FROM THE PACIFIC TO THE TROPICAL FORESTS: NETWORKS OF SOCIAL INTERACTION IN THE ATACAMA DESERT, LATE IN THE PLEISTOCENE1 DESDE EL PACÍFICO A LA FORESTA TROPICAL: REDES DE INTERACCIÓN SOCIAL EN EL DESIERTO DE ATACAMA DURANTE EL PLEISTOCENO FINAL

1 An earlier version of this article was presented at the International Workshop “Caravan Archaeologies: En Route to the Past, Present and Future”, Pica, Chile (May 2017). This manuscript was evaluated by external reviewers and edited by the editors of the Chungara Editorial Committee. 2 Instituto de Alta Investigación, Universidad de Tarapacá, Arica, Chile. *Corresponding author: calogero_santoro@yahoo.com; franciscorothhammer@gmail.com 3 Departamento de Oceanografía, Universidad de Concepción, Concepción, Chile. emgayo@uc.cl 4 Centro de Ciencia del Clima y la Resiliencia (CR), Santiago, Chile. 5 Department of Anthropology, The Pennsylvania State University, Pennsylvania, USA. jmcapriles@gmail.com 6 Centro de Estudios Avanzados en Zonas Áridas (CEAZA), Coquimbo, Chile. marcelo.rivadeneira@ceaza.cl 7 Departamento de Biología Marina, Universidad Católica del Norte, Coquimbo, Chile. 8 Departamento de Biología, Universidad de La Serena, La Serena, Chile. 9 Laboratoire de Préhistoire et Technologie, Université Paris Nanterre, Paris, Francia. herreragodoy@hotmail.com 10 Camino al Volcán 33411, San José de Maipo, Región Metropolitana, Chile. valentina.mandakovic@yahoo.es 11 Décima Avenida 1230, depto. 47, San Miguel, Santiago, Chile. monicarallo@yahoo.com. 12 Department of Geology & Environmental Earth Science, Miami University, Oxford, Ohio. rechja@miamioh.edu 13 Departamento de Antropología, Universidad de Tarapacá, Arica, Chile. barbara_cases@yahoo.es; dani.valenzu@gmail.com 14 Museo Municipalidad de Pica, Pica, Chile. geoglifo2@gmail.com 15 Programa Doctorado en Antropología UCN-UTA, Universidad Católica del Norte, San Pedro de Atacama, Chile. olguinlaura.o@gmail.com 16 Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina. laborrero@hotmail.com 17 School of Anthropology, University of Arizona, Tucson, USA. arqueo.paulaugalde@gmail.com 18 Departamento de Ecología & Centro UC Desierto de Atacama, Pontificia Universidad Católica de Chile, Santiago, Chile. clatorre@bio.puc.cl 19 Instituto de Ecología y Biodiversidad (IEB), Santiago, Chile. 20 Department of Anthropology, Trent University, Peterborough, Ontario, Canada. paulszpak@trentu.ca Volumen 51, N° 1, 2019. Páginas 5-25 Chungara Revista de Antropología Chilena

For scattered populations of hunter-gatherers, the need and the ability to maintain networks of social interaction is crucial to assure access to both utilitarian and non-utilitarian goods and to gather survival information, especially during times of shortage (Dillehay 2013;Veth 2008;Whallon 2006). Here we attempt to show that the hunter-gatherer societies that initially inhabited the Pampa del Tamarugal (PdT) basin, an intermediate depression between the Pacific coast and the Andes (Figure 1) in the hyper arid core of the Atacama Desert, needed to develop and maintain extensive mechanisms for the transference of non-utilitarian and utilitarian goods (found in the archaeological record) as well as partners and information, which in turn allowed them to create social networks, essential for the sustainability of their sociocultural systems within hyper arid environments by the end of the Pleistocene. Hence, people that initially settled within the Atacama Desert, albeit likely at low population densities, should have developed strong and dynamic social networks at local, regional and supraregional scales.
Unlike highly institutionalized later Andean circulation (e.g., Dillehay 2013; Dillehay and Núñez 1988;Núñez and Nielsen 2011;Salomon 1985), the transference of items and movement of people among late Pleistocene hunter-gatherer societies could have been operated under less structured systems. Particularly, the transport of low-frequency items may have occurred by means of several mechanisms of both direct and indirect acquisition. Direct mechanisms may include: the incidental acquisition of goods fixed in circuit of residential and logistic mobility (Binford 1979), for instance through transhumance (Dillehay 2013;Lynch 1971); as well as non-utilitarian mobility by long-and short-distance special-purpose trips (Gould and Saggers 1985;Grove 2014;Lovis et al. 2005;Newlander 2012;Pearce 2014;Speth et al. 2013;Whallon 2006). Indirect mechanisms may included: direct and indirect (down-the-line) economic exchange of utilitarian and non-utilitarian items (Dillehay 2013;Laguens et al. 2007;McBryde 1984;Mulvaney 2002;Silveira et al. 2010;Tripcevich 2010;Veth 2008;Wiessner 2002;Wilkins et al. 2010); flux of resources as part of marriages (bride price/ service) (Walker et al. 2011); and incidental deposition of allochthonous goods as a result of visits and other sporadic and non-systematic activities (e.g., Gusinde All of these mechanisms imply social interaction and the setting of social networks, which in turn allowed the procurement of exotic goods of prestige and sacred importance, but also expanding kinship and socio-politic relationships and getting ecological, geographical, and social information for "insurance purposes" (Grove 2014:3573), all of which have evolutionary advantages for social and biological reproduction (Cummings 2014;Dalton 1977;Gamble 1982;Gamble 1999;Gould and Saggers 1985;Hamilton et al. 2007;Newlander 2012;Pearce 2014;Politis 1996;Speth et al. 2013;Whallon 2006). The initial bands of huntergatherers (possibly ca. 25 people), such as those who initially explored the open landscapes of the PdT, required to constantly move and eventually join other groups to avoid extinction (Kelly 1995;Tolan-Smith 2003). Several small interconnected groups moving separately over a large territory increase the likelihood that they can learn more about a region and later share this knowledge, which would clearly improve their survival and adaptation capacities (Kelly 1995(Kelly , 2003Pearce 2014). Societies subject to extreme insecurities of key resources such as water, like those that inhabited the hyper arid ecosystems of the Atacama Desert, were more dependent on long-distance social networks (Gould and Saggers 1985:122).
The problem is the difficulty to identify and distinguish archaeologically direct and indirect mechanisms of acquisition / transfer of goods (Close 2000;Costin 2001). Although the presence of exotic goods in hunter-gatherer contexts is often explained by mobility, there is an increasing archaeological, ethnographic and ethnohistorical recognition that hunter-gatherers practiced various forms of exchange (Bate 1986;Cummings 2014;Dalton 1977;Delage 2018;Dillehay 2013;Gusinde 1991;Jordan 2014;McBryde 1984;Standen et al. 2018;Veth 2008). Our objective is not to discriminate the type of specific mechanism of acquisition / transference of goods that operated among late Pleistocene PdT hunter-gatherers because the availability of material is limited. Rather, based on ongoing interdisciplinary studies that integrated the available archaeological, isotopis and paleoecological data, we attempt to show the different geographical scales in which the networks of interactions functioned within these societies. We propose three kinds of network for resource management and other social material and immaterial requirements: local, regional and supra-regional scales. We aim to reconstruct and discuss how people integrated vastly different territories and eventually built dynamic socio-cultural interaction networks in the Atacama Desert during the late Pleistocene.

