The phylogenetic significance of colour patterns in marine teleost larvae

Ichthyologists, natural-history artists, and tropical-fish aquarists have described, illustrated, or photographed colour patterns in adult marine fishes for centuries, but colour patterns in marine fish larvae have largely been neglected. Yet the pelagic larval stages of many marine fishes exhibit subtle to striking, ephemeral patterns of chromatophores that warrant investigation into their potential taxonomic and phylogenetic significance. Colour patterns in larvae of over 200 species of marine teleosts, primarily from the western Caribbean, were examined from digital colour photographs, and their potential utility in elucidating evolutionary relationships at various taxonomic levels was assessed. Larvae of relatively few basal marine teleosts exhibit erythrophores, xanthophores, or iridophores (i.e. nonmelanistic chromatophores), but one or more of those types of chromatophores are visible in larvae of many basal marine neoteleosts and nearly all marine percomorphs. Whether or not the presence of nonmelanistic chromatophores in pelagic marine larvae diagnoses any major teleost taxonomic group cannot be determined based on the preliminary survey conducted, but there is a trend toward increased colour from elopomorphs to percomorphs. Within percomorphs, patterns of nonmelanistic chromatophores may help resolve or contribute evidence to existing hypotheses of relationships at multiple levels of classification. Mugilid and some beloniform larvae share a unique ontogenetic transformation of colour pattern that lends support to the hypothesis of a close relationship between them. Larvae of some tetraodontiforms and lophiiforms are strikingly similar in having the trunk enclosed in an inflated sac covered with xanthophores, a character that may help resolve the relationships of these enigmatic taxa. Colour patterns in percomorph larvae also appear to diagnose certain groups at the interfamilial, familial, intergeneric, and generic levels. Slight differences in generic colour patterns, including whether the pattern comprises xanthophores or erythrophores, often distinguish species. The homology, ontogeny, and possible functional significance of colour patterns in larvae are discussed. Considerably more investigation of larval colour patterns in marine teleosts is needed to assess fully their value in phylogenetic reconstruction.


INTRODUCTION
The pelagic larval stages of most marine fishes inhabit an evolutionary arena distinct from that of adults, and morphological specializations that presumably enhance survival in the planktonic realm have evolved in numerous teleost groups (Moser, 1981;Moser et al., 1984). In a few cases, marine fish larvae and adults are so different morphologically that they were initially classified as separate genera or families Johnson et al., 2009). Distinctive pigment patterns are among the transient features that characterize the pelagic larval phase of marine fishes , and Kendall, Ahlstrom & Moser (1984) included pigment patterns in their list of characters of early life-history stages commonly utilized in taxonomic and systematic studies of fishes. Although pigment patterns were noted by  to be among the most useful larval characters at specific and generic levels, only melanophores, not other types of chromatophores, were discussed. Chromatophores of poikilothermic vertebrates are dermal pigment units that comprise light-absorbing erythrophores (red/orange pigment), xanthophores (yellow pigment), and melanophores (dark brown/black pigment), as well as light-reflecting iridophores (structural colour, often silver/blue but many colours possible) -e.g. Bagnara & Hadley (1973), Grether, Kolluru & Nersissian (2004). Colour patterns in adult marine fishes are well documented, particularly those of tropical reef fishes, but pigment other than melanin in marine fish larvae has received little attention. This is not surprising because erythrophores and xanthophores fade upon conventional preservation (melanophores generally do not), and plankton samples are typically preserved soon after capture. Only when larvae have been observed, illustrated, or photographed prior to preservation has colour in marine teleost larvae been reported in the scientific literature (e.g. Brownell, 1979;Baldwin & Smith, 2003;Yasir & Qin, 2007;Baldwin et al., 2009a;Miller, 2009;Wittenrich, Baldwin & Turingan, 2010).
The transient colour patterns in marine fish larvae generally bear little resemblance to those of adults and may result from different ontogenetic (larval and adult) populations of chromatophores (Nakamura et al., 2010). The ontogeny of pigment patterns in marine fishes is poorly understood relative to that of many freshwater fishes, especially zebrafishes (Danio spp.), which have been studied extensively (e.g. Johnson et al., 1995;Parichy et al., 2000;Parichy, 2003Parichy, , 2006Kelsh, 2004;Budi, Patterson & Parichy, 2011). In Danio, the colour pattern of the recently hatched fish transforms directly into the adult colour pattern through incorporation of embryonic chromatophores and differentiation of new chromatophores from stem cells at metamorphosis (Parichy, 2003(Parichy, , 2006. There is no pelagic larval stage in Danio and most other freshwater fishes comparable to that in most marine fishes, and there is no accompanying distinctive pigment phase between the recently hatched and adult stages (Bagenal & Nellen, 1980;. Colour variation in this 'extra' pigment phase in marine fish larvae was the focus of this study. The presence of a specialized larval stage in certain freshwater fishes that are evolutionarily derived from marine fishes corroborates the distinction between early life stages of most marine and freshwater fishes. For example, larvae of basal freshwater percoids typically lack the head spination characteristic of pelagic larvae of basal marine percoids, but young Lates from Lake Tanganyika retain head spination that evolved in their marine, Indo-Pacific ancestors (Kinoshita & Tshibangu, 1997).
Colour patterns in the young of some freshwater fishes are highly conserved and thus of little potential phylogenetic value. For example, Quigley et al. (2004) noted that the young of several Danio species have virtually indistinguishable pigment patterns, and Kelsh (2004) noted the same for five Danio species and Tanichthys albonubes. In contrast, Baldwin & Smith (2003) and Baldwin et al. (2009aBaldwin et al. ( , 2011 highlighted the utility of patterns of erythrophores and xanthophores in species identification of larval Gobiidae and Apogonidae, and Baldwin et al. (2011) suggested that chromatophore patterns may be of value in resolving the generic classification of western Atlantic Apogonidae. The potential utility of larval colour patterns at higher taxonomic levels has not been properly investigated.  noted that pigment (melanophore) patterns in fish larvae are of limited use in systematic studies in part because convergence has resulted in the occurrence of strikingly similar patterns in unrelated groups. Convergent evolution can be detected if pigment characters are examined in a phylogenetic context. A phylogenetic tree that includes clades composed primarily of marine teleosts was used herein to examine the distribution of nonmelanistic chromatophores in larvae among major groups of marine teleosts. The exclusion of clades of freshwater fishes renders the tree but a partial view of teleost phylogeny, but in this first attempt to provide comparative information on colour patterns among marine teleost larvae, it is desirable to contemplate the results from both broad and more focused phylogenetic perspectives. Numerous recent hypotheses of phylogenetic relationships among various groups of teleosts were utilized for comparisons at lower taxonomic levels. The purposes of this paper were to describe the distribution of nonmelanistic chromatophores among a broad spectrum of marine teleost larvae and comment on the potential of selected chromatophore patterns to inform phylogeny.

MATERIAL AND METHODS
This study was based largely on colour photographs of fish larvae collected off Belize, Central America. Larvae were collected in a plankton net of 505 mm mesh fitted onto a 0.5 ¥ 1 m rectangular frame and deployed from a dock at Carrie Bow Cay (16°48.5′N, 88°05′W). Specimens were submerged in a photo tank, photographed with a Nikon D1 or Fuji FinePix S3 digital camera, and then tissue sampled for DNA analysis prior to preservation. Species identification of larvae was accomplished by matching cytochrome oxidase-c subunit I (COI) sequences (DNA barcodes) COLOUR IN MARINE TELEOST LARVAE 497 of larvae to those of known adults . To date this protocol has resulted in the identification of larvae of approximately 170 Caribbean fish species. A few larvae had no species-level matches in the Smithsonian DNA database or the Barcode of Life Database (BOLD; http://www.boldsystems.org/views/ login.php) and are identified only to genus in the figure legends. Taxonomic coverage was increased by examining images of marine fish larvae from specimens collected off the east coast of South Africa, Florida Straits, Hawaii, and Cozumel, as well as several Indo-Pacific species reared from aquarium specimens. Connell's (2007) website of early lifehistory stages of South African marine fishes is impressive in scope, and although few of his images are reproduced here, numerous references to his site and uniform resource locators (URLs) to specific pages are provided herein. All photo editing was carried out by the author. Images of most larvae were cut from their original photographic backgrounds and placed on a uniform background using the MaskPro 4 plug-in for Adobe Photoshop with the aid of a Wacom table and stylus. In many cases, the transparent fins of larvae were difficult to see, and shapes of fins may be approximations. Photo credits are provided in figure legends, and affiliations for contributors other been used interchangeably, as have 'orange pigment ' (or 'red pigment') and 'erythrophores'. The presence or absence of yellow and orange pigment was equated with the presence or absence of xanthophores and erythrophores, respectively, and the presence or absence of light-reflecting pigment was equated with the presence or absence of iridophores. Assessments of types of nonmelanistic chromatophores were based solely on examination of fresh specimens and colour photographs, not histological examination.
The classification of Wiley & Johnson (2009) was selected for use in this study. This classification was chosen over other more commonly used fish classifications because it is a Linnaean classification based on monophyletic groups. It was essential in this work to examine the distribution of nonmelanistic chromatophores in fish larvae among major teleost groups from an evolutionary perspective, and constructing a teleost phylogeny from the Wiley & Johnson (2009) synapomorphy-based classification was easily accomplished ( Fig. 1).

