Molecular phylogeny of the gobioid fishes(Teleostei: Perciformes: Gobioidei)

Christine E. Thacker*

Vertebrates-Ichthyology, Natural History Museum of Los Angeles County, 900 Exposition Blvd., Los Angeles, CA 90007, USA

Received 2 July 2001; revised 29 January 2002

Abstract

The phylogeny of groups within Gobioidei is examined with molecular sequence data. Gobioidei is a speciose, morphologically

diverse group of teleost fishes, most of which are small, benthic, and marine. Efforts to hypothesize relationships among the gobioid

groups have been hampered by the prevalence of reductive evolution among goby species; such reduction can make identification of

informative morphological characters particularly difficult. Gobies have been variously grouped into two to nine families, several

with included subfamilies, but most existing taxonomies are not phylogenetic and few cladistic hypotheses of relationships among

goby groups have been advanced. In this study, representatives of eight of the nine gobioid familes (Eleotridae, Odontobutidae,

Xenisthmidae, Gobiidae, Kraemeriidae, Schindleriidae, Microdesmidae, and Ptereleotridae), selected to sample broadly from the

range of goby diversity, were examined. Complete sequence from the mitochondrial ND1, ND2, and COI genes (3573 bp) was used

in a cladistic parsimony analysis to hypothesize relationships among the gobioid groups. A single most parsimonious topology was

obtained, with decay indices indicating strong support for most nodes. Major phylogenetic conclusions include that Xenisthmidae is

part of Eleotridae, and Eleotridae is paraphyletic with respect to a clade composed of Gobiidae, Microdesmidae, Ptereleotridae,

Kraemeriidae, and Schindleriidae. Within this five-family clade, two clades are recovered. One includes Gobionellinae, which is

paraphyletic with respect to Kraemeriidae, Sicydiinae, Oxudercinae, and Amblyopinae. The other contains Gobiinae, also para-

phyletic, and including Microdesmidae, Ptereleotridae, and Schindleriidae. Previous morphological evidence for goby groupings is

discussed; the phylogenetic hypothesis indicates that the morphological reduction observed in many goby species has been derived

several times independently.

� 2002 Elsevier Science (USA). All rights reserved.

Keywords: Gobioidei; Odontobutidae; Eleotridae; Eleotrinae; Xenisthmidae; Gobiidae; Gobiinae; Gobionellinae; Sicydiinae; Oxudercinae;

Amblyopinae; Kraemeriidae; Microdesmidae; Ptereleotridae; Schindleriidae; Molecular phylogeny; Miniaturization; Reduction

1. Introduction

Gobioidei includes an estimated 2121 species in 268

genera or 23% of perciforms (Nelson, 1994). Gobies are

widely distributed throughout the tropical, subtropical,

and temperate regions of the world, in freshwater and

nearshore marine habitats. They are a prominent com-ponent of many fish faunas, but because goby species

are generally cryptic and difficult to sample, the biology

of the group is understudied. The majority of gobies are

benthic, often living in burrows, but the group also in-

cludes nektonic reef-dwellers, planktonic species, and

estuarine representatives with the ability to breathe air.

Most gobies attain a small adult size; the largest species

may reach a length of 100 cm or more but the majority

are 10 cm or less. Compared to other perciforms, gobies

are not only small but also often morphologically re-

duced, with many species possessing simplifications and

losses in various aspects of morphology. Gobioidei in-cludes the most extreme case of vertebrate paedomor-

phosis, the genus Schindleria (Johnson and Brothers,

1993), and many other gobies exhibit lesser degrees of

morphological reduction (Iwata et al., 2001; Matsubara

and Iwai, 1959; Springer, 1983). These factors have all

hindered studies of goby phylogeny, and relationships

both within and among goby groups are mostly unre-

solved.

Molecular Phylogenetics and Evolution 26 (2003) 354–368

www.elsevier.com/locate/ympev

MOLECULARPHYLOGENETICSANDEVOLUTION

* Fax: 1-213-748-4432.

E-mail address: thacker@nhm.org.

1055-7903/02/$ - see front matter � 2002 Elsevier Science (USA). All rights reserved.

doi:10.1016/S1055-7903(02)00361-5

The current classification of gobies reflects the un-certain knowledge of goby relationships. As with many

large vertebrate groups, the trend in gobioid classifica-

tion has been to identify groups of genera or species

which share some distinct morphological characters and

elevate them to family rank. This approach has resulted

in a classification in which many small families have

been subdivided from the largest groups; often mor-

phological character evidence is presented to supportmonophyly of the defined groups and sometimes trees

are presented (Gill and Hoese, 1993; Harrison, 1989;

Hoese and Gill, 1993; Rennis and Hoese, 1987; Springer,

1973, 1983), but cladistic analyses of goby taxa are rare

(Larson, 2001; Murdy, 1989; Parenti and Thomas, 1998;

Thacker, 2000). The group is diagnosed by more than 20

apomorphic characters, but its sister group is unknown

(Winterbottom, 1993). Several schemes for gobioidhigher classification have been proposed, including two

(Miller, 1973), six (Hoese, 1984; Hoese and Gill, 1993;

Pezold, 1993), eight (Nelson, 1994), or nine (Thacker,

2000) families, many with included subfamilies (taxo-

nomic history is reviewed in Harrison, 1989 and Akihito

et al., 2000). The classification given in Table 1 is a

composite of recent classifications, with the approximate

number of genera and species in each named taxon givenafter the taxon name.

All of the named taxa in Table 1 except Odontob-

utidae, Butinae, and Gobionellinae may be diagnosed by

at least one character (Hoese, 1984; Hoese and Gill,

1993; Pezold, 1993). Additionally, several studies have

described variation in characters that are potentially

useful for elucidating phylogeny among the larger

gobioid groups. Gosline (1955) described a variety ofosteological characters for representatives of Eleotridae,

Gobiidae, Kraemeriidae, and Microdesmidae, and rec-

ommended that Microdesmidae be included in Gobioi-

dei. Takagi (1989) surveyed the sensory canal system of

55 genera of Japanese gobioids, and Akihito (1986) ex-amined the morphology of the sensory canals as well as

the suspensorium, branchial apparatus and pectoral

girdle. Birdsong et al. (1988) identified patterns in the

spinous dorsal fin pterygiophore formula and select

other characters of the axial skeleton for over 200

gobioid genera that they used to delineate groups of

genera within the described families and subfamilies.

Harrison (1989) used characters of the palatopteroqua-drate complex in the suspensorium to hypothesize rela-

tionships among groups of genera in the gobiid

subfamilies.

In spite of all the morphological character data that

has been identified, no cladistic analysis of the large-

scale relationships, among the gobioid families and

subfamilies, has been presented. There is general agree-

ment that the more reduced and simplified gobies are themost derived. Characters such as the reduction of the

epurals and lateral line and the loss of the anterior

branchiostegal ray, infraorbital bones, endopterygoid,

basibranchials 2–4, and various sensory canals have all

been used to define goby groups (Hoese, 1984). The

most extreme example of morphological reduction

among gobies, and among vertebrates generally, is seen

in the genus Schindleria. As adults, the two Schindleria

species resemble larval gobiids, possessing larval char-

acters such as a transparent body, functional pronephric

kidney, tubular heart and many losses and reductions in

the skeletal system. Schindleria has been placed in Go-

bioidei based on otolith morphology, egg morphology,

presence of a sperm duct gland and skeletal characters

including similarities between the caudal skeleton of

Schindleria and that of larval gobioids (Johnson andBrothers, 1993), but the sister taxon to Schindleria

within Gobioidei has not been determined.

Schindleria is an extreme example of a trend towards

reduction often described in studies of gobioid rela-

Table 1

Classification of groups within Gobioidei, with number of genera and species given following each taxon name

Family Subfamily Species Reference

Rhyacicthyidae (1 genus; 1 species) Miller (1973)

Odontobutidae (3 genera; 4–5 species) Hoese and Gill (1993)

Eleotridae (35 genera; est. 150 species) Hoese and Gill (1993)

Butinae (13 genera) Hoese and Gill (1993)

Eleotridinae (21–22 genera) Hoese and Gill (1993)

Xenisthmidae (5 genera; 19 species) Springer (1983)

Gobiidae (212 genera; est. 1875 species) Pezold (1993)

Oxudercinae (10 genera; 34 species) Murdy (1989)

Amblyopinae (12–13 genera; est. 30 species) Murdy and Shibukawa (2001)

Sicydiinae (5–6 genera; est. 100 species) Parenti and Thomas (1998)

Gobionellinae (56 genera) Pezold (1993)

Gobiinae (109 genera) Pezold (1993)

Kraemeriidae (2 genera; 8 species) Gosline (1955)

Microdesmidae (5 genera; 30 species) Thacker (2000)

Ptereleotridae (5 genera; 30 species) Thacker (2000)

Schindleriidae (1 genus; 2 species) Johnson and Brothers (1993)

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 355

tionships. Xenisthmidae, Kraemeriidae, Microdesmidae,and some Gobiidae also exhibit reduction but the af-

fected structures are not always the same. Therefore, one

question that may be asked concerning gobioid interre-

lationships is whether or not reduction has occurred as a

single gradual trend or several times independently in

different goby groups. In this study, DNA sequence data

are used as a character source to investigate relation-

ships among gobioid families, subfamilies and genera.Sequence data are attractive for resolving gobioid rela-

tionships because they are independent of the reduction

that can confound morphological character analyses.

The aims of this study were to provide a new view of

goby phylogeny on a broad scale, to allow reinterpre-

tation of previously described morphological character

data and to clarify relationships among groups where

morphological character data have proved insufficient.This study will also serve as a step towards assembling

large scale total evidence phylogeny for the group, pro-

viding a framework for future studies using both mo-

lecular and morphological data. Representatives of all

the named goby taxa listed in Table 1 were included,

with the exception of the monotypic Rhyacichthyidae.

