Mapping the chronology of an ancient cosmovision: 4000 years of continuity in Pecos River style mural painting and symbolism

Abstract

Forager societies in southwest Texas and northern Mexico painted polychromatic Pecos River style murals in limestone rock shelters containing well-preserved archaeological assemblages. To establish the temporal context of the murals, we obtained 57 direct radiocarbon dates and 25 indirect oxalate dates for pictographs across 12 sites using plasma oxidation and accelerator mass spectrometry. Bayesian modeling estimates that Pecos River style began between 5760 and 5385 calibrated years before the present (cal B.P.) and probably ended in 1370 to 1035 cal B.P. Painting spanned a duration of 4095 to 4780 years (68.3%). Stratigraphic and iconographic analyses revealed that eight of the murals were created as compositions adhering to a set of rules and an established iconographic vocabulary. Results suggest consistent messaging throughout a period marked by changes in material culture, land use, and climate. We propose that Pecos River style paintings, embedded in a cultural keystone landscape, faithfully transmitted a sophisticated metaphysics that later informed the beliefs and symbolic expression of Mesoamerican agriculturalists.

Radiocarbon dating of rock art provides evidence for a pan–New World metaphysics and iconographic system at 6000 years.

INTRODUCTION

Rock art is an informative and nuanced class of material culture, yet it is often overlooked by archaeologists because it is difficult to secure in time. Pictographs (paintings) and petroglyphs (engravings) on geological surfaces provide insights into the cosmologies, rituals, sociopolitical organization, and economies of cultures worldwide. With advances in pictograph dating (1–7) and interpretation (8–11), this imagery can now be incorporated alongside other avenues of archaeological knowledge to achieve a deeper understanding of the human experience.

Pecos River style (PRS) pictographs of southwest Texas and northern Mexico are an intricately detailed form of parietal art containing an assemblage of iconographic elements that constitute a sophisticated graphic vocabulary (Fig. 1). The complex, often monumental murals are densely concentrated into an ~8000 km² area situated between the desert-adapted cultures of Mexico to the south and the Plains cultures to the north. Known as the Lower Pecos Canyonlands (LPC), the region contains evidence of more than 12,500 years of hunter-gatherer occupation (12). The association of information-laden murals with an exceptionally well-preserved record of forager lifeways provides an unprecedented opportunity to answer questions about human cognition, social complexity, artistic development, belief systems, information exchange, and landscape-marking. Establishing the temporal context of PRS is a prerequisite for leveraging the full interpretive potential of this sophisticated iconographic system. In this study, we developed a chronological model to determine the start-stop date for PRS painting within the broader context of material culture.

Fig. 1. PRS pictographs.

(A) Zoomorphic, enigmatic, and anthropomorphic figures at 41VV1230. (B) Distribution of PRS pictographs noted in yellow shading. Study area (red box) and location of Hall’s Cave in the paleoclimate discussion (blue dot). (C) Dated pictograph sites (red triangles); other archaeological sites mentioned in the text (blue dots). Map services and data available from the US Geological Survey, National Geospatial Program.

We obtained 57 radiocarbon dates for 53 figures from 12 sites, making PRS the most securely dated rock art province in the Americas. Our research design began with an iconographic analysis to identify diagnostic PRS motifs and their geographic distribution across the region. Detailed stratigraphic studies using digital microscopy and Harris matrices were used to establish the painting sequence of murals and identify paint sampling locations. We used plasma oxidation to extract organic carbon from paint samples for accelerator mass spectrometry (AMS) radiocarbon dating (5, 13, 14). Dated oxalate minerals found in accretion layers both below and above paint layers provided a second method to corroborate direct paint dates.

We constructed a chronological model that indicates that hunter-gatherer painters began creating PRS murals between 5760 and 5385 calibrated years before the present (cal B.P.) (68.3%) and continued to do so for more than 4000 years. This iconographic system persisted through changes in technology, paleoclimatic conditions, and land use. We identified complex compositions with interwoven figures and a rule-bound color application sequence during a stratigraphic study of the paintings. Radiocarbon dates within murals are statistically indistinguishable, aligning with the stratigraphic study, suggesting that each mural was painted within a relatively short period, perhaps during a single event.

Prior research has demonstrated parallels between cosmological concepts portrayed in PRS murals and the myths and cosmologies of later Mesoamerican agriculturalists (8, 15–19). These interpretive studies are contributing to ongoing discussions into the existence, distribution, and antiquity of a pan-Mesoamerican or perhaps pan–New World cosmovision (20–24). The chronological model for PRS painting sets mural production within a temporal framework, supporting the hypothesis that it is an early graphic representation of an entrenched and pervasive foundation of beliefs and practices that subsequently informed the ideological universe of agricultural societies in Mexico. Continuity in style, motifs, color, the mural-making process, and the ontological character of images ensured fidelity in the transmission of this sophisticated metaphysics. The LPC with its geomorphologically unique and densely painted landscape was a place of high cultural salience—a cultural keystone place. Placing the rich symbolic content of PRS art into a temporal context contributes to global studies of hunter-gatherer complexity, the origins and tenacity of myth, and the transmission of knowledge through rock art.

Lower Pecos Canyonlands

The LPC is a Cretaceous limestone tableland cut by three rivers and hundreds of narrow canyons (25). Centered at 29°46′40″N and 101°14′10″W, between 350- and 400-m elevation, the 12 sites in this study were painted in the tributary canyons of the Pecos River, the Devils River, and the Rio Grande, separated by a distance of 70 km east-west and 12 km north-south.

