In southern New England, typological distinctions between Niantic, Hackney Pond, and Shantok ceramics have been used to describe changing settlement patterns for the Late Woodland (500-350 Years B.P.) and Contact (post A.D. 1600) Periods. Based on the initial typologies developed by Rouse (1947) the Shantok ceramic tradition was also considered an ethnic marker of the Mohegan and Pequot tribes based on material recovered from Fort Shantok. Re-examination of stylistic data have suggested that levels of stylistic similarity between late ceramic types actually limit their use as ethnic markers in reconstructing Contact Period settlement patterns (McBride 1990; Lizee 1994). In this study, neutron activation analysis is employed to determine if compositional profiles correspond with identified stylistic types. The distribution of compositional groups within the region prove to be useful in describing changes in settlement during the Late Woodland and Contact Periods for southeastern Connecticut. Results of this study suggest that cultural factors underlying the evolution of the historic Pequot and Mohegan tribes, and locations of focal village sites, also had an impact on access to clay resource zones at the time of European Contact.

In southeastern Connecticut the final indigenous ceramic horizon is defined by three stylistic types: 1) Niantic; 2) Hackney Pond; and 3) Shantok (Figure 1). Similarities between types include use of incising and shell stamping in horizontal, vertical, and opposed oblique bands in the area of the collar. While the three types share some degree of stylistic similarity they are also characterized by gross differences in physical attributes that are identifiable using optical techniques: Niantic ceramics are usually shell tempered (Rouse 1947; Lavin 1980); Hackney Pond ceramics are described as lacking in temper (McBride 1990; Lavin 1986); and Shantok ceramics are always described as shell tempered and are further characterized by a "soapy" surface texture (Rouse 1947).

Niantic, Hackney Pond, and Shantok ceramic types have been interpreted as representing two contemporary traditions, Windsor (Niantic and Hackney Pond) and Shantok. Niantic and Hackney Pond ceramics are considered to represent late stylistic and technological developments within the local Windsor ceramic tradition. The classification of Shantok as a distinctive tradition is based on the high density of shell temper, thick collars, and castellations (Rouse 1947). Niantic and Hackney Pond ceramics lack the distinctive morphological stylistic attributes of Shantok (frilled collars and castellations) but are similar in the number and kinds of design elements used in decorations, shell-stamped or incised interlocking triangular designs being common to all three types.

In the initial typology for southern New England ceramics, Rouse (1947) suggested the Shantok ceramic tradition represented an ethnic marker for Mohegan and Pequot tribal affiliation. More recent archaeological and ethnohistoric investigations have suggested that the Shantok Tradition originated as a post-Contact (ca. 1637 A.D.) ceramic type produced exclusively by the Mohegan Tribe, while the Pequots continued to utilize Niantic and Hackney Pond variants of the Windsor tradition (McBride 1990; Lizee 1994). Stylistic analysis of contemporary Windsor and Shantok assemblages has documented stylistic similarity between these traditions. In southeastern Connecticut, stylistic sharing has been interpreted as an indicator of interaction and information exchange between historic tribal groups. Distinctions in ceramic paste and vessel morphology, however, suggest that clay resources and vessel construction techniques were unique to each tradition. The distributions of Windsor and Shantok Tradition ceramics in southern New England show a considerable geographic overlap during 16th- through 17th-centuries, which may account for stylistic similarity but not for differences in technology or morphology. Compositional analysis provides a new strategy for examining changes in settlement and exchange between the groups which became known as the Pequots and the Mohegans during the 17th-century (Figure 2). In this study, neutron activation analysis of 70 sherds from nine Late Woodland and Contact Period sites is used to identify compositional groups within a sample of Windsor and Shantok Tradition ceramics. Because Shantok and Hackney Pond ceramics are considered technologically distinct, neutron activation analysis provides a useful technique for assessing the association between clay resource zones and tribal territories that have been suggested by stylistic and ethnohistoric data (Rouse 1947; Williams 1972; McBride 1990).

