The emergence in eastern Africa of the Homo and Paranthropus lineages as well as the extinction of Australopithecus afarensis in the Late Pliocene (between 3 Ma and 2.6 Ma) coincide with a period of vegetation change that was occurring in the region. Paleoenvironmental proxies derived from marine and lake sediments—including pollen, phytoliths, and stable carbon isotopes of plant wax biomarkers—indicate that vegetation oscillated between more wooded and more grass-covered conditions in relation to insolation changes caused by Earth's orbital cycles (Bonnefille et al., 2004; Feakins et al., 2005, 2013; Lupien et al., 2021; Yost et al., 2021). Overlying these orbital-scale oscillations is a region-wide trend of increasing grasses utilizing the C4 photosynthetic pathway and arid-adapted plants (Feakins et al., 2005; Bonnefille, 2010; Uno et al., 2016; Bonnefille and Bourel, 2021). From the Late Pliocene to the Early Pleistocene, C4 plants increased to a similar degree at many paleoanthropological sites in eastern Africa (Fillion and Harrison, 2023). Although there was a general trend for the spread of C4 grasses across eastern Africa, at any given time, local proportions of grassy versus woody vegetation varied by location. For example, areas in the Lower Awash Valley generally had more grassy vegetation than those in the Turkana Basin because grasses were already more prevalent in the Awash earlier in the Pliocene (Cerling et al., 2011; Levin et al., 2011; Fillion and Harrison, 2023). The spread of C4 grasses is apparent in the dietary guild structure of the mammal communities that inhabited eastern Africa during the Pliocene. Among large herbivores, grazing taxa became more prevalent from the Pliocene to the mid-Pleistocene (Cerling et al., 2015; Faith et al., 2019, 2024). Dietary shifts among the herbivores in response to vegetation change are also reflected in the carnivore guild. The stable carbon isotope composition of carnivoran tooth enamel in the Turkana Basin is increasingly positive between 4 Ma and 1.5 Ma, indicating greater consumption of C4-consuming herbivores (Hopley et al., 2022).

Vegetation change would have had an important impact on hominin adaptation and ecology. The transition from wooded environments to those with more extensive grass cover is hypothesized to have been an important driver of novel hominin adaptations, speciation, and extinction (Vrba, 1988; Bobe and Behrensmeyer, 2004; Potts, 2013; Antón et al., 2014; Domínguez-Rodrigo, 2014; Uno et al., 2016; but see Maxwell et al., 2018). Decreasing tree cover and the spread of C4 grasses would have affected the exploitation of food resources by hominins (Copeland, 2009; Lee-Thorp et al., 2012; Stewart, 2014), thermoregulation (Wheeler, 1991; Pruetz, 2007; Pruetz and Bertolani, 2009), locomotion and ranging behavior (Rodman and McHenry, 1980; Hunt, 1994, 1998; Pontzer et al., 2009), and access to trees for shelter and predator avoidance (Sept et al., 1992; Hernandez-Aguilar, 2009; Koops et al., 2012). Evidence from eastern African paleoanthropological sites suggests that hominin species turnover is associated with local vegetation change. In the Lower Awash Valley in Ethiopia, early Homo occurs between 2.8 Ma and 2.4 Ma in environments that are more grass-covered than those in which Au. afarensis occurred (Robinson et al., 2017; Alemseged et al., 2020). Similarly, in Ethiopia's Omo Valley, the first occurrence of Paranthropus ∼2.7 Ma is also associated with an expansion of C4 grasses (Negash et al., 2020).

