Phylogenetic Interrelationships of Ginglymodian Fishes (Actinopterygii: Neopterygii)

The Ginglymodi is one of the most common, though poorly understood groups of neopterygians, which includes gars, macrosemiiforms, and “semionotiforms.” In particular, the phylogenetic relationships between the widely distributed “semionotiforms,” and between them and other ginglymodians have been enigmatic. Here, the phylogenetic relationships between eight of the 11 “semionotiform” genera, five genera of living and fossil gars and three macrosemiid genera, are analysed through cladistic analysis, based on 90 morphological characters and 37 taxa, including 7 out-group taxa. The results of the analysis show that the Ginglymodi includes two main lineages: Lepisosteiformes and †Semionotiformes. The genera †Pliodetes, †Araripelepidotes, †Lepidotes, †Scheenstia, and †Isanichthys are lepisosteiforms, and not semionotiforms, as previously thought, and these taxa extend the stratigraphic range of the lineage leading to gars back up to the Early Jurassic. A monophyletic †Lepidotes is restricted to the Early Jurassic species, whereas the strongly tritoral species previously referred to †Lepidotes are referred to †Scheenstia. Other species previously referred to †Lepidotes represent other genera or new taxa. The macrosemiids are well nested within semionotiforms, together with †Semionotidae, here restricted to †Semionotus, and a new family including †Callipurbeckia n. gen. minor (previously referred to †Lepidotes), †Macrosemimimus, †Tlayuamichin, †Paralepidotus, and †Semiolepis. Due to the numerous taxonomic changes needed according to the phylogenetic analysis, this article also includes formal taxonomic definitions and diagnoses for all generic and higher taxa, which are new or modified. The study of Mesozoic ginglymodians led to confirm Patterson’s observation that these fishes show morphological affinities with both halecomorphs and teleosts. Therefore, the compilation of large data sets including the Mesozoic ginglymodians and the re-evaluation of several hypotheses of homology are essential to test the hypotheses of the Halecostomi vs. the Holostei, which is one of the major topics in the evolution of Mesozoic vertebrates and the origin of modern fish faunas.


Introduction
A very important step in the evolution of the actinopterygian fishes is the origin of the Neopterygii, with the acquisition of a better control of the movements of both dorsal and anal fins, resulting in an improvement in their swimming capabilities. They additionally acquired several modifications in the skull, which allowed the evolution of different feeding mechanisms and consequently the colonization of new ecological niches. All of these characters represented major improvements, so that the Neopterygii became the dominant group of fishes (and, thus, taxonomically of vertebrates in general), and they also include the vast majority of the modern fishes, the teleosts. Among basal neopterygians, the family {Semionotidae has played a critical role when trying to understand the origin and relationships of the other neopterygian lineages. Regan [1] considered {Semionotidae to represent the ancestral stock from which all other neopterygian lineages, including teleosts, had evolved. Brough ([2]: p. 108) proposed that most, if not all holosteans arose from the families {Semionotidae and {Eugnathidae independently. Danil'chenko [3] and McAllister [4] classified the {Semionotidae within an order Amiida or Amiiformes distinct from the Lepisosteiformes, but Gardiner placed them together with the Lepisosteidae in a superfamily Semionotoidea [5] or order Semionotiformes [6]. Patterson [7], after including the dapediids in {Semionotidae, concluded that semionotids represent a grade-group (para-or polyphyletic) and placed them as basal halecostomes of uncertain relationships. Recent phylogenetic analyses have demonstrated the monophyly of a major clade including {Semionotidae, Lepisosteidae, and {Macrosemiidae (Figs. 1, 2) [8][9][10][11][12][13].
Originally based on the Upper Triassic genus {Semionotus from central Europe, the family {Semionotidae has become a ''wastebasket'' taxon for many taxa of basal neopterygians that cannot be confidently assigned to any of the other groups, spanning from the latest Permian to the late Cretaceous. However, although it turned into one of the most diverse taxon of fossil neopterygians, semionotid monophyly, the interrelationships of the taxa included and even their alpha taxonomy has not been satisfactorily established so far. The family {Semionotidae was created by Woodward [14] to include {Semionotus, {Dapedium, {Tetragonolepis, and the perleidiforms {Pristisomus and {Cleithrolepis. {Lepidotes, which was previously considered to represent its own family {Lepidotidae Owen, 1860 [15], was later added to {Semionotidae by Woodward [16], together with a variety of genera, including some other perleidiforms like {Colobodus. With the time, {Semionotidae became even larger, containing about 20 genera of diverse basal neopterygians, and both the concept of the family as well as its phylogenetic relationships became more and more confused. Lehman [17] and Wenz [18] [8] within and outside the family is not clear [19][20][21][22][23]. The most recent taxonomic hypothesis for {Semionotidae is that of Wenz [21], who proposed a new arrangement of the taxa included according to the number and disposition of suborbital bones, though she did not provide a new formal diagnosis for the family. Therefore, the family {Semionotidae has still neither been satisfactorily defined, nor diagnosed.
Grande [13] reorganized some of the genera previously classified in the family Lepisosteidae in a new family {Obaichthyidae. Therefore, I will use the informal name of gars in reference to both lepisosteids and obaichthyids.

Cladistic Analysis
Phylogenetic relationships are explored through parsimony analysis. A data matrix with a total of 90 characters and 37 taxa was assembled using Mesquite Version 2.73 [40] (see list of material examined in Text S1 and data matrix in Text S2). The data matrix is also availabe in Morphobank (http://www. morphobank.org/). Tree search was performed with PAUP* Version 4.0 beta version [41] and TNT version 1.1 [42]. All characters were considered unordered and given equal weight. All of the studied taxa have been included independently of the amount of missing information (missing data due to lack of information or inapplicable characters varying between 34% for {Isanichthys and 3% for Lepisosteus or {Dentilepisosteus). Most parsimonious trees were obtained both in PAUP* and TNT through heuristic search with random addition sequence, 10000 replicates and tree bisection and reconnection branch swapping. Furthermore, the data matrix was analysed in TNT with the ''new technology approaches'' (ratchet, sectorial searches, tree drifting, and tree fusing). The number of trees held at each iteration was set at 1 and 10 for different runs with both programs, but the results were identical. Distribution of characters and character changes have been analysed in PAUP* through accelerated and decelerated transformations (ACCTRAN and DELTRAN respectively; see list of synapomorphies in Table S1). Branch support was evaluated through decay indexes for each node (Bremer support) and Bootstrap and Jackknife methods. Both Bootstrap and Jackknife analyses were also run in PAUP* and TNT through heuristic search with 10000 replicates and simple addition sequence.
Based on the results of the cladistic analysis, taxonomic decisions were made within the framework of Phylogenetic Systematics and, thus, the taxa defined herein represent monophyletic groups. All generic diagnoses are based on unambiguous synapomorphies only. To facilitate identifications, additional distinctive combinations of features are also provided. Higher rank taxa are here named based on stem-based definitions according to de Queiroz & Gauthier [43]. The diagnoses proposed for the taxa above the generic rank are based on unambiguous and ambiguous synapomorphies. Among them, the unambiguous synapomorphies are indicated with an asterisk ''*'' and the ambiguous synapomorphies with ''(ACCTRAN)'' or ''(DEL-TRAN)'' depending on the optimization method (in all cases, the precise direction of change is given in the list of synapomorphies in Table S1). The character number and state is given between brackets for all characters included in the diagnoses.
Out-groups. In contrast to previous phylogenetic studies of ''semionotiforms'' and lepisosteiforms [8,10,12], which used a hypothetical ancestor, real Stensiö, 1921 [45], from the Early Triassic of Spitzbergen, and {P. stoschiensis Stensiö, 1932 [46], from the Early Triassic of East Greenland. Other species have subsequently been added to this genus by Piveteau [47], Teixeira [48], Lehman [49], Beltan [50], and Su [51]. In a revision of the actinopterygian fishes from the Middle Triassic of northern Italy and the Canton Ticino (Switzerland), Lombardo [52] argued that the genus {Perleidus should be restricted to the type species {P. altolepis. In particular, {P. woodwardi and {P. stoschiensis have, according to this author, a very different pattern of bones in the ethmoid region of the skull; {P. woodwardi would further differ in having a different kind of caudal fin (abbreviated heterocercal vs. hemiheterocercal in {P. altolepis). However, apart from differences in the anatomical nomenclature used by Stensiö [45,46] and Lombardo [52], and the different interpretation of certain bones (in particular the antorbital, interpreted as a rostral by Stensiö) I do not find major differences in the pattern of skull bones in the three species {P. altolepis, {P. woodwardi, and {P. stoschiensis. Quite the opposite, the skull osteology is strikingly similar, supporting the referral of the three species to the same genus. Also, Lombardo [52] argued for the absence of epaxial fin rays in the caudal fin of {P. woodwardi, but the caudal fin is not completely preserved in any of the specimens of this species studied by Stensiö and the photograph in Stensiö ([45]: pl. 33) provides no evidence for an heterocercal tail, as indicated by Lombardo ([52]: 357). Consequently, I consider {P. altolepis and {P. stoschiensis as the best described species of {Perleidus. Accordingly, the morphological characters were scored on the basis of descriptions by Lombardo [52] and Stensiö [46] and figures of the respective species.
The cladistic analysis by Arratia [57] shows that the teleosts split in two lineages at the base of Teleostei. One lineage is represented by the extinct {Siemensichthys-Group and the other, leading to the living teleosts, includes {Pholidophorus at its base. The genus {Siemensichthys Arratia, 2000 [57], was chosen to represent the {Siemensichthys-Group [57]. {Siemensichthys is represented by two species from the Late Jurassic of Southern Germany: {S. macrocephalus (Agassiz, 1834 [58]) and {S. siemensis Arratia, 2000 [57]. Among them, the type species {S. macrocephalus, originally thought to represent the genus {Pholidophorus [58], is the most completely known and, thus, it was chosen to represent the genus in the present cladistic analysis. Scorings for this species are based on Arratia [57] and direct observation of the holotype BSPG AS I 1134. Arratia [57] discussed in detail the problems concerning the poor definition of the order Pholidophoriphormes Berg, 1940 [59], the family Pholidophoridae Woodward, 1890 [14], and even the genus {Pholidophorus Agassiz, 1832 [25]. The author demonstrated that a monophyletic {Pholidophorus is restricted to the type species {Ph. latiusculus and {Ph. bechei, as previously suggested by Nybelin [60] and proposed by Zambelli [61]. Scorings for this genus are based on the descriptions of {Ph. bechei by Nybelin [60], Patterson [62] and Arratia [57,63]. {Leptolepis coryphaenoides is also included because it shares with the more advanced teleosts several synapomorphies that are absent in {Pholidophorus or the {Siemensichthys-Group [64]. The scorings for {L. coryphaenoides are based on Patterson [62] and Arratia [57,63].
The To avoid misinterpretations concerning the relationships between the out-group taxa and the ingroup, the analysis was run leaving the outgroup in an unresolved polytomy at the base of the trees. However, due to the possible close relationship between {Dapedium and the ''semionotiforms'', this genus was not defined as outgroup in PAUP*.
Apart from the three Chinese taxa, which are poorly described in the literature and material of which was not available for this study, the remaining eight of the 11 ''semionotid'' genera are included in the analysis.  [70] are also part of the in-group. Also, a new Chinese taxon very recently described, Luoxiongichthys hyperdorsalis Wen et al. 2012 [71] is here included in the in-group, because, according to my own observations, the fish is a ''semionotiform'', although the authors of this taxon classified it in the Halecomorphi.
