Reconciling molecules and morphology: Molecular systematics and biogeography of Neotropical blennies (Acanthemblemaria)

Abstract

Neotropical reef fish communities are species-poor compared to those of the Indo-West Pacific. An exception to that pattern is the blenny clade Chaenopsidae, one of only three rocky and coral reef fish families largely endemic to the Neotropics. Within the chaenopsids, the genus Acanthemblemaria is the most species-rich and is characterized by elaborate spinous processes on the skull. Here we construct a species tree using five nuclear markers and compare the results to those from Bayesian and parsimony phylogenetic analyses of 60 morphological characters. The sequence-based species tree conflicted with the morphological phylogenies for Acanthemblemaria, primarily due to the convergence of a suite of characters describing the distribution of spines on the head. However, we were able to resolve some of these conflicts by performing phylogenetic analyses on suites of characters not associated with head spines. By using the species tree as a guide, we used a quantitative method to identify suites of correlated morphological characters that, together, produce the distinctive skull phenotypes found in these fishes. A time calibrated phylogeny with nearly complete taxon sampling provided divergence time estimates that recovered a mid-Miocene origin for the genus, with a temporally and geographically complex pattern of speciation both before and after the closure of the Isthmus of Panama. Some sister taxa are broadly sympatric, but many occur in allopatry. The ability to infer the geography of speciation in Acanthemblemaria is complicated by extinctions, incomplete knowledge of their present geographic ranges and by wide-spread taxa that likely represent cryptic species complexes.

Full text

Reconciling molecules and morphology: Molecular systematics and biogeography
of Neotropical blennies (Acanthemblemaria)
Ron I. Eytan
a,
⇑
, Philip A. Hastings
b
, Barbara R. Holland
c
, Michael E. Hellberg
a
a
Department of Biological Sciences, 202 Life Sciences Building, Louisiana State University, Baton Rouge, LA 70803, USA
b
Scripps Institution of Oceanography, University of California San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA
c
School of Mathematics & Physics, University of Tasmania, Private Bag 50, Hobart, TAS 7001, Australia
a r t i c l e i n f o
Article history:
Received 27 June 2011
Revised 18 September 2011
Accepted 19 September 2011
Available online 20 October 2011
Keywords:
Acanthemblemaria
Chaenopsidae
Blennies
Reef fishes
Neotropics
Species trees
Caribbean
Tropical Eastern Pacific
Divergence time estimation
Correlated characters
Morphology versus molecules
a b s t r a c t
Neotropical reef fish communities are species-poor compared to those of the Indo-West Pacific. An
exception to that pattern is the blenny clade Chaenopsidae, one of only three rocky and coral reef fish
families largely endemic to the Neotropics. Within the chaenopsids, the genus Acanthemblemaria is the
most species-rich and is characterized by elaborate spinous processes on the skull. Here we construct
a species tree using five nuclear markers and compare the results to those from Bayesian and parsimony
phylogenetic analyses of 60 morphological characters. The sequence-based species tree conflicted with
the morphological phylogenies for Acanthemblemaria, primarily due to the convergence of a suite of
characters describing the distribution of spines on the head. However, we were able to resolve some of
these conflicts by performing phylogenetic analyses on suites of characters not associated with head
spines. By using the species tree as a guide, we used a quantitative method to identify suites of correlated
morphological characters that, together, produce the distinctive skull phenotypes found in these fishes. A
time calibrated phylogeny with nearly complete taxon sampling provided divergence time estimates that
recovered a mid-Miocene origin for the genus, with a temporally and geographically complex pattern of
speciation both before and after the closure of the Isthmus of Panama. Some sister taxa are broadly sym-
patric, but many occur in allopatry. The ability to infer the geography of speciation in Acanthemblemaria is
complicated by extinctions, incomplete knowledge of their present geographic ranges and by
wide-spread taxa that likely represent cryptic species complexes.
Ó2011 Elsevier Inc. All rights reserved.
1. Introduction
Reef communities harbor the greatest marine fish diversity of
any oceanic ecosystem (Sale, 2002). Biodiversity of reef fishes is
highest in the Indo-West Pacific and decreases longitudinally to
the east and west, with the Neotropics being species-poor in com-
parison (Bellwood and Wainwright, 2002; Briggs, 1974; MacPher-
son et al., 2009; Mora et al., 2003). A major exception to this
pattern is the Blennioidei, a group of small, bottom-dwelling rocky
and coral reef fishes. Blennies are a species-rich group composed of
six families. Of those, the Labrisomidae, the Dactyloscopidae, and
the Chaenopsidae are the only reef fish families entirely or largely
endemic to the New World (Bellwood and Wainwright, 2002;
Hastings, 2009).
1.1. Acanthemblemaria
Acanthemblemaria (Metzelaar, 1919) is the most species-rich
genus of chaenopsids, as well as one of the most species-rich gen-
era of Neotropical blennies (Hastings, 2009; Hastings and Springer,
2009b). All members in the genus are small (1.2–3.5 cm standard
length) and are obligate dwellers of vacated invertebrate holes on
shallow (<1–22 m) rocky and coral reefs (Stephens, 1963). As cur-
rently recognized, Acanthemblemaria includes 22 species, 10 in the
Tropical Eastern Pacific and 12 in the Tropical Western Atlantic
(Hastings, 2009). Since the comprehensive treatment of the family
Chaenopsidae by Stephens (1963), more named species have been
added to Acanthemblemaria than to any other chaenopsid genus.
Much of this growth has been due to the recognition that several
species with broad distributions contain cryptic, often allopatric
taxa (Hastings and Robertson, 1999a; Hastings and Springer,
2009a,b; Lin and Galland, 2010).
The generic name Acanthemblemaria comes from the Greek
Akanthos-, or thorn. The name is apt, as Acanthemblemaria blennies
are typified by the presence of spinous processes on the frontal
bones (Metzelaar, 1919; Smith-Vaniz and Palacio, 1974; Stephens,
1055-7903/$ - see front matter Ó2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2011.09.028
⇑Corresponding author. Present address: Yale University, Department of Ecology
and Evolutionary Biology, Environmental Science Center, 21 Sachem Street, #358D,
New Haven, CT 06520-8106, USA.
E-mail addresses: ron.eytan@gmail.com (R.I. Eytan), phastings@ucsd.edu (P.A.
Hastings), Barbara.Holland@utas.edu.au (B.R. Holland), mhellbe@lsu.edu (M.E.
Hellberg).
Molecular Phylogenetics and Evolution 62 (2012) 159–173
Contents lists available at SciVerse ScienceDirect
Molecular Phylogenetics and Evolution
j o u r n a l h o m e p a g e : w w w . e l s e v ie r . c o m / l o c a t e / y m p e v

1963). Morphological characters related to head spination repre-
sent the majority of the characters used to infer the interspecific
relationships in the group (Hastings, 1990). Recent molecular phy-
logenies of the genus (Eytan, 2010; Lin and Hastings, 2011) recov-
ered Acanthemblemaria as monophyletic, but also recovered
conflicts with the morphological hypothesis of Hastings (1990),
where taxa with clear affinities based on cranial morphology were
not closely related in the molecular phylogeny.
The characters used for inferring phylogenetic relationships
must be independent of one another (Kluge, 1989). Suites of mor-
phological characters that evolve in concert violate this dictate.
Such correlated evolution is most likely to occur when a set of
characters underlie a functionally adaptive phenotype or common
developmental pathway (Emerson and Hastings, 1998). Such suites
of correlated characters can mislead phylogenetic analyses because
they track adaptive history instead of phylogeny (Holland et al.,
2010; McCracken et al., 1999) or because they are developmentally
linked to other characters (Schlosser and Wagner, 2004; West-
Eberhard, 2003). In practice, it is difficult to determine the under-
lying nature of character correlations. This is because a suite of
characters that are highly correlated with one another are ex-
pected to produce the same result as a suite of independent char-
acters with good phylogenetic signal: strong support for a given
clade (Shaffer et al., 1991).
Here we test whether the homoplastic morphological charac-
ters related largely to head spination in Acanthemblemaria are cor-
related with one another independent of the phylogeny, and if
accounting for that correlation can reconcile the molecular and
morphological hypotheses for the genus. We reconstruct the spe-
cies tree of the genus Acanthemblemaria using five nuclear mark-
ers and employ Bayesian relaxed clock divergence dating to
determine the age of the group and timing of speciation among
the members of the genus. We also examine the historical bioge-
ography of the genus, with the aim of elucidating the geography
of speciation in the group. Of particular interest is whether speci-
ation in this clade has occurred primarily between ocean basins
on either side of the Isthmus of Panama, or within the basins
themselves.
2. Materials and methods
2.1. Taxon sampling
Between one and five individuals from 16 of the 22 named
Acanthemblemaria species, as well as one undescribed species
and four outgroup taxa, chosen based on Hastings (1990) and
Almany and Baldwin (1996), were included in the study
(Table S1). Where possible, individuals were sampled from more
than one population. Of the taxa included, six are putative trans-
isthmian geminates (Hastings, 1990; Hastings and Springer,
1994), with two geminate pairs in the ingroup and one in the out-
group. Whole fishes were stored individually in 95% ethanol or
salt-saturated DMSO at ÿ80 °C. Voucher specimens for some spe-
cies are present in the SIO Marine Vertebrate Collection and photo
vouchers from freshly collected specimens of others have been
submitted to the Dryad repository.
2.2. DNA extraction, PCR and sequencing
DNA was extracted using the Qiagen (Valencia, CA) QIAMP DNA
Minikit. The polymerase chain reaction (PCR) was performed to
amplify five genetic markers (Table 1): nuclear protein-coding
genes recombination-activating gene 1 (rag1), titin-like protein
(TMO4C4), melanocortin 1 receptor (MC1R), SH3 and PX domain
containing three gene (SH3PX3), and intron V from nuclear
a
-tropomyosin (atrop). PCR amplification of the full-length rag1
molecule was not possible for some taxa. In these cases, a set of
internal primers were developed for the study and used to amplify
rag1.
Amplicons were purified with a Strataprep PCR Purification Kit
(Stratagene, La Jolla, CA) or directly sequenced without cleanup in
both directions on an ABI 3100 or 3130 XL automated sequencer
with 1/8 reactions of BigDye Terminators (V3.1, Applied Biosys-
tems) with the amplification primers, or internal primers as indi-
cated in Table 1.
2.3. Sequence alignment and model selection
Sequences for the four protein-coding genes were aligned
using MUSCLE (Edgar, 2004) as implemented in Geneious v3.6
(Drummond et al., 2007). The
a
-tropomyosin sequences, which
contained numerous gaps, were aligned in BAli-Phy v2.0.1 (Su-
chard and Redelings, 2006) using the GTR substitution model,
gamma distributed rate variation, and the default indel model.
BAli-Phy was run four times to ensure concordance among runs.
All the samples of the Markov chain taken before convergence,
as determined by stationarity in the Markov chain, which was
visualized in Tracer v1.5 (Rambaut and Drummond, 2010), were
discarded as burnin. The consensus alignment from the run with
the highest posterior probability was used for subsequent analy-
ses and all positions with posterior probabilities less than 0.95
were discarded.
Ten different partitioning strategies were evaluated for both the
species tree and concatenated analyses. These partitioning strate-
gies ranged from treating all genes as a single partition, to each
gene and codon position given its own partition (Table 2). Models
of sequence evolution for each strategy were determined using
jModelTest (Posada, 2008) and the AIC, while partitioning strate-
gies were determined using 2 ln Bayes factors (Kass and Raftery,
1995) with the modification of Suchard et al. (2001), implemented
in Tracer v1.5.
2.4. Bayesian species tree and divergence dating analyses
2.4.1. Species tree estimation
Species tree analyses were conducted using the ⁄BEAST package
in BEAST v1.5.4 (Heled and Drummond, 2010). Sequences were
grouped by nominal species for the analyses. Trees and clocks were
unlinked among all genes, with each gene region dated using the
uncorrelated log normal distribution (UCLD) (Drummond et al.,
2006) and the calibrations detailed below. The datasets were run
twice for 100,000,000 generations, sampling every 5000. Conver-
gence onto the posterior distribution for the estimated topology
was assessed using the ‘‘compare’’ and ‘‘cumulative’’ functions in
Are We There Yet? (AWTY) (Nylander et al., 2008). Convergence
onto the posterior distribution for parameter estimates was as-
sessed by effective sample size (ESS) values greater than 250, as
determined in Tracer v1.5 (Rambaut and Drummond, 2010). A
time-calibrated phylogeny of the concatenated dataset was also
constructed in BEAST, using the same calibrations and run condi-
tions as for the species tree.
2.4.2. Divergence dating
Priors on the time to most recent common ancestor (TMRCA)
for two species pairs separated by the Isthmus of Panama were
specified. The first species pair considered, Acanthemblemaria
betinensis and Acanthemblemaria exilispinus, occur in <1 m of water
and are restricted to areas close to the Isthmus (Hastings, 2009).
These distributions suggest that their progenitor was split close
to the final closure of the Isthmus. The calibration was given
an exponential prior with a mean of 7 million years and a zero
160 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