Laetoli, a paleoanthropological site in northern Tanzania, presents another opportunity to examine vegetation change and its association with hominin turnover during the Late Pliocene. Two of the fossil-bearing sedimentary units at Laetoli, the Upper Laetolil Beds (ULB), dated to 3.83–3.65 Ma, and the Upper Ndolanya Beds (UNB), dated to 2.66 Ma (Deino, 2011), span an important period of climate and environmental change in the Late Pliocene. During this time period, Homo and Paranthropus emerged and Au. afarensis went extinct. Hominin species turnover occurred locally at Laetoli: Au. afarensis in the ULB was replaced by Paranthropus aethiopicus in the UNB (Leakey, 1987; Harrison, 2011a). The UNB provides one of the earliest occurrences of Para. aethiopicus in the fossil record and is the only known occurrence of this species outside of the Omo-Turkana Basin (Harrison, 2011a; Wood and Patterson, 2020). Reconstructing the paleo-vegetation of the ULB and UNB is important for understanding the environmental contexts in which Au. afarensis and early Paranthropus occurred and provides additional evidence for how local-scale paleoenvironments responded to global and regional climate change during this crucial period in hominin evolution.

Although the paleoenvironment and paleoecology of Laetoli have been extensively studied, it remains unresolved how the vegetation of the UNB differed from that of the ULB. The balance of evidence from multiple paleoecological proxies indicates that the ULB and UNB were both woodland-grassland mosaics in which tracts of grasslands and dense woody cover occurred simultaneously (Harrison, 2011b, 2017; Scott, 2012; Su and Harrison, 2015; Forrest et al., 2018; Fillion et al., 2022; Harrison et al., 2022). Thus, how vegetation might have changed at Laetoli is a question of the degree of the relative extent of grasses versus woody vegetation (Harrison, 2017; Harrison et al., 2022). However, different proxies point to alternative conclusions about how the mosaic paleoenvironment of Laetoli may have changed between the deposition of the ULB and UNB. On the one hand, evidence from stable isotopes of paleosol carbonates (Cerling, 1992), bovid ecomorphology Harrison, 2011b, 2017; Scott, 2012; Su and Harrison, 2015; Forrest et al., 2018; Fillion et al., 2022; Harrison et al., 2022(Kovarovic and Andrews, 2007, 2011; Bishop et al., 2011; Forrest et al., 2018), stable isotopes of the tooth enamel and the mesowear of large herbivores (Kingston and Harrison, 2007; Kaiser, 2011; Kingston, 2011), the composition of the large mammal community (Kovarovic et al., 2002; Su, 2005, 2011; Su and Harrison, 2007), the phytolith assemblage (Rossouw and Scott, 2011), and insect trace fossils (Genise and Harrison, 2018) suggest the UNB was more arid and had a greater extent of grass cover than the ULB. On the other hand, the stable isotope composition and increased thickness of ostrich eggshells (Harrison and Msuya, 2005; Kingston and Harrison, 2007; Kingston, 2011) and the composition of the terrestrial gastropod (Tattersfield, 2011) and micromammal (Reed, 2011; Reed and Denys, 2011) communities instead indicate that the UNB was at least as densely wooded as the ULB and may have been wetter.

It has been suggested that some proxy evidence derived from large herbivores that suggests an increase in grass vegetation in the UNB, such as mesowear, stable carbon isotopes of tooth enamel (δ13Cenamel), and ecomorphology, is heavily influenced by the appearance of novel taxa in the UNB (Harrison, 2017). These taxa, including a new species of the hipparionin equid Eurygnathohippus and new alcelaphin and antilopin bovid species, are inferred to have been specialized for grazing in grassland-dominated environments (Kingston and Harrison, 2007; Kovarovic and Andrews, 2007, 2011; Kaiser, 2011; Kingston, 2011). It is unlikely that these taxa originated at Laetoli; rather, these grassland specialists presumably emerged elsewhere in eastern Africa. When these taxa arrived at Laetoli, they would have been more capable of exploiting the grassy component of the vegetation mosaic that existed at Laetoli, even if grass cover did not increase much between the ULB and the UNB. Thus, if evidence suggesting that the UNB had more grass cover is primarily the result of novel grazing specialists, this could indicate that grass cover was increasing in the broader eastern African region, but not necessarily locally at Laetoli.