Detailed information on the studied material and the literature consulted for each taxon is included in Text S1. Most of the taxa included in the in-group have been studied first hand and specific literature was mainly consulted to complete information and reconcile the interpretation of several anatomical features.

Anatomical Nomenclature
Skull bones are generally named according to the use of most authors in actinopterygians. The bones carrying the infraorbital sensory canal anterior to the orbit are referred to as 'anterior infraorbitals' following Wenz [21,72] and López-Arbarello & Codorniú [22]. The ossifications of the palatoquadrate are named according to Arratia & Schultze [73]. The distinction of nontritoral, moderately tritoral and strongly tritoral dentitions is based on Jain [74]. Fringing fulcra are named according to Patterson [75]. Scutes, unpaired and paired basal fulcra are identified according to López-Arbarello & Codorniú [22]. More specific problems of anatomical nomenclature related to discussions of homology will be explained in the following section 'Discussion of characters'.

Nomenclatural Acts
The electronic version of this document does not represent a published work according to the International Code of Zoological Nomenclature (ICZN), and hence the nomenclatural acts contained in the electronic version are not available under that Code from the electronic edition. Therefore, a separate edition of this document was produced by a method that assures numerous identical and durable copies, and those copies were simultaneously obtainable (from the publication date noted on the first page of this article) for the purpose of providing a public and permanent scientific record, in accordance with Article 8.1 of the Code. The separate print-only edition is available on request from PLoS by sending a request to PLoS ONE, 1160 Battery Street, Suite 100, San Francisco, CA 94111, USA along with a check for $10 (to cover printing and postage) payable to ''Public Library of Science''.
In addition, this published work and the nomenclatural acts it contains have been registered in ZooBank, the proposed online registration system for the ICZN. The Zoo Bank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ''http://zoobank.org/''. The LSID for this publication is: urn:lsid:zoobank.org:pub:BFFD7527-33BA-41D5-AF0F-CFD43625FDBE.

Discussion of Characters
Among basal neopterygians, ''semionotids'' are one of the most morphologically conflicting groups. ''Semionotiforms'' show morphological affinities with both halecomorphs and teleosts [7], and have been regarded as ancestors of at least some halecostome fishes [1,2]. Establishing the phylogenetic relationships of these fishes has been a challenge and this is largely due to the poor knowledge of the homology and evolution of several morphological characters.
The 90 parsimony-informative characters used in the present cladistic analysis are listed in this section. Some of the characters are newly proposed, while others are taken from previous authors. In the latter case, the source is clearly indicated. Wiley [76] performed several cladistic analyses of the phylogenetic relationships of gars with other neopterygians, and within the Lepisosteidae. For the purposes of this study, I took characters from his analysis of the relationships of chondrosteans, gars, amiids, and teleosts (indicated with a number followed with ''a''), and from his analysis of the relationships of Lepisosteus and Atractosteus to the Halecostomi and Chondrostei (indicated with a number followed with ''b'').
Newly proposed characters or characters significantly modified from previous authors deserve special discussion, and are, thus, explained in detail. Character state ''0'' does not necessarily represent the plesiomorphic condition because character polarity was determined by rooting the tree [77]. Character 1. Relative position of the dorsal fin.
0. Dorsal fin contained between pelvic and anal fins. 1. Dorsal fin opposite to anal fin. 2. Dorsal fin opposite to pelvic fins. 3. Dorsal fin originates anterior to pelvic fins and extends opposite to anal fin.
Cavin & Suteethorn ( [10]: 347) regarded the ''elongated body with the dorsal and anal fins located far backward, close to the caudal peduncle'' as a synapomorphy shared by gars and {Isanichthys. In the latter taxon, the dorsal and anal fins are not as remote as normally in the gars. However, among the studied taxa, only in the gars and {Isanichthys are the dorsal and anal fins fully opposite to each other and located backward.
The position of the dorsal fin relative to the pelvic and anal fins is a discrete feature, which is easy to evaluate. Quite the opposite, identifying cylindrical or elongated bodies is usually problematic and rather subjective. {Pliodetes specimens are never preserved in lateral view, but in dorsal or dorsolateral view and, thus, this fish apparently shares with the gars a cylindrical body shape. The only known specimen of {Isanichthys shows a very long and shallow body, approximately equally deep throughout the thoracic region, suggesting a circular cross section [10]. However, the fish is completely preserved in lateral view, not twisted as usually happen with fishes with circular bodies (like almost all the specimens of {Pliodetes) and, thus, the condition in {Isanichthys is doubtful. Similar doubts come up when trying to evaluate the condition of a possibly cylindrical body in other fishes with elongated bodies. Evaluating the feature ''elongated body'' also becomes problematic when trying to draw the line between elongated and not The presence of a quadratojugal is considered primitive in actinopterygians [7]. In basal actinopterygians the quadratojugal is a plate-like dermal ossification placed lateral to the quadrate and tightly bound to the preoperculum, the maxilla and the posterior margin of the quadrate in a very rigid cheek unit (e.g. see detailed descriptions in Gardiner [79] or Arratia & Schultze [73]). In these fishes, the quadratojugal carries a distinctive vertical pit line [73]. The quadratojugal is thus a superficial bone involved in the very rigid upper jaw and the sensory system. Above this primitive level, different conditions are found among neopterygians.
In ''semionotiforms'' and {Dapedium the bone identified as a quadratojugal is a splint-like dermal ossification lying along the dorsal margin of the preoperculum, with an anterior articular head that buttresses the articular process of the quadrate and a posterior spine-like portion. The symplecticum articulates between the quadrate and this posterior spine-like portion of the quadratojugal. Therefore, the condition in ''semionotiforms'' (State 1; Fig. 3A) is markedly different from that in basal actinopterygians and this splint-like quadratojugal plays a very different role in the skull. This splint-like bone is well inside the skull and is involved in the suspension of the lower jaw buttressing the palatoquadrate and transmitting forces between the quadrate and the preoperculum.
Although the topographic homology between the plate-like quadratojugal of basal actinopterygians and the splint-like quadratojugal of several neopterygians was proposed by Hammarberg [80] and supported by Patterson [7], it was questioned by Arratia & Schultze [73], who first expressed doubts about the homology among at least some of the different bones identified as quadratojugal in different osteichthyan lineages.
The macrosemiids have a splint-like quadratojugal, the most anterior portion of which is partially fused to the quadrate; the spine-like posterior portion is free (state 2; Fig. 3B) ( [39]; pers. obs.). In the gars the quadratojugal is also an independent splintlike bone with an articular head and a spine-like posterior portion, but it is notably larger than in other neopterygians (state 1). In teleosts, there is no independent quadratojugal, but the quadrate forms a spine-like posterior process, which has been considered homologous to the splint-like quadratojugal of ''semionotiforms'' and other neopterygians [7]. According to this hypothesis of homology, the quadratojugal is completely fused to the quadrate in teleosts (state 3; Fig. 3C).
The homology between the splint-like quadratojugal of Lepisosteus and the spine-like posterior process of the quadrate of teleosts has been supported by several authors [7,39,76,[81][82][83][84][85][86][87], but it has been questioned by Arratia & Schultze [73] and Arratia [88]. The similarity between the partially fused quadratequadratojugal complex of macrosemiids and the quadrate of basal teleosts is noteworthy ( Fig. 3B-C). Strikingly similar is also the development of the quadratojugal of Lepisosteus and the posterior process of the quadrate in teleosts. Hammarberg ([80]: p. 315) noted that in Lepisosteus platostomus ''Das Quadratojugale erscheint im 18.3-mm-Stadium als ein ä usserst dünner Knochenstab, der dicht and dem lateroventralen Rand des vorderen Teils des Palatoquadratum gerade hinter dem Unterkiefergelenk liegt'' (the quadratojugal appears in the state of 18.3 mm as a very thin rod of bone, which is positioned close to the lateroventral margin of the palatoquadrate, just behind the mandibular joint). In teleosts, the posterior process of the quadrate ossifies independently: ''… the posteroventral margin of the pars quadrata … close to the symplectic ossifies first, followed by the membranous ossification of the posterior process; the perichondral ossification of the body of the quadrate follows next'' ( [73]: pp. [67][68]. This early membranous ossification of the posterior process of the quadrate of teleosts further resembles the early ossification of the quadratojugal of Lepisosteous both morphologically and topologically (compare [73]: fig. 44B with the description in [80] and the photograph in [13]: fig. 25B). Although I have not found a separate or partially fused quadratojugal in a teleost, I defined an independent character state 3 assuming the homology of the posterior process of the quadrate of teleosts and the splint-like quadratojugal of other neopterygians. Since the character is unordered, this character state 3, which is restricted to the teleosts, does not affect the relationships within the in-group in this analysis, but allows a phylogenetic test for this hypothesis of primary homology. However, since only a few teleosts are here included as out-group taxa, this question of homology cannot be solved in the present phylogenetic study and should be tested in a more comprehensive cladistic analysis of basal neopterygians.
Finally, within Halecomorphi a small plate-like quadratojugal has been identified in one specimen of {Watsonulus by Olsen [53] and doubtfully in {Thomasinotus by Lehman [49], which would represent a condition similar to that in basal actinopterygians. However, no quadratojugal is present in the specimens of {Watsonulus described by Lehman [49] or the acid prepared specimens illustrated by Grande & Bemis [54]. Therefore, and considering that the quadratojugal is absent in all other known halecomorphs [7,54], the putative quadratojugal in {Watsonulus [53]  This character is the result of merging character 8 of Grande & Bemis [54] and character 2 of Grande [13]. In the first of these characters Grande & Bemis [54] distinguished between two degrees in the strength of ornamentation of the dermal bones of the skull: weak and/or fine (their character state 0) and strong, coarse (their character state 1). In his character 2 Grande ( [13]: 742) distinguished between the presence and absence of ''large, firmly anchored, pointed conical teeth covering the dermal bones of the skull''. As shown by Grande [13] this strongly toothed ornamentation is rare among actinopterygians, known only in the Cretaceous gars, in the clupeomorph Denticeps and in the {''paleonisciform'' {Coccolepis.
Wiley [76] interpreted the presence of single pair of extrascapulars (vs. two pairs in gars) as a synapomorphy of amiids and teleosts. However, more basal actinopterygians have a single pair of large extrascapular bones, as is the case in {Perleidus and {Watsonulus.
The number of extrascapular bones within a species might be variable and, thus, the condition should be checked in several specimens when possible. For example, some specimens of {Lepidotes mantelli have three pairs of extrascapular bones (NHMUK PV P.6336) while others have four pairs (NHMUK PV P.6933, 11832), and one specimen has three extrascapulars on one side of the skull and four on the other side (NHMUK PV P.20673a). Despite this variability, the patterns defined as the three states of this character were found to be stable within a species among the taxa studied here.
Character 21. Posterior extension of parietals median to the single pair of laterally placed extrascapular bones. 0. Absent. 1. Present.