Late Woodland Ceramics (ca. 500-350 Years B.P.)
In southern New England, the Late Woodland Period is marked by the appearance of a ceramic horizon referred to as Niantic (Rouse 1947). Niantic ceramics are characterized by smoothed interior and exterior surfaces, low relief collars, elongated globular bodies, and decorative motifs consisting of interlocking triangles in the area of the collar (Figure 1). Identified types within the Niantic horizon include Niantic-stamped, stamp/drag, linear dentate, incised, and punctate (Lavin 1986). These types are contemporary with the Owasco, Castle Creek, and Bainbridge types of the Owasco Tradition found in New York State and are stylistically similar (Ritchie and MacNeish 1949). The Niantic type dominates Late Woodland assemblages across southern New England. In this region the primary differences among Niantic assemblages are in terms of temper. Examples recovered from interior and riverine regions are usually mineral tempered, while assemblages recovered from the coast are predominantly shell tempered.

Contact Period Ceramics (post- A.D. 1600)
The early part of the Contact Period saw the appearance in southeastern Connecticut of two contemporary ceramic types, Hackney Pond and Shantok. These types can be distinguished from one another by morphological and gross technological attributes. Shantok vessels are castellated, while Hackney Pond ceramics are not. Shantok vessels have thick collars with applied decorations in the form of lobes (Rouse 1947), while Hackney Pond vessels have low relief collars and lack applied elements (McBride 1990; Lizee 1994). Shantok ceramics are usually tempered with medium to fine grained shell and Hackney Pond ceramics are described as having little or no discernable temper. While the morphological differences are obvious and easily recognized, the technological attributes used to distinguish the two types are more difficult to discern and lack precision.

Niantic, Hackney Pond, and Shantok ceramics share a basic suite of decorative motifs, suggesting that they are all derived from the local Windsor Tradition. Variability within the Late Woodland Period is largely restricted to morphological changes which appear to intensify during the Final Woodland Period (ca. A.D. 1600), giving rise to late Windsor Hackney Pond and Shantok types. From this perspective, Niantic types are best considered the stylistic ancestors of both Hackney Pond and Shantok ceramics. Stylistic similarity between the Hackney Pond and Shantok types is obscured by differences in morphology and technology. Lizee (1989) has proposed reclassification of late Windsor Tradition ceramics to include Shantok as a castellated sub-type rather than as a tradition. Because the key distinctions between Shantok and Hackney Pond ceramics are morphological and technological, the task of classifying small collections from sites lacking collars often hinges on the presence or absence of shell temper.

In Late Woodland and Contact Period ceramic studies, Shantok ceramics have been used as an ethnic marker of the Mohegan tribe (Rouse 1947; Williams 1972). Excavations conducted at the principal Mohegan village, Fort Shantok (Site 86-02), and assemblages from several coastal sites in the region had been used to associate the Shantok type with the historic Mohegan Tribe (Williams 1972). More recent archaeological surveys of southeastern Connecticut have identified a second contemporary, and possibly earlier, type referred to as Hackney Pond. McBride (1990) and Lizee (1994) have both concluded that, while Hackney Pond ceramics have been associated with historic Pequot occupations, they have also been recovered from sites predating European Contact (ca. A.D. 1614). In central Massachusetts, Byers and Rouse (1960) classified late non-castellated vessels as a separate tradition which they labeled 'Guida'. In this respect, Hackney Pond ceramics are not as precise an ethnic marker as Shantok. Hackney Pond ceramics have been recovered from a range of site types and ecozones in southern New England. Shantok ceramics are more limited in distribution to large palisaded villages in coastal and riverine settings.

Materials Studies in the Northeast
To date, no compositional studies of Windsor ceramics have been conducted at the regional level. Preliminary studies of single sherds and single assemblages have provided little useful data in the identification of regional compositional groups (Philpotts and Wilson 1994). The closest region that has been the subject of compositional studies lies to the north in Quebec and southern Ontario (Trigger et al. 1980; Crepeau and Kennedy 1990). X-ray fluorescence and neutron activation analysis have been applied to Iroquoian assemblages with varying degrees of success.

Most recently, Crepeau and Kennedy (1990) have reported a high degree of compositional homogeneity within sites and few clear distinctions between sites occupied by the Saint Lawrence Iroquois. Chemical homogeneity within assemblages suggested that most Late Woodland pottery was of local origin, and that distinctions between ethnic (Iroquois) groups was problematic at the material level. The results for the Saint Lawrence River valley were found unusable for purposes of describing protohistoric settlement or trade patterns. Compositional profiles could not be formed to distinguish Saint Lawrence Iroquois from other Iroquoian or Algonquian groups.

