Evaluating Dietary Diversity among Andean Central Altiplano Early Camelid Pastoralists Using Stable Isotope Analysis

ABSTRACT Stable isotope studies have revolutionised our understanding of food webs, subsistence, mobility and their change through time. In the Andes, dietary stable isotopic studies have often focused on timing the adoption of maize as a staple food and identifying camelid pastoralism in selected valleys of the Pacific coast. Few studies have focused on highland societies and understanding pastoralist communities in particular, but the underlying assumption has been that camelid herders had essentially narrow and specialised diets. To evaluate this assertion, we analyzed dietary carbon and nitrogen stable isotopes of 27 human and 23 faunal specimens from nine archaeological sites located in the Central Altiplano of Bolivia supported by 27 radiocarbon dates. Our results suggest important diversity characterised both human and animal isotopic ecology between 1300 BCE and 1200 CE. The collected data also demonstrates that maize was not regularly consumed, suggesting interregional food exchange between this region of the Altiplano and its neighboring lowland inter-Andean valleys likely postdated 1100 CE. Consistent with reliance on cultivated crops such as chenopods and tubers and wild fauna including rodents, birds and fish, early camelid herders in the Andean Central Altiplano relied on a generalised form of pastoralism.


Introduction
Stable isotope analysis has revolutionised our understanding of subsistence transformation through time by facilitating qualitative and quantitative reconstructions of the diets and environments of ancient peoples (Lee-Thorp 2008;Schoeninger and Moore 1992;Ventresca Miller and Makarewicz 2017).Because the biochemical composition of vertebrate skeletal tissues is directly related to their diet and local ecology, stable isotope analysis can substantially increase our understanding of food webs, localised environmental change, socio-economic behaviour and mobility through space and over time (Janzen, Balasse, and Ambrose 2020;Otaola, Ugan, and Gil 2018;Pilaar Birch 2013;Zavodny et al. 2014).In the Andes, a number of archaeological dietary stable isotopic studies have focused on timing the agricultural reliance on maize (Finucane, Agurto, and Isbell 2006;Gil et al. 2014;Hastorf 1990;Hastorf and Johannessen 1993;Killian Galván, Samec, and Panarello 2016;Somerville et al. 2015;Tung and Knudson 2018), characterising the mobility strategies between the coast and interior valleys (Knudson, Aufderheide, and Buikstra 2007;Santana-Sagredo et al. 2019;Santana-Sagredo et al. 2021;Torres-Rouff, Pestle, and Gallardo 2012), and identifying camelid pastoralism in selected valleys of the Pacific coast (Dufour et al. 2014;Gayo et al. 2020;Szpak et al. 2016;Szpak and Valenzuela 2020).As these and other studies have allowed a greater appreciation of human subsistence diversity in relation to environmental and socioeconomic variability, various questions that could be addressed with these methods emerged (Dufour et al. 2020;Samec et al. 2020;Szpak et al. 2020;Takigami et al. 2020).For instance, how was animal husbandry as initially practiced by highland communities complemented by other subsistence activities, and how did it change over time?
Here we use stable isotope analysis to assess the nature of economic resource utilisation, subsistence and mobility by early pastoralist communities in the Bolivian Central Altiplano, a large high-elevation grassland, initially occupied by mobile foragers specialised in hunting wild camelids and eventually replaced by domestic camelid herders (Capriles 2014;Capriles 2017).The transition between foraging and herding occurred approximately 3500 years ago when an archaeological cultural complex known as Wankarani emerged and dispersed across the Central Altiplano (Bermann and Estévez Castillo 1995;Capriles 2011;Fox 2010;McAndrews 2005).Although increasing archaeological research focused on Wankarani sites has allowed a better understanding of its economic and settlement patterns, how this early pastoralist society led to successful adaptations to the harsh conditions of the Andean highlands, including its degree of reliance on wild and cultivated resources, remains uncertain.

