Introduction
Preserving and identifying intact Late Pleistocene and Early Holocene deposits in the southeastern United States remains one of the most persistent challenges in North American archaeology (Williams and Stoltman, Reference Williams, Stoltman, Wright and Frey1965; Dunnell, Reference Dunnell1990; Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999; Anderson et al., Reference Anderson, Smallwood and Miller2015; Miller et al., Reference Miller, Smallwood, Tune, Miller, Smallwood and Tune2022). The region is dominated by low-relief, deflated uplands and active fluvial systems, and as a result, most Late Pleistocene cultural materials occur as surface finds or mixed palimpsests lacking stratigraphic integrity (Dunnell, Reference Dunnell1990; Holliday and Miller, Reference Holliday, Miller, Waters, Goebel and Graf2013). As a result, the behavioral and chronological frameworks that define early human settlement across the South Atlantic Coastal Plain depend heavily on the few sites where depositional stability can be demonstrated rather than assumed. Demonstrating how and where intact cultural horizons survive is therefore a critical prerequisite for making inferences about the lifeways of people who first inhabited the region that has the densest concentration of early sites in North America (Anderson et al., Reference Anderson, Miller, Yerka, Gillam, Johanson, Anderson, Goodyear and Smallwood2010; Miller et al., Reference Miller, Smallwood, Tune, Miller, Smallwood and Tune2022).
With few deeply stratified rockshelters outside the Cumberland Plateau and Appalachian Highlands (Miller and Gingerich, Reference Miller and Gingerich2013) and a lack of loess sequences characteristic of other regions (Holliday, Reference Holliday1997; Mandel, Reference Mandel2008), geoarchaeological approaches to identifying site formation processes have become essential to evaluating archaeological site integrity in the American Southeast (Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999). Contemporary approaches to evaluating the impact of postdepositional processes rely extensively on combining analyses of artifact fabric, sedimentary structure, and spatial analysis to distinguish primary cultural horizons from secondary reworked deposits, particularly when absolute dating and stratigraphic markers are sparse. Historically, studies in the American Southeast that have applied multimethod approaches on Late Pleistocene and Early Holocene archaeological sites have focused on alluvial terraces (Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999), caves and rockshelters (Sherwood, Reference Sherwood2001; Sherwood et al., Reference Sherwood, Driskell, Randall and Meeks2004), and even sinkholes (Halligan et al., Reference Halligan, Waters, Perrotti, Owens, Feinberg, Bourne-Worster and Fenerty2016; Halligan, Reference Halligan2021). Yet few have done so in upland, non-fluvial settings. Consequently, for the Atlantic Plain, there is a dichotomy in landforms that are typically examined for Late Pleistocene and Early Holocene sites: deeply buried alluvium or surface scatters on upland deposits. However, we argue here that the range of landforms capable of preserving Late Pleistocene occupations across the Southeast remains poorly defined, and there are a variety of landforms not typically examined and overlooked in archaeological surveys that may yield preserved deposits that date to the Pleistocene. As a case study, the Allendale-Brier Creek Paleoindian quarry complex along the Savannah River has produced sites in a variety of landscape settings that contain datable archaeological deposits (Goodyear and Sain, Reference Goodyear, Sain, Goodyear and Moore2018; Moore et al., Reference Moore, Brooks, Daniel, Ivester, Feathers, Barbour, Goodyear and Moore2018; Fig. 1). Here, we discuss our analysis of one such site, the Swag site (38AL137), that contains preserved, Pleistocene-aged deposits in a novel setting—an upland basin pond or marsh.
Location of the Allendale-Brier Creek Complex in the American Southeast.
In 1979, Albert Goodyear and Tommy Charles conducted an extensive archaeological survey of Allendale County adjacent to the Savannah River to find the extent of “Coastal Plain chert” quarries that they attributed to the Oligocene Flint River Formation (Goodyear and Charles, Reference Goodyear and Charles1984). They identified 25 archaeological sites, including the Topper site (38AL23), the Big Pine Tree site (38AL143), and the Charles site (38AL135), all of which contain Late Pleistocene– and Early Holocene–aged deposits (Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999; Waters et al., Reference Waters, Forman, Stafford and Foss2009; Russell, Reference Russell2015). The Swag site (38AL137) was also recorded during the Goodyear and Charles (Reference Goodyear and Charles1984) survey and was initially identified as a quarry-related site based on observable chert boulders and a dense lithic scatter. The site lies within an upland depression underlain by fine-grained, poorly drained sediments. Preliminary testing revealed a discrete buried artifact concentration containing early-stage lithic reduction (Goodyear and Charles, Reference Goodyear and Charles1984). The site’s geomorphic context, distinct from the alluvial terrace and source-bordering dune settings typical of other Late Pleistocene and Early Holocene sites in the area, provided an opportunity to examine how intermittently saturated pond-margin environments influence archaeological preservation on the coastal plain. In other words, while most of the sites in the Allendale-Brier Creek Complex were found adjacent to major rivers or tributaries of the Savannah River (Goodyear, Reference Goodyear and Gingerich2018), or along the rims of Carolina Bays (Brooks et al., Reference Brooks, Taylor and Grant1996; Moore et al., Reference Moore, Brooks, Daniel, Ivester, Feathers, Barbour, Goodyear and Moore2018), the location of the buried deposits of the Swag site are novel: they are in the bottom of a basin that was once an upland pond or marsh.
