*1.2. The Brewery Context*

The excavated area of the brewery constitutes approximately 500 square meters encompassing four distinct spaces (Rooms 1–4) as well as an additional set of smaller rooms to the southwest (Figure 3). Distinct rooms were dedicated to grinding maize, preparing comestibles, boiling the mash, and fermenting and storing the brew. The boiling room contained the remains of between eight and twelve hearths along the northwest wall, each framed by two upright stones placed 50 cm from the wall to support a ceramic boiling vat. The hearths contained the burnt remains of fuel ash, identified principally as dung, including guinea pig as well as camelid sources. Trash pits in the floor of the boiling room contained the remains of large numbers of the spent drupes of the Peruvian pepper berry, *S. molle* [16]. The fragmented remains of large cooking vessels were also recovered from this space, with sooted base fragments and clear evidence of use. One of these fragments was collected in situ and wrapped in foil for residue analysis, reported below.

**Figure 3.** Map of the brewery at Cerro Baúl, illustrating the fermentation/storage area (Room 1), boiling area (Room 2C), and the grinding and food preparation rooms (Rooms 4, 43A). Other rooms may have been used for storage or access to brewing facilities.

The adjacent fermentation or storage room contained the remains of several oversize jars, originally lined up along the northwest wall. These jars held up to 150 L of brew each and were decorated as Wari personages, with a face on the neck of the jar, or with a chevron headband painted on the rim of the neck as shown in Figure 2. Another large vat found in situ embedded in the floor was over a meter in diameter and may have held up to 1000 L of liquid. The fermentation room and the adjacent grinding rooms were the areas where most of the fineware fermentation vats and drinking and serving vessels were found smashed on the floors [10]. These serving vessels include cups of ca. 300 mL volume, keros (drinking mugs) of upwards of 2-L volumes, open bowls, and restricted jars. Four of the sherds from the fermentation/storage vessels were exported for residue sampling as described later. The highly decorated nature of the assemblage, the comparative high quality of the ceramic materials, and the highly privileged location of the brewery indicates it was an elite production space. While evidence of smaller scale brewing has been identified in intermediate elite contexts at the adjacent site of Cerro Mejia and Cerro Trapiche and in the houses of Tiwanaku leaders, we have not seen contemporary commoner houses with these production contexts.

Radiocarbon dates and evidence of modification of space in the brewery indicate a long occupation, with at least two phases of construction. Radiocarbon dates range from 1400 ± 60 BP to 900 ± 40 BP uncalibrated [10], which span the range of occupation of the Wari settlement as a whole. When the site was abandoned ca 1050 CE, the brewery was intentionally burnt to the ground and the fineware ceramic drinking vessels were smashed in the smoldering flames. As the fire extinguished, seven semi-precious stone and shell bead necklaces were placed atop the ashes and covered with sediment to cap the ritual closure of the facility. The large fermentation jars in the central storage room were also broken and pieces were dispersed throughout the brewery. These drinking cups, serving wares, and fermentation

vats were the samples for the brewery ceramic paste sourcing analysis discussed next, some of which are displayed in Figure 2.

#### **2. Materials and Methods**

#### *2.1. Ceramic Paste Analysis Methods*

A sample of 20 ceramic sherds from the Baúl brewery were analyzed by instrumental neutron activation analysis (INAA) at the Missouri University Research Reactor (MURR) and another 19 samples from nearby consumption contexts were analyzed by both INAA at MURR and by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at the Elemental Analysis Facility (EAF) at The Field Museum [19–21]. We also analyzed an additional 41 sherds from other summit contexts on Cerro Baúl for a total of 80 ceramic samples from the top of the mountain citadel (Table 1).The INAA dataset also contains comparative ceramic samples of 10 sherds from Formative sites in the local Moquegua region, 51 sherds from the Wari site of Cerro Mejia, 44 samples from Wari and Tiwanaku sites on the slopes of Cerro Baúl, as well as 29 ceramic sherds from the Wari heartland [19].

**Table 1.** Summary table of ceramic sourcing samples from Cerro Baúl and comparative sites.


Sharratt et al. [22] conducted a clay survey in 2005, providing 50 clay locales that were analyzed via LA-ICP-MS. In conjunction with this analysis, they analyzed an additional 29 Tiwanaku sherds from the local region and compared them with 19 of the 29 sherds from the Wari heartland to provide the link between local clay sources and ceramic production workshops [20–22]. This analysis complements the INAA data and the 19 samples from Baúl consumption contexts outside the brewery mentioned earlier.

INAA analytical methods are detailed in Glascock [23], while we describe ICP-MS methods here. A Varian inductively-coupled plasma-mass spectrometer (ICP-MS), equivalent to the Varian 810 instrument, was used at the EAF (quadrupole, auto-optimized spectrometer on 55 selected isotopes). The facility uses a New Wave UP213 (helium carrier gas, 213 nm laser operated at 0.2 mJ and a pulse frequency of 15 Hz) laser in conjunction with the LA-ICP-MS to introduce solid samples. Methodology follows that presented previously [20,22].

