Refuge Series Overview

Refuge wares were first identified on the lower Savannah River at the Savannah National Wildlife Refuge. This is an ambiguous "type" as most people who use the name seem to believe that dentate stamping or "Allendale Punctate" is the defining characteristic and thus any time that decoration is seen the pottery is called "Refuge" and interpreted as an Early Woodland diagnostic. In fact the dentate stamp decoration makes up less than 1% of the pottery found at the type site. So when it is found in large numbers elsewhere it may well represent a different type of a different age and culture.


Refuge ceramics were recognized by Waring (1968b) as an intermediate series between Stallings and Deptford, based on the excavation of four 5 foot squares opened in 6 inch levels to a depth of seven feet into a small shell midden (38JA5) on the South Carolina side of the Savannah River. The site was located on the Savannah National Wildlife Refuge, hence the origin of the site and series name. Five Refuge types were identified and described by Waring (1968b:200) based on a sample of 683 sherds: Refuge Simple Stamped (N=200; 29.3 percent of Waring's sample); Refuge Plain (N=403; 59.0 percent); Refuge Dentate Stamped (N=45; 6.6 percent); Refuge Punctated (N=33, 4.8 percent); and Refuge Incised (N=2, 0.3 percent). Recognition of Refuge Phase sites has been highly confused, however, primarily because most of the Refuge types (as defined) are similar or identical to established types in the Thom's Creek and Deptford series. Only Refuge Dentate Stamped, of the five original Refuge types, for example, can be unambiguously sorted, primarily because the characteristic exterior finish does not occur on earlier or later wares in the immediate region.

Formal type descriptions of Waring's five Refuge types were published by DePratter (1979:120-123), based on collections from the mouth of the Savannah River (including from the Refuge site). Two phases were proposed, Refuge I (1000-800 BC) and Refuge II (800-600 BC), the former characterized by plain, incised, simple stamped, and punctated types and the latter by the addition of dentate stamping and the disappearance of punctations (DePratter 1979:113, 117). DePratter's (1979:120-121) Refuge Punctated and Refuge Incised types, as defined, are indistinguishable from Thom's Creek punctated and incised types (e.g., Trinkley 1980a:353), and the Refuge I phase may actually reflect the adoption of Thom's Creek wares and manufacturing technology on the north Georgia coast. Refuge Dentate Stamped appears somewhat later, during what DePratter (1979:117) called late Refuge I through Refuge II.

At Clear Mount on Groton Plantation in Allendale County, South Carolina, and from other sites in the area, Peterson (1971a, 1971b) documented the presence of a Thom's Creek assemblage intermediate between Stallings and Refuge. Wares identifiable as Thom's Creek Reed Separate Punctate were replaced by distinctive Refuge types, with surface finishes characterized by dentate stamping, and by what Peterson (1971b:77) called Irregular Punctate, which is similar or identical to Stoltman's (1974:276277) Allendale Punctate. The Refuge Dentate Stamped and Allendale Punctate types are gradually replaced by simple stamped wares at Groton Plantation, a trend also noted by DePratter (1979:117-118) from the mouth of the Savannah.

At Mattassee Lake along the lower Santee a large sample (n=102) of Refuge Dentate Stamped sherds with a temperless to clay-grog tempered paste were found, a paste previously unreported within the Refuge series. The occurrence of "grit and sand in considerable quantities" (DePratter 1979:121) characteristic of Refuge wares from the lower Savannah River was not at all evident. What these marked differences in tempering between the two areas means in cultural terms is presently unknown. No abraders were noted in the Mattassee Lake Refuge sherd assemblage, unlike the situation along the lower Savannah where these tools are common (DePratter 1979); this absence may be due to the nature of clay-grog paste which, lacking much sand, was not suitable for abrading. The clay-grog paste in the dentate stamped sherds from the lower Santee is for all practical purposes identical to the paste in the Hanover/Wilmington wares from the same area, and a similar manufacturing technology is evident. The series may be related, or evolve from one to the other.