The apparent discrepancies in vegetation reconstructions from different paleoenvironmental proxies may also be due to the spatial scale at which proxies record environmental information (Harrison, 2017). Large herbivores, which range over large distances and can occur in a wide diversity of habitats, provide paleoenvironmental evidence of a coarser spatial resolution (Harrison, 2017). In contrast to large mammals, terrestrial gastropods and small mammals inhabit much smaller ranges and tend to have narrower habitat preferences and ecological tolerances (Reed, 2011; Tattersfield, 2011). Small mammal carcasses and gastropod shells may be transported by avian raptors or fluvial actions, respectively, but while some small mammal remains at Laetoli are likely derived from owl pellets, other specimens were likely buried in their burrows (Denys, 1985; Reed and Denys, 2011). There is no evidence of the distant transport of either small mammals or gastropods. Thus, evidence from gastropods and small mammals, at least at Laetoli, provides fine-grained spatial resolution (Harrison, 2017). Similarly, while ostriches are capable of traversing long distances, their eggs are not typically transported far from the nest site and therefore reflect local conditions (Harrison and Msuya, 2005; Kingston, 2011; Harrison, 2017).

Thus, the question remains, did the vegetation of the UNB differ from that of the ULB, and if so, how did it differ? To investigate potential vegetation change at Laetoli, this study uses multiple proxies to infer the diets of large herbivores (Artiodactyla, Perissodactyla, and Proboscidea) and reconstruct the dietary guild structures of the ULB and UNB large herbivore communities. A multiproxy approach allows for potential differences due to scale to be addressed, while the approach of using community-wide dietary change to reconstruct paleoenvironment recognizes recent work suggesting that it is communities, rather than individual species, that best reflect differences in vegetation (Bedaso et al., 2010; Robinson et al., 2021; Sokolowski et al., 2023).

Herbivores can be broadly categorized based on the proportion of grass consumed, along a continuum ranging from browsers and frugivores through mixed feeders to grazers (Hofmann and Stewart, 1972; Kay et al., 1980; McNaughton and Georgiadis, 1986; Gagnon and Chew, 2000; Cerling et al., 2003; Sponheimer et al., 2003). The diets of large herbivores, reconstructed using a variety of proxies, reflect different habitat types and have been widely applied to investigate vegetation change during the Plio-Pleistocene of eastern Africa (Andrews et al., 1979; Spencer, 1995; Uno et al., 2011; Bedaso et al., 2013; Bibi et al., 2013; Cerling et al., 2015; Lüdecke et al., 2016; Wynn et al., 2016; Rowan et al., 2017; Blondel et al., 2018; Dumouchel and Bobe, 2020; Manthi et al., 2020; Negash et al., 2020; Dumouchel et al., 2021). Examining shifts in the dietary guild structure of large herbivore communities is useful for reconstructing past vegetation even if the diets of individual herbivore lineages do not necessarily track vegetation change (Bedaso et al., 2010; Robinson et al., 2021). Investigating community-wide change is particularly relevant following recent work suggesting that few herbivore species are good ‘indicator species’—that is, species whose presence indicates a particular type of vegetation (Sokolowski et al., 2023). While the presence of a particular herbivore species or dietary guild may not be strong evidence for reconstructing paleoenvironments, how different species are assembled to form communities remains a powerful tool for investigating ecological change.

In this study, three proxies—hypsodonty, mesowear, and δ13Cenamel—are used to infer the diets of large herbivores and compare the dietary guild structures of the ULB and UNB communities. Hypsodonty is the condition wherein the tooth crown is apicobasally tall such that upon tooth eruption, some of the crown remains within the jawbone and erupts later in life (Janis, 1988). It is an adaptation to high rates of tooth wear that occur when consuming foods such as grasses that are intrinsically abrasive and/or that tend to accumulate extrinsic abrasive particles, such as dust or grit (Damuth and Janis, 2011). Mesowear is a measure of the relative contribution of attrition, or tooth-on-tooth wear, and abrasion, or food-on-tooth wear, which influence the shape of tooth cusps over the course of an individual animal's lifetime (Butler, 1972; Fortelius and Solounias, 2000). While browsing diets tend to generate more attritive wear, grazing diets generate more abrasive wear (Fortelius and Solounias, 2000). δ13Cenamel reflects the stable carbon isotope composition of the foods an individual animal consumed during dental enamel formation. The stable carbon isotope composition of plants that utilize the C3 photosynthetic pathway, which includes trees, shrubs, and forbs, is distinguishable from the stable carbon isotope composition of plants that utilize the C4 photosynthetic pathway, which in eastern Africa are predominantly grasses except at high elevations (Bender, 1971; Tieszen et al., 1979; O'Leary, 1988; Farquhar et al., 1989). Plants that utilize a third photosynthetic pathway, Crassulacean acid metabolism, have a similar stable carbon isotope composition as C4 plants, but these were likely not major contributors to the biomass of vegetation in eastern Africa during the Pliocene and are thus not considered here (O'Leary, 1988; Kingston, 2011).

