Ancient Conflagration: a Reconstruction of a Middle Formative Fire at the Chiripa Mound, Bolivia

Abstract

In this study, we assess competing interpretations of a burnt ceremonial structure from the terminal Middle Formative period (ca. 300–100 BCE) by analyzing the stepped platform mound at Chiripa, Bolivia, through a systematic reconstruction of the fire that destroyed it. We developed a model of potential fire pathways, their social contexts, and material indicators. Our approach contrasts incipient fires from accident or arson to planned fires initiated for functional or social ends. We assessed these pathways for the Chiripa mound fire through experimental, geoarchaeological, faunal, and botanical data. Experiments were aimed at deducing the temperature, duration, and oxidation conditions of the fire. The fire temperature and duration were approximated from geoarchaeological analyses of construction materials in comparison with controls, and thermal alteration of faunal bone. Fuels were reconstructed through paleoethnobotanical analysis of charred remains from discrete areas within the burnt structure. We conclude that an intentional fire burned the structures on the Chiripa mound to temperatures of 700 °C or higher under oxidizing conditions for several hours. The pattern of heat-altered materials recovered would have required a substantial supplemental fuel load. At the 3840 masl elevation of Chiripa, the effective control of a high temperature oxidizing fire demonstrates technical expertise in fire management. Our findings indicate the fire appears intentional, likely a ritual event. We believe the structures were burned to facilitate a socio-political change during a period of social transition at the end of the Middle Formative period in Bolivia.

Introduction

The recovery of a heavily burnt occupation deposit is an evocative—even visceral—experience. As Tringham (2005, p.107) details, encountering charred remains can conjure images of a structure up in flames, of people drawn to or alarmed by the display, smell, and noise of the conflagration. In 2006, excavators with the Taraco Archaeological Project (TAP) excavating on the ceremonial mound at Chiripa, Bolivia, revealed layers of ash, concentrations of charcoal, and bright heat-reddened walling from a small earthen structure that inspired just such an image of a building destroyed by flames (Hastorf et al., 2008). Prior research at the Chiripa mound also recovered other burning levels in the quadrangle of structures surrounding a central recessed courtyard from the final Middle Formative period (ca. 300–100 BCE)(Bandy, 1999b; Bennett, 1936; Browman, 1978a, b, 1989; Hastorf, 1999a; Kidder, 1956; Whitehead, 1999).

The Middle Formative Chiripa community was among the largest in the southern Lake Titicaca region (Bandy, 2001, 2004a), and enjoyed access to a wide range of food stuffs, luxury goods, and imports (Bandy, 2004b; Browman, 1989; Chávez Mohr, 1988; Hastorf, 2003; Hastorf et al., 2008; Moore et al., 1999). While there are sunken courts built across the landscape during this time, the stepped platform mound at Chiripa is a unique locale in this region. It is therefore puzzling why some of the ceremonial structures on the mound were burned when the settlement appeared to be thriving. After the fire, the circle of structures was not rebuilt, but infilled to create a level surface surrounding the interior plaza (Hastorf, 1999a; Hastorf et al., 2008; Kidder, 1956). This leads us to query if the structures were destroyed from purposeful or accidental action.

This article investigates the origins and burning conditions of the fire that engulfed one structure on the Chiripa mound, Structure 4. We discuss the implications of this fire from competing interpretations and their cultural meanings for the Chiripa residents. In order to test potential interpretative scenarios, we conducted open air and laboratory experiments to determine the material effects of different burning conditions. We analyzed walling, botanical remains, and bone from Structure 4 to reconstruct the temperature, duration, and oxidation conditions of the fire. We further analyzed botanical remains to reconstruct potential fuel sources. These data allow us to reconstruct the fire and assess the potential significance of the blaze in Bolivian prehistory.

Burning Models: Interpreting Structural Fires in the Archaeological Record

Archaeologists have developed interpretative criteria to assess the cultural significance of fires in regions where burnt structures are a common feature of the archaeological record (e.g., Cameron, 1990; Chapman, 1999a, b, Demarest et al., 1997; Gheorghiu, 2019; Harrison, 2008; Inomata, 1997; Ivanova, 2007; Johnston et al., 2019; Stevanović, 1997; Tringham, 2000, 2005, 2013; Twiss et al., 2008; Verhoeven, 2000, 2002). Much of this literature centers on house fires. Here we adapt these arguments to the civic-ceremonial context of the Chiripa mound by omitting arguments specific to domestic occupancy. Archaeological considerations of burnt structures fall into two general categories: incipient fires, where the blaze started by accident or arson, and planned fires, where the fire was set alight on purpose for functional or cultural purposes. We emphasize planning rather than intentional fire (contra Tringham, 2005) because discerning evidence for what a planned event intended in the past remains a considerable research challenge. Incipient structural fires are deduced when the only fuel source is flammable architecture and the structure’s contents. Incipient fires are also indicated when the distribution of the fire, and the structure’s contents, appear unplanned by the occupants. Both accidental and incipient arson fires are expected to burn unevenly and may show evidence of being quenched. During an incipient fire, occupants may not have time to remove items from the burning structure, and burned artifacts may therefore represent normal use.

Experimental burning studies provide an excellent source of information for reconstructing the temperature, origin, and fuel for archaeological fires. Drawing on these studies, we can identify the degree of planning that went into a fire. Experimental burning of thatch roofed structures provides a baseline for our expectations for the behavior of the roof fire at Chiripa. A rectilinear wattle and daub structure with a thatched roof in South Africa burned to over 1000 °C for 7 min, and the roof collapsed after 20 min (Friede & Steel, 1980). In the same study, a round thatch structure burned for significantly longer, had glowing ash after 24 h, and was still hot 48 h later. The earthen walls of Chiripa structures make a wattle and daub experiment an applicable comparison to this study.

In a European experiment, a wattle and daub house with a decaying thatch roof was ignited from an internal hearth, and within 20 min, the roof thatch was consumed and collapsed (Bankoff & Winter, 1979). The wooden roof beams smoldered along the house perimeter for over 30 h, but the earthen walling was largely unaffected, and little fired daub was produced. A similar fire pattern was found in experiments with thatch roofed earthen structures in the North American Southwest (Icove et al., 2006). Thatch roofs burned through and quickly fell while earthen walls and floors showed little alteration. Icove and colleagues (Icove et al., 2006) concluded that burning thatch roofing was insufficient to alter earthen architectural elements in their color or chemistry. They suggest that large, additional fuel loads would be required to produce the archaeological patterns they observed including oxidization patterns, fire scars on walls, and sintered earthen materials. In an experimental study in Ukraine where timber was used as kindling, almost nine times more wood was needed to burn the structures than was needed to build them initially (Gaydarska & Chapman, 2020; Johnston, 2020).

Where a roof remains intact, heat will concentrate and develop characteristic fire patterns that move down the wall as long as the structure holds (Harrison, 2013; Icove et al., 2006). These patterns may be referred to as a heating front or thermal zone. The depth of the heat front down the wall is an indicator of fire intensity in a closed fire. In the experimental burning studies, the thatch roofs fell quickly, and heat largely emanated upwards to the upper portions of the rooms, out the top, and only then into the lower walls or floors.

Sediments change structure in response to heat by first sintering as adjacent particles merge ca. 400–600 °C, and then vitrification between 700 and 950 °C when crystalline materials melt (Rye, 1981, pp.106–109; Carter & Norton, 2013, 438–45). Sintering and vitrification produce a stable reduction of pore space, and alter soil-forming minerals, particularly those with metal oxides (Mataix-Solera et al., 2011). The resultant fired earth aggregates are commonly recovered from water flotation processing. In contrast to the preparation of construction materials, evidence for fire management and preparation of the room contents can bolster interpretations of ritually prepared fires.

Accidental fires are evinced when the fire appears to emerge from an established burning installation (Harrison, 2013; Twiss et al., 2008). At Formative Chiripa, cooking fires were generally kept between 100 and 300 °C to avoid charring food (Moore et al., 1999). At these lower temperatures, fires produce characteristic charcoal, soot, and ash residues (Mentzer, 2014). A hearth fire can get out of control when the fire extends to roofing. Mudbrick structures have high thermal inertia which limits heat conduction making heated adobe unlikely to propagate fires (Harrison, 2013). However, soot build-up or unanticipated flares can spread fire upwards towards the roof, which may leave a visible trail. In an abandoned contemporary house on the Taraco Peninsula, a domestic hearth produced a trail of soot almost 1 cm thick from the hearth to a small window near the roof, but the adobe bricks underneath were barely altered (Goodman Elgar, 2008). If sparks or flames ignited this soot, it could carry the blaze upwards to the flammable thatch roof.

In contrast to accidental hearth fires, arson fires generally originate from torching flammable structural components from outside the structure, notably the roof. Archaeologists have considered burnt sites, particularly houses, as evidence for arson arising from conflict in diverse areas including North America (Lambert, 2002; Snead, 2008), Mesoamerica (Demarest et al., 1997; Inomata, 1997, 2014), Eastern Europe (Chapman, 1999b; Ivanova, 2007; Johnston et al., 2019), and also the Northern Titicaca Basin (Levine, 2013; Stanish & Levine, 2011). Arson may be suspected when a burnt structure contains culturally valuable items that are generally curated in other contexts. Arsonists may torch multiple disconnected buildings and the extent of a fire can therefore also support interpretations of arson (Levine, 2013; Stanish & Levine, 2011; Tringham, 2013). However, incipient fires may also start from unusual events such as a lightning strike. In a thatch-roofed mudbrick building, a lightning strike could ignite the roof and may be indistinguishable from arson.

In contrast to incipient fires, planned fires enable the participants to adjust fuels to reach higher temperatures and achieve more complete combustion than incipient fires (Johnston, 2020; Shevchenko & Gaydarska 2020; Stevanović, 1997). Room contents need not represent normal activities as items may be added or removed prior to burning. Reasons for intentional burning include practical issues such as fumigation, improving future construction, and social concerns related to ritual renewal that draw on the transformative power of fire (Chapman, 1999a; Tringham, 2005, 2013). Fires for fumigation would presumably concentrate in storage features with burning located in discrete, predictable areas. In order to prevent such fires from getting out of control, the fuel load would be relatively small, and the temperature and duration also controlled.

