychotherapy notes are a specific legal category

Under the HIPAA Privacy Rule, psychotherapy notes are a defined and narrow thing: a clinician's notes analyzing the contents of a session, which are kept physically or electronically separate from the rest of the record. They are given extra protection: they are generally excluded from the individual's right of access, and disclosure typically requires a specific authorization rather than being covered by a general consent to release records.

Two conditions matter. If the notes are not kept separate, they are not psychotherapy notes and get no special treatment. And the category expressly excludes medication and prescription monitoring, session start and stop times, treatment modality and frequency, test results, and any summary of diagnosis, functional status, treatment plan, symptoms, prognosis, and progress. In other words, everything a payer or a court usually wants is in the progress note by definition, not in the protected category.

Some clinicians use this well, keeping working hypotheses separate from the chart. Others treat it as a place to write things they would not want the client to see, which is a poor reason and often an ethical problem in its own right.

A subpoena arrives

This is the moment students most reliably get wrong, and the error is expensive.

A subpoena issued by an attorney is not a court order. It is a demand from one party in a dispute. A court order is signed by a judge. The difference determines what you are permitted to do, because client communications in clinical work are generally privileged, and the privilege belongs to the client, not to you.

The sequence when one arrives:

1. Do not send the records. Not that day, not as a courtesy, not because the letterhead is intimidating.
2. Tell your supervisor and the agency's legal counsel. Same day.
3. Notify the client, unless doing so is legally prohibited. They may want to authorize release, and they may want to fight it. It is their privilege to assert or waive.
4. If the client authorizes release, release only what the authorization covers, and no more.
5. If they do not, counsel asserts privilege on the client's behalf, and the question goes to a judge. A protective order limiting who sees what is often the practical outcome.
6. Comply with a court order, and produce exactly what it specifies.

The Code of Ethics is aligned with this in standard 1.07 on privacy and confidentiality: disclose the least amount necessary, and where compelled by law, request that the court withdraw the order or limit it as narrowly as possible.

The note as a clinical act

One more thing, easily missed. Clients increasingly read their own notes, and some clinicians now write them collaboratively, reading the assessment paragraph aloud at the end of a session. It takes ninety seconds, it catches errors, and it does something to the relationship that is hard to get any other way: the client learns exactly what is being said about them and can correct it.

It also disciplines the writing. It is remarkably difficult to write client is resistant to treatment recommendations when you know you are going to read it to her at 4.55.

Common misconceptions

• Notes are primarily for billing and liability. The first job is continuity: could the next clinician pick this up. That test resolves most inclusion questions.
• Clinical language sounds professional. Words like hostile, manipulative, and non-compliant are conclusions dressed as observations, and they cannot be defended when someone asks what you actually saw.
• Anything I label psychotherapy notes gets extra protection. Only notes kept separate from the record qualify, and the category excludes almost everything payers and courts request.
• A subpoena must be answered by sending the file. An attorney's subpoena is not a court order, the privilege belongs to the client, and the correct first move is to send nothing and consult.
• Vague notes protect the client. They protect nobody. A vague note cannot show that risk was assessed, cannot support your reasoning, and leaves the next clinician starting over.

Putting it together

• A note exists first for continuity, then for reviewable clinical reasoning, accountability, and communication.
• SOAP separates subjective from objective; DAP merges them into data. Both work when the entries are quotations, observations, and numbers with dates.
• Describe behaviour rather than characterizing the person, record the negatives you checked, and always write the because.
• Apply the minimum necessary standard to information about people who are not your client.
• Psychotherapy notes are protected only when kept separate, and the category excludes diagnosis, plan, progress, times, and modality.
• An attorney's subpoena is not a court order: send nothing, consult counsel, notify the client, and let the privilege be asserted or waived by its owner.

Sources

1. National Association of Social Workers. (2021). Code of ethics, standards 1.07 and 3.04. NASW. socialworkers.org
2. U.S. Department of Health and Human Services. (n.d.). HIPAA and mental health information. HHS. hhs.gov
3. Wikipedia contributors. (n.d.). SOAP note. Wikipedia. en.wikipedia.org
4. Kagle, J. D., and Kopels, S. (2008). Social work records (3rd ed.). Waveland Press.