The Pampa del Tamarugal
The Atacama Desert expands across the western flank of the Andes, extending from southern Peru to the Copiapó valley (~17°-27° S). Across these >1,400 km, the relief rises from the Pacific coast to the highelevation Altiplano, and is accompanied by an abrupt elevation-precipitation gradient with widespread hyper aridity across the coast and low-elevation areas (0-2,000 masl), arid climates prevail along the pre-Andean cordillera (2,000-4,000 masl), with semi-arid but cold conditions over the Altiplano (>3,500 masl).
The PdT is a multi-aquifer system of more than 4,000 km 2 that lies within the central valley in the hyper arid core of the Atacama Desert. Here, the extreme negative hydrological balance leads to a landscape that is practically devoid of biotic and water resources. Nevertheless, the localized discharge of surficial and groundwater resources generates discrete oases for life. In fact, within deeply incised valleys (or quebradas) that drain the precordillera are dense riparian ecosystems that occur alongside perennial/ephemeral streams (Gajardo 1994;Luebert and Pliscoff 2006;Villagrán et al. 1999). Typically, these ecosystems in the northern Atacama (17º-21º S) incorporate endemic and native herbaceous and woody taxa such as Escallonia angustifolia, Baccharis spp., Morella pavonis (syn. Myrica pavonis), and Schinus molle, among others. Except for a few rodents, lizards, birds and amphibious species, the native faunal assemblage in these formations is limited (Marquet et al. 1998). Within the central valley, the localized evaporation and outcropping from shallow groundwater table maintains salt pans (salares) along the western margin of the PDT with phreato-halophytic vegetation consisting of Prosopis trees (P. tamarugo and P. alba), shrubs (Tessaria absinthioides) and herbaceous taxa (i.e., Distichlis spicata, Atriplex atacamensis) (Briner 1985;Luebert and Pliscoff 2006).
Paleoecological reconstructions for the ecological and hydrological setting of our study area show that during the Pleistocene-Holocene transition (13,000-9,000 cal yr BP) the hyper arid core of the Atacama Desert was a hospitable environment for humans. During this period, the PdT contained key resources (fresh water, plants, and game, plus good quality lithic raw materials) for the settling of small groups of huntergatherers that centered their life within this ecosystem. The PdT may have served also as an ecological corridor connecting five different ecosystems on the western Andean slope: (a) the Pacific coast and Cordillera de la Costa 60-80 km to the west of the PdT, (b) the quebradas that drain the Andes 60-70 km to the east, (c) the Altiplano in the high Andes 80-150 km to the east, (d) the exoreic valleys to the north, and (e) the Loa River and the Salar de Atacama basin towards the south. Ecosystems between the Pacific coast and the tropical forests were more continuous than today, facilitating a long distance network of interaction and making transcontinental movement less difficult, both of which were crucial during this epoch of very low population density .
Indeed, late Pleistocene human occupations of the PdT coincided in large part with a boom in local ecosystem productivity that arose from wetter conditions over the Andes (Gayo et al. 2012;Nester et al. 2007;Pfeiffer et al. 2018). These conditions occurred during the second stage of the Central Andean Pluvial Event (CAPE II; 13,000 -9,800 cal yr BP, (Gayo et al. 2012), in which amplified precipitation in the high Andes induced significant positive hydrological budgets within the hyper arid core of the Atacama between 18° and 25°S (Gayo et al. 2012;Nester et al. 2007;Sáez et al. 2016). As a result, excess runoff from the highlands produced fertile green oases with abundant biotic resources along the low-elevation desert (Gayo et al. 2012;Nester et al. 2007). The associated archaeological sites we have studied were synchronous with CAPE II, although paleoclimate data suggest equivalent paleoecological and hydroclimatic conditions during the first CAPE stage (17,500-14,200 cal yr BP) yet no archaeology date to this period has been identified. Anomalous earlier and later dates of marine shell, wood, and plant remains are not considered within this time frame because of the marine reservoir effect and the presence of old -subfossil-wood (Table 1; Herrera 2017;Joly et al. 2017;Latorre et al. 2013).

