Flying locusts perform a characteristic gliding dive in response to predator-sized stimuli looming from one side. These visual looming stimuli trigger trains of spikes in the descending contralateral movement detector (DCMD) neuron that increase in frequency as the stimulus gets nearer. Here we provide evidence that high-frequency (>150 Hz) DCMD spikes are involved in triggering the glide: the DCMD is the only excitatory input to a key gliding motor neuron during a loom; DCMD-mediated EPSPs only summate significantly in this motor neuron when they occur at >150 Hz; when a looming stimulus ceases approach prematurely, high-frequency DCMD spikes are removed from its response and the occurrence of gliding is reduced; and an axon important for glide triggering descends in the nerve cord contralateral to the eye detecting a looming stimulus, as the DCMD does. DCMD recordings from tethered flying locusts showed that glides follow high-frequency spikes in a DCMD, but analyses could not identify a feature of the DCMD response alone that was reliably associated with glides in all trials. This was because, for a glide to be triggered, the high-frequency spikes must be timed appropriately within the wingbeat cycle to coincide with wing elevation. We interpret this as flight-gating of the DCMD response resulting from rhythmic modulation of the flight motor neuron's membrane potential during flight. This means that the locust's escape behavior can vary in response to the same looming stimulus, meaning that a predator cannot exploit predictability in the locust's collision avoidance behavior.

Despite recent papers on problems associated with full-model and stepwise regression, their use is still common throughout ecological and environmental disciplines. Alternative approaches, including generating multiple models and comparing them post-hoc using techniques such as Akaike's Information Criterion (AIC), are becoming more popular. However, these are problematic when there are numerous independent variables and interpretation is often difficult when competing models contain many different variables and combinations of variables. Here, we detail a new approach, REVS (Regression with Empirical Variable Selection), which uses all-subsets regression to quantify empirical support for every independent variable. A series of models is created; the first containing the variable with most empirical support, the second containing the first variable and the next most-supported, and so on. The comparatively small number of resultant models (n = the number of predictor variables) means that post-hoc comparison is comparatively quick and easy. When tested on a real dataset – habitat and offspring quality in the great tit (Parus major) – the optimal REVS model explained more variance (higher R²), was more parsimonious (lower AIC), and had greater significance (lower P values), than full, stepwise or all-subsets models; it also had higher predictive accuracy based on split-sample validation. Testing REVS on ten further datasets suggested that this is typical, with R² values being higher than full or stepwise models (mean improvement = 31% and 7%, respectively). Results are ecologically intuitive as even when there are several competing models, they share a set of “core” variables and differ only in presence/absence of one or two additional variables. We conclude that REVS is useful for analysing complex datasets, including those in ecology and environmental disciplines.