Seven new species within western Atlantic Starksia atlantica, S. lepicoelia, and S. sluiteri (Teleostei, Labrisomidae), with comments on congruence of DNA barcodes and species

Abstract Specimens of Starksia were collected throughout the western Atlantic, and a 650-bp portion of the mitochondrial gene cytochrome oxidase-c subunit I (COl) was sequenced as part of a re-analysis of species diversity of western Central Atlantic shorefishes. A neighbor-joining tree constructed from the sequence data suggests the existence of several cryptic species. Voucher specimens from each genetically distinct lineage and color photographs of vouchers taken prior to dissection and preservation were examined for diagnostic morphological characters. The results suggest that Starksia atlantica, Starksia lepicoelia, and Starksia sluiteri are species complexes, and each comprises three or more species. Seven new species are described. DNA data usually support morphological features, but some incongruence between genetic and morphological data exists. Genetic lineages are only recognized as species if supported by morphology. Genetic lineages within western Atlantic Starksia generally correspond to geography, such that members of each species complex have a very restricted geographical distribution. Increasing geographical coverage of sampling locations will almost certainly increase the number of Starksia species and species complexes recognized in the western Atlantic. Combining molecular and morphological investigations is bringing clarity to the taxonomy of many genera of morphologically similar fishes and increasing the number of currently recognized species. Future phylogenetic studies should help resolve species relationships and shed light on patterns of speciation in western Atlantic Starksia.


Introduction
Th e description of six new species of Caribbean Starksia by Williams and Mounts (2003) capped more than 100 years of systematic research on this New World labrisomid genus. It would have been reasonable to assume after such eff ort that there is little about the group left to discover. But the utilization of modern DNA barcoding techniques in taxonomic studies is revealing the need to reanalyze existing species classifi cations of many groups of animals and, in combination with traditional morphological analyses, resulting in the recognition of numerous new species (e.g., Crawford et al. 2010, Hebert et al. 2004, Pauls et al. 2010, Pöppe et al. 2010, Ward et al. 2008, Zemlak et al. 2009). Western Atlantic shorefi shes are no exception (e.g., Tornabene et al. 2010;Victor 2007Victor , 2010. Particularly for small cryptic reef fi shes such as Starksia blennies, we do not know where we stand in terms of understanding species diversity, and our current concepts may be surprisingly incomplete. Starksia fi shes inhabit shallow to moderately deep (to ca. 30 m) rock and coral reefs in the western Central Atlantic and eastern Pacifi c oceans. Th ey are small (Atlantic species are generally < 40 mm SL) and cryptic, but they often exhibit bright orange or red coloration in life. Twenty-one species are currently recognized in the western Atlantic (Williams and Mounts 2003), six of which are considered members of the S. ocellata species complex (Greenfi eld 1979).
Th e purpose of this paper is to describe the systematic results of our recent genetic and morphological investigations of western Atlantic Starksia, work that was prompted by our discovery of incongruences between preliminary genetic data and the current species classifi cation. We describe seven new species within S. atlantica, S. lepicoelia, and S. sluiteri and provide keys to the species of each of those species complexes. We provide photographs of living and preserved pigment patterns to help in future identifi cations of the included species and in distinguishing them from western Atlantic Starksia species likely to be discovered in the future. Finally, we discuss geographical distributions of Starksia species and comment on congruence between DNA barcoding data and morphologically recognizable species.

Materials and methods
Specimens used in this study were collected from Barbados, Belize, Bahamas, Curacao (Netherland Antilles), Florida, Honduras, Panama (Atlantic), Saba Bank (Netherland Antilles), St. Th omas (U.S. Virgin Islands), Tobago (Trinidad and Tobago), and Turks and Caicos. Th at material and additional museum specimens examined are listed in the appropriate species and comparisons sections. Starksia specimens included in the genetic analysis but not in the species accounts are tabulated in Appendix 1. Institutional abbreviations for collections follow Sabaj Pérez (2010).
Specimens were collected with quinaldine sulfate, rotenone, or clove oil using snorkel gear or scuba depending on depth. Field protocol involved taking digital color photographs of fresh color patterns and subsequently a tissue sample (muscle, eye, or fi n clip) for genetic analysis. For many, particularly small specimens, it was necessary to remove the posterior 1/3 to 1/2 of the body to obtain enough tissue for genetic analysis. Voucher specimens were preserved and later used to investigate diagnostic morphological features of each recovered genetic lineage. Field measurements of standard length (SL), to the nearest 0.5 mm, were made by viewing specimens against a plastic ruler under a dissecting microscope. Lengths of voucher specimens generally were not re-measured in the lab because many vouchers are now incomplete specimens. Th ose that were measured in the lab were measured to the nearest 0.1 mm with digital calipers or with the aid of an ocular micrometer in a dissecting microscope. Lengths of head (HL) and genital papilla were measured to the nearest 0.1 mm with the same ocular micrometer and microscope. To ensure that we were not introducing bias due to shrinkage of specimens after preservation, head length as a percentage of SL was calculated only for specimens in which both measurements were made from preserved specimens. Counts of dorsal-, anal-, and caudal-fi n rays were made from digital radiographs of specimens, from preserved specimens, or from photographs of voucher specimens taken prior to dissection. We followed Böhlke and Springer (1961) in counting the last two segmented rays of the dorsal and anal fi ns separately. Lateral-line scales were not counted because too many scales are missing on most specimens. Th is is likely due to the long time the specimens were held for processing prior to preservation and the physical manipulation of the specimens during processing. Pores from the circumorbital ossifi cations are either uniserial or paired; the positions of any paired pores are described based on their position relative to the orbit as though it were a clock; on the left side, for example, a pair of pores at 3 o'clock is on the posterior margin of the orbit, a pair at 6 o'clock is on the ventral margin.
Molecular techniques employed at the Smithsonian are as described below. Methods utilized to sequence DNA from specimens from Barbados, Honduras, Panama, and St. Th omas are as outlined in Victor (2010). Tissue samples for molecular work were stored in saturated salt buff er (Seutin et al. 1990) or in 95% ethanol. Genomic DNA was extracted from up to approximately 20 mg minced preserved tissue via an automated phenol:chloroform extraction on the Autogenprep965 (Autogen, Holliston, Massa-chusetts) using the mouse tail tissue protocol with a fi nal elution volume of 50 μL. For polymerase chain reaction (PCR), 1 μL of this genomic DNA was used in a 10 μL reaction with 0.5 U Bioline (BioLine USA, Boston, Massachusetts) Taq polymerase, 0.4 μL 50 mM MgCl 2 , 1 μL 10× buff er, 0.5 μL 10 mM deoxyribonucleotide triphosphate (dNTP), and 0.3 μL 10 μM each primer FISH-BCL (5'-TCAACYAATCAYAAAGA-TATYGGCAC) and FISH-BCH (5'-TAAACTTCAGGGTGACCAAAAAATCA). Th e thermal cycler program for PCR was 1 cycle of 5 min at 95°C; 35 cycles of 30 s at 95°C, 30 s at 52°C, and 45 s at 72°C; 1 cycle of 5 min at 72°C; and a hold at 10°C. Th e PCR products were purifi ed with Exosap-IT (USB, Cleveland, OH) using 2 μL 0.2× enzyme and incubated for 30 min at 37°C. Th e reaction was then inactivated for 20 min at 80°C. Sequencing reactions were performed using 1 μL of this purifi ed PCR product in a 10 μL reaction containing 0.5 μL primer, 1.75 μL BigDye buff er, and 0.5 μL BigDye (ABI, Foster City, California) and run in the thermal cycler for 30 cycles of 30 s at 95°C, 30 s at 50°C, 4 min at 60°C, and then held at 10°C. Th ese sequencing reactions were purifi ed using Millipore Sephadex plates (MAHVN-4550; Millipore, Billerica, Massachusetts) per manufacturer's instructions and stored dry until analyzed. Sequencing reactions were analyzed on an ABI 3730XL automated DNA sequencer, and sequence trace fi les were exported into Sequencher 4.7 (GeneCodes, Ann Arbor, MI). Using the Sequencher program, ends were trimmed from the raw sequences until the fi rst and last 10 bases contained fewer than 5 base calls with a confi dence score (phred score) lower than 30. After trimming, forward and reverse sequences for each specimen were assembled. Each assembled pair was examined and edited by hand, and each sequence was checked for stop codons. Finally the consensus sequence (655 bp) from each contig was aligned and exported in a nexus format (sensu Swoff ord 2002).
A neighbor-joining tree (Saitou and Nei 1987) and distance matrix were generated using Paup*4.1 (Swoff ord 2002) on an analysis of Kimura two-parameter distances (Kimura 1980). Th e neighbor-joining tree is not intended to refl ect phylogenetic relationships. Th e labels for each entry on the tree is our DNA number, and we include that number in the material examined sections and fi gure captions. Abbreviations used in DNA numbers refl ect geographical location: BAH -Bahamas, BAR -Barbados, BLZ -Belize, BRZ -Brazil, CUR -Curacao, FLA -Florida, HON -Honduras, PAN -Panama, SAB -Saba Bank (Netherland Antilles), STVI -St. Th omas Virgin Islands, TCI -Turks and Caicos, TOB -Tobago. COI sequences are deposited in Genbank (accession numbers HQ543038-HQ543055, HQ571151-HQ571164, HQ600864-HQ600963).

