Taxon sampling.—Sixty-three taxa were analyzed in the current phylogenetic study. To provide a test of scorpaenoid monophyly and relationships, 12 outgroups were included: Anthiadidae, Bathymasteridae, Bembropidae, Bovichtidae, Cirrhitidae, Epinephelidae, Niphonidae, Percidae, Psychrolutidae, Serranidae, and Trachinidae. The topology was rooted with Cirrhitus (Centrarchiformes), which has long been closely allied with, but excluded from, the mail-cheeked fishes (Gill, 1888; Smith and Wheeler, 2004; Near et al., 2013). The 51 traditional scorpaenoid terminals analyzed in this study included all 20 scorpaenoid families (clade sensu Imamura, 2004; classification sensu Eschmeyer et al., 2017).

Character sampling.—A total of 5,393 morphological and molecular characters were scored. These data included 5,280 aligned nucleotides from five nuclear and three mitochondrial loci. Previous studies have shown the effectiveness of the included molecular loci for resolving relationships among percomorphs generally or mail-cheeked fishes specifically (e.g., Chakrabarty et al., 2011a, 2011b; Li et al., 2011; Near et al., 2012a, 2012b; Girard and Smith, 2016). The molecular terminals analyzed in the present study and GenBank accession numbers corresponding to the gene fragments sequenced are listed in Supplemental Table 1 (see Data Accessibility). For these analyses, the 134 novel DNA sequences were combined with previously published DNA sequences from the following sources: Smith and Wheeler (2004, 2006); Smith and Craig (2007); Smith et al. (2009, 2016); Near et al. (2012a, 2013, 2015); Wainwright et al. (2012). The molecular matrix was 90% complete at the amplicon level and 82% complete at the cell or individual-base-pair level (Supplemental Table 1; see Data Accessibility). These molecular data were simultaneously analyzed with a morphological dataset (Supplemental Table 2; see Data Accessibility; Appendix 1) composed of 113 characters that was built from multiple sources, but focused on the work of Imamura (2004). Details about additional sources of morphological data are listed along with all character descriptions in Appendix 1. The morphological matrix was 92% complete at the cell or individual character level (Supplemental Table 2; see Data Accessibility).

Acquisition of nucleotide sequences.—Fish tissues were preserved in 95% ethanol prior to extraction of DNA. Nuclear and mitochondrial DNA was extracted from muscle using a DNeasy Tissue Extraction Kit (Qiagen). The polymerase chain reaction (PCR) was used to amplify all gene fragments. Double-stranded amplifications were performed in a 25 μL volume containing one Ready-To-Go PCR bead (GE Healthcare), 1.25 μL of each primer (10 pmol), and 2–5 μL of undiluted DNA extract. Primers and primer sources are listed in Table 1. Amplifications for all novel fragments except Glyt and zic1 were carried out using the following temperature profile: initial denaturation for 360 sec at 94°C; 36 cycles of denaturation for 60 sec at 94°C, annealing for 60 sec at 46–51°C (see Table 1 for primary annealing temperature for each locus), and extension for 75 sec at 72°C; with a final terminal extension for 360 sec at 72°C. For Glyt and zic1, the following temperature profile was used: initial denaturation for 180 sec at 94°C; ten cycles of denaturation for 45 sec at 94°C, annealing for 45 sec at 57–58°C (see Table 1 for core annealing temperature for each locus), and extension for 75 sec at 72°C; 30 cycles of denaturation for 45 sec at 94°C, annealing for 30 sec at 55–57°C (see Table 1 for secondary annealing temperature for each locus), and extension for 75 sec at 72°C, with a final terminal extension for 360 sec at 72°C. The double-stranded amplification products were desalted and concentrated using AMPure (Beckman Coulter). Both strands of the purified PCR fragments were used as templates and amplified for sequencing using the amplification primers and a Prism Dye Terminator Reaction Kit v1.1 (Applied Biosystems) with minor modifications to the manufacturer's protocols. The sequencing reactions were cleaned and desalted using cleanSEQ (Beckman Coulter). The nucleotides were sequenced and the base pairs were called on a 3730 automated DNA sequencer (Applied Biosystems) or by Beckman Coulter Genomics (Danvers, MA). Contigs were built in Geneious v8.1.8 (Kearse et al., 2012) using DNA sequences from the complementary heavy and light strands. Sequences were edited in Geneious and collated into fasta text files. The novel sequences were submitted to GenBank and assigned accession numbers MF966393–MF966401 and MF991301–MF991425.

Table 1. 

Primers, PCR conditions, and substitution models for each amplicon analyzed in the current study.

Morphological investigation.—Ethanol preserved, cleared and stained, and dried skeletal material was examined using several stereomicroscopes with varying magnification and lighting regimes. Photographs documenting the anatomy of specimens were taken under normal visible lighting with either a Nikon D800 with an AF-S VR Micro-NIKKOR 105mm f/2.8G IF-ED lens or a Lumenera INFINITY2-5 digital CCD camera attached to a Nikon SMZ-18 stereomicroscope. Fluorescent images were taken either using the D800 with lighting from two NightSea BlueStar flashlights (Lexington, MA) and filtered through the Nikon light-shading plate from the Nikon SMZ-18 microscope or using the Nikon SMZ-18 stereomicroscope with either a Nikon P2-EFL GFP-B or P2-ELF RFP filter cube. Drawings of specimens were aided with a camera lucida arm attachment.

Phylogenetic analyses.—A likelihood analysis was used to analyze the molecular and morphological data. For this analysis, each of the eight amplicons was aligned individually in MAFFT v7.017 (Katoh et al., 2002) using default values. The maximum-likelihood dataset was broken into 18 partitions. Two partitions were designated for the mitochondrial (12S, tRNA-Val-16S, and 16Sar-br) and nuclear (28S) ribosomal fragments. Fifteen partitions covered the three codon positions in each of the five protein-coding genes: mitochondrial (cytochrome oxidase I) and nuclear (Glyt, histone H3, TMO-4c4, and zic1). The 18th partition was the morphological dataset (Supplemental Table 2; see Data Accessibility). The optimal nucleotide substitution model for each molecular partition was determined empirically (Table 1) by comparing different models under an Akaike information criterion as executed in jModelTest v0.1 (Posada, 2008), and the morphological model was set to Mkv (Lewis, 2001). The datasets were coded, concatenated, examined, and analyzed (ancestral-state reconstructions) in Mesquite v3.04 (Maddison and Maddison, 2017). Maximum-likelihood analyses were conducted in GARLI v2.01 (Zwickl, 2006). The tree with the best likelihood score from 40 independent analyses was selected as the preferred hypothesis. A non-parametric maximum-likelihood bootstrap analysis was conducted for 200 random pseudoreplicates to assess nodal support. For the support analysis, terminals represented solely by morphological data were excluded (Eschmeyer and Perryena) because they have excessive missing data in the combined analysis; the optimal tree with these terminals excluded was otherwise identical to the complete phylogeny (results not shown). We recognize two levels of nodal support: ≥70% bootstrap support represents a moderately supported node or clade and ≥95% bootstrap support represents a well-supported node or clade.

