Sample bias in the distribution and abundance of Midwestern fluted bifaces

Sample bias in the distribution and abundance of Midwestern fluted bifaces

Shott, Michael J


Maps of fluted-biface distributions are a common subject of archaeological interpretation. Often we regard patterns on such maps solely as the product of Paleoindian land use and population distribution. But maps are drawn from samples of fluted bifaces, samples that are subject to bias. I use Anderson and Faught’s (2000a) database to interpret fluted– biface counts by county in seven midwestern states. In complex but significant ways, modern population distribution and number of recorded sites of all periods correlate with fluted-biface count, but cultivated acreage does not. Observed distributions are not only the result of how many Paleoindians lived or traveled where but also of how many people might seek or have sought fluted bifaces.

Introduction: Fluted-Biface Distribution

Separately, Mason (1958) and Quimby (1958) pioneered Paleoindian distributional studies in the Midwest. Plotting the distribution and abundance of fluted bifaces (calling them “points” is a functional inference that is often, but not always, reasonable) in the Great Lakes region, they demonstrated an association with the fossil remains of Pleistocene mammoths and mastodonts. Thus was the Mason-Quimby Line born, marking the same northern limit in the distribution of bifaces and megafauna across the middle of Michigan’s lower peninsula.

There followed from Mason and Quimby’s lead many Paleoindian distributional studies at various scales. State surveys were undertaken across the United States, often using counties as units of observation. In the Midwest, surveys were reported in Indiana (Tankersley et al. 1990), Iowa (Morrow and Morrow 1994), Minnesota (Higginbottom 1996), Missouri (Chapman 1975), and Ohio (Seeman and Prufer 1982). (Shott [2001] discussed earlier surveys of several of these states.) At larger scales, surveys were conducted of Ohio and neighboring states (Lepper 1983), the eastern seaboard (Brennan 1982, where states and provinces served as units of observation), and the southern Plains (Blackmar 2001). At a subcontinental scale, Anderson and Faught’s (1998, 2000a, 2000b) ambitious survey stands out for its geographic breadth, its fine-scaled data, and its use of GIS technology.

Starting with Mason and Quimby, one general purpose of distribution studies of all scales was to infer the character of Paleoindian land use and adaptations. In the process, patterns in the abundance and, especially, distribution of bifaces were treated as a faithful register of patterns of Paleoindian land use. From Mason’s and Quimby’s gross patterns of human-animal associations, Anderson and Faught (1998, 2000a) progressed even to inferences about tempo and mode of colonization of the Americas from the distribution and abundance of fluted bifaces. Even more impressive is Steele et al.’s (1998) recent simulation that used sophisticated analytical methods and fine-grained environmental data to model Paleoindian colonization mode. Thus, where we recorded many bifaces, we reasoned that many Paleoindians gathered and acted. Fluted bifaces were equated with the footprints of Paleoindians themselves. This much is undeniable, if recorded specimens are a representative sample of all fluted bifaces discarded by Paleoindians.

Like all archaeological research, artifact distributional studies have strengths and weaknesses. Among the former, their unit of observation is irreducible physical objects whose reality is not in question. Thus, they avoid the problematics inherent in the sometimes necessary construction of “sites.” From the start, however, archaeologists acknowledged the troubling possibility that the accumulated record is not a direct but a refracted reflection of past patterns of land use. Many factors can complicate the relationship between the archaeological record as laid down in the past and the record that we document. This is no more than the recognition, made long ago by Cowgill (1970), that archaeology’s physical-finds population need not be faithfully reflected in our accumulated physical-finds samples.

The recorded distribution and abundance of fluted bifaces is one example of a physical-finds sample, accumulated from various sources. Most fluted bifaces were found on exposed ground surfaces, usually cultivated ones in the Midwest, but concentrations of them were recovered in specific excavations. Surface survey was conducted over enormous areas by thousands of individuals of widely varying backgrounds from unreconstructed pothunter to professional archaeologist. Excavation usually-but not always-was performed by professionals in, trivially, much smaller areas. Both quality and quantity of the resulting samples differ between surveys and between all surveys and excavation. Thus, the aggregate fluted-biface sample is the composite of many samples that differed greatly in design, scale, and purpose, depending on who took them, how, and where. None of this is to deny that the accumulated sample preserves much evidence of prehistoric pattern, merely that it is not purely the result of that pattern. Instead, the accumulated sample is the joint product of the archaeological record itself (Cowgill’s physical-finds population) and how we have sampled it. No doubt the accumulated record preserves the pattern in past action; it is merely a question of how much of the pattern is aboriginal and how much the unintended byproduct of sampling.

Simple counting gives us Cowgill’s physical-finds sample, which generally grows in size and may change in character over time. (It may not always grow, since some collections or information about them can be lost.) Yet if we fail to gauge how representative the sample is of the population, we assume that it is faithful in all salient respects. The assumption appeals, but the voluminous sampling literature and the experience of archaeologists cast doubt on it. Of course, we can never be certain of the population’s character because we can never know it completely. But to disregard the sample’s representativeness is equivalent to supposing that if we do not know about an illness, it cannot harm us. Instead, we must gauge the sample’s character, not assume it.

Thus, Mason (1962:235) admitted that the distribution of modern population might affect the recorded distribution and abundance of fluted bifaces, even if he discounted the possibility. Similarly, Anderson and Faught (2000:509) acknowledged the possible effects of “differences in visibility or recording effort” across the United States but also considered them slight. Earlier, Anderson (1990:171, Fig. 2) discounted sample effects in the observed distribution of fluted bifaces precisely because that distribution was highly variable. Yet this quality of the distribution seems at least as likely to implicate sample effects as to preclude them. At best, it is inconclusive. Steele et al. (1998:295) called “grossly simplifying” their assumption that the observed distribution of fluted bifaces was free of sample bias, but they seemed nevertheless to discount the possibility of bias. Natural scientists face comparable sampling problems but recognize the need to gauge the reliability of their samples, not just assume it (Kodric-Brown and Brown 1993). Archaeologists have yet to assimilate the same lesson.

In the Midwest, the virtues of fluted-biface samples were debated at length in Ohio and neighboring states during the 1980s (Lepper 1983, 1985; Seeman and Prufer 1984). Briefly, Lepper argued that visibility and, especially, modern population affected the size and character of the sample of fluted bifaces. Seeman and Prufer agreed that modern population and fluted-biface counts were correlated but largely discounted that relationship for interpretive purposes. Paleoindian distributional studies remain popular, but the question of sample bias remains unresolved or at least in contention.

Here, I study three factors that may complicate the relationship between fluted– biface samples and the original populations from which they were drawn: modem population distribution, patterns of modern land use, and intensity of archaeological scrutiny. All else equal, the more people who live in an area, the more collectors that population includes. In turn, the more collectors there are, the more of the archaeological record they are apt to find in the aggregate. Since many if not most fluted bifaces and other prehistoric remains are found on surfaces, the more exposure there is, the more “sites” and artifacts might be found. Finally, the more archaeological research carried out in an area, the likelier that Paleoindian sites and fluted bifaces will be found. Sampling bias introduced by these factors, separately or in combination, has been examined in studies in the Midwest and elsewhere (Blackmar 2001:75-76; Largent et al. 1991:330; Lepper 1983, 1985; cf. Seeman and Prufer 1984).

