Journal of Mammalian Evolution

, Volume 14, Issue 3, pp 163–181

Dental Microwear Texture Analysis of Varswater Bovids and Early Pliocene Paleoenvironments of Langebaanweg, Western Cape Province, South Africa

Authors

    • Department of AnthropologyUniversity of Arkansas
  • Gildas Merceron
    • Biozentrum Grindel and Zoological MuseumUniversity of Hamburg
  • Robert S. Scott
    • Department of AnthropologyUniversity of Arkansas
Original Paper

DOI: 10.1007/s10914-007-9050-x

Cite this article as:
Ungar, P.S., Merceron, G. & Scott, R.S. J Mammal Evol (2007) 14: 163. doi:10.1007/s10914-007-9050-x
  • 256 Views

Abstract

The extensive early Pliocene mammalian assemblages at Langebaanweg hold the potential to provide important information about paleoenvironments of the southwestern tip of Africa, an area that today consititutes the Fynbos Biome. We here add to a growing body of literature on the paleoenviornments of the site with an examination of dental microwear textures of bovids from the Varswater Formation. Microwear texture analysis is a new, automated and repeatable approach that measures whole surfaces in three dimensions without observer error. A study of extant ruminants indicates that grazers have more anisotropic microwear surface textures, whereas browsers have more complex microwear surface textures. Fossil bovids recovered from the Muishond Fontein Pelletal Phosphorite Member vary in their microwear textures, with some taxa falling within the extant browser range, some closer to extant grazers, and others in between. These results are consistent with scenarios suggesting mosaic habitats including fynbos vegetation, some (probably C3) grasses, and woodland elements when these fossils were accumulated.

Keywords

Dental microwearBovid dietsLangebaanweg paleoecologyPaleoenvironments

Introduction

Dental microwear, the study of microscopic scratches and pits that form on a tooth’s surface as the result of its use, is a valuable tool for reconstructing the diets of fossil mammals. Dental microwear has been examined for mammals representing a broad range of taxa, from primates (e.g., Jacobs 1981; Strait 1993, 1998; Daegling and Grine 1994; Lucas and Teaford 1994; Teaford et al. 1996; Ungar 1996, 1998; Ungar and Teaford 1996; King et al. 1999a; Rafferty et al. 2002; Leakey et al. 2003; Ungar et al. 2004; Godfrey et al. 2004; El Zaatari et al. 2005; Merceron et al. 2005a), to perissodactyls (e.g., Hayek et al. 1991; MacFadden et al. 1999; Solounias and Semprebon 2002; Kaiser and Solounias 2003), artiodactyls (e.g., Solounias et al. 1988; Solounias and Moelleken 1993; Hunter and Fortelius 1994; Rivals and Deniaux 2003; Semprebon et al. 2004; Franz-Odendaal and Solounias 2004; Merceron et al. 2004a; Merceron and Ungar 2005; Merceron et al. 2005b; Merceron and Madelaine 2006; Schubert et al. 2006), rodents (Gutierrez et al. 1998; Lewis et al. 2000; Hopley et al. 2006), carnivorans (Van Valkenburgh et al. 1990; Anyonge 1996), proboscideans, (Capozza 2001; Filippi et al. 2001; Green et al. 2005) and others (Krause 1982; Biknevicius 1986; O’Leary and Teaford 1992).

Dental microwear has proven to be particularly useful for distinguishing between grazing and browsing ungulates (Solounias and Moelleken 1992a, b, 2004; Merceron et al. 2005b). This has potentially important implications for reconstructing paleoenvironments (Merceron and Ungar 2005; Schubert et al. 2006; Merceron and Madelaine 2006; Teaford et al. in press). While there is no one-to-one correspondence between diet and habitat, there is a reasonable association between grazing and grass availability and between browsing and the presence of woody cover (e.g., Estes 1991; Kingdon 1997; Sponheimer et al. 1999). Indeed, grazers, browsers and mixed feeders tend to inhabit more open, closed and mixed or seasonal environments, respectively (Gagnon and Chew 2000). Thus, studies of dental microwear of ungulates can, especially when considered alongside other lines of evidence, offer clues concerning environments of the past.

