The Seagull Bay site—Clovis technology from American Falls on the Eastern Snake River Plain

Introduction

The position of Idaho’s Snake River Plain midway between the southern portal of the ice-free corridor and landfalls available along the Washington and Oregon coasts for immigrants arriving from Asia or Beringia made an ideal location for the earliest colonists to spread throughout the New World. Colonists drifting eastward from the Pacific coast would soon discover that the corridor running from Farewell Bend on the Oregon–Idaho border to South Pass in the Wyoming Basin was the easiest route to the plains. Similarly, people funneling out of Beringia by way of the ice-free corridor might radiate westward along the route staked out by the Anzick, Fenn, Simon, and East Wenatchee Clovis caches.

It is therefore not surprising that most of the Clovis period projectile points recorded in Idaho cluster along the Snake River Plain and adjacent uplands. This article describes one such cluster of Clovis material on the upper Snake River Plain where the Portneuf River and Bannock Creek join the Snake River and American Falls Lake. For clarification, American Falls Reservoir is a larger body of water today and is distinguished from the Pleistocene period body of water referred to as American Falls Lake. Since documentation began in the early 1960s, Clovis period lithic technology has been discovered in this area along with the remains of significant (currently nonassociated) Pleistocene megafauna.

Background

The 228 km² reservoir behind American Falls Dam began to fill in 1927. Shoreline erosion and littoral processes quickly exposed a stratigraphic sequence rich in vertebrate fossils. The sequence was ordered into five major units labeled A–E (Hopkins and Butler, 1961) (see Figure 1). Unit A, the uppermost, is a nonfossiliferous stratum of Late Pleistocene and Holocene loess. The B and E units have high concentrations of Rancholabrean fauna while the intervening C and D units do not. Unit C has been studied for its invertebrate and floral remains while Unit D exhibits a “scant” vertebrate fauna (Pinsof, 1992).

OPEN IN VIEWER

Figure 1.

Generalized profile of stratigraphic units from American Falls near Seagull Bay. Adapted from Hopkins and Butler (1961: 10).

Both the B and E units represent transported and redeposited faunas, although occasional partially complete animals have been recorded. Even these are thought to have arrived in their present positions as “bloat and float” specimens. Pinsof (1992: 3) cites estimates made between 1955 and 1961 that “average erosion rate of the bluffs bordering the reservoir was approximately 5 meters per year.” Unit B and E faunas are thoroughly commingled on the exposed beaches. Clovis lithic scatters and isolates recorded during reservoir surveys have eroded out of the Unit A loessal cap. To date, no direct association has been found between lithics and extinct faunas. In addition, there have been no archaeological excavations carried out at American Falls, only surface survey and collection under the management of the Bureau of Reclamation.

In the surrounding areas, the environment was similar to American Falls Lake in plant and animal life. Swan Lake is a shallow depression in the Portneuf range near Red Rock Pass, the outlet from which late Pleistocene Lake Bonneville flooded into Marsh Creek and the Snake River basin. The lake lies at about 1560 m above sea level. Fossil pollen, spores, and seeds from a core taken here recorded a coniferous forest between 12,090 and 10,800 BP. Pinus contorta and Pinus flexilis were more abundant than today, and the climate was cooler and wetter than present.

The complete skeleton of an extinct peccary, Platygonus compressus, found nearby in the Cache Valley arm of Lake Bonneville, indicates the lake bed supported a grassland fauna at 11,340 ± 50 BP (McDonald and Anderson, 1975). Warmer temperatures between 10,800 and 10,300 BP raised the lower timberline and brought steppe vegetation upslope. After 10,300 BP Douglas fir became more common, Pinus contorta less common, and sagebrush steppe dominated lower foothills and intermontane valleys. Rapid desiccation began at Lake Bonneville 10,800 years ago and the lake was nearly dry 500 years later (Bright, 1966). Between 10,300 and 8400 BP, increasing warmth and dryness had adjusted surrounding vegetation to its present state. Dry, arid conditions similar to the present persisted until 3100 BP, followed by slightly cooler temperatures and an expansion of grassland and contraction of sagebrush between 3100 and 1700 BP. Dry, arid conditions returned by 1700 BP and continue to the present.

