ADI-R scoring follows a defined sequence. A trained interviewer codes a caregiver’s responses to 93 items, a selected subset of those codes transfers to a diagnostic algorithm, and the total for each symptom domain is compared against an empirically derived cutoff. A classification of autism requires meeting or exceeding every domain cutoff, including at least one point for developmental abnormality evident at or before 36 months.
That is the short answer. The longer answer, the one that matters when you are writing reports, training new staff, or explaining results to a referring provider, involves coding conventions, code conversions, two families of algorithms, and several interpretation decisions the algorithm cannot make for you. This guide works through each step. For a broader orientation to the instrument, its history, and where it sits in an evaluation battery, start with our ADI-R overview for clinicians. Here, we stay on scoring.
The ADI-R in One Paragraph
The Autism Diagnostic Interview-Revised (ADI-R) is a standardized, semi-structured caregiver interview published by WPS and authored by Michael Rutter, Ann Le Couteur, and Catherine Lord. It descends from the research interview described by Lord and colleagues in 1994 and is appropriate for children and adults with a mental age above 2 years, 0 months. The interview quantifies developmental history and current behavior across three functional domains: language and communication, reciprocal social interaction, and restricted, repetitive, and stereotyped behaviors and interests. It is a diagnostic instrument rather than a screener, a distinction we cover in screeners vs. diagnostic tools for autism.
How ADI-R Scoring Works, Step by Step
Every ADI-R score sheet reflects the same five-step path:
- Interview and code. The clinician conducts the full interview, probing each of the 93 items until enough behavioral detail emerges to assign a code. Codes reflect the caregiver’s description of behavior, not the caregiver’s own labels or conclusions.
- Transfer selected items. Only a subset of items appears on any given algorithm. The clinician transfers those item codes to the algorithm form.
- Convert codes to scores. Algorithm scoring applies standard conversions, described below, so that each item contributes 0, 1, or 2 points.
- Sum the domains. Converted scores are totaled within each algorithm domain.
- Compare totals to cutoffs. Each domain total is compared to its published cutoff. The pattern across domains, not any single number, drives the classification.
Notice what is absent from this sequence: there is no overall ADI-R score, no standard score, and no percentile. The instrument yields categorical results. That design choice shapes nearly everything about interpretation, and we will return to it.
The Interview Behind the Scores
Scoring quality is set during the interview itself, so the structure is worth reviewing. The 93 items open with background and early developmental history, then move through language and communication, social development and play, and interests and behaviors, with responses organized into content areas that feed the domains. The publisher estimates 90 to 150 minutes for administration and scoring combined.
| Feature | ADI-R |
|---|---|
| Format | Semi-structured, investigator-based caregiver interview |
| Items | 93, with standardized probes |
| Respondent | Parent or caregiver familiar with early development and current behavior |
| Administration and scoring time | 90-150 minutes (publisher estimate) |
| Appropriate for | Children and adults with a mental age above 2 years, 0 months |
| Output | Categorical algorithm results, not normed scores |
| Qualification | WPS Level C, plus instrument-specific training |
Two administration points bear directly on score validity. First, the ADI-R is investigator-based: the interviewer, not the caregiver, assigns each code based on described behavior, which is why standardized training matters as much as the protocol itself. WPS requires Level C qualification to purchase the instrument and offers a structured training program in administration and coding. Many clinics also have new interviewers double-code alongside an experienced administrator before independent use. Second, the respondent must actually know the person’s early history. When the available caregiver joined the child’s life later, or early records are thin, the historical items become harder to code with confidence, and the report should say so.
Coding Conventions: What the Numbers Mean
Each ADI-R item is coded on an ordinal 0-to-3 scale. A code of 0 indicates that the behavior specified in the item is not present. Higher codes indicate definite abnormality of the specified type, with 3 reserved for the most severe or pervasive presentations. Three additional codes sit outside the ordinal scale and flag responses it cannot capture, for example, an atypicality of a different kind than the item describes, an item that does not apply to that individual, or information the caregiver cannot supply. These are coded 7, 8, and 9.
The algorithm then applies two conversions that surprise many first-time users:
- Codes of 3 convert to 2. On the algorithm, a 3 contributes the same two points as a 2. The severity distinction is preserved in the interview protocol for clinical use, but the algorithm counts definite abnormality without extra weighting.
- Codes of 7, 8, and 9 convert to 0. Responses outside the ordinal scale contribute nothing to domain totals.
Both conventions are documented in the peer-reviewed literature as standard algorithm practice. The practical consequence is that a domain total can understate clinical severity, and an unusually high number of 8 and 9 codes can quietly lower totals. A well-written report notes when either situation applies.
One more layer sits underneath the codes. Most items are coded for more than one time frame. “Current” codes generally reference roughly the most recent three months. Other codes capture whether a behavior was “ever” present, and several social and communication items are keyed to the developmental window between the 4th and 5th birthdays, coded as the most abnormal presentation in that period. Which time frame feeds the algorithm depends on which algorithm you are running, and that brings us to the cutoffs.
