Reading a claim about your mind

Sooner or later you'll meet a claim about your own mind that you'd like to act on: an app that says it's clinically proven, a headline about stress, a friend who swears by something. This last lesson is for that moment, and for a decision about your own life that comes after it.

It uses two habits and makes one refusal. The first habit is the institute's sort, which the closing lessons of Memory, Note-Taking and Sleep all used. The second is this course's question, what the comparison group got, and this lesson joins it to the sort. The refusal comes near the end, with the people who can do what a course cannot.

Four specimens, sorted

The sort has three bins, and it is used here rather than taught again. Measurable: a study could check the claim. Unmeasurable as stated: a word in it has no agreed measure, so nothing could count against it. False: it has been checked and it did not hold.

The four specimens below were written for this lesson, so none is a quotation from anywhere, and each verdict is this course's reading of the lessons it draws on.[11]

One: "Stress is bad for you." Unmeasurable as stated. Which stress, how much, bad for what, and compared with whom? Lesson 7 showed what it takes to pin a sentence like this down. Keller and colleagues had to split it into two survey questions, how much stress people reported and whether they believed it affected their health, and in the full text neither question on its own "independently predicted premature mortality". Only the combination did, and the authors write that the study "is unable to establish a causal relationship".[9] So the specimen isn't false. It's the outline of a claim, and nobody can test it until someone fills in the stress, the outcome and the comparison.

Two: "Most people who go through something traumatic never really get over it." False, as a statement of what lesson 7's review found. Galatzer-Levy, Huang and Bonanno pooled 54 studies of people after "potentially traumatic events", and this course read the abstract. The resilient course was the commonest, at an average of 65.7 percent across populations; recovery came next at 20.8 percent, then the chronic course at 10.6 percent and delayed onset at 8.9 percent.[8] Lesson 7 glossed resilience as staying roughly steady and recovery as a drop, then a climb back. The shares varied from study to study, and the review traced most of that to "substantive population differences".[8] So the specimen describes real people, only not most of them, and no average can say which course one person is on.

Three: "Cold showers build your resilience to stress." Measurable: a trial could put some people on cold showers and others on warm ones, and measure stress, or resilience as lesson 7's review did, in both. This course read nothing on cold water, so that's the whole verdict: measurable, and this course read nothing on it. Knowing that a claim could be tested tells you nothing about whether anyone has tested it, or how it came out. Notice too that the specimen has no comparison in it. Cold showers against what: warm ones, none at all, ten minutes of anything else? That is where the next habit comes in.

Four: "Meditation is as good as medication, so try an app before you bother a doctor."

You've split claims like this before. Sleep's closing lesson had one welded by "so", and lesson 7 took apart the idea that resilience is rare and can be trained. They matter more here, because the second half is so often advice about your mind, and the measured half lends it a standing the evidence never gave it. Watch for "so", "which means" and "that's why" in a sentence about your mind, and sort each side on its own.

The question that finishes the sort

Lesson 2 set this course's question beside the four the earlier courses on the Core added: what did the comparison group get? Joining it to the sort is simpler than it sounds.

For a measurable claim, the sort ends by asking what would settle it. For a claim about a method, the answer always has a comparison group in it, because "works" means "does better than something". So, in this course's way of joining the two, the comparator question is the sort's last step made specific.[11]

Try it on a sentence you might meet on a phone: "Clinically proven to reduce anxiety." The sort says measurable. The question then splits it into several different claims that share one set of words. Against a waiting list, it says the app beat being told to wait. Against an attention control, it says the app beat time and contact with no active ingredient. Against therapy, it says the app matched a therapist. Those are three findings, and a waiting list is the comparison lesson 2 ordered as the easiest to beat.

The question can also change what kind of verdict a claim gets. Suppose the only trials behind "as good as therapy" used a waiting list. Nobody's shown the claim wrong, so it isn't false. In this course's terms it's measurable and not yet measured, which is a different thing to know, and a more useful one.[11]

The course's main figures, by what the other group got

Every figure this course has printed came with its comparison, or with a note that the source did not give one. Lined up by that comparison, they make the question visible.

The table runs in lesson 2's order, from the comparisons that got least to those that got most. The order, the rule for placing rows and the last column are this course's; the figures and the comparisons are the sources'.[11] The rule: where a source's comparison groups were a mix, or were described only loosely, the row sits at the easiest comparison the source's words allow, because that is the one that could have flattered it most. A placebo pill is lesson 2's placebo control, given in a drug trial, so those rows sit just above the ones where the other group got another treatment. Fincham's breathwork row comes last, because its abstract does not say what the controls got.

