Evi A.D. Paemelaere^1*, Craig Guyer¹ & F. Stephen Dobson^1,2*

Predation is one of the major selective agents influencing evolution of color patterns. Cryptic color patterns decrease detection probability by predators, but their concealing function depends on the background against which patterns are seen; therefore, habitat use and color patterns are tightly linked. in many anole species, females exhibit variation in dorsal color patterns; the drab and perhaps cryptic colors of the patterns suggest a predator avoidance function behind this polymorphism. We tested whether these different color patterns experience different predation rates depending on their micro- habitat. We expected each pattern to form at least one optimal combination with a typically used micro-habitat that would result in lower predation compared to other morphs in the same micro-habitat. We tested this hypothesis for anoles at La Selva, Costa Rica, using clay models resembling a common species at this site: Norops humilis. The first experiment tested for variation in predation on various substrates. We included leaf litter, live leaves, and two size classes of woody stems, using 44 models for each pattern substrate combination. A second experiment tested effects of perch height (10 and 60cm) and diameter (<2cm and >5cm), with 50 models for each pattern perch  combination. We found differences in predation rates between the morphs depending  on their micro-habitat. Specifically, the striped morph had a significant advantage over the others on green leaves. in the second experiment, striped morphs showed significantly lower predation on low than on high perches, irrespective of perch diameter. Reticulated models had an advantage over other morphs on thin stems for the first experiment, where models were placed about 60cm high. Diameter did not have a significant effect on predation for reticulated morphs when height classes were combined. Dotted models did not experience an advantage over the other morphs in any of the treatments. in leaf litter and on thick perches no morph had any advantage over another, and leaf litter predation rates were generally low. These results support a role for predation in maintaining multiple female morphs within small Costa Rican anoles, such as N. humilis.

Key words: female polymorphism, Norops humilis, selection, predation, La Selva Biological Station, habitat use.

En los animales, los patrones dorsales de coloración a menudo se asocian con la protección contra la depredación. Con el fin de analizar el papel que tienen los micro-hábitat en  la  depredación  de  lagartijas  hembras  que  presentan variaciones en los patrones de  coloración dorsal (colores grises, crípticos),  se realizó en la Estación Biológica La Selva, Costa Rica un estudio con el objetivo de comprobar si los patrones de diferentes colores provocan cambios en las tasas de depredación en función de a su micro-hábitat; se esperaba que cada patrón formara una  combinación óptima utilizando el micro-habitat y que permitiera de esta manera reducir los niveles de depredación. Para evidenciar esta hipótesis se utilizaron modelos de arcilla que asemejan a la especie Norops humilis propia del lugar. Se realizaron dos estudios, el primero analizó modelos colocados en cuatro sustratos diferentes: hojarasca, hojas vivas y dos clases de tamaño de tallos leñosos, los cuales representaban cada hábitat donde es posible observar esta especie. Un segundo experimento estudió el efecto de la  altura y el diámetro de la percha en la depredación, para ello se colocaron los modelos sobre tallos de diferentes alturas y diámetros. Fue posible observar en el primer  experimento que el morfo rayado  tuvo  una  ventaja  significativa  sobre  los  demás morfos en las hojas verdes y que los modelos reticulados tuvieron una ventaja sobre otros morfos en tallos delgados. El segundo estudio mostró que los morfos rayados tienen una tasa de depredación baja en perchas altas, independientemente del diámetro de la percha. Fue posible comprobar que los morfos punteados no experimentaron ventaja sobre otros morfos en ninguno de los dos estudios.

Palabras clave: poliformismo, Norops humilis, selección, depredación,  Estación Biológica La Selva, hábitat.

