The fossil record of Gonorynchiformes

The fossil record of gonorynchiform fishes provides key information on the diversity, palaeobiogeography, and phylogeny of the group. The first mention of fossil Gonorynchiformes dates back to Cuvier in the early 19th century, and there is still a need for a critical review of the earliest descriptions and of some key taxa today. Fossil gonorynchiform fishes are known from the earliest Cretaceous (Berriasian-Valanginian) to the earliest Miocene, and the clade has several extant representatives. To date, the fossil record has yielded only about 18 genera and 35 species of Gonorynchiformes. With only 46 known localities, their fossil record is relatively poor compared to that of other groups of Ostariophysi. The distribution of these localities is heterogeneous in both space and time. Debates on the phylogenetic status of Gonorynchiformes have mainly focused on the identity of the basal-most members of the clade and on its sister group. Unfortunately, very few large-scale phylogenetic studies have included the fossil representatives of the clade. Using known fossil occurrences and several phylogenetic proposals, we conducted an exploratory diversity analysis. A traditional taxic approach shows that gonorynchiform diversity rose steadily during the Early Cretaceous and reached a peak in the Aptian-Cenomanian interval. It then declined slightly towards the end of the Cretaceous and it decreased further at the dawn of the Cenozoic. This apparent low diversity level is only interrupted by relative diversity peaks in the first half of the Eocene and in the Oligocene. In the absence of fossils after the earliest Miocene, diversity estimates are conjectural for most of the Neogene. We found a close similarity of the estimates obtained with alternative phylogenetic hypotheses, meaning that the differences among these phylogenies have virtually no impact on inferred diversity patterns. Our diversity analysis points to some major gaps in the known fossil record, and it calls for the integration of most (if not all) fossil taxa in phylogenetic analyses.


Introduction
Fossil Gonorynchiformes have been known for a long time. Cuvier (in Cuvier and Brongniard 1822) first recognized the similarity between some fossil bones from the Paris Basin (Tertiary of France) and the living Gonorynchus, described by Gronovius in 1763 (non-available name after Nelson 1994). After the recognition of Gonorynchiformes as a taxonomic entity (Gosline 1960, Greenwood et al. 1966, one of the most important published works was certainly that of Rosen and Greenwood (1970), who moved Gonorynchiformes from Clupeiformes to Otophysi.
The monophyly of the order Gonorynchiformes is now well established (Fink and Fink 1981, Blum 1991a, Poyato-Ariza 1996a, Grande and Poyato-Ariza 1999, Lavoué et al. 2005, but the phylogenetic position of many fossil forms remains unclear. This is because some recently described taxa have not been included in a cladistic analysis yet, and also because a critical review of the earliest descriptions is still awaited.
Here we review all taxa that have been assigned to Gonorynchiformes, and we discuss their occurrence and validity. Because there is no consensus about the systematic position of several gonorynchiform species, we have chosen to present the type genus first, and then other genera in alphabetical order. The same choice was made for the species within each genus. Recent species are just listed and briefly commented on. This review is not exhaustive and it does not include all published accounts on fossil Gonorynchiformes. Instead, we selected some major studies dealing with the anatomy and/or phylogeny of the group. In the same way, synonymy lists are limited to contrasting opinions.
In a second part, we present the spatio-temporal distribution of fossil gonorynchiforms, we review some major phylogenetic proposals, and we provide an exploratory diversity analysis.
A single species belongs to this genus, †Aethalionopsis robustus (Traquair 1911). It comes from the famous dinosaur-bearing locality of Bernissart (Belgium). The age of the fossils is middle Barremian to earliest Aptian, as estimated from angiosperm pollens (Yans et al. 2005(Yans et al. , 2006. Gaudant (1966) erected the genus †Aethalionopsis, which he considered close to †Anaethalion (within Anaethalionidae), the taxon to which the Bernissart specimens were first assigned. †Aethalionopsis robustus is known from numerous complete specimens, reaching 10 to 40 cm in standard length.
According to Gaudant (1968), this genus is represented by at least two other species. Indeed, Bassani and Erasmo (1912) and Erasmo (1915) assigned to †Anaethalion robustus several specimens from the Aptian-Albian (Early Cretaceous) of Castellamare and Pietraroia (Italy). However, these fossils differ from the type-species by the composition of their dorsal and anal fins, and they represent another, yet unnamed, species of †Aethalionopsis (Gaudant 1968). The third species, †A. valdensis, was described by Woodward (1907) from the Berriasian-Hauterivian of Sussex, England. Taverne (1981) revised the anatomy of †Aethalionopsis robustus and he designated a lectotype. This author first emphasized its similarities with some Gonorynchiformes, such as †Dastilbe, †Tharrhias and †Parachanos. However, he did not mention the Italian and English occurrences. For Taverne (1981), †Aethalionopsis differs from all other gonorynchiforms by some plesiomorphic characters, such as the shortened mandible and the lack of fusion of the third and fourth infraorbitals.
