Show the full module text
Module 5 — The Instruments You Are Likely to Meet
What each tool actually measures, who fills it in, and why no single one of them is required — including the one most people believe is.
The Assessment · Part Two — The process
The big idea
Short on capacity today? The big idea: Somewhere between booking and the first appointment, most people go looking for the list. Which test is the real one? Which result decides it?
There is no such test, and this module exists to say so precisely rather than reassuringly.
No instrument is mandatory for an adult autism or ADHD diagnosis. Not the observation schedule, not the caregiver interview, not any rating scale, and not neuropsychological testing. The clinical guidance that governs adult assessment in England says “consider using” — a permissive verb — and then names several tools in the same breath, without ranking them. What guidance does forbid is the opposite error: a diagnosis “should not be made solely on the basis of rating scale or observational data.”
So the question is not which tool you get. It is whether the tools you get are asking different people, in different ways, about different things — and whether the person reading them can hold the disagreements. That is what this module equips you to see.
Step 1 — The lesson
A. Three kinds of tool, and the question nobody thinks to ask
Every instrument in this field is one of three things, and the difference matters more than the brand name.
Observation. The clinician sets up situations and codes what you do. You are not asked to report anything; you are watched doing it. Measures behaviour now, in a room with one stranger, for an hour or two.
Interview. A trained person works through a defined set of questions and codes the answers — either with you, or with someone who knows you. Measures a life as described, reaching backwards.
Rating scale. Somebody ticks boxes. Cheap, fast, and entirely dependent on who is holding the pen. Measures a perception.
Then the second axis, which is the one people miss: who supplies the information. You, an informant, or the clinician. The same instrument can exist in more than one version — and which version you are handed changes what it measures.
When two tools disagree, it is usually because they asked different people different questions in different modalities. That is signal, not error. Module 1’s whole argument is that the assessor’s job is to weigh those disagreements, and this is where the disagreements come from.
Which gives you the single most useful question to put to a service about any form you are sent: “Which version of this is it, and who fills it in?”
Diagram — A · Watched, asked, ticked. Every instrument in this field is one of three things, and the difference matters more than the brand name. Observation: the clinician sets up situations and codes what you do. You are not asked to report anything; you are watched doing it. It measures behaviour now, in a room with one stranger, for an hour or two — which on a timeline running from childhood to this hour is a very short block at the far right. Interview: a trained person works through a defined set of questions and codes the answers, either with you or with someone who knows you. It measures a life as described, reaching backwards across the whole span — as reported, though, not as recorded. Rating scale: somebody ticks boxes. Cheap, fast, and entirely dependent on who is holding the pen. It measures a perception, which is why it has no block on the timeline at all. Then the second axis, which is the one people miss: who supplies the information — you, an informant, or the clinician. The same instrument can exist in more than one version, and which version you are handed changes what it measures. The SRS-2 adult forms exist in two, a self-report and a relative or other report. Which gives you the single most useful question to put to a service about any form you are sent: which version of this is it, and who fills it in? And when two tools disagree, it is usually because they asked different people different questions in different modalities. That is signal, not error. An assessment is sound when several independent lines of evidence, gathered in different ways from different people, point the same way — and when the person weighing them can say honestly what happened where they did not.
B. What you might actually meet
Not a required list. A field guide. Any competent adult assessment might use two or three of these, or none of them, and still be sound.
Diagram — B · A gate with no fence. Somewhere between booking and the first appointment, most people go looking for the list. Which test is the real one? Which result decides it? There is no such test. No instrument is mandatory for an adult autism or ADHD diagnosis: not the observation schedule, not the caregiver interview, not any rating scale, and not neuropsychological testing. The clinical guidance that governs adult assessment in England says consider using — a permissive verb — and then names several tools in the same breath, without ranking them. What guidance does forbid is the opposite error: a diagnosis should not be made solely on the basis of rating scale or observational data, and biological tests, genetic tests and neuroimaging are not to be used routinely for diagnostic purposes. So the question is not which tool you get. It is whether the tools you get are asking different people, in different ways, about different things — and whether the person reading them can hold the disagreements. If a service tells you an assessment is not valid without a particular instrument, that is a claim about their protocol, not about the evidence.
