at SciVerse ScienceDirect

Journal of Archaeological Science 40 (2013) 3117e3128

Contents lists available

Journal of Archaeological Science

journal homepage: http : / /www.elsevier .com/locate/ jas

Statistical means for identifying hunteregatherer residential features in a lithiclandscape

Christopher Morgan a,*, Molly Boeka Cannon b, Benjamin Fowler b

aDepartment of Anthropology, University of Nevada, Reno,1664 N. Virginia St., Reno, NV 89557-0096, USAbAnthropology Program, Utah State University, 0730 Old Main Hill, Logan, UT 84322-0730, USA

a r t i c l e i n f o

Article history:Received 12 January 2013Received in revised form20 March 2013Accepted 8 April 2013

Keywords:Stone circleRock ringSpatial statisticsGroup sizeResidential mobilitySite structureSite formation

* Corresponding author. Tel.: þ1 775 682 8992; faxE-mail addresses: ctmorgan@unr.edu (C. Morg

(M.B. Cannon), b.fowler@aggiemail.usu.edu (B. Fowler

0305-4403/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jas.2013.04.009

a b s t r a c t

Techniques are described for extracting circular rock features from landscapes dominated by clasts of thesame type from which cultural features are composed, using as a test case a large stone circle residentialsite in western Wyoming, USA. Methods consist of point plotting all relevantly-sized culturally andnaturally-deposited clasts in the field and identifying potential cultural features using point densityanalyses tools in ArcGIS. Potential rings are either accepted or rejected as cultural features by comparingclast frequency, density and distribution in internal, feature-ring, and external spatial buffers to eth-noarchaeological data recording stone circle size and morphology and to similar data generated from acontrol sample of off-site, naturally-occurring clasts. The results of the analysis are used to discuss groupsize, mobility type, and duration of site occupation and to explore problems of assessing such at surfacearchaeological sites resulting from palimpsest-type site formation processes.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Circular rock features are common to many types of archaeo-logical sites, especially those associated with hunteregatherergroups, but their objective identification is often confounded byfaint outlines, poor preservation, palimpsest-like site formationprocesses, post-depositional disturbances, and difficulty ofdiscernment in landscapes dominated by naturally-occurring clasts(Scheiber and Finley, 2010a; Seymour, 2009). Because of this, andthe fact that hunteregatherers nearly always make structures atresidential sites (sensu Binford, 1980, 1990), this study seeks toanswer modest yet nonetheless critical empirical questions aboutresidential feature identification using straightforward morpho-logical analyses and geostatistical methods. These types of ques-tions are often overlooked and their answers often rely on intuitionin hunteregatherer archaeology: what comprises rock-ringed res-idential features and how can such features be objectively dis-cerned from the natural distribution of rocks often found on thesurface of archaeological sites? The subject is important not only tobasic methods of site recording and mapping (Hester et al., 2008),

: þ1 775 327 2226.an), molly.cannon@usu.edu).

All rights reserved.

but also toward more fundamental questions related to deter-mining group size, group composition, and intensity of site occu-pation and re-occupation over time (Diehl, 1992; Moore, 1998;Smith, 2003). These empirical determinations are of courseessential to answering larger questions concerning degrees ofmobility and population size (Binford, 1990; Bocquet-Appel et al.,2005; Casteel, 1979; Kelly, 1992; Schreiber and Kintigh, 1996;Zorn, 1994) which are in turn critical to understanding the funda-mental human ecology of groups who built and used such features(e.g., Bettinger, 1977; Grayson and Cannon, 1999; Hardesty, 1983;Winterhalder et al., 1988).

This research focuses on spatial data collected from 48TE479, astone circle (oftentimes referred to in the vernacular as “tipi rings”,though the superstructures associated with such features werelikely of considerable variation) site in the Gros Ventre River valleyin western Wyoming, USA; it is thus perhaps most applicable toanalyses focusing on North America’s Great Plains, where suchsurface features are relatively common (Frison, 1983; Kehoe, 1960;Kornfeld et al., 2010). Such features are structurally simple, con-sisting typically of little more that 1e3 courses of locally-availablecobbles and small boulders arranged in a circular manner, withdiameters ranging anywhere from about 2.5 to 8 m (Mobley, 1983).They almost never contain subsurface features or deposits (Kehoe,1983). Similar features are found throughout western NorthAmerica, where clastic site surfaces are common and where

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283118

residential features are often difficult to objectively identify withinsuch landscapes. For instance, rock rings associatedwith residential(per Binford, 1980) group aggregations and storage are common inthe piñon ecozones (Bettinger, 1989), valley floors (Eerkens, 2008),and alpine elevations (Bettinger, 1991, 2008; Thomas, 1982) of theGreat Basin. Similar features are found in the higher elevations ofWyoming’s mountain ranges (Morgan et al., 2012a; Scheiber andFinley, 2010b) and in California’s Sierra Nevada (Morgan, 2008,2012). In fact, due to their simple construction and associationwithmany types of human activities, the methods described in thispaper might be applied to any of those situations where humansmade and used rock rings, whether it be Neolithic Europe (Blot,1991; Eogan, 1964), Sub-Saharan Africa (Anquandah, 1986; Soper,1977), prehistoric Japan (Komai, 1961), or the deserts of Australia(Mathews, 1895; O’Connell, 1987), but its principal focus is on thecircular or semicircular structures most often made and used bymobile hunteregatherer groups (Binford, 1990; Diehl, 1992;Whiting and Ayers, 1968).

Despite their widespread distribution, surprisingly little previ-ous research has been specifically conducted on the problem ofidentifying and documenting stone circles and other rock rings,particularly in the last decade or so, a situation Scheiber and Finley(2010a) link to the proliferation of management-oriented archae-ology into those regions where rock rings are most common, but aphenomenon also likely linked to the paucity of artifacts and sub-surface deposits, and hence the perceived research value, typicallyassociated with such sites, at least on the Great Plains (Kehoe,1983). Historically, most metric research on Great Plains stonecircles has focused on basic description, middle-range type ques-tions regarding feature function, and intrasite spatial analysis(Davis, 1983a; Kehoe, 1983). For instance, Smith (1974) and Aaberg(1975) provide basic recording instructions for Great Plains stonecircles, emphasizing the benefits of point-plotting the individualrocks comprising such features. Hoffman (1953), Kehoe (1958) andMalouf (1961) use empirical observation and ethnographic analogyto equate stone circle sites with short-term residential groupings ofPlains populations. Among the first to systematically addressfunction and the metric morphology of stone circles was Kehoe(1960), who suggested smaller rings may pre-date the acquisitionof the horse by Blackfoot groups, a hypotheses later questioned byLarson (1979). Partly in an effort to address this hypothesis, Mobley(1983) used non-parametric statistics to argue that there are indeedthree significant stone circle diameter classes in NewMexico, but isequivocal about whether these classes have temporal or functionalsignificance (see also Larson, 1981; Quigg, 1981; Roll, 1981; Wilson,1983). In a similar vein, Davis (1983b) uses multivariate analyses ofinternal and external diameters, number of rocks, distance tonearest adjoining stone circle, rock density and feature shape forstone circles in Wyoming in an attempt to link circle morphologywith ethnohistoric function. Like Mobley, he is equivocal about hisfindings and suggests that stone circles, regardless of size ormorphology, cannot be excluded as possible tipi rings (see alsoFinnigan, 1980). Corroborating an earlier observation by Frison(1967), Finnigan (1981) measured rock densities in differentdirectional segments of stone circles in Alberta and found thatdensities are greater on the windward side of such features, indi-cating their function as weights holding down the skins of tipis ortents. Moving beyond the analysis of stone circles themselves,others have performed intrasite spatial analysis of artifact scatters,living floors, and refuse deposits, assessing their spatial associationwith stone circles in attempts to elucidate different activity areas(Reher, 1983; Smith et al., 1995) and the symbolic use of space(Oetelaar, 2000). In a return to the basics of site mapping in thedigital age, Scheiber and Finley (2010a) have used high-precision,differently-correctable GPS, remote sensing and radiocarbon

dating to record spatially-based attribute data for individual fea-tures and the rocks comprising these features in the BighornCanyon area of Montana and Wyoming, linking these features toboth the Late Plains Archaic (ca. 500 BC) and arguably themigrationof the Crow into the region in the Late Prehistoric (after AD 700).

