Courtship Displays Are Condition-Dependent Signals That Reliably Reflect Male Quality in Greek Tortoises, Testudo graeca
Abstract
The courtship behavior of male Greek tortoises (Testudo graeca) is based on a multiple signaling system that involves tactile, visual, olfactory, and acoustic signals. In some recent studies on marginated (Testudo marginata) and Hermann's tortoises (Testudo hermanni) we showed that male mounting success was significantly and positively correlated to the intensity of courtship displays. This was due to the number of rams and bites and the number of calls emitted during mounting, which are considered to be condition-dependent signals that reliably convey information on male quality. In this correlative study, we analyzed relationships between male morphology, hematological profile, courtship intensity, vocalizations, and mounting success in a group of 104 Greek tortoises breeding in seminatural enclosures. As expected, our study showed that mounting success of males was positively correlated to the number of rams and the frequency of interactions during courtship and negatively correlated to the call duration. Moreover the hematocrit was positively correlated with the call rate and the number of rams. Therefore, courtship signals exhibited by male Greek tortoises, including vocalizations, reliably reveal different components of male condition, and females may use these signals to choose high-quality partners, as previously observed in marginated and Hermann's tortoises.
One of the more interesting topics in sexual selection is the evolution of courtship signals, which represent the key cues in mate choice processes of many animals (reviewed in Andersson 1994; Bradbury and Vehrencamp 1998; Searcy and Nowicki 2005). The well-known “handicap principle” or the “good genes” hypothesis (Zahavi 1975) suggests that individuals of greater fitness are able to signal their quality through handicapping behaviors or morphological traits that actually lower their viability. According to this theory, exaggerated secondary sexual ornaments, such as large morphological traits, bright coloration, or redundant song repertoires (Andersson 1994), may act as honest, condition-dependent signals, and only individuals of superior quality are able to express these costly ornamentations (Grafen 1990). Generally, costs resulting from producing extravagant sexual traits may be direct, such as increased predation or parasitism, or indirect, when negative effects on viability arise through physiological consequences, such as increased energy expenditure or reduced immune competence (indirect or physiological costs; Møller and Saino 2004; Kotiaho 2001; Setchell et al. 2009). This model of sexual selection postulates that female mate-choice is just based on these costly and therefore honest signals of male quality (Trivers 1972; Zahavi 1977; Halliday 1978; Galeotti et al. 2005c). This allows females to select mates that either directly or indirectly provide fitness benefits such as avoidance of parasite transmission, increased investment in offspring, or “good genes” for vigor and health that will be passed on to offspring (Zahavi 1975; Hamilton and Zuk 1982; Andersson 1994; Able 1996; Setchell et al. 2009). Therefore, it is crucial to understand how and what information can be transferred from a signaler (usually a male) to a receiver (usually a female) during courtship (e.g., Andersson 1994; Bradbury and Vehrencamp 1998; Arnqvist and Rowe 2005; Searcy and Nowicki 2005).
Testudinidae represent a suitable group to investigate and test handicap theories and good genes models, since their courtship and copulatory behavior is elaborate, and based on a multiple signaling system involving visual, olfactory, tactile, and acoustic displays sorted in stereotyped sequences, which are largely shared among different species (Weaver 1970; Carpenter and Ferguson 1977; Sacchi et al. 2003; Galeotti et al. 2005b). The main courtship displays exhibited by male tortoises include bites and rams given to the carapace of females when attempting to convince them to accept mounting, and vocalizations produced during the mount. Interestingly enough, vocalizations seem to be strictly associated with mounting, because this is the main context in which tortoises vocalize (Carpenter and Ferguson 1977; Ernst and Barbour 1989; Galeotti et al. 2005a, 2005b). In addition, tortoises show a promiscuous mating system and females copulate with several different males during the same breeding season. Since males do not bring any direct benefit to females, such as nuptial gifts, territorial access, or parental care (Weaver 1970; Ernst and Barbour 1989; Kuchling 1999; Sacchi et al. 2003), their only possible contribution is related to genetic benefits to progeny. In this scenario, females can increase the overall fitness of their offspring by promoting sperm competition (Madsen et al. 1992; Birkhead et al. 1993; Yasui 1997). In addition, they may exert a postcopulatory cryptic choice (Eberhard 1996), thus raising the probability that eggs are fertilized by sperm from high-quality males.
