the social networks and structural variation of ... · 44). simply, social capital is a set of...

6
The social networks and structural variation of Mississippian sociopolitics in the southeastern United States Jacob Lulewicz a,1 a Department of Anthropology, Washington University in St. Louis, St. Louis, MO 63130 Edited by David G. Anderson, University of Tennessee, Knoxville, TN, and accepted by Editorial Board Member Elsa M. Redmond January 16, 2019 (received for review October 24, 2018) Network approaches in archaeology offer a promising avenue for facilitating bottom-up, comparative approaches to sociopolitical or- ganization. While recent applications have focused primarily on migration and demographic trends, identity and identity politics, and the dynamics of geopolitical and regional interaction, little in the way of comparative sociopolitical organization has been attempted. In this study, I present an alternative approach to the use of sociotypological models across southern Appalachia. In particular, I demonstrate the value in employing network analyses as a mode of formally and quantitatively comparing the relational structures and organizations of sociopolitical landscapes; in this case, those tradi- tionally characterized as constellations of chiefdoms. By approaching southern Appalachian histories through the relationships upon which social, political, and economic institutions were actually built, I move the study of southeastern political systems beyond the use of models that emphasize the behaviors of elites and the ruling class as inspired by the ethnographic and ethnohistoric records. To these ends, using a robust regional ceramic dataset, I compare network histories and political landscapes for the southern Appalachian region between ca. AD 800 and 1650. The results of these analyses contribute insights to the study of small-scale political organizations by demonstrating that (i ) as chiefdoms developed, leaders drew on preexisting social and political conditions; (ii ) while networks of chiefly interaction were defined by instability, wider networks of interaction were much more durable; and (iii ) quantitative network analyses and qualitative ethnohistoric accounts can articulate with one another to shed light on indigenous political organization. social network analysis | sociopolitical organization | archaeology | southeastern United States I n AD 1540, Spanish explorer Hernando de Soto, traveling across the southernmost stretch of the Appalachian Mountains, encountered an indigenous political landscape characterized by a number of autonomous villages loosely bound together under the influence of a single powerful chief, then residing at the town of Coosa in what is today northern Georgia (Fig. 1) (1, 2). Indeed, the ethnohistoric record of the southeastern United States is replete with descriptions of indigenous social and political organization (see refs. 1 and 35) and has been employed extensively in archaeological contexts as a model for reconstructing political landscapes reaching back centuries before European contact. Of particular import has been the use of the ethnohistoric record as a model (or at least a preliminary basis) for identifying, characterizing, and interpreting the emergence and dynamics of complex social and political orga- nizations in the archaeological record of the Southeast. In this study, I present an alternative, comparative approach to the use of socio- typological models in the southern Appalachian region. In particu- lar, I demonstrate the value in employing network analyses as a mode of formally and quantitatively comparing the relational structures and organizations of sociopolitical landscapes; in this case, those traditionally characterized as constellations of chiefdoms. The quantitative analyses employed here, which shed light on the development of complex societies like chiefdoms, are replicable and testable and serve to inform our understanding of structural dynamics that are often difficult to detect through more qualitative approaches. The purpose of this paper is not to negate, confirm, or even critically contemplate the concept of the chiefdom. Nor is it the purpose of this study to evaluate the use of the chiefdom concept in navigating the deep indigenous histories of the southeastern United States. Rather, the purpose is to explore fundamental variability among political landscapes and, more specifically, to explore the re- lational foundations through which sociopolitical institutions were inevitably constituted, generated, maintained, and transformed. As such, the major implications of this study include (i ) an understanding that as regional chiefdoms developed, community leaders drew on preexisting social and political conditions to advance claims to power and status; (ii ) the realization that although networks of chiefly in- teraction and engagement through warfare, trade, and strategic diplomacyexperienced cycles of development and collapse (e.g., refs. 68), wider networks of intimate relationships were much more durable; and (iii ) the demonstration that ethnohistoric accounts of the chiefdom of Coosa encountered by Spanish explorers (which have been used intensively to interpret archaeological patterns) articu- late well with the archaeological network evidence for the develop- ment of region-wide networks during this protohistoric period. Chiefdoms, as commonly defined, are a generalized societal type that refers primarily to hierarchical political organization based on kinship. Such generalized patterns have been well documented Significance An alternative approach to the use of sociotypological models in interpreting sociopolitical organization is presented. The devel- opment of analytical and conceptual strategies to compare and describe sociopolitical organizations is a cornerstone of anthro- pological inquiry. This study contributes to such disciplinary foci by presenting a formal, comparative approach to the study of sociopolitics, particularly contributing to discussions of the evo- lution and development of organizational complexity in small- scale, nonstate societies. Beyond archaeology however, this study demonstrates the value of a deep, historical perspective on the evolution of human networks more broadly and highlights the value of a productive relationship between contemporary social, economic, and political theory and the historical sciences for the investigation of both past and present societies. Author contributions: J.L. designed research, performed research, analyzed data, and wrote the paper. The author declares no conflict of interest. This article is a PNAS Direct Submission. D.G.A. is a guest editor invited by the Editorial Board. Published under the PNAS license. See Commentary on page 6519. 1 Email: [email protected]. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1818346116/-/DCSupplemental. Published online February 19, 2019. www.pnas.org/cgi/doi/10.1073/pnas.1818346116 PNAS | April 2, 2019 | vol. 116 | no. 14 | 67076712 ANTHROPOLOGY SEE COMMENTARY Downloaded by guest on September 4, 2020

