Last year, I attended a symposium hosted by Peter Kappeler at the German Primate Center on the topic of “social complexity”. A bunch of evolutionary and behavioral ecologists from different backgrounds got together to argue about stuff like ‘How should we define social complexity?’, ‘Is the brain size of a species a good of measure of social complexity? (or anything at all?) and “Why does Germany have so much more science funding than us?”
This was actually one of the best conferences I attended, I met a bunch of people whose work I’d read, and I started useful collaborations, including one with a Msc student working on neophobia, who convinced me that bats reacting to novel objects was actually interesting and helped me write this paper. With others from Damien Farine’s lab, I wrote my opinion about the conference here.
What’s more socially complex: a bee or a chimp? (Answer: vampire bat). Biologists tend to define social complexity in such as way that their animal is one of the ‘complex ones’. One possible solution to this that I liked for talking about social complexity was the terms proposed by Dieter Lukas and Tim Clutton-Brock: “organizational complexity” (exemplified by cooperative breeders and eusocial insects which have a division of reproduction and labor) and “relational complexity” (exemplified by animals with individualized relationships like some primates, elephants, and of course vampire bats). They published evidence for these distinct dimensions of complexity in mammals here. Others suggested other frameworks.*
The invited speakers were asked to write up their talks as papers for this special issue in Behavioral Ecology and Sociobiology. I gave a talk about vampire bats and I contributed to an invited talk/paper on social complexity across bats, led by Jerry Wilkinson.
Here’s the link to that paper, Kinship, association, and social complexity in bats (if that link doesn’t work look under “publications” above). The basic idea is that comparing social networks across species is actually quite difficult and rarely done because different studies measure ‘groups’ and ‘associations’ differently across species. But bat researchers tend to do the same thing: we individually mark bats, then observe which bats are in a roost across different days, and when we mark them, we collect a tissue sample to estimate their genetic relatedness. Several people have done this over several years. So Jerry gathered all the data together from different researchers that have done long-term studies** and we did the same basic social network analyses in each species to see if anything interesting came up.
To be honest, there wasn’t anything too surprising if you know the social and genetic structures of these different bat species, but it was quite nice to put it all together in one place and to measure all these species using the same metrics. For some species, it did change my picture of their social structures as being a bit more ‘messy’ that I thought. I also noticed was that we actually had different conclusions about relatedness and association for some species than previous published analyses, suggesting that the details of how you measure relatedness and association can determine what you conclude. Another lesson was that the link between relatedness and association can depend a lot on what your null model accounts for. This is something Damien has often written about. For example, if two individuals are always seen together but always in the same roost, then do they actually prefer roosting near each other or do they just prefer the same roost and they don’t actually care about each other? It’s actually easier to infer social structure for animals that switch roosts and move around because you can account for spatial effects. Another example is that a null model that does not account for time effects could lead to the idea that two individuals are highly associated simply because they both died in the first year of the study. Social networks are inherently correlational, and it’s quite easy to draw the wrong conclusions if you don’t think about and correct for these kinds of biases.
Another nice thing that you don’t see in the paper is that we ran all the same analyses in both Matlab (‘Socprog’ by Hal Whitehead) and R (‘asnipe’ by Damien Farine) to test if both packages gave the same results and if not, why not. Given all the possible ways to do things, not all of the analyses we did ended up in the paper, and I think there’s a lot more that could possibly be tested. For example, if we could get reliable maternity data, perhaps we could test for evidence of within-group maternal inheritance of associations (if you’re reading this and want to see if this analysis is feasible, feel free to email me).
*_One could also look at animal social complexity not as just understood by us biologists but more from the perspective of mathematicians who study ‘complexity science’. Liz Hobson and others at the amazing and lovely Santa Fe Institute wrote a paper on this (preprint here).
** Jerry Wilkinson had data on evening bats in the USA. Wilkinson and Kisi Bohn’s had data on greater-spear nosed bats in Trinidad. Mirjam Knörnschild, Linus Günther, Barbara Caspers, Martina Nagy, and Frieder Mayer provided data on sac-winged bats and proboscis bats in Costa Rica. Gloriana Chaverri had data on disc-winged bats in Costa Rica. Gerald Kerth had data on Bechstein’s bats in Germany. Jorge Ortega had data on Jamaican fruit bats kin Mexico. Krista Patriquin had data on Northern long-eared bats in Canada. Bryan Arnold (Pallid bats) and Dina Dechmann (Lophostoma silvicolum, the bats that live in active termite nests) also contributed data but were not included in the study because the data were too sparse to estimate good networks. Victoria Flores and Rachel Page will soon be adding the frog-eating bat to this comparative dataset.