The LPC forms the southwestern edge of the Texas Edwards Plateau. It is a transition zone where three ecological regions meet: the Edwards Plateau to the north and east, the Chihuahuan Desert to the west, and the Tamaulipan thorn scrub to the southeast (Fig. 2). Shallow alkaline soils support semiarid savannah vegetation dominated by short grasses and microphyllous shrubs in the uplands, small trees in sheltered canyons, and desert-adapted agave (Agave lechuguilla), sotol (Dasylirion texanum), yucca (Yucca spp.), and prickly pear (Opuntia spp.) in open areas. The climate is semiarid with hot summers averaging 29.7°C and mild dry winters averaging 10.7°C. The average annual precipitation is 469 mm, typically occurring in bimodal spring-summer peaks but with high interannual variability (26).

Fig. 2. Lower Pecos Canyonlands.

(A) Devils River. (B) Field of D. texanum, an evergreen desert rosette that was an important resource for food and basketry. (C) 41VV40 (left) and 41VV39 (right) in a tributary canyon of the Devils River. Pictographs at 41VV39 were radiocarbon dated to 1510 to 1380 cal B.P. (D) Located in Eagle Nest Canyon, 41VV167 excavation conducted by the Ancient Southwest Texas Project, Texas State University. Note units up to 4-m-deep containing thick strata of thermally altered rock and fiber. Pictographs at 41VV167 were radiocarbon dated to 3560 to 3455 cal B.P.

Hundreds of rock shelters contain an unusually complete record of hunting, gathering, and fishing lifeways. Evidence for hunting large game has been recovered from deposits dating to ~12,500 to 10,500 cal B.P., but these lifeways ended by ~10,000 cal B.P. with a marked shift to the use of earth ovens for pit baking of plant food (12, 27). For the Holocene, diet consisted of small game, fish, and plant foods including nuts and fruits from small trees, prickly pear fruit and pads (nopales), and agave and sotol baked in pit ovens (28, 29). Processing agave, yucca, and sotol also provided raw materials for a fiber industry used in food gathering, as well as in ceremonies and production of monumental parietal art (30). Basketry technology suggests connections extending into southern Coahuila, Mexico (31). Portable art recovered from deposits includes elaborately painted pebbles and small clay figurines (Fig. 3) (32, 33).

Fig. 3. Examples of material culture from rock shelter contexts.

Lower Pecos rock shelters preserve many objects that may have been used in a ritual context. (A) Cloud blower or pipe (41VV188). (B) Cane flute (41VV87). (C) Small lightweight basket made with a hoop and fiber netting (41VV162). (D) Rabbit jawbone/scarifier (41VV87). (E) Close-twined carrying basket fragment (41VV113). (F) Two examples of portable art: clay figurine (41VV82) and painted pebble (41VV160). PHOTO CREDITS: AMISTAD NATIONAL RECREATION AREA [(A) AMIS#14716, (B) AMIS#2794, (C) AMIS#29450, (D) AMIS#2836, and (F) AMIS#27620]; WITTE MUSEUM, SAN ANTONIO, TEXAS, wittemuseum.org (E).

PRS murals cover the walls and ceilings of at least 150 rock shelters north of the Rio Grande, with likely as many in the less-studied areas in adjacent Coahuila, Mexico (34). They are technically produced and ambitious in scope. The larger panels can measure 150 m long and 15 m high, requiring ladders or scaffolding to produce. Previous stratigraphic analyses using digital microscopy, coupled with iconographic studies, suggest that some of the murals are planned, narrative compositions, rather than random collections of images painted across time (8, 18, 35). They include anthropomorphic, zoomorphic, and enigmatic figures, as well as elaborate conflations of all three (Fig. 4). Anthropomorphs range in size from 10 cm to 8 m tall and are portrayed with an assemblage of accoutrements and paraphernalia. Deer, felines, birds, and snakes are among the most common animals represented in the paintings, and to a lesser degree, there are images resembling insects, such as caterpillars, butterflies, bees, and dragonflies.

Fig. 4. Pecos River style attributes.

(A) Anthropomorph from 41VV83 with many of the attributes characteristic of PRS. (a) Anthropomorphs with U-shaped heads are common; however, they are also portrayed with headdresses resembling rabbit ears, antlers, and feathers. (b) Atlatls, usually loaded with a dart, and (c) wrist adornments are associated with the figure’s dominant hand, whereas (d) elbow adornments are typically portrayed on the nondominant arm. (e) Darts or spears, (f) staffs, and (g) power bundles are also associated with the nondominant arm. (h) Waist tassels and (i) hip clusters are among other diagnostic PRS attributes. (B) This digital illustration represents a 39-m section of the 67-m mural at 41VV83 and reveals the complexity of PRS paintings. (C) Orthophotograph generated from an SfM (structure from motion) three-dimensional model of 41VV83. ILLUSTRATION CREDIT: C.E.B.

The PRS palette included earth colors in shades of red, orange, yellow, black, and white. Choosing from among this range of colors, muralists composed symbols that serve as divine beings, things, places, sensations, and experiences. Human and animal forms provided a frame upon which artists added semantically charged visual attributes, such as headdresses of varying types; adornments attached at the wrist, elbow, waist, or hip; and paraphernalia, including atlatls (spear-throwers), darts, rabbit sticks, staffs, and power bundles (Fig. 4). These attributes functioned as a graphic vocabulary, and their arrangement in the mural conveyed a meaning (18). PRS painters constructed messages with this repertoire of colors and motifs to relate myths, prescribe rituals, and provide social cohesion (8, 17, 36, 37).

