BOVID LIMB ELEMENT MORPHOLOGIES, LOCOMOTOR REPERTOIRE, HABITAT PREDICTIONS, AND ECOLOGICAL MODELING OF THE UPPER LAETOLIL BEDS AT LAETOLI IN NORTHERN TANZANIA by TRACEY REN Âƒ E LANCASTER B.A., University of Northern Colorado, 2011 A thesis submitted to the Faculty of the Graduate School of the University of Colorado i n partial fulfillment of the requirements for the degree of Master of Arts Department of Anthropology 2014
ii 2014 TRACEY RENÂƒE LANCASTER ALL RIGHTS RESERVED
iii This thesis for the Master of Arts degree by Tracey Ren ÂŽ e Lancaster has been approved for the Department of Anthropology by Charles Musiba Chair Zaneta Thayer David Tracer May 2 nd 2014
iv Lancaster, Tracey R. (Master of Arts, Anthropology) Bovid Limb Element Morphologies, Locomotor Repertoire, Habitat Predictions, and Ecological Modeling of the Upper Laetolil Beds at Laetoli in Northern Tanzania Thesis directed by Associate Professor Charles Musiba ABSTRACT Paleoecological models are integral to paleoanthropological research. By discovering s elective forces that influenced early hominin evolution, researchers are able to discern ecological circumstances under which early hominins evolved and also better understand what caused the hominin lineage to evolve. Paleoecological models can provide important environmental details that may help researchers understand why certain morphological characters appear in the fossil record. Since paleoecological models are so important to understanding hominin evolution this study creates a paleoecological model of Laetoli, a unique Pliocene hominin site in northern Tanzania. This study uses e comorphology which is a taxon free method that a nalyzes morphological characters to determine the habitats to which fauna are adapted. This study in particular examines morphological variations in bovid astragal i from the Upper Laetolil Beds (3.85 3.63 Ma) and determines that bovids and the associated h ominins present at the site were occupying open environments with pockets of forest The form and content of this abstract are approved. I recommend its publication. Approved: Charles Musiba
v ACKNOWLEDGEMENTS I would first and foremost like to thank Dr. Charles Musiba my adviser and thesis chair for his support and guidance through all stages of this thesis and throughout my master's program This thesis and the opportunities I have had during my graduate work would not have been poss ible without him. I would also like to express my gratitude t o the Tanzanian government and the Smithsonian Museum of Natural History for granting me access to the collection s To my committee members, Dr. Zaneta Thayer and Dr. David Tracer, I am grateful for your comments and suggestions To Connie Turner, thank you for your unconditional help a nd encouragement throughout this process F inally I cannot express my deep est gratitude to Nate Miller, f or keeping my head above water, and to my family for their unwavering support I am sincerely indebted to all of these people.
vi TABLE OF CONTENTS CHAPTER I INTRODUCTION 1 The Study S ite: Laetoli 3 Why Bovids? 8 Biases in Paleoecological Models 8 II. PALEOECOLOGICAL MODELS OF EAST AFRICA & LAETOLI 12 Micro & Macrobotanical Analyses : Polle n, Phytoliths, and Macrofossils 12 Stable Isotope Analyses 14 Faunal Assemblages Analyses 16 Taxonomic Methods 17 Taxon Free Methods 22 II I MATERIALS AND METHODS 24 Linear Measurements 26 Statistics 28 Habitat Cate gories 28 Confounding Factors 2 9 III. RESULTS 30 Descriptive Statistics 30 Principal Components Analysis 32 Discriminant Function Analysis Modern Astragali Collect ion 33 Misclassifications 37
vii Uncertain Classifications 40 Fossil Astragali Collection 42 Uncertain Classifications 42 I V. DISCUSSION 45 V. CONCLUSION 47 REFERENCES 5 0
viii LIST OF TABLES TABLE 3 .1 Taxonomy and Habitat Categories of Modern Bovid Collection 24 2 6 4 .1 Means of Habitat Category by Variable 3 1 4 2 Multivariate Statistics Showing the Means of the Habitat Categories are Significantly Different 3 2 4 .3 Principal Component Analysis of All Variables 32 4 .4 Results of DFA on Individual Variables 3 4 4 .5 Results of DFA on Grouped Variables 35 4.6 Results of DFA with All Variables Included 35 4.7 Summary Table of Modern Collection DFA With All Variables Included 36 4.8 Misclassifications of Modern Collection DFA 40 4.9 Uncertain Classifications for Modern Collection at the .10 Level 41 4.10 Results of DFA on Fossil Collection at the .05 Level 42 4.11 Summary of Uncertain Classifications for F ossil Collection at the 0.05 L evel 45
ix LIST OF FIGURES FIGURE 1.1 Map of Laetoli in Northern Tanzania 5 1.2 Map of L ocalities at Laetoli 6 1.3 Stratigraphy of Laetoli 7 3.1 Astragali Linear Measurements 27 4 .1 Loading Plot of Principal Components Analysis 33
1 CHAPTER I INTRODUCTION Paleoecological modeling is an integral aspect of paleoanthropological research. Especially when new fossil hominin species are discove red, special attention is given to the paleoecological forces that infl uenced the evolution of these species. Recent fossil hominin discoveries illustrate this point because each new hominin species has at least one article published analyzing the paleoeco logy of the new species (Brunet et al 2002, Berger et al 2010, Haile Selassie et al 2010, Rightmire et al 2006 Senut et al 2001, Ward et al 1999, White et al 2010; WoldeGabriel et al. 2010 ). However, d ebates continue because unusual morphological ch aracters present in some hominins leave researchers with many questions Why does a bipedal Ardipithecus ramidus also have a divergent big toe? Why are Paranthropus molars so large and why did they need such strong chewing muscles? These questions, among many others, are complex, and researchers today are only just beginning to answer them One of the ways to continue illuminating these topics is to better understand the selective pressures under which these species evolved. Since no species can escape the ecological pressures of its environment discovering factors that influenced early hominin evolution may elucidate what caused the human lineage to evolve, as well as why certain morphological characters are seen in the fossil record. To further our understanding of the environments in which early hominins evolved t his study constructs a paleoecological model of a paleoanthropological site in northern Tan zania Laet oli Specifically, this study uses ecomorphology to analyze bovid astragali morphology as a proxy for the habitats that bovids may have occupied at Laetoli
2 during the Upper Laetolil Beds (3.85 3.63 Ma) (Deino 2011) 109 bovid astragali were recovered from the Upper Laetolil Beds at localities 7 8 and 9 from the 2010 2013 field seasons. In this study, I create a model which discerns the types of habitats bovids occupied at the site. The accuracy of the model is determined by a comparative sample of extant bovid astragali, which are from the Smithsonian Museum of Natural History (n=177) The comparative sample is classified to the genus and species level and the habitat preference is known for each specimen (T able 3 .1). Then, a discriminant function analysi s is completed for the modern collection, which predicts the habitat category of the bovids based on the morphology of their astragali. Next, a misclassification rate is determined for the modern collection which in turn makes it possible to evaluate the accuracy of the fossil collection's discriminant function analysis The method that this study employs is e comorphology which is also referred to as functional morphology. It is a paleoecological modeling method bas ed on the fundamental principle that an organism's postcranial elements reflect the terrain and the substrate in which the organism evolved (Plummer et al 2008). Many studies have used this method to discover bovid a nd cervid habitats (Curran 2012; DeGusta and Vrba 2005 2003; Kappelman 199 1; Kappelman et al 1997; Klein et al 2010, Musiba and Magori 2003; Plummer and B ishop 1994; Plummer et al. 2008 ). This method has proven especially useful when applied to bovid astragali ( Bishop et al. 2011, DeGusta and Vrba 2003, Plummer et al 2008 ) T his is due to their ubiquity in the fossil record in East Africa. Collectively, these studies provide a growing body of evidence that bovid and cervid postcrania reflect the habitats in which the se animals live. Therefore, since ecomorphology has proven to be an useful method and the bovid astragali has been
3 shown to be a reliable habitat indicator, this study uses bovid astragali to construct a paleoecological model of the habitats bovids occupied during the Upper Laetolil Bed s (3.85 3.63 Ma) at Laetoli, n orthern Tanzania. The Study Site: Laetoli Laetoli is a rich, fossiliferous, Plio Pleistocene site where fossil hominins such as Australopithecus afarensis Australopithecus aethiopicus, and anatomically modern Homo sapiens have been discovered. Additionally, hominin footprints and animal trackways fossilized in volcanic ash, and vertebrate faunal assemblages have been discovered at this site ( Harrison 2011 a Leakey and Harris 1987, Musiba 1999). It is situated in the southe rn part of the East Afric a n R i ft Valley on the drainage between Olduvai Gorge and Lake Eyasi, and within the Ngorongoro Conservation Area, Tanzania (Fi gure 1 .1 ). The East African Rift Valley is an area with past and present day volcanism, which can be use ful for dating f ossils because sometimes fossils are found between volcanic layers that span a relatively short time period, making some fossil dates very temporally specific. Furthermore Laetoli is unique in its taphonomic history from other hominin si tes in East Africa. Taphonomic studies of the Upper Laetolil Beds indicate the presence of hot and dry as well as wet environments (Musiba et al. 2007, Musiba personal communication) which indicates a complex environment with either different habitats coe xisting at the same time or short term environmental changes Moreover, Laetoli is a hominin site that had different environmental pressures than other East African sites because it lacked permanent water sources (Harrison 2011a Musiba 1999 ). Steams, rive rs, and other water sources would have been seasonal (Ditchfield and Harrison 2011).