Phylogenetic analyses.—The likelihood analysis resulted in a single optimal tree (Fig. 3). Most of the 59 nodes recovered in the bootstrap analysis (analysis excluding Eschmeyer and Perryena) were moderately to well supported with 36 (61%) nodes being moderately supported by a bootstrap value ≥70 and 24 (41%) nodes being well supported by a bootstrap value ≥95. In this analysis, the traditional cottoid and zoarcoid representatives were nested within the Scorpaenoidea (sensu Imamura, 2004), sister to the Congiopodidae. This congiopodid + zoarcoid + cottoid clade was recovered as the sister group to all non-congiopodid scorpaenoid fishes. Many traditionally recognized family-level clades (sensu Eschmeyer et al., 2017) that were represented by more than a single terminal were recovered as monophyletic: Congiopodidae, Neosebastidae, Pataecidae, Peristediidae, Platycephalidae, Setarchidae, and Synanceiidae. In contrast, the Bembridae, Scorpaenidae, Sebastidae, Tetrarogidae, and Triglidae were recovered as para- or polyphyletic. As has been seen in previous morphological and molecular analyses, the core scorpionfishes included Caracanthus (Smith and Wheeler, 2004, 2006; Shinohara and Imamura, 2005; Smith and Craig, 2007; Lautredou et al., 2013) and the traditional Setarchidae (Imamura, 2004; Smith and Wheeler, 2004, 2006; Shinohara and Imamura, 2005). Similarly, hoplichthyids were recovered as more closely related to the Triglidae + Peristediidae than to platycephalids (see also Imamura, 1996, 2004; Lautredou et al., 2013). In contrast to morphological phylogenies (Ishida, 1994; Imamura, 2004) and some molecular phylogenies (Lautredou et al., 2013), but in agreement with other molecular phylogenies (e.g., Smith and Wheeler, 2004, 2006), the live-bearing rockfishes (e.g., Helicolenus, Sebastes) were nested deeply in the scorpionfishes as opposed to a stem grade. Despite these similarities to other studies, the combined analysis of morphological and molecular data is frequently in conflict with the traditional taxonomy and prior phylogenetic work. The traditional Peristediidae was nested within the Triglidae (see also Portnoy et al., 2017), Bembradium was allied with Plectrogenium (see also Imamura, 1996, 2004), Tetrarogidae was widely polyphyletic within a “stonefish” clade (see also Mandrytza, 2001; Smith and Wheeler, 2004; Honma et al., 2013), and representatives of the Scorpaenidae and Sebastidae were variously distributed within a “scorpionfish” clade. If we compare our results to a recent classification of fishes (Betancur-R. et al., 2017), we note that their Percoidei, Platycephaloidei, Serranoidei, Scorpaenidae, Serranidae, Tetrarogidae, and Triglidae were recovered as polyphyletic. Their Scorpaenoidei was the only relevant higher-level grouping with multiple families that was recovered as monophyletic in our analysis. The differences between our results and previous hypotheses necessitate the familial taxonomic revisions presented herein (Fig. 3; Appendix 2). All of the remaining results and discussion below will use the revised classification unless noted otherwise.

Fig. 3. 

Optimal cladogram resulting from the partitioned likelihood analysis of the dataset composed of 113 phenotypic and 5,280 nucleotide characters. Clades with 50% bootstrap support are retained and identified with their support. Nodes with bootstrap support of 95% were marked with an “*”. Family-level classification is designated on the right. Dashed branches indicate terminals that were represented only by morphological data that were excluded from the bootstrap analyses. Family-level-phylogeny of scorpaenoids and their sister group in gray box in upper-right corner.

Description of the lachrymal saber.—The lachrymal saber is a complex feature that involves a series of modifications to the circumorbitals, maxilla, adductor mandibulae, and associated tendons. Among mail-cheeked fishes, there are some modifications to these lachrymal-saber elements (e.g., adductor mandibulae variation as illustrated by Yabe [1985] and Mandrytza [1990]), but we treat this element as a single feature in this study. Future studies looking broadly at the phylogeny of mail-cheeked fishes should consider the historical independence of several aspects of the lachrymal saber to assess their individual phylogenetic significance. For example, many triglids have additional tendinous connections between the adductor mandibulae and maxilla that appear anatomically similar and potentially convergent with some synanceiids while lacking the remainder of the lachrymal-saber modifications (Yabe, 1985; Mandrytza, 1990; pers. obs.).

As expected, the circumorbitals of synanceiids lie lateral to the suspensorium (Figs. 4, 5). As with most mail-cheeked fishes, the third circumorbital is firmly attached to the preopercle posteriorly forming the characteristic suborbital stay. The circumorbitals are attached anteroventrally to the maxilla and anterodorsally to the lateral ethmoid via the lachrymal. Both anterior lachrymal attachments have substantive soft-connective tissue, which enables the necessary rotation of the lachrymal saber while also holding the bones in close association. The circumorbitals of mail-cheeked fishes frequently have elongations of their lateral-line canal pores that create the group's characteristic bony spines and knobs (Smith, 2005). In all species with a lachrymal saber, the ventral spines associated with the second pore on the first circumorbital (lachrymal) are enlarged (Figs. 4–7). As noted in revisions and artificial keys separating various mail-cheeked fishes (e.g., Poss, 1999), the lachrymal is highly mobile and hinged to the lateral ethmoid dorsally in synanceiids. As the element is relatively free to rotate, it abuts, but is not as firmly bound to, the second circumorbital as is otherwise common among scorpaenoid fishes. This freedom allows for the rotation of this element such that the typical ventral lachrymal spines, including the larger spine, can be rotated laterally and projected outward making the “saber” (Fig. 5).

Fig. 4. 

Lateral view of a cleared-and-stained specimen of the synanceiid Paracentropogon, CAS_SU 68769. Images highlight the (A) resting position of the lachrymal saber (arrow) along the side of the waspfish's cheek and the (B) locked-out position where the lachrymal saber extends laterally from the specimen. The rotation of both the first and second circumorbital are visible in the lower image.

Fig. 5. 