Characterization of paint ingredients

PRS pigments have been the subject of numerous studies. X-ray diffraction at six rock art sites in the region identified inorganic minerals as the color-producing ingredient in PRS paint: red, orange, and yellow (hematite, maghemite, goethite, lepidocrocite, magnetite, and ferrihydrite) and black (manganite and pyrolusite) (38). White pigment is also present but has not been characterized. Laser ablation inductively coupled plasma mass spectrometry on paint samples from four rock art sites, including one in this study (41VV576), suggested that the most likely source for PRS red and yellow pigments was limonite siltstones found in the bottom of canyons (39). Elemental analyses using portable x-ray fluorescence conducted at 13 rock art sites across the region, including two of our study sites (41VV167 and 41VV1230), confirmed that red and yellow PRS paintings were made with iron minerals and black PRS paintings were painted with manganese mineral pigments (40). Fourier transform infrared spectroscopy identified oxalate mineral accretions associated with PRS pictographs at 20 sites in the region, including three in this study (41VV76, 41VV167, and 41VV576) (41). Oxalate accretions are ubiquitous on the limestone rock shelter walls in the region.

PRS paint samples have also been the subject of numerous organic analyses to identify binders, emulsifiers, and/or vehicles. Early ancient DNA research identified ungulate (deer or bison) as the binder in PRS paint samples (42); however, when an attempt was made to replicate these studies with more specific primers to identify species, there was no amplification of the DNA (43). Gas chromatography–mass spectrometry analysis of untreated paint samples attempted to identify fatty acids as a possible binder, but the results were inconclusive (44). Raman spectroscopy detected CH-stretching bands associated with organic compounds in a sample of black paint from 41VV576. Another sample of red paint from the same site showed NH, CHN, ─C═C─C═C─ conjugation, and an aromatic quinonoid functional group (45).

The presence of organic ingredients in the paint has been established. Paint samples treated with base washes before plasma oxidation have measurable levels of organic carbon for radiocarbon dating, whereas control samples of unpainted rock analyzed with the same procedures contain negligible organic carbon (table S1). Thus, dateable carbon is inherent in the paint and not in the unpainted rock. Stable carbon isotope values for this organic material extracted from PRS paint samples range from −20 to −26‰ (14). Ethnographic texts and experimental archaeology provide insights into possible paint ingredients (46–48). North American Indigenous groups used fat in deer bone marrow as a binder to adhere mineral pigment particles together and saponin-rich plants, such as the C3 Yucca constricta, as a vehicle or emulsifying agent. Although not conclusive, stable carbon isotope values are consistent with a combination of bone marrow and yucca.

RESULTS

Iconographic and paint stratigraphy analyses

We identified 134 sites containing one or more recurring motifs unique to PRS, including rabbit-eared headdresses, antlers tipped with dots, single-pole ladders, crenellated arches with portals, impaled dots, stylized dart tips, winged anthropomorphs with antlers, speech breath, and power bundles, a prevalent PRS motif. More than 60% of the sites incorporate anthropomorphs with power bundles extending from their nondominant arm (defined as the arm not wielding an atlatl). The motif includes a plant-, animal-, or human-like shape located at the distal end of long parallel lines that are connected perpendicularly to a staff or dart. While there is variability in the form of the distal end, it is most frequently an ovoid shape resembling a spiny seed pod (Figs. 5 and 6).

Fig. 5. Power-bundle motif at 41VV584 and 41VV167.

(A) At 41VV584, an anthropomorph (A9) with a power bundle. This is the oldest power-bundle image directly dated in this study with a radiocarbon date of 4470 ± 100 radiocarbon years before the present (RCYBP) (CAMS ID 188179, sample no. 3, 5730 to 5300 cal B.P., 95.4% probability, table S1). The weighted average of the stratigraphically related pictographs at this site is a better measure of the age of this motif than a single measurement. The chronological model estimates painting at 41VV584 to 5460 to 5320 cal B.P. (68.3% probability). (B) At 41VV167, a much larger anthropomorph (A1) with a headdress resembling rabbit ears and dots emanating from the mouth denoting speech breath is also depicted with a power bundle. For A1, we obtained a direct radiocarbon date of 3380 ± 110 RCYBP (CAMS ID 191210, sample no. 4, 3900 to 3300 cal B.P., 95.4% probability, table S1). The chronological model estimates painting at 41VV167 to 3560 to 3455 cal B.P. (68.3% probability). ILLUSTRATION CREDITS: C.E.B.

Fig. 6. Power-bundle motif at 41VV286 and 41VV1230.

(A) At 41VV286, a winged anthropomorph (A4) with a power bundle was radiocarbon dated as 2960 ± 100 RCYBP (CAMS ID 191217, sample no. 1, 3370 to 2860 cal B.P., 95.4% probability, table S1). The chronological model estimates painting at 41VV286 to 2995 to 2800 cal B.P. (68.3% probability). (B) At 41VV1230, an anthropomorph with red antler tines tipped with black dots (A5) was radiocarbon dated as 2050 ± 45 RCYBP (CAMS ID 190057, sample no. 4, 2125 to 1870 cal B.P., 95.4% probability, table S1). The chronological model estimates painting at 41VV1230 to 1990 to 1925 cal B.P. (68.3% probability). This is one of the youngest dated power bundles in the study. ILLUSTRATION CREDIT: C.E.B.

We selected eight PRS sites containing multiple intersecting figures for stratigraphic analysis. Each mural incorporates anthropomorphs wielding power bundles and at least two additional diagnostic PRS motifs and is distributed geographically across the region. We also radiocarbon dated four additional PRS sites (three have power bundles) to construct a chronological model that incorporates data from all 12 sites. A stratigraphic analysis of intersecting paint layers at the eight sites established the murals’ painting sequence and informed our selection of paint sample locations for radiocarbon dating.

