Dipsadoboa tissue samples obtained through fieldwork and loans from museum collections were processed, and mitochondrial DNA (mtDNA) was analyzed. The samples represented all recognized species and covered the known distributional range of the genus across sub-Saharan Africa. Information on tissue samples and their associated voucher specimens, where available, is provided in Suppl. material 1: table S1. Additionally, sequences of Dipsadoboa were retrieved from GenBank to complement the newly generated dataset. In total, the final dataset comprised new genetic material from 152 Dipsadoboa samples plus 39 sequences from GenBank. The outgroup included multiple sequences and species from 25 colubrid genera (Suppl. material 1: table S2). To leverage available Dipsadoboa data from GenBank and complement previous phylogenetic work on the group, data were generated for three mitochondrial markers (16S ribosomal RNA [16S], cytochrome b [Cyt-b], and NADH dehydrogenase subunit 4 [ND4], including flanking transfer RNAs [tRNAs]; Suppl. material 1: table S2). DNA was extracted using commercial DNA extraction kits following the manufacturers’ protocols. Primer sequences and polymerase chain reaction (PCR) conditions are listed in Suppl. material 1: table S3. Raw sequences were checked and edited using Geneious Prime v. 2024.0.5 (http://www.geneious.com/) and aligned using the MUSCLE plugin for Geneious Prime. All new sequences have been deposited in GenBank (Suppl. material 1: table S2).

Phylogenetic relationships were inferred using both maximum likelihood (ML) and Bayesian inference (BI) approaches for the concatenated dataset of all three loci (16S, Cyt-b, and ND4; 2,227 bp). The ML analysis was also run for each locus separately to produce individual gene trees. The ML analyses were performed in IQ-TREE v. 2.3.4 (Trifinopoulos et al. 2016). For the concatenated dataset, three partitions were used, and 1,000 bootstrap replicates were run with the ultrafast bootstrap approximation method (Hoang et al. 2018). The best-fitting substitution models were selected automatically for each partition. Bootstrap support (BS) values of 95% or higher were considered to provide strong support (Minh et al. 2013). For Bayesian inference, MrBayes v. 3.2.7 (Ronquist et al. 2012) was used with the concatenated dataset divided into three partitions. The BI analysis ran for 10 × 10⁶ generations using Metropolis-coupled Markov chain Monte Carlo (MCMC), sampling every 1,000 generations. Convergence and effective sample sizes (ESS > 200) were assessed in TRACER v. 1.7.2 (Rambaut et al. 2018). A 25% burn-in was discarded, and a 50% majority-rule consensus tree was generated in MrBayes. The substitution model space was set with the parameters lset nst=mixed rates=invgamma, and posterior probabilities (PP) ≥ 0.95 were considered to indicate strong support.

Given the uncertainty surrounding genetic boundaries among putative taxa (species or subspecies) and the limitations of relying exclusively on mitochondrial DNA (Wüster 2025), multiple species delimitation approaches were applied to infer the most reliable species limits. These alternative delimitation schemes were subsequently used in the biogeographic analyses.

Species delimitation was inferred for both 16S [430 bp] and Cyt-b [1,020 bp] using the ASAP and ABGD methods (Puillandre et al. 2012, 2021). Pairwise sequence divergences based on current taxonomy were estimated using SpeciesIdentifier v. 1.8 (Meier et al. 2006) to generate frequency distributions of intra- and interspecific sequence divergence for the Cyt-b gene. Because this analysis cannot accommodate missing data, the leading and trailing ends of the sequences were trimmed to 650 bp, and seven sequences were removed entirely because of their short reads (one each of D. montisilva, D. u. viridiventris, and D. werneri; two each of D. underwoodi and D. weileri). Given that genetic divergences are expected to be low within species but high between species, intraspecific and interspecific divergence values were expected not to overlap (the ‘barcode gap’; Meier et al. 2006).

The divergence threshold separating intra- from interspecific comparisons provides an initial guideline for species delimitation (Lefébure et al. 2006). To complement the SpeciesIdentifier analyses, individuals were assigned to putative species based on Cyt-b ASAP and ABGD results, as well as currently recognized taxa, to evaluate their taxonomic validity (16 taxa in total). All species of Dipsadoboa were included, except D. riparia, for which Cyt-b data are unavailable. Finally, net uncorrected pairwise sequence divergences (p-distances) were estimated separately for 16S [430 bp] and Cyt-b [650 bp] in MEGA v. 10.1.7 (Kumar et al. 2018) to quantify intra- and interspecific variation among strongly divergent mitochondrial lineages. To reduce analytical uncertainty and facilitate comparison between analyses, the same trimmed datasets as in SpeciesIdentifier were used. This enabled direct comparison of interspecific divergence values across species within the genus, including the focal taxa, with the frequency distributions obtained from SpeciesIdentifier.