Here, we combine sediment characterization, artifact fabric analysis (orientation and inclination), refit studies, and optically stimulated luminescence (OSL) dating to assess the depositional and postdepositional history of the Swag site. This integrated approach allows us to (1) determine whether the buried artifact horizon represents a minimally disturbed occupation surface; (2) identify the geomorphic and pedogenic processes that led to its formation; and (3) assess the broader potential of upland depressions as locations of stratigraphically intact Late Pleistocene archaeological deposits. Beyond the regional significance of documenting a buried fluted-point component consistent with Clovis technology, this study presents a methodological model for assessing context and preservation in humid, sandy settings where stratified hunter-gatherer sites are notoriously rare. More broadly, predictive models that guide archaeological surveys are only as good as their input (Verhagen and Whitley, Reference Verhagen and Whitley2012), and we argue that the Swag site provides a valuable data point to demonstrate that the interior and immediate periphery of upland basins have the potential to contain preserved archaeological sites that date to the Late Pleistocene and Early Holocene.
Background
The study area lies within the upper Atlantic Coastal Plain of southwestern South Carolina, a low-relief landscape underlain by a southeastward-thickening wedge of Cretaceous and Cenozoic sediments deposited during repeated marine transgressions and regressions (Cooke, Reference Cooke1936; Prowell, Reference Prowell1996). During the Quaternary, the Savannah River and its tributaries dissected this depositional surface (Leigh, Reference Leigh2008). During glacial periods, continental winds from the west and northwest were more arid, and when coupled with the braided river regimes, led to source-bordering dunes along the major rivers of the South Atlantic Coastal Plain. This wind-driven activity also helped create and reshape the shallow, oval upland depressions known as “Carolina Bays” (Brooks et al., Reference Brooks, Taylor and Grant1996; Ivester et al., Reference Ivester, Leigh and Godfrey-Smith2001). Consequently, Pleistocene-aged archaeological sites in the region are most often preserved within river terraces or source-bordering dune fields, where episodic burial offered some protection from surface processes (Moore et al., Reference Moore, Brooks, Daniel, Ivester, Feathers, Barbour, Goodyear and Moore2018). The overwhelming majority of the Late Pleistocene and Early Holocene archaeological record, however, consists of surface finds recovered from deflated or minimally aggrading upland settings, where shallow burial and reworking have stripped away primary depositional context (Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999; Anderson et al., Reference Anderson, Smallwood and Miller2015).
The Allendale-Brier Creek Complex is well known for its abundance of high-quality lithic raw material (Goodyear, Reference Goodyear and Gingerich2018; Saunders, Reference Saunders2023). Quarry and workshop sites such as Topper (38AL23) and Big Pine Tree (38AL228) have produced large Late Pleistocene and Early Holocene assemblages, yet most occur in slope or terrace settings where colluvial movement and pedogenic mixing complicate stratigraphic interpretation (Waters et al., Reference Waters, Forman, Stafford and Foss2009; Miller, Reference Miller2010; King, Reference King2012; Weidman, Reference Weidman2013; Russell, Reference Russell2015; Sain, Reference Sain2015; Smallwood, Reference Smallwood2015). Upchurch (Reference Upchurch, Goodyear and Charles1984) identified the lithic raw material Oligocene-aged chert outcrops of the Flint River Formation. However, while the “Flint River Formation” (Cooke, Reference Cooke1936) is frequently used in the archaeological literature, the name has been abandoned in modern geologic usage, as Huddlestun (Reference Huddlestun1993) demonstrated that the original unit comprised an unrelated assemblage of residual cherts and siliceous sediments of differing ages and origins. The formation name does not appear in Prowell’s (Reference Prowell1996) nomenclature for South Carolina or in Huddlestun and Summerour’s (Reference Huddlestun and Summerour1996) for Georgia. Consequently, the age and origin of the lithic material of the Allendale-Brier Creek Complex is unresolved.
The Swag site occupies a bowl-shaped basin in the uplands adjacent to the Savannah River (Fig. 2). The area is mapped to the Late Eocene Barnwell Group, a coarsening-upward succession that reflects shallow-marine deposition (Cooke, Reference Cooke1936; Huddlestun and Hetrick, Reference Huddlestun and Hetrick1979, Reference Huddlestun and Hetrick1986; Prowell, Reference Prowell1996). Prowell (Reference Prowell1996) identified an “Upland Unit” at the highest interfluves, which rests unconformably on the underlying Barnwell Group (Prowell, Reference Prowell1996). The Upland Unit has not been definitively dated, and could date to the Oligocene, Miocene, or Pliocene (Prowell, Reference Prowell1996, p. 4). The basin is drained by an outlet channel connecting the basin to an oxbow of the Savannah River that formed during the last glacial maximum (Goodyear and Sain, Reference Goodyear, Sain, Goodyear and Moore2018; Goodyear and Brooks, Reference Goodyear and Brooks2019). The outlet channel exposes a chert outcrop at 36 m above sea level (m asl), tentatively attributed to the Clinchfield Formation (Utley Limestone Member; Huddlestun and Hetrick, Reference Huddlestun and Hetrick1986; Prowell, Reference Prowell1996).