The sample of ceramics submitted for neutron activation analysis was drawn from collections housed at the Laboratory of Archaeology, located at the University of Connecticut. The sample included 70 archaeological specimens from nine sites and five raw samples from four clay sources. Archaeological and ethnohistoric research in southern New England provides a dataset of well documented Contact Period Mohegan and Pequot sites with associated ceramic assemblages (Williams 1972; McBride 1990). Documented Pequot (N=2) and Mohegan (N=1) village sites dating to the 17th-century were examined and compared with ceramics from five contemporary seasonal camps and modern samples collected from raw clay sources (Table 1). A fourth Contact Period village site located on Block Island was also included in the sample. Historic documents indicate Block Island inhabitants had social and economic ties with both the Pequots and Mohegan groups. Archaeological specimens include members from the Niantic (n=9), Hackney Pond (n=17), and Shantok (n=25) types with an additional 19 specimens of uncertain typological affiliation.

Analytical Procedure
Neutron activation analysis (NAA) was performed at the Missouri University Research Reactor (MURR) using procedures described in detail by Glascock (1992). Ceramic samples were first irradiated for five seconds, then allowed to decay for 25 minutes before acquiring gamma spectra on a high resolution germanium detector for 720 seconds. This count identified the short-lived elements aluminum (Al), barium (Ba), calcium (Ca), dysprosium (Dy), potassium (K), manganese (Mn), sodium (Na), titanium (Ti), and vanadium (V). Samples were then allowed to decay for two weeks, after which the entire batch of samples was subjected to a 24 hour irradiation. After irradiation, samples decayed for seven days and were then counted for 2000 seconds on a high-resolution germanium detector coupled to an automatic sample changer. This count yielded determinations of seven medium halflife elements: arsenic (As), lanthanum (La), lutetium (Lu), neodymium (Nd), samarium (Sm), uranium (U), and ytterbium (Yb). After an additional three week decay, a final 10,000 second count was conducted on each sample. The last measurement yielded 17 long halflife elements: cerium (Ce), cobalt (Co), chromium (Cr), cesium (Cs), europium (Eu), iron (Fe), hafnium (Hf), nickel (Ni), rubidium (Rb), antimony (Sb), scandium (Sc), strontium (Sr), tantalum (Ta), terbium (Tb), thorium (Th), zinc (Zn), and zirconium (Zr). Two irradiations and three counts provided concentration data for 33 elements. These data were tabulated and stored in dBase III and Lotus 1-2-3 files for further analysis.

Data recorded in parts per million for each element were transformed to log base 10 values. This transformation compensates for the differences in magnitude between major elements, such as Al and Fe, and trace elements. After verifying the quality of the data and carrying out the necessary transformations, a series of pattern-recognition and group evaluation analyses was undertaken. The goal of such analyses is to recognize compositional subgroups that are archaeologically meaningful (Bishop and Neff 1989). Realizing this goal entails successive formulation and testing of working hypotheses regarding the number and make-up of groups in the compositional data. Working hypotheses can be derived from a preliminary sort of a ceramic collection, which may suggest potential groups that constitute a starting point for analysis; from provenience information; from pattern-recognition techniques such as hierarchical cluster analysis; or from a combination of these sources of information.

The basic approach to pattern recognition used in this study was to inspect the multidimensional compositional space defined by elemental concentrations through two-dimensional windows (bivariate scatterplots). Scatterplots of the original elemental concentrations were examined, but derived spaces defined by linear transformations of the original data provided more information (Davis 1986). In the approach used here, principal components analysis (PCA), the eigenvectors of the variance-covariance or correlation matrix of the data, which are by definition orthogonal, become the new reference axes. The new axes are arranged in order of decreasing amount of variance subsumed, so crucial discontinuities or geographic trends in the data tend to appear on the first few principal components, the rest being attributable to "noise." PCA, in contrast to hierarchical cluster analysis, is highly sensitive to the "shape" defined by patterns of interelement correlation: the components themselves are defined as the axes along which the dataset as a whole is elongated; furthermore, subgroups of the data with distinctive patterns of interelement correlation will be elongated in comparison to the dataset as a whole, thus underscoring their coherence as a group. In the present case, analysis was based on a 28-element variance-covariance matrix of all available elements except Ni, Ca, Ba, Mn, and Sr, using all 70 archaeological and five raw material samples.

