Comparative Quantification of Intralocational, Interlocational, and Interspecific Variability in Stag Beetles (Coleoptera: Lucanidae) and the Questions of Phenotypic Plasticity and Species Selection

Abstract

Stag beetles show a great intraspecific variation. The morphological variation of a species at a single location is a reflection of the species developmental plasticity (DP) and that across locations is a measure for the species adaptive plasticity (AP). Variations of external morphology (developmentally highly variable) and reproductive organ (developmentally highly stable) were studied in each of 514 locations comprising 187 species to determine intralocational, interlocational, and interspecific variability. Dealing with naturally occurring variation, extra attention was paid to the normality of each data set and the validity of partitioning each variation component. With this, the comparative quantification of DP and AP and its presentation as broad-sense heritability were made possible. The results indicate that intraspecific evolution begins with the diversification in genitalia morphology through selection on individual phenotypes as the traditional theory dictates while DP of external morphology sustains the morphological robustness of species. More significantly, there was a highly significant difference in DP of external morphology among species and the interspecific, broad-sense heritability of DP in body and mandible length was very high, indicating that DP is firmly a genetic trait on its own. Furthermore, DP (of body length) influences the species geographic range (GR) more significantly than the species mean or other measures. DP in external morphology and GR being a highly heritable emergent trait and an emergent property, respectively, at the species level, these findings suggest that DP plays a significant role in interspecific evolution through species selection rather than selection on individual phenotypes.

RESUMEN

Existe una gran variación intraespecifica en los “Ciervos Volantes”. La variación morfológica de una especie en un sólo sitio refleja la Plasticidad de Desarrollo (DP) de la especie, mientras que la variación morfológica a través de varios sitios es una medida de la Plasticidad Adaptiva (AP) de la misma. Se estudiaron la variación de la morfología externa (altamente variable) y de los órganos reproductivos (muy estables) de 187 especies en 514 sitios para determinar la variabilidad intrasitio, entre-sitios e interespecífica. Para manejar la variación natural se confirmó la normalidad de cada conjunto y la validez de la separación de cada uno de los componentes de la variación. Esta confirmación permitió la cuantificación de la DP, AP y la estimación de la heredabilidad en sentido amplio. Los resultados indican que la evolución intraespecifica comienza con la diversificación de la morfología del aparato genital a través de la selección fenotípica individual según la teoría tradicional, mientras DP de la morfología externa sostiene la robustez morfológica de las especies. Los hallazgos destacan la alta significancia de la DP de la morfología externa entre especies y de la heredabilidad interespecifica en sentido amplio de la DP de la morfología externa del largo del cuerpo y de la mandíbula; sugiriendo definitivamente que la DP es sólo un rasgo genético. Además, la influencia de la DP (del largo del cuerpo) sobre el Rango Geográfico (GR) es mayor que el promedio de la especie y las otras medidas. La DP de la morfología externa es un rasgo emergente y el GR una propiedad emergente al nivel de especies. Ésto sugiere que DP juega un papel clave en la evolución interespecífica a través de la selección por especie en lugar de la selección sobre fenotipos individuales.

The changes in the environment are increasingly rapid as we move into the Anthropocene age (Crutzen 2006). The capacity of the individuals of any species to adapt rapidly to changing conditions is critical for survival under fast-changing scenarios. Phenotypic plasticity provides individuals or species with the ability to rapidly adapt to new circumstances, and hence should act as measure of protection against extinction (Pigliucci 2005, Pfennig et al 2010, Storz et al 2010, Moczek et al 2011).

Phenotypic plasticity (or environmental responsiveness) is the ability of an organism to react to an external environmental input with a change in form (West-Eberhard 2003, among many similar definitions [Debat and David 2001]). Throughout nearly the entire period of incorporating genetics theory into the synthetic theory of evolution, many have considered phenotypic plasticity a nuisance because it obscures the true expression of genotypes (e.g., Falconer 1952). However, with the recognition of the importance of development and environmental effects on phenotypes, phenotypic plasticity is increasingly recognized as having a significant function in morphological evolution (West-Eberhard 2003, Piesma and Drent 2003, Pigliucci 2005, Gilbert and Epel 2009, Pfennig et al 2010, Moczek et al 2011). However, the genetics of phenotypic plasticity and its role in species evolution have not been clearly established.

When dealing with the populations of a species, phenotypic plasticity can be divided into two categories—developmental plasticity (DP) and adaptive plasticity (AP). DP is the responsiveness of an individual to variation in the microenvironment, usually within the same location. This is caused by the interaction of the species developmental program with the microenvironmental effects. AP is the species responsiveness to macroenvironmental differences, usually associated with each particular location. AP is considered to be composed of an irreversible component (genetic) and a reversible component (an extension of DP; based on the observations of Debat and David 2001 and Piersma and Drent 2003).

While a developmental program is embedded in every member of species, the DP seen in a population is not reducible to the phenotype of a single individual in the population. If there is a significant difference in DP among species, that difference cannot be reduced to selection on individual phenotypes (Piersma and Dret 2003). This inevitably leads us to conclude that there is selection at the population level. This may be species selection.