During the Late Pleistocene, grazing species dominated the fauna of the upper Snake River plain. Most abundant in the local faunas at Massacre Rocks, American Falls Lake, Rainbow Beach, and Duck Point were several species of bison, camel, dire wolf ground sloth, horse, llama, mammoth, mastodon, and saber-toothed cats. The vegetation around the ancient bed of American Falls Lake was sagebrush and chenopod steppe. Browsers such as mastodon and ground sloth were scarcer and confined to birch/willow/sedge habitat in the riparian strips along the Snake River (Pinsof, 1992). Collection of paleontological specimens from around American Falls Reservoir began in the early 1930s and continues to the present, with approximately 12,000 specimens currently housed at the Idaho Museum of Natural History. The Snake River has occupied its present channel here since the Late Pleistocene.

Two Clovis points have been recovered from a site near the Rainbow Beach locality. A third Clovis point has been recorded, though not fully documented, from lower Bannock Creek (Butler, 1965), and a fourth has been reported but not recorded or recovered from the lower Portneuf River (Butler, 1968). Finally, an eroding cache of bifacial performs of possible Clovis age has been partially reported from the confluence of the Portneuf River and American Falls Reservoir (Butler, 1969). This cluster of Clovis diagnostics along the southeastern edge of American Falls Lake hints that the Pleistocene megafauna recorded in paleontological gravels remained sufficiently concentrated here 13,000 years ago to attract Clovis hunters.

Seagull Bay

The Seagull Bay site lies in a horizon designated Upper Fossiliferous that is well exposed along the southern margin of the American Falls Reservoir. The reservoir occupies much of the bed of ancient American Falls Lake, a natural impoundment created by a lava flow that blocked the Snake River at ca. 72,000 ± 14,000. The lake was breached by the floodwaters of the Bonneville Flood about 17,500 years ago (Hart et al., 2004; Janecke and Oaks, 2011; Jarrett and Malde, 1987; Rhode and Louderback, 2007; Scott et al., 1983). The Upper Fossiliferous deposit caps lacustrine clays and is capped by the Michaud gravels, deposited in depth near the mouth of the Portneuf River. The Upper Fossiliferous horizon has seven radiocarbon ages ranging from 21,500 to 38,000 years ago, and contains a diverse Rancholabrean fauna. The post-Bonneville bed of American Falls Lake probably comprised an archipelago of smaller ponds and marshes that supported waterfowl as well as large grazers, browsers, carnivores, and scavengers. None of these species has been dated to the terminal Pleistocene in the American Falls locality. However, they are known from surrounding areas such as Cache Valley, the Wasden site, and Kelvins Cave (Reid, 2017; Henrikson et al., 2017). The relative contribution to Clovis diet of Pleistocene megafauna, especially Columbian mammoths, continues to be debated by archaeologists. The fact that four Clovis points and a possible Clovis biface cache have been identified in the midst of this Terminal Pleistocene ecosystem suggests that the American Falls locality may eventually provide direct evidence bearing on the issue.

The first Clovis period find recorded from Seagull Bay was on 27 October 1973 by David W Corliss (1973) as part of a shoreline survey of American Falls Reservoir. It was described as a surface scatter of quartzite, ignimbrite, and chert flakes; a white quartzite endscraper; and a gray obsidian fluted point. The endscraper and fluted point were collected. A sketch shows that the scatter extended about 45 m north–south and 90 m east–west along two wave-cut contours between the reservoir margin and a wave-cut “cliff” of stratified sand and silt. The site lies immediately east of the mouth of Seagull Creek. The site record says paleontological specimens of short-faced bear, bison, and horse have been recovered from the same locality about 90 m to the east. The site was then recorded as 10PR89. A table of the recorded surface finds is listed in Table 1.

OPEN IN VIEWER

Table 1.

Clovis artifacts from 10PR89.

