409

Multi- Scale and Nested-Intensity SamplirgTechniques for Archaeological Sunrey

Oskar BurgerThe University of New MexicoAlbuquerque, New Mexico

Lawrence C. Todd

Paul BurnettColorado State UniversityFort Collins, Colorado

Tomas J. StohlgrenU. S. Geological Survey

Fort Collins Science Center. Fort Collins. Colorado

Doug StephensU. S. Forest ServiceChadron, Nebraska

This paper d.iscwsses sornpling tecl,tniqwes for archaeohgicnl suvyey tbnt are d,irected. toward,wnlwntirug the properties of swfnce ortifuct d.istribwtiorus. The sarnpling tecbniques we experi-rnented. with consist of o rnulti-scnle sornpling plot d.weloped. in plant ecology and, the wse of nnested,-intensity suwey design. We present reswlts frorn the initinl npplication of these noethod;.Tbe snrnpling technique we bowowed. frorn plnnt ecol.ogy is the Mod.ifl.ed.-Whittaker rnwlti-scole snrnpling pht, wbich gathers obsewations at the spntial scnles of I sq rn, 10 s4 zM, 100 sqrn, nnd. 1000 sq m.Nested.-intensity swveys gnther obsewntions on tbe st.vne snrnple wnits ntrnaltiple resolutiotos. We cornpare the reswlts of n closely-spnced. wnlking sav'vey n crawling swr,-uey, and. a. test excovntion to n d,epth of 10 crn. These techniques were applied, to ten 20 x 50rn suruey plots dixribwted. oper A.n a.rea. of 418 hn near tbe Hwd.son-Meng Bison Bonebed. inxw l{ebraskn. These nppn oncbes cntc signfficontly irnproue tbe nccwrncy of suwey d.ato. Ourreswlts show thot high-resolwtion couernge techniques overlook rnore rnoterial than nrchoeoh-gists haue suspected. The conobined. npproncLtes of rnwlti-scnle nnd. nested.-intensity sarnplingprovide naa tools to irupn ore owr abiliry tu investignte the properties of sutfnce record.s.

Introduction of the diversity of potential applications of survey darArchaeological survey has changed dramatically over the the challenges associated with obtaining and underr

years. V\4rile at one time the need to survey had to be jus- ing them, multiple concerns may compete for priortified, the concern with regional patterning has continual- the selection of field techniques. Surveys may need t<ly developed and today survey is among the most funda- er very large areas in little time as well as account for tmental techniques of archaeological inquiry (Ammerman fects of a variety of taphonomic factors that influenrI98f ; Banning 2002; Schiffer, Sullivan, and I(linger regional record. Ideally, the data supporting inter1978). The goals of survey projects and the demands tions and management decisions are efficiendy gatplaced on survey data have become more diverse and com- but also of sufficient quality to accurately represent ttplicated with time. The observations made by crews walk- tributional properties of the regional record.ing parallel lines over the landscape must address issues Both management and research perspectives (altlranging from finding and protecting sites from damage to these need not be mutually exclusive) can benefitinvestigating the nature of past land-use strategies. Because large spatial samples that reveal associations betwee

410 Sa*pli"g Techniqwes for Archaeological Swwey/Bwrger, Tod.d' Bwrnett, Stnhlgren, and, Stephens

Figure I. Location of the Oglala National Grassland, Nebraska within the United States

ography and material culture and a thorough understand-ing of the processes that influence the distributions thatsurveys document. Accordingly, the numerous goals of ar-chaeological survey are here grouped into discovery-basedand property-based modes of investigation. Discovery-based surveys identify geographical aspects of the surfacerecord by locating and describing clusters of artifacts. Aproperty-based approach focuses on evaluating accurary ofmethod and technique and formational aspects of the re-gional record (Ebert and Kohler I9BB; Shott 1995; Wand-snider and Camilli L992). The capacity to analyze and in-terpret the contexts of discovery could be improved byconducting more detailed analyses of the factors that influ-ence the surface samples that we use to infer spatial rela-tionships. This requires embracing an explicidy experimental approach and integrating property-based investiga-tion as one of several phases in a survey project (Given etal,. L999; Schiffer, Sullivan, and Klinger l97B). Siteless or

distributional surveys have made major contributions inthis regard (Ebert 1992;Foley I98I; Thomas 1975). Bysampling regions as opposed to "sitesl' these surveys at-tempt more holistic and accurate interpretations of land-scape records.

We present two additions to the conventional surveytool*it that are aimed at developing property-based inves-tigation. The first is dre use of a multi-scale sampling plotdev,eloped in plant ecology. The second is a nested-intensi-ty survey design that covers each plot with more-than-oneobserver resolution. These techniques were applied in anarchaeological survey on the Oglala National Grassland(ONG) of ltw Nebraska (rrc. r). A property-based ap-proach is valuable for investigating geomorphologically ac-tive and topographically diverse archaeological landscapeslike the ONG (rrc. z), but the basic principles are applica-ble to many other sampling situations whether landscape-oriented or intra-site.

South Dakota

Wyoming

OglalaNationalGrassland Nebraska

I

Colorado

Kansas

Figure 2. The Oglala National Grassland landscape.

The technique we borrowed from plant ecology is theModified-Whittaker multi-scale vegetation sampling plot(rrc. l). This sampling design was applied to archaeologi-cal survey because ofthe increased accurary ofthe plant di-versity samples gathered with this method, the experimen-tal control that the plot provides, and because of its multi-scale design. Plants and artifacts share some basic distribu-tional properties, so it follows that a survey design that ishighly usefirl in plant species surveys may also be usefi.rl forsampling artifact distributions. "r{rtifacts share many prop-erties with plants, in having small unit size in relation to avery large spatial contexq and also by having a patchy dis-tribution" (Foley l9&l: 174). The size of the Modified-S4rittaker plot is 1000 sq m, which is a very small areacompared to conventional designs. Orton (2000: BB) con-siders survey units that are 500 to 1000 m on a side as"small" archaeological sample units. Thus, we were initial-ly hesitant to apply the Modified-S4rittaker plot direcdy toarchaeological survey. fu we became more aware of the im-provement in the accurary and reliability of plant speciessurveys conducted with this method, howeveE we decidedto test its utility in an archaeological case study.

