The Archaeology of an Agave Roasting Location

Mary Van Buren, James M. Skibo, and Alan P. Sullivan III

Rockpiles and roasting pits, distinctive features of fields on mid-bajada slopes in Zone 2 of the Marana Community, form extensive agricultural complexes for cultivating and processing agave (Chapter 7). Site AZ AA:12:205 is one of the smaller examples, containing only a few rockpiles and one roasting pit. The site was investigated in order to clarify the nature of the artifact assemblage and activities associated with such small sets of features.


Site 205 is situated on a gently sloping ridge between two northeast to southwest trending washes (Fig. 8.1). On-site vegetation is dominated by bursage and creosote bush, but saguaro and cholla also occur, and leguminous trees line the adjacent drainages. Site features consist of a roasting pit and three rockpiles (Fig. 8.2). Associated with these is an artifact assemblage comprised primarily of plain ware pottery (870 sherds), although a handful of decorated ceramics (21 sherds) and lithics (55 specimens) were also recovered. Temporally diagnostic designs could not be discerned on the decorated sherds. The artifacts were scattered over 1230 square meters, mostly in and around the roasting pit. No remains of dwellings or temporary shelters were apparent on the surface, nor were any revealed during excavation.


Although mapping and collecting by Northern Tucson Basin Survey crews had provided detailed information about surface characteristics of rockpile fields, understanding the activities that occurred at these locales required more intensive study. To further investigate the agricultural nature of these sites and the presumed cultivation of agave, and to provide data regarding their precise functions and periods of use, a sample of rockpiles and 16 roasting pits at a number of locations were excavated during the spring of 1984 and 1985 with the aid of the University of Arizona undergraduate field class. As a part of this research program, the excavation of Site 205 was motivated by three concerns. The first was to identify the botanical resources produced and processed at the site. A second goal was to provide additional information about the function of rockpiles. The foremost aim was to clarify the nature of the occupation and artifact assemblage so as to better understand how such small sites articulated with the larger subsistence and settlement system.

Surface Collection and
Stratigraphic Tests

A grid system of 2-m squares was imposed over the site and an intensive surface collection was conducted. Excavation of the roasting pit and rockpiles was preceded by systematic random testing on nonfeature areas in order to evaluate the relationship between surface and subsurface remains (Fig. 8.2). One to three sherds were found in six 1-m by 1-m units, and others had no cultural material. The stratigraphy of these test pits was fairly uniform. Reddish brown sediment containing pebbles and small cobbles was underlain by a solid layer of caliche from 5 cm to 27 cm below the surface. In one unit, however, ashy fill occurred both above and below a broken layer of caliche. Seven sherds were recovered from the upper 30 cm. These results provided convincing evidence that surface remains were reliable indicators of subsurface deposits, and that few artifacts would be overlooked if only features visible on the surface were excavated.

Figure 8.1. Location of site AZ AA:12:205 in the Marana survey area.
Other archaeological sites are depicted as irregularly shaped outlines.

Figure 8.2. Map of site AZ AA:12:205.



The three rockpiles at Site 205 were located 18 m, 33 m, and 38 m southwest of the roasting pit. Each was approximately 1 m in diameter and 10 cm to 15 cm high. The rocks comprising the piles were identical to those littering the surface of the rest of the site and do not seem to have been selected on the basis of any distinctive characteristics. The rockpiles were bisected (Fig. 8.3) and completely dismantled after profiles were drawn, and the surrounding soil was then stripped to reveal any signs of related features such as postholes or brush structures for directing runoff. In all cases the rockpiles consisted of a layer of rocks underlain by 2 cm to 10 cm of fine, grayish brown soil intermixed with pebbles and small cobbles. A caliche layer immediately beneath this soil was not broken or otherwise modified by human activity. No artifacts were recovered from the rockpiles. The units excavated adjacent to the rockpiles revealed a similar soil and caliche sequence.

Roasting Pit

On the ground surface the roasting pit appeared as a shallow depression surrounded by a band of pebbles and small cobbles. This rock ring was not intentionally constructed, but resulted from the prehistoric excavation of the pit and the subsequent removal of its contents. The depression measured approximately 8 m in diameter and contained soft, ashy fill. The highest concentration of surface artifacts occurred in this area. Excavation revealed that actual pit dimensions beneath the surface depression were approximately 2.25 m in diameter and 1 m in depth (Fig. 8.4). The pit was slightly conical and was not lined with rocks. Fill consisted of dark, ashy sediment mixed with fire-cracked

Figure 8.3. Cross section of a bisected rockpile, showing the stratigraphy
and the relationship to the underlying caliche layer at site AZ AA: 12:205.