Recent studies in marginated and Hermann's tortoises (Testudo marginata and Testudo hermanni) indicate a relationship between male courtship signals and their mounting success (Sacchi et al. 2003; Galeotti et al. 2005b, 2005c). In particular, 3 features of male calls (call rate, fundamental frequency, and duration) of marginated tortoises strongly correlated with body condition and male mounting success (Sacchi et al. 2003). In Hermann's tortoises, mounting success was positively correlated to the number of sexual interactions per hour, the number of bites given to the female during interactions, and to the call rate and frequency-modulation range. In addition, these courtship features reflect physiological attributes relevant to fitness, being associated with hematocrit and leukocyte concentration (Galeotti et al. 2005b). This suggests that courtship signals are condition-dependent traits potentially conveying information about the general quality and health of males in socio-sexual contexts, and females may use these multiple traits to choose high-quality partners (Galeotti et al. 2005b).
The Greek tortoise (Testudo graeca), like the other 2 species of Mediterranean Testudinidae native to Europe, maintains nonexclusive home ranges, and exhibits a promiscuous mating system with multiple copulations. Sexual dimorphism is noticeable, since males are smaller than females, possess longer tails, and have a concave plastron (Willemsen and Hailey 2003; Kaddour et al. 2008).
During the breeding season, usually in April–May and September–October (Willemsen and Hailey 2003), males follow females, encircling them, biting their limbs and ramming their carapaces, attempting to mount. When females are successfully mounted, mounting males emit a regular series of shrill vocalizations. As in other Testudinidae, sperm may be stored in female genital tracts for as long as 3–4 years (Swingland and Stubb 1985; Kuchling 1999), thus setting the optimal scenario for the evolution of sperm competition and/or female cryptic choice.
In this study, we investigated whether behavioral traits, and particularly acoustic signals, exhibited by male Greek tortoises during courtship reliably reflect aspects of male general health and body condition, and what effect these signals may have on male mounting success. According to the previous results obtained for marginated and Hermann's tortoises, we expected that Greek tortoise courtship signals are condition dependent and, consequently, males that signal at higher rates should also gain greater mounting success.
METHODS
We carried out this study during spring and summer 2002 and 2003, at the CARAPAX European Centre for Tortoise Conservation, located at Massa Marittima (Tuscany, central Italy). At this facility, 8000 individuals of several tortoise species reproduce in enclosures, in seminatural conditions and at high densities (Donato Ballasina, 2002, pers. comm.). Data were collected from a sample of 104 Greek tortoises (49 males and 55 females). Tortoises were maintained in 4 different large enclosures, 600 to 750 m2 (Table 1). The mean population density within enclosures was 4 individuals/100 m2 (range: 2.9–4.5), and the sex ratio was 1.05 males per female (range 0.35–1.75; Table 1). To recognize individuals, we marked them by painting a number on the carapace.
Courtship Observation and Male–Female Interactions Recordings
Tortoise courtship behavior consists of a complex sequence of displays. According to our previous observations, we identified 9 behavioral displays in the courtship of Greek tortoises, which are similar to those we described for marginated and Hermann's tortoises (see Sacchi et al. 2003; Galeotti et al. 2005b).
We observed male courtship by continuous sampling of a focal male courting a female, for a total of 144 hours, between 0900 and 1800 hours. During sampling, we recorded the occurrence and duration of each display performed by focal males. Overall, we recorded 90 interactions (49 in 2002 and 41 in 2003) from 21 male Greek tortoises (mean ± SE = 4.3 ± 0.8, range 1–13).
Morphological Measures
For each focal male, we measured 8 morphological features (in mm) including carapace length, width, and height; plastron length and width; and the length, width, and height of the head. We used weight (g) × (carapace length)−3 (Peterson 1996) as a measure of male body condition instead of the weight, because this latter characteristic is greatly influenced by carapace size. Finally, we calculated the mean ratio between male and female body size, computed across all male–female interactions.
Since morphological measures of carapace and plastron were highly intercorrelated (|rp| > 0.60, p < 0.001, n = 21, we used a principal component analysis (PCA) to reduce these 5 measures to a single “carapace size” component. PCA procedure extracted only one component (eigenvalue > 1), which accounted for 88% of the cumulative variance and was positively related to each of the 5 morphological variables (all 5 variables entered PCA with component loadings > 0.78). We used the same procedure to reduce to a single “head size” component the 3 head variables, which were also highly intercorrelated (|rp| > 0.59, p < 0.005, n = 21). PCA procedure extracted only 1 component (eigenvalue > 1), which accounted for 79.4% of the cumulative variance, and was positively related to each of the 3 morphological variables of the head (all 3 variables entered PCA with component loadings > 0.81).
