Background Color Matching in Texas Horned Lizards (Phrynosoma cornutum)

Animals often tend to blend in, rather than stand out, with their colors and patterns matching the overall colors and patterns of their background (Endler, 1990; Diamond and Bond, 2013). The signal of background color matching is strongest in species that are more sedentary (Thayer, 1909; Egan et al., 2016), in species that have high reproductive isolation between contrastingly colored populations (Rosenblum, 2008; Robertson and Rosenblum, 2009; Sobel et al., 2010; Rosenblum and Harmon, 2011), and in areas where the background surroundings are most uniform in color (Hoekstra, 2010). These phenotype-to-environment color adaptations have important implications for conservation and the maintenance of biodiversity, as organisms adapted to match their respective environmental background colors have been shown to decline in number, phenotypic diversity, and genetic diversity when the matching background environment becomes rare or changes into a different hue (Majerus, 2009; Baling et al., 2016).

Historically, quantitative background-color matching studies of wild fauna have been largely limited to those researchers who have access to reflectometers (or spectrophotometers), and to organisms and their substrates that can easily be manipulated (captured, procured, and transported back to laboratory facilities) for color analyses (Benson, 1933; Dice and Blossom, 1937; Hoekstra and Nachman, 2003; Rosenblum, 2006). An alternative approach to using expensive spectrophotometers in controlled lighting and laboratory settings is with digital photography (Stevens et al., 2007; Troscianko and Stevens, 2015; Samia et al., 2015).

One limitation of using ordinary cameras and red, green, and blue (RGB) displays to measure background-color matching in nature is that such devices do not capture the ultraviolet (UV) portion of the spectrum that tetrachromats (e.g., many birds) can perceive (but see Troscianko and Stevens, 2015 for a workaround). However, using everyday RGB devices and methods for measuring crypsis can often serve as an adequate rough index for overall patterns in nature because most background features of terrestrial environments (e.g., soil, leaves, leaf litter, bark, and other dry vegetation) do not reflect UV (Endler, 1993), and species that predominantly display inconspicuous earthy colors (i.e., yellowish to brown and reddish brown) do not reflect UV and typically match the spectrum of their habitat background (Finger and Burkhardt, 1994). Therefore, most species that appear to mimic the coloration of inedible structures in their respective background habitats to avoid predation are not likely to have UV-reflective patches of coloration that are visible to tetrachromats but hidden to trichromats such as humans (Honkavaara et al., 2002). Such overall patterns of cryptically colored species lacking UV-reflective patches to match their backgrounds has been experimentally observed in inconspicuously colored birds (Finger and Burkhardt, 1994) and in cryptic insects that are predated by diurnal, visually oriented predators (Church et al., 1998; Lyytinen, 2001; but also see Lyytinen et al., 2001; Endler and Mappes, 2004). Therefore, among animal species studied thus far, most examples of apparent crypsis in the RGB spectrum retain their camouflage in the UV spectrum.

Among the most cryptically colored of terrestrial vertebrates in North America are the horned lizards of the genus Phrynosoma. Biologically, horned lizards are ecological specialists; of the 96 species of lizards in the American Southwest (as defined by Jones and Lovich, 2009), the nine species of horned lizards are perhaps the group that depends most on crypsis and remaining motionless at the approach of a predator (Bundy and Neess, 1958; Norris and Lowe, 1964; Pianka and Parker, 1975). Although not all members and populations of the genus are in serious decline, Texas Horned Lizards (Phrynosoma cornutum) are a threatened species in the State of Texas and a species of special concern in the State of Oklahoma (Price, 1990; Carpenter et al., 2014). Texas Horned Lizards have disappeared from the eastern part of their range in Texas and Oklahoma and have declined and disappeared from other areas in Texas, and therefore their conservation is urgent (Price, 1990; Donaldson et al., 1994; Henke, 2003; Haney et al., 2022).

