A new species of green-eyed Cordylus Laurenti, 1768 from the west-central highlands of Angola, and the rediscovery of Cordylus angolensis (Bocage, 1895) (Squamata: Cordylidae)

Cordylus angolensis (Bocage, 1895) was described 128 years ago on the basis of a single specimen collected at Caconda in the west-central highlands of Angola. Additional specimens referred to this species were collected at ‘Mombolo’ (also in the central highlands) during the Vernay Angola Expedition in 1925. As the holotype was apparently destroyed in the fire of 1978 at the Museu Bocage in Lisbon and no additional specimens have been collected, its taxonomic status and phylogenetic relationships has remained uncertain. The species has eluded all efforts aimed at its re-discovery in the vicinity of the type locality, with a single specimen from near Condé, north of Mombolo—collected in 1970 by Wulf Haacke—the only other specimen of Cordylus known from west-central Angola. Recent field work in the Angolan highlands resulted in the collection of a series of specimens from Taqueta Mountain (west of Caconda), Monte Verde (Sandula, ‘Mombolo’) and Uassamba (Vondo). A phylogenetic analysis, using three mitochondrial and six nuclear genes, indicated the existence of two distinct species-level lineages in the Angolan highlands. These two species are allopatric and morphologically distinct, differing especially in terms of their colour patterns, eye colour and certain scalation characteristics. We therefore confirm that C. angolensis is a valid species and designate a neotype, and describe a new species, Cordylus momboloensis sp. nov.


Introduction
Cordylus Laurenti, 1768 is the most speciose and geographically widespread genus of girdled lizards (Cordylidae), comprising 21 species of mainly rupicolous lizards that occur from the south-western tip of South Africa northwards through Botswana and Namibia to Angola, and elsewhere to Mozambique, Zimbabwe, Zambia, Malawi, south-eastern Democratic Republic of the Congo, Tanzania and Kenya, with an apparently isolated population in southern Ethiopia (Branch 1998;Broadley and Branch 2002;Branch et al. 2005;Greenbaum et al. 2012;Bates et al. 2014;Stanley et al. 2016;Marques et al. 2019).
Cordylus originally encompassed all heavily armoured members of the family Cordylidae, but a phylogenetic analysis by Stanley et al. (2011) revealed that this grouping was paraphyletic and the genus was consequently divided into six genera, five of which were new.Cordylus was reduced to only 20 species, with three additional species, Cordylus marunguensis Greenbaum et al., 2012, Cordylus namakuiyus Stanley et al., 2016, and Cordylus phonolithos Marques et al., 2019 described subsequently.Two species in the C. minor FitzSimons, 1943group, namely C. cloetei Mouton & Van Wyk, 1994and C. aridus Mouton & Van Wyk, 1994, were recently synonymised with C. minor (Tolley et al. 2021).Stanley et al. (2011) identified two geographically distinct clades within Cordylus, a predominantly East African group (that also included the Angolan Cordylus machadoi Laurent, 1964) containing species found from the South African province of Mpumalanga northwards to Ethiopia (see Broadley and Branch 2002), and a Cape group containing species restricted mainly to the Western and Eastern Cape provinces of South Africa (see Branch 1998;Bates et al. 2014).Within the more northerly East African group, Cordylus vittifer (Reichenow, 1887) was recovered as sister to all other species, with Cordylus jonesii (Boulenger, 1891), C. machadoi and Cordylus rhodesianus (Hewitt, 1933) forming a clade sister to the remaining eastern and northern species.Subsequently, more species-rich phylogenies returned the Angolan C. machadoi as sister to the East African group (Greenbaum et al. 2012;Stanley et al. 2016;Marques et al. 2019).However, the phylogenetic position of another Angolan species of Cordylus, C. angolensis Bocage, 1895, remained unknown as the latter was not sampled in any of these studies.
Four species of Cordylus have been recorded from Angola, namely C. angolensis and three closely related species restricted to the south-western parts of the country, namely C. machadoi, C. namakuiyus and C. phonolithos.Bocage (1895) examined three specimens of Angolan Cordylus, all of which he considered similar to Cordylus cordylus (Linnaeus, 1758).He described Zonurus (= Cordylus) angolensis on the basis of one of these specimens, sent from Caconda (13°46'S, 15°05'E) in the west-central part of Angola by the Portuguese collector José Alberto de Oliveira Anchieta .The other two specimens without specific localities were assigned to Z. cordylus.Perhaps because the description of Z. angolensis is contained under the heading "Zonurus cordylus" and Bocage (1895: 25) noted that it was "included in our catalogues provisionally under the name Z. angolensis" (translated from French), he neglected to mention this species in his subsequent list of type specimens in the Lisbon collection (Bocage 1897).The holotype of Z. angolensis is presumed to have been destroyed in the fire of 1978 at the Museu Bocage in Lisbon (Almaça and Neves 1987;Madruga 2011).Broadley (1965Broadley ( , 1966) ) split the genus Cordylus into two groups on the basis of habitat: an essentially East African group of mainly tree-living species, and a South African-Zimbabwe group of more strongly depressed rupicolous taxa (see also Visser 1971).Although nothing was known about its habitat use, Broadley (1965) assigned C. angolensis-presumably on geographical grounds-to the more northerly group of partly arboreal species.Broadley (1971: 22) considered the taxonomy of the northern forms of Cordylus and noted "I have examined the type and only known specimen of Zonurus angolensis Bocage from Caconda, Angola, and it is probably synonymous with C. t. tropidosternum Cope, having the three outer rows of ventrals keeled and mucronate and the tail almost equal to snout-vent length in spite of the tip being missing."Broadley and Branch (2002: 10) noted that C. angolensis appeared to be a valid species, but "only three or four specimens are available, so this species should be redescribed when more material becomes available from Angola".The additional specimens they referred to were three lizards from Mombolo in the collection of the American Museum of Natural History (New York) collected in Angola during the Vernay Angola Expedition in 1925 (Bogert 1940) and another specimen (Ditsong National Museum of Natural History [formerly Transvaal Museum], Pretoria) from near Condé further north in the country (D.G.Broadley in litt. 17 April 2009).Greenbaum et al. (2012) examined three (AMNH R4733-5) Mombolo specimens for which they provided various measurements and scale counts.Although they referred their material to C. angolensis because it was similar to Bocage's (1895) holotype with regard to colour pattern, head dimensions, lack of a loreal scale, presence of a small row of scales between the anal plates, and in having similar numbers of transverse rows of ventrals, they noted that their specimens had only 11-12 longitudinal rows of ventrals compared to 16 in the holotype.Reissig (2014) also provided scalation data for 'C.angolensis', but without reference to museum material, and mapped it at the two localities mentioned above.
Cordylus vittifer machadoi was described by Laurent (1964) on the basis of a male holotype (see fig. 26 in Ceriaco et al. 2020d) and a juvenile paratype (Fig. S1), both collected at an elevation of 1800 m a.s.l. at Leba (15°03'S, 13°16'E) on the Humpata Plateau, Huíla Province, Angola.This locality is about 240 km south-west of Caconda, the type locality of C. angolensis.A few additional nearby localities for C. machadoi were added by Stanley et al. (2016).Laurent (1964) considered his Angolan lizards to be closely related to South African populations of C. vittifer (Figs 1, S2), despite them occurring about 1600 km apart, as they shared the characteristically elongated scales of the first row of transverse dorsals (see FitzSimons 1943, p. 461 and fig. 368; this feature is also present in C. namakuiyus and C. phonolithos).Cordylus v. machadoi was distinguished from other C. vittifer on the basis of its greater numbers of longitudinal dorsal scale rows (26-28 versus 20-24) and a narrower rostral scale (2.32-2.34times wider than deep versus 3-4 times wider).Reichenow (1887), in his original description of Zonurus vittifer (Fig. S2), reported only 20-22 longitudinal dorsal scale rows and did not mention the size or shape of the rostral, and it therefore seems that Laurent (1964) was referring to Loveridge (1944a, see table on p. 15) and possibly also FitzSimons (1943) when comparing machadoi to vittifer.It should be noted that concepts of species boundaries within Cordylus around the time of Laurent's (1964) description were very different from what they are now.At that time, biological rather than lineage-based species concepts were often followed, and morphological similarity was used to rank populations, sometimes resulting in morphologically distinct and geographically disparate populations being considered subspecies.Both FitzSimons (1943) and Loveridge (1944a), for example, recognized several subspecies of C. cordylus, often isolated, occurring from the Cape Peninsula (C.c. cordylus) to as far north as Zimbabwe (i.e., C. c. rhodesianus = C. rhodesianus) and central Namibia (C.c. pustulatus [Peters, 1862] = Namazonurus pustulatus); and Loveridge (1944a) considered Z. angolensis a subspecies of C. cordylus, despite a geographical gap of about 2000 km.
Material from "Damaraland" (north-western Namibia), identified as Zonurus griseus Cuvier, 1829 by Pe-ters (1869) and Z. cordylus by Boulenger (1885), was referred to Cordylus cordylus angolensis by Loveridge (1944a) and Mertens (1955Mertens ( , 1971)), although this region is close to southerly records of C. namakuiyus in Angola (Lobón-Rovira et al. 2022b) and C. cf.machadoi in adjacent northern Namibia (Stanley et al. 2016).However, the Damaraland material cannot be assigned to any species of Cordylus with certainty as Peters (1869) did not provide morphological data, while Boulenger's (1885) description included material (probably C. cordylus) from South Africa, and he does not mention the characteristically elongated first row of transverse dorsals that characterises the C. machadoi group (C.machadoi, C. namakuiyus, C. phonolithos) and C. vittifer.Peters' specimens were probably deposited in the collection of the Museum für Naturkunde der Humboldt-Universität, Berlin, but most Namibian Cordylus specimens there are without specific localities and it was not possible to trace any specimens specifically from Damaraland (F.Tillack in litt. 2 February 2012).Mertens (1955Mertens ( , 1971) ) suggested that "C.c. pustulatus" may be a junior synonym of "C.c. angolensis", but N. pustulatus is in fact morphologically distinct and restricted to central Namibia (Branch 1998).Monard (1937) referred Bocage's (1895) two Angolan "C.cordylus" without localities to C. vittifer, but it is not clear why he was of that opinion.Perhaps he had in fact examined some of the Angolan material and had observed the elongated first row of transverse dorsals that characterises C. vittifer and the C. machadoi group.Nevertheless, Loveridge (1944a) disagreed with Monard and suggested, apparently on the basis of their assumed geographical proximity, that Bocage's Cordylus were all referable to C. c. angolensis.Loveridge (1944a: 39) also noted that Boulenger's (1897) record of a Zonurus cordylus from the Aruwimi River (a tributary of the Congo River) in Democratic Republic of the Congo "may be angolensis or else an undescribed form, it certainly would not be typical cordylus." Laurent (1964) felt that Monard (1937) could not have assigned the two Angolan specimens (lacking locality details) to C. vittifer based on the information presented by Bocage.Bocage's (1895) description of these specimens includes sparse details and the scalation details given do not identify the specimens with certainty as C. machadoi, C. namakuiyus or C. phonolithos.However, the same details do not unequivocally refer these specimens to C. angolensis as suggested by Loveridge (1944a), and Bocage's specimens may likely represent the first Angolan records of either C. machadoi or C. namakuiyus, considering that his main provider of Angolan material, José de Anchieta, collected extensively within the geographical range of these two species, and not C. phonolithos (see below) which is restricted to Serra da Neve.The problem is now academic, as these specimens, along with the holotype of C. angolensis, were apparently destroyed in the fire at the Museu Bocage in Lisbon in 1978.Although Bocage's description of C. angolensis is brief and unillustrated, there is little reason to suggest that his species is synonymous with C. machadoi, C. namakuiyus or C. phonolithos.
A molecular phylogeny conducted by Stanley et al. (2016) on cordylids from south-western Angola deter- mined that C. machadoi occurred at high elevations on the Humpata Plateau, while populations in the Pro-Namib, at lower elevations in the more arid western parts, represented a new species, C. namakuiyus.Morphologically, the latter species differed most notably from C. machadoi on account of its extensive osteodermal armour, and possession of a large interparietal scale that completely separates the anterior parietals (in C. machadoi the interparietal is small and does not fully separate the anterior parietals).The status of Namibian populations from Baynes and Otijihipa mountains referred to as C. machadoi (e.g., Branch 1998;Stanley et al. 2016, as C. cf.machadoi) remains unclear as they have not been subjected to genetic analysis, although newly-collected Cordylus namakuiyus material from the highlands of Iona National Park (2033 m elevation), near the southern border of Angola (Lobón-Rovira et al. 2022b), suggests that the material from Baynes and Otjihipa mountains may belong to this species.Specimens were recently collected in the Otjihipa Mountains and will be subjected to genetic analysis to examine their status (F.Becker pers comm.2022).
A third species in the C. machadoi group, namely C. phonolithos, was described by Marques et al. (2019).It was based on a subadult male (holotype) and a juvenile (paratype) collected at Serra da Neve inselberg, north of the other two species and about 200 km west of Caconda.The new species was shown to be the sister species to C. machadoi + C. namakuiyus.The authors named several morphologically diagnostic features that distinguished their new species from the latter two species, but with a larger sample at hand, most of these characters appear to be variable (see below; P. Vaz Pinto and J. Lobón-Rovira, unpubl.data).
Fortunately, the holotype of C. angolensis was examined at the Museu Bocage in 1968 by the late Dr Donald Broadley (Natural History Museum, Bulawayo) before it was apparently destroyed in the fire a decade later.Broadley's scale counts agree with data presented by Bocage (1895), but he also recorded additional morphological data (see below).A series of five specimens from Mombolo on the plateau in west-central Angola, collected on 11 September 1925 by H. and A. Chapman during the Vernay Angola Expedition, were initially catalogued at the American Museum of Natural History (New York) as "Cordylus tropidosternum jonesii".Although similar to C. angolensis, to which they were referred by Broadley and Branch (2002) and Greenbaum et al. (2012), it was evident that these specimens differed with regard to certain scalation characteristics.Differences in morphology between C. angolensis and the Mombolo samples mentioned above led us to hypothesise that at least two distinct species were present in the central-western highlands of Angola.
Recent field work in central-western Angola resulted in the collection of a series of specimens from Taqueta Mountain (c. 100 kms west of Caconda), Monte Verde (Sandula, 'Mombolo') and Mount Uassamba (Vondo).In order to examine the evolutionary relationships between these populations, we conducted a phylogenetic analysis using three mitochondrial and six nuclear genes.A de-tailed morphological evaluation was also conducted, indicating the existence of two distinct species, one of which we refer to C. angolensis, while the other represents a new species (initially referred to as C. 'Mombolo') which we describe below.

