The origins of preferences that drive the evolution of arbitrary sexual signals have been hotly debated for over 150 years. An emerging but little-tested theory, efficient coding theory, proposes that male visual courtship displays are adapted to pre-existing processing biases shaped by the statistical properties of the natural environment. Natural scenes show strong spatial correlations with average amplitudes of spatial frequencies falling with an average spectral slope of -1, and humans have been shown to prefer random amplitude spectrum images that possess similar slopes. It has been proposed that other animals may also prefer the statistics of their natural environment and that this preference drives the evolution of sexual signaling displays. Here, we measure the spectral slope of the male display pattern of the Australian peacock jumping spider Maratus spicatus and test for a general preference towards that slope. We present spiders (male, female and juvenile) with random images of the male slope of -1.7 compared to a) a shallower slope of -1.0 and b) a steeper slope of -2.3. Spiders spent more time oriented towards the shallower slope than towards the male slope and spent the same amount of time oriented towards the male slope and the steeper slope. Our results indicate that spiders, like humans, prefer the average natural slope of -1, suggesting that this is likely the slope typically found in their natural habitat. Rather than exploiting a potential processing bias, it seems that males have evolved slopes that contrast with the visual background to enhance conspicuousness.

Mulitspectral imaging (salticidVirtualFilters_530nm.mat, imageSets.mat)

Multispectral imaging was done using a multisprectral camera containing three bird-based optical filters (U, M and L) previously described in Tedore and Nilsson (2019) and two additional filters previously described in Glenszczyk et al. (2021). By taking a weighted sum of these pre-existing camera filters, we calculated computational filters to generate new spectral sensitivities matching the spectral sensitivity of the salticid green receptor, which typically peaks around 530 nm (De Voe 1975; Yamashita and Tateda 1976; Blest et al. 1981; Zurek et al. 2015; Glenszczyk et al. 2021). The computational filter technique is further described in Tedore and Nilsson (2021) and Glenszczyk et al. (2021).

Display distance (FirstDanceDist.txt)

We analyzed videos of mating experiments for the distance at which males first lifted their abdomen towards the female using ImageJ on screenshots of the video. 

Data for analysing preference for random noise image stimuli with spectral slopes of -1.0, 1-.7 and -2.4 (PreferenceTestInfos.txt, BORIS Data)

In each trial, a spider was presented with a pair of random images, one of them always representing the spectral slope of the courtship display of M. spicatus males (-1.7) and the other one representing either a shallower slope of -1.0 or a steeper slope of -2.3. Every spider was tested with both combinations at intervals of 7 days (±1 day), and the amount of time the spiders looked at each image was recorded. Video analysis was performed with BORIS software (Friard and Gamba 2016). We scored the viewing direction (left image, right image) by following the direction of a bright stripe naturally occurring on top of the prosoma.

Reflection data of grayscale after printing (Reflection Grayscale)

We measured the reflectance of printed grays having RGB values of 0 to 255 in increments of 5. For measurement, printed grays were illuminated by a xenon light source (PX-2) channeled through a solarization-resistant fiber optic cable oriented 45° relative to the surface normal. Reflectance was measured relative to a PTFE diffuse white standard (WS-1) using a solarization-resistant fiber optic cable oriented normal to the surface being measured. This fiber channeled reflected light into a spectrometer (QE Pro; all sourced from Ocean Optics, Ostfildern, Germany). 

R Core Team. 2019. R: A Language and Environment for Statistical Computing. https://www.R-project.org/.

Matlab. 2020. version 9.9.0 (R2020b). Natick, Massachusetts: The MathWorks Inc.