On anthropogenic mounds, a mudbrick structure may be burned to prepare a foundation to support the next building level, or to reuse fired architectural elements in future building projects (Chapman, 1999a; Shaffer, 1993). Fires intended to improve the structural properties of mudbrick require high temperatures to promote partial melting of the minerals, which necessitates a significant fuel load (Johnston, 2020; Kruger, 2015). In this interpretation, the burned layer will underlie subsequent occupation. Likewise, fired structural fragments in new constructions support an argument of burning to create construction materials.

Finally, planned burning of ceremonial structures may be a component of a society’s ritual life cycle. In this model, a planned burn will end a structure’s use-life. This has been extensively explored in Neolithic Europe where researchers apply this model of ritual burning to houses (e.g., Bailey, 1990; Chapman, 1999a; Stevanović, 1997; Tringham, 2000, 2005, 2013). For instance, Bailey (1990, p. 28) states, “one may speak of the life-cycle of a house: a house is born, it lives, it dies, it is buried or cremated, and its spirit is remembered after its death.” Interpretations of ritual closure in Neolithic houses emphasize the performative aspect of house burning (Tringham, 2005) and the sacrifice of material goods (Chapman, 1999a). These activities are most often related to memorializing activities. Given that the Chiripa mound structures are not interpreted as domestic houses, this model may be adapted to this architecture. The structures surrounding the sunken court were regularly burned, and replastered each time, like a cleaning process (Bandy, 1999b, 2001; Hastorf, 2003). The larger scale fires that consumed Structure 4 may also be ascribed a ritual cleansing purpose if we can substantiate a planned fire.

In order to disentangle the various interpretative possibilities of the Chiripa fires, properties such as temperature, evidence for management, fire origin, fuel sources, multiple ignition points, and the spatial distribution of burnt remains must be garnered (Icove et al., 2006, Shevchenko & Gaydarska, 2020). From Table 1, we can predict that incipient fires can be differentiated from planned fires by the absence of evidence for added fuels, fire management, or preparation of the structure before burning. An interpretation of arson is favored where the fire appears to have originated outside the structure, while accidents are more likely when the fire centers on an internal combustion feature. Cooking fires have notably low temperatures and higher temperatures are considered accidental. Fumigation fires should be concentrated in storage features. In contrast, a fire to prepare construction materials requires temperatures high enough to alter the sediment across the structure and also evidence for use of heat altered materials.

Table 1 Characteristics of incipient and planned fires for prehistoric structures

The Chiripa Mound on the Taraco Peninsula

The Taraco Peninsula, Bolivia, curves towards the west into the southern Wiñaymarka section of Lake Titicaca (Fig. 1). The semiarid grasslands surrounding the lake provide a resource-rich environment within the broader Altiplano region with lacustrine resources, freshwater springs, arable loess soils, and grazing lands (Argollo et al., 1996). The circum-Titicaca region, located in the northern part of the Altiplano, became increasingly populated during the Formative period (1500 BCE–400 CE, Table 2), or the Andean Neolithic. During this time, residents settled in villages, and developed a mixed subsistence base that included rainfed farming of chenopods, highland tubers and wild plants (Browman, 1989; Bruno & Whitehead, 2003; Bruno, 2008, 2014; Miller et al., 2021; Whitehead, 2006, 2007), camelid pastoralism, hunting, and fishing (Browman, 1989; Capriles et al., 2007, 2014; Moore et al., 1999, 2010).

Fig. 1

Lake Titicaca basin. The sites of Chiripa and Kala Uyuni are on the Taraco Peninsula

Table 2 Southern Lake Titicaca basin Cultural Chronology. Based on Bayesian analysis within the site of Chiripa using stratigraphy and ceramics. Dates calibrated with ShCal13 curve. Late Formative chronology refined in Marsh et al. (2019), the Early and Middle Formative chronology refined by Capriles (Capriles & Hastorf, n.d.)

Over the Formative period, settled agrarian communities became densely clustered around the lake from ca. 1500 BCE (see Bandy & Hastorf, 2007; Hastorf, 2006; Janusek, 2008; Stanish, 2003; Fig. 1; Table 2). Settlements flourished on the Taraco Peninsula during the Early and Middle Formative period and the Chiripa stepped platform mound was a central venue (Bandy, 2001, 2004a; Beck, 2004; Hastorf, 2003). We believe the Chiripa settlement emerged as a local center during the Early to Middle Formative period with evidence of a dense residential population, sunken court, and a stepped platform mound (Bennett, 1936; Browman, 1978a, b, 1989; Chávez Mohr, 1988; Hastorf, 1999a, b, 2003; Kidder, 1956; Portugal Ortiz, 1975). Mounds denote multi-lineage gathering spots during this time throughout the Andean region and are rare in the Titicaca basin, making Chiripa stand out as a Formative period ceremonial center. The Formative period ends with the precipitous growth of the nearby Tiwanaku center, in a valley south of Lake Titicaca, a reduction in the population size at the Chiripa settlement, and the adoption of Tiwanaku practices on the mound (Hastorf, 1999a, b; Janusek, 2004, 2008; Kolata, 2003; Stanish, 2003).

The Chiripa stepped platform mound that rises five meters above the surrounding landform (Bennett, 1936; Hastorf et al., 2008). The mound’s construction history begins with a circle of rough stone and mud structures detached and irregular, called lower “houses” (Bennett, 1936). These were filled in with sediment and a second group of 14 clearly coordinated, well-built structures were built around the central court, again out of mudbrick and stone. This second group of small rectilinear structures was initially called “upper houses” by Bennett (1936) (Kidder, 1956; Fig. 2). To avoid an overemphasis on domesticity, we refer to these buildings on the mound as structures, rather than houses, as their construction and contents are unlike domestic structures from the Andean Formative. The upper structures on the mound have unusual architecture with small enclosures or “bins” interspersed with niches around the interior walls (Fig. 2). Their assemblages include polychrome and incised ceramics, trumpets, food stuffs, and sub-floor burials. These assemblages lead Chávez Mohr (1988) to suggest that the mound was a temple or ceremonial space, which has been confirmed by subsequent investigations (Bandy, 2001; Hastorf, 1999a, 2003; Roddick & Hastorf, 2010; Roddick, 2002). Burials on the mound concentrate in fill between the two construction phases (Bennett, 1936; Blom & Bandy, 1999; Machicado Murillo, 2009).

Fig. 2

Idealized reconstruction of the second level of the Chiripa mound structures (after Janusek, 2008, Fig. 3.5, and Hastorf et al., 2008, Fig. 3.1). Kidder labeled these structures with numbers and letters (see Bandy, 1999a). The materials analyzed in this study derive from the excavation of Structure 4 on the center left and highlighted with a red border (Hastorf et al., 2008)

During its use-life, the Chiripa mound contained the largest and most complex architecture yet recovered from the southern Titicaca Basin. For comparison, the nearby contemporary settlement of Alto Pukara was about half the size of Chiripa and contained a leveled hilltop with only two simple rectilinear buildings (Beck, 2004). In the northern Lake Titicaca Basin, the Late Formative sites of Taraco (Peru) were larger and had more substantial stone architecture than Chiripa (Levine, 2012, 2013; Plourde & Stanish, 2006) (Fig. 1). Bandy (2001, 2006; see also Hastorf, 2003) proposed that the structures surrounding the sunken court on both levels of the Chiripa mound were associated with lineages or other segmentary groups who engaged in ceremonies, communal feasting, and exchange of exotic goods to leverage their engagement throughout the greater region. Beck called these house societies, again supporting this sense of self-identifying groups coming together physically in these structures (2004). There is no clear surface evidence for social hierarchy, but rather redundancy, as is also noted at many ceremonial centers throughout the Andes, from Caral to Aspero and La Galgada. While each excavated structure at Chiripa was very similar architecturally, they were pronounced differences in subfloor grave goods. Burials within the structures may have commemorated the death of important members of these lineages (Machicado Murillo, 2009).

The stratigraphic sequence in excavated lower mound structures includes superimposed plastered floors and walls. The floors have a repetitive sequence of thin layers of yellow clay over a thin cultural fill or sand, upon which a shallow fire was lit, and subsequently a new floor layer or clay and sand was laid to start the sequence again (Bandy, 1999a, 43). In one structure, this sequence was repeated eight times, leading Bandy to suggest that this regularity represents a ritual practice associated with floor/interior renewal (1999a, 45). An example of this sequence has been observed in a thin section of Structure G where a shallow in situ fire comprised a 1.5 cm sequence of charcoal and ash lenses with artifact fragments aligned horizontally across the floor, over which a very thin layer of sand was laid (Goodman Elgar, 1999). The burning events in these sequences are thin 1–2 cm layers with abundant charcoal and ash. The preservation of charcoal indicates that the fires in the structures did not exceed 500 °C when carbon was converted to CO₂ (Heiri et al., 2001). Midden deposits were placed on the mound exteriors between the structures and have notable charcoal lenses (Bandy, 1999b; Hastorf et al., 2008). We therefore conclude that small-scale burning was a common practice for both the structure’s interiors and their more public exteriors.

Structure 4

We focus here on excavations from Structure 4 in the upper level of the mound at Chiripa. Several other similar structures from this level have been excavated since the 1930s with some evidence of burning noted (Fig. 2; Bennett, 1936; Browman, 1978a, b, 1989; Chávez Mohr, 1988; Hastorf, 1999a; Hastorf et al., 2008; Kidder, 1956; Portugal Ortiz, 1975). However, the burning events were never investigated nor described. Differences in recording between projects limit direct comparison of previously excavated structures. Following the fire, or fires, in this upper level of the mound, the structures were infilled and covered to encircle part of the mound to create a unified surface. This later surface surrounded the central sunken court during the late Middle Formative phase, and was resurfaced two more times in that same phase (Bandy, 1999b; Hastorf et al., 2008). During the subsequent Middle Horizon phase, the residential population of Chiripa waned, and settlement shifted to the southern side of the Taraco Peninsula and beyond to the site of Tiwanaku (Bandy & Hastorf, 2007; Bandy, 2001, 2004a; Janusek, 2008).