Digital microscopy of intersecting paint layers and construction of Harris matrices established the stratigraphic sequence and the relationship between and among figures within the rock art panels selected for analysis. We analyzed 2206 photomicrographs at 588 points of intersecting paint involving 256 figures across the eight sites (table S2). Some photomicrographs produced inconclusive results as a result of poor image quality or heavy accretions obscuring the paint layers; however, the effort resulted in 535 successful determinations, and only 53 analyses were indeterminate. The results of this stratigraphic study agree with Boyd and Cox’s previous analysis of the 41VV124 mural (8). Painters at each site consistently followed the same paint color application sequence, moving from darker to lighter hues (Fig. 7A). For example, red superimposes black, yellow superimposes both red and black, and white superimposes all colors. Only 10 of the 535 analysis locations deviated from this paint color application sequence. Because painters held to this organized layering of colors, many figures are sandwiched together, that is, a paint layer of one figure is painted both over and under another figure (Fig. 7). In this way, polychromatic figures across the murals were woven together to form a complex tapestry of intertwined images, evidence that supports a single painting event.

Fig. 7. Analyzing mural stratigraphy.

(A) Paint sequence determined from 535 stratigraphic analyses across eight sites. (B) Photomicrographs at 50×. (C) Digital illustration of a 7-m section of the 62-m mural at 41VV576. See also fig. S3. To describe the paint stratigraphy, we focus on the area marked by the green box. (D) Photograph of the area outlined in green. (E) Digital illustration demonstrating the painting sequence from darker to lighter hues. The black paint of A14 and E1 was applied first, followed by the red paint of A12 and E1 and, lastly, the yellow paint of A14. We identified this same painting sequence across 70 analysis locations involving 23 figures in this section of the mural. See table S1 for these data. (F) Harris matrix diagram showing the stratigraphic relationship among three figures (A12, A14, and E1). Paint layers in a Harris matrix are ordered from top (T) to bottom (G). The arrows point from layers above to layers below. For example, Yellow_A14 is over Red_E1, and Red_E1 is over Black_A14. The two figures are interwoven, and they are both over and under each other. See fig. S4 for the complete Harris matrix for this section of Jackrabbit Shelter. ILLUSTRATION CREDIT: C.E.B.

Radiocarbon dating

We obtained 57 direct radiocarbon dates for pictographs and 25 oxalate dates at 12 sites across the region. Pictograph radiocarbon results are summarized in Fig. 8. All 57 radiocarbon measurements are listed in table S1. For paint samples, stable carbon isotope values (δ¹³C) were assumed to be −25‰, as extracted amounts of carbon were too small to split for isotope ratio mass spectrometry. For oxalate samples, δ¹³C values were assumed to be −11‰, the average value measured by isotope ratio mass spectrometry for LPC calcium oxalate samples (41). Calibration and chronological modeling were accomplished using OxCal computer software version 4.4.4 with IntCal20 curve data (49, 50). For all 52 control samples, organic carbon levels were negligible (≤1%) (table S1). Thus, there is negligible physical or chemical contamination in the unpainted limestone substrate adjacent to paint sample locations.

Fig. 8. Direct pictograph dates and modeled age estimates.

(A) C-14 date (RCYBP) weighted averages are reported for the eight sites with replicate measurements. Four of the sites have only one radiocarbon measurement. Posterior density estimates are listed from model 1. (B) The brackets along the left-hand side and the OxCal commands define the model exactly. (C) PRS pictographs from 41VV1230.

Oxalate accretion dates

Obtaining dates on oxalate mineral accretions lying above and below dated paint layers provided minimum/maximum dates, giving further confidence in the accuracy of age results presented here. In all instances, underlying oxalate layers were older and overlying oxalate accretions were younger than direct paint dates (table S1). This correctly ordered microstratigraphy of the accretion and paint layers supports the validity of both dating methods. Figure 9 shows the stratigraphic sequence of accretion and paint dates from 41VV576. A radiocarbon date for an oxalate sample is a weighted average of the deposited layers’ ages. If layers are thicker in one place than another, it results in disparate ages. Overlying oxalate accretions provide minimum ages for rock paintings as any mixture of the outer layers is still younger than any underlying paint. Underlying oxalate layers provide maximum ages for rock paintings as the mixture of the underlying layers is still older than any overlying paint. Reported oxalate dates do not reflect a single event of formation but instead represent the weighted average age of the oxalate crust material.

Fig. 9. Direct and indirect dating of pictographs at 41VV576.

(A) Cross section (0.2-cm width) of black Mn-pigmented paint on the limestone substrate. Mineral accretions are observed both overlying and underlying the black paint layer. (B) Cream-colored mineral accretion above and below a red ocher–pigmented paint layer (0.2-cm width). X-ray diffraction and infrared spectroscopy confirmed the presence of calcite, whewellite, and gypsum. (C) We obtained five direct paint dates shown in blue. Underlying oxalate accretion layers are older (orange), and overlying oxalate accretion dates are younger (yellow). (D to G) Sampling locations for dated pictographs at 41VV576.

Pictograph dates and Bayesian chronological modeling

On the basis of mural stratigraphy results at eight sites, paint samples from individual figures are not from different painting events as previously supposed but instead are replicate samples from a single mural painting and, thus, can be averaged. Weighted averages are reported for each site where multiple paint samples were directly dated using the R_Combine function of OxCal. For the direct paint dates at each site, the chi-square test statistic was less than the critical value at the 5% significance level, demonstrating that there is no significant difference in the measurements. The observed variation is expected because of random error, suggesting that individual paintings are coeval. One date from 41VV38 was excluded as an outlier. The remaining 56 radiocarbon dates were included in all models.

On the basis of Bayesian chronological modeling, PRS painting in the region began between 5760 and 5385 cal B.P. (3815 to 3435 cal BC) and probably ended in 1370 to 1035 cal B.P. (580 to 915 cal AD) (68.3%). On the basis of a span analysis, PRS painting had a duration of 4095 to 4780 years (68.3%). Estimated ages for PRS pictographs from the chronological model are italicized in the Discussion as interpretive 68.3% posterior density estimates, with the conservative 95.4% posterior intervals also provided in Fig. 8.

