Subspeciation befogged by the “Seligmann effect”: the case of Laudakia stellio (Reptilia: Sauria: Agamidae) in southern Sinai, Egypt

A study aimed at clarifying the taxonomic status of the variable hardun population of the south Sinai mountains, hitherto included in Laudakia stellio brachydactyla, encountered the confounding role of the “Seligmann effect”—the biometrical difference between original (whole)‐tailed and regenerated (or broken)‐tailed individuals. We morphologically examined museum specimens from the Negev, Israel, and Sinai, Egypt (N = 117; including 58 focal specimens used in all analyses). We used 23 conventional mensural, meristic, and qualitative characters (besides sex). Biometry was statistically analysed, separately for each sex, within and between two geographically disparate populations constituting potential subspecies, typical L. s. brachydactyla in the Negev and north Sinai, and the questioned south Sinai population. Multivariate analyses poorly separated the two geographical populations. Within each of these, the original‐tailed and regenerated‐tailed subsamples significantly differed in some characters (“Seligmann effect”). After excluding the regenerated‐tailed specimens, principal components analysis separated the two populations at conventional subspecies level. The two significantly differ in several characters, in one without overlap. Hence the south Sinai population is described as L. s. salehi Werner n. ssp. The morphology‐based key distinguishing them was tested on additional specimens and yielded geographically coherent results for all specimens. Minor directional asymmetry of the subdigitalia was left‐dominant and highly significant only in the regenerated‐tailed subsample (lumping sexes and regions). Associations of significant inter‐character correlations differed a little between the two subspecies, indicating a possibility of functional and ecological differences.


Introduction
The hardun Laudakia stellio (Reptilia: Sauria: Agamidae) is the most conspicuous reptile in assorted vertical environments in the eastern Mediterranean, from Corfu, Salonika and the Cyclades through the Turkish Islands, Anatolia (up to 39uN), Syria, northern Iraq, synonymy (see Wermuth 1967) we limit to the bare minimum necessary for showing the nomenclatural history of the population(s) involved.

General research strategy and preliminary results
Like Almog et al. (2005), we relied on available museum material, thus on morphology in terms of conventional mensural, meristic, and qualitative characters (Daan 1967; Baran and OZ 1985;Baig 1992). We did not recruit the sophisticated set of characters used by Moody and Boehme (1984), expecting the main relevant phenomena to be reflected in the simpler basic set (Almog et al. 2005). Moreover, we excluded as redundant some characters reputedly correlating with others, such as the sublabial count (that parallels the supralabial count). Two bilateral meristic characters were recorded on both sides . Together with ratios computed between characters, we recorded 23 characters, besides sex.
We limited statistical sophistication to the minimum yielding solid results. To reduce the problem of allometry, we initially excluded young specimens and expressed measurements as percentages of body length. But to improve sample size, we initially included the many specimens with broken or regenerated tails (Arnold 1984;Loumbourdis and Kattoulas 1986;Schall et al. 1989;Lapid and Werner 1991). In a second round we excluded all specimens lacking complete tails, in consideration of the ''Seligmann effect'', implying biometrical differences between the original-tailed and regenerated-tailed subsamples (Almog et al. 2005).
We first mapped the specimens and excluded those north of latitude 31u309N (070 Old Israel Grid) because according to preliminary information (Werner 1971a;unpublished data) this approximated the northern boundary of L. s. brachydactyla. Next we applied both cluster analysis and principal components analysis to the included material, probing for its north-south differentiation. When this failed ( Figure 1) we tentatively subdivided the area at latitude 30u029400N (940 Old Israel Grid) in the middle of a 50-100 km gap between northern and southern collection sites, in our data as well as those of Saleh (1997), assuming that the gap was real and hence that any north-south differentiation would likely occur across it ( Figure 2). Here we term these areas respectively ''Negev'' in the north and ''Sinai'' in the south, although our boundary differs from the political frontier. Statistical analysis then proceeded in steps: (1) testing for intra-individual variation, directional asymmetry, to see whether bilateral characters should be used separately or averaged, and whether asymmetry itself was a significant taxonomic character (Seligmann 1997). Because asymmetry may vary between sexes and between related taxa  we did this separately for each sex of each potential subspecies. (2) Testing intrapopulation variation, sexual dimorphism (Beutler 1981;Shine 1989;Abo-Taira et al. 1996), to see for which characters the sexes could be pooled. Because sexual dimorphism could vary between related agamid taxa (Werner 1971a) we did this separately for each potential Figure 2. Map of the study area (from Werner 1988) delimited towards north by a broken line and subdivided by a second broken line into northern ''Negev'' and southern ''Sinai'', with localities of the focal specimens (solid circles) and human settlements for orientation (open squares), explained in Werner (1987Werner ( , 1988 subspecies. (3) Because subspecies cannot be defined if most of the geographic variation of characters is discordant (Inger 1961), we tested for correlation within all pair-wise combinations of the 23 characters. (4) Principal components analysis (PCA; separately for each sex), not only for the whole material, but also separately for only the original-tailed specimens, to see whether specimens cluster in geographical coherence. (5) Statistical testing of the differences between the original-tailed and regenerated-tailed specimens (Negev and Sinai separated), separately for each participating character, to identify taxonomically confusing characters. (6) Statistical testing of the differences between the potential subspecies (Negev and Sinai), separately for each participating character, to identify diagnostic characters. This includes statistical reduction of the relevant characters, for each sex and subspecies, which enables constructing a key. Finally, (7) testing the conclusions by applying the resulting key to additional specimens excluded from the detailed analyses due to juvenile size, bad preservation, missing parts, or timing of specimens becoming available.