4 The fossil bovid astragali collection that this study uses w ere obtained from localities 7 8, and 9 (Figure 1.2) and were all found in the Upper Laetolil Beds (Figure 1. 3) which date to 3.85 3.63 Ma (Deino 2011) Assemblages from the Upper Laetolil Beds are noteworthy because they date to a time in which fossil faunal evidence at other sites is rare (Harrison 2011 a ). Finally, all of the fossil astragali are well preserve d. Weathering is limited to stage I for most of the assemblage. In the entire assemblage, there were only 28 specimens with weathering in areas that affected the measurements, and only those were not included in this study The total number of fossil astragali that this study uses is 109.
5 Figure 1 .1 : Map of Laetoli in Northern Tanzania (taken from Musiba et al. 2007 )
6 Figure 1. 2: Map of L ocalities at Laetoli (taken from Musiba et al. 2007 )
7 Figure 1. 3: Stratigraphy of Laetoli (taken from Deino 2011 )
8 Why Bovids? Using bovids as habitat predictors is not a new approach in paleoanthropology ( Bobe et al. 2007; DeGusta and Vrba 2005, 2003; Gentry 2011, 1987, 1985; Gentry and Gentry 1978; Harris 1991; K appelman 1991, 1984; Kappelman et al. 1997; Klein et al. 2010; Kovarovic and Andrews 2011, 2007 ; Musi ba 1999; Musiba and Magori 2003 ; Musiba et al. 2007; Plummer and Bishop 1994; Plummer et al. 2008; Shipman and Harris 1988; Spencer 1997; Su and Harrison 2 007; Vrba 1985, 1980 ) Bovids are used quite frequently in paleoecological modeling because they are sensitive indicators of environments (Vrba 1976). This is because bovids practice resource partitioning (Spencer 1997). This means that bovids are speciali sts occupying distinct niches, so even subtle environmental changes should be reflected in bovids. An assumption present in any study that uses bovids as environmental indicators is that fossil bovids were similarly sensitive to environment changes as extant bovid tribes are today (Spencer 1997). However, this assumption should not be a problem for studies that construct paleoecological models for at least the last 4 million years because the bovid tribes living today were established by the Pliocene (K ingdon 1982). Biases in Paleoecological Models For all paleoecological models it is important to keep in mind that there are biases embedded in these and that they can never be one hundred percent accurate. One of the most important reasons that all pal eoecological models are flawed is the biased nature of the fossil record. For example, there is a bias toward animals adapted to lacustrine and alluvial settings (Domingeuz Rodrigo and Musiba 2010). This means that animals adapted to those habitats will be overrepresented in the fossil record, while
9 animals adapted to other habitats will be underrepresented. This is simply the nature of the fossil record, and it is very important to bear in mind when creating paleoecological models Another reason that al l paleoecological models are flawed is taphonomic processes, which are processes that a lter preservation of fossils Processes such as hydraulic transport, trampling, weathering, and biochemical alterations are among just a few of the processes that can affe ct the preservation quality of fossil assemblage s Taphonomic processes may cause fossils to move from original death locations, which may result in the loss of small bone elements. These processes may also cause we athering to such a degree that fossil s disintegrate Numerous researchers have highlighted the importance of thoroughly studying the taphonomic processes that affect fossil hominin sites ( Dominguez Rodrigo and Musiba 2010, Kidwell and Holland 2002 Kingston 2007 ). Without understanding the pr ocesses that have further biased the fossil record, interpretations are likely to be incorrect. Likewise, a similar reason why paleoecological models are flawed is because of time averaging. Time averaging has to do with the amount of time in which a pa rticular assemblage becomes buried. If rapid burial occurs, there will be no time for animals to become buried later in the same sediments and thus the assemblage has little time averaging However, most assemblages accumulate more slowly, and these can b e significantly time averaged. Assembla ges that accumulate for a period of hundreds or thousands of years can mask changes in environment or faunal composition at smaller resolutions (Kidwell and Holland 2002).
10 For example highly time averaged assemblage s will not reflect seasonal migrations (Dominguez Rodrigo and Musiba 2010). Many animals live in one habitat during the wet season and another in the dry season. S easonal migrations are difficult to discern in an assemblage that is time averaged over a per iod of hundreds or thousands of years. Moreover, highly time averaged assemblages will mask short term environmental changes. For example, Dominguez Rodrigo and Musiba (2010) point out that the Serengeti Mara area was first described in the 1800's as open grassland. In the early 1900's, the area became dense woodland and a few decades later became open grassland again. These quick environmental changes were due to changes in the density of certain animal communities and the presence or absence of wildfires. These quick environmental changes would be completely masked in highly time averaged assemblages. Furthermore, yet another reason why paleoecological models are flawed is because of collection bias. Collection bias exists because there is no standard col lection protocol for paleoanthropologists in the field ( Alemseged et al. 2007). As a result, researchers follow the guidelines of team leaders who decide what is and what is not important to collect (Behrensmeyer et al. 2007). Even teams that strive to col lect every fossil will miss many. Experimental seeding studies such as Wandsnider and Camilli (1992) and Burger et al. (2004) show that even intensive surveys will tend to miss low density items. Some researchers have highlighted the pervasiveness of colle ction bias by showing how it prevents robust comparative studies of faunal assemblages from different localities, even if those are within the same site ( Alsemseged et al. 2007; Eck 2007). An issue to bear in mind when creating paleoecological models is related to data scale. One way in which researchers create paleoecological models is by synt hesizing
11 lines of evidence that use different method s While this technique is useful, it also must be used with caution. Data obtain from marine and some terre strial sediments indicate global climatic processes that are affected by Milankovitch cycle s including the eccentricity of Earth's orbit around the Sun obliquity of the tilt of Earth's axis, and precession of Earth's axis. These data examine processes oc curring at the global scale. Data obtained from stable isotopes analyses of fossil herbivore teeth or of phytoliths provide evidence for processes occurring on local or habitat scales (Kingston 2007). It can be useful to synthesize data from local and glob al scales; however, this process tends to detract from the rigor of the original data, and these syntheses tend to be based more on presumptions than hard evidence (Kingston 2007). Yet, despite the biases and flaws embedded in paleoecological models rese archers in the field of paleoanthropology continue creating them. This is because paleoecological models are the only way to understand the selective forces guiding hominin evolution. Therefore, in order to make the most accurate paleoecological models pos sible, researchers need to bear in mind how these issues may affect interpretations of their data.
12 CHAPTER II PALEOECOLOGICAL MODELS OF EAST AFRICA & LAETOLI Despite biases embedded in paleoecological models they are important for understanding the selective forces that influenced hominin evolution. The methods that are used to create these models provide details of hominin environment interactions on the global or regional scale or can provide detail s of interaction s at the local or habitat scale. W hen researchers are interested in global or regional scale interactions they most commonly employ stable isotope analyses of marine and terrestrial sediments ( Cerling and Hay 1986, deMenocal 2007, Norton Griffiths 1975, Potts 2007, Quade and Levin 2013, S ikes 1994). When researchers are interested in local scale interactions they commonly employ methods such as micro and macrobotanical analyses (Bamford 2011; Bonnefille 2010, 1984; Bonnefille and Riollet 1987, 1980; Rossouw and Scott 2011), stable isotop e analyses (Kingston 2011) and faunal assemblage analyses ( Bishop 20 11; Bishop et al. 2011; DeGusta and Vrba 2003; Denys 2011; Frost 2007; Harrison 2011b, 2011c; Hernesniemi et al. 2011; Kappelman 1991; Klein et al. 2010; Krell and Schawaller 2011; Lewis and Werndelin 2007; Musiba and Magori 2003; Plummer and Bish op 1994; Plummer et al. 2008; Reed 2007; Su and Harrison 2007; Werdelin and Dehghani 2011) Micro & Macrobotanical Analyses : Polle n, Phytoliths, and Macrofossils Analyses of micro and macrobo tanical remains can provide details about environments and habitats present at early hominin sites. Fossil pollen analyses allow researchers to examine grains of pollen deposited into soil when the stratigraphic layers were created In general, fossil poll en analyses reflect a larger area than macrobotanical
13 remains because they may be carried long distances with the wind (deMenocal 2007). Bonnefille and colleagues have used this method to determine East African paleoenvironme nts ( Bonnefille 2010, 1984; Bon nefille and Riollet 1987, 1980 ). In a study examining Laetoli fossil pollen Bonnefille and Riollet (1987) conclude that the data indicate a paleoenvironment similar to that of t he present day Serengeti Plains, which is considered open grassland. However other studies have criticized Bonnefille and Riollet's conclusions. For instance, Musiba et al (2007) point out their study ignores a significant amount of unidentified pollen. Musiba and Magori (2003) argue that the unidentified pollen probably indicat es Laetoli was a mosaic environment with more diversity of habitats than the present day Serengeti Plains. Yet, a later study by Bonnefille (2010) maintains that the fo ssil pollen evidence indicates the area was dry open grassland. A r ecent phytolith an alysis indicate s shift s from mostly C 3 plants in the lower part of the Upper Laetolil Beds to more C 4 plants in the upper part of the Upper Laetolil Beds (Rossouw and Scott 2011). C 3 plants use the most common photosynthetic pathway, which incorporates car bon dioxide into a three carbon compound. C 3 plants are the most common plants around the world and are best adapted to wet conditions. C 4 plants are plants that use a different photosynthetic pathway and modifications to the photosynthetic pathway allow these plants to thrive in drier conditions. Therefore many C 3 plants indicate wetter environment s while many C 4 plants indicate drier environment s With this in mind, t he analysis of phytoliths at Laetoli (Rossouw and Scott 2011) indicate s that the Upper Laetolil Beds was a time with a combination of C 3 and C 4 plants with wetter conditions in the lower part of the Upper Laetolil Beds and drier conditions in
14 the upper part of the Upper Laetolil Beds Thus, phytoliths at Laetoli show increasing aridity bet ween 3.85 3.63 Ma. While fossil pollen and other microbotanical remains generally indicate a wider environmental context, macrobotanical remains reflect local habitats since these elements are less likely to be deposited long distances from their original locations (deMenocal 2007). Macrobotanical remains from t he Upper Laetolil Beds indicate a complex, mosaic environment (Bamford 2011) Fossil leaves, f ruits, and seeds indicate forested habitats and dry open woodlands. Overall, the micro and macro botanical evidence indi cates that Laetoli was either grassland similar to the present day Serengeti Plains, or grassland with pockets of wooded and forested habitats Stable Isotope Analyses Stable isotope analyses allow researchers to discern details ab out paleoenvironments ranging from cycles of warm cool and wet dry to the kinds of plants that dominant an area. Examinations of c arbon i sotopes allow researchers to determine the types of foods an organism ate during its lifetime. This is because carbon e xists in nature in a few forms (i.e. isotopes). All of the forms have the same number of protons, but the number of neutrons varies based on the climate in which the plant g rows. Therefore if researchers can determine the ratio of one carbon isotope to ano ther, a probable environment can be determined. The process is similar for oxygen isotope analyses. However, instead of indicating which plants were most abundant, oxygen isotopes indicate general warming or cooling cycles and wetter or drier cycles based on ocean temperatures (Potts 2007). This is based on the chemical and physical reactions of water vapor with warm, cool, wet, or dry air.