The scores on both carapace size and head size components for each tortoise were used to examine their relationships with blood parameters, courtship intensity, call features, and mounting success.
Courtship Intensity Estimate
Courtship intensity was measured for each male using the frequency of sexual interaction per hour, i.e., the number of interactions per hour of observation, the mean number of bites and rams, and the call rate produced by males during each interaction. Call rate was quantified as the number of vocalizations emitted per 5 seconds of mount. Since male Greek tortoises bit females only in < 3% of interactions, while they rammed females in all interactions, the number of bites was excluded from the subsequent analyses.
Mounting Success Estimate
To estimate the mounting success of each male, we measured: 1) the total number of mounts performed during all interactions, 2) the mean number of mounts per interaction, and 3) the percentage of different females mounted during the breeding season. Female percentages were used because female availability differed among enclosures. Since these variables were highly intercorrelated (|rp| > 0.76, p < 0.0001, n = 21), we reduced them to a single “mounting success” factor by using a principal component analysis. PCA procedure extracted only 1 component (eigenvalue > 1), which accounted for 70% of the cumulative variance, and was positively related to each of the 3 variables (they entered PCA with component loadings > 0.52). The scores of the mounting success component for each tortoise were used to investigate relationships with body condition, blood parameters, courtship intensity, and call features.
The mean scores of mounting success and the mean values of courtship variables among all males within each enclosure did not significantly correlate with the corresponding values of population density and sex ratio (all |rp| < 0.44, p > 0.10). Therefore, no relationship could be found between composition of experimental groups within enclosures and courtship behavior of male tortoises, and consequently, data collected from different enclosures have been pooled.
Call Recording and Sonographic Analysis
We recorded tortoise vocalizations by using a Sony TCD-D7 DAT tape recorder connected to a SHURE shotgun microphone with a hypercardioid capsule. Recordings were made from close distances (5–15 m) and were undertaken without disturbing the sexual interactions.
We collected 106 recordings from 14 of the 21 male Greek tortoises that courted females during behavioral observations (6 in 2002 and 8 in 2003), obtaining on average 7.6 ± 1.8 SE recordings for individual (range 2–25). Seven males did not achieve any mountings, and thus it was impossible to record vocalizations from them.
Only recordings with high signal to noise ratios from each male (minimal background noise and high sound intensity) were analyzed by using AVISOFT SAS-LAB pro software (Specht 1993). Vocalizations in the 0–4 kHz frequency range were analyzed using a sampling rate of 8000 sample/s, bandwidth 80 Hz, frequency resolution 30 Hz, and time resolution 16 milliseconds. Overall, we analyzed 667 calls (50.7 ± 25.3 SE calls per male, range 3–169). For each call, we selected and measured directly on the screen the following structural features (Fig. 1): the mean duration in milliseconds, the mean time interval in seconds between calls, the mean fundamental frequency in Hz, and the range of frequency modulation in Hz (thereafter ΔF).
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Hematological Parameters
Blood samples were taken from the jugular vein of 13 males whose vocalizations had been recorded during behavioral observations and collected in 75-µl heparinized capillary tubes and smears. All males were sampled for blood only once at the end of observations (late June). A single sample provides reliable values, since blood condition of individuals did not vary significantly within the same season (Christopher et al. 1999; Mader 2000).
To measure hematocrit (hereafter packed volume cells, PVC), capillary tubes containing blood samples were centrifuged for 5 minutes at 7826 × g, and PVC was expressed as the volume of the part of the capillary occupied by blood cells × (blood volume in the capillary)−1 (Saino et al. 1997; Ots et al. 1998).
Leukocytes and red blood cells were counted after blood smears were air-dried, fixed in methanol, and stained using the May-Grunwald/Giemsa method. Blood smears were randomly scanned at ×630 magnification under oil immersion following standard routines (Canfield 1988; Latimer and Brienzle 2000). In each microscope field (mean = 21 ± 1 SE; range 14–34 fields) red blood cells were counted and leukocytes classified as lymphocytes, monocytes, eosinophils, heterophils, and basophils. In each smear we counted 150–200 leukocytes and the corresponding red blood cells. This allowed the number of leukocytes of each type per 10,000 red blood cells to be calculated. Even when fewer leukocytes were counted, this method has been shown to give significantly repeatable within-blood smear measures of leukocyte concentration (Saino et al. 1995). Finally, heterophil to lymphocyte ratios (H∶L) were calculated based on cell count results. H∶L is a reliable measure of the stress response, and there is also evidence that this ratio reflects immunosuppression (Case et al. 2005; Davis et al. 2008).