Networks of Interaction
The archaeological evidence summarized below is presented to discuss multi-scalar networking practiced by hunter-gatherers occupying the PdT during the end of the Pleistocene, which we tentatively define as local (within the PdT), regional (coastal and highland), and supra-regional (trans-Andean) networks of interaction. Archaeological data come from one quarry (Cerro Chipana) linked to a workshop open camp in Quebrada Chipana (Chipana-1), and several open camps, all of which have been excavated and well-dated at the distal section of the alluvial fans of Quebrada Guatacondo: Pampa Ramaditas-4 (PR4), Pampa Ramaditas-5 (PR5), Pampa Ramaditas-7 (PR7), and Quebrada Maní: Quebrada Maní 12 (QM12), Quebrada Maní 32 (QM32), Quebrada Maní 35 (QM35), and Salar de Llamara 17 (SLl17). Forty-eight AMS dates have been obtained from these sites, excluding the last two sites (not yet dated), and these fluctuate between 12,420 to 9370 cal yr BP. These sites show that these people were well-adapted to the Atacama Desert as they contain abundant stone tools of different local and extra-local raw material and technology, processed plant and animal fibers, modified sea shells, and faunal remains associated, in some cases, with hearths and other features.

Local network
The local network corresponds to a social space that has been archaeologically identified within the southern section of the PdT circumscribed between the mouth and proximal section of the fan of the quebradas that drain into the eastern margin of this basin, and their distal section. Within a radius of 15 to 20 km we have inventoried five large archaeological areas: (1) the proximal section of the alluvial fan of Quebrada Maní (QM), (2) the distal section of the alluvial fan of the same quebrada and the canyon to the north -Pintados, (3) the distal alluvial fan section section of Quebrada Guatacondo known as Pampa Ramaditas (PR), (4) the quarry and work shop of Quebrada and Cerro Chipana (Chi) and (5) Salar de Llamara (SLl), all of them covered with open camps numbered as QM12, QM32, QM35 PR4, PR5, PR7, Chi-1, and SLl17, (Figure 1). The camps were part of a complex socioeconomic mobility system that took advantage of local resources generated by the and contemporaneous paleowetland and riparian ecosystems where the sites used to be located, which were available in the PdT during CAPE II (Boyd and Richerson 2005;.
The hunter-gatherers who inhabited these sites manifested technical knowledge and a concomitant socio-economic and technological organization, which included a wide range of lithic reduction techniques encompassing both simple and highly complex attributes, and the management of locally available resources, listed below:

(a) Lithic raw material
Silicified fine-grained cherts, fine-grained sedimentary dark rocks (lutites), and quartzitic sandstone  This study Herrera 2017 Herrera 2017 Herrera 2017 Herrera 2017 Herrera 2017 This study Herrera 2017 This study Herrera 2017 This study Herrera 2017Herrera 2017Latorre et al. 2013 This study sourced from the Altos de Pica Formation (in the northeast section of the PdT) are of good knapping quality and are commonly found on the surface at camp localities and dispersed over the landscape (Latorre et al. 2013). Another common lithic raw material is a white opaline silica (Blanco and Tomlinson 2013) that was quarried and collected from Cerro Chipana, a huge quarry site next to Chipana-1 (Figure 2). At this particular workshop raw material was transformed into advanced stages of large biface artifacts (Herrera 2017), which were then exported to neighboring sites like PR5, PR7, QM12, QM32, QM35 and reworked; creating a circulation of material between 20 and 30 km from the Cerro Chipana quarry.

(b) Animal resources
Bone archaeological remains show that large artiodactyls, including guanaco (Lama guanicoe) and possibly vicuña (Vicugna vicugna), along with medium-sized rodents such as tuco tuco (Ctenomys sp.) and chinchilla rats (Abrocoma sp.), as well as a few birds were consumed. It is possible that these animals were locally available and hence locally hunted, as they could be attracted to the paleowetlands, making the camps strategic locations for foraging wild resources. Some of the camelid remains consist of neonates, suggesting that hunters might have been present during the early wet season (December-January) when these animals are born. This does not rule out the entry of fauna from the highlands, as is suggested by isotope analyses on animal fiberdiscussed below -that indicate highlands origin for some of the bone remains. Moreover, the fragmented nature of the faunal assemblages suggests intensive consumption of the animals as well as the weathering of the bones by salts as these were exposed to hyper arid conditions.

(c) Plant fiber yarn
Yarn and other remains of processed plant fibers were found at excavated stratigraphic levels from QM32. A few elements (n=8) of cotton and another unidentified plant fiber, locally cut and gathered were employed to prepare knots, twisted fiber and Z spun yarn, made through further processing or disintegration of the fiber. One sample of cotton (monochrome strand Z) was isotopically analyzed and had a δ13C value of −24.1 ‰. Relative to cotton fibers from the Azapa valley dated to the Formative Period (average δ 13 C = −23.2 ‰) this lower value for the single PdT fiber may be indicative of the wetter conditions in this region as discussed previously. The δ 13 C values of C3 plants (such as cotton) are sensitive to water availability, with low water availability producing relatively high (less negative) δ13C values in plant tissues (Farquhar and Richards 1984;Stewart et al. 1995).