Molecular analysis
Sampling Between 2016 and 2021, specimens of Cordylus were collected across Angola, focusing on the central highlands.A total of 25 specimens, comprising 12 C. angolensis-like specimens and 13 specimens of the C. machadoi group, were collected as vouchers.Specimens were euthanised by injection of tricaine methanesulfonate (MS222) (Conroy et al. 2009).After euthanasia, specimens were fixed in 10% formalin, after which they were transferred to 70% ethanol for long-term storage in the Museo Nacional de Ciencias Naturales (MNCN), Spain, Port Elizabeth Museum, Gqeberha (PEM), South Africa and Fundação Kissama (FKH), Angola.For phylogenetic analyses, liver or muscle samples were collected and stored in 96% or 99% ethanol.Localities were recorded by Global Positioning System (WGS84 co-ordinate system).

Molecular data
Phylogenetic analysis was used to provide a phylogenetic context and support the morphological findings.For that purpose we generated 213 new Cordylus sequences from 24 individuals from across Angola following Stanley et al. (2011), summarised in Table 1.DNA was extracted using EasySpin Genomic DNA Tissue Kit (Citomed, Portugal), following the manufacturer's protocols.PCR amplifications were performed to amplify three mitochondrial genes (16S, 12S and ND2) and six nuclear genes (RAG-1, c-mos, BDNF, PRLR, MYH2 and Kif24) following Stanley et al. (2011Stanley et al. ( , 2016)).Primer and PCR conditions are presented in Table 2.For phylogenetic comparisons we combined the newly-generated sequences and complemented this with previously published sequence data of 60 Cordylus individuals (Stanley et al. 2011(Stanley et al. , 2016;;Greebaum et al. 2012;Nielsen and Colston 2014;Marques et al. 2019), using Namazonurus campbelli (FitzSimons, 1938) as outgroup, and Pseudocordylus + Ouroborus as sister groups (Stanley et al. 2011; Table 1).Samples of 20 of the 21 known species (excluding only C. nyikae Broadley & Mouton, 2000) were included in the analysis.Sequences were cleaned and inspected visually using GE-NEIOUS Prime v2021.1.1 (http://www.geneious.com),and aligned using MUSCLE with the same software.Finally, sequences were concatenated using MESQUITE v3.6 (Madisson and Madisson 2019).All sequences have been deposited in GenBank (Table 1).
The identity of an individual from Morro do Pundo, Bocoio (P1-279 in Table 1), for which only a tail tip was obtained, was determined (as C. 'Mombolo') using the 16S marker under the conditions as defined above.This sample was not included in the phylogenetic analysis, but its locality was plotted on the distribution map below.

Material examined
This study was based on material in the herpetological holdings of the following museums/collections: American Museum of Natural History, New York (AMNH); Ditsong National Museum of Natural History (formerly Transvaal Museum), Pretoria (TM); Florida Museum of

External morphology
Specimens were examined using binocular dissecting microscopes (up to 40 or 48 times magnification).Measurements were performed using digital calipers (0.01 mm) or a ruler (1 mm), often under magnification.[Cope, 1869]) were examined for comparison (Appendix 1).These numbers include specimens for which details of head scalation only were available from photographs (two C. machadoi, two C. namakuiyus, five C. phonolithos; see below).Males were usually identified by their well-developed femoral pores (weakly developed with small or absent secretory plugs in females) and the presence of differentiated femoral scales (absent in females).However, because some females of C. vittifer possess differentiated femoral scales (De Waal 1978), all specimens of this species were also dissected and examined for presence of ovaries and/or oviducts in females, and testes and/or sperm ducts in males.
Measurements were performed on the right side of the body unless damaged.The following were measured: snout-vent length (SVL, from tip of snout to anterior margin of vent, with lizard on its back and flattened); tail length (posterior margin of vent to tail tip, original tails unless indicated); head length (tip of snout to posterior margin of lateral temporals); head width (widest part in the temporal area); head height (midpoint of eye from top of head to bottom of lower jaw); ear length (greatest distance more-or-less dorso-ventally); eye length (excluding small scales around the eye); nostril-eye distance (shortest distance between); eye-ear distance (shortest distance); snout-eye length (tip of snout to anterior margin of eye, excluding small scales around the eye); internarial distance (shortest distance between nostrils); inter-orbital distance (measured between the middle of the eye sockets); snout-arm length (from tip of snout to anterior margin of forelimb); axilla-groin distance (posterior edge of forelimb insertion to anterior edge of hindlimb insertion); forearm (inner part of elbow fold to tip of claw of fourth (longest) finger; 4 th (longest) toe of hindlimb.
Scale counts were, unless indicated, performed on both sides of each lizard.The following counts were taken: supralabials (all scales bordering the upper lip, excluding the rostral, but including the scale at the corner of the mouth; the second last supralabial is the largest); infralabials (all scales bordering lower lip, excluding the mental, the posterior one situated at the corner of the mouth below and partly behind the posterior supralabial); sublabials (large scales in contact with the infralabials, the posterior one situated near the ear opening below and partly behind the posterior infralabial); number of chin shields in contact with anterior sublabials; supraoculars (large scales); supraciliaries (narrow and elongate scales above the eye and in contact with the much larger supraoculars, excluding tiny granules and postocular scales that may be in contact with the posterior supraocular); lo-