DISTRIBUTION OF NONMELANISTIC CHROMATOPHORES
IN BASAL MARINE TELEOSTS AND NEOTELEOSTS (FIG. 1) Larvae of most basal teleosts -elopiforms, albuliforms, anguilliforms, and clupeiforms -do not exhibit xanthophores, erythrophores, or iridophores (Figs 2,3). Exceptions occur within the anguilliform families Ophichthidae, Muraenidae, Congridae, and Nettastomatidae. Miller, D'Avella & Tsukamoto (2010: figs 1, 2) described xanthophores along the bodies of two unidentified ophichthid leptocephalus larvae videotaped swimming at night off Hawaii (Fig. 4A, B). An image of the ophichthid Neenchelys (Miller, 2009: fig. 57A) has yellow pigment on the snout, anterior portion of the oesophagus, and on the gut swellings (Fig. 5A). Another ophichthid leptocephalus, Myrichthys breviceps, captured off Belize bears similar conspicuous xanthophores on the gut swellings (Fig. 4C). Tawa et al. (2012: fig. 2) published a colour image of a muraenid leptocephalus (Strophidon ui) that exhibits xanthophores in front of and behind the eye, and Miller (2009) provided images (Fig. 5B, E, F herein) of muraenid leptocephali with similar yellow pigment adjacent to the eye (and in one case also scattered on the head). Miller (2009) noted the presence of yellow pigment on the dorsal surface of the eye in some congrids and ophichthids (Fig. 5D) and yellow pigment on the snout and anterior portion of the oesophagus in the nettastomatid Saurenchelys (Fig. 5C). Identification of more anguilliform larvae is needed to determine the taxonomic distribution of 498 C. C. BALDWIN Figure 1. Teleost phylogeny based on the classification of Wiley & Johnson (2009). Only marine taxa for which colour in larvae was examined are included. The number of squares associated with each teleost order is equivalent to the number of species examined. Open squares indicate the absence of xanthophores, erythrophores, and iridophores. Yellow squares indicate the presence of one or more of these types of chromatophores. The shaded rectangle denotes Johnson & Patterson's (1993)  500 C. C. BALDWIN xanthophores, but the presence of yellow pigment on gut swellings in ophichthids, on the snout and anterior oesophagus in ophichthids and nettastomatids, in front of and behind the eye in muraenids, and dorsal to the eye in congrids and ophichthids might represent diagnostic patterns and therefore warrant additional study. Most leptocephali collected off Belize lack yellow pigment, yet many are members of families discussed above that have it. Anguilliform leptocephali from Belize that lack yellow pigment (Fig. 2) include Gymnothorax moringa (Muraenidae), Moringua edwardsi (Moringuidae), Chilorhinus seunsoni (Chlopsidae), and Ahlia egmontis, Aprognathodon platyventris, Myrophis punctatus, and Myrophis platyrhynchus (Ophichthidae). Based on the absence of xanthophores in larval albuliforms and elopiforms, it is reasonable to assume that their absence is ancestral for anguilliforms. The absence of yellow pigment in leptocephali of Moringua edwardsi and Synaphobranchidae (Miller, 2009) provides corroborative evidence based on the basal positions of Moringuidae and Synaphobranchidae in the molecular anguilliform phylogeny of Tang & Fielitz (2012). Anguilliform taxa that exhibit yellow pigment in the leptocephalus stage -some Congridae, Nettastomatidae, Ophichthidae, and muraenine Muraenidae -occupy more distal phylogenetic positions in the order (Tang & Fielitz, 2012), but they do not constitute a monophyletic assemblage. It seems likely that xanthophores in larvae evolved independently within the various families of Anguilliformes that exhibit them.
Little information is available on the presence or absence of nonmelanistic chromatophores in larvae of basal marine neoteleosts (Fig. 1). Recently hatched larvae of one phosichthyid stomiatiform from off South Africa lack erythrophores and xanthophores, whereas a preflexion larva of a melanostomiatid has yellow pigment on the head and body (Connell, 2007; see links to images in Appendix). Two aulopiform families (Synodontidae and Giganturidae) also have larvae that lack orange or yellow coloration (Figs 6A-C, 7A), but an unidentified aulopiform larva has vibrant erythrophores and xanthophores on enlarged pectoral fins as well as on the dorsal and caudal fins (Fig. 7B). A preflexion myctophiform larva from off South Africa has yellow pigment on the head and body, but yellow pigment is not as evident in a larger preflexion larva (Connell, 2007;Appendix). A postflexion larval myctophiform from the Florida Straits lacks erythrophores and xanthophores (Fig. 7C). In gadiforms, preflexion larvae of an unidentified gadid from South Africa have numerous xanthophores on the head and body (Connell, 2007;Appendix), whereas the bregmacerotid Bremaceros from Belize lacks orange and yellow pigment (Fig. 6D). Both lampridiform larvae examined (a trachipterid and Lampris) have nearly the entire body and some fins covered with erythrophores (Fig. 8). The single stephanoberyciform larva examined, the bizarre 'mirapinnid' larva of the whalefish family Cetomimidae (Johnson et al., 2009), has yellow pigment on the body and fins (Fig. 9), and the single zeiform examined, Zeus faber (Connell, 2007;Appendix), also has a small bit of yellow pigment in both pre-and postflexion stages that is mixed with and largely occluded by melanophores. All preflexion and postflexion beryciform larvae have erythrophores, xanthophores, iridophores, or some combination of those chromatophores (Fig. 10).
To summarize the comparative information for basal marine teleosts and neoteleosts, several basal neoteleost orders are similar to anguilliforms in having some larvae that exhibit nonmelanistic chromatophores and others that do not. Xanthophores are the only types of chromatophores observed in larval anguilliforms and in the single larval stomiatiform examined that has nonmelanistic chromatophores. Xanthophores are lacking in larval elopiforms, albuliforms, and the few marine clupeiforms examined. Xanthophores are not lacking in all otomorphs, however, as they have been documented in freshwater Cyprinidae (e.g. Johnson et al., 1995;Parichy et al., 2000;Parichy, 2003). Erythrophores are first observed phylogenetically in marine fish larvae in aulopiforms and are prominent in lampridiforms and some beryciforms. Possibly erythrophores in larvae are phylogenetically significant at the level of Eurypterygia (Aulopiformes and above, Fig. 1). Iridophores first appear phylogenetically in larval beryciforms and could provide corroborative evidence for Johnson & Patterson's (1993) Euacanthopterygii (Beryciformes and above, shaded rectangle in Fig. 1). Iridophores are lacking in larvae of the marine elopomorphs, otomorphs, and other basal neoteleosts examined; however, they are present in some recently hatched freshwater Otomorpha (e.g. Danio -Johnson et al., 1995;Parichy et al., 2000;Parichy, 2003). As discussed below, an ontogenetic transition involving iridophores in atheriniforms, beloniforms, gasterosteiforms, and mugiliforms, which are members of the percomorph series Smegmamorpharia, may help diagnose that group. Within percomorphs (Mugiliformes and above, Fig. 1), erythrophores, xanthophores, iridophores, or some combination of them are present in larvae of every group and in almost every species examined, and the remainder of this paper is devoted to descriptions of percomorph colour patterns and discussions of their potential phylogenetic significance at various taxonomic levels.

PERCOMORPHACEA -SMEGMAMORPHARIA
Johnson & Patterson's (1993) Smegmamorpha comprise seven orders of percomorph fishes, but the  504 C. C. BALDWIN monophyly of the group has been questioned. For example, Springer & Orrell (2004) did not find support for it based on extensive analysis of the gill-arch musculature and skeleton. Smegmamorph orders for which larvae were examined in this study are Atheriniformes, Beloniformes, Gasterosteiformes, and Mugiliformes. Larval mugilids and exocoetids examined share a striking ontogenetic transition in colour pattern. Small larvae are covered with pale orange/yellow pigment and melanophores but also bear some conspicuous iridophores (Fig. 11A, B). In larger larval specimens, the orange colour is no longer present, and the fishes are silvery and covered almost entirely with iridophores ( Fig. 11C, D). This ontogenetic transition was not observed in any nonsmegmamorph fishes and would appear to provide corroborative evidence for a close relationship between Mugiliformes and Atherinomorpha, which comprises Atheriniformes and Beloniformes. Stiassny (1990Stiassny ( , 1993 proposed a sister-group relationship between mugiliforms and atherinomorphs despite strong evidence also suggesting that mugilids are closely related to perciforms. This same colour transition occurs in hemirhamphids and belonids but apparently at a later stage of development. Larvae of Platybelone and Hemirhamphus are covered with erythrophores and xanthophores, respectively, and bear scattered iridophores (Fig. 12A, B, F). A juvenile Hemirhamphus still bears xanthophores, but the abdominal region is silvery (Fig. 12C), and adults of both genera are entirely silver. Larval gasterosteiforms also have yellow/orange chromatophores and melanophores covering most of the body as in young mugilids and beloniforms (Fig. 12D, E), and although they never undergo the transition to a silvery body, at least some gasterosteiform larvae have silvery pigment on the abdomen (Fig. 12E, G). Other gasterosteiform larvae known only from early preflexion stages -Fistularia, Aulostomus, Aeoliscus -also exhibit numerous xanthophores mixed with melanophores (Connell, 2007; Appendix), but later larval stages are needed to determine whether or not they develop iridophores. Some larval beloniforms may have a different larval trajectory in that they appear to lack the early yellow/orange stage. For example, a 5.6 mm Notochord Length (NL) larva of Oxyporhamphus micropterus from off South Africa is mostly pale, has melanophores and iridophores along the entire dorsal margin of the body and along a portion of the ventral midline, and has a silvery/blue gut (Connell, 2007;Appendix). Possibly there is a small bit of yellow pigment on the oesophagus and gut, but it was difficult to determine if this is dermal pigment or part of the gut contents. It is also possible that the yellow/ orange stage develops after notochord flexion, as all mugilid, beloniform, and gasterosteiform larvae in which that colour phase was observed had undergone flexion. A 13-mm Standard Length (SL) Hirundichthys has only melanophores and iridophores (Fig. 13A), but younger larvae are unknown. Larval atherinids have few or no orange or yellow chromatophores (smallest specimen examined 7.0 mm SL - Fig. 13B) and have a dense covering of iridophores on the gut. Note the striking similarity between a 14-mm  SL atherinid larva and an unidentified beloniform larva in general appearance and the presence of iridophores on the gut that reflect bright blue (Fig. 13C, D).

