Collected Papers on Caddoan Archaeology in the Upper Sabine River Basin, Northeastern Texas

Well preserved faunal and floral materials recovered from the Taddlock and Spoonbill sites in the Sabine River Basin of East Texas provide one of the first opportunities to document the subsistence strategies of sedentary hamlet occupations during the Early Caddoan period. The Taddlock site, dated ca. A.D. 940-1000, has a large faunal assemblage indicating a generalized and balanced exploitation of small and large mammals, reptiles, and fish. The two components at Spoonbill, dated ca. A.D. 970 ± 65 and A.D. 1260 ± 65, are characterized by an extensive floral sample of wild plant foods, seeds, and maize. At both sites, maize constitutes less than 10% by weight of the total plant food remains. The Early Caddean inhabitants exploited a wide variety of animal and plant food, but at this time maize was likely one of several main sources of food energy rather than the focus of a specialized economy.

This part of Northeast Texas has a highly significant and diverse archaeological record, one that has intrigued professional and a vocational archaeologists alike for at least 75 years (e.g., Pearce 1920;Johnson and Jelks 1958;Johnson 1962;Granberry 1985;Friedell and Skinner 1995). However, we still know very little about the prehistoric and early historic Caddoan groups who lived in the basin, and unfortunately it has been a number of years since dedicated archaeologists, professional or avocational, turned their attention to this region; on the other hand, looters and vandals who want to make a profit from their plunder of the past have not overlooked the region.
Thus, the publication of this compilation of papers serves two purposes: first, to make accessible in one document an integrated and coherent series of papers that illustrate the interesting and dynamic nature of Caddoan archaeology in the Upper Sabine River basin, and second, to foster a renewed interest in studying the regional Caddoan archaeological record. Hopefully, this will help to effectively communicate the results of archaeological investigations to interested members of the public and the Caddo Tribe ·RAiNs co.·-· '"l.   800 -1400). This site is located on an upland projection adjacent to the margins of Lake Fork Reservoir in Rains County, Texas. It was discovered in 1975 by the Archaeology Research Program at Southern Methodist University during a survey of the upper end of the thenproposed Lake Fork Reservoir : Figure  13 ). SMU tested the site the same field season, recovering 75 artifacts in the excavation of 12 systematically placed postholes across the site (Bruseth et at. 1977: 165). Ceramics and one dart projectile point found in the postholes suggested that the site was occupied during Archaic and Caddoan periods.
The site was not recommended for excavations during the mitigation phase of the archaeological research at Lake Fork Reservoir , because it would not be directly affected by the reservoir inundation. Indirect affects of the completed reservoir have been severe, however. Subsequent to reservoir filling in 1982 the area around Lake Fork has been witness to considerable lakeshore development. This part of the lakeshore has been partially bulldozed for a new road connecting a series of staked-out house lots. That activity has exposed several new archaeological sites, and disturbed 41RA65 (and 41RA66, 400 m to the westnorthwest of 41RA65). This disturbance was noted during a 1986 reconnaissance of Lake Fork Reservoir, and led us to return to the site to assess its condition. We found that the site had been stripped bare of vegetation and most of the A horizon on the point of the upland projection had been pushed downslope by a bulldozer. Quantities of cultural material were exposed on the cleared surface and in the fill downslope. The site was also being used as a convienent fishing spot. Thus, the in situ site deposits have probably been removed entirely as a result of land development arising from lake construction. Our work at the site was a belated attempt to recover as much useful information about the archaeological record here as possible before all evidence of it is erased. SETTING 41RA65, located approximately 6 km north of the community of Emory, Texas, and 5 km south of the Rains and Hopkins County line, is situated along an interstream divide between the Garrett Creek Valley and Lake Fork Creek. Garrett Creek is a permanent flowing stream which originates in SW Hopkins County along margins of the Blackland Prairie. It flows in a northwestsoutheast direction to its present confluence with Lake Fork Creek near the Rains-Woods County line below the site. In the last 5-6 km of its course it runs in an old Lake Fork Creek channel that parallels the steep, dissected north valley wall of the Lake Fork Creek floodplain. The combined Lake Fork Creek-Garrett Creek floodplain ranges between 1-2 km in width, then constricts to half that width at the confluence of the two streams near less resistant Eocene age Wilcox Group bedrock outcrops.
The Garrett Creek floodplain by the site is about 800 meters wide, and is marked by three distinct channel meanders from previous Garrett Creek courses. One old course of Garrett Creek is situated at the base of the upland projection about 40 m from the site. At the time of the prehistoric it is likely that this Garrett Creek channel was the primary stream course. The proximity of a stable, elevated landform with a permanent stream course is a rare topographic setting in this area where streams slowly meander through wide valleys and only infrequently flow near valley walls. It is an uplandfloodplain edge setting that was commonly selected for prehistoric settlement locations during certain periods of time (Bruseth and Perttula 1981: 133-138).
This part of the Lake Fork Creek-Garrett Creek drainage is in the Oak-Hickory or Post Oak Savannah biotic association. This association is a relatively narrow woodland band, and a natural transition zone between ·the Blackland Prairie to the west and the more mesic Oak-Hickory-Pine Forest of Pineywoods to the east. Upland areas were primarily covered with a widely . spaced medium-tall to tall post oak-blackjack oak savannah overstory, and pecan, oaks, hackberry, elm, sweetgum, and other hardwoods grew near the riverbank, and in the floodplains of the two streams. 1jhe interested reader is referred to Bruseth et al. (1977:5-19) for further details on the natural biotic communities in this section of the l:.ake Fork Creek basin.
41RA65 is situated on shallow Woodtellloarn soils averaging 27 em (10 in) in thickness (Lane 1977:23;Bruseth et al. 1977:165). In SMU's survey the site was recognized only as a scatter of ceramics and lithic cultural materials over a 300 x 60 m area of the upland projection. The total extent of the midden deposits now known to be present on the site is difficult to determine because of bulldozer disturbances, but probably covered at least 2000 sq m on the crest of the landform.
Similar types of prehistoric sites include 41RA66, 41RA79, and 41RA83 within a one km radius of 41RA65, and others are known within a 5 krn radius . The Gilbert Site (41RA13), which also has an Early Caddoan component (Jelks 1967:185), is about 1.2 km southwest of 41RA65.

INVESTIGATIONS
Our investigations at the site were limited to a surface collection of those areas where large quantities of cultural material had been exposed by bulldozing and road traffic. Because of this disturbance no attempt was made to either carry out any excavations, or employ systematic surface collection procedures. Instead we concentrated on: (1) collecting .all observable lithic tools and pottery sherds, regardless of size, and (2) selectively gathering lithic debitage and fire cracked rock. In the case of lithic debitage, our grab sample was accumulated with the intent of first documenting the diversity in raw material types present in the site assemblage; secondly, to amass a relatively representative sample of debitage classes; and finally, to gather a large enough overall sample that usemodified and/or small intentionally modified tools might be included by chance. Because of their bulk, no attempt was made to collect representative samples of fire cracked rock.
It should be pointed out that immediately prior to our investigations, a church group from Sulphur Springs, Texas had spent approximately one hour on the site collecting projectile points. About fifteen whole projectile points were removed from the site by this party, but we were able to ascertain that broken tools, lithic debitage, and ceramic sherds were not collected PREHISTORIC ARTIFACTS Over 500 artifacts were recovered in surface collections at 41RA65 (Table 1). Time -diagnostic lithic and ceramic artifacts present suggest that two components can be defined ) and Early Caddoan Period 1/11 (A.D. 800 -1400) in affiliation. Given the topographic setting of the site, the shallow cultural deposits, ;md the nature of the collection, 2 · it is virtually impossible to assign each artifact class or typological category to one compodent or the other. Associated chronological and typological information presented in , Johnson (1962), and Thurmond (1981,1985) was important in reasonably segregating certain artifact classes, and developing more specific temporal estimates for the two components at 41RA65.

10
• The X41RA9 site number was assigned by SMU POTTERY A total of 221 pottery sherds were recovered from the site. Grog (crushed sherds)-grit tempered pottery represents 86.4% of the sample, while bone tempered pottery accounts for only 13.6% (Table 2). The two pottery categories were further subdivided by differences in the apparent density of the temper in the paste (e.g. Rogers et al. 1985). Without attempting a detailed quantification of temper particles visible in sherd and also acknowledging the difficulties enumerated by Sheperd (1976:26) in assessing temper densities, we have separated the sherds into fine and coarse tempered wares. These divisions approximate the differences between 10% and 30% temper by volume employed by Rogers et al. (1985: Figure 2) in their study of temper quantification.
Coarse bone tempered pottery represents 5.4% of the ceramic assemblage, while the coarse grog-grit tempered class accounts for 30.3%. Sixty-four percent of the ceramics from 41RA65 are fine tempered pottery sherds; 8% of these are tempered with bone, the remainder having been tempered with grog and grit aplastics. Plain and decorated body sherds from each of these categories were measured for thickness, demonstrating that coarse groggrit tempered sherds are thicker than the other pottery groups (Table 3).
Variations in sherd thickness within the four groups, in combination with differences in surface treatment, suggest that two distinctive wares are present in the 41RA65 assemblage. They are: (1) a relatively thin bone and grog-grit tempered pottery of both fine and coarse paste, with slipped, engraved, and incised rim and body  (2) a relatively thick, coarse grog-grit tempered pottery decorated only with large punctations. The first group, comprising 69.7% of the ceramic assemblage, clearly belongs to the Caddoan ceramic tradition. The remainder of the ceramics, the second group, does resemble the Caddoan ceramics in a general sense, but here are suggested to actually represent an Early Ceramic Period ceramic assemblage at 41RA65. The identification of an Early Ceramic assemblage is based on recognized differences in sherd thickness between the two groups, paste characteristics, and similarities to the Resch Site, a well-described Early Ceramic Period component in the Upper Sabine Basin (Webb et al. 1969:18-43).
The Early Ceramic Period is characterized in East Texas by the development of two different ceramic technologies: a sandy paste ware and a grog-and sometimes bone-tempered ware (Story 1981:146). The latter is often identified as Williams Plain (Brown 1971:42-58), a thick (10 mm or more) ceramic ware dominated by "flowerpots" and simple bowls with flat disk bases. The thickness criteria developed in Southeast Oklahoma and Southwest Arkansas for catetgorizing and/or recognizing Williams Plain (e.g. Rohrbaugh 1985;Schambach 1982) often are not applied uniformly in East Texas studies, making direct comparisons difficult between different Early Ceramic Period assemblages.
In general, Williams Plain seems to occur in greater frequencies per site and in more sites north of the Sabine and Sulphur Rivers, while sandy paste wares are common from south of the Sabine River to the Gulf Coast (Story 1981:146). However, Williams Plain ceramics are not common in Upper Sabine Basin components dating to the Early Ceramic Period Webb et al. 1969).
The Early Ceramic component ceramic assemblage at the Resch Site is comprised of sand, bone, and clay tempered wares with only a limited amount of decorated pottery. The Resch ceramics average 7.69 mm in thickness, ranging between 7.0-8.3 mm for the different wares (Webb et al. 1969: Table 1). The Williams Plain identified from the site included clay and bone tempered ceramics between 7-9 mm in thickness. The decorated pottery includes Tchefuncte Stamped, Churupa Incised, Marksville Incised var. Yokena , Troyville Stamped, and Marksville Stamped, all Lower Mississippi Valley types dating prior to ca. A.D. 500 (Williams and Brain 1983!Figure 12.1). While the four corrected radiocarbon dates from Resch range from 500 B.C. to A.D. 125 (e.g. Webb et al. 1969:95), their stratigraphic contexts are too mixed for these to be regarded as suitable dates.
Defined Early Ceramic components at Lake Fork Reservoir are suspected to date later than those at Resch because the ceramic assemblages consist mainly of horizontally incised decorative motifs confined to vessel rims, along with other incised and punctated motifs (Bruseth and Pertulla 1981:Table 5-9). This method of 3 ceramic decoration is analogous to Coles Creek Incised. Diagnostic Coles Creek Incised pottery, including var. Coles Creek and var. Greenhouse, have been recovered from Grace Creek and Resch (Jones 1957;Webb et al 1969). These varieties of Coles Creek Incised date to ca. A.D. 700 -850 in the Lower Mississippi Valley (Williams and Brain 1983).

Caddoan Ceramic Assemblage
Only eleven sherds in the Caddoan ceramic assemblage were decorated ; this amounts to 7.1% of the sample. An additional 3.2% have a hematite derived red slip or film, and 15% were burnished ( Table 4). Because of differential preservation of sherds it is likely that the percentages of slipping and burnishing are underrepresented in the assemblage. Partially eroded and water worn sherds were noted in the collection. and Pertulla 1981: Table 5-4). The red slipped pottery from 41RA65 is present in both bottle and bowl vessel forms similar to whole vessels recovered at the Yarbrough Site on the Sabine River in Van Zandt County, Texas (Johnson 1962:Figure 22c, g).
The three engraved sherds are represented by two direct standing rims (Brown 1971: Figure 2) and a body sherd from an undetermined vessel form (Figure 1c, d) .
One of the engraved rim sherds derives from a carinated bowl, a common vessel form for Sanders EQgraved (Suhm and Jelks 1962:137). Motifs identified include diagonal engraved lines along the rim panel, and an engraved-filled triangle (probably pendent from the rim or from horizontal engraved lines). The sherd and motif are too small to discern the complete decorative element. Engraved-filled triangles are an important stylistic marker for post-A.D. 1000 Caddoan sites in the Lake Fork Reservoir (Bruseth and Perttula 1981 :86).
Incised vessels are represented by 7 sherds (Figure 1e, T he high frequency of plain rims , indicating undecorated vessels, and the low overall representation of decorated sherds, is compatible with other Early Caddoan occupations in the Upper Sabine Basin (Bruseth and Pertulla 1981;Duffield 1961;Johnson 1962). The red slipped pottery is classifiable as Sanders Plain (Brown 1971), a consistent component in ca. A.D. 1000 -1400 context in many Caddoan sites in the Western Gulf Coastal Plain (Ferring and Perttula n.d.). The identification of Sanders Plain in the assemblage at 41RA65 need not imply that an A.D. 1200 -1400 Sanders phase or Early Caddoan Period II component (e.g. Thurmond 1985: 189) is present because red slipped plain wares are common from ca. A.D. 900 in the Lake Fork Creek basin archaeological sites. For example, at the Taddlock Site (41WD482), radiocarbon dated between ca. A.D . 960 -1150, slipped pottery accounts for 3.5-6.7% of the 2,895 rim and decorated ceramics from the three middens of this Pecan Grove phase component (Bruseth 1h). Both diagonal and cross-hatched incised decorative elements common to the type Canton Incised (Suhm and Jelks 1962:23) are present. At the Osborn site (41WD73) in Lake Fork Reservoir, diagonal incised motifs accounted for 60% of the Canton Incised sherds. The Osborn assemblage has been dated to A.D . 775 ± 68 corrected). By contrast, diagonal incised Canton Incised represented less than 40% of all Canton Incised sherds at the ca. A.D. 960 -1150 Taddlock occupation, and only 20% from the later Spoonbill (41WD109) component (Bruseth and Perttula 1981:Table 5-8). While changes in the frequency of diagonal and cross-hatch incised motifs apparently have temporal significance in the Upper Sabine Basin, the sample from 41RA65 is too small to be compared with the Lake Fork Reservoir sites in this respect. Only one small punctated sherd of a fine grog-grit temper was recovered in the 41RA65 surface collection. The decoration is apparently a row of fingernail punctations. Punctations in an Early Caddoan context do occur in association with incised motifs, as well as an independent decorative element A variety of rim shapes are present in the Early Caddoan ceramic assemblage (Table 5). Standing rimsstraight or vertically oriented -account for 92% of the rims in the small sample. A direct rim is one that has no change in thickness or orientation in the vessel contour (Brown 1971:19), while beveled, rolled, and thickened rims are elaborations to the body of the rim.  Ten of the rims are plain, though three have lip notches or scallops (Figure lb). The vessel with lip notching is a small simple bowl. Vessel lip notching has been noted in an Early Caddoan context at the Yarbrough (Johnson 1962:Figure 23i), Limerick (Duffield 1961:88), and Sanders (Krieger 1946 Sites in East Texas. Occasionally lip notching is associated with the addition of a thin strip of clay added to the interior of the vessel to create an interior thickened rim, though this was not done on the 41RA65 specimen. The scalloped rims have been thickened at the lip, but do not have an interior thickened or collared profile. Scalloped rim Sanders Plain and other undecorated Caddoan wares are present in the large collection from the Sanders Site (Krieger 1946:Plate 24b, c).

Possible Early Ceramic Assemblage
The small punctations were apparently made with a sharp pointed implement that did not displace the still-plastic clay on the vessel surface ( Figure 1g). This design is present on a carinated bowl.
The one rim is from a plain carinated bowl ( Figure  la). The bowl has a direct and standing rim (Table 5).