Emphasis was placed on the families and subfamilies

that have been postulated to be more derived (Kra-emeriidae, Schindleriidae, Microdesmidae, Ptereleotri-

dae, Gobiinae, Gobionellinae, Sicydiinae, Oxudercinae,

and Amblyopinae), based on a character of the bran-

chial skeleton, loss of the anterior branchiostegal ray

(Hoese, 1984). Additionally, most members of these

putatively more derived taxa are reduced in size com-

pared to eleotrids, odontobutids, and rhyacichthids and

possess various other reductions in morphology.The complete sequence of three mitochondrial genes

(ND1, ND2, and COI) was used as the character source

for this analysis. Mitochondrial DNA sequence is useful

for resolving phylogenetic relationships due to its pat-

tern of inheritance (maternally inherited, without re-

combination) and rapid rate of change compared to

nuclear genes. Several previous studies have used mito-

chondrial sequence data to resolve relationships amongactinopterygian and chondrichthyian species (Block et

al., 1993; Chow and Kishino, 1995; Finnerty and Block,

1995; Kocher et al., 1995; Kocher and Stepien, 1997;

Tang et al., 1999; Wiley et al., 1998, 2000). Most of these

studies have involved freshwater fishes, and many use

ribosomal DNA sequence as a data source. The ND1,

ND2, and COI genes were chosen because unlike the

ribosomal DNAs, these genes are protein-coding genes.In ribosomal DNAs, the translated RNA is the final

functional product and insertions and deletions which

affect the stem and loop structure of the RNA are

common, rendering alignment of sequences for phylo-

genetic analysis particularly difficult. In protein coding

genes, insertions and deletions are much rarer and when

they do occur, they generally involve addition or loss of

a codon; knowledge of this constraint and the ability toperform alignments based on translated amino acid se-

quence makes alignment much less ambiguous. The

entire sequence of three genes (3573 bp total) was used

to provide a large enough amount of sequence data to

provide adequate resolution of relationships at this

broad scale. Mitochondrial genes are also appropriate

choices for resolution of relationships within Gobioidei

based on saturation patterns; saturation is not acute inmitochondrial genes for divergences less than approxi-

mately 100 million years ago (Mindell and Thacker,

1996). Goby fossils are scarce, but are not known from

earlier than the Eocene (Miller, 1973; Patterson, 1993).

Perciforms, the larger group of which Gobioidei is a

part, are not present prior to the upper Cretaceous

(80million years ago; Patterson, 1993).

2. Materials and methods

Fresh and ethanol-preserved tissues for DNA se-

quencing were obtained from several sources (Table 2).

In most cases, only one individual of each species was

sequenced, largely due to the scarcity of available tis-

sues. In ten cases two individuals were sequenced: El-eotris sandwicensis, Gnatholepis cauerensis, Amblygobius

phalaena, Microdesmus longipinnis, Risor ruber, Nema-

teleotris magnifica, Ptereleotris zebra, Pandaka lidwilli,

Ctenogobius saepepallens, and Kraemeria cunicularia.

Three individuals of Gnatholepis thompsoni were se-

quenced. When more than one individual was examined,

the sequences were very similar but not identical, and in

the phylogenetic hypothesis they were recovered to-gether. Individuals of 67 species representing 51 genera

were sequenced; eight eleotridid and one xenisthmid

species were included, and the odontobutid Odontobutis

obscura was designated as the outgroup in the analysis.

A previous study of relationships of eleotrids (Hoese

and Gill, 1993) included morphological character data

that indicated that rhyacichthids and odontobutids are

the primitive sister taxa to other gobies, and that botheleotridid subfamilies form an unresolved trichotomy

with the rest of Gobioidei.

Total genomic DNA was extracted from tissues using

the QIAquick Tissue Kit (Qiagen, Chatsworth, CA) and

quantified by running 5 ll of each extraction with 1 ll ofloading dye on a 1.5% low melting point agarose gel

stained with ethidium bromide. In some cases, for am-

plification of the ND1 and ND2 genes, hotstart XLPCR was performed using primers L3827 and H6313

(Sorenson et al., 1999) and Taq rTth XL polymerase

with AmpliWax PCR Gems (Perkin–Elmer, Foster City,

CA). The PCR was performed with a profile of 94 �C for

5min, followed by 16 cycles of 94 �C/30 s denaturation,50–53 �C/20 s annealing and 70 �C/4min extension, then

21 cycles of the same profile but with 30 additional

356 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

Table 2

Species sequenced for this study

Species Source GenBank Accession Nos.

Odontobutidae

Odontobutis obscura Akihisa Iwata, Japan AF391330, AF391402, AF391474

Eleotridae: Eleotrinae

Eleotris sandwicensis Small bottom trap, stream, North Oahu, Hawaii AF391333-4, AF391405-6, AF391477-8

Erotelis smaragdus Bottom tow, Twin Cays, Belize AF391355, AF391427, AF391499

Hypseleotris aurea Peter Unmack, Gascoyne River, WA, Australia AF391392, AF391464, AF391536

Hypseleotris compressa Peter Unmack, Ross River, Qld., Australia AF391366, AF391438, AF391510

Hypseleotris klunzingeri Peter Unmack, Barcoo River, Qld., Australia AF391393, AF391465. AF391537

Mogurnda adspersa Peter Unmack, Ross River, Qld., Australia AF391367, AF391439, AF391511

Ophieleotris aporos Peter Unmack, Ross River, Qld., Australia AF391368, AF391440, AF391512

Philypnodon grandiceps Peter Unmack, Glenelg River, Vic., Australia AF391386, AF391458, AF391530

Xenisthmidae

Xenisthmus sp. Mark Westneat, Santa Cruz Island, Solomon Islands AF391372, AF391444, AF391516

Gobiidae: Gobionellinae

Acanthogobius flavimanus Scott Matern, Sacramento River Delta AF391381, AF391453, AF391525

Awaous guamensis Brent Tibbats, Guam AF391338, AF391410, AF391482

Chaenogobius annularis Ho Young Suk, Korea AF391365, AF391437, AF391509

Ctenogobius saepepallens Plankton tow, Carrie Bow Cay, Belize AY077595-6, AY077602-3, AY077609-10

Eucyclogobius newberryi CAS 86280; San Gregorio Creek, California AF391361, AF391433, AF391505

Evorthodus minutus Jim Van Tassell, Mazatlan, Mexico AY077593, AY077600, AY077607

Gillichthys mirabilis Nancy Aguilar, California AF391340, AF391412, AF391484

Gnatholepis cauerensis Quinaldine, Moorea, Society Islands AF391364 & 75, AF391436 & 47,

AF391508 & 19

Gnatholepis scapulostigma Quinaldine, Moorea, Society Islands AF391376, AF391448, AF391520

Gnatholepis thompsoni Quinaldine, Carrie Bow Cay, Belize AF391343-4, AF391415-6, AF391487-8,

AY077594, AY077601, AY077608

Gobiopterus semivestita Peter Unmack, Milingandi Creek, NSW, Australia AF391387, AF391459, AF391531

Mugilogobius sp. Brent Tibbats, Guam AF391356, AF391428, AF391500

Mugilogobius rivulus Peter Unmack, Leaders Creek, NT, Australia AY077592, AY077599, AY077606

Pandaka lidwilli Tony Gill, Innes Park Creek, Qld., Australia AY077590-1, AY077597-8, AY077604-5

Stenogobius hawaiiensis Brent Tibbats, Guam AF391349, AF391421, AF391493

Typhlogobius californiensis Nancy Aguilar, California AF391345, AF391417, AF391489

Gobiidae: Gobiinae

Amblyeleotris wheeleri Quinaldine, Moorea, Society Islands AF391383, AF391455, AF391527

Amblygobius nocturnus Quinaldine, Moorea, Society Islands AF391379, AF391451, AF391523

Amblygobius phalaena Quinaldine, Moorea, Society Islands AF391369 & 78, AF391441 & 50,

AF391513 & 22

Asterropteryx semipunctatus Quinaldine, Moorea, Society Islands AF391377, AF391449, AF391521

Barbulifer ceuthoecus Quinaldine, Carrie Bow Cay, Belize AF391353, AF391425, AF391497

Bathygobius cocosensis Quinaldine, Rangiroa, Tuamotu Atolls AF391388, AF391460, AF391532

Bathygobius curacao Quinaldine, Pelican Cays, Belize AF391354, AF391426, AF391498

Cabillus tongarevae Quinaldine, Moorea, Society Islands AF391382, AF391454, AF391526

Callogobius sclateri Quinaldine, Moorea, Society Islands AF391390, AF391462, AF391534

Coryphopterus dicrus Kathleen Cole, Carrie Bow Cay, Belize AF391395, AF391467, AF391539

Coryphopterus hyalinus Kathleen Cole, Carrie Bow Cay, Belize AF391326, AF391398, AF391470

Coryphopterus personatus Kathleen Cole, Carrie Bow Cay, Belize AF391325, AF391397, AF391469

Coryphopterus punctipectophorus Kathleen Cole, Carrie Bow Cay, Belize AF391396, AF391468, AF391540

Ctenogobiops feroculus Quinaldine, Moorea, Society Islands AF391363, AF391435, AF391507

Eviota afelei Quinaldine, Moorea, Society Islands AF391391, AF391463, AF391535

Fusigobius neophytus Quinaldine, Moorea, Society Islands AF391374, AF391446, AF391518

Fusigobius signipinnis Mark Westneat, Santa Cruz Island,

Solomon Islands

AF391370, AF391442, AF391514

Gobiodon histrio Rob Reavis, Captive stock AF391360, AF391432, AF391504

Gobiosoma macrodon Colette St. Mary, Florida AF391348, AF391420, AF391492

Lophogobius cyprinoides Kathleen Cole, Florida AF391362, AF391434, AF391506

Priolepis cincta Quinaldine, Moorea, Society Islands AF391385, AF391457, AF391529

Priolepis eugenius David Greenfield, Hawaii AF391329, AF391401, AF391473

Risor ruber Colette St. Mary, Florida AF391351-2, AF391423-4, AF391495-6

Valenciennea strigata Quinaldine, Moorea, Society Islands AF391384, AF391456, AF391528

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 357

seconds of extension added at each step. These long

(� 2500 bp) fragments were quantified on a 1.5% low

melting point agarose gel stained with ethidium bro-

mide, bands were visualized and photographed under

UV light, cut from the gel and DNA purified from the

bands using the QIAquick gel extraction kit (Qiagen,

Chatsworth, CA). The long PCR fragments were used as

template for four shorter PCR reactions using the pri-mer pairs: L3827/H4644; L4500/H5191; L5219/H5766;

and L5758/H6313 (Sorenson et al., 1999). These ampli-

fications were performed with AmpliTaq or AmpliTaq

Gold DNA polymerase (Perkin–Elmer, Foster City,

CA). PCR was performed with a profile of 94 �C for

3min, followed by 35 cycles of 94 �C/15 s denaturation,50–55 �C/20 s annealing and 70 �C/1min extension.

