Strategies for 14C Dating the Oxtotitlán Cave Paintings, Guerrero, Mexico

ABSTRACT Oxtotitlán Cave paintings have been considered among the earliest in Mesoamerica on stylistic grounds, but confirmation of this hypothesis through absolute dating has not been attempted until now. We describe the application of advanced radiocarbon strategies developed for situations such as caves with high carbon backgrounds. Using a low-temperature plasma oxidation system, we dated both the ancient paint and the biogenic rock coatings that cover the paint layers at Oxtotitlán. Our research has significantly expanded the time frame for the production of polychrome rock paintings encompassing the Early Formative and Late Formative/Early Classic periods, statistically spanning a long era from before ca. 1500 cal B.C. to cal A.D. 600. Los murales de la Cueva de Oxtotitlán, de acuerdo con criterios estilísticos, han sido considerados entre los más tempranos de Mesoamérica. Sin embargo, hasta la fecha esta hipótesis no había sido corroborada mediante fechamiento absoluto. En este trabajo se describe la aplicación de técnicas de radiocarbono avanzadas, las cuales han sido desarrolladas para lugares como cuevas con un elevado fondo de carbón. Fechamos tanto la pintura antigua como los recubrimientos biogénicos que cubren las capas de pintura utilizando un sistema de oxidación de plasma a temperatura baja. Nuestras investigaciones han ampliado de manera significativa el intervalo temporal de la pintura mural policroma en Mesoamérica, abarcando los periodos del Formativo Temprano al Formativo Tardío/Clásico Temprano, desde antes de aproximadamente 1500 cal a.C. hasta 600 cal d.C.

FIGURE 1. Map of the location of the Oxtotitlán cave paintings, Guerrero, Mexico, and other significant Formative period archaeological sites. Oxtotitlán is part of the larger site complex of Quiotepec-Oxtotitlán.
C-1, Figure 2) captured everyone's attention. With a dramatic placement high on the cliff face above the south arm of the cave, the painting has stylistic and iconographic similarities with monuments from major urban Olmec centers on the Gulf Coast of Mexico. The Gulf Coast Mexico Olmec built some of Mesoamerica's earliest cities, the dynastic seats of Mesoamerica's first kings (Clark and Pye 2006;Coe 1995;Coe and Diehl 1980;Cyphers 1997Cyphers , 2004aCyphers , 2004bDrucker 1952;Drucker et al. 1959;González Lauck 1996, 1997. Gulf Olmec cities are renowned for their synthesis of monumental sculpture and urban design. The Oxtotitlán painting presented a new dimension both in media and in geographical provenience, inviting questions about the artists of Oxtotitlán and their relationship with the Gulf Coast. One of the largest but most enigmatic polychrome paintings covers a rock face (Panel C-2, Figure 3) at ground level in front of the south rockshelter entrance, directly below the enthroned figure (C-1). Although a considerable amount of paint loss from natural exfoliation of the rock surface has made it difficult to ascertain what the painting represents, areas of red, green, black, and yellow paint are still visible. Grove's (1970) earlier research reported black jaguar spots on the lower part of the rock face, possibly referencing deep-rooted jaguar-focused rain ceremonialism. Though difficult to discern with the naked eye, these spots are visible in computationally enhanced images of the painting.
Other polychrome panels painted across the front of the cave have been neglected in comparison because they are less ambitious and less visually accessible. Nevertheless, the Mexican government project (Instituto Nacional de Antropología e Historia, INAH) to clean the paintings of dirt and graffiti under the direction of Sandra Cruz Flores (Cruz Flores 2003Flores , 2004Flores , 2005 has revealed new visual details in these panels. In addition, recent developments in imaging have further assisted our efforts in new documentation. Specifically, Hurst and Ashby identified the outline of a human head behind a large shield (Panel 4-05, Figure 4) in front of the north shelter. The shield image, previously designated as Painting 8 by Grove (1970), is not a stand-alone symbol but is part of a larger scene similar in scale to the C-1 painting.