Results
A neighbor-joining tree derived from western Atlantic Starksia COl sequences is shown in Fig. 1. Th irteen of the 21 currently recognized western Atlantic Starksia species are represented in the tree: S. atlantica, S. culebrae, S. elongata, S. fasciata, S. guttata, S. hassi, S. lepicoelia, S. multilepis, S. nanodes, S. occidentalis, S. ocellata, S. sluiteri, and S. starcki. Four species, S. culebrae from the U.S. Virgin Islands, S. guttata from Tobago, S. occidentalis from Belize, and S. ocellata from Florida, cluster on the tree but represent genetically distinct lineages. Th ose results support Greenfi eld's (1979) recognition of a S. ocellata species complex with several allopatric component species. Similarly, S. atlantica, S. lepicoelia, S. nanodes, and S. sluiteri comprise multiple, geographically distinct, genetic lineages, suggesting that they also represent species complexes comprising multiple allopatric species. We do not deal further with the S. nanodes complex in this paper because no genetic data is available from the type locality, Bahamas, and we are thus uncertain if any of the four genetic lineages on the tree (Barbados, Belize, Panama, and Saba Bank) represents S. nanodes Böhlke and Springer 1961. We also do not deal further with fi ve species, S. elongata, S. fasciata, S. hassi, S. multilepis, and S. starcki (but see discussion of S. fasciata under the S. sluiteri complex section). Each of those species is represented in our material from only one geographical location, and material from additional geographic locations is needed to determine if they represent species complexes. We note that our material of S. elongata, S. fasciata, S. hassi, and S. multilepis is from the type localities of those species or relatively close by; the type locality of S. starcki, however, is Florida, and our specimen is from Belize.
Th e multiple genetic lineages within S. atlantica, S. lepicoelia, and S. sluiteri are the focus of the species treatments below. For each complex, we discuss congruence of the component genetic lineages with results of our morphological investigation. When diagnostic morphological features (primarily pigment) support the genetic data, we recognize genetic lineages as species. Greenfi eld (1979) noted that the ability to identify individuals of the S. ocellata complex to species based on morphology without prior knowledge of locality supports the recognition of the component populations as species vs. subspecies. We concur, and believe that the addition of the COl data strengthens this argument. Th ere are no available names for new species within any of S. atlantica, S. lepicoelia, and S. sluiteri complexes, and the seven unnamed species discovered are described herein as new. Keys to the species of the S. atlantica, S. lepicoelia, and S. sluiteri complexes are provided. We suggest that readers use the taxonomic key to western Atlantic Starksia provided by Williams and Mounts (2003) to identify S. atlantica, S. lepicoelia, and S. sluiteri and the keys in this paper to distinguish the members within each complex. Note that the sixth couplet of the Williams and Mounts (2003) key contains an error: 6b should lead the user to couplet 10, not 9 as indicated. Th e geographical locations listed for each species in our keys are the type locality plus any additional localities for which we have genetic data. Additional collecting and study are needed to determine the distributions of all western Atlantic Starksia species. Distance matrices for intra-and interspecifi c variation in COl sequences for the S. atlantica, S. lepicoelia, and S. slui-teri species complexes are provided in tables within the text. A distance matrix for all lineages is in Appendix 2. Longley (1934) described Starksia atlantica from a single specimen from Andros Island, Bahamas. Th e neighbor-joining tree derived from COI sequences (Fig. 1) includes fi ve distinct genetic lineages in the S. atlantica complex. Th e lineages from Barbados (BAR) and Panama (PAN) are known only from larvae or juveniles and are not discussed further. Th e Panama lineage is highly divergent in COl, and it likely represents a cryptic species within S. atlantica or one of the eight western Atlantic Starksia species not identifi ed in our material. Th e other three lineages-Curacao (CUR), Saba Bank (SAB), and Bahamas/Turks and Caicos/Belize (BAH/ TCI/BLZ) comprise specimens originally identifi ed as S. atlantica on the basis of absence of an orbital cirrus. (Note: Williams and Mounts (2003) correctly noted the absence of an orbital cirrus as diagnostic for S. atlantica in their key to western Atlantic Starksia [p. 147], but they erroneously stated "orbital cirri present" in their treatment of the species [p. 160].) Within the BAH/TCI/BLZ lineage, there are three sublineages, two from Belize and one from Bahamas/Turks and Caicos Islands (or four if the latter is viewed as two). We have identifi ed the specimens from Bahamas and Turks and Caicos as S. atlantica (Longley) based on the type locality (Bahamas) and pigment pattern, specifi cally the presence of two or three rows of block-like blotches on the trunk that are irregular in size and shape (Böhlke and Springer 1961). We found no consistent diff erences between specimens from the Bahamas and Turks and Caicos.