' C G C C T G T T TAT C A A
reals; suboculars (all scales bordering eye and in contact with the supralabials, but excluding the preocular and any postoculars; when present, a fourth subocular is invariably situated near the back of the eye); preoculars; rows of gulars between posterior angles of jaws, excluding tiny granules; dorsals, transverse rows (from immediately behind occipitals to vent); dorsals, longitudinal rows (counted midway between fore-and hindlimbs, excluding granular or minute scales on either side if present); ventrals, transverse rows (from axilla to groin; the most anterior row curves anteriorly and the most posterior row curves posteriorly); ventrals, longitudinal rows (counted midway between fore-and hindlimbs; the most lateral row consists of scales generally at least half as wide as the adjacent inner ones; when this row is not clearly distinguishable, the formula '12 + 2' is used, indicating a possible extra row on either side); subdigital lamellae on 4 th (longest) toe (the most basal scale counted is fully present on the digit); femoral pores; differentiated femoral scales (generation glands).
The interpretation of 'occipital' requires explanation.In the genus Cordylus, occipitals are typically defined as the scales posterior to, and in contact with, the pair of posterior parietals, plus those scales in contact with the posterior upper temporals [on either side of the posterior parietals] (e.g., FitzSimons 1943).This is quite straightforward in some species, such as C. vittifer (Fig. S2), as the occipitals are often distinct, of similar size and present as a more-or-less straight row across the back of the head; and also, in this species, the row of occipitals is followed immediately behind by the first transverse row of dorsals which consists of mostly elongated scales (longer than those of the next row of dorsals) and has the appearance of a collar.However, the interpretation of what comprises occipitals in most species of Angolan Cordylus is complicated by the presence of what appear to be incomplete rows of these scales, i.e., rows that start with a scale situated behind the posterior upper temporal, followed by one or more scales on other side (and in contact behind the posterior parietals), but the 'row' is then interrupted medially by the intrusion of large scales from the row behind.We refer to the interrupted row as 'occipitals' ('pre-occipitals' in Stanley et al. 2016) and the undivided row behind, which has some scales in contact with the parietals medially, as 'post-occipitals' (Fig. 2).The posterior part of the post-occipitals aligns, more-or-less, with the posterior part of the posterior lateral temporals.In the three species of the C. machadoi group, the two rows of 'occipitals' are both situated anterior to the 'collar' represented by the first transverse row of dorsals, which, as in C. vittifer, is easily recognized by its much-elongated scales.In C. namakuiyus and C. phonolithos the first 'row' of occipitals is incomplete, but in C. machadoi it is often continuous throughout.The situation is somewhat more difficult to interpret in the two west-central highlands species as neither has the transverse row of much-elongated dorsal scales.In the Taqueta Mountain population of Angolan Cordylus (C.angolensis) there is a continuous row of occipitals (the next row is considered the first transverse row of dorsals), the posterior ends of which are more-or-less aligned with the posterior parts of the posterior lateral temporals; but in other populations from the west-central highlands of the country (C.'Mombolo') there is a medially-interrupted row of occipitals followed by a row of post-occipitals (posterior edges extending slightly beyond, or aligned with, the posterior edges of posterior lateral temporals).

Comparison with related species
All material referable to C. angolensis, as well as that of C. 'Mombolo', was compared to one another and samples of other Angolan  Bauer et al. 1994) and the species varies considerably in terms of scale characters and colour pattern (e.g., De Waal 1978;Jacobsen 1989), samples were selected from six areas representing its total range, with populations assigned to the three known varieties (A, B, C) where possible (Fig. S3).A number of scalation and body size characters were compared (see below).

Osteological analysis
For osteological descriptions of the west-central Angolan cordylid material, microComputed Tomography (CT) datasets of C. 'Mombolo' (AMNH R47333-4 and TM 46476) and C. angolensis (MNCN 50648) were produced using Phoenix V|Tome|X M (Nanoscale Research Facility, University of Florida), Phoenix V|Tome|X S (Microscopy and Imaging Facility, American Museum of Natural History) and Zeiss Xradia 510 Versa (Centro de Instrumentación Científica of Granada [CIC], Spain) systems.Separate scans of the head and body were performed to maximize the resolution of the cranial datasets.Xray settings were modified to maximize resolution, contrast and signal to noise (Table S1).X-ray radiographs were converted to tomograms using filtered back-projection, with Datos|X R (Waygate Technologies) and Reconstructor Scout-and-Scan Control System (Zeiss) software.Comparative CT datasets for the other three species of Angolan Cordylus and nine species of east African Cordylus were produced at the institutions listed above, or obtained from www.morphsource.org.All CT data were processed in VGStudioMax 2022.1 (Volume Graphics, Heidelberg, Germany) to digitally isolate individual cra-nial bones, postcranial skeleton and osteoderms to facilitate osteological descriptions and comparative analysis.Total osteoderm and postcranial skeleton volumes were recovered for all scanned Angolan cordylid material and the osteoderm/postcranial skeleton volume ratio plotted against SVL to provide an allometric assessment of the degree of armour for each species.All resulting tomogram data are freely available on www.morphosource.org(Table S1).

Species assessment
We follow a lineage-based species concept whereby a species is represented by an independently evolving metapopulation lineage (Frost and Hillis 1990;de Queiroz 1998de Queiroz , 2007)).The species delimitation approach followed in this study is based on an integrative approximation (Vieites et al. 2009;Padial et al. 2010;Lobón-Rovira et al. 2022d), considering as separate species those lineages that are well supported based on phylogenetic, morphological, and biogeographic/ecological differences.

Phylogenetic analysis
Both the BI and ML analyses were largely concordant, yielding the same well-supported topology.Phylogenetic analyses demonstrate that Angolan Cordylus is sister to the C. jonesii-C.rhodesianus group (PP: 0.99, BS: 87) (Fig. 3).

Morphological evaluation
Length of scales in the first transverse row of dorsals (Figs 4,S2).In C. machadoi, C. namakuiyus and C. phonolithos, and C. vittifer, all or most of the scales in the first transverse row of dorsals are about 1.5-3 times longer than those in the row behind.In some cases, this row appears as a distinct collar.In C. angolensis and C. 'Mombolo', as well as all other species of Cordylus (e.g., FitzSimons 1943;Broadley and Branch 2002), the first transverse row of dorsals is similar to the second, with most or all scales of about the same length.
Presence or absence of a loreal scale (Figs 5,7).A loreal shield (usually distinctly smaller than the preocular) is usually present in most species compared here, as well as C. cordylus (see also FitzSimons 1943;Mouton 1987;Bates 2007;Broadley and Branch 2002), but always absent in C. angolensis and C. 'Mombolo'.A loreal was absent in eight out of 35 C. vittifer variety B (partly fused with preocular on one or both sides of head in two specimens; absent on left side only in one specimen) and partly fused with the preocular on the right side of the head in one (of 18) C. vittifer from KwaZulu-Natal.
Nasal scales (Figs 5,7).In all Angolan species, C. jonesii, and C. rhodesianus the nasal scale is divided posteriorly (i.e., longitudinally), while it is undivided in C. tropidosternum (see also Broadley and Branch 2002) and C. vittifer (but divided longitudinally in NMB R8768).In a couple of C. 'Mombolo' (i.e., TM 46476 and right side of PEM R025217) the nostril appears to be in contact with the preocular.There are also differences in the position of the sulcus -it is situated between the outer center of the nostril and the preocular in C. angolensis and C. 'Mombolo', between the outer center of nostril and loreal in C. machadoi, C. phonolithos and C. namakuiyus, between outer center of nostril and suture between loreal and first supralabial ('curved sulcus' according to Broadley and Branch 2002) in C. rhodesianus, and is present as a groove (also referred to as a 'curved sulcus' by Broadley and Branch 2002) running from above the nostril down to the outer center of the loreal in C. jonesii.(Figs 5,7).Broadley and Branch (2002) noted that the position of the nostril within the nasal proved to be a useful taxonomic character.The nostril of C. angolensis and C. 'Mombolo' is pierced halfway up the posterior edge of a large, slightly domed nasal, and is well separated from the first supralabial (usually by more than the greatest extent of the nostril).In C. machadoi, C. phonolithos and C. namakuiyus the situation is similar and the nostril is also well separated from the first supralabial -lower edge of nostril separated from first supralabial by a distance of at least ¾ its own diameter, whereas in C. vittifer the nostril and first supralabial are in close proximity -usually (87.5%, n = 72) separated by a distance of <½ diameter of nostril (first supralabial is in contact with the nostril in two specimens, and in NMB R8864 the nasal rim and first supralabial are fused).In C. marunguensis (south-eastern Democratic Republic of the Congo) the nostril is situated centrally on the lower margin of the nasal, separated from the first supralabial by a narrow rim (Greenbaum et al. 2012).In C. cordylus the nostril is situated in the lower, posterior part of the nasal, and frequently bordered below by a small scale (FitzSimons 1943;Bates 2007).In C. jonesii and C. rhodesianus the nostril is pierced centrally in the posterior half of a semi-divided nasal; whereas in C. tropidosternum it is pierced in the postero-inferior corner of the nasal and bordered by the first supralabial (Fig. 7; Broadley and Branch 2002).

