Introduction

In the study of evolutionary processes, contemporary biology tends to focus on multilevel causation (Martínez and Esposito 2014). With the end of the twentieth century, the bottom-up genocentric view of life processes started to be substituted by a more comprehensive approach which takes into account that not only genetically linked traits can co-evolve and that this is the case especially regarding behavior (Ryan 1998). In the following, we first briefly review some cases of complex behavior in reptiles which demonstrate an unexpected richness of semiotic capacities in these traditionally underestimated animals and then present a case study of the relationship between turtle head coloration in connection with mating behavior strategies and sexual size dimorphism (SSD).

Exposed surfaces of animals, i.e., their semantic organs, form a potent communicative interface (Kleisner 2008a, b, 2015; Maran and Kleisner 2010; Kleisner and Maran 2014). The inner potentialities of an organism can enter the senses of another living being when effectively expressed on the outer surfaces of the former and meaningfully perceived by the latter. Yet though conspicuous color marking is found in many species of turtles (Chelonia), the forces which select for the coloration are not well understood (Liu et al. 2013). The Four-eyed turtle (Sacalia quadriocellata), for example, has two eyespots behind its eyes (Liu et al. 2009) and these spots could either serve as a deterrent to predators or have a sexually selected function. Turtle body coloration, especially dull hues, may serve as camouflage, thus helping to protect their carriers against potential predators, and some turtles are even capable of gradually changing their color over a period of several weeks (Rowe et al. 2009). Conspicuous seasonal color dichromatism linked to sexual processes has been documented in Painted terrapins (Callagur borneoensis) (Moll et al. 1981) and Chinese red-necked turtles (Mauremys nigricans) (Anders 2012). Male Red-eared sliders (Trachemys scripta) dramatically change their color as they get older and larger (Lovich et al. 1990b). This transformation is accompanied by a shift in their mating strategies so that while the younger, conspicuously colored males engage in ritualized courtship behavior characterized by waving front limbs, the older, dull colored males often adopt more aggressive behavior such as biting (Thomas 2002). Darker melanin coloration of male Hermann’s tortoises (Testudo hermanii) has been positively correlated with aggressive behavior against conspecifics (Mafli et al. 2011). Conspicuous stripe pattern in genus Trachemys may serve as a species-isolating mechanism (Legler 1990) and the same has been hypothesized for genus Graptemys (Vogt 1993). As shown in Trachemys scripta elegans, dichromatism can exist in the ultraviolet range of light spectra invisible to human observers (Wang et al. 2013a). The visual apparatus of turtles is among the most complex in vertebrates: it includes seven cone types, with one type sensitive to UV light (Loew and Govardovskii 2001). Moreover, conspicuous patterns on turtles are correlated with the state of their immune system and thus carry information about the individual’s state of health (Polo-Cavia et al. 2013; Ibáñez et al. 2014). It is yet unclear, however, whether turtles themselves are able to perceive and interpret different coloration as a signal and whether coloration in general is related to sexual communication.

Umwelt Bias

When studying communication in nature, we should always keep in mind that we are human and our perceptual and cognitive abilities are thus limited in a particular way. On the other hand, the straw man of unwarranted anthropomorphism can restrict scientific interpretation of the way in which other animals experience the world ad minimum. Perhaps as a result of a traditional emphasis on the superiority of human skills and our social and cultural organization, it took us a long time to realize that the difference between humans and animals is not categorical. Nowadays, we are thus finally in position to appreciate that both extremes, that is, both excessive anthropomorphisation of animals and its opposite, can be and have been opportunistically used and misused in argumentation in politics, sociology, popularization, and even environmental conservation (Lestel 2011; Stella and Kleisner 2010; Tønnessen 2010).

Recently, Uexküll’s theory of Umwelt has attracted the attention of contemporary biologists who realized that the functionality of epigamic displays, warning signals, but also some other displays and behaviors, is significantly co-determined by the relevant organisms’ perception and cognition (Chittka and Brockmann 2005; Ryan 2011; Prum 2012; Ryan and Cummings 2013).

In studying the coloration of turtles, one thus ought to take into account the Umwelt structure of the species in question. We view the head ornaments in turtles as semantic organs (Kleisner 2008a, b, 2011, 2015; Maran and Kleisner 2010) which function via the meaning ascribed to them in an Umwelt-specific interpretation. The existence of semantic organs is objective in the sense that all members of a community (e.g., a species) view them as meaningful objects and relate to them accordingly. Semantic organs are thus objective but not universal entities because their functionality is constrained by the perceptual and cognitive abilities of a certain community of animal interpreters (e.g., turtles, turtle predators, turtle prey, and other inhabitants of turtle biotopes).