August 1969, Cowell Memorial Hospital

A graduate student at Berkeley named Prosenjit Poddar told his psychologist, Lawrence Moore, that he intended to kill a young woman when she came back from Brazil. He did not name her in that session, but she was easily identifiable: Tatiana Tarasoff, who had rejected him.

Moore did something. He telephoned the campus police, then wrote to them asking that Poddar be committed for observation. The police picked Poddar up, questioned him, found him rational, and released him after he promised to stay away from her. Moore's supervising psychiatrist then directed that no further action be taken and that Moore's letter be destroyed. Nobody told Tatiana Tarasoff or her family.

On 27 October 1969, Poddar killed her at her home.

Her parents sued the university. The California Supreme Court decided the case twice. In 1974 it found a duty to warn the intended victim. It then reheard the case and issued the 1976 opinion that carries the name, Tarasoff v. Regents of the University of California, reformulating the obligation as a duty to exercise reasonable care to protect the foreseeable victim. Warning is one way to discharge it. So is notifying police, so is initiating commitment proceedings, so is whatever the circumstances reasonably require. Justice Tobriner's sentence is the one everybody remembers: the protective privilege ends where the public peril begins.

Two things follow that are routinely got wrong. Tarasoff is a California decision, and other states have gone their own ways: some impose a duty, some permit disclosure without requiring it, some limit it tightly to an identified victim and an explicit threat, and a few have declined to recognize one at all. You need to know your own state's statute, and you should look it up before you need it. And Tarasoff is about a specific configuration: a serious threat of violence against a reasonably identifiable person. It is not a general licence to break confidentiality when a client worries you.

The Code, and what changed in 2021

The NASW Code of Ethics is organized as six values, a set of ethical principles, and then numbered standards that are the working text. The most recent revision, effective in 2021, made two changes worth knowing. It added explicit language on professional self-care, moving it from a personal virtue to an ethical expectation. And it revised the cultural competence standard to speak of cultural humility and to name the obligation to work against oppression and racism, rather than treating competence as a credential a clinician attains and keeps.

Learn the numbers of the standards you will use most: 1.01 commitment to clients, 1.03 informed consent, 1.04 competence, 1.06 conflicts of interest and dual relationships, 1.07 privacy and confidentiality, 3.01 supervision, 3.09 commitments to employers, and 4.05 impairment.

Boundaries: crossing and violation

These two words are used interchangeably and should not be.

A boundary crossing is a departure from the usual frame that is not exploitative and may be clinically indicated: attending a client's graduation, accepting a hand-knitted scarf at termination, giving a lift to a stranded client in a rural county with no bus. A boundary violation exploits the relationship for the worker's benefit or foreseeably harms the client. Sexual contact with a current client is the clearest case and is prohibited absolutely.

The Code does not ban dual relationships outright, because in rural communities, small cultural or linguistic communities, recovery communities, and the deaf community, an absolute ban would mean no services. Standard 1.06 requires you to avoid dual relationships where there is a risk of exploitation or harm, and where they are unavoidable, to set clear, appropriate, and culturally sensitive boundaries.

Four questions decide most real cases.

1. Whose need does this meet? If the honest answer includes yours, stop.
2. Would I do this for every client with the same presentation? Exceptions are where trouble lives.
3. Would I write it in the note? Reluctance to document is diagnostic.
4. Would I tell my supervisor before rather than after? Same test, different direction.

Why this matters: Serious violations almost never begin with the serious act. They begin with a small unexamined exception followed by a reason not to mention it.

Informed consent is a process

Standard 1.03 makes consent a continuing obligation rather than a signature. It covers purpose, risks, limits arising from third-party payers, costs, reasonable alternatives, the client's right to refuse or withdraw, and the timeframe.

