MIDWEST PROPERTY TYPES
The Earliest Americans theme study employs a single system for the classification, identification and description of Paleoindian property types (Anderson 2000). According to the Secretary of the Interior's Standards, "a property type is a grouping of individual properties based on shared physical or associative characteristics [that] link the ideas incorporated in the theoretical historic context with actual historic properties that illustrate these ideas." Property types in this study include isolated finds (ineligible for NHL listing), caches, bone beds and kill sites, human burials, rock art and other petroglyphs or pictographs, quarries and workshops, and occupations.
For purposes of NHL recording, documentation, and management, property types are grouped into two major classes, sites and districts. Sites are deposits containing artifacts, features, or biocultural evidence associated with one or more Paleoindian components. As such, the term retains its traditional archaeological usage, albeit here with the inclusion of isolated finds as a particular subset. Districts are defined as multiple contiguous or discontiguous deposits associated with one or more Paleoindian components. They are groupings of sites or components.
Identifying Property Types for the Midwest
Organic preservation is quite poor generally in the Midwest. With few exceptions as noted above, the Paleoindian record is a record of stone tools and features. Stone tools occur in sets that we identify as assemblages and ordinarily treat as registers of single or few brief occupations to particular places. Stone-tool assemblages are described by their size and composition (Shott in press); analysis is the measurement and interpretation of assemblage size and composition.
The traditional archaeological view is that assemblages are categorical things, whether ethnic calling cards or functional toolkits, assemblage variation a simple thing. Assemblage analysis in this tradition typically involves simple inspection of tool frequencies because patterns of variation and the formation processes underlying them are equally simple, variation is categorical, and assemblages fall neatly into types. Although the number of and meaning assigned to types differ, most studies identify at least quarries where raw material was obtained and reduced toward finished tools, residences where many Paleoindians performed many tool-using activities, and hunting camps where smaller groups of men sought and killed game. Although not recorded, there may especially be where women or men used stone tools for plant procurement or process. There is no doubt that quarries existed nor that Paleoindians killed animals and established homes, and they did these things at different places in the landscapes. The question instead is how these activities accumulated an archaeological record over long periods. The traditional view assumes no reoccupation of places, or at least reoccupation for purposes different from the original, and a simple and direct relationship between tools and activities. It creates a Paleoindian past that consists of neat ethnographic tableaux. Thus, traditional archaeological understanding of assemblages is essentialist, regarding them as fundamental kinds of entities that represent entire cultures or discrete parts of them.
But processes that governed assemblage variation are the complex product of interactions between activity and formation processes. Places could have been occupied at various times by groups of various size and composition for various purposes. Tool types varied in their use lives, mapping relations and consequent patterns of covariation with one another. The tools that we find together and take to represent an assemblage may or may not have been used and discarded in the course of a single occupation of a place. However, their present propinquity may result from their having been associated initially complex, overlapping deposits of some number of brief occupations undertaken in approximately the same location for various purposes and at unknown times. Great Lakes Paleoindian assemblages exhibit continuous variation between assemblage "types" that is consistent with time-averaging (Shott 1997; cf. Muller 1999). There is nearly as much variation within assemblage types as between them. However Shott (1997) tested the validity of assemblage types was tested, it was found wanting. Assemblage variation is real and informative but does not correspond neatly to types. However, if we assume a priori that assemblage types exist, and if we use analytical methods commensurate with that assumption, we always will find types.
Isolated finds are individual artifacts that have been demonstrated to be Paleoindian in age "through typological or other analyses" (Anderson 2000:5). Midwestern state journals and newsletters are filled with accounts of isolated Paleoindian finds. Back issues of Wisconsin Archaeologist are a particular source of such reports. Their number and distribution certainly inform about the incidence and distribution of tool-using activity across the Paleoindian landscape. No doubt some fluted discoveries are legitimate isolated artifacts, lost in hunting or by some other accident. Such discoveries form a regional background distribution to the better known elements of the record that we know as sites. By definition, individual artifacts are the units of observation and description in the case of isolated finds. Thus, isolated finds are the sole exception to the received view that the Paleoindian record occurs in the form of discrete sites.
Yet some, perhaps many, isolated finds are the tips (no pun intended) of larger icebergs (Bintliff et al. 1999). Often locations are revisited, other Paleoindian tools are found, isolated finds growing into assemblages qua sites.
Caches are special deposits where tools and other goods were placed deliberately. Collins (1999:147-167) defined votive, ritual and utilitarian classes of caches. Among the latter, he (1999:176) distinguished seasonal from insurance caches, both useful where people were scattered thinly across unstable habitats. Obviously, purely food caches are difficult if not impossible to recover archaeologically, since food long since has perished. Fisher's (1987) Great Lakes mastodont meat caches are possible but not yet confirmed. Otherwise, there is no Midwestern evidence for Paleoindian food caches.
Relatively few in number, highly concentrated so as to occupy small spaces, and deliberately concealed by burial, caches pose acute sampling problems, practically defying discovery via systematic survey. Indeed, no Paleoindian caches anywhere were discovered other than accidentally, either when dislodged by plowing or construction, or in the course of excavation of larger sites.
Therefore, few Paleoindian caches are known in the Midwest. Excluding possible caches of nondiagnostic tools noted above, most relevant assemblages are from Ontario (Deller and Ellis 1984; Deller and Ellis 1992:26,97; Storck and Tomenchuk 1990). In the Midwest, Rummels-Maske (Anderson and Tiffany 1972), a Clovis or Gainey-affinity in eastern Iowa, is the sole apparently major early or middle Paleoindian cache. Significantly, it was not found by archaeologists and is known only because it was reported at once. No doubt other, perhaps many, caches have been discovered but not reported, so we cannot know how unusual Rummels-Maske is. Rummels-Maske appears to be a utilitarian cache (sensu Collins 1999:175) of fluted bifaces or fragments. Any organic objects that may have existed decomposed before discovery. Although the size and composition of cache assemblages varies considerably across North America (see above), Rummels-Maske is not exceptional in the kind or number of its contituents.
Among late Paleoindian industries, Dalton caches, comparatively abundant in the southeast, extend into southern Missouri and southwestern Illinois. The American Bottom's Nochta yielded four Dalton adze blanks in an apparent cache (Higgins 1990:43, Plate 9). Nearby Jens's Feature 30 contained one Dalton point, one adze, six end scrapers and two other unifaces (Walthall and Holley 1997:156-157). In far southern Illinois, Olive Branch apparently produced two caches; one was not reported in detail but the other contained an adze blank, seven biface preforms and one retouched flake (Gramly 1995:42, Fig. 19). Other Dalton caches have been found in southern Illinois but remain sketchily documented (Walthall and Koldehoff 1998:262-263). Chapman (1975:102) reported a cache of nine Dalton bifaces from St. Francois County, southeast of St. Louis.
Bone Beds/Kill Sites
The only documented Midwestern Paleoindian bone bed is Missouri's Kimmswick (Graham et al. 1981) unless we accept the possible evidence of proboscidean butchery from Michigan (Fisher 1987) and the Chesrow Complex of southeastern Wisconsin (Overstreet 1993). Only Holcombe (Fitting et al. 1966) among well-associated Midwestern Paleoindian sites has yielded bone of any kind, and that calcined cervid (probably caribou) bone (Cleland 1966) from a single feature. Ontario's Gainey Phase Udora (Storck and Spiess 1994) and several northeastern sites also have yielded animal bones in small quantities. Whatever big-game orientation Paleoindians may have maintained, in the faunal record, fish and small game are as common as caribou.
Kill sites are assemblages that are dominated by fluted bifaces and channel flakes, and/or that lie above or near strand lines, valley bluffs or other locations thought to be especially suitable for spying and ambushing game; places where Paleoindians killed and butchered animals and left behind tools and debris. Deller and Ellis (2000:150-187) argued at length for Parkhill's Areas B and C as kill sites, and Fisher (Storck 1997) also is interpreted as a kill site. In the Midwestern United States, Barnes (Voss 1977), and Holcombe also suggest use as kill sites.
In a material record impoverished by organic decay, there is no definite evidence of early and middle Paleoindian human remains anywhere in the Midwest. Caches may sometimes consist of offerings placed with human remains; Ontario's late Paleoindian Crowfield site (Deller and Ellis 1984) is the best such example from near the Midwest. As above, possible cremations of late Paleoindian Eden-Scottsbluff affinity were found in northern Wisconsin (Mason and Mason 1960; Ritzenthaler 1972) and upper Michigan (Buckmaster and Paquette 1988).
Petroglyphs and other forms of rock art are not uncommon in the Midwest, although all seem much more recent than the Paleoindian period to judge from settings, motifs, styles and subjects, and cultural associations (e.g., Lothson 1976; Swauger 1984; Zurel 1999). There is no reason to doubt that Midwestern Paleoindians executed rock art. Alas, rock art is ravaged by exposure and weathering and art the age of Paleoindian cultures will preserve only under special conditions not yet found in the Midwest. Taphonomic studies suggest that much prehistoric rock art is destroyed by natural agents that act over time (Bednarik 1994). Typical is Swauger's assumption that Midwestern rock art older than 3,000 years would have eroded away by now (Swauger 1984:267). This unsupported dismissal of potential data is all too characteristic of Paleoindian studies. In fact, any Midwestern site with the potential for yielding evidence of rock art, either still in place or in a datable secondary deposit would be of national significance. If and when Swauger's assumption is shown incorrect, we may gain wonderful insights into the Paleoindian symbolic world.
