A new way of analysing the social networks that link individual animals to each other has been tested on wild great tits by Oxford University researchers.
How animals associate in groups can have important consequences in terms of the health and survival of both individuals and whole populations, and can influence factors such as the spread of disease and the ability to find food or mates.
But revealing the networks underlying animal societies is a challenge when a large amount of fieldwork data consists of a long stream of automated observations of the times and locations of individuals, leaving scientists to try and reconstruct the 'big picture' of how individuals are connected.
The new approach can automatically identify periods of intense social activity within a large number of observations – in this example around one million observations of wild great tits (Parus major). This makes it possible to examine these periods in greater detail and calculate which individuals are real 'friends', rather than random passers-by, and even which are looking to pair up and mate.
A report of the research is published in this week's Journal of the Royal Society Interface.
'If you think of the data about you in Facebook it records things like who you are friends with, where you've been, and what you share with others,' said Ioannis Psorakis of Oxford University’s Department of Engineering Science, who led the research. 'What we have shown is that we can analyse data about individual animals, in this case great tits, to construct a "Facebook for animals" revealing who affiliates with who, who are members of the same group, and which birds are regularly going to the same gatherings or "events."'
The researchers team tested the new technique on data from two breeding seasons of wild great tits (August 2007 to March 2008 and August 2008 to March 2009). The data came from transponders attached to thousands of birds and sensors that logged when individuals appeared at any one of 67 bird feeders spread throughout Wytham Woods, near Oxford.
The researchers found that their predictions from this data about which birds were 'friends' that regularly foraged for food together, as well as which birds were starting the process of pairing up or were already in a pair, matched visual observations made by zoologists.
'What we've shown is that our technique can extract information about the networks that bind individuals together by sampling and analysing their mobility patterns,' said Ioannis Psorakis. 'Our approach makes it possible to look at huge amounts of data without having to decide what time resolution is best to extract meaning – the model evaluates this automatically. This is just the first example of how zoologists are beginning to use our method to explore social networks of animals in a "big data" context.’
Early results from the work with great tits suggest that individual birds do not participate in flocks at random, but have a bias towards other members of the population they interact with. The majority of networks extracted using the approach are strongly clustered, and in such tight bird communities, individuals forage together and interact with their current or future mating partner.
This approach is being used not only in great tits, but also in a mix of wild-bird social networks, exploring the animal sociality at an inter-species level. Some of the most important future steps in this work are to combine the 'social' information available through this method, with other types of information: for example combining it with genetic data is enabling researchers to explore the genetic basis of sociality: do genetically similar individuals attract each other, or is it the other way round? Can we find specific areas of the genome that account for gregariousness?
The work could also help researchers understand how information spreads through animal populations. Tits are a famous exemplar of social learning: their habit of pecking open milk bottles on doorsteps to get access to cream spread rapidly through England in the mid 20th century. The new approach is helping scientists to test how specific social structures help or hinder the spread of novel information from individual to individual.
The research team was led by Professor Ben Sheldon of the Department of Zoology and Professor Stephen Roberts of the Department of Engineering Science and included Dr Iead Rezek of the Department of Engineering Science. The work was supported by Microsoft Research.
A report of the research, entitled 'Inferring social network structure in ecological systems from spatiotemporal data streams', is published in Journal of the Royal Society Interface.