# A Reasonable Approximation

### Officially faceblind

I realized a few years ago that I was at least somewhat faceblind/prosopagnosic. A while back I took an online test out of curiousity, and scored low. They said that if I was in London and interested in further tests, I should leave my email address. A few days ago I went in for those tests, and now I have a PhD student (Katie) also telling me I'm faceblind. Which makes it official, I guess.

Next she wants to run EEGs on me, which should be cool. That will help work out where my brain is going wrong, in the long chain between "photons stimulate nerve endings in my eyeballs" and "I recognize a face" (whatever that means). Also, apparently there's a phenomon which sounds to me like blindsight, where some prosopagnosics' brains are clearly reacting to faces on some level that doesn't reach their consciousness. She wants to learn more about that too.

What follows is discussion of the tests, my scores, and what they mean. I've been given a powerpoint with my scores reported as percentiles, along with absolute scores and average control scores. 2% or lower is counted as "impaired". Percentiles are only given as integers, or as "<1%". On the day, Katie also gave me some numbers in terms of standard deviations (σ). Under a normal distribution, 2.5% would be approximately -2σ, but I'm not sure any of these results will be normally distributed, so I don't know if σ scores really tell me anything.

A note: if you think you might be faceblind, and you'd be interested in getting more detailed tests, it might be a good idea not to read the below. I expect it wouldn't significantly bias the results if you did, except for one bit that I've rot13ed. But I don't trust myself to make that call. If you're in London, you can take the above test like me and see what happens. Otherwise I'm not sure how you'd go about getting more tests.

The object/face recognition tests were "memorise these things, then we show you a sequence of things and you have to say if each of these things was a thing in the first set". The things were houses, cars, horses, and bald women's faces. I was bad at all of these: 4% for cars, 2% for houses, and <1% for horses and women. (Average score was higher for women than horses, and my score was higher for horses than women, so I'm worse at women than horses. I think Katie told me I was somewhere between -5σ and -6σ for women. Under normality, -5σ is one in thirty million, but this is clearly not normal.) So it seems I have some level of general object agnosia, but more specific prosopagnosia on top of that.

I was 11% for reading emotions from eyes, which is a point for Team Phil Does Not Have Aspergers (some of my friends are divided about that). In fact, the average score is 26 and I scored 23, and there were a few cases where I said an answer, then thought "wait no it's this" and didn't say anything because I wasn't sure if I should. (I was speaking my answer and Katie was recording it. I had to choose from four emotions, so I'm not sure why this wasn't recorded by a computer like most of the other tests.) So plausibly I'm actually above 11%.

I was <1% at famous face recognition, recognising five out of fifty that I'd been exposed to, out of sixty in total. (I got Jvyy Fzvgu, Uneevfba Sbeq, Dhrra Ryvmnorgu, Ebova Jvyyvnzf, and surprisingly Ovyy Pyvagba.) It seems that controls tend to get above 40, so even counting that "exposed to" is vague, I did really badly at this. I think Katie said I was -9σ, which would be one in 10^19 under normality.

I'm <1% at the Cambridge Memory Test for Faces, which is the one I linked above. I actually scored worse in the lab than online. (47% versus 58%, IIRC, with a control average of 80%, and 60% indicating impairment. But the lab score I've been given is 34 against control of 58, so it's clearly been adjusted.) There could be any number of reasons for this, including "chance". But when I took it online, I often thought that one of the faces looked a little like Matt Damon, and chose that one. I like to think that "mistaking people for Matt Damon" is the way forward in face recognition.

I was somewhat okay at half of the Cambridge Face Perception Test. In this one, they showed me a face at an angle, and below it the same face face-on, six times, with varying degrees of modification. I had to order them according to how similar each was to the original face, within a minute. For half the tests, the faces were all upside down. For all of the tests, they all looked incredibly similar and my instinctive reaction was WTF.

On the upright test, I got <1%. On the inverted test, I got 7%. One strategy I used a few times was to focus on the lips, specifically on the size of the dip in the bow. I just ordered them according to that. I guess it helps, but I found it a lot easier for inverted faces.

Doing better at inverted would seem to suggest that I'm doing some kind of holistic face processing that goes wrong and blocks off later avenues for perception. Buuut, objectively I scored worse on the inverted faces, just not as much worse as average, so I'm not sure if this does suggest that. (And I'm not sure it is "objectively" - if all the faces had been assigned to the other condition, would my scores have changed?)

Hypothetically, high scores on both tests could indicate my problem is with memory, not initial perception. The low score here doesn't mean I don't have a problem with memory, but it does seem to hint that I do have a problem with initial perception. And I suspect the famous faces test points at me also having a memory problem.

Posted on 19 January 2016