The first of these factors, population, bears lengthier examination than the others. The vast majority of the modern midwestern population are not collectors, so population size per unit area does not directly measure collection effects. Effects are best measured directly, by counting the number of collectors who operate in particular areas and the size of their collections. For the modern day, so much is knowable if practically unknown across the Midwest. Partly, this is because some collectors do not wish to be known but mostly because archaeologists have not made the effort to know them. Moreover, historical changes in the size and distribution of the collecting population since substantial Euroamerican settlement of the Midwest began nearly 200 years ago do not register in the size and distribution of the contemporary collector population. Once we take seriously the need to census collectors and collections, we can directly measure their effects. We may find that the numbers of collectors are proportional to modern population size or we may not.

But modern (and historical) population is known through census records. In the absence of direct evidence, censuses are reasonable proxies for collection effects. In the United States, gauging collection effects is rare, if it has ever been attempted, but in one study of the distribution and abundance of Middle Pleistocene British handaxes, most were “spatially associated with the major [modern] centres of population” (Hosfield 1999:25). Modern population may be a proxy for collection and the landscaping that often reveals archaeological evidence, variables considered more directly influential on fluted-biface counts. Yet if those counts covary with it, surely this reveals some form of sample bias, whether from collecting and land alteration or not. The only alternative is to suppose that, somehow, Paleoindian and modern populations chance to be distributed in the same way despite the undeniably vast differences in their sizes. Not only do the attractions of settlement depend very much on economic, technological, and broader cultural contexts, as different as can be imagined between late Pleistocene Paleoindians and the United States today, empirically this is equivalent to arguing that Paleoindians lived mostly in and near major cities. Also, modern population is sedentary. Paleoindians probably were highly mobile, occupying many places and much territory in the course of their annual rounds. If their population was much smaller than is ours, their scale of residential land use at the level of family or co-residential group was much greater. It seems highly doubtful that Paleoindian and modern population distributions should be similar either for cause or by chance. It follows that modern population distribution can be treated substantially as independent of its Paleoindian equivalent. If the fluted-biface distribution covaries with modern population, to that extent it is independent of Paleoindian population distribution, and a bias is revealed in the fluted-biface sample.


Recording Units and Study Area

To study the relationship between fluted bifaces and modern practices, I must first choose a scale at which to record data. Lepper (1983) counted fluted bifaces by physiographic province in a study area that included Ohio, Indiana, Michigan, and neighboring states. Other studies (e.g., Blackmar 2001) used physiographic provinces along with counties as observation units, and still others simply used counties (e.g., Largent et al. 1991; Meltzer and Bever 1995; Seeman and Prefer 1982, 1984). There are good reasons to correlate biface distribution with physiography, but the small number and large size of provinces hampers statistical analysis (Seeman and Prefer 1984:227). By contrast, counties are many in number, small in size relative to the study area, and arbitrary in location and boundaries with respect to the archaeological record. Therefore, counties are useful contrivances by which to measure the distribution and abundance of bifaces and other variables of interest across the Midwest.

My study area is the midwestern states of Indiana, Iowa, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. County-level data on number of fluted bifaces were published for Indiana (Tankersley et at. 1990), Iowa (Morrow and Morrow 1994), Minnesota (Higginbottom 1996), Missouri (Chapman 1975), and Ohio (Seeman and Prufer 1982). Older or partial data exist for Michigan and Wisconsin, as referenced in Anderson and Faught (2000b). Most of Michigan, Wisconsin, and Minnesota was ice-free at the time of human arrival (Dyke and Prest 1987), but few fluted bifaces are reported in their northern sections (including all of Michigan’s Upper Peninsula) and it is unlikely that these areas were occupied extensively by Paleoindians. Therefore, they were omitted. Thus, in Michigan the study area limit approximated the Mason-Quimby Line (Figure 1). In Wisconsin, fluted-biface survey data in the larger research project of which this study is a part (Shott 2001) were available only in Regions 6, 7, and 9 (Boszhardt 1991; Overstreet 1991a, 1991b), also shown in Figure 1. By default, therefore, all of northern Wisconsin was omitted, as was its southwestern corner. Fluted bifaces are reported in northern Minnesota, but they are few in number and some are only ambiguously fluted (e.g., Higginbottom 1996:Fig. 35), a problem with Higginbottom’s catalogue (not his efforts), acknowledged but otherwise ignored. Although not alone a reason to omit northern Minnesota, its counties also are more irregular in shape and variable in size than elsewhere in the Midwest, so the value of counties as de facto observation units is low there. The study area appears in white in Figure 1. It numbers 529 counties, although for some analyses only subsets of counties could be used, as described below.

Wiant (1993) reported the distribution and abundance of Illinois Paleoindian sites, but there is no equivalent information for fluted bifaces themselves. Like southwestern Wisconsin, therefore, Illinois was omitted from the larger study of which this article is an outgrowth (Shott 2001) and is not included here. My original plan was to use updated biface counts but, as subsequent analysis shows, results differed little from Anderson and Faught’s (2000b) figures. Ultimately, I used the latter source, which also included southwestern Wisconsin and all of Illinois. These areas could have been included in analysis even though they could not be included in the larger study (Shott 2001); by the time I reached this conclusion, however, analysis was well advanced.

Fluted Bifaces

As above, all data are taken from Anderson and Faught (2000b; see also 1998:Table 1). Because my concern is the effects of modern factors on the sample of known fluted bifaces, it was unnecessary to distinguish Clovis, Folsom, or other fluted-biface taxa. Clovis and Folsom distributions may differ, at least in detail, where both types are common (Blackmar 2001:74; Largent et al. 1991:329; Meltzer and Bever 1995:60). Yet there is no consensus on the presence of Clovis bifaces in the Midwest or their typological affinities to Gainey bifaces. In the Midwest, Folsom bifaces are found chiefly in the west (Munson 1990) and are rare eastward, where Barnes and Cumberland types are common. Therefore, all fluted bifaces were grouped as a single category for analysis, found in Column G of Anderson and Faught’s (2000b) data file.

Perhaps midwestern state surveys vary in quality. Certainly they vary in age. Indiana, Iowa, Minnesota, and Wisconsin data were compiled since 1990, so are reasonably current. But Ohio data are from Seeman and Prefer (1982). Missouri data are even older, being reported in Chapman (1975:Fig. 4-3), itself drawn from information collected as long ago as the late 1960s. (Counts from Missouri and several other states were supplemented in Anderson and Faught [2000b] by Munson [1990].) O’Brien and Wood (1998:57) acknowledged that more fluted bifaces have been found in Missouri but did not report exact figures. Similarly, much Wisconsin data are from Stoltman and Workman (1969). Michigan data are derived from Mason (1958), supplemented by more recent personal communications to Anderson and Faught. Thus, my data are from most, but not all, midwestern counties, even though Anderson and Faught reported fluted-biface counts for all counties.