Lateral (A), rostral (B), and dorsal (C) views of the lachrymal saber in the Soldierfish (Gymnapistes marmoratus, AMNH 31009). Arrow highlights both the resting and locked lachrymal saber in the various angles.

The lachrymal saber can be held in an outwardly projected or lateral position by a locking mechanism that functions somewhat like a ratchet and pawl that relies on the friction between highly textured (often ridged) modifications on both the medial surface of the lachrymal and the lateral surface of the maxilla. These modifications include a bony spur that ranges from a peg to a cup on the inner surface of the lachrymal (Fig. 6A) and a ridged bony protuberance on the anterolateral end of the maxilla (Fig. 6B). The lachrymal's medial projection acts as the pawl when the saber is extended laterally. This lachrymal pawl, when in a locked position, is tightly connected to a highly sculptured protuberance on the anterior end of the maxilla that acts somewhat like a ratchet allowing the lachrymal to be held firmly in place (Fig. 7). While the lachrymal is free to rotate, its motion is highly constrained by a notable amount of soft connective tissue.

Fig. 6. 

Skeletal images of the key lachrymal saber components. (A) Lateral image of the lachrymal, second circumorbital (CO2), and third circumorbital (CO3) in Minous quincarinatus, KUI 41397. Circular inset of upper image shows medial view with medial protuberance (PP). (B) Lateral image of the maxilla in Apistus carinatus, KUI 41400. Circular inset of lower image shows closeup of the maxillary protuberance (MP). These two protuberances interact to lock the lachrymal saber.

Fig. 7. 

Composite dorsal images of various specimens of Paracentropogon highlighting the morphology of the components of the lachrymal saber with the locked position on the left side and the resting position on the right side. Note that the anteriormost component of the palatine has been made largely transparent in the images to allow for a better examination of the maxilla. The upper image (A) represents the anatomy with all elements marked, and the lower image (B) represents the image with the circumorbitals shadowed to visualize underlying muscles. Muscle attachments on removed bones are denoted with pale circles at the interface. Abbreviations: A1 or A2 represent the major subdivisions of the adductor mandibulae; CO3–circumorbital 3; CO4–circumorbital 4; LAP–levator arcus palatini; MP–maxillary protuberance. Illustration by Clara Richardson.

Manipulations of ethanol preserved and cleared and stained specimens indicate that the lachrymal saber is likely rotated and controlled by movements of the maxilla. The maxilla can be rotated by contractions of the various subunits of the adductor mandibulae A₁ division that have diversified among synanceiids. These separate muscle subdivisions can facilitate the engagement of the locking mechanism as well as outward rotation of the lachrymal (Figs. 4, 5) and, to a lesser degree, the second circumorbital (Fig. 4B). The number, size, and placement of adductor mandibulae A₁ division subunits and their associated tendinous connections vary tremendously among synanceiids (Fig. 8; Mandrytza, 1990). The traditional adductor mandibulae A₁ division in synanceiids has between two and five discernible subunits that variously originate caudally from the preopercle, the suborbital stay of the third circumorbital, the ventral surface of the third circumorbital, and/or the ventral surface of the second circumorbital. These muscles attach anteriorly to the maxilla via a diversity of tendons (Figs. 7, 8; Yabe, 1985; Mandrytza, 1990; Ishida, 1994). As these tendons can be homologous with ligaments in other fish groups (Johnson and Patterson, 2001; Datovo and Vari, 2013), individual representatives of these elements are often referred to as ligaments. As emphasized in Yabe (1985) and Mandrytza (1990), the proliferation of adductor mandibulae A₁ division subunits and their associated tendons/ligaments in synanceiids requires a more taxonomically comprehensive study across percomorphs generally and scorpaenoids specifically before the specific homology of the elements can be determined.

Fig. 8. 

Lateral view of the non-circumorbital components of the lachrymal saber in composite images of the various specimens of Paracentropogon above (A) and the various specimens of Apistus below (B). The comparisons are shown to illustrate variation in this system between species (e.g., subdivisions and attachment points of the adductor mandibulae). Muscle attachments on removed bones are denoted with pale circles at the interface. Abbreviations: A1 or A2 represent the major subdivisions of the adductor mandibulae; LAP–levator arcus palatini; MP–maxillary protuberance. Illustration by Clara Richardson.

Many aspects of the lachrymal saber vary: the number of adductor mandibulae A₁ division subunits, the caudal origination locations of the adductor mandibulae A₁ division subunits, the number and placement of tendons inserting on the maxilla, the length of the lachrymal spines, and the shape and size of the lachrymal and maxillary components of the locking mechanisms. Despite this substantial variation, all of the synanceiids have the core modifications to the circumorbitals, maxilla, adductor mandibulae, and associated tendons that constitute this novel defensive mechanism.

In addition to anatomical investigations using typical LED “daylight” lighting, one specimen was examined under fluorescence (Synanceiidae: Centropogon australis; Fig. 9). The lachrymal saber in this species biofluoresced in the green spectrum (Fig. 9B), whereas other regions of the head fluoresced red (Fig. 9C). Compared to visible light (Fig. 9A), stereomicroscopic images using fluorescence (Fig. 9B) highlight regions associated with the lachrymal saber that fluoresce in the green spectrum. Additional inspection (Fig. 9C) illuminated with two NightSea BlueStar flashlights and filtered through the Nikon light-shading plate highlights both the green lachrymal saber element and the red fluorescence associated with eye rings in this species of stonefish. Considerably more species need to be examined to assess the diversity of species that have biofluorescent lachrymal sabers or even whether additional species have this visual modification.

Fig. 9. 

Lateral (A and B) and rostral (C) images of Centropogon australis, KUI 41409. (A) Image represents a visible light image of Centropogon with the lachrymal saber in the resting position. (B) Image represents the identical placement of the image in panel A under fluorescent light with GFP filter under the Nikon SMZ-18 microscope. This image shows the bright green biofluorescence visible on the lachrymal saber. (C) Image shows a rostral view of the same individual under NightSea BlueStar flashlight illumination and the light-shading plate from the Nikon SMZ-18, which is not as restricted as the microscope filter. In this image, the green (lachrymal saber) and orange (dorsal surface of head) fluorescent emissions are visible in the waspfish.