(e) Pigments
Pigments were also circulated within the PdT, although their origins are still unknown. No chemical analyses have been carried out to identify their origin and compare them with the chemical composition of the early pigments mined from the interior of the coast of Taltal south of the PdT (Salazar et al. 2011), and contemporaneous with their camps.

Regional networks
Based on material introduced from the coast and the high Andes, found in small quantities in most of the open camps and the workshop-quarry, we estimate that people from the PdT organized or were a part of regional social network of interaction with access directly or indirectly to the Pacific coast and the Andes ( Figure  1). We also estimate that the acquisition of foreign items was embedded or was part of more complex social interactions, not materialized in archaeological material, which are discussed below.
west. From this network hunter-gatherers obtained a selection of shells from marine benthic mollusks and cobbles, that were modified.
(a) Shells The following small shells have been found in the PdT: Nassarius gayi (cf. Priene scabrum), cf. Leukoma thaca, Crepipatella dilatata, and litorinids ( Figure 3). Fissurella crassa and Concholepas concholepas of larger size have also been identified. These ecofacts and artifacts were found in the stratigraphy and the surface of QM12, QM32, and QM35. In Chipana-1 we found only C. concholepas. Additionally, the small species Felicioliva peruviana, cf. P. scabrum littorinids was found on the surface of all of these sites except Chipana. At PR4, PR5, and PR7 sites, no marine shells have been identified. With the exception of C. concholepas and L. thaca these species are for the most part infrequently found in Holocene archaeological shell middens of the hyper arid coast of northern Chile (Bird 1943). Yet, they are common in modern natural assemblages (Guzmán et al. 1998;Laudien et al. 2007), as well as in the Pleistocene fossil record of Peru and northern Chile Paskoff et al. 1995;Rivadeneira and Carmona 2008).
N. gayi ( Figure 3a) is a non-edible small snail (up to 14 mm in length) that inhabits sandy and muddy sediments from 0-15 m depth with a modern distribution from Isla Lobos de Afuera in Peru (6ºS), to Cape of Horn in southern Chile (55ºS) (Guzmán et al. 1998;Marincovich 1973;Reid and Osorio 2000). Its presence in coastal shell middens of northern Chile has been described for Taltal between 8,000-5,600 cal yr BP (Olguín 2013). This is the most represented mollusk at the QM12 site (n=30). All complete specimens have a distinctive human-made hole in the body of the shell, and were probably used as ornaments. One of these was directly dated to 12,200 cal yr BP (a 400 years reservoir correction was applied to the dated sample) (Latorre et al. 2013), and was coeval with the rest of the dated materials of the site.
F. peruviana (= Oliva peruviana, Figure 3b) is a nonedible taxon that inhabits subtidal water in soft bottoms, between 4-10 m depth, with a modern distribution from Sechura Bay, Peru (5ºS) to Lota in southern Chile (37ºS). These taxa reach up to 55 mm in length (Guzmán et al. L. thaca (= Prothaca thaca, Figure 3c) is an edible taxon that inhabits soft bottoms, from the intertidal to 15 m depth, reaching sizes up to 108 mm length (Huber 2010). This species is distributed from Chicama, Peru (7ºS) to the Chonos Archipelago in southern Chile (43ºS) (Alamo and Valdivieso 1997). This taxon is currently exploited by the artisanal fleet and consumed by humans along the coast of Chile (Jerez and Figueroa 2008). The oldest record of this species in shell middens of Chile is ca. 9,000 cal yr BP, in Los Vilos in central Chile (Báez et al. 2003), but its presence in shell middens of northern Chile starts at ca. 7100 cal yr BP (I-9817) in Camarones (Niemeyer and Schiappacasse 1979).
C. concholepas or loco, the Chilean abalone, is an edible taxon native to the Pacific coast of Chile and Peru with a modern distribution between 5ºS to 55ºS. The species inhabits the low intertidal zone to 50 m depth (Häussermann and Försterra 2009) and reaches sizes up to 179 mm in length (Vermeij 2012). The oldest record of loco in the shell middens of Chile is between 12,570-12,140 cal yr BP at the Los Rieles site in central Chile (Jackson et al. 2012).

(b) Cobbles
Rounded, and often highly elongated cobbles (Figure 4), were most likely selected by humans from Pacific coastal environments and transported to the PdT by direct or indirect mechanisms to be used as hammer instruments due to their hardness. These cobbles are highly polished, likely due to abrasion by waves on the coast. Such well-rounded and hard cobbles are uncommon in the PdT as large and prolonged stream discharge events, critical for rounding and polishing stones, are infrequent. The PdT is too proximal to the adjacent Andes for such cobbles to form. Indeed, the majority of well-rounded cobbles found profusely in the PdT streams are typically comprised of soft ignimbrite lithologies, which are unusable as hammer stones.

Highland network
Andean regional networks of interactions resulted in human transportation of raw materials (obsidian, chalcedony, and camelid fiber) to the PdT sites. Moreover, highland projectile point types have been identified, indicating that the acquisition of physical material may have been linked to the transferring of certain technological knowledge, among other social businesses.