Position of the nostril
The position of the nostril relative to the suture between rostral and first supralabial is also informative.In C. angolensis and C. 'Mombolo' the entire nostril is situated posterior to the suture, as is the case in C. tro pi doster num and C. vittifer, whereas in all three species of the C. machadoi group at least some part of the nostril is sit- Proximity of the frontonasal to the frontal scale (Fig. 4).In most species the frontonasal and frontal are usually separated by the prefrontals.However, these scales were in contact in 38% of C. jonesii and 29% of C. rhodesianus (Broadley and Branch 2002).They were separated in all C. 'Mombolo' specimens, but in contact in two (Fig. 4A 4).In C. angolensis there is always a regular, uninterrupted row of six occipitals in broad contact (narrowly medially in CHL 615).
In C. 'Mombolo' a medially-interrupted row (i.e., 2-3 small scales on either side of the head) is always present, as well as a row of 5-6 post-occipitals.A medially-interrupted row of small occipitals also characterises C. namakuiyus (2-4 on either side of head; plus 6-8 post-occipitals; the holotype has 2-3 occipitals on either side of the head and 7 post-occipitals), and C. phonolithos (3-5 occipitals on either side of head; plus 6-7 post-occipitals; the holotype has 4 occipitals on either side of the head, but on the right side there is an additional two scales between the occipitals and the 10 post-occipitals), but in C. machadoi there may be a continuous row of 6-8 occipitals, plus a continuous row of 6-8 post-occipitals (Fig. 4), although the holotype has four occipitals on the left and three on the right, narrowly interrupted medially by the fifth (from the left) of seven post-occipitals (Laurent 1964 2019) indicated that the posterior parietals of C. phonolithos were fully separated by a large scale that is also in contact with the interparietal.However, such a scale was absent in all seven specimens (P to V) of this species photographed for Fig. 4.
Femoral pores and differentiated femoral scales (generation glands) (Table 4, Table S2).Males of all species have large secretory femoral pores.One male C. machadoi (PEM R19782) also has a differentiated scale      (Broadley and Branch 2002).The venter of most species is white to cream, but in C. machadoi it is sometimes mainly yellow (Fig. S4) or a mixture of yellow and cream, and it may be cream-yellow in C. namakuiyus.
Colour of the iris of the eye (Fig. 5).This character needs to be photo-recorded using live specimens.In most of the Angolan species (C.angolensis and the three species of the C. machadoi complex) the iris in life is brown (Fig. 5).However, in C. 'Mombolo' most of it is greenish (blueish-green to some eyes), with a narrow brown ring around the black pupil.This eye colour is a diagnostic feature that easily separates the latter and its sister species  (Broadley and Branch 2002).In C. vittifer the gulars are usually smooth except for the lateral ones below the ear openings which are often weakly and obtusely keeled (e.g., holotype, Fig. S2), but all scales may be smooth as in var.C; occasionally the laterals are distinctly keeled, and in a few cases most other gulars are also weakly or even moderately (e.g., NMB R8773) keeled.
Cranial skeleton (Fig. 8, Table S3).CT scans of Angolan Cordylus and select members of the East African Cordylus clade (see Table S1) reveal some consistent skeletal differences between the two morphotypes from west-central Angola.The maxilla of C. 'Mombolo' is deeper and the maxilliary process more pronounced than observed in C. angolensis, causing a steeper, shorter articulation with the nasals.The posteriomedial process of the parietal differs between C. angolensis and C. 'Mombolo', the former having a narrow and biconcave "Y"-shaped projection that articulates with the sagittal crest of the supraoccipital, while the latter's process is less laterally compressed, forming an "M"-shaped projection.The posteriomedial process of the members of the C. machadoi complex is variably bifurcated, and the shape of this process is highly variable within these and other species of Cordylus (Lang 1991;Stanley et al. 2016)    Bocage's (1895) description of a single type specimen (MBL 429, adult male) of C. angolensis was fairly terse and he did not provide any illustrations.He described the dorsal parts of the head and back as being a variety of brown and black on a pale tan, with two longitudinal series of small irregular whitish spots along the back, the tail brown, with venter white.Regarding scalation he noted only that: Internasal (= frontonasal) rhomboid, in contact with the anterior part of the frontal, thus separating the prefrontals; loreal absent; preocular large and in contact with the posterior part of the nasal; nasal large, slightly domed, pierced halfway up the posterior edge; gular with quadrangular scales, smooth, narrow, juxtaposed, of various shapes, and significantly smaller than in C. cordylus; dorsals in 26 longitudinal and 24 transverse rows; ventrals in 16 longitudinal and 27 transverse rows; two large pre-anal scales separated on the median line by two small plates; pre-anal pores 6 [presumably number of femoral pores on each thigh]; snout-vent length = 74 mm; tail length = 78 mm (incomplete).Head width/ SVL = 24.3%;head length/SVL = 31.1% [based on head and body measurements provided].
In 1968 Dr Donald G. Broadley (in litt., 18 January 2012) examined the holotype (MBL 429) of C. angolensis at the Museu Bocage in Lisbon and recorded the following data: nasals in broad contact, loreal absent (fused with preocular); frontal and frontonasal in contact, suboculars 3, supraciliaries 3, gulars 26, dorsals in 24 transverse and 26 longitudinal rows, ventrals in 27 transverse and 16 longitudinal rows, femoral pores 6 on each thigh; outer three rows of ventrals keeled and mucronate; SVL = 78 mm; tail length 74 mm tail (broken).As mentioned earlier, Broadley (1971: 22) noted that the tail of the holotype was "almost equal to snout-vent length in spite of the tip being missing".
According to Madruga (2011: 4), despite the fire at Museu Bocage in 1978, "many of Bocage's papers, folder, manuscripts and correspondence still exist today".From February to April 2012 MFB corresponded with Catarina Madruga (Museu Bocage) who searched through Bocage's correspondence and notes in the hope of finding information and/or diagrams of the holotype of C. angolensis, but nothing was found.The holotype of C. angolensis is no longer available, so in order to stabilise the taxonomy we designate a neotype that is in agreement with Bocage's (1895) description of the species (see above).
The new specimens from Taqueta Mountain were collected about 100 km west of Caconda and are assigned to C. angolensis on the basis of their similarity to the holotype as described by Bocage (1895) and Broadley (pers.comm.)(see details above).Their colour patterns are similar, as are several of the scale counts.Bocage (1895) noted that the frontonasal was in contact with the frontal in the holotype; and while this was the case in only two of the five new specimens, this is clearly a variable character.Bocage's (1895) holotype of C. angolensis reportedly had 16 longitudinal rows of ventrals, whereas the Taqueta Mountain specimens examined here all had 14 rows.Such discrepancies probably reflect the likelihood that different observers count these rows differently, e.g., some observers may exclude a poorly differentiated lateral row that others count.A similar problem may occur with regards to the longitudinal rows of dorsals (26 in holotype versus 20-23 in new material), transverse rows of ventrals (27 versus 22-24) and gulars across the throat (26 versus 20-21).
Neotype.The holotype (MBL 429) is no longer in existence, and therefore we propose and describe one of the new specimens as neotype: MNCN 50648, an adult male from Taqueta Mountain, Benguela Province, Angola (-13.7778°,14.1794°; 2112 m a.s.l.) collected 10 August 2017 by Afonso Vaz Pinto and Pedro Vaz Pinto.Cordylus angolensis is distinguishable from other members of its genus by the following combination of characters: (1) back dark brown with a paravertebral series of pale markings; (2) top of head plain brown or with occasional pale blotches; (3) iris of the eye brown; (4) scales of the first transverse row of dorsals similar in appearance to those of the row behind; (5) loreal shield absent; (6) nostril pierced in the posterior part of a large nasal, situated behind the suture of rostral and first supralabial, usually well separated from both the first supralabial and the preocular; (7) a regular row of six enlarged, non-spinose occipitals; (8) Frontonasal separated from the frontal by a pair of prefrontals (each of which usually exceeds it in size) or in contact; (9) Anterior pair of parietals usually in contact anteriorly; (10)  Its status as a distinct species is supported by monophyly with high levels of support from a suite of three mitochondrial and six nuclear markers (see above); and it differs from C. 'Mombolo' (see below), the most similar species genetically and morphologically, by an uncorrelated ND2 p-distance of 9.22% (Table 3).[Boulenger, 1910], and C. vittifer [which occasionally has a loreal]) in lacking a loreal.It differs from C. ukingensis and C. macropholis by virtue of its smooth (rarely with a few weak keels on a few scales) versus strongly keeled (even spinose in C. macropholis) gulars, as well as by having its nostril pierced near the middle of the posterior part of the nasal scale (versus infero-posteriorly).Differs from C. tropidosternum by having smooth versus keeled gulars, and having the nostril well separated from the first supralabial (not in contact or near-contact); from C. rhodesianus by having distinctly rugose versus finely rugous to smooth upper head shields, 24-25 versus 25-29 transverse rows of dorsals, and a straight versus curved sulcus dividing the posterior part of the nasal; from C. jonesii by having its nostril pierced halfway up the posterior edge of the nasal rather than towards the centre, and a straight versus curved sulcus dividing the posterior part of the nasal; and from C. marunguensis which has the nostril pierced centrally on the lower margin of the nasal.Distinguished from C. vittifer and C. machadoi, C. namakuiyus and C. phonolithos by always lacking a loreal scale, having most or all scales of the first transverse row of dorsals of similar length (rather than longer) than those of the next row, and by having a pair of paravertebral rows of pale greenish-cream spots or blotches versus a lack of these.Most similar to C. 'Mombolo' (see below).