LITHIC ARTIFACTS
The lithic artifact assemblage from the site consists primarily of lithic debitage, cores, and bifacial tools (Table 1). A few unifacially worked flake tools ace present, while the remainder of the lithic tools include an arrowpoint, and a small fragment of a polished axe.
The majority of the lithic debitage and tools are on Ogallala chert, followed by quartzite, and other types of chert (Table 6). With few exceptions, these lithic raw materials are locally available in upland stream divides and ridges as gravels of palm and fist-sized cobbles (Bruseth and Perttula 1981:Table 6-7). Quartzite and ferruginous sandstone also occur in outcrops of Eocene age Weches and Reklaw formations in the Upper Sabine Basin (Perttula et al 1986:449).
The unmodifted lithic debitage from 41RA65 has been categorized following the method of debitage analysis formulated by Sullivan and Rozen (1985: Figure  2). This approach differs from others employed in East Texas because it is based on dimensions of flake margins, positive percusion features, and the presence of a point of applied force (i.e. presence of a striking platform) that can Four punctated sherds were recovered from vessels that were manufactured using a thick coarse grog-grit temper/paste. The punctations appear to be aligned in parallel rows, probably on the rim panel (Figure lf  be assessed independently of hypothetical reduction sequences or particular methods of tool production (Sullivan and Rozen 1985:758). Categories recognized include complete flakes, broken flakes, flake fragments, and debris ( The largest category of flake type in the assemblage is debris, lithic artifacts lacking a striking platform, a bulb of percussion, and with margins that are not intact. These artifacts are generally small in size, and analogous to the term chip used in some East Texas typologies. Debris accounts for 51.1% of the debitage in the assemblage, followed by flake fragments, broken flakes, and complete flakes (Table 7).
There is little difference between flake types and the percentage of flakes with cortex, indication that the different kinds of flakes were produced as part of the overall process of tool and core reduction carried out on the site. The high proportion of debris, flake fragments, and broken flakes at 41RA65 also suggest that both tool manufacture and core reduction activities were important lithic technological activities. This is further attested to by the frequency of cores, and broken bifaces that are discarded manufacturing failures. The comparability in proportions of cortical debitage among the four flake types would be expected in a lithic technology where cobbles and partially cortical cores were specifically reduced in order to produce useable, complete flakes for tools. This interpretation seems to be supported if one notes the high relative frequency of complete chert flakes in the lithic debitage (Table 7}, and the percentage of complete flakes classified as utilized pieces (Table 8). Complete flakes are two to five times more likely to be selected for use as tools than the other types of flakes in the 41RA65 assemblage.
The utilized pieces constitute a morphologically diverse set of flake tools, ranging from those exhibiting a few small continous flake scars along lateral and/or distal  (Figure 2h, i) to those where retouched flake scars are more uniform, larger in size, and continous along the margins of the tool (Figure 2g). In the latter case, the uniform flake removals imply more extensive use and/or intentional retouching of the tools. The one arrow point from 41RA65 ( Figure 2a) has a contracting stem, straight-concave blades, and distinct, slightly barbed shoulders. It is similar to Form 2 arrow points defined at Lake Fork Reservoir (Bruseth and Perttula 1981:Figure 6-1), and to the Minter type identified by Johnson (1962:250). In both cases, these contracting stem arrow points occur on archaeological sites dating after A.D. 800, but probably before ca. A.D. 1400.
Two of the projectile point/knives (PP/K) were manufactured of a coarse grained, non-heat treated, red quartzite, while the others were made of locally available Ogallala chert. All of the PP/K are classified as the Gary type, var. Camden (Schambach 1982). Schambach defined three varieties of Gary points -var. Camden, var. LeFlore, and var. Garythat represent modal trends in thickness and stem width which have chronological implications. The var. Camden Gary points are the thinnest, and have the narrowest stem width of the three varieties, and are considered the latest expression of the type. Schambach's findings are summarized in Table 9. The modal stem width of 1.4 em, and modal thickness of 0.6 em, in the small sample of Gary points from 41RA65 is generally consistent with the metrical parameters of the var. Camden (Table 9). Moreover, its presence is contextually appropriate for an Early Ceramic Period component at the site.
Based on attributes of blade asymmetry, serration (  tached to a foreshaft. Evidence of edge crushing and step fractures on the lateral margins suggests these implements were used in cutting and scraping tasks. The presence of impact fractures on two of the specimens (see Figure 2b) also indicates the tools were used as projectiles; the Gary PP/K is truly a multi-purpose tool.
Ovoid to triangular bifaces are common at 41RA65. These bifaces range from pieces without well defined working edges and/or zones of utilization on edges or faces (thick bifaces) to those with regular shapes and well defined working edges across the tool. Of the latter (thin bifaces), eight of twelve are fragmentary pieces that were apparently broken during fmal manufacture. Wear patterns are identifiable on several specimens (though wear patterns are difficult to observe on the coarse grained quartizites collected in local gravels), primarily low to medium angle unifacial flaking (Bruseth and Perttula 1981 :Table 6-5). Items included within this class represent both completed, but fragmentary, bifacial tools, and bifaces that are preforms which have not yet been proximally modified for hafting. In general, thin bifaces are manufacturing by-products in the production ofPP/K. The thick bifaces are not finished tools. They represent instead initial attempts in the reduction of cobble-sized pieces of raw material, for the eventual production of PP/K, that were discarded before substantial thinning could be accomplished. Knapping failures, raw material inclusions, and poor quality raw materials all are contributing factors in these thick bifaces not being completed. Cortex remnants on the thick bifaces also suggest that pieces of raw material were small in size, ranging from 4-10 em in length and 3-6 em in width. The cores from 41RA65 are generally about the same size, and hardly suitable for the production of large tools, but flakes for unifacial tools, arrow points, and some of the PP/K are possible to manufacture with cobbles of these size ranges.
While both thin and thick flakes are uniformly manufactured from local upland gravels (Table 6), one thin biface of a distinctive white chert was noted. This white chert, which has macroscopically visible black inclusions, resembles Frisco chert from the Arbuckle Mountains in Southeastern Oklahoma (Banks 1984), but its specific origin has not been traced as yet. This lithic raw material has been noted at a number of prehistoric sites in the Upper Sabine Basin, albeit in an Archaic context . Cherts from the Arbuckle Mountains region were noted also in Lake Fork Reservoir sites dating from ca. 4000-400 B.P., but this non-local material was not as common as raw materials from the Edwards Plateau or the Ouachita Mountains (Bruseth and Perttula 1981:Table 6-10).
A total of nine cores were recovered from the surface collection at 41RA65. Eight of the cores are of Ogallala chert, and one is a black chert similar to the black variety of Big Fork chert (Mallouf 1976:49-50). The nearest source of Big Fork chert is in Red River gravels below the Kiamichi River confluence in the vicinity of the Sam Kaufman Site (Skinner et al. 1969). This particular core is a small river cobble, 3 em in length and width, with gravel cortex covering approximately 50% of the cobble.
The cores definitely of local chert were collected and brought to the site where initial to final manufacturing took place. Flakes have been removed from unprepared platforms, whether single, multiple, or opposed in orientation. None of the cores have been depleted since gravel cortex is visible on both the dorsal and ventral surfaces of the cobbles, but because of their small size it would be difficult to remove many additional flakes to thin the cobble without actually splitting the cobble inadvertently. One of the cores actually represents a substantial remnant of a split and fractured core.
One fragment of a polished axe was found at 41RA65. The axe was manufactured from a locally available ferruginous sandstone.
Quartzite and chert cobbles were also utilized in cooking activities. Nine rocks from the surface collection ( Table 6) show evidence of heating sufficient to have fractured the cobbles from exposure to a heat source.

TEMPORAL AND FUNCTIONAL CONTEXT
No radiocarbon dates have been obtained from 41RA65. Based on the styles of the PP/K and the ceramic wares recovered from the disturbed surface of the site it is apparent that the site was occupied during both the Early Ceramic and Early Caddoan Periods.
The Early Ceramic Period in the Upper Sabine Basin, as well as in East Texas generally, is poorly known at present (Perttula et al. 1986:53-54). It represents a particularly important cultural expression, however, because it bridges the gap in time between non-sedentary hunter-gatherers and the evolution of sedentary horticulturists in the region (e.g. Story 1985). Within the Upper Sabine Basin components belonging to this time period have been identified at several sites, including an Early Ceramic midden excavated at the Resch Site (Webb et al. 1969), a small component at the Yarbrough Site, Area A (see Johnson 1962:206), and six components at Lake Fork Reservoir .
None of· the components at Lake Fork contained Williams Plain ceramics, the thick grog tempered ware usually considered a diagnostic for the Early Ceramic Period. Additionally, several components are suspected to date after ca. A.D. 500 because of the presence of Friley arrow points, other arrow points, and ceramic assemblages dominated by horizontally incised decorative motifs confined to vessel rims. A date of A.D. 775 ± 68 from the Osborn Site was the only radiocarbon date obtained from the Early Ceramic components. The Osborn occupation probably marks one of the latest Early Ceramic components (Bruseth and Perttula 1981:141), or one of the earliest Early Caddoan Period I occupations, in the reservoir. A re-examination of the cultural content of the Early Ceramic assemblages from Lake Fork suggest that based on temporal and stylistic groups, components of the period recognized there overlap with both Early Ceramic and Early Caddoan Period I assemblages defmed elsewhere in East Texas (fhurmond 1981, 1985).
The Early Ceramic component at 41RA65 is estimated to date from ca. A.D. 200 -700, based primarily on the occurrence of Gary var. Camden PPIK. Associated with these tools is a lithic artifact assemblage not significantly different from that of the local Late Archaic in that it is dominated by bifacial cutting and scraping tools and quantities of cores. The frequencies of bifaces, cores, and PP/K in Late Archaic and Early Ceramic assemblages at Lake Fork Reservoir are quite different from Early and Late Caddoan assemblages (Bruseth and Perttula 1981:Table 6-4). It is suspected that the majority of the bifacial tools, cores, and debitage are associated with the Early Ceramic occupation, rather than with the Early Caddoan settlement at 41RA65. Certainly, the predominate utilization of locally available lithic raw materials such as Ogallala chert, petrified wood, and quartzite is consistent with the Early Ceramic pattern of lithic raw material use noted in Lake Fork Reservoir (Bruseth and Perttula 1981:Table 6-9). The utilization of non-local materials for tool manufacture was found to be quite low in Early Ceramic Period lithic assemblages, about three to six times lower than in Early Caddoan sites. Early Ceramic groups employed local materials in such high frequencies due in part to territorial constraints, reductions in exploitation range, and localized interregional exchange (Perttula 1984).
The lithic assemblage recovered would seem to suggest that one of the primary activities at the site during the Early Ceramic occupation includes lithic tool production and maintenance, hunting, butchering and cutting of meat products, and some plant food processing. Cooking and storage of plant and animal foods are indicated by the use of ceramic containers, as well as by the presence of fire cracked rock.
While the midden deposits at 41RA65 probably date primarily to the Early Caddoan Period, Early Ceramic settlements in the Upper Sabine Basin do have earthen middens. At the Howle, Block I (41WD74), and Osborn Early Ceramic components earthen middens were present with storage and trash pits, and other indications of relatively permanent multi-seasonal occupations. These types of occupations have been interpreted as singlefamily homesteads occupied for one to two generations (Bruseth and Perttula 1981:141). With only limited evidence on settlement available from 41RA65 we are hesitant to ascribe to it a particular settlement type or function. This is an analytical effort that requires extensive archaeological investigations and regional landuse information.

10
In the Early Caddoan occupation of the Upper Sabine Basin the most common types of settlement are locales containing one to three middens per site. These middens represent small habitation areas of fairl y brief occupational span: perhaps 50 -100 years if the chronological record from the Taddlock Site, a well preserved example of such sites, is accurate. Three radiocarbon dates from Taddlock document an occupation dating from A.D. 1037 ± 74 to A.D. 1070 ± 77 (Bruseth and Perttula 1981:48-53). The middens have been found either to cover house locations or represent concentrated trash deposits of broken ceramic vessels, bones, plant remains, and other refuse. Because of the fact that the midden deposits at 41RA65 were disturbed, there seemed little way to determine its nature without designing an excavation program to look for subsurface features and in situ cultural remains on the site. Since the midden appeared to have been relatively large it is probable that a large part of it is composed of sheet trash dispersed across the flatter parts of the landform, though domestic structures were likely present at the site.
Early Caddoan midden sites in the Upper Sabine Basin are usually classified as sedentary hamlets and farmsteads. Not only are they located near arable soils, but flotation of midden deposits at a number of sites have documented that maize (Zea mays L.) is ubiquitous, even though other wild plant foods were also important parts of the economy . In addition to nuts 'and maize, seeds from 15 wild plant species were recovered in flotation samples from the Early Caddoan component at the Spoonbill Site, about 20 km east of 41RA65.
The Early Caddoan settlement at 41 RA65 took place about 650-800 years ago, ca. A.D. 1150 -1300. The presence of Sanders Engraved, Sanders Plain, scalloped rim bowls, and Canton Incised in the ceramic assemblage seems congruent with this general placement. though a temporal estimate of ca. A.D. 1200 -1400 could be entertained (e.g. Thurmond 1985: 189). Our suggested chronological context for the settlement at 41RA65 is founded primarily on a proposed seriation of radiocarbon dated Early Ceramic and Early Caddoan component design elements from Lake Fork Reservoir sites (see Bruseth and Perttula 1981:Table S-9). Of particular interest are changes in the relative proportions of Canton Incised, Davis Incised, Sanders Engraved, and punctated designs ( Table 10). The admittedly small ceramic sample from 41RA65 is placed in the seriation between the Hines Site (41WD450) and the Spoonbill Site, Early Caddoan occupations with house construction features and other evidence for permanent settlement (Bruseth and Perttula 1981:21-26, 40-48).
Caddoan components investigated at Lake Fork. Reservoir

CONCLUSIONS
While the Lake Fork. Creek area has been the scene of relatively intense archaeological investigations since the rnid-1970s it is fair to say that the regional archaeological  :484) support the Spoonbill dates.
The Hines and Spoonbill dates effectively bracket the period between ca. A.D. 1150 -1300 as the span of time in the Early Caddoan Period 1/II when it is most likely that 41RA65 was occupied. It is doubtful, however, that the site was continously occupied throughout that period. Instead, any one Early Caddoan settlement at the site was probably of short duration, though we have no concrete evidence that the site was sequentially re-occupied within the Early Caddoan Period. We suggest the possibility of multiple occupations from the fact that the midden deposits are fairly extensive compared to other Early record is structured more in terms of basic-temporal spatial analyses than it is to broader conceptualizations of cultural change. That is not to suggest that advances in processual studies, i.e. questions concerning settlement pattern types, subsistence, and models of cultural systems, have not taken place in the area, or in the Upper Sabine Basin in general. Rather, significant information on the diversity and variability in the regional archaeological record, particularly for the Caddoan sequence, has -7 been obtained here that is quite comparable to important developments elsewhere in East Texas and the Caddoan Area. Future prospects are exciting. At the same time, we are faced with relatively pedestrian, but integral, questions of chronological context because of a poorly developed sequence of radiocarbon dates, as well as from an overall lack of interest in refining chronological problems (Thurmond 1985:188). The need for chronological control is still a critical parameter in developing a more systematic understanding of prehistoric adaptations in East Texas since a local chronology is necessary if questions about cultural change are to be addressed. Our concern with process must be tempered with this grounding in chronology; the study of Caddoan archaeology needs to proceed simultaneously at both levels.
In this paper we have endeavored to present information on site assemblage character and age from a multi-component Early Ceramic and Early Caddoan settlement in the Upper Sabine Basin. The site is not pristine, far from it, but we believe that the site contains data of local and regional importance. This is especially so because of the continuing research emphasis on changing land use in East Texas and the Upper Sabine Basin (e.g. Thurmond 1981). Construction and development activities associated with the impoundment of waters for the Lake Fork Reservoir unfortunately led to major disturbances at the site. 41RA65 was known from earlier surveys; it is a moot point now as to whether or not the site warranted additional investigations before construction activities and inundation proceeded. However, we believe that the continued study of such unprotected sites as 41RA65 is essential in the Upper Sabine Basin. Without further work at sites of these time periods and character, a full understanding of settlement and land use through time will not be possible before further development activities permanently destroy the archaeological record of the region.

INTRODUCTION
Because faunal and floral preservation is generally poor in the acidic soils of East Texas, the consideration and interpretation of prehistoric subsistence patterns in this region of the Caddoan area rests more on the available ethnographic information than on the archaeological record itself. Excavations at the Taddlock (X41WD39) and Spoonbill (41 WD1 09) sites in Wood County on Lake Fork Creek in the Upper Sabine River Basin produced well preserved and abundant faunal and floral remains from Early Caddoan period Sanders focus context (Krieger 1946). The material recovered from these two sites affords considerable insight into the nature of Early Caddoan subsistence at a time of funda-mental culture change in East Texas prehistory (Story 1981 :149).
In this paper we document and explicate the subsistence strategies of sedentary hamlet occupations during the Early Caddoan period: first, by summarizing the faunal and floral data from the two sites, and second by comparing these sites with other published Caddoan sites of this and later periods.