Acanthuriformes
Acanthuriformes are represented among the larvae examined by two species of Acanthurus surgeonfishes from Belize (Fig. 14) and an earlier stage of an unidentified species of Acanthurus from the Florida Straits (Fig. 15). The specialized, pelagic 'acronurus' larval stage (e.g.  is mostly transparent with a band of silver-or blue-reflecting iridophores from the dorsum to the pelvis that encompasses the orbit and gut. There is little if any orange or yellow pigment in the acronurus stage, but an image of preflexion larvae of Acanthurus mata (Connell, 2007;Appendix) shows xanthophores on the upper jaw and an internal streak of xanthophores beneath the anterior section of the notochord. Relationships of acanthuriforms are unclear, but a close relationship with tetraodontiforms has been proposed (e.g. Tyler, 1980;Rosen, 1984). Tetraodontiform larvae (see 'Tetraodontiformes' below) typically have iridophores on the gut and sometimes over much of the body, but they are otherwise very different from acanthuriforms in having much more colour (xanthophores or bronze iridophores), melanophores (sometimes in striking patterns), or both.

Blenniiformes
Blenniiformes are represented in the Belize material by larval chaenopsids and labrisomids. Larvae of both families have an elongate body and share the following chromatophore pattern: melanophores on the ventral midline associated with the anal-fin base and internally along the dorsal margin of the vertebral column, orange or yellow pigment on the temporal region of the head, internal orange pigment along the vertebral column, and orange chromatophores at the base of the caudal fin (Fig. 16). Colour patterns vary little among some genera (e.g. Acanthemblemaria of the Chaenopsidae, Labrisomus of the Labrisomidae - Fig. 16A-D), but the single larval specimen of Paraclinus (Labrisomidae) differs in having prominent orange chromatophores mixed with the melanophores along the anal-fin base (Fig. 16E). Additional Paraclinus species are needed to determine whether or not this is a generic pattern. The chromatophore pattern on the body could not be interpreted well in an in situ image of a larval fish from Hawaii identified by G. D. Johnson (pers. comm.) as a blenniid (image not reproduced here), but the fish is unlike other blennioids in having enlarged pectoral fins with xanthophores on the membranes between fin elements.

Caproiformes
Caproiformes are represented in the images examined only by preflexion Antigonia rubescens from South Africa (Connell, 2007;Appendix). Recently  hatched larvae have xanthophores on the body, but no information is available for larger specimens.

Carangiformes
Carangiformes are not common in larval-fish collections off Belize, but an 8.0-mm SL Trachinotus (Carangidae) has most of the body covered with orange chromatophores and silver-reflecting iridophores (Fig. 17). A colour image of a young carangid, Selene (Fig. 18C), shows yellow/orange chromatophores on a somewhat silvery body as well as on most fins. A 5.1-mm preflexion larva of the coryphaenid, Coryphaena hippurus, has xanthophores mixed with melanophores from the tip of the snout posteriorly to the base of the incipient caudal fin (Fig. 18A). A postflexion Coryphaena hippurus larva (Fig. 18B) and a reared postflexion rachycentrid, Rachycentron canadum (D. Benetti, pers. comm.), have some yellow pigment mixed with melanophores on the body, but the solid yellow ground coloration present in preflexion Coryphaena is absent. Colour patterns are not known for other carangiforms (Nematistiidae, Echeneididae), and no conclusions can be drawn regarding the phylogenetic significance of colour for the order or for families, genera, and species within. The bars of pigment in postflexion Coryphaena (Fig. 18B) resemble those of Selene in having some yellowish coloration mixed with melanophores, and this character could be significant at the ordinal level. Phylogenetic analysis of both mitochondrial (Miya, Satoh & Nishida, 2005) and nuclear (Little, Lougheed & Moyes, 2010) data suggested a close relationship between carangiforms and pleuronectiforms, but there is nothing obvious in the colour patterns of larvae to support this (see 'Pleuronectiformes' below). Nuclear DNA data have also suggested a relationship between carangids and istiophorids (Little et al., 2010), which is discussed under 'Scombriformes' below.

Gobiesociformes
Gobiesociformes comprise the Gobiesocidae, Callionymidae, and Draconettidae. Colour information in larvae is available for several callionymids and one gobiesocid. Wittenrich et al. (2010) described morphological development of laboratory-reared Synchiropus splendidus and noted a complete colour transition from yellow in early-stage larvae to nearly solid orange at 8 days posthatching (Fig. 19A, B). Connell (2007) provided a colour image of a 1.8-mm NL larva of Callionymus sp. from South Africa (Fig. 19C) and one of a 1.6-mm NL larva of Draculo (Appendix) that are nearly identical to the similar-sized (2 days posthatching) larva of Syn. splendidus (Fig. 19A) in having the head and body covered in prominent xanthophores except for the posterior fin fold. The colour pattern in larger larvae of the South African Callionymus is unknown, but postflexion larvae of Callionymus bairdi from Belize ( Fig. 19D) are nearly solid orange like Synchiropus. Additional material is needed to determine whether the observed yellow-toorange colour transition is diagnostic of callionymids or if it is also present in other gobiesociform families. A 4.0-mm SL larva of the gobiesocid Acyrtops from Belize has yellow/orange coloration on the head and scattered over most of the body (Fig. 19E), possibly a very similar pattern as in the callionymids but with the erythrophores contracted. Larger and smaller    specimens of these taxa are not available. The only other percomorph larvae examined that were as extensively covered by erythrophores as postflexion callionymid larvae are some apogonids and the goby Priolepis (see 'Perciformes' and 'Gobiiformes' below). Apogon aurolineatus is particularly similar to Callionymus bairdi in having a bright orange body with yellow first dorsal and pelvic fins. The presence of similar chromatophore patterns in these presumably distantly related taxa is best interpreted as convergence.
Pigment in larvae does not appear to provide any evidence for a close relationship between gobiesociforms and blenniiforms as proposed by Miya et al. (2003) and Springer & Orrell (2004). The extensive coverage of yellow/orange pigment in the gobiesociforms examined is very different from the more restrictive colour pattern observed in blenniiforms; however, information on colour is needed for larvae of more taxa, especially those blenniiforms hypothesized by Stepien et al. (1997) to be more basal members of the group than labrisomids and chaenopsids.