The palaeoenvironment in Bernissart was most probably fresh water with episodic connections to the sea (Marliere and Robaszynski 1975, Martin and Bultynck 1990, Grande and Bemis 1998. Grande (1999a) suggested that †Aethalionopsis may have been an euryhaline taxon with a wide salinity tolerance. Because †Aethalionopsis was found within a mixed marine and freshwater fauna, isotopic analyses would be necessary to infer the habitat of this taxon. †APULICHTHYS Taverne, 1997 †Apulichthys gayeti Taverne, 1997 type and only species

1980
Elopiforme: F. Medizza and L. Sorbini, p. 133, fig. 1997 †Apulichthys gayeti: L. Taverne,p. 403;figs. 1 to 13,table. †Apulichthys gayeti is a small (less than 12 cm in total length) marine gonorynchiform from the late Campanian-early Maastrichtian of Porto Selvaggio, near Nardo, Apulia, southern Italy. The age of the fossil locality was determined by nannofossils, and the ichthyofauna suggests a shallow platform environment (Sorbini 1978, Medizza andSorbini 1980). Only four more or less complete specimens are currently known. The family †Apulichthyidae was erected for this monospecific genus, which is regarded as the sister taxon of all other fossil and modern gonorynchoids (Taverne 1997). †CHANOIDES Woodward, 1901 †Chanoides was named for †Clupea macropoma Agassiz, 1839/44, a fossil found in the famous Eocene site of Monte Bolca, Italy (Woodward 1901). Because of some anatomical similarities with the extant Chanos, Woodward (1901) placed both †Chanoides and Chanos in the family Albulidae. Patterson (1984a) showed that †Chanoides macropoma is neither an albulid nor a chanid, but an ostariophysan fish. A second species of †Chanoides has been described recently as †Chanoides chardoni (Taverne 2005). It comes from the late Campanian-early Maastrichtian of Nardo, Apulia, southern Italy. The species †Chanoides striata, originally described by Weiler (1920), was moved to †Neohalecopsis by Weiler (1928), whereas †Chanoides leptostea was transferred to †Coelogaster (see below). A yet older species of †Chanoides from the Santonian of southern Italy (Apricena) is currently under study by Taverne and De Cosmo. †CHANOPSIS Casier, 1961 †Chanopsis lombardi Casier, 1961 type and only species 1961 †Chanopsis lombardi: E. Casier,p. 60;figs. 17 and 18; Casier (1961) described †Chanopsis lombardi based on some isolated bones of a single individual that was about one metre long. The fossils were found in the Wealdian (perhaps Albian according to Taverne 1984) Loia strata, Democratic Republic of Congo. This species was tentatively assigned to the family Chanidae by Casier (1961), but Taverne (1984) showed that it belongs in Osteoglossidae, on the basis of the peculiar enlarged shape of the frontals, the large opercle, the small or absent subopercle, and the anterior position of the autosphenotic, among other characters. Poyato-Ariza (1996a) agreed with the conclusions of Taverne (1984).
As for all other Lebanese fossil species, the palaeoenvironment was marine (Hückel 1970, Cappetta 1980 This eel-shaped species has been reported from Santonian marine strata at Sahel Alma, Lebanon. It is the smallest known †Charitosomus, with a maximal total length of about 8 cm. As for all fish from Sahel Alma, anatomical study is rather difficult due to their state of preservation (Pictet andHumbert 1866, Woodward 1901), but this species was redescribed recently (Gayet 1993b) on the basis of numerous specimens and acid preparation. This author noticed the possible presence of a supraneural 1 in †Charitosomus lineolatus, as observed in †Hakeliosomus and †Chanoides chardoni (Gayet 1993b, Taverne 2005. This could be interesting to investigate because the absence of supraneural 1 is the rule among all other Ostariophysi (Fink and Fink 1981). †Charitosomus major Woodward, 1901Woodward, 1901 †Charitosomus major: A.S. Woodward,p. 272;table 15,fig. 3.
This large †Charitosomus species, also from the Santonian Sahel Alma locality in Lebanon, can reach a total length of 20 cm. Only five specimens are presently known and described (Gayet 1993b). Like †Charitosomus lineolatus, two body shapes are present and they could represent a case of sexual dimorphism.
Type species for the genus and the only valid species of †Dastilbe according to Davis and Martill (1999) and Brito and Amaral (2008). It has been reported from black shales in the Aptian Muribeca Formation of Riacho Doce, Sergipe-Alagoas Basin, northeastern Brazil. Santos (1990) cited †Dastilbe crandalli in the Aptian Cabo Formation (Brazil). Because he differentiated †D. crandalli from †D. elongatus by the number of pectoral fin rays, Blum (1991a) reported the former species from the Crato Formation, Araripe Basin, Brazil. Davis and Martill (1999) analysed the very abundant specimens of †Dastilbe from the Nova Olinda Member of this formation. Berthou (1990) suggested a latest Aptian to early Albian age for the sequence containing the Nova Olinda Member. Numerous specimens are known, and they range up to 7 cm in total length. Otoliths rarely occur in situ but can be abundant as isolated microfossils (Davis and Martill 1999). In contrast with Blum (1991a) and Martill (1993), who suggested that †Dastilbe was a non-marine fish, Davis and Martill (1999) speculated that †D. crandalli was an anadromous fish tolerant to hypersalinity. †Dastilbe batai Gayet, 1989Gayet, 1989 Dastilbe batai: M. Gayet,p. 22; table 1, fig. 1. This species was erected by Gayet (1989) on the basis of a single, complete specimen from the Aptian-Albian beds of Río Benito, south