In this paper we look at the dental microwear of bovids from early Pliocene deposits at Langebaanweg in the Western Cape Province of South Africa with an eye toward the paleoenvironments of the site. While preliminary work has already been reported based on more conventional microwear techniques (Merceron and Ungar 2005), we here present data collected using a new and radically different approach, dental microwear texture analysis. This new approach has proven to be a valuable addition to the microwear arsenal, distinguishing living primates with known differences in diet (Scott et al. 2006; Ungar et al. in press) and fossil hominin species (Scott et al. 2005). Still, despite hints that this might also be useful for studies of bovids (Ungar et al. 2003), no microwear texture analysis has been published to date to determine whether this technique can distinguish extant grazers from browsers, or whether it can be used to infer food preferences of fossil ungulates. This paper presents the first dental microwear texture analysis of extant ruminants, and uses resulting data as a baseline for interpretation of diets of early Pliocene bovids from Langebaanweg.

Langebaanweg paleoenvironments

The fossil deposits at Langebaanweg are superlative for the study of Mio-Pliocene environmental dynamics. Thousands of vertebrate specimens have been recovered from these deposits over the past half century, yielding the oldest known substantive Pliocene fossil assemblage in southern Africa. Most of these come from the Varswater Formation; more specifically, from two river channel deposit exposures (Beds 3aS and 3aN) associated with the Muishond Fontein Pelletal Phosphorite Member (MPPM). The MPPM lithostratigraphic unit was laid down during the early Pliocene transgression, and dates to circa 5 Mya. While the temporal relationship between the two beds is unclear, Bed 3aS was claimed by Hendey (1981) to be slightly older than 3aN. Hendey further argued that Bed 3aS may contain some fossils reworked from the underlying Quartzose Sand Member (also Varswater Formation).

Researchers have looked to both the palynological and faunal records from Langebaanweg to reconstruct paleoenvironments at the site. Much of this work reflects efforts to put the southwestern tip of Africa into the context of the wide-scale expansion of C4 grasses that was occurring elsewhere at the time (Cerling et al. 1997). Fossil pollens have provided some intriguing glimpses, but most work to date has focused on the diets of the remarkable vertebrate assemblage, with studies ranging from taxonomic uniformitarianism, to hypsodonty indices, stable isotope analyses, dental mesowear, and dental microwear.

Hendey’s (1981, 1983, 1984) pioneering work offered a first glimpse at environments around the Western Cape during the earliest Plicoene. He noted, based largely on taxonomic uniformitarianism, that most herbivores in the assemblage were grazers or mixed feeders. Occurrences of alcelaphines, bovines, and equids suggested to Hendey more open country, though reduncine and boselaphine elements reflected some woodland or forested patches. He suggested that dense forest undergrowth would have been minimal given a paucity of smaller grazing ungulates at Langebaanweg compared with Miocene occurrences further north in the southern Namib Desert. He concluded that the faunal assemblages at Langebaanweg were an “indication of the environmental transition which was underway during the Early Pliocene,” involving “a progressive development of grasslands at the expense of forests” (Hendey 1984).

Klein (1982) noted the catastrophic mortality profiles of fauna such as Sivatherium, Giraffa, and Damalacra species at the site; a pattern that likely resulted from drowning events during peak flooding. Klein (1982) proposed on this basis that these animals must have lived close to a river, perhaps in gallery forests or associated floodplain grasslands. These animals may have been, according to Hendey (1983) attracted to vicinity of the proto-Berg River by its fresh water. The mouth of the river would have been near to the site, as would the shoreline given higher sea levels than today (Hendey 1983; Roberts and Brink 2002).

Scott (1995), based on palynological samples from the Varswater Formation, helped flesh out the vegetal context at Langebaanweg. While Scott’s samples were not rich in pollen, some interpretation was possible. The dominant type was Ranunculaceae (buttercup), probably occurring locally as aquatic or semiaquatic plants in a swampy basin according to Scott (1995). Tree pollen types were also found in small numbers, indicating the occurrence of some woodland in the general surroundings. Still, the samples also contained open vegetation elements, including fynbos shrubs. This is not surprising, as fynbos vegetation was present earlier as well, along with subtropical forests during earlier Elandsfontein Formation times (Coetzee and Rogers 1982).

The take home message of paleoenvironmental reconstructions offered through the 1990s was a mixed setting, not inconsistent with some spread of well-watered grasslands to the Langebaanweg area by the early Pliocene. A series of more recent studies by Franz-Odendaal et al. (2002, 2003, 2004) has brought new analytical tools to bear on the questions regarding details of the paleoenvironment of Langebaanweg, and has greatly improved our understanding of the site.