Communication in Non-Turtle Reptiles

In reptiles, the role of communication and other behaviors has traditionally been underestimated. Although many breeders of reptiles have been familiar with these phenomena for a long time, the scientific community often accuses them of being unduly sentimental because breeders’ observations tend to be anecdotal and are rarely presented in scientific literature. Moreover, we tend to neglect the impact of captive environment on the wellbeing of reptile species. By enriching the captive environment, we can encourage natural behavior, promote the psychological wellbeing of captive animals, reduce their boredom, etc. But, as Burghardt (2013) noted, this step would be impossible without resorting to some degree of anthropomorphic modeling, as introduced in his concept of ‘critical anthropomorphism’ which refers to Uexküll’s Umwelt theory (Burghardt 1991). In a similar way, we can talk about Umwelt-modeling as applied within current biosemiotics. Reptiles share many traits with birds and mammals, including sophisticated communication, problem solving, parental care, play, and complex sociality. For instance, some species of agamas and iguanas live in large, hierarchically ordered social groups. They even have ways of solving conflicts which arise within a group by communication based on head movements: it is a sort of ‘reptile diplomacy’. Leal and Powell (2011) have shown in a series of experiments with Anolis evermanni that lizards exhibit unexpected behavioral flexibility – traditionally ascribed mostly in birds and mammals – such as solving novel motor tasks using multiple strategies, reversal learning, and rapid associative learning.

It used to be generally believed that many complex behaviors which are in mammals and birds linked to maternal care and social behavior of juveniles are highly exceptional in the reptile taxa. Yet reptiles which guard their nests exhibit highly advanced forms of parental behavior (Rosenblatt 2003). Nest building and guarding is well known, e.g., in King cobras (Ophiophagus hannah) (Veselovský 2005) and maternal behavior linked to guarding the neonates has also been documented in rattlesnakes (Crotalus atrox) (Price 1988). Moreover, various types of social behavior in neonates are found across reptile taxa, a fact which challenges the alleged superiority of mammalian and avian behavior patterns (Burghardt 1977). An interesting case of social behavior in hatchlings has been described for example in Green iguanas (Iguana iguana), where juveniles from different clutches join together and form complex social groups which migrate together (Burghardt et al. 1977). Paternity analysis of stable aggregations of Australian lizards (Egernia stokesii) has revealed that these reptiles remain in groups formed on a family basis throughout their lives. Only reproductive mates are not related. Upon reaching maturity, young lizards Egernia stokesii seek genetically unrelated partners and establish new family groups consisting of the next generation of individuals (Gardner et al. 2001). Noble et al. (2014) have shown in an experimentally designed study that young male skinks (Eulamprus quoyii) learn faster than adults how to solve novel instrumental and association tasks if a conspecific demonstrator is used. Well known and well documented is maternal care in Crocodilians. In alligators and crocodiles, females defend their nests during the incubation of eggs and crocodile females even stay close to the nest for most of the incubation time. Towards the end of this period, hatchlings start to vocalize. This prompts the female to aid the juveniles and the yet unhatched animals are gently liberated from their eggs by the female’s jaws cracking the shell. She then transports the hatchlings in her mouth to a nearby water source. Juveniles then stay with the female for different lengths of time depending on the species and they communicate by vocal signaling (Garrick and Lang 1977).

Acoustic signaling is present in several taxa of reptiles. In Crocodilians, vocalization is present not only in juveniles but also in adults, who vocalize during courtship by vibrating their body surface in water. Crocodilians also produce other sounds, such as clapping of jaws or air expulsion through the mouth which results in bellowing, growling etc. (Garrick and Lang 1977). Chinese alligators (Alligator sinensis) form choruses involving both sexes to gather individuals for mating (Wang et al. 2009). Gans and Maderson (1973) divided the sound-producing mechanisms of reptiles in three categories: (1) non-modulated hissing by massive air expulsion, (2) vocalizations proper, such as modulated oral expulsions, and (3) mechanisms involving integumentary modifications. As mentioned above, acoustic signaling is found in a number of reptile groups, including geckos (Mercellini 1977; Frankenberg 1982), Pygopodids (Weber and Werner 1977), and strangely enough even snakes (Young 2003). Some of these signals play a role in intraspecific communication, others in heterospecific signaling and mimicry. Interestingly, the non-vocal marine iguanas (Amblyrhynchus parvulus) of the Galapagos recognize the alarm calls of Galapagos mockingbirds (Nesomimus parvulus) and respond to it by anti-predatory behavior (Vitousek et al. 2007). This is, among other things, convincing evidence of good auditory discrimination in these reptiles.

