The ADI-R Test: The Autism Interview About Your Childhood
Ninety-three questions, up to two and a half hours, and they are not asked of you. They are asked of someone who knew you when you were four.
The ADI-R Test is the half of a formal autism assessment that most people are unprepared for. Not because it is difficult, but because of who it involves. You cannot do it alone. It needs a parent, a carer, or someone else who watched you grow up and can remember what you were like before you learned to compensate.
For a four-year-old being assessed, that is straightforward. For a thirty-eight-year-old whose parents have died, it is the thing that stops the process.
What is the ADI-R Test?
The ADI-R Test, properly the Autism Diagnostic Interview Revised, was developed by Michael Rutter, Ann LeCouteur and Catherine Lord, and is published by Western Psychological Services. It is a standardised interview conducted by a trained clinician with someone who knows the person being assessed well, almost always a parent.
Ninety-three items. They sit across eight content areas: background and family, an overview of current behaviour, early development and developmental milestones, language acquisition and any loss of language or skills, current language and communication, social development and play, interests and behaviours, and clinically relevant behaviours such as aggression, self-injury and possible epileptic features.
Those feed into three domains that map onto the diagnostic criteria:
- Language and communication
- Reciprocal social interaction
- Restricted, repetitive and stereotyped behaviours and interests
A fourth consideration is onset, specifically whether developmental abnormality was evident before 36 months.
Unlike most instruments described on this site, the ADI-R Test does not produce a scale score compared against norms. It produces a categorical result: whether cutoffs in each domain were met.
The reference period
Here is the feature that surprises people. The diagnostic algorithm does not ask what someone is like now. It asks about the period when behaviour was most unusual, typically between the fourth and fifth birthdays.
There is a reason for that. Autism is developmental, and the clearest signal is usually in early childhood, before compensation, schooling and social pressure have reshaped behaviour. Asking about age four gets closer to the underlying pattern than asking about age forty.
It also means the interview depends on a memory that may be thirty or more years old.
What happens during the ADI-R Test
Expect two hours or more. During the ADI-R Test the clinician works through the questions in order, asking your informant for specific examples rather than general impressions. The probing is relentless by design.
A question will not be “was he sociable as a child.” It will be closer to: when he was four or five, did he bring things to show you, just to share them rather than to ask for help. Did he point at aeroplanes. When you came home, did he look up. How did he play with other children at that age, specifically, can you describe an occasion.
Informants frequently find this exhausting and sometimes upsetting. Parents are being asked, in effect, to re-examine years they may have found difficult, and some feel they are being examined themselves. They are not. A good clinician will say so at the start.
How the ADI-R Test is scored
The ADI-R Test is coded item by item, usually 0 to 3, where 0 means the behaviour described was not present and 3 means it was present and severe. Some codes sit outside that range for cases where the question does not apply.
The coded items feed a comprehensive algorithm form, which generates one of five age-specific algorithms: two diagnostic algorithms based on developmental history, and three current-behaviour algorithms used for planning support rather than for diagnosis.
For a result consistent with autism, cutoffs must be met in all three domains, plus the onset criterion. Meeting two of three is not a partial result. It is a result that does not meet criteria, which is one of the reasons the ADI-R Test is a demanding instrument to pass.
Internal consistency across domains has been reported between 0.54 and 0.84, which is a wide range and weak at the lower end.
How accurate is the ADI-R Test?
The accuracy of the ADI-R Test varies sharply by age, and the pattern is the opposite of what people assume.
In school-age children, the instrument performs well, particularly at distinguishing autism from intellectual disability without autism. Used together with an observational assessment, agreement with clinical diagnosis reaches 80 percent or better in some samples.
In toddlers it performs less well. Research has repeatedly found that very young children have not yet developed enough repetitive behaviour to meet the third domain, so strict application of the cutoffs misses cases. Agreement with other measures in toddlers has been reported as poor.
In adults, there is a specific and well-documented problem.
The two halves of the gold standard fail in opposite directions
This is worth putting plainly, because it rarely appears anywhere.
In adults, the ADOS-2 Test has high sensitivity and poor specificity. It catches nearly everyone and flags a great many people who are not autistic. The ADI-R Test in adults has the reverse profile: good specificity, poor sensitivity. It rarely says yes when the answer is no, and frequently says no when the answer is yes.
Used together, those two error patterns partly cancel out, which is the argument for using both. But it also means that whichever one your service happens to rely on more heavily will shape your outcome. If an adult assessment leans on developmental history and your informant’s recall is thin, a real diagnosis can be missed. If it leans on observation, the opposite error becomes likelier.