Phrynosoma cornutum have highly variable dorsal coloration and anecdotally appear to covary with geographic location and prevailing substrate type (Price, 1990), although the alleged concealing coloration has not been tested. Among the other proposed but untested crypsis adaptations for P. cornutum are that the species is adapted to terrestrial habitats with sun-bleached plant stems, as the discrete color pattern of a ubiquitous white vertebral stripe spanning the dorsum suggests a match to these stems (Fig. 1; Sherbrooke, 2002). Understanding whether Texas Horned Lizards match their substrate is potentially important because several zoos and state wildlife agencies are releasing both wild-caught and captive-raised offspring into areas where Texas Horned Lizards have been extirpated. In one study that tracked 57 translocated individuals for three active seasons, only 2 individuals remained by the end of the study and at least 60% of 55 mortalities were lost to predation (N. Rains, pers. com.; Miller et al., 2020). Such reintroduction efforts are currently being done without any knowledge of the role that geographic color pattern adaptation and crypsis play in survival, and so it is possible that the high predation levels might arise in part from lizards being introduced into areas where they do not closely match the substrate.

In this study, we used a method for measuring animal-background contrast with digital photographs called the color overlapping index (COI; Samia et al., 2015) to test the hypotheses that 1) Texas Horned Lizards' color matches the substrates on which they are found and 2) the vertebral white stripe helps conceal the lizard. Specifically, we asked three questions: 1) Do Texas Horned Lizards from specific populations match their own substrates better than substrates from other populations? 2) Do vertebral stripes of Texas Horned Lizards match nearby plant stems better than other nearby objects (i.e., soil and gravel) in their environment? and 3) Do the vertebral stripes of Texas Horned Lizards from specific populations match their own habitat's plant stems better than the plant stems from other Texas Horned Lizard populations?

We took RAW photographs of 32 individual lizards with color standards in the field from 20 May to 31 August 2018 in seven Texas counties: Karnes (n = 10), Rolling Plains Quail Research Ranch (RPQRR) of Fisher County (n = 11), Dimmit (n = 5), Presidio (n = 3), Irion (n = 1), Jeff Davis (n = 1), and Crockett (n = 2) (Fig. 2, Table S1). Lizards from the San Antonio Zoo (n = 12; originally from the Chaparral Wildlife Management Area in the South Texas Plains ecoregion) were also photographed against white sand for use in the color calibrated vs. uncalibrated comparison (see below). To make photographs standardized, higher quality, and with less noise, we shot all images outdoors with automatic white balance, in situ (precisely where lizards were found) with ISO-200 at aperture f/14 with a shutter speed of 1/400 sec. These are optimal settings chosen for shooting in variable amounts of sunlight, with higher depth of field (for background comparisons), and for freezing the occasional moving object, respectively.

We did not use flash to mitigate any unnatural shadows, modifications to the habitat's natural lighting spectrum, and color metamerism, a phenomenon that renders dissimilar wavelengths of color similar to the observer, or vice versa (Stevens et al., 2007; Samia et al., 2015; Fairchild and Heckaman, 2016; Qiu et al., 2017). We achieved correct color values in the field by placing a 24-patch color standard, the X-Rite ColorChecker Passport (photo version), within ∼10 cm of each individual lizard in the photo frame, and using the ColorChecker Camera Calibration software version 1.0.0 (X-Rite) and the X-Rite ColorChecker plug-in in Adobe Lightroom to create digital negative (DNG) profiles and adjust field photographs to their true color values (Myers, 2009; Sakai et al., 2013; Varghese et al., 2014). Lizards included in the analyses were not handled before photographing to not influence stress-induced color change, which has been documented in this and other lizard taxa (Parker, 1938).

We sampled research-grade in situ photographs of lizards taken in daylight with no apparent flash from citizen scientists via the iNaturalist repository of data. We used 93 photographs from across the species range in Texas, southwestern New Mexico, and Chihuahua, Mexico (Fig. 2, Table S1).

