Museum specimens and comparisons

We examined all Bassaricyon specimens in the collections of the American Museum of Natural History, New York, USA (AMNH); Academy of Natural Sciences, Philadelphia, USA (ANSP); Natural History Museum, London, UK (BMNH); Museo de Zoología, Universidad Politecnica, Quito, Ecuador (EPN); Field Museum of Natural History, Chicago, USA (FMNH); Biodiversity Institute, University of Kansas, Lawrence, USA (KU); Los Angeles County Natural History Museum, Los Angeles, USA (LACM); Museum of Comparative Zoology, Harvard University, Cambridge, USA (MCZ); Museo Ecuatoriano de Ciencias Naturales, Quito, Ecuador (MECN); Museum of Vertebrate Zoology, University of California, Berkeley, USA (MVZ); Naturhistoriska Riksmuseet, Stockholm, Sweden (NMS); Museo de Zoología, Pontificia Universidad Católica del Ecuador, Quito, Ecuador (QCAZ); Royal Ontario Museum, Toronto, Canada (ROM); Biodiversity Research and Teaching Collections, Texas A&M University, College Station, USA (TCWC); Museum of Zoology, University of Michigan, Ann Arbor, USA (UMMZ); National Museum of Natural History, Smithsonian Institution, Washington, D.C., USA (USNM); Peabody Museum of Natural History, Yale University, New Haven, USA (YPM); and Museum für Naturkunde, Humboldt Universität, Berlin, Germany (ZMB). These holdings include all type specimens in the genus and represent the great majority (well over 95%) of olingo specimens in world museums. We also had access to published information on a few additional specimens in museum collections in Colombia and Bolivia (Saavedra-Rodríguez and Velandia-Perilla 2011, Anderson 1997). Tissue samples are stored in the frozen tissue collections of the MVZ, ROM, USNM (including specimens to be accessioned at QCAZ), the New York State Museum, Albany, New York, USA (NYSM), and the Museum of Texas Tech University, Lubbock, Texas, USA (TTU) (Table 1).

Values from external measurements of 95 specimens are presented to provide an appreciation of general body size and lengths and proportions of appendages. Values (in mm) for total length and length of tail are those recorded by collectors on labels attached to skins; subtracting length of tail (abbreviated TV) from total length produced a value for length of head and body (HB). Values for length of hind foot (HF), which includes claws, were either obtained from skin labels or from our measurements of dry study skins; those for length of external ear (E), or pinna, come from collector’s measurements recorded on skin labels or in field journals (we assume, but are not certain for all specimens, that ear-length measurements represent the greatest length from the notch to the distal margin of the pinna).

Morphological terminology follows Evans (1993) and Ahrens (2012). Craniodental variables were measured by the first author with digital calipers to the nearest 0.1 mm. Single-tooth measurements are measured on the crown. All measurements of length are in millimeters, and measurements of mass are given in grams. Only fully adult, wild-collected specimens that are sufficiently intact were included in our morphometric analyses. A total of 115 specimens were included (51 male, 64 female). The classification of ‘‘adult’’ was applied generally only to skulls in which the full dentition is completely erupted, and in which the basilar (basioccipital-basisphenoid) suture (synchondrosis) in particular is obliterated via ossification. Variables measured include maximum crown widths (W) of premolars (p1, p2, p3, p4, P2, P3, P4, with lower case designating lower teeth and uppercase designating upper teeth) and molars (m1, m2, M1, and M2); maximum crown lengths (L) of the larger premolars and molars (P4, M1, M2, m1, and m2); condylobasal length (CBL), zygomatic width (ZYG), breadth of braincase (BBC), external width across the canines (CC), and length of the maxillary toothrow, C-M2 (MTR), all as defined by Kennedy and Lindsay (1984); and four posterior skull measurements: greatest width across the postdental palatal shelf (WPP), length of the postdental palate behind an imaginary line delineated by the back of the second molars (LPP), anteroposterior length of the auditory bullae including the eustachian tube (LAB), and the dorsoventral diameter inside the external auditory meatus (EAM). Unless explicitly noted, all reported metrics (and resulting statistical and multivariate comparisons) refer only to fully mature (adult) specimens, as judged by direct examination of skulls. Because some olingo taxa demonstrate significant sexual dimorphism in cranial measurements, patterns of morphometric variation in males and females were compiled and analyzed separately. Principal Component Analysis (PCA) and Discriminant Function Analyses (DFA) were undertaken using a combination of cranial and dental measurements indicated in tables and in the text, selected to sample craniodental size and shape, and to maximize sample size. All measurement values were transformed to natural logarithms prior to multivariate analysis. Principal components were extracted from the covariance matrix. The software program Statistica 8.0 (Statsoft Inc., Tulsa, Oklahoma, USA) was used for all multivariate analyses.

