ADOS-2 Test Examples: Activities, Modules, and Scoring

The Autism Diagnostic Observation Schedule, widely known as the ADOS-2, is a semi-structured assessment in which a trained clinician guides a person through a series of social activities, conversations, and play-based tasks designed to bring out behaviors associated with autism spectrum disorder (ASD).¹ Rather than a pencil-and-paper quiz with right or wrong answers, the ADOS is essentially a structured social situation. A clinician observes how someone responds to specific prompts, manages back-and-forth conversation, uses gestures and eye contact, and engages with toys or materials. What the clinician is watching for, and how the session actually unfolds, varies depending on the person’s age and language ability.

What the ADOS Session Actually Looks Like

An ADOS session typically lasts around 40 to 60 minutes and takes place in a room with a table, chairs, and a selection of materials chosen for the module being used. The clinician is not running a test in the way most people picture one. There are no flashcards to get right, no timed puzzles, no score displayed at the end of a round. Instead, the clinician creates social situations and watches how the person navigates them. Does the child look at the clinician’s face when something surprising happens? Does the teenager pick up on a conversational cue and follow it, or does the topic shift feel awkward and forced? Does the adult describe emotions in a way that connects to the listener, or does the account feel flat and rehearsed?

The clinician follows a loose script of activities but has room to adjust pacing and emphasis. Each activity is called a “press” because it presses on a particular social or communicative behavior. If a young child does not respond to one kind of social bid, the clinician tries another approach before moving on. This flexibility is by design: the goal is to give the person the best opportunity to show what they can do, so the clinician can accurately code what they actually observe.

Modules and Who Gets Which One

The ADOS-2 has five modules, and the choice depends on the person’s age and expressive language level. You do not take all five. You take one.

Toddler Module: For children between roughly 12 and 30 months of age. This module uses toys like bubbles, a ball, a shape sorter, and a wind-up toy, and relies heavily on free play and caregiver interaction.
Module 1: For children older than 30 months who do not yet use phrase speech consistently. Activities focus on play, response to name, and shared enjoyment of simple cause-and-effect toys.
Module 2: For children of any age who use phrase speech but are not yet fluent. Activities include interactive play, a birthday party role-play scenario, and conversation about daily routines.
Module 3: For children and younger adolescents who are verbally fluent. Tasks include telling a story from a picture book, describing social situations, and conversation about emotions and friendships.
Module 4: For older adolescents and adults with fluent speech. The session looks more like a structured interview, with questions about relationships, work or school, plans for the future, and feelings about social situations.

The original ADOS-G described each module as roughly 30 minutes long, with four modules covering different expressive language levels.² The Toddler Module was developed later specifically because the existing modules were not capturing the earliest signs of autism in very young children. In validation testing, the Toddler Module was used in 360 evaluations with 182 children and showed strong ability to discriminate between children with ASD, those with other developmental delays, and typically developing children.³

Examples of Specific Activities and What Clinicians Watch For

If you are preparing for an ADOS assessment for yourself or your child, the activities might sound deceptively simple. Here are some concrete examples drawn from how the modules are structured.

In Module 1 or the Toddler Module, the clinician might blow bubbles and then suddenly stop. The “press” here is whether the child looks at the clinician’s face, reaches toward the bubbles, or makes a sound or gesture to request more. A typically developing child will usually look to the adult and signal in some way. A child on the autism spectrum might stare at the bubble wand without redirecting attention toward the person holding it, or might try to grab the wand themselves rather than communicating a request. The clinician might also bring out a balloon, blow it up, let it deflate noisily, and watch whether the child seeks shared enjoyment by looking at the adult and smiling, or whether the child’s reaction stays self-contained.

In Modules 2 and 3, one common activity involves the clinician setting out miniature figures, a toy car, plates, and cups, and inviting the child to play “make-believe.” The clinician watches whether the child creates a narrative (perhaps the figure drives the car to a party), assigns roles, and invites the clinician into the story. Less imaginative or more rigid play, like lining the figures up by size or repeatedly opening and closing the car door, gets coded differently.

In Module 3, a picture book with no words might be placed in front of the child, who is asked to tell a story based on the pictures. The clinician is not evaluating vocabulary or reading ability. They are watching whether the child describes characters’ feelings, explains motivations, and looks at the clinician while narrating, or whether the description stays concrete and action-based (“the frog jumped, then the boy fell”) without reference to emotions or social meaning.

Module 4 looks and feels different. The clinician and the person sit across from each other, and the conversation covers topics like friendships, loneliness, plans for the future, and what makes the person feel stressed or happy. There are also tasks: the person might be asked to make up a story using a set of random objects, or to describe a picture showing a social scene. The clinician is coding whether the person spontaneously offers emotional content, uses varied facial expressions that match what they’re saying, and picks up on conversational cues like humor or rhetorical questions.

How the Scoring Works

After the session ends, the clinician reviews notes and assigns scores to individual items across two main domains: Social Affect (which collapses what used to be separate communication and social interaction scores) and Restricted and Repetitive Behaviors. Each item is typically scored on a scale from 0 to 3, where 0 means no evidence of abnormality and higher numbers reflect increasingly atypical behavior. These item scores feed into an algorithm that produces a total score, which is compared against cutoffs for “autism” and “autism spectrum.”

