Save this PDF as:

Size: px
Start display at page:



1 Herpetologica, 59(1), 2003, by The Herpetologists League, Inc. A CRITICALLY ENDANGERED NEW SPECIES OF CNEMIDOPHORUS (SQUAMATA, TEIIDAE) FROM A CERRADO ENCLAVE IN SOUTHWESTERN AMAZONIA, BRAZIL GUARINO R. COLLI 1,3,GABRIEL C. COSTA 1,ADRIAN A. GARDA 1,KÁTIA A. KOPP 2, DANIEL O. MESQUITA 1,AYRTON K. PÉRES, JR. 1,PAULA H. VALDUJO 1, GUSTAVO H. C. VIEIRA 1, AND HELGA C. WIEDERHECKER 1 1 Departamento de Zoologia, Universidade de Brasília, Brasília, DF Brazil 2 Departamento de Ciências Biológicas, Universidade Federal de Santa Maria, Santa Maria, RS Brazil ABSTRACT: We describe a new species of Cnemidophorus from a Cerrado enclave in southwestern Amazonia, Rondônia state, Brazil. This species is apparently endemic to Cerrado enclaves in the vicinity of the city of Vilhena, a region under intensive anthropic pressure due to the expansion of soybean plantations. A discriminant analysis indicated that femoral pores and scales around the tail are the best discriminators among Brazilian species of Cnemidophorus. A naïve Bayesian network constructed with categorical (mostly coloration) variables indicated that the new species had high conditional probabilities of dorsolateral fields absent, vertebral field spotted, and paravertebral lines absent. The analyses revealed clear distinctions between species of Cnemidophorus that range north and south of the the Amazon River. The new species may have evolved as a result of vicariance, following the isolation of peripheral enclaves of Cerrado in southwestern Amazonia after the late Pleniglacial. The restricted range in small areas, under extreme human pressure around Vilhena, makes this species one of the most critically endangered elements of the Brazilian herpetofauna. Key words: Amazonia; Brazil; Cerrado; Cnemidophorus; Extinction; Lizard; Rondônia RANGING from northern United States to central Argentina, Cnemidophorus has the widest geographic distribution among all teiid genera (Wright, 1993). Currently, 57 species are known in the genus (Uetz et al., 1995). All South American species of Cnemidophorus are assigned to the C. lemniscatus species group (Burt, 1931; Wright, 1993), which comprises 17 species: C. arenivagus Markezich, Cole, and Dessauer, 1997; C. arubensis Van Lidth de Jeude, 1887; C. cryptus Cole and Dessauer, 1993; C. gramivagus McCrystal and Dixon, 1987; C. lacertoides Duméril and Bibron, 1839; C. leachei Peracca, 1897; C. lemniscatus (Linnaeus, 1758); C. littoralis Rocha, Araújo, Vrcibradic, and Costa, 2000; C. longicaudus (Bell, 1843); C. murinus (Laurenti, 1768); C. nativo Rocha, Bergallo, and Peccinini-Seale, 1997; C. nigricolor Peters, 1873; C. ocellifer (Spix, 1825); C. pseudolemniscatus Cole and Dessauer, 1993; C. serranus Cei and Martori, 1991; C. vacariensis Feltrim and 3 CORRESPONDENCE: , Lema, 2000; and C. vanzoi (Baskin and Williams, 1966). That about 50% of the species in the Cnemidophorus lemniscatus group has been described in the last 15 yr testifies to the paucity of knowledge of the alpha-level diversity in the neotropics and the growing taxonomic activity in the area. Indeed, until recently, all populations of Cnemidophorus in northern South America were assigned to a single species, C. lemniscatus. Several studies have indicated that C. ocellifer is also a complex of species (e.g., Rocha et al., 1997, 2000; Rodrigues, 1987). In addition, the availability of tools to probe molecular markers has only recently allowed the detection of morphologically cryptic species, such as C. cryptus (e.g., Cole and Dessauer, 1993; Sites et al., 1990). During the course of a study on the Cerrado enclaves in Rondônia, Brazil, we collected a relatively large series of an undescribed species of Cnemidophorus. A few individuals of this species had been previously collected by Vitt and Caldwell (1993) during a survey sponsored by the 76

2 March 2003] HERPETOLOGICA 77 FIG. 1. Vegetation map of type locality of Cnemidophorus parecis. Polonoroeste program along the BR-364 road in Rondônia (Nascimento et al., 1988; Vanzolini, 1986). The individuals were referred by Vanzolini both to Cnemidophorus ocellifer (Vanzolini, 1986) and to an unnamed member of the ocellifer species group (sic) (P. E. Vanzolini, personal communication, in Vitt and Caldwell, 1993). Herein, we describe this new species, making detailed comparisons with other Brazilian congeneric species. MATERIALS AND METHODS We collected lizards with a shotgun, pitfall traps, and drift fences on 4 10 November 1998 and 21 August 22 September 1999, around the city of Vilhena (12 43 S, W), Rondônia state, Brazil (Fig. 1). The vegetation in the study area consists of a mosaic of Cerrado enclaves (Eiten, 1972) within a matrix of Terra Firme Amazonian Forest (Pires and Prance, 1985) and human-modified areas consisting of soybean plantations and groves of pine trees. We found lizards exclusively in Cerrado enclaves with sandy soils around the city of Vilhena, even though we sampled the largest patches of Cerrado enclaves throughout Rondônia, including those around the cities of Pimenta Bueno and Guajará-Mirim. We obtained data on C. cryptus, C. gramivagus, C. lemniscatus, C. littoralis, and C. ocellifer from specimens housed in the Coleção Herpetológica da Universidade de Brasília (CHUNB) (Appendix I). The first three species occur predominantly in Amazonian Savannas north of the Amazon River (e.g., Ávila-Pires, 1995; Cole and Dessauer, 1993), whereas the last two occur exclusively south of the Amazon (e.g., Cei, 1993; Rocha et al., 2000). It should be noted that C. cryptus is a provisional name for a complex of parthenogenetic, unisexual populations in the Amazon Basin and that much work is