Upload: others

Post on 17-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

The social networks and structural variation ofMississippian sociopolitics in the southeasternUnited StatesJacob Lulewicza,1

aDepartment of Anthropology, Washington University in St. Louis, St. Louis, MO 63130

Edited by David G. Anderson, University of Tennessee, Knoxville, TN, and accepted by Editorial Board Member Elsa M. Redmond January 16, 2019 (received forreview October 24, 2018)

Network approaches in archaeology offer a promising avenue forfacilitating bottom-up, comparative approaches to sociopolitical or-ganization. While recent applications have focused primarily onmigration and demographic trends, identity and identity politics,and the dynamics of geopolitical and regional interaction, little in theway of comparative sociopolitical organization has been attempted.In this study, I present an alternative approach to the use ofsociotypological models across southern Appalachia. In particular, Idemonstrate the value in employing network analyses as a mode offormally and quantitatively comparing the relational structures andorganizations of sociopolitical landscapes; in this case, those tradi-tionally characterized as constellations of chiefdoms. By approachingsouthern Appalachian histories through the relationships uponwhich social, political, and economic institutions were actually built,I move the study of southeastern political systems beyond the use ofmodels that emphasize the behaviors of elites and the ruling class asinspired by the ethnographic and ethnohistoric records. To theseends, using a robust regional ceramic dataset, I compare networkhistories and political landscapes for the southern Appalachianregion between ca. AD 800 and 1650. The results of these analysescontribute insights to the study of small-scale political organizationsby demonstrating that (i) as chiefdoms developed, leaders drew onpreexisting social and political conditions; (ii) while networks ofchiefly interaction were defined by instability, wider networks ofinteraction were much more durable; and (iii) quantitative networkanalyses and qualitative ethnohistoric accounts can articulate withone another to shed light on indigenous political organization.

social network analysis | sociopolitical organization | archaeology |southeastern United States

In AD 1540, Spanish explorer Hernando de Soto, travelingacross the southernmost stretch of the Appalachian Mountains,

encountered an indigenous political landscape characterized by anumber of autonomous villages loosely bound together under theinfluence of a single powerful chief, then residing at the town ofCoosa in what is today northern Georgia (Fig. 1) (1, 2). Indeed, theethnohistoric record of the southeastern United States is repletewith descriptions of indigenous social and political organization (seerefs. 1 and 3–5) and has been employed extensively in archaeologicalcontexts as a model for reconstructing political landscapes reachingback centuries before European contact. Of particular import hasbeen the use of the ethnohistoric record as a model (or at least apreliminary basis) for identifying, characterizing, and interpretingthe emergence and dynamics of complex social and political orga-nizations in the archaeological record of the Southeast. In this study,I present an alternative, comparative approach to the use of socio-typological models in the southern Appalachian region. In particu-lar, I demonstrate the value in employing network analyses as amode of formally and quantitatively comparing the relationalstructures and organizations of sociopolitical landscapes; in this case,those traditionally characterized as constellations of chiefdoms.The quantitative analyses employed here, which shed light on

the development of complex societies like chiefdoms, are replicable

and testable and serve to inform our understanding of structuraldynamics that are often difficult to detect through more qualitativeapproaches. The purpose of this paper is not to negate, confirm, oreven critically contemplate the concept of the chiefdom. Nor is it thepurpose of this study to evaluate the use of the chiefdom concept innavigating the deep indigenous histories of the southeastern UnitedStates. Rather, the purpose is to explore fundamental variabilityamong political landscapes and, more specifically, to explore the re-lational foundations through which sociopolitical institutions wereinevitably constituted, generated, maintained, and transformed. Assuch, the major implications of this study include (i) an understandingthat as regional chiefdoms developed, community leaders drew onpreexisting social and political conditions to advance claims to powerand status; (ii) the realization that although networks of chiefly in-teraction and engagement—through warfare, trade, and strategicdiplomacy—experienced cycles of development and collapse (e.g.,refs. 6–8), wider networks of intimate relationships were much moredurable; and (iii) the demonstration that ethnohistoric accounts ofthe chiefdom of Coosa encountered by Spanish explorers (which havebeen used intensively to interpret archaeological patterns) articu-late well with the archaeological network evidence for the develop-ment of region-wide networks during this protohistoric period.Chiefdoms, as commonly defined, are a generalized societal type

that refers primarily to hierarchical political organization based onkinship. Such generalized patterns have been well documented

Significance

An alternative approach to the use of sociotypological models ininterpreting sociopolitical organization is presented. The devel-opment of analytical and conceptual strategies to compare anddescribe sociopolitical organizations is a cornerstone of anthro-pological inquiry. This study contributes to such disciplinary fociby presenting a formal, comparative approach to the study ofsociopolitics, particularly contributing to discussions of the evo-lution and development of organizational complexity in small-scale, nonstate societies. Beyond archaeology however, this studydemonstrates the value of a deep, historical perspective on theevolution of human networks more broadly and highlights thevalue of a productive relationship between contemporary social,economic, and political theory and the historical sciences for theinvestigation of both past and present societies.

Author contributions: J.L. designed research, performed research, analyzed data, andwrote the paper.

The author declares no conflict of interest.

This article is a PNAS Direct Submission. D.G.A. is a guest editor invited by the Editorial Board.

Published under the PNAS license.

See Commentary on page 6519.1Email: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1818346116/-/DCSupplemental.