The upper-level structures are substantial constructions, elaborated with painted walls, internal niches, plaster floors, and unusual, slotted doorways with sliding doors (Bennett, 1936; Browman, 1978a; Chávez Mohr, 1988; Hastorf, 1999a; Hastorf et al., 2008; Kidder, 1956; Portugal Ortiz, 1975). In Structure 4, the yellow plaster floor is tempered with coarse plant fragments (Davenport, 2010), which is similar to fiber-tempered Early Chiripa ceramics (Steadman, 1999). The upper structures only opened onto a surface overlooking the central sunken court (Fig. 2). The small interiors of these structures could only accommodate a few people and the sliding doors further indicate hidden, controlled access (Bandy, 2001, 2004a). The surrounding small chambers, or bins, within each of the 14 structures are too small for an adult to enter (Bennett, 1936; Hastorf et al., 2008; Kidder, 1956).

Several structures were burned at the end of their life. Bennett (1936, 424) observed of Structure 1, “The floor…was covered by a thick layer of ash and charcoal, burnt clay, bone and stone tools, and many fragments of plain and painted bowls…The rest of the structural fill was burnt adobe. The charcoal gave the impression of a thatch roof which had been burnt and collapsed.” A similar assemblage was recovered in the structure Bennett (1936) designated “House 2” although he notes that the wall plaster appeared “fired and slightly polished” (p.426). Bennett notes that burned deposits were traced across the ca. 30 m trench his team cut across the mound. Unfortunately, both structures excavated by Bennett have been subsequently destroyed, and his trench was not accessible for later excavation (Bandy, 1999a).

In 2006, Structure 4 was partially excavated to the floor in the central room and five bins were identified, three of which were sampled (Fig. 3, Hastorf et al., 2008). Excavations were initiated in 2 × 2 m units and each discrete layer or feature was given a locus number. These loci were resolved to events using a Harris matrix system (Fig. 4, Harris 1979). For each excavated locus, all uncollected sediment was screened through a 0.635-cm mesh, and systematic sediment samples for water flotation, soil, archive, and phytoliths were collected. Excavators also collected grab samples of fired walling fragments. We made opportunistic use of these collections seeking combustion evidence from fills, ash deposits, and burned materials from the central room, and two bins (R202, R207/8) as summarized in Table 3. The cultural fill event (D511) over Structure 4 was included because it contained burned fragments of collapsed walling from the fire-associated wall fall event (D514). Ash was distributed discontinuously across the central room and was divided into three events (D528, D516, D521). Below these strata in the central room, the excavators recovered a series of fills and burned areas (D515, D535, D517, D523, D524) before reaching the floor (D525, D538, D534, D539). The southwest bin had two fill layers (D544, D546) and a floor (D547), while the northwest bin (R207/8) contained fill (D533) and a floor (D 534). Figure 5 illustrates the post burning fill layers with three discrete orange flooring events.

Fig. 3

Plan of structure 4 (after Hastorf et al., 2008, Fig. 3.1)

Fig. 4

Chiripa Structure 4 Harris matrix of excavated events (adapted from Hastorf et al., 2008, Fig. 3.120, revised in 2012). Events are identified by D in this structure (see Table 3)

Table 3 Summary of fire reconstruction analyses from the 2006 Chiripa excavations. Analysis key: Bots, macrobotanical analyses; Bone, faunal analysis; SMM, soil thin section analysis; WDXRF, wavelength-dispersive X-ray fluorescence EXP, experimental (Goodman Elgar et al., 2015)

Fig. 5

2006 excavated profile wall showing three lighter burning layers (photograph by C. Hastorf)

Fire Reconstruction

From the excavation evidence and preliminary studies of mudbricks (Davenport, 2010), we developed a hypothesis for the burning conditions that destroyed Structure 4. The ash events D516, D521, and D528 indicate that the organic fuel was burnt to ash, which occurs above 500 °C (Canti, 2003). Hard, red mudbrick fragments indicate that the fire burned under oxidizing conditions for long enough to penetrate more than 10 cm into the adobe bricks. A temperature of over 500 °C is unlikely to be sustained by roof fall alone as this is a relatively modest fuel load. We therefore hypothesized that the Structure 4 fire was a planned fire with a supplemental fuel load placed in the central room where the rubified walling concentrated (Fig. 5). To test this hypothesis, we drew on three archaeological datasets (bone, sediment, and botanical assemblages) in addition to experiments with burning contemporary and archaeological mudbrick. These data allowed us to reconstruct the burning conditions including oxidation conditions, temperature range, and potential fuels. We also drew on prior wavelength-dispersive X-ray fluorescence (Goodman Elgar et al., 2015) and thin section analyses (Davenport, 2010).

We conducted experiments on heat alteration of mudbrick by an open-air fire on the Taraco Peninsula, following Berna and colleagues (Berna et al., 2007), and in the laboratory through controlled firing experiments in a muffle furnace at Washington State University. A study of thermal alteration of archaeological bone from Structure 4 provided a means to reconstruct temperature that is independent of the sediments. Thin sections of archaeological and contemporary mudbricks were analyzed for evidence of heat alteration. In addition, fuels were assessed through paleoethnobotanical analyses of plant remains encountered in the flotation samples from the interior of Structure 4 and one of the bins (Langlie, 2009). Goodman Elgar developed an interpretive framework for differentiating roofing from Altiplano fuels, and applied this framework to deposits within the fills, floors, and bins of Structure 4 based on a range of published research (see Table 11). We evaluated this evidence within the context of the burning scenarios considered above.

Thermal alteration of organics begins by 200 °C, producing char and soot, and full combustion occurs by 500 °C producing silicate ash (Ball, 1964; Canti, 2003; Heiri et al., 2001; Nelson & Sommers, 1996). Exposure to heat alters sediment composition, color, and structure. Composition changes result from the volatilization of soil components as heat increases: water is released by 105 °C, organic carbon is converted to carbon dioxide at 500 °C, and mineral carbon (calcium carbonate) combusts at 900 °C (Ball, 1964; Heiri et al., 2001; Nelson & Sommers, 1996). Ash will melt to produce ash opal which is much more durable than ash, but this is also burned off over 900 °C. Under oxidizing conditions, sediments rubify from unheated colors to browns, burgundies, and bright reds with increased heat exposure (Canti & Linford, 2000). Sediments exposed to moderate heat for short periods are generally blackened with burned organics, while sediments under high heat or oxidation conditions become higher chroma—brighter and redder (Berna et al., 2007, p.365; Canti & Linford, 2000; Mirti, 1998). As the heat and fire exposure time increase, sediments continue to redden as fuel is converted to ash (Ketterings & Bigham, 2000, 1826). However, under reducing conditions, metal oxides turn gray and then darken toward dark brown and black. Heated color generally reflects mineral composition, but the use of color for fire reconstruction can be complicated by mineralogy, original firing time, and conditions, as well as analyst’s color perception (Canti & Linford, 2000; Mirti, 1998; Shepard, 1956, 102–106). Firing color is relatively stable and can be used as a proxy for firing temperature but must be corroborated by other evidence, as we have done in this study.

Geoarchaeological Analyses

Aims and Methods

Geoarchaeological analyses were conducted on mudbrick samples collected from wall fall within the central room and one bin of Structure 4 as well as from a column recovered from the wall (sample 06CHP9B). As mentioned, wall fall samples were opportunistic samples collected by the excavators and therefore may not represent all aspects of the fire that destroyed the structure. We also assessed contemporary mudbricks from three sources (EXPAC, EXQSP, EXSEDE) as reference material, as described in Goodman Elgar et al. (2015). This study draws on prior thin section, sediment, and wavelength-dispersive X-ray fluorescence (WD-XRF) analyses. Seven mudbricks were assessed for composition and manufacturing methods by Davenport (2010). Thirteen archaeological samples from Structure 4 and triplicate subsamples of the three contemporary mudbricks were characterized by WD-XRF for 28 elements (Goodman Elgar et al., 2015).

The composition and manufacturing of the seven mudbricks analyzed by Davenport (2010) have two textures (silty fine sand and silty sand with gravel), and two types of plant temper (chopped fine pieces ca. 1 cm in length, and coarse plant temper over 3 cm in length). The coarse plant temper is quite similar to Middle Chiripa ceramics from Chiripa which have dense plant temper that is probably Stipa ichu (Kunth) grass (Roddick, 2009, p.219–221; see also Steadman, 2007). WD-XRF identified two geochemical compositional groups (Goodman Elgar et al., 2015), but they did not correspond to the manufacturing groups. This demonstrates that the mudbricks in Structure 4 were prepared from multiple localities by several makers, following more than one recipe. Although the local soils are light beige, all samples assessed by WD-XRF have enough iron and manganese to produce pronounced chroma shifts from fire and heat. Mudbricks with coarse plant temper are likely to show much more alteration of the internal structure of the mudbrick than sand tempered bricks from the conversion of organic matter to CO₂ gas.

For the present study, we conducted open fire and muffle furnace experiments on all three contemporary mudbricks and one archaeological mudbrick sample. We compared these to the colors and temper of 12 archaeological samples from the central room, and two archaeological samples from the bin. Finally, thin sections of 16 mudbricks and duplicate samples of the contemporary mudbricks were analyzed in thin section for heat alteration considered below.

Experimental Open-Air Fire

Aims and Methods

We built an experimental fire pit using contemporary and archaeological mudbrick, and conducted an open-air fire at the Santa Rosa Community Center at ca. 3840 masl. Air pressure and oxygen are lower at high elevation, which has the effect of reducing boiling temperature but increasing time to combustion. For instance, at 4267 masl, water boils at 87 °C rather than the 100 °C boiling point of water at sea level (USDA, 2011). We anticipated that mudbricks adjacent to a fire would be notably altered by heat and flame and that mudbricks placed in the center of the fire would be comparable to the rubified Structure 4 walling.

Materials included modern mudbricks from three different sources (EXPAC, EXQSP, EXSEDE) and one unfired archaeological mudbrick from Event 533, Locus 8676 (Table 3; see Goodman Elgar et al., 2015). EXPAC samples were from 250 m south of the Kala Uyuni site and were silty clay loam with inclusions of small pebbles and grass or chaff temper. EXSEDE samples were from the Santa Rosa Community Center, Bolivia, and made from silty clay loam with small pebbles, grass temper, and minor inclusions of dung. EXQSP sample were from about 500 m down slope from the community buildings, and made from silty loam with small pebbles and grass temper.

