# FORAGER MOBILITY, CONSTRUCTED ENVIRONMENTS, AND EMERGENT SETTLEMENT HIERARCHY: INSIGHTS FROM ALTIPLANO ARCHAEOLOGY

FORAGER MOBILITY, CONSTRUCTED ENVIRONMENTS, AND

EMERGENT SETTLEMENT HIERARCHY:

INSIGHTS FROM ALTIPLANO ARCHAEOLOGY by

W. Randall Haas, Jr.

A Dissertation Submitted to the Faculty of the

SCHOOL OF ANTHROPOLOGY

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2014

2

The University of Arizona

Graduate College

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by W. Randall Haas, Jr. titled “Forager Mobility, Constructed Environments, and

Emergent Settlement Hierarchy: Insights from Altiplano Archaeology” and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

Date: 23 July 2014

Mark S. Aldenderfer

Date: 23 July 2014

Steven L. Kuhn

Date: 23 July 2014

Mary C. Stiner

Date: 23 July 2014

J. Stephen Lansing

Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

Date: 23 July 2014

Dissertation Director: Steven L. Kuhn

Date: 23 July 2014

Dissertation Director: Mark S. Aldenderfer

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgment of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

3

SIGNED: W. Randall Haas, Jr.

4

ACKNOWLEDGMENTS

I am truly grateful to the innumerable individuals who helped me on the path to composing this dissertation. My committee including Drs. Mark Aldenderfer, Steven Kuhn,

Mary Stiner, and Stephen Lansing have offered exceptional guidance, insight, patience, and opportunities throughout the course of this work. I am particularly grateful to my advisors, Drs.

Aldenderfer and Kuhn. Dr. Aldenderfer introduced me to Andean archaeology when he invited me to assist in a subsurface testing project in the Rio Ramis region, Puno in 2008. My subsequent work in the region would not have been possible without the many intellectual and logistical doors that he opened. Dr. Kuhn has provided a constant source of wisdom especially as

I endeavored to view hunter-gatherer archaeology through the lens of complex systems theory.

Many of the theoretical ideas and mathematical treatments have origins in conversations with him whether in his office, at professional meetings, or by video conferencing across the country or globe.

As archaeological data and politics created twists and turns in the direction of my research, many other School of Anthropology faculty members contributed guidance along the way. In particular, Drs. David Killick, Barbara Mills, John Olsen, David Raichlen, Michael

Schiffer, and James Watson have been particularly influential in helping me work through theoretical, methodological, and logistical problems. Dr. Greg Hodgins, Becky Watson, Marcus

Lee, and others at the Arizona Accelerator Mass Spectrometry lab assisted with 14C processing.

This research has also benefited from conversations with many other faculty members and fellow students inside and outside of the classroom. The 2007 School of Anthropology cohort and other fellow students too numerous to mention individually have offered incredible friendship, support, and intellectual stimulation throughout this process. I look forward to many more years of friendship and collaboration with them.

Field seasons long and short were made possible with support from the Collasuyo

Archaeological Research Institute (CARI). For field and laboratory assistance, sharing materials, and sharing in great memories, I thank Dr. Elizabeth Arkush, Cecilia Chavez Justo, Erika Brandt,

Karl la Favre, Dr. Elizabeth Klarich, Cynthia Klink, Brieanna Langlie, Dr. Abigail Levine, Dr.

Aimme Plourde, Dr. Carol Schultze, Dr. Charles Stanish, Dr. Matthew Warwick, and Matthew

Velasco. Ceci Chavez analyzed the ceramics, and Brieanna analyzed a sample of macrobotanical remains recovered from Soro Mik'aya Patjxa. Brieanna also generously allowed us to use her brand new, custom-built flotation machine.

I would like to extend a special thank you to my field mentor and great friend, Dr. Nathan

Craig. Dr. Craig patiently introduced me to Andean fieldwork, took me to visit remote sites, and provided council at various stages of my research. I am a great admirer of the depth and breadth of his work. Many of the findings presented here are built on foundations laid by his impressive efforts.

Other Peruvian colleagues have supported field efforts. Carlos Viviano Llave served as the project co-director for the Proyecto Arcaico Tardio del Ilave (PATI). His hard work, knowledge of bureaucracy, and good spirit was indispensable. I thank Dr. Rafael Vega Centano for putting me in touch with Carlos. The community of Mulla Fasiri permitted excavation on their land, and the communities of Totorani and Ilave made me feel welcome during my stay. I

am immeasurably grateful to Albino Pilco Quispe, Virginia Incacoña, Mateo Incacoña Huaraya,

Daniel Pilco Incacoña, Nestor Condori Flores, Dario Pilco Incacoña, and Karen Pilco Incacoña who provided support in the field and laboratory. They patiently helped me better understand and navigate the cultural and natural landscapes of the region, and they became some of my greatest teachers and friends along the way.

Field research was supported by a National Science Foundation Doctoral Dissertation

Improvement Grant (award BCS-1311626) and an American Philosophical Society Lewis and

Clark Fund for Exploration and Field Research. Additional financial support for radiocarbon dating was provided by the School of Anthropology and the Graduate Professional and Student

Council at The University of Arizona. This document was drafted during the support of a

Marshall Foundation dissertation writing fellowship.

On a more personal note, I express my gratitude to the many family members and loved ones who selflessly supported me in many ways on the path to graduate school and completing this dissertation. My mother, Vicki Rush; my father, Randy Haas Sr.; my brother, Jared Haas; my partner Lauren Hayes; and many others including grandparents, aunts, uncles, cousins, and friends provided the inspiration and encouragement to pursue this unconventional adventure.

Daisy and Nellie probably also deserve acknowledgment for their patience and uncanny ability to keep spirits high.

5

Dedicated to the people of Soro Mik'aya Patjxa.

6

7

CONTENTS

The Statistical Structure of Human Settlement-Size Variation and Its Explanation.................30

Hunter-Gatherer Site Formation and Settlement-Size Variation: Five Models.........................35

Appendix I: Power-Law Scaling in Prehistoric Hunter-Gatherer Settlement Systems.................68

8

Hypothesis 3: Artifact-per-site distribution tends toward a power-law function................119

9

Case Study: A Late Archaic Period Hunter-Gatherer Settlement System in the Lake Titicaca

10

R Code for Continuous Power-Law Tail Distribution Plausibility Test.............................168

R Code for Akaike Information Criterion Analysis (Continuous Data).............................171

Appendix III: Hunter-Gatherers of the Eve of Complexity: Investigations at Soro Mik'aya Patjxa,

11

ABSTRACT

This dissertation examines human settlement-size variation through the lens of huntergatherer archaeology. Research article 1 presents an analysis of prehistoric hunter-gatherer settlement patterns from a wide range of environmental contexts and in the absence of socioeconomic complexity. Hunter-gather settlement size variation is found to exhibit heavytailed statistical structure that is consistent with the statistical structure of modern settlement-size variation, supporting claims that socioeconomic complexity is not requisite for the formation of so-called settlement-size hierarchies in human societies. Following insights from hunter-gatherer anthropology, complex systems research, and ecology, research article 2 proposes that the structure of hunter-gatherer site-size variation is an emergent property of obligate tool use among mobile hunter-gatherers. As materials are moved, modified, and deposited on the landscape, they effectively subsidize the costs of future land use at those locations, which results in additional material deposition, attracting future use, and so on. Using an agent-based model, it is demonstrated that this recursive niche-construction behavior is sufficient to generate the heavytailed property of hunter-gatherer site-size variation. The working model is then used to predict other dimensions of hunter-gatherer settlement structure related to artifact clustering and site occupation histories. Research articles 2 and 3 present test results based on Late Archaic Period

(7,000-5,000 B.P.) settlement patterns in the Lake Titicaca Basin, Peru. Good agreement is found between the predictions and empirical observations suggesting that ecological niche construction may have played a significant role in structuring hunter-gatherer mobility and land use, which in turn may have created a context for emergent settlement hierarchies.

12

INTRODUCTORY STATEMENT

Variation in population size may be among the most important drivers of human behavioral variation through time and across space (Bettencourt and West 2010; Carneiro 2000;

Ortman et al. 2014; Shennan 2001; Stiner 2001). In agricultural and industrial societies, populations tend to be non-linearly partitioned among settlements such that extremely small settlements are extremely frequent and extremely large settlements extremely rare. This *heavy-*

*tailed *population structure tends to correspond with variation in many other dimensions of human behavior including innovation, infrastructure, inequality, crime, pollution, and disease to name a few (Bettencourt, Lobo, and Strumsky 2007; Bettencourt, Lobo, Helbing, et al. 2007;

Bettencourt and West 2010; Flannery 1998; Hoch 1976; 1987; Ortman et al. 2014; Jones et al.

2008; Richerson 2013; Shennan 2000; 2001; Walmsley and Lewis 1989). Understanding how and why cities, towns, and villages come to exhibit extreme size variation can therefore offer key insights into the dynamics of many important dimensions of human behavior. Yet, scholars continue to grapple with fundamental problems related to the characterization and explanation of human settlement-size variation despite more than a century of theoretical and empirical research spanning many disciplines, many world regions, and great time depths (Auerbach 1913; Adamic

2011; Batty 2008).

Dominant behavioral models suggest that settlement-size variation is an emergent property of complex economic and political processes including food production, centralized manufacturing, political organization, and warfare (e.g., Ames 2008; Drennan and Peterson

2004; Griffin and Stanish 2007; Johnson 1977; Krugman 1996a; Stanish 2003), but recent

13 studies suggest that more fundamental ecological processes may be at work (Decker et al. 2007;

Hamilton et al. 2007). The first research article of this dissertation, presented in Appendix I, explores this possibility by analyzing the structure of settlement-size variation in prehistoric hunter-gatherer settlement systems that existed in a wide range of environmental contexts and in the unequivocal absence of socioeconomic influences from agricultural and industrial societies.

The analysis reveals that hunter-gatherer settlement-size variation exhibits the same heavy-tailed statistical structure that has been used to define settlement-size hierarchies in complex societies.

These results therefore support the idea that the so-called hierarchical structure of settlement-size variation is founded on principles that transcend complex socioeconomic processes. However, it also raises the question of what those ecological principles are and how they logically connect to settlement-size variation.

The second paper of this dissertation, found in Appendix II, presents a working model of hunter-gatherer mobility that attempts to account for the observed settlement structure among hunter-gatherer populations. The model posits that by habitually moving, altering, and depositing material resources on landscapes, hunter-gatherers construct ecological niches that attract reuse of those locations, which results in additional material deposition, and so on in recursive fashion.

Given this behavior, some hunter-gatherer sites are expected to grow in runaway fashion as they accrue culturally deposited materials over long, discontinuous spans of use. An agent-based model shows that this simple feedback behavior is sufficient to generate the heavy-tailed statistical structure of hunter-gatherer settlement-size variation. The model is then extended to

14 predict other properties of hunter-gatherer settlement patterns related to artifact clustering and site occupation span.

The predictions of this model are tested against archaeological data from the Andean highlands of Lake Titicaca Basin, Peru, 7,000-5,000 cal. B.P. The high-elevation grassland environment of the Andean altiplano is relatively homogeneous (Winterhalder and Thomas 1978) and therefore ideal for exploring how foragers construct their environments. Excavations at the site of Soro Mik'aya Patjxa (SMP), summarized in research article 3 (Appendix III), were undertaken to test the prediction that the largest settlements in a settlement system represent long occupation spans that can approach the full duration of a settlement system's existence rather than short-term aggregations. The field effort yielded over 80-thousand artifacts and thirteen clearly defined cultural features including storage pits and human burials. Radiocarbon dates reveal an occupation span of more than a millennium, from 7,900 to 6,700 cal. B.P. As the earliest securely dated open-air site in the Lake Titicaca Basin, SMP offers novel insights into the behaviors of hunter-gatherers who lived on the eve of low-level food producing economies. The sheer quantity of materials, the intensive use of informal groundstone, extreme dental attrition, the absence of evidence for investment in permanent structures or features, and radiocarbon assays converge to suggest that SMP was a residential location that was repeatedly used by residentially mobile hunter-gatherers for more than a millennium. These excavation results along with settlement-pattern analyses show agreement between the predicted and observed structural properties of the hunter-gatherer settlement system, suggesting that working niche-construction

15 model offers a plausible account of residential mobility and settlement-size variation among Late

Archaic Period hunter-gatherers of the Lake Titicaca Basin.

Critical readers are likely to observe that other models could also account for the observed variation in hunter-gatherer settlement size. It is therefore important to describe at the outset how I came to emphasize the niche-construction model of forager mobility and came to exclude other potential models. Drawing insights from hunter-gatherer anthropology, complex systems research, and ecology, I began with five initially plausible candidate models. In the interest of theoretical parsimony, these models were kept as simple as possible (sensu Fogelin

2007; Marquet et al. 2014). Three of the models are null models that explore the degree to which random processes can account for the observed variation. To be sure, the use of random behaviors in the three null models is not to suggest that hunter-gatherer behavior is random. To the contrary, their use suggests that mobility decisions are so complex and contingent as to be appropriately modeled as random. The first of the null models, which is the simplest, posits that variable residence times emerge from residential moves that are randomly distributed through time. They are independent of the decisions of other foragers and the timing of other residential moves. The second two null models relax the constraints of the first model. The *social-bias *

*model* imagines that forager residential moves are biased by the locations of other foragers. The

*niche-construction model* imagines that forager residential moves are biased by previously deposited cultural materials. In each model, foragers iteratively occupy the locations of other foragers or previously deposited cultural materials that are randomly selected from the respective populations.

16

Two deterministic models are also given consideration. The first posits that huntergatherer settlement-size variation reflects ecological patch structure. More productive ecological patches simply draw greater occupation intensity and therefore generate larger archaeological sites. The second posits that hunter-gatherer site-size reflects topographic structure that results from the geomorphic processes. Larger sites simply reflect topographic contexts that have remained relatively resistant to erosion and burial.

The justifications for each model along with preliminary assessments of their efficacy are detailed below in the Literature Review. In short, the random residential move model fails to predict the heavy-tailed structure observed in the empirical data. The social-bias model fails to predict the long occupation spans observed among hunter-gatherer sites. The ecological patch model fails to predict settlement-size variation in homogenous environments. And, finally, the geomorphic process model fails to predict spatial associations between geologic structure and artifact distributions. Only the niche-construction model survived this preliminary testing, therefore justifying further attention. The preliminary considerations are by no means exhaustive, nor is the list of candidate models, but I argue that the approach offers a parsimonious starting point toward explaining an important dimension of human behavior that, to my knowledge, has not seen previous attention in hunter-gatherer archaeology.

To the extent that the working niche-construction model of forager mobility accounts for hunter-gatherer settlement-size variation, the research presented here holds implications for our understanding of hunter-gatherer mobility and the emergence of socioeconomic complexity in human societies. Research to date on hunter-gatherer mobility has tended to emphasize the role

17 of natural-environment structure in predicting land-use patterns. Such studies have shed considerable light on hunter-gatherer behavioral variation, but the findings presented here suggest that the structure of the natural environment is limited in its ability to account for settlement-size variation within hunter-gatherer settlement systems. The working model instead suggests that foragers played a non-trivial role in constructing the landscapes through which they move. It is suggested that such niche-construction behaviors—with an emphasis on those associated with alterations to the abiotic environment—require more-explicit consideration in hunter-gatherer mobility theory.

An emergent property of the niche-construction model of forager mobility is heavy-tailed statistical structure in settlement-size variation. This structural property holds potential implications for our understanding of how similar statistical structure emerges in the settlementsize variation of agricultural and industrial societies. The statistics are more than mere triviality.

They describe the extreme nature of settlement-size variation in human settlement systems including extremely large settlements that are often orders of magnitude larger than most other settlements in the system. On one hand, such large population centers are engines of cultural innovation and manufacturing. On the other hand, they can be loci of relatively high disease, crime, and inequality rates. Understanding the dynamics that drive superlinear growth in human settlement size therefore contributes to our understanding of these many other important dimensions of human behavior. Previous research would seem to suggest that the extreme settlement-size variation is an emergent property of complex socioeconomic processes. However, the research presented here suggests that the opposite may be true. Given population growth, the

18 posited mobility could be expected to result in decreasing inter-occupation intervals and increasing co-resident population sizes among settlements. With sufficiently large overall populations, settlement population sizes would mirror the statistical structure of cultural-material distributions amongst sites. The quantitative increases in co-resident population sizes would have necessitated qualitative changes in social and economic organization (Carneiro 2000). Thus, the structure of hunter-gatherer site-size variation within hunter-gatherer settlement systems could have created a groundplan for the emergence of hierarchical socioeconomic structure in complex societies.

LITERATURE REVIEW

In order to contextualize the behavioral significance of settlement-size variation in human societies, this literature review begins with a survey of research that relates settlement size to other human behaviors. I then describe current statistical methods for characterizing and interpreting settlement-size variation. Statistical models are reviewed along with the generative mechanisms behind them. I then explore how scholars have previously characterized and explained human settlement-size variation, concluding that we currently lack a firm understanding of settlement-size structure among hunter-gatherer societies. Finally, because

Archaic Period hunter-gatherer societies of the Lake Titicaca Basin serve as the major case studies that inform the conclusions found here, I present an overview of relevant archaeological research in the region.

19

**Human Behavior and Settlement-Size Variation**

Despite the difficulties in predicting human behavior, some behaviors show strikingly regular relationships with population, or settlement size. Some of these behaviors, especially those that reflect basic human needs, simply vary isometrically with population. That is, they are proportionate to population. For example, settlement population size linearly predicts housing and employment in several modern nations (Bettencourt, Lobo, Helbing, et al. 2007). However, other human behaviors vary allometrically with settlement size. That is, they scale non-linearly with population. Thus while population size can directly account for much of the variation in such behavioral traits, other population-related factors affect the scale of the outcomes. A notable example is found in cultural innovation. Scholars have demonstrated that cultural innovation rates increase nonlinearly with increases in human population. Stephen Shennan (2001:15; see also Richerson 2013; Shennan 2000) suggests that “...demographic fluctuations may be the single most important factor in explaining how and why the emergence of modern human culture occurred when it did.” Increases in human population are also thought to have played a key role in major human economic transformations such as major dietary shifts including the origins of agriculture (Stiner 2001). Although such studies refer to population broadly across evolutionary time and continental space, the same processes are at work among sub-populations. Among modern cities, cultural innovation as measured by patent rates and numbers of inventors has been shown to increase superlinearly with urban population size (Bettencourt, Lobo, and Strumsky

2007). Communities with larger populations tend to generate proportionately more innovation than do settlements with smaller populations. Other modern human behaviors that covary

20 superlinearly with settlement size include wages, cost-of-living, gross-domestic product, electrical consumption, disease transmission, crime, pollution, communication, walking speeds, and road-network accessibility (Batty 2013; Bettencourt et al. 2010; Hoch 1972; 1976; Jones et al. 2008; Schläpfer et al. 2014; Walmsley and Lewis 1989). Sublinear relationships, or those that proportionately decrease with settlement population, tend to include infrastructural features such as road surface, length of electrical cables, number of gas stations, and number of social institutions (Bettencourt, Lobo, Helbing, et al. 2007; Carneiro 1987; 2000).

These examples reveal the predictable nature of relationships between settlement size and many important dimensions of human behavior. Luis Bettencourt and Geofferey West (2010:912) suggest that “...size is the major determinant of most characteristics of a city; history, geography and design have secondary roles.” To be sure, this is not to suggest that population size *causes* the behaviors of interest. Rather, population serves as a particularly useful proxy for the intensity of human and human-environment interactions that come with costs and benefits to the inhabitants of human settlements (Bettencourt et al. 2010:6). Nonetheless, if we can better understand what drives settlement-size variation in human societies, we will have made significant strides toward understanding the dynamics behind many other dimensions of human behavior.

**Univariate Statistical Structure and Its Interpretation**

Many real-world behaviors are best predicted by their previous state as opposed to independent variables (Levin 2003). When this is the case—when some some variable of interest depends largely on its previous state—distinct statistical properties can be captured in the

21 structure of a single variable. This section attempts to explain how we can detect and interpret such statistical properties. Settlement-size variation in human societies is the variable of interest.

Settlement-size variation in agricultural and industrialized societies is extreme. Yet, this extreme variation is remarkably regular in statistical structure. Scholars seem to agree that *power-law* and

*lognormal* distributions can serve as plausible and parsimonious statistical descriptions for human settlement-size variation (Clauset et al. 2009:26; Decker et al. 2007; Mitzenmacher

2001). Power-law functions are those in which “...the probability of measuring a particular value of some quantity varies inversely as a power of that value” (Newman 2005:323), or

*p*

( *x*)∝*x*

−α

, where *x* is some quantity and alpha, α, is a scaling parameter that usually falls in the range of

2 < α < 3. Related statistical functions include Lévy, Mandelbrot (fractal dimension), Pareto,

Simon, Yule, Zeta, and Zipf functions. In archaeology, power laws have been referred to as the rank-size rule. It should also be noted that archeological literature has tended to confuse powerlaw and lognormal distributions (Griffin 2011:882). Lognormal functions are those in which the exponentials are normally distributed. Put conversely, logarithmically transformed normal distributions describe lognormal distributions.

*p*

( *x*)∝ 1 exp

[−(ln *x*−μ )

2

σ 2

2

] , where μ is the log-mean and σ is the log-standard deviation.

Both power-law and lognormal distributions share the property of *heavy-tailed* structure such that small occurrences—in this case, settlements—are very frequent while extremely large

22 occurrences are extremely rare. In other words, heavy-tailed distributions are those in which

“with overwhelming probability,” the values of few occurrence contribute significantly to the sum of values for all occurrences while the values of most occurrences contribute negligibly to the sum (Sornette 2006:79). For example, we could say that the New York City metropolitan area with approximately 23-million people contributes significantly to the total U.S. population of

314-million people while most other U.S. population centers contribute negligibly to the total.

In practice, it is often difficult to confidently distinguish among heavy-tailed statistical forms, especially when sample sizes are small (Clauset et al. 2009) as is often the case in archaeological research. The more generic term of heavy-tailed distribution is preferred in such cases. Although discrimination among heavy-tailed distributions can be difficult, heavy-tailed distributions are readily distinguished from more-familiar normal (i.e., Gaussian) and exponential distributions, which are exponentially bounded in their range. The largest and smallest observations in such numeric sets rarely deviate by more than a factor or two from their means (Clauset et al. 2009).

Interest in identifying heavy-tailed statistical structure is more than statistical triviality.

Many scholars argue that the distinction is imminently important for substantive research. For example, Aaron Clauset (2009:1) states,

*Not all distributions fit [a Gaussian] pattern, ...and while those that do not are often considered problematic or defective for just that reason, they are at the same time some of the most interesting of all scientific observations. The fact that they cannot be characterized as simply as other measurements is often a sign of complex underlying processes that merit further study.*

23

Clauset's quote (see also Barabási 2005; Bentley and Maschner 2008; Brown et al. 2005;

Mandelbrot 1982) point to two advantages to exploring heavy-tailed statistical structures in empirical phenomena. On the practical side, such analyses can furnish considerable predictive power. I illustrate this by way of a hypothetical archaeological scenario. Imagine a researcher who has designed an archaeological settlement survey of 40 km

2

. In the pilot phase of the project, the researcher initially surveys a random sample of 10 km 2 . In doing so, they document

10 residential sites that belong to a prehistoric settlement system of interest. The architecture and geology are such that house foundations are well-preserved and visible on the surface. House counts for each of the ten sites are as follows: 1, 1, 3, 2, 1, 3, 1, 2, 1, and 1. The researcher summarizes the house-count data with a mean and standard deviation in their summary report.

The average is 2 and the standard deviation is 1 house per site. This does not seem an unreasonable characterization of the data given that the full range of variation in the numeric set is captured by the mean plus one standard deviation.

The researcher is now ready to put these pilot data to predictive work. Given that 10 sites were found in the 10-km

2

pilot area and lacking additional information, they reasonably estimate finding 30 additional sites with residential architecture in the remaining 30 km 2 . Given the assumption of normality, these 30 new sites are most-likely to produce 2 houses each for a total of 60 houses. To get a sense of the error in this estimate, a simple simulation is used to derive

likely lower and upper house counts a . The researcher generates 10,000 simulated surveys each

a R code: houses<-c() #creates an empty set of house counts to be populated.

for (i in 1:10000){ #indicates that the following code within the brackets should be iterated 10,000 times.

x<-rnorm(30,2,1) #30 house-count values are randomly drawn from a normal distribution with a mean of 2 and sd of 1.

h<-c(h,sum(x))} #the house counts are summed for the 30 sites and appended to the set of previously derived

24 consisting of 30 residential sites. The house counts for each site are drawn at random from a normal distribution with a mean of 2 and standard deviation of

1 house per site. The simulation produces a mean of 60 houses with a standard deviation of 5. Thus, the researcher concludes that there is a 95-percent probability of finding between 50 and 70 houses in the

remaining survey area (Figure 1). The field strategy is

designed and resources are allocated accordingly.

Soon into field recording, it is found that some sites have much greater house counts than expected.

For example, 10 houses were located at the second site recorded, which is more than eight standard deviations greater than the inferred mean. The researcher chalks

*Figure 1: Simulated survey results in a hypothetical predictive model using the assumption of normality. Each simulation locates 30 sites and sums the house counts, which are drawn from a normal distribution with a mean of 2 houses per site and a sd of *

*1. The simulation is repeated 10,000 times to examine the probability of deriving various total house counts. *

*95 percent of the counts fall between *

*50 and 70 houses.*

this up to anomaly and continues recording. By the end of the survey, however, the supposed anomalies add up. The 60 newly recorded sites produced

140 houses—double the predicted upper limit. Unfortunately, the error in prediction resulted in incomplete site documentation and reallocation of research resources, which detracted from other areas of research. What went wrong?

house counts.

25

The numbers presented above were not simply pulled from thin air. They were drawn

from a power-law probability function with a scaling parameter of 2 and a minimum of 1 b ,

precisely the type of distribution that has been proposed to describe human settlement-size variation. In a parallel universe, the researcher determined that the pilot data are better characterized by a power-law function. Running a simulation with the assumption of a powerlaw probability distribution, they found a 95 percent chance that the total house count for 30 sites would be greater than or equal to 70 houses, which was the upper value of the previous estimate

under the assumption of a normal distribution c . Moreover, they estimated a 51-percent chance

that the sum would be more than double the most likely value predicted under the assumption of normality, which was 60 houses. With such information, the observed 140 houses was not a surprise to the researcher, and well-informed decisions were made for time and resource allocation.