offset of 3.1 million years. This prior represents the most recent
possible split for the geminates at the close of the Isthmus, but al-
lows for a split prior to the closure, although with decreasing
probability back in time.
The second pair of geminates considered, A. rivasi and A. cas-
troi, have a Galápagos–Caribbean distribution (Hastings, 2009).
While the most recent possible split between these two would
have been the closure of the Isthmus and the earliest possible
split the rise of the Galapagos (at most 17 million years ago; Wer-
ner and Hoernle, 2003), the split most probably occurred between
those dates. A truncated normal prior for the split time of A. rivasi
and A. castroi was specified. A minimum offset of 3.1 million
years, representing the most recent possible split for the species
pair, was used. The mean and standard deviation were set at 10
and 3.52, respectively, which gave a 95% confidence interval of
3.1 and 16.9 million years for the prior. The third pair of gemi-
nates, the outgroup taxa Ekemblemaria myersi and E. nigra, were
not used to calibrate a molecular clock because the E. myersi spec-
imen used in this study was not collected in Panama, but further
north.
2.5. Analysis of morphological data
A modified version of the morphological matrix from Hastings
(1990) was analyzed. Acanthemblemaria stephensi and A. atrata
were not sampled for the species tree analyses, as tissues were
not available, and were removed from the matrix. Three taxa were
added to the morphological matrix (A. n. sp.,Protemblemaria bicir-
rus, and Cirriemblemaria lucasana) and scored for the set of 60 char-
acters from Hastings (1990). The new matrix was analyzed in a
Bayesian framework using MrBayes v.3.1.2 (Ronquist and Huelsen-
beck, 2003) and the Mkv model for morphological data (Lewis,
2001). In MrBayes all characters were set as variable and unor-
dered, save for three that were ordered in Hastings (1990): charac-
ter 2 (number of spines on the nasal rami (excluding AFO process)),
character 3 (process on the nasal bones anterior to the first antero-
frontal sensory pore (AFO process)), and character 7 (anterolateral
extent of the frontal ridge). The MrBayes analyses were run twice
with four heated chains (temp = 0.1) for 10,000,000 generations,
sampling every 1000. Convergence onto the posterior distribution
for the model parameters and topology was assessed using ESS
Table 1
Primer sets and PCR conditions used in this study.
Marker Primer name Primer sequence References
ATROP ATROP-L GAG TTG GAT CGC GCT CAG GAG CG Hickerson and Cunningham (2005)
ATROP-H CGG TCA GCC TCC TCA GCA ATG TGC TT Hickerson and Cunningham (2005)
RAG1 RAG1Of2 CTG AGC TGC AGT CAG TAC CAT AAG ATG T Taylor and Hellberg (2005)
RAG1F.4.27 AGCTGTAGTCAGTAYCACAARATG This study
RAG1S2F CCG AGA AGG CTG TAC GTT TCT CTT Taylor and Hellberg (2005)
RAG1S1R CCT GCC AGC ACA GAA ACA GAC ATA Taylor and Hellberg (2005)
RAG1R1.539.519 CAG GAC AGT TCT GAG TTT GGC This study
RAG1F3.519.539 GCC AAA CTC AGA ACT GTC CTG This study
RAG1S2R CATTACCGGCTTGAGCTTCATCCT Taylor and Hellberg (2005)
RAG1F4.1129.1148 ATGAATGGGAACTTTGCCCG This study
RAG1S3F GCT CAT GAG GCT CTA TAT TCA GAT G Taylor and Hellberg (2005)
RAG1Or2 CTG AGT CCT TGT GAG CTT CCA TRA AYT T Taylor and Hellberg (2005)
SH3PX3 SH3PX3_F461 GTATGGTSGGCAGGAACYTGAA Li et al. (2007)
SH3PX3_R1303 CAAACAKCTCYCCGATGTTCTC Li et al. (2007)
TMO4C4 TMO-F2 GAKTGTTTGAAAATGACTCGCTA Near et al. (2004)
TMO-R2 AAACATCYAAMGATATGATCATGC Near et al. (2004)
MC1R MC1RFor ATGGAAATGACCAACRGGTCCYTGC This study
MC1RRev CARGGTTYTMCGCAGCTCCTGGC This study
MC1RF477 TCCAGCATCCTCTTCATCG This study
MC1RR243 AGCATACCTGGGTGAACGTC This study
MC1RR907 CGTAAATGAGCGGGTCGATGA This study
MC1RR649 TATGAAGGTAGAGCACCGC This study
PCR conditions
RAG1,TMO4C4,SH3PX3,MC1R: One cycle of 94 °C for 2 min, 50 °C for 90 s, 72 °C for 2 min followed by 38 cycles of 94 °C for 45 s, 50 °C for 1 min, and 72 °C for 90 s, and a final
cycle of 94 °C for 40 s, 50 °C for 1 min, and 72 °C for 10 min.
ATROP: One cycle of 94 °C for 2 min, 62 °C for 1:30, 72°for 2 min followed by 38 cycles of 94 °C for 45 s, 62 °C for 1 min, and 72 °C for 45 s, and a final cycle of 94 °C for 45 s,
62 °C for 1 min, and 72 °C for 10 min.
Table 2
Partitioning strategies evaluated for this study.
Model Name Partition description Number of partitions
1 FULL All included nucleotide positions 1
2 SNMAT SRD06 model for nDNA, mtDNA,
a
-trop concatenated 3
3 SNMIE SRD06 model for nDNA, mtDNA,
a
-trop intron and exon 4
4 GENES Partitioned by gene region,
a
-trop concatenated 5
5 NMAT nDNA by codon, mtDNA by codon,
a
-trop concatenated 5
6 NMIE nDNA by codon, mtDNA by codon,
a
-trop intron and exon 6
7 SGAT SRD06 model for each locus,
a
-trop concatenated 9
8 SGIE SRD06 model for each locus,
a
-trop intron and exon 10
9 GCAT Each locus by codon position,
a
-trop concatenated 13
10 GCIE Each locus by codon position,
a
-trop intron and exon 14
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 161

