Anyone who's ever worked on animal behaviour knows that animals,irrespective of whether they are vertebrates or invertebrates, do whatever they please, whenever they please, often completely ignoring the experiment that you've spent weeks painstakingly designing. Much of this spontaneous behaviour generated by animals is at best confusing because it doesn't seem to be linked to any external stimulus – under identical conditions animals often respond in entirely different ways. This variability is often attributed to `noise' within the nervous system: for example, variation in the pattern of neural and muscle activity when carrying out a specific action. The implication is that in the absence of such noise, the animal's behaviour would lack this variability. A recent paper by Alexander Maye and his colleagues from Germany and the USA set out to determine whether noise in the nervous system really is sufficient to account entirely for the spontaneity of animal behaviour.

To assess this spontaneity they recorded the turning behaviour of fruit flies, Drosophila melanogaster, flying whilst attached to a torque monitor in a homogenous arena free of any external cues. When the flies attempted to turn either left or right they generated rapid changes, or spikes, in the yaw torque. By monitoring these spikes and the time between them, the team determined the spontaneous changes in direction the flies made. They compared the behaviour of the flies in the homogenous arena to flies in an arena with a single black stripe, upon which they could fixate, or in an arena with a uniformly textured surface, which allowed the flies to fly straight but not fixate.

Maye and his colleagues proposed several hypothetical mechanisms that could account for the observed behavioural spontaneity. Each hypothetical mechanism makes specific predictions about the temporal patterns of yaw spikes that the flies produce. They began by testing whether noise is sufficient to produce the patterns of inter-spike intervals (ISIs) that they observed under all three conditions. If noise is sufficient to generate the spontaneous behaviour of the flies, then the ISIs should not differ from a random process. However,the ISIs generated by the flies under all three conditions were significantly different from those of a random process, leading the team to conclude that noise could not be sufficient to explain the flies' spontaneous behaviour.

Next, Maye and colleagues tested whether the flies' behaviour could be generated by a process in which each subsequent step is only dependent on their current state – a `memoryless' process – which had been previously used as a model for spontaneous behaviour. This possibility was also eliminated by comparing the sequences of ISIs from the flies under the three different conditions with randomly shuffled sequences of ISIs. Another model that assessed the fluctuations within the ISI time series suggested that there were strong non-linear relationships contained within the flies'behaviour that could partially account for the behavioural patterns. In non-linear systems, the responses to one set of inputs can't be used to predict those to a different set of inputs. Finally, using computer-modelling techniques usually applied to weather forecasting, the team showed that non-linearity was not sufficient to fully capture the flies' behaviour: the non-linearity also needed to operate under specific conditions.

This study convincingly shows that non-linear processes and not random noise accounts for the spontaneity of the flies' behaviour. Yet, as the authors themselves point out, the source of this non-linearity within a fly's brain remains unclear. With such a clear behavioural paradigm, however, it should be possible to tease apart the components of the neural circuitry involved, and that's just what Maye and his colleagues suggest they will do next.

Maye, A., Hsieh, C.-h., Sugihara, G. and Brembs, B.(
). Order in spontaneous behavior.