You are on page 1of 27

Society for American Archaeology

Communities, Settlements, Sites, and Surveys: Regional-Scale Analysis of Prehistoric Human Interaction Author(s): Christian E. Peterson and Robert D. Drennan Source: American Antiquity, Vol. 70, No. 1 (Jan., 2005), pp. 5-30 Published by: Society for American Archaeology Stable URL: http://www.jstor.org/stable/40035266 . Accessed: 06/04/2011 11:21
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=sam. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to American Antiquity.

http://www.jstor.org

^^KTMIIMWHS

COMMUNITIES, SETTLEMENTS, SITES, AND SURVEYS: REGIONALSCALE ANALYSIS OF PREHISTORIC HUMAN INTERACTION
ChristianE. Petersonand RobertD. Drennan

The study of developing complex societies can fruitfullyfocus on the human interactions that define communities,which have always been at the heart of settlementpattern research. Yetlittle attention has been paid to how communitiesof varying scales can actually be identified in archaeological survey data. Most often sites have simply been assumed to correspond to communities,although this practice has been criticized. Methods are offeredto delineate communitiesat different scales systematically in survey data, and their implicationsfor field data collection strategies are explored comparatively for cases from northeast China, Mesoamerica, and the northernAndes. Las interaccioneshumanasque definenlas comunidadessiemprehan sidofundamentals para las investigacionesdepatrones de asentamiento, conformanun enfoqueproductivopara el estudiodel desarrollode las sociedades complejas.Sin embargo, y los metodospara identificarlas comunidadesde variados tamahos con datos de reconocimientoarqueologico no han sido se muydiscutidos.Generalmente asume que un sitio arqueologicocorrespondea una comunidadhumana,aunqueesta prdcde tica ha sido criticada. Este articulo ofrece unaperspectivabasada en la representacion distribucionesde poblacion como una superficiecuya altura varia segun la densidadde ocupacion.Asumiendoque la interaccionen general disminuyecon distancia, las comunidadesde interaccionpueden ser delineadas sistemdticamentecomo las bases de picos aislados en esta topografiaartificial.Se exploranlos requisitosque tiene estaforma de andlisis en cuanto a la recoleccion de informacionen el campo. Ejemplosdel andlisis de patronesde distributionde ocupacion en China (region Chifeng),Mesoamerica(Vallede Oaxaca), y los Andes colombianos(Alto Magdalena) demuestranque el metodo sugerido ayuda a delinear comunidadesde una variedadde tamahosy que tiene la capacidad de no delinear comunidadescuando esta estructuraespecifica de interaccion no esta presente.

in areconstituted thepatterned whichare interactions betweenhouseholds, central Communities to everydaylife in many societies in all partsof the world.It is in thismatrixof interaction thatthe forces thatproducesocial change are generated,and the qualitativesocial changes can often studiedby archaeologists be viewed as interacthe emergenceof new ways of structuring tion in communities.This has long been recognized in archaeological analysis and recently renewed attentionto communitiesdemonstrates the continuedrelevanceof this perspective(e.g., CanutoandYaeger2000; Flannery1976;Hegmon 2002; Kolb and Snead 1997; Kowalewski2003; and Kuijt2000; RogersandSmith 1995;Schwartz Wilk andAshmore Falconer1994; Trigger1968; 1988). has Theword"community" beentakento mean

a numberof differentthings, rangingfrom Murdock's (1949:79-90) classic definitionbased on co-residence and face-to-face interaction, to or "ideational," "imagined," "emic"communities (Hegmon2002; Knapp2003; Isbell 2000; Marcus 2000; Preucel 2000; Yaeger and Canuto 2000). Whateveraspect of communityis focused on, its study in prehistoryrequiresa means to identify communitiesin the archaeological record.This is especially challengingfor the kind of emic comfromresidencediscussedby Marmunitydivorced cus (2000). Our focus here is on behavior, not to beliefs; what we offer is an approach delineatratherthan ing communitiesof social interaction communitiesof the mind.In the absenceof modern technologiesof transportation communiand the costs and inconvenienceof interaction cation, increase substantiallywith distance. Economic

PA Christian E. Peterson and Robert D. Drennan Departmentof Anthropology,Universityof Pittsburgh, Pittsburgh, 15260 (cepst22@pitt.edu, drennan@pitt.edu) AmericanAntiquity,70(1), 2005, pp. 5-30 Copyright2005 by the Society for AmericanArchaeology
5

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

is practicality not the only basis on which households make their decisions, but facilitatingdaily of patterns activitiesis amongtheirchief concerns in decidingwhereto locatetheirresidences.Tothe extent that daily interactionsare important, then, householdscan be expected to locate their residences close to those of other households with whom they frequentlyinteract.Althoughmisgivings aboutsuchprinciplesarecommon,thosewho have expressedsuch misgivings often in fact rely heavily on distance-interaction assumptions (e.g., Yaeger 2000:126, 130-131). It is reasonableto contexts,patexpectthat,especiallyin premodern ternsof interaction, thus social communities, and will be broadlyreflectedin patterns spatialdisof tribution residence. of Murdock'sfocus on daily face-to-faceinteraction providesa logical point of departure the for identification social interaction of archaeological the communities, although existenceof suchsmall local communities cannot simply be assumed. Rather,their presence must always be demonstrated.When they are present,they are entities withinwhichvariations thenature households in of andin householdactivitiesandinteractions be can and much that has been labeled investigated, "household consistsof preciselysuch archaeology" At the same time, small-scalecominvestigation. munitiesbecome the units of analysis at a larger scale, where study can focus on variationsin the natureof communitiesandthe patterns interacof tionsbetweenthem.Thesepatterns maypermitthe identification yet largersocial communities of entities to which we are accustomedto applying terms such as "district," and "polity," others,but which exist in fact, like small local communities, in thepatterns interaction of betweensmallerunits. Inconsidering districts polities,we approach or the scale of regionalsettlementstudy. spatial The notion of community,then, as we use it boundto a particular scale here,is notstrictly spatial betweenthatof the householdand the region. Its is essence, rather, in patternsof intensityof interactionacross space. These patternsof interaction come into focus at differentscales to reveal specific structures exist simultaneously a given that in Thisis not to say,however,thatthereis any region. standard of scalesatwhichsuchstructures set must exist everywhere. the very processof disIndeed, coveringthe scales at whichcommunitystructures

of interactionform in differenttimes and places makes a major contributionto the comparative Our studyof complexsocietiesandtheirformation. small local comanalysisbegins by investigating munitiesof people in virtuallydaily face-to-face interaction;it proceeds to progressively larger scales. Althoughit has not been framedin quite this way, such familiaranalyticaltools in regionalsettlementanalysisas rank-sizegraphsand site-size are histograms reallyaimedat studyingthe variety of local communitiesandthe natureof theirinteractions.Theseandothersuchanalysesmakesense only if the units of analysiscan be meaningfully A thoughtof as humansocial communities. onefor-one correspondencebetween archaeological sitesandsmalllocalcommunities oftenassumed, is and archaeologicalsites have often been used as the basic unitsfor such analyses.Criticshaveconthe vincingly undermined automaticassumption of site-community uponwhichthis correspondence is based(e.g., Dunnell 1992;Dunnelland practice Dancey 1983;Ebert1992). Since such corresponof dencemayormaynotexist,thedelineation local interaction communitiesis an analyticalquestion in its own right, answerableby focusing on how themselvesacrossthe landscape, people distribute in as bestreflected archaeologically howtheirmaterial remainsaredistributed acrossthe landscape. This reflectionis, of course, distortedto varying degreesby a numberof factors.Humanactivities in laterperiodscan alterthe materialrecord left from earliertimes. Changesoccurringduring that a singleperiodcanproduce palimpsests aredifand Natural ficultto disentangle interpret. processes transportmaterialtraces of human activity and depositsedimentsthathide suchtraces.These and otherprocessesobscureandconfuse the regionalscale patterning need to recognize and create we on their own. The more ephemeralthe patterns remains,the morelikely theyareto archaeological be severely affected by such factors. Efforts to understandand cope with these problems (e.g., Ebert 1992:27-34; Holdawayand Fanning2004; papers in Rossignol and Wandsnider1992) are worthwhile,but arenot the subjectof this paper. In a preindustrial society,one aspectof agrarian be can economicpracticality consistently expected to spread householdsbroadlyacrossthelandscape: thelabordemandsof cultivation. Considering only

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

this factor, households might reasonably be expectedto place theirresidencesdirectlyon the land that they farm so as to minimize the effort involved traveling in fromtheirhomesto theirfields (Chisholm 1970; Drennan 1988; Stone 1993). Pullingin the oppositedirectionarethe economic of with practicalities interactions otherhouseholds, which arefacilitatedif the interacting households are located in close proximityto each other.This is trueregardless theprecisenature theseinterof of actions. They may have an economic character, such as specializedproductionand exchange, or of laborif thisis notorgacoordination agricultural nized entirely at the household level. They may in or involveparticipation religiousritual otherpublic ceremonies.They may have political implications, or they may be social in the most direct the sense- forexample,maintaining bondsof kinWe ship, findingmates, exchanginginformation. havethis broadrangeof activitiesof diversekinds in mind when we refer,loosely, to social interaction.A local community formedwhenthisrange is of social interactions is intensely concentrated within a single well-definedgroupof households thatinteractonly much less intenselywith households outsidethe group.Such a pattern interacof tions wouldencourageall householdsin the group formto locatetheirresidencesin close proximity, fromothersuch clusing a spatialclusterseparate ters acrossa region.These wouldbe recognizable in the archaeological recordas a series of clusters While this is not of material remainsof habitation. the only way interactions between households it might be patterned, is certainlya common pattern. The presence of such a patterncannot be assumed,but it can be sought as a fundamental analyticaltask. Small Local Communities The prehistoric Hongshan period (4500-3000 B.C.) in the Chifengregionof easternInnerMongolia, northChina,providesan example of a disof tribution occupationin which a pattern local of communities to exist, butthe exact way in appears which they should be delineatedis not obvious. Systematic archaeological survey of 765 km2 between 1999 and 2001 has provideda basis for information aboutprehistoric popreconstructing ulations and their distributions(Chifeng 2003a,

Figure 1. Distribution of collection units in the Chifeng survey area which contained Hongshan ceramics. The inset shows individual collection units.