Figure 8.4. Excavated roasting pit at site AZ AA: 12:205, showing the size and
depth, fire-cracked rock, and ashy fill containing charred vegetal material.

Table 8.1. Wood Charcoal Recovered by Flotation
from the Roasting Pit at Site AZ AAA 2:205

rocks and charcoal. Toward the bottom, the fill became black and greasy in texture. A solid mass of fused organic matter measuring about 30 cm. in diameter was at the base of the feature. Within the pit were 145 sherds, 16 percent of the total pottery recovered at Site 205.

The roasting area at Site 205 is small compared to excavated pits at several large rockpile fields in the Marana Community. S. Fish, P Fish, and Madsen (1985; Chapter 7 in this volume) note that pit size tends to follow field size. Excavation did not reveal any stratigraphic traces of repeated use, nor were multiple pits present. However, the large amount of ash and firecracked rock at the site, as well as the quantity of artifacts, implies repeated use.

Flotation samples from the roasting pit indicated that only agave was processed, a pattern typifying excavated pits in the northern Tucson Basin. Charred agave was recovered from 23 percent of the samples. Exploitation appears to have focused on the roasting of agave hearts (83% of identified plant parts) and, perhaps, leaves (17%). The predominant fuel species, amounting to 95 percent of the wood charcoal, were ironwood, mesquite, and palo verde (Table 8.1).

A radiocarbon assay on agave yielded a calibrated age of 1040 90 years B.P. (Beta 10801), indicating that the site was used sometime between A.D. 894 and 1148 (one sigma). The midpoint of this determination, A.D. 1011, would place the site somewhat earlier than the large agave fields associated with the Marana Community, which date primarily to the early Classic period (S. Fish, R Fish, Miksicek, and Madsen 1985).


Tool and Debitage Analysis

A total of 55 flaked stone artifacts was recovered from surface collections and excavation at Site 205. According to the classification methods employed here(Sullivan 1983; Rozen 1984; Sullivan and Rozen 1985), this small collection comprised 49 pieces of debitage divided among five categories, two unifacially retouched pieces, and four core tools. The two tools with positive and negative percussion features were found in widely separated units.

Most of the debitage is quartzite (83.7% of the collection), with rhyolite (10.2%) and limestone (6.1%) composing the remainder. Rhyolite and limestone were brought to the site from sources elsewhere. Complete flakes (51.0%) and split flakes (18.4%) dominate the debitage; flake fragments (12.2%), debris (12.2%), and broken flakes (6.1%) equal less than one-third of the assemblage (30.5%). A comparatively high combined frequency for complete flakes, split flakes, and debris (81.6%) strongly suggests an emphasis on hard hammer core reduction (following Rozen 1984: 588).

Patterns of Artifact Selection

In general, it appears that relatively thick debitage, which may have been more resistant to stress failures, was selected for use in activities related to agave harvesting and processing. Half of the debitage with edge damage is rhyolite (although it equals only 10.2% of all debitage), and four out of five pieces of rhyolite debitage show edge damage. Rhyolite seems to have been selected because either its edges are sharper or it holds a sharper edge longer. It is likely that the rhyolite debitage was produced elsewhere and transported to Site 205, because no rhyolite cores were recovered and four of the five rhyolite artifacts were complete flakes.

The two unifacially retouched tools and the four core tools from Site 205 share some functional characteristics in spite of their typological differences. In contrast to the eight pieces of edge-damaged debitage, these relatively "massive" artifacts appear to have been designed not for resistance to lateral stresses, but rather for their resistance to perpendicularly transmitted stress. Similar core tools or "pulping planes" are consistently found in large rockpile fields of the Marana Community (Chapter 7).


A review by Doelle of the material correlates of ethnographically documented Southwestern agave exploitation makes no mention of ceramics used in harvesting or roasting activities. However, jars were used in secondary processes such as boiling, storing processed products, and rehydrating dried agave prior to consumption (Doelle 1980: 95-97).

