Introduction
Heat-hardened amorphous fragments of earth are a common component of many archaeological sites on the SW US coastal plains and in northeast Texas, USA (Fig. 1). A number of names have been applied in the literature to these artifacts (Costa et al., Reference Costa, Frederick, Gregory and Oksanenin press), such as burned clay, fired clay (Ricklis et al., Reference Ricklis, Weinstein and Wells2012), clay balls, baked clay, clay lumps, and clay nodules (Turpin, Reference Turpin2004), but here they will be referred to as burned clay objects (BCOs). Unlike burned daub, which is heat-hardened earthen plaster typically associated with wattle and daub buildings, BCOs generally lack impressions of timber or other organic material such as grass, and instead consist of irregularly shaped, rounded to angular fragments of oxidized or reddened, slightly hardened earth generally less than about 7 cm in diameter. A few sites, such as the Morhiss Site (41VT1; Campbell, Reference Campbell, Fox and Hester1976; Dockall and Black, Reference Dockall and Black2024) have large (< 10 cm) spherical to ellipsoidal objects that appear to have been molded by hand from moist mud (Fig. 2). Post-excavation soil micromorphology and experimental work at 41DL203 near Dallas, Texas (Tinsley and Dayton, Reference Tinsley and Dayton2011) implied based on BCO internal oxidation–reduction patterns that some of the larger tetrahedron-shaped fragments were once part of larger, up to 10-cm diameter, spherical objects that had since disintegrated, even though no complete, spherical objects were present among the artifact assemblage.
Choropleth map of Texas showing the number of sites reported to contain burned clay of any kind, by county. The higher number of sites around Houston, Corpus Christi, and Brownsville most likely reflect a higher number of archaeological investigations in those counties. Data and map compiled by August Costa and Arlo McKee.
Photograph from the Works Project Administration (WPA) excavations at the Morhiss Site (41VT1) showing a concentration of cracked, but still intact, roughly spherical burned clay objects (BCOs) just above the board with the feature label. We think the largest BCO here is approximately 10 cm in diameter. The white arrows point to fragments of disarticulated BCOs. Photograph courtesy of the Texas Archeological Research Laboratory, The University of Texas at Austin. 41VT1-385.
Unfortunately, the fact that irregular shaped burned earth fragments can be created by natural processes without human agency has resulted in some archaeologists ignoring burned clay altogether and refusing to accept any burned clay as the result of human agency. Wildfires are capable of burning dead tree stumps and roots in situ, as well as the earth adhering to the roots of the root plates of wind-upturned or toppled trees, thereby creating burned clay of various morphologies, entirely in the absence of human agency. Ironically, burned clay created by forest clearance, specifically the uprooting of trees and burning the stumps can be one of the more pervasive forms of evidence of Historic period forest clearance (e.g., Cummins et al., Reference Cummins, Frederick, Costa, King, Barrett and Mangum2023). As a result of this equifinality issue, archaeological attention to these artifacts has been minimal with the exception of a few stalwart proponents such as Leland Patterson (Reference Patterson, Highley and Hester1980, Reference Patterson1989; Patterson et al., Reference Patterson, Hudgins, Gregg and McCLure1987, Reference Patterson, Hudgins, Gregg, Kindall, McCLure and Neck1993, Reference Patterson, Hudgins, McCLure, Kindall and Gregg1994) who documented prehistoric burned clay objects at archaeological sites across the upper Texas Coast.
The clearest sign that burned clay objects may be artifacts of human agency is their association with prehistoric occupational debris in ancient occupation layers. In such contexts, BCOs have been observed in a variety of arrangements, such as discrete concentrations typically designated features/hearths/ovens (Fig. 2), scatters across occupation surfaces, or rarely, dense accumulations typically called middens (Fig. 3).
(A) Profile photograph of the burned clay midden exposed at Microartifact Column 1 at the Lodoso Site, from which four of the dated BCOs were collected (white dashed line rectangles, left side of image). Yellow dashed line denotes the location of soil micromorphology block 2, a scan of which is shown in (C). (B) Same photograph as top left but with the soil micromorphology blocks isolated prior to collection. (C) Scan of the polyester-embedded soil micromorphology block 2. (D) Scan of a selection of larger burned clay objects from 41NU114 that have been polyester embedded and slabbed with a rock saw. CC denotes a carbon core, and OC is oxidized core.
In prehistoric North America, burned clay objects have been most frequently reported from the Gulf Coast (Texas to Florida) and California, with Poverty Point Objects (PPOs) being the only formally recognized variant. PPOs occur in association with the Poverty Point culture, best represented in the Lower Mississippi River valley, and occur in a wide range of molded shapes such as biconical, cylindrical, tetrahedral, and spherical, which are often embellished with broad deep grooves or other forms of surface design such as cane impressions, narrow spiral grooves, or perforations (Ford et al., Reference Ford, Phillips and Haag1955; Ford and Webb, Reference Ford and Webb1956; Gibson, Reference Gibson1983, Reference Gibson2000, 2007; Hays et al., Reference Hays, Weinstein and Stoltman2016; Henry et al., Reference Henry, Ortman, Arco and Kidder2017; Hays, Reference Hays2019). By comparison, the Texas BCOs discussed here are, as one colleague has described them, ill-formed and expediently shaped (Turpin Reference Turpin2011). In both cases, PPOs and the Texas BCOs discussed here, archaeologists generally assume that these objects are artifacts of cooking and were most likely used as heat transfer devices that could have been employed in a variety of cooking processes such as baking in earth ovens, or boiling water. In Texas, BCOs are most common in regions that lack significant rock (Fig. 1) and hence are viewed as expedient replacements for rocks in cooking (faux rocks).