No matter how old the data, their accuracy and reliability sometimes are doubtful. In Higginbottom’s (1996) survey, some fluted bifaces were unavailable for study and hence are hearsay occurrences. There also is the question of authenticity. In recent years, chertknapping has become a cottage industry capable of “muddling the prehistoric record” (Whittaker and Stafford 1999:203) by representing newly made specimens as ancient. Absent detailed study, no one knows how common this practice is in the Midwest. At least two “Clovis” bifaces previously reported at Wisconsin’s Kouba site are identified as modern frauds (Kouba 1985). At least one Minnesota fluted biface is questioned on similar grounds (Higginbottom 1996:MFP.HE.1, Fig. 20). Higginbottom’s (1996:Fig. 1) results suggested that the provenience or authenticity of nearly 30 percent of all reported fluted points was unreliable to some degree.

Anderson and Faught’s data necessarily are drawn from many sources of varying currency, using various discovery and reporting methods, and on a continental scale. It is no surprise that they do not agree entirely with values for fluted bifaces compiled only for the midwestern states and using some sources that are quite recent. One reason for differences is currency of information. For instance, Anderson and Faught used Seeman and Prufer’s (1982) data for Ohio. That source recorded 14 fluted bifaces in Muskingum County, yet Carskadden et al. (2000) reported 126 there. Neither do Anderson and Faught’s data include the fluted bifaces found in more recent excavations at Nobles Pond (Seeman 1994), Paleo Crossing (Brose 1994), or Sheriden (Tankersley 1999). Recent Wisconsin work has reported impressive numbers of fluted bifaces in Pierce County (Amick et al. 1999:Table 2), which in any event lies outside this study’s sample. Trempealeau County has no fluted bifaces in Anderson and Faught’s data, but nine were found there in recent years (Amick et al. 1999). Based on a good 1969 survey, Anderson and Faught recorded one biface for Calumet County, yet at least four were found there at Aebischer (Mason 1988:212). In St. Louis County, Martens was recorded too recently for listing in Chapman’s (1975:Fig. 4-3) survey of Missouri fluted bifaces. Its 24 fluted bifaces (Morrow 1996:Table 13) presumably are not among the 60 that Chapman reported and that Anderson and Faught recorded there. Anderson and Faught used Higginbottom (1996) for Minnesota data. Anfinson (1997:28-29) reported fluted bifaces in Nobles and Waseca Counties, as did Jenks (1937:42-43) in Anoka and Fillmore Counties, that Higginbottom did not report. In Michigan, Anderson and Faught reported no bifaces for Clinton County, where at least eight have been found on the surface at Leavitt (Shott 1993). They reported six bifaces from Monroe County, where the Grogitsky site has yielded at least nine (Zurel 1979) and several others are known, and six from Midland County, where at least 14 were found apparently on the surface at Barnes (Wright and Roosa 1966). Indeed, it may be desirable, if difficult, to distinguish counts of fluted bifaces found in survey and excavation, since these forms of fieldwork have different sampling characteristics.

Different typologies used to define fluted bifaces also create some differences. For instance, Seeman and Prufer (1982) reported 184 bifaces for Coshocton County, whereas Lepper (1988:37) reported 410, but the two studies used different criteria to define fluted bifaces (M. Seeman, personal communication 2001). Like (usually rising) changes in counts through time, these are honest consequences of the history of Paleoindian research. They are not to be lamented but must be acknowledged. To the extent that we continue to use continental-scale fluted-biface compilations, we also must control for typological differences among primary sources and the different numbers of bifaces that they yield.

Although Illinois was omitted from this study, the challenge of compiling accurate, complete information on fluted-biface distribution and abundance is nowhere more prominent than in one case from that state. Anderson and Faught recorded no fluted bifaces in DeWitt County, in central Illinois. On their distribution maps (e.g., Anderson and Faught 2000:Fig. 1), it is a void. Yet between roughly 1910 and 1960, a county resident recorded 332 fluted bifaces there (Munson and Tankersley 1991). Perhaps some were fakes. Perhaps others mistakenly were recorded more than once. Perhaps some specimens were misidentified as fluted. If as many as half of the recorded total are legitimate, then DeWitt County is transformed from a void to having among the largest concentrations of fluted bifaces in the Midwest. On distributional maps, it would change from blank space to continental hotspot. Illinois is perhaps the first midwestern state where we should renew systematic efforts to compile flutedbiface counts.

In the Midwest especially but also elsewhere, most recorded bifaces were found through surface collection. Excavation is a far more intensive effort that samples smaller areas and is undertaken less often by amateurs than professionals. Therefore, incidences of fluted bifaces from survey and excavation are not comparable. For instance, Anderson and Faught reported 40 bifaces for Genesee County, Michigan. The documented number is considerably higher, yet all but a few were found in extensive excavations at Gainey and Butler (Simons 1997; Simons et al. 1984). In Macomb County, Michigan, the Holcombe monograph (Fitting et al. 1966) was a model for its time yet did not distinguish between bifaces found on the surface and in excavation, and both professional and amateur work there involved extensive excavation. It would be difficult to determine how many fluted bifaces from Holcombe and nearby Paleoindian assemblages were found on the surface.

Clearly, the midwestern fluted-biface record is a historical product, its size depending on when figures were compiled. From state historic preservation office (SHPO) and state archaeologist records, Figure 2 plots the combined total of Paleoindian sites (not bifaces) recorded by decade from 1910 to 2000 in Indiana, Iowa, Michigan, and Wisconsin. Data were compiled from available sources by Shott (2001). The Ohio SHPO organized its figures differently, reporting Paleoindian sites as being recorded only from 1970 to 2000. Considering Prufer and Baby’s (1963) data, these figures must indicate not date of discovery but of registration in Ohio files. The pronounced increase since the 1970s likely is the result of increased research, especially contract work, and perhaps better documentation of local collections. Discoveries declined in the 1990s, so perhaps sampling has reached a point of diminishing returns. (Yet Texas figures for Clovis bifaces nearly doubled from 1985 to 1995 [Meltzer and Bever 1995:47], and Folsom counts apparently continued to rise into the 1990s [Largent 1995:323].) Figure 2 also suggests regional differences. On the western Plains, Paleoindian site discovery peaked in the 1930s and 1950s (Seebach 2000:Fig. 3), relatively quiescent times in the Midwest. Those peaks were due to drought and the increase in surface visibility it caused across the Plains (Seebach 2000:1). The fluted-biface sample from the entire United States not only is a historical product, but it is a complex one whose causes and sources vary by region.

Elsewhere there are even greater differences between Anderson and Faught’s fluted-biface counts and those used by others (e.g., Blackmar 2001:71-72), probably for various reasons. None of this is to criticize Anderson and Faught, both because it is easier to compile more complete information for smaller areas and because their efforts are systematic. Indeed, none of us knows exactly how many fluted bifaces have been found in each midwestern county, because some are reported only in unpublished records, others in obscure publications, and many not at all. Anderson and Faught used the best available, reasonably comprehensive sources from across the United States. Pointing to current information is fair, but it serves no legitimate purpose to criticize their data for shortcomings that, in some degree, are shared by even the most current sources. Indeed, whatever pattern resides in the distribution and abundance of midwestern fluted bifaces is robust; revising Anderson and Faught’s fluted-biface counts to account for the recent data cited above did not alter any statistical results in this study (e.g., between their and revised biface counts, r=.90, p=.00; r^sub s^=.99, p=.00). Their data are convenient, available to all, and certainly are a reasonably accurate and reliable source for purposes of this study, although the details of their distributions (e.g., size and location of concentrations) are open to question. Therefore, I report results obtained using Anderson and Faught’s data.