Congiopodidae.—The Congiopodidae is a small Southern Hemisphere endemic family that has been recovered in a diversity of phylogenetic placements (Ishii and Imamura, 2008). Traditional higher-level phylogenetic studies (e.g., Regan, 1913; Greenwood et al., 1966) placed the congiopodids among the mail-cheeked fishes in their own order or suborder. Using morphological data and explicit phylogenetic analyses (Fig. 1), Ishida (1994) placed congiopodids sister to Synanceiidae (sensu Imamura, 2004) using five shared myological features and Imamura (2004) placed congiopodids sister to a clade composed of Gnathanacanthidae and Pataecidae (both sensu Imamura, 2004) using preopercular spination, the loss of the uroneural, and the separation of the transversus ventralis posterior. In contrast, Mandrytza (2001) recognized a congiopodoid clade (composed of a separate Congiopodidae and Zanclorhynchidae equivalent to our Congiopodidae) independent of his Scorpaenoidea. Mandrytza (2001) presented a variety of results, but his various explicit morphological analyses either allied his Congiopodoidei with a restricted Cottoidei (not including or associated with the Anoplopomatoidei, Hexagrammoidei, and Zoarcoidei) or in a polytomy with Anoplopomatoidei, Cottoidei + Hexagrammoidei, Hoplichthyoidei, Normanichthyoidei, and Scorpaenoidei (all sensu Mandrytza, 2001). Mandrytza (2001) also provided evidence that Perryena was not a congiopodid; instead, he suggested that this genus was a member of the Tetrarogidae (sensu Mandrytza, 2001). This finding was largely corroborated in a morphological analysis of a subset of scorpaenoids by Honma et al. (2013) who created a new family, Perryenidae, for this genus because of the non-monophyly of their two putative tetrarogids (Perryena and Tetraroge) included in their analysis. Earlier analyses of DNA-sequence data have resulted in even more hypotheses than the diversity of morphological studies. Smith and Wheeler (2004) recovered a separate Congiopodus (sister to Neosebastidae) and Zanclorhynchus (sister to the Notothenioidei). Li et al. (2009) recovered Zanclorhynchus sister to all other scorpaeniform fishes. Smith and Wheeler (2006) recovered Congiopodus sister to Bembridae. Smith and Craig (2007; Fig. 2A) recovered a clade composed of Congiopodus + Zanclorhynchus sister to Neosebastidae. Lautredou et al. (2013; Fig. 2B) recovered Zanclorhynchus sister to a clade composed of Plectrogeniidae + Scorpaenidae + Synanceiidae. Near et al. (2015) recovered a clade composed of Congiopodus + Zanclorhynchus sister to a clade composed of Bembridae, Bembropidae, and Scorpaenidae. Finally, Smith et al. (2016) recovered a clade composed of Congiopodus + Zanclorhynchus sister to a clade composed of the Plectrogeniidae + Scorpaenidae + Synanceiidae. We recover (Fig. 3) the Congiopodidae sister to a clade composed of the cottoids + zoarcoids, which is most similar to the findings of Mandrytza (2001). Our results corroborate the findings of Mandrytza (2001) and Honma et al. (2013) that separate Perryena from other traditional tetrarogids. Our sister-group relationship between congiopodids and cottoids + zoarcoids is supported by the loss of the metapterygoid lamina (character 32, state 1), the loss of one branchiostegal ray (character 37, state 1), the loss of a tooth plate on the third epibranchial (character 40, state 1), the separation of the lateral extrascapular into two elements (character 45, state 1), the fusion of the lower hypural plate and the parhypural (character 71, state 1), the origin of the levator operculi on the pterotic and posttemporal (character 83, state 1), the presence of adductores I–III (character 91, state 1), and the obliquus superioris not extending to the neurocranium (character 103, state 1). The overwhelming majority of explicit analyses and our results recover a monophyletic Congiopodidae that includes Zanclorhynchus; we recover 12 characters supporting the monophyly of this family (Appendix 2). Further, most recent studies separate the Congiopodidae from the Synanceiidae. The placement of this clade is inconsistently recovered across studies, and it remains difficult to know whether the family is more allied with the cottoid or scorpaenoid components of the scorpaeniform tree. We recommend treating this clade as a separate family within its own suborder Congiopodoidei.

Bembridae, Hoplichthyidae, Platycephalidae, and Triglidae.—Traditionally, the Bembridae, Hoplichthyidae, and Platycephalidae have been treated as a single evolutionary unit (Matsubara, 1943; Washington et al., 1984; Fig. 1). As first noted by Imamura (1996), explicit analyses of “platycephaloid” relationships do not recover bembrids in a clade with the hoplichthyids and platycephalids. Instead, Imamura's (1996) study demonstrated that Platycephalidae was sister to a clade of Hoplichthyidae + Triglidae. These three families were then hypothesized to be sister to Bembras, and that clade was subsequently hypothesized to be sister to Parabembras (Imamura, 1996; Fig. 1). Our results support Imamura's (1996) hypothesis that Platycephalidae is sister to a clade composed of Hoplichthyidae + Triglidae. However, our results place this Hoplichthyidae + Platycephalidae + Triglidae clade sister to all other scorpaenoids. This relationship supports the view that the less flattened bembrids are more closely related to scorpionfishes and stonefishes than to hoplichthyids, platycephalids, and triglids.

The Bembridae is a small, deep-water Indo-Pacific marine family whose limits have varied across studies. Jordan and Hubbs (1925) first separated the Parabembridae from the Bembridae, but this separation was not followed in most subsequent studies. For example, Washington et al. (1984) and Nelson (2006) included Bambradon, Bembradium, Bembras, Brachybembras (not discussed in Washington et al. [1984]), and Parabembras in their Bembridae. Imamura (1996) not only recognized a separate Bembridae and Parabembridae, but he also classified one traditional bembrid, Bembradium, as a member of the Plectrogeniidae. Molecular studies have not sufficiently sampled the Bembridae, and they have recovered their included bembrids sister to Congiopodidae, Hoplichthyidae, Platycephalidae, Plectrogeniidae, Synanceiidae, or a clade composed of the Bembropidae + Scorpaenidae (Smith and Wheeler, 2004, 2006; Smith and Craig, 2007; Lautredou et al., 2013; Near et al., 2013; Smith et al., 2016; Betancur-R. et al., 2017). Molecular studies that have included Bembras and Parabembras have consistently resulted in these genera forming a clade (Near et al., 2013, 2015; Betancur-R. et al., 2017), thus not requiring the recognition of a separate family-level status for Parabembridae. The current study is the first study with molecular data that included Bembradium, Bembras, and Parabembras. Given the alignment of Bembradium with Plectrogenium and the separate grouping of Bembras and Parabembras, we recommend classifying Bambradon, Bembras, Brachybembras, and Parabembras as a revised Bembridae (Appendix 2). Further, and as discussed below, our findings support the results of Imamura (1996) who placed Bembradium and Plectrogenium in the Plectrogeniidae. We recovered the Bembridae sister to a clade composed of Neosebastidae + Plectrogeniidae + Scorpaenidae + Synanceiidae (Fig. 3). This sister group relationship is supported by the loss of the lateral-line canal on the pterotic (character 22, state 0), two spines on the first dorsal-fin pterygiophore (character 56, state 0), and the presence of an adductor dorsalis (character 100, state 0).