The taxa and sequences included in our analysis are listed in Table 1. Our choice of taxa outside of Bassaricyon was guided by the findings of Koepfli et al. (2007), Fulton and Strobeck (2007), and Eizirik et al. (2010). These studies provide strong statistical support for relationships and divergence dates within Procyonidae and Carnivora based on >6, 000 bases of DNA and fossil evidence. We chose one mitochondrial marker and one nuclear marker used in these and many other mammalian studies, in order to capture the evolutionary histories of these distinct genetic systems. Although deeper relationships within the order Carnivora cannot be resolved solely by using these two genes, we are confident that they provide the appropriate level of support to resolve species-level relationships within this group of procyonids (Koepfli et al. 2007). As our specific goal was to estimate the timing of divergence within Bassaricyon, and our reduced dataset did not provide enough support to resolve deeper nodes in Caniformia, we decided to use highly supported divergence date estimates from Koepfli et al. (2007) and Eizirik et al. (2010) as priors in our analysis.

DNA extraction

Tissues from fresh and frozen specimens were processed using a Qiagen DNeasy kit (QIAGEN, Valencia, CA, USA) to obtain genomic DNA. The sample from the skull of KU 165554, a museum specimen of Bassaricyon gabbii, was taken from the turbinate bones and extracted following the method of Wisely et al. (2004). Including this turbinate sample of Bassaricyon gabbii, we successfully extracted DNA from eight individuals of Bassaricyon (four Bassaricyon medius, one Bassaricyon alleni, and two Bassaricyon neblina sp. n.). All pre-PCR protocols were conducted in an isolated ancient DNA laboratory facility located in a separate building from the one containing the primary DNA laboratory.

DNA Sequencing

Mitochondrial gene, cytochrome b: For cytochrome b (1140 bp), polymerase chain reaction (PCR) and sequencing reactions were carried out with primers LGL 765 and LGL 766 from Bickham et al. (2004) and using a thermal cycler (MJ Research, Waltham, MA, USA) under the following conditions, repeated for 35 cycles: denaturation at 92°C for 1 min, annealing at 50°C for 1 min, extension at 72°C for 1 min. The PCR reagents in a 25 μL reaction were 0.2 μL AmpliTaq (5 units μL-1, Applied Biosystems, Foster City, CA, USA), 1μL per primer (10 μM), 2.5 μL dNTP (2 μM), 2 μL MgCl2 (25 mM), 2.5 μL AmpliTaq Buffer (Applied Biosystems), 2μL BSA (0.01 mg/μL), 1 μL gDNA and 12.8 μL sterile water. To amplify DNA from turbinate samples, PCR and sequencing were carried out with internal primers designed for this study using sequences generated from the tissue samples; the reverse primer, H151949Pro (5’ CTCCTCAAAAGGATATTTGYCCTCA 3’), located at 14611 – 14636 on the Nasua nasua mitochondrial genome, was used with LGL 765. A new forward primer, BAS420F (5’ TCAGACAAAATCCCCTTCCA 3’), position 14825 - 14845 on the Nasua nasua mitochondrial genome, was used with LGL 766. The thermal cycle regime was modified to 50 cycles; reagents were as above.

Nuclear intron, Cholinergic Receptor Nicotinic Alpha Polypeptide 1 precursor (CHRNA1): For CHRNA1 (347 bp), we used the primers described by Lyons et al. (1997) and the thermocycling conditions consisted of an initial denaturation (95°C for 10 min), followed by 30 cycles of denaturation at 95°C for 30 s, annealing at 54°C for 30 s and extension for 72°C for 45 s, and final extension of 72°C for 5 min. Reagent volumes were the same as for cytochrome b amplification (above), except 2μL of gDNA was added for CHRNA1 amplification, decreasing sterile water volume to 10.8μL. We were unable to sequence the nuclear intron from the turbinate bone sample.

Each PCR was conducted with negative and positive controls to minimize risk of spurious results from contamination or failure of the reaction. A 2μL sample of the PCR product was stained with ethidium bromide and run on an agarose gel with a 1 kb ladder. The gel was placed under UV light to visualize the PCR products. Polymerase chain reaction products were amplified for sequencing using a 10 μL reaction mixture of 2 μL of PCR product, 0.8 μL of primer (10 μM), 1.5 μL Big Dye 5 x Buffer (Applied Biosystems), 1 μL Big Dye version 3 (Applied Biosystems), and 4.7 μL sterile water. The reaction was run using a thermal cycler (MJ Research) with denaturation at 96°C for 10 s, annealing at 50°C for 10 s and extension at 60°C for 4 min: this was repeated for 25 cycles. The product was cleaned using sephadex filtration method and sequences for both strands were run on a 50 cm array using the ABI PRISM 3100 Genetic Analyzer (Applied Biosystems).

Molecular analysis

Sequences were aligned and edited in Sequencher version 4.1.2 using the implemented Clustal algorithm and the default gap penalty parameters (Gene Codes Corporation, Ann Arbor, MI, USA, http://www.genecodes.com).