(a) Obsidian
Obsidian is a high-quality volcanic raw material typically found in the highlands. At least five different obsidian sources have been characterized in the neighboring southern Altiplano, including Pelun, Salar de Tara, Cerro Kaskio, Laguna Blanca/Zapaleri and Caldera Vilama (Capriles et al. 2018;Seelenfreund et al. 2010;Yacobaccio et al. 2004). The closest known obsidian source to Quebrada Maní is Cerro Kaskio, situated approximately 150 km to the east. Cueva Bautista, a high-elevation temporary camp -located at the same latitude of our study area-was occupied around the same time as the late Pleistocene occupations of the PdT. It is located 15 km away from Cerro Kaskio, and contains abundant evidence of Cerro Kaskio obsidian (Capriles et al. 2016).
Energy-dispersed x-ray fluorescence and instrumental neutron activation analyses suggest that the QM12 obsidian flake and a distal fragment of a triangular blade projectile point from SLl17 correspond to two different but still unidentified obsidian sources. Furthermore, a retouched obsidian flake from a somewhat later (early Holocene) human occupation in Quebrada Blanca, an open camp located between the PdT and Cueva Bautista , was sourced to Salar de Tara (Figure 1).
The date of obsidian flake from QM12, as it was recovered from the A horizon beneath a desert pavement, would fluctuate between ca. 12,200 and 11,900 cal yr BP, corresponding to the temporal period defined for the site. At SLl17 (a site located in Salar de Llamara, a late Pleistocene fresh water basin, hydrographically linked to Quebrada Mani, and currently a dry salt pan) the blade of a large obsidian projectile point, of still unknown chronology, was recovered from the surface, embedded in the saltpan next to other lithic debitage of different raw materials (Figure 5a). As these former lagoon ecosystems collapsed at ca. 10,000 to 9000 cal yr BP (Pfeiffer et al. 2018), it is likely that this obsidian point was brought before that time, roughly contemporaneous or slightly later than the occupation of QM12. Transportation of small quantities of obsidian from the high Andes was not rare at open camps in Pacific coastal region of southern Peru (Rademaker et al. 2013), contemporaneous with QM12, which are thought to be part of exchange networks that included the circulation of processed fish (Reitz et al. 2016).
Similarly, south of the PdT, late Pleistocene hunter-gatherers living on pre-Andean ecological zones (2000-3800 masl) also obtained obsidian from highland environments linked to paleolake habitats that were gradually colonized based on a seasonal network of interaction (Grosjean et al. 2005;Loyola et al. 2017;Núñez et al. 2002;Núñez et al. 2018).

(b) Chalcedony
Chalcedony is another raw material with no local sources, which is consistent with its rare occurrence in PdT archaeological sites. We found on the surface of PR5 the base of a projectile point manufactured on a fine-grained brownish lustrous chalcedony (Figure 5b). A few other flakes of this same exogenous raw material have been found in QM32, QM34 and QM35. The second most abundant raw material in the late Pleistocene occupation of Cueva Bautista in the highlands of Bolivia, following Cerro Kaskio obsidian, is a fine-grained jasper-like lustrous chalcedony that ranges in color from beige to burgundy. This chalcedony is macroscopically very similar to the raw material found in the PdT, but as chalcedony can be macroscopically similar and come from different sources, further analyses are required to confirm that beyond its rarity, it was a highland raw material.

(c) Projectile point types
Another proxy for the Andean networks of interaction is represented by projectile point types (Patapatane and Tuina) found both in stratigraphy and the surface from open camps in the PdT: QM12, Chipana-1, and QM35. These projectile point types are common in Early Archaic sites in the highlands of northern Chile. The Patapatane type is common in Andean archaeological sites from southern Peru and northern Chile (e.g., Asana, Caru, Patapatane, Hakenasa, and Las Cuevas). The Tuina type form is common in archaeological sites from the Central Atacama (e.g., Tuina, Tambillo, Punta Negra, and Alero El Pescador sites). It has been also recognized in northwestern Argentina (Herrera 2017;Herrera et al. 2015;Latorre et al. 2013;Núñez et al. 2016;Osorio, Steele, et al. 2017;Santoro and Núñez 1987).

(e) Yarn of animal fiber
Yarn and other remains of processed animal fibers were found at excavated stratigraphic levels from QM32. The major technical innovation was the processing of animal fleece and the production of primary textile structures, such as yarn, felted fleece, and other evidence of animal fiber processing (n=126). Raw materials may have included fibers from guanaco, vicuña, and rodents, along with human hair. Technical action involved cut-shearing, as shown in fiber with few modifications related to procurement or extraction possibly from animal hides; felted fleece, felt stitching, fleece spinning, and hand spinning and twisting or plying, as shown by single yarns, 2 ply yarns and knotted threads. Specifically, a sample of 2-ply yarn found in situ in QM32 was dated to 11,270 cal yr BP (Table 1).
Carbon isotopic analyses of these fibers (fleece and yarn) show a very high contribution of C3 plants, consistent with a highland origin. Considering the prevailing wetter conditions in the high Andes at this time (Gayo et al. 2012), we would expect the Altiplano to be characterized by very few C4 plants (Still et al. 2003). We analyzed 12 samples of animal fiber, 10 possibly camelids (vicuña and guanaco) and 2 rodents (chinchilla or vizcacha). The fiber samples analyzed include 3 yarn, 5 fleece, 2 twisted fleece, 1 felted fleece, 1 stitched fleece with remains of red pigment. Comparatively, camelid bone collagen from Tiwanaku in the Lake Titicaca basin produced comparable δ13C and δ15N values relative to these animal fibers from PdT (Szpak 2013). Adjusting for hair-collagen differences in isotopic discrimination (Szpak et al. 2014) the δ13C values range between −21.0 and −18.2 ‰ and the δ15N values range between +6.2 and +8.7 ‰ ( Figure 6, and Supplementary Material 1).