The Hoplichthyidae is an Indo-Pacific marine family of 17 species that has been variously classified among mail-cheeked fishes but has been allied generally with the platycephalids. Traditional higher-level phylogenetic studies (e.g., Gill, 1888; Regan, 1913; Matsubara, 1943; Quast, 1965; Washington et al., 1984) placed the hoplichthyids as a separate family, closely related to the platycephalids. Greenwood et al. (1966) had a similar classification, but they treated the hoplichthyids as their own suborder, Hoplichthyoidei. Winterbottom (1993) suggested that the hoplichthyids might even be close relatives to the gobioids, so the family's placement has been historically varied. As noted above, Imamura (1996) first suggested that hoplichthyids are more closely related to the triglids than the platycephalids, and his results suggested that Hoplichthys was sister to a clade composed of Peristedion + Satyrichthys. Molecular data, beginning with Smith and Wheeler (2004), have suggested alternative placements for the hoplichthyids ranging from a close relationship with Ovalentaria to sister to the Bembridae, Normanichthyidae, Synanceiidae, a clade composed of Bembridae + Platycephalidae, a clade composed of the “cottoids and allies” (Anoplopomatoidei, Cottoidei, Gasterosteioidei, Hexagrammoidei, and Zaniolepidoidei) + Triglidae, a clade composed of Cottoidei + Gasterosteioidei + Hexagrammoidei + Triglidae + Zoarcoidei, and a clade composed of the “cottoids and allies” + Scorpaenidae + Triglidae (Smith, 2005; Smith and Wheeler, 2006; Smith and Craig, 2007; Lautredou et al., 2013; Near et al., 2013, 2015; Smith et al., 2016; Betancur-R. et al., 2017; Fig. 2). Our results (Fig. 3) place the Hoplichthyidae sister to the Triglidae, similar to the findings of Imamura (1996, 2004; Fig. 1). This sister group relationship is supported by the presence of tubercles on the neurocranium (character 13, state 1), the loss of one postcleithrum (character 48, state 1), the increase in the number of free pectoral-fin rays to three or more (character 49, state 3), the fusion of the cartilaginous caps on the anterior portion of the pelvis (character 51, state 1), the presence of a hyohyoides inferioris (character 84, state 1), the attachment of dorsal elements of pelvic-fin muscles to the pectoral girdle (character 97, state 1), and the obliquus superioris bypassing and lying ventrally to Baudelot's ligament (character 104, state 1). As with the Congiopodidae, the placement of the hoplichthyids is inconsistent across studies. Most studies ally the hoplichthyids more with the Triglidae (often combined with the included “cottoids and allies”). Given the ambiguity, we recommend treating the Hoplichthyidae, diagnosed by 20 morphological synapomorphies (Appendix 2), as a separate scorpaenoid family.

The Platycephalidae is a modestly large family of 84 species that is found in brackish and marine environments in the Indo-Pacific region (Nelson, 2006). As has been found in previous studies, the Hoplichthyidae and Platycephalidae were recovered as independent monophyletic groups with the traditional limits (Keenan, 1991; Imamura, 1996; Appendix 2; Fig. 3). The interrelationships of the family have been discussed above, and our results support the findings of Imamura (1996) that the Triglidae + Hoplichthyidae is recovered as the sister group to the platycephalids. This sister group relationship is supported by the presence of a tooth plate on the second epibranchial (character 39, state 1). Additionally, our limited sampling of platycephalids corroborates the phylogeny and classification presented in Imamura (1996).

The Triglidae is a large family of 171 species that are found in tropical, temperate, and deep-water habitats across all oceans (Nelson, 2006). Traditional classifications (e.g., Gill, 1888; Matsubara, 1943; Washington et al., 1984; Imamura, 1996; Eschmeyer et al., 2017) often treat the Peristediidae and Triglidae as independent, closely related families. Our study corroborates the findings of Smith (2005) and Portnoy et al. (2017) that place the traditional peristediids (our peristediines) within the Triglidae. Other than the placement of peristediines inside the triglids, our hypothesized relationships support the phylogenies of Imamura (1996) and Richards and Jones (2002). Similarly, our phylogeny supports the phylogeny of Portnoy et al. (2017), including the placement of peristediines inside the Triglidae. This placement and resulting expansion (and monophyly) of the Triglidae is supported by both morphological and molecular data, and our study recovers seven characters supporting the monophyly of this expanded family (Appendix 2).

Neosebastidae and Plectrogeniidae.—The placement of Neosebastidae and Plectrogeniidae has been historically problematic; they are often represented as early diverging scorpaenoid lineages or “ancestral” forms (Matsubara, 1943; Imamura, 1996). Although the current study is the first study to unite these two families (Fig. 3), Imamura (2004), Smith and Wheeler (2004, 2006), Smith and Craig (2007), and Smith et al. (2016) have often found them relatively closely related. In this study, this sister-group relationship was supported by the separation of the first and second hypurals (character 68, state 0).

The Neosebastidae is a predominantly anti-tropical Indo-Pacific marine family of 18 species that has been often separated from the core scorpionfishes in a separate family or subfamily in modern phylogenies and classifications (Imamura, 2004; Motomura, 2004; Nelson, 2006). Beginning with Matsubara (1943) and supported by Washington et al. (1984) and Nelson (2006), the Neosebastidae has been treated as a separate subfamily (Neosebastinae) of the Scorpaenidae. Matsubara (1943) hypothesized that the neosebastids were allied with the sebastines, and Ishida (1994; Fig. 1) allied the neosebastids with the setarchines and recognized them as a separate family. Imamura (2004; Fig. 1) also recognized the clade as a separate family and suggested a non-scorpaenid relationship for the neosebastids; he resolved them with the more flattened scorpaenoids in the families Bembridae, Hoplichthyidae, Platycephalidae, Plectrogeniidae, and Triglidae. Molecular studies have grouped the neosebastids with several non-scorpaenoid groups (e.g., Acanthistius or bembropids; Smith and Wheeler, 2006; Smith et al., 2016) or with congiopodids (Smith and Wheeler, 2004; Smith and Craig, 2007). Our finding of a Neosebastidae + Plectrogeniidae clade adds additional complications to the placement of the Neosebastidae, but this result is closer to the findings of Imamura (2004) and several molecular studies (e.g., Smith and Wheeler, 2004, 2006) that have plectrogeniids among the closest relatives of the neosebastids.