FIGURE 2.
Photograph of an enthroned figure painted in Olmec style sitting on a throne (Panel C-1) and its context above the south rockshelter grotto. The base of the figure is 9 m above the rockshelter floor. Sample location indicated by arrow.

EXPERIMENTAL STRATEGIES
The collage of paintings at Oxtotitlán is complex, but with no visual means to establish temporal information such as superimposed paint layers or temporally sensitive elements. Radiocarbon dating the paints was the key to unraveling the relationships among paintings inscribed in the cave. Our study focused on Panels C-1, C-2, and 4-05 because of intriguing questions about the subject matter and the availability of green, red, and black paints that could potentially contain organic carbon for radiocarbon dating. A principal objective of the study was to establish chronologies of individual rock paintings at the site either by direct radiocarbon dating paint components or by constraining ages by 14 C dating the natural, biogenic coatings that encrust the paint layers.
Thin (∼ 1 mm) coatings on limestone surfaces within the Oxtotitlán site were visually observed on painted and unpainted panels. One of the most common constituents of natural rock coatings on limestone and sandstone inside dry rockshelters and openair sites is calcium oxalate. Oxalate-rich coatings occur globally and have been identified at a multitude of rock art sites, usually via chemical studies of rock paints. There is considerable evidence that oxalates in rock coatings are biogenic, the byproduct of bacteria (Bonaventura et al. 1999;Hess et al. 2008;Rusakov et al. 2015), fungi (Gadd et al. 2014;Monte 2003;Ortega-Morales et al. 2016), lichens (Del Monte and Sabbioni 1987;Russ et al. 1996), and mixed microbial communities ( Gorbushina 2007). Evidence suggests that all carbon in oxalate biofilms originates from atmospheric carbon dioxide (Beazley et al. 2002) and that, once formed, there is no carbon exchange with the substrate (Watchman 1993). As such, radiocarbon ages of oxalates would correlate with periods when the microbial communities flourished on the rock surfaces in the past (Caneva 1993;Russ et al. 2000) and thus date the formation of the rock coating.
At sites where oxalates cover or encapsulate ancient rock paints, and the stratigraphy of the coating relative to the paint layers can be discerned, 14 C analysis of the oxalate can be used to determine minimum or maximum ages for paintings. This approach has been used at sites worldwide since Watchman (1991) first radiocarbon dated oxalate coatings in Australia. Since then, 14 C ages of oxalate rock coatings have been used to constrain the age of ancient paintings at numerous sites in Australia (David et al. 2013;Watchman et al. 2000;Watchman et al. 2005), as well as sites in Africa (Mazel and Watchman 2003), the Iberian Peninsula (Ruiz et al. 2012), North America (Whitley 2013), and South America (Steelman et al. 2002).
Our sampling strategy at Oxtotitlán focused mainly on black paints because of the possibility that the pigments were prepared using pyrolyzed carbon (PyC), the byproduct of incomplete combustion of plants or animal fats and referred to as charcoal, soot, carbon black, or lampblack. Many black paints used to create parietal rock art in antiquity were constructed from manganese and iron oxide minerals (Chalmin et al. 2003;Chalmin et al. 2006;Hyman et al. 1996;Koenig et al. 2014;Littmann 1975;Magaloni, Newman, et al. 1995;Vázquez et al. 2008;Zoppi et al. 2002); however, PyC was also commonly used, thus providing a source of carbon for direct radiocarbon analysis (Baker and Armitage 2013;Bonneau et al. 2011;López-Montalvo et al. 2014;Magaloni 2001;Morwood et al. 2010;Valladas et al. 1992;Valladas et al. 2001).
For red and green paints, the pigments are invariably mineralbased; however, in order to produce a paint, it is necessary to suspend the mineral particles in a liquid medium (vehicle) that also serves to bind pigments to the substrate (binder). It is generally hypothesized that carbonaceous substances such as seed oils, plant resins, vegetable gums, animal fats, or beeswax were added to create a liquid paint mixture, thereby providing a source of organic carbon for radiocarbon dating (Ilger et al. 1995;Magaloni 1998;Morwood et al. 2010;Russ et al. 1990;Wright 2008Wright , 2010. Here we employed a low-temperature oxygen plasma system to isolate carbonaceous carbon used in the original paint mixtures from carbon-rich minerals in the substrate and natural rock coatings, allowing for direct radiocarbon dating of the artifacts (Rowe 2009). Excited oxygen in the plasma system reacts with organic carbon yielding carbon dioxide (CO 2 ). The reactions occur below the decomposition temperatures of carbon-rich minerals such as calcite (CaCO 3 ), the primary component of limestone and present in sandstone. Both of these rock types are common substrates used for ancient rock paintings. Oxalate minerals whewellite (CaC 2 O 4 ·H 2 O) and weddellite (CaC 2 O 4 ·2H 2 O) are prevalent in natural rock coatings and are also stable in the plasma and so do not interfere with the radiocarbon analysis (Armitage et al. 2001;Ilger et al. 1995).