of Bata, Equatorial Guinea. It was distinguished from †D. crandalli by the anterior position of the pelvics, with only seven lepidotriches, and by anal fins with eight lepidotriches. Poyato-Ariza (1996a), followed by Davis and Martill (1999), considered the diagnostic characters of this species not convincing, and the later authors considered it synonymous with †D. crandalli. The position of the pelvic fins relative to the dorsal fin is a variable feature of little diagnostic value, and the fin ray counts overlap with the variation observed among the Brazilian specimens of †D. crandalli. Some of the numerous isolated bones surrounding the holotype may belong to this species, but the poor state of preservation may hamper any robust taxonomic decisions (Brito and Amaral 2008). In any case, the specimen from Bata currently represents the only occurrence of †Dastilbe in Africa (but see Dietze 2007). †Dastilbe batai was found in black shales. The same level yielded two clupeomorphs, †Ellimma goodi (referred to the marine †Ellimmichthys by Chang and Grande 1997) and †Clupavichthys dufouri (Gayet 1989). Clupavidae seem to be marine [for example, Clupavus is known in marine deposits in Portugal and Morocco (Gayet 1981, Taverne 1977]. Other fossil fish, including †Parachanos, also occur in bituminous shales in several fossil localities close to Rio Benito (Casier and Taverne 1971). They are mostly marine, and the palaeoenvironment corresponds to a calm bay temporarily lagoonal (Weiler 1922). †Dastilbe elongatus Santos, 1947Santos, 1947 †Dastilbe  (Santos 1947, Duarte andSantos 1993), and in a manuscript that remained unpublished. Santos (1947) described and figured this species on the basis of five specimens from the laminated limestone of the Aptian Crato Member, Santana Formation, Araripe Basin, northeastern Brazil. However, he did not provide any diagnostic character. These specimens seem to be lost (Davis and Martill 1999). †Dastilbe elongatus was also listed in a review of the Araripe fish fauna (Santos and Valença 1968). Blum (1991a) distinguished †D. elongatus and †D. crandalli on the basis of the count of pectoral fin rays and of total length (respectively 13 rays and TL up to 20 cm for †D. elongatus, and 10 rays and TL up to 6 cm for †D. crandalli). These characters were criticized by Poyato-Ariza (1996a) and Davis and Martill (1999). Taverne (1981) hypothesized that †Dastilbe crandalli and †D. elongatus may represent distinct populations of a single species. Davis and Martill (1999), Dietze (2007) and Brito and Amaral (2008) found no skeletal criterion for distinguishing these two species and they regard †D. elongatus as a junior synonym of †D. crandalli, an opinion that we follow here. Poyato-Ariza (1996a) considered the specimens of †Dastilbe crandalli from Riacho Doce and specimens from the Crato member as a distinct species, based on their absence of distema in the caudal skeleton. †"Dastilbe minor" Santos, 1975Santos, 1975 †Dastilbe minor: R. da Silva Santos, unpublished doctoral thesis. 1990 †Dastilbe minor Santos, 1975: R. da Silva Santos, p. 267, listed 1996a †Dastilbe minor?: F. J. Poyato-Ariza, p. 43, listed.
In his unpublished doctoral thesis, Santos (1975) proposed under the name †Dastilbe minor a new species from the Marizal Formation (Lower Cretaceous), Bahia State, Brazil. He quoted this name again many years later in another paper (Santos 1990). Unfortunately, that species has never been diagnosed or described, only listed, so that it can be regarded as nomen nudum (Brito and Amaral 2008). †Dastilbe moraesi Santos, 1955Santos, 1955 †Dastilbe moraesi: R. da Silva Santos, p. 19; tables 1 and 2. Santos (1955) reported this species from the Aptian (or late Barremianearly Aptian, see Brito and Amaral, 2008) of the Areado Formation, Presidente Olegario, State of Minas Gerais, Brazil. Some cranial characters seem to relate this species more closely to †Tharrhias than to †Dastilbe (Poyato-Ariza 1996a), but the count of vertebrae differs between these two genera (respectively 50 and 36 for the single specimen examined by this author). Davis and Martill (1999) agreed with Poyato-Ariza (1996a) by considering this taxon valid, awaiting a more complete study, whereas Brito and Amaral (2008) consider it as a junior synonym of †D. crandalli. †Dastilbe sp. Malabarba et al. (2002) reported †Dastilbe from the Aptian/Albian outcrops of the Maceió Formation, Alagoas Basin, northeastern Brazil, and Soares and Calheiros (1991) reported †Dastilbe sp. from Aptian-Albian bituminous shales at Rio Largo, Muribeca Formation, in levels younger than those in which †D. crandalli was found. According to Soares and Calheiros (1991), these bituminous shales were formed in a paralic flysch depositional environment. †ECTASIS Jordan and Gilbert, 1919 †Ectasis proriger Jordan and Gilbert, 1919 type and only species 1919 †Ectasis proriger: Jordan and Gilbert (in Jordan), p. 62; pl. 22.
A single specimen from the Pliocene of the Los Angeles clay shale, at the Third Street Tunnel, was first assigned with doubts to the family Gonorynchidae by Jordan and Gilbert (in Jordan 1919). Later on, Jordan (1921a) placed †Ectasis in Elopidae, close to Elops.

Gonorynchiformes? incertae sedis
A single articulated skeleton from the Early Cretaceous freshwater deposits of Kyushu, Japan, was reported and figured by Yabumoto (1994) as "Gonorynchiformes? incertae sedis" (probably meaning "Gonorynchiformes? indet."). According to , based on the specimen figured by Yabumoto, the form of the body is gonorynchid-like.
