Chapter 2
Review of the Archaeological Record
Kenneth M. Ames

Review of the Archaeological Record: Continuities, Discontinuities and Gaps

Radiocarbon Dates and population Dynamics in Time and Space

Evidence from Radiocarbon Dates.
Archaeologists have long been interested in population changes on the Plateau. Early workers were aware, for example, of greater numbers of sites post-dating 2000 years ago, and assumed that reflected significant population growth. Coupled with this has been an interest in population dispersion: i.e. where were people concentrated during particular times during the Holocene and why? Of course, population dispersion is related to settlement patterns, and the long-standing interest in establishing when the Early Modern (AD 1720 – AD 1850) settlement pattern system developed. This interest has been hampered until the last decade or two because most archaeology on the Plateau was focused in areas to be flooded by dams. Thus while there was considerable knowledge about occupation along the rivers, there was correspondingly little about occupations elsewhere. This has changed, particularly in some portions of the Central Basin, and the Southern Highlands.

Earlier estimates of population size and dispersion were based on the spatial and geological distributions of diagnostic artifacts, house pits, and sites, as well as the numbers and sizes of house pits and sites. All of these measures are affected by a variety of site formation processes, including erosion. Hammatt (1976), for example, suggests that variation in site numbers during some periods in the Lower Snake River region are the result of the river's depositional cycles. Researchers have more recently turned to radiocarbon dates as a surrogate measure of human populations. The basic assumption in using radiocarbon dates as an indirect measure of human numbers is that there is a relationship between the number of people in an area, or during a period of time, and the amount of archaeological charcoal that is present, i.e. more people = more charcoal. There are two additional assumptions: there is a direct, positive relationship between the amount of archaeological charcoal and the number of radiocarbon dates; and that there has been no loss of charcoal with time. (e.g. Rick 1987. See also Erlandson and Moss [1999] who discuss some of the formation process issues associated with use of radiocarbon dates.).

Chatters (1995) has pointed out that destruction of old archaeological charcoal by a variety of processes and sampling error will produce a sample of 14C dates heavily biased towards recent than older dates. With regards to sampling error, the available sample of early sites primarily reflects a combination of skill and luck, in the sense that there has been no structured search for "old dirt" across the Plateau. Thus, old deposits are likely to have been "under–sampled." To control for loss of charcoal with increasing age resulting from site formation and destruction processes, Chatters proposes a correction factor. However, as yet, he has not provided the empirical basis for that correction. On that ground alone, I elected not to correct for loss. Further, the inherent "spikiness" of a plot based on a small sample will be exaggerated by any correction factor. Finally, the correction factor removes us one more step from the original data and may itself be subject to unknown sources of error. On the other hand, I agree with Chatters that great caution is needed in interpreting 14C data sets.

For this report, a database of 969 radiocarbon dates was developed. These dates are primarily from the radiocarbon databases of the Washington and Oregon SHPOs. These databases are for their entire states. The dates used here are from the counties covered by this study: in Washington, all counties east of the Cascade mountains, while in Oregon, the counties along the Columbia and Snake Rivers. To this sample was added the catalogue of radiocarbon dates in Sappington (1994) for the Clearwater River drainage. The Washington/Oregon databases are current to 1998, Sappington's to 1994. The Oregon database did include Paulina Lake (Connolly 1999). The Washington database was further limited by removing all except charcoal dates. Thus, for example, the data base for this report does not include the early shell dates from Marmes (Sheppard et al. 1987), or the bone dates from Lind Coulee. The Oregon database did not provide information on dated material, so the dates used here still may include shell, soil, and other materials in Oregon. The exclusion of shell and bone dates from the database reduces the number of very early dates by removing, for example, nine early shell dates from Marmes, two bone dates and a soil date from Lind Coulee, among others. Altogether, 80 shell dates were removed, spanning a period from 1210±70 to 10810±275. The majority of shell dates predate 4000 BP. Archaeologists chose shell to date when they need a date and charcoal is unavailable. The removal of shell and other non-charcoal dates is appropriate for a number of reasons, including the underlying assumption of these analyses that the amount of archaeological charcoal is a surrogate measure of human numbers. The graphs also do not include the very early dates from Cooper's Ferry listed in Table 1.

All dates were calibrated using the OxCal version 3.3 calibration program. This program was selected because of ease of data entry and reading results. Only two-sigma age ranges were used. Where dates were calibrated in the sources, but only single sigma age ranges provided, the dates were recalibrated. One set of graphs use uncalibrated dates, for reasons explained below.

The dates are graphed here in three ways. In the first pair of graphs (Figures 13 14) the uncalibrated intercept dates are plotted in 50-year increments. In the first of these, a two–unit (100 years) moving–average line was fitted to the raw counts. In the second, the percentage of the total dates was plotted/50 year period. In the third graph (Figure 15) the total sample of calibrated dates is graphed as high-low graphs. In the fourth graph (Figure 16), the number of dated sites/250 year period is plotted. Usually when this is done (e.g. Chatters 1995), intercept dates are used, and the number of intercept dates/time unit is charted. However, this procedure can be biased when sites have produced a large number of dates in a particular time span. The alternative is to count sites with dates during a particular period; thus a site with one date and another with 300 during the period each count as one. With age spans, the situation is a little more complicated. In this case, we used 250-year spans, and counted sites with calibrated age spans that fell entirely within a 250-year period. If a site had a single date, with a wide sigma, it could fall into several 250-year periods. It was counted as one site in each increment. This has the effect of exaggerating the presence of that site, but to eliminate dates with wide age-ranges would eliminate many more early dates than had already been removed. Using age spans also has the effect of smoothing curves, and making them more conservative. Finally, a series of radiocarbon date plots developed by Chatters (1995) (Figures 17 & 18), Hess (1997) (Figures 19 & 20), and Ames (Ames 1991) (Figures 21 & 22) are presented and discussed.

The uncalibrated dates were plotted for two reasons: first, they are the raw data upon which this entire discussion is based. Secondly, most of the discussion in the literature is based on calibrated dates. However, calibrating dates (converting radiocarbon dates to calendar dates) is not a straightforward matter, despite the ready availability of calibration programs, such as OxCal. In analyses of the kind discussed here, researchers increasingly use calibrated dates so that the temporal duration of demographic events revealed by plotting the dates is closer to their actual date and duration. However, the act of calibration does not ensure that the patterns in the data reflect events in the past. It may, but calibrating radiocarbon dates can itself create patterning, for reasons relating ultimately to long–term variations in atmospheric ¹⁴C.

The question then is whether this particular data set is robust enough to support inferences about demographic trends or shifts. As would be expected from a small sample the temporal distribution of dates before 5000 BP is spotty (there are only 92 charcoal dates earlier than 5000 BP). Some of the peaks before 5000 BP can be partially accounted for by sampling: multiple early dates from Hatwai, Paulina Lake (Connolly 1999), and 35JE49 (Horne 1995). Five sites are reflected in the peak of dates at c. 9800. This peak includes multiple dates from Hatwai. The peak at 8000 BP also includes a range of sites, including Paulina Lake, Wildcat Canyon, Five Mile Rapids, and others, but also multiple dates from single sites such as Paulina Lake. Site sampling affects the graph. The gaps between dates also probably reflect sampling, particularly given the small sample of excavated sites, and the likely loss of datable charcoal. However, some of the gaps could also be "real" in the sense that they point to a break in the record.

The vast majority of dates fall after 4000 BP (2400 BC); 52 % of all dates are later than 2000 BP (AD 1), and 47 % post-date 1700 BP (c. AD 350). While this overall distribution of dates probably reflects human population sizes, it also is a consequence of other factors as discussed above. The numbers of dates begins to increase irregularly after 4500 BP. This increase occurs in three episodes, one just before 4000 BP, a second, 3500 and 2500 BP, and a third between 1700 and about 700 BP. There is a major dip in dates just before 2000 BP. This dip in dates corresponds to a "kink" in the 14C curve (Figure 23) that may partially account for the relatively few dates in that period, suggesting that the pattern should be regarded with caution.

Turning from the raw dates to the calibrated dates, in Figure 16, the numbers of dated sites are steady between c. 11,000 BC until 8000 BC, when there may have been a slight decline (remember that this plot was constructed in a way that minimized spikiness). This decline is followed by a period between 8000 BC and 4250 BC of higher, but slightly fluctuating numbers. The number of dated sites begins to rise exponentially until about 3000 BC where there is a brief plateau in numbers for 750 years. They then increase, reaching a bimodal peak between 1250 BC and 500 BC. The numbers of dated sites plummet sharply, achieving a nadir between 250 BC and AD 250. They increase again rapidly, reaching a second, complexly bimodal peak between 250 AD and 1000 AD, after which they plummet. The complexity of this latter period is also reflected in the raw dates.

Chatters (1995) plotted calibrated intercept dates in 20-year spans (Figure 17). His graph contains two lines, one representing the calibrated dates, the second the dates corrected for loss of charcoal. His unadjusted plot is similar in its overall shape to my plot of uncalibrated dates, though quantification techniques differ. His correction for progressive loss of charcoal with time has a significant impact on the graph, increasing its overall spikiness before 4500 BP. The apparently small increase in dates around 4000 BP becomes quite large in his graph. The other impact of the correction is to lessen the impact that the large number of post-2000 BP dates has on Figures 13, 14, and 16. The demographic implication of these three figures is that human populations on the Southern Plateau were several orders of magnitude larger during the past 2000 years than they had been previously, the base population was much higher and the region had experienced exponential population growth between 3500 BC and 500 BC. Chatters' figures suggest that while populations were overall higher in the last 4500 years or so, the base population was not orders of magnitude larger than it had been previously.

Most recently, Hess (1997), following Chatters' approach, plotted first the calibrated intercept dates for dates from the Washington and Oregon SHPO radiocarbon databases. That graph (Figure 19) is very similar to the one produced for this study (Figure 13). He then plotted a least-squares best-fit line to the first graph, and plotted the residuals (Figure 20). He also used Chatters' correction factor for charcoal loss with increasing age. Hess' plot shows an early Holocene peak at c. 6000 – 5500 BC. The residuals then fluctuate, but slowly decline (rather than remain steady). Hess' graph shows a marked Holocene nadir at about 4800 – 4600 BC. Hess' early Holocene peak may be an artifact of multiple dates from single sites. His plot does show exponential growth at 3500 BC with a higher subsequent population, as well as fluctuations over the past 3000 years.

