MIDWEST REGIONAL CONTEXT
Who first settled North America and when? Native Americans and later Old World migrants did not conceive of the question, or answered it in ways consistent with their understanding of the world. Columbus failed to pose the question because he failed to recognize the New World. Long ago, Acosta identified America's pioneers as Asian migrants (Huddleston 1967). Scholars have sought to answer the second question using whatever evidence was at hand and in the nineteenth century, flirtation with the prospect of an American Paleolithic gave way to the view that New World human antiquity was short. The Clovis and Folsom discoveries revolutionized understanding of the depth of human antiquity here and archaeologists have since recognized what we call Paleoindians as early and perhaps the first human occupants of the Americas. When radiocarbon dating became common in the 1950s, archaeologists could fix approximate ages to Paleoindian occupation, ages that exceeded 10,000 years. If the New World could not boast the old one's antiquity (nor any reliable evidence of evolutionary ancestors of biologically modern humans), at least it had a respectable one of its own.
Today Paleoindians are among North American archaeology's most popular subjects. This owes partly to their intrepid settlement of the vast hemispheric land mass, partly to the undoubted technical virtuosity that Paleoindians displayed in their stone-tool industries, partly to the role they may have played in the demise of Pleistocene megafauna, but perhaps to their sheer antiquity more than any other reason. As archaeology has matured, so too have Paleoindian studies, which now have surpassed preoccupation with strict chronological questions though these remain urgent issues. Some believe that New World human antiquity exceeds the time range traditionally ascribed to Paleoindians: they believe that Paleoindians are not the oldest people of the New World. Others hold instead that Native American Indians' ancestors always have been here.
In these unsettled and contested times, the National Park Service and the Society for American Archaeology have undertaken the Earliest Americans Theme Study "to identify, evaluate, and nominate archeological properties associated with the initial peopling of America" (Archaeological National Historic Landmark Committee 1997:7) as National Historic Landmarks (NHLs). The goals include systematic compilation of archaeological evidence of earliest occupation, updating earlier records where circumstances warrant, and dissemination of results to scholarly, native and general audiences (ibid).
The United States is sufficiently large to justify subdivision and concentration of effort. This document is a critical synthesis of Paleoindian archaeology in the American Midwest. For its purposes, the Midwest encompasses Illinois, Indiana, Iowa, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. Its southern boundaries are natural to a point, the Ohio River to Pennsylvania. Its northern boundary is the Canadian border along the Great Lakes. So defined, the Midwest includes the upper Mississippi Valley to its major tributaries, and the American portion of the Great Lakes basin. It grades to plains-prairie climate and biota to the west; farther east it borders and partly encompasses the Appalachian Plateau, and resembles the climate and biota of the east. It is hackneyed to call the Midwest the crossroads of North America, but the region spans major biotic, climatic, and modern cultural boundaries. It also spans important prehistoric cultural boundaries.
Such study areas are defined by modern political boundaries that are arbitrary with respect to drainage, climate and biota and to the cultural units (territories, ranges) defined by other earlier occupants. Although Paleoindian cultures cannot be reduced to their natural settings, we reason, legitimately, that natural features and prehistoric cultural boundaries bear on our subject, and because any boundaries drawn in our current state of ignorance are apt to be arbitrary; they may as well be modern political ones.
While the American Midwest lacks the remoteness in space and national culture that makes regions like the Southwest redolent of earlier times, the Midwest witnessed the entire span of North American prehistory. By common measures, it boasts an archaeological record as abundant and diverse as any other area. More Paleoindian sites are recorded, more fluted bifaces known, in the Midwest than in most other areas. Today the Midwest has more farmland than any other region, and cultivated farmland is exposed land that is well suited to archaeological study. Thus, the Midwest perhaps has the best combination of evidence and availability of any American region. Here the Paleoindian record can be well documented and can answer the questions about the past that continue to occupy us. As Paleoindian research continues, the Midwest surely will play a critical role in the process.
Ellis and Deller (1997:2) reluctantly called "phases" the basic time-space units of Great Lakes Paleoindian archaeology, taking pains to equate these with "archaeological culture." "Phase" and other systematic terms derive from McKern's taxonomy (1939), whose legacy, for better or worse, persists. There is broad agreement that components are recognizable distinct occupations of "sites" and that patterns like Paleoindian are the broadest time-space units. By McKern's definition, the intermediate units differed only by degree of similarity in assemblage or trait-list composition (Trigger 1989:190), and each higher unit vaguely covered larger time-space scales.
Under the heading of "Basic Archaeological Units" Willey and Phillips (1958) identified nothing between component and phase, defined as per McKern. They (1958:22) defined phase as a "unit possessing traits sufficiently characteristic to distinguish it from [similar] units…spatially limited to the order of magnitude of a locality or region, and chronologically limited to a relatively brief interval of time." In effect, phases are constructed empirically from components, and are cultural units with time-space integrity for analytical purposes. Willey and Phillips (ibid) also considered phase the most "practicable and intelligible unit of archaeological study."
As theory develops and fieldwork progresses, of course, phase definitions can change in their meaning and scale. But we speak legitimately of the Gainey or Parkhill Phases. Archaeologists sometimes use "complex" in favor of "phase"; neither McKern nor Willey and Phillips used this term. Kidder et al.'s definition of "complex" (quoted in Overstreet 1991a:209) implies temporal integrity no less than "phase."
Chronology is archaeology's chief preoccupation and prerequisite. The Earliest Americans Theme Study recognizes three major Paleoindian intervals, a possible pre-Clovis one, and a post-Paleoindian one equivalent to what most archaeologists call the Early Archaic Period. Midwestern Paleoindian chronology rests on radiocarbon (and, secondarily, thermoluminescent) dating, typological cross-dating with well dated assemblages and industries elsewhere, geochronology and seriation. Despite the many sources of evidence and inference, Midwestern Paleoindian chronology is highly imperfect, to say the least. Few doubt that Paleoindian cultures arose more than 11,000 rcbp and endured at least for centuries. But many continue to doubt that Paleoindians were the first people in the New World, instead claiming that they were preceded by ancestral cultures reaching back perhaps millennia. Thus, the antiquity of first human occupation must be distinguished from Paleoindian antiquity.
Pre-Clovis Claims in the Midwest. Paleoindian archaeology long has been enmeshed in claims for pre-Clovis human occupation of North America. The quality of most such claims is dubious, their staying power slight; the "shelf life" of pre-Clovis claims hovers around 10 years (Meltzer 1995:22). Midwestern places and assemblages have not figured prominently in such debates. Among Meltzer's (1995) sources published between 1964 and 1990 that advocated pre-Clovis claims, only the "Imlay Complex" of Michigan's Thumb (Baggerly 1954), championed by Krieger (1964:35, Fig. 1), lay in the Midwest. Alas, the "Complex" consisted of glacially worked but naturally fractured chert fragments; it soon lapsed into deserved obscurity. Fiedel's (2000) recent synthesis would add only the Big Eddy site, discussed below.
Rowlett (1981) reported a small assemblage of possible flake tools stratified beneath a fluted biface at Shriver in northwestern Missouri. Geochronology and thermoluminescence dating suggested that the assemblage predates 13,000 B.P., but Shriver is not radiocarbon dated, nor are all of the illustrated specimens (Rowlett 1981:Fig. 5) obviously tools or even flakes (cf. Lepper 1999:367). Moreover, the ostensibly pre-Clovis artifacts and context "may in fact relate to the Early Paleoindian…component" (J. Morrow 1996:91).
Near the Upper Mercer sources in Coshocton County, Ohio, Eppley Rockshelter returned a 12,185±130 B.P. date (UCLA-2589C) (Brush 1993:6; see also Brush 1990:238) from feature charcoal. The feature contained no artifacts, but an end scraper was found nearby. A second sample from the same feature returned a considerably more recent date of 9,890±100 B.P. (UCLA-2859E). The latter may owe to organic contamination after excavation (Brush 1993:6), but on available evidence the earlier date cannot be accepted without question.