In other cases, particularly amplifications of the COIgene, PCR reactions were performed directly from ge-

nomic DNA with the goby-specific primers listed in

Table 3, using the enzymes and PCR profile given

above. PCR products were run out on a low melting

point agarose gel, visualized and photographed, then cut

out and purified with the QIAquick kit. Using the same

primers (1 lM rather than 10 lM solution) the short

PCR fragments were cycle sequenced using rhodaminedye terminator/Taq FS or Big Dye terminator ready

reaction kits (Perkin–Elmer, Foster City, CA) and run

on an ABI 377XL automated sequencer. Both the heavy

and light strands were sequenced separately for each

short PCR fragment. The resultant chromatograms for

the heavy and light strands were reconciled in Sequence

Navigator (Perkin–Elmer, Foster City, CA), or Se-

quencher (Gene Codes, Ann Arbor, MI) to check

basecalling, translated to amino acid sequence using the

universal mtDNA code, and aligned by eye. There were

no ambiguities or gaps in the alignment; all the gapspresent in the final matrix were due to missing data and

Table 2 (continued)

Species Source GenBank Accession Nos.

Gobiidae: Oxudercinae

Periophthalmus barbarus Nancy Aguilar, Nigeria AF391339, AF391411, AF391483

Pseudapocryptes elongatus CAS 90433, Yangon Fish Market, Myanmar AF391394, AF391466, AF391538

Scartelaos histophorus Nancy Aguilar, Australia AF391346, AF391418, AF391490

Gobiidae: Amblyopinae

Odontamblyopus rubicundus CAS 90432, Yangon Fish Market, Myanmar AF391371, AF391443, AF391515

Gobiidae: Sicydiinae

Sicyopterus lagocephalus Quinaldine, stream, Moorea, Society Islands AF391389, AF391461, AF391533

Stiphodon elegans Brent Tibbats, Guam AF391350, AF391422, AF391494

Microdesmidae

Cerdale floridana Plankton tow, Carrie Bow Cay, Belize AF391337, AF391409, AF391481

Gunnellichthys monostigma Yuji Ikeda, Japan AF391373, AF391445, AF391517

Microdesmus bahianus Plankton tow, Carrie Bow Cay, Belize AF391347, AF391419, AF391491

Microdesmus longipinnis Richard Heard, Gulf Coast of Mississippi AF391341-2, AF391413-4, AF391485-6

Ptereleotridae

Nemateleotris magnifica Aquarium supplier AF391327-8, AF391399-1400, AF391471-2

Ptereleotris microlepis Quinaldine, Moorea, Society Islands AF391380, AF391452, AF391524

Ptereleotris monoptera Aquarium supplier AF391357, AF391429, AF391501

Ptereleotris zebra Aquarium supplier AF391358-9, AF391430-1, AF391502-3

Kraemeriidae

Kraemeria cunicularia Akihisa Iwata, Japan AF391331-2, AF391403-4, AF391475-6

Schindleriidae

Schindleria pietschmanni Plankton tow, Kaneohe Bay, Oahu AF391335, AF391407, AF391479

Schindleria praematura Plankton tow, Palmyra Atoll, Line Islands AF391336, AF391408, AF391480

Unless otherwise indicated, tissues were collected by the author and where known the collection method is indicated. CAS indicates the specimen

was from the tissue collection of the California Academy of Sciences, San Francisco; other species are uncataloged holdings of the Natural History

Museum of Los Angeles County. Species are grouped by family and subfamily, and separate GenBank accession numbers are given for each gene.

Table 3

Goby-specific primers used for amplification of ND1, ND2, and COI

genes

Primer Sequence

GOBYL3543 GCAATCCAGGTCAGTTTCTATC

GOBYH4389 AAGGGGGCYCGGTTTGTTTC

GOBYL4201 GTTGCMCAAACMATTTCHTATGAAG

GOBYH4937 GGGGTATGGGCCCGAAAGC

GOBYL4919 CCCATACCCCGAAAATGATG

GOBYH5513 GAGTAGGCTAGGATTTTWCGAAGYTG

GOBYL5464 GGTTGAGGRGGCCTMAACCARAC

GOBYH6064 CTCCTACTTAGAGCTTTGAAGGC

GOBYL6468 GCTCAGCCATTTTACCTGTG

GOBYH7127 ACYTCTGGGTGACCAAAGAATC

GOBYL7059 CCCTGCMGGTGGAGGAGACCC

GOBYH7696 AGGCCTAGGAAGTGTTGAGGGAAG

GOBYL7558 TTTGCWATTATGGCWGGATTTG

GOBYH8197 ATTATTAGGGCGTGGTCGTGG

All primers are given in the 50–30 direction.

358 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

Fig. 1. Molecular phylogeny of Gobioidei. This hypothesis is based on the complete sequence of three mitochondrial genes (ND1, ND2, and COI), a

total of 3573 bp, of which 2012 were parsimony-informative. The length is 30,268 steps, with a CI of 0.159, a RI of 0.416 and a RC of 0.066. Numbers

on nodes indicate decay index values, and roman numerals indicate clades mentioned in the text. Brackets on the right side indicate familial and

subfamilial classification: species are classified into the top grouping unless otherwise indicated with boldface or asterisks to the right of the name.

Note that in many cases, these bracketed groups are not monophyletic, they serve merely to identify the current classification of included species.

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 359

treated as such (as ? rather than a new character state) inthe analysis. A three base pair indel (AAC) was present

just prior to the stop codon at the end of the ND1 gene

in O. obscura; because this indel was present in none of

the other taxa (autapomorphic), it was removed from

the matrix for analysis rather than introducing gaps in

all other species. Aligned nucleotide sequences were

exported from Sequencher as NEXUS files.

All parsimony analyses were performed usingPAUP*, version 4.0b4a (Swofford, 1998). One thousand

replications of a heuristic search were run, using TBR

branch swapping. The data were designated as equally

weighted, following K€aallersj€oo et al. (1999) and Brough-

ton et al., 2000). Decay indices (Bremer, 1988) were

calculated with PAUP* and TreeRot v.2 (Sorenson,

1999). O. obscura was designated the outgroup taxon

and used to root the tree. As described above, mor-phological evidence indicates that this is the most

primitive of the taxa considered. The molecular data

were not partitioned into separate genes for analysis; all

data were combined in a total evidence analysis, such

that the resultant hypothesis best explains all the data

(Barrett et al., 1991; Brower, 1996; Eernisse and Kluge,

1993; Kluge, 1998; Nixon and Carpenter, 1996) and

because it has been shown that homoplasy or misleadingsignal takes a more complicated pattern than can be

represented in process partitions such as separate genes

(DeSalle and Brower, 1997; Siddall, 1997).

3. Results

Of the 3573 bp that make up the ND1, ND2, andCOI genes, most were successfully sequenced for most

taxa. Small gaps in the sequence, due to uncertainties in

reading or reconciling the chromatograms, are present

in the sequences for Acanthogobius flavimanus, Aster-

ropteryx semipunctatus, Bathygobius curacao, Cerdale

floridana, Erotelis smaragdus, Eucyclogobius newberryi,

Eviota afelei, Evorthodus minutus, Gobiosoma macrodon,

Mugilogobius sp., Priolepis eugenius, Stenogobius ha-

waiiensis, Stiphodon elongatus, and Valenciennea strig-

ata, one of the three G. thompsoni, one of the two N.

magnifica, and both of the two C. saepepallens. Larger

gaps, caused by failure to amplify one of the seven short

PCR fragments, are present in sequences for A. flavim-

anus, Amblygobius nocturnus, A. semipunctatus, Awaous

guamensis, Bathygobius cocosensis, B. curacao, Barbu-

lifer ceuthoecus, Cabillus tongarevae, Callogobius scla-

teri, Chaenogobius annularis, E. smaragdus, E. afelei,

Fusigobius neophytus, F. signipinnis, Gobiopterus semi-

vestita, G. macrodon, Gunnellichthys monostigma, Lop-

hogobius cyprinoides, Odontamblyopus rubicundus, O.

obscura, Ophieleotris aporos, Priolepis cincta, P. euge-

nius, Pseudapocryptes elongatus, Ptereleotris microlepis,

P. monoptera, Schindleria praematura, S. hawaiiensis,

and V. strigata, both specimens of C. saepepallens, K.cunicularia, P. lidwilli, and P. zebra and one of the two

specimens sequenced for A. phalaena, E. sandwicensis,

M. longipinnis, N. magnifica, and R. ruber. In no case did

a sequence have more than 40% missing data, and all

but five had less than 30% missing data. Missing data

were indicated by gaps in the data matrix and coded as

missing data (?) rather than new states.

A single most parsimonious cladogram was obtainedfrom parsimony analysis of the aligned nucleotide se-

quences (Fig. 1). This phylogeny has a length of 30,268

steps (2012 of 3573 characters were informative), con-

sistency index of 0.159, retention index of 0.416 and

rescaled consistency index of 0.066. Decay indices indi-

cate strong support for most nodes, ranging from one

for the clade containing Ptereleotridae, Schindleriidae,

G. monostigma, and Fusigobius signipinnis, to 292 be-tween species of Schindleria. Most decay index values

ranged from 4 to 53.

4. Discussion

4.1. Odontobutidae, Eleotridae, and Xenisthmidae

The molecular phylogenetic hypothesis supports the

monophyly of a large group consisting of the gobioid

families Microdesmidae, Ptereleotridae, Kraemeriidae,

Gobiidae, and Schindleriidae to the exclusion of Eleo-

tridae, Xenisthmidae and Odontobutidae (clade I in Fig.

1). Morphological character evidence concurs with this

grouping; Microdesmidae, Ptereleotridae, Kraemerii-

dae, Gobiidae, and Schindleriidae all have five bran-chiostegal rays, rather than six as seen in

Rhyacichthyidae, Eleotridae, Odontobutidae, and Xe-

nisthmidae. The five families lacking the anterior bran-

chiostegal ray are also hypothesized to be more derived

than the remaining families based on characters in-

cluding loss of the endopterygoid and dorsal postclei-

thrum, absence of infraorbitals, lack of ossification in

the scapula (in most species) and separation of theoculoscapular sensory canal into anterior and posterior

portions (Akihito, 1986; Hoese, 1984). All of these

characters have some variation in their distribution but

are mostly restricted to the five most derived families

and exemplify the typical pattern in gobies: losses and

reductions are generally found in derived taxa.