Published online February 19, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1818346116 PNAS | April 2, 2019 | vol. 116 | no. 14 | 6707–6712

ANTH

ROPO

LOGY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Page 2: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

across the southeastern United States and especially the southernAppalachian region (1, 2, 8–11). A product of mid-20th-centuryneoevolutionism (see refs. 12 and 13), the use of sociopoliticaltypes has dominated archaeological interpretations of social andpolitical organization for over a half-century, not only across the USSoutheast, but across the world. Although these approaches con-tributed invaluably to advancing understandings of indigenous po-litical histories, they lack the appropriate conceptual mechanisms toaccount for significant variation between cases. Indeed, theseneoevolutionary frameworks have elicited countless responses forrefinement and nuance (e.g., refs. 14–16) as well as total aban-donment (e.g., ref. 17). Other authors continue to propose com-parative approaches that move beyond categorical concepts andoutdated classificatory systems (e.g., ref. 18). Even so, we continueto lack any standard mode of comparing the underlying relationalstructures of political landscapes. Recent network approaches inarchaeology (e.g., refs. 19–21) provide a promising avenue in fa-cilitating a comparative approach to sociopolitical organization.Although recent applications have focused primarily on migrationand demographic trends (e.g., refs. 22 and 23), on identity andidentity politics (e.g., refs. 24 and 25), and on the dynamics ofgeopolitical and regional interaction (26, 27), little in the way ofcomparative sociopolitical organization has been attempted.Network analyses and frameworks highlight the role of rela-

tionships in generating particular social, political, and economicforms. Indeed, they are explicitly anticategorical (28). With thisunderstanding, we can move beyond black-box approaches thatattempt to explain social phenomena through the use of top-down sociopolitical models and toward approaches that allowus to parse these kinds of structures into their constituentelements (28). By approaching the sociopolitical histories ofsouthern Appalachia from the bottom-up, through the relation-ships upon which social, political, and economic institutions arebuilt, I move the study of southeastern political systems beyondgeneralizations that emphasize the political activities of elites asinspired by the ethnographic and ethnohistoric records.To these ends, I construct a network history for the southern

Appalachian region between ca. AD 800 and 1650. I compare net-works that straddle two major critical transitions. The first of thesetransitions, ca. AD 1050 to 1150, is traditionally characterized by theemergence of chiefdom-level political organizations associated with

classic Mississippianization that includes increased sociopoliticalcomplexity, institutionalized socioeconomic inequality, new ritualand religious practices, and intensified agricultural economies (29).For the southern Appalachian region, this includes the emer-gence of Etowah, what would have been the largest and mostcomplex community in the region and likely coeval with the earlyemergence of Mississippian cultural practices at the Macon Plateausite in central Georgia (8, 30, 31). The second transition is charac-terized primarily by the abandonment of the community at Etowahca. AD 1325 to 1350 and the subsequent reorganization of thesouthern Appalachian settlement landscape. In contrast to the de-velopmental sequence of Etowah and northern Georgia, suchregional hierarchical organizations and intensive communityaggregation seem to have been absent from the historical trajectoryof communities in eastern Tennessee (32–34). To explore thesepolitical landscapes before and after these transitions, I draw on adatabase of roughly 350,000 ceramics from nearly 100 archaeolog-ical components from across southern Appalachia (Fig. 1). Al-though the political landscapes spanning this 850-y period have beendescribed as being dominated by chiefdom (or incipient chiefdom)political organizations, I explore variation in the actual networktopologies and structural characteristics of these landscapes.

Results and DiscussionTwo different kinds of networks are evaluated and compared. Oneset of networks is built from information on the highly visibledecorative attributes adorning the exterior of ceramic vessels. Theother set is built on low-visibility technological choices, particularlytemper selection (the aplastic materials added to clay paste to alterits thermal and mechanical properties). High-visibility attributesreflect patterns of social signaling (e.g., refs. 25, 26, and 35),whereas lower-visibility technological choices reflect more intimatenetworks of interaction and information exchange through com-munities of practice (e.g., refs. 25, 36, and 37). As such, the net-works considered in this study represent, first and foremost, therelational fields within which potters were embedded. Throughboth diachronic and synchronic comparisons of these differentkinds of networks, I demonstrate the complexity of the relationalfoundations of political landscapes and highlight the ways in whichdifferent kinds of relationships may be structured independently ofone another to generate sociopolitical forms. A more in-depth,detailed overview of ceramic data and variables is provided in SIAppendix, SI Materials and Methods.In interpreting the network graphs in Fig. 2 and the variables

related to network characteristics presented in Table 1, I draw on arobust body of literature from social economics, political science,and economic policy that correlates network structures with partic-ular kinds of social capital (e.g., refs. 38–43). At its foundations,social capital describes relations of trust and reciprocity, the normsand networks that enable people to act collectively, and the insti-tutions that generate and sustain these norms and networks (38, 43,44). Simply, social capital is a set of “relations that enable actors togain, maintain, or expand access to resources” (ref. 42, p. 112) andreferences the “resources embedded in one’s social network that canbe accessed or mobilized through ties in the network” (45). As anarchaeology of sociopolitics is undoubtedly concerned with the dis-tribution, access, and control of social, political, and economic re-sources (either material or immaterial), such concepts adapted toarchaeological network analysis can serve as a powerful conceptualand analytical framework. However, the use of a social capitalframework in interpreting the networks presented here is merely apreference and should not be taken as prescriptive. I follow (andexpand upon) recent applications of this framework in organizingarchaeological network analyses (e.g., refs. 26 and 46) but do notpresent such a framework as a new model of generalized typologies.Although I choose to use network analyses to formally compare theavailability and potential advantages of different kinds of socialcapital across the southern Appalachian political landscape, the

Fig. 1. The southern Appalachian region showing the location of sitesused in this study. Labels indicate the most well-studied and well-knownMississippian mound sites in the region: Etowah (9Br1), Wilbanks (9Ck5),Bell Field (9Mu101), Little Egypt (9Mu102), Sixtoe Field (9Mu100), Hixon(40Ha3), Dallas (40Ha1), Hiwassee Island (40Mg31), DeArmond (40Re12),Toqua (40Mr6), and Martin Farm (40Mr20).