Three parts get skipped. Capacity: if a client cannot understand the information, you seek permission from an appropriate proxy and still inform the client to the extent they can understand. Recording and observation: a session recorded for supervision requires consent, obtained beforehand, with the client free to say no. And the student clause, which is the one that applies to you: clients have to be told when services are being provided by a student, and by whom you are supervised. Saying it plainly costs nothing and hiding it costs everything if it surfaces later.

Consent gets renewed whenever something material changes: a new modality, a group, a recording, a change of supervisor, an unpaid balance affecting continuation.

Confidentiality, and its four exceptions

Standard 1.07 sets the rule and the exceptions. The rule is that you disclose only with valid consent. The exceptions are compelling professional reasons, and there are four you will meet.

Mandated reporting deserves a paragraph because students expect certainty and the standard is lower. It is reasonable suspicion, not proof. You are not the investigator, and waiting until you are sure is itself a failure. Tell the client you are making a report, unless doing so would increase risk to a child, and tell them before you file rather than after. The relationship sometimes survives that conversation and sometimes does not; conducting it honestly is the only version that gives it a chance.

Telehealth, where the rules are newer

The technology standards published jointly by NASW, ASWB, CSWE, and the Clinical Social Work Association set expectations that a video platform does not automatically satisfy.

• Licensure follows the client. You must generally be licensed in the jurisdiction where the client is physically located at the time of the session, not where you are sitting. A client who takes a session from a car in another state has created a problem you need to know about.
• Verify location every session. Ask where they are. It sounds fussy and it is the basis for both the licensure question and the emergency plan.
• Have an emergency plan on file. Address, a named local contact, and the local emergency number, agreed at the outset. If a client becomes unsafe on a video call, you cannot walk them down a corridor.
• Both ends need privacy. Yours is your responsibility; theirs is a conversation, especially where a household member may be part of the problem.
• Say what the platform is and is not. Encryption, recording, and what happens to the data.
• Digital boundaries. Searching for a client online, accepting a social media request, or texting outside agreed channels are all boundary questions, and the first one requires a clinical reason and a note.

Working a dilemma

Take one. You are an intern in a small county behavioural health office. Your client, Jonah Reese, 34, discloses that his sister's partner has been hitting her, that she has a two-year-old, and that he is telling you only because you promised confidentiality. Nothing has been said about the child being hurt.

The sequence, not the answer.

1. Identify the actual question. Not should I break confidentiality, but does this meet my state's reporting threshold, given that the disclosure concerns a child in a household where partner violence is occurring, and that states differ on whether exposure to such violence is reportable.
2. Locate the rules. The state statute first, then agency policy, then the Code standard 1.07, then any relevant case law.
3. Consult, today. This is not a decision to make alone, and it is exactly the kind that supervision exists for.
4. Consider who else is affected. Jonah, his sister, the child, and the future of Jonah's own treatment.
5. Decide, act, and document the reasoning, including the consultation and the rule you relied on.
6. Tell Jonah what you are doing and why, before you do it if that does not increase anyone's risk.

Notice that the sequence protects the client rather than the clinician, and that step three is not optional. An intern who resolves this alone has made a second error whatever the first answer was.

Common misconceptions

• Tarasoff created a duty to warn. The 1976 rehearing reformulated it as a duty to exercise reasonable care to protect, of which warning is one option among several.
• Tarasoff is the law everywhere. It is a California decision. States impose, permit, limit, or decline the duty, and you need your own state's statute before the situation arises.
• All dual relationships are prohibited. The Code prohibits those risking exploitation or harm. An absolute ban would end services in rural, cultural, linguistic, and recovery communities.
• Informed consent is the form signed at intake. It is a continuing process that is renewed whenever the modality, the recording, the supervision, or the financial arrangement changes.
• You must be certain before making a mandated report. The standard is reasonable suspicion. Investigating first is a failure of the duty, not diligence.