Southeastern Wisconsin's Chesrow Complex is a candidate for pre-Clovis, or at least early Paleoindian, occupation of the Midwest. Partly on typological grounds, Overstreet (1993) suggested that it preceded Clovis or regional middle Paleoindian taxa. Chesrow bifaces indeed are "highly variable" (Overstreet 1993:62) in technology and typology, but they resemble best late Paleoindian basally thinned, small lanceolate bifaces like Quad and Hi-Lo. Specimens (Overstreet 1993:62, Pl. I-VII) are lanceolate with edge and base grinding that varies in extent and degree, basally thinned or fluted, usually accomplished in several flake removals, sometimes shouldered or notched, or bearing basal ears. Most are made of local cherts, and some have collateral flaking. The Chesrow report is commendably well illustrated, and includes specimens from several nearby assemblages. Judging from their appearance and dimensions (Overstreet 1993:Table 3), these are not Holcombe bifaces (cf. Overstreet 1993:77). They seem much more like Hi-Lo ones, which themselves show "considerable morphological variation" (Ellis and Deller 1982:7), although Hi-Los have more pronounced shoulders. On haft elements, Hi-Los are ground, thinned to fluted, sometimes eared and sometimes laterally thinned below the shoulders (Ellis and Deller 1982:10); many Chesrow bifaces possess these attributes, including lateral thinning (Overstreet 1993:Pl. VIId, j).
On available evidence, Chesrow assemblages are few in number, limited in distribution, and documented chiefly from uncontrolled surface collection. Save distribution, they share these properties with the other components of the Midwestern Paleoindian record. Chesrow assemblages are confined to southeastern Wisconsin (Overstreet 1993:Fig. 3). Their association with proboscideans is inferred, not direct, and their typological affinities lie with late Paleoindian industries. This need not exclude association with mastodonts, but by Overstreet and Stafford's (1997:70) reckoning, would postdate mammoths. Local proboscidean remains with associated tool are mammoths.
Big Eddy near Rodgers Shelter in southwestern Missouri is perhaps the Midwest's most promising deeply stratified site from which the timing and sequence of Paleoindian and possible pre-Clovis occupations might be established (Lopinot et al. 1998; Ray et al. 1998). It has yielded acceptable dates for Dalton/San Patrice and for early and middle Paleoindian occupations (N. Lopinot personal communication, 2000; Lopinot et al. 1998:Table 1; Ray et al. 1998:77), which overlie what the excavators identify as a Gainey occupation. Below the latter is a small cobble and flake assemblage that itself overlies charcoal dated to 12,940±120 B.P. (B-109008) (Lopinot et al. 1998:219-220; Ray et al. 1998:80). There also are more recent dates from the same contexts, so the integrity of dated samples and the age of the deposit remain uncertain. Indeed, a recently obtained date (AA-35460) from slightly deeper than B-109008 is slightly more recent than the radiocarbon age of that sample, and another deeper one (AA-29021) is considerably more recent (Lopinot personal communication, 2000; Lopinot et al. 1998:Table 7.1). There is "a good possibility" (Lopinot et al. 1998:220) that an intact, pre-Clovis deposit exists at Big Eddy. Until more is known about the site, this evidence is suggestive but not conclusive.
Radiocarbon Chronology. The western and eastern North American Paleoindian records differ in important respects. The west has comparatively few sites that possess contextual integrity, the east many sites that possess little integrity. C. Haynes (1993) concluded that most Clovis sites date to between roughly 11,500 and 10,800 rcbp, most Folsom ones to between roughly 10,900 and 10,300 rcbp. In a recent revision that added some assays and removed others, R. Taylor et al. (1996) were able to separate Clovis and Folsom occupation somewhat better (see especially 1996:Fig. 7).
Radiocarbon dates are reported from many fewer eastern Paleoindian sites than western ones (Ellis et al. 1998:Table 1; Fiedel 1999:Fig. 5). Omitting sites whose dated samples may not be associated with the Paleoindian occupation (e.g., Dutchess Quarry Caves, Michaud), C. Haynes (1993:223) found that eastern sites occurred between roughly 10,600 and 10,200 rcbp, although some individual radiocarbon dates are considerably older than this range (e.g., from Whipple). Although Folsom bifaces are rare in the east, C. Haynes (1993:223) judged eastern Paleoindian occupation to be "penecontemporaneous with Folsom sites in the west", a judgment consistent with then available evidence. If it is correct that Paleoindians arrived in the eastern U.S. from the west, and if it is correct (as existing evidence would suggest) that Paleoindian occupations of the west do not antedate 11500 rcbp, then eastern Paleoindian sites should not pre-date 11500 rcbp and perhaps should be no older than 11300 or even 10500 rcbp. The validity of the logical statement is unimpeachable but as a scientific statement it is merely one unverified hypothesis among many other possible hypotheses.
In southwestern Missouri, Big Eddy is a deeply stratified deposit distinguished most by late Pleistocene and early Holocene deposits. Between its possible pre-Clovis deposits and its certain early Holocene ones, Lopinot et al. (1998; see also Ray et al. 1998) reported a suite of dates, and several others have been obtained more recently (N. Lopinot, personal communication 2000). Dates generally run consistently with depth (Lopinot et al. 1998:Table 7.1). There is Middle Paleoindian occupation at Big Eddy, represented by what the excavators identify as Gainey bifaces and one Folsom specimen. Most are out of context (Lopinot et al. 1998:212) but a fluted biface base and possible midsection (Lopinot et al. 1998:Fig. 8.48d, e) were recovered in situ, and several other likely Paleoindian tools were found at similar depths (Lopinot et al. 1998:215). To judge from the illustration and description, the Gainey assignment seems reasonable.
Big Eddy's Middle Paleoindian strata are dated by a suite of radiocarbon assays. Alas, within those strata the dates do not pattern clearly with depth (Table 8) (Lopinot, personal communication 2000; Lopinot et al. 1998:Table 7.1, 218) even if dates from the entire stratigraphic sequence are reasonably concordant. Three dated samples were located within 3 cm of the Gainey biface; Lopinot et al. (1998:218) associated the specimen with only one, 10,710±85 B.P., which does not nearly overlap at two δ with either the older or younger nearby date but apparently is the closest horizontally to the biface (Lopinot et al. 1998:Table 7.1).
Northeastern Ohio's Paleo Crossing (Brose 1994) is a Gainey-affinity occupation (Barrish 1995) quite similar to one reported earlier from Nobles Pond (Seeman et al. 1994b) a short distance southeast. One feature there yielded radiocarbon results that suggest, at face value, a four-millennium span between its filling and sealing, a span that brackets the Paleoindian interval. It does not, however, include any dates that lie in that interval. A second Paleo Crossing feature identifed as a post mold yielded a result of 12,250±100 B.P. (AA8250; Brose 1994:63) that lies closer to the acceepted Paleoindian interval but still is early (Brose 1994:64-65). Individual charcoal fragments from this post mold were assayed (AA8250B-F) to produce two distinct populations (Brose 1994:65). Samples AA-8250B and AA-8250F are indistinguishable statistically and yield an average date of 12,083±105 B.P. using Stuiver and Reimer's (1993) CALIB 3.0.3 program, which again is early for most Paleoindian archaeologists to accept. However, samples AA8250C-E produced an average of 10,979±81 B.P. (Using a different averaging technique, Brose [1994:65] reported an average that differs only trivially.) This Paleo Crossing date conforms to current understandings of eastern North America Paleoindian chronology and stands as the evidence for the age of Paleoindian occupation in the Great Lakes.
Even this conclusion, however, must be qualified by the observation that the dated context produced populations of charcoal of two statistically distinct ages, such that natural agents may have disturbed the original context and introduced either older or younger charcoal. The averaged 10,979±81 post mold date calibrates to the surprisingly old 12,989(12,897)12,805 B.P. at one sigma using Stuiver and Reimer's (1993) bidecadel data set. Stuiver and Reimer (1993:8-9) recommend this data set "for most non-marine samples" but it extends only to 9,840 cal B.P. or 11,440 cal B.P., depending on the dendrochronological sequence used. It is extended to reach and surpass the dates reported for Paleo Crossing only by statistical modeling of coral, not tree-ring, data (Edwards et al. 1993; Stuiver and Reimer 1993:0[sic]). Few if any eastern North American Paleoindian dates have been reported using modern calibrations, and the Paleo Crossing result is best reported in radiocarbon years to facilitate comparison to other sites.
