Introduction and Objectives
In July 1996 a disarticulated skeleton was found in Kennewick, WA within a few meters of shore in Lake Wallula at Columbia Park (Chatters 1997a). A single 14C measurement on a total amino acid extract from part of the fifth metacarpal bone of the skeleton yielded a date of 8410±60 yr. B.P.1 (UCR3476/CAMS-29578) (Taylor et al. 1998). Because the skeleton was found disarticulated in a secondary context, it is important to determine where the skeleton was originally buried. Such information may provide additional insight into the age of the remains and potentially how they were buried and modified after burial. Preliminary geoarchaeological study at Columbia Park (Huckleberry et al. 1998; Wakeley et al. 1998) suggests that the skeleton eroded out of the edge of a stream terrace during a period of high water in late winter and early spring of 1996. Although Huckleberry et al. (1998) and Wakeley et al. (1998) confirmed that the overall geological context of the site is congruent with the single 14C age of the skeleton, it was not possible to determine from where exactly in the streambank the skeleton was derived. However, Huckleberry et al. (1998) and Wakeley et al. (1998) proposed that sediment concretions on the skeleton probably correlate to a concretion-bearing stratum at approximately 60-130 cm below the modern surface. Chatters (1997b,c) earlier proposed this idea but further hypothesized that the skeleton came from a more specific depth in the upper part of the concretion-bearing stratum based on granulometric data. A preliminary chronology for the site (Stafford 1998; Wakeley et al. 1998) indicates that the terrace deposits located above the shore of Lake Wallula at the site range several thousands of years in age. The age of sediments within the concretion-bearing stratum could potentially differ significantly in age over a short vertical distance depending on depositional and erosional history. Consequently, it is critical that we test hypotheses regarding the geological history of the site and the specific location of the buried skeleton prior to its disturbance.
In this study, we attempt to better define the original position of the skeleton within the stream terrace at Columbia Park (herein "site") by matching sediments from the skeleton and site through a combination of physical and chemical tests. Our ability to correlate accurately the human remains with site stratigraphy relies on complete characterization of sediments from both skeleton and site. Whereas this study provides good sedimentological control for the human remains, the horizontal and vertical sedimentological variation at the site has yet to be precisely defined. Moreover, only a discontinuous set of control samples are available from the skeleton's discovery position. Although this limits our ability to determine the original provenience of the skeleton with a high degree of precision, we believe that there is adequate sedimentological control to test the hypothesis that the skeleton was derived from the concretion-bearing stratum. Whether or not there is adequate sedimentological control from the site to test Chatter's (1997c) hypothesis that the skeleton comes from the upper part of that stratum is less certain.
This report is divided into seven parts. Following the introduction, we review what is presently known of the stratigraphy at Columbia Park based largely on the work of Huckleberry et al. (1998) and Wakeley et al. (1998). We then describe the sediments adhering to the skeleton. This is followed by an overview of the methods employed to sample and analyze the sediments. We then present the results of the physical and chemical analyses and our interpretations of the data. We conclude with a discussion of the implications of our results regarding the age of the skeleton and its burial history and provide suggestions for future work.
The Kennewick Man discovery site is situated on a low Columbia River stream terrace at river mile 331.6 at an elevation of 104 m above sea level. The terrace is on the south side of the Columbia River between the mouths of the Yakima and Snake rivers and inset into a much higher Pleistocene terrace underlying the town of Kennewick. The higher terrace is associated with one of the many catastrophic glacial outburst floods (Baker et al. 1991; Bretz 1969; O'Connor et al. 1995) that formed the Channeled Scabland, the last of which occurred no later than 13000 yr. B.P. The terrace at Columbia Park post-dates the last Missoula flood, and was created by the Columbia River during a time of deglaciation in the northern Cordillera.
Limited stratigraphic information for the terrace is based on a preliminary geoarchaeological study led by the U.S. Army Corps of Engineers Waterways Experiment Station (WES) during December 1997 (Huckleberry et al. 1998; Wakeley et al. 1998). WES designated a 345 m reach of streambank as the study area. Horizontal survey control was established for the study area with the downstream end marked CPP000 and the upstream end CPP345 (Figure 1). The skeleton was recovered along the shore between CPP057 and CPP093. Huckleberry et al. (1998) subdivide the terrace deposits into Lithostratigraphic Units I and II whereas Wakeley et al. (1998) subdivide the terrace deposits into Units I-VI (Table 1). Despite differences in nomenclature, there are no significant differences on how the two research teams characterize and interpret the stratigraphy at Columbia Park. Huckleberry et al. (1998) preferred to limit the unit designations to the largest, mappable lithologic units until more detailed study of the terrace was performed. In this report, we present both systems of stratigraphic nomenclature (Figure 2), but prefer to use the terms "Lithostratigraphic Units I and II" given that their boundaries are well defined at the skeleton discovery site.
Lithostratigraphic Unit I is middle to late Holocene in age (less than ~7000 yr. B.P.) as evidenced by tephra and a few 14C dates on shell and sediment humates. It is probable that the reworked tephra clasts in the lower part of Lithostratigraphic Unit I at the skeleton discovery position are the same as that at CPP334, but because it is found in a possibly reworked context, it only provides a maximum limiting age, i.e., the lower part of the deposit is no older than 6700 yr. B.P. There are also two 14C dates on shell from the lower and middle parts of Lithostratigraphic Unit I at CPP005 and ~CPP200 that date 6510±60 yr. B.P. (Beta-113838) and 6090±80 yr. B.P. (Beta-113977), respectively (Wakeley et al., 1998). Although the shell dates may be erroneous due to reservoir effects (Taylor 1987:131; Stafford 1998), they are congruent with the stratigraphic location of Mazama tephra, geomorphic position of the terrace, and degree of soil development. Hence, Lithostratigraphic Unit I began to form approximately 7000 years. B.P. after the Columbia River downcut, and deposition on the resulting terrace became dominated by windblown, reworked alluvium (Huckleberry et al. 1998).