We first extracted color data for images by using the rectangular selector tool in ImageJ to select both an area of each lizard's body and an equiareal portion of the substrate on which it was found (Fig. 3). The substrate portion was picked randomly by using the Grid plug-in for ImageJ with the “random offset” on, with the cells of the grid sized to the same square pixel area as the lizard portion. All grid cells immediately adjacent to the lizard with less than ∼15% nonsoil debris (e.g., vegetation) and in the same lighting as the lizard selection were considered eligible for color analysis. All eligible cells were then assigned a number and picked using a random number generator to mitigate sampling bias. Then we obtained color frequencies for the selected portions of the images using the histogram function of the Color Inspector 3D v. 2.3, a plug-in of ImageJ (Barthel, 2006). We used the default setting of 30 color intervals (i.e., number of color cells in the histogram display function of the Color Inspector 3D plug-in) for the color frequency outputs of all photographs sampled in the study, to keep constant any decrease or increase in percent color-pattern match among all samples. Samia et al. (2015) showed that this setting was adequate for ranking differences in color-matching percentages without being either too sensitive or dismissive in detecting color-pattern similarities, and thus, evoked parsimony. Although departures from the color interval default results in varying outputs for color-matching percentages, the results are not significant if the departures are modest, as demonstrated by Samia et al. (2015).

We then pasted color frequency tables from the lizards and corresponding substrates into Excel spreadsheets, and we quantified the percentage of background-color matching by using the COI (Samia et al., 2015; R Core Team, 2022). The COI is an index that calculates the percent overlap between color classes in two or more images and is based on the Renkonen similarity index used in community ecology (Renkonen, 1938; Wolda, 1981). The COI is defined quantitatively as the sum of the lowest relative frequencies among color classes shared by animal and substrate (Samia et al., 2015). Its formula is written as: COI = ∑min(p_1i, p_2i) × 100, where p_1i and p_2i are the relative frequencies of the ith color class of the animal (p₁) and the substrate (p₂) and COI scores range from 0 to 1. Scores near 0 display high contrast between lizard and substrate and very little color overlap, whereas higher scores indicate more extensive color matching. More precisely, the COI is a proxy for measuring the percent overlap in color pattern because the frequencies of individual color classes (counted in numbers of pixels) is one method for quantifying color pattern. In instances where only two objects were compared for their COI, we used the “pairs” mode of the COI function, and when we compared objects across photographs we used the “fuzzy” mode of the COI function (Samia et al., 2015). When using fuzzy mode, we averaged all pairwise COI values for each photograph so that data were independent for statistical tests. The COI code for R, including the .R files, along with instructions and an example, can be found in the supplementary data folder S2, and should be saved in the same directory before importing to R/RStudio to run COI.

There are several potential biases inherent to digital photographs that could potentially affect calculated COI scores. Most of these potential biases can be mitigated if photographs are taken in RAW mode and appropriate standards are included in the photo (Stevens et al., 2007). Colors can be calibrated, or standardized, across photos using a color standard placed within the photo. Because of the inherent nonlinearity of many camera responses to light intensity across the three wavebands, RGB, it is standard procedure to linearize the RGB values with light intensity before comparing them. Linearization returns one-to-one relationships between the amount of light intensity in nature across the RGB spectrum and the camera (Stevens et al., 2007). Variability in environmental lighting and white-balance settings on the camera can also change color, for instance with improper white balance rendering photos too yellow or too blue. These considerations may not always be possible in the field, however, and they are not available for photographs on citizen science sites like iNaturalist. We therefore compared the COI scores for photographs that were 1) color calibrated and uncalibrated, 2) taken under variable lighting conditions (sunny, shaded, cloudy), 3) linearized vs. unlinearized, and 4) auto white balance vs. neutralized white balance to determine the suitability of using photographs such as those from iNaturalist, for which it is not possible to use these standardization methods. We used linear regression and t-tests when assumptions were met to make these comparisons. Otherwise, Mann-Whitney tests were used to make comparisons.