For Bassaricyon, we included all newly sequenced and previously available sequences for cytochrome b and CHRNA1 (for cytochrome b this included five individuals of Bassaricyon medius, two Bassaricyon alleni, one Bassaricyon gabbii and three Bassaricyon neblina sp. n.; for CHRNA1 this included five individuals of Bassaricyon medius, two Bassaricyon alleni, and three Bassaricyon neblina sp. n.) (Table 1).

Maximum Parsimony, Maximum Likelihood and Bayesian analyses were performed for each gene and a concatenation of the two genes to check for any incongruence in structure and support of the Bassaricyon clade. All Bayesian and Maximum Likelihood phylogenetic inferences were carried out using the Cipres Portal (Miller et al. 2009). Indels were treated as missing data or non-informative data for all of the analyses as in previous molecular phylogenetic studies of procyonids (Koepfli et al. 2007; Eizirik et al. 2010).

Pairwise distances for cytochrome b were generated using the Kimura 2-parameter model using MEGA4 (Tamura et al. 2007).

The branch and bound search method implemented in the software package TNT (Goloboff et al. 2008) was used for the maximum parsimony analyses. Parsimony bootstrap support was generated using the heuristic search method with 100 random stepwise additions for 1000 replicates.

Maximum Likelihood analysis was conducted using the software package GARLI 0.96b (Zwickl 2006). The genetic-like algorithm was used to simultaneously find the topology, branch lengths and substitution model parameters with the greatest log-likelihood (lnL) for each dataset. Bootstrap support was generated with 1000 replicates and two independent searches per replicate.

jModeltest version 0.1.1 (Posada 2008) was used to find the best model of sequence evolution. We chose to partition the cytochrome b data in order to minimize the number of parameters and to account for differences in base composition and substitution rates among the different codon positions (Corse et al. 2013). The software PartitionFinder (Lanfear et al. 2012) was used to determine the best partitioning scheme, and for the cytochrome b, the scheme with 1^st, 2^nd, and 3^rd codon positions partitioned separately was selected. The best fit model under the Bayesian information criterion (BIC) for cytochrome b for the first and second codon position partitions was HKY + G + I (Hasegawa et al. 1985), and for the third codon position, the best model under BIC was TrN + I + G (Tamura and Nei 1993). The model chosen for CHRNA1 was K80 + G (Kimura 1980). The parameters were then applied in MrBayes version 3.1p (Huelsenbeck and Ronquist 2001). The model parameters were set to nst = 2 using a gamma distribution for CHRNA1, nst = 2 and the rate parameter invariant with a gamma distribution for the cytochrome b 1st and 2nd codon partitions, and nst = 6 with a gamma distribution and rate parameter invariant for cytochrome b 3rd codon partition. Since this version of MrBayes did not include the specific model selected for the cytochrome b 3rd codon position partition, we opted for using a more complex model (nst = 6) following the results of Huelsenbeck and Rannala (2004). The Bayesian analysis was run using 5, 000, 000 generations along four chains with 2 replicates at a temperature of 0.05. The convergence between the two replicates run in MrBayes was assessed by the average standard deviation of split frequencies (ASDSF) between runs. After 5, 000, 000 generations, the ASDSF was 0.003. Sample frequency was set to 1000 with a burn-in of 1, 250.

Molecular divergence estimates were generated in BEAST (Drummond and Rambaut 2007). The following calibration nodes were included based on Eizirik et al. (2010): Nasua – Bassaricyon truncated normal mean 7.2 million years ago (mya) (± 1.7 s.d.); -Potos truncated normal mean 16.2 mya (± 2.5 s.d.); Procyonidae normal mean 20.7 mya (± 4.0 s.d.); Procyonidae-Mustelidae-Ailuridae-Mephitidae normal mean 30 mya (± 7.0 s.d.); Phoca-Mirounga normal mean 20 mya (± 6 s.d.); Caniformia normal mean 48 mya (± 6.5 s.d.). The molecular clock was estimated using the uncorrelated lognormal setting, operators were left to the default setting, and trees were searched using the Yule process. The substitution and clock models were left unlinked, partition tree model was linked, and the models for the two gene partitions were: cytochrome b (1), (2), and (3) => TN93 + I + G (all parameters unlinked) and CHRNA1 K80 + G (HKY + G). In order to evaluate the effects of the priors on the divergence time estimates, we carried out a run using an empty alignment but with the same settings and compared it to our results, with the outcome indicating that the priors are not having an especially strong effect on the estimated divergence times (Drummond et al. 2006).

To infer geographical range evolution of procyonids we used the Maximum Likelihood model of dispersal-extinction-cladogenesis (DEC) implemented in Lagrange v. 20130526 (Ree and Smith 2008). The BEAST chronogram tree was trimmed to keep one representative per procyonid species, and two additional lineages, one representing Mustelidae and one representing Mephitidae. Six general geographic areas were used to characterize the distribution ranges: Eurasia, North America, Central America, Chocó, Andes, and Amazonia. The branches of the mustelid and mephitid lineages were treated as belonging to the ancestors of the families and their hypothesized distributions are according to previous ancestral range estimations (Koepfli et al. 2008, Sato et al. 2009). Reconstruction of potential ancestral area combinations and dispersal scenarios took into account realistic dispersal routes (e.g., allowing Eurasia to connect only with North America) and the geological history of the region (e.g., formation of the Panama Isthmus during the late Miocene and Pliocene; Weyl 1980, Almendra and Rogers 2012).