Refuge assemblages, which are characterized by punctate, random punctate, twig impressed, dentate, and simple stamped finishes, have been dated to between 1000 and 600 BC along the lower Savannah River. Stratigraphic evidence from the two wildlife refuge sites and from Clear Mount indicates that the punctated and dentate stamped finishes, as well as simple stamped and plain, occur early in the period, and are replaced by assemblages dominated by plain and simple stamped finishes. These distinctions are marked by the creation of two subphases, each two hundred years in length, to accommodate the initial early Woodland period locally. Key diagnostic indicators of the Refuge I phase (ca. 1000-800 BC) in the Middle Savannah ceramic sequence include the Refuge Punctate (vars. Refuge, Allendale) and Dentate Stamped types. The dentate stamped finish is characterized by linear, typically parallel arrays of small rectangular impressions or dentates that occur over the vessel's exterior surface. Geometric linear arrangements of dentates, usually forming triangles, or random arrangements of dentates, exhibiting no obvious pattern, occur much less commonly. Two varieties of Refuge Punctate are evident, the first (var. Refuge) characterized by randomly spaced punctations typically applied with a reed. The absence of linear or geometric arrangement is what distinguishes this ware from earlier Thom's Creek Punctate varieties. Refuge Punctate var. Allendale is characterized by small, irregular and frequently angled punctations that appear to have been made by a bunch of straw or small twigs (Stoltman 1974:276).

The Refuge II phase (ca. 800-600 BC) in the Middle Savannah ceramic sequence is characterized by plain and simple stamped ceramics. There is no evidence to suggest that the earlier punctated and dentate stamped finishes carried over into this period. Because simple stamping occurs widely during the Woodland period locally, components of this period are difficult to recognize unless fairly large assemblages are available. Refuge Simple Stamped pottery is characterized by carelessly executed and applied U- and V-shaped longitudinal grooves. The stamp impressions are frequently irregularly spaced with respect to each other and haphazardly applied, although care in stamp execution and application may vary considerably. V-shaped impressions dominate assemblages, although U-shaped grooves may occur as a minority finish. Parallel stamping tends to be the principal method of application, although cross stamping commands an appreciable minority of the sherds in many assemblages. The vessels are occasionally lightly to extensively smoothed after stamping.

Refuge wares seem to have a distribution similar to that of the previous Stallings and Thom's Creek series, and appear to evolved directly from them. Refuge components have been documented along both the Georgia and South Carolina coast and well into the interior of the Coastal Plain in both states (Anderson 1975; DePratter 1976, 1979). Along the lower Santee River in South Carolina major Refuge assemblages have been examined at the Mattassee Lake and Minim Island sites in recent years (Anderson et al. 1982; Brockington and Espenshade 1989), and work in the Sea Island area of Georgia indicates many Refuge period sites lie buried in the marsh, having been covered by rising sea levels (DePratter 1976, 1977). Although regional variation in Refuge assemblages has seen little examination, Refuge assemblages along the Santee River appear to be characterized by a higher incidence of dentate stamping, and a much lower incidence of simple stamping, than Refuge assemblages found along the Savannah River. At least some of the Refuge materials along the Santee, furthermore, are characterized by grog-tempering, something not observed at all during this period along the lower Savannah (Anderson 1982).

Initial Early Woodland assemblages elsewhere in the Savannah River Valley differ from the situation in the middle part of the basin to varying degrees. In the mouth of the Savannah sequence three Refuge phases have been advanced (DePratter 1979). Plain and simple stamped finishes are observed during all three phases, which are differentiated by the occurrence of punctated and incised finishes (Refuge I, ca. 1100-1000 BC), dentate stamping (Refuge II, ca. 1000-900 BC), and linear check and check stamping (Refuge III, 900-400 BC). Given the absence of absolute dates in or near the basin supporting subdivisions as fine grained as those advanced for the Refuge I and II phases, or for such an early appearance of Deptford as suggested by the range offered for the Refuge III phase, a more conservative approach has been taken in the definition of the two Initial Early Woodland Refuge subphases advanced in the middle Savannah sequence.

No evidence for Refuge or other major Initial Early Woodland period ceramic assemblages have been found to date to the north of the Fall Line in the Savannah River basin (Anderson 1988c). Virtually nothing, in fact, is currently known about the Initial Early Woodland period in the upper part of the basin. A continuation of typical Coastal Plain sequences has been documented at several sites near the Fall Line (Ferguson and Widmer 1976), but beyond this point little is known.

Primary references: Edwards (1965:24); Waring in Williams (1968:200); Peterson (1971a, 1971b); Anderson 1975a, 1975b; DePratter (1976, 1977, 1979:122-123); Trinkley (1980a); Widmer 1976a; Herold and Knick (1978, 1979a); Lepionka (1980, 1981, n.d.); Anderson et al. (1982:264-268).