Jowrnal of Field. Archa.eology@l. 29, 2002-2004 4Ll

An additional methodological issue carne to light whenone of our colleagues in plant ecology wanted to knowhow archaeologists estimate the number of artifacts theymiss, or walk past, when conducting a pedestrian survey.We realized that we had no way to reliably answer thisquestion with conventional techniques (Banning 2002:62) and thus we elected to conduct a nested-intensity sur-vey to address this issue. A nested-intensity survey consistsof covering the same sample units with different resolu-tions. We covered each of the I0 Modified-Whittaker plotsin this study with a fine-grained walking survey) a crawlsurvey, and then conducted test excavations ofthe top 10cm. Comparing the results of the samples obtained by eachtechnique provides a means for investigating the effects ofbasic decisions on the estimates concerning the overallpopulation. The intensive coverage provides high-resolu-tion samples of limited areas that can be used to evaluatethe accurary ofour approach.

Before discussing the survey in greater detail, we shouldemphasize that the techniques of multi-scale and nestedsampling are strategies for systematic sub-sampling thatcan be used to augment traditional suwey practice. These

4L2 Sarnpli.ng Techni.qwes for Archaeohgicwl SwweyfBwrger, Tod.d., Bwmett, Stohlgren, and' Stepherus

41 m

A B

Figure 3. The Modified-Whittaker multi-scale sampling plot. A) The layout of the 20 x 50 m plot. Then u m b e r e d p l o t s ( f t o l 0 ) a r e 0 . 5 x 2 m , t h e A a n d B p l o t s a r e 2 x S m , a n d t h e C p l o t i s 5 x 2 0 m ; B )Plot layout with guides for arranging subplots. The location ofeach subplot is indicated as a distance inmeters from the anchor corner, marked by a 0 m in the lower right corners of the K plot and subplot C.

are not considered replacements for conventional tech-niques that gather coarser-grained samples of larger spatialextent. Certain aspects of survey certainly require largecoarse-grained samples. The small area covered by this ap-proach could not be used on its own to achieve an under-standing of a regional record but dris small-scale samplingplot is ideal as an experimental framework for property-based investigations. The nature and significance of theitems found by discovery-based surveys could be placed in-to more informative contexts if they were complementedwith high-resolution samples aimed at evaluating the prop-erties ofthe record itself. The techniques we present are in-tended to enhance discovery by focusing on the evaluativegoals of survey which are often not given as much empha-sis in the design of regional sampling schemes.

The Modified-Whittaker Plot and Multi-ScaleSampling

The Modified-Slhittaker multi-scale sampling plot is animprovement over conventional rangeland techniques forstudying plant diversity (Stohlgren, Falkner, and Schell1995; Stotrlgren, Bull, and Otsuki 1998). In comparativestudies, the Modified-Whittaker plot outperformed tradi-tional plant sampling designs by documenting more plantspecies, capturing more rare species, and more accuratelyrepresenting the relative abundances of species in the com-munity. The traditional techniques that typify plant species-richness surveys are analogous to those in archaeology.Conventional plant survey methods involve documentingspecies within quadrats placed along a transect, whereas ar-chaeological surveys consist of long linear transects walked

K

B

by surveyors who continually search the ground surface forcultural material. Unlike plant suwey techniques, archaeo-logical surweys are often conducted without a specific re-gard for spatially defirred sample units. The transect-basedplant sampling methods over-represented the dominantplants while failing to capmre representative proponionsof non-native or rare species (the many species that eachcontribute less than lolo ofthe total vegetation cover) andoften missed them altogether (Stohlgren, Bull, and Otsuki1998: 168). The trarsect methods were also poor predic-tors of the total number of species, missing about half ofthe native and non-native plant species in all habitat types.Generally, similar observations can be made ofarchaeolog-ical survey medrods.

Wandsnider and Camilli (1992) dernonstrated that asirnilar bias often results from pedestrian transect methodsused in archaeological suwey, which tend to over-representdominant parts of the surface record. Wandsnider andCamilLit (1992: 184) invesrigation of survey accuracyfound drat "survey with an acceptable transect interval of15 m will intercept, at most, 6-137o of the members of alow-density artifact population, but only some of theseitems will actually be found." A tralsect spacing of 5 m wasmore like\ to find dustered rather than isolated artifacts.In an experiment involving seeded artifacts (painted wash-ers and nails), crews recovered 820lo of the clustered mate-rial compared to just 167o of the isolated material (Wand-snider and Camilli 1992). It is therefore likely that tradi-tional pedestrian sur-veys result in shaky inferences aboutlandscape patterns due to a lack of understanding of theproperties of the sample.

The improved performance of the Modified-Whittakerplot over conventional plant sampling is based on the spa-tia.l layout ofthe subplots (rrc. 3a). The 0.5 x 2 m subplots(rrc. 3e, numbered I to I0) are arranged to reduce t}reamount of spatial autocorrelation in the sample, which isthe principle that two points dose to one another are morelikely to be similar than two points that are further apart.\A4thin the 1000 sq m plot (K), these I sq m subplots arearranged to improve the accuracy of their spatial coverage.In archaeology, this is usefi.rl as a metlod for gathering sys-tematic small-scale samples of the surface record and alsooffers an ideal affangement for test excavations. Standardpedestrian transects, which are essentially "numerous

quadrats placed end-to-end" (Onon 2000; 90), have ahigh degree of spatial autocorrelation. The Modified-S4rittaker plot's multi-scale design can be used to counterthis problem and results in more representative samples(Stotrlgren, Falkner, and Schell 1995).

The incremental increase in the spatial scales of the sub-plots within the Modified-S/hittaker ftame leads to its in-

Joamal of Fietul Archaeokgyflol. 29, 2002-2004 413

creased utility for evaluating population variab.les. The tenI sq m subplots are placed along the internal boundary ofthe 1000 sq m plot and along the external edge ofthe 100sq m C subplot (rrc. 3a). Subplots A and B are 2 x 5 m indimension (10 sq rn) and are in opposite corners of theModified-\4/hittaker plot. The central rectangle, plot C,covers an area of 100 sq m and measures 5 x 20 m (rrc.

3A). The exact distances between subplots are presented inFigure 3b, which is based on tracing the perimeter of theplot with a I00 m tape measure anchored at the lower rightcorner of the plot. All of the subplots within the Kplot arenon-overlapping so tiat ttre properties of each samplingunit can be evaluated independendy (Stohlgren, Fa.lkner,and Schell 1995). This allows the properties ofthe sampledpopulation to be anallzed with respect to changes ir.r spa-tial scale. Combining this design with siteless suwey mayaid in eliminating the bias of favoring obtrusive and denseponions of the archaeological record.

