What I am fascinated by is why animals do the things they do. This means I study animal behaviour, and the ecological and evolutionary forces that underpin the amazing variation we see in the natural world. I have a particular interest in social behaviour, so why two or more animals interact, what happens when they do so, and what consequences this has for their lives and the evolution of populations. I’ve grouped my research into three broad sub-sections, each with links for further reading. They are “Social networks of animals“, “Evolution when animals interact“, and “Among-individual differences“.


Social networks of animals

Animals interact in diverse ways, making analysing social interactions consistently across species and contexts difficult. Thankfully, within sociology the techniques of “social network analysis” were developed to study human social interactions. A social network is a way of representing interactions among individuals, by representing individuals as “nodes” or “vertices” and linking them with “edges, “links” or “ties” connecting them if they interact. They are very analytically tractable, and so have been seized upon by those studying non-humans to help us understand animal social structure. Animal social behaviour has of course been studied for a very long time, with advances in quantitative genetics and the appreciation of “indirect genetic effects” (see below) being used to understand the evolution of traits involved in social interactions. There are therefore multiple fields alongside social network analysis that seek to understand the social worlds of animals. I, along with Dr. Andrew McAdam, have reviewed how these different fields have taken different approaches, and investigated different sets of questions, but have much to teach other. If both fields can appreciate that animal’s social and non-social traits co-vary in dynamic ways, we should achieve a greater understanding of social interactions and their ecological and evolutionary consequences.

MM net SN
A cricket social network (males fighting each other in 2006)

Initially the use of social network analysis tended to focus on highly social animals that live in groups, and interact in various cooperative ways, e.g. dolphins and chimpanzees. However, essentially all organisms interact, when they mate, fight, move, cooperate and compete with each other. Hence there is no reason to limit the use of social network analysis to only highly social animals. I have applied social network analysis to multiple generations of the cricket Gryllus campestris, to study whether they social networks are stable across generations, despite the complete turnover of individuals, and to investigate classic questions in sexual selection using the tools of network analysis. Interestingly, cricket networks are relatively stable across generations, despite all crickets dying 9 months before the next generation starts forming the new network, and differences in population size across years is more important for differences between networks. Meanwhile, the network structure indicates that promiscuous males tend to mate with promiscuous females, which, as female crickets store sperm and place males in sperm competition with each other, the males that mate the most may not be as successful as it appears, as they may lose some paternity to others the female mates with. Males who mate infrequently may instead guard their mates more, meaning they mate less often as well. Those males that mate a lot are also the males that fight a lot, suggesting among-individual differences in “quality”, rather than trade-offs for different strategies. Finally, males don’t seem able to avoid being in sperm competition with rivals by fighting them, as males that fought more were also in more intense sperm competition.


All work on the crickets was with Prof. Tom Tregenza and Dr. Rolando Rodríguez-Muñoz, who make up the Wild Crickets project.

While studying the social networks of crickets, I came across several really cool methods for analysing social networks that people studying humans were using, but hadn’t really been adopted by those studying other animals. Myself and Dr. Matthew Silk thought this was a shame, and so we have written a couple of review articles, firstly reviewing how stochastic actor-oriented models can be used to analyse social networks changing through time (co-written with Dr. Amiyall Ilany and Tom Tregenza), and then reviewing how exponential random graph models (ERGMs) can be used to understand animal social structure. I’m always on the look-out for the best methods to analyse complex animal social structures, and am currently working with Matt and Dr. Julian Evans to compare ERGMs to other commonly used methods of network analysis. Stay tuned for those results, vast simulations are running on a super computer!

Evolution when animals interact

Most of the time in evolutionary biology, we consider how an organism’s traits relate to its fitness, and how its genes influence the expression of that trait, and thus predict how that trait may evolve. However, when animals interact with one-another (which they tend to do all the time, such as when competing for resources) this picture gets a lot more interesting/complicated (depending on what side of the bed one got up from that morning).

ML selec diag
Multi-level selection can act at a range of spatial scales, and may be strongest locally

Firstly, when animals interact in groups, traits of those groups can influence the fitness of the individuals. For instance, the degree of cohesion shown by a hunting pack, which is not a trait one can attribute solely to a single individual, will influence the amount of food the pack members get, and so their reproductive output. This demonstrates that selection is “multilevel”, acting on genes, aggregations of those genes in individuals, and aggregations of individuals into groups. I have demonstrated that this extends to individuals that do not live in groups, but to solitary animals that can be aggregated at different spatial scales. My research on North American red squirrels (Tamiasciurus hudsonicus) demonstrated that selection is for earlier breeding in the spring at a local level (so you want to breed earlier than others within 130m) but at larger spatial scales selection is for later breeding, presumably as breeding very early in the Yukon spring is very difficult. Local selection for early breeding was more intense when population densities were high, suggesting the benefit of early breeding is to allow pups to acquire a vacant territory. Meanwhile, selection on growth rate is for faster growth at all scales, and is unaffected by population density, suggesting absolute growing speed is what is key, without regard for what others are doing.