Starksia atlantica Species Complex
Th e two Belize sublineages diff er from other members of the S. atlantica complex by the presence of regular, vertical, brown bars on the trunk separated by narrow white interspaces and a well-defi ned horseshoe-shaped blotch on the cheek. Although those two sublineages are genetically similar to S. atlantica, we recognize the two lineages from Belize as a distinct species based on their strikingly diff erent pigment pattern and geographic separation. We found no consistent morphological variation between the two Belize sublineages and treat them as a single new species. Two specimens of this new species were illustrated as S. atlantica by Greenfi eld and Johnson (1981: Fig. 3A,B), who noted consistent diff erences in pigmentation on the body between their material from Belize and Honduras and the description of pigmentation for S. atlantica by Böhlke and Springer (1961). Th e other two genetic lineages of S. atlantica ( Fig. 1) are from Curacao (CUR) and Saba Bank (SAB). Th e Curacao specimens have a distinctive pattern of pigment on the cheek and pectoral-fi n base, and we recognize them as a distinct species. Th e single sequence from Saba Bank likely represents a new species (Fig. 1), but additional material is needed to confi dently assess its status (see "Remarks" under "Starksia sp." below). We describe two new species within the S. atlantica complex, S. sangreyae from Belize and S. springeri from Curacao. Specimens examined ranging from 12.0 to 19.0 mm SL; HL 29-34% SL (31% in holotype); male genital papilla adhered to fi rst anal spine proximally; papilla length between two-thirds and three-quarters length of fi rst anal spine, 0.6-1.0 mm; some females with very small genital papilla.
Pigment. Vertical brown bars present on trunk separated by narrow white interspaces; anteriormost 6 bars relatively uniform in all specimens; posterior bars often irregular or incompletely formed. A thick horseshoe-shaped blotch of pigment present on cheek. Bright orange pigment present on distal portions of pectoral-fi n rays, and pale orange pigment usually present on distal portions of posterior anal-, caudal-, and soft dorsal-fi n rays. Color pattern sexually dimorphic: males with pale red heads (vs. females without red coloration); relatively poorly defi ned horseshoe-shaped blotch of pigment on cheek that fades posteriorly (well-defi ned horseshoe-shaped blotch on cheek that is sometimes mirrored on operculum and pectoral fi n base); body bars tan and usually with some gold or green color in life (darker and without green/gold color but some posterior bars often with some orange pigment); body bars usually terminating ventrally dorsal to ventral midline (body bars usually extending to ventral midline); blotches of tan/gold pigment on base of dorsal fi n associated with body bars, and no tan/gold color present on anal fi n (bright orange markings on base of dorsal fi n associated with body bars and several bright orange spots on base of anal fi n); and large dark spot, roughly diameter of pupil or larger, on trunk at posterior end of dorsal fi n (two large dark spots on trunk, one at posterior end of dorsal fi n similar in size to that of males, and smaller spot at posterior end of anal fi n).
Color in preservative. Vertical bars on trunk, horseshoe-shaped blotch of pigment on cheek, and spot at posterior end of dorsal fi n (and anal fi n in females) retained in preservative; margins of at least some body bars in females with small dark spots; prominent patches of melanophores on jaws and gular region, and scattered pigment (heavier in females) on rest of head; dorsal fi n ranging from overall dusky to having concentrations of pigment on base of fi n associated with body bars; caudal fi n with light pigment on outer rays, and pectoral fi n with scattered melanophores over entire fi n; pelvic fi n clear.
Etymology. Th e species name is in honor of Mary Sangrey for her many years of work coordinating the intern program at the Smithsonian's National Museum of Natural History. Mary brought the intern application of the second author to the fi rst author's attention and took the fi rst steps toward procuring funding for Castillo's internship.
Distribution. All material that we examined is from Belize. Th e range of the species also apparently includes Honduras, as Greenfi eld and Johnson (1981) noted that a specimen of S. atlantica from Honduras has regular vertical bars of pigment on the body.

Diagnosis.
A species of Starksia distinguished by the following combination of characters: no orbital cirrus; trunk with irregular dark blotches on pale background; pectoral-fi n base with relatively straight margins defi ning pale gap that separates two dark blotches; cheek with distinctive dark and pale markings: anterior portion of cheek with prominent dark blotch, anteroventral and posterior margins of blotch well defi ned by pale regions; posterior pale area on cheek bordered posteriorly by thin, dark, anteroventral-to-posterodorsal streak of pigment along distal edge of preopercle.
Specimens examined ranging from 12.0 to 19.0mm SL; HL 25-32% SL (32% in holotype); genital-papilla length in 15.0-mm SL paratype 0.3 mm, one-fourth length of fi rst anal spine (broken); papilla adhered to spine proximally. Note: the presence of a small but measurable genital papilla on 15.0-mm SL paratype suggests that it is a male: although female Starksia sometimes have a small genital papilla, the 19 mm female holotype does not. As noted below, the 15 mm paratype has a pupil-size dark spot at posterior base of anal fi n, which usually characterizes females. We tentatively recognize this paratype as a male.
Pigment. (Note: a fi eld photograph of the 12.0-mm SL paratype is a dorsal view of poor quality, and only the head remains as a preserved voucher. Th e following   description is based on the 15.0-mm SL paratype and the 19.0-mm SL holotype.) Trunk with irregular dark blotches on pale background, most blotches consisting of orange chromatophores and melanophores in paratype; two dark spots present on trunk, large one at posterior end of dorsal fi n (larger than pupil diameter) and smaller spot at posterior end of anal fi n. Paratype with pale orange and brown pigment on head; tips of jaws with dark pigment in both paratype and holotype, but rest of jaws and gular region distinctly barred in holotype, mottled with tiny spots in paratype; cheek with distinctive dark and pale markings: anterior portion of cheek with prominent dark blotch, its anteroventral and posterior margins well defi ned by pale regions; posterior pale area on cheek bordered posteriorly by thin, dark, anteroventral-to-posterodorsal streak of pigment along distal edge of preopercle. Bright orange markings present on bases of dorsal fi n and anal fi ns, sometimes occurring in pairs; bright orange pigment also present on distal portions of pectoral-fi n rays; pale orange pigment present distally on at least some rays of soft dorsal, caudal, and anal fi ns; pectoral-fi n base with relatively pale gap separating two dark blotches, margins of gap relatively straight; dark blotches on pectoral-fi n base comprising orange chromatophores and melanophores. Color in preservative. (Note: pigmentation on trunk in preservative based on the only entire specimen, female holotype.) Trunk with irregular dark blotches on pale background; spots at posterior ends of dorsal and anal fi ns retained in preservative. Dark markings on head described above retained in preservative, mottled jaws and gular region of male(?) paratype strikingly diff erent from barred markings on female holotype; top of head in both specimens covered with scattered melanophores; dark and pale regions on cheek and pectoral-fi n base retained in preservative. Anal and pectoral fi ns with lightly scattered melanophores; caudal fi n with light pigment on outer rays; pelvic fi n clear.
Etymology. Named in honor of Victor G. Springer, Senior Scientist Emeritus, Smithsonian National Museum of Natural History, for his contributions to the systematics of blennioid fi shes, including Starksia, and for advice and friendship he has bestowed upon the fi rst author.
Distribution. Known only from Curacao, Netherland Antilles.

Remarks.
A DNA sequence from a single specimen collected at Saba Bank (Netherland Antilles) is genetically distinct from the other members of the S. atlantica species complex (SAB 0601019, Fig. 1). Our material includes color photographs of 9.0-and 15.1-mm SL females and the preserved 9.0 mm specimen (USNM 388032). Presumably the 9.0 and 15.1 mm specimens are the same species as the specimen represented on the tree, but we do not have tissue samples of either for genetic analysis or a preserved voucher of SAB 0601019 for morphological analysis.
Trunk pigment in the images and preserved specimen is similar to that of S. atlantica from the Bahamas and S. springeri from Curacao (i.e., mottled vs. barred as in S. sangreyae), but the Saba specimens lack the horseshoe-shaped blotch of pigment on the cheek characteristic of S. atlantica and the distinctive dark and pale markings on the cheek of S. springeri. Th e blotches of trunk pigment in the Saba Bank specimens are neither conspicuously block-like nor clearly organized in horizontal tiers as they are in S. atlantica. Specimens from Saba Bank presumably represent another new species within S. atlantica, but additional specimens are needed for comparative purposes and description. Members of the S. atlantica complex are diagnosed by the absence of an orbital cirrus. Starksia sangreyae is distinct in having regular vertical body bars separated by narrow pale interspaces and a well-defi ned horseshoe-shaped blotch on the cheek. Starksia springeri, S. atlantica, and the specimens from Saba Bank have irregular dark blotches on a pale background on the trunk, the blotches better defi ned in our S. atlantica material than in the other species and often more clearly arranged in two or three horizontal tiers. Starksia springeri, S. atlantica, and the Saba Bank specimens can be distinguished on the basis of pigment patterns on the cheek: specimens from Saba Bank lack cheek blotches; S. atlantica has a horseshoe-shaped blotch on the cheek; and S. springeri has a prominent dark blotch on the cheek bordered anteroventrally and posteriorly by pale areas and a thin, dark, anteroventral-to-posterodorsal streak of pigment along the distal edge of the preopercle. Although S. sangreyae and S. atlantica are easily separated based on trunk pigment, we note that both have a horseshoe-shaped blotch of pigment on the cheek; the blotch is most prominent and best defi ned in S. sangreyae females, often completely faded in preserved S. sangreyae males. Starksia atlantica and S. springeri can be separated based on pigment on the pectoral-fi n base: in S. atlantica, the pale gap between two blotches of darker pigment has wavy margins, whereas in S. springeri, the margins of the pale gap are relatively straight. Starksia springeri has XVIII dorsal spines vs. usually XIX in the other species (Table 1), but we have only one entire specimen of S. springeri on which to base counts. No other signifi cant diff erences were found in numbers of fi n rays or vertebrae among species of the S. atlantica complex.