Accurate Samples as Tools for BvaluatingDiscovery

One major value of a nested-intensiq' suwey is that itprovides archaeologists widr the rneans to evaluate t}te ac-curary of a surface sample in terms of the number of itemsthat surveyors did not discover. High-resolution coveragecan be used to investigate how the variables of the sampledeviate from those ofthe population. "[T]he systematic ac-quisition of the whole archaeological record, including thelow-density areas, can yield more valid and reliable insightsinto the nature of the archaeological record, even the na-ture of archaeologica.l sites" (Dunne.ll and Dancey 1983:274) . Lkewise, Cowgill pointed out that the "most obvi-ous way to get relatively good information on even verytiny occurrences (insofar as they are visible at all on the sur-face) is to increase the intensity ofsurvey by spacing surveyteam members so closely that nearly al.l tiny occurrenceswill be spotted' (1990: 257). S4.rile Cowgill was not in-tentionally referring to crawl surveys (personal communi-cation, 2001), archaeologists should have an understand-ing for just how closely suweyors need to be spaced in or-der to actually find "nearly all tiny occurences."

In order to evaluate the properties of a large spatial sam-ple, archaeologists may find it usefirl to include intensivelysarnpled surwey plots, even if it means reducing their over-all spatial coverage. In excavations, archaeologists havetended to sacrifice area in favor of more thorough record-ing strategies, producing smaller samples at higher resolu-tion. Such fine-grained samples favor documenting con-text over obiect discoveirr and are based on the realizationdrat there is more information contained in the relation-ships between objects and their surroundings than in the

414 Sawtpkng Tbchniqwes for Archneological Swwey/Bwrger, Tod.d., Bwrnett, Stohlgren, and. Stephens

Figure 4. Colorado State University sflrdents conducting a crawl survey.

objects themselves (Clarke 1977). Similarly, the mesh sizesused to screen archaeological deposits have become pro-gressively smaller: rl4-inch dry screen was once the normand now many archaeologists water screen at least a por-tion of their deposits through r/r6-inch mesh. Tlansectspacing can be considered as analogous to screen sizes inthe context of survey methods. Tiaditional techniques havefocused on the highly dense aspects of the record, whilemuch of the rest has slipped by in the gaps between tran-sects.

The ONG Survey Project and Method

Our study area is the grassland surrounding the Hud-son-Meng Bison Bonebed in lrw Nebraska. The survey wasconducted as part of the Colorado State University archae-ological field school in cooperation with the United StatesForest Service during the summers of I99B-200I on theOglala National Grassland (ONG Survey Project). The re-gion consists of rolling northern mixed prairie, pine-treecrowned ridges, and badland exposures (rrc. z). Chippedstone is by far the dominant artifact class and the prehis-toric record was generally dominated by hunter-gadrererbehavior patterns. The initial goals of the survey were toevaluate dre utility of the Modified-Whittaker plot for ar-chaeological survey and to investigate the effects of ob-ser-ver intensiw on artifact recoverv.

N e st e d, - I nt e n s i.ty S az'np kng

Nested-intensity sampling covers the same section of

ground at progressively finer resolutions. For the ONGsurveys) the first coverage involved walking one set of par-allel transects over the entire 1000 sq m I( plot with about7O cm between surveyors. The guideline for crew spacing

was that each surveyor could touch the shoulder of the next

person in the survey line. The second coverage consisted ofa different survey team crawling, with shoulders touching,over subplots I to 10, A, B, and C (rrc. +). The total area

crawled was 130 sq m (nrc. 3e: subplots I to I0 are I sqm, A and B are l0 sq m, and C is I00 sq m). All discoveredartifacts from both surveys were left in place and pains weremade to return them to their original locations after

recording. As a result, the crawling survey would rediscov-er many of the artifacts recorded by the walking survey. Af-

ter the walking and crawling surveys were completed andall discovered artifacts had been recorded, the top I0 cm ofsediment of each 0.5 x 2 m subplot (subplots I to l0) wereexcavated and passed through r/8-inch screen.

The excavation is considered to be a measure of the pop-ulation of artifacts that could potentially be exposed and

discovered. The subsurface test is necessary because ar-chaeologists often use surface artifact densities as predic-tors of subsurface densities, but several processes can ex-

Thble l. Summary data for documented chipped stone in all Modified-Shittaker plots. Plotlocations are shown in Figure 5. Means are calculated by taking the total for each column anddividing it by the total number of subplots. For subplots I-10, each total is divided by I00;for subplots A and B, totals are divided by 20; for subplot C and plot K, the totals are dividedby I0.

Jowmal of Field. Archneohgy/Vol. 29, 2002-2004 4Ls

mented with two approaches to plot set-up. The first washighly controlled and time-consuming in that the exactprovenience of each corner of each subplot was fit to theUTM grid with precision mapping equipment, locus GPSreceivers or a total stationl both provide sub-centimeter ac-curacy. When the multi-scale plot was first used we wereconducting a number of experiments with survey intensityand spatial variability that demanded such accuracy. As theproject developed, however, expediency of set-up becamemore of a concern and a more "low-tech" and time efficientmethod was adopted that involves tape measures and acompass. To ensure consistency, the plots were set up ac-cording to the arrangement in Figure 3b, which is includ-ed as a guide for any future applications of this method. AModified-V\4rittaker plot can be set up in under 30 minutesusing the low-tech approach.

Results

Ten Modified-Whittaker plots were placed within a 418ha area of grassland (prc. s). No statistically-derived sam-pling scheme was used to determine the placement of theplots on the landscape, but special efficrt was made to se-lect locations that differed in topographg ground surfacevisibility, surface chipped stone density, and degree of dis-turbance from grazing catde. Some of the plots wereplaced where artifacts were known to be abundant, butothers were in locations drat would frequendy escape ar-chaeological analyses due to high vegetation cover or lowartifact density. Thble I shows the variability in chippedstone recovery and provides an initial summary of the rela-tionship between method, spatial scale, and artifact densi-ty. Several ofthe plots yielded artifacts in subsurface testswhen few or none were seen on the surface. None of the

pose artifacts near the surface, so passing the earth througha screen measures artifact density in the near-surface tapho-nomically active zone orTAZ (Lyman L994: 405).