Four radiocarbon ages based on organics extracted from sediment in vibracore CPC059.5 provide age estimates for Lithostratigraphic Unit II (Figure 1) (Wakeley et al. 1998). One sample from approximately 1.2 m depth below the surface (approximately 0.5 m below the top of Lithostratigraphic Unit II) dated 9010±;50 yr. B.P. (WW-1626). The three other samples located at 2.5 m, 3.1 m, and 3.5 m below the surface yielded 14C ages of 12460±50 yr. B.P. (WW-1737), 15330±60 yr. B.P. (WW-1627), and 14560±50 yr. B.P. (WW-1738), respectively. Hence, the 14C dates are progressively older with increasing depth except for the bottom two samples. Given the mutual proximity of these two deepest samples and the mottled appearance of sediments from CPC059.5 at this depth, it is probable that the dates are reversed due to stratigraphic mixing. Of more direct significance to the skeleton is that the latest 14C date from CPC059.5 is derived from the bottom part of the concretion-bearing stratum hypothesized to have contained the buried human remains, and that this date is similar to the 14C date for the skeleton (8410±60 yr. B.P.). However, the accuracy of the sediment dates is questionable (Stafford 1998). Most troubling is that the bottom two ages predate the last of the glacial outburst floods (O'Connor et al. 1995) but appear to come from relatively low energy overbank alluvial deposits. In addition, Stafford (1998) argues that because the type of organic fraction dated in these samples is not specified, the actual significance of these ages cannot yet be determined. Despite these uncertainties, the overall geomorphology, pedology, and tephrachronology are consistent, at least at timescales of 1,000 years. We interpret Lithostratigraphic Unit II as a post-outburst flood deposit formed by the Columbia River during the Pleistocene-Holocene transition and continuing until just prior to the 6700 yr. B.P. Mt. Mazama eruption.
Preliminary observations of the sediment adhering to the human remains suggested that two kinds of sediments were present: 1) light-colored and hard and 2) darker-colored and friable. By far the most common sediments were contained within light gray (2.5Y 7/2; dry) calcitic concretions. The concretions were discontinuously distributed over the surface of the bones giving a lumpy oatmeal appearance (Figure 3). Some of these concretions would flake off, and the bottom sides previously in contact with the bone were darker in color: (2.5Y 5/2; dry). Less common was the darker friable sediment, commonly located in protected locations, e.g., between metatarsal bones and inside cranium. Some of these sediments contained common very fine roots (Figure 3) suggesting they were residual from the original terrace deposits. Also, some of the friable sediment on the exterior of the post-cranial elements was noticeably darker (2.5Y 6/2; dry) than adjacent concretions. The darker color may have been caused by lower calcite and/or higher organic matter content, relative to the light-colored hard sediment.
We hypothesized that the light-colored, hard (concretionary) sediment was derived from the stratum in which the bones were originally buried, and that the dark-colored, friable sediment was derived from either/or both the original burying stratum but with less calcite, or the modern alluvium that became attached to the skeleton as it was reworked in the shallow water. We extracted sediment samples in order to test these hypotheses.
Removal from Skeleton
On Friday, February 26, 1999, Vicky Cassman (conservator) evaluated the list and ranked specimens according to her assessment of the ease with which sediment could be removed with the least amount of damage to the bones. The highest ranked specimens were photographed as a record of their pre-removal condition. Strategies for removing the sediment from the bone were discussed, including scraping with bamboo or bone pointed tools, cutting with "exacto" knife, or vibrating and scraping with small drill. Physical anthropologists Jerry Rose and Joseph Powell, and Vicky Cassman recommended using a Dremel (brand) drill as the technique least likely to impact any bone. A new drill was purchased: Dremel Multi-Pro, 2-speed modal 285 type 5, with attachments of Dremel flex shaft (model 25T2) and multi-speed foot pedal (over-rides the 2-speed function and gives continuous speed control). The bits used were of various shapes and made of steel, carbide steel, and carbide steel embedded with diamond chips.
Sediments extracted from the bones were given the catalog number of the individual specimen. For example, sediment extracted from the left fibula was called 97.L.21d, meaning the sediment came from the left (L) fibula (21) fragment (d). An additional notation was added because multiple samples were sometimes taken from one specimen and needed to be stored and labeled separately. We devised a descriptive notation that was added to the catalog number. For example, 97.L.21d exterior, 97.L.21d marrow cavity, 97.L.21d dark represent three different sediment samples taken from the same specimen.
The greatest concern in removing the sediment from the specimen was to leave all bone in place. No part of the bone could be removed along with the sediment. To facilitate this, the physical anthropologists and conservator checked every sediment clump under magnification. All agreed that no bone was adhering to the sediment collected. After sediment was removed from a specimen, the bone was again photographed as a record of the changes made.
The specimens from which sediment was extracted are listed in a master inventory by catalog number, element, place of extraction on the bone, weight, destination (University of Washington or Washington State University), and analyses (see Appendix F: Final Sediment Sample Inventory - CENWW Kennewick Collection and Addendum). All 36 samples were placed in separate plastic vials, each with a foil-backed label. The sediment is grouped by provenience, e.g., exterior of bones, marrow cavity, interior vault of the cranium, and by visible traits, e.g., darker in color.
The sediment taken from the exterior of the bones seemed to be held tightly in place by cement. We focused on sampling the right and left tibia that had concretions on the outside described as an "oatmeal" texture, the right and left fibula and the right humerus that had nodules on the exterior shaft that looked much like concretions, the left ulna that had exterior color ranging from very light to dark reddish that contrasted with the color of the sediment nodules, the sacrum, and an unidentifiable fragment that also had bulbous nodules that looked like they could be easily removed.
We also collected sediment from the foot bones. Two metatarsal bones were held together by significant amount of sediment adhering to their exteriors. The large quantity of sediment preserved between the two metatarsals and the sediment's soft appearance made them a prime candidate for sediment collection.
All long bones had marrow cavities filled with sediment. Long bones that were broken at angles and at the widest part of the shaft were the right and left tibia, right femur, and left humerus, and thus we selected these specimens for sediment extraction.