Each of the three dietary proxies records information on different temporal scales (Davis and Pineda-Munoz, 2016). Hypsodonty is a specialized trait that reflects dietary changes in herbivore lineages over tens to hundreds of thousands of years. Mesowear reflects an individual animal's diet over the course of its adult lifetime. δ13Cenamel records dietary information during the period of enamel formation and reflects a young individual's diet during its early developmental period. Thus, hypsodonty indicates the type of diet a herbivore is adapted to consume, mesowear indicates the average diet a herbivore consumed during its life, and δ13Cenamel indicates the diet a herbivore consumed during a short-term preadult period. Due to these differences in scale, the inferred diet of a herbivore taxon may differ between proxies. However, these potential discrepancies also offer insight into how scale impacts dietary inferences. To infer the diets of the large herbivores at Laetoli, results from all three proxies are used, but scale differences are taken into account by giving more weight to results from proxies that record the foods an animal actually consumed during its lifetime (i.e. mesowear and δ13Cenamel) over the proxy that reflects the diet a species is adapted to consume (i.e. hypsodonty).

This study takes a community-wide approach to using the diets of large herbivores to reconstruct potential vegetation change. First, to determine whether dietary behavior changed among the large herbivores at Laetoli, taxon-average hypsodonty, mesowear, and δ13Cenamel values are compared between the ULB and UNB large herbivore communities. Proxy results are discussed in terms of how their differences in scale might offer insights into interpretations of dietary change. One question is whether potential differences in community-wide proxy results are driven primarily by novel taxa in the UNB or whether they reflect dietary change among taxa that occur in both units. To investigate this, the hypsodonty, mesowear, and δ13Cenamel values of herbivore taxa in the ULB and UNB that are shared between both units are compared to taxa that only occur in the UNB. Second, the results from all three proxies are combined to determine the dietary guild structure of the large herbivore communities at Laetoli. Despite the possibility that different proxies might infer different dietary behavior, previous work has shown that using multiple proxies to infer diet leads to better outcomes than classifying diets based on one proxy alone (Fraser and Theodor, 2011). Differences in the dietary inferences made based on each proxy are noted and discussed. Finally, the dietary guild structures of the ULB and UNB are compared to each other and to communities of large herbivores in Africa today.

Based on previous work indicating a shared trend of increasing C4 vegetation at most major eastern African paleoanthropological sites during the Late Pliocene to Early Pleistocene (Bonnefille, 2010; Levin et al., 2011; Uno et al., 2011; Negash et al., 2020; Lupien et al., 2021; Fillion and Harrison, 2023), and in accordance with evidence from the majority of proxies from Laetoli, we hypothesize that grassy vegetation was more predominant in the UNB than in the ULB. Specifically, we predict that each of the three dietary proxies, hypsodonty, mesowear, and δ13Cenamel, will be higher in the UNB herbivore community than in the ULB community, indicating an overall increase in grass consumption at the community level. We further predict that when all proxies are combined to determine the dietary guild structure of the herbivore communities, the UNB will have more grazers and fewer mixed feeders and browsers/frugivores than the ULB. Finally, we predict that the UNB community will be more similar to modern African herbivore communities in environments with a high degree of grass cover (e.g. communities in grasslands and woody grasslands) than the ULB community.