Materials and methods
The following abbreviations are used: %D, relative asymmetry, the difference between R and L counts, as a percentage of the average of R and L counts (a negative value implies Lside dominance); FMR, female to male ratio, female RA length as a percentage of male RA length (Fitch 1981); L, left side; PCA, principal components analysis; PERCRA, percentage of RA (Werner 1971b); R, right side; RA, rostrum-anus length (Werner 1971b); SD, standard deviation. Museum abbreviations follow Leviton et al. (1985).

Summary of material examined
We examined 117 ''Laudakia stellio brachydactyla sensu lato'' specimens from the Negev (Israel) and Sinai (Egypt), including 58 focal specimens used in all tests and the holotype of L. s. brachydactyla, and for comparison also 44 L. s. vulgaris from Egypt, including the neotype. The specimens are briefly listed by taxon and locality in the Appendix.

Characters
Sex. In Laudakia stellio males are easily identified by their preanal field of callous glandular scales and, especially, longitudinal band of similar glandular scales in mid-belly (Boulenger 1885;Almog et al. 2005). In a sample of 265 adults from Israel and Sinai (Kosswig et al. 1976) and Egypt no ambiguous specimens were encountered (Y. L. Werner, unpublished data). For statistics sex was coded: 1, male; 2, female; and also 5, juvenile.
Mensural characters, including computed ones (11). N RA: from tip of snout to cloacal slit, to nearest 0.5 mm.
All the following measures were processed, and are presented, as PERCRA. The measures (except tail length) were taken by callipers to nearest 0.1 mm. N Tail length: from cloacal slit to tip (only if original), to nearest 0.5 mm. N Head length: from tip of snout to rear border of ear, parallel to long axis (by Goren and Werner's 1993 modified callipers). N Head width: at maximum width, including laterally protruding spiny scales. N Head index: (100 head length)/head width. N Head depth: with closed mouth at maximum height. N Head flatness: (100 head length)/head depth. N Forelimb length: from axilla to tip of claw of longest finger, along mid-width of limb, in two sections: to and from pin in elbow joint. N Hindlimb length: from groin to tip of claw of fourth toe, along mid-width of limb, in two sections: to and from pin in knee joint. N Interlimb ratio: hindlimb length/forelimb length. N Fourth toe length: from insertion of fifth toe to tip of claw.
N Supralabialia (L, R): counted from the most posterior scale sharing the typical appearance of the supralabials, to (and excluding) the rostral.
N Relative asymmetry (%D) in supralabials: the difference between R and L counts, as a percentage of the average of R and L counts.
N Dorsalia size index: count5number of dorsal scales in a longitudinal row, counted along a 10-mm window in a thin plastic ruler, in the largest scaled area vertebrally or paravertebrally, halfway between fore-and hindlimbs. Scale size5(10/count). Index5100(scale size)/(trunk length). Trunk length5RA minus head length (Moody and Boehme 1984).
N Ventralia size index: count5number of ventral scales in an (oblique) transverse row, along 10 mm (method as above), in the middle of the belly (from the midline laterad, located as above). Scale size510/count. Index5100(scale size)/head width.