Species selection can be defined simply as high-level selection that is not reducible to causes operating at conventional lower levels of selection of individual phenotypes (Gould 2002, p. 661). It can be further defined as selection resulting from heritable differences in net diversification rates among species, when the causal basis of these rate differences is not a simple function of differential survival and reproduction at the individual level (Jablonski 2008, Rabosky and McCune 2010). In early days of the modern Synthesis, species selection was treated only as a theoretical consideration, not an actual entity (Fisher 1958). Some real cases were later recognized, but its occurrence was very rare (Allmon 2009) and, as such, the concept was regarded as of little evolutionary importance (Dietrich 2010). The paucity of evidence to support higher level selection has been attributed to both an overly strict definition of the phenomenon (Gould 2002, p. 662) and deficiencies in the search for evidence of the process (Gould 2002, p. 710, Erwin 2010).

Thus, 1) How heritable is plasticity?, 2) How is plasticity related to species selection?, and 3) How frequent is species selection in nature? are the questions to be answered. To tackle these issues, an extensive number of samples from genetically related families raised or collected in many different environments is necessary. This depends on low-tech, tedious observations that require long periods of dedication by those involved, “a combination that is sometimes difficult to justify to funding agencies when compared with more ‘high-tech’ science” (Pigliucci 2005).

Japan has a culture of marveling at many aspects of insects (Kawahara 2007): there are hundreds of ardent collectors and investigators of stag beetles (Coleoptera: Lucanidae) who dedicate untold hours to their collection and description. This cadre of amateur biologists was the core from which it was possible to build a highly comprehensive species-level taxonomy and a network survey of geographic distribution in this family.

Stag beetles show a great intraspecific variation (Leuthner 1885, Arrow 1951, Otte and Stayman 1979, Mizunuma and Nagai 1994). Fujita (2010), after summarizing the experiences of the hundreds of stag beetle breeders as well as the results from dozens of scientific studies, concluded that body size is determined by both the quantity and quality of nutrition available during the larval stage and also the length of the larval stage that is influenced by the temperature regime. The size of male mandibles is critically influenced by the temperature and the length of prepupal stage, in addition to the same nutritional and larval stage length factors as in the case of body size. Intraspecifically, there is little genetic difference in this developmental pattern. The size of genitalia is little influenced by external and internal factors (Eberhard et al 1998, Kawano 2004). Consequently, genitalia length is developmentally relatively stable, body length is highly variable, and mandible length is extremely variable (Kawano 2000, 2006).

The external morphology of beetles does not change once they emerge from the pupa. Fujita (2010) and the experience of many breeders indicate that genetic variation within a species from a single location is small in stag beetles. Hence, the adult morphological variation of a species within a location must be principally due to expression of the developmental program embedded in each member of a given species, and hence can be taken as measure of the species DP. Similarly, assuming that the magnitude of effects unrelated to adaptation, such as genetic drift, is comparatively small, the variation across different locations is a measure of the species AP.

It is difficult to design and carry out experiments to study the specific mechanisms of plasticity, adaptation, and species selection. Instead, based on the existence of extensive data accumulation of naturally occurring variation in external morphology and reproductive organ, I analyzed the hierarchical relationships of DP, AP, and species geographic range (GR) to answer the questions regarding plasticity and species selection in a quantitative, statistical manner.

Materials and Methods

Sample Collection

The sample collection started in 1965 in Sapporo, Japan, modestly by myself catching stag beetles in the forest. It continued intermittently to the present 50 yr, increasingly involving more collaborators and expanding to covering most stag beetle areas in Asia. In most locations, samples were collected within an area of radius <50 km. Each of these locations was assigned as one point within the GR, which often exceeds a radius of >1,000 km for a single species. Throughout the whole period of collection, careful attention was paid to collect all the available samples as representatives of naturally occurring population without bias to particular types.

From this collection, four large genera, i.e., Lucanus, Odontolabis, Prosopocoilus, and Dorcus, were used for the study. Those species that provided >15 male individuals per location for the measurement of body and mandible length and >10 male individuals per location for the measurement of genitalia length were used for the analysis. There were a total of 12,176 individuals in 514 species-locations comprising 187 species for body and mandible length measurement and 4,853 individuals, available for destructive sampling, in 340 species-locations comprising 116 species for genitalia length measurement resulted. Detailed data by species are given in Supplementary Data (online only).

Species Identification

I used solely a monograph by Fujita (2010), not only because this monograph was the most comprehensive of the kind but also because possible confusion could be minimized by referring to one single central document.