6708 | www.pnas.org/cgi/doi/10.1073/pnas.1818346116 Lulewicz

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Page 3: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

analyses and comparative approach presented here can be effectivelydeployed within a range of conceptual, middle-range frameworks inevaluating sociopolitical organization.

Communities of Practice (Low-Visibility Attributes). For networksbased on temper, the same general topology is retained throughoutthe entire 850-y period (Fig. 2, Left). What the visual assessment ofthese networks makes clear is that choice of temper is highly cor-related with geographic proximity, as there is a clear distinction ofsubgraphs between northern Georgia nodes (Fig. 2, red) andeastern Tennessee nodes (Fig. 2, green) across all three periods.There are, however, some significant temporal patterns that can behighlighted. Each side of the earliest-period graph represents analmost-complete subgraph (one for sites in Georgia and one forsites in Tennessee) in which connections are virtually ubiquitousamong all sites, indicating a high degree of similarity in temperusage. Even if driven primarily by raw-material variability, interac-tions resulting in the transfer of either technological information,

finished products, raw materials, or people seem to have beenlimited during this period. However, during the following period,AD 1050 to 1325, although the general structure characterized bytwo geographically defined subgraphs is maintained, many moreconnections between the two subgraphs are established. Most ofthese connections are funneled through the large, politico-religiouscenter of Etowah (9Br1, depicted as a red star in Fig. 2).The same general structure that characterized the AD 1050 to

1325 period continued to characterize the AD 1325 to 1650 period.Again, although two distinct subgraphs remain, connections be-tween the two subgraphs remain heightened, as evidenced by higheraverage similarity coefficients and an increased number of tiesbetween subgraphs compared with the earliest period, AD 800 to1050. Thus, even after the abandonment of Etowah (ca. AD 1325)and the major reorganization of the sociopolitical landscape, theintegrity of patterns of interactions producing these networks wasretained. Notably, for the region as a whole, the overall similarityvalues between all sites were heightened compared with previous

Fig. 2. Networks based on temper (Left) and surface decoration (Right) for all three time periods considered here. Sites located in northern Georgia areshown in red. Sites located in eastern Tennessee are shown in green. The site of Etowah (9Br1) is depicted as a red star. Ties are determined using the strengthof Brainerd-Robinson coefficients of similarity between individual sites. Node and tie positions are determined using a spring-embedded multidimensionalscaling algorithm. A heavily annotated version of this figure depicting network features discussed in the text can be found in SI Appendix.

Lulewicz PNAS | April 2, 2019 | vol. 116 | no. 14 | 6709

ANTH

ROPO

LOGY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Page 4: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

periods, indicating more intense or frequent interactions betweennorthern Georgia and eastern Tennessee communities.While general network topologies remained similar, the qualities

of the relationships comprising these networks changed significantly.In the first period, no communities emerged as particularly central.This also seems to be the case for the third period, AD 1325 to 1650.However, in the second period, AD 1050 to 1325, more nodes ex-hibit centrality scores above the SD than in any other time period. Itis clear that there is a relationship between the emergence ofchiefdomlike political entities (ca. AD 1050) and changes to regionalnetworks. The sociopolitical institutions that characterized the pe-riod between AD 1050 and 1325 clearly facilitated the emergence ofcentral actors in their social networks. Given the topology of thesenetworks, central actors drew on networks of interaction at the re-gional scale, establishing connections and relationships with com-munities from across the southern Appalachian region as opposedto their local networks alone. These patterns are consistent withextant narratives (e.g., refs. 8 and 47) that argue for a shift towardnetwork-based political strategies (sensu ref. 48) just after the turn ofthe first millennium AD. Thus, as leadership roles became re-gionally centralized, political actors drew on preexisting networks ofrelationship to establish a landscape of exclusive interactions. Giventhe stability of network topology through time, it can be argued thatthe network structures through which hierarchical political organi-zations and socioeconomic inequality emerged were established longbefore the emergence of the institutions themselves.