The os coxae had two colors of sediment that we called dark and light. The dark-colored sediment was also observed on the left tibia and left fibula, which we sampled. For these specimens the dark sediment was collected and stored separately from the light sediment. We hypothesized that the dark sediment had a different origin than the lighter colored sediment, and hence would need to be tested separately.
The cranium contained, by far, the most sediment of all the specimens in the collection, but was the most difficult to extract. The Dremel drill was slid into the interior vault of the cranium through the foramen magnum after the conservator built a protective collar for the opening. Sediment was dislodged from various locations, but removed only after the drill was extracted, and the cranium was tilted. Sediment then fell out onto a piece of acid-free paper and was transferred to a vial. Both soft and somewhat cemented sediment was observed in the cranium. Both were recovered and stored separately.
The sediment on the bones was not all extracted with equal effort. On the exterior of bones, the sediment that formed lumps and "oatmeal" texture was firmly attached to the bone surface, and the Dremel drill had to pulverize the sediment to remove it. We suspected the cement was silica (in addition to carbonate), because the carbide steel drill bits were ground to a smooth surface very quickly. In fact, the diamond-tipped drill bits were the only ones that seemed to last longer than about one hour. The diamonds from these bits could have been dislodged and found their way into the sediment, so diamonds in any mineralogical analysis should be ignored. The pulverization of the drill bit rendered the sediment sample useless for granulometric analysis. Some of the sediment was held together with cement and internally cohesive but was already detaching from the bone surface. These concretions or "flakes" were loosened on the end still attached to the bone, and gently removed. Only large pieces (e.g., > 5 mm) of sediment extracted as a whole (and not pulverized) were used for granulometry. These pieces were also analyzed as "concretions" in the thermogravimetric analysis, where they were compared to concretions pulled from the bank samples.
of Site Sediments
The skeleton was recovered in shallow water a few meters offshore between CPP057 and CPP093 (Huckleberry et al. 1998). We selected sediment samples from stratigraphic column CPP054 and vibracore CPC059.5 for our control samples because they contained the most samples in close proximity to the skeleton recovery area. At CPP054, Lithostratigraphic Unit I extends from 0 to 70 cm below the modern surface and is sampled at 10 cm intervals. Because we are testing the hypothesis that the skeleton was derived from the concretion zone, i.e., the upper part of Lithostratigraphic Unit II or Wakeley et al.'s (1998) Stratum IV, we analyzed all samples from the 70-130 cm depth range (Figure 2). In addition we selected samples from Lithostratigraphic Unit I (Wakeley et al.'s Strata I, II, and III) and from the lower part of Lithostratigraphic Unit II (Wakeley et al.'s Stratum V) for comparative purposes. Two samples of modern shore deposits from the top of core CPC054 were also analyzed. Sediments from the site were shipped in sealed plastic bags from WES in Vicksburg, MS and air-dried in the Geoarchaeology Laboratory at the University of Washington.
A simple form of thermogravimetric analysis is called Loss-on-Ignition (Stein 1984). This procedure calls for sediment to be weighed before and after burning at 550°C, and again after burning at 1000°C. The technique works well for sediments with less than 5% clay. Sediments with more than 5% clay do not give accurate results because the heating drives off interstitial water held in clays that can be misinterpreted as organic matter or carbonates. The samples found on the skeleton and in the streambank contain very little clay (Appendix A), so loss-on-ignition provides an accurate measurement of the percentages of organic matter and carbonate. Stein performed these analysis in a standard laboratory furnace (Table 3; Appendix C) in conjunction with a more detailed thermogravimetric analysis described below.
New technology is available that allows for a much more accurate representation of the weight loss occurring during the heating of sediment (Brown 1988; Speyer 1994, Wendlandt 1986). Rather than measuring the weight loss at only three designated times (before and after 550° and after 1000°), a Thermogravimetric Analyzer (TGA) can make continuous measurements of weight-loss in a controlled stream of oxygen. This analyzer provides not only the weight loss from oxidation of organic matter and carbonate, but also a graphic representation of that weight loss over time. This allows for a more detailed comparison of sediment samples. Samples with similar mineralogical and organic content (i.e., from the same layer) should produce identical TGA graphs.
Five representative sediment samples from the skeleton were selected for loss-on-ignition and TGA analysis (Table 3; Appendix C) to represent the different kinds of sediment observed on the skeleton. Nine samples from the discovery site were selected to represent strata described in the terrace profile.
Loss-on-ignition and TGA data were derived from one procedure. Between .05 and .09 g of pulverized sediment sample was placed in a Perkin Elmer (brand) 7 Series/Unix TGA 7; Thermogravimetric Analyzer. The TGA is controlled by a computer which is programmed to begin heating at a specified temperature and to continue to heat at a constant rate to a final temperature. The heating program used in this study called for a 5 minute period of acclimation at 30° C, followed by a 60 minute period (rising at rate of 1° C per minute) from 30° to 1000°. Weight measurements are taken multiple times per minute providing an accurate distribution of weight loss as temperature increases. The standard TGA software program controls the instruments, the heating and cooling, the calculations and graphic output. The machine used in this analysis was operated by Stein, under the supervision of Dr. Brian Flinn. It is housed in the Thermal Analysis Lab - Material Science and Engineering, University of Washington. The loss-on-ignition data (percent weight) was calculated by the computer using the TGA data, involving the weight of the sample before the burn, minus the weight loss at 550° C and the weight loss at 1000°C (Table 3).
Samples subjected to X-ray analysis were selected to cover the range of sediment types from the skeleton as well as the site samples. The data was collected by Dr. Brian Flinn and Kyle Flanigan of the Department of Materials Science and Engineering at the University of Washington who produced a report included as Appendix D. It includes individual X-ray patterns presented as graphs for each sample tested, as well as a combination of graphs from all the skeleton samples, and a combination of graphs from all site samples. Lastly, the graphs for site sediments are compared with one sample (a concretion) from the skeleton. The individual graphs are analyzed by a computer program called JADE. The program instructs the computer to identify the minerals associated with all peaks whose intensity falls above 10% of the highest peak. This procedure allows one to identify the major crystalline minerals in the sample. If no peaks fall above 10% of the height of the highest peak, then only the highest peak is identified.