Quarries and Workshops
Prominent Midwestern Paleoindian quarry/workshop sites include Welling (Prufer and Wright 1970) at Ohio's Upper Mercer source, Silver Mound (Hill 1994) near western Wisconsin's Hixton source, and Ready (J. Morrow 1996) near Burlington sources at the confluence of the Illinois and Mississippi Rivers. Major Paleoindian sites that probably include quarry/workshop components also occur near Ontario chert sources (Storck 1997). Like the Welling site, Honey Run (PiSunyer et al 1967) and McConnel (Prufer 1963) are two of the many late Paleoindian workshops that lie near the well-surveyed Upper Mercer outcrops in Coshocton County, Ohio. This makes it likely that poor sampling is responsible for the failure to identify major Paleoindian workshops near Hornstone sources or Attica chert sources in southern and central Indiana, respectively, or near the Bayport chert source near Saginaw Bay in Michigan, or near the Flint Ridge sources in central Ohio (Tankersley 1990) although local Middle Woodland mining of the latter outcrops, along with their prehistoric projectile point collecting may well have obscured such evidence.
To varying degrees but always considerably, Midwestern quarries/workshops have been analyzed usefully (Prufer and Wright 1970; J. Morrow 1996; Koldehoff 1983; Hill 1994). Yet more synthetic work would be more valuable still by comparing and contrasting quarry/workshop assemblages for their nature and scale of tool production. We may find that similar reduction practices characterized all quarries/workshops or that reduction changed throught time or varied with natural properties like fracture mechanics or the size and form of cobbles. Certainly, experiments (Bradbury and Franklin 2000) suggest that the latter properties greatly influence reduction practices. Such studies at Midwestern Paleoindian quarries/workshops should be supplemented by survey or excavation of natural outcrops to sample the range of variation in original cobble size and form. Welling, Honey Run and McConnel are just a few workshops among the many that probably exist at the Upper Mercer source area (e.g., Lepper 1988). That area is perhaps the best candidate for extensive and detailed study of Paleoindian quarry practices and reduction technology.
In the Midwest, relatively extensive Paleoindian occupations include Gainey (Simons 1997; Simons et al. 1984) and probably Grogitsky (Zurel 1979) in Michigan, Paleo Crossing (Brose 1994) and Nobles Pond (Seeman 1994; Seeman et al. 1994a,b) in Ohio, Big Eddy and Martens in Missouri, and Aebischer (Mason 1988) in Wisconsin. Smaller but significant and documented occupations include Leavitt (Shott 1993) in Michigan, and Bostrom (Tankersley 1995) and CB-North (Evans et al. 1997) in Illinois.
The Midwestern Paleoindian record is an historical document compiled by many types of investigation over a century or more. We must use it to the fullest, by studying individual tools, assemblages and deposits, and regional distributions of assemblages. But we misuse the record if we do not assay both its strengths and weaknesses. Cowgill (1970:163) distinguished between the available material evidence and the collections and observations we make from it. He called the former archaeology's physical-finds population, the latter its physical-finds sample. We have only the latter, which certainly grows in size and may change in character over time, and we must estimate how well the sample represents the population. This is "representativity."
Halén's (1994:29-35; see also Kristiansen 1985) representativity model included numerical and essentially categorical components, identified various dimensions of representation of the material record in archaeological data, and proposed measures of them. Dimensions include time, space, degree of preservation, classification effects on recognition and abundance of legitimate categories, kind and amount of data acquired previously, and incidental and deliberate modern destruction. Halén's numerical representativity signifies what fraction or percentage of the population the available sample comprises. Sample representativeness includes fraction (How big is the sample?) and composition (How typical is the sample in proportional abundance?). If sample fraction is difficult to estimate, its composition seems more difficult still. The high esteem in which collectors hold fluted bifaces and the heavy contributions that they have made to the recorded sample suggests that those artifacts are represented disproportionately to their abundance in the original physical-finds population. But it remains very difficult to estimate how disproportionate is the overrepresentation of fluted bifaces.
Yet even the recorded physical-finds sample is a social product. More than a century ago, the British prehistorian Daniel Wilson described then-extant stone tool collections in the Ohio Valley. An unknown but probably small fraction of the collections were of Paleoindian provenance, but Wilson's account is at once illuminating and sobering. Compared to Europe, Wilson (1876 I:56) considered "the abundance of flint and stone implements in the virgin soil of the New World...almost marvelous." Visiting the Flint Ridge quarries in Ohio, he amassed a sizable collection within two hours (ibid:8). Wilson inspected many private collections from near Cincinnati, one of whose "enthusiastic" owners reported finding as many as 70 stone tools in a day's collecting (ibid:79). Wilson also met dealers in antiquities, proving if nothing else that the market scourge beset North American archaeology from its beginnings. He observed ruefully of a specimen purchased from a dealer that "information on the locality and the circumstances attendant to its discovery could not be obtained" (ibid:60), a conclusion too familiar to modern archaeologists and understandable to them if not to Wilson; the antiquities trade destroys, not preserves, information. Even much more recently, Springer found it necessary to hurry to the field each spring "to precede the activities of collectors" (1985:10-12).
We know neither the size nor the current location of Wilson's collections, so cannot determine how many Paleoindian specimens it contains. A Paris museum holds at least 16 fluted bifaces, most from the Midwest, among a large collection of North American artifacts amassed in the first half of the twentieth century (Smith 1961). Documentation is poor, but at least some specimens were purchased from dealers. The British Museum holds at least seven fluted bifaces from Midwestern states (N. Ashton, personal communication 2000) in its substantial North American artifact collections.
One Illinois case suggests how abundant the Paleoindian record once was (Munson and Tankerley 1991). Thomas Kiley of DeWitt County in central Illinois became interested in archaeology in 1909. He began compiling records of artifacts found there, developing a special affinity for "grooved" points. His descriptions make it clear that these were fluted bifaces which, by Kiley's account, comprised about 1% of all bifaces found. But 1% of many bifaces is a large number. Over the next 50 years, Kiley faithfully recorded information that came to him (Who knows how many fluted bifaces found in DeWitt County escaped his notice?) when fluted bifaces were discovered in fields, in streambanks, in building excavations. One, for instance, was found "three feet deep in clay" (Munson and Tankersley 1991:4). Many of the finds were incidental discoveries by local residents, but Kiley reported that serious collectors traveled to the area in search of specimens. Kiley's efforts indeed were "the oldest and longest fluted point survey" (ibid:6) on record, and they documented an astonishing 332 fluted bifaces in this one county.
The implications of such data are clear for settlement interpretations, since DeWitt County would not by any current understanding be considered prime Paleoindian habitat. Another implication is that the closer or longer we look, the more we find. Kiley's records, after all, were compiled over 50 years, a period much longer than archaeologists can devote to survey areas. About 330 bifaces recorded in 50 years gives an annual yield of 6.6 bifaces. In three typical years, the yield would reach about 20 bifaces, respectable but not remarkable. Thus, the number is a product of the length of study as much as any other factor. The unsuspected number of fluted bifaces in one obscure Illinois county suggests that the patterns we see of both abundance and distribution of Paleoindian "sites" are not well controlled for sample error and other biases. By all means we should compile and examine those patterns, but always from a "source-critical" (Baudou 1985) perspective. Another example is the apparently marked concentration of Folsom sites in the southwestern corner of Iowa (Billeck 1998; T. Morrow and J. Morrow 1994). Perhaps the documented distribution faithfully reflects the underlying archaeological one, but Billeck (1998:401) attributed it at least in part to the sustained, systematic work there of Paul Rowe over forty years.
The character of the record includes not just its size but also its distribution and contents. The record of known sites may be biased by systematic factors (Lepper 1983b) as well as chance. In Michigan, Barnes was found because it lay about 100 m from Wallace Hill's house. Leavitt was found because the Leavitt family collected their fields for years, and Donna Leavitt Sanford later joined the Michigan Archaeological Society and reported the site. Gainey and Butler were found because they were about 2 km from D.B. Simons' house, and he could revisit them repeatedly. Holcombe was found because Jerry DeVisscher lived nearby and reported because he knew Wayne State University archaeologists. Local resident D. Wymer found several Gainey sites in south-central Michigan. In Wisconsin, the large Silver Mound collection took Gary Steele 20 years to accumulate (Hill 1994:224). Most fluted bifaces by far near St. Louis were found by private collectors, and many of their collections already had been dispersed, the information they contained lost, by the mid-20th century (Smail 1951:11).
By most recent count, Illinois and Missouri together have yielded about 500 fluted bifaces (Anderson and Faught 2000). Yet half a century ago, Smail (1951) documented more than 500, just from the St. Louis environs alone, and much of the modern total probably includes bifaces from major sites (e.g., J. Morrow 1996) found since Smail's survey. Thus, much of Smail's evidence is lost to modern study. None of these sites was found in systematic survey. Absent such survey over large areas, we simply never will know the true distribution and abundance of Paleoindian remains across the landscape. Collectors have made fine contributions to the accumulated record, but not even their efforts are a substitute for the systematic, large-scale work needed.
Paleoindian tools and sites are widely distributed across the Midwest, and are or were more abundant there than often supposed. Apart from these valid generalizations we have much to learn about the validity of accumulated patterns and the nature of original ones. The Paleoindian archaeological record patterns according to original physical-finds population distribution among other factors. The latter include the modern taphonomic agents noted above, as well as natural agents like slope erosion and alluviation.
"Sites." "Site" is an inference or construction, not an empirical property of the archaeological record (e.g., Shott 1995b). For purposes of this study, I take sites as reported at face value. One way to estimate the number of Paleoindian sites (the physical-finds population at the site level) is from available published sources. Peru's (1967:7) map of Allegan and Kent Counties, Michigan was unscaled, so that source was omitted. Table 19 compiles figures on Paleoindian sites in several reported survey areas across the Midwest.
Table 19. Paleoindian Site Density Estimates from Survey Data in the Midwest.[Long description]
Mallam (1971) reported seven or nine Paleoindian sites depending on how locations are aggregated as sites. Area surveyed was taken as the entire Upper Iowa River watershed. Similar areas were mapped by Peru (1967:Fig. 1), Wendt (1985:Fig. 1), Smith (1990:Fig. 7), Lepper (1988:Fig. 1) and Harrison et al. (1977:Fig. 16), and reported by Munson and Downs (1968:122). It would be surprising if as much as half of each area was surveyed (e.g., Jackson 1998:4), but in the absence of precise figures mapped area was treated as surveyed area. Site figures are minimum values as well. Several sources reported only sites that yielded fluted bifaces, but Harrison et al. (1977) and Wendt (1985) defined "Paleoindian" more liberally. Only fluted bifaces and sites from which they derive were used from those sources, as were Munson and Downs's (1968) Paleoindian but not Plano figures. Coshocton County, Ohio sites were tabulated from Lepper (1988:Fig. 3). Henry and Nichols's (1963:Fig. 52) eastern Illinois survey area is omitted from uncertainty over its size, but it shows at least four and perhaps seven fluted-biface sites in an area that measures about 6 km2. Jonathan Bowen (personal communication, 1988) amassed considerable information on Paleoindian and Early Archaic artifacts and sites in northern Ohio, but those data could not be included.
