CARS-2 Scoring Guide: ST vs HF, Cutoff Scores, and Age Ranges

CARS-2 scoring follows a consistent sequence. A trained clinician rates 15 items on a scale of 1 to 4, using half-point midpoints where a child falls between anchors, then sums those ratings into a Total Raw Score between 15 and 60 and compares the total against cutoff values. Which cutoffs apply depends on the version: on the Standard Version (CARS2-ST), totals of 30 and above fall in the autism spectrum range, while on the High-Functioning Version (CARS2-HF), the threshold is 28.

That version distinction is where most scoring questions, and most scoring errors, begin. The two forms look alike on the page, yet they carry different item content and different cutoff scores. Accurate scoring starts before the first item is rated.

This guide covers version selection, item-level mechanics, the severity ranges for each form, and the CARS-2 age range. If your team is still deciding which instrument belongs in your assessment battery, start with our comparison of the CARS-2 and ADOS-2, which covers tool selection. This article assumes the CARS-2 is already on your desk and goes deep on scoring alone.

The CARS-2 in Brief

For readers newer to the instrument: the Childhood Autism Rating Scale, Second Edition (CARS-2) is a clinician-completed rating scale authored by Eric Schopler, Mary Van Bourgondien, Glenna Wellman, and Steven Love, published by WPS in 2010 as an update to the original CARS. It is intended for individuals ages 2 years and older. Ratings draw on direct observation combined with caregiver and record information, and per the publisher, the rating itself takes about 5 to 10 minutes once that information has been collected. The kit contains two scored 15-item forms, the CARS2-ST and the CARS2-HF, plus one unscored form, the Questionnaire for Parents or Caregivers (CARS2-QPC).

Everything in Childhood Autism Rating Scale scoring builds on this structure.

CARS-2 ST vs HF: Choosing the Right Version

The version decision is the first scoring decision, governed by three factors: age, expressive language, and estimated cognitive ability.

Per the publisher’s criteria:

A practical shorthand: the ST is the default, and the HF applies only when all three of its conditions are met at once. A verbally fluent 5-year-old is still rated on the ST. So is a 9-year-old with limited spoken language, regardless of cognitive scores.

The HF form exists because the second edition set out to capture autism in older, verbally fluent individuals. Several of its items are reworked accordingly, with attention to areas such as social-emotional understanding and restricted interests rather than the early-childhood behaviors emphasized on the ST.

Two habits prevent most downstream problems. When cognitive estimates are borderline, document the basis for your version choice in the report. And never score a completed form against the other version’s cutoffs. In developmental-behavioral practice, most CARS-2 interpretation problems I encounter trace back to version confusion, not arithmetic.

One clarification about the QPC, since it prompts frequent questions: it produces no score. It structures what parents and caregivers report, and the clinician weighs that information when making the 15 ratings on the ST or HF.

Scoring Mechanics: Items, Anchors, and the Total Raw Score

Each of the 15 items is rated on a 4-point scale in which 1 reflects behavior within typical limits for the child’s age and 4 reflects severely atypical behavior for age. Raters may use half-point midpoints (1.5, 2.5, 3.5) when a child’s presentation falls between anchors, which gives seven possible values per item. The age-referencing matters: the question is always whether behavior departs from what is expected at this child’s age.

The items sample a broad set of domains, including relating to people, imitation and social-emotional understanding, emotional response, body and object use, adaptation to change and restricted interests, visual and listening responses, responses to taste, smell, and touch, fear or anxiety, verbal and nonverbal communication, activity level, and the consistency of intellectual response, ending with the clinician’s general impression.

Good ratings rest on adequate information. The manual expects the rater to draw on direct observation plus what caregivers, teachers, and records contribute, with the QPC available to structure the caregiver piece. A rating made from a thin observation and nothing else is a weak rating, whatever the arithmetic says afterward.

Scoring itself is straightforward. Sum the 15 item ratings to produce the Total Raw Score, which ranges from 15 to 60. The manual also provides conversions from raw scores to standard scores and percentile ranks based on a clinical sample of 1,034 individuals with autism spectrum disorders. Those converted scores describe where an individual stands relative to others diagnosed with ASD, not the general population, so they are best used to characterize severity within the spectrum.

CARS-2 Cutoff Scores and Severity Ranges

Here is the center of CARS-2 score interpretation, and the place where precision matters most: the two forms use different cutoff scores and different severity bands. Do not blend the tables.

