The cultural and chronological context of early Holocene maize and squash domestication in the Central Balsas River Valley, Mexico

The Xihuatoxtla Shelter (964 m asl) is formed by the overhang of an enormous boulder (17 × 16 × 14 m) and contains ≈75 m² of protected floor space inside the dripline (Fig. 2 and Fig. S1). It is located 260 m from Barranca Xihuatoxtla near the point where the now-seasonal stream opens out onto a flat basin. The nearly 1-m-deep stratigraphic sequence (Fig. 3) is anchored by 4 radiocarbon dates on charcoal: (1) 1200 ± 40 B.P. (1240–1000 cal B.P.) from layer B, 20–30 cm below surface, (2) 2790 ± 40 B.P. (2970–2780 cal B.P.) from layer B, 30–40 cm below surface, (3) 4730 ± 40 B.P. (5590–5320 cal B.P.) from layer C, 49 cm below surface, and (4) 7920 ± 40 B.P. (8990–8610 cal B.P.) from layer D, 65 cm below surface. The cultural deposits extended another 35 cm below the 8990–8610 cal B.P. level, but we were unable to date any of the samples recovered from these earlier deposits. Though starch grains and phytoliths were well preserved in stone tool residues, and phytoliths commonly occurred in the shelter sediments, pollen grains and plant macrofossils other than wood charcoal were scarce (28). No faunal remains were preserved. The sedimentary sequence at Xihuatoxtla can be divided into 5 major and discrete stratigraphic units (Fig. 3).

Layer A (≈0–10/15 cm below surface) is a 10- to 15-cm-thick deposit of very compact tan silt that contained numerous pre-Columbian potsherds (n = 414) and a modest number (n = 26) of lithic artifacts, primarily debitage but including a small tanged obsidian point and an obsidian blade fragment. A small number of bottle glass fragments (n = 3) are the only indication of recent disturbance (all counts given here and below are for units 1 and 2, contiguous 1 m² blocks).

Layer B (≈10/15–45/50 cm b.s.) is a 30- to 40-cm-thick deposit of clayey silt with small amounts of angular roof-fall that changes from gray/brown to brown in color with depth. Sherd densities vary from 110 sherds per 5-cm level per square meter at the top of the layer to 25 sherds per 5-cm level per square meter at the bottom. Lithic densities averaged 27 specimens per 5-cm level per square meter. Obsidian artifacts included 12 prismatic blade fragments, 1 tanged point, 1 tabular wedge (piece esquillée), and 9 flakes. Other chipped stone artifacts of local materials included 1 steep scraper, 1 flake knife, 1 graver, 1 chopper, and 1 small bipolar core. Also recovered from the layer were 3 handstones (manos, small implements used for grinding plant and possibly other kinds of materials), 2 milling stone base fragments, and 1 shaped metate fragment.

Layer C (≈45/50–60 cm b.s.) is an 8- to 12-cm-thick deposit of large angular roof-fall blocks (cobble sized) in a matrix of gray/brown clayey silt. The density of lithic artifacts (119 per 5-cm level per square meter), absence of obsidian blades, and the AMS ¹⁴C date of 4730 ± 40 B.P. (5590–5320 cal B.P.) indicate that this layer dates to the preceramic, although there is some intrusion of sherds from the ceramic occupation above (n = 36). The chipped stone tools recovered from layer C include 2 scraper planes, 1 borer, 1 flake knife, 1 bifacial core, 2 bipolar cores, 1 small obsidian biface fragment, and 4 used flakes. There were also 6 handstones (see Fig. 4 and Fig. S2A) and 1 possible milling stone base fragment recovered from layer C.

Layer D (≈60–70 cm b.s.) is an 8- to 10-cm-thick deposit of brown clayey silt with some angular roof-fall blocks. There were 251 chipped stone artifacts recovered (63 per 5-cm level per m²) and no ceramics. The chipped stone tools recovered from layer D include 1 cobble spall chopper, 1 concave scraper, 2 gravers, 5 flake knives, and 1 biface fragment. Also recovered were 2 handstones, and 1 complete and 2 fragmentary milling stone bases (see Fig. 4 and Fig. S2 B and C).