In animals, dorsal color patterns are often associated with protection from predation. Flashing coloration in prey species often func- tions as a warning signal for the predator, while drab color generally decreases visibility (Poulton 1890, Cott 1940, Edmunds 1974). Perception  of  color  patterns,  however,  depends  on the background that the animal is seen against (Endler 1978, Ruxton et al. 2004, Bond 2007), so that visibility is not inherent to a color pat- tern. Reduced visibility to predators (camouflage) can actually be obtained through color patterns that fail to stand out from the background (crypsis) or patterns that deflect the predator’s vision from the contours of its prey. in the latter case the colors need not be similar to the background (Cott 1940, Endler 1978, Merilaita & Lind 2005, Cuthill et al. 2007). A camouflaging color pattern can serve a generalist role, forming a compromise by providing camouflage against a variety of backgrounds, but not maximizing camouflage for each back- ground separately (Merilaita et al. 1999, Merilaita et al. 2001). Alternatively, a pattern can be specialized for maximum concealment for a specific micro-habitat element and individuals may select different microhabitats within the heterogeneous environment, so that a variety of  color  patterns  can  be  maintained  within a population (Edmunds 1974, Endler 1978, Endler 1984, Hedrick 1986). in the latter case, predation in heterogeneous environments may result in color polymorphic populations (Ford 1945, Cain & Sheppard 1954).

Anoles (Squamata: Dactyloidae) are known to vary in dorsal patterns within populations (e.g. Fitch 1975, Schoener & Schoener 1976). Moreover, these pattern variations are primarily limited to females, and the same patterns re-occur in many species of anoles throughout their range (Savage 2002, Paemelaere et al. 2011a). These patterns vary in shape and do not appear to differ in color. Common female-only morphs or gynomorphs are the cream-colored vertebral striped and reticulated  patterns.  The  variety  of  patterns may also include a male-like or andromorph pattern, which differs among species. The color pattern of these anoles consists of varieties of brown and green, suggesting that these patterns help  reduce  visibility  to  predators  (Collette 1961, Fitch 1975, Macedonia 2001), but the maintenance of this polymorphism has rarely been studied.

In Norops sagrei (Duméril & Bibron 1837) frequency dependent selection appears to play a role in the maintenance of three female morphs (Calsbeek et al. 2010), but for N. humilis (Peters 1863), no support was found for this mechanism (Paemelaere et al. 2011b). A comparative study found that female polymorphism in anoles is associated with perch use (Paemelaere 2010). Previous studies indicated that female morphs select different perches. in particular, striped females of N. sagrei appear to select thinner perches (Schoener & Schoener 1976),  while  in  N.  polylepis  (Peters  1874) striped females select higher perches (Steffen 2010). Both studies tested the hypothesis that dorsal pattern polymorphism in anoles is maintained through differential concealment among alternative morphs that select different back- grounds. Although perch use appears to differ among female morphs, no study has addressed the effect of predation on different morphs.

Methods have been developed to quantify how easily an individual is detected by a predator, depending on the background the individual is seen against (Endler 1990, Endler & Mielke 2005). While these techniques allow for relatively easy assessment of background matching for color polymorphism, such techniques have not been well developed for pattern polymorphism (Endler & Mielke 2005). Clay models have gained popularity to assess the interactions between predator and prey because predation is difficult to study under natural circumstances (Brodie 1993, Bittner 2003, Howe et al. 2009, Noonan & Comeault 2009, Steffen 2009). The advantage of clay models is that individual variation, such as size and behavior, can be removed. Under the hypothesis of background matching by morphs, we expected to find different predation rates under a variety of pattern-background combinations. Specifically, we expected each morph to match at least one background for which predation of this morph was lower than for other morphs.

Perch substrate: The first study took place in June, 2007. Clay models were placed on four different substrates that represented habitat elements where the polymorphic anoles were seen: leaf litter, green (live) leaves, stems less than 2cm diameter and stems more than 5cm diameter. All stems were woody, to avoid effects of green versus brown stems. On live leaves and stems, models were placed as close to 60cm high as possible, because this is the average perch height of adult N. humilis (Talbot 1979). Each pattern-substrate combination received 44 models, for a total of 528. 11 study plots each contained four replications of each of 12 possible combinations of dorsal pattern and substrate. A random number was assigned to  every  combination  using  ‘sample(48)’  in R Version  2.8.1 (R Development Core Team 2008), and used as the order in which the combinations were placed within the plot. Models were placed in six rows. The minimum distance between models was two meters, but this distance increased if the pre-determined perch type was located at a slightly greater distance. Models were checked every third day. Models with predation were removed from the plot and damaged  models  were  repaired  by  smoothing the surface. Because damage from ants increased over time, all models were collected after 12 days.