The open fire experiment was conducted on July 24, 2009 (Fig. 1). A 1 × 2 m sack of sun-dried cow’s dung was procured locally and used as fuel for the experiment. Pre-Columbian people would have used camelid dung rather than European cow dung. Camelids are no longer herded locally and camelid dung was not available. Taraco Peninsula residents informed us that camelid dung is preferred because it burns hotter and cleaner than cow dung. According to the Servicio Nacional de Meterología e Hidrología de Bolivia, on July 24, 2009, the nearest weather station at El Alto recorded temperatures between 12.4 and − 4.4 °C. The lowest temperature had not been reached by midnight when the experiment was concluded. Nevertheless, frigid air temperatures cooled the fire rapidly after we stopped adding fuel at the experiment’s end.

Modern mudbricks were placed in a ring approximately 0.75 m across and the Chiripa mudbrick was placed among them (Fig. 6). The enclosure was filled with cow dung and lit with cane alcohol at 6:10 pm (Fig. 7). Temperatures were measured with a grill thermometer to a maximum temperature of approximately 650 °F/343 °C (Table 4) which precluded determining the actual maximum temperature achievable from the open fire experiment. Readings were taken in 50°F increments and converted to Celsius. The thermometer was initially placed on a mudbrick along the edge of the fire (Fig. 8). Temperatures were recorded in increments until the maximum temperature of 650°F/343 °C was reached. Temperature was then recorded at intervals for 3.5 h during which time fuel was added to maintain the fire (Table 4 and 5). A formal assessment of the ashes was not conducted since cow’s dung is not a pre-Columbian fuel.

Fig. 6

a Experimental fire ring with archaeological sample noted by white arrow (left); b experimental fire showing dung arrangement when fire was started by Franz Choque (center); c experimental fire in full blaze observed by Di Hu and Nichole Davenport (right)

Fig. 7

Thin sections of contemporary mudbricks heated in open air fire (5 cm × 7 cm). A EXQSP1 subsample of heat exposed surface; B EXPAC2 subsample of heat exposed surface; C EXQSP5 subsample of exterior surface; D EXQSP6 subsample of exterior surface; E EXSEDE1A subsample of heat exposed surface; F EXSEDE1B subsample of exterior surface

Table 4 Open air fire thermal color changes on mudbricks

Fig. 8

Contemporary burned mudbrick micromorphological features. EXPAC2: A charred plant temper (xpl); B rubified groundmass (xpl); C cross section of grass temper with bright white birefringence characteristic of unaltered plants (xpl); EXPAC6: D clay nodule; E plant temper with morphology characteristic of plant pseudomorphs; F sesquioxide impregnation of silty groundmass

Table 5 Open air fire thermal color changes

Results

The experimental fire reached the maximum temperature we could measure within an hour. The cow dung gave off little smoke or smell once it was burning (Fig. 6). Additional dung was required to maintain the fire at 36 min, 59 min, and 170 min when the fuel was expended. Despite abundant fuel, the fire was difficult to maintain as ash built up. The fire went out and had to be restarted after 80 min, and it did not regain the highest measurable temperature. After we consumed all the dung, the fire was allowed to burn out. It cooled until we ended observations after 5.75 h when it had reached 66 °C. The fire was then quenched with water.

Combustion was very efficient and very little unburnt dung residue was found in the light gray ashes when reviewed the following morning. Contemporary mudbricks changed color on the surface adjacent to the fire after 5 min of heat exposure and started to brown within 15 min. All three brick types displayed graded color changes from surface to interior. The maximum penetration of high chroma reddening was in EXPAC which rubified to 2 cm and a further 2 cm was darkened. EXQSP only darkened to 2 cm, and EXSEDE had a dark surface 2 mm rubified interior and was darkened to 2.5 cm. The archaeological mudbrick did not change color after 27 min and was placed in the middle of the fire with the thermometer placed on top. However, even under these conditions, the archaeological mudbrick did not rubify, and only cracked after 52 min at approximately 260 °C.

The groundmass of mudbricks observed in thin section ranges from light brown to red (Table 7). The light brown bricks (8676/1, 8573) also had smooth, subangular voids, preserved organics and highly birefringent minerals. The brick fragment from D515 was dark colored with graded color variation typical of a heating front, highly birefringent minerals, and preserved plants and charcoal. The remaining samples (Loci 8615, 8645, 8673, and 8676/2) demonstrated high chroma colors, rounded voids typical of gas expansion, minerals with low birefringence, and no observed organics.

Discussion

The dung fire required consistent management to stay lit, and even with several attendants, it had to be restarted. This observation supports our contention that a hot fire requires management at high elevation. After the final heating and cooling, the contemporary adobes were only thermally altered along the margin adjacent to the fire, and surprisingly the archaeological sample did not rubify at all. Since the geochemical properties of the archaeological and contemporary mudbricks are quite similar (Goodman Elgar et al., 2015), similar patterns of heat alteration would be expected. It is possible that the archaeological sample contained internal moisture (March et al., 2014). The thermal alteration patterns identified on the mudbricks surrounding the dung fire parallel observations of adobe kilns by Rye (1981:100). However, counter to our expectations, none of the contemporary mudbricks lining the fire pit demonstrated thermal alteration into the bricks comparable to the rubified Structure 4 walling in terms of high chroma red or depth of reddening (e.g., sample 06CHP9B). This indicates that despite abundant fuel, ample oxygen, and several people attending the fire, the open fire failed to achieve sufficient thermal intensity to penetrate and alter the bricks (Keeley, 2009). We therefore conclude that a closed combustion environment within Structure 4 was required to achieve the temperatures that would burn through the mudbricks and completely combust the plant remains within some mudbricks as observed by Davenport (2010). This experiment substantiates our assertion that an expedient fire from roof burn could not produce the alteration of earthen walling observed in Structure 4.

Experimental Controlled Firing

Aim and Methods

Refiring is a well-established technique in ceramic analyses for determining firing temperature by generating control samples fired at known temperatures and assessing these for temperature dependent characteristics such as color, porosity, and surface features for comparison to archaeological samples (Rye, 1981; Tite & Maniatis, 1975). When sediments are heated, their color changes depending on temperature, heating time, organic content, iron oxide composition, and available oxygen during burning (Berna et al., 2007, p. 365; Canti & Linford, 2000; Ketterings & Bigham, 2000; Rye, 1981). For the Chiripa fire reconstruction, a control series was made by heating small sediment blocks in a furnace to build on the field experiment using the same contemporary mudbricks and archaeological mudbrick sample used in the open pit fire. Subsamples were fired in 100 °C increments for 4 h from 400 to 1000 °C in a Thermolyne™ benchtop muffle furnace in the Geoarchaeological Laboratory at the Washington State University, Pullman, for a total of six firings. A sample was removed after each firing and cooled, and the dry Munsell color was recorded (Rye, 1981, p. 119; Table 6).

Table 6 Dry Munsell colors of experimentally fired archaeological and contemporary mudbrick

The dry Munsell colors of 11 archaeological mudbrick samples from six stratigraphic layers in Structure 4 and the southeast bin were determined (Tables 6 and 7). Samples from loci 8547 and 8615 each had more than one mudbrick fragment, and these mudbricks were recorded separately. In addition, samples 8625, 8547-A, and 06CHP9B displayed different Munsell colors on the exterior of the mudbrick to the colors apparent on the interior of the mudbrick which were also recorded separately.

Table 7 Dry Munsell colors of recovered archaeological mudbrick samples

Results

Unfired experimental mudbricks ranged in color from very pale brown (10YR 7/4) for EXSEDE and the archaeological sample to light brown (7.5YR 6/3) for EXQSP and EXPAC. All mudbricks responded to experimental firings with a progressive color shift towards red, and there were subtle color variations between samples (Table 6). We tracked broad color changes by fired hue (2.5 YR, 5YR, 7.5YR, 10YR). The contemporary bricks shifted from 7.5YR to 5YR at 700 °C, and from 5YR to 2.5RY (red) at 1000 °C. The archaeological sample shifted from 10 to 5 YR at 400 °C and remained in the 5YR color range until 1000 °C when the color shifted to redder 2.5 YR colors.

The colors of the archaeological samples fell within the range of the experimentally fired bricks except for samples 8625-exterior, 8625-interior, 8627, and 8615/2 which were browner or grayer (Table 7). Two archaeological bricks appeared unfired, one from D517 and one from the southeast bin (D544). Two samples fell in between the unfired bricks and those of the 400 °C colors and were assigned to a category of > 400 °C (8625-interior (D511) and 8615/2 (D544). One sample from the southwest bin (D5444) appeared fired to 400 °C, and two samples appeared fired between 500 and 700 °C (one from D516, one from D524). All five samples from D516 and one from D524 were 5YR, interpreted as heated between 700 and 900 °C. Finally, two samples displayed darker colors not observed in the firing experiments. The exterior of a brick from D511 (sample 8625-interior) was a dark grayish brown and a brick from D517 (sample 8627) was grayish brown. These darker colors are more typical of reducing conditions.

Discussion

In the furnace firing experiment, the archaeological and contemporary mudbrick samples follow a general pattern of light brown when unfired, to deeper browns at 400 °C, to reddish-yellow between 500 and 900 °C, to red at 1000 °C. The airflow in muffle furnaces promotes oxidizing conditions, and this is observed in the progressive shift to red. The similarity in color changes between the contemporary bricks and archaeological samples reflects their origins in Taraco topsoil.

The archaeological sample colors generally fall into those observed for the experimental firings. However, the pattern is heterogeneous even between samples from the same 1 × 1 m unit. The southeast bin samples were unfired or fired under 400 °C. The other brick that matched unfired colors came from the same event as mudbricks that were fired much hotter, up to 900 °C. Events D511 and D517 indicate lower temperatures probably under reducing conditions from less oxygen and may have been buried in rubble. The sample from D524 indicated different firing conditions on the surface (400 °C) to the sample interior (600–700 °C). Since the fire was on the inside of the building, this pattern does not conform to expectations. In contrast, archaeological mudbrick samples from ashy event D516 corresponded to much higher temperatures, ranging from 500 to 900 °C. The high temperature must have been sustained for a considerable time to penetrate all the way through these mudbrick samples.