Burned clay also has been described as a by-product of cooking elsewhere. Australian Aborigines are known to have fired soil peds to use in earth oven cooking (Beveridge, Reference Beveridge1869) in a manner similar to how BCOs have been interpreted along the Texas coast. Huatia cooking in the central Andes of Peru is an expedient earth oven cooking method used to bake tubers such as potatoes, which results in fired soil clods and peds (Sayre and Rosenfeld, Reference Sayre, Rosenfeld and Staller2021).
The work that is presented here chronicles efforts to date prehistoric activity at an archaeological site, the Lodoso Site (41NU114), which is located near Corpus Christi, Texas (Fig. 4). The site was initially reported by Patterson and Ford (Reference Patterson and Ford1974) as a small surficial scatter of lithic flakes, BCOs, and snail shells representing a pre-contact period occupation. Trenching by Abbott and Morris (Reference Abbott and Morris2018) as part of an archaeological survey in advance of the widening of US Highway 77 by the Texas Department of Transportation (TxDOT), and subsequent excavations by Feit and Goldstein (Reference Feit and Goldstein2019), revealed that the site contained multiple prehistoric occupations. These included two large, burned clay middens near the surface, beneath which lay at least two discrete stratigraphically isolated occupations that contained burned clay scatters and spatially discrete hearth-like features. The analysis and reporting of the site is, in part, focusing on the most ubiquitous artifacts, namely BCOs, and their potential utility in understanding prehistoric behavior in this region. Critical to achieving this is establishing their age and relationship to other units within the site.
Map of 41NU114 showing the approximate location of the large burned clay object middens, and the excavation units. BT = backhoe trench, MAC = Microartifact Column Blocks are areas scraped with a Gradall searching for features. Inset at bottom left shows the approximate locations of places mentioned in the text.
As with many Texas archaeological sites, charcoal can be uncommon, which leaves researchers searching for other ways to date prehistoric occupations. For burned clay, two methods have been employed previously. Turpin (Reference Turpin2004, Reference Turpin2011), investigating the only other site with a burned clay midden (41AT232) in Atascosa County (Texas), used AMS radiocarbon dating of organic matter within the BCOs and bulk soil organic matter humate dates collected from the same depth range to assess the age of the prehistoric features and occupation. It has been shown that carbon incorporated during BCO formation could remain within clays even when fired up to 1000°C (well above temperatures attained during burning in antiquity; Johnson et al., Reference Johnson, Clark, Miller-Antonio, Robins, Schiffer and Skibo1988). The rationale here is similar to direct radiocarbon dates on prehistoric pottery (cf., Hedges et al., Reference Hedges, Tiemei and Housley1992), but as with radiocarbon dates on bulk soil, the source of the carbon dated is unknown and the accuracy of the dates questionable. The ages obtained from the burned clay objects were in correct stratigraphic order, between ca. 1925 and 3130 years BP, unlike the bulk soil humate ages, and led Turpin (Reference Turpin2004, Reference Turpin2011) to conclude that these artifacts represent human activities during the Middle Archaic to Late Archaic periods.
Luminescence dating (thermoluminescence (TL) and optically stimulated luminescence (OSL)) has been used to date a variety of artifactual materials ranging from heat-altered natural rocks including flint and fire cracked rock used in cooking, to manufactured heated earthen materials such as brick and pottery, as well as by-products such as slag (e.g., Wintle, Reference Wintle2008; Bailiff, Reference Bailiff and Bateman2019). Hypothetically, BCOs should be ideal candidates for luminescence dating, given that they can be relatively large and were heated sufficiently to transform the clays to a hardened brick-like state. Experimental firing of clay balls in Mexico within an open wood fire showed temperatures ranged from 502°C to 714°C (Simms et al., Reference Simms, Berna and Bey2013) and well above the 325°C and 375°C TL peaks in quartz targeted for luminescence dating (McKeever, Reference McKeever1984). It should be expected therefore that firing would have reset the TL/OSL signal. However, a recent paper that aimed to refine the age of Poverty Point sites in the lower Mississippi valley Louisiana rejected all 30 of the early TL dates on PPOs obtained by Huxtable et al. (Reference Huxtable, Aitken and Weber1972) because the dates conflicted with “reasonably established radiocarbon-based chronologies at Poverty Point or elsewhere” and implied that many of these artifacts were insufficiently heated to result in incomplete bleaching (Kidder and Grooms, Reference Kidder and Grooms2024, p. 115).
The only previous luminescence dating of Texas BCOs was conducted by James Feathers in association with the data recovery investigations at the Dimond Knoll site (41HR796; Feathers, Reference Feathers2022—unpublished report in possession of the authors), which was excavated by Coastal Environments, Inc., on the behalf of TxDOT. Feathers (Reference Feathers2022) used OSL to date four burned clay objects from a single feature (Feature 17) for which there was also a charcoal radiocarbon date. All four BCOs were dated using coarse grains, and one was also dated using fine grains. The results indicated that three of the four BCOs had been sufficiently heated to reset the luminescence signal. Of the three that were sufficiently reset (UW4127, UW4129, UW4130), the ages obtained using the central age model (CAM) were statistically identical with a weighted average of 780 ±40 years (2022 datum) or AD 1250 ± 40. The radiocarbon date from the same feature was 737–900 cal BP (Beta-582406, semi-carbonized hard wood, 900 ± 30 14C yr BP, 1σ) which at one standard deviation is slightly older than the OSL age but overlaps at two standard deviations. This example suggests that OSL dating of BCOs can, when the sample has been sufficiently heated to reset the traps of interest, return ages comparable to radiocarbon dating of charcoal.