15 Using these principles, s ome stable isotope analyses exami ne the effect of global climate on h ominin evolution in Ea st Africa Samples for global or regional analyses generally include ice cores, marine sediments, and some large lakebed sediments. These data reflect global scale hominin environment interactions (Kingston 2007). Regional stable isotope analyses of East A frica find the East African Pliocene had similar seasonality as today (Cerling and Hay 1986, Norton Griffiths 1975) but also environmental instability due to orbital variations (deMenocal 2007). These orbital variations are related to Milankovitch cycle s where changes in Earth's eccentricity, obliquity, and precession move in predictable cycles. Moreover, a trend of increasing aridity throughout the Pliocene is apparent in the marine sediments from East Africa (deMenocal 2007) However, the trend toward increasing aridity in East Africa was not completely permanent Short term climatic fluctuations resulted in quick environmental changes (deMenocal 2007 ). Furthermore, despite their geographic proximity, stable isotope analyse s of sediments from Ethiopia, Kenya and Tanzania find that those regions were influenced by different climatic cycles during the Miocene (approx 23 5 Ma) (Quade and Levin 2013) The expansion of C 4 grasses in Kenya and Ethiopia also indi cates increasing ar idity in the transition from the Miocene to the Pliocene (Quade and Levin 2013) Yet, not all stable isotope analyses look at whole regions. Some examine Laetoli to obtain data that is reflective of local environments and habitats. Stable isotope anal yse s of fossil herbivore teeth show an environment with both C 3 and C 4 plants present (Kingston 2011). This could indicate either grassland, woodland or forested habitat s These results show a C 3 to C 4 ratio generally seen in large herbivores indicating either
16 mixed grazing and browsing feeding strategies or heavy reliance on C 3 browsing (Kingston 2011). This provide s evidence that the environment was probably more wooded since C 3 browsing feeding strategies tend to occur in wetter, more woo ded habitats However, this could also mean that the habitats and niches present have no modern analogue s (Kingston 2011). Importantly, though these data also show an increase in the number of grazing bovids in the transition from the Upper Laetolil Beds to the Ndolanya Beds (F igure 1. 3) Though these data indicate an increase in grasslands and thus increasing aridity over time, this could also be a product of taphonomic bias es in the collection. All in all these studies show that during the Pliocene the East African environment was in constant flux. There was environmental instability due to orbital variations that affec t cycles of wet dry and warm cool (deMenocal 2007), but the seasonality was probably similar to today ( Cerling and Hay 1986, Norton G riffiths 1975). These studies also show that the environment at Laetoli at this time was lightly wooded but contained some element of C 4 plants which are indicative of the presence of grassland. Therefore, overall East Africa and Laetoli in the Pliocene were experiencing increasing aridity. Faunal Assemblage Analyses One of the more common method s to create paleoecological models are analyses of vertebrate faunal assemblages recovered from archaeolo gical and paleoanthropological sites. That is the method that this study uses. Researchers can choose to complete faunal analyses that are taxomonic or taxon free. Taxonomic methods analyze species abundance, diversity and turnover patterns. Taxon free methods, such as
17 ecomorphology, analyze diet, locomotion, favored terrain, and habitat diversity ( Reed et al 2013). Taxonomic Methods Taxonomic faunal assemblage analyses estimate pr obable environment or habitat based on species diversity abundance or turnover patterns It, therefore, relies on classifying fauna to the generic or species level. In East Africa, m ost taxonomic faunal analyses use mammal ian remains Commonly, researchers use remains of primates (Frost 2007 ; Harrison 2011b 2011c ), carnivores (Lewis and Werndelin 2007, Su a nd Harrison 2007 Werdelin and Dehghani 2011 ), bovids (Bobe et al 2007, Gentry 2011, Musiba et al. 2007, Su and Harrison 2007), suids (Bishop 2011), rhinoceros (Hernesniemi et al. 2011), and micromammals (Denys 2011; Krell and Schawaller 2011; Reed 2011, 2007 ; Reed and Denys 2011 ). Some s tudies that taxonomically analyze mammalian fauna from the Upper Laetolil Beds look at community structure One such study finds that the mammalian community structure is stable during the Upper Laetolil Beds and that th e environment is mosaic (Su and Harrison 2007). Based on this evidence, it appears that local ecological pressures at Laetoli had a greater influence over the composition of local fauna th an global or regional pressures since global climatic fluctuations did not change the faunal composition. Similarly speciation and extinction of primates in East Africa does not coincide with global climatic shifts (Frost 2007) The underrepresentation of cercopithecids in the lower po rtion of the Upper Laetolil Beds compared to the upper portion indicates changes in the open to closed aspect of the environment (Harrison 2011b ) although this could
18 also be due to taphonomic biases The presence of galagidae ( Harrison 2011c ) indicates th ere were habitats present that contained some trees or bushes, providing more evidence for wooded or forested habitats at Laetoli. Thus taxonomic primate analyses of the Upper Laetolil Beds provide evidence for an environment in flux with some wooded or f orested habitats Instead of primate diversity, abundance, and community structure, some researchers examine carnivore community structure The community structure of carnivores in East Africa indicates speciation events not associated with climatic shift s (Lewis and Werdelin 2007) which is similar to the community structure of primates (Frost 2007) C arnivores at Laetoli show continuity in community structure through the Upper Laetolil Beds and the Ndolanya Beds ( F igure 1. 3) (Wedelin and Dehghani 2011) This indicates the selective pressures on carnivores were relatively stable. T his doe s not mean that the environment was stable because carnivore community structure may stay relatively stable during environment fluctuations Thus the evidence from carniv ore assemblages does not aid in determining what environments were present at Pliocene Laetoli. On the other hand, taxonomic analyses of bovids do provide details about environments present at Laetoli and East Africa. B ased on species abundance and diver sity of bovids at Turkana, Kenya and Hadar, Ethiopia it is apparent that the environment in those region s became drier after 3.4 Ma (Bobe et al 2007) At Laetoli, bovid species abundance and diversity indicates neither open grassland nor woodland forest which means there was a complex, mosaic environment present (Musiba et al. 2007) Moreover species diversity and abundance of bovids at Laetoli is significantly
19 different from species abundance and diversity of bovids at Shungura Ethiopia, Koobi Fori, K enya, and even the geographically close Olduvai Gorge Tanzania (Gentry 2011) This ties in well with the regional stable isotope evidence indicating that Tanzania, Kenya, and Ethiopia were being influenced by different climate cycles despite their geograp hic proximity (Quade and Levin 2013). More t axonomic bovid analyses show increasing aridity throughout the Upper Laetolil Beds and that Laetoli was mosaic with open and wooded habitats, including dense woodlands and riverine forest (Su 2011) Thus, the bov id evidence shows increasing aridity during the Pliocene and the presence of a complex, mosaic environment at Laetoli. Primates, carnivores, and bovids are not the only fauna that researchers use to create paleoecological models. Sometimes suids (pigs) are used as environmental proxies. Suid species diversity and abundance at Laetoli indicates that there were multiple different habitats (Bishop 2011) Additionally, sometimes rhinocerotidae (rhinoceros) species diversity and abundance are used as environm ental proxies. Rhinoceros at Laetoli show an increase in grazing feeding strategies (Hernesniemi et al. 2011). This provides compelling evidence for an expansion of grasslands during the Pliocene, since grazers primarily inhabit grasslands. Overall, the su id and rhinoceros remains indicate similar environmental pressures as the bovids remains. They all show an increase in aridity during the Pliocene and the presence of a complex, mosaic environment at Laetoli. Large mammals such as primates, carnivores, bo vids, suids, and rhinoceros are useful environmental proxies. But some researchers also recognize the utility of micromammal remains as environmental proxies Micromammals provide habitat specific data since the boundaries of their habitats are more constr ained than large
20 mammals (Denys 2011). Micromammal community structure throughout East Africa indicates drying trends throughout the Pliocene (Reed 2007). The structure of micromammal genera at Laetoli indicates habitats similar to the present day Serenget i Plains (Reed 2011 Reed and Denys 2011 ). In addition, beetle species diversity and abundance indicates habitats at Laetoli that were primarily grasslan d (Krell and Schawaller 2011). Interestingly micromammals species abundance and diversity at Laetoli i s distinct from micromammal species abundance and diversity at Hadar and Adu Asa, Ethiopia (Denys 2011). This is further evidence that fits well with the stable isotope analysis that found that Tanzania, Kenya, and Ethiopia were influenced by different cli matic cycles despite their geographic proximity (Quade and Levin 2013). Thus micromammal species abundance and diversity indicate s increasing aridity an d grasslands throughout East Africa during the Pliocene and that Laetoli was similar to grassland envir onments like the present day Serengeti Plains. In addition to examining species diversity and abundance of faunal assemblages, sometimes researchers examine dental mesowear in species to determine feeding strategies. Dental mesowear examines wear patterns and abrasions on teeth, which indicates feeding strategies in fossil species because different feeding strategies leave distinc t wear and abrasion patterns The dental mesowear of mammals from Laetoli shows an absence of grazers and a large number of browsers during the Pliocene (Kaiser 2011 ). A large number of grazers generally indicates a predominance of grasslands so the absence of grazers may imply that the environment was not predomina ntly grassland, although there could be taphonomic biases affe cting the preservation of grazers. Dental mesowear of mammals at Laetoli also shows decreasing diversity of feeding niches from
21 the Upper Laetolil Beds to the Ndolanya Beds ( Kaiser 2011 ), which provides evidence that a diverse and mosaic environment during the Upper Lae toli Beds became less diverse and less mosaic at the Plio Pleistocene transition The d ental mesowear data also shows change s in feeding strategy in the bovid tribes Alcelaphini and Hippotragini (Kaiser 2011). These data show that those trib es practiced an intermediate feeding strategy of grazing and browsing in the Pliocene while their living relatives are specialized grazers. This illustrates an extremely important assumption found in all taxonomic faunal analyses. This assumption is that the species in the fossil record practice the same behaviors as their counterparts living today. The issue with this assumption is that evolution is a continuous cycle so changing environmental pr essures over millions of years may mean that there were fos sil communities in the past that have no modern analogues (Plummer et al. 2008) The dental mesowear of bovid tribes at Laetoli illustrates how feeding strategies have changed since the Plioc e ne and, therefore how other behaviors may have also changed Thus, taxonomic faunal analyses cannot get around that fundamental weakness. Another fundamental weakness of taxonomic faunal analyses is that errors are introduced into the data because fauna are identified to the generic or species level. Morphological v ariations are fairly apparent at the family level; however, those variations in morphology become more subtle at the generic and species level Thus, generic and species level morphologies are easier to misidentify, and errors may be introduced into analys es To combat these weaknesses, some researchers choose to use taxon free methods.