Comparisons with other
Description of the neotype.MNCN 50648 (Figs 4-6, 8-11; Tables 5, 6).External morphology: Head and body moderately depressed.SVL 88.4 mm.Tail 62.8 mm (tip missing), 0.71+ times as long as SVL.Head 24.6 mm, 1.18 times as long as wide (20.9 mm); head depth 10.6 mm, 50.6% head width.Upper head shields coarsely rugose.Nasals in broad contact; frontonasal quadrangular and slightly smaller than a prefrontal; prefrontals in contact with one another (separating frontonasal from frontal), and each one in contact with the frontal, first supraocular, preocular and a large nasal on either side; frontal in contact with first and second supraoculars, followed by a pair of frontoparietals in median contact; interparietal flask-shaped, narrowing and extending anteriorly to completely separate the anterior parietals; each anterior parietal smaller than a posterior parietal; a straight row of six rugose but non-spinose occipitals.Four supraoculars and three supraciliaries.Nasals large, with nostril pierced more-or-less centrally in the posterior part of the scale; nostril separated from preocular by a distance of more than half its diameter, distinctly separated from first supralabial by a distance almost equal to the height of the latter scale (and almost equal to the greatest extent of the nostril); nasal divided posteriorly at the level of the middle of the nostril.Loreal absent.Lower eyelid with 9-10 vertical septa.Three suboculars, well separated from the lip, with another large scale bordering the eye between posterior subocular and postocular.Rostral semi-divided dorsally and about twice as broad as deep; supralabials 6; infralabials 6; sublabials 5. Mental almost twice as broad as long.Gulars vary considerably in size and shape, from rectangular (especially on the sides) to square and oval, mostly smooth (a few centrally have feeble blunt keels medially) and at most subimbricate, forming about 12 transverse rows (from first row in line with angle of jaws to last distinct row posterior to chin shields), and 20 longitudinal rows between posterior angles of jaws; 5 chin shields (including tiny median scale anteriorly) in contact with 1 st pair of sublabials; two pairs of distinctly enlarged chin shields posterior to these.Dorsal scales rectangular, rugose, moderately keeled (less so medially), seldom spinose or mucronate, not serrated at the posterior edges; laterals oval, juxtaposed, rugous, sharply keeled and moderately spinose, no additional spines on the free end of scales; dorsals plus laterals in 25 transverse rows and 20 longitudinal rows (vertebral scales slightly smaller); on the central part of the belly the mesial ventrals are rectangular (transversely) and larger than others, the next row on either side with slightly rectangular scales, including the next two rows on either side; ventrals mostly smooth, but 2-3 lateral rows on either side with some obtusely keeled scales; ventrals in 22 transverse and 14 longitudinal rows (plus a row of oval, keeled scales on either side); a pair of enlarged and somewhat oval pre-cloacal plates is followed anteriorly (before the ventrals) by two transverse rows of much smaller irregular scales.
Scales on limbs above are large, strongly keeled and spinose; scales under fourth (longest) toe 14 on left and right feet; femoral pores 6 on left thigh, 5 on right, with distinct plugs of yellowish secretion; differentiated glandular femoral scales on thigh 21 on left, 20 on right.Tail with whorls of large, elongate, strongly keeled, spinose (spines directed backwards and longest superolaterally), weakly serrated, scales; supracaudals strongly keeled throughout most of their length, subcaudals basally distinctly keeled mainly on the distal half.Scales on palms of hands and soles of feet moderately to obtusely keeled; supradigital scales of hands smooth to weakly keeled, of feet smooth to moderately keeled; subdigital scales of hands weakly keeled, of feet moderately keeled.
Colour: The colour pattern of the neotype in life is similar to that described for the holotype (see above).The back is a mixture of light and dark brown, with some of the darker markings forming ill-defined transverse bands, and there is a series of greenish-cream blotches, mostly elongated, arranged in close proximity paravertebrally (about eight pairs).Top of the head is a mixture of pale and dark brown, with an ill-defined pale median band longitudinally.The upper parts of the tail are brown, with some keels dark brown.Belly is dirty white and the throat cream.Bocage (1895) recorded the markings on the back as 'whitish', but this may have been in reference to preserved material that had faded somewhat.
Cranial skeleton (Fig. 8, Table S1): Segmented mesh file of MNCN 50648 can be found here http://tinyurl.com/CordylusAngola.Morphosource link to tomogram stacks doi.org/10.17602/M2/M529985.The scales of the dorsal and temporal regions of the skull and the ventrolateral aspects of the jaws are underlain with rugose osteoderms.These osteoderms fuse to the parietal, frontal and postor-bital bones, although the mesokinetic and metakinetic joints appear unobstructed and flexible.Lateral maxilla and anterior aspect of the premaxilla lack osteoderms.The parietal is pentagonal, with five osteoderms that underlie the parietal shields fused to its dorsal surface, and a narrow, bifid, 'Y'-shaped medioposterior process that articulates with the sagittal crest of the supraoccipital.Three large osteoderms are fused to the frontal, which is unpaired and clasped by the parietal at its posterolateral edge.The upper temporal fenestra is obscured anteriorly by a large osteoderm fused to the dorsal surface of the postorbital bone, and posteriorly by two unfused rectagonal osteoderms that overlie the squamosal.Premaxilla is unpaired and contains seven pleurodont teeth, with a dorsal process that extends posteriorly to intersect the nasals, which themselves overlie the frontal.The maxillae are typically scinciform, with a deeply grooved crista dentalis, a deep fossa that lodges the lacrimal sac, 9 (left) or 11 (right) lateral foramina, and 22 (left) or 23 (right) teeth.Teeth display pleurodont attachment and are unicuspid, with a slight concave surface where they connect with the mandibular teeth.No palpebral is present but the prefrontals possess a small protuberance, forming a shelf that directly underlies the anteriormost superorbital osteoderm.The jugal is triangular in cross-section and asymmetrically T-shaped, with a tapering anterior process and a broad, truncated posterior process that extends along and past the posterior edge of the maxilla.The lacrimal is small, flattened and oval, slightly bicuspid anteriorly, the two processes meeting prefrontal cusps to bracket the lacrimal duct.Pterygoids are edentate and extend back to connect with the quadrates, becoming C-shaped in cross-section posterior to the epipterygoid condyle.The squamosal is curved and blade-like, circular in cross-section anteriorly, becoming flattened posteriorly, where it articulates with the cephalic condyle of the quadrate and the supratemporals.Supratemporals are flattened, ovoid and not fused with the elongate paraoccipital processes.The posterior aspect of the prootic not fully fused with the otooccipital, resulting in a deep groove along the dorsal aspect of the paraoccipital processes.Quadrates very broad with a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin.The supraoccipital has a strong sagittal crest that extends posteriorly to contact the ventral surface of the medioposterior process of the parietal.The prootic bears an extended alar process and a well-developed, rhomboid christa prootica, and a very weak supratrigeminal process.Basipterygoid processes are well developed and flattened.The lower jaw possesses a large adductor fossa, a highly flattened and medially extended retroarticular process, a medially open Meckelian canal that is Osteoderms: (Fig. 9) The dorsal trunk is covered in rectangular, dorsomedially keeled, imbricate osteoderms, each roughly three times as long as wide.Dorsal osteoderms are arranged in whorls and become progressively more oval and better separated laterally.The nuchal osteoderms are spined posterior to the tympanic opening.Ventral osteoderms are delicate and plate-like, and con-centrated in the gular, antero-pectoral and abdominal regions.No precloacal osteoderms.The forelimbs are covered in keeled, imbricate, rhomboid osteoderms, except for the axillary, antecubical and palmar regions, which are unarmoured.The hindlimbs are covered in rhomboid osteoderms, except for the ventral femoral, popliteal and plantar regions.Hindlimb osteoderms are unkeeled on the anterior thigh, and become more spinose posteriorly and distally.The caudal osteoderms are large, robust and arranged in imbricated whorls, feebly keeled and mucronate along the dorsal and ventral aspects, becoming more heavily spined laterally.4-6; Tables 5, 6).External morphology: Detailed morphometrics and mensural data for the new material is presented in Tables 5-6.All additional material agrees in general with the holotype and neotype, but differs as follows: Less than half the length of the nasals are separated by the frontonasal in CHL 613; frontonasal pentagonal in MNCN 50649; size of the frontonasal varies from much smaller than a prefrontal scale (CHL 613), only slightly smaller (CHL 611), about equal to or only slightly larger (CHL 615) or distinctly larger (MNCN 50649); frontonasal and frontal in narrow (CHL 611) and broad (MNCN 50649) contact, separating frontonasal and frontal.Shape and size of the interparietal varies: usually kite-shaped (with most acute angle anteriorly) and separating more than Habitat.Although no specific habitat information was available in Bocage's (1895) original description, the morphology of this species suggested that it is rupicolous, as are most other congeners (Branch 1998).The type locality of Caconda, where José de Anchieta was based, lies in relatively flat terrain between 1500 and 1700 m a.s.l.The local habitat there today comprises mostly agricultural fields, interspersed with very occasional, scattered, small granite outcrops.In the past, the region of Caconda was probably covered by well-developed miombo woodlands, characterised by dominance of tree species such as Julbernardia paniculata, Brachystegia spiciformis and Brachystegia spp.(Barbosa 1970).Nevertheless, the species has never again been found in the vicinity of Caconda, but was instead found almost 100 km to the west on Taqueta Mountain, at 2112 m a.s.l., where the habitat is typical of the Angolan highlands (Fig. 11B).All specimens were found in cracks and crevices in granite outcrops present on moderate to steep slopes, surrounded by well-developed montane grassland in proximity to Afromontane forest patches in deep ravines.