When a liquid product was made from agave (see Castetter and others 1938), impermeable containers were used. For instance, Felger and Moser (1970) state that the Seri sometimes fermented agave leaves in sea turtle shells to make wine, and Doelle (1980: 96) mentions that watertight baskets were used to catch the juice extracted from pounded leaves. Roasted agave was occasionally boiled by the Pima to produce syrup, and many Apache groups as well as the Thrahumara made fermented drinks from the roasted crowns (Castetter and others 1938: 49, 60-63).

Presumably the technology involved in processing liquid from agaves would parallel that used to process saguaro fruit into syrup (Crosswhite 1980) and might leave comparable traces in the archaeological record (following Raab 1973). The relatively abundant pottery at Site 205 could indicate a processing of agave that involved the extraction of liquids. Alternatively, the sherds could have been the remains of vessels broken during the course of water transport and domestic tasks associated with resource production and processing activities. In both cases, breakage of whole vessels would account for the presence of sherds. By determining the number, type, and spatial distribution of vessels, more precise information might be gained about how ceramics had been used at the site, the configuration of activity loci, and the manner in which sherds had entered the archaeological record.

Ceramic Analysis

The separation of sherds, into groups representing parent vessels was based on paste characteristics (type, size, and density of temper and the presence, color, and configuration of the carbon streak), surface finish, color, and thickness (Sullivan 1983). Small, severely eroded sherds, and completely charred ones (20% of the total) were not assigned to vessels. Initial homogenous batches were then sorted into vessel fragments, or aggregates of batches that exhibited a degree of internal variability but which were very different from one another. These larger groupings indicated the minimum number of vessels.

The results of the analysis contradicted original assumptions about the assemblage. The ceramics recovered from Site 205 were not the remains of whole vessels used and broken at the site, but instead represented a tool kit composed of sherds that were transported to the site for use in agave processing. Several lines of evidence support this conclusion. The number of vessels represented at Site 205 is large compared to expectations for such a small, special purpose site. Based on a conservative count, portions of 82 different vessels were identified. None of these vessels

Table 8.2. Weights and Numbers of Sherds for 82 Vessel
Fragments from the Ceramic Assemblage at Site AZ AA:12:205

are even partially reconstructible, and conjoinable sherds are rare (7% of the total). Of the 82 vessel fragments identified, 61 (74.4%) are represented by 9 or fewer sherds, and the mean number of sherds per vessel is only 8.7 (Table 8.2; Fig. 8.5).

More informative than sherd counts, however, are the weights of vessels, because they more accurately reflect the degree to which these vessel fragments approximate whole pots. At Site 205, 72 of the vessel fragments weigh 57 g or less. The range is wide, however, with the smallest fragment weighing only 2.1 g and the largest 522.4 g. In comparison, the average weight of 20 whole

Figure 8.5. Scattergram of the weight and number of sherds
of the 82 refitted vessel fragments from site AZ AA:12:205.

Table 8.3. Weights of 20 Whole Vessels
from southern Arizona

and nearly intact vessels from southern Arizona in the Arizona State Museum collection is 1401.9 g (Table 8.3) Ten bowls, with a mean weight of 874 g, weigh substantially less than jars, which average 1919.9 g. The smallest bowl weighs 347 g and the largest jar weighs 4180.4 g, 89.4 percent of the total weight of vessel fragments recovered from Site 205. Only three vessel fragments weigh as much as or more than the smallest bowls in the whole vessel sample (Fig. 8.5.). Rim sherds included with these vessel portions, however, indicate that these are not small, nearly intact bowls, but heavy, relatively large jars.

The low weights of vessel segments from Site 205, the large number of vessels represented, and their fragmentary condition support the contention that the ceramics from this site are the remains of broken or discarded sherds rather than broken vessels. Data collection techniques were through, and this finding was not the result of an incomplete recovery of artifacts. All surface sherds were collected, the features were excavated and screened, and testing revealed the few artifacts occurred outside

Figure 8.6. Sherd tools from roasting pits. The proximal ends of a and b from site AZ A-k-12:205
(fragments of larger sherd tools) and the right side of b display rounded edges due to wear. c is a complete
tool from site AZ AA-12:470 and shows wear along its proximal edge and both lateral edges. Length of c is 20.8 cm.

these contexts. In addition, the geological surface on which the site is located is stable. Although some downhill movement of artifacts appears to have occurred, the sherds show no traces of having been subjected to fluvial action strong enough to move them far from the area in which they were deposited (Skibo 1987).