Following Wandsnider and Camilli (L992), surfaceitems were separated according to the manner in whichtheywere discovered. The locations of all items found dw-ing the systematic walking survey were marked with a redpin flag. Each red-flagged item was then recorded by ateam of two to three people seated or kneeling around thepin-flagged artifacts. The locations of artifacts found non-systematically or during the recording process) weremarked with a blue pin flag and also recorded. Due in partto our non-collection recording strategy) a number of at-tributes were documented for each item including length,width, thickness, inclination, orientation, and a variety ofother descriptive codes consistent with the recording strat-egy used in the excavation of the Hudson-Meng BisonBonebed (Todd l9B7). As a result, the ground surfacearound the red-flagged (systematically discovered) artifactswas intensively scrutinized during the recording process,providing opportunities to make blue-flag (nonsystematic)discoveries. We presumed that the intensive double-cover-age of the area provided by the systematic and nonsystem-atic coverage would lead to a nearly complete document ofthe surface record and implemented the crawl survey to testthat assumption. We did not think that the crawl surveywould add a significant number of artifacts. As discussedbelow, we were very wrong.

Sa,ti"g Up the Mod.if,td.-Whittaher Plots

Because of the elaborate internal stmcture of the Modi-fied-Whittaker plot, the set-up process needs to be as effi-cient as possible. Over the course of our study we experi

Wi,ttoberSu@lots l to 10

TilE----dn11rl B*cftrf,ttSwbplotsA and B-

w6lk (,rawl

Sabpht CWelk ( )rawl,

I( pltt

Walh

86727It2l l

I

z

0a

54 1606 80 32 70 00 00 00 00 00 0

NRIPWIRINWIROUTccOlwHOr33N6N6BSWTRIWTROTTOI

MeanI Standard error

0000

00000

0.03 0.50.0 0.2

0.5 2.L0.3 1.3

6.2 r7.85.4 15.8

46I000

00000

LJ+L

66t337

0l3I

8 3 80 l0 00 00 0t ?

0 00 00 00 022

259.0

92.886.r

i1

ar "'\.,,,0-.\ l\ \r. lt / \ 'a.r,

\ . / ) rA \

l:'' \,.::,", 'a;\k.M Iil!)f=/rwJ' \ - .

) ) , \U\x ) [^\ r^-?\/ /,l lV(r'\Vrvv" ) .( a /rl_) ,i\r o 1 /

\ * ' I t l N \ - - ;

t / / \ / ) \ ( \ c c o r , l

.rdso;r- , f'\rzeo-\-'r \ l^

"1, \e n g r \ ( \ / \

st /.,j---\,Jrrol / /N ,,.,_,/ .J /-_,____-/-

-\-__\ |

) \ 0 1 0 0 2 0 0 s (- / )

416 Sarnpli.ng Techniqwes for Archaeological Suwey/Bwrger5 Tod.d' Bwmet4 Stoblgren, and. Stepbens

Figure 5. The locations of dre ten Modified-Whittaker plots used in this study. The contour interval is 5 m.

plots were placed where a previous site had been recorded,but all plots were within the limits of previously surveyedareas. I{ere we focus on the results of the study's property-based investigations and emphasize the plot's applicabilityto archaeology as an experimental sampling design.

Two Exarnphs: PhtsNRTP ond.WRINFor purposes of illustration we have selected two sam-

ple plots for detailed description. Plot NRIP was locatedon a prominent, flat-topped hill where a high density ofsurface artifacts was known to exist. Our initial focus wasto evaluate how much material evades discovery from fine-grained methods in order to gain perspective on how many

artifacts are overlooked by standard transects. Of the tenplots, NRTP had the highest density of surface materialsrecorded. In the 1000 sq m of the I(plot, the walking sur-vey found 366 items systematically (red-flag discoveries)and this number increased to 867 when the blue-flaggeddiscoveries were included. Thus, even at this close spacing,more than half of the artifacts were found while recordingafter the systematic walking phase was completed. Thecrawl survey found 3.5 times more material (244 artifacts)than the walking survey (65 artifacts) in 130 sq m (sub-plots I to I0, A, B, and C).

The substantial increase in items found by the crawlingsurvey is evident in dre distributions ofartifacts recovered

20m

Figure 6. Plot NRIP. An example of how the record changes with observer intensity. The contours arebased on the chipped stone recovered in the subsurface tests of subplots I to l0; the interval is 25 flakes.The walking survey covers the entire 1000 sq m. The crawling survey covers the 130 sq m in the subplots(delineated rectangles).

by the different methods (rrc. o).1 Artifact clusters wereidentified during the crawl survey that were not apparentin the combined systematic and nonsystematic distribu-tions from the walking survey (rrc. Oa-n). The nonsys-tematic coverage, while adding many additional finds,tended to highlight existing clusters rather than identifiznew ones. This is especially evident when the densities insubplot C are considered. The crawl survey transformed arelatively diffuse scatter into a dense cluster (Frc. 6c).

NRTP also had a dense subsurface artifact population.

l. The arrangement of the subplots is slighdy different for NRTP thanfor the other plots in the study. For NRIP, all l0 of the numbered I sqm subplots are on the perimeter ofthe K plot but for all others subplots7-10 are placed on the edge ofthe C subplot as pictured in Figure 3A.This is an adjustment made to the Modified-Srhittaker plot by Stohlgren,Bull, and Otsuki (f 998) in order to further reduce the spatial autocorre-lation among subplots. We shifted to the adjusted version in the summerof2000. The plot depicted in Figure 3,\ could be referred to as ttre "re-

vised Modified-S/trittaket'' but such terminology seems excesstve.

Jowrual of Field. Archneohgy/Vol. 29, 2002-2004 4L7

The contour lines in Figure 6 represent subsurface chippedstone density. In the I0 test units excavated at NRIP a to-taL of 2342 pieces of chipped stone were recovered. Thislocation can be thought of as an obtrusive high-density ar-tifact cluster that would likely have been located andrecorded as a "site" by conventional methods.

The location for plot \ IRIN was determined by theneeds of an interdisciplinary study of damage by grazing,and we did not intentionally seek a location with culturalmaterials. WRIN was placed inside a fenced area con-

structed to keep catde away from the Hudson-Meng BisonBonebed and visitor center. Although an archaeologicalsurvey conducted prior to the construction of the fencerecorded no artifacts, \ IRIN produced the second highestartifact density of the I0 plots. The walking survey at 70cm spacing discovered 27 cl'ttpped stone artifacts (in 1000sg m); the crawling survey found I0 (in I30 sq m), and thesubsurface test found 66 (in 10 sq m).

1 0

€ p o

4r8 S^*plirg Techniqwes for Arcbaeolngicnl Swwey/Bwrger, Tod.d' Bw"nett, Stohlgren, and. StEhens

Figxe 7. The chipped stone distributions fromnvo walking surveys on plot \4|RIN drat tookplace one year apart. The contours representsubsurface chipped stone density; the interval isone chipped stone artifact.