Perch diameter and height: In July 2007, a second study investigated the effect of perch height and diameter on predation. Models were placed vertically on stems of different diameters and at different heights. The diameter of the stem was lower than 2cm or thicker than 5cm. Models were placed at 10cm or 60cm height  from  the  forest  floor.  Combinations were randomized and placement depended on the availability of perches as described above. For each of 12 combinations of pattern, diameter and height, 50 models were prepared for a total of 600 models. They were placed in plots as described above and left for four days before collecting all of the models and checking them for signs of predation.

Both experiments were carried out in disturbed forest (abandoned plantations, secondary growth). All three common polymorphic species were encountered in this type of habitat.  Plots  were  separated  by  at  least  100m. Upon collecting the models, they were placed in Ziploc bags and transported to the lab, where predation marks were recorded. in both experiments, predation rates per pattern per back- ground were summed over the full duration of the experiment. Only V or U-shaped imprints and small symmetric slits and punctures were scored as predation (Brodie 1993, Saporito et al. 2007, Steffen 2009) (Fig. 2). Ant damage was not scored as predation. When multiple predator marks were seen on one individual, a single predation event was scored. Occasionally tails were found to have fallen off of the model onto the ground. Marks on these tails were  not  included  in  the  analysis.  Models that were not retrieved were considered lost. Analyses were run with and without scoring the lost models as predation events. in addition, analyses were ran with serially attacked models removed, because attacks on neighboring models could result from easier detection after one of the models had been approached.

In the first experiment, absence or presence of predation on the model was compared for the different pattern-background combinations. We used a randomization test of independence with  5 000  iterations  partitioned  per  background (McDonald 2009). The strength of the association was reported as Cramér’s phi (φ_c), which ranges between 0 and 1, but 0.5 would be considered a strong association, 0.3 medium and 0.1 weak. in the second experiment, this test  was  also  used  to  determine  differences in predation at differing perch heights and on differing perch diameters for each morph. The same randomization test was also used for each experiment to determine differences in predation rate on morphs, independent of the perch they  were  on. All  analyses  were  completed with proc freq in SAS/STAT ® software, Version 9.1.3 for SAS System for Windows.

Perch substrate: Of 515 retrieved models, 111 (21%) showed signs of predation. Overall, predation was highest on dotted (andromorph) models, moderate for reticulated gynomorphs, and  lowest  for  striped  gynomorphs  (Dotted: 29.8%, Reticulated: 19.5%, Striped: 15.6%, Χ²_{(2, N=515)}=9.06, p=0.01, φ_c=0.13). For models placed on live leaves, the number of predation events differed significantly  between  morphs; striped morphs experienced less predation than expected,  while  the  opposite  was  true  for both other morphs (Χ²_{(2, N=129)}=6.45,  p=0.04, φ_c=0.22; Fig. 3). On stems with a diameter less than  two  centimeters,  the  reticulated  model had a lower attack rate and the dotted morph a much higher attack rate than expected (Χ²_{(2, N=131)}=8.30,  p=0.02,  φ_c=0.24).   Predation  in leaf  litter  was  generally  very  low  and  did not  significantly  differ  among  morphs  (Χ²_{(2, N=123)}=0.81,  p=0.67,  φ_c=0.08).  Similarly,  no significant  difference  was  found  in  predator attacks on models placed on stems greater than five centimeters in diameter (Χ²_{(2, N=132)}=4.26, p=0.12,  φ_c=0.18).  When  repeating  the  analysis  with  lost  models   scored  as  predation, results were virtually unchanged (not shown). Removing  serially  attacked  models  resulted in similar patterns  of predation rates, but low sample  size  resulting  in  insignificant  differences (not shown).