Of course there are other statistical functions besides normal and power-law functions that could also be considered when the statistical structure of the data is unknown. Deciding which merit attention can depend on a number of empirical observations and theoretical insights, b R code for generating random numbers from a power-law function following the transformation method

(Newman 2005) rpl<-function(n,alpha, xmin,t='continuous'){ #creates the random power-law function (rpl) with parameter specifications

r<-runif(n,0,1) #draws n values at random from a uniform probability distribution between 0 and 1.

x<-xmin*(1-r)^(-1/(alpha-1)) #transformation applied to the random numbers

if(t=='discrete'){x<-as.integer(x)} #Converts the data to integer data if specified. Otherwise the data are continuous.

list(n=n,alpha=alpha,xmin=xmin,type=t,x=x)}#Reports the results.

c R code: h<-c() #creates an empty set of house counts to be populated.

for (i in 1:10000){ #indicates that the following code within the brackets should be iterated 10,000 times.

x<-rpl(30,2,1,t='discrete')$x #30 house-count values are randomly drawn from a power-law distribution with a scaling paramer of 2 and a minimum value of 1.

h<-c(h,sum(x))} #the house counts are summed for the 30 sites and appended to the set of previously derived house counts.

26 but assuming that some variables of interest could potentially exhibit heavy-tailed statistical structure, there is much predictive power to be gained by exploring the plausibility of relevant models.

Clauset's quote presented above also alludes to more theoretically substantive gains from the consideration of heavy-tailed statistical structure in empirical data. Statistical probability distributions result from generic underlying behaviors that are often well understood by statisticians and statistical physicists. These generic processes can be used to guide the construction and evaluation of more proximate models—human behavioral models in the case of this research.

Normal distributions tend to emerge when many independent events converge to generate variation around a typical value in a population. This is referred to as *Gaussian process*. Human height serves as a classic example of a Gaussian phenomenon that generates a normal distribution. A Centers for Disease

Control and Prevention data sample of the heights of

6,219 U.S. adult individuals (age > 18) shows the signature bell curve of a normal distribution when

plotted as a histogram (Figure 2). Mean height is 167

cm and the standard deviation is 10 cm. This distribution suggests that variation in U.S. adult height

*Figure 2: U.S. height distribution based on a Centers for Disease *

*Control and Prevention sample of *

*6,219 individuals all over 18 years of age taken between 2013 and 2014. *

*The distribution illustrates a classic normal distribution, suggesting that *

*Gaussian process can account for human-height variation.*

is the result of many independent factors that contribute to any one individual's height. This conclusion would be reasonable given our understanding that an individual's adult height is the

27 product of a complex array of genetic and environmental factors.

Poisson and exponential/geometric distributions are also linked to generic mechanisms that are well understood and referred to as *Poisson processes *(Barabási 2005). These are processes that entail a constant probability of occurrence for a given event. Poisson models are used with great success in modeling the frequencies and inter-interval times of many human behaviors such as traffic flows, accident frequencies, telephone call frequencies, and radioactive decay. If the probability of occurrences is constant, the number of occurrences per unit of observation will follow a Poisson distribution. The inter-event, or waiting times will tend toward an exponential distribution if the quantity being measured is a continuous variable or a geometric distribution if the quantity being measured is a discrete (i.e., integer) variable (Johnson et al.

2005).

Accordingly, if some archaeological data set reveals a Poisson, exponential, or geometric distribution, we can infer a Poisson process, which should give us a clue toward developing and evaluating candidate behavioral models. If a candidate behavioral model does not readily reveal an element of Poisson process, then we might consider it unlikely to account for the observed variation. To exemplify this logic by way of archaeological scenario, imagine that a forager band of more-or-less constant size revisits a rockshelter regularly throughout its collective existence of hundreds or thousands of years. Natural-resource configurations generally determine the rate at which the group uses the shelter in a given year, decade, or century, and myriad environmental

28 and social contingencies account for the variation in the number of visits in any specific span of time. Given this model, the number of visits per century could be expected to follow a Poisson probability distribution, and inter-interval occupation times would be expected to follow an exponential distribution. Although these hypotheses cannot be tested directly in this archaeological example, radiocarbon dates taken from annual materials combusted in hearths could be used to test the hypothesis of Poisson process assuming reasonably unbiased preservation. If the inter-interval distribution of radiocarbon dates reveals exponential structure, we could assume Poisson process and conclude that the distribution of visitation of the shelter was effectively random. Otherwise, we could use the absence of Poisson structure to rule out behavioral models that have Poisson processes built in.

Gaussian and Poisson processes cannot readily explain the extreme variation found in heavy-tailed distributions. For heavy-tailed lognormal distributions, the generative mechanism goes by the names of multiplicative process or law of proportionate effect (Mitzenmacher

2001:235–237). In multiplicative processes, the proportion of growth or fragmentation of some entity is independent of its current size, but the absolute growth is dependent. Imagine, for example an organism that weighs 20 g and another that weighs 10 g. Both grow by 10 percent of their current mass resulting in 2 and 1 g of growth, respectively. Even though the proportion of growth is the same and therefore independent of body mass, the total growth of the first organism is twice that of the second and therefore dependent on body mass. If some set of values iteratively experiences growth or fragmentation by some random fraction of its current size, a lognormal distribution emerges. Fission-fusion cycling in human population dynamics provides

an example of a multiplicative process model that has seen anthropological application (Griffin

2011; Tuncay 2008).

29

There is considerably more debate as to the types of mechanisms that can and cannot generate power-law distributions (Mitzenmacher 2001; Newman 2005). Models that have been proposed—many of which are synonymous or closely related—include preferential attachment, self-organized criticality, Yule process, combination of exponentials, inverses of quantities, correlated random walks, percolation, highly optimized tolerance, random extremal process, and task queuing (Barabási 2005; Mitzenmacher 2001; Newman 2005). Definition and review of each of these processes is beyond the scope of this work, and the interested reader is directed to the citations above for more information. This discussion focuses on the process of *preferential *

*attachment* because it it posited to be of particular relevance to the problem at hand. Preferential attachment also goes by the names of Pareto process, cumulative advantage, the Gibrat principle, the rich-get-richer phenomenon, Yule process, and the Matthew effect. Such processes are simply those in which growth predicts growth. Wealth and scholarly citations are two common examples of preferential attachment. The more wealth one has the more wealth one is likely to generate.

The more a scholarly paper is cited, the more citations that paper is likely to attract. Such recursive processes generate power-law structure in many empirical phenomena. Conversely, if we identify power-law structure in empirical data, it is possible—though not guaranteed—that preferential attachment figures into the behavioral mechanism that generated the pattern. In the last section of this review, I use the preferential attachment concept to inform two models of hunter-gatherer site formation.

30

*The Statistical Structure of Human Settlement-Size Variation and Its Explanation*

Scholars have given variable consideration to statistical models in their characterization and interpretation of human settlement-size variation, though this has generally come from nonarchaeological sciences (Bentley and Maschner 2008:248). At least for modern nations, most recent research seems to indicate that lognormal statistical distributions describe the bulk of settlement-size variation, but the upper tail of the distribution, which includes the largest settlements (i.e., cities), tends to deviate significantly from lognormality. These values are better characterized by a power-law function whether settlement size is measured by city population, metropolitan-area population, or areal extent (Batty 2008; Clauset et al. 2009; Cristelli et al.

2012; Decker et al. 2007; Eeckhout 2004; Gabaix 1999a; González-Val 2010). Earlier studies, though founded on statistical methods that have since been found to be problematic (Clauset et al. 2009; Edwards et al. 2007; White et al. 2008), have suggested that various combinations of power-law and lognormal models also apply to other times and nations as well (Berry 1961;

Fujita et al. 1999; Gabaix 1999a; Krugman 1996a; Zipf 1949). Archaeological research appears to bolster this claim with observations of apparent heavy-tailed site-size scaling in a variety of cultural and temporal contexts, all of which represent sedentary agricultural and industrial societies (Blanton 1976; Drennan and Dai 2010; Drennan and Peterson 2004; Hodder 1979;

Inomata and Aoyama 1996; Johnson 1977, 1980; Laxton and Cavanagh 1995; Ortman et al.

2014; Paynter 1982; Pearson 1980; Peterson and Drennan 2005; Savage 1997). However, few of these studies have explicitly tested for heavy-tailed structure (but see Brown and Witschey

2003).

Satisfactory behavioral models to account for this aspect of human demography remain elusive. The problem persists even a century after Felix Auebarch (1913) first grappled with it

31

(Adamic 2011; Batty 2008). For example, economic geographer and Nobel laureate in economic sciences, Paul Krugman, attempted to explain this heavy-tailed statistical structure only to conclude that in city-size distributions, we are faced with “...one of the most overwhelming empirical regularities in economic science...with no good theory to account for it” (Krugman

1996b; see also Krugman 1996a). Dominant thinking in the latter half of the last century has tended to view this structure through the lens of complex socioeconomic dynamics including central-place economics and political administration sensu Christaller (1966). In this view, settlement-size variation is seen as an emergent property of (1) agricultural subsistence, which tends to disperse people across landscapes, (2) craft manufacturing and service industries, which simultaneously tend to nucleate populations, and/or (3) political organization and warfare, which drives fission-fusion cycling and political hierarchies among populations (Ames 2008;

Beckmann 1958; Beckmann and McPherson 1970; Drennan and Peterson 2004; Eeckhout 2004;

Flannery 1998; Fujita et al. 1999; Gabaix 1999b; Griffin and Stanish 2007; Inomata and Aoyama

1996; Johnson 1977; Krugman 1996a; Parr 1969; Steponaitis 1981). Such thinking has led to the use of the term settlement-size hierarchy to describe otherwise continuous variation in settlement size.

Without doubt, such dynamics contribute proximately to structural properties of human settlement systems. However, recent research following largely from insights in complex systems research (sensu Bentley and Maschner 2008; Lansing 2003; Mitchell 2009) seems to

32 suggest more fundamental ecological processes driving the structural properties of human settlement-size variation. Ethan Decker et al. (2007) suggest that the structure of modern settlement-size variation may be an emergent property of human interaction networks within which people and resources flow (see also Fujita et al. 1999). Additional support for this idea is found in studies of ethnographic hunter-gatherer settlement systems. Because hunter-gatherer economies tend to lack agricultural production, specialized manufacturing, hierarchical political structure, and warfare (Kelly 1995), they present ideal case studies for exploring how individuals distribute themselves among settlements in the absence of complicated socioeconomic structure.

Marcus Hamilton and colleagues (2007) found that the frequency distributions of hunter-gatherer group sizes exhibit heavy-tailed statistical structure and, like Decker et al., they suggest that the statistical property may be an emergent property of socioeconomic optimization in humanecological networks:

*[Hunter-gatherer] human social systems are hypothesized to reflect optimized networks of flows of essential commodities: food, other material resources, genes and culturally transmitted information. Individual foragers should maximize fitness by participating in social networks of exchanges that optimize the flow of resources. *

*However, in density-dependent populations, individuals face tradeoffs between resource availability and competition from conspecifics, leading to optimization principles acting to regulate interactions and therefore network organization. It is these density-dependent tradeoffs that lead to the complex hierarchical structures we*

*report here *(Hamilton et al. 2007:2199–2200)*.*

In addition, Clifford Brown and colleagues (2006) and Matthew Grove (2010) have shown that the duration of settlement occupation among !Kung hunter-gatherers is heavy tailed such that short occupation spans are extremely frequent and extremely long occupation spans extremely rare. Either or both of these proximate behaviors—differential group formation or site-

33 occupation duration—could conceivably generate variation in the sizes of human settlements as detected archaeologically (Surovell 2009). This follows from the intuitive premise that larger group sizes and longer occupation spans should generate larger archaeological footprints—a premise supported by John Yellen's (1977) ethnoarchaeological work showing that group size

and occupation span generally tend to predict the size of !Kung logistic-camp sites (Figure 3).

The observation of nonlinear scaling in ethnographic hunter-gatherer group size and siteoccupation duration may offer insights into the behavioral basis for settlement-size variation in hunter-gatherer and complex societies alike. It may be that heavy-tailed statistical structure in group formation and/or site occupation span created a framework onto which hierarchical settlement structure would eventually be built in more complex societies.

While hunter-gatherer ethnography provides important observations relevant to the problem at hand, scholars have suggested that we exercise caution in assuming that ethnographic foraging societies were economically independent of more complex societies (Wobst 1978).

“[L]ong before anthropologists arrived on the scene, hunter-gatherers had already been contacted, given diseases, shot at, traded with, employed and exploited by colonial powers, agriculturalists, and/ or pastoralists. The result has been dramatic alterations in hunter-gatherers' livelihoods” (Kelly 1995:25). The possibility therefore remains that the observed variation in ethnographic hunter-gatherer settlement sizes was influenced by social and economic articulation with agropastoral and industrial societies.

The hunter-gatherer archaeological record offers us the only empirical solution to exploring the structural properties of hunter-gatherer settlement systems in the unambiguous

*Figure 3: Settlement size as predicted by occupation span in !Kung logistical camps. *

*Data from Yellen (1977).*

34

35 absence of influences from more economically complex societies. To be sure, archaeological data are fraught with their own limitations, most notably the persistent problem of equifinality— the idea that multiple behaviors can generate a given material outcome (Lyman 2004).

Nonetheless, archaeological settlement data offers an independent and complementary line of evidence to the ethnographic record of hunter-gatherer settlement systems.

Settlement-pattern analysis is common practice in hunter-gatherer archaeological research, but little attention has been paid to the characterization and interpretation of huntergatherer site-size variation (c.f. Bettinger 1981, 1991; Thomas 1972, 1973, 1974, 1983, 1988;

Thomas and Bettinger 1976; Zeanah 2002). The omission is somewhat surprising given that settlement-size analysis is common practice for prehistoric agricultural and state settlement systems (Drennan and Peterson 2004). Whatever the case, archaeological settlement-size analysis is required to resolve the question of heavy-tailed settlement-size structure in huntergatherer settlement systems. This is the subject of research article 1 presented in Appendix I.

**Hunter-Gatherer Site Formation and Settlement-Size Variation: Five Models**

How can we best explain the heavy-tailed statistical structure of hunter-gatherer settlement-size variation? In this section, I draw from hunter-gatherer, complex-systems, and ecological theory to identify a set of candidate models with potential for explaining the empirical phenomenon. In the interest of theoretical parsimony, I begin by considering three *null models*,* * or models that rely on random processes to generate patterns of interest. Scholars have observed that random processes can account for variation in such diverse phenomena as the frequency of biological taxa, forest fire size, earthquake magnitudes, and wealth (Newman 2005). An

36 example of a particularly compelling null model can be found in molecular genetics, where scholars have argued that predictable levels of variation in the frequencies of genetic alleles can emerge simply through random mutation and genetic drift (Kimura 1968, 1985). While natural selection can drive the dominance of certain alleles in populations, the null model suggests the possibility that alleles can also come to dominate a population simply through stochastic effects that emerge from repeated transmission of genetic material between generations. Because some allele frequencies can be explained without recourse to Darwinian selection—i.e., they are adaptively neutral alleles, such models are considered a special case of null models termed

*neutral models*.

Anthropologists have found considerable utility in applying null models to understanding cultural variation. Steven Lansing and colleagues (2008) tested a neutral model of social dominance among Indonesian patrilines. Analysis of Y-chromosome DNA reveals genetic variation that is statistically indistinguishable from neutral model predictions in 88% of the 41 communities investigated. In other words, it appears that social dominance is not stable within

Indonesian patrilines but instead drifts among them over time. Frasier Neiman (1995) shows that the frequencies of stylistic variation observed in prehistoric ceramic traditions of the Eastern

Woodlands of North America may be understood as a result of neutral processes of cultural transmission. Similarly, Alexander Bentley and colleagues (Bentley et al. 2004, 2007; Bentley and Shennan 2005; Hahn and Bentley 2003; Herzog et al. 2004) show that null models of cultural transmission may account for frequency variation, or relative popularity, in cultural traits such smoking, baby names, ceramic designs, music, and dog breeds. To provide an example of

37 specific relevance to hunter-gatherer behavior, Jeffrey Brantingham (2003) shows that random walks can account for lithic raw material diversity in archaeological assemblages. He also shows that morphological variation in material assemblages follows from various combinations of stochastic and directed processes (Brantingham 2007; Brantingham and Perreault 2010).

Following the lead of such studies, the first three models considered here are null models of forager mobility. Each of these models considers a version of random mobility under different sets of behavioral constraints that are informed by hunter-gatherer ethnography. For the purpose of this discussion, I term these null models the *independent residential move*,* social-bias*, and

*niche-construction* models of hunter-gatherer mobility.

Although it is critical to consider null explanatory models for empirical behaviors, hunter-gatherer anthropology demonstrates that deterministic models require careful consideration as well. It is entirely possible that externalities could have shaped the observed settlement-size variation. I explore two such models that offer initially plausible mechanisms for generating heavy-tailed statistical structure in archaeological site-size variation. I refer to these models as the *environmental structure* and *geomorphic process* models for the purpose of this discussion. The mechanics of and theoretical basis for each of the five models is discussed in turn, and each is given preliminary evaluation in light of archaeological observations.

*Independent Residential Move Model*

The independent residential move model is the simplest model explored here. It posits that the timing of *n *residential moves is randomly distributed over time span *t*. Each residential move is independent of previous moves, independent of the locations of other foragers, and the

38 result of a complex set of factors that are effectively random from our point-of-view (see Bentley and Shennan 2003:462–463 for a similar null model). The time that elapses between each residential move therefore defines the occupation duration for each site. For example, if a forager commences occupation of site A at year 1.10 and leaves to occupy site B at year 1.50, then the occupation span of site A is simply 1.50-1.10=0.40 years. This model describes a Poisson process, which results in an exponential distribution of inter-interval times (Barabási 2005)—in this case, an exponential distribution of site-occupation times.

The process and outcome can be illustrated with a simple simulation d . Imagine a forager

who moves their residence 30 times over a 10-year period. Drawing move times at random from a uniform distribution between 0.00 and 10.00 years gives residential moves at the following years: 0.27, 0.48, 0.51, 0.78, 1.25, 1.34, 1.52, 1.90, 2.35, 2.71, 2.93, 3.15, 3.66, 4.09, 4.23, 4.23,

4.39, 4.64, 5.40, 6.50, 6.51, 6.60, 6.77, 6.92, 8.14, 8.19, 8.31, 8.63, 8.66, and 9.73 years. Taking the elapsed time between each sequential pair of residential move times generates the residence

times summarized in the histogram in Figure 4. The result reveals site occupation times with

right skewed exponential structure and values ranging from 0.01 to 1.22 years. Exponential structure is obtained regardless of the number of foragers involved, the time frame examined, or the number of residential moves within that time frame. We would therefore expect to find an d R code used to generate synthetic site occupation times for a forager whose residential moves conform to a

Poisson process is as follows: pf<-function(n){ #n is the number of residential moves and hence the number of archaeological sites created.

m<-sort(runif(n,0,1),decreasing=F) #generates a random uniform distribution of residential move times, *m*, between 0 and 1 sorted in ascending order.

t<-m[2:n]-m[1:n-1] #Finds the time, *t*, elapsed between each residential move, *m*, and the prior move.

list(m=m, t=t)} #Lists the results.

39

*Figure 4. Sample residence time data generated from a simple Poisson-process model of *

*residential mobility. The model posits that *n* residential moves are randomly distributed *

*throughout a given time span. The example shown here shows inter-move intervals for 30 residential moves randomly spread out over a 10-year period. The result is an exponential distribution of residence times. *

exponential distribution in archaeological measures of occupation duration if hunter-gatherer residential moves are marked by “random,” independent decisions.

While this simple random model predicts a fair degree of variation in residence times, the exponential structure and its discrete-data analog—geometric structure—are highly unlikely to explain variation in the archaeological proxies of occupation intensity, which instead tend to reveal heavy-tailed statistical structure (see Appendix I). Consideration of alternative models is therefore required.

*Social-Bias Model*

The social-bias model is a null model that adds the constraint of partial dependence on the locations of other foragers. Ethnographers frequently observed that residential movement

40 among mobile foragers served not only subsistence interests but also social interests. For example, Julian Steward (1955; 1970) remarked that Great Basin Shoshone “people were always eager to associate with one another...” for various reasons including communal hunts, dancing, gambling, healing, and visiting. Similarly, John Yellen's (1977) ethnographic observations hint at the important role that social interaction played in structuring the mobility decisions of !Kung foragers. While camp variation was “...determined primarily by environmental variables...”, the occupation of Dobe—the largest camp by far—was “...based largely on social factors: desire to visit other !Kung, pick up or leave behind individuals or families, visit Bantu cattle trading posts, or check in at the anthropologist's camp.” Similar observations, often couched in terms of fission-fusion dynamics, are also evident in many other ethnographic accounts (e.g., Damas

1969; 1972; Leacock 1969; Rogers 1969; 1972; Slobodin 1969; Silberbauer 1972; Turnbull

1968; Woodburn 1968; Yellen 1977). More recently, Hamilton and colleagues' (2007) extensive analysis of data from 339 ethnographic hunter-gatherer societies observed that forager groups consisted of six hierarchically organized divisions that were, at least in part, related to periodic population aggregations. The diverse social-interaction goals that have been evident among ethnographic foragers were certainly important among prehistoric foragers as well (Gamble

1999; MacDonald and Hewlett 1999; Stiner and Kuhn 2006; Whallon 2006; Wobst 1974).

It is therefore reasonable to consider a model in which the residential moves of huntergatherer populations served not only to connect foragers to natural resources but also to one another. In the null version of such a model, foragers simply choose a member of the population at random and reside at their location. Of course, it is logically impossible that all residential

moves are determined by the locations of other individuals because such behavior would ultimately result in all individuals of the population residing permanently in a single location—

41 an unsustainable and unrealistic scenario. Instead, we imagine that each of *n* foragers moves to the location of a random individual with probability, *s*, and to a random location on the landscape with probability 1 - *s*. The latter behavior effectively models situations in which social interaction interests are outweighed by social conflict and/or the draw of natural resources.

Importantly, this model generates a feedback loop in site occupation because locations with greater numbers of individuals will tend to attract the occupation of more individuals leading to growth in recursive fashion—an example of emergent preferential attachment behavior. Socialbias behavior is therefore expected to generate occasional settlements with extremely large aggregations of individuals relative to other settlements.

The model behavior and outcomes can be illustrated by way of a simulation example.

Imagine 25 foragers/forager groups initially occupy distinct random locations on a landscape. At each time step, there is a 95-percent chance that each will move to the location of another forager chosen at random from the population and a 5 percent chance that they will move to a novel location on the landscape. The behavior is iterated many times, and the person-time accumulated

at any given location is tallied. Figure 5 summarizes the results of a simulation using these

parameters after the creation of 500 settlements. The simple random behavior generates considerable variation in settlement sizes with person-time of occupation ranging from 1 to 731.

Importantly, the statistical structure of the distribution is extremely right skewed, or heavy tailed.

If “random” social bias accounts for settlement-size variation, we would therefore expect to find

42

*Figure 5. Settlement-size results from a simulation in which the residential moves of virtual foragers are biased by residential locations of other foragers. In this example, 25 foragers move to the location of a random forager with 95-percent probability and to a random location on the landscape with 5-percent probability. The person-time at each of 500 sites are summarized in this histogram.*

heavy-tailed statistical structure in archaeological indices of settlement occupation intensity. The social-bias model therefore appears to be initially plausible.

*Niche-Construction Model*

I now turn to a third null model that implements a slightly different behavioral constraint but also generates preferential attachment behavior. In addition to social motivations for mobility, hunter-gatherer ethnography also suggests that cultural materials can play an important role in the mobility decisions of mobile foragers. In *Archaeology of Place*, Lewis Binford (1982) suggested that “cultural geographies” play an important role in shaping how hunter-gatherers use landscapes. Binford observed that Nunamiut logistical camps created at time 1 could become residential loci at time 2 and vice versa. Although Binford did not explicitly propose an

43 explanation for why foragers should reuse locations, he presumably assumed that the materials and infrastructure deposited at a settlement drew foragers to reoccupy locations when in pursuit of natural resources in the vicinity.

Insights from ecological research give theoretical traction to Binford's cultural geography concept. Ecologists have increasingly come to emphasize that organisms shape the ecological niches to which they adapt and this includes the abiotic structure of natural environments (Laland et al. 2007; Odling-Smee et al. 1996; 2003; 2013). Hunter-gatherers alter their ecological niches whenever they move, alter, and deposit materials on the landscape. Assuming that such environmental modifications can help to subsidize the costs of later land-use practices, there may be an evolutionary fitness advantage to reusing locations. These ethnographic and ecological observations justify a model of forager mobility that is biased by culturally deposited materials. I therefore present a null version of such a model, which, for the purpose of this discussion, I term the *niche-construction model of forager mobility*. The model assumes that each residential move decision is, in part, influenced by the locations of previously deposited cultural materials or infrastructure. With probability *m*, a forager moves to the location of a previously deposited material chosen at random from the population of previously deposited materials, and with probability 1 - *m*, the forager moves to a random location on the landscape. The latter behavior effectively simulates decisions where culturally deposited materials are either irrelevant to a task at hand or outweighed by the pull of natural resources.

Similar to the social-bias model discussed previously, the material bias model is expected to generate feedback behavior in the deposition of cultural materials. As more materials are

deposited at a given settlement, the likelihood of reoccupation increases, which increases the likelihood of additional material deposition, and so on in recursive fashion. Thus we could expect superlinear growth in site size and heavy-tailed statistical structure in the person-time of occupation.

44

The working model and expected settlement-size outcomes can be illustrated by way of simulation. In the simulation, a forager/forager group initially occupies a random location on the landscape, and at each time step, they (a) deposit a unit of cultural material, be it a tool or structure, and (b) occupy the location of a randomly selected material with probability *m* or a random location on the landscape with probability 1-*m*. The behavior is repeated many times,

and the person-time of occupation is tallied for each site that is created. Figure 6 summarizes the

*Figure 6. Settlement-size results from a simulation in which the residential moves of virtual foragers are biased by the locations of previously deposited cultural materials. In this example, one forager/forager group moves to the location of a randomly chosen culturally deposited material with 95-percent probability and to a random location on the landscape with 5percent probability. The person-time at each of 500 sites are summarized in this histogram.*

45 results for 500 simulated settlements using an *m *value of 95 percent. As with the results generated in the previous social-bias model, the niche-construction model produces heavy-tailed statistical structure. The same general structure emerges regardless of the value of *m *(see

Appendix II for a more detailed analysis of this model).

