We conducted a survey of bats on the Caribbean island of Nevis from 24–29 April 2016. Live traps used included a harp trap and 6 m or 12 m mist nets set singly in the forest understory or on a triple-high telescoping pole system. Traps were regularly monitored for the first 2–3 hours after sunset when bat activity is the highest after they leave their roosts to feed. Individuals not kept as part of the representative collection documenting the species diversity were released at point of capture. A combined scientific research and export permit (F002) was issued through the authority of the Nevis Historical and Conservation Society. An Animal Use Protocol (2016-01) was obtained from the Royal Ontario Museum Animal Care Committee. An import permit (#2016-02101-4) was authorized by the Canadian Food Inspection Agency. Use of wild mammals followed the guidelines of the American Society of Mammalogists (Sikes et al. 2016).

The cytochrome c oxidase subunit 1 (CO1) gene is the best represented molecular marker for Chiroderma on the genetic sequence database GenBank (www.ncbi.nih.gov/genbank). There are 117 samples from nine countries in Central and South America (Brazil, Ecuador, El Salvador, French Guiana, Guatemala, Guyana, Mexico, Panama, and Suriname). We add 26 new sequences to bring the sample size to 143 sequences representing 12 countries in the Neotropics, including Venezuela, Peru, and Nevis, and five species in the genus (Appendix 1). There are no tissue samples or nucleotide sequences on GenBank of any genes for the recently described Chiroderma vizottoi (Taddei and Lim 2010). Outgroup taxa were other genera in the subtribe Vampyressina Baker et al., 2016 (Platyrrhinus incarum Thomas, 1912 and Uroderma bilobatum Peters, 1866) of the New World leaf-nosed bats, for direct comparison to Baker et al. (1994) in their analysis of Cytb. Alternative phylogenetic relationships within the subtribe are given by Baker et al. (2016) and Rojas et al. (2016). We also analyzed Cytb, but there are only 11 sequences on GenBank, although we did add one new sequence of Chiroderma trinitatum gorgasi from Panama (Appendix 2).

Molecular methods for new sequences of CO1 follow the protocol for DNA extraction, PCR amplification, and automated nucleotide sequencing outlined in Lim (2017). For Cytb, extraction, amplification, and sequencing followed Lim et al. (2008). Base calls were confirmed with bidirectional sequences and aligned using Sequencher version 4.8 (Gene Code Corporation, Ann Arbor, Michigan). Phylogenetic and molecular evolutionary analyses were conducted using MEGA version 6 (Tamura et al. 2013). For a robust comparison of phylogeny, we used parsimony as a method that minimizes evolutionary change without an explicit model of evolution and maximum likelihood as a probabilistic method with an explicit model of evolution. Maximum parsimony used the subtree pruning regrafting inference method with 500 bootstrap replicates to test branch supports. Maximum likelihood used the Tamura 3-parameter substitution model and gamma distributed rates with invariant sites for COI as determined by the best fit test. For Cytb, the Tamura Nei model and gamma rates were the best fit. Tree inference used nearest neighbor interchange heuristic inference with 500 bootstrap replicates. Genetic distances were calculated with the Tamura 3-parameter model with gamma distributed rates among sites for the larger COI dataset.

Morphological and morphometric comparisons included 138 specimens from five species of Chiroderma, including two C. improvisum, four C. doriae, seven C. salvini, 58 C. trinitatum, and 66 C. villosum (Appendix 3). We also analyzed the holotypes of C. trinitatum gorgasi Handley, 1960 and C. trinitatum trinitatum Goodwin, 1958, but did not have specimens of the most recently described C. vizottoi. Only adults (defined as having closed cranial sutures and complete epiphyseal ossification of metacarpal and phalanx joints) of both sexes were examined. Specimens are deposited in the following institutions; Royal Ontario Museum (ROM, Toronto, Canada); National Museum of Natural History (USNM, Washington, DC, USA); American Museum of Natural History (New York, USA); Texas Tech University (Lubbock, USA); and Field Museum of Natural History (Chicago, USA).

Measurements defined below were taken with digital calipers accurate to 0.01 mm following the descriptions of Handley (1960): forearm length (FA); greatest length of skull (GSL); interorbital width (IOW); postorbital width (POW); braincase width (BCW); condyloincisive length (CIL); zygomatic breadth (ZB); width across upper molars (M-M); width across upper canines (C-C); and length of maxillary toothrow (C-M). An analysis of variance (ANOVA) for each measurement and a multivariate analysis of variance (MANOVA) were performed to examine the significance of morphometric divergence among species of Chiroderma. The level of significance was p = 0.05 for all statistical tests. The homoscedasticity of each variable was tested using Bartlett’s test with the R package mvoutliers. Statistical analyzes were performed using R 3.1.0 (R Core Team 2005) and PAST 2.17. Variables were log-transformed and a correlation matrix was used in a Principal Components Analysis (PCA) to assess phenetic differences in multivariate morphological space.