Every row was read at abstract level except these: Mayo-Wilson and Hoge, the abstract in full; Goyal, the abstract and one paragraph of the Discussion; Kirsch and Cipriani, the abstracts and parts of the full texts; Stone, the abstract, Introduction, Discussion and Conclusions, and the Results paragraphs on the mixture model. The minus signs are the sources' own and do not all mean the same thing: some reviews report improvement as a fall in symptoms, and in the head-to-head rows the sign says which treatment came out ahead. So compare the sizes and set the signs aside.

Now read the first column against the second. Against another treatment the figures are small, 0.03 to 0.23 in size. Against anything less they run from 0.27 to 0.79 with no clear order among them: the waiting-list rows hold both the largest and the smallest. So the step lesson 2 predicted is roughly there, between another treatment and everything else, though the order is this course's.[11] Stone's row is in percentage points and Fincham's has no known comparison, so neither is in those ranges. And the table cannot show more than that, because its rows differ in method, people and outcome as well as in comparison, so the comparison is one of several things changing at once.

Then read the last column. None of its entries says anything about one reader. Each licenses a statement about groups, in a stated setting, against a stated comparison.

Two of the course's findings are not in the table, because nobody in them was assigned to a comparison group at all. Keller's hazard ratio of 1.43, from a paper lesson 7 read in the abstract and, in the full text, its results, sensitivity analysis and discussion, compared people who reported a lot of stress and believed it harmed them a lot with people who reported almost no stress and hardly any such belief, as they happened to answer a survey.[9] And the 65.7 percent from specimen two is a share, a base rate, with nothing set against it.[8] For both, the question's answer is that no comparison group was assigned, and that answer is a finding in itself: these are an association and a base rate, not the effect of anything.

What this course can't tell you

The question many readers bring to a course like this one is about themselves. Is what I'm living with ordinary stress, or something more?

This course can't tell you whether what you have is ordinary or clinical, and here's who can.

The reason is in the guidelines this course has read all along, in the sentences they give clinicians. For generalised anxiety disorder (GAD), CG113 tells a clinician, "For people who may have GAD, conduct a comprehensive assessment that does not rely solely on the number, severity and duration of symptoms, but also considers the degree of distress and functional impairment. [2011]"¹ For depression, NG222: "Conduct a comprehensive assessment that does not rely simply on a symptom count when assessing a person who may have depression, but also takes into account severity of symptoms, previous history, duration and course of illness."²

History first. NG222 tells clinicians to ask about "any history of mood elevation (to determine if the depression may be part of bipolar disorder)".² Lesson 1 noted that bipolar disorder is treated differently, and a reader looking only at their low spells might never think to raise the high ones.

Then the body. For people who arrive at an emergency department with a panic attack, CG113 says they should "undergo the minimum investigations necessary to exclude acute physical problems".¹ Lesson 1 explained why this course will not tell you chest pain is probably anxiety, and no course can examine anyone.

Then risk. NG222: "Always ask people with depression directly about suicidal ideation and intent. [2009]"² A clinician asks that question of a person and can act on the answer, while a page can only print the numbers, which is why they're at the top of every lesson.

And time. CG113's first step for anxiety includes "active monitoring", which lesson 1 glossed as a clinician checking back at a planned point.¹ A course cannot notice whether you're getting better or worse.

There's one more reason, and it comes from the evidence itself. Lesson 3 read a finding that for depression, "guided iCBT was associated with overall better outcomes in patients with baseline PHQ-9 greater than 9", while in milder symptoms the difference between guided and unguided was small.[5] The PHQ-9 is the nine-item depression questionnaire lesson 3 met. And NG222's table for more severe depression moves self-help down, with a note that "the potential advantages of providing other treatment choices with more therapist contact should be carefully considered first."² This course reads that as a reason the evidence cannot be used to skip the step it depends on:[11] which evidence applies to you depends on how severe things are, and that's the very thing the guidelines say needs an assessment.

The refusal runs both ways. This course will not tell you it's just stress, and it won't tell you you're ill, because neither is something a reader of averages can know about one person.

Who can

• If you're not safe right now: the lines in the callout at the top of this page.
• In the UK, urgent but not an emergency: the NHS says to "call 111 and select the mental health option".[3]
• In England: your GP, or NHS Talking Therapies, where "you can refer yourself to NHS talking therapies without speaking to a GP". To use it, "you need to be registered with a GP and must be aged 18 or over (or 16 or over in some areas)".[3]
• Elsewhere: a doctor, usually the one you'd see first for anything. As lesson 1 said, this course cannot map your health system.