THE SITES, NATURAL SETTING , AND CULTURAL BACKGROUND
The Taddlock and Spoonbill sites are located in the Oak-Hickory biotic association or Post Oak savannah in East Texas. This association is a narrow swath of woodland and is a natural transition zone between the more xeric Blackland Prairie to the west and the more mesic Oak-Hickory-Pine Forest or Pineywoods to the east (Fig. 1 ). Medium-tall to tall broadleaf deciduous forests characterize the Oak-Hickory forest association, with post oak and blackjack oak most common (Gould 1969). The Blackland Prairie to the west consists of medium-tall and dense little bluestem grassland (Kuchler 1964:76). The Pineywoods to the east of the sites are similar to the Oak-Hickory forest with the exception that shortleaf and loblolly pine are also dominant species. The presence of pine is usually assumed to represent a subclimax or fire disclimax (Mahler 1973) and thus due to the less mature nature of the association.
The climate of the Upper Sabine River Friends of Northeast Texas Archaeology, Special Publication No. 1 (1995) 0 llackland Prairie Fig. 1. Major biotic association in the Upper Sabine River Basin (after Kuchler 1964) and sites mentioned in the text.
Basin is humid subtropical with average winter temperatures of 47° F and summer temperatures of 83° F; droughts are not uncommon . Precipitation averages around 43 inches per year, with seasonal variations in amounts. The majority of precipitation falls in the spring and fall months. The growing season ranges over 230 days. The Early Caddoan period in East Texas dates from ca. A.D. 750 to A.D. 1350 (Story and Valastro 19n). In the Upper Sabine River Basin , sites of this period are included in the recently defined Pecan Grove phase (Bruseth and Perttula 1981 :141-142). Excavations at 16 components of this phase were conducted between 1975 and 1979, with extensive investigations primarily carried out at the Taddlock and Spoonbill sites in 1979 (Bruseth et al. 19n;Bruseth andPerttula 1980, 1981 ).
Both sites apparently consist of a small number of house structures, a " plaza" between houses where features and activity areas are present, and adjacent trash middens or refuse pits with good to excellent ecofactual preservation. Pecan Grove phase settlements are present on both major and minor tributaries of the Sabine River and on the Sabine River itself (Skiles et al. 1980: Figure 5). Intra-and inter-site complexity during the Pecan Grove phase was minimal, with most sites either similar to Taddlock and Spoonbill in the nature of their occupation or apparently short-term, limited activity, sites 16 scattered around the permanent settlements. This suggests that such sites represent the lowest level of regional Caddoan settlement and may be referred to as farmsteads (Story 1981 : 150). Four C-14 dates from Taddlock indicate an occupation dating from ca. A.D. 940 to A. D. 1 000, while five dates from Spoonbill in combination with associated artifact styles support components dated ca. A.D. 970 ± 65 and A.D. 1260±65.

FAUNAL AND FLORAL ANALYSIS
Excavation and analytical procedures at these two sites have been described elsewhere (Bruseth and Perttula 1981 :11 , 40-53, 117). Suffice it to state that all soil excavated was screened through 1/1 6 inch mesh, and uniform volumetric flotation samples were removed for processing from all excavation units, including general matrix and feature matrix. The flotation samples were processed by a watersluicing device (Bruseth and Carter 1980) that passed soil through four sizegraded screens: one heavy fraction grade: 2.0 mm mesh; and three light fraction grades: 1.68 mm, .707 mm, and .25 mm mesh (for the purposes of this paper all material separated by screen size has been combined ; further details are on file at the Archaeology Research Program, Southern Methodist University).
Only the faunal material from Taddlock and the floral material from Spoonbill will be discussed in this paper. This is because only small quantities of floral material were recovered from Taddlock (summarized in Table  5 below), and less than 200 poorly preserved faunal elements were found at Spoonbill. It is unfortunate that complementary subsistence information could not be secured from each individual site, but the strong patterning between Taddlock and Spoonbill in the nature of the site occupation and topogi'aphic/environmental location support the comparative analysis of subsistence strategies during the Pecan Grove phase on the basis of the two partly contemporaneous sites.
A well preserved trash midden (Midden A) was excavated at the Taddlock site. A total of 63,463 faunal elements below the disturbed plowzone were recovered , primarily from the 1/16 inch mesh residue. The majority of the flotation samples were not processed due to a shortage of funds, but several were examined which contained quantities of small elements such as fish vertebrae. Based on the 70% of the midden that was excavated, the total midden faunal content of 1/16 inch material is estimated at 90,000 identifiable and unidentifiable elements. Faunal material is present at approximately 5,900 elements per cubic meter of cultural deposits.
Animals from all vertebrate classes are represented; mammals, reptiles, and fish were particularly important resources (Table  1 ). Those elements considered identifiable were classified to the lowest taxonomic level possible (see discussion in Butler and Perttula 1981 :117).
Although mammal bones do not make up the largest portion of the sample, mammals were important to the site's inhabitants since they contributed proportionately a large amount of meat per animal ( Table 2). Four mammals were most common: rabbit (eastern cottontail and swamp), jackrabbit, squirrel, and deer. These represent about 90% of the identified mammal sample.
White-tailed deer is the most frequent mammal species represented by element and certainly contributed the greatest quantity of meat to the diet, about 55% of the total estimated meat yield. The total number of deer elements at Taddlock is not, however, in the high proportion seen at other Caddoan sites (Byrd 1980;Doehner et al. 1978:Tables 53 and 54; Henderson1978: Table 37). Based on dental wear, four age groups are represented: individuals between one and eleven months, 11J2 to 2 years, 4 to 5 years, and 5 to 7 years old at the time of death. The sample is too small for quantitative interpretations of relative age distributions, but there is no evidence for preference of a certain age group (cf. Emerson 1980).
The abundance of smaller species, especially squirrels, jackrabbits, and rabbits, indicates a regional and/or local abundance in the Oak-Hickory savanna habitat. Over 80% of the Minimum Number of Individuals (MNI) for mammals are accounted for by the smaller species (less than 5 kg). This suggests a low availability of larger species (Bayham 1979).
Of the birds from Taddlock, only the wild turkey is represented in quantity. Males were preferred, though females and juveniles are 17 also present. Seven of the identified bird species are permanent East Texas residents, while the others, such as the pied-bill grebe, green-wing teal, and yellow-shafted flicker, all winter in East Texas and are there from approximately September to April.
The large number of turtle shell recovered indicates that turtles were an important economic resource. Both aquatic and land turtles were exploited, with the mud turtles and box turtles most frequent, but the snapping and slider turtles contributed the most meat to the diet ( Table 2).
Several fish species were found at the site, all of which are common to rivers and streams of East Texas. The gar elements are either from shortnose or spotted gar and were probably collected in the summer through early fall after spawning. Bowfins like relatively clear, quiet waters with abundant aquatic vegetation and reach 15 pounds in weight (Byrd 1980:259). Both catfish and sunfish/bass can be found in a range of aquatic habitats, from small, clear streams to large rivers and sloughs (cf. Smith 1975: Table 14). These two species groups accounted for approximately 60% of the identifiable fish sample and about the same amount of meat as the suckers, the largest of the fish species represented in the Taddlock sample.
All of the species of animals recovered from Taddlock are either year-round or seasonal inhabitants of East Texas. The habitat range of species indicates that most animals were associated with an open Oak-Hickory forest with many grassland areas. More extensive grassland areas were probably exploited for jackrabbits and prairiechickens. The forested bottom lands were exploited for various medium-sized mammals, as well as for birds. Fish were taken from Lake Fork Creek and its sloughs. The boundary or edge areas between the wooded bottomlands and the more open upland woods and grasslands were exploited for deer, terrestrial turtles, and wild turkey. Since these habitats are linear in orientation, with little width to each, most of the species of animals recovered from the site could have been procured within a relatively short distance of it.
The absence of three species at Taddlock, dog, bear, and bison, is notable given the historic utilization of these species by Caddoan peoples of East Texas (Swanton 1942:     (1975) and Smith (1978) for deer ''T • Trace '• ·Averages ol the valuee presented in Smith (1975: Tables 33 and 34) 134-137). The presence of dog at the nearby Steck ( Fig . 1) and Arnold sites (a contemporaneous occupation 40 km to the north [ Henderson 1978]), does show it was utilized in the region, at least for hunting purposes. Dog burials were common at these sites. Larger mammalian species with low reproductive rates, such as bear, do not seem to have been as frequently utilized during the Mississippian period in the southeastern United States as smaller species (Robison 1982), while bison was never very common in the East and North Texas prairies. In fact, it is only after ca. A.D. 1200 that bison populations seem to have increased in the Blackland Prairie to the west (Lynott 1980:99), or at least that different procurement strategies focusing on bison developed at that time (Ferring 1982). In any case, there is no evidence for the exploitation of bison in the Sabine Basin until the eighteenth century (Lorrain 1967:Table 11 ), even though the Blackland Prairie is only 25 km away as the crow flies from these sites.
At Taddlock, the wide and diverse range of species present shows the relative balance in which classes were utilized and a broadspectrum exploitation of the surrounding area. A diversity measure was utilized to quantify the variability in the faunal population: = 1.00-S. The diversity index Aw is determined by summing the squared percentage of each variable (S}, in this case classes mammals, birds, reptiles, and fish, and subtracting that number from 1.00. The resultant figure will range from 0.0 (homogenous population) to 1.00 (heterogenous population). Using the MNI of the different classes as the variables in the computations, a diversity index of .675 is obtained. This indicates a considerably heterogenous faunal population at the Taddlock site.
The intensity of animal groups exploited varied, however. Fish and turtle were intensively collected as were deer, squirrel, rabbits, and turkey. On the other hand, though many birds were locally available, including migratory fowl, their rarity suggests that they were not even important seasonal resources. Mammals contributed the greatest percentage of total meat to the Early Caddoan population at Taddlock (Table 3), with fish ranked second, followed by birds and amphibians/reptiles. Based on projected meat yields, the ranking of the 1 0 major animal species at Taddlock is as follows: deer, 55.5%; carp sucker, 7.1 % ; catfish, 6.3% ; opossum, 5.1 %; turkey, 3.2%; squirrel, 3.1%; jackrabbit, 2.1 %; freshwater drum, 1.8%; raccoon, 1 .8%; and beaver, 1.5%. Certain behavioral characteristics of the vertebrates present at the Taddlock site suggest it was utilized on a year-round basis. The gar and hibernation habits of the raccoon and some of the turtles indicates these animals were most likely procured during the late spring, summer, or early fall periods. A young fawn that died around June shows a summer occupation. Deer crania with shed antlers, plus the occurrence of migratory fowl like the grebe, teal, and flicker, point to a late fall , winter, and early spring occupation. In combination, this information is indicative of a yearround permanent settlement, probably over a relatively short span of time, perhaps two to three generations, by occupants of the Pecan Grove phase.
All floral remains recovered from the Spoonbill site were recovered through the flotation of 268,000 cm 3 of soil from six features. Evaluation of the data must take into consideration the many different factors pertaining to plant carbonization (see Dennell 1976). The vast majority of carbonized plant remains belong to plant foods with dense, inedible parts such as large fru it pits, nut and acorn shells, and corncobs and/or cupules, rather than to nondense plant foods with a high water content or those plant foods normally ingested in entirety. Determinations of economic importance, as well as quantitative treatments, must therefore be assessed by as many means as possible, including the total range of the plant resources utilized, their context, and the type of inferred activities in-volved in collecting and processing specific resources.
The identified floral material from Spoonbill was carbonized through in situ burning and derives from 6 of the 11 cultural features recorded at the site; the flotation residue from the other five features has yet to be studied at this date. Two of the six features are claylined hearths which contained the majority of the nutshell and seeds; three are basinshaped pits between 15-25 em in depth, and one resembles a smudge-pit (Binford 1972: 41;Guy 1981 :1 05) and contained the majority of the corn cupules and at least six corncob sections and carbonized cornstalk (Bruseth and Perttula 1981 :46).
Quantities of carbonized seeds were recovered from the Spoonbill site (Table 4). Identified forest fruits such as grape, berry, persimmon, and passionflower were probably gathered along streams and in open woods and thickets. The fruits are available in the late summer and fall, and were likely stored and dried for winter use. Persimmon fruits could have been made into beer, bread, or eaten raw. One grain, marsh millet, was recovered. Marsh millet was presumably available during the fall in specific aquatic locations such as sloughs and small remnant channel lakes in the alluvial bottomlands of Lake Fork Creek and the Sabine River. Knotweed and chenopods are herbaceous annuals that produce an abundance of seeds in the middle to late summer. The small number recovered suggests they may be incidental anthropogenic plants rather than food products. Wild beans are annuals that also grow on disturbed ground and open areas. The seeds could be eaten raw or boiled when available in the summer; roasting and cooking the flour into gruel has also been suggested (Shea 1980). Since beans generally preserve poorly (Gasser and Adams 1981: 183-184), the significant numbers found at Spoonbill probably do represent food remains.
Sunflower and marshelder are present in small numbers. Marshelder is a weedy annual that grows in small, concentrated patches along streams, sloughs, and in disturbed and  1978:309). Boiling or roasting the seeds produces a valuable oil high in carbohydrates and fats. The Spoonbill achenes measured have a mean length of 2.9 mm (Crane 1982), approximately the size of modern sumpweed species rather than cultigens (Yarnell 1978:297, Table 2). Sunflowers have an extremely high calorie, protein, and fat composition, particularly when processed as a flour (Watt and Merrill 1963). Carbonized nut remains are common at Spoonbill and Taddlock, and at a number of Early Caddoan sites in the Sabine River Basin  Table 7-7). Oak, hickory, pecan, and walnuts were utilized, with hickory accounting for over 99% of the Spoonbill sample by weight. Hickory is the most abundantly represented nut species at the George C. Davis site on the Neches River, occupied ca. A.D. 780-1260 (Jackson 1981).
Acorns are more common only at Taddlock and other Early Caddoan Pecan Grove phase sites dating prior to ca. A.D. 1100. Both black oak (i.e., willow oak and blackjack) and white oak (white oak and post oak) groups were utilized. Since acorn nutshells are less dense than hickory nutshells, their presence is probably well under-represented (ct. Chapman and Shea 1981 ).
Corn is the only tropical cultigen recovered from Early Caddoan flotation samples in the Sabine Basin; the common bean and squash are present in small quantities at the nearby Late Caddoan Steck site ( Fig. 1 ). Maize is present in flotation samples from every Pecan Grove phase site excavated in the Sabine Basin (Bruseth and Carter 1980;Bruseth and Perttula 1981 :Table 7-8), and in context at least, is comparable to the widespread utilization of nuts. Its frequency within sites and features is relatively low, however, suggesting other factors should be taken into consideration in evaluating its importance from the paleobotanical record. Corn may be overrepresented relative to other plant remains when the high correlation between smudge pits and carbonized corn cobs is noted. This points to a higher probability of preservation when the " food stuff" is also used as a fuel. In a preservation context similar to the nutshells, the kernels and cupules seem to represent traces of food preparation/consumption and incidental inclusions resulting from proc-essing activities such as corn parching. In these instances, the quantity and density of nutshells is considerably more than corn at Spoonbill and Taddlock. The fact that only one of the six features from Spoonbill had corn in quantity, while all the features had a considerable amount of nutshells, points to a lesser utilization of corn relative to wild plant foods such as nuts.
The common bean does not preserve very well, so its absence is not unexpected on this basis. The bean was certainly utilized in the region after A.D. 1350. Since the common bean was apparently the last of the tropical cu ltigen triad of maize-beans-squash to be successfully introduced and adapted to the environment of the Eastern United States, temporal considerations may also account for its absence. Yarnell (1976:272) indicates that the earliest evidence for the common bean in the east is about A.D. 1050.
The triad of maize-beans-squash is nutritionally sound, with high protein quality, even though the individual constituents contain incomplete or low values in essential amino acids (Wing and Brown 1979). Corn is high in calories and carbohydrates, while beans are generally high in calcium and phosporus, critical nutrients during pregnancy and lactation.
The Caddo are reported to have cultivated two crops of corn each year. The " little corn " was planted in April and harvested in May, while the "flour corn" was planted in the late spring and harvested at the end of July (Swanton 1942:130). The "little corn" was similar to a popcorn, while the "flour corn" may have been a flint corn. It is unclear what type of corn was utilized during the Pecan Grove phase. No complete cobs were recovered, so patterns of row numbers or cob shapes could not be obtained. Information on kernel shape was prohibited by the fact that only four were recovered. Other than the George C. Davis (Jones 1949;Ford 1974) and Hanna sites (Shea 1980), there is a lack of regional data on Caddoan maize. Both Eastern Eight Row, Midwest 12 Row, and North American Pop have been identified, with 8 and 1 0-rowed corn most frequent. Until the types of corn present at a series of Caddoan sites can be positively identified, it will be impossible to evaluate subsistence models that rely on the Caddoan ethnographic record (Keller 1977).

CONCLUSIONS
The exploitation of animal food sources was primarily for protein. Animal protein supplies all of the essential amino acids in about the same proportion as they are needed by the human body, thus promoting the effcient utilization of energy. The variety of protein sources utilized insured an adequate protein supply with little chance of deficiencies.
The acquisition of carbohydrates and fat energy sources comes primarily from the plant foods. Fats, the most concentrated dietary source of energy, come mainly from hickory, pecan, black walnut, and the oily seed plants such as marshelder and sunflower. Nuts and maize occurred in all contexts sampled, but it is probable that the degree of utilization is not comparable even if the amount cannot be quantified. The utilization of maize has been estimated at 30-40% of the diet of Caddoan groups (Story 1981 :148), but it was not specified whether this applied to both the Early and Late Caddoan occupations in East Texas.
It is important to stress the complementary nature of corn and nut utilization (Keller 1977;. Nuts contain large quantities of calories, protein, and fats, while corn is high in carbohydrates and calories. In additipn, the vegetable oil produced by the boiling of hickory nuts is extremely high in (Watt and Merrill 1963). The nut resources, importantly, are relatively generalized in nutritive value. They contribute to all the energy classes being considered. Seeds are low-level energy producers, while maize js clearly selected for its abundance in carbo-!'lydrate production.
Combined with other plant and animal foods that contribute certain vitamins and other nutrients, the judicious combination of these food sources would result in an extremely well balanced diet. To summarize, the diet of the Early Caddoan inhabitants of the Sabine River Basin depended upon a wide variety of animal and plant foods . The procurement of animal foods was principally during the summer and fall, though some hunting activities were conducted year-round. The procurement of nuts was a fall activity, and a substantial contributor to the diet. Certain native plants like sumpweed and goosefoot were utilized as sources of oil and carbohydrates. Maize was present prior to the eighth century A.D. in the Sabine Basin , but was not the major contributor to the diet until some centuries later, and beans and squash were not apprently utilized at this time.