Whether a hypothetical compositional group represents a raw material "source" should be judged in light of information on the relative density of data points in different regions of the compositional space. In the present case, one means of group evaluation was to plot probability ellipses representing a constant (two dimensional) Mahalanobis distance from a group centroid on the various principal component plots. Mahalanobis or generalized distance is a multivariate extension of the univariate standardized distance (z-score). Like a z-score, Mahalanobis distance takes into account not just a point's Euclidean distance from a group centroid but also the dispersion of data points out from the centroid in the direction of the point of interest.

P-values, which indicate the probability of observing the same or a larger Mahalanobis distance, are easily obtained. In the present study, these p-values are based on the 28 elements retained for analysis. With a fairly small specimen-to-variate ratio, as in this case, individual points have a magnified effect on group shape, thus inflating p-values derived from Mahalanobis distance. To ameliorate this stretchability problem, we omitted each individual reference group member from the group before calculating its own Mahalanobis distance and associated p-value.

The Thames group is the largest compositional group identified in the study and includes 47 members (62% of total sample). Six sites are represented in the group (1-04, 61-49, 72-91, 75-06, 86-02, and RI 118) (Table 4). These sites are largely restricted to the area between the Connecticut and Thames Rivers (Figure 4). Although Hackney Pond and Shantok stylistic types are both present in the group, it is dominated by the Shantok type (74% of group members). One group member from Site RI 118 (Block Island) may represent a movement of pottery between the mainland and the islands of Long Island Sound.

The Thames group is the most well defined group recognized in this analysis. The tight clustering of data points shown in Figures 5-7 may reflect a limited number of clay sources from which residents of Fort Shantok (Site 86-02) obtained their clay. Alternatively, collection of similar clays at several locations coupled with uniform paste preparation technology may contribute to the group profile. One possible interpretation for compositional homogeneity may be the increased focus of settlement around palisaded villages during the mid- to late- 17th-century. Ethnohistoric research has identified Fort Shantok as a Mohegan site and Fort Mystic as a Pequot site (Figure 4). Ceci (1990) has suggested that fully sedentary villages were a post-contact phenomenon that represent an economic adjustment to trade with Europeans. This may account for the limited spatial distribution of both the Shantok Tradition and the Thames compositional group.

The Mystic compositional group includes 13 members and sherds from only four sites (59-19 [Fort Mystic], 86-02 [Fort Shantok], 72-31, and 72-200). Three of these specimens are shell tempered and the remainder have no temper. The ceramics from Site 72-31 dominate the Mystic reference group (76% of the group) and include both Niantic Stamped and Hackney Pond types. This may suggest an association with raw material resources lying east of the Thames River. This area includes both historic and modern day Pequot territory. Only one example of the Shantok Castellated type (Sample JML024) was included in the Mystic compositional group (Table 4 and Figure 4).

The Thames and Mystic compositional groups are well separated from each other in multivariate space. The separation between the two groups shown on the PCA plot (Figure 3) is attributable largely to differing rare earth element concentrations; as shown in Figure 5, rare earth element concentrations are consistently lower in the Mystic reference group. Moreover, the two groups define different axes of correlation on the rare earth element plot (Figure 5), suggesting that raw material differences rather than just paste preparation differences or proportional variation within a single clay deposit separate the two groups.

A collection of five mineral tempered sherds and one untempered sherd share a broadly similar compositional profile that is distinct from that of the Thames and Mystic reference groups (Figures 3 and 5-7). The Mineral Temper group includes sherds from four sites (61-49, 72-31, 75-06, and 86-02 [Fort Shantok]). The small number of specimens in this proto-group precludes a more detailed discussion of its typological characteristics. However, it is important to note that all six specimens were associated with assemblages dominated by shell temper. Considering the wide distribution of proveniences in the Mineral Temper group and its chemical variability, this group may represent use of similar temper materials or clays with similar non-plastic minerals which are common at a number of locations across southern New England.