Comparisons among Species of the Starksia atlantica Complex (Figs 4-5,
A photograph of a specimen identifi ed as S. atlantica from St. Croix, U. S. Virgin Islands (taken by William Smith-Vaniz) shows irregular block-like blotches on the body arranged in roughly 3 horizontal tiers, wavy margins on the pale gap that separates two darker areas on the pectoral-fi n base, and an irregular horseshoe-shaped blotch of pigment on the cheek. Th e U.S. Virgin Islands are thus likely part of the geographical distribution of S. atlantica Longley. Several USNM specimens identifi ed as S. atlantica from Navassa Island exhibit pigmentation that is somewhat intermediate between that of S. atlantica and S. sangreyae: bars of pigment are present on the trunk anteriorly as in S. sangreyae, but trunk pigment is more block-like posteriorly as in S. atlantica; Navassa specimens also have an irregular horseshoe-shaped blotch on the cheek as in S. atlantica. Further genetic and morphological investigation should help clarify species issues of S. atlantica from Navassa Island.

Starksia lepicoelia Species Complex
Böhlke and Springer (1961) described S. lepicoelia on the basis of numerous specimens from the Bahamas and one from St. John, U.S. Virgin Islands. Th e presence of a simple cirrus above the eye, two externally obvious pelvic-fi n rays, a completely scaled belly or at least posterior half scaled, and usually 17 anal-fi n soft rays are diagnostic of the species. Six genetic lineages in our data set cluster in the S. lepicoelia complex (Fig. 1). Th ere are no photographs or vouchers of the Barbados specimens (BAR on tree), and that lineage is not discussed further. Clearly it represents either a cryptic species within S. lepicoelia or one of the eight species of western Atlantic Starksia not identifi ed in our material. Two of the S. lepicoelia lineages are from the Bahamas/Turks and Caicos (BAH/TCI), and although sequence divergence for the two is 4-6% --much higher than typical intraspecifi c variation in western Atlantic Starksia -we were unable to fi nd consistent morphological diff erences between them and tentatively recognize them together as S. lepicoelia (Fig. 6). A fourth genetic lineage comprises specimens from Belize (BLZ), and a fi fth, specimens from Panama (PAN). Although those lineages diff er by only about 1% sequence divergence in COl, they are easily distinguished by color pattern. We describe the specimens from Belize and Panama as two new species. A sixth genetic lineage is represented in our tree by a single specimen from Saba Bank, Netherland Antilles. Based on that specimen and several lots of non-voucher material, we recognize the Saba Bank population as a fourth species within the S. lepicoelia complex.  Diagnosis. A species of Starksia distinguished by the following combination of characters: orbital cirrus present; belly scaled; trunk pale (pale red in life), without distinct bars or other markings; lips peppered with white spots in life; lacrimal region with single row of small white spots in life; jaws usually with lightly scattered melanophores in preserved specimens, without distinct banding or dark bars; entire gular region usually covered with scattered melanophores; total dorsal elements usually 27; total vertebrae usually 32; dorsal spines + anal soft rays + vertebrae modally 75.
Specimens examined ranging from 9.5 mm to 24.0 mm SL; HL 30-36% SL; length of male genital papilla two-thirds to equal length of fi rst anal spine, papilla 1.0-1.8 mm and free from spine.
Pigment. Both males and females with pale red to reddish brown trunk; indistinct vertical bars, if present, more prominent dorsally; two small (less than half pupil diameter) dark spots on posterior portion of trunk, one at posterior end of dorsal fi n and one at posterior base of anal fi n. Both sexes with pale red heads, scattered small white spots on anterior portions of lips, and single row of white spots beneath eye on lacrimal region; white spots representing absence of chromatophores in areas otherwise covered with pale orange to red pigment; eye with six or seven white spots around pupil, spots separated by darker areas (eff ectively a candy-stripe pattern). Males with prominent dark blotch on cheek and with small white spots extending from anterior portions of lips to posterior portions of jaws; females without dark cheek blotch and usually with larger white spots, blotches, or bands on posterior portions of jaws. Males with red pigment on dorsal fi n confi ned to blotches at base and little red pigment on rest of fi n and other median fi ns (but with scattered melanophores on dorsal, caudal, and anal fi ns); females with red pigment extending onto entire dorsal fi n and with prominent orange/red pigment on caudal and anal fi ns (but without prominent melanophores); males with yellowish brown pectoral fi n, females with pale orange to orange pectoral fi n; pelvic fi n clear.
Juvenile (BLZ 8123) color pattern: trunk pale orange, with some yellow mixed in; head with dark bar from anterior portion of eye to upper and lower lips; black triangle of pigment beneath eye; and black cap of pigment on head that extends anteriorly to vertical through middle of eye. Dorsal, anal, and caudal fi ns pale orange; bases of several dorsal-fi n elements with darker blotches of orange pigment; most anal-fi n elements with melanophore at base (typical of blennioid larvae), bases of about half of anal-fi n elements also with prominent orange spot.
Color in preservative. Males mostly pale, except with very dark blotch on cheek; trunk, belly, jaws, gular region, branchiostegals, operculum, top of head, nape, and all fi ns except pelvics with scattered melanophores, pigment on trunk fairly heavy in one male. Some females very pale, with only a few melanophores on gular region, cheek, branchiostegals, and on all fi ns except pelvics; other females with poorly formed dark blotch on cheek, fairly heavy pigment on gular region, branchiostegals, belly, dorsal fi n, and anal fi n; and lightly scattered melanophores on trunk, jaws, operculum, top of head, nape, caudal fi n, and pectoral fi n; pigment on head and nape usually lighter in females than in males.
Only anterior portion of body remains in juvenile voucher specimen (BZE 8123): body mostly pale; black cap of pigment on head, dark bar from anterior portion of eye to upper and lower lips, and black triangle of pigment beneath eye present in preservative.
Etymology. Th e species name is in honor of Lee A. Weigt, Head of the Smithsonian's Laboratories of Analytical Biology, in recognition of his contributions to the DNA barcoding of fi shes and his contributions to fi sh-collecting eff orts in Belize, Curacao, Florida, Tobago, and Turks & Caicos Islands.
Distribution. Known only from Belize, Central America. Paratypes Diagnosis. A species of Starksia distinguished by the following combination of characters: orbital cirrus present; belly scaled; trunk pale to tan (dark orange/tan to bright orange in life), without distinct bars or other markings; lips without conspicuous white spotting, distinct banding, or dark bars-usually with lightly scattered mel-anophores in preserved specimens; total dorsal elements 27; total vertebrae usually 32; dorsal spines + anal soft rays + vertebrae modally 75.
Specimens examined ranging from 18.5 mm to 22.5 mm SL; HL 34-38% SL; male genital-papilla length between two-thirds and three-fourths length of fi rst anal spine, papilla 1.0-1.25 mm and free from spine.
Pigment. Trunk dark orange/tan to bright orange, color nearly uniform-i.e., without indistinct dark bars and pale areas; two small (less than half pupil diameter) dark spots on posterior portion of trunk, one at posterior end of dorsal fi n and one at posterior base of anal fi n. Both sexes with orange heads, a few small pale spots on lips and lacrimal region, and six or seven white spots around pupil, spots separated by darker areas (eff ectively a candy-stripe pattern). Males with prominent dark blotch on cheek and uniformly orange/tan lips; females without dark blotch on cheek and with mottling of orange and pale blotches on lips. Males with red pigment on dorsal fi n largely confi ned to blotches at base and little red pigment on rest of fi n and other median fi ns (but with numerous melanophores on dorsal, caudal, and anal fi ns); females with bright orange spotting on dorsal, anal, and caudal fi ns (but without prominent melanophores except one dark spot sometimes present in anterior portion of spinous dorsal); males with yellowish brown pectoral fi n, females with orange pectoral fi n; pelvic fi n clear.
Color in preservative. Males tan, usually with fairly heavy pigment on head, trunk, and dorsal-, anal-, outer caudal-, and posterior portions of pectoral-fi n rays; prominent dark blotch on cheek retained in preservative; no dark spots, streaks or bars on lips. Females mostly pale, sometimes with noticeable concentrations of melanophores on cheek, jaws and gular region, but no prominent dark cheek blotch; lightly scattered melanophores usually present on branchiostegals, opercle, belly, median and pectoral fi ns; no conspicuous pattern of dark and pale blotches on lips, but light bar present across lips just posterior to symphysis and sometimes a few spots present just anterior to end of upper and lower jaws; posterior tips of upper and lower jaws usually pale.
Etymology. Named in honor of Jeff rey T. Williams, Smithsonian's National Museum of Natural History, in recognition of his work on blennioid fi shes, including Starksia. Jeff 's fi eld-collecting eff orts at Saba Bank, Tobago, and Turks and Caicos resulted in numerous specimens utilized in this study.
Distribution. Known only from Saba Bank, Netherland Antilles. Diagnosis. A species of Starksia distinguished by the following combination of characters: orbital cirrus present; belly scaled; trunk pale to dark tan (dark orange/tan to bright orange in life), without distinct bars or other markings; lips without conspicuous white spotting in life; ventral surface of lower jaw of males with one to three dark blotches or bars in preserved specimens, lips without distinct banding or dark bars; dorsal-fi n elements usually XX,7 -27 total; vertebrae usually 10+22=32; dorsal spines + anal soft rays + vertebrae modally 75.
Specimens examined ranging from 16.0-22.0 mm SL; HL 32-36% SL (32); male genital-papilla length between one-half and three-fourths length of fi rst anal spine, papilla 0.6-1.9 mm and free from spine. Pigment. Color in life known only for two females. Trunk dark orange/tan to bright orange, color nearly uniform or with indistinct dark bars and pale areas; two small (less than half pupil diameter), inconspicuous dark spots on posterior portion of trunk, one at posterior end of dorsal fi n and one at posterior base of anal fi n. Head orange, mottled with white patches; a few small, pale spots present on lips and lacrimal region; eye with six or seven white spots around pupil, spots separated by darker areas (eff ectively a candy-stripe pattern). Bright orange spotting on dorsal, anal, and caudal fi ns, and some orange pigment on pectoral fi n; pelvic fi n clear.