REGIONAL AND TEMPORAL CONSIDERATIONS
There is little systematically collected or published paleobotanical and faunal information in the Caddoan area that can contribute to understanding changes in subsistence strategies. Some regions have better faunal and floral records than others, particularly the dry Ozark bluffshelters utilized by the Arkansas and White River Valley Caddoan groups after ca. A.D. 900 (Sabo 1982;Wyckoff 1980), though the majority of this information has only recently been restudied from 1930s museum-curated food remains (Cleland 1965;Ford 1981 ;Fritz 1981 ).
An examination of gram weight percentage compositions of plant foods from selected published Caddoan sites (Table 5) suggests that there is a decrease in the representation of wild plant resources such as nuts, and a corresponding increase in the frequency of cultigens, maize and beans, in the Late Caddoan period (see Chapman and Shea 1981 : Table 3 for analogous temporal changes in the Little Tennessee River Valley). Early Caddoan subsistence data from the Hanna and Crenshaw sites (see Byrd 1980;Shea 1980; in the Middle Red River Valley are directly comparable to the data from the Sabine River Basin (i.e., a generalized strategy based on wild plants and a wide variety of animal resources, supplemented with cultigens such as maize and squash). In the early Caddoan sites examined, maize accounts for less than 10% by weight of the total floral samples.
From the Roden site, a fourteenth to seventeenth century settlement on Red River, the subsistence data are very different. The analysis of skeletal lesions, caries frequencies, and the wear patterns on teeth led Rose et al. (1981 :125) to conclude that there was ''a large carbohydrate component in the diet, while the scanning electron microscope does not indicate the use of nuts and a low consumption of unprocessed plant fiber. These data indicate that maize formed a significant part of the diet at the Roden site." In the Arkansas Basin and Western Ozark Highlands of Oklahoma, Arkansas, and Southwest Missouri, tropical cultigens such as maize, squash, common bean, and bottle gourd occur more frequently and consistently than wild plant foods after ca. A.D. 900. On this basis, it has been argued that the subsistence strategy of the Caddoan inhabitants of this region was a fully developed agricultural economy supplemented with the exploitation of woodland animal species like deer and turkey (Saba 1982;Wyckoff 1980). Native cultigens such as sumpweed and sunflower were also utilized, along with other possible native cultigens like goosefoot, amaranth, knotweed, giant ragweed, and maygrass. After ca. A.D. 1300, bison exploitation became more frequent while the evidence of sedentary settlements along the western flanks of the Ozarks became sparse, suggesting the adoption of seasonal bison procurement and semisedentary farming (Wyckoff 1981 ).
There is every reason to infer that substantial regional differences in the timing and development of horticultural and agricultural strategies exist within the Caddoan area. This is based solely on the considerable regionally specific ecological and environmental diversity within the Trans-Mississippi South. The adoption of an agricultural strategy will be variable in benefits across environmental zones and through time, such that the change to an agricultural economy was probably not synchronic within or between regions in the Caddoan area. Instead, this change occurred at different times in a time-transgressive process as secure yields of cultigens were reached that surpassed the potential of natural resources in the different regions. Therefore, it is entirely possible that agricultural economies may have existed among Caddoan populations in the Western Ozark Highlands at the same time a more generalized horticultural and wild resources strategy was in use in the Western Gulf Coastal Plain.
Returning to the Sabine River Basin, Late Caddoan and historic Caddoan sites like Steck and Gilbert have lower diversity indices, with deer and turkey present in high frequencies. Age culling seems to have been practiced in the exploitation of deer, with primarily the 1 112 to 3 year-old classes being taken.
This selection far exceeds the proportion in which this age group occurs in a natural population (Emerson 1980: 126) and is more comparable to the pattern of faunal exploitation in the Central Mississippi Valley Mississippian period with the dominance of deer and turkey than to the Lower Valley with its emphasis on fish and a lesser dependence on deer and turkey (Springer 1980:201-202).
The eighteenth century Gilbert site ( Fig. 1) further exemplifies the changing cultural strategy of faunal exploitation. The majority of the deer killed at Gilbert were between 1112 and 3 years of age ( + 100 MNI); deer, moreover, represented over 80% of the MNI and were the major source of food at the site (Lorrain 1967:225, 232).
Obviously, the regional data base is scanty and high priority should be placed not only on the excavation of sites with good ecofactual preservation, but also on the systematic analysis and publishing of Caddoan subsistence data. Whether the suggested subsistence strategy changes are accurate can hardly be evaluated at this point, though this should certainly be a priority. The relative utilization and contribution of maize, as well as models that account for the development of agricultural economics (e.g., Rindos 1980), remain to be assessed and tested within any one region of the Caddoan area.

ACKNOWLEDGMENTS
We would like to thank Dr. Barbara H. Butler for the faunal analysis of the Taddlock material and Ms. Cathy Crane for the floral analyses. Mr. Bob Skiles participated in the excavations at the two sites, and his good spirits and dedication to Wood County archaeology are appreciated. Appreciation also goes to Dr. Laverne Herrington of the Texas Antiquities Committee for her efforts to secure financial assistance for work at Spoonbill and also for her help in the field investigations. Lastly, the comments of the two anonymous reviewers were helpful, but we remain responsible for any errors of interpretation or omission .

13.
Three-dimensional contour maps relative artifact densities were obtained with the aid of a synagraphic computer oiapping program (SYMAP) developed by the Laboratory for Computer Graphics and Spatial Analysis, Harvard University. 14.
Bruseth and , Figure 4- The site is situated at the tip of upland projection overlooking the Sabine River floodplain, but extends into the floodpl · to within ca 30 meters of the river bank ( Figure   1). The Lake Fork Creek channel is app ximately one km east of the site.
While the site was an improved lpasture for many years prior to 1975 and to the present, it had been previously In fact, this cultivation may have contributed to its initial identification in the early 19fs (see below), as well as its subsequent partial burial. The upland sandy soils derive frof' the Queen City Formation, and these are highly susceptible to erosion and colluvial dowrtwasting. Colluvial deposition seems to have been a prominent factor in the burial of culf.ru materials along valley margins and lower footslopes elsewhere in the Upper Basin , and the site's topographic position suggests that both uvial and colluvial deposition is responsible for the burial of the floodplain cultural d ts at the Carlisle site.
The Carlisle site was initially rded in 1930 by A.T. Jackson as a "dense midden deposit; many mussel shells" on the Meredith farm .
When the site was re-recorded in 1975, 4 e midden deposits were not visible on the surface and were exposed only in coring activitif s near the bank of the Sabine River. The midden deposits (here labelled Area B) were covf with ca 20 em of sterile overburden (Skiles et al. 1980). A second area of concentrat cultural deposits was identified on the adjacent upland projection elevated about five m a ve the Sabine River floodplain (Area A).
Friends of Northeast Texas Archrernogy, Special Publication No. 1 (1995) Surface Site Limit Test excavations were carried out lin both areas of the site in 1975 by Skiles. In Area A Skiles excavated six lxl m test units to sample the deposits on the upland landform, and two 50x50 em shovel tes, were also excavated there in 1986 (Figure 2). Although no obvious feature] or concentrations of cultural materials were encountered in the Area A excavations, mr t of the materials recovered (such as pieces of daub, a mud-dauber nest, and several large sherds from refired brushed and incised vessels) suggest that a Caddoan upon the crest of the upland projection.

ARTIFAI ASSEMBlAGES
A wide variety of artifacts was tound at the Carlisle site in the 1975 and 1986 investigations ( Table 1). Plain and deco ted ceramic sherds and lithic debitage were the most common types of artifacts present at the site, followed by unifaciallithic tools, bifaces and biface fragments, and dan projectile ' ints. Most of the materials were collected from the Area A knoll and the general surfa , particularly the lithic tools and debris, while ceramic sherds and daub comprised 85 ent of the artifacts from the Area B midden (see Table 1).  Sweeping, curvilinear brushing is present on another 26 percent of the brushed sherds, and all these are from a distinctive vessel heavily tempered with bone (instead of the grog used with almost all the rest of the sherds from Carlisle). Finally, one carinated bowl was decorated with curvilinear and horizontal brushing marks on the rim.

Ceramics
Plain rims from Area B are predominantly standing and direct types (see Table 6).
Lip-notched and scalloped-rim bowls are also present; these types of lip and rim treatment are notable in Middle Caddoan ceramic assemblages in the Upper Sabine River basin.
Small pieces of burned clay and daub were recovered from both Areas A and B at Carlisle (see Table 1). These are generally rounded and eroded pieces of clay that had been applied to the walls of structures, or were used to line hearths, and became fire-hardened through hearth cooking and/or structure burning. The daub has grass and stick impressions on them.

FAUNAL ANALYSES
A small but extremely diverse faunal assemblage was obtained in the excavations of    Perttula 1993). The assemblages are diverse, indicating that an assortment of upland, riverine, and aquatic species were exploited for food, with deer the most important mammal species, but and fishes . I also were valued supplements to the Caddoan diet.

MUSSEL SHELL ANALYSES
A total of 133 identifiable mussel shells were recovered from the Ardt B midden. About 60 percent of the mussel shell were not identifiable to species, being represented only by pseudocardinal teeth. The most common mussel shell species included Amblema plicata, Quadrula quadrula, and Tritogonia verrucosa, but a number of other species were identified in the assemblage (fable 8).

I 55
In general, the mussel shell species represented at the Carlisle site prefened muddy and slow moving water from medium-sized streams and rivers, although a fJ w species preferred clear water with sandy bottoms. Both stream conditions can be found both the Sabine River and Lake Fork Creek. j  1975 and 1986. Since much of the material derives from surface collections, and functional relationships between material remains are based in large measure upon the regional overview of artifact sequences for Northeast Texas proposed by Story (1990).

Area. A
The earliest occupation at the Carlisle site occurs on the upland projectidn (Area A Over 51 percent of all the dart points from the Carlisle site are Gary vpr. LeFlore (dated ca. 450 B.C. to AD. 250 by Schambach [1982]) and var. Camden (ct A.D. 250-750) projectile points from Area A (see Table 2). _ This suggests that a substantial 56 Early Ceramic period occupation was present on the upland landform. No features were noted in the Area A excavations that relate to the Early occupation, but much df the lithic debris, broken and finished tools, as well as the unifacial tools, probably can be associated with this occupation. Similar types of Early Ceramic period components are common in the Sabine and Sulphur River drainages, namely archeological deposits with large numbers of Gary points and other lithic tools, no ceramics, and no features (see Fields et al. 1992;Perttula et al. 1993). They to represent intensively, but intermittenly, utilized places where tool manufacture and refurbishing activities took place along with the procurement and processing of animal and plant food resources.
A more substantial Caddoan occupation is also present in Area A of the Carlisle site. 1be test excavations there encountered evidence that a structure probably stood on the upland projection: pieces of daub, a mud-dauber's nest, and several large sherds from brushed vessels that appear to have been refired during structure burning. Additionally, 31 sherds from a large incised/brushed jar were found on what appears to have been a living surface (or house floor?) at about 30-40 em below surface (see Figure 11). Ninety percent of the vessel is present, and all of the sherds were recovered at a common depth in U 4S The same types of ceramic decorative styles and vessel forms noted in Area A are present in Area B (see Table 5). By far the most common vessel form present was a cooking jar with an everted rim, and these were decorated with cross-hatched incised lines and punctated marks on the rim and vertically brushed bodies. The punctated marks were commonly applied on an appliqued fillet at the rim/body juncture (see Figure 12). Plain sherds are common in the Area A and B ceramic assemblages (see Table 5). In nearby Three Basins subcluster sites of the Titus phase, like Goldsmith ( 41 WD208) and Steck (41WD529), brushed utility wares are not particularly common (Thurmond 1990;Perttula, Skiles, and Yates in press), and engraved sherds are four to five times more common in sherd assemblages. In Lake Fork Reservoir, brushed utility wares are extremely rare, and occur only in Late Caddoan Titus phase contexts .

7E. The Area
On the south side of the Sabine River, however, at sites such as Bryan Hardy The Bryan Hardy site is undated, but an initial examination of the excavated ceramic assemblage suggests a probable date range between about A.D. 1200-1400 :81).
The A.D. 1410 +/-60 date from the Carlisle site seems consistent witll the frequency of brushed ceramics, the presence of interior thickened rims, and the recovery of Maxey Noded Redware vessels from the site. A thermoluminescence (lLM) date of ca.
A.D. 1280 (Alpha-2398) was obtained from an interior thickened Sanders Plain vesfel at site 41WD117 on Big Sandy Creek, while another 1LM date of ca. A.D. 1400 (Mpha-2397) was secured on a Maxey Noded Redware vessel from another site in that drainage (Perttula et al. 1986:484). Similar interior thickened rims and lip notches have also been noted in the ceramics at the nearby Yarbrough (41VN6) and Limerick (41RA8)  The Carlisle occupation represents a ca. A.D. 1400 small farmstead or houJplace that shares more similarities in ceramic styles with sites on the Sabine River than it does with generally contemporaneous Titus phase occupations upstream in the Lake Fork P"eet drainage. Considerable refinement in cultural assemblage character and chronolbgical 59 sequences are still necessary, however, to understand more adequately the regional significance and social differentiation of the Caddoan use on this part of the Sabine River itself.
There is still a great need for the development of a reliable chronological framework for the Caddoan period occupations in the Upper Sabine River basin (see Story 1990).
Isolating distinctive chronological components in space and time, combined with the identification of discrete single component assemblages, has to be done if archaeological units are to be related to regionally meaningful socio-cultural entities (Johnson 1987), and if we are to move past simple and basic settlement patterning questions.
Every effort should be made to investigate depositional contexts such as those at Carlisle where ecofactual remains might be preserved in cultural association. Certainly sites such as Taddlock, Spoonbill, and Carlisle exist where well-preserved subsistence data can be obtained, but these types of sites have not really been the focus of intensive study in the Upper Sabine River basin. Obviously, the systematic recovery and analysis of faunal and floral remains will contribute immeasurably to the full consideration of Upper Sabine River basin Caddoan lifeways.
Finally, an understanding of the regional paleoenvironmental and geomorpholpgical record is an integral aspect of attempts to conceptualize prehistoric cultural adaptions.
Moreover, these types of investigations may help to locate contexts such as those at Carlisle where buried archaeological deposits are present Currently, the overall paleoenvironmental I record for Northeast Texas is poorly known (Bryant and Holloway 1985;Story 1990), although the potential to recover significant information on Late Holocene environments for the basin is good (e.g., Perttula et al. 1986:322).
In each case, the potential exists with the data base already in hand to carry through exciting and useful research endeavors in Caddoan archaeology in the Upper Sabine River basin. The problem now is to tum that potential into reality by considering broader concepts of cultural change beyond simply basic temporal-spatial analyses. The Carlisle site contains much of the data we need to forge new understandings of Northeast Texas prehistory.

REFERENCES CITED
Bruseth, James E. and Timothy K. Pernula 1981 Prehistoric Settlement Patterns at Lake Fork Reservoir. A.D. 800-1600) of the area. Both structural and burial mounds are known in a ca. 300 kilometer stretch of the Sabine River basin between Lake Tawakoni and Bend Reservoir, particularly being constructed and used during the period between ca. A.D.
1000-1400. The larger mound sites, containing multiple mounds and associated settlements, are apparently regional civic-ceremonial centers. However, whether a hierarchy of contemporaneous civic-ceremonial centers existed at any time during the Caddoan settlement of the Sabine River basin is still a matter of speculation because only one (the Hudnall-Pirtle site) of the known mound centers have been dated by absolute methods.

Introduction
Mound groups constructed and used by Caddoan groups represent a unique but poorly studied cultural resource in Northeast Texas. Important prehistoric Caddoan social, ceremonial, and political centers in the region may be represented by as many as 105 single and multiple mound sites (Perttula 1993a(Perttula , 1993b. Both structural and burial occur as distinct mound types, with burial mounds the mortuaries for the elite members of Caddoan cultural groups, and the structural mounds served as platforms for the Friends of Northeast Texas Archaeology, Special Publication No. 1 (1995) construction (and deliberate destruction) of specialized structures or dwellings (Story 1990:340-341). The study of the civic and ceremonial nature of the Caddoan mound centers is important for understanding the development of Caddoan culture because they can provide us with data on how the emergence and elaboration of socio-political complexity is related to cultural change over time in the region.
Although Caddoan mound sites have been reported in the Sabine River Valley since the early 1900s (Pearce 1920), the only professional investigations of such sites has been primarily restricted to relatively unsophisticated trenching of the mound deposits carried out by A.T. Jackson in the 1930s for the University of Texas (see Guy 1990). Since that date, new mound sites have been reported (see Malone 1972;Webb et al. 1969;Jensen 1968aJensen , 1968bPerttula and Skiles 1987;Bruseth 1991), but with the
The present research on Caddoan mound groups in the Sabine River Valley and tributaries is thus a first step towards developing a sound data base on the location, character, contextual integrity, and current preservation condition of known and potential mound sites, especially the important multiple mound groups. This data base serves as an integral part of the Northeast Texas Preservation Plan for archeological resources (e.g., Kenmotsu and Perttula 1993), will help to document and support National Register nominations for mound complexes in the Sabine River Valley, and will be important for focusing study on the broader questions concerning the complex socio-political developments of prehistoric Caddoan societies in the Sabine River Valley{cf. Story 1990;Perttula 1989aPerttula , 1993a.