Bioclimatic range modeling

Vouchered localities of occurrence for Bassaricyon used in our analyses were extracted from museum specimen labels, often as clarified by associated field notes and journals, and from definitive published accounts. Gazetteers published by Paynter (1982, 1993, 1997), Stephens and Taylor (1983, 1985), Fairchild and Handley (1966), Handley (1976), Voss (1988), and Voss et al. (2002) were especially helpful in georeferencing Neotropical expedition and collecting localities represented in museum collections.

We used Maximum Entropy Modeling (Maxent) (Phillips et al. 2006) to predict the geographic range of the geographic range of the four Bassaricyon species at broad scales based on vouchered localities (Appendix 2) and 20 environmental variables representing potential vegetation and climate. For potential vegetation we used the 15 major habitat types classified as ecological biomes (Olson et al. 2001). For climate we used 19 BIOCLIM variables representing annual trends, seasonality, and extremes in temperature and precipitation across Central and South America (derived from Hijmans et al. [2005] as described at http://www.worldclim.org/bioclim.htm). We used all vouchered specimen localities to train the final model (excluding published records based only on visual observations). We also tested overall performance by running 10 model iterations while randomly withholding 20% of the points as test locations. To produce geographic ranges showing presence/absence of a species we used the average equal training sensitivity and specificity for the 10 test models as our probability cutoff value (Phillips et al. 2006).

Phylogenetics

With the largest molecular sampling effort to date, we show that Bassaricyon is well resolved as a monophyletic genus (cf. Nyakatura and Bininda-Emonds 2012) within the family Procyonidae. All of our analyses resolve Bassaricyon as a clade with bootstrap and probability values of 100%. The sister genus to Bassaricyon is Nasua, a relationship consistently recovered in our analyses with 100% support. The divergence between Bassaricyon and Nasua was estimated at 10.2 million years old (mya) (95% Confidence Interval [CI] = 7.6 – 12.7 mya), consistent with previously published results (Koepfli et al. 2007, Eizirik et al. 2010).

The family Procyonidae is well resolved as monophyletic (100% bootstrap and probability values) with a divergence date of 21.4 mya (CI 18.1 – 25.0 mya), in agreement with the divergence estimate of 22.6 mya (CI 19.4 – 25.5 mya) by Eizirik et al. (2010). Eizirik et al. (2010) had a more constrained confidence interval on the age of this divergence, due to the incorporation of genes that are more informative at deeper nodes in the tree. We chose CHRNA1 and cytochrome b with a focus toward resolving relationships within Bassaricyon; these markers are far more useful for determining relationships in recent radiations within Procyonidae than the deeper relationships within Carnivora. The only part of the Procyonidae where CHRNA1 and cytochrome b did not provide sufficient resolution to re-construct recently published multi-gene topologies (Koepfli et al. 2007, Eizirik et al. 2010) was the divergence between the two species of Bassariscus, and Procyon. In our BEAST chronogram the divergence for Bassariscus is 7.6 mya (CI 4.8 – 10.6 mya) but the branch leading to their divergence has no support, and therefore is collapsed in the phylogeny (Figure 1; see also Koepfli et al. 2007, Eizirik et al. 2010). The other procyonid genera are well-supported monophyletic groups; according to our chronogram Procyon lotor and Procyon cancrivorus diverged 4.2 mya (CI 2.3 – 6.5 mya) and Nasua narica and Nasua nasua diverged 5.6 mya (CI 3.5 – 7.9 mya).

The concordance of our recovered topology and estimates of genetic divergence with previous phylogenetic studies of the Procyonidae suggests that data from cytochrome b and CHRNA1 across sampled taxa have provided a well-supported framework in which the species relationships and divergence dates within Bassaricyon can be reliably assessed. Previous molecular phylogenetic studies have included either only one species (e.g., Ledje and Arnason 1996a, 1996b, and further studies using the same sequences, see below), identified as “Bassaricyon gabbii” (Genbank identifier X94931), but actually representing Bassaricyon neblina sp. n.; or, two species (Koepfli et al. 2007), identified as Bassaricyon alleni (correctly, sample from Amazonian Peru) and “Bassaricyon gabbii” (actually Bassaricyon medius orinomus, from Panama). Koepfli et al. (2007) gave the divergence estimate for these latter two taxa (i.e. Bassaricyon alleni and Bassaricyon medius orinomus) as 2.5–2.8 mya (CI 1.2–5.0 mya). Our results indicate that the earliest divergence within Bassaricyon, corresponding to the split between the ancestors of Bassaricyon neblina sp. n. and other Bassaricyon, occurred 3.5 mya (CI = 2.1 – 5.2 mya). Sequence divergence in cytochrome b between Bassaricyon neblina sp. n. and other Bassaricyon (including specimens of Bassaricyon medius medius collected in regional sympatry with Bassaricyon neblina sp. n. in the Western Andes of Ecuador) is 9.6-11.3% (Table 2). Cytochrome b sequence divergences between Bassaricyon gabbii, Bassaricyon medius, and Bassaricyon alleni are 6-7% (Table 2). Subspecific distances (see Systematics, below, for discussion of subspecies boundaries) are 1.6-2.0% within Bassaricyon medius (between Bassaricyon medius medius and Bassaricyon medius orinomus) and 1.6% within Bassaricyon neblina sp. n. (between Bassaricyon neblina neblina subsp. n. and Bassaricyon neblina osborni subsp. n., the two subspecies for which we have molecular data).

