Updated: Apr 3, 2019
By analysing fluctuations in experimental measurements, we find that schooling (when fish move together in a coherent direction) is actually a noise-induced effect. Read-on to find out what this means, both for fish and for the field behavioural inference in general...
Can we infer the mechanism(s) by which fish school by simply watching them in the lab? (Or, more accurately, filming them and then analysing the resulting movies). This is the business of so-called behavioural inference, where a frequent challenge is how to deal with noise in the data. In this context, we asked whether, rather than being simply an irritant and obscuring an underlying 'signal', such noise can actually encode important information? The answer, as we show in our latest preprint, is yes.
Looking at collective alignment in groups of the fish Etroplus Suratensis, we found that schooling bears all the hallmarks of a noise-induced effect. Moreover, the fluctuations that we find when measuring group alignment (or average direction of motion) can be traced back to the residual, or left-over, noise that hasn't been 'averaged out'. This means that the smaller the group size, the larger the noise, but the more coherently they school!
Such behaviour is understood in the theoretical literature, but empirical evidence of this kind is very rare. As a result, this places a tight constraint on the types of underlying interactions that might take place between individual fish, but are otherwise impossible (read: very hard) to measure. In particular, we use computer simulations to demonstrate that fish likely interact according to a disarmingly simple pairwise protocol: with specified rates, a given fish either chooses another fish and copies their direction of motion, or simply changes to move in a different at random.
Notably, these results are at odds with the celebrated Vicsek model of 'flocking', which demonstrates a density-dependent phase transition, and is based on fish performing an imperfect rolling average of their neighbours' directions.
So, what are the take home messages? Well, fish don't appear to be able to perform local averages and instead adopt a simpler set of behavioural 'rules', achieving coherent group motion via a novel noise-induced mechanism. Also, in the more general context of behavioural inference, rather than trying to separate the 'signal from the noise', to quote Nate Silvers' eponymous book, the signal is the noise.