Similar research patterns pertain outside the Great Plains. Interms of function, for instance, Bettinger (1989) used rock ringdiameter and artifact associations correlated with ethnographicinformation to identify ring function in Owens Valley, in easternCalifornia. He argues smaller diameter rings are storage features andlarger ones with milling equipment and heterogenous artifact as-sociation are domiciles; other possible functions include sweatlodges andmenstrual huts. In a similar vein, Baker (2003) comparedthe morphology of rock rings in western Colorado to ethnohistoricrecords to argue that the features he identified there likely representUte menstrual huts. Seymour (2009) used similar ethno-archaeological methods to identify faint Apache wickiup footprintsin Sonora, Mexico (see also Donaldson andWelch, 1991). One of themost direct (yet also perhaps unnecessarily complex) attempts toquantitatively assess rock ring morphology as an indicator of func-tion comes fromoutside the realmof hunteregatherer archaeology:in Great Britain, Patrick andWallace (1982) used Fourier analysis toassess degrees of circularity of stone circle sites arguably linked toarchaeoastronomical observation.

Moving beyond simple morphological-functional studies,Simms (1989) used ethnoarchaeology and intrasite spatial analysesto elucidate mobility type and duration of occupation at a Shosh-onean wickiup site in Eastern Nevada. Bettinger (1975) took thesetypes of analyses one step further by estimating population sizefrom rock ring surface area in eastern California. Counterintuitively,some of the more recent and quantitative approaches toward res-idential feature identification consist of those focusing on geo-statistical analysis of artifact distributions in the absence of directevidence for residential feature construction (Stiger, 2006; Surovelland Waguespack, 2007) or those employing remote sensing tech-niques (Finley and Scheiber, 2007; Morgan et al., 2012b). There hasconsequently been a fair amount of description and attempts atdetermining rock ring function, mainly by employing ethno-archaeological techniques, but very little focus on objectivelyidentifying features from surface data, suggesting intuition trumpsmetric analysis in many site recording and reporting situations.Many who have stood around what appears to be a faint rock circleor a cluster of cobbles with their colleagues, shrugging theirshoulders as to whether or not they had a feature on their handslikely know this is often the case.

Within this context, the goal of this paper is to provide simplequantitative and statistical means of objectively and confidentlyidentifying circular surface residential features in any variety ofgeomorphic contexts using methods that should be transparentand easily replicable to those with a working grasp of basic map-ping techniques, simple spatial statistics and access to industry-standard GIS software. It uses as a test case a large stone circlesite along the Gros Ventre River in WesternWyoming, where clastseroding out of a Pleistocene fluvial terrace have hindered past at-tempts to determine the actual number of residential features atthe site. Though methodologically oriented, it concludes with abrief consideration of the contributions of objective feature iden-tification toward interpreting group size and composition, durationof occupation, degrees of residential mobility and the constraints ofascertaining such from surficial archaeological deposits.

2. Site description

The focus of this study is site 48TE479, originally recorded in1971 (Love, 1971). It was then described as a site containing at least

Fig. 1. Photograph of stone circle at 48TE479 (foreground) with students mapping the site in the background (Photo by C. Morgan).

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e3128 3119

16 stone circles atop a high terrace above an outside, erodingmeander with a commanding view of the westward-flowing GrosVentre River (Fig. 1). For the most part, this original description isapt. The site measures 162 m NEeSW by 45 m NWeSE and consistsof numerous stone circles made up of cobble concentrations andsingle-course, ring-like arrangements of local quartzite cobbles, aswell as a very light-density surface scatter of lithic artifacts (Fig. 2).Intensive surface inventory identified 30 tested quartzite cobbles,

Fig. 2. Site map showing distribution of su

59 coreereduced quartzite flakes, five obsidian flakes, four chertflakes and two basalt flakes. All of the quartzite is endemic to thesite and eroding out of the Pleistocene fluvial terrace on which thesite is located; the remaining lithic material is extralocal. Thepaucity of artifacts at the site is not surprising: like many stonecircle sites in the region, abundant features associated with fewartifactual remains are common and often interpreted as evidenceof high-frequency residential moves during the Archaic (Frison,

rface clasts and sample subdivisions.

Fig. 3. Site map showing results of point density analyses and feature centerpoints.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283120

1983) and by Late Prehistoric and ethnographic groups like theCrow and the Eastern Shoshone, especially following the intro-duction of the horse (Horde, 1982; Kornfeld et al., 2010; Loendorfand Weston, 1983; Morris et al., 1983; Smith et al., 1995). Impor-tantly, through multiple episodes of site recording, the number ofstone circles identified at the site has varied from as few as 16 to asmany as 33. This discrepancy results mainly from different re-searcher’s intuition as to what exactly determines whether or not agrouping of quartzite cobbles is actually a residential feature, whichof course has a direct effect on the number of circles identified eachtime the site was re-recorded. These discrepancies provided theinitial impetus to conduct this study.

3. Methodology

The goal of this analysis is to identify rock ring features in aclastic landscape using simple methods for identifying patterningin spatial point data. Point patterns were extracted via field andanalytical techniques described below.

3.1. Field methods

Field methods consisted of intensive surface survey and point-plotting of artifacts and individual rocks across the surface of thesite, which is characterized by sparse, low grasses and generallyexcellent ground visibility. The survey was conducted via line-abreast, compass oriented transects, with spacing of no morethan 1m between transects. All artifacts and natural clasts between7 cm (about the size of a fist) and 21 cm in diameter (the largestclast identified at the site) were marked with pin-flags and theircenterpoints mapped with a sub-cm accuracy robotic total station.Clast centerpoints were approximated by placing the total stationreflector rod on top of the visually-estimated center of identifiedclasts. These dimensionless points were used as proxies for thefairly small and more-or-less consistently-sized site clasts in allsubsequent spatial analyses. Due to the high numbers of pin flagsrequired to map the site (3404 individual points), several survey-mapping rounds were required to complete the task. A lightningstorm occurred during this process, resulting in a rapid take-downof the total station. This resulted in a small error (approximately20e30 cm) being generated between northern and southern pointplots; for this reason, the point data across the site are notcontinuous, but rather must be considered two independent sam-ples (Fig. 2) due to the possibility that this error could skew theresults of subsequent geostatistical analyses. In order to generate acontrol sample of naturally-occurring clasts, three distributions ofgeologically-deposited cobbles eroding out of a gentle slope some80 m outside the northwestern site boundary were mapped usingthese same methods.

3.2. Analytical methods

Analytical methods consist of techniques used to extract clusters(per Clark and Evans, 1954) of rocks and ring-like structures fromthe spatial point data for all rocks mapped in the field and to assesswhether identified spatial patterning in these point data conformto both ethnographic and archaeological accounts of residentialstone circles and also to simple mathematical models developed toaccount for the human use of space and human modification to thenatural distribution of cobbles at the site. These methods are sub-divided into a set of four assumptions described below. Associatedanalytical methods were conducted using ESRI ArcGIS 10.0 soft-ware and the Spatial Analyst Extension contained in this software.Analyses were performed separately on North, South and Controlsamples, with results from the Control samples used to develop

expectations for what natural rock concentrations should look likewhen compared to cultural ones.

Assumption 1: Residential features should be marked by agreater density of clasts than that of the surrounding landscape anddistributed in circular-shaped clusters. To identify such clusters, thePoint Density tool for circular neighborhood was used to generate araster showing the magnitude of rock point density per squaremeter in a 4 m neighborhood search radius, this representing themaximum outside radius of Great Plains stone circles (i.e., amaximum 8 m diameter) (Finnigan, 1981; Mobley, 1983).

