The order Characiformes (Teleostei: Otophysi) is one of the most diverse freshwater fish groups. It contains around 1400 living species in South and Central America and Africa. Their fossil record starts in the Cretaceous on both continents and also in Europe. Here, we describe and discuss the occurrences of new characiform fish teeth from Provence (Maastrichtian, S. France). Five morphological types are recognized. They belong to possibly three different taxa, and they are regarded as Characiformes indet. However, two of them have resemblances to alestin fishes and could be related to the African family Alestidae. The characiform fishes from Provence are among the oldest known in Europe, together with a freshwater characiform fish occurring in Romania, and the recently described marine fish Sorbinicharax from Italy. The biogeographical history of characiform fishes has been intensively discussed during the last three decades. The group is generally accepted to be Gondwanan and its diversification linked with the break-up of this continent, with two main scenarios depending on whether the group is archaeoor telolimnic. Some authors also propose a Pangaean origin. The recent discoveries of Sorbinicharax and of the fossils from Provence change our view on the Cretaceous characiform diversity and their early ecology, and they also enable us to re-evaluate the proposed biogeographical scenarios, reinforcing the hypothesis of the telolimny of the group. The order Characiformes (Teleostei: Otophysi) includes the well-known tetras and piranhas. It contains around 1400 living species. About 1200 are South and Central American, the rest are African. Research over the last decade has hardly improved our knowledge of their intrarelationships; but many nodes of their phylogenetic tree are still debated and the monophyly of the order Characiformes itself is contested by some authors. For example, Peng et al. (2006) suggest that the order Gymnotiformes nests within the order Characiformes on the basis of molecular data. However, most specialists now agree about the monophyly of several taxa within the order, notably the four African families, i.e. Distichodontidae and Citharinidae which are also considered to be sister-groups (Vari 1979; Fink & Fink 1981; Orti & Meyer 1997), Hepsetidae (monogeneric with Hepsetus), and Alestidae (Paugy 1986, 1990; Murray & Stewart 2002; Zanata & Vari 2005). The family Alestidae corresponds to the African members of the nonmonophyletic Characidae sensu Greenwood et al. (1966), the polyphyly of which is now widely accepted (e.g. Buckup 1998; Lucena & Menezes 1998; Toledo-Piza 2000; Zanata & Vari 2005). The family Characidae as currently defined includes only American fishes. Whether they are based on morphological (Vari 1979, 1995; Buckup 1998) or DNA studies (Orti & Meyer 1997; Calcagnotto et al. 2005), the various phylogenetic analyses agree about the sister relationship of African and South American taxa in three cases and also about the fact that, of the transatlantic groups they form, one sits at the stem of the tree of the order and the two others at its crown. (1) Distichodontidae þ Citharinidae (African) is the sister group of all the other characiform fishes (with basal South American members), except for Uj (1990) who proposed a sister relation with the family Alestidae, thus forming with Erythrinidae, Hepsetidae and Ctenoluciidae the sister group of all the other characiform fishes. (2) The families Hepsetidae (African) and Erythrinidae (South American) form a monophyletic group either alone (Uj 1990; Orti & Meyer 1997), or including other families, i.e. Ctenoluciidae (Vari 1979; Buckup 1998), Ctenoluciidae and Lebiasinidae (Calcagnotto et al. 2005). Roberts (1972) remarks that among characiforms, only hepsetids and erythrinids From: CAVIN, L., LONGBOTTOM, A. & RICHTER, M. (eds) Fishes and the Break-up of Pangaea. Geological Society, London, Special Publications, 295, 155–164. DOI: 10.1144/SP295.10 0305-8719/08/$15.00 # The Geological Society of London 2008. are ambush predators with the particular behaviour of nest building and parental care. (3) The family Alestidae (African) is the sister-group of Acestrorhynchus and forms a monophyletic group with Ctenoluciidae and Brycon þ Salminus according to Orti & Meyer (1997) and to Calcagnotto et al. (2005), whereas it is the sister group of Brycon þ Chalceus according to Lucena (1993), and of Brycon and many others including Acestrorhynchus, Hepsetidae and Erythrinidae according to Buckup (1998). Uj (1990) only proposed a sister relation with the African monophyletic group