The phylogenetic hypothesis is rooted with a single

odontobutid, O. obscura, so the monophyly of Odon-tobutidae could not be assessed. Hoese and Gill (1993)

provide characters diagnosing a group consisting of all

gobioids except Odontobutidae and Rhyacichthyidae:

expansion of the procurrent cartilages anteriad to sup-

port the anterior procurrent caudal rays; scapula re-

duced, such that dorsalmost pectoral radial extends past

scapula and often extends to cleithrum; two radials

360 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

(rather than three) in the pterygiophore of the first ele-ment of the second dorsal fin; and absence of trans-

forming cteni on the scales. Within the clade of gobioids

exclusive of Odontobutidae and Rhyacichthyidae, the

family Eleotridae (subfamily Eleotridinae; no members

of Butinae were included) is paraphyletic with respect

not only to Xenisthmidae but also to the rest of the

gobioid families examined. Hoese and Gill (1993) named

the family Odontobutidae, delineated the subfamiliesEleotridinae and Butinae within Eleotridae and diag-

nosed Eleotridinae based on attachment of the adductor

mandibulae tendon on the shaft of the maxilla rather

than to a process at the anterior end, and posterior ex-

pansion of the procurrent cartilages over the tips of the

epurals. Of the eleotridines examined, not all of the

groups delineated by Birdsong et al. (1988) are mono-

phyletic. Members of the Eleotris (Eleotris and Erotelis)and Gobiomorphus (Mogurnda and Philypnodon) groups

(neither diagnosed by a synapomorphy) do not group

together. Miller (1998) synonomized Eleotris and Erot-

elis, based on several morphological characters the two

share; they differ morphologically only in scale size, but

in the molecular hypothesis a group containing both

genera would be paraphyletic. O. aporos is the only

member of the Dormitator group examined; this groupdiagnosed by a strongly recurved first hemal spine that

almost touches the second, and in the molecular hy-

pothesis, is the sister taxon to the Gobiomorphus group

member Mogurnda. There is especially strong support

(decay index value of 101) for a monophyletic genus

Hypseleotris. The Hypseleotris group, containing Hyps-

eleotris and Hemieleotris, is distinguished by possessing

a cyprinid-like body shape with an elongate body cavityand a high number (8–11) of anal pterygiophores pre-

ceding the first hemal spine (Birdsong et al., 1988); in the

molecular hypothesis this group is sister to the other

Gobiomorphus group member, Philypnodon.

The only included member of the family Xenisthmi-

dae examined, Xenisthmus sp., is nested within the pa-

raphyletic Eleotridinae, sister to the pair of species

Mogurnda adspersa and O. aporos. The family Xe-nisthmidae comprises one of Birdsong et al.�s (1988)

groups, the Xenisthmus group, diagnosed by several

characters: an ossified rostral cartilage; ventral lip with

free ventral margin extending across dentary symphysis;

ascending process of premaxilla greatly reduced or ab-

sent; and basibranchial two absent (Springer, 1983,

1988). All but the first two characters are reductive

features for xenisthmids; in addition, some xenisthmidshave also lost the pterosphenoids and basibranchials 3

and 4, and in the miniature Tyson (20mm standard

length or less) the spinous dorsal fin, extrascapulars,

lacrimal, exoccipital condyles, infrapharyngobranchials

2 and 4, gill rakers and scales are also absent (Gill and

Hoese, 1993; Springer, 1983, 1988). Xenisthmidae is one

example of reduction among gobioids, exhibiting a

mosaic of reductive and non-reductive morphologicalcharacters.

Akihito et al. (2000) performed a molecular phylo-

genetic analysis, in which sampling was concentrated in

Eleotridae (including both Eleotridinae and Butinae of

Hoese and Gill, 1993), but which also included repre-

sentatives of Xenisthmidae, Odontobutidae, Gobiidae,

Kraemeriidae, Microdesmidae, and Ptereleotridae.

Their analysis included the 1140 bp of the mitochondrialcytochrome b gene, and was not cladistic; instead, they

produced unrooted networks using both neighbor-join-

ing and maximum likelihood methods. They did not

consider the relationships of each species as revealed in

their analysis, rather subdividing their sampled taxa into

six ‘‘clusters,’’ each containing two to eight species, plus

the pair O. obscura and Xenisthmus sp. The results

presented in their trees agree with the current analysis insome respects, including that Eleotridae is paraphyletic

andMogurnda and Ophieleotris are closely related. Their

hypotheses disagree with this one in the placement of

Xenisthmus: it is sister taxon to Odontobutis in their

hypothesis, within Eleotridae here. Wang et al.�s (2001)molecular hypothesis, based on cladistic analysis of

1078 bp of the mitochondrial 12S and tRNAVAL genes,

shows a monophyletic Eleotridinae. Within it, theirhypothesis agrees with this one in some respects, in-

cluding a monophyletic Hypseleotris, and a sister taxon

relationship between Mogurnda and Ophieleotris. How-

ever, Wang et al.�s (2001) hypothesis differs from this

one in the relationships among the genera Hypseleotris,

Eleotris, and Philypnodon. In their hypothesis, Hypsel-

eotris is sister to the pair Eleotris+Philypnodon, unlike

this hypothesis which indicates that Hypseleotris andPhilypnodon are sisters, to the exclusion of Eleotris.

4.2. Gobionellinae, Kraemeriidae, Sicydiinae, Oxuderci-

nae, and Amblyopinae

The molecular phylogeny includes a clade consisting

of Gobionellinae, Kraemeriidae, Sicydiinae, Oxuderci-

nae, and Amblyopinae (clade II in Fig. 1; here this cladeis referred to as the ‘‘expanded monophyletic gobionel-

line clade’’ or ‘‘expanded monophyletic Gobionellinae’’;

when the term Gobionellinae is used alone it is sensu

Pezold, 1993). Within the expanded monophyletic

gobionelline clade, two smaller clades are present: one

containing both Mugilogobius species, Gillichthys mira-

bilis, Typhlogobius californiensis, C. annularis, E. new-

berryi, A. flavimanus, G. semivestita, P. lidwilli, and thekraemeriid K. cunicularia (clade IIA in Fig. 1). In Lar-

son�s (2001) revision of Mugiligobius and evaluation of

relationships among selected gobionelline genera, all of

these species are placed in her ‘‘Mugilogobius clade’’

except Acanthogobius (incertae sedis within Gobionelli-

nae) and Kraemeria (not examined) Sister to this clade is

one (clade IIB in Fig. 1) containing three smaller clades,

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 361

one including A. guamensis, S. hawaiiensis, and the sic-ydiines Stiphodon elegans and Sicyopterus lagocephalus,

and sister to that group a clade containing the three

Gnatholepis species examined, G. thompsoni, G. scapu-

lostigma, and G. cauerensis, as well as C. saepepallens

and E. minutus. A close relationship between the sic-

ydiines, Awaous and Stenogobius has been postulated

previously; they have been included together in Harri-

son�s (1989) ‘‘Ctenogobius lineage,’’ and in the ‘‘Ste-nogobius clade’’ of Larson (2001). Awaous and

Stenogobius were also shown to be closely related to

Sicydiinae by Parenti and Thomas (1998). Sister to both

these clades is a clade including the amblyopine O. ru-

bicundus, and the oxudercines Scartelaos histophorus, P.

elongatus, and Periophthalmus barbarus. These genera

are all included in Harrison�s (1989) ‘‘Oxyurichthys

lineage’’; Murdy (1989) and Murdy and Shibukawa(2001) also indicated that Oxudercinae and Amblyopi-

nae are probably closely related.

Within clade IIA, the Gobionellus group member

Mugilogobius is basal to representatives of a mix of

Birdsong et al.�s (1988) Chasmichthys, Gobiopterus,

Astrabe, and Acanthogobius groups and Krameriidae

(Kraemeria group). Three of the species (G. mirabilis, C.

annularis, and E. newberryi) are members of the Chas-

michthys group, a group diagnosed by insertion of the

first dorsal spine into interneural space 4 or 5, and

whose members also feature high vertebral counts

ð13� 17þ 18� 22 ¼ 32� 38Þ and a temperate north-

ern Pacific distribution. In this hypothesis the Chas-

michthys group is paraphyletic with respect to the

Astrabe (T. californiensis) group; Astrabe group genera

share reduced eyes and posterior displacement or loss ofthe spinous dorsal fin, and are distributed in the same

regions as the Chasmichthys group. The Acanthogobius

group (A. flavimanus), is sister to the Gobiopterus group

(P. lidwilli and G. semivestita), which is itself paraphy-

letic with respect to Kraemeriidae. Acanthogobius group

genera share a unique dorsal fin pterygiophore pattern:

3-1221110, and are found in the temperate western Pa-

cific. The Gobiopterus group is not diagnosed butmembers share a 10þ 15 ¼ 25 vertebral count and are

restricted to the Indian Ocean and Indo-Pacific regions

(the species sequenced, G. semivestita, is a temperate

Australian estuarine goby). Thus, the four groups are all

distributed in the temperate margins of the Pacific and

Indian Oceans. Kraemeriidae is also distributed

throughout the western Pacific. Kraemeriids attain a

maximum length of 40mm and exhibit reduced char-acters such as three pectoral radials (rather than four), a

single epural, fusion of all the hypurals into a single

plate and all skeletal elements slender and weakly ossi-

fied (Matsubara and Iwai, 1959). With the exception of

the reduction in pectoral radials, all of these reductive

characters are present in other gobioids. Akihito et al.�s(2000) molecular analysis placed Kraemeria in a cluster

with the microdesmid Gunnellichthys and the ptereleo-trid Ptereleotris. The disagreement between Akihito et

al.�s (2000) hypothesis and this one is probably due to

sampling: they did not include any gobiine gobiids in

their hypothesis.