The ranges below reflect the manual’s values as reported in the peer-reviewed literature.

CARS2-ST severity ranges

CARS2-HF severity ranges

Notice what the two-point difference does in practice. A total of 29 sits below the cutoff on the ST but above it on the HF. A total of 35 falls in the mild-to-moderate band on the ST and in the severe band on the HF. The number alone tells you nothing until you know the form, which is why every report should state the version next to the score.

Two further notes. The half-point structure means totals often land exactly on a band boundary, at values like 29.5 or 33.5; boundary scores call for corroborating evidence, not a coin flip. And published research continues to examine alternative thresholds, including a lower CARS2-ST value studied for capturing milder presentations in screening contexts. A cutoff is an evidence-based decision aid; the clinician, not the table, carries the interpretive responsibility.

CARS-2 Age Range: Who the Instrument Covers

The CARS-2 age range starts at 2 years and extends upward through adolescence and into adulthood. Within that span, age interacts with version choice in one specific way: before age 6, the ST is the only option, and from 6 onward, the HF becomes available when the language and cognitive criteria are met.

The CARS-2 is not designed for children under 2; concerns in that window belong in developmental surveillance and screening pathways, followed by diagnostic evaluation with tools built for toddlers. If you are mapping which instruments cover which ages across your intake population, see our reference chart [LINK WHEN LIVE: Autism Assessment Age Ranges: A Quick Reference Chart].

For practices serving a wide age span, this reach is one of the instrument’s most practical features. The same structure covers the 3-year-old referred from early intervention and the 10-year-old referred after years of subtle presentation in a mainstream classroom, with the version system adjusting item content and cutoffs to fit.

Interpreting CARS-2 Scores in Practice

A CARS-2 total does three things well. It quantifies the degree to which observed and reported behavior departs from age expectations across 15 domains. It places that quantity in a severity band with published cutoffs. And it gives a psychologist, a BCBA, and a pediatrician reading the same chart a shared, structured way to talk about presentation.

It is equally worth being clear about what a total does not do. The CARS-2 is a rating scale: it structures clinical observation and informant information, and the ratings themselves involve clinician judgment. Two thoughtful raters working from different observations can land on different totals. A score at or above a cutoff is not a diagnosis, and a score below a cutoff does not rule one out. Autism remains a clinical diagnosis made against DSM-5 criteria by a qualified professional synthesizing history, observation, and functional impact. For where rating scales sit on the wider map of instruments, our overview of screeners versus diagnostic tools draws those lines in detail.

In a complete evaluation, the CARS-2 typically sits alongside a structured developmental history such as the ADI-R, direct structured observation such as the ADOS-2 (our guide to ADOS-2 scores and interpretation covers that side), and adaptive behavior measurement. Convergence across methods is what gives a diagnostic conclusion its footing; divergence is a finding in itself and usually a prompt for more information.

Finally, plan how the number will be communicated. Families tend to hear scores as verdicts, so lead with the band and its meaning rather than the decimal, name the version, and frame the result as one structured piece of a larger clinical picture. Referral partners and payers need the same three elements in writing: the version administered, the total and its band, and the information sources behind the ratings. A reported “CARS-2 of 31” with no version attached is an incomplete data point.

Comparing CARS-2 Scores Over Time

Many practices re-administer rating scales at reassessment, so CARS-2 totals from different time points often sit side by side in progress reviews. Before reading a change score, walk through four questions.

First, was the same version used both times? A child can cross the age 6 boundary, or gain verbal fluency, between evaluations. When the version changes, the cutoffs and item content change with it, and the totals are no longer on a common scale.

Second, who rated, and from what information? CARS-2 ratings reflect the rater’s observations and judgment as well as the child’s behavior. Different raters, different settings, or a different depth of caregiver input can shift totals for reasons unrelated to the child’s development. Holding the rater and procedure constant strengthens any comparison; when that is not possible, note it in the report.

Third, how big is the change relative to the instrument’s grain? Items move in half-point steps, so small total-score shifts can reflect measurement noise as easily as clinical change. Read one-point differences with humility.

Fourth, remember that ratings are age-referenced. A child who gains skills while age expectations also rise may show a stable total that conceals real growth, or the reverse.