Social Signaling (High-Visibility Attributes). Similar to the generaltrend identified in the networks based on temper, the network to-pologies for social signaling practices generally retain their struc-tures throughout the 850-y period considered here (Fig. 2, Right). Incontrast to the structures represented by the distribution and use oftemper, the structures related to signaling practices are much lesssegregated by geography and, indeed, much less polarizing in gen-eral. In each of the three networks, no distinct subgraphs emerge asthey did in the temper networks. In the first period, AD 800 to 1050,the network is characterized by three weak subgroups that are allstill heavily connected to one another, all of which include sites fromboth eastern Tennessee and northern Georgia. In the second pe-riod, AD 1050 to 1325, the integration between sites is further en-hanced as any indication of subgroups disappears and a structurecharacterized by a general mixture of both strong and weak tiesemerges. Although the average tie strength remains the same as the

tie strength for the previous network, the similarity cutoff is signif-icantly higher. Both similarity cutoffs, however, are substantiallyhigher than the similarity values characterizing the networks basedon temper alone.While the integration between sites is strengthened from the

first to the second time period, the integration of the overallnetwork becomes more compartmentalized as chiefdoms fill thelandscape. This is especially apparent in reviewing the geodesicdistance variables in Table 1. Compared with the first period,geodesic distances between nodes (the number of nodes one mustpass through to move between any two nodes) are lengthenedsubstantially in the second period, indicating that information,resources, and interactions would have been more restricted andless freely flowing between communities. Combined with an in-crease in the number of actors with heightened betweenness val-ues, there would have been more potential for information,interactions, and network flows to have been subject to controland manipulation than they previously would have been.In the final period considered here, after the collapse of

Etowah as a major regional center (ca. AD 1325), the average tiestrength increases further, indicating the continued dissolution ofdifferences between northern Georgia and eastern Tennesseedecorative traditions. In addition to the increased similarity be-tween the two localities, overall network integration and co-hesiveness increases, with greatly declining geodesic distancesindicating the opening up of network relationships and interac-tions and a lower degree of control and potential exclusivitywithin the network. The number of central nodes also continuesto drop in this final period, with no nodes exhibiting heightenedcentrality values. Interestingly, this increased network cohesive-ness, likely reflecting a shift in patterns of social signaling towardone that emphasizes region-wide participation in some loose orsituational organization, occurs during the period within whichthe Coosa chiefdom would have likely begun to crystallize.

Social Capital and Comparative Sociopolitical Networks. All three ofthe networks based on the use of temper are defined by bonding(i.e., strong) ties within subgraphs and by weak, bridging ties be-tween subgraphs. All three of the networks based on surface dec-orations are defined by both strong and weak ties, linking nodes inlong strung-out networks indicative of bridging structures andcapital. Bridging ties (long, weak ties) tend to rely not on the rule ofenforcement or informal collective actions like bonding ties would,

Table 1. Metrics and measurements for networks depicted in Fig. 2

Measure

Temper Surface

AD 800–1050 AD 1050–1325 AD 1325–1650 AD 800–1050 AD 1050–1325 AD 1325–1650

BR cutoff for graph 80 70 100 120 155 145Average tie strength (BR) 75 72 108 116 116 127Network degree centrality 13 6 10 13 9 11Node degree centralities > 1σ 0 7 0 3 0 0Network betweenness centrality 2 1 8 3 0 7Node betweenness centralities > 1σ 0 2 0 3 4 1Network eigenvector centrality 6 3 8 7 3 4Node eigenvector centralities > 1σ 1 2 1 3 0 0Longest geodesic distance 11 18 13 14 21 9Average geodesic distance 4.06 8.61 5.31 5.41 7.71 4.51Compactness/cohesion 0.514 0.231 0.323 0.277 0.218 0.334Fragmentation/breadth 0.486 0.769 0.677 0.723 0.782 0.666Max k-core 17 15 10 5 9 9Nodes in max k-core, % 63 38 55 19 23 50Nodes in k-core of ≥3, % 67 100 61 85 74 78

Although the networks in Fig. 2 were constructed by binarizing ties based on a subjective BR similarity cutoff value, measures of centrality, geodesicdistance, compactness, fragmentation, and k-cores are based on weighted ties between nodes, or the actual BR values, not binarized matrices of ties. Units forgeodesic distances and k-cores are node counts. BR, Brainerd-Robinson coefficient of similarity.

6710 | www.pnas.org/cgi/doi/10.1073/pnas.1818346116 Lulewicz

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Page 5: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

but instead depend on more formalized institutions such as legal orreligious organizations (ref. 38, p. 392). What this may indicate is thepresence of formal institutions situated at the regional scale thatserved to integrate populations from eastern Tennessee and north-ern Georgia as evidenced by the distribution of decorative traditionsacross each locality. Further, bridging capital is argued to emergefrom weak ties across society in which individuals and organizationalbehavior are embedded (40), but which is nevertheless held togetherthrough diverse group memberships and the sharing of some com-mon values (ref. 42, p. 115). In general, weak ties like those consti-tuting the signaling networks exist among members of social groupsthat are heterogeneous, having different social identities (ref.42, p. 115). These relationships rely on earned trust among looselyconnected peoples rather than on ascribed trust among a homog-enous, strongly related group (ref. 40 and ref. 42, p. 115). All threesignaling networks are consistent with characteristics of bridgingnetworks and are likely good indicators of loose relationships,interactions, and exchanges through which generalized trust amonggroups was established. These patterns of generalized trust, andthe regional institutions maintaining and generating these patterns,seem to have remained stable for at least the 850-y period con-sidered here.Bonding capital emerges from strong social ties which are based