What can we conclude about demographic changes on the Plateau from all of this? Before answering that question, I want to raise two other factors that might account for some of the apparent patterning. Site formation processes is one of these factors. Fluctuations in conditions conducive to or inimical to the preservation of sites will be reflected in the graph. Hammatt (1976) argues, for example, that the relatively low numbers of Tucannon sites on the Lower Snake River is a reflection of alluvial cycles, not human numbers. Presumably, since this sample is for a broad region, it should be relatively insensitive to such local conditions.

Another, and important, factor affecting the shape of this graph is settlement patterns, or put another way, our monitoring position. If a small number of people are widely dispersed on the landscape, they will generate more sites; if a larger number are concentrated in a few places, they will produce fewer sites. While our underlying assumption is that number of people = amount of charcoal, our measurement of that will be affected by archaeological sampling practices. Archaeologists may only date a few samples from a massive, deep residential site, while they may run samples from each of many small sites, producing more dates for the smaller population. Finally, if we sample one landform more than we do others, patterning in our sample may reflect fluctuating uses of that landform, not numbers. For example, if the bulk of the radiocarbon sample is taken from river canyons, then the numbers of dates will go down during periods when settlement patterns are focused in under sampled upland areas.

We can return to the question, then, what do these analyses tell us about demographic processes on the Plateau? First, it is clear that the sample is sensitive to plotting technique. This is sharpest in the contrasts between Figure 16, which is based on age ranges, and the figures based on intercepts. The apparent pattern will also be affected by the interval in which dates are plotted (20 years, 50 years, 250 years). This sensitivity to quantification technique suggests that these data are not robust, and conclusions must be drawn cautiously. Given these cautions, the analyses taken together seem to show that populations were lower during the Early Holocene, perhaps falling (or fluctuating) to a low point around 4800 BC. Exponential growth may have begun after 3500 BC, depending on whether Chatters' correction is used or not. Schalk et al. (1995) using dates based on obsidian hydration rinds suggests that the increase in population in the Middle Holocene is real, but that it began about 5000 BC, at least in northeast Oregon. This technique may provide an independent check on the 14C–based curves. Finally, all analyses indicate some degree of demographic flux during the past 3000 years, and even within the last 1500 years.

A more refined analysis could, and should, be done, in which the geographic distribution of dates during particular periods is analyzed as a measure of changing dispersion during different periods. Such an analysis was also beyond the temporal constraints on this study. Reid (1991a, Figure 24) has contrasted the numbers of radiocarbon dates in the uplands and canyon bottoms for the Lower Snake River drainage for the last 5000 years, suggesting the patterns may point to shifts towards the uplands during dry periods, and back towards the canyon bottoms in wet periods. There is evidence for population shifts in the southern highlands.

Based on evidence produced by the Pipeline Expansion Project through central and northeast Oregon, Schalk et al. (1995) conclude that there was what they term a "pause in land use" between 3200 and 2000 BC. Their evidence suggests the region for which they have data was not used or was only lightly used during that period, which corresponds to the beginning of Pithouse 2.

For a more recent period, Endzweig (1994) suggests, based on radiocarbon dates and regional climate records, that there were shifts in population distribution in the southern highlands during the past 2000 years. Endzweig (1994), Pettigrew and Hodges (1995), Schalk et al. (1995), and Hess (1997) all suggest that there may have been a general shift from villages scattered through the uplands to villages concentrated along the rivers, during the past 2000 years (which may explain the increased number of dates, given that most excavated sites are residential sites along the rivers). The uplands were not abandoned, but use may have shifted more to logistical task groups. Schalk et al. (1995) also argue for the formation of major population aggregations – large villages – at this time, using the Miller site (Cleveland 1976, Schalk 1983a) near the confluence of the Columbia and Snake rivers as an example of a very large village that formed at this time. Radiocarbon date plots for the Chief Joseph Reservoir also show gaps in that record (Figure 25).

In reviewing the causes archaeologists propose for these changes, they are all fundamentally ecological, except for the last 1000 years or so when the Numic expansion is sometimes invoked (e.g. Reid 1991a). As far as I am aware, no one has invoked major population incursions to explain periods of growth. However, swift changes in subsistence organization of the kind Bettinger (1994) proposed to mark the Numic expansion in the Great Basin have been proposed (see below).

Other Long-term Evidence for Population Distribution
As noted above (Section 4) the central Columbia Basin is the only portion of the region that appears to have been at least partially unoccupied for any period during the Holocene. During Period 1B (11,000 – 4500 BC), population distribution appears to have been controlled by effective moisture (Ames 1988b), with sites located in areas we would expect to have been moister, or areas that had moist, edaphic conditions (e.g. Lind Coulee). As the climate became increasingly warm and dry after c. 9000 BC, people may have withdrawn from most of the central Basin. One site, Meyers Cave, has a projectile point sequence that appears to span the last 7000 years or so, but it, again, may reflect local edaphic conditions (Bryan 1955). Surveys of the central Basin (Greene 1976, Chatters 1984, Galm et al. 1981) have failed to find materials that date to the period between ca. 9500 BC (or before) and perhaps 4000 BC. Galm et al. (1981) note some materials that may be as old as those latter dates, but it does not appear to be until after c. 1000 BC that people begin to make extensive use of this dry, interior area. As discussed above, the available evidence suggests that what little activity there was in this area prior to c. 4000 BC occurred within about a mile of the Columbia River.

Discussion
While we do not have strong control over the human demography of the Plateau, it is clear from the evidence cited above, and yet to be discussed below, that people have shifted around on the Columbia Plateau landscape during the Holocene. However, there seems to be no evidence of a major Plateau–wide break in occupation during that time.

Material Culture

As observed above, little of Early Modern (AD 1720 – 1850) material culture on the Plateau has received the kind of archaeological or anthropological attention that would allow homologies in the form of both active and passive styles to be traced back in time. Further, the available archaeological sample of perishables for the Plateau is quite small (e.g. Mills and Osborne 1952, Cressman 1960, Swanson 1962). In contrast, on the Northwest Coast, for example, it is possible to trace some styles back in time as much as 4000 years using a combination of materials from both dry (e.g. Holm 1990) and wet sites (e.g. Croes 1989). Even there, however, the archaeological sample of objects bearing design motifs is quite small (Ames and Maschner 1999), limiting the kinds of inferences that can be made. Further, such analysis requires visits to museums and curatorial centers, and study of the objects themselves. Such an analysis has never been done for the Southern Plateau, to my knowledge, though decorated objects do occur archaeologically. In fact, there has been little attention to material culture among archaeologists working on the Plateau.

Early workers (e.g. Daugherty, B. R. Butler, Nelson, Swanson) viewed material cultural style as informing about historical relationships, with an emphasis on temporally sensitive artifacts. However, as noted above, this interest did not translate into a research program that emphasized material cultural distributions in time and space. In the late 1960s and early 1970s, with the shift in interest to questions about land-use, subsistence, and ecology, artifacts were de-emphasized, except as they could inform about subsistence practices. While this may be read as a criticism, it is not. No one could anticipate the necessity of studies such as this one. Additionally, a thorough, coherent study of material culture changes on the Plateau would require a great deal of time and resources to accomplish, absent a long research tradition of such work. Such a study is beyond the scope of the present effort, though it would be very germane. As a consequence, it is not possible to reliably trace most elements of Early Modern material culture back in time, except for a few artifact types that have received emphasis, and except for a very few studies that have focused on doing that.

Yent (1976) analyzed the recovered assemblages from the Wawawai site on the Lower Snake River. Her interest was to compare assemblages that pre and post-dated contact and to test Leonhardy and D. Rice's distinctions between the Harder (500 BC – AD 1000), Piqunnin (AD 1000 – 1720) and Numipu (post – 1720) phases. The site contained three components assignable to the Early Modern (AD 1720 – 1850) and Late Modern (post– 1850) periods. The most recent post-dated AD 1860, a second was thought to date between c. AD 1840 and 1860, and the oldest was dated to the late 1700s based on the presence of trade goods. The pre-contact assemblages were derived primarily from houses and were radiocarbon dated to c. AD 1000, if not somewhat earlier (Yent 1976). Yent also used earlier assemblages from Alpowai (Brauner 1976) in her analysis. These latter assemblages were also from houses and dated to as early as c. AD 1 – 100. Her analysis was restricted to non-perishable tools, such as projectile points, cobble tools, etc.

She concludes there is strong, overall continuity among these assemblages during the period covered by her analysis. The only trends or changes she observed include:

• Changes in the relative proportions of medium and small projectile points (see below);
  
• Absence of L-shaped scapula awls at Wawawai, while they are present earlier at Alpowai;
  
• Matting needles less common at Wawawai;
  
• House pits at Alpowai change from round to oval, and become shallower. Earlier house pits are c. 1 m deep, while later ones are about .6 m – .8 deep;
  
• The changes in house pit profile (becoming shallower) and hearths are absent at Wawawai where house form is stable.

The change in house pit form is widespread on the Plateau and is discussed below in the section (Section 6.4) on houses¹². The changes in projectile point forms are also discussed in the next section. The other trends she observes (excluding the presence of trade goods) could be due to sampling, since only two sites are involved in her study. The artifacts she examined are utilitarian, profoundly utilitarian. It is quite unlikely that these tools would display active style. Any stylistic variation in their form is likely to have been passive style. Given their utilitarian nature it could be argued that the continuity she observed is continuity among a restricted range of activities, among a restricted range of functional alternatives, and therefore the continuity reflects analogies rather than homologies. Many of the forms are widespread and are common types of stone tools, again probably reflecting a limited array of options (there are only so many ways to make an end scraper). However, the continuities she observes include almost the entire material record, and the only changes are ones which occur across the entire Plateau, or which may simply be the result of sampling.

Collins (1997) examined changing gender roles on the Southern Plateau, using burial data. She divided grave goods into utilitarian and non-utilitarian artifacts. While burial practices are outside the scope of this study, she also concludes that there is continuity over this lengthy period.

Projectile Points, Styles, and Function
As in many places in North America, projectile points have been the focus of most analyses of artifact style because they vary in time and space, and, in the absence of other chronological controls, are essential to dating archaeological sites. There are debates in the archaeological literature over what this variability means. Some of it may be functional, while some of it is certainly due to repair and reworking after the object has broken in use (e.g. Flenniken and Raymond 1986, Thomas 1986), changes in hafting methods (e.g. Musil 1988), and in weapon systems (e.g. shift from atlatl to bow and arrow). However, archaeologists also sometimes argue that contemporaneous variation (usually in hafting element) that cannot be explained functionally may reflect ethnic differences (e.g. Reid 1991a). Thus, the presence of the small Desert Side Notched point of the Great Basin is often taken on the Plateau to be evidence for either contact with, or the presence of, Numic speakers including Shoshone and Northern Paiute.

