Lookalikes

Can we really see the features of parents in little babies?

Genetically it makes sense that babies look a bit like their parents - they do, after all, share 50% of their genome wity each parent. Eye colour, nose shape, etc, are going to be obvious indicators. It makes evolutionary sense, too: in the days before paternity testing, it may have been necessary to convince a father that the child was indeed his and therfore worthy of his time, effort and resources.

A quick survey of the panel of old wives reminds us of these common phrases:
"He looks just like his dad."
"She's got your eyes."

Science Dad is sceptical. Whenever I hear these phrases applied to a child in front of me,  I often don't agree, and wonder if it's just something people say to makeconversation, alleviate potential concerns of dubious paternity, or at least to strengthen the parent/child bond. I'm more inclined to believe the null hypothesis:
"All babies look more or less the same"

Science Mum is adamant she can identify these features. An experiment is needed!

I scanned through my facebook friends (only the ones that my wife hasn't met) looking for pictures of mums, dads and babies. I found 4 good candidates (4 mothers, 3 fathers, 4 babies), with a handful of pictures for each person to avoid too much repetition. I then coded up a little survey and tested her.

It looked a bit like this (I've blurred it to protect identities):

The answer was the left option 25% of the time, the right option 25% of the time, and the middle option ("neither") 50% of the time.

Science Mum picked the correct option 68% of the time - much better than as if by chance. That said, she hadn't twigged that there were only 4 children - she thought there were a lot more! Her comparison was therefore on features alone, she hadn't recognised the children.

I tweaked the survey to see if she could match each parent to one of the four children. It looked more like this:

This time she got the right answer 75% of the time.

She was quickly identifying the children using subjective criteria: "plain one", "better looking one" etc. She had failed to do this in the previous version of the experiment, but now seeing the children side by side she was able to build quite a reliable face recognition index. Soon it didn't matter which of the photos of the child were used, she could tell them all apart easily (much better than I was able to do). Matching to the parents was then a matter of comparing obvious features such as nose shape, hair colour etc. Two of the children were easily matched to their parents, but the other two were harder because the dads looked similar.

The conclusion: Science Mum did better than Science Dad had expected. She's clearly superior at face recognition, which perhaps gives her a more extensive feature set for making comparisons between faces. Nevertheless she wasn't getting 100% accuracy, so Science Dad reckons he's still won the argument.