Preliminary Analyses and Sample Selection
To distinguish between inorganic, mineral-based pigments and PyC pigments, we analyzed the paintings in situ using a handheld portable X-ray fluorescence (HHpXRF) spectrometer. This technique has proven effective in the study of rock art primarily because it is noninvasive and provides qualitative/semiquantitative data on chemical elements in the paints (Koenig et al. 2014;Newman and Loendorf 2005;Roldán et al. 2010;Rowe et al. 2011). The method is especially useful for discriminating between mineral-based and carbon-based pigments, and thus for identifying specific paintings that could potentially be 14 C dated (Beck et al. 2013;Koenig et al. 2014;Rowe et al. 2011). While there has been controversy surrounding the use of HHpXRF for quantitative analyses of archaeological materials (Shackley 2010), our goal was simply to establish whether manganese or iron concentrations in paints were above the natural background levels of the bedrock.
A handheld Innov-X Alpha Series pXRF with a silver anode X-ray tube source and a Si-PIN diode detector was operated in "soil mode" using 40 kV excitation energy with 30-second analysis times. A Hewlett Packard iPAQ personal digital assistant was interfaced to the instrument to control the analyses and to store data. We analyzed multiple locations on each painting at the site and then analyzed non-painted surfaces adjacent to the paintings to establish the background composition of the limestone substrate and natural rock patinas.
Results of these analyses showed elevated iron for red paints and elevated copper for green paints, compared to the background. Nevertheless, manganese and iron concentrations in all black paintings at the site measured below the detection limit of the instrument or were statistically the same as the background compositions, i.e., ≤ .014 ± .004 % (McPeak et al. 2013). This fact is in contrast to analyses of black paints in the Lower Pecos Canyonlands, Texas, where measured concentrations of manganese in black paintings were consistently and significantly greater than ambient concentrations (Koenig et al. 2014). At the Oxtotitlán site, it was clear that red and green pigments are mineral-based but that the black paints are most likely carbon-based, results that are consistent with Mesoamerican colorant manufacture used in later periods (Magaloni 2001).
We focused on three areas at the site to collect samples for further laboratory study to explore the feasibility of obtaining 14 C ages of the paintings. We collected paint samples from Panel C-1 (enthroned figure, Figure 2), Panel C-2 (jaguar figure, Figure 3), and Panel 4-05 (shield figure, Figure 4). We selected areas on the paintings where small chips were in the process of natural exfoliation to minimize damage to the imagery, and that we judged free of possible contamination from graffiti, cleaning, or restoration. Samples were removed by slight prying using a pre-cleaned knife blade and allowed to drop onto a sheet of aluminum foil that was then wrapped, labeled, and placed in a plastic sample bag. Samples were collected such that the paint layers remained intact on a small section of the substrate. We also collected non-painted samples directly adjacent to sampling areas to provide chemical and stratigraphic information on the background that included the natural rock coating and the basal limestone.

Laboratory Analyses
To characterize the paints and natural rock coatings, we used Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) Spectroscopy, Environmental Scanning Electron Microscopy with Energy Dispersive X-ray Spectroscopy (ESEM-EDS), and optical microscopy of polished thin-sectioned samples.