We calculated the COI of lizard coloration to soil coloration within 36 color-calibrated photographs and compared them with the COI calculated from the same uncalibrated photographs using linear regression. We also compared COIs between lizards and soils across photographs that were color calibrated and uncalibrated using a Mann-Whitney test for 10 lizards from a single population in Karnes City, Texas. In these comparisons, we freehand selected the entire bodies of lizards (instead of the usual smaller rectangular selections) and stored them as XY coordinates (in .roi files) and reimported them for use in both color-calibrated and uncalibrated sets to ask if color calibration significantly changed the number of colors detected on the lizards using a t-test.

We photographed a true-to-scale hand-painted pewter model of P. cornutum (Horny Toad Connection) in the same place (total n = 20 times) but variously under sunny (n = 7), shaded (n = 3), and cloudy (n = 10) conditions, using the same camera and settings as the actual lizard photographs taken in the field. We made the same freehand dorsal selection of the model in each photograph and saved them (as .roi files) in ImageJ, and we calculated a matrix of all pairwise COIs between models in the 20 photographs.

We linearized the seven photographs of the pewter lizard model taken in sunny conditions by creating DNG files of the RAW photographs with Adobe DNG converter and set the base tone curve to “linear” in Adobe DNG profile editor, before saving them as DNG profiles and exporting the linearized photographs from Adobe Lightroom. We then compared the COI results with the nonlinearized versions of the same photographs.

We used the same sample of seven photographs of the model taken in sunny conditions to standardize white balance in postproduction. We neutralized the white balance by selecting both the white balance eyedropper tool in Adobe Lightroom and the neutral white balance patch (i.e., the middle, notched patch) on the landscape row of the X-Rite ColorChecker Passport Creative Enhancement Target to compare COIs of the same selections with the out-of-the-box “auto white balance” setting typical in many digital cameras.

We chose four habitats in Texas that differed in their soil substrate color and for which we had 10 individuals photographed: 1) Alpine (brown soil), 2) Freer (tan soil), 3) Karnes City (gray soil), and 4) RPQRR (red soil) (Fig. 4). We obtained photographs from iNaturalist for Alpine and Freer and used our own photographs for Karnes City and RPQRR. We scored them for their COIs both within their respective habitats and between the other three habitats. We used Mann-Whitney and Kruskal-Wallace tests for these comparisons. To get an estimate for the overall trend in background color matching for P. cornutum, we used 75 pictures from iNaturalist and 25 of our own pictures to calculate the COI between lizard coloration and the soil within their own photographs.

We analyzed a sample of 23 photographs from iNaturalist that contained dead grass stems and P. cornutum from each of three major phylogeographic clusters described using mitochondrial and nuclear microsatellite loci in Williams et al. (2019) (northern, n = 9, western, n = 7, and southern, n = 7) to test 1) whether vertebral stripes of P. cornutum match nearby plant stems better than other nearby objects (i.e., soil and gravel) in their environment and 2) whether the vertebral stripes of P. cornutum from each group match their own group's plant stems better than the plant stems from the other P. cornutum populations. We freehand-selected the vertebral stripe of each lizard, along with two to five nearest (to the lizard) bleached grass stems, and all the soil and gravel substrate in the frame in ImageJ. We then calculated the COI scores of each lizard's vertebral stripe to the grass stems and the soil/gravel substrate. Then we calculated the COI scores of each lizard's vertebral stripe to the grass stems of the other lizards within its own group, and between the other two groups. We used Mann-Whitney tests for these comparisons.