Franz-Odendaal et al. (2002) first examined stable isotopes in enamel of representative ungulates, including giraffids, hippopotamids, suids, and bovids from bed 3aN. Their results offered little evidence of C4 plant consumption. The authors concluded that, given earlier interpretations of grazing-dominated fauna, Langebaanweg probably retained C3 grasslands in the early Pliocene. Further support for the availability of grasslands then came from the study by Franz-Odendaal et al. (2003) of mesowear in ‘Eurygnathohyppus’ cf. ‘E.’ baardi. Results suggested that these MPPM equids were predominantly grazers, with some variation (specimens from Bed 3aS perhaps took a bit more browse than those from 3aN). Franz-Odendaal and Solounias (2004) then analyzed Sivatherium hendeyi, using hypsodonty indices, dental mesowear, and dental microwear to reconstruct the diets of these giraffids. Results suggested mixed feeding, indicating the presence of at least some browse in the area during MPPM times.

Despite this recent flurry of activity, work on the bovid fauna at Langebaanweg has remained limited. This is surprising because bovids are often considered excellent paleoenvironmental indicators (e.g., Vrba 1980, 1985; Kappelman 1984; Harris 1991; Plummer and Bishop 1994; Spencer 1997; Sponheimer et al. 1999; Schubert et al. 2006). These ungulates tend to be ubiquitous at African Pliocene fossil sites, and modern taxa can usually be placed in conventional dietary categories that to at least some degree reflect habitat preferences. Thus, available bovids from Langebaanweg are the focus of this study.

Materials and methods

Specimens used in this study were all housed at the IZIKO South African Museum in Cape Town. All available maxillary and mandibular first and second molar teeth from MPPM bovids identified in the collections to species or genus were examined. Specimens were inspected first by hand lens and light microscopy if necessary to determine suitability for microwear analysis. Taphonomic damage is especially common for fossil bovid molars as their thin bands of occlusal surface enamel are frequently chipped or broken, and microwear is not preserved. Accordingly, the vast majority of the Langebaanweg bovids examined did not possess unobscured antemortem microwear. Criteria for assessing taphonomic damage followed Teaford (1988) and King et al. (1999b). We found in total 100 specimens suitable for microwear analysis (only one tooth per named specimen was included in this study to avoid weighting the sample). These include individuals representing Damalacra (n = 25), Gazella cf. G. vanhoepeni (n = 4), and Raphicerus paralius (n = 2) from Bed 3aS and Damalacra (n = 7), Kobus subdolus (n = 21), Mesembrioportax acrae (n = 25), and Simatherium demissum (n = 16) from Bed 3aN. Our Damalacra samples may include both D. acalla and D. neanica (Gentry 1980; Vrba 1998; Brink personal communication), but if so, their molar teeth are too similar for us to reliably assign them to one or the other.

In addition, some extant ruminants were analyzed to serve as a baseline for interpretation of Langebaanweg fossil bovid microwear patterns. The baseline series included wild-shot museum specimens from the US, South Africa, and France (see Table 1). Many of these were molded and impressions made available to us by Blaine Schubert (see Schubert et al. 2006 for details). The extant sample included 58 specimens representing nine extant taxa chosen for their range of diets and habitat preferences. These include both open-country grazers (Hippotragus niger, Kobus leche, Oryx gazella, and Redunca arundinum) and mixed or more closed habitat browsers (Litocranius walleri, Sylvicapra grimmia, Tragelaphus strepsiceros, Cephalophus monticola, and Capreolus capreolus; e.g., Estes 1991; Tixier and Duncan 1996; Kingdon 1997; Gagnon and Chew 2000).
Table 1

Descriptive microwear texture statistics for extant Bovidae

Taxon

Statistic

n

Complexity

Anisotropy

Collection

Capreolus capreolus

Mean

10

3.45

0.0048

INRA

Median

 

1.31

0.0048

Standard Deviation

 

4.10

0.0014

Skewness

 

1.33

0.0700

Cephalophus monticola

Mean

8

3.67

0.0043

SAM, AMNH

Median

 

3.11

0.0036

Standard Deviation

 

1.64

0.0024

Skewness

 

1.21

0.1300

Hippotragus niger

Mean

5

1.51

0.0073

AMNH

Median

 

1.37

0.0078

Standard Deviation

 

0.70

0.0014

Skewness

 

1.63

−1.5900

Kobus leche

Mean

5

1.45

0.0056

TM

Median

 

1.17

0.0049

Standard Deviation

 

1.03

0.0021

Skewness

 