What to carry forward

• Tarasoff arose from a real sequence of decisions in 1969, and the 1976 opinion imposes a duty to protect rather than specifically to warn.
• Check your own state's statute on the duty and on mandated reporting before you need either.
• The 2021 Code added professional self-care as an ethical expectation and reframed cultural competence around humility and anti-oppressive practice.
• A boundary crossing may be indicated; a violation exploits or foreseeably harms. Ask whose need it meets, whether you would do it for everyone, whether you would write it down, and whether you would tell your supervisor first.
• Consent is continuing, covers third-party limits and recording, and requires telling clients when they are being seen by a student.
• Telehealth licensure follows the client's location, which is why you ask where they are every single session.

Sources

1. National Association of Social Workers. (2021). Code of ethics of the National Association of Social Workers (rev. ed.). NASW. socialworkers.org
2. Tarasoff v. Regents of the University of California, 17 Cal. 3d 425 (1976). CourtListener
3. NASW, ASWB, CSWE, and CSWA. (2017). Standards for technology in social work practice. socialworkers.org
4. Reamer, F. G. (2018). Social work values and ethics (5th ed.). Columbia University Press.

Twenty-two minutes on a Thursday

Your supervision hour runs like this. You read out your twelve open cases. Your supervisor asks about the two with court dates. You both look at the waiting list. She signs the two forms she needs signed, and it is over in twenty-two minutes.

You have just had a case management meeting. You have not had supervision, and the difference is not pedantic: nothing that happened in that room could have caught the thing you are getting wrong.

Here is the test. In those twenty-two minutes, did anyone look at what you actually did in a session? Not what you reported doing. What you did.

Three functions, and the one that gets eaten

Alfred Kadushin's division of supervision into three functions is the standard framework in social work, and its value is that it makes an absence visible.

Under pressure the administrative function expands until it fills the hour, because it has deadlines and the other two do not. That is how a supervision hour becomes twenty-two minutes of list-reading, and it is the single most common structural failure in clinical supervision.

The point: If your supervision has no educational content and no supportive content, it is not supervision, and saying so out loud to your supervisor is a reasonable professional act.

Self-report is the wrong data

The central problem in supervising clinical work is that the supervisor was not in the room. What they get is your account, and your account is filtered by what you noticed, what you understood, and what you would rather not raise.

This is not a character flaw. When researchers compare what clinicians say they do with recordings of what they actually do, the two diverge substantially, and they diverge in a consistent direction: clinicians report delivering more of the difficult components than the recordings show. You cannot report the thing you did not notice yourself doing.

So good supervision runs on live material.

• Recordings, audio or video, with consent, which is the strongest form and the one clinicians resist most.
• Live observation or co-working, where the setting allows.
• Process recordings, which are the social work profession's own instrument and are underrated. You reconstruct a segment of a session in three columns: what was said, what you were thinking and feeling as it was said, and your analysis afterwards. It is laborious and it makes visible things nothing else does, particularly the gap between what you felt and what you did.

Bring the case you are avoiding. This is the single most useful habit in supervision, and everyone knows which case it is: the one you have not written up, the one you keep at the end of the list, the client you feel a small unexplained relief about when they cancel. That relief is data.

Reflective practice, specifically

Donald Schon's distinction, published in 1983, is worth having because it names two different skills. Reflection-in-action is thinking while doing: noticing mid-session that the last four questions were yours and none were hers, and changing course before the hour ends. Reflection-on-action is thinking afterwards, in the note, in supervision, in the car.

Reflection-on-action can be taught with process recordings. Reflection-in-action is developed by having reflected-on so many sessions that the noticing starts to arrive earlier. There is no shortcut, and this is one of several places in this course where the honest answer is that the thing cannot be acquired by reading.

One phenomenon to know by name. Parallel process is the observation that the dynamic between clinician and client sometimes reappears in the dynamic between clinician and supervisor. A worker who feels helpless with a demanding client may present that case in a way that leaves the supervisor feeling pressed to solve it immediately. Naming it in the room, gently, is often the fastest route into what is happening in the case.