A useful refinement is the Calibrated Severity Score (CSS), which adjusts the raw total to account for the person’s age and language level. Because a 3-year-old with limited speech and a 10-year-old who speaks in full sentences are taking different modules, their raw totals are not directly comparable. The CSS places everyone on a 1-to-10 scale. Research has shown that the CSS is more stable over time than raw scores, remains more consistent across different modules, and is less influenced by the person’s verbal ability, making it a more valid measure of autism severity.⁴

One thing worth understanding: a score above the cutoff does not automatically mean a person receives an autism diagnosis. The ADOS is one piece of evidence. A clinician integrates it with developmental history, caregiver reports, cognitive testing, and clinical judgment. The ADOS is designed to observe social and communicative behaviors within a standardized context, and clinicians who use it report that its biggest strength is exactly that: it provides a structured opportunity to see ASD-specific behaviors that might not emerge in a typical office visit.⁵

Accuracy of the ADOS-2

A meta-analysis examining the ADOS-2’s performance in children found sensitivity between 89% and 92% and specificity between 81% and 85%.⁶ In plain terms, the ADOS-2 correctly identifies roughly 9 out of 10 children who have autism, and correctly rules it out in about 8 out of 10 children who don’t. A large German validation study confirmed these findings, showing that both the original ADOS and the updated ADOS-2 algorithms maintained diagnostic validity across a sample of over 800 individuals.⁷

Those numbers are strong, but they also mean the ADOS-2 is not perfect. Roughly 1 in 10 people with autism can score below the cutoff and be missed, while roughly 1 in 5 to 6 people without autism might score above it. This is why the assessment process typically includes more than just the ADOS.

Why the ADOS Is Usually Paired With a Parent Interview

The ADOS captures what a clinician sees in a single structured session. It does not capture what happens at home over years. The Autism Diagnostic Interview-Revised (ADI-R) fills that gap: it is a lengthy structured interview with a caregiver that covers early development, language history, social behavior, and repetitive interests. Research on combining the two shows that together they achieve balanced sensitivity (around 77%–80%) and specificity (around 87%–90%) in young children with unclear developmental presentations.⁸ The two instruments are complementary: the ADOS tends to be more sensitive (better at catching true cases), while the ADI-R adds specificity (better at ruling out non-cases).⁹

Interestingly, a study that used machine-learning classifiers found that when models were built on both ADOS and ADI-R data, the models relied primarily on ADOS items. In the adolescent and adult sample specifically, the classifier using only ADOS items actually outperformed the one using both instruments.¹⁰ This does not mean the ADI-R is useless, but it does raise questions about when it adds meaningful diagnostic information beyond what the ADOS already provides, particularly for older individuals.

How the ADOS Performs Differently for Females

One of the most actively discussed issues in autism assessment is whether the ADOS systematically underdetects autism in girls and women. A large study comparing ADOS-2 scores by sex found that males had significantly higher total Calibrated Severity Scores on average than females (roughly 7.5 versus 6.3), with medium effect sizes for Social Affect and smaller differences for Restricted and Repetitive Behaviors.¹¹ This could mean females genuinely show fewer observable behaviors, or it could mean the instrument is less sensitive to how autism presents in females.

A more recent item-level analysis of Module 3 found that on four social communication items, females tended to be underscored, meaning they were rated as showing fewer autistic features than males with equivalent underlying levels of autism traits. The researchers speculated this may reflect masking (where a person consciously or unconsciously compensates for social difficulties) or clinician bias in scoring.¹² For women and girls being assessed, this is worth knowing: a score below the cutoff does not necessarily mean autism is absent, especially if masking is a factor. A thorough clinician will weigh the ADOS alongside self-report, caregiver history, and their own clinical impression.

Module 4 data for adults shows a similar pattern. One study at a specialized clinic found outstanding diagnostic performance for adults with IQ above 70 overall, but for females specifically, the revised algorithm scores were lower and did not correlate with ADI-R scores the way males’ scores did.¹³ This disconnect reinforces the idea that the ADOS may not fully capture how autism manifests in some women.

Adults Seeking a Late Diagnosis

If you are an adult who suspects you might be autistic and is looking into assessment, the ADOS-2 Module 4 is the version designed for you. The session feels less like a medical examination and more like a semi-structured conversation with some odd prompts thrown in. You might be asked to describe a time you felt overwhelmed, to explain what friendship means to you, or to make up a story from a set of unrelated objects on the table. The clinician is quietly coding your eye contact, the range and timing of your gestures, whether your facial expressions track with your speech, and how you handle unexpected shifts in the conversation.

Research confirms that Module 4 has solid diagnostic accuracy for adults without intellectual disability.¹⁴ A multi-site validation study of a revised Module 4 algorithm found that the update improved sensitivity (fewer missed cases) but at the cost of somewhat lower specificity (more people without autism scoring above the threshold). Performance was strongest for females, people with verbal IQ below 85 or above 115, and those aged 16 and older.¹⁵

Adults who are good at social compensation often worry that they will “pass” the ADOS even though they find social interaction genuinely difficult. This is a real concern. A skilled evaluator accounts for it by looking beyond whether you can produce socially appropriate behavior in the moment and attending to how naturally and effortlessly you produce it. Still, there is no guarantee the instrument catches everyone, and some clinicians will lean more on developmental history and self-report when they suspect masking is at play.