Despite the importance of projectile points, there has been remarkably little analytical attention paid to them on the Plateau. Lohse's work (Lohse 1985, 1995) remains the only formal analysis of a large collection of Plateau projectile points. The foliate Cascade point is an important example here. As a type, it was first defined by B. R. Butler (1961), and the definition refined by Nelson (1969). The type is central to both the notion of the Old Cordilleran Culture concept, and Leonhardy and D. Rice 's Cascade phase (7000 – 4500 BC). It is a major time-stratigraphic marker. However, foliate points have a very broad time span in the Pacific Northwest, sometimes being recovered in sites much younger than the end of the Cascade phase (e.g. Baxter 1986). This sometimes has been interpreted as "cultural conservatism" (B.R. Butler 1962). However, often points are classed as Cascade points without reference to the original type definition, nor, aside from Lohse' analysis, have there been studies of variability within the class "Cascade Point." Thus, there is no guarantee that different collections of artifacts classed as Cascade points contain typologically comparable specimens. These comments can be extended to almost all classes of projectile points on the Plateau, and have been (Lohse 1995).

However, the picture is not quite so bleak. First, Lohse (1985) did an extensive analysis of a large sample of projectile points, which has internal consistency. Second, most archaeologists working on the Lower Snake River during the 1960s – early 1990s were trained at Washington State University, primarily by Frank C. Leonhardy and his students, producing a common approach to classification. While this does not ensure typological consistency, it indicates at least a common approach. Unfortunately, there has been no study of Eastern Plateau projectile points similar to Lohse's.

Lohse (1985) defined 19 projectile point styles for the Western Plateau (essentially our South-central Plateau) and established their broad temporal distribution (Figure 26). This temporal distribution of types, very broadly defined (i.e. large stemmed, Cascade, Large Side-Notched, etc) extends across the Plateau, though details differ. Windust (stemmed, lanceolate) points (Figure 9) are the earliest, as they are everywhere on the Plateau. Lohse recognized three varieties of Cascade point (Figure 26), although some analysts might include his Cascade B as a Windust variant. Lohse regards his Cascade C as the "classic" Cascade point of B. R. Butler (1961) and Cascade A as a common variant. Both were in use a long time. Evidence from Marmes Rockshelter and Paulina Lake indicate a relatively smooth shift from Windust to Cascade forms during this period of time (Figures 27 and 28).

Cold Springs side notched points are the first notched points on the Plateau and are present in deposits above the Mazama ash, though they also occur below Mazama ash in sites in Hells Canyon (Pavesic 1971, Randolph and Dahlstrom 1977). Lohse (1995) and others regard them as the Plateau variant of Northern Side notched points (see Lohse 1995), which also occur at this time in the Southeastern Plateau, where they are the only trait separating the Early Cascade (7000 – 5700 BC) and Late Cascade (5700 – 4500 BC) subphases in Leonhardy and D. Rice 's original formulation (Leonhardy and D. Rice 1970, Bense 1972). The style appears to have originated in the Great Basin. Lohse (1995) regards them as part of a widespread shift to notched hafting elements that begins in the Early Holocene. The foliate lanceolate points drop out of collections after 2000 BC (but not completely, e.g. Ames et al. 1981¹³). Medium-sized stemmed and corner notched forms appear and persist until c. AD 1, when they are replaced by a variety of small stemmed and notched forms.

As noted above, in the Eastern Plateau, where the sequence is longer, Windust points may be present as early as 11,900 BC, though certainly by 11,000 BC. Cascade points are present at Hatwai in sediments dating as early as 10,000 BC. The major difference in regional projectile point sequences, however, appears to be during the Middle Holocene, or the Late Cascade(5700 – 4500 BC)\Tucannon (4500 – 500 BC) phases, when at least one point type not included in Lohse's analyses is common. According to Leonhardy and D. Rice , the type projectile points for the Tucannon phase include a point with an expanding stem, side-corner notching, and short barbs (Leonhardy and D. Rice 1970, 11). This is Lohse' Columbia Corner notched A (Lohse 1985, 349). Leonhardy and D. Rice also describe a point with "a short blade, varying shoulders and contracting stem." Kennedy (1976) found this point to be most common along the Lower Snake River below its confluence with the Clearwater River. These points can be assigned to Lohse's Nespelem Bar type. However, assemblages from about 17 miles below Clarkston (Brauner 1976) up the Clearwater, and into Hells Canyon, contain a very different point, with low side-corner notches, a variable blade, thick cross-section, no shoulders and marked ears on the base, that Ames (Ames 1984, Ames et al. 1981) has dubbed the "Hatwai-eared point". Brauner (1976) first described them at Alpowai, and thought them reminiscent of Elko-eared points in the Great Basin. However, in an analysis using Thomas' approach for measuring Great Basin points (Thomas 1981), Ames (n.d.) has shown the points are metrically not Elko-Eared, though formally similar. This point spans a period from as early as 3800 BC to about 1000 BC.

In addition to this east-west difference, Lohse (1985) notes north/south differences in projectile point distributions on the Plateau at about this period, with Rabbit Island stemmed points more common in sites in the central and northern portions of the Southern Plateau, while Columbia Corner Notched are more characteristic of assemblages on the Southern Plateau. We will return to geographical variation below.

Figure 26 suggests a relatively orderly change of projectile point forms over time, with most forms having some temporal overlap. These changes follow broader patterns of replacement of projectile point styles in western North America (e.g. Lohse 1995), including the appearance of a range of small points around AD 1, although larger forms persist for a few hundred years. These small points are clearly associated with the introduction (or acceptance) of the bow and arrow at that time. However, there are some problems with these inferences.

Deciding whether a point is an arrow point or a dart (atlatl) point depends on the artifact's size, including its weight, shoulder width, and neck width. Thomas (1978) metrically analyzed arrows and darts in a museum collection to determine if points could be consistently assigned to one class or the other on a metric basis. More recently Shott (1997) investigated the utility of four measurements (length, shoulder width, thickness and neck width) to separate arrow from dart points, and determined that shoulder width is the best single measurement for separating arrow from dart points. Mean shoulder width for arrows in his study was 23.1± 4.6 mm, and for darts it was 14.4 mm ±3.4 mm, and for neck widths it was 9.8±2.6 and 15.2 ± 3.3 respectively. The mean measurements for neck width and maximum width are presented in Figure 26 (Lohse's closest measurement to maximum width is shoulder width. From his diagram of where he took measurements, he generally measured maximum width across the shoulders [Lohse 1985]). From these measurements, Plateau dart and arrow points appear to be quite small relative to specimens elsewhere in North America. I did not, however, apply Shott's classification functions to Lohse's measurements to test that suggestion. The small size of these points is also suggested by Ames' analysis of Hatwai-eared points (Ames 1990). Using Thomas' measurements for distinguishing arrows and darts, Hatwai-eared points are arrow size except in their thickness. While it is unlikely that they are all arrow points, this does suggest that hunting tackle on the Plateau has been relatively small and light for the last 11,000 years.

There are some tantalizing correspondences between the timing of the introduction of general projectile point styles and the possible demographic shifts discussed previously. Side and corner notched points first appeared on the Plateau at about the same time as the major episode of population growth around 3500 BC, and the bow and arrow came into use in the period (250 BC – AD 250) between the two apparent peaks in population (or aggregation) during the late Holocene.

Microblades
Microblades are small blades usually struck from a prepared core. At the Ryegrass Coulee site (discussed below), complete blades are less than 2 cm in length. These small blades were hafted into the ends and/or sides of bone, antler and wooden handles and shafts to produce an array of cutting and piercing tools. To the north, microblades are present in interior British Columbia, and on the central and northern Northwest Coast, and are, in fact, diagnostic of the Archaic period (10,000 – 6500 BC) in that region. They are also found in Alaska as well as Siberia (see papers in Carlson and Dalla Bona 1996). They have great antiquity in Northeast Asia. The circum-North Pacific distribution of microblades plays an important role in some scenarios of the peopling of the New World (Dixon 1999, also see papers in West 1996). On the northern Northwest Coast, they are generally absent, with some exceptions, after 4000 BC (Ames and Maschner 1999). Microblades occur on the southern Northwest Coast between about 1500 BC and 600 BC. The temporal distribution of microblades in interior British Columbia is the subject of some ambiguity (Pokotylo and Mitchell 1998). They may be present as early as 6500 BC and fall out of use after 2000 BC. However, there is also some evidence for microblades there during the last 2000 years (Pokotylo and Mitchell 1998).

Microblades and microblade cores¹⁴ are not found throughout the Southern Plateau. They occur in sites along the Upper Columbia River upstream from Vantage and in sites in Grand Coulee (Galm et al. 1981). In the Chief Joseph Reservoir sites, they date between c. 5000 BC to 1000 BC (Campbell 1985). The largest assemblage of microblades on the Southern Plateau was recovered at the Rye Grass Coulee site (Munsell 1968) where 235 microblades and seven complete and fragmentary cores were recovered. The microblades were associated with a late Vantage/Cascade phase (5700 – 4500 BC) assemblage that included Cascade points, Mahkin Shouldered (Lohse 1985) and Cold Springs Side Notched points, edge-ground cobbles and milling stones. The site produced three radiocarbon dates, two on charcoal, one on shell. The earliest charcoal dates are from a composite sample recovered immediately below what was identified in the field as Mazama ash. The date (UW-114) is 6790 ± 340 (cal 2 sigma age span of 6400 – 4900 BC¹⁵), and given the sigma is reasonable for an early date associated with Mazama ash. The shell date (UW-113) is 6480±80. The second charcoal date is 3525±145 (UW-112) (cal 2 sigma age span of 2300 – 1500 BC). This is taken by Munsell to closely date the site's last occupation. While they are uncorrected, these dates are generally contemporaneous with the dates on microblades recovered farther upstream in the Chief Joseph Reservoir.

Microblades and microblade cores have also been recovered farther west in two sites near Mt. St. Helens in the southern Washington Cascade Mountains. Occupations at both sites are contemporaneous with Ryegrass Coulee. Twelve cores and 43 blades were recovered at Layser Cave (Daugherty et al. 1987a) and 19 cores and 23 flakes were recovered from the Judd Peak Rockshelters (Daugherty et al. 1987b). The oldest date at Layser Cave is 6650±120 b.p (WSU 3593) which calibrates to 5780 – 5360 BC. The lowest date at Judd Peak is 5970±100, (5250 – 4550 BC). The Layser Cave represents perhaps 1000 or 2000 years of occupation. The Judd Peak occupation spans the rest of the Holocene. Microblades are present through its deposits.