DISCUSSION

Results demonstrate that PRS pictographic murals were painted as single entities distributed across the study area during the last half of the Holocene. Diagnostic motifs of the style occur throughout the time span of pictograph production, as does the sequential ordering of color layers. With this information, we examine PRS imagery within the broader context of paleoenvironment, demography, and technology and explore possible drivers for its persistence across millennia.

Paleoclimate and cultural chronology

Middle Holocene (8000 to 4000 cal B.P.)

Nomadic foragers began transforming the LPC into a painted landscape during the Middle Holocene. Between ~8000 and 6000 cal B.P., the region experienced a general drying trend, with short fluctuations in temperature and effective moisture, resulting in the loss of vegetation cover and soil erosion (51–54). After ~6000 cal B.P., stable isotope data from Hall’s Cave (41KR474) register a slight rise in mesoscale convective storm activity and a modest increase in C3 plants on the landscape. A brief spell of very dry but cooler conditions occurred around 5900 cal B.P. when bison moved into the Edwards Plateau (55). Intermittent catastrophic flooding on the Pecos River began around 5100 cal B.P. (56). The region remained generally arid, punctuated by occasional storms and massive floods, creating high interannual variability in landscape productivity.

A rise in occupation around the rivers from ~5800 to 5000 cal B.P. is indicated by a summed probability distribution (SPD) of regional radiocarbon dates from rock shelter excavations (Fig. 10 and Supplementary Text) (57). This agrees with component counts and stone tool frequencies recovered from 55 dated LPC sites (58). Turpin (37, 59, 60) proposed a scalar stress model to explain the onset of PRS production, arguing that increasing aridity and a reduction in upland resources forced population packing along the rivers. According to this hypothesis, PRS emerged as a visual form of ritual communication used to provide guidance and promote unity.

Fig. 10. SPD and changes in the projectile point technology.

(A) SPD of radiocarbon dates (blue) from LPC rock shelter excavations, with a 500-year (yr) moving average trendline in purple (57). Technological changes in projectile points shift during PRS pictograph production: (B) Bell; (C) Pandale; (D) Langtry; (E) Castroville; (F) Ensor. PHOTO CREDIT: E. PREWITT.

Our chronological model estimates the inception of PRS at 5760 to 5385 cal B.P. At roughly the same time, projectile point technology changes from relatively thin, broad points to the Pandale type, a beveled narrow/lanceolate style (59). Geographically restricted to the LPC and adjacent areas, a technologically similar precursor has not been identified (Fig. 10).

41VV584 (5460 to 5320 cal B.P.) is one of the earliest dated PRS murals in our study, located in a Pecos River tributary (fig. S1). Although portions of the mural are degraded, 47 polychromatic figures remain intact and extend 30 m along the shelter wall. At 5.7 m tall, artists would have needed scaffolding or ladders to reach the upper registers. Just upstream is 41VV576 (4825 to 4625 cal B.P.), where 267 documented PRS figures span more than 62 m, reaching a height of 3 m (figs. S2 and S3). On the basis of our iconographic and stratigraphic analysis, these early murals reveal a painting tradition with an established graphic vocabulary, a rule-bound color application sequence, and a compositional structure.

Late Holocene dry period 1 (4000 to 3200 cal B.P.)

As indicated by a hiatus in deposition at Hall’s Cave and a shift back to C4/CAM isotope ratios in leaf wax data, even drier conditions returned by ~4000 cal B.P. and prevailed until ~3200 cal B.P. (54). Major flooding continued on the Pecos River until ~3200 cal B.P. (51, 56). Thinner, contracting stem Langtry and Val Verde points, also geographically restricted, supplant the Pandale point (59). Between ~5000 and 3500 cal B.P., the SPD trendline rises slightly and then falls to lower levels (Fig. 10).

41VV1573 (3815 to 3580 cal B.P.) and 41VV167 (3560 to 3455 cal B.P.) date to this drier period. 41VV167 is an amphitheater-like rock shelter measuring 60 m wide and 25 m deep (Fig. 2). Recent excavations at this site have unearthed a stratigraphic sequence beginning with a Younger Dryas component and continuing through the Holocene. Twelve earth oven facilities, primarily used to bake agave and sotol, date from ~10,500 to 660 cal B.P. Analyses of these ovens identified a time around 4000 cal B.P. when reuse of the rock heating elements intensifies as a result of increased cooking of agave and sotol. Rapid reuse of the rock heating elements would cause large accumulations of earth oven debris over a short span of time (29). Alternatively, PRS murals may have been the focus of periodic aggregations, as first proposed by Turpin (36). Intensified use of earth oven facilities may not be connected to continuous rock shelter occupation but rather to seasonal population aggregations for ceremonial purposes and the production of rock art.

Late Holocene mesic interval (3200 to 2200 cal B.P.)

Although the interval is poorly dated, proxy climate records from the LPC and surrounding regions indicate a shift to mesic conditions during the Late Holocene (54, 61, 62). Pollen from Bonfire Shelter (41VV218) and changes in Hall’s Cave leaf wax stable isotope ratios indicate a wetter environment (51, 54). Major flooding ceases on the Pecos River as indicated by fine-grained, slack-water flood deposits between ~3200 and 2000 cal B.P. (56). Bison populations expand into the Edwards Plateau at 3160 and 2425 cal B.P. (55). Evidence for bison in the LPC is known from five sites, including an 80-cm-thick deposit of burned bison bone at Bonfire Shelter dating to 2590 cal B.P. (27).