Geographic Range

One highly useful feature of this monograph is the detailed recording of locations where each species occurs, by referring to the actual specimens and observations of 115 individual collectors (including myself) and five public museums. By the nature of data, this type of information cannot be perfect (missed locations and uneven area specification of the locations); yet, it is one of the most reliable and usable data sets dealing with insects of low to medium mobility. The calculation of GR of each species was made by simply multiplying the latitudinal distance between the northernmost and southernmost spots with the longitudinal distance between the westernmost and easternmost spots. The actual work was conducted by determining the distance in 100 km on a large map (1: 15 millions). I assigned 100 km × 100 km to the species whose distribution was given by one single location, following the fact that there are several relatively common species that are well known to be occurring only in a small area of no more than 100 km × 100 km but not limited to one spot within the area, such as L. koyamai (Akiyama et H. Hirasawa, 1990), O. baderi (Bomans et Bartolozzi, 1990), P. taroni (Lacroix et Ratti, 1983), or D. negrei (Lacroix, 1978). The whole scheme may be fairly “low-tech”; yet, the difference between small and large GR is sufficiently large to offer robust statistical comparisons.

Morphological Measurement

I measured the specimens with a slide-caliper to the nearest 0.1 mm for body length (the distance from the front of the head [excluding mandibles] to the tip of the elytron along the center line of the body) and mandible length (from the base to the tip). Using the specimens available for destructive sampling, I extracted the male copulating organ (penis; the part chitinously hardened inside the male copulating organ; see the electronic version of Kawano (2004) for the detailed figures of the part actually measured) from fresh or water-softened samples and measured the length of the straightened (manually by fingers or tweezers) penis under a low-magnifying microscope with a slide-caliper to the nearest 0.1 mm (Kawano 2006).

Test of Normality

The biological measures may not be normally distributed. In addition, the standard deviations (SDs) and the coefficient of variability (CV) calculated from these measures may also deviate from a normal distribution. To determine the degree of deviation from normal distribution, two indicators—W, the sample-number-free measure for normal distribution, and P, the sample-number-dependent measure for the probability the sample population coming from normal distribution, were computed using the Shapiro–Wilk test (Shapiro and Wilk 1965). I subjected all of the more than two thousand data sets used in this study to this test.

Statistical Analysis

In a standard analysis of variance (ANOVA) for a variation scheme with a large number of entries with a varying number of samples in each entry (e.g., Goulden 1952, Snedecor 1956), mean squares for the entry give an estimate for σ²e + n_oσ²E, where σ²e is the error component given by the variance of samples within entry and σ²E is the variance of entries. n_o is given as [1/(α − 1)](T − Σn_i²/T), where T is the total number of samples and α is the number of entries and n_i is the number of samples in the ith entry (e.g., Snedecor 1956).

In ANOVA for intraspecific, interlocational variation, locational difference of the same species is the entry and the individual variation of the same species within location is the error component (ANOVA–I). The intralocational variation is a measure of the interlocational difference. Broad-sense heritability for intraspecific, interlocational variation in each species is given by the simple, classic equation: h²= σ²G/(σ²G + σ²e) (e.g., Allard 1960), where σ²G corresponds to the intraspecific, interlocational variance and σ²e to intraspecific, intralocational variance, thus separating the locational effect from the full intraspecific variation. σ²e is composed of the species reaction to microenvironmental difference (DP), the genetic component, and random error. In traits that are developmentally highly variable but morphologically uncomplicated, such as body and mandible length, the proportion of random error in σ²e is comparatively low and the genetic variation within a species within a given location is expected to be small in most species. σ²G represents AP; however, σ²G includes both irreversible (genetic) and reversible (an extension of DP) components and, as such, may not truly merit the term “heritability.” This point will be further discussed with the results obtained. In order to obtain meaningful estimates for σ²G, only those species with data from more than either 5 or 10 locations were used for analysis.

Extending the same scheme to interspecific variation in each genus, species difference is the entry and the intraspecific, interlocational variation is the error component (ANOVA-II). In this case, the interlocational variation is the measure of interspecific difference. This procedure offers a test of species difference not only in terms of the species mean but also of the intraspecific variability per se, which is in turn a measure for DP. SD and CV in each location were used as the measure and, as such, the interspecific difference of DP can be tested and quantified. In this scheme, the calculated broad-sense heritability may be slightly underestimated or overestimated because the error component includes a small fraction of genetic variation and the interspecific variation includes nongenetic variation caused by higher order interactions. This point will be further discussed with the results.

Glossary for Statistical Measures

So as to better understand the hierarchical data presentation and the variation of variation scheme, the terminology is clarified in this section. Throughout the study, species mean is calculated as the mean of all location means in each trait in each species (Tables 1, 4, 5, 6, and 9). This avoids the possible overrepresentation of certain locations with a large number of samples, which may occur when the overall mean of all individuals or weighted means are used. Similarly, the genus mean is given as the mean of all the species means (Table 4).

Table 1.

Mean body, mandible, and genitalia length and their mean variation

Genus	Body length			Mandible length			Genitalia length
	Mean (mm)a	SD (mm)b	CV (%)b	Mean (mm)a	SD (mm)b	CV (%)b	Mean (mm)a	SD (mm)b	CV (%)b
Lucanus	33.5	4.5	12.6	15.7	4.0	23.4	5.55	0.31	5.5
Odontolabis	39.1	4.8	11.9	12.0	4.4	34.6	5.83	0.29	4.9
Prosopocoilus	28.4	4.3	14.9	11.6	4.1	34.8	4.77	0.31	6.5
Dorcus	29.8	4.5	14.4	9.8	3.1	30.0	4.74	0.30	6.2

^a Mean of species means.