Given that the social-bias and niche-construction models produce similar statistical structure in settlement-size variation, it is important to consider ways to distinguish between them. A key distinction is related to differences in the predicted occupation histories of sites in each model. In the social-bias model, we would expect that settlement occupation is continuous over the span of occupation. The reason for this expectation follows from the assumption of social bias. Once a location is completely abandoned, there is no reason for foragers to return to the particular location if their mobility decisions are biased by the locations of other foragers.

Conversely, the niche-construction model predicts discontinuous occupations of settlements because foragers are expected to return to settlements even after abandonment to take advantage of materials and infrastructure that are left behind. Thus the niche-construction model predicts discontinuous occupation of settlements.

A number of archaeological observations would seem to support the niche-construction model's expectation of discontinuous occupation. Chronometric analyses of several huntergatherer sites from diverse environmental and geographic contexts reveal long-occupation spans of centuries or millennia without evidence of appreciable residential permanence—i.e., they evidence high degrees of residential mobility (Aldenderfer 1998; Mitchell et al. 2011; Ortmann

2010; Saunders et al. 1997; 2005; Stiger 2001). It is difficult to account for this dimension of

46 variation with the social-bias model, but it is readily explained by the niche-construction model. I therefore conclude that the niche-construction model offers a more promising starting point for explaining hunter-gatherer settlement-size variation.

*Environmental Structure Model*

Models that predict hunter-gatherer land-use and diet from environmental structure have been enormously successful (Binford 2001; 1980; Bird and O’Connell 2006; Jochim 1976; 1981;

Kelly 1995; Kelly et al. 2013; Kuhn 1995; Yellen 1977). Following from such explanatory successes, environmental structure can be hypothesized to determine hunter-gather settlementsize variation. In the Andes region of South America, which provides empirical cases for the current study, archaeologists have demonstrated that environmental structure predicts many aspects of Archaic Period demography and subsistence behavior (Aldenderfer 1989; 1998; Klink

2005; Lynch 1971; MacNeish et al. 1983; Rick 1980; Sandweiss et al. 2009). Grove's (2010) observation that the occupation spans of !Kung camps were power-law distributed (see also

Brown et al. 2006) led him to suggest underlying power-law structure in natural resource patches. Premo (2006) also considered a variant of this problem, demonstrating through agentbased modeling that patchy ecological conditions can generate highly skewed artifact distributions. Finally, ecological studies have shown that some species of plants and animals tend to generate fractal structure in their spatial distributions (Brown et al. 2002; Bonabeau et al.

1999; Li 2000) suggesting that ecological patch structure could conceivably produce fractal settlement-size structure in human settlement systems.

The archaeological implications of such environmental models are relatively straightforward. If hunter-gatherer site size is a passive reflection of ecological patch structure,

47 then the locations and sizes of economically relevant resource patches should predict the locations and sizes of sites. Kuznar (1989) and Aldenderfer (1998) have shown that localized wetlands known as bofedals were among the most economically important ecological patches in the Andean highlands because bofedals provide critical habitat for the wild camelids and deer that dominated Archaic diets. Accordingly, we might expect that bofedal size and density predicts size variation among Archaic sites in that region. However, in the particular sub-region of interest to be discussed below, bofedales are extremely rare, making them unlikely candidates for explaining settlement-size variation. Other sources of environmental variation are not readily apparent in the study area. Modern environmental structure may be a poor proxy for prehistoric environmental structure, but reconstructing paleoenvironmental structure at a spatial scale that is commensurate with the spatial scale of the settlement pattern data would be exceedingly difficult or impossible due to methodological limitations. Moreover, it seems unlikely that the environmental changes that have been inferred (see Baker et al. 2001; Rigsby et al. 2003) would have significantly altered the major subsistence resources available to hunter-gatherers. Given the uncertainties associated with paleoenvironmental reconstruction at the resolution required here, more tractable solutions are first given consideration.

*Geomorphic Process Model*

A second candidate deterministic model that seems initially plausible for generating settlement-size variation is differential erosion and burial. Because hydrologic processes are

48 fractal in nature, they tend to generate power-law scaling in the statistical properties of topographic relief (Dodds and Rothman 1999; 2000). It is therefore reasonable to hypothesize that power-law scaling in site-size distributions is an emergent property of differential erosion and burial. A simple expectation of this model would be a close association between artifacts and geologic contexts. I evaluated this proposition qualitatively during visits to the many Archaic

Period sites in the Ilave region and the areas between them. However, I observed considerable variation in the distributions of artifacts within geologic contexts. Artifact densities seemed to rise and fall independently of topographic relief and geologic units. Although more rigorous analysis could be used to explore the perceived lack of relationship, my observations lead me to conclude that geology is unlikely to satisfactorily explain hunter-gatherer site-size variation in the Titicaca Basin.

In sum, insights from previous research lead to the consideration of five explanatory models to account for size variation among hunter-gatherer settlements. Although a fair degree of site-size variation can be generated from the independent residential move model, which is the simplest of the candidate set, it does not generate the heavy-tailed structure that is observed empirically. Conversely, the social-bias and niche-construction models are null models that do predict such heavy-tailed statistical structure. A shortcoming of the social-bias model is its inability to predict the discontinuous reuse of settlements that is commonly observed archaeologically among hunter-gatherer sites. The niche-construction model can, however, explain this property and is therefore considered the stronger account of the observed variation.

49

Two deterministic models were also given consideration. The environmental structure model posited that ecological patch structure predicts settlement-size variation. However, the general homogeneity of the Titicaca Basin altiplano environment within which the full range of site-size variation occurs cannot account for the settlement-size variation. Paleoenvironmental structure offers a more appropriate comparison, but adequate reconstruction at the appropriate scale is currently not possible. Finally, geomorphic process suggests an initially plausible explanation for settlement-size variation, but qualitative observations on the absence of spatial relationships between artifact distributions and geologic units leads to the rejection of the candidate model.

Evaluation and comparison of the five candidate models therefore lead me to conclude that the niche-construction model of forager mobility is the most promising and therefore merits additional exploration. Appendix II presents details the working niche-construction model and its predictions.

**Archaic Hunter-Gatherers of the Lake Titicaca Basin and their Context**

The Lake Titicaca Basin Archaic Period serves as the primary empirical case for the analysis presented here. It is particularly suited to exploring the working niche-construction model because (1) previous archaeological fieldwork has generated robust baseline data with good temporal control, (2) the relatively homogeneous environmental structure of the highelevation grassland allows us to isolate non-environmental effects on settlement-size variation, and (3) hierarchical settlement structure is well known among prehistoric agropastoral settlement systems in the region (Albarracin-Jordan 1996; McAndrews et al. 1997; Stanish 2003) giving us

50 a baseline for what settlement-size variation should look like in the region under conditions of socioeconomic complexity.

The Lake Titicaca Basin landscape of Peru and Bolivia lies at over 3,800 masl and is dominated by expansive rolling hills grasslands dissected by streams and flanked by mountains.

The region is one of few in the world to witness the endogenous emergence of residential sedentism, food production, and socioeconomic complexity (Feinman and Marcus 1998; Smith

1995). Specifically, the state of Tiwanaku existed between 1,500 to 1,000 years ago and was characterized by intensive agriculture, monumental architecture, social stratification, longdistance exchange, colonial expansion, and complex craft economies that included textile, metal, and ceramic production (Janusek 2004; Kolata 1993; Moseley 1992; Stanish 2003). Many of these economically complex behaviors can be traced to the preceding Formative periods, 3,500-

1,500 B.P. (Bandy 2004; 2006; Browman 1981; Capriles et al. 2008; Hastorf 1999; Janusek

2004; Kolata 1993; Plourde and Stanish 2006; Schultze et al. 2009; Stanish 2003; Stanish et al.

1997; 2005). Given the socioeconomic complexity evident in the Tiwanaku and Formative periods, it is not surprising that the associated settlement-size distributions exhibit heavy-tailed structure (Albarracin-Jordan 1996; McAndrews et al. 1997; Stanish 2003). Following centralplace economic models and local ethnographic analogs, McAndrews and colleagues (1997) interpret the size structure to economic integration among a nested hierarchy of political units

(see also Albarracin-Jordan 1996). More recently, Arthur Griffin (Griffin 2011; Griffin and

Stanish 2007) has shown through computer simulations that heavy-tailed settlement-size structure can emerge from complex interactions of ecological structure, economic

51 complementarity, and peer-polity competition. What previous research has yet to address is the degree to which such dimensions of socioeconomic complexity are necessary to generate heavytailed settlement-size scaling in the region (but see Haas and Tagliabue 2012). One robust way to answer this question is to evaluate Archaic Period settlement structure that existed prior to sedentism, agriculture, and specialized craft production.

Until relatively recently, little was known about the culture histories that preceded and gave rise to Formative Period settlement patterns in the Titicaca Basin. In 1994 and 1995, Mark

Aldenderfer directed an intensive, systematic pedestrian survey of a 41-km 2 area in the Ilave region on the western side of Lake Titicaca Basin with the goal of locating and examining pre-

Formative Period sites (Aldenderfer 1996). Survey crews documented 468 archaeological sites, and recovered 100-percent of stone tools visible on the surfaces of each site. The surface contents of all of these sites include scatters of flaked stone artifacts such as projectile points, bifacial preforms, scrapers, cores, flake tools, and debitage all made from fine-grained volcanics, cherts, and quartzites (see also Capriles Flores et al. 2011; Cipolla 2005; Klink 2005; La Favre

2011). Formative and Terminal Archaic Period surface assemblages include similar lithic assemblages along with obsidian, undecorated ceramics, and groundstone artifacts. Hundreds of temporally diagnostic projectile points from sites dating to the Early (11,500-9,000 B.P.), Middle

(9,000-7,000 B.P.), Late (7,000-5,000 B.P.), and Terminal Archaic (5,000-3,500 B.P.) periods were recovered and offer a degree of temporal control over the settlement patterns (Klink and

Aldenderfer 2005).

52

Subsequent surveys in other regions of the Titicaca Basin have documented many additional Archaic sites and have permitted a robust picture of Archaic land-use patterns and demography. The Late Archaic Period appears to have represented a population peak as indicated by high site densities and diagnostic projectile point counts (Cipolla 2005; Craig 2005; 2011;

Klink 2005; La Favre 2011). Throughout the Archaic, settlement remained biased toward higher elevations away from Lake Titicaca Basin margins. Few Archaic Period sites and artifacts have been located in surveys conducted adjacent to the lake shore (Bandy 2006). However, growing

Archaic populations from the Early through Late Archaic periods gradually spilled over into lower elevations of the Basin (Cipolla 2005; Klink 2005). After what appears to have been a short period of population decrease during the Terminal Archaic Period, a time of incipient food production, the demographic center of gravity underwent a major shift to the margins of Lake

Titicaca in the Formative Period, a time when climatic conditions ameliorated, the lake rose to its modern levels, and agricultural production and pastoralism were fully underway (Cipolla 2005;

Craig 2005; 2011; Craig et al. 2010; Klink 2005).

Limited excavations at four Ilave Basin sites—Pirco, Jiskairumoko, Kaillachuro, and

Qillqatani—provide a more-detailed view of Archaic period economic organization in the

Titicaca Basin. Shallow deposits and a general absence of cultural features in the 1.4-ha Late

Archaic Period site of Pirco led Nathan Craig (2005) to conclude that it was likely a place of short-term residence. Aldenderfer (1989) arrived at a similar conclusion based on limited Late

Archaic Period findings at the nearby rockshelter site of Qillqatani, which produced two Late

Archaic dates in its lower levels. Unfortunately, the excavations at Pirco did not produce any

53 direct dates. The well-dated Terminal Archaic Period sites of Jiskairumoko and Kaillachurro revealed more complex deposits with small burial mounds, pit houses, and pit features indicating semi-sedentary habitation during that time. Botanical evidence reveals incipient agriculture

(Murray 2005; Rumold 2010), and exotic goods including obsidian, gold, and lapis lazuli reveal increasingly complex socioeconomic organization (Aldenderfer et al. 2008; Craig 2005; Craig et al. 2010; Tripcevich 2007; Tripcevich and Mackay 2011).

Unfortunately, subsistence-related data are extremely limited for the Titicaca Basin Late

Archaic Period. The abundance of projectile points and other lithic artifacts recovered from surface surveys would seem to suggest heavy investment in hunting. Faunal materials from Late

Archaic levels at Qillqatani indicate that deer and wild camelids comprised an important part of the diet. Aldenderfer (1989:122) observed that groundstone was rare at Qillqatani as did Craig

(2011:375) at Pirco leading both to suggest that seed grinding was a relatively unimportant activity at those sites during the Late Archaic Period. The absence of storage pits would seem to corroborate these observations.

Although great strides have been made in hunter-gatherer research in the Lake Titicaca

Basin, the foregoing discussion makes clear that much work is still needed, especially for the

Late Archaic Period and earlier periods (Aldenderfer and Flores Blanco 2011). Settlement pattern analyses are relatively robust for the region, but targeted excavations remain limited. Prior to the research presented here, only two direct dates could be attributed to the Late Archaic Period, and directly dated features and open-air sites remained elusive. However, in this dissertation, I present new excavation results from the previously unexcavated site of Soro Mik'aya Patjxa

(SMP, Ilave 95-259) located in the Ilave Basin. Seventeen radiocarbon dates taken from secure feature contexts place the occupation securely across the Middle/Late Archaic Period boundary

54 with the dates weighted toward that latter end of the occupation. The excavations produced 13 clearly defined cultural pit features, 14 human burials, and 80 thousand artifacts including lithics, groundstone, pigment stone, carbon, and bone. The archaeological findings offer important insights into Middle and Late Archaic Period settlement, subsistence, and social organization.

SMP appears to have been repeatedly occupied for over a millennium by residentially mobile hunter-gatherers whose diets were heavily reliant on ground seeds. Evidence of violent trauma suggests that Late Archaic Period populations experienced a degree of social competition that may have been related to the population peak identified by other researchers. These results and interpretations are detailed in Appendix III.

REFERENCES CITED

Adamic, Lada

2011 Complex systems: Unzipping Zipf’s law. *Nature* 474(7350): 164–165.

Albarracin-Jordan, Juan

1996 Tiwanaku Settlement System: The Integration of Nested Hierarchies in the Lower

Tiwanaku Valley. *Latin American Antiquity* 7(3): 183–210.

Aldenderfer, Mark S.

1989 Archaic Period “Complementarity” in the Osmore Drainage. In *Ecology, Settlement, *

*and History in the Osmore Drainage, Peru*, edited by Don S. Rice, Charles Stanish, and

Phillip R. Scarr, pp. 101–128. BAR International Series 545. B.A.R., Oxford.

1996 *Reconocimiento arqueologico de la Cuenca del Rio Ilave. *Unpublished Report submitted to the Instituto Nacional de Cultura, Lima, Peru.

1998 *Montane Foragers: Asana and the South-Central Andean Archaic*. University of Iowa

Press, Iowa City.

Aldenderfer, Mark S., Nathan M. Craig, Robert J. Speakman, and Rachel Popelka-Filcoff

2008 Four-Thousand-Year-Old Gold Artifacts from the Lake Titicaca Basin, Southern Peru.

*Proceedings of the National Academy of Sciences* 105(13): 5002–5005.

Aldenderfer, Mark S., and Luis A. Flores Blanco

55

2011 Reflexiones para Avanzar en los Estudios del Periodo Arcaico en los Andes Centro-Sur.

*Chungara: Revista de Antropología Chilena* 43(1): 531–550.

Ames, Kenneth M

2008 The Archaeology of Rank. In *Handbook of Archaeological Theories*, edited by R.

Alexander Bentley and Hebert D.G. Maschner, pp. 487–513. AltaMira Press.

Auerbach, Felix

1913 Das gesetz der bevolkerungskonzentration. *Petermanns Geographische Mitteilungen*

59: 74–76.

Baker, Paul A., Geoffrey O. Seltzer, Sherilyn C. Fritz, Robert B. Dunbar, Matthew J. Grove,

Pedro M. Tapia, Scott L. Cross, Harold D. Rowe, and James P. Broda

2001 The History of South American Tropical Precipitation for the Past 25,000 Years.

*Science* 291(5504): 640–643.

Bandy, Matthew S.

2004 Fissioning, Scalar Stress, and Social Evolution in Early Village Societies. *American *

*Anthropologist* 106(2): 322–333.

2006 Early Village Society in the Formative Period in the Southern Lake Titicaca Basin.

*Andean archaeology* 3: 210–236.

Barabási, Albert-László

2005 The Origin of Bursts and Heavy Tails in Human Dynamics. *Nature* 435(7039): 207–

211.

Batty, Michael

2008 The Size, Scale, and Shape of Cities. *Science* 319(5864): 769–771.

2013 A Theory of City Size. *Science* 340(6139): 1418–1419.

Beckmann, Martin J, and John C McPherson

1970 City Size Distribution in a Central Place Hierarchy: An Alternative Approach. *Journal *

*of Regional Science* 10(1): 25–33.

Beckmann, Martin J.

1958 City Hierarchies and the Distribution of City Size. *Economic Development and *

*Cultural Change* 6(3): 243–248.

Bentley, R. Alexander, Matthew W. Hahn, and Stephen J. Shennan

2004 Random Drift and Culture Change. *Proceedings: Biological Sciences* 271(1547): 1443–

1450.

Bentley, R. Alexander, Carl P. Lipo, Harold A. Herzog, and Matthew W. Hahn

2007 Regular Rates of Popular Culture Change Reflect Random Copying. *Evolution and *

*Human Behavior* 28(3): 151–158.

Bentley, R. Alexander, and Herbert D.G. Maschner

2008 Complexity Theory. In *Handbook of Archaeological Theories*, edited by R. Alexander

Bentley and Hebert D.G. Maschner, pp. 245–270. AltaMira Press.

Bentley, R. Alexander, and Stephen J. Shennan

2003 Cultural Transmission and Stochastic Network Growth. *American Antiquity* 68(3): 459–

485.

2005 Random Copying and Cultural Evolution. *Science* 309(5736): 877–879.

56

Berry, Brian J. L.

1961 City Size Distributions and Economic Development. *Economic Development and *

*Cultural Change* 9(4): 573–588.

Bettencourt, Luís M. A., José Lobo, Dirk Helbing, Christian Kühnert, and Geoffrey B. West

2007 Growth, Innovation, Scaling, and the Pace of Life in Cities. *Proceedings of the *

*National Academy of Sciences* 104(17): 7301–7306.

Bettencourt, Luís M. A., José Lobo, Deborah Strumsky, and Geoffrey B. West

2010 Urban Scaling and Its Deviations: Revealing the Structure of Wealth, Innovation and

Crime across Cities. *PLoS ONE* 5(11): e13541.

Bettencourt, Luis M.A., José Lobo, and Deborah Strumsky

2007 Invention in the City: Increasing Returns to Patenting as a Scaling Function of

Metropolitan Size. *Research Policy* 36(1): 107–120.

Bettencourt, Luis, and Geoffrey West

2010 A Unified Theory of Urban Living. *Nature* 467(7318): 912–913.

Bettinger, Robert L.

1991 *Hunter-Gatherers: Archaeological and Evolutionary Theory*. Plenum Press, New York.

1981 Settlement Data and Subsistence Systems. *American Antiquity* 46(3): 640–643.

Binford, Lewis R.

1982 The Archaeology of Place. *Journal of Anthropological Archaeology* 1(1): 5–31.

1980 Willow Smoke and Dogs’ Tails: Hunter-Gatherer Settlement Systems and

Archaeological Site Formation. *American Antiquity* 45(1): 4–20.

2001 *Constructing Frames of Reference: An Analytical Method for Archaeological Theory *

*Building using Hunter-Gatherer and Environmental Data Sets*. University of California

Press, Berkeley.

Bird, Douglas W., and James F. O’Connell

2006 Behavioral Ecology and Archaeology. *Journal of Archaeological Research* 14(2): 143–

188.

Blanton, Richard E.

1976 The Role of Symbiosis in Adaptation and Sociocultural Change in the Valley of

Mexico. In *The Valley of Mexico: studies in pre-Hispanic ecology and society*, edited by

Eric R. Wolf. University of New Mexico Press, Albuquerque.

Bonabeau, Eric, Laurent Dagorn, and Pierre Fréon

1999 Scaling in Animal Group-Size Distributions. *Proceedings of the National Academy of *

*Sciences* 96(8): 4472 –4477.

Brantingham, P. Jeffrey

2003 A Neutral Model of Stone Raw Material Procurement. *American Antiquity* 68(3): 487–

509.

2007 A Unified Evolutionary Model of Style and Function based on the Price Equation.

*American Antiquity* 72(3): 395–416.

Brantingham, P. Jeffrey, and Charles Perreault

2010 Detecting the Effects of Selection and Stochastic Forces in Archaeological

Assemblages. *Journal of Archaeological Science* 37(12): 3211–3225.

57

Browman, David L.

1981 New Light on Andean Tiwanaku: A Detailed Reconstruction of Tiwanaku’s Early

Commercial and Religious Empire Illuminates the Processes by Which States Evolve.

*American Scientist* 69(4): 408–419.

Brown, Clifford T., Larry S. Liebovitch, and Rachel Glendon

2006 Lévy Flights in Dobe Ju/’hoansi Foraging Patterns. *Human Ecology* 35(1): 129–138.

Brown, Clifford T., and Walter R. T. Witschey

2003 The Fractal Geometry of Ancient Maya Settlement. *Journal of Archaeological Science*

30(12): 1619–1632.

Brown, Clifford T., Walter R. T. Witschey, and Larry S. Liebovitch

2005 The Broken Past: Fractals in Archaeology. *Journal of Archaeological Method and *

*Theory* 12(1): 37–78.

Brown, James H., Vijay K. Gupta, Bai-Lian Li, Bruce T. Milne, Carla Restrepo, and Geoffrey B.

West

2002 The Fractal Nature of Nature: Power Laws, Ecological Complexity and Biodiversity.

*Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences*

357(1421): 619 –626.

Capriles Flores, José M., Sergio Calla Maldanado, and Juan Albarracín-Jordán

2011 Tecnología Lítica y Estrategias de Subsistencia Durante los Períodos Arcaico y

Formativo en el Altiplano Central, Bolivia. *Chungara: Revista de Antropología Chilena*

43(1): 455–468.

Capriles, José M., Alejandra I. Domic, and Katherine M. Moore

2008 Fish Remains from the Formative Period (1000 Bc–Ad 400) of Lake Titicaca, Bolivia:

Zooarchaeology and Taphonomy. *Quaternary International* 180(1): 115–126.

Carneiro, Robert L.

1987 The Evolution of Complexity in Human Societies and Its Mathematical Expression.

*International Journal of Comparative Sociology* 28.

2000 The Transition from Quantity to Quality: A Neglected Causal Mechanism in Accounting for Social Evolution. *Proceedings of the National Academy of Sciences* 97(23): 12926–

12931.

Christaller, Walter

1966 *Central Places in Southern Germany*. Trans. Carlisle W. Baskin. Prentice-Hall,

Englewood Cliffs N.J.

Cipolla, Lisa M.

2005 Preceramic Period Settlement Patterns in the Huancané-Putina River Valley, Northern

Titicaca Basin, Peru. In *Advances in Titicaca Basin Archaeology-1*, edited by Charles

Stanish, Amanda B. Cohen, and Mark S. Aldenderfer, pp. 55–64. Cotsen Institute of

Archaeology at UCLA, Los Angeles.

Clauset, Aaron, Cosma Rohilla Shalizi, and M. E. J Newman

2009 Power-Law Distributions in Empirical Data. *SIAM Review* 51: 1–43.

Craig, Nathan M.

2011 Cultural Dynamics, Climate, and Landscape in the South-Central Andes During the

58

Mid-Late Holocene: A Consideration of Two Socio-Natural Perspectives. *Chungará* 43:

367–391.

2005 The Formation of Early Settled Villages and the Emergence of Leadership: A Test of

Three Theoretical Models in the Rio Ilave, Lake Titicaca Basin, Southern Peru.

Unpublished Ph.D. Dissertation, University of California, Santa Barbara.

Craig, Nathan M., Mark S. Aldenderfer, Paul Baker, and Catherine A. Rigsby

2010 Terminal Archaic Settlement Pattern and Land Cover Change in the Rio Ilave,

Southwestern Lake Titicaca Basin, Peru. In *The Archaeology of Anthropogenic *

*Environments*, edited by Rebecca M. Dean. Center for Archaeological Investigations,

Occasional Paper No. 37. Southern Illinois University Center for Archaeological

Investigations, Carbondale.

Cristelli, Matthieu, Michael Batty, and Luciano Pietronero

2012 There is More than a Power Law in Zipf. *Scientific Reports* 2.

Damas, David

1969 Characteristics of Central Eskimo Band Structure. In *Contributions to Anthropology: *

*Band Societies*, edited by David Damas, Bulletin No. 228:pp. 116–141. Anthropological

Series 84. National Museums of Canada, Ottawa.

1972 The Copper Eskimo. In *Hunters and Gatherers Today: a Socioeconomic Study of *

*Eleven Such Cultures in the Twentieth Century*, edited by M. G Bicchieri, pp. 3–50. Holt,

Rinehart and Winston, New York.

Decker, Ethan H., Andrew J. Kerkhoff, and Melanie E. Moses

2007 Global Patterns of City Size Distributions and Their Fundamental Drivers. *PLoS ONE*

2(9): e934.

Dodds, Peter Sheridan, and Daniel H. Rothman

1999 Unified View of Scaling Laws for River Networks. *Physical Review E* 59(5): 4865.

2000 Scaling, Universality, and Geomorphology. *Annual Review of Earth and Planetary *

*Sciences* 28: 571–610.

Drennan, Robert D., and Xiangming Dai

2010 Chiefdoms and States in the Yuncheng Basin and the Chifeng Region: A Comparative

Analysis of Settlement Systems in North China. *Journal of Anthropological Archaeology*

29(4): 455–468.

Drennan, Robert D., and Christian E. Peterson

2004 Comparing Archaeological Settlement Systems with Rank-Size Graphs: A Measure of

Shape and Statistical Confidence. *Journal of Archaeological Science* 31(5): 533–549.

Edwards, Andrew M., Richard A. Phillips, Nicholas W. Watkins, Mervyn P. Freeman, Eugene J.

Murphy, Vsevolod Afanasyev, Sergey V. Buldyrev, M. G. E. da Luz, E. P. Raposo, H. Eugene

Stanley, and Gandhimohan M. Viswanathan

2007 Revisiting Levy Flight Search Patterns of Wandering Albatrosses, Bumblebees and

Deer. *Nature* 449(7165): 1044–1048.