2003b; Chinese-American2002; Linduff et al. 2004). Both relativeand absolutepopulationestiand matesfor this surveyareahavebeenproduced, the methodology utilized has been fully documented (Drennanet al. 2003a). These estimates, calculatedfrom areas of surface artifactscatters and surfaceartifactdensities,are the basis for all in statements below aboutpopulations the Chifeng region.The Chifengsurveyrecoveredsurfaceartifactsby collectionunitsconsistingof areasof 1 ha to orless definedin thefieldaccording fieldboundaries, streams,roads,or otherconvenientfeatures a of thelandscape. Sometimes collectionunitis spafrom others,but sometimesseveral tially separate are spatiallycontiguous.Such sets of contiguous collectionunitsarewhatarchaeologists usuallycall but "sites," the unitof datacollectionandanalysis in Chifeng is not the site but the collection unit et (Drennan al. 2003b). Figure 1 shows the distributionof collection wererecovunitswhereHongshan periodartifacts ered in the Chifeng surveyarea.Figure2a shows how contiguouscollection units combineto form "sites" as usually defined. By this traditional scheme,one would identify34 sites in this partof the survey area, and sites so defined have often been implicitly assumedto representmeaningful humancommunities. Among these 34 sites, however,some pairsareseparated distancesof only by 50 m, which is surelynot farenoughto createany on impedimentto face-to-faceinteraction a daily basis. Faced with such a situation,some analysts

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 2. Subjectively defined clusters of collection units at increasing scales: (a) groups of contiguous collection units would traditionally be called "sites";(b) clusters of very closely spaced "sites"might better represent local communities; (c) more inclusive clusters of "sites" could also be interpreted as local communities; (d) such clusters also appear to exist at an even larger scale.

havefelt the need to combinesuch closely spaced occupationareas in pursuitof more meaningful unitsof analysis.In Figure2b, we have combined areasseparated only such shortdisoccupation by tancesinto largerunits, in much the way Blanton et al. (1982:40) combined"sites"in the Valley of Oaxaca analysis. for Thesecombinations reducethe numberof analyticalunits to 26. It would be possible to continueto formmoreinclusivegroupings from occupationareas separated only slightly by greaterdistances.The 12 units in Figure 2c are each almost 1 km across,andthe fourin Figure2d are over 2 km from one end to the other.These largest clusters probablyexceed the size within which daily face-to-face interactionis likely to occuramongall households, theclustersin Figbut ure 2c may well representsmall local communities of no morethanroughly40 households(based

to referred on the meansof estimatingpopulations above).LikeBlantonet al. (1982:40),we haveidentified these clusters in purely subjectivefashion, relying primarily on distances of separation between collection units, a process that can be accomplishedefficientlyand more objectivelyby buffersaround collectionunitswith a GIS drawing program. Evenwithanexplicitlydefinedbuffer, however, the choice of buffer distance remains arbitrary. distanceis not the only releMoreover,separation vantvariable.The interaction "pull"of a commuwith a large numberof inhabitantswill be nity greaterthanthat of a smallercommunity(Anaya etal. 2003:181;HodderandOrton1976:188;Olsson 1965;Plog1976).Thisresultsfromthefactthat, other things being equal, more people produce more interpersonal interactions. Whatever the

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

Figure 3. Four topographic surfaces produced from the same set of measurements by different interpolation algorithms: inverse distance squared, kriging, Shepard's method, and minimum curvature (left to right).

natureof such interactions, odds arethatthere the will be more of them across spatial separations This whenthe number peopleinvolvedis larger. of observationdoes not depend on assumptionsof in but functional communities, specialization larger ratheris a simple numericalexpectation.Functional specializationor various kinds of centralthe izationcould,however, strengthen effect.As we observations into to clusterarchaeological attempt human interactionunits, we should, meaningful then,seek analyticaltools thatreflectnotjust sepsizes. Herewe but aration distances alsopopulation units must exist at do not assumethat interaction any particularscale; the presence or absence of such units at differentscales is an essentialquestionto be investigated. Thustheanalytical approach takenmustbe capableof failingto findthemwhere thereis notactuallyanyevidencethattheydo exist. The approach followed here begins by representthe distribution peopleacrossa landscapeas of ing to a surfacewhose elevationis proportional local density. population thatproducesuch graphicrepresenPrograms tationsarewidelyavailable usedin applications and thatrangefrom datadisplayto topographic mapthat ping.Users,however,often seem unaware the complexity of the interpolationand smoothing difmakesit possibleto producerather algorithms ferentsurfaces fromthe samedata(Figure3)- differences that would be worrisome enough in mappingand trulyfrighteningas a basis for data analysis. If the featuresof a surface are used to defineanalyticalunits, as below, we mustbe sure real data structure are not and thatthey represent mathematical artifactsof the interpolation simply algorithm producing the display. Ebert (1992:173-1 85) makesa similarpointaboutthe zbut scalesin suchrepresentations, we thinkhe gives

up too easily in the effortto make systematicuse of them(see below).Intheanalysespresented here, and we use a straightforward tightly constrained approach to smoothing. The approach is best explainedby continuingwith the analysis of the Hongshanexamplefrom above. the Figure4 showsa surfacethatrepresents distribution of Hongshan occupation across the Chifeng survey region that appearsin Figure 1. Similardisplayshavepreviouslybeen used in settlementanalysis,although theydo not all represent the datain the same ways nordo theyeven all representthe same kindsof data.Forthis reason,it is to important be quiteclearabouthow suchsurfaces are derived.The surfacein Figure4 is based on a regularlyspaced grid of z-values at 100-m intervals. These values are surfacedensitiesof Hongrecordedin systematic shanceramics(sherds/m2) The examplein Figure5 shows survey. hypothetical how sherddensitiesin threecollectionunitsbecome z-values at 100-mintervals,effectivelyrasterizing the sherddensity data.The value for each 100-m cell is the sum of the productsof the surfacesherd densitiesby collectionunit andthe corresponding areas of fractionsof collection units in the cell. Cells with no collectionunitsorpartsof collection units (where Hongshan sherds were found) are, assignedvalues of 0.00. naturally, Areasanddensitiesof surfacesherdshaveoften been used as archaeologicalproxy measures of local populationdensities, and the applicationof this principleto the Chifengregionis discussedby Drennanet al. (2003a). Followingthis reasoning, we take surfaceslike thatin Figure4 to represent The series of regional occupationaldistribution. isolatedsharpoccupation peaksrisingfroma complaneshownin bird'spletelyflat(i.e., unoccupied) as view in Figure4, could also be represented eye

10

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 4. Unsmoothed surface representing Hongshan occupation in the Chifeng region. (See Figure 1 for scale.)

a contourmap (Figure6), which providesa basis for systematicallyclusteringcollection units into meaningful groupings. Occupationalpeaks are sometimessingle, contiguousoccupiedareasand collectionunits.An sometimesclustersof separate selected low contourlevel will outappropriately linethebasesof thepeaksandshowclustersof multiple collectionunits.The heavyline in Figure6 is a cutoffcontourthatdefinesclustersrangingup to nearly 1 km across, which would seem near the upper limit for daily face-to-face interaction.A lower cutoff contour would form groupingstoo in largeto interpret these termsas local communi-

ties. The heavy line also satisfyingly groups togetheroccupationsseparatedby distancesthat would seem only trivialimpedimentsto frequent a interaction; highercontourwould fail to do this. distributions density as Visualizing occupational surfaces, then, can facilitate the recognition of that patterns conpeaksthatrevealthe interaction stitutesmall local communities.The correspondthese mapsmakeit possibleto delineate ing contour communities in a systematic way. The analysis remainssubjective,in thatit providesno absolute for or objectivecriterion selectinga cutoffcontour. This selection, however,is not entirelyarbitrary.