Color in preservative.
Trunk ranging from pale to dusky, belly with fairly heavy pigment in males and some females even if trunk pale. Males usually with prominent dark blotch on cheek (largest male, USNM 399909, PAN1419, with dark spots on cheek but no conspicuous blotch), females without dark cheek blotch. Underside of lower jaw with one to three dark spots or bars in males, middle one (situated roughly beneath a vertical through pupil) darkest and sometimes the only one noticeable; anterior marking, if present, sometimes extending onto lower lip as a few dark dots; no dark spots, streaks, or bars on lips in either sex, but portions of lips uniformly covered with melanophores in males and with at least a few spots in females; females usually with patch or bar of pigment (small and faint in some specimens) extending from lacrimal region across both lips. In males, branchiostegals dusky, upper part of cheek, opercle, and top of head pale to dusky; in females, head mostly pale, with isolated patches of spots on cheek, opercle, top of head, and branchiostegals. Dorsal, anal, caudal, and pectoral fi ns dusky in males, mostly pale in females with a few scattered spots on some fi ns.
Etymology. Named in recognition of the contributions by D. Ross Robertson of the Smithsonian Tropical Research Institute to the understanding of the diversity of shorefi shes of the New World and his generous facilitation of collecting in Panama.

Distribution. Known only from Panama (Atlantic)
Comparisons among Species of the Starksia lepicoelia Complex (Figs 10-11 Comparisons. Starksia lepicoelia and S. starcki are the only previously described western Atlantic Starksia with the combination of an orbital cirrus, two externally obvious pelvic-fi n rays, and a scaled belly (Williams and Mounts 2003). Starksia starcki is easily distinguished from the species of the S. lepicoelia complex by the presence of eight or nine irregular dark bars on the body and usually 19 segmented anal-fi n rays.  In life, S. weigti is easily distinguished from S. lepicoelia, S. williamsi, and S. robertsoni by the conspicuous pale round spots on the lips. In preservative, S. lepicoelia males are distinctive in having at least some very dark spots, streaks, or bars on the lips and lower jaw, and S. robertsoni males have at least one (up to three) dark spots or bars on the ventral portion of the lower jaw (but not on the lips). Although the diff erences are subtle, preserved males of S. williamsi typically can be separated from preserved males of S. weigti in having the lips uniformly covered with melanophores except for the pale anterior tips. In S. weigti males, lip pigment is variable, but there are usually one or two thin, faint, poorly formed bars of pigment following the pale anterior portions of the lips; posteriorly, the lips may be uniformly covered with melanophores as in S. williamsi or be quite pale.
Preserved female S. lepicoelia also have a distinctive lip pattern-alternating pale and dark areas. Although this banding pattern appears to be present in color images of S. williamsi, S. weigti, and S. robertsoni, it is not present in preserved females of those species, suggesting that in S. lepicoelia the banding comprises both chromatophores and melanophores whereas in females of the other species it comprises only chromatophores and thus is not retained in preservative. As in males, diff erences in head pigment between preserved female S. williamsi and S. weigti are subtle, but S. williamsi females have a relatively well-formed bar of pigment from the anterior portion of the lacrimal across both lips, whereas S. weigti females typically have only a light scattering of melanophores on the upper lip beneath the anterior portion of the lacrimal. Additionally, S. williamsi females tend to have a bit of dark pigment at the posteroventral corner of the orbit and another bit just ventral to posteriormost point of orbit; S. weigti females usually have more widely scattered pigment on the cheek --sometimes in a fairly cohesive spot. Th e head pigment of female S. robertsoni is very similar to that of S. williamsi, but modal diff erences in fi n-ray counts separate them, and they are geographically distinct. Specifi cally, S. williamsi-from the eastern Caribbean-typically has XIX,8 dorsal-fi n elements, whereas S. robertsoni-from Panama-typically has XX,7.
We examined color photographs and numerous preserved specimens from St. Croix, U.S. Virgin Islands, but we do not have genetic data for that material. Fresh specimens lack the diagnostic white spots on the lips of S. weigti. Preserved specimens most closely resemble S. lepicoelia in pattern of pigment on the lips and lower jaw, with females typically having at least some alternating pale and dark areas (nearly identical to that of S. lepicoelia in some specimens, not distinctive at all in others). Although most males have fairly uniform pigment on the lips and lower jaw, at least some males have the distinctive dark bars, spots, or streaks characteristic of male S. lepicoelia. If the St. Croix specimens represent one of the known S. lepicoelia species, it seems likely  based on geography and pigmentation that they are S. lepicoelia. However, we note that S. lepicoelia typically has 28 total dorsal elements and 17 anal-fi n soft rays, whereas the St. Croix specimens (15 counted) typically have 27 and 16, respectively (but 28 dorsal elements and 17 anal rays are not uncommon counts). Additional investigation, including genetic analysis, is needed. Lips with few or no white spots in life; dorsal-fi n spines modally XIX (also see "Comparisons," above) ..........S. williamsi (Saba Bank, Netherland Antilles)