Based on color, we observe hot spots in the central room along the southern wall and along the western wall. In event D516, temperatures above 500 °C correspond well with the field descriptions, which interpreted the deposit as ash, because organic carbon would have been consumed at these temperatures. In contrast, the southern bin, and some bricks from D511, D514, and D524, only experienced temperatures below 400 °C. Finally patches of the room interior experienced reducing conditions likely from being buried in wall fall. It should be emphasized that the opportunistic samples assessed here were collected as mudbrick specimens, rather than evidence of fire conditions. Therefore, these observations reflect the available evidence of maximum temperatures from the samples collected and may not represent all the burning environments in Structure 4. It is therefore necessary to corroborate these patterns with other lines of data.

Thin Section Analyses

Twenty-two thin sections of mud brick were analyzed for evidence of heat alteration for this study. Samples represent six stratigraphic events from the central room (n = 12), one event (n = 4) from the northwest bin, and two samples each from the contemporary bricks (n = 6) (Table 3), which includes previously analyzed bricks (Davenport, 2010). Samples were analyzed for composition, structure, and evidence for temper under plane polarized (ppl) and crossed polarized light (xpl) at 50–600 × magnification in the Geoarchaeology Labs and the Franceschi Microscopy and Imaging Center, both at Washington State University. Thin section observations follow Stoops (2003) with reference to archaeological studies (Berna et al., 2007; Canti, 2003; Canti & Linford, 2000; Mallol et al., 2007; Matthews, 2010; Mentzer, 2014). Thin section observations concentrated on color/opacity of the groundmass and metal oxides, evidence for a heating front, void shape, birefringence of clays, and the condition/presence of plant temper.

Mudbricks are very dense, especially if they have sintered, and are therefore difficult to impregnate with the resin used to preserve sediments for thin sectioning. Spectrum Petrographics (Vancouver, WA) and Applied Petrographic Services, Inc. (Hempfield Township, PA) were able to prepare usable slides. However, small domains of incomplete impregnation and air bubbles are common processing artifacts in this slide collection and may appear in photomicrographs. However, processing artifacts did not significantly limit analyses as they are easily identified.

A heating front or thermal alteration zone of color change was observed in the experimental open fire and is typical for earthen material in contact with heat (Rye, 1981, p. 100). Andean mudbricks are commonly grass-tempered, which produces subangular voids referred to as plant pseudomorphs because the shape derives from plant material that has decayed (Matthews, 2010). Oblong voids with rounded edges may occur when such features are heated and the trapped gas expands rounding the void. These must be differentiated from voids which normally have rounded termini, such as root channels. Clay in Taraco soils is normally pale yellow to orange but heat commonly shifts clay color redder and increases birefringence. The structure of minerals also changes with heat (Grapes, 2011), which alters birefringence. Moderate heat may cause organized crystals to fracture. At melting temperatures, crystals in minerals lose their structure and become glass, which is anisotropic (dark) under cross-polarized light. The birefringence of mineral inclusions and coarser sands were assessed for evidence of anisotropy. Fresh plant remains are also highly birefringent as they dull with heating until charcoal is produced around 400 °C (Matthews, 2010). If heating exceeds 500 °C, macrocharcoal should not be present. These micromorphological features therefore provide clues into heating conditions.

Contemporary Mudbricks

The heating front is clearly visible in heat exposed subsamples of our experimental mudbricks with coarse grass temper (Fig. 7A,B). The exterior edge samples do not display color alteration. Dark colored nodules of metal oxides (sesquioxides) are clearly visible in all thin sections. However, the nodules are darker and more opaque in EXQSP and EXPAC samples (Fig. 7C,D), sometimes referred to as a ferruginous matrix, which can reflect a shift in iron oxidation resulting in hematite (Röpke & Dietl, 2017). The sesquioxides in EXSEDE thin sections display a broader range of colors and are less opaque than those in the other mudbricks and did not show a heating front (Fig. 7E,F).

When a heating front was observed in the block specimens, the grass temper under magnification was notably charred (Fig. 8A). Bricks without a heating front in the blocks (hand specimen) contain well-preserved plant remains with tissue morphology under magnification (Fig. 8C,E). The heating front is also clearly visible in the bright reddening of the groundmass in EXPAC2 (Fig. 8B). In contrast, the groundmass observed in the exterior sample of the same brick has a low golden birefringence (Fig. 8D). Clay nodules show a color shift towards red but the birefringence was relatively low (Fig. 8D), and sesquioxide concentrations were not notably opaque (Fig. 8F). These features were also identified in EXQSP while EXSEDE did not contain clear indications of heat alteration. The heating effects observed in the contemporary mudbricks indicate that the experimental fire did not heat the bricks over 500 °C. Char forms between 300 and 400 °C and would be converted to ash over 500 °C. It is also clear that that the heat did not penetrate deeply into the mudbricks.

Archaeological Mudbricks

Based on the refiring experiments, we suggested that the archaeological mudbrick samples experienced heterogeneous firing temperatures even within the same deposit since the deposits included wall fall. Thin section observations support this assertion as seen both in the mudbrick blocks (hand specimens), and in their micromorphological features. The thin sections in Fig. 9 are arranged from top to bottom from less heat-altered to more heat-altered based color and void shape. A mudbrick from the upper burnt construction fill within the central room (D523) was light colored with well-preserved coarse plant temper observed as large subvertical voids (Fig. 9A). A second brick from this context did not have coarse plant temper, and the colors indicate heat alteration. However, the structure and voids were not notably altered. A similar pattern was observed in samples from the northwest bin (D533), with one fine-textured brick not notably heat-altered (Fig. 9B). Another sample demonstrated a heating front and deep colors but nevertheless retained structural integrity (Fig. 9D).

Fig. 9

Archaeological mudbricks in thin section (A 2.5 × 4.5 cm, all others 5 × 7 cm). A Slide 8020A (D523); B slide 8676A (D533); C slide 8020B (D523); D slide 8676B (D533); E slide 8625A (D511); F slide 8627 (D517); G slide 8625B (D511); H slide 8545 (D514)

Several thin section samples preserved the brick surface that was exposed to the fire. Figure 9E is from the fill over Structure 4 (D511). The surface of this brick was fine textured without plant temper similar to the silty barro used to surface contemporary buildings (Goodman Elgar, 2008). There were few voids in the bright red surface layer which appears to have limited the flow of oxygen into the brick resulting in reducing conditions. Stress fractures in the form of large vertical cracks were clearly visible. Similar characteristics were observed in a brick from cultural fill deeper within deposit D517 with a fine textured surface, deep color patterns, and visible cracks (Fig. 9F). However, the surface was orange-yellow rather than red. Slide 8625B (Fig. 9G) was also collected from D511, and was deeply rubified with large, rounded voids indicating temperatures above 500 °C. Similarly, the wall fall D514 sample was also deeply rubified and had large oblong voids with rounded termini (Fig. 9H).

Plant remains in bricks without color alteration were withered within rectilinear channels typical of plant pseudomorphs (Fig. 10A). The groundmass was golden brown and clay nodules were limpid yellow to golden brown (Fig. 10B,C), which is typical of topsoil from the Taraco Peninsula (Goodman Elgar, 2008). Plant remains within darkened mudbricks were charred or absent, and the channels were rounded rather than rectilinear (Fig. 10D). The groundmass was notably redder and brighter than the unfired bricks (Fig. 10E,F,G). Clay nodules were redder (Fig. 10F) and sesquioxide nodules dark and opaque (ferruginous; Fig. 10H). In the rubified bricks, the birefringence was extremely bright and red (Fig. 10J,L) while sesquioxides were notably black and opaque (Fig. 10K).

Fig. 10

Photomicrographs of archaeological mudbrick micromorphological features. Slide 8020A (D523) A coarse plant temper in unaltered sediment (ppl); B golden brown groundmass and small clay nodule (xpl); slide 8676A (D533) C deformed clay nodule with limpid yellow to strong brown birefringence (xpl); slide 8020B (D523) D browned plant temper in rounded channels within red-brown groundmass (ppl); E reddish brown groundmass (xpl); 8676B (D533) F dull red deformed clay nodule in reddish brown groundmass (xpl); slide 8625A (D511) G dark reddish-brown groundmass (xpl); H browned grass temper in subrounded channels with dark, opaque sesquioxide nodule above left (ppl); slide 8627 (D517) I anisotropic (browned) grass temper in rounded channel within dense, red brown groundmass (xpl); slide 8625B (D511) J bright, deep red groundmass (xpl); K opaque sesquioxide impregnation (xpl); slide 8454 (D514) L deformed clay infilling with intense red birefringence (xpl)

Discussion

Features of the contemporary mudbricks clearly show heat alteration, but these are limited to areas that have changed color. The main features observed, apart from color change, were a red, birefringent groundmass, and charring of grass temper. These features are consistent with temperatures between 300 and 400 °C when plants char. Morphological changes in voids were not observed in the contemporary mudbricks.

The archaeological mudbricks documented a range of firing conditions. The material characteristics of unfired bricks are similar to local topsoil and had well-preserved plant pseudomorphs from grass temper. In contrast, bricks with notable color alteration also had internal features demonstrating heat alteration such as bright, reddened groundmass and darker, opaque sesquioxides. The vertical cracking observed in several bricks indicates structural changes from heating. The bricks with the most profound color alteration were bright red and had rounded voids with no plant remains. This indicated that any plant temper and organics had burned off and the resulting CO₂ gas caused morphological changes in the hot bricks. The jumbled distribution of bricks with different heating conditions appears to reflect wall collapse, and the subsequent leveling of Structure 4.

The archaeological thin sections therefore provide further evidence that the burning conditions within Structure 4 were not similar to an open-air fire. Parts of the structure were exposed to temperatures over 500 °C sustained long enough for organics to completely combust and to alter the physical structure of adjacent bricks. Complete combustion takes approximately 4 h (Ball, 1964; Heiri et al., 2001), and could not have been sustained by incipient burning of the roof and beams. It is also clear that to sustain such temperatures, the fire must have been enclosed and the oxidation conditions controlled. The technical skills needed to produce these firing conditions are the same as those needed for pottery production. Since Formative peoples produced high-quality polychrome pottery, we know the technologies were within their craft repertoire. The mudbrick evidence indicates that knowledgeable agents managed the Structure 4 fire, and these people may indeed have been potters.