Figure 5. Three hypothetical collection units rasterized into 100-m cells. Final values appear in the center of each cell on the right, based on the surface sherd densities and collection unit areas indicated at the left.

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

11

Figure 6. Contour map of the occupational peaks in one section of Figure 4. Collection units with Hongshan ceramics are in gray. The chosen cutoff contour is the heavy line, which indicates clusters of collection units.

Just as differentphotographers might not focus a camera identically, different analysts might not choosepreciselythe samecutoffcontour. Contours too highortoo low to plausiblyidentifylocal communities, however,areas easily andsurelyrejected as arepicturesbadly out of focus. Thereare 125 local communitiesso definedin thesurveyarea,whosetotalpopulation estimated is at 4,000 to 8,000 inhabitants(see discussion of estimationin Drennanet al. 2003a). In population termsof spatialextent,these communitiestend to be a few hundred metersacross;the smallestconsistof singlecollectionunits,whoselongestdimension is less than 100 m, while a few areas muchas 800 m fromone end to the other.The largest,then, the approach maximumfeasible areawhich daily face-to-face interactionmight cover. Most have fewer than 50 inhabitants(Figure 7). We might characterize them as hamlets or small dispersed villages, althoughmany are clearly no more than the isolated farmsteadsof single families. A few standout as exceptionallylarge,with populations of up to perhaps500 people. Patternslike that in

Figure 7 are sometimes taken to indicate settlementhierarchy some measureof regionalcenand tralization. rank-sizegraphfor these 125 local A communities(Figure 8) shows a very significant fromlog-normality, its convexshape but departure widecentralization. indicates oppositeof regionthe The A value (see Drennanand Peterson2004) for this rank-sizepatternis .301 (p < .01). Such convex shapesareoften producedwhen severalindependent systems are combined into a single to analysis.We mightattempt look for evidenceof theseindependent ata spatialscale smaller systems than the whole region, but larger than the local communitygroupingswe havejust defined. Larger-Scale Structure If the surfacerepresented Figure4 were mathein the "smoothed," basesof all occupational matically onesto merge peaksmightexpand,allowingnearby into just such more inclusive groupings. One to approach smoothingwould be to resamplethe cell aboveatincreasing sizes. raster imagediscussed

12

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 7. Histogram of estimated populations for the 125 possible Hongshan local communities in the Chifeng region.

This would producea smoothersurface,but only with considerable loss of detail.A more effective means of smoothing one thatretainsdetail vital to the analysis below- is a weighted averagein which the value for each 100-m cell becomes the average of all the values of all the cells, each to according someinverse powerof its disweighted tance.If the chosenpoweris high,thenthe weights decreaserapidlywith distanceandlittlesmoothing occurs, while a lower power produces stronger smoothing. There is no meaningful difference between surfacesfor powersgreaterthan about4 because,forall practical purposes, powersthishigh no smoothing.At the other end of the represent the scale,a powerof 0 produces ultimate smoothing in smoothing, flatsurface. vara 9 illustrates Figure ious degreesof smoothingof the Hongshanoccupation,with powers rangingfrom 4 (virtuallyno to smoothing) .001 (very strongsmoothing). As we look at the increasinglysmoothedsurfaces, we begin to see the emergence of spatial structure a largerscale. At a power of 2, small at basal flanges begin to appeararoundthe occupational peaks. These broadenat a power of 1 into distinctlyfunnel shapesthatmight help us define the more inclusive clusters referred to above. Choosinga very low contourlevel on this surface again delineatesclustersof occupation,although thepattern morecomplicated is (Figure10).Toward delinthenorthwest the surveyarea,thiscontour of eates clustersof the small communitiesidentified

Figure 8. Rank-size graph for the 125 possible Hongshan local communities in the Chifeng region. A value and 67 percent confidence zone are determined as suggested by Drennan and Peterson (2004).

It in the less-smoothedtopography. seems reasonable to take these as communitiesas well, in the sense as we have used the same social interaction word "community" before, but at a larger scale thanthe small local communitiesalreadydefined. Towardthe southeastone very large clusteris delineated,but a look at the detailsof the power 1 surface 9) (Figure revealsthatthisis nota veryaccuof ratecharacterization this partof the landscape. In fact,occupational peaksvery similarto those of the northwestoccurin the southeastas well. They are, however, close enough together that the has smoothingof the topography resultedin a very fromwhich surface of the"flat" uplift slightgeneral the peaks rise sharplyup. Because the "flat"surface is slightly higher in the southeastern partof the surveyarea,we canchoose anadditional higher contourlevel to define the bases of the six major peaksin this sector(Figure11),again occupational a scatterof small Hongshancommunities leaving not combined into the larger scale clusters.The betweenthesetwo specificdetailsof thetopography contourlevels in the southeastern of the surpart that vey area(Figure9) reveal"valleys" dividethis areabetween the majoroccupationalpeaks. Takwe ing thesevalleysas boundaries, canassignmany of the smaller scattered settlements to clusters

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

13

Figure 9. Surfaces representing Hongshan occupation in the Chifeng region. Smoothing increases from top to bottom, with inverse distance powers of 4, 2, 1, .5, .25, and .001, respectively. (See Figure 1 for scale.)

aroundthe largerpeaks, forming what might be called "districts" units of social interaction like the largercommunitiesin the northwest,but with anaddedterritorial dimension(Figure12).Wethus the occupational interpret valleys as low points in the interaction landscape,producedby less inter-

action between clusters than within them. These valleys, then, are analogousto the flat areassepabut ratingthe largercommunitiesin the northwest, the makeit possibleto delineatemoresharply they territorial boundaries betweenthe communitiesin the southeast.

14

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 10. A very low contour line for the Hongshan power 1 surface (Figure 9) encircles the bases of the occupational peaks where they rise from the flat unoccupied plane.

Figure 12. Contours describing the topography between the lower and higher lines of Figures 9 and 10, and "district" boundaries following the occupational "valleys" in the southeast.

Figure 11. A higher contour line delineating the bases of Hongshan occupational peaks rising from the uplifted southeastern corner (dashed line) of the power 1 surface (Figure 9).

of The distribution the estimated of populations definedin thepower or the65 communities districts 1 smoothedsurfaceshows a largenumberof communitieswith quitesmallpopulations (Figure13). to These 51 communities largelycorrespond small local communities as originally defined in the unsmoothedsurface,modifiedonly by a few clussmallcommunities. teringsof pairsof neighboring The 14 largestof the 65 communities(with estiover 150) areclearlysetoff from matedpopulations the smaller ones in the frequency distribution (dashedline in Figure 13). It is these 14 that suggest the existence of furthercommunitystructure above the level of the small local community on interaction a dailybasis. involvedin face-to-face of It appears not all the Hongshaninhabitants that the regionwere involvedin suchlargerscale communities. The 51 smaller communitiesfrom the power 1 surface, which do not cluster strongly together, although substantiallymore numerous for account onlyabout thanthelarger communities, In of one-fourth the estimated regionalpopulation. most of the higher-order communities,into which of the otherthree-fourths the regionalpopulation the areorganized, smalllocal communities occupy

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

15

theroleof building comblocks,as thehigher-order munitiesconsistof clustersof up to 10 smalllocal communities in the (although twoinstances, higherorder communities are single settlements).The clustersare spreadacrossdistancesof as much as 3 km, a spatialscale thatseems too largeto involve face-to-face interaction on a daily basis. We attribute formationof these clusters to more the intensiveinteraction among the small local communitieswithineach clusterthanwith communities outsideit. Meaningful Units of Analysis The communitiesdefinedthusfarcan be the units of further formalanalysesatdifferent scales.When distance-interaction are systematically principles as distributions, above,there appliedto settlement is some reason to believe that the resultingunits human communities. Units do, in fact,approximate of this sort shouldprovidea much sounderfoundationfor rank-sizegraphsand othersuch analyses thansites as traditionally definedin the field. As noted above, a rank-sizegraphfor the 125 Hongshansmall local communitiesdefinedin the Chifeng surveyregion shows a convex patternof considerablestrengthand statisticalsignificance (Figure8). This suggests very little centralization attheregionalscale,butrather presenceof multhe tiple independentsystems (Johnson 1977, 1980, 1981; Paynter 1983). To pursue this possibility, rank-sizegraphscould be constructedseparately for each of the higher-order communities(Figure 14). These are small samples,so significancelevels will be low, but almostall show a good degree of centralization eitherprimate with or patterns patterns not significantlydifferentfrom log-normal. The picturethatemerges,then,is one of a seriesof communities modestsize (populations of between about150 and650 in an areaaround3 km across), most composedof multiplesmall local communities thatclustertogether.Internally, these higherordercommunities tendto be centralized with one localcommunity considerably of larger population thanthe others. Backattheregionalscaleof analysis, higherthe ordercommunitiesare units whose relationships can be investigated. rank-sizegraphof these 14 A communitiesagain shows a strong higher-order and highly significantconvex pattern(Figure 15;

Figure 13. Stem-and-leaf plot of estimated populations for the 65 clusters of Hongshan occupation defined from the power 1 smoothed surface. The 14 higher-communitiesfall below the dashed line.