Trans-Andean network
Data for a trans-Andean connection comes from a piece of wood (ecofacts) found in layer 3 of QM32 (the larger one measuring 8 cm long and weighing 21.6 gr.). The poor-preservation prevented the determination of whether this pieces formed part of any artifact. There is no direct dates for this wood fragments, but several radiocarbon samples from stratigraphic layer 3 of QM32 site were dated to the late Pleistocene (Table 1).
Anatomical analysis on the larger wood fragment identified it as exotic silk floss tree (cf. Ceiba spp. very similar to Ceiba speciose; Figure 7). Although, a tangential section was observed due to the poor condition of the tissue, it was possible to identify the following anatomical characteristics: simple vessel perforations plates, alternate inter-vessel pits, polygonal shape of alternate pits, fibers of very thin walled, large rays commonly > 10 seriated, ray height > 1mm, rays with procumbent, square and upright cells mixed throughout the ray, sheet cells (IAWA-Committee 1989;Richter and Dallwitz 2000 onwards;Tortorelli 1956). These anatomical features are unequivocally attributable to Ceiba spp.
Ceiba spp. (silk floss tree or palo borracho) is a native Malvaceae (subfamily Bombaicodeae) species from the Atlantic seasonal semi-deciduous forest and tropical savannas from eastern South America, where mean annual rainfall ranges between 1,500 and 1,000 mm/yr, and the mean annual temperature exceeds 18°C (Cruz Ruggierom et al. 2002;Killeen et al. 1998). Although paleoclimate data suggest a two-to-four-fold increase in rainfalls during the CAPE II event (Gayo et al. 2012;Latorre et al. 2006;Pfeiffer et al. 2018), the magnitude for such change was not enough to support local populations of Ceiba spp. either at the PdT or over the western Andean slope (Figure 7). But more importantly, paleoclimate transformations in the PdT were related to reduced temperatures and positive hydroclimate anomalies brought about amplified surface and groundwater discharge instead of increased local precipitation (Gayo et al. 2012;Pfeiffer et al. 2018). The fact that hyper arid conditions (i.e. no local rainfalls) remained stable during the CAPE II at PdT, then the record of Ceiba spp. woodfragments in the archaeological context of QM32 necessarily implies intended anthropogenic transport.