The other clade within the expanded monophyletic

Gobionellinae, clade IIB, includes the gobioid subfam-

ilies Sicydiinae, Oxudercinae and Amblyopinae. Within

clade IIB is a clade containing the Gobionellus groupmembers Gnatholepis and Ctenogobius in addition to

Evorthodus, a genus not classified by Birdsong et al.

(1988), but there indicated to possibly be related to

Gobionellus group genera. Awaous, Stenogobius, and the

sicydiines S. elegans and S. lagocephalus are also in-

cluded in this clade; Stenogobius is in the Gobionellus

group, and the remaining three genera are included in

Birdsong et al.�s (1988) Sicydium group. The Gobionellusgroup contains ten genera, phenetically united by a

combination of dorsal fin, vertebral and caudal fin

characters, and is distributed broadly through the tro-

pics and subtropics. Birdsong et al. (1988) indicate that

although the group contains marine representatives,

most members are found in estuarine or freshwater.

Morphological evidence for the close relationship of

Sicydiinae and Gobionellus group genera is found in thepalatopterygoquadrate complex in the suspensorium, as

described by Harrison (1989). Harrison describes several

apomorphic conditions of the palatine, ectopterygoid

and quadrate; Awaous, Stiphodon, Sicyopterus, Ste-

nogobius, Gnatholepis, Evorthodus, and Ctenogobius are

all part of a group characterized by a long palatine,

which extends towards or meets the quadrate. Awaous,

Stiphodon, and Sicyopterus additionally share similari-ties in external morphology and skeletal characters in-

cluding a spatulate posterior process on the pelvis.

Sicydiinae is diagnosed by several morphological char-

acters: palatine bone with long dorsal process that ar-

ticulates with the lateral ethmoid, no differentiation

between articular and ascending processes on the pre-

maxilla, tongue fused to floor of mouth, thick, branched

pelvic rays; pads at the tips of the pelvic spines, and theproximal ends of the pelvic spine and first pelvic ray

close together, and separated by a gap from the other

pelvic rays; in all genera except Sicyopus the upper jaw

teeth are tricuspid and found in several rows (Harrison,

1989; Hoese, 1984; Parenti and Maciolek, 1993). The

two sicydiines considered in this analysis are sister taxa,

and Sicydiinae is sister to the pair S. hawaiiensis and A.

guamensis. The relationship of Awaous to the sicydiineshas been debated: Harrison (1989) considers Awaous to

be the sister taxon to Sicydiinae based on the presence of

the spatulate posterior pelvic process, the lack of an

ossified scapula, a long palatine, a dorsal fin pterygio-

phore pattern of 3-12210, a single epural, and similar

head neuromast patterns. All but the spatulate pelvic

process are found in other goby groups; the spatulate

362 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

pelvic process is also seen in the gobionelline Tukugo-

bius, and although Birdsong et al., 1988) place Awaous

in the Sicydium group, they mention that one author

believes Awaous to be closely related to the Gobionellus

group (including Mugilogobius, Stenogobius, Gnathol-

epis, and Ctenogobius). Larson�s (2001) hypothesis con-curs with this one in that Awaous and Stenogobius are

sister taxa, as well as Evorthodus and Gnatholepis (she

did not consider Ctenogobius or the sicydiines). Parentiand Thomas (1998)�s cladistic analysis of morphology

indicates the the sister taxa to Sicydiinae are Tukugobius

and Rhinogobius, and the next most proximal sister taxa

are a trichotomy of Awaous, Gnatholepis, and Stenogo-

bius, followed by Evorthodus and the oxudercine

Pseudapocryptes. Tukugobius and Rhinogobius were not

included in this molecular analysis, but overall, the re-

sults accord well with Parenti and Thomas (1998), andin broad respects with Harrison (1989). Unlike Harrison

(1989), the molecular phylogeny indicates that Ste-

nogobius is included in the clade with Awaous and the

sicydiines, and there are also differences between inter-

pretations of the placement of Harrison�s (1989) �Cte-nogobius lineage.� This molecular analysis concurs with

previous morphological studies (Harrison, 1989; Parenti

and Thomas, 1998) in the conclusion that Gobionellinaeis paraphyletic with respect to Sicydiinae.

Gnatholepis and Ctenogobius are part of the �Cte-nogobius lineage� of Harrison (1989), and additionally

share anteroposteriorly elongate quadrate lamina as well

as the elongate palatine; the molecular hypothesis indi-

cates that these genera are closely related, specifically

that Gnatholepis is sister to Ctenogobius plus Evorthodus.

Harrison (1989) also indicates that Gnatholepis, Cte-

nogobius, Evorthodus, and Stenogobius share a similar

arrangement of suborbital neuromasts. The molecular

hypothesis differs from the hypothesis presented by

Harrison (1989), in which Stenogobius is the primitive

sister taxon to a clade containing the �Ctenogobius line-age� and his �Oxyurichthys lineage,� and Awaous and the

Sicydiinae are sister to that clade. Instead, the molecular

data indicate that the �Ctenogobius lineage� genera aresister to a Stenogobius/Awaous/Sicydiine clade. The

disagreement may be due in large part to rooting. If

Harrison�s (1989) hypothesis is rooted in the same way

as the expanded monophyletic gobionelline clade (clade

II in Fig. 1), between the �Ctenogobius lineage� and

�Oxyurichthys lineage,� the results are in agreement with

the molecular phylogeny, with one small exception:

Stenogobius would be sister to Awaous+Sicydiinae inHarrison�s (1989) hypothesis, but Stenogobius+Awaous

is sister to Sicydiinae in the molecular hypothesis.

Harrison�s (1989) �Oxyurichthys lineage� includes the

gobionelline genus Oxyurichthys, and the subfamilies

Oxudercinae and Amblyopinae; this group shares the

presence of a very short, stubby palatine, and he hy-

pothesizes that it is the sister to the �Ctenogobius lineage.�

His hypothesis requires that the long palatine is reversedin the �Oxyurichthys lineage�; in the molecular hypothe-

sis, the taxa with long palatines are closely related, to the

exclusion of the �Oxyurichthys lineage� taxa, implying

that the short palatine was derived independently and

not secondarily lost. In the molecular hypothesis, Ox-

udercinae is paraphyletic with respect to Amblyopinae.

In addition to the short palatine configuration de-

scribed by Harrison (1989), Amblyopinae and Oxud-ercinae share a tongue fused to the floor of the mouth

(also seen in Sicydiinae but considered a homoplasy by

Parenti and Maciolek (1993)) and elongation of the

frontal bones (Hoese, 1984; Murdy, 1989). Amblyopines

are elongate, burrowing fishes found in estuaries and

river mouths with extremely reduced, dorsally placed

eyes. Oxudercines are commonly known as mudskip-

pers; they inhabit soft bottomed and mangrove swamphabitat in the Indo-Pacific and West Africa and many

species are capable of aerial respiration and terrestrial

locomotion. A cladistic hypothesis of relationships has

been presented for Oxudercinae (Murdy, 1989). In

Murdy�s hypothesis Oxudercinae is diagnosed by five

characters including a complex arrangement of the

dorsal neurocranial bones and eyes (including the large

lateral sphenotic process and anterodorsally placed eyes,characters used to diagnose Oxudercinae by Hoese,

1984), extension of the anterior nostril into a flap that

overlaps the upper jaw, venteroposterior process of

palatine greatly reduced (this describes the same condi-

tion that Harrison (1989) calls a short, stubby, palatine),

reduced and vertically oriented ascending processes of

the premaxilla, and a single (or rarely two) anal-fin

pterygiophore anterior to the first hemal spine (thischaracter is not unique to Oxudercinae). Part of the

complex neurocranial character is the elongation of the

frontal bones that is observed to a lesser extent in Am-

blyopinae, and two amblyopine genera (Brachamblyopus

and Trypauchen) share another of the diagnostic oxu-

dercine characers, the reduction of a venteroposteriorly

directed process on the palatine that overlaps or joins

the ectopterygoid, as well as a similar dorsal fin ptery-giophore formula. However, Murdy did not propose a

close relationship between Oxudercinae and Amblyopi-

nae; instead, he noted similarities between Oxudercinae,

Sicydiinae and several gobionelline genera including

Ctenogobius, Gnatholepis, Mugilogobius, Oxyurichthys,

and Evorthodus. In contrast to Murdy�s analysis, the

molecular hypothesis indicates that Oxudercinae is pa-

raphyletic with respect to Amblyopinae.The molecular hypothesis also disagrees with Murdy�s

placement of genera within Oxudercinae. Murdy gives

the following relationship: (Pseudapocryptes (Scartelaos

and Periophthalmus)), an arrangement which is not

congruent with the molecular hypothesis placement of

Periophthalmus plus Pseudapocryptes as sister to the pair

Scartelaos plus the amblyopine Odontamblyopus. As

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 363

with the differences between the Harrison (1989) hy-pothesis and this one, the disagreement may partially be

attributed to a different rooting of the oxudercine clade.

Murdy (1989) postulated that Evorthodus was most clo-

sely related to Oxudercinae based on three characters of

the teeth, branchial apparatus and the retractor dorsalis

muscle; Evorthodus was used as the proximal outgroup,

along with the sicydiines Sicydium and Stiphodon, the

Gobionellus group gobionellines Gnatholepis, Ctenogo-

bius, Mugilogobius, and Oxyurichthys, and the amblyo-

pines Trypauchen, Brachamblyopus, and Gobioides.

Oxyurichthys was not examined in this study, and the

sampling within Sicydiinae and Amblyopinae differs, but

the molecular phylogenetic hypothesis supports a sister

taxon relationship between Oxudercinae +Amblyopinae

and a clade containing Gnatholepis, Ctenogobius, Ev-

orthodus, Awaous, Stenogobius and the sicydiines Sicy-

opterus and Stiphodon. The different relationships within

Oxudercinae obtained in the molecular hypothesis do

not substantially change interpretation of the evolution

of some ofMurdy�s (1989) characters. Murdy interpreted

the reduction in the ascending process of the premaxilla

(his character 4) as being primitively absent (process not

reduced), then present (reduced) in Pseudapocryptes and

Scartelaos, then secondarily lost (process regained) inPeriophthalmus. The molecular hypothesis supports

similar interpretation: the process is primitively present,

then reduced in Pseudapocryptes and Scartelaos, and

retained or regained in Periophthalmus. Similarly, Murdy

(1989) discussed characters of the branchial apparatus

and jaws (large, lattice-like fifth ceratobranchials, dorsal

rather than lateral articulation of the epibranchials to the

infrapharyngobranchials and large, recurved canineteeth internal to the symphysis of the lower jaw) that are

not coded in his phylogenetic analysis but are shared by

Evorthodus and all Oxudercinae except Periophthalmus

and Periophthalmodon. He interpreted retention of the

more generalized state in Periophthalmus and Perioph-

thalmodon as a reversal to the primitive condition; in the

molecular hypothesis the optimization of these charac-

ters is more ambiguous and difficult to assess withoutdenser sampling within Oxudercinae, but suggests that

the conditions in Evorthodus and most oxudercines are

independently derived. Thus, the molecular hypothesis

supports Murdy�s (1989) conjecture that the similarities

in branchial structure and dentition between Evorthodus

and most Oxudercines may be due to convergence; both

occupy soft bottomed habitats and may use the teeth for

burrowing and the complex branchial structures forstraining out ingested substrate.