Other Variables

Population was measured directly from census records (U.S. Census Bureau 2000). (St. Louis City and County, Missouri, were reported separately but combined for analysis.) But population can differ greatly even between neighboring counties, and collectors travel to their field sites without regard to county or even state lines. Thus, two hypothetical counties that have equally small modern populations may attract different numbers of collectors if one county is surrounded by other small counties but the other adjoins a major city. Therefore, I smoothed each county’s population by calculating the mean of its population and the population of all bordering counties, including those that bordered it only at a single corner point and those from adjacent states included in the study. Obviously, counties that are larger or more irregular in form tend to border more counties. Yet the effect of this fact is uncertain; it may raise or lower smoothed values. Whatever the case, in no analysis did original and smoothed population pattern differently with biface count. Therefore, I report results of analysis using only original, unsmoothed figures.

Lepper (1983:Table 1) used census values from sources nearest in time to the publication date of fluted-biface surveys. Certainly population has increased in recent decades, but only modestly in the Midwest compared with other parts of the United States. Indeed, many rural counties probably have lost, not gained, population in that interval. On balance, use of most recent census figures assumes that changes in modern population distribution since the dates of various state surveys do not significantly affect results. To test for census-age effects, I also obtained earlier census figures for two states, Ohio and Missouri, and report their use below.

Land use was measured as cultivated acreage as of 1997 (National Agricultural Statistics Service 1999), on the reasoning that exposure greatly increases the probability of discovery and that, at least in the Midwest, cultivation is the most common form of exposure. I also measured land use as the proportion of each county under cultivation, or “relative acreage.” Survey effort or intensity was measured indirectly as the number of recorded archaeological sites. Websites for Iowa (Office of the State Archaeologist 2001), Minnesota (Office of the State Archaeologist of Minnesota 2001), and Missouri (Archaeological Survey of Missouri 1997) reported archaeological sites per county. Iowa and Minnesota data are for all sites, including historic ones (S. Anfinson and C. Eck, personal communications 2001). Missouri data apparently also included historic sites. No comparable figures could be found for other states. Thus, survey intensity figures are not strictly equivalent between states, a fact acknowledged but otherwise ignored.

Variable Distributions

The frequency distribution of fluted bifaces per county is highly skewed (Figure 3a, which omits the four counties-Adams, Coshocton, and Ross in Ohio and Harrison in Indiana-that have more than 60 bifaces). The high count makes significance likelier and, not surprisingly, the distribution is nonrandom when compared to the Poisson model (chi^sup 2^=3392.6, p

Modern population also is skewed (Figure 3b). Testing the distribution for randomness is pointless, considering the very high values involved and the improbability that people today are randomly distributed across the Midwest. Smoothed population values are similarly skewed. Cultivated acreage per county is roughly normal in distribution (Figure 3c, d). Finally, total number of recorded archaeological sites per county also is skewed (Figure 3e).

Variable Scale and Transformation

Many variables are interval in scale and so are suited to analysis using parametric statistics. Yet parametric statistics are not robust measures of correlation in skewed data. Another way to seek patterning is to reduce the variables from ratio or interval to ordinal scale and to measure association, not correlation, between them. Accordingly, for some analyses, the interval-scale fluted-biface distribution was reduced to four ordinal classes. Simple quartiles cannot be defined in interval variables unless their boundaries chance to fall at interval boundaries. The 529 counties in the sample include 231 (43.6 percent) with no fluted bifaces; exact quartiles cannot be drawn from these data. Instead, I defined ordinal classes on the reasoning that successive intervals in skewed data should span wider ranges when possible (Table 1). Frequencies sum to 529 cases for bifaces and modern population but fewer cases for land use and survey effort because relevant figures were not available for all states.

Class intervals were defined as follows. Log-transformations of population and survey effort are roughly normal in distribution, so intervals defined by standard deviation of the log-transformed variables were used. Both acreage variables are roughly normal in distribution, so standard deviation of the untransformed variable was used to define ordinal intervals. For log-population, acreage, and log-sites, therefore, cases whose value is = mean+1 s.d. the fourth. Table 1 shows the ordinal classes defined for each variable and the number of cases falling in each.

Analysis: Biface Frequency Distribution

Two hundred ninety-eight of the 529 counties in the study area have at least one fluted biface. Thus, fluted-biface ubiquity is 298/529=56.3. Ubiquity is a limited but robust measure. Calculated separately for Clovis and Folsom bifaces in each of four regions of Blackmar’s (2001:73) southern Plains study, ubiquity did not attain this value, although Clovis ubiquity in the High Plains region was 51. Overall Clovis ubiquity was 42 and Folsom ubiquity 28. But results are not comparable to this study because Blackmar distinguished Clovis and Folsom bifaces, while all fluted specimens here were treated as a class. Comparing Blackmar’s Figures 2 and 3, a total of 223 of 436 counties (ubiquity=223/436=51.1) contained at least one fluted biface. Thus, southern Plains and midwestern ubiquity values are similar.

Bifaces and Population

As above, both variables are skewed. Natural-logarithm transformation of either or both variables changed their separate distributions and the form of the plot of biface count on population. Yet log-transformation did not significantly alter statistical results. Therefore, Pearson’s r and the nonparametric Spearman’s rho (r^sub s^) were used to measure correlation between original, untransformed variables.

Also above, I raised the question of using the most recent census figures. To gauge the problem, I obtained 1980 Ohio (nearest to Seeman and Prufer’s [ 1982] survey) and 1970 Missouri (between Chapman’s [1975] study and its 1969 origin) census figures. In both states, the earlier and most recent census values were highly correlated (r=.99, p=.01). When Ohio and Missouri values were analyzed together, the same result was found. Population obviously differs between censuses, but values are highly correlated between censuses 20 and more years apart. I further tested the relationship by separately correlating earlier and most recent Ohio and Missouri census figures with biface counts. In Missouri, biface count correlated identically with the two population figures; in Ohio results differed trivially. In other states, the 1990 census might be used although even then, Iowa and Minnesota surveys were published in the mid-1990s, about as close to the most recent as the 1990 value. Yet the 1990 census is more recent than either used for Ohio and Missouri so is likely to correlate even better with recent figures. I conclude that recent figures are a valid population measure whatever the date of a fluted-biface survey.

Across all cases, biface count and modern population are correlated (r=.15, p=.00; r^sub s^=.42, p=.00), but a cross-plot defies interpretation (Figure 4). Apparently, modern population does not correlate very much with biface count. Yet four cases-Adams, Coshocton, and Ross Counties, Ohio, and Harrison County, Indiana-are conspicuous outliers that have many fluted bifaces but small populations. (As above, the most recent report of at least the Coshocton County value is even higher.) Coshocton and Harrison Counties both contain major chert sources, so quarrying might partly explain their abundance of fluted bifaces. Adams and Ross Counties were surveyed closely for Paleoindian sites in the 1960s and 1970s (Lepper 1985:243), so intensity of archaeological effort might help explain their abundance. Removing these cases as legitimate outliers for these reasons, correlation between bifaces and population increases slightly (r=.28, p=.00; r^sub s^=.42, p=.00) but remains ambiguous (Figure 5). Still, modem population and fluted bifaces seem weakly related using parametric measures.

