Warning signals are often characterized by highly contrasting, distinctive and memorable colors. Both chromatic (hue) and achromatic (brightness) contrast contribute to signal efficacy, making longwave colored signals (red and yellow) that generate both chromatic and achromatic contrast common. Shortwave colors (blue and ultraviolet) do not contribute to luminance perception, yet are also common in warning signals. The presence of UV aposematic signals is paradoxical as UV perception is not universal, and evidence for its utility is at best mixed. We used visual modeling to quantify how UV affects signal contrast in aposematic butterflies and frogs. We found that UV only appreciably affected visual contrast in the butterflies. As the butterflies, but not the frogs, have UV-sensitive vision these results support the notion that UV reflectance is associated with intraspecific communication, but appears to be non-functional in frogs. Consequently, we should be careful when assigning a selection-based benefit from UV reflectance.

To capture reflectance values across an ecologically relevant spectrum, we took calibrated photographs in both human visible (VIS = ~400-700 nm) and ultraviolet wavelengths (UV = ~300-400 nm), following methods outlined in Yeager and Barnett (2020). In short, we took all digital images using a tripod mounted, UV-sensitive, full spectrum quartz converted Canon EOS 7D that was combined with a metal body NIKKOR EL 80 mm lens. For human-visible spectra we fitted the lens with a Baader UV-IR blocking filter (allowing transmission of 420-680 nm), and for the UV photographs we fitted a Baader UV pass filter (allowing transmission of 320-380 nm). We photographed each subject in both human visible and UV wavelengths, under natural downwelling illumination that was representative of the covered canopy forests where both butterflies and frogs occur. All images were saved in RAW format and included a 10% and a 77% reflectance standard that allowed for color calibration and scaling.

We used the MICA toolbox in ImageJ v1.52k to linearize, align, and combine our paired VIS and UV photographs into a series of multispectral images (Schneider et al. 2012; Troscianko and Stevens 2015). We used the 10% and 77% reflectance standards to linearize the images, and each of the photo pairs were aligned manually. We then manually selected regions of interest (ROIs), from each multispectral image, by selecting up to six of the strongest UV reflecting regions (UV+), and up to six similarly sized and shaped adjacent regions that did not reflect UV (UV-).