Because physiological and life history strategies of extant and fossil vertebrates are recorded in their hard tissues, such as bone and teeth (Bromage et al., 2009, Chinsamy-Turan, 2005 and Klevezal, 1996), paleohistology allows reconstruction of fitness components in extinct vertebrates. Life history traits are often associated with changes in the individual's energy budget, such as the channelization of resources away from growth toward reproduction (Ricklefs, 2007, Roff, 2002 and Stearns, 1992). These changes in the energy budget affect the rate of bone tissue apposition and leave marks on the growing bones, which allow reconstruction of certain life history traits (Castanet, 2006 and de Margerie et al., 2002, de Margerie et al., 2004). Teeth provide additional information about life history strategies. Because mammalian dental tissue formation is periodic, cyclic variations during the ameloblast secretory activity lead to the formation of incremental marks in the enamel (Boyde et al., 1988). These allow determination of the rate and duration of dental growth and development, thus providing insight into developmental pathways and into life history traits of mammals (Beynon et al., 1998, Dean, 2006, Dean et al., 2001, Schwartz et al., 2002 and Smith et al., 2007).

Because of the trade-off between growth and reproduction, mammals usually attain sexual maturity when final body size is reached. Therefore, the beginning of a dense outer circumferential layer (OCL) of organized lamellar tissue in the cortex of long bone shafts, formed when growth rate suddenly decreases, is considered as a proxy of the onset of fecundity (Chinsamy-Turan, 2005 and de Margerie et al., 2002).

Estimate of age at sexual maturity in M. balearicus and in G. borbonica was based on counts of annual growth marks (lines of arrested growth ‘LAGs’) in transversal sections of long bone midshafts (skeletochronology, see Castanet, 2006, Castanet et al., 2004, Chinsamy-Turan, 2005 and Erickson, 2005).

Fossil material: We sectioned 60 femora of M. balearicus from four sites (Cova de Llenaire, Es Bufador, Cova de Moleta and Son Maiol). The sample comprises all ages, from neonate through very old individuals. Not all thin sections were sufficiently well preserved as to allow analysis of the microstructure, so that we used 18 well-preserved specimens that represent almost all ages. Material of G. borbonica comes from two Pliocene sites of Teruel (La Puebla de Valverde) and Soria (Layna), Spain. The sample comprises 21 long bones (humeri, femora, radius and tibiae) of different age stages: five juvenile individuals and 13 adult individuals. Three individuals lack their epiphyses, so that no information about their age was available from macroscopic observation. Our descriptions of bone tissues are based on the typological classification established by de Margerie et al., (2002).

In mammals, tooth eruption is an adaptive trait tightly associated with life history (Schultz, 1935, Schultz, 1956 and Smith, 2000). The eruption of first permanent molars is commonly associated with weaning (Franz-Odendaal et al., 2003, Macho, 2001, Macho and Williamson, 2002, Schwartz et al., 2002 and Smith, 1989). We used first permanent lower molar (M₁) eruption time in M. balearicus as a proxy for age at weaning. M₁ eruption time in M. balearicus was estimated by calculating Crown Formation Time (CFT) in a slightly worn first lower molar (MBCN15801, Cova de Moleta, Majorca), using the incremental structures of enamel tissue (Beynon et al., 1998, Boyde, 1963, Bromage, 1991, Dean et al., 2001 and Smith, 2008) observed in a thin section of the tooth cut longitudinally in the buccolingual plane at the mesial cusps (Fig. 1a). The method used to calculate CFT follows Jordana and Köhler (2011). Enamel prisms (Fig. 1b, A) mark the path of growth; incremental features (Fig. 1b, B) represent the position of the developing enamel front. Thus, the time required to form the distance comprised by the complete path of an incremental feature (Fig. 1b, C), is the same time required to form a prism from enamel-dentine junction to enamel surface (Fig. 1b, A). The latter is calculated by dividing the length of a prism by the average Daily Secretion Rate (DSR). DSR was calculated measuring the distance between adjacent fine incremental features (daily growth features) in different regions of the crown (Fig. 1c). Previously, this distance was compared along the inner and outer enamel and no significant differences were found. This process is repeated along the tooth crown in order to calculate the increase in crown height from cuspal region to cervical region as a function of time.

Additionally, we obtained X-ray images (Fig. 2) of an ontogenetic series of mandibles from M. balearicus (Cova de Moleta, Majorca), which allowed us to determine the extent of M₁ crown that is formed at the postnatal stage, since the formation of this molar already starts at the prenatal stage. Mandibles were scanned with a Siemens Sensation 16 CT-scan at Hospital Mútua de Terrassa (Spain) with an interslice space of 0.2 mm. Images were analysed using the software OsiriX v.3.5.1.