Assumption 2: Residential features should be distributed in ringor circular-shaped patterns. To identify such annulus-shaped pat-terns, the Point Density tool for annulus was used to generate araster showing the magnitude of rock point density per squaremeter in a 1e4 m neighborhood search radius, this approximatingthe minimum and maximum outside radius of Great Plains stonecircles used to hold down the edges of tents and tipis (Brumley andKooyman, 1978; Finnigan, 1981). The net result of both types ofdensity analysis is the identification of circular patterns of rocksand partial and complete rings of rocks within these clusters. Fromthese data, centerpoints of circular clusters and rings (both com-plete and partial) were determined by plotting a point in the centerof each cluster or ring. This was accomplished by dividing eachmaximum cluster or ring diameter and its perpendicular axis by

0 1.5 3 4.5 60.75Meters

a b

Fig. 4. Maps showing 1.25, 4 and 5.25 m buffers for: (a) Feature 0S; (b) edited buffers for Feature 4S.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e3128 3121

two. These centerpoints are considered centers of potential culturalfeatures (Fig. 3).

Assumption 3: Stone circles should be marked by a greaterfrequency and density of clasts in a radius conforming to ethno-graphic and archaeological evidence for rock ring clast density inthe geographic area in which the analysis is being performed. Forthe northern Great Plains, averageminimum internal ring diameteris about 2.5 m (i.e., a 1.25 m radius) and maximum external ringdiameter is about 8 m (i.e., a 4 m radius) (Adams, 1978; Finnigan,1981; Kehoe, 1960; Mobley, 1983). To generate frequency anddensity data conforming to test these expectations at 48TE479,multiple ring buffers with 1.25, 4.0 and 5.25m radii were generatedand queried to identify the number of rocks in each buffer. Thoughoverlapping rings are rare in the region (Kehoe, 1983), to allow forthe possibility that rocks may have been re-used, buffers werecreated for each potential feature centerpoint (Fig. 4a), meaningclasts could be associated with more than one centerpoint. In thecase of incomplete rings, buffers were edited to generate minimumenclosing area arcs containing all clasts in a given buffer to ascer-tain the area by which to generate point density and performgeostatistical analysis (e.g., the external, 3rd buffer in Fig. 4b). Forthe control sample, the smaller feature size and dearth of rocks

rock (typ.)

Control 01 Control 02

(scale is the s

0 10.5

Fig. 5. Maps showing 1.25 and 3

surrounding these natural features required generating only twobuffers with 1.25 and 3 m radii (Fig. 5). These smaller buffers weregenerated because using the larger site sample buffers would resultin disproportionately large areas containing no clasts outside of thenatural, 3-m diameter distribution of rocks in the control samples.These larger buffers would consequently result in artificially lowclast densities that reflect more the effect of sample area size thanthe actual distribution of clasts (this is the same reason that sitesample buffers were edited to generate minimum enclosing arcs inthe case of incomplete rings). Such large buffers would also intro-duce unnecessary bias by increasing the effect of sample areaconfiguration and size (per Ebdon, 1976) in subsequent statisticalanalyses (see below).

Assumption 4: Deliberately-made features should show evi-dence of clast clustering and naturally-occurring features shouldtypically show random distributions of clasts (Finnigan, 1981;Patrick and Wallace, 1982), or even patterned or regular clast dis-tributions (Waters and Kuehn, 1996). To test this assumption, thenearest-neighbor (NN) statistic, a simple, intuitive spatial statisticwith a long history of use in archaeology (Clark, 1972; Durand et al.,1992; Hodder and Hassell, 1971; Hodder and Orton, 1976;Ladefoged and Pearson, 2000; Washburn, 1974) was used to

Control 03

ame for all)

2 3Meters

m buffers for control sample.

Table 1Control sample analysis results.

Feature Buffer Freq. Density/m2 Obs. dist. Exp. dist. NN stat. Z score p-value NN det. Z-score det. Conclusion Ring? Explanation

1C 1 21 4.278 0.331 0.263 1.255 1.406 0.079 Dispersed Not significant Random Reject Type 1 & 3 errors1C 2 49 1.080 0.315 0.342 0.920 0.960 0.168 Slightly clustered Not significant Random2C 1 17 3.463 0.320 0.296 1.080 0.576 0.282 Slightly clustered Not significant Random Reject Type 1, 2 & 3 errors2C 2 36 0.793 0.398 0.439 0.906 0.961 0.168 Slightly clustered Not significant Random3C 1 10 2.037 0.377 0.399 0.943 0.305 0.379 Slightly clustered Not significant Random Reject Type 1 & 2 errors3C 2 38 0.837 0.359 0.372 0.964 0.373 0.354 Random Not significant Random

Fig. 6. Graphic depiction of clast frequency, density and NN statistics in control samplebuffers.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283122

generate NN statistics, z-scores and p-values for point distributionsin the buffers around each hypothetical feature centerpoint. The NNstatistic, dobs/dran, is generated by dividing the observed meanEuclidian distance between each point and its nearest neighbor(dobs) by an expected random distribution (dran) derived, tradi-tionally (though alternative methods have been developed), by thefollowing formula: dran ¼ 0.5O(n/a). This is simply one-half thesquare root of the observed point density, or the number ofobserved points [n] divided by the size of the study area [a]. NNstatistics less than one indicate clustering, a statistic of one in-dicates a random distribution, and statistics greater than oneindicate regular or dispersed distributions (Clark and Evans, 1954).Associated z-scores and p-values indicate the confidence level atwhich the null hypothesis (that observed points are randomlydistributed) is rejected: typically z-scores between 1.96 and �1.96are rejected because they result in p-values of less than 0.05.

Because the size and shape of the study area can have significanteffects on expected distributions (dran) and hence on NN statisticaloutcomes, NN statistics are sometimes seen as less robust thansome other geostatistical methods (Ebdon, 1976), a situationarguably compounded by how expected random values aregenerated (Pinder et al., 1979). Their appeal, however, is in howstraightforward they are to calculate and interpret (Connolly andLake, 2006:164). Further, Morgan (2009) has shown that the wayexpected random distributions are modeled has little or no sig-nificant effect on archaeological interpretation. To compensate forthe second problem, that of edge effects and study area configu-ration potentially skewing analysis results, this analysis employs avisual basic script developed by Sawada (2002) which generateshypothetical random NN values using polygon shapefiles to deter-mine area (in this case the aforementioned buffers) and containscode to correct for edge effects. Using this script, NN statistics weregenerated for each of the three buffers surrounding each hypo-thetical feature centerpoint and the two buffers surrounding eachof the control samples.

4. Results

The results of the analysis are presented in subsections forcontrol and site samples. The former are used to develop a set ofexpectations for accepting or rejecting rings as residential featuresin the site sample.

4.1. Control sample results

Control sample results are presented in Table 1 and Fig. 6. Clastfrequency and density are inversely related: clast frequency in-creases and clast density decreases as a function of the larger areaof Buffer 2 in all three samples. On average, Buffer 2 frequencies are2.75 times greater than Buffer 1 frequencies. Average Buffer 1densities are 3.67 times greater than those in Buffer 2. Importantly,NN statistics concentrate near values of 1, particularly in the outerbuffer. These statistics generally indicate random distributionsregardless of buffer size; z-score and p-values support this

conclusion. These data are interpreted as representative of thenatural densities, frequencies and distributions of clasts in the sitevicinity; they therefore provide a measure by which to assesswhether natural or cultural processes (sensu Schiffer,1987) resultedin the distribution of clasts within the site.

These control sample data result in four conditions by whichfeatures in the two site samples might be rejected as cultural res-idential features: (Type 1 Error) NN statistics do not indicate clus-tering at a confidence level of 95% (i.e., p-values � 0.05) in Buffer 2(which ostensibly contains the ring of rocks used to secure tipi or

Table 2Site sample analysis results.