As much as anything, Table 19 suggests how rare are probabilistic data on Paleoindian site frequency, available in sources consulted only for southeastern Michigan (Peebles and Shott 1981) and southeastern Wisconsin (Goldstein 1987). The received wisdom is that Paleoindian sites are rare, may owe to sample effects or taphonomy in Plains studies (Sheehan 1995). Table 19 shows that site density is higher in both systematic surveys reported than in any selective one. Partly this may owe to uncertainty in areas surveyed in sources that report selective survey, as above.
Whatever the case, in selective surveys mean Midwestern Paleoindian site density=0.0149/km2 (s.d.=0.0140). Little of eastern North America has been intensively surveyed for Paleoindian sites, so comparative data are few. Yet McAvoy's (1992:12, 23) survey in southern Virginia found 23 Paleoindian sites in 780 km2, yielding a density of 0.0295 sites/km2. This figure is higher than all but one Midwestern value reported in Table 19, and nearly twice the Midwestern mean. It might be discounted because McAvoy's survey area deliberately included the environs of the famous Williamson site, known to be thick with Paleoindian workshops. Yet Lepper's Coshocton County study area is similarly situated near Upper Mercer quarries (e.g., Welling [Prufer and Wright 1970]). Many Nottoway sites apparently were not quarries or workshops and McAvoy may have surveyed more of his area than did most Midwestern sources. In probabilistic surveys, mean site density=0.0693/km2 (s.d.=0.0653). For both selective and systematic survey, the standard deviation nearly equals the mean, so parameter estimates are very imprecise.
Even McAvoy's figures may underrepresent Paleoindian site density. In Ontario, Jackson (1998:4-5) found 24 Paleoindian sites in 83 km2. The resulting density is 0.2893 sites/km2, more than twice the nearest figure and nearly an order of magnitude greater than results obtained in most selective surveys. Yet Jackson could survey only 25% of this area (still a substantial area). Similarly, sustained work by Ellis, Deller and colleagues near the south shore of Lake Huron in Ontario (Ellis and Deller's 2000 Figure 1.2) shows 17 Paleoindian sites in an area 20 km in diameter. There is no reason not to suppose the that the Midwestern record is equally abundant.
Table 19 data are subject to revision as fieldwork continues. Harrison et al.'s Kishwaukee Basin survey area in northern Illinois is especially instructive in this respect. A subsequent report there showed that much less than the entire area was surveyed (Springer 1985:Fig. 3). Moreover, the later report included 24 Paleoindian sites, defining Paleoindian as liberally as did the original authors (ibid: Fig. 4). The combination of a smaller area surveyed and more sites found would increase site density. Finally, Harrison et al.'s survey area subsequently yielded yet another Clovis-Gainey site [Koldehoff 1999:12]. To be conservative, only estimates from Harrison et al. (1977) appear in Table 19.
Since the Midwest's northern margins were icebound or only recently deglaciated during the early and middle Paleoindian periods, the northern one-third of Minnesota and Wisconsin, and all of the Upper Peninsula and the northern one-third of Michigan's Lower Peninsula are omitted. This yields an area for the Midwest of 1,063,702 km2. Estimated mean density from selective surveys yields its own estimate of 1,063,702 x 0.0149±0.0140 sites/km2 = 15,849±14,892 Paleoindian sites. The figure from probabilistic surveys yields an estimate of 1,063,702 x 0.0693±0.0653 sites/km2 = 73,715±69,460 Paleoindian sites. At the upper end, there may be nearly 150,000 Midwestern Paleoindian sites, which seems impossibly high; at the lower end there may be as few as 1,000. Perhaps a reasonable if broad estimate would be a range from the minimum of 2,000 to the mean of about 17,000 from selective surveys.
Indiana, Iowa, Michigan, Ohio and Wisconsin SHPOs responded to queries with figures on Paleoindian sites, and Wiant (1993) for Illinois and Seeman and Prufer (1982) for Ohio supplemented these figures. Combined, areas of these states habitable by Paleoindians comprise about 732,000 km2. From above density estimates, Paleoindian sites over this area might total from about 2,900 to 8,300. Together, these states report 1,676 Paleoindian sites, although Indiana's total of 469 may include many late Paleoindian sites. This total is about three-quarters of the conservative estimate of all Paleoindian sites in those states but barely 20% of the higher estimate. Still, considering the degree to which modern land use has destroyed the Midwestern archaeological record, such fractions of the estimated total are encouragingly large.
Discoveries in Time
Kristiansen (1985:8) gauged representativity via "finds curves," essentially histograms of the number of sites discovered by time interval. Curves showing a past peak followed by a steady decline to the present suggest that the subject is well sampled. But curves that continue to rise to the present suggest that many, perhaps most, subjects remains to be found.
Table 20 shows Paleoindian sites recorded by twentieth century decade for several Midwestern states, compiled from figures kindly provided by SHPOs. (A few Wisconsin records from the 1890s were omitted.) Sources were careful to explain that "Paleoindian" was defined broadly or narrowly at different times in the past century, so some sites may be very late Paleoindian or Early Archaic in age. The Ohio SHPO reported no Paleoindian sites recorded before 1980. For the state where Shetrone (1936) long ago reported 140 fluted bifaces and Prufer and Baby (1963) documented an impressive Paleoindian data base, this seems an artifact of record-keeping.
Table 20. Paleoindian Sites Reported by Decade in Selected Midwestern States
Ohio has by far the most sites, geographically larger Iowa the fewest. There is a perfect rank-correlation between 1990 state population (United States Bureau of the Census 1999) and number of Paleoindian sites (Fig. 1). In area Ohio is the smallest state, in population the largest. Modern population seems to bear on the probability of discovering Paleoindian sites. Excluding Ohio as above, a "finds curve" of these data (Fig. 2) shows very low recording rates until the 1960s with a steeply rising rate through the 1980s. The rise is especially steep between the 1960s and 1970s, when modern preservation law and the enormous survey efforts it inspired took effect. The modest decline in the 1990s suggests either that the Paleoindian site sample is approaching completion or that systematic survey declined then. Whichever is the case, even the 1990s figure is high, suggesting that many Paleoindian sites remain to be found before the upper tail of the distribution is reached.
Table 20 compiles 1,268 of the 1,676 sites reported in these states for which date of recording is known. Probably some records are apocryphal, or of poor quality regarding location, contexts, associations and artifact assemblages. Some, perhaps many, records register classification effects whereby Early Archaic diagnostics are classified as Paleoindian. (This problem may be especially acute in Indiana, where as many as 565 Paleoindian sites of whatever description but only 48 early Paleoindian ones are recorded and date of recording is reported only for 182 (K. Tinkham, personal communication 16 March 2000). Yet a considerable portion of the total probably are early or middle Paleoindian sites, indicating a substantial archaeological record across the Midwest. Well studied and published Midwestern Paleoindian assemblages, by contrast, practically can be counted on two hands. They comprise a tiny fraction of even the recorded total, so Table 20 points out how much work remains to do just with existing records.
Geomorphology and Geological History
Large rivers and their floodplains grace the Midwestern landscape. As much as anything, the Midwest is a land of major rivers. Alas, this quality imposes its archaeological cost, since Paleoindian sites that occur in floodplains may be deeply buried. Big Eddy (Lopinot et al. 1998) is merely the most prominent example of such buried Paleoindian deposits. CB-North in the American Bottom is buried by aeolian and alluvial deposits (Evans et al. 1997). In the Ohio Valley, Manning has Early Archaic and perhaps Paleoindian deposits, to judge from Occupation 2's soil-humate date of 10,240±110 B.P. (B-17774) (Lepper 1994:146). Also in the Ohio Valley, Sandy Springs (Cunningham 1974; Seeman et al. 1994a) is a surface Paleoindian site, but one that lies on a Pleistocene terrace. Geoarchaeological studies of portions of the Mississippi Valley show that much of the Paleoindian surface there is deeply buried or eroded. In one study area in Iowa and Illinois, nearly two-thirds of deposits of Paleoindian age are buried or gone (Bettis and Benn 1989:86). From Billeck's (1998:405) account, Missouri River floodplain surface deposits in western Iowa all postdate Paleoindian occupation. In the Cairo Lowlands just downstream of the Ohio-Mississippi confluence, practically the entire modern surface postdates 3,000 B.P., such that "sites that date earlier than the Late Archaic period are unlikely to be present" (Lafferty 1998:132; see also O'Brien and Wood 1998:58). Thus, the American Bottoms area is rich in sites but surface surveys there revealed not a single Paleoindian assemblage (Munson and Harn 1971:4, 21), yet CB-North shows that they exist. The Mississippi is merely the largest of Midwestern rivers, and has buried or destroyed much of its Paleoindian surface.
Until detailed geological histories are compiled for other floodplains the most conservative assumption is that their Paleoindian surfaces also are essentially destroyed. The absence of Paleoindian sites in surface contexts of geologically active regions says nothing about Paleoindian occupation since the occupational record may have existed but been destroyed. Similarly, wetlands are or were common in the Midwest; ordinarily they are not considered likely places in which to find any, let alone Paleoindian, sites. Yet Clark (1982:118, 121) found several late Paleoindian sites in central Wisconsin wetlands only recently drained at the time of his survey, which suggests that remaining wetlands may contain many Paleoindian sites yet undiscovered.