Another proto-group, listed under Inland Composition (Table 4) includes two archaeological specimens (both from Site 75-06), and raw clay samples from two interior locations (71-10 and 72-C1). These specimens exhibit low concentrations of iron and other transition metals along with a high concentration of Zr and Hf (Figures 6 and 7). Enrichment of these two elements may be attributable to higher quantities of silt-size grains in the paste. Enrichment of Hf and Zr might also reflect derivation from older sediments. These specimens are also distinguished by enrichment of heavy rare earths, such as lutetium, compared to the Thames or Mystic reference groups (Figure 5). The two archaeological specimens in this proto-group are body sherds associated with a Shantok Castellated pot recovered from Site 75-06; the fact that they closely match compositional signatures of natural clay sediments found at Site 71-10, an interior site, may indicate some movement of ceramic materials downstream from the interior to estuarine zones. The vast majority of other specimens from Site 75-06 are of Thames composition (Table 4).

Two chemically similar specimens (JML034 and JML037) suggest the possibility that broader sampling would identify a compositional profile distinctive to the islands of Long Island Sound (these are listed under "Island Composition" in Table 4). This proto-group includes raw clay from Fisher's Island, New York and an archaeological specimen from site RI 118, located on Block Island, Rhode Island. The sherd from Block Island is classified as an example of the Shantok Castellated type based on optical and morphological attributes. It is of interest to note that two other sherds submitted for analysis from site RI-118 included one member of the Thames reference group and one unassigned specimen. While three sherds cannot be used to address vessel and population movements in Long Island Sound, the compositional diversity in the very small sample from this site may suggest consumption of pots from several points of origin.

The meager compositional evidence from Block Island accords with some suppositions regarding the early Contact Period history of the island. Site RI-118 is a documented fortified village occupied by a group referred to as the Manissee. In the early to mid-17th- century Block Island was an important trading center for Native Americans who may have traveled from both coastal Connecticut and Long Island to engage in trade and shell procurement associated with wampum production. Ethnohistoric sources also indicate that the Block Island Indians were engaged in economic trade with the Mohegans, Pequots, and Narragansetts at different times during the 17th-century (McBride 1990). It is tantalizing to find such a diversity of compositions represented at this site of known commercial importance.

Unassigned specimens in the analyzed sample include the sherd from Site RI 118 already mentioned (which may represent an imported vessel from Long Island or some part of southern New England not yet sampled), along with one raw clay sample from Windsor, Connecticut. Windsor is located on the Connecticut River in north-central Connecticut, approximately 45-kilometers north of the study region and may lie in a clay resource zone that was not accessed by people living in southeastern Connecticut.

Assuming that the historical framework illustrated in Figure 1 is basically correct, the compositional results suggest that divergence of the Windsor Tradition into Hackney Pond and Shantok sub-traditions was accompanied by differentiation in ceramic resource procurement practices. The Mystic compositional group, which contains most of the Niantic type specimens, is tentatively linked with resources lying east of the Thames River (Figure 4). After divergence of the two daughter types, the clay resources of this zone continued to be used in both Hackney Pond and Shantok types. Shantok decorative practices were followed in several zones, as indicated by the fact that Shantok type specimens in this study occur in both main compositional groups and two of the three proto-groups (Table 4). However, the main focus of Shantok production shifted west of the Thames River, and may have been largely restricted to the vicinity of Fort Shantok, which was occupied by the Mohegan tribe during the 17th-century. In contrast, the Hackney Pond type is represented in both main compositional groups and one of the proto-groups (Table 4), which suggests that vessel forming and decorative practices associated with this type were also disseminated widely. The seventeen Hackney Pond specimens in this study include shell, mineral, and non-tempered variants, a greater range of tempers than in the Shantok type. There is no evidence that the Hackney Pond decorative type is concentrated in one restricted geographic zone, as is the case with the Shantok vessels made in the vicinity of Fort Shantok. Thus, the tendency toward reliance on specific, restricted resources by makers of Shantok vessels was not matched among the makers of Hackney Pond vessels.

Complementary ethnohistoric and archaeological data suggest that at or near the time of contact with Europeans (circa A.D. 1614) the Mohegans distanced themselves as an ethnic and political group from a population that shared a common ancestry. This group became known as the Pequots, and they were located in southeastern Connecticut. Economic competition for European trade goods as well as social and political conflicts between Mohegans and Pequots in the early- to mid- 17th-century may have limited the transmission of technological information as tribal boundaries formed. This social distancing is marked historically by warfare between the Mohegans and Pequots (A.D. 1637), and later the Mohegans and Narragansetts (A.D. 1643) (De Forest 1852; Williams 1972; McBride 1990). Changes in economic, social, and political dynamics between Native American groups at the time of European Contact may partially explain the morphological and compositional changes in ceramics.