We obtained the highest bootstrap and posterior probability support values (100% and 1.0 respectively) for relationships within Bassaricyon with every method of phylogenetic inference that was used in this study. The single exception was that the topology that recovered the node uniting Bassaricyon alleni and Bassaricyon medius to the exclusion of Bassaricyon gabbii was assigned a slightly lower Bayesian posterior probability value of 0.97, but all other methods lent full support to this topology (Bassaricyon gabbii, (Bassaricyon medius, Bassaricyon alleni)). These results were also well-supported by our comparisons of morphological characters and together lend strong support for this scenario as being an accurate representation of the evolutionary history of diversification within Bassaricyon.

Biogeography

The historical biogeographic reconstruction for the Procyonidae using the DEC model sets Central America as the likely center of origin of crown-group procyonids (Figure 2) (though we note that the family has many extinct, Eocene to Miocene representatives in North America and Europe). Major splits within the family appear to have occurred in Central America previous to the formation of the Panamanian Isthmus, and all the dispersal events resulting in the extant species have occurred within the last 10 million years. All those dispersal events involving southward movements seem to have occurred up to circa 6 mya, coinciding with the initial uplift of the Panamanian Isthmus, and, presumably once it was consolidated, with the Great American Biotic Interchange (GABI) (Figures 1–2). The clade containing all olingo species is likely to have originated directly as a result of the formation of the Panamanian Isthmus, and provides evidence of a complex pattern of dispersal events out of Central America (Figure 2).

Morphology and morphometrics

Our study of Bassaricyon taxonomy originally began with close examination of craniodental traits of museum specimens, which quickly revealed to the first author the existence of Bassaricyon neblina sp. n., which is highly distinctive morphologically. Close examinations of skins and skulls revealed clear differences in qualitative traits, and in external and craniodental measurements and proportions, between the four principal Bassaricyon lineages identified in this paper (which we recognize taxonomically as Bassaricyon neblina sp. n., Bassaricyon gabbii, Bassaricyon alleni, and Bassaricyon medius; Figures 3–5). Externally, these especially include differences in body size, pelage coloration, pelage length, relative length of the tail, and relative size of the ears (Figure 3, Table 5). Craniodentally, these especially include differences in skull size, relative size of the premolars and molars, configuration of molar cusps, relative size of the auditory bullae and external auditory meati, and the shape of the postdental palatal shelf (Figures 4–5, Tables 3–4). These and other differences are discussed in greater detail in the species accounts provided later in the paper.

Principal component analyses of cranial and dental measurements support our molecular results in clearly identifying a fundamental morphometric separation between the Olinguito (Bassaricyon neblina sp. n.) and all other Bassaricyon taxa, in separate comparisons involving both males and females (Figures 6–7, Appendix 1). When first and second principal components are juxtaposed in a bivariate plot, Olinguito specimens demonstrate clear morphometric separation from all other Bassaricyon, despite overlap between these clusters in body size (as indicated by overlap in the first principal component, on which all or most variables in the analysis show positive [males] or negative [females] loadings). Despite smaller average body size compared to other Bassaricyon, the morphometric distinctness of Olinguito specimens is reflected not especially in small size but rather primarily by separation along the second principal component, indicating trenchant differences in overall shape and proportion, especially reflecting consistent differences in the molars, auditory bullae, external auditory meati, and palate, in which the Olinguito differs strongly and consistently from other Bassaricyon (Figures 6–7, Tables 3–4, Appendix 1).

The lower elevation olingo taxa Bassaricyon gabbii, Bassaricyon medius, and Bassaricyon alleni are not separable in most principal component analyses of craniodental measurements (e.g., Figures 6–7), but discriminant function analyses of craniodental measurements (e.g., Figure 8, showing separation of male skulls) separates them into discrete clusterings with few misclassifications, and identifies some of the more important craniodental traits that help to distinguish between them (Appendix 1). These (and other, qualitative) craniodental distinctions are complemented by differences in pelage features and genetic divergences that we discuss below.