"Mountainous" may seem an oxymoron in Midwestern context, yet old, small mountain ranges are scattered across it. The mountainous Appalachian uplands are geologically dynamic in ways that can expose but also obscure the Paleoindian physical-finds sample. Among the reasons for the apparent paucity of Paleoindian sites there is geological taphonomy, not just the reluctance of Paleoindians to venture there. In part, the paucity of Paleoindian sites in Appalachia "is likely the result of geological active landscapes" (Tankersley et al. 1996:93). Southern Illinois, southwestern Indiana and western upper Michigan also are mountainous; the number of Paleoindian sites recorded in those areas, even where considerable (Wiant 1993:Fig. 1), may underestimate the true abundance of sites. (McCracken 1986).
Modern land use influences kind and amount of archaeological remains known. Extent of gravel mining and even method of mining influenced the number of Paleolithic artifacts discovered in southern England (Hosfield 1996). The extent and duration of cultivation affected the number and condition of Danish mounds preserved for recording (Baudou 1985). As in Scandinavia, cultivation and other kinds of land use in the Midwest promote discovery of some kinds of remains and damage or destroy others (Halén 1994:30; Kristiansen 1985:8).
The Midwest is an agricultural region and much of its land is under cultivation, yet pockets of covered ground exist in and near cities and extensive tracts of forest and pasture are interspersed through the region. The northern halves of Minnesota, Wisconsin and Michigan mostly are forested, as is much of southern Missouri. The Shawnee National Forest occupies considerable land in southern Illinois. Much covered ground near cities already is altered sufficiently to destroy whatever archaeological remains it may have possessed. But forests and pasture often are unaltered, their archaeological record largely intact; indeed, many such tracts never were plowed so their archaeological deposits are better preserved than in plowed fields.
Yet such covered ground poses special challenges to discovery. Shovel-testing is the preferred survey method in such contexts. It is a perfectly valid sample method but highly inadequate at discovery; shovel-testing will discover some sites but fail to discover most. Indeed, controlled comparison of shovel-testing and surface survey suggest that shovel-testing discovers about one site for every 14 that surface survey would if survey tracts were cultivated (Shott 1985:78-79). Shovel-testing's inadequacies are particularly acute where sites are small in area and sparse in artifacts (e.g., Lewis 1993), qualities that describe a fair part of the Paleoindian record (Shott 1997; Jackson 1998).
The Appalachian Plateau seems impoverished in Paleoindian sites (Anderson and Faught 1998:Figs. 1-2; Seeman and Prufer 1982:Fig. 2). This condition may reflect a small physical-finds population but also the taphonomic effects of a low rate of cultivation and surface exposure there. Certainly the perception of a sparse record is belied by the surprising number of sites in one Pennsylvania county in Appalachia, a number revealed only by sustained efforts of local collectors (McCracken 1986).
The Appalachian Plateau covers only the southeastern corner of Ohio among Midwestern states. But the argument can be extended to any areas where ground cover is forest or pasture which, as above, are common in the Midwest. If type and amount of ground cover influence the probability of site discovery and the size and composition of the accumulated physical-finds sample (Lewis 1993:51), then we must control for their effects on the Paleoindian sample.
An apparent concentration of fluted bifaces near St. Louis (e.g., Smail 1951) may register the effects of St. Louis collectors, not Paleoindian site distributions. Thus, O'Brien and Wood (1998:58) interpreted the distribution of fluted bifaces in Missouri as some compound product of original distribution and varying degrees of collection intensity. Similarly, Henry and Nichols (1963:122) suggested that the observed distribution of Paleoindian sites in Vermilion County, Illinois, may owe as much to where collectors worked as to original site distributions. As above, the abundance and distribution of Paleoindian sites and artifacts changed considerably in northern Illinois between Harrison et al.'s (1977) and Springer's (1985) studies.
Taphonomic Effects Upon Resource Abundance and Distribution
Geology, land use and collecting history bear on the accumulated Paleoindian record. These effects urge caution in interpreting tool and site distributions, and due regard for taphonomy. Indeed, Graham and Lundelius (1994:290) counseled no less in the interpretation of Pleistocene fossil distributions across the Midwest. If modern land use and other factors that condition distribution of research affect the size and character of fossil collections, no less seems reasonable to assume for archaeological distributions.
Lepper compiled fluted-biface frequencies in Indiana, Kentucky, Michigan, Ohio, Tennessee and West Virginia and considered the possibility that they were partly determined by amount of cultivated land and modern population. He (1983b:276) concluded that "the documented distribution of fluted points is a contemporary phenomenon" although his data measured biface frequency, not distribution. Lepper's analysis stood in part on Ohio data reported by Seeman and Prufer (1982). Referring to that state, Seeman and Prufer (1984) disputed Lepper's judgment. They noted that Lepper's data treatment minimized the number of independent observations (i.e., gross physiographic zones rather than the more numerous and smaller counties). Recompiling fluted-biface frequencies by county, Seeman and Prufer (1984:228, Figs. 1-2) suggested at least that fluted-biface frequency was influenced by modern population.
Obviously, the relationship is crude since some collectors travel great distances to seek artifacts; the margins of large cities should not necessarily yield the greatest number of artifacts just by virtue of their populations. Extent of cultivation may not covary with archaeological remains in part because most Ohio counties (except near cities and in the southeast) are extensively cultivated in the first place. These qualities render moot Seeman and Prufer's (1984:228) regression analysis to show the failure of modern population to completely determine biface frequency. Lepper did not claim the "straightforward situation" (ibid) charged of him.
One way to gauge representativity of the Paleoindian physical-finds population is to measure the distribution of remains on a regional scale. Random distributions imply either randomness in the underlying population or some combination of patterned original distribution and random taphonomic effects. Nonrandom distributions are more problematic to interpret, since nonrandom patterning may register the properties of the physical-finds population or again, some combination of original distribution and modern taphonomic effects that are themselves nonrandom.
Table 21 shows fluted bifaces or sites per county for states having available data. Some data are from published sources, others from SHPOs. County totals for Michigan and Wisconsin are for counties below the late Pleistocene glacial front. For Michigan, this roughly corresponds to the famous Mason-Quimby Line. For bifaces, Iowa values are combined Clovis and Folsom specimens, not either separately (T. Morrow and J. Morrow 1994). (T. Morrow and J. Morrow's [1994:47] Folsom map seems to show 28 specimens, yet they [1994:48] reported 27.) The Poisson model tests such frequency data by quadrat (county in this case) for randomness, using the χ2 statistic (Harvey 1966). Frequencies are compiled from 0-5, with all higher values grouped as 6+. Thus, the threshold χ2@.05, df=6=12.6. All results are significantly nonrandom. Both bifaces and sites are distributed nonrandomly, probably in clustered distributions.
Table 21. Frequency Distributions of Fluted Bifaces or Sites per County[Long description]
Physical-Finds Population: Fluted Bifaces
Trends in sample rate can be supplemented by gauging sample fraction. One way is to estimate the size of the physical-finds population, expressed as number of either sites or tools, especially fluted bifaces. As above, our imperfect knowledge of the range of diagnostic Paleoindian tool types makes it practically impossible to estimate the size of the Paleoindian tool population across the Midwest. Even estimating either the number of sites or of fluted bifaces is difficult enough, but at least worth the effort. It belabors the point to note that the following estimates are of unknown accuracy. Only the heuristic nature of this exercise justifies the attempt.
The physical-finds population of fluted bifaces is determined by three quantities: number of tool-using Paleoindians, rate of discard (Ammerman and Feldman's  "dropping rate") per person per unit time (discard/person/time) and the duration of the Paleoindian period or any of its subdivisions. We can estimate the size of Paleoindian populations by intelligent use of ethnographic data on population density and its relationship to environmental variables that can be measured or inferred for Paleoindian contexts. Discard rate is more difficult to estimate, since we utterly lack ethnographic data for the two quantities that determine it: number in use per person and object use life.
Shott (n.d.) discussed these questions at greater length. Table 22 shows different results depending upon whether only men (Scenarios I and II) or men and women (Scenarios III and IV) used fluted bifaces. The result is estimates of the physical-finds population, not the human population. If the middle Paleoindian period's duration were much shorter, the estimated number of fluted bifaces would be less. If use life were longer or dropping rate lower, again the estimate would decline.
Table 22. Estimated Number of Early and Middle Paleoindian Fluted Bifaces Discarded in the Midwest.
Yet input values either were given (phase duration) or estimated conservatively. The figure of approximately 8,500 fluted bifaces recently documented east of the Mississippi River (Anderson and Faught 1998:170) is impressive but comprises a tiny fraction of the lowest value estimated in Table 22. The 8,500 bifaces also includes early and late Paleoindian bifaces, not just the middle Paleoindian ones the subject of these calculations. Anderson and Faught's (2000) most recent data include 2,551 fluted bifaces from Midwestern states, the great majority identified as "Clovis-like." If C. Haynes's (1966; cited in Anderson and Faught 1998:170) estimate of millions of fluted bifaces seems high, Anderson and Faught's compilation is less than 1% of the minimum estimate. This is a criticism of neither source, which had other purposes than this exercise's. Anyway, Anderson and Faught compiled empirical figures, not estimates; 4,250 bifaces is probably a suitably large sample for many analytical purposes. Yet it may be a small fraction of the total made, used and discarded.
Thematic Research Issues for the Midwest
The Midwestern Paleoindian record is sufficiently large and diverse to accommodate research projects from many empirical and theoretical approaches. It would be folly to propose a comprehensive set of questions to guide all future research. Instead, the following questions and problems might help direct Midwestern Paleoindian research and provide a framework for NRHP and NHL evaluation of Midwestern sites. They also identify some kinds and amounts of evidence needed for their answers. The questions complement the research themes presented in the national context. Obviously, there is some overlap between themes owing to their complex and related natures.
On a continental scale the Midwest literally was in the middle of the peopling process; although the earliest Midwestern prehistory cannot be learned in isolation, the history of peopling other parts of the continent take second place here.