Species in the Plectrogeniidae are relatively widespread with collections ranging from the western Indian Ocean to Hawaii despite the family including just four species (Imamura, 1996; Nelson, 2006; Eschmeyer et al., 2017). Fowler (1938) first emphasized the distinctiveness of Plectrogenium. Matsubara (1943) supported Fowler's (1938) assertion, suggesting that the genus may represent the ancestral condition of some of the deeper water, flattened scorpaenoids. He noted that Plectrogenium shared the loss of the gas bladder and the presence of several rows of prominent head spines and notched pectoral fins with some scorpaenids (e.g., Sebastolobus) while also showing characteristics in common with the bembrids. Washington et al. (1984) further corroborated this hypothesis by pointing to similarities in the scales and caudal fin between Parabembras and Plectrogenium. Subsequently, Imamura (1996, 2004) supported a placement of an expanded Plectrogeniidae (including Bembradium) sister to the clade composed of Bembridae + Hoplichthyidae + Platycephalidae + Triglidae that was united by the presence of a posterior pelvic fossa. Molecular studies have not fully supported or refuted these morphological hypotheses. Smith and Wheeler (2004) and Smith and Craig (2007) recovered Plectrogenium sister to Bembridae, Lautredou et al. (2013) recovered Bembradium sister to Synanceiidae (their analysis did not include any bembrids), and Smith et al. (2016) recovered Plectrogenium sister to the Scorpaenidae. This is the first molecular study to include both Bembradium and Plectrogenium, and we found a unique relationship for plectrogeniids sister to Neosebastidae. The monophyly of the Plectrogeniidae (Appendix 2) is supported by five characters and corroborates Imamura's (1996) hypothesis and the recognition of this distinct family.

Scorpaenidae and Synanceiidae.—The scorpaenids and synanceiids are the most species-rich clades of scorpaenoids, and the species in these families have often been classified together in whole or in part. Gill (1888) grouped these clades together to the exclusion of all other mail-cheeked fishes. Regan (1913) modified Gill's (1888) arrangement and united scorpaenids, synanceiids, and triglids in his Scorpaeniformes (their studies did not include any representatives of the Neosebastidae or Plectrogeniidae). Matsubara (1943; Fig. 1B) distributed the core scorpionfishes across his “Cocotropus-stem,” “Scorpaena-stem,” and “Sebastes-stem” based primarily on circumorbital differences. His “Cocotropus-stem” was composed of the Japanese synanceiids. His “Scorpaena-stem” was composed of Plectrogenium and all non-sebastine scorpaenids. Finally, his “Sebastes-stem” was composed of the Neosebastidae and Sebastinae. The classification of Washington et al. (1984) largely followed Matsubara (1943) except that they combined his “Scorpaena-stem,” “Sebastes-stem,” Synanceiinae, Apistus, and Cheroscorpaena into their more inclusive Scorpaenidae. They distributed the remaining synanceiids across four additional families (Aploactinidae, Gnathanacanthidae, Pataecidae, and Tetrarogidae) and recognized Caracanthus as a distinct family from the Scorpaenidae. Ishida's (1994) phylogenetic hypothesis (Fig. 1C) and classification recognized 12 families. As noted above, this phylogeny included Congiopodidae, Neosebastidae, and Plectrogenium nested among the included representatives of the Scorpaenidae and Synanceiidae. Ishida's (1994) major groupings largely followed Washington et al. (1984) except that Ishida often recognized clades at higher taxonomic levels. Ishida (1994) elevated Washington et al.'s (1984) Apistinae, Neosebastinae, and Setarchinae to the family level. Ishida's (1994) Sebastidae included Washington et al.'s (1984) Plectrogeniinae, Sebastolobinae, and Sebastinae. Ishida's (1994) classification included a Synanceiidae that incorporated Washington et al.'s (1984) Choridactylinae, Minoinae, and Synanceiinae. Finally, his Scorpaenidae was restricted to Washington et al.'s (1984) Pteroinae and Scorpaeninae. As noted by Smith and Wheeler (2004), a computer-aided re-analysis of Ishida's (1994) matrix recovered many equally optimal trees that were shorter than the tree presented by Ishida (1994). This large assortment of most parsimonious trees resulted in a poorly resolved phylogeny with just three of his families represented by more than one species being recovered as monophyletic: Aploactinidae, Congiopodidae, and Pataecidae. Relative to Ishida (1994), Imamura (2004; Fig. 1D) increased the taxon sampling among closely related groups (e.g., Platycephalidae, Triglidae) and recovered a largely complementary phylogenetic classification with a few changes. Imamura (2004) recognized a Sebastolobidae at the family level and relegated the Setarchidae of Ishida (1994) to a clade within Scorpaenidae. Subsequent work by Shinohara and Imamura (2005) also placed Caracanthus into the Scorpaenidae. Most recently, Honma et al. (2013) recognized a new family Perryenidae for a member of Mandrytza's (2001) Tetrarogidae. This new family and Mandrytza's (2001) earlier treatment of Eschmeyeridae as an additional monotypic family based on former waspfishes casts doubt on tetrarogid monophyly. The proliferation of new waspfish families and Smith and Wheeler's (2004) re-analysis of Ishida's (1994) dataset that recovers 5–6 distinct clades of tetrarogids (sensu Ishida, 1994) highlights that traditional “stonefish” taxonomy is becoming complicated with substantial evidence for tetrarogid polyphyly and a diversity of families with three or fewer species (traditional Apistidae, Eschmeyeridae, Gnathanacanthidae, Pataecidae, and Perryenidae).

As seen with the morphological studies, molecular phylogenies that have included representatives of both the Scorpaenidae and Synanceiidae have recovered varied phylogenetic results, but the clades themselves have been largely repeated. These studies have also echoed some morphological results. For example, molecular studies, like morphological studies, consistently recover a polyphyletic Tetrarogidae (sensu Ishida, 1994) and Plectrogenium separate from the sebastines (Smith and Wheeler, 2004, 2006; Smith and Craig, 2007; Smith et al., 2016). The molecular results have also largely recovered reciprocally monophyletic scorpaenid and synanceiid clades. Smith and Wheeler (2004) recovered independent clades of the Scorpaenidae and Synanceiidae with a diversity of included taxa (including non-scorpaeniforms) in the MRCA. Smith and Wheeler (2006) recovered the scorpaenids sister to epinephelids and synanceiids sister to the triglids with an MRCA that includes the “cottoids and allies” as well as scorpaenoid and serranid fishes. Smith and Craig (2007) recovered relationships similar to Smith and Wheeler (2006) except that the MRCA excluded the Epinephelidae and included the Anthiadidae, Niphonidae, Percidae, Serranidae, and Trachinidae. With more families sampled, Lautredou et al. (2013) and Smith et al. (2016) recovered a clade composed of Plectrogeniidae + Scorpaenidae + Synanceiidae. Finally, Betancur-R. et al. (2017) recovered a clade composed of Scorpaenidae + Synanceiidae, but their analysis only included two synanceiids and did not include any congiopodids, neosebastids, or plectrogeniids, so their phylogeny is of limited comparative value. Our current analysis is the first to recover a clade composed of Neosebastidae + Plectrogeniidae + Scorpaenidae + Synanceiidae, and this clade was supported by the loss of the fourth circumorbital (character 9, state 1).