Mesic interval deposits contain larger, wide-bladed dart points associated with bison hunting (59). These projectile points, belonging to the Castroville series (Fig. 10E), have been recovered in components throughout the LPC and as far as the eastern and northern edges of the Edwards Plateau. The presence of wide-bladed dart points and intermittent bison incursions signals changes in land use and the possible influence of bison hunting groups in the LPC, causing disruption or expansion of local and regional social networks. We recorded PRS production from two mural sites, 41VV286 (2995 to 2800 cal B.P.) and 41VV1205 (2935 to 2790 cal B.P.), with dates close to the estimated onset of mesic conditions. While these two murals contain many of the same diagnostic elements identified at earlier sites, including power bundles, single-pole ladders, stylized darts, and speech breath, here, we see the introduction of winged anthropomorphs with antlers. Both sites contain this motif, as do other PRS sites not included in this study (Fig. 6A) (63).

Late Holocene dry period 2 (2200 to 1200 cal B.P.)

The end of the short mesic period and a return to hot, generally dry conditions are marked by a hiatus in deposition at Hall’s Cave, minimal Pecos River flooding, and a drop in grass and tree pollen percentages (51, 54, 56). Turpin (36) proposed a second rise in riverine population density by ~2300 cal B.P., based on the number of components at upland archaeological sites and in projectile point frequencies. The SPD trendline rises steeply at ~2200 cal B.P. (Fig. 10). By ~2300 cal B.P., the broad-bladed projectile points associated with bison hunting are replaced by a smaller design, the Ensor-Frio series, which had a wide distribution throughout south and central Texas and northern Mexico (59).

PRS production continues with 5 of the 12 study sites in the chronological model falling into this time frame (Fig. 8). The largest of these later murals, 41VV1230, contains ~194 figures—96 anthropomorphs, 48 zoomorphs, and 50 enigmatic figures. The painting sequence, as determined by 215 microscopic observation points, was black, red, yellow, and white. This is the same sequence observed in the earliest mural at 41VV584, demonstrating a continuity of paint color layering. PRS core motifs, including power bundles, single-pole ladders, stylized dart tips, and speech breath, are also present at both 41VV1230 (1990 to 1925 cal B.P.) and 41VV584 (5460 to 5320 cal B.P.). Appearing at three sites (41VV38, 41VV76a, and 41VV1230) in the PRS repertoire of dated motifs are red antlers with black dots at the end of each tine (Fig. 6B). This is the earliest dated occurrence of the antlers with dots motif.

Lacking an accurate chronological framework and emphasizing environmental drivers, it was thought that PRS production started because of the Middle Holocene dry period and faded because of the interruption of the Late Holocene mesic interval (60). We now have evidence of PRS production continuing for at least 800 years after the mesic interval. The chronological model provides an estimate of 1370 to 1035 cal B.P. for the end of PRS production, raising the possibility that the practice of mural painting continued until the introduction of the bow and arrow.

Although we see a temporal relationship between dry conditions, a rise in the SPD trendline, and the onset of PRS production, demonstrating the cause and effect among these factors is outside the scope of this study. Population pressure and resource intensification in riverine areas may have contributed to the conditions under which seasonal ritual aggregations and PRS production began (29). However, why did PRS mural production continue for more than 4000 years amid changes in climate, demography, lithic technology, and land use? And why would nomadic hunter-gatherers invest time, resources, and labor to paint complex and often monumental murals along rock shelter and canyon walls within the LPC?

Places of power and knowledge

PRS mural production persisted in a limited and remarkably unique geographic area. We suggest that a possible explanation for its continued production is tied to the region’s distinctive landscape of deep canyons, sinkholes, caves, rock shelters, permanent springs, and winding rivers. In Indigenous thought, these geomorphological features are ancestral places of emergence and are imbued with power and agency (64). They are experienced as places filled with potency, portals to ancestral knowledge, and locations to conduct ritual activity (65). Once the efficacy of the place has been established, people return to acquire more knowledge and interact with the forces dwelling there. With each subsequent human and supernatural interaction, more power accrues (66).

Rock art is often concentrated in these sacred places (65). Painted onto or engraved into the living landscape, images are considered embodiments of supernatural entities, sources of power, and a vehicle to transmit knowledge (35, 65, 67–69). The creation of PRS murals by ~175 generations of artists suggests that the LPC was perceived as a place of supernatural power and high cultural salience, fitting the definition of a cultural keystone place (70). From an Indigenous perspective even today, the paintings are not simply passive props but living entities directly communicating sacred ancestral knowledge (18, 35). Continuity in style, motifs, and process across four millennia implies high fidelity in the transference of that knowledge (71).

PRS painters abided by the same rules in color application order, selected motifs from within an established inventory, and created complex compositions of interwoven figures requiring a substantial investment in time, materials, and labor. While new motifs enter the repertoire of images and experimentation with existing motifs is evidenced, core pictographic elements diagnostic of the style persist in form and context. We propose that the constancy in PRS mural production, location, and content communicated a deeply ingrained Archaic core cosmovision that eventually was manifested in the symbolism and belief systems of later Mesoamerican agriculturalists.

Visible remains of an Archaic cosmovision?

For decades, scholars have observed an ancient cultural unity in the ideological, philosophical, and religious beliefs and rituals of Mesoamerican peoples, including creation stories, cyclical time, quadripartite cosmovision, and complex calendrical systems (20–23). López Austin (21) proposed that these shared concepts, which he termed “el núcleo duro” or the hard core, were highly resistant to change and virtually unchangeable. Rice (24) has suggested that this deeply rooted and widely shared substratum of beliefs and practices dates back several thousand years and was developed before people entered the region we now refer to as Mesoamerica, suggesting pan–New World origins. The paucity of surviving material evidence from the Archaic period has limited the evaluation of these ideas (72). As a result, little research has been done on the consequences of such profound antiquity for Mesoamerican ideologies (24). PRS pictographs have provided an opportunity to explore the timing and graphic expression of this ancient cosmovision.