Colour
Colour of live individuals, in the field or laboratory, was recorded with a Leica M2 camera (assorted lenses) on Kodachrome diapositive film, with a millimetre ruler bearing a series of Ostwald standard colour patches next to the animal (examples in Werner 1995).

Statistics
Basically we used parametric statistics. To test differences between independent samples we applied t-tests for equal or for unequal variances according to the significance (P,0.05) of Levene's Test, equal variances not assumed. The significance of DA was tested by the one-sample t-test.
Bivariate correlations between all characters (and also altitude) were performed separately for males and females, also separating between Negev and Sinai populations (but pooling tail states), via a bivariate matrix.
For multivariate separation of the Negev and Sinai populations, we separately subjected all males (N532) and females (N527) having all data to two procedures. Initially, cluster analyses were done with the MVSP package, using the Squared Euclidian Distance and Minimum Variance (Ward's) Clustering method. When it appeared advisable to select characters for cluster analysis, we applied the ''sort'' function to all characters and evaluated the distribution of each character over the two (geographically) potential subspecies. Finally, we applied PCA; we repeated these tests including only individuals with an original tail having all data (N537).
Other than MVSP, all calculations were by the SPSS statistical software, version 12.0 but graphs were done using Microsoft Excel (version 2003).

Intra-individual variation: asymmetry
Directional asymmetry occurred only in the subdigitalia (not in supralabialia) and only in the regenerated-tailed subsample; it was left-dominant and, for pooled sexes and regions, highly significant (Table IA). Its possible occurrence was tested also separately in the eight character-taxon-sex combinations (original-tailed and regenerated-tailed specimens pooled) and was never significant, though for the subdigitalia of the Negev males P50.076 (Table IB). Although the numbers of subdigitalia differed significantly (or nearly so) between the Negev and Sinai (on both the left and right sides, and in both sexes), the degree of asymmetry did not (Table IC).

Intra-taxon variation: sexual dimorphism
Upon trying to sex 117 presumed Laudakia stellio brachydactyla (sensu lato) specimens by visual examination, they readily segregated into 58 males, 48 females and 11 juveniles. Males were only insignificantly larger (in RA) than females; in the Negev subsample FMR597.8 and in Sinai FMR595.3. The list of characters showing significant sexual dimorphism differed little between the Negev and Sinai (Table II). Both the Negev and Sinai males had larger heads (especially broader and longer) than females but only in the Negev was the male head broader in shape (smaller head index). Both Negev and Sinai males had longer tails than females but in Sinai the difference was greater. Negev and Sinai females both had larger ventralia but this was significant only in Negev. In both Negev and Sinai limiting the comparisons to original-tailed specimens enhanced the difference in limb lengths (especially fourth toe, accompanied by subdigital count), greater in the males.

Correlation among characters
We tested for correlation in all the pair-wise combinations of the 23 characters (excluding sex but including computed ones) separately in each sex and potential subspecies-the Negev and Sinai populations. However, tail states were pooled because of sample size. The more conspicuous, statistically significant, and presumably more important correlations are presented comparatively in Table III. All characters were involved, to varying extent, in correlations with other characters. Of special interest is the ramified involvement of the dorsalia pattern, as this is a key character in defining L. s. brachydactyla.
Among the few differences shown between the sexes, that in the correlation of RA length (mm) with head width (PERCRA) seems instructive: in the Negev this correlation was negative in females, expressing the expected negative allometry of the head in adults, but positive in males, expressing continued allometric increase in head width.  The differences between the Negev and Sinai include some character groups that raise functional and ecological questions; for example, the occurrence mainly in Sinai males of a negative correlation of all limb lengths (PERCRA) with RA length (mm), while retaining a relatively stable interlimb ratio.
The correlations of morphology with the environmental key factor, altitude, were also tested and are of special interest (not shown in Table III). These occurred only in males, and differently in the Negev and Sinai. In particular, the dorsalia pattern correlated with altitude negatively in the Negev (r520.549, P50.034, N515) but positively in Sinai (r50.808, P50.001, N512). In the females, both from the Negev and from Sinai, no characters correlated with altitude.