None of this is a criticism of the instrument; it is the honest arithmetic of any judgment-based rating scale used longitudinally, and it is the measurement question our field has been working through in subjective versus objective measurement in autism care and in the six-month reassessment problem. If serial CARS-2 ratings are part of your reassessment workflow, standardize the rater, the setting, the information sources, and the version documentation, and interpret trends rather than single-step changes.

Frequently Asked Questions

What does a CARS-2 score of 30 mean?

It depends on the version. On the CARS2-ST, 30 is the cutoff score itself and marks the start of the mild-to-moderate symptoms range (30 to 36.5). On the CARS2-HF, 30 also falls within the mild-to-moderate range, but that band begins at 28. Report the version with the score, and treat totals near a cutoff as a prompt for careful synthesis with the rest of the evaluation.

What is the autism cutoff score on the CARS-2?

There are two, one per version. The CARS2-ST cutoff is 30, with 30 to 36.5 indicating mild-to-moderate symptoms and 37 to 60 indicating severe symptoms. The CARS2-HF cutoff is 28, with 28 to 33.5 indicating mild-to-moderate symptoms and 34 to 60 indicating severe symptoms. Research has also explored alternative thresholds for screening purposes; cutoffs guide judgment rather than replace it.

What ages is the CARS-2 designed for?

The publisher lists the CARS-2 for ages 2 years and up, reaching through adolescence into adulthood. Children under 6 are always rated on the Standard Version. From age 6 onward, the High-Functioning Version becomes an option for verbally fluent individuals with estimated IQ above 80. For children younger than 2, use developmental surveillance and screening pathways rather than the CARS-2.

Who can administer and score the CARS-2?

WPS lists the CARS-2 at qualification Level C, and in practice it is administered and interpreted by psychologists, developmental-behavioral pediatricians, and other clinicians with graduate-level training in assessment. Parents and caregivers contribute through the unscored QPC, but the clinician makes the ratings. Whether a CARS-2-anchored evaluation can support a formal diagnosis in your setting depends on state licensure and payer rules; see our reference on diagnostic authority.

Is the CARS-2 alone enough to diagnose autism?

No single instrument is. ASD is a clinical diagnosis made by a qualified clinician against DSM-5 criteria, built from developmental history, direct observation, and information about functioning across settings. The CARS-2 contributes a structured severity rating to that synthesis. Most complete evaluations pair it with a developmental interview, a structured observation measure, and adaptive or cognitive testing.

Can I compare a CARS2-ST score with a CARS2-HF score?

Not directly. The forms differ in item content, cutoff scores, and severity bands, so totals do not translate across versions. When a child moves from the ST to the HF between evaluations, describe change domain by domain in narrative form rather than subtracting totals, and make the version explicit in every report.

How long does CARS-2 scoring take?

Per the publisher, making the 15 ratings takes roughly 5 to 10 minutes once the needed information has been collected. The substantive time is in the collection: observing the child, gathering caregiver input (often through the QPC), and reviewing records. Budget the evaluation around information quality rather than form completion.

Where Objective Eye-Tracking Data Fits Alongside the CARS-2

Everything above describes a well-built rating scale doing what rating scales do: converting structured observation and informant report into a quantity a clinician can interpret. The strength of that quantity depends on the observation and judgment behind it, which is why experienced raters and consistent procedures matter so much.

Some evaluation teams now add a different kind of measurement alongside those ratings: instrument-based data that does not pass through a rater. 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. Using eye-tracking technology, it measures how a child visually attends to social information and summarizes those measurements as objective data for the clinician’s review. It is available by prescription and used under the direction of qualified clinicians.

The device does not make a diagnosis, and it does not replace the CARS-2 or any other component of a complete evaluation. It complements clinical judgment by contributing a measurement stream that does not depend on rating decisions, which some teams find useful as convergent evidence next to rating-scale results, particularly when findings are mixed. For a fuller discussion of where objective measurement fits across assessment and progress monitoring, see our clinician’s guide to objective outcomes measurement in ABA care.

The Takeaway

Sound CARS-2 scoring comes down to three habits. Choose the version deliberately: the ST by default, the HF only for verbally fluent individuals 6 and older with estimated IQ above 80. Rate from adequate information, using observation, caregiver input through the QPC, and records. Then interpret the total against the correct table, remembering that 30 and 28 are different thresholds and that a cutoff informs a clinician rather than deciding for one. Used this way, the CARS-2 earns its place among the most practical severity measures available to pediatric and ABA teams.