About one hundred kilometers west, in north-central Ohio, Sheriden Cave has yielded many radiocarbon dates on archaeological and paleontological samples (Holman 1997; McDonald 1994; Tankersley 1999, personal communication 2000) (Table 8). Pending revision of stratigraphic interpretation and reporting of newly acquired dates (Tankersley, personal communication 2000), Sheriden's cultural assemblage includes one Paleoindian biface identified as Holcombe in type and a worked-bone object, possibly a bone point (Tankersley 1999:70). Among Unit IIb's three populations of radiocarbon dates, Tankersley (1999:71) favored the 11,200 yr B.P. average of three. The biface was found in underlying Unit IIc, so should antedate that figure. That unit also apparently contains three populations among its set of at least nine dates, even the most recent of which is surprisingly old for inferred Holcombe occupation. But Tankersley (1999:71) noted that Holcombe was not yet dated directly. Fluoride relative dates agree with radiocarbon ones (Tankersley 1999:71-72).
Sheriden's faunal assemblage is large and diverse; detailed description of its association with cultural remains is eagerly awaited, although at least some dated faunal samples suggest butchering and hence direct human association (Tankersley, personal communication 2000). Nevertheless, association must be demonstrated on archaeological and taphonomic grounds. Cultural deposits seem bracketed between ca. 11,200-10,600 rcbp. They may be associated with caribou and Holocene species, but also with extinct Pleistocene species like giant beaver (Castoroides ohioensis) and peccary (Platygonus spp., Mylohyus spp.). Upon full reporting, Sheriden should be of great significance to Midwestern Paleoindian chronology, paleoenvironmental inference, and the study of human agency in Pleistocene extinctions.
Two western New York State sites yield radiocarbon dates relevant to Midwestern chronology. Arc returned dates from several strata which bracket the Gainey-affiliate Paleoindian occupation of the site and dates at between 11,700 and 10,300 rcbp (Ellis et al 1998:158). A suite of radiocarbon dates on bone/ivory and plants from Hiscock fall between 11,400 and 10,200 rcbp (Laub et al 1996; Tankersley et al 1998: Table 2). The Hiscock site’s dates on mineralized bone generally agree with radiocarbon dates there and at Big Eddy, Paleo Crossing and Sheriden sites as well.
When sites have only one or few radiocarbon dates, the accuracy of individual assays and the details of occupational history cannot be gauged (Bird and Frankel 1991:179-180). Big Eddy and Sheriden stand out for the few aberrant dates among them. Radiocarbon dating is inherently probabilistic (Shott 1992a) and fraught with problems of association and calibration. In the interests of full reporting and to aid interpretation of Paleoindian chronology, we should report unaccepted radiocarbon results. Gainey, for instance, returned a date of 2830±175 B.P. on a pooled sample that evidently contained at least some charcoal more recent than the Paleoindian occupation (DIC-1564) (Simons et al. 1984:22). Unacceptably recent dates also were obtained at the nearby Parkhill Phase Leavitt site (Shott 1993:21), this time using the AMS technique with individual charcoal pieces identified as spruce.
Table 9 compiles unaccepted dates from several important Midwestern and Ontario Paleoindian sites. It excludes dates earlier than early or middle Paleoindian times from Big Eddy and Eppley. Paleo Crossing dates are excluded even though charcoal from a single probable postmold yielded two statistical populations circa 12,150 and 10,980 rcbp and assays from the lowest and uppermost stratum of another feature are statistically distinct from one another and the postmold dates (Brose 1994:63, 65). Paleo Crossing is geologically complex, yet no assay there clearly is aberrant. Intrusion, ambiguous association, or dating errors combine to explain the dates shown in Table 9. The nine assays there are gratifyingly few compared to the 13-15 (two are late Paleoindian in age) apparently reliable dates from Big Eddy, the 24-26 (two are pre-12,000 rcbp) from Sheriden, and Paleo Crossing's date. Accepting all Big Eddy, Sheriden and Paleo Crossing dates gives a figure of 42 acceptable dates against Table 9's nine unaccepted ones, a better average than radiocarbon ordinarily achieves (Buck et al. 1994:232). Yet it is possible that unacceptable dates from other Paleoindian sites have not been published. AMS dates seem disproportionately represented in Table 9; for Leavitt at least, AMS assays were obtained while the method still was somewhat untested.
Table 9: Unaccepted Dates from Midwestern and Nearby Paleoindian Sites.[Long description]
Thermoluminescence (TL) dating is less precise than is radiocarbon dating in the Paleoindian time range (Aitken 1990:141; Dunnell and Feathers 1994:116) and it is not used often in North America. TL dates of ca. 12,400 B.P. and 11,400 B.P. were obtained from burned chert samples at Gainey. Both were reported with very large measures of uncertainty. (Perhaps the only other eastern North American TL-dated specimen is a fluted biface of uncertain typological affinity from Massachusetts' Wapanucket Locus 8 [Robbins 1980:Plate 22C] TL-dated to 9,000 B.P. [Robbins 1980:281, 290] from a context of equally uncertain age in which Paleoindian and Archaic deposits are mixed.) For the Midwest, we can do no better than to echo Allen's (1994:342) judgment that "Obtaining paired samples of luminescence dates and radiocarbon determinations from diverse sites where reliable samples for both technqiues occur…is a major priority." Given the poor resolution of the radiocarbon timescale during the period, Paleoindian sites which yield artifacts susceptible to thermoluminescent dating and which have samples appropriate for radiocarbon dating are likely to be of national significance in their ability to improve these techniques and correlate their applications.
Geochronology. The Midwest's Late Pleistocene history is well known, its glacial deposits reasonably well ordered and dated. Archeologists use this knowledge to infer maximum and minimum ages of archaeological deposits, and the gross patterning of these archaeological data on dated geological features illuminates some of the temporal and topographic limits to patterns of human activity.
Mason (1958, 1962) saw that Michigan fluted bifaces were confined to the southern half of the lower peninsula. He correlated the distributions with late Pleistocene glacial features, concluding that Paleoindians entered Michigan no earlier than approximately 12,000 B.P., when the ice front stood at the approximate northern boundary of the fluted-biface distribution. Quimby (1958) then noted similar patterning in the distribution of Late Pleistocene fauna, and the Mason-Quimby line was born. Since Mason’s work, the northern limit of fluted-biface distribution has crept northward (Cleland et al. 1998).
Following Mason, similar studies appeared for Ohio (Prufer and Baby 1963; see also Seeman and Prufer 1983), and Indiana (Dorwin 1966). Studies of smaller scope concerned parts of southwestern lower Michigan (Peru 1965, 1967) and eastern Illinois (Henry and Nichols 1963). To the south, the latest Pleistocene features and the earliest human presence diverge in time. The validity and precision of chronological inferences in early studies are less important than the fact that they were reasonable efforts to ground Paleoindian chronology, and that they demonstrated some interpretive value to apparent patterning in the distribution of Paleoindian remains.
Based on the association of sites with fossil standlines of known age most archaeologists agree that Great Lakes Paleoindian occupation occurred between roughly 11,000 and 10,000 B.P. (Ellis and Deller 1997). Although data remain limited, regional specialists subdivide this interval into three successive phases—Gainey, Parkhill, and Crowfield—each marked by a diagnostic fluted-biface type and, to some extent, diagnostic tool kits. Gainey Phase occupation is thought to be the earliest, and to date to ca. 11,000 B.P. with an uncertainty of several centuries.