For COI, the 657 basepairs (bp) at the 5’ end were available for most (82%) of the specimens analyzed. The complete 1140 bp of Cytb were available, including the newly generated sequence, for most (75%) of the specimens analyzed. The topology of the Chiroderma COI maximum likelihood tree identified six primary terminal clades with (1) C. salvini as sister species to all other taxa; (2) C. improvisum and (3) C. villosum as sister species; and (4) C. doriae sister to (5) C. trinitatum trinitatum with (6) C. trinitatum gorgasi sister to these taxa (Fig. 3; Suppl. material 1: Fig. S1). These phylogenetic relationships were supported by bootstrap values ≥85 and were congruent with the maximum parsimony tree (Suppl. material 2: Fig. S2), which had bootstrap values ≥73. Not surprisingly for linked mtDNA loci, the same interspecific topology was recovered by the smaller Cytb dataset analyzed by maximum likelihood (Suppl. material 3: Fig. S3) and maximum parsimony (Suppl. material 4: Fig. S4), except for lower bootstrap supports. The unexpected result was the paraphyly of C. trinitatum in relation to C. doriae. The sister-group relationship of C. t. trinitatum and C. doriae was well supported by values ≥73 in all molecular analyses.

Interspecific genetic distances of the larger COI dataset ranged from 11.3% between C. doriae and C. salvini to 2.5% between C. doriae and C. t. trinitatum (Table 1). The sequence divergence between C. t. trinitatum and C. t. gorgasi was 3.9%. Intraspecific distances were 1% within C. villosum, 0.9% within C. t. trinitatum, and 0.2% within C. doriae, but three taxa were represented by only one sample.

Cranial and body measurements for the six taxa of Chiroderma identified in the molecular analyses are shown in Table 2. Chiroderma trinitatum gorgasi and C. trinitatum trinitatum are the smallest members of the genus, whereas C. improvisum is the largest for most measurements. In the PCA, there are three main groups of species (Fig. 4). The first group is formed by the smaller taxa C. t. gorgasi and C. t. trinitatum. The second group has species with medium size, C. villosum and C. salvini, and the third group is formed by the largest species of the genus, C. doriae and C. improvisum. The first and second principal components (PC1 and PC2) explained 94.5% of the total variation. PC1 shows a pattern in general size variation and is explained mostly by C-M, C-C, and FA. PC2 has positive loadings for most measurements, especially IOW, with the exception of C-M, C-C, M-M, and ZB that have negative loadings (Table 3). All the species seem to occupy the entire range of PC2, indicating that the contrast among measurements is negligible and it is not responsible for the separation of groups.

All variables had p > 0.05 for Bartlett’s test of homoscedasticity, indicating constant variances (p values: FA = 0.06, GSL = 0.25, IOW = 0.59, POW = 0.3l, BCW = 0.06, CIL = 0.45, ZB = 0.08, M-M = 0.32, C-C = 0.08, and C-M = 0.06). The MANOVA and the ANOVA demonstrated that C. t. trinitatum and C. t. gorgasi are significantly different from the other taxa of Chiroderma (P < 0.001) for all measured variables (Appendix 3). By contrast, C. t. gorgasi and C. t. trinitatum are not significantly different from each other (p = 0.16, F = 56.0). However, the ANOVA showed that one measurement, BCW (p = 0.01; F = 62.0), was significantly larger for C. t. trinitatum than for C. t. gorgasi. All other cranial measurements had smaller mean values for C. t. trinitatum than for C. t. gorgasi.

Although similar in size, C. t. trinitatum has a more robust breadth of the braincase than C. t. gorgasi. Chiroderma t. trinitatum also has an accessory cusp on the second lower premolar, which is absent in C. t. gorgasi (Fig. 5). In the genetic analyses, C. t. trinitatum is well supported as the sister species to C. doriae and does not share a most recent common ancestor with C. t. gorgasi. We consider this as a previous example of a cryptic species and therefore now recognize C. gorgasi as a distinct species from C. trinitatum. Sáez and Lozano (2005: 111) considered cryptic species to be “groups of organisms that are morphologically indistinguishable from each other, yet found to belong to different evolutionary lineages”. They also stated that “after detailed comparisons of morphological and non-morphological features, we can often establish key morphological characters for their identification. In those cases, we can then refer to pseudo-cryptic or pseudo-sibling species”. Because Handley’s original description was qualitative and univariate, we offer an amended description of this taxon.