Because of marked and consistent differences in body size between the two regional populations of Bassaricyon medius (one distributed in western South America, the other primarily distributed in Panama), we choose to recognize these two as separate subspecies (Bassaricyon medius medius and Bassaricyon medius orinomus, respectively, Tables 6–7). The Olinguito likewise shows a clear pattern of geographic variation, with different regional populations in the Northern Andes showing consistent differences in craniodental size and morphology (Figures 9–10, Table 8, Appendix 1), as well as pelage coloration and length. We recognize four distinctive subspecies of the Olinguito throughout its recorded distribution, as discussed in the description of Bassaricyon neblina sp. n., below. Two of these subspecies are included in our genetic comparisons; genetic comparisons involving the remaining two subspecies remain a goal for the future.

Bioclimatic range modeling

Distribution models for all species are judged to have performed well based on their high values for ‘area under the curve’ (AUC) and unregularized test gain (Table 9), as well as their fit of the final prediction to the locality data (Figures 11–12). There was relatively low impact of withholding test data from these models, as indicated by the low Mean Test AUC values. These values are lowest for Bassaricyon alleni, probably reflecting its larger distribution relative to the variation of environmental data (Phillips et al. 2006). The strongest environmental predictors for Bassaricyon neblina sp. n. were seasonal variation in temperature (suitability declines with higher variation, after sharp threshold) and the temperature of the wettest quarter (negative relationship). The annual range of temperatures was the most important predictor for the Bassaricyon gabbii and Bassaricyon medius distributions (both sharp negative relationships). Bassaricyon alleni was the only one of the four species to have an ecological biome ranked as one of the top predictors (Tropical Moist Broadleaf Forests as highly suitable).

The full Maxent distribution models predict the suitability of habitat across South and Central America (Figure 11). To make the binary prediction maps (Figure 12) we excluded areas with high probability that were disjunct from areas where specimens have been recorded (e.g., western Venezuela excluded from the map for Bassaricyon neblina sp. n., central and eastern Brazil excluded from the Bassaricyon alleni map, northern Central America excluded from the Bassaricyon medius map, South America excluded from the Bassaricyon gabbii map). For Bassaricyon neblina sp. n. we excluded areas of high probability from the Eastern Cordillera of Colombia and the Andes of southern Ecuador and northern Peru because of the lack of specimens. Likewise, predicted suitable habitat for Bassaricyon gabbii in northern Central America (Honduras, Guatemala) remains unverified by specimen data. Although there are two recent unconfirmed records in the region (Ordóñez Garza et al. 1999–2000), the specific locations of these sightings did not fall in areas predicted as suitable habitat by our models. Finally, the exact area of transition between Bassaricyon gabbii and Bassaricyon medius in Panama remains unclear. All of these regions should be considered high priority areas for future surveys, especially areas identified as potential Bassaricyon neblina sp. n. habitat (see Discussion, below).

The range of Bassaricyon neblina sp. n. is typical of many Andean species in being restricted to wet cloud forest habitats, which are limited in area and also under heavy development pressure. In comparing recent land use (Eva et al. 2004) of suitable historical Bassaricyon neblina habitat, we found that 42% of suitable habitats have been converted to agriculture or urban areas, and 21% remain in natural but largely unforested conditions. Thus we predict that only 37% (40, 760 km²) of appropriate Olinguito habitats remain forested.

Subspecies of Bassaricyon neblina

Reproductive isolation and genetic divergence of Bassaricyon neblina

Information from sympatric occurrences and captive breeding demonstrates that the Olinguito, Bassaricyon neblina, is reproductively isolated from other species of Bassaricyon and clearly constitutes a distinct “biological species” (i.e., sensu Mayr 1940, 1942).

In Ecuador we documented the Olinguito (Bassaricyon neblina neblina) in regional sympatry with the Western lowland olingo, Bassaricyon medius medius; we recorded the two species at localities less than 5 km apart (i.e., at Otonga and San Francisco de las Pampas) during fieldwork in August 2006. The ecogeographic relationship between the two species is probably one of elevational parapatry or limited elevational overlap along the western slopes of the Andes. Bassaricyon medius medius extends into the elevational range of Bassaricyon neblina, perhaps especially in areas where cloud forests have been cleared for human settlement, agriculture, and pastoralism (Figure 17).

Ingeborg Poglayen-Neuwall (pers comm. to R. Kays, 2006) informed us that an adult female zoo animal named “Ringerl” (Figure 15; also figured by Poglayen-Neuwall 1976), which we can now identify as an Olinguito (Bassaricyon neblina osborni), was moved among several zoos during the 1970s because it would not successfully breed with other captive olingos (i.e., not Bassaricyon neblina), most of which were apparently Bassaricyon alleni (see Poglayen-Neuwall 1976).

