|
nif:isString
|
-
We conducted our experiment at La Selva Biological Station (LSBS) from May 17–29, 2008. LSBS encompasses 1500 hectares of primary and secondary lowland tropical wet forest (sensu [17]) on the Caribbean slope of Costa Rica. Using a mold formed from a preserved specimen of D. pumilio, we crafted frog replicas from precolored, non-toxic Sculpey® brand polymer clay (Polyform Products Company, IL, USA). Modeling clay is often used in predation experiments because predators that attack a replica leave identifiable impressions in the clay (e.g., [2], [11], [13]).
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0048497.t001 Number of replicas deployed, recovered, and attacked. We manufactured 200 frog replicas that resembled the local D. pumilio form with a red body and blue limbs (i.e., an aposeme assumed to be familiar to LSBS predators) and 200 frog replicas that resembled a form of D. pumilio novel to LSBS but found on Isla Colon, Bocas Del Toro, Panama, with a yellow body, orange limbs and black spots [18] (Fig. 1). This aposeme is unlike any other frog at LSBS [19] and is assumed to be unfamiliar to predators at LSBS. In addition, we made 200 brown replicas colored to resemble one of several small Eleutherodactylus spp. leaf-litter frogs found at LSBS (i.e., a cryptic form familiar to predators at LSBS) and 200 black replicas (i.e., a cryptic form unfamiliar to predators at LSBS; [19]). We used a black permanent marker to place black eyespots on the replicas and to draw the black dorsal spots on the Isla Colon replicas. Hereafter, for brevity, we refer to each replica color form by its main body color (red, yellow, brown, and black, respectively). Saporito et al. [15] found that Sculpey® clay exhibits low ultraviolet (UV) reflectance. Because previous studies have shown that D. pumilio also has low reflectance in the UV range [20], we followed Saporito et al. [15] and mixed clay colors in proportions that best matched live frog colors according to the visual assessment of the authors (three males with normal color vision). We also compared spectrometric measurements of our red clay (the main color of the dorsum of the local aposematic form) against spectrometric measurements of the dorsum of live red-and-blue D. pumilio morphs taken by Summers et al. [20]. Reflectance values of live frogs remain near zero until a sharp peak between 625 and 675 nm (Fig. 12 in ref. [20]), matching reflectance spectra for the red clay, which remain relatively flat until a sharp peak around 625 nm (Fig. S1). Given the similarity in reflectance spectra for red clay and live frogs, we assume that predators at La Selva are likely to be “familiar” with the red replicas and will treat them like live, local frogs.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0048497.g002 Relative contrast estimates for the three visual contrast methods under an avian visual model, scaled to 1.Estimates are not directly comparable among methods. Color forms assigned different letters differ significantly from each other after Bonferroni correction (Table S1).
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0048497.g003 Observed-minus-expected avian attacks.Expected attack distributions were generated using the three visual contrast methods. Negative observed-minus-expected values signify that a color form was attacked less than expected and vice versa. Colors follow the legend in Fig. 2. Stars denote significant departures from the expected attack distribution. ‘*’: P<0.05, ‘**’: P<0.01. Crosses denote significant departures from the expected attack distribution from pairwise comparison of red vs. yellow replica attack rates. ††: P<0.01, n.s. : not significant.
We conducted six predation trials over 12 days with each trial in a unique location along LSBS’s dendritic trail system. Each trial consisted of ten 60 m transects; transects were spaced 50 m apart. Each transect contained 40 regularly spaced frog replicas placed directly onto the forest floor in a regular, alternating color order, haphazardly with respect to the background substrate. The density of replicas along study transects was realistic given the high density of D. pumilio in nature [21]. To aid our recovery searches, we ran clear fishing line approximately one meter above each transect and placed each frog replica approximately one meter to the right of the fishing line. We collected the replicas after they had been in the forest for 48 hours. We inspected each recovered replica for evidence of predation and classified predation events as avian, rodent, or un-attributable. Avian predators left beak shaped piercings and impressions in the clay; rodent predators left easily identifiable incisor marks (Fig. S2). During collection, we recorded how many seconds it took for a human researcher to locate each individual frog replica. These search time data were collected as a proxy for replica conspicuousness, at least to the human eye (see Text S1). If a replica was not found after 180 seconds, it was considered missing. We chose 180 s as our cutoff because a February 2008 pilot study determined that replicas not found by 180 s would likely not be found at all. Following recent studies, missing replicas were not included in subsequent analyses since their fates could not be determined reliably [15], [22].