necessary to clarify their taxonomic status. We recorded the following meristic variables in each specimen: supralabials (number of enlarged scales along the upper jaw, total on both sides), infralabials (number of enlarged scales along the lower jaw, total on both sides), chinshields (most anterior pair of chinshields separated from infralabials by row of small scales), gular folds (number of folds in gular region), supraoculars (number of supraoculars on right side), parietals (number of parietals plus interparietal scale), scales around midbody (counted midway between fore- and hindlimbs, excluding ventrals), transverse rows of ventrals (counted along the midline, from gular fold to anterior margin of hindlimbs), ventrals in transverse row (counted midway between fore- and hindlimbs), femoral pores (total number on both sides), prefemorals (number of enlarged scales on anterior aspect of thigh, counted midway between the hip and the knee, on a row from femoral pores to granules on dorsal aspect of thigh), prefemoral rows (counted from hip to knee), infratibials (number of enlarged scales on longitudinal row from knee to base of first metatarsal), preanals (number of enlarged scales on preanal plate, from level of medialmost femoral pores to vent), fourth finger lamellae (counted under the finger), fourth toe lamellae (counted under the toe), scales around tail (counted on fifth transverse row), and dorsals (counted along the midline, from occiput to first transverse row of scales around tail). We used a discriminant function analysis (Tabachnick and Fidell, 2001) to inves-

3 78 HERPETOLOGICA [Vol. 59, No. 1 tigate differences among species in meristic characters. To identify the most powerful meristic discriminators of the species of Cnemidophorus, we used a stepwise discriminant function analysis of meristic characters (Tabachnick and Fidell, 2001), using the METHOD STEPWISE option in PROC STEPDISC of SAS (SAS Institute Inc., 1988). We evaluated linear discriminant functions through posterior probability error-rate estimates, based on crossvalidation (SAS Institute Inc., 1988). We also recorded the following categorical variables: lower lateral fields (LLF, one on each side, area between ventral scales and lower lateral stripes: absent, light, spotted, or dark), upper lateral fields (ULF, one on each side, area between upper lateral and dorsolateral stripes: absent, light, spotted, or dark), dorsolateral fields (DLF, one on each side, dark area between dorsolateral and paravertebral stripes: absent, light, spotted, or dark), vertebral fields (VTF, one on each side, middorsal area between paravertebral stripes: absent, light, spotted, or dark), vertebral line (VTL, light stripe from interparietal scale to base of tail: absent, interrupted, continuous, or double), paravertebral stripes (PVS, one on each side, light stripe from parietal scale to first third of tail, between vertebral and dorsolateral fields: absent, interrupted, or continuous), dorsolateral stripes (DLS, one on each side, light stripe from superciliaries to first third of tail, between dorsolateral and upper lateral fields: absent, interrupted, or continuous), upper lateral stripes (ULS, one on each side, light stripe from suborbital region to hindlimb, between upper and lower lateral fields: absent, interrupted, or continuous), lower lateral stripes (LLS, one on each side, light stripe from axilla to hip, between lower lateral field and ventrals: absent, interrupted, or continuous), lateral spots (LTS, rounded light areas on flanks: absent, present), hindlimb spots (HLS, rounded light areas on hindlimbs: absent, present), chinshield contact (CHS, degree of contact between most anterior pair of chinshields: no contact, contact smaller than half of their lengths, or contact greater than half of their lengths), semicircles (SMC, degree of contact between supraoculars and medial head scales: no contact, no contact with semicircles isolating first supraocular, supraoculars contacting frontal and parietals, supraoculars contacting frontal, supraoculars contacting parietals), dorsal caudals (DCA, keels on dorsal, caudal scales, from most anterior third of tail: absent, present), ventral caudals (VCA, keels on ventral, caudal scales, from most anterior third of tail: absent, present), and preanal spur (PAS: absent, present). To model the conditional dependencies of each categorical variable within each species, we constructed a naïve Bayesian model using the software UNBARB (available at ftp://ftp.cic.unb.br/pub/cic/wagner/ software/bnets/unbarb.zip). In this model (Fig. 2), the variable species influences the chance of the occurrence of states of each categorical variable, assuming that categorical variables are conditionally independent for a given species and that the joint distribution of all variables satisfies the following: k 1 k i i 1 P(S, C,...,C) P(C S) P(S), where S is species, C is a categorical variable, P(C i S) is the posterior probability of C i given S, and P(S) is the prior probability of S (Cowell et al., 1999). We estimated P(S) and P(C i S) using Bayesian statistics (Bernardo and Smith, 2000; DeGroot, 1970) and the software Hugin Lite v. 5.3 (Jensen and Nielsen, 1999). Throughout the text we report means 1 SD. SPECIES DESCRIPTION Cnemidophorus parecis sp. nov. Holotype. CHUNB (Figs. 3, 4), adult male, from Fazenda Cachoeira ( S, W), 50 km NW of Vilhena, Estado de Rondônia, Brazil, elevation about 612 m, collected on 5 November 1998 by Guarino R. Colli and students. Paratopotypes. (107) All specimens from Cerrado enclaves around Vilhena: CHUNB , , collected on November 1998 by same collectors; CHUNB , collected on August September 1999 by same collectors.