values in Tracer v1.5 (Rambaut and Drummond, 2010), and the
‘‘compare’’ and ‘‘cumulative’’ functions in AWTY (Nylander et al.,
2008), respectively.
2.6. Identification of correlated incongruent morphological characters
The method of Holland et al. (2010) was used to identify mor-
phological characters that are incongruent with the molecularly-
derived Acanthemblemaria phylogeny and correlated with one an-
other because of homoplasy. Following Holland et al. (2010) we
constructed a matrix of ‘‘excess’’ distances between each pair of
morphological characters (the excess distance is the parsimony
score of the two characters taken together minus the parsimony
score of each character taken individually). Zeros in the matrix
indicate compatible pairs of splits. The dissimilarity matrix was
visualized in SplitsTree4 (Huson and Bryant, 2006). UPGMA was
used to identify a maximal clique of compatible characters. This
is heuristic in that it is not guaranteed to find the maximum clique,
so the procedure was repeated 100 times with different random
orderings of the characters. The size of the largest clique was com-
pared to those found for 100 shuffled alignments created following
the second shuffling procedure described by Holland et al. (2010).
This procedure creates shuffled alignments that have the same par-
simony score on the sequence-based trees as the unshuffled mor-
phological data. Each of the 100 shuffles was based on a different
tree from the posterior distribution of the species tree analysis.
This gave a null distribution of clique sizes that allowed us to as-
sess if the maximal clique found in the unshuffled data was larger
than would be expected by chance conditional on the level of
agreement between the morphological characters and the se-
quence-based trees. If the clique is larger than expected by chance
this is interpreted as evidence for convergent evolution in the mor-
phological data.
3. Results
3.1. Molecular data, partitioning strategy, and convergence criteria
The five nuclear gene regions were successfully amplified in all
taxa for a total alignment length of 3790 bp. The lengths of the
aligned sequences, as well as the proportion of variable and parsi-
mony informative sites for each marker, can be found in Table 3.
All sequences have been submitted to GenBank with accession
numbers JN897037-JN897271. Using 2 ln Bayes factors, the GCIE
partitioning strategy was selected. For each of the analyses
(time-calibrated species and concatenated trees, and the morpho-
logical tree) convergence diagnostics (AWTY results and ESS values
>250) indicated that convergence onto the posterior distribution
had occurred.
3.2. The species tree estimate for Acanthemblemaria yielded a well-
supported phylogeny but it was in significant conflict with the
morphological hypothesis
3.2.1. Comparison of species tree with Hastings (1990)
The Bayesian species tree estimate yielded a well-resolved
topology with 13 of 19 nodes supported by Bayesian posterior
probability (BPP) values greater than 0.95 (Fig. 1A). However, many
of the well-supported nodes conflicted with the morphological
hypothesis of Hastings (1990) (Fig. 1B) and the Bayesian estimate
of the morphological data inferred in this study (Fig. 1C).
As in Hastings (1990),Acanthemblemaria was recovered as
monophyletic in the species tree analysis, here with high support
(BPP = 1.0) (Fig. 1A). Hastings’ phylogeny was highly nested, show-
ing a progression from A. chaplini and A. greenfieldi at the base of
the tree, through the Caribbean Acanthemblemaria taxa, to the
‘‘hancocki species group’’ at the crown (Fig. 1B). In the ⁄BEAST spe-
cies tree, two major clades, here denoted as Clade I and Clade II,
were recovered with high support (Fig. 1A). Each of these clades
contained a pair of transisthmian sister species, both of which were
recovered with BPP of 1.0. Neither of these transisthmian pairs was
basal to the other taxa in their respective clades. The relationships
of each of these two pairs of geminate taxa to the other members of
their respective clades received high support, but for both there
was less than 0.95 posterior support (Fig. 1A).
Clade I was composed of a majority of Eastern Pacific taxa, Clade
II of mostly Caribbean taxa. In Clade I, a monophyletic group of
taxa that occurs in the Eastern Pacific, with the exception of the
geminate A. rivasi, was found. This clade, A. crockeri + ‘‘the hancocki
species group’’ (sensu Hastings, 1990), was also recovered by
Table 3
Lengths of aligned sequences and the proportion of variable and parsimony informative sites for each marker and partition.
Included length % Variable sites (no. variable sites) % PI sites (no. PI sites)
Gene region
RAG1 1503 bps 15.17 (228) 11.38 (171)
MC1R 855 bps 16.61 (142) 12.28 (105)
SH3PX3 741 bps 16.87 (125) 12.96 (96)
TMO4C4 411 bps 24.82 (102) 18.98 (78)
a
-tropomyosin 280 bps 20.71 (58) 13.93 (39)
TOTAL 3790 bps 17.28 (655) 12.9 (489)
Partition
RAG1 (1) 501 bps 6.59 (33) 5.39 (27)
RAG1 (2) 501 bps 5.19 (26) 3.39 (17)
RAG1 (3) 501 bps 33.73 (169) 26.55 (133)
MC1R (1) 285 bps 4.91 (14) 3.86 (11)
MC1R (2) 285 bps 2.46 (7) 1.4 (4)
MC1R (3) 285 bps 42.81 (122) 36.14 (103)
SH3PX3 (1) 247 bps 2.43 (6) 2.02 (5)
SH3PX3 (2) 247 bps 1.21 (3) 0.81 (2)
SH3PX3 (3) 247 bps 46.96 (116) 38.06 (94)
TMO4C4 (1) 137 bps 12.41 (17) 8.76 (12)
TMO4C4 (2) 137 bps 5.84 (8) 4.38 (6)
TMO4C4 (3) 137 bps 56.2 (77) 45.99 (63)
a
-Tropomyosin (I) 92 bps 56.52 (52) 42.39 (39)
a
-Tropomyosin (E) 188 bps 3.19 (6) 1.06 (2)
Total 3790 bps 17.28 (655) 12.9 (489)
162 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

Hastings. However, the species tree analysis recovered the trans-
isthmian geminates A. castroi and A. rivasi as sister to the remaining
species in the clade, with A. crockeri nested within the ‘‘hancocki
species group’’, but with poor support.
In Clade II, the well-supported relationship between the
geminate taxa A. betinensis and A. exilispinus was also recovered
in the Hastings (1990) analysis. However, many of the other
relationships within Clade II conflicted with the morphological
phylogeny. The A. maria/A. spinosa split was not recovered in
the species tree, nor was the ‘‘aspera species group’’ of (A. medusa,
(A. aspera,A. paula)). Instead, A. spinosa was found to be sister to
A. aspera and A. paula, while A. maria was sister to the unde-
scribed Acanthemblemaria species (not included in Hastings,
1990). A. medusa, which was placed as sister to A. aspera and A.
paula in the ‘‘aspera species group’’ based on morphological data,
was found to be sister to A. maria and A. n. sp., albeit with a BPP
of 0.86.
3.2.2. Comparison of species tree with Bayesian estimates of
morphology
The phylogeny based on Bayesian inference of morphological
data closely mirrored the parsimony analysis of Hastings
(1990), although support was poor for many of the nodes
(Fig. 1C). All the relationships and clades inferred by Hastings
were recovered here with the exception of the hancocki/stephensi
split, as the latter taxon was not included in this study. The
undescribed Acanthemblemaria species, which was not included
in Hastings (1990), was recovered here as a member of the ‘‘han-
cocki species group’’. Also, as in Hastings (1990), the morpholog-
ical tree inferred here was highly nested, with the same
progression of taxa.
3.3. Identification of correlated incongruent morphological characters
The median size of the maximal clique within the 60 morpho-
logical characters was 24 characters. It is possible to get large cli-
ques of compatible characters by chance, but never as large as
the one recovered for the unshuffled data: the shuffled data pro-
duced maximal cliques of size 9–22 (median 16). (Recall that the
shuffling procedure of Holland et al. (2010) is guaranteed to pro-
duce shuffled characters with the same level of incongruence to
the molecular trees as the unshuffled characters.) This suggests
that convergent evolution has occurred amongst the morphologi-
cal characters.
A. exilispinus
A. rivasi
A. castroi
A. macrospilus
Ekemblemaria
A. balanorum
A. paula
A. chaplini
A. greenfieldi
A. crockeri
A. spinosa
A. aspera
A. maria
A. stephensi
A. hancocki
A. betinensis
A. medusa
A. atrata
Hypothetical Second Outgroup
E. nigra
A. exilispinus
A. rivasi
C. lucasana
A. castroi
A. macrospilus
E. myersi
A. balanorum
A. paula
A. chaplini
A. greenfieldi
A. crockeri
A. spinosa
A. aspera
A. maria
P. bicirrus
A. n. sp.
A. hancocki
A. betinensis
A. medusa
0.9
0.45
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
0.86
0.88
0.77
1.0
0.89
0. 2
0. 97
0. 83
0. 98
0. 51
0. 52
0. 67
0. 92
0. 64
1.0
0. 89
0. 64
1.0
0. 94
0. 99
0. 71
A. chaplini
A. greenfieldi
A. n. sp.
A. hancocki
A. macrospilus
A. rivasi
A. castroi
A. balanorum
A. crockeri
A. spinosa
A. maria
A. exilispinus
A. betinensis
A. paula
A. aspera
A. medusa
Cirriemblemaria
Protemblemaria
E. nigra
E. myersi
A
B C
I
II
0. 008
“hancocki species group” “aspera species group”
“hancocki species group” “aspera species group”
Fig. 1. Molecular and morphological hypotheses of the phylogeny of Acanthemblemaria, with Tropical Eastern Pacific (TEP) and Caribbean taxa in bold or normal font,
respectively. (A) Bayesian species tree estimated in ⁄BEAST. Posterior probabilities are shown at all nodes and branch lengths are in units of substitutions per site. The
majority of taxa in Clade I (five versus three) occur in the TEP, while Clade II consists of primarily Caribbean species (seven versus one). (B) Morphological phylogeny inferred
using maximum parsimony from Hastings (1990). (C) 50% majority rule consensus tree from the Bayesian estimate of the morphological dataset. Posterior probabilities
greater than 0.5 shown at nodes. The ‘‘hancocki’’ and ‘‘aspera’’ species groups, sensu Hastings (1990) are enclosed by boxes.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 163