And a boundary: supervision is not therapy. It examines your reactions insofar as they affect the work. When the material is about your own history rather than the case, the correct move is a referral to your own therapist, and a supervisor who blurs that line has created a dual relationship with a power differential attached.

Three injuries with three different causes

These terms are used interchangeably and they describe different things with different remedies.

The burnout row is the one most often misread. Christina Maslach's research programme, running since the 1970s, locates burnout in the mismatch between a person and six areas of their working life: workload, control, reward, community, fairness, and values. It is an occupational phenomenon with occupational causes. The World Health Organization classifies it that way too, as a syndrome resulting from chronic workplace stress that has not been successfully managed, rather than as a medical condition of the individual.

That matters because of what agencies do with it. Offering a resilience workshop and a yoga voucher to a team carrying 38 cases each is a way of converting a structural problem into a personal one. The 2021 Code added professional self-care as an ethical expectation for social workers; it did not relieve employers of responsibility for the conditions that make self-care necessary at that scale. Both things are true and only one of them appears on the poster in the break room.

Vicarious traumatization is the concept described by Lisa McCann and Laurie Anne Pearlman in 1990, and its distinctive feature is that it is a change in worldview rather than a symptom cluster. The worker who has spent three years in child protection and now cannot let their own child walk to school is not exhausted. Their beliefs about how safe the world is have been revised by evidence, and the evidence was real.

What to actually do

1. Watch the caseload mix, not just the count. Twelve trauma cases is a different job from twelve mixed cases, and the mix is something a supervisor can change.
2. Use a measure. The Professional Quality of Life scale gives separate scores for compassion satisfaction, burnout, and secondary traumatic stress. Taking it twice a year is more informative than asking yourself how you are doing.
3. Protect the recovery interval. Not the annual holiday: the fifteen minutes after a risk session, the walk between the last home visit and the office, the refusal to open the laptop again at 9 pm.
4. Keep something in the week that is unrelated to suffering. This is the part that reads as a platitude and is not: workers who lose every non-clinical part of their identity are the ones who leave.
5. Know the impairment standard. Code standard 4.05 says that when personal difficulties interfere with professional judgement, you take remedial action, which may include reducing work, seeking help, or stepping back. Recognizing that in yourself is a professional competence rather than a failure.
6. Say it out loud in supervision. The supportive function only exists if somebody uses it.

Common misconceptions

• Supervision means reviewing cases with a senior colleague. Case review is the administrative function. Without examination of actual practice and attention to the worker, two thirds of it is missing.
• My account of the session is good enough data. Clinicians reliably report delivering more of the difficult components than recordings show, and nobody can report what they did not notice.
• Burnout is a personal resilience problem. It is an occupational phenomenon arising from workload, control, reward, community, fairness, and values, and self-care advice offered in place of a caseload change privatizes a structural failure.
• Vicarious trauma is just a severe form of burnout. It is a change in the worker's beliefs about safety, trust, and control, and it can occur in someone who is not exhausted at all.
• Supervision is a place to work on my own issues. It examines your reactions where they affect the work. Beyond that it is therapy, and a supervisor providing it has created a dual relationship.

The takeaway

• Supervision has administrative, educational, and supportive functions, and the administrative one expands to fill the hour unless somebody stops it.
• Self-report is the wrong data. Recordings, observation, and process recordings are how a supervisor sees the work rather than a description of it.
• Bring the case you are avoiding; the relief you feel when a client cancels is clinical information.
• Reflection-in-action and reflection-on-action are different skills, and only the second can be taught directly.
• Burnout comes from working conditions, secondary traumatic stress from trauma content, and vicarious traumatization from cumulative change in belief.
• Manage the caseload mix, measure, protect recovery intervals, and know the impairment standard before you need it.