Some of Murdy�s characters obtain a less parsimo-

nious interpretation on the molecular phylogeny. Peri-

ophthalmus and Scartelaos are both amphibious and

also share a character of the metapterygoid, a dermal

cup that functions as a moisture reservoir for the eyes,

and separate dorsal fins (his characters 24–27). The

molecular hypotheses indicates that these characterswere either derived independently in Periophthalmus and

Scartelaos, or primitively present and lost in Pseudapo-

cryptes and the amblyopine Odontamblyopus. Murdy

(1989) also pointed out that Oxudercinae shares with

both the Gobionellus group and the Sicydium group the

same vertebral number ð10þ 16 ¼ 26Þ and dorsal fin

formula (3-12210). In the molecular hypothesis these

characters diagnose the expanded monophyletic gobio-nelline clade, and are altered in clade IIA exclusive of

Mugilogobius. The Acanthogobius, Astrabe, and Chas-

michthys group genera in clade IIA (Acanthogobius,

Eucyclogobius, Chaenogobius, Gillichthys, and Typhlo-

gobius) are somewhat to very elongate, with elevated

vertebral counts and often with posterior displacement

of the dorsal fin. The Gobiopterus group genera (Gobi-

opterus and Pandaka) and Kraemeria have similar, butnot identical, dorsal fin and vertebral characters as

compared to Oxudercinae, Sicydiinae, and the Gobio-

nellus group. The major conclusion to be drawn from

the analysis of Oxudercinae and Amblyopinae in this

analysis is that Amblyopinae is nested within Oxud-

ercinae and both are within a paraphyletic Gobionelli-

nae. Akihito et al.�s (2000) molecular analysis also

supported a close relationship between Oxudercinae,Amblyopinae, within a paraphyletic Gobionellinae, but

their sampling in these groups (single representatives of

both Oxudercinae and Amblyopinae) is not dense en-

ough to address the question of Oxudercine paraphyly.

Wang et al.�s (2001) molecular hypothesis agrees well

with this one: they hypothesize that the Gobiopterus

group genera are sister to Oxudercinae, which is sister to

the pair Sicydiinae plus Stenogobius.

4.3. Gobiinae, Microdesmidae, Ptereleotridae, and

Schindleriidae

The molecular phylogeny indicates an expanded

monophyletic Gobiinae, including Microdesmidae,

Ptereleotridae, and Schindleriidae (clade III in Fig. 1;

here this clade is referred to as the ‘‘expanded mono-phyletic gobiine clade’’ or ‘‘expanded monophyletic

Gobiinae’’; when the term Gobiinae is used alone it is

sensu Pezold, 1993). Monophyly of Gobiinae is sup-

ported by the presence of a single anterior interorbital

pore (rather than a pair of pores) and a single epural;

most gobiines also share a dorsal fin pterygiophore

pattern of 3-22110 or 3-221100 (Pezold, 1993). Two

epurals are found in most gobionellines, oxudercinesand amblyopines, but kraemeriids and most sicydiines

have only one. Members of Microdesmidae and Pterel-

eotridae have single epurals; in Schindleriidae epurals

are absent. Dorsal fin pterygiophore patterns vary

among these three families, and only in some Ptereleo-

tridae (Parioglossus group) is the gobiine pattern of 3-

22110 found (Birdsong et al., 1988). The single anterior

364 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

interorbital pore characteristic of gobiines is found onlyin some ptereleotrine genera; other ptereleotrines have a

pair of pores (Klausewitz and Conde, 1981; Randall and

Allen, 1973; Randall and Hoese, 1985; Rennis and

Hoese, 1985, 1987). Microdesmidae and Schindleriidae

lack head sensory pores entirely.

Most of the gobiine genera sampled for the molecular

phylogeny are part of Birdsong et al.�s (1988) Priolepis

group. The Priolepis group is a large assemblage (54genera) of reef-dwelling gobies that primarily inhabit the

Indo-Pacific but are also found in the eastern Pacific,

Caribbean, and Atlantic. This group is diverse and

phenetically united by the dorsal fin formula of 3-22110,

one epural, two anal-fin pterygophores preceding the

first hemal spine, and vertebral counts of 10þ 16 ¼ 26.

Priolepis group genera included in this hypothesis are

Amblyeleotris, Amblygobius, Asterropteryx, Cabillus,Callogobius, Coryphopterus, Ctenogobiops, Eviota, Fusi-

gobius, Gobiodon, Lophogobius, Priolepis, and Valen-

ciennea. Other gobiid groups included in the molecular

phylogeny are Bathygobius (genus Bathygobius) and

Gobiosoma (genera Barbulifer, Gobiosoma, and Risor);

along with the Priolepis group, these groups include

most gobiine genera (the Gobius, Kellogella, Microgo-

bius, and Pomatoschistus groups are not considered;these four groups include eleven genera). Four Coryph-

opterus species and two species each of the genera

Amblygobius, Bathygobius, Fusigobius, and Priolepis are

included, and in all cases except Fusigobius the genera

are monophyletic. The molecular phylogeny indicates

that the Priolepis group is paraphyletic with respect to

Bathygobius group, Gobiosoma group, and the families

Microdesmidae, Ptereleotridae, and Schindleriidae.The most basal clade in the expanded Gobiinae (clade

IIIA of Fig. 1) includes Priolepis, Cabillus, and Bathy-

gobius. The Bathygobius group differs from the Priolepis

group in vertebral number (Bathygobius group genera

usually have one more caudal vertebra, for a count of

10þ 17 ¼ 27). Bathygobius is widely distributed in

tropical waters; one old world (B. cocosensis) and one

new world (B. curacao) species are included in the mo-lecular hypothesis. A close relationship between Bathy-

gobius and Cabillus is additionally supported by the

hypothesis of Gill (1994); both genera share a morpho-

logical character, presence of paired lateral protuber-

ances near the anterior nostrils.

Two large clades comprise the rest of the expanded

monophyletic Gobiinae: one containing several Priolepis

group genera as well as Ptereleotridae, Schindleriidaeand some Microdesmidae (clade IIIB of Fig. 1), and a

second including the remainder of Microdesmidae, the

Gobiosoma group genera and two Priolepis group gen-

era, Coryphopterus and Lophogobius (clade IIIC of Fig.

1). Interestingly, these clades differ in the geographical

distribution of their members: clade IIIB includes genera

found in the old world, and clade IIIC genera are all new

world (except Priolepis, which is also found in the At-lantic; the only other new world representative in the

expanded monophyletic Gobiinae is one of the Bathy-

gobius species, found in clade IIIA).

Within the expanded monophyletic gobiine clade,

Ptereleotridae is nested within clade IIIB and Microde-

smidae is split between clade IIIB and IIIC. These

groups have been previously placed as subfamilies in the

same family (Hoese, 1984), a grouping which neitherthis molecular analysis nor morphological phylogeny

(Thacker, 2000) supports. The character previously used

to unite these groups is the presence of an elongate,

posterior process on the pelvis; such a process is present

in Ptereleotridae but not Microdesmidae. Other char-

acters used to diagnose a Ptereleotridae +Microdesmi-

dae clade, including unfused pelvic fins, lateral

compression of the head and body, a single epural andreduction of the articulation between the palatine and

lateral ethmoid, are widely distributed and plesiomor-

phic among gobioids (Thacker, 2000). The molecular

phylogeny indicates not only that Microdesmidae and

Ptereleotridae are not sister taxa, but also that neither

family is monophyletic. A nonmonophyletic Ptereleo-

tridae is a result that conflicts with previous morpho-

logical analyses. Rennis and Hoese (1987) providedseveral diagnostic characters for the family: the elongate

pelvic process, a single pterygiophore preceding the first

hemal spine, fused premaxillary processes and separate

dorsal fins (the latter two are not unique to Ptereleo-

tridae). Morphological characters suggest that Nemate-

leotris is the most primitive ptereleotrid genus and

Ptereleotris the most derived (Rennis and Hoese, 1987).

The molecular hypothesis does indicate a monophyleticPtereleotris. Nemateleotris and Ptereleotris are included

in two different groups (Parioglossus and Ptereleotris,

respectively) by Birdsong et al. (1988). These groups

differ in dorsal-fin pterygiophore formula: Parioglossus

group genera have the common gobiid condition of 3-

22110, while Ptereleotris features the unique 3-32010.

The hypothesis of a nonmonophyletic Microdesmidae

also conflicts with a previous morphological phyloge-netic analysis (Thacker, 2000), and, unlike the dis-

agreements between the molecular phylogeny and

morphology-based hypotheses of relationships for Ox-

udercinae, Sicydiinae, and Gobionellinae, this disagree-

ment cannot be explained by a change in rooting. Three

of the five microdesmid genera were included in this

study. One, the Indo-Pacific Gunnellichthys, is included

in clade IIIB; the other two, the new world Cerdale andMicrodesmus, are placed in clade IIIC. Two species of

Microdesmus are included and they are recovered

together, sister to Cerdale, but only distantly related to

Gunnellichthys. The morphological characters used to

diagnose Microdesmidae are: the presence of an anterior

projection on the maxilla, overlapping the premaxillary

processes; loss of the medial process of the palatine that

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 365

articulates with the lateral ethmoid; presence of a tiny,slender pelvis, with pelvic intercleithral cartilage deeply

cleft; presence of a continuous dorsal fin, rather than

separate spined and rayed portions; an elevated vertebral

number and an elongate body. Of these characters, two

are unique novelties (maxilla projection and pelvis

morphology), one is a loss (loss of palatine process), and

three are not unique to Microdesmidae (single dorsal fin,

body elongation and elevated vertebral number). It ispossible that these characters are all functionally asso-

ciated with the burrowing and feeding (egg predation)

habits of these fishes and thus the result of convergence.