The Olinguito differs from congeners (Bassaricyon alleni, Bassaricyon medius, and Bassaricyon gabbii) by 9.6–11.3% in base-pair composition of the (mitochondrial) cytochrome b gene (Table 2), a level of divergence consistent with that separating biological species in many groups of mammals, including carnivores (Baker and Bradley 2006). For comparison with other procyonids, this level of genetic distinction is equivalent to the 10–11% divergence between Procyon lotor and Procyon cancrivorus, sympatrically-occurring raccoons traditionally classified in separate subgenera (Goldman 1950, Helgen and Wilson 2005), and comparable to the 9–13% divergence between Nasua narica and Nasuella olivacea (Helgen et al. 2009), coatis traditionally classified in separate genera (Hollister 1915, Decker and Wozencraft 1991, Wozencraft 1993, 2005).

Karyotype: The karyotype of an adult female Olinguito(Bassaricyon neblina osborni, then identified as “Bassaricyon gabbii”, with 2n = 38, as in all procyonids) was reported and discussed (but not described in detail) by Wurster-Hill and Gray (1975), and figured by Nash (2006). This was based on a captive animal originally captured from mountains in the vicinity of Cali in Colombia (Figure 15).

Description: The Olinguito is the smallest species of Bassaricyon, both in skull and body size (Tables 3, 5), and is thus, on average, the smallest living procyonid (matched only by small individuals of the Ringtail, Bassariscus astutus). The tail averages 10% longer than the head-body length (Table 5). The pinnae are proportionally much smaller in Bassaricyon neblina than in other Bassaricyon, appearing shorter and rounder, and standing out less conspicuously on the head; they are also more heavily furred and usually fringed with a paler, contrasting border of buffy or golden fur. The dorsal fur is dense, long, and luxurious, with the longer hairs measuring 30–40 mm in length (usually much shorter in other Bassaricyon, at least in the predominantly lowland taxa Bassaricyon medius and Bassaricyon alleni, but reaching 25 mm in the highest-elevation populations of Bassaricyon alleni on the eastern versant of the Andes). The hairs of the dorsum, crown, upper limbs, and tail are golden-orange, with grey bases and dark red-brown or blackish-brown tips, generating a distinctly dark, often red-brown appearance, more striking than the relatively drab fur colors (more tan or yellowish-brown to grayish-brown) of other Bassaricyon (Figure 3). The fur of the cheeks, chin, venter, and underside of the limbs is yellow to the bases, often washed with orange. The fur of the face in front of the eyes is shorter and gray or buff with black tipping, sometimes with a pale cream ring around the eyes. The hairs of the tail are strongly tipped with gold, or with both golden and blackish-brown tipping. In contrast to specimens of other Bassaricyon, the tail is not conspicuously banded, though when viewed in the right light, a banding pattern of alternating golden and brown hues is weakly apparent in some specimens. A white terminal tail tip is present in a minority of individuals.

Like other Bassaricyon, the cranium of Bassaricyon neblina is long relative to its width, with a moderately long and broad rostrum, an elongate and somewhat globose braincase with a smooth dorsal surface, and moderately developed postorbital processes.

In Bassaricyon neblina, the temporal ridges do not meet to form a sagittal crest, even in older animals. The postdental palate is usually flared laterally, but is smoothly parallel-sided, tapers posteriorly, or bears only weaker bony flaring in other Bassaricyon (Figures 4–7, 19). At its more extreme development (e.g., in FMNH 70726), the portion of the bony palate sitting behind M2 is almost continuously joined to the postdental palate by a continuous shelf of bone, rather than bearing a deep excavation separating the molar-bearing portion of the bony palate from the postdental shelf (Figure 19). The auditory bullae are very small in the Olinguito relative to other Bassaricyon, both in length and vertical inflation, and the external auditory meatus is considerably narrower in diameter, on average (Figures 4–7). The median septal foramen of the anterior palate (Steno’s Foramen), between the paired incisive (or anterior palatal) foramina, is usually well-developed. The mandible is proportionally less elongate than in other Bassaricyon, with a proportionally larger and more vertically-oriented coronoid process (Figures 4–5). The first two upper premolars are caniniform, similar in size and shape to those of other Bassaricyon. P3 is usually relatively smaller in Bassaricyon neblina than in other Bassaricyon. P4 is similar in structure to congenersbut is relatively larger with a more bulbous protocone and more prominent metacone. M1 and M2 are proportionally lengthened and considerably more massive in appearance, especially relative to skull size, than in other Bassaricyon. p4 is variable in size among Bassaricyon neblina subspecies, generally smaller than other Bassaricyon in Bassaricyon neblina ruber and Bassaricyon neblina hershkovitzi, but proportionally quite large in Bassaricyon neblina neblina. m1 is relatively much larger in Bassaricyon neblina than in other Bassaricyon;each of the four major cusps that define the subrectangular shape of this tooth are massive and bulbous, and the posterior portion is especially broadened, with the metaconid and hypoconid particularly large and laterally expanded relative to congeners. m2 is also often expanded in size in Bassaricyon neblina relative to other Bassaricyon.