Analysis of variance (ANOVA) tests for difference in means of one variable by treatment or levels of another. Biface count varies by population classes (F=7.7, p=.00) (Table 2). Standard deviations exceed means; data are highly variable and thus complex, a finding reached throughout this study that advises great caution in the use of mean values for description. Student’s t-tests for difference between successive population classes mostly are significant (Table 2).

Ordinal data like the classes defined above are suitable for analysis of association. Classes of fluted bifaces and modern population are associated strongly (chi^sup 2^=90.5, df=5, p=.00), even measured by Cramer’s V (=.24, p=.00) to minimize the effects of large counts (Table 3). No expected cell counts were less than 5. Direction or sign of standardized residuals shows that small modern population is associated with low fluted-biface counts, large modern population with high biface counts. To determine if these results are robust, I further reduced each ordinal variable by combining the two smallest classes with one another, and the two larger with one another. Again, association was highly significant (chi^sup 2^=48.9, df=1, p=.00; Fisher’s Exact Test p=.00, V=.30, p=.00) (Table 4).

Modern population and fluted-biface count are associated. Scale and numerical pattern of relationship are not measured in these tests, nor do they have predictive value for individual counties. But we should not expect precise relationships to register in data subject to so many uncontrolled biases of unknown magnitude any more than we should expect parents’ height to determine entirely adult height in their children. The pattern of association alone is sufficient to demonstrate that modern population contributes to but does not determine the number of fluted bifaces recorded per county. Still, it is worth examining original data for complex patterning that may reside in them.

Parsing the data set might control for some complications and reveal patterns in subsets of data that are not visible in all data. For instance, counties having no fluted bifaces tend to have small populations but, trivially, vary much more in population than in bifaces, since they are identical in the latter respect. Counties having few bifaces also tend to be small (but larger on average than those having none) but again vary much more in population than in number of bifaces. Thus, variation is constrained in fluted bifaces. It seems sensible to examine the relationship between bifaces and modern population where the range of variation is greater in bifaces.

Therefore, Figure 6 plots bifaces against population for the middle two biface– count ordinal classes (1-2 and 3-7 bifaces) and the lowest population class. In this population range, population and biface Class 2 are independent 11, p=.76; r^sub s^ -.22, p=.46) but population and biface Class 3, which occupies a wider range than Class 2, are correlated (r=.82, p=.01; r^sub s^=.83, p=.01). Figure 6 also shows separate regression lines on population for biface Classes 2 and 3. (Regression lines in these and subsequent figures have little predictive value and appear only to illustrate patterns of covariation.) If rather more equivocally, population Class 4 and biface Classes 2 and 3 show similar patterns (Figure 7). Correlation in biface Class 2 is practically nil (r=-.04, p=.85; r^sub s^=-.25, p=.24) but diffuse if weakly significant in biface Class 3 (r=.28, p=.20; r^sub s^=.34, p=.11).

Figure 8 plots biface Class 4 against population Class 3. Coshocton and Ross Counties, Ohio, and Harrison County, Indiana, again are conspicuous outliers. Removing them as special cases for reasons cited previously, biface count and population are highly correlated (0=.48, p=.00; r^sub s^=.48, p=.02) (Figure 9). Finally, Figure 10 plots bifaces against population in each variable’s highest class. St. Louis City and County, Missouri, perhaps forces the correlation, which nevertheless is highly significant (r=.63, p=.00), although not in r^sub s^ (=.22, p=.26).

Overall, population and biface count pattern equivocally. But subsets of the

variables correlate quite strongly. The pattern is complicated by other factors, the underlying distribution and abundance of fluted bifaces perhaps chief among them, but modern population and biface-count are related.

Bifaces and Land Use

As with modern population, bifaces correlate with cultivated acreage (r=-.15, p=.00; r^sub s^=-.17, p=.00) (Figure 11). Yet the relationship is negative: more bifaces occur with less, not more, ground exposure. Omitting the same counties-Adams, Coshocton, and Ross in Ohio and Harrison in Indiana-that are outliers to this relationship as they were between bifaces and population scarcely affects pattern or results (r=-.18, p=.00; r^sub s^=-. 17, p=.00). The surprising inverse relationship suggests either the counterintuitive notion that more bifaces are found where exposure is less or that the relationship between acreage and bifaces is equivocal. Similarly, biface count patterns significantly but negatively with acreage classes (F=5.7, p=.00); higher classes have fewer fluted bifaces (Table 5). Successive acreage classes differ significantly in biface count (Table 5). Finally, biface and acreage classes are associated (chi^sup 2^=25.7, p=.00; V=. 13, p=.00) but standardized residuals show an inverse, not direct, pattern of association between bifaces and acreage (Table 6). High acreage values generally occur with low biface counts, low acreage values with high biface counts. Standardized residuals are lower than those found in association between bifaces and population, and biface Class 2 is largely unassociated with acreage. Plotting and correlating biface and acreage classes like biface and population ones yield no clearly significant patterns.

Lepper (1983:273) found a correlation with fluted-biface count and percentage, not absolute amount, of cultivated acreage per county. In heavily agricultural counties, absolute cultivated acreage varies much more with county size than with land use. Relative acreage (cultivated acreage/total acreage) controls for county size so perhaps is a better land-use measure. Obviously, it has a theoretical maximum value of 1.0 and an empirical maximum in these data of .93.

Absolute and relative acreage are highly correlated (r=.82, p=.00), suggesting that biface count patterns no more clearly with the latter measure. Across the study area, perhaps county size varies less than does extent of cultivation, such that absolute or relative acreage cultivated measure nearly the same thing. Indeed, relative acreage correlates weakly with biface count (r=-.14, p=.01; r^sub s^=.12, p=.01). Biface count patterns significantly but negatively with relative-acreage classes (F=4.9, p=.00), yet only the two highest biface classes differ significantly in mean value (Table 5). Finally, biface and relative-acreage classes are associated (chi^sup 2^= 19.7, p=.02; V=. 11, p=.02), but again standardized residuals largely suggest an inverse relationship: as relative-acreage rises, biface-count declines (Table 6). Most standardized residuals again are lower than comparable values for the association of biface-count and modern population. As with absolute acreage, plots of relative-acreage classes and biface-count classes yield no clear patterns.

Elsewhere, biface counts were higher in areas having more surface visibility (Blackmar 2001:75-76), consistent with Lepper’s finding. Moreover, land-use effects on the probability of discovery are suggested in the Plains (Seebach 2000), so exist in places. Biface count and land use, in both absolute and relative terms, seem largely independent. This may be due in some degree to the somewhat restricted range of variation in midwestern land use, since it is the most extensively cultivated section of the United States. But that range still is significant. In the Midwest, biface-count and land use are legitimately independent.