The genus Dipsadoboa is sister to Crotaphopeltis and belongs to the subfamily Colubrinae (Zaher et al. 2019). More specifically, Dipsadoboa is assigned to the clade referred to as “Colubrinae 2” sensuZheng and Wiens (2016), which includes several African genera. The divergence time from Crotaphopeltis was previously addressed by Engelbrecht et al. (2021), and therefore, the timing of lineage diversification within Dipsadoboa was estimated using the same fossil calibration points and protocol defined by these authors (Suppl. material 1: table S2). Accordingly, a time-calibrated gene tree was estimated in BEAST v. 2.6.3, as specified below. Newly generated sequences for Dipsadoboa were combined with additional sequences obtained from GenBank, resulting in a concatenated dataset partitioned by gene.

Once congruence between the time-calibrated tree topology of Engelbrecht et al. (2021) and the concatenated dataset was corroborated (Suppl. material 1: fig. S1), secondary calibration points were applied to a pruned dataset (selected target species, including all Dipsadoboa species identified by the species delimitation analyses and Crotaphopeltis as the outgroup) to reduce missing data and avoid unrealistic age estimates. Best-fit substitution models were obtained using the bModelTest package implemented in BEAST 2 (Bouckaert and Drummond 2017). An uncorrelated lognormal clock prior was selected for the three genes. For each gene partition, among-site rate heterogeneity was modeled using a gamma distribution with four rate categories. A birth–death speciation process was specified as the tree prior, with random starting trees. Calibration priors were constructed using normal distributions, with mean and sigma values adjusted so that the distributions matched the 95% credible intervals reported in the primary study, as follows: (C1) the divergence between Crotaphopeltis and Dipsadoboa, estimated at 21–29 million years ago (Mya) (mean = 25.0 Mya, SD = 2.5 Mya; 95% CI: 20.9–29.1 Mya), and (C2) the divergence between Crotaphopeltis degeni and C. tornieri 1, estimated at 9–16 Mya (mean = 12 Mya, SD = 1.8 Mya; 95% CI: 9.4–15.0 Mya). Uninformative priors (gamma distribution with parameters α = 0.1 and β = 10) were assigned to the molecular clock parameters, which were estimated during the analysis. Ten independent MCMC runs were conducted, each for 200 million generations and sampled every 10,000 generations. Convergence and adequate sampling were assessed in TRACER v. 1.7.2, and the first 10% of samples were discarded as burn-in. Post-burn-in trees from all runs were combined using LogCombiner v. 2.3.0, and a maximum clade credibility (MCC) chronogram was generated in TreeAnnotator v. 1.8. The resulting dated phylogeny was visualized in FigTree (Rambaut 2009).

Historical range evolution and biogeographic origins of Dipsadoboa were reconstructed using the BioGeoBEARS package in R v. 4.4.0 (R Core Team 2021). The time-calibrated MCC tree was pruned to retain only ingroup taxa. Because some nodes received low support, the topology at these nodes was considered unresolved, and they were collapsed to account for phylogenetic uncertainty. The present distribution of Dipsadoboa was divided into four major regions, reflecting environmental and geological features across sub-Saharan Africa, but extending from East Africa southward to middle Mozambique following the Rift Valley division. Three regions correspond to the major Afrotropical forest subdivisions recognized by Linder et al. (2012) and Couvreur et al. (2008): (i) the Guineo-Congolian forests (the Congolian Region of Linder et al. 2012), encompassing the Lower Guinean forests and the Congo Basin and extending from Nigeria eastward to Uganda and southward to Angola; (ii) the Upper Guinean forests (Guinean Region of Linder et al. 2012), extending from Senegal to Togo; and (iii) the Eastern Afromontane forests, comprising forested areas east of the eastern branch of the African Rift System, from Mozambique to Kenya. An additional region was delineated to represent the moist savannas and woodlands described by Sayre et al. (2013) and treated as a major bioregion (the Zambezi Region) by Linder et al. (2012). This region is referred to here as (iv) the Southern Savannas and includes the Miombo and mixed woodlands located south of the Congo Basin and west of the Malawi Rift. This regionalization allows not only the geographic origin of the genus, as proposed by Rasmussen (1993), to be tested but also the assessment of whether the evolutionary origin of the group is associated primarily with closed-canopy forest habitats or with more open moist savanna and woodland ecosystems. The maximum allowed range size was set to three regions. Ancestral range estimation was conducted under the Dispersal–Extinction–Cladogenesis model (DEC; Ree and Smith 2008), which accounts for vicariance and ‘subset’ sympatric speciation; DIVALIKE (Ronquist 1997), which accounts for vicariance; and BAYAREALIKE (Landis et al. 2013), which accounts for widespread sympatric speciation but not vicariance. All models were implemented with the +J parameter, allowing for founder-event speciation and providing adequate statistical comparison in situations in which most species have geographic ranges restricted to single areas (Matzke 2022).