Distichodontidae þ Citharinidae. Because of the Afro-South American distribution of living characiform fishes and the age of the fossils, a Gondwanan origin of the order is usually accepted (see also discussion). Depending on the ability of early members to stand salty waters, two different palaeobiogeographical scenarios have been proposed. That living characiform fishes are strictly freshwater fishes can be interpreted as reflecting the archaeolimny of the group (the archaeoand telolimny were defined by Patterson in 1975, according to primarily or secondarily fresh water habitat). This implies a vicariant scenario for the characiform fish historical biogeography. This hypothesis was first proposed by Lundberg (1993), who stressed two main points: (1) the diversification of the main taxa in the order Characiformes might have taken place on Gondwana before its break-up; and (2) early African characiform fishes would have to have suffered great extinctions in order to explain the lack of the many sister-groups of the South American families on this continent. A vicariant interpretation could also involve a primary break-up of Gondwana with possible temporary secondary connections between the lands formed, rather than just a unique break-up event. This would allow a better explanation of the differences between the American and African diversity and also the existence of three trans-Atlantic groups, due to three vicariant events (Patterson 1984; Fink et al. 1984; Maisey 1993, 2000; Lundberg et al. 1998). However, a second interpretative scenario takes shape if the order is telolimnic. The hypothesis of the telolimny of the group is supported by the brackish or marine habitat of several basal otophysan fishes and a few extinct characiform fishes: Lusitanichthys and Salminops (Cenomanian, Portugal), which are considered to be basal characiform fishes (Gayet 1981, 1985) or basal ostariophysan fishes possibly at the stem of the Otophysi (e.g. Fink & Fink 1981; Fink et al. 1984; Patterson 1993); Chanoides (Campano-Maastrichtian and Eocene, Italy), which is a primitive otophysan fish (Patterson 1984; Taverne 2005); Santanichthys (Aptian, Brazil), which is the earliest otophysan fish known yet, and should possibly be placed at the stem of the order Characiformes (Filleul & Maisey 2004); Sorbinicharax (CampanoMaastrichtian, Italy), which is a characiform fish recently described from marine Tethyan waters (Taverne 2003); and some Eocene and Oligocene characiform fossils from France that are described from brackish environments (Cappetta et al. 1972; Gaudant 1980). Moreover, many gonorynchiform fishes and the representatives of two siluriform families inhabit coastal marine waters or brackish waters: the Plotosidae and the Ariidae to which are related some of the oldest catfish fossils. In the opinion of some authors (e.g. Chardon 1967; Gayet 1982, 1986), this reflects the telolimny of the order Characiformes (among others), and supports the hypothesis of marine dispersal between the newly separated continents of the fragmented Gondwana. Based on the same evidence, Calcagnotto et al. (2005) also call on a probable early ability of characiform fishes to withstand salty waters, to build an alternate palaeobiogeographical scenario to the vicariance with possible migration from one plate to another. They note that this scenario could appear to be relatively more parsimonious, contra Patterson (1984), Fink et al. (1984), and Filleul & Maisey (2004), among others. Here, we describe and discuss the attribution of new characiform fishes from Provence (Maastrichtian, Southern France). They are among the oldest known, notably in Europe, together with the Campano-Maastrichtian marine fish Sorbinicharax from Italy (Taverne 2003), and a freshwater Maastrichtian characiform fish reported from Romania (Grigorescu et al. 1985). For a long time, the Romanian occurrence was the only one known north of the Tethys Sea for the Cretaceous, and it was logically interpreted as resulting from the trans-Tethyan dispersal(s) that affected vertebrate continental faunas during the Late Cretaceous (e.g. Gheerbrant & Rage 2006). Geological context and age The characiform teeth were collected by one of us (X.V.) from two new localities of the Provence Basin: Verane and Les Pennes-Mirabeau (Fig. 1). The fossiliferous deposits are composed of marly horizons with carbonate rich palaeosols (nodules and root traces). They are located just above the Rognac Limestone (dated as Maastrichtian by Garcia & Vianey-Liaud 2001). The material has been obtained by screen-washing the deposits. O. OTERO ET AL. 156