Artifact type	Count	Material type	Sourced?	Date found
Projectile Point	1	Obsidian	Yes	1973
Projectile Point	1	Obsidian	Yes	1992
Endscraper	1	Obsidian	No	1973
Biface fragment	1	Quartzite	No	1992
Biface, late stage	1	Obsidian	No	2018
Core	5	Quartzite	No	1992
Core	1	Obsidian	No	2018
Expedient tools	2	Quartzite	No	1992
Formal flake tool	1	Obsidian	No	2018
Flakes	54	Quartzite (2 chert)	No	1992

The site was revisited 19 years later and the 10PR89 site form updated on 25 September 1992 (Douglas, 1992). It had an area of 4200 m². Five cores, 54 flakes, 1 biface edge fragment, 2 expedient flake tools, 1 point fragment, and 1 complete Clovis point were recorded. Except for two chert flakes, the toolstone for all specimens was described as quartzite. The Clovis point was collected and is described below. From Douglas’s site record, it is clear that she was aware that this was the second Clovis point found at 10PR89. However, in the report of survey results (Bruder et al., 1999), another Clovis point was mentioned from Rainbow Beach, a mile north of the Seagull Bay site. The citation is a personal communication from David Corliss in 1975, apparently to one or more of the four-chapter authors. Because the citation occurs in the context of a summary of the Paleoindian results of the survey and fails to mention the Clovis point that Corliss did find at Seagull Bay in 1973, we suspect that this Rainbow Beach reference is an error and actually refers to the earlier Seagull Bay specimen. The 1973 Idaho Museum of Natural History site record and the 1992 Intermountain Antiquities Computer System (IMACS) record that updates it mentions two different Clovis points from the same lithic scatter. There is no ASI record for a Clovis point from Rainbow Beach. The 1992 paleontological survey recorded a Bison antiquus phalange in redeposited Pleistocene gravels at site 10PR88 as well.

The Seagull Bay site continues to produce nondiagnostic material where these Clovis projectile points have been found. In 2018, this included a late stage biface (not clearly Clovis), a formal flake tool, and an obsidian core. The Seagull Bay site, 10PR89, with its continued erosion of artifacts to the present, strongly suggests a significant buried Clovis component.

Description of the Clovis points

10PR89 (1973 collection)

There is a remnant detachment scar from a large percussion flake blank on the first face that was fluted (see Figure 2, bottom). The final thinning on the blade was done by selective, comedial pressure flaking. There are two channel flakes on this face; both step terminated in the flat area left by the remnant detachment scar. These channel flakes originated at detachment points that are proximal to the existing basal margin (using radial fissures to estimate detachment points). Therefore, part of the base was broken when the channel flakes were detached or the base was reworked following fluting.

OPEN IN VIEWER

Figure 2.

Clovis projectile points from Seagull Bay, American Falls Reservoir, Idaho. Top point is 1992 specimen; bottom point is 1973 specimen. Line illustrations by Sarah Moore and adapted from Reid et al. (2015).

Two channel flakes were detached from the second face to be fluted. One traveled far down the blade and ended in a hinge termination. The second removed much of the scar from the first flute but did not travel as far and ended in a small step termination. The detachment point of this last channel flake remains, indicating that it was removed from the extant basal margin. None of the four channel flake terminations was removed by subsequent flaking. Some marginal pressure flaking carries into the channel flakes, but the flutes are largely unmodified. The lateral margins were heavily ground with edge-parallel striations extending for 28.4 mm on one side and 27.3 mm on the other (Table 2).

OPEN IN VIEWER

Table 2.

Measurements of the Seagull Bay Clovis points (from Titmus and Woods, 1991, in mm).