In macrosemiids and a few ''semionotiforms'' the extrascapular bones are represented by one pair of small lateral elements only (Fig. 4) [39]. These bones are placed lateral to the parietals, and the median section of the supraoccipital commissure is enclosed in the posterior portion of the parietal bones. Olsen & McCune [8] interpreted the condition in macrosemiids as homologous with the two pairs of extrascapulars in gars. According to this hypothesis, a fusion of the medial pair of extrascapulars with the parietals is assumed. The fusion of extrascapulars with the parietals has been reported for several taxa (see discussion in Bartram [39]: 143) and  [39] it is present in some, though not all specimens of the Chinese species of {Sangiorgioichthys [34]. However, no macrosemiid demonstrates direct evidence of this fusion. Even if a fusion is assumed, it is not possible to be certain about the actual number of possibly fused extrascapulars.
0. Length of parietals less than half but more than one-third the length of frontals. 1. Length of parietals about half the length of frontals. 2. Length of parietals less than one-third the length of frontals.
Character 23. Length of frontals (from [74]; modified from [54]: character 34). 0. Frontals less than 3 times longer than their maximum width. 1. Frontals 3 or more times longer than their maximum width. Independently of a more or less developed inter-orbital constriction, the frontals are subrectangular in most basal actinopterygians and in most basal neopterygians (Fig. 5B). In {Semionotus and other ''semionotiforms'' the antorbital portion of the frontal narrows gradually anteriorly (Fig. 5C). In most macrosemiids, including the taxa considered in the present analysis, the frontals narrow abruptly and become almost tubular in the antorbital portion of the skull, enclosing the anterior portion of the supraorbital sensory canal (Fig. 5D) [39]. In {Semionotus bergeri, {S. capensis and {Luoxiongichthys the antorbital portion of the frontal is expanded laterally (Fig. 6). This expanded area has a triangular shape following the anterior rim of the orbit posteriorly and the series of anterior infraorbitals ventrally. Such an expansion is absent in the other studied taxa.
Character 28. Nasals long and narrow.
0. Supraorbitals do not contact infraorbitals at the anterior rim of the orbit. 1. Supraorbitals contact infraorbitals, closing the orbit.
For this and the following characters related to the circumborbital bones, a brief explanation is necessary concerning the chosen anatomical nomenclature and the homology of certain bones. Starting at the anterodorsal corner of the orbit and in clockwise direction, the following bones are here distinguished in the circumborbital series of the ''semionotiforms'', macrosemiids and gars: supraorbitals, dermosphenotic, infraorbitals, anterior infraorbitals, toothed infraorbitals, antorbital, and rostral (Fig. 7). Normally in neopterygians, the circumborbital series includes only supraorbital, dermosphenotic, infraorbital (including the so-called postorbitals, suborbitals and lacrimals), antorbital, and rostral bones (e.g. Fig. 7A). Anterior infraorbitals and toothed infraorbitals are unique features of the ''semionotiforms'', macrosemiids and gars (see below), with the latter bones being a unique specialization of the gars [8,76].
Perleidiforms and other basal neopterygians, as well as a few taxa considered as advanced stem neopterygians have a series of supraorbital bones forming the dorsal rim of the orbit between the nasal and the dermosphenotic [90]. Accordingly, the bones forming the dorsal rim of the orbit and placed lateral to the frontals and anterior to the dermosphenotic are here identified as supraorbitals, though in ''semionotiforms'', macrosemiids and gars the skull is elongated anteriorly and, thus, the nasals are far from the orbit and do not articulate with the supraorbital bones ( Fig. 7B-D). Under this topographic criterion, the identification of the supraorbital bones largely depends on the identification of the dermosphenotic. Poplin [91] summarized the problems concerning the identification of the dermosphenotic bone in nonteleostean actinopterygians. However, as a single bone placed at the posterodorsal corner of the orbit, laying on the sphenotic, and carrying the last portion of the infraorbital sensory canal, the identification of the dermosphenotic in ''semionotiforms'', macrosemiids or gars is usually not problematic ( Fig. 7B-D).
Anteroventral to the dermosphenotic follows the series of dermal bones associated with the infraorbital sensory canal, which border the orbit posteriorly and ventrally. Following the dermosphenotic these bones have been named postorbitals and suborbitals in Amia (e.g. [54,56]) and in Lepisosteus (e.g. [86]). They were called circumborbitals (e.g. [76,92]), infraorbitals (e.g. [93]), or subinfraorbitals and postinfraorbitals (e.g. [94]) in gars. In ''semionotiforms'' and macrosemiids they have generally been named infraorbitals (e.g. [7,18,[20][21][22][23]29,33,37,39,72,74]) but were also called circumborbitals in earlier works (e.g. [95,96]). Although one or more suborbital, a jugal, and one or more postorbital have been distinguished in this series, the number of infraorbital bones is highly variable among actinopterygians and individual homologies cannot be established [86]. The association of each of these bones with particular neuromasts of the infraorbital sensory line does not provide a valid criterion of homology because the number of neuromasts in this sensory line is variable, even between species of the same genus [85]. Furthermore, their number was shown to be variable between the left and right sides of the same specimen of L. platostomus [80].
Developmental studies [56,80,85,97] demonstrated that all the ossifications associated with the infraorbital line occur in connection with one or more neuromasts and go through the same developmental process. Therefore, serial homology (sensu [98]) can be assumed for the whole series from the rostral to the dermosphenotic. Also, some correspondences can be recognized in the development of these dermal ossifications in Amia [56] and Lepisosteus [80,85,97]. The rostral and the antorbital bones appear simultaneously and are among the first elements to ossify. The dermosphenotic appears much later than the rostral and the antorbital, but slightly earlier than one or more infraorbitals immediately below it. The series of infraorbital bones between the antorbital and the dermosphenotic gradually appears in caudally directed succession, starting with the few most anterior elements, which appear concurrently with the rostral and the antorbital.
The most anterior bones in the circumborbital series can further be distinguished because of their relationship with the sensory canals: the rostral with the ethmoidal commissure, the antorbital with the anterior connection between the infra-and supraorbital canals. Similarly, the dermosphenotic, as mentioned before, carries the last portion of the infraorbital sensory canal. Conversely, apart from their sometimes clearly defined position relative to the orbit and their peculiar morphology in some taxa, there is no valid criterion distinguishing individual elements among the infraorbital bones placed between the antorbital and the dermosphenotic. Therefore, taxic primary homology (sensu De Pinna [98]) is here accepted for the rostral, the antorbital and the dermosphenotic individually, and the series of infraorbital bones between the antorbital and the dermosphenotic as a whole.
In ''semionotiforms'', macrosemiids and gars, however, the anterior infraorbitals and toothed infraorbitals ( Fig. 7B-D) can be distinguished clearly within the series of infraorbital bones, on the bases of their morphology and position. These terms are thus being used to indicate these bones, which are only found in ''semionotiforms'', macrosemiids and gars, but individual homologies are not assumed. In the elongated ethmoid region of the skull of these fishes, the series of infraorbital bones starts far beyond the anterior border of the orbit, where it is represented by the so-called anterior infraorbitals and toothed infraorbitals in Lepisosteidae and Obaichthyidae, or by the anterior infraorbitals only in the ''semionotiforms''.
The term 'anterior infraorbitals' (after [21]) refers to the infraorbital bones placed anterior to the anterior border of the orbit and posterior to the antorbital, which do not contribute to the orbital margin ( Fig. 7C). Different names have been used for these bones in the literature: preorbitals [95], lacrimals [8,76], or anterior infraorbitals [21,72], among which the latter is preferred here because it highlights the homology of these bones with the other infraorbital bones (serial homology; see above).
The 'toothed infraorbitals' (after [76]), are placed between the antorbital and the anterior infraorbitals in lepisosteids and obaichthyids (Fig. 7D). These toothed dermal bones are rigidly attached to the ectopterygoid and pierced by the infraorbital sensory canal [13]. They have been regarded as 'maxillary bones' [99], 'lacrimals' [80,85], or 'infraorbitals' (Aumonier [97], who proposed their homology with the more posterior infraorbital bones surrounding the orbit [76]). The maxilla, which is extremely reduced, is fused to the most posterior toothed infraorbitals in lepisosteids (at some stage between the 75-150 mm specimens in L. osseus and between the 85-125 mm specimens in L. platostomus; data from [85]). The number and shape of the anterior infraorbitals is variable among taxa, but stable within a species. The number of toothed infraorbitals varies during the ontogeny [85], and their possible inter-and intraspecific variability in adults is unknown.
Character 30. Ventral border of infraorbital series flexes abruptly dorsally at the anterior margin of the orbit.
The circumborbital series of bones in lepisosteids and obaichthyids is peculiarly shaped, probably in relation to feeding adaptations [100]. In these fishes, the infraorbital bones at the anterior portion of the orbit become very narrow and the ventral border of the series flexes dorsally rather abruptly, following the orbit and the rounded coronoid process of the lower jaw (Fig. 7D). The lower jaw is then free to effectively move in a rapid strike [100]. Generally in basal actinopterygians (e.g. perleidiforms, ophiopsids, macrosemiids; [39,42,49,101], basal halecomorphs [18,42,54,102] and in basal teleosts [88] the supraorbitals are relatively small bones. This is also the case in many ''semionotids'' (e.g. In most ''semionotiforms'', the dorsal border of the anterior portion of the circumborbital series describes a convex curve, while the ventral border follows an only slightly concave curve. Accordingly, the depth of the anterior infraorbitals decreases gradually anteriorly, so that the most posterior anterior infraorbital is the deepest among these elements. In the macrosemiids the series of anterior infraorbitals is almost straight and, thus, the bones are all approximately equally deep. In contrast, in {Lepidotes  fig. 2A), the ventral border of the anterior portion of the circumborbital series follows a deep concave curve and the depth of the anterior infraorbitals becomes gradually larger anteriorly, so that the most anterior infraorbital is the deepest among these elements (Fig. 8).
Character 36. Relative size of the infraorbital bone (or bones) at the posteroventral corner of the orbit.  summarized in the three character states described above, which account for the condition observed in the studied taxa. In most ''semionotiforms'', the infraorbitals forming the posterior border of the orbit are relatively small bones, which are dorsoventrally elongated and sometimes almost reduced to a tube around the infraorbital sensory canal ( In most neopterygians the area of the cheek lateral to the quadrate is naked or protected by suborbital bones (see character 36). In lepisosteids, obaichthyids, {Pliodetes and {Araripelepidotes however, the series of infraorbital bones expands posteriorly and ventrally, covering the quadrate laterally (Fig. 7D).
Jain & Robinson [105] and Wenz [21] first attempted to classify the ''semionotids'' according the number and arrangement of suborbital bones. Wenz [21] presented three character states, which are equivalent to character state 0, 2 and 3 as defined here. Later, Cavin & Suteethorn [10] first included this character in a cladistic analysis using the three character states defined by Wenz and a fourth state representing the absence of suborbital bones. The same character was more recently used in the cladistic analysis of Cavin [12].
The presence or absence of suborbital bones is here represented with a separate character (41), because the presence of suborbital bones is independent of their number and arrangement. On the other hand, two character states have been added to represent the observed variability better. Several taxa have a stable number of two ({Lepidotes minor and the two species of {Macrosemimimus among the species included in this analysis). On the other hand, {Araripelepidotes, {Neosemionotus and {Tlayuamichin have three or four suborbital bones arranged in a series, but limited to the area posterior to the orbit. In all the fishes presenting the character state 2 (several suborbitals arranged in one row extending below the orbit), and only in these fishes, the suborbitals cover the quadrate bone laterally and, thus, this character state and the character 40 together account for the character 19 of Cavin ([12]; see comments above).