Multivariate analyses
As mentioned above, initially we applied phenetic cluster analysis, separately to males and females, using the pool of original-tailed and regenerated-tailed specimens for which a maximum number of characters was recorded, and dropping some specimens and some characters so as to get maximum representation without replacing missing data points by averages. All dendrograms (not shown) failed to display any geographical coherence of the branches. The same lack of geographical resolution prevailed with PCA. But when we considered the ''Seligmann effect'' (Table IV) and limited the material in PCA to originaltailed specimens (Figure 3), the Negev and Sinai subsamples of the males became almost separate, and those of the females, completely separate (though adjacent). The 75% rule for separation of subspecies (Mayr 1969) is fulfilled. The confusing effect of the regeneratedtailed specimens, which had foiled the earlier analyses, is exemplified in Figure 4, a histogram showing the bimodal distribution of toe length, one mode due to original-tailed specimens, the other due to regenerated-tailed ones. Thereupon the south Sinai population is described here as a new subspecies, which will be further validated below.     (Israel) and Sinai (Egypt). Bimodal distribution of one character, fourth toe length, one mode due to regenerated-tailed specimens, the other due to original-tailed specimens (N513). Supralabialia (L/R), 12/12; dorsalia pattern, 10; dorsalia size index, 2.47; dorsal enlarged scales juxtaposed in orderly transverse rows, 18 between axilla and groin, each row subdivided into left and right halves by a band of small scales (smaller than the ones between successive tubercle rows) running along the vertebral line. Ventralia size index, 3.62; mid-abdominal band of glandular scales moderately well defined; preanal glandular scales, 60; subdigitalia, 19/19. Coloration (in alcohol): head uniformly grey (no markings around eyes or ears or on throat). Dorsally, body and limbs dark grey; transverse tubercle rows light cream, creating cross-banding; no dorsal light blotches. Limbs grey. Underparts cream-grey; throat somewhat darker. Tail dorsally with 16 grey cross-bands, poorly demarcated from the cream intervals; ventrally yellowish cream in anterior third, thereafter, gradually, the dorsal grey cross-bands form rings.

Variation
The mensural and meristic variation of the focal type specimens are presented in Tables V and VI. The type series (N563) varies in the array of dorsal transverse rows of juxtaposed enlarged scales. A variable number of anterior rows tend to be replaced by scattered tubercles or tubercle clusters that may or may not be arranged in transverse rows. When the set is sufficiently complete for counting, which occurs mainly in males, there are 16-19 (usually 18-19) rows between axilla and groin.
Coloration (in alcohol, based on specimens in which coloration was reasonably well preserved, N557) varies greatly. Dorsally, the coincidence of light cross-banding with the full set of transverse rows of enlarged scales as in the holotype occurs only in 4/22 paratypes. In others only every other such scale row is light-coloured (4/22 paratypes, females), possibly asymmetrically alternating over left and right half-rows (2/22), or the narrow light cross-bands are joined by a light vertebral band (4/22, males) that sometimes shows a few blotch-like accentuated areas (3/22); additional variants occur. Ventrally the gular area is spotless as in the holotype in 23/49 paratypes. Otherwise the grey throat carries darker markings of varying density and shape-tiny dots, dashes, rarely even marbling (20/ 49) or, exceptionally, lighter markings (6/49). The belly is spotless as in the holotype in only 4/21 paratypes, usually it is dotted with single dark scales of varying intensity and density (17/21). The limbs are unmarked, dark, in only one paratype; mostly they are crossbanded dark and light (12/21) or only the forelimbs are (9/21). The number of dark crossbands on the complete tail ranges from 12 to 17 (usually 14 or 15) but the last few may be represented by a long black tail tip.
The live coloration (from diapositives of three males and two females, from assorted localities) is more vivid (Figure 6). In the males, the general background of granules is grey. Mid-dorsally, the light-coloured enlarged scales are orange and the remaining enlarged scales are metallic-bluish lighter grey. On the flanks, clusters of tubercles are light creamyellow. On the head, the eye and ear are framed orange to varying extent. Light orange may be dominant in the cross-banded hindlimbs and anterior part of the tail. The females may be somewhat less colourful, especially on the head, and seem to lack the bluish component of those enlarged dorsal scales that are metallic-grey (the others being orange). In one captive female photographed comparatively on a warm and on a cool day, the grey ground colour was much darker on the cool day.