The long and contested history of pre-Clovis claims and evidence was reviewed briefly above, more thoroughly by Dincauze (1984) and Fiedel (2000). If the Chesrow Complex indeed is early Paleoindian in age, the Midwest's first human occupation occurred as early as 13,500 rcbp. Until this possibility is confirmed, there are claims but no conclusive evidence of people in the Midwest before circa 11,000 rcbp, roughly the end of the early Paleoindian period. One of the beauties of archaeology is that this judgment can be mooted in the time required to discover and document a site, so it is never more than a contingent judgment.
From wherever they came, the first people surely reached the Midwest on foot or in very small watercraft. How they moved in and occupied the Midwest remains unknown. Martin's (1973) melodramatic wave-advance model rests on dubious assumptions about population size, distribution, hunting rates and practices, colonization rates, and the adaptability of animals to hunting; in any event, there is no Midwestern evidence for the suggestion that many people were crowded into an assault front that ravaged vulnerable megafauna . It is more reasonable to assume that Paleoindians colonized the Midwest intelligently. The available evidence and radiocarbon chronology are far too inadequate to reveal details of the timing, let alone the manner and direction, of colonization. Anderson and Gillam's (2000:Figs. 2-3) suggestion of first entry from the Missouri-Mississippi confluence area eastward and northward seems reasonable.
After the first people colonized the Midwest, their Paleoindian descendants occupied it for at least centuries. In that span, what were the trends in settlement and population? Was human population small until the Midwest was filled at some threshold value? Did it increase thereafter and, if so, how fast? Such questions are valid not just for historical but also theoretical reasons, for understanding the subsequent history of colonizing populations.
Again leaving aside proboscidean sites, Table 8's radiocarbon dates are a rough index of Paleoindian population trends through time, but the evidence comes from three sites only, almost all from two, and is the product of justified but nonrandom selection. The nature and scale of perceptible chronological trends depend upon our ability to measure and resolve past time. Radiocarbon dating probably is too coarse to resolve time sufficiently to gauge Paleoindian population trends. A more promising approach is refined typological dating, especially if trends are expressed in continuous terms in metric variables. Ellis and Deller (1997:5-8) proposed as much for Great Lakes fluted bifaces, postulating an empirical trend to reduced size of fluted bifaces through time. Variables that define the trend—principally thickness, base width, and basal concavity—are not subject to change during the use life of specimens, so the apparent trend is not the by-product of the differential reduction of individual tools. As we find, excavate and study more Paleoindian lithic assemblages, we might observe similar trends in other types. Doing this, we might resolve time in near-continuous terms and so begin to measure population changes through time, among other things.
Creation of Social Institutions
Archaeology in the 1960s was a sincere attempt to plumb the material record for its social dimensions. It foundered largely on the difficulties of teasing social units from remains governed by complex formation processes, not by any theoretical deficiency (Shott 1998). One early extension to lithics was in Voss's (1976:269-272) analysis of fluted bifaces and channel flakes from the Barnes site (see also Wilmsen and Roberts 1978). His cluster analysis identified four social units there.
Shaping the Political Landscape
In preindustrial societies, politics involves alliance and accommodation for various material and social purposes. Ethnography documents patterns of interaction between forager cultures on near-continental scales, interaction often symbolized in sumptuary items. Presumably, the scale of distribution of such items reflects the scale of political alliance and the pattern of distribution—abundance with distance from sources—reflects its character.
Hayden (1982) used this general approach to argue that Paleoindian colonists of the midcontinent required active negotiation of political alliances as they expanded and even as they simply persisted in their ordinary ranges. In his view, as population and social density increased during the Paleoindian period, band ranges declined. Thus, progressively through the Paleoindian interval material markers of social identity—practically, stone tool types or "styles"—should have grown more localized, the number of styles greater on a regional scale. In part Hayden grounded his thesis in a view of alliance and formation of political landscapes as insurance on the scale of bands.
Uneven Midwestern evidence makes it difficult to assess Hayden's thesis for the early and middle Paleoindian phases. Yet the character—scarcity, fine workmanship, size and thinness—and distribution of Sloan points or "blades" in midcontinental Dalton assemblages suggest that they might have circulated among Dalton bands as material symbols of alliance (Walthall and Koldehoff 1998).
Toolstone and Paleoindian Mobility. To move beyond ethnographic analogies, we must use existing theory that links environmental parameters to forager mobility (e.g., Kelly 1995:111-160) and so reasonably infer Paleoindian mobility practices on theoretical grounds. Then we must develop the theory of technological organization (e.g., Shott 1986a) that links mobility practices to materials remains, and so test mobility theory in archaeological evidence. Paleoindian studies are ideally suited to the problem.
Mobility is a set of practices expressed in frequency, the number of residential moves made per unit time, and magnitude, the cumulative distance covered in those moves (Shott 1986a:21-22). Both in theory and practice (Shott 1986a, 1986b, 1989b) frequency and magnitude have different effects and groups can differ inmobility frequency or magnitude despite having similar ranges. In the Great Lakes, mobility frequencies reveal greater differences than do mobility magnitudes suggesting they reflect differences in how often people moved to acquire tool stone rather than differences in how far people moved to acquire it.
Almost always, archaeologists assume direct acquisition by users, as opposed to trade. But direct acquisition is ambiguous if only some members of a society actually acquired toolstone and then distributed it to their fellows. Thus, the direct-indirect distinction applies at the level of minimum social groups; from the start, toolstone distribution is socially implicated.
Direct acquisition may be purposive involving travel expressly to obtain stone. Binford (1979:259) considered this practice rare in forager cultures, arguing that toolstone acquisition was embedded or accomplished incidental to other activities during travel. Embedded acquisition must be sometimes but not always the practice, extended travel to stone sources equally possible. Ethnographic documentation of purposive travel for food and other resources is legion, the observation banal. Toolstone is a common good like food, so it should not surprise that people sometimes travelled to acquire it. More likely, most travel was for many reasons and embedded toolstone acquisition no more nor less than other activities. Indirect acquisition involves exchange for social or economic reasons, the latter varying greatly in scale and pattern (e.g., "down-the-line," gravity). Society-wide large-scale and individual small-scale exchange can be distinguished, as well perhaps as Meltzer's (1989:23) gift and ceremonial exchange.
Since toolstone acquisition implicates social conditions, it reflects them in the archaeological record. Consequently, archaeologists long have struggled to identify the often subtle differences between variants of direct and indirect acquisition. Sometimes we distinguish between possibilities via properties of assemblages like abundance, technology and context. If more toolstone is acquired directly than indirectly, a source's abundance might imply direct acquisition. If sources occur in archaeological deposits in proportion to their natural abundance (perhaps adjusted for quality), then direct embedded acquisition is implicated; people simply acquired stone incidentally to other land uses. Meltzer (1989: Table 2.1) has catalogued general Criteria than can affect the presences, abundance and typological richness, stylistic properties and reduction processes of local/nonlocal stone (within reason, the more such archaeological Criteria applicable, the better). All else being equal, the greater the distance between source and archaeological occurrence, the likelier that toolstone was exchanged (see Brose 1989). What is local for high-latitude foragers may be quite exotic for densely settled farmers. Even in roughly similar contexts, archaeologists use many different threshold values between local and nonlocal, although ca. 30-40 km seems common.
Meltzer's (1989) Table 2.1 is a catalogue of often undetermined factors that determine the presence, abundance and typological richness, stylistic properties and reduction process of local/nonlocal stone. Within reason, the more archaeological Criteria used the better.
Toolstone Distribution. Most Paleoindian stone tools were made of chert, although metamorphics like quartzite sometimes were used. Major Midwestern toolstones include Burlington chert in the St. Louis area, Hixton silicified sandstone, quarried at Silver Mound in western Wisconsin, Attica chert in western Indiana, Harrison County chert from southern Indiana, and Upper Mercer and Flint Ridge cherts from eastern Ohio. Pennsylvanian chert very similar to Upper Mercer outcrops in Moline, Illinois and in southern Missouri (Ray 1983), and was dominant in Paleoindian assemblages as distant as Aebischer in eastern Wisconsin (Mason 1988). It may be the "dark, almost black" (Clark 1985:118) chert of another possible Gainey biface from that area although Moline chert is nodular rather than tabular (Birmingham and Van Dyke 1981:348). Collingwood chert, near southern Georgian Bay in Ontario, was a common Paleoindian toolstone in Ontario, but occurs rarely in neighboring American states. Many cherts were distributed in secondary deposits like glacial drift or the channels of major streams (Shott 1986b:125-126). To use distance-to-source we must be sure of toolstone and sources. Most identifications are visual, which is problematic for some of the sources: Flint Ridge and Upper Mercer are notoriously variable in color, luster and other visible properties. Short of neutron-activation analysis or the study of diagnostic fossil inclusions, many other sources cannot be distinguished.
Ingbar (1994) examined the distribution and relative abundance of chert sources across the landscape, varying only mobility frequency and size of tool inventory between three different siluations. He found that source proportions varied widely between "sites" and that frequently no tool from sources used at a "site" was left there as evidence of use. Ingbar (1994:46) concluded "there is not a good correlation between raw material source proportions and territory" and that range is underestimated by the distance between material and cultural occurrence of sources found in many assemblages. Bradbury and Carr's (2000) recent simulation included debris as well as tools and showed that replacement (using tool reserves kept ready for use versus returning to source, reduction technology), and artifact use life affected the abundance and distribution of toolstone independently of "mobility."
Of course, range can be as easily overestimated by wrongly inferring that exchanged stone was obtained directly, even if sources are properly identified. The Gainey assemblage (Simons et al. 1984) lies over 300 km from its chief source of toolstone. Sheriden's chief source lies 400 km away (Tankersley 1999:70). Such distances are reported elsewhere (Paton 1994:177), and the quantity of this chert and the full range of reduction products represented make direct acquisition the likeliest mode here. Taking this distance as the maximum travelled in a circular range yields roughly a 70,000 km2 area. Curran and Grimes (1989) inferred a Paleoindian range encompassing much of New England and roughly 140,000 km2 on the strength of toolstone distribution in five major assemblages. Even the lower of these two values greatly exceeds the highest reported ethnographically (Kelly 1995:Table 4-1).