The analysis of the data from [lRlN demonstrates that

samples of the surface record in this area can change in veryshort time periods. Because of the grazing study, WRINwas resurveyed after one calendar year had passed (alongwith other experimental plots) in order to compare short-term changes in range condition and other factors. Thechipped stone was also re-recorded. The results of dresetwo surveys were plotted, one on top of the otheq in orderto compare two documents of the same area gathered atequal intensities (r'rc. 7). The chipped stone observed dur-ing the respective surveys produced two entirely differentarchaeological documents, although only one year hadpassed and the plot lies in an ungrazed setting. At dris

point, the discrepancies between the two samples can mostlikely be attributed to a combination of factors. The arti-facts that individual surveyors are like$ to see at a giventime will vary with changes in sunlight intensity and theirexperience level. It is also lil<ely that the visible surfacerecord had altered its composition due to variables such assmall-scale vegetation changes and sediment cover.

To document the effects of grazing on the movementand visibility of archaeological materials we seeded andprovenienced individually numbered aluminum clasts onthe surface of\4{RIN (and other plots) to track their move-ment over time. The clasts were flat rectangles and trape-zoids or cylinders between 16 and 45 mm in length. Theclasts had moved over the course of a year. The largestmovement on SIRIN was 9 cm, but almost all moved 3 cmor less. Of the 20 aluminum clasts on the surface of\4lRIN,6 were not relocated by the second survey. This suggeststhat the differences in results between the nvo surveys weredue to changes in the record itself, and cannot be explainedby crew experience alone, because the aluminum clasts arelarge, shiny, and very conspicuous against the brown sub-strate. It is clear that the taphonomic agents that rearrange,obscure, and expose surface artifacts are active over shortperiods of time and in settings that appear to be relativelyundisturbed. Understanding variability in these propertiesof the archaeological record is difiicult, but these datademonstrate how property-based investigations can beused to evaluate the relative stability ofthe surface artifactpopulation.

Wnlhing vs. Crowli,ngAchieving a reliable measure of survey accuracy required

a sample that was nearly equivalent to the actual surfacerecord. While artifact size and limits of visual acuity requirethat items will always be missed, those documented on thecrawl survey appear to approximate closely the observablesurface record. No nonsystematic discoveries were made

while recording the items found by the systematic crawl,implyng that nearly all of the artifacts on the surface werefound in the first survey pass. The results of the crawlingsurvey are considered a close estimate of the total numberof artifacts on the surface at the time of the survey and canbe used to evaluate the accurary ofcoarser-grained cover-

age ofthe sarne area.In comparing the walking to the crawling surveys (Thble

2), only the 130 sq m area that both methods cover was in-

cluded (subplots I to I0, A, B, and C). The systematic and

nonsystematic discoveries are combined for each subplot.

Comparison of the techniques indicates that the crawling

survey increased the total number of chipped stone artifacts

discovered per square meter by 3620/o (rmnn z). The spe-

Table2. Comparison of total and average chipped stone (CS) discovered bywalking and crawling surveys, with same area covered (130 sq m). Plotlocations are shown in Fisure 5.

cific value of 3620/o is not meant to be generalized to oth-er contexts, but it is clear that pedestrian surveys of anytransect width are missing many artifacts in most settings.The magnitude of this increase was greater than many ar-chaeologists (ourselves included) might have assumed andthe implications need to be considered.

The size of artifacts alone does not explain the increasein items found crawling. For the walking survey, the aver-age maximum length of artifacts discovered during the sys-tematic pass is L7.3 mm, but the nonsystematic finds wereonly L2.7 mmlong on average. Thus, the initial survey passby the walking crew favored larger items. The artifactsfound by the crawling survey averaged I3.8 mm long,which is not significandy smaller than the combined aver-age of 14.6 mm for all discoveries made by the walking sur-vey. Although non-systematic coverage recovered small ar-tifacts, the crawl survey led to a more accurate portrayal ofthe surface record.

Linhing N exed, S nrnplesRegression analysis of each survey demonstrates that the

ability to predict the properties of the sub-surface archaeo-logical record is much greater for the crawling survey thanthe walking survey. The crawling survey could explainT2o/oof the variance in the top I0 cm whereas the walking sur-vey explained only 24o/o.2 Interestingly, no combination ofground surface visibility, vegetation height, or plant bio-mass improved the amount of explained variability. V\4renarchaeologists are deciding where to place test excavationunits, often the only clues available are depositional setting

2. If the variables are log-transformed, the crawling survey explains54% of the variance in the near-surface record and the walking survey ex-olains 187o.

Jowrwnl of Field. Archneology/Vol. 29, 2002-2004 4L9

and surface artifact density. This provides a starting pointfor understanding the degree to which the surface record"predicts" the subsurface record in different settings. Thesurface and near-surface records are not separate entitiesand differences between them, in some instances, may bemore a function of sample accuracy than the nature of ar-chaeological deposits.

While this suggests that high resolution samples canpredict some of the variation in the near-surface record, thearchaeological reality is that in situations such as pipelinesurveys, archaeologists must often attempt to predict thenature ofdeposits far below the surface based on surfaceproperties. Such situations are difficult because ofthe ex-pense of conducting deep test excavations combined withthe problematic reliance on visible distributions as indica-tors of materials far below the surface. We do not suggestthat crawling surveys have the ability to predict artifactdensities far below the surface, and ideally evaluations ofburied deposits must be augmented with exposed verticalfaces or test excavations and trenches. Furthermore, theserelationships will likely change significandy with largersample sizes and especially in different depositional con-texts and high surface density settings. The large number ofzeroes in the dataset indicates that NRIP, the high surfacedensity plot, had an especially strong influence on the out-comes of the regressions. In spite of this, expanding ourunderstanding of the factors that condition variance in sur-face and subsurface samples will be usefirl for property-based evaluations and for comparing samples across con-texts.