Variation in additional new material. (Figs
Distribution.Although the original type locality is given as Caconda (Huíla Province), efforts over more than one century to rediscover the species at or near that locality have failed.The fact that Anchieta was based at Caconda for more than 20 years (Banha de Andrade 1985), and the holotype appears to be the only specimen collected, suggests that the species was likely rare there in those days possibly due to scarcity of rocky habitat, which subsequently must have been further impacted by human encroachment and use of rocks for construction.Alternatively, the holotype may have also been collected on the mountains criss-crossed by the 19 th century land routes linking Caconda to the major coastal city of Benguela, a route that was often followed by Anchieta himself and his assistants.While conducting surveys in this region in 2017, and specifically targeting the mountains thought to have been along those historical routes, we made the discovery of a population of C. angolensis at Taqueta Mountain.This new locality is currently the only site where this species Table 5. Meristic data for scalation characters in Cordylus angolensis and Cordylus momboloensis sp.nov.from Angola.When different, the number of scales on the left and right sides of the head are separated by a slash symbol (/).For occipitals, when the row is medially interrupted by post-occipitals, the number of scales on the left and right sides of the head respectively are separated by a plus sign (+).  is known to occur with any certainty (Fig. 12).Nevertheless, it is likely to be present on various rocky hills on the highlands west of Caluquembe, along the ridge of the southern Angolan escarpment, where suitable habitat is present.
Conservation.This species is known from only one locality, and little is known about its biology or population trends.Nevertheless, it should be noted that montane habitats in Angolan highlands are currently under anthropogenic pressure due to widespread excessive burning, cut ting of remnant forests, and overgrazing of grasslands, all of which may negatively affect the species.Cordylus momboloensis sp.nov. is distinguishable from other members of its genus by the following combination of characters: (1) back dark brown with a paravertebral series of pale markings; (2) top of head with pale blotches; (3) iris of the eye mostly pale green (bluegreen to some eyes), with a brown ring around the pupil; (4) scales of the first transverse row of dorsals similar in appearance to those of the row behind; (5) loreal shield absent; (6) nostril pierced in the posterior part of a large nasal, situated behind the suture of rostral and first supralabial, always well separated from the first supralabial, and usually separated from the preocular; (7) an interrupted row of non-spinose occipitals consisting of 2-3 scales on either side of head; (8) a row of 5-6 non-spinose post-occipitals, the median scales of which are in contact with the posterior parietals and separate the occipitals; (9) Frontonasal separated from the frontal by a pair of prefrontals, each of which exceeds it in size; (10) Anterior pair of parietals usually in contact anteriorly; (11) dorsolateral and lateral scales weakly to moderately spinose; (12) tail spinose, more weakly so distally; (13) dorsal scale rows transversely 22-24; (14)  Its status as a new species is supported by monophyly with high levels of support from a suite of three mitochondrial and six nuclear markers (see above); and it differs from C. angolensis, the most similar species genetically and morphologically, by an uncorrelated ND2 p-distance of 9.22% (Table 3).
Comparisons with other Cordylus species.The new species differs from all other Cordylus in having nine (rather than seven, rarely six) premaxillary teeth, and from most other congeners (except C. angolensis, C. ukingensis, C. macro pholis, and C. vittifer [occasionally has a loreal]) in lacking a loreal.It differs from C. ukingensis and C. macropholis by virtue of its weakly keeled or smooth versus strongly keeled (even spinose in C. macropholis) gulars, as well as by having its nostril pierced near the middle of the posterior part of the nasal scale (versus infero-posteriorly).Differs from C. tropidosternum by having smooth or weakly keeled versus keeled gulars, and having the nostril well separated from the first supralabial (not in contact or near-contact); from C. rhodesianus by having distinctly rugose versus finely rugose to smooth upper head shields, 22-24 versus 25-29 transverse rows of dorsals, and a straight versus curved sulcus dividing the posterior part of the nasal; from C. jonesii by having its nostril pierced halfway up the posterior edge of the nasal rather than towards the centre; and a straight versus curved sulcus dividing the posterior part of the nasal; and from C. marunguensis which has the nostril pierced centrally on the lower margin of the nasal.Distinguished from C. vittifer, C. machadoi, C. namakuiyus and C. phonolithos by always lacking a loreal scale, having most or all scales of the first transverse row of dorsals of similar length (rather than longer) than those of the next row, and by having a pair of paravertebral rows of pale greenish-cream spots or blotches versus a lack of these.
The new species is most similar to C. angolensis, but distinguished from it as follows: in live specimens the iris of the eye is largely pale green to blueish-green versus brown; top of head distinctly marked by pale blotches versus mostly plain brown; throat usually dirty white versus cream; a medially-interrupted row of occipital scales (2-3 on either side of head) is present, never forming a continuous row, each scale usually smaller than a post-occipital versus a continuous row of 6 occipitals across the back of the head; lower numbers of ventral scale rows longitudinally (usually 12-13 vs 14); relatively longer forearms (forearm/SVL 0.20-0.25 versus 0.15-0.19);higher numbers of generation glands in males (25-37 per thigh versus 19-25); and higher numbers of premaxillary teeth (nine versus seven).Description of holotype.PEM R25217 11,13,Tables 5,6).External morphology: Head and body moderately depressed.SVL 85.1 mm.Tail 77.5 mm (tip truncated and healed), 91.1% SVL.Head length 25.3 mm, 1.19 times as long as wide (21.4 mm), head depth 9.2 mm, 43.2% head width.Upper head shields coarsely rugose, not striated.Nasals in moderate contact, their posterior halves separated by the frontonasal, which is quadrangular and about half the size of a prefrontal; prefrontals in moderate contact with one another, and contacting the supranasals, preoculars and anterior supraoculars on either side; frontal in contact with first and second supraoculars, followed by a pair of frontoparietals in median contact; interparietal quadrangular, separating posterior one-third of anterior parietals, and anterior onethird of posterior parietals, the latter being distinctly larger than the anterior ones; occipitals two on one side of head, three on the other, interrupted by a slightly curved row of six post-occipitals; occipitals and post-occipitals rugose, occasionally striated but essentially non-spinose.Four supraoculars and three supraciliaries (postocular also in contact with posterior supraocular).Nasals large, with nostril pierced more-or-less centrally in the posterior part of the scale; nostril separated from preocular by a distance of about half its diameter, distinctly separated from first supralabial by a distance of more than half the height of the latter scale (i.e., at least half the greatest extent of the nostril); nasal divided posteriorly at the level of the middle of the nostril, and also divided below the nostril, forming a small postnasal.Loreal absent.Lower eyelid with 6-7 vertically enlarged septa; suboculars 4 left and 3 right, well separated from the lip.Rostral 2.3 times as broad as deep; supralabials on left side six (four anterior to median subocular), right side five (three anterior to median subocular); infralabials 6; sublabials 5. Mental 1.9 times as broad as long; gulars vary considerably in size and shape, from elongate, sometimes rectangular, on the sides of the throat, to squarish, oval and irregular, sometimes pentagonal or hexagonal, smooth and at most subimbricate, forming about 12 transverse rows (from first row in line with angle of jaws to last distinct row posterior to pair of sublabials), and 26 longitudinal rows between posterior angles of jaws; 5 chin shields (including a tiny elongate scale medially) in contact with 1 st pair of sublabials.Dorsal scales closely-set but often juxtaposed, rectangular, rugose, moderately keeled (less so on middle of back), some are also weakly spinose dorso-laterally, occasionally weakly serrated at their posterior edges; laterals occasionally subimbricate, often oval, rugose, sharply keeled and moderately spinose, no additional spines or serrations on the free ends of scales; dorsals (including laterals) in 22 transverse rows (excluding one half-row) and 23 longitudinal rows (including row of small vertebral scales which is interrupted at parts); on the central part of the belly the paired rows of mesial ventrals are rectangular (transversely), the next row on either side with only slightly transversely rectangular scales, the others being squarish, with two rows of lateral ventrals on either side consisting of longitudinally-rectangular scales; ventrals mostly smooth, but the outermost lateral row on either side, and an additional row of 'pseudo-ventrals' on either side, with some obtusely keeled scales; ventrals in 22 transverse and 12 longitudinal rows (excluding a row of oval, keeled scales on either side best considered 'pseudo-ventrals'); a pair of enlarged pre-cloacal plates is followed anteriorly (before the ventrals) by 2-3 transverse rows of much smaller plates.
Scales on upper parts of hindlimbs are large, strongly keeled and strongly spinose; scales of forelimbs large, moderately keeled and moderately spinose; scales under fourth (longest) toe 13 on left foot, 14 on right; scales under fourth (longest) finger 10 on left hand, 11 on right; femoral pores 7 on each thigh, of moderate size with distinct plugs of yellowish secretion; differentiated glandular femoral scales on thighs 26 on left, 27 right.Tail whorled, dorsally with large, elongate, rugose, strongly keeled, spinose, weakly serrated scales, with spines directed backwards and longest superolaterally; supracaudals strongly keeled throughout most of their length, subcaudals basally distinctly keeled mainly on the distal half of tail.Scales on palms of hands obtusely keeled, on soles of feet moderately keeled; supradigital scales of hands weakly keeled, of feet moderately keeled; subdigital scales of hands and feet weakly keeled or smooth.
Colour of holotype (in life): The upper parts of the head, back, flanks, limbs and tail were mostly dark brown, with irregular pale or light greenish to blueishgreen markings on either side of the midline (about six pairs) and on top of the head; venter whitish to cream; throat dirty white (Fig. 13).After preservation: As above, but back with indications of dark bands, and pale dorsal markings and venter have faded to a dirty white colour.
Colour of paratypes (in life): Colouration of the Sandula and Vondo paratypes was similar to that of the holotype.The belly had a slightly blueish tinge.
Colour of paratypes (in preservative): AMNH R47331 is the most distinctly marked of the old 'Mombolo' series (Fig. 14): The upper parts of the head, back, flanks, limbs and tail are medium to dark brown; there are five dark transverse bands across the back from nape to tail, separated by narrower bands of paler colour, including three pairs of large and distinct cream-coloured paravertebral blotches, together with a series of less distinct but similarly pale blotches dorso-laterally; and a few cream spots are present on the top of the head.In the other AMNH specimens the upper parts of the head, back, flanks, limbs and tail are medium to dark brown, but without distinct dark markings or pale spots (pattern may have faded).Venters of all AMNH specimens are cream to pale tan.
Cranial skeleton: (Fig. 8; segmented mesh files of paratype AMNH R47334 and TM 46476 can be found here (tinyurl.com/CangolensisAMNH)and here (tinyurl.com/CordylusAngola2). Morphosource links to tomogram stacks (https://doi.org/10.17602/M2/M530010and https://doi.org/10.17602/M2/M530028).Consistent characters between both specimens are reported here, with any differences noted.The scales of the dorsal and temporal regions of the skull and the ventrolateral aspects of the jaws overlie rugose osteoderms.These osteoderms fuse to the proximal parietal, frontal and postorbital bones, although the mesokinetic and metakinetic joints appear unobstructed and flexible.Lateral maxilla and anterior aspect of the premaxilla lack osteoderms.The parietal is pentagonal, with five osteoderms that underlie the parietal shields fused to its dorsal surface, and tapers to a bifid medioposterior process, which articulates with the sagittal crest of the supraocciptial.Three large osteoderms are fused to the frontal, which is unpaired and clasped by the parietal at its posterolateral edge.The upper temporal fenestra is obscured anteriorly by a large osteoderm fused to the dorsal surface of the postorbital bone, posteriorly by two unfused rectagonal osteoderms that overlie the squamosal.Premaxilla is unpaired and contains nine pleurodont teeth, with a dorsal process that extends posteriorly to intersect the nasals (absent due to damage in AMNH R47334), which themselves overlie the frontal.The maxillae are typically scinciform, with a deeply grooved crista dentalis, a deep fossa that accommodates the lacrimal sac, and 20-21 (TM 46476) to 18-19 (AMNH R47334) teeth.Teeth display pleurodont attachment and are unicuspid, with a slight concave surface where they connect with the mandibular teeth.No palpebral is present but the prefrontals possess a small protuberance, forming a shelf that directly underlies the anteriormost superorbital osteoderm.The jugal is triangular in cross-section and asymmetrical-ly T-shaped, with a tapering anterior process and a broad, truncated posterior process that extends along and past the posterior edge of the maxilla.The lacrimal bone is small, flattened and oval, slightly bicuspid anteriorly, the two processes meeting prefrontal cusps to bracket the lacrimal duct.Pterygoids are edentate and extend back to connect with the quadrates, becoming C-shaped in cross-section posterior to the epipterygoid condyle.The squamosal is curved and blade-like, circular in cross-section anteriorly, becoming flattened posteriorly, where it articulates with the cephalic condyle of the quadrate and the supratemporals.Supratemporals are laterally flattened, ovoid and not fused with the elongate paraoccipital processes.The posterior aspect of the prootic in the smaller AMNH R47334 is not fully fused with the otooccipital, resulting in a deep groove along the dorsal aspect of the paraocciptal processes.Quadrates very broad with a pronounced ridge and concave region at the entire lateral edge in TM 46476 (right quadrate of AMNH R47334 partially damaged).The supraoccipital has a strong sagittal crest that extends posteriorly to contact the ventral surface of the medioposterior process of the parietal.The prootic bears an extended alar process and a well-developed, rhomboid christa prootica, and a very weak supratrigeminal process.Basipterygoid processes are well developed and flattened.The lower jaw possesses a large adductor fossa, a highly flattened and medially extended retroarticular process, a medially open Meckelian canal that is closed posteriorly by a large splenial, and a dentary with a strong subdental shelf; 22 (AMNH R47334) and 25-26 (TM 46476) mandibular teeth, and 6-7 dentary foramina.
Osteoderms (Fig. 9): The dorsal trunk is covered in rectangular, dorsomedially keeled, imbricate osteoderms, each roughly three times longer than its width.The dorsal osteoderms are arranged in whorls, becoming more oval and better separated laterally.The nuchal osteoderms are spined posterior to the tympanic opening.Ventral osteoderms are absent in AMNH R47333, and delicate and plate-like in TM 46476, grading from rhomboid in the gular region to mildly imbricated and cycloid in the pec-toral region, square and non-imbricate in the abdominal region, to pentagonal in the cloacal region.The forelimbs are covered in keeled, imbricate, rhomboid osteoderms, except for the axillary, antecubical and palmar regions, which are unarmoured.The hindlimbs are covered in rhomboid osteoderms, except for the ventral femoral, popliteal and plantar regions.Hindlimb osteoderms are unkeeled on the anterior thigh, become more spinose posteriorly and distally.The caudal osteoderms are large, robust and arranged in imbricated whorls, feebly keeled and mucronate along the dorsal and ventral aspects, becoming more heavily spined laterally.
Variation in additional material (n = 2).(Figs 4-6, 8-9, 15; Tables 5, 6) External morphology: TM 46476 (female).SVL 106.0 mm.Original tail 106.7% SVL.Head 1.25 times as long as wide, head height 0.469 times head width.Nasals in broad contact.Interparietal bulb-shaped with a long neck, completely separating the anterior parietals.Anterior and posterior parietals similar in size and shape.Nostril in contact with preocular, nasal scale apparently not divided, and no postnasal present.Lower half of preocular with a vertical suture medially, suggesting partial fusion with loreal.Lower eyelid with five vertically enlarged septa.Rostral 2.8 times as broad as deep.Supralabials 5 left, 7 right.Mental 1.49 times as long as broad.Gulars mostly rectangular and obtusely and feebly keeled, imbricate to subimbricate, forming nine transverse rows (from first row in line with angle of jaws to last distinct row posterior to pair of sublabials), and 23 longitudinal rows between posterior angles of jaws.Four chin shields in contact with first pair of sublabials.Dorsals in 24 transverse and 20 longitudinal rows.Ventrals in 22 transverse and 12 (plus one row of 'pseudo-ventrals' on either side) longitudinal rows.Pair of cloacal plates preceded by two transverse rows of much smaller plates.Scales under fourth toe 13 (both sides), under fourth finger 12 (both sides).Femoral pores 7 left, 5 right; no differentiated femoral scales.Subdigital scales of feet strongly keeled and spinose, especially posteriorly.
Colour of additional material (in life).TM 46476: Upperparts medium brown with a paired series of irregular cream paravertebral blotches from nape to base of tail; top of head with a few small, scattered, cream markings.FKH 0129 was similar to the type series.
Etymology.The name 'momboloensis' derives from the local name ("m'bolo") for Mombolo which in Umbundo -main language in central Angola -is applied to bread or staple food.It is a reference to the fact that the region in question consists of a fertile and relatively large highland plateau (surrounded by mountains, in Luanza Sul Province).Most of the available specimens, including the holotype as well as the first series of specimens of this species collected during the Vernay Angola Expedion in 1925, were collected in Mombolo.
Habitat.Although no specific habitat information is available for the material housed at the AMNH, it is like- ly that it may have been obtained at the same site as the recent material from Sandula.At both the Sandula and Vondo localities, the species proved to be rupiculous and specimens were found sheltering inside cracks and crevices in rocky outcrops (igneous rocks in Sandula and granite in Vondo) present in fairly steep slopes above 1900 m a.s.l.(Fig. 11D).The surrounding habitat at Sandula and Vondo is montane grassland.The specimen from Condé was reportedly found in a rock outcrop surrounded by grassland and sparse miombo savannah (see Fig. 15B).
Distribution.This species was collected at Sandula (Monte Verde) and Vondo (Mount Uassamba), both in Cuanza Sul Province on high mountain peaks at 1975-2095 m a.s.l., and an individual was briefly observed at Morro do Pundo (-12.4464°,13.9171°) in Benguela Province, at approximately 1000 m elevation near the town of Bocoio (Fig. 12).The Bocoio specimen quickly disappeared into a deep crevice and only a tail tip was secured.The latter was analysed for 16S and the specimen was referred to this species.The two peaks at Sandula and Vondo are situated about 110 km apart in highlands that include several mountain chains that have never been surveyed before, suggesting that the species is widespread on the west-central Angolan plateau in suitable mountain habitat, while the specimen from 3 km W of Condé places the species further north at elevations of about 1236 m a.s.l. in miombo woodlands (Fig. 12).
The AMNH material was collected by 'H.Chapman and A. Chapman' in 1925 and simply labeled as 'Mombolo'.This location, however, is rather vague as the region of Mombolo covers about 1000 km 2 , and included three farms belonging to the Chapman family.A very detailed discussion on the problems in relation to another rupiculous lizard, Afroedura bogerti, collected under identical circumstances (Loveridge 1944b), is provided by Branch et al. (2017).It can be noted that Hill and Carter (1941) stated that "Boulton and Charles Chapman went to Namba in the Mombolo region."Two of Chapman's farms were located on the separate mountain chains of Namba and Monte Verde (Sandula), suggesting that the geographic location could be confidently narrowed down to those two sites.However, several recent surveys conducted on both mountains have so far resulted in Cordylus being found only at Sandula, suggesting that this was the original locality where the Chapman specimens were collected.