4 March 2003] HERPETOLOGICA 79 FIG. 2. Diagram depicting naïve Bayesian network constructed to evaluate conditional probabilities of 16 categorical variables from species, assuming independence among categorical variables. See text for abbreviations. Diagnosis. A species distinguished from all other members of the lemniscatus group by the following combination of characters: (1) large size, maximum SVL 90 mm; (2) scales around tail; (3) femoral pores; (4) dorsals; (5) dorsolateral fields absent; (6) vertebral field spotted; (7) paravertebral stripes absent (Fig. 4); and (8) preanal spurs absent. Cnemidophorus parecis differs from C. lemniscatus in having supralabials (14 16 in C. lemniscatus), infralabials (15 19), femoral pores (44 54), scales around tail (24 30), dorsolateral fields absent (dark), vertebral field spotted (dark), paravertebral stripes absent (continuous), vertebral line absent (two continuous stripes) (Fig. 5), hindlimb spots absent (present), and ventral caudals smooth (keeled). Cnemidophorus parecis differs from C. gramivagus in having femoral pores (40 51 in C. gramivagus), scales around tail (20 32), lower lateral fields light (absent), upper lateral fields light (dark), dorsolateral fields absent (dark), vertebral field spotted (dark), paravertebral stripes absent (continuous) (Fig. 5), lower lateral stripes interrupted (absent), hindlimb spots absent (present), and ventral caudals smooth (keeled). Cnemidophorus parecis differs from C. cryptus in having supralabials (14 17 in C. cryptus), 8 10 transverse rows of ventrals (7 8), femoral pores (38 48), scales around tail (22 30), upper lateral fields light (dark), dorsolateral fields absent (dark), paravertebral stripes absent (continuous) (Fig. 5), and ventral caudals smooth (keeled). Cnemidophorus parecis differs from C. littoralis in having scales around tail (25 34 in C. littoralis), upper lateral fields light (dark), paravertebral stripes absent (interrupted), vertebral line absent (interrupted) (Fig. 5), and dorsal caudals keeled (smooth). Cnemidophorus parecis differs from C. ocellifer in having 8 10 transverse rows of ventrals (5 8 in C. ocellifer), femoral pores (16 28), scales around tail (21 30), and lateral spots absent (present) (Fig. 5). Description of holotype. Rostral pentagonal, higher than wide, visible from above, bordered posteriorly by nasals. Nasals in contact along midline, each nasal divided by an oblique suture. Nostril in lower part of suture, directed posterolaterally, longer than tall. Frontonasal sub-

5 80 HERPETOLOGICA [Vol. 59, No. 1 FIG. 3. Cnemidophorus parecis, holotype, CHUNB 09811, adult male, SVL 86 mm. FIG. 4. Top: Cerrado enclave in Vilhena, Rondônia, Brazil. Middle: Adult female of Cnemidophorus parecis. Bottom: Adult male of C. parecis (holotype, CHUNB 09811). hexagonal, suture with nasals and loreals forming a semicircle, sutures with prefrontals forming a wide angle. Prefrontals quadrilateral, with medial suture twice as long as that between nasals, in contact laterally with loreal, first supraciliary, and first supraocular. Frontal approximately pentagonal, longer than wide, and wider anteriorly. Sutures with prefrontals and supraoculars roundish, those with frontoparietals forming a straight line. Frontal in contact laterally with first, second, and third supraoculars. Frontoparietals pentagonal, longer than wide, with long, straight medial suture; in contact laterally with third supraocular and small scales bordering supraocular region posteriorly. Interparietal hexagonal, bordered at each side by hexagonal parietals, divided into two irregular scales by an oblique suture. Occipital scales irregular, posterior to interparietal and parietals, variable in size, including two moderately enlarged, medial scales, separated by a smaller scale. Four supraoculars on each side, second and third largest, followed posteriorly by group of small scales. Second, third, and fourth supraoculars separated from supraciliaries by row of small scales. Supraciliaries seven on each side, highest anteriorly, third largest, others subequal. Loreal single, large,

