Home
About
Blog
Info hubs 
Autistic burnout
AuDHD
Autism strengths
Autism tests
Autism assessments
Store
Contact
The Reading the Mind in the Eyes Test (RMET) is a 36-question task that explores how you recognize emotions and mental states from photographs of people’s eyes. For each photograph, you choose one of four descriptions. Your score shows how often your answers match the test’s scoring key.
Can the RMET tell you whether you are autistic? No. Although it is widely used in autism research, a high score does not rule out autism, and a low score does not confirm it. The test captures a narrow part of social perception—not your overall empathy, compassion, or ability to understand other people.
If you are wondering, “Could I be autistic?”, questionnaires about your everyday experiences offer a more relevant starting point. The AQ and RAADS-R explore autistic traits, while the CAT-Q explores social masking and camouflaging. All three are free to complete and receive your automated scores, although none can diagnose autism on its own.
Basic information
Test items: 36 scored items in the revised adult version
Timing: The standard test has no time limit
Type: Performance task involving mental-state recognition from eye photographs
Developers: Simon Baron-Cohen and colleagues
Publication:Original version: 1997; revised adult version: 2001
Score: 0–36; one point for each answer matching the scoring key
Or choose a questionnaire about autistic traits or masking.
What does the RMET measure?
The RMET was developed as a task involving theory of mind: thinking about other people’s mental states. However, matching a word to a photograph captures only a small part of understanding another person. The original revised-test paper itself distinguished identifying a mental state from understanding what caused it.[1]The Reading the Mind in the Eyes Test revised version (Baron-Cohen et al., 2001)
A meta-analysis found that RMET performance was somewhat more closely related to emotion perception than to other theory-of-mind measures. Vocabulary also mattered, while relationships with measures of cognitive and affective empathy were weak. The authors therefore questioned interpreting it as a broad theory-of-mind measure.[2]Sty in the Mind’s Eye (Kittel, Olderbak, & Wilhelm, 2022; first published online in 2021)
Your score is not a measure of your kindness, compassion, or worth as a friend or partner.
Instructions
For each photograph, read all four words and select the one you think best matches the person’s thoughts or feelings. If more than one seems possible, choose the closest match.
- Complete the practice item, if provided, to become familiar with the task.
- Work through the photographs at your own pace. The standard RMET is not a speed test.
- The standard test includes a glossary. If a word is unfamiliar, consult the official adult test materials.
- Wait until you have finished before opening the answers below.
The task format, glossary, and absence of a time limit are described in published research.[3]Predictors of performance on the Reading the Mind in the Eyes Test (2020) Translations are available from the Autism Research Centre.
What does my RMET score mean?
Your total is the number of answers that match the scoring key, out of 36. A higher score means you matched more of the test’s intended answers.
Can a high score rule out autism?
No. Recognising expressions in these photographs does not rule out autism. The RMET does not assess the full range of experiences relevant to an autism assessment.
Does a low score mean I am autistic?
No. Difficulty with this task is not specific to autism. A low score also does not establish that you lack empathy.
Research-group averages are not personal diagnostic thresholds. In particular, results from a small study using a culturally adapted version should not be presented as universal “autistic” and “non-autistic” scoring ranges. The Autism Research Centre cautions that its tests are not diagnostic by themselves.[4]Autism Research Centre: Frequently asked questions
If autism is your question, consider your experiences over time rather than treating this score as a verdict. The questionnaires below can help you reflect on a broader range of experiences.
What does research say about the RMET?
Group differences do not make it a diagnostic test
A 2019 meta-analysis of 18 studies found that non-autistic comparison groups scored higher on average than autistic groups. However, a difference between group averages does not establish a cutoff that can diagnose an individual.[5]The Reading the Mind in the Eyes Test in autism-spectrum disorders: A systematic review and meta-analysis (Peñuelas-Calvo et al., 2019)
Consistency does not establish what the score measures
Kittel and colleagues’ meta-analysis reported acceptable average internal consistency (α = .73). That concerns how the items work together; it does not establish that the test measures overall empathy, everyday social understanding, or diagnostic status.