The records and collections at the Texas Archeological Research Laboratory and the
University of North Texas were reviewed, and interviews were conducted with local avocational archaeologists and collectors, to develop an initial inventory of known and potential Caddoan mound groups in the Sabine River Valley. Based on leads gathered in those repositories, such as landowner names of farms provided in manuscripts and land survey field notes, a cursory examination of county records and archives (such as the land deed records of the 1920s-1930s), as well as local histories (e.g., Wolden 1932), the inventory process was considered to be relatively comprehensive for Northeast Texas.
Information on selected Northwest Louisiana mound sites in the Sabine River Valley was supplied by the Louisiana Department of Culture, Recreation and Tourism, Office of Cultural Development, Division of Archaeology, and from published sources (e.g., Girard 1991).
This work was supplemented by relocating known and recorded mound sites in the Sabine River Valley. This was done to obtain more detailed locational information on the mounds, as well as gather data on associated material culture assemblages. The multiple mound sites, or potential multiple mound sites, known in the Sabine River Valley were given precedence in guiding the survey effon because of their regional archaeological significance (Pentula 1989a:1, 4).
Limited subsurface testing was conducted at four mound sites (41PN8, 41SM54, 41SY46, and 41UR30) to obtain, if feasible, temporally and functionally diagnostic prehistoric artifacts from selected sites and/or specific areas within sites, and also to obtain suitable materials such as charcoal or thermoluminescence samples for dating (see Pentula 1989a:35-40). Where possible, surface collections were made at individual sites, and notes were maintained for each site specifying the location and extent of surface-exposed anifactual materials.

Discussion of Results
A total of 37 possible mound sites have been identified in the Sabine River Basin of Northeast Texas and Northwest Louisiana, three in Louisiana and the remainder in Texas (see also Pentula 1989a:43-91; Girard 1991). They are distributed over a ca. 300 kilometer stretch of the Sabine River Valley and its tributaries, from Lake Tawakoni in the west to below Toledo Bend Reservoir in the south (Figure 1). The potential and known mound sites represent a significant span of the regional prehistoric archaeological record, from ca.     ( Figure 2). Significant pothunting has occurred at 41 UR30 in the last five years, and a large Early Caddoan cemetery has been destroyed through this work.
One of the mounds at the Hudnall-Pirtle site (41RK4), then known as the Bivins-Flanagan mound, was investigated about 1960 by Buddy C. Jones (see Davis et al. 1971 The Archaeological Conservancy (Bruseth 1991).
In the 1960s the University of Texas and Southern Methodist University conducted testing and excavation projects at several mound sites at Toledo Bend Reservoir, most notably at the Lafitte (41SY15) [Scurlock 1964] and the Coral Snake Mound (16SA48) sites. The latter is an Early Ceramic or Woodland Period burial mound (McClurkan et al. 1966(McClurkan et al. , 1980Jensen 1968a;Story 1990). Methodist University also conducted minimal excavations at the Jamestown (41SM54) and Cox (41WD349) mound sites in the late 1970s, but the notes, artifacts, and photographs from that work have been lost or misplaced.
In the early 1980s, the Brittain (41SY42) and Beauchamp Creek (41SY46) mound sites were reported in or adjacent to the U.S. Forest Service's Sabine National Forest in Shelby County, Texas (see Figure 2). A small test hole was excavated by John Ippolito  (1988), Forest Archeologist, in one of the reported mounds at the Beauchamp Creek site, but nothing of significance was apparently recovered from this work other than to demonstrate that it was of artificial construction.

SABIN£ RIVER BASIN EOG£
Almost all the Sabine River Basin mound sites have been potted or vandalized at one time or another from the late nineteenth through the twentieth century. However, few have been as extensively disturbed by looters as have Caddoan cemetery sites throughout the region (Perttula 1989b), with the notable exception of the above-mentioned looting at the Boxed Springs (41UR30) mound site where a cemetery containing more than 125 individuals was found adjacent to one of the mounds.
Consequently, many of the mounds themselves still possess some degree of overall contextual integrity, and in cases where off-mound habitation areas exist, they have not yet been seriously damaged by looting activities. Apparently, with the exception of the Boxed Springs cemetery, cemetery areas in off-mound habitation settings are relatively uncommon on Sabine River'Valley mound sites.

Regional Spatial Patterns
The Sabine River Basin is divided into Upper, Middle, and Lower basins utilizing physiographic, geomorphological, and geological criteria (e.g., Kier et al. 1977). The Upper Basin, part of the East Texas Embayment, includes the area from the headwaters of the Sabine River to the western edge of the Sabine Uplift, while the Middle Basin is entirely within the area effected by the Sabine Uplift. The Lower Basin is that part of the Sabine River Valley below the Sabine Uplift and extending to the Gulf of Mexico (see Gibson 1978) [ Figure 3].    Table 1). Both the Jamestown (41SM54) and Boxed Springs {41UR30) sites contain evidence for extensive settlements associated with the mounds, and thus were clearly not vacant Caddoan community centers (see Story 1990:341).
The key multiple mound sites in the Upper Sabine Basin are the Jamestown (41SM54) and Boxed Springs (41UR30) sites. This is suggested based on the number, size, internal arrangement and spacing, and presumed character of the mounds at each of the sites, as well as the extent of the associated settlements ( + 15 acres) [Perttula 1989a:67-70, 78-80). The regular spacing of mound centers along the major streams and tributaries in the Upper Sabine Basin also hints at the existence of locally Creek, and Flat Fork Creek (see Table 2). All mound sites in the Middle Sabine  Table 1). On the basis of the ceramic assemblage and calibrated radiocarbon dates of AD. 1158 +/-70 (Beta-43539) and A.D. 1174 +/-70 (Beta-43540), the Hudnall-Pirtle mound site was apparently occupied during the Early Caddoan Period (Bruseth 1991), but all the others were probably constructed and utilized during some portion of the Late Caddoan Period (ca. AD. 1400-1600) {see Table 1].
Few other mound sites in the Middle Sabine Basin can be positively identified as having an Early or Middle Caddoan Period occupation, and thus the Hudnall-Pirtle mound ' center is clearly an isolated premier or key mound group in this part of the basin. Indeed, the si.ze of the site and its associated settlement ( + 60 acres), the number of flat-topped platform mounds and conical mounds (three and five, respectively), and the likely presence of a plaza area (Bruseth 1991), all indicate that the Hudnall-Pirtle site is the most prominent Caddoan mound center in the Sabine River Basin of Northeast Texas and Northwest Louisiana. Story (1990:325)  Sabine River floodplain, more than 60 kilometers south of the next nearest mound, the Anthony site (16SA7 or X16SA40). Neither site has been professionally investigated, and the available information about them is rather limited (Perttula 1989a:47, 56-57;Story 1990:279).

Social and Temporal Considerations
It is presumed that the use of mounds by Cad.doan peoples represents deliberately patterned cultural behavior expressing social, religious, and symbolic principles shared by related groups and communities (see Rogers 1989;Kay et al. 1989). For instance, Sabo and Early (1988:98) suggest that social or ceremonial activity at these individual centers [Caddoan mound sites in the Arkansas River and tributary valleys] promoted the solidarity of the local communities responsible for the construction and maintenance of these centers, in addition to providing contexts for the expression of important aspects of social structure such as systems of ranking (e.g., Brown 1971;Rogers 1982Rogers , 1983. We may suggest also that the Network of mound centers ... promoted socially integrative activity on a larger level than the local corporate group; that is, solidifying geographically separated, small corporate groups into a single, regional community [brackets added; the Sabine River Basin, the Coral Snake burial mound site (16SA48) [see Jensen 1968aJensen , 1968bStory 1990:282-289 for further details].
The identifications of premier or key mound centers is based on a number of variables, including: mound sizes, internal mound arrangement and spacing, the existence of plazas, and inferred mound functions (i.e., as mantles over burials, as bases for specialized buildings such as charnel houses or temples, or to mantle sub-surface features).
The large and internally complex mound centers at the Jamestown, Boxed Springs, and Hudnall-Pirtle sites are considered the premier centers in the Sabine River Valley. They are Early to Middle Caddoan Period mound centers with four to eight mounds each (Figure 8).
Possible contemporaneous Early to Middle Caddoan Period mound centers that are considered subsidiary or secondary nodes in such a hypothetical hierarchical system are the Cox and Colony Church (41RA31) sites, and possibly the Lee Joyner or Seaton Bros.
( 41RA38) sites (see Figure 7). Subsidiary mound centers are identified as those sites containing multiple mounds, but lacking the complex internal arrangements and spacing of mounds and plaza noted for the premier or key mound centers, nor do they have the variety. of mound types identified or postulated for the Jamestown, Boxed Springs, or Hudnall-Pirtle sites (see Table 1).
Late Caddoan Period mound sites are common in the Middle Sabine River Basin, particularly in southwestern Harrison County and Shelby County, Texas (see Figure 8).  Thurmond 1985Thurmond , 1990Perttula 1992, 1993a), but the cultural-taxonomic systematics in this part of Northeast Texas require considerable refinement and redefinition (Story 1990:167-168 and Perttula et al. (1986:57) suggest on the basis of limited ceramic evidence that the AN. Vickery site (41WD11) may have been occupied during this period. The lack of mounds in this part of the basin does not mean that the area was unoccupied ca. A.D.
1400-1600, because the regional density of Late Caddoan habitation sites is much more substantial in parts of the Lake Fork and Big Sandy Creek drainages within the basin than was the Early and Middle Caddoan Period settlement (e.g., Perttula and Gilmore 1988;Perttula et al. , 1993Thurmond 1985Thurmond , 1990.  Trends in mound construction and use in the Sabine River Basin are similar to those outlined by Thurmond (1990:234-235) Webb 1959).
In the period between ca. A.D. 800-1200 in the Cypress Creek Basin, mound groups were classified by Thurmond (1990:234) into large, multi-mound centers and small single-mound components. Thurmond (1990:234) states that: "It seems quite possible that a hierarchical system of centers is represented, with each succeeding level serving a broader area, and the whole integrated into a regional network of interaction and redistribution." Such a hierarchical system of mound centers is also postulated in the Sabine River Basin between ca. A.D. 800-1200 (or as late as A.D. 1400), with the Jamestown, Boxed Springs, and Hudnall-Pirtle sites representing the apex or top level of the hierarchical system. The regional network may have included groups in both the Upper and Middle Sabine River Basins, based on the spacing considerations discussed above (see Figure 9), but was also integrated as well into a broader system of interaction and redistribution that probably extended throughout larger portions of Northeast Texas and Northwest Louisiana. It is interesting to note, therefore, that the spacing between the major Early Caddoan Period mound centers of Crenshaw, Gahagan, and Mounds Plantation on the Red River in Northwestern Louisiana and Southwestern Arkansas is about 80 km (Webb and McKinney 1975:122), and that the Hudnall-Pirtle site, the premier Early Caddoan Period mound center in the Sabine River Basin, is about the same distance from both Mounds Plantation and Gahagan.

91
The term local network employed herein is analogous to the affiliated group defmed by Story and Creel (1982:32 and Figure 8) for the Neches River Late Caddoan Period archaeological record. According to Story and Creel (1982), the afflliated group represents the archaeological concept of a number of constituent groups who shared a similar socio-political organization, had similar intergroup interaction and settlement patterns, and were integrated in a hierarchical structure by the temple-residence complex center. This center contained an earthen mound where the paramount leader of the affiliated group resided. The constituent group would consist of the lesser centers (without mounds), domiciles, cemeteries, farmsteads, hamlets, and villages that are socio-politically unified into the afflliated group. With the exception that the subsidiary centers in the Sabine River Basin during the Early, Middle, and Late Caddoan Periods have at least one earthen mound, they may be considered to be representative of a level of socio-political integration generally compatible with Story and Creel's (1982) model of a constituent group lesser center.
The patterned arrangement of mounds within a site, consistent means or methods of mound construction, the recurrent use of a restricted space for mound construction, structure dismantling, as well as other ritualized activities in Caddoan mound centers, highlight the contextual importance of mound use in Caddoan culture (e.g., Sabo and Early 1988:99;Story 1990:339-342;Perttula 1992). One particularly intriguing aspect of the use and patterned arrangement of mounds in the Southeastern United States is the correlation of mounds with celestial orientations, and with a consistent unit of spacing referred to as the Toltec module (Sherrod and Rolingson 1987).
Some of the major Caddoan mound sites studied by Sherrod and Rolingson (1987: Figure 12). There is also a stellar alignment with Vega, "one of the brightest stars seen from the northern hemisphere" (Sherrod and Rolingson 1987:29). The Toltec module standard of measure, 47.5 meters, correlates well with the spacing of only three of the other six mounds at the Jamestown site (see Figure 12).
Other evidence of a complex hierarchical structure beyond the number, size, plan, and complexity of mound centers is obtained by examining mortuary behavior in mound contexts (see Thurmond 1990:235). The recovery of burials with elaborate funerary offerings and exotic goods as grave associations with paramount individuals, typically adult males, in mound contexts has been argued by Caddoan archaeologists (see Brown 1971;Rogers 1982;Sabo and Early 1988) to represent high-status burials and the existence of a ranked class structure. With the exception of the centrally located sub-mound burial pit from one mound at the Boxed Springs site (see Perttula 1989a:78-80; Story 1990), which because of its position and burial accompaniments has been interpreted to be a high status burial, it is not possible at present to conclusively demonstrate either that the premier centers in the Sabine River Basin all contain evidence of high status burials, or that the sociopolitical interpretations embedded in the hierarchical classification of mounds are realistic. Attempts to demonstrate that the temporal, spatial, and functional differences in mound centers within the Sabine River Basin are the result of sociopolitical distinctions will require a more comprehensive investigation of these mound sites before these proposed archaeological units can be fully related to regionally and locally meaningful cultural variability (e.g., Johnson 1987  It is important to reiterate that the majority of the mound sites known and/or investigated in the Sabine River valley appear to be generally intact Although it is the case that these Caddoan mound sites have been cleared and plowed over the last 100 years or so, and have been the scene of occasional vandalism and looting activities, they have not as a group been subjected to the intensive and systematic disturbances caused by looting on Caddoan cemetery sites that contain certain artifacts which gamer a high price on the antiquities market in Texas, Arkansas, and other states (e.g., Early 1989;Perttula 1989b).
The reasons behind why Caddoan mound sites are not extensively disturbed have not been thoroughly studied, but one primary reason may be the pothunter's perception that these mound sites generally do not contain the types of easily worked archaeological deposits common at aboriginal cemeteries, nor do they necessarily contain the number and variety of contexts from which high-priced lootable goods can be quickly obtained.
Since Caddoan mound sites as a group appear to retain a measure of integrity not typically held by other types of Caddoan sites in Northeast Texas and Northwest Louisiana, it is critical that measures be implemented by State and Federal agencies, the professional archaeological community, avocational archaeologists, and intere sted landowners, to insure the preservation and protection of as many of these important sites as possible. A site protection and preservation plan for Caddoan mounds in the Sabine River Valley, and indeed for all of Northeast Texas and Northwest Louisiana, should develop goals and policies that will help determine which are the mound sites most vulnerable to destruction, and which are the sites of most importance for immediate, short-term, and long-term protection.
The Jamestown (41SM54) and Boxed Springs (41UR30) sites, two of the three premier Caddoan mound centers in the Sabine River Valley of Northeast Texas, need top priority actions to insure that they will be protected and preserved, if at all feasible. The other premier mound center, Hudnall Pirtle (41RK4), is owned by The Archaeological Conservancy.
Hand in hand with the development of a protection and preservation plan is the development of an active research program at these Caddoan mound sites (cf. Story 1991;Perttula 1993a). The research program should consist of excavations at a sample of the different types of Caddoan mound sites in the Sabine River Valley as a means to assess their internal character and integrity, and also to examine the temporal, functional, and cultural context of the mounds, middens, borrow pits, and other types of features preserved in the archaeological record.
It is also important that these research efforts include the development of contemporary maps depicting the current condition of the properties, and these maps should be bolstered with the analysis of previous and recent aerial photographs of the sites.
The research efforts need to be of sufficient scale that radiocarbon or thermoluminescence dates from interpretable mound or nonmound, fill zone, or sealed contexts be obtained from as many Caddoan mound sites as possible.

Known collections of cultural materials from Caddoan mound sites in the Sabine
River Valley need to be thoroughly reanalyzed, photographed, and described so that this updated information becomes a useful aspect of the data base for the study of Caddoan archaeology in Northeast Texas and Northwest Louisiana. Along with the research efforts proposed for the Caddoan mound sites themselves, it is important to initiate intensive survey investigations in the areas immediately proximal to the mound sites to identify the contemporaneous Caddoan villages, hamlets, and farmsteads of those groups who constructed and used the mound sites for civic and ceremonial purposes. The intensity and scope of such a survey should be addressed as an aspect of the protection and preservation plan which may need to be developed for specific Caddoan mound sites in the region.

Acknowledgments
The work reported on herein was funded by the Texas Historical Commission

Patterns of Prehistoric Lithic Raw Material Utilization in the Caddoan Area: The Western Gulf Coastal Plain
TIMOTHY K. PERTIULA

Introduction
The prehistoric inhabitants of the Caddoan area utilized a wide variety of raw materials in the manufacture of lithic tools. Raw materials from as far away as the Texas Panhandle (Aiibates chert and Tecovas jasper), southern Kansas (Kay County chert and Flint Hills chert), Nebraska (Nehewaka chert), and Tennessee (Dover chert) have been recovered in excavated lithic assemblages. Only in the last few years, however, have archaeologists begun to study the variability in Caddoan raw material types as a key to understanding prehistoric behavior (e.g., Banks and Winter 1975;Mallouf 1976;Vehik and Galm 1979;Wright 1980;Wyckoff 1963Wyckoff , 1965Wyckoff , 1967Wyckoff , 1968. This paper examines the lithic raw material data from a series of sites in the Lake Fork Reservoir in the Western Gulf Coastal Plain of Texas ( Bruseth andPerttula 1979, 1981). The differential utilization of local versus non local raw materials is evaluated, and the patterns of raw material use in Archaic and Caddoan lithic assemblages are examined to elucidate questions of regional interaction and tool use.