Layer E (≈70–95 cm b.s.) is an 8- to 25-cm-thick deposit of brown clayey silt with more angular roof-fall blocks than layer D. It rests on eroded bedrock (layer F) that slopes from 78 to 95 cm below surface moving from south to north (see Fig. 3). There were 179 chipped stone artifacts recovered (30 per 5-cm level per square meter), including the base of a stemmed, indented base point, the distal end of a thick lanceolate point (Fig. 5), 1 graver, 1 spokeshave, 1 flake knife, 9 used flakes, 5 thinning flakes with ground platforms, 2 core fragments, and 3 bipolar cores. Four handstones were recovered from this earliest occupation layer (Fig. 4 and Figs. S2D and S3).

Nearly all of the handstones and milling stone bases (Fig. 4) used for plant processing consisted of river cobbles and boulders that were modified only by use. One milling stone base fragment came from a stone slab (Fig. S2C) rather than a river boulder that was also modified only by use. Wear polish can occur on one or both faces of a handstone and on the tool edges as well. One handstone from layer E (318e) appears to have seen considerable use as a pestle. A number of the milling stone fragments exhibit polish on their flat or slightly concave surfaces as well. Striations are rare on both handstones and milling stone fragments but where they occur they are aligned in parallel to each other, suggesting a back and forth motion. An exception is the complete milling stone recovered from layer D (316d) that exhibited randomly oriented striations near the edges of a shallow depression 11 cm in diameter. The use of unmodified river cobbles as handstones and unmodified river boulders or stone slabs as milling stone bases is a common pattern observed in early Holocene sites in the semiarid Mexican highlands in the Tehuacan Valley (29) and at Guilá Naquitz in Oaxaca (1). Similar ground stone tools were also recovered in early Holocene contexts at the Santa Marta Rock Shelter (30), a site in the Central Chiapas Depression that sits at an elevation of 860 m in a zone of tropical deciduous forest.

Chipped stone tool manufacturing at Xihuatoxtla was an important activity during the preceramic period and continued into the ceramic period, albeit with less intensity (Table S2). A wide range of raw materials was used, including chalcedony, chert, quartzite, rhyolite, basalt, sandstone, and obsidian. All but the obsidian appear to be locally available. At the base of the sequence in layer E (≈75–95 cm below surface), evidence for bifacial reduction in the production of projectile points is indicated by thinning flakes oftentimes containing ground platforms. The preparation of the edge of a spear point preform by grinding is a way of increasing platform strength and allowing larger thinning flakes to be removed. Bifacial thinning flakes with ground platforms are characteristic of Paleoindian and Early Archaic periods (before ≈8000 cal years B.P.) (31), and their appearance in layer E is consistent with its estimated age (older than ≈9000 cal years B.P.) The basal half of a stemmed, indented base point and the distal end of a diamond cross-sectioned point were also recovered from this layer (Fig. 5). The point base is similar to single examples classified as Pedernales points from the Guilá Naquitz site in Oaxaca and from the Tehuacan Valley in contexts dated between ≈9510 and 7742 cal years B.P. (1, 29). Evidence of bifacial reduction continued in layer D (≈60–75 cm below surface) with the recovery of additional thinning flakes and also one biface preform fragment. Bifacial flaking occurred in layer C and in layer B associated with ceramics, albeit at reduced levels and with no platform preparation by grinding.

Of equal or greater importance in stone tool manufacturing was core reduction for flakes that could be used with little or no modification. Cores were reduced opportunistically or by bipolar methods. The resulting flakes were used as knives, scrapers, borers, spokeshaves, and gravers and represented the majority of the tools recovered in the preceramic deposits. Tools recovered from ceramic-aged deposits were mainly made on obsidian prismatic blades, although some tools and the bulk of the chipping debris were of local nonobsidian materials.

As reported in a companion paper (28), starch grains identified as maize were recovered, often in large numbers, from most artifacts identified as grinding stones found throughout the Xihuatoxtla stratigraphic sequence. All 8 of the grinding tools and both of the chipped stone tools examined for starch grains in levels contemporary with or below the ¹⁴C date of 7920 ± 40 B.P. (8990–8610 cal years B.P.) yielded maize starch. Maize phytoliths were recovered from both sediments and stone tools throughout the sequence, including those beneath and from the level that produced the 7920 ± 40 ¹⁴C B.P. date. Phytoliths identified as domesticated Cucurbita, possibly C. argyrosperma, were also found in direct association with the 7920 ± 40 ¹⁴C B.P. date at Xihuatoxtla and in levels below (see ref. 28 for details on starch grain and phytolith recovery and identification). Thus, the evidence firmly indicates that maize and squash were domesticated by 8990–8610 cal years B.P., the earliest date yet recorded for maize.