Gobiiformes
Gobiiformes examined include members of the Gobiidae, Eleotrididae, and Microdesmidae. Larvae of all exhibit erythrophores or xanthophores, or both, and some have iridophores, but there are no obvious diagnostic patterns for the order, major clades within the order, or the highly diverse Gobiidae. Preliminary comparative data suggest that colour patterns may be most informative at intergeneric, generic, and species levels.
Larvae of the microdesmid genera Microdesmus and Cerdale have nearly identical patterns of orange/ yellow chromatophores ( Fig. 20A-C). Hoese (1984) united Microdesmus, Cerdale, and three other microdesmid genera with Ptereleotris and its allies in an expanded family Microdesmidae, but Smith & Thacker (2000) recognized the Microdesmidae and Ptereleotrididae as separate families. Thacker's (2003Thacker's ( , 2009) molecular phylogenetic analyses did not recover a monophyletic group comprising microdesmids and ptereleotridids, but the molecular analysis of Thacker & Roje (2011) did, with ptereleotridids paraphyletic with respect to microdesmids. Superficially, the larval Ptereleotris ( Fig. 20D) is quite different from larval Microdesmus and Cerdale in that it does not have an elongate body with an obvious series of erythrophores internally along the vertebral column and above the gut. The available colour images of Ptereleotris, however, are of specimens that are mostly opaque (vs. transparent), and although internal erythrophores are present above the gut and along the lateral midline posteriorly, it is not possible to determine the extent of these series. Colour patterns in the three genera are otherwise quite similar, with all having a series of erythrophores along the ventral midline (except directly beneath the swim bladder), orange or yellow chromatophores in a series along the dorsal midline posteriorly, and at least some orange pigment externally on the central portion of the caudal peduncle. This pattern was not observed in other gobiiforms, and may provide corroborative evidence for a monophyletic group comprising the two families. The hypothesis that Coryphopterus gobies are more closely related to Microdesmus and Cerdale than microdesmids are to Ptereleotris (Thacker, 2003) would not appear to be supported by colour patterns in larvae. Coryphopterus gobies have a distinctive pattern of erythrophores on the trunk and lack the chromatophore pattern described above for microdesmids (see last paragraph of Gobiiformes section, below). Coryphopterus shares with microdesmids the presence of erythrophores mixed with melanophores along the anal-fin base, but this pigment is also present in Ptereleotris and even the distantly related eleotridids. Miller (1998) synonymized the eleotridid genera Eleotris and Erotelis, and colour patterns in larvae of the two genera are strikingly similar and distinctive among gobiiforms (Fig. 20E, F). Thacker's (2003) molecular analysis did not recover the two as a monophyletic group; rather, they are successive sister groups to all other gobiiforms excluding the Odontobutidae. The more comprehensive molecular phylogeny of Thacker (2009) corroborates a common ancestry for the two genera, with Erotelis embedded within the Eleotris clade. In addition to having similar patterns of erythrophores/xanthophores, Eleo- tris and Erotelis are the only two gobiiform larvae examined that have a swath of blue-reflecting iridophores at the base of the caudal fin.
Morphological and molecular data support a close relationship between Ctenogobius and Gnatholepis (Harrison, 1989;Thacker, 2003Thacker, , 2009. As noted by Baldwin & Smith (2003), larval Ctenogobius saepepallens and Gnatholepis thompsoni are similar in having a prominent, narrow bar of orange pigment on the body just posterior to the anal-fin base that is lacking in other gobiids (Fig. 21A, B). Baldwin & Smith (2003) also noted that the orientation of this bar is slightly different in the two species and that it may not be homologous. Gobionellus oceanicus has a wide bar of pale orange pigment on the caudal peduncle that, because of the long anal fin in this species, is located at the posterior base of the anal fin as the narrow bar is in Ctenogobius and Gnatholepis (Fig. 21C). The phylogenetic significance within the Gobiidae of orange pigment at the posterior base of the anal fin extending dorsally from the ventral midline is unclear. Observations of colour patterns in larvae of more genera of Gobionellus-group gobies (Birdsong, Murdy & Pezold, 1988) is needed, including Evorthodus, which Thacker (2003Thacker ( , 2009 proposed as the sister group of Ctenogobius. Postlarval Ctenogobius boleosoma (Wyanski & Targett, 2000) shares with Ct. saepepallens an erythrophore at the tip of the lower jaw, a vertical reddish-orange streak between the thoracic and abdominal regions, and the orange bar posterior to the anal-fin base. Identification of other Ctenogobius larvae is needed, but the combination of these pigment characters may be diagnostic of the genus.
Nes and Psilotris are seven-spine gobies of the 'Gobiosoma group' (Birdsong et al., 1988) that share with Varicus, Chriolepis, and Gobulus the absence of head pores (Böhlke & Robins, 1968). Rüber, Van Tassell & Zardoya (2003) recovered this group as monophyletic, minus Varicus, which was not included in their analysis. As noted by Smith & Baldwin (1999), larvae of Nes and Psilotris are so similar that at first these authors did not recognize them as distinct taxa. The chromatophore pattern is nearly identical in the two genera and distinctive among gobiids examined (Fig. 21D, E). Identification of additional goby larvae, including Gobulus and Chriolepis, is needed to determine whether the pattern is unique to Nes and Psilotris or perhaps to a larger group. Rüber et al. (2003) hypothesized that the relationships of Nes are as follows: (Nes(Gobulus(Chriolepis + Psilotris))).
At the generic level, colour patterns in larvae are useful in diagnosing Bathygobius and Coryphopterus. Bathygobius larvae have orange/yellow chromatophores on the dorsal and ventral portions of the trunk (in association with melanophores) that extend toward, and often meet at, the lateral midline ( Fig. 22A-C). Although the colour (yellow or orange)  and density of pigment differ among species, the net effect is distinctive (Baldwin & Smith, 2003;Tornabene et al., 2010). All known Coryphopterus larvae have diagonal bars of orange pigment on the trunk, the height of the bars and how far the series extends anteriorly varying among species (Baldwin & Smith, 2003;Fig. 22E-H). Colour patterns in larvae of genera hypothesized to be closest to Bathygobius (Glossogobius, Istigobius, and Callogobius; Tornabene & Pezold, 2011, and references therein;Thacker, 2009) and Coryphopterus (Lophogobius; Thacker, 2003Thacker, , 2009) are unknown. The unique generic patterns described above could diagnose larger groups. Monophyly of Thacker's (2003) clade IIIA, which includes Bathygobius and Priolepis, would not appear to be supported by larval colour patterns. Larval Priolepis (Fig. 22D) is distinctively orange.

Labriformes
Labriformes comprise the Cichlidae, Embiotocidae, Labridae, Odacidae, Pomacentridae, and Scaridae. The monophyly of the group is questionable (Wiley & Johnson, 2009, and references therein), and molecular data suggest that scarids are embedded within the Labridae (Westneat & Alfaro, 2005;Choat et al., 2012). Marine labriforms for which colour patterns were assessed for larvae are pomacentrids, labrids, and scarids. All exhibit nonmelanistic chromatophores, usually erythrophores, but the patterns are not diagnostic of the order. Based on existing comparative material, colour patterns appear useful in diagnosing some families and genera. All scarid larvae examined are united in having a linear series of erythrophores along the ventral midline of the trunk from beneath the operculum to the anus, a linear series of erythrophores along the anal-fin base, a roughly linear series of erythrophores above the anal-fin base that curves dorsally on the caudal peduncle where it is continuous with erythrophores on the lateral midline, and erythrophores on the caudal fin (Fig. 23). This pattern is present in Cryptotomus, Scarus, and Sparisoma, which differ from one another in (1) the extent of orange coloration on the caudal fin (with streaks of bright orange chromatophores on the ventral lobe in Cryptotomus, Fig. 23A, vs. orange pigment more scattered or paler in the other genera); (2) organization of erythrophores and melanophores above the anal-fin base (somewhat   23D) and Sparisoma radians (Fig. 23E) is unique among labriforms and may indicate that these species are more closely related to one another than either is to Sparisoma chrysopterum (Fig. 21C).
Larval pomacentrids examined have silvery iridophores on the gut and at least some erythrophores or xanthophores on the trunk and fins (Fig. 24). Colour patterns of larvae of the four genera examined are distinctive. Stegastes larvae have a swathe of orange pigment on the trunk from just behind the eye to the anterior-most part of the caudal peduncle, where the swath ends abruptly in a near vertical line ( Fig. 24A-C). Differences among Stegastes species include the presence or absence of erythrophores on the pectoral fin and along the spinous dorsal-, pelvic-, and anal-fin bases. Colour information was available for only one species of Chromis (Fig. 24D) and one of Abudefduf (Fig. 24E), but patterns in these genera are distinct from one another and from Stegastes. Chromis lacks colour on the fins and has erythrophores restricted to the posterior portion of the trunk and caudal peduncle, whereas Abudefduf is distinctive in having xanthophores on most of the trunk (mixed with melanophores) and conspicuously yellow first dorsal and pelvic fins. Like Stegastes, pigment on the trunk in Abudefduf ends abruptly on the anterior portion of the caudal  peduncle. Reared larvae of Amblyglyphidodon ternatensis (Fig. 24F) are similar to Abudefduf in having prominent yellow pigment on the dorsal and pelvic fins. They are distinctive in having yellow pigment covering the dorsal portion of the head, including the dorsal portion of the orbit, and also the anterior portion of the anal fin. The molecular phylogeny of Cooper, Smith & Westneat (2009) suggests that Stegastes, Chromis, Abudefduf, and Amblyglyphidodon are members of four distinct evolutionary assemblages (Stegastinae, Chrominae, Abudefdufinae, Pomacentrinae, respectively), a hypothesis that is not contradicted by colour patterns in larvae. Larval Abudefduf and Amblyglyphidodon are the most similar of the four in terms of coloration, and Abudefdufinae and Pomacentrinae are sister groups according to Cooper et al. (2009). Acquisition of colour information for larvae of additional pomacentrids is needed to determine whether or not the colour patterns identified herein characterize the subfamilies (or some subset of them) and whether aspects of the colour pattern in Abudefduf and Amblyglyphidodon represent a synapomorphy of Abudefdufinae and Pomacentrinae.  (2005) did not include Ha. maculipinna, but it suggests that Halichoeres is paraphyletic without the inclusion of numerous other genera, including Thalassoma and Anampses. The presence of erythrophores in the three genera on the upper jaw and gut, combined with the presence of distinct dark markings on the dorsal and anal fins, constitute a unique pattern within labriforms that may support a close relationship among these genera.
Xyrichtys larvae are elongate and consistently have erythrophores on both jaws and behind the eye. The rest of the body is pale except for a blotch of colour on the caudal peduncle ( Fig. 26A-C). In Xyrichtys novacula (Fig. 26A) this blotch is always yellow, whereas in Xyrichtys splendens (Fig. 26B) and Xyrichtys martinicensis (Fig. 26C) it is orange. Differences in size and shape of the orange blotch distinguish these two species. There is little if any intraspecific variation in chromatophore pattern among the individuals of each Xyrichtys species examined.
Doratonotus and Lachnolaimus are monotypic genera, and their larvae are clearly distinct from one another and other labrids (Fig. 26D, E). Doratonotus larvae have internal erythrophores along the vertebral column and along the myosepta of the posterior third of the trunk and, as noted above, lack erythrophores on the jaws. Lachnolaimus larvae are very different from those of other labrids in having almost the entire body covered with orange/yellow chromatophores -mostly xanthophores anteriorly and erythrophores posteriorly.