Cavin & Suteethorn [10] and Cavin [12] considered the pattern of suborbitals in {Isanichthys equivalent to the mosaic of suborbitals present in other taxa like the lepisosteids because there are two rows of suborbitals in the ventral region of the cheek in this fish. However, at least one specimen of {Lepidotes gigas (BSPG 1940-I-8; Fig. 8) and one specimen of {L. elvensis (MNHN JRE-250), species that normally present a single series of suborbital bones, also have irregularly arranged suborbitals in the ventral region of the cheek. Therefore, the pattern in {Isanichthys is here considered a normal deviation from character state 2.
Similarly, although in most specimens of {Sangiorgioichthys sui the series of suborbital bones is interrupted by an enlarged infraorbital that reaches the preoperculum, thus separating the suborbitals placed posterior to the orbit from the one or two elements placed lateral to the quadrate, a few specimens show a continuous series of suborbitals, like in the cases represented by the character state 2 ( The number of suborbital bones is much lower in {Obaichthys (two or three) and {Dentilepisosteus (three or four) than in the lepisosteids. However, since these few suborbitals are irregular in shape and size and irregularly arranged [13], the condition in the two obaichthyid genera is considered here homologous to the mosaic of suborbitals normally present in the gars.
Character 43. Independent of the total number, there is a large suborbital covering almost the whole area between the infraorbital bones and the preoperculum. Independent of the total number of suborbital bones, different patterns of suborbitals have been observed in those fishes with more than one suborbital bone. In gars and the ''semionotiforms'' with a mosaic of suborbitals, the suborbial bones are irregular in size and shape and no pattern can be defined, apart from the mosaic itself (state 0). However, three patterns steadily repeat in those fishes with more than one suborbital arranged in a row. In fishes such as {Lepidotes minor or {Macrosemimimus, the first (most dorsal) suborbital is relatively small, ovoid to subrectangular in shape and longitudinally elongate, and the second is notably the largest in the series and covers almost the whole area between the infraorbital bones and the preoperculum (Fig. 10A). This pattern also occurs in {Sangiorgioichthys, although this fish has a series of suborbitals arranged in one row.
Character 44. First and last suborbitals are larger than the other suborbitals. Olsen & McCune [8] considered the elongate nasal process of ''semionotiforms'', macrosemiids, gars, and Amia as a derived condition. Developmental evidence summarized by Wiley [76] suggests that the nasal processes of the premaxillae of Amia and gars are derived independently. However, due to the presence of a nasal process in most extinct halecomorphs and ''semionotiforms'', for which ontogenetic or developmental data are not available, the homology between the nasal process of gars and Amia should be tested in a cladistic analysis. Testing this hypothesis of homology is, however, not the purpose of the present analysis, since it would require a different data set including a much wider array of halecomorphs and other basal neopterygians. Therefore, pending further research, the homology of the nasal processes in all neopterygians is here assumed.
Favouring this assumption of homology, Patterson [7] pointed out the morphological, topographical and functional similarities of the nasal process of gars and Amia. In these fishes and in ''semionotiforms'' the nasal process lines the nasal pits, sutures with the frontal, and is perforated by the olfactory nerve.   Fig. 11). In {Araripelepidotes and {Pliodetes, the maxilla is very reduced but it is still an independent bone with a well-developed articular process (state 1; Fig. 9B). In the lepisosteids the maxilla is atrophied and fused to the ''toothed infraorbitals'' (State 2) [76,92].
The jaws of {Araripelepidotes are very peculiarly shaped [19]. They are well preserved and nicely exposed in the acid prepared specimens MNHN BCE-335 and BCE-336 (Fig. 9B). In these two specimens, the maxilla is a relatively small bone, the main body of which is laterally compressed, with convex dorsal and posterior borders, and a concave ventral border in MNHN BCE-335, but notably straight ventral border in MNHN BCE-336. The maxilla becomes rapidly shallower and laterally expanded anteriorly forming a dorso-ventrally compressed and anteriorly rounded medial process. The maxilla of ''semionotiforms'' is normally elongate, its depth being no more than half of its length. In a few taxa however, the maxilla is posteriorly expanded forming a deep plate (e. This character is taken from Thies [19] and refers to the acuminate process extending backwards from the ventral border of the dentary in {Lepidotes and other ''semionotiforms''. Cavin & Suteethorn ([10]: character 5) modified this character and considered the condition of the dentary of {Araripelepidotes as homologous to the condition in {Lepidotes as described by Thies [19]. However, the authors do not discuss this hypothesis in any detail and there is no comparable morphological structure or any evidence supporting the homology of the highly modified dentary of {Araripelepidotes (or any portion of it; Fig. 9B; [106]) with the posteroventral process of the dentary in other ''semionotiforms''.
The character was further modified by Cavin [12] by adding a character state 2 representing the condition of the dentary of gars. However, there is no evidence of homology for the condition in gars, the dentary of which extends to the posterior border of the lower jaw dorsal to the angular, and the condition in ''semionotiforms'' as defined here and described by Thies [19], which refers to a process extending backwards ventral to the angular. Only in {Dentilepisosteus, in addition to the expanded portion dorsal to the angular that normally occurs in gars, there is a short posteroventral process ( [13]: fig. 488), which closely resembles the posteroventral process of the dentary in {Lepidotes and it is thus here considered homologous to the latter. In the out-groups {Perleidus and {Watsonulus, and in {Araripelepidotes the preoperculum is a dorsoventrally elongated bone, which has no anteroventral arm (state 0; see [19]: figs. 1-2). In most ''semionotiforms'', as well as in Amia and basal teleosts, the preoperculum is a crescent-shaped bone and there are no welldefined dorsal and anteroventral arms (state 1; Fig. 7A-C). Distinctively in {Pliodetes, the preoperculum is L-shaped, with well defined dorsal and anteroventral arms forming an approximately right angle (state 2, Fig. 12; see [21]: figs. 5-7). The condition in gars resembles that of {Pliodetes, but the dorsal arm is variably reduced in the different taxa, and the anteroventral arm is notably larger than in ''semionotiforms'' (Fig. 7D; [13]). In ''semionotiforms'' the suboperculum has a well-developed ascending process, which is absent in non-neopterygian actinopterygians (Figs. 7, 10, 11). The distribution of this character among neopterygians is poorly known, although an ascending process is present in {Dapedium and Amia.
Character 65. Shape of ascending process of the suboperculum. The shape and relative height of the ascending process of the suboperculum is variable among ''semionotiforms''. The ascending process is usually narrow and acuminate towards the dorsal tip in most cases, but it is unusually broad and with rounded dorsal end in {Lepidotes maximus, {L. laevis, the lepisosteids, and {Dentilepisosteus.
Character 66. High ascending process of the suboperculum. 0. Less than or equal to half of the length of the dorsal border of the bone (Fig. 8). 1. More than half of the length of the dorsal border of the bone (Fig. 10).
In addition to the variation in shape, the height of the ascending process is usually less than half of the length of the dorsal border of the suboperculum in most taxa, but it is unusually high in {Lepidotes minor, {Tlayuamichin, {Macrosemimimus, {Paralepidotus, and {Semiolepis.
Character 67. Suboperculum less than half the depth of the operculum.
0. Absent (Fig. 7D, 12). 1. Present (Figs. 7A-C, 8, 10, 11). The depth of the suboperculum is normally less than half of the depth of the operculum, but the bone is deeper in most of the taxa with shallow opercula (character 54). Although characters 54 and 58 are based on relative measurements, the two characters are The presence of an interoperculum is a synapomorphy of Neopterygii. The bone has been secondary lost independently in Lepisosteidae and Siluridae (Teleostei). Wenz [21] mentioned the presence of an interoperculum in {Pliodetes. However, after detailed observation of the specimens of {Pliodetes in the MNHN (Paris), there is no independent interoperculum in this fish. The preoperculum of {Pliodetes is a robust L-shaped bone, which is firmly attached to the suboperculum. The preopercular canal is deeply excavated close to the anterior and dorsal margin of the preoperculum, and several branches of the main canal exit the bone through a series of relatively large pores aligned almost parallel to the dorsal border of its ventral (horizontal) arm (se holotype MNHN GDF-1275 in Fig. 12A). In some specimens, the ventral arm ventral to this series of pores is detached from the rest of the bone, thus resembling and independent interoperculum (e.g. MNHN GDF-1276 in Fig. 12B).
The presence of an independent interoperculum in obaichthyids has been clearly illustrated by Grande  The interoperculum is longitudinally elongated, deepest posteriorly at the suture with the suboperculum, and narrowing gradually in anterior direction. It places medial and ventral to the preoperculum and usually extends all along the horizontal arm of the latter (state 0; Figs. 7A, C, 8, 10, 11). Thus, the anterior border of the interoperculum is close to the posterior end of the lower jaw, to which it is connected through a ligament in Recent fishes. Bartram [39] noted that the interoperculum in macrosemiid fishes is smaller than usual and places well behind the lower jaw (state 1; Fig. 7B  Bartram ([39]: 218) discussed the peculiar condition of the uppermost caudal fin ray in macrosemiids, which ''does not insert beneath the squamation proximally, but remains superficial, and is not sharply delimited from the axial lobe scales''. He considered this condition as primitive and reported the same phenomenon in Lepisosteus osseus and {Acentrophorus varians, and partially in {Dapedium orbis and a species of {Caturus. Such a scale-like ray is also present in several ''semionotiforms'' (Fig. 13A).
Character 80. A constant number of exactly eight lepidotrichia in the lower, non-axial lobe of the tail (from [39]). Resembling the case described before, the gars also present a constant number of rays in the lower lobe of the caudal fin, but there are six in this case. Comparing the specimens of Lepisosteus osseus illustrated in the figures 89 and 94 in [13], the lateral line and the hinge-line or limit of the body lobe are good indicators of the limit between the upper caudal fin rays, which articulate with the hypurals, and the lower caudal fin rays, which articulate with the parhypural and precaudal haemal spines.
Character 82. Body lobe scale row (modified from [29] fig. 8) first noticed the variation related to the row of scales bordering the axial lobe of the tail in some ''semionotiforms''. In these fishes, like {Sangiorgioichthys sui, there is a complete row of elongated scales between the last scale of the lateral line and the uppermost caudal fin ray (Fig. 13A). In {Macrosemimimus fegerti and other taxa like {Tlayuamichin or {Semiolepis, there is an incomplete row of scales at the margin of the body lobe, in addition to the complete row described before (Fig. 13B). Normally in actinopterygians most of the scales of the body are articulated through the so called peg-and-socket articulation consisting in a dorsal spine-like peg protruding from the dorsal border of the scale (Fig. 15), which fits in a conical socket excavated in the medial surface of the scale. In some ''semionotiforms'' the scales have only very reduced pegs and sockets or this articulating structure is completely absent. The peg-and-socket articulation explained above function in dorso-ventral direction and is there is an anterior area, without processes, which is overlapped by the adjacent scale ( Fig. 15A;

Description of the Results of the Cladistics Analysis
The ''new technology'' and heuristic searches in TNT and PAUP* produced equivalent results. The heuristic search in PAUP* produced 88 most parsimonious trees (MPT) of 272 steps (CI = 0.4191; RI = 0.7304; HI = 0.5809; RC = 0.3061). The Strict Consensus Tree (SCT) of these MPT is identical in both PAUP* and TNT and is represented in Figure 16, in which Bootstrap and Bremer values are given above and below the branches leading to each of the well-supported nodes, respectively. A detailed list of synapomorphies is provided in Table S1.