Discussion
Traveling, networking for moving objects, people, and ideas have been key social interactive behaviors in human history (Clarkson et al. 2017;Dillehay 2013;Dillian and White 2010;Hirth and Pillsbury 2013). In Sout America, it seems that since the early colonization of the continent people were linked to a broad network that directly and indirectly connected human groups from a wide range of dispersed ecosystems. This strategy allowed them to get access to a variety of exotic utilitarian (subsistence) and non-utilitarian resources, which in turn helped them to cope with environmental instabilities and food shortages; to reinforce social linkages, political alliances, and kinship extension (including mate exchange); and to avoid conflicts; all of which contributed to social reproduction. The value of transporting non-utilitarian and utilitarian goods is not intrinsic to the good itself. After all, this analytical dichotomy between "necessity and sumptuary" resources is culturally mediated. As the flow of utilitarian and non-utilitarian resources and objects are embedded in social and cultural practices, the decisions of movement do not only concern subsistence (Gould and Saggers 1985;Lovis et al. 2005;Newlander 2012;Politis 1996;Whallon 2006). Local and exotic goods found in the late Pleistocene archaeological record of the PdT were used for a variety of social purposes and functions, including food, tools, personal adornment, and other objects of unknown function.
Almost all the mollusks found are small and many have small holes, which suggests that they were not a primary food source, but selected for other purposes, such as ornamental and ritual with the exceptions of C. concholepas, Fissurella spp., and L. thaca. The small size of N. gayi likely precluded its use as a food source. It was, however, an ideal primary material for body adornment (Correa 2016). F. peruviana was used as a funerary offering in the Late Archaic Tiliviche cemetery (Standen and Núñez 1984), and became very common since the Formative period ca. 3600 cal yr BP onwards in funerary contexts of northern Chile. These items were brought from the Pacific coast 60-80 km away, and its special purpose is also suggested by its low frequency. Other item brought from the coast, but with utilitarian purpose were the hard-rounded cobbles. They were Figure 6. Carbon and nitrogen isotopic compositions of animal fiber samples from Pampa del Tamarugal. For comparative purposes, the standard bivariate ellipse for camelid bone collagen from Tiwanaku is shown to represent camelids consuming a high proportion of C3 plants. The δ 13 C values of the Tiwanaku camelid bone collagen have been adjusted by -1.3 ‰ to account for the difference in diet-tissue discrimination between hair and bone collagen as described in the text.
Composición isotópica de carbono y nitrógeno de muestras de fibras animales de Pampa del Tamarugal. Con fines comparativos, se muestra la elipse bivariada estándar de colágeno de huesos de camélido de Tiwanaku, que consumían una alta proporción de plantas C3. Los valores de δ 13 C del colágeno óseo de Tiwanaku se han ajustado en -1,3 ‰ para dar cuenta de la diferencia en la discriminación entre el tejido y la dieta entre el cabello y el colágeno óseo como se describe en el texto. used as percussion tools in preparing stone artifacts. Suitable stones for hammering are not geologically available in the PdT.
Items from the highlands included obsidian, chalcedony, and camelid fiber, also found in small quantities. As good lithic raw material is locally available, the introduction of foreign stones was not conditioned by the absence of this kind of supply. Animal fibers were also locally available but, according to isotope analyses, some fiber for the production of primary textile structures came from the highlands. Given that different stages of the operational sequence for yarn production were identified, we deduce that the raw material was imported and then locally processed. This interpretation is reinforced by the presence of textile structures made out of local plant fiber (cotton and other unknown taxon), which indicates that these people had the technological knowledge to produce threads, the primary structure for fabrics.
The piece of wood of a tree species (Ceiba spp.) that today occurs in the tropical and subtropical forests to the east of the Andes, ca. 600 km away, is an item that does not fill the absence of wood raw material. By the end of the Pleistocene the PdT had a good coverage of trees with appropriate wood for tool making. Besides, Ceiba spp. provides a soft and light wood which function in this context is unknown.
The particular set of exotic items found in the PdT, although in small quantities, imply that people were involved in broad social networks that may have included exchange, logistic movements, marriages, and sporadic visits, among others. For the acquisition of these materials, they needed to establish links with other populations, and therefore, with other cultural worlds, languages, landscapes, and ideologies.
Like some contemporaneous hunter-gatherers (Politis 1996), the late Pleistocene hunter-gatherers in the PdT needed to manage different scales of territory, that ranged from a local social territory well known by a band, to very distant lands, known and visited by possibly few members of the group. Within this vast territory, people obtained their supplies from different resources to be incorporated into their socio-cultural subsystems. The mechanisms to access and obtain goods, nevertheless, remain a dilemma. As it has been noticed by several investigators the detection of exotic or imported goods is not enough to distinguish between direct access by mobility and indirect access by exchange (e.g., Costin 2001, Close 2000, Hodder 1982Pallo and Borrero 2015). As far as we know, however, archaeological indicators evaluated to distinguish both mechanisms are not conclusive. For example, the ratio between the distance to the source of origin and the curve of frequency of the imported objects, from short distances versus objects of long-distance sources do not necessarily represent the extremes of mobility and exchange patterns, respectively. Australian ethnographic and archaeological cases, among others, provide abundant data that show that long-distance movements covering several hundred km were practiced to obtain socially valuable objects (and therefore not plentiful), along with short and longdistance exchanges for other important items (Gould and Saggers 1985;McBryde 1984;Mulvaney 2002;Veth 2008). Next, we hypothesize the possible ways these mechanisms occurred.
For the local network, the current archaeological data from the PdT allow us to suggest that people managed three basic levels of territorial information as suggested by Rockman (2003:4): locational, limitational and social knowledge. At the locational level, people certainly recognize the places and physical characteristics of local resources for subsistence (i.e., lithic outcrops, wood, game, plant, wetlands). At the limitational level people were familiar with the range of distribution and reliability of local resources, the costs of acquisition, and its usefulness. At the third level, the PdT people created a social knowledge, as they settled and moved throughout the landscape revisiting more regularly camps located in the proximal section of the fluvial fan of the quebradas that drain into the PdT. Their familiarity with this territory increased substantially, conceptually defining the components they recognized, those that were utilized and those that were not, creating a complex human landscape. Due to the predicted low population density in the PdT by the end of the Pleistocene, we estimate that the procurement, translation and integration of local resources by hunter-gatherer societies could have occurred through both logistical and residential mobility.
Regional networks of interaction required different sorts of knowledge and movements. Information would likely have been gathered from social networks linked to other social groups, settled along the Pacific coast or the high Andes. The introduction of low quantities of high-quality Andean lithic raw materials may have occurred through mobility of the PdT groups to those external ecosystems. Alternatively, people may have accessed to the exotic items through large-scale network of exchange or down-the-line trade. For obsidian found in late Pleistocene open camps of the southern Peruvian Pacific coast, which originated more than 130 km toward the Andes, it has been suggested that it was obtained through the exchange of processed fish within a regional network (Rademaker et al. 2013;Reitz et al. 2016). Exchange or down-the-line trade in intermediate territories is a feature assigned to well-established Holocene hunter-gatherers in South America (Dillehay 1999). Considering that the long distance transported items had no food value, and some of them were nonutilitarian and were low frequency items, it may indicate that the purpose for making the effort of accessing coastal and Andean goods was probably motived by social gatherings (e.g., marriages and information exchange). This means that in these and other social gatherings, exotic items could be exchanged to reinforce social bonds that were crucial for the sustainability of late Pleistocene hunter-gatherers living in the shifting and challenging ecosystems of the PdT. The finding of exotic material itself, however, should not be considered as conclusive evidence to distinguish between exchange and direct access. Nonetheless, it should be considered, however, that less than 100 km separate the PdT and the Pacific coast to the west, and the Andes to the east, thus it may be possible that human groups directly reached the resources of these ecosystem through logistic mobility circuit (Figure 1). This interpretation would need research with a larger geographic scope.
Late Pleistocene hunter gatherers of the PdT managed to obtain resources located beyond the Andes into the tropical lowlands, as is indicated by the taxonomic identification of tropical wood (Ceiba spp.). In other words, it seems likely, that people were part of larger, complex long-distance networks that enabled the circulation of such materials. More importantly, though, was the possibility of the existence of extended networks of interaction for sharing knowledge and for establishing social bonds beyond the band level. Very recent genetic research that examined South American population structure and pre-Hispanic population displacements, utilizing mitochondrial DNA haplogroups of 4200 subjects of extant mixed populations of Mexico, Costa Rica, Venezuela, Colombia, Ecuador, Peru, Bolivia, Brazil, Argentina and Chile, showed that the territories of southern Peru, northern Chile and highland Bolivia were colonized by genetic flows from the Amazon basin (Rothhammer et al. 2017), which in turn reinforces the (Lathrap 1970) hypothesis that the upper Amazon was peopled by upstream population movements using the Amazon river network as a waterway.

