Visual modeling
To investigate how the presence of UV reflectance affects visual
contrast between adjacent UV+ and UV- regions we compared the responses
of a UV-sensitive visual model to those of a VIS-sensitive visual model
(Troscianko and Stevens 2015; Yeager and Barnett 2020). These models
were both generated using the UV sensitive, tetrachromatic, vision of
the Eurasian blue tit (Cyanistes caeruleus , Paridae), which has
single cone λmax of 573 nm (LWS), 508 nm (MWS), 413 nm
(SWS), 372 nm (UVS), and double cones with λmax of 565
nm (D) (Hart et al. 2000). The UV-sensitive model included the LWS, MWS,
SWS, and UVS cones spanning 300-700 nm, whereas the VIS-sensitive model
used the LWS, MWS, and SWS cones (excluding the UVS cone), to cover
400-700 nm. We also included the response of the D cone in both visual
models.
We converted each multispectral image into relative cone capture rates
using the MICA toolbox in ImageJ v1.52k (Schneider et al. 2012;
Troscianko and Stevens 2015). Visual contrast was calculated as ‘just
noticeable differences’ (JNDs) using the receptor-noise-limited model
(Vorobyev and Osorio 1998). A JND of 1 represents the theoretical visual
discrimination threshold below which two colors cannot be distinguished
(Vorobyev and Osorio 1998). Conversely, JNDs >3 are
increasingly more easily discernable (Vorobyev and Osorio 1998). We
calculated chromatic (hue) contrast from the responses of the single
cones and calculated achromatic (luminance) contrast from the response
of the double cone (Hart et al. 2000; Vorobyev and Osorio 1998). In both
cases, we used Weber fractions of 0.05 (Hart et al. 2000; Troscianko and
Stevens 2015; Vorobyev and Osorio 1998). We hypothesized that if UV
reflectance is an important component of the signal, chromatic contrast
would be perceivably higher in the UV-sensitive model than is the
VIS-sensitive model (Yeager and Barnett 2020).