The same five skeleton sediment samples and nine site sediment samples selected for thermogravimetric analysis were also used in the X-ray diffraction analysis (Table 2). Each sediment sample was ground in an agate mortar, placed within a sample holder, and received one hour and 13 minutes of machine time (see Appendix D for details of the procedure). The diffraction signal was plotted versus the 2q (degree).
The volume-weighted mean diameter is a measure of the average particle-size whereas span and uniformity provide measures of the spread of the distribution (variance), or in this case, the degree of sorting. Samples with higher span and uniformity values have a greater range of particle sizes. Because span and uniformity are both measures of sorting, their coefficients of variation for strata and skeleton are essentially the same. Consequently, in discussing sorting patterns between strata and skeleton, only measured values of span are presented.
Several sediment samples had repeat measurements performed on separate subsamples (Appendix A: Table A.1), thus allowing for a qualitative assessment of the reproducibility and the inherent range of variability in the Mastersizer laser method. In general, the range of particle-size means for repeat runs on the same sample is less than 12 µm. This is in comparison to a range of 94.2 µm for particle-size mean for all runs (minimum 15.2 µm; maximum 109.4 µm). An exception is CPC059.5, 0-10 cm which has two subsamples measured with particle-size means of 37.4 µm and 60.4 µm, or a difference of 23 µm. The source of variation in repeat measurements is due to both variation in subsampling and the sediment:water flux across the laser beam (Chappell 1998). Nonetheless, most repeat samples had measured particle-size means that differed less than 10% of the entire range of values. Given the relatively low amount of variation in particle-size means for samples from the site and skeleton, 10% variability in the laser method can be significant and limit correlation.
Skeleton Sediments: There is very little variation in the grain-size distributions of sediments removed from the skeleton (Appendix A: Table A.2). The range of particle-size means is 63.5-83.8 µm, and the coefficient of variation for the five sample means is 0.11. The dark sediment from the right leg (R.21a) is coarsest in texture; the finest sediment comes from the external concretions (L.20c+L.15b+L.15a). All of the sediments are relatively well sorted. Values for span range 2.5-3.0 with a mean of 2.7; the coefficient of variation for the span means is only 0.08. It is worth noting that the greatest difference in particle-size mean is between sediments from the concretions and the dark sediments. There is an inadequate number of sediment samples from the skeleton to determine if the difference is statistically significant.
Site Sediments: The site sediments were subdivided into two categories: Lithostratigraphic Unit I and Lithostratigraphic Unit II. These units are used because their boundaries are well defined, and it is easy to assign samples to their respective deposit. Moreover, by grouping samples into these two primary deposits rather than the six-stratum system of Wakeley et al. (1998), the sample size is increased per stratum thus facilitating statistical comparison. Overall, the range in particle sizes in both lithostratigraphic units (n=13) is not large (Appendix A: Table A.2). Particle-size means range 15.2-81.8 µm with a mean of 52.9 µm and a coefficient of variation of 0.38. The mean of the span values is 3.0 (range 2.3-4.1); the coefficient of variation is 0.18. The beach sediments, however, are quite distinct with a particle-size mean of approximately 100 µm (Appendix A: Table A.3). When the samples are subdivided into Lithostratigraphic Unit I (n = 3) and II (n = 10), the means of the particle-size means, span, and uniformity are greater for Lithostratigraphic Unit I than Unit II (Appendix A: Table A.3). Several of the samples from CPC059.5, however, are from positions well below the surface of Lake Wallula and could not represent a potential point of origin for the skeleton. If only samples from Lithostratigraphic Unit II located above the reservoir lowstand (n=7) are included, there is little change in the relative means and variances; the means of particle-size mean, span, and uniformity are still greater for Lithostratigraphic Unit I (Appendix A: Table A.4). However, statistical trends are limited with such small sample sizes.
In an effort to increase sample size, we added Chatter's (1997b) granulometric data. Although we do not know from where specifically Chatter's samples were collected, he notes the stratigraphic column was located within two meters of the skull and thus somewhere between CPP054 and CPP093. It is possible to assign the appropriate lithostratigraphic unit to Chatter's samples based on his stratigraphic profile which shows the concretion-bearing stratum. Samples located above the concretion-bearing stratum are part of Lithostratigraphic Unit I; those located within and below the concretion-bearing stratum are part of Lithostratigraphic Unit II. With the combined data sets, the sample size increases to six for Lithostratigraphic I and 18 for Lithostratigraphic Unit II (Appendix A: Table A.5). The range in particle-size means for Lithostratigraphic Units I and II are 54.8-79.3 µm and 42.0-81.8 µm, respectively. Although the mean of particle-size means remains higher for Lithostratigraphic Unit I, the means of span and uniformity are essentially the same. Both t and F tests indicate that there is no statistical difference at a 95% confidence interval between Lithostratigraphic Units I and II particle-size means, span, and uniformity. A summary of granulometric statistics is provided in Appendix A: Table A.6.
Comparison of Skeleton and Site Sediments: Because there is no statistical difference in particle size mean and sorting between Lithostratigraphic Units I and II, it is not possible to confidently match skeleton sediments to either unit based on granulometry alone. For example, the particle-size mean for the concretions on the skeleton is 63.5 µm which falls within the 95% confidence interval of Lithostratigraphic Unit I (58.5-79.9 µm) and not Lithostratigraphic Unit II (50.9-60.4 µm). This would suggest that the skeleton came from above the concretion-bearing stratum. However, given that the variation in repeat measures of particle-size mean using the Malvern laser spectrophotometer is generally 12 µm and can be as high as 23 µm, these differences in particle-size mean are not adequate to confidently correlate skeleton sediments to any stratum at the site, let alone a particular depth within a stratum. In contrast, the particle-size mean of the dark sediment from the skeleton (83.8 µm) falls outside the confidence intervals of both units and suggests a post-erosional origin, i.e., it adhered to the bone while it was reworked in the shallow shorezone in 1996. Nonetheless, the difference is not great and repeat sample variation could place it within either confidence interval of Lithostratigraphic Unit I or II.