Furthermore, in addition to visual and auditory channels, reptiles also often use chemical communication. In 1971, Regnier introduced the term ‘semiochemical’ to describe chemical signals which mediate information between organisms and suggested that there may be many different types of both intraspecific and interspecific semiochemical communication (Regnier 1971). In reptiles, we find two kinds of chemoreception. The well-known tongue flicking represents vomerolfaction, a phenomenon which in reptiles has been more extensively studied than the other kind of chemical reception, namely olfaction. Reptiles, with the exception of crocodilians, use their well-developed semiochemical perception for various purposes, including prey and predator detection, species and individual recognition, mate choice, alarm signaling, and territoriality (Mason and Parker 2010). At this point, let us just conclude that literature and research focusing on semiochemical communication is extensive (see, e.g., Halpern and Martínez-Marcos 2003; Houck 2009; Mason and Parker 2010; Brykczynska et al. 2013; Robinson 2014).

Visual signals such as gestures and specific postures are used by Crocodilians to communicate or avoid direct confrontation with conspecifics and to alert heterospecifics before attacking (Garrick and Lang 1977). Head et al. (2005) hypothesized that skinks (Eulamprus heatwolei) use vision to find females and then apply chemoreception to evaluate the female’s sexual receptivity. Many lizard species communicate by various movements of their bodies or limbs. For example, both sexes of the phrynosomatid lizard Sceloporus graciosus perform so called ‘push-up’ displays, which include body posture (flattening, arched back), head movements, and limb extension. Body postures seem to be a strong and discrete signal, while head movements and leg extension are graduated, whereby different numbers of repetitions of leg movements or leg extensions in different situations take on different meanings (Martins 1994). The green anoles (Anolis carolinensis) use different head movement displays accompanied by colorful dewlap extension in three different situations: when directed to females, when directed to males, and in undirected displays (Jenssen et al. 2012). The third situation is especially interesting because undirected displays differ according to activities, such as environment monitoring or migrating through space. Moreover, particular displays which are usually employed in directed communication can be also used without any specific addressee. It would thus seem that undirected displays function as expressions of the internal state of the specimens (Jenssen et al. 2012) and semiotic co-option (Kleisner 2011; Maran and Kleisner 2010) of undirected behavior may potentially lead to the formation of directed displays. In various reptile taxa, postures and gestures such as head movement, dewlap extensions, lateral body compressions, etc., are associated with colorful pattern exposure (Brattstorm 1974).

In many cases, colorful patches on animal surfaces serve as indicators of the bearer’s quality. This can be explained by the presence of various pigments in color patterns, such as melanins, pterins, porphyrins, flavonoids, and psittacofulvins, which often function as antioxidants (McGraw 2005). By displaying these pigments, animals demonstrate their ability to produce or sustain these substances in levels sufficient to maintain their inner physiological function but also the outer signaling function. The most extensively studied pigments responsible for condition signaling are carotenoids. Animals do not synthesize these organic terpenoids themselves: they are received only in plant nutrition. Access to carotenoids then directly influences the quality (hue, brightness) of yellow, orange, and red color of animal tissues, which are further subjected to sexual selection (Inouye et al. 2001; Faivre et al. 2003; Saks et al. 2003). Pterins have similar optical properties as carotenoids (Martínez and Barbosa 2010) and in female Scleropus virgatus, one can observe during their reproductive period the appearance of pterin-based orange spots which serve as a signal towards males. The quality of these orange spots positively correlates with antioxidant contents of the egg yolk, which further affects the phenotype of their offspring (Weiss et al. 2011). A comparison between two species of anoles Norops sagrei and Norops humilis shows that both pterins and carotenoids contribute to final coloration and the resulting color is a combination of both (Steffen and McGraw 2007). Stuart-Fox and Moussalli (2008) suggest that in chameleons of the Bradypodion genus, conspicuous social signals drive the evolution of rapid color change with strong selection for signal detectability. To achieve higher reliability of signal, multimodal signaling is often used because multimodal signals have a bigger impact on the receiver than each signal on its own (Rowe 1999).