Sources

1. Maslach, C., and Leiter, M. P. (2016). Understanding the burnout experience: Recent research and its implications for psychiatry. World Psychiatry, 15(2), 103-111. PubMed Central
2. McCann, I. L., and Pearlman, L. A. (1990). Vicarious traumatization: A framework for understanding the psychological effects of working with victims. Journal of Traumatic Stress, 3(1), 131-149.
3. Kadushin, A., and Harkness, D. (2014). Supervision in social work (5th ed.). Columbia University Press.
4. Schon, D. A. (1983). The reflective practitioner: How professionals think in action. Basic Books.
5. Wikipedia contributors. (n.d.). Compassion fatigue. Wikipedia. en.wikipedia.org

A photograph printed on ordinary paper

Session twelve. Renata Marchetti puts a piece of paper on the desk before she sits down. It is a photograph printed at home, slightly crooked, of a nine-year-old in a red kit on a wet field. Taken from the touchline, on a Saturday in October, by someone who drove there on the highway.

She says: I did not know what else to bring.

What happens in the next forty minutes is not a formality and it is not a wind-down. It is the last piece of clinical work in this case, and done badly it can undo a share of the eleven sessions before it.

Endings are treatment, not administration

Two reasons, and the second is the one people miss.

The first is that consolidation happens at the end or not at all. What a client takes away is not the sum of the sessions; it is the account they leave with about what happened and why. That account is constructed in the final sessions, and if you do not construct it deliberately, one gets constructed anyway, often a poor one.

The second is that endings land on top of every previous ending in a person's life. For a client whose father left without a conversation, or whose last three workers were reassigned mid-case, or who has been discharged from four services for missing appointments, this ending carries all of them. It is one of the few opportunities in clinical work to do a familiar thing differently: to end on time, in the open, having said what it was.

Remember: The ending is the part of the treatment the client will describe to the next clinician.

Six kinds of ending

Standard 1.16 of the NASW Code of Ethics governs this territory. It requires termination when services are no longer needed or no longer serve the client's interests. It requires reasonable steps to avoid abandoning clients still in need, including notification and arrangements for continuity. And it treats termination for non-payment as permissible only under conditions: the financial arrangement was clear, the client does not pose an imminent danger to self or others, and the clinical and social consequences have been considered.

Abandonment is a real professional finding, not a figure of speech. Its usual shape is not a dramatic walkout; it is a case that goes quiet, a clinician who is busy, and a chart with nothing in it for four months.

The ending most readers will do first

Your placement ends in May whether or not the work does. This is the ending you are most likely to face before any other, so plan it now.

Say the date at the first session. Not vaguely: I am a student here until the twelfth of May, so we have about nineteen weeks. Clients handle a known limit well and handle a surprise limit badly, and a client who learns in April that you were always leaving has learned that you withheld something for months.

Then, roughly four sessions out, name it again and start the ending work. Do not compress it into the last ten minutes of the last session, which is what anxious students do, and do not add extra sessions to postpone it, which is what anxious clinicians do.

The last four sessions, concretely

1. Name the ending and ask what it brings up. Some clients say nothing much, and some say something you did not expect. Both answers are information.
2. Review with the measures. Renata's PHQ-9 went 18, 16, 14, 11, 8. Her panic frequency went five a month to one. Put the numbers in front of her; a chart of five points does more than a paragraph of encouragement.
3. Attribute the change correctly. This is the step with a mechanism behind it. Ask what she did. When a client attributes improvement to their own actions rather than to the clinician or the medication, maintenance is more likely. So when Renata says you helped me so much, the accurate reply is not modesty: I gave you a straw and a piece of paper. You drove the on-ramp four times when you were terrified.
4. Build relapse prevention in her handwriting. Early warning signs, in her own words. The first two things she would start avoiding. What she would do in week one of a slide, written as instructions to herself. What her threshold for coming back is.
5. Say what the relationship was. Not a speech. Something true, brief, and specific: you told me things it took nerve to say, and I will remember the supermarket queue.
6. Practicalities. How to get back in, how long the record is kept, whether you can be contacted, what happens if she moves.