Feature Buffer Freq. Density/m2 Obs. dist. Exp. dist. NN stat. Z score p-value NN det. Z-Score det. Conclusion Ring? Explanation

0S 1 0 0.000 0.000 0.000 N/A N/A N/A None None None Accept Meets all criteria0S 2 101 2.227 0.251 0.352 0.711 5.149 < 0.0001 Clustered Significant Clustered0S 3 9 0.248 1.001 0.787 1.272 1.337 0.091 Dispersed Not significant Random1S 1 2 0.407 0.644 1.098 0.587 1.015 0.155 Clustered Not significant Random Accept Meets all criteria1S 2 49 1.080 0.292 0.518 0.563 5.321 <0.0001 Clustered Significant Clustered1S 3 1 0.028 N/A N/A N/A N/A N/A None None None2S 1 3 0.611 0.177 0.834 0.212 2.335 0.010 Clustered Significant Clustered Accept Meets all criteria2S 2 52 1.146 0.289 0.502 0.576 5.330 <0.0001 Clustered Significant Clustered2S 3 4 0.110 2.730 0.700 3.901 9.753 N/A Dispersed Significant Dispersed3S 1 1 0.204 N/A N/A N/A N/A N/A None None None Accept Meets all criteria3S 2 62 1.367 0.360 0.457 0.787 2.932 0.002 Clustered Significant Clustered3S 3 17 0.468 0.573 0.454 1.262 1.779 0.038 Dispersed Significant Dispersed4S 1 1 0.204 N/A N/A N/A N/A N/A None None None Accept Meets all criteria4S 2 56 1.235 0.352 0.482 0.730 3.526 0.000 Clustered Significant Clustered4S 3 7 0.193 1.168 0.594 1.967 4.250 <0.0001 Dispersed Significant Dispersed5S 1 5 1.019 0.624 0.604 1.033 0.126 0.450 Random Not significant Random Accept Meets all criteria5S 2 42 0.926 0.345 0.564 0.612 4.358 <0.0001 Clustered Significant Clustered5S 3 8 0.220 0.710 0.520 1.365 1.717 0.043 Dispersed Significant Dispersed6S 1 2 0.407 1.328 1.097 1.211 0.519 0.302 Dispersed Not significant Random Accept Meets all criteria6S 2 93 2.050 0.305 0.368 0.829 2.912 0.002 Clustered Significant Clustered6S 3 15 0.413 0.477 0.565 0.844 0.994 0.160 Clustered Not significant Random7S 1 21 4.278 0.222 0.263 0.843 1.243 0.106 Clustered Not significant Random Reject Type 2 & 3 errors7S 2 51 1.124 0.380 0.507 0.750 3.102 <0.0001 Clustered Significant Clustered7S 3 20 0.550 0.384 0.368 1.043 0.325 0.372 Random Not significant Random8S 1 23 4.686 0.202 0.251 0.805 1.627 0.052 Clustered Significant Clustered Accept Meets all criteria8S 2 102 2.249 0.284 0.350 0.811 3.376 <0.0001 Clustered Significant Clustered8S 3 16 0.440 0.555 0.755 0.734 1.725 0.042 Clustered Significant Clustered9S 1 20 4.075 0.285 0.271 1.054 0.472 0.319 Random Not significant Random Accept Meets all criteria9S 2 83 1.830 0.233 0.391 0.596 6.488 <0.0001 Clustered Significant Clustered9S 3 12 0.330 0.419 0.926 0.453 3.093 0.001 Clustered Significant Clustered10S 1 1 0.204 N/A N/A N/A N/A N/A None None None Accept Meets all criteria10S 2 31 0.683 0.507 0.666 0.762 2.282 0.011 Clustered Significant Clustered10S 3 5 0.138 2.355 2.207 1.067 1.951 0.026 Dispersed Significant Dispersed11S 1 10 2.037 0.336 0.400 0.840 0.868 0.193 Clustered Not significant Random Accept Meets all criteria11S 2 36 0.794 0.433 0.613 0.706 3.142 0.001 Clustered Significant Clustered11S 3 20 0.551 0.540 0.632 0.854 1.077 0.141 Slightly clustered Not significant Random12S 1 8 1.629 0.430 0.455 0.944 0.267 0.394 Random Not significant Random Reject Type 1 error12S 2 160 3.527 0.255 0.276 0.921 1.775 0.037 Random Significant Random12S 3 43 1.183 0.427 0.463 0.923 0.839 0.200 Random Not significant Random13S 1 10 2.037 0.407 0.400 1.017 0.089 0.464 Random Not significant Random Accept Meets all criteria13S 2 101 2.227 0.254 0.352 0.721 4.981 <0.0001 Clustered Significant Clustered13S 3 76 2.092 0.281 0.326 0.860 2.070 0.019 Slightly clustered Significant Clustered14S 1 3 0.611 0.544 0.834 0.652 1.031 0.151 Slightly clustered Not significant Random Accept Meets all criteria14S 2 70 1.543 0.343 0.428 0.802 2.914 0.002 Clustered Significant Clustered14S 3 40 1.101 0.296 0.379 0.781 2.336 0.010 Clustered Significant Clustered15S 1 4 0.814 0.565 0.693 0.815 0.627 0.265 Clustered Not significant Random Reject Type 4 error15S 2 33 0.727 0.523 0.643 0.813 1.844 0.0325 Clustered Significant Clustered15S 3 28 0.770 0.624 0.531 1.174 1.533 0.0625 Dispersed Significant Dispersed16S 1 6 1.222 0.564 0.540 1.043 0.181 0.428 Random Not significant Random Accept Meets all criteria16S 2 58 1.279 0.293 0.473 0.619 5.073 <0.0001 Clustered Significant Clustered16S 3 26 0.716 0.441 0.387 1.138 1.189 0.117 Dispersed Not significant Random17S 1 3 0.611 0.915 0.834 1.097 0.287 0.387 Random Not significant Random Accept Meets all criteria17S 2 72 1.587 0.228 0.282 0.811 2.835 0.002 Clustered Significant Clustered17S 3 54 1.487 0.242 0.299 0.809 2.395 0.008 Clustered Significant Clustered18S 1 8 1.629 0.401 0.455 0.880 0.577 0.281 Slightly clustered Not significant Random Reject Type 1, 2 & 3 errors18S 2 5 0.110 0.929 1.444 0.643 1.340 0.090 Clustered Not Significant Random18S 3 0 0 N/A N/A N/A N/A N/A None None None0N 1 6 1.222 0.568 0.540 1.051 0.213 0.416 Slightly clustered Not significant Random Accept Meets all criteria0N 2 101 2.227 0.259 0.352 0.735 4.716 <0.0001 Clustered Significant Clustered0N 3 7 0.193 1.402 1.107 1.267 1.164 0.122 Dispersed Not significant Random1N 1 25 5.093 0.265 0.240 1.108 0.939 0.174 Slightly dispersed Not significant Random Accept Meets all criteria1N 2 149 3.285 0.251 0.287 0.873 2.772 0.003 Slightly clustered Significant Clustered1N 3 38 1.046 0.492 0.525 0.938 0.632 0.264 Slightly clustered Not significant Random2N 1 3 0.611 1.336 0.834 1.601 1.780 0.038 Dispersed Significant Dispersed Accept Meets all criteria2N 2 43 0.948 0.433 0.556 0.778 2.530 0.006 Clustered Significant Clustered2N 3 31 0.853 0.430 0.448 0.959 0.382 0.351 Random Not significant Random3N 1 25 5.093 0.224 0.240 0.935 0.567 0.285 Slightly clustered Not significant Random Accept Meets all criteria3N 2 65 1.433 0.385 0.445 0.865 1.908 0.028 Slightly clustered Significant Clustered3N 3 7 0.193 1.920 1.398 1.373 1.622 0.052 Dispersed Not significant Random4N 1 2 0.407 1.795 1.096 1.637 1.565 0.058 Dispersed Not Significant Random Reject Type 1 error4N 2 77 1.697 0.408 0.406 1.004 0.064 0.474 Random Not significant Random4N 3 25 0.688 0.760 0.628 1.210 1.726 0.042 Dispersed Significant Dispersed5N 1 12 2.444 0.309 0.360 0.857 0.845 0.198 Slightly clustered Not significant Random Reject Type 1 & 4 errors