2.02

1.3600

Litocranius walleri

Mean

9

2.33

0.0020

MCZ, AMNH

Median

 

2.04

0.0017

Standard Deviation

 

0.87

0.0012

Skewness

 

1.43

1.0600

Oryx gazella

Mean

6

1.99

0.0062

SAM, AMNH

Median

 

1.39

0.0070

Standard Deviation

 

1.22

0.0024

Skewness

 

0.86

−1.8200

Redunca arundinum

Mean

4

0.93

0.0058

MCZ, AMNH

Median

 

0.92

0.0061

Standard Deviation

 

0.23

0.0030

Skewness

 

0.21

−0.5100

Sylvicapra grimmia

Mean

7

2.17

0.0035

MCZ

Median

 

1.65

0.0033

Standard Deviation

 

1.81

0.0020

Skewness

 

0.95

0.6700

Tragelaphus strepsiceros

Mean

4

4.63

0.0019

TM

Median

 

4.45

0.0018

Standard Deviation

 

1.82

0.0008

Skewness

 

0.38

0.2900

AMNH American Museum of Natural History, New York; INRS Institut National de la Recherche Agronomique, Tolouse; MCZ Museum of Comparative Zoology, Cambridge, MA; SAM South African Museum, Cape Town; TM Transvaal Museum, Pretoria

High resolution replicas were prepared following conventional procedures (Ungar 1996). Teeth were cleaned with cotton swabs soaked in acetone, and crown surface molds were prepared using President’s Jet regular body polyvinylsiloxane dental impression material (Coltène-Whaledent Corp.). Positive replicas were poured using Epotek 301 high-resolution epoxy resin and hardener (Epoxy Technologies Corp.). It was not necessary to coat casts with a conductive material or mount them on stubs as would be typical of SEM based microwear analyses.

Each replica was then mounted on the stage of a Sensofar PLμ confocal imaging profiler (Solarius, Inc). This white-light scanning confocal microscope collects 3D point clouds representing a surface, and does so rapidly and with better axial resolution than a standard laser based confocal system. A 100× objective was used, and surface elevations for each specimen were collected at a lateral (x, y) interval of 0.18 μm with a vertical resolution of 0.005 μm. The resulting lateral resolution exceeds that typical of SEM based studies of specimens scanned at 500× (e.g., Grine et al. 2006; Ungar et al. 2006). Data were collected on Facet 1, located on the distobuccal enamel band of the mesial cuspid of M1 or M2, or the mesiobuccal enamel band of the mesial cusp of M1 or M2 (Janis 1990; Merceron et al. 2005b). The center of the facet, or field closest to it that preserved antemortem microwear was sampled. The work envelope for each surface was 100 × 140 μm, represented by nearly 432,000 points.

Each point cloud was analyzed using scale-sensitive fractal analysis (SSFA) software (ToothFrax and SFrax, Surfract Corp.). The fundamental premise of this approach is that a surface can look different at different scales—an asphalt road may appear smooth to a motorist driving along it, but rough and bumpy to an ant trying to cross it. Our group has to this point identified five types of microwear texture attributes that can distinguish species with different diets (Ungar et al. 2003, 2007; Scott et al. 2005, 2006). We focused this study on two variables expected to be especially useful for distinguishing grazers and browsers: area-scale fractal complexity (Asfc) and length-scale anisotropy of relief (epLsar). We will refer to these as complexity and anisotropy, respectively.

Complexity is measured as change in surface roughness at different scales. Asfc is basically the slope of the steepest part of a curve fit to a plot of measured relative area over the range of scales at which those measurements are made (see Figs. 1 and 2). The steeper the slope, the more complex the surface (see Scott et al. 2006 for details). A surface dominated by pits of various sizes, or pits and scratches will tend toward high complexity.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig1_HTML.gif
Fig. 1

Dental microwear texture analysis attributes. Plots of anisotropy (top) and complexity (bottom) calculated for a “pitted” surface (a) and a more striated one (b). The mean vector length for the top images (individual vectors indicate relative lengths of transects sampled at 1.8 μm) is used to calculate epLsar, and the maximum slopes in the bottom image indicate Asfc. See the text for further explanation.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig2_HTML.gif
Fig. 2

Diagrammatic representations of surfaces with different levels of anisotropy and complexity.

Anisotropy is a measure of orientation concentration of surface roughness. epLsar is a measure of differences in lengths of depth profiles sampled across a surface at different orientations at a given scale; in this case a sample interval of 5° and a scale of 1.8 μm (see Scott et al. 2006). Average relative lengths of profiles and their orientations (treated as vectors) are normalized using the exact proportion method and a mean vector length is calculated. A surface dominated by striations running parallel to one another will have high anisotropy (see Figs. 1 and 2).