This paper presents the results of a pilot study that dated a suite of seven BCOs collected from different stratigraphic positions within the Lodoso Site (41NU114), a prehistoric camp in Nueces County, Texas. One BCO was collected from near the modern ground surface, four were collected from a burned clay midden, and two from discrete occupations beneath the midden. These were dated using TL and OSL to (1) assess the suitability of BCOs for luminescence dating, (2) establish the age of the midden, and (3) test interpretations of the midden deposit. Interpretation of the results of this work are in part based upon the BCO ages with respect to a suite of four single grain OSL ages obtained from the sedimentary matrix, and with respect to paired AMS radiocarbon ages obtained from charcoal collected from the same levels as the BCOs.
Study area
The Lodoso site, 41NU114, is a multicomponent prehistoric campsite situated adjacent to Petronila Creek, 37 km (23 miles) west of Corpus Christi, Texas, near Driscoll, Texas (Fig. 4). It lies upon and beneath the first terrace, which lies about 6 m above the creek channel. The modern floodplain is inset into and lies below the first terrace. The prehistoric cultural deposits were situated immediately adjacent to the scarp separating the floodplain and the first terrace. Fieldwork identified two informal stratigraphic units, referred to here as Depositional Unit A and Depositional Unit B. The core of the site beneath the first terrace is comprised of Depositional Unit A, which was exposed to a depth of about 1.6 m by trench7 (Fig. 5), which extended across the scarp separating the first terrace from the floodplain. Because Unit B overlies Unit A and is inset into it, Unit B is thought to be younger (Fig. 6).
Drawing of the stratigraphy exposed by the east wall of backhoe Trench 7, annotated to highlight the two major stratigraphic units with respect to the single grain OSL sediment ages. Note that Trench 7, shown here, did not expose either of the middens (see Figure 4), but the approximate stratigraphic position of the middens is shown on the right side of the illustration.
OSL Equivalent Dose (D e) data from BT7. Mean is mean D e after outlier removal (this is not the D e used for age calculation; see text for details), OD is overdispersion and “n” the number of accepted measured grains. The abanico plots centered are on lowest D e extracted using Finite Mixture Model (FMM). Grains falling outside this lowest FMM component are shown in red with the next lowest FMM D e component shown with the second gray rectangle (see text for details).
Methods
Sedimentary profile dating
Four samples were collected to date the sedimentary deposits exposed by Backhoe Trench 7 by hammering 18-cm (7-inch) long galvanized fencepost pipe into the clean trench walls. Two samples were collected from Depositional Unit A (Shfd21006, Shfd21007) and two were from Depositional Unit B (Shfd21005, Shfd21008) and they were prepared and measured for single grain OSL dating (SG-OSL) in the Sheffield University Luminescence laboratory. Sample dose-rates were determined from elemental data from bulk material as measured by ICP-MS and ICP-OES (Table 1). These elemental data were converted to dose-rates using data from Guérin et al. (Reference Guérin, Mercier and Adamiec2011). Dose-rates were attenuated for particle size, density, and a paleomoisture value based on present-day ones with a 3% uncertainty (Table 1; Aitken, Reference Aitken1998). A cosmic dose contribution was calculated using the expression found in Prescott and Hutton (Reference Prescott and Hutton1994). Decay chain equilibrium was assumed not measured, however U:Th ratios are within the expected range for crustal material and taken to indicate no leaching or concentration of the more mobile U.
OSL-related data for BT7 with finally accepted De and Age shown in bold. Proportion is proportion of data falling into this component as determined by the Finite Mixture Model with a sigma-b run with of 0.2; components representing <10% data not shown.
a n is the number of De determinations accepted after screening. nm is the number of measured grains.
b D e values were determined using the Finite Mixture Model (FMM; Roberts et al., Reference Roberts, Galbraith, Yoshida, Laslett and Olley2000). The reported error is the standard error. Numbers shown in bold are those used for final reported ages.
c Radionuclide concentration measurement uncertainties are 10% for K and 5% for Th and U. Beta and gamma dose rates were calculated using the conversion factors of Guérin et al. (Reference Guérin, Mercier and Adamiec2011).
d Cosmic dose rates (Gy/ka) were calculated according to Prescott and Hutton (Reference Prescott and Hutton1994).
e Luminescence ages, rounded to the nearest 10 years, are expressed as thousands of years before AD 2024. Error is 1 sigma. Ages in bold are final reported ages.