Instrument. ADOS-2. Kind. Observation. Who supplies the information. Clinician codes; you take part. What it is reaching for. Current social communication and restricted/repetitive behaviour in a structured setting. Module 4 is the one used with verbally fluent adolescents and adults.
Instrument. ADI-R. Kind. Interview. Who supplies the information. A caregiver who knew you in early childhood. What it is reaching for. Developmental history against autism criteria.
Instrument. DISCO. Kind. Interview. Who supplies the information. Clinician with you and/or an informant. What it is reaching for. A broad developmental and behavioural picture; named in English guidance alongside the two above.
Instrument. 3Di-Adult. Kind. Interview. Who supplies the information. An informant. What it is reaching for. A brief DSM-5-mapped developmental account (Module 4).
Instrument. MIGDAS-2. Kind. Interview (qualitative). Who supplies the information. Clinician with you; separate parent/caregiver and teacher questionnaires. What it is reaching for. A narrative description rather than a score, built through sensory-based materials. Has a dedicated adult form for verbally fluent adults.
Instrument. DIVA-5. Kind. Interview (semi-structured). Who supplies the information. Clinician with you, and an informant where there is one. What it is reaching for. Adult ADHD criteria, mapped item by item to DSM-5, in childhood and now.
Instrument. SRS-2. Kind. Rating scale. Who supplies the information. Adult forms exist in two versions: self-report, and relative/other report. What it is reaching for. Autistic traits and social responsiveness, quantified.
Instrument. SPM-2. Kind. Rating scale. Who supplies the information. An adult version exists; the only self-report form found in the literature read here is the adolescent one — ask which form you are given and who completes it. What it is reaching for. Sensory processing across everyday environments.
Instrument. CAARS-2. Kind. Rating scale. Who supplies the information. Self-report form; observer versions are the usual pairing — ask. What it is reaching for. Adult ADHD symptom dimensions.
Instrument. Brown EF/A Scales. Kind. Rating scale. Who supplies the information. Ask which version you are given. What it is reaching for. Executive function and attention as experienced day to day.
Instrument. AQ · RAADS-R · CAT-Q · EQ · GQ-ASC. Kind. Rating scale (screening). Who supplies the information. You. What it is reaching for. Trait level, camouflaging, empathy, gendered presentation — starting points, not conclusions.
Three notes on that table.
On the screeners. The last row is the territory of the Autistic Self-Discovery and ADHD Self-Discovery courses, which teach each one with its limits shown. No cut-off is quoted here on purpose. If an assessment begins with one of these, that is normal; if it ends with one, that is not an assessment.
On MIGDAS-2. Its publisher describes it as a qualitative interview process for children, adolescents and adults, in which systematic interviews with parents, caregivers, teachers and the person being evaluated produce a narrative description; it uses ordinary sensory-based materials rather than a fixed set of tasks. It is unusual in this list because its output is prose rather than a number, which some people find far easier to recognise themselves in. It is also unusual in a second way: a search of the biomedical literature index used for this module returned no independent validation study for it. That is worth knowing and it is not a reason to refuse it — a qualitative interview is not the kind of thing a sensitivity figure describes well.
On Brown EF/A. The same search returned nothing indexed for it either. Again: ask.
C. The instrument you have heard of, and what the adult evidence says
The ADOS-2 is the tool most people have heard of and the one most often described as the gold standard. Module 1 already gave the correction; this is the fuller version, because the belief is persistent and it costs people money.
In a study of 88 adults referred to a specialist adult autism service, ADOS-2 Module 4 scores “failed to predict the diagnostic outcome” — the authors’ own words. Overall sensitivity was 92 per cent and specificity 57 per cent. It caught nearly everyone who went on to be diagnosed, and it also caught a great many people who did not. The paper’s conclusion is a single sentence: the authors “recommend clinicians are cautious when interpreting results of the ADOS-2 assessment.”