Researchers have noted parallels between metaphysical concepts portrayed in PRS murals and the myths and cosmologies of later Mesoamerican agriculturalists hundreds of kilometers to the south. In 1974, before advances in radiocarbon dating pictographs, Kelley (15) recognized similarities between PRS and Mesoamerican ceremonial art. Thinking PRS to be much younger than we know it to be today, he attributed the parallels to diffusion and suggested that PRS painters were attempting to mimic the ceremonial art more fully developed to the south. More recently, Tate (19) proposed that the Aztec sign for Chicomoztoc, their primordial place of origin, was influenced by a toponymic sign, a place name for a geographical or cosmological feature, created by PRS artists to the north. She suggests that the PRS toponym, a distinctive crenellated arch and portal motif, signifies the generative undulating canyons, caves, and waterways of the Lower Pecos region from which all life emerged. Analyses conducted by Boyd and her collaborators (8, 16–18, 35) have identified patterns in PRS pictographs that equate in detail to creation myths and cosmological constructs of the 14th to 16th century Aztec (Nahua) and the present-day Huichol (Wixárika). On the basis of their analysis, they maintain that PRS murals are graphic manifestations of metaphysical concepts traditionally associated with complex agricultural societies in Mesoamerica, evidence of López Austin’s el núcleo duro and suggestive of a pan–New World cosmovision. Whether because of preservation bias or the ideology’s initial codification, Boyd and Cox (8) proposed that PRS murals are the oldest surviving visual manifestation of this ancient and enduring cosmovision, spanning not only time but linguistic and geographic boundaries. Our chronological model places PRS murals within a temporal framework that supports this hypothesis.

With more than 134 documented PRS sites, it is unlikely that the 12 dated murals represent the first or last painting event. During this project, as additional sites were radiocarbon dated, we added to our data and further refined our chronological model to obtain a “believable story” by combining radiocarbon dates with prior archaeological information such as stratigraphy (73). Interpretive age estimates will change as further information becomes available. Advances in the identification of organic paint ingredients will bolster radiocarbon results, and additional stratigraphic studies will expand our understanding of compositional murals in the region.

Our data demonstrate that by 5760 to 5385 years ago, PRS murals are embedded in the Lower Pecos landscape as a sophisticated and well-established iconographic system. It was not only an art tradition that persisted for more than 4000 years but also a belief system and worldview that informed the content and technical execution of the murals. These results reveal a rule-based iconographic system guided by a metaphysical viewpoint that has endured over millennia.

MATERIALS AND METHODS

Experimental design

We developed a research protocol to ensure accurate and reliable results. Building a chronological model for PRS began with an analysis of pictograph panel content and structure. We considered pictograph sites on the basis of the presence of recurring symbols and diagnostic PRS motifs distributed across the region. We included 12 PRS murals in our study ranging in size and complexity.

Because of the complexity and density of overlapping polychromatic images, digital microscopy was used to record stratigraphic relationships between PRS figures (fig. S2). Stratigraphic determinations were entered into Harris Matrix Composer software, a program designed to organize complex stratigraphic sequences for archaeological excavations into a diagram representing relative time (74). We adapted the software to map PRS mural stratigraphy, the proxy for a mural’s painting sequence (8). This guided paint sample selection for radiocarbon dating.

To overcome challenges associated with dating pictographs, we used two independent methods to obtain ages for PRS paintings. The first method dates the organic carbon in paint binders/vehicles/emulsifiers to provide a direct age for paintings. The second method dates carbon in calcium oxalate mineral accretions to provide minimum and maximum ages for the paintings.

Iconographic analysis

Before the start of the radiocarbon dating campaign, personnel from Shumla Archaeological Research & Education Center (Shumla) completed baseline documentation of 236 rock art sites in the region. Baseline documentation included high-resolution photography (three-dimensional photogrammetry models and gigapixel panoramas or GigaPans) of rock art panels, as well as an iconographic inventory of motifs and attributes diagnostic of regional rock art styles. The resultant iconographic data were entered into a searchable, region-wide rock art database (75). From 134 sites with PRS iconography, we selected eight study sites for stratigraphic analysis on the basis of geographic distribution and the presence of diagnostic attributes unique to PRS anthropomorphs.

Paint stratigraphy analysis

We used true-color and DStretch–color-enhanced (76) GigaPans as maps of pictograph panels to identify locations of intersecting paint layers within, between, and among figures (fig. S2). DStretch is a plug-in to ImageJ, a free, open-source image processing program, that creates false color enhancements to make obscured or faint images more visible. With DStretch-enhanced GigaPans serving as base layers in Adobe Photoshop software, we inserted a layer containing the figure identification codes and a layer marking potential locations for digital field microscopy. These intersecting locations were circled and labeled to denote the figure and colors involved (e.g., Red_E33, Yellow_A1, and Black_A1) (fig. S2A).

The panel map, printed from the annotated GigaPan image as a large format scroll, served as a reference in the field to guide and track digital microscopy (fig. S2B). Using a handheld DinoLite Edge digital microscope, we captured photomicrographs of the intersecting paint layers at 50× to 200× (Fig. 7). We photographically documented each analysis location, including context photos of the overall figure and macrophotographs indicating the exact placement of the microscope.

In the lab, we examined the digital photomicrographs and macrophotographs of intersecting paint layers to determine the stratigraphic sequence or over/under relationships between the paint layers (e.g., Red_A023 over Black_E002). To avoid bias, at least two researchers made independent determinations for each analysis location. When the determinations did not match, a third person made another independent observation as a tiebreaker. Inconclusive results were reported as indeterminate.

We entered the results of the stratigraphic analysis into Harris Matrix Composer version 2.0.1 software. The top unit (T) in the matrix represents the interface of the topmost layer of paint with the atmosphere. The ground unit (G) represents the interface of the bottommost layer of paint with the shelter wall. Each arrow in the matrix diagram represents the stratigraphic relationship between paint layers and these two units. After inserting color units for a figure, we began adding units for figures it either superimposes or by which it is superimposed (fig. S4). We used the matrix to reveal interwoven paint layers, where one figure is both over and under another figure by virtue of the painting sequence.