Paleoindian typology largely involves fluted bifaces and archaeologists doubtless invest more thought in fluting technology that its ancient practitioners did. Barrish (1995:40-53) and J. Morrow (1996:159-167) described the history of Midwestern Paleoindian typological studies, yet the established typological tradition has not solved our problems. As much today as thirty years ago "Formidable obstacles block the path of anyone attempting to assign specific type designations to individual fluted points" (Stoltman and Workman 1969:193).
Midwestern archaeologists commonly lament the "uncleanliness" of typological results as though it were an undesirable flaw. To J. Morrow (1996:158) we lack "clearly defined, quantitative measures of segregating the different [fluted biface] types." There remains much disagreement about what defines a type and whether individual fluted bifaces belong to one type or another (e.g., Barrish 1995:19-20; Lepper 1999:372). Compare Howard (1988) to J. Morrow (1996:175-186) on what Clovis is and whether Ready site bifaces are Clovis.
Whether we reveal or construct types, typological approaches assume that all change is discrete. Where we draw the line between successive types in a time series of continuous variation is an arbitrary act. Where we seek continuous, not discrete patterns of variation obliterated in typological approaches, we create many transitional specimens that resist classification.
Paleoindian typology derives from the western taxonomic units Clovis and Folsom. As in other respects, the west exerts influence for largely historical reasons: it is where fluted points first were found in association with megafauna. Midwestern archaeologists recognize the following fluted-biface types, arranged in (possibly overlapping) order by time.
To J. Morrow and T. Morrow (1996:18-19), Clovis bifaces are the product of transverse percussion flaking from isolated platforms. Traditionally, archaeologists consider Clovis fluting to occur from bases steeply beveled to one face, but lacking prepared "nipple" platforms. J. Morrow and T. Morrow (1996) argued that Clovis fluting instead was from centrally bevelled bases that possessed nipples. Collins (1999:Fig. 3.3b-d) illustrated Clovis preforms from Texas that were beveled to one face and that lacked prepared platforms. Basal concavity is shallow, and the resulting bifaces are thick. Apparently, Clovis bifaces were fluted near the middle, not end, of the production process. Gainey bifaces are the product of medial percussion flaking. They were fluted by indirect percussion from bases beveled to one face, then the other, and also with prepared platforms. Fluting occurred late in the production process. Resulting concavities are deeper and Gainey bifaces are thinner overall. J. Morrow and T. Morrow's study was admirably detailed. As a minor criticism, the isolated lateral-margin platforms left to aid thinning and finishing that they identified with Clovis characterizes Ontario Parkhill assemblages (e.g., Deller and Ellis 1992:31-34).
Clovis bifaces (Barrish 1995:29-30; Howard 1990; J. Morrow 1995; J. Morrow and T. Morrow 1996) are comparatively large, wide, thick and therefore heavy. Sides are subparallel to slightly excurvate. Haft margins and modestly concave bases are heavily ground. Maximum width position (distance from the base to the point of maximum width) is near or below the mid-point. Fluting is from prepared bases, either simply beveled (Howard 1990:255) or with a prepared platform (J. Morrow 1995:171). Fluting usually is on both faces, can be multiple (Howard 1990:255) or single, and usually is fairly wide and short. Fluting may occur at an intermediate, not late, production stage (J. Morrow 1995:171; J. Morrow and T. Morrow 1996).
Gainey bifaces (Barrish 1995; Ellis and Deller 1997; J. Morrow 1996; J. Morrow and T. Morrow 1996; Shott 1986b; Simons et al. 1984) also are relatively large, wide and thick, with subparallel sides, and ground haft margins and modestly concave bases. Basal concavity is arched (Ellis and Deller 1997:2). Usually single, relatively short and wide flutes occupy both faces. Flutes may have been guided by fairly prominent median ridges produced by preform thinning on both faces. Basal finishing is by the Barnes technique (Ellis and Deller 1997:Table 1; J. Morrow 1996:176). Earlier-stage preforms and unfinished fluted preforms are rare at the Gainey type site and in most other assemblages, so fluting production details are poorly understood (J. Morrow and T. Morrow 1996). Gainey in the Midwest is roughly equivalent to Bull Brook in the northeast (J. Morrow 1996:175; Shott 1986b:111) but perhaps also Vail and Debert there (J. Morrow 1996:184).
Clovis and Gainey. However much they differ in typological methods and conclusions, most Paleoindianists agree that Clovis and Gainey are best distinguished as types or modes (Barrish 1995:55; Howard 1990: J. Morrow 1996:193-194; J. Morrow and T. Morrow 1996; Stoltman 1993:58). Alas, they disagree on how and why the modes differ. To Barrish (1995) Clovis is more selective of toolstone and more sparsely distributed in fewer sites than is Gainey, but is difficult to distinguish on technological grounds. To J. Morrow (1996; see also T. Morrow 1996b:3; J. Morrow and T. Morrow 1996), the key difference is in fluting technology, where Gainey follows Folsom techniques of fluting from prepared platforms, yet some typological disputes persist. Certainly in respects besides fluting, Gainey and Folsom bifaces differ very much. Ultimately following Roosa (1965), some argue that Clovis simply is not found in the Midwest (Howard 1990; Stoltman 1993). This makes major Midwestern Paleoindian assemblages near St. Louis (e.g., Ready, Martens) Gainey, not Clovis, in affinity. J. Morrow (1996) is the chief advocate of the opposing view that Clovis occurred in the Midwest, and that St. Louis-area sites are its chief expression.
Parkhill Phase Barnes bifaces are thinner and narrower, not necessarily shorter, than Clovis and Gainey ones (Ellis and Deller 1992, 1997; Shott 1986b). Margins are excurvate and bases eared, so points have a fishtailed form. Basal concavities are modest to deep and straight rather than arched. Barnes bifaces are sufficiently similar to Cumberlands to be considered roughly equivalent. Barnes bifaces are more common in the Great Lakes, Cumberlands in the Ohio Valley and points south (Justice 1987:Map 6). Similar points extend east into the Northeast (Dincauze, this volume).
Folsom bifaces (J. Morrow 1996:177; T. Morrow 1996b; Munson 1990) are shorter and thinner than Gainey bifaces, have deeper basal concavities, straight to contracting margins, and long, wide, usually single flutes that occupy most of both faces. Ahler and Geib (in press) identified a thin, uniform transverse section from tip toward base as a key Folsom attribute. Munson (1990) plotted the distribution of Folsom bifaces across the region, confining it largely to the western half of the Midwest; he correlated it with the Prairie Peninsula (see also Justice 1987:Map 7). Folsom traditionally has been associated with bison hunting, a judgment strengthened by Ahler and Geib's (in press) argument that Folsom bifaces were highly specialized, renewable bison-hunting points. As a generalization subject to revision, Folsom indeed seems most abundant in the western Midwest, especially in western Iowa (Billeck 1999; T. Morrow 1996b) and rare further east.
Crowfield bifaces (Ellis and Deller 1984:97, 1997:3) are wide and remarkably thin, with contracting margins, and arched basal concavities and often sharp, wide or obtuse-angled tips. Individual flutes vary in length and especially width, and both faces tend to have several flutes. Crowfield bifaces often are pentagonal (sometimes called 'pumpkin-seed') in outline form, probably an incidental consequence of their near-constant wide tip angle, maintained in resharpening by oblique faceting (Ellis and Deller 1997:5), and their near-constant corner angles. Crowfields are best known and most abundant at the eponymous site, but occur elsewhere. Significantly, Ellis and Deller's (1997:Table 2) regional compendium identified Crowfield assemblages principally in Ontario, although they encompassed Vermont's Reagen (Ritchie 1953, DD:this volume) site as well. No sites were identified in the Midwest, although Prufer and Baby (1963:Fig. 4) illustrated one apparent Crowfield biface from Ohio. Surely they occur there and elsewhere but the distribution does seem centered on Ontario.