The sister taxon to Microdesmidae is not known.

Most previous hypotheses have been based on a Mic-

rodesmidae+Ptereleotridae clade, with Ptereleotridae

being primitive. Based on several characters of the pal-

atopterygoquadrate complex, Harrison (1989) proposedthat Microdesmidae is distinct from Ptereleotridae and

is the sister taxon to a clade containing the Sicydiinae

and some Gobionellinae. In this hypothesis, Microde-

smidae, Sicydiinae, and Gobionellinae are not closely

related. Rather, the new world microdesmids (Cerdale

and Microdesmus) are sister to the Gobiosoma group

gobiines and the new world Priolepis group gobiines.

The old world Gunnellichthys is found in a clade con-taining Ptereleotridae, old world Priolepis group gobi-

ines and Schindleriidae.

The molecular topology indicates that Schindleria is

sister to Gunnellichthys. This grouping is particularly

interesting, since the schindleriids have only recently

been classified within Gobioidei (Johnson and Brothers,

1993) and present a particularly difficult problem for

traditional morphological systematic studies because oftheir extreme paedomorphic reduction. Schindleriidae

contains two species, S. praematura and Schindleria

pietschmanni, both of which were sequenced for this

study. Schindleria adults resemble larvae in all respects

except for gonad maturation, which occurs at 11–15mm

standard length. Johnson and Brothers (1993) were able

to find morphological characters that indicated Schin-

dleria was related to the Gobioidei, including similaritiesbetween Schindleria and gobioid larvae, but were not

able to determine its sister group. Several characters

indicated that Schindleria was related to the more de-

rived gobies, but the majority of these characters were

reductions or losses that could be independantly derived

as a result of ontogenetic truncation. One character is

shared by Schindleria and Microdesmidae: a configura-

tion of the pharyngobranchials in which the second liesfully anterior to the third and articulates with it only at

the tip (Johnson and Brothers, 1993). This articulation

condition is also found in Xenisthmus, a taxon distantly

related to these taxa according to the molecular topol-

ogy, which has a very elongate and modified second

pharyngobranchial. The molecular phylogeny supports

the morphological character evidence of Johnson and

Brothers (1993); further evidence for the placement ofSchindleria with Microdesmidae is the observation that

the larvae of Gunnellichthys are superficially very similar

to Schindleria. Both are elongate, with a continuous

dorsal fin, a pointed snout, large eyes and a similar

overall morphology (Thacker, pers. obs.).

In addition to the new world microdesmid genera, the

new world gobiine clade (clade IIIC in Fig. 1) includes

three Gobiosoma group genera (Barbulifer, Risor, andGobiosoma) and two Priolepis group genera (Coryph-

opterus and Lophogobius). Together, the Gobiosoma

group, Microgobius group, and the genus Ophiogobius

comprise the tribe Gobiosomini: the American seven-

spined gobies (Birdsong, 1975). As originally proposed,

Gobiosomini includes the genera Aruma, Barbulifer,

Bollmannia, Chriolepis, Eleotrica, Enypnias, Everman-

nichthys, Ginsburgellus, Gobiosoma, Gobulus, Gymneleo-

tris, Microgobius, Nes, Palatogobius, Pariah, Parrella,

Psilotris, Pycnomma, Risor, and Varicus. These genera

share a vertebral formula of 11þ 16� 17 ¼ 27� 28 and

a dorsal fin pterygiophore pattern of 3-221110, (there is

some variation within the genus Evermannichthys) and

comprise most of the gobioid fauna of the tropical

eastern Pacific, western Atlantic and Caribbean. All the

genera except Microgobius, Parrella, Bollmannia, andPalatogobius share a specialization of the caudal fin in

which the two hypural elements (composing fused hyp-

urals 1–2 and 3–4) are fused to each other and to the

terminal vertebral element. On the basis of this caudal

character, Birdsong et al. (1988) delineated those genera

as the Gobiosoma group and placed Microgobius, Par-

rella, Bollmannia, and Palatogobius in another group, the

Microgobius group. Members of the Gobiosoma grouppresent in this study are B. ceuthoecus, R. ruber, and G.

macrodon. These genera form a clade, sister to the genera

Coryphopterus and Lophogobius. Thacker and Cole

(2002) examined the phylogeny of Coryphopterus and

outgroups based on both molecular and morphological

data. Their analysis agrees with the results seen in this

molecular hypothesis: the four Coryphopterus species

examined have the same relationships as seen in Thackerand Cole (2002), Lophogobius is sister to Coryphopterus,

and a nonmonophyletic Fusigobius is more distantly re-

lated. In addition to molecular characters, morphologi-

cal characters congruent with the sister taxon

relationship of Coryphopterus and Lophogobius include

the presence of a fleshy ridge or crest on the dorsal sur-

face of the head (also seen in Rhinogobiops nicholsii), and

a similar gonad structure, related to their protogynoushermaphroditism (also seen in Fusigobius [Cole, 1988]).

5. Conclusions

This analysis provides a broad view of gobioid in-

terrelationships that has been impossible to reconstruct

366 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368

with morphological data. The molecular hypothesisprovides broad taxonomic sampling, using a character

set that may be examined in all the taxa. It accords

generally well with many disparate previous phyloge-

netic studies, both those based on morphological char-

acters and those based on smaller molecular data sets.

Overall, the molecular phylogeny reveals not only that

the larger gobioid groups (Eleotridae, Gobiinae, and

Gobionellinae) are paraphyletic with respect to thesmaller ones (Xenisthmidae, Oxudercinae, Amblyopi-

nae, Sicydiinae, Ptereleotridae, Microdesmidae, Kra-

emeriidae, and Schindleriidae), but also provides a

framework for an interpretation of morphological

character evolution, and in particular reductive evolu-

tion, in this group. The pattern of reduction and sim-

plification observed in gobies is confirmed as a general

trend: more derived taxa exhibit greater morphologicalreduction. However, specific instances of drastic reduc-

tion such as those seen in Xenisthmidae, Schindleriidae,

Kraemeriidae and some Gobiidae such as Pandaka

(subfamily Gobionellinae) and Priolepis (subfamily

Gobiinae) are manifested differently and are indepen-

dently derived, indicating that reduction is a recurrent

phenomenon among gobies.

Acknowledgments

This study would not have been possible without the

generosity of those who supplied goby tissues for DNA

sequencing: Nancy Aguilar, Kathleen Cole, Tony Gill,

David Greenfield, Richard Heard, Yuki Ikeda, Akihisa

Iwata, Scott Matern, Rob Reavis, Colette St. Mary HoYoung Suk, Brent Tibbatts, Jim Van Tassell, and Peter

Unmack. I also thank Dave Catania and Ramona

Swenson for collecting and curating ethanol-preserved

gobies in the collection of the California Academy of

Sciences. Field collections in Hawaii and the Line Islands

were assisted by Bret Danilowicz, Shawn Doan, Sharon

Kobayashi, Theresa Martinelli, and Bruce Mundy. Field

collections in French Polynesia were assisted by AndrewThompson and Daniel Geiger. Field collections in Belize

were assisted by David Smith, Carole Baldwin, and

Kathleen Cole. I thank Michael D. Sorenson, David P.

Mindell, Jennifer Ast, David Kizirian, Derek Dimcheff,

Stacie Novakovic, Alec Lindsay, and Tamaki Yuri for

expert instruction in and assistance with amplification,

sequencing, and molecular data analysis techniques. Part

of the work for this project was performed at the Uni-versity of Michigan Museum of Zoology, in the lab of

David Mindell. This work was supported in part by the

Carl L. and Laura Hubbs Fellowship. This is contribu-

tion No. 3 of the W.M. Keck Foundation Program in

Molecular Systematics and Evolution at the Natural

History Museum of Los Angeles County.

References

Akihito, P., 1986. Some morphological characters considered to be

important in gobiid phylogeny. In: Uyeno, T., Arai, R., Taniuchi,

T., Matsuura, K. (Eds.), Proceedings of the Second International

Conference on Indo-Pacific Fishes. The Ichthyological Society of

Japan, Tokyo, pp. 629–639.

Akihito, Iwata, A., Kobayashi, T., Ikeo, K., Imanishi, T., Ono, H.,

Umehara, Y., Hamamatsu, C., Sugiyama, K., Ikeda, Y., Sakam-

oto, K., Fumihito, A., Ohno, S., Gojobori, T., 2000. Evolutionary

aspects of gobioid fishes based upon a phylogenetic analysis of

mitochondrial cytochrome b genes. Gene 259, 5–15.

Barrett, M., Donoghue, M.J., Sober, E., 1991. Against consensus. Syst.

Biol. 40 (4), 486–493.

Birdsong, R.S., 1975. The osteology of Microgobius signatus Poey

(Pisces: Gobiidae), with comments on other gobioid fishes. Bull.

Florida State Mus., Biol. Sci. 19 (3), 135–187.

Birdsong, R.S., Murdy, E.O., Pezold, F.L., 1988. A study of the

vertebral column and median fin osteology in gobioid fishes with

comments on gobioid relationships. Bull. Mar. Sci. 42 (2), 174–214.

Block, B.A., Finnerty, J.R., Stewart, A.F.R., Kidd, J., 1993. Evolution

of endothermy in fish: mapping physiological traits on a molecular

phylogeny. Science 260, 210–214.

Bremer, K., 1988. The limits of amino acid sequence data in

angiosperm phylogenetic reconstruction. Evolution 42, 795–803.

Broughton, R.E., Stanley, S.E., Durrett, R.T., 2000. Quantification of

homoplasy for nucleotide transitions and transversions and a

reexamination of assumptions in weighted phylogenetic analysis.

Syst. Biol. 49 (4), 617–627.

Brower, A.V.Z., 1996. Combining data in phylogenetic analysis (a

response to Huelsenbeck et al., 1996). Trends Ecol. Evol. 11, 334–

335.

