Colour pattern edge contrast statistics can predict detection speed and success at ecologically relevant viewing distances in triggerfish (Rhinecanthus aculeatus)

Abstract

Edge detection is important for object detection and recognition. However, we do not know whether edge statistics predict the detection of prey by non-human predators. Understanding the link between image statistics and animal behaviour is crucial and of increasing importance given the growing availability of image analyses and their application across non-human visual systems. Here, we investigated whether Boundary Strength Analysis (BSA), Local Edge Intensity Analysis (LEIA) and the Gabor Ratio (GabRat) could predict the speed and success with which triggerfish (Rhinecanthus aculeatus) detected patterned circular stimuli against a noisy visual background, in both chromatic and achromatic presentations. We found that individual pattern statistics could only explain up to 2% of the variation in detection time, whereas PCA regression analysis considering all edge statistics simultaneously explained up to 6% of the variation. This suggests that other factors explained more behavioural variation than individual edge statistics. Furthermore, different statistics significantly correlated with detection speed depending on treatment, viewing distance, and changes in fish response over time, while highlighting the importance of considering spatial acuity and relevant viewing distances in the study of visual signals. Our results demonstrate the need for broad and unbiased approaches for identifying task-specific predictive relationships between pattern statistics and animal behaviour using image statistics capturing different aspects of colour patterns. We require robust statistical approaches to investigate correlations between ecological effect and the ever-increasing dimensionality and size of datasets in the field of visual ecology, rather than pre-emptively narrowing down the choice of image statistics unless warranted by specific hypotheses. Summary statementCorrelations between edge detecting colour pattern statistics and animal behaviour are complex. Specifically, correlations are unlikely to be explained by single image statistics and depend upon observer distance.