Andean Pastoralism and Interregional Complementarity
Pastoralism is an economic subsistence system that is fundamentally, but not exclusively, based on the management, production and consumption of herding animals (Capriles and Tripcevich 2016;Chang and Koster 1986).The incorporation of domesticated animals marks a distinctive transition in human-animal relations and pastoralist societies vary substantially in terms of animals involved, degree of mobility and sociopolitical organisation (Marshall and Hildebrand 2002).Most pastoralist societies explicitly choose to differentiate themselves from farming and foraging communities because their primary subsistence activities are focused on feeding and safeguarding their herds in relation to environmental variability (Khazanov 1984).Nevertheless, the actual reliance, consumption and dependence of pastoralists on their domesticated herds varies cross-culturally (Cribb 1991).On one hand, specialised pastoralist societies focus on producing primary or secondary products and services such as meat, dairy, fat, wool, transportation and dung used for fuel and fertiliser, and therefore, often engage in symbiotic relationships with agricultural communities for exchanging complementary subsistence staple foods and products.On the other hand, generalist pastoralists often supplement animal husbandry with various other subsistence activities including crop cultivation, opportunistic hunting and gathering and even trade.Although most pastoralist societies fit somewhere in between these two extremes, the specialisation gradient serves as a helpful analytical framework for evaluating the degree of socioeconomic dependence that specific societies developed in relation to their animal herds as well as their eventual transformations over time (Marshall 1990).
In the Andes, ethnohistoric documentation suggests a form of specialised pastoralism developed as part of vertical complementarity, a socio-ecological system through which certain polities accessed resources from different ecological zones by means of population outposts.For instance, the Lupaqa settled in the Northern Altiplano around Lake Titicaca, defined themselves as pastoralists, and managed some of the largest herds in the Andes as well as differentiated themselves from the Uru fishermen of the lake, who were marginalised precisely because of their reliance on fish, waterfowl and other wild foods (Murra 1968).The Lupaqa also had permanent colonies in the inter-Andean valleys to the east and west of the lake to guarantee reliable access to essential goods not grown in the highlands, such as maize, chili peppers and coca (Murra 2002).Similarly, various other highland groups specialised in camelid pastoralism accessed non-highland goods through direct exchange facilitated by trains of llama caravans (Browman 1990).These societies were largely transformed after the Spanish conquest, but highly specialised Andean pastoralist communities persisted over time (Browman 1987).For instance, the herders of Paratia, a community situated west of Lake Titicaca and above the cultivation zone, had to either exchange or sell their camelid-derived products to supplement their subsistence (Flores-Ochoa 1979).Nevertheless, it remains largely unknown how these socio-ecological systems developed and transformed over time in the deeper archaeological past.
Located in the Bolivian Andes, the Central Altiplano extends between 14°and 22°south latitude, 66°and 71°west longitude and 3600 and 4200 m above sea level and is demarcated by a series of mountain ranges, highland plains and wetlands that drain in the Desaguadero River and Lake Poopó (Bermann and Estévez Castillo 1995;Capriles, Calla Maldonado, and Albarracin-Jordan 2011;Fox 2007).The climate of the Central Altiplano is semi-arid and marked by a strong contrast between the dry (May-October) and wet (November-April) seasons, which varied considerably in the course of the Holocene (Rigsby et al. 2005).Currently, the local vegetation is dominated by tussock grasslands and scrubby shrublands in addition to aquatic plants in marshes and bodies of water (Cuenca Sempertegui et al. 2005).
Archaeological research suggests that during the Archaic Period (8000-1500 BCE) under prevailing arid conditions, the highlands were occupied by dispersed bands of hunter-gatherers (Capriles et al. 2018;Osorio et al. 2017).The Wankarani archaeological cultural complex emerged in the Central Altiplano during the subsequent Early Formative Period (1500-500 BCE) featuring the establishment of mounded settlements containing overlapping residential areas with abundant domestic refuse including fragmented ceramics, bone and stone tools (Bermann and Estévez Castillo 1995;Ponce Sanginés 1970).The Wankarani are often cited as an example of a specialised herding society but evidence for limited engagement with chenopod and tuber dryland cultivation has been reported (Langlie et al. 2011;McAndrews 2005).During the Late Formative Period (500 BCE-500 CE), the dispersed settlements of the Wankarani persisted and might have contributed to the pastoralist foundation of the Tiwanaku state, which expanded into the Central Altiplano between 500 and 1100 CE (Beaule 2002;Bermann and Estévez Castillo 1993;Browman 1997;Janusek 2008).Following the disintegration of the Tiwanaku state, during the Late Intermediate Period or LIP (1100-1450 CE), the Central Altiplano was occupied by decentralised lineagebased pastoralist communities coalesced into larger segmentary polities, which were subsequently incorporated by the Inca through both conquest and diplomatic alliances (Catacora et al. 2002;Cruz et al. 2017;Michel López 2008).
Ethnohistoric research suggests that during the XVI century Spanish conquest, the Central Altiplano was populated by large Aymara speaking pastoralist polities subjugated by the Inca such as the Carangas, Soras and Quillacas (Gyarmati and Condarco Castellón 2014;Medinacelli 2010;Pärssinen 2005).Facilitated by llama caravans, these specialised herding communities exchanged staple dietary highland goods such as dry meat and salt for lowland foods such as maize and chili peppers, with farming populations settled in lowland valleys towards the east and west of the Altiplano (Lima Tórrez 2014; Núñez and Dillehay 1995; Saintenoy, González Estefane, and Uribe Rodríguez 2019).Although ethnohistoric accounts mention the establishment of highland colonies in these valleys (Durston and Hidalgo 1997), it remains uncertain when these patterns of social and resource complementarity emerged.