Going to one of them isn't giving up on self-help, and it is not a sign that self-help or you have failed. In CG113, assessment is where care begins: its first step for generalised anxiety disorder is "Identification and assessment; education about GAD and treatment options; active monitoring".¹

Three things people get wrong

"If a method has evidence, it'll work for me." It's natural to hear "it works" as a promise, because that is how it is usually said. But every figure in the table is a gap between two groups. On lesson 2's conversion, an effect of 0.3 puts the average treated person ahead of about 62 percent of the comparison group, and that is not 62 percent of people helped.[11] Even for antidepressants, whose average advantage over placebo both sides of lesson 8's dispute measure, Stone and colleagues concluded, from a model they fitted, that it's "best understood as affecting a minority of patients".[10] The reverse holds too: a method that does not help you has not been shown wrong, and you have not been shown to be a particular kind of case.

"Needing more than self-help means self-help failed, or I did." It's an easy thing to feel, because self-help puts the work in your hands. But CG113 builds the next step in on purpose. Its step 3 is for "an inadequate response to step 2 interventions or marked functional impairment", so a clinician can place someone at step 3 without step 2 first.¹ The Cochrane authors on self-help for anxiety, Mayo-Wilson and Montgomery, say it "may be useful for people who are not able or are not willing to use other services", and in the same sentence that "for people who can access it, face-to-face cognitive behavioural therapy is probably clinically superior."[5]

"Somebody must have checked the app." This course read nothing on how app stores or anyone else vets mental-health apps, so it cannot tell you who, if anyone, checked yours. What it read was the evidence on programmes that were tested, and the Cochrane review of self-help for anxiety is dated: its searches ran to 1 January 2013, so it says nothing about any app released since. It reports that "Most interventions tested are not available to consumers", and that "Evidence regarding harm was lacking."[5] So the app you meet may not be one that was tested, and even for the ones that were, that review found little evidence either way on harm. Lesson 3's exercise, holding a self-help book or app up to NICE's specification, shows the kind of check you can run yourself.

Practice

Connections

Back. To the whole course. Lesson 1 drew the line between ordinary and clinical, and its reasons to get help are the first ending of the decision. Lesson 2 gave the question the table is sorted by. Lesson 3 read the evidence closest to what you've been doing, and where a guide matters most. Lesson 4 separated a package that works from a single technique tested on its own. Lesson 5 asked one question three ways and traced a press release. Lesson 6 showed a comparison across trials that misleads, a trial five years later that tested it for anxiety disorders, and who a method is not for. Lesson 7 followed three popular ideas and watched an association harden into a cause. Lesson 8 set one number beside two readings, and showed that a threshold is a judgement and an average can hide a group.

Forward. The comparator question is what Statistics for Citizens, The Scientific Method and Medical Literacy will need from this course; all three are planned later on the Core path and not yet written. Stepped care is for Medical Literacy too, and the difference between a package and a component is for Habits and Addiction, also planned.

Go deeper

• NICE NG222, section 1.2, which holds the recommendations on assessment and risk that this lesson quoted. This course read sections 1.1 and 1.2 in full, among others. Read 1.2.6 to 1.2.9 to see what an assessment is meant to cover, in the guideline's own words.
• NICE CG113's appendix, "Assessing generalised anxiety disorder", read in full for this course. The clearest account of what a clinician weighs for anxiety.
• NHS Talking Therapies, for readers in England. This course read the opening and the section on how to get them.
• Mayo-Wilson and Montgomery, 2013, the Cochrane review of self-help for anxiety, free at PubMed Central. The abstract was read in full here. Its conclusions are a fair account of what self-help can and can't claim, from the people who pooled it.