6 March 2003] HERPETOLOGICA 81 FIG. 5. Dorsal coloration patterns in six species of Cnemidophorus from Brazil. (A) C. littoralis, CHUNB 08308, (B) C. ocellifer, CHUNB 12964, (C) C. parecis, CHUNB 11652, (D) C. lemniscatus, CHUNB 01101, (E) C. cryptus, CHUNB 08491, (F) C. gramivagus, CHUNB in contact with nasal, frontonasal, prefrontal, first supraciliary, frenocular, first subocular, and second, third, and fourth supralabials on the left side, and third and fourth supralabials on the right side. Frenocular narrow, higher than wide, in contact with subocular, loreal, first supraciliary, and small scales in ocular region. Suboculars three, large, second largest, all in contact with supralabials. A continuous keel runs from frenocular to about midlength of second subocular. Postoculars small, arranged in two rows with 3 4 larger scales. Lower eyelid with semiopaque disc, formed by transversally enlarged, convex scales. Supralabials six on each side, fifth and sixth below center of eye; followed by series of small scales extending to commissure of mouth. Temporal region with irregular scales, granular centrally, moderately enlarged peripherally. Ear opening large, oblique, with smooth margins, anterior margin forming a semicircle, posterior margin straight. Tympanum recessed in a short auditory meatus. All dorsal and lateral head scales juxtaposed, smooth. Symphysal anteriorly ellipsoid, in posterior contact with first infralabials and postsymphysal, forming two wide angles. Postsymphysal single, pentagonal, in contact with first and second infralabials, followed by three pairs of enlarged chinshields. First pair of chinshields in ample contact along midline, first and second pairs in contact with infralabials, third pair separated from infralabials by row of small scales, followed posteriorly by enlarged scales. Medial scales on chin small, smooth, elongate, arranged in longitudinal, posteriorly divergent rows; increasing in size and becoming roundish posteriorly. Infralabials six, followed posteriorly by series of small scales extending to commissure of mouth; fifth infralabial below center of eye. Gular region divided in two areas: anterior region with enlarged, rounded scales, in transverse rows, delimited posteriorly by line uniting lower margin of ear openings; posterior region covered with granules, in transverse rows, bordered posteriorly by antegular fold. Gular and antegular folds marked by granules; scales between the two folds larger, irregular, slightly imbricate.

7 82 HERPETOLOGICA [Vol. 59, No. 1 TABLE 1. Canonical discriminant analysis of meristic characters of six species of Cnemidophorus. Canonical variable Variable Canonical statistics Eigenvalue Cumulative proportion r 2 F P Supralabials Infralabials Chinshields Gular fold Supraoculars Parietals Scales around midbody Ventral rows Transverse rows of ventrals Femoral pores Prefemorals Prefemoral rows Infratibial rows Preanals Fourth finger lamellae Fourth toe lamellae Scales around tail Dorsals Species C. cryptus C. gramivagus C. lemniscatus C. littoralis C. ocellifer C. parecis Raw canonical coefficients Canonical variable Canonical variable Means Canonical variable Canonical variable Scales on nape and sides of neck similar to dorsals. Dorsals and scales on flanks granular, rounded, smooth, subimbricate; 214 scales from nape to base of tail; 111 scales in transverse row around midbody (excluding ventrals). Ventrals large, smooth, rectangular (wider than long), imbricate, in 31 transverse rows; 10 ventral scales in transverse rows across midbody. Ventral scales separated from scales on flanks by row of moderately large scales. Preanal plate with five rows of enlarged scales, surrounded laterally and posteriorly by small scales. Preanal spurs absent. Femoral pores in a continuous row along each thigh, medially with a short gap; 13 pores on each side. Scales on base of tail rectangular, longer than wide, smaller than ventrals, keeled dorsally, smooth ventrally, in transverse rows. All transverse rows continuous around tail, except first two rows incomplete ventrally. Tail scales becoming longer and narrower posteriorly; subcaudal scales becoming keeled posteriorly. Tail broken, regenerated from 37th transverse row of scales. Limbs with large, smooth, imbricate scales on dorsal aspect of upper arms, anterodorsal aspect of forearms, anteroventral aspect of thighs, and ventral aspect of lower legs; elsewhere scales small, granular. Larger scales on upper arms in longitudinal rows. Forearms with one row of enlarged scales, wider than long. Anterior scales on thigh decreasing in size proximally. Lower legs with three rows of enlarged, hexagonal scales. Ventral aspect of hands and feet granular; three enlarged tubercles at base of pollex. Subdigital lamellae single; lamellae under both fourth fingers 14; under left fourth toe 25; under right fourth toe 24. Color in life. Head brownish green dorsally, lemon-green laterally; labial regions and ventral aspect of head yellowish (Fig. 4). Anterior two-thirds of dorsum brownish green; remaining dorsal areas, dorsal aspect of tail, hindlimbs, and forearms brown. Anterior aspect of thighs, proximal aspect of lower legs, lower