Diagram — C · The marks on the door jamb. The ADOS-2 is the tool most people have heard of and the one most often described as the gold standard. In a study of 88 adults referred to a specialist adult autism service, ADOS-2 Module 4 scores failed to predict the diagnostic outcome — the authors' own words. Overall sensitivity was 92 per cent and specificity 57 per cent. It caught nearly everyone who went on to be diagnosed, and it also caught a great many people who did not: at 57 per cent specificity, of every hundred adults who did not go on to be diagnosed, forty-three scored above the line anyway. The paper's conclusion is a single sentence — the authors recommend clinicians are cautious when interpreting results of the ADOS-2 assessment. A second finding is stranger and more instructive. Using data from 2,310 participants aged 4 to 72, 46 per cent of them autistic, a machine-learning study built classifiers from the combined ADOS and ADI-R data and from ADOS data alone. For adolescents and adults, specificity was lower than for children; and in that sample the classifier using only ADOS items performed better than the classifier using both instruments, leading the authors to infer that particularly in adolescents and adults the ADI-R may play a lesser role than current behaviour observations. Read those two studies together and the conclusion is not that the ADOS-2 is bad, or that the ADI-R is unnecessary. It is that the combination everyone calls the gold standard was validated mainly in children, that in adults its parts do not behave the way the reputation says they do, and that no study has established any pairing of tools as the required one. There is a practical corollary: if you cannot produce a caregiver for an ADI-R, which many adults cannot, the adult evidence is the reason that is not the obstacle it sounds like.
The second finding is stranger and more instructive. Using data from 2,310 participants aged 4 to 72 — 46 per cent autistic, 54 per cent not — a machine-learning study built classifiers from the combined ADOS and ADI-R data, and from ADOS data alone. For adolescents and adults, specificity was lower than for children. And in the adolescent and adult sample, the classifier using only ADOS items performed better than the classifier using both instruments. The authors’ inference is worth quoting: particularly in adolescents and adults, “the ADI-R may play a lesser role than current behavior observations.”
Read those two together and the conclusion is not “the ADOS-2 is bad” or “the ADI-R is unnecessary”. It is this:
The combination everyone calls the gold standard was validated mainly in children. In adults, its parts do not behave the way the reputation says they do — and no study has established any pairing of tools as the required one.
There is a practical corollary. If a service tells you an assessment is not valid without a particular instrument, that is a claim about their protocol, not about the evidence. And if you cannot produce a caregiver for an ADI-R — which many adults cannot — the literature above is the reason that is not the obstacle it sounds like.
D. Neuropsychological testing is not required
This one deserves its own section because it is where the money goes.
Neuropsychological testing means timed, performance-based tasks: continuous performance tests, working-memory tasks, processing-speed measures, executive-function batteries. They are sometimes sold as the objective layer under all the questionnaires — the part that cannot be faked.
They do not do that job. In a study of 108 patients in a clinical psychiatric population, of whom 60 were diagnosed with ADHD, researchers examined eight neuropsychological tests alongside a structured diagnostic interview and a self-report scale. The finding, in the authors’ own words: “Neuropsychological tests have a poor ability to discriminate between patients diagnosed with ADHD and patients not diagnosed with ADHD.”
All the instruments showed poor discriminative ability except the interview. DIVA reached sensitivity 90.0 and specificity 72.9. Adding continuous-performance-test variables to the DIVA raised specificity to 83.3 — a real gain, and a supplementary one. The structured conversation outperformed all eight tests.
Diagram — D · Money in, tape out. Neuropsychological testing means timed, performance-based tasks: continuous performance tests, working-memory tasks, processing-speed measures, executive-function batteries. They are sometimes sold as the objective layer under all the questionnaires — the part that cannot be faked. They do not do that job. In a study of 108 patients in a clinical psychiatric population, of whom 60 were diagnosed with ADHD, researchers examined eight neuropsychological tests alongside a structured diagnostic interview and a self-report scale. The finding, in the authors' own words: neuropsychological tests have a poor ability to discriminate between patients diagnosed with ADHD and patients not diagnosed with ADHD. All the instruments showed poor discriminative ability except the interview. The structured interview reached sensitivity 90.0 and specificity 72.9, and adding continuous-performance-test variables to it raised specificity to 83.3 — a real gain, and a supplementary one. The structured conversation outperformed all eight tests. The same logic runs on the autism side, where guidance is explicit in a different direction: do not use biological tests, genetic tests or neuroimaging for diagnostic purposes routinely as part of a comprehensive assessment. So neuropsychological testing can be genuinely useful — for a suspected learning difference, for an occupational-health question, for a specific functional query about memory or processing speed. It is not a diagnostic requirement for adult autism or adult ADHD, and a service that says otherwise should be asked which finding it is relying on.
