This paper was written thanks to a pre-doctoral subsidy from the Fundación Atapuerca assigned to the IPHES (Instituto de Paleoecología Humana y Evolución Social) and to projects CGL 2012-38358 and CGL 2009-12703-C03-03 of the Spanish Ministry of Economy and Competitiveness, and SGR 2009-324 of the Generalitat de Catalunya, as well as to the help of the Junta de Castilla and the great field work carried out by the research teams at each site.

J.M. L.-G. benefited a Beatriu de Pinós postdoctoral fellowship (2011BP-A00272) from the Generalitat de Catalunya, a grant co-funded by the European Union through the Marie Curie Actions of the 7^th Framework Program for R&D.

1 - Introduction

1The first appearance data (FAD) for Mimomys savini in Europe date to ca. 1.8 Ma (Topachevskij & Skorik, 1977; Jánossy, 1986; Rekovets, 1994; Fejfar et al., 1998), and the species survived until 0.6 Ma (Fejfar et al., 1998). In the Iberian Peninsula this taxon was documented for the first time around 1.6 Ma in the Guadix-Baza Basin (Agustí, 1986; Agustí et al., 1987; 2010). Its earliest record coincides approximately with the first presence of hominids on the European continent, i.e. with the Homo antecessor remains found at Gran Dolina, Atapuerca (Bermúdez et al., 1997; Carbonell et al., 2008), and with the lithic industries and cut-marks at the sites in the Orce region in the Guadix-Baza Basin (Martínez-Navarro et al., 1997; Oms et al., 2000; Toro et al., 2010).

2Mimomys savini is the oldest known representative of the so-called water vole lineage, all the members of which are characterized by rather large m1 teeth. Lower first molars of extant Arvicola species, which are descendants of, or at least forms closely related to, Mimomys savini (see discussion below), can reach lengths of as much as 4 mm.

3To date, Mimomys savini is the last known representative of the genus Mimomys. The evolution of this genus shows a trend that brings its molar morphology successively closer to that of the genus Arvicola. Mimomys savini is considered by many authors to be the ancestor of Arvicola (Chaline & Sevilla, 1990; Viriot et al., 1990; Chaline et al., 1999).

4Due to their rapid evolutionary changes and dispersal, rodents have become a particularly useful tool in worldwide biozonation (arvicolines in the Holarctic region), separating the Pleistocene into the Villanyian, the Biharian and the Toringian stages (in the sense of Fejfar and Heinrich (1980) and Heinrich (1987) among others). The Biharian, which corresponds to the Microtus-Mimomys superzone (characterized by the coexistence of these two genera), is subdivided into two rodent zones: the lower is the Mimomys savini + Mimomys pusillus zone, and the upper is the Mimomys savini zone (fig. 1) (Fejfar et al., 1998). In the biochronological subdivision defined for the Iberian Peninsula, Mimomys savini is present in the Iberomys brecciensis biozone, the Allophaiomys chalinei biozone, the Allophaiomys lavocati biozone and the Allophaiomys pliocaenicus biozone (Cuenca-Bescós et al., 2010b).

5The oldest known populations of Mimomys savini in Europe date to the Early Pleistocene period and have been recorded at Tiligul (Topachevskij & Skorik, 1977) and Zhevakhova Gora in the Ukraine (Rekovets, 1994), Villány 5 in Hungary (Jánossy, 1986), Fuente Nueva 3 and Barranco León in Spain (Agustí, 1986; Agustí et al., 1987; 2010), and levels TD 3-4 (the current TD 4 was divided into TD 3 and TD 4 in the earliest papers) of Gran Dolina in Spain (Cuenca-Bescós et al., 1995; 1999; Cuenca-Bescós & García, 2007; Cuenca-Bescós et al., 2010b). The most recent populations correspond to the Early-Middle Pleistocene transition and have been recorded at several sites such as Zapadnye Kairy in the Ukraine (Markova, 1982, 2004), Kärlich E and F in Germany (von Koenigswald & van Kolfschoten, 1996), Voigtstedt in Germany (Maul, 2002), West Runton in England (Stuart, 1981), and level TD 8 of Gran Dolina in Spain (Cuenca-Bescós et al., 2010b).

6In this article we present the results of our studies on the evolution of the molar morphology of Mimomys savini from the Iberian Peninsula. On the basis of these results, we discuss the phylogenetic relationship between Mimomys savini and the genus Arvicola.