Holcombe bifaces are best known and most abundant at the type site and nearby locations of southeastern Michigan (Fitting et al. 1966). Usually composed of Bayport chert, Holcombe bifaces are relatively narrow, small lanceolates whose basal treatment ranges from thinning to multiple fluting (Fitting et al. 1966:Figs. 6-7). Many specimens apparently are preforms and many finished ones are proximal fragments, but as best can be determined flutes, when present, are not especially long. To judge from illustrated specimens (Fitting et al. 1966:Fig. 6o), at least some Holcombe bifaces were fluted from prepared platforms. Although Holcombe bifaces are widely distributed at least in southeastern Michigan, the type's full range is unknown. Justice (1987:Map 5) confined it to the Great Lakes basin. Prufer and Baby (1963:Fig. 1) showed a possible Holcombe biface from Ohio, but Seeman and Prufer's later update excluded the Holcombe type (1983:157). McKibben may be a Holcombe assemblage, but illustrated specimens do not confirm the assignment (cf. Prufer and Sofsky 1965:12, Fig. 1a-h). Holcombe does not seem be be common in adjacent Ontario, where perhaps coeval technologically different Crowfield industries occur. In Ohio, Sheriden's unillustrated Holcombe biface (Tankersley 1999:70) may be as early as 10,900 rcbp. The Bostrom site in Illinois may partly be a Holcombe assemblage but likelier is earlier (Tankersley 1995; cf. Koldehoff 1999:2, J. Morrow 1996:108).
In the Midwest, Clovis and Gainey bifaces tend to be made on what are considered good toolstones (the highly variable Burlington for Clovis, Upper Mercer or Moline for Gainey), Folsom, Barnes and Crowfield bifaces on a wider variety of often local cherts. Thus, the earliest fluted types might be defined by toolstone as much as by size, form and technology. Patterns in archaeological distribution of toolstones are also used to infer settlement mobility, range size and other practices that might define phases.
Late Paleoindian Types
Dalton bifaces are late Paleoindian forms whose basal treatment ranges from thinning to short fluting. Basal curvature is pronounced. In plan form, Daltons are lanceolate but with somewhat expanding stem margins, usually have weak shoulders, and almost always have well defined corners that approximate ears. Dalton bifaces often are finely serrated. The Dalton horizon dates to ca. 10,500-10,000 rcbp (Goodyear 1982). O'Brien and Wood (1998:80) would extend its lower boundary to 10,900 rcbp. Walthall and Koldehoff (1998:259) described the distribution of Dalton assemblages, which seem most abundant by far in the central Mississippi Valley; northeast Arkansas in particular boasts many Dalton assemblages. Yet Daltons occur to some extent in the Midwest, especially in its western portion. There are many in Missouri, especially south of the Missouri River (Chapman 1975:98-125; O'Brien and Wood 1998:73). Sizable Dalton assemblages were found at Rodgers Shelter (Kay 1982) and at nearby Montgomery (Collins et al. 1983:30-31). Indeed, Rodgers and Graham Cave (cf. O'Brien and Wood 1998:76-79) were instrumental in dating the Dalton horizon. Along with the typologically related and perhaps coeval San Patrice type, Dalton bifaces and other tools were found at Big Eddy (Lopinot et al. 1998:163-174, Figs. 8.34, 8.35, 8.37). Associated radiocarbon dates there (ibid: Table 7.1) cover the Dalton span noted above. The existence of at least 59 Dalton sites in four relatively small sections of southwestern Illinois also hints at considerable occupation (Walthall and Koldehoff 1999:28). Dalton assemblages are distinguished not just by Dalton bifaces qua points but also by large, thick, oval to subrectangular beveled bifaces that probably functioned as adzes (Morse and Goodyear 1973).
Among other Late Paleoindian types, Agate Basin bifaces seem to fall between 10,500-10,000 rcbp, Hell Gap bifaces between 10,000-9,500 rcbp, and Cody-Complex Eden-Scottsbluff specimens between 9,400-9,000 rcbp (Buckmaster and Paquette 1996:38-44). Broadly consistent with this view, Salzer's (1974:43) northern Wisconsin Flambeau Phase of Agate Basin affinity originally was dated to ca. 10,000-9,300 rcbp. Yet Salzer (1974:44-45) assigned his Minoqua Phase of Cody affinity to a subsequent Early Archaic time range. Radiocarbon and stratigraphic confirmation of these assignments remain elusive (Mason 1997:105-106), and a date of 9410±110 on Eden-Scottsbluff material from Maine (Petersen 1995:211) seems congruent more with Buckmaster and Paquette's's than Salzer's Eden-Scottsbluff chronology. If this chronology is correct, then Agate Basin and Dalton are contemporaneous if not sympatric in distribution (O'Brien and Wood 1998:86). At Metzig Garden in eastern Wisconsin an Eden-Scottsbluff occupation overlaid possibly an Agate Basin/Plainview one (Behm 1986:3, Fig. 5), supporting the chronological relationship between these taxa. Agate Basin material is comparatively sparse in the midcontinent. O'Brien and Wood (1998:86-87) attributed few known Missouri bifaces to the Agate Basin taxon, and also questioned the Agate Basin assignment of Cherokee Sewer specimens from Iowa (Anderson and Semken 1980). Fishel (1988) plotted Agate Basin's distribution chiefly in the western half of the Midwest, roughly comparable to the Folsom distribution (Munson 1990) and overlapping with Dalton (cf. Justice [1987:Map 9], who extended the distribution much further east). Apparent Agate Basin locations include Patrow in northern Minnesota (Neumann and Johnson 1979:Fig. 6d), Cherokee in northwestern Iowa (Anderson and Semken 1980), northwestern lower Michigan's Samels Field (Cleland and Ruggles 1996), central Illinois (Luchterhand 1970:Fig. 13; Munson and Downs 1968:125-126), near Marquette in upper Michigan (Buckmaster and Paquette 1996) and across central and northern Ohio (Prufer and Baby 1963:20, Figs. 10, 20). Tomak (1994:120, Figs. 7.4-7.5) reported Agate Basin bifaces from southern Indiana's Alton, but illustrated specimens seem ambiguous. Cody Complex material has been found at Renier (Mason and Mason 1960) on Wisconsin's Door Peninsula, Pope in Wisconsin (Ritzenthaler 1972), and at Gorto (Buckmaster and Paquette 1988) in Michigan's nearby upper peninsula. All three assemblages apparently are from cremation mortuary contexts. Other finds near Gorto (Clark 1989), at Patrow (Neumann and Johnson 1979:Fig. 6c), in northern Wisconsin (Behm 1986; Clark 1982:Figs. 1-2; Kuehn 1998:464; Mason 1985) and at Sawmill in Ohio (Prufer and Baby 1963:32-33) document a wider range of Cody occupation. Metzig Garden is particularly significant because it contains undisturbed sub-plowzone deposits that include as many as 18 late Paleoindian features (Behm 1986:4). Alex (1980:114) reported one Eden-Scottsbluff biface from excavation context in Iowa.
The fluted and late Paleoindian point typology and the time series based upon it are poorly grounded in radiocarbon chronology. What grounding there is comes from sites in the west and, increasingly, the northeast. Only Paleo Crossing, Sheriden and Big Eddy among major assemblages with clear typological affinities are dated, and then not without questions, by radiocarbon.
Other Phase Diagnostics
Blade Technology. Green's 1963 report from Blackwater Draw made prismatic blades and polyhedral cores further Clovis diagnostics. In North America, Clovis blade technology is found most in the southern plains. Collins' (1999; see also Parry 1994) recent assay of blade technology rightly abandoned the formal definition of blades by arbitrary values of length-width ratios for one that emphasizes technology over product. Blade technology included selection for isotropic chert, core preparation and maintenance, and soft-hammer or indirect percussion. It made efficient use of stone and yielded sharp, relatively uniform products capable of modification into a range of finished tools. Perhaps a better definition of blades than the traditional one is that they are flakes struck from prepared conical or wedge-shaped polyhedral cores, whose length is only weakly constrained by their width and thickness. Blades possessed small platforms, thin percussion bulbs, smooth interior surfaces, few, subparallel exterior facets normal to the longitudinal axis, and often curvature.