Sanger (1968) suggests that the Plateau microblade tradition was distinct from those on the coast or to the north, although it must, he thought, ultimately derive from the north. Microblades have sometimes been attributed to Athabascan speakers (their presence reflecting a passive style of tool manufacture associated with Athabascans (see discussion in Pokotylo and Mitchell 1998), and their temporal and geographic distribution taken as a measure of the movement of that language family. However, there is also an array of functional, formal and technological issues that surround them (Campbell 1985 Hicks 1997, papers cited in Pokotylo and Mitchell 1998, papers and bibliography in Carlson and Dalla Bona 1994) – it is not altogether clear how and why they were used, and under what circumstances.

In the Chief Joseph Reservoir sample, their distribution seems to be restricted to sites that are classed as field camps and stations/locations (following Binford's terms [Binford 1981]), not residential sites. However, these sites are more generalized than the specialized localities expected of collectors. Campbell (1985) suggests that microblades might have been employed because they were an efficient means of transporting raw material and producing expedient tools. Hicks (1997) was unable to find support for this argument.

Other Chipped and Ground Stone Material Culture
Aside from projectile points, there are really only three categories of artifacts that are distinctive or unique to the Windust (11,000 – 7000 BC)/Cascade (7000 – 4500 BC) phases: small hand milling stones, edge–ground cobbles, and small, finely–made bone needles. The Cascade phase is further distinguished by a heavier reliance on basalt for tool stone than during preceding and subsequent periods and the presence of the Levallois prepared core technique.

Small milling stones are relatively common in Windust sites sometimes in large numbers (e.g. Warren et al.1963, Connolly 1999), and occur in Cascade assemblages (Bense 1972). They drop out of assemblages in the Middle Holocene, at about the same time that mortar bases and pestles appear in the record. It is usually assumed that this pattern reflects a shift in plant processing, an assumption which may or may not be justified. Large, heavy mortars, mortar bases and pestles are associated with some of the earliest houses on the Plateau, and are present in the record after c. 4000 – 3500 BC. Pestles are generally columnar in profile and not extensively shaped and decorated. Mortar bases are often slabs of stone with the mortar depression in the center. These also occur in other contexts besides houses.

In addition to its milling stones, the pre-Mazama occupation at Paulina Lake also yielded large numbers of what the investigators term "abraders." Some of these are flat stones with abrasion striae, while others are grooved pieces of pumice (Connolly 1999). This is virtually unique for a pre-Mazama assemblage, but suggests extensive working of perishable materials. Some of the abraders with striations may also have served other functions. Such tools do occur occasionally in subsequent assemblages. However, while rare, they occur throughout western North America.

Edge–ground cobbles were key attributes of the Cascade phase (7000 – 4500 BC) (Leonhardy and Rice 1970) and diagnostic artifacts of the Old Cordilleran Culture (B. R. Butler 1961). They do occur in Windust assemblages (at Marmes, for example [D. Rice 1972]). They also occur in assemblages dating to the last 2500 years as well (e.g. Greene 1976, Yent 1976). However, as with Cascade points, the type "edge-ground cobble" may not be consistently applied. Sims (1971), following Sprague and Combes (1966), distinguishes two kinds of cobbles with edge working: edge–ground and edge–battered and proposes several uses for these types, including root grinding and hide working for the former, and as cores for the latter. His distinction is rarely made in subsequent work, however, and so it is difficult to know whether all edge–ground cobbles through time are typologically the same. In early assemblages they are generally taken to indicate plant processing.

Grooved stones, sometimes called "bola" stones, are also associated with Windust (11,000 – 7000 BC) and Cascade (7000 – 4500 BC) phase components (Leonhardy and D. Rice 1970, Figure 3; Bense 1972) and contemporaneous manifestations (e.g. Five Mile Rapids [Cressman et al. 1960). Their function is presently unknown, although they may have been net weights (e.g. Hess 1997). They also may have been bola stones. They are not present in later assemblages. However, a notched (by chipping) cobble net weight is present at Hatwai, in a deposit that probably predates 10,000 BC. Net weights have also been recovered at Kettle Falls, on the Columbia River in the Northern Plateau. These predate the Mazama ash fall. Net weights are very rare in Plateau assemblages until the last 2000 to 3500 years (Johnston 1987).

Generally, most chipped–stone tools on the Plateau are made from a variety of stones archaeologists commonly call "cryptocrystaline"– cherts. Basalt tools that are not cobble tools are rare, except in the Cascade phase (7000 – 4500 BC), when fine–grained basalts were quarried (e.g. Womack 1977) and used, often to produce foliate bifaces of varying form and sizes. While Cascade points were sometimes made of basalt, they are most commonly of chert. Basalt Windust points also occur. The reasons for this preference for basalt are unknown and somewhat controversial (e.g. Andrefsky 1995, Reid 1997). Additionally, the amount of basalt in Cascade assemblages is variable (e.g. Andrefsky 1995). The debate between Andrefsky and Reid also relates to interpretations of Cascade phase (7000 – 4500 BC) mobility: whether Early Cascade (7000 – 5700 BC) peoples were focused towards the river bottoms (Bense 1972) or uplands (see also Morrsion 1996). Muto (1976) originally proposed that the Levallois technique was used as a method of working basalts. Use of the technique appears also to be highly variable, but it seems never to be the most common method of producing and working cores. For example, Bense (1972), in her synthesis of Cascade phase assemblages from the Lower Snake River, reports only nine Levallois cores.

There has been little formal attention to chipped–stone manufacturing techniques on the Southern Plateau. Some local studies have invested considerable effort in it (e.g. Womack 1977), but there have been no general synthetic studies. One of the distinctions Leonhardy and D. Rice (1970) draw between the Tucannon phase (4500 – 500 BC) and earlier phases is in what they see as the craftsmanship of chipped stone tools. Chipped–stone tools from the Windust (11,000 – 7000 BC) and Cascade (7000 – 4500 BC) phases are generally much better made than those of the Tucannon phase, particularly projectile points. (However, assemblages of all of these phases have large numbers of utilized flakes.) They concluded from this, and other evidence, that there was no continuity between the Cascade and Tucannon phases. Tucannon phase chipped–stone technology, at least as reflected in its bifaces, is quite opportunistic (e.g. Nelson 1991). Hatwai–eared points, for example, can be quickly and easily made; they are, in a sense, disposable projectile points. However, these changes are those many lithics specialists expect with increased sedentism (e.g. Morrow and Jeffries 1989, but see also Kelly 1992), which is associated with the Tucannon phase (see below).

Bone Technology
Bone tools are present throughout Plateau prehistory. Early Modern Plateau material culture was characterized by a single point (barb) harpoon, leisters and the needles, shuttles, etc. required to make and maintain nets. A range of awls is present after Windust, but assemblages vary in the types present. Lyman (1976) summarizes the temporal distributions of some bone tool types from the Lower Snake River region. Ulna awls, metapodial awls, and antler wedges are present in the Cascade phase (7000 – 4500 BC), while metapodial awls and wedges are present in the Tucannon phase (4500 – 500 BC). The Harder phase (500 BC – AD 1000) is marked by metapodial awls, L–shaped scapula awls, and perforated elk teeth. The Piqunnin phase (AD 1000 – 1720) (then poorly represented) had no bone tools, while only antler wedges are present in the Early Modern period (post – 1720). However, as Yent (1976) notes, bone tools may have been among the first tools replaced by metal tools obtained through trade.

The temporal distribution of bone tools is difficult to evaluate, since they are generally rare in assemblages, even in those from deposits with good bone preservation. Thus, while compilations such as Lyman's are informative, they are problematic. Further, many bone tool types, such as metapodial awls, are widely distributed in North America and in the world. Scapula awls, on the other hand, are much less common. Such tools are probably analogous cultural traits where encountered, rather than homologous ones. Thus, their presence provides little information of the kind needed here.

Very small bone needles have been recovered in the early deposits at Marmes (D. Rice 1972) and in association with the Buhl burial in southern Idaho (Green et al. 1998). The Buhl burial is earlier than the time under consideration here, but the grave goods associated with her are typologically Windust. Such small needles do not occur in subsequent assemblages.

Randolph and Dahlstron (1977) recovered what they regarded as leister parts and fishhook barbs in pre-Mazama deposits at the Bernard Creek Rockshelter in Hells Canyon. On the Upper Columbia River, toggling harpoons (both the valves¹⁶ and points) are reported to be present in Kartar phase (4500 – 1500 BC) assemblages, although they are exceedingly rare until the Coyote Creek phase (AD 1 – 1800). Barbed points are present at Lind Coulee, and in Kartar and Hudnut (1500 BC – AD 1) components. Generally, in the Pacific Northwest, including the Plateau and Coast, large, barbed points are the earliest such tools found. They are usually too fragmentary to determine whether they are barbed points, or harpoon heads. Toggling harpoons generally appear across the entire region after 6500 BC (Ames and Maschner 1999).

Jorgensen (1980) lists the "crutch–handled" digging stick as a significant trait of the Columbia Plateau. While they do also occur in the Great Basin, they are primarily found throughout the Plateau. These are often made of hardwood with an antler handle (the crutch part). The handles were a common grave good in late pre-contact graves (Hayden and Schulting 1997). The earliest handle that I am aware of is a decorated one recovered at the Hymer site (Draper 1986b) on the Upper Columbia. On the basis of associated projectile points, Draper dates the site to c. 1850 – 500 BC. A significant bone artifact that I cannot trace through time is the small, hollow bone tube traditionally used as part of the stick game (Brunton 1998). These are also present in late graves (e.g. Hayden and Schulting 1997), but have not been traced back through time, if that could reliably be done.

Discussion
Most of the diagnostic artifacts of the Windust (11,000 – 7000 BC) and Cascade\Vantage (7000 – 4500 BC) phases do not carry forward through time. When they do, edge–ground cobbles, for example, questions about typological consistency cloud the issue. On the other hand, the pattern that emerges is of material culture replacement. Tool forms generally do not drop abruptly from the record. Projectile points seem to typify the history of material culture on the Plateau, as in the overlap between the use of darts and bows and arrows, and the overlap between foliate lanceolate points and notched points. Formal variation in projectile points also displays an interesting geographic dimension, which Reid (1991a) has suggested may reflect broadly defined social groups. The limited spatial and temporal distribution of microblades on the Plateau (in contrast with the Cascades) is also interesting, if poorly understood. Forms generally persist for very long periods. Of course, this may in part be a reflection of the capacity of archaeological classifications to measure the kinds of change of interest here.