on social identities such as family, kinship, gender, ethnicity, or or-ganizational culture (ref. 42, p. 114). As a consequence of cohesion,bonding capital generates a type of trust that is prescribed tomembers of a group. Such groups tend to generate high levels oftrust, cooperation, and organization, facilitating collective action andlearning. Modes of production are often based on bonding ties thatmay be reinforced through practices of sharing resources andredistributing profits (49). Because networks based on temper exhibitmany of the structural characteristics of bonding networks, they arelikely good referents for networks of kin or other strong relationshipsgenerating high levels of trust and reciprocity. The scale and struc-ture of such intimate relationships as kin ties, lineage affiliations, andmodes of intergenerational enculturation seem to have remainedintact in the context of major sociopolitical transformations. While atsome level the very relationships constituting the social, political, andeconomic form of society were redefined as part of new sociopolit-ical institutions and practices, the most intimate ties and relation-ships among members of southern Appalachian societies seem tohave been successfully enacted over at least 850 y. This suggests thedurability of some very basic social institutions, including regionalorganizations of kinship groups, patterns of marriage exchange, andcontinued face-to-face interaction through communal events. In-deed, the stability of these institutions likely worked to mediate theuncertainty that political transformations may have generated.What is clear is that southern Appalachian societies were defined

by both bridging and bonding forms of capital and would have hadboth at their disposal. In the context of modern economic studies, ithas been argued that economies need both types of social capital(42). It is necessary for a minimum amount of bonding capital to bepresent for bridging capital to emerge—that is, bridging capitalbuilds on the social capabilities generated by bonding capital andnetworks. While bridging capital is undoubtedly more beneficial toeconomic and sociopolitical development, as it connects broader,more diverse groups to one another, bonding capital provides thefertile ground upon which bridging capital can develop (ref. 42,p. 116). Although sociopolitical institutions were substantiallytransformed at roughly AD 1050 and again at ca. AD 1325, accessto both bonding and bridging capital remained constant acrosssouthern Appalachian populations.

ConclusionsThe political histories of chiefs and their political strategies, asdefined through the use of ethnographic models, do not necessarilyreflect the wider social histories of southern Appalachian peoples.By accessing the underlying relationships through which southern

Appalachian society and politics were enacted, this study pro-vides insight on how we might reconceptualize our archaeologicalnarratives of sociopolitical organization and change, especially interms of the scale, structure, and stability of southeastern chiefdoms.First, it is clear that Etowah established its regional prominence, atleast in part, through relationships that spanned beyond local con-texts—at regional and continental (46) scales—leveraging relation-ships forged between groups living across both eastern Tennesseeand northern Georgia. As such, in considering the geographic con-tainer and social boundaries within which we situate our narrativesof political systems, we must look beyond traditional scales ofanalysis such as the individual polity or river valley. Further, wecannot assume that long ties between regions are the exclusive realmof elite interactions of alliance building. Indeed, the ceramic datasetsI have employed here point toward these elite interactions as beingunderlain by wider networks of intimate relationships that werelikely in place before new political institutions were ever built.Second, the bottom-up network approach to political organi-

zation and activity employed here allows for a finer-resolutioninterrogation of the heterogeneity of social and political structuresthrough time. When the Spanish encountered southern Appalachiansocieties in the 16th century, the political landscape was composedof autonomous communities linked together through loose politicalties, with the chief of the most powerful town at Coosa wieldingsignificant influence over these communities. Through the study Ihave just presented, we know that this 16th-century landscape wascharacterized by loose, open networks of information and resourceflows. Although these ethnohistoric accounts of political structureare often extended into the past, what has been demonstrated hereis that the political landscape 200 y before the crystallization of 16th-century political relationships (a landscape similarly defined bychiefdom political structures) seems to have been markedly differ-ent, characterized by restricted network flows, numerous centralactors, and the structural potential for the monopolization of socialand political relationships. Even though the kinds of social capitalavailable to southern Appalachian communities remained stableover 850 y, it is clear that these pools of capital were leveraged inunique ways and in unique sociohistorical contexts. Hernando deSoto and other Spanish explorers would not have encountered thesame political landscape had they arrived 200 y earlier.The third way that this study contributes to the study of south-

eastern chiefdoms is in its implications for understanding the sta-bility of such political structures. Extant narratives pose thatchiefdom political structures—and the relationships that definedthem—were generally unstable institutions (6, 7). The study pre-sented here, however, demonstrates that the networks underlyingpolitical strategies, figures, and institutions were some of the mostenduring, stable features of southern Appalachian societies. Assuch, it would seem that it was specific elite, political practices andbehaviors that were unstable, not the structures of relationshipsthrough which these behaviors would have been enacted.

Materials and MethodsData. The ceramic data used here totals to 365,331 sherds from 84 componentsacross southern Appalachia, including 276,626 sherds from 43 componentsacross eastern Tennessee and88,705 sherds from41 components across northernGeorgia spanning between ca. AD 800 and 1650 (SI Appendix). Raw attributedata and counts for temper and surface decoration were synthesized fromnumerous published archaeological reports (SI Appendix, Table S1) and areincluded as SI Appendix, Tables S2–S7. Ceramic assemblages were partitionedinto time periods based on a regional ceramic seriation utilizing correspon-dence analysis, frequency seriation, and Bayesian chronological modeling (50,51). Discrete, subcommunity assemblages (e.g., pit features, house floors,midden layers) were used to conduct the seriation; however, assemblages be-longing to the same temporal components determined through seriation wereaggregated to the scale of the site. That is, a single site may be present as anode in more than one of the networks because multiple temporal compo-nents were represented at the site.