Coloration in the Archosaurian lineage played an important role in the evolution of key bird traits such as feathers and sexual selection was probably its major driver (e.g. Foth et al. 2014). We had already noted that especially Crocodilians display many behaviors similar to those of birds and mammals. Surprisingly, rather unexpected behavioral features such as aimless play (Lazell and Spitzer 1977; Burghardt 2005) and tool use (Dinets et al. 2013) have also been described in Crocodilians. In fact, Crocodilians are the closest living relatives of birds, since they belong to the same reptile evolutionary lineage of Archosaurs (see e.g. Sereno 1991), while all other reptiles belong to Lepidosauria. Nonetheless, Crocodilians are not a good candidate for studying the communicational role and evolution of conspicuous coloration in the Archosaurian lineage simply because they are all dull colored. Turtles, on the other hand, are the closest relatives of Archousaurs (Crawford et al. 2014; Shaffer et al. 2013; Wang et al. 2013b) and they seem to be more a promising candidate because many of them do possess conspicuous coloration patterns on their shells and soft tissues.

Communication in Turtles

Until recently, turtles were usually seen as nothing but well-adapted, silent, armored creatures with sophisticated physiology and unknown evolutionary origin, who just roam around trying to find some food. Lately, some highly interesting behaviors such as spatial and social learning has been documented in a few specimens of red-footed tortoise (Wilkinson et al. 2007, 2010, 2013; Mueller-Paul et al. 2014) but the artificial set-up in which these observations were made is probably rather unlike their natural conditions and the study used a small number of specimens. Some turtle enthusiasts, however, have observed some spontaneous higher cognitive abilities such as play (Burghardt 2005) or altruism (Petr Velenský and Jiří Moravec, personal communication) earlier. For example, when a turtle meets with an accident and end up turned on its back, potential help of conspecifics who help it return to a carapax-up position is crucial to its survival. Otherwise, a turtle lying belly-up would die of overheating or exhaustion. For the helper, on the other hand, helping its consepecific back to a belly-down position is not immediately profitable, especially in desert tortoises which live in environments where there is little food to go around. Yet it has been shown that in desert tortoises (Gopherus agassizii), this altruistic action is in the wild stimulated by the production of a suppliant sound (Patterson 1976). Turtles show ability to learn novel behavioral skills (Davis and Burghardt 2007), to learn about visual cues from experienced conspecifics (Davis and Burghardt 2011), and to retain these visual tasks for long periods of time (Davis and Burghardt 2012).

In turtles, maternal care – another example of altruistic behavior – was long thought to include only nest-site selection and urination on the clutch and the nest (e.g. Hamilton et al. 2002). Nonetheless, nest guarding has been recently described in Gopherus agassizii (Agha et al. 2013) and the mud turtle (Kinosternon) buries itself along with its clutch of eggs and remains buried during the entire incubation period (Rosenblatt 2003). Escalona et al. (2009) used GIS framework modeling to characterize the patterns of nest distribution in the freshwater Podocnemidid turtle. Within the clutching beaches they studied, the authors tried to distinguish between random nesting and cases of nest-site selection, i.e., deliberate choice of nest location. Their results, however, showed that neither random nesting nor nest-site selection provide the right model to describe Podocnemis nesting pattern. By systematizing their hypothesis in a testable network, they realized that the nest distribution must be due to the operation of another mechanism. Finally, they arrived at a hypothesis of ‘social facilitation’, which claims that the nest pattern results from social interactions between the Podocnemis females. The term ‘social facilitation’ was first used in 1924 by Floyd Henry Allport but earlier psychological publications had also contributed to this subject (Strauss 2002). Social facilitation model has also been proposed for sea turtles (Chelonia mydas) (Carr and Hirth 1960) and Carreta carreta (Owens et al. 1982). Escalona et al. (2009) found that unlike the models of random nesting or nest-site selection, the hypothesis of social facilitation provides a good explanation of existing data and very well describes the variability in the spatial distribution of clutches over three seasons. Nevertheless, it was still unclear how this social facilitation is maintained and implemented on inter-individual level. Ferrara et al. (2012) had recently published a revolutionary paper on turtle vocalizations, in which they offer the first evidence of post-hatching parental care in turtles. In their paper, the authors describe 11 different types of sounds which females and juveniles produce to communicate with one another. They show that these sounds are used by adults and juveniles as a signal to congregate for mass migration from nesting areas to feeding grounds tens of kilometers away. Vocal communication could thus potentially play a role in mediating the socially facilitated nesting in females if we view it as a sort of a consensual decision to gather a ‘turtle assembly’. Moreover, synchronic hatching mechanism mediated by acoustic signaling has been described in Carettochelys insculpta (Doody et al. 2012).