2 - Geographical, geological and chronological background

7This paper is based on the data obtained from Mimomys savini remains from Barranco León (level D), Fuente Nueva 3, Gran Dolina (levels TD 4B, TD 5b, TD 5a, TD 6-3, TD 6-2 and TD 6-1), Vallparadis (layer 10) and La Boella (level 2) sites (fig. 2).

2.1 - Fuente Nueva 3 and Barranco León

8Fuente Nueva 3 (FN-3) and Barranco León (BL) are located within the Guadix-Baza Basin, which forms part of the Betic Mountain Range in the northern part of the province of Granada (Spain) (fig. 2). These sites fall within the Allophaiomys lavocati biozone for the Iberian Peninsula (Agustí et al., 2010; Cuenca-Bescós et al., 2010b; Oms et al., 2011), which corresponds to the upper part of the Early Pleistocene.

9The first data on the rodent fauna of Barranco León were published by Agustí et al. (1987); the first on Fuente Nueva 3 by Martínez-Navarro et al. (1997).

10The section at Fuente Nueva 3 presents a sedimentary record of some 20 meters of thickness, and the studied section of Barranco León presents a 25-m-thick sedimentary record. Both consist of a single stratigraphic unit that is Early Pleistocene in age. The samples analyzed for this article come from the layers of red and white clays of Fuente Nueva 3 and from level D of Barranco León (known in the first excavations as level 5) (fig. 2) (Toro et al., 2010).

11Palaeomagnetic data have revealed that these two sites date back to before the Jaramillo event (Oms et al., 2000), while ESR dating has shown that level D of Barranco León dates back to 1.46 ± 0.38 Ma, and Fuente Nueva 3 to a period between 1.34 and 1.67 Ma (Duval, 2008). Recently, however, a new dating result of 1.19± 0.21 Ma was published for Fuente Nueva 3 (Duval et al., 2011), thus giving a new chronological range between 1.46 ± 0.38 Ma and 1.19± 0.21 Ma.

2.2 - Gran Dolina

12Gran dolina is one of the sites that form part of the archaeo-palaeontological complex of Atapuerca. It is located in a cave formed as a continuation of one of the fracture planes of the Upper Cretaceous layers of the Sierra de Atapuerca. Broadly speaking, this sierra, situated approximately 15 km to the east of the city of Burgos (Spain), is a small rocky promontory situated at the north-eastern limit of the hydrographic basin of the River Duero.

13The sedimentary filling of Gran Dolina is divided into 11 levels (from TD 1 to TD 11). The Matuyama-Brunhes magnetic polarity reversal has been located in level TD 7 (fig. 2). The material under study in this paper corresponds to the section between the top of TD 4 (TD 4b) and the top of TD 6 (Parés & Pérez-González, 1995).

14Levels TD 5 and TD 6 are subdivided into various sublevels, of which TD 5b, TD 5a, TD 6-3, TD 6-2 and td 6-1 provided remains of Mimomys savini.

15Chronologically, studied levels encompass a period extending from approximately 1 Ma to 0.78 Ma (Falguères et al., 1999; Berger et al., 2008; Cuenca-Bescós et al., 2010b).

2.3 - Vallparadis

16The site of Vallparadis is located in the north-east of the Iberian Peninsula in the province of Barcelona (Spain). It displays a faunal association characteristic of the Lower Pleistocene. The sedimentary sequence is divided into two parts separated by an erosive discontinuity (unit 5) and is formed by the intercalation of fluvial and alluvial sedimentation (fig. 2). The palaeontological and archaeological remains are found in almost all the sequence (Madurell-Malapeira et al., 2010; Martínez et al., 2010; Minwer-Barakat et al., 2011). The archaeological excavation focused mainly on the richest unit, unit 7 (Martínez et al., 2010).

17Unit 7 presents two archaeological layers: layer 10 and layer 10c.

18Layer 10c has been interpreted as an accumulation produced by massive transport due to a high-density fluid with erosive capacity, whereas layer 10 has been interpreted as flood muds deposited above the previous layer (fig. 2) (Madurell-Malapeira et al., 2010; Martínez et al., 2010).

19US-ESR dating carried out on layer 10 yielded a chronology of 0.83 ± 0.07 Ma (Martínez et al., 2010).