3.4. A. spinosa and A. medusa were responsible for the majority of
incongruence between molecules and morphology
The splits network of the morphological tree with the sequence-
based tree revealed many areas of agreement between the two
trees, as indicated by strictly bifurcating splits, including the entire
‘‘hancocki species group’’ clade (Fig. 2A). The two taxa that were
responsible for the majority of the conflict between the two trees,
as visualized by conflicting networks of splits, were A. spinosa and
A. medusa. For both of these taxa a relatively large number of extra
splits had to be traversed to unite them with clades specified by
either the morphological or molecular phylogeny. When both of
these taxa were removed from the tree, conflicting splits disap-
peared from the splits networks (Fig. 2C).
3.5. Evidence of a mix of historical and convergent signal for the
placement of A. spinosa, but not A. medusa
We investigated the characters responsible for the conflict be-
tween the morphological and molecular trees and the source of
the incompatible splits. In the case of A. spinosa, the A. maria/A.
spinosa split that was recovered from the morphological matrix
was supported by six characters (Table 4A). However, six other
characters in the morphological matrix were incompatible with
the A. maria/A. spinosa split (Table 4B). When a maximum parsi-
mony (MP) tree was inferred using only these incompatible char-
acters, six most parsimonious trees were found, all supporting
the clade (A. aspera, (A. paula,A. spinosa)) (not shown.) This clade
was found in the species tree as well, although in the species tree
the sister relationship was (A. spinosa, (A. aspera,A. paula)). A single
character in the morphological matrix (57; posterior pair of antero-
frontal pores fused into a single medial pore) was a synapomorphy
for the clade (A. aspera,A. paula,A. spinosa).
The inclusion of A. medusa in the ‘‘aspera species group’’, which
consists of (A. aspera,A. medusa,A. paula), was supported by two
characters in the morphological matrix (Table 4C). However, the
morphological dataset contained five characters in conflict with
the ‘‘aspera species group’’ (Table 4D). Unlike the conflicting char-
acters for the A. maria/A. spinosa split, the maximum parsimony
trees inferred from the characters conflicting with the ‘‘aspera spe-
cies group’’ did not recover the clade found in the species tree: (A.
medusa, (A. maria,A. n. sp.)). Instead, all the MP trees recovered a
clade consisting of A. aspera,A. chaplini,A. greenfieldi, and A. medusa
and no morphological characters supported the clade found in the
species tree.
3.6. Time calibrated phylogenies recovered a mid-Miocene origin for
Acanthemblemaria
3.6.1. Species tree
The dated species tree analysis found that Acanthemblemaria
originated in the mid-Miocene, with a complex pattern of specia-
tion within the genus both before and after the closure of the Isth-
mus of Panama (Fig. 3 and Table 5). The time to most recent
common ancestor (TMRCA) of Acanthemblemaria was recovered
with a mean of 13.1 mya and lower and upper confidence levels
of 7.4 and 20.9 mya, respectively.
Three out of six terminal splits in Acanthemblemaria were in-
ferred to have occurred prior to the closure of the Isthmus of Pan-
ama. Two of these three ingroup splits were the transisthmian
geminates A. castroi/A. rivasi and A. betinensis/A. exilispinus with
mean split times of 4.6 and 4.2 mya, respectively. The third termi-
nal split prior to the closure of the isthmus, that of A. chaplini/A.
greenfieldi, had a mean divergence date of 8.2 mya, but was not sig-
nificantly older than either of the geminate taxa. In addition to
those three splits, two clades that did not include transisthmian
geminates were also found to have split prior to the closure of
the isthmus. The (A. spinosa, (A. aspera,A. paula)) clade had a mean
TMRCA of 7.7 mya and the (A. medusa, (A. maria,A. n. sp.)) clade had
mean divergence time of 8.3 mya (Table 5).
For three pairs of terminal taxa, a split after the closure of the
Isthmus of Panama could not be rejected. The mean TMRCA for
two of those splits, A. aspera/A. paula and A. maria/A. n. sp. were
similar, 5.3 and 5.7 mya, with lower confidence limits of 2.63
and 2.76 mya, respectively. In contrast, the third split, A. balano-
rum/A. macrospilus, was substantially younger, with a mean
inferred divergence time of 2.7 mya. The clade to which those
two species belong, (A. hancocki, (A. crockeri, (A. balanorum,A.
macrospilus))) was also inferred to have diverged after the closure
of the Isthmus (3.9 mya, but with a lower confidence limit of
1.9 mya).
3.6.2. Concatenated tree
The time-calibrated estimate of the phylogeny from the concat-
enated dataset yielded a well-supported phylogeny that was con-
gruent with the species tree, both in topology and support, as
well as divergence times of major clades and splits (Fig. 4 and Ta-
ble 5). For the splits and clades that were shared between the spe-
cies tree and concatenated analyses (i.e. all interspecific splits)
divergence dates were in agreement. However, there was a trend
towards older divergence estimates from the species tree analysis
compared to the concatenated analysis, although it was not signif-
icant (Table 5).
The concatenated analysis revealed substantial divergence
times among populations for six nominal species: A. chaplini,A. riv-
asi,A. medusa,A. maria,A. paula, and A. spinosa, where the mean
TMRCA between populations within species was at least 1 mya
for all taxa (except A. rivasi at 0.97 mya). The most extreme exam-
ple comes from A. chaplini. Individuals sampled from Bocas del
Toro, Panama and the Abacos in northwest Bahamas were deeply
diverged from A. chaplini sampled from New Providence, in the
central Bahamas (Fig. 4). The mean TMRCA for the intraspecific
split in A. chaplini was 5.06 mya, with lower and upper HPDs of
2.77 and 7.95 mya, respectively (Table 5). This split time was sig-
nificantly older than the one between the A. chaplini individuals
from Panama and the northwest Bahamas (Table 5 and Fig. 4). As
opposed to A. chaplini, the other species with substantial intraspe-
cific divergence did not have significantly different split times be-
tween populations.
4. Discussion
4.1. Acanthemblemaria – molecules versus morphology
Our phylogenetic reconstruction of the genus Acanthemblemaria
based on molecular data was in significant conflict with the phylo-
genetic estimate of the group based on morphological data (Figs. 1
and 2). Our results also conflicted with those from a recent total
evidence analysis of relationships within the Chaenopsidae based
on one mitochondrial marker, four nuclear markers and 148 mor-
phological characters (Lin and Hastings, 2011). That study sampled
fewer species within Acanthemblemaria and because of conflicts
among genetic markers the morphological signal was dominant
within this portion of the chaenopsid phylogeny (Lin and Hastings,
2011, Fig. 6) resulting in a hypothesis of relationships resembling
the morphological analysis of Hastings (1990).
Two species were responsible for most of the conflict between
the molecular and morphological phylogenies – A. medusa and A.
spinosa (Fig. 2). A. spinosa (the ‘‘spinyhead blenny’’) has an elabo-
rate suite of spinous processes on several bones of the head. A.
maria has the most elaborate spinous processes in the group
164 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

A. spinosa
A. maria
A. n. sp.
A. exilispinis
A. betinensis
A. medusa
A. paula
A. aspera
Cirrieblemaria
Protemblemaria
E. myersi
E. nigra
A. chaplini
A. greenfieldi
A. balanorum
A. macrospilus
A. crockeri
A. hancocki
A. castroi
A. rivasi
A. spinosa
A. maria
A. n. sp.
A. exilispinis
A. betinensis
A. paula
A. aspera
Cirrieblemaria
Protemblemaria
E. myersi
E. nigra
A. chaplini
A. greenfieldi
A. balanorum A. macrospilus
A. crockeri
A. hancocki
A. castroi
A. rivasi
A. exilispinis
A. betinensis
A. mariaA. n. sp.
A. paula
A. aspera
A. chaplini
A. greenfieldi
E. myersi
E. nigra
Cirrieblemaria
Protemblemaria
A. castroi
A. rivasi
A. hancocki
A. crockeri
A. balanorum
A. macrospilus
A
B
C
Fig. 2. Split networks of the morphological tree with the species tree. Splits that agreed between the two trees are indicated by strictly bifurcating splits. Conflicting splits are
represented as a network of edges. (A) Split network for all taxa. (B) Split network after the removal A. medusa. (C) Split network after the removal of A. medusa and A. spinosa.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 165

(Böhlke, 1961) and a gross skull morphology similar to A. spinosa
(Smith-Vaniz and Palacio, 1974). Analyses based on morphological
data recovered A. maria as the sister species to A. spinosa, both
in this study (Fig. 1C), and in Hastings (1990) (Fig. 1B). This
inferred sister relationship between A. maria and A. spinosa was,
unexpectedly, not reflected in the genetically-based species tree,
where A. spinosa was recovered as sister to A. aspera and A. paula
(Fig. 1A).
The A. maria/A. spinosa clade recovered from the analyses of the
morphological data was supported by six characters (Table 4A).
Five of these come from three bones in the skull: the frontals
and the two infraorbitals (Table 4A and Hastings (1990)). These
may be functionally constrained in an unknown way, may share
a common developmental pathway, or the character states could
have been scored erroneously by Hastings (1990).
Six morphological characters were incompatible with an A. mar-
ia and A. spinosa clade (Table 4B). A parsimony analysis of these six
characters recovered the clade (A. aspera,A. paula,A. spinosa),
which was also found in the species tree analysis (Fig. 1A). How-
ever, in contrast to the species tree, the parsimony analysis recov-
ered A. spinosa sister to A. paula, with A. aspera sister to these two
taxa (not shown). Only one of those six characters relates to spines
(Table 4B) and its state is shared by A. paula and A. spinosa (Has-
tings, 1990). Taken together with the convergent character states
of skull bones in A. maria and A. spinosa, this result gives credence
to the idea that suites of characters relating to spinous processes
have evolved multiple times in Acanthemblemaria. These results
suggest that although there was strong support in the morpholog-
ical data for the sister relationship of A. maria and A. spinosa, there
was also some support for the (A. aspera,A. paula,A. spinosa) clade,
but it got ‘‘outvoted’’ in the morphological analyses.
In contrast to A. spinosa, the placement of A. medusa in the mor-
phological analyses does not appear to be caused by convergence.
The morphological phylogeny places A. medusa sister to A. aspera
and A. paula, in the ‘‘aspera species group’’ (Fig. 1A and B, and Has-
tings, 1990). This group is supported by two synapomorphies, both
related to the lacrimal bone (Table 4C). However, more characters
did not support the ‘‘aspera species group’’ than did; five in total
(Table 4D). When parsimony trees were constructed using these
five characters, the clade found in the species tree (A. maria,A. me-
dusa,A. n. sp.) was not recovered (not shown). These results show
that there was not strong support for the ‘‘aspera species group’’
sensu Hastings (1990) in the morphological data. However, in con-
trast to A. spinosa, there was little if any support for an alternate
placement of A. medusa.
Suites of characters can cause substantial errors in phylogenetic
analyses based on morphology because they can create the illusion
that relationships are supported by more independent characters
than is the case. Known suites of correlated characters point to
the role of natural selection in the repeated evolution of function-
ally adaptive phenotypes and/or the role of common developmen-
tal mechanisms (Emerson, 1982; Emerson and Hastings, 1998;
Holland et al., 2010; McCracken et al., 1999).
The function of the spinous processes on the skull bones of
Acanthemblemaria is not known. Acanthemblemaria blennies spend
most of their lives in vacated invertebrate holes (Böhlke, 1957;
Böhlke and Chaplin, 1993). As such, the heads of these fishes are
frequently the only exposed part of their bodies and thus likely tar-
gets for (possibly convergent) selective pressure. Skull morphology
does not appear to be important in feeding behavior, nor does it
influence predation success (Clarke et al., 2009, 2005; Finelli
et al., 2009). There may be selection for skulls that efficiently block
the blenny shelters as a means of defense against predators (Lind-
quist and Kotrschal, 1987). Defense against conspecifics seems
more likely, as shelters may be limiting (Hastings and Galland,
2010) and A. spinosa individuals can use the head spines to wedge
Table 4
(A–D) The list of characters found which support or conflict with the placement of A. spinosa and A. medusa in the morphological phylogeny. Character numbers, names, and states
are from the morphological data matrix used in this study, which was based on Hastings (1990).
Character
number
Character names and states
A. Characters supporting A. maria/A. spinosa split
4 Lateral supratemporal ridge: spines present medially
5 Posterior extent of the frontal ridge: to lateral supratemporal ridge
7 Anterolateral extent of the frontal ridge: confluent with the dorsoposterior margin of the postorbital
27 Orbital margin of the postorbital: serrations or spines present
30 Dorso-posterior margin of the postorbital: a row of laterally projecting spines present, contiguous with a row of spines on the frontal wedge
48 Shape of the proximal dorsal-fin pterygiophores (at the level of the mid-spinous dorsal fin): a single central strut present with a flat sheet of bone
both anteriorly and posteriorly
B. Characters incompatible with A. maria/A. spinosa split
8 Central area of the frontal wedge: an open swath with no spines or ridges present (maria) OR spines or ridges present (spinosa)
31 Shape of the junction of the circumorbitals: entire, the lacrimal and postorbital both extending to the posterior angle (spinosa) OR the postorbital
excluded from the posterior angle of the circumorbitals (maria)
42 Neural spur, a lateral projection on the anterior portion of the neural arch: present on one to four caudal vertebrae (spinosa) OR absent from all caudal
vertebrae (maria)
47 Posterior inner margin of the pelvis: no ossified threads present (spinosa) OR two central threads of bone present (maria)
56 Modal number of common pores: one (spinosa) OR two or more (maria)
57 Posterior pair of anterofrontal pores: fused into a single medial pore (spinosa) OR separate (maria)
C. Characters supporting A. medusa as part of the ‘‘aspera species group’’
21 Ventral margin of the lacrimal: three or four blades present
23 Ventral margin of the lacrimal at the third anterior infraorbital pore: a distinct notch present
D. Characters incompatible with A. medusa as part of the ‘‘aspera species group’’
1 Anterior margin of the nasal bones: smooth (medusa and aspera) OR spines or serrations present (paula)
7 Anterolateral extent of the frontal ridge:confluent with the middle of the supraorbital flange, at or anterior to the second supraorbital sensory pore
(medusa and aspera) OR confluent with the lateral edge of the supraorbital flange, at or posterior to the first supraorbital sensory pore (SOl) but
anterior to the frontal/postorbital juncture (paula)
8 Central area of the frontal wedge: an open swath with no spines or ridges present (aspera and medusa) OR spines or ridges present (paula)
44 Epipleural ribs: present on all precaudal vertebrae (within one before to one after the last precaudal vertebra) (medusa and paula) OR absent from
two or more posterior precaudal vertebrae (aspera)
45 Hypural five: ossified, autogenous (paula) OR unossified or not autogenous (aspera and medusa) (Pleisomorphic condition uncertain)
166 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