Avian Visual-model-based Estimates of Replica Conspicuousness: To account for differences in visual conspicuousness among replica color forms, we took spectrometric measurements of the clay color mixtures used in the replicas as well as leaf litter samples collected from the replica transects. At each transect, a 25 cm×25 cm quadrat was thrown haphazardly onto the forest floor, and four representative leaves were selected from this quadrat. On each leaf we measured reflectance at three locations (averaging three spectrometer readings per location), and then calculated an average reflectance measurement for each leaf (Fig. S3). To quantify clay colors, we measured five samples of each standardized color mixture (red, orange, yellow, blue, brown, and black) used in the replicas, and one pen mark from a black permanent marker on a yellow clay background. We averaged three spectrometer readings per clay sample and then averaged across the five samples to obtain one spectrum representing each standard color mixture. For all color measurements, we used an Ocean Optics USB-2000 spectrometer with an R-400 reflectance probe and PX-2 pulsed xenon light source, and Optics OOIBase 32 v2.0.6.5 software (Ocean Optics, Inc., FL, USA, 2002). We used white and dark standards to calibrate the spectrometer before measurements were taken.
(ii) Replica conspicuousness to the avian eye: We focus on the avian predator assemblage because birds were likely the most common visual predators of D. pumilio at LSBS [15], [23], [24]. The avian visual system has both double-cones, which seem to be used in brightness and motion detection, and four single-cone classes that are used for color discrimination [25]. In general, bird eyes can be divided into two types according to their single-cone classes [26]: V-type and U-type; the latter type has greater sensitivity to UV wavelengths [26]. We used a model of the UV-insensitive, V-type avian visual system for this study [26], [27], assuming that the ambient light profile of LSBS’s forest understory matched the UV-poor “forest shade” profile of other lowland tropical forest sites [28]. Results for the U-type avian visual system are qualitatively similar and not presented below. Moreover, our results do not change qualitatively if an alternative ambient light profile [28] is used in our models instead. Animals may use chromatic information independent of brightness, brightness information independent of chromaticity, or both chromatic and brightness information to detect prey against the visual background [29]. Therefore, we calculated visual contrast between frog replicas and the leaf litter background in three ways by extracting from the cone excitation data: (1) only chromatic information, (2) only brightness information, and (3) both chromatic and brightness information. The computer code for extracting color information from the reflectance spectra is available upon request. Cone sensitivity functions [26] were derived from cone excitation values for ten avian species with a V-type eye (see Table 1 in ref. [26]). Chromaticity-only contrast: We transformed the four avian single-cone excitation values for each reflectance spectrum into three-dimensional coordinates in a tetrahedral color space using a method developed by [26]. We then calculated color contrast as the Euclidean distance between clay color and individual leaf color in this color space.Brightness-only contrast: To calculate brightness contrast, we calculated the difference in double-cone excitation between clay colors and individual leaf colors. Double-cone excitation values came from the reflectance spectra and a cone function derived from empirical measurements of double cone photoreceptor sensitivities and oil droplet absorbance spectra [30], [31], [32].Chromaticity+brightness contrast: Because chromaticity and brightness contrasts were estimated using different cone types, we normalized contrasts under each method to one. Then, under the assumption that chromaticity contrast and brightness contrast contribute equally to object-background discrimination, we summed the chromaticity and brightness contrast proportions to obtain a composite contrast measure.Hereafter, we refer to these three distance-based contrast measures as the “chromaticity,” “brightness,” and “composite” measures, respectively. To calculate an overall visual contrast score for each replica color and contrast method (chromaticity, brightness, and composite), we averaged the clay/leaf contrast scores across all leaf samples. Because aposematic replicas incorporated more than one clay color, we estimated whole-replica contrast under each contrast method by calculating the contrast between each individual clay color and the leaf litter background, and then taking a weighted average of these contrast values according to the area covered by each color on the replica (areas measured from digital photographs of frog replicas viewed dorsally).
We used a Chi-square (χ2) test of independence to test for differences in absolute predation rates (i.e. regardless of contrast) among color forms. Then, to ask whether predation rates are proportional to the contrast of each form (i.e. do differences in attack rates match differences in conspicuousness? ), we generated an expected attack distribution given replica contrast for each color form under each contrast method according to the equation:where i is the contrast method (chromaticity, brightness, or composite), j is the replica color (red, yellow, brown, or black), Ei,j is the expected number of attacks, Rj is the number of replicas recovered, Ci,j is the visual contrast value, and α is a constant of proportionality that ensures that the total number of expected attacks is equal to the total number of observed attacks. Finally, we compared the expected attack distributions to the observed distribution using a χ2 test of independence.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}