Discussion

Currendy, there is litde consensus on the ideal transectspacing for pedestrian surveys. In many instances this may

Walkini suwty Crawknl suwe,y WaJh ps. rawlModifud-Wbittakrplnt TitalCS Ay/ sq m TotalCS Av/ sq rn Netincrease%in

NRTP 65 0.50 244 1.88 L79 375\4RIN 6 0.05 l0 0.08 4 L67\ 4 ' R O U T 0 0 4 0 . 0 3 4 xc c O t 2 0 . 0 2 7 0 . 0 5 5 3 5 0w H 0 r 0 0 0 0 . 0 0 0 -

3 3 N 6 1 0 . 0 1 3 0 . 0 2 2 3 0 0N 6 B S 0 0 0 0 . 0 0 0 -W T R I 0 0 0 0 . 0 0 0 -W T R O 0 0 0 0 . 0 0 0 -T T 0 I 0 0 0 0 . 0 0 0 -

Total, mean 74 0.06 268 0.2I I94 362x Note that four artifacts were found crawling and zero were found walking. This is asignificant increase but a percentage cannot be calculated.

42O Sr*plirg Techniqwes for Arcbaeological Swwey/Bwrger, Tod.d' Burnett, Stoblgren, nnd. Stephens

be a project-specific decision dependent on variables suchas funding, time constraints, and the goals of dre project.When different methods are used in different projects,however, dre results are difficult to compare. For perspec-tive, Schiffer and Wells (1982: 353) present a table con-taining crew spacing and other details of 12 surveys thattook place in the Southwest United States between 1974and 1980. During this time, transect widths varied from 4to 50 m but averaged ca. 25 m. Current archaeologicalpractice still exhibits this range of variability. Surveys thatclaim complete coverage of a region generally use widetransects, but are often geared towards documenting high-ly visible features (compared to chipped stone) such asburied structures, temples, or cities. \44rile all major setde-ments may indeed be found by such surveys, the notion of100% coverage requires that a substantial portion of thesurface record be written off as insignificant. To return tothe screen size comparison, claims of 100% coverage areakin to using a very large screen size in an excavation andasserting that everything of importance is larger than thegaps in the mesh.

Compared to the norms of current practice, our 70 cmtransect width is absurdly narrow and should yield highlyaccurate results in spite of the limited spatial coverage. Theresults of the walking survey with 70 cm spacing, howev-er, explained only a fraction of the variance in the subsur-face records and a tremendous amount of surface chippedstone was overlooked. Wider transect spacing progressive-ly lessens the ability to address basic parameters of the tar-get population. This suggests that an optimal strategy mayinvolve multiple phases of survey one of which should behighly consistent and precise, while another would focuson large coarse-grained samples. The need to protect sitesfrom land disturbance and the fact that researchers willnever know where all cultural propefties are located re-quires the use of conventional pedestrian transects. In thecase of land management districts which must regularlycontract for surveys, implementing just a few Modified-Shittaker plots in each survey would over the course oftime lead to a far better understanding of the regionalrecord. The methods oudined in this article would be es-pecially usefirl if applied and experimented with in a widerrange ofcontexts.

Area and.Intensity

Two primary causes of systematic surveys failing to dis-cover cultural materials can be identified. One is the sacri-fice of area: the smaller the spatial extent of the regionalsample, the more material within that region will escapediscovery. The second is the sacrifice of intensity: the low-er the observational intensity, or grain, used within the sur-

veyed area, the less material within that area will be ob-served. Sampling schemes employed in survey projectsshould consider both of these sacrifices. This is a difficultproblem because both sacrifices appear to be inevitable,while neither is acceptable. S/hich is the"better" way to notfind artifacts: to not look in enough places, or to not lookclosely enoughf S4rile this is a trade-off (Schiffer and WellsL9B2: 347), nested-intensity surveys and multi-scale sur-vey designs introduce ways to make quantitative sense outof these two respective sacrifices.

In considering the sacrifice of area, the problems are theplacement of the plots on the landscape and the spatial ex-tent of the sample units. Some of our plots demonstratethat if only small areas are crawled, the area sacrificed couldlead to a complete absence of discovery. The increase initems found by the crawling survey demonstrates that thesacrifice of intensity can be significant. When walking sur-veys fail to find materials, this may be because nothing waspresent or because nothing was identified. An area crawledwith nothing discovered is not a wasted effiort, because theconfidence of knowing that no artifacts were visible at thetime of the survey is in itself valuable. Without conductinga limited arnount of survey at this intensity, the simplequestion of "how much material is present within the sur-veyed areaf" cannot be addressed. This point can also bemade in statistical terms. A sample is representative if itsvalues approximate those of the population. In the case ofstandard archaeological surface samples, basic propertiessuch as mean and median artifact density cannot be calcu-lated reliably (Van de Velde 2001: 25).

Archaeological surveys frequently claim to haveachieved complete coverage of a region. In this sense, theintent is to demonstrate that the entire region has been ob-served so that the researcher does not have to infer region-al properries based on limited samples. Coarse-grainedcoverage of large regions is necessarywhen the goal is iden-tifying relationships benveen setdements is, and such sur-veys are often focused on the discovery of higtrly visiblefeatures. The claim that surveys result in I007o coverage isoverly optimistic, however. Interpretations and manage-ment decisions based on pedestrian samples should be tem-pered by the properties of survey data. In any setting, ar-chaeologists need to be more explicidy aware of the infer-ences that can be reliably made from a sample given the na-ture ofthe coverage. Again, dre techniques presented heremay be usefirl for reaching more accurate evaluations of thenature of archaeological survey data drat can be used tocomplement coarse-grained coverage.

Other approaches to surface survey have emphasizedcollection strategies, such as Van de Velde's (200I) point-sampling approach that collects artifacts in 2 m diameter

circles on a systematic grid. Artifact collection is not ap-propriate in all circumstances, but Van de Velde applies sys-tematic collection to the problem of obtaining accuratesamples. I^4ren archaeologists must conduct research aswell as preserve and manage the record they study, it makeslitde sense to destroy the record whiLe also managing it.Obviously, some artifacts require analysis that cannot beconducted in the field and must be collected. In other cas-es, the record is in dar.rger of destruction and collectionmay be necessary Van de Velde's systemic collection strat-egy is a novel approach, but isolated 2 m circles will not re-veal the distributional properties of a regional record norcan they be used to evaluate methodological accuracy orstability in surface distributions. The Modified-\A4rittakerploCs advantages over such approaches include the ex-panded spatial sample, improving the probability ofrecording isolated and rare finds, and the ability to investi-gate the record at mt tiple scales (Stolrlgren et al. 1997).