The estimate of M₁ eruption time in G. borbonica, as a proxy for age at weaning, was not possible due to the lack of dental material.

Histological slides, both of bones and teeth, were made following standard procedures (Chinsamy-Turan, 2005 and Klevezal, 1996) and examined under transmitted and polarized light. Specimens labeled ICP are housed at the Institut Català de Paleontologia, Universitat Autònoma de Barcelona, Bellaterra, Spain. Specimens labeled MBCN are housed at the Museu Balear de Ciències Naturals, Sóller, Majorca (Spain).

Life history strategies are shaped by two factors, their scaling with body mass and their ecological context (resource availability, extrinsic mortality). Life history traits are fitness components because they regulate the balance between reproduction and growth and/or maintenance. They determine demographic characteristics such as population growth rate, generation time (the weighted mean age of mothers at offspring birth in a given population, Leslie, 1966), recruitment of juveniles, mortality rates, and others. Life history traits, hence, provide important insights into the ecology as well as into the vulnerability of species (Brown and Sibly, 2006, Calder, 1984, Palkovacs, 2003, Ricklefs, 2007, Roff, 2002 and Stearns, 1992).

Ungulates are large mammals, which is reflected in most of their life history traits (Peters, 1983). They are characterized by low fecundity (only one or two offspring/year), a long potential life span, and strong iteroparity (Gaillard et al., 2000). Environmental variability and density-dependence, however, are documented to cause considerable phenotypic flexibility such as temporal variability in certain fitness components, which is especially obvious for age at sexual maturity (Gaillard et al., 2000 and Saether, 1997). Age at maturity is a critical trait as it marks the moment in an individual's life at which resources are shifted away from growth toward reproduction (Kirkwood, 2005 and Stearns, 1992). The age at first reproduction is found to provide reliable information about the ranking of a given species along the slow-fast continuum of life history traits (Gaillard et al., 2000). Among populations, however, there is a certain degree of variability in this trait due to the dependence of female size and body condition on resource availability and population density, which may vary over time and geographic area. Thus, populations in habitats with low resource levels frequently show late primiparity (Gaillard et al., 2000 and Saether, 1997). Age at first conception in females of Gazella, for instance, oscillates between 0.5 and two years (Mendelssohn et al., 1995, Nowak, 1999 and Wronski and Sandouka, 2008), with G. dorcas from resource poor desert environments showing the broadest span and the longest delay, from 190 days (Cassinello, 2005 and Ralls et al., 1980) to 660 days (Furley, 1986 and Yom-Tov et al., 1995). Despite these variations, ungulates are rather conservative in their life history strategies as summarized in their life history graph (Fig. 5), taken from Gaillard et al. (2000). In general terms, therefore, fitness components in ungulates are predictable.

Our results show that, indeed, our continental G. borbonica conforms to the predictions (Fig. 5). It is similar in size to Gazella dorcas, measured by the length of the radius (G. borbonica 132 mm, # ICP 28290, La Puebla de Valverde; G. dorcas, 131.3 mm, # 7536 HH). Taking into account that G. borbonica is not an extreme desert form like some living species such as Gazella dorcas (Andrés and DeMiguel, 2008, Kurtén, 2009 and Suc et al., 1995), it should be expected that G. borbonica reaches maturity well within the first year of life. Indeed, the first resting line appears in the OCL after overall growth has ceased.

M. balearicus, however, does not fit the predictions. The mean body mass (close to 25 kg at Cova de Moleta and Cova Estreta) (Köhler and Moyà, 2004) is comparable to that of Gazella rufifrons, which attains sexual maturity after 270 days (Walther, 1990), and smaller than that of the relative Rupicapra rupicapra (24–50 kg) which reaches maturity at two years (Nowak, 1999). Instead, growth decreased in M. balearicus at 8 years, at the earliest, indicating that it attained maturity at that age or later, with an important individual variability (Fig. 5). The high plasticity in this fitness component is likely related to environmental variability and density dependence, factors found to commonly underlie variability in this trait in ungulates (Gaillard et al., 2000 and Saether, 1997), and which are accentuated on small islands (MacArthur and Wilson, 1967).

Certain characteristics point to a correlated delay in at least two other fitness components, namely age at weaning and life span. Thus, the eruption of the first molar (M₁) in M. balearicus took place at the age of one year, contrasting with M₁ eruption in extant bovids, which ranges from 1 to 6 months (Smith, 2000). Owing to the straight correlation between M₁ eruption and weaning in mammals (Franz-Odendaal et al., 2003, Macho, 2001, Macho and Williamson, 2002, Schwartz et al., 2002 and Smith, 1989), our results strongly suggest a delayed age at weaning for M. balearicus.