Stable Isotope Analyses and Subsistence Reconstruction
To improve our current understanding of the development of early camelid pastoralist communities in the Andean highlands we used dietary stable isotope analyses.Variation in stable carbon isotope ratios (δ 13 C) can be used to distinguish between the different photosynthetic pathways of plants.Most (Farquhar, Ehleringer, and Hubick 1989;Kohn 2010;O'Leary 1988).The Andean landscape has plants from all three of these categories; whereas C 3 plants predominate in the highland Andes, C 4 plants are present with increasing frequency as altitude decreases, and stable isotope values can indicate which resources were used by these pastoralist communities over time (Berryman 2010;Thornton et al. 2011;Yacobaccio, Morales, and Samec 2009).By studying carbon isotope values from camelids and other fauna we can characterise their diets, and in turn the carbon isotope values of humans relate to the plants people consumed as well as the diets of the animals they consumed.
Archaeobotanical data from the Altiplano focuses on the plant species recovered from human settlements and may not represent the diversity of taxa consumed by pastoral peoples if they had highly specialised/narrow diets, frequently moved across different ecosystems and/or utilised more wild species than agriculturally focused communities.For example, remains from chenopods and tubers (C 3 Andean highland domesticates) dominate archaeobotanical assemblages from the Formative Periods within the Lake Titicaca Basin (Bruno 2014), with maize (an introduced C 4 plant) appearing in notable quantities only beginning in the Late Formative and particularly increasing during the Tiwanaku period (Logan, Hastorf, and Pearsall 2012).Within the Andean context, maize has been given exceptional attention due to its sociopolitical uses in multiple, large-scale polities (Wari, Tiwanaku, Chimu, Inka).Importantly, maize cannot be grown in most areas of the Altiplano environment, but it was actively exchanged into the Tiwanaku center during its height (Hastorf et al. 2006).The presence of enriched carbon isotope values for camelids and pastoral peoples from the Tiwanaku period could indicate the incorporation of maize into their diets but may also point to the use of other wild C 4 or CAM plant species.However, the presence of C 4 or CAM signatures in human samples from the Formative Periods may indicate the importance of wild plant species in pastoral diets.
To contextualise our faunal and human isotopic data, we rely on baseline data collected from other Bolivian highland studies.For instance, a small number of modern plant samples from the Lake Poopó region (elevation 3686 m) provides some guidance for expected carbon values for plants grown in the Central Altiplano.Three Chenopodium quinoa samples have a mean δ 13 C of −25.6‰ and a standard deviation (SD) of 1.1‰, that clearly separate from four Zea mays samples (mean δ 13 C = −12.3‰,SD = 0.5‰), along the expected C 3 /C 4 isotopic groupings (Miller et al. 2021).Similarly, a set of modern and a few archaeological plant samples from the southern Lake Titicaca basin (Chiripa, Bolivia; elevation 3850 m) suggests highland crops follow the expected photosynthetic pathway structure of modern C 3 plants and have low carbon values, such as tubers (Solanum tuberosum and Oxalis tuberosa; mean δ 13 C = −26.7‰),chenopods (Chenopodium spp.; mean δ 13 C = −23.9‰)and beans (Fabaceae; mean δ 13 C = −26.9‰)(Miller et al. 2021).Amaranth species have been reported to utilise either C 3 or C 4 pathways (Cadwallader et al. 2012) and two Lake Titicaca regional samples exhibited values typical of C 3 plants (modern Amaranthus δ 13 C = −28.3‰;archaeological Amaranthus δ 13 C = −22.8‰)(Miller 2005;Miller et al. 2021).The modern Cactaceae species sampled, which are expected to use the CAM pathway, all show quite high δ 13 C values (Cactaceae mean δ 13 C = −12.7‰),and the C 4 plant maize, Zea mays, follows with relatively higher carbon isotope values (mean δ 13 C = −11.3‰).These values are for modern plants and therefore their carbon values are ∼1.5‰lower (more negative) than ancient plants would have been due to the Suess effect (modern burning of fossil fuels altering atmospheric carbon) (Keeling 1979).
Additionally, carbon stable isotope values can help distinguish between terrestrial and aquatic habitats: marine and some freshwater fish, including those from the Lake Titicaca basin, can be enriched in the heavier carbon isotope and can have isotopic values similar to C 4 plants (Miller, Capriles, and Hastorf 2010).Consequently, we expect that if Altiplano herders practiced a specialised form of pastoralism with narrow diets based on C 3 plants and C 3 -plant consuming camelids, their δ 13 C values should be characterised as slightly less negative than the camelid average.Generalised forms of pastoralism characterised by a wider dietary range including the consumption of freshwater fish might involve higher δ 13 C values.Furthermore, interregional exchange and C 4 and/or CAM plant consumption should be identifiable by relatively high δ 13 C values.Distinguishing specific resource use can be a challenge if the resources overlap isotopically, which could be the case for many Altiplano freshwater aquatic foods and plants such as maize (Miller, Capriles, and Hastorf 2010), making the incorporation of other data sets such as archaeobotanical and faunal data particularly important, as well as the potential for emerging methodologies to improve our resolution of ancient diets in areas with the possibility of mixed resource use (Miller et al. 2021).
Skeletal tissue isotopic values reflect various dietary components.Collagen, the proteinaceous organic portion of bone and teeth, preferentially incorporates dietary protein contributions whereas apatite carbonate, derived from the inorganic hydroxyapatite matrix of bone and teeth, reflects total dietary sources (Ambrose and Norr 1993;Tieszen and Fagre 1993).An offset of approximately +5‰ in δ 13 C between the organic collagen fraction of bones and teeth and the average consumed diet has been observed (Sullivan and Krueger 1981).The carbon isotopic signature of omnivores reflects both the plant proteins as well as the proteins from any animals consumed, and as a result plants consumed by primary consumers (herbivores) are reflected in the signature of higher trophic level consumers, such as humans.The inorganic fraction of bones and teeth incorporate the carbon from all dietary sources and therefore reflect the range of foods consumed over the period that tissue was created, and this fraction also shows a dietary offset, ranging from +9 to +12‰ (Krueger and Sullivan 1984;Lee-Thorp, Sealy, and van der Merwe 1989).A combination of bone collagen and apatite data could help to further determine if a few or multiple food sources were present in studied samples (Froehle, Kellner, and Schoeninger 2012).
The nitrogen isotope values (δ 15 N) of consumers reflect their position in a food web because δ 15 N values increase by +3-4‰ per trophic level but can also vary in relation to latitude, altitude, climate and phenotypic variability (Schoeninger, DeNiro, and Tauber 1983;Schoeninger and DeNiro 1984).Except for nitrogenfixing flora, most plants have δ 15 N ranges of +2-6‰ (DeNiro and Epstein 1981;Schoeninger and DeNiro 1984).Nevertheless, empirical research suggests that δ 15 N values can also be enriched as a consequence of specific physiological responses to aridity and other microenvironmental conditions (Szpak et al. 2013).Furthermore, the use of fertilisers, such as sea bird guano but also dung, can enrich the nitrogen content of plants, as documented in the north coast of Peru and the Atacama Desert (Santana-Sagredo et al. 2021;Szpak et al. 2012).
Modern plant samples collected from the Lake Poopó region (Miller et al. 2021) were fertilised with llama dung and therefore δ 15 N values are potentially altered due to fertiliser effects, so results should be interpreted with caution in their applicability to past conditions.The three Chenopodium quinoa samples have an average δ 15 N = +8.1‰,but the standard deviation is 2.3‰, while the four maize samples have an average δ 15 N = +8.6‰and a standard deviation of 1.2‰, demonstrating the large variation in nitrogen isotope values possible within the same species, grown in the same area, under the same conditions (Miller et al. 2021).Isotopic analysis of modern plants from Chiripa in the Lake Titicaca basin further demonstrate a wide range of nitrogen isotope values across terrestrial and aquatic plants, with δ 15 N values ranging from −5.0 to +15.2‰, while most modern domesticated crops have δ 15 N values around +7‰ (Miller 2005;Miller et al. 2021).Specifically, tubers had an average δ 15 N of +7.7‰ (Solanum tuberosum and Oxalis tuberosa), chenopods (Chenopodium spp.) average δ 15 N = +7.1‰and maize (Zea mays) average δ 15 N = +7.0‰,while beans, which fix atmospheric nitrogen, have an average δ 15 N of +0.4‰ (Fabaceae) (Miller et al. 2021).Based on this information, we expect that specialised altiplano pastoralists had δ 15 N values reflecting a predominantly terrestrial animal diet.More specifically, if camelid meat was the primary food source, human carbon isotopic values should be a trophic level above the camelid δ 15 N average value.Depending on additional resources consumed, generalised forms of pastoralism could involve proportionally lower or higher δ 15 N values.Finally, significant dietary variations including the consumption of agricultural products cultivated with manure as well as the potential consumption of imported marine resources should be reflected in high δ 15 N values.