Catalog	Length	Maximum width	Thickness between flutes	Ridge thick A	Ridge thick B	Flute length A	Flute length B	Flute width A	Flute width B	Basal indentation	Basal width
1 FS3	73.5	27.5	7.1	7.1	6.5	9.4	28.6	12.6	19.9	5.9	26.0
none	54.05	29.3	4.8	5.1	4.5	13.5	34.4	22.8	15.8	4.65	29.3

The blade exhibits 10× grinding and rounding on flake ridges and in some of the flake scars at 10× magnification. This wear is clearly cross-cut by pressure flaking around the tip. This is clear evidence for resharpening of the blade; likely after transport created the wear on the surface of the blade. This flaking intrudes into the channel flakes on first face that was fluted but not on the opposite face. This reworking created inflection points at the blade/ground stem juncture on both margins. The widest area on the point is now at the base, suggesting that this point was resharpened to exhaustion as a projectile point.

Clovis point sourcing

Analyses of obsidian were performed at Geochemical Research Laboratory by Dr Richard Hughes on a QuanX-EC™ (Thermo Electron Scientific Instruments Corporation) edxrf spectrometer equipped with a silver (Ag) x-ray tube, a 50 kV x-ray generator, digital pulse processor with automated energy calibration, and a Peltier cooled solid state detector with 145 eV resolution (FWHM) at 5.9 keV. The x-ray tube was operated at differing voltage and current settings to optimize excitation of the elements selected for analysis. In this case, analyses were conducted for the elements rubidium (Rb Ka), strontium (Sr Ka), yttrium (Y Ka), zirconium (Zr Ka) niobium (Nb Ka), and barium (Ba Ka). Iron versus manganese (Fe Ka/Mn Ka) ratios also were computed for each specimen.

Trace element measurements in Table 3 (except Fe/Mn ratios) are expressed in quantitative units (i.e., parts per million (ppm) and weight percent composition (%)), and matches between the artifacts (unknowns) and known geologic obsidian chemical groups were made on the basis of correspondences (at the two-sigma level) in diagnostic trace element concentration values (in this case, ppm values for Rb, Sr, Y, Zr, Nb, and, when necessary, Ba, Ti, Mn and total iron) that appear in the literature (Ambroz, 1997; Hughes, 1986a, 1986b, 1986c, 1988, 1989, 1993; Hughes and Nelson, 1987; Nelson, 1984; MacDonald et al., 1992; Skinner, 1983, 1997) and certain other unpublished data on obsidians in the region (Hughes, 1988, 1989, 1993). Artifact-to-obsidian source (i.e., geochemical type-sensu; Hughes, 1998), correspondences were considered reliable if diagnostic mean measurements for artifacts fell within 2 standard deviations of mean values for source standards. We use the term “diagnostic” to specify those trace elements that are well measured by x-ray fluorescence and whose concentrations show low intra-source variability and marked variability across sources. In short, diagnostic elements are those concentration values allowing one to draw the clearest geochemical distinctions between sources (Hughes, 1990; Hughes and Smith, 1993). Zn and Ga ppm concentrations are not considered “diagnostic” because they do not usually vary significantly across obsidian sources (see Hughes, 1984). This is particularly true of Ga, which occurs in concentrations between 10 and 30 ppm in nearly all parent obsidians in the study area. Zn ppm values are infrequently diagnostic; they are always high in Zr-rich, Sr-poor peralkaline volcanic glasses, but otherwise they do not vary significantly between sources in the study area.

OPEN IN VIEWER

Table 3.

Quantitative composition estimates for obsidian Clovis points from Seagull Bay (10PR89).

Catalog number	Zn	Ga	Rb	Sr	Y	Zr	Nb	Ba	Ti	Mn	Fe2O3T 2O3T	Fe/Mn	Obsidian Source (chemical type)
10PR89-001	nm	nm	180 ± 4	26 ± 3	59 ± 3	228 ± 4	43 ± 3	882 ± 12	nm	nm	nm	42	Walcott Tuff, American Falls, ID
10PR89-002	nm	nm	274 ± 4	8 ± 3	230 ± 3	324 ± 4	296 ± 3	nm	nm	nm	nm	nm	Big Southern Butte
U.S. Geological Survey reference standards
RGM-1 (meas.)	nm	nm	146 ± 4	110 ± 3	25 ± 3	222 ± 4	8 ± 3	806 ± 10	nm	nm	nm	63	Glass Mountain, CA
RGM-1 (rec.)	32	14	149	108	25	219	9	807	1600	279	1.86	nr	Glass Mountain, CA

Note: Values in parts per million (ppm) except total iron (in weight %) and Fe/Mn intensity ratios; ± equals expression of x-ray counting uncertainty and regression fitting error at 120–360 second livetime; nm = not measured. Data from Hughes (2006, 2008) and Reid et al. (2008: Table 1) and Reid et al. (2015: Table 1).