^b Mean of DP measures.

Species AP measures (F, SD, CV, and h²) were obtained from δG or δ²G of ANOVA-I (Tables 2, 3, 7, and 8).

Table 2.

F (and d.f. and P) values in ANOVA-I for interlocational variation (a measure for AP)

Species	F and P values () in ANOVA for intraspecific, interlocational variation
	Body length (B)	Mandible length (M)	Genitalia length (G)
D. titanus	7.94 (31, 624; < 0.001)	9.11 (31, 624; < 0.001)	39.51 (20, 326; <0.001)
D. reichei	14.95 (19, 536; <0.001)	9.13 (19, 536; <0.001)	77.89 (18, 288; <0.001)
D. taurus	3.12 (13, 384; <0.001)	2.00 (13, 384; 0.020)	21.37 (13, 250; <0.001)
D. antaeus	2.29 (13, 299; 0.007)	3.28 (13, 299; <0.001)	9.11 (9, 115; <0.001)
D. parryi	2.29 (11, 260; 0.011)	1.82 (11, 260; 0.051)	19.80 (9, 188; <0.001)
O. cuvera	4.30 (19, 470; <0.001)	1.52 (19, 470; 0.074)	12.09 (17, 211; <0.001)
O. mouhoti	1.48 (13, 376; 0.122)	0.94 (13, 376; 0.511)	9.51 (9, 189; <0.001)
O. dalmani	5.64 (12, 359; <0.001)	11.47 (12, 359; <0.001)	10.02 (11, 143; <0.001)
P. astacoides	6.51 (14, 428; <0.001)	4.45 (14, 428; <0.001)	2.38 (13, 193; 0.006)
P. giraffa	2.14 (9, 182; 0.028)	2.48 (9, 182; 0.011)	8.88 (9, 111; <0.001)
P. buddha	8.02 (9, 180; <0.001)	2.64 (9, 180; 0.005)	26.75 (9, 103; <0.001)
D. curvidens	3.98 (9, 238; <0.001)	3.43 (9, 238; 0.001)
O. siva	0.52 (9, 216; 0.859)	1.52 (9, 216; 0.142)

Table 3.

Heritability (h²) for intraspecific, interlocational variability (broad-sense intraspecific heritability for AP)

Species	h2 for Body length (B)	h2 for Mandible length (M)	h2 for Genitalia length (G)	Number of locations for B & M measurement	Number of locations for G measurement
D. titanus	0.254	0.284	0.705	32	21
D. reichei	0.340	0.231	0.826	20	19
D. taurus	0.076	0.037	0.529	14	14
D. antaeus	0.056	0.092	0.400	14	10
D. parryi	0.055	0.037	0.492	12	10
O. cuvera	0.121	0.025	0.432	20	18
O. mouhoti	0.018	0.000	0.328	14	10
O. dalmani	0.142	0.273	0.413	13	12
P. astacoides	0.162	0.108	0.089	15	14
P. giraffa	0.056	0.072	0.654	10	10
P. buddha	0.290	0.087	0.695	10	10
D. curvidens	0.109	0.091		10
P. siva	0.000	0.023		10

Table 4.

Variations of species mean and intraspecific, intralocational variation

Genus	Body length; CV (%) of			Mandible length; CV (%) of			Genitalia length; CV (%) of
	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV
Lucanus	17.6	39.2	26.8	39.6	56.7	36.5	11.3	26.3	22.7
Odontolabis	29.6	39.8	19.7	33.0	52.7	32.9	29.6	38.8	19.8
Prosopocoilus	25.2	35.2	17.6	46.1	49.2	23.3	27.6	27.9	15.9
Dorcus	30.9	48.3	25.6	38.5	48.7	26.0	29.6	35.8	20.0

For species DP measure in SD, (1) SD of whole individuals within species, (2) mean of all intralocational SDs within species, (3) weighted mean of intralocational SDs within species, (4) δe of ANOVA-I (SD of the population where all the locational means are adjusted to the same value), were calculated. While ① (1) represents the total phenotypic variability of species, it is a mixture of DP and AP, and the actual values were nearly always greater than the others. (2) may suffer from overrepresentation by locations of a small number of samples with skewed random effect. (3) and (4) are nearly of similar meaning. Actual computation gave very similar values for (3) and (4), and accordingly the subsequent analyses gave very similar results. I selected (4) for a possible better representation of intralocational variation free from locational effect in this paper (Tables 1, 4, 5, 6, 7, and 9, and Figs. 2, 3, and 5). For species DP measure in CV, this SD and species mean (mean of all location mean) were used (Tables 1, 4, 5, 6, 7, and 9, and Figs. 2, 3, and 5).

Table 5.