I have stressed throughout this section that material culture has not received much emphasis in regional archaeological work, despite reports filled with artifact classifications. One area of material culture, houses, is an important expectation. The next section treats houses, community patterns, residential mobility, and settlement patterns. These topics are inextricably mixed in the regional literature.

Houses, Communities, and Mobility patterns

Houses
The appearance of pit dwellings on the Plateau is presently widely seen as evidence for changes in subsistence and mobility strategies, including increased semisedentism. In the past, they were viewed as the diagnostic indicator of the appearance of the "Plateau Pattern," indicating the historical appearance of Early Modern Plateau culture. The presence of houses and villages was assumed to indicate the land-use practices of the immediate pre-horse period, with winters spent in villages and the summers having people dispersed across the landscape in camps and task groups. However, as the initial age of the structures has been pushed back , and their widespread distribution in western North America more fully appreciated, that view has become much less tenable. The timing of the appearance of houses, how they appear, and the associated mobility and subsistence strategies are central issues in Plateau archaeology and very germane to the issues addressed in this study.

Early Modern Houses
H. Rice's (1985) is the most complete study of Plateau houses and other structures. H. Rice distinguishes among long lodges, conical dwellings, excavated dwellings (subterranean and semi subterranean), plank houses, and temporary shelters. He also describes what he terms attendant structures: arbors, scaffolds, and racks. The most common form of long lodge is what he calls an inverted V lodge.

Inverted V-lodges were pole structures with rounded ends. The frame of the house was a series of crossed poles tied together, with their ends placed in the ground. A more elaborate arrangement of poles at the ends of the house and a ridgepole connecting the paired poles provided stability. They were covered with mats of bark or tule. In the southern Plateau, they were often built over a pit about a meter deep. The interior was open, since there were often no central support poles or subdivisions. This open interior usually had a row of hearths in its middle. The houses could be 1.2 m (4 ft) to 1.8 m (6 ft) high and over 9 m (30 ft) long. Some of these structures may have had a central row of interior poles supporting the roof ridge, a variant that H. Rice terms a "double lean–to." Other variants of the long lodge include the flat-topped mat lodge (built at fishing sites), and a gable-roofed form with vertical walls. This latter form was found primarily among the Tenino, Yakima, and Klikitat. Long lodges were the most common form of winter dwelling during the Early Modern (AD 1720 – 1850) Period and were used well into the Late Modern Period (AD 1850 – present).

Conical dwellings include both hide and mat covered structures. Hide–covered structures are tipis, and are commonly thought to have been introduced from the Plains, as part of the cultural complex accompanying the horse. The mat lodge is not a mat-covered tipi. It is built around a foundation cone of three or four poles, which are covered with tule or cattail mats, over which more poles are placed to hold the mats down. While they were used as houses, H. Rice also notes they were used also for storage. The structures were sometimes placed over shallow excavations 30 cm (12 in.) or so deep. The structures range in diameter between about 3 m (10 ft.) to some 6 m (20 ft.). During the Early Modern Period (AD 1720 – 1850), they were used widely on the southern Plateau, particularly as a summer dwelling, and as a house for family sized groups.

H. Rice describes two basic types of excavated dwellings: subterranean and semi subterranean, the former fully underground, the latter only partially. Pithouses are not well documented for the Early Modern Period, when they were not the primary form of dwelling on the Southern Plateau. Ray (1939) provides measurements for some of these structures in the Early Modern Period, and they range in size from 3 m (10 ft.) to 6.1 m (20 ft.) in diameter, and from 1.2 m (4 ft.) to 3 m (10 ft.) in depth. Subterranean houses were deep enough that the roof was laid on the ground, resting on a support of timbers, the whole being covered over by dirt. According to H. Rice 's sources, this kind of structure was variously used as a menstrual hut, a sweat lodge, and a dormitory for boys and unmarried men. It is also possible that such structures were used as storage facilities.

Semi subterranean houses appear to have been far more common. In more northerly areas, probably with heavy snowfall, they had quite substantial roofs, and heavy interior roof supports (Figure 29) while in the Southern Plateau and deep canyons, the roof support was much lighter (Figure 30). In some areas of the South-central and Southwestern Plateau, a dome or hemispherical roof was placed over the pit (H. Rice 1985). Houses with substantial roofs had roof entrances, those with lighter roofs, side entrances.

H. Rice also describes plank houses, which were restricted to the eastern end of the Columbia Gorge, and on down the river. These houses were built of a frame of posts with cedar cladding. Unlike coastal areas, they had bark roofs (see also Hajda 1994). While plank houses were geographically restricted, planking and split wood were used in structures across the Southern Plateau (see Galm and Masten 1985). In addition to these houses, H. Rice discusses temporary shelters (small mat lodges), and other structures. Finally, he develops detailed expectations about how these various structures might appear archaeologically. While I will not rehearse those expectations here, I will refer to them in the following discussion. Of particular concern will be the presence/absence of a pit, the planview, and cross–section of the pits, its size and evidence of a superstructure.

The Archaeological Record of Houses on the Plateau
The earliest documented structure in the region is what is probably a small hut or wind break exposed at the Paulina Lake site (Connolly 1999). It may have been ovoid in planview. The structure had a central hearth. Excavators also recovered the charred fragments of at least five posts. The dwelling is part of a complex of features that included many manuports that might have been used to weigh down the covering. Age ranges for the hearth dates are 10,040 – 9970 BC and 9980 – 9680 BC. Dates for the posts are 9850 – 9490 BC, 9530 – 9440 BC, and 9500 – 9380 BC. While the dates on the posts are slightly younger than the hearth dates, the post dates are extremely tight, and may be taken to date the structure. The older hearth dates may be a consequence of old wood. As has been noted previously, The Paulina Lake Windust occupation contains an array of artifacts, including stemmed and foliate lanceolate points, abraders, milling stones and manos. While it shares these artifacts with other, contemporaneous sites, this is the only excavated Windust site that is clearly residential. The poor organic preservation has probably cost us an interesting bone and antler assemblage, given the large number of abraders, presumably used to shape bone, antler and wood. In any case, it is Connolly's thought that this is a long-term residential base (Connolly 1999), a reasonable inference. Similar but somewhat younger structures are reported for the Upper Columbia region (Chance and Chance 1985b, Chatters 1986a).

The oldest pithouses in western North America are found not on the Columbia Plateau, but in southwestern and south-central Wyoming (Larson 1997), where 28 sites produced some 45 structures. While one structure has a date of 6400 – 5700 BC (7160±150, Larson 1997, Table 3), the majority of many of the dates fall between 4800 BC – 3000 BC (6000 and 4500 B.P). A few structures are present through the rest of the sequence in that area. The early dwellings have interior hearths, as well as interior and exterior storage pits. In these respects they differ from many, but not all, of the earliest houses on the Plateau. Similarities with the earliest Plateau structures include ground stone tools (80% of structures), occasionally, an array of bone tools and ornaments. The ground stone tools appear to be associated with seed processing (she does not specify the type of tools). In reviewing the evidence for seasonality of occupation, she suggests the houses may have been occupied in the summer and then used as storage caches during the rest of year. She also notes similar houses in other parts of the western USA. Among these is a structure at the King's Dog site in Surprise Valley, California, just south of the border of California and Oregon that is dated to c.4850 – 4050 BC (O'Connell 1975, 33).

The oldest pit structures on the Southern Plateau overlap with these in age. The Johnson Creek site (Pettigrew and Hodges 1995), in the southern uplands, produced the oldest dated pit structure on the Plateau. It is an oval structure about 5m by 4m, and 30 to 50 cm deep. It is dated by four dates, the oldest of which is 5500 – 4300 BC (5960±250, Pettigrew and Hodges 1995) from what they interpret as a structural member. The other three are younger and date close to 3800 BC. House 6 at Hatwai also dates to c. 4600 – 2900 BC (Ames et al. 1981). House 1 at Givens Hot Springs in southern Idaho dates to c. 4000 – 2600 BC (Green 1993). After this, a series of sites have pit dwellings that date between 4000 BC and c. 2000 BC, though the bulk of the age spans begin around 3000 BC and end at c. 2500 BC (Ames 1991, Chatters 1995).

Chatters (1989, 1995) and Ames (1988a, 1991) have argued that pit house construction ceased on the Plateau around 1800 BC. Both Chatters' and Ames' plots of dated house floors (Figures 18 & 22) show that gap. After that gap, pithouse construction resumes, and pithouses become ubiquitous in the record, although Ames (1991) argues that there are subsequent fluctuations in pit house construction. However, as observed above with regard to the variation in radiocarbon dates, some of that apparent fluctuation may be due to shifts in settlement patterns into and away from the river canyons. Some archaeologists (e.g. Reid 1991a, Schalk et al. 1998) are skeptical of the reality of the gap at 1800 BC and argue that pithouse construction is continuous after they first appear. Schalk et al. (1998) rightly note that there are many gaps in the regional and local radiocarbon dates (e.g. Figure 25) and the gap in house dates could just be an artifact of that. There is also debate over the causes of their initial appearance, and I will return to that issue below.

Virtually all of the structures before AD 500 are semi subterranean pithouses. The earliest long lodge in the archaeological record was located in the Calispell Valley of northeastern Washington and dates to c. AD 500 (Ames 1991). The appearance of long lodges in the record should be marked by depressions longer than wide, and oval in planview. The appearance of oval plan views in the millennium after AD 1 occurs in as disparate areas as the Lower Snake River (Brauner 1976, Yent 1976) and the southern uplands (Endzweig 1994), and marks the presence of long houses. Small, shallow circular depressions also occur, probably indicating the use of conical mat lodges. Evidence for superstructures is virtually always absent. This has suggested to most archaeologists working on the Southern Plateau that roofs were therefore light. There is little evidence for the deep subterranean houses that H. Rice describes.