Survey Effort

Paleoindian sites and fluted bifaces might be found in the course of research or contract fieldwork. Survey effort is difficult to measure directly without knowing number of person-days of archaeological survey or excavation in each county. In Texas, Meltzer and Bever (1995) used a crude estimate of effort: number of recorded sites per county of all ages, not just Paleoindian ones. Absent better measures, I use it here as well.

In the midwestern study area, biface count and number of archaeological sites are correlated (r-.36, p=.00; r^sub s^=.42, p=.00). Two Missouri counties, St. Louis and Franklin, are conspicuous outliers at high biface values (Figure 12). Correlation declines slightly but remains significant upon their omission (r=.31, p=.00; r^sub s^=.41, p=.00). Largent et al. (1991:329) noted a similar correlation in Texas for Clovis bifaces only, as did Meltzer and Bever (1995:59) between Paleoindian sites, not bifaces, and total recorded sites.

In ordinal data, biface count patterns positively with number of recorded sites (F=8.3, p=.00) (Table 7): the more sites recorded per county, the more fluted bifaces also are recorded. Student’s t-tests between successive pairs of site-number levels also are significant in most cases, and always when significance level p=.10 (Table 7). Association between ordinal classes of both variables is strong (chi^sup 2^=52.5, p=.00; V=.25, p=.00), although expected values are

As with other sample factors, recorded sites were plotted against fluted bifaces for all combinations of ordinal classes on both variables. Pattern is weaker than in modern population, but again seems meaningful in higher classes on both variables, that is, at the upper end of the range in bifaces and recorded sites. Because the latter is known only for three states, sample size is smaller and significance more difficult to attain. Pooling classes partly remedies this deficiency. In recorded-site Classes 2-4 and biface Class 4, correlation is weakly significant (r=.45, p=. 10; r^sub s^=.44, p=. 11) but perhaps forced by St. Louis County, Missouri (Figure 13).

Joint Causality of Population and Survey Effort

Fluted-biface count covaries with modern population and recorded sites, but little with cultivated acreage. Biface count seems subject to sample effects. Yet strength and direction of covariation differ, being stronger in modern population than in recorded sites. Also, patterns are clearer at the upper ends of all variables’ ranges. This conclusion further justifies the omission of northern parts of Minnesota, Wisconsin, and Michigan, because values for all variables are very low in those areas.

Because biface count covaries positively with modern population and site count, I now consider these two variables jointly. Because biface Classes 3 and 4 covary most strongly with population Classes 3 and 4 and site-count Classes 2-4, I confine analysis to these classes. Site count is known only for three states, which reduces sample size. Multiple regression of biface count on population and site count is questionable because of sample size and the rather wide dispersion of cases evident in crossplots. Partial correlation parses the contribution of population and site count to biface count. Controlling for site count, biface count and population correlate well (r=.59, p=.06, but n=9). Controlling for population, biface and site counts correlate more weakly (r=.31, p=.36, n=9). As far as this small subsample permits, the effect of site count on biface count seems mediated by modern population.


Modern population and survey effort affect the size and character of the midwestern fluted-biface sample. Cultivated acreage is not clearly related to biface count. At least this factor can be eliminated as a source of bias, surely some good news to those who dislike the conclusion that the fluted-biface sample is biased. Still, we cannot attribute the distribution and abundance of fluted bifaces just to Paleoindian population distribution, colonization routes, and the like. To some significant extent, fluted-biface distribution and abundance owe instead to how many people reside in counties today and to how many other sites are recorded there. Where more people live, there is more opportunity to find bifaces, and where more archaeological sites are recorded, more fluted bifaces are found. The sophistication of recent models is no solution to the serious sampling problems that they elide. Thus, Steele et al.’s (1998) technical virtuosity is vitiated by their own virtues, principally the correlation of fine-grained environmental data with a fluted-biface distribution and abundance subject to unknown but certainly significant sample bias. The bias documented here has implications for any study that cannot demonstrate freedom from it.

Results therefore cast doubt on the uncritical interpretation of distribution data but not on their value. On the contrary: we need more, not less, information of the sort Anderson and Faught (2000b) compiled. Its critical evaluation to account, for instance, for the effects of modern population should reveal even more meaningful patterns than heretofore detected. As both the fluted-biface sample and our ability to account for sample effects increase, perhaps robust patterning will emerge in the evidence. By now, it is commonplace to observe that most fluted bifaces have been found in the East, not the West (e.g., Anderson and Faught 2000a:510). Yet many more people have looked for them in the East than in the West. The population effect documented in this study and others urges caution in our judgments about where the balance of the fluted-biface distribution lies. It may very well be in the East, but until we account for the effect of larger modern population there, we cannot be certain.

Modern population and other factors affect the quality of fluted-biface samples but do not entirely determine them. Much if not most patterning in biface distribution and abundance inheres in the record itself. It is just that some, perhaps much, patterning is due to sample effects. No longer is the question whether fluted-biface samples are biased but, rather, how and how much. Treating the observed distribution and abundance of fluted bifaces as an unbiased record of Paleoindian population distribution is no different than measuring intelligence by SAT scores. Surely intelligence is one factor among others-class background, gender, preparation, education-that influences scores. Yet not even aptitude tests’ most ardent defenders any longer claim that scores measure intelligence, because the effects of other factors are well documented.

Now we must attend to identifying biases and controlling them. Like this one, Hosfield’s (1999) Paleolithic study used stone tools as observation units. Archival sources on private collectors and their collections, their location, and duration of collection demonstrated measurable effects on the distribution and abundance of artifacts (1999:45-47) that Hosfield considered necessary to control before accepting artifact distribution as a Paleolithic phenomenon. Indeed, Hosfield also modeled the aggregate effect of land use, industrial activity, and other factors that influence the discovery of artifacts. In his study, sample bias was recognized but also measured and controlled. Once we make the effort to collect relevant information, there is no reason that we cannot do the same. As a start, I suggest the following:

1. To control manifest sample bias in the Midwest, first we must gauge accurately the true abundance of fluted bifaces across the region and elsewhere, probably via several large-scale probabilistic surveys in selected parts of the Midwest. Indeed, we should conduct several kinds of intensive surveys. First, we might survey areas where many Paleoindian sites and bifaces are known to exist. Counties like Coshocton and Muskingum in Ohio, Genesee and Monroe in Michigan, and Mills in Iowa come to mind but are by no means the only candidates. There, we should repeatedly survey as much ground as possible. This is not the place to discourse on survey methods (e.g., Shott in press); it suffices to note that reliable results are achieved only after five or more surveys following an equal number of cultivations. We also should survey single tracts in the same season but after each measurable rainfall, because cultivated surfaces weather and expose more artifacts with each storm.

2. We also should conduct probabilistic surveys of larger areas, perhaps watersheds, even where concentrations of Paleoindian remains are not reported. These surveys are likely to yield more variable results, with considerable evidence found in some places but little in others. Probabilistic surveys were fashionable in the 1970s but no longer are. This is perverse and tragic because the need for such work persists and its value is undeniable.