Lulewicz PNAS | April 2, 2019 | vol. 116 | no. 14 | 6711

ANTH

ROPO

LOGY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020

Page 6: The social networks and structural variation of ... · 44). Simply, social capital is a set of “relations that enable actors to gain, maintain, or expand access to resources”

Network Analysis. Much like ceramic data are used here as a proxy for re-lationships and connections, measures and indices of network structures areused as proxies for the qualities of those relationships and, more specifically, asproxies for the kinds of social capital available to southern Appalachian societiesin responding to and mediating sociopolitical change. Similarities in the pro-portions of tempering agents across assemblages were used to construct net-works based on communities of practice. Similarities in the proportions ofsurface decorations across assemblages were used to construct the signalingnetworks. A range of quantitativemeasures andmetrics were evaluated to fullydescribe network structures and node positions (SI Appendix, SI Materials andMethods). Ceramic assemblage similarity data were prepared by executingscript written by Peeples (52) using R (version 3.3.2) statistical software. Visu-alization of network graphs was made in UCINET (version 6) (53) and NetDraw(54). The networks in Fig. 2 were produced using binarized ties based on asubjective cutoff of similarity values between nodes. That is, only the strongerties were used to produce network visualizations, the purpose of which wasto highlight the basic network structure while maintaining a single network

component. This process is described in more detail in SI Appendix. Nodes andties were positioned in graph theoretic space using a spring-embedded multi-dimensional scaling algorithm, the default algorithm for node positions inUCINET. Network- and node-level metrics were based on weighted tie values(the actual similarity values between nodes, not a binarized matrix) and calcu-lated using UCINET (53) and R script written by Peeples (55). Ceramic attributes,methods of determining similarity, the process of constructing networks, andmodes of network evaluation are reported in more detail in SI Appendix.

ACKNOWLEDGMENTS. I thank the University of Georgia Laboratory ofArchaeology; the McClungMuseum of Natural History & Culture; the Antonio J.Waring Laboratory of Archaeology; Jennifer Birch, Victor Thompson, SteveKowalewski, Lynne Sullivan, and Tim Baumann; and two anonymous re-viewers and a PNAS member editor for their valuable and constructivereviews, which have contributed greatly to the improvement of this man-uscript. Funding for this project was from the National Science Foundation(Award 1644359).

1. Hudson C, Smith M, Hally D, Polhemus R, DePratter C (1985) Coosa: A chiefdom in thesixteenth-century southeastern United States. Am Antiq 50:723–737.

2. Smith M (2000) Coosa: The Rise and Fall of Southeastern Mississippian Chiefdom(Univ Press of Florida, Gainesville, FL).

3. Hudson C (1990) The Juan Pardo Expeditions: Exploration of the Carolinas andTennessee, 1566-1568 (Univ of Alabama Press, Tuscaloosa, AL).

4. Knight VJ, Jr (1990) Social organization and the evolution of hierarchy in south-eastern chiefdoms. J Anthropol Res 46:1–23.

5. Widmer R (1994) The structure of southeastern chiefdoms. The Forgotten Centuries:Indians and Europeans in the American South, 1521-1704, eds Hudson C, Tesser CC(Univ of Georgia Press, Athens, TN), pp 125–155.

6. Anderson D (1994) The Savannah River Chiefdoms: Political Change in the LatePrehistoric Southeast (Univ of Alabama Press, Tuscaloosa, AL).

7. Hally D (1996) Platform mound construction and the instability of Mississippianchiefdoms. Political Structure and Change in the Prehistoric Southeastern UnitedStates, ed Scarry J (Univ Press of Florida, Gainesville, FL), pp 92–127.

8. King A (2003) Etowah: The Political History of a Chiefdom Capital (Univ of AlabamaPress, Tuscaloosa, AL).

9. Hally D (1994) The chiefdom of Coosa. The Forgotten Centuries: Indians and Europeansin the American South, 1521-1704, eds Hudson C, Tesser CC (Univ of Georgia Press, Athens,TN), pp 227–256.

10. Hally D (2008) King: The Social Archaeology of a Late Mississippian Town inNorthwestern Georgia (Univ of Alabama Press, Tuscaloosa, AL).

11. Peebles C, Kus S (1977) Some archaeological correlates of ranked societies. Am Antiq42:421–448.

12. Chapman R (2003) Archaeologies of Complexity (Routledge, London).13. Trigger B (1989) A History of Archaeological Thought (Cambridge Univ Press, Cam-

bridge, UK).14. Beck R (2003) Consolidation and hierarchy: Chiefdom variability in the Mississippian

Southeast. Am Antiq 68:641–661.15. Blitz J (1999) Mississippian chiefdoms and the fission-fusion process. Am Antiq 64:

577–592.16. Feinman G, Neitzel J (1984) Too many types: An overview of sedentary prestate so-

cieties in the Americas. Adv Archaeol Method Theory 7:39–102.17. Pauketat T (2007) Chiefdoms and Other Archaeological Delusions (Rowan Altamira,

New York).18. Smith M (2012) The Comparative Archaeology of Complex Societies (Cambridge Univ

Press, Cambridge, UK).19. Brughmans T, Collar A, Coward F (2016) The Connected Past: Challenges to Network

Studies in Archaeology and History (Oxford Univ Press, Oxford).20. Knappett C (2013) Network Analysis in Archaeology: New Approaches to Regional

Interaction (Oxford Univ Press, Oxford).21. Mills B (2017) Social network analysis in archaeology. Annu Rev Anthropol 46:

379–397.22. Mills B, et al. (2013) Transformation of social networks in the late pre-Hispanic US

Southwest. Proc Natl Acad Sci USA 110:5785–5790.23. Mills B, et al. (2018) Evaluating Chaco migration scenarios using dynamic social net-

work analysis. Antiquity 92:922–939.24. Borck L (2017) Lost voices found: An archaeology of contentious politics in the

Greater Southwest, A.D. 1100-1450. PhD dissertation (University of Arizona, Tucson,AZ).

