How we work in groups depends on the task and on what personalities there are. Sometimes we need strong leaders to direct us. Sometimes we just need to all be on the same wavelength. Does the same apply to animal groups?

To find out, Alex Kotrschal from Stockholm University’s Zoology department and I are working to systematically organise groups of guppies, sorting them according to their behaviour in groups.

It’s hard when we can’t give animals names and questionnaires. Instead, we use a system like league tables in sport. We track the motion of 16 groups of 8 guppies, and rank them according to how well co-ordinated the groups were. Then, to mix things up, we randomly swap guppies from groups next to each other in the ranking.

 

Then we do it all again, and again... Alex K films the groups in the daytime, the computers track the videos overnight, then over breakfast I run some code to analyse the tracks. By (usually) 9am Alex K has the new rankings, and can start the next round of mixing.

What does this achieve? In our new paper, we simulate this process under various assumptions for how individuals with different behavioural traits work together. In all cases, the variation of the behaviour between groups is expected to increase with each round. This is because the mixing allows different combinations of individuals to interact. In many cases, the method also sorts the individuals according to their traits.

Preliminary experimental results indicate that our guppies are becoming sorted. This will allow Alex K to selectively breed the individuals that are consistently in the most co-ordinated groups, to find out just how heritable this behaviour is. If it works, we could breed super-coordinated guppies!

But this is just one of the benefits of sorting animal groups. Researchers can also find out what other traits are correlated with group behaviour. Also, by analysing the data in detail it should be possible to infer how group behaviour depends on individuals - for example, whether it is influenced by strong leaders or weakest links. We believe that our method could sort out many such challenges in studying animal behaviour.