Faunal Analyses

Aims and Methods

Thermal alteration of bone may be observed as surface alteration or it may penetrate into the bone. Sooty bones suggest temperatures between 300 and 500 °C when soot is converted to carbon dioxide. Calcined bone is observed as chalk white, blue, or gray and is formed from ca. 600 °C when organic carbon has been burned off and carbonate apatite remains. By 900 °C, carbonate burns off, and only traces of calcium oxide (quicklime) remain (Figueiredo et al., 2010; Nicholson, 1993; Shipman et al., 1984). Bone density is also altered by heat as calcined bones are highly porous, but bone density increases again by 1200 °C (Figueiredo et al., 2010). Bones burned at moderate temperatures become brittle and break easily when pressed or compacted (Stiner et al., 1995), which further aids in their identification. Since cooking temperatures are below the temperatures required for calcination, calcined bones are usually associated with secondary burning of a structure, site maintenance, or deliberate burning in a hearth.

The faunal analyses presented here are adapted from the analyses of Katherine Moore published in Hastorf et al. (2008). Burning conditions were assessed on washed bone from excavated and analyzed contexts in the field by recording evidence of heat alteration, and the proportion of altered bone was quantified by number of pieces and weight/fragment. Four events from the central room of Structure 4 (D511, D514, D515, D517) and two events from the northwest bin (R207/8, Fig. 3) were analyzed for evidence of burning and calcination (Table 8).

Table 8 Burning and fragmentation of animal bone by event (adapted from K. Moore in Hastorf et al., 2008, Table 4.1) (Calc. calcined, unb. unburnt)

Results

Heat alteration was observed in all analyzed contexts except for the lower above floor fill (D533) of the northwest bin (R207/8). Burnt bone was most significant in D514, the wall fall layer, and D515, the cultural fill layer immediately below the ash layers but well above the floor, perhaps part of the roof collapse. Calcined bone was best represented in D511, the cultural fill over the collapsed structure, and in D517, the burnt cultural fill layer just above the floor within Structure 4.

Discussion

In order to compare the intensity of heating between events, the ratio of unburnt to calcined bone weight per fragment was calculated (Table 8). The bone in the cultural fill in the center of Structure (D517) was subjected to heat in excess of 600 °C for long enough to calcine a substantial portion of the bone. Moving from the center towards the south wall, the temperature decreased to 300–500 °C when bone is altered but most of the organic carbon remains. The farthest westernmost excavated portion of the room also had a concentration of calcined bone again suggesting temperatures over 600 °C.

The presence of bone in the fill above the floors indicates that it was deposited well above the floor, presumably over the fuel or even after the building was destroyed, during the closure fill event before the preparation of the new post-structure surface. This may be considered structured deposition (Garrow, 2012; Roddick & Hastorf, 2010), which in this context would support an interpretation of repeated ritual burning. Large camelid bone was also recovered from the terminal fills of the nearby Choquehuanca sunken civic-ceremonial court at Chiripa (Hastorf, 1999a) suggesting the same end deposition activity. Animal offerings may have had ritual significance for Formative people.

Altiplano Fuel Sources

Based on the temperature the mudbricks of Structure 4 burned, we anticipated finding evidence of supplemental fuel in botanical samples from the structure. We anticipated that fuel from an incipient fire would only include flammable structural materials, predominantly roofing, but a planned fire would include supplemental material added as fuel. Fuel choices may depend on availability and procurement costs, manageability, combustion properties, and combustion byproducts. Here we focus on charred plant remains and dung as evidence of fuel. Common Altiplano fuels include wood, sedges, totora reeds, grasses, shrubs, and camelid dung, while cactus and herbaceous plants are less common (Arnold, 1993; Goldstein, 2011; Johannessen & Hastorf, 1990; Pearsall, 1988; Sillar, 2000, Winterhalder et al., 1974; Wright et al., 2003). Table 9 summarizes paleoethnobotanical evidence for Altiplano fuels and construction materials from prior studies. Several Altiplano plants would have arrived in archaeological contexts via multiple and sometimes indistinguishable pathways including as food, fodder, fiber, and fuel as well as accidental transport by wind or on clothing.

Table 9 Altiplano construction and fuel-related plants ubiquity expressed as percentage presence, N number of samples analyzed

Wood

Currently, trees are very sparse in the Altiplano and palynological studies from lake sediments indicate very few trees in the region before the Formative period (Gosling & Williams, 2013; Paduano et al., 2003). The primary trees in the Altiplano are Polylepis spp. (keñua, kewiña) and Buddleja spp. (kiswara). These trees are rarely used as fuels due to their slow growing nature and sparse distribution (Ansión, 1986). Whitehead (2006, 2007) noted a sharp decline in wood charcoal remains at Chiripa sites between Early Formative and Middle Formative periods as the regional population increased (Bandy, 2001, 2004a, b). This is consistent with research on the most studied Altiplano tree, Polylepis spp., which is notably sensitive to human disturbance (Domic et al., 2014; Hensen, 2002).

Wood is also used for durable products and architectural elements including cooking wares, doors, tools, beams, and purlins (see Ansión, 1986; Capriles & Flores Bedregal, 2002; Table 10; Nair et al., 2018). The relative scarcity and high utility of tree wood in the Altiplano makes it unlikely to have been a common fuel during the Middle Formative phase. As in other areas, large roof beams would have been particularly valuable materials that are likely to have been salvaged from structures that were going to be abandoned and reused. This suggestion is supported by radiocarbon dates from wood beams at the Formative site of Taraco, Peru, in the northern Titicaca Basin (Stanish & Levine, 2011), which indicate that beams were considerably older than the associated grass and sedge roofing. Nevertheless, wood has advantages over other fuels in its potential for high burning temperatures and long burning times (Winterhalder et al., 1974). Therefore, the use of wood as a fuel may have been considered an ostentatious display in the Altiplano, and could suit a ceremonial burning scenario.

Table 10 Standardized densities of fuel and architecture related plant species by room sector in Structure 4 (N = 18)

Woody shrubs are more common than trees in the Altiplano and are also burned for fuel such as t’ola (Baccharis microphylla Kunth) and Tetraglochin cristatum (Britton) Rothm (Bruno, 2008; Wright et al., 2003).

With this in mind, we differentiate between wood beams and wood fuel including trees and bushes in this study. Heat alteration was most visible in the center of the chamber (Hastorf et al., 2008), and we anticipate this is where the fire started. If a wood fire were kindled in the structure, it would have burned the roof beams. Therefore, we should find a higher concentration of wood at the building’s edges from the beams as ethnoarchaeological structural fire studies illustrate wood charcoal concentrates along external walls where the beams terminate (Bankoff & Winter, 1979; Friede & Steel, 1980; Icove et al., 2006). Wood concentrations in the structure’s center would suggest added fuel. We therefore predict that wood derived exclusively from roofing should concentrate along the external walls in Structure 4 (Fig. 3).

Sedges and Grasses

In the Altiplano, sedges and grasses are traditionally used for roofing, matting, boats, camelid forage/fodder, and, in the case of totora (Schoenoplectus californicus) rhizomes, also food (Table 11). However, both matting and foods can be expected to produce regular, low concentrations in macrobotanical assemblages because matting does not contain many seeds (Browman, 1989), and totora is unlikely to be preserved through charring, since the rhizomes can be eaten raw. Contemporary potters collect the straw of grasses, including ichu (Festuca spp.), for fuel in the Andes (Arnold, 1993, 55; Sillar, 2000). Thus, many pathways are possible for the introduction of sedges and grasses into Formative contexts.

Given that straw fuel is used to fire ceramics, stalks might be able to produce initial fires with enough intensity and duration to alter earthen walling under specific conditions but would most likely turn to ash in an enclosed hot fire. However, as previously noted, falling thatch fires burn out quickly and leave little trace on earthen construction materials. The basal temperatures of experimental grass fires were too low to alter sediments, ca. 225–400 °C (Korfmacher et al., 2003). While grass fire temperatures may spike, heat is not sustained long enough to alter sediment (Bento-Gonçalves et al., 2012). We therefore expect that a large supplement of grass, sedge, and/or straw fuel would have been required to produce the fire patterns observed in Structure 4, and if so, this might be visible in high stem and stalk concentrations.

Chenopods

Along these same lines, chenopods also have overlapping uses. While quinoa and kañawa are most often interpreted as food in the archaeological record, wild and domestic Chenopodium spp. are also consumed by foraging camelids (Hastorf & Wright, 1998). Additionally, Bruno (2008:212) observed contemporary Taraco residents burning quinoa stalks, which could introduce chenopod seeds into the burned record. Depending on the context, chenopods could indeed be interpreted as human food, dung burning, or stalks used directly as fuel.

Dung

In xerophytic, high elevation environments, dung is the dominant fuel, and introduces small carbonized herbaceous seeds into the archaeological record (Bruno & Hastorf, 2016; Hastorf & Wright, 1998; Miller, 1984; Miller & Gleason, 1994; Miller & Smart, 1984). In the Altiplano, dung was a preferred fuel source for domestic cooking fires (Hastorf & Wright, 1998; Johannessen & Hastorf, 1990), and thus, archaeobotanical remains often preserve the partially masticated and digested remnants of grazing (Miller & Smart, 1984). We anticipate that camelid dung was an available resource used for fuel in Middle Formative times since camelids were widely utilized at Chiripa (Hastorf et al., 2008; Moore et al., 1999). Dung fuel is well-documented by paleoethnobotanists from the Formative period to the modern era in the Altiplano (Browman, 1989; Bruno, 2008; Bruno & Hastorf, 2016; Hastorf & Wright, 1998:213; Langlie, 2020; Whitehead, 2006, 2007). It is thus plausible that Chiripa residents used dung to fuel the burning of Structure 4.