A 2024 review examined 1,461 studies using the RMET. 63% did not report validity evidence in any of the six categories examined, and reported evidence frequently failed to meet accepted standards. The authors raised substantial concerns about interpreting RMET scores. Missing reporting does not itself prove that every study is invalid, but it limits confidence in the conclusions drawn.[6]Construct validity evidence reporting practices for the Reading the Mind in the Eyes Test (Higgins et al., 2024)
Researchers disagree about how far these criticisms should go. Murphy and Hall argued that the review underweighted evidence linking RMET scores to other meaningful measures and outcomes. Higgins and colleagues defended their concerns in a subsequent response. The debate remains open.[7]How a strong measurement validity review can go astray (Murphy & Hall, 2024)[8]A response to Murphy and Hall (Higgins et al., 2025)
Some answer choices are ambiguous
A 2026 study of 640 university students found that performance was associated with familiarity with the answer words. On a quarter of the items, an alternative answer was rated similarly to the intended answer as a description of the photograph. Scores showed a small association with only one of two self-report cognitive-empathy measures. These findings raise questions about the items and their interpretation; the study was not a clinical test of autism diagnostic accuracy.[9]A multi-site psychometric evaluation of the Reading the Mind in the Eyes Test – Revised (Hafner et al., 2026)
The multiple-choice format affects performance
In a 2023 study of 70 autistic and 71 non-autistic adults, both groups performed much better on a 10-item multiple-choice version than when asked to generate their own descriptions. However, group differences remained after controlling for verbal IQ, and both formats were related to another theory-of-mind measure. The task is therefore influenced by its format, but these findings do not support describing it as merely a vocabulary test.[10]Response format changes the reading the mind in the eyes test performance of autistic and non-autistic adults (Lim et al., 2023)
Alexithymia may also matter
In a 2016 study, autistic participants and non-autistic participants matched for alexithymia did not differ on RMET performance, although they differed on another social-cognition task. This suggests that difficulty identifying and describing emotions is relevant to interpretation. It does not establish that alexithymia explains every autism-related difference reported in the research.[11]Theory of mind is not theory of emotion (Oakley et al., 2016)
Photographs are different from everyday interactions
The RMET presents isolated, still images with four possible labels. In everyday interactions, we usually have a whole face, voice, body language, and context. A correct answer also does not reveal whether you recognised the expression immediately, analysed it deliberately, or guessed.
Use the result as a prompt for reflection, rather than a verdict on your ability to understand or care about other people.
Our experiences with the RMET
These are personal reflections, rather than research findings or diagnostic examples.
Ember:
I scored 35/36, which is above the average ranges reported in some studies. I tend to consciously analyze facial expressions by looking for specific cues around the eyes and eyebrows.
One limitation I notice is that the test uses only Caucasian faces. There also appear to be gendered patterns in how some of the emotions are represented.
Kendall:
When I first took the test, I scored 19/36. Eight months later, I scored 27/36.
I often used process of elimination or guessed when I wasn’t sure. I may also have remembered some of my previous mistakes, which could explain part of the improvement on my second attempt.
Without the word choices, many of the photographs would have been very difficult for me to interpret. For me, one useful aspect of the test was identifying emotions and mental states that I may need to take extra care interpreting.
Eva:
When I first took the test several years ago, I remember getting only a few items wrong. When I retook it in 2024, however, I got 6 items wrong, despite feeling confident in my answers.
What interested me most was the pattern of my mistakes. In the past, I tended to interpret ambiguous facial expressions more negatively. This time, I often interpreted expressions as more positive or innocuous than the intended answers.
My experience also highlights something important about the RMET: recognizing mental states from isolated photographs of the eyes may not reflect how we understand people in real-life social situations, where we usually have access to the whole face, body language, voice, and context.
Comments
Let us know what you think!