The ADI-R Diagnostic Algorithm: Domains and Cutoff Scores
The diagnostic algorithm draws on developmental history, using “ever” codes and the most-abnormal-4-to-5 codes. Its domains mirror the ICD-10 and DSM-IV conception of autism, with a fourth criterion for early onset. The published cutoffs, confirmed in the peer-reviewed literature, are:
| Algorithm domain | Score range | Cutoff |
|---|---|---|
| A. Qualitative abnormalities in reciprocal social interaction | 0-30 | 10 |
| B(V). Qualitative abnormalities in communication, verbal individuals | 0-26 | 8 |
| B(NV). Qualitative abnormalities in communication, nonverbal individuals | 0-14 | 7 |
| C. Restricted, repetitive, and stereotyped patterns of behavior | 0-12 | 3 |
| D. Abnormality of development evident at or before 36 months | Onset items | 1 |
A few notes on reading this table correctly:
Verbal status is defined by the instrument, not by impression. The verbal communication algorithm applies when the individual uses phrases of at least three words, sometimes including a verb, on a daily basis. This is established by a specific language item early in the interview. Verbal individuals are scored on the full communication item set; nonverbal individuals are scored on the nonverbal communication items only, against the lower cutoff of 7.
Classification requires every cutoff, not most of them. The algorithm classifies an individual as autism only when the social, communication, and restricted-repetitive totals each meet or exceed their cutoffs and at least one onset criterion is met. A child who scores 24 on social interaction and 12 on communication but 2 on restricted and repetitive behavior does not meet the algorithm, and the report should present that plainly rather than rounding up.
Exceeding a cutoff by a wide margin is clinically informative but algorithmically identical. A social total of 11 and a social total of 28 both satisfy domain A. Domain totals describe the breadth of reported abnormality across items; they are not severity scales, and the manual provides no normative interpretation of distance above the cutoff.
The algorithm is anchored to earlier diagnostic frameworks. The domain structure follows ICD-10 and DSM-IV criteria for autism. DSM-5-oriented ADI-R algorithms have been proposed in the research literature, but in standard clinical use, the clinician, not the form, maps ADI-R findings onto current DSM-5 criteria.
Diagnostic Algorithm vs. Current Behavior Algorithm
The ADI-R ships with five age-specific algorithms calculated from a single algorithm form: two diagnostic algorithms and three current behavior algorithms.
| Diagnostic algorithms | Current behavior algorithms | |
|---|---|---|
| Time frame | Developmental history: “ever” and most-abnormal-4-to-5 codes | Present functioning, generally the recent months |
| Number of versions | Two, selected by age | Three, selected by age |
| Primary use | Formal diagnostic evaluation | Treatment and educational planning |
| Compared to cutoffs | Yes | Used descriptively for planning |
The distinction matters most in two situations. For an older child or adult, current behavior may look quite different from the preschool years, particularly after years of intervention. The diagnostic algorithm intentionally reaches back to the early presentation, because that history carries the diagnostic signal. The current behavior algorithm, by contrast, describes the person in front of you now, which is the more useful frame for building treatment targets or informing an IEP discussion. One version of the diagnostic algorithm applies from age 4 years, 0 months onward; the other covers younger children, down to the instrument’s mental-age floor. Selecting the wrong version, or reporting current behavior totals against diagnostic cutoffs, is among the more common scoring errors seen in re-reviewed records.
From Algorithm Classification to Clinical Interpretation
The algorithm’s output is a classification, and a classification is not a diagnosis. Autism spectrum disorder is diagnosed by a qualified clinician applying DSM-5 criteria to the whole picture: caregiver interview, direct observation, developmental and medical history, and cognitive and language testing where indicated. The ADI-R contributes one structured, quantified account of that picture. It contributes well, which is why the instrument has anchored research and specialty practice for three decades, but it was never designed to stand alone.
In practice, interpretation runs along three lines.
Convergence with direct observation. Because the ADI-R samples caregiver report, most diagnostic evaluations pair it with an observational measure, most often the ADOS-2. When interview and observation agree, confidence rises. When they diverge, the divergence itself is informative: a child may suppress repetitive behaviors in a novel clinic room, or a caregiver may under-recognize social differences that are obvious in direct interaction. Our guide to ADOS-2 scores and interpretation covers the observational side of that pairing.
Item-level review. Domain totals compress a great deal of information. Two children can post identical social totals through entirely different item patterns, one driven by limited peer interest and one by absent social reciprocity with familiar adults. Before writing the interpretation section of a report, it is worth rereading the contributing items, because they, not the totals, describe the child.
Translation for the receiving team. ADI-R results routinely travel to people who never administer the instrument: BCBAs building programs, school teams, primary care physicians. A useful report states which algorithm was run, whether the verbal or nonverbal communication cutoff applied, each domain total against its cutoff, and what the clinician concluded after integrating everything else. ABA providers who want a working knowledge of the observational companion measure can start with our ADOS-2 explainer for ABA providers. Companion scoring guides for other instruments in the typical battery are being published alongside this one.