Eeckhout, Jan

2004 Gibrat’s Law for (All) Cities. *The American Economic Review* 94(5): 1429–1451.

La Favre, Karl

59

2011 Limiting Population Size in Hunter-Gatherer Societies: An Analysis of Projectile Points from the Lake Titicaca Basin, Perú. Unpublished master’s thesis, University of California,

Los Angeles.

Feinman, Gary M., and Joyce Marcus (editors).

1998 *Archaic states*. School of American Research Press, Santa Fe.

Flannery, Kent V.

1998 The Ground Plans of Archaic States. In *Archaic states*, edited by Gary M. Feinman and

Joyce Marcus, pp. 15–57. School of American Research Press, Santa Fe.

Fogelin, Lars

2007 Inference to the Best Explanation: A Common and Effective Form of Archaeological

Reasoning. *American Antiquity* 72(4): 603–625.

Fujita, Masahisa, Paul R Krugman, and Anthony Venables

1999 *The Spatial Economy: Cities, Regions and International Trade*. MIT Press, Cambridge,

Mass.

Gabaix, Xavier

1999a Zipf’s Law and the Growth of Cities. *The American Economic Review* 89(2): 129–132.

1999b Zipf’s Law for Cities: An Explanation. *The Quarterly Journal of Economics* 114(3):

739–767.

Gamble, Clive

1999 *The Palaeolithic Societies of Europe*. Cambridge University Press, Cambridge.

González-Val, Rafael

2010 The Evolution of U.S. City Size Distribution from a Long-Term Perspective (1900-

2000). *Journal of Regional Science* 50(5): 952–972.

Griffin, Arthur F.

2011 Emergence of Fusion/Fission Cycling and Self-Organized Criticality from a Simulation

Model of Early Complex Polities. *Journal of Archaeological Science* 38(4): 873–883.

Griffin, Arthur F., and Charles Stanish

2007 An Agent-Based Model of Prehistoric Settlement Patterns and Political Consolidation in the Lake Titicaca Basin of Peru and Bolivia. *Structure and Dynamics: eJournal of *

*Anthropological and Related Sciences* 2(2): 1–47.

Grove, Matt

2010 The Quantitative Analysis of Mobility: Ecological Techniques and Archaeological

Extensions. In *New Perspectives on Old Stones: Analytical Approaches to Paleolithic *

*Technologies*, edited by Stephen J. Lycett and Parth R. Chauhan, pp. 83–118. Springer,

New York.

Haas, W. Randall, Jr., and Jacopo Tagliabue

2012 Prediciendo la Coalescencia en los Períodos Formativo y Tiwanaku en la Cuenca de

Titicaca: Un Modelo Simple Basado en Agentes (Predicting Formative and Tiwanaku

Period Coalescence in the Titicaca Basin: A Simple Agent-based Model). In *Arqueología *

*de la Cuenca del Titicaca, Perú*, edited by Luis Flores Blanco and Henry Tantaleán, pp.

243–260. Instituto Francés de Estudios Andinos; Cotsen Istitute of Archaeology at UCLA,

Lima.

60

Hahn, Matthew W., and R. Alexander Bentley

2003 Drift as a Mechanism for Cultural Change: An Example from Baby Names.

*Proceedings: Biological Sciences* 270: S120–S123.

Hamilton, Marcus J., Bruce T. Milne, Robert S. Walker, Oskar Burger, and James H. Brown

2007 The Complex Structure of Hunter-Gatherer Social Networks. *Proceedings of the Royal *

*Society B: Biological Sciences* 274(1622): 2195–202.

Hastorf, Christine Ann (editor).

1999 *Early Settlement at Chiripa, Bolivia: Research of the Taraco Archaeological Project*.

Contributions of the University of California Archaeological Research Facility 57.

Archaeological Research Facility, University of California, Berkeley.

Herzog, Harold A., R. Alexander Bentley, and Matthew W. Hahn

2004 Random Drift and Large Shifts in Popularity of Dog Breeds. *Proceedings: Biological *

*Sciences* 271: S353–S356.

Hoch, Irving

1972 Income and City Size. *Urban Studies* 9: 299–328.

1976 City Size Effects, Trends, and Policies. *Science* 193(4256): 856–863.

1987 City Size and US Urban Policy. *Urban Studies* 24(6): 570–586.

Hodder, Ian

1979 Simulating the Growth of Hierarchies. In *Transformations: Mathematical Approaches *

*to Culture Change*, edited by Colin Renfrew and Kenneth L Cooke, pp. 117–144.

Academic Press, New York.

Inomata, Takeshi, and Kazuo Aoyama

1996 Central-Place Analyses in the La Entrada Region, Honduras: Implications for

Understanding the Classic Maya Political and Economic Systems. *Latin American *

*Antiquity* 7(4): 291–312.

Janusek, John Wayne

2004 Tiwanaku and Its Precursors Recent Research and Emerging Perspectives. *Journal of *

*Archaeological Research* 12(2): 121–183.

Jochim, Michael

1976 *Hunter-Gatherer Subsistence and Settlement: A Predictive Model*. Academic Press,

New York.

1981 *Strategies for Survival: Cultural Behavior in an Ecological Context*. Academic Press,

New York.

Johnson, Gregory A.

1977 Aspects of Regional Analysis in Archaeology. *Annual Review of Anthropology* 6: 479–

508.

1980 Rank-Size Convexity and System Integration: A View from Archaeology. *Economic *

*Geography* 56(3): 234–247.

Johnson, Norman L., Adrienne W. Kemp, and Samuel Kotz

2005 *Univariate Discrete Distributions*. 3rd ed. Wiley-Interscience, August 30.

Jones, Kate E., Nikkita G. Patel, Marc A. Levy, Adam Storeygard, Deborah Balk, John L.

Gittleman, and Peter Daszak

61

2008 Global Trends in Emerging Infectious Diseases. *Nature* 451(7181): 990–993.

Kelly, Robert L.

1995 *The Foraging Spectrum: Diversity in Hunter-Gatherer Lifeways*. Percheron Press, New

York.

Kelly, Robert L., Todd A. Surovell, Bryan N. Shuman, and Geoffrey M. Smith

2013 A Continuous Climatic Impact on Holocene Human Population in the Rocky

Mountains. *Proceedings of the National Academy of Sciences* 110(2): 443–447.

Kimura, Motoo

1968 Evolutionary Rate at the Molecular Level. *Nature* 217(5129): 624–626.

1985 *The Neutral Theory of Molecular Evolution*. Cambridge University Press, Cambridge.

Klink, Cynthia J.

2005 Archaic Period Research in the Río Huenque Valley, Peru. In *Advances in Titicaca *

*Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 13–24. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Klink, Cynthia J., and Mark S. Aldenderfer

2005 A Projectile Point Chronology for the South-Central Andean Highlands. In *Advances in *

*Titicaca Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 25–54. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Kolata, Alan L.

1993 *The Tiwanaku: Portrait of an Andean Civilization*. Blackwell, Cambridge, MA.

Krugman, Paul

1996a *The Self-Organizing Economy*. Blackwell Publishers, Cambridge.

1996b Confronting the Mystery of Urban Hierarchy. *Journal of the Japanese and *

*International Economies* 10(4): 399–418.

Kuhn, Steven L

1995 *Mousterian Lithic Technology: An Ecological Perspective*. First Edition. Princeton

University Press, Princeton N.J., April 3.

Kuznar, Lawrence A.

1989 The Domestication of Camelids in Southern Peru. In *Ecology, Settlement, and History *

*in the Osmore Drainage, Peru*, edited by Don S. Rice, Charles Stanish, and Phillip R.

Scarr, pp. 167–182. BAR International Series 545. B.A.R., Oxford.

Laland, Kevin N., Jeremy R. Kendal, and Gillian R. Brown

2007 The Niche Construction Perspective: Implications for Evolution and Human Behaviour.

*Journal of Evolutionary Psychology* 5(1): 51–66.

Lansing, J. Stephen

2003 Complex Adaptive Systems. *Annual Review of Anthropology* 32: 183–204.

Lansing, J. Stephen, Joseph C. Watkins, Brian Hallmark, Murray P. Cox, Tatiana M. Karafet,

Herawati Sudoyo, and Michael F. Hammer

2008 Male Dominance Rarely Skews the Frequency Distribution of Y Chromosome

Haplotypes in Human Populations. *Proceedings of the National Academy of Sciences*

105(33): 11645–11650.

Laxton, R.R., and W.G. Cavanagh

62

1995 The Rank-Size Dimension and the History of Site Structure from Survey Data. *Journal *

*of Quantitative Anthropology* 5: 327–358.

Leacock, Eleanor

1969 The Montagnais-Naskapi Band. In *Contributions to Anthropology: Band Societies*, edited by David Damas, Bulletin No. 228:pp. 1–17. Anthropological Series 84. National

Museums of Canada, Ottawa.

Levin, Simon A.

2003 Complex Adaptive Systems: Exploring the Known, the Unknown and the Unknowable.

*American Mathematical Society* 40: 3–19.

Li, Bai-Lian

2000 Fractal Geometry Applications in Description and Analysis of Patch Patterns and Patch

Dynamics. *Ecological Modelling* 132(1-2): 33–50.

Lyman, R. Lee

2004 The Concept of Equifinality in Taphonomy. *Journal of Taphonomy* 2(1): 15–26.

Lynch, Thomas F.

1971 Preceramic Transhumance in the Callejon de Huaylas, Peru. *American Antiquity* 36(2):

139–148.

MacDonald, Douglas H., and Barry S. Hewlett

1999 Reproductive Interests and Forager Mobility. *Current Anthropology* 40(4): 501–524.

MacNeish, Richard S., Robert K. Vierra, Antoinette Nelken-Terner, and Angel Garcia Cook

1983 *Prehistory of the Ayacucho Basin, Peru: The Preceramic Way of Life*. Vol. IV.

University of Michigan Press, Ann Arbor.

Mandelbrot, Benoit

1982 *The Fractal Geometry of Nature*. W.H. Freeman, San Francisco.

Marquet, Pablo A., Andrew P. Allen, James H. Brown, Jennifer A. Dunne, Brian J. Enquist,

James F. Gillooly, Patricia A. Gowaty, Jessica L. Green, John Harte, Steve P. Hubbell, James

O’Dwyer, Jordan G. Okie, Annette Ostling, Mark Ritchie, David Storch, and Geoffrey B. West

2014 On Theory in Ecology. *BioScience* 64(8): 701–710.

McAndrews, Timothy L., Juan Albarracin-Jordan, and Marc Bermann

1997 Regional Settlement Patterns in the Tiwanaku Valley of Bolivia. *Journal of Field *

*Archaeology* 24(1): 67–83.

Mitchell, Melanie

2009 *Complexity: A Guided Tour*. Oxford University Press, Oxford [England]; New York.

Mitchell, Peter, Ina Plug, Geoff Bailey, Ruth Charles, Amanda Esterhuysen, Julia Lee Thorp,

Adrian Parker, and Stephan Woodborne

2011 Beyond the Drip-Line: A High-Resolution Open-Air Holocene Hunter-Gatherer

Sequence from Highland Lesotho. *Antiquity* 85(330): 1225–1242.

Mitzenmacher, Michael

2001 A Brief History of Generative Models for Power Law and Lognormal Distributions.

*Proceedings of the Annual Allerton Conference on Communication Control and *

*Computing* 39: 182–191.

Moseley, Michael

63

1992 *The Incas and their Ancestors: The Archaeology of Peru*. Revised. Thames and Hudson,

New York.

Murray, Andrea

2005 Chenopodium Domestication in the South-Central Andes: Confirming the Presence of

Domesticates at Jiskairumoko (late Archaic-Formative), Peru. Unpublished Master’s

Thesis, University of California, Fullerton.

Neiman, Fraser D.

1995 Stylistic Variation in Evolutionary Perspective: Inferences from Decorative Diversity and Interassemblage Distance in Illinois Woodland Ceramic Assemblages. *American *

*Antiquity* 60(1): 7–36.

Newman, M. E. J.

2005 Power Laws, Pareto Distributions and Zipf’s Law. *Contemporary Physics* 46(5): 323–

351.

Odling-Smee, F. John, Kevin N Laland, and Marcus W Feldman

2003 *Niche Construction: The Neglected Process in Evolution*. Princeton University Press,

Princeton.

1996 Niche Construction. *The American Naturalist* 147(4): 641–648.

Odling-Smee, John, Douglas H. Erwin, Eric P. Palkovacs, Marcus W. Feldman, and Kevin N.

Laland

2013 Niche Construction Theory: A Practical Guide for Ecologists. *The Quarterly Review of *

*Biology* 88(1): 3–28.

Ortman, Scott G., Andrew H. F. Cabaniss, Jennie O. Sturm, and Luís M. A. Bettencourt

2014 The Pre-History of Urban Scaling. *PLoS ONE* 9(2): e87902.

Ortmann, Anthony L.

2010 Placing the Poverty Point Mounds in Their Temporal Context. *American Antiquity*

75(3): 657–678.

Parr, John B.

1969 City Hierarchies and the Distribution of City Size: A Reconsideration of Beckmann’s

Contribution. *Journal of Regional Science* 9(2): 239–253.

Paynter, Robert

1982 *Models of Spatial Inequality: Settlement Patterns in Historical Archeology*. Academic

Press, New York.

Pearson, Charles E.

1980 Rank-Size Distributions and the Analysis of Prehistoric Settlement Systems. *Journal of *

*Anthropological Research* 36(4): 453–462.

Peterson, Christian E., and Robert D. Drennan

2005 Communities, Settlements, Sites, and Surveys: Regional-Scale Analysis of Prehistoric

Human Interaction. *American Antiquity* 70(1): 5–30.

Plourde, Aimeé, and Charles Stanish

2006 The Emergence of Complex Society in the Titicaca Basin: The View from the North.

*Andean archaeology*(3): 237–257.

Premo, Luke

64

2006 Patchiness and Prosociality: Modeling the Evolution and Archaeology of Plio-

Pleistocene Hominin Food Sharing. Unpublished Ph.D. Dissertation, The University of

Arizona, Tucson.

Richerson, Peter

2013 Human Evolution: Group Size Determines Cultural Complexity. *Nature* 503(7476):

351–352.

Rick, John W.

1980 *Prehistoric Hunters of the High Andes*. Academic Press, New York.

Rigsby, Catherine A., Paul A. Baker, and Mark S. Aldenderfer

2003 Fluvial History of the Rio Ilave Valley, Peru, and Its Relationship to Climate and

Human History. *Palaeogeography, Palaeoclimatology, Palaeoecology* 194(1-3): 165–185.

Rogers, Edward S.

1969 Band Organization among the Indians of Eastern Subarctic Canada. In *Contributions to *

*Anthropology: Band Societies*, edited by David Damas, Bulletin No. 228:pp. 1–17.

Anthropological Series 84. National Museums of Canada, Ottawa.

1972 The Mistassini Cree. In *Hunters and Gatherers Today: a Socioeconomic Study of *

*Eleven Such Cultures in the Twentieth Century*, edited by M. G Bicchieri, pp. 90–137.

Holt, Rinehart and Winston, New York.

Rumold, Claudia U.

2010 Illuminating Women’s Work and the Advent of Plant Cultivation in the Highland

Titicaca Basin of South America: New Evidence from Grinding Tool and Starch Grain

Analyses. Unpublished Ph.D. dissertation, University of California, Santa Barbara.

Sandweiss, Daniel H, Ruth Shady Solís, Michael E Moseley, David K Keefer, and Charles R

Ortloff

2009 Environmental Change and Economic Development in Coastal Peru Between 5,800 and

3,600 Years Ago. *Proceedings of the National Academy of Sciences* 106(5): 1359–1363.

Saunders, Joe W, Rolfe D Mandel, Roger T Saucier, E. Thurman Allen, C. T Hallmark, Jay K

Johnson, Edwin H Jackson, Charles M Allen, Gary L Stringer, Douglas S Frink, James K

Feathers, Stephen Williams, Kristen J Gremillion, Malcolm F Vidrine, and Reca Jones

1997 A Mound Complex in Louisiana at 5400-5000 Years Before the Present. *Science*

277(5333): 1796–1799.

Saunders, Joe W., Rolfe D. Mandel, C. Garth Sampson, Charles M. Allen, E. Thurman Allen,

Daniel A. Bush, James K. Feathers, Kristen J. Gremillion, C. T. Hallmark, H. Edwin Jackson, Jay

K. Johnson, Reca Jones, Roger T. Saucier, Gary L. Stringer, and Malcolm F. Vidrine

2005 Watson Brake, a Middle Archaic Mound Complex in Northeast Louisiana. *American *

*Antiquity* 70(4): 631–668.

Savage, Stephen H.

1997 Assessing Departures from Log-Normality in the Rank-Size Rule. *Journal of *

*Archaeological Science* 24(3): 233–244.

Schläpfer, Markus, Luís M. A. Bettencourt, Sébastian Grauwin, Mathias Raschke, Rob Claxton,

Zbigniew Smoreda, Geoffrey B. West, and Carlo Ratti

2014 The Scaling of Human Interactions with City Size. *Journal of The Royal Society *

65

*Interface* 11(98): 20130789.

Schultze, C. A., C. Stanish, D. A. Scott, T. Rehren, S. Kuehner, and J. K. Feathers

2009 Direct Evidence of 1,900 Years of Indigenous Silver Production in the Lake Titicaca

Basin of Southern Peru. *Proceedings of the National Academy of Sciences* 106(41): 17280–

17283.

Shennan, Stephen

2000 Population, Culture History, and the Dynamics of Culture Change. *Current *

*Anthropology* 41(5): 811–835.

2001 Demography and Cultural Innovation: A Model and Its Implications for the Emergence of Modern Human Culture. *Cambridge Archaeological Journal* 11(01): 5–16.

Silberbauer, George B.

1972 The G/wi Bushmen. In *Hunters and Gatherers Today: a Socioeconomic Study of Eleven*

*Such Cultures in the Twentieth Century*, edited by M. G Bicchieri, pp. 271–326. Holt,

Rinehart and Winston, New York.

Slobodin, Richard

1969 Leadership and Participation in a Kutchin Indian Trapping Party. In *Contributions to *

*Anthropology: Band Societies*, edited by David Damas, Bulletin No. 228:pp. 56–89.

Anthropological Series 84. National Museums of Canada, Ottawa.

Smith, Bruce D.

1995 *The Emergence of Agriculture*. Scientific American Library series 54. Scientific

American Library, New York.

Sornette, Didier

2006 *Critical Phenomena in Natural Sciences: Chaos, Fractals, Selforganization, and *

*Disorder: Concepts and Tools*. Springer, Berlin; New York.

Stanish, Charles

2003 *Ancient Titicaca: The Evolution of Complex Society in Southern Peru and Northern *

*Bolivia*. University of California Press, Berkeley.

Stanish, Charles, Amanda B. Cohen, Edmundo de la Vega, Elizabeth Arkush, Cecilia Chávez,

Aimeé Plourde, and Carol Schultz

2005 Archaeological Reconnaissance in the Northern Titicaca Basin. In *Advances in Titicaca *

*Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 289–316. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Stanish, Charles, Edmundo De la Vega M., Lee Steadmann, Cecilia Chávez Justo, Kirk Lawrence

Frye, Luperio Onofre Mamani, Matthew T. Seddon, and Percy Calisaya Chuquimia

1997 *Archaeological Survey in the Juli-Desaguadero Region of Lake Titicaca Basin, *

*Southern Peru*. Field Museum of Natural History.

Steponaitis, Vincas P.

1981 Settlement Hierarchies and Political Complexity in Nonmarket Societies: The

Formative Period of the Valley of Mexico. *American Anthropologist* 83(2): 320–363.

Steward, Julian H.

1955 *Theory of Culture Change: The Methodology of Multilinear Evolution*. University of

Illinois Press, Urbana.

66

1970 The Foundations of Basin-Plateau Shoshonean Society. In *Languages and Cultures of *

*Western North America*, edited by Earl Herbert Swanson, pp. 113–151. Idaho University

Press, Pocatello.

Stiger, Mark

2001 *Hunter-Gatherer Archaeology of the Colorado High Country*. University Press of

Colorado, Boulder.

Stiner, Mary C.

2001 Thirty Years on the “Broad Spectrum Revolution” and Paleolithic Demography.

*Proceedings of the National Academy of Sciences* 98: 6993–6996.

Stiner, Mary C., and Steven L Kuhn

2006 Changes in the “Connectedness” and Resilience of Paleolithic Societies in

Mediterranean Ecosystems. *Human Ecology* 34(5): 693–712.

Surovell, Todd A.

2009 *Toward a Behavioral Ecology of Lithic Technology: Cases from Paleoindian *

*Archaeology*. University of Arizona Press, October 1.

Thomas, David Hurst

1972 A Computer Simulation Model of Great Basin Shoshonean Subsistence and Settlement

Patterns. In *Models in Archaeology*, edited by David L. Clarke, pp. 671–704. Methuen,

London.

1973 An Empirical Test for Steward’s Model of Great Basin Settlement Patterns. *American *

*Antiquity* 38(2): 155–176.

1974 An Archaeological Perspective on Shoshonean Bands. *American Anthropologist* 76(1).

New Series: 11–23.

1983 *The Archaeology of Monitor Valley 1: Epistemology*. Vol. 58(1). Anthropological Paper of the American Museum of Natural History. American Museum of Natural History, New

York.

1988 *The Archaeology of Monitor Valley 3: Survey and Additional Excavations*. Vol. 66(2).

Anthropological Papers of the American Museum of Natural History. American Museum of Natural History, New York.

Thomas, David Hurst, and Robert L. Bettinger

1976 *Prehistoric Piñon Ecotone Settlements of the Upper Reese River Valley, Central *

*Nevada*. Vol. 53(3). Anthropological Papers of the America Musueum of Natural History.

American Museum of Natural History, New York.

Tripcevich, Nicholas

2007 Quarries, Caravans, and Routes to Complexity: Prehispanic Obsidian in the South-

Central Andes. Unpublished Ph.D. dissertation, University of California, Santa Barbara.

Tripcevich, Nicholas, and Alex Mackay

2011 Procurement at the Chivay Obsidian Source, Arequipa, Peru. *World Archaeology* 43(2):

271–297.

Tuncay, Çağlar

2008 A Theoretical Approach for Pareto–Zipf Law. *International Journal of Modern Physics *

*C: Computational Physics & Physical Computation* 19(8): 1175–1181.

67

Turnbull, Colin M.

1968 The Importance of Flux in Two Hunting Societies. In *Man the Hunter*, edited by

Richard B Lee and Irven DeVore, pp. 132–137. Aldine Publishing Company, Chicago.

Walmsley, D. Jim, and Gareth J. Lewis

1989 The Pace of Pedestrian Flows in Cities. *Environment and Behavior* 21(2): 123–150.

Whallon, Robert

2006 Social Networks and Information: Non-“Utilitarian” Mobility among Hunter-Gatherers.

*Journal of Anthropological Archaeology* 25(2): 259–270.

White, Ethan P., Brian J. Enquist, and Jessica L. Green

2008 On Estimating the Exponent of Power-Law Frequency Distributions. *Ecology* 89(4):

905–912.

Winterhalder, Bruce, and R. Brooke Thomas

1978 *Geoecology of Southern Highland Peru: A Human Adaptation Perspective*. University of Colorado, Institute of Arctic and Alpine Research, Boulder.

Wobst, H. Martin

1974 Boundary Conditions for Paleolithic Social Systems: A Simulation Approach. *American*

*Antiquity* 39(2): 147–178.

1978 The Archaeo-Ethnology of Hunter-Gatherers or the Tyranny of the Ethnographic

Record in Archaeology. *American Antiquity* 43(2): 303–309.

Woodburn, James

1968 Stability and Flexibility in Hadza Residential Groupings. In *Man the Hunter*, edited by

Richard B Lee and Irven DeVore, pp. 103–117. Aldine Publishing Company, Chicago.

Yellen, John

1977 *Archaeological Approaches to the Present: Models for Reconstructing the Past*.

Academic Press, New York.

Zeanah, David W.

2002 Central Place Foraging and Prehistoric Pinyon Utilization in the Great Basin. In *Beyond*

*Foraging and Collecting: Evolutionary Change in Hunter-Gatherer Settlement Systems*, edited by Ben Fitzhugh, pp. 231–256. Kluwer Academic/Plenum Publishers, New York.

Zipf, George Kingsley

1949 *Human Behavior and the Principle of Least Effort: An Introduction to Human Ecology*.

Addison-Wesley Press, Cambridge.

APPENDIX I:

POWER-LAW SCALING IN PREHISTORIC HUNTER-GATHERER SETTLEMENT

SYSTEMS

68

Authors and Affiliations:

W. Randall Haas, Jr.

a

Mark S. Aldenderfer a,b

Cynthia J. Klink

Greg J. Maggard c d a School of Anthropology, The University of Arizona, Tucson, AZ 85701-0030; b School of Social

Sciences, Humanities, and Arts, University of California, Merced, CA 95343;

Oneonta, Oneonta, NY 13820, d University of Kentucky, Lexington c

SUNY College at

Contributions: WRH designed the research; carried out field, laboratory, and analytical work; and wrote the paper. MSA, CJK, and GJM carried out field and laboratory work.

69

ABSTRACT

Settlement size predicts the rates and magnitudes of many human behaviors such as technological innovation, disease, crime, and inequality. Yet, the factors that drive settlementsize variation remain poorly understood. Size variation among economically integrated settlements tends to be heavy tailed such that the smallest settlements are extremely common and the largest settlements extremely large and rare. The upper tail of this size distribution is often formalized as a power-law function. Explanations for emergent power-law structure in human settlement systems tend to emphasize complex socioeconomic processes including agriculture, manufacturing, and warfare, but the degree to which heavy-tailed settlement-size structure existed in the absence of complex behaviors remains unclear. By examining eight archaeological hunter-gatherer settlement systems spanning three distinct environmental contexts, this analysis finds that power-law models offer plausible and parsimonious statistical descriptions of huntergatherer settlement-size variation. We suggest that the heavy-tailed statistical structure of huntergatherer settlement-size variation may have laid a structural foundation for emergent settlement hierarchies in subsequent times.

INTRODUCTION

Human population density is often cited as a significant driver of behavioral variation

(Carneiro 2000; Shennan 2000; Stiner 2001). Within modern human settlement systems, populations vary considerably from settlement to settlement. This variation tends to be heavy tailed such that extremely small settlements are extremely frequent and extremely large

settlements extremely rare. The high degree of variation relates in predictable ways to the dynamics of disease transmission, wealth inequality, technological innovation, crime, and even

70 seemingly trivial behaviors such as walking and talking speeds (Bettencourt, Lobo, Helbing, et al. 2007; Bettencourt, Lobo, and Strumsky 2007; Bettencourt et al. 2010; Jones et al. 2008).