Many turtles produce vocal sounds. It is not clear, however, whether these vocalizations are related to social behavior or result simply from hyper-excitation connected with massive expiration and carry no message at all (see a review in Gans and Maderson 1973). When a Platysternon megacephalum is captured and taken from water, it produces with its opened giant jaws a roaring sound which is thought to be a warning aposematic signal (Sachsse 1969). In turtles, sound is produced not only by vocalization but also by mechanical friction of scaled surfaces, which results in a production of a stridulating sound (Gans and Maderson 1973). Vocalization in turtles is well observable in mating tortoises due to the loudness (e.g. Geochelone gigantea, Frazier and Peters 1981) and duration of call, which negatively correlates with mating success in Testudo marginata (Sacchi et al. 2003). Giles et al. (2009) show that both males and females of the Side-necked turtle (Chelodina oblonga) can produce a variety of sounds but males produce longer calls lasting up to several minutes. Ferrara et al. (2014) claim that many hitherto unexplained aspects of social behavior of aquatic turtles could be explained if vocalizations were taken into account.

Semiochemical communication plays an important role in turtle sociality. Male Gopherus agassizii have striking chin glands which play a role in building their social structure (Alberts et al. 1994). Not only the glands, but also urine and excrements play an important role in conspecific and heterospecific recognition in the Gopherus tortoises (Patterson 1971). During the mating season, chemical stimuli produced by specialized glands serve as sex-recognition cues in Sternotherus odoratus (Eisner et al. 1977). In Mauremys leprosa, both sexes use chemoreception to recognize the presence of conspecifics and different preferences for chemical cues during different seasons of the year have been demonstrated in both males and females (Muñoz 2004). In an experiment using the native Iberian turtle (Mauremys leprosa) and the alien Red eared slider turtle (Trachemys scripta), Polo-Cavia et al. (2009) concluded that reactions of the two species to the presence of chemical stimuli representing the other species can differ dramatically. Trachemys scripta neither avoided nor preferred the scents of Mauremys leprosa, but Mauremys leprosa avoided the scents of the Trachemys scripta and both species preferred the scent of conspecifics (Polo-Cavia et al. 2009). In their native range, Trachemys scripta share habitat with numerous different species but Mauremys leprosa do not. The difference between their reactions to the abovementioned semiochemical stimuli could thus therefore be due to their historical experience with heterospecific semiochemicals and its integration to the Umwelten of the two species. Head movements during courtship dances may help spread the semiochemicals from males to females (Auffenberg 1966; Polo-Cavia et al. 2009).

Courtship behavior is an important source of information which provides us with clues to understanding turtle communication as such. Recently, courtship behavior was extensively reviewed by Liu et al. (2013), which is why in the following we present only a brief overview. As mentioned above, turtles use during their courtship both acoustic and chemical cues. At this point, however, let us rather focus on the tactile and visual behavior which complements this acoustic and semiochemical communication. First of all, it should be noted that not only the male, but also female turtles are active in courtship behavior (Lovich et al. 1990a). Female receptivity in turtle courtship forms a continuum, meaning that females either allow or reject male advances immediately or let the males continue their courtship and eventually participate in courtship and copulation (Ferrara et al. 2009). Liu et al. (2013) recognize six general courtship display types in freshwater turtles: (1) Biting, which may be light and harmless or strong enough to cut the skin. Biting may be directed to the soft tissues or the shell. (2) Nudging and (3) rubbing are head-mediated tactile behaviors, in freshwater turtles often directed to the chin, cloaca, carapace, or the plastral bridge of the female. This behavior can also serve to transmit the semiochemicals. (4) Head movements consist of three categories of behavior: vertical vibration of the head, horizontal head swinging, and head movement on the female carapace. All three types of head movements can occur independently of each other and the presence of these movements correlates negatively with foreclaw displays (Liu et al. 2013). (5) Foreclaw displays include gentle stroking of the female’s head by front limbs. In the Emydidae family, this kind of display became ritualized and is called titillation. From a phylogenetic perspective, the evolution of foreclaw displays and head movements is linked to the disappearance of biting (Liu et al. 2013). (6) Gulping is a behavior in which the male expels water by mouth or nose so as to create currents around the eyes and face of the female. Similar types of behavior have been described in tortoises (Auffenberg 1977), though there are some differences which could be ascribed to environmental constraints (e.g. in tortoises, semiochemical communication is more contact-dependent because the relevant chemicals are not soluble in water). Berry and Shine (1980) have shown that male mating strategy across the whole order Chelonia correlates with habitat type and SSD. In species where the males are larger than females, they observed combats and forced inseminations, but where females are the larger sex these behaviors are absent and female choice seems to be of primary importance (Berry and Shine 1980). Sexual size dimorphism correlates with mating strategy also when only freshwater turtles are considered (Liu et al. 2013).