themselves into shelters and temporarily prevent extraction by lar-
ger conspecifics (PAH, pers. observ.).
It seems unlikely that A. maria and A. spinosa are subject to
exactly the same selective pressures. A. maria occurs in high-
energy environments on the reef crest or in shallow water and
generally does not live in live or standing dead corals, nor does
it shelter in holes high up off the reef substrate (Clarke, 1994;
Greenfield, 1981; Greenfield and Johnson, 1990); Eytan and
Hellberg, unpub. data). A. spinosa, on the other hand, is found
in deeper, lower energy sections of the reef, typically in live or
standing dead coral not close to the reef substrate (Clarke,
1989, 1994, 1996; Greenfield and Greenfield, 1982; Eytan and
Hellberg, unpub. data).
Alternatively, convergence in the skull spines of A. maria and A.
spinosa may have arisen due to a common pattern of heterochrony.
All Acanthemblemaria species have spinous processes on the frontal
bones, but with differences in the degree of spination. A common
pathway could underlie the development of spines in all species
and different phenotypes arise due to differences in developmental
timing. In the case of A. maria and A. spinosa, hypermorphosis,
where there is a delay in the offset of a developmental process,
could give rise to the extreme spination found in these species.
As suggested by Emerson and Hastings (1998), this could be tested
by studying the ontogenetic trajectory of spine development in a
number of different Acanthemblemaria species to determine the
onset and offset of these traits.
4.2. Acanthemblemaria diversity
Our results demonstrate that Acanthemblemaria species diver-
sity is presently under-described. The molecular phylogenies in-
ferred in this study supported the inclusion of the undescribed
species from Isla Margarita (A.n.sp.) as belonging to Acanthembl-
emaria (Figs. 1A and 4). In addition, two other lineages were iden-
tified as possible undescribed taxa. The first represents a
population of A. rivasi from coastal Venezuela. Acero (1984) noted
diagnosable differences between A. rivasi populations from the
southern and southwestern Caribbean and those from Central
America, where the species was originally described by Stephens
(1970). Acero found that A. rivasi individuals from Colombia and
Venezuela have significantly different numbers of total dorsal fin
and segmented anal fin elements from those in Costa Rica and Pan-
ama. In addition, individuals from Venezuela have a pattern of
bright blue dots on the head that is less prominent in Central
American populations. These meristic and color differences be-
tween A. rivasi populations, together with the reciprocal mono-
phyly of Venezuelan and Panamanian A. rivasi populations based
on the concatenated dataset (Fig. 4), the population in coastal Ven-
ezuela likely represents an undescribed species.
Another undescribed species, sister to A. chaplini, was found in
the concatenated phylogeny. A. chaplini from New Providence,
Bahamas, was recovered as sister to A. chaplini individuals from
the Abacos in the Bahamas and Panama (Fig. 4). These last two
3.0
PLEISTP L I O C E N EM I O C E N EOLIGO
E. nigra
A. exilispinus
A. rivasi
C. lucasana
A. castroi
A. macrospilu
s
E. myersi
A. balanorum
A. paula
A. chaplini
A. greenfieldi
A. crockeri
A. spinosa
A. aspera
A. maria
P. bicirrus
A. n. sp.
A. hancocki
A. betinensis
A. medusa
012345678910111213141516171819202122232425
Fig. 3. Time-calibrated species tree, with branch lengths in units of millions of years. Support values for the species tree are the same as those from Figure 1A. Node bars
indicate the 95% upper and lower HPDs for node heights. The vertical dashed line indicates the final closure of the Isthmus of Panama, 3.1 mya.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 167

were separated from the New Providence individual by a long
branch, with a mean TMRCA of 5 my, which was deeper than that
of some nominal congeners (Table 2 and Fig. 4). This is despite the
much greater distance between the Abacos and Panama
(2000 km) than the Abacos and New Providence (130 km).
The Abacos and New Providence are separated by the deep waters
of the Northeast Providence Channel, which may help maintain the
deep genetic divergence between the two populations. However,
the Caribbean Sea between the Bahamas and Panama is not shal-
low, discounting the possibility that water depth alone is responsi-
ble for the isolation of these lineages.
Because New Providence is the type locality for A. chaplini
(Böhlke, 1957), the individuals from the Abacos and Panama
should be described as a new species. A species similar to A. chap-
lini,A. cubana, was recently described from Cuba (Garrido and
Varela, 2008). A. cubana lives in sympatry with A. chaplini on Cuban
reefs and is distinguished from the latter by slight differences in
papillae. Given the slight differences between A. cubana and A.
chaplini, it is not clear if the former is a valid species. However,
those subtle differences may represent a deeply divergent lineage,
such as the one we found in this study. Without further
examination it is difficult to determine the validity of A. cubana,
whether it represents one of the two lineages we have sampled
here, or if it belongs to a third, unsampled, lineage. To resolve this,
sympatric A. cubana and A. chaplini individuals should be collected
and analyzed genetically.
4.3. Biogeography and timing of speciation in Acanthemblemaria
Our divergence dating recovered a mid-Miocene origin for the
genus Acanthemblemaria and extant species pairs were found to
have diverged both before and after the closure of the Isthmus of
Panama (Figs. 3 and 4). In addition, we found that sister taxa had
a variety of geographic distributions, from broadly sympatric to
completely allopatric (Figs. 5 and Hastings, 2009).
The Isthmus of Panama has long been recognized as a major dri-
ver of allopatric marine speciation in the Neotropics (Hastings,
2000, 2009; Jordan, 1908; Knowlton et al., 1993; Lessios, 2008;
Lessios et al., 2001). However, its importance in the diversification
of reef fishes has not been consistent across groups. Taylor and
Hellberg (2005) found that for the Neotropical goby genus Elacati-
nus, the Isthmus of Panama was associated with two splits and that
Table 5
Estimated divergence times for selected nodes in the species tree (top) and concatenated tree (bottom). All times are in millions of years and bold values indicate splits inferred to
have occurred prior to the final closure of the Isthmus of Panama, 3.1 mya.
Node Mean Lower 95 HPD Upper 95 HPD
Species tree divergence times
Root 29.07 16.74 47.97
Acanthemblemaria 13.15 7.75 21.44
A. betinenis/A. exilispinus 4.16 3.1 6.29
A. castroi/A. rivasi 4.63 3.1 7.31
A. spinosa (A. aspera,A. paula)7.71 4.12 12.56
A. chaplini/A. greenfieldi 8.2 4.45 13.42
‘‘barnacle blennies’’ 9.63 5.47 15.73
A. aspera/A. paula 5.33 2.63 8.9
Clade I 10.53 6.14 17.25
Clade II 10.74 6.14 17.25
Ekemblemaria myersi/E. nigra 8.46 4.12 14.46
Cirriemblemaria lucasana/Protemblemaria bicirrus 12.08 6.04 20.44
A. maria/A. n. sp. 5.74 2.76 9.62
A. medusa, (A. maria/A. n. sp.) 8.31 4.53 13.57
‘‘hancocki species group’’ 3.91 1.94 6.5
A. balanorum/A. macrospilus 2.71 1.08 4.71
A. crockeri, (A. balanorum/A. macrospilus) 3.54 1.78 6.01
A. medusa, A. n. sp., A. maria, A. exiispinus, A. betinensis 9.6 5.54 15.38
Concatenated tree divergence times
Root 24.93 14.82 38.06
Acanthemblemaria 11.47 7.11 17.4
A. betinenis/A. exilispinus 3.96 3.1 5.72
A. spinosa (A. aspera,A. paula)7.17 4.25 10.98
‘‘barnacle blennies’’ 8.61 5.35 13.12
A. aspera/A. paula 5.1 2.94 7.99
Clade I 9.7 6.08 14.81
Clade II 9.4 5.79 14.24
Ekemblemaria myersi/E. nigra 7.49 3.97 11.96
Cirriemblemaria lucasana/Protemblemaria bicirrus 10.97 5.99 17.3
A. maria/A. n. sp. 5.46 3.06 8.55
A. medusa, (A. maria/A. n. sp.) 7.61 4.56 11.67
‘‘hancocki species group’’ 3.66 2.01 5.73
A. balanorum/A. macrospilus 2.61 1.32 4.16
A. crockeri, (A. balanorum/A. macrospilus) 3.2 1.73 5.03
A. medusa, A. n. sp., A. maria, A. exiispinus, A. betinensis 8.75 5.43 13.3
A. paula TMRCA 1.45 0.62 2.46
A. aspera TMRCA 0.58 0.13 1.16
A. spinosa TMRCA 1.61 0.7 2.7
A. maria TMRCA 1.14 0.47 1.96
A. medusa TMRCA 1.39 0.55 2.44
A. cf. chaplini TMRCA 0.72 0.21 1.37
A. rivasi/A. cf. rivasi 0.97 0.38 1.71
A. rivasi s.l./A. castroi 4.36 3.1 6.57
A. chaplini/A. cf. chaplini 5.06 2.77 7.95
A. chaplini s.l./A. greenfieldi 7.64 4.54 11.74
168 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