flanks, and lateralmost longitudinal rows of ventrals vivid blue. Remainder of belly and ventral aspect of limbs and tail immaculate white. Anterior aspect of dorsum with two subtle paravertebral rows of brown spots. Dorsolateral stripe whitish, continuous from supraciliaries to first third of tail; lateral stripe whitish, continuous from suborbital region to hip; lower lateral stripe whitish, interrupted, extending from axilla to hip. Color in fixative (ethanol, after formalin). Head brownish green dorsally, bluish laterally; labial regions and ventral aspect of head whitish. Anterior two-thirds of dorsum brownish blue; remaining dorsal areas, dorsal aspect of tail, hindlimbs, and forearms brown. Remaining areas of the body as in life. Variation. Sexes are dimorphic in

8 March 2003] HERPETOLOGICA 83 FIG. 6. Scores on the first two canonical discriminant axes based on meristic variables in six species of Cnemidophorus from Brazil. adult color pattern (Fig. 4). Dorsal surface of head brownish green in males, brown in females. Anteriormost two-thirds of dorsum greenish in males, whereas females have only a narrow, green blotch, restricted to the first third of the dorsum. Sides of the head lemon-green in males, but brown in females. Labial regions and throat yellowish in males and whitish in females. Anterior aspect of thighs, proximal aspect of lower legs, and lower flanks vivid blue in males and whitish in females. Overall, juveniles of both sexes resemble adult females in coloration, but the dorsum is predominantely brown. Etymology. The name parecis refers to the highlands in western Brazil, Chapada dos Parecis, that separate the Paraguai and Amazon basins and include the known range of the species. The Chapada dos Parecis was named after the Paresi indigenous people, a branch of the Aruak language, who inhabit the region. Distribution and ecology. Known only from the type-locality (Fig. 1). Cnemidophorus parecis is one of the most abundant lizard species in the Cerrado enclaves around Vilhena. It is sympatric with the polychrotids Anolis meridionalis and Polychrus acutirostris, the skinks Mabuya frenata and M. nigropunctata, the gymnophthalmids Micrablepharus atticolus and Pantodactylus schreibersii, and the teiids Ameiva ameiva, Kentropyx vanzoi, and Tupinambis merianae. Ecological data on the species are summarized in Mesquita (2001). The microhabitats used most frequently by C. parecis are the open ground (with grasses) and the area under schrubs (Fig. 4). Individuals are active during midday, with a mean cloacal temperature of 38.2 C. The diet consists mainly of termites, ants, spiders, and beetles. Clutch size varies from one to two eggs. RESULTS AND DISCUSSION The first two canonical discriminant variables explained approximately 88% of the total variation in the meristic variables (Table 1). The first canonical variable, representing mainly a contrast between femoral pores and transverse rows of ventrals (Table 1), indicated a clear separation between species from Amazonian Savannas

9 84 HERPETOLOGICA [Vol. 59, No. 1 TABLE 2. Meristic characters of six species of Cnemidophorus. Values indicate x 1 SD, with range in parentheses. Species Variables C. cryptus (n 80) C. gramivagus (n 64) C. lemniscatus (n 54) C. littoralis (n 53) C. ocellifer (n 101) Supralabials Infralabials Chinshields Gular folds Supraoculars Parietals Scales around midbody Ventral rows Transverse rows of ventrals Femoral pores Prefemorals Prefemoral rows Infratibial rows Preanals Fourth finger lamellae Fourth toe lamellae Scales around tail Dorsals (14 17) (13 17) (1 3) (2 5) (98 118) (26 32) (7 8) (38 48) (5 9) (9 13) (7 12) (3 4) (14 18) (28 33) (22 30) ( ) (13 16) (13 18) (4 4) (95 120) (29 33) (8 8) (40 51) (4 9) (10 13) (9 14) (14 21) (27 34) (20 32) ( ) (14 16) (15 19) (3 6) (98 126) (26 31) (8 8) (44 54) (5 8) (9 13) (9 12) (14 19) (24 31) (24 30) ( ) (13 17) (10 13) (96 123) (30 38) (8 10) (28 36) (4 11) (8 14) (6 13) (16 24) (29 37) (25 34) ( ) (11 16) (10 14) (0 2) (3 4) (81 141) (25 31) (5 8) (16 28) (4 8) (8 11) (7 10) (13 20) (24 38) (21 30) ( ) C. parecis (n 108) (11 14) (10 15) (1 3) (4 5) (2 5) (96 127) (29 35) (8 10) (25 33) (5 12) (10 14) (8 12) (4 5) (13 19) (23 31) (34 47) ( )