The informant problem
This is the ADI-R Test barrier that affects adults more than any psychometric issue, and it is the thing people most often ask about.
The instrument requires someone who remembers your early childhood in detail. For many adults seeking assessment, that person is unavailable. Parents may have died, may be estranged, may have dementia, or may simply not remember. Some adults were in care. Some have parents who are themselves likely autistic and who recall nothing unusual because nothing seemed unusual to them.
There are also parents who do remember and do not want to. A diagnosis can read as an accusation about how a child was raised, and some informants minimise accordingly.
What services do about this varies. Some proceed without an informant and rely on school reports and the applicant’s own developmental account. Some use an abbreviated history-taking instead of the full interview. Some will not assess adults without an informant at all, which is worth checking before you join a waiting list.
Who the ADI-R Test is for
The ADI-R Test is used for children and adults with a mental age above roughly two years. Published age ranges run from toddlerhood to old age.
In practice it is most often used in three situations: a full diagnostic assessment of a child, a complex case where the picture is unclear and developmental history will settle it, and research, where it is near-universal because standardisation matters more than speed.
It is less likely to be used in a straightforward adult assessment, where time pressure and the informant problem together make it impractical. Many adult services use a structured developmental history interview of their own design instead.
Strengths and weaknesses of the ADI-R Test
Strengths
- Captures developmental history, which no self-report instrument can
- Not vulnerable to masking, because it asks about behaviour before compensation developed
- Standardised and widely used, so results travel between services and studies
- Requires specific examples rather than impressions
- Covers onset, which is a diagnostic criterion other instruments skip
- Strong performance in school-age children, especially alongside observation
Weaknesses
- Requires an informant who knew you as a young child
- Poor sensitivity in adults, with one study classifying only 55 percent correctly
- Underperforms in toddlers, who often lack the repetitive behaviours the third domain needs
- Ninety to 150 minutes is a substantial demand on clinician time
- Internal consistency as low as 0.54 in some domains
- Depends on recall that may be decades old and emotionally loaded
- All three domains must be met, so near misses count as negatives
Controversies around the ADI-R Test
Retrospective memory
The whole ADI-R Test rests on an informant accurately recalling specific behaviours from a particular two-year window, often long ago. Memory does not work that way. It reconstructs. It is shaped by what has happened since, and it is influenced by knowing what the interview is for. This is the central methodological criticism and it has no clean solution.
Whether the cutoffs suit DSM-5
The algorithms were built around earlier diagnostic criteria. Work on DSM-5 based algorithms has continued, including recent revisions for children aged 4 to 17, which indicates the original thresholds were not a perfect fit for current criteria.
Equity of access
An instrument that requires an available, willing and lucid parent systematically disadvantages people who were in care, who are estranged from family, whose parents have died, and who are older. Those are not small groups, and the barrier is practical rather than clinical.
The cost of the pair
Using both this and an observational assessment properly takes four hours of trained clinician time or more. That is a real part of why adult autism waiting lists are measured in years, and why many services have moved to shorter protocols of their own design.
How the ADI-R Test compares to other instruments
| Instrument | Asks | Of whom | Covers |
|---|---|---|---|
| ADI-R Test | Structured interview | A parent or carer | Your childhood |
| ADOS-2 Test | Structured observation | You, watched | Your behaviour now |
| M-CHAT Screening Test | Short questionnaire | A parent | A toddler, now |
| RAADS-R Test | Self-report | You | Your whole life, as you recall it |
| RAADS-14 Screen | Self-report | You | A brief lifetime screen |
| AQ-50 Test | Self-report | You | Current traits |
| AQ-10 Test | Self-report | You | Whether to refer at all |
| CAT-Q Test | Self-report | You | How much you mask |
The closest thing to a self-administered substitute is a lifetime-format questionnaire. The RAADS-R Test asks whether each item was true now, true only now, true only before 16, or never true, which is a crude version of the same developmental question. It is self-report and therefore vulnerable to everything the interview avoids, but it is something you can bring to an assessment when no informant exists.
If more than autism may be involved, which is common in adult referrals, our Neurodivergent Test screens seventeen conditions and separates traits from impact. For a shorter autism-specific screen, the Free Autism Test takes about six minutes.
Frequently asked questions about the ADI-R Test
Can I take the ADI-R Test online or by myself?
No. It is an interview conducted by a trained clinician with someone who knew you as a child. There is no self-administered version and there cannot be, because the whole design rests on an external account of your early development.
What if my parents have died, or we are not in contact?
