Today I got the news that my new paper has been accepted for publication, it should be available soon. This paper started as a hunch at the end of my PhD, where I noticed a strange pattern in some data on a fellow lab members poster. I did some digging through my old data and found the same pattern, and it turned out to be caused by an undiscovered, fundamentally important spiking feature of a neuron that has been studied in detail for more than 25 years.
Target detection in insects is a small field, and there are still relatively few identified target-detecting neurons in the literature. Perhaps the most well studied target detecting neuron in any insect is the dragonfly Centrifugal Small Target Motion Detector (CSTMD1), which David O'Carroll first recorded and dye filled in the early 1990s. As with most electrophysiological studies, almost every study of target-detecting neurons in insects quantifies responses by computing spike rates - where spikes are counted over a window of time and the mean spike frequency over that time is calculated. However many neurons do not communicate via spike rates, and some controversial neuroscientists even argue that no neurons communicate via spike rates.
When presented with an appropriate stimulus, the spike rate of a neuron will change. This means that if our goal is to determine how certain stimulus parameters effect a neurons activity, comparing spike rates across different conditions makes a lot of sense. However we often overlook the fact that there can be a huge difference between neuronal activity and neuronal coding - quantifying neuronal activity is not synonymous with quantifying neuronal coding. Spike rates alone are only directly informative about neuronal coding if we assume that the neuron studied is using a rate code, and in my opinion, that assumption is often made with absolutely zero justification.
In my paper we investigated the spike trains of dragonfly visual neurons in detail, focussing on CSTMD1. We found that unlike other neurons studied, CSTMD1 produces unique spike trains, consisting of periodic spike bursts rather than a steady pattern of spikes. This finding implies that CSTMD1 might use a burst code instead of a rate code. Burst codes are very rarely mentioned in insect vision, but they have several significant advantages over rate codes, and some of these would be extremely useful in a highly dynamic scenario such as target pursuit. We analyse responses of CSTMD1 through both spike rate-based and spike burst-based coding schemes in a simplified target detection task, and found that bursting results in a much faster and more accurate representation of target movement.