Statistical analyses were conducted on the extant species to determine whether taxa reported to be grazers differed significantly from those reported to be browsers. Nested analysis of variance models were used in this study with species as the subordinate level of classification nested within broad diet category. All data were rank-transformed before analysis to mitigate possible effects of violation of assumptions inherent to parametric statistical procedures (Conover and Iman 1981). Complexity and anisotropy were considered separately. These tests allowed us to determine whether the variation between the diets for each microwear attribute was greater than that expected given variation among the taxa within the diet categories. If so, values for each fossil taxon could then be compared with those for the extant species to determine whether the Langebaanweg bovids could be placed reliably within the grazer or browser category.

Results

Descriptive and analytical statistics are presented in Tables 1, 2, and 3, and illustrated in Figs. 3, 4, 5, 6, and 7.
Table 2

Nested analyses of variance on extant taxa

Source

SS

df

MS

F

p

Complexity (ASFC ranked data)

Diet

3,425.814

1

3,425.814

16.259

0.000

Species (diet)

3,318.400

7

474.057

2.250

0.046

Error

10,324.192

49

210.698

  

Anisotropy (epLsar ranked data)

Diet

5,101.437

1

5,101.437

30.284

0.000

Species (diet)

3,249.244

7

464.178

2.756

0.017

Error

8,254.079

49

168.451

  
Table 3

Descriptive microwear texture statistics for fossil Bovidae

Taxon

Statistic

n

Complexity

Anisotropy

Damalacra ssp. (Bed 3aN)

Mean

7

2.16

0.0050

Median

 

2.05

0.0052

Standard deviation

 

0.78

0.0016

Skewness

 

−0.74

−0.1800

Damalacra ssp. (Bed 3aS)

Mean

25

3.68

0.0038

Median

 

3.71

0.0037

Standard deviation

 

1.54

0.0014

Skewness

 

0.11

−0.2100

Gazella cf. G. vanhoepeni

Mean

4

3.36

0.0039

Median

 

2.52

0.0036

Standard deviation

 

3.38

0.0018

Skewness

 

1.04

0.8200

Kobus subdolus

Mean

21

3.65

0.0040

Median

 

3.12

0.0034

Standard deviation

 

2.52

0.0022

Skewness

 

1.42

0.9500

Mesembriportax acrae

Mean

24

4.06

0.0031

Median

 

4.09

0.0032

Standard deviation

 

2.10

0.0018

Skewness

 

0.89

1.1200

Raphicerus paralius

Mean

2

7.73

0.0016

Median

 

7.73

0.0016

Standard deviation

 

1.57

0.0001

Skewness

   

Simatherium demissum

Mean

15

3.16

0.0046

Median

 

2.64

0.0049

Standard deviation

 

2.04

0.0019

Skewness

 

1.14

−0.2800

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig3_HTML.gif
Fig. 3

Photosimulations of microwear surface texture of a grazer (above) and a browser (below) generated from point cloud data. The sampled area is 100 × 140 μm.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig4_HTML.gif
Fig. 4

Mean complexity (Asfc) histograms comparing extant grazers and browsers. Lines associated with each histogram bar indicate one standard error above and below the mean.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig5_HTML.gif
Fig. 5

Mean anisotropy (epLsar) histograms comparing extant grazers and browsers. Lines associated with each histogram bar indicate one standard error above and below the mean.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig6_HTML.gif
Fig. 6

Photosimulations of microwear surface texture of fossil bovids generated from point cloud data. The sampled area is 100 × 140 μm.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-007-9050-x/MediaObjects/10914_2007_9050_Fig7_HTML.gif
Fig. 7

Mean anisotropy values plotted against mean complexity values for the extant taxa and individual fossil species.

Extant ruminant microwear textures

The extant ruminant species identified as browsers (Litocranius walleri, Sylvicapra grimmia, Tragelaphus strepsiceros, Cephalophus monticola, and Capreolus capreolus) all had higher average Asfc and lower average epLsar values than those considered to be grazers (Hippotragus niger, Kobus leche, Oryx gazella, and Redunca arundinum; Figs. 3, 4, and 5; Table 1).