10 March 2003] HERPETOLOGICA 85 TABLE 3. Stepwise discriminant analysis of meristic characters from six species of Cnemidophorus, with error rate estimates based on cross validation. Step Variable entered F P Error rate Femoral pores Scales around tail Dorsals Ventral rows Fourth toe lamellae Transverse rows of ventrals Infralabials Parietals Supralabials Infratibial rows Prefemorals Chinshields Fourth finger lamellae Gular fold Scales around midbody Preanals Supraoculars (C. cryptus, C. gramivagus, and C. lemniscatus) and the remaining species (C. littoralis, C. ocellifer, and C. parecis) (Fig. 6). Amazonian Savanna species, with positive means on the first canonical variable (Table 1), are characterized by a high number of femoral pores and fewer transverse rows of ventrals (Table 2). The second canonical discriminant variable, representing a contrast between transverse rows of ventrals versus supralabials and fourth toe lamellae (Table 1), indicated a clear separation among the three species that occur south of the Amazon River (Fig. 6). Cnemidophorus parecis, with the highest mean on the second canonical variable (Table 1), is characterized by numerous transverse rows of ventrals and few supralabials and fourth toe lamellae (Table 2). Cnemidophorus littoralis, with the lowest mean on the second canonical variable (Table 1), is characterized by few transverse rows of ventrals and numerous supralabials and fourth toe lamellae (Table 2). Cnemidophorus ocellifer (and the Amazonian Savanna species) has intermediate values for these variables. The stepwise discriminant analysis of meristic characters selected 17 variables as the most powerful discriminators among the species of Cnemidophorus (Table 3). Femoral pores was selected first, correctly classifying 82.4% of the individuals, followed by scales around tail and dorsals, whose addition improved the classification criterion to 87.6% and 90.4%, respectively (Table 3). To determine if important variables were excluded from the analysis because of correlations among the original variables (Tabachnick and Fidell, 2001), we removed femoral pores and repeated the analysis. Scales around the tail were then entered in the first step, correctly classifying 56.3% of the lizards (error rate 0.437, F , P 0.001). We then removed scales around the tail and repeated the analysis. This time, infralabials was the first variable to be selected, correctly classifying about 60% of the lizards (error rate , F , P 0.001). These analyses indicated that femoral pores and scales around the tail are the meristic characters that best discriminated the species of Cnemidophorus and that their discriminatory power was high. The naïve Bayesian model indicated that C. parecis determined high conditional probabilities ( 95%) of 12 categorical variables: dorsolateral fields absent, vertebral field spotted, paravertebral lines absent, lateral line continuous, lower lateral line interrupted, vertebral line absent, lateral spots absent, hindlimb spots absent, preanal spur absent, supraoculars contacting frontal and parietals, ventral caudals smooth, and dorsal caudals keeled (Table

11 86 HERPETOLOGICA [Vol. 59, No. 1 TABLE 4. Conditional probabilities of states for qualitative characters for six Brazilian species of Cnemidophorus, assuming a naïve Bayesian model. Species Variables C. cryptus C. gramivagus C. lemniscatus C. littoralis C. ocellifer C. parecis Lower lateral fields Upper lateral fields Dorsolateral fields Vertebral field Paravertebral stripes Dorsolateral stripes Upper lateral stripes Lower lateral stripes Vertebral line Lateral spots Hindlimb spots Chinshield contact Preanal spur Semicircles Ventral caudals Dorsal caudals D (55.95%) D (95.24%) D (96.43%) D (71.43%) C (96.43%) C (97.59%) C (96.39%) C (65.06%) A (67.86%) A (64.63%) A (98.78%) (96.39%) A (98.78%) IFP (95.29%) K (98.78%) K (98.78%) A 95.59% D (94.12%) D (95.59%) D (89.71%) C (91.18%) C (95.52%) I (53.73%) A (91.04%) A (95.59%) A (96.97%) P (93.94%) (97.01%) A (69.70%) IFP (94.20%) K (98.48%) K (98.48%) L 48.28% D (70.69%) D (94.83%) D (93.10%) C (94.83%) C (89.47%) C (73.68%) C (64.91%) C (93.10%) A (57.14%) P (98.21%) (87.72%) P (51.79%) IFP (93.22%) K (98.21%) K (98.21%) L 50.88% D (92.98%) D (64.91%) D (64.91%) I (94.74%) C (82.14%) C (96.43%) I (66.07%) I (89.47%) A (98.18%) A (98.18%) (80.36%) A (98.18%) IFP (58.62%) S (98.18%) K (98.18%) L 71.43% D (46.67%) D (43.81%) L (80.00%) I (56.19%) C (64.42%) C (52.88%) A (56.73%) A (94.29%) P (90.29%) A (98.06%) (98.08%) A (99.03%) Co (51.89%) S (99.03%) K (99.03%) L (91.96%) L (91.07%) A (96.43%) Sp (96.43%) A (97.32%) C (84.68%) C (96.69%) I (95.50%) A (97.32%) A (99.09%) A (99.09%) (44.14%) A (99.09%) IFP (95.58%) S (99.09%) K (99.09%) Note: values in bold indicate conditional probabilities greater than 95%. D: dark; L: light; Sp: spotted; A: absent; P: present; C: continuous; I: interrupted; K: keeled; S: smooth; Co: complete; IFP: incomplete, supraoculars contacting frontals and parietals; : contact between first pair of chinshields greater than half of their lengths.