In Michigan, early Paleoindian Gainey Phase assemblages are dominated by Upper Mercer chert. Subsequent Parkhill Phase assemblages are dominated by the local Bayport chert, occasionally supplemented by Upper Mercer and rarely perhaps by Collingwood chert. This shift is interpreted as evidence for lessened mobility in the sense of more restricted range. Most archaeologists argue that later Paleoindian cultures, certainly Early Archaic ones, were even less mobile (i.e., ranged more narrowly). Yet many Early Archaic Michigan (LeCroy) bifaces were made on Upper Mercer chert. Some might argue that Upper Mercer was obtained via trade by Early Archaic groups, but that its abundance among industrial debris in assemblages like Gainey's testifies to direct acquisition. Until more Early Archaic assemblages are excavated in Michigan, northern Ohio and Indiana, we will not know enough about their industrial character to settle the question.
Ingbar's simulation, and perhaps Curran and Grimes' data, show that any single system of use and technology can create wide intra-assemblage variation in source presence and proportion. Only when important assumptions are met can material distance serve as a measure of range. We already have the theory (Nelson 1991; Shott 1986a) and have begun to develop the methods (Brosowske 1996; Shott 1986b, 1989b) to distinguish magnitude and frequency. We must get beyond sole, even primary, reliance upon distance-to-source because it is an inherently crude measure.
Developing the American Economy
If Paleoindians were the first Americans and Midwesterners, they developed the first economies here. Economy includes socially organized subsistence. We know little about how subsistence was socially organized. We know litle more about what was eaten. In the Northeast, it is now formulistic to describe Paleoindian subistence as broad as it was once described as narrowly focused on megafauna. In the Midwest, most archaeologists still view Paleoindians as hunting specialists (e.g., Ellis and Deller 2000:255). Caribou figure prominently in most reconstructions. Indeed, Jackson and Thatcher's (1997) volume practically enshrined this view. Mason (1981:99) reported a fluted biface lodged in an elk rib (Cervalces sp.). Nevertheless, exclusive focus on large mammals is questionable (even the earliest Clovis sites in the southern Plains yielded a variety of small mammals and reptiles, although bison and ground sloth were present [Ferring 1995:277]). Lacking "better-preserved representative diet samples, we cannot assess the relative importance of hunting and other Midwestern Paleoindian subsistence activities" (Fiedel 2000:77).
Smaller mammals and other classes are much less abundant in the archaeological record, though fairly abundant in the fossil one (e.g., McDonald 1994). Udora, a Gainey-affinity assemblage in southwestern Ontario has yielded hare (Lepus spp.), Arctic fox (Alopex lagopus) and caribou (Rangifer tarandus) (Storck and Spiess 1994:126-128). Fourteen bones, most calcined and all apparently mammalian, were found at Halstead, and cervid and unidentified mammal bone was found at Sandy Ridge, both Gainey-affinity and also in southwestern Ontario (Table 9; Jackson 1998:27, 56-57). None was recovered from feature context, but calcining suggests human agency and Halstead and Sandy Ridge are single-component Gainey deposits. Possible antler and other elements identify at least one cervid, likely Odocoileus sp. but perhaps Rangifer sp. One uncalcined incisor is from Castor sp. Ohio's Sheriden Cave yielded probable Paleoindian artifacts in deposits rich in Pleistocene and Holocene fauna that include reptiles and fish as well as small mammals, caribou, deer, stag moose, and giant beaver (Holman 1997; McDonald 1994; Tankersley 1999a:68-69). One mammal long bone was worked into a point, but other remains may combine natural and cultural occurrences. The Chesrow occupation at Lucas yielded unidentified calcined bones (Overstreet 1998:39). Hiscock's culture-bearing deposit also contains a surprisingly diverse biota with mastodonts, cervids and California condor (Gymnogyps californianus) (Steadman 1988). The late Paleoindian Holcombe Beach site yielded caribou, probably R. tarandus (Cleland 1965). More broadly, Kuehn's (1998) survey of late Paleoindian subsistence evidence from northern Wisconsin revealed a broad range of Holocene species. Although there are natural occurrences of caribou in the Midwest (e.g., McAndrews and Jackson 1988) there are no known kill or butcher sites of caribou or other game (O'Connell et al. 1992).
There are few if any plant remains from Midwestern Paleoindian sites. Chenopodium (Chenopodium spp.) and perhaps acorn (Quercus spp.) occurred in late Paleoindian deposits, and wild grape (Vitis spp.) in middle to late Paleoindian contexts at Big Eddy (Lopinot et al. 1998:284). Small amounts of hickory ( sp.) shell were recovered from a Paleoindian feature at CB-North in the American Bottom (Evans et al. 1997:170). Paucity of evidence owes partly to the few intact Paleoindian deposits like features ever examined thoroughly for subsistence evidence using fine screening or flotation. Where this has been attempted (e.g., at PaleoCrossing [Brose 1994]) no floral remains were recovered other than fragments of possible posts.
The Midwestern Paleoindian record substantially is a lithic one. To determine Paleoindian economic development, our first task is to identify the economic properties of toolstone. One is abundance although Paleoindian consumption was far less than the supply. Abundance can be measured by number of cobbles or tabular pieces, by their volume or weight. Cobble size and form constrain size and form of finished objects, but mostly affect technology and efficiency of reduction (Bradbury and Franklin 2000). We must thoroughly sample outcrop and secondary deposit sources to ascertain the natural ranges of variation in these properties. Studying only those samples recovered from cultural sites, selectively acquired and used by Paleoindian and later knappers, has skewed the original districution. Finally, fracture and other mechanical properties of toolstones (McCutchen and Dunnell 1998), these bear on the selection, technology, tool design, use and discard of tools.
Expanding Science and Technology
How Paleoindians fluted can be understood, in part from Midwestern evidence. Even now we cannot agree on fluting failure rates (Ellis and Payne 1995; Storck 1983). Length and width of flute channels perhaps rose from Gainey to Parkhill Phases, then graded into basal thinning by the late Paleoindian period. Since extent of fluting varied over time, the reason for fluting perhaps varied also, at least by degree. In some sense, fluting may have been done to increase the surface area of contact between point and foreshaft, thus to improve fixing in hafts. Residue analysis might be attempted on flute channels, although we need not expect blood residues or antibodies there unless Paleoindians mixed animal blood to form mastics. If fluting ability were valued arbitrarily, then the best fluters would enjoy social advantage, perhaps excluding others from the activity as a way to reinforce status or even to control the distribution of fluted bifaces.
A great unanswered question in Paleoindian studies is how fluted bifaces were used as projectiles. Did they tip hand-held spears, thrown ones, or atlatl-launched darts? Hutchings's (1997) method poorly distinguished dart and arrow use, but seems able to distinguish dart use from use as hand-held or thrown spear. It should be applied to Midwestern Paleoindian fluted bifaces.
Technological Organization. Technology involves not just the tools people made or how they used them, nor just the social dimensions of production and use. It involves organization as well, how tool production and use are integrated into larger cultural wholes. Studies in the Midwest showed how Paleoindians planned their tool use to avoid shortages as they moved away from sources (Ellis 1984), and how mobility, especially its frequency, influenced tool design and the kind and number of tools used (Shott 1986a, 1989b). Midwestern Paleoindian sites can demonstrate via organizational studies that technology is deeply embedded in cultural contexts. This lesson may encourage us to view our own technology properly in its social context.
Transforming the Environment
Our view of hunter-gatherers vacillates between the romance of Rousseau—primitive paragons of virtue in sweet harmony with their environment–and the gloomy judgment of the cynic—that despoilation is a timeless human activity. In the 1980s it became fashionable to argue that even hunter-gatherers were polluters: our equals in kind who lacked only the means to wreak environmental havoc so thoroughly as we. But people always have been ambivalent about changing the environment, and much of our history celebrates triumphs which control and thus transform nature. If hunter-gatherers are not gentle stewards of nature neither are they passive ciphers in it.
Anthropologists have come to recognize how hunter-gatherers among others used fire to transform their environments, sometimes subtly and always in complex ways. Perhaps Paleoindians did as much, a practice whose effects would be hard to distinguish among the major climatic and historical changes wrought at Pleistocene's end. Otherwise the question of environmental effect reduces to the possible role that Paleoindians played in the extinction of Pleistocene megafauna. After decades of avid searching, however, there are surprisingly few certain human associations with proboscideans. Fisher's and colleagues' taphonomic virtuosity makes a suggestive but not conclusive case. Liberally accepting all such claims, half or more of analyzed proboscideans remain natural deaths without human intervention. Mammoths had rather narrow ecological tolerances that almost certainly doomed them in the late Pleistocene's climatic furies even had humans never reached North America. Broader diet and habitat preferences made mastodonts a hardier species, but one already experiencing natural stresses that invited extinction (King and Saunders 1984:331) before humans arrived. At Pleistocene's end, the Midwest was populated also by short-faced bears, sloths, horses, giant beavers and other megafauna now extinct. These too have been found at Sheriden (Tankersley 1999), yet we do not routinely ascribe their demise to Paleoindian hunters. Moreover, we have direct evidence that Midwestern and nearby Paleoindians hunted caribou, hare and Arctic fox, yet these species survive today.
Significant Research Concerns for the Midwest
Typology. The theoretical basis of Paleoindian fluted-biface typology is largely underspecified. We must construct typologies (not reveal types), defined by fairly strong patterns of association between metric and discrete variables; at least broadly, fluted biface typology captures and summarizes simultaneous variation in biface technology, size and form.