Measurement Considerations for Repeated Use
Every instrument has a design envelope, and the ADI-R is the diagnostic workup. Three of its design features are worth understanding before anyone asks it to do a different job, such as demonstrating progress across a six-month authorization period.
First, the diagnostic algorithm is historical by construction. “Ever” codes and the most-abnormal-4-to-5 codes cannot decline with treatment, because they describe the past. A child could make substantial gains and produce an unchanged diagnostic algorithm, which is exactly what the algorithm is supposed to do.
Second, the interview quantifies caregiver recollection. That is a strength for diagnosis, since caregivers hold years of observation no clinic visit can reproduce, but recall of early milestones becomes less precise as time passes, and repeated administrations sample the same memory rather than new information.
Third, the output is categorical. The ADI-R provides no normed change score, no standard error of measurement around a scale total, and therefore no principled way to say that a movement of a few points is meaningful improvement. Add the 90-to-150-minute administration, and frequent re-administration becomes hard to justify on both psychometric and practical grounds.
None of this is a flaw. It is a scope statement, the same kind that applies to any well-built instrument. Teams that need to show change over a treatment cycle typically keep the ADI-R in its diagnostic lane and pair it with measures built for serial use. We have written about that measurement gap in the six-month reassessment problem and about the field’s broader movement in subjective vs. objective measurement in autism care.
Where Objective Measurement Fits Alongside the ADI-R
The ADI-R quantifies one indispensable information source: what a caregiver has observed over the child’s lifetime, coded by a trained interviewer. A complete evaluation also benefits from data that do not pass through anyone’s recall or coding, and this is where objective physiological measurement has begun to enter diagnostic practice.
The EarliPoint System is an FDA-cleared medical device indicated for use as a tool to aid qualified clinicians in the diagnosis and assessment of Autism Spectrum Disorder (ASD) in children 16 to 95 months old who are at risk based on concerns shared by a parent, caregiver, or healthcare provider. It uses eye-tracking to measure a child’s moment-by-moment visual engagement with social information while the child watches a series of short videos, producing quantitative measurements for the clinician’s review. It is available by prescription and is used under the supervision of a qualified clinician.
The relationship to the ADI-R is complementary, not competitive. The interview captures history and caregiver-observed behavior; eye-tracking captures a direct biological measurement of social attention on the day of testing. Neither replaces clinical judgment, and the EarliPoint System complements clinical judgment rather than substituting for it. For clinicians evaluating the evidence, our summary of the two large JAMA studies of the technology’s diagnostic accuracy is a good place to start.
Frequently Asked Questions
What is a positive ADI-R score?
There is no single positive score. The diagnostic algorithm yields a classification of autism when the totals for reciprocal social interaction, communication, and restricted and repetitive behavior each meet or exceed their cutoffs, and at least one point is scored for developmental abnormality evident at or before 36 months. Missing any one cutoff means the algorithm does not produce the classification, though the clinician still interprets the full profile.
What are the ADI-R cutoff scores?
On the standard diagnostic algorithm, the cutoffs are 10 for reciprocal social interaction, 8 for communication in verbal individuals, 7 for communication in nonverbal individuals, 3 for restricted and repetitive behaviors, and 1 for abnormality of development evident at or before 36 months.
How long does the ADI-R take to administer and score?
The publisher estimates 90 to 150 minutes for administration and scoring combined. Interviews with talkative historians or complicated developmental courses run longer, and experienced administrators tend to be faster without sacrificing probe quality.
Who can administer the ADI-R?
WPS sells the ADI-R at qualification Level C, and valid administration requires completing standardized training in the interview and its coding conventions. In most settings, administrators are licensed psychologists, developmental pediatricians and other physicians, or clinicians with comparable graduate training in assessment who have completed ADI-R-specific training.
What is the difference between the ADI-R and the ADOS-2?
The ADI-R is a caregiver interview that quantifies developmental history and reported behavior. The ADOS-2 is a direct observational assessment in which the clinician elicits and codes behavior firsthand. They sample different information sources, and many diagnostic centers administer both within a single evaluation.
Can the ADI-R alone diagnose autism?
No. The algorithm produces a classification that informs diagnosis. Autism spectrum disorder is a clinical diagnosis made by a qualified professional who integrates interview findings with direct observation, history, and DSM-5 criteria.
Can the ADI-R be used to track treatment progress?
It was not built for that purpose. The diagnostic algorithm rests on lifetime and early-childhood codes that cannot change with treatment, and the instrument yields categorical results rather than normed change scores. Teams demonstrating progress typically pair diagnostic instruments with measures designed for serial use, from standardized rating scales and direct behavioral data to objective tools used under clinician supervision.
Closing Thoughts
ADI-R scoring rewards precision at every step: disciplined probing, faithful coding, correct conversions, the right algorithm for the age and question, and honest reporting of totals against cutoffs. Get those steps right and the instrument does what it has done since 1994, which is to turn a caregiver’s knowledge into structured diagnostic evidence. Keep its scope in view, pair it with direct observation and, where appropriate, objective measurement, and the resulting evaluation serves the child, the family, and every clinician who reads the report afterward.