22 Taxon Free Methods Ecomorphology is a commonly employed taxon free method It is sometimes referred to as functional morphology, and it predicts that the morphological characters of an organism will reflect the habitat in which that organi sm lives. DeGusta and Vrba (2005, 2003) demonstrate that bovid postcrania l morphology reflects the terrain and habitats in which bovid s live This is because bovids f rom different habitats have adaptations that are reflective of different locomotor strategies for predator avoidance. For example, bovid s with adaptations to open habitats have postcranial morphologies that propel them in a straight and fast manner across their habitats On the other hand, bovid s with adaptations to forested habitats have postcr anial morphologies enabling quick direction changes to maneuver around trees. Although e c omorphology is not a new method, t here have been many ecomorphological studies completed recently. Some analyze the functional morphology of bovid crania l elements to determine dietary strategies (Spencer 1997) Bovid f emora and metapodials can also be used as environmental proxies, and th ese postcranial elements from Laetoli provide support for the environment being mosaic or e cotone (Musiba and Magori 2003) Other studies focusing on identifying past environments have used bovid femora (Kappelman 1991, Musiba and Magori 2003), metapodials ( Klein et al 2010, Musiba and Magori 2003 Plummer and Bishop 1994 ), astragali ( Bishop et al. 2011, DeGusta and Vrba 2003, Plummer et al 2008), phalanges ( Bishop et al. 2011, DeGusta and Vrba 2005), radii (Bishop et al 2011), and finally the calcanei o f cervids, which are a sister taxa of bovids (Curran 2012).
23 Ecomorphological analyse s of bovid postcranial elements at Laetoli indicate woodland and forest habitats (Bishop et al 2011). In addition, the environment at the time of the Upper Ndolanya Beds ( Figure 1.3) was more open than the environment during the Upper Laetolil Beds and was an environment that shift ed between bushland, woodland, and grassland (Kovaric and Andrews 2011) Thus the ecomorphological analyses of Laetoli provide evidence that the environm ent was complex, contained forest or woodland habitats and also might show short term environmental changes during the Pliocene. Taxon free methods are informative of many facets of paleo environment s Yet, a n important weakness of taxon free methods such as ecomorphology is that analyses can only indicate a range of behaviors that an animal may have practiced (Reed et al 2013). This means that there are overlapping behaviors that may be impossible to discern from one another Anot her important weakness of taxon free methods such as ecomorphology is that morphology tends to respond slower to changes in ecological pressures than behavior or physiology (Reed et al 2013). Therefore, rapidly changing environmental conditions may not be directly reflected in t he hard tissues that fossilize. Despite these two weaknesses, ecomorphology is a robust and commonly used method that addresses the weaknesses of taxonomic methods. Therefore, this study uses ecomorphology to determine the paleoenvi ronment of Laetoli at 3.85 3.63 Ma.
24 CHAPTER III MATERIAL S & METHODS The fossil bo vid astragal i collection (n=109 ) has been obtained from Dr. Charles Musiba at the University of Colorado Denver This collection was recovered during the 2010 2013 field seasons from localities 7 8 and 9 All astragali in the fossil collection are from the Upper Laetolil Beds (3.85 3.63 Ma) which makes this study a fairly temporally specific habitat re construction. Additionally, a comparative collection of extant African bovid astragali (n=177 ) has been obtained from the Smithsonian Museum of Natural History. Table 3 1: Taxonomy and Habitat Categories of Modern Bovid Collection Subfamily Tribe Species N Habitat Alcelaphinae Aepycerotini Aepyceros melampus (Impala) 5 Open Alcelaphini Alcelaphus buselaphus (Hartebeest) 7 Open Connochaetes taurinus (Brindled Gnu) 3 Open Damaliscus lunatus (Tiang) 12 Open Damaliscus pygargus (Bontebok) 13 Open Antilopinae Antilopini Antidorcas marsupialis (Springbuck) 2 Intermediate
25 Table 3 .1 continued: Taxonomy and Habitat Categories of Modern Bovid Collection Subfamily Tribe Species N Habitat Gazella dorcas (Dorcas Gazelle) 4 Open Gazella granti (Grant's Gazelle) 13 Open Gazella rufrifrons /thomsoni (Thompson's Gazelle) 18 Open Litocranius walleri (Gerenuk) 7 Habitat Neotragini Madoqua kirkii (Kirk's Dikdik) 13 Intermediate Neotragus moschatus (Suni) 1 Forest Neotragus pygmaeus (Royal Antelope) 1 Forest Ourebia ourebia (Oribi) 12 Open Raphicerus campestris (Steinbuck) 5 Open Raphicerus sharpei (Sharpe's Grysbok) 4 Intermediate Bovinae Tragelaphini Tragelaphus euryceros (Bongo) 2 Intermediate Tragelaphus scriptus (Bushbuck) 7 Forest Tragelaphus spekei (Sitatunga) 2 Intermediate Tragelaphus strepiceros (Gre a ter Kudu) 3 Intermediate
26 Table 3 .1 continue d: Taxonomy and Habitat Categories of Modern Bovid Collection Subfamily Tribe Species N Habitat Hippotraginae Hippotragini Oryx gazella (Southern Oryx) 4 Open Reduncini Kobus ellipsiprymnus (Waterbuck) 10 Intermediate Kobus kob (Kob) 13 Intermediate Kobus leche (Lechwe) 1 Intermediate Kobus megaceros (Nile Lechwe) 1 Intermediate Kobus vardoni (Puku) 1 Intermediate Redunca arundinum (Southern Reedbuck) 3 Intermediate Redunca fulvorufula (Mountain Reedbuck) 4 Intermediate Redunca redunca (Bohor Reedbuck) 9 Open Linear Me asurements There are eight linear measurements on the astragali that have been taken on each specimen (Figure 3.1) The measurements have been chosen carefully to depict size and shape variation and are adopted from DeGusta and Vrba ( 2003) Six ratios have also been create d from the linear measurements to better depict shap e variation instead of size variation All measurements were taken in millimeters with Mitutoyo digit al calipers with a .01 mm error and all were taken by the author to avoid inter observer error.