This is the most common question about this instrument and there is no universal answer. Some services proceed with school reports, medical records and your own developmental account. Some use an older sibling, an aunt, a long-standing family friend. Some will not assess without an informant. Ask the service before you join the list, because it determines whether the referral is viable.
Can a partner or spouse be the informant?
For the current-behaviour sections, often yes. For the diagnostic algorithm, no, unless they knew you as a young child. The algorithm asks about ages four to five, and a partner who met you at twenty-five cannot answer those questions.
How long does it take?
Ninety minutes to two and a half hours including scoring. Expect a long appointment, and expect your informant to find it tiring.
What if my parent does not remember?
Say so. A clinician would far rather record that a question could not be answered than code a guess. Bring documents to fill the gaps, and remember that an informant who cannot recall is a common situation, not a failure.
What if my parent downplays everything?
It happens. Often enough that assessors expect it. Some parents genuinely did not notice. Some noticed and feel implicated. Some are autistic themselves and found nothing unusual. If you are worried about this, tell the clinician privately beforehand. They can weight the account accordingly and probe harder for specifics.
Does the ADI-R Test diagnose autism on its own?
It is not supposed to. It produces one structured input. Best practice combines it with direct observation such as the ADOS-2 Test, plus clinical judgement and consideration of alternative explanations.
Why does it ask about age four or five rather than now?
Because autism is developmental and the signal is clearest before compensation sets in. A forty-year-old has had decades to build workarounds. A four-year-old has not.
Is it accurate for adults?
Less so, and this is well documented. One study of adults found good specificity but poor sensitivity, with 55 percent correctly classified, and attributed the problem to reliance on decades-old recall. A negative result in an adult assessment is weaker evidence than it looks.
How much does it cost?
Nothing to you in a public health system. Privately it is bundled into an assessment fee running into four figures. The kit costs clinicians around $503, with the full training package considerably more.
Who can administer it?
Qualified clinicians with specific training. The publisher restricts purchase to a professional qualification level, and research use requires a further reliability standard. Your GP cannot do it.
My child met two of three domains. What does that mean?
Formally, criteria were not met. In practice it is information rather than a verdict, and a good assessment will weigh it against observation, school reports and clinical judgement rather than treating the algorithm as final. Ask for the reasoning in writing.
Is there a shorter version?
No official short form in general use, though many services have developed their own structured developmental history interviews to use where the full instrument is impractical. Those are not standardised between services, which is a trade-off worth knowing about.
If you are preparing for one
If an ADI-R Test is booked, start collecting now, before the appointment. School reports from the earliest years you can find. Health visitor or paediatric records. Baby books. Home video, which is unusually valuable because it shows behaviour rather than describing it.
Talk to your informant beforehand. Not to rehearse, which is unhelpful, but so the questions are not a surprise and so they can start remembering. Two hours of cold recall produces thin answers.
And if you have no informant, say so at referral rather than at the appointment. It changes what the service needs to plan for, and finding out on the day wastes an assessment slot that someone has waited a long time for.
Sources
- Rutter M, LeCouteur A, Lord C. Autism Diagnostic Interview, Revised (ADI-R). Torrance, CA: Western Psychological Services, 2003. Publisher page, the source of the item count, timing, domains and qualification details above.
- Fusar-Poli L, Brondino N, Rocchetti M, et al. Diagnosing ASD in adults without ID: accuracy of the ADOS-2 and the ADI-R. Journal of Autism and Developmental Disorders, 2017. DOI. The adult accuracy study behind the 55 percent figure.
- Lampinen LA, et al. DSM-5 based algorithms for the Autism Diagnostic Interview-Revised for children ages 4 to 17 years. Journal of Child Psychology and Psychiatry, 2025. DOI · Full text. On updating the algorithms for current criteria.
- Risi S, Lord C, Gotham K, et al. Combining information from multiple sources in the diagnosis of autism spectrum disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 2006;45(9):1094-1103. DOI. On why the interview and the observation are used together.
- Lecavalier L, Aman MG, Scahill L, et al. Validity of the Autism Diagnostic Interview-Revised. American Journal on Mental Retardation, 2006;111(3):199-215. DOI. The internal consistency figures.
Search the literature yourself
- PubMed: ADI-R validity in adults and ADI-R and ADOS used together.
- Google Scholar: ADI-R sensitivity and specificity.
- NICE CG142: Autism spectrum disorder in adults, on what an assessment should include.
- National Autistic Society: diagnosis. Practical guidance, including for adults without an informant.
- Wikipedia: Autism Diagnostic Interview. Background and history.