Lophiiformes
Lophiiformes comprise more than a dozen percomorph families, but larvae of only one species from Belize, Antennarius pauciradiatus (Antennariidae), have been identified (Fig. 27A). Images of colour patterns in several unidentified lophiiform larvae from the Florida Straits were used for comparative purposes (Fig. 27B-D). In Ant. pauciradiatus the distended skin forming the characteristic lophiiform 'balloon' around the head and body is lightly covered with erythrophores in the lower jaw and gular regions. The three unidentified lophiiform larvae have different colour patterns from Ant. pauciradiatus and one another, two of them exhibiting xanthophores (Fig. 27C, D) and one of them seemingly lacking xanthophores/erythrophores and exhibiting only blue iridophores (Fig. 27B). Lophiiforms, which form part of Rosen & Patterson's (1969) and Patterson & Rosen's (1989) Paracathopterygii, were hypothesized to be percomorph fishes in the molecular phylogeny of Miya et al. (2003). The 'paracanthopts' examined for colour patterns in larvae are gadids (Fig. 6D), lophiiforms, and ophidiiforms (see below). With the limited amount of material available, the only observation relevant here is that unlike the gadid Bregmaceros, which lacks xanthophores, erythrophores, and iridophores, all lophiiforms (and ophidiiforms) examined have one or more of those types of chromatophores. Like numerous other acanthomorph orders, lophiiforms are currently considered incertae sedis within the Percomorphacea (Fig. 1), but the hypothesis of Holcroft & Wiley (2008) that lophiiforms are closely related to tetraodontiforms is discussed below (see Tetraodontiformes).

Ophidiiformes
Ophidiiformes were also hypothesized to be percomorphs rather than paracanthopterygians by Miya et al. (2003Miya et al. ( , 2005. Wiley & Johnson (2009) noted that no convincing evidence for the monophyly of the order (ophidioids + bythitoids) exists. Colour information in larvae is known for two ophidioids from Belize, Carapus bermudensis (Carapidae) and   526 C. C. BALDWIN orange/yellow chromatophore patterns (Fig. 28). In situ images of two Hawaiian ophidiids that appear to be larval Brotulataenia and Lampogrammus (Fig. 29) show strikingly beautiful larvae with numerous xanthophores on the body and fins, as well as erythrophores on the fins in the former. Erythrophores in Parophidion are confined to the head and trunk and do not extend onto the median fins. No conclusions about the potential phylogenetic significance of colour patterns in larval ophidioids can be drawn at this time, and colour patterns, if any, in young bythitoids are unknown.

Perciformes
Perciformes remain a conglomerate of families and incertae sedis genera that are not united by synapomorphies (Wiley & Johnson, 2009), and larvae of the group exhibit a diverse array of chromatophore patterns. At the interfamilial level, the Chaetodontidae and Pomacanthidae have larvae that are covered with iridophores -silver/bronze in Holacanthus, Chaetodon, and Pomacanthus arcuatus, silver/blue in Pomacanthus paru (Fig. 30). Although not as conspicuous in Holacanthus (Fig. 30A) and Pomac. paru (Fig. 30D) as in Chaetodon and Pomac. arcuatus,   larvae of all specimens of all three genera have xanthophores, erythrophores, or both, minimally on the jaws, snout, base of dorsal fin, and caudal peduncle. It does not appear that chaetodontids and pomacanthids resemble smegmamorphs in transitioning from an early stage featuring predominantly xanthophores/erythrophores to a later stage featuring primarily iridophores, as preflexion Chaetodon marleyi and Pomacanthus rhomboides have few if any xanthophores or erythrophores (Fig. 31). The preflexion Chaetodon and Pomacanthus are quite similar in having a broad band of iridescent silvery blue iridophores around the trunk. Chaetodontids and pomacanthids have been considered to be so closely related in the past that until Burgess ' (1974) publication, pomacanthids were classified as a subfamily of the Chaetodontidae. The families were still considered closely related by Tyler et al. (1989), who delineated a monophyletic group comprising chaetodontids, pomacanthids, and Drepane (Drepaneidae) based on configuration of the ethmoid bone. Colour in larval Drepane is unknown, but the presence in larvae of that genus of a colour pattern comprised largely of iridophores with xanthophores/ erythrophores positioned as described above for chaetodontids and pomacanthids could provide corroborative evidence for the monophyly of the group. Recent molecular studies have challenged the hypothesis of a close relationship between chaetodontids and pomacanthids. Using mitochondrial genes, Bellwood, van Herwerden & Konow (2004) hypothesized that chaetodontids are more closely related to scatophagids than to pomacentrids, and Holcroft & Wiley (2008) hypothesized that chaetodontids and scatophagids are part of a large group also comprising acanthuroids, lophiiforms, and tetraodontiforms but not pomacanthids. Colour information for larval scatophagids is not available, but colour in larvae would not appear to support a closer relationship between chaetodontids and acanthurids (Figs 14,15) than between chaetodontids and pomacanthids (Fig. 30). Larval chaetodontids bear little resemblance to larval lophiiforms (Fig. 27), but larval tetraodontid tetraodontiforms examined are similar to chaetodontids and pomacanthids in having the body covered with bronze/gold iridophores and the abdominal region silvery (see 'Tetraodontiformes' below).
At the familial level, larval Apogonidae have nearly the entire body covered with erythrophores. With the exception of Priolepis gobies and Callionymus dragonettes, no other larvae examined are covered so completely with erythrophores, and this feature may be synapomorphic for the family or some subset of it. Colour patterns in western Atlantic larval Apogon, Astrapogon, and Phaeoptyx (Fig. 32) have  been described (Baldwin et al., 2009a;. Patterns of melanophores, in conjunction with orange/yellow chromatophores, diagnose the three genera. Within Apogon, colour patterns delineate several species groups that may be meaningful in the generic classification of the group (Baldwin et al., 2011).
Larval Gerreidae examined lack erythrophores and xanthophores on most of the body, but some pale yellow pigment is usually present over the gut and swimbladder and sometimes on the dorsal portion of the head (Fig. 33). Determining whether this pattern, combined with the distinctive arrangement of melanophores at the bases of the median fins, characterizes Eucinostomus (larvae of four species identified) or the entire family requires additional material.
Late-stage larvae of the Lutjanidae are easily recognized by the cap of silver iridophores over the gut and opercular region, a mostly clear trunk, and xanthophores associated with the dorsal fin (or its base) and sometimes caudal peduncle (Fig. 34). The pattern of xanthophores appears species specific within Lutjanus. Presence of the same general pattern of iridophores and xanthophores in Ocyurus chrysurus (Fig. 34F) may lend support to the hypothesis based on morphological (including larval) and molecular data that Ocyurus is a synonym of Lutjanus (Domeier & Clarke, 1992;Chow, Clarke & Walsh, 1993;Clark, Domeier & Laroche, 1997), but information on colour patterns in larvae of other lutjanid genera is needed to determine at what taxonomic level the colour pattern in Lutjanus is significant. In general appearance, including the presence of melanophores dorsally on the head, a silvery gut, pigment associated with the pelvic-fin spine, and usually orange or yellow chromatophores on the caudal peduncle, larval Lutjanus is very similar to larvae of the grouper genus Mycteroperca (see 'Scorpaeniformes' below).
At the generic level, larval Haemulon (Haemulidae) has a stripe of xanthophores/erythrophores mixed with dark melanophores on the posterior half of the trunk (Fig. 35A, B). Among perciforms examined, larval Calamus is most similar in having a swathe of xanthophores/erythrophores on the posterior portion of the trunk mixed with melanophores (Fig. 35C, D). There are no chromatophores on the posterior portion of the trunk in the haemulid Anisotremus (Fig. 35E), but several species of the haemulid genus Pomadasys are very similar to Haemulon (e.g. Pomadasys commersonnii, Connell, 2007; Appendix). There is no consensus based on morphological or molecular data for inter-relationships of haemulids and sparids among perciforms.
A larva tentatively identified as an opistognathid (Fig. 36A) and the sciaenid Odontoscion (Fig. 36B) lack orange/yellow chromatophores on the trunk and have xanthophores confined to the head and gut region. A larval Mullidae, Upeneus parvus, does not resemble any other perciform larvae examined in that it is covered with melanophores and blue iridophores (Fig. 36C). Based on mitochondrial and nuclear DNA data, Smith & Craig (2007) suggested a close relationship between mullids and a nonperciform family, Dactylopteridae (Dactylopteriformes are incertae sedis in percomorphs in the classification of Wiley & Johnson, 2009). Although colour information is lacking, dactylopterid larvae have huge head spines that are lacking in Upeneus. Possibly the dense covering of iridophores in Upeneus larvae will be of value in identifying its closest relatives in the future. Preflexion larvae of Oplegnathus (Oplegnathidae), Pempheris (Pempheridae), Neoscorpis (Scorpididae), and several other perciform families from off South Africa have a considerable amount of yellow pigment on the trunk (Connell, 2007; Appendix), but larger specimens are needed for comparisons with other postflexion perciform larvae.