The same logic runs on the autism side, where guidance is explicit in a different direction: do not use biological tests, genetic tests or neuroimaging for diagnostic purposes routinely as part of a comprehensive assessment.
So: neuropsychological testing can be genuinely useful — for a suspected learning difference, for an occupational-health question, for a specific functional query about memory or processing speed. It is not a diagnostic requirement for adult autism or adult ADHD, and a service that says otherwise should be asked which finding it is relying on.
E. Convergence is what makes an assessment sound
Here is the whole module in one movement.
Module 1 set out the five evidence sources: your self-report, your developmental history, informant accounts, structured instruments, and direct observation. Sections B to D of this module have now shown that each individual source is weaker than its reputation. Self-report bends with what you currently believe. Recall is poor in everyone. Informant accounts are strong when they confirm and much weaker when they do not. The best-known observation schedule had 57 per cent specificity in adults. Neuropsychological tests could not tell the groups apart at all.
Every one of those is a real limitation, and none of them is an argument against assessment. Because the claim was never that any single source is decisive. A published set of clinical recommendations for diagnosing autism in adults puts it plainly: clinicians “should utilize a multi-trait assessment method.”
An assessment is sound when several independent lines of evidence, gathered in different ways from different people, point the same way — and when the person weighing them can say honestly what happened where they did not.
That is why the answer takes hours instead of minutes. It is also why “which test did you use?” is a less revealing question than “what did you do when the sources disagreed?”
Diagram — E · One line of email. One: for every form you are sent, ask which version it is and who fills it in. Self-report or informant report changes the meaning of the result entirely; this is the single highest-value question in the module and it takes one line of email. Two: do not shop for a service by instrument list. A longer list is not a better assessment. What you are looking for is more than one kind of evidence — something observed, something asked, something from someone else — not more brands. Three: if you cannot provide a caregiver, say so, and do not treat it as fatal. The caregiver interview is one route to developmental history, not the only one, and in the adult data it carried less weight than current observation. Four: do not pay extra for neuropsychological testing on the assumption that it settles anything. Ask what specific question it will answer that the rest of the assessment cannot; a suspected learning difference or a workplace question about processing speed is a fine reason, and to be thorough is not. Five: take any screener scores you already have, with dates, and expect them to be a starting point rather than a conclusion. Six: if you score below a cut-off and the assessor still proceeds, that is not a mistake — cut-offs are calibrated on populations, not on you. And the question that reveals most about a service is not which test did you use, but what did you do when the sources disagreed?
→ The two courses named in section B each open with the same module, on how to read a screener at all: Autistic Self-Discovery, Module 1 and ADHD Self-Discovery, Module 1.
F. What helps
1. For every form you are sent, ask which version it is and who fills it in.
Self-report or informant report changes the meaning of the result entirely. This is the single highest-value question in the module and it takes one line of email.
2. Do not shop for a service by instrument list.
A longer list is not a better assessment. What you are looking for is more than one kind of evidence — something observed, something asked, something from someone else — not more brands.
3. If you cannot provide a caregiver, say so, and do not treat it as fatal.
The caregiver interview is one route to developmental history, not the only one, and in the adult data it carried less weight than current observation. Module 4 covers the alternatives.
4. Do not pay extra for neuropsychological testing on the assumption that it settles anything.
Ask what specific question it will answer that the rest of the assessment cannot. If there is a good answer — a suspected learning difference, a workplace question about processing speed — that is a fine reason. “To be thorough” is not.
5. Take any screener scores you already have, with dates — and expect them to be a starting point.
The Self-Discovery courses teach each instrument with its limits shown, which is the right frame to arrive with.
6. If you score below a cut-off and the assessor still proceeds, that is not a mistake.
Cut-offs are calibrated on populations, not on you. Section C is what happens when a threshold is treated as an answer.
Module 6 is the outcome conversation — what happens in the session where you are given the answer, what a yes contains beyond the word, and why a no is more revisable than a yes.
Up next
Module 6 — The Outcome Conversation