Dry dung has beneficial burning properties and can sustain temperatures of 900 °C with relatively even and smoke-free burning (Shepard, 1956, Fig. 4 p. 103; Rice, 1987 157–8; Sillar, 2000). Sillar (2000) found that contemporary potters in the Cuzco region preferred dung for ceramic fires because of even burning temperatures. Contemporary Altiplano potters used cow dung for ceramic fires in the 1940s (Tschopik, 1950). Roddick (2009, pp. 281) found that many Middle Formative ceramics were fired between 550 and 650 °C under oxidizing or partly oxidizing conditions, which he found consistent with dung fuel(see also Sillar, 2000). Dung collection and curation require technical knowledge of animal husbandry practices and burning properties (Sillar, 2000; Winterhalder et al., 1974). Sillar (2000, 49–50) notes that camelid dung collection “is facilitated by the tendency of camelids to a restricted area, often near a water source, where they defecate in communal piles.” After drying in the sun, people can easily collect these territory-marking piles of dung pellets from their herds (Pearsall, 2000; Sillar, 2000:49–50).

After conducting experimental botanical analysis of the seeds found in modern llama dung, Hastorf and Wright (1998:222) concluded that many small herbaceous seeds recovered in archaeobotanical samples at Altiplano sites likely come from burnt dung. Additionally, in parts of the world where dung burning is common, many carbonized seeds recovered from paleoethnobotanical samples are partially fragmented, which is interpreted as the remnants of the plants consumed by herd animals (Miller, 1984). We therefore expect to find a high concentration of burned small herbaceous seeds and partially fragmented seeds if dung fueled the Structure 4 fire.

Paleoethnobotanical Analyses

Aims and Methods

Since geoarchaeological analyses indicated a high sustained temperature for Structure 4, the aim of the paleoethnobotanical analyses was to assess the type of supplemental fuel used to alight the structure. During the 2006 excavations, a bulk sediment sample of approximately 10 L was recovered from each excavated locus (Pearsall, 2015; Lennstrom & Hastorf, 1992) and floated in the field using a modified SMAP (Hastorf, 1999b; Watson, 1976) with a daily poppy seed test to ensure effective flotation recovery (Wagner, 1982). Eighteen flotation samples were analyzed from Structure 4 fills, floors, and bins. Langlie conducted this analysis with a binocular light microscope with 5 to 40 × magnification in the McCown Archaeobotanical Laboratory, University of California Berkeley (2009). Carbonized seed specimens were identified to family, genus, and species where possible. Wood and parenchyma (starchy plant storage tissue, usually evidence of tubers in the region) were identified for only the > 2 mm fraction; unidentifiable fragments were excluded from further analysis.

Analysis

Overall, we consider wood, straw, and dung to be the most likely fuels for the Chiripa Structure 4 fire. If the only fuel was roofing, then wood should concentrate in the structure’s perimeter, and ichu and totora should be distributed throughout the structure’s deposits. We propose to differentiate wood fuel from beams by a spatial concentration of wood remains. Likewise, supplemental straw (sedge, grass, and chenopod) fuel might be detected as a stem concentration or the seeds of these species in the central room fill. Finally, dung could be detected through identification of dung fragments, and also high quantities or concentrations of small herbaceous seeds, particularly from Relbunium sp. and Malvaceae seeds (Hastorf & Wright, 1998).

To assess the relative fuel amendments, we used ubiquity expressed as percentage presence and standardized density expressed as specimens per liter of soil. We used ArcGIS Pro to map the standardized densities of wood, camelid forage, and human foods across Structure 4 (Fig. 10).

Results

Of the plant taxa identified at Chiripa, only species related to fuels and architecture are presented in Table 10 and Fig. 10. The remaining primarily food-related species are presented elsewhere (Langlie, 2009). Chenopods have the highest ubiquity and standardized density, while Fabaceae, Malvaceae, and Poaceae have high ubiquity values, but comparatively lower standardized densities. Wood was found in half of the samples. Wood was identified in high densities in three samples, with very low densities elsewhere. Relbunium sp. and Cactaceae were sparsely represented. No straw or dung was identified, but traces of these fuels may be found in seed remains as suggested below.

Figure 11 and Table 10 present the results of the spatial distribution of the plant standardized densities of fuel, food, and architectural remains. Interestingly, wood only concentrated in one bin in the western portion of the structure. Excavators indicated that it looked like a wood roof beam was burnt in this location. Wood was very sparse in the central room. Small amounts of dung and forage seeds were found including seeds of Fabaceae (Trifolium amabile), Malvaceae, and Relbunium sp. These were found in comparatively low quantities across the structure. Cyperaceae and Poaceae seeds were found in similarly low quantities. The co-occurrence of Cyperaceae and Poaceae in two samples along with significant amounts of camelid forage plants is evidence of forage/dung, rather than roofing material. Small amounts were found in the western bin, as well as in two samples in the central room. Human food plant remains, Chenopodium spp., and Cactaceae seeds were the predominant plant remains in both the central room and bins (Fig. 11B).

Fig. 11

Standardized density distributions of macrobotanical remains across Structure 4 surfaces (see Table 11); A dung taxa: Fabaceae, Malvaceae, and Relbunium sp. seeds; B food species: Chenopodium spp. and Cactaceae seeds; C thatch: Poaceae and Cyperaceae seeds; D wood fragments of tree and shrub, > 2 mm in size

Discussion

The paleoethnobotanical results provide support for our contention that supplemental fuel was added to alight Structure 4. The high ubiquity values of Malvaceae, Poaceae, and Fabaceae point to dung use for fuel. While ubiquity values were high, standardized densities of these same taxa were low. Low standardized densities of forage species were found in both the central room fill and the bins pointing to the possibility that dung may have been added in this location and turned to ash during burning. While grass and sedge seeds can be interpreted as evidence of dung burning and/or roofing collapse, the low quantities of these taxa make interpreting the origins of the use of these seeds difficult.

The modest, disbursed quantity of human food remains, particularly chenopods, was surprising, but not wholly unexpected given that we know quinoa was stored in other bins on the mound, supporting the hypothesis that bins were used as special food storage locations (Bruno & Whitehead, 2003; Chávez Mohr, 1988). On the other hand, chenopods are also camelid forage in the Andes, and their recovery in all 18 analyzed samples in this study could be interpreted as further evidence of dung burning. It is possible that future analysis of phytoliths and fecal struvites across Structure 4’s floor could clarify the hypothesis that dung was a significant fuel amendment in the fire.

In contrast to the seed remains, which were found throughout the structure, wood was found in only half of the analyzed samples. Furthermore, it concentrates in high densities only in one bin located along the periphery of the structure. These findings conform to the model of wooden beams collapsing and burning at the edge of the structure.

It is notable that the overall densities of herbaceous seeds and thatch materials are low compared to residential contexts at the site (Bruno, 2008). One explanation is that given the evidence for temperatures exceeding 600 °C discussed above, much of the botanical matter inside the central portion of Structure 4 would have been burnt to ash in such a hot blaze, also noted in the red sediment collapsed on the floor. The residual botanical assemblage in the central room reflects those pockets that experienced more reducing environments in their combustion after the roof fell in. Nevertheless, the high ubiquity of chenopods, Poaceae, Malvaceae, and Fabaceae, as well as the distribution of these small herbaceous seeds, provides tentative support for supplemental dung fuel added into Structure 4, not wood or grass.

Discussion

Initially, we hypothesized that the high chroma rubified mudbricks recovered from the central room of Structure 4 indicated a sustained, high temperature fire under oxidizing conditions. We suggested that thatch from a burning roof would not sustain such a fire long enough to alter the sediments as much as they did, as reducing conditions would have occurred once the roof fell. Therefore, we hypothesized that a supplemental fuel source produced the conditions observed in Structure 4. Supplemental fuel is a hallmark of a planned fire, and we therefore proposed that Structure 4 was intentionally burned in antiquity. Our analyses of sediment, bone, and plant remains confirm a sustained high temperature fire under oxidizing conditions from a supplemental fuel. We therefore consider the fire was both planned and managed.

The open pit experiment with dung fuel demonstrated that even an open-air fire requires maintenance and some degree of skill to keep burning. A burning time of over 5 h with a generous dung supply and several attendants was insufficient to reproduce the rubified Structure 4 mudbricks. High chroma red was only observed at exterior corners, and beneath the surface on two bricks to a depth of 2 cm. We were able to replicate the sediment alteration seen in Structure 4 bricks in a laboratory muffle furnace at sustained temperatures under oxidizing conditions over 500 °C. Furthermore, thin sections of rubified archaeological bricks show morphological changes in void shape and complete combustion of organics, which requires that the high temperatures be sustained for several hours. An explanation for how the Chiripa sediments became highly burnt is therefore required.

The most expedient explanation for the fire conditions is that the fire was ignited as a closed system with the roof intact. We interpret our paleoethnobotanical results as indicating that a fire was set in the central room with totora stalks and dung fuel. This would create the oxidizing conditions necessary to initiate the observed color changes and concentrate the heat within a thick, walled structure (Canti & Linford, 2000; Harrison, 2013; Icove et al., 2006). Structure 4 had only a single entrance towards the courtyard interior. We cannot evaluate if this would provide enough oxygen to maintain the fire at this high elevation, although the open-air experiments indicate this is a concern. Favorable oxidation conditions could be created from an external source such as holes in the roof, bellows, or blow tubes (Berna et al. (2007). No such technologies have been demonstrated on the Altiplano, although blow tubes were used with kilns on the Peruvian coast in Moche times (200 BCE–1000 CE) as demonstrated in a ceramic vessel on display at the National Museum of Archeology and Anthropology in Lima, Peru (see Shimada et al., 2007).

This study confirmed that the structure fire was sustained in excess of 600 °C, and may have reached 900 °C. Concentrations of calcined bone and sediment refiring colors and mineral alteration indicate that the Structure 4 fire was sustained at 600–900 °C long enough to alter both the bone and sediments in the central room while the bins experienced lower temperatures. Therefore, we can reject the functional argument that the fire was planned to fumigate the bins. The temperatures and conditions we identified are significantly higher than those required for cooking, or to produce the small reflooring fires recovered previously in a different mound structure. The temperature, oxidation condition, and duration of the Structure 4 fire are similar to ceramic firing. A substantial fuel supply and experienced fire tenders would be needed to sustain these conditions at this high elevation locale. Although we have no evidence to suggest that potters burned down this structure at Chiripa, they would have possessed the technical knowledge necessary to produce this fire. The fine polychrome Middle Formative Chiripa ceramics demonstrate that ceramic firing technology was well developed, although ceramic production sites in the region have yet to be identified (Roddick, 2002, 2009; Steadman, 1999, 2007).