Samples were prepared following the procedure to extract and clean quartz outlined in Bateman and Catt (Reference Bateman and Catt1996). Measurements were carried out on a size range of 180–212 μm. The samples underwent measurement using a Risø DA-15 luminescence reader with a dual laser single grain attachment and radiation doses were administered using a calibrated 90Sr beta source. Grains were mounted in 300-µm pits with 100 pits per 9.6 mm stainless steel aliquot. A focused 532 nm Nd:YVO4 laser provided the stimulation. Quartz single grain OSL measurements were at 125°C, with stimulation over 1s and measured through a Hoya U-340 filter. For each OSL measurement, signal was integrated from the first 0.06s of stimulation data with a background signal subtracted from an integral of the last 0.3s of stimulation data. Measurement was with the Single Aliquot Regenerative (SAR) protocol, with five regeneration points (including zero and a recycled point; Murray and Wintle, Reference Murray and Wintle2000, Reference Murray and Wintle2003). Within SAR, a preheat of 260°C for 10s was used, this having been arrived at experimentally from a dose recovery preheat plateau test. Within the SAR protocol a cut-heat of 160°C for 0s was applied prior to the measurement of Tx. Quartz grains were accepted if the recycling was within 1.0 ± 0.2, recuperation was <5%, error on the test dose was <20%, and the SAR growth curve fitted all SAR regeneration points within errors. Up to 1500 grains per sample were measured in order to obtain at around 50 grains that met these quality control criteria. Shfd21008 had slightly fewer accepted grains due to a large number of saturated grains being present (see discussion below).
Analysis of resultant D e distributions for each sample showed appreciable scatter (high Overdispersion – OD; Table 1) and positive skewing (Fig. 6) indicating deposition after only partial bleaching of the sediments. This accords with the fluvial terrace depositional setting with likely overbank deposition during turbid flood events. The uppermost sample (Shfd21005) collected from only 20 cm depth included appreciable numbers of zero-dose grains indicating grains that recently had been exhumed and buried by active pedoturbation (Bateman et al., Reference Bateman, Frederick, Jaiswal and Singhvi2003). Zero-dose grains were excluded from the age calculation of sample Shfd21005 in order to derive a sediment burial age. For this sample it is possible the positive skewing is not from partial bleaching at deposition, but some upward movement of older grains associated with pedoturbation, or a combination of both. Below this sample zero-dose grains were almost entirely absent (n = 1 for Shfd21006). Sample Shfd21008 also contained a number of grains in full saturation indicating inclusion in this sample of some material that had not been reset at all (Table 1).
Given the potential for both pedoturbation and incomplete resetting, selection of the D e component, which represents the true sediment burial age is difficult. In such cases, one approach is to assume that the D e component representing the most data (the dominant component) incorporates the grains least affected by pedoturbation and which were best bleached at deposition. To this end, the Finite Mixture Model (FMM; Roberts et al., Reference Roberts, Galbraith, Yoshida, Laslett and Olley2000) was used and ages derived from the dominant FMM component (Table 1). For samples Shfd21005 and Shfd21006, ages were based on the youngest FMM component because this was also the dominant one. For sample Shfd21007, the second lowest FMM D e component was used because it was marginally dominant, and the resultant age conformed with the profile stratigraphy. For sample Shfd21008, where there were no zero-grain D e values and, based on the number of saturated grains very poor bleaching prior to burial had taken place, the age was based on the youngest FMM component even though this was not the dominant component. Such an approach generated an age that was conformable to observed site stratigraphy.
Burned clay object dating (14C, TL, and OSL)
As noted previously, burned clay features at the site occur in two forms: (1) large (>10 m) dense concentrations of baked clay fragments that comprise a midden and that are thought to be the remains of earth-oven baking, and (2) discrete concentrations of relatively large BCOs that were stratigraphically beneath the midden and are interpreted to be vestiges of thermal features such as hearths or ovens. The seven BCO samples dated here (DRI Sample Numbers CB001 to CB007) were obtained from both contexts and recovered from 50 × 50 cm column samples excavated in 5-cm levels that were primarily collected for microartifact analysis. These bulk soil samples were subsequently wet sieved through a 1-mm mesh and the material retained on the screen was dried and then sieved through 4-mm, 2-mm, and 1-mm sieves.
Radiocarbon dating
A single charcoal fragment was collected from the same level as each of the luminescence-dated burned clay objects and submitted to the Northern Arizona University ACE isotope lab for AMS radiocarbon dating. There, the charcoal samples were subjected to an acid–base–acid treatment to remove inorganic and exogenous carbon (Santos and Ormsby, Reference Santos and Ormsby2013) and subsequently dried, weighed in tin rolling capsules, and graphitized on an AGE3 automated graphitization system (IonPlus, Switzerland). The graphite was then pressed into aluminum cathodes and analyzed on a MICADAS (Mini Carbon Dating System, IonPlus, Switzerland) 14C-AMS system. Data preparation was performed in the IonPlus software BATS (Wacker et al., Reference Wacker, Christl and Synal2010). The results of the AMS dating are presented in Table 2. Of the seven charcoal samples submitted for dating, one sample, the match for luminescence-dated burned clay object CB002, did not survive pretreatment.
Results of radiocarbon dating charcoal samples.
a Dates calibrated with the IntCal20 dataset using Calib 8.20 (Reimer et al., Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey and Butzin2020).
b Calculated as median probability age plus 74 years (luminescence ages were measured in 2024).