Topographic map of the Swag site (South Atlantic Coastal Plain, USA) with test unit locations. Cross section for Figure 3 is marked A–A′.
Figure 2 Long description
Topographic map of the Swag site on the South Atlantic Coastal Plain, USA. The map features contour lines indicating elevation changes. Key elements include a chert outcrop, an outlet channel and a Pleistocene oxbow. Test unit locations are marked as 2015-TU7, 2015-TU8, 2015-TU9, 2015-TU10 and 2017-TU1. A dashed line labeled A–A′ indicates a cross-section. The map includes a scale bar showing distances of 0, 100 and 200 meters and a north arrow for orientation. The site boundary is outlined and the map highlights the relationship between the chert outcrop and the outlet channel.
We mapped the sediments across the drainage at nine locations derived from eight bucket auger cores (P5-P12) and an excavation unit (2015-TU8) (Fig. 3, Supplementary Table S1). At the center of the basin, near 40 m asl, the sediments match the “brick-red to reddish brown residuum” of the Tobacco Road Sand from the Barnwell Group described by Huddlestun and Summerour (Reference Huddlestun and Summerour1996, p. 66), although here the sediments are gleyed with display features produced by a water table perched at the base of the basin. The facies adjacent to the basin are not gleyed and conform more closely Huddlestun and Summerour’s description. Around the basin periphery, the sediments are consistent with the Upland Unit of Prowell (Reference Prowell1996), resting unconformably on the Tobacco Road Sand. Colluvial sands from the Upland Unit have been redeposited into the center of the basin. On the periphery of the basin, the soils are consistent with the Blanton and Bonneau series soils (Paleudults), which typically contain a surface A horizon overlying an argillic horizon (Rogers, Reference Rogers1990; Soil Survey Staff, 2025a, 2025b). While the interior of the basin is mapped as Cowarts series soils (Soil Survey Staff, 2025c), the sediments are more consistent with Rutledge series (Humaquepts) that consist of a surface A horizon overlying gleyed sediments (Soil Survey Staff, 2025d).
Elevation profile and geomorphic cross section for the Swag basin (South Atlantic Coastal Plain, USA) illustrating the locations of bucket auger and test unit profiles.
Figure 3 Long description
The elevation profile and geomorphic cross section of the Swag basin illustrate the sediment layers and soil series across the drainage. The top section shows a profile from point A to A prime, with elevations marked at 53 meters above sea level at point P10, 36 meters above sea level at point P8 and 40 meters above sea level at point P6. The profile includes points P5 through P12, with a chert outcrop in the outlet channel labeled as Utley Formation Limestone. The horizontal scale is marked in meters with vertical exaggeration of five times. Below the profile, the cross section displays various soil series and sediment layers. On the left, the Bonneau Series is characterized by sandy, siliceous, subactive thermic Arenic Paleudults, with layers labeled A and 2Bt. Adjacent to it, the Blanton Series consists of loamy, siliceous, semiactive thermic Grossarenic Paleudults, with layers labeled Bw1 and Bw2. The Rutledge Series, located centrally, includes loamy, siliceous, semiactive thermic Fluvaquentic Endoaquepts, with layers labeled A, Bg and Cg. Wetland soils are noted as gleyed and redoximorphic, with the Upland Unit described as fine marine sands or redeposited colluvium. The Tobacco Road Formation is indicated at the base. On the right, another section of the Bonneau Series is shown, with layers labeled A1, Ab2, Ab3, Bw2 and Bw3. The diagram uses symbols to represent sand, loamy sand and sandy clay loam, with a vertical scale in centimeters ranging from 0 to 160.
The closest well-dated fossil pollen locality is White Pond near Columbia, South Carolina, approximately 160 km northeast of the study area (Watts, Reference Watts1980; Krause et al., Reference Krause, Russell, Zhang, Williams and Jackson2019). Krause et al. (Reference Krause, Russell, Zhang, Williams and Jackson2019) found that vegetation from ∼30,000–19,700 cal yr BP was dominated by open Pinus–Picea forest (70–95% Pinus) under cold, dry conditions, with a brief decline in Quercus around 26,000 cal yr BP coinciding with Heinrich event H2. Beginning around 19,700–16,300 cal yr BP, Quercus and Carya gradually increased as Pinus began declining, and this hardwood expansion accelerated between 16,300 and 10,400 cal yr BP, with Picea peaking and then collapsing, and mesic hardwoods (Ostrya/Carpinus, Fagus, Ulmus, Fraxinus) rising to their highest levels. By 12,000–10,400 cal yr BP, Quercus dominated (up to 59%), while Pinus and Picea reached minimum values. Krause et al. (Reference Krause, Russell, Zhang, Williams and Jackson2019) argued that the onset of the Holocene at 10,400 cal yr BP brought a steplike Pinus resurgence (25–45%) alongside continued high Quercus and declining mesic hardwoods. After 6400 cal yr BP, Pinus increased further (55–80%) at the expense of Quercus (dropping to 10–25%), establishing the pine-dominated landscape that persists to the present. Within the study area, the basin and surrounding land-holdings managed by the Archroma Corporation are operated as a commercial Pinus plantation that has undergone multiple harvest rotations over the past three decades.