Understanding how extreme settlement-size variation emerges and persists in human societies is therefore relevant to understanding variation in many other human behaviors. Yet, the behavioral basis for extreme settlement-size variation remains poorly understood (Adamic 2011; Batty

2008; Gan et al. 2006; Krugman 1996). Scholars commonly link the variation to central-place theory (sensu Christaller 1966), mapping settlement size onto political and economic function in hierarchically organized settlement systems. In archaeology, settlement-size hierarchy is often considered a defining characteristic of complex, state-organized society and a consequence of complex socioeconomic processes involving various combinations of agriculture, specialized craft production (i.e., manufacturing), elite competition, and warfare (Flannery 1998; Johnson

1980; Brown and Witschey 2003; Stanish 2003; Drennan and Peterson 2004). Economists and geographers have long considered settlement-size hierarchy integral to economic systems— simultaneously an emergent property and driver of economies of scale (Krugman 1996; Fujita et al. 1999; Christaller 1966).

With more than a century of research in several disciplines, scholars have characterized the extremeness of settlement-size variation in a number of ways. A common approach has been to subjectively partition perceived modes in settlement-size histograms into size tiers (Ames

2008; Flannery 1998; Stanish 2003). A more rigorous alternative approach follows from the

71 observation that settlement-size variation in complex societies tends toward a continuous powerlaw function where the size of a settlement is proportional to the inverse of its rank—rank 1 being the largest settlement and rank *n *the smallest (Clauset et al. 2009). The probability density function of a power-law model with scaling exponent, *α,* is expressed as follows: *f*

(*x*)∝ *x*

− *α*

(Newman 2005). Because size and frequency are inversely related, power-law size structure gives the impression of pyramidal size grading, or hierarchical structure.

A number of graphical and statistical procedures exist to test for power-law structure in empirical data. Perhaps the most widely used cross-disciplinary approach—and the one employed here—assesses the absolute and relative statistical fit between settlement-size data and a candidate set of statistical models. This method of identifying a statistical model to describe empirical data has the advantage of providing theoretical links among models and generic mechanisms capable of producing them (Newman 2005; Mitzenmacher 2001). In identifying plausible and implausible statistical models, valuable insights into the relevance of more proximate behavioral models can be gained.

Current studies of human settlement-size variation are overwhelmingly biased toward modern and historical settlement systems of western cultures (Batty 2008; Christaller 1966;

Paynter 1982; Richardson 1973; Zipf 1949). Considerable attention has been given to the structure of U.S. settlement systems, for example, which appear to exhibit power-law structure in the distribution of city sizes whether city size is measured in terms of population or areal extent

(Batty 2008; Clauset et al. 2009; Decker et al. 2007; Krugman 1996; Zipf 1949). Archaeological research extends the scope of settlement-size studies to include the settlement systems of

72 prehistoric non-western cultures, especially state-organized societies such as Maya,

Mesopotamia, and Tiwanaku (Brown and Witschey 2003; Inomata and Aoyama 1996; Stanish

2003; Wright 1977). To a lesser extent, non-state agricultural societies have also been examined

(Bandy 2006; Peterson and Drennan 2005).

Comparable analyses of settlement-size structure among hunter-gatherer societies are currently lacking. The reason perhaps stems from conventional notions that hierarchical structure of any form is antithetical to egalitarian hunter-gatherer economies. Whereas the socioeconomic organization of complex societies tends to include residential sedentism, agricultural dependence, specialized manufacturing, political hierarchy, and/or warfare, hunter-gatherer socioeconomic organization tends to emphasize high residential mobility and resource pooling to access spatially and temporally disparate wild resources (Kelly 1995; Kuhn and Stiner 2001).

Because heavy-tailed settlement-size variation is often seen as an index of socioeconomic complexity, analyses of settlement-size variation has tended to remain in studies of agrarian and industrial societies. Yet, few studies have explicitly tested the assumption that heavy-tailed settlement-size variation is exclusive to complex societies. Bridging the empirical gap is essential to understanding the degree to which extreme settlement size variation is a cause or consequence of emergent socioeconomic complexity in human societies. If heavy-tailed structure is symptomatic of complexity as previous research would seem to suggest, then we would expect heavy-tailed size structure to be absent in the settlement systems of hunter-gatherer societies.

Alternatively, there is reason to suspect power-law structure in hunter-gatherer settlement-size variation even in the absence of socioeconomic complexity. Some gregarious

73 animals appear to exhibit power-law structure in their group sizes (Bonabeau et al. 1999), and some predatory animals appear to exhibit power-law structure in the distribution of *waiting times*

*—*the amount of time spent in a given location before moving to another (Wearmouth et al.

2014). Analogous forms of either of these behaviors—differential aggregation and/or sedentism

—could conceivably generate power-law structure in archaeological site-size variation in huntergatherer settlement systems.

Hamilton et al. (2007) make the surprising observation that group-size variation among

339 ethnographic hunter-gatherer settlement systems is characterized by log-linear structure.

Though the semi-quantitative data structure may or may not reflect power-law scaling, it is indicative of heavy-tailed structure and suggests the possibility of power-law structure in huntergatherer-group size variation. Moreover, two studies argue that power-law scaling characterizes waiting times in ethnographic !Kung foraging patterns (Brown et al. 2006; Grove 2010). While these novel studies make key contributions to our understanding of hunter-gatherer settlement structure, three fundamental data biases constrain interpretation. First, all ethnographic huntergatherers were economically integrated with sedentary agricultural and industrial societies to some degree (Kelly 1995; Wobst 1978), leading us to question if the observed patterns would have emerged in the absence of such socioeconomic relations. Second, in the case of Hamilton et al.'s study, the semi-quantitative data available to the scholars constrains their statistical interpretation. Third, the waiting time analyses use statistical procedures that have since been shown to produce biased results (Clauset et al. 2009; Edwards et al. 2007).

74

Archaeological examination of hunter-gatherer settlement systems offers one means to address ethnographic data limitations. The great time spans covered by archaeological data offer greater control over economic factors such as the availability of domesticated foods. Moreover, settlement-size data are quantitative and do not have the same mathematical constraints as ethnographic group-size data. However, because archaeological data are fraught with their own biases—most notably the persistent problem of equifinality (Lyman 2004)—such observations are appropriately considered an independent and complementary line of evidence to ethnographic observations.

In this analysis, we examine the statistical structure of settlement-size variation among eight prehistoric hunter-gatherer settlement systems spanning more than 4,000 years and three distinct environmental contexts in highlands Peru, coastal Peru, and interior U.S. Southwest.

Settlement size is measured in terms of areal extents and artifact counts. The analysis shows that incipient forms of power-law structure existed among hunter-gatherer settlements systems across diverse environmental contexts. Below, we describe the materials, methods, and results and conclude with a discussion of theoretical implications for hunter-gatherer mobility and emergent complexity in human societies.

MATERIALS AND METHODS

To test the hypothesis of power-law structure in hunter-gatherer settlement-size variation, we examine the size distribution of prehistoric archaeological sites deposited by eight huntergatherer settlement systems in three distinct geographic contexts. This section describes the sample, the indices of settlement size, and the procedure used to test for power-law scaling.

75

**Archaeological Sample**

Eight New World archaeological settlement systems and their environmental contexts are summarized in this section. All settlement systems represent unequivocal hunter-gatherer economies marked by high residential mobility and ephemeral residential architecture.

Agricultural neighbors were either absent or highly unlikely in all cases. The purpose in describing the environmental contexts of the settlement systems is to illustrate the extreme environmental variation represented in our sample. Although some aspects of environmental structure such as temperature and precipitation have varied significantly through time, modern conditions offer apt proxies of prehistoric conditions for our purpose because the range of variation through time within each setting is small compared to the range of variation among them. Environmental contexts range from 16º south latitude to 33º north latitude, sea level to over 3,800 masl, temperate to seasonal weather regimes, and grassland to desert biomes. The economic and environmental scope of the sample serves to explore the generality of settlementsize structure within a narrow hunter-gatherer economic regime.

The first environmental context discussed is the western Lake Titicaca Basin in the Andes

Mountains of highlands Peru. Elevations range from 3,800 masl at Lake Titicaca to 6,400 masl at the peak of Cerro Janq'u Uma. Human populations intensively inhabit the lower elevations in an environment known as the *altiplano—*a vast expanse of rolling-hill grasslands dissected by perennial rivers and flanked by mountains (Winterhalder and Thomas 1978). Precipitation varies from approximately 300 to 900 mm/yr depending on elevation, local physiography, and lowfrequency climatic conditions. Mean daily temperature lows and highs range from -10ºC to 19ºC.

76

The altiplano archaeological data come from two sub-regions of the Ilave drainage in the western Titicaca Basin—The Río Ilave Basin and the middle Río Huenque Valley. The Ilave

Basin study area is centered at 16º12'40”S, 69º43'20”W (WGS 1984). Elevations range from approximately 3,830 to 3,900 masl with adjacent mountains to 4,600 masl. In a 41-km 2 sample area, Aldenderfer and colleagues recorded 90 archaeological sites with Archaic period huntergatherer artifacts (Aldenderfer and Flores Blanco 2011). In addition, the first author revisited 24 of those sites and recorded 6 new sites.

The Río Huenque study area occurs in a tributary to the Río Ilave centered at 16º45'50”S,

69º43'40”W (WGS 1984) (Klink 2005). The 33-km 2 sample area occurs in a relatively restricted valley where elevations range between approximately 3,940 and 4,070 masl. Surrounding mountains rise to 5,100 masl. Klink recorded 139 archaeological sites with hunter-gatherer artifacts.

Hunter-gatherer populations occupied the Lake Titicaca Basin for at least 4,000 years

(Aldenderfer 2008), and a detailed projectile point typology allows us to divide the long time frame into Early (11,500-9,000 cal. B.P.), Middle (9,000-7,000 cal. B.P.), and Late (7,000-5,000

cal. B.P.) Archaic periods (Figure 7) (Klink and Aldenderfer 2005). All of these periods represent

hunter-gatherer economies reliant on vicuña (a wild camelid), taruca (Andean deer), wild seeds, and wild tubers (Aldenderfer 1998). The three temporal contexts in conjunction with the two environmental sub-contexts constitute six archaeological settlement systems. Settlement-size metrics for all altiplano sites include areal extents reported by field crews and temporally diagnostic projectile point counts tabulated specifically for this study.

77

The second environmental context considered is Jequetepeque coastal plain and foothills of northern Peru, 1,400 km northwest of the altiplano study area. Mean daily temperature lows and highs range from 16ºC to 31ºC. The extremely arid environment rarely receives more that 50 mm of rainfall per year, but the Pacific littoral and lush alluvial plains offer highly productive localized resource zones with diverse and often-abundant marine and terrestrial resources.

Dillehay and Maggard (Dillehay et al. 2003;

Maggard 2011) conducted archaeological settlement surveys covering 70 km 2 centered at

7º9'11"S, 79º22'31"W (WGS 84). The efforts recorded 126 hunter-gatherer sites with material evidence of *Paijan culture*—a hunter-gatherer tradition marked by distinctive flaked stone technologies that persisted from approximately

11,000-8,500 cal. B.P. (Dillehay 2008; Maggard

2011).

*Figure 7: Examples of temporally diagnostic projectile points used to (1) assign archaeological sites to settlement systems and (2) measure site size in terms of artifact counts. Top row: Titicaca *

*Basin Late Archaic Period. 2 nd row: *

*Titicaca Basin Middle Archaic Period. 3 rd row: Titicaca Basin Early Archaic Period*

*(images reproduced with permission of *

*Dr. Chris Loendorf). 4*

*Middle Archaic. 5 th th*

* row: Gila River *

* row: Jequetepeque *

*Paijan.*

78

The third environmental context considered is located in the Middle Gila River of the

U.S. Southwest at 33º8'53"N, 111º51'10"W, approximately 6,000 km to the northwest of the

Jequetepeque region. The Sonoran Desert environment averages 200 mm of precipitation per year. Mean daily temperature lows and highs range from 3ºC to 40ºC. Surface water is scarce and ephemeral. Major hunter-gatherer subsistence resources include bighorn sheep, whitetail deer, rabbit, mesquite seeds, and cactus fruit. Gila River Indian Community archaeologists report

Middle Archaic Period (5,000-4,000 cal. B.P.) projectile-point counts for 50 archaeological sites in a 591-km 2 area (Loendorf and Rice 2004). These counts comprise the eighth and final case study investigated here.

**Measuring Settlement Size**

Decker et al. (2007) show that power-law scaling can be observed in a number of settlement-size metrics for modern cities. In this study, we use two comparable, archaeologically tractable metrics of settlement size—areal extent and artifact count. A site's areal extent is considered a relative proxy for co-resident population size. In general, the more individuals that contemporaneously occupy a location, the more horizontal space they tend to require. A site's artifact count is considered a relative proxy for person-hours of occupation. In general, the greater the number of individuals that occupy a settlement and the longer it is occupied, the greater the deposition of cultural materials. Although specific activities may also affect site areas and artifact counts, we assume they are reasonable proxies for occupation intensity, and this assumption finds empirical support in the ethnoarchaeological work of Yellen (1977).

79

For seven of the eight settlement systems analyzed, we compile site-area estimates as reported by field analysts. Site-area estimates are not available for the Gila River case. Because occupations unrelated to a given period of interest can inflate site-area estimates, we employ several bias-control measures. For the altiplano settlement systems, a site's area is included in a given analysis only if 50 percent or more of the diagnostic projectile points for that site can be assigned to the period of interest. In doing so, it is likely that the site's area is primarily a function of occupation during the period of interest.

In the Jequetepeque case where flaked stone tool traditions are more constrained in time, we take three approaches to control for site-area inflation. Each approach reflects a tradeoff between sample confidence and sample size. First, we examine all sites with one or more Paijan artifacts. Second, only sites identified as single component with Paijan artifacts are examined.

Third, sites with one or more Paijan artifacts and excluding sites with ceramic artifacts are examined because ceramics are associated with later agricultural occupations.

Surface-artifact count provides the second estimate of settlement size. To avoid count over-estimation from extraneous periods, only temporally diagnostic artifacts for the periods of interest are used. Temporally diagnostic artifacts include projectile points in all cases. In the

Jequetepeque case, other temporally diagnostic tool types allow for the consideration of additional metrics including (a) total flaked stone tool counts and (b) projectile-point and *limace* counts. Limaces are a unifacial, steep-edged flaked stone tools of uncertain function and unique to Paijan culture (Maggard 2011). Other flaked stone tools include tool preforms and scrapers.

The total flaked stone tool counts include all bifaces plus limaces. In the north coast of Peru,

bifacial flaked stone tools generally do not extend beyond the period of Paijan occupation and therefore can be considered temporally diagnostic. In highlands Peru and the U.S. Southwest, this is not the case—bifacial lithic technology extends well beyond periods of interest and therefore are not temporally diagnostic at the level required of this analysis.

80

**Power-law Analysis of Settlement-Size Variation**

Each data set is analyzed in six steps in an effort to reject the hypothesis of power-law scaling. First, the structure is subjectively assessed using cumulative density function (CDF) plots with logarithmic axes. Power-law distributions generate linear trends in such plots (i.e., they are log-linear) (Newman 2005). As long as log-linearity is approached over some portion of the data, we move to more-rigorous assessments of power-law structure.

Second, we use maximum likelihood estimation (MLE) to find the best-fit model parameters for each candidate statistical model (Clauset et al. 2009). For site-area data, which consist of continuous data measured in square meters, the candidate set of statistical models that we consider include normal, exponential, lognormal, and power-law (Pareto) distributions.

Because artifact-count data are discrete integer data, we consider Poisson, geometric, and powerlaw (Zeta) distributions. Poisson and geometric distributions are discrete-data analogs to normal and exponential distributions, respectively (Johnson et al. 2005). Each of these statistical models has seen explicit or implicit use in the study of human settlement-size variation (e.g., Brown et al. 2006; Drennan and Peterson 2004; Johnson 1980) and therefore require consideration in our effort to reject power-law hypotheses.

81

Third, we assess the statistical plausibility of each model fit to the data using

Kolmogorov-Smirnov (KS) measures of statistical distance and Monte Carlo simulation as described by Clauset et al. (2009). The tests quantify the probability that the difference between the data and a given best-fit statistical model results from statistical chance alone. If the probability is low (*p < *0.10), then there is a high likelihood that the model and the data do not represent the same statistical structure, and the model is rejected. If a best-fit model and the data do not produce a statistically significant difference (*p > *0.10), then the model is considered plausible.

Fourth, we compare the relative information content of each statistically plausible model using Akaike information criterion (AIC) and AIC weights following the method described by

Edward's et al. (2007). Models that generate weak AIC weights (*w* < 0.10) are rejected in favor of those that produce strong AIC weights (*w* > 0.10).

Unlike other statistical distributions, theoretical power-law distributions are scale invariant meaning that they obtain over an infinite range of scales. Real-world phenomena, however, invariably have finite size limits that restrict the potential range of applicability of power-law scaling (Newman 2005). As a result, even in cases where power-law models are rejected for the full range of data, the possibility remains that power-law structure pertains to some upper-tail fraction of the data (Newman 2005). Defining this range is an analytical problem that must be addressed. The fifth step applies the iterative KS-test method of Clauset et al. (2009) to find the most-probable threshold value, *x*

*min*

, for the hypothesized power-law tail of a given data sample. We then use MLE to find a best-fit power-law model for the tail and a Monte Carlo

82 simulation to assess the statistical plausibility of the model. An upper-tail power-law model is rejected if the difference between the data and the model are unlikely to be explained by statistical chance (*p < *0.10).

In some cases, the methods described above yield a p-value in the range of statistical plausibility but a power-law scaling parameter that exceeds the upper limit of acceptable values

(*α >* 3). Such values are theoretically problematic because they describe distributions that are not scale invariant and thus converge on non-power law distributions (Brown et al. 2005). Moreover, such values are greater than those found to describe settlement structure in complex societies.

For these reasons, an otherwise statistically plausible power-law model is rejected if the scaling parameter is not less than three.

RESULTS

For the discrete artifact count data, CDF plots reveal clear log-linearity in all cases

suggesting potential power-law structure (Figure 8). The Monte Carlo and KS-test procedure

Conversely, the procedure is unable to reject power-law structure in seven of ten cases. A controlled analysis of the synthetic data (see supplementary material) demonstrates that the method is highly unlikely to derive the observed results by statistical chance. We therefore conclude that power-law scaling provides a plausible and parsimonious characterization of the size distribution of hunter-gatherer settlements when settlement-size is measured in terms of artifact counts.

*Figure 8. Cumulative density function plots for site-artifact counts (discrete data) on log-log axes. The log-linear structure is consistent with power-law structure.*

83

84

*Table 1. The prevalence of plausible statistical models for the full data*

*range of the eight settlement systems. See Table 2 for source data.*

**data type**

continuous

(site area)

**statistical model**

*n*

**proportion of total**

normal 1 0.1

exponential 6 0.5

lognormal 5 0.4

power law (Pareto) 0 0.0

none 0 0.0

discrete

(artifact count)

Poisson total 12* 1.0

1 0.1

geometric 0 0.0

power law (Zeta) 7 0.6

none 2 0.3

total 10* 1.0

*total exceeds the number of test cases due to multiple plausible statistical models for individual settlement systems.

For the continuous site-area data, CDF plots reveal convex structure over the full range of

the data, suggesting an absence of power-law structure (Figure 9). Moreover, the more rigorous

methods do not favor power-law distributions in any of the results. Exponential and/or lognormal distributions are favored in a total of seven of the nine test cases. A normal distribution is favored in one case, and one case did not produce a plausible or parsimonious fit to any of the models considered. It is clear that power-law functions are highly unlikely to characterize the full range of settlement-size data when settlement size is measured by areal extent.

Although power-law models cannot account for the full range of settlement-area data, the possibility remains that power-law scaling describes upper-tail values. Model plausibility values indicate that we cannot rule out power-law structure in the upper tails for 3 of the 9 cases.

However, a controlled analysis of the synthetic data demonstrates a high probability that this

85

*Table 2. Best-fit statistical models for each archaeological test case. Supported models include those that are statistically plausible and parsimonious given the alternative models. Tables s1 and s2 of the supplementary materials present analytical results that support the conclusions summarized in this table.*

**test case environment**

Lake Titicaca altiplano

Gila River

Sonoran

Desert

Jequetepeque coastal plain

**settlement system**

Ilave Basin

Late

Archaic

Río

Huenque

Late

Archaic

Ilave Basin

Middle

Archaic

Río

Huenque

Middle

Archaic

Ilave Basin

Early

Archaic

Río

Huenque

Early

Archaic

Middle

Archaic

Paijan

**metric**

site area

(ha)

**constraints**

site area

(ha) projectile point count sites > 50% Late Archaic points temporally diagnostic projectile points site area

(ha) projectile point count site area

(ha) sites > 50% Late Archaic points temporally diagnostic projectile points sites > 50% Middle

Archaic points projectile point count site area

(ha) projectile point count site area

(ha) projectile point count site area

(ha) projectile point count projectile point count temporally diagnostic projectile points sites > 50% Middle

Archaic points temporally diagnostic projectile points sites > 50% Early Archaic points temporally diagnostic projectile points sites > 50% Early Archaic points temporally diagnostic projectile points temporally diagnostic projectile points stone tool counts

**sum of values**

20.3 ha

**site count**

51

463 pps* 70

10.0 ha

235 pps

3.2 ha

64 pps

8.0 ha

148 pps

1.4 ha

33 pps

6.1 ha

96 pps

94 pps

38

83

16

36

20

64

9

20

19

55

50

**supported statistical model(s)**

exponential power law exponential lognormal

Poisson normal power law exponential lognormal none exponential lognormal power law plausible exponential lognormal implausible implausible power law plausible power law

**power-law tail**

implausible plausible implausible implausible implausible implausible plausible plausible plausible all sites with one or more

Paijan-diagnostic artifacts excluding sites with non-

Paijan artifacts

254.6 ha 126

17.6 ha 23 none exponential plausible implausible excluding sites with ceramics

54.6 ha 91 lognormal plausible all flaked stone tools 646 tools 126 none

Paijan points and limaces 237 tools 65 power law plausible plausible

Paijan points 163 pps 45 power law plausible

*pps=projectile points

86

*Figure 9. Cumulative density function plots for site-area (continuous) data on log-log axes. The upwardly convex data structure suggests an absence of power-law structure for the full range of data.*

result could be obtained by statistical chance alone (see supplementary material). Given the sample sizes, likely models for the full range of the data, and their parameter values, the method is highly likely to erroneously identify power-law structure in the upper tails of the data.

87

In conclusion, we are unable to reject power-law scaling in any of the three environmental contexts and eight settlement systems examined when settlement size is measured by artifact counts. However, we reject power-law structure for settlement-size variation that is measured by areal extent. Instead, the finding of exponential/lognormal structure indicates that the structure of site-area variation lacks the extreme variation observed in the artifact-count data and the site-area variation of complex societies.

DISCUSSION

How can we explain the observed similarities and differences in settlement-size structure among such fundamentally disparate economic systems? The extreme variation in the areal extents of settlements is complex societies can be readily understood as a function of extreme variation in contemporaneous settlement populations because more co-resident individuals simply occupy more horizontal space. The relatively constrained variation observed in huntergatherer site areas suggests that co-resident group size were more constrained. This may be partly a function of the fact that hunter-gatherer populations tend to be smaller than agricultural populations in a given environment, which constrains the upper limits of co-resident group sizes and thus the horizontal space occupied at any given moment. More importantly, hunter-gatherer subsistence resources cannot be intensified to the same degree as agricultural resources, further limiting co-resident group sizes.

A site's total occupation time, conversely, is not so constrained in hunter-gatherer settlement systems. A given settlement could experience extreme material growth through longterm use whether continuous or incremental. In light of attenuated site-area variation, the

88 extreme artifact-count variation is best understood as the result of variation in total occupation duration as opposed to variation in group size. In other words, whereas site-size variation in the settlement systems of complex societies appears to be a synchronous phenomenon with the full range of site-size variation extant at one point in time, in hunter-gatherer settlement systems it appears to have been asynchronous with the full range of site-size variation distributed through time.

These conclusions suggest a trajectory for the emergence of settlement-size hierarchies in human societies. As hunter-gatherer land use intensified and residential mobility decreased, asynchronous power-law structure may have given way to the synchronous structure that characterizes settlement hierarchies in settled agrarian and industrial societies. If so, settlementsize hierarchy may appropriately be thought of as creating a context for emergent socioecomomic complexity rather than the other way around. This finding does not undermine previous models that use complex socioeconomic behaviors such as agriculture, manufacturing, and warfare to predict settlement-size hierarchy (e.g., Brown and Witschey 2003; Griffin and

Stanish 2007; Krugman 1996), but it does suggest that such complex behaviors may be proximate to more fundamental social behaviors.

Hunter-gatherer research now faces the challenge of explaining the structural properties described here. We currently lack models of hunter-gatherer mobility, social interaction, and/or site formation that explicitly predict the observed structure of settlement-size variation.

Efficacious models will be those that predict (1) power-law statistical structure in the sizedistributions of settlements as measured by cumulative occupation time and (2) exponential or

89 lognormal structure in terms of co-resident group-size variation. We hope that future modeling efforts take strides to logically link hunter-gatherer behavior to the observed structural properties that fundamentally shape the organization of human societies.

REFERENCES CITED

Adamic, Lada

2011 Complex systems: Unzipping Zipf’s law. *Nature* 474(7350): 164–165.

Aldenderfer, Mark S.

1998 *Montane Foragers: Asana and the South-Central Andean Archaic*. University of Iowa

Press, Iowa City.

2008 High Elevation Foraging Societies. In *Handbook of South American Archaeology*, edited by Helaine Silverman and William H. Isbell, pp. 131–143. Springer, New York NY.

Aldenderfer, Mark S., and Luis A. Flores Blanco

2011 Reflexiones para Avanzar en los Estudios del Periodo Arcaico en los Andes Centro-Sur.

*Chungara: Revista de Antropología Chilena* 43(1): 531–550.

Ames, Kenneth M

2008 The Archaeology of Rank. In *Handbook of Archaeological Theories*, edited by R.

Alexander Bentley and Hebert D.G. Maschner, pp. 487–513. AltaMira Press.

Bandy, Matthew S.

2006 Early Village Society in the Formative Period in the Southern Lake Titicaca Basin.

*Andean archaeology* 3: 210–236.