The large number of very old animals as seen in the extremely worn teeth housed in the many collections with remains of Myotragus, is not known from any continental site. Bone histology also provides evidence that Myotragus reached surprisingly old ages for wild ungulates, as shown in Fig. 6. Moreover, in agreement with previous findings from living ruminants (Carranza et al., 2004 and Veiberg et al., 2007), the outstanding increase in molar crown height (hypsodonty) in this species might not have evolved, at least not primarily, in response to diet but it rather increased the individual's life expectancy by extending the lifetime of the feeding apparatus. Altogether, this strongly suggests a high survival rate and an unusually long lifespan for this insular bovid (Köhler, 2010).

Contrary to our fossil continental gazelle, hence, the life history graph of the insular M. balearicus completely differs from the generalized graph for extant ungulates. It shows an important delay in all those fitness components that can be reconstructed.

Generation time is a fitness component with a great impact on population growth (Gaillard et al., 2000) and, hence, a key trait in the threat-rating system for long-lived and late-maturing organisms (Oates, 2006). We calculated generation time for M. balearicus using the equation by Pianka (2000):

T = (α + ω)/2

where T is generation time, α age at first reproduction, and ω age at last reproduction (here age of the oldest individual). This equation is only a very crude way to estimate T, and there are several problems involved in its use. Thus, for ungulate populations both the age at first reproduction and the age at last reproduction is available in a few exceptional cases only (for instance, Loison et al., 1999). Another source of error is the fact that senescence is pervasive among ungulates, with old females showing senescence in both survival and reproduction (Gaillard et al., 2000). The formula, however, calculates the midrange and, hence, does not take into account that the mean age of mothers at offspring birth is skewed towards younger ewes. Generation time calculated with this equation, hence, must be expected to be higher than estimates obtained from field data. This is indeed the case when compared with data yielded by the allometric relationship between generation time and adult body mass in ungulates based on field observations: Ln (mean age of mother at time of birth) = 0.967 + 0.247 Ln (adult body mass) (Gaillard, 2007). Thus, the formula based on vital rates (age at first and last reproduction) provides an average generation time of 10 for ungulates of the size class of Myotragus (14–30 kg), while the body mass based formula yields an average generation time of only 5.6 for the same size class. Nevertheless, though generation time and other biological times correlate positively with body mass, there is an important variation even among populations (Gaillard, 2007) reflecting selection under different environmental conditions irrespectively of body mass (Veiberg et al., 2007). Thus, only the formula based on vital rates provides information about differences in generation times between species of comparable size. In a first attempt to estimate generation time in a fossil mammal, hence, we will use Pianka's formula to provide a rough idea of the ranking of Myotragus along the slow-fast continuum in life history traits.

The late onset of maturity and the outstanding longevity in M. balearicus provide the extraordinarily long generation time of approximately 17 years (based on an old individual with a minimum of 26 LAGs and assuming that it attained sexual maturity at a minimum age of 8 years; Fig. 6). Seventeen years is much longer than the average generation time of other bovids (Gaillard, 2007 and Pianka, 2000) and is comparable only to estimates for long-lived and slow-maturing mammals such as large primates (Oates, 2006). A long generation time (K-species) makes sense in the context of resource limitation under insularity (MacArthur and Wilson, 1967). However, as in the case of primates, it makes the species vulnerable to those factors that increase mortality rate, such as predation pressure or hunting, because it imposes long time periods of recovery from declines in population size. We suggest that the long generation time of M. balearicus was the likely cause underlying its fast extinction after the arrival of man some 3000 years ago.

We analysed the evolution of fitness components in two fossil bovids under different selective regimes: the insular M. balearicus and the mainland G. borbonica. The former taxon evolved under chronic resource limitation and low extrinsic mortality, whereas the latter evolved under high extrinsic mortality. Results have been compared with those from studies of extant large herbivores. Ungulates are rather conservative in their life history strategies; however, environmental variability and density dependence are documented to cause important temporal variability in certain fitness components, which is especially obvious for age at sexual maturity. Our results show that G. borbonica conforms to the predictions for ungulates of similar body size; M. balearicus, however, completely differs from the generalized life cycle for extant ungulates, as it shows an important delay in all those fitness components that can be reconstructed from the fossil remains, namely age at weaning, age at sexual maturity, life span and generation time. This variation in life history strategy makes sense in the context of resource limitation under insularity.