In 2015, we initiated a 102 acre survey to relocate and delineate quarry-related sites that were identified by Goodyear and Charles (Reference Goodyear and Charles1984). The survey resulted in 459 shovel tests (>75% positive for cultural materials) and documented a near-continuous scatter of lithic debris across the survey area. The survey area encompassed the Swag site, which extends from the chert cobbles at the outlet channel, and was initially mapped by Goodyear and Charles (Reference Goodyear and Charles1984) based on a limited number of bucket auger tests and flakes visible in the adjacent road. We placed shovel tests around the periphery of the mapped site boundary, including the eastern edge of the site boundary near the center of the basin. After identifying buried deposits beyond the mapped site boundary, we excavated two test units adjacent to the Swag site. The first (2015-TU8) produced overshot flakes and the second (2015-TU7) produced a distal preform with overshot flaking consistent with Clovis lithic technology (Bradley et al., Reference Bradley, Collins, Hemmings, Shoberg and Lohse2010; Smallwood, Reference Smallwood2012; Supplementary Fig. S1).
In 2016, we expanded the 2015-TU8 test unit into a 12 m2 excavation block in the basin floor (Fig. 4). The upper sediments consist of dark grayish brown (10YR 3/2) sand that ends abruptly at 25 cm below ground surface, transitioning to gray (2.5Y 6/1) sand below (Fig. 5). At 65 cm, a clear boundary marks a shift to light brownish gray (10YR 6/2) sandy clay loam with weak subangular blocky structure and strong brown (7.5YR 5/6) masses of iron accumulation.Footnote 1
Topographic map illustrating excavation unit locations indicating the two excavation units (N500E399 and N501E399) analyzed in this study (Swag site, South Atlantic Coastal Plain, USA).
Figure 4 Long description
Topographic map illustrating the Swag site in the South Atlantic Coastal Plain, USA. The map highlights two excavation units, N500E399 and N501E399, within a study area. Contour lines indicate elevation changes, with a chert outcrop and an outlet channel marked. A dirt road is visible, providing access to the site. Insets show detailed layouts of excavation units 2015-TU8 and 2017-TU7, with the study area clearly marked. A scale bar indicates distances and a north arrow provides orientation.
Soil profile from the north wall of test unit 2015-TU8 (Swag site, South Atlantic Coastal Plain, USA).
We interpret these sediments as colluvial deposits from the Upland Unit that washed into the basin center, burying the Tobacco Road Formation sands. A perched water table subsequently developed within these sands, driving the formation of the observed redoximorphic features. Lithic artifacts recovered from the basin interior block were located within the gray sands and included a fluted preform base (Art# N501E399-B(84); Fig. 6), early-stage reduction and modified flake debitage, expedient tools, early-stage bifaces, scrapers/planers, adze-like bifaces, and modified prismatic blades, including some with denticulated edges (Supplementary Fig. S2).
Fluted preform base (N501E399-B(84); Swag site, South Atlantic Coastal Plain, USA).
In 2017, we placed a second test unit (2017-TU1) on the basin margin approximately 72 m southwest of the excavation block, following a bucket auger core (2015-P8) that recovered lithic flakes from 40–80 cm below ground surface. The upper sediments at 2017-TU1 consist of a surface A horizon (0–28 cm) of dark grayish brown sand (10YR 3/2), below which yellowish-brown (10YR 6/4) sand grades to light yellowish brown (10YR 5/4). At 90 cm, a clear boundary marks the transition to a light brownish gray (10YR 6/2) loamy sand with weak subangular blocky structure and strong brown (7.5YR 5/6) masses of iron accumulation. These sediments are consistent with colluvial sands from the Upland Unit unconformably redeposited on the Tobacco Road Formation along the toe slope of the basin. Test unit 2017-TU1 produced a Taylor Side-Notch point 50 cm below surface, stratigraphically overlying a deeper horizon of lithic artifacts (Supplementary Fig. S3). Taylor points are characteristic of the Early Archaic period in the South Atlantic Coastal Plain and have been dated to 11,690–10,240 cal yr BP (Coe, Reference Coe1964; Daniel, Reference Daniel2021; Thulman, Reference Thulman, Miller, Smallwood and Tune2022). The stratigraphic relationship documented in 2017-TU1 (an Early Archaic Taylor point overlying a deeper lithic horizon) indicates that multiple cultural components are preserved along the southern pond margin. In 2018 and 2019, we expanded excavations on the basin margin to excavate a 14 m2 excavation block and continued surveying the area surrounding the Swag site.
Here, we focus our analysis on units N500E399 and N501E399 from the 2015-TU8 excavation block in the basin floor. These units were selected because they appear to contain a single, dense horizon of lithic debris that contains the fluted preform base (Art# N501E399-B(84)). These two excavation units are notable because they are the two densest excavation units recorded in the entire history of excavations (∼700 m2 excavated) at the Allendale-Brier Creek Complex (Goodyear, A., personal communication, 2026). Yet their location is more than 200 m away and located at a higher elevation than the closest available outcrop of chert; thus, all chert here is necessarily anthropogenic.