Together, the ethnohistoric, archaeological, and compositional data suggest that the divergence of Shantok from Hackney Pond encompassed diversification of both decorative practices and resource procurement patterns. Initially, biased sampling of the Fort Shantok assemblage may have produced a sort of bottleneck effect, as a result of which the ceramics produced by the smaller group (Mohegans) appear to diverge from the classic Windsor forms and at the same time show an overall reduction in clay resources. The compositional profiles of specimens from the Fort Shantok assemblage appear to be made from a more restricted range of raw materials than the Hackney Pond ceramics found at contemporary Pequot sites. Reduced contact with surrounding Native American groups in the early to mid- 17th-century may have accentuated the divergence by chance processes alone (drift). Selection may also have been involved to the extent that increased sedentism among the Mohegans favored reliance on a restricted range of ceramic resources, whereas continued mobility among the Pequots (following the Pequot War of 1637) permitted use of a wider range of materials.

In sum, the compositional diversity and subgrouping tendencies identified in this study may be explained in part by changes in economics, settlement, and social organization during the early- to mid- 17th-century in southeastern Connecticut. The identification of compositional groups in the sample population of Windsor and Shantok Tradition ceramics may suggest that village formation and competition in the Contact Period had the effect of limiting access to clay resource zones for at least one group (Mohegans). In this case, the Thames compositional group is linked with the historic territory of the Mohegan Tribe (west side of Thames River) and the Mystic group is linked with Pequot territory (east side of Thames River). To both the historic and present-day groups the Thames River serves as a recognized boundary (Figure 4). These changes in social and economic organization during the contact period are also supported by decreased proportions of nonlocal lithic materials coming into southern New England (Feder 1984).

The results of this study indicate that ceramic sourcing provides a means to monitor intergroup contact, trade, and settlement during the Late Woodland and contact periods. While much effort has been directed towards reconstructing the culture histories of the 17th-century Mohegan, Pequot, Niantic, and Narragansett tribes, little effort has been directed towards identifying when these groups formed and how these proto-historic groups were distributed with regard to different resource zones. Neutron activation analysis provides one method for identifying patterns in archaeological data that cannot be determined using less precise optical mineralogical or stylistic comparisons.

Acknowledgements. Neutron activation analysis was conducted at the University of Missouri Research Reactor, located in Columbia, Missouri. The National Science Foundation (BNS 91-02016) and Department of Energy (FG07-80ER10725) provided support for equipment, labor, and irradiations. The authors greatly appreciate the assistance from Cynthia Lewis Hays in the laboratory phase of the analysis. Robert Dewar, Michael Graves, Janet Walker, and three anonymous reviewers offered comments that greatly enhanced the quality of this paper. Kevin McBride and the Mashantucket-Pequot Tribe provided several samples used in this study. Nicholas Bellantoni, Connecticut State Archaeologist, provided the Fort Shantok samples from the collections of the Connecticut State Museum of Natural History. Tara L. Prindle provided the vessel illustrations used in Figure 1.

Bishop, R. L. and H. Neff
1989 Compositional data analysis in archaeology. In Archaeological Chemistry IV, edited by R. O. Allen, pp. 576-586. Advances in Chemistry Series 220, American Chemical Society, Washington, D.C.

Byers, D. S. and Irving Rouse
1960 A Re-examination of Guida Farm. Bulletin of the Archaeological Society of Connecticut 30:3-43.

Ceci, Lynn
1990 Native Wampum as a Peripheral Resource in the Seventeenth-Century World System. In The Pequots of southern New England, edited by L. M. Hauptman and J. D. Wherry. University of Oklahoma Press, Norman.

Crepeau, Robert and George Kennedy
1990 Neutron Activation Analysis of Saint Lawrence Iroquoian Pottery. Man in the Northeast 40: 65-74.

Davis, J. C.
1986 Statistics and Data Analysis in Geology. John Willey and Sons, New York.

De Forest, John W.
1852 History of the Indians of Connecticut. Wm. Jas. Hamersley, Hartford.