Statistical Analysis
One-way analysis of variance (ANOVA) and discriminant function analysis (DFA) were used to check for the possible differences in call features among males, and to identify those variables that better distinguish different individuals. Multiple regression analyses (MRA stepwise forward method) were used to analyze relationships between courtship variables and call features and morpho-physiological parameters, using the former as dependent variables and the latter as predictors. The same analysis was then performed to test the separate and combined effect of morpho-physiological variables (body condition, carapace and head size, male/female size ratio, PVC, and leukocyte concentrations), courtship measures, and call features on male mounting success. To achieve normality, all counts were log transformed (Sokal and Rohlf 1980). All statistics were 2-tailed and were performed with SPSS 15.0 statistical package. Unless otherwise stated, values are means ± SE.
RESULTS
Call Features Description
During each mount, male Greek tortoises emit a sequence of monotonous calls that sound like moans or groans, with regular intervals between them (Fig. 1). They were highly modulated in frequency and showed a clear harmonic structure, with up to 19 harmonics observed in some sonograms; call features were not intercorrelated (Pearson correlation coefficients, all p-values > 0.12). Mounting calls emitted by male Greek tortoises showed intermediate features between marginated and Hermann's tortoises (see Table 2 for details).
All calls differed significantly among individuals (Table 3). The 4 functions generated by DFA were highly significant, the first and second accounting for 94% of the variance. Fundamental frequency and call duration appeared to be the most important variables in discriminating among individuals' vocalizations. However, the high values of Wilks λ for each function indicated that they had low discriminant efficiency. Indeed, only 53.4% of vocalizations of male Greek tortoises were correctly assigned to the signaller. Thus, mounting calls hardly conveyed information about signaller identity.
Relationships Between Morphological Variables and Blood Parameters
Lymphocytes constituted the bulk of white blood cells (> 55%), while basophils were the minority (< 3% of counted leukocytes, Table 4) and were thus excluded from subsequent analyses. H∶L was 0.35 ± 0.04 on average (range 0.16–0.65; Table 4) and the mean PVC value was 0.27 ± 0.01 (range 0.20–0.34; Table 4).
Neither PVC nor H∶L of male Greek tortoises correlated with body condition or with carapace or head size, and no multiple regression analyses produced any significant models. Among leukocytes, only monocytes were related to morphological variables, significantly decreasing in relation to carapace size (F1,11 = 5.318, p = 0.042, R2 = 0.33, b = −0.57). Since monocytes tend to increase in sick tortoises (Work and Balazs 1999; Jacobson 2007), this finding may suggest that smaller males were in worse health condition than larger males.
Relationships Between Morpho-Physiological Measures, Courtship Variables, and Call Features
We did not detect any significant relationship between morphological measures and courtship variables. Among blood parameters, PVC was positively related to all courtship variables, particularly to the call rate (F1,10 = 9.800, p = 0.011, R2 = 0.49, b = 0.70; Fig. 2) and to the number of rams (F1,11 = 5.478, p = 0.039, R2 = 0.33, b = 0.58; Fig. 3), while the relationship with frequency of interaction per hour was only marginally significant (F1,11 = 4.758, p = 0.052, R2 = 0.30, b = 0.55). These findings suggested that males displaying at higher courtship rates were also in better physiological condition. No other blood parameters were related to courtship variables.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Call features did not correlate with male morpho-physiological measures, although head size had a marginal negative relation with the fundamental frequency (F1,12 = 4.200, p = 0.063, R2 = 0.26, b = −0.51). Thus, males with larger heads tend to utter lower-frequency calls.
Among call features, only the call duration was negatively related to the number of rams (F1,12 = 10.607, p = 0.007, R2 = 0.47, b = −0.68), suggesting that males displaying higher ramming rates also emitted shorter calls.
Effects of Morpho-Physiological Measures, Courtship Variables, and Call Features on Mounting Success
Among morpho-physiological measures, only PVC correlated positively with male mounting success (F1,11 = 6.156, p = 0.031, R2 = 0.36, b = 0.60).