Archaeological deposits are rarely pristine snapshots of past behavior (Shott, Reference Shott2006). Artifacts are subjected to spatial distortions caused by both cultural and natural processes that alter their position, quantity, and condition from the moment of deposition onward (Schiffer, Reference Schiffer1983, Reference Schiffer1988). Schiffer (Reference Schiffer1995) distinguished between cultural behaviors (C-transforms) such as trampling, refuse disposal, or modern plowing, and natural processes (N-transforms) like bioturbation, fluvial movement, pedoturbation, and aeolian activity. As examples for how to untangle these processes to make inferences about human behavior, Surovell et al. (Reference Surovell, Waguespack, Mayer, Kornfeld and Frison2005) and Surovell (Reference Surovell2022) used strike, dip, artifact conjoins, and spatial analysis at Barger Gulch Locality B to reconstruct postdepositional disturbances to a Folsom component. Elsewhere in eastern North America, Gingerich (Reference Gingerich2007, Reference Gingerich, Clark and Gingerich2022) used this same methodology to identify a possible structure and activity areas at the Shawnee-Minisink Clovis site. Similarly, Miller (Reference Miller2010) and Smallwood (Reference Smallwood2015) examined formation processes at the Topper site using spatial analysis, artifact density, and refit distributions. Smallwood identified a vertically discrete Clovis workshop floor with knapping loci that showed minimal horizontal displacement, while Miller found that preservation was differentially distributed across the excavation block, with some areas showing disturbance. More recently, Jennings et al. (Reference Jennings, Anderson, Smallwood, Plummer, Miller, Anderson and Goodyear2025) used these methods to identify evidence for the oldest structure in the American Southeast at the Topper site.
Unlike the landforms dominating the Topper site, the Swag landform represents a pond-margin basin that accumulated colluvial and minor aeolian sediments under low-energy conditions during the Late Pleistocene. Such settings are rarely examined for Late Pleistocene and Early Holocene archaeological research, yet they may provide some of the most stable sedimentary environments on the South Atlantic Coastal Plain (Goodyear, Reference Goodyear, Bonnichsen and Turnmire1999; Moore et al., Reference Moore, Brooks, Daniel, Ivester, Feathers, Barbour, Goodyear and Moore2018). However, there is evidence for multiple occupations at the margins of the basin and contemporary ground disturbance from agriculture, road construction, and land modification to drain the basin could have redeposited and reburied artifacts. By integrating sedimentological, geochronological, and artifact-fabric analyses, we argued that the study area within the Swag site represents a minimally disturbed archaeological deposit. Moreover, this research demonstrates how similar upland depressions may contribute to the identification of well-preserved Late Pleistocene and Early Holocene archaeological sites across the region.
Methods
During excavations, we screened the surface A-horizon sediments through 6.35 mm (¼-inch) wire mesh. All sediments below the A horizon were excavated in 5 cm arbitrary levels and screened through 3.175 mm (⅛-inch) wire mesh. All artifacts measuring 2 cm or larger were piece-plotted in three dimensions. The dip (inclination) and strike (orientation) were recorded for each piece-plotted artifact using an inclinometer and compass. We expect undisturbed deposits to have a random distribution of strike measurements, whereas a fluvial process, such as winnowing, would orient artifacts to the direction of the water flow (Bertran and Texier, Reference Bertran and Texier1995; Byers et al., Reference Byers, Hargiss and Finley2015). Conversely, we expected the dip measurements to be nonrandom, because artifacts discarded on a flat surface should remain flat if postdepositional processes have not disturbed them (Surovell et al., Reference Surovell, Waguespack, Mayer, Kornfeld and Frison2005). Results were visualized with rose diagrams and dip histograms to evaluate the artifact fabric across the artifact horizon. We used Pearson’s goodness-of-fit test to assess statistical significance (Plackett, Reference Plackett1983).
We conducted a refit analysis for all artifacts 2 cm or larger to test for vertical or horizontal displacement. A total of 13 participants worked on the refit study for a combined 70 hours, at which point we reached a point of diminishing returns in identifying new conjoins consistent with Laughlin and Kelly (Reference Laughlin and Kelly2010). Vertical offsets greater than a few centimeters would indicate downward movement through bioturbation or drift, whereas close refit clusters suggest artifacts remained near their original depositional positions. Artifact counts were also tabulated by depth to produce vertical density profiles for each unit, allowing comparison of artifact concentrations with sedimentary boundaries and assessment of whether secondary peaks reflected localized disturbances such as root penetration or krotovina. Chronological control for the deposits was established through OSL dating of quartz grains collected from freshly cleaned profiles on the west wall of excavation unit N501E399 from the 2015-TU8 excavation block. Samples were extracted from just above and below the primary artifact concentration using opaque PVC tubes to prevent light exposure. The resulting ages provide maximum and minimum limits on sediment deposition and, by extension, on the age and stability of the artifact horizon. Spatial, sedimentary, and chronometric datasets were integrated using R (R Core Team 2025) and QGIS (QGIS Development Team, 2024) to compare artifact fabric, refit, and OSL data in three dimensions. The spatial data and R code used to conduct this analysis are available via GitHub (https://github.com/archaeoshane/Shores_etal_QR).