Likewise, the span value for the concretion on the skeleton (2.83) falls within the 95% confidence interval for Lithostratigraphic Unit I (2.31-3.59) and Lithostratigraphic Unit II (2.82-3.20), and thus sorting does not help in matching the skeleton sediment to a stratum. Furthermore, the variation in repeat sampling could place the concretion sample outside either confidence interval. In sum, granulometry alone is inadequate to pinpoint the origin of the skeleton at the site. We cannot reject Lithostratigraphic Unit I or II as sources for the skeleton based on granulometry.
Despite our inability to correlate sediment from the skull to Lithostratigraphic Unit I or II based on micromorphology, it does appear that sediment from inside the skull is residual from the terrace (rather than beach sediment incorporated into the skull in 1996). Samples 251 and 251a from the cranium are "not dissimilar from that of Samples 245, 246, and 250" (Appendix B). The cranium sediments contain secondary calcite that has been partially leached. Thus, the thin-section analysis was successful in identifying in situ, pedogenically altered sediment inside the cranium, but it does not help in our correlation of the skeleton to a particular depth and stratum at the site.
All sediment samples had very small weight losses between temperatures of 70°C to 550°C (representing oxidation of organic matter), and all but a few also had small weight losses between temperatures of 550°C to 1000°C (representing oxidation of carbonates) (Figure 6; Table 3; Appendix C). Fortunately, several samples have distinct and comparable derivative weight-loss curves providing a means for correlating sediments between the skeleton and the site.
Skeleton Sediments: Four of the sediment samples taken from the skeleton have relatively low amounts of organic matter (between 1.2% and 1.8%)(Table 3). Such levels of organic matter are commonly found in fine textured alluvial sediments due to incorporation of detrital organic matter (Stein 1992). However, there are possible alternative sources for the organic matter including modern rootlets such as were observed on many bones and in the adhering sediment, and/or soluble organic acids produced in the modern surface soil and leached downward or residual in buried A horizons. The dark-colored sediment removed from the pelvis has the highest measured organic matter content (2.3%). It also has a derivative curve of weight loss that displays a distinctive shape, different from the other skeletal samples. The shape displays three small dips between temperature 70°C and 550°C. The first dip probably represents the weight loss from evaporating water held in the lattice structure of clays and organic matter; the second and third dips represent the oxidation of organic matter. Higher organic matter content is also supported by the relatively dark color of the sediment. This suggests that the dark sediment may have originated from a layer close to the modern A horizon or alternatively from the modern shore sediments.
Weight loss due to oxidation of carbonates is more varied and dramatic than that for organic matter. The loss-on-ignition results for the concretion (97.I.25c) indicate a very large weight loss after 550°C, suggesting that the concretion is half carbonate by weight (Table 3). The metatarsal sediment (97.L.24c) contains intermediate amounts of carbonate (18.5%), as does the sediment from the marrow cavity of the femur (97.R.18a)(18.4%). These intermediate values indicate that carbonate is present but not concentrated into a nodule or concretion, and explain why the sediment, although not completely indurated with carbonate, may have remained with the bone even after exposure to the swash and backwash of shallow waters of Lake Wallula. Sediment in the cranium (97.U.1a) has only 6.7% carbonate, which explains why it was softer and more easily extracted. Micromorphological examination indicates that the sediment within the cranium contains partially leached secondary calcite (Appendix B). Given that there is less carbonate in the cranium sediments than in sediment adhering to other bones, the sediment inside the cranium may have been protected from carbonate illuviation, i.e., percolation of calcite-enriched waters was impeded by the cranium and limited to a few openings in the skull. The least amount of carbonate from any sediment extracted from the skeleton was measured in the dark sediments on the pelvis (3.7%) suggesting that it may have originated from closer to the modern surface, or more likely, from the shore of Lake Wallula.
Site Sediments: Organic matter content decreases with depth in the upper 90 cm of CPP054 (Table 3); this includes all of Lithostratigraphic Unit I and the upper 20 cm of Lithostratigraphic Unit II. The highest value of organic matter (2.2%) is found in the modern A horizon near the surface at 10-20 cm, the upper part of Lithostratigraphic I (Wakeley et al.'s Stratum I). The next highest values are 2.1% and 1.7% and found in samples from 30-40 cm and 50-60 cm representing the middle part of Lithostratigraphic Unit I (Wakeley et al.'s Stratum II or III). Organic matter content ranges 1.4% to 1.8% in the upper part of Lithostratigraphic Unit II (Wakeley et al.'s Stratum IV) within the concretion-bearing stratum. The lowest organic matter content (1.0%) is from a calcitic concretion at 80-90 cm and reflects more of the diluting effect of calcite rather than a real drop in organic matter content at that depth. Two samples from below the concretion-bearing stratum (Wakeley et al.'s Stratum V) have comparable organic matter values of 1.6 and 1.7%. All of these values are reasonable for fine-textured alluvium affected by pedogenesis and comparable to sediments adhering to the skeleton.