When Intellectual Disability Is Also Present

Diagnosing autism in someone who also has an intellectual disability is tricky, because some behaviors that look like autism, such as limited eye contact or repetitive movements, can also stem from general cognitive delay. One study of adults with intellectual disability found the ADOS was extremely sensitive (100% in testable cases) but had poor specificity (only 45%), meaning it correctly caught every person with autism but also flagged many who did not have it.¹⁶ The ADOS essentially becomes over-inclusive in this population, calling too many people positive.

For very young children with developmental levels below 12 months, the ADOS-2 Toddler Module still managed to differentiate ASD from general developmental delay, with autism symptoms showing up regardless of how delayed the child was cognitively.¹⁷ This finding is practically important because it means clinicians can use the Toddler Module even for children whose developmental level falls below the officially recommended cutoff, potentially getting them access to early intervention sooner.

Remote and Telehealth Versions

The COVID-19 pandemic forced a rapid rethink of autism assessment, since the ADOS requires face-to-face interaction with standardized materials. The result was the Brief Observation of Symptoms of Autism (BOSA), developed by the same team behind the ADOS. The BOSA uses a kit of materials that is sent to the family’s home. A caregiver interacts with the child using the materials while a clinician watches over video and codes behaviors using their ADOS training.¹⁸

An evaluation of an online protocol combining the BOSA with a caregiver interview achieved roughly 95% sensitivity and 70% specificity, with about 80% overall agreement between the remote and face-to-face evaluators.¹⁹ A separate pilot study validated a brief remote observation against the in-person ADOS-2 in 30 children and found a strong correlation (r = .75) and acceptable sensitivity (about 89%) and specificity (80%).²⁰ These remote tools are not yet considered replacements for the full in-person ADOS, but they offer a viable option when distance, waitlists, or health concerns make face-to-face evaluation impractical.

Cultural and Language Adaptations

The ADOS was developed and normed primarily in English-speaking Western populations, which raises questions about how well it works in other cultural contexts. Translation is not just about swapping words; the social expectations baked into the tasks matter too. When the ADOS-2 was translated into Afrikaans and tested with a community in South Africa’s Western Cape, researchers found that the social interaction demands were broadly appropriate, but specific features of the local dialect (“Kaaps”) needed to be accounted for during administration.²¹ A Brazilian Portuguese adaptation similarly found good equivalence with the original English version, though some semantic differences surfaced during back-translation.²²

This matters practically: if you or your child is being assessed in a language or cultural context different from the one the ADOS was built for, it is worth asking the clinician whether they are using a validated translation and whether they have experience with your cultural background. Eye contact norms, for instance, vary across cultures, and a clinician unfamiliar with those norms could misinterpret a culturally typical gaze pattern as socially atypical.

Conditions That Can Mimic Autism on the ADOS

One practical concern for anyone going through an assessment is whether a different condition might produce a false positive. Social anxiety, ADHD, and language disorders can all affect how a person behaves in a structured social observation. A study comparing children referred to an autism clinic who ended up diagnosed with ASD, ADHD, or anxiety disorders found that communication and social interaction items on diagnostic criteria largely distinguished the autism group from the other two. However, items in the restricted and repetitive behavior domain did a poorer job of separating them.²³ A person with severe social anxiety, for example, might show reduced eye contact and stilted conversation during an ADOS session, which could superficially look like autism. This is why the broader assessment context, including developmental history and information about anxiety symptoms, is critical for distinguishing what the ADOS score actually reflects.

Women with anorexia nervosa are another group in which ADOS scores sometimes raise autism flags. A study that administered Module 4 to 16 women with anorexia found that the assessment picked up social communication and repetitive behavior patterns overlapping with autism features.²⁴ Whether these reflect genuine co-occurring autism or whether the social withdrawal and rigid thinking patterns of anorexia simply mimic autism on the instrument is still debated. For clinicians, the takeaway is that an elevated ADOS score in the context of an active eating disorder needs careful interpretation.

Sensory Behaviors and Repetitive Interests During the Assessment

Parents sometimes ask what “restricted and repetitive behaviors” look like in the room during an ADOS. In practice, the clinician is watching for things like hand-flapping, spinning objects, lining up toys, fixating visually on specific parts of a toy rather than using it functionally, or returning repeatedly to the same object or topic despite the clinician redirecting. One detailed case description noted constant motor stereotypies including flapping and mannerisms, repetitive non-directed vocalizations, unusual visual fixation, and tactile sensory interests toward surfaces and non-functional parts of objects.²⁵ These behaviors are scored separately from social communication, and together the two domains produce the overall algorithm score. In some individuals, particularly verbally fluent older children and adults, repetitive behaviors during the session can be subtle or even absent: the person might only show restricted interests through their conversational topics rather than through visible motor behaviors.