Results
A total of 1337 artifacts larger than 2 cm were piece-plotted in units N500E399 and N501E399. Vertical artifact densities display a single, well-defined concentration approximately 6 cm thick in both units (Fig. 7A). Artifact counts begin to increase at roughly 99.50 m elevation, peak near 99.44 m, and decline sharply below 99.38 m. This discrete peak corresponds to the same stratigraphic horizon in both units, indicating lateral continuity of the deposit across the excavation block within the gleyed, subsurface soil horizon. A minor, secondary density increase was detected approximately 10 cm higher in unit N501E399. This smaller secondary peak is associated with a krotovina infilled with lighter-colored sediment and likely represents a tree-throw or animal burrow rather than a second cultural occupation.
(A) Vertical density profiles for units N500E399 and N501E399. (B) Distribution of strike (orientation) measurements. (C) Distribution of dip (inclination) measurements. Swag site (South Atlantic Coastal Plain, USA).
Figure 7 Long description
A showing two horizontal bar charts. Left chart title: N500E399. The vertical axis label is Elevation (m) with tick labels 99.4, 99.6, 99.8. The horizontal axis label is n with tick labels 0, 50, 100, 150. Bars are shown at multiple elevation levels between about 99.4 and 99.5. Right chart title: N501E399. The vertical axis label is Elevation (m) with tick labels 99.4, 99.6, 99.8. The horizontal axis label is n with tick labels 0, 20, 40, 60. Two callouts label Secondary Peak and Primary Peak. B showing a circular rose chart with compass labels N, NE, E, SE, S, SW, W, NW around the circle. Wedge sectors radiate from the center. C showing a vertical bar chart. The horizontal axis label is Dip with category labels 0 to 15, 16 to 30, 31 to 45, 46 to 60, 61 to 75, 76 to 90. The vertical axis label is n with tick labels 0, 100, 200, 300. Bar heights are approximately: 0 to 15 near 300; 16 to 30 near 300; 31 to 45 near 230; 46 to 60 near 100; 61 to 75 near 50; 76 to 90 near 70.
For the artifact strike analysis, we failed to reject the null hypothesis (χ2 = 10.95; df = 5; P = 0.052; Fig. 7B). Although the results were not statistically significant, we cautiously interpret the orientations as broadly consistent with a random distribution. The slightly elevated counts in the 0–29° and 30–59° bins may reflect minor alignment with the landform’s general north–south slope. Alternatively, they may stem from measurement error, with the recorder inadvertently conforming to the site grid or rounding to a cardinal direction similar to what Bertran and Texier (Reference Bertran and Texier1995, p. 527) observed at Pincevent. Either way, there is no clear evidence of strong preferred orientation. Inclination data are strongly nonrandom (χ2 = 390.6; df = 5; P < 0.0001; Fig. 7C), with the 0-15° and 16-30° bins significantly overrepresented. The dominance of nearly horizontal dip values indicates that artifacts rest in stable, flat positions consistent with primary deposition on a low-gradient surface rather than postdepositional tilting or burial compaction.
We identified 40 conjoining artifact pairs distributed throughout the artifact horizon. Vertical distributions of refits averaged 3.2 cm, with a median of 2.1 cm, while horizontal separations averaged 48 cm (Fig. 8). Most refits occurred within a single stratigraphic level or within one contiguous 5 cm increment, suggesting that individual reduction events remained spatially coherent after burial. A small subset of refits linking the primary and weak secondary density peaks indicates limited vertical displacement by localized bioturbation but not wholesale mixing. The overall pattern of tight vertical clustering and short vertical refit distances supports the interpretation that the artifact horizon represents an in situ depositional surface.
Planview (A) and profile (B) of units N500E399 and N501E399 (Swag site, South Atlantic Coastal Plain, USA) showing the location of plotted artifacts, refit pairs, and optically stimulated luminescence (OSL) dates.
Figure 8 Long description
Panel A, titled ′Planview,′ displays a scatter plot with the horizontal axis labeled ′Easting′ ranging from 399.00 to 400.00 and the vertical axis labeled ′Northing′ ranging from 500.0 to 502.0. The plot features a dense cloud of gray points representing artifact locations, overlaid with black lines connecting specific points. Open circles mark significant locations, including ′OSL-1 and OSL-2′ and ′N501E399-B(84) Fluted Preform.′ The connected lines suggest spatial relationships among artifacts, with a concentration of points forming a coherent pattern. Panel B, titled ′Profile View,′ presents a scatter plot with the horizontal axis labeled ′Northing′ ranging from 500.0 to 502.0 and the vertical axis labeled ′Elevation′ ranging from 99.4 to 99.8. This plot shows a dense distribution of gray points, with black lines connecting specific points, indicating elevation changes across the northing. Notable locations are labeled with ′OSL-1 10,435 ± 715′ and ′OSL-2 16,910 ± 1,145,′ along with ′N501E399-B(84) Fluted Preform.′ The plot illustrates how elevation varies in relation to artifact distribution, with a clear pattern of connected points suggesting a spatial coherence. There is no legend provided and the visual encoding differentiates artifact locations and significant markers.
The lithic assemblage from units N500E399 and N501E399 is dominated by early-stage and core reduction debitage. Of the 1337 piece-plotted artifacts (≥2 cm), approximately 65% have cortex present and consist of flakes, flake fragments, biface thinning flakes, and retouched or expedient tools. The predominance of early-stage reduction debris, combined with few expedient and formal tools, is consistent with a short-term occupation focused on tool production, maintenance, and limited on-site use.