2.4 - La Boella

20The site of La Boella (Barranco de la Boella, La Boella Ravine) is situated in the north-east of the Iberian Peninsula in the province of Tarragona, near the municipality of La Canonja. Two sites are currently being excavated within the ravine of La Boella, El Forn and La Mina, the first of which yielded the remains under study in this paper. The stratigraphy of the ravine is divided into 4 units (fig. 2). The material under study comes from level 2 at the top of unit 3.

21The studies carried out on this site date it around 0.7 Ma (Saladié et al., 2008; Vallverdú et al., 2008).

3 - Methodology

22The remains were recovered by water-screening the sediment with both Freudenthal tables (Daams & Freudenthal, 1988) and a tower of sieves. Which equipment was used depended on the richness of the sediment. The sediment from sites very rich in remains, such as Gran Dolina, was processed using the tower of sieves. For sediments with few remains we used the Freudenthal table, which makes it possible to wash larger samples in less time. For the washing tables, mesh sizes of 5 mm (concentrate did not contain microfaunal remains), 2 mm and 0.5 mm were used.

23Remains of Mimomys savini were extracted from the 2-mm concentrate. Lower first molars (m1) and upper third molars (M3), from adult specimens only, were used for the study in order to prevent the high morphological variability of juvenile forms from biasing the results. We regarded m1 and M3 as belonging to adult animals when enamel-free areas were visible at the occlusal surface and roots were closed or in the process of closing (fig. 3).

24On these teeth, the presence or absence of enamel islets was recorded for m1 and M3, and of the Mimomys ridge for m1. Measurements were taken of the length (L) and the width (W) of m1 and of the enamel thickness, as shown in figure 3. The SDQ (enamel differentiation index) was calculated as:

25where N refers to the number of dentine fields of the studied tooth; teet (trailing edge enamel thickness) refers to the maximum thickness of the posterior enamel band; and leet (leading edge enamel thickness) refers to the maximum thickness of the anterior enamel band of each dentine field (Heinrich, 1978, 1987; von Koenigswald & van Kolfschoten, 1996; modified by Cuenca-Bescós et al., 2010a).

26All the measurements were taken on the occlusal plane of the molars with the image-treatment software Adobe PhotoShop CS5, using photographs from an Infinity X camera coupled to an Olympus SZ-40 binocular microscope. This software ensures highly reliable measurements made on flat surfaces.

4 - Material

27The characters of Mimomys savini m1 are:

28- an occlusal surface with a posterior lobe,

29- five alternating triangles and an anteroconid complex,

30- the presence of roots (except in the earliest stages of life),

31- enamel with Mimomys-type differentiation (Heinrich, 1978, 1987; von Koenigswald et al., 1994; von Koenigswald & van Kolfschoten, 1996),

32- cement in the re-entrant angles,

33- T4 and T5 widely confluent,

34- depending on the age of the populations, a shallow or absent BRA3; and the neck of the head of the anteroconid complex tends to be narrow (fig. 3).

35Most of the remains from Barranco León and Fuente Nueva 3 show dark coloration, and in many cases they are fractured and relatively fragile. In the absence of a detailed taphonomic study, this is interpreted as being a consequence of diagenetic processes in acid environments. It is significant that the remains come from layers showing the dark coloration characteristic of high organic matter content, which would have been responsible for the acidification of the environment. The sample from Fuente Nueva 3 and Barranco León consists of 132 m1 and 32 M3.

36The material under study from Gran Dolina was recovered in the course of a drilling carried out at the site during the 1990s (Cuenca-Bescós et al., 1995, 1999). Material from more than one level of Gran Dolina was studied (TD 4B, TD 5b, TD 5a, TD 6-3, TD 6-2 and TD 6-1). The total number of remains from Gran Dolina is 364 m1 and 28 M3.

37The sample from Vallparadis used in this study is made up of 22 m1. It corresponds to remains recovered between 2005 and 2007 (Martinez et al., 2010).

38Finally, as regards the remains studied from La Boella, the sediment is very poor in microfaunal remains, and the few remains that are recovered tend to be highly fractured. Normally only remnants of enamel are recovered. For this reason, the sample used in this study was restricted to just two complete m1 recovered during the 2010 campaign.

5 - Results

39As the chronology of Fuente Nueva 3 includes that of Barranco León and in the light of the fact that they occur within the same geological context, the samples from these two sites were grouped together.