F (and d.f. and P) values in ANOVA-II for interspecific variation

Genus	Body length			Mandible length			Genitalia length
	Value	SD	CV	Value	SD	CV	Value	SD	CV
Lucanus	15.65	23.67	12.94	17.43	18.71	11.74	5.12	1.45	1.80
	(37, 23;	(37, 23;	(37, 23;	(37, 23;	(37, 23;	(37, 23;	(20, 20;	(20, 20;	(20, 20;
	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	0.207)	0.099)
Odontolabis	57.11	15.24	7.37	7.60	6.25	8.02	79.86	2.66	0.91
	(37, 99;	(37, 99;	(37, 99;	(37, 99;	(37, 99;	(37, 99;	(29, 69;	(29, 69;	(29, 69;
	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	0.001)	0.601)
Prosopocoilus	37.13	17.00	2.80	28.77	19.93	4.76	95.83	1.86	0.74
	(52, 74;	(52, 74;	(52, 74;	(52, 74;	(52, 74;	(52, 74;	(30, 54;	(30, 54;	(30, 54;
	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	0.018)	0.812)
Dorcus	21.54	19.98	7.40	9.49	8.38	3.10	27.12	13.69	1.72
	(57, 129;	(57, 129;	(57, 129;	(57, 129;	(57, 129;	(57, 129;	(33, 80;	(33, 80;	(33, 80;
	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	<0.001)	0.026)

Table 6.

Heritability (h²) for interspecific variability (broad-sense interspecific heritability for species mean and DP measures)

Genus	Body length; h2 for			Mandible length; h2 for			Genitalia length; h2 for
	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV	Species mean	Intraspecific, intralocational SD	Intraspecific, intralocational CV
Lucanus	0.903	0.935	0.883	0.912	0.918	0.872	0.684	0.205	0.297
Odontolabis	0.941	0.804	0.647	0.657	0.602	0.669	0.965	0.367	0.000
Prosopocoilus	0.939	0.873	0.436	0.923	0.890	0.617	0.973	0.238	0.000
Dorcus	0.869	0.860	0.678	0.732	0.705	0.408	0.894	0.806	0.243

Table 7.

Correlation between DP and AP measures (correlation coefficient, [d.f., P])

Correlation between		Body length	Mandible length	Genitalia length
Species DP measure	Species AP measure
Among species with >10 locationsa
SD	SD	0.166 (11, 0.587)	0.228 (11, 0.453)	0.431 (9, 0.186)
CV	CV	0.544 (11, 0.055)	−0.177 (11, 0.563)	0.511 (9, 0.108)
Among species with >5 locationsb
SD	SD	0.309 (25, 0.117)	0.269 (25, 0.174)	0.613 (16, 0.007)
CV	CV	0.527 (25, 0.005)	0.017 (25, 0.932)	0.544 (16, 0.020)

^a Thirteen species for body and mandible length and 11 species for genitalia length.

^b Twenty-seven species for body and mandible length and 18 species for genitalia length.

Variation of variation was given as CV of DP measures (SD and CV) in Table 4. It is also handled as F values of ANOVA-II for DP measurers (Table 5) and h² of DP measures given from ANOVA-II (Table 6). In Fig. 3, the magnitude of interspecific variation of DP (variation of variation) is compared with that of AP.

GR was given as log₁₀(1,000 km²) (Table 9 and Fig. 5).

Results

Test of Normality

The intraspecific, intralocational variation (Supplementary Data [online only]) in body and genitalia length showed near normal distribution in most cases, indicating that the variation at this level can be used as: 1) a very reliable measure for DP; and 2) the error basis for the significance test at the next level (ANOVA-I) as well as 3) for the subsequent computation of intraspecific heritability. The same variation in mandible length had somewhat lower W and P values as expected from the highly allometric nature of mandible morphology; yet, the deviation was not statistically significant in most cases, suggesting that the same statistical treatments as in body length would be acceptable.

The intraspecific, interlocational variation (Supplementary Data [online only]) showed near normal distribution in all the three traits. Especially encouraging were the high W and P values in SD and CV, not only because these are the measures for the interlocational variation but also because they constitute the error basis for the significance test of the interspecific difference in DP (ANOVA-II) and the subsequent estimate of interspecific heritability of DP.

The interspecific variations (Supplementary Data [online only]) in species mean and SD showed somewhat low W and P values; yet, they did not highly significantly deviate from normal distribution in most cases. This suggests that the significance test for the interspecific effect as well as the estimation of interspecific variance and the subsequent computation of heritability were acceptable. Highly encouraging is the very high W and P values of CV, indicating that the interspecific variation of CV essentially follows normal distribution and the following estimation of heritability is conceptually highly appropriate. The use of CV is the core of the quantitative analysis and presentation of DP in this paper.

The variations of species GR given as km² showed very low W and P values, indicating that they are not normally distributed. At the same time, the logarithmic transformation of the same measure gave W and P values acceptably high for the proper probability test of the correlations involving GR (Supplementary Data [online only]). Besides, GR given in logarithmic scale would more naturally represent our perception of the effects of distance, as the area encompassed within the geographic range increases with the square of the radius.