Two OSL samples bracket the artifact-bearing horizon (Table 1). The lower sample (BG4681) yielded an age of 16,910 ± 1145 years ago, while the upper sample (BG4680) returned an age of 10,435 ± 715 years ago. Together, these results indicate that sediment accumulation at the site occurred throughout the Pleistocene–Holocene transition. With only two OSL samples, these ages provide a broad temporal bracket rather than a detailed time–depth model. We cannot rule out short-term hiatuses in deposition within this interval; however, there is no sedimentological evidence for erosional unconformities or high-energy reworking within the dated profile. We emphasize that the Clovis affiliation of the Swag assemblage rests primarily on lithic typology (a fluted preform base consistent with regional Clovis technology) rather than on high-precision chronometry, although the age range is broadly consistent with the appearance of Clovis technology (Waters et al., Reference Waters, Stafford and Carlson2020). The OSL ages provide a broad Late Pleistocene–Early Holocene bracket, within which the occupation is interpreted as Clovis age, but cannot by themselves pinpoint the assemblage to a narrow interval within that range.
Optically stimulated luminescence (OSL) ages on quartz grains from the Swag site (South Atlantic Coastal Plain, USA).
Table 1 Long description
Quartz OSL results for two sediment samples from the Swag site report depth, lab ID, aliquots analyzed, grain size, equivalent dose, overdispersion, radionuclide content, water content, cosmic and total dose rates, and resulting ages. Sample OSL 1 at 0.41 m (BG4680) used 38 of 40 aliquots of 250 to 150 micrometer grains and yields an age of 10,435 years plus or minus 715. Sample OSL 2 at 0.52 m (BG4681) used 38 of 40 aliquots of the same grain size and yields an older age of 16,910 years plus or minus 1,145. Equivalent dose rises from 11.58 Gy to 26.49 Gy with depth, while total dose rate increases from 1.11 to 1.57 milligray per year; cosmic dose rate is the same for both at 0.20 milligray per year. Overdispersion is similar for both samples at about 22 percent, suggesting comparable spread in equivalent dose values. Uranium, thorium, and potassium concentrations are higher in the deeper sample, consistent with its higher dose rate. Ages are reported with stated uncertainties and should be interpreted in light of the measurement and model choices used to derive equivalent dose and dose rate.
a Aliquots used in equivalent dose calculations vs. original aliquots measured.
b Equivalent dose calculated on a pure quartz fraction with about 40–100 grains/aliquot and analyzed under blue-light excitation (470 ± 20 nm) using single-aliquot regeneration protocols (Murray and Wintle, Reference Murray and Wintle2003). The central age model of Galbraith et al. (Reference Galbraith, Roberts, Laslett, Yoshida and Olley1999) was used when overdispersion was <25%.
c Values reflect precision beyond instrumental errors; values ≤ 25% (1σ) indicate low dispersion in equivalent dose values and unimodal distribution.
d U, Th, and K content analyzed by inductively coupled plasma mass spectrometry by ALS Laboratories, Reno, NV; U content includes Rb equivalent. Cosmic dose rate from Prescott and Hutton (Reference Prescott and Hutton1994).
e Systematic and random errors calculated in quadrature at 1σ. Datum year is 2010 CE.
Taken collectively, the sedimentological observations, artifact fabric, refit data, and OSL chronology point to a discrete, minimally disturbed occupation surface preserved within a low-energy pond-margin environment. The gleyed, fine-textured sediments appear to have restricted vertical migration and protected the deposit from extensive pedoturbation. Localized disturbances such as root burning or small burrows are identifiable but limited in scale and effect. Thus, the Swag site preserves a rare example of a Late Pleistocene lithic assemblage largely in its original depositional position, one of the few such contexts documented to date in the American Southeast.
The single-component artifact horizon documented in units N500E399 and N501E399 contrasts with the stratigraphy observed in 2017-TU1, where a Taylor Side-Notched point was recovered above a deeper lithic horizon. This indicates that while the analyzed block preserves a discrete Late Pleistocene occupation surface, other areas of the upland basin contain multiple superimposed components. These findings highlight the potential for additional stratified cultural deposits elsewhere around the basin.
Discussion
The stratigraphic and sedimentological data from Swag demonstrate that intact Late Pleistocene cultural deposits can occur within upland depressions. The gleyed matrix and low-energy depositional context indicate a hydrologic regime distinct from the terrace and slope settings that dominate the regional Late Pleistocene and Early Holocene record. In other words, rather than forming within high-energy alluvial fills or source-bordering dunes, the Swag sediments accumulated gradually in a perched, poorly drained depression where seasonal saturation inhibited artifact migration and disturbance. Within the basin floor units (N500E399-N501E399), this low-energy, water-saturated environment sealed a single occupation surface, even though other parts of the basin margin (e.g., 2017-TU1) preserve multiple stratigraphically separated components. We suggest the artifacts in the basin floor reflect Clovis flintknapping directly on a frozen pond with discrete knapping episodes occurring on the ice, which subsequently melted and deposited the artifacts into the center of the pond. Our reasoning is that the extreme density of lithic materials at the time of the site’s occupation would have made walking, sitting, sleeping, and most domestic tasks nearly impossible Alternatively, a period of aridity resulted in the shallow pond drying up and allowing Clovis groups to occupy the basin floor. A post-Clovis increase in moisture may have consistently inundated the basin, resulting in subsequent populations only occupying the basin margin during the Early Archaic (Taylor), which may explain differences in occupation episodes in and around the upland basin.
