General accounts. The quadrate and vertebrae of specimen MML 1234 were found associated – although not articulated – in the same bed and within an area of approximately two square meters. The location of the elements recovered suggest that they were preserved articulated and that dislocation occurred as a consequence of recent weathering.

Quadrate (Fig. 2). The right quadrate is almost complete except for a portion of the anterolateral tympanic rim. Its maximum height is 56 mm. The suprastapedial process is long and contacts ventrally with a low and stout infrastapedial process. The oval and elongated tympanic meatus is completely enclosed by the suprastapedial process. The suprastapedial process is parallel-sided for most of its length and flares slightly in a ventral direction, ending in two distinctive lateral and medial blunt and swollen crests as in H. arambourgi (Bardet et al., 2005a). In lateral view, the lateral crest is deflected anteriorly, and together with the medial crest, encloses a triangular depression. The suprastapedial process is ventrally confluent, but not fused, with a short and stout infrastapedial process. The stapedial pit for the reception of the internal process of the extrastapes is relatively small and oval, and its margins are poorly defined. There is a well-defined ventromedial articulation area for the pterygoid ramus (Fig. 2A, D, F). The median ridge is single, sharp and raising vertically. As it was described by Polcyn et al. (2012) in H. arambourgi, the median ridge turns posterodorsally at about the level of the stapedial pit. The median ridge turns sharply in a dorsal direction, separating the median surface of the quadrate from the swollen cephalic head. The median ridge and the anterior edge of the tympanic meatus enclose an elongated depression ending at about the level of the dorsal edge of the meatus. The stapedial pit is located on the dorsal portion of this depression (Fig. 2A). The posteroventral edge, as described by Polcyn et al. (2012) in H. arambourgi, is sculpted by a series of vertical grooves interpreted as the osteological correlates of a strong ligamentous attachment with the posterior rim of the mandibular glenoid. The tympanic ala is almost complete except for a small anterodorsal portion. The dorsal portion of its rim is uniformly curved to define a large tympanic area in such a way that the dorsal articular surface of the quadrate is not exposed in lateral view. A tympanic disc is almost complete except for its anterodorsal portion, which is broken. Its rugose texture and color make it clearly distinguishable from the bony tympanic rim. The most conspicuous feature of this disc is the presence of a perforation in its center and other minor pits and shallow furrows on its external surfaces, probably resulting from the activity of bioerosive agents. The ventral quadrate condyle is convex in all other aspects and, as in H. arambourgi (Bardet et al., 2005a) it is perpendicular to the proximal head.

Vertebrae. Twelve vertebrae were preserved, five cervicals and seven dorsals. However, the neural arches were not preserved on any of them. Although not articulated, they were found in close proximity. Four vertebrae bear hypapophyses. A fifth vertebra lacking most of its centrum could correspond to a posterior cervical, or an anterior dorsal, based on the location and morphology of the synapophyses. In three of the hypapophysis-bearing centra, the peduncles are stout and located posteriorly on the ventral surface of the centra (Fig. 3A–C), while in the remaining it is notably less developed suggesting that this vertebra corresponds to one of the last of the cervical series (Fig. 3E–H). As this is the most completely preserved centrum and the only one in which the total length could be measured, description of the cervical vertebrae is mainly based upon it. This centrum measures 59.02 mm long. The two articulating surfaces are dorsoventrally compressed. The cotyle is roughly rectangular, with a height/width ratio of 0.54. The synapophyses are large, stout, located close to the cotyle and posteroventrally projected. In dorsal view, the posterior projections of the synapophyses produce an angle of approximately 92° between the main axis of both synapophyses. Their anterior surfaces bear an anterodorsal and a posteroventral blunt crest forming a deep embayment. The anterodorsal crest curves slightly upward, continuing as a blunt ridge to merge with the robust bases of the prezygapophyses (Fig. 3H). In posterior view, the distal edges of the synapophyses are well below the ventral rim of the centrum (Fig. 3E). The condyle is roughly rectangular in outline (width = 31.07; height = 19.41 mm), and slightly obliquely orientated. This articulating surface is separated from the main body of the centrum by a slight constriction. In lateral view, the central articulations are obliquely inclined relative to the vertical vertebral axis (Fig. 3H).

Seven vertebrae from the dorsal series are preserved. Five of them, identified as anterior dorsals, bear robust synapophyses located on the anterior part of the centrum (Fig. 4A–F). In posterior view their ventral edges do not reach below the ventral surface of the centra. Their articulating surfaces are, as in the cervicals, dorsoventrally compressed with an average height/width ratio of 0.67. Another vertebra lacks synapophyses so its relative position in the dorsal series is difficult to determine. However, based on the centrum length, the location of the bases of the missing synapophyses, and the condyle compression, it could also correspond to the anterior dorsal series. One dorsal vertebra has dorsoventrally compressed transverse processes, located at half of the centrum height; the distal facets for the rib articulations are roughly elliptical in outline. The articulating surfaces are deep with a height/width ratio of 0.90 (Fig. 4G and H). This vertebra probably corresponds to the posterior dorsal series.