3. Completing both kinds of intensive surveys, we then can calibrate our systematic results to the sample documented by collectors, itself differently systematic in the aggregate, not unsystematic. Because collectors have the advantage of location and local knowledge, their samples surely are now and will be larger than our systematic ones. This difference is less important than determining whether our systematic samples differ from theirs in character. If the two agree in distribution and both are independent of modem population and other possible biases, then we will have compiled a reliable fluted-biface (and site) sample. If they do not, then our probabilistic and other intensive surveys perhaps are more reliable samples, as they will be free of biases like the distribution of modem population. As daunting as they are, the analytical challenges of such research pale by comparison to practical ones. Finding the necessary funds and institutional support will be the greatest challenge of all, as both funding sources and institutions are oriented toward smaller-scale and shorter-term research.

4. Generally, the Midwest is extensively cultivated. Although extent of cultivation did not covary with biface count in this study, it is worth examining this relationship on a larger scale. Wider nationwide variation in extent of cultivation might free this variable to covary with biface count more than in the Midwest. At the same time, those who doubt sample bias have the opportunity thereby to discount cultivation’s effect, if it does not register on that larger scale.

5. But local collectors have found most fluted bifaces by far. Because they are many, energetic, and widely distributed, collectors also will find most bifaces in the future. Both to gauge the size of the fluted-biface population and to compare the number and distribution of collectors to the general population, there is no substitute for massive, systematic canvassing of the thousands of private collections across the Midwest, a task that the professional community has not adequately attempted. A few collectors are apt to resist, but most seem willing, some eager, to share their knowledge and collections, so long as we treat them with due respect. Collectors and archaeologists are not natural enemies.

There is no doubt that “continued compilation of primary data on Paleoindian projectile points… is absolutely crucial to understanding human occupations” (Anderson and Faught 1998:163) during colonization of the New World. In this connection, Anderson and O’Brien’s (1998) plea bears repeating. All future discoveries should be so compiled. But merely increasing the already-considerable corpus of data is not sufficient. Large samples that are biased in unknown ways and to unknown degrees are no better-merely bigger-than small samples possessing the same qualities. Nor are larger samples necessarily less biased just by virtue of their size. Besides compilation, we must critically evaluate existing information for its biases. All archaeological data are biased to some extent; fluted-biface distributions are not uniquely deficient in this respect. Critical evaluation requires that we consider sampling effects from the distribution of modern population and other sources. When they are found, we must gauge their magnitude and try to control for them. We cannot avoid bias but neither should we ignore it for this reason. Instead, we must understand it. With the benefit of this knowledge, then we can interpret the true or underlying distribution and abundance of fluted bifaces as a register of Paleoindian action, not partly of modern sample effects.


Colleen Eck and Scott Anfinson advised in the use of state records from Iowa and Minnesota, respectively. Comments from David Anderson and Mark Seeman improved the article, although I alone am responsible for its content and those readers do not necessarily share my views. This article originated as an extension of a synthesis of midwestern Paleoindian archaeology for the “Earliest Americans Theme Study,” a joint Society for American Archaeology-National Park Service project. I thank David Brose for inviting my participation. John Doershuk and Timothy Weitzel of Iowa’s Office of the State Archaeologist drafted Figure 1 using ArcView 3.2.


Amick, Daniel S., Robert F. Boszhardt, Kenneth Hensel, Matthew G. Hill, Thomas J. Loebel, and Dean Wilder

1999 Pure Paleo in Western Wisconsin. Report of Investigations No. 350. Mississippi Valley Archaeology Center, University of Wisconsin-La Crosse. Anderson, David G.

1990 The Paleoindian Colonization of Eastern North America: A View from the Southeastern United States. In Early Paleoindian Economies of Eastern North America, edited by Kenneth B. Tankersley and Barry L. Isaac, pp. 163-216.

Research in Economic Anthropology, Supplement 5. JAI Press, Greenwich, Connecticut.

Anderson, David G., and Michael K. Faught

1998 The Distribution of Fluted Paleoindian Projectile Points: Update 1998. Archaeology of Eastern North America 26:163-187.

2000a A North American Paleoindian Projectile Point Database Website. Online: http: // 2000b Palaeoindian Artefact Distributions: Evidence and Implications. Antiquity 74:507-513.

Anderson, David G., and Michael J. O’Brien

1998 Missouri Paleo-Indian Projectile Point Recording Project: A Call for Data. Missouri Archaeological Society Quarterly 15(3):4-9.

Anfinson, Scott F.

1997 Southwestern Minnesota Archaeology: 12,000 Years in the Prairie Lake Region. Minnesota Historical Society, St. Paul.

Archaeological Survey of Missouri

1997 Recorded Archaeological Sites by County in Missouri. Jefferson City. Online:

Blackmar, Jeannette M.

2001 Regional Variability in Clovis, Folsom, and Cody Land Use. Plains Anthropologist 46:65-94.

Boszhardt, Robert F.

1991 Paleoindian Study Unit: Region 6, Western Wisconsin. The Wisconsin Archeologist 72:155-200.

Brennan, Louis A.

1982 A Compilation of Fluted Points of Eastern North America by Count and Distribution: An AENA Project. Archaeology of Eastern North America 10:27– 46.

Brose, David S.

1994 Archaeological Investigations at the Paleo Crossing Site, a Paleoindian Occupation in Medina County, Ohio. In The First Discovery ofAmerica: Archaeological Evidence of the Early Inhabitants of the Ohio Area, edited by William Dancey, pp. 61-76. Ohio Archaeological Council, Columbus.

Carskadden, Jeff, Mark Faulhaber, and Brian DaRe

2000 A Survey of Fluted Points from Muskingum County, Ohio. Ms. on file, Muskingum Valley Archaeological Survey, Zanesville, Ohio.

Chapman, Carl H.

1975 The Archaeology of Missouri, Volume 1. University of Missouri Press. Columbia.

Cowgill, George L.

1970 Some Sampling and Reliability Problems in Archaeology. In Archeologie et Calculateurs: Problemes Semiologiques et Mathematiques, edited by Jean– Claude Gardin, pp. 161-175. Centre National de la Recherche Scientifique, Paris.

Dyke, Arthur S., and Victor K. Prest

1987 Late Wisconsinan and Holocene History of the Laurentide Ice Sheet. Geographie Physique et Quaternaire 41:237-263.

Fitting, James E., Jerry DeVisscher, and Edward J. Wahla

1966 The Paleo-Indian Occupation of the Holcombe Beach. Anthropological Papers No. 27. Museum of Anthropology, University of Michigan, Ann Arbor.

Higginbottom, Daniel K.

1996 An Inventory of Fluted Projectile Points from Minnesota. Paper presented at the 54th Plains Anthropological Conference, Iowa City.

Hosfield, Robert

1999 The Palaeolithic of the Hampshire Basin: A Regional Model of Hominid Behaviour during the Middle Pleistocene. BAR British Series 286. Archaeopress, Oxford.

Jenks, Albert E.

1937 Minnesota’s Browns Valley Man and Associated Burial Artifacts. Memoirs No. 49. American Anthropological Association, Menasha, Wisconsin.

Kodric-Brown, Astrid, and James H. Brown

1993 Incomplete Data Sets in Community Ecology and Biogeography: A Cautionary Tale. Ecological Applications 3:736-742.

Kouba, Theodore F.