27 The measurement on the medial side is: Medial Length (LM): maximum proximal distal length The measurements on the lateral side are: Lateral Length (LL): maximum proximal distal length Distal Thickness (TD): anterior posterior width on distal end Intermediate Thickness (TI): minimum anterio r posterior width of the intermediate surface between the proximal and distal ends Proximal Thickness (TP): anterior posterior width of proximal end Finally, the measurements on the anterior side are: Intermediate Length (LI): minimum proximal distal length. Distal Width (WD): medial lateral width on distal most end Intermediate Width (WI): minimum medial lateral width on intermediate surface, including any tubercles or projections Additionally, the following ratios were used in this study: LL/TI, LL/TP, LI/WD, LI/WI, LM/WI, and LM/WD. Although this study does not include class five bovids, which are bovids weighing more than 250 kg, allometry may be a factor skewing the results of an ecomorphological study (DeGusta and Vrba 2 003, Plummer et al. 2008). Thus, ratios are used to help depict shape variation in the collections, not size variation Figure 3.1 : Astragali Linear Measurements (taken from Charles Musiba's personal archive)
28 Statistics After completing the measurements of the fossil collection at the University of Colorado Denver paleoanthropology lab and completing measurements of the modern collection at the Smithsonian Museum of Natural History, I input the data into JMP w here explor atory data analyses and a principle component analysis have been completed to investigate the distribution and structure of the data. Then a discriminant function analysis has been performed which predicts the habitats bovids inhabit based on the morphology of their astragali. Completing the discriminant function analysis on the modern collection produces a misclassification rate This, in turn, allows the researcher to predict the accuracy of the discriminant function analysis that is completed on the fossil collection The discrim in ant function analysis uses the canonical discriminant functions that it created Habitat Categories The habitat categories that this study uses are adopted from Kappelman et al (1997), DeGusta and Vrba (2003), and P lummer et al (2008). The difference between this study's habitat categories and the others is that this study uses three habitat categories while the others use four The habitat categories that this study uses are as follows: Open open country, plains arid country, edge or ecotone areas Intermediate bushland, tall grass, woodland, swamp, hilly areas, and riverine or near water habitats Closed forested habitats
29 The reason that this study uses only one intermediate category instead of light cover an d heavy cover like Kappelman et al (1997), DeGusta and Vrba (2003), and Plummer et al (2008) is because there tends to be a significant overlap between the light cover and heavy cover categories (Musiba 1999). Confounding F actors All bovid astragali, both extant and fossil, were excluded from this study if they belong to class five bovid s, which are bovids weighing over 250 kg DeGusta and Vrba (2003) note that heavy body weight and large body size can be a confounding factor in ec omorphological studies Some variability in the morphological dimensions of the astragali is due to heavy bod y weight or large body size ; therefore, there is chance that if the heaviest and largest bovids were included in a discriminant function analysis the results would reflect body weight or size not adaptations to habitat (DeGusta and Vrba 2003)
30 CHAPTER IV RESULTS Before completing any statistical analyses I check ed my data for errors in codes and measurements. For the modern collection, I obtained the habitat categories for species from DeGusta and Vrba (2003) Kingdon (2004), and Plummer et al (2008). Importantly, there were discrepancies between some sources in the habitat category to which Ourebia ourebia and Madoqua kirkii belong DeGu sta and Vrba (2003) consider Ourebia o urebia to live in an intermediate habitat while Kingdon (2004) writes that this species lives primarily in grasslands, preferring flat areas with short grass. I have given preference to Kingdon's habitat categories wh en discrepancies between my sources arose, so I changed Ourebia o urebia to the open habitat category. In addition, DeGusta and Vrba (2003) consider Madoqua kirkii a forest dwelling bovid, but Kingdon (2004) describes Madoqua kirkii as occupying distinct re gions of bushland. Since bushland falls into the intermediate habitat category I changed Madoqua kirkii to the intermediate category. I also checked the data for measurements errors and took out one Alcelaphus buselaphus specimen because of a measurement error. Descriptive Statistics Next, I completed descriptive statistics to better visualize the distribution of the data. Table 4.1 summaries the means of each habitat category. Various multivariate statistics have been completed to determine whether there is a significant difference between the means of the habitat categories. The null hypothesis is that there is no difference between the means The results of these tests are summarized in table 4.2. T ogether, tables 4.1 and 4.2 show that there is a statistically significant difference among
31 the three habitat categories. That means that the variables show a real difference in the morphology of bovid astragali for those that are adapted to open, intermed iate, and closed habitats. Table 4.1: Means of Habitat C ategory by V ariable Variable Mean of Open Mean of Intermediate Mean of Closed LM 32.69 36.39 32.76 LL 35.15 38.97 35.08 TD 15.63 48.52 14.51 TI 17.90 19.49 17.41 TP 13.13 14.26 13.03 LI 27.30 31.24 25.83 WI 20.93 22.17 20.51 WD 20.95 23.30 19.83 LL/TI 197.26 200.95 200.82 LL/TP 271.89 279.96 269.27 LI/WD 131.92 136.57 132.51 LI/WI 132.26 143.31 132.88 LM/WI 157.88 167.29 163.29 LM/WD 157.46 159.43 166.07 Total n 112 57 8
32 Table 4.2: Multivariate S tatistics Showing the Means of the Habitat Categories are Significantly Different Test Value Approx. F Num DF Den DF Pr > F Wilks' lambda 0.4327 5.9818 28 322 <.0001 Pillai's Trace 0.6627 5.7354 28 324 <.0001 Hotelling Lawley 1.0901 6.2349 28 279.74 <.0001 Principal Components Analysis Next I also completed a principal component analysis ( PCA ) A PCA generates components and indicates which variables are most important in the study First I completed a PCA on both the linear variables and the ratio variables. The linear variables yie ld 2 components The first component explains 86% of the variability, whil e the second component explains 12% of the variability. For the PCA with the ratio variables, after rotation the fir st component accounts for 47% of the variability and the second component accou nts for 33% of the variability. Then I completed a PCA on all of the variables together. This also yields two components, the first of which explains 64.6% of the variability in the data set and the second of which explains 14.2% of the variability (Table 4.3 ) Each PCA indicates that all of the variables are important to the analysis. The loading plot of this PC A is shown in Figure 4.1 Table 4 .3 : Princ ipal Component Analysis of All V ariables Component 1 Component 2 Eigenvalue 9.0477 1.9842 % Variability explained 64.6% 14.2%
33 Figur e 4 .1 : Loading Plot of P rincipal C omponents A nalysis D iscriminant F unction A nalysis Modern Astragali Collection The first step to completing the discriminant function analysis (DFA) of the modern astragali collection is to run each variable separately in the DFA This in combination with the results of the principal components analysis indicate s the importance of each variable in the study Misclassification rates range from 34% to 67% (T able 4 4 ).
34 Table 4 .4 : Results of DFA on Individual V ariable s Variables Percent Misclassified LM 64.48 LL 62.3 TP 67.76 TI 65.03 TD 61.75 LI 59.02 WI 63.39 WD 66.12 LL/TI 63.93 LL/TP 67.21 LI/WD 48.63 LI/WI 34.97 LM/WI 40.44 LM/WD 54.64 After completing the DFA on single variables, I then completed a DFA on the linear variables and anoth er DFA on the ratio variables (T able 4 5 ).
35 Table 4 .5 : Results of DFA on Grouped V ariables Percent Misclassified Eigenvalue % Explained Variability Linear V ariables 27.87 Factor 1: 0.2333* Factor 2: 0.0902 Factor 1: 72.10* Factor 2: 27.89 Ratio V ariables 24.59 Factor 1: 0.3499 Factor 2: 0.1326 Factor 1: 72.51* Factor 2: 27.48 i ndicates statistical significance at the 0 .05 level Next, I completed a DFA with all of the variables (Table 4.6 ). I chose to do this because the PCA indicate s that all of the variables are important for explaining the variation in the data. T he classification rat e is approximately 82%, which is very high Additionally, the first factor is statistically significant at the 0.05 level. Table 4 .6 : Results of DFA with All V ariables I ncluded Percent Misclassified Eigenvalue % Explained Variability 12.43 % Factor 1: 0.821 Factor 2: 0. 268 Factor 1: 75.34 Factor 2: 24.65 indicates statistically significant at the 0.05 level Table 4.7 shows the summary table of the DFA with all variables included in the analysis It shows the total number of specimens per habitat ca tegory for the modern collection and shows which habitat category the DFA predicts the modern specimen s to fall based on the measurements. In some cases, the specimens are predicted to fall in their correct habitat category. Other times, the specimens are incorrectly classified.