Ceramic samples were laser-ablated with a spot size of 150 microns and a dwell time of ninety seconds. Each sample was ablated ten times and a total of 55 elements was measured, using 29Si as an internal standard to control for time variability in ablation efficiency and resulting signal strength. Concentrations were calculated using NIST standards n610, n612, and Brick Clay (n679) via the approach first presented by Gratuze, Blet-Lemarquand, and Barrandon [24]. Large temper grains in the ceramic were avoided when positioning the laser so as to focus on the clays used in production. Ablation took place on sherd cross-sections to avoid pigments and treatments on ceramic surfaces.

Statistical analysis was performed using SPSS and Gauss Runtime statistical routines developed by MURR.A principal components analysis was carried out on the 16 elements that best produced group separation in the INAA and ICP-MS data [20]. We converted concentration values to base-10 logarithms to improve normality for trace element data and to minimize scaling differences of trace elements. We conducted a principal components analysis using elements from the variance–covariance matrix in order to examine the multivariate patterns in the data.

#### *2.2. Residue Analysis Methods*

In order to provide a baseline for residues within Wari ceramic vessels, Donna Nash conducted an experimental archaeology process for the production of *chicha de molle*. In collaboration with ethnographic informants, Nash prepared three types of *chicha* in newly fabricated clay vessels: *chicha de jora* (*maíz*), *chicha de molle*, and *chicha de maíz y molle*. Maize was germinated, then ground and boiled. *Molle* was winnowed and the seeds steeped in hot water for a short period of time. The liquid was then left to repose in clay fermentation vessels for five days before tasting. The *chicha* was left in the vessels for another five days, its maximum shelf life before consumption, and the empty clay vessels for each type of *chicha* were broken into sherds. Several years later, the clay vessel sherds were subjected to chemical analysis of residues embedded in the ceramic pores, as described herein. Likewise, five fragments from clay vessels from the ancient brewery at Cerro Baúl were subjected to the same residue analysis to ascertain which brew was chemically most similar to the archaeological residues embedded in the ceramic vessels.

Direct analysis in real time mass spectrometry (DART-MS) was employed to characterize potsherd residues and spent *molle* drupes recovered from the Cerro Baúl brewery, which were compared to experimental archaeology controls (produced by Nash and her ethnographic informants as described above).Analyses were carried out using a DART ionization source (IonSense, Saugus, MA) on a JEOL AccuTOF mass spectrometer, additional background on the DART-MS may be found elsewhere [25]. Spectra were collected in negative ion mode with the DART ion source at 500 ◦C using helium as the DART gas. Orifice 1 on the AccuTOF was set to −30 cV to minimize fragmentation; the ring lens voltage and orifice 2 were set to −5 V. The temperature at orifice 1 was held at 150 ◦C for analysis. The DART grid voltage was kept at the default value of −530 V. The mass spectrometer RF ion guide ("peaks voltage") was set to 800 V to maximize sensitivity above *m*/*z* 80. Mass calibration was carried out using PEG-600 in methanol during each acquisition. The mass resolving power was approximately 6000.

Individual *molle* drupes were held in a pair of tweezers and introduced into the space between the DART ion source and the AccuTOF mass spectrometer orifice. No sample preparation was required. For the ceramic materials, a small amount of the surface was removed using a rotary grinding tool

and placed into a small autosampler vial with a few drops of HPLC grade methanol. This ceramic slurry was introduced into the space between the DART ion source and the mass spectrometer orifice on the closed end of a capillary melting point tube. The methanol served primarily to allow a significant quantity of ceramic slurry to be subjected to ionization rather than as a solvent for the residue [26]. DART-MS has a distinct advantage over the traditional residue analysis methods like gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) in that it requires no sample preparation, or as in the case here, very little. The speed of analysis—just a few seconds in the ion source—comes at a cost: molecules can only be identified based on their molecular mass. Fragmentation can be accomplished by using a higher voltage on Orifice 1, but interpretation is complicated by the sheer number of ions produced from the mixture being ionized. This analysis was an initial attempt to determine if the pertinent residues could be distinguished with DART-MS, an approach that uses minimal solvent, no additional reagents, and requires very little time. It is, one could argue, a sustainable initial approach to screening ceramics prior to further study with more traditional methods like GC-MS or LC-MS.

The sensitivity of the DART-MS method is dependent upon the amount of material presented for ionization. For the ethnographic samples, the control ceramic and *chicha de jora* vessels were sampled both before and after being scrubbed with a brush under running water to simulate the process of field cleaning of ceramic samples. Each sample was run in triplicate unless otherwise noted. Compounds were identified in the spectra using modified versions of the databases compiled by one of us (JMH) of compounds that have been previously identified and isolated in *Schinus* spp. and in maize, from Reaxys and/or NAPRALERT web-based tools [27]. The modified databases removed compounds present as glycosides (or linked to other sugars) as these compounds do not readily ionize by DART-MS.

Identifications are based on the observed peaks being within 10 millimass units (±1 in the second decimal place) of the calculated exact mass of the compounds in the database. Each compound is identified based on only one mass, as the deprotonated species (molecule less one hydrogen atom, or M–H anion); as multiple compounds can share the same formula and thus the same mass, identifications must also be based on what compounds are present from this small subset derived from web-based tools and the analysts' (RAA and JMH) discernment in relation to the mass spectra.