Although {Dapedium was not defined as an outgroup, it joined the polytomy formed by the ingroup and the other out-group taxa at the base of the tree. Therefore, {Dapedium is not more closely related to ''semionotiforms'' than to halecomorphs or teleosts in this analysis.
In the SCT the ingroup form a well-defined monophyletic group with seven unambiguous and 10 ambiguous synapomorphies, Bremer value of 4 and Bootstrap of 76. Except for {Neosemionotus, all the other ingroup taxa split in two major clades indicated at nodes A and B in the Figure 16.
Four unambiguous and six ambiguous synapomorphies define the clade at Node A, which is supported with decay index of 1 and Bootstrap value of 44. Three monophyletic groups form a polytomy at the base of this clade. {Lepidotes minor Agassiz, 1833 [58], from the British Purbeck is recovered as the sister group of {Tlayuamichin itztli López-Arbarello & Alvarado-Ortega, 2011 [32], from the Early Cretaceous of Mexico. This relationship has bootstrap value of 77, decay index of 2, and two unambiguous and one ambiguous synapomorphies. The monophyly of the new genus from Europe, which is described in [70] Fig. 17).
The close relationships between {Pliodetes and {Araripelepidotes and the gars is very strongly supported with Bremer value higher than 4, Bootstrap value 99, and 15 unambiguous and 12 ambiguous synapomorphies at node C in Figure 16. A sister-group relationship between {Pliodetes and {Araripelepidotes is recovered in 82% of the MPTs (Fig. 17). Above these two taxa, the monophyly of the Lepisosteiformes sensu Grande ( [13]; i.e. Obaichthyidae and Lepisosteidae) is recovered with eight unambiguous and 14 ambiguous synapomorphies, decay index of 2 and Bootstrap value of 83. The family Obaichthyidae Grande, 2010 [13] is not monophyletic in the strict consensus tree or in the majority rule consensus (Figs. 16, 17), but the monophyly of the family Lepisosteidae is confirmed with bootstrap value of 97, decay index higher than 4 and nine unambiguous and seven ambiguous synapomorphies. In 82% of the MPTs {Isanichthys is the sister group of the clade formed by {Pliodetes and {Araripelepidotes and the gars.
The relationships of {Neosemionotus are not resolved in the strict consensus tree, but this taxon is the sister group of the clade defined at node B, leading to the gars, in 89% of the MPTs (Fig. 18).

Comparison and Discussion of Previous Phylogenetic Hypotheses for ''Semionotiform'' Fishes
As a pioneer in the field, the cladistic analysis of Olsen & McCune [8] was the first study to demonstrate the sister group relationships between gars, ''semionotids'' and macrosemiids in a clade they named Semionotiformes. Subsequently, the monophyly of this clade was confirmed by each and every cladistic analysis, in which ''semionotiform'' fishes have been included (Figs. 1, 2) ([9-13,108] and the present analysis). Similarly, the monophyly of gars was first demonstrated by Wiley [76] and subsequently confirmed by every cladistic analysis, with the most recent study [13] breaking with the idea of gars being more plesiomorphic than other neopterygians. Classified in the family Lepisosteidae in its own subclass Ginglymodi the gars were thought to be more plesiomorphic than the other neopterygians because they lack an interoperculum [75]. However, Grande [13] demonstrated that an interoperculum is present in the Cretaceous gars, which he classified in a separate family {Obaichthyidae. Reinforcing this evidence the cladistic analysis performed here confirms the presence of an interoperculum in the lineage leading to gars, which includes the fossil stem taxa {Pliodetes (without independent interoperculum) and {Araripelepidotes (with independent interoperculum). The present analysis also shows (like that of Cavin [12]) that the gars actually represent the crown group in one of two main clades at the base of the in-group (the clade defined at node B; see discussion of phylogenetic relationships below).
With the exception of the studies of Cavin & Suteethorn [10] and Cavin [12], the analysis presented here cannot be compared with previous cladistic analyses, apart from those patterns of higher-level relationships, because the data matrices are very different. The present analysis is the most comprehensive study of ''semionotiform'' fishes. The analysis is based on 30 [12] also included numerous ingroup taxa (28), representing almost all genera of gars, macrosemiids and ''semionotids'', but the various analyses performed by this author are based on only 42 and 45 informative characters, respectively (vs. 90 informative characters included in this analysis) and have several problems that will be discussed in detail below. Finally, except for the analysis of lepisosteids and their fossil relatives by Grande [13], this is also the first cladistic analysis of ''semionotiform'' interrelationships using real out-group taxa instead of hypothetical ancestors [8,10,12].
Phylogenetic analyses of Cavin [12]. Although based on less than half the number of characters, the cladistic analyses presented by Cavin [12] have 20 (out of 28) in-group taxa in common with the analysis presented herein. However, the relationships proposed by Cavin for several of these taxa are very different and, thus, deserve detailed discussion. Before discussing the differences between both studies it is worth noting that Cavin [12] already shows that the Cretaceous {Neosemionotus, {Araripelepidotes and {Pliodetes are stem taxa on the lineage towards Lepisosteidae and, thus, more closely related to the gars than to {Lepidotes or {Semionotus (Fig. 2). Cavin also recovered the basal position of the Middle Triassic {Sangiorgioichthys among ''semionotiforms'', macrosemiids and gars, but apart from these agreements the phylogenetic relationships proposed for the other taxa in common with the current study are controversial.
According to Cavin [12] the macrosemiids are the sister group to all ''semionotiforms'' and gars, which contrasts with the more derived position, well nested within the main clade defined at Node A in my analysis (compare Figs. 2, 16). One of the three analyses performed by Cavin ([12]: fig. 2), which he chose for the discussion of relationships, also produced two main clades representing the taxa more closely related to {Semionotus than to gars on the one side Further major discrepancies concern the relationships of {Isanichthys, and {Paralepidotus. According to Cavin {Isanichthys is more closely related to {Lepidotes or {Semionotus than to the gars. In my analysis, {Isanichthys is more closely related to the gars than to {Semionotus, and although the relationships between {Isanichthys, {Lepidotes and the gars are not resolved in the strict consensus tree, {Isanichthys is more closely related to the gars than to {Lepidotes in 82% of the MPTs (Fig. 17).
In Cavin's analysis {Paralepidotus is more closely related to the species of {Lepidotes and {Isanichthys than to {Semionotus or the macrosemiids, but the opposite relationships were produced by my analysis. In the latter, {Paralepidotus is placed in the major clade defined at Node A and including the species of {Semionotus and the macrosemiids, among other taxa, but not {Lepidotes or {Isanichthys. Moreover, {Paralepidotus is included in a monophyletic group, which is the sister group of the {Semionotus (Fig. 16).
Besides the different treatment of some characters (see comments in the discussion of characters above), I disagree with Cavin [12] in the scoring of a number of taxa. According to my own study of the same specimens and literature of {Lepidotes lattifrons Jain & Robinson, 1963 [105], from the Oxford Clay, I was not able to confirm the scoring of at least 10 out of 29 characters Cavin scored for this species. The main problem with this species is that it is represented by two completely disarticulated specimens and a third fish with articulated postcranium preserved in lateroventral view and almost completely disarticulated skull. Therefore, many of the characters scored for {L. lattifrons seem to be based on reconstructions or assumptions about how the complete articulated fish would have looked.  fig. 1a) is noteworthy. The material of {L. tendaguruensis was first identified as {L. minor [110] and I will explain in the following section (first level beta-taxonomy) that, based on the available material, the two species are almost indistinguishable. Most of the features proposed by Arratia & Schultze [109] as diagnostic for {L. tendaguruensis are also present in {L. minor, and only after thorough analysis and comparison was I able to confirm the validity of the former species on the basis of two characters (see below). Cavin ([12]: supplementary material) did not examine the specimens of {L. tendaguruensis first hand, and, based on [109], he scored four characters with different states for {L. tendaguruensis and {L. minor, his characters 11, 20, 27, and 42 (first cladistic analysis). Thus, based on the evidence available to him, he scored the frontals as being narrower anteriorly than posteriorly in {L. tendaguruensis (ch. 25 (1)

tendaguruensis.
Cavin's character 27 refers to the presence of tritoral dentition, which he considered absent in {L. tendaguruensis, as reported by Arratia & Schultze [109]. However, it will be explained below that the condition is unknown in this species, because coronoid and pterygoid dentitions are not preserved in the known specimens of this species. Finally, although there are ''dorsal ridge scales lacking a posterior spine'' (Cavin's ch. 42(0)) in {L. tendaguruensis according to Arratia & Schultze ([109]: p. 138), I was not able to verify this feature with certainty in the poorly preserved specimens representing this species (see [109]: figs. 2-3).
Other disagreements with character scorings of Cavin [12] will not be discussed in detail here because the main problem I find in his analyses is the use of an artificial hypothetical outgroup. Cavin [12] does not explain how he inferred the hypothetical ancestor, and applying the outgroup algorithm of Maddison et al. [111] it is not possible to recover the same ancestral states that he used in his analyses. The use of hypothetical ancestors has fallen into disuse and Bryant ([112]: p. 345) has shown that ''the use of a priori hypothetical ancestors as additional terminal taxa is either potentially problematic (outgroup comparison), or invalid (ontogenetic and paleontological methods)''. Therefore, and since Cavin [12] also included two real out-group taxa in his data matrices (Amia calva and {Leptolepis coryphaeonides), I re-analysed his data matrices excluding the hypothetical outgroup and the results are shown in Figure 18. Thus, using real outgroups for the data matrix presented by Cavin [12] leads to significantly lower resolution of the tree, and most of the inconsistencies between his analyses and the one presented here are no longer present in the strict consensus tree.

First Level Beta-taxonomy
Based on the results of the present cladistic analysis described above, the following taxonomic changes are proposed ( Table 1). The generic diagnoses are based on unambiguous synapomorphies only but, additionally, distinctive combinations of features are provided to facilitate identifications.
Additionally, the following combination of features is distinctive of {Lepidotes: large fusiform fishes with body depth c. 35% of the standard length (SL) and head length c. 30% SL; pelvic, dorsal and anal fins placed in the posterior half of the body, the pelvic fins inserting at c. 55% SL, dorsal fin inserting at c. 65% SL, and anal fin inserting at c. 75% SL; the presence of a single pair of extrascapular bones (ACCTRAN); numerous suborbital bones of variable size and shape, arranged in a series, which extends ventral to the orbit covering the quadrate laterally; thick ganoid scales with strongly developed longitudinal articulation through large dorsal and ventral anterior processes.
Remarks. {Lepidotes has been one of the largest ''wastebasket'' genera of Mesozoic actinopterygians and most of the species previously referred to this genus either represent independent taxa or should be regarded as nomina dubia.