Mean Variability

Intraspecific, intralocational variation, given as CV, was in general low for genitalia length, high for body length, and extremely high for mandible length (Table 1). Figure 1 offers a visible example of the highly stable nature of genitalia length, the high variability of body length and the extremely variable length of the mandible within species. The great variability of body and mandible morphology is maintained throughout the vast geographic range of the species (4,000 latitudinal km × 5,750 longitudinal km), a clear sign that the intraspecific variability itself is a species-level trait. The genitalia size diversifies in different locations, but its variability stays small and unchanging throughout the geographic range.

Fig. 1.

Intraspecific, intralocational variation of adult male body, mandible, and genitalia (extracted from the body; shown below) of P. giraffe in four locations.

Fig. 2.

Comparison among intralocational, interlocational and interspecific variations.

Fig. 3.

Comparison of the magnitude of AP with the interspecific variation of DP (variation of variation).

Magnitude of Variations at Three Levels

Figure 2 presents the relative magnitude of 1) intraspecific, intralocational variation (all species mean of DP measure in CV), 2) intraspecific, interlocational variation (mean of AP measure in CV among species with >10 locations), and 3) interspecific variation (all species CV of species means) in the three traits studied. This visually offers the basis for discussing the validity of separating the variation into the distinct components such as DP, AP and interspecific variation.

Variation of Variation

Figure 3 demonstrates the much higher magnitude of “variation of variation” (given as interspecific CV of DP measure in CV) compared with the mean magnitude of AP in the three traits. This similarly offers a basis for discussing the interspecific variation of intraspecific, intralocational variability (DP) relative to that of intraspecific, interlocational variation (AP).

Heritability of AP

F and P values of ANOVA-I indicated that the effect of location is highly significant in genitalia length in most species (Table 2). Thus, heritability estimated with ANOVA-I (the relative magnitude of interlocational variation to intralocational variation) for genitalia length was high in most species (Table 3). The heritability estimated in a similar manner for body and mandible length was much lower and not statistically significant in most species.

DP

The intralocational variation for body and mandible length was much larger than the interlocational variation (Figs. 1 and 2) Assuming from the breeding experiences (e.g., Fujita 2010) that genetic variation within a location of a single species is very small compared with the nutritionally induced variation in stag beetles, the relatively large intralocational variation suggests that DP is the main component of intraspecific, intralocational variation in these two traits, Hence, intraspecific, intralocational variation can be regarded as measure for DP in body and mandible length. Comparisons of two species in the same genus in the same location offer schematic examples of large interspecific differences in intraspecific variability in body and mandible length (Fig. 4): this indicates large species differences in DP.

Fig. 4.

Intraspecific, intralocational variation in body (B), mandible (M), and genitalia (G) length: Comparisons between a highly variable and a less variable species at the same location in four genera studied.

Fig. 5.

Correlations of species DP measure (intraspecific, intralocational CV) in body length with geographic range (GR).

Comparison of interspecific variation measures (CV of species mean, CV of intraspecific, intralocational SD, and CV of intraspecific, intralocational CV) gives further insights (Table 4). The CV of SD for body length, was invariably greater than the CV of the species mean, suggesting that interspecific variability of intraspecific variability (DP) can be greater than that of mean value itself. A similar variation scheme was observed in mandible length. This leads to further questions of the role of DP in adaptation and its relation with GR.

Heritability of DP

F and P values of ANOVA-II indicate, as expected, that the species mean values in the three traits were highly significantly different among species. More importantly, the species variability itself in body and mandible length differed highly significantly between species (Table 5). Heritability values calculated from ANOVA-II for the species mean were invariably very high in all the three traits in all the genera (Table 6). Similarly, the heritability of intraspecific variation (species DP measures in SD and CV) in body and mandible length in all the genera was very high. This clearly indicates that DP of external morphology is on its own a highly genetic trait at the species level. In contrast to this, the heritability of intraspecific variation for genitalia length was generally low and especially that for CV was not statistically different from 0 in all the genera.

Relation of DP to AP

In those species with >10 sampling locations (13 and 11 species for body and mandible measurement and for genitalia measurement, respectively), the estimation of AP may be reasonably accurate, but the degrees of freedom for the succeeding correlation analyses are small and, even more critically, these species pertain by nature to large GR, resulting in more constricted relationship analysis involving AP. In those species with 5 or more sample locations (32 and 18 species, respectively), the estimation of AP may be less reliable but the degrees of freedom for correlation analyses are larger and these species may better represent the variation range of species GR, and hence may offer a more representative relationship analyses involving AP. Even with these limitations, in which the correlation tends to be underestimated rather than overestimated, species DP was positively correlated with species AP in body length and genitalia length (Table 7).

Relation of AP to GR

Under the same limitation of relationship analysis involving AP, species AP, especially in body and mandible length, was significantly correlated with species GR (Table 8).

Table 9.