The distinction between diet categories is confirmed by the Nested ANOVA results (Table 2). The “big story” for both variables is the higher level differences between grazers and browsers, though species also differed significantly on the subordinate level (within the categories). Looking first at complexity, browsers have significantly higher Asfc values than grazers (p = 0.000). Variation between species within diet categories was also significant (p = 0.046). The same pattern of differences was evident for anisotropy. In this case, grazers have significantly higher epLsar values than browsers (p = 0.000). Variation between species within diet categories was again also significant (p = 0.017).

Fossil bovid microwear textures

The fossil bovid microwear data include species with averages for Asfc and epLsar that fall within the ranges of the extant browsers, the extant grazers, and between the two (Figs. 6 and 7; Table 3). Aniostropy is plotted against complexity in Fig. 7. Mean values for the fossil species can be compared with those for the extants in AsfcepLsar bivariate space. Only Gazella cf. G. vanhoepeni has both mean Asfc and epLsar values within the ranges of the extant grazers. Damalacra from Bed 3aN has a mean Asfc value within the grazer range, but an epLsar value between that of extant grazers and browsers. Simatherium demissum has both Asfc and epLsar values between those of extant grazers and browsers. Finally, Damalacra from Bed 3aS, Kobus subdolus, Mesembrioportax acrae, and Raphicerus paralius all have mean values within the extant browser space. Raphicerus paralius has a rather high mean Asfc value even compared with extant browsers, though we would not take too much stock in this given their dismal sample size (n = 2).

Concerning possible differences between the two MPPM river channel deposits, there is little evidence for a separation between Bed 3aN and Bed 3aS taxa. Still, Damalacra from Bed 3aN does appear to be shifted in the “graze” direction relative to the 3aS Damalacra sample. We further note that that there is at this point little evidence for a systematic difference in proportions of D. acalla and D. neanica between the beds that might explain this.

Discussion

These results have important implications both for the use of microwear texture analysis to reveal diets of fossil bovids and for the paleoenvironments of Langebaanweg.

Dental microwear and ruminant diets

Basic theory

While the basic contrasts between grazing, browsing, and mixed feeding (Hofmann and Stewart 1972) are clearly oversimplifications (Bodmer 1990; Gordon and Illius 1994), they do have value for this study, as grasses and browse differ in their nutrient content and physical properties (Shipley 1999). Grasses tend to have very thick cell walls dominated by cellulose (Demment and Vansoest 1985). In addition, veins of monocot grasses tend to be parallel to the long axes of their blades, making these structure “notch insensitive,” wherein it becomes especially difficult to transmit stresses across those blades to effect fracture (Vincent 1990). Individual grass blades may also be too thin to store sufficient energy to allow cracks to spread freely through them (Sanson 2006). As a final note, grasses may have very high concentrations of abrasive silica, as endogenous phytoliths, and/or exogeneous grit adherent to individual blades (Baker et al. 1959; McNaughton et al. 1985; Sanson et al. 2007).

Browse is much more difficult to characterize because it includes a broader variety of food items. Still, browse items tend to have thinner cell walls and more readily accessible nutrients. They are often weaker than grasses, affording less resistance to fracture propagation. Browse components may be more brittle, requiring less work to propagate a crack through them. On the other hand, some browse items are also harder than grass, requiring greater stress to initiate a fracture.

The basic contrast between “stress delimited” (hard) and “displacement delimited” (tough) foods (Ashby 2005) can have important implications for the biomechanics of chewing (Lucas 2004), and by implication, patterns of dental microwear. Tough grasses would tend to involve more lateral movement of opposing occlusal surfaces relative to one another (e.g., “shearing,” or grinding), whereas harder, more brittle foods may require more vertical contacts of opposing surfaces with food items crushed between them. More parallel tooth–food–tooth interactions should produce microscopic striations as opposing teeth slide past one another, dragging abrasives across their surfaces (Rensberger 1973; Franz-Odendaal et al. 2003). This can also cause small, prism sized pits due to the phenomenon of “prism plucking” (Walker 1984; Teaford and Runestad 1992). Browse foods, on the other hand, can vary more in size and shape, with larger, harder, and more brittle items requiring more crushing expected to lead to pits of a variety of shapes and sizes.