ATR-FTIR.
We used a Perkin-Elmer Spectrum 100 with a diamond-zinc selenide composite ATR that allowed

Advances in Archaeological Practice
A Journal of the Society for American Archaeology May 2017 FIGURE 5. An example of an Attenuated Total Reflectance-Fourier transform Infrared (ATR-FTIR) Spectroscopy spectrum from the analysis of a sample collected near Panel C-1 (black spectrum). Also shown are overly spectra from the analysis of a calcium oxalate standard (blue) and a calcium sulfate standard (red) demonstrating that the coating is primarily oxalate and sulfate.
microgram-sized samples to be analyzed. The ATR-FTIR analysis of the natural rock coating consistently showed the presence of oxalates and sulfates, the former based on the broad carbonyl peak at 1,610 cm −1 and sharp diagnostic peaks at 1,300 cm −1 and 790 cm −1 , and the later based on the broad peak at 1,110 cm −1 and a sharp peak at 655 cm −1 ( Figure 5). A peak at 1,400 cm −1 indicative of carbonate was also detected in several samples, likely from inclusion of calcite from the limestone substrate.
ESEM-EDS. The micromorphology, microstratigraphy, and elemental compositions of paints and coatings were established using a Philips XL 30 ESEM operated at 30 keV in the "wet mode." We analyzed sample surfaces and cross-sectional views of bifurcated samples, as well as powdered aliquots scraped from samples using a needle.
ESEM images revealed that the pigment used in the black colorant from the C-1 painting was distinctly different from the pigment used for Panel 4-05. In the former, the paint layer appeared amorphous under high magnification; moreover, the paint was not incorporated within the coating or substrate, but instead formed a distinct stratum between the coating and basal limestone. EDS analysis of the pigment material showed that it was predominately carbon (Figure 6). These results are consistent with a petroleum-based pigment, either asphalt or processed bitumen (Argáez et al. 2011), a result that precludes using the black paint sample from Panel C-1 for 14 C analysis. Mesoamerican peoples used bitumen for many purposes, such as decoration, as a sealant, and as an adhesive, especially in Mexico's southern Gulf Coast lowlands where the material occurs naturally (Wendt and Cyphers 2008). A rare example of bitumen (chapopote) wall paint is noted at the Mayan site of Bonampak in Chiapas, Mexico (Magaloni 2001:181-184).
The pigment in the shield figure (Panel 4-05) consisted of small, black particles with dimensions ∼5 μm × ∼ 20 μm that were dispersed within the coating (Figure 7). These particles were also carbon-based, as demonstrated in the EDS analysis of the particles (Figure 7), with estimated carbon/oxygen (C/O) ratios between .2-.3. This C/O ratio, together with particles sizes > .2 μm 2 , suggests that the particles are PyC with size and compositional properties between charcoal and soot (Hedges et al. 2000;Preston and Schmidt 2006). ESEM images of cross-sectioned samples further demonstrated that the natural coating is heterogeneous with a mixture of microcrystals and platy crystals. The morphologies and elemental concentrations are consistent with whewellite and/or weddellite for the microcrystals, while the platy crystals, along with the presence of sulfur, indicates gypsum (CaSO 4 ·2H 2 O) (McPeak et al. 2013;Russ et al. 1999). Optical microscopy of polished thin-sections revealed that the thickness of the oxalate/sulfate rock coating was ∼ .5 mm and completely covers the paint layers.

AMS 14 C Dating Strategies
Based on the analyses described above, we concluded that there were three possible options for establishing direct or relative 14 C ages of several of the paintings: (1) direct dating of PyC pigment from the shield figure, (2) direct dating of a carbonaceous vehicle/binder used to construct red and red/green paints from the C-1 and C-2 paintings, and (3) radiocarbon analysis of the oxalate coatings superimposing paints to obtain minimum ages.