Correlation of body length measures with GR (correlation coefficient [d.f. and P])

Genus	Species mean	Species intralocational SD	Species intralocational CV
Lucanus	0.559 (36; <0.001)	0.692 (36; <0.001)	0.703 (36; <0.001)
Odontolabis	0.673 (36; <0.001)	0.746 (36; <0.001)	0.666 (36; <0.001)
Prosopocoilus	−0.059 (51; <0.674)	0.149 (51; 0.288)	0.327 (51; 0.017)
Dorcus	0.292 (56; 0.026)	0.370 (56; 0.004)	0.405 (56; 0.002)
All genera	0.271 (185; 0.0002)	0.448 (185; <0.0001)	0.495 (185; <0.0001)

Table 8.

Correlation between AP measures and GR (correlation coefficient, [d.f., P])

Correlation between		Body length	Mandible length	Genitalia length
Species AP measure	Species GR
Among species with >10 locationsa
SD	GR	0.567 (11, 0.044)	0.585 (11, 0.036)	0.474 (9, 0.141)
CV	GR	0.517 (11, 0.071)	0.519 (11, 0.069)	0.417 (9, 0.202)
h2	GR	0.441 (11, 0.132)	0.516 (11, 0.071)	0.489 (9, 0.127)
Among species with >5 locationsb
SD	GR	0.497 (25, 0.008)	0.532 (25, 0.004)	0.567 (16, 0.014)
CV	GR	0.441 (25, 0.022)	0.531 (25, 0.004)	0.504 (16, 0.033)
h2	GR	0.423 (25, 0.028)	0.557 (25, 0.003)	0.558 (16, 0.016)

^a Thirteen species for body and mandible length and 11 species for genitalia length.

^b Twenty-seven species for body and mandible length and 18 species for genitalia length.

Relation of DP to GR

Both measures of species DP (SD and CV) in body length were highly significantly correlated with species GR in the over-all comparison (encompassing all the four genera) and in individual analysis of three of the four genera (Table 9; Fig. 5). In the genus Prosopocoilus, the correlations were somewhat less significant but were still positive. These correlations were much stronger than that of the correlation between the species means and GR (Table 9). In mandible and genitalia length, the correlations were generally lower and not consistent. These clearly indicate that DP of body size influences the species GR more significantly than the body size itself or other measures.

Discussion

Legitimacy of Plasticity Measures and Heritability

Dealing with naturally occurring variation, clear-cut separation of genetic variation at every level of variation hierarchy cannot be expected. Intraspecific, intralocational variation may include genetic variation. The magnitude of genetic variation within location may not be the same as that across locations. The range of environmental variation within location that is associated with DP may not be the same as that across locations. These factors have to be considered in order to evaluate the legitimacy of the DP measures used in this study. The overwhelmingly larger variation caused by varying nutrition and temperature regimes compared with the variation caused by different families from the same location offsets the concern with differential intralocational genetic variation. The adoption of SD and CV adjusted over all locations (④ of “Glossary for statistical measures and heritability” section) would largely resolve the possible complication that might arise from differential intralocational environmental ranges. Thus, the DP measures used in this study can be considered as a reliable representation of true DP.

Intraspecific, interlocational variation, used as a measure for AP, comprises error, genetic variation (the irreversible component of AP) and the interaction of the species developmental program with locations (the extension of DP or the reversible component of AP). The low intraspecific variability of genitalia length (Table 1) suggests that error and environmentally induced variation (reversible variation) are a small part and genetic variation (irreversible variation) is the main component of interlocational variation in this trait. This probably supports the terminology “heritability” and the conceptual basis for determining broad-sense heritability of AP for genitalia length.

Interspecific variation encompasses the true interspecific genetic variation plus parts of the variation elements mentioned above. Yet, as the interspecific variation is much larger than the interlocational variation (Figs. 2 and 3), which represents much of the nongenetic variation contained in interspecific variation, it can be interpreted that a large part of the interspecific variation is genetic. Here again, the terminology “heritability” and the underlying concepts are appropriate for determination of broad-sense heritability of DP in body and mandible length. The overwhelming significance of these primary effects (Tables 2 and 5) and the convincingly high values of estimated heritability (Tables 3 and 6) suggest that they provide a reliable basis for discussion.

Intraspecific Evolution

Following the traditional understanding of natural selection, evolution starts from an intraspecific genetic diversification through adaptation to different environmental conditions, often provided by different geographic locations and occasionally influenced by random drift. Natural selection on individual phenotypes and random effects on individual genotypes are the driving force of evolution at this level. The diversification of genitalia length in this study fits precisely with this understanding. The broad-sense intraspecific heritability for interlocational variability (AP) is always high (Table 3) and DP, measured by intraspecific, intralocational SD and CV, is consistently low in this trait (Table 1). Selection on individual phenotypes has been the main driving force in intraspecific evolution of genitalia morphology.

On the other hand, the generally low intraspecific heritability of body and mandible length for intraspecific, interlocational variation (AP) (Table 3) indicates that the role of these traits in intraspecific evolution is limited. The very high DP (Table 1) may function primarily as a suppressing agent for morphological diversification: the individuals are able to adapt their phenotype to variation in the environment with no genetic changes. Thus, intraspecific evolution in this group of beetles in the case of genitalia has been through individual selection that has led to specific genital traits well adapted to the local conditions. In the case of DP, evolution has been toward the ability to adapt and modify the phenotype to the prevailing conditions. This plasticity enables the species to maintain its morphological robustness.