36 Table 4.7 : Summary Table of Modern C ollection DFA W ith All Variables I ncluded Predicted Group Open Intermediate Closed % Correct Total Group Open 112 0 0 100 112 Actual Intermediate 16 41 0 71.9 57 Closed 6 0 2 25 8 A closer examination of the species in the modern collection that are misclassified by t he DFA reveals more details about the accuracy of the model. Misclassifications Table 4.8 summarizes the species that are misclassified by the DFA There are only ei ght species in the modern assemblage that are misclassified, and only 23 out of 177 specimens are misclassified. Some of the misclassifications are easily explained. For example, Connochates tarinus and Tragelaphus strepsiceros are very close to the weight cut off that this study employs Class five bovids, those that weigh above 250 kg, are excluded from this study but the weight range for males and females of many species can be significantly different. Connochates tarinus males can weigh between 165 290 kg and females can weigh between 140 260 kg (Kingdon 2004). Tragelaphus strepsiceros males can weigh between 190 315 kg and females can weigh between 120 215 kg (Kingdon 2004). The data at the Smithsonian, where the modern collection was measured, did n ot indicate the weight or sex of the specimens, which could mean that some of the
37 specimens belonging to Connochates tarinus and Tragelaphus strepsiceros could weigh more than 250 kg Thus, their astragali morphologies could be reflective of heavy body wei ght and not adaptations to habitat. T he other misclassifications are of species that fall significantly below the weight cut off. Therefore, the next area to examine is the habitat within which the misclassified species live Many bovids move throughout a few different habitats, so while the habitat categories strive to be as accurate as possible, they are necessarily generalized and simplified versions of reality. Because of this oversimplification, a species may be correctly assigned to the open categor y while retaining morphological characters that are r epresentative of an intermediate habitat That scenario make s predicting the habitat of certain bovids more difficult than others. Kobus kob is a good example of this. This species lives in plains and r olling hills that are near water sources (Kingdon 2004). DeGusta and Vrba (2003) Plummer et al (2008) and this model include Kobus kob in the intermediate habitat category Yet, this species is one that lives within and contains morphological adaptations to both flat and hilly areas. Therefore, it seems reasonable that just under half of the specimens belonging to this species in the modern collection are misclassified ( Table 4.8 ) For similar reasons Kobus megaceros is also misclassified. DeGusta and Vrba ( 2003), Plummer et al (2008) and this study include Kobus megaceros in the intermediate category. The discrimi nant function analysis predicts this species inhabi a t s open habitats This discrepancy also makes sense because Kobus megaceros lives in grasslands that are situated between swamps and flooded grasslands (Kingdon 2004). While this species primarily lives in an intermediate habitat, it clearly spends time in
38 grasslands. Therefore, it is reasonable that this species a lso has morphological characters that are adapted to either or both habitat s Another misclassified species is Redunca arundinum This species is classified in DeGusta and Vrba (2003), Plummer et al (2008), and in this study as living in an intermediate habitat Redunca arundinum lives in grasslands, woodlands, and areas with scrubs (Kingdon 2004) A ll of the specimens belonging to this species (n=3) are incorrectly classified as belonging to the open habitat category (Table 4.8 ) Again, it is reasonable that a bovid that live s within a wide range of habitats also contain s morphological characters are reflective of other habitats, too. The other species that are misclassified are slightly more complex to analyze. Redunca fulvorufula which is the mountain reedbuck, inhabits grassy, flat areas within rocky or mountainous terrain (Kingdon 2004). All of the specimens belonging to this species (n=4) are incorrectly classified as belonging to the open habitat category (Table 4.8) DeGusta and Vrba (2003) Plummer et al (2008) and this study include Redunca fulvorufula in the intermediate habitat category. This discrepancy can be explained because despite living in hilly or mountainous habitats Redunca fulvorufula prefers the flatter and grassier areas within these habitats. Therefore, the morphological characters that the discriminant function analy sis recognizes make this species fit more closely with the open category, despite definitively belonging to the intermediate category. Another miscla ssification in the modern collection is Tragelaphus spekei This species lives in shrub filled areas that are close to swamps or marshes (Kingdon 2004). DeGusta and Vrba (2 003), Plummer et al. (2008) and this study include this species in the intermediate category since swamps are part in the intermediate habitat category. Yet,
39 Tragelaphus spekei is an interesting species because it spends much of its time in the swamp, unlike any of the other species included in the modern collection With this in mind, i t become important to remember the fundamental principle of ecomorphology, which is that bovid postcranial elements reflect the terrain and the substrate in which the bovid lives (Plummer et al 2008). For a semiaquatic bovid like Tragelaphus spekei that means that the morphological characters of the astragali are in fact more representative of a watery habitat rather than an open versus closed habitat like the other bovids included in this study The final misclassification is Tragelaphus scriptus This species lives in thick cover, usually around the edges of forests, although at night it move s into open areas to find food (Kingdon 2004). DeGusta and Vrba (2003), Plummer et al (2008) and this study include Tragelaphus scriptus in the closed habitat cat egory There could be two reasons why this species is misclassified and is predicted to be part of the open category The first reason is because although this species spends its time primarily in the forest, it feeds at night in open areas. Due to this fe eding strategy, Tragelaphus scriptus could have morphological characters more representative of its nighttime feeding strategy, rather than where it spends most of its time during the day. The second reason that this species could be misclassified is becau se the sample size of the modern bovids belonging to the closed habitat category is fairly small in comparison to the rest of the sample (n=8). Because the sample size is so small, the model is not as robust in predicting closed habitats as it is for predi cting open habitats (n=112). However, this study include s all of the extant Africa bovids adapted to forested habitats that the Smithsonian Museum of Natura l History has in its collection
40 Table 4 .8 : M isclassifications of Modern C ollection DFA Species Ac tual Predicted # Misclassified Connochaetes tarinus Open Intermediate 1 out of 3 Tragelaphus speke i Intermediate Open 1 out of 2 Tragelaphus strepsiceros Intermediate Open 1 out of 3 Kobus kob Intermediate Open 6 out of 13 Kobus megaceros Intermediate Open 1 out of 1 Redunca arundinum Intermediate Open 3 out of 3 Redunca fulvorufula Intermediate Open 4 out of 4 Tragelaphus scriptus Closed Open 6 out of 7 Uncertain classifications In the modern collection (n=177), there are only 11 uncertain predictions at the 0.1 0 probability level and 13 at the 0.05 level (T a ble 4.9 ) The small number of uncertain specimens indicates that this is a robust model for predicting habitat s based on astragali morphology. Some of the uncertain predictions are the same species as the misclassifications. This includes Connochaetes tarinus Tragelaphus spekei and Kobus kob The other uncertain classifications, Ourebia ourebia, Redunca redunca, Madoqua kirki, and Raphicerus sharpie were not p art of the misclassifications. Some of the uncertain classifications may reflect heavy body weight or inhabiting multiple habitats. For example, Connochaetes tarinus males are extremely close to the weight c ut off of 250 kg. Species such as Kobus kob, inhabit multiple habitats, as previously discussed in the misclassifications section. In addition, Tragelaphus spekei is the semi aquatic bovid
41 that lives primarily in swamps and whose astragali morphology previ ously discussed in the misclassifications section probably reflects the watery habitat in which it lives. However, d etermining the reason for the uncertain classification of species such as Ourebia ourebia and Redunca redunca is difficult because each spe cies inhabits primarily grassland. Similarly Madoqua kirki and Raphicerus sharpei primarily inhabit intermediate habitats Therefore, the reason for their uncertain classifications could have to do with a number of factors including dependency on water sources or traits that could keep species close to a few different habitats Table 4.9 : Uncertain Classifications for M odern Collection at the .10 L evel Actual Predicted Probability Species Open Open 0.59 Connochaetes tarinus Open Open 0.88 Redunca redunca Open Open 0.87 Ourebia ourebia Intermediate Open 0.64 Tragelaphus spekei Intermediate Open 0.88 Kobus kob Intermediate Open 0.66 Kobus kob Intermediate Open 0.79 Kobus kob Intermediate Open 0.66 Kobus kob Intermediate Intermediate 0.50 Kobus kob Intermediate Intermediate 0.84 Madoqua kirkii Intermediate Intermediate 0.80 Raphicerus sharpei
42 Fossil Astragali Collection The results of the modern collection are used as a base model to e valuate the results and the accuracy of the fossil astragali collection So the next step that I have completed in this study is another discriminant function analysis ( DFA ) this time on the fossil astragali and the resu lts are summarized in table 4.10 The total sample size of the fossil astraga li is 109 specimens. Table 4.10 : Results of DFA on Fossil Collection at the .05 Level Predicted Habitat Total Number Assigned Percent # Uncertain Classifications Open 64 58.72% 3 Intermediate 14 12.84% 1 Closed 31 28.44% 1 The results show that more than half of the fossil collection contains morphological characters similar to bovids adapted to open habitats. Almost a third of the collection shares morphological characters similar to bovids adapted to closed habitats. Finally, only a few share morphological characters with bovids adapted to intermediate habitats. The results also show that there are only five uncertain classifications out of a total sample size of 109 fossil specimens at the 0.05 probability level (Table 4.9) U ncertain Class ifications Table 4.10 summarizes the uncertain predictions for the fossil collection at the 0.05 probability level The number of uncertain predictions does not change in the fossil collection between the 0.05 and 0.10 level. Analyzing the r esults of the uncertain predictions from the fossil collection is slightly different from analyzing of the modern collection uncertain predictions because the accuracy of the predictions in the modern
43 collection can be assessed since the genus and the spec ies of the specimens are known Knowing the genus and species of the modern bovids allows researchers to know definitively the habitat to which the bovid is adapted In the fossil collection the genus and species is unknown. One of the reaso ns why not ass igning a genus or species to the fossil collection adds to the robustic ity of the model is that it does not assume that species today are practicing the same behaviors as their counterpart s in the Pliocene Therefore to evaluate the accuracy of the fossil results this model relies on the classification rate of the modern collection The classification rate of approximately 82% in the modern collection (Table 4.6 ) shows that the results for the fossil collection are fairly accurate, although there will be a few misclassifications. F or the modern collection those specimens that are misclassified are some of the specimens with uncertain classifications such as Connochaetes tarinus, Tragelaphus spekei, and Kobus kob It is important to note, however, that just as in the modern collection, some of the misclassification in the fossil collection may not overlap with the uncertain classifications. The uncertain classifications that do not overlap with misclassifications are harder to identify in the fossil coll ection Thus this shows that some but not all of the fossil collections uncertain classifications will be those that are misclassified. Unfortunately with the fossil collection there is no way to know exactly which specimens are the mis classified astrag ali. The model however, provides probabilities about which astragali may be misclassified. Some have higher probabilities than others of being misclassified but some have 100% probability of being classified accurately (n=64) Additionally, since there are only 5 uncertain classifications at the 0.05 probability level, that means that