In our study, we use data on sexual behavior (Liu et al. 2013) and data on size sexual dimorphism (Ceballos et al. 2013) to test whether in freshwater turtles, conspicuous coloration is related to behavioral strategies. Evolution of SSD could be the result of developmental heterochrony (Gibbons and Lovich 1990), sexual selection, and environmental pressures (Cox et al. 2007; Lindeman 2008). Here we hypothesize that species with biased SSD are species where mate choice plays a role and mating behavior has driven the evolution of colorful patterns for communication. Because SSD is correlated with mating behavior, we further predict that the more gentle and elaborate the behavior during the mating, the more colorful should be the species. Finally, we discuss the role of head color patterns in turtle communication during mating.

Material and Methods

Data Acquisition

Data on sexual behavioral traits was taken from Liu et al. (2013). Courtship and mating behavior of freshwater turtles was binary coded. Behavioral traits consisted of data on rubbing (rub), biting (bite), gulping (gulp), nudging (nudge), foreclaw display (FD), and head movement (HM). Altogether, we analyzed data on 35 species.

Data for SSD was taken from Ceballos et al. (2013). The data was based on an estimated sexual size dimorphism index distributed from −0.072 to 0.407. Negative values of SSD represent male-biased body size dimorphism (males larger than females), while positive values represent female-biased body size dimorphism (females larger than males). Species with equal average body size of both sexes would have SSD = 0 but in our sample, we found no such case.

In our study, we focused specifically on turtle head coloration. Coloration was considered conspicuous when (1) warm and bright colors were present, or (2) the predominant color was pale, while bright coloration formed well-defined ornaments (sharply bordered spots, stripes, blotches, eyespots). We did not, however, include cases where a contrasting pattern of head coloration was clearly due to counter-shading effect (Thayer 1896; Komárek 2003).

Information about the type of coloration (conspicuous or dull) in a particular species was acquired from photographic material and recorded in binary code. Head coloration was coded based on expert evaluation of one of the authors (JB). Most species were evaluated for coloration based on pictures published in the Terralog book series (Vetter 2004, 2005, 2011; Vetter and van Dijk 2006). Information on Australian turtle species was taken from sources listed at http://reptilesofaustralia.com/turtles/turtles.htm#.U-OvI_l_uSo (accessed August 7, 2014) because Terralog volume with Australian taxa is not yet published. Altogether, our study included 35 species and we used only species where all the relevant information on behavioral traits and SSD was available (see Table 1).

Table 1 List of species, taxonomic categorization, and type of coloration

Statistical Analysis

All analyses were performed in R-software for statistical computing, version 3.0.2 (R Development Core Team 2013). To control for phylogenetic relationships, we used a matrix based on taxonomical categories (suborder, superfamily, family, subfamily, genus, species) converted to a tree using R functions ‘as.phylo’ and ‘compute.brlen’; Ape library (Paradis et al. 2004). To calculate the length of branches, we used the default setting for the ‘compute.brlen’ function.

To avoid co-linearity between variables represented by behavioral traits, we performed Principal Coordinate Analysis (PCoA, Ape library) with binary method for computing matrix distances and Cailliez method of correction for negative eigenvalues (Cailliez 1983). The first two principal coordinate axes, which explained 55 % of variability in behavioral data, were extracted and used in further analysis. The Kendall tau rank correlation coefficient was used to explore the relationship between behavior and head coloration. To test the relationship between behavior and SSD and head color while controlling for phylogeny, we used Phylogenetic ANOVA (R function phylANOVA within phytools library) (Revell 2012).

Results

The Principal Coordinate Analysis of six behavioral traits resulted in two PCo axes (Fig. 1). The first axis (PCo1) explained 33 %, the second (PCo2) 23 % of all variability. The first axis accounted mainly for the aggressive/non-aggressive continuum of sexual behavior (non-aggressive meaning foreclaw display and gulping, aggressive being rubbing, biting, and nudging). Head movements were responsible for most variation along the second PCo axis.