Pleuronectiformes
Pleuronectiformes are united as a monophyletic group in part based on ontogeny (transformation from a bilaterally symmetrical larva to an asymmetrical adult), and one or more elongate anterior dorsalfin elements characterize larvae of several families (Hensley & Ahlstrom, 1984). Colour patterns in larvae should prove useful phylogenetically at some levels. The presence of bright orange erythrophores on the median fins and fin bases, pelvic fin, head, and lateral midline (where they are small to minute) in the bothids Bothus maculiferus and Bothus ocellatus, combined with the absence of melanophores, may be synapomorphic for the genus (Fig. 37A, B). Superficially different from Bothus larvae because of the presence of numerous melanophores, larvae of the paralichthyid genus Syacium (e.g. Fig. 37C-F) are quite similar to Bothus in nonmelanistic colour pattern. Larvae of Syacium papillosum and three unidentified species have orange/yellow pigment on the median fins and fin bases, pelvic fin, head, and lateral midline, but also in a distinct patch on the dorsal portion of the gut. A single, small Citharichthys larva (Paralichthyidae) appears to have a similar pattern, but erythrophores are tiny (or contracted), and it is difficult to discern their precise distribution (Fig. 37G). Bothids and at least some paralichthyids (including Citharichthys and Syacium) consistently appear as sister groups or components of a slightly larger monophyletic group based on morphology and molecules (Cooper & Chapleau, 1998;Berendzen & Dimmick, 2002;Pardo et al., 2005;Azevedo et al., 2008), and the general colour pattern described above may help define a clade that includes bothids and those paralichthyids. However, the more distantly related Cynoglossidae (one Symphurus examined) has a similar colour pattern, with orange pigment present on the median fins and fin bases, head, and lateral midline (Fig. 37I). The single Achiridae examined (Trinectes) is heavily covered with melanophores, largely obscuring the nonmelanistic colour pattern, but pale orange pigment is visible on the median and pelvic fins (Fig. 37H). An in situ image of an unidentified pleuronectiform larva from off Hawaii, possibly a bothid, has the same general colour pattern  534 C. C. BALDWIN observed in most pleuronectiforms -i.e. yellow/orange pigment on the median fins (dorsal and anal fins almost completely covered with pigment), dorsal-and anal-fin bases, pelvic fin, head, and lateral midline (Fig. 38).
Species differences in the material examined are evident in the configuration of erythrophores, Bothus serving as an excellent example (Fig. 37A, B). In the absence of melanophores, preserved larvae of Bo. ocellatus and Bo. maculiferus lack diagnostic pigment patterns, but fresh specimens are easily distinguished by the pattern of erythrophores (Bo. ocellatus with more orange markings along the dorsal-and anal-fin bases than Bo. maculiferus and with erythrophores on the base of the caudal fin, in dashes along the dorsal and ventral body margins posteriorly, and in lines between those dashes and the orange markings along the bases of the dorsal and anal fins vs. none of these markings in Bo. maculiferus). As noted under 'Carangiformes', a relationship between pleuronectiforms and carangiforms as proposed by Little et al. (2010) would not appear to be supported by colour patterns in larvae, nor would a proposed relationship (Smith & Craig, 2007) with Xiphias gladius (see 'Scombriformes' below).

Scombriformes
Scombriformes are represented in the material examined by larvae of the Gempylidae, Istiophoridae, Scombridae, Sphyraenidae, and Xiphiidae. Early postflexion larvae of Xiphias (Fig. 39A) and Sphyraena (Fig. 39B) have extremely similar colour patterns -yellow extending from the snout to the caudal peduncle with numerous melanophores mixed in. A larval scombrid examined, by contrast, is mostly pale, with only a small amount of yellow pigment over the gut and pale orange pigment along the anal-fin base and lateral midline (Fig. 39C, see also http:// vertebrates.si.edu/fishes/larval/perci.html for an image of a preflexion Thunnus that is similar). A postflexion Auxis from off South Africa has some reddish pigment over the gut but is otherwise similarly pale (Connell, 2007;Appendix). A larval gempylid is similar to the scombrids examined in having a mostly pale head and trunk, but the spinous dorsal fin is densely pigmented with melanophores and bronze iridophores (Fig. 39D). A larval istiophorid examined does not resemble any of the other scombriform larvae in that nearly the entire body is covered with blue-reflecting iridophores (Fig. 39E). The relationship of billfishes to scombrids has been the subject of much controversy (e.g. Collette et al., 1984;Johnson, 1986;Orrell, Collette & Johnson, 2006;Little et al., 2010), and larval colour patterns do not appear to shed much light on the matter. The most recent molecular hypotheses (Orrell et al., 2006;Little et al., 2010) suggest that billfishes are not the closest relatives of scombrids. Orrell et al. (2006: fig. 3) proposed a sister-group relationship between Sphyraena and Xiphias + Istiophoridae, and, if apo- morphic, the similar colour patterns in larval Sphyraena and Xiphias could provide further support for this hypothesis.

Scorpaeniformes
Scorpaeniformes traditionally have comprised the scorpaenids and their mail-cheeked relatives (e.g. Nelson, 2006), but Wiley & Johnson (2009) followed Imamura & Yabe (2002) in placing scorpaenoids, platycephaloids, and the Serranidae in the Scorpaeniformes. Some molecular data conflict with this hypothesis and the proposed monophyly of the com-ponent suborders -Scorpaenoidei and Serranoidei (Smith & Wheeler, 2004;Smith & Craig, 2007). All scorpaeniforms for which colour in larvae has been examined -the scorpaenids Scorpaena, Scorpaenodes, Dendrochirus; the peristiid, Peristedion; and the serranids Diplectrum, Serranus, Hypoplectrus, Gonioplectrus, Mycteroperca, Bathyanthias, Liopropoma, Pseudogramma, Rypticus, and Pseudanthias -have orange/yellow chromatophores on the body. The three scorpaenid genera have erythrophores or xanthophores on an enlarged pectoral fin, the placement and extent of the colour aiding generic recognition: in Scorpaenodes nearly the entire fin is covered with xanthophores and melanophores with sometimes the very proximal area clear (Fig. 40A); in Scorpaena yellow or orange chromatophores -and sometimes dense melanophores -cover the proximal portion of the fin, and the distal portion is clear (Fig. 40B-D); and in Dendrochirus xanthophores form a band across the central portion of the fin (Fig. 40E). The pectoral fin superficially looks larger in Scorpaenodes than in the other genera because the distal portion is heavily pigmented. Within Scorpaena, the colour of the chromatophores on the pectoral fin and whether or not they are mixed with melanophores allow species recognition (Fig. 40B-D). In Peristedion, the pectoral fin is also enlarged, the upper rays extremely so, and there appear to be erythrophores mixed with melanophores on the upper elongate rays (Fig. 41).
Diplectrum, Serranus, and Hypoplectrus (serranine serranids) likewise have erythrophores or xanthophores, always mixed with melanophores, on the pectoral fin and are further similar to Scorpaena larvae in having yellow/orange chromatophores on the head and on the anterior and midlateral portions of the trunk (Fig. 42). The pectoral fin is not enlarged in the serranines, but it is enlarged and highly pigmented in some epinepheline serranids, including the genus Rypticus ( Fig. 43A-C). Distinct patterns of orange/ yellow chromatophores and melanophores on the enlarged fin characterize genetic lineages/species of Rypticus. Larvae of another epinepheline serranid examined, Pseudogramma, have an enlarged pectoral fin, but have only pale orange or yellow coloration vs. bright orange/yellow as in Rypticus (Fig. 43D). The presence of an elongate, pigmented pectoral fin is unusual among larval percomorphs examined and may be phylogenetically significant in uniting scorpaenids and serranids at the ordinal level.
Larval Pseudogramma also share with at least some Rypticus a similar pattern of prominent orange/ yellow chromatophores on the posterior portion of the trunk: one blotch is present centrally on the caudal peduncle, and two blotches precede it, one on the dorsal portion of the trunk and one on the ventral portion ( Fig. 43B-D). Pseudogramma and one species   of Rypticus (Fig. 43B, D) have additional orange/ yellow chromatophore blotches further anteriorly. An image of the Pacific Pseudogramma polyacantha shows a nearly identical colour pattern to that observed in the Atlantic Pseudogramma gregoryi except that the chromatophores are xanthophores rather than erythrophores (Fig. 44). Additional specimens of Pseudogramma polyacantha are needed to determine whether the colour is always yellow, but in multiple colour images of Pseudogramma gregoryi examined from Belize larval-fish collections, the colour is always orange.
Larvae of other epinepheline serranids examinedthe Spanish flag Gonioplectrus, the grouper Mycteroperca, and the liopropomin basses Bathyanthias and Liopropoma (Figs 45, 46) -do not have enlarged pectoral fins. Gonioplectrus has orange pigment on and at the base of the elongate second dorsal-fin spine, in a large blotch over the opercular region and base of pectoral fin, and along the entire length of the elongate pelvic-fin spine (Fig. 45). The pelvic-fin spine is also orange in settlement-stage Mycteroperca bonaci larvae (Fig. 45). Gonioplectrus has only melanophores in a blotch on the caudal peduncle, whereas Mycteroperca has erythrophores in a similar arrangement and posi-tion. Larval Bathyanthias from the Florida Straits and reared Liopropoma rubre larvae have two extremely elongate dorsal-fin spines, the anterior of which has pigmented swellings along its length (Fig. 46). In Bathyanthias, the swellings are small and orange, whereas in Liopropoma they are very large, and each is encircled with a prominent band of yellow pigment. Larvae of both genera have a stripe of erythrophores from the snout to the eye and additional erythrophores on the tip of the lower jaw and on the head behind the eye. Bathyanthias has orange pigment on the second dorsal, caudal, anal, and pelvic fins, but this pigment is not evident in available photographs of Liopropoma. Larvae of one anthiine serranid, Pseudanthias cooperi, do not greatly resemble other serranids in colour pattern: they are mostly pale, with a bright orange blotch at the posterior end of the lower jaw, pale xanthophores on the head, and pale xanthophores and erythrophores on the trunk (Fig. 47).