(continued on next page)

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e3128 3123

Table 2 (continued )

Feature Buffer Freq. Density/m2 Obs. dist. Exp. dist. NN stat. Z score p-value NN det. Z-Score det. Conclusion Ring? Explanation

5N 2 58 1.278 0.437 0.473 0.924 1.001 0.158 Slightly clustered Not significant Random5N 3 65 1.789 0.343 0.356 0.963 0.498 0.309 Random Not significant Random6N 1 2 0.407 1.969 1.096 1.795 1.955 0.025 Dispersed Significant Dispersed Reject Type 4 error6N 2 54 1.190 0.350 0.414 0.846 1.969 0.024 Slightly clustered Significant Clustered6N 3 50 1.376 0.389 0.384 1.0140 0.166 0.433 Random Not significant Random7N 1 2 0.407 1.675 1.097 1.528 1.296 0.097 Dispersed Not significant Random Accept Meets all criteria7N 2 57 1.257 0.405 0.478 0.848 2.011 0.022 Clustered Significant Clustered7N 3 6 0.165 0.508 0.444 1.144 0.593 0.277 Slightly dispersed Not significant Random8N 1 21 4.278 0.252 0.264 0.955 0.355 0.361 Random Not significant Random Accept Meets all criteria8N 2 54 1.191 0.319 0.492 0.649 4.504 <0.0001 Clustered Significant Clustered8N 3 6 0.165 0.757 0.561 1.348 1.430 0.076 Dispersed Not significant Random9N 1 6 1.222 0.431 0.425 1.015 0.075 0.470 Random Not significant Random Accept Meets all criteria9N 2 70 1.543 0.301 0.428 0.703 4.360 <0.0001 Clustered Significant Clustered9N 3 13 0.358 0.473 0.785 0.603 2.348 0.009 Clustered Significant Clustered10N 1 1 0.204 N/A N/A N/A N/A N/A None Not applicable None Accept Meets all criteria10N 2 43 0.948 0.356 0.478 0.746 2.879 0.002 Clustered Significant Clustered10N 3 9 0.248 0.777 0.944 0.823 0.874 0.191 Clustered Not significant Random11N 1 10 2.037 0.384 0.400 0.960 0.215 0.415 Random Not significant Random Accept Meets all criteria11N 2 97 2.139 0.276 0.360 0.767 4.057 <0.0001 Clustered Significant Clustered11N 3 12 0.330 0.887 0.902 0.983 0.094 0.462 Random Not significant Random12N 1 21 4.278 0.293 0.263 1.113 0.903 0.183 Slightly dispersed Not significant Random Reject Type 1 error12N 2 186 4.101 0.261 0.255 1.020 0.496 0.309 Random Not significant Random12N 3 56 1.541 0.380 0.431 0.882 1.465 0.071 Slightly clustered Significant Clustered13N 1 4 0.815 0.922 0.693 1.330 1.122 0.131 Dispersed Not significant Random Accept Meets all criteria13N 2 150 3.307 0.259 0.286 0.904 2.094 0.018 Slightly clustered Significant Clustered13N 3 19 0.523 0.606 0.706 0.859 1.007 0.157 Slightly clustered Not significant Random14N 1 3 0.611 1.156 0.834 1.386 1.142 0.127 Dispersed Not significant Random Accept Meets all criteria14N 2 127 2.800 0.264 0.312 0.845 3.116 0.001 Clustered Significant Clustered14N 3 9 0.248 0.613 0.662 0.925 0.375 0.354 Slightly clustered Not significant Random15N 1 1 0.204 N/A N/A N/A N/A N/A None Not applicable None Accept Meets all criteria15N 2 71 1.565 0.364 0.425 0.856 2.126 0.017 Slightly clustered Significant Clustered15N 3 15 0.413 0.565 0.709 0.797 1.292 0.098 Clustered Not significant Random16N 1 0 0 N/A N/A N/A N/A N/A None Not applicable None Reject Type 1 & 4 errors16N 2 64 1.411 0.389 0.403 0.964 0.495 0.310 Random Not significant Random16N 3 62 1.706 0.384 0.434 0.883 1.535 0.062 Slightly clustered Not significant Random17N 1 15 3.055 0.222 0.317 0.701 1.998 0.022 Clustered Significant Clustered Reject Type 1 error17N 2 57 1.256 0.431 0.477 0.903 1.273 0.101 Slightly clustered Not significant Random17N 3 17 0.468 0.933 0.783 1.190 1.281 0.100 Slightly dispersed Not significant Random18N 1 2 0.407 1.682 1.09 1.534 1.313 0.094 Dispersed Not significant Random Reject Type 4 error18N 2 62 1.367 0.361 0.456 0.791 2.871 0.002 Clustered Significant Clustered18N 3 85 2.340 0.315 0.364 0.865 2.101 0.017 Slightly clustered Significant Clustered19N 1 10 2.037 0.368 0.399 0.922 0.420 0.337 Slightly clustered Not significant Random Reject Type 1 error19N 2 94 2.072 0.359 0.365 0.982 0.305 0.379 Random Not significant Random19N 3 45 1.238 0.365 0.411 0.887 1.267 0.102 Slightly clustered Not significant Random20N 1 0 0.000 N/A N/A N/A N/A N/A None Not Applicable None Accept Meets all criteria20N 2 12 0.265 0.506 0.787 0.643 2.088 0.018 Clustered Significant Clustered20N 3 0 0.000 N/A N/A N/A N/A N/A None Not applicable None21N 1 0 0.000 N/A N/A N/A N/A N/A None e No neighbors Not applicable None Accept Meets all criteria21N 2 92 2.028 0.380 0.455 0.834 2.781 0.003 Clustered Significant Clustered21N 3 51 1.404 0.430 0.439 0.979 0.245 0.403 Random Not significant Random22N 1 5 1.018 0.385 0.6035 0.638 1.373 0.0847 Clustered Not significant Random Reject Type 4 error22N 2 76 1.675 0.285 0.3728 0.767 3.558 <0.0001 Clustered Significant Clustered22N 3 56 2.286 0.349 0.3732 0.935 0.818 0.206 Slightly clustered Not significant Random23N 1 26 5.297 0.236 0.235 1.004 0.036 0.486 Random Not significant Random Accept Meets all criteria23N 2 133 2.932 0.257 0.305 0.845 3.198 0.001 Clustered Significant Clustered23N 3 59 1.976 0.361 0.404 0.893 1.375 0.085 Slightly clustered Not significant Random24N 1 2 0.407 1.364 1.096 1.244 0.599 0.274 Dispersed Not significant Random Reject Type 4 error24N 2 94 2.072 0.278 0.365 0.761 4.085 <0.0001 Clustered Significant Clustered24N 3 79 3.033 0.326 0.318 1.025 0.380 0.351 Random Not significant Random25N 1 0 0 N/A N/A N/A N/A N/A None Not applicable None Reject Type 1 & 4 errors25N 2 17 0.581 0.585 0.752 0.777 1.552 0.060 Clustered Not significant Random25N 3 32 2.520 0.382 0.357 1.068 0.657 0.255 Slightly dispersed Not significant Random26N 1 1 0.203 N/A N/A N/A N/A N/A None Not Applicable None Reject Type 1 & 4 errors26N 2 13 0.462 0.992 0.861 1.152 0.926 0.177 Slightly dispersed Not significant Random26N 3 28 1.195 0.373 0.542 0.688 2.731 0.003 Clustered Significant Clustered27N 1 2 0.407 1.846 1.096 1.679 1.669 0.047 Dispersed Significant Dispersed Type 4 error27N 2 58 1.746 0.330 0.407 0.810 2.514 0.006 Clustered Significant Clustered27N 3 53 3.411 0.279 0.300 0.931 0.853 0.196 Slightly clustered Not significant Random

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283124

tent edges); (Type 2 Error) the frequency of clasts in Buffer 2 is lessthan 2.75 times the ring density in Buffer 1, meaning there arefewer clasts in the buffer that should contain evidence of deliberaterock placement than within the center of the purported feature

(this still allows for internal features like hearths to be present inidentified residential features); (Type 3 Error) using the same logic,the density of clasts in Buffer 1 is more than 3.67 times the densityof clasts in Buffer 2 (meaning there are more rocks in the center of

0 1 2 3 40.5Meters (scale is the same for all)

Fig. 7. Maps showing three distinctive stone circles at site 48TE479.