no sister taxa were transisthmian geminates. Instead, the Risor
clade was divided by the Isthmus, as was a basal Elacatinus species,
which was sister to the rest of the genus. Likewise, Rocha et al.
(2008) found that for Haemulon grunts there was limited support
for the Isthmus playing a role in generating diversity. A single pair
of geminate taxa was recovered in their analysis, while two pairs of
taxa proposed by Jordan (1908) to be geminates were not. How-
ever, they did recover sister clades sundered by the Isthmus (Rocha
et al., 2008). Craig et al. (2004) reported findings similar to Rocha
et al. and cautioned that inadequate taxon sampling my lead to
erroneous conclusions regarding the role of the closure of the Isth-
mus in recent speciation events.
Our results are similar to these three studies, but with a more
complicated pattern. We recovered two pairs of geminate taxa: A.
betinensis and A. exilispinus, and A. castroi and A. rivasi (Figs. 1, 3
and 4). Both pairs were sister to other clades or pairs of species,
and neither was basal in the phylogeny. We also recovered a basal
split in Clade I between A. greenfieldi and A. chaplini and the ‘‘han-
cocki species group’’. Therefore, the Caribbean taxa were not
monophyletic. This split in Clade I was quite old, with a mean
TMRCA of 10.5 my and 9.7 my, respectively, and matched the
TMRCA of Clade II (Table 2). The sister relationship between A.
greenfieldi and A. chaplini and the ‘‘hancocki species group’’ was
surprising, as they are well separated by morphology and by dis-
tribution (Hastings, 1990; Smith-Vaniz and Palacio, 1974). Given
the age of this split and difference between these species, Clade
I may have been larger in the past, with subsequent extinctions,
as suggested by the distributions of A. chaplini and A. greenfieldi
(see below).
Both Taylor and Hellberg (2005) and Rocha et al. (2008) found
that the majority of taxa in their studies diversified within ocean
basins. However, the geography of speciation differed between
Elacatinus and Haemulon.Taylor and Hellberg (2005) found that
Caribbean Elacatinus species diversified in allopatry and that sister
taxa had either allopatric or micro-allopatric distributions. In con-
trast, Rocha et al. (2008) found that most sister taxa and closely re-
lated species had sympatric distributions.
In this study, we found a combination of both patterns. The dis-
tributions of sister taxa and sister clades overlapped substantially
in some cases, while others were allopatric (Fig. 5). The three
Caribbean clades (A. spinosa, (A. aspera,A. paula); A. medusa, (A.
maria,A. n. sp.); A. chaplini,A. greenfieldi) varied in their extent of
3. 0
0123456789101112131415161718192021
A. maria BA
E. myersi PPAN
E. nigra CPAN
A. exilispinus PPAN
A. rivasi CPAN
C. lucasana MEX
A. crockeri MEX
A. castroi GAL
A. betinensis CPAN
A. macrospilus MEX
A. rivasi CPAN
A. maria BE
A. balanorum MEX
E. myersi SAL
A. hancocki PPAN
A. aspera HN
A. balanorum MEX
A. paula BE
A. paula BA
A. chaplini BA
A. greenfieldi BE
A. crockeri MEX
A. spinosa PR
A. greenfieldi BE
A. spinosa CU
A. paula BA
A. aspera SXM
A. maria STX
P. bicirrus MEX
A. n. sp. VE
A. maria CPAN
A. paula BE
A. spinosa SXM
A. medusa VE
A. cf. chaplini BA
A. cf. chaplini CPAN
A. exilispinus PPAN
A. maria SXM
A. hancocki CR
A. betinensis CPAN
A. spinosa HN
A. cf. rivasi VEN
A. spinosa VE
A. medusa SXM
A. macrospilus MEX
A. medusa CU
A. n. sp. VE
PLEISTP L I O C E N EM I O C E N E
Fig. 4. Time-calibrated Bayesian phylogeny of the concatenated dataset, with branch lengths in units of millions of years. Branches subtending nodes with <0.95 BPP are
light; all others are bold. Node bars indicate the upper and lower 95% HPDs for node heights and the vertical dashed line indicates the final closure of the Isthmus of Panama,
3.1 mya. Locality abbreviations are listed after species names and are as follows: BA: Bahamas, BE: Belize, CPAN: Caribbean Panama, CR: Costa Rica, CU: Curaçao, GAL:
Galapagos, HN: Honduras, MEX: Pacific Mexico, PPAN: Pacific Panama, PR: Puerto Rico, SAL: El Salvador, STX: St. Croix, SXM: Saint Maarten, VE: Venezuela. All locality
information can be found in the Appendix.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 169

range overlap (Fig. 5). The species in the A. spinosa, (A. aspera,A.
paula) clade had the largest degree of range overlap (Fig. 5A). A. as-
pera and A. spinosa co-occur over a large portion of their respective
ranges. A. paula was found in close sympatry with these species in
two locations: the Belizean barrier reef and New Providence in the
Bahamas. Since its description, A. paula has been considered a
A. chaplini
A. greenfieldi
A. paula
A. spinosa
A. aspera
B
C
A. maria
A. n. sp.
A. medusa
A
Fig. 5. The confirmed distributions (in yellow, blue, or a star for point localities) and degree of range overlap (in green) for the three Caribbean clades. (A) A. spinosa, (A. aspera,
A. paula). (B) A. medusa, (A. maria,Acan. n. sp.). (C) A. chaplini,A. greenfieldi. Distribution information comes from Smith-Vaniz and Palacio (1974),Dennis et al. (2004, 2005),
Hastings and Robertson (1999b), as well as subsequent examination of museum specimens and personal observations by RIE. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)
170 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