Secondly, when animals interact, they can influence each other’s traits. If this effect has a genetic basis (i.e. individuals who influence others more have relatives who do likewise) then some of an individual’s traits are determined by the genes of those it interacts with, not just its own genes. This can have fascinating and counter-intuitive consequences for evolution. For instance, in red squirrels I have found that individuals that tend to give birth early also cause their neighbours to give birth later, but only at high population densities. At low densities squirrels seem to leave each other alone a bit more. When selection is for early breeding, the next generation of squirrels will both have genes to breed earlier, but also to suppress their neighbours’ breeding more. This means the average date of breeding actually won’t change as much as we expect it to, as squirrels are suppressing each other more in each generation. Amazingly, this means that the change in parturition date is expected to be faster at low densities, despite selection being weaker, than at high densities, when selection is strongest; mind blown!

IGEs on PD Fig2 Results summ4

I’ve also been looking at how the previous owner of a squirrel’s territory can influence it, “from beyond the grave”. Results are not finalised yet, but I can give you a sneak peak: it looks like males cache the most white spruce cones (Picea glauca – a red squirrel’s favourite food), and leave the most behind when they die, meaning the next owner of the territory has more cones, can give birth earlier, and has a higher lifetime fitness.


Its pretty amazing that a dead squirrel can have such an effect on a live one! Meanwhile, squirrels have the most cones at intermediate ages, and so it is indeed the squirrels that die at intermediate ages that leave behind the most cones for the next owner. How these effects of “temporal neighbours”, alongside the effects of “spatial neighbours” identified above, influence evolutionary trajectories appears to be really interesting, with much still to be discovered! 

All work on red squirrels has been with Andrew McAdam , as well as all members of the Kluane Red Squirrel Project.

Among-individual differences

Behaviour of an animal could in theory change completely from one moment to the next: a songbird could switch from a loud territorial song to sudden fearful flight just as quickly as it can register a potential predator. This then suggests that animals should be able to adjust their behaviour to match any given situation. However, they do not. Research in the last 20 years has regularly found that individuals are limited in their behavioural repertoires compared to what occurs across an entire population, and that some individuals are consistently risk-taking, or active, or aggressive, compared to other individuals. A large sub-field has therefore developed to explain these among-individual differences, also known as “personalities” or “coping styles”.

In crickets for instance, some individuals emerge quite quickly out of shelter in a lab, while others may not emerge at all, while once out, some crickets run around a lot, while others do very little. I, working with Dr Morgan David, demonstrated these differences are somewhat consistent over the adult lifetime of individuals, which they need to be to be related to stable life-history strategies, one of the requirements of several leading theories for the evolution and maintenance of among-individual differences. In fact, among-individual differences tend to get greater as individuals age. These among-individual differences in activity have consequences in the wild too; crickets that ran around a box in the lab more also ran around in the wild more the next day. I also found that those that were consistently more active also had shorter lives, suggesting all that activity lead them into trouble. They did however have equal lifetime mating success, suggesting they got more done in their shorter time on earth. Interestingly, the time it took for a cricket to leave a shelter in the lab was not related at all to the time it took an individual to leave its burrow after a disturbance in the wild. We must be measuring different things in the wild and the lab for these two behaviours, even thought they look superficially very similar. This work was conducted with Adèle James.

Reading and writing about among-individual differences all the time, even among clones raised in identical environments, got me thinking that they should be considered the norm, not an unusual phenomenon requiring explanation. But could there be something that causes among-individual differences to arise out of essentially equivalent beginnings? Working with Joseph Burant, and Matthew Brachmann, I have written a forum piece, discussing how insights from chaos theory, and the study of complex systems, could bring light to this problem.

Lorenz Attractor
A Lorenz attractor, showing how adjacent trajectories can quickly diverge, in a deterministic system

Essentially, a chaotic system is one where greatly divergent results can arise from very similar starting points. Weather systems are a good example, as even utterly tiny errors of measurement can mean models give inaccurate predictions if we extend them too far into the future. Something similar might be happening as animals develop: small differences in early environments, parental provisioning, or differences introduced by genuine stochasticity, could lead to adults that behave quite differently from one-another. I’d love it if someone could directly test this idea!