40The data were arranged in accordance with the available physical datings, from oldest to youngest: Barranco León and Fuente Nueva 3, Gran Dolina TD 4B, Gran Dolina TD 5b, Gran Dolina TD 5a, Gran Dolina TD 6-3, Vallparadis, Gran Dolina TD 6-2, Gran Dolina TD 6-1 and La Boella.

41The results show a reduction in the percentage of individuals with the Mimomys ridge over the evolution of the species and changing values for the percentage of the population with the enamel islet, with the maximum values associated with the oldest sites (Barranco León and Fuente Nueva 3). The enamel islet on M3 only occurs in populations older than 1 Ma (Barranco León and Fuente Nueva 3), whereas the enamel islet on m1 and the Mimomys ridge are retained until at least 0.78 Ma (fig. 4 and tab. 1).

42The SDQ index of the species for this period does not show a clear trend. It oscillates between values of 160 and 140 (fig. 5 and tab. 2).

43The length and the width of m1 show a clear trend to increase during this period, with the exception of the L values obtained from La Boella (fig. 6 and tab. 2).

6 - Discussion

6.1 - Evolution of Mimomys savini in the Iberian peninsula

44This study shows that the evolution of Mimomys savini in the Iberian Peninsula is characterized by a gradual disappearance of ancestral characters in the m1 (Mimomys ridge and enamel islet) and M3 (enamel islet) and an increase in size. The specimens from La Boella have not been taken into account for the conclusions of this paper because the sample was not large enough for the statistics to be considered reliable.

45The enamel islet in M3 is not present in populations dating to less than 1 Ma, while the percentage of individuals with the enamel islet in m1 shows a gradual decrease in populations. This evolutionary decrease had previously been noted for this species by Cuenca-Bescós et al. (1999), who ascertained that in Lower Pleistocene populations of the percentage of individuals with the enamel islet on m1 is greater than 14%, whereas it is lower in Middle Pleistocene populations. In our study these percentages are not attained, which could be due to the different size of the sample or to the inclusion of juvenile individuals in Cuenca-Bescós et al. (1999). Nevertheless, the trend towards a decrease in the number of individuals with the enamel islet on m1 could be clearly made out in both cases: i.e. both the study by Cuenca-Bescos et al. (1999) and the present investigation. Both characteristics, the enamel islet and the Mimomys ridge, are lost during the individual’s life (due to wear). The reduction in the population with these characteristics indicates the gradual disappearance of these characters in the evolution of the species (given that the palaeontological population is a reflection of the original biological population).

46These changes constitute a progressive process over the whole stratigraphic range of M. savini and represent an evolution from an archaic m1 (showing similarities to the m1 of Mimomys ostramosensis) to a m1 which - except for the presence of roots - is very similar to the first populations of Arvicola (Arvicola cantianus) studied at Meisenheim I (van Kolfschoten, 1990) and at a number of localities of the Holsteinian, Eemian, Saalian and Weichselian of central Europe and the Pannonian Basin (Heinrich, 1982). The set of evolutionary changes undergone by Mimomys savini represents a clear evolutionary trend, a phenomenon noted in other arvicoline genera such as Ondatra (Martin, 1979), Kislangia and the theridomyids (Vianey-Liaud, 1972; Chaline & Sevilla, 1990; Agustí et al., 1993; Chaline et al., 1993).

6.2 - Relationship between the genus Arvicola and Mimomys savini

47In arvicolines the increase in size at the lineage level has been studied for Kislangia cappetai - Kislangia gusii and Mimomys occitanus - Mimomys pliocaenicus (Agustí et al., 1993), as well as for Mimomys occitanus - Mimomys savini (Viriot et al., 1990; Néraudeau et al., 1995; Chaline et al., 1999). As the evolution of Mimomys savini molars involves a loss of Mimomys-type characters and an increase in size that brings it closer to the size and molar morphology of the earliest populations of Arvicola, an evolutionary relationship between these taxa (Mimomys savini-Arvicola cantianus) seems to be clear. It is corroborated by the chronology of the records and the fact that the evolution of the genus Arvicola also shows an increase in size. This relationship has already been suggested by many authors (e.g. Viriot et al., 1990; Chaline et al., 1999),