As the taxonomy of this group remains partially unresolved, the results of the ASAP and ABGD analyses of the Cyt-b dataset were selected as the most conservative and reliable. Thus, the 14 clusters, plus D. riparia (not available in the Cyt-b dataset) and the nominal taxa D. brevirostris and D. guineensis, assigned according to the proximity of the type localities, were used. This resulted in 17 tips for the time-calibrated tree and biogeographic reconstruction.

All phylogenetic trees estimated from individual genes using maximum likelihood (ML) and Bayesian inference (BI) analyses showed topologies similar to those of the concatenated analyses (Fig. 1), although minor differences were detected between the Cyt-b and ND4 mitochondrial gene trees and the 16S phylogenetic reconstruction (Suppl. material 1: figs S1, S2); the latter was poorly supported at most nodes (Suppl. material 1: fig. S2), in agreement with the proposal that the use of 16S for phylogenetic reconstructions in squamates should be viewed with caution and may be misleading (Main et al. 2024). In all analyses, the monophyly of the genus was strongly supported (BS = 100 / PP = 1), and it was recovered as sister to the genus Crotaphopeltis (Fig. 1, Suppl. material 1: fig. S3). All phylogenetic analyses recovered four major well-supported clades and one deeply divergent lineage corresponding to D. kageleri. The ASAP and ABGD analyses recovered 14 clusters in the Cyt-b dataset (excluding D. riparia, which was not available in Cyt-b) and 18 clusters in the 16S dataset (excluding D. kageleri and D. werneri, which were not available in 16S; Fig. 1). The results from the Cyt-b dataset are consistent with the frequency histogram of sequence divergences generated in SpeciesIdentifier, which indicates that values > 8% correspond to interspecific divergence (Fig. 2). The frequency histogram of sequence divergences did not show a distinct “barcode gap,” but instead showed an area of overlap between inter- and intraspecific values in the range of 4–5%. This overlap, however, was attributed only to D. duchesnii and D. brevirostris/D. guineensis, which had sequence divergences that fell within the intraspecific range (Fig. 2). A “gray zone” in the intraspecific region between 5% and 8% was ascribed to the pronounced genetic structure of D. montisilva, D. shrevei, D. viridis, and D. weileri (Fig. 2). Among the interspecific comparisons, the lower range (e.g., 8.5–9.5%) can be attributed to D. underwoodi vs. D. cf. underwoodi and D. unicolor viridiventris vs. D. weileri. Importantly, both approaches (i.e., ASAP and the p-distance frequency distributions) suggest that D. duchesnii, D. brevirostris, and D. guineensis are conspecific, although this hypothesis requires additional validation (Figs 1, 2).

The basal node strongly supports (BS = 100 / PP = 0.99) the monophyly of the genus Dipsadoboa, while the first clade—sister to all other Dipsadoboa and containing the ‘aulica’ group and the deeply divergent D. kageleri—is not strongly supported (BS = 68 / PP = 0.64; Fig. 1). Within the ‘aulica’ group, two mitochondrial lineages were identified (Cyt-b net uncorrected p-distance 12.9%; Suppl. material 1: table S4), corresponding to the nominal D. aulica and D. flavida from the eastern region of South Africa, Mozambique, and Tanzania, respectively, on the southern and northern banks of the Zambezi River (Figs 1, 3). The monophyly of both D. aulica and D. flavida was highly supported (BS = 100 / PP = 1), and the two species were identified as sister species.