Sherds as Tools

At Site 205 the unusual nature of the ceramic assemblage indicates that sherds, rather than pots, were transported to the site for use in resource processing activities. No intact sherd tools were found, perhaps because they were discarded only after their use-lives had expired. Sherds were probably gathered from trash middens associated with nearby residential sites.

Physical traces on the sherds themselves, and to some extent their spatial contexts, provide information about how these artifacts were used. The concentration of sherds around and within the roasting pit and the presence of burned sherds suggest that most were used directly in the roasting process, perhaps in some cases to keep the agave from contacting the coals or soil. Although ethnographic accounts reveal that sherds were used to parch seeds (Castetter and Underhill 1935: 25; Goodyear 1975: 171), there is no evidence that sherds were used for this purpose at Site 205. Parching food over an open fire might soot the exterior of a sherd, but would not turn it completely gray, a common attribute of pottery at the site and one that is only produced in a nonoxidizing atmosphere (P. Rice 1987: 335, 343).

Sixteen vessel fragments also exhibit distinctive forms of edge wear unlike that produced by fluvial abrasion (Skibo 1987). Most of these have rounded edges that are sometimes associated with parallel striations on the sherd surface (Fig. 8.6). These artifacts appear to have been used as scrapers or scoops, perhaps in the processing of agave flesh or fiber, but more likely in the construction of the pit or in the removal of the hot, ashy contents.

In addition to sherds modified by use, seven sherds from five vessels have been purposefully shaped by chipping or grinding. Four are small disks, one is perforated, and two are square. The function of these artifacts in relation to agave cultivation and processing remains enigmatic.

Only five vessel fragments weighed 200 g or more (Table 8.2). All these heavier fragments have some form of edge wear even though only one-third of all vessel fragments from the site were worn. A complete scoop shaped from a large plain ware sherd was recovered from a roasting pit at a large rockpile field, AZ AA:12:470 (Fig. 8.6c). It weighed 277 g and had obvious traces of abrasion. Large, heavy sherds appear to have been purposefully selected for such tools.

If large sherds were brought to the site as protective coverings or as tools, several possibilities may explain the presence of the numerous small sherds. Small sherds could have been detached from larger ones when the pit was sealed with rocks and soil or when the agave was removed; in either case the potential for sherd breakage is high. If large sherds were used to excavate a pit or to clean it out, small pieces could easily have been snapped off the larger ones. In each case, the small sherds would have been incorporated readily into the matrix of the pit, whereas the larger sherds may have been saved and used on another occasion.


Excavation results indicate that the activities that occurred at Site 205 were limited in scope and of short duration. Apparently the site was used only for the cultivation and processing of agave. Agave remains were found in the roasting pit, and if the roasted agave was subsequently converted into syrup or wine, such secondary processing probably did not take place at the site.

No evidence of domestic or maintenance activities was observed, suggesting that the site was used for only short periods of time. Structural remains could not be discerned, and the site also lacked middens composed of domestic debris. The narrow range of artifact types in the assemblage supports the conclusion that agave processing was not associated with the seasonal occupation of a field house or with routine domestic tasks that would be performed during the course of a protracted stay.

The location of the site away from habitations could be due to a number of factors. It might have been more convenient, for instance, to process the agave where it was harvested and then to transport the processed product, rather than the raw material, back to the habitation locus. The depletion of firewood in the immediate vicinity of the residential area (Shackley 1983) as well as the desire to locate grimy roasting pits away from residences (Ambler 1987) may have influenced the spatial separation of agave processing from domestic environs.

The investigation of Site 205 provides some insight into the exploitation of agave by the Tucson Basin Hohokam. However, implications of the ceramic analysis go beyond the identification of the subsistence practices that occurred at this small site. Potsherds are invariably regarded by archaeologists as simply the remains of broken containers, despite ample ethnographic evidence of extensive recycling (R Rice 1987: 294). By assuming that variability in the attributes and distribution of sherds is directly linked to the prehistoric use of containers, archaeologists may be ignoring a wide range of functionally diagnostic activities associated with the secondary use of sherds as tools.

Next: Chapter 9 | Previous: Chapter 7

UAP Online Books  |   UAP Home Page

The University of Arizona Press, 2/10/01 5:04PM