Luminescence dating of BCOs
Seven BCOs were prepared for luminescence measurement at the DRI Luminescence Research Laboratory in Reno, Nevada (Rodrigues, Reference Rodrigues2024). A Dremel tool was used to remove the outer 2 mm of each BCO to exclude light-exposed material. The outer 2 mm of each BCO was used to determine the dose rate contribution derived from the objects themselves. The interior portion of the sample was crushed carefully in a mortar and pestle and then treated with 10% HCl and 30% H2O2 to remove carbonates and organic material, respectively. The remaining sample was wet sieved to isolate the 32- to 63-µm polymineral fraction. For luminescence measurements, grains were mounted on 10-mm stainless steel discs using silicone adhesive sprayed through a 4-mm mask (∼3000 grains per aliquot).
All luminescence measurements were conducted on a Lexsyg Research Instrument housed at the DRI Luminescence Research Laboratory. OSL measurements were conducted using blue laser diodes with peak emission at 458 nm and power of 45 mW/cm2 at the sample position and infrared (IR) laser diodes with peak emission at 850 nm and power of 145 mW/cm2. Both OSL and TL signals were detected through a filter with a 365-nm detection window using a H7360-02 (UV/VIS) photomultiplier tube. Artificial irradiation was administered using a calibrated 90Sr beta source (for regenerative dosing) and 241Am alpha source (for alpha efficiency measurements) integrated in the Lexsyg Research Instrument.
TL glow curves were measured by heating aliquots to 450°C at a rate of 2°C/s in a N2 atmosphere under vacuum. A preheat of 200°C preceded each measurement. A single aliquot regenerative dose protocol (SAR-TL, adapted from Murray and Wintle, Reference Murray and Wintle2000, Reference Murray and Wintle2003) was deemed suitable for measurement of the equivalent dose (D e) after testing for large and irreversible sensitivity changes between measurement of the natural and first regenerative dose using a SAR-SARA protocol. The stable portion of the glow curves suitable for analysis was identified using a plateau test (Aitken, Reference Aitken1985). The signals were integrated within the identified plateau range (typically 250–300°C, Supplementary Fig. 1), and the obtained data were fitted with an exponential function. Interpolation of the natural signal to the exponential regenerative curve yielded the SAR-TL D e of the sample. A SAR-TL dose recovery test on CB001 yielded a ratio of given to measured dose of 1.05 ± 0.1 (n = 10). TL anomalous fading was measured by giving equal beta doses to several aliquots of CB001, CB004, and CB007 that were previously annealed to 450°C and then storing the discs at room temperature for various time periods spanning 2 days. The TL signal was measured for each aliquot after storage and plotted against the logarithm of storage time to produce a log-linear curve. Within the TL measurement uncertainties, no anomalous fading could be detected for any of the samples.
OSL D e measurements were carried out using a modified SAR protocol (Murray and Wintle, Reference Murray and Wintle2000, Reference Murray and Wintle2003; Fig. 2) with IR stimulation preceding blue stimulation to isolate the signal from quartz (post-IR-OSL; Roberts and Wintle, Reference Roberts and Wintle2001). The SAR protocol used for D e measurements included a preheat of 200°C and cut heat of 160°C determined on the basis of a preheat plateau test conducted on CB001. IR stimulations were carried out at 50°C for 250s and subsequent OSL stimulations were carried out for 100s at 125°C; OSL signals were integrated from the first 0.42s with background signals subtracted from the final 20s of stimulation. Notably, the IR-stimulated signal in these samples was weak compared to the subsequent OSL signal, suggesting that feldspars constituted a limited component of this grain-size fraction. A post-IR-OSL dose recovery test on CB001 yielded a ratio of given to measured dose of 0.98 ± 0.05 (n = 10). Post-IR-OSL anomalous fading rates were also measured for samples CB001, CB004, and CB007 using the method of Auclair et al. (Reference Auclair, Lamothe and Huot2003). Results yielded fading rates consistent with zero for each sample, suggesting that the post-IR-OSL protocol adequately isolates a stable quartz OSL signal for these samples.
Dose response curves for each aliquot were fitted using an exponential curve. Routine screening criteria included rejection of aliquots that exhibited the following behavior: (1) poor signals as judged from net natural test dose signals that were less than three times the standard deviation of the background signal; (2) failure to reproduce, to within 10%, the same sensitivity-corrected luminescence signal from identical regeneration doses given at the beginning and end of the SAR sequence; (3) a test dose error of >10%.; and (4) recuperation >10% of the sensitivity-corrected natural signal. Owing to the low overdispersion and symmetrical TL and post-IR-OSL D e distributions for all samples (e.g., Fig. 7, Supplementary Fig. 3), the D e values used for age calculation were modeled using the Central Age Model (CAM) of Galbraith et al. (Reference Galbraith, Roberts, Laslett, Yoshida and Olley1999).
Abanico plots showing the post-IR-OSL (left) and TL (right) D e distributions for CB001. The gray band is centered on the CAM D e estimate. Abanico plots for the remaining six samples can be found in the Supplemental Text.