The remaining species are divided into three major clades. The ‘werneri’ group (BS = 88 / PP = 0.81; Fig. 1) includes three recognized species, D. werneri, D. montisilva, and D. shrevei, with > 11% net uncorrected p-distance for Cyt-b among them (Suppl. material 1: table S4). Dipsadoboa werneri and D. montisilva from East Africa are more closely related (BS = 100 / PP = 0.99), and together they form a sister clade to D. shrevei (BS = 100 / PP = 0.99). The latter is distributed across the wet miombo and mixed woodlands of southern Africa (Fig. 4).

The phylogenetic reconstruction recovered the ‘duchesnii’ clade as a sister group to the ‘unicolor’ group with moderate support (BS = 87 / PP = 0.81) and included all members of the ‘duchesnii’ group: D. duchesnii, D. brevirostris, and D. guineensis. However, all phylogenetic reconstructions recovered four mitochondrial lineages, including these taxa, with low genetic divergence among them for Cyt-b (net uncorrected p-distance < 5.0%; Suppl. material 1: table S4), within the intraspecific threshold indicated by SpeciesIdentifier (Fig. 2). Furthermore, the analysis recovered D. brevirostris from southeastern Nigeria (near the type locality in southwestern Cameroon) and D. guineensis from Guinea and Liberia as sister taxa, and together they form a clade that is sister to D. duchesnii (Figs 1, 3).

The ‘unicolor’ major group recovered the ‘underwoodi’ clade, morphologically a member of the ‘unicolor’ group, as a well-resolved clade sister to the other members of the ‘unicolor’ group (BS = 100 / PP = 0.94; > 15% Cyt-b net uncorrected p-distance from all other members of the group). This clade comprises two highly divergent mitochondrial lineages (Cyt-b net uncorrected p-distance 8.1%; Suppl. material 1: table S4), including the nominal D. underwoodi, which is documented from the western Congolian and eastern Lower Guinean forests, and a second lineage named herein as D. cf. underwoodi, genetically documented from the Upper Guinean forests (Figs 1, 4). Finally, a large clade including all the remaining members of the ‘unicolor’ group (BS = 99 / PP = 0.91) was recovered with two well-defined subclades (Cyt-b net uncorrected p-distance > 10% between clades; Figs 1, 5; Suppl. material 1: table S4). The first clade includes the nominotypical D. unicolor unicolor from the Upper Guinean forests and D. viridis (BS = 96 / PP = 1) from West and Central Africa, which have a large genetic divergence between them (Cyt-b net uncorrected p-distance 11.4%; Suppl. material 1: table S4). The lineage ascribed to D. u. unicolor includes material from Guinea and Liberia, showing low intraspecific genetic divergence. In contrast, D. viridis exhibits pronounced genetic structuring across its distribution, with five mitochondrial lineages showing some biogeographic structure (Figs 1, 5). One of these lineages is attributed herein to the recognized subspecies D. v. gracilis from the eastern and central Congo Basin (Figs 1, 5).

The third clade within the ‘unicolor’ group represents the least-resolved clade in the study. It includes three recognized taxa (D. weileri, D. u. viridiventris, and D. riparia), as well as one additional mitochondrial lineage previously assigned by Figueroa et al. (2016) and later reanalyzed by Branch et al. (2019) as D. weileri and by Trape et al. (2023) as “D. aff. weileri,” but here referred to as Dipsadoboa sp. “NG/CM” and documented from three sites along the Cameroon Volcanic Line in Nigeria and Cameroon (Figs 1, 5). However, D. riparia formed a long branch embedded within D. u. viridiventris, and this common clade received only moderate support (Fig. 1). Notably, D. riparia is the only species for which only 16S genetic data are available. The other three taxa diverged from each other by > 9% net uncorrected p-distance for Cyt-b (Suppl. material 1: table S4).

Although all described taxa were highly or moderately supported as monophyletic clades, D. unicolor was the only species recovered as paraphyletic. The nominotypical subspecies, D. u. unicolor from the Upper Guinean forests, was strongly supported as sister to D. viridis, including sympatric individuals from Liberia, whereas D. unicolor viridiventris from the Albertine Rift was recovered as closely related to the Upper Guinean D. riparia, with both taxa forming a clade that is sister to the Central African D. weileri (Figs 1, 5). These two sister clades diverged from each other by 12.05% net uncorrected p-distance for Cyt-b (Suppl. material 1: table S4).