Starksia sluiteri Species Complex
Metzelaar (1919) described Brannerella sluiteri from two specimens from Bonaire, Netherland Antilles. Longley (1934) synonymized Brannerella with Starksia Jordan and Evermann (type species Labrisomus cremnobates Gilbert, from the eastern Pacifi c). Böhlke and Springer (1961) concurred with Longley's synonymy, noting that Brannerella is distinctive in a single character, and generic recognition of one-character diff erences would require the erection of several new genera within Caribbean Starksia. Our material includes three genetic lineages originally identifi ed as S. sluiteri based on the taxonomic key of Williams and Mounts (2003) -one from Curacao, one from Tobago, and one from Belize/Honduras/Panama. Specimens in all three lineages modally have 13 pectoral-fi n rays, 20 or fewer dorsal-fi n spines, and two or three rows of dark spots or blotches along the body-features typical of S. sluiteri. We have identifi ed our genetic lineage from Curacao (CUR in Fig. 1) as S. sluiteri (Metzelaar) based on geography and morphology. In particular, the second row of dark markings (middle row when there are three) are distinctly round in S. sluiteri and in our Curacao speci-mens, whereas those markings are usually vertically elongate in our specimens from Belize (BLZ), Honduras (HON), and Panama (PAN). Additionally, although Metzelaar (1919) illustrated a male specimen in his original description, he did not mention any round, pale markings on the head-prominent diagnostic features in males of our specimens from Tobago (TOB) that are lacking in our male S. sluiteri from Curacao. We recognize the genetic lineage from Tobago, as well as that from Belize/Honduras/ Panama, as new species within the S. sluiteri complex and provide descriptions below. Böhlke and Springer (1961) noted that counts of dorsal-and anal-fi n elements in specimens of S. sluiteri they examined from off Colombia and Venezuela (XIX dorsal spines and 15-16 anal rays) diff er from those given by Metzelaar (XX and 17). Based on pigment, their Colombian and Venezuelan specimens appear to be S. sluiteri. Our specimens from Curacao, as well as Böhlke and Springer's two Venezuelan specimens (USNM 195750), have XIX dorsal spines and 15-16 anal rays. Th ere is thus a discrepancy between counts in our material and those reported by Metzelaar for the holotype. We examined a photograph of the holotype, and there appear to be XX dorsal-fi n spines as noted by Mezelaar; XX is likely a non-modal count for S. sluiteri. We note that there is more variation in dorsal-and anal-fi n counts in some Starksia species than suggested by Metzelaar's description; for example, S. greenfi eldi has XVIII-XX dorsal spines, 7-9 dorsal rays, and 14-16 anal rays. Diagnosis. A species of Starksia distinguished by the following combination of characters: orbital cirrus present; two to three rows of dark blotches on side of body, blotches in middle row (or ventral row if only two rows) mostly circular, never vertically elongate or oval; white (or pale), mostly round spots (absence of melanophores against a darker background) on at least portions of cheek, opercle, and gular region, this spotting pattern more prominent in males; males with dark blotch of pigment on anterior portion of spinous dorsal fi n; fi rst anal-fi n spine one-half to three-quarters length of male genital papilla; belly naked.
Pigment. Head and body pale yellow to pale orange, generally more orange in males, more yellow in females; posterior margins of most body scales covered with yellow or orange chromatophores mixed with melanophores, resulting in background pattern of chain-link or diamond-shaped markings. Two or three rows of dark markings on trunk in mature specimens, markings diff use in some specimens: dorsalmost row with 7-10 roughly square blotches that extend onto bases of dorsal-fi n elements (another dark blotch on nape in line with this row of markings); second row with 6-7 circular blotches situated just above lateral midline; lower row, if present, with 1-4 diff use, round to oblong blotches. A few to many white, mostly round spots on at least portions of cheek, opercle, and gular region and sometimes lower jaw; this pattern resulting from the absence of melanophores against a darker background and typically signifi cantly more prominent in males. Males also diff ering from females in having dark blotch of pigment on anterior portion of spinous dorsal fi n. Distinctive, dark-orange markings usually present on proximal portion of dorsal fi n where dark blotches in dorsalmost row of markings on body extend onto dorsal fi n; where those dark blotches extend onto two (vs. one) dorsal-fi n element, dark orange markings distinctly paired. Orange pigment also present on distal portions of pectoral-fi n rays and lighter orange pigment present on at least distal portions of second dorsal-, caudal-, and posterior anal-fi n rays; sometimes orange blotches present intermittently along lengths of second dorsal-, caudal-, and anal-fi n rays forming wavy stripes or bars of pigment on those fi ns. Orange pigment present on top of head, in bright ring around eye, and on nasal cirrus. Some specimens with dark orange pigment on snout, in blotches radiating from pupil, on operculum, and on dorsal portions of pectoral-fi n base. In one specimen most chromatophores on head and body yellow to yellowish orange, but those on nasal cirrus, around eye, and on fi ns distinctly orange.
Color in preservative. Diagnostic dark markings on trunk present as described above; diagnostic white, round spots on head described above present as distinctive pale markings in preserved specimens-head markings especially prominent in large males; trunk largely tan and peppered with dark dots, especially along posterior margins of scales; lips with mottled or barred pigment pattern; a fairly uniform covering of melanophores on snout, branchiostegals, pectoral-fi n base, and belly; eye sometimes surrounded by dark ring of pigment; top of head and nape usually darker than rest of head, pigment on nape usually in form of dark saddle extending over dorsal midline; two concentrations of melanophores usually visible on brain; dorsal and anal fi ns dusky, dark body blotches in upper row usually extending onto base of dorsal fi n; dorsal fi n of males with dark blotch between spines II-IV; caudal-fi n rays edged with dark pigment, outer rays with more uniform scattering of melanophores; proximal portion of pectoral fi n covered with scattered melanophores, distal portion with dark edging along rays; males sometimes with pigment on membranes between some pectoral rays distally; pelvic fi n clear.
Etymology. Th e species name is in honor of David W. Greenfi eld, in recognition of his work on blennioid fi shes, particularly his work on the Starksia ocellata complex.  Table 3. Colombia (Isla Providencia): MZUSP 107860, 1 (not a DNA voucher). Panama (San Blas Islands): USNM 399918, PAN 018.