Gonorynchidae indet. The posterior part of an individual with 23 vertebrae and the caudal skeleton, from the late Albian to middle Cenomanian of Kipala, Kwango, Democratic Republic of Congo, was assigned to a new species of †Charitosomus, †C. hermani by Taverne (1976a). However, it seems more appropriate to attribute this caudal skeleton to an indeterminate gonorynchid fish. Even if close to fossil genera from the Middle East, it differs by the fusion of the hypural 1 and 2, both attached to the terminal centrum. In all studied specimens of †Hakeliosomus, †Charitosomus and †Charitopsis, the first hypural is free from the second, and it articulates with the terminal centrum. The whole fauna from Kipala suggests a lagoonal palaeoenvironment (Taverne 1976a). †GORDICHTHYS Poyato-Ariza, 1994 †Gordichthys conquensis Poyato-Ariza, 1994 type and only species
First mentioned as an undetermined teleost close to †Ascalabos (Leptolepis) voithi (Sanz et al. 1988), and as a teleost incertae sedis (Poyato-Ariza and ), this small fish was recognized as a chanid and described in detail by Poyato-Ariza (1994). It comes from the freshwater late Hauterivian-early Barremian beds of Las Hoyas, Cuenca Province, Spain (Poyato-Ariza 1994, Poyato-Ariza et al. 1998. After his cladistic analysis of chanids, Poyato-Ariza (1996a, b) arranged †Gordichthys and †Rubiesichthys in the new subfamily †Rubiesichthyinae. This result is based on five synapomorphies, including two unique derived characters (acute angle between the preopercular limbs in adults and presence of the posterior process of the first supraneural). The numerous specimens currently known are all less than 4 cm in standard length. Géry, 1964 Only one Recent species, Grasseichthys gabonensis Géry, 1964, is reported from the Cuvette Centrale, Democratic Republic of Congo and from the Ivindo Basin, Gabon (Roberts 1972). No fossil species is known for this genus. †HAKELIOSOMUS Gayet, 1993b  1 to 3. †Hakeliosomus hakelensis has been reported from the marine early Cenomanian strata of Haqil, Lebanon. The species has long been considered as belonging to †Charitosomus (Davis 1887, Woodward 1898, Patterson 1970. However, acid preparation of numerous specimens allowed a redescription of the species and its transfer into the monospecific genus †Hakeliosomus (Gayet 1993b). Grande and Poyato-Ariza (1999) found no justification for this taxonomic choice. Phylogenetically, †Hakeliosomus is close to †Charitosomus and †Charitopsis. However, it differs from all species of †Charitosomus and from †Charitopsis by the smooth border of its subopercle (that is spinous in these two genera). Also, a supraneural 1 seems to be present, as in some species of †Charitosomus (e.g., †C. lineolatus). For these reasons, we believe that it represents a valid genus, although Grande and Grande (2008) argued that it is a synonym of †Ramallichthys. †HALECOPSIS Woodward, 1901  This species was reported from the marine Ypresian London Clay (southeast England), from the "Argile des Flandres" (Belgium), from the Clay of Hemmoor (northwest Germany), and from northern France. Its osteology was studied by Delvaux and Ortlieb (1888), Woodward (1901), and especially Casier (1946Casier ( , 1966. Jordan (1910) suspected some relationship with the chanid †Dastilbe, but Schaeffer (1947) questioned this affinity. Casier (1966) was the first to suggest gonorynchid affinities. For Patterson (1984b), †Halecopsis and †Neohalecopsis (see below) are chanids rather than members of the family †Halecopsidae, which is not well defined. Because the specimens of these two genera are poorly preserved, they were not included into the phylogenetic analysis by Poyato-Ariza (1996a). However, because they lack two chanid synapomorphies (wide frontals and expanded operculum), that author suspected that they may be closer to Gonorynchus than to Chanos. Neither Grande (1999a) nor  mentioned this teleost. Taverne and Gayet (2006) confirmed the gonorynchoid status of †Halecopsis insignis, and they placed it between †Apulichthys and the other families of the suborder. †JUDEICHTHYS Gayet, 1985 †Judeichthys haasi Gayet, 1985 type and only species Only one specimen from the marine early Cenomanian of Ramallah, Judea Mounts, Israel, is presently known. A new family, †Judeichthidae, was erected for this taxon (Gayet 1985). Grande (1996) rejected the validity of both this family and this genus. This author included both †Judeichthys and †Ramallichthys into †Charitosomus. More recently, Grande and Grande (2008) argued that †Judeichthys should be synonymized with †Ramallichthys.

GRASSEICHTHYS
There is a controversy for the identity of the two posterior pharyngeal patches of conical teeth. For Grande (1996), these teeth belong to the entopterygoid and to the basibranchials, as in all gonorynchiforms. However, the specimen of †Judeichthys was apparently not subject to post-mortem displacement because the two mandibles are nearly superposed. It is therefore difficult to imagine how the right endopterygoid only could have been turned out in a movement placing the patch of teeth at a level as low as the second basibranchial. In addition, two teeth belonging to the upper patch are placed outside the entopterygoid (Gayet 1985). Gonorynchus and †Notogoneus have elongated skull, and Grande (1996) argued that the shortness of the anterior part of the skull of †Judeichthys is similar to the 'derived' Phractolaemus. However, there was no direct evidence of a relationship between these two taxa.
Although it would be out of place to discuss all the characters of †Judeichthys here, we propose not to maintain the family †Judeichthyidae, but we do maintain the generic status of †Judeichthys.