Batty, Michael

2008 The Size, Scale, and Shape of Cities. *Science* 319(5864): 769–771.

Bettencourt, Luís M. A., José Lobo, Dirk Helbing, Christian Kühnert, and Geoffrey B. West

2007 Growth, Innovation, Scaling, and the Pace of Life in Cities. *Proceedings of the *

*National Academy of Sciences* 104(17): 7301–7306.

Bettencourt, Luís M. A., José Lobo, Deborah Strumsky, and Geoffrey B. West

2010 Urban Scaling and Its Deviations: Revealing the Structure of Wealth, Innovation and

Crime across Cities. *PLoS ONE* 5(11): e13541.

Bettencourt, Luis M.A., José Lobo, and Deborah Strumsky

2007 Invention in the City: Increasing Returns to Patenting as a Scaling Function of

Metropolitan Size. *Research Policy* 36(1): 107–120.

Bonabeau, Eric, Laurent Dagorn, and Pierre Fréon

1999 Scaling in Animal Group-Size Distributions. *Proceedings of the National Academy of *

*Sciences* 96(8): 4472 –4477.

Brown, Clifford T., Larry S. Liebovitch, and Rachel Glendon

2006 Lévy Flights in Dobe Ju/’hoansi Foraging Patterns. *Human Ecology* 35(1): 129–138.

Brown, Clifford T., and Walter R. T. Witschey

90

2003 The Fractal Geometry of Ancient Maya Settlement. *Journal of Archaeological Science*

30(12): 1619–1632.

Brown, Clifford T., Walter R. T. Witschey, and Larry S. Liebovitch

2005 The Broken Past: Fractals in Archaeology. *Journal of Archaeological Method and *

*Theory* 12(1): 37–78.

Carneiro, Robert L.

2000 The Transition from Quantity to Quality: A Neglected Causal Mechanism in Accounting for Social Evolution. *Proceedings of the National Academy of Sciences* 97(23): 12926–

12931.

Christaller, Walter

1966 *Central Places in Southern Germany*. Trans. Carlisle W. Baskin. Prentice-Hall,

Englewood Cliffs N.J.

Clauset, Aaron, Cosma Rohilla Shalizi, and M. E. J Newman

2009 Power-Law Distributions in Empirical Data. *SIAM Review* 51: 1–43.

Decker, Ethan H., Andrew J. Kerkhoff, and Melanie E. Moses

2007 Global Patterns of City Size Distributions and Their Fundamental Drivers. *PLoS ONE*

2(9): e934.

Dillehay, Tom D

2008 Profiles in Pleistocene History. In *Handbook of South American Achaeology*, edited by

Helaine Silverman and William H. Isbell, pp. 29–43. Springer, New York NY.

Dillehay, Tom D., Jack Rossen, Greg Maggard, Kary Stackelbeck, and Patricia Netherly

2003 Localization and Possible Social Aggregation in the Late Pleistocene and Early

Holocene on the North Coast of Perú. *Quaternary International* 109–110(0): 3–11.

Drennan, Robert D., and Christian E. Peterson

2004 Comparing Archaeological Settlement Systems with Rank-Size Graphs: A Measure of

Shape and Statistical Confidence. *Journal of Archaeological Science* 31(5): 533–549.

Edwards, Andrew M., Richard A. Phillips, Nicholas W. Watkins, Mervyn P. Freeman, Eugene J.

Murphy, Vsevolod Afanasyev, Sergey V. Buldyrev, M. G. E. da Luz, E. P. Raposo, H. Eugene

Stanley, and Gandhimohan M. Viswanathan

2007 Revisiting Levy Flight Search Patterns of Wandering Albatrosses, Bumblebees and

Deer. *Nature* 449(7165): 1044–1048.

Flannery, Kent V.

1998 The Ground Plans of Archaic States. In *Archaic states*, edited by Gary M. Feinman and

Joyce Marcus, pp. 15–57. School of American Research Press, Santa Fe.

Fujita, Masahisa, Paul R Krugman, and Anthony Venables

1999 *The Spatial Economy: Cities, Regions and International Trade*. MIT Press, Cambridge,

Mass.

Gan, Li, Dong Li, and Shunfeng Song

2006 Is the Zipf Law Spurious in Explaining City-size Distributions? *Economics Letters*

92(2): 256–262.

Griffin, Arthur F., and Charles Stanish

2007 An Agent-Based Model of Prehistoric Settlement Patterns and Political Consolidation in

91 the Lake Titicaca Basin of Peru and Bolivia. *Structure and Dynamics: eJournal of *

*Anthropological and Related Sciences* 2(2): 1–47.

Grove, Matt

2010 The Quantitative Analysis of Mobility: Ecological Techniques and Archaeological

Extensions. In *New Perspectives on Old Stones: Analytical Approaches to Paleolithic *

*Technologies*, edited by Stephen J. Lycett and Parth R. Chauhan, pp. 83–118. Springer,

New York.

Hamilton, Marcus J., Bruce T. Milne, Robert S. Walker, Oskar Burger, and James H. Brown

2007 The Complex Structure of Hunter-Gatherer Social Networks. *Proceedings of the Royal *

*Society B: Biological Sciences* 274(1622): 2195–202.

Inomata, Takeshi, and Kazuo Aoyama

1996 Central-Place Analyses in the La Entrada Region, Honduras: Implications for

Understanding the Classic Maya Political and Economic Systems. *Latin American *

*Antiquity* 7(4): 291–312.

Johnson, Gregory A.

1980 Rank-Size Convexity and System Integration: A View from Archaeology. *Economic *

*Geography* 56(3): 234–247.

Johnson, Norman L., Adrienne W. Kemp, and Samuel Kotz

2005 *Univariate Discrete Distributions*. 3rd ed. Wiley-Interscience, August 30.

Jones, Kate E., Nikkita G. Patel, Marc A. Levy, Adam Storeygard, Deborah Balk, John L.

Gittleman, and Peter Daszak

2008 Global Trends in Emerging Infectious Diseases. *Nature* 451(7181): 990–993.

Kelly, Robert L.

1995 *The Foraging Spectrum: Diversity in Hunter-Gatherer Lifeways*. Percheron Press, New

York.

Klink, Cynthia J.

2005 Archaic Period Research in the Río Huenque Valley, Peru. In *Advances in Titicaca *

*Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 13–24. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Klink, Cynthia J., and Mark S. Aldenderfer

2005 A Projectile Point Chronology for the South-Central Andean Highlands. In *Advances in *

*Titicaca Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 25–54. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Krugman, Paul

1996 *The Self-Organizing Economy*. Blackwell Publishers, Cambridge.

Kuhn, Steven L, and Stiner, Mary C.

2001 The Antiquity of Hunter-Gatherers. In *Hunter-gatherers: An Interdisciplinary *

*Perspective*, edited by Catherine Panter-Brick, Robert Layton, and P. Rowley-Conwy, pp.

99–142. Bioscocial Society Symposium Series 13. Cambridge University Press, New York.

Loendorf, Christopher R, and Glen Rice

2004 *Projectile Point Typology: Gila River Indian Community, Arizona*. Gila River Indian

Community, Cultural Resource Management Program, Sacaton.

92

Lyman, R. Lee

2004 The Concept of Equifinality in Taphonomy. *Journal of Taphonomy* 2(1): 15–26.

Maggard, Greg J

2011 Las Ocupaciones Humanas del Pleistoceno Final y el Holoceno Temprano en la Costa

Norte del Perú. *Boletín de Arqueología PUCP* 15: 1–23.

Mitzenmacher, Michael

2001 A Brief History of Generative Models for Power Law and Lognormal Distributions.

*Proceedings of the Annual Allerton Conference on Communication Control and *

*Computing* 39: 182–191.

Newman, M. E. J.

2005 Power Laws, Pareto Distributions and Zipf’s Law. *Contemporary Physics* 46(5): 323–

351.

Paynter, Robert

1982 *Models of Spatial Inequality: Settlement Patterns in Historical Archeology*. Academic

Press, New York.

Peterson, Christian E., and Robert D. Drennan

2005 Communities, Settlements, Sites, and Surveys: Regional-Scale Analysis of Prehistoric

Human Interaction. *American Antiquity* 70(1): 5–30.

Richardson, Harry

1973 Theory of the Distribution of City Sizes: Review and prospects. *Regional Studies* 7:

239–251.

Shennan, Stephen

2000 Population, Culture History, and the Dynamics of Culture Change. *Current *

*Anthropology* 41(5): 811–835.

Stanish, Charles

2003 *Ancient Titicaca: The Evolution of Complex Society in Southern Peru and Northern *

*Bolivia*. University of California Press, Berkeley.

Stiner, Mary C.

2001 Thirty Years on the “Broad Spectrum Revolution” and Paleolithic Demography.

*Proceedings of the National Academy of Sciences* 98: 6993–6996.

Wearmouth, Victoria J., Matthew J. McHugh, Nicolas E. Humphries, Aurore Naegelen,

Mohammed Z. Ahmed, Emily J. Southall, Andrew M. Reynolds, and David W. Sims

2014 Scaling Laws of Ambush Predator “Waiting” Behaviour are Tuned to a Common

Ecology. *Proceedings of the Royal Society B: Biological Sciences* 281(1782): 20132997.

Winterhalder, Bruce, and R. Brooke Thomas

1978 *Geoecology of Southern Highland Peru: A Human Adaptation Perspective*. University of Colorado, Institute of Arctic and Alpine Research, Boulder.

Wobst, H. Martin

1978 The Archaeo-Ethnology of Hunter-Gatherers or the Tyranny of the Ethnographic

Record in Archaeology. *American Antiquity* 43(2): 303–309.

Wright, H T

1977 Recent Research on the Origin of the State. *Annual Review of Anthropology* 6(1): 379–

93

397.

Yellen, John

1977 *Archaeological Approaches to the Present: Models for Reconstructing the Past*.

Academic Press, New York.

Zipf, George Kingsley

1949 *Human Behavior and the Principle of Least Effort: An Introduction to Human Ecology*.

Addison-Wesley Press, Cambridge.

SUPPLEMENTARY MATERIAL

Data analyses used to support the reported results in the main body of this paper are presented here along with a meta-analysis of the procedure's efficacy.

**Supplementary Analytical Results**

Table s1 summarizes each data sample, MLE model-fit parameters, goodness-of-fit for each model, and their statistical plausibilities. For cases where more than one statistical model was identified as plausible for the full range of the data, Akaike information criterion (AIC) and

AIC weights (*w*) were used to identify the most parsimonious solution following the method of

Edwards et al. (2007). The numerical results are presented in Table s2 and summarized in Table

*Table s1. Maximum likelihood estimated model parameters and statistical plausibility of their fits to the data. Plausible models (p > 0.10) are highlighted.*

**case metric constraints**

*Σ*

**sites statistical model model parameters KS**

**D**

*p*

Titicaca Basin

Ilave Late

Archaic site area sites > 50%

Late Archaic points

20.3 ha 51 normal exponential lognormal

11.3 ha 11

*µ=*3983, *σ=*4462

*λ=*2.51x10

-4

0.21 0.00

0.08 0.72

0.13 0.02

*µ=*7.70, *σ=*1.34

power law *α*=1.18, *xmin*=9 0.43 0.00

power-law tail *α*=3.27*, *xmin*=6231 0.07 0.98

94

**case metric**

projectile point count

**constraints**

temporally diagnostic projectile points

*Σ*

**sites statistical model**

463 pps 70 Poisson

**model parameters KS**

**D**

*λ=*6.61

*p*

0.52 0.00

geometric power law

*prob=*0.13

*α*=1.68, *xmin*=1

371 pps 19 power-law tail *α*=2.13, *xmin*=6

0.25 0.00

0.08 0.11

0.07 0.62

Titicaca Basin

Huenque Late

Archaic

Titicaca Basin

Ilave Middle

Archaic site area projectile point count sites > 50%

Late Archaic points temporally diagnostic projectile points site area sites > 50%

Middle

Archaic points

10.0 ha 38 normal exponential lognormal

*µ=*2637, *σ=*1919

*λ=*3.79x10

-4

0.11 0.35

0.12 0.35

0.13 0.14

*µ=*7.53, *σ=*0.94

power law *α*=1.47, *xmin*=236 0.29 0.00

6.9 ha 15 power-law tail *α*=3.91*, *xmin*=3204 0.13 0.48

235 pps 83 Poisson *λ=*2.83

0.23 0.81

81 pps 11 geometric power law

*prob=*0.26

*α*=1.83, *xmin*=1 power-law tail *α*=10.75*, *xmin*=7

3.2 ha 16 normal exponential lognormal

*µ=*2017, *σ=*1098

*λ=*4.96x10

-4

*µ=*7.25, *σ=*1.18

0.45 0.00

0.12 0.01

0.02 0.95

0.14 0.54

0.27 0.03

0.16 0.00

projectile point count temporally diagnostic projectile points

1.9 ha 6 power law *α*=1.24, *xmin*=30 0.41 0.01

power-law tail *α*=6.25*, *xmin*=2632 0.17 0.50

64 pps 36 Poisson geometric power law

*λ=*1.78

*prob=*0.36

*α*=2.35, *xmin*=1

Titicaca Basin

Huenque

Middle

Archaic site area sites > 50%

Middle

Archaic points

41 pps 13 power-law tail *α*=3.08*, *xmin*=2

8.0 ha 20 normal exponential

*µ=*4040, *σ=*4702

*λ=*2.48x10

-4 lognormal power law

*µ=*7.61, *σ=*1.34

*α*=1.31, *xmin*=94 power-law tail *α*=2.98, *xmin*=5479 5.7 ha 6 projectile point count temporally diagnostic projectile points

148 pps 64 Poisson geometric power law

*λ=*2.31

*prob=*0.30

*α*=2.06, *xmin*=1

114 pps 30 power-law tail *α*=2.57, *xmin*=2

Titicaca Basin site area sites > 50% 1.4 ha 9 normal *µ=*1527, *σ=*1362

0.47 0.00

0.59 0.00

0.07 0.11

0.06 0.48

0.21 0.02

0.12 0.84

0.14 0.43

0.29 0.02

0.10 0.87

0.33 0.07

0.51 0.00

0.10 0.02

0.03 0.97

0.19 0.48

95

Ilave Early

Archaic

Gila River

Middle

Archaic

Jequetepeque

Paijan

**case**

Titicaca Basin

Huenque Early

Archaic

**metric**

projectile point count

**constraints**

Early

Archaic points temporally diagnostic projectile points

*Σ*

**sites statistical model**

exponential

0.9 ha 3

33 pps 20 Poisson geometric power law

**model parameters KS**

**D**

*λ=*6.55x10

-4

*p*

0.10 1.00

lognormal power law

*µ=*6.71, *σ=*1.41

*α*=1.27, *xmin*=30

0.18 0.55

0.35 0.08

power-law tail *α*=3.17*, *xmin*=2201 0.13 0.71

*λ=*1.65

*prob=*0.38

*α*=2.37, *xmin*=1

0.51 0.01

0.61 0.00

0.07 0.31

site area projectile point count sites > 50%

Early

Archaic points temporally diagnostic projectile points power-law tail *α*=2.37, *xmin*=1

6.1 ha 19 normal *µ=*3200, *σ=*2727 exponential *λ=*3.12x10

-4

2.5 ha 3

96 pps 55 Poisson geometric power law

*λ=*1.75

*prob=*0.36

*α*=2.32, *xmin*=1

0.07 0.31

0.18 0.12

0.11 0.87

lognormal power law

*µ=*7.57, *σ=*1.18

*α*=1.34, *xmin*=121

0.15 0.27

0.34 0.00

power-law tail *α*=6.60*, *xmin*=7477 0.13 0.65

0.48 0.00

0.60 0.00

0.04 0.53

projectile point count temporally diagnostic projectile points power-law tail *α*=2.32, *xmin*=1

94 pps 50 Poisson *λ=*1.88

geometric power law

*prob=*0.35

*α*=2.25, *xmin*=1 power-law tail *α*=2.25, *xmin*=1

0.04 0.53

0.44 0.01

0.57 0.00

0.07 0.14

0.07 0.14

site area all sites with one or more

Paijandiagnostic artifacts

254.6 ha 126 normal exponential

238.7 ha 53 lognormal power law power-law tail

*µ=*20205, *σ=*55992

*λ=*4.95x10

-5

*µ=*8.39, *σ=*1.73

*α*=1.26, *xmin*=100

*α*=1.87, *xmin*=7209

0.36 0.00

0.32 0.00

0.08 0.08

0.28 0.98

0.06 0.97

0.18 0.05

excluding sites with non-Paijan artifacts

17.6 ha 23 normal exponential lognormal power law

*µ=*7674, *σ=*6447

*λ=*1.30x10

-4

*µ=*8.41, *σ=*1.18

0.14 0.47

0.19 0.03

0.24 0.05

*α*=1.40, *xmin*=418

8.8 ha 5 power-law tail *α*=6.91*, *xmin*=15200 0.13 0.62

excluding 54.6 ha 91 normal *µ=*6005, *σ=*11561 0.30 0.00

96

**case metric**

stone tool count

**constraints**

sites with ceramics all flakestone tools

Paijan points and limaces

*Σ*

490 tools

237 tools

**sites**

33

65

**statistical model**

exponential lognormal power law power law power-law tail

Poisson geometric power law

**model parameters KS**

**D**

*λ=*1.67x10

-4

*p*

0.23 0.00

*µ=*7.79, *σ=*1.40

*α*=1.31, *xmin*=100

0.08 0.15

0.29 0.00

0.10 0.32

42.7 ha 27 power-law tail *α*=2.42, *xmin*=6250

646 tools

126 Poisson geometric

*λ=*5.13

*prob=*0.16

*α*=1.73, *xmin*=1

*α*=2.21, *xmin*=5

*λ=*3.65

*prob=*0.22

*α*=1.88, *xmin*=1

0.48 0.00

0.30 0.00

0.07 0.06

0.07 0.36

0.44 0.00

0.38 0.00

0.05 0.34

power-law tail *α*=1.88, *xmin*=1

Paijan points 163 pps 45 Poisson *λ=*3.62

geometric *prob=*0.22

power law *α*=1.85, *xmin*=1 power-law tail *α*=1.83, *xmin*=1

*model considered implausible due to alpha values greater than 3.

0.05 0.34

0.41 0.00

0.39 0.00

0.07 0.34

0.07 0.34

97

*Table s2. Results of Akaike information criterion (AIC) and AIC weight (w) analysis. The analysis only includes cases where two or more statistical models were found to be statistically plausible. Favored results (w > 0.10) are highlighted. *

**case metric constraints plausible statistical models AIC AIC weight**

Rio Huenque Late

Archaic site area sites > 50% Formative points normal exponential

686 0.06

677 0.63

Rio Huenque Middle

Archaic

Ilave Basin Early

Archaic

Rio Huenque Early

Archaic site area site area site area sites > 50% Formative points sites > 50% Formative points sites > 50% Formative points lognormal exponential lognormal normal exponential lognormal normal exponential lognormal

680 0.31

374 0.72

377 0.28

159 0.06

152 0.77

157 0.17

359 0.04

347 0.75

352 0.21

**Performance Assessment**

This meta-analysis serves two purposes. First, the tests serve to demonstrate that the methods function as intended. Second, the analysis identifies potential errors given the alternative models considered, sample sizes, and probable model parameters. Robust archaeological samples are extremely difficult and costly to generate from sites deposited thousands of years ago by small-scale hunter-gatherer populations who left few archaeologically stable materials on the landscape. The archaeological samples used in this study are therefore small by statistical standards. Nonetheless, because of the scientific and cultural importance of the samples and the costliness of obtaining them it is appropriate to explore them as they exist with emphasis on understanding their limitations.

98

The meta-analysis consists of nine tests—one for each of the statistical models considered in this study. In each test, random samples are drawn from known statistical distributions and analyzed with the same code used to analyze the empirical data. Sample sizes and the parameters for the synthetic data are selected at random from the set of sample sizes and parameters observed in the empirical data. The procedure is iterated 100 times to evaluate how many correct and incorrect model identifications are observed. The fraction of correct and incorrect results gives us a probabilistic sense of the efficacy of the procedure, which we can then use to evaluate the robustness of the conclusions reached for the empirical data.

The discrete data results are summarized in Table s3. The results demonstrate that the procedure is (1) efficacious in identifying the underlying structure of discrete data, (2) highly unlikely to fail to identify power-law structure given power law data, and (3) highly unlikely to find power-law structure given non-power law data.

The procedure correctly identifies 90 percent of the synthetic power law data as consistent with power-law models. Moreover, power-law structure is identified in the upper tails of 97 percent of the synthetic power-law samples. Power-law structure was never incorrectly identified in Poisson data samples, and only once (1 percent) in the geometric data samples.

These meta-analysis results suggest an exceedingly small chance that the 7 test cases identified as consistent power-law models came from Poisson or geometric data. Moreover, the fact that 10 of the 100 of the synthetic power law models were misidentified as inconsistent with all of the considered models suggests that the 2 of 8 discrete empirical cases found to be

inconsistent with all considered models could conceivably be power-law distributed data that were misidentified by the procedure (*x*

*2*

*=*1.43, df=1, *p=*0.23).

99

*Table s3. Probabilities of identifying data models from random samples drawn from known data models—discrete-data cases. Based on 100 simulations per data model using sample sizes and model parameter values observed in the empirical data.*

**known discrete data model analytical conclusions power law power-law tail Poisson geometric none**

power law

Poisson geometric

0.90

0.00

0.01

0.97

0.00

0.29

0.00

0.87

0.37

0.00

0.00

0.74

0.10

0.13

0.01

The continuous-data experimental results are presented in table s4. The results demonstrate that the procedure is (1) efficacious in identifying the underlying structure of continuous data given the models considered, (2) highly unlikely to fail to identify power-law structure given power law data, and (3) unlikely to find power-law structure given non-power law data. The procedure correctly identifies each respective data model in most cases. Given samples drawn from power-law models with parameter values in the range of the empirically estimated values, the procedure correctly identifies 96 percent as consistent with power-law models. Moreover, power-law structure is identified in the upper tails of 98 percent of the synthetic power-law data samples.

100

*Table s4. Probabilities of identifying data models from random samples drawn from known data models—continuous-data cases. Based on sample sizes and model parameter values observed in the empirical data. 100 simulations per data model.*

**known continuous data model analytical conclusions power law power-law tail normal exponential lognormal none**

power law normal exponential lognormal

0.96

0.05

0.13

0.13

0.98

0.04

0.32

0.73

0.00

0.57

0.14

0.03

0.00

0.29

0.85

0.30

0.04

0.25

0.42

0.87

0.00

0.29

0.02

0.00

However, the procedure also incorrectly identifies models with low frequency. Because the empirical data show that power-law structure is highly unlikely to obtain over the full range of empirical data but plausible for the upper tails of the data, we are most concerned here with how likely power-law structure is to be identified in the upper tails of non-power-law samples.

The procedure incorrectly finds power-law structure in the upper tails of 73 percent of the lognormal data samples. The same misidentification occurs in exponential samples 32 percent of the time. Normally distributed data generate upper tails that are mistakenly identified as powerlaw distributed 4 percent of the time.

Recall that the archaeological site-area data produced 3 of 9 cases with plausible powerlaw structure in the upper tails of the distributions. Given that (a) exponential, lognormal, and normal structure is found to be consistent for the full range of data 6, 5 and 1 times, respectively, in the 9 continuous archaeological data samples and (b) the proportion of times we expect those distributions to generate upper tails identifiable as power-law distributed, we would expect incorrectly identified power-law structure in the upper tails in

(0.32∗6/9+0.73∗5/9+0.04∗1/9+1.00∗1/9)∗9=7 of 9 data samples. This expectation more than accounts for the 3 of 9 archaeological data samples with plausible power-law tails, leading

101 us to conclude that the observed plausibility of power-law structure in the upper tails may simply be an artifact of sample uncertainty.

**Reference Cited**

Edwards, Andrew M., Richard A. Phillips, Nicholas W. Watkins, Mervyn P. Freeman, Eugene J.

Murphy, Vsevolod Afanasyev, Sergey V. Buldyrev, M. G. E. da Luz, E. P. Raposo, H. Eugene

Stanley, and Gandhimohan M. Viswanathan

2007 Revisiting Levy Flight Search Patterns of Wandering Albatrosses, Bumblebees and

Deer. *Nature* 449(7165): 1044–1048.

APPENDIX II:

FORAGER MOBILITY IN CONSTRUCTED ENVIRONMENTS

102

103

ABSTRACT

As obligate tool users, humans habitually reconfigure material-resource distributions. It is proposed here that such resource restructuring may have played an important role in shaping hunter-gatherer mobility decisions and the emergent macro-structure of settlement patterns. This paper presents a model of hunter-gatherer mobility in which modifications of places, including previously deposited cultural materials, bias future mobility decisions. With the aid of an agentbased model, this niche-construction model is used to deduce six hypotheses for the structure of hunter-gatherer settlement patterns. The predictions are tested against archaeological data from a hunter-gatherer settlement system in the Lake Titicaca Basin, Peru, 7,000-5,000 cal. B.P. Good agreement is found between the predicted and empirical patterns demonstrating the model's efficacy and suggesting an explanation for key structural properties of hunter-gatherer settlement systems that, in some cases, facilitated the emergence of socioeconomic complexity in human societies.

INTRODUCTION

The structure of the natural environment accounts for many dimensions of forager mobility and land use. Environmental proxies such as biomass productivity, precipitation, temperature, and latitude are demonstrably successful in predicting population size, residentialmove frequency, residential move distances, territory size, and group size (Binford 2001;

Hamilton et al. 2007; Kelly 1995; Kelly et al. 2013; Smith 1981). They also provide great insight into broad archaeological settlement and subsistence patterns across environmental gradients

104

(Bettinger et al. 2009; Morgan 2009; Richerson et al. 2001; Winterhalder et al. 2010; Zeanah

2002). Despite these explanatory successes, considerable settlement variation within environmental contexts remains unexplained by environmental structure that is external to hunter-gatherer behavior. It is axiomatic that archaeological materials are unevenly distributed within resource patches; they occur in material clusters, or sites, of variable size. Often such clusters do not clearly correspond to environmental features within broader environmental contexts. Archaeologists commonly call upon an array of explanations to account for a site's particular location and material density. Burial, erosion, collection bias, looting, and naturalresource configurations (extant or extinct) are a few of the more-commonly invoked explanations

(Grøn 2012; Schiffer 1987). Yet, such post-hoc interpretations offer limited predictive power, thus limiting their theoretical utility. On one hand, we could endeavor to construct complicated models of site formation, integrating the many external variables described above in the hope of arriving at a satisfactory level of predictive utility. On the other hand, we might concede that the environmental and social contingencies that go into settlement location and occupation intensity are largely beyond archaeological detection and satisfactory explanation. A possibility between these theoretical extremes is suggested here.

