Penacchio, O., Lovell, P. G., Cuthill, I. C., Ruxton, G. D., & Harris, J. M. (2015). Three-Dimensional Camouflage: Exploiting Photons to Conceal Form. American Naturalist, 186(4), 553-563. https://doi.org/10.1086/682570 Publisher's PDF, also known as Version of record Link to published version (if available): 10.1086/682570 Link to publication record in Explore Bristol Research PDF-document This is the final published version of the article (version of record). It first appeared online via University of Chicago Press at http://dx.doi.org/10.1086/682570. University of Bristol - Explore Bristol Research General rights This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/ vol. 186, no. 4 the american naturalist october 2015 Three-Dimensional Camouflage: Exploiting Photons to Conceal Form Olivier Penacchio,1,*P.GeorgeLovell,1,2 Innes C. Cuthill,3 Graeme D. Ruxton,4 and Julie M. Harris1 1. School of Psychology and Neuroscience, University of St. Andrews, South Street, St. Andrews, Fife KY16 9JP, United Kingdom; 2. Division of Psychology, Social and Health Sciences, Abertay University, Dundee DD1 1HG, United Kingdom; 3. School of Biological Sciences, University of Bristol, Life Sciences Building, 24 Tyndall Avenue, Bristol BS8 1TQ, United Kingdom; 4. School of Biology, University of St. Andrews, Dyers Brae, St. Andrews, Fife KY16 9TH, United Kingdom Submitted June 28, 2014; Accepted April 1, 2015; Electronically published August 20, 2015 Online enhancement: appendixes. Dryad data: http://dx.doi.org/10.5061/dryad.pt532. abstract: (reviewed by Stevens and Merilaita 2011). In general, how- Many animals have a gradation of body color, termed fl “countershading,” where the areas that are typically exposed to more ever, camou age must conceal three-dimensional (3D) ob- light are darker. One hypothesis is that this patterning enhances vi- jects within a 3D world. sual camouflage by making the retinal image of the animal match The basic camouflage strategies of background match- that of the background, a fundamentally two-dimensional theory. ing and disruptive coloration interfere with object detec- More controversially, countershading may also obliterate cues to tion through contrast and outline coherence (Stevens and three-dimensional (3D) shape delivered by shading. Despite rely- Merilaita 2011). However, 3D objects provide many addi- ing on distinct cognitive mechanisms, these two potential functions hitherto have been amalgamated in the literature. It has previously tional cues that aid in their detection and recognition via not been possible to validate either hypothesis empirically, because shape. Some species use binocular vision for this, but its util- there has been no general theory of optimal countershading that ity falls off with distance (e.g., Harris 2004); for humans, it is allows quantitative predictions to be made about the many environ- useful only up to a distance of ca. 6 m (Cutting and Vishton mental parameters involved. Here we unpack the logical distinction 1995). For animals with smaller eye separation but similar between using countershading for background matching and using it resolution, the distance will be even smaller. There are, how- to obliterate 3D shape. We use computational modeling to deter- ever, many monocular cues to shape and depth, the strongest mine the optimal coloration for the camouflage of 3D shape. Our coming from surface shading (Gibson 1979; Lovell et al. model of 3D concealment is derived from the physics of light and fl informed by perceptual psychology: we simulate a 3D world that 2012). For example, even though the re ectance of the cyl- incorporates naturalistic lighting environments. The model allows inder in figure 1a is uniform, because of the directionality us to predict countershading coloration for terrestrial environments, of the lighting the upper part is noticeably brighter than the for any body shape and a wide range of ecologically relevant param- lower (referred to as “self-shadow”). Thus, matching the color eters. The approach can be generalized to any light distribution, in- of the background does not effectively conceal the object. cluding those underwater. However, if the animal’s pattern is the inverse of the self- Keywords: countershading, background matching, obliterative shad- shadow, then the two cancel, resulting in constant luminance ing, camouflage, shape-from-shading. and perfect self-shadow concealment. This is the basis for the long-influential, but controversial, theory that countershad- ing functions to obliterate shape-from-shading cues, render- Introduction ing a prey animal perceptually flat to a viewer (Poulton 1890; Visual