12 March 2003] HERPETOLOGICA 87 4). Further, only C. parecis determined high conditional probabilities of dorsolateral fields absent, vertebral field spotted, and paravertebral lines absent (Table 4). The analysis also indicated that species from Amazonian Savannas can be clearly separated from species south of the Amazon River by the presence of keeled subcaudals in the former (Table 4). Our results revealed a greater similarity between C. parecis and the other species ranging south of the Amazon River (C. littoralis and C. ocellifer). We also verified that, besides the presence of anal spurs in males, the species of the C. lemniscatus complex can also be distinguished from their congeneric southern relatives by the presence of keeled subcaudals, a larger number of femoral pores, and fewer transverse rows of ventrals. This distinction suggests that C. parecis may be a closer relative of the southern species of Cnemidophorus. A number of palynological and sedimentary studies indicate that, in Rondônia, open vegetation replaced the forest in the recent past, approximately 3000 ybp (Absy and Van der Hammen, 1976; de Freitas et al., 2001; Sifeddine et al., 2001; Van der Hammen and Absy, 1994). Therefore, present day enclaves of Cerrado in Rondônia may be relicts of a past open vegetation landscape that ranged from Central Brazil to Rondônia. Following the breakup of the connections between Rondônia and the Central Brazil open landscapes, C. parecis presumably differentiated as a consequence of the geographical isolation. This view is supported by the presence of a depauperate Cerrado lizard fauna in the Rondônia enclaves (Vanzolini, 1986; Vitt and Caldwell, 1993). However, the affinities of C. parecis can only be ascertained with confidence through rigorous phylogenetic analysis. Apparently, C. parecis is an endemic of a small group of Cerrado patches with sandy soils around the city of Vilhena. Rondônia is the most deforestated state in Amazonia, having lost about one-third of its total vegetation cover, whereas the municipality of Vilhena has less than 10% of the original vegetation. The Brazilian legislature (Código Florestal) stipulates that 80% of each rural property must be preserved (Reserva Legal), but a state law passed recently in Rondônia reduced this percentage up to 20% in highly productive regions, favoring the establishment of large properties. The Cerrado patches around Vilhena are critically threatened by the expansion of soybean plantations, favored by the gentle topography that aids mechanization, the low risks of drought, and the recent development of two exportation corridors. The first is the Ferronorte Railroad, connecting Santa Fé do Sul in São Paulo to Alto Taquari in Mato Grosso. The other is the Hidrovia do Rio Madeira, connecting Porto Velho in Rondônia state to Itacoatiara in Amazonas state, along the Madeira and Amazon Rivers. Therefore, C. parecis is apparently at a critical risk of becoming extinct within a very short period of time. Because the presumed range of C. parecis is not included in any conservation unit, being entirely within privately owned land, urgent measures are needed to protect the species. Acknowledgments. This work was developed under the project Estrutura e dinâmica da biota de isolados naturais e antrópicos de Cerrado: lições para a biologia da conservação, funded by Programa Nacional da Diversidade Biológica-PRONABIO, MMA- MCT-CNPq-GEF-BIRD. S. F. Balbino, C. G. Batista, C. A. Freitas, D. Diniz, the staff of IBAMA at Vilhena, Joécio Costa, Aparecido Donadoni, Vicente de Paulo Campos Godinho, and M. M. Utumi assisted in fieldwork. We thank Dr. Wagner Teixeira da Silva for assistance with Bayesian statistics, Zenilton J. G. Miranda for drawing the holotype, and three anonymous reviewers for their insightful comments on the manuscript. This work was supported by graduate student fellowships from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior to A. A. Garda (# /99-9), P. H. Valdujo, D. O. Mesquita, and G. H. C. Vieira; from Conselho Nacional do Desenvolvimento Científico e Tecnológico to A. K. Péres, Jr. (# /99-9) and H. C. Wiederhecker (# /00-7); and by a research fellowship from Conselho Nacional de Desenvolvimento Científico e Tecnológico CNPq to G. R. Colli (# /88-1). LITERATURE CITED ABSY, M. L., AND T. VAN DER HAMMEN Some paleoecological data from Rondônia, southern part of the Amazon basin. Acta Amazonica 6: ÁVILA-PIRES, T. C. S Lizards of Brazilian Amazonia (Reptilia: Squamata). Zoologische Verhandelingen, Leiden 1995: BERNARDO, J.M.,AND A. F. M. SMITH Bayesian Theory. Wiley, New York, New York, U.S.A.