Previous microwear studies

Results presented here accord with previous studies that have distinguished browsing and grazing ungulates on the basis of dental microwear. As one would predict, grazers tend to have occlusal surfaces dominated by microscopic scratches, whereas browsers tend to have more pits. This was demonstrated using conventional feature-based microwear analyses in the classic works of Solounias and his colleagues (e.g., Solounias et al. 1988; Solounias and Moelleken 1992a, b, 1993). In those studies, microwear surfaces were imaged at 500× using a scanning electron microscope and the lengths and breadths of individual features were digitized. The average numbers of scratches and pits (distinguished by length-to-breadth ratios of greater and less than 4:1, respectively) separated modern grazers (with more scratches) and browsers (with more pits). Mixed feeders tended to show bimodal distributions.

This basic dietary dichotomy is so unambiguous that grazers and browsers can be distinguished by counting numbers of scratches in a field of view using a binocular light microscope at low magnifications (e.g., Solounias and Semprebon 2002; Rivals and Deniaux 2003; Semprebon et al. 2004). Grazers have more striations than browsers. Other techniques combining low-magnification light microscopy (Merceron et al. 2004a, b) or white light confocal microscopy (Merceron and Ungar 2005; Schubert et al. 2006) with semiautomated image analysis have also been effective in distinguishing extant grazing and browsing ungulates. Once again, grazers tend to have more striations, whereas browser surfaces have relatively more pits.

Microwear texture analysis and ruminant diets

Despite small extant baseline sample sizes, microwear texture analysis also clearly distinguishes grazers from browsers. Bovids identified as grazers have significantly more anisotropic surfaces, whereas ruminant browsers have significantly more complex surfaces. All of the grazers have higher mean epLsar values than any of the browsers, and all of the browsers have higher mean Asfc values than any of the grazers. In other words, there is no overlap between the taxa in anisotropy-complexity bivariate space. This translates roughly, for those used to thinking about features rather than textures, to more parallel striations for the grazers, and more features of varying sizes and shapes for the browsers (but see below). It is not surprising then, that the species identified here as having the highest and lowest Asfc values (Tragelaphus strepsicerous and Redunca arundinum, respectively) also have the highest and lowest mean pit percentage values for these same specimens (see Merceron and Ungar 2005).

Dental microwear texture analysis may also hold the promise of allowing distinction of finer resolution differences between ruminants. For example, Tragelaphus strepsicerous has the highest mean Asfc value, and a rather catholic diet (including leaves, herbs, fruits, succulents, vines, tubers and flowers), whereas other browsers with less complex surfaces tend toward more leaves. In contrast, the tall grass specialist, Redunca arundinum has the lowest complexity, whereas grazers with more variable diets (e.g., Hippotragus niger and Oryx gazella) have higher Asfc values (Estes 1991; Kingdon 1997; Fortelius and Solounias 2000; Gagnon and Chew 2000). Further study of larger samples including more ruminant species with known subtle differences in diet should give us a better idea of the potential resolution of microwear texture analysis for identifying fine-scale diet differences.

Microwear features and surface textures

Results presented here accord to some extent, though not fully, with feature-based microwear analyses of these same specimens (Merceron and Ungar 2005). This is not surprising, as there is no one-to-one correspondence between surface texture attributes and mean feature size, shape or density. Area-scale fractal complexity is a measure of how rapidly surface roughness changes with scale of observation. Thus, high pit percentages can correspond to either high or low Asfc values depending on the complexity (e.g., density, range of sizes and shapes) of features on that surface. Further, none of the feature attributes reported by Merceron and Ungar (2005) is analogous to texture anisotropy.

Feature-based and texture-based microwear analyses differ fundamentally in their approaches to characterizing microwear surfaces. Microwear texture analysis characterizes surfaces in three dimensions without the need to measure individual features, or to classify them as discrete types. A 4:1 length-to-breadth ratio, for example, may be of heuristic value for distinguishing scratches from pits, but it is inherently arbitrary because feature lengths and breadths vary along a continuum. Characterizations of whole surfaces rather than tallying features may also have more biological meaning, as there is surely no one-to-one correspondence between individual features and single feeding events. Finally, a scale-sensitive approach affords the benefit of examining patterns across a range of scales, potentially taking advantage of signals evident at both higher and lower magnifications.

Dental microwear of the bovids from Langebaanweg

The bovids of Langebaanweg MPPM Bed 3aS and Bed 3aN show a broad range of microwear texture patterns. While none studied evince the obligate grazer pattern seen in Redunca arundinum, Gazella cf. G. vanhoepeni and Damalacra from Bed 3aN do appear closer to the extant grazer cluster than the living browsers. Other taxa, such as Mesembrioportax acrae, Damalacra from Bed 3aS, and Raphicerus paralius fall squarely within the browser category (though we are hesitant to make much of the R. paralius results given a sample size of n = 2). Still others, such as Simatherium demissum and Kobus subdolus have values more clearly intermediate between extant grazers and browsers.