16

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 14. Rank-size graphs for 11 of 14 Hongshan higherorder communities (those comprised of three or more small local communities). A values and 67 percent confidence zones are determined as suggested by Drennan and Peterson (2004).

Figure 15. Rank-size graph for 14 Hongshan higher-order communities. A value and 67 percent confidence zone are determined as suggested by Drennan and Peterson (2004).

A = .614, p < .01). This might be interpreted a as lackof integration a scalethatencompassed mulat communities.In occupational tiple higher-order surfaces withevengreater smoothing to a power (up of .001 areshownin Figure9) theindividual higherordercommunitiesremainclearlyevident.Justas therank-size wouldleadus to believe,no singraph gle large communityforms a much higher peak whose basal area "captures" others,even though the peaks do blend together more from the smoothing. An alternative interpretation might be thatthe communities the southeastern in higher-order portion of the surveyregionwere partof a yet larger systemwhose centerlies well outsidethe areasurof of veyed. Evaluation this interpretation, course, would requireadditionalinformation.If a much morepopulouscommunity existedoutsidethe surareato the southeast(one which might have vey servedas a centerintegrating communitiesin the the surveyarea),its additionto the datasetwould create a higher peak in the smoothedsurface a whose slopes would engulf the higher-order peak communitiesalreadyidentified.If such structure communities existed,thenthe higher-order would, in turn,have become buildingblocks in that yet in structure, just the sameway thatthe larger-scale intersmalllocalcommunities dailyface-to-face of actionwerethebuildingblocksof thehigher-order communities districts. delineation comand The of in interaction structure smoothedsurfaces munity then, can representingoccupationaldistribution, continuestepwisein thisfashionto largerscalesof analysis, with greatersmoothingof the surfaces allowingdelineationof unitsfor analysisat larger scales. In a studythat,like this one, focuses on delinat eating spatialpatterning differentscales, Ebert (1992:173-185) discourageswhathe calls "semianalysis"based on surfacesor contourmaps representingartifactdensities.He worriesespecially thatshorter longerz-axes show detailsof spatial or atdifferent scaleswithoutproviding any patterning to differentiatebetween them. Varyingthe way scale of the z-axis is indeeda clumsy way to seek such patterning; progressivesmoothing, as sughere, reveals the phenomenaof different gested spatial scales much more effectively. Like the choice of cutoffcontoursfor definingclusters,the decisionas to whichdegreesof smoothingto focus

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

17

on is subjective,althoughby no means arbitrary. The two least-smoothed surfacesin Figure9 show essentially the same small-scalepatternof local communities. thirdandfourth The throwthelargerof scale patterning higher-order communitiesinto sharprelief. In the fifth surface,the higher-order communitiesbegin to bluras the still larger-scale structure extendingbeyondthe surveyarea,more evident in the sixth surface,begins to come into focus.Comparing thesesurfaces a provides reliable at way to home in on phenomena threeclearlydifferent sociospatialscales. That these phenomena inventedand imposedupon arenotjust structures the data is shown by the following examples, in which surfaces with different characteristics lead of unequivocally to verydifferent descriptions interaction structure. A Highly Integrated System We can only speculateaboutthe possibility of a scale centralizedsystem extendingbeyond largertheChifengstudyarea,butone is actuallyobserved for early hierarchical societies in the (2125 km2) of Valleyof Oaxacain thesouthern highlands Mexico (Blantonet al. 1982; Kowalewskiet al. 1989). Theunsmoothed occupational (power4 orgreater) surfacesfor Rosariophase (700-500 B.C.) Oaxaca showclearlydefinedsmalllocal communities, each consistingof a single contiguousoccupation of area,or occasionallya combination two orthree distancesfrom othnearbyareas,at considerable ers (Figure 16). Much like the set of Hongshan local communities,the vast majorityof those in Rosariophase Oaxacaare small hamletsor farmsteads(Figure17) withfewerthanabout50 inhabto estimatesmade itants,according the population by Blanton et al. (1982) and Kowalewskiet al. (SanJose (1989). Oneespeciallylargecommunity Mogote) standsout in the Rosariophase distribuIts tion, with over 500 inhabitants. populationis similarto thatof the largestHongshanlocal community,but,while thereare otherHongshanlocal in withpopulations thehundreds, San communities Jose Mogote is uniquein Oaxaca;no othercomits munityeven approaches size. Althoughits population is similar to the largest Hongshan local San community, JoseMogoteis a clusterof 10 separatearchaeological occupationareasspanninga distanceof roughly2 km, morethantwice the dis-

tance across the largestHongshanlocal commuof nity.Thiswouldsuggestthattheinhabitants San Jose Mogote were not as tightly bound together through interaction as those of Hongshan communities. structure also visiblein themore is Larger-scale smoothed surfaces representing Rosario phase occupation.In the surfaceproducedby an inverse distancepowerof .5 (Figure16), SanJose Mogote 28 "captures" smaller local communitiesin one arm of the roughlyY-shapedvalley. This largerscale interaction communityhas the same territorial character the Hongshandistrictsdiscussed as are with a above;its boundaries easily demarcated contour atthelevelof theinflection line pointwhere the surfacerises up fromthe nearlyflatzero-occupation-level plane(Figure18). The estimated populationof this San Jose Mogote districtis almost twice thatof the largestHongshandistrict(1,100 versus650), but its constituent local communities are spreadover a much largerterritory some 25 kmfromone endto theother,compared no more to than5 km for Hongshan. Thecharacteristics theHongshan of surface that enabledus to identifydistrictboundaries altoare getherabsentfromthe Rosariosmoothedsurface: the San Jose Mogote districttakes the form of a single very tall occupationalpeak, not of several peaksseparated lowersaddlesandvalleys.This by indicationof a well-integrated systemis bolstered to rank-sizegraph(Figby the log-normal primate ure 19;A - -.674) for the 29 local communitiesin the San Jose Mogote district.In termsof internal then,the San Jose Mogote districtand integration, theHongshan districts similar. are UnliketheHongshandistricts,however,the San Jose Mogote districtdoes not have neighborsof comparable size. The only other clustering suggested by the smoothed surfaceis a combinationof two local in communities the southeastern of theValley. part While this higher-order communityhas a population in the rangeof Hongshanhigher-order communitypopulations,the San Jose Mogote district is morethanfivetimeslarger(Figure20), andthese are the only two higher-ordercommunities the smoothedRosario surfacesuggests. The remainder of the local communities, like a substantial of number Hongshan localcommunities, small, are do not cluster,and are apparently involvedin not structure. This picture, anylargerscale interaction

18

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 16. Surfaces representing Rosario phase occupation in the Valley of Oaxaca. Smoothing increases from top to bottom, with inverse distance powers of 4, 2, 1, .5, .25, and .001, respectively. (See Figure 18 for scale.)

of only two higher-order communitieswith vastly sistent with warfareand raiding (Blanton et al. different is at variance withthenotion 1999:42-44;MarcusandFlannery 1996:139-144; populations, of three competing chiefdoms at a standoff in Spencer and Redmond 2003:32-34). Rivalry RosariophaseOaxaca,although is perfectlycon- between multiple higher-ordercommunities of it

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

19

Figure 17. Histogram of estimated populations for the 65 Rosario phase local communities in the Valley of Oaxaca.

roughlyequal statuswould be a much more plaudissibleinterpretation theHongshan of settlement tribution thanof Rosariophase Oaxaca.

Figure 19. Rank-size graph for the 35 local communities within the San Jose Mogote district. A value and 67 percent confidence zone are determined as suggested by Drennan and Peterson (2004).

At ologicalremains. a slightlylargerscale,Parkinclus"settlement son(2002:409^15) hasidentified takenhere ters"from such surfaces.The approach Data Collection, Sites, and Landscapes is more like that envisionedby Cherryin that it in Thereis precedent archaeology analyseslike reliesnot on pointlocationsof sites butdirectlyon for areas the one described here. Cherry (1983:395), for areasanddensitiesof surfaceartifacts. Larger for surfaces defin- of higherdensityoccupation higher producelarger, usingdensity example,imagined froma continuous distribution arche- peaks, which more readily"capture" of smaller,less ing "sites" outlyingunits,thusincorporating denselyoccupied the higherlevels of interaction by produced larger RenfrewandLevel (1979) recognized populations. in therelevance larger of politpopulations defining thatlarger centerswould ical boundaries, assuming oneswould. than dominate territories smaller larger of Theproduction occupational densitysurfaces for multiscalaranalysis requirescertainkinds of for data,andthishasimplications how dataarecollected in the field on archaeologicalsurvey.This, in turn,raises some contentiousissues of concepunits.The aims tual, analytical,and observational andvocabulary ouranalysisthusfarwill surely of have caused advocatesof siteless surveyor landexamto scape archaeology class it asjust another of outmoded settlementarchaeology.It may ple come as a surprisethat, while we would unhesipresentedhere "settletatinglylabel the approach we mentarchaeology," also see it as fullyconsistent It with "landscape" approaches. is also in harmony Figure 18. Rosario phase local communities with contour com- with rejection of the "notionsite" as the fundacutoff showing how they combine into higher-order munities. and mentalunit of observation analysis.