Stromateiformes
Stromateiformes are represented in the Belize material by one 5.0-mm SL unidentified specimen that resembles an illustration of a 5.1-mm SL larva of Cubiceps pauciradiatus (Lamkin, 2006) in pattern of melanophores. The DNA barcode for the Belize larva is approximately 3% different from that of Cu. pauciradiatus, which is greater than typical intraspecific divergence . Further study is needed, but the DNA data clearly identify the larva as a nomeid. It is mostly pale, but has xanthophores mixed with chromatophores on the top of the head and in a series over the swimbladder and gut (Fig. 48). Two species of Stromateidae from South Africa (Connell, 2007) have xanthophores mixed with melanophores in preflexion stages, and one postflexion specimen has most of the head and trunk covered with xanthophores and large dark spots.

Tetraodontiformes
Tetraodontiformes are represented in the material examined by larval monacanthids and ostraciids (Balistoidei), and tetraodontids and a molid (Tetraodontoidei). Larval Canthigaster rostrata and Sphoeroides (Tetraodontidae) have strikingly similar colour patterns in that most of the body is covered with iridophores that reflect bronze/gold/blue colours ( Fig. 49A-C). A similar shimmering colour pattern is present in larval Ranzania, a molid (Fig. 49D). Colour pattern in larvae of these tetraodontoids, which is very different from that observed in larval ostraciids and monacanthids, may provide corroborative evidence for the monophyly of the Tetraodontoidei (sensu Holcroft, 2005) or a less inclusive clade within that    assemblage. Monocanthus ciliatus larvae are mostly yellow, with xanthophores present on the head, trunk, and spinous dorsal and pelvic fins, and larval Aluterus is less colourful but has some yellow pigment on the head, trunk, and spinous dorsal and caudal fins (Fig. 50A, B). Larval ostraciids have a pale orange/yellow coloration overlain by dark dots or large dark spots (Fig. 50C, D). Relationships of tetraodontiforms to other acanthomorph fishes are unresolved, but taxa hypothesized to be close relatives include zeiforms, acanthuriforms, and lophiiforms (Wiley & Johnson, 2009, and references therein). There is little resemblance between tetraodontiform larvae examined and Z. faber (Connell, 2007;Appendix), and, as noted above (see Acanthuriformes), there is little similarity between acanthurid and tetraodontiform larvae. There is at least superficial resemblance between tetraodontid and chaetodontid/ pomacentrid larvae in that the pigment is primarily in the form of iridophores (Figs. 30, 49).
The tetraodontiform/lophiiform hypothesis (Holcroft & Wiley, 2008;Miya et al., 2010) is intriguing in light of colour patterns and other aspects of larval morphology: compare a preflexion larval ostraciid from South Africa with an unidentified lophiiform (Fig. 51). In both, but in larvae of no other teleosts examined, the trunk is contained within an inflated, threedimensional sac that is covered with xanthophores. Aboussouan & Leis (2004) noted the presence of a 'dermal sac' in tetraodontids, diodontids, molids, ostraciids, and balistids but not in triacanthodids, triacanthids, and monacanthids. Pietsch (1984: 320) described lophiid and other lophiiform larvae as having the epidermal layer of the head and body 'greatly distended by transparent, gelatinous connective tissue'. The inflated condition in tetraodontiforms appears to be restricted to the preflexion stage, whereas that in lophiiforms persists after flexion (compare Figs 50C, D and 51; see also Connell, 2007).

Trachiniformes
Trachiniformes, which currently comprise the fishes traditionally included in the Trachinoidei (Chiasmodontidae, Champsodontidae, Pinguipedidae, Cheimarrhichthyidae, Trichodontidae, Creediidae, Percophididae, Leptoscopidae, Trachinidae, and Uranoscopidae) and Ammodytidae, were considered likely to be paraphyletic by Wiley & Johnson (2009). Among trachinoid larvae from off South Africa (Connell, 2007;Appendix), an early-stage larval uranoscopid bears little resemblance to an early-stage chiasmodontid and a champsodontid. The last two are similar in having only melanophores and lacking obvious nonmelanistic colour in the preflexion stage, whereas the larval uranoscopid has distinctive blotches of xanthophores. Mito (1962) described a reared series of Champsodon snyderi and noted that the species has only melanophores throughout larval development. Information on colour patterns in postflexion larvae of other champsodontids might be useful in commenting on the proposed relationship between champsodontids and scorpaeniforms (Mooi & Johnson, 1997). As noted (see 'Scorpaeniformes' above), the latter have distinctive patterns of erythrophores/xanthophores.