Background
Those areas characterized by a lack of naturally occurring lithic raw materials (Mississippi Valley lowlands) or by their presence in redeposited gravel beds (the Western Gulf Coastal Plain) are particularly suitable for the study of lithic raw material util ization. Local raw materials occur in gravel deposits of the Uvalde Gravels.
These gravels were derived from extensive erosion of the High Plains in the Pleistocene and are composed principally of small quartzites, chert, and petrified wood cobbles.
Nonlocal raw materials are easily isolated through visual inspection, and locations of local raw material outcrops or ·gravel deposits, if present, can be predicted if the area geomorphology is known.
There are four particular uplift areas adjacent to the Western Gulf Coastal Plain that provide a number of lithic raw materials pertinent to this study. They are the Ozark, Ouachita, Arbuckle, and Edwards areas (Figure 7 material utilization was relatively homogeneous temporally and spatially (e.g., Vehik 1979:490,494).

A Model Of Nonlocal Lithic Raw Material Utilization
At present, prehistoric lithic raw material utilization in the Caddoan area, as for most other areas of the Southeast, is imperfectly understood. More detailed raw material identification procedures need to be implemented (cf. Luedtke 1979) across a wide area of the Gulf Coastal Plain and adjacent uplifts to confirm the visual identifications presented in most recent Caddoan studies (Bobalik 1977;Calm 1978;Mallouf 1976). Refinements in data are necessary, but perhaps of equal importance are developments in model building.
The model that follows is concerned with the relationship between changes in lithic raw material utilization and changes in settlementsubsistence systems. Following precepts initially suggested by Grady ( 1978) for the Archaic period in Texas, the model is further expanded to deal with the Early Ceramic and Caddoan occupations of the Western Gulf Coastal Plain (see Bruseth and Perttula 1981:111 and Table 6-8).
Grady (1978) has argued that differences in lithic raw material utilization . in the Archaic period in east central Texas relate directly to the establishment of more regionally specialized exchange systems through time. He postulates that during the Late Archaic period (ca. For this discussion, an exchange network refers to the specific exch_ ange of goods, information, etc. , between linked groups or series of groups. This pattern is seen by Grady to be the result of the evolutionary development of a Primary Forest Efficiency economic strategy (Caldwell 1958) in the area during the Archaic Period. In the Early and Middle Archaic, the generalized economy of hunting and gathering involved a wide territory of effective group exploitation. Interaction with adjacent groups therefore involved the incorporation of larger areas into an overall exchange network composed of a number of cultural groups.
The effective network was synonymous with the maximum range of preferred exploitation, which was a geographically extensive area. The generalized nature of these Early and Middle Archaic economies, combined with a mobile settlement system and a flexible exchange system, allowed a simple linear exchange pattern to develop with members of adjacent interaction spheres. Some of these areas would have possessed fine-grained silicates--superior to local _ cherts--that the exchange network would have made available. Assuming that lithic raw materials were critical resources and high quality cherts were preferred for tool manufacture because of their ease of working, then Early /Middle Archaic tool assemblages would be expected to include significant percentages of non local raw materials.
During the Late Archaic in the Western Gulf Coastal Plain, native economies were expanding and diversifying (e.g,, Christenson 1980). Exotic cucurbits and wild plant foods may well have been incorporated into the diet, since they were in adjacent areas at this time. The intensive foraging characteristics of this adaptation (Winters 1974) combined with apparent population growth resulted in definable terri-tories within the major streams in the area. Thus, a relatively stable and compact pattern of less mobile populations may be characteristic.
The reduction of exploited territory had ramifications in the exchange system. Whereas Early /Middle Archaic period exchange was suggested to have taken place · across larger areas and different exchange systems, Late Archaic exchange became more regionally oriented, less flexible, and generally confined within a common economic sphere composed of a series of territorial groups with complementary adaptive strategies ( Grady 1978). Thus, exchange of non local lithic raw materials and tool forms would be replaced by materials that reflect the raw materials available within a much-reduced area. Less access to nonlocal materials is compensated for by a more well-developed, stabilized pattern of exchange that ensures constant access to local critical resources.
Exchange patterns during this period should emphasize local raw materials, and it is expected that tool assemblages will consist of higher percentages of local materials and tool forms than the preceding periods.
In the Midwest, a h ighly developed pattern of exchange of trad e goods is characteristic of the Late Archaic period (Winters 1968). Marine shell, non local cherts , and elements of the lapidary industry occur in burial associations and indicate occasional wide-ranging interregional exchange. Whether such a pattern of exchange also ex ists in the Late Archaic in the Western Gulf Coastal Plain is not known (Story 1981) .
Early Ceramic (A.D. 1-800) raw mate r ia l utilization patte rns are suggested to be comparable to Late Archaic components primarily as a result of the same restricted territorial exploitation/exchange relationships and territorial autonomy characteristic of both periods. No centralized loci for the handling, controlling, and dispersing of non loca l materials are known, and it is likely that the generally low degree of hierarchical differentiation reflected in the archaeological record will correlate with lower frequencies of exchanged valuables (e.g., Hodder 1980}. Of course, where there is a greater burial and/or settlement evidence of hierarchy, higher frequencies of nonlocal goods will be expected . Such would seem to be the case along the lower Sabine a nd Angelina rivers (Jelks 1965}.
The innovation of mound construction in the Early Caddoan period had an important effect on the character of exchange in the Western Gulf Coastal Plain . Evidence of elaborate ceremonial activities in east . Texas around A. D. 800 (Story 1972) indicates the development of a highly sophisticated and well-organized exchange network. This network probably encompassed the majority of the Caddoan area. The increased organizational complexity and hierarchical differentiation within this period, and the well-developed sociopolitical integration so characteristic of the Early Caddoan period, are reflected in the archaeological record by elaborations in social and ceremonial activities in which the procurement of nonlocal raw materials (at least in the major mound centers} plays a large role. While it is not clear how raw materials were exchanged on an intraregional basis, it is likely that the mound centers had a vital function in their procurement and ultimate distribution to lower-level components (villages and hamlets} within the overall system. On an interregional level, exchange was probably between distinct territorial units (phases?) within the larger regional area of cultural groups with similar adaptive strategies (Brain 1976; , as well as with different cultural groups (e.g., The Mississippi Valley).
With the cessation of most mound construction in east Texas in the Late Caddoan Period (ca. A.D. 1300--1400) it is presumed that exchange once again became more localized (Story 1981) • Without the mound centers' functioning as foci of resource procurement and redistribution, particularly from distant (i.e., Ozark Highlands) resource locations, a more localized and intraregionally based, but not necessarily less complex, exchange network developed.
The shift toward a smaller network had its basis in the relatively specialized nature of Late Caddoan agricultural economies.
These economies stressed cohesive social networks with strong, intraregional community ties geared to the development of local economic security.
lntraregional exchange was both necessary and sufficient, while interregional exchange controlled by local and/or regional elites remains characteristic only in the major alluvial valleys. Hence, it is to be expected that in those areas whe re mound construction ceased, such as the area under consideration here , local assemblages should exhibit higher frequencies of local raw materials than in the preceding Ea rly Caddoan phase.
Within the ma jor river systems of the Caddoan a rea , such as the Red , Ouachita , and Little Rivers , t he exchange network p r esumably rema ined unc hanged up to abou t A.D . 1700, judging f rom the continuation of earthwork construct ion a nd e la borate ceremonia l activities (Wyckoff 1974  Th is distribution indicates a much wider exchange networ k than e xisted during all previous periods except for the Early and Middle Archa ic. Without detailed knowledge of how horses affected exchange patterns or the supposed role of the Wichita and Plains Apache as middlemen, distributors, and procurers of these non local raw materials, (see Hofman 1978), the exact nature of the exchange process cannot now be specified. Nevertheless, it appears that the exchange network during the historic period may be of a different order of magnitude than that of preceding periods. This is reflected in the wider movement of resources and in the differences in the resources being exploited.
There were two main patterns in the historic exploitation and e xchange of lithic materials in the southern Plains and adjacent woodlands.
The first is the direct exploitation of a lithic raw material source by a small segment of a larger cultural group, with exchanges of lithic materials taking place as groups met at particular prescribed times of the year (see Hartley and Miller 1977:258;Reher and Frison 1980: 130;Stanford 1978) .
The second pattern consists of indirect exploitation and localized exchange within particular expanded networks and zones of distribution.
The first pattern seems to be characteristic of the more mobile bison-hunting inhabitants of the mixed-grass and short-grass high plains of Texas and Oklahoma, and the second pattern of the sedentary bison and/or deer hunting horticulturists of the prairie-woodland interface ( Norteno focus). The former pattern would be manifest in the archaeological record as site assemblages where distant nonlocal lithic materials occur predominantly or exclusively (see Ferring and Vernon 1978 : 346;Hughes and Willey 1978:253, 281-282). The latter pattern would be much like that of earlier ceramic phases with high percentages of local raw materials and corresponding lower frequencies of nonlocal materials when compared to the more mobile populations.
In summary, a cyclical pattern of nonlocal raw material utilization is envisioned, with quantitative changes through time. Higher percentages of non local materials are expected, then, in Archaic II (late Middle Archaic), Early Caddoan, and Historic components, and correspondingly tower in Late Archaic, Early Ceramic, and Late Caddoan components in the Lake Fork area specifically. The applicability of such a model to the Western Gulf Coastal Plain as a whole will be considered below.
These changes are argued to be the result of corollary developments in subsistence strategies, elaborations in social and ceremonial activities and related exchange activity, and territorial constraints.
Utilizing only sites from a limited area, the present analysis essentially provides sufficient control over the range of raw materials (both local and · non local sources) available for utilization at any one temporal period. Granted that local raw materials were abundant and apparently easily procured from within a 10-15 km radius from any site in the Lake Fork Reservoir, general changes in raw material utilization between and within particular periods are most likely the result of prescribed patterns of resource utilization and exploitation .

The Lake Fork Area
The archaeological assemblages considered herein consist of a series of spatially and temporally discrete components of Archaic through Late Caddoan period occupations recovered from 1976 and 1978 excavations at Lake Fork Reservoir in northeast Texas (Figure 7.1), on Lake Fork Creek, a major tributary to the Sabine River. The project was conducted by the Archaeology Research Program, Southern Methodist University, and was funded by the Sabine River Authority and the Texas Antiquities Committee under a state of Texas Antiquities Permit. A complete description of the project area, the sites, and the field and laboratory methods has been provided elsewhere ( Bruseth and Perttula 1979, 1980.
The analysis of lithic raw materials from Lake Fork archaeological assemblages began with a preliminary sorting of lithic tools and debris. A number of working raw material categories were established. These categories were checked by consultations with Larry D. Banks, U.S . Army Corps of Engineers Archaeologist in Dallas, Texas.
Banks pointed out raw material identification errors, provided descriptive and geological information, and indicated source locations.
Following this, relevant literature concerning lithic raw materials was examined to specify, where possible, the area of natural occurrence (e.g., certain river gravels, formation outcrops, or quarries), within particular regional lithic raw material areas, of each of the defined stone types.
. Twenty raw material types were identified in the Lake Fork archaeological assemblages. On the basis of macroscopic observation, each type was separated by color, texture and grain size, translucency, and inclusions. It was possible to specify natural source distributions for most raw material types and to suggest probable natural source distributions for .the remainder. However, the degree of confidence in assigning source areas varies somewhat with raw material categories.
Thirteen types are categorized as nonlocal, given the location of the Lake Fork area relative to natural source distributions. The remaining seven raw material types are locally available in Yalde gravels.
Two raw material ·types each have been identified in Lake Fork assemblages from the Ozark and Arbuckle regions, one type from the Edwards Plateau region, and eight types from the Ouachita Mountain region (Table 7.1). Except for the chert types in the Edwards Plateau and Ozark regions, the remainder of the non local raw materials could be obtained in Red River gravels approximately 100 km north of the Lake Fork Creek area.
The Boggy, Washita I K iamichi, Little I and Red Rivers carry Ouachita and Arbuckle lithic raw materials as gravels. Edwards Plateau chert outcrops are approximately 200-250 km or more southwest (Henry et al. 1980), while Ozark lithic raw materials are carried in Arkansas River (and related stream) gravels 250-300 km north of Lake Fork Creek Reservoir. Raw material descriptions of the nonlocal types are provided in Table 7.1.  1978:47). Distinguished from quartzites by its smoother, more lustrous, finer grained texture.
Uvalde gravels. Because of the widespread occurrence of the gravels across the southern High Plains and adjacent physiographic provinces, quartzite occurring in Lake Fork assemblages is attributed to local sources.    1980).

Temporal Trends
Based on the analysis of over 25,000 pieces of lithic debris and 1 ,000 tools from 30 prehistoric components, clear trends in the differential utilization of nonlocal raw materials are evident. Though the representation of nonlocal lithic raw materials differs from tools to lithic debris in absolute frequency for all components , the same gene ral trends are shown (Tables 7. 2 and 7. 3).
. All r Archaic components have low percentages of non local lithic debris (X=t 0%) : mucfi lower 'than for tools at each component. There is an-absolute decrease in the utilization of nonlocal raw materials from the Late Archaic :· period (ca. 2000 B.C.-1 A.D.) through the Early Ceramic ·period ( 1-800 ··.A.D.), when the trend reaches its nadir.
The differential representation of nonlocal raw materials from tools to lithic debris is also present ' in Early Ceramic components but is not as pronounced as in Late Archaic (Archaic Ill) components. An absolute increase in nonlocal raw materials is clearly correlated with Early Caddoan (ca. A.D . 800-1350) components, generally two or three times higher in magnitude. In fact, approximately half of the raw material utilized in tool production is nonlocal, and up to 80% of the small projectile points were manufactured from nonlocal materials. From other lines of evidence, it is suggested that nonlocal raw material utilization peaks ca. A.D. 1250, or •the later part of the Early Caddoan period as defined here (e.g., . The utilization of nonlocal .lithic raw materials decreases again in the Late Caddoan period (A. D. 1350-1600). The analysis of the lithic raw materials from the Gilbert site,· located·. in the upper portion of Lake Fork Reservoir, indicates an increase In the utilization of nonlocal materials from the  The utilization of raw materials from the differe nt lithic regions s hows a number of apparent changes through time (Table 7.4). First, Edwards and Ouachita materials dominate every temporal period. Ozark region cherts have the lowest representation, which is not unexpected given the long distance ( 300 km) . from Lake Fork. Nevertheless, Ozark materials are a consistent component in all assemblages with mediocre to good sample sizes. Cherts from the Arbuckle region, which would be available in the Red River gravels downstream from the Washita a nd Blue Rivers (ca. 100 km north of Lake Fork), consistently rank third in utilization behind Edwards and Ouachita types. While sample sizes obviously have an · influence on fluctuations, the presence of Ozark region materials, . especially Keokuk chert, does seem to be more frequent in the later portions (after A.D. 1200) of the Early Caddoan period , and the Late Caddoan period, while Arbuckle material in general seems to increase at the beginning of the Early Caddoan period ( Bruseth and Perttula 1981: Tables 6-1 0).
It would be worthwhile to consider what these fluctuations represent in terms of interregional exchange and cylical patterns of raw material availability (e.g., Wright and Zeder 1977:238). Such patterns of different regional emphasis should become more apparent when better dated, and larger and comparable samples can be obtained. Ogallala chert, quartzite, and petrified woods are the three commonly utilized local raw materials throughout all temporal periods at Lake Fork (Table 7. 5). Jasper and chalcedony are only infrequently represented, primarily during the Early Caddoan period. It is likely that the , lpw _representation of jasper and chalcedony is due, in part, n_ot_ to tl)eir absolute-in Uvalde Gravels, but to the fact that n;tos1;_ pebbles· are too small for effective prehistoric utilization. The notable increase in the frequency of jasper utilization during the Early Caddoan _period may reflect procurement from another nonlocal source. Jasper, both red and yellow, is common in Red River gravels (Mallouf 1976:51) and could have been included in transactions involving other nonlocal materials traded from the Red : River Valley to the interior areas East Texas. Except for the apparent higher utilization of petrified wood during the Archaic II period, the utilization of the three main local raw material types i!' " consistent · through time . However, differences in frequency between components demonstrated to be contemporan"'ous parts of a larger site cluster (i.e., WD 39A/WD39B and WD 83A/WD83B) indicate that dhf erences in utipzation of local materials were also, in part, the result o functional parameters and -different disposal modes rather than solely temporal factors.

Functional and Morphological Considerations in the Differential Utilization of Raw Materials
Different tool classes appear to be correlated with particular raw material varieties. The majority of cores and large blfacial tools  (bifaces, axes, etc.) are manufactured from local coarse-grained cherts, quartzite, and petrified wood, while the tool classes requiring sharp edges were mainly manufactured from fine-grained cherts. The differences in the utilization of raw materials seem to relate to the kind of durability of the cutting edge desired. An edge· produced on coarsegrained material is more durable than an edge produced on fine-grained material.
The utilization of nonlocal raw material is such that the majority of tools requiring a sharp cutting edge (retouched pieces, flake tools, projectile points, etc.} are expected to be of this material. Furthermore, the relative scarcity of non local material should result in the maximization of Its use through resharpening and/or secondary tool utilization. Tools requiring a more durable edge would be expected to be made of local materials such as quartzite and petrified wood. To investigate the correlation of raw material types with particular tasks, each defined tool class was examined in terms of local versus nonlocal raw material. The consideration of raw material differences by tool types is complementary to the functional analyses also carried out as part of the lithic analysis (Bruseth and Perttula 1979:214-263) since the results can be applied to further test the initial inferences about the functional variability in Lake Fork lithic assemblages (e.g., Greiser and Sheets 1979). The mechanisms of resource procurement and the consequent availability of the "most" suitable raw material greatly depends on prevalent patterns of exchange. While some raw materials appear to have been more suitable for particular tasks than others, the predictable availability of various raw materials influenced the ultimate selection of cherts.