The paleoethnobotanical results indicate that wood is an unlikely fuel source, since wood fragments only concentrate in one bin. A similar pattern was observed in experimental house fires where burnt roof beams landed at the edges (Bankoff & Winter, 1979; Friede & Steel, 1980; Icove et al., 2006). The ubiquitous but low quantities of small herbaceous seeds and chenopods may be interpreted as a dung fuel source. Dung was certainly in use at Chiripa for fuel (Bennett, 1936; Browman, 1989; Bruno & Hastorf, 2016; Whitehead, 2006, 2007). The overall seed densities recovered from Structure 4 were low, and the evidence for dung fuel is not strong. Due to the high temperatures, botanical matter would be converted to ash, as observed in events D516, D521, and D528. Future phytolith and fecal struvite analyses from the ash levels may provide further support for dung fuel for the Structure 4 fire. In contrast, bone (or potentially meat on the bone) was deposited in the roof or above the burning deposits and not on the floor. This curious position supports intentional consumption in the blaze, such as a meat/bone offering, which was recovered elsewhere at Chiripa.

If the Structure 4 fire was intentionally started in the central room and managed, as we propose, the question remains, why were some of the upper-level structures burned and then buried? This fire was not an isolated event as Structures 1 and 2 also burned, and charred materials were found across the upper level (Bennett, 1936). Clues can be gleaned from the observation that no new structures were built on the mound in the aftermath of the blaze despite centuries of post-fire use, as all structures were systematically filled in creating a level U-shaped surface (Hastorf et al., 2008). After the fire, the building remains were leveled and a prepared orange surface laid over the perimeter of the central sunken court for subsequent use.

Although we proposed that arson fires are incipient, there is a possibility that the planned fire that destroyed the upper level was malicious. The nature of prehistoric intergroup conflict in the Andes is debated, and competing models include ritualized warfare in the tinku tradition, militaristic warfare (ch’axwa), and warfare for conquest (e.g., Arkush, 2009; Arkush & Stanish, 2005; Levine, 2013; Ogburn, 2012; Platt, 1987; Stanish & Levine, 2011). The best regional evidence for conflict in the Formative, and the role of fire in it, is from the northern Titicaca Basin site of Taraco, Peru. Regional surveys and excavations documented a proliferation of settlements with sunken courts during the Middle Formative (1350–500 BCE) (e.g., Cohen, 2010; Levine, 2012; Plourde, 2006; Stanish et al., 2014). By the end of this period, the Peruvian sites of Taraco and Pukara became notable regional centers (Levine, 2012). However, in the first century C.E., a fire scorched the residential area of Taraco (Peru) torching roofs and rubifying walls (Levine, 2013; Stanish & Levine, 2011). After the Taraco (Peru) fire, that sector was not immediately rebuilt, residents engaged in less long-distance trade, and produced poorer masonry. Stanish and Levine (2011) associated these trends with the waning of Taraco (Peru) as a regional center as the competing site of Pukara ascended to political dominance.

There is currently no direct archaeological evidence of intergroup conflict in the southern Titicaca Basin where this study was conducted during the Middle Formative. On the Taraco Peninsula, there is evidence for increasing population during this time (Bandy, 2001, 2004a), which can lead to circumscription of resources and motivate conflict (Carniero, 1970, 1988). However, there are no indicators of warfare noted in human remains, settlement patterning, or weaponry, in contrast to the northern Titicaca Basin.

Our findings point to ceremonial closure and renewal as the explanation for incineration of this civic-ceremonial structure. The case for ritual burning of the upper-level rests on an understanding of the Chiripa site as a longstanding locus for ceremonialism and community cohesion, which continued throughout the Middle Formative (Bandy, 1999b, 2001, 2006; Beck, 2004; Chávez Mohr, 1988; Hastorf, 1999a, 2003). Hastorf (2003) contrasts the relatively public, communal activities in and around the sunken courts with the relative secrecy of these small upper structures with their sliding doors. The small-scale burning events within the structures were interpreted as ritualized events of the social group (Bandy, 2001, 2006). Hastorf (2003) interprets the upper level as being conceived of as a single construction project with significant investment of labor and planning that helped forge the identity of the expanding Chiripa population. The participation of different regional groups in building the upper structures is supported by the differences in manufacturing compositions and techniques documented between the mudbricks used to build the different structures (Davenport, 2010, Goodman Elgar et al., 2015). The different mudbrick manufacturing sequences provide evidence of segmentary lineages or families participating in its construction and use. Despite variable construction materials, the final architecture of these structures was remarkably consistent. The destruction by fire, infilling the structures, leveling, and creating a single surface encircling the sunken court are all representative of an emerging unification of the Chiripa community. The era of subgroups using small structures ended a practice of social separation. The subsequent phase marked a political transition towards a more unified ceremonial locale. Understood in this context, the upper structure conflagrations heralded political change as well as spiritual transition.

The conspicuous consumption of fuel suggested by paleoethnobotanical analysis also points to a ritual purpose. In the Andean highlands, fuels have other uses such as using straw for fodder and in mudbricks, and dung for fertilizer (Winterhalder et al., 1974; Arnold, 1993, 55–6; Sillar, 2000). The high intensity burning evidenced in Structure 4 represents a costly activity that consumed valued resources. The choice to burn Structure 4 therefore indicates the high value of the burning event to those who set it. This activity is more consistent with ceremonial activity than with arson, where fuels are generally expedient (e.g., roofing) and high temperatures are not expected. A similar study in Ukraine found that significantly more wood was needed to burn structures down than was needed to build them in the first place (Gaydarska & Chapman, 2020; Johnston, 2020). Such a conspicuous consumption of fuel led researchers to similarly conclude that the fire was intentionally set. At Chiripa, the bone in the upper fill levels may also be significant as ritual offerings.

Finally, ritual burning is consistent with the enduring use of the Chiripa mound for cleansing, curation, and burial (Machicado Murillo, 2009), and its potential as a central locus for community identity and memory (Hastorf, 2003). Differences between structures in the number of burials and their grave goods illustrate differentiation between Middle Formative sub-groups within the community. In this context, ritual activities in this upper level memorialize not only the religious aspirations of the Chiripa people, but also manifest their internal political relations. These events might in fact have memorialized the mound, re-marking it as a place of special value and perpetuating its relevance.

The dynamics of commemorative social practices within sacred places and ritual architecture is an enduring focus in the Andes (e.g., Allen, 2012; Angelo, 2014; Moore, 2010; Roddick, 2013; Vega-Centeno, 2007). The syntheses of agency, structure, and memory found in practice theory provide an emphasis on the dynamics of culture making (Moore, 2010). The study of prehistoric Andean monuments has been enlivened by contextually informed actors working collectively in performative acts that reference the past within the exigencies of the present. For instance, in his reconstruction of Cerro Lampay, Vega-Centeno (2007) suggests that the food remains recovered within the structural foundations indicate that construction and closure were as much part of the ceremonial life of its builders as was the use of the structure. An emphasis on dynamic culture making is apt for interpretations of this burnt horizon at Chiripa which, in its novelty, presents an active rupture with centuries of practice. The fire would have been visible for several kilometers and seen across the southern portion of the lake if it burned at night. If we agree that the structures of the Chiripa mound were used by lineages, then we suggest that the act of burning/cleansing/transforming these sacred places was an important political act by individual lineages. Burning the Chiripa structures may have been a social signal of change at Chiripa. In engulfing some of the upper structures in flames, tensions arising from inequities between the lineages proposed by Bandy (2001, 2006) may have been temporarily eased through an experience of loss, offering, and transformation experienced by the community.

Conclusions

In this study, we brought together independent lines of evidence to reconstruct the fire that consumed Structure 4 at Chiripa. Geoarchaeological and faunal evidence support a sustained temperature in excess of 600 °C. Under Altiplano conditions, this required a great deal of planning and tending. The botanical data provides tentative evidence that dung was used as fuel, and the thatch roofing also burned. There is no evidence that significant amounts of wood were added, or if it was, it burnt completely to ash. Experimentally, we found that an open-air dung fire required regular tending or it went out. The hottest part of the blaze centered in the main room rather than the ancillary bins. A sustained high temperature fire under oxidizing conditions indicates that the Structure 4 fire was both planned and managed over several hours.

We suggest that this evidence supports an interpretation of planned burning for ritual purposes. We suggest that the Chiripa community ceremonially closed the upper-level structures before embarking on new practices in the later Middle Formative (Hastorf, 2003; Hastorf et al., 2008). As noted, the burned level on the Chiripa mound is not dissimilar in intensity to the vitrified patterns observed at nearby Taraco, Peru, which was interpreted as arson by Stanish and Levine (2012). However, there are major differences in these two burning events leading us to a different interpretation for Chiripa. At Taraco, Peru, it was the residential sector that was burnt, valuable wooden beams were left in place, and the site was not immediately rebuilt or used, leading researchers to suggest the fire there was the result of raiding and conflict, and subsequently abandoned. The Chiripa mound fires were restricted to individual structures and the platform was resurfaced and continued to be used ceremonially. Based on scarce artifact densities, the structures were cleared before the burning events, further indicating that residents planned the fire of Structure 4.

The burning of a structure on the Chiripa mound would have been an event to behold. The spectacle lasted several days so residents across the region would have known that Chiripa was burning. Bright yellow and orange flames would have been visible at night from many kilometers away and plumes of smoke would have billowed into the atmosphere during the day. For those that attended the fire, the sheer heat that it generated would have been impressive. The large amount of fuel consumed undoubtedly required collection before this event could occur. Further experiments are needed to estimate how much fuel would be needed to experimentally burn a structure like the one considered here (e.g., following Johnston, 2020).

The intentional kindling of the upper levels of the Chiripa mound point to participatory dynamics, such as those identified by Vega Centano at Cerro Lampay (2007) and represent social and political transitions activity within this enduring community. The burning event did not mark the end of the Chiripa mound’s ceremonial use-life but instead marked a shared closure of the prior ceremonial organization and the beginning of a new community dynamic. During the Formative, regular ceremonial reconstruction, reconstruction and reorganization of civic ceremonial space clearly helped to maintain the vitality of the Chiripa mound and its supporting community over its centuries of use.