Because evidence for superstructures is usually absent, archaeologists have turned to the pit of the house for information about house form and household organization. Plateau pit structures lack the defined floors, hearths and other fittings of Mogollon pit dwellings, for example. Floors are invariably earthen, and sometimes remarkably difficult to recognize in the field. Hearths can be present or absent. When present, they are often only a stain. Structures sometimes have what appear to be stabilizing walls of stone or shell. Some structures have earthen interior benches; many do not. In some sites, such as Hatwai, the structures contain numerous large stones as site furniture, while others may have almost no site furniture. In 1991, Ames examined a sample of 226 house pits (Ames 1991). They displayed considerable variability through time (table 4). Area was calculated using either the diameter of a circular pit, or the length and width. Volume was estimated with area and depth. This sample was collected in the late 1980s and does not include the Johnson Creek house, for example. Nevertheless, it is large enough, particularly for the later periods, that it should still accurately reflect temporal trends.

The earliest houses are rather moderate in area, though the sample is small. The Johnson Creek structure would no doubt lower the mean. Houses during the next period have the largest mean area of any period. Variability in area, as measured by the standard deviation, is similar to that of the next two periods. Between 4000 BC and AD 1, mean house size shrinks, while the standard deviation remains stable. Houses begin to become both larger in area and more variable after AD 1. Volume follows the same general trends, although variability in volume (pit depth) begins to increase after 1250 BC. In addition, minimum volumes become quite small after 2500 B.C. and very small in the last 1000 years. Variability in volume becomes quite strong in that same period, probably reflecting the simultaneous presence of long lodges (some of which could be quite long) and small, conical mat lodges. The changes in mean house sizes and in the amount of variability among houses during any given period must, in part, relate to shifts in household organization (Ames 1991, Reid 1991a). Long lodges could almost be extended infinitely (H. Rice 1985) as groups fluctuated in size through time. They also could be large enough, and sometimes were, to house an entire village in one structure. They were much more flexible in this way than a pithouse could be. Another reason for increasing variation in pit size would be increased functional differentiation of the pits. Many smaller pits classed as house pits may have been storage pits. Evidence presented above indicates that pithouses were used for storage.

Differences in house sizes and form could also reflect emerging status differentials, as Hayden (e.g. 1997b) has argued for the Keatley Creek site on the Northern Plateau. He argues that status inequality emerged across the Plateau during the Late Pacific Period. Archaeologists on the Northwest Coast also argue that differentials of house sizes can reflect status differentials (Coupland 1985) between house groups.

Community size, organization, and structure are usually measured by the number of contemporaneous houses a village contained and the spatial arrangements among the houses (e.g. Warren 1960). All the available evidence suggests that until the last two millennia or so, communities consisted of only one, or perhaps two or three houses (Ames 1991). Large aggregations do not appear to develop on the Plateau until the last 1500 years or so. This occurs seemingly simultaneously on both the Southern Plateau (e.g. Cleveland 1976, 1978; Schalk 1983a) and in British Columbia (Hayden 1997b). Villages were not structured spatially, as they were on the Northwest Coast (Ames and Maschner 1999), though houses were usually laid out along the river bank (Warren 1960).

Mobility Patterns
In 1972, Bense argued that Cascade phase (7000 – 4500 BC) mobility patterns were already very like those of the Nez Perce during the Early Modern (AD 1720 – 1850) period. However, more recent evidence strongly indicates otherwise. In 1988, Ames argued that both Windust (11,000 – 7000 BC) and Cascade peoples were very mobile, and that individual groups probably utilized rather large regions. Obsidian sourcing data from Paulina Lake, for example, indicate the mean distance from which the stone was transported to the site was 50 km, with some distances even greater (Connolly 1999) (Figure 11), in contrast to post-Mazama transport distances, which were much smaller (Figure 31).

Ames also found in his comparative analysis of Windust (11,000 – 7000 BC) and Cascade (7000 – 4500 BC) assemblages that while Cascade period patterns seemed to be those of midlatitude foragers, shifting residential bases to resources, Windust strategies were somewhat more collector-like. Recent work in the southern upland supports that inference, particularly the discovery of the Windust residential camp at Paulina Lake (Pettigrew and Hodges 1995, Schalk et al. 1995, Hess 1997, Connolly 1999).

In contrast, Late-Cascade/Vantage (5700 – 4500 BC) phase mobility and residential strategies are not well known, and there is little evidence for residential camps of any kind, for storage, or other appurtances of collector life. Presumably, people remained foragers during this period.

Several researchers (Campbell 1985c, Lohse and Sammons–Lohse 1986, Ames 1991, Chatters 1995) have concluded that the mobility patterns associated with the earliest houses differed to some degree from Late Pacific (AD 500 – first contact) or Early Modern (AD 1720 – 1850) strategies. Ames (1988a, 1991) and Chatters (1989, 1995) recognize two separate episodes of house pit construction, which Chatters has termed Pithouse 1 and Pithouse 2, while Campbell (1985c) and Lohse-Sammons-Lohse seem to see continuity with subsequent patterns, albeit quite attenuated.

Pithouse 1 appears to represent a period of sporadic house pit construction on the flood plains of the main rivers, as well as upland tributaries. Lohse and Sammons-Lohse, and Chatters see it as what might be termed "settled foraging," or "settled immediate–return." People practiced residential mobility, but were able to maintain residential bases in one place over several years. Campbell argues that they were collectors, albeit weak ones, who exploited rather large territories. She argues that there was some storage, though direct evidence is again weak. Non-residential sites are rare, but do occur in a variety of environments (Ames 1991, Chatters 1995). There are, however, few clear-cut stations or seasonal camps, in contrast to later times. Although this is variable, most residential sites contain mortars and pestles, sometimes quite large ones, sometimes many large ones – indicating some investment in the residential locality.

Pithouse 2 is markedly different. Initially, houses were smaller and occupational duration shorter (only a few years). Annual occupation of residential sites may have been shorter. In Pithouse 1, most investigators conclude the houses were sometimes lived in year–round, while in Pithouse 2, most dwellings were winter houses. There is stronger evidence for logistical mobility, both in the frequency with which residential sites are shifted²⁰, and in the far greater diversity of other sites on the landscape. There are specialized task-oriented sites in the dry central Basin and elsewhere (e.g. Chatters 1980, Campbell 1985c). Pithouse 2 is also accompanied by evidence for storage, greater numbers of fishing tackle and so on. Thus, residential and subsistence patterns are much closer to the historic forms, though, again, there is increasing variation in house size and form after about AD 500.

Discussion
Pithouse 1 appears to have evolved out of previous land use patterns. While Campbell (1985c) and Lohse and Sammons–Lohse (1986) do not consider its beginnings, that conclusion seems implicit in their thinking. Both Chatters and Ames are quite explicit. Ames suggests that pithouses were probably in sporadic use from the earliest occupations in the region. The earliest pit structures in the Western Hemisphere are reported from the coast of Peru (Sandweiss et al. 1998), where they are contemporary with early Windust sites here. That and the widespread distribution of such structures in northeast Asia suggests that either pithouse construction was a part of the cultural repertoire of some of the earliest peoples to enter the Americas, or it has been reinvented many times. Ames also sees the appearance of pit houses on the Plateau as the local aspect of a very wide spread development in western North America. This event is represented by, among other things, the pit houses in Wyoming and shell middens on the Northwest Coast (Ames and Maschner 1999). Ames has not yet, however, demonstrated that. Chatters (1995), in contrast, explains these shifts in terms of regional environmental changes specific to the Plateau. Ames argues that, ultimately, the appearance of pithouses was caused by population growth, while Chatters argues that dramatic population growth followed the beginnings of Pithouse 1.

The not–quite–complete abandonment of pithouses must represent a shift back to full forager mobility. Chatters explains it as one consequence of a major population collapse that was in turn caused by environmental changes that were occurring during the latter years of Pithouse 1, but which reached a "critical mass" at the end of the period. He suggests the crucial shift was increasing seasonality of resource availability, not an overall decline in productivity. He sees the evolution of Pithouse 2 as a very rapid reorganization of the adaptation as a consequence (Chatters 1995, 390). Other workers, writing before the formulation of the earlier pithouse period (e.g. Galm et al. 1981, Schalk and Cleveland 1983, Campbell 1985c) argue that it was the result of population growth, not a response to environmental stress. Ames has yet to publish a model to explain these changes.

These changes in residential strategies are both marked and not marked, by material culture changes. On the Lower Snake River, for example, Pithouse 1 includes both Late Cascade and Tucannon assemblages. Sites such as Hatwai and Alpowai will have both Cascade points and the later side and corner notched forms that Leonhardy and D. Rice (1970) thought to indicate a different cultural tradition. Most of the point styles of the early Holocene²¹, with few exceptions²², do not survive Pithouse 1. Cold Springs side notched points are present before Pithouse 1 and do not persist into Pithouse 2. On the other hand, Nespelem Bar points correspond temporarily with Pithouse 1 and last into Pithouse 2. Hatwai–eared points are present in both Pithouse 1 and early Pithouse 2. Rabbit Island Stemmed A and Columbia Corner Notched A appear at about the end of Pithouse 1 and persist well into Pithouse 2. Edge ground cobbles become less common to rare, with occasional exceptions. Microblades disappear from the western Plateau not long after the beginnings of Pithouse 2. On the other hand, mortars and pestles appear in Housepit I assemblages and persist through the rest of the Holocene.

In sum, there were changes in material culture associated with both Pithouse 1 and Pithouse 2. The changes in projectile point styles probably reflect changes in hunting tackle that were already underway, and which continued. The turnover in projectile point styles is relatively much more rapid after 2000 BC than before. The appearance of mortars and pestles seems to have been an innovation, though pestles are occasionally reported earlier (Bense 1972), and occur widely throughout North America.

Thus, while some of these changes appear to have been abrupt²³, they do not seem analogous to either the Thule or postulated Numic spreads. This does not preclude Simms' "demographic fluidity." Brauner (1976), in his initial discussion of what Ames later called Hatwai–eared points, noted what he saw as similarities between Elko-eared points of the Great Basin, and the Hatwai–eared forms. He suggested this formal similarity might reflect increased contact between Plateau and northern Great Basin peoples in the central uplands. He argued that the Tucannon phase (4500 – 500 BC)was marked by an increased upland subsistence focus, which led to this increased contact. While Ames shows elsewhere (Ames 1990) that Hatwai-eared and Elko points are metrically not the same, this does not rule out the possibilities of contact and movement.