May 2017 Advances in Archaeological Practice
A Journal of the Society for American Archaeology  For these experiments, we employed a low-temperature oxygen plasma methodology for three primary purposes: (1) to oxidize and extract carbon from the PyC pigment in the shield figure for direct AMS 14 C analysis; (2) to evaluate whether an organic binder/vehicle was used to produce the red and red/green composite paints, and, if present, extract the organic carbon for radiocarbon analysis; and (3) to pretreat the oxalate samples by removing organic contaminants from the mineral coatings prior to the 14 C analysis (Rowe 2009;Steelman and Rowe 2012).
The low-temperature plasma oxidation system creates electronically excited oxygen molecules within an enclosed glass chamber that react with organic matter at a low temperature (∼100°C), yielding carbon dioxide and water. The excited oxygen does not react with oxidized forms of carbon such as carbonates and oxalates, and the low-temperature prevents thermal decomposition of these compounds (Armitage et al. 2001;Ilger et al. 1995). Thus, the excited oxygen reacts solely with reduced organic carbon yielding CO 2 that can be cryogenically isolated on a liquid nitrogen cold finger that is flame-sealed for 14 C analysis. This technique was originally developed for extracting organic carbon in prehistoric rock paints for 14 C analysis, especially those on carbonate surfaces (Russ et al. 1990) and has proved to be effective for radiocarbon dating these artifacts (Rowe 2009;Steelman and Rowe 2012).
We prepared the paint and oxalate samples for the plasma treatment by removing loose detritus from the surfaces with light scrubbing using ultrapure water (18.2 M ) and drying them in a 60°C oven overnight. Sample surfaces, including paints and oxalate coatings, were separated from the limestone substrate using a razor knife. The resulting powders were analyzed via ATR-FTIR to confirm the presence of oxalate and evaluate the amount of carbonate included in the sample. The samples were then exposed to an oxygen plasma discharge following the procedures described by McDonald et al. (2014).
The sample from the shield figure had a 3 cm × 3 cm surface area with approximately two-thirds of the sample covered with pigment while the remaining one-third was void of paint and contained only the natural rock coating. The paint and the unpainted surfaces were removed separately, providing one sample that was a mixture of paint and natural coating and a second consisting of only the coating, both from the same specimen. The samples were scraped from the substrate while we observed them under a dissection microscope to ensure that the two areas were not cross-contaminated.

Advances in Archaeological Practice
The two samples were treated separately using a lowtemperature oxygen plasma system. The 1.5 hr plasma oxidation of the paint sample yielded 40 μg of datable carbon (in the form of carbon dioxide) that was sealed in a glass ampoule and sent to the Center for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory (CAMS-LLNL) for 14 C analysis. Afterward, we continued plasma exposure of the paint sample until no more carbon was extracted, thereby removing residual organic matter from the remaining oxalate. The "cleaned" oxalate sample was processed for AMS 14 C analysis as described below.
Plasma oxidation of the unpainted portion of the sample yielded ≤10 μg carbon, insufficient for an accurate 14 C measurement. Thus, we assume that the carbon extracted from the paint was predominately from the PyC pigment, with minimal background contamination. Nevertheless, plasma treatment of background samples from other areas at the site contained significant amounts of carbon.
(2) Direct 14 C Dating of a Binder/Vehicle in Paints from Panels C-1 and C-2. A second strategy for determining the chronologies of the rock paintings was via plasma extraction of organic matter that might have been added to the inorganic pigments in the original paint mixture. It is generally surmised that mineral pigments such as copper and iron in green and red paints, respectively, needed to be suspended in an oily (organic) medium (vehicle) to produce a paint and to bind the pigments to the rock surface (binder). If such a vehicle/binder was an organic substance, and if the carbon could be extracted using low-temperature plasma oxidation, then we could 14 C date the paint (Rowe 2009;Steelman and Rowe 2012).
We selected paint and background samples from two areas at the Oxtotitlán site for these experiments, a red-green composite paint from Panel C-1 (enthroned figure) and a red paint from Panel C-2 (jaguar figure). Unfortunately, plasma treatment of the non-painted, background samples collected near the paintings yielded considerable amounts of carbon, equivalent to that produced by the treatment of paint samples. Organic substances can occur on rock surfaces with oxalates due to microbial activity that results in the formation of oxalates (Spades and Russ 2005). We concluded that there was excessive organic matter in the natural rock coating (i.e., the background) that precludes obtaining reliable 14 C dates from a binder or vehicle (Livingston et al. 2009).
As with the paint sample from the Panel 4-05 shield figure, we continued to oxidize the C-1 and C-2 paint samples until no measurable carbon dioxide was produced during the plasma reaction. This effectively removed the organic matter in the paint and coating, leaving only the oxalate component of the biofilm that covered the painting to be further processed for 14 C analysis.
(3) 14 C Dating of Oxalate Rock Coatings. We 14 C dated five calcium oxalate coating samples, three of which occurred on top of paint layers and two from unpainted surfaces adjacent to rock paintings including the Panel C-1 enthroned figure and Panel 4-05 shield figure. The oxalate coating superimposes paint layers (Figure 8), indicating that the production of the rock art would have occurred prior to the oxalate deposition; therefore, the oxalate 14 C ages represent minimum ages of the paintings.
We prepared the samples for 14 C measurement by first removing loose detritus from sample surfaces with light scrubbing using ultrapure water and drying them overnight in a 60°C oven. Coatings and paints were scraped from the substrate using a razor knife under magnification with a dissection microscope. The powdered samples were then reacted in the low-temperature oxygen plasma, first to measure the quantity of organic matter so as to