Interspecific Evolution

The generally higher interspecific variability of body and mandible length than that of genitalia length (Table 4) indicates that the external morphology plays an important role. The consistently higher interspecific variation of body and mandible length variation (intraspecific, intralocational SD) than the variation of body and mandible length mean values (Table 4) further suggests that DP is probably of greater importance than the mean value in this level of evolution. Furthermore, the high interspecific heritability of intraspecific, intralocational variability (Table 6) supports the concept that DP is of great significance in interspecific morphological evolution. DP can only be observed at the population and cannot be evaluated from the individual phenotype; thus, the unit of interspecific evolution in DP is the population and not the individual.

In contrast, the low interspecific heritability of intraspecific, intralocational variability indicates that DP has not played a major role in evolution for genitalia length. Selection on individual phenotypes has been the main process throughout the intra and interspecific evolution of genitalia morphology.

Species Selection

In discussing species selection or clade selection, irreducibility to lower levels is a requirement for the process, in which the traits or properties in question have to be true species level traits or emergent properties, not aggregate traits (e.g., Lieberman and Vrba 2005). Emergent properties are properties of a physical system that cannot be reduced to the properties of its constituent parts at lower levels, while aggregate traits are species-level features that are the sum of lower-level parts (Dietrich 2010). Being irreducible to the phenotypes of lower level individuals, GR may be a clear case of an emergent property because it is a result of complex and nonlinear processes and not a simple addition of selection on individual phenotypes (Jablonski 2008, Erwin 2010). The species mean should be treated as an aggregate trait, not only because it is a simple linear sum of the component individuals, but also because one directional selection on individual phenotypes will lead to the same one directional change in the mean value. DP can be measured through a mathematical processing of component individuals; yet, it is not a linear sum of individuals and does not respond linearly to one directional selection on individual phenotypes. As such, selection on DP is not reducible to the selection on individual phenotypes. DP at the species level in this study fulfills the requirements to be considered as an emergent trait.

Furthermore, if the DP in a given populations is sufficiently broad that any individual can express a nearly optimal phenotype in the environment in which it finds itself, then there is little or no variation in the ability to survive and hence there is minimal divergent selection. This clarifies the whole concept of DP being a species character and not an individual trait (DeWitt and Scheiner 2004, Storz et al 2010).

It is understandable that the DP measures in body length are more significantly related to GR than the same measures in genitalia and mandible because body size is a more direct measure of the general metabolic activity that may be crucial in expanding into new environments or retreating from changing environments for a particular species. The low to nonexistent correlation of genitalia measures with GR may be due to a limited association with general metabolism. The general absence of correlation of mandible length measures with GR suggests that mandible size may be more closely related to sexual selection than to general metabolism.

The high heritability of DP measures indicates that developmental plasticity is definitely genetically based and is subject to natural selection. Species with higher DP may more readily expand their territory to new environments and persist under adverse conditions. Higher DP may provide an adaptive advantage.

This view is supported by observation that DP measures (SD and especially CV) in body length are consistently correlated more significantly with species GR than with the species mean (Table 9). This correlation between two emergent properties is more significant than a correlation between an aggregate trait and an emergent property, and is a clear sign of a species-level process.

Answers to the Questions

1. How heritable is the plasticity? The present study firmly established a quantum genetic variation in DP and quantified the magnitude of this variation. How this variation originated and inherited would be a great challenge to the future studies.

2. How is the plasticity related to species selection? Vrba (1984) proposes that species selection can only be strictly demonstrated by distinct rates of speciation and/or extinction among closely related species. The present study does not directly provide this evidence. However, I suggest that distinct GRs among related species can be considered as a legitimate proxy for the different speciation rate. Then, this study offers a direct relationship of species DP with species GR, through the positive relationship of DP to AP and of AP to GR. Furthermore, if I follow the slightly less strict criterion for species selection by Lloyd and Gould (1993) or Gould (2002, P662) as a correlation between species-level traits and emergent fitness at the species level, then present correlation between DP and GR fits perfectly with the criteria for species selection.

3. How frequent is species selection in nature? The stag beetle family Lucanidae is a medium-size family in the Coleoptera with 1,414 species known to be described at the time of the monograph by Fujita (2010). In addition to the four genera used in this study, there are many smaller genera in which I have observed a similar species difference in intraspecific variability. It is reasonable to presume that species selection is a common occurrence in these genera as well. Thanks to the many hundreds of enthusiasts, Lucanidae is probably the most taxonomically and ecologically studied family among the families of Coleoptera. If the same level of species identification and geographic range survey are extended to other larger families, many more cases of species selection would be detected.

Acknowledgments

I thank the hundreds of collaborators, dealers, and collectors of stag beetles who made the accumulation of specimens used in this study possible. The great majority of the specimens used for the measurements in this study is stored in my collection, which I intend to donate to a public institution in due time, if there is a taker. I am grateful to J. H. Cock who thoroughly reviewed the manuscript.