25. Peeples M (2018) Connected Communities: Networks, Identity, and Social Change inthe Ancient Cibola World (The Univ of Arizona Press, Tucson, AZ).

26. Birch J, Hart J (2018) Social networks and northern Iroquoian confederacy dynamics.Am Antiq 83:13–33.

27. Hart JP, Birch J, Gates St-Pierre C (2017) Effects of population dispersal on regionalsignaling networks: An example from northern Iroquoia. Sci Adv 3:e1700497.

28. Emirbayer M, Goodwin J (1994) Network analysis, culture, and the problem of agency.Am J Sociol 99:1411–1454.

29. Anderson D, Sassaman K (2012) Recent Developments in Southeastern Archaeology:From Colonization to Complexity (Society for American Archaeology Press, Wash-ington, DC).

30. Schroedl G (1994) A comparison of the origins of Macon Plateau and Hiwassee Islandcultures. Ocmulgee Archaeology 1936-1986, ed Hally D (Univ of Georgia Press, Ath-ens, TN), pp 138–143.

31. Williams M (1994) The origins of the Macon Plateau site. Ocmulgee Archaeology1936-1986, ed Hally D (Univ of Georgia Press, Athens, TN), pp 130–137.

32. Schroedl G, et al. (1990) Explaining Mississippian origins in east Tennessee. TheMississippian Emergence, ed Smith B (Univ of Alabama Press, Tuscaloosa, AL), pp175–196.

33. Schroedl G (1998) Mississippian towns in the eastern Tennessee Valley. MississippianTowns and Sacred Spaces, eds Lewis R, Stout C (The Univ Press of Alabama, Tusca-loosa, AL), pp 64–92.

34. Sullivan L (2016) Reconfiguring the Chickamauga basin. New Deal Archaeology inTennessee, ed Dye D (The Univ of Alabama Press, Tuscaloosa, AL), pp 138–170.

35. Bliege Bird R, Smith E (2005) Signaling theory, strategic interaction, and symboliccapital. Curr Anthropol 46:221–248.

36. Crown P (2014) The archaeology of crafts learning: Becoming a potter in the PuebloanSouthwest. Annu Rev Anthropol 43:71–88.

37. Gosselain O (2000) Materializing identities: An African perspective. J ArchaeolMethod Theory 7:187–217.

38. Adger N (2003) Social capital, collective action, and adaptation to climate change.Econ Geogr 79:387–404.

39. Crowe J (2007) In search of a happy medium: How the structure of interorganizationalnetworks influence community economic development strategies. Soc Networks 29:469–488.

40. Granovetter M (1985) Economic action and social structure: The problem of em-beddedness. Am J Sociol 91:481–510.

41. Putnam R (2000) Bowling alone: America’s declining social capital. Culture andPolitics, eds Crothers L, Lockhart C (Palgrave Macmillan, New York), pp 223–234.

42. van Staveren I, Knorringa P (2007) Unpacking social capital in economic development:How social relations matter. Rev Soc Econ 65:107–135.

43. Woolock M, Narayan D (2000) Social capital: Implications for development theory,research, and policy. World Bank Res Obs 15:225–249.

44. Ostrom E, Ahn T (2008) The meaning of social capital and its link to collective action.Handbook on Social Capital, eds Svendsen GT, Svendsen GL (Edward Elgar, Chelten-ham, PA), pp 17–35.

45. Lin N (1999) Building a network theory of social capital. Connections 22:28–51.46. Lulewicz J, Coker A (2018) The structure of the Mississippian world: A social network

approach to sociopolitical interactions. J Anthropol Archaeol 50:113–127.47. Cobb C, King A (2005) Re-inventing Mississippian tradition at Etowah, Georgia.

J Archaeol Method Theory 12:167–193.48. Blanton R, et al. (1996) A dual-processual theory for the evolution of Mesoamerican

civilization. Curr Anthropol 37:1–14.49. Repetti M (2002) Social relations in lieu of capital. Res Econ Anthropol 21:43–59.50. Lulewicz J (2018) Network histories of southern Appalachia, AD 600-1600. PhD dis-

sertation (University of Georgia, Athens, TN).51. Lulewicz J (2018) Radiocarbon data, Bayesian modeling, and alternative historical

frameworks: A case study from the US Southeast. Adv Archaeol Pract 6:58–71.52. Peeples MA (2011) R script for calculating the Brainerd-Robinson coefficient of simi-

larity and assessing sampling error. Electronic document. Available at www.mattpeeples.net/BR.htmlhttp://www.mattpeeples.net/br.html. Accessed October 1, 2017.

53. Borgatti SP, Everett MG, Freeman LC (2002) UCINET for Windows: Software for socialnetwork analysis (Analytic Technologies, Lexington, KY), Version 6.0.

54. Borgatti SP (2002) NetDraw: Graph visualization software (Analytic Technologies,Lexington, KY), Version 6.

55. Peeples MA (2017) Network science and statistical techniques for dealing with un-certainties in archaeological datasets. Electronic document. Available at mattpeeples.net/netstats.html. Accessed December 15, 2018.

6712 | www.pnas.org/cgi/doi/10.1073/pnas.1818346116 Lulewicz

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

4, 2

020