Discussion

The two Clovis points from the beach at Seagull Bay are the third and fourth Clovis points recovered in wave-bitten, reservoir-margin contexts in Idaho and eastern Oregon. They add two more obsidian sources, Big Southern Butte and Walcott Tuff, to the roster of Clovis toolstones. These two sources have not been previously documented for Clovis period obsidian sources. The Big Southern Butte is approximately 65 km north from the Seagull Bay site; while the Walcott Tuff source is approximately 1 to 5 km surrounding American Falls. Both of these sources are relatively easy to access. The Big Southern Butte obsidian source is higher quality than the Walcott Tuff and is composed of rhyolitic domes, flows, pyroclastic debris, and basalt flows. The features are mostly Early Pleistocene along the eastern Snake River Plain with subvolcanic to volcanic features. The Walcott Tuff is locally available as obsidian welded and bedded tuffs. The obsidian tuff or upper section of the Walcott Tuff is thickest close to the Snake River and ranges in size from about 10 m thick near the American Falls dam to 4 m thick near upper Ferry Hollow (for more detailed information, see Carr and Trimble, 1963).

The observations summarized here from the American Falls reach suggest that Clovis populations ranged across the whole of the Snake River plain (see Figure 2). When these data are viewed in the larger framework of Clovis finds across the state of Idaho (Reid et al., 2015), the points are continuously distributed across southern Idaho but absent from the central batholithic core and from the northern counties above the Clearwater basin. An earlier suggestion that Clovis points were clustered on the western Snake River plain below Auger Falls, with Folsom points clustered on the eastern or upper Snake River plain must be revised. It may still be true that Folsom points cluster in the bunchgrass basalt plains of the upper Snake River. Historically, this part of the Snake River basin has provided the best bison habitat and Folsom populations focused on bison. The first researchers to comment on the east/west patterning of Clovis and Folsom fluted points along the Snake River Plain was Robert Yohe and Jim Woods (2002).

Butler (1965) identified a fluted point found in a plowed field a quarter mile east of Bannock Creek, south of Michaud Flats, as the first Clovis point reported from southeastern Idaho. He gave measurements of 9.6 × 3.4 × 0.9 cm with flute lengths of 3.2 and 3.8 cm. The material is described as a dark green obsidian. The specimen showed no evidence of grinding in the haft area, which Butler attributed to plow damage. The reason for this attribution is not clear; the photograph in his Figure 3 does not appear to be severely plow-scarred. The whereabouts of the point is unknown and no site form was recorded for it. Butler mentioned another Clovis point found in 1965 on the lower Portneuf River. It was found on a beach in a surface context “with some immature elephant teeth” (Butler, 1968). This is the site referred to as “Sampson” by Yohe and Woods (2002); however, it is not included in the inventory of the Archaeological Survey of Idaho.

OPEN IN VIEWER

Figure 3.

Map of Clovis site in Idaho and Eastern Oregon with obsidian sources. Adapted from Reid et al. (2015). Red star indicates Seagull Bay site.

A biface cache of possible Clovis age was described by (Butler, 1969) eroding out of a wave-cut terrace at American Falls Reservoir near the mouth of the Portneuf River. Three large bifacial performs of the same “light, silvery-grey quartzite” were recovered embedded in chunks or peds of indurated sediment that matched a dark stratum in the adjacent 3-m cut-bank. The dark unit extended from 0.4 to 1.6 m below surface. “Fossil bone” was exposed in a lighter unit below the dark unit. Butler said that in their size and shape, these bifaces resembled similar forms from the Simon cache. This cache has been referred to as the Higgins Cache or Higgins Find and after an exhaustive search of all Idaho museum collections has yet to be found and is assumed to be in private collector’s hands. The only known image of these is shown in Figure 4.