Fig. 8. Graphic depiction of clast frequency, density and NN statistics in three clearlyrecognizable stone circle buffers.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e3128 3125

the feature than in the purported ring or stone circle); and (Type 4Error) Buffer 3 frequency or density is greater than Buffer 2. Thoughno data could be collected for a third buffer in the control sample(the rock distributions here were too small), increasing clast fre-quencies and/or densities outside of possible features (i.e., in Buffer3) in the site samples would indicate that any increased frequencyor density in Buffer 2 could potentially be the product of naturalprocesses endemic to densely-packed clastic landscapes. If any ofthese four conditions are met in the site sample, then the feature inquestion is rejected as a legitimate cultural feature.

4.2. Site sample results

Circular and annulus point density analysis resulted in thegeneration of 47 potential residential features at the site (Table 2).Of these, 30 features met all four criteria and were accepted aslikely representing legitimate stone circle residential features.Importantly, in cases where ring-like structures were extremelyobvious (e.g., at Features 2S, 0N, and 14N) (Fig. 7), both frequencyand density increase in Buffer 2 (where clast density decreased inthe control sample), NN statistics indicate robust and significantclast clustering in Buffer 2 (where random distributions wereidentified in the control sample), and clast frequency and densitywere particularly low in Buffer 3 (Fig. 8). Seventeen features wererejected due to their failure to meet one or more of the criteriadescribed in the preceding section (Table 3). Type 1 and Type 4errors were the most common (at 10 each), with four featuresrejected for generating both Type 1 and 4 errors. Five features wererejected for only Type 1 errors and six were rejected for only Type 4errors. The remaining two rejected features were rejected for Type1, 2 and 3 and Type 2 and 3 errors, respectively.

5. Discussion

Methodologically, three main considerations stand out in thisanalysis. The first is the necessity of point plotting all artifacts andecofact-sized clasts at sites where feature identification is prob-lematic. Though certainly more time consuming than othermethods, point data have the utility of allowing precise geo-statistical analyses at multiple scales and can easily be transformedinto gridded or other types of lower-resolution data. Second is thenecessity of generating a control sample of naturally-occurringclasts from the same geomorphic setting as the site in question inorder to develop a comparative and quantitative measure of thenatural frequency, density and distribution of clasts contained

Table 3Frequency of error types resulting in feature rejection.

Feature Type 1Buffer 2clustering

Type 2Interiorfrequency

Type 3Interiordensity

Type 4Exteriordensity

7S X X12S X15S X18S X X X4N X5N X X6N X12N X16N X X17N X18N X19N X22N X24N X25N X X26N X X27N XTotal 10 2 2 10

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283126

therein. The last is that the methods described here do not operatein an empirical vacuum, but rather rely on prior knowledge offeature size and construction in the area inwhich suchmethods areemployed; in this case by the ethnographic and archaeological dataon stone circle size and morphology used to develop the expecta-tions and buffer sizes used in the analysis.

The utility of the rules for rejecting potential cultural featureswas found to be variable. Based on the frequency of which rulesresulted in rejecting potential cultural features at the site, it is clearthe NN statistic used to identify clustering in Buffer 2 (i.e., Type 1errors) and increased frequency and/or density in the buffer (Buffer3) outside of hypothetical features provide the most robust meansfor either accepting or rejecting clast accumulations as culturalfeatures. The comparative frequency (Type 2 errors) and density(Type 3 errors) of clasts in internal (Buffer 1) and hypothetical rockring (Buffer 2) buffers were found to be less robust, resulting in therejection of only two possible rings. Though perhaps less criticalthan the other two rules, these data are nonetheless important dueto the fact that they result in a more conservative and stringentmeans by which to assess whether rock features were culturallyderived.

In terms of human ecology and population size, it appears thereare 30 residential features at the site. Historically, tipis in the greaterregion held between eight and 16 individuals (Laubin and Laubin,1977; Voget, 2001), meaning as many as 240e480 people couldconceivably have occupied the site at one time. This high number isdoubtful, however, because 48TE479 appears to have been occupiedseveral times (unlike many stone circle sites on the Plains), withseveral partially disarticulated and overlapping stone circles char-acterizing the site. In fact, the dearth of artifacts at the site andpalimpsest-like nature of stone circle site formation processes isconsistent with several short-term yet fairly large residential groupaggregations characteristic of the high residential mobility (sensuBinford, 1980) documented on the northern Plains during the LatePrehistoric andHistoric periods (Kornfeld et al., 2010). The relativelylarge diameters of stone circles at the site are arguably consistentwith such a temporal assignation (Kehoe, 1960).

The problem of determining group size during repeated occu-pations at a site such as this, however, is clearly related to the factthat it is a surface site (sensu Tainter, 1998) representing more thana single occupation. Not knowing exactly howmany episodes of siteoccupation occurred at the site limits inferences as to how manystone circles may have been occupied contemporaneously (e.g.,

Wandsnider, 1996, 2008). Without greater temporal control,through radiocarbon assay, diagnostic artifacts, or other methodsused to date individual stone circles (all lacking at 48TE479),palimpsest-like surface sites such as 48TE479 still face the long-running and vexing problem of dating High Plains surface de-posits (a similar problem is common to much of the surfacearchaeology of the Great Basin). Post-depositional processes alsoaffect the ability to infer group size at sites such as this, with at leastone rock feature (4N) likely representing a legitimate stone circle(Fig. 2), but one so disturbed by surface processes that it failed tomeet the deliberately-conservative and stringent criteria discussedabove. At sites with less evidence of overlapping rings and lessevidence of surface disturbance, as is fairly common on the Plains(Kehoe, 1983), such complications become less problematic. In anyevent, despite difficulties of assessing contemporaneity of stonecircle construction and use at 48TE479, the methods developed inthis paper do provide an objective means of determining featurefrequency and distribution in areas where determining such hasproven difficult in the past.

6. Conclusions

This study has developed straightforward techniques for iden-tifying spatial patterning in point data that are consistent with anattribution of cultural, rather than natural site formation processes.These techniques compare frequency, density and distribution ofthese data to naturally-occurring clast distributions, to expecta-tions derived from ethnoarchaeological data and to models pre-dicting how humans modified and used residential space. Theutility of this method is the objective identification of residentialfeatures in landscapes and settings where determining whethernatural or cultural processes resulted in the distribution of clasts atarchaeological sites is confounded by high-density clastic site sur-faces. Using these conservative measures, these methods may beused to better inform reconstructions of group size, composition,and type of mobility, with the caveat that determining the exactnature and timing of different site formation processes at surfacesites continues to confound efforts to accurately make these typesof reconstructions. Though the latter is clearly beyond the scope ofthis paper, the methods it presents represent an important firststep in accurately and reproducibly identifying and quantifyingsurface features at hunteregatherer residential sites.

Acknowledgments

Special thanks are extended to Bridger-Teton National Forestarchaeologist Jamie Shoen, who initiated and helped support ourresearch at the site. Thanks also go to 2010 Utah State Universityarchaeological field school students and staff. Any errors or omis-sions, however, are our own. Partial funding was provided by NSFgrant #960077 for data collection and analysis.

References

Aaberg, S., 1975. Comprehensive stone circle site mapping. Archaeol. Mont. 16, 1e12.Adams, G., 1978. Tipi Rings in Southern Alberta. The Alkali Creek Sites, Lower Red

River, Archaeol. Survey of Alberta. In: Occassional Paper No. 9. Edmonton.Anquandah, J.R., 1986. Investigating the stone circle mound sites and art works of

Komaland, N. Ghana. Nyame Akuma 27, 10e13.Baker, S.G., 2003. Historic Ute Archaeology: interpreting the last hour wickiup.