The time-calibrated trees (Fig. 6, Suppl. material 1: figs S3, S4) were largely congruent with the topology inferred from the concatenated dataset, supporting the monophyly of Dipsadoboa, the four main groups, and D. kageleri as a distinct lineage on a long branch. However, the monophyly of D. u. viridiventris was not supported with respect to D. riparia, and the two taxa were collapsed with their sister species D. weileri for the biogeographic analysis.

Divergence time estimates indicated that after the split of Dipsadoboa from Crotaphopeltis (~25 Mya), diversification within the genus began in the early Miocene, approximately 22 Mya (95% highest posterior density [HPD]: 19.0–25.5 Mya; Fig. 6, Suppl. material 1: fig. S4). This initial split separated the ‘aulica’ group from the remaining lineages, although the position of D. kageleri, which diverged during the early diversification of the genus, received lower support. Subsequently, the ‘werneri’ group diverged approximately 19.5 Mya (HPD: 15.7–23.9 Mya). This was followed by the separation of the ‘duchesnii’ group from the clade comprising the ‘unicolor’ group at approximately 17.0 Mya (HPD: 13.0–21.2 Mya). Afterward, the ‘unicolor’ group started to diverge approximately 14.5 Mya (HPD: 10.7–18.4 Mya), when the ‘underwoodi’ subclade separated from the other members of the ‘unicolor’ group. This was followed by a radiation during the mid–late Miocene of the remaining members of the ‘unicolor’ group. Finally, the calibrated tree recovered the most recent diversification within the ‘duchesnii’ group, occurring during the Plio-Pleistocene between ~4 and 2 Mya (Fig. 6, Suppl. material 1: fig. S4).

The DEC+J model was selected as the best-fitting biogeographic model, although DIVALIKE+J and BAYAREALIKE+J showed comparable statistical support (Table 1). All biogeographic inferences are based on the DEC+J reconstruction. According to this model, East Africa is the most probable origin of the genus Dipsadoboa (probability: 0.31), although a combined origin in southern and eastern Africa or spanning East to Central Africa cannot be entirely ruled out (probabilities: ~0.22; Fig. 6).

Within the genus, the ‘aulica’ and ‘werneri’ groups are both inferred to have originated in East Africa, with probabilities of 0.76 and 0.53, respectively (Fig. 6, Suppl. material 1: fig. S4). In contrast, the clade that includes the ‘duchesnii’ and ‘unicolor’ groups likely originated in the tropical forests of Central Africa (probability: 0.77; Fig. 6, Suppl. material 1: fig. S4). Biogeographic reconstructions suggest that each of these groups subsequently originated within this region and diversified along the Upper Guinean and Lower Guinean-Congolian forests. The ‘duchesnii’ group seems to have originated in Central Africa (probability: 0.85; Fig. 6). A similar scenario was recovered for the ‘unicolor’ group, which most likely originated within the wider Congo Basin, with a probability of ~0.70 (Fig. 6, Suppl. material 1: fig. S4). These results suggest at least three independent colonization events of the Upper Guinean forests from the Lower Guinean-Congolian forests, whereas other events driven by climatic refugia occurred within the wider Congo Basin (Fig. 6B).

The results recover Dipsadoboa as a monophyletic genus, sister to Crotaphopeltis, a relationship that has been proposed in most previous phylogenetic analyses (e.g., Zheng and Wiens 2016; Branch et al. 2019; Engelbrecht et al. 2021; Conradie et al. 2026), but which contrasts with the results reported by Zaher et al. (2019) and Trape et al. (2023), who recovered the genus as paraphyletic. These findings are consistent with the morphological distinctions between the two genera previously highlighted by Laurent (1951) and Rasmussen (1993). Furthermore, the phylogenetic analysis partially agrees with the morphological cladogram proposed by Rasmussen (1993). Notably, five major mitochondrial clades were recovered, four corresponding to the morphological clusters (Rasmussen 1993) of the ‘aulica’, ‘werneri’, ‘duchesnii’, and ‘unicolor’ groups, and one corresponding to the distinct D. kageleri lineage. In contrast to Rasmussen (1993), the phylogenetic placement of D. kageleri differs from both its morphological assignment to the ‘werneri’ group and its placement in the ‘werneri’ group in a previous phylogenetic analysis, which had very low support (Branch et al. 2019). Instead, this species clusters within the ‘aulica’ group. Nevertheless, the node that defines its placement shows low support and should therefore be considered unresolved. Several taxonomic inconsistencies were also detected in previously published datasets, including cases of apparent misidentification (e.g., “D. weileri” or “D. aff. weileri” MVZ 253208 in Figueroa et al. 2016; Branch et al. 2019; Trape et al. 2023; Arroyo-Lambaer et al. 2025, which is reassigned here as Dipsadoboa sp. “NG”; or “D. duchesnii” or “D. cf. duchesnii” CAS 254086 in Portik et al. (2016) and Trape et al. (2023), respectively, which is reidentified here as D. underwoodi). Based on these results, some taxonomic actions that require further investigation are proposed below (Table 2).