Subsamples of the BCOs and external dose rate samples were dried and milled to a fine flour consistency and sent to ALS Geochemistry for geochemical analysis of U, Th, and K2O. Subsamples used for U and Th measurement were fused with lithium borate and measured with ICP-MS. K2O was measured from the sample with ICP-AES and converted to %K. Dose rates (Gy/ka) were calculated using the conversion factors of Liritzis et al. (Reference Liritzis, Stamoulis, Papachristodoulou and Ioannides2013). Dose rate calculations assumed radioactive equilibrium in the 238U and 232Th decay chains. Alpha efficiency values were determined on all samples with an 241Am alpha source using conventional SAR-based methods for either post-IR-OSL or TL measurements (e.g., Kreutzer et al., Reference Kreutzer, Schmidt, DeWitt and Fuchs2014). Alpha and beta dose attenuation factors were based on Brennan et al. (Reference Brennan, Lyons and Phillips1991) and Guérin et al. (Reference Guérin, Mercier, Nathan, Adamiec and Lefrais2012), respectively. The fractional gamma dose rate contribution to each sample from both the BCO itself and the surrounding sediments was calculated following Aitken (Reference Aitken1985, Appendix H). For samples taken within 30 cm of the surface, gamma dose rates were also adjusted using fractional corrections based on Aitken (Reference Aitken1985, Appendix H) to account for an incomplete gamma field. Alpha, beta, and gamma dose rates were adjusted for estimated moisture content of the sample using laboratory measured gravimetric moisture contents. Cosmic dose rates (Gy/ka) were calculated according to Prescott and Hutton (Reference Prescott and Hutton1994). Dose rate and final age calculations were made using DRAC (Durcan et al., Reference Durcan, King and Duller2015).
Results
Results of dating the late Quaternary alluvial deposits
The reported ages, as shown in Table 1, are in stratigraphic order and range from approximately 140 to 6600 years in age.
Depositional Unit A
The age of Unit A is provided by two single grain OSL ages obtained from Profile 1 in Trench 7 (Fig. 5). The deepest sample was collected about 40 cm below the base of the midden, around the same depth of a discrete prehistoric occupation approximately 1 m below the ground surface. This sample was analyzed using single grain method and returned a dominant finite mixture model age of 6600 ± 550 years (Shfd21007). A second sample was collected from near the base of the midden, at a depth of 64 cm and it returned an age of 2760 ± 200 years (Shfd21006). The base of Unit A was not exposed by the site excavations, and the initial age of sedimentation is presently unknown.
Depositional Unit B
The first terrace is draped by a thin veneer of overbank alluvium deposited by Petronila Creek, and a thicker inset of this deposit was exposed at the north end of Trench 7 (Fig. 5). This deposit is stratified where Unit B thickens beneath the scarp separating the first terrace from the floodplain and gradually thins and becomes increasingly massive away from the shoulder of the scarp, eventually losing all evidence of depositional bedding about 5 m to the south. Two OSL ages were collected from Unit B: one from a thick, stratified exposure (Profile 2), at the north end of trench 7, and another from the massive veneer facies at the south end of Trench 7 (Profile 1). The sample from Profile 2 was collected from a discrete sand bed (Zone 6) and at a depth of about 50 cm and returned an age of 140 ± 30 years (Shfd21008). The sample collected from the massive terrace veneer facies of Unit B returned a dominant finite mixture model age of 200 ± 30 years (Shfd21005). Considered together, these ages indicate that the upper part of unit B was deposited in the Historic period. The actual age of the base of Unit B is presently unknown but younger than the youngest sample in Unit A, which is 2760 ± 200 (Shfd21006) and the youngest dated burned clay object in Microartifact Column 1.
Results of dating the burned clay objects
The paired TL and post-IR-OSL ages for the BCOs (Table 3) are consistent within the reported 1σ uncertainties. However, the SAR-TL results were less precise than their post-IR-OSL counterparts (Table 3). Their narrow D e distributions (Fig. 7 and Supplementary Text) suggest that the BCOs were fired at sufficiently high temperatures to reset both TL and OSL signals. The reported ages are generally in stratigraphic order and range from approximately 2 to 5 ka.
Summary of BCO equivalent dose, dose rate, and age data.
a n is the number of D e determinations accepted after screening.
b D e values were determined using the Central Age Model (CAM) (Galbraith et al., Reference Galbraith, Roberts, Laslett, Yoshida and Olley1999). The reported error is the standard error.
c Radionuclide concentration measurement uncertainties are 10% for U, Th, and K. Beta and gamma dose rates were calculated using the conversion factors of Liritzis et al. (Reference Liritzis, Stamoulis, Papachristodoulou and Ioannides2013).
d Alpha dose rates were calculated using measured alpha efficiency values and the grain size attenuation factors of Brennan et al. (Reference Brennan, Lyons and Phillips1991).
e Beta dose rates were calculated using the grain size attenuation factors of Guérin et al. (Reference Guérin, Mercier and Adamiec2011).
f Cosmic dose rates (Gy/ka) were calculated according to Prescott and Hutton (Reference Prescott and Hutton1994).
g Luminescence ages, rounded to the nearest 10 years, are expressed as thousands of years before AD 2024. Error is 1 sigma.
Discussion
The results of this work can be assessed in several different ways. First is the direct comparison of the TL and OSL dates obtained from each BCO with respect to the radiocarbon age obtained from charcoal collected from the same level (Table 4; Fig. 8). The second way of assessing the results is whether the ages occur in correct stratigraphic order.
Contexts from where the dated BCOs were collected for this study. Soil Micromorphology sample M2 is shown in Figure 3c. The Microartifact columns are 50 cm × 50 cm units excavated in 5-cm levels that were wet sieved through a 1-mm mesh to allow comparison of the depth variation in small artifacts (microartifacts, < 2 mm) with large artifacts (6.25 mm). The dated BCOs were recovered from the microartifact samples.