Fig. 1
figure 1

Biplot of Principal Coordinate Analysis of mating behavior traits (scaled and centered). The first and second axes explain 33 and 23 % percent of behavioral data respectively. The first axis stands for aggressive/non-aggressive mating behavioral continuum. Vectors denote the following behavioral traits: rubbing (rub), biting (bite), gulping (gulp), nudging (nudge), foreclaw display (FD), and head movement (HM)

When the scores of PCo axes were correlated with head coloration, only PCo1 turned out to be significantly associated with head color (PCo1: N = 35, p = 0.033, τ = −0.304; PCo2: N = 35, p = 0.569, τ = 0.081). Aggressiveness in courtship and mating behavior thus showed significant negative correlation with conspicuous head color pattern. Once the phylogenetic structure of data was taken into account, the correlation between PCo1 and head coloration was no longer significant (F = 6.57, p = 0.1) but the trend was still present (Fig. 2).

Fig. 2
figure 2

Boxplot showing scores of the first PCo axis for turtle species with conspicuous and dull head coloration. Whiskers denote minimum and maximum values. Top and bottom of the box represent +/− standard deviation, while the inner band indicates the mean value. Conspicuously colored head of Batagur borneoensis (left) and dull colored Actynemys marmorata (right) are depicted below. Original drawings by Lucie Čermáková

Further, we tested for any non-random association between SSD and conspicuousness of head coloration. We found a statistically significant negative correlation between SSD and head coloration (F = 15.2, p = 0.014, corrected for phylogeny) (Fig. 3).

Fig. 3
figure 3

Boxplot showing values of sexual size dimorphism (SSD) for turtle species with conspicuous and dull head coloration. Whiskers denotes minimum and maximum values. Top and bottom of the box represent +/− standard deviation. The inner band shows the mean value

Discussion

The biological meaning and evolution of conspicuous coloration in freshwater turtles has so far been studied rather sporadically. Our study focused on a possible role of head coloration in mating and courtship behavior of freshwater turtles. We explored the idea that the level of complexity in external appearance may reflect the animal’s experience of the world, in particular the richness of the animal’s Umwelt, its communication capacity, or richness of semiotic bonds with other living beings and the environment (Portmann 1960; von Uexküll 1921; Kull 2010; Kleisner 2008b, 2011; Maran and Kleisner 2010; Maran 2011). We provide the very first evidence to support a claim that vivid head coloration in freshwater turtles is associated with sexual behavior and SSD.

Our results show that there is a relationship between head coloration and SSD. In particular, an increase in male-biased SSD (i.e., if the male is larger than the female) is associated with a reduction of head color ornaments. And conversely, in species with female-biased SSD (larger females, smaller males), we found a tendency to develop colorful markings on the head. Conspicuous head coloration tends to be present in species which employ ‘gentle’ types of sexual behavior. In other words, species whose males rely on foreclaw display and gulping, rather than rubbing, biting, and nudging, tend to display a conspicuous pattern on their heads.

It is worth noting that on their own, these markings are usually not sexually dimorphic, that is, the same color ornament is present in both males and females. While selection pressure acting on one of the sexes may be responsible for the appearance of head color patterns, they are inherited by both sexes. This is clearly the case in turtles, since in most of their species, heterogamety does not occur and sex is determined by the temperature to which eggs are exposed during incubation period. Two turtles with almost identical genetic background can thus be of a different sex (Vogt and Bull 1982). It would be interesting to test whether the probability of sexual dichromatism increases in turtle species where sex is genetically determined.

