The Ultimate Coalition: Averaging the Portraits of UK Political Leaders


The image above may look strangely familiar to you. That’s because it’s a facial average of the leaders of the main political parties in this week’s general election in the UK. If you’ve been following UK politics, you have probably seen these people many times in the media, leading to an involuntary familiarity with them.

A facial average like the one above is created by digitally altering each person’s face to a matching position and expression, and then morphing them all together to create an average.

This process minimises all of the differences contained in each individual picture – for instance lighting, expression and background. It captures the unique features of each individual, and produces an image of them as one coherent identity.

This picture illustrates the average pictorial representation of David Cameron (Conservatives), Ed Miliband (Labour), Nick Clegg (Liberal Democrats), Nigel Farage (UKIP) and Natalie Bennett (Green Party).


Depending on your level of familiarity with each of the candidates, one individual or a particular feature, may appear to dominate the image – like Nigel Farage’s eye bags, Ed Miliband’s hair-line or David Cameron’s forehead. This demonstrates that having prior experience (i.e. seeing someone’s face a lot) influences the way in which we perceive the world around us. You have may experienced this effect, for instance, when you meet someone and notice similarities between other people you know.

Here’s another example, of all of the prime ministers since 1900: you might notice the ghost of John Major’s glasses.


The images were generated using a software called psychomorph. The program enables the creation of clear facial averages by processing the texture of the face indepedent of things like facial expression and camera angle. This is achieved by placing multiple fixed points onto the face’s features to create a shape template, seen below.


Geometrically, these feature-related points form lot’s of tesselating triangles, which are then used to manipulate the pixels – as if a rubber sheet were being stretched onto another shape. This template can then be used to morph all of the images in the average to a common expression and angle, which avoids producing a blurry or messy image.

The programme was developed in javascript by Dr Bernie Tiddeman at Aberystwyth University in order to generate stimuli for various academic experiments. If you want to try doing this yourself the programme is available on his website for non-commercial usage, along with a detailed wiki.

Statistically, these face averages are likely to be judged as more attractive than any individual party leader. This would be especially true for someone who has not seen any of the candidates before – although you may disagree depending on your personal perceptions! This could occur because potentially unattractive – but often unique – features are averaged out, creating a less distinctive, but generally more attractive, composite.

An evolutionary explanation can be offered to account for this attractive-average effect. It has been observed that facial averageness is associated with a greater genetic diversity in the genes that code for immune response. Selecting a sexual partner with a more “average” face could therefore benefit any future offspring’s immune system, giving an evolutionary advantage to finding these faces attractive.

Recently, these averaging techniques have demonstrated more practical applications when applied to images of the same person. Researchers at the University of York have shown that individuals are able to extract an average representation of a person’s face. Participants were shown pictures of several unfamiliar people, and later asked to identify the images they had seen. Participants in this experiment claimed to have seen the average image of these faces – despite only being shown the individual pictures. This offers the possibility that we become familiar with someone through automatically storing an average representation of people we are familiar with in our brains.

Averages of your own face can even be used to improve your smartphone’s facial recognition capabilities. Through combining as many images of your face as possible, the average created reduces all of the idiosyncrasies, usually found in each individual picture, to a minimum. This reduces the work an algorithm has to do in order to correctly identify you through the phone’s camera, and therefore increases its reliability. This was demonstrated by investigators at York FaceVar Lab using the Samsung Galaxy range’s security system — a face average improved the ability of users to unlock their phones with the front-facing camera.

Facial perception can also influence the actual decision that we make about people, even in elections. Researchers at Princeton University asked participants who were completely unfamiliar with candidates in gubernatorial (elections, to judge the candidates on their competence, based only on their faces. The participants only saw each face for a fraction of a second. The results of these competence judgements actually predicted the outcome of the elections for 68% of state candidates and 72% of senate elections. This either suggests that a person’s face influences voting behaviour, or that a face tells us something about a candidate’s suitability for the position. Either way someone’s face affects the way we judge him or her more than we might think – and the idea that the public might vote on someone’s face, rather than their policies is slightly disconcerting.

About the author: Alexander Irvine is a Research Assistant in the Eye Movement and Attention Laboratory at the University of Aberdeen. He can be contacted at [email protected]. The article originally appeared in The Guardian and has been republished here with Irvine’s permission.