Figure 1. (a) Moths have persistent positional offsets relative to flowers. The offset differs between individuals and between different bouts of feeding. (b) A slope less than 1 indicates a mean-reverting (memory) process. (c) Feedback control diagram for flower-tracking with an internal positional reference point. (d) Screen-feeding, with a jump in flower position to disambiguate positional or velocity-integration memory. Pilot data shows that moths respond to moving projected flowers while screen-feeding.
We have performed behavioral measurements on hawkmoths to verify the existence of “position-locking”. This is when hawkmoths maintain a fixed positional offset relative to the center of a flower while hover-feeding. Having analyzed the timeseries of these offsets, these timeseries indicate a mean-reverting character. Additionally, moths feeding multiple times have a new positional offset every time they feed. These observations indicate the presence of a memory that re-forms at the start of a feeding bout and persists only during uninterrupted feeding.
The next step is to identify whether this memory is a positional snapshot, or is the outcome of a velocity integrator which is made zero at the start of feeding. In order to test these two alternatives, we are using a screen with a projected flower, from which hawkmoths can feed. Using this screen, the projected flower position is jumped a discrete amount. If moths respond with a complete compensation, the memory is a positional snapshot. Alternatively, if moths have only a transient response to a jump without a significant change in position, this would indicate a velocity-integrator based memory. We have performed pilot experiments for this study, and we are currently in the process of collecting a full dataset. This is important because proportional-derivative control on position and proportional-integral control on velocity can appear very similar in response to continuous inputs, but they have very different biological sources.