Subsistence and Economy

Introduction
Changes in subsistence have played a major role in archaeologists' explanations of culture change on the Plateau. Reconstructions of ancient economies are based on a mix of evidence, including artifacts, settlement patterns, and faunal and floral remains. Some of this evidence has already been reviewed. This section focuses primarily on faunal and floral remains. Actually, it focuses on faunal remains. Extensive archaeobotanical work on the Plateau is in its infancy (e.g. Lepofsky et al. 1996), and there is no data set of sufficient temporal and geographic scope to be useful here. Inferences about plant use are based almost exclusively on assumptions about artifact functions. There is some direct evidence that will be reviewed below. There are also problems with the record of animal exploitation, which will be briefly discussed below.

Research into Plateau subsistence patterns has been guided by two intertwined sets of questions: 1) When and why did the Early Modern (AD 1720 – 1850) subsistence economy develop? 2) How and why did Plateau subsistence economies change through time? While these seem like the same questions, in ways they are not. In the first, subsistence economy is a marker of the Plateau Pattern, much the same as winter villages are a marker. In this case, the diagnostic traits are heavy reliance on salmon (e.g. Sanger 1967), (or, alternatively, on roots and\or salmon [e.g. Ames and Marshall 1980, Thoms 1989, Peacock 1998]) and storage.

The second approach is more generally grounded in hunter-gatherer theory (e.g. Schalk and Cleveland 1983, Atwell 1989), and drawing on evolutionary ecology (e.g. Hess 1997), although work on the first question can also be so grounded, approaching the first question using the methods of the second approach (e.g. Thoms 1989).

The data requirements for each approach have been different. In the first, at least in the early years of research on the Plateau, the presence of large numbers of salmon bones alone was evidence for the historic pattern (e.g. Cressman et al. 1969), even though the assemblage was perhaps 9000 to 10,000 years old, and subsequent deposits at the site (Five Mile Rapids) lacked salmon bones, or any evidence for fishing. The second approach focuses more on documenting and explaining variability in the record.

At the same time, recovery and analytical techniques have changed. I have been told, for example, that on some early projects the excavators only collected materials they thought identifiable. In reports, species were often listed as present or absent, without information as to which bones were present or how many. In some projects, excavated dirt was dry–screened through 1/4 in. mesh, while in others it was water–screened, probably with different recovery outcomes. More recently, sediment is screened through smaller meshes (most commonly 1/8 in. mesh). The change in mesh size alone has significant effects on recovery, particularly of small–boned animals and many fish species.

The mesh size issue also has important implications for interpreting faunal collections made over many years. Chatters (1995), for example, proposes marked differences in hunting patterns between his Pithouse 1 and Pithouse 2 periods. He argues that Pithouse 1 diets were broader and more inclusive than those of the later period. The faunal data he uses are central to his conclusions that there was a major reorganization of subsistence practices after Pithouse 1. Schalk et al. (1998) demonstrate that much, if not all, of this pattern is the result of differing screen sizes: most of the Pithouse 2 assemblages were screened using 1/8 in. mesh, while most of the Pithouse 1 faunal assemblages were screened through 1/4 in. mesh. They also show a significant difference in salmon recovery using the different mesh sizes.

There has also been an increasing awareness of site formation and taphonomic issues, such as differential bone preservation (e.g. V. L. Butler and Chatters 1994). One issue, for example, is whether a low frequency of salmon cranial bones relative to vertebrae can be used as an indicator of salmon storage. V. L. Butler and Chatters discovered that cranial bones have lower bone densities than vertebrae, and so are more likely to be lost from the archaeological record. Caution in interpreting cranial bone/vertebrae ratios is necessary. On the other hand, V. L. Butler (1990) was able to show that the salmon bones recovered at Five Mile Rapids by Cressman had been butchered by people, rather than being a natural accumulation (Schalk and Cleveland 1983).

Finally, monitoring position becomes an overwhelming issue. Hunter-gatherer economies are not site specific; resources are drawn from a range of habitats. Further, we know that during the Early Modern Period (AD 1720 – 1850) at least, processed foods were traded widely on the Plateau (e.g. Anastasio 1975). The economy in that period was regional in scope. Archaeological sites reflect the subsistence economy in one place, and sometimes at only one period of the year. This issue becomes particularly pertinent here because one aspect of ancient Plateau economies that seems to be emerging as an issue is regional diversity. The general assumption has long been that subsistence economies were uniform across the Plateau. This appears not to be the case, at least during Pithouse 1 times, as will be briefly discussed below. With this preamble, I will briefly summarize a conservative reconstruction of the history of subsistence economies and then equally briefly address salmon, roots, and some problems in interpreting the record.

Plateau Subsistence Economies (Ames et al. 1998)
Early Period IB (Windust) economies harvested a range of animals, including both large (e.g. elk, deer and bison) and medium mammals (e.g. marmots, rabbits) (Lyman 1983, Atwell 1989). Evidence for animal use derives from faunal remains. They probably also harvested plants, but the only evidence for this are manos, small milling stones (presumably for seeds) and edge ground cobbles. Fishing is indicated by the presence of salmon bones at Five Mile Rapids and a notched net sinker at Hatwai.

Later Period IB (Cascade\Vantage) economies are similar, except the medium mammals drop from the economy. Atwell (1989) and others (e.g. Schalk and Cleveland 1983) see this period as marked by a very diffuse, broad economy, while Chatters (1995) suggests it to be somewhat narrower than the following Pithouse 1 (Period II [4500 – 1500 BC) economies. It is during Period II that there may be marked regional differences in economies. Atwell compared faunal remains from the Lower Snake River–Clearwater (Hatwai), with those from the Chief Joseph Reservoir and the Wells Reservoir projects on the Upper Columbia. The Hatwai fauna (which included Hatwai and Alpowai) reflect a strong, almost exclusive focus on deer (74%)²⁴, and to a lesser extent, elk (8%), and canids (5%). Fish remains were rare. At Wells Reservoir at this time, the economy was extremely diffuse, with a broad range of harvested resources, including rabbits (16%), deer, antelope, salmonids, minnow, elk, and suckers in descending order of frequency. The contemporary Kartar phase (4500 – 1500 BC) fauna from Chief Joseph reservoir is about half deer (45%), minnows (16%), salmon (12%), and marmots (8%). Chatters reconstructs the economy of this period as extremely diffuse (he did not include the Hatwai data, while Hatiuhpuh was not available at the time Atwell did his work, and Chatters could not use all of it). This period is also thought to represent a period of increased use of roots because of the presence of hopper mortar bases and pestles in most of the houses of the period (Ames and Marshall 1980, Lohse and Sammons-Lohse 1986).

There is little disagreement among researchers that the subsequent Period III (1500 BC – AD 1720) is marked by a very broad and diverse subsistence base. It is generally argued that it was during this period that heavy reliance on salmon developed and that there is good evidence for storage. Reid (1991a), for example, reviews the available evidence for storage facilities in the Southeastern Plateau, finding that storage pits become numerous at this time and storage caves come into use (Figure 32). It is also during Period III that we see the evidence of increased bison exploitation between c. 500 BC and 500 AD, an issue reviewed above.

Salmon
The available evidence is that fishing, including salmon fishing, occurred continuously on the Plateau from its earliest occupation. I have already described the earliest netweights. Salmon are present during Period IB (11,000 – 4500 BC) at a number of sites along the Columbia, Lower Snake River, and Upper Columbia. Hess (1997) reviews these data, finding fish bones, including those of salmon, reported at 12 of the 50 sites of the period he examined. Additionally, he found mollusks at 11 of the 50, indicating use of riverine resources by Period 1B peoples. On the other hand, a great many sites along the rivers have no salmon bones at all.

Fish, including salmon, are present in all subsequent periods. Using Atwell's figures (Atwell 1989), fish, including salmon, are present in small numbers in Period II (4500 – 1500 BC) assemblages along the Lower Snake River, very common in Intermediate Period assemblages in the Wells Reservoir, and moderately common in Kartar (4500 – 1500 BC) assemblages in the Chief Joseph Reservoir. They become more common in Period III (1500 BC – AD 1720) assemblages in the Southeastern Plateau . These figures are difficult to compare to other areas, because most of these sites were excavated prior to the wide spread use of 1/8" mesh.

In a review of the distribution of fishing tackle, Johnston found that most of it post–dates 2000 years ago. Johnston organized his data temporally by millennia, not by phase. Of 2130 reported netweights on the Southern Plateau, almost 86% post–date AD 1. Interestingly, 53% of the total date to between AD 1 and AD 1000. While quite rare before AD 1, they are present in assemblages after 7000 BC. He documents harpoons as early as 6000 – 5000 BC, and the earliest toggling harpoon he dates between 3000 and 2000 BC. His data did not include the Chief Joseph project where toggling harpoons may be present in the Kartar phase (4500 – 1500 BC). However, bone tools have a similar frequency distribution to that of netweights, not becoming common until after AD 1. He also reviewed the temporal distribution of storage pits with salmon or fish remains. With the exceptions of a couple at Marmes Rockshelter and Five Mile Rapids, all such features post–date AD 1. In sum, Johnston's review suggests that reliance on fishing, presumably salmon fishing, increased dramatically after AD 1.

Johnston's conclusions accord well with the conclusions of others (e.g. Galm et al. 1981, Schalk and Cleveland 1983). Chatters places increased reliance on salmon fishing and storage earlier, with the beginnings of his Pithouse 2 period. Ames (Ames 1991) also places it at about that same, earlier time. There are, however, problems with Johnston's data.

One of these problems is he does not control for site numbers/period. Thus, his data show a decline in the number of netweights after 1000 AD. There is also a decline in the number of radiocarbon dates during that same period. This pattern could be due to increased site loss due to modern development, and therefore fewer netweights and charcoal, or, conceivably, to fewer people. Campbell (1989) has challenged the idea that the heavy reliance on salmon in the 19^th century characterized earlier periods. She suggests that with population loss due to epidemics, people focused more heavily on salmon in the 18^th and 19^th centuries, an argument echoing one made earlier by Craig and Hacker (1947) and Hewes (1947, 1973). Schalk (1986) reviews this notion and finds it unlikely. Schalk also rejects Hewes' suggestion that peoples on the Plateau may have depressed salmon numbers through over-exploitation. But, in any case, Johnston's figures may be sample–size dependent.

In sum, then, there is evidence for fishing on the Plateau perhaps as early as 10,000 BC, but definitely from 7000 BC on. Fish production may have been intensified as early as around 1800 BC but certainly after AD 1. There is some discussion over whether fishing practices of the mid to late 19^th century and early 20^th centuries reflect the intensity of fishing before AD 1800.