Tbe General Uti.liry of Mwlti-Scale Sam.plingAlthough the ONG Survey ?roject area is a hunter-

gatherer record composed almost entirely ofchipped stoneanifacts, the utility of the techniques presented here is ap-plicable in a variety of archaeological contexts. A-rchaeolo-gists snrdying complex societies often deal with extensivescatters of ceramic sherds. Attempts at making inferencesfrom drese surface scafters and their associated features de-veloped into the conceptual cornerstone for vimrally allcontemporary survey projects (e .g., Willey 1953). Follow-ing dris tradition, setdement pattern analysis led to a vari-ety of highJevel interpretations regarding prehistoric pop-ulation densities, movementsr and cultural changes thatwere large\ based on the analysis of surface scafters (e.g.lSanders, ?arsons, and Sandey 1979). These estimates arebased on calculating the number of different ceramic g,pes,sherd density, and site size. Multi-scale techliques could bevaluable in such situations because thev allow for freouen-cies ofdifferent ceramic rypes to be invesdgated as spatialscale increases. Application of the Modfied-\44rittaker plotcould enhance and refine the data gathered during the pioneering surwey projects conducted in these regions andimprove comparisons between sites of different qpes orbeween regions.

It might be argued that this method is usefi.rl for low ar-tifact densities in arid landscapes but could not be used inregions ofeither extremely high artifact loads or high veg-eation cover. Cenainly, with the restraints of time andmoney, piece-plotting each individual artifact may not beviable in all situations. The advantages of multi-scale sam-pling are not dependent on accurate artifactJevel prove-nience, however. Archaeologists frequendy work il areas

Jowmal of Field Arcbaeohgy/Tol. 29, 2002 2004 421

where the ground surface is blanketed in anifacts. Howshould such populations be sampledf Very dense artifactscatters cafl also be sampled with intensive strategies andmay especially benefit from multi-scale techniques. Themulti-scale framework of the Modified-I4/hittaker plot canbe used for projecting the propenies of the sarnple to thecomplete population.

To apply the Modrfied-\44rittaker plot to the samplingof dense surface records, the abundance and diversity ofeach artifact qpe would fust be tallied for each of the sub-plots and the K plot. These data can then be plotted withthe spatial scales of I, 10, 100, and 1000 sq m on the x-ax-is with either the number ofobserved artifacts or tie num-ber of anifact qpes on the y-aris. By fitting a ftmction tothese points,3 the area-density relationship can be project-ed to make quantitative estimates for how the sampleshould behave at scales larger than can be feasibly observed.This function could be used to estimate the number of in-dividuals for a cenain qpe or to project the diversity oftheassemblage in a manner analogous to the plant species-areacuwe (Begon, Harper, and Townsend 1996: 86I-871;Mueller-Dombois and Ellenberg 1974: 5l). As area in-creases, so will the diversity of the sample. This is a func-tion of the larger sample size and the higher probability offinding rare specimens in a larger area, but the exact natureofthe relationship will depend on the heterogeneity of thepopulation (Ebert 1992: 213-244). Artifact diversityshould also increase with area and anallzing distributionsin an area-diversity perspective may be a usefirl avenue forcomparing qpologica-lly diverse archaeological landscapes.

Conclusion

In this paper we have attempted to show how propeny-based investigations of the archaeological record can bemore filitfully implemented by modiSring a few aspects oftraditiona.l sunzey practice. By combining this approachwith existing norms, the context of the items discovered isenhanced by incorporating a more accurate understandingof the nature of the population that is being sampled. Thisis usefirl for pragmatic concerns such as the effect of tran-sect spacing on anifact density, and can also be used formonitorhg change over time on an archaeological land-scape. Understanding how processes such as large herbi-vore grazing influence the distribution of archaeologicalmaterials is an imponant management concern (e .g., Minard 2003) and is also necessary for archaeological inter-

3. Most computer programs that can form scetter plots such as Excelcan fit finctions to the data distribution and provide the resulting equa-tion ofthe firnction. A power function is the most appropriate for appli-catioff like d1is.

422 Se?npling Trcbniqua for Arcbaeol.ogical Suwry/Bwryer5 Tod.d., Bruv'ex, St1blgt'sn, end StePhens

pretations of spatial relationships among anifacts and fea-tures, For instance, we found that single seasons ofgrazingcan seriously disturb artifact distributions and the ap-proach we used to monitor t}tem can facilitate the devel-opment of a framework for estimating long-term effects.We also found that using high-resolution coverage can leadto imponant revelations, such as the magnitude of adfactloss given certain techniques and the degree to which thesurface record can change over short time scales. Theseprocesses need to be investigated with an experimental ap-proach and the Modified-\44uttaker plot provides an idealframework fcr doing so.

A number of conclusions could be made. Perhaps ar-chaeologists should resign themselves to the fact that sur-face samples gleaned by survey will not match the proper-ties of the surface record, and we should proceed as nor-rnal, simply regarding this as a cautionary ta.le . On the oth-er hand, many archaeologists would argue for the need tocontinually seek new ways of improving manage ment andresearch techniques in the attempt to extract behavioral in-formation frorn surface rnaterials. We prefer the latter op-tion because a failure to understand the properties of thedistributions captured by a surwey will lead to a failure tounderstand the processes behind them.

Acknowledgments

The participants in the Colorado State University fieldschool frorn 1999 to 2001 were entiusiastic, dedicated,and most of al.l, great to work with. We are indebted tothem for making this research possible. We are also thank-fi.rl for volunteers from previous field schools who returnedto take part in the research. LuAnn Wandsnider providedusefirl commentary and an excellent foundation for our ex-periments with method and discovery |udson Finley andDave Rapson made valuable contributions to early draftsofthis paper. Three anonymous reviewers offered especial-ly valuable Lnsights and editorial suggestions.

Oshar Bwrger is a grad.utLte stad.ent a.t the Univetsity of NewMarin interested in la.ndscape nrcbaeokgl nnd. und.erctand.-ing tbe ffia of hwruan bebevim on the e?ulironznent. Thisrnanuscl'ipt arose jlorn bis Ma.stefs resenrch, clmpleted. in thesprin! 0f 2002 a.t C1loradt State Unfuersity. Mailing Ad-d.ress: DEa.rt nent of An hropzlagy, Unfuer:ity of Nat Menico,Albaqwerqae, NM 87 I 3 I - 1086. E-wail: oshar@unw.ed.w

Lnwrence C. Todd. orgcotized. tbe ONG rne*rcb projea andit interested in ta,phonontic research across spa,tit l sca'les tt7td,the d.wel.opment of integrttted mubi-d.istiplinnry researcb andm a,n a,g ew ent p € rsp e c, n e s.

Pa.ul Burnett i"s a. nrtdua.te stwdent at Colnradn Stnte Uni-

wrsiry and is d.nekping applications ofthese techniques for re-searcb in bigb-ahitwd.e sexings.