Materials and Methods
Archaeological excavations conducted at nine different sites located in the Wankarani hinterland of the Bolivian Central Altiplano provided the human and faunal remains analyzed for this study (Figure 1).Human remains originate from seven sites and faunal remains from four sites.These sites correspond to Archaic, Early Formative, Late Formative, Tiwanaku and Late Intermediate periods.KCH20 is an openair Archaic Period residential camp where a hearth feature containing stone tools and fragmented remains of wild fauna, but no human remains, were recovered and dated to 7000 BCE (Capriles et al. 2018).San Andrés, Chuquiña, La Barca and Pusno are Early Formative Wankarani mound sites containing overlapping circular residential structures built with stone foundations, mud-brick walls, food remains, and occasionally, human burials (Bermann and Estévez Castillo 1995;Fox 2007).KCH11, KCH21 and KCH56 are Wankarani residential bases occupied during the Late Formative Period and contain domestic structures, interior and exterior storage features, and abundant plant and animal remains, including a preponderance of domesticated camelids but also fish, aquatic birds and mid-sized rodents (Capriles 2011;Capriles 2017).Pusno, KCH11 and KCH21 also contained human burials, including some that postdated their residential occupations.Finally, Jachakala is a large, dispersed settlement that articulated interregional exchange during the Tiwanaku Period as suggested by an abundance of basalt hoes from this region that made their way to the Titicaca basin (Beaule 2002;Bermann and Estévez Castillo 1993).
Skeletal samples from humans were selected from available burial contexts and faunal remains derive from sites where detailed zooarchaeological analyses were conducted.Basic bioarchaeological analyses of the human remains were carried out to estimate their sex and age at death (Buikstra and Ubelaker 1994).Most selected human samples were adults (n = 20), but a few juveniles (n = 4) are also present, and almost all human samples are bone, with only two individuals having a tooth sampled.Direct accelerated mass spectrometry (AMS) radiocarbon dating of most human specimens was conducted to precisely place these individuals in time and additional radiocarbon dates on charcoal from contexts near the undated human remains are also included.Radiocarbon dates were calibrated using the SHCal20 calibration curve (Hogg et al. 2020) in Oxcal 4.4.2(Figure 2) (Bronk Ramsey 2020).
Bone and tooth samples were cleaned and prepared for isotopic analysis following published protocols for both collagen and apatite fractions (Koch, Tuross, and Fogel 1997;Sealy et al. 2014;Yoder and Bartelink 2010).Isotope values are described as parts per mil (‰) and presented in delta (δ) notation in relation to the standards of Vienna-Pee Dee Belemnite (VPDB) for δ 13 C and atmospheric nitrogen (AIR) for δ 15 N.The primary focus of this work is on δ 13 C and δ 15 N from bone collagen, but we also discuss δ 13 C results from a sub-set of bone apatite samples.
Isotopic samples were analyzed at two lab facilities.At Duke University's Environmental Stable Isotope Laboratory (DEVIL), bone collagen samples were processed for δ 13 C and δ 15 N using a Carlo Erba Elemental Analyzer with zero-blank autosampler connected to a Conflo III interface, coupled with a Thermo Finnigan Delta Plus XL continuous flow mass spectrometer.Samples were analyzed with international reference materials including IAEA N-1, USGS 25, USGS 26 for nitrogen and NBS-22, Sucrose ANU, and PEFI foil for carbon as well as internal standards Costech acetanilide, Duke urea, Duke sucrose and USGS 26.External precision (relative to reference materials) is approximately ±0.2‰ at one standard deviation while internal precision (relative to internal standards) is generally better.At University of California, Berkeley's Center for Stable Isotope Biogeochemistry (CSIB) bone collagen samples were analyzed for δ 13 C and δ 15 N using a Vario Isotope Cube Elemental Analyzer coupled with an IsoPrime IRMS.Samples were analyzed with NIST standards (SRM 1547 peach leaves, SRM 1577c bovine liver), IAEA-CH6 (sucrose) and internal project standards.Long term precision for δ 13 C is ±0.1‰ and δ 15 N is ±0.15‰.Bone apatite carbonate samples were analyzed for δ 13 C and δ 18 O at UC Berkeley's Laboratory for Environmental and Sedimentary Isotope Geochemistry (LESIG) with NBS19, two lab standards, internal standards and LESIG reported long-term external precision of ±0.04‰.Collagen preservation was assessed by standard measures of percent (%) yield, %C, %N and atomic C:N ratios (DeNiro, Schoeninger, and Hastorf 1985;van Klinken 1999).
To describe food web and ecological relationships we applied single and bivariate statistics, grouping data taxonomically, temporally and spatially following standard representation of dietary stable isotopic data.Furthermore, for human and faunal samples with both organic and inorganic δ 13 C values we evaluate dietary data patterns in relation to controlled-feeding study dietary groups using data published by Froehle, Kellner, and Schoeninger (2010).Finally, we compare our results with existing dietary stable isotopic studies from neighboring regions.