One near complete specimen from late Campanian-early Maastrichtian marine deposits in Nardo, Apulia, Southern Italy, was named †Lecceichthys wautyi by Taverne (1998). Close to †Notogoneus and Gonorynchus, it was placed in the Gonorynchidae, as the basal sister group of these two genera. †NEOHALECOPSIS Weiler, 1928. †Neohalecopsis striatus (Weiler, 1920) type and only species 1920 †Chanoides striata: W. Weiler, p. 4. 1928 †Neohalecopsis striatus ( This taxon is based on a single specimen from the Oligocene Septarientones, Flörsheim, Rheinhessen (Germany). The fossil was first described as †Chanoides striata by Weiler (1920), and it was later assigned to a new monospecific genus, †Neohalecopsis, supposed close to †Halecopsis. Casier (1946) erected the family †Halecopsidae for these two genera (see †Halecopsis). However, the skull and the general morphology of †Neohalecopsis greatly differ from †Halecopsis, and they seem to indicate chanid affinities instead (Taverne and Gayet 2006 Type species of the genus, †Notogoneus osculus was first reported from the early Eocene of Wyoming (Green River Formation and Bridger Formation). †Notogoneus osculus was apparently not common in Eocene lake faunas, and it was not a near-shore species (Grande 1999a). Hundreds of nearly complete specimens are known. This species has been described by several authors, including Cope (1885), Whitfield (1890), Perkins (1970) and Grande (1984). The known representatives of †Notogoneus osculus are between 20 and 80 cm long. Isolated scale fragments from the early Eocene freshwater deposits of the Coalmon Formation, northern Colorado, have been assigned to †Notogoneus sp. cf. N. osculus (Wilson 1981). †Notogoneus alsheimensis (Weiler, 1942) fig. 20.
Otoliths from the late Oligocene of Niederrhein, Germany, have been reported as a species of †Notogoneus by Schwarzhans (1974). Nolf (1985) first agreed with this determination, but further research indicates that these otoliths now appear to belong to a perciform fish from the family Acropomatidae (D. Nolf, personal communication to L. Taverne, 2004 Gaudant (1981a), these layers were deposited in the slightly brackish waters of a lagoon that had some connection to the sea. Only two imperfect articulated skeletons and some isolated bones are known. Gaudant and Burkhardt (1984) also reported †Notogoneus cf. cuvieri from the "marnes grises" (grey marls) at Altkirch (Haut-Rhin, France), a site dated as early Oligocene (Sittler 1965(Sittler , 1972. These †Notogoneus specimens represent about two thirds of the ichthyofauna at this locality. The skeletons are more or less fragmentary, but the bones are usually found in connection. The fish fauna, the invertebrates, the plant remains, as well as some isotopic analyses suggest a low salinity for the palaeowater at Altkirch (Gaudant and Burkhardt 1984). †Notogoneus fusiformis Schwarzhans, 1994Schwarzhans, 1994 †Notogoneus fusiformis: W. Schwarzhans, p. 57; fig. 40. Schwarzhans (1994) described this species from the late Oligocene (Chattian) of Niederrhein (Germany), based on isolated otoliths. The generic determination seems correct (D. Nolf, personal communication to L. Taverne, 2004). Interestingly, Schwarzhans (1994) described the palaeoenvironment as "a shallow marine setting". †Notogoneus gracilis, Sytchevskaya, 1986Sytchevskaya, 1986 †Notogoneus gracilis: E.K. Sytchevskaya, p. 51; figs. 15 and 16; tables 8 and 9. Sytchevskaya (1986) described a new species of a small †Notogoneus from the late Paleocene or early Eocene Boltyshka Basin, Ukraine. This taxon is represented by numerous partial or complete specimens of small size, between 5 and 9 cm . †Notogoneus gracilis is the oldest record of the genus in Europe. †Notogoneus janeti Priem, 1908Priem, 1908 †Notogoneus janeti: F. Priem, p. 133; pl. 3, figs. 2 and 3.
This species was erected for a single specimen from the late Eocene or early Oligocene "Marnes bleues supragypseuses" of the Paris Basin, France. It was described and figured by Signeux (1961). †Notogoneus janeti was regarded as a junior synonym of †Notogoneus cuvieri by Gaudant (1981a Numerous complete specimens from the early Miocene (Aquitanian) freshwater "Hydrobienschichten" deposits in Germany were assigned to †Notogoneus longiceps by Weiler (1963). This species is also present at Godrastein-bei-Landau (Palatinat, Germany), in the lower part of the Hydrobienschichten layers (Brelie et al. 1973). According to Gaudant (1981b), isolated otoliths from the same areas were also attributed to this species by these authors and by Malz (1978a, b).
Despite the presence of the amphibians Rana, †Palaeobatrachus and Salamandra, as well as of freshwater gastropods (Planorbidae and Lymnaeidae), Gaudant (1981b) argued for oligohaline conditions during the formation of the Hydrobienschichten deposits, because strictly marine fishes are absent. †Notogoneus montanensis  †Notogoneus montanensis: L. Grande and T. Grande, p. 614; figs. 1 to 4; table 1.  described this new species of †Notogoneus from the Campanian Two Medecine Formation, northwestern Montana (USA). It is based primarily on a single skeleton that is less than 5 cm in total length and that is missing much of the skull. It represents the earliest ascertained occurrence of †Notogoneus in North America, and the earliest known freshwater gonorynchiform from this continent. The palaeoenvironment is described as a small lacustrine or waterhole environment by Varrichio and Horner (1993). †Notogoneus parvus Hills, 1934Hills, 1934 †Notogoneus parvus: E.S. Hills,p. 164;figs. 8 and 9;pl. 20.
This species, about 20 cm in total length, is known from late Eocene or early Oligocene freshwater deposits of Redbank Plains, southern Queensland, Australia. It was first described by Hills (1934) on the basis on five nearly complete specimens preserved as moulds in a limonitic mudstone. However, the absence of denticles on the scales suggests that the Australian specimens may belong to another genus. †Cyprinus squammosseus Blainville . †Notogoneus squammosseus, from the late Oligocene gypsum quarry of Aixen-Provence (southern France), is known from several complete specimens up to 77 cm in total length. They were re-described by Gaudant (1981c), who suggested a brackish palaeoenvironment for the deposits, based on the fish assemblage. †Notogoneus sp.