DISCUSSION
The first broad comparative survey of nonmelanistic colour in marine teleost larvae presented herein has revealed a striking array of colour patterns that may inform phylogeny at multiple taxonomic levels. The absence of xanthophores, erythrophores, and iridophores in pelagic larvae of most basal marine teleosts examined, and their presence in many basal neoteleosts, in all percomorph orders examined, and in nearly every percomorph species examined, suggest a phylogenetic trend (Fig. 1). It is premature to make specific hypotheses about the evolution of the three types of chromatophores in larvae of marine teleosts, but in the limited material examined, xanthophores are the only type of nonmelanistic chromatophore present in larval marine anguilliforms and stomiatiforms. Erythrophores appear at the level of Eurypterygia, and iridophores first appear in larval euacanthopterygians. The situation in Anguilliformes and basal neoteleosts, in which larvae of some members of an order exhibit nonmelanistic pigment but others do not, suggests that nonmelanistic chromatophores have arisen independently (or were lost) multiple times. Considerably more information on colour patterns in more larvae is needed to determine the taxonomic distribution of the various types of chromatophores among teleosts, as is investigation of the cellular basis of observed chromatophore types. In this study, presence or absence of chromatophores was assessed from visual examination of fresh specimens and colour photographs, but histological studies are needed to confirm that the presence or absence of a particular nonmelanistic pigment corresponds to the structural presence or absence of the vertebrate chromatophore unit that typically contains that pigment. Biological and physical factors that might affect the development or display of chromatophores also warrant study. For example, thyroid hormones are known to have a role in regulating larval-fish development, including pigment (e.g. Brown & Kim, 1995, and references therein). However, experimental manipulation of exogenous hormones suggests they may affect timing of pigment development and density/shape of pigment features but not the types of chromatophores that develop (Reddy & Lam, 1992; COLOUR IN MARINE TELEOST LARVAE 549  Brown & Kim, 1995;Clement, Lichtenbert & Kohler, 2001). It seems unlikely that physical factors related to habitat ecology have much bearing on the presence or absence of nonmelanistic chromatophores, as larvae of most marine teleosts inhabit a similar planktonic realm. In the material examined for this study, fish larvae such as those of clupeids, engraulids, elopiforms, and anguilliforms that lack nonmelanistic chromatophores were collected in the same plankton samples as larvae that have them.
Apomorphic characters of chromatophore patterns rather than the presence or absence of chromatophore types may have the greatest phylogenetic potential in teleosts. Pigment patterns of adults have been incorporated into phylogenetic studies of fishes at subgeneric, generic, and familial levels (e.g. Mabee, 1995;Ayache & Near, 2009;Layman & Mayden, 2012), but nonmelanistic colour patterns of marine fish larvae have not. Among the percomorphs examined, colour patterns in marine fish larvae may have phylogenetic significance at several taxonomic levels: interordinal -Mugiliformes/Beloniformes, Tetraodontiformes/ Lophiiformes; interfamilial -Chaeotodontidae/Pomacentridae, Bothidae/Paralichthyidae, Chaenopsidae/ Labrisomidae, Callionymidae/Gobiesocidae; familial -Apogonidae, Eleotrididae, Microdesmidae, Scaridae, Scorpaenidae, Tetraodontidae; intergeneric -Abudefduf/Amblyglyphidodon, Anampses/Halichoeres/ Thalassoma, Ctenogobius/Gnatholepis/Gobionellus, Nes/Psilotris, Lutjanus/Ocyurus; and generic -Bathygobius, Coryphopterus, Scorpaena, Rypticus, Xyrichtys. Slight differences in generic colour patterns of larvae -including whether the pattern comprises xanthophores or erythrophores -often distinguish component species (e.g. those of Xyrichtys, Pseudogramma, Scorpaena). Matsumoto (1965) noted that vesicles that contain yellow and red pigments are sometimes found in the same cell in freshwater Xiphophorus, suggesting that the distinction between xanthophores and erythrophores is artificial. The presence in some marine genera of species with nearly identical patterns of pigment except for the colour of the chromatophores (orange or yellow) suggests that xanthophores and erythrophores are distinct. Erythrophores in fishes reportedly are red/orange because of carotenoids ingested in the diet, whereas the yellow pigment of xanthophores (from pteridines) is metabolized endogenously (Grether et al., 2004, and references therein). Whether the origins of yellow and orange pigments in marine fish larvae also are endogenous and exogenous, respectively, does not appear to have been investigated. It seems remarkable that nearly identical patterns of pigment that differ only in colour (e.g. Xyrichtys, Fig. 26A, C; Pseudogramma, Fig. 44) would have different biochemical origins. Future studies that investigate the biochemistry of pigments in marine fish larvae might also compare the chemistry of pigments in basal teleosts such as eel leptocephali, basal neoteleosts such as whalefish, and percomorphs. Similarities or differences in the biochemistry of the pigments may shed light on the homology and evolutionary history of larval-fish pigments among marine teleosts.  and Mabee (1995) noted that individual variation and convergent evolution of some patterns are considered obstacles for phylogenetic use of pigment characters because of difficulties in establishing hypotheses of homology. Numerous specimens of many species were examined in this study (Appendix), and individual variation in nonmelanistic chromatophore patterns is low. Differences frequently appear to reflect the expanded or contracted state of the chromatophores rather than their pattern (Fig. 52). Assessing homology of nonmelanistic chromatophores in larvae among major groups of marine teleosts through phylogeny is complicated by the absence of a comparable, transient, pigment phase in most freshwater fishes. Larvae of marine Clupeiformes examined, for example, lack nonmelanistic chromatophores, but most Otomorpha are freshwater inhabitants. The presence of xanthophores and iridophores in larval cypriniforms such as freshwater Danio is well documented (Johnson et al., 1995;Parichy et al., 2000;Parichy, 2003Parichy, , 2006. What is the relationship between the pigment pattern of larval freshwater fishes and that of pelagic marine fish larvae? As shown in the ontogenetic series of numerous marine teleosts photographed by Connell (2007), the pigment pattern of fully developed marine fish larvae is preceded by a very different pigment phase in recently hatched larvae. For example, a reared myctophid series shows yellow pigment on the head and body in yolksac larvae, but all yellow pigment has disappeared by 4.3 mm NL. Acanthurid acronurus larvae examined have iridophores but no erythrophores or xanthophores, yet preflexion Acanthurus mata exhibit xanthophores (see 'Acanthuriformes' above). Studies on aquarium-reared callionymids revealed a colour transformation in larval Syn. splendidus from yellow in the preflexion stage to bright orange after flexion ( Fig. 19; Wittenrich et al., 2010). Possibly the early pigment pattern in marine fish larvae is the ontogenetic counterpart of the larval pigment pattern in freshwater fishes. Assembling ontogenetic series of marine fishes that incorporate chromatophore patterns from recently hatched to adult stages, investigating the development of the pattern at each stage, and comparing the development of pigment among marine and freshwater fishes would shed light on the homology of pigment patterns among life-history stages and among related taxa. Such ontogenetic series also warrant further study as a potential source of phylogenetic characters. Mabee (1995) highlighted the phylogenetic significance of pigment-pattern development in the evolution of centrarchid sunfishes, which are freshwater fishes that exhibit direct development of the adult pigment pattern from the early larval stage. The ontogeny of pigment patterns in marine fishes may be an even riper source of phylogenetic information yet to be tapped. In addition to the colour transitions in myctophids, acanthurids, and callionymids mentioned above, the ontogenetic transition in mugilids and some beloniforms from a stage dominated by xanthophores, erythrophores, and melanophores to one dominated by iridophores is an example. This transition also suggests that the presence of iridophores may be homologous in the two groups because the condition was derived from a similar ontogenetic transformation. Incorporating developmental studies of colour patterns in future phylogenetic treatments of the topic could thus provide both crucial information on character homology and new phylogenetic characters. Yasir & Qin (2007) described developmental changes in pigment patterns of the marine anemonefish Amphiprion ocellaris from embryo to larva and noted that a better understanding of the process of pigment formation is needed. Embryologically, neural crest cells, which give rise to all four types of chromatophores, migrate to the dermis, and the density and distribution of the chromatophores are integrated to produce colour patterns in vertebrates (Hawkes, 1974;Bagnara, 1987;Hall, 1999). Holt (2011) noted that the pigment patterns of fishes are formed during embryogenesis and are replaced by the adult pattern at metamorphosis. Studies of freshwater Danio indicate that the adult colour pattern results from incorporation of both embryonic chromatophores and differentiation of new chromatophores from stem cells at metamorphosis, and the ratio of contribution by the two may vary significantly among species (Parichy, 2003(Parichy, , 2006. The existence of an additional colour phase (between the recently hatched and adult phases) in pelagic larvae of marine fishes would appear to complicate this scenario. Furthermore, juveniles of many marine fishes exhibit yet another colour phase that is different from the pelagic larval and adult phases (e.g. Haemulidae, Labridae, Pomacentridae, Pomacanthidae). The ontogeny of chromatophore patterns has been studied in few marine fishes, but Nakamura et al. (2010 and references therein) have investigated development of the adult pigment pattern in Pleuronectiformes. Two types of melanophores and xanthophores/erythrophores are hypothesized to exist in pleuronectiforms -those that develop in the larval stage prior to metamorphosis and those that develop afterwards and form the adult pigment pattern. Flatfishes exhibit an unusual ontogeny in which larvae are bilaterally symmetrical, including the pigment pattern, but metamorphosis involves migration of one eye to the other side of the head resulting in a bilaterally asymmetrical juvenile and adult. There is typically little or no pigment on the 'blind' side but the 'eyed' side is well pigmented. Whether chromatophores of pelagic larvae in other marine fishes are succeeded ontogenetically by adult versions of those pigment cells is unknown, but this would be a reasonable hypothesis based on the different pigment patterns exhibited in larval and adult stages. Other features that are present in pelagic larvae of some marine fishes, such as elongate fin rays and head spines, disappear or transform upon metamorphosis to the juvenile stage. Studies in marine fishes that would explain the origin of adult pigment patterns (i.e. from existing neural crest lineages or newly differentiated stem cells) are lacking.
If the colour patterns exhibited by marine fish larvae are present only during the pelagic larval period, are they adaptations that have functional significance? Grether et al. (2004) discussed the potential significance of the different pigments in adult freshwater fishes, but only one of their hypotheses seems potentially relevant to pelagic marine fish larvae -'spectral fine-tuning '. Grether et al. (2004) noted that orange and yellow pigments have different absorptive properties that may enable spectral fine-tuning, which, in turn, may enhance the ability to blend into the background. Orange coloration is not restricted to fishes in a plankton sample: numerous invertebrates such as the larvae of shrimps and crabs look very similar. Indeed, the guts of numerous freshly caught marine fish larvae are orange because of diet (see Figs 27D,39D,41,46). Carvalho, Zuanon & Sazima (2006) provided examples of transparency and similar colour patterns in freshwater fishes and crustaceans that travel as a group and suggested that these organisms may be utilizing a type of protective association known as numerical mimicry as a means of avoiding potential predators. Regarding the bright orange coloration in larvae of the mandarinfish, Syn. splendidus, Wittenrich et al. (2010) noted that Lindquist (2002) and Young & Bingham (1987) hypothesized that orange may be an aposematic warning colour in the larvae of some marine organisms.
In summary, colour patterns in marine teleost larvae emerge as an intriguing new source of potentially valuable phylogenetic information, but considerably more data are needed to fully assess their significance. An increased global focus on documenting colour patterns in larvae prior to preservation would facilitate incorporating information from larval colour patterns into more systematic studies, and developmental, histological, and biochemical studies that shed light on the ontogeny of colour patterns in early life history stages would improve our understanding of ontogenetic pigment phases, characters, character transformations, and character homology. ACKNOWLEDGEMENTS David Smith and Lee Weigt contributed to the field work that resulted in the Belize larval-fish images in numerous ways. The molecular data that enabled identifications of larvae analysed in this paper were published by Weigt et al. (2012), and a complete list of acknowledgements for the Belize field work is pro- Guigand, and he provided contact information for those photographers and Donald Hughes. I thank Allan Connell, Cedric Guigand, Joshua Lambus, Matthew D'Avella, Michael Miller, Donald Hughes, Chris Papero, Todd Gardner, Matthew Wittenrich, Robert Cowen, Daniel Benetti, and Dallas Alston for allowing me to use their images of fish larvae or assisting me in locating images. The addition of numerous fish species not represented in the Belize material greatly increased the taxonomic scope of the paper, and many of the additional images are so stunning that it was a privilege to study and incorporate them into this work. Ross Robertson provided a portable Wacom tablet and provided instructions on how to remove images from their original backgrounds. Without this editorial procedure, many of the Belize larval-fish images would not have been publishable. Sandra Raredon and Jeff Williams provided initial guidance on the best photographic equipment and equipment set-up for photography in the field. Wen-Hsiung Li, Chief Editor of Zoological Studies, and Aoi Sasaki of TERRAPUB (publisher of Aqua-BioScience Monographs), allowed the images in Figure 4B and Figure 5,