camouflage is the use of color and/or pattern to con- Thayer 1896; Cott 1940). fi ceal an object, rendering it more difficult to detect or rec- There are also other hypotheses speci c to concealment ognize. Most of the theory and research on camouflage con- of 3D objects, but these remain largely untested. This article cerns concealment of objects in a two-dimensional (2D) plane anchors theories of 3D concealment in what can be pre- dicted from optical physics and human perceptual psychol- * Corresponding author; e-mail: [email protected]. ogy. We present a computational modeling framework that Am. Nat. 2015. Vol. 186, pp. 553–563. q 2015 by The University of Chicago. allows us to test logically whether theories are mutually ex- 0003-0147/2015/18604-55587$15.00. All rights reserved. clusive. This is based on a simulated 3D world that allows DOI: 10.1086/682570 for naturalistic lighting environments and delivers realistic This content downloaded from 137.222.120.101 on May 10, 2016 08:40:32 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 554 The American Naturalist abc Figure 1: a, A uniform cylinder exhibits shading when lit from above. b, When the cylinder’sreflectance is the inverse of that shading, the two cancel, but the cylinder is still visible. c,Thereflectance may additionally be chosen so that the radiance from the body also matches that of the background, resulting in both perfect obliterative shading and background matching. effects that take account of the complex interplay between viewed from different angles. Previous studies have often 3D object, light direction, and environment. Using this, not clearly delineated between these potential mechanisms, we make predictions about what data we need to test hy- and, to our knowledge, no previous work has made specific potheses around countershading as effective camouflage. predictions that could distinguish between the hypotheses. Recovering Shape and Depth Information Possible Mechanisms by Which Countershading from 2D Retinal Images Can Act as Camouflage Shape-from-shading is inherently ambiguous because the Matching Multiple Backgrounds. Countershading might oc- pattern of stimulation at the retina is the product of shape, cur as a “side effect” of matching backgrounds that are dif- reflectance, and lighting, all of which can vary, and different ferent according to the angle of viewing (Wallace 1889). For combinations of these can lead to the same image on the example, if the background when viewed from above is dark retina (e.g., Curran and Johnston 1996). Nevertheless, hu- (e.g., ground) and that viewed from below is light (e.g., sky), mans are highly sensitive to it, although our perceptions then this predicts a dark dorsum and a light ventrum. On can be biased by variations in the light direction (e.g., Nefs this account, countershading is not a mechanism to defeat et al. 2005). There is some evidence for the use of shading cues from 3D form but conforms to the general principles cues in other animals (e.g., pigeons; Cook et al. 2012). Any of background matching. Countershading would then be animal using shape-from-shading as a cue to shape should common simply because there are many examples where be fooled by countershading, acting to reduce the shape cue backgrounds are sky and ground and so have these proper- (Tankus and Yeshurun 2001). The inverse also follows: if ties of lighter-background-from-below/darker-background- countershading disguises 3D form, then the viewer is likely from-above, with a consistent body side being viewed against to have developed perceptual mechanisms that derive shape each (e.g., pelagic fish, seabirds). from shading. However, there is a logical counterargument to this pro- The way in which texture and pattern distort in the 2D posal. No matter how light the animal, it cannot be light retinal projection of a 3D object is also a useful source of in- enough to match the bright sky. The radiance from the sky formation on depth and shape (Gibson 1979; Knill 2001). is orders of magnitude higher than the radiance reflected Camouflage for countershading and camouflage for texture from the underside of an animal, because, even if colored matching need not be mutually exclusive and can operate in white to reflect light maximally, an animal’s underside is illu- concert (e.g., the countershaded and textured pattern of a minated only by the (far weaker) indirect light that has it- leopard pelt). We demonstrate that countershading is effec- self been scattered and reflected by the ground. Only in cer- tive from most viewing directions but that texture matching tain marine animals can a match to such downwelling light