The weight-loss derivative curves for samples from the upper part of the profile show two marked increases in slope in the temperature range of 450° to 550°C. The first increase probably represents the weight loss from evaporating water held in the lattice structure of clays and organic matter; the second increase represents the oxidation of organic matter, but a kind of organic matter different from that observed in the derivative curve for dark sediment from the pelvis. This second peak is not evident in the sample from 50-60 cm, nor in the samples from 70-80 cm, 80-90 cm, or 0-10 cm (core). There is a slight expression of this second peak in the sediment sample from CPP054, 95-135 cm. The curves for the two samples from Lithostratigraphic II located lower in the profile at depths of 30-40 cm and 60-66 cm in CPC059.5 (Wakeley et al.'s Stratum V) are slightly different from all other site samples. These curves have a dramatic peak representing the weight loss from evaporating water held in the lattice structure of clays and organic matter, with a less dramatic peak for the oxidation of organic matter. As mentioned before, the scale of the derivative curve varies for each sample, yet the overall shapes of the curves can be compared. Accordingly, these lowermost thermogravimetric samples are unique and contrast with sediments higher up in the profile, thus supporting Wakeley et al.'s separate stratum designation (Stratum V). In sum, the thermogravimetric curves indicate that although the amount of organic matter and carbonate varies between samples, the kind of organic matter and mineralogy present is similar in samples above 135 cm depth which includes all of Lithostratigraphic Unit I and the upper part of Lithostratigraphic Unit II (Wakeley et al.'s Strata I, II/III, and IV) but different for samples below 135 cm depth (middle and lower part of Lithostratigraphic Unit II and Wakeley et al.'s Stratum V).
Carbonate values for site sediment samples range widely with depth (Table 3; Figure 6), although there is a gradual increase with depth above 90 cm depth. The lowest carbonate value (2.8%) is found in the modern A horizon near the surface at 10-20 cm, the upper part of Lithostratigraphic I (Wakeley et al.'s Stratum I). This increases to approximately 8.7% at 70-80 cm depth. Carbonate values for the concretion-bearing stratum are highest in the sequence, but are exaggerated by the fact that for sample CPP054, 80-90 cm one concretions indurated with carbonate was selected for individual analysis and not surprisingly yielded carbonate contents close to 50%. This concretion was selected as a comparison for the concretion extracted from the skeleton. Carbonate values for samples from 70-80 cm (8.7%) and 95-135 cm (5.0%) better represent the range of values for sediments surrounding the concretions. The sediment sample taken from CPC059.5, 0-10 cm must have contained a small concretion given its carbonate content of 35.0%. Thus, the concretion-bearing stratum (upper part of Lithostratigraphic Unit II or Wakeley et al.'s Stratum IV) seems to contain a large amount of carbonate, but it is concentrated into concretions that are not easily measured in sample sizes as small as dictated by TGA (.07 grams). Finally, carbonate content for the two deepest samples (Wakeley et al.'s Stratum V) contain the lowest amounts of carbonate of all of the samples (1.8%) and appear to have experienced either little calcification or have been effectively decalcified.
Changes in the rate of weight loss at temperatures between 600°C and 800°C produced a distinctive peak in the derivative curve for samples in Lithostratigraphic Unit I and the upper part of Lithostratigraphic Unit II. This suggests that although the amount of carbonate varies between these samples, the kind and form of carbonate seems to be similar. Again, however, the samples from Wakeley et al.'s Stratum V are different with the same peak being less distinct and occurring closer to the 600°C temperature. Hence, Stratum V appears unique in both its carbonate and organic matter assemblage.
One other moderately defined trend is evident in the thermogravimetric data. The temperature corresponding to the peak of the derivative curve, i.e., the maximum rate of weight-loss, for sediment samples from the site generally increases with depth, at least within approximately the upper one meter of the terrace. Because these temperatures are greater than 600° C, we believe they relate to the carbonate content in the sediment, although it is uncertain if this represents differences in carbonate chemistry or crystal size.
Comparison of Skeleton and Site Sediments: Some distinctive comparisons can be made between the sediment extracted from the skeleton and site. First, the organic matter and carbonate contents of the concretions extracted from the skeleton (97.I.25c) and the sample located in the upper part of Lithostratigraphic Unit II (CPP054, 80-90 cm) are almost identical with respect to their TGA curves. This suggests that the concretion on the skeleton was formed when the human remains were within the upper part of Lithostratigraphic Unit II. The skeleton could, however, have come from any depth within the concretion-bearing stratum, i.e., the upper part of Lithostratigraphic Unit II above Wakeley et al.'s Stratum V.
Second, the other four sediment samples extracted from the skeleton are comparable to each other and to two site samples (CPP054, 80-90 cm and CPC059.5, 0-10 cm). in terms of carbonate content. This suggests that the sediment adhering to the human remains is only that sediment with the highest levels of carbonate. In other words, the carbonate is one of the reasons that the sediment is still adhering to the skeleton. Without the carbonate holding the sediment to the bone, the sediment would have washed away in the waters of Lake Wallula.
Third, the dark sediment from the skeleton (97A.I.17a; os coxsae) is dark because it contains higher levels of organic matter, levels comparable to the A horizon of the modern surface soil. Its TGA derivative curve, however, has a distinctive third peak suggesting that the sample has an additional kind of clay mineral or organic matter not present in the A horizon. We believe the dark sediment most likely became attached to the bones while they were laying in the shallow shorezone of Lake Wallula.
Finally, the samples extracted from the skeleton (other than the dark sediments) have no organic matter signature in the TGA curve data. This suggests that the weight loss recorded up to 550°C was not the result of organic matter oxidation. It was more likely related to the weight loss from evaporating water held in the lattice structure of clays. This suggests that the skeleton was located well below the modern A horizon.
As mentioned above, the temperature at which the maximum rate of weight-loss occurs in the site sediments tends to increase with depth. A linear regression analysis of the maximum weight-loss temperatures for the site sediments above 135 cm depth (Figure 5) results in an R2 value of 0.64. The four skeleton sediment samples have fairly consistent temperatures of peak weight-loss in the 700°-780° C range. Using the linear regression curve, this suggests that skeleton sediments came from a depth of 62-110 cm below the surface. This includes the bottom 8 cm of Lithostratigraphic Unit I and the upper 40 cm of Lithostratigraphic Unit II. Although this suggests a more specific depth for the skeleton within the concretion-bearing stratum, the evidence should be treated with caution because we do not know the cause of this trend in maximum weight-loss temperatures with depth.