44 there are 105 specimens that have a probability of more than 95% that they are classified in the correct habitat category. Overall, these results from the fossil collection indicate that the classifications are very accurate. Table 4.11 : Summary of Uncertain Classifications for F ossil C ollection at the 0.05 L evel Predicted Habitat Probability Alternative Predicted Habitat Probability Open .84 Intermediate 0.16 Open 0.54 Intermediate 0.45 Open 0.59 Intermediate 0.41 Intermediate 0.57 Open 0.43 Closed 0.68 Intermediate 0.31
45 CHAPTER V DISCUSSION The results from the discriminant function analysis of the modern collection of bovid astragali indicate that the model for determining the habitats that may have been available to bovids at Pliocene Laetoli is robust and accurate T he accuracy rate for this model is approximately 82%, which without jackknifing, bootstrapping, and other similar statistical procedu res that o ther models have done is a relatively high rate. The reason that this model did not do jackknifing and bootstrapping is because the sample size is not large enough to do jackknifing and the sample size is large enough that bootstrapping was not necessary. The fossil collection results indicate that o ver half of the specimens share morphological similarities to open habitat bovids. Approximately one third of the fossil specimens share morphological similarities t o closed habitat bovids. Finally, a small number of the specimens (approximately 12%) share morphological similarities to the intermediate habitat bovids T he data indicate s that Laetoli during the Upper Laetolil Beds (3.85 3.63 Ma) was a complex environment. T he habitat s available to bov ids were mostly open. This means that the habitats were primarily grasslands, plains, arid country, ecotone regions, or a combination of these habitats. Because there is approximately one third of the fossil collection that belongs to the closed habitat ca tegory, this indicates that although the habitat was not primarily forested there were pockets of forested habitat s spread throughout the site. The very small number of specimens belonging to the inte rmediate habitat category shows that habitats that were woodland, hilly areas, swamp, bushland or near water habitats were perhaps present at Laetoli but rare. This percentage
46 could be infla ted due to taphonomic processes since the intermediate habitat category includes near water habitats and there is a bia s in the fossil record that animals adapted to alluvial and lacustrine environments will be over represented, while those who are not tend to be under represented (Dominguez Rodrigo and Musiba 2010)
47 CHAPTER VI CONCLUSION To sum, this study shows how a model may be constructed that can determine which habitats bovids may have occupied at a site by analyzing bovid astragali morphology and using a modern bovid collection as a control for determining the accuracy of the fossil results. Other studies have s ho wn this to be a reliable method (DeGusta and Vrba 2003, Plummer et al. 2008), and some have applied this model to fossil collections to determine paleoenvironments at important paleoanthropological sites (Bishop et al. 2011 Curran 2012, Kappelman 1991, Klein et al. 2010, Musiba and Magori 2003 ). For this study, the classification rate is approximately 82%. The results of the ecomorphological study indicate that the habitats bovids occupied at Laetoli 3.85 3.63 Ma was mostly open, with pockets of forested habitats spread throughout. This study also reinforces the importance of paleoecological modeling in paleoanthropological research. Paleoecological models help to illuminate selective pressures that influenced hominin evolution. Understanding more details about selective pressures at Laetoli is important because Laetoli is a unique paleoanthropological site in East Africa Fossil hominins, vertebrate faunal assemblages, and fossilized hominin and animal trackways have been discovered there Laetoli is also a site that lacked permanent water sources (Harrison 2011a Musiba 1999 ). Moreover, t he time in whi ch the Upper Laetolil Beds date (3.85 3.63 Ma) is also a time in which fossil faunal evidence at other sites is rare ( Harrison 2011 a ). Overall, the evidence points toward a paleoenvironment at Laetoli that was complex and mosaic with pockets of forested habitats within more open areas, and this
48 study adds to that growing body of evidence ( Bamford 2011; Bishop et al. 2011 ; Harrison 2011a, 2011b 2011c ; Kaiser 2011; Kingston 2011; Kovaric and Andrews 2011; Musiba and Magori 2003; Su 2011; Su and Harrison 2007 ). The evidence for a complex, mosaic environment at Laetoli is significant for understanding hominin environment int eractions. Bovids and hominins lived in the same area so the sel ective pressures influencing bovids should have similarly affected early hominins. Since bovids are sensitive indicators of environmental changes (Spencer 1997) and hominins are not, analyses of bovids provide less ambiguous reconstructions of habitats and environments at a site. Moreover, since Laetoli is a unique East African hominin site due to its lack of permanent water sources (Harrison 2011 a Musiba 1999 ), this study provides further evidence that there were different selective pressures at Pliocene Laetoli than at other early hominin sites. Future research should primarily be concerned with determining how the Laetoli paleoenvironment was mosaic. It is possible that the environmenta l signature from the Upper Laetolil Beds is mosaic due to short term environmental changes. The Upper Laetolil Beds reflect about two hundred thousand years, which may be enough time averaging in this collection to obscure short term environmental fluctuat ions On the other hand, it is also possible that Laetoli was a complex environment with pockets of different habitats coexisting at the same time. Future research should conduct similar paleoecological studies with better stratigraphic control. A more det ailed analysis with tighter temporal control could be conducted if the layers that astragali were recovered from within the Upper Laetolil Beds were known. Additionally, future research of the paleoecology of East Africa should seek middle range theories t hat aid in the synthesis of different lines of data that examine the world at different scales This will enable more
49 accurate generalizations of region s from local, site specific evidence recovered from sites such as Laetoli
50 REFERENCES Alemseged Z, Bobe R, Geraads D. 2007. Comparability of fossil data and its significance for the interpretation of hominin environments. In: Bobe R, Alemseged Z, Behrensmeyer A, editors. Hominin Environments in the East African Pliocene, The Netherlands : Springer p. 159 182. Bamford M. 2011. Fossil Leaves, Fruits, and Seeds. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Pa leoanthropology. New York : Springer. p. 235 252. Behrensmeyer A, Alemseged Z, Bobe R. 2007. Finale and Future: Investigating faunal evidence for hominin paleoecology in East Africa. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands : Springer. p. 333 346. Berger L, de Ruiter D, Churchill S, Schmid P, Carlson K, Dirks P, Kibii J. 2010. Australopithecus sediba : A new species of Homo like Australopith from South Africa. Science. 328, 195 20 4. Bishop L. 2011. Suidae. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York : Springer. p. 327 337. Bishop L, Plummer T, Hertel F, Kovarovic. 2011. Paleoenvironments of Laetoli, Tanzania as Determined by Antelope Habitat Preferences. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoeco logy, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 355 366. Bobe R, Behrensmeyer A, Eck G, Harris J. 2007. Patterns of abundance and diversity in late Cenezoic bovids from the Turkana and Hadar Basins, Ken ya and Ethiopia. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands : Springer p. 129 158. Bonnefille R. 1984. Palynological research at Olduvai Gorge. National Geographic Society Resear ch Reports. 17: 227 243. Bonnefille R. 2010. Cenezoic vegetation, climate changes and hominid evolution in tropical Africa. Global and Planetary Change. 72:390 411. Bonnefille R, Riollet G. 1980. Palynologie, vegetation et climates de Bed I et Bed II a Olduvai, Tanzania. Preceedings of the Eighth PanAfrican Congress of Prehistoric and Quaternary Studies. Nairobi.
51 Bonnefille R, Riollet G. 1987. Palynological spectra from the Upper Laetoli Beds. Laetoli. In: Leakey M, Harris J, editors. A Pliocene Sit e in Northern Tan zania. Oxford : Clarendon Press p. 52 61 Brunet M, Guy F, Pilbeam D, Mackaye HT, Likius A, Ahounta D, Beauvilain A, Blondel C, Bocherens H, Bolsserle J, De Bonis L, Coppens Y, Dejax J, Denys C, Duringer P, Elsenmann V, Fanone G, Fronty P, Geraads D, Lehmann T, Lihoreau F, Louchart A, Mahamat A, Merceron G, Mouchelin G, Otero O, Campomanes P, Ponce De Leon M, Rage J, Spanet M, Schuster M, Sudre J, Pascal T, Valentin X, Vignaud P, Virlot L, Zazzo A, Zollikofer C. 2002. A new hominid from the Upper Miocene of Chad, Central Africa. Nature. 418, 145 151. Burger O, Todd L, Burnett P, Stohlgren T, Stephens D. 2004. Multi scale and Nested Intensity Sampling Techniques for Archaeological Survey. Journal of Field Archaeology. 29: 409 423. Cerl ing T, Hay R. 1986. An isotopic study of paelosol carbonates from Olduvai Gorge. Quaternary Research. 25: 63 78. Curran S. 2012. Expanding ecomorphological methods: geometric morphometric analysis of Cervidae post crania. Journal of Archaeological Sciences. 39: 1172 1182. DeGusta D, Vrba E. 2003. A method for inferring paleohabitats from funcational morphology of bovid astragali. Journal of Archeological Sciences. 30: 1009 1022. DeGusta D, Vraba E. 2005. Methods for inferring paleohabitats from fun ctional morphology of bovid phalanges. Journ al of Archaelogical Science. 32: 1099 1113. Deino A. 2011. 40 Ar/ 39 Ar Dating of Laetoli, Tanzania. In: Harrison T, editors. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geoch ronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York: Springer. p. 77 97. deMenocal P. 2007 African climate change and faunal evolution during the Plio Pleistocene. Ea rth and Planetary Letters. 220: 3 24. Denys C. 2011. Rodents. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 15 53. Ditchfield P, Harrison T. 2011. Sedimentology, Lithostratigraphy and Depositional History of the Laetoli Area. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology New York : Springer p. 47 76.
52 Dominguez Rodrigo M, Musiba C. 2010. How Accurate are Paleoecological Reconstructions of Early Paleontological and Archaeological Sites? Journ al of Evolutionary Biology. 37: 128 140. Frost S. 2007. African Pliocene and Pleistocene cercopithecid evolution and global climatic change. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands : Springer p. 51 76. Gen try A W 1985 The Bovidae of the Omo Group Deposits, Ethiopia (French and American collections). In: Coppens Y, Howell F, editors. Les Fanes Plio Pl ÂŽistocÂnes de la Basse VallÂŽe de L'Omo (Âƒthiopie). Tome I. Les PÂŽrissodactyles les Artiodactyles (les Bovida e) Expediti on Internationale 1967 76. Paris : Âƒditions du Centre National de la Recherche Scientifique. p. 119 191. Gentry A W. 1987. Plioc e ne Bovidae from Laetoli. In: Leakey M, Harris J, editors. Laetoli, a Pliocene S ite in Northern Tanzania. Oxford : Clar endon Press p. 378 408. Gentry A W 2011. Bovidae. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology New York : Springer. p. 363 465. Gentry AW, Gentry A. 1978. Fossil Bovidae (Mammalia) of Olduvai Gorge, Tanzania, Parts I II. Bulletin of the Brithish Museum of Natural History (Geology), London. 29, 30: 289 446; 1 83 Haile Selassie Y, Saylor B, Deino A, Alene M, Latimer B. 2010. New hominid fossils from Woranso Mille (Central Afar, Ethiopia) and taxonomy of early Australopithecus American Journal of Physical Anthropology. 141, 406 417. Harris J. 1991. Koobi For a Research Project, volume 3: Fossil Ungulates: Geology, Fossil Artiodactyles, and Paleoenvironments. Oxford: Clarendon Press. Harrison T. 2011 a Laetoli Revisited: Renewed Paleontological and Geological Investigations at Localities on the Eyasi Plateau in Northern Tanzania. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer. p. 1 15. Harrison T. 2011b Cercopithecids (Cercopi thecidae, Primates). In: Harrison T. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology a nd Paleoanthropology. New York : Springer p. 83 139.