Radiocarbon Dating and Chronology
Our chronology is based on 15 AMS radiocarbon dates directly conducted on human bone from the sites of Chuquiña, Pusno, Jachakala, La Barca and KCH11.In addition, we rely on 11 additional dates of charcoal from contexts associated with other analyzed human and faunal remains from Chuquiña, Pusno, San Andrés, KCH11, KCH21 and KCH56, to provide strong chronological control for our study sample (Table 1).In addition, a direct AMS date on a wild guinea pig bone from KCH20, dating to approximately 7000 BCE, strengthens the chronology of the samples included in this study.Radiocarbon dates in relation to the regional cultural chronology are used to present and discuss the results of the dietary stable isotope research.

Dietary Stable Isotopes
Our collagen carbon (δ 13 C coll ) and nitrogen (δ 15 N coll ) stable isotope dataset corresponds to 23 faunal (Table 2) and 27 human specimens (Table 3) summarised in Table 4. Faunal remains derive from Archaic (n = 4), Late Formative (n = 16) and Tiwanaku (n = 3) period contexts.All faunal samples had good C:N ratios, ranging between 3.0 and 3.6, within the accepted range for biogenic collagen (DeNiro, Schoeninger, and Hastorf 1985;DeNiro & Schoeninger 1983).Human remains correspond to Early Formative (n = 17), Late Formative (n = 5 samples from 2 individuals), Tiwanaku (n = 4) and Late Intermediate (n = 1) Period contexts.All analyzed human samples had good C:N ratios, between 3.0 and 3.3 (Schwarcz and Nahal 2021), but sample P2 from Pusno, produced poor yield and its exceptionally low %C and %N content suggest degradation, and it was removed from subsequent discussion.In addition, we reported 28 paired apatite carbon (δ 13 C ap ), 20 of which also included oxygen (δ 18 O ap ) stable isotopes for most faunal and a few human specimens.The δ 18 O ap values, related to variable sources of water, are reported but not discussed in this text (Knudson 2009).