Chipped Stone Tool Types
The patterns of raw material uti I ization by defined tool types is comparable through all temporal periods in the Lake Fork area. Particular assemblage data are summarized in Table 7. 6.
Retouched pieces are predominantly nonlocal materials capable of producing a sharp cutting edge (cf. Wright 1980 : 212). Large projectile points (projectile points and/or hafted · cutting tools), on the other hand, were predominantly manufactured from coarse-grained quartzites, Ogallala chert, and petrified wood. Even within the Ceramic phase components , large projectile points were consistently manufactured from loca l materials.  There are no major differences in the utilization of nonlocal raw materials for the manufacture of small projectile points from the Early Ceramic through the Late Caddoan period; however, the frequency of its use is approximately 4 to 8 times as high as for bifaces, and 2 to 3 times as high as for large projectile points (Table 7. 6). Bifaces are uniformly manufactured from local raw materials (approx. 75%-100%) for all temporal periods. The functional analysis indicated that bifaces were _ utilized in a variety of ways (tasks), all probably of a generalized nature ;.-and bifaces were probably designed to have a durable working edge suitable for multiple ifunctlons. The ·local raw materials are most adequate for this type of edge.
Other tool forms such as drills, notches, perforators, and scrapers begin to appear after the Early Ceramic period, and they are almost exclusi'(ely made of nonlocal materials. The dependence on local raw materials for both utilized bifaces and large projectile points, combined with the variety . of functions represented by these tools ( Bruseth and Perttula · 1981 : Table , 6-6), argues for a relatively generalized pattern, one not altered even In the Ea.rly Caddoan period when the utilization of non local . raw materials is at a maximum.
Large "projectile points" were designed to be equally useful in both projecti te · and cutting functions.
The choice of raw materials for these forms (i.e., local coarse-grained materials) was thus a compromise to accomodate multiple functions. The specialized nature of small projectile points contrasts with bifaces/large projectile points not only in function but also in the utilization of nontocal raw materials. The small points from all ceramic components are predominantly nonloca l raw materials. The factors that contribute to the overall stability of the system are more than purely functional considerations, i.e. , selection for different edge characteristics .
There is no reason to infer that the criteria for choice of r aw material for projectile point manufacture were solely functional. The piercing quality of locally produced projectile points is probably comparable to those made on non local materials, at least judging from the presence of locally made small projectile points throughout a ll components and temporal periods in Lake Fork and other adjacent a r eas (Doehner and Larson 1978;Lynott 1975). It could be suggested that nonlocal raw materials were easier to manipulate than local materials . However, the presence of locally manufactured small projectile points of coarse-grained quartzites and petrified woods indicate that this is not necessarily the case.
It may be that when a cobble or blank of nonlocal raw material was secured, the design was to maximize its utilization by manufacturing a number of flakes suitable for use as cutting tools, rather than manufacturing a limited number of large toots.
Thus, with limited availability of an exotic raw material, its use is primarily restricted to tool forms that most efficiently utilize this material, while other tool forms will continue to be manufactured on local materials .
The importance of unretouched and retouched cutting tools in a total tool assemblage has . been aptly commented upon by Callahan in his Living Archaeology project: Aside from hafted knives and celts, our most useful little tool was the so-called "waste" flake.
We used biface thinning flakes for stripping bark for our packs and gathering baskets, for processing . cordage for fish line, for cutting meat and wild vegetal goods to size, for skinning and .butchering small game, for scraping deer hides, for processing bark strip . cordage for our shelter and raft, for carving our bone harpoons and fish hooks, for arrow shaft and foreshaft modification, for · bow planing and scraping, and for gathering virtually all the sweet flag thatching for our shelter. For jobs, such flakes were in most -cases the only tools · used.
That is to unmodified or slightly modified biface thinning flakes were a primary priority of our camp (1974:7).
With the advent of small projectile points, multiple tool needs could be satisfied with one blank or cobble; particularly so when· the availability of non local raw materials increased during the Early Caddoan period. In summary, it Is suggested that the raw material arrays during the different periods in the Lake Fork area reflect functional and maximization considerations ln. conjunction with increased availability of nonlocal materials. Local materials were consistently utilized for large tools such as bifaces and projectile point/hafted cutting tools, while nonlocal materials were . utilized for smaller tools with sharper, but less durable, cutting edges. . Differences in the presence and availability of nonlocal raw materials through time did not significantly modify this pattern.

'
Only celts were manufactured of exotic raw materials, primarily from quartzltic sandstone from the Jack Fork Formation in the Ouachita Mountains. The remainder of the groundstone tools were manufactured exclusively from local ferriginous sandstone, hematite. and quartzite (Table 7 . 6).
The similarity of celts from Early and Late Caddoan components to celt forms at the Sam Kaufman Site (Fer ring 1969:89), the lack of spalls indicative of initial manufacturing processes, and the extensive reworking, all argue for the · celts entering Lake Fork assemblages in a blank or finished state (see also Wright 1980:226). Type 2 celt forms at Sam Kaufman (the forms found in Lake Fork Reservoir) are not common nor· extensively worked, perhaps indicating that the majority of these forms were carried elsewhere, while the more common Type 1 celts at Sam Kaufman were those locally favored and not considered exchangeable goods.