20

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 20. Stem-and-leaf plot of estimated populations for all communities defined in the Rosario power .5 surface. The San Jose Mogote district, with its very large population, is at the bottom.

The Chifengdatasetis not builton the concept of archaeological eitheras observational or "site," unit. Rather, dataconsist of a set of the analytical continuousartifactdensity values spreadsystematically across a landscape,quite comparableto those used by Ebert,for example,in the approach he wastempted call "antisite to (Ebert archaeology" 1992:70).Regionalsurveyin Chifengconsistedof complete and systematiccoverage, not implying total recoverybut simply that the entire area (as opposedto scattered sampleblocks)was surveyed.

As usualin suchsurveys,worknormally proceeded from one cultivatedfield to the next, by walking and the next, and so on. These fields, along with other geographicfeatures,provide a naturaland convenient way to partitionthe landscape into recordableunits. The landscapeis the object of study,and these spatialtractsbecome the unitsof 1 observation thestandardized-hacollectionunits that appearin the precedingdiscussion(Drennan followed in Chifeng et al. 2003b). The procedure amounts to superimposingan irregulargrid of The 1-hasquareson the landscape. approximately boundariesof collection units indicatedby solid lines in Figure21 delimitthe actualextentsof surrecoveredon survey.In each collecface artifacts tion unit a systematiccollection was made of all artifactsin a measuredarea, so a density value were could be calculatedfor each unit.If artifacts buttoo sparsefor systematiccollectingto present a be practical, uniformverylow densityvaluewas et assigned(Drennan al. 2003b). This procedureis precisely analogous to the common practiceof excavatingby grid squares. Thereis no reasonwhy an excavationgrid could not be madeof a seriesof irregularly spacedwavy lines, except that it would be extremelyinconvenientto do so. At the regionalscale, however,it is 1 thesurveying of layingoutregular -hasquares task that kilometers becomes of acrosshundreds square at and inconvenient time-consuming. Fortunately, the regional scale, this is not necessary,because most regions,like Chifeng and Oaxaca,are natugrid units by the field rally dividedinto irregular streams,roads,andotherfeaturesthat boundaries, form the fundamental organizationof collection units. It is only necessaryto recordthem as they are encountered (quitequickly and easily on aeror ial photographs satelliteimagery)andonly then if archaeological remainsareactuallypresent.Just as excavation grids sometimes consist of 1-m and sometimesof 2-m squares, sometimes squares, of smalleror largerunits, surveycollection units couldbe smalleror largerthan 1 ha, dependingon and the level of resolutionattainable neededfor a particular study. Ouranalyseshavevariedin scalefromthe local communitiesor discommunity,to higher-order tricts,to structure involving multipledistricts.In and Oaxacaat least, the local communiChifeng ties we haveidentifiedarevariable,andgeneraliz-

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

21

Figure 21. A collection unit "grid" superimposed on a small portion of the satellite image of the Chifeng region. The density of Hongshan sherds (per m2)is shown for each collection unit where any archaeological materials were found. The hypothetical collection units where archaeological material might have been found but was not are indicated by dashed lines.

ing aboutthem necessitatesworking with a fairsized sample one large enough to encapsulate extremesof variationtogetherwith common patterns.This would not have been possible in these two caseswithsurveydatafrommuchless than100 km2, and choosing the "right"100-km2 survey block would really have required knowing the resultsof ouranalysisbeforehand. larger-scale The we have looked at extendthroughat least patterns several hundredsquarekilometers.We are thus clearlyin the realmof regional-scaleanalysis. Dunnell and Dancey (Dunnell 1992; Dunnell andDancey1983)advocated "sitelesssurvey" techfor recordingartifactdensitiesas they varniques ied across a landscape.They found it practicalto map individualartifactlocationsin an areaof .08 km2.Ebert(1992) mappedindividual artifact locations in a 6.25-km2sample from a largerregion, andCamilliandEbert(1992) did the same in 4.48 were km2,althoughin high-densityareasartifacts counted by 1-m grid square instead of being in usedin Chifeng mapped place.Thefieldmethods (Drennanet al. 2003b) to producethe datasetwe have analyzed are simply an extension of these methodsto a yet largerscale.The notionof a range of scales of study,fromvery small at high resolution to very large at lower resolution,is a wellestablished principle in many fields, including

archaeology(e.g. Kowalewski1990). In practice, there is confusion aboutjust how big "large"or "regional" scale is in archaeology. Kvamme (2003:435, 438), for example, refers to "large regions"and"largeburiedcultural archaeological The that landscapes" consist of "tensof hectares." withherethus tworegional we surveys haveworked farcoveredareasof 765 km2and2, 125km2, respecis 's tively.Kvamme "large region" roughlyanorder of magnitude larger than Dunnell's study area; Ebert' is anotherorderof magnitudelargerthan s Kvamme's;ChifengandOaxacaaretwo ordersof the s. thanEbert' Obviously, same magnitude larger methodology cannot be applied to this range of At scales acrossfourordersof magnitude. thetruly artifact mapregionalscale,thedetailof individual ping appropriately gives way to the lower resoluin tion pictureof densitymeasurements gridunits by field-by-fieldsurvey.This is forced provided et (cf. uponus by practicality Cherry al. 1988:161), but more importantly,field-by-field resolution, while too low for the study of a few hectares,is fully sufficientfor studyingpatternsacross hundreds of squarekilometers.Even contemplating mappingindividualartifactlocationsacross such a large arearecalls UmbertoEco's (1994) elaborationof JorgeLuis Borges's(1960) conceitof the folly of mappingthe Empireat a scale of 1:1. Althoughit is not widely recognized,pioneering "settlementpattern"studies in the Basin of et Mexico (Sanders al. 1979)andtheValleyof Oaxaca (Blantonet al. 1982; Kowalewskiet al. 1989) were alreadyemployingmuchthis same approach in some40 yearsago.Although bothstudies, groups to of contiguouscollection units were referred as "sites,"the basic unit of observationwas not the site butthe field, or collection.A successful"fieldby-field"techniquewas devised aftera field procedure focused on locating sites was rejected et (Sanders al. 1979:20-30).Oncethefield-by-field was completed,"sites"were identified post survey hoc based on artifactdensities, assessed subjectively in the field.The methodologyutilizedin the Valleyof Oaxacaimprovesuponthatof the Basin surfacecolof Mexico principally regularizing by assessments out so as to movechronological lecting The Chifeng of the field and into the laboratory. methodologycontinuesin this directionby incorporatingsystematicartifactcollection to provide assessmentof artimorerigorousandquantitative

22

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

fact densities. Survey of the Nemea Valley in et Greece(Cherry al. 1988;Cherry al. 1991)also et unitspreciselyanal"tracts" spatial proceeded by ogous to "collectionunits"in Chifeng.As in the Basin of Mexico and Valley of Oaxaca surveys, Cherryet al. defined"sites"post hoc. Systematic into the surfacecollecting was also incorporated NemeaValleysurvey,althoughin a somewhatdifferentformthanused in Chifeng. To be sure, many,and perhapsmost, regional settlementsurveyscarried in recentyearshave out backward into uncriticalutilizationof the slipped notion "site"in ways thatdegradedatacollection and conceptuallyundermineanalysis:by recording only pointlocationsfor sites in the firstplace, examinetheentirelandby failingto systematically for archaeological remains,by not assessing scape artifact densities,andin otherways. Suchpractices have perhapsbeen encouraged publicationsin by which sites are represented dots on maps even as when field survey was not a process of "finding sites"(e.g., Sanderset al. 1979). None of the settlementstudiesreferred aboveusedsitesas a unit to of observation; or "sites," "settlements," "communities"werein all cases delineated posthoc as clustersof high-density artifacts a particular of period. In this light, the fundamentaldistinctions some have made between settlementarchaeologyand various approachesto siteless survey seem con1998). siderablyoverdrawn (e.g., Wandsnider similarlines, we find no necessaryconAlong flict betweenthe kind of studywe engage in here in anda varietyof landscapeapproaches archaeolin landscapearchaeology,"ourgoal is to ogy. If, distinguish,from among the many simultaneous dancesthathaveoccurred, those specificdancesof interestto a particular researchquestion"(Wandsnider1998:87),thenpatterns social interaction of are one of those dances. This dance is different fromthoseof constructed, or conceptual, ideational Anscheutzet al.; KnappandAshlandscapes(e.g., more 1999), but in no way denies theirexistence or importance. is just different simultaneous, It but different. Althoughthe level of resolutionis lower (and the scale larger), analysispresented hereis funthe a clusteranalysisbased on a continudamentally ous scale of artifact density values, like the "siteless" or "antisite" approaches of Dunnell (1992), Ebert (1992), and others. Communities