Tomas J. Stoblgren is a research sci.entist fm the U. S. Geo-hgicd Suwey in Fofi Coll,ins, Cokradn. He xwdies pla'nt bio-dfuersiry and. satnpkng d.etigns in North Awericnn emslstems.

Dowg Stephens is a Foreo Semice arcbaeohgist and. overcawthe researcb on tbe Oglala Natinnal Grosslandl He is inter-ested. in intetratin! feld techniqua tbnt ,neet bztb ,na'nage-ment amd. research requiremmtl

Ammerman, Alben J.198I "Surveys and Archaeological Researchl' Anwul Rwiew of

A4thropokg! \0: 63-88.

Banning, X. B.2002 Archazohgical Szn T New York: Kluwer Academic.

Begon, Michael, |ohn L. Haryef and Colin R. Townsend1996 Ecology: Intl.fu luak, Po?uk tions, and. Corntnunities, 3rd edn.

Oxford: Blackwell Science.

Clarke, David L.1977 "Sp^tt i Information in Archaeolog,v," in David L. Clark,

ed., S?Mial Artkraolagy. I-ondon: Academic Press, I-32.

Cowgill, George L.1990 "Toward Refining Concepts of Full-Coverage Suveyi in

Susan K. Fish and Steve A. Kowalewski, eds., Tbe Archaeokg! of Regions. Washrngton D.C.: Smithsonian ImtitutionPress,249-259.

Dumrell, Robert C., and WilJiarn S. Dancey1983 "The Siteless Survey: A Regional Scale Data Collection

Strategy;' in M. B. Schiffer, ed., Advances in ArchaakgialMahod and. Tbmry YoL 5. New York: Academic Press,267-287.

Ebet, lames L\992 Distribatiowl Archneokgy Albtqretque: University of

New Mexico Press.

Eben, James I., and Timodry A. Kohler1988 'The Theoretical Basis ofArchaeological Predictive Mod-

eling and a Consideration ofAppropriate Data-CollectionMethodsj' in W J. Judge and L. Sebastian, eds., Q"atfi-in! the Present d.nd. Pled.icrin4 tbe Pqit: Theory, Metbod, antlAlplica.tion of Arcbaeological Pred th,e ModeLing. Der.er:U.S. Bureau ofl-and Management, 9l-172.

Folel Roben A.l98t "Off-Site Archaeology: An Alternative Approach for the

Short-Sitedj' in Ian Hodder, Glen Isaac, and N. Ham-JJi'ornd, eds., Pasterns of tbe Pa.st: Esta.ls in Honor of Dadd L.Clarhe. Cambridge: Cambridge University Press, 157 I83.

Given, Michael, A. Bernard Knapp, Natl-ran Meyer, Timodry E. Gre-gory VasiJiki IGssianidou, iay Noller, Lisa Wells, Neil Urwin, andHaddon Wright

1999 "The Sydney Cypnrs Survey Project: An InterdiscipliiaryInvestigation of l,ongrfbrm Change in dre North CenualTioodos, C1prus,"/ozmal ofFieLl Arcboeokgy 26: 19-39.

Lyman, R l-ee1994 Uel'tebrne Tapbonowy Cambridge: Cambridge Universiry

Press.

Minard, Arne2003 "Unwelcome Bison May Face a DNA Test," Srit t 299:

1835-1836.

Joumal of Ficld. Archaeohgypol. 29, 2002-204 423

Mueller-Dombois, Dieter, and Heinz Ellenberg- 1974 AiM and Msthodr of Uegotdriitu Ecok4!. New York: Iohn

Stlley arrd Sons.Onon, Clive

2OOO Sernpling in Archazokgy. Canbidge: Cambridge Universi-' ty Press.

Sanders, \44lliam T., I. R. Parsons, and R S. Sandey1979 The Ba.titt. of Matia: The Ealagical Procexes in the h,olnti.on

of a Cirilizatioo, New York: Academic Press.

Schiffer, Michael B., and Susan ]. Wells1982 'Archaeological Surveys, Past and Futurej' in R. H.

McGuirc and M. B. Schiffel eds., Hohohan and P*ayn:Pr?hittory 0f Soatbwestent Arizdfi.a. Nevt York: Academic

Schitrer, Michael B., A. P Sullivan, ard T C. Ktinger1978 -Ihe Design ofArchaeological Surveys i WorM Archazokgy

I0: I-28.Shon. Michael I

1995 'Reliability ofArchaeological Records on Cultivated Sur-faces: A Michigan Case Sndyi Joamal of Field Archazokgy22: 475490.

StohlgrerL Tomas J., K, A. BuI, and Y. Otsuki1998 "Comparison of Rarrgelartd Vegetation Sampling Tech-

niques in the Crntra.l Grxslatdsi Joanal of B.ange Man-dgemcrrt 5I: 16+172,

Stol gren, Tomas J., M. B. Falkner, and L, D. Schell1995 "A Modified-t/trittaker Nested Vegetation Sampling

Metrcdj' Vegetati.o I17: I13-121.Stohlgren, Tomas J., G. W Chong, M. A. Kalkhan, and L. D. Schell

1997 'Mu.lti-scale Sampling of Plant Diversity: Effects of Mini-mum Mapping Unit Size;' Etokticd.l Apphfuti| Tl1064=1074.

Thomas, David H.1975'Nonsite Sampling inArchaeology: Up the CreekWidrout

a Sitel" in I. \ I Mueller, ed., ScnEling in Arcbaeology.'luc-son: University ofArizona Press, 61 81.

Todd, lawrence C.1987 'Thphonomy of the Homer II Bone Bedl' in George C.

Frison and Lawrcnce C. Todd, eds., Tbe IIoner Site, TheType Sin ofthe Ady Cahural Cntxphx. Otlatdo AssociatedPress, 107-198.

Van de Velde, Pieter2001 "An Extensive Altenative to Intensive Suwey: Point Sam-

pling in the Riu Mannu Svrey Ptqect, Sardtnal' Joumal0f Mcd.iterad.ned.n Arcbdzohgy 14: 25 52.

Wandsnider, LuAnn, and Eileen Camilli1992 'The Character of Surface Archaeologica.l Deposits and Its

Iniuence on Survey Accutacyj' Joamal of Field Archazohgry19:169-188.

. \44lley, cordon R.1953 Prehistoric Senlewert Pqfiemt in the Vinl Vallzy, Peni. Bw

ra* ofArncrican. Ethnakgy Bullain No, I55. Washington

. D.C.: Smithsonian Institutron.