Despite continued iterative improvement, the march toward a monophyletic taxonomy based on morphological and molecular data has been incompletely accepted by the major fish classifications (e.g., Nelson, 2006; Nelson et al., 2016; Eschmeyer et al., 2017). For example, Nelson (2006) largely follows the pre-phylogenetic study of Washington et al. (1984) except for the placement of Ishida's (1994) Tetrarogidae in their Scorpaenidae. Nelson et al. (2016) followed Nelson (2006) except they placed Caracanthus in the Scorpaenidae and recognized Eschmeyeridae as a separate family from their Tetraroginae. Curiously, they left the Tetraroginae within the Scorpaenidae and continued to recognize Perryena in the Congiopodidae despite the evidence for both tetrarogine changes being presented in the same study (Mandrytza, 2001). In contrast, Eschmeyer et al. (2017) largely followed Ishida (1994) except for the placement of Caracanthus in the Scorpaenidae (presumably following Shinohara and Imamura, 2005) and the recognition of a separate Plectrogeniidae (presumably following Imamura, 1996), Eschmeyeridae (presumably following Mandrytza, 2001), and Perryenidae (presumably following Honma et al., 2013). Betancur-R. et al. (2017) largely followed Eschmeyer et al. (2017) including the retention of a non-monophyletic (in their study) Scorpaenidae. Presumably, their minimal changes relative to Eschmeyer et al. (2017) are due to limited sampling where only 11 of their 21 platycephaloid, scorpaenoid, or trigloid families were examined. All of these previous studies highlight the need to combine molecular and morphological data to generate a complete and holistic phylogenetic hypothesis that can become stable and more widely accepted.

The Scorpaenidae is a worldwide marine family of 370 species that have been collected in environments ranging from shallow to deep water and from the poles to the tropics (Nelson, 2006; Eschmeyer et al., 2017). This group includes the traditional Caracanthidae, Scorpaenidae, Sebastidae, and Setarchidae (sensu Ishida, 1994; Appendix 2) and includes animals with reproductive modes ranging from more traditional broadcast spawning to live birth (Breder and Rosen, 1966; Muñoz, 2010). Most inexplicit and explicit morphological studies and one large-scale molecular study generally have resolved the rockfishes (Sebastinae) as an ancestral or stem grade within the scorpaenoid radiation (Matsubara, 1943; Ishida, 1994; Imamura, 1996, 2004; Lautredou et al., 2013). In contrast, Smith and Wheeler (2004, 2006), Smith and Craig (2007), Smith et al. (2016), and Betancur-R. et al. (2017) have recovered the Sebastinae as a deeply nested lineage. This revised phylogenetic hypothesis implies that the scorpaenoids originated in warmer waters and transitioned to deeper (e.g., Setarchinae) and colder habitats (e.g., Sebastinae) rather than the previous hypotheses that would necessitate transitioning from cooler waters into more temperate and tropical regions. This more traditional hypothesis may have been largely driven by the evolutionary perspective that the colder, overwhelmingly North Pacific cottoids and allies were the closest allies to a sebastine-stem scorpaenoid radiation (Smith and Busby, 2014). This Sebastinae was resolved as the sister group to a clade composed of Adelosebastes + Sebastolobus, which have generally been allied with or nested among the core sebastines in previous morphological and molecular studies (Figs. 1, 2). As might be expected with the revised placement of these core rockfishes, we recover this clade nested within a larger assemblage that includes all other sampled deeper and cooler water genera (i.e., Ectreposebastes, Pontinus, Setarches, and Trachyscorpia). In this study, this colder-habitat clade of scorpaenids was recovered as the sister group to a Scorpaenodes + Pteroinae clade. This is in contrast to several previous studies (e.g., Imamura, 2004; Smith and Craig, 2007) that have found multiple clades of deep-water scorpaenoids sister to the pteroine lionfishes and allies (Figs. 1, 2). One of the most consistent results across scorpaenid studies is the sister-group relationship between Scorpaenodes and pteroine lionfishes (e.g., Ishida, 1994; Imamura, 2004; Lautredou et al., 2013; Smith et al., 2016; Betancur-R. et al., 2017). The final scorpaenid clade recovered in our analysis is a primarily tropical and subtropical clade composed of Caracanthus, Pteroidichthys, Scorpaena, Scorpaenopsis, and Taenionotus. This clade was sister to the cold-water scorpaenoids + lionfishes and allies and has been consistently recovered in molecular studies (Smith and Craig, 2007; Lautredou et al., 2013; Smith et al., 2016; Betancur-R. et al., 2017). In contrast, morphological studies have typically recovered these fishes as a grade with Scorpaenodes and Pteroinae (and potentially other genera) nested within the group (Ishida, 1994; Shinohara and Imamura, 2005). It is clear that the inversion of scorpaenid relationships with sebastines deeply nested with the family that is recovered in this combined study and several other molecular studies (Smith and Wheeler, 2004, 2006; Smith and Craig, 2007; Smith et al., 2016; Betancur-R. et al., 2017) has dramatically altered the polarity of morphological transformations and the impact this has on the evolutionary relationships in this commercially important clade.