Sources

1. National Institute for Health and Care Excellence, Generalised anxiety disorder and panic disorder in adults: management, CG113, 2011, updated
   2020. Read: the Recommendations chapter to 1.3.41, and the appendix on assessing GAD. Supports: recommendation 1.2.6, the emergency-department sentence on physical problems, the stepped-care model with step 1's entry and active monitoring, the reason for step 1 (1.2.10), and step 3's entry.
2. National Institute for Health and Care Excellence, Depression in adults: treatment and management, NG222, 2022. Read in part: the definitions, sections 1.1 and 1.2 in full, 1.4.1 to 1.4.21, section 1.5 with Table 1, and 1.6.1 with Table 2's row order and its guided self-help and group exercise rows. Supports: recommendations 1.2.6, 1.2.7 and 1.2.8, the group mindfulness row's caution, the note on more severe depression, and recommendation 1.4.12.
3. NHS, "Where to get urgent help for mental health", read on 23 September 2026; and NHS, "NHS talking therapies", read in part: the opening and "How to get talking therapies". Supports: the two sentences on mental health emergencies, 111's mental health option, and self-referral in England with who can use it.
4. P. Cuijpers and colleagues, "Cognitive behavior therapy vs. control conditions, other psychotherapies, pharmacotherapies and combined treatment for depression", World Psychiatry 22(1), 2023, doi 10.1002/wps.21069. Read: the abstract. Supports: the table's 0.79 row with its controls and its 0.06 row against other psychotherapies.
5. E. Mayo-Wilson and P. Montgomery, "Media-delivered cognitive behavioural therapy and behavioural therapy (self-help) for anxiety disorders in adults", Cochrane Database of Systematic Reviews 2013, CD005330, doi 10.1002/14651858.CD005330.pub4 (read: the abstract in full); E. Karyotaki and colleagues, JAMA Psychiatry 74(4), 2017, doi 10.1001/jamapsychiatry.2017.0044, and JAMA Psychiatry 78(4), 2021, doi 10.1001/jamapsychiatry.2020.4364 (read: the abstracts). Supports: the table's 0.67, -0.23 and 0.27 rows with their comparisons, the search date, the conclusion on who self-help may suit, availability to consumers and harm, and guided against unguided by severity.
6. A. J. Clegg and colleagues, "Exercise for depression", Cochrane Database of Systematic Reviews 2026, CD004366, doi 10.1002/14651858.CD004366.pub7. Read: the abstract. Supports: the three exercise rows, their certainty grades and the head-to-head trial counts.
7. J. Galante and colleagues, Nature Mental Health 1(7), 2023, doi 10.1038/s44220-023-00081-5 (read: the abstract); M. Goyal and colleagues, JAMA Internal Medicine 174(3), 2014, doi 10.1001/jamainternmed.2013.13018 (read: the abstract, and one paragraph of the Discussion); E. A. Hoge and colleagues, JAMA Psychiatry 80(1), 2023, doi 10.1001/jamapsychiatry.2022.3679 (read: the abstract in full); G. W. Fincham and colleagues, Scientific Reports 13(1), 2023, doi 10.1038/s41598-022-27247-y (read: the abstract); and N. T. Van Dam and colleagues, "Mind the Hype", Perspectives on Psychological Science 13(1), 2018, doi 10.1177/1745691617709589 (read: the abstract, introduction and the three "Consensus about..." sections). Supports: the mindfulness, meditation, MBSR and breathwork rows, the reading of specimen four's first half, and Van Dam's phrase.
8. A. M. Kunzler and colleagues, "Psychological interventions to foster resilience in healthcare professionals", Cochrane Database of Systematic Reviews 2020, CD012527, doi 10.1002/14651858.CD012527.pub2; and I. R. Galatzer-Levy, S. H. Huang and G. A. Bonanno, "Trajectories of resilience and dysfunction following potential trauma", Clinical Psychology Review 63, 2018, doi 10.1016/j.cpr.2018.05.008. Read: both abstracts. Supports: the resilience-training row and checkpoint with its conclusion, and the four trajectory shares in specimen two and the review's account of their spread.
9. A. Keller and colleagues, "Does the perception that stress affects health matter? The association with health and mortality", Health Psychology 31(5), 2012, doi 10.1037/a0026743, PMC3374921. Read: the abstract, and in the full text the results with Table 3, the sensitivity analysis and the discussion. Supports: specimen one's two questions and their result, the authors' sentence on causation, and the hazard ratio with its reference group.
10. I. Kirsch and colleagues, PLoS Medicine 5(2), 2008, doi 10.1371/journal.pmed.0050045; A. Cipriani and colleagues, Lancet 391(10128), 2018, doi 10.1016/S0140-6736(17)32802-7; and M. B. Stone and colleagues, BMJ 378, 2022, e067606, doi 10.1136/bmj-2021-067606. Read: Kirsch's and Cipriani's abstracts in full and parts of each full text, as lesson 8 lists them; Stone's abstract, Introduction, Discussion and Conclusions, and the Results paragraphs on the mixture model; not the Methods. Supports: the two antidepressant rows, the shares of Large responses, and the sentence on a minority of patients.
11. This course's own constructions, labelled where they appear. The four specimens and their sorting, including specimen three's verdict; the extension of the NHS's sentence from emergencies to ordinary appointments; the joining of the comparator question to the sort, and the reading that a claim can be measurable and not yet measured; the table's order, its rule for placing mixed and loosely described comparisons, its last column, and the reading of its ranges; the list of what an assessment has that a reader alone does not; the reading that the evidence depends on a severity only an assessment can judge; the 62 percent conversion, from lesson 2; and the view that a question kept open has costs of its own.