Conclusions
The archaeological evidence presented here shows that various exotic, utilitarian, non-utilitarian and edible goods were introduced and consumed in the PdT (lithic raw materials, wood, animal and plant fibers for tool elaboration, and mollusks for food and other purposes). They were carried from the Pacific coast, the Andean highland and the tropical forest, and were integrated to the resources gathered locally within the PdT. The specific mechanisms for the circulation of these items are uncertain. The distribution of material may have occurred through long-term incursions of people from the PdT, and the encounter with individuals from other ecosystems inside or outside the PdT. These encounters could have happened within rather informal networks of interaction, as suggested by the scarcity of the exotic objects. Connections with the tropical forest, more than 600 km away, were the consequence of even larger networks of interaction that involved several groups moving within and between ecosystems. Although most of the identified resources could be classified within the "utilitarian" world, there are elements that do not hold nutritional value or cannot be used as raw materials for manufacturing tools. Thus, they were carried because of other social purposes and functions. Besides the circulation of exotic items, the encounter with other peoples should have served to share knowledge and experience on plants and animal behavior, geography, information in managing raw materials and technological practices, as well as marriage arrangements, collaborative social actions. Moreover, the down-the-line transportation of goods should have propitiated also the migration of women and men that interacted along the way from the tropical forest to the Pacific and vice versa.
One of the marks of the functioning of a long-range system of interaction is that beyond the economically necessary transported goods, the distribution of alternative items is also served. Their significance, however, is not obvious, as they may be related to spiritual or ritual necessities or may be the basis for different classes of alliances. The maintenance of local, regional, and supra-regional networks of interaction should have required different kinds of mobility practices involved in the social life of the groups who initially colonized the Atacama Desert. It should be mentioned that rainfall in the highlands during the late Pleistocene was two or three times more abundant than today. These climatic conditions reduced the stark discontinuities of the ecological zones involved in the networks of interaction utilized by hunter-gatherers between the Pacific coast and the tropical forests. Fresh water and natural routes such as rivers and available plants and animals would have made travel more feasible. As connectivity was better during the late Pleistocene, it is likely that the continuous landscapes with different resources facilitated a pan-Andean connection beyond the Altiplano. In other words, the late Pleistocene-early Holocene inhabitants of the PdT had the possibility to be part of a broad exchange network that allowed them access to a diversified list of rather conspicuous items brought from both the eastern and western slopes of the Andes.
In sum, the circuit of movement implied in the circulation of the exotic items is still uncertain, but it can be imagined that people at least managed a local circuit of mobility within the PdT of ~200 km in length (north to south) and ~100 km in width (east to west). Connections with the external world such as the coast (ca. 60-80 km to the west) and the highland (ca. 80-150 km to the east) may have involved direct incursions to these ecosystems or the encounter with other groups of people in intermediate territories. Connections with the tropical forest, more than 600 km away, were possibly the consequence of down-the-line exchange networks that involved several groups in the circulation. Explanations for the introduction of exotic "non-utilitarian" objects, with non-dietary value, into the PdT would be that: (a) as exotic goods they had intangible meanings and were used within socio-cultural performances (i.e., birth and death of group member; marriage; investiture of a new leader; healings); (b) by forming part of these and other behavioral patterns, the action of introduction of these material objects allowed them to create bonds with outside groups to cope with social strategies for the survival and reproduction of the groups; (c) exchange included also sharing meanings imbedded in these objects and transferring information about the outside world. Archaeological excavations in the studied sites were authorized by the Consejo de Monumentos Nacionales: Ord. Nº4379/12, Nº003352/13, Nº001478/15, Nº2203/17, and Nº4379/12.

Stable Isotope Analysis -Sample Preparation
Samples were cleaned of any visible foreign matter with tweezers and then sonicated in Type I water for 1 h, then in 2:1 chloroform:methanol for 1 h. After removing the solvent the samples were air dried under normal atmosphere for at least 24 h.

Stable Isotope Analysis -Instrumentation and Analytical Uncertainty
Carbon and nitrogen isotopic and elemental compositions were determined using a Nu Horizon continuous flow isotope ratio mass spectrometer. Sample measurements were calibrated relative to VPDB (δ 13 C) and AIR (δ 15 N) using USGS40 and USGS41a (Table S1; Qi et al., 2003;Qi et al., 2016). The following standards were used to monitor accuracy and precision (Table S2). The isotopic compositions for the internal keratin standard represent long-term averages.  Table S3 summarizes the mean and standard deviation of carbon and nitrogen isotopic compositions for all check standards, as well as the standard deviation for all calibration standards -the mean of the calibration standard for an individual run is predetermined to calibrate the data. One sample (#529) was analyzed in duplicate. The difference between the duplicate measurements was 0.15 ‰ for δ 13 C and 0.14 ‰ for δ 15 N.

Analytical Uncertainty
Standard uncertainty was calculated using the method present by Szpak et al. (2017), which largely follows Magnusson et al.'s (2012) approach. Standard uncertainty was determined to be ±0.16 ‰ for δ 13 C and ±0.15 ‰ for δ 15 N.

Results
The carbon and nitrogen isotopic and elemental compositions of all samples analyzed are presented in Table S4.