27 without mimicking expressions, 30 with -- which is probably a small enough spread to be explainable by chance.
Oddly, though, there's almost no overlap between the answers I got wrong: only one mistake appears on both lists. This might indicate that different reasoning pathways are being engaged, but a sample size of one is far from conclusive; did anyone else notice something similar going on?
I seem to recall this site having a high population with some sort of Autism Spectrum Disorder, and so you might be interested in an online test of eye-reading ability, which seems to be linked to ASD.
But once you take it, just read your score and open the test up in a new tab- don't see which eyes you got wrong. Then retake the test, but imitate the photo each time, and then guess.
Does your score increase? Compare the lists of eyes you got wrong (the benefit of doing this in two tabs). Anything interesting?
The test is here.
From The Last Psychiatrist.