Natural history: Our field observations document that Bassaricyon neblina is nocturnal, arboreal, frugivorous, and probably largely solitary (compiled during July and August 2006 at Otonga Forest Reserve in Ecuador: 00°41’S, 79°00’W; for faunal and floral context see Freiberg and Freiberg 2000, Nieder and Barthlott 2001, Jarrín-V 2001). It occupies cloud forest canopies and is an adept leaper. It has a single pair of mammae and probably raises one young at a time. Notes associated with AMNH 14185, the first specimen to arrive in a museum, mention that it was “shot at 2 pm [an error for 2 am?] in high trees while coming down mountain to feed on guavas; strictly nocturnal.”

An adult female Olinguito (an animal named “Ringerl”, Bassaricyon neblina osborni, Figure 15) that lived at the Louisville Zoological Park and the National Zoological Park in Washington during 1967–1974 made vocalizations different from those of other Bassaricyon according to Poglayen-Neuwall (1976). Poglayen-Neuwall (1976) figured a picture of this animal in characteristic estrus behavior and in various other circumstances (see below for more discussion of this captive Olinguito).

Previous identifications and references:

Though described taxonomically for the first time in this paper, the Olinguito (heretofore misidentified as other species of Bassaricyon) has been represented in museum collections for more than a century, has been exhibited in zoos, has had its karyotype published, and has been included in published molecular phylogenetic studies.

Olinguito museum specimens previously reported in the literature include specimens from Gallera, Colombia, mentioned by Allen (1912, 1916) (AMNH 32608 and 32609, as “Bassaricyon medius”); a specimen from Santa Elena, Colombia, reported by Anthony (1923) (AMNH 42351, as “Bassaricyon medius”), specimens from “El Duende Regional Reserve” (2200 m asl; 04°02'55.6"N, 76°27'28.4"W)” and “Los Alpes, Florida, 2250 m asl” in Valle del Cauca Department, Colombia (mammal collection of the Universidad del Valle, Cali, Colombia, specimen numbers 12736, 13700) discussed by Saavedra-Rodríguez and Velandia-Perilla (2011) (as “Bassaricyon gabbii”); and a skull from San Antonio, Huila Department, Colombia (FMNH 70727) figured by Prange and Prange (2009) (as “Bassaricyon gabbii”, designated above as the holotype of Bassaricyon neblina hershkovitzi). One Olinguito specimen, AMNH 32609, bears an unpublished scientific name, “Bassaricyon osborni”, written on the skull and on the tags, apparently during the early twentieth century—correctly reflecting an understanding that the specimen represented an undescribed species. This appellation (a “manuscript name”) is likely attributable to J.A. Allen or H.E. Anthony (seemingly too early to be G.H.H. Tate). In any case, the name was never published, and by 1923, Allen had passed away (in 1921) and Anthony had decided that the specimen in question was best referable to Bassaricyon medius (see Anthony 1923). We have chosen to validate this name under our own authorship, above, as a subspecies of Bassaricyon neblina.

Mejía Correa (2009) reported camera-trap photos of a species of Bassaricyon at Munchique in Colombia; these records presumably represent Bassaricyon neblina osborni, the only Bassaricyon recorded at Munchique.

Molecular data for Bassaricyon neblina from a cell line were first generated and used in a phylogenetic study of carnivore relationships by Ledje and Arnason (1996a, 1996b), apparently the same animal whose karyotype was reported and discussed by Wurster-Hill and Gray (1975) (also Nash 2006). DNA sequence data (12S rRNA, cytochrome b) from this sample, available on Genbank, have been used in various other published studies (e.g., Flynn and Nedbal 1998, Koepfli and Wayne 1998, Emerson et al. 1999, Flynn et al. 2000, Gaubert et al. 2004, Marmi et al. 2004, Flynn et al. 2005, Fulton and Strobeck 2007, Yonezawa et al. 2007, Wolsan and Sato 2010, Nyakatura and Bininda-Emonds 2012, but not in some important studies, e.g., Koepfli et al. 2007, Agnarsson et al. 2010). This cell line apparently originated from the zoo animal “Ringerl” (discussed by Poglayen-Neuwall [1976]), an adult female Olinguito (Bassaricyon neblina osborni, originally from mountains near Cali, Colombia), apparently exhibited at the Louisville Zoo, National Zoo, Tucson Zoo, Bronx Zoo, and possibly Salt Lake City Zoo during the late 1960s and 1970s (Ingeborg Poglayen-Neuwall, pers. comm. to R. Kays, 2006). Ivo Poglayen-Neuwall (in litt. to C.O. Handley, Jr., 6 November 1964) mentioned another Bassaricyon, a young adult male at the Louisville Zoo, also from Cali, received in 1964, that seems also to have been an Olinguito (“shows the following unusual physical features: (1) strikingly round head… (2) very short, round ears! (3) rather short tail (no amputation!)”). This latter animal seems not to be discussed in Poglayen-Neuwall’s various publications on olingos, and it is unclear what became of it.