Conservation.
Little is known about the biology of this species, but habitat at the four sites where it has been found appears to be relatively undisturbed.Given the current knowledge, if an assessment was to be done, it is likely to result in a classification of Least Concern based on IUCN (2022a) Standards and Petitions Committee criteria.and colonise their current territory.Cordylus jonesi and C. rhodesianus form part of the continuous corridor of the eastern clade, co-occuring with C. vittifer in the south and C. tropidosternum in the north (Fig. 16).Cordylus jonesii has a broader distribution than most Cordylus species and, like the similarly widespread C. tropidosternum, is an ecologically unusual cordylid, often eschewing the typical cordylid rock-crevice retreat sites for tree bark and hollow tree limbs in mopane woodlands (Branch 1998).If the reduced reliance on rocky habitat seen in C. jonesii represents the plesiomorphic condition for this clade, the ancestors of Angolan Cordylus may have been able to disperse through the dry woodlands of the southern African Miocene via Botswana, Zimbabwe, Zambia and northern Namibia (Bamford 2000), reverting to obligate rupicoly upon reaching the rocky habitats of the Angolan escarpment and coastal desert.The current diaspora in the northern Cordylus clade implies that the intermediate habitat must have become unsuitable following the westwards dispersal, with aridification of the Kalahari in the mid-to late-Miocene providing one explanation of how this isolation could have occurred (Hoetzel et al. 2015).

Key to the genus Cordylus in Angola
Recent studies on other rupicolous lizards in the south-western quarter of Angola have indicated similar distribution patterns to Cordylus.In both Angolan Afroedura and Cordylus, the main divide splits one clade in the west-central highlands and one clade widely distributed along the south-western coastal plain and escarpment.The six species of Angolan flat geckos (Branch et al. 2021;Conradie et al. 2022b) show a distribution pattern similar to the five species of Cordylus (Fig. 12): in the west-central highlands the A. bogerti group is similarly distributed to the C. angolensis group, and in the south-western group the other three species of Afroedura have a similar distribution to the C. machadoi group (high elevations at the top of the southern escarpment: A. vazpintorum and C. machadoi; Serra da Neve inselberg: A. praedicta and C. phonolithos; widely distributed on the semi-arid coastal plain: A. donveae, A. vazpintorum, C. namakuiyus).Hemidactylus benguellensis is another gecko restricted to south-western Angola (Lobón-Rovira et al. 2021), with a distribution similar to the C. machadoi group as its range includes the lowlands (similar to the range of C. namakuiyus), highlands (similar to C. machadoi) and Serra da Neve (habitat of C. phonolithos).
These similar distribution patterns strongly suggest that the same evolutionary forces may have shaped the speciation history of these genera in Angola.The Angolan west-central highlands, corresponds to the marginal mountain chain or ancient massif (Huntley 2019), and includes most of the country's highest peaks and Afromontane ecosystems.These highlands are recognized as an important centre of biodiversity endemism, as part of a continent-wide archipelago of relic montane grasslands and forests (Huntley 2019).Radiation of saxicolous lizards speciating allopatrically in 'rock islands' has been well documented (e.g., Branch et al. 2021;Conradie et al. 2022b), and our data suggest that the species currently present in the Angolan central highlands, such as C. angolensis and C. momboloensis sp.nov., represent rela-tively old lineages, while speciation in the south-western clade is more recent and may not be fully resolved.
Despite several recent studies on the herpetofauna of Angola (see references in Introduction), more surveys will be required to allow an evaluation of the true distribution of both C. angolensis and C. momboloensis sp.nov., and to investigate the possibility of additional congeners occurring in the country.The highlands of central Angola are still relatively poorly known biologically and barely explored, but it is likely that future work will reveal similar speciation patterns in other rupiculous taxa, which may significantly contribute to a better interpretation of Angolan biogeography in general.Our findings underline the importance of the Angolan west-central highlands in functioning as a regional reservoir of biodiversity, and highlight the need to conserve its highly threatened montane habitats.

Conclusion
Our evaluation of genetic and morphological data indicates that two species of allopatric, rupicolous Cordylus, C. angolensis and C. momboloensis sp.nov., occur in the west-central highlands of Angola.Cordylus angolensis was until now known only from the holotype collected 128 years ago.We also show that these two species form a clade most closely related to the three species of the C. machadoi complex in south-western Angola, and together the Angolan assemblage of Cordylus is a distinct clade most closely related to east African congeners, especially C. jonesii + C. rhodesianus.