Animal Resources
Our faunal sample consists of 6 camelids, 4 tuco-tuco gophers, 3 guinea pigs, 5 birds and 5 freshwater fish (Figure 3; Tables 2 and 4).The six camelid remains were identified as domesticated llamas (Lama glama) based on osteometry and contextual evidence (Capriles 2011).The camelid δ 13 C coll values range from −19.6‰ to −15.2‰, with an average of −17‰ and a SD of 1.4‰, indicating most diets were predominantly C 3 -plant-based, with some individuals consuming small amounts of CAM, C 4 plants and/or freshwater resources.The camelid δ 15 N coll values range from 8.1 to 11.1‰ with an average of 9.9‰ (SD = 1.1‰), which indicates the plant resources they generally consumed varied between 4 and 8‰ δ 15 N, which are consistent with both terrestrial and aquatic Andean environments (Juengst et al. 2021;Miller, Capriles, and Hastorf 2010;Pestle et al. 2017).Two mid-sized rodent taxa were sampled for this study: cuy/guinea pigs (Cavia spp.) and tuco-tuco gophers (Ctenomys opimus).Although together these rodents have the largest range of stable isotope data for both carbon and nitrogen, indicating diverse dietary preferences, there are some concrete interspecific and intraspecific patterns.The two most similar Cavia samples, average −21.1‰ for carbon, and 4.5‰ for nitrogen, and most likely represent wild cuy (Cavia tschudii).These specimens were recovered from a feature containing dozens of wild guinea pig bones at KCH20.In contrast, the third guinea pig is from a Tiwanaku period context and had some of the highest δ 13 C (−9.8‰) and δ 15 N (10.8‰)values reported in the assemblage.It is possible that its enriched isotopic composition derives from the consumption of various sources including maize or other C 4 /CAM plants, and possibly gnawing of animal bones, as suggested by the fact that approximately 2.9% of all the camelid remains from the zooarchaeological study included rodent gnawing marks (Capriles 2011).Therefore, this specimen is either a commensal wild cuy or a domesticated guinea pig (Cavia porcellus).
The four tuco-tuco samples also show diversity, with an average δ 13 C coll value of −16‰ (SD = 1.4‰).Their nitrogen values range from 5‰ to almost 14‰.If we exclude the extremely high nitrogen value, the other three samples have an average δ 15 N coll value of 7.5‰ (SD = 1.9‰).This variability and range indicate the ability of tuco-tucos to successfully subsist on different diets that include various food sources (Takenaka et al. 2020).These values might also derive from commensal behaviour, as  these animals were probably snared from the vicinity of the sites to which they were probably attracted as potential sources for food.Two different bird taxa were sampled, teals (Anatidae) and coots (Rallidae).The teals are represented by three samples, two have very similar carbon isotope values (−22.1‰ and −21.2‰) and also very similar nitrogen values (6.5‰ and 6.7‰), while the third has a very different δ 13 C value of −15.1‰ and a δ 15 N value of 8.2‰.The significant difference in isotopic values (and therefore dietary differences) between these birds, suggests that at least two species may be represented.Given that there are over half a dozen different teal species that inhabit highland wetlands of the Central Altiplano including a few migrant speciates, this is not surprising.The two coots have very similar δ 13 C values, which average −18.9‰ and very similar δ 15 N values that average to 8.9‰.The coots generally have higher nitrogen values than the teals suggesting they are regularly consuming resources from a higher trophic level.
All fish specimens were identified as killifishes (Orestias agassii) and while most of them cluster around a δ 13 C coll value of −21‰, one sample had a very high value (−14.6‰),perhaps suggesting a dietary shift with age and/or size of the individual, but most likely, a distinct species.The δ 15 N coll values range from 5.1‰ to 7.0‰ and the average nitrogen value for these fish is 6.2‰, which is comparable with the trophic position of the genus in Lake Titicaca (Miller, Capriles, and Hastorf 2010;Miller et al. 2021).Although Orestias agassii is the only species currently represented in Lake Uru-Uru, zooarchaeological analysis from the study sites identified considerable variation within the size and morphology of killifish specimens (Capriles 2011).

Humans
The Early Formative sample of human remains consists of 17 individuals (n = 14 adults, n = 3 juveniles) from four Wankarani mound sites (Figure 3; Tables 3  and 4).Excluding individual P2 from Pusno, the 16 Early Formative humans have a mean δ 13 C coll of −16.5‰ (SD = 1.6‰) and a mean δ 15 N coll of 10.9‰ (SD = 1.6‰) (Figure 3).Although the Early Formative includes our largest temporal sample, no clear effects were observed for sex or age.Nevertheless, there is plants and/or animal proteins such as some birds and fish.Individuals with high δ 15 N coll values likely consumed more meat in their diets or more animals that were higher on the food chain (such as birds and fish) compared to the individuals on the lower end of the nitrogen data range and whom may have had diets based on plants and herbivores.
The Late Formative period is represented by a small sample with only two adult individuals from the KCH21 site.However, these two people were sampled for multiple tissues (n = 5 samples total) to provide dietary data from different periods during their lifetimes.Individual L1003 was an adult female and we sampled both the crown of her first permanent molar to capture her diet during infancy including breastfeeding and weaning, and then the same tooth was sampled at the root tip to study her later childhood diet.The tooth dentin crown sample has a δ 13 C coll value of −18.5‰ and a δ 15 N coll value of 12.7‰, followed by her tooth dentin root sample having a carbon value of −19‰ and a nitrogen value of 10‰.The 2.7‰ drop in δ 15 N between the crown and root samples is consistent with the weaning process as an infant has breastmilk removed from the diet (Beaumont et al. 2013;Eerkens et al. 2018).A rib bone sample from the same individual, L1003, has a δ 13 C coll value of −18.4‰ and a δ 15 N coll value of 10‰, suggesting that her diet may have been relatively consistent between later childhood and her adult years.The crown from a permanent first molar from the L503 individual had a δ 13 C coll value of −18.2‰ and a δ 15 N coll value of 13.4‰, while its tooth root had a δ 13 C coll value of −18.6‰ and a δ 15 N coll value of 10.7‰.The decline in nitrogen across L503's tooth samples is consistent with the expected change of an infant being weaned off milk.Both individuals' isotopic values from later childhood/adulthood are very similar and are consistent with diets dominated by C 3 plants and animals feeding mostly on C 3 plants.
The four people sampled from the Tiwanaku period were all adults (3 males, 1 female) and show similar carbon values to the Late Formative individuals but differ in their nitrogen values.The Tiwanaku period individuals average δ 13 C coll is −17.5‰(SD = 0.6‰), and average δ 15 N coll is 12.7‰ (SD = 1.4‰).Again, these isotopic values suggest diets dominated by C 3 plants such as tubers and chenopods, and C 3 -consuming animals such as camelids, with possible addition of some small rodents, birds and fish (given the higher nitrogen isotope values observed for these Tiwanaku-period people).Interestingly, the carbon isotope values do not indicate significant consumption of maize (C 4 plant).Finally, a Late Intermediate Period individual from Pusno (P1) included the highest δ 13 C coll (−11.2‰) and one of the highest δ 15 N coll (+14.4‰)values in the sample.Chronologically, this is the most recent sample in our study and its high isotopic values suggest that this individual consumed a higher proportion of CAM or C 4 plants such as maize, and potentially other resources including higher proportions of animal protein, which might signal increased interregional mobility.