Numerous specimens, lacking specific diagnostic characters, have been assigned to †Notogoneus sp. These specimens come from the late Paleocene freshwater deposits of Alberta (Paskapoo Formation), Canada (Wilson 1980); from the middle Eocene (Lutetian) of Brasles, Aisne Departement, France (subopercle and scales, Gaudant 1981b); from the early Oligocene potassic basin of Alsace, France (caudal skeleton associated with the cyprinodontid †Prolebias, Gaudant 1981d); from the early Oligocene of the Isle of Wight, England (Gaudant 1981b); and from the early Oligocene of Hoeleden, Brabant, in Belgium (Nolf 1977, Gaudant 1981b A new, undescribed genus, based on two specimens, was reported by Applegate (1996) from the middle-late Albian Tlayúa quarry, near Tepexi de Rodríguez, Puebla, Mexico. There is currently no information about the morphology of the fossil specimens, and only size and body shape were used to tentatively assign it to †Notogoneus. The co-occurrence of coral reef, freshwater and open ocean taxa suggests that the paleoenvironment was a lagoon protected behind a reef barrier (Espinosa-Arrubarrena and Applegate 1996). If the specimens really belong to †Notogoneus, they would represent the earliest occurrence of the genus.
? † Notogoneus 1999 Gonorynchidae: L. Grande and T. Grande, p. 620. 1999a ? †Notogoneus: T. Grande, p. 437. Some isolated bones from late Paleocene to middle Eocene freshwater oilshale deposits in Longkou county (Shandong Province, China) have been attributed to a gonorynchiform by . The material, still to be studied in detail, probably represents a species of †Notogoneus (Grande 1999a). Such an occurrence is not unexpected, given the wide geographic and temporal distribution of †Notogoneus. †PARACHANOS Arambourg and Schneegans, 1935a Weiler (1922) to erect a new species: †Leptosomus aethiopicus. Arambourg and Schneegans (1935a, b) later removed this species from the genus †Leptosomus because they regarded the Rio Benito specimens as juveniles of a new genus they just discovered in Gabon: †Parachanos. This taxon was found in the sub-littoral sandstones of the Cocobeach series, Gabon. Like Chanos, †Parachanos was placed among the clupeiforms at that time (Arambourg and Schneegans 1935a). Later recognized as a chanid, †Parachanos then became the focus of another debate: its similarity with †Dastilbe.
The spatial distribution of localities is also unbalanced (see also Grande 1999a andFara et al. 2007 for summary maps). More than half the sites are located in Europe, and this is accompanied by a higher taxonomic diversity, because the majority of fossil genera and species are European (Figs. 6.1 and 6.2). The oldest representative of the group ( †Rubiesichthys) is known from Europe, as well as the most basal taxa. The Italian taxon †Chanos leopoldi is currently the oldest known record of Chanos, which is the only extant gonorynchiform genus with fossil representatives. All fossil Chanos species are restricted to Europe (although the extant Chanos is Indo-Pacific), and they range apparently from the Early Cretaceous to the Oligocene.
Only three fossil genera and three species have been reported from Africa ( Fig. 6.3), but their precise taxonomic status remains unresolved ( †Parachanos, †Dastilbe, and †Charitosomus). Therefore, gonorynchiform fish are undoubtedly present in Africa since Aptian-Albian times, but the exact nature of the taxa is open to debate. Nothing is known about the origin of the two living African families, Kneriidae and Phractolaemidae, which lack a fossil record. Occurrences in the Middle East (Fig. 6.3) suggest that gonorynchiform diversity was apparently higher than in Africa, and they provide important information on the Tethyan biogeography in the early Late Cretaceous.  The most remarkable feature of the gonorynchiform fossil record in North and Central America is that it documents only a few species of a single genus, †Notogoneus (Fig. 6.4), unless the yet undescribed Notogoneuslike form from Mexico represents another genus. Given the size of the continent, the well-known geology, the intensive sampling and the numerous fossil fish localities, this very low gonorynchiform diversity is puzzling, but it certainly represents a real biological pattern. Asia and Australia have very poor but interesting gonorynchiform fossil records (Fig. 6.4). For example, the oldest known occurrence of the group may be a gonorynchid-like specimen from the earliest Cretaceous (Berriasian-Valanginian) of Japan. Also, there is currently no palaeobiogeographic model to explain the presence of an Eocene/Oligocene freshwater †Notogoneus in Australia. In turn, there is no doubt that more Asian and Australian fossils would provide key phylogenetic and palaeobiogeographic information for the group.
The fossil record of Gonorynchiformes in South America (Fig. 6.5) is limited taxonomically (only two genera, †Dastilbe and †Tharrhias), spatially  (most records are from northeastern Brazil), and temporally (Aptian-Albian). The distribution of †Dastilbe seems to be linked with regional events during the opening of the South Atlantic Ocean. However, the precise palaeobiogeographical history of †Dastilbe awaits a better understanding of the relationships among its various species (Maisey 2000). The synoptic range-chart in Fig. 6.6 summarizes the temporal distribution of all fossil genera considered valid in this work.