Fluted-biface typology must be grounded in theory of biface size and form and production technology governed by functional or social constraints. We should ground our typologies in theories of whole-object form, and the functional requirements of hafted bifaces probably used as weapon tips. Such approaches require us to determine which variables register which functional constraints, and how they register them and to distinguish functional from stylistic variables if in fact they are different. Over time, ratio variables under functional constraints should trend monotonically stylistic variables should vary at random. Categorical variables under functional constraint should fix at constant proportions over time; stylistic ones should form the chronologically normal curves of mid-twentieth century models (Beck 1998). In fluted bifaces, typology usually rests upon gross size, plan form, and details of fluting and base form. But until we identify relevant functional variables, we cannot be certain if these measure functional or other design constraints. Hughes's (1998) consideration of engineering theory led her to identify mass, perimeter, and cross-section area as important functional variables in points. Paleoindian biface mass sometimes is reported, but the other variables rarely if ever are.
J. Morrow (1996:159-173) noted the deficiencies of earlier typologies and was careful to remove reduction effects from her own typology of biface finished form (1996:193-194). She also advocated replicable typologies based on quantification. In her own typology, Morrow used four variables (basal concavity, base width, maximum thickness on the flute, and maximum width) to compute two ratios (concavity-to-base-width, maximum-thickness-to-maximum-width), and defined her types on the strength of patterns in the ratios. In this way, she (1996:194-200) distinguished Clovis from Gainey from Vail/Debert. Of necessity, Gainey was represented chiefly by the Lamb site (Gramly 1999) because too little data were available from the type site. Two ratios seem reasonable but were not justified by any theory that linked variables to technology, form or intended or actual use. Also, if the variables that underlay the ratios were correlated, then the patterns in them identified as typological may be continuous; these types may be arbitrary subdivisions of continuous variation.
Barrish (1995) compared bifaces from the early Paleoindian Paleo Crossing, Nobles Pond and Gainey sites, the latter subdivided into Upper Mercer and Ten Mile Creek specimens. First she (1995:Figs. 12-19) compared the assemblages by basic metric variables. Then Barrish (1995:92-93) performed principal-component analysis on metric variables except length, because length is affected by resharpening. Finally she (1995:96) distinguished the three assemblages via discriminant analysis. Barrish (1995:104) concluded interestingly that Paleo Crossing, Nobles Pond and Gainey formed a time series within a broader Gainey type, thus combining discrete and continuous approaches.
Recognition of data limitations (Barrish's multivariate analyses were performed on 27 specimens apportioned among three assemblages) suggests small samples might be unusually affected by sample error. Barrish omitted length from analysis because it is affected by reduction, but she included position of maximum width (distance from base to point of maximum width) and flute length, which also can be sensitive to resharpening. Principal component 1 was size (Barrish 1995:Table 10) as it often is in metric data. Barrish (1995:93) interpreted component 2 as a haft dimension involving flute length and basal concavity (Simons et al. [1984:268] noted a similar correlation between flute width and concavity), but specimens that scored low on the component seemed heavily reworked (1995:Fig. 21). If their original flute length was reduced by resharpening, then measured flute length is influenced by resharpening and should be excluded from analysis. Barrish's discriminant analysis did not include thickness, although this variable is as significant as those included. Her study did demonstrate significant metric differences between Gainey subtypes defined by different toolstones.
Ellis and Deller (e.g., 1997) did not use multivariate methods to classify fluted bifaces, but did use many technological and formal variables to justify the three successive early to middle Paleoindian biface types Gainey, Barnes and Crowfield. Their study ordered the types in a time series independently confirmed by contextual data (geochronology, toolstone availability and selection, typological cross-dating). It also demonstrated that toolstone selection, resharpening, function and discard patterns do not produce or even complicate the time series. Thus, Ellis and Deller argued that patterning owed to time, not extraneous factors. Their views on discard (1997:6) should be strengthened by a more thorough consideration of discard processes and their effects on tool size and form (e.g., Shott 1989a:173-183). On balance, however, Ellis and Deller's typology is grounded in a relatively comprehensive set of relevant shape and size variables.
Darwent et al. (2000) performed cladistic analysis on attributes from 500+ southeastern fluted and other bifaces. Results (2000:7) suggested a great deal of "homoplasy"—convergence, independent invention or diffusion of attributes—much surely for functional reasons. Cladograms rather freely mixed defined types (e.g., one "clade" grouped Cumberlands, most Quads and some Daltons) that, at least in the Midwest, contextual evidence clearly distinguishes. Darwent et al. did not entirely control for the confounding effects of resharpening but cladistics is an innovative way to perceive and interpret historical patterns of change in biface morphometrics. It may revise type definitions and reveal unsuspected complexities in the time order of Paleoindian biface traditions.
The Midwest probably has the greatest expanse of cultivated land in the United States, and most Paleoindian sites were found by surface inspection of plowed fields. But cultivated surfaces are complex sampling domains subject to considerable random and systematic variation between plowings. Further, no one surface collection is faithful to a site's properties. Several documents of one Michigan record varied enough to make it change from "hunting camp" to "processing site" to "armament site" (Shott 1995b:Fig. 7) using Judge's (1973) taxonomy. A considerable literature shows that an adequate sample of a site requires at least five collections of its surface (e.g., Shott 1995b).
Archaeologists' systematic surveys across the Midwest occasionally discover Paleoindian sites. But practically all professional surveys involved single passes across the surface. Collectors often are local and energetically visit the same site many times. It is no surprise, therefore, that collectors have discovered more Paleoindian sites and probably more fluted bifaces by far than have professional archaeologists. Southwestern Ontario near Toronto probably boasts the largest concentration of Paleoindian sites and certainly the fullest range of variation in assemblage size, density and location in areas near the Midwest. The record (e.g., Ellis and Deller 1997) is the product of decades of sustained collaboration by local collectors and nearby professional archaeologists. Perhaps the southwestern Ontario Paleoindian record truly is exceptional. Until similar efforts are made across the American Midwest, however, we never will know. Paleoindian sites often require two, three or more visits before the discovery of the first diagnostic artifact. Morrow-Hensel and Gail Stone in western Wisconsin were visited by local collectors many times before significant collections were amassed (D. Amick, personal communication 2000). Missouri's Walter site was found only on the second visit there (Nichols 1970). Among the nine Paleoindian sites in southern Wisconsin's Yahara Valley, only one would have been found in a single-pass survey (Wendt 1985:245). It took eighteen years of collection to discover fewer than 10 fluted bifaces among the 19 Paleoindian and late Paleoindian bifaces found near Danville in eastern Illinois (Henry and Nichols 1963). Hawk's Nest required "Repeated surface collection" over four years to yield a Gainey assemblage of about 80 tools (Amick et al. 1997:4). Anderson in northern Illinois required "more than two decades" (Koldehoff 1999:12) of collecting to produce a modest but significant assemblage.
Also in northern Illinois, Kishwaukee Basin sites tend to be diffuse and relatively unproductive. "In order to acquire a reasonable idea of the sequence and intensity of occupation, numerous visits are necessary" (Springer 1985:10). Culloden Acres was known since the 1960s and presumably was visited many times before the first diagnostic Paleoindian artifacts were found there (Ellis and Deller n.d.:8). Discovery of other small Ontario sites required intensive survey sustained over many years (Ellis 1994:423). To Jackson (1996:13), "literally years" of repeated survey were required to locate, define and compile substantial collections from Ontario Paleoindian sites. Twenty years of collection were necessary before so many as 25 fluted bifaces were found at Martens (O'Brien and Wood 1998:60). In Illinois "intensive, multi-year surveys are necessary when studying the remains of mobile foragers since the archaeological signature of such groups is only faintly represented...and is likely to be missed during short term suveys" (Walthall and Koldehoff 1999:30). Elsewhere, Virginia's Nottoway Valley survey spanned nearly 30 years, 12 of which involved intensive efforts (McAvoy 1992:13).
We might survey the regional Paleoindian record using both conventional and new techniques. The former includes intensive surface survey of cultivated fields. The latter includes extensive mechanical excavation, for which efficient methods have been developed to process (Steinberg 1996; Van Horn 1988). Using a purpose-made machine in Denmark, Steinberg (1996:Fig. 2) sampled plowzone volumes vastly greater than would be practical by hand. Using similar methods in the Midwest, we could efficiently define the regional dimensions of the Paleoindian record free of the constraints but not the virtues of the site concept. Thus we could identify concentrations that might legitimately be treated as sites but also measure continuous background densities of Paleoindian remains. Such methods would demonstrate resourcefulness and technological ingenuity. We too can be innovative under difficult conditions.
Obviously, the accumulated Midwestern Paleoindian record is substantial. As above, there should be many opportunities to collect regional assemblage data. The nearest extant Midwestern version seems Lepper's (1988) report on Coshocton County, Ohio assemblages near Upper Mercer chert sources. Regional survey may be selective or probabilistic in design. The effort to compile the necessary data may require some fundamental reordering of institutional arrangements for fieldwork. I argued (Shott 1992b) that general preservation needs would best be served by reorganizing toward sustained regional survey, not project-specific measures. This is the archaeological equivalent of fire management and systematic prevention over putting out brushfires.
Improved survey standards will have useful consequences and would produce regional-scale assemblage data, heretofore unknown in the Midwest. Paleoindian land use and cultural organization were regional in scale and a single system obviates the problems of typological comparability. Such data exist in the southwest (Judge 1973) and southeast (McAvoy 1992), but not in the Midwest. Sustained fieldwork in southwestern Ontario is approaching this goal, but elsewhere in the region the nearest equivalent is studies that combine primary sources of varying quality and detail (Shott 1997).
Another likely advantage of survey standards would be correcting small site underrepresentation in conventional survey (Shott 1985). Illinois survey records suggested that small sites of every age are discovered at lower rates than larger ones (Lewis 1993:48). Find spots of single fluted bifaces seem fairly common among Paleoindian sites, although they can become sites yielding many artifacts upon repeated inspection. Many Paleoindian sites have few artifacts and are small in extent, and they may be underrepresented in professional survey and study. In Ontario, small Paleoindian sites were most commonly found only after intensive survey (Ellis 1994:423). To Ellis and Deller (n.d.:3-4) the understandable emphasis on large sites biases the accumulated data base and somewhat distorts our view of Paleoindian cultures. Small Paleoindian sites may reveal brief occupational episodes and exhibit greater integrity in their spatial distributions.