Southwest. Lore 69, 1e34.Bettinger, R.L., 1975. Late prehistoric and historic structures in Owens Valley,

Eastern California. J. Calif. Anthropol. 2, 198e204.Bettinger, R.L., 1977. Aboriginal human ecology in Owens Valley: prehistoric change

in the Great Basin. Am. Antiq. 42, 3e17.Bettinger, R.L., 1989. The Archaeology of Pinyon House, Two Eagles, and Crater

Middens: Three Residential Sites in Owens Valley, Eastern California. AmericanMuseum of Natural History, New York.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e3128 3127

Bettinger, R.L., 1991. Aboriginal occupation at high altitude: Alpine villages in theWhite Mountains of California. Am. Anthropol. 93, 656e679.

Bettinger, R.L., 2008. High altitude sites in the Great Basin. In: Fowler, C.S.,Fowler, D.D. (Eds.), The Great Basin. School for Advanced Research Press, SantaFe, New Mexico, pp. 87e97.

Binford, L.R., 1980. Willow smoke and dog’s tails: hunter-gatherer settlement sys-tems and archaeological site formation. Am. Antiq. 45, 4e20.

Binford, L.R., 1990. Mobility, housing, and environment: a comparative study.J. Archaeol. Res. 46, 119e152.

Blot, J., 1991. The stone circle of Urdanarre Sud 1 (excavation 1989). Munibe 43, 1e6.Bocquet-Appel, J.-P., Demars, P.-Y., Noiret, L., Dobrowsky, D., 2005. Estimates of

Upper Palaeolithic meta-population size in Europe from archaeological data.J. Archaeol. Sci. 32, 1656e1668.

Brumley, J.H., Kooyman, B.P., 1978. Excavations of Stone Circles at EhPb-1. In:Archaeological Survey of Alberta Occasional Paper. Archaeological Survey ofAlberta, Edmonton.

Casteel, R.W., 1979. Human population estimates for hunting and gathering groupsbased upon net primary production data: examples from the Central Desert ofBaja California. J. Anthropol. Res. 35, 85e92.

Clark, D.L., 1972. Models in Archaeology. Methuen, London.Clark, P.J., Evans, F.C., 1954. Distance to nearest neighbor as a measure of spatial

relationships in populations. Ecology 35, 445e453.Connolly, J., Lake, M., 2006. Geographical Information Systems in Archaeology.

Cambridge University Press, Cambridge, U.K.Davis, L.B., 1983a. Introduction to tipi ring problems and research. Plains Anthropol.

28, 1e5.Davis, W.E., 1983b. A morphological analysis of stone circles from the Copper

Mountain Project, Shoshoni, Wyoming. Plains Anthropol. 28, 71e79.Diehl, M.W., 1992. Architecture as a material correlate of mobility strategies: some

implications for archaeological interpretation. Behavior Sci. Res. 26, 1e35.Donaldson, M.W., Welch, J.R., 1991. In: Beckett, P.H., Mogollon, V. (Eds.), Western

Apache Dwellings and Their Archaeological Correlates. Coast Publishing andResearch, Las Cruces, New Mexico, pp. 93e105.

Durand, S.R., Pippin, L.C., Spennman, D.H.R., 1992. A pragmatic application of thenearest neighbor statistic. J. Field Archaeol. 19, 263e274.

Ebdon, D., 1976. On the underestimation inherent in the commonly used formulae.Area 8, 391e399.

Eerkens, J.W., 2008. Rock-ring features on the shores of Owens lake and implica-tions for prehistoric geophyte processing and storage. Proc. Soc. Calif. Archaeol.21, 180e184.

Eogan, G., 1964. The excavation of a stone alignment and circle at Cholwichtown,Lee Moor, Devonshire, England. Prehist. Soc. Proc., 25e38.

Finley, J.B., Scheiber, L.L., 2007. Documenting Domestic Landscapes at Crow StoneCircle Sites in Bighorn Canyon, Southern Montana and Northern Wyoming,Bighorn Canyon National Recreation Area. Lovell, Wyoming and Weatern Na-tional Parks Association, Tucson, Arizona.

Finnigan, J.T., 1980. Interpreting tipi ring structure. Na’pao: Sask. Anthropol. J. 10,1e6.

Finnigan, J.T., 1981. Regular or random: the significance of tipi ring rock distribu-tions. In: Wilson, M., Road, K.L., Hardy, K.J. (Eds.), Megaliths to MedicineWheels: Boulder Structures in Archaeology, Proc. 11th Annual Chacmool Con-ference. The Archaeological Association of the University of Calgary, pp. 69e78.

Frison, G.C., 1967. The Piney Creek Sites, vol. 3. University of Wyoming Publications,Wyoming, pp. 1e92.

Frison, G.C., 1983. Stone circles, stone-filled fire pits, grinding stones and high plainsarchaeology. Plains Anthropol. 28, 80e91.

Grayson, D.K., Cannon, M.D., 1999. Human paleoecology and foraging theory in theGreat Basin. In: Beck, C. (Ed.), Models for the Millenium. University of UtahPress, Salt Lake City, pp. 141e151.

Hardesty, D.L., 1983. Ecology, economics, and evolutionary explanation in economicprehistory. In: Bronitsky, G. (Ed.), Ecological Models in Economic Prehistory.Anthropological Research Papers No. 29. Arizona State University, Tucson,pp. 4e17.

Hester, T.R., Shafer, H.J., Feder, K.L., 2008. Field Methods in Archaeology, seventh ed.Left Coast Press, Walnut Creek, California.

Hodder, I., Hassell, M., 1971. The non-random spacing of Romano-British walledtowns. Man 6, 391e407.

Hodder, I., Orton, C., 1976. Spatial Analysis in Archaeology. Cambridge UniversityPress, Cambridge.

Hoffman, J.J., 1953. Comments on the Use and Distribution of Tipi Rings in Montana,South Dakota and Wyoming. In: Anthropology and Sociology Paper 14. Mon-tana State University, Missoula.

Horde, D.M., 1982. Stone circle research and the ethnographic record. South DakotaArchaeol. 6, 33e46.

Kehoe, T.F., 1958. Tipi rings: the “Direct Ethnological” approach to an archaeologicalproblem. Am. Anthropol. 60, 861e873.

Kehoe, T.F., 1960. Stone Tipi Rings in North-Central Montana and the AdjacentPortion of Alberta, Canada: Their Historical, Ethnological and ArchaeologicalAspects. In: Bureau of American Ethnology Anthropological Bulletin No. 173,Anthropological Papers No. 62, pp. 421e473. Washington, D.C.

Kehoe, T.F., 1983. A retrospectus and commentary. Plains Anthropol. 28, 326e342.Kelly, R.L., 1992. Mobility/sedentism: concepts, archaeological measures, and ef-

fects. Annu. Rev. Anthropol. 21, 43e66.Komai, K., 1961. On the stone circle of Sawsuchi in Iwate prefecture. Kokogaku

Zasshi 47, 65e67.

Kornfeld, M., Frison, G.C., Larson, M.L., 2010. Prehistoric Hunter-Gatherers of theHigh Plains and Rockies. Left Coast Press, Walnut Creek, California.

Ladefoged, T.N., Pearson, R., 2000. Fortified castles in Okinawa Island during theGusuku period, AD 1200e1600. Antiquity 74, 404e412.

Larson, T.K., 1979. Dated stone circle sites in Wyoming. Wyo. Archaeol. 22, 9e17.Larson, T.K., 1981. Dated stone circle sites in Wyoming. In: Wilson, M., Road, K.L.,

Hardy, K.J. (Eds.), Megaliths to Medicine Wheels: Boulder Structures inArchaeology, Proceedings of the 11th Annual Chacmool Conference. TheArchaeological Association of the University of Calgary, pp. 93e100.

Laubin, R., Laubin, G., 1977. The Indian Tipi: Its History, Construction, and Use.University of Oklahoma Press, Norman.

Loendorf, L.L., Weston, L.O., 1983. An examination of tipi rings in the BighornCanyon-Pryor Mountain area. Plains Anthropol. 28, 146e155.