micro-endemic species, thought to only occur in a small area in Be-
lize (Hastings, 2009; Johnson and Brothers, 1989). The species is
very small (18 mm maximum standard length), lays few eggs (less
than five per brood), and is a habitat specialist (Clarke, 1994;
Greenfield and Greenfield, 1982; Johnson and Brothers, 1989), giv-
ing credence to the idea that its ability to colonize new regions is
poor. Here we document a 1500 km range extension for the spe-
cies, showing that A. paula’s distribution is much larger than previ-
ously thought.
A. aspera,A. paula, and A. spinosa demonstrate fine scale habitat
partitioning where they co-occur. In Belize, each species is found
on a different section of the reef, spanning a depth gradient from
<1–6 m in A. paula, 3–12 m in A. spinosa, and 5–22 m in A. aspera
(Clarke, 1994; Eytan and Hellberg, unpub. data). Where they co-oc-
cur, these species partition out the substrate by hole size, coral
type, and shelter height, in some cases all co-occurring on the same
stand of coral (Clarke, 1994; Eytan and Hellberg, unpub. data). This
fine scale partitioning could be an example of ecological character
displacement permitting co-existence of closely related species
(Bay et al., 2001; Robertson, 1996). Alternatively, these species
may have diverged in parapatry with disruptive selection due to
competition for shelters driving speciation. However, evidence to
support either hypothesis is lacking, and further study is war-
ranted to address this question.
The A. medusa, (A. maria,Acan. n. sp.) clade also had a broad dis-
tribution and often overlapping ranges, but sister taxa do not. A. n.
sp, recovered as sister to A. maria, has never been recorded east of
Isla Margarita (Ramjohn, 1999), nor has it been recorded as far
west as Los Roques, Venezuela (Cervigón, 1991), suggesting that
its distribution may be quite restricted, as it is only known from
a small area. Additional sampling may change this, though. While
its range is close to that of A. maria, the two taxa do not overlap,
but have abutting distributions (Fig. 5A).
The sister pair of A. chaplini and A. greenfieldi exist in complete
allopatry with disjunct ranges (Fig. 5C). A. chaplini is found in Flor-
ida and the Bahamas, as well as further south in Panama (Hastings
and Robertson, 1999b). Meanwhile, A. greenfieldi is found in the
central and western Caribbean, in between the two regions where
A. chaplini is found. A Panama–Florida distributional tract may not
be uncommon though, as it has been found in Elacatinus gobies
(Taylor and Hellberg, 2005, 2006), several peripheral freshwater
fishes (Gilmore and Hastings, 1983), and in the coral Acropora pal-
mata (Baums et al., 2005). These two species have the oldest diver-
gence time of any Acanthemblemaria sister taxa (Figs. 3 and 4,
Table 2). It may be that extensive extinctions have occurred since
these taxa split, perhaps in the eastern Caribbean or Caribbean
coast of South America, resulting in the observed allopatric
distributions.
In contrast to the old split between A. chaplini and A. greenfieldi,
the ‘‘hancocki species group’’ in the eastern Pacific is a young clade.
The mean TMRCA of the included taxa in the group was estimated
to be 3.91 or 3.66 my for the species tree and concatenated analy-
ses, respectively (Table 2). However, the lower 95% HPD was as
young as 1.9 mya. This suggests diversification of this species
group occurred after the closure of the Isthmus of Panama. The sis-
ter taxa in this group, A. macrospilus and A. balanorum occur in
sympatry in southern Mexico (Fig. 2.1.2 in Hastings, 2009). In the
Gulf of California, A. balanorum partially overlaps the range of the
recently described sister species of A. macrospilus (A. hastingsi Lin
and Galland, 2010; not included in this study). As in the A. spinosa,
(A. aspera,A. paula) clade, where A. hastingsi and A. balanorum co-
occur, they partition out the available habitat along a depth gradi-
ent (Lindquist, 1985).
Determining the geography of speciation for any taxonomic
group is difficult because current species distributions may not re-
flect those at the time of speciation (Losos and Glor, 2003). In the
case of Acanthemblemaria, this is exacerbated by evidence that
extinction (Clarke, 1996; Eytan and Hellberg, 2010), poorly known
geographic ranges (Dennis et al., 2004, 2005; Hastings and Robert-
son, 1999b; this study), and the presence of cryptic taxa (Hastings,
2009; Hastings and Springer, 2009a; Lin and Galland, 2010; this
study) may be common in this genus.
5. Conclusions
In this study, three lineages were recovered as possible new
species, which would bring the membership of the genus to 25
taxa and make Acanthemblemaria one of the most species-rich
clades of Neotropical reef fishes. We found that some of the head
spines characteristic of Acanthemblemaria have evolved repeatedly,
leading to conflict between the morphological and molecular phy-
logenies of the group. This was typified by A. spinosa and A. maria,
both of which have elaborate spinous processes, but were not
recovered as sister to each other in the molecular phylogenetic
analyses. Multiple skull bones appear to have evolved in concert,
perhaps due to selection acting on constrained developmental
pathways. Bayesian divergence dating found that the genus di-
verged in the mid-Miocene. A complex pattern of clades was recov-
ered, diverging both before and after the closure of the Isthmus of
Panama, almost entirely within present-day ocean basins. While
several clades have overlapping ranges, most sister taxa occur in
allopatry. The exception was the A. spinosa, (A. aspera,A. paula)
clade, which exists in sympatry. Fine scale habitat segregation
may allow for co-existence of these taxa, and warrants further
study.
Acknowledgements
We thank B. Smith-Vaniz and John Lundberg for allowing use
of images from the Proceeding of the Academy of Natural Sci-
ences of Philadephia. We thank J. Carlson for providing illustra-
tions. We thank R. Clarke, B. Holt, and E. Whiteman for field
collections. We thank D. Warren, A. Caballaro, R. Colin, G. Ja-
come, Joe Neigel and the Neigel lab, Howard Lasker and the Las-
ker Lab, the crew of the R.V. Walton Smith, and the staffs at
Nature Foundation SXM and the STRI Bocas del Toro Research
Station, and Dolphin Encounters in Nassau for providing help
in the field. We thank G. Bernardi, J. Wares, and two anonymous
reviewers for comments that significantly improved the manu-
script. We thank the governments of the Bahamas, St. Maarten,
Belize, Panama, and Curacao for providing research permits in
order to collect samples. Support for this research was provided
to RIE by LSU BioGrads, the McDaniel Travel Fund, an LSU
Graduate School travel award, a Sigma Xi Grant in Aid of
Research Award, an ASIH Raney Fund Award, a Lerner-Gray
Grant for Marine Research, a STRI short-term fellowship, and
the Association of the Marine Labs of the Caribbean, and to
MEH by the National Science Foundation (OCE-0550270 to
MEH and Iliana Baums). This study was conducted under IACUC
protocol No. 05-040 at Louisiana State University. Portions of
this research were conducted with high performance computa-
tional resources provided by the Louisiana Optical Network
Initiative (http://www.loni.org).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ympev.2011.09.028.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 171