Roots
Recent research has shown the importance of roots to Modern (AD 1720 – present) Plateau economies (e.g. Marshall 1977, Hunn and French 1981, Thoms 1987, Hunn 1990, Peacock 1998, and references in all). On the Southern Plateau, these included camas (Camassia quamash), cous (Lomatium cous), several other members of the species Lomatium including Lomatium canbyi, bitterroot (Lewisia rediviva), and other members of the Lily family. Ongoing ethnographic research on both the Southern Plateau (e.g. Marshall 1977, Hunn 1990) and Northern Plateau (e.g. Peacock 1998) also emphasizes the cultural centrality of root collection, cooking and serving.

As observed above, there is very little direct archaeological evidence for plant use. Surrogate measures are therefore used. These include artifacts thought to be plant-processing tools (e.g. mortars and pestles) and features produced by cooking. Among these latter are earth ovens, in which the roots are steamed. Cooking in earth ovens produces pits filled with, or surrounded by, thermally altered rock, charcoal from the fuel and the wet plant material used to shield the roots from heat and produce steam, and, sometimes, the charred roots themselves. The charcoal can be dated to provide reliable dates on when the oven was used.

The largest sample of dated ovens was excavated in the Calispell Valley of northeastern Washington, a valley that has always lacked a salmon run. Thoms and Burtchard excavated a large number of ovens as part of a project there in the mid-1980s (Thoms and Burtchard 1986, 1987). Thoms embodied the data into his dissertation on root-crop intensification by hunter-gatherers generally. The dated ovens span the last 6300 years. There are sporadic dates between about 4300 BC and 1800 BC (Thoms, 1987, 441). Dated ovens peak at 1250 BC, and then decline sharply, reaching a nadir at c. AD 1, after which their numbers increase to AD 500 where they essentially stabilize. The low point in dated ovens corresponds to the same low point in radiocarbon dates in Chatters' and Ames' date samples (Ames, 1991, Chatters 1995) that could be a product of fluctuation of atmospheric ¹⁴C. It could also reflect a population decline. Peacock (1998) dated a series of ovens in southern British Columbia and reviewed other projects there. In that area, the bulk of earth ovens post-date c. 350 BC, and their numbers actually peak at AD 1 (although the sample of dated ovens is relatively small). Elsewhere, in the southern Willamette Valley, Connolly has dated ovens as c. 8000 to 9000 BC, but the great bulk of dated ovens post-date 4500 BC (Connolly et al. 1997). His dates tend to cluster between 4500 and 2000 BC, and after AD 500, although he has a number of ovens dating to c. 1000 BC. Thus the available evidence (which is thin), including both mortars and pestles and dated ovens, indicates the heaviest use of roots after 4500 BC. However, the evidence also indicates at least some use extending back several millennia.

Finally, Peacock (1998) notes one difference between Early Modern root processing areas and older ones. The older ones sometimes contain storage facilities, which are not present in the more recent collection areas.

Discussion
It is clear that, at the most general level of inference, Plateau economies approximated their Early Modern (AD 1720 – 1850) form by c. AD 1, although many crucial economic subsistence practices had become important much earlier. It is also clear that there was considerable regional local variation in economy, which is not well understood by archaeologists. Finally, any inferences that can be made, or conclusions relevant to this study, are severely limited by the nature of the database, and by methodological issues. These later problems are not restricted to the collection and analysis of fauna from archaeological sites; they are, perhaps, just more visible.

Exchange and Interaction

Trade was deeply embedded in Plateau life (e.g. Anastasio 1975, Erickson 1990, Stern 1998b). The region appears to have been integrated into a single interaction sphere with strong ties to the Northwest Coast, the Great Basin, California, and, at least after adoption of the house, the Great Plains. The interaction sphere had at least one, if not more, major nodes. The primary center for exchange on the Southern Plateau was at The Dalles, Oregon, at the upstream end of the Columbia River Gorge and close to some of the best salmon fishing places on the globe. Other centers for trade and exchange were at Lytton – Lilloett at the confluence of the Fraser and Thompson Rivers in British Columbia (Galm 1994b, Hayden and Schulting 1997), Kettle Falls on the Columbia River in far northeastern Washington and at Wenatchee, Washington, near a low pass across the Cascades to Puget Sound (Galm 1994b) (Figure 33). A great range of objects was exchanged, including processed foods (Anastasio 1975), wealth items (Hayden and Schulting 1997), and obsidian and shell, among other things (Figure 34).

The distribution and age of marine shell in archaeological deposits on the Plateau is currently the best-controlled measure of the extent and timing of interaction. There is increasingly intense research on sourcing obsidian, but obsidian is generally rare in assemblages north of the Columbia River. Central and southern Oregon and southwestern Idaho have many exposures of obsidian that were quarried and used over the past 10,000 years or more. Obsidian moved from central Oregon as far north as Puget Sound, southern Vancouver Island, and in one instance, the central British Columbia coast between c. 4800 and 2500 BC (Carlson 1994)²⁵. Chert is the most common tool stone used on the Plateau, but it has proven very difficult to source.

Galm (1994b) reviews the evidence for the presence of obsidian on the Plateau, and finds that it has been present there through most of the sequence. Most Plateau obsidians originate in central and southern Oregon, southwestern Idaho, and northern California. He postulates a long-term exchange relationship with those directions, perhaps at its most intense during the late part of Period IB (Late Cascade\Vantage (5700 – 4500 BC). This connection might be marked, according to Galm, by not only obsidian but by the presence of Northern Side Notched and Cold Springs Side notched points. This is similar to Brauner's suggestion (Brauner 1976) that the similarities between Hatwai–eared points and Elko–eared points also reflect strong southward interaction. He suggests the trade in obsidian was in blanks, and perhaps finished points. There appears to be more obsidian present in sites in later periods and perhaps an expanded trade in obsidian (e.g. Hess 1997). Galm resists plotting the spatial distribution of obsidian by source.

Marine shell was also traded into and around the Plateau from at least 7500 B.C. on. Erickson (1990) exhaustively reviews this trade. Four kinds of marine mussel shells are present in Plateau sites: Olivella biplicata, Dentalium pretosium, Haliotis fulgens, and Glycermis subobsoleta. Anastasio (1975) also lists abalone. Olivella (Olivella biplicata) is present at Marmes Rockshelter in association with Windust phase (11,000 – 7000 BC) burials (D. Rice 1972). Olivella is virtually the only marine shell present on the Plateau until c. 1000 BC – AD 1. The olivella shells show little modification. Erickson (1990) and Galm (1994b) suggest it entered the Plateau primarily from the south, ultimately from California. Erickson notes that the presence and distribution of olivella on the Plateau is quite similar to that in the Great Basin during the same period. However, olivella lives along the Pacific Coast from Vancouver Island to Baja California, although, according to Erickson, it is more common along the Oregon and California coasts than to the north.

After 1000 BC, dentalium becomes increasingly frequent in sites. The primary sources for dentalium in western North American are beds off the western coast of Vancouver Island, where the shell was collected during the Early Modern Period (AD 1720 – 1850) by NuuChahNuulth people. There are no dentalium processing sites in that area, however (Ames and Maschner 1999). Other marine mollusks appear, and there is a greater variety of shaped objects. There may also have been trade in shell adz blades from the coast beginning perhaps 1500 years earlier (Benson 1986). Assemblages are dominated by dentalium until c. AD 1350 – 1750, when the number of marine shell objects and their diversity again increases markedly. The number of forms of beads and other objects also increases. Dentalium was a major trade item in western North America, and was traded across the Rockies and to the Plains. The shift from olivella to dentalium suggests a change in at least some external trade connections around 1000 BC to 1 AD, if not earlier.

In addition to shell, a great many other kinds of objects circulated, including native copper items; objects of steatite, serpentine and nephrite, including adze blades; and whale bone clubs. Hayden and Schulting (1997) document the distribution of these and other objects in Late Period III (AD 1000 – 1720) and Early Modern (AD 1720– 1850) contexts across the Plateau, demonstrated the centrality of The Dalles and the Lytto–Lillooett nodes. However, they also include the distributions of other widespread items of Plateau material culture, such as digging stick handles, arguing they are status markers. Hayden and Schulting suggest that the distributions of these objects reflect the development of status inequalities on the Plateau, which, in turn, would lead to the formation of an interaction sphere. If they were correct in this, then the increasing numbers and variety of shell ornaments entering the Plateau after 1350 AD would suggest that regional elites formed just before that date.

The archaeological data on exchange are limited, despite its importance in the Early Modern Period. What little evidence there is suggests that interaction beyond the Plateau may have been orientated towards the south and west before about 1000 B.C. After that date, exchange with the Northwest Coast became more important, and connections to the south apparently less so. However, the data are not robust.

¹²Reid (1991a) challenges her conclusions, arguing there is greater diversity of house sizes in the Harder phase than during the preceding Tucannon, or subsequent Piqunnin (Late Harder, in Yent's terms).
¹³The presence of Cascade points in younger deposits at Hatwai could be due to sediment mixing caused by construction of houses. There is also evidence at Hatwai of the deliberate collection of older artifacts by more recent occupants.
¹⁴The presence of microblade cores in archaeological deposits is necessary to demonstrate the presence of microblade technology.
¹⁵These calibrated dates should be viewed with great caution, although they are generally comparable with dates on microblades elsewhere in the Pacific Northwest).
¹⁶The object illustrated as a valve in Lohse 1984, plate 3-8, x, pg. 165, does not appear to be a valve, however.
¹⁷All these dates are from Connolly (1999 Table 9.5). I use his calibration results. He reports only the single sigma and intercept calibrations.
¹⁸The proportions of this structure, its shallowness, and overall planview actually fit H. Rice's archaeological correlates for long houses, not a semi subterranean pit house. Unfortunately, there is insufficient super structural evidence to reconstruct the rest of the dwelling.
¹⁹The younger dates from the site were averaged, producing a date of 5089±44, which was then calibrated.
²⁰Residential sites also appear to be occupied for shorter periods of time (Ames 1991).
²¹Cascade and Mahkin stemmed, using Lohse' terms (Lohse 1985) throughout (Figure 26).
²²As noted above, foliate points do persist in many areas. However, it is an open question whether such points are Cascade points.
²³While I agree with Chatters that the changes between his Pithouse 1 and Pithouse 2 were abrupt, I do not necessarily agree with his explanation. I am particularly skeptical of the population collapse.
²⁴Atwell's figures are all Number of Identified Specimens, or NISP.
²⁵Hess (1997) dismisses Carlson's summary of obsidian sources and distributions on the Northwest Coast, but without adequate explanation (citing two papers by Richard Hewes).

Kennewick Home | Ames Home | Next Section >>