Comparison of the burned clay object TL, OSL and paired 14C ages.
a Calibrated age range + 74 years (median probability + 74 years)
b 14C Age minus TL age in years
c 14C Age minus OSL age in years
Direct comparison of the BCO radiocarbon–luminescence pairs
The interpretive frame for this experiment is typical of archaeological work. Each set of dated items (specifically each burned clay object and matching charcoal) were derived from a single 5-cm level. Items recovered from a single level are generally assumed to be the same age, but there are various scenarios that can result in this assumption being incorrect. In every case the luminescence ages obtained from the burned clay objects are older than the radiocarbon age obtained on charcoal from the same level, with the difference ranging from as little as 144 years to as much as 1138 years for TL, or 364–1280 years for OSL. The average TL age is 716 ± 376 years older, and the average OSL age is 696 ± 313 years older.
The source of this age discrepancy is unclear but there are several possibilities. First, the association between the dated charcoal and the BCOs is relatively weak. Unlike Feathers (Reference Feathers2022) work at 41HR796, where each dated BCO was derived from what appeared to be a discrete single-use thermal feature, that is unlikely to be the case here. If this feature is the residue from earth oven cooking, the thermal refuse (the BCOs, charcoal, ash, and other food-related materials) would be scattered around the actual oven, and residues from different oven events comingled. If the sedimentation rate is slow, which appears to be the case, then there can be tens or hundreds of years between items adjacent to each other on the surface of the oven discard area. Second, the charcoal that was dated was not identified to species, nor were these items necessarily short-lived specimens (like seeds). It is also possible that the radiocarbon ages may result from an old wood effect, but this would, if anything, make the radiocarbon ages too old.
Aside from issues with the radiocarbon dates, the fact that every radiocarbon date is younger than the BCO luminescence ages may indicate other possible explanations. First, if the paleo-moisture content used in the luminescence age calculations, determined and in part predicated by conditions at the time of sampling, is higher than average for the last 3 ka, then the age calculations would be too old. The values used in the calculation of the BCO ages ranged between 13% and 19% and are similar to but lower than the moisture values determined by Sheffield for the sedimentary matrix SG-OSL ages, which were 16–27%.
The functional question is whether these values are reasonable for the Late Holocene. The weather during the excavation was quite wet and reflected in the site name (lodoso, Spanish for “muddy”), so it is possible that conditions during the excavation were wetter than the last 3 ka average. For these samples, a 1% decrease in estimated moisture content leads to approximately a 1% reduction in calculated age. To reconcile the OSL/TL and radiocarbon ages within 1σ uncertainty, an estimated moisture content decrease of 0–23% is required; this range narrows to 0–14% for agreement at the 2σ level. Examination of Figure 9 reveals that the age difference for samples in the upper 50 cm of the profile appear relatively similar, between 544 and 750 years (excluding the anomalously large difference between the TL and charcoal for sample CB005), and this pattern is not repeated by the sub-midden BCO dates. The midden is within a paleosol A-horizon formed at the top of Depositional Unit A (DU-A), which is buried by Depositional Unit B, and the presence of this horizon indicates a declining sedimentation rate towards the end of DU-A sedimentation. Hence it is likely that materials in the midden experienced a greater range of moisture variability owing to drying and wetting, being close to the ground surface.
Plot of the ages obtained during this study. Solid trend line is for the matrix single grain OSL dates. The dashed trend line is for the TL-BCO luminescence dates. Error bars are shown for the luminescence ages except where error bars were smaller than the icon. Errors are not shown for the radiocarbon ages because they are smaller than the icon. With the exception of CB003 and CB001, both of which have luminescence and radiocarbon ages that are out of stratigraphic order, the BCO luminescence ages demonstrate the midden is a time-transgressive deposit and not the product of a single burning event.
This introduces yet another possibility, which is that the sample depths within the midden do not reflect the depth these samples were at during formation of the midden. The SG-OSL matrix ages indicate that Depositional Unit B is less than 200 years old, which means that sample CB002, which was recovered from a depth of 27.5 cm, was within 5 cm of the ground surface as recently as 200 years ago and for as long as nearly 1800 years. Hence all the samples in the midden may have been exposed to greater cosmogenic dose than assumed based on the recovery depth. We note, however, that given the small contribution of the cosmogenic dose to the total dose rate, calculating ages based on linear versus instant overburden accumulation would only lower the calculated ages by less than 1%.
Stratigraphic interpretations
The shallowest BCO dated was recovered from the top of Depositional Unit B. The SG-OSL matrix ages consistently placed in the last 200 years (Shfd21005 and Shfd21008). The TL/OSL ages obtained from the BCO, and the charcoal recovered from that level returned ages between 2300 and 2940 years ago, which indicates that these materials are out of stratigraphic position. When the TL/OSL/14C ages for CB001 are viewed in light of the samples derived from the underlying midden, they suggest that these items most likely were derived from near the base of the midden, at or slightly below samples CB005. There was no evidence observed in the field that indicated these items were out of place other than Depositional Unit B, which appeared to entirely post-date the prehistoric occupations at the site.