Although Mexico played a central role in New World agricultural origins, contributing a large number of important crop plants in addition to maize (9), little research directed toward understanding early human occupation and agricultural evolution has been carried out in its lowland tropical regions. Our investigations in northern Guerrero show that this region of tropical southwest Mexico supported persistent human occupation beginning by at least 9,000 years ago and possibly a few thousand years earlier, if some of the stone tools recovered from the surface of the Temaxcalapa Shelter turn out to indeed be Clovis in age. The archaeobotanical (28) and stone tool evidence from Xihuatoxtla indicates that plant exploitation and cultivation were important activities from the earliest stages of the site's occupation.

Xihuatoxtla is a small shelter that appears to have been repeatedly visited by small groups of people who stayed for significant periods of time (several weeks or more) based on the nature and density of materials left at the site. During the preceramic period, these groups likely shifted locations seasonally. Our data on early settlements from the Central Balsas region are insufficient to document this pattern in any detail. Nonetheless, the evidence we do have suggests that preceramic groups were occupying different settings and were engaged in different subsistence pursuits. The much larger El Abra Shelter, only 7 km distant from Xihuatoxtla and occupied over much of the same time span, included both small and large mammal bone in the limited sample of undisturbed deposits that were excavated in that site. The near absence of grinding stones at this site suggests that activities were directed more toward hunting than plant exploitation and cultivation. The surface collections from the Temaxcalapa Shelter, located at a higher elevation (1,350 m asl) than Xihuatoxtla (964 m asl) and in more broken terrain (Fig. 1), contained Archaic point types (Fig. 5) and other chipped stone tools (n = 38) and flakes (n = 128) but no grinding tools, hinting again at a different set of activities carried out there than at Xihuatoxtla.

Our paleoecological data from nearby lakes (Fig. 1) indicate that utilization of resource-rich lacustrine environments was an early and important component of subsistence and probably seasonal settlement cycles (27). At Laguna Tuxpan, phytolith data indicate that humans were exploiting lake edge settings and modifying the local vegetation, including with fire, during the early Holocene. Phytoliths of maize and domesticated Cucurbita plus Zea pollen were also recovered and minimally dated to 6000 cal years B.P., with a likely actual early Holocene age (27). At the Ixtacyola lake bed, chipped stone artifacts occurred in lake edge sediments to a depth of 4.3–4.4 m, 10 cm below a date of 6290 ± 40 ¹⁴C years B.P. (7280–7170 cal B.P.) (27). These tools are accompanied by a major increase in pollen disturbance indicators and charcoal in the paleoecological records. The paleoecological data suggest that maize and squash were being planted in the productive soils near lake edges that were exposed during the dry season as lake levels fell. These practices would have provided attractive yields for minimal effort.

Our data from the Central Balsas provide a firm cultural and chronological context for early domesticated maize and a species of squash in their postulated Mexican homelands. They also add to the increasing archaeological and molecular evidence attesting to a primary role of seasonal tropical forests from Mexico to Brazil in the origins and dispersals of New World domesticated crops (e.g., 9, 15, 23, 26, 32).

Four contiguous 1-m² units were opened at one edge of the Xihuatoxtla Shelter. The excavations were largely carried out with hand tools in 5-cm levels and the deposits screened through 1/4- and 1/16-inch screens. Major artifacts were measured in situ, and those that were potentially used for plant processing were immediately sealed in plastic bags to avoid any possibility of contamination of plant residues. Sediment samples from the immediate vicinity of potential plant processing tools were taken and cataloged along with the associated tool. A 20 × 20-cm-column sample was removed from the north wall of unit 1 to a depth of 1 m for the purpose of extracting plant microfossils—phytoliths, pollen, and starch grains—if present.

We thank 3 anonymous reviewers for their helpful comments on the manuscript. We also thank the Instituto Nacional de Antropología e Historia for permission to carry out the research and Froylan Cuenca, the Xihuatoxtla Shelter landowner, for his support. This research was funded by the National Science Foundation (Grant BSC 0514116), the National Geographic Society (Grant 7748–04), the Wenner-Gren Foundation (Grant 7299), the Smithsonian National Museum of Natural History, Smithsonian Tropical Research Institute, and the College of Liberal Arts, Temple University.