Combining Collagen and Apatite Datasets
We sampled the inorganic δ 13 C ap from all 23 faunal specimens as well as from four of the Late Formative and Tiwanaku human individuals (with one, L1003, sampled twice) to further our dietary reconstructions (Figure 4).Adding regression lines to the resulting dataset shows how animal and human diets are dynamically composed of multiple food sources of varying isotopic ranges (Froehle, Kellner, and Schoeninger Table 4. Summarised results of stable isotope analysis by taxa and in the case of humans by temporal period including maximum (Max) and minimum (Min), mean and standard deviation (SD).2010; Kellner and Schoeninger 2007).Our data mostly falls around the lower regression line, indicating that although some C 4 or CAM plants could have occasionally been consumed, C 3 resources dominated protein and non-protein sources for many of these species.Humans are also closer to the C 3 protein line and are clustered to the left of that line, indicating diets that were almost exclusively C 3 foods both in terms of protein sources and their overall diet (carbohydrates, fats and proteins).Interestingly, the camelids are further to the right on the C 3 protein regression line, indicating that their diets had some 'CAM' or 'C 4 ' contributions, but this could be also explained by feeding on aquatic plants (Miller, Capriles, and   Hastorf 2010; Miller et al. 2021).The one outlying sample that lies just above the C 4 /marine protein line is the (likely domesticated) guinea pig.

Discussion
This study provides the first broad assessment of diet based on both human and animal stable isotope analyses for the Andean Central Altiplano.Regarding camelid herds, the specimens included in our study have nitrogen and carbon isotope values consistent with C 3 -plant based diets from the highlands albeit potentially with some amount of C 4 and/or aquatic plants (Thornton et al. 2011).Camelids with a wide range of δ 13 C values have been recorded from a number of Andean sites and are largely attributed to herd management and animal feeding strategies (Berryman 2010;Finucane, Agurto, and Isbell 2006;López et al. 2017;Samec et al. 2020).In contrast, camelids herded in coastal sites have even higher nitrogen and carbon stable isotope values (Dufour et al. 2020;Szpak et al. 2020;Szpak and Valenzuela 2020).If the camelids from this study were used as pack animals or were herded over areas with large altitudinal differences, they may have consumed more C 4 and/or aquatic plants and thus had more enriched δ 13 C and δ 15 N values (Finucane, Agurto, and Isbell 2006;Miller, Capriles, and Hastorf 2010;Szpak and Valenzuela 2020).Nevertheless, the enriched carbon values found in some of the camelid samples studied here may reflect the consumption of various highland resources including CAM succulents and highland aquatic resources (Cadwallader et al. 2012;Miller, Capriles, and Hastorf 2010;Miller et al. 2021).
Regarding human subsistence, based on previous archaeological and isotopic research in the ancient Andes, we hypothesised that if the early Altiplano pastoralist communities specialised in managing and consuming domesticated camelids, their diet breadth should be relatively narrow and show relatively low δ 13 C and δ 15 N values.Generalised pastoralists on the other hand, would have relied on a wider dietary breadth including agricultural and other faunal resources that should have resulted in wider isotopic ranges.Our data largely confirmed that the Central Altiplano early pastoralists had δ 13 C values reflecting a primarily C 3 dependent diet, but with important variability within these choices and likely adaptability over time (Figure 5).Consistent with zooarchaeological and paleoethnobotanical analyses that suggested Central Altiplano early camelid pastoralists utilised a wide range of dietary resources, isotopic results suggest consumption of C 3 herbivores such as domesticated llamas, but also C 3 plants including chenopods (such as quinoa, kañawa) and tubers (such as potatoes and ocas) as well as wild resources.For instance, about a third of the Early Formative people's diets reflect the consumption of birds, fish and other C 4 /CAM/freshwater consumers.Furthermore, a few of the human δ 13 C values, like those from San Andrés (e.g.Sa5: −13.5‰,Sa4: −14.2‰) and a few from Chuquiña (e.g.C20: −13.9‰), suggest significant consumption of 13 C enriched foods, likely as a result of consuming freshwater fish.Certainly, future research should further explore the nature of the observed variability and its potential association with specific intra-regional patterns of mobility, resource availability and the potential effects of climate change in shaping dietary change over time.
Concerning the question about the emergence of interregional complementary exchange identified in the ethnohistoric sources, our data do not provide strong evidence for the regular consumption of maize or any C 4 plants until the end of Tiwanaku.Previous work has also failed to identify consistent evidence of maize in Wankarani settlements and with Figure 5. Carbon and nitrogen stable isotopes of human remains in relation to time and paleoenvironmental data derived from diatom analysis from Lake Titicaca (Bruno et al. 2021).
the exception of a few shell ornaments, archaeofaunal evidence for interregional trade is limited (Capriles 2017).These results further verify paleoethnobotanical research that underscores the importance of locally grown foods over time and the rarity of exotic cultigens, including maize, before the Late Intermediate Period (Bonzani 2021;Langlie and Capriles 2021).Nevertheless, the domesticated guinea pig from KCH21, likely fed with maize, is intriguing and might have been an exotic import.More research might verify if individuals of higher status could have accessed and consumed more maize, perhaps in the form of chicha beer, which at the time, was fundamentally a ritual beverage (Berryman 2010;Hastorf et al. 2006).In contrast, the human sample from Pusno (P1) dated to the very beginning of the Late Intermediate Period, suggests that interregional staple food exchange, including an emphasis on maize, might have begun quite early after the disintegration of Tiwanaku, which is consistent with growth of caravan networks and nodes in oases and valleys towards the interior of the Atacama Desert (Capriles et al. 2021;Santana-Sagredo et al. 2019;Torres-Rouff et al. 2015).Indeed, the consumption of maize and other lowland foods such as chili peppers and squash likely became feasible only after the establishment of strong networks of interregional exchange no earlier than the Late Intermediate Period (Korstanje 2015;Langlie and Capriles 2021;Santana-Sagredo et al. 2021b).