May 2017 Advances in Archaeological Practice
A Journal of the Society for American Archaeology  Bronk Ramsey 2009, 2013 with IntCal13 curve data from Reimer et al. (2013).
c The large uncertainty is due to a small sample size (40 μg of C).
evaluate the feasibility of directly dating organic matter in the paint, and second to eliminate the reduced carbon in the samples that might interfere with 14 C dates of the oxalate coating.
Once cleaned via the oxygen plasma treatment, aliquots of the samples were analyzed using ATR-FTIR to test for the presence of carbonates. Samples were then placed in pre-cleaned glass centrifuge tubes with ∼3 mL of 1.0 M phosphoric acid (HPLC grade) and ultra-sonicated for 45 minutes to remove carbonates. The tubes were centrifuged, the supernatant was removed, and then the tubes were rinsed twice using ultrapure water. Samples were dried overnight in a 90°C oven, and aliquots were reanalyzed using an ATR-FTIR to confirm that all carbonates had been removed. The powdered samples were then sent to CAMS-LLNL for combustion of the oxalates to carbon dioxide, followed by reduction to graphite for an AMS 14 C target.

RESULTS
All the black paintings in the Oxtotitlán site were constructed from pigments other than manganese or iron, based on the absence of these elements in the HHpXRF analyses on site. This finding suggests that the black pigments were produced from carbon-based materials. We also ascertained in the laboratory that the rock coatings that cover the paintings are composed mainly of calcium oxalate and calcium sulfate, the former having a biogenic origin. We measured the radiocarbon age of one prehistoric paint sample, three oxalate coatings that superimpose paint layers, and two oxalates samples collected adjacent to the rock art (Table 1). An overarching goal of the study was to establish the age of the iconographic Panel C-1, but unfortunately this was not accomplished. We discovered that the black pigment used to construct the artifact was carbon-based but composed of asphalt or processed bitumen and thus could not be radiocarbon dated. This is, to our knowledge, the first identification of a petroleum-based pigment used in prehistoric rock art. Although the age of the oxalate coating next to the painting dated to 3705 ± 30 yr B.P., the oxalate in Panel C-1 occurs both above and below the paint layer and thus cannot constrain the age of the paint (Figure 9).
The experiment to radiocarbon date the Panel 4-05 shield figure appears reliable. The pigment was identified as PyC with particle size and composition between charcoal and soot. This organic carbon in the paint was directly radiocarbon dated as 1980 ± 190 yr B.P. The oxalate layer covering the paint was radiocarbon dated as 1370 ± 30 yr B.P., which is stratigraphically consistent since the coating occurs on top of the paint layer and so should be younger. The results of the two oxalate ages from Panel C-2 (3195 ± 30 yr B.P. and 1410 ± 35 yr B.P.), both of which were superimposing paint layers, indicate that these artifacts were produced prior to 3,195 years ago as a minimum age.
The radiocarbon ages from all oxalates at the site range from 590 to 3705 yr B.P. This disparity in ages is not unusual within a single site. For example, oxalate ages measured by Ruiz et al. (2012) at the Spanish site Cueva del Tio Modesto ranged from 2800 to 5210 yr B.P., and the range of two samples at Abrigo de los Oculados was 2610 to 4675 yr B.P. As oxalate layers form over time, a weighted average of the dated material is obtained. If layers are thicker in one place than in another area, disparate ages would result.