Our finding here contrasts with the colluvial and alluvial dynamics at nearby sites like Topper and Big Pine Tree, where Late Pleistocene and Early Holocene occupations were preserved through repeated burial by slopewash and fine-sand aggradation along the Savannah River terraces (Waters et al., Reference Waters, Forman, Stafford and Foss2009). Those terraces relied on episodic sediment inputs for rapid burial, whereas Swag achieved preservation through the opposite mechanism: persistent saturation rather than periodic burial. This contrast underscores the diversity of site-formation processes across the Atlantic Coastal Plain, which vary with local geomorphic context and produce patchy archaeological visibility (Daniel et al., Reference Daniel, Goodyear, Moore, Thulman and Wilkinson2025). Elsewhere, aeolian sand ridges preserve stratified Paleoindian and Early Archaic assemblages through windblown sand accumulation during Late Pleistocene drought intervals (Daniel et al., Reference Daniel, Moore and Caynor2013), while pond-margin aggradation, as at Swag, represents a less-recognized but equally effective preservation mechanism. Test unit 2017-TU1 further shows that colluvial sedimentation around the basin was continuous enough to stack cultural horizons without mixing, reinforcing the interpretation that the Swag pond margin acted as a long-term sediment trap archiving successive occupations across the Pleistocene–Holocene transition.
The recent identification of a probable Clovis hearth and structure at Topper’s upland locality provides direct evidence of domestic organization and short-term habitation (Anderson et al., Reference Anderson, Smallwood, Goodyear and Walters2016; Jennings et al., Reference Jennings, Anderson, Smallwood, Plummer, Miller, Anderson and Goodyear2025). Swag currently lacks architectural features but mirrors this scale of occupation, containing dense, localized knapping debris within a confined area consistent with brief residential or task-specific use. Its preservation in a hydrologically stable setting suggests similar small campsites and/or activity areas may remain undetected elsewhere on the Coastal Plain, particularly within Carolina Bays, especially those where archaeological deposits have been identified on interfluves and source-bordering dunes. In other words, if archaeological sites have been located on the margins of the basin, then it is worth examining deposits within the basin for preserved archaeological deposits as well.
These results also speak to a broader sampling bias: many shallow, sandy sites long dismissed as mixed or undifferentiated colluvial or aeolian deposits in fact contain stratified Late Pleistocene and Holocene sequences (Michie, Reference Michie1990; Miller, Reference Miller2010; Daniel et al., Reference Daniel, Moore and Caynor2013; Moore et al., Reference Moore, Brooks, Daniel, Ivester, Feathers, Barbour, Goodyear and Moore2018). More importantly, the archaeological deposits at the Swag site add another dimension to this bias by demonstrating that the interior of hydrologically restricted upland depressions can preserve archaeological deposits. In sum, the Swag site complements the terrace-based landscape-use models from Topper and Big Pine Tree, indicating that preservation of early archaeological horizons depends less on elevation or proximity to major drainages than on localized sediment dynamics and hydrologic regime.
Conclusion
Geoarchaeological evidence from the Swag site demonstrates that upland depressions on the South Atlantic Coastal Plain can preserve Late Pleistocene and Early Holocene deposits under conditions markedly different from those at better-known alluvial terraces and source-bordering dune deposits in the Atlantic Coastal Plain. At the Swag site, fine-grained, gleyed sediments in the center of the basin record persistent saturation with low-energy deposition, producing a stable matrix that limited artifact displacement. Artifact fabric and refit data confirm relatively intact cultural materials, while OSL ages bracket the deposits within the Pleistocene–Holocene transition, consistent with the broader regional record.
Viewed alongside the alluvial–colluvial sequences at Topper and Big Pine Tree and the aeolian sand-ridge deposits around Carolina Bays, the Swag site adds to the range of geomorphic processes capable of preserving Late Pleistocene surfaces across the Atlantic Coastal Plain. Early site preservation in the American Southeast should therefore be evaluated through sedimentological, artifact-fabric, and chronometric data together, and this study offers a replicable framework for identifying intact Pleistocene-aged components in geomorphically subtle settings. Even the interior of small, hydrologically stable upland basins can yield high-resolution records of human activity within the dynamic landscapes.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/qua.2026.10109.
Acknowledgments
We would like to thank the Archroma Corporation for supporting archaeological research and Al Goodyear for thoughtful conversation and good barbecue. We would also like to thank Chris Moore for coming out to help us sample for OSL dating. We could not have done this project without Ashley Smallwood, Tom Jennings, Bill Covington, Joan and Ernie Plummer, Wes Muckenfuss, Leslie Page, DuVal Lawrence, Carol Reed, Judy Kendall, KC Jones, and the legion of field school students and volunteers from Mississippi State University, the University of West Georgia, and the University of Tennessee. We collectively adore Jenalee Muse for providing a “home away from home” when her husband, David, would invite all of his archaeology friends over. Finally, we would like to thank two anonymous reviewers for their suggestions that substantially improved this article.
Funding statement
Funding for field and lab work was provided by the Topper Fund at Mississippi State University, and we are grateful to all of the donors who have contributed.
Open access
