Super Facial Recognition and Natural History: the Results

If you didn't read the post I put up requesting help for a 'research ' project, you can read it here. Thanks to everyone who contributed. It's not come back with anything conclusive, if anything the results are quite difficult to explain but I will try!

To summarise the data: 61 people took part with scores ranging from 29 to 70. My score of 68 was beaten only by three people putting me in the top 7% (thanks for spotting that mistake), at least I'm not making it up when I recognise people all the time! The mean of the whole data set was 58.8, considerably higher than the 53 which was noted as the mean on the website.

In the above chart, you can see the data is split between three categories. This therefore requires a one-way ANOVA (assuming normally distributed data). The data wasn't normally distributed and no amount of data-transformations was going to do anything about that. So we had to go with the non-parametric Kruskal-Wallis test (X2=3.65, P=0.16, n=61). And no significant differences were found. I was intrigued to compare just the 'Naturalists' with 'Pan-listers' using a Mann-Whitney test and the same was true but this result was approaching significance (Z=1.82, P=0.07, n=61). 

This would be quite interesting if the 'Non-naturalist' category wasn't sitting right in the middle. All I can think is, is that many of the PSL and naturalist types just did the test while perhaps the people who didn't see themselves as naturalists only did the test if they thought they had pretty good facial recognition. In fact, this was the group that had the least submissions. All of the groups were way above the mean on the website of 53. I believe my data collection method was causing an intrinsic bias across all three classes as only people who thought they might score well were entering. Who knows. All thoughts on this are most welcome.

I'll stick to the ecology.

But is there a correlation with PSL list size and test score I hear you say? Well, no is the answer (F=0.05, P=0.82, n=21). A big fat no at that. That line can barely get more level.