Mounting success was positively related to the number of rams (F1,19 = 21.167, p < 0.001, R2 = 0.53, b = 0.73; Fig. 4) and to the frequency of interaction per hour (F1,19 = 16.059, p = 0.001, R2 = 0.46, b = 0.68; Fig. 5), but not to call rate.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Call features were in general unrelated to male mounting success, except for call duration, which appeared to negatively affect it (F1,12 = 5.606, p = 0.036, R2 = 0.32, b = −0.56; Fig. 6).
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0840.1
Taken together, all these findings suggested that males in better physiological condition, courting females more intensely, and emitting shorter calls attained higher mounting success than other males.
Finally, an overall multiple regression analysis carried out on mounting success of male Greek tortoises using morpho-physiological measures, courtship variables, and call features as independent variables produced a highly significant model (F3,10 = 7.078, p = 0.008), based exclusively on frequency of interaction per hour (b = 0.37) and rams (b = 0.42), which explained 68% of the total variance (R2).
DISCUSSION
This study showed that courtship variables, including call rate, exhibited by Greek tortoises were condition-dependent signals, which are associated with male mating success in this species. The results of this study are thus in accordance with previous findings on marginated and Hermann's tortoises (Sacchi et al. 2003; Galeotti et al. 2005b).
First, we found that PVC was strongly related to all courtship signals. Specifically, males performing a higher number of interactions, giving more rams to females during courtship, and vocalizing at higher rates during mounting, showed higher PVC values. Actually, hematocrit is a reliable indicator of the aerobic capacity of individuals, that seemingly influences fitness through the effects on locomotor performance and courtship endurance (Chappel et al. 1995, 1997; Peterson 2002). In addition, low values of PVC are indicative of anemia induced by bacterial infections and gastrointestinal disorders, including parasitism and hemorrhage (Dein 1986). Hematocrit may also be related to nutritional deficiencies of protein and minerals such as iron or copper (Sturkie and Griminger 1986).
Secondly, the PVC value, the frequency of interactions, and the number of rams were all positively associated with mounting success of male tortoises. Therefore, by choosing males providing highly energetic displays during courtship, female Greek tortoises are also selecting healthy males (Alexander 1975; Borgia 1979; Borgia et al. 1985).
In our previous studies on marginated and Hermann's tortoises we showed that call features (call rate and duration) strongly correlated with various aspects of male quality (weight and body size, general health condition) and with male mounting success (Sacchi et al. 2003; Galeotti et al. 2005b), suggesting a possible adaptive role of vocalizations in conveying crucial information to females during courtship. The communication function of call rate is now confirmed also for the Greek tortoise, although this call feature did not apparently relate to male mounting success in this species.
Finally, we did not find any significant relationship between call duration and morpho-physiological measures, as by contrast occurred in the Hermann's tortoise (Galeotti et al. 2005b), although call duration tended to decrease with increasing PVC and to increase with monocytes. This class of leukocytes is related to chronic infections and increases with antigenic stimulation (Work and Balazs 1999; Work et al. 2001; Jacobson 2007). In addition, monocytes are involved in granuloma and giant cell formation and are specifically associated with responses to bacterial infections and metazoan parasites (Campbell 1996). Thus, males in better condition tended to emit shorter calls and, in fact, we found that call duration negatively affected mounting success. These tendencies were consistent with our previous findings (Galeotti et al. 2005b), and the lack of significant relationships between call duration and blood parameters might be due to the lower number of Greek tortoise individuals analyzed with respect to Hermann's tortoises (13 vs. 49).
In conclusion, this study supports and generalizes the hypothesis that courtship displays and vocalizations exhibited by male Testudinidae are condition-dependent signals that have probably evolved via female choice for healthy and resistant males, which are able to display at higher rates during courtship.
Sonograms of typical mounting calls of the Greek tortoise, showing the acoustic parameters considered in this study. D: duration, I: interval, FF: fundamental frequency, ΔF: range of frequency modulation.
Relationship between hematocrit and call rate of male Greek tortoises.
Relationship between hematocrit and the number of rams given by male Greek tortoises to females during courtship.
Relationship between the number of rams given by male Greek tortoises to females and their mounting success.
Relationship between frequency of interaction per hour and mounting success of male Greek tortoises.
Relationship between duration of call emitted during mounting by male Greek tortoises and their mounting success.