Description. See
Specimens examined ranging from 9.0-19.0 mm SL; HL 29-33% SL (29%); length of male genital papilla 19-22% SL; papilla adhered to fi rst anal-fi n spine and extending well beyond it, spine two-thirds to three-quarters length of papilla.
Pigment. Head and body pale orange; posterior margins of most body scales covered with yellow or orange chromatophores mixed with melanophores, resulting in background pattern of chain-link or diamond-shaped markings. Two rows of dark markings on trunk: dorsal row with 9 roughly circular blotches that extend onto bases of dorsalfi n elements (another dark blotch on nape in line with this row of markings); ventral row with 6-7 blotches along middle of trunk, at least some vertically elongate to oval in shape; blotches generally not round, although one or more within row may be roughly so. Females with small dark spots on cheek, operculum, branchiostegals, lower jaw, gular, and pectoral-fi n base; spots smaller than pupil (several would fi t in pupil) but much larger than tiny dark dots that pepper most of head and trunk; males with dark, fat, crescent-shaped marking on cheek; orange chromatophores associated with head markings in both sexes. Both males and females lacking dark blotch of pigment on anterior portion of spinous-dorsal fi n. Prominent orange markings present on bases of dorsal-fi n elements above dark blotches along dorsal portion of trunk; where dark blotches extend onto bases of two dorsal-fi n elements, orange markings distinctively paired; other orange pigment including chromatophores on top of head, around eye, on nasal cirrus, and on tips of pectoral-, dorsal-, caudal-, and anal-fi n rays; those on pectoral fi n bright orange.
Color in preservative. Diagnostic dark blotches on trunk present as described above; diagnostic small dark spots on head in females and large blotch on cheek in males also distinctive in preserved specimens; body overall tan to dark tan. Males with uniform scattering of spots on lips and rest of head and pectoral-fi n base; dorsal, caudal, anal, and pectoral rays dusky -i.e., with pigment on membranes between fi n rays. Females with dark spots on lips, chin, snout, circumorbitals, and pectoral-fi n base; top of head and nape densely covered with melanophores; dorsal, caudal, anal, and pectoral rays edged in dark spots, but little or no pigment on membranes between fi n rays. Dark blotches on dorsal portion of trunk extending onto dorsal-fi n rays in both sexes; belly pale to lightly pigmented; pelvic fi n clear.
Etymology. Named in honor of Michael A. Lang, Director of the Smithsonian Marine Science Network (MSN) and Smithsonian Science Diving Program, in gratitude for the support MSN has provided for our Caribbean fi sh diversity studies and in recognition of the contributions Michael has made to science diving.
Remarks. A tissue sample from a single specimen off Honduras (HON 050 on tree in Fig. 1) produced a COl sequence very similar to those of our Belize specimens, and one from Panama (PAN 018) is approximately 1% diff erent. Th e Honduras specimen (Fig. 14A) has the diagnostic pigment on the cheek of male S. langi, and the Panama specimen (Fig. 14B) has the diagnostic small dark dots of female S. langi. We recognize the Honduras and Panama specimens as S. langi.
We lack tissue samples of Colombian specimens, but the fi ve specimens in UF 223370 from Cayos del Este (San Andrés) and a 16-mm SL specimen from Isla Providencia (Fig. 14C) appear to have the vertically elongate pigment blotches on the trunk diagnostic of S. langi. Pigment is somewhat faded in the UF specimens, but the 16mm SL female in the lot has dark spots on the head as in female S. langi. Although we include "Colombia" in the distribution list of this species above, we note that the Colombian specimens are from the Archipelago of San Andrés, Providencia, and Santa Catalina, a group of islands nearly 800 km from Colombia but only 220 km from Nicaragua. We have no material from continental Colombia, but S. sluiteri replaces S. langi off Venezuela.
Comparisons among Species of the Starksia sluiteri Complex (Fig. 15 Starksia langi is easily distinguished from S. greenfi eldi and S. sluiteri based on pigmentation of the trunk, head (females), and fi rst dorsal fi n (males). Th e trunk pigment of S. langi comprises both larger and more prominent markings than that of S. greenfi eldi and S. sluiteri, and only in S. langi are the markings in the second row vertically elongate (generally round in the other species and sometimes considerably more diff use in S. greenfi eldi). Starksia greenfi eldi lacks dark markings on the head in both sexes, and S. sluiteri lacks them in females; S. langi males have a prominent dark blotch on the cheek, and females have numerous small, discrete, dark spots. Males of S. langi lack a dark blotch on the anterior portion of the dorsal fi n, whereas this blotch is present in S. greenfi eldi and S. sluiteri.
Starksia greenfi eldi can be distinguished from S. langi and S. sluiteri by the white (or pale), mostly round spots (absence of melanophores against a darker background) on at least portions of cheek, opercle, and gular region. Th is pattern is present in both sexes but is often much more prominent in males. Williams and Mounts (2003) noted that S. sella, another species of Starksia known only from Tobago, has small pale spots on the head, but that species lacks dark blotches along the trunk, lacks a dark blotch in the anterior dorsal fi n of males, and may be larger (Williams and Mounts specimens of S. sella are 13.7-27.7 mm SL, our specimens of S. greenfi eldi are 11.0-23.0 mm SL).
S. sluiteri (Metzelaar) is most easily distinguished from S. langi by having the second row of trunk blotches almost perfectly round (vs. vertically elongate), in lacking conspicuous dark spots on the head (females), and in having a dark marking on the anterior portion of the dorsal fi n (males). From S. greenfi eldi, S. sluiteri diff ers in lacking pale round spots on the head. Although S. sluiteri and S. langi have very similar   chromatophore patterns, S. sluiteri appears to have more orange pigment on the second dorsal, caudal, and anal fi ns. In their descriptions of S. leucovitta, S. melasma, S. multilepis, S. rava, and S. sella, Williams and Mounts (2003) noted that those species belong to the S. sluiteri complex. Large genetic distances separate the species of the S. sluiteri complex, and our S. multilepis samples from Brazil are nearly as similar genetically to S. sluiteri as S. langi is (Fig. 1). We have no tissue samples of the other proposed members of the S. sluiteri complex for comparative purposes. Th ose species are not very similar to S. sluiteri in trunk pigment, particularly in lacking any bold markings. Starksia fasciata from the Turks and Caicos Islands (TCI 9204, TCI 9349, TCI 9350) is embedded within our S. sluiteri complex (Fig. 1), and S. fasciata is morphologically similar to species in that complex (Fig. 15). In Williams and Mounts (2003) diagnostic key, S. fasciata and S. sluiteri are in the same couplet, separated by pattern of pigment on the trunk (bars of trunk pigment in the former, rows of dark blotches in the latter). Male and female S. fasciata from the Turks and Caicos Islands (Fig. 15) are very similar to male and female S. langi from Belize in head pigmentation and in having prominent orange markings along the base of the dorsal fi n. More material is needed to determine if S. smithvanizi, a species that Williams and Mounts (2003) considered part of the S. fasciata complex, also is genetically aligned with the S. sluiteri complex. We reiterate that our neighbor-joining tree (Fig. 1) is not intended to refl ect phylogenetic relationships, and a species-level phylogeny derived from multiple genes should help resolve species and supra-specifi c relationships in the S. sluiteri complex.
Museum specimens examined from the Lesser Antilles (Dominica) and Puerto Rico appear to be S. sluiteri based on trunk pigment (round vs. elongate blotches in the second row of markings) and no conspicuous round pale spots on the cheek. Th e pigment is somewhat faded in those specimens, however, and more material, including tissue samples for genetic analysis, is needed. Two female specimens from Navassa (USNM 361059) are not S. sluiteri, as the markings in the second row of trunk blotches are elongate, not round. However, those markings are rectangular in the Navassa specimens, and the markings in the upper row are square-much more so than in our material of S. langi from the western Caribbean. Th e larger of the two females has some dark spots on the head as in S. langi. More material is needed. Other museum material examined (e.g., the UF specimens from Antigua and Mexico) are too faded to identify to species. Head without conspicuous pale round spots S. sluiteri (Netherland Antilles) Gilbert (1965) and Greenfi eld (1979) noted that some species of Starksia can only be distinguished on the basis of color patterns-i.e., they exhibit no other morphological diff erences except sometimes modal diff erences in counts. Greenfi eld (1979) surmised that color patterns on the lips and sides of the head may be important in species recognition in blennioid fi shes, which often live in cryptic habitats, in some cases (e.g., some chaenopsids) with only the heads typically visible. Our morphological investigation of the multiple genetic lineages within S. atlantica, S. lepicoelia, and S. sluiteri resulted in similar fi ndings-i.e., most of the member species within the three complexes are distinguished from one another solely on the basis of pigment patterns, sometimes only diff erences in pigment on the lips and cheeks. All diff erences in counts are modal. Morphological diff erences other than pigmentation separate some of the species complexes; for example, members of the S. atlantica complex lack an orbital cirrus, and those of S. lepicoelia have a scaled belly. Genetic divergence among species within each complex is generally smaller than that between complexes: 2-14% within S. atlantica, 1-9% within S lepicoelia, and 7-19% within S. sluiteri vs. 17-22% between S. atlantica and S. lepicoelia, 17-24% between S. lepicoelia and S. sluiteri, and 17-23% between S. atlantica and S. sluiteri (Tables 4-7). Th e genetic distances separating species of the S. lepicoelia complex are particularly small, and those species are separated on the basis of minor diff erences in pigmentation on the head. Larger genetic distances separate most species of the S. sluiteri complex, and more prominent diff erences in trunk pigmentation separate some of those species. Th ere is thus a correlation between small diff erences in COl sequences and minor diff erences in pigmentation, suggesting that pigment patterns may be among the fi rst morphological changes accompanying speciation in Starksia. Greenfi eld (1979) did not have the benefi t of genetic data for comparative purposes, but our COl data for four species in his S. ocellata complex (Fig. 1, Appendix 2) support his decision to recognize species almost entirely on the basis of minor diff erences in pigment. Although species recognition based on such limited morphological data may in general be a questionable practice, the congruence between Greenfi eld's (1979) S. ocellata species and the COl data supports this practice in Starksia.