The genus {Lepidotes Agassiz, 1832 [25] was erected for two fish specimens from the Posidonienschiefer (Toarcian) at Ohmden near Boll in Germany. Some years later Agassiz found this fish indistinguishable from a specimen from the Lias (Toarcian) of La Caine in France, which had already been named {Cyprinus elvensis Blainville, 1818 [69] and, thus, he put the two species in synonymy, but kept the name {Lepidotes gigas for this taxon [58]. Later Quenstedt [113] proposed the combination {Lepidotes elvensis for the German and French nominal species. Although the three specimens of the French species at the Muséum National d'Histoire Naturelle in Paris (the holotype MNHN JRE-545 and two other specimens MNHN JRE-250, 254) are more poorly preserved than the German material, a few anatomical differences support the validity of two different species (Fig. 20). Accordingly, {L. gigas Agassiz, 1832 [32] and {L. elvensis (Blainville, 1818) [69] differ in the general shape of the skull, the number of supraorbital bones (2 vs. 3 respectively), the relative size of the first, most dorsal suborbital bone, which is relatively trapezoidal and largest in {L. gigas while triangular, narrowing posterodorsally in {L. elvensis. Additionally, the maxilla is somewhat larger and the snout a little longer in {L. elvensis than in {L. gigas. Though the number of anterior infraorbital bones is the same, the frontal and most anterior supraorbital are slightly differently arranged in these species, so that the frontal extends over three anterior infraorbitals in {L. elvensis, but over two anterior infraorbitals in {L. gigas. Although {L. elvensis has been erroneously cited as the type species of the genus (e.g. [16,18,95,109] [114] from Germany differ from the previously described species in having strongly serrated scales and tritoral dentition [115,116]. The dermal bones in the skull of {L. bülowianus including all circumborbital and suborbital bones are furthermore very densely ornamented with ganoine tubercles, which are absent in {L. semiserratus. The precise limits and relationships between these four coeval species of {Lepidotes need further study. A thorough revision of all the available material has not been done so far and it is not yet possible to assert if these {Lepidotes species mirror the endemism shown by plesiosaurs, ichthyosaurs and marine crocodiles within in the lower Toarcian seas of Western Europe. However, based on the published material, they seem to follow the three or four marine reptile zones proposed by Godefroit [117] and Maisch & Ansorge [118]. Diagnosis. Three or more pairs of extrascapular bones; in the series of suborbital bones that extend ventral to the orbit covering the quadrate laterally, the first and last suborbitals are the largest; dentition extremely tritoral; strong knob-like anteroventral process in posttemporal bone; orbital sensory canal present; middle pit line contained in a groove excavated in dermopteroticum and parietal.

Genus
Additionally, the following combination of features is distinctive of {Scheenstia: large fishes with fusiform bodies with body depth c. 40-45% of the standard length (SL) and head length c. 30% SL; pelvic, dorsal and anal fins placed in the posterior half of the body, the pelvic fins inserting at c. 50-53%, dorsal fin inserting at c. 65-70% SL, and anal fin inserting at c. 75-78% SL; infraorbitals at the posterior border of the orbit longer than deep; maxilla edentulous, very short and deep (ACCTRAN), ends at the level or before the anterior border of the coronoid process; thick ganoid scales with vertical peg-and-socket articulation variably developed and very well developed longitudinal articulation through large dorsal and ventral anterior processes.
Remarks. The close relationship between {Scheenstia zappi and {Lepidotes was already put forward by López-Arbarello & Sferco [37]. These authors also discussed the close resemblance between this fish and the large tritoral forms that have been referred to {Lepidotes (i.e. {L. laevis, {L. mantelli, {L. maximus), which are now transferred to {Scheenstia based on the derived characters shared by them and the type species of this genus, {S. zappi. {Scheenstia mantelli (Agassiz, 1833) [58] is certainly the best known among these tritoral fishes [95]. {Scheenstia laevis (Agassiz, 1837) [58] is best known from an excellently preserved though incomplete specimen described by Saint-Seine [121].
Probably the largest and more impressive species in this genus is {Scheenstia maximus (Wagner, 1863) [119] (Fig. 21). The type material of this species was stored at the Bayerische Staatssammlung für Palä ontologie und Geologie in Munich when Wagner ([119]: 19) described the species, but it is unfortunately lost. The material was most probably destroyed during the Second World War, as many other specimens in this collection, the house of which was severely bombed. The type material included two fragmentary specimens containing several articulated scales. Wagner described only one of them, consisting of a fragment (approximately 61 cm high x 37 cm long) of a large fish including several articulated scales mostly exposed in medial view, though at  least some of them exhibited their lateral surface. Unfortunately, Wagner did not illustrate the specimen, which is neither figured nor described in any other publication. However, the characteristics described by Wagner for the scales in the type specimen, perfectly match the scales in the specimens SMF P.325 and SMF P.2386 of the Senckenberg Museum in Frankfurt, which has been studied by Jain [103]. According to Wagner the type specimens were found in the Solnhofen limestones of Kelheim, Solnhofen and Eichstä tt. The two almost complete specimens in the Senckenberg Museum come from the Solnhofen limestones at Langenaltheim, which represents the same depositional centre as the locality of Solnhofen and are well correlated with the equivalent outcrops at Kelheim in the Rueppellianus Subzone, and with those of Eichstä tt in the Hybonotum Zone (lower Tithonian; [122]). Therefore, the specimen SMF P.2386 (Fig. 21) is here designated neotype of {Scheenstia maximus (Wagner, 1863) [119] new combination, to provide objective evidence for this species and avoid confusion over its characteristics.
A second species described by Wagner [119], {Scheenstia decoratus from Solnhofen (Hybonotum Zone, Solnhofen Formation; early Tithonian; [122]) is represented with a rather complete specimen only (Fig. 22). Although the skull is only partially preserved, the holotype is generally very similar to the recently described {S. zappi from Schamhaupten (Beckeri Zone, Rögling Formation; latest Kimmeridgian; [122]), but differs from this species in the ornamentation of the skull bones, which is made up of densely arranged broad tubercles and ridges that reach the free margin of the suborbital and infraorbital bones producing a crenulated border, very different from the much more sparsely and smaller tubercles with no ridges in {S. zappi, the lower jaw is notably more robust and the scales more strongly serrated in {S. decoratus than in {S. zappi. Due to the incomplete preservation it is not possible to take exact measurements in the holotype and so far only known specimen of {S. decoratus, but the body is somewhat more slender and the head was certainly smaller than the head of {S. zappi. Although none of the fins in {S. decoratus is complete enough to allow detailed comparison, further differences in the body are the total number of vertical rows of scales (38 vs. 37), the number of inverted rows of scales forming the body lobe of the tail (8 vs. 10). Two poorly known species from the German ''Wealden'', {S. degenhardti Branco, 1885 [120] and {S. hauchecornei Branco, 1885 [120] are tentatively referred to {Scheenstia, thought they need detailed revision.
Etymology. From the Ancient Greek ''calli-'', beautiful, and Purbeck, the current name of the area inhabited by the fish.
Diagnosis. The following combination of features is distinctive of {Callipurbeckia: medium size semionotiform fishes with fusiform bodies with body depth c. 45% of the standard length (SL) and head length c. 30% SL; pelvic, dorsal and anal fins placed in the posterior half of the body, the pelvic fins inserting at c. 50%, dorsal fin inserting at c. 65% SL, and anal fin inserting at c. 75% SL; skull bones ornamented with tubercles; single pair of extrascapular bones; two suborbital bones, a small oval dorsal suborbital and a much larger ventral suborbital filling most of the area between the infraorbitals and preoperculum; maxilla deep, forming a more or less circular plate; dentition moderately tritoral; conspicuous dorsal ridge of scales; ganoid scales with well developed vertical and longitudinal articulation with large dorsal peg and large dorsal and ventral anterior processes.
Remarks. Agassiz [58] coined the binomen {Lepidotes minor for a species commonly found in the Purbeck sequences at Swanage, which he represented with a specimen in the collection    [32] from the Albian of Mexico. Although this sister-group relationship is as strong as the relationship shown by other species within a single genus, several apomorphic features of {Tlayuamichin (see diagnosis in [32]) in addition to the geographic and chronostratigraphic differences between the two species support the establishment of separate genera.
Among the species referred to this genus, the ''semionotiforms'' from the Upper Saurian Bed (Late Jurassic: Tithonian) in Tendaguru, Tanzania, representing the species {Lepidotes tendaguruensis Arratia & Schultze, 1999 [109], were originally referred to {Lepidotes minor [110] and after studying the material at the Museum für Naturkunde in Berlin, I find little evidence supporting different species. The few known specimens are very poorly preserved and it is not possible to corroborate several detailed anatomical features proposed in the diagnosis of {C. tendaguruensis. Since no detail structures like the distinct sockets are observable in the bone identified as the epiotic in {C. tendaguruensis, I am not sure if this element actually represents this bone. Nonetheless, even accepting this interpretation of the bone, a digitated posterior process is common to all ''semionotiforms'' for which the epiotic is known (see Cavin [12]: character 1). The series of supraorbitals is surely incompletely preserved in {C. tendaguruensis and at least one supraorbital is missing anteriorly (see [109]: figs. 4, 6-7). Therefore, there are certainly more than two supraorbital bones in these fishes and there are three supraorbitals in {C. minor (NHMUK PV P1118, P8047, P36080). The relative size and shape of the two suborbital bones, as well as the teeth on the premaxilla and dentary are basically the same in both nominal species.
A very peculiar feature reported in {C. tendaguruensis is the presence of two most anterior infraorbital bones horizontally placed, one dorsal to the other. According to my observations the two ''anterior infraorbital bones or antorbitals, rectangular shaped and placed above each other'' ( [109]: 138) in MB. f.7048 rather represent two fragments of the most anterior anterior infraorbital, dorsal and ventral to the sensory canal, which is deeply excavated in the infraorbital bones of {C. tendaguruensis (see [109]: fig. 6). Accordingly, {C. tendaguruensis would have three and not four anterior infraorbitals, as is the case in the neotype of {C. minor. However, the anterior region of the skull of {C. tendaguruensis is not completely preserved in any of the known specimens and, thus, the exact number of anterior infraorbitals is unknown.
Arratia & Schultze ( [109]: p. 145) described two rows of teeth on the dentary (their dentalosplenial) of {C. tendaguruensis, which would also represent a remarkable feature. The presence of two rows of teeth on the dentary, a lateral row of small pointed teeth plus an inner row of much more robust fangs is a unique feature of Lepisosteus and Atractosteus and unknown in ''semionotiforms'', which have only one row of teeth on the dentary (see character 85 in this cladistic analysis and [13]: character 39). However, the only well preserved dentary of {C. tendaguruensis in the specimen MB. F.7043 has a single row of teeth (pers. obs.; see also [109]: figs. 5A and 10C). On the other hand, tritoral dentition is alleged to be absent in {C. tendaguruensis, but the tritoral teeth of {C. minor, as in many other semionotiforms, are not dentary teeth, but on the coronoid bones (NHMUK PV P.29399, 17329), and they were described in detail by Jain ([74]: 30). Therefore, since no coronoid bone is preserved in any of the specimens of {C. tendaguruensis, there is no evidence for the alleged absence of a tritoral dentition.
As mentioned before, the specimens of {C. tendaguruensis are poorly preserved and I was not able to confirm the anterior membranous outgrowths of the hyomandibula described and illustrated by Arratia & Schultze ([109]: compare the photograph of the cast in fig. 6A with the interpretative drawing of same cast in fig. 7B). Similarly, the series of postcleithra and the dorsal ridge scales are incompletely preserved. On the other hand, the alluded absence of fringing fulcra in {C. tendaguruensis is erroneous. Fringing fulcra are present at least in the pectoral (MB. f.7040) and dorsal (MB. f.7041) fins of this species (Fig. 24).