Radiocarbon dating

Early experimental efforts to date PRS pictographs demonstrated that it was possible to extract and radiocarbon date organic material in inorganic-pigmented pictographs. In the 1990s, Rowe (14) obtained 29 radiocarbon dates for organic binders/vehicles/emulsifiers from 16 inorganic-pigmented PRS paintings. Steelman et al. (14) evaluated legacy PRS dates on the basis of contextual, compositional, and measurement elements, concluding that these experimental results were problematic and should not be used to draw archaeological conclusions. (i) No backgrounds or control samples of unpainted rock were analyzed. (ii) No chemical pretreatment was used to remove any potential contamination before plasma oxidation. (iii) Iconographic information was not collected from sampling locations. Over the past 35 years, improvements in experimental procedures (outlined below) and refinement of the method have allowed the application of the method to answer global archaeological questions.

Traditional radiocarbon procedures using acid-base-acid washes and combustion are often not able to collect measurable carbon from miniscule amounts of organic binders/emulsifiers/vehicles embedded in a primarily inorganic matrix (Fig. 9, A and B). In contrast, the plasma oxidation technique oxidized organic carbon under mild conditions below the decomposition temperature of carbon-containing minerals such as limestone (CaCO₃) and whewellite (CaC₂O₄·H₂O) so that acid pretreatment was not necessary before plasma oxidation. This reduced the amount of organic material lost during chemical pretreatment and allowed the AMS radiocarbon dating of inorganic-pigmented pictograph samples.

Oxalate accretions cover most of the limestone substrate on which the paint was applied and have subsequently formed over the murals after they were completed. Oxalate carbon in mineral accretions can be used to estimate indirect ages for pictographs (5, 77–80). The oxalate dates provided a second method to check the accuracy of direct paint dates.

Sample collection

Paint samples, ranging in size from 1 to 2 cm², contained two potentially datable components: organic binders/emulsifiers used by the artists to make the paint and oxalate minerals encasing the paint layers. Paint samples were collected high on the wall, inaccessible to sheep and goats (and humans) that might rub against the paintings. We used individual sterile scalpel blades to collect paint samples onto prebaked (550°C) aluminum foil, which were folded and placed in labeled plastic bags. We also collected control samples (backgrounds) of unpainted rock adjacent to paintings to investigate levels of organic contamination in the rock substrate. In the laboratory, we separated the outer mineral accretion layer, paint layer, and underlying accretion layer into three subsamples by microexcavating with a scalpel blade under a stereoscope at a 10× magnification. Any visible physical contaminants (rootlets, plant fibers, spider webs, etc.) were removed with microtweezers.

Chemical pretreatment

For paint and control samples, we washed the powdered solid in 5 ml of 1 M sodium hydroxide solution in centrifuge tubes placed in an ultrasonic water bath at 50°C for 1 hour. Samples were rinsed with Optima-grade water and filtered onto sterilized quartz-fiber filters (prebaked at 550°C). Solid samples were then dried at 110°C.

For overlying and underlying oxalate accretion layers, we conducted four sequential weak acid washes (5 ml of 1 M phosphoric acid) for 1 hour each to remove calcite. Samples were rinsed with Optima-grade water, filtered onto sterilized quartz-fiber filters, and dried at 110°C. We analyzed the accretion layers, both before and after acid treatment, with x-ray diffraction and/or Fourier transform infrared spectroscopy to confirm the complete removal of carbonate minerals and the continuing presence of oxalate minerals.

Plasma oxidation

We used a radio frequency capacitively coupled plasma to oxidize organic material in samples to carbon dioxide. For paint samples, this carbon dioxide was reduced to graphite for carbon isotope measurement. For oxalate samples, we used the plasma oxidation technique to completely remove any organic contamination before combustion of the oxalate mineral, graphitization, and AMS radiocarbon dating. Isotope measurement was conducted at the Center for Accelerator Mass Spectrometry (CAMS), Lawrence Livermore National Laboratory.

Controls

To check for background carbon contamination in the limestone substrate, we analyzed control samples collected from adjacent unpainted rock surfaces using the same laboratory process (base pretreatment and plasma oxidation) as for paint samples. If we had found notable levels of contamination in the background samples, we would not have been able to report radiocarbon dates for the paintings as there would be no way to distinguish or separate the organic carbon that was associated with the painting event and the contamination. Unfortunately, this type of control sample check is not done in many rock art dating studies.

Statistical analysis

Bayesian chronological modeling

We constructed a chronological model to query the start, end, and span of pictograph production. OxCal, a radiocarbon calibration and Bayesian chronological modeling software program, used Markov chain and Monte Carlo sampling methods to generate a representative sample of solutions and a range of probability outcomes (49). These models combine radiocarbon dates (likelihoods) and archaeological evidence (prior beliefs) to produce age ranges (posterior density estimates) (73). We constructed three different models as a form of sensitivity analysis to see how archaeological inferences can affect model results (see the Supplementary Materials). Model 1 incorporated archaeological information (prior beliefs) from paint stratigraphy by calculating a single weighted average (R_Combine) for each site (Fig. 8). Model 2 placed direct paint dates from each site into different phases. Model 3 was the simplest model and placed all PRS dates into a single phase. The convergence (C) values for the elements in each model were all >95, illustrating that the models are robust. The models showed good agreement (A_model) between the chronometric dates and model assumptions. All three models produced similar posterior density estimates. Model 1 (A_model = 99.4) was preferred because it incorporated robust contextual information from the field work. Modeled age estimates are italicized in the text.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S4

Tables S1 and S2

OxCal Code

References