Figure 3 .
Figure 3. Maximum likelihood (ML) phylogenetic tree for Cordylus based on concatenated dataset, with Bayesian inference (BI) support overlaid.Numbers above the key nodes indicate BI posterior probability (≥ 0.95 were considered supported), while below the key nodes indicate ML bootstrap values (≥ 95% were considered supported).Major Angolan clades are highlighted in different colours.
C. angolensis.Based on colour photographs in Reissig (2014) and Fig. 1, the iris in C. jonesii, C. tropidosternum, C. rhodesianus and C. vittifer is brown to orange-brown (distinctly orange in C. tropidosternum, Fig. 1E).Texture of gular scales (Fig. 6).Gulars or throat scales are smooth in C. angolensis but weakly keeled at least laterally in most C. 'Mombolo' specimens'.In one specimen (MNCN 50760) of C. 'Mombolo' most of the gulars are obtusely keeled (Fig. 6).Judging from Fig. 6, the gulars are smooth in C. machadoi and C. phonolithos, but in C. namakuiyus two of the three specimens have obtuse keeling laterally on the throat.Most of the gulars are obtusely keeled in C. tropidosternum, smooth in C. rhodesianus, while in C. jonesi they are smooth mesially but feebly keeled laterally Figs 4-11; Tables5, 6

Figure 8 .
Figure 8. Computed tomography images depicting the cranial osteology of the five species of Cordylus from Angola.

Figure 9 .
Figure 9. Computed tomography generated images showing the osteology of the five Angolan Cordylus species.Osteoderms are highlighted in blue-green.

Figure 10 .
Figure 10.Neotype (MNCN 50648) of Cordylus angolensis from Taqueta Mountain, Benguela Province, Angola.A dorsal aspect, B ventral region, C top of the head, D left side of the head, E ventral side of the head.Photos: Pedro Vaz Pinto.

Figure 12 .
Figure 12.Geographical distribution of Cordylus angolensis (dark blue) and C. momboloensis sp.nov.(green) in Angola.The distribution of the three species of the C. machadoi group (C.machadoi [red], C. namakuiyus [pale blue], C. phonolithos [yellow]) is also indicated.Type localities are indicated by stars; for C. angolensis, the original type locality of Caconda is indicated by a black star, while the neotype locality is indicated by a dark blue star.Locality records are from the present study (see species accounts and Appendix 1) as well as Bocage (1895), Laurent (1964), Stanley et al. (2016), Marques et al. (2019), and Lobón-Rovira et al. (2022b).

Figure 13 .
Figure 13.Holotype (PEM R25217) of Cordylus momboloensis sp.nov.from Sandula (Mombolo), Cuanza Sul Province, Angola.A dorsal and ventral aspects, B lateral (right), dorsal and ventral aspects of the head, C underside of the foot, D femoral region of the legs.Although not evident because of the way the holotype was photographed, its tail is intact.Photos: Pedro Vaz Pinto.

Figure 16 .
Figure 16.Geographical distribution of the genus Cordylus, restricted to sub-Saharan Africa.In the key, Angolan species are indicated by an asterisk.Polygons for non-Angolan species are reported from IUCN (2022b).
photographic images of Angolan specimens; Frank Tillack (Museum für Naturkunde der Humboldt-Universität, Berlin) for information about the holotype of C. vittifer and for supplying photographs of this specimen; Aaron Bauer (Villanova University, Villanova) for providing copies of old literature; René Navarro (University of Cape Town) for preparing the C. vittifer map; Luis Verissimo (Catholic University of Angola, Luanda) for information on locality names, habitat and elevations in Angola; Werner Conradie (Port Elizabeth Museum, Bayworld), Krystal Tolley (South African National Biodiversity Institute, Cape Town) and John Measey (Stellenbosch University, Stellenbosch), field

Table 1 .
Cordylidae samples included in this study with museum voucher numbers/field numbers and GenBank accession numbers for nine genes.Voucher specimens are indicated with museum acces- sion numbers: CAS -California Academy of Sciences, San Francisco; CHL -Colecção Herpelogica do Lubango; FKH -Fundaçao Kissama Private Collection, Luanda; MCZ -Museum of Comparative Zoology, Harvard; MNCN -National Museum of Natural Science, Madrid; NMB -National Museum, Bloemfontein; PEM -Port Elizabeth Museum, Gqeberha; UTEP -University of Texas, El Paso.According to W.Conradie (pers.comm.2022),C. machadoi WRB 0040 and 0041 are duplicates of PEM R19782 and R19784 respectively.

Table 2 .
PCR primer details used for this study.
Relative size of the frontonasal scale (Fig.4).In C. machadoi, C. namakuiyus and C. phonolithos the frontonasal is always larger than either of the prefrontals, whereas it is always smaller in C. 'Mombolo' and usually smaller in C. vittifer (of equal size in NMB R4591, NMB R10301-2, 10329; larger in NMB R8149, 8740; absent in NMB R8866).The frontonasal varies in size in C. angolensis (smaller than a prefrontal in three specimens, bigger in two [Fig.4B,C]), C. rhodesianus (smaller than a prefrontal in three specimens, of equal size in NMB R10031, slightly larger in NMB R10034) and C. jonesii (smaller in three specimens, larger in NMB R8652, 8655).

Table 4 .
(Broadley and Branch 2002)for Cordylus from the Angolan clade and various species in the East African clade(including C. jonesii-C.rhodesianus).The first row of data refers to the range of values, the second row is the mean, and the third row is the standard deviation.For C. angolensis, data for the (now lost) holotype, as recorded by Bocage (1895) andBroadley (pers.comm.), was not used in calculations (see discussion in species account), but is indicated in square brackets when it exceeds the ranges recorded here; for C. phonolithos, data is for the holotype and paratype(Marques et al. 2019)and two new specimens (MNCN 50651-2).(M=male,F=female,J=juvenile.-1meansthatsamplesizeforaparticularcharacter was one less than that indicated in the second column, -2 means two less, -3 means three less.*IncludesFKH0129 in Fig.4; **Holotype, paratype, and CHL 848, 851, 852, 1056 and MNCN 50652 in Fig.4).Free State] and var.C: large and secretory; KwaZulu-Natal: moderate to large and secretory).Differentiated femoral scales are present in males of all species, but absent in all females except those of C. vittifer variety C. Cordylus jonesii, C. rhodesianus and C. tropidosternum all have similar numbers of femoral pores per thigh (3-8) to the Angolan species; but often lower numbers of generation glands per thigh (3-21)(Broadley and Branch 2002), although a couple of specimens examined for the present study had high counts (C.rhodesianus: 23-26).scalerowslongitudinally(Table4,TableS2).All newly-collected specimens of C. angolensis have 14 longitudinal rows of ventral scales versus the 11-14 (usually 12-13) in material assigned to C. 'Mombolo'.Bocage (1895) recorded 16 ventral scale rows for the holotype of C. angolensis, but he may have included an extra row of small 'pseudoventrals' on either side.There was much variation in the other species: 11-18 in C. 'Mombolo' is a mixture of pale and dark brown, the darker scales tending to form vague transverse bands.In both species there is always a paired series of irregular, pale, cream-greenish to blue markings/blotches paravertebrally along the back, from the neck to the base of the tail (see figures below).In C. angolensis the markings tend to be smaller and less extensive compared to C. 'Mombolo', which also has more extensive pale markings on the top of the head.The belly is dirty white, but while the throat is of a similar colour in C. 'Mombolo', it is cream in C. angolensis.In C. machadoi, C. namakuiyus and C. phonolithos the central part of the back, top of limbs and top of head are mostly plain medium brown, and the flanks tan to cream-yellow; in the latter two species the flanks are paler, and C. phonolithos has vaguely marked, broad, cream dorsolateral stripes.In Bates et al. 2014;Reissig 2014)5][16][17][18][16][17][12][13][14][14][15][16][17][18] mean  16.2; mean 14.3;  sometimes 18; var.C: 14).Other scale counts (Table 4, TableS2).There is considerable variation in most other meristic scale characters.Regarding the number of gular scales in contact with the first pair of chin shields, C. angolensis (5-6, mean 5.2) and C. 'Mombolo'(4-5, mean 4.5) had the highest counts, and C. vittifer var.A had the lowest (2-3, mean 2.3), but these were the only populations that did not overlap in this character.Colour pattern (in life; Figs1, 4, 6 and other figures below).The back of both C. angolensis and C. C. vittifer the dorsum is brown to orange-brown, usually with darker markings, and may contain dark (occasionally pale) lateral stripes and/or a pale vertebral stripe (e.g.,Bates et al. 2014;Reissig 2014).Cordylus tropidosternum is dark brown above, often with pale or darker markings that may form ragged dorsolateral bands; C. jonesii is light brown with darker blotches, while C. rhodesianus is yellow-brown to black with pale (sometimes bright yel-low) and dark blotches Stanley et al., 2011) large rupicolous Cordylus with a moderately depressed head and body.Referred to Cordylus (rather than any other species of Cordylidae) by the following combination of characters: head distinct from body, two pairs of large and well developed limbs (body serpentiform, head indistinct from body, and limbs rudimentary in Chamaesaura Schneider, 1799), scales on back large and keeled (granular in Platysaurus Smith, 1844, partly granular in Pseudocordylus Smith, 1838 and Hemicordylus Smith, 1838), non-spinose occipitals (spinose in SmaugStanley et al., 2011), 23-24 transverse dorsal scale rows (40-43 in NinurtaStanley et al., 2011;  31-46 in Karusasaurus Stanley et al., 2011; 15-16 in  Ouroborus Stanley et al., 2011); loreal absent (present in Smaug, Nirnuta, Ouroborus, Karusasaurus, and NamazonurusStanley et al., 2011).

Cordylus momboloensis sp. nov.
Loveridge (1944a)a) this species is difficult to infer and threats cannot be fully assessed at this time, but a Data Deficient categorisation, based on IUCN (2022a) Standards and Petitions Committee criteria, may apply.Remarks.It was not possible to locate the specimen of 'Zonurus cordylus' from Aruwimi (presumably Aruwimi River, a tributary of the Congo River) in northern Democratic Republic of the Congo as documented byBoulenger (1897).However, we rejectLoveridge's (1944a)inclusion of this record under C. c. angolensis.If the locality is indeed valid and a population of Cordylus does/did occur in the vicinity of this river, it is likely -as suggested byLoveridge (1944a)-to represent an undescribed species.The Aruwimi River enters the Congo River at Basoko and extends eastward to a position north of Bunia (near Nizi) where it continues, as the Nizi River, to the vicinity of Jiba near the western shores of Lake Albert.The nearest other cordylids are C. tropidosternum (see map in Greenbaun et al. 2012) from northern Tanzania, about 700 km southeast from the nearest part of the Aruwimi/Nizi River, and Katanga Province (D.R.C.) about 900 km to the south.