There was no significant difference in COI scores of lizards and the soil in their pictures (n = 36 lizards) using color-calibrated or uncalibrated photos. Linear regression between calibrated and uncalibrated photos had an R² of 0.99 and a slope that did not differ significantly from 1 (y = 1.004x − 0.21, F_1,34 = 2,602.8, P = 1 × 10⁻³³, P = 0.43 for a slope ≠ 1).

We then compared COI values calculated between pictures that were calibrated with those that were uncalibrated for 10 individuals from Karnes City. The average COI difference between calibrated and uncalibrated was low (X = −1.26; range = −1.88–3.90) and median COI values were almost identical between calibrated and uncalibrated comparisons (56.96 in calibrated, 56.52 in uncalibrated; Mann-Whitney: W = 100, n₁ = n₂ = 10, P = 0.73). The average number of colors each lizard had on its dorsal aspect was also not different between calibrated (89.8 ± 7.66 SE) and uncalibrated photos (85.3 ± 6.68 SE) (t = 2.10, df = 18, P = 0.66). These analyses suggest that color calibration of photos does not appreciably change the calculated COI values, so we performed the rest of the analyses on noncalibrated photos.

COI values were significantly higher when the lizard model was photographed under sunny conditions compared with cloudy and shade (median COI sunny = 83.05; range = 78.59–85.55, n = 7; median COI cloudy/shade = 75.16, range = 67.52–80.41, n = 13; Mann-Whitney: W = 115, P = 0.001). Linearized photographs of the model did not differ in their calculated COI values from nonlinearized photographs (nonlinearized median = 83.05, range = 78.59–85.55; linearized median = 82.17, range = 66.89–85.65; Mann-Whitney: W = 59, P = 0.44). Photographs of the model had higher COI scores when white balance was standardized (neutralized) compared with nonstandardized (automatic white balanced) photographs (nonstandardized median COI = 83.05; range = 78.59–85.55; standardized median COI = 84.64, range = 83.11–87.32, Mann-Whitney: W = 36, P = 0.04).

The COI of lizards was higher (median = 65.57, range = 26.66–86.47) with soil from their own picture than the habitat soil average (median = 45.97, range = 20.0–62.86) (Mann-Whitney: W = 1095, P = 4.5 × 10⁻⁷). The overall average within-habitat COI was higher than the average between-habitat COI, indicating that lizards more closely matched soil in their own habitat than in other habitats (Mann-Whitney: W = 744, n₁ = n₂ = 40, P < 0.001) (Fig. 5). The four sites differed significantly in their within-habitat (Kruskal-Wallis: H = 22.36, df = 3, P = 5.5 × 10⁻⁵) and between-habitat COI (Kruskal-Wallis: H = 13.23, df = 3, P = 0.004) (Fig. 5). The within-habitat COI was higher than the between-habitat COI for Alpine, Karnes, and RPQRR (P < 0.0001 in all cases) but not for Freer (P = 0.20) (Fig. 5). Using photographs from Texas, New Mexico, and Chihuahua, the average COI score for a lizard to its own soil substrate was X = 60.92 ± 1.37 SE (n = 100) and ranged from 26.7 to 86.5 with 81% of the comparisons at a COI of 50 and above.

Stripe-to-stem COI scores were higher than stripe-to-soil/gravel COI scores (median stripe-to-stem = 47.98, range = 15.29–76.21, median stripe-to-soil/gravel = 4.80, range 0–21.43, Mann-Whitney: W = 280, P = 1.1 × 10⁻⁸). Similar to the results for soil, individual lizards matched stems in their own picture more closely (median = 47.98, range = 15.29–76.21) than to others within their cluster (median = 26.85, range = 8.0–49.02) (Mann-Whitney: W = 353, P = 5 × 10⁻⁵). The between-cluster stripe-to-stem COI values (median = 25.43, range = 3.15–47.17), however, were comparable in magnitude to the within-cluster stripe-to-stem COI values (median = 26.85, range = 8.0–49.02) (Mann-Whitney: W = 858, P = 0.50). A closer look at the data revealed that the western cluster stripe-to-stem COI scores were significantly higher than both the northern (Mann-Whitney: W = 53, P = 0.02) and the southern lizards (Mann-Whitney: W = 70, P = 0.03) (Fig. 6). Northern and southern lizards were similar (Mann-Whitney: W = 87, P = 0.29) (Fig. 6).