These results suggest a fairly mixed setting at Langebaanweg during MPPM times. The fossil bovids do not show a predominantly “obligate graze” microwear texture signal. This is consistent with models suggesting that paleoenvironments at Langebaanweg were not dominated by widespread open grasslands. Still, some taxa do have values closer to that of extant grazers, portending at least some availability of grasses, or at least of food items that would result in similar microwear textures. On the other hand, some of the MPPM bovids do have microwear texture values within the browse range. This suggests the availability of browse, and paleoenvironments in which such foods would have been available (e.g., some woodland and/or fynbos shrubland).

Results presented here also allow us to compare bovid microwear textures between the two river channel exposures. These offer little evidence of consistent differences between Bed 3aS and Bed 3aN. The one possible hint that these may sample slightly different paleoenvironments is that the Damalacra sample has shifted from a more browser like to a more grazer like microwear signal (if Bed 3aS is indeed the earlier river channel).

Results from the fossil bovid microwear texture analysis can be considered in the context of previous efforts to understand the paleoenvironments of Langebaanweg. Original interpretations of Hendey (1981, 1983, 1984), based largely on taxonomic uniformitarianism suggested a somewhat open setting with forested patches. This accorded well with the idea of a transitional habitat, intermediate between more closed Miocene settings and more open ones later in the Plio-Pleistocene.

More recent work has begun to put together a more comprehensive picture of Langebaanweg paleoenvironments. Scott’s (1995) palynological study suggested the presence of aquatic or semiaquatic plants in a swampy basin, with fynbos vegetation and woodland in the general surroundings during Varswater times. And there is little doubt that the area was relatively well-watered at the time (Hendey 1983; Roberts and Brink 2002).

Stable isotope analyses of the enamel of various ungulates from Bed 3aN show little evidence for C4 plant consumption (Franz-Odendaal et al. 2002), indicating that Langebaanweg probably did not experience the wide-scale expansion of C4 grasses that was occurring elsewhere at the time (Cerling et al. 1997). The lack of grazer-dominated microwear texture signals is certainly consistent with this.

That said, some other lines of evidence are consistent with the presence of at least some grasses. These include evidence of grazing for ‘Eurygnathohyppus’ cf. ‘E.’ baardi based on mesowear analysis (Franz-Odendaal et al. 2003) and evidence of mixed feeding for Sivatherium hendeyi based on hypsodonty indices, dental mesowear, and dental microwear (Franz-Odendaal and Solounias 2004). Evidence from the microwear texture analysis of the bovids is also consistent with at least some grazing and presumably mixed feeding.

In sum, the microwear texture data for MPPM bovids, taken along with other lines of evidence, point to a mosaic paleoenvironment including fynbos vegetation, some (probably C3) grasses, and woodland elements. This seems likely for both Beds 3aS and 3aN. Still, there is a hint that the two river channel deposits may sample slightly different paleoenvironments, as both ‘Eurygnathohyppus’ cf. ‘E.’ baardi mesowear and fossil bovid microwear textures of Damacra suggest slightly more browse in Bed 3aS compared with Bed 3aN.

Dental microwear texture analysis holds the potential to reveal important new information about the diets of fossil mammals at Langebaanweg and other sites. Much more work remains to be done, both on expanding this avenue of research to include other taxa, and on better understanding how dental microwear textures relate to diets of ruminants. Of particular relevance to Neogene sites in the Western Cape Province is the need to determine how fynbos relates to microwear textures. As Franz-Odendaal and Solounias (2004) duly noted, “determining the wear features that fynbos produces is not easily accomplished.”

Acknowledgements

We thank Dave Roberts, James Brink, Kaye Reed and Matt Sponheimer for discussions related to this project. We also thank Thalassa Matthews and Richard Smith for their kind invitation to participate in the workshop that prompted us to conduct the research described in this paper. We are grateful to the reviewers for their helpful comments on an earlier version of this manuscript, to Blaine Schubert for his help collecting molds of many of the extant specimens and to Margaret Avery and Graham Avery for allowing us to study the fossil materials in their care. Research described in this paper was funded by US National Science Foundation Grants BCS 0222176, 0315157 and 0510038 and a Fyssen Foundation Postdoctoral Fellowship to GM.

Copyright information

© Springer Science+Business Media, LLC 2007