Discussion and conclusions
Th ere is not, however, universal congruence between genetic divergence and recognizable morphological diff erences in our data set. One S. greenfi eldi specimen, TOB 9312, is 2% diff erent from other S. greenfi eldi, and one S. fasciata, TCI 9204, is 2% diff erent from other S. fasciata. Both of those values are high for intraspecifi c variation in fi shes in general (often well less than 1%), but we fi nd no morphological evidence Table 4. Average (and range) Kimura two-parameter distance summary for the Starksia atlantica species complex based on cytochrome c oxidase l (COl) sequences of individuals represented in the neighborjoining tree in Figure 1. Intraspecifi c averages are shown in bold. n/a = no average (one specimen). BAR -Barbados, SAB -Saba Bank, PAN -Panama.

1%
(1) --robertsoni 7% (6-7) 7% (7-8) (1-2) 7% (7) n/a n/a supporting the genetic divergences. Similarly S. sangreyae comprises two sublineages that are as genetically distinct in COl (2-3%) as S. sangreyae is from S. atlantica (2-3%), yet no consistent morphological diff erences were discovered, not even minor diff erences in color pattern. Even more puzzling, the two genetic sublineages of S. lepicoelia are 4-6% diff erent in COl, yet we found no morphological diff erences between them (Fig. 6). Very little material of one of those lineages is available, and further investigation is needed. Specimens in the two lineages were taken in the Bahamas at the same station, in 20-40 ft. of water off Great Stirrup Cay.
In contrast to the examples above, very little sequence divergence in COl exists between S. sangreyae from Belize and S. atlantica from Bahamas/Turks and Caicos (2-3%), yet those species are easily distinguished on the basis of trunk pigment. Similar incongruences between COl data and morphology have been documented. For example, Baldwin et al. (2009b) found two morphological (pigment) variants of the goby Coryphopterus venezuelae, yet those morphs are not genetically distinct. Victor (2010) pointed out incongruences between COl data and morphologically recognizable species in greenbanded gobies (Elacatinus spp.). Specifi cally, he noted that E. multifasciatus from the eastern Caribbean and E. panamensis from Panama are morphologically extremely similar, but exhibit 11.3% sequence divergence in COl; he further noted that despite prominent diff erences in color pattern between E. rubrigenus and E. panamensis, those species exhibit only 3.3% sequence divergence in COl.
Th ose examples notwithstanding, the general congruence between COl lineages and morphologically recognizable species in western Atlantic Starksia is remarkable, and we have found the same to be true in our genetic and morphological investigations of other shorefi sh genera (e.g., Baldwin et al. 2009a, Baldwin et al. 2009b, Tornabene et al. 2010. A paper summarizing Smithsonian investigations of western Central Atlantic shorefi sh diversity and the utility of DNA Barcoding in this work is in preparation. Cases in which incongruences exist between genetic and morphological data ultimately will be further investigated; because DNA barcoding involves sequencing a Table 6. Average (and range) Kimura two-parameter distance summary for the Starksia sluiteri species complex based on cytochrome c oxidase l (COl) sequences of individuals represented in the neighborjoining tree in Figure 1. Intraspecifi c averages are shown in bold.

1%
(0-2) Table 7. Range Kimura two-parameter distance summary for the Starksia atlantica, S. lepicoelia, and S. sluiteri species complexes based on cytochrome c oxidase l (COl) sequences of individuals represented in the neighbor-joining tree in Figure 1. Within-complex ranges are shown in bold.

S. atlantica complex (n=26)
S. lepicoelia complex (n=25) S. sluiteri complex (n=20) S. atlantica complex 2-14% --S. lepicoelia complex 17-22% 1-9% -S. sluiteri complex 17-23% 17-24% 7-19% relatively short segment of a single mitochondrial gene, adding additional genetic data may help resolve some confl icts. On the morphological side, adding information from early life history stages may be of value: the pelagic larval stages of many marine fi shes off er a suite of characters for study not present in adults. A striking element of our COl data for Starksia (Fig. 1) is the correlation between genetic lineages and geography within the S. atlantica, S. lepicoelia, S. sluiteri, and S. ocellata species complexes. Specimens from Bahamas, Belize, Curacao, Saba Bank, and Tobago never occur in more than one genetic lineage within each complex, yet the species complexes themselves are broadly distributed (Fig. 16). Starksia nanodes also appears to be a broadly distributed species complex, with geographically distinct genetic lineages in Panama, Barbados, Saba Bank, and Belize (Fig. 1). Greenfi eld (1979) proposed superspecies status (sensu Amadon 1966, Mayr 1963 for the S. ocellata complex based on its six allopatric component species, and the S. atlantica, S. lepicoelia, and S. sluiteri species complexes described herein could be categorized likewise (we note, however, that the superspecies category has not been widely adopted in systematic treatments of fi shes). It is not clear what evolutionary mechanisms are driving speciation within Starksia, but the life history of the group is characterized by a short pelagic phase of about two weeks (Victor, unpublished data). Although pelagic larval duration (PLD) is not always a good indicator of genetic structure (e.g., Bowen et al. 2006), a short PLD combined with restricted movement of adults may support the evolution of numerous allopatric species within a group by restricting gene fl ow among populations. It is premature to conduct a phylogeographic analysis of western Atlantic Starksia, but we concur with Greenfi eld (1979) that the division of some Starksia species into multiple allopatric component species is not typical of western Atlantic shorefi shes in general. As noted by Floeter et al. (2008), Briggs' (1974 two major biogeographic provinces of the Caribbean (western Caribbean plus Florida and West Indian/eastern Caribbean) are largely supported by recent genetic and biogeographical studies. Starksia is not the only exception to this general trend. Colin (2010) described fi ve eco-morphological suites of western Atlantic Elacatinus goby species that are similar to our Starksia species complexes in that each comprises multiple species usually with allopatric distributions, and the suites themselves are broadly distributed. Considerably more studies of diversity and distribution of speciose genera of small, cryptic, Caribbean reef fi shes and other Caribbean marine life are needed to determine if there are subdivisions of the major biogeographic provinces and, if so, what evolutionary mechanisms may be supporting them. Rocha et al. (2005) suggested that ecological speciation, in which natural selection in diff erent environmental conditions in adjacent locations may drive populations along separate evolutionary pathways, could help explain high levels of species diversity in marine fi shes in the absence of suffi cient physical barriers to account for that diversity. Colin (2010) suggested that faunal breaks in Elacatinus species may correlate well with observed ocean currents, and he proposed to further investigate known fi sh distributions and actual dispersal potential as estimated from satellite-tracked current drifters. For Starksia, future investigation must include more taxonomic and geographic coverage. Increased sampling will assuredly result in the recognition of new species and likely of new species complexes. Th e faunal breaks that separate members of the species complexes are unknown. In S. atlantica and S. lepicoelia, our specimens from Bahamas and Turks and Caicos represent the same species, and in S. sluiteri, specimens from Belize, Honduras, and Panama appear to be the same. Specimens in close proximity geographically thus tend to cluster into recognizable species. As better coverage is attained, it will be interesting to see if the same geographical boundaries characterize more than one of the species complexes or if the boundaries are diff erent for each. Likewise it will be interesting to compare geographic boundaries of Starksia species with faunal breaks in other reef fi shes such as Elacatinus. Future phylogenetic studies in which relationships among species and species complexes of Starksia and other groups are hypothesized should help shed light on patterns of speciation in small reef fi shes of the western Atlantic.
Because we do not know how much more investigation is required to obtain a reasonably complete picture of Starksia biodiversity and biogeography, the words of Winston Churchill included as an epigraph in this paper seem particularly appropriate. Th e study of Starksia must continue. (2-3)