The quadrate of specimen MML 1234 is characterized by the following combination of character states: extremely elongate, ventrally flaring suprastapedial process that meets a low, broad and shelf-like infrastapedial process below; small and weakly emarginated stapedial pit; well-defined ventromedial articulation for the pterygoid ramus; posterior enclosure of an elongated tympanic meatus formed by the elongated suprastapedial process. This unique combination of quadrate features (Bardet et al., 2005a, Polcyn and Everhart, 2008 and Polcyn et al., 2012) allows referring specimen MML 1234 to Halisaurinae.

Within halisaurines, the MML 1234 quadrate differs from the quadrate of E. sternbergii by the general shape of the tympanic rim. In E. sternbergii the quadrate is rounded, with a regularly convex tympanic rim (cf. Bardet and Pereda Suberbiola, 2001, Fig. 2; Bardet et al., 2005a, Fig. 5E; Wiman, 1920, Fig. 5) whereas in MML 1234 the tympanic rim is roughly triangular. In E. sternbergii the infrastapedial process is strongly projected medially and relatively low in posterior view (cf. Bardet and Pereda Suberbiola, 2001 and Wiman, 1920, Fig. 5) when compared to that of MML 1234, H. platyspondylus (cf. Holmes and Sues, 2000, Fig. 4), H. arambourgi (cf. Polcyn et al., 2012, Fig. 2) and P. ortliebi (Lingham-Soliar, 1996, Fig. 5; MSF personal observations). Specimen MML 1234 differs from P. ortliebi by the shape and extension of the tympanic meatus and the shape of the suprastapedial process. In P. ortliebi the tympanic meatus is extremely long and narrow, with its anterior and posterior edges almost straight and parallel (MSF personal observations) while in MML 1243 the anterior edge of the tympanic meatus is slightly convex and the meatus outline is roughly oval (Fig. 5). The suprastapedial process of MML 1243 is distally less swollen than in P. ortliebi. The medial crest on the ventral ending of the suprastapedial process is anteromedially projected in P. ortliebi in such a way that it remains exposed in a medial view of the quadrate. Contrarily, in MML 1234 the medial projection of the crest is less pronounced and almost hidden in medial view.

In posterior view, the dorsal portion of the suprastapedial process of MML 1234 resembles that of H. platyspondylus (cf. Holmes and Sues, 2000, Fig. 5) and differs by its robustness from those of H. arambourgi (cf. Bardet et al., 2005a, fig. 8H; Polcyn et al., 2012, Fig. 2E), E. sternbergii (cf. Bardet and Pereda Suberbiola, 2001, Fig. 2; Wiman, 1920, Fig. 5) and P. ortliebi (MF pers. observ). Thus, in the MML 1234 quadrate the width of the dorsal portion of this process is more than half the width of its ventral ending, whereas in H. arambourgi, P. ortliebi and E. sternbergii the dorsal portion of the suprastapedial process is mediolaterally compressed (Fig. 5). The quadrate morphology of MML 1234 also resembles that of H. platyspondylus (Fig. 5B) in the outline of the tympanic rim. Thus, in both forms the anterodorsal tympanic rim is more tightly curved than the rest of the rim.

The cervical and anterior dorsal vertebral morphologies, characterized by dorsoventrally compressed vertebral centra, subrectangular outline of the articulating surfaces, and synapophyses flaring distally and projecting below the ventral rim of centra, also confirm the halisaurine affinities of specimen MML 1243. When relative values of the condyle compression of MML 1234 are plotted with those of other mosasauroids also characterized by condyle compression, they fall within the range of other halisaurines reported in literature (Bardet et al., 2005a and Caldwell and Bell, 1995), and also close to the isolated cervical vertebra of Halisaurus sp. from Perú (Fig. 6). The remaining cervical vertebrae lack most of their centra and condyle compression could only be estimated by means of the height/width ratios. The values of this ratio range from 0.5 to 0.63.

The discovery of a halisaurine in northern Patagonia is significant as it documents the widespread distribution of halisaurines during the Late Maastrichtian, and also the southernmost certain occurrence of the group. As a consequence of Atlantic transgressions during the Maastrichtian, northern Patagonia was flooded and transformed into an archipelago in a sea named the Kawas Sea by Casamiquela (1973). Paleobiogeographic analysis of the Maastrichtian marine faunas from this area, based mainly on mollusk and decapod crustaceans, revealed affinities with southern high latitude faunas (Weddellian Province, Zinsmeister, 1982). However, near the K/Pg boundary the Danian, Weddellian forms were replaced by others typical from low latitudes such as northern Brazil, the Caribbean and northern Africa (Aguirre-Urreta et al., 2008). Although fragmentary, Maastrichtian mosasaur evidences recovered up to now show no evidences of the endemism, at least at a genus level, that characterized the Weddellian Province. Thus, Late Maastrichtian rocks deposited in the Kawas Sea yielded the same mosasaur taxa known to occur worldwide such as Plioplatecarpus, Mosasaurus and Prognathodon (Fernández et al., 2008); specimen MMLP 1234 confirms the presence of Halisaurus. Up to now, no evidence of durophagous forms such as Globidens and Carinodens, which seem to have been widespread during the Latest Cretaceous (Bardet et al., 2005b and Polcyn et al., 2010), has been found. Further exploration of the Jagüel Formation, and more complete materials are needed to test whether Late Maastrichtian mosasaur faunas are cosmopolitan or if endemism are represented at species levels.