References
Acero, A.P., 1984. The Chaneopsine blennies of the southwestern Caribbean (Pisces:
Clinidae: Chaenopsinae). II. The genera Acanthemblemaria,Ekemblemaria and
Lucayablennius. Revista de Biologia Tropical 32, 35–44.
Almany, G.R., Baldwin, C.C., 1996. A new Atlantic species of Acanthemblemaria
(Teleostei: Blennioidei: Chaenopsidae): morphology and relationships.
Proceedings of the Biological Society of Washington 109, 419–429.
Baums, I.B., Miller, M.W., Hellberg, M.E., 2005. Regionally isolated populations of an
imperiled Caribbean coral, Acropora palmata. Molecular Ecology 14, 1377–1390.
Bay, L.K., Jones, G.P., McCormick, M.I., 2001. Habitat selection and aggression as
determinants of spatial segregation among damselfish on a coral reef. Coral
Reefs 20, 289–298.
Bellwood, D.R., Wainwright, P.C., 2002. The history and biogeography of fishes on
coral reefs. In: Sale, P.F. (Ed.), Coral Reef Fishes: Dynamics and Diversity in a
Complex Ecosystem. Academic Press, San Diego, pp. 5–32.
Böhlke, J., 1957. The Bahaman species of emblemariid blennies. Proceedings of the
Academy of Natural Sciences of Philadelphia 109, 25–60.
Böhlke, J.E., 1961. The Atlantic species of the Clinid fish genus Acanthemblemaria.
Notulae Naturae of the Academy of Natural Sciences of Philadephia 346, 1–7.
Böhlke, J.E., Chaplin, C.G.C., 1993. Fishes of the Bahamas and Adjacent Tropical
Waters. University of Texas Press, Austin.
Briggs, J.C., 1974. Marine Zoogeography. McGraw-Hill Book Company, New York.
Cervigón, F., 1991. Los peces marinos de Venezuela. Fundación Científica Los
Roques.
Clarke, R.D., 1989. Population fluctuation, competition and microhabitat
distribution of 2 species of tube blennies, Acanthemblemaria (Teleostei,
Chaenopsidae). Bulletin of Marine Science 44, 1174–1185.
Clarke, R.D., 1994. Habitat partitioning by chaenopsid blennies in Belize and the
Virgin Islands. Copeia, 398–405.
Clarke, R.D., 1996. Population shifts in two competing fish species on a degrading
coral reef. Marine Ecology – Progress Series 137, 51–58.
Clarke, R.D., Buskey, E.J., Marsden, K.C., 2005. Effects of water motion and prey
behavior on zooplankton capture by two coral reef fishes. Marine Biology 146,
1145–1155.
Clarke, R., Finelli, C., Buskey, E., 2009. Water flow controls distribution and feeding
behavior of two co-occurring coral reef fishes: II. Laboratory experiments. Coral
Reefs 28, 475–488.
Craig, M.T., Hastings, P.A., Pondella, D.J., 2004. Speciation in the Central American
Seaway: the importance of taxon sampling in the identification of trans-
isthmian geminate pairs. Journal of Biogeography 31, 1085–1091.
Dennis, G.D., Hensley, D., Colin, P.L., Kimmel, J.J., 2004. New records of marine fishes
from the Puerto Rican Plateau. Caribbean Journal of Science 40, 70–87.
Dennis, G.D., Smith-Vaniz, W.F., Colin, P.L., Hensley, D.A., McGehee, M.A., 2005.
Shore fishes known from islands of the Mona Passage, Greater Antilles with
comments on their zoogeography. Caribbean Journal of Science, 41.
Drummond, A.J., Ho, S.Y.W., Phillips, M.J., Rambaut, A., 2006. Relaxed phylogenetics
and dating with confidence. Plos Biology 4, 699–710.
Drummond, A.J., Ashton, B., Cheung, M., Heled, J., Kearse, M., Moir, R., Stones-Havas,
S., Thierer, T., Wilson, A., 2007. Geneious v3.6. <http://www.geneious.com>.
Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and
high throughput. Nucleic Acids Research 32, 1792–1797.
Emerson, S.B., 1982. Frog postcranial morphology: identification of a functional
complex. Copeia 1982, 603–613.
Emerson, S.B., Hastings, P.A., 1998. Morphological correlations in evolution:
consequences for phylogenetic analysis. The Quarterly Review of Biology 73,
141–162.
Eytan, R.I., 2010. Population Genetics and Systematics of a Species-Rich Clade of
Neotropical Reef Fishes, the Tube Blenny Genus Acanthemblemaria (Teleostei:
Blennioidei). Ph.D. Dissertation. Louisiana State University, Baton Rouge,
Louisiana.
Eytan, R.I., Hellberg, M.E., 2010. Nuclear and mitochondrial sequence data reveal
and conceal different demographic histories and population genetic processes
in Caribbean reef fishes. Evolution 64, 3380–3397.
Finelli, C., Clarke, R., Robinson, H., Buskey, E., 2009. Water flow controls distribution
and feeding behavior of two co-occurring coral reef fishes: I. Field
measurements. Coral Reefs 28, 461–473.
Garrido, O.H., Varela, C., 2008. Sobre la especie Acanthemblemaria chaplini (Pisces:
Chaenopsidae) en Cuba, con la descripción de una especie nueva. Solenodon 7,
29–36.
Gilmore, G.R., Hastings, P.A., 1983. Observations on the ecology and distribution of
certain tropical peripheral fishes in Florida. Florida Scientist 46, 31–51.
Greenfield, D.W., 1981. The blennioid fishes of Belize and Honduras, Central
America, with comments on their systematics, ecology, and distribution
(Blennidae, Chaenopsidae, Labrisomidae, Tripterygiidae). Fieldiana Zoology, 1–
106.
Greenfield, D.W., Greenfield, T.A., 1982. Habitat and resource partitioning between
two species of Acanthemblemaria (Pisces: Chaenopsidae), with comments on the
chaos hypothesis. The Atlantic Barrier Reef Ecosystem at Carrie Bow Cay, Belize,
1: Structure and Communities, pp. 499–507.
Greenfield, D.W., Johnson, R.K., 1990. Community structure of western Caribbean
blennioid fishes. Copeia, 433–448.
Hastings, P.A., 1990. Phylogenetic relationships of tube blennies of the genus
Acanthemblemaria (Pisces, Blennioidei). Bulletin of Marine Science 47, 725–
738.
Hastings, P.A., 2000. Biogeography of the tropical Eastern Pacific: distribution and
phylogeny of chaenopsid fishes. Zoological Journal of the Linnean Society 128,
319–335.
Hastings, P.A., 2009. Biogeography of new world blennies. In: Patzner, R.A.,
Gonçalves, E.J., Hastings, P.A., Kapoor, B.J. (Eds.), The Biology of Blennies.
Science Publisher, Enfield, New Hampshire, pp. 95–118.
Hastings, P.A., Galland, G.R., 2010. Ontogeny of microhabitat use and two-step
recruitment in a specialist reef fish, the Browncheek Blenny (Chaenopsidae).
Coral Reefs 29, 155–164.
Hastings, P.A., Robertson, D.R., 1999a. Acanthemblemaria atrata and
Acanthemblemaria mangognatha, new species of eastern Pacific barnacle
blennies (Chaenopsidae) from Isla del Coco, Costa Rica, and Islas
Revillagigedo, Mexico, and their relationships with other barnacle blennies.
Revue Français d’Aquariologie 25 (1998), : 107–118.
Hastings, P.A., Robertson, D.R., 1999b. Notes on a collection of chaenopsid blennies
from Bahia Azul, Bocas del Toro, Caribbean, Panama. Revue Francaise
d’Aquariologie Herpetologie (Nancy) 26, 33–38.
Hastings, P.A., Springer, V.G., 1994. Review of Stathmonotus, with redefinition and
phylogenetic analysis of the Chaenopsidae (Teleostei: Blennioidei). Smithsonian
Contributions to Zoology, 1–48.
Hastings, P.A., Springer, V.G., 2009a. Recognizing diversity in blennioid fish
nomenclature (Teleostei: Blennioidei). Zootaxa 2120, 3–14.
Hastings, P.A., Springer, V.G., 2009b. Systematics of the Blenniodei and the included
families Dactyloscopidae, Chaenopsidae, Clinidae and Labrisomidae. In: Patzner,
R.A., Gonçalves, E.J., Hastings, P.A., Kapoor, B.J. (Eds.), The Biology of Blennies.
Science Publisher, Enfield, New Hampshire, pp. 3–30.
Heled, J., Drummond, A.J., 2010. Bayesian inference of species trees from multilocus
data. Molecular Biology and Evolution 27, 570–580.
Hickerson, M.J., Cunningham, C.W., 2005. Contrasting quaternary histories in an
ecologically divergent sister pair of low-dispersing intertidal fish (Xiphister)
revealed by multilocus DNA analysis. Evolution 59, 344–360.
Holland, B.R., Spencer, H.G., Worthy, T.H., Kennedy, M., 2010. Identifying cliques of
convergent characters: concerted evolution in the cormorants and shags.
Systematic Biology 59, 433–445.
Huson, D.H., Bryant, D., 2006. Application of phylogenetic networks in evolutionary
studies. Molecular Biology and Evolution 23, 254.
Johnson, G.D., Brothers, E.B., 1989. Acanthemblemaria paula, a new diminutive
chaenopsid (Pisces, Blennioidei) from Belize, with comments on life history.
Proceedings of the Biological Society of Washington 102, 1018–1030.
Jordan, D.S., 1908. The law of geminate species. American Naturalist 42, 73–80.
Kass, R.E., Raftery, A.E., 1995. Bayes factors. Journal of the American Statistical
Association 90, 773–795.
Kluge, A., 1989. A concern for evidence and a phylogenetic hypothesis of
relationships among Epicrates (Boidae, Serpentes). Systematic Zoology 38, 7–25.
Knowlton, N., Weigt, L.A., Solorzano, L.A., Mills, D.K., Bermingham, E., 1993.
Divergence in proteins, mitochondrial-dna, and reproductive compatibility
across the Isthmus of Panama. Science 260, 1629–1632.
Lessios, H., 2008. The great American schism: divergence of marine organisms after
the rise of the Central American Isthmus. Annual Review of Ecology, Evolution,
and Systematics 39, 63–91.
Lessios, H.A., Kessing, B.D., Pearse, J.S., 2001. Population structure and speciation in
tropical seas: global phylogeography of the sea urchin Diadema. Evolution 55,
955–975.
Lewis, P.O., 2001. A likelihood approach to estimating phylogeny from discrete
morphological character data. Systematic Biology 50, 913–925.
Li, C.H., Orti, G., Zhang, G., Lu, G.Q., 2007. A practical approach to phylogenomics:
the phylogeny of ray-finned fish (Actinopterygii) as a case study. BMC
Evolutionary Biology, 7.
Lin, H.-C., Galland, G.R., 2010. Molecular analysis of Acanthemblemaria macrospilus
(Teleostei: Chaenopsidae) with description of a new species from the Gulf of
California, Mexico. Zootaxa 2525, 51–62.
Lin, H.C., Hastings, P.A., 2011. Evolution of a neotropical marine fish lineage
(subfamily Chaenopsinae, Suborder Blennioidei) based on phylogenetic analysis
of combined molecular and morphological data. Molecular Phylogenetics and
Evolution 60, 236–248.
Lindquist, D.G., 1985. Depth zonation, microhabitat, and morphology of 3 species of
Acanthemblemaria (Pisces, Blennioidea) in the Gulf of California, Mexico.
Marine Ecology – Pubblicazioni Della Stazione Zoologica Di Napoli I 6, 329–344.
Lindquist, D.G., Kotrschal, K.M., 1987. The diets in 4 pacific tube blennies
(Acanthemblemaria, Chaenopsidae) – lack of ecological divergence in syntopic
species. Marine Ecology – Pubblicazioni Della Stazione Zoologica Di Napoli I 8,
327–335.
Losos, J.B., Glor, R.E., 2003. Phylogenetic comparative methods and the geography of
speciation. Trends in Ecology & Evolution 18, 220–227.
MacPherson, E., Hastings, P.A., Robertson, D.R., 2009. Macroecology of marine fishes.
In: Witman, J., Roy, K. (Eds.), Marine Macroecology. University of Chicago Press.
McCracken, K.G., Harshman, J., McClellan, D.A., Afton, A.D., 1999. Data set
incongruence and correlated character evolution: an example of functional
convergence in the hind-limbs of stifftail diving ducks. Systematic Biology 48,
683–714.
Metzelaar, J., 1919. Report on the fishes collected by Dr. J. Boeke in the Dutch West
Indies 1904–1905. With comparative notes on marine fishes of tropical West
Africa. Rapport Visscherij en de industrie van Zeeprodukten in de Kolonie
Curaçao, Gravenhage.
Mora, C., Chittaro, P.M., Sale, P.F., Kritzer, J.P., Ludsin, S.A., 2003. Patterns and
processes in reef fish diversity. Nature 421, 933–936.
172 R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173

Near, T.J., Bolnick, D.I., Wainwright, P.C., 2004. Investigating phylogenetic
relationships of sunfishes and black basses (Actinopterygii: Centrarchidae)
using DNA sequences from mitochondrial and nuclear genes. Molecular
Phylogenetics and Evolution 32, 344–357.
Nylander, J., Wilgenbusch, J., Warren, D., Swofford, D., 2008. AWTY (are we there
yet?): a system for graphical exploration of MCMC convergence in Bayesian
phylogenetics. Bioinformatics 24, 581.
Posada, D., 2008. jModelTest: phylogenetic model averaging. Molecular Biology and
Evolution 25, 1253–1256.
Rambaut, A., Drummond, A.J., 2010. Tracer v1.5. <http://beast.bio.ed.ac.uk/Tracer>.
Ramjohn, D., 1999. Checklist of coastal and marine fishes of Trinidad and Tobago.
Marine Fishery Analysis Unit, Fisheries Division, Ministry of Agriculture, Land
and Marine Resources, Trinidad and Tobago. Fisheries Information Series 8, p.
151.
Robertson, D.R., 1996. Interspecific competition controls abundance and habitat use
of territorial Caribbean damselfishes. Ecology 77, 885–899.
Rocha, L.A., Lindeman, K.C., Rocha, C.R., Lessios, H.A., 2008. Historical biogeography
and speciation in the reef fish genus Haemulon (Teleostei: Haemulidae).
Molecular Phylogenetics and Evolution 48, 918.
Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference
under mixed models. Bioinformatics 19, 1572–1574.
Sale, P.F., 2002. Coral Reef Fishes: Dynamics and Diversity in a Complex Ecosystem.
Academic Press, San Diego, California.
Schlosser, G., Wagner, G.P., 2004. Modularity in Development and Evolution.
University of Chicago Press.
Shaffer, H.B., Clark, J.M., Kraus, F., 1991. When molecules and morphology clash: a
phylogenetic analysis of the North American ambystomatid salamanders
(Caudata: Ambystomatidae). Systematic Zoology 40, 284–303.
Smith-Vaniz, W.F., Palacio, F.J., 1974. Atlantic fishes of the genus Acanthemblemaria,
with description of three new species and comments on Pacific species
(Clinidae: Chaenopsinae). Proceedings of the Academy of Natural Sciences of
Philadelphia 125, 197–224.
Stephens, J.S., 1963. A revised classification of the blennioid fishes of the American
family Chaenopsidae. University of California Publications in Zoology 68, 1–165.
Stephens, J.S., 1970. Seven new chaenopsid blennies from the Western Atlantic.
Copeia, 280–309.
Suchard, M.A., Redelings, B.D., 2006. BAli-Phy: simultaneous Bayesian inference of
alignment and phylogeny. Bioinformatics 22, 2047–2048.
Suchard, M., Weiss, R., Sinsheimer, J., 2001. Bayesian selection of continuous-time
Markov chain evolutionary models. Molecular Biology and Evolution 18, 1001.
Taylor, M.S., Hellberg, M.E., 2005. Marine radiations at small geographic scales:
speciation in neotropical reef gobies (Elacatinus). Evolution 59, 374–385.
Taylor, M.S., Hellberg, M.E., 2006. Comparative phylogeography in a genus of coral
reef fishes: biogeographic and genetic concordance in the Caribbean. Molecular
Ecology 15, 695–707.
Werner, R.K., Hoernle, K., 2003. New volcanological and volatile data provide strong
support for the continuous existence of Galápagos Islands over the past 17
million years. International Journal of Earth Sciences 92, 904–911.
West-Eberhard, M.J., 2003. Developmental Plasticity and Evolution. Oxford
University Press, USA.
R.I. Eytan et al. / Molecular Phylogenetics and Evolution 62 (2012) 159–173 173