Distribution, ecology and biology
Comparison of the collection sites of the focal specimens on the map (Figure 2) with the biogeographical map (Werner 1988, p 374) shows that most were on the South-Sinai  mountain massif constituting the Irano-Turanian region enclave. Individual localities were at altitudes ranging form 400 to 2100 m, mostly .1000 m (one at 50 m). Outside this massif the subspecies seems to occur sparsely ( Figure 4) and this scarcity of records is probably not merely due to biased collecting effort, as shown by maps for other lizard species (Werner and Sivan 1994;Segoli et al. 2002). Bodenheimer (1937) regarded L. stellio in Sinai as ''known so far from the higher elevations of the Sinai massif only''. Saleh (1997), too, commented, ''In southern Sinai, apparently restricted to higher altitude where it usually prefers vertical walls and runs into crevices when disturbed''. Danin (1987) characterized the magmatic massive of south Sinai: area 1300 km 2 ; altitude 1500-2640 m; mean annual precipitation 70-100 mm (some of this, snow on high mountains); mean annual temperature 9-15uC; commonest plant-the shrub Artemisia herba-alba.

Etymology
Laudakia stellio salehi is named for Professor Mostafa A. Saleh, Department of Zoology, Al Azhar University, Cairo, Egypt, who recognized the distinction of this population and partly characterized it morphologically and ecologically (Saleh 1997). Moreover, ''saleh'' means righteous, which seems appropriate for a lizard from Mt Sinai.

Comparison between subspecies
Saleh (1997) already noted that ''Individuals from the high mountains of south Sinai … have darker coloration and longer tails and their enlarged mid-dorsal scales are only feebly keeled''. Here we present, beyond the geographical coherence of the PCA scattergrams, the descriptive statistics of characters that significantly differed between the subspecies, within Tables V (males) and VI (females), specifying the levels of significance. In one character, tail length, there was no overlap of the range of variation between the two subspecies in the males, and almost none in the females. Although the phenomena of directional asymmetry and of sexual dimorphism differed little between the two subspecies, the assemblages of correlated characters, including asymmetry, differed between them, as reported above, accentuating the distinction of these subspecies as biological entities. Moreover, the two subspecies differ in qualitative terms: foremost, the array of enlarged keeled dorsal scales. While in L. s. brachydactyla the left and right transverse rows of enlarged scales merge medially with a broad continuous vertebral band of fairly similar enlarged scales, in L. s. salehi these left and right rows tend to be separated by small scales that at least in most males form a continuous narrow vertebral band. The dorsal coloration, variable as described above, never displays the bold four to five yellow or orange blotches that adorn typical L. s. brachydactyla (Werner 1995, p 55, Figure 2;Saleh 1997, Plate 27;Amitai and Bouskila 2001, p 131). The gular coloration, of possible social role, is in L. s. salehi usually uniform or with dark markings, while in L. s. brachydactyla it is usually with light spots (10/13 specimens), rarely uniform or with dark spots (2/13; 1/13). In L. s. vulgaris the throat is usually dark with lighter spots, at least anteriorly (36/42), rarely uniform (5/42) or marbled (1/42). The subspecifically distinct scale and colour patterns appear already in juveniles of 44 mm RA, possibly earlier. We applied this key to 103 adult and young specimens that had not participated in the main analyses. Juveniles under 90 mm RA usually could not be sexed with certainty and so their mensural characters were useless but their scalation and colour pattern enabled identification. All specimens were easily identified to subspecies though five of those 22 identified as L. s. brachydactyla diverged in one character or another and could also be interpreted as intermediates, four with the more northern L. stellio ssp., and one with L. s. salehi.

Discussion
Taxonomy was confounded, and its clarification delayed, by the ''Seligmann effect'', the phenomenon that among conspecific lizards the tail-losers differ biometrically from the tailretainers (Seligmann 2001;Seligmann et al. 2003;Almog et al. 2005). In other words, taillosers (as a statistical group), which appear to be relatively accident-prone, also differ morphologically. Hence it is only natural that including them would affect the statistics of biometry. There may be here a general caveat for taxonomic work in comparable cases. However, it is yet unclear whether the difference between the two behaviouralmorphological types within a taxon is genotypic or perhaps phenotypic, as it could be affected by environmental factors during embryo development (briefly discussed in Razzetti et al. forthcoming). After proper consideration of the ''Seligmann effect'', the morphological distinction of the Sinai harduns from the Negev harduns became clear, enabling, or requiring, the description of the former as a new subspecies. The occasional occurrence of apparent intermediate specimens near subspecies range boundaries is of course compatible with the subspecies concept (Mayr 1969). The subspecies concept has long been criticized (Wilson and Brown 1953) but Laudakia stellio does seem to tend to classical subspeciation. For example, the very distinctive L. s. picea of the Jordanian basal desert (Werner 1992) has recently been found to be linked to a neighbouring population by intermediate specimens (A. M. Disi, personal communication).