as haveappeared clustersof higherdensityvalues. denWe have also seen largetractswhereartifact sities are effectively zero. These "vacant"areas separatewhat archaeologistshave traditionally called sites. These areas are, of course, not truly in senseof theterm.If we looked vacant thestrictest in practically fieldin Chifeng,we any long enough Thereis, could probablyfind at least one artifact. however, a very real and dramatic difference between fields where artifacts are exceedingly scarce and otherswhere densitiesare substantial. These latterare the potentialsettlementlocations among sedentarypeoples, while the formermay evidence well represent interestingand important of "off-site"(i.e., awayfromresidence)activities. that It shouldcome as no surprise the archaeorecordfor a sedentary populationhas this logical on characteristic. Sedentary peoplescanbe counted of to dispose of sizeablequantities garbagein and betweenthe aroundtheirresidences.The contrast of remainsaround dwellingsand sedentary density activitiesat some disthose produced scattered by tance from them will be sharp.For this reason,a low-resolutionlarge-scalepictureof artifactdensities provides primarilyan indicationof where people lived. The whole issue of where people "lived"is much more complicatedfor residence patterns with a higher degree of mobility. The archaeologicalsignaturesof residence are much of moreephemeral, largerproportions materand ial remainsaregenerallydepositedawayfromresidentiallocations.As Cherry(1983:396) implies, it is no coincidencethat concernaboutan undue attenand focus on "settlements" aboutinadequate tion to remains outside settlements, however fromthose who study defined,has come primarily less sedentary peoples(Chang1992;Dunnell1992; Dunnell and Dancey 1983; Ebert 1992; Foley are and 1981). "Settlements" "communities" not, however, "crippledanalytic units" (Wandsnider read1998:101),but insteadare real phenomena, in context.Theirtraces observable ethnographic ily area fundamental of thelandscape evidence part of critical importance to the study of social interaction. As noted above, communities,as a particular humaninteraction,can never way of structuring be assumed to exist, nor can they be simply assumednot to exist. The questionis an empirical one. If tracesof communitiescan be found, then

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

23

of somethinghas been learnedaboutthe structure socialinteraction. haveargued bothHongWe that shanChifengandRosarioOaxacashow clearevidence of a structure small local communities of embedded in larger-scalepatterning.If we had failedto findsuchcommunitystructure, would we also have learnedsomethingaboutsocial organization in these two regions. The conclusion we arrived at is only convincing if the analytical approachwe employedcould have failed to find evidenceof communitystructure none existed. if A finalcase shows thatpreciselythis is possible. Failing to Find Communities The surfaces Figure22 represent in RegionalClassic period(A.D. 1-900) occupationin a 317 km2 in surveyregionin the Alto Magdalena the Andes of southwestern Colombia(Drennan2000; Drennan, ed. 1985; Drennanet al. 1991; Drennanand Quattrin1995). In contrastto the cases we have looked at thusfar,the unsmoothed(power4) surfacedoes notconsistof widelyseparated peaks,but rather denselyconcentrated of occupational spikes whose bases are crowdedso closely togetherthat oftentheycannotbe separated. Thusit is impossilike ble to definesmalllocal communities those of Chifengor Oaxacafromthis surfacewith a cutoff contour. low contourin Figure23 delineatesa The wide varietyof differentkinds of phenomena.Of the separateoccupationareas delineated,a very are largeproportion so small thatthey likely representsingle isolatedhouseholdsor groupsof no more than two or three households. Some areas as couldplausiblybe interpreted small local comzones of sparse munities,mostly in the peripheral but occupation, otherlargeareasof uninterrupted reachover9 km2andsome6.5 kmfrom occupation the one endto the other fartoo largeto represent so sortof smalllocal communities easily identified for ChifengandOaxaca.Selectinga highercutoff contour high enough to fragment these large - resultsin theomissionof numerous smaller areas or shorterspikes. The Alto Magdalenaevidence, then, providesno indicationof the kind of suprastructure seen to be central householdinteraction in to local social organization Chifeng and Oaxaca.Itis thusclearthatouranalysisdoesnotassume structure of a priori the presenceof a particular small local communities.In the Alto Magdalena,

we have failed to consistentlyfind the clustersof high artifactdensity values argued to represent small local communitiesin Chifengand Oaxaca. The power 1 smoothed surface (Figure 22), interaction struchowever,does show larger-scale ture,as was thecase withChifengandOaxaca.The occupational topography of this survey region resembles that of the southeasternsector of the boundariescan Chifeng example, and "district" be definedalongoccupational againreadily valleys (Figure24). In this instance,the entiresurveyarea is easilydivided districts into someappar(although extend beyond the survey limits); there are ently no isolatedhigher-order communities therewere as in the northwestern sectorin Chifeng.Reinforcing our interpretation such occupational that topograterritorial social units,a single cenphy represents ter with architectural sculpturalevidence of and occursin eachof the fourmore politicalleadership or less complete districtsidentifiedin Figure 24 (Drennan 2000; Drennan and Quattrin 1995). Unlike the Hongshanhigher-order communities and districtsin Chifeng,the Alto Magdalenadistrictsdo notappear be constructed smalllocal to of but builtupdirectlyfromindicommunities, rather vidualhouseholds. ForboththeAlto Magdalena Chifeng,modand ernoccupational distribution community and structurerelateto each otherin ways very like whatwe haveargued the archaeological for cases discussed here. For Chifeng,the separatepeaks rising from the flatplanein the unsmoothedsurfacein Figure 25 reflectthe well-definedsmall local communities visible on the modernlandscape(Figure26). Evenwherethesecommunities mostnumerous are andmost closely spaced,theyremainseparate and identifiable.Social, political,and economic interactionis todaystronglyfocusedwithinthese communities, where the local representatives of are larger-scale politicalorganization found,within which landtenureand agricultural are production organized,and where specializedproductionand exchangetakeplace.Residentsof thesesmalllocal are in communities boundtogether relationships of economic interdependencethrough a vigorous commercein local goods and services including farm produce, repairs to industrially produced For goods, and even restaurants. the Alto Magmodern dalena,on the otherhand,the unsmoothed occupationalsurface (Figure 27), much like the

24

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 22. Surfaces representing Regional Classic period occupation in one survey area of the Alto Magdalena. Smoothing increases from top to bottom, with inverse distance powers of 4, 2, 1, .5, .25, and .001, respectively.(See Figure 23 for scale.)

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

25

Figure 23. Cutoff contour that fails to delimit clear local communities in the unsmoothed surface for Regional Classic occupation in the Alto Magdalena (Figure 22).

RegionalClassicperiodone fromFigure22, has a multitude tiny separatespikes,which represent of individualfarmsteads(Figure 28). The kinds of interactionthat producethe Chifeng community Land ownership patternare absentor attenuated. andthe organization agricultural of are production not communityfunctionsbut insteadare centered in individual households.Modernpoliticalauthorat ity has effectively no representation this very local level in theAlto Magdalena, sharpcontrast in

is localcommerce very to Chifeng. Andsmall-scale poorlydeveloped. do interaction patterns exist today Larger-scale in both Chifeng and the Alto Magdalena. In Chifeng,districtsarebuilt up of small local communitiescenteredon a largevillage ortown.There strucof is thusa three-tiered hierarchy interaction turein the modernChifeng surveyregion:household, small local community,and large village or town. In the Alto Magdalena,in additionto the

Figure 24. Contour map of Regional Classic power 1 smoothed surface (Figure 22). District boundaries along occupational valleys are indicated with heavy lines.

26

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

Figure 25. Unsmoothed surface and contour cutoff delimiting small local communities for modern occupation in the Chifeng region.

widely dispersedoccupationof families living on the land they own and farm, there are two small townswherecommercial politicalfunctions and are concentrated. Some of thefunctionsof thesetowns arelike those carriedout at virtuallyall villages in of Chifeng. Interchange locally producedgoods, however,is minimal;most commerceinvolvesthe distribution manufactured distilled) goods of (or

fromthe cities andthe sale of the local coffee crop to outsidebuyers.The residentsof these towns do form local communities, which can be seen as in largerclustersof occupation Figure27. Outside these towns, however,there is little sign of local ruralhouseholdsparticipate communitystructure; in town functionsdirectlyand individually. They do not clusterinto small local communitiesthem-

Figure 26. Small local communities in modern Chifeng.

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

27

Figure 27. Unsmoothed surface and contour cutoff that fails to clearly delimit small local communities for modern occupation in part of the Alto Magdalena.

selves, since their level of interactionwith their is neighbors low.(Andtheydo notformthe sprawling areasof contiguousoccupationseen in Figure to 23, which areprobablyattributable the gradual of artifacts thepalimpsest and effect shifting surface of shifting dispersed household locations over structure interaction of time.)Thusthehierarchical in theAlto Magdalenasurveyregionhas only two tiers:householdand small town.