While it is acknowledged that the determinants of initial site selection may be empirically inaccessible, subsequent site-selection may be relatively predictable if we consider that foragers tend to reuse locations. John Yellen observed that !Kung logistic-foraging groups regularly reused previous camps to take advantage of existing brush structures (Yellen 1977). In other words, prior cultural modifications to the !Kung landscape biased subsequent land-use decisions.

105

A more explicit treatment of this type of behavior is found in Lewis Binford's (1982)* *

*Archaeology of Place*. Binford stated that, “...aside from certain 'absolute' characteristics of biogeography within the region, *there is always a 'cultural geography'*...” (Binford

1982:6[emphasis in original]). In particular, he observed that Nunamiut logistical camps used at time 1 could become residential camps at time 2 and vice versa.

Putting such observations to theoretical work may seem problematic for several reasons.

First, “cultural geographies” tend to be assumed rather than explicitly grounded in higher level theory. Binford, for example, does not specify why foragers should reuse previous site locations, though the underlying logic is ostensibly pragmatic reuse of culturally deposited materials and infrastructure. Second, the observations are often embedded in models that emphasize the effects of external environmental structure on forager mobility. It is therefore difficult to disentangle the relative roles that natural and constructed environments play in shaping behavior and material outcomes. Third, “cultural geographies” present an analytical challenge. How does one test for relationships between human behavior and environmental structure that is created by that very human behavior? More-conventional analyses do not suffer from this seemingly circular problem because forager behavior and natural environment are independent variables. In so-called cultural geographies, dependent and independent variables are inextricable. What is needed to theoretically ground the cultural geography concept and put it to work in behavioral prediction is

(1) a theoretical umbrella that makes sense of why mobile foragers should ever re-occupy specific locations on landscapes and (2) a formal model and analytical tools to generate testable predictions for recursive human behavior and its material correlates.

106

A WORKING NICHE-CONSTRUCTION MODEL OF FORAGER MOBILITY

Anatomically modern humans are obligate tool users. I use the term tool broadly here to include the range of extrasomatic materials that humans modify and rely on to maintain homeostasis. Implements and constructions of stone, fiber, shell, bone, skin, and earth are used to regulate body temperature, obtain calories, court, and parent. Given the requirement of tool use, search-and-handling costs involved in obtaining the requisite raw materials and modifying them into useful objects can affect human fitness. Individuals who can minimize such costs can put the time savings to work in more-direct fitness pursuits. One way that foragers can minimize the search-and-handling costs of tool production is through the organization of mobility. Often in hunter-gatherer research, the cost of material acquisition is modeled in terms of natural-resource distributions on a pristine landscapes. However, humans routinely rearrange landscapes, transporting and modifying raw materials to create implements, storage pits, huts, hearths, and other cultural features. Assuming that at least some of these translocated materials and constructions retain utility after abandonment, whether intentional or unintentional, the expanded distributions of materials and labor have the potential to reduce future search-and-handling costs for those materials and products. In other words, by moving and modifying materials on the landscape, people actively construct the ecological niches to which they adapt (Kendal et al.

2011; Odling-Smee et al. 1996).

To the extent that there is benefit in recycling previously constructed facilities or previously deposited materials, a forager's use of space may become biased to the locations of those materials. If a brush hut, grinding slab, and flaked stone are deposited at some location

107 within a hunting patch at year 1, and a portion of those materials retains utility, then a future forager exploiting the same patch at year 2 could benefit by reusing that location. In turn, the forager is likely to deposit additional materials and invest additional energy into infrastructure, thereby subsidizing additional future use of the location, and so on. Foragers can thus become locked in to using otherwise unexceptional locations, potentially generating considerable site growth. Such site reuse could be quite persistent as well. Culturally deposited materials may loose utility by natural decay, consumption, or economic obsolescence, but serial deposition of materials and/or the deposition of relatively durable materials could be expected to continually draw individuals to a specific location over the duration of the economic systems existence, which in many cases could extend for hundreds or thousands of years or more if the lengths of archaeological periods offer some indication.

Of course, it is logically impossible that all mobility decisions are constrained by the landscape alterations of previous foragers. Natural-resource opportunities and competition are expected to drive occupation of novel locations. Historical contingencies such as point-of-entry into a resource patch or chance resource encounters with mobile prey are expected to condition where some occupations fall. But though the location may begin its importance as a quirk of history, humans may add to the attraction of the place through subsequent alterations that increase its future attractiveness. In recursive fashion, subsequent occupation may further increase the site's attractiveness creating a runaway site growth. Thus small differences in the initial conditions of site use can lead to radically different outcomes in overall occupation intensity. This working model is therefore non-deterministic in the sense that it does not predict

108 absolute locations or sizes of particular settlements—properties that are expected to be sensitive to initial conditions (sensu Beekman and Baden 2005; Bentley and Maschner 2003; 2008; Brown et al. 2005; Lansing 2003; Lansing and Downey 2009). Instead, the recursive nature of the model behavior holds implications for *structural properties* of variation, or relational patterns amongst entities.

GENERATING HYPOTHESES

Some of the test implications for the working niche-construction model are readily deduced from simple logical arguments. However, other arguments are not so readily formulated due to the nature of recursive processes. Each time cultural materials are deposited on a physical landscape, the fitness landscape is potentially altered. Landscape structure is therefore a moving target. How does one test for spatial relationships between an organism and environmental variables that are created by that very organism? This may seem a circular problem, but the recursive nature of the model does generate distinct, observable properties. Such properties can be inferred with the help of *agent-based models *(ABM). ABM is a computational method specifically designed to explore how iterated micro-level behaviors give rise to macro-level structure, or *emergent properties* (Bankes 2002; Premo 2006; 2007).

The ABM presented here serves to operationalize the conceptual model by exploring the net distribution of materials deposited by foragers whose mobility decisions are biased toward previously deposited cultural materials. In the ABM, a forager initially occupies a random

location on a pristine landscape (Figure 10). At each time step, the forager (a) deposits a unit of

cultural material on the landscape and (b) makes a decision about where to reside. With

109

*Figure 10: Graphical user interface for the agent-based model implementation of the working niche-construction model of forager mobility. Each small yellow dot in the model world (center) represents a forager-deposited material. This example shows a model run after the deposition of *

*100 sites in 1,998 time steps, or ticks. *All ABM is performed using NetLogo programming language (Wilensky 1999).

probability *m*, the forager decides to reside at the position of a previously deposited material.

Higher values of *m* reflect greater probability that a given occupation decision will be determined by the location of a previously deposited material. If *m* equals 0.95, for example, there is a 95percent chance at any given time step that the virtual forager will occupy the location of a culturally deposited material.

To keep the model simple, the target material is selected at random from the population of previously deposited materials. This is not to suggest that a forager's material choices in reality are random. To the contrary, the use of random model behavior simply suggests that an

110 individual's material choice may be so complicated and contingent that it is appropriately modeled as a random process. In some instances, the material that determines the residence decision is at the forager's current residential site, in which case, the forager simply continues to occupy the same location. In other cases, the targeted material is located elsewhere in the model space, requiring a residential move.

The probability that an occupation decision is not determined by the location of a previously deposited material is 1 - *m*. In such instances, the virtual forager moves their residence to a previously unoccupied location on the virtual landscape. If *m = *0.95, for example, there is a 5-percent chance that the virtual forager occupies a novel location at any given moment. To isolate the effects of the niche-construction behavior, exogenous natural resources are assumed to be randomly distributed on the landscape. Again, this is not to suggest that realworld natural resource distributions are necessarily random. Rather, it is assumed that resource distributions combined with the foragers' decisions to pursue them may be so complicated and contingent as to be appropriately modeled as random.

It is difficult to know a priori what range of biases, *m*, are plausible. We can readily imagine that a forager's bias toward previously deposited cultural materials would depend on a number of factors related to environmental structure. In environments where resource availability is limited, foragers could be expected to rely more heavily on previously deposited resources, thus requiring a model with a relatively high value of *m*. In resource rich environments, we might expect that forager land use is less influenced by previously deposited cultural materials thus requiring a model with a relatively low value of *m*. Rigorous analyses of the relationship between

111 environmental structure and the relative weight of niche-construction behavior could ultimately prove insightful. However, the more immediate task at hand is to explore the degree to which niche-construction behavior is at all relevant to understanding forager mobility and settlement structure. I therefore explore a range of *m* values including a strong case (*m = *0.95), a moderate case (*m = *0.50), and a weak case (*m *= 0.05) of niche-construction behavior.

As materials are deposited on the virtual model landscape, their positions are altered slightly to model the physical constraint that materials are unlikely to occupy the exact same horizontal space. The direction of the displacement is drawn at random from a uniform distribution. The distance of displacement is more difficult to model lacking theoretical guidance on how humans deposit objects across space. At a general level, there should be a higher probability of artifacts falling closer to the point of occupation than farther. For the purpose of this analysis, I explored two different statistical models for artifact dispersion—a normal distribution and an exponential distribution. The mean of the normal distribution is simply 0. The standard deviation of the normal distribution and the rate parameter of the exponential

distribution are specified by the analyst using the *jitter* variable in the ABM (see Figure 10). This

variable is specified in arbitrary model units. Higher values create greater spatial dispersion. The model world is a torus so that material displacements that extend beyond the model boundary simply cross into the opposite boundary. This strategy mitigates boundary effects.

Table 3 presents the basic NetLogo programming code with explanations in order to

illustrate the simplicity of the programmed behavior (see supplementary material for the full code). All of the major forager and artifact behaviors are modeled with nine lines of code.

112

*Table 3. Basic NetLogo ABM code used to operationalize the working niche-construction model.*

5

6

1

2

3 line

4

7

8

9 to go

ask foragers [

NetLogo code

ifelse p-move-to-material >= random 101

[set target [who] of one-of materials explanation

Initiates sequence.

Calls the foragers (in this case, one forager).

If the probability, *m*, of moving to a material is greater than or equal to a random number between 1 and 100, execute the following...

Set target destination to be a randomly selected material that was previously deposited.

move-to material target]

[setxy random-xcor random-ycor]

Move to the location of the target material.

If the probability, *m*, of moving to a material was less than the random number generated above, move to a random location.

hatch-materials 1[ Forager deposits a unit of material at their current location.

rt random 360 fd random-exponential jitter]] The material is moved in a random direction to a distance that is randomly selected from a normal or exponential distribution.

end Ends sequence.

Although the full code includes many more lines (91 in total), the additional code does not affect forager or artifact behavior, instead serving to initiate the model, record information, allow for parameter adjustment, and adjust visualization properties. The analyst can also control the length of a given simulation. Time steps, or *ticks*, are in arbitrary units. A model can be set to run for a specified number of ticks, until a certain number of sites are deposited, or until a certain number of two-dimensional sites (those with more than one artifact needed to define an area) are deposited.

It should be noted that the fundamental mechanics of the working ABM are very similar to the neutral model of cultural transmission presented by Bentley and colleagues (Bentley et al.

113

2004; 2007; Bentley and Shennan 2005; Hahn and Bentley 2003; see also Lansing and Cox

2011). Their model offers an adaptively neutral explanation for the distribution of stylistic variation in artifact assemblages. For cultural traits such as names or pottery styles, individuals either copy an extant variant with probability *p* or they invent a novel variant with probability

*1 - p*. Stochastic effects in the cultural transmission process alone generate dominant variants over time. In the same way, stochastic effects in material-biased mobility are hypothesized to generate the observed variation in site sizes in the working model presented here.

In sum, the simple working ABM for forager mobility posits that a forager continually decides where to reside on a virtual landscape, and that decision is biased by the location of previously deposited cultural materials. While the forager moves through the virtual landscape, s/he deposits cultural materials at a constant rate. The resultant distribution of those materials are used to predict structural properties of archaeological settlement patterns.

MODEL HYPOTHESES

In this section, I use the working niche-construction model to deduce six hypotheses for structural properties of archaeological hunter-gatherer settlement patterns. In general, the hypotheses are related to the spatial and frequency distributions of artifacts and sites in a given hunter-gatherer settlement system. Several of these hypotheses predict axiomatic properties of the archaeological record. Yet, I suggest that such properties are important for what they can tell us about hunter-gatherer behavior and our assumptions about them. The proposition of a model to account for the archaeological patterns does not imply that the current model offers the only explanation each of the patterns. To the contrary, alternative explanations can be readily

114 imagined to explain some or all of the hypotheses presented below. The current model's efficacy depends on (a) its ability to predict the range of properties and (b) its ability to do so more parsimoniously than alternative models. This paper offers a starting point for thinking about the behaviors that drive several important dimensions of hunter-gatherer site variation. It is hoped that the paper will stimulate the formalization of additional tests and other models.

*Hypothesis 1: Cultural materials are spatially clustered.*

Because the working model posits that previously deposited cultural materials should attract subsequent reoccupation and, therefore, deposition of more cultural materials, resultant artifact distributions should tend to cluster. Only in the weakest cases of niche-construction should material distributions approach random distributions, and they should never reveal spatial

dispersion. Figure 11 shows ABM-generated spatial clustering of 1,000 cultural materials

deposited under different values of *m*. As *m* increases, the points qualitatively reveal greater clustering. Such variable degrees of clustering can be quantified. Spatial-data distributions can theoretically range from *dispersed* to *random* to *clustered* within a defined area (Lee and Wong

2001). At the extreme of perfect dispersion, distances among all pairs of data points are

maximized as in a grid of non-overlapping points that completely fill some space (Figure 12). At

the other extreme of perfect clustering, distances among all pairs of data points are minimized, approaching zero. Between these two extremes, distances among pairs of points can approach random distributions such that there is equal probability of any two points being maximally or minimally distant from one another.

115

*Figure 11: The effect of *m* on material clustering. As *m* increases from left to right, *R

*decreases. All values of *m* produce *R* ratios in the range of clustering, though small *m* *

*values generate results that are indistinguishable from random (left). *R* ratios never fall *

*significantly in the range of dispersion (*R>*1).*

*Average-nearest-neighbor-distance (ANND) analysis *can be used to evaluate clustering objectively. This is a method for assessing the degree to which spatial clustering and dispersion can be distinguished from spatially random distributions (Baxter 2003:163–169; Lee and Wong

2001:72–78). The ANND ratio, *R*, is the ratio of observed and expected ANNDs, *d̅*

*obs*

* *and *d̅*

*exp*

, respectively.

*R*

=

¯*d*

*obs*

¯*d* exp

.

The observed ANND for some set of nearest-neighbor distances for each of *n *data points is found as follows:

¯*d*

*obs*

=

*n*

# ∑

*i*

=1

*d i*

The expected ANND, which assumes that points are randomly distributed in some area, *a*, is found as follows:

116

*Figure 12: Interpretation of nearest-neighbor ratios, R, demonstrated with three synthetic data sets, each consisting of 100 points. The p *

*values show that the first and last distributions have *R* ratios that are *

*significantly different from random (*p<*0.10). The center distribution, *

*with x and y coordinates drawn at random from a uniform distribution *

*between 0 and 1, is indistinguishable from random (*p>*0.10).*

¯*d* exp

= 1

2

# √

*n*

/*a*

, where *n* is the number of points and *a* is the area of interest. Clustered distributions have *R* values that approach 0, random distributions produce values that approach 1, and dispersed distributions produce values that approach 2.149.

For this analysis, the probability that *R* is significantly different from the expected

random value is given as a p-value derived by Monte Carlo simulation 1 . The p-value reflects an

average p-value taken from 100 independent KS-tests, each comparing the observed distribution

1 It is common practice in ANND analysis to use a z-score to assess the statistical significance of departure from random. Z-scores assume a normal distribution, and this assumption is likely violated in most cases of twodimensional nearest-neighbor distances. The violation of the assumption could result in spatially random distributions erroneously identified as showing significant non-randomness. The Monte-Carlo method employed here avoids this problem by using nonparametric comparisons between the simulated and empirical data. All calculations were coded in R statistical computing language using a combination of base and custom code (The

R Foundation 2009). All code can be found in the supplementary material at the end of this document.

117 of nearest-neighbor distances to a distribution of nearest-neighbor distances derived from a random synthetic distribution of the same number of points and same area as the test case. The synthetic nearest-neighbor distances are derived from point distributions with x and y coordinate values randomly drawn from a uniform distribution with a minimum value of 0 and maximum

value equal to the square root of area, *a*. To illustrate the efficacy of the procedure, figure 12

gives calculated *R* and *p* values for the synthetic clustered, random, and dispersed spatial-data

distributions. Figure 11 gives calculated *R *and *p *values for ABM artifact distributions. All three

ABM examples generate *R *ratios in the range of clustering (*m < *1), though the case of weak niche-construction (*m = *0.05) is statistically indistinguishable from random.

Baxter (2003:164–165) reviews key methodological and interpretive pitfalls of using nearest-neighbor analysis to characterize archaeological data. Of particular relevance to this analysis are *boundary effects*. Whether by post-depositional alternation or research design, the boundaries of a study area may be arbitrary relative to the system that generated the point pattern of interest. This boundary condition can bias ANND estimates by arbitrarily excluding nearest neighbors that fall outside the study boundary. Boundary effects therefore tend to overestimate

ANNDs. This directional bias has consequences for ANND interpretation. While it is possible that a random distribution could be erroneously identified as dispersed or a clustered distribution as random or dispersed, boundary effects will never result in the incorrect identification of a dispersed or random distribution as clustered. Accordingly, boundary effects are not a concern for empirical distributions that are identified as significantly clustered. As we will see, the

118 archaeological results presented here are highly and significantly clustered and therefore do not require additional statistical treatment to mitigate boundary effects.

In sum, the working niche-construction model predicts that artifact distributions among prehistoric settlement systems ought to reveal spatial clustering. As long as the nicheconstruction behavior is sufficiently strong, and the sample size sufficiently large, the degree of clustering should be statistically distinguishable from random. Archaeological data should never generate an *R* ratio that is significantly dispersed.

*Hypothesis 2: Material-per-site counts have high dynamic range.*

*Dynamic range* refers to the ratio of the largest and smallest values in some numeric set

(Newman 2005). For example, documented human heights range from 60-260 cm, giving a dynamic range of 4.3 meaning that the tallest documented human is 4.3 times taller than the shortest. No matter how many individuals are added to this sample, the dynamic range will never change appreciably. This is a small dynamic range in comparison to the populations of incorporated settlements in the U.S., which in 2010 ranged from 4 to 9,000,000 people, giving a dynamic range of 2,250,000. Because the niche-construction model under investigation here involves recursive process such that a site's size predicts its growth, we would expect some sites to grow in runaway fashion while others remain extremely small. Thus, high dynamic range in archaeological material counts per site is expected when niche-construction behavior is strong.

The ABM is used to illustrate the expected dynamic range under various strengths of niche-construction behavior. When *m* is weak (*m = *0.05), the dynamic range is only 4 artifacts after the creation of 1,000 sites. When *m* is moderate (*m = *0.50), the dynamic range is 51 after

119 the creation of 1,000 sites. When *m *is strong (*m = *0.95), the dynamic range is 9,025 after the generation of 1,000 sites. Moreover, the dynamic range is expected to covary with sample size.

The ABM continually generates small sites, but the largest sites will continue to grow as long as the simulation continues, thereby increasing the dynamic range. We therefore expect high dynamic range in artifact-per-site counts when niche-construction behavior is moderate to strong.

If inter-site artifact quantities exhibit low dynamic range, we conclude that niche-construction behavior is either weak or absent.

*Hypothesis 3:* *Artifact-per-site distribution tends toward a power-law function.*

Though hypothesis 2 predicts the total range of variation in site-based artifact counts, it says nothing of the structure of variation between the extremes. The ABM records material counts for all sites that are created giving us the ability to generate predicted frequency distributions for material-counts per site. I use the ABM to predict the statistical structure of sitebased artifact counts under the conditions of weak, moderate, and strong niche construction.

Figure 11 serves as a qualitative reference for the spatial distribution of artifacts. To generate

quantitative predictions for each of the three cases, 1,000 virtual archaeological sites are

generated with material counts tabulated for each. Figure 13 summarizes a random sample of

site-based artifact counts for each case using three visualization techniques—a standard histogram a histogram of log-transformed values, and a cumulative density function (CDF) plot with logarithmic axes. Sample sizes of 70 sites for each case are used to facilitate comparison with the archaeological case study to be described below. The distributions all share several key features. The standard histograms are highly right skewed, the histograms of the log-transformed

120

*Figure 13: Predicted distributions for material-count variation among sites in hunter-gatherer settlement systems. Each column presents a different graphical summary, including a standard histogram (left), log-transformed histogram (center), and rank-size plot with logarithmic axes *

*(right). Each row represents a different value of niche-construction bias, *m.* All three cases *

*share the same general properties. The standard histograms are highly right skewed. The logtransformed histograms are moderately right skewed. The CDF plots are approximately loglinear. These properties are qualitatively consistent with power-law structure.*

121 values are moderately right skewed, and the CDF plots reveal log-linear structure. These qualitative features are commonly used to detect power-law structure (Bentley et al. 2004;

Newman 2005). Accordingly, if the archaeological artifact distributions result from the posited niche construction behavior, we should expect to observe the same qualitative features in artifactper-site counts.

However, such qualitative properties are insufficient to confidently conclude power-law structure. Small sample sizes, for example, can result in the erroneous identification of plausible power-law fits if not checked against more rigorous procedures. I therefore use the methods described by Clauset et al. (2009) and Edward's et. al. (2007) to determine if a power-law model is statistically plausible and parsimonious relative to alternative models of comparable simplicity

(i.e., those that have few free parameters). The method proceeds as follows: First, we decide upon a set of candidate statistical models that could conceivably describe the data. The choice of models follows from subjective assessment of the distributions, theoretical reasoning, and/or the suggestions of previous research. Here, I consider Poisson, geometric, and Zeta power-law models because they pertain to discrete data, and previous research has either explicitly or implicitly suggested that their continuous-data analogs (normal, exponential, and Pareto powerlaw distributions, respectively) could characterize variation in human settlement-size variation or in processes that have similar underlying mechanics (Bentley and Shennan 2005; Brown and

Witschey 2003; Brown et al. 2006; Clauset et al. 2009). Power-law functions are only considered for the full range of data, not only the upper tail as is the case in some studies (Clauset et al.

2009).

122

Second, we solve for best-fit parameters for each of the candidate models using maximum likelihood estimation (MLE). Third, we determine if the observed data could have plausibly come from each of the best-fit statistical models. The method of determining statistical plausibility uses KS tests and Monte Carlo simulation of synthetic datasets as described by

Clauset et al. A given model is rejected if there is a 10-percent chance or less (*p < *0.10) of deriving the ABM data from a given best-fit statistical model. All other outcomes are considered statistically plausible. Fourth, if two or more models are statistically plausible, we use Akaike information criteria (AIC) weights to decide if any are more parsimonious than the others. Those that generate high AIC weights (*w > *0.10) are considered parsimonious relative to the alternatives. Models that generate low AIC weights (*w < *0.10) are rejected. All calculations were performed in R statistical computing language (R Foundation 2009), and all code is available in the supplementary materials.

The statistical structure of each of the three ABM results was characterized using these methods. For a given case of 1,000 ABM-generated sites, the tests were bootstrapped 100 times, each time drawing a sample of 70 sites (the sample size of the archaeological case study below) to explore the range of potential outcomes due to statistical chance. The results indicate that the power-law models offer plausible and parsimonious fits to the simulated data in the majority of tests for each of the three cases. When *m=*0.05, 78 percent of the results are consistent with power-law models, zero are consistent with the alternatives, and 22 percent failed to identify a plausible model. When *m=*0.50, 70 percent of the results are consistent with power-law structure, zero percent are consistent with the alternatives, and 30 percent failed to identify a plausible

123 model. When *m=*0.95, 80 percent of the results are consistent with power-law structure, 2 percent are consistent with Poisson models, none are consistent with the geometric models, and

18 percent failed to identify a plausible model.

While each of the three ABM cases is most consistent with a single distribution class, the results generate variation that could prove insightful in interpreting archaeological distributions.

Power-law variation is captured in a scaling parameter, *alpha*, which describes the slope of the probability distribution of values (Bentley and Maschner 2008; Brown et al. 2005; Newman

2005). Larger values of alpha suggest greater weight toward small data values while larger values of alpha suggest greater weight toward larger data values. Logically, we should expect higher values of *m* in the ABM to generate lower alpha values because increasing nicheconstruction behavior should tend to generate relatively few small sites and larger large sites.

The bootstrap results on the ABM data bear this out. Mean alpha values+sd are 4.7+0.6 when

*m=*0.05, 2.4+0.2 when *m=*0.50, and 1.8+0.1 when *m=*0.95.

The posited niche-construction behavior therefore predicts that artifact-per-site distributions should be statistically consistent with power-law statistical models and inconsistent with Poisson and geometric statistical models. In other words, if we find that archaeological sitebased artifact counts are consistent with a power-law statistical model and more parsimonious than the alternatives, the current hypothesis is supported. Conversely, if we find that Poisson or geometric models are plausible and more parsimonious, the current hypothesis is rejected. If empirical results are inconsistent with all three statistical models, conclusion is indeterminate.

124

*Hypothesis 4:* Material quantities per site can be heterogeneous even in homogeneous environments.

Although it is reasonable to expect that localized resources such as springs, lithic quarries, and rock shelters predict hunter-gatherer sites of high occupation intensity, the nicheconstruction model adds that we should also expect considerable site-size variation even in the absence of patchy, localized natural-resource distributions. This follows from the premise that resource patches are effectively constructed by hunter-gatherers. Conversely, the null hypothesis predicts greater site-size uniformity in homogenous environments than in heterogeneous environments (Jochim 1981:155–159).

*Hypothesis 5:* Artifact quantity partially covaries with occupation span.

As a result of the feedback behavior posited for the working niche-construction model, foragers are expected to become locked into reusing places on the landscape. In addition, the model posits that new sites are continually created as foragers occupy novel locations to exploit natural resources independently of culturally deposited materials. As a result, the most material rich sites should always exhibit long occupation spans that may approach the full temporal range of a settlement system's existence while those of small sites should vary widely. To be clear, occupation span simply refers to the difference between the first and last occupations. It does not speak to how occupation is distributed within that span—i.e., whether or not the occupation is continuous or discontinuous.