Feder, Kenneth L.
1984 Pots, Plants, and People: The Late Woodland Period of Connecticut. Bulletin of the Archaeological Society of Connecticut 47: 99-112.

Glascock, M. D.
1992 Characterization of Archaeological Ceramics at MURR by Neutron Activation Analysis and Multivariate Statistics. In Chemical Characterization of Ceramic Pastes in Archaeology, edited by H. Neff, pp. 11-26. Prehistory Press, Madison.

Harbottle, G.
1976 Activation analysis in archaeology. Radiochemistry 3:33-72. The Chemical Society, London.

Lavin, Lucianne
1980 Analysis of Ceramic Vessels from the Ben Hollister Site, Glastonbury, Connecticut. Bulletin of the Archaeological Society of Connecticut 43:3-41.

1986 Pottery classification and cultural models in southern New England prehistory. North American Archaeologist 7(1):1-13.

Lizee, J. M.
1989 Niantic, Hackney Pond and Shantok: A reconsideration of the Late Woodland and Contact Period Ceramic typology for southern New England. Paper presented at the 1989 Annual Meeting, Northeastern Anthropological Association, Montreal.

1994 Prehistoric Ceramic Sequences and Patterning in southern New England: The Windsor Tradition. Unpublished Ph.D. dissertation, Department of Anthropology, University of Connecticut, Storrs.

McBride, Kevin A.
1984 Prehistory of the Lower Connecticut River Valley. Ph.D. dissertation, University of Connecticut, Storrs, University Microfilms, Ann Arbor.

1990 The Historical Archaeology of the Mashantucket Pequots, 1637-1900. In The Pequots of southern New England, edited by L.M. Hauptman and J.D. Wherry, pp. 96-116. University of Oklahoma Press, Norman.

Neff, H. and R. L. Bishop
1988 Plumbate origins and development. American Antiquity 53(3):505-522.

Philpotts, Anthony R. and Nancy Wilson
1994 Application of Petrofabric and Phase Equilibria Analysis to the Study of a Potsherd. Journal of Archaeological Science 21: 607-618.

Ritchie, W. A. and R. S. MacNeish
1949 The Pre-Iroquoian pottery of New York State. American Antiquity 15: 97- 124.

Rouse, Irving
1947 Ceramic traditions and sequences in Connecticut. Bulletin of the Archaeological Society of Connecticut 21: 10-25.

Steponaitis, V. P. and M. J. Blackman
1981 Chemical characterization of Mississippian pottery. Paper presented at the 38th Annual Meeting of the Southeastern Archaeological Conference, Asheville, North Carolina.

Trigger, B. G., L. Yaffe, M. Diksic, J. L. Galinier, H. Marshall, and J. F. Pendergast
1980 Trace Element Analysis of Iroquoian Pottery. Canadian Journal of Archaeology 4: 119-145.

Williams, L. E.
1972 Fort Shantok and Fort Corchaug: A Comparative Study of Seventeenth Century Culture Contact in the Long Island. Doctoral dissertation, New York University, New York.

Figure 1: Illustrated chronology of Niantic, Hackney Pond, and Shantok ceramic types.

Figure 2: Map of New England showing location of study region and sites from which specimens were selected for neutron activation analysis.

Figure 3: Plot of principal components 1 and 3 of the 75-specimen data set. Ellipses represent 90% confidence level for membership in the four groups. Symbols differentiate the compositional groups.

Figure 4: Distributions of sites included in the Mystic and Thames reference groups. Symbols differentiate the compositional groups.

Figure 5: Plot of lanthanum and lutetium base 10 log concentrations in dataset. Ellipses represent 90% confidence level for membership in the three groups. Symbols differentiate the compositional groups.

Figure 6: Plot of uranium and iron base 10 log concentrations in dataset. Ellipses represent 90% confidence level for membership in the three groups. Symbols differentiate the compositional groups.

Figure 7: Plot of hafnium and zirconium base 10 log concentrations in dataset. Ellipses represent 90% confidence level for membership in the three groups. Symbols differentiate the compositional groups.

TABLE 1: Ceramic traditions and assemblages.

TABLE 2: Principal components of variance-covariance matrix of total New England data set.

TABLE 3: Eigenvectors and factor loadings for the first ten principal components derived from the PCA summarized in Table 2.

TABLE 4: Compositional group assignments and % p-values for each specimen compared to the Thames reference group.