Therefore, the only two features that distinguish {C. tendaguruensis from {C. minor are a comparatively short preoperculum that does not reach the dermopterotic and the ventroposterior expansion of the infraorbital bone placed at the posteroventral corner of the orbit, as noted by Arratia & Schultze [109]. The shape of the infraorbital bones is somewhat variable individually in all ''semionotiforms'' I have examined and one to one relationships of homology cannot be established for the individual bones in the infraorbital series (see the above discussion of characters). However, all infraorbital bones from the posteroventral to the anteroventral corner of the orbit reach the depth of their adjacent elements in the series in {C. minor, but not in {C. tendaguruensis, in which the infraorbital bone at the posteroventral corner of the orbit expands ventroposteriorly respect to the circumference drawn by the other infraorbital bones (compare [109]: fig. 7A with Fig. 24). Based on the two latter features, the species named by Arratia & Schultze is here confirmed as valid and based on its close resemblance with {C. minor it is referred to {Callipurbeckia gen. nov.
Although it needs thorough revision, another species of this genus is probably {''Lepidotes'' notopterus Agassiz, 1833 [58]. According to Woodward [95] this species might occur in the Wealden Formation. However, the type specimens described by Agassiz came from the Solnhofen limestones. I have observed several unstudied specimens from the Solnhofen limestones, which are almost indistinguishable from {C. minor (e.g. MB. f.17878). I was not able to find any of the type specimens in the Natural History Museum and it is not clear whether the fish from the Wealden described and figured by Woodward actually represents this species or a different, still unnamed taxon (note the important chronostratigraphic difference between the Wealden and Solnhofen formations). Although strikingly similar, {C. notopterus apparently differs from {C. minor in some morphometric proportions and a few meristic and osteological features (pers. obs.).
Within the clade Lepisosteiformes, the relationships between {Lepidotes, {Scheenstia, {Isanichthys and the remaining studied lepisosteiforms are unresolved in the strict consensus tree. Nonetheless, in 82% of the MPTs, {Isanichthys is more closely related to Lepisosteus than to {Lepidotes, and {Scheenstia and {Lepidotes are sister groups. The latter relationship suggests that the family Lepidotidae Owen, 1860 [15], probably represents a natural group, but the present analysis does not provide enough evidence supporting this hypothesis.
One of the most interesting results of the analysis is the rearrangement of many of the species so far classified in the genus {Lepidotes. A monophyletic {Lepidotes Agassiz, 1832 [25], is restricted to a few species from the Early Jurassic of central Europe, two of which have been included in this analysis: {L. gigas Agassiz, 1832 [25], and {L. semiserratus Agassiz, 1836 [58]. Most of the species previously referred to this genus that were included in this analysis do not join the monophyletic {Lepidotes, but other recently defined taxa (Fig. 16 [37], and, thus, based on six unambiguous synapomorphies and very high Bootstrap and Bremer values, these three species are here refer to {Scheenstia. On the other hand, as explained before, {''Lepidotes'' minor represents an independent genus {Callipurbeckia gen. nov., which is more closely related to {Semionotus and the macrosemiids within the Semionotiformes than to {Lepidotes. The close relationship of {Pliodetes and {Araripelepidotes with the lepisosteids and obaichthyids sensu Grande [13] is very strongly supported. Most of the synapomorphies proposed by Grande [13] for his Lepisosteiformes are endocranial features unknown in {Pliodetes and {Araripelepidotes (Grande's characters 2,32,59,60,[63][64][65]77). However, among the derived lepisosteiform features according to Grande [13] and although the junction between the supraorbital and infraorbital canal occurs in the dermosphenotic of {Araripelepidotes (AMNH 11813), the supraorbital canal does not penetrate the parietals in {Araripelepidotes or {Pliodetes (Fig. 9C-D; [21]: 112; pers. obs.). Furthermore, although the junction occurs in different bones, the general pattern followed by the supraorbital, infraorbital and temporal canals is basically the same in {Araripelepidotes, {Pliodetes and the gars. On the other hand, {Pliodetes shares with the gars some typically lepisosteiform features like the L-shaped preoperculum, the nasal processes of the premaxillae forming an external dermal component of the skull roof and bearing the supraorbital sensory canal, a mosaic of suborbital bones, and the absence of an independent interoperculum [13,76]. {Pliodetes further presents two of the synapomorphies proposed by Grande [13] for the Obaichthyidae: large conical teeth firmly anchored to the surface of most of the dermal bones of the skull and rostral region elongated well anterior to the lower jaw symphysis by over 50% of the mandibular length (Grande's characters 2 and 4 respectively). Also, the flank scales of {Araripelepidotes and {Pliodetes closely resemble the scales of obaichthyids in forming one or two large prominent spines at their posterior margin (Fig. 14). Consequently, according to the evidence discussed above, I consider {Araripelepidotes and {Pliodetes as basal gars and propose the name Lepisosteoidei for the clade defined at Node C in Figure 16.
Within Lepisosteoidei, the close relationships of {Obaichthys and {Dentilepisosteus with the Recet gars is very well supported. This arrangement is acknowledge as the Lepisosteiformes by Grande [13], but according to this study it represents an infra-ordinal rank and is here regarded as a superfamily Lepisosteoidea. The family {Obaichthyidae Grande, 2010 [13] is not recovered as a monophyletic group in this analysis (Figs. 16,17). Nevertheless, Grande's data matrix is more adequate than the matrix used for this study to solve the relationships within Lepisosteoidea because it includes more lepisoteoid taxa and more characters that are significant to establish those relationships. Therefore, I have no reason to question the results of the analysis carried out by Grande [13] and I accept the sister group relationships between {Obaichthys and {Dentilepisosteus in the clade {Obaichthyidae. Similarly, the family Lepisosteidae is here accepted in the more restricted sense of Grande [13], for which very high Bremer and Bootstrap values were obtained (Fig. 16).
Lepisosteoidea. The clade including all taxa more closely related to {Obaichthys or to Lepisosteus than to {Pliodetes or {Lepidotes.

Character Evolution in Ginglymodi
In addition to the comments already made in the section ''Discussion of Characters'', the evolution of certain characters deserve further and more detailed discussion. Two main features are distinct and stable among Ginglymodians: the presence of anterior infraorbitals and the absence of gular plates. The gulars however are also absent in other neopterygians like aspidorhynchids or osteoglossomorphs and more advanced teleosts [9,63], but the anterior infraorbitals represent a very interesting feature uniquely derived in Ginglymodi. In Neopterygii the infraorbital bones are serial homologous and they develop in relation to the organs of the infraorbital sensory canal. The development of the dermal bones of the infraorbital sensory canal in Amia calva was described in detail by Pehrson [56] and can be summarized as follows. The formation of the canal bones in the skull of Amia calva starts early in the anterior part of the canal system and proceeds posteriorly. Pehrson defined two stages in the formation of these dermal ossifications. In the first stage the osteoblasts are formed and migrate under the epidermis to form the primary blastemas under each separate sense organ, and a stratum of osteoblasts under the future canal. The next stage is the formation of the secondary blastemas as a result of the gathering in the mesenchyma of the previously formed osteoblasts. These secondary blastemas do not always arise in connection with each separate sense organ and a single secondary blastema may be connected with more than one sense organ.
The rostral, antorbital and first infraorbital (lacrimal) bones (Fig. 7) develop first and nearly simultaneously. The primordia for the antorbital and first infraorbital (lacrimal) are already visible in a 11.5-mm specimen. The antorbital primordium is associated with the sense organs 3 to 6. The first infraorbital (lacrimal) primordium is associated with the organs 7 and 8. The two bones are already formed in a 12 mm specimen. Also in a 12 mm specimen, each of the two first sense organs on each side in the infraorbital series appear in connection with a separate, blastematic rostral primordium in the second stage of development.
These four primordia will later fuse to form a single cylindrical rostral bone. The development of the more posterior infraorbitals and the dermosphenotic proceeds gradually posteriorly. A rudiment of the first of these elements is found under the sense organ 9 in a 12 mm specimen, and in a 13.8 mm specimen the primordium for the following bone is formed under organ 10. In a 16.1 mm specimen the primordia for the postorbitals, except the last element, and the dermosphenotic are formed, and, thus, the dermosphenotic forms earlier than the infraorbital bone immediately below it [56].
In the case of Lepisosteus there are three main reference works concerning the development of the dermal bones in the skull: Hammarberg ([80]; L. platostomus), Aumonier ([97]; L. osseus), and ( [85]; L. osseus and L. platostomus). Among them, Hammarberg [80] includes the more detailed and complete description of the development of the bones around the infraorbital sensory canal. The first bones to develop in this series in Lepisosteus are the rostral, the antorbital, and the toothed infraorbitals. On each side, the rostral primordium appears in a 18.7 mm specimen of L. platostomus, in connection with the first neuromast in the infraorbital series [80]. In a 33.4 mm specimen the primordial rostral had extended backwards up to the third neuromast, the ethmoidal connection between the two infraorbital lines is established in a 44.2 mm specimen, and the cylindrical rostral is almost completely formed in a 65.4 mm specimen [80]. In L. osseus, Jollie [85] found the first evidence of the rostral in a 29 mm specimen. The antorbital primordium appears in a 18.7 mm specimen of L. platostomus and is associated to the neuromasts 4 to 7, and it is a well-formed tubular Y-shape bone in the 65.4 mm specimen [80]. In L. osseus the antorbital (lateral rostral in Jollie [85]) appears in a 28 mm specimen, below and anterior to the already formed first toothed infraorbital, and is associated to 4 or 5 neruomasts [85].
The first, most anterior primordial elements of toothed infraorbitals have no teeth and appear rather rapidly. There are already four primordia in a 18.7 mm specimen of L. platostomus [80]. The formation of the remaining toothed infraorbitals proceeds more slowly posteriorly. There are seven primordia in a 25 mm specimen and 11 primordia in a 49 mm specimen of L. osseus [85], and 13 and 14 primordia on each side of a 65.4 mm specimen of L. platostomus [80]. The teeth of the toothed infraorbitals form independently of the bones in the mouth margin below them. The teeth attach later to the toothed infraorbitals, starting at about the stage of a 39 mm specimen [84]. The vestigial maxilla also attach to the series of toothed infraorbitals at some stage between 75 and 150 mm specimens in L. osseus, and 85 and 125 mm specimens in L. platostomus [85].
The more posterior bones in the infraorbital series appear as a different series, which starts forming in a 54.2 mm specimen of L. platostomus [80], and this series is complete in a 75 mm specimen of L. osseus and an 85 mm specimen of L. platostomus [85]. The dermosphenotic forms some time before the infraorbital bones below it, and its blastema is found in the 54.2 mm specimen of L. platostomus [80].
