It’s a little frightening (for me at least) to think that 10 years ago next month (a decade!) I started my MSc in Psychological Research Methods at Exeter University. The course was (and I believe pretty much still is) a load of taught modules combined with a dissertation that was done primarily over the summer term. My project was on identifying the sex of a cat by looking at only its face; both with and without training (essentially a perceptual learning task). I remember quite clearly taking a taxi out to Cats Protection shelter on a snowy January morning to take photos of different cats. Here are two examples of the session:
One of these is male and the other female. Any guesses? (answer is at the bottom of the post).
Over that summer I gave people 40 pictures of cat faces (20 male and 20 female) and asked them to say whether they thought the cat was male or female. Accuracy was 53.94% that whilst somehow turned out to be statistically better than chance (50%), suggests that people really can’t do this very well.
In order to run the experiment I created a program in Visual Basic to present pictures and collect responses from participants. As a sort of training exercise, I tried converting the VB program into a Java applet for web-use. Somehow, not sure how, I managed it and the experiment is still around today (Web Cat Face Experiment) although I’ve long since forgotten how the code works.
Anyway, occasionally data files get emailed to me from the system from where people have attempted the task. I decided to take a quick look at these data files. A total of 31 responses where there. The average performance? 51.13%.
(The cat on the left is female, the one of the right is male).