CONCLUSIONS
This study provided an opportunity to apply two rock art dating strategies on a single sample, thereby providing a test of these dating techniques. We identified a paint sample from Panel 4-05 that contained both a pigment made from pyrolyzed carbon and a biogenic rock coating that superimposed the paint layer. We used a low-temperature oxygen plasma to isolate the carbon in the paint from the carbon-rich limestone substrate and oxalate coating to obtain a direct measurement of the radiocarbon age of the pigment. Furthermore, the oxygen plasma treatment "cleaned" the residual sample of organic carbon without loss of biogenic oxalate, a loss that would have occurred using traditional wet chemistry acid-base treatments. Only a direct acid treatment to remove carbonates was necessary to prepare the oxalates for the radiocarbon assay. The resulting radiocarbon ages of the two separate materials from Panel 4-05 were stratigraphically consistent, with the age of paint (1980 ± 190 yr B.P.) being older than the biogenic rock coating that covers the

Advances in Archaeological Practice
A Journal of the Society for American Archaeology May 2017 FIGURE 9. Optical micrograph of a polished thin-section of a sample from Panel C-1. The paint layer in this sample occurs within the oxalate coating and so we cannot deduce a relative age of the paint layer.
paint (1370 ± 30 yr B.P.). We conducted a direct test with a positive result from two methods most often used to determine or constrain rock art chronologies.
The consistency of the measured radiocarbon ages of the PyC in the paint and the biogenic coating provides confidence in the data and the methodology. Previously, Armitage et al. (2001) directly tested the plasma oxidation applied to PyC by radiocarbon dating charcoal in rock paint from Mayan hieroglyphs that contained calendar dates. The measured radiocarbon ages of the charcoal proved to be older than the calendar dates by between 110 and 140 years. While this is not a significant discrepancy in terms of rock art, these results do reinforce caution when evaluating radiocarbon ages from PyC in rock art paints. The most pressing problems are from the use of "old wood" to produce the pyrolyzed carbon (Schiffer 1986), "old charcoal" in constructing the paints, and the necessity of analyzing very small samples (Armitage 2001).
Another unique strategy used here was the oxygen plasma treatment of oxalate samples to remove organic contaminants in lieu of using acid-base methods. This protocol was also applied to oxalate coatings covering rock paints from Panel C-2, where paints underlie the rock coating. The radiocarbon age for one sample was the oldest measured at the site, suggesting that the paints were applied prior to 3195 ± 30 yr B.P. A second sample from the same panel yielded an age of 1410 ± 35 yr B.P. This significantly younger age does not necessarily compromise the older date. Oxalate biofilms form over long periods of time, and the radiocarbon age of a particular coating sample reflects the average of the periods of production of the organisms that produce the oxalates.