Phylogenetic Relationships
In recent years, debates on the phylogenetic status of gonorynchiforms have mainly focused on the identity of the basal-most members of the clade and on its sister group (e.g. Poyato-Ariza 1999, Lavoué et al. 2005). We shall briefly review the main phylogenetic hypotheses in order to test whether conflicting proposals affect diversity patterns inferred from the fossil record. Very few large-scale phylogenetic studies have included the fossil representatives of Gonorynchiformes. The hypothesis by Gayet (1993c) needed to be updated after the discovery of new taxa such as †Gordichthys   (Poyato-Ariza 1994), †Apulichthys and †Lecceichthys (Taverne 1997(Taverne , 1998. Grande and Poyato-Ariza (1999) provided the most comprehensive phylogenetic analysis of fossil and extant Gonorynchiformes, although †Apulichthys and †Lecceichthys could not be included at that time. We summarize below these three proposals as indented lists, and we present the corresponding phylogenetic trees (phylogenetic hypotheses projected on a time scale) in Figs. 6.7-6.9. Poyato-Ariza (1996a) first provided a cladistic analysis of extant and fossil Gonorynchiformes, but it is not included here because it focused primarily on Chanidae and was later updated in Grande and Poyato-Ariza (1999). Gayet (1993c) first proposed a sister-group relationship between Gonorynchus and †Notogoneus, and she was followed by subsequent workers Superorder Ostariophysi sensu Rosen and Greenwood 1970 Series Anatophysi sensu Rosen and Greenwood 1970 Order Gonorynchiformes sensu Rosen and Greenwood 1970 Gayet 1993c (Fig. 7 Gayet (1993b) named the Gonorynchoidei in her taxonomic scheme, with the exception of †Hakeliosomus, which was not considered a valid taxon by Grande and Poyato-Ariza (1999). The family names †Judeichthyidae and †Charitosomidae have no more reason to exist, as observed by . Another similarity between the hypotheses of Gayet (1993c) and Grande and Poyato-Ariza (1999) is the sister-group relationship between Gonorynchus, †Notogoneus and the Middle East taxa on the one hand and the extant Kneriinae and Phractolaeminae on the other hand. The monophyly of the clade including all these extant African freshwater taxa is well supported by osteological, myological, and molecular data (Howes 1985, Gayet 1993c, Fink and Fink 1996, Grande and Poyato-Ariza 1999, Lavoué et al. 2005. Within this clade, however, there is some disagreement between the results of osteological studies (Grande 1996, Grande andPoyato-Ariza 1999) and those of a recent study based on whole mitogenome sequences (Lavoué et al. 2005), in which Grasseichthys and Cromeria are regarded as sequenced sister groups to the clade (Kneria+Parakneria). Such a conflict has a great heuristic value in the systematics of Gonorynchiformes. Taverne's (1997Taverne's ( , 1998 definition of Gonorynchoidei was similar to that of Gayet (1993c) and Grande and Poyato-Ariza (1999), but he added †Lecceichthys, †Apulichthys and †Halecopsis to that group. His definition of Chanoidei made this clade paraphyletic. The identity of the basal-most members of Gonorynchiformes is a recurrent debate. Poyato-Ariza (1996a) and Taverne (1997Taverne ( , 1998 placed †Aethalionopsis among the Chaninae, but the analysis of Grande and Poyato-Ariza (1999) excluded this taxon from the Chanidae, a result in accordance with Gayet's (1993c) early hypothesis. The relationships of †Tharrhias, †Dastilbe, and †Parachanos are similar in the three phylogenetic schemes reviewed here. Although Gayet (1993c) placed Chanos outside the " †Tharrhias group", this genus certainly belongs to the clade composed of †Dastilbe, †Parachanos, †Prochanos, and †Tharrhias (Taverne 1997, 1998, Grande and Poyato-Ariza 1999. A vast majority of osteological studies (Fink and Fink 1981, Gayet 1993c, Poyato-Ariza 1996a, Taverne 1997, 1998, Grande and Poyato-Ariza 1999 agree on the relationships of extant forms, with the structure {Chanos[Gonorynchus(Phractolaemidae+Kneriidae)]}. However, this configuration has been challenged recently by the mitogenomic study of Lavoué et al. (2005), in which the arrangement {Gonorynchus[Chanos(Phractolaemidae+Kneriidae)]} was found. Fig. 6.7 Phylogenetic tree obtained according to the phylogenetic hypothesis by Gayet (1993c).  Taverne (1997Taverne ( , 1998 Interestingly, such a structure is in accordance with the early osteologybased study by Greenwood et al. (1966).

Diversity Dynamics
From the problems outlined above, it is clear that both the fossil record of Gonorynchiformes and our understanding of their phylogenetic relationships are not yet ripe for carrying out a robust diversity study. The following attempt should be regarded as an exploratory analysis and a base for future work.
In this review, we count 18 genera and 35 valid species of fossil gonorynchiforms between the Early Cretaceous and the Miocene. Diversity dynamics was estimated by different techniques at the species and generic levels. First, we assessed diversity by simply counting observed ranges within each stratigraphic stage. This traditional "taxic approach" (Levinton 1988) yields a raw diversity estimate. The latter was corrected for the "Lazarus effect" (Jablonski 1986, Fara 2001, that is, a taxon is counted as present between its first and last occurrence even if no fossil record is actually known in the intervening stages. Taxa whose estimated ages cross a boundary between two time intervals were counted as present in both intervals. Second, we used the "phylogenetic approach" (Smith 1994). This method extends the observed stratigraphic ranges of taxa with the "ghost lineages" in order to confirm the predictions of a phylogenetic hypothesis (Smith 1988, Norell 1993. It supposes that sister taxa originate at the same time. We have applied this approach to the phylogenetic hypotheses by Gayet (1993c), Taverne (1997Taverne ( , 1998, and Grande and Poyato-Ariza (1999) to see how these different proposals affect inferred diversity patterns. For comparison purposes, only taxa common to these three phylogenetic schemes were retained in the analysis. The lack of resolution in some parts of Grande and Poyato-Ariza's (1999) cladogram did not affect the calculations because it concerns a series of either Aptian or mostly Cenomanian taxa.