Vast artifact collections molder in barn lofts and basements across the Midwest. We must systematically inventory the invaluable data that such old collections possess. Doing so will require effort sustained over many years, and funding to support the necessary work by students and SHPO offices. Such efforts must aim toward the documentation of collections in their entirety, not just their Paleoindian components. There are practical and ethical obstacles to purchasing collections, but owners might be encouraged to donate them to responsible repositories via the inducement of tax deductions.
Collections are lost to archaeology far too easily and often. Surely hundreds of collections are at risk today; our inaction will allow the problem to solve itself, but only at great cost to the interests of knowledge and preservation. There is no time to waste in rescuing collections from neglect and eventual oblivion. As part of the NHL Theme Study, the Society for American Archaeology and the National Park Service might launch a collections documentation and preservation project consistent with the public education and outreach goals espoused by both organizations.
In the absence of systematic data, we cannot know how many fluted bifaces, let alone other Paleoindian tools, are held by universities and museums and how many by private collectors. A rough estimate of the proportions might be gained by comparing the number of Midwestern Paleoindian sites found by archaeologists and by collectors; we are apt to learn that archaeologists control 5% or less of the accumulated sample. That is approximately the percentage of the sample of diagnostic artifacts identified in one Swedish study area about which professional archeologists were aware. They control about that percentage of diagnostic Neolithic artifacts in one Swedish study area (Karsten 1990:36).
One way to gauge the compositional representativity of Paleoindian assemblages would be to compare tool type proportions in systematic collections to private collections. Probably fluted bifaces would be overrepresented in the latter. For instance, most fluted bifaces in the Leavitt assemblage (Shott 1993) were found by collectors in the family over decades; only a few were found in systematic work there. Obviously, we cannot know how many fluted bifaces were found by other collectors and then lost to documentation. Yet even this measure would be limited because the available excavation sample surely is not probabilistic and therefore may be unrepresentative. We tend to excavate where we know or expect that there are many artifacts to find, and such locations are not necessarily typical.
Another interesting exercise would be to compile tool counts by type from Midwestern and other Paleoindian caches. Obviously, caches often contain fluted biface but also other tools. Some, like Udora (Storck and Tomenchuk 1990), contain no fluted bifaces at all. At any rate, type proportions from caches might reflect proportions in use free of the effects of use rates and use life. If type proportions from private collections and perhaps from systematic surface surveys differ from cache proportions, they may be biased and the sites from which they were drawn a nonrepresentative sample of Paleoindian assemblages.
National Historic Landmark Criterion 6 and National Register Criterion D / Registration Requirements
For specific properties, NRHP and NHL evaluation should proceed by first completing a property designation matrix, which provides a basic overview of condition and research potential under National Historic Landmark Criterion 6 and National Register Criterion D, and can serve as a guide to the preparation of detailed nomination statements. Procedures by which the matrix is to be used are provided in the Evaluation Criteria Matrix / Registration Requirements section above. How specific research questions and themes outlined in the matrix apply to Midwestern Paleoindian properties, however, were discussed previously in this regional context.
NHL Property Types and Integrity Levels for the Midwest
Linked with the evaluation of specific NHL property classes and categories is an assessment of their integrity. Property integrity refers to the physical condition of the remains under investigation, i.e., their preservation, context, and ability to contribute important information under Criterion 6 for NHL designation and Criterion D for the National Register. To be considered eligible for designation as an NHL, Paleoindian properties must possess deposits with sufficient integrity to yield information capable of identifying discrete periods of occupation or utilization, property function or type, and must have clearly defined boundaries.
Three levels of integrity are recognized in the NHL theme study: High, Moderate, and Low.
High Integrity. Properties possessing high integrity are potential NHLs or have national-level NRHP significance. Such sites have clearly identified Paleoindian component(s) in secure context, and with precise calendric dating. That is, the geologic and sedimentary context of the assemblage(s) are well documented, with sources of intrusion or disturbance recognized and controlled, and the age of the deposits ascertained using one or more absolute dating procedures, such as radiocarbon or thermoluminescent dating. Sufficient age determinations must, however, have been obtained from samples in secure context to ensure confidence in the results. Individual dates, accordingly, or even large numbers of dates from controversial associations, will probably not be considered sufficient, unless supported by other kinds of evidence, such as unambiguous geological or biotic associations. In the Midwest, sites with high integrity and national level significance include Big Eddy, and perhaps Sheriden Cave and Chesrow Complex sites.
Moderate Integrity. Properties of moderate integrity are potential NHLs or have national- or state-level NRHP significance. Sites with Moderate integrity have Paleoindian component(s) that are to some extent mixed with later materials, in moderately secure context, and with relative rather than absolute dating. Thus, geologic and sedimentary context may be somewhat uncertain, with some mixing or reworking of the deposits. Control for disturbance is less secure. The age of the deposits is also somewhat less secure, and may depend upon stratigraphic relationships or typological cross-dating with materials securely dated elsewhere. Many midwestern sites possess moderate integrity are widespread, and include most assemblages found on conflated, usually plowed, surfaces, where distinguishing Paleoindian remains from materials dating later is sometimes difficult or impossible. Nevertheless, such assemblages often retain great research potential and considerable sub-plowzone integrity. Moderate-integrity sites are many, but Gainey and perhaps Grogitsky in Michigan, Nobles Pond, Paleo Crossing and perhaps Sandy Springs in Ohio, Aebischer in Wisconsin, Lincoln Hills/Ready, Mueller and Martens in Illinois, and perhaps Walter in Missouri.
Low Integrity. Properties whose integrity is Low are not considered NHL candidates. If they were to be considered eligible for inclusion on the NRHP, it would probably be at the state or local level of significance. Low-integrity sites have presumed Paleoindian components that are in highly disturbed context, and whose age may be uncertain or questionable. Lithic scatters lacking diagnostics, absolute dates, or sound stratigraphic contexts are examples of such sites, as are sites with diagnostics whose deposits are severely disturbed or are thoroughly mixed with materials of later periods. Sites yielding low numbers of Paleoindian points as well as later materials in surface context, would tend to have Low integrity. Many Midwestern Paleoindian sites that might have possessed higher integrity now possess little or none. These include Holcombe in Michigan, Bostrom and CB-North in Illinois and Rummels-Maske in Iowa because the sites either are destroyed or completely recovered.
Isolated diagnostic projectile point finds, abundant in the Midwest, are a special property class of great importance for research purposes. Yet they are not considered eligible for inclusion on the NRHP, unless the artifact itself is of exceptional significance. However, groups of culturally related but otherwise isolated Paleoindian remains found in connection with diagnostic land-forms or other paleogeological, geomorphological, or paleoenvironmental contexts may be nominated as contributing properties within a district. That is, isolated finds, taken collectively, may under certain conditions (i.e., high density, significant paleoenvironmental associations) be considered important enough to warrant inclusion as part of an NHL.
Finally, although most other Midwestern Paleoindian sites possess moderate integrity, some might qualify for NHL status individually or as districts. Individual sites of moderate integrity but arguably great significance include Gainey (perhaps the largest Paleoindian excavation in North America), Paleo Crossing, which is reasonably well dated for a site that falls in the moderate-integrity category, Nobles Pond, also extensively excavated and subject to perhaps the most intensive taphonomic efforts of any North American Paleoindian site (Seeman et al. 1994b), Lincoln Hills/Ready, and perhaps Aebischer and Mueller. But some Midwestern sites are linked thematically despite the distance separating them. Silver Mound in western Wisconsin and Coshocton County, Ohio sites like Welling, Honey Run, McConnel and McKibben all are Paleoindian quarry/workshop assemblages. Recognized as individual NHLs, but discussed comparatively in each nomination, the properties might attract research along common lines and produce educational and interpretive results of common interest.
Although NHL designation is not the same thing as NRHP status, any successful NHL nomination will also have to meet NRHP Criteria. NRHP significance Criteria are well established, and only briefly summarized here. Possessing any of the following qualities probably would qualify sites for inclusion on the NRHP:
To these attributes can be added Glassow's (1977) Criteria by which site significance can be assessed, as discussed in this and the national context chapter. The presence of any of the following characteristics tends to automatically make a Midwestern site yielding Paleoindian materials ineligible for inclusion on the NRHP:
Given how rare Paleoindian sites are in the Midwest, full justifications should also be provided in cultural resource management studies detailing why Paleoindian properties cannot yield information important to history or prehistory.
Modoc Rockshelter and Graham Cave already are Midwestern Paleoindian NHLs (Archeological National Historic Landmark Committee 1997:Appendix 3). They are extremely important sites but, as neither are predominantly Paleoindian, do not exhaust the prospects for Midwestern representation among NHLs. Table 23 lists other prospective Midwestern Earliest American NHL properties. Big Eddy clearly possesses high integrity and is perhaps the strongest Midwestern candidate for recognition. Sheriden also might be considered for NHL status. If uncertainty about the age and typological affinities of Chesrow diagnostics can be dispelled, then Chesrow Complex sites like Lucas should receive recognition as an NHL district. Metzig Garden preserves intact late Paleoindian deposits, including features, that surely merit NHL recognition. Indeed, the concentration of late Paleoindian Agate Basin and Eden-Scottsbluff components near Lake Winnebago suggests the wisdom of a thematic regional NHL in that area.
* = Possible NHL candidate
designated an NHL
As elsewhere, Midwestern Paleoindian studies are short on "good, hard dirt archeology" (Collins 1999:40). We should dig more sites. We should survey more widely but also thoroughly, both probabilistically and selectively. When we survey, we should repeatedly visit the same places, so that "findspots" grow into "sites" or collectors' few tools grow into large assemblages. Once we have compiled large assemblages, we must analyze them thoroughly. All Paleoindian scholars know of important assemblages never studied or published. In the meantime, perhaps we might become more circumspect in imposing patterns that we favor or imagine upon a Paleoindian archaeological record of unknown size and character, biased in unknown ways to unknown degrees.