53 Harrison T. 2011c Galagid ae (Lorisoidea, Primates). In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York: Springer. p. 75 82. Hay R.L. 1987 Geology, Dating, and Palynology: Geology of the Laetoli area. In: Leakey MD, Harris JM, editors. Laetoli A Pliocene Site in Northern Tanzania. Clarendon Press: Oxford. P. 23 47. Hernesniemi E, Giaourtsakis I, Evans A, Fortelius M. 2011. Rhinocerotidae. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 275 294. Kaiser T. 2011. Feeding Ecology and N iche Partitioning and of the Laetoli Ungulate Faunas. In: Harrison T. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 329 354. Kappelman J. 1984. Plio Pleistocene environments of Bed I and lower Bed II, Olduvai Gorge, Tanzania. Paleogeography, Paleocli matology, and Paleoecology. 48: 171 196. Kappelman J. 1991. The Paleoenvironment of Kenyapithecus a t Fort Ternan. Journal of Human Evolution. 20: 95 129. Kappelman J, Plummer T, Bishop L, Duncan A, Appleton A. 1997. Bovids as indicators of Plio Pleistocene paleoenvironments in East Africa. Journal of Human Evolution. 32: 229 256. Kidwell S, Holland S. 2002. The Quality of the Fossil Record: Implications for Evolutionary Analyses. Annual Review of Ecological Systems. 33: 561 588. Kingdon J. 1982. East African Mammals, Volume IIIC and IIID: Bovids. Chicago: University of Chicago Press. Kingdon J. 2004. The Kingdon Pocket Guide to African Mammals. Princeton : Princeton University Press Kingston J. 2007. Shifting Adaptive Landscapes: Progess and Challenges in Reconstructing Early Hominid Environments. Yearbook of Physical Anthropology. 5 0: 20 58. Kingston J. 2011. Stable Isotopic Analyses of Laetoli Fossil Herbivores. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 293 328.
54 Klein R, Franciscus R, Steele T. 2010. Morphometric identification of bovid metapodials to genus and implications for taxon free habitat reconstruction. Journal of Archaeological Sciences. 37: 389 401. Kovarovic K, Andrews P. 2007. Bovid postcranial ecomorphological survey of the Laetoli pal eoenvironment. Journal of Human Evolution. 52: 663 680. Kovarovic K, Andrews P. 2011. Environmental Change within the Laetoli Fossiliferous Sequence: Vegetation Catenas and Bovid Ecomorphology. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology New York: Springer. p. 367 380. Krell F T, Schawaller W. 2011. Beetles (Insecta: Coleoptera). In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 535 548. Leakey M, Harris J. editors. Laetoli: A Pliocene Sit e in Northern Tanzania. 1987. Oxford: Clarendon Press Lewis M, Werdelin L. 2007. Patterns of change in the Plio Pleistocene carnivorans of eastern Africa. In: Bobe R, Alemseged Z, Behrensmeyer A, editors. Hominin Environments in t he East African Pliocene The Netherlands : Springer p. 77 106. Musiba C. 1999. Laetoli Pliocene paleoecology: a reanalysis via morphological and behavioral approaches. Ph.D. Dissertation, University of Chicago, Chicago, Illinois. Musiba C, Magori C. 2003. Laetoli Paeloecology : Predictive behavioral ecology model based on functional morphology and sediment proxy data. In: Lane P, Mapunda B, editors. Tanzania Cultural Heritage. British Archaeology Institute in East Africa Special Publication. London : University Press Musiba C, Magori C, Stoller M, Stein T, Branting S, Vogt M, Tuttle R, Hallgrimsson B, Killindo S, Mizambwe F, Ndunguru F, Mabulla A. 2007. Taphonomy and paleoecological context of Upper Laetoli Beds (Localities 8 & 9), Laetoli in northern Tanzania. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands : Springer p. 257 278. Norton Griffiths M, Herlocker D, Pennycuick L. 1975. The patterns of rainfall in the Serengeti Ecosystem, Tanza nia. East African Wildlife. 13: 347 374. Plummer T, Bishop L, Hertel F. 2008. Habitat preference for extant bovids based on astragalus morphology: operationalizing ecomorphology for paleoenvironmental reconstructions. Journal of Archaelogical Scie nces. 35: 30 16 3027.
55 Plummer T, Bishop L. 1994. Hominid paleoecology at Olduvai Gorge, Tanzania as indicated by antelope remains. Journal of Human Evolution. 2 9: 321 362. Potts R. 2007. Environmental hypotheses of Pliocene human evolution. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands: Springer. p. 25 50. Quade J, Levin N. 2013. East African Hominin Paleoecology: Isotopic Evidence from Paleosols. In: Sponheimer M, Lee Thorp J, Reed K, Ungar P, editors. E arly Hominin Paleoecology. Boulder : University Press of Colorado. Reed D. 2007. Serengeti micromammals and their implications for Olduvai paleoenvironments. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East Af rican Pliocene. The Netherlands : Springer p. 217 256. Reed D. 2011. Serengeti Micromammal Communities and the Paleoecology of Laetoli, Tanzania. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 253 263. Reed D, Denys C. 2011. The Taphonomy and Paleoenvironmental Implications of the Laetoli Micromammals. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 265 278. Reed K, Spencer L, Rector A. 2013. Faunal Approaches in Early Hominin Paleoecology. In: Sponheimer M, Lee Thorp J, Reed K, Ungar P, editors. Early Hominin Paleoecology. Boulder : University Press of Colorado p. 3 34. Rightmire G, Lordkipanidze D, Vekua A. 2006. Anatomical descriptions, compartive studies and evolutionary significance of the hominin skulls from Dmanisi, Republic of Georgia. Journal of Human Evolution. 50, 115 141. Rossouw L, Scott L. 2011. Phytoliths and Pollen, the Microscopic Plant Re mains in Pliocene Volcanic Sediments Around Laetoli, Tanzania. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoa nthropology. New York : Springer p. 201 215. Senut B, Pickford M, Gommery D, Mein P, Chebio K, Coppens Y. 2001. First hominid from the Miocene (Lukeino Formation, Kenya). Earth and Planetary Sciences. 332, 137 144.
56 Sikes N. 1994. Early hominid habitat preferences in East Africa: paleosol carbon isotope evidence. Journal of Human Evolu tion. 27: 25 45. Shipman P, Harris J. 1988. Habitat preference and paleoecology of Australopithecus boisei in Eastern Africa. In: Grine F. Aldine de Gruyter, editors. Th e Evolutionary History of the Robust" Australopithecines. New York : Aldine Transaction p. 343 381. Spencer L. 1997. Dietary adaptations of Plio Pleistocene Bovidae: implications for hominid habitat use. Journal of Human Evolution. 32: 201 228. Su D. 20 11. Large Mammal Evidence for the Paleoenvironment of the Upper Laetolil and Upper Ndolanya Beds of Laetoli, Tanzania. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 1: Geology, Geochronology, Paleoecology, and Paleoenvironment. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 381 392. Su D, Harrison T. 2007. The paleoecology of the Upper Laetoli Beds at Laetoli: A reconsideration of the large mammal evidence. In: Bobe R, Alemseged Z, Behrensmeyer A editors. Hominin Environments in t he East African Pliocene. The Netherlands : Springer p. 279 313. Vrba E. 1976. The fossil Bovidae of Sterkfontein, Swartkrans, and Kromdraai Transvaal Museum Memiors. 21: 1 166. Vrba E. 1 980. The significance of bovid remains as indicators of environment and predation patterns. In: Behrensmeyer A, Hill A, editors. Fossils in the Making. Chicago : University of Chicago Press p. 247 271.. Vrba E. 1985. Paleoecology of early Hominidae, with special reference to Sterkfontein, Swartkrans, and Kromdraii. In: Coppens Y, editor. L'Environment des Hominid ÂŽ s aux Plio Pl ÂŽistocÂne. Paris: Masson. Wandsnider L, Camilli E. 1992. The Character of Surface Archaeological Deposits and its Influence on Survey Acc uracy. Journal of Field Archaeology. 19: 169 188. Ward C, Leakey M, Walker A. 1999. The new hominid species Australopithecus anamensis. Evolutionary Anthropology. 7, 197 205. Werdelin L, Dehghani R. 2011. Carnivora. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human Evolution in Context. Volume 2: Fossil Hominins and Associated Fauna. Vertebrate Paleobiology and Paleoanthropology. New York : Springer p. 189 232. White T, Asfaw B, Beyene Y, Haile Selassie Y, Lovejoy CO, Suwa G, WoldeGa briel G. 2010. Ardipithecus ramidus and the Paleobiology of Early Hominids. Science. 326, 75 86.
57 WoldeGabriel G, Ambrose S, Barboni D, Bonnefille R, Bremond L, Currie B, DeGusta D, Hart W, Murray A, Renne P, Jolly Saad M, Stewart K, White T. 2010. The Ge ological, Isotopic, Botanical, Invertebrate, and Lower Vertebrate Surroundings of Ardipithecus ramidus Science. 326, 65e1 65e5.