The next four dated burned clay objects in this succession were recovered from the midden, and such deposits often exhibit complex taphonomic pathways (Schiffer, Reference Schiffer1987; Leach and Bousman, Reference Leach, Bousman, Nickels, Bousman, Leach and Cargill2001). In the field, several possible scenarios were entertained for the formation of the midden, ranging from the burned clay being attributable to a single heating event, such as a burned house, to a time-transgressive midden composed of cooking debris that had been created at different times over a long period. Open air earth oven cooking features such as burned rock middens, which are common in central and west Texas, often exhibit episodic use over thousands of years (Nickels et al., Reference Nickels, Leach, Cargill, McRae, Bousman, Nickels, Bousman, Leach and Cargill2001; Koenig and Miller, Reference Koenig and Miller2023), and radiocarbon ages obtained from BCOs at site 41AT232 (Turpin, Reference Turpin2011), one of the only other closely studied burned clay middens, spanned about 1500 years between 1847 to 3360 years cal BP (1921–3434 years before 2024).
All three dating methods (TL, OSL, and 14C) suggest that the midden here at 41NU114 was formed by separate, discrete burning events that spanned a period between approximately 800 and 1000 years, sometime approximately 1500–3000 years ago, which supports the diachronic midden hypothesis. Each dating method returned slightly different age ranges (e.g., 826 years (14C), 800 years (OSL), and 930 years (TL)), and absolute periods (e.g., 1480 and 2306 years ago (14C), 2050 to 2850 years ago (OSL), and 2010 to 2940 years ago (TL)). When dating charcoal alone, issues associated with the old wood effect could account for a significant part of this age range, but the luminescence dates derived directly from the burned clay objects side steps this issue and records the burning event that last heated each clay object, hence this approach offers a distinct interpretational advantage. When the paired BCO luminescence and radiocarbon ages are viewed in stratigraphic order (Fig. 9), three of the four appear to be in the correct age order, but sample CB003/ACE10475.1.1 appears to be out of order, yielding ages 460 years (14C), 330 years (TL), and 340 years (OSL) older than the underlying sample (CB004). This is most likely due to post-depositional disturbance.
The two stratigraphically discrete occupations beneath the midden appear to have occurred at least 1000 years apart and possibly separated in time by as much as 2000 years. The upper of the two occupations, at a depth of approximately 100.5 cm, yielded ages between 4336 and 4700 years. The relative concordance in all three ages suggests that this is a discrete occupation rather than a palimpsest. The lower occupation, at 140.5 cm, on the other hand, yielded a charcoal age that was 1130–1280 years younger than the BCO luminescence ages. There was no apparent pedogenic evidence indicative of this occupation being situated on a surface with a long period of time averaging, such as the presence of a buried soil, so the age discrepancy is surprising. But at this time the difference in age between the AMS and BCO luminescence ages could be due to this surface being a palimpsest.
Finally, when the results of the TL and post-IR-OSL results are analyzed by regression, the results (R 2 = 0.822) indicate that the age of the baked clay objects are in age order and reflect that the midden was formed over a period of about a millennium between 2,000 and 3,000 years ago (Fig. 9).
When all of the dating results are plotted together, the TL/post-IR-OSL ages indicate a relatively linear sedimentation rate, which is slightly out of sync with the sediment OSL dates (Fig. 9). The likely reason for this is that there are minor differences between the thickness of the Unit B veneer between the profile where the sediment dating samples were collected (Trench 7) and the two profiles from which the baked clay objects were collected, which were located off of Trench 6. Specifically, the veneer of Unit B, and the thickness of the midden in Trench 7 where three of the single grain OSL sediment samples were collected was slightly thinner than the profile where the baked clay objects were sampled from the midden in Trench 6. Alternatively, the age structure of Unit A in Trench 7 is slightly different for that in Trench 6.
Conclusions
The data presented here employs luminescence dates from two different labs on two different materials and the results are in good agreement. These results indicate that both TL and OSL dates obtained from baked clay objects can yield accurate ages for the creation or last use of the object when they have been heated sufficiently to reset the TL-OSL signal. The results obtained by Feathers (Reference Feathers1997) from the Dimond Knoll site indicate that not all BCOs from the same feature share the same thermal history and some may not be adequately reset, which is apparent from the data recovered from such samples during processing for luminescence dating. The results also suggest that the taphonomic processes operative in middens may have resulted in dating complications due to mixing by ancient inhabitants or fauna, as well as the comingling of artifacts and charcoal from different cooking events and periods. Comparison of the BCO luminescence dates with paired radiocarbon-dated charcoal revealed that the luminescence dates are all older than the charcoal, and the source of this discrepancy is presently unknown, but possibly an artifact of a drier than estimated paleo-moisture history.
In radiocarbon dating, to obtain dates relevant to a specific event, it is critical to date materials that can be tied to that event. For this reason, charred material recovered from archaeological features (in this case cooking installations created by people) is preferred. The luminescence dates obtained here directly date the creation/use of the primary artifacts of human activity at this site and should constitute best evidence of the timing of human activity at this site. Moving forward, additional work is needed to understand the differences observed in the radiocarbon and luminescence dating.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/qua.2025.10058.
Acknowledgments
The fieldwork reported here was done by AmaTerra Environmental, Inc. (now ERG), supervised by Rachel Feit and Amy Goldstein, under contract to the Texas Department of Transportation. Post-excavation work has been coordinated by Aaron Norment (Amaterra-ERG), in consultation with Eric Oksanen and Brittney Gregory. The manuscript was also improved by the comments of 2 anonymous reviewers and volume editor Paul Hanson.
Competing interests
The authors declare that we are unaware of any potential competing interests that could interfere with the objectivity or integrity of a publication.
Open access