1985 Letter. The Wisconsin Archeologist 66:77.

Largent, Floyd B.

1995 Some New Additions to the Texas Folsom Point Database. Plains Anthropologist 40:69-71.

Largent, Floyd B., Michael R. Waters, and David L. Carlson

1991 The Spatiotemporal Distribution and Characteristics of Folsom Projectile Points in Texas. Plains Anthropologist 36:323-340.

Lepper, Bradley T.

1983 Fluted Point Distributional Patterns in the Eastern United States: A Contemporary Phenomenon. Midcontinental Journal of Archaeology 8:269-285.

1985 The Effects of Cultivation and Collecting on Ohio Fluted Point Finds: A Reply to Seeman and Prefer. Midcontinental Journal of Archaeology 10:241-250.

1988 Early Paleo-Indian Foragers of Midcontinental North America. North American Archaeologist 9:31-51.

Mason, Richard P.

1988 Preliminary Report on the Fluted Point Component at the Aebischer Site (47 Ct 30) in Calumet County, Wisconsin. The Wisconsin Archeologist 69:211– 226.

Mason, Ronald J.

1958 Late Pleistocene Geochronology and the Paleo-Indian Penetration of the Lower Michigan Peninsula. Anthropological Paper No. 11. Museum of Anthropology, University of Michigan, Ann Arbor.

1962 The Paleo-Indian Tradition in Eastern North America. Current Anthropology 3:227-278.

Meltzer, David J., and Michael R. Bever

1995 Paleoindians of Texas: An Update on the Texas Clovis Fluted Point Survey. Bulletin of the Texas Archeological Society 66:47-81.

Morrow, Juliet E.

1996 The Organization of Early Paleoindian Lithic Technology in the Confluence Region of the Mississippi, Illinois, and Missouri Rivers. Ph.D. dissertation,

Department of Anthropology, Washington University, St. Louis. Morrow, Toby A., and Juliet E. Morrow

1994 A Preliminary Survey of Iowa Fluted Points. Current Research in the Pleistocene 11:47-48.

Munson, Patrick J.

1990 Folsom Fluted Projectile Points East of the Great Plains and Their Biogeographical Correlates. North American Archaeologist 11:255-272.

Munson, Patrick J., and Kenneth B. Tankersley

1991 Early Paleoindian Settlement in the Grand Prairie: A Case Study in De Witt County, Illinois. Paper presented at the 56th Annual Meeting of the Society for American Archaeology, New Orleans.

National Agricultural Statistics Service

1999 Census ofAgriculture Volume 1: United States Summary and State Data. United States Department of Agriculture, Washington, D.C.

O’Brien, Michael J., and W. Raymond Wood

1998 The Prehistory of Missouri. University of Missouri Press, Columbia.

Office of the State Archaeologist

2001 State of Iowa Archaeological Site Counts by County and by Township Section. Iowa City. Online: htm.

Office of the State Archaeologist of Minnesota

2001 Distribution of Recorded Archaeological Sites in Minnesota. St. Paul. Online:

Overstreet, David F.

1991a Updated Paleoindian Study Unit for Region 7. The Wisconsin Archeologist 72:201-244.

1991b Paleoindian Study Unit: Region 9, Southeastern Wisconsin. The Wisconsin Archeologist 72:265-366.

Prufer, Olaf H., and Raymond S. Baby

1963 Palaeo-Indians of Ohio. Ohio Historical Society, Columbus.

Quimby, George I.

1958 Fluted Points and Geochronology of the Lake Michigan Basin. American Antiquity 19:317-331.

Seebach, John D.

2000 Drought or Development? Patterns of Paleoindian Site Discovery in the Western United States. Paper presented at the 65th annual meeting of the Society for American Archaeology, Philadelphia.

Seeman, Mark F.

1994 Intercluster Lithic Patterning at Nobles Pond: A Case for “Disembedded” Procurement Among Early Paleoindian Societies. American Antiquity 59:273– 288.

Seeman; Mark F., and Olaf H. Prufer

1982 An Updated Distribution of Ohio Fluted Points. Midcontinental Journal of Archaeology 7:155-169.

1984 The Effects of Cultivation and Collecting on Ohio Fluted Point Finds: A Cautionary Note. Midcontinental Journal of Archaeology 9:227-233.

Shott, Michael J.

1993 The Leavitt Site: A Parkhill Phase Paleo-Indian Occupation in Central Michigan. Memoirs No. 25. Museum of Anthropology, University of Michigan, Ann Arbor.

2001 Midwestern Paleoindian Context. Earliest Americans Theme Study, edited by David S. Brose. Society for American Archaeology and the United States National Park Service, Washington, D.C.

in press Data Quality in Survey Sampling. Archaeological Review from Cambridge 19.

Simons, Donald B.

1997 The Gainey and Butler Sites as Focal Points for Caribou and People. In Caribou and Reindeer Hunters of the Northern Hemisphere, edited by Laurence J. Jackson and Paul T. Thacker, pp. 105-131. Avebury, Aldershot, England.

Simons, Donald B., Michael J. Short, and Henry T. Wright

1984 The Gainey Site: Variability in a Great Lakes Paleo-Indian Assemblage. Archaeology of Eastern North America 12:266-279.

Steele, James, Jonathan Adams, and Tim Sluckin

1998 Modelling Paleoindian Dispersals. World Archaeology 30:286-305.

Stoltman, James B., and Karen Workman

1969 A Preliminary Study of Wisconsin Fluted Points. The Wisconsin Archaeologist 50:189-214.

Tankersley, Kenneth B.

1999 Sheriden: A Stratified Pleistocene-Holocene Cave Site in the Great Lakes Region of North America. In Zooarchaeology of the Pleistocene/Holocene Boundary, edited by Jonathan Driver, pp. 67-75. BAR International Series 800. Archaeopress, Oxford.

Tankersley, Kenneth B., Edward E. Smith, and Donald R. Cochran

1990 Early Paleoindian Land Use, Mobility and Lithic Exploitation Patterns: An Updated Distribution of Fluted Points in Indiana. North American Archaeologist 11:301-319.

United States Census Bureau

2000 County Population Estimates for July 1, 1999 and Population Change for April, 1999 (CO-99-2). Population Estimates Program, Population Division, Washington, D.C. Online:

Whittaker, John C., and Michael Stafford

1999 Replicas, Fakes, and Art: The Twentieth Century Stone Age and Its Effects on Archaeology. American Antiquity 64:203-214.

Wiant, Michael D.

1993 Exploring Paleoindian Site Distribution in Illinois. Illinois Archaeology 5:108– 118.

Wright, Henry T., and William B. Roosa

1966 The Barnes Site: A Fluted Point Assemblage from the Great Lakes Region. American Antiquity 31:850-860.

Zurel, Richard L.

1979 A Paleo-Indian Site in Monroe County, Michigan. Paper presented at the annual meeting of the Eastern States Archaeological Federation, Ann Arbor.

Michael J. Shott

Department of Sociology, Anthropology and Criminology

University of Northern Iowa Cedar Falls, IA 50614-0513

Copyright Office of the State Archaeologist Spring 2002

Provided by ProQuest Information and Learning Company. All rights Reserved