Differences in minor crystalline minerals do occur in the other samples. Although many of the small peaks cannot be identified mineralogically, they do serve as a reference for qualitative comparisons. If each sample from the skeleton is compared to samples from the site, then some likely matches can be identified (see Appendix D-20 to D-24). Areas of the curves that are instructive for comparison are labeled A, B, C, D, E, F, and G. We note the following observations:
Obviously these observations are based on only a small sample from a sequence of heterogeneous deposits, and the XRD traces have not been analyzed to determine the most probable crystalline phases associated with each peak. Despite these restrictions, we believe that the XRD data supports the hypothesis that the skeleton comes from the concretion-bearing stratum, i.e., the upper part of Lithostratigraphic Unit II or Wakeley et al.'s Stratum IV. There is no complete, identical match between the XRD graphs, but taken as a whole, the XRD data suggest that the most likely match for the mineral fraction of the sediment adhering to the skeleton is with the sample from CPP054, 70-80 cm, and less likely but still strong matches with sediment from CPP054, 80-90 cm and 95-135 cm.
Other data are presented by the analysts in Appendix D, pages D-25 to D-34, (e.g., probable crystalline phases associated with each peak in each sample), and are available for anyone interested in further analysis.
Given that the skeleton contains calcitic concretions on the exterior of the bones, it seems intuitively obvious that the concretion-bearing stratum in the upper part of Lithostratigraphic Unit II is the probable origin of the skeleton. However, good science is based on repeated testing of hypotheses. In order to increase our confidence that the skeleton indeed was buried in the upper part of Lithostratigraphic Unit II, we need to try and falsify that hypothesis. All of the laboratory analyses performed in this study provide new data for characterizing sediments adhering to the human remains and from the stratigraphy at Columbia Park, but some of the analyses proved more useful than others for testing the hypothesis that the skeleton was derived from the concretion-bearing stratum located in the upper part of Lithostratigraphic Unit II.
The laboratory analyses presented here support interpretations by Huckleberry et al. (1998) and Wakeley et al. (1998) that the site was formed by low energy overbank deposition by the Columbia River with some eolian inputs in the upper deposits, and that these sediments have been affected by soil formation. Lithostratigraphic Unit II contains bedded alluvium that is significantly mixed and modified by secondary accumulations of pedogenic calcite, and to a lesser degree, clay. Allophane was identified in the exterior concretion from skeletal element I.25C and is a noncrystalline clay that forms from the weathering of volcanic tephra (Birkeland 1984:100). The most abundant source of tephra at the site is from the Mt. Mazama eruption of 6700 yr. B.P. This suggests that 1) the skeleton was buried at a depth below the Mt. Mazama tephra which is in the lower part of Lithostratigraphic Unit I (Figure 2), and 2) a substantial part of the soil formation at the site including formation of the calcitic concretions in Lithostratigraphic Unit II occurred after 6700 years. B.P. This suggests that the top of Lithostratigraphic Unit II is not a truncated paleosol and that the contact between the two units is conformable (Huckleberry et al. 1998). Although the upper part of Lithostratigraphic Unit II may have been subject to soil formation prior to burial by Lithostratigraphic Unit I, much of the pedogenesis occurred during the middle to late Holocene during or after deposition of Lithostratigraphic Unit I. This pedogenesis resulted in the modification of both deposits. Thin-section analysis (Appendix B) suggests that sediments from both above and below the contact of Lithostratigraphic Units I and II have been decalcified, suggesting more recent leaching that was very effective. Given climate shifts during the middle and late Holocene in the Columbia Basin (Chatters 1998), such changes in the local soil-forming environment is not surprising.
Although the main purpose of the laboratory analyses is to determine the spatial and temporal context of the skeleton, our analyses indicate that some of the physical properties of the sediments are not useful for distinguishing deposits or correlating site and skeleton sediments. For example, the granulometric data demonstrate that differences in particle size and sorting in Lithostratigraphic Units I and II are so small that they cannot be used confidently to match the skeleton with any particular depth. The granulometric data do indicate, however, that the dark sediment on the skeleton was probably deposited after the skeleton eroded out of the streambank and reworked in the shallow shorezone of Lake Wallula. Likewise, micromorphological analysis suggests that sediment inside the cranium has been affected by pedogenesis and is therefore derived from the terrace. Micromorphology samples from Lithostratigraphic Units I and II, however, are not greatly different from each other, and consequently do not help us in determining the original provenience of the skeleton. The trace-element chemistry also is relatively uniform with depth at the site, thus precluding correlation of skeleton sediment with Lithostratigraphic Unit I or II. A large amount of yttrium in the dark sediment on the skeleton, however, suggests that this sediment is not derived from the stream terrace and more likely comes from the modern shore.
Thermogravimetric analysis proved more helpful than granulometry, micromorphology, and trace element analysis in our effort to match skeleton sediments to the site. Specifically, the weight-loss curves from the concretions in the upper part of Lithostratigraphic Unit II match the concretions on the skeleton, indicating they have the same carbonate and organic matter content. Other sediment extracted from the skeleton has values of carbonate suggestive of a cementing agent, no doubt one of the reasons that the sediment is still adhering to the skeleton. Sediment from the skeleton is also low in organic matter, and display no organic matter weight-loss peak in their TGA derivative curve. This suggests that the skeleton did not reside in layers close to the present surface of the site, i.e., the modern A horizon. Lastly, the TGA data support inferences based on granulometric and trace element analysis that that the dark sediment found on the skeleton does not come from the streambank. These sediments have organic matter levels higher than any sample from Lithostratigraphic Unit II, and a distinctive extra peak within the organic matter weight-loss curve that distinguishes it from the modern A horizon at the surface.
The X-ray diffraction data provide evidence of subtle differences in mineralogy at different depths of the site and thus provide additional evidence for correlating site and skeleton sediments. Some of vertical differentiation is due to the presence of secondary minerals such as calcite and allophane confirming field observations of pedogenesis. The XRD technician believes that the mineralogy of the sediment sample from CPP054, 70-80 cm below the surface, best matches the mineralogy of sediment from the skeleton. The second best match for the skeleton is CPP054, 95-135 cm below the surface. He sees differences, however, between the mineralogy of the sediment from CPP054, 80-90 cm, a depth intermediate between the two samples ranked as closest matches with the skeleton samples. Unfortunately, it is not possible to quantify the differences between the mineralogy of these strata, and thus we are uncertain how much confidence to place in the correlation of the skeleton to CPP054, 70-80 cm. Our conservative interpretation is that the mineralogy supports the hypothesis that the skeleton is derived from the upper part of Lithostratigraphic Unit II, but that mineralogical analysis alone, of only a few samples, cannot be used to discern a burial depth at a resolution of 10 cm within Lithostratigraphic Unit II.