The developmental patterns summarized before show that all the ossifications associated with the infraorbital sensory canal undergo the same process and serial homology can be assumed for the whole series from the rostral to the dermosphenotic [98]. However, due to their topographic relationships and early and simultaneous ontogenetic appearance, individual homology is accepted for the rostral and antorbital bones in Amia and Lepisosteus as already proposed by Hammarberg [80], Patterson [62], and Jollie [85], independently of the number of neuromasts associated with each bone. Similarly, the individual homology for the dermosphenotic bone in these taxa is supported by its position and out of turn development compared with the other infraorbital bones. The other infraorbital bones including the anterior infraorbitals, but not the toothed infraorbitals, developed gradually in the series and individual homologies cannot be established for any of them in particular. However, the topographic relationships of the ginglymodian anterior infraorbitals are unique among actinopterygians. Based on this topographic criterion, the hypothesis of primary homology has been proposed and tested in the cladistic analysis, resulting in an unambiguous and uniquely derived synapomorphy of the Ginglymodi. Therefore, within this clade, secondary homology is accepted for this portion of the infraorbital series, taken as a whole and restricted to the area between the antorbital and the first infraorbital bone forming the rim of the orbit.
In Lepisosteus, the most posterior toothed infraorbitals form later (at 60 to 65.4 mm stages) and ventral to the first, most anterior infraorbitals (at 54.2 mm stage) [80]. Therefore, there are two independent series: the series of toothed infraorbitals and the series of infraorbital bones, including the anterior infraorbitals (Fig. 7). It has been interpreted as a novelty of gars and, although serial homology with the infraorbital series ventral and posterior to the orbit is indicated by their development, this series of toothed infraorbitals has no known homologous structures in other actinopterygians. As shown by the cladistic analysis, the series of toothed infraorbitals appeared only once in the Lepisosteoidea and also represent a case of secondary homology.
A Splint-like quadratojugal is a unique feature of the Ginglymodi and their probably stem-taxon {Dapediidae. According to Patterson [83], the evolutionary trend in teleosts is towards the complete fusion and reduction of the quadratojugal, which might be limited to the spine-like posterior process of the compound quadrate in advanced teleosts, and a similar trend is observed in some semionotiforms [39]. The plate-like quadratojugal of basal actinopterygians contribute to the rigid upper jawcheek-palatoquadrate complex. The upper jaw becomes free and mobile in neopterygians and there are changes in the mode of suspension of the lower jaw in these fishes. Patterson [75] reinterpreted the ''symplectic'' of basal actinopterygians like {Pteronisculus, {Boreosomus and {Australosomus [135][136] and chondrosteans as an interhyal and proposed that the symplectic is a synapomorphy of the Neopterygii. I agree with Patterson and find no sustainable evidence for a symplectic outside Neopterygii. In neopterygians, the symplectic develops from the antero-ventral portion of the hyomandibular cartilage and contributes to the suspension of the lower jaw directly or via the quadrate and/or the quadratojugal [7]. The direct contribution to the suspension of the lower jaw occurs in the halecomorphs, in which the symplectic articulates directly with the lower jaw, as well as the quadrate. In the non-halecomorph neopterygians the symplectic contributes to the suspension indirectly. In teleosts the symplectic fits into a medial groove formed by the spine-like posterior process of the quadrate and the body of the quadrate. In ginglymodians and dapediids the symplectic articulates with the quadratojugal only (lepisosteids) or with the quadratojugal and the quadrate. In the latter case, the quadratojugal is a buttress firmly bound to the articular process of the quadrate (sometimes even partially fused to it) and the two bones form a medial groove that receives the symplectic. Among ginglymodians, the trend in lepisosteids is towards the enlargement of the quadratojugal, which becomes a bridge bone supporting the quadrate at its anterior end, and receiving the support of the symplectic at its posterior end. The suspension of the lower jaw is displaced forwards in lepisosteids; the quadrate places anterior to the orbit (which is a synapomorphy cladistic analysis shows that this rostro-caudal mode of scale articulation first appeared in the clade ({Semionotiformes, Lepisosteiformes) and is absent in the most basal ginglymodians. In both lineages, the anterior ventral process is secondarily reduced: in the Lepisoteoidei among lepisosteiforms and in the macrosemiids among semionotiforms.
Jain [74] distinguished three kinds of dentition in ''semionotids'': non-tritoral, moderately tritoral, and strongly tritoral. According to Jain [74] tritoral dentition is recognized by the combination of four characters: ''Firstly, the width of the tooth relative its height, some non-tritoral teeth do have tumid crowns, but these are set on moderately long pedicles. Secondly, the shape of the crown, those of the tritoral species being typically broad and with a very bluntly conical termination when newly erupted. Thirdly, the relative thickness of the enamel, which is thick in the non-tritoral forms, thin on the tritoral teeth. Fourthly, the wear on the teeth, which is absent in non-tritoral forms, variably developed in tritoral species, perhaps due to different rates of tooth replacement and to types of diet. All four of these characters must be used in deciding on the nature of the inner dentition'' ( [74]: 30). To distinguish between moderately and strongly tritoral dentitions, Jain ([74]: Table 9) also used other morphological characters of the lower jaw and palate, which he found associated to the type of dentition: the depth of the jaw symphysis, the thickness and relative size of the tooth-bearing area of the coronoid bones, the presence of co-ossified vomers, and the relative length of the tooth-bearing areas on the vomers. The association of these features with one or the other kind of dentition is however ambiguous, and some of them are rarely preserved or visible in the fossils. Although tritoral and semitritoral dentitions may occur together with co-ossified vomers, fishes like {Lepidotes microrhis [72] or the basal teleost have co-ossified vomers, but lack tritoral dentition, and fishes like {Tlayuamichin, {Macrosemius or {Paralepidotus have semitritorial dentition and separate paired vomers. The actual shape of the coronoid bones or their toothbearing areas, as well as the relative length of the tooth-bearing areas on the vomers are only rarely observable because these bones are usually partially to mostly hidden. However, the depth of the jaw symphysis does seem to be positively correlated with the presence of extremely tritoral dentition. Although {Scheenstia zappi has strongly tritoral dentition, but moderately deep jaw symphysis, the other fishes with this kind of dentition ({Scheenstia maximus, {S. laevis, {S. mantelli and {Macrosemimimus lennieri) have very deep mandibular symphyses. Although it is interesting to explore the potential co-occurrence of the four characters proposed by Jain [74] in his Table 9 in further detail, the tooth morphology alone is sufficient to distinguish between a strongly tritoral dentition, in which there is no styliform tooth, and a moderately tritoral dentition, in which the marginal teeth are styliform and the teeth are gradually more tritoral towards the midline.
Woodward [95] and Jain [74] analysed the variation in the shape and proportion of the mouth and jaw bones in several species that at those times were classified in the genus {Lepidotes, including several taxa studied for this analysis: {Lepidotes elvensis and {L. semiserratus, {Callipurbeckia minor, {C. notopterus, {Scheenstia mantelli, and the type specimens of {''Lepidotes'' toombsi (junior synonym of {Macrosemimimus lennieri according to [70]). Woodward proposed the general tendencies through time towards more tritoral dentition, shorter jaws, smaller mouth, increasing number of extrascapulars and suborbitals and a nearly straight interfrontal suture. After measuring as many specimens as possible, Jain was unable to confirm the first three and the last of these tendencies and I find no clear tendencies regarding those features either. However, the shape of the lower jaw is highly variable among ginglymodians and since variation in the jaw lever ratios has been shown to reflect variation in jaw closing speeds, force and modes of feeding [147], this observed morphological variation is most probably related with differences in the diet (e.g. those fishes with strongly tritoral dentition have very deep coronoid processes and very massive lower jaws). In an attempt to account for such a meaningful morphological variation, biomechanical comparisons were challenged by the difficulties of establishing the proper point of insertion of the ligaments and muscles involved in the lower jaw lever model of fishes in these Mesozoic fossils. Furthermore, many fossil ginglymodians are fully articulated and actually conceal the jaw joint, so that the most anterior tooth is only rarely preserved in situ and complete, making it impossible to measure the out-lever moment arm of the jaw.
Jain [74] agreed with Woodward [95] concerning the putative increase in the number of suborbital and extrascapular bones. I already commented on the variation on the number and arrangement of suborbital bones in my data set, and rather than an increase in number, the observed tendencies relate to the pattern of suborbitals. On the other hand, the variation in the number of extrascapular bones, although taxonomically useful because the number of extrascapulars is stable for a given genus, it shows no phylogenetic signal or temporal correlation.  [8]. Thies [19] proposed the presence of inconspicuous dorsal ridge scales, without a posterior spine, as a diagnostic feature of {Lepidotes. The present analysis shows that, except for {Neosemionotus, the relationships of which are not resolved, conspicuous dorsal ridge scales occur only in semionotiforms. The presence of dorsal ridge scales with a low posterior spine has a rather patchy distribution within this clade (present in {Sangiorgioichthys, {Semiolepis, {Callipurbeckia and {Tlayuamichin), but ridge scales with high posterior spines are uniquely derived in {Semionotus.

Conclusions
The Neopterygii are the largest and most important group of fishes, and they also include the teleosts and thus some 50% of modern vertebrate diversity. However, the origin and early evolution of neopterygians is still poorly understood, as are the interrelationships of their main lineages. This situation is the consequence of several interrelated factors: the unresolved question of the out-group to Neopterygii, many problematic cases of homology among basal neopterygians, and the still very incomplete knowledge of several basal lineages, most of them extinct, which is not necessarily due to the lack of well preserved fossils, but rather the paucity of studies. The ''semionotiforms'' have certainly been one these so far poorly known fossil groups. The main goal of this and my previous taxonomic work on ''semionotiforms'' [22][23]32,34,[36][37]70] has been to change this situation providing empirical information as much detailed as possible on the anatomy of these fishes. The phylogenetic relationships obtained through the cladistic analysis presented here, as any other cladogram, only represent a hypothesis that will hopefully be improved by future research.
The question of the living sister group of teleosts is usually referred to as the ''gar-Amia-teleost'' problem (i.e., the ''Halecostomi vs. Holostei''), because the living sister group of teleosts would be Amia under the Halecostomi hypothesis, or the group (Amia, Lepisosteus) under the Holostei hypothesis. As a result of my research on ''semionotiforms'', I found that the ''Halecostome paradigm'' has been misled by the erroneous interpretation of the evolution of certain morphological characters. Under the ''Halecostome paradigm'' interpretations of primary homology and character polarity have mainly been based in comparisons with modern teleosts (e.g. [7,148,149]). With the increasing knowledge of the anatomy of Palaeozoic and early Mesozoic actinopterygians, and the incorporation of this information in cladistic analyses, we now begin to understand the right direction of change in the evolution of several morphological characters, which leads many authors to return to the Holostei hypothesis of Huxley [150]. Recent morphological studies like the ones by Hurley et al. [108] and Grande [13] are very important because they have seriously questioned the ''Halecostome paradigm''. In particular Grande [13] provided thorough and detailed anatomical information on lepisosteiforms, which is essential and cannot be ignored in future research around the ''gar-Amia-teleost'' problem. This work completes Grande's study providing anatomical information on basal ginglymodians and, thus, is a contribution to our understanding of the origin and phylogenetic relationships of basal neopterygians. My studies on Mesozoic ginglymodians led me to confirm Patterson's [7] observation that these fishes show morphological affinities with both halecomorphs and teleosts. Therefore, although the hypothesis of the Holostei is still far from being demonstrated, we are at the beginning of a new and fruitful era in palaeoichthyological research. Detailed anatomical studies of non-teleostean actinopterygians (e.g. [8,13,39,54,62,[151][152]) are going to be the foundation for the re-evaluation of our current hypotheses of homology and the compilation of large data sets, which are the only valid way to test the hypotheses of the Halecostomi vs. the Holostei. In particular, more studies on Triassic and Jurassic Neopterygians are utterly needed.