Conclusions
In this study, we reported dietary stable isotopic data from samples of human and animal archaeological skeletal remains from the Central Altiplano of Bolivia.Direct AMS radiocarbon dates verify the samples originate from a temporal range spanning between 1300 BCE and 1200 CE.The relatively broad dietary breadth documented in the human carbon and nitrogen stable isotopes is consistent with previously reported zooarchaeological and paleoethnobotanical data from the region indicating that from early on these groups practiced a generalised form of pastoralism with a strong reliance on camelid meat but also on other wild and domesticated plant and animal foods (Capriles 2014;Langlie and Capriles 2021).These results contrast with the ethnohistorically documented system of vertical complementarity as strong evidence for the consumption of maize is limited to the most recent human sample reported here.Therefore, maize does not seem to have been a significant component of the diet until the LIP confirming interregional exchange for staple goods developed later in time in this region.For the Central Altiplano peoples most dietary protein was derived from herbivore animals such as domesticated llamas, and C 3 cultivated crops such as chenopods and tubers but also the consumption of wild resources including various amounts of freshwater fish, birds and plants.Over time, there is a persistence in this wide range of dietary sources and some temporal fluctuations potentially associated with resource intensification possibly connected to regional mobility patterns, population size and even climate change.
Our data suggest the early Central Altiplano pastoralists relied on a form of generalised pastoralism that mostly involved local plant cultivation and procurement of regionally available wild resources.Nevertheless, unlike the Carangas, Soras, Quillacas, and other ethnohistorically and ethnographically documented herding communities, they did not practice interregional complementarity.Beginning during the Early Formative Period, Wankarani communities structured residential mobility patterns tied to locally available resources.The integration of the central highlands and their local-scale subsistence systems into a larger political economy during the Tiwanaku Period does not seem to have involved strong processes of dietary change, but more research and larger sample sizes are needed to verify this proposition.Settlement patterns, site layouts and artefact analysis from Central Altiplano sites suggest that although the impact of the state was not dramatic, there were some significant differences between the Formative and Tiwanaku periods (Capriles 2017).As an intermediate region between the Titicaca basin, the Pacific western valleys and the tropical eastern valleys, the Central Altiplano herding communities could have benefited from increased traffic and regional interaction (Albarracin-Jordan 2007;Bonzani 2021;Browman 1997;Hastorf et al. 2006;Núñez and Dillehay 1995).Settlements such as Jachakala and KCH11, might have worked as network nodes for llama caravans moving to and from Tiwanaku and other consumption loci such as Cochabamba, Pica or the Azapa valley.Some of the larger sites might have served as gathering places for fairs, where exchanging prestige goods, consumption rituals (possibly involving maize chicha) and incipient political integration occurred.Nevertheless, the dietary stable isotopes along with other lines of archaeological evidence indicate that several of the resources consumed in earlier time periods continued to be utilised with subtle changes in the regional economic subsistence.
Bolivian Ministerio de Culturas y Turismo, and facilitated by the Inti Raymi Mining Company, Universidad Mayor de San Andrés, Universidad Técnica de Oruro, and the people of the communities of Chuquiña, La Joya, Iroco, Cochiraya and Chuzequeri.

Figure 1 .
Figure 1.Study area including the Central Altiplano, Wankarani early pastoralist sites (white circles) and archaeological sites that were sampled for this study (red circles).

Figure 2 .
Figure2.Calibrated radiocarbon dates of samples and contexts associated with this study including sum probability density (SPD) of the Central Altiplano using Oxcal 4.4.2(Bronk Ramsey 2020) and SHCal20(Hogg et al. 2020).

Figure 3 .
Figure 3. Bivariate plot showing carbon and nitrogen stable isotopes of bone collagen categorised by taxa and human samples by period.

Figure 4 .
Figure 4. Bivariate plot showing carbon isotopic results on bone collagen and bone apatite in relation to the predicted food consumption lines (following Froehle, Kellner, and Schoeninger 2010; Kellner and Schoeninger 2007).

Table 2 .
Carbon and nitrogen stable isotope results of faunal remains.

Table 3 .
Carbon and nitrogen stable isotope results of human samples.also a good deal of diversity in the isotopic values recorded within sites.For example, Chuquiña, the site with the largest number of human burials (n = 11), includes δ 13 C coll values that range from −17.9‰ to −13.9‰ and δ 15 N coll values that range from 8.7‰ to 13.9‰.Many of the Early Formative individuals' δ 13 C values cluster around −17‰, suggesting their diets were based almost exclusively on C 3 plants and C 3 consuming herbivores, which is consistent with most of the fauna reported above.A few individuals have values around −14‰, which indicates their diets had significant contributions from CAM or C 4