The results support Rasmussen’s (1993) suggestion that the Lower Guinean forests (Area II in Rasmussen 1993) within the Congo Basin are the main diversity hotspot for Dipsadoboa, while also highlighting the Eastern Afromontane region (Area VII in Rasmussen 1993) as an important region for its diversification. The ancestral range reconstruction indicates the Eastern Afromontane region as the most probable region of origin of the genus (Fig. 6B), thus suggesting a forest-dwelling origin of the group. This conclusion is reinforced by the sister genus Crotaphopeltis, for which East and Southern Africa have been proposed as the regions of origin (Engelbrecht et al. 2020). Other studies have suggested ancestral origins of African forest clades in the Eastern Afromontane region (Fjeldså and Lovett 1997; Fjeldså and Bowie 2008). The DEC+J model emerged as the best-fitting scenario, revealing low dispersal (d = 0.002), negligible extinction, and a significant role of founder-event speciation (j = 0.18), suggesting that rare long-distance dispersal followed by lineage isolation (vicariance) shaped the current distribution of Dipsadoboa across African forests and savanna woodlands (Ree and Smith 2008).

Diversification of Dipsadoboa seemed to occur independently between the Eastern Afromontane forests and southern savannas and within Central Africa (Fig. 6). Furthermore, the model recovered two independent divergence events occurring between Eastern Afromontane pockets and southern savannas in the early and mid-Miocene. This pattern is consistent with the tectonic evolution of the East African Rift, which began developing during the mid–late Miocene, with further rift propagation continuing into the Plio–Pleistocene (McConnell 1972; Macgregor 2015; Michon et al. 2022). It is also consistent with Miocene climatic fluctuations and the associated shifts in the distribution of forest and savanna habitats across Africa, which likely promoted lineage isolation and diversification, particularly through the persistence of climatic refugia on East African inselbergs (e.g., Loader et al. 2014; Couvreur et al. 2021).

Climatic fluctuations seem to have also played an important role in shaping diversification within the Congo Basin, where at least four diversification events are inferred to have occurred during the mid- and late Miocene (18–5 Mya). These results are consistent with studies of other forest-dwelling vertebrate groups (Portillo et al. 2015, 2018; Nicolas et al. 2020; Onditi et al. 2021; Lobón-Rovira et al. 2024), suggesting that climatic refugia and savanna expansion drove multiple vicariant events across the Congo Basin, with subsequent secondary contacts during forest expansions.

The ancestral range reconstruction for the Upper Guinea region suggests three independent colonization events followed by isolation from Central African lineages. However, a broader ancestral distribution across West and Central Africa cannot be excluded, with subsequent isolation driven by climatic refugia in the region, which is well known for its pronounced climatic fluctuations (Ernst et al. 2025).

Overall, although the ancestral range reconstruction supports founder-event speciation (dispersal followed by isolation) as the most plausible scenario, the strong geographic structuring observed in several lineages indicates that vicariant processes were also important, particularly because of climatic refugia in the Congo Basin and the origin and expansion of the Rift Valley between East and Southern Africa. Together, these results suggest that both dispersal events and historical isolation have shaped the diversification and geographic patterns of Dipsadoboa in Africa. However, the biogeographic results should be interpreted with the caveat that the phylogenetic reconstructions are based on mitochondrial DNA data; therefore, the topologies and time estimates would greatly benefit from the inclusion of nuclear markers in future studies for more robust interpretations.