The evolution of head color patterns could be causally linked with behavioral strategies derived from variation in SSD among turtles (Ceballos et al. 2013; Liu et al. 2013). Sexual size dimorphism is generally seen as an adaptation of the sexes to their distinct sexual roles and their subsequent ecological diversification (Fairbairn 2007). In reptiles, there are three evolutionary pressures which affect the ultimate level of SSD: 1) sexual selection of the size of males, 2) fecundity selection favoring larger female size for larger maternal size, and 3) natural selection for resource partitioning (Cox et al. 2007) or habitat use (Lindeman 2008). Analyses of the evolutionary rate of body size change in freshwater turtles by Ceballos et al. (2013) show that in some families, the main movers of SSD evolution are the males (in Chelidae, Chelydridae, Emydidae, Geoemydidae, and Kinosternidae), while in other families, SSD evolution is driven by females (in Podocnemidae and Trionychidae). Berry and Shine (1980) have demonstrated that habitat use correlates with SSD. The males of bottom-walking and semi-aquatic turtle species are larger than or as large as females, while in aquatic swimmers, females are usually larger than males. Puts (2010) came up with a ‘dimensionality hypothesis’ which states that with increase in the number of dimensions of the environment we witness a decrease in aggressive behavior (contests) and the appearance of more elaborate ornaments due to mate choice (see also Bartholomew 1970; Stirling 1975). This could also be the case in freshwater turtles, since in aquatic swimmers (3D), which tend to have female-biased SSD, one encounters gentle mating behavior and according to our results also conspicuously colored head patterns, which contrasts with bottom-walkers and semi-aquatic turtles (2D). Alternatively, Gibbons and Lovich (1990) suggest that intraspecific differences in SSD could be the result of a developmental shift in the timing of maturity (heterochrony). Ontogenesis may also influence coloration and mating strategies as described above in the Introduction (Lovich et al. 1990b; Thomas 2002). At the moment, however, we are not in position (as many before us, including Blackenhorn 2005; Cox et al. 2007) to distinguish which mechanism of SSD evolution came first (ecological, behavioral, or developmental divergence of the sexes). To understand the causal hierarchy between SSD, head coloration, and mating behavior, further evidence is needed, such as for instance a detailed ancestral reconstruction of target traits.

Possible weakness of our findings may be that after the application of a phylogenetic correction, we obtained null results for correlation between mating behavior and head coloration with positive trend. Note, however, that correlation between head coloration and SSD remained significant even after phylogenetic correction. Does that imply that we ought to reject the hypothesized correlation between head coloration and sexual behavior? Not really. It only indicates that the link between conspicuous coloration and mating behavior is phylogenetically constrained (at least as shown by our dataset). These results therefore do not necessarily imply that the phenomenon itself is somehow less important or less surprising. Shared phylogeny does not automatically explain the ultimate evolutionary function of any display. Phylogeny may account for the fact that different displays co-evolved within the same lineage, but it does not explain why the displays retain their function and do not disappear.

We suggest that the evolution of head coloration in turtles could be explained in terms of evolutionary scaffolding and semiotic cooption/selection processes (Kleisner 2015; Maran 2015; Maran and Kleisner 2010; Hoffmeyer 2007, 2008, 2014a, b). Similar evolutionary processes probably led to the appearance of various semantic organs including the human face, wing ornaments of butterflies, coat patterns of mammals, etc. First of all, conspicuous head ornaments did not appear strictly speaking de novo, out of the blue. They may have arisen by a scaffolding process from a basic pattern, which originally emerged as a self-organizing cell structure (such as the striped pattern in most Emydid turtles) but we find similar body patterns also in other families, such as Chelidae, Trionychidae, Geoemydidae, and Kinosternidae. This phase of the phenotypic change might have been affected by various genetic/epigenetic, developmental, and environmentally induced processes and their combinations. These changes in the phenotype, i.e., the modified stripes or patches, can but need not correlate with the animal’s state of health and its biological quality (i.e., fitness indicators). They can remain neutral or be selected for crypsis. Secondly, the scaffolded and variously shaped patterns can be subsequently internalized within the Umwelten of a community of perceivers (such as conspecifics), associated with some quality (e.g. males with vividly colored patches are less aggressive), and co-opted for a new biological function. The relative simplicity of this two-tier model also explains how conspicuous head patterns may have evolved independently in different turtle lineages. Moreover, the likelihood of a repeated emergence of conspicuous head patterns can be relatively high when we take into account the shared basic patterns (spots or stripes) in turtles and subsequent multiple semiotic cooption of these basic patterns for signaling purposes.

To sum up, we believe that the relationship between conspicuous head coloration and SSD in turtles is most likely influenced by all three of the abovementioned selection pressures involved in SSD evolution (environment, development, and behavior). Different habitats simply enable or hamper visual communication between sexes, which affects the Umwelten and the behavioral repertoire of the species in question. This relation to habitat use implies that bottom-walkers and semi-aquatic species should be less colorful than species residing in other habitats (cf. Berry and Shine 1980). This explanation is parsimonious because one can well imagine that the environment at the bottom of a pond or in the marshland eventually leads to a deterioration of organisms’ ability to visually distinguish colorful signals and in case of contest or forced mating, it is more difficult to escape. Reptile behavior and communication should no longer be underestimated and one ought to make assumptions similar as in the endotherms.