This study tested the hypotheses that 1) Texas Horned Lizards' body color matches the variably colored soil substrates on which the lizards are found throughout their geographic range (Price, 1990) and 2) that the vertebral white stripe helps conceal lizards by mimicking nearby sun-bleached grass stems (Sherbrooke, 2002). The results of this study support these hypotheses, using the RGB color spectrum. Color matching is only one of several traits horned lizards use to blend into their backgrounds, as cryptic coloration is always coupled with one or more behaviors. Horned lizards flatten their body against the substrate, and they have white, lateral fringe scales to avoid casting a shadow (Sherbrooke, 2003). Dorsal patterns bear contrasting black and white and sometimes yellow color patterns that appear to break up lizard outline and shape. Horned lizard antipredator defenses occur in a sequence: 1) evading visual detection via immobility and crypsis, 2) escaping via locomotion, and 3) using behavioral resistance via posturing, puffing their lungs full of air, poking skins of predators with sharp cranial horns, and squirting blood from a sinus behind the ocular cavity (Middendorf and Sherbrooke, 1992; Sherbrooke, 2003, 2008; Young et al., 2004). Although the relative contributions to survival of these three levels of defense remain unknown, many aspects of their anatomy, color pattern, and behavior suggest that the first level, escaping the detection of visual predators, has had a profound, multifaceted influence on their evolution (Norris and Lowe, 1964; Sherbrooke, 2002). Mirkin et al. (2021) demonstrated that gray-colored foam models of P. cornutum were attacked by birds on reddish substrate significantly more often than color-matched reddish models on the same substrate, although the two treatment groups were tested at different times. Future tests should use models painted different colors at the same time to further test the importance of background color matching for P. cornutum.

Although human vision is very good at detecting colors in the RGB spectrum, other taxa such as birds and some snakes that hunt by sight such as Coachwhips (Masticophis flagellum) can also detect UV (Secor and Nagy, 1994; Werler and Dixon, 2000; Macedonia et al., 2009; Hill, 2010). Font and Molina-Borja (2004) summarized a list of 22 known lizards with UV-reflective patches. This list and more recent reports (e.g., Pérez I De Lanuza and Font 2014, Assis et al., 2020, 2021) collectively show that most of these UV-reflective color patches are not on the dorsum or other places on the body that would be detectable to visual predators (Fleishman et al., 1993) but are typically on the throat (e.g., gular patches and dewlaps), chest, or low lateral and ventrolateral flanks (Stoehr and McGraw, 2001). Of the few lizard species with UV-reflective patches on visible (from above) parts of the body, at least one—the brown morph of Anolis conspersus—spends its time on tree bark that remarkably reflects UV light at the same wavelength as the lizard, thus matching its background (Macedonia, 2001). Given that lizards are a common diet item for predatory birds and snakes it seems unlikely that P. cornutum would have patches conspicuous to UV-sensitive predators, but this remains to be tested.

There are two examples of apparent plant-stem litter and plant-stem shadow mimicry in the genus. One species, the Flat-Tailed Horned Lizard (P. mcallii), lives in the Sonoran Desert and bears a thin, black dorsal stripe that is hypothesized to resemble plant-stem shadows on the sand (Sherbrooke, 2002). In contrast, P. cornutum has a white vertebral stripe and is found in the major grasslands of the southwestern deserts and the south-central Great Plains (Sherbrooke, 2002). Unlike the dunes habitat of P. mcallii, P. cornutum microhabitats tend to accumulate much litter in the form of windblown grass, forb, and shrub stems, usually of highly reflective tan or grayish-white color, in varying stages of decomposition and bleaching by solar radiation (Sherbrooke, 2002). The white stripe of P. cornutum more closely matched grass-stem litter than other objects in the environment and had a moderately high stripe-to-stem COI average, suggesting the importance of this feature in the environment. Western lizards also color match to nearby blanched grass-stem litter significantly better than the northern and southern lizards match to their respective grass stems. With lower amounts of vegetation in the deserts of the western range, it is possible that nature selects for higher levels of crypsis when hiding underneath sparser shrubs and bunch grasses than in the two grassier eastern (i.e., northern and southern) clusters. However, this is conjecture and warrants further testing.