Communities and Prehistoric Social Change The surfacesfor Chifengandthe Alto Magdalena of (Figures25 and 27) representthe distribution in modernoccupation acrossthe landscape exactly the same way the surfacesdiscussedearlierrepreof sent prehistoric occupation.The patterns interaction that can be directly observedtoday agree well with the structures revealedby applyingdis-

Figure 28. Modern dispersed settlement in the Alto Magdalena.

28

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005

scalelandtance-interaction evidenceis partof theregionalprinciplesto these modernsur- relevant faces. The approaches have proposedfor pre- scape, available for archaeological study with we historic cases rely on precisely these same appropriate analyticaltools. Thereis no reasonto supposethatthese principles. We Acknowledgments. thank Stephen Kowalewski for proare in than viding data from the Valley of Oaxaca survey in electronic principles anyless applicable prehistory in modern-day ruralChina or Colombia;in fact, form and Gregory Indrisanoand William Honeychurchfor thecompleteabsenceof modern and inviting us to participatein the electronic symposiumat the transportation communication wouldonlymeanthat Society for AmericanArchaeology meeting in Montrealfor technologies which this paperwas originallywritten.Additionalthoughts distance-interaction had principles a greater impact were provoked by Simon Holdaway, Jeffrey Parsons, and on howpeoplechoseto distribute across otherparticipantsin the discussion at that symposium. themselves a landscape. Distance-interaction principles,then,can plauReferences Cited and usefully be appliedto the interpretation sibly of prehistoric and settlementdistributions the pur- AnayaHernandez, for Armando,StanleyP. Guenter, MarcU. Zender of defininginteraction communities differat pose A Model 2003 SakTz'i' , A ClassicMayaCenter: Locational ent scales. The result is not only to provide an LatinAmericanAntiquity Based on GIS and Epigraphy. 14:179-191. interestingbasis for substantivecomparison of ancientcommunities(Drennanand Haller 2004; Anschuetz, Kurt R, Richard H. Wilshusen, and Cherie L. Scheick Drennan Peterson and 2005), butalsoto bringmore 2001 An Archaeology of Landscapes:Perspectivesand more Directions. Journal of Archaeological Research sharplyinto focus the need to understand 9:157-311. the natureof the interactions that produced fully Richard GaryM. Feinman, A. Blanton, E., Stephen Kowalewski, and sustained these communities. As has often andLindaM. Nicholas scale analysishas 1999 AncientOaxaca:TheMonteAlbdnState.Cambridge provedto be the case, regionalUniversityPress,Cambridge. us aboutthe nature prehistoric of socienlightened Richard StephenKowalewski, E., GaryFeinman,and eties while simultaneouslyhelping to formulate Blanton, JillAppel 1982 MonteAlbdn's Hinterland,Part I: The Prehispanic more questionsthatcan only be answeredthrough Patternsof the Centraland SouthernPartsof Settlement intensivestudy at smallerscales. While we have the Valley Oaxaca,Mexico.Memoirsof the Museumof of suggestedgeneric descriptionsof interaction patAnthropology, Universityof Michigan,No. 15. scalepicture will Borges,JorgeLuis terns,thelow-resolution regional1960 Del Rigoren la Ciencia.InEl Hacedor,p. 103. Emece nottell us whatactivitieswereorganized (andhow) Editores,BuenosAires. withincommunities, how these led to particular Camilli,Eileen L., andJamesI. Ebert or kindsof interaction, turnproducing interacin Surface the 1992 ArtifactReuse andRecyclingin Continuous and for LandUse Distributions Implications Interpreting tion structures to differentkinds of compeculiar In Patterns. Space,Time, Archaeological and Landscapes, munities. Answering these questions strikes edited by JacquelineRossignol and LuAnnWandsnider, the pp. 113-136. PlenumPress,New York. directlyat understanding forces that produce social change and requiresmore detailedbut still Canuto,MarcelloA., andJasonYaeger(editors) A Per2000 TheArchaeologyof Communities: New World individual comfairlyextensivemeansof sampling spective.Routledge,London. munities,combinedwith still moreintensivestud- Chang,Claudia The Ethnoarchaeology 1992 ArchaeologicalLandscapes: ies of their constituent household units. The of PastoralLandUse in the GrevenaProvinceof Northemergenceof large-scalesocial formations(those ern Greece. In Space, Time,and ArchaeologicalLandoften referredto by terms such as "chief dom"or scapes, edited by Jacqueline Rossignol and LuAnn Wandsnider, 65-89. PlenumPress,New York. pp. is a processof differentiation, the heart at "state") JohnF. of which is changein the natureand operationof Cherry, 1983 Frogs Round the Pond: Perspectives on Current the smallerscale communitiesand householdsof Archaeological Survey Projects in the Mediterranean Region. In ArchaeologicalSurveyin the Mediterranean which they are composed. Comparingand conArea,editedby DonaldR. KellerandDavidW. Rupp,pp. theselarge-scalesocialformations, they as trasting Series 155, Oxford. 375^16. BAR International have existed throughtime and space, dependson Cherry,JohnE, JackL. Davis, Anne Demitrack,Eleni Mantand zourani,ThomasF. Strasser, LaurenE. Talalay ourabilityto findthesecommunities the archaein 1988 Archaeological Landscape: Surveyin anArtifact-Rich ological record(if they existed)by recoveringthe A MiddleNeolithicExamplefromNemea,Greece.Amerinteractionstructures that constitutedthem. The ican Journalof Archaeology92:159-176.

Peterson and Drennan]

REGIONAL-SCALE ANALYSIS OF HUMAN INTERACTION

29

JohnR, JackL. Davis, andEleni Mantzourani Cherry, 1991 Landscape Archaeology as Long-TermHistory: NorthernKeos in the CycladicIslandsfrom EarliestSettlementuntilModernTimes.Monumenta Archaeologica, 16. UCLA Institute Archaeology, Angeles. of Los ArchaeologicalSurveyTeam ChifengCollaborative 2003a Neimenggu Chifengdiqu1999 nianquyuxingkaogu diaochabaogao [RegionalArchaeologicalSurveysin the Chifeng Region, Inner Mongolia in 1999]. Kaogu 2003(5):24-34. Collaborative Chifeng International ArcheologicalResearch Project 2003b RegionalArcheologyin EasternInnerMongolia:A Science Press,Beijing. Methodological Exploration. American ChineseTeam ChifengCollaborative Archeological 2002 NeimengguChifeng diqu quyu xing kaogu diaocha jieduan xing baogao (1999-2001) [A Phased Reportof Full-Coverage Regional Archeological Surveyin Chifeng, Inner KaoguYangjiu Mongolia( 1999-200 1)]. InBianjiang [Researchon China'sFrontier Archaeology],Vol.1., pp. 357-368. Science Press,Beijing. Chisolm,Michael 1970 RuralSettlement Land Use. Aldine, Chicago. and RobertD. Drennan, 1988 HouseholdLocationand CompactversusDispersed in Mesoamerica. Household In Settlement Prehispanic and in Past:CaseStudiesin the Community theMesoamerican MayaArea and Oaxaca, edited by RichardR. Wilk and WendyAshmore,pp. 273-393. Universityof New Mexico Press,Albuquerque. 2000 Las SociedadesPrehispdnicasdel Alto Magdalena. Instituto Colombiano Antropologia Historia, de e Bogota. RobertD. (editor) Drennan, 1985 Regional in de Archaeology theValle la Plata, Colombia: A PreliminaryReport on the 1984 Season of the Proyecto Arqueologico Valle de la Plata. Museum of of Technical Anthropology, University Michigan, Reports, No. 16. RobertD., andMikaelJ. Haller Drennan, 2004 EarlyStates and Also-Rans:Comparative DevelopmentalTrajectories. Paperpresentedat the AnnualMeetMontreal. Archaeology, ing of the Society for American Elizabeth RobertD., LuisGonzaloJaramillo, Ramos, Drennan, and Carlos AugustoSanchez,Maria Angela Ramirez, Carlos A. Uribe 1991 RegionalDynamicsof Chiefdoms in the Valle de la Plata, Colombia. Journal of Field Archaeology 18:297-317. RobertD., andChristian Peterson E. Drennan, SettlementSystemswith 2004 Comparing Archaeological Rank-SizeGraphs:A Measureof Shape and Statistical Confidence. Journal of Archaeological Science 31:533-549. The Set2005 Early ChiefdomCommunitiesCompared: Recordfor Chifeng,theAlto Magdalena, tlementPattern and andtheValleyof Oaxaca.In Settlement, Subsistence, editedby RichardE. Blanton.Cotsen Social Complexity, Instituteof Archaeology,University of California,Los Angeles, in press. Drennan, Robert D., Christian E. Peterson, Gregory G. Indrisano, Teng Mingyu, Gideon Shelach,ZhuYanping, M. Katheryn Linduff,andGuo Zhizhong 2003a Chapter4: Approachesto Regional Demographic In Reconstruction. RegionalArcheologyin EasternInner Science Press, Exploration. Mongolia:A Methodological Beijing. RobertD., andDale W. Quattrin Drennan,