48However, the discovery of the new species Arvicola jacobaeus dated to approximately 1.2 Ma (Cuenca-Bescós et al., 2010a) raises a number of doubts regarding this evolutionary relationship (fig. 7). This record showed that the first representatives of Arvicola are similar in age to Mimomys savini. As this article reveals, the dental morphological characters (SDQ index, the presence of an enamel islet and Mimomys ridge) and size of M. savini are different from those of Arvicola jacobaeus during this period. Among M. savini populations, a higher percentage of individuals than in modern ones possess an enamel islet (Arvicola jacobaeus does not have an enamel islet), and their average length is approximately 3.2 mm, whereas in Arvicola jacobaeus smallest length values are 3.48 mm (Cuenca-Bescós et al., 2010a). The data obtained in this study, together with the classification of A. jacobaeus, thus force us to reconsider the evolutionary relationship between Mimomys savini and Arvicola.

49Various alternatives explaining the relation between Mimomys and Arvicola have been proposed. Chaline & Sevilla (1990) and Horàček (1990) argue that Mimomys savini gave rise to Arvicola terrestris. Rabeder (1981) proposes that the origin of Arvicola sapidus can be traced to Mimomys savini, whereas Arvicola terrestris would be the evolutionary descendant of Mimomys milleri. By contrast, Ruiz Bustos (1999) proposes the existence at 1.6 Ma of the species Arvicola deucalion, an ancestor of Mimomys tornensis. This theory raises a problem, because Arvicola deucalion is considered as a synonym of Allophaiomys deucalion by most authors who also uphold the phylogenetic relationship between these two species, Mimomys tornensis - Allophaiomys deucalion (Rabeder, 1981; Garapich & Nadachowski, 1996; Chaline et al., 1999), or at least between Mimomys sp. and Allophaiomys deucalion. This phylogenetic relationship between Mimomys tornensis and Allophaiomys deucalion is defended by Chaline & Sevilla (1990), who also suggest that Mimomys savini is the ancestor of Arvicola.

50As mentioned above, the increase in size during the evolution of Mimomys savini also occurs in Arvicola. Additionally, the morphological similarities between the two taxa would imply a phylogenetic relationship between them.

51When the SDQ values obtained in this study are compared with those from various populations of Arvicola throughout Europe (fig. 8), both Arvicola jacobaeus and the oldest populations of Arvicola cantianus display similar values to those of Mimomys savini. Moreover, the most modern populations of Mimomys savini (Gran Dolina TD 6-2, TD 6-1 and Voigtstedt) show a trend towards a reduction in this index, which is continued in Arvicola populations up to recent times. The continuous evolution of this character would be an indicator that Mimomys savini is the ancestor of Arvicola cantianus, as proposed by Chaline et al. (1999). Taking into account that the youngest populations of M. savini have length and SDQ values similar to the earliest A. cantianus populations, both species follow the same evolutionary trend for these characters (Abbassi & Desclaux, 1996; Escudé et al., 2008a, 2008b), the morphology of their m1 is similar, and they are only clearly distinguished by the absence of roots.

7 - Conclusions

52In the Iberian Peninsula, Mimomys savini displays clear evolutionary trends: the successive reduction or loss of archaic characters in stratigraphically consecutive samples, such as the enamel islet on M3, which is only present in populations older than 1 Ma, and the enamel islet and the Mimomys ridge on m1.

53In the course of its evolution, this species also undergoes an increase in size. The only character that seems not to follow a clearly directed evolutionary trend in M. savini is the SDQ value, which only shows oscillatory variations.

54The occurrence of Arvicola jacobaeus during similar periods to Mimomys savini requires the assumption of a common ancestor shared by Mimomys savini and the genus Arvicola, as already argued by Cuenca-Bescós et al. (2010a). However, this paper has demonstrated that the youngest populations of Mimomys savini show a trend towards a decrease in the SDQ index, which is continued by the oldest Arvicola populations. The evolution of this character supports Mimomys savini as a possible ancestor of the Arvicola genus, as proposed by Chaline et al. (1999), or at least of Arvicola cantianus. However, the origin of Arvicola jacobaeus still remains unexplained.

55Everything seems to indicate that two groups with different origins have been classified as Arvicola: on the one hand Arvicola jacobaeus, which shares a common ancestor with Mimomys savini, and on the other hand Arvicola cantianus and descendants, whose ancestor would be Mimomys savini.