Collins (1999) described the time-space distribution of blade technology. He saw little evidence in midwestern assemblages, but considered blades from and near Meadowcroft possible Clovis antecedents. This view does not account for the persistent doubts about Meadowcroft's age nor its small assemblage's affinities to late, not early, Paleoindian industries. As above, Parry (1994) considered Meadowcroft specimens to be biface thinning flakes, not true blades at all. Judging from illustrations can be problematic, but blades and polyhedral cores may occur in several Midwestern assemblages, chiefly from the St. Louis area (e.g., Martens [J. Morrow 1996:Figs. 71-76], Faust [Koldehoff 1994:Fig. 3], and perhaps Mueller [Koldehoff 1977:27]). A possible blade core and blades were found in situ in buried deposits at Big Eddy (Lopinot et al. 1998:212-214, Fig. 8.49e, k). T. Morrow (1996a) reported a Clovis blade from southeastern Iowa. A few blades may pass unnoticed in large assemblages, and only recently have many midwestern archaeologists looked closely for them. Moreover, end scrapers and other tools probably made from blades abound in midwestern assemblages, so the distribution of Paleoindian blade technology may be wider than it seems, their paucity in midwestern Paleoindian assemblages more apparent than real. Only thorough inspection and publication of assemblages will determine if, for whatever reason, unmodified Paleoindian blades are confined to the St. Louis area or are more widely distributed in the Midwest. Blades sometimes were used without modification, but most were destined for working into endscrapers and other tools. It would be interesting to estimate how many tools could be fashioned from a typical blade, and the probability that a blade would be worked down to bifaces, endscrapers, or other tools.
In a carefully and tightly constructed argument, Jackson (1998:49-50) used both discrete and metric variables to infer the affinities of two small Ontario assemblages that lacked fluted bifaces. Because they also lacked Onondaga chert, Jackson reasoned on comparative contextual grounds that the assemblages were earlier, not late, Paleoindian. Among early Paleoindian variants, he inferred Gainey affinity by discrete (the presence of quartz and bipolar cores) and metric (channel flake width) variables. Jackson also used negative evidence (absence of channel-flake points and other possible Parkhill Phase diagnostics) to infer Gainey affinity. Across the Midwest, use of quartz does not seem an exclusive Gainey Phase preference. Leavitt may be of Parkhill affinity, and it contains a quartzite fluted biface (Shott 1993:Fig. 6.4e). Certainly Hixton silicified sandstone, mechnically similar to quartzite if not quartz, is a common toolstone for Gainey bifaces at Silver Mound near the source assemblages (Hill 1994:230-231) but also for late Paleoindian types like Plainview and Agate Basin (Hill 1994:Figs. 6-7). Such materials do not seem confined to one particular Paleoindian phase. Obviously, in some cases geochronology can at least set limits on the possible age and therefore affinities of Paleoindian assemblages or at least the deposits in which they occurred.
Spurred end scrapers may be Paleoindian diagnostics (Rogers 1986; cf. Morris and Blakeslee 1987). In Shawnee Minisink's large assemblage, at least some spurs were fortuitous (Rule and Evans 1985:214). In the Midwest, spurring occurred on some specimens at Martens and Bostrom (e.g., J. Morrow 1996:601, 619) and at Leavitt (Shott 1993:Fig. 5.9c, 5.12a). Spurring at Leavitt seemed largely a by-product of resharpening (Shott 1993:72-73), at Parkhill a combination of resharpening and deliberate design (Ellis and Deller 2000:106-110). The frequent appearance of spurs on Paleoindian scrapers is undeniable, but the status and function of spurs remain in question. After all, end scraper spurs are described in ethnographic accounts (Shott 1995a:60) among people who definitely were not Paleoindians. For the Great Lakes, Ellis and Deller (1988) proposed a number of types considered diagnostic of Paleoindian age. Most occur in low frequencies and narrow geographic ranges so are not necessarily found in Paleoindian assemblages across the Midwest, and are defined empirically via an eclectic set of technological, formal and use Criteria. Yet Jackson (1998:49-50, 86) found such types useful in distinguishing Gainey and Parkhill assemblages.
Jackson (1998:122-127) also came down to the level of individual tools in an attempt to distinguish Gainey and Parkhill phase end scrapers. In general, metric variables distinguished only ambiguously. Divergence or expansion angle of the opposing sides of end scrapers, a blank attribute, seem to differ between Gainey (values ca. 25°) and Parkhill (values ca. 30°) phases (Jackson 1998:122). Jackson could not distinguish Ontario Gainey and Parkhill Phase assemblages by length, weight or bit length (1998:125-126). He concluded that "dimensions do not appear to be a sensitive or reliable phase indicator" (1998:126). This is no surprise, since most such dimensions vary with amount of reduction experienced by specimens. Yet discrete variables distinguished phases fairly well. Jackson (1998:126-127) noted much higher frequencies of spurring, especially right-side spurring, and of double-pair notching in Gainey than Parkhill end scrapers.
Geography and Environmental Conditions for the Midwest
Paleoenvironment encompasses the landscapes and landforms, the climates, and the biota that Paleoindians encountered across the Midwest. In its northern half, land forms are chiefly Wisconsinan-age glacial ones; nearly all of Minnesota and Michigan, most of Wisconsin and parts, mostly northern fringes, of Iowa, Illinois, Indiana and Ohio were glaciated during at least the last Wisconsinan advance. Glacial landforms include eskers, till plains and moraines. Related landforms are the strandlines—beach ridges—of peri-glacial lakes. Climate involves patterns in the abundance and distribution of temperature and precipitation. Biota are plants and animals native to the Midwest. Although biota were in part determined by climate, historical factors and ecosystem dynamics mean that, especially for animals, it is difficult to interpret their occurrence and distribution just in terms of climate.
Paleoindian habitats have no strict (in some cases and respects not even rough) modern analogues. This is truer the further to the north that we go. Yet for purposes of description, and interpretation, it is helpful to point out that Midwestern Paleoindian habitats were roughly similar to modern ones found much further north. To some extent, the northern Midwest passed gradually from high Arctic to tundra to Subarctic conditions during the Paleoindian period, although any but the very earliest Paleoindians encountered tundra or parkland only in narrow periglacial zones. Further south glacial effects were attenuated but nevertheless felt. Very broadly, then, a modern latitudinal gradient from, say, northern Minnesota to the Arctic Ocean encompasses—over a wider geographic range—some of the environmental variation that, at Pleistocene's end, was squeezed into the northern Midwest.
The chief issue regarding landform relates to the shapes and levels and flows of the Great Lakes. Although these aspects of the lakes seemed inalterable enough to serve as the "fixed and permanent" boundary between the United States and Canada, their current characteristics are less than 3,000 years old. If we recall their origins as, essentially, large puddles left behind by the Laurentide glacier, we understand how the Great Lakes' extent, levels, exact locations and outlets varied through time at Pleistocene's end as first the Laurentide glacier retreated then readvanced, then began its final retreat. Because Superior was glaciated during much of the Paleoindian period and Ontario falls outside the Midwest, emphasis here is on the middle lakes of Michigan and Huron.
Early twentieth century research established a basic sequence of fluctuations in the Michigan-Huron basin, today in reality a single lake (Larson T.2; Jackson et al. 2000; Larson 1987). For the early Paleoindian period, these basins were occupied by a series of periglacial lakes higher than the modern level. A brief low stage ("Kirkfield Low Stage") was followed at the approximate start of the middle Paleoindian period by Main Lake Algonquin, slightly higher than the modern lake levels. Circa 10,300 rcbp (Jackson et al. 2000) this high stage fell abruptly, exposing vast landscapes previously inundated. Then, the Great Lakes were comparative puddles, and Great Lakes states, especially Michigan, had much more dry land than they do now. The slow rise to the Nipissing high of ca. 6,000 ya, reaching the earlier Algonquin stage, began during the late Paleoindian period.