1995 Social Inequalityand Agricultural Resourcesin the Vallede la Plata,Colombia.In TheFoundations Social of Inequality,edited by GaryM. Feinman,and T. Douglas Price,pp. 207-331. PlenumPress,New York. Drennan, Robert D., Teng Mingyu, Christian E. Peterson, GideonShelach,Gregory Indrisano, Yanping, G. Zhu Katheryn M. Linduff,Guo Zhizhong,andManuelA. RomanLacayo 2003b Chapter3: Methods for ArcheologicalSettlement Study.In RegionalArcheologyin EasternInnerMongolia:A Methodological SciencePress,Beijing. Exploration. Dunnell,RobertC. 1992 The Notion Site. In Space, Time,andArchaeological Landscapes,edited by JacquelineRossignol and LuAnn Wandsnider, 2\-A\. PlenumPress,New York. pp. Dunnell,RobertC, andWilliamS. Dancey 1983 The Siteless Survey:A RegionalScale Data Collection Strategy. Advancesin ArcheologicalMethodand In Theory,Vol.6, pp. 267-387. AcademicPress,New York. Ebert,JamesI. 1992 Distributional of Archaeology. University UtahPress, Salt LakeCity. Eco, Umberto 1994 OntheImpossibility Drawinga Mapof theEmpire of at a Scale of 1 to 1. In How to Travelwith a Salmonand OtherEssays, pp. 95-106. Harcourt Brace,New York. KentV (editor) Flannery, 1976 The Early MesoamericanVillage.Academic Press, New York. Foley,RobertA. 1981 Off-SiteArchaeology:An Alternative Approachfor the Short-Sited. Patternsin thePast,editedby IanHodIn der,pp. 157-183. Cambridge Press,Cambridge. University Hegmon,Michelle 2002 Conceptsof Community Archaeological in Research. In Seeking the CenterPlace: Archaeologyand Ancient Communities the Mesa Verde in Region,edited by Mark D. Varienand RichardH. Wilshusen,pp. 263-379. UniversityUtahPress, Salt LakeCity. Hodder,Ian,andClive Orton 1976 SpatialAnalysisinArchaeology.Cambridge University Press,Cambridge. Holdaway,Simon, and Patricia Fanning 2004 Developinga Landscape Historyas Partof a Survey New SouthWales,AusStrategy: ExamplesfromWestern tralia.Paperpresented the AnnualMeetingof the Sociat ety forAmericanArchaeology,Montreal. Isbell,WilliamH. 2000 WhatWe ShouldBe Studying: The "Imagined ComIn munity"andthe "Natural Community." TheArchaeolA editedby ogy of Communities: New World Perspective, Marcello CanutoandJasonYaeger, 243-366. RoutA. pp. ledge, London. A. Johnson,Gregory 1977 Aspectsof RegionalAnalysisinArchaeology. Annual Reviewof Anthropology 6:479-508. 1980 Rank-SizeConvexityandSystemIntegration: View A fromArchaeology. EconomicGeography56:234-347. 198 1 Monitoring and ComplexSystemIntegration Boundary Phenomenawith SettlementSize Data. In Archaeoeditedby logical Approachesto the Studyof Complexity, S. E. van derLeeuw,pp. 144-188. Universiteit Amsvan Amsterdam. terdam, Knapp,A. Bernard 2003 The Archaeology of Community in Bronze Age PolitikoPhoradesin Context.AmericanJournal Cyprus: of Archaeology107:559-580.

30

AMERICAN ANTIQUITY

[Vol. 70, No. 1 , 2005]

and Knapp,A. Bernard, WendyAshmore 1999 Archaeological Constructed, Landscapes: ConceptuIn alized,Ideational. Archaeologiesof Landscapes:Coneditedby WendyAshmoreandA. temporary Perspectives, Bernard Knapp,pp. 1-30. Blackwell,Oxford. Kolb,MichaelJ., andJamesE. Snead 1997 It's a Small WorldAfterAll: Comparative Analyses of CommunityOrganization Archaeology.American in 62:609-628. Antiquity Kowalewski,StephenA. 1990 Scale and Complexity:Issues in the Archaeologyof the Valley of Oaxaca.In Debating Oaxaca Archeology, edited by Joyce Marcus,pp. 207-370. Anthropological Papers,Museumof Anthropology, Universityof Michigan, No. 84. 2003 WhatIs the Community? LongView fromOaxThe aca, Mexico. Social Evolutionand History2:4-34. Kowalewski,StephenA., Gary M. Feinman,LauraFinsten, RichardE. Blanton,andLindaM. Nicholas 1989 MonteAlbdn's Hinterland, PartII: PrehispanicSettlementPatternsin Tlacolula,Etla, and Ocotldn,the Valley of Oaxaca, Mexico. Memoirs of the University of No. MichiganMuseumof Anthropology, 23. Kuijt,Ian (editor) 2000 LifeInNeolithicFarmingCommunities: Social Organization, Identity and Differentiation. Kluwer Academic/PlenumPublishers, New York. Kvamme,KennethL. 2003 Geophysical Surveys as LandscapeArchaeology. American 68:435-457. Antiquity and Linduff,Katheryn RobertD. Drennan, GideonShelach M., 2004 EarlyComplexSocieties in NE China:The Chifeng International Collaborative ResearchProArchaeological ject. Journalof FieldArchaeology29:45-73. Marcus,Joyce 2000 Towardsan Archaeology of Communities.In The editedby M.A. Canutoand Archaeology Communities, of J.Yaeger,pp. 231-342. Routledge,London. Marcus,Joyce, and KentV. Flannery 1996 ZapotecCivilization: How UrbanSocietyEvolvedin Mexico's Oaxaca Valley. ThamesandHudson,London. Murdock,GeorgePeter 1949 Social Structure. MacMillan,New York. Olsson, Gunnar 1965 Distanceand HumanInteraction: ReviewandBibA liography. Regional Science Research Institute, Philadelphia. RobertW. Paynter, 1983 Expanding the Scope of Settlement Analysis. In ArchaeologicalHammersand Theories,edited by J. A. Moore and A.S. Keene, pp. 233-375. Academic Press, New York. Plog, Stephen 1976 Measurement Prehistoric of Interaction betweenCommunities.In TheEarly MesoamericanVillage,editedby KentV.Flannery, 255-372. AcademicPress,NewYork. pp.

Preucel,RobertW. 2000 Making Pueblo Communities:ArchitecturalDiscourse at Kotyiti, New Mexico. In TheArchaeologyof A Communities: New World Perspective,editedby Marcello A. CanutoandJasonYaeger,pp. 58-77. Routledge, London. Rogers,J. Daniel, andBruceSmith(editors) and Households. Uni1995 Mississippian Communities veristyof AlabamaPress,Tuscaloosa. (editors) Rossignol,J., andL. Wandsnider Plenum and 1992 Space,Time, Archaeological Landscapes. Press,New York. WilliamT., JeffreyR. Parsons,andRobertS. Santley Sanders, 1979 The Basin of Mexico: Ecological Processes in the AcademicPress,New York. Evolutionof a Civilization. Schwartz,G. M., and S. E. Falconer(editors) Smith1994 ArchaeologicalViews from the Countryside. D.C. sonianInstitution Press,Washington, Spencer,CharlesS., andElsa M. Redmond 2003 Militarism, Resistance,andEarlyStateDevelopment in Oaxaca, and Mexico.SocialEvolution History2:25-70. Stone, GlennD. 1993 AgrarianSettlementand the SpatialDispositionof In Labor. SpatialBoundariesand Social Dynamics:Case Studies Societies,editedbyAugustin fromFood-Producing Holl and Thomas E. Levy, pp. 273-393. International Ann in Monographs Prehistory, Arbor. Trigger,BruceG. of 1968 The Determinants SettlementPatterns.In Settlement Archaeology,edited by K. C. Chang, pp. 53-78. NationalPressBooks, PaloAlto, California. LuAnn Wandsnider, Land1998 RegionalScale ProcessesandArchaeological editedby Ann scapeUnits.In UnitIssues inArchaeology, F. RamenovskyandAnastasiaSteffen,pp. 87-102. Universityof Utah Press,Salt LakeCity. Wilk, RichardR., andWendyAshmore(editors) 1988 Household and Communityin the Mesoamerican Past: Case Studies in the Maya Area and Oaxaca. Universityof New Mexico Press,Albuquerque. Yaeger,Jason in of 2000 The Social Construction Communities the Clasof sic MayaCountryside: Strategies Affiliationin Western A Belize. In TheArchaeology Communities: New World of Perspective, edited by Marcello A. Canuto and Jason Yaeger,pp. 123-142. Routledge,London. Yaeger,Jason,andMarcelloA. Canuto In 2000 Introducing Archaeologyof Communities. The an Archaeology of Communities,edited by Marcello A. CanutoandJasonYaeger,pp. 1-15. Routledge,London.

ReceivedApril 21, 2004; Revised October13, 2004; Accepted October21, 2004.

You might also like