Perch  diameter  and  height:  Of  600 models,  593  were  retrieved  and,  of  these, 6.7 % showed signs of predation. No significant difference in predation was found for morphs  placed  at  perches  less  than  two  or more than five centimeters diameter (Dotted: Χ²_{(1, N=196)}=0.88, p=0.35, φ_c=-0.07;  Reticulated: Χ²_{(1, N=198)}=0.07, p=0.79,φ_c=0.02;  Striped: Χ²_{(1,  N=199)}=0.00,  p=0.95,  φ_c=0.00;  Fig.   4). When comparing predation between low  and high perches, dotted and reticulated morphs did not  significantly  differ  in  predation  (Dotted: Χ²_{(1, N=196)}=0.10, p=0.76, φ_c=-0.02; Reticulated (Χ²_{(1,  N=198)}=1.73,  p=0.19,  φ_c=-0.09).  Striped morphs,  on  the  other   hand,  were  predated significantly less on the  lower perches (Χ²_{(1, N=199)}=4.95, p<0.03,  φ_c=-0.15). in  our study, predation did not differ between female pattern morphs (Χ²_{(1,  N=593)}=0.66, p=0.72). Repeating the analysis with lost models scored as predation did not change the conclusions.

The microhabitat hypothesis for maintaining female polymorphism predicts that predation differs based on pattern-background combinations. Hence, survival would be highest for females selecting the background that provides the best concealment for their pattern. The prolonged survival and consequent higher lifetime reproductive output of these females could be sufficient for polymorphism to be maintained. Our first experiment showed that clay models representing female morphs of common anoles at La Selva differ in the rate of  predator  attacks,  depending  on  the  substrate they were placed on. We also found a higher overall predation rate on dotted morphs compared to the other two, but this difference was not seen in the second experiment. In the second experiment, striped morphs exhibited lower predation rates on relatively low perches, with a non-significant qualitative trend in this direction by reticulated morphs.

The difference between the first and second experiment could be due to differences in the predation rates. Predation in the first study was high compared to the typical predation rate of 5-10% on clay models (Bittner 2003, Husak et al. 2006). This rate was found in studies leaving models in the field between four and seven days and thus the difference in predation rate may be due to the relatively long time period (12 days) during which models in the first experiment were left in the field. indeed, predation rate per day was 1.75%, so that four days would result in a 7% predation rate. The second study, in which models were left out for four days, predation rate was 6.7%, and thus consistent with predation rates generally seen in clay model studies.

The first study showed that the reticulated morph had a significant advantage over striped and dotted morphs on thin stems, at least when placed at 60cm high (where the second experiment showed that % predation was relatively high). When low (10cm) and high (60cm) heights were combined in the second study, no significant effect of diameter was seen on predation rates for the reticulated morph, though there was a nearly doubled rate of predation at the greater height. The striped morph had lower predation than expected on all substrates, but it had particular advantage over the other two on green leaves. Although the brown coloration of all morphs stands out against the green background of these leaves, the stripe may function as a disruptive pattern (Cott 1940), rendering it more effective on backgrounds that are in strong contrast with the brown body color (Stevens & Cuthill 2006). For stems, the typical, elevated perch for N. humilis, striped morphs were found to experience significantly less predation on low stems, compared to higher perches. Predation on the dotted morph was generally higher than expected. The higher predation rate on the dotted morph in the first study suggests that this andromorph pattern may be the least concealing.

Overall, our results supported the hypothesis of differential predation on morphs based on the background against which they were seen. The second prediction of the micro- habitat hypothesis is that individuals select the background that provides the best concealment for their pattern. Future research should test whether these anoles are indeed selecting the perches where predation is lowest for their dorsal pattern.

We are grateful to M. Wooten and B. Zinner  for  comments  on  earlier  versions  of the manuscript. L. Pastorello provided assistance with statistical procedures. M. Paemelaere helped with the design of the mold and construction  of  models,  and  A.  Liner  and R.  Conway  assisted  in  the  field.  We  thank R. Saporito for helpful tips on clay model experiments. We also owe our gratitude to the staff of the La Selva Biological station. This research was funded by the Organization for Tropical  Studies  and  the Auburn  University Graduate School. This research was approved by IACUC, permit number 2007-1177, and by the Costa Rican Ministry of Environment and Energy (MINAE), permit number 01521.

Received 12-IV-2012. Corrected 20-IX-2012. Accepted 22-X-2012.