Neural coding in drosophila: How do small population codes underpin visually guided behaviour?

All organisms wishing to survive and reproduce must be able to respond adaptively to a complex, changing world. As computational power is constrained by the energy cost, mechanisms that are parsimonious yet robust are favoured. An example of this can be seen in the fruitfly, where a group of only 42 visually responsive neurons are known to be critically involved in visually guided behaviours. These cells have recently had their visual receptive fields mapped by Seelig and Jayaraman (2013, Nature, 503, 262–266). We show, with simulations of these small populations of neurons, how they can account for seemingly complex visual behaviour, such as bar fixation (Osorio, 1990, J Exp Biol, 149, 281–292) and pattern discrimination (Ernst & Heisenberg, 1999, Vis Res, 1999, 39, 3920–3933). We further investigate the information encoded in these populations, showing that the information needed to distinguish patterns differing in parameters such as size orientation and elevation is implicitly encoded in the population activity of these neurons. We therefore suggest that higher order modules that specifically extract such parameters are unnecessary. We discuss the implications in terms of future behavioural work and perceptual coding more generally.