Conclusions
A tentative model of lithic raw material utilization applicable to the non-Red River · watersheds of the Gulf Coastal Plain of the Caddoan Area has been outlined.. The model is based on assumptions concerning interareab:. exchange, hypothesized subsistence and territorial changes .through time, and preliminary evidence about the utilization of nonlocal raw materials in lithic assemblages. The lithic raw material data from Lake Fork assemblages were reviewed with regard to the requirements of the model parameters and ancillary functional considerations.
Many ·factors influence the kind and quantity of stone tools that appear at a given component within any one temporal period. Comparability between assemblages is essential to properly evaluate the changes in lithic raw material utilization through time.
Differing emphases on tool maintenance or tool manufacture at components can also skew lithic raw material representation.
All Caddoan period components seem, based on flake size (e.g., Raab et al. 1979) and lack of cores, to lhaye emphasized tool maintenance rather than on-site Higher frequencies of cores and petrified wood at Archaic components may ·indicate slightly different emphases; i.e., different reductive strategies suitable for the hard-to-work . silicified woods. Debltage:tool ratios (Table 7. 7) are consistent within all temporal periods for both local debitage:tool and nonlocal debitage:tool components. This suggests similar patterns of maintenance and stone In addition, the consistently low debitage: tool ratios of non local raw materials, when compared to local materials, indicate that nonlocal raw material arrived at sites in blank or finished form and was subjected to little additional alteration or recycling.
Further research on the problem of lithic raw material utilization changes · needs to take all these considerations into account, particularly comparability, adequate temporal control, and small sample sizes.
Our research procedure c-ould certainly have benefited from better control over these parameters.
The study of procurement patterns and modes of transportation of local and nonlocal materials has deliberately not been stressed In the Lake Fork study. No clear differentiation between direct and Indirect procurement processes (cf. Klinger and Mathis 1978:61) is known because of the many intervening and unknown factors that affect formation of the archaeological record.
The process of reduction conducted at outcrop locales, the possibility of finished forms or preforms being introduced into sites, and the differing manufacturing processes occurring at sites due to functional reasons ( Raab et al. 1979) all need to be considered before probable procurement methods can be specified.
Only one outcrop locale of local material is known in the Upper Sabine River Basin I approximately 40 km south of Lake Fork Creek; there are probably many others that have gone unnoticed.
Malone states that The coarse-grained I sugarlike quartzite occurs in outcrop areas in the form of large boulders. These cover 3 or 4 acres in several places near the center of the reservoir (Mineola Reservoir) areas.
Site 41VN39 shows signs of removal of considerable amounts of material.
Partially decorticated cores I as well as numerous flakes I were present (1972 : 32).
It is unclear how applicable the proposed lithic raw material utilization model is to East Texas. High-quality I fine grained silicates are common only at the northern end of the area I especially in the gravels of the Red River and on its terraces. Stream valleys to the south and west of the Red River I however I are primarily within the area of the Uvalde gravels.
Not only are Uvalde Gravels dispersed throughout upland gravel deposits and Pleistocene terrace remnants I but drainage systems cutting across the area carry this reworked material as river gravels. It is not known how far east this sheet wash material was carried during the Pliocene and Early Pleistocene. Menzer and Slaughter (1971 :220) suggest the border was possibly the Mississippi Embayment. Gravel deposits in western Louisiana are not comparable in composition to Uvalde . Gravels, being made up of 90% chert pebbles that come primarily from Paleozoic outcrops in the Ouachita Mountains (Woodward and Gueno ,1941 :37).
Tertiary uplands from Natchitoches to Shreveport I Louisiana, are barren of gravel. Quite possibly then, the Sabine Uplift of· eastern Texas (Sellards et al. 1932), which partly separates the East Texas embayment from the Mississippi Embayment I provided the limiting factor to the eastward spread of Uvalde Gravels I except for material later carried as river gravels. This area, then, is suspected to be the area of primary relevance for further testing and revising of the model.
The prehistoric inhabitants of Lake Fork utilized varying quantities of nonlocal high-quality cherts mainly available within a 100 km radius (Table 7. 8). Nearby areas such as Lake Lavon on the East Fork of the Trinity River (Lynott 1975) , Cooper Lake on the Upper Sulphur ( Doehner, Peter and Skinner 1978), Cedar Creek Lake . on Cedar Creek (Story 1965)', and areas on the upper and middle Sabine (Duffield 1959;Malone 1972:34) utilized ·approximately 90%-98% local Ogallala chert, quartzite, and petrified wood. The high quality cherts, available only 50-100 km .away in Red River Gravels, are · noticeably absent during the Early Caddoan (or Neo-American) period. Surrounding areas of high    1970:27-29;Mallouf 1976:45) that pertain to this temporal period . Occupants· within the Blackland Prairie during the A.D. 800-1300 period thus had limited interaction with prehistoric Caddoan and Austin/Toyah phase cultural groups who inhabited areas with high-quality cherts. Evidence of interaction (the presence of sherds) is more common, but in no case abundant.
In any event, what interaction there was probably ended by approximately A.D. 1000 (M. Lynott, personal communication).
A tentative "boundary" between Caddoan groups and the occupants of the Blackland Prairie can be suggested from this evidence.
High-quality cherts were distributed from the Red River area and Central Texas to the Caddoan groups farther south and east, such as Lake Fork, bypassing the intermediate and adjacent groups.
The lowland areas of eastern Arkansas have been the focus of research efforts dealing with the procurement and utilization of nonlocal lithic raw materials (House 1977;Klinger and Mathis 1978:59-63) . Coming to· grips with basic questions 1 such as the function of exchange in adaptive strategies, i.e., why exchange occurs or why there were different emphases in the utilization of raw materials during certain periods, are only some of the research problems that can be dealt with in the study of lithic raw materials (House 1977:375-376).
Raw material studies comparable to Webb's (1977:53) for the Poverty Point complex will be necessary within this region of the Southeast before some of the apparently shifting patterns in resource procurement are clarified. Brain's (1976) discussion of the varying emphasis on social and ceremonial activities over time in the Southeast is pertinent to any evaluation of what factors contributed to cultural decisions concerning lithic raw material utilization and exchange in general. The exchange network for Poverty Point, for example, is much more expanded than the Caddoan network discussed in this paper. Lithic materials from Missouri I Illinois I Ohio, Tennessee, and other distant sources, such as the Great Lakes 1 seem to be common at the Poverty Pol nt Site (Webb 1977: Fig. 28). However I more recent study of the Poverty Point lithic material (S. 1981} suggests that the variety of lithic raw materials present at the site results not from the procurement of nonlocal materials but from the wide variety of lithics found in local Pleistocene gravels. No comprehensive study of non local materials from the primary Caddoan centers, such as Spiro, Battle, or Davis, have been conducted, so this picture may be illusionary. Spiro appears to have a high variability of raw materials from sources such as Missouri and Tennessee and may be comparable to Poverty Point when fully studied (cf • . Brown 1976). Nevertheless, the implications of this behavior in studies of exchange and adaptation will remain unclear until such analysis is begun.
Hopefully, this study will contribute to further and more detailed studies, which, by taking a perspective that relates processes of subsistence, settlement, lithic assemblages, and changes in adaptation, may eventually lead to a more refined and comprehensive understanding of the archaeological record in the Caddoan area.
Acknowledgments I would like to thank James E. Bruseth for hi s comments a nd advice on the lithic analysis of the Lake Fork material, and the initial opportunity to work on the Lake Fork project. Dr. S. Alan Skinner served as Principal Investigator for the project.
Without the cooperation and assistance of Larry Banks and Kate Huckabay in the sorting and classification of lithic raw material types, this study would not . have been possible. .1 would also like to thank Kathy Boswell for typing this paper. Any errors of interpretation are a result of th is author•s shortcomings. 124 CONCLUSIONS Late Caddoan Titus phase sites are common in the Wood County part of the Upper Sabine Basin, primarily in the Dry Creek and Caney Creek drainages and along Lake Fork Creek between the confluences of these two creeks (Slciles et al. 1980: Figure 6; Thurmond 1985: Figure 6). The settlements are distributed almost exclusively along tributaries and headwater areas rather than along major streams. In the Caney Creek drainage, Titus phase sites are known at least as far north as the Attaway site (41HP15) near Como. Similar Late Caddoan settlement patterns are apparent in Harrison and Rusk counties on creeks like Hatley Creek in the South HallsvilleMineProject (LaVardera 1985),Martin Creek (Clark and Ivey 1974),and Potters Creek directly to the oorth (Webb etal. 1969). Thurmond (1981 : (Granberry 1985) in the Upper Sabine Basin that date before A.D. 1500 (Story 1981:149;Thurmond 1981:Tab1e 52).
The basic type of Titus phase site in the Three Basins subcluster is a small settlement of one to several homesteads or farmsteads. These are marked by trash and household middens that must have been deposited during an occupational episode of atroost20 to 50 years (see Good 1982). The character, extent, and content of local Late Caddoan trash middens (fable 10) reflect a basic similarity not only in the length and type of occupation, but also in the activities relating to refuse disposal at that time.
The house at the Goldsmith site was probably destroyed by the road construction, since it was probably located between the trash midden and the cemetery. It is also unfortnnate that an accurate estimate of the number of burials in the cemetery cannot be made since ideally, the relative duration of the hamlets (settlements) can be measured by the number of graves in the cemeteries and relative temporal placement of each hamlet can be detennined by seriating mortuary assern blages [Shafer 1981:156].
The ecofactual remains from trash middens would contribute important subsistence information on the Titns phase if a larger, statistically reliable sample could be obtained (e.g., Grayson 1984;Jones etal. 1983). This could be accomplished  .
with the excavation of about20 ro40percentofthe midden (4 to 8 m 1 ) if the faunal densities from our worlc at the Goldsmith site are representative. Subsistence remains with interpretive significance for the Titus phase are still limited to the Steck site (Perttulaet al. 1983), although well-preserved remains are known at several other sites that have as yet received little professional anention. Floral evidence suggests that the tropical cultigen maize (Zea mays L.) is a dietary staple, and beans (P haseolus vulgaris) were also an important food source. Nuts and seeds available in local environmental settings were also gathered, but may have been ofless importance in the Titus phase than they were between about A.D. I 000 and 1400 (Crane 1982;. Vertebrate species identified from the trash midden at the Steck site include deer, turkey, cottontail, jackrabbit, squirrel, and beaver, as well as several dog burials. Turtle and fish remains were also present but were relatively uncommon compared to the mammals and birds ; deer and turkey were the dominant species at the Steck site.
The Late Caddoan archeological sites in the Dry Creek and Caney Creek drainages oftheLakeFork basin have been included in the recently defined Cypress Cluster, which is thought to be the archeological manifestation of a series of social groups banded together in a socio-political structure analogous ro and at least partially contemporaneous with that of the Hasinai to the south and the Kadohadacho to the northeast. Four subclusters .. . are believed to represent the individual component groups comprising this affiliated group [Thurmond 1985:196].
The Three Basins subcluster comprises Titus phase occupations on Caney and Dry creeks in the Upper Sabine River Basin, and headwater areas of Little Cypress, Big Cypress, and White Oak creeks. More recently, sites attributable to the Three Basins subcluster have been identified in the Big Sandy Creek valley . The suggested dating of the Cypress Cluster to between A.D. 1600 and 1700 (Thurmond 1985:192) is as yet untested by absolute dating methods. A corrected date of A. D. 1470±80from the Steck site may mean that the Three Basins subcluster developed at an earlier date than is consistent with current chronological frameworks for the Cypress Cluster. Obtaining thermoluminescence and radiocarbon dates from sites such as Goldsmith or Steck is vital to the acquisition of new information on the development of the Cypress Cluster and the Titus phase. The primary units of analysis-the hamlet cemeteries-limit to some extent the precision and reliability of diachronic analyses in the Titus phase. The likelihood or possibility of frequent shifting of settlements on a generational basis suggests that a large number of sites would be identified within the four sutx:lusters that are actually sequent to (rather than contemporaneous with) other known sites. Synchroneity between and within subclusters is much more difficult to demonstrate with a series of archeological components in this siruation than are intraphase diachronic changes.
Nevertheless, the definition of the Cypress Cluster, an archeological unit with specific implications for the recognition of sociopolitical groupings, indicates that Late Caddoan sites in the Upper Sabine River Basin can be studied within an analytical framework superseding basic settlement patterning questions. This is because recognition of the Cypress Cluster is an initial attempt in East Texas archeology to relate archeological units to regionally meaningful sociocultural variables that have specific archeological implications (e.g., Thurmond 1985). If differences between the Dry Creek and Caney Creek localities have more than temporal significance, that is, if they represent archeological manifestations of contemJX)raneous constituent groups (Story and Creell982) within the Three Basin subcluster, sites such as Goldsmith will be important contextual Wlits for investigation in further studies of Late Caddoan prehistory. Their short occupation spans, their potential for developing close-order and fme-scale seriations, and the possibility of obtaining economic information at an analytically useful level (the individual farmstead or homestead), are ideal for investigating intraregional and interregional settlement, sociopolitical and adaptive variability between about A.D. 1400 and 1700. The individlf.l farmstead or homestead is the most common type of Late Caddoan settlement in the Upper Sabine Basin, but one that until recently has received little professional archeological scrutiny. Hypotheses about the formation, development, and eventual disintegration of Late Caddoan sociopolitical groupings in East Texas can be evaluated, where appropriate, with the type of Late Caddoan archeological record kn?wn to exist at sites such as Goldsmith throughout the Dry and Caney Creek Attempting to model archeological contemporaneity for the Titus phase as a whole will certainly require a more chronological data base than now exists if sociopolitical interpretations and considerations of adaptation are to be seriously considered, for as Story has pointed out, Perttula, SkUes, and Yates-Goldsmith Site certainly the Caddoan area is neither environmertally or culturally homogenous. Different processes could have been in operation simultaneously and it must be established, not assumed, that the sequence of cultural change in one locale applies to another.
We hope our research at the Goldsmith site will contribpte to a better understanding of the Three Basins subcluster of the Titus phase in the Dry Creek basin of East Texas, and serve to initiate more systematic and lbng-term research on Late Caddoan archeology in the Upper Sabine Basin and throughout East Texas.
Another Look at an Eighteenth-Century Archaeological Site in Wood County, Texas TIMOTHY K. PERTTULA AND Bon D. SKILES* T HE FRENCH PRESENCE IN EAST TEXAS DURING THE t:IGIITEENTH century is less well known from an archaeological or archival standpoint than is the Spanish. Although it is known that the French maintained several trading establishments within this part of the state, concrete evidence of these places is presently lacking. Archaeological sites that are the material remains of these trading establishments offer the best, and perhaps the last, remaining opportunities to understand and assess the mutual effects of acculturation and adaptation on French entrepreneurs and native groups such as the Caddo and Wichita living in the area.
If the eighteenth-century site known as the Woldert site is the location of the French trading post called Le Dout, the site is especially important to future historical and scientific research dealing with European-Indian contact and interaction. If, on the other hand, the Woldert site represents an aboriginal encampment occupied after initial European settlement and exploration, study of the archaeological record provides an excellent opportunity to address how Caddoan or Wichita societies changed because of their dealings with the French and Spanish. Whichever is the case, the Woldert locality offers rich archaeological, ethnographic, and archival evidence from which to study processes of culture change and thereby gain a clearer and more detailed perspective on Indian-European interaction and adaptation in eighteenth-century Texas.
The Upper Sabine Basin of East Texas in the eighteenth century was part of the area between the Upper Angelina and the Red River de-Friends of Northeast Texas Archaeology, Special Publkation No. 1 (1995) scribed by Herbert E. Bolton as uninhabited territory. 1 The original inhabitants were probably Caddoan-speaking groups, agricultural peoples who had lived in the area for several thousand years before the Europeans arrived in Texas. 2 By the time French and Spanish settlers began to colonize the Red River Valley and parts of East Texas circa these agricultural tribes were organized into entities known as confederacies. 4 The principal confederacies were the Kadohadacho and Hasinai. 5 They lived on the Red River and on the Neches and Angelina rivers, respectively, in East Texas and adjoining areas of Arkansas, Louisiana, and Oklahoma.
Except for scattered archival and archaeological evidence, little is known about what happened to the Caddoan groups who had lived in the Upper Sabine Basin before 16go. 6 When Anglo-American pioneer settlers reached this part of the East Texas frontier about 1840, the only Indian groups living there were Shawnee, Delaware, Choctaw, Quapaw, and Cherokee groups who had originally resided east of the Mississippi River. 7 Southern Wichita-speaking groups such as theTawakoni, Taovayas, and Yscani had moved into the Upper Sabine Basin from the Arkansas River Valley in the middle of the eighteenth cen-and archaeological records 11 is explored here with a reexamination of the eighteenth-century site now named the Woldert site. 11 Known since the 187os, and described in this journal in 1952, the site has generally been overlooked since it was first discovered.
The Woldert site is intriguing for several reasons. First, it is located north of the Sabine River and near to known historic eighteenthcentury Caddoan and Wichita settlements. Second, the large quantity of European manufactured goods found there suggests the existence of a sizable encampment. Finally, available archival information describes one or two eighteenth-century French "factories" or trading posts that were located in the vicinity.' 4 Thus, the site could conceivably represent an important aboriginal settlement as well as a French trading establishment set up to deal in the deer and peltry trade.' 5 East Texas include the Natchitoches Archives, the Records of the Cabildo (the Spanish administrative bureaucracy in Louisiana) available at the New Orleans Public Library in New Orleans and the Spanish Governor Dispatches in the Howa rd-Tilton Memorial Library at Tulane University in New Orleans. See A. 0. Hebert, "Resources in Louisiana Depositories for the Study of Spanish Activities in Louisiana," in The Spanish in Ill '-Mississippi Valley, 1762-18o4, ed. In 1952 the Southwestern Historical Quarterly published a short article by Albert Woldert on a possible Indian battle in Wood County, Texas. 16 Found at the Woldert site (41 WD333) were copper and . brass fragments, iron knives, iron hatchets, glass trade beads, lead balls, broken and smashed French flintlock musket barrels (estimated to date about 1770), an undated silver coin, and a copper cross inscribed with "Holy Mother" in Spanish. "Tomahawks" and stone arrowheads were also reported to have been found in association with these Europeanmanufactured goods. 17 These artifacts were found near a large artesian spring 18 on Mill Race Creek about two miles south of Hainesville, Texas ( fig. 1), and about ten miles east of Mineola, Texas. Because of the large number of battered and broken old gun barrels found around the spring, Woldert concluded that late in the eighteenth century "Indians living near the large spring may have been surprised by armed soldiers rushing down the high hill and may have retreated eastward up Mill [Race] Creek valley, or perhaps a battle could have begun toward the east and ended at or near the spring where most of the relics were found ." 19 To unravel circumstances in which the historic materials were found, it is necessary to evaluate the site's local legend and lore, which may help to determine its archaeological context and research potential. 20 The collection of artifacts described by Woldert has been recently studied and photographed by the authors. Woldert was able to examine the gun barrels and some gun parts in the collection of Frank Haines, the son of Christian Haines, on whose property they were discovered, but he did not describe the artifacts in any detail. The collection belonging to Frank Haines was handed down to his daughter, Ruth Haines DavisY Our aim in restudying the collection was to determine the age and context of the artifacts, including those apparently not mentioned by Woldert, as well as their origin and use. Specifically we hoped to determine whether the site's artifacts of European manufacture had been or employed in a manner consistent with either Indian or Eu-  ropean patterns of usage. 22 Various alterations in the manufacture, form, or apparent function of European-derived material remains are utilized hy archaeologists to infer aspects of aboriginal change in acculturative situations. The dating of European-manufactured goods such as knives, glass trade beads, and flintlock guns has been considerably refined since the making chronological estimates fairly precise for archaeological sites presumed to date to the eighteenth century.
Woldert noted that the artifacts were found and collected from various farms on the W. M. Kern and W. M. Patton surveys in a .two-milewide area around the spring ( fig. 1). Local information indicates, however, that the majority of the guns were actually found accidentally by ditchdiggers working on Christian Haines' water mill.
According to a local resident, Johnie Moody, about 1870 Haines built a water mill in the Hainesville area. This mill was powered by water from an artesian spring only one-half mile northeast of the mill site on Mill Race Creek. Irishmen hired by Haines to dig the race ditch exposed at least twenty-five "antique" rifles, enough to cover the bottom of a wagon bed. 24 These same ditchdiggers uncovered an unmarked Caucasian burial in 1874 on the Joe Moody Farm, located in the W. M. Kern Survey. 25 This burial presumably predates the Anglo-American settlement of the redlands and Hainesville areas, which began around 184 1. 26 The burial had been placed in a split and hewn log used as a wood coffin. Local legend has it that this burial was the body of Sieur de La Salle, the French explorer who was murdered somewhere in East Texas in March, 1687, while searching for the Mississippi River. 27 This putative associa-22 J elfrey Brain, Tunica Treasure, Papers of the Peabody Museum of Archaeology and Ethnology, Harvard University, Vol. 71 (Cambridge; Peabody Museum of Archaeology and Eth· nology, 1979), 271-274. Oiscussing histo ric contact period sites, Brain distinguishes artifacts of aboriginal manufacture, which represent traditional techniques o f manufacture, use, and function (such as pottery and stone tools), from those of aboriginal and European manufacture, which required new materials and techniques of manufacture, form, and function. These could include such unmodifit:d items as brass gorgets, tinkling cones, axes, hoes, and firearms. When these are found on an archaeological site, they can represent various stages of innovation and aboriginal acculturation as well as evidence of European habitation. Thus, it is essen· tial that a knowledge of artifact context and association be obtained from a site to sort out European influence on aboriginal peoples from an actual occupation by a European group. 2 ' Ibid. , Johnie (Mrs. A. L.) Moody, "Reminiscence of Hainesville," in Chips of Wood County, comp. Adele W. Vickery (Minc:ola; Adele W. Vickery, 1gug), Part 2, 3 (quotation), 4· 2s Wood County Democrat (Quitman), Aug. 6, 1go8. 26 Wood County, t8JO-I900 (Quitman;Wood County Historical Society, 1976).
the actual location where Sieur de La Salle was murdered is unknown, current thinking suggests it was in the vicinity of the Trinity River, at least 120 miles south of the Woldert site. See RobertS. Weddle, Wildemess Manhunt: The Spani<h Search for LaSalle (Austin; Univer-tion with La Salle 28 was based primarily on the fact that many old guns had recently been found on the nearby Haines farm. A tgo8 article in the Wood County Democrat had noted that the guns and other relics found around there, and more especially [that) the guns were all bent or broken, showing the fact that the superstition of the Indians had been aroused owing to the fact that the bullets corning out of these guns killed their comrades, and they could not understand why such weapons should so mysteriously kill when there was nothing about them that they could see, to produce death. Mr. C. H. Haines, in his lifetime gathered up quite a number of these bent and broken guns. 29 The obvious antiquity of the burial and the rifles found nearby suggesting La Salle's demise by Indians who had then broken the Frenchmen's guns, contributed by the 1940s to the general notion that the area was the site of an Indian battleground. Woldert appears to have been convinced by the county lore. In a 1946 article the Mineola Monitor described the Haines collection as "old gun barrels which had been battered and broken in two, indicating a fight to the finish with Indians. It was the practice of the Indians in the early days to destroy the mysterious death-dealing firearms whenever they captured them." 30 Woldert's study of these guns ultimately removed the La Salle association because Woldert felt that the flintlocks in the collection dated to the late 17oos. 31 The evidence summarized in Woldert's article, combined with local legend, does indicate that a considerable quantity of European goods has been found around the artesian spring and at other places on Mill Race Creek. The number of guns found together in the mill race ditch implies a cache of guns, 32 though it is still unclear whether the cache was deposited by Indians or Europeans. Giving a European attribution to the burial on the joe Moody Farm seems logical in view of its apparent age and unusual mode of interment, but its association with the gun cache is still unknown.
Aboriginal and European-manufactured goods in the Haines collection are a composite of occupations on Mill Race Creek beginning sity of Texas Press, 1973); and Henri J outel, A Journal of the Last Voyag-e Perforrn'd by Monsr. de ln. Sale, lo the Gulph of Mexico ... (1714;reprint, New York: Corinth Books, 1962), 102-104. 28 Wood Cou•uy Democrat (Quitman), Aug. 6, 1908; Moody, "Reminiscence of Hainesville," 4· 29 Wood C01mty Democrat (Quitman), Aug. 6, 1908. 30 Miruwln. Monitor, Nov. 22 , 1946. 31 Woldcrt, "Relics of Possible Indian Batdc in Wood County, Texas," 487. "Caches of flintlock guns have been found al several aboriginal Indian sites of eighteenthcentury age. Sec T. M. Hamilto n (comp.), "Im.lian Trade Guns," Misso-uri ArclttU!ologist, XXII (Dec., 1g6o), 150-171; J ay C . Blaine and R. K[iug] Harris, "Guns," Bulletin of lht Texas Archeological Society, XXXVII (1967), 33-86. about 10,ooo years ago." None of the aboriginal artifacts included in the collection are definitely of eighteenth-century manufacture. However, ground celts and clay pipes of similar form and manufacture have been recovered from such historic period sites in East Texas. 34 A fragmentary clay pipe of "ring-base" style from the Woldert site has been recovered from the possible site of the French post (dating from the 1730s to the 1770s) among the The remainder of the collection is dominated by artifacts of eighteenth-century French manufacture. In addition to the fourteen octagonal gun barrel and muzzle sections, a number of other gun parts and iron, brass, or copper artifacts were present. Unfortunately, the glass trade beads, coins, and cross mentioned by Woldert are missing from the collection.
The guns represented in the Wolden site collection include at least four TypeD French fusils or light muskets. 56 The French guns are colonial frontier Aintlock muskets manufactured between 1730 and 1765. These types of guns were made in France for trade, where they were shipped to the Colony of Louisiana for eventual distribution to Indians in exchange for pelts, bear oil, horses, and other supplies and as annual presents. 37 · The Type D gun is noted in several Wichita and Caddoan eighteenth-century archaeological sites in East Gun-barrel bore measurements and lead ball diameters suggest that the guns were 28 to 32 calibre, 39 the standard bore for French trade guns of the eighteenthcentury.40 Foliate scrolls and hunting scenes were engraved on the two side plates; these were common French designs for flintlocks of the 1730S. 41 A bow-quiver design was engraved on one of the four cast brass butt plates. 42 Other gun parts include a trigger guard mainsprings, two gun cocks, and a breech plug. The barrels include octagonal breech sections and round muzzle sections. None appear to have been flattened or deformed from breakage, though such flattened and reworked barrel sections have been reported from Indian sites where the barrels were made into diggers, stakes, fieshers, and scrapers. 44 One of the iron tools in the collection resembles an adze or scraper, and it has been reworked from a piece of metal scrap 40 millimeters wide. Similar artifacts, formed from barrel hoops, were recovered from the nearby Gilbert site. 45 Two single-bitted iron axes are in the collection, one of which appears to have also been used as a wedge. These small axes reportedly cost fifteen sols apiece in the eighteenth ccntury, 46 roughly equivalent to the cost of two deer pelts in the French Louisiana trade. 47 There are several fragments of cast brass kettles, including a repaired and riveted kettle bail ear. 48 These types of kettles, with a fiat bottom, straight sides, and a diameter of up to 50 centimeters are quite common trade items at the 1730-1763 Tunica Indian Trudeau site. 19 The three iron knives in the collection are very similar to French clasp knives, though no names of the French manufacturers are discernible on the heavily rusted blades. Knives of this type were popular trade items and have been found at a number of Caddoan and Wichita sites contemporaneous with the Woldert The only known aboriginal groups living in the Sabine Basin during the middle of the eighteenth century were the Nadaco (or Anadarko) Caddo on the Sabine River near the Rusk and Panola county and various Wichita groups on the Upper Sabine and its tributaries near the Blackland Prairie in Rains County ( fig. 2). Wichita groups included the Taovayas, Tawakoni, and Yscani and the separate group of Kichai, though, according to De Mezieres, in 1770 there were other Wichita groups living on the Sabine River. 5 2 Sometime between 1542 and 1717 Nadaco settlements were split between those on the Sabine River and newer settlements in the vicinity of the Nacao Caddo in the Angelina River A 1717 map 54 locates the southern Nadaco group near the Hasinai. Also depicted on art! Type A, Variety 1 kettles de fined by Brain, Tunica Treasure, 165. Similar kettles were also found at the Womack site; see Harris, Harris, Blaine, and Blaine, "A Prelirninary Archeological and Documentary Study of the Womack Site, Utmar County, Tt!xas," Figure Present evidence suggests that Vial crossed the Sabine River in southwestern Wood County near its confluence with Grand Saline Creek, 64 about twenty miles west of the Wolden site. From the Sabine River crossing southeast to the Nadaco village was 25.5 leagues. The village had between thirteen and fifteen houses scattered over 3 leagues, but these were evidently located along tributaries of the Sabine River because the second Sabine River crossing by Vial was 5 or 6leagues to the east. 65 The presumed location of the Nadaco "village" in 1788 is in the vicinity of Tatum and Carthage, Texas-the same part of the Sabine River Basin in which post-168o historic Caddoan archaelogical sites (referred to as the Kinsloe Focus) have been found. 66 Known Kinsloe Focus sites extend from Longview, Texas, to the vicinity of Carthage, and all are located on tributaries to the Sabine River. This territorial relationship need not necessarily imply that the Kinsloe Focus is the archaeological correlate of the Nadaco Caddo, though the initial archaeological and ethnographic associations are supportive.
The American State Papers, Documents of the Congress of the United States (t78o-t815), include several valuable documents by John Sibley concerning eighteenth-century French trading establishments among aboriginal groups on the Sabine River. 67 French traders lived among the many different aboriginal groups in the Caddoan area, operating as traders and coureur du bois from Natchitoches, New Orleans, or other French enclaves.filj Gregory notes that "French hunters operated from the Natchitoches post on a sort of share cropper basis. Men were outfitted with French firearms, flints, powder and shots to go to the vicinity of the Wichita to hunt hides. several times been at the place; the French flag used to be hoisted there, and there are the remains of the buildings and works now to be seen; and that the Dout is about 150 miles northwest from Natchitoches. 77