Reading the Mind in the Eyes
Guess emotion from the eyes
About this trainer
You are shown a close-up photo cropped to just the eye region of a face, with four emotion or mental-state words around it (for example: jealous, panicked, arrogant, hateful). Your job is to pick the single word that best describes what the person is thinking or feeling. The standard set runs 36 photos.
What it develops
It trains social cognition: reading subtle emotional and mental states from the eyes alone, plus the emotion vocabulary you need to label those states precisely. In practical terms you are exercising rapid, fine-grained facial-cue interpretation.
History
It began as a psychology research instrument, not a game. Simon Baron-Cohen's group introduced the original 'Eyes Test' in 1997 as an advanced theory-of-mind measure for adults, then released a revised version in 2001 after the first one was criticised. It spread fast, has been used in well over 250 studies and translated into many languages, and the eyes-and-words format was later adapted into popular online quizzes and brain-training drills.
Who created it — and when
Simon Baron-Cohen and colleagues at the University of Cambridge. The original test is from 1997; the widely used revised version (Baron-Cohen, Wheelwright, Hill, Raste and Plumb) was published in 2001 in the Journal of Child Psychology and Psychiatry.
How to train
Look at the eyes first and form an impression before you read the four words, so the labels do not bias you. Pay attention to concrete cues, the openness of the lids, the angle of the brows, the direction of gaze, and rule out the options that clearly do not fit. When you miss one, stop and learn the word: many errors are vocabulary gaps, not perception gaps, so building a richer feelings vocabulary directly raises your score.
How long to practise
Short sessions work best: one pass of 36 faces takes only a few minutes. Two or three focused sessions a week is plenty; reviewing the items you got wrong matters more than sheer volume.
Evidence base
What is solid: the test reliably separates groups on average, autistic adults tend to score lower than non-autistic adults, and meta-analyses report a small female advantage. With practice you also get better at the task itself. What is shaky: a growing body of work questions what it actually measures, arguing it may tap emotion-word knowledge and recognition more than genuine 'mind-reading', and reports weak internal consistency and a poor factor structure, with some items behaving badly. Links to self-reported empathy are inconsistent, and the well-known finding that oxytocin boosts Eyes-Test scores has struggled to replicate. Treat claims that it trains broad real-world social skill or empathy as unproven.
Recommendations
Treat every miss as a vocabulary lesson, not just a wrong guess, and look up the exact word before moving on.
FAQ
Is a low score a sign of autism?
No. It was built as a group-level research measure, not a diagnostic tool. Scores vary widely among people without autism, and on its own a low score tells you very little about any individual.
Will practising it make me better at reading people in real life?
Mostly it makes you better at this specific task. You will likely improve at picking the labelled word from the eyes, but evidence that this transfers to everyday social skill or empathy is weak and contested.
Why are some answers so hard even when the eyes look obvious?
Often the bottleneck is the words, not the face. The options include subtle, uncommon terms, so if you do not precisely know a word like 'contemplative' or 'apprehensive' you can read the emotion correctly and still choose wrong.
Variants
There is a child version with simpler wording, and translated editions in many languages with locally validated answer keys. Related drills use whole faces instead of just the eyes, dynamic video clips of expressions, or voice and body-posture cues, and some apps add a timer or feedback after each item.