Lizards had significantly higher COI scores when compared with soils in their immediate surroundings, i.e., in their own photo, than they did when compared with soils where other lizards were found in their population. A similar pattern was observed for stripe-to-stem COIs. It is possible that there is always inherently a higher COI between two or more objects within a single photograph than for objects in different photographs (e.g., because of nearby objects reflecting light off each other). Alternatively, it is possible that horned lizards choose the most optimal color-matching microhabitats for enhanced crypsis, as in some squamate reptiles with generally less visual acuity than lizards that show this behavior (Kravchuk and Watson, 2020).

This study demonstrated that lighting conditions (i.e., sunny) and standardizing white balance across photographs significantly increased color-matching (COI) scores, whereas color calibration and linearization did not significantly change COI scores. One review of the COI method reported that using partial (e.g., rectangular) selections yields lower COIs than whole-body polygonal selections (Cordeiro De Moura et al., 2019). Although we attempted to only use photographs taken in sunny conditions, the COI values reported in this study are probably conservative since we used rectangular selections for most analyses and were not able to standardize the white balance for photographs from iNaturalist. Nevertheless, the implications of being able to use uncalibrated photographs of organisms potentially opens an enormous resource for organismal color researchers in citizen science projects like iNaturalist and other repositories where scientific naturalists upload field data and photographs. We therefore believe that the method used here is useful for generating additional hypotheses and tests that could later be tested in more technical detail using calibrated photographs, simulations of various visual systems, differences in hue and saturation, and the degree of achromatic and chromatic contrast. The COI method used herein to measure color-pattern matching should convince researchers and citizen scientists of its verifiability and reproducibility compared with other colorimetric methods and of the usefulness of capturing photographic data as raw material for investigating color-related questions in ecology, evolution, and conservation.

Any conservation plan concerning Texas Horned Lizards might make it a priority to include crypsis (with appropriate soils and grasses) in its strategy (Jacobs, 1991). An organization involved in reintroductions might aim for making sure their lizards' colors have a COI of at least 45% to the local soils since this was the average between-photograph COI within the same habitat, or at least choose sites with the highest possible color match, all else being equal. The importance of color matching for reintroductions should be further tested by measuring the survival of various colors of live P. cornutum released at sites.

This work was supported by a grant from the Horned Lizard Conservation Society and grants from the Andrews Institute of Mathematics & Science Education and Research and Creative Activities Fund at Texas Christian University (TCU). We thank H. Breckenridge and M. and T. Hunt for lodging and hospitality at the 505 in Kenedy and W. Phelps for help and support of our studies of Texas Horned Lizards in Karnes County. We also thank R. McClure, S. Mirkin, and M. Lee (San Antonio Zoo) for photographing lizards for the color calibration tests; B. Bourassa and A. Leung for assistance with the map; A. Johnson for assistance with R; and B. Lawrence, R. Hartdegen (Dallas Zoo), and D. Rollins for assistance at the RPQRR. We thank D. Biffi for constructing the figures. This work was conducted under the Institutional Animal Care and Use Committee approval from TCU (18-06) and a scientific research permit from Texas Parks and Wildlife (SPR-0613-073).

Supplementary data associated with this article can be found online at http://dx.doi.org/10.1670/22-008.s1.