The Synanceiidae, a family of 133 species, is primarily a marine clade with a few fresh- or brackish-water representatives (e.g., Gymnapistes, Neovespicula) that is distributed from the western Indian Ocean to the South Pacific Ocean (Eschmeyer and Rama-Rao, 1973; Nelson, 2006). The largest taxonomic change we are recommending in this study is the consolidation of the traditional Apistidae, Aploactinidae, Eschmeyeridae, Gnathanacanthidae, Pataecidae, Perryenidae, Synanceiidae, and Tetrarogidae (all sensu Eschmeyer et al., 2017) into a monophyletic Synanceiidae. This expanded Synanceiidae is diagnosed by the evolution of the lachrymal saber as well as five additional morphological transformations (Appendix 2). Additionally, Leis and Rennis (2000) provided evidence from larval morphology that separates these fishes from the remainder of the core scorpaenoids. We recommend this higher-level change because of the proliferation of family-level names that are already emanating from the former Tetrarogidae (i.e., Eschmeyeridae, Perryenidae) and that are likely to continue. Our results, previous molecular studies (Smith and Wheeler, 2004, 2006; Smith and Craig, 2007; Smith et al., 2016), and the computer aided re-analysis of Ishida's (1994) scorpaenoid study in Smith and Wheeler (2004) that have all sampled multiple species of tetrarogids (sensu Eschmeyer et al., 2017) have suggested that upwards of four or five additional small or monogeneric families will likely be needed to generate a monophyletic taxonomy. Instead of describing a diversity of new waspfish families, the alternative strategy is chosen here where the diversity of existing comparatively small families of stonefishes and waspfishes can be consolidated into a single, well supported, consistently supported, and taxonomically stable family with the retention of subfamilies as warranted and needed (e.g., Apistinae, Aploactininae, Pataecinae, Synanceiinae). Further, our classification largely returns the subfamilial taxonomy to that recommended by Matsubara (1943). Our results recover the typical placement for the Apistinae as the earliest diverging lineage in the Synanceiidae. Regan (1913) included this group among his Scorpaenidae, which was composed of our Scorpaenidae, Apistus, Erisphex (an aploactine), and six genera of “tetrarogids” (sensu Ishida, 1994). Matsubara (1943), Ishida (1994), and Imamura (2004) all treated Apistus as the earliest branching lineage in his clade that is largely equivalent to our Synanceiidae. The placement of Apistinae is one of the most consistently recovered relationships in scorpaenoid phylogenetics, but it is important to note that Washington et al. (1984) highlighted a number of features that potentially group the Apistinae with the Triglidae: a bilobed gas bladder with an intrinsic muscle, elongate pectoral-fin rays (also found in hoplichthyids, Choridactylus, Inimicus, and Minous), and an expansion of the circumorbitals. Thus, continued work is warranted. Our phylogeny, like previous morphological studies (Ishida, 1994; Imamura, 2004), recovers a monophyletic Synanceiinae deeply nested within the Synanceiidae. Our study recovers a clade composed of the Aploactininae and Pataecinae sister to the restricted Synanceiinae. Other than his inclusion of the Congiopodidae in this clade, Ishida (1994) recovered this same relationship. Finally, we have a grade composed of the various “tetrarogid” or formerly “tetrarogid” genera and Gnathanacanthus as a diversity of lineages more closely related to Synanceiinae than to Apistinae.

Evolution of the Scorpaenoidei.—The combined morphological and molecular phylogeny presented herein provides an opportunity to reconcile the often conflicting datasets and look at the implications for this updated hypothesis on the evolution of this species-rich clade. One of the major findings in this study was the discovery of the lachrymal saber. As noted above, this specialization is hypothesized to have a primarily defensive role. The maxillary rotation of the lachrymal saber has two major anti-predator impacts. First, it expands the width of the head by projecting the spine(s) outward (Figs. 4–7). This expansion increases the rostral width of the fish by 10–25% and would greatly increase the gape required by a would-be predator (Price et al., 2015). Second, the presence of an outwardly directed and sharp spine should reduce predation because of the potential for the saber to pierce a would-be predator. Cowan (1969) described a similar defensive role for the outward projection of the preopercular spines in the closely allied psychrolutids that have enlarged antler-like modifications (e.g., Enophrys, Icelinus; Yabe, 1985). In addition to its role in avoiding or reducing predation, it is possible that the lachrymal saber is used for intraspecific competition. As seen in the evolution of antlers and horns (Chapman, 1975), the lachrymal saber could play a role when synanceiids compete for mates or territories. The one synanceiid species examined for biofluorescence (Centropogon australis) had a green fluorescent lachrymal saber that contrasted with non-fluorescent or red-fluorescent regions on the head of these animals (Fig. 9). Recent studies (e.g., Sparks et al., 2014; Anthes et al., 2016; Gruber et al., 2016) have demonstrated that a number of fish groups have green and red fluorescence that appears to be playing an ecological and/or evolutionary role. As such, it is possible that the synanceiids could be advertising or highlighting this specialization with this fluorescence in a similar role as the bioluminescence associated with the defensive dorsal spines in etmopterid sharks (Claes et al., 2013) or that the lachrymal saber is involved in intraspecific competition and mate choice where synanceiid species are advertising their sabers to conspecifics.

In addition to exploring the evolution of the lachrymal saber, the revised scorpaenoid hypothesis has implications for the evolution of viviparity in this clade. The deeply nested placement of Sebastinae within the Scorpaenidae corroborates Wourms' (1991) assertion about the evolution of live birth and corresponding intermediate stages in the transition from an ovuliparous (classical oviparous) ancestor among the non-scorpaenid scorpaenoids. As noted by Smith and Wheeler (2004), there appears to be an evolutionary transition from a more common ovuliparous ancestor in the platycephalids, synanceiids, and triglids to an oviparous species that releases fertilized eggs within a gelatinous egg mass in genera such as Dendrochirus, Pterois, Scorpaena, Scorpaenodes, Scorpaenopsis, and Sebastolobus (Wourms, 1991; Koya and Muñoz, 2007). This intermediate reproductive mode is further modified in Helicolenus where species in the genus have internal fertilization and zygoparity where fertilized ova are held by the mother before being released into the ocean (Wourms, 1991). The reproductive mode of Hozukius is unknown, but it shares a II-3 ovarian type with most scorpaenids (e.g., Caracanthus, Dendrochirus, Helicolenus, Scorpaena; Cole, 2003; Koya and Muñoz, 2007). The more derived viviparous sebastine rockfishes (Sebastes and Sebastiscus) have a type II-1 ovarian type (Koya and Muñoz, 2007). This suggests that Hozukius is likely to be more similar to Helicolenus or Scorpaena and not be live bearing. The evolution of reproductive modes and live birth in scorpaenoids has been discussed in considerably more detail in other studies (Wourms, 1991; Koya and Muñoz, 2007; Muñoz, 2010; Pavlov and Emel'yanova, 2013), but their interpretations have implicitly or explicitly relied on the hypothesized placement of sebastines at the base of the scorpaenoid tree. The inversion of the phylogeny of scorpaenoids proposed in this study is more consistent with traditional views on the evolution of viviparity where there is a transition from external fertilization to internal fertilization with the retention of either the developing eggs or embryos within the mother (Wourms, 1991; Wourms and Lombardi, 1992). These are two examples demonstrating the impact of the proposed phylogeny on the evolution of the scorpaenoid fishes. We hope that this revised hypothesis for the relationships of scorpionfish and allies will allow researchers to test additional evolutionary hypotheses for this important percomorph clade.