Conclusions and Recommendations
Our study supports the conclusions of Huckleberry et al. (1998) and Wakeley et al. (1998) that the skeleton was eroded out of the streambank at Columbia Park from a depth below the Mazama ash which, if a primary deposit, means the skeleton is older than 6700 yr. B.P. The hypothesis that we tested in our study is that the skeleton was derived from the concretion-bearing stratum that corresponds to the upper part of Huckleberry et al.'s Lithostratigraphic Unit II and Wakeley et al.'s Stratum IV. At the skeleton discovery location, this corresponds to a depth of approximately 70-140 cm below the surface. We cannot reject Lithostratigraphic Unit I as a source for the skeleton based on granulometry, micromorphology, or trace-element chemistry. We can, however, reject Lithostratigraphic Unit I as a source for the skeleton based on thermogravimetry (organic carbon and carbonate content) and X-ray diffraction (mineralogy) analysis. The data support the interpretation that the skeleton was derived from the concretion-bearing stratum. At our sedimentological control site, CPP054, this represent the upper 70 cm of Lithostratigraphic Unit II (see Wakeley et al. 1998: Appendix D:CPP054) or the deposit referred to as Stratum IV by Wakeley et al. 1998.
We cannot reject Chatter's (1997c) hypothesis that the skeleton came from the upper 30 cm of Lithostratigraphic Unit II corresponding to a depth of 70-100 cm below the surface at CPP054 (Figure 2) based on any of our physical and chemical laboratory analyses. That the skeleton comes from this depth is best supported by the XRD data which suggests that the sediments on the skeleton best match those at 70-80 cm depth, but variability in XRD in adjacent samples, e.g., 89-90 cm, does not match the skeleton sediments, and it is unlikely that the skeleton was confined to only a 10 cm layer. Thus, the XRD data alone is inadequate to confirm this hypothesis. The thermogravimetric data confirm that the concretions on the skeleton are the same as those from the site, and further suggest that that the skeleton comes from the upper 50 cm of Lithostratigraphic Unit II, i.e., 70-120 cm below the modern surface at CPP054. Thus, we conclude that the best estimate for the original depth of burial for the skeleton at Columbia Park is the upper 50 cm of Lithostratigraphic Unit II or Wakeley et al.'s Stratum IV. There is, however, insufficient information within these sedimentological analyses to infer the mode of emplacement (i.e., burial) of the skeleton within these deposits.
If indeed the skeleton comes from this depth at Columbia Park, then there are two stratigraphic ages that help to confine the age of the skeleton: the Mazama tephra and a 14C date from ca. 130 cm depth (Figure 2). Small fragments of Mazama tephra were identified in the lower part of Lithostratigraphic Unit I at CPP054 (Huckleberry et al. 1998:Appendix A, CPP054). It is not known how exactly the Mazama ash relates to these sediments at CPP054 (Stafford 1998). If the tephra represents primary airfall, then the age of the sediments is 6700 yr. B.P. If reworked by wind and water, then the sediments at this depth could be over a thousand years younger than 6700 yr. B.P. The AMS 14C date (WW1626), derived from organic sediment, is 9010±50 yr. B.P. Because the organic fraction dated is not specified in Wakeley et al. (1998), the accuracy of this age is uncertain (see Site Stratigraphy above). If we assume that the Mazama tephra is in situ, and that the 14C age is correct, then the geologically correlated age for the skeleton is 6700-9000 yr. B.P. This age is congruent with the single bone date from the skeleton of 8410±60.
Given the significance of these human remains, the spatial and temporal context of the skeleton must be defined at a high level of confidence. Our study provides additional contextual information for the skeleton but is constrained by a lack of stratigraphic and sedimentological control at the site. Our correlations can be strengthened by further stratigraphic analysis of the site. Although the skeleton discovery position is buried under rock riprap, excavations into the terrace adjacent to the riprap would provide more detailed information about Lithostratigraphic Units I and II and hence better horizontal and vertical stratigraphical control of the discovery site. To date, only a partial two-dimensional view of the stratigraphy is available.
Even if our correlations of site and skeleton sediments prove to be correct, the resulting age inferences are only as sound as the quality of the samples taken at the site. The 6700-9000 yr. B.P. correlated age estimate for the skeleton is based on assumptions that the Mazama ash is primary and that the 9010±50 yr. B.P. 14C age is correct. These two assumptions need to be tested. That would be accomplished by 1) excavating the site adjacent to the riprap to determine the stratigraphic context of the Mazama ash in the vicinity of the skeleton discovery position, 2) determining what organic fraction was dated on sample WW1626 by the U.S Geological Survey Climate History/Hazard Program 14C Laboratory in Reston, VA, and 3) dating different organic fractions from several sediment samples in the upper 50 cm of Lithostratigraphic Unit II. Further 14C dating of sediments should involve detailed sampling from a single vertical exposure within the terrace as close to CPP054 and CPC059.5 as possible. These analyses will improve the contextual resolution of the human remains found at Columbia Park to a point beyond what we have provided.
We thank Drs. Lillian Wakeley and William Murphy of the U.S. Army Corps of Engineers Waterways Experiment Station in Vicksburg, MS for rapidly providing us with sediment samples from Columbia Park. We also appreciate the help of the WSU Pedology and Quaternary Studies Laboratory staff (Dr. Alan Busacca and Ms. Sandra Lilligren), Dr. Charles Knaack of the WSU Geoanalytical Lab, ICP unit, Dr. Brian Flinn and Kyle Flanigan, of the UW Department of Materials Science and Engineering, and Dr. Paul Goldberg for his analysis of the thin-sections.