Love, C., 1971. 48TE479 Archaeological Site Record, On File, United States ForestService. Bridger-Teton National Forest, Jackson, Wyoming.

Malouf, C.I., 1961. The tipi rings of the high plains. Am. Antiq. 26, 381e389.Mathews, R.H., 1895. The bora, or initiation ceremonies of the Kamilaroi tribe.

J. Anthropol. Inst. G. B. Irel. 24, 411e427.Mobley, C.M., 1983. A statistical analysis of tipi ring diameters at sites near Santa

Rosa, New Mexico. Plains Anthropol. 28, 101e112.Moore, K.M., 1998. Measures of mobility and occupational intensity in highland Peru.

In: Rocek, T.R., Bar-Yosef, O. (Eds.), Seasonality and Sedentism, ArchaeologicalPerspectives from Old and New World Sites, Peabody Museum of Archaeologyand Ethnology. Harvard University, Cambridge, Massachusetts, pp. 181e197.

Morgan, C., 2008. Reconstructing prehistoric hunter-gatherer foraging radii: a casestudy from California’s southern Sierra Nevada. J. Archaeol. Sci. 35, 247e258.

Morgan, C., 2009. Climate change, uncertainty and prehistoric hunter-gatherermobility. J. Anthropol. Archaeol. 28, 382e396.

Morgan, C., 2012. Modeling modes of hunter-gatherer food storage. Am. Antiq. 77,714e736.

Morgan, C., Adams, R., Losey, A., 2012a. High altitude residential occupations inWyoming’s Wind River range. N. Am. Archaeol. 33, 35e79.

Morgan, C., Pack, R., Adams, R., Cannon, K., 2012b. Applications of LiDAR to theRemote-Sensing of High-Altitude Residential Features in Wyoming’s WindRiver Range, Poster presented at the 2012 Society for American ArchaeologyMeetings, Memphis, Tennessee.

Morris, E.A., Mayo, D., Blakeslee, R.C., Brower, P.W., 1983. Current perspectives onstone ring structures in Northeastern Colorado. Plains Anthropol. 285, 45e58.

O’Connell, J.F., 1987. Alyawara site structure and its archaeological implications. Am.Antiq. 52, 74e108.

Oetelaar, G.A., 2000. Beyond activity areas: structure and symbolism in the orga-nization and use of space inside tipis. Plains Anthropol. 45, 35e61.

Patrick, J.D., Wallace, C.S., 1982. Stone circle geometries: an information theoryapproach. In: Heggie, D.C. (Ed.), Archaeoastronomy in the Old World. Cam-bridge University Press, Cambridge, pp. 231e264.

Pinder, D., Shimada, I., Gregory, D., 1979. The nearest-neighbor statistic: archaeo-logical application and new developments. Am. Antiq. 44, 430e445.

Quigg, J.M., 1981. Stone circle excavations in Alberta to 1978: a summary. In:Wilson, M., Road, K.L., Hardy, K.J. (Eds.), Megaliths to Medicine Wheels: BoulderStructures in Archaeology, Proceedings of the 11th Annual Chacmool Confer-ence. The Archaeological Association of the University of Calgary.

Reher, C.A., 1983. Analysis of spatial structure in stone circle sites. Plains Anthropol.28, 192e222.

Roll, T.E., 1981. Tipi rings of North-Central Montana: expression and analysis. In:Wilson, M., Road, K.L., Hardy, K.J. (Eds.), Megaliths to Medicine Wheels: BoulderStructures in Archaeology, Proceedings of the 11th Annual Chacmool Confer-ence. The Archaeological Association of the University of Calgary, pp. 79e92.

Sawada, M., 2002. Nearest Neighbor Analysis/Event-Event Distances, ElectronicDocument. http://arcscripts.esri.com/details.asp?dbid¼12227 (accessed03.01.09.).

Scheiber, L.L., Finley, J.B., 2010a. Domestic campsites and cyber landscapes in theRocky Mountains. Antiquity 84, 114e130.

Scheiber, L.L., Finley, J.B., 2010b. Mountain Shoshone Technological Transitionsacross the Great Divide. In: Scheiber, L.L., Mitchell, M.D. (Eds.), Across a GreatDivide: Continuity and Change in Native North American Societies. Universityof Arizona Press, Tucson, pp. 128e148.

Schiffer, M.B., 1987. Formation Processes of the Archaeological Record. University ofNew Mexico Press, Albuquerque.

Schreiber, K.J., Kintigh, K.W., 1996. A test of the relationship between site size andpopulation. Am. Antiq. 61, 573e579.

Seymour, D.J., 2009. Nineteenth-century Apache wickiups: historically documentedmodels for archaeological signatures of the dwellings of mobile people. An-tiquity 83, 157e164.

Simms, S.R., 1989. The structure of the Bustos Wickiup site, eastern Nevada. J. Calif.Great Basin Anthropol. 11, 2e34.

Smith, C.S., 2003. Hunter-gatherer mobility, storage, and houses in a marginalenvironment: an example from the mid Holocene in Wyoming. J. Anthropol.Archaeol. 22, 162e189.

Smith, C.S., McNees, L.M., Reust, T.P., 1995. Site structure of two buried stone circlesites, southern Wyoming. Plains Anthropol. 40, 5e21.

Smith,M.B.,1974. Rapidmethod for recording stone circles. Archaeol.Mont.15, 47e54.Soper, R.C., 1977. The stone-circle graves at Ng’amoritung’a, southern Turkana

District, Kenya. Azania 12, 193e208.Stiger, M., 2006. A Folsom structure in the Colorado Mountains. Am. Antiq. 71,

321e351.

C. Morgan et al. / Journal of Archaeological Science 40 (2013) 3117e31283128

Surovell, T.A., Waguespack, N.M., 2007. Folsom hearth-centered use of space atBarger Gulch, locality B. In: Brunswig, R.S., Pitblado, B. (Eds.), Emerging Fron-tiers in Colorado Paleoindian Archaeology. University of Colorado Press,Boulder, pp. 219e259.

Tainter, J.A., 1998. Surface archaeology: perceptions, values, and potential. In:Sullivan, A.P. (Ed.), Surface Archaeology. University of New Mexico Press,Albuquerque, pp. 169e179.

Thomas, D.H., 1982. The 1981 Alta Toquima Village Project: a Preliminary Report.Desert Research Institute Social Sciences and Humanities. Publications 27.

Voget, F.W., 2001. Crow. In: DeMallie, R.J. (Ed.), Plains. Smithsonian Institution,Washington, D.C, pp. 695e717.

Wandsnider, L., 1996. Describing and comparing archaeological spatial structures.J. Archaeol. Method Theory 3, 319e384.

Wandsnider, L., 2008. Time-averaged Deposits and Multitemporal Processes in theWyoming Basin, Intermontane North America: a Preliminary Consideration ofLand Tenure in Terms of Occupation.

Washburn, J., 1974. Nearest neighbor analysis of Pueblo IeIII settlement patternsalong the Rio Puerco of the East, New Mexico. Am. Antiq. 39.

Waters, M.R., Kuehn, D., 1996. The geoarchaeology of place: the effect of geologicalprocesses on the preservation and interpretation of the archaeological record.Am. Antiq. 61, 483e497.

Whiting, J.W., Ayers, B., 1968. Inferences from the shape of dwellings. In:Change, K.C. (Ed.), Settlement Archaeology. National Press, Palo Alto, California,pp. 117e133.

Wilson, M.C., 1983. A test of the stone circle size-age hypothesis: Alberta andWyoming. Plains Anthropol. 28, 113e138.

Winterhalder, B., Baillargeon, W., Cappelletto, F., Daniel, J., Prescott, I.R., 1988. Thepopulation ecology of hunter-gatherers and their prey. J. Anthropol. Archaeol. 7,289e328.

Zorn, J.R., 1994. Estimating the population size of ancient settlements:methods, problems, solutions, and a case study. Bull. Am. Sch. Oriental Res.295, 31e48.