During most of the Paleoindian period, Lake Erie fluctuated within its modern boundaries; only after the Paleoindian period were large parts of Lake Superior ice-free (Larson 1987:28-30). Central to this sequence and its associated chronology was the concept of the "hinge-line" running roughly from Green Bay across Saginaw Bay and southeast into Ontario. North of this line, the modern land surface was subject to exponential isostatic rebound, which raised earlier landforms like glacial-lake strandlines above their original elevations. South of it, however, rebound did not occur; strandline elevations there would not have changed. If a strandline south of the hinge line today stands at, say, 175 masl [meters above sea level], then during the Pleistocene it stood at 175 masl. North of the hinge line, however, its elevation today would range from 175 masl upward.
A "revisionist" (Morgan et al. 2000:15) view developed in the 1980s. In this model "the hinge line…is an invalid concept" (Larson 1987:2) and isostatic rebound simply rises exponentially everywhere that Wisconsinan ice previously covered. If "revisionists" are correct, Main Lake Algonquin was confined much further north in both the Michigan and Huron basins. During the middle Paleoindian period, Lake Michigan first was higher—larger—than it is now, then considerably smaller and lower. Nowhere during the middle Paleoindian period did Lake Huron extend nearly as far south as it does today (Larson 1987:Figs. 16-17).
The different views are important not just in general terms but because many Great Lakes Paleoindian sites have been dated indirectly by their location on strandlines of known (or presumed) age. Fitting et al. (1966:1-4) used geochronology to infer Holcombe's age by associating its strandline with an extension of the Main Lake Algonquin. This view is not tenable, but the concentration of middle Paleoindian sites in southwestern Ontario (e.g., Deller and Ellis 1992; Ellis and Deller 1997:Fig. 5, 2000) long have been dated by association with strandlines of Main Lake Algonquin that themselves may or may not date to between roughly 11,000-10,300 rcbp. Yet recent syntheses of relevant geological data preserve the Ontario geochronology (Jackson et al. 2000; Morgan et al. 2000). Whether Main Lake Algonquin or not, there were "higher water levels in the southern Huron basin during the period from ca. 11,000 to 10,500 B.P." (Morgan et al. 2000:15).
However we understand the geological dynamics of the Great Lakes, Paleoindians witnessed the rise or fall of lakes, events that were instantaneous in geological terms and probably unfolded in years or perhaps decades. Consider for a moment how remarkable it would be if, say, Lake Michigan receded today, stranding Chicago and Milwaukee or, alternatively, if it rose, transforming them into Midwestern Venices if not deep lakebed. Paleoindians experienced comparable and comparably momentous changes. Exposed lakebeds increased available land and triggered rapid, complex patterns of biotic expansion and succession. These were disturbed habitats par excellence. Yet from an archaeological perspective, the complex fluctuations of Great Lakes levels came at a price: some early Paleoindian sites might be inundated or severely reworked.
Pollen Data and Vegetation
Shane (1994:11-13) summarized vegetation conditions and trends for the eastern Midwest, principally Ohio and Indiana. Her Phases IIa (13,000-11,000 rcbp; 15,631-13,000 B.P.) and IIb (11,000-10,000 rcbp; 13,000-11,450 B.P.) are roughly equivalent to the early and middle Paleoindian phases, respectively. Phase IIa began with essentially periglacial climate and vegetation, the latter dominated by spruce. A strong warming trend then caused hardwoods like oak and ash to increase at the expense of spruce, especially outside the Appalachian Plateau. Net above-ground productivity (NAP) declined in abundance from west to east, indicating more open conditions on the western till plain. Phase IIa vegetation communities were "without modern analogs" (Shane 1994:12). Inferred paleoclimatic conditions suggested mean January temperatures near -8-12°C and mean July temperatures of 23-21°C from west to east. A different longitudinal trend obtained in precipitation, which apparently rose to the east.
Phase IIb was "extremely complex" (ibid) in all climatic and vegetational properties, characterized less by average values than abrupt changes and major differences across time and space. This is no surprise for a period roughly commensurate with the Younger Dryas event. In this interval, the till plain and Appalachian Plateau experienced rather different sequences of change. In the former, Phase IIb opened with some apparent cooling, marked by a spike in spruce pollen from 10% to 30% in less than a century. Just as abruptly, spruce then was replaced by pine in an apparent warming trend perhaps also influenced by the historical factor of pine's slow colonization rate. July temperatures declined slightly, January ones more substantially, and precipitation declined across the till plain. In the Appalachian Plateau, no such dramatic changes occurred, although hardwoods and pine gradually replaced spruce through time. July temperatures were slightly cooler than previously, but January ones considerably warmer. Southern Wisconsin data from ca. 11,000 rcbp–the transition from Phase IIa to IIb—show spruce and net above—ground productivity (NAP) dominance, with minor representation of pine and hardwoods, essentially "open spruce parklands or woodlands" (Fredlund et al. 1996:88).
The best recent midcontinental paleoenvironmental research is from Canada, not the United States. For southwestern Ontario and adjacent parts of Michigan and Ohio, Muller (1999:21-46) compiled and synthesized data from 130 pollen cores. His chief conclusion was that spruce parkland became a nearly closed spruce forest before pine replaced spruce ca. 10,500 rcbp. That is, forests closed before they became dominated by pine. Otherwise, Muller's synthesis corroborates earlier studies, and suggests that middle Paleoindian people experienced rather dramatic climatic and biotic changes from spruce parkland to closed spruce, then pine, forests. Significantly, he suggested (1999:45) that Gainey Phase middle Paleoindian occupation occurred in spruce parkland trending southward, by extension, to spruce forest in Ohio but that subsequent Parkhill Phase middle Paleoindians, at least in southwestern Ontario and, again by extension, across central lower Michigan, occupied a rough ecotone between "closing…spruce-dominant forest to the south, and an open spruce-parkland" to the north.
More broadly, spruce exhibited a pronounced longitudinal gradient across the Midwest ca. 12,000 rcbp from higher to lower values from east to west (Jackson et al. 1997: Fig. 13). At the same time, the Midwest was wetter in general than it is today and certainly than it would be during the subsequent Younger Dryas (ibid:57). On balance, late Pleistocene vegetation communities were "without modern analogs over much of eastern North America…[and] In the mid-continent, these assemblages suggest Picea-dominated woodlands" (ibid:62) with some hardwoods.
If our chronological inferences are correct, early Paleoindian people who may have appeared in the Midwest occupied relatively stable periglacial habitats trending to boreal forest and hardwood forests to the south. By contrast, middle Paleoindian people lived in a dynamic, rapidly changing environment. Across the region, they encountered spruce forests or parklands with some hardwood constituents and sedges and grasses. Roughly similar to modern habitats near Hudson Bay, the northern Midwest nevertheless differed in important respects; no modern habitats are equivalent to what middle Paleoindians found, nor did they experience the stasis characteristic of the post-Hypsithermal Holocene. In environmental terms, these people lived in proverbially interesting times. Postglacial environment adjustments were occurring, and noticeable differences in climate and biota may have occurred within individual lifetimes.
By the later stages of the middle Paleoindian phase, continued change in the maturing postglacial ecosystems had erased the spruce parklands of the Great Lakes and replaced them with landscapes more similar to modern boreal forests. These probably varied latitudinally, those to the north being more open.
Paleoindians surely used wood and plant fibers and must have eaten unknown amounts of plant food. Yet, their diets were most affected by the abundance and distribution of animals. Empirical data from northern Ohio (McDonald 1994) to southwestern Missouri sites like Boney Springs, Koch Springs and Trolinger Springs (McMillan 1976; O'Brien and Wood 1997:45-48) revealed a diverse late Pleistocene bestiary across the Midwest. It included extinct taxa like proboscideans, sloth, giant beaver, tapir and horse, extant but now-exotic taxa like moose and caribou, and many extant taxa like white-tailed deer and small mammals. It also included a surprisingly diverse assemblage of reptiles and amphibians (e.g., Holman 1997).