13 88 HERPETOLOGICA [Vol. 59, No. 1 BURT, C. E A study of the teiid lizards of the genus Cnemidophorus with special reference to their phylogenetic relationships. Bulletin of the United States National Museum 154: CEI, J. M Reptiles del Noroeste, Nordeste y Este de la Argentina. Museo Regionale di Scienze Naturali, Torino. COLE, C. J., AND H. C. DESSAUER Unisexual and bisexual whiptail lizards of the Cnemidophorus lemniscatus complex (Squamata: Teiidae) of the Guiana region, South America, with descriptions of new species. American Museum Novitates 3081:1 30. COWELL, R. G., A. P. DAWID, S. L. LAURITZEN, AND D. J. SPIEGELHALTER Probabilistic Networks and Expert Systems. Springer-Verlag, New York, New York, U.S.A. DE FREITAS, H. A., L. C. R. PESSENDA, R.ARAVENA, S. E. M. GOUVEIA, A. S. RIBEIRO, AND R. BOULET Late Quaternary vegetation dynamics in the southern Amazon Basin inferred from carbon isotopes in soil organic matter. Quaternary Research 55: DEGROOT, M. H Optimal Statistic Decisions. McGraw-Hill, New York, New York, U.S.A. EITEN, G The Cerrado vegetation of Brazil. The Botanical Review 38: JENSEN, F., AND L. B. NIELSEN Hugin Lite, Version 5.3. Hugin Expert A/S, Aalborg, Denmark. MESQUITA, D. O Uma análise comparativa da ecologia de populações do lagarto Cnemidophorus (Squamata: Teiidae) do Brasil. Unpublished Master s Thesis, Universidade de Brasília, Brasília, Distrito Federal. NASCIMENTO, F. P., T. C. S. ÁVILA-PIRES, AND O. R. CUNHA Répteis Squamata de Rondônia e Mato Grosso coletados através do Programa Polonoroeste. Boletim do Museu Paraense Emílio Goeldi, nova série, Zoologia, Belém 4: PIRES, J. M., AND G. T. PRANCE The vegetation types of the Brazilian Amazon. Pp In G. T. Prance and T. E. Lovejoy (Eds.), Key Environments: Amazonia. Pergamon Press, Oxford, U.K. ROCHA, C. F. D., H. G. BERGALLO, AND D. PECCI- NINI-SEALE Evidence of an unisexual population of the Brazilian whiptail lizard genus Cnemidophorus (Teiidae), with description of a new species. Herpetologica 53: ROCHA, C. F. D., A. F. B. ARAÚJO, D.VRCIBRADIC, AND E. M. M. COSTA New Cnemidophorus (Squamata; Teidae) from coastal Rio de Janeiro State, southeastern Brazil. Copeia 2000: RODRIGUES, M. T Sistemática, ecologia e zoogeografia dos Tropidurus do grupo Torquatus ao sul do Rio Amazonas (Sauria, Iguanidae). Arquivos de Zoologia, São Paulo 31: SAS INSTITUTE INC SAS/STAT User s Guide, Release 6.03 Edition. SAS Institute Inc., Cary, North Carolina, U.S.A. SIFEDDINE, A., L. MARTIN, B.TURCQ, C.VOLKMER- RIBEIRO, F.SOUBIES, R.C.CORDEIRO, AND K. SU- GUIO Variations of the Amazonian rainforest environment: a sedimentological record covering 30,000 years. Palaeogeography, Palaeoclimatology, Palaeoecology 168: SITES, J. W., JR., D. M. PECCININI-SEALE, C. MO- RITZ, J.W.WRIGHT, AND W. M. BROWN The evolutionary history of parthenogenetic Cnemidophorus lemniscatus (Sauria, Teiidae). I. Evidence for a hybrid origin. Evolution 44: TABACHNICK, B. G., AND L. S. FIDELL Using Multivariate Statistics. Allyn & Bacon, Needham Heights, Massachussetts, U.S.A. UETZ, P.,T.ETZOLD, AND R. CHENNA The EMBL Reptile Database. embl-heidelberg.de/ uetz/livingreptiles.html. VAN DER HAMMEN, T.,AND M. L. ABSY Amazonia during the last glacial. Palaeogeography, Palaeoclimatology, Palaeoecology 109: VANZOLINI, P. E Levantamento Herpetológico da Área do Estado de Rondônia sob a Influência da Rodovia BR 364. Conselho Nacional do Desenvolvimento Científico e Tecnológico, Assessoria Editorial, Brasília, Distrito Federal, Brasil. VITT, L. J., AND J. P. CALDWELL Ecological observations on Cerrado lizards in Rondônia, Brazil. Journal of Herpetology 27: WRIGHT, J. W Evolution of the lizards of the genus Cnemidophorus. Pp In J. W. Wright and L. J. Vitt (Eds.), Biology of Whiptail Lizards (Genus Cnemidophorus). The Oklahoma Museum of Natural History, Norman, Oklahoma, U.S.A. Accepted: 22 April 2002 Associate Editor:Joseph Mendelson III APPENDIX I Specimens Examined Cnemidophorus cryptus (80). BRAZIL: Amapá: Macapá, CHUNB , , ; Tartarugalzinho, CHUNB Cnemidophorus gramivagus (64). BRAZIL: Amazonas: Humaitá CHUNB , ; Ilha da Botija CHUNB Cnemidophorus lemniscatus (54). BRAZIL: Pará: Santarém, CHUNB Cnemidophorus littoralis (53). BRAZIL: Rio de Janeiro: Barra de Maricá CHUNB , , , 24117, Cnemidophorus ocellifer (101). BRAZIL: Bahia: Ibipeba, CHUNB , 03265, , 03358, 03360, , 03372, , 03378, 03401, 03405, 03414, 03421, 08481, ; Irecê, CHUNB , , 03359, , , 03377, , 03382, ; Lençóis, CHUNB 05591; Paulo Afonso, CHUNB ; Salvador, CHUNB , 08152; Pernambuco: Buíque, CHUNB 23529; Exu, CHUNB 12855, 13590; Petrolina, CHUNB , , , , 13614, 24109, Cnemidophorus parecis (108). BRAZIL: Rondônia: Vilhena, CHUNB , ,