Endings that go wrong, and the tells

• The fade-out. Sessions get further apart, then stop. Nothing is said. Nothing is consolidated, and the client is left to decide privately whether they were dropped.
• The extra session that keeps arriving. Usually the clinician's difficulty, not the client's. If you find yourself finding reasons to continue, take it to supervision that week.
• Abrupt ending after a good session. Frequently a rupture nobody named, sometimes a fear of dependence, occasionally a client who got what they came for and is not interested in ceremony. Reach out, ask, and do not assume the worst reading or the best.
• The flight into health. A sudden announcement that everything is fine, arriving in the same session as something painful. Worth naming gently: I noticed that came up straight after we talked about your mother.
• Relief. Sometimes the clinician is glad it is over. That is worth examining in supervision rather than acting on, and it is very common with clients who were hard work.

When the client disappears

About one client in five leaves adult psychotherapy earlier than the clinician thinks they should, on the meta-analytic estimate by Joshua Swift and Roger Greenberg. It is not an exceptional event and it should not be treated as an insult.

What you do is a sequence you can defend later.

1. Attempt contact by the means the client agreed to. Two or three attempts across two or three weeks, not eight in a day.
2. Send a letter that says three things: that you have not heard from them, that they can return, and how. No reprimand.
3. If there was any risk indicator, raise it with your supervisor and follow the agency's protocol rather than deciding alone.
4. Write a closing summary: presenting problem, what was done, where things stood, what remains, and the outreach you attempted with dates.
5. Leave the door open in the record and in the letter.

Those steps are also what distinguishes a defensible closure from abandonment, and the dates are the part that does the work.

The last ten minutes

Renata's, more or less.

You: you brought a photograph. Tell me about the drive.
Renata: I nearly went the town way. I sat at the junction and I thought, if I go the town way I will go the town way for the rest of my life.
You: so what happened.
Renata: I indicated.
You: I want to be clear about something before you go. I did not do that. You did that, at a junction, on your own, with nobody watching.
Renata: you gave me the straw.
You: I gave you a straw. You drove.
Renata, after a pause: what happens if it comes back?
You: some of it probably will, for a while. That is why your page says what it says. Read the top line to me.
Renata: if I catch myself planning a route, that is the sign.
You: that is the sign. And you know how to get back in.

Then it ends, at the time it was always going to end, and she takes the photograph with her.

Common misconceptions

• Termination is the administrative closing of a case. It is the phase in which change is consolidated and attributed, and it can undo a share of the preceding work if it is skipped.
• Ending should be raised near the end. A session limit or a placement end date belongs in the first session, and the ending work starts roughly four sessions out.
• Attributing progress to the client is modesty. Clients who attribute change to their own actions are more likely to maintain it, which makes the attribution a clinical move rather than a courtesy.
• A client who stops attending has rejected the work. Roughly one in five adult clients ends earlier than the clinician expects, and the correct response is documented outreach and an open door.
• Adding a few more sessions is the safer, kinder choice. Sometimes. Often it is the clinician's difficulty with endings, and it should go to supervision before it goes into the diary.

What to remember

• The ending is clinical work: consolidation happens there, and it lands on top of every previous ending in the client's life.
• Name a limit at the start, and begin the ending work about four sessions before the last one.
• Review with the measures, attribute the change to what the client actually did, and write relapse prevention in their handwriting.
• Say something true and specific about the relationship, and cover the practicalities of getting back in.
• Standard 1.16 requires reasonable steps against abandonment, and abandonment usually looks like a quiet case and an empty chart.
• When a client vanishes, attempt contact, write, consult if there is risk, complete a closing summary with dates, and leave the door open.

Sources

1. National Association of Social Workers. (2021). Code of ethics, standard 1.16 on termination of services. NASW. socialworkers.org
2. Swift, J. K., and Greenberg, R. P. (2012). Premature discontinuation in adult psychotherapy: A meta-analysis. Journal of Consulting and Clinical Psychology, 80(4), 547-559.
3. Walsh, J. (2007). Endings in clinical practice: Effective closure in diverse settings (2nd ed.). Lyceum Books.
4. Wikipedia contributors. (n.d.). Psychotherapy. Wikipedia. en.wikipedia.org