Chow, S., Kishino, H., 1995. Phylogenetic relationships between tuna

species of the genus Thunnus (Scombroidei: Teleostei): inconsistent

implications from morphology, nuclear and mitochondrial ge-

nomes. J. Mol. Evol. 41, 741–748.

Cole, K.S., 1988. Patterns of gonad structure in hermaphroditic gobies

(Teleostei: Gobiidae). Environ. Biol. Fish. 28, 125–142.

DeSalle, R., Brower, A.V.Z., 1997. Process partitions, congruence, and

the independence of characters: inferring relationships among

closely related Hawaiian Drosophila from multiple gene regions.

Syst. Biol. 46 (4), 751–764.

Eernisse, D.J., Kluge, A.G., 1993. Taxonomic congruence versus total

evidence, and amniote phylogeny inferred from fossils, molecules

and morphology. Mol. Biol. Evol. 10 (6), 1170–1195.

Finnerty, J.R., Block, B.A., 1995. Evolution of cytochrome b in the

Scombroidei (Teleostei): molecular insights into billfish (Istiophor-

idae and Xiphiidae) relationships. Fish. Bull. 93, 78–96.

Gill, A.C., Hoese, D.F., 1993. Paraxenisthmus springeri, new genus

and species of gobioid fish from the west Pacific, and its

phylogenetic position within the Xenisthmidae. Copeia 1993 (4),

1049–1057.

Gill, H.S., 1994. Phylogenetic relationships of the members of the

Bathygobius and Priolepis groupings (sensu Birdsong et al., 1988)

which possess a longitudinal papillae pattern. In: Proceedings of

the Fourth Indo-Pacific Fish Conference: Systematics and Evolu-

tion of Indo-Pacific Fishes, Bangkok, Thailand, pp. 40–59.

Gosline, W.A., 1955. The osteology and relationships of certain

gobioid fishes, with particular reference to the genera Kraemeria

and Microdesmus. Pac. Sci. IX, 158–170.

Harrison, I.J., 1989. Specialization of the gobioid palatopterygoqua-

drate complex and its relevance to gobioid systematics. J. Nat.

Hist. 23, 325–353.

Hoese, D.F., 1984. Gobioidei: relationships. in: Moser, H.G. (Ed.),

Ontogeny and Systematics of Fishes. Spec. Pub. Am. Soc. Ichthy.

Herp. No. 1. Allen Press, Lawrence, Kansas, pp. 588–591.

C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368 367

Hoese, D.F., Gill, A.C., 1993. Phylogenetic relationships of eleotrid

fishes (Perciformes: Gobioidei). Bull. Mar. Sci. 52 (1), 415–440.

Iwata, A., Hosoya, S., Larson, H.K., 2001. Paedogobius kimurai, a new

genus and species of goby (Teleostei: Gobioidei: Gobiidae) from

the west Pacific. Rec. Aust. Mus. 53, 103–112.

Johnson, G.D., Brothers, E.B., 1993. Schindleria: a paedomorphic

goby (Teleostei: Gobioidei). Bull. Mar. Sci. 52 (1), 441–471.

K€aallersj€oo, M., Albert, V.A., Farris, J.S., 1999. Homoplasy increases

phylogenetic structure. Cladistics 15, 91–93.

Klausewitz, W., Conde, B., 1981. Oxymetopon cyanoctenosus n. sp., un

nouvel Eleotride des Philippines, avec une etude comparee du genre

(Pisces, Perciformes, Gobioidei, Eleotridae). Revue Francaise

d�Aquariol. 8 (3), 67–76.

Kluge, A.G., 1998. Total evidence or taxonomic congruence: cladistics

or consensus classification. Cladistics 14, 151–158.

Kocher, T.D., Conroy, J.A., McKaye, K.R., Stauffer, J.R., Lockwood,

S.F., 1995. Evolution of NADH dehydrogenase subunit 2 in east

African cichlid fish. Mol. Phylogenet. Evol. 4 (4), 420–432.

Kocher, T.D., Stepien, C.A., 1997. Molecular Systematics of Fishes.

Academic Press, New York.

Larson, H.K., 2001. A revision of the gobiid fish genus Mugilogobius

(Teleostei: Gobioidei), and its systematic placement. Rec. West.

Aust. Mus. Suppl. No. 62, 233 pp.

Matsubara, K., Iwai, T., 1959. Description of a new sandfish,

Kraemeria sexradiata, from Japan, with special reference to its

osteology. J. Wash. Acad. Sci. 49 (1), 27–32.

Miller, P.J., 1973. The osteology and adaptive features of Rhyacichthys

aspro (Teleostei: Gobioidei) and the classification of gobioid fishes.

J. Zool. Lond. 171, 397–434.

Miller, P.J., 1998. The west african species of Eleotris and their

systematic affinities (Teleostei: Gobioidei). J. Nat. Hist. 32, 273–

296.

Mindell, D.P., Thacker, C.E., 1996. Rates of molecular evolution:

phylogenetic issues and applications. Ann. Rev. Ecol. Syst. 27,

279–303.

Murdy, E.O., 1989. A taxonomic revision and cladistic analysis of the

oxudercine gobies (Gobiidae: Oxudercinae). Rec. Aust. Mus.

Suppl. 11, 1–93.

Murdy, E.O., Shibukawa, K., 2001. A review of the gobiid fish genus

Odontamblyopus (Gobiidae: Amblyopinae). Ichthyol. Res. 48 (1),

31–43.

Nelson, J.S., 1994. Fishes of the World, third ed. Wiley, New York.

Nixon, K.C., Carpenter, J.M., 1996. On simultaneous analysis.

Cladistics 12, 221–241.

Parenti, L.R., Maciolek, J.A., 1993. New sicydiine gobies from Ponape

and Palau, Micronesia, with comments on systematics of the

subfamily Sicydiinae (Teleostei: Gobiidae). Bull. Mar. Sci. 53 (3),

945–972.

Parenti, L.R., Thomas, K.R., 1998. Pharyngeal jaw morphology and

homology in sicydiine gobies (Teleostei: Gobiidae) and allies. J.

Morph. 237, 257–274.

Patterson, C., 1993. Osteichthyes: Teleostei. In: Benton, M.J. (Ed.),

The Fossil Record 2. Chapman and Hall, London, pp. 622–656.

Pezold, F., 1993. Evidence for a monophyletic Gobiinae. Copeia 1993

(3), 634–643.

Randall, J.E., Allen, G.R., 1973. A revision of the gobiid fish genus

Nemateleotris, with descriptions of two new species. Quart. J.

Taiwan Mus. 26 (3–4), 347–367.

Randall, J.E., Hoese, D.F., 1985. Revision of the dartfishes, genus

Ptereleotris (Perciformes: Gobioidei). Indo-Pacific Fishes 7, 1–36.

Rennis, D.S., Hoese, D.F., 1985. A review of the genus Parioglossus

with descriptions of six new species (Pisces: Gobioidei). Rec. Aust.

Mus. 36, 169–201.

Rennis, D.S., Hoese, D.F., 1987. Aioliops, a new genus of ptereleotrine

fish (Pisces: Gobioidei) from the tropical Indo-Pacific with

descriptions of four new species. Rec. Aust. Mus. 39, 67–84.

Siddall, M.E., 1997. Prior agreement: arbitration or arbitrary? Syst.

Biol. 46 (4), 765–769.

Sorenson, M.D., 1999. TreeRot, Version 2. Boston University, Boston,

MA.

Sorenson, M.D., Ast, J.C., Dimcheff, D.E., Yuri, T., Mindell, D.P.,

1999. Primers for a PCR-based approach to mitochondrial genome

sequencing in birds and other vertebrates. Mol. Phylogenet. Evol.

12 (1), 105–114.

Springer, V.G., 1973. The osteology and adaptive features of Rhyac-

ichthys aspro (Teleostei: Gobioidei) and the classification of

gobioid fishes. J. Zool. Lond. 171, 397–434.

Springer, V.G., 1983. Tyson belos, new genus and species of Western

Pacific Fish (Gobiidae, Xenisthminae), with discussions of Gobioid

osteology and classification. Smithsonian Contributions to Zool-

ogy No. 390, 40 pp.

Springer, V.G., 1988. Rotuma lewisi, new genus and species of fish

from the southwest Pacific (Gobioidei, Xenisthmidae). Proc. Biol.

Soc. Wash. 101 (3), 530–539.

Swofford, D.L., 1998. PAUP* vesion 4. Sinauer, Sunderland, MA.

Takagi, K., 1989. Cephalic sensory canal system of the gobioid fishes

of Japan: comparative morphology with special reference to

phylogenetic significance. J. Tokyo Univ. Fish. 75, 499–568.

Tang, K.L., Berendzen, P.B., Wiley, E.O., Morrissey, J.F., Winterbot-

tom, R., Johnson, G.D., 1999. The phylogenetic relationships of the

suborderAcanthuroidei (Teleostei: Perciformes) based onmolecular

andmorphological evidence.Mol. Phylogenet.Evol. 11 (3), 415–425.

Thacker, C.E., 2000. Phylogeny of the Wormfishes (Teleostei:

Gobioidei: Microdesmidae). Copeia 2000 (4), 940–957.

Thacker, C.E., Cole, K.S., 2002. Phylogeny and evolution of the gobiid

genus Coryphopterus. Bull. Mar. Sci, in press.

Wang, H.-Y., Tsai, M-P., Dean, J., Lee, S.-C., 2001. Molecular

phylogeny of gobioid fishes (Perciformes: Gobioidei) based on

mitochondrial 12S rRNA sequences. Mol. Phylogenet. Evol. 20 (3),

390–408.

Wiley, E.O., Johnson, G.D., Dimmick, W.W., 1998. The phylogenetic

relationships of lampridiform fishes (Teleostei: Acanthomorpha),

based on a total-evidence analysis of morphological and molecular

data. Mol. Phylogenet. Evol. 10 (3), 417–425.

Wiley, E.O., Johnson, G.D., Dimmick, W.W., 2000. The interrela-

tionships of acanthomorph fishes: a total evidence approach using

molecular and morphological data. Biochem. Syst. Ecol. 28, 319–

350.

Winterbottom, R., 1993. Search for the gobioid sister group (Actin-

opterygii: Percomorpha). Bull. Mar. Sci. 52 (1), 395–414.

368 C.E. Thacker / Molecular Phylogenetics and Evolution 26 (2003) 354–368