OPEN IN VIEWER

Figure 4.

Higgins Biface Cache recovered from near mouth of Portneuf River, American Falls, Idaho. Image from Butler (1969: 66).

A cache of nine obsidian bifaces was reported by Pavesic (1966) from a remnant of the Sterling terrace at the northwest end of American Falls Reservoir. A later analysis showed that the obsidian sources included Big Southern Butte, Bear Gulch, and Teton Pass but not the locally outcropping Walcott Tuff (Hughes and Pavesic, 2005), as originally surmised. It is not possible to tell from the single low-quality photograph (Pavesic, 1966) whether the bifaces display any Clovis characteristics such as fluting or overshot flaking but several do appear to exhibit bold flake removals. This cache is unavailable for study at this time.

Of all the Clovis and potentially Clovis age finds and caches reported here, only the two points from Seagull Bay have actually been fully recorded, collected, and curated. As mentioned, the Bannock Creek and Portneuf River finds were in private collections at the time they were reported and suffer from incomplete recordation or nonavailability for further study, or both.

Summary

The association of Clovis points with extinct Pleistocene megafaunal species, especially Columbian mammoth, continues to stimulate debate over the possible contribution of these pachyderms to Paleoindian diets, osteological raw materials for tools, and processes of landscape learning helpful to a colonizing population (Haynes, 2006). In Idaho, the evidence of mammoth remains and Clovis diagnostics overlaps only in the American Falls locality. The Tolo Lake locality on the Nez Perce Prairie included remains of mammoths but the only cultural remains in the vicinity were of late Holocene age. A single Clovis point has been recorded near Tolo Lake at the upper end of Butcher Creek. However, a careful technological analysis revealed that although it was presented to professional archaeologists in good faith by the owner; however, it is a replica that was made earlier in the 20th century before becoming a family heirloom. The lateral edges have polish from an emory cloth, offset flutes from being fluted in a lever device, and transmedial lateral flaking unlike Clovis points.

The only two mammoths in Idaho that has been radiocarbon dated have ages of 14,770 ± 150 BP (Beta-89262) (Grove Mammoth site) and 11,830 ± 30 BP (Beta-484099) (Kimberly Mammoth site). The Grove Mammoth site in Kamiah, Idaho, was recovered in 1957 and later excavated in 1997 at a field school by the University of Idaho. The context of the excavations suggested fluvial transport in a tributary of the Clearwater River near Kamiah (Plastino et al., 1997). An ambiguous quartzite chopper was recovered in an uncertain relationship to the bones. The Kimberly Mammoth site was first discovered in 2004 with recovery of a six-foot-long tusk, a large chipped basalt cobble, and some fragmentary vertebrae (Taylor, 2018). An excavation by Idaho State University in 2018 recovered a section of shoulder and is still ongoing under the direction of Speer. No clear cultural materials have been found yet at the site. Ground penetrating radar has revealed the potential for two mammoths spread across the site.

None of the extinct fauna represented at the Seagull Bay, Rainbow Beach, and Dam sites at American Falls has radiocarbon ages younger than 21,000 years. The only radiocarbon age in Idaho that is directly associated with Clovis diagnostics comes from a hearth in the headwaters of Salmon Falls Creek at Heil Pond (10OE9965) (Murphey, 1985). No extinct faunal remains were associated with either of these sites (for more information about these sites, see Reid et al., 2015).

It is important to note here that the greatest potential for discovering Clovis period hunter-gatherer interaction with extinct Pleistocene fauna may very well be at the American Falls locality. The clustering of Clovis points and late Pleistocene megafauna suggests that the American Falls locality warrants systematic monitoring during Bureau of Reclamation reservoir drawdowns and will continue to produce Clovis period cultural material. The Clovis horizon is widely distributed across the Snake River Plain and adjoining uplands, and Clovis technology in Idaho does not appear regionally distinctive or isolated.