The FAUNMAP compilation (Graham and Lundelius 1994) shows that the Late Pleistocene Midwest of the early and middle Paleoindian periods contained both a full range of small and medium modern and Pleistocene fauna. Of course, each taxon had its own distribution and patterns of association. These data reveal complex patterns of association, distribution and change determined by climate, community relationships, and history. Taxa fluctuated in abundance and range individually, not as integrated communities. Thus, caribou occurred as far south as Alabama, modern taxa now confined to the east occurred well to the west, and modern plains taxa occurred as far east as Virginia (FAUNMAP Working Group 1996:1601). Over the span of the late Pleistocene and early Holocene, faunal communities were "continually emergent" (ibid), their taxonomic richness and composition only weakly correlated with climate. Such biogeographic furies made the late Pleistocene Midwest far more dynamic than the region is today, and its spatial heterogeneity was at any time then greater than in the Holocene (ibid:1603-1604). This quality makes faunal communities of Paleoindian age hard to infer from paleoclimatic data alone, and underscores the value of additional empirical evidence derived from Paleoindian sites.
All herpetofauna from late Pleistocene deposits at Sheriden are modern and suggest that "a boreal climate was not occurring in northwestern Ohio" (Holman 1997:1) circa 11,700 ya. That far north that early in time, climate apparently had ameliorated beyond the range of boreal habitats. Similarly, fossil Coleoptera–beetles–in southwestern Ontario suggest late Pleistocene climates more temperate than pollen data there would indicate (Muller 1999:38-38). The apparent disjunction between more climatically and microenvironmentally differing controlled vegetation and more adaptable fauna is further evidence that late Pleistocene habitats were unlike any that exist today and faunal communities available to Paleoindians, have completely inadequate modern analogues. Indeed, Midwestern Paleoindians may have encountered and exploited habitats more diverse and spatially complex and heterogeneous than any extant in North America. Subsistence evidence, is sparse but broadly consistent with this conclusion.
Paleoenvironmental Summary. We engage in paleoenvironmental inference to situate Paleoindian cultures in environmental context. Yet knowledge of late Pleistocene environment across the Midwest may not yield better understanding of Paleoindian cultures' material context. As above, late Pleistocene faunal communities were emergent, and their character in specific time-space settings is difficult to infer absent local empirical data, which does not exist for most of the Midwest. Specific environmental and biotic parameters largely elude us, and we are forced to rely upon environmental generalizations that had little bearing on the lives of Paleoindians. Thus any Midwestern Paleoindian site able to yield data in context will be of national significance.
Inference of mobility parameters or population from general ecological parameters like net primary productivity or secondary biomass (Kelly 1995) assumes a correlation between climate and biota that we must deny for the Paleoindian case. The absence of modern analogues in paleoenvironments compromise much of the value of studying those environments in the first place. We are forced to conclude that Paleoindians adapted to habitats of unknown character in ways that we cannot know because our theory does not take account of such habitats. Thus, resolution of biotic data will be essential to more fully understand Paleoindians' economy and, in turn the functional nature of the sites they occupied.
People and Proboscideans
When people reached the Midwest, mammoths (Mammuthus sp.) and mastodonts (Mammut americanum) inhabited the region. Obviously, the determination of whether Paleoindians hunted or scavenged these animals in the eastern U.S. would reveal many environmental contexts and temporal ranges of human occupation and it would certainly have profound implications for understanding the role of human societies in biological extinctions of the Late Pleistocene. And while it would naturally bear witness to Paleoindian diet and economy, as it is supposed to do in the west, the available evidence may be skewed toward the large kills, however intermittent or occasional, because "very large-bodied prey are greatly overrepresented relative to their actual importance in the diet" (O'Connell et al. 1992). Even if people hunted or scavenged proboscideans, they probably ate many other foods as well.
There are no constraints on the time-space distribution of proboscidean fossils in the southern Midwest. Across the northern tier of states, however, late Pleistocene proboscideans cannot predate ice retreat, so their fossil distribution is constrained in time. Yet half or more of dated proboscidean occurrences in Ohio are after human colonization (McDonald 1994:27-28). Fluted bifaces are largely confined to south of Michigan's Mason-Quimby line (Cleland et al. 1998:Fig. 3) and proboscideans entirely are (Abraczinskas 1993:Fig. 1). Human-proboscidean association at Hiscock may be partly a lag deposit owing to geological, not cultural, processes (Laub et al. 1988). Proboscideans may have been associated with stone tools at Koch Springs in western Missouri, although the association seems secondary (McMillan 1976:84, 92). Indeed, no direct evidence short of unmistakable human modification of bone—establishing association with bones, not necessarily living animals—or the smoking gun of stone tools protruding from proboscidean bone makes the case. Elsewhere, undocumented but tantalizing hints of association are part of Midwestern folklore; most archaeologists know a farmer who knew a farmer who had a father who said that his grandfather had a friend who long ago found "spear points" with fossils of an antediluvian beast. What, for instance, can we make of an Illinois collector's 1921 account that three fluted bifaces "were found with a tooth as big as your fist" (Munson and Tankersley 1991:3), unless the tooth was a mastodont molar?
The best Midwest associations are at Hebior, Kimmswick, Martins Creek and Schaefer which yielded proboscidean remains with stone tools of various kinds, although only Kimmswickhas a fairly large assemblage (Brush and Smith 1994; Graham et al. 1981; Overstreet 1996, 1998; Overstreet and Stafford 1997). But Kimmswick's faunal assemblage is large and diverse; stone tools there might be associated with mastodonts, with other animals, or with none at all. Many earlier and poorly documented excavations (J. Morrow 1996:85-88) suggest that the original Kimmswickassemblage was larger still. Tools and proboscidean bones may be associated at Willard in Ohio (Lepper 1999:Table 3) and at Boaz (Palmer and Stoltman 1976), but the latter's 1897 discovery and much later "documentation" render the association unknowable. Unquestionable Paleoindian tools were found at Kimmswickin the 1970s, but poorly documented work there nearly a century ago also may have recovered tools (J. Morrow 1996). McDonald (1994:28) reported possible human-proboscidean associations in Ohio. Fisher's various sites yielded no stone tools but taphonomic evidence suggestive of human agency (Fisher 1987:Table I; Kapp et al. 1990). The Rappuhn Mastodont site in Michigan also falls in this group (Kapp 1986; Wittry 1965). Just beyond the Midwest, there is tantalizing but inconclusive evidence from Ontario (McAndrews and Jackson 1988:170). Finally, at many sites there is no evidence of human association (e.g., Fisher 1987:Table I).
Efforts of Fisher (e.g., 1987) and Overstreet (e.g., 1993, 1998) stand out in the taphonomy of Midwestern proboscidean sites. Fisher reported proboscideans, mostly mastodonts, from southern Michigan and neighboring states, that bore evidence of hunting or scavenging. More recently, Overstreet reported a concentration of proboscidean occurrences in southeastern Wisconsin and adjacent Illinois. Taphonomic study revealed: 1) patterns of skeletal disarticulation plausibly explained as butchering, and articulation of elements plausibly explained as manageable cuts; 2) possible cut marks often crossing joints or where large muscles attached to bone, at Mud Lake for instance (Mason 1997:Fig. 5.6); 3) specimen age and inferred season of death skewed toward healthy adults in late fall (not always [Kapp et al. 1990], when hide and meat were in prime condition, fat was abundant, and animals were healthy; and 4) in some cases, possible caching of butchered segments using weighted intestines as sinkers. No stone tools or other conclusive evidence of human presence were found at 10 (Fisher 1987:Table 1) inferred kills. In Wisconsin, fragmentary tools were found at Schaefer and Hebior (Overstreet 1998:42-43).
G. Haynes (1991) considered the inferred butchering process to obtain bone tools improbable, since stone presumably was needed and, obviously, would work well enough in butchery themselves. The taphonomy of cut marks is a subject of dispute. Patterns of disarticulation seem determined more by skeletal anatomy than any difference between human and natural agents (Hill 1979; O'Connell et al. 1992:333). Most of Fisher's excavations measured in the tens of m2, yet ethnoarchaeological data on large-animal butchering suggest that resulting scatters measure in the hundreds of m2 (O'Connell et al. 1992:351). Small excavations leave much play for sampling error and cast doubt on any inference of butchering from absence of strategic skeletal elements.