Third, we computed minimal species diversity of Gonorynchiformes with the method described by Fara (2004). This approach is an intermediate between the taxic and the phylogenetic approaches. Figure 6.10 shows the raw species and genus diversity estimated with the taxic approach. These two estimates are very similar because most genera have only one species represented in each stratigraphic stage. Gonorynchiform diversity rose steadily during the Early Cretaceous and reached a peak in the Aptian-Cenomanian interval. The high diversity level in the Aptian occurred both in Gondwana ( †Dastilbe, †Parachanos, †Tharrhias) and in southern Europe ( †Rubiesichthys, †Gordichthys, †Aethalionopsis, Chanos), whereas the Cenomanian diversity maximum is dominated by Tethyan forms ( †Ramallichthys, †Judeichthys, †Hakeliosomus, †Charitopsis, †Charitosomus). Gonorynchiform diversity then declined slightly towards the end of the Cretaceous and dropped at the beginning of the Cenozoic. This apparent low diversity level is only interrupted by relative peaks in the first half of the Eocene (Chanos, †Notogoneus, †Halecopsis, †Coelogaster) and, to a lesser extent, in the Oligocene (Chanos, †Notogoneus, †Neohalecopsis). This uneven Cenozoic diversity pattern occurred in Europe, North America, and Asia. After the last record of †Notogoneus in the Aquitanian (earliest Miocene) gonorynchiform diversity dropped to a minimum for the rest of the Neogene (Fig. 6.10).
This literal reading of the fossil record cannot be taken at face value, however. It is always possible that gonorynchiform diversity stabilized or decreased in a more or less regular pattern since the Cenomanian, and that new taxa wait to be discovered in Palaeocene, late Eocene, and Mio-Pleistocene deposits. Figure 6.11 shows the diversity estimates computed with the phylogenetic approach based on the three phylogenetic hypotheses mentioned above (curves), together with the raw diversity of the genera common to these phylogenetic proposals (histogram). The diversity level is of course higher in the curves derived from the phylogenetic approach because of the addition of range extensions. The most remarkable feature is the similarity of the three estimates ( Fig. 6.11), meaning that the differences across the studied phylogenies have virtually no impact on inferred diversity patterns. The same is true for the estimates of minimal species diversity (not shown here). Figure 6.11 also shows a limit of the phylogenetic approach in the case of the Gonorynchiformes. The post-Cenomanian diversity level is remarkably low because several taxa were not included in most phylogenetic analyses (e.g., †Coelogaster, †Apulichthys, †Halecopsis, †Chanoides). Updated and comprehensive phylogenetic analyses will certainly help to resolve this issue (see Poyato-Ariza et al., this volume).
This brief overview of diversity is limited by the low number of fossil taxa currently known, and it calls for caution when analysing fossil diversity patterns for Gonorynchiformes. These fossil diversity patterns strongly contrast with the relatively high diversity observed among modern gonorynchiform fishes. Unless most of these extant forms have originated in a sub-Recent radiation, this pattern is suggestive of a bias in the fossil record, especially for the Neogene (see also Fara et al. 2007).

Historical Biogeography
The historical biogeography and associated palaeoenvironments of Gonorynchiformes have been recently studied in detail by Grande (1999a), Fig. 6.10 Diversity dynamics of Gonorynchiformes during geological times as assessed by the taxic approach. The histogram represents the total raw diversity at the genus level (gray bars) and species level (dotted bars). The minimal species diversity patterns (not shown here) are virtually identical. Abbreviations of international stratigraphic stages (x-axis) are from Harland et al. (1990), except T-C-S, Turonian-Coniacian-Santonian; La-Ser, Langhian-Serravallian; To-Mss, Tortonian-Messinian; PLI, Pliocene; QUA, Quaternary.

Fig. 6.11
Phylogenetic hypotheses and diversity estimates. The histogram represents the raw diversity of the genera common to the phylogenetic studies of Gayet (1993c), Taverne (1997Taverne ( , 1998, and Grande and Poyato-Ariza (1999). The curves correspond to the diversity estimated from these three phylogenetic proposals using the phylogenetic approach. Note the similarity of these estimates, regardless of the selected phylogenetic scheme. Abbreviations as in Fig. 6.10. and we shall not duplicate this work here. She used cladistic vicariance biogeography on available fossil data based on the phylogenetic hypothesis by Grande and Poyato-Ariza (1999). She found that the historical biogeography of Gonorynchiformes is complex and probably results from various episodes of vicariance and dispersal (Grande 1999a).

Conclusions
Although they are known since Cuvier, gonorynchiform fishes have still many secrets to unveil. Their fossil record is crucial to this endeavour, and it certainly provides key information that complements recent advances in molecular and developmental biology. However, our current knowledge of Gonorynchiformes makes it difficult to quantify precisely several aspects of their evolutionary history. Whether for biogeography or diversity analysis, the known fossil record and the understanding of their phylogenetic relationships are still insufficient to draw robust inferences and models. In particular, our diversity analysis calls for a special effort in sampling several parts of the fossil record, and it pleads for the integration of most (if not all) fossil gonorynchiform taxa in the phylogenetic analyses of the order.