Drosophila escape behaviours in spatially complex environments

Abstract

Escape behaviours are severely time-constrained due to fast predatory strikes. They thus provide a useful model to understand how different behaviour strategies are computed over a limited number of synapses. Broadly, one can propose two general strategies of escape: to escape away from the threat or to escape toward safety. While not mutually exclusive, these two strategies are fundamentally distinct in the computations they require. Escaping toward a safe place has better long-term value but might take up valuable time since it requires complex computations such as evaluating shelter locations and available escape routes. Mice escape directly towards a learned shelter location by encoding a shelter-directed vector in their brain. The proposed model behind this shelter-direction vector is closely related to models of heading direction proposed for various circuits in other animal species, including fruit flies. This thesis investigates whether fruit flies show shelter directed escape behaviours. I tested multiple shelter designs and found that flies bias their escape trajectory towards shelters, but this bias is weak and, in the short term, does not lead to optimal safety-directed trajectories