The presence of a tympanic disc seems to be a widespread feature among mosasauroids as it has been recorded in phylogenetically distant taxa (Fig. 1). The tympanic disc has been mentioned and described as a calcified tympanic membrane (e.g., Lingham-Soliar, 1994, p. 200; Russell, 1967, p. 58; Vaughn and Dawson, 1956). However, other authors identified the tympanic disk as an expanded portion of the calcified extracollumella (Caldwell et al., 2007, McDowell, 1967, Polcyn, 2010 and Polcyn, 2011). Polcyn (2011) interpreted the tympanic disc as a calcified extracolumellar cartilage forming a relatively stiff and functional pseudo-tympanum suspended by the true tympanum. The tympanic disc of specimen MML 1234, as others described in the literature, is restricted and fills the tympanic conch. Of the two main mechanisms for sound transmission, i.e. tympanic middle ear pathway and bone conduction, the morphology of the bony middle ear elements of mosasauroids is consistent with the typical tympanic middle ear configuration (Clack and Allin, 2004 and Nummela and Thewissen, 2008). The non-mammalian tympanic middle ear, which arose independently several times during tetrapod evolution (Clack and Allin, 2004 and Christensen-Dalsgaard and Carr, 2008), is a highly adaptable system (Hetherington, 2008) and consists of a tympanic membrane, supported by bones and ligaments, and an extracolumella and stapes that cross an air filled cavity towards the inner ear.

The general pattern of the middle ear of mosasauroids resembles that of most lizards but there are some striking differences particularly in the location and pattern of the sound-receiving elements. In lizards, except in the forms in which the tympanic membrane was secondarily lost or reduced (e.g., Chamaeleonids and the earless monitor lizard Lanthanotus, Clack and Allin, 2004 and McDowell, 1967), key elements of the sound conducting apparatus include the inner bony stapes and the outer slim cartilaginous extracolumella attached to the tympanic membrane. The tympanic membrane is partially enclosed by the quadrate. It is attached by means of ligaments to the posterior margin of the quadrate anteriorly, and to the retroarticular process of the lower jaw ventrally (Rieppel, 2002 and Rieppel and Zaher, 2000) (Fig. 7C). Contrarily, in mosasauroids and as a result of the bony enclosure the middle ear is conspicuously more massive. Although in some lizards, such as Varanus or Tupinambis (Fig. 7A and B), the quadrate suprastapedial process projects posteriorly above the stapes, this ventral projection is not as pronounced as in mosasauroids. It should be noted that even the mosasauroids with a short suprastapedial process (e.g., tylosaurines, Fig. 7D), it is still relatively long compared to that of Varanus or even Tupinambis. A phylogenetic tree including the relative development of the suprastapedial process reveals that the most extreme bony closure of the tympanic conch – whether by fusion or confluence of the supra- and infrastapedial processes evolved independently several times during the evolutionary history of mosasauroids (Fig. 1). This extreme enclosure suggests a general trend towards an acoustic isolation of tympanic cavities. Another difference between the middle ear of most lizards and mosasauroids is the lateral expansion of the extracolumella to form a tympanic disc in the latter. A similar tympanic disc has been described in the middle ear of Lanthanotus borneensis, which lacks the tympanic membrane (McDowell, 1967).

On the other hand, the middle ear of mosasauroids resembles that of turtles. As reported by McDowell (1967), Russell (1967), and Hetherington (2008), in both lineages the middle ears are enclosed by bone and separated from the mandible (Fig. 7E and F). Recently, Christensen-Dalsgaard and Carr (2008) and Christensen-Dalsgaard et al. (2012) demonstrated that the middle ear of the aquatic turtle Trachemys scripta is most sensitive to sound underwater and that their sensitivity depends on the large middle ear, which has a cartilaginous tympanic disc, and not on the tympanic membrane as the key sound receiver. Willis et al. (2013) increased taxon sampling to non-aquatic turtles and demonstrated that the pattern of the middle ear is similar in extant turtles regardless their ecological niche or phylogenetic position. They proposed that enlarged and isolated middle ear cavities acting as resonator enhanced hearing in both underwater and air hearing conditions.

Based on the morphological similarity between the middle ear among turtles (i.e. enlarged and bony enclosed middle ear, laterally expanded extracolumella forming a disc, and tympanic area separated from the lower jaw), it appears likely that the middle ear of mosasauroids functioned as that of extant turtles. In other words, that the key sound receiver was the tympanic disc (= expansion of the extracolumella), the bony tympanic conch acting as a resonator.