DISCUSSION
Our research significantly expands the periods of cave use at Oxtotitlán, and the sequence of dates shows that episodes recurred over time adding to our appreciation of Mesoamerican sacred landscapes and ritual practice. The radiocarbon date from the oxalate covering a paint layer demonstrates that the polychrome painting tradition extends back much earlier than the Formative period (1500-400 B.C.) date assigned on stylistic grounds to the bird-costumed, enthroned figure's portrait (Panel C-1). Radiocarbon results on oxalate of 1520-1410 cal B.C., a terminus ante quem marker, point to polychrome painting (Panel C-2), in the early part of the Early Formative period, the earliest evidence for this medium in Mesoamerica.
The large polychrome painting designated Panel C-2 testifies to ancient ceremonialism. Grove (1970) identified what he thought were jaguar spots arrayed across the panel when he viewed it in the late 1960s, possibly aligning the cave with the jaguar character. Our enhanced photographs support Grove's identification of the darker markings visible in images as jaguar spots (Figure 3). If the identification is correct, the panel may refer to rain rituals focusing on the jaguar of the kind practiced historically in this area, including at Oxtotitlán's nearest settlement Acatlán (Díaz

May 2017
Advances in Archaeological Practice A Journal of the Society for American Archaeology Vázquez 2003;Gutiérrez and Pye 2010). Thus, we propose that Oxtotitlán was a significant regional ritual center from the beginning of the Formative period and perhaps even earlier.
The Early Formative, or earlier, oxalate date for Oxtotitlán's Panel C-2 suggests that an elaboration of rituals at regional nature shrines, with a sacred template of mountain-cave-water source, preceded the human-built Formative ceremonial centers that often incorporated or imitated these features. Early offerings at El Manatí's natural hill-spring complex, across Mexico in Veracruz, date to about the same time or earlier. Note that those first offerings to the spring water included concretions probably formed by water in a cave and transported to El Manatí (Ortiz and Rodríguez 1999).
In contrast, the shield figure painting (Panel 4-05) is uncommon among Formative period imagery.  (Hajovsky 2012(Hajovsky , 2015Nicholson 1959). Among earlier paintings from the first to sixth centuries A.D., shield-bearing figures include tomb paintings at Palenque where nine painted individuals accompany what is believed to be a dynastic founder to the Underworld, a sacred place related to caves (Robertson 2000).
Cave painting is present across Mesoamerica, and beyond, from the Archaic period; yet at the caves in Guerrero, pictographs transformed into a new art form that was the foundation for muralism in Mesoamerica. During the Formative period in Guerrero, sites such as Oxtotitlán were using polychrome paint technology and complex figurative scenes in pictographs prior to the innovation and widespread adoption of plaster technology. We suggest that mural painting on plaster, now recognized as a major Mesoamerican medium, might have begun with cave painting and that polychrome art was one of the significant innovations in cave painting in the Formative period from the start. To date, the earliest murals known are Maya wall paintings from San Bartolo, Guatemala, ca. 300-100 B.C., which use Olmec maize god imagery (Saturno et al. 2005;Saturno et al. 2006;Taube et al. 2010), perhaps indicating that both ideological and technological innovations spread from Formative period sites like Oxtotitlán. Later painted works, including the well-known murals from Bonampak (ca. 790 A.D.) (Miller and Brittenham 2013) and Cacaxtla (ca. 650 to 950 A.D.) (Brittenham 2015), as well as Late Classic cave painting at Naj Tunich (Stone 1995), elaborate on scenes of ceremonialism, palace, rulership, war activities, and ritual associated with rain and fertility; these examples demonstrate continuing use of painted cave walls and human-built spaces as a medium for engaging occult religious and political content.
In sum, the dated samples from the Oxtotitlán paintings provide evidence for a tradition of polychrome rock art preceding intensive human-built ceremonial complexes in Guerrero or the Gulf Coast regions. At Oxtotitlán Cave, we trace a lineage extending from the earliest dated polychrome painting related to jaguar imagery (Panel C-2) to the more elaborate Olmec Formative-style enthroned bird-costumed human image (Panel C-1). Paint and oxalate dates suggest that a newly identified image (Panel 4-05) of a human brandishing a shield likely post-dates the Olmec-style enthroned bird-costumed human portrait (Panel C-1). They document the significance of the cave as a site of continued painting and associated ritual through periods of dramatic social change at the end of the Late Formative period and into the Early Classic period. The radiocarbon dates of this study will enable a more contextualized analysis of the Oxtotitlán iconography and the longevity of ritual practice at landscape shrines outside of habitation areas.