The ABM can be used to illustrate this expectation. The simulation shown in figure 10,

with *m=*0.95, serves as the example. For this test, the model was run until 1,000 sites were

125

deposited. Figure 14 shows occupation spans for each

simulated site plotted as a function of material count. As reasoned above, the sites with fewer artifacts—those toward the left side of the plot—tend to exhibit more variation in occupation span than sites with many artifacts, which tend to exhibit longer occupation spans that approach the full duration of the simulation.

An ideal archaeological test of this relationship would involve comparison of site occupation spans and artifact counts for many sites in a given settlement system. Such robust archaeological inventories are difficult to muster. However, it may be possible to partially test the hypothesis by examining the occupation spans of the most artifact rich sites in a

*Figure 14: Hypothesized partial relationship between site occupation span and material quantity. Materialrich sites are expected to have longoccupation spans that approach the full duration of the settlement system's existence while materialpoor sites have variable occupation spans. Graphic generated from an *

*ABM trial that produced 1,000 sites *

*and used an *m*=0.95.* given settlement system. Without ecological inheritance or highly localized natural resources, there would be little reason for hunter-gatherers to return to a specific location after abandonment. Given the assumption that mobile foragers move at least once per year and often more frequently (Kelly 1995), any given site for residentially mobile foragers should only represent a single occupation lasting no more than a year. For sites that are not clearly attached to some localized resource such as a spring or rockshelter, chronometric analysis of archaeological should tend to reveal short occupation spans in which the error terms of all chronometric data

126 overlap each another. Conversely, the working model predicts that the occupation spans of many sites in the settlement system, especially the most artifact rich sites, should represent superannual occupation spans extending multiple human generations or even the full duration of a settlement systems' existence. This pattern could be expected to occur in both exceptional and unexceptional natural-resource locations.

*Hypothesis 6:* Site-area variation follows a lognormal distribution.

The working niche-construction model holds implications for site-area variation in hunter-gatherer settlement systems as well. Again, the recursive nature of the niche-construction model makes it difficult to use simple logic to predict the expected size distribution. ABM is therefore used to deduce expected structure. Site-area distributions are simulated for strong, moderate, and weak values of *m*, and using normal and exponential artifact-displacement

functions. All results reveal generally similar structural properties. Figure 15 summarizes the

simulation results with histograms. The qualitative structure is consistent with logormal and exponential distributions—the standard histogram is highly right skewed, the histogram of logtransformed values is approximately normal with slight left skew, and the CDF plot with logarithmic axes is upwardly convex.

The more-rigorous quantitative methods reveal that lognormal distributions are favored regardless of *m *or the artifact-displacement model used. The normal distribution of artifact displacement produces a lognormal sitea-area distributions in approximately 50 percent of the simulated site samples. When exponential functions are used to model artifact displacement, lognormal distributions are favored in approximately 70 percent of the cases. In all other

127

*Figure 15: ABM-generated distributions for site-area variation in hunter-gatherer settlement systems using an exponential artifact displacement function. Each column presents a different graphical summary, including a standard histogram (left), log-transformed histogram (center), and cumulative density function (CDF) plot on logarithmic axes (right). Each row represents *

*different degrees of niche-construction bias, *m. *The standard histograms are highly right *

*skewed. The log-transformed histograms are approximately normal with a slight left skew. The *

*CDF plots are upwardly convex. These qualitative features are therefore predicted features of the archaeological site-area distributions.*

128 simulated site-area distributions, none of the alternative models are found to be plausible.

Accordingly, given the working niche-construction model, we should expect archaeological sitesize distributions in hunter-gatherer settlement systems to be consistent with lognormal distributions. If the archaeological distribution is consistent with a normal, exponential, or power-law distribution, then the hypothesis is rejected. If the archaeological distribution is inconsistent with all of the considered models, than conclusion is indeterminate.

CASE STUDY: A LATE ARCHAIC PERIOD HUNTER-GATHERER SETTLEMENT

SYSTEM IN THE LAKE TITICACA BASIN, PERU

Although any hunter-gatherer settlement system could be used to test the hypotheses presented above, an ideal case study should occur in a relatively homogeneous resource environment and should offer reasonable temporal control at a regional scale. The Lake Titicaca

Basin Late Archaic Period settlement system generally meets these criteria and is used as the test case here. The Late Archaic Period (ca. 7,000-5,000 cal. B.P.) precedes the Terminal Archaic

Period (5,000 -3,5000 cal. B.P), a period of low-level food production (Bruno 2006; Craig et al.

2010). The Late Archaic Period subsistence economy is characterized by mobile hunting and gathering with emphasis on vicuña (a wild camelid), taruca (Andean deer), tubers, and small seeds (Aldenderfer 1989; Aldenderfer 1998).

The Late Archaic material repertoire appears to have been fairly limited. Flaked stone technology is the most archaeologically visible material. The flaked stone technology emphasizes bifacial projectile points and scrapers. Late Archaic projectile points are distinctive in their size and form (Klink and Aldenderfer 2005). They tend to be larger than those of any

129 other period in the region. In the western Titicaca Basin, there are two general forms—a stemmed form (type 4D) and a concave-base form (type 3F)—that were presumably used with spear and/or atlatl delivery systems. Informal groundstone was used in low quantities (Craig

2011; Rumold 2010), and permanent architecture is unknown from surface surveys and limited excavations (Cipolla 2005; Craig 2011; Klink 2005). Ceramic technology is unknown in Late

Archaic material assemblages. Though this relatively limited material repertoire likely reflects limited investigation and issues of preservation to some extent, it is clear from comparison with later periods (c.f., Aldenderfer et al. 2008; Hastorf 2008) that it also reflects a fairly limited technological tradition geared toward a mobile settlement system.

The specific archaeological case used here draws on settlement-survey and excavation data collected in the lower Ilave River drainage centered at S16º 13', W69º 45', 3,860 masl (WGS

84), west of Lake Titicaca (Figure 16). In 1994 and 1995, Mark Aldenderfer conducted a

systematic pedestrian survey covering 41 km 2 centered on the Ilave Basin (Aldenderfer et al.

1996; Craig 2011). The survey was specifically focused on locating Archaic sites, and therefore the research design was limited to geologic settings conducive to Archaic site preservation including relict alluvial terraces that have been relatively stable since Archaic occupation

(Rigsby et al. 2003). Within these geologic settings, systematic pedestrian transects at 15-m spacing were used to locate archaeological materials. The positions of all encountered sites and isolated artifacts were recorded by GPS, and temporally diagnostic materials were systematically collected. Importantly, and unlike in other parts of the world, flaked stone artifact collecting is not a regional pastime in the Lake Titicaca Basin. Accordingly, temporally diagnostic artifact

130

*Figure 16: Late Archaic projectile point locations used in average-nearest-neighbordistance analysis for the Ilave River study area .The inset shows a closeup of one area to give a sense of the spatial patterning.*

counts are unlikely to be affected by looting. Site areas were estimated by pin-flagging the boundaries of sites and estimating their north/south and east/west dimensions by compass and pace. The survey resulted in the collection of 247 Late Archaic projectile points from 58 sites.

All sites are located near the margins of relict alluvial terraces. All share easy access to

perennial rivers within 2 km or less and are situated in *altiplano *habitat (Figure 17). The

altiplano environment is characterized by expansive rolling hills grasslands dissected by perennial rivers and flanked by mountains (Winterhalder and Thomas 1978). Due to high elevation, ca. 3,800 masl, the vegetative landscape is relatively homogenous. *Bofedales *are

131

*Figure 17: The Ilave Basin study area in the relatively homogeneous altiplano environment.*

perhaps the most economically important source of environmental variation in the altiplano.

Bofedales are* *localized wetlands with relatively high forage productivity that is important for wild camelid subsistence. Although Aldenderfer (1989) shows that bofedales were economically important to Archaic foragers in other parts of the south-central Andes, they are rare in the study area and are therefore unlikely to have significantly influenced land use in the sub-region under investigation.

The site of Pirco (Ilave 95-169), covering 1.4 ha, is the largest Late Archaic Period site discovered during Aldenderfer's survey (Craig 2011). Systematic surface collection produced 105 projectile points, of which 58 could be assigned to the Late Archaic Period. Nathan Craig conducted limited excavations at Pirco, which yielded organic-stained soils, abundant lithics, minor amounts of groundstone, and a secondary human burial. Ceramics and archaeologically

132 detectable residential structures were absent. Unfortunately, the excavations did not produce any datable carbon. From the sum of the evidence, Craig concludes that Pirco was a residential site used repeatedly by mobile foragers.

The previous field investigations by Aldenderfer and Craig serve as the foundation for testing the hypotheses derived from the working model. However, model testing requires additional data. For hypothesis 1 (spatial clustering of artifacts), point-provenience data are required on temporally diagnostic artifacts. For isolated occurrences and sites with single Late

Archaic point occurrences, Aldenderfer's site proveniences can serve as artifact point proveniences for the purpose of this study. However, point-provenience data are lacking for sites with two or more Late Archaic points because the research design at the time did not call for plotting all artifact locations within sites. Accordingly, I revisited all sites that were recorded as having two or more Archaic projectile points to systematically collect and point-provenience projectile points. Although the previous surface collection necessarily decreased the artifact densities at those sites, 18 years had elapsed between collections, and it seemed reasonable to expect that a sufficient number of points would have surfaced in the meantime and thus minimized the source of sampling bias. Previous research on repeated surface collections shows this to be a reasonable assumption especially during the first few collection episodes (Stiger

2001). Indeed, the collection effort generated 117 Late Archaic Period projectile points from 25 sites. All point-provenience information was collected using single-frequency gps with a 6-m error term.

133

Testing hypothesis 2 requires estimates of the upper and lower bounds of material counts among Late Archaic sites. Ideally, such artifact count estimates should come from well-dated sites. In addition, testing hypothesis 5 requires knowledge of site-occupation spans.

Unfortunately, none of the Late Archaic Period sites are well dated. I therefore conducted limited excavations at the site of Soro Mik'aya Patjxa (SMP; Ilave 95-259). After Pirco, SMP is the largest Late Archaic site in the sample covering 0.3 ha and producing 25 Late Archaic projectile points in Aldendenderfer's survey and an additional 13 in my surface collections. The excavation effort covered 50 m 2 of surface area and 21 m 3 of sediment. Abundant carbon was recovered from secure feature contexts including human burials and shallow pits. Seventeen carbon samples were submitted to the University of Arizona Accelerator Mass Spectrometry lab. The chronometric results are presented below.

*Hypothesis 1 Test Results*

Hypothesis 1 predicted that culturally deposited materials should tend to cluster in geographic space. ANND analysis is used to test the hypothesis. To control for time, I restrict the analysis to Late Archaic Period projectile points. These include systematically collected isolated projectile point finds recorded by Aldenderfer and those that I systematically located and point provenienced in previously recorded sites. Together, the sample represents a systematic surface collection of the entire study area. Because my collections were made on previously collected sites, depressed artifact counts risk artificially inflating nearest neighbor distances, thus making the distribution appear more dispersed than it actually is. This only becomes a concern if results

134 are dispersed or random; it does not affect clustered results. The total sample includes the

locations of 134 Late Archaic projectile points (see Figure 16).

The resultant ANND for the archaeological sample is 97m. The expected ANND is 276 m. The resultant *R* ratio is 0.35 with a p-value of 0.00 indicating strong and significant tendency toward clustering. Although Aldenderfer's survey strategy attempted to exclude geologic contexts not conducive to Archaic site preservation, we might be concerned that the survey tracts that failed to yield any Late Archaic artifacts actually have problematic geologic contexts, thus artificially depressing the *R* ratio. Removal of these survey tracts from the analysis reduces the

41-km 2 are to 21 km 2 and only slightly affects the *R *ratio, but does not affect the pattern of extreme and significant clustering (dexp=198 m, dobs=97m, *R=*0.49, *p=*0.00). The hypothesis of spatial clustering of hunter-gatherer artifacts is therefore supported.

*Hypothesis 2 Test Results*

Hypothesis 2 predicted that material counts among sites should exhibit high dynamic range due to the feedback behavior that emerges from the model. To test the prediction, we require estimates of the upper and lower ends of material counts per site in hunter-gatherer settlement systems. Toward estimating the low end, isolated artifact occurrences including isolated Late Archaic projectile points are known in the study area. If we are liberal in what we consider a site, these could constitute the lower end of the range. If we are more demanding in what minimally constitutes a hunter-gatherer site, sites with tens to hundreds of artifacts with multiple artifact classes (e.g., flaked stone, groundstone, ocher) are present and could be considered the lower estimate.

135

Estimating the upper end of the dynamic range was accomplished by estimating total artifact counts at the partially excavated site of SMP, the second largest Late Archaic Period site documented in the study area. Over 80,000 artifacts including flaked stone, groundstone, pigment stone, carbon, and bone were recovered from 50 m 2 of excavation. When projected to the rest of the 2,800-m

2

site, the total artifact estimate is 3.1 million + 60 thousand

estimates are admittedly rough, they are sufficient to generate a minimum estimate of the dynamic range in artifact counts for Late Archaic sites in the Ilave study area. Using the most conservative estimates—a minimum site size of 1,000 artifacts and a maximum of 3 million—the dynamic range is 3,000 artifacts. This extremely high minimum estimate for dynamic range is consistent with the expectations of the working niche-construction model.

*Hypothesis 3 Test Results*

Hypothesis 3 predicted that the distribution of artifact-per-site counts should be consistent with a power-law function. The test for this prediction uses projectile points collected in systematic, 100-percent surface collections made in the Ilave study area. Artifact counts are

limited to Late Archaic projectile points in order to control for time. Figure 18 summarizes the

data using histograms. Qualitatively, the histograms show all of the characteristics of power-law structure including extreme right skew in the standard histogram, moderate right skew in the

2 To estimate the total artifact count for the site, we first examined the density distribution of artifacts in each excavation unit, which cover no more than 1 sq. m each. Akaike information criteria suggests that an exponential distribution with a rate of 8.93x10

-4

offers the best fit (99.9%) to the data over normal, lognormal, and power law alternatives. KS test comparison of the maximum-likelihood estimated (MLE) model parameters with the raw data suggests marginal difference (D = 0.1208, p-value = 0.10), which we consider sufficient for the purposes of estimation. The total artifact count estimates were derived using a random exponential function with 2,800 iterations (2,800 m 2 site) and the MLE estimated rate parameter for the exponential function. This was iterated 10,000 times to generate an error term. The result is a normal distribution of estimates with a mean of 3.1 million and standard deviation of 59 thousand artifacts.

136

*Figure 18: Distribution of Late Archaic projectile points per site in the Ilave study area. The data reveal structural properties that are consistent with those observed in the ABM simulated *

*data (compare with Figure 13).*

histogram of log-transformed values, and log-linearity in the CDF plot (compare with Figure 13).

More rigorous quantitative analysis using MLE, KS tests on Monte Carlo simulations, and AIC weights reveals that the distribution is most consistent with a power-law function. Only a best-fit power-law model generated a statistically plausible result (*D=*0.08, *p=*0.11). Poisson and geometric distributions were determined to be statistically implausible (*D=*0.52, *p=*0.00 and

*D=*0.25, *p=*0.00, respectively). It is noteworthy that the best-fit power-law scaling parameter,

*alpha*, as determined by maximum likelihood estimation is 1.6, which is comparable to the alpha value produced for the ABM-simulated strong case of niche construction (*alpha=*1.8+0.1). This would seem to indicate a high propensity for site reuse in the empirical case.

*Hypothesis 4 Test Results*

Hypothesis 4 predicted extreme site-size variation in the absence of patchy, localized natural resources. The null hypothesis is that settlement occupation intensity in homogeneous

137 environments would similarly be homogeneous. As described above, the high-elevation environmental structure of the altiplano case study is relatively homogeneous. No particular locations within the study area stand out in terms of having unusual natural resources. Water is abundant throughout the region and perennial rivers are accessible within 2 km of all points in the study area. Localized lithic quarries were not observed in the study area, though river-gravel sources of low-quality materials are readily accessible from all locations in the study area.

Although some rockshelters are present in the Uncallane River portion (see Figure 16), Archaic

Period occupation was not observed in any of them (Aldenderfer et al. 1996). Bofedales are largely absent from the study area. If the altiplano structure determined variation in site occupation intensity, then we would expect site-occupation intensities and corresponding artifact quantities to reveal a similar degree of homogeneity.

Yet, as discussed in the previous two hypotheses, the range of variation in occupation intensities in the homogeneous environment is extreme. The dynamic range of total artifact counts per site is on the order of thousands to millions, and the artifact-count distribution is consistent with a power law function such that no sites appear to deviate appreciably from the general trend, suggesting that this range of variation is not biased by anomalously small or large sites. Rather, all site-size variation fits comfortably into a continuous but highly skewed probability function with high dynamic range. It therefore appears that the altiplano environmental structure cannot readily account for the observed archaeological variation.

A potential flaw in the test is the assumption that modern environmental context serves as a reasonable proxy for the prehistoric environmental structure. Perhaps environmental patches

138 were present at the time of occupation but are no longer observable today. While the paleoenvironment of the Late Archaic Period certainly did exhibit significant differences from the modern environment (Baker et al. 2001; Rigsby et al. 2003), it seems unlikely that the differences would have significantly altered the types and structure of natural-resources available to Late Archaic foragers in ways that would affect the patterns of interest here.

*Hypothesis 5 Test Results*

Hypothesis 5 predicted that the artifact-rich sites located in unexceptional environmental locations should exhibit multi-annual occupation spans and potentially multigenerational occupation spans that could encompass the entirety of the settlement system's existence. The current cultural chronology of the Lake Titicaca Basin suggests that the Late Archaic Period persisted for approximately 2,000 years (Klink and Aldenderfer 2005), which gives a sense of an upper end of what we could expect in this particular case.

The site of SMP was selected for excavation to test this hypothesis. SMP is located on the edge of an alluvial terrace near the center of the vast Ilave Basin pampa, or grassland. The site is among the most intensively occupied Late Archaic sites in study area. Excavations produced an abundance of carbon from secure pit-feature contexts that are used to test the current hypothesis.

Table 4 presents the results of seventeen

14 C assays. Calibrated median dates range from 7,900-

6,700 cal. B.P., suggesting a minimum occupation span of 1,200 years. Considering the apparently mundane nature of the environmental context coupled with the high residential mobility of the site's hunter-gatherer occupants, the occupation span of more than a millenium is surprising. However, it is consistent with the expectations of the working niche-construction

139

*Table 4: Radiocarbon age determinations for the site of Soro Mik'aya Patjxa (n=17).*

lab ID material

AA102848 wood

1

AA102854 wood

AA102843 wood provenience area unit level feature

1 61 2 10

7

1

19

55

1

1

14

16 radiocarbon B.P.

age

5891 error

49

5914

5924

35

48

95% calendar B.P.

2 max median min

6850 6713 6566

6843

6883

6733 6660

6748 6656

AA102828 wood

AA102851 bark?

AA102859 wood

AA102858 wood

AA102834 parenchyma

AA102855 parenchyma

AA102837 parenchyma

AA102829 parenchyma

AA102835 twig

AA102842 twig

AA102827 grass stem

1

7

7

7

1

7

1

1

1

1

1

25

11

26

23

33

19

48

25

33

52

22

1

1

1

1

2

1

1

1

2

1

1

3

9

18

15

6

14

13

3

6

13

2

5940

5957

5983

5996

6002

6003

6089

6103

6148

6157

6401

49

48

47

51

48

50

49

48

50

49

50

6891

6904

6945

6966

6972

AA102831 grass stem

AA102826 parenchyma

1

1

27

22

1

1

5

2

6458

6631

71

50

7499

7580

AA102838 twig 1 48 1 13 7090 59

1

2

All dates generated using AMS on charred materials from lower levels of feature contexts

OxCal v4.2.3 (Bronk Ramsey 2013); r:5 IntCal13 atmospheric curve (Reimer 2013).

8013

6981

7157

7159

7170

7175

7425

6768 6664

6788 6670

6822 6679

6836 6693

6842 6729

6843 6725

6959 6800

6980 6808

7053 6896

7063 6903

7338 7255

7369 7255

7517 7435

7914 7794 model, which suggests that the exceptional quality of the environment derives from the human occupation itself.

*Hypothesis 6 Test Results*

Hypothesis 6 predicted that site-area variation should be most consistent with a lognormal statistical model over normal, exponential, and power-law alternatives. Fifty-one Late

Archaic Period sites were analyzed for statistical structure in site-area variation. To avoid overestimation of site areas due to inflation from non-Late Archaic occupations, the analysis only includes sites for which 50-percent or more of the temporally diagnostic projectile points could

140

be assigned to the Late Archaic Period. Figure 19 summarizes the results using three types of

histograms. Qualitatively, the results appear to be consistent with the simulation results, evidencing an extremely right skewed histogram, a normal but slightly left skewed histogram of

log-transformed values, and an upwardly convex CDF plot (compare with Figure 15). However,

the more rigorous tests indicate that the expectation of lognormality is unmet. Of the candidate statistical models, only a best-fit exponential model generated a statistically plausible fit to the data. We therefore reject the prediction of lognormal structure in site-area variation.

It is currently unclear what factors drive this deviation from expectation, but the deviation does indicate either a model shortcoming or data flaw. That an exponential distribution offers a better fit to the empirical data than does the predicted lognormal structure suggests that the bias is systematic with larger sites disproportionately affected. This is because lognormal distributions tend to have heavier statistical tails than exponential distributions. In other words, exponential distributions tend to be more attenuated in their variation. This is somewhat surprising because it

*Figure 19: Late Archaic site-area variation in the Ilave study area. The structural properties of *

*the distributions are qualitatively similar to those predicted by the ABM (see Figure 13). *

*However, more objective tests indicate the the distributions are inconsistent with the predicted distributions.*

141 suggests that the archaeological variation is more constrained than the model variation. One possible model flaw is the use of normal and exponential distributions for modeling artifactdisplacement distances in the ABM. Other artifact-displacement models may be more realistic, but lacking theoretical guidance on how artifacts should be dispersed, it is difficult to implement alternatives. Another distinct possibility is systematic bias in how site-areas are recorded.

Establishing site boundaries is a necessarily subjective task, and it is possible that the boundaries of larger sites are systematically under-estimated due to relatively low thresholds in artifact density for defining the boundary's cutoff point. Alternatively, it may be that the temporal-control method used to exclude extraneous occupation from site-area estimation systematically biases the sample. Theoretical and empirical analyses of artifact dispersion and methodological bias or consideration of alternative models may ultimately be required to answer these questions. For now, I simply acknowledge the shortfall and leave the problem to future research.

SUMMARY AND CONCLUSION

Humans fundamentally expand the spatiotemporal distributions of cultural materials on the landscapes they occupy. In doing so, they potentially reduce future costs of material acquisition and the production of implements and structures. Inheritors of the landscape therefore benefit from previous uses. Because such material subsidies potentially free up time and energy for reproductive pursuits, evolutionary theory predicts that the culturally deposited materials should be incorporated into the mobility strategies of tool-using foragers, thus biasing occupations to previously occupied locations.

In turn, when foragers reoccupy locations to take advantage of previously deposited materials, they probabilistically deposit additional materials, which attracts future use of the location, resulting in more deposition, and so on. Subtle variations in initial conditions—i.e., point-of-entry and chance resource encounters—that are often beyond empirical detection can

142 cause certain locations to be occupied or passed up. But once a particular location is occupied for the first time, forager use may lock in to resuing it over the long term. This positive-feedback niche construction behavior predicts distinctive structural properties in archaeological settlement patterns.

ABM was used to derive logically sound predictions from the model of recursive nicheconstruction behavior. Drawing insight from the elegant mechanics of the neutral model of cultural transmission described by Bentley and colleagues (Bentley and Shennan 2005; Bentley et al. 2004; Bentley et al. 2007; Hahn and Bentley 2003), the virtual forager simply decides at each time step to use the location of some previously deposited material with probability *m* or a novel location on the landscape with probability 1-*m*. This simple decision process effectively models the posited niche-construction behavior by biasing land use to the locations of previously deposited cultural materials.

The ABM and simple logical arguments were used to make six structural predictions for archaeological settlement patterns. The Late Archaic Period hunter-gatherer settlement system of the Lake Titicaca Basin served as the test case. The analysis found empirical support for five of the six hypotheses and a lack of support for the remaining hypothesis. (1) Spatial clustering of artifacts, (2) high-dynamic range in site-based artifact counts, (3) power-law structure in site-

143 based artifact counts, (4) heterogeneous site-size variation in homogenous natural environments, and (5) long occupation spans in environmentally unexceptional locations were deduced from the working niche-construction model. Empirical data support these expectations. Only the prediction of lognormally distributed site-area variation found a lack of empirical support.

Nonetheless, given the preponderance of support for the model predictions, I conclude that working niche-construction model offers a plausible and parsimonious understanding of mobility among tool-using foragers.

DISCUSSION

Although the niche-construction model of forager mobility presented here accounts for several dimensions of hunter-gatherer settlement structure, its implications require additional testing against archaeological and ethnographic data. I have argued that the model predictions are consistent with at least five structural features of Late Archaic Period settlement systems in the

Lake Titicaca Basin. The tests presented here are necessarily limited by time, space, the data at hand, and my own areas of specialization. Yet, I hope the predictions will be tested against additional hunter-gatherer cases and that new predictions will emerge with consideration of other methods and material classes. The model's efficacy will ultimately depend, in part, on the results of such tests.

As a first approximation, we can look to published literature on other archaeological hunter-gatherer settlement systems. Regarding the hypothesis of high dynamic range in artifact counts, previous excavations at other open-air hunter-gatherer sites have recovered samples of thousands of artifacts, which likely reflect total material counts in the hundreds of thousand or

144 millions. Surface and subsurface samples recovered from the Tenderfoot hunter-gatherer site in the Gunnison Basin, Colorado, 8,600-1,400 cal. B.P., generated over 67-thousand artifacts

(Stiger 2001). Excavations at the Barger Gulch Locality B Folsom site excavations in Middle

Park, Colorado, ca. 12,400 cal. B.P., have yielded over 65-thousand artifacts (Surovell et al.

2005). Undoubtedly, archaeologists could enumerate many more open-air hunter-gatherer sites with artifact quantities in the tens of thousands to millions of artifacts. Such large artifact quantities in individual open-air sites are consistent with the prediction of high dynamic range in artifact counts.

Regarding the hypothesis of power-law scaling in artifact quantities among huntergatherer sites, a study of eight hunter-gatherer settlement systems shows that power-law scaling may also characterize diverse settlement systems from highlands Peru, coastal Peru, and the interior desert of the U.S. Southwest (see Appendix I). The cases span a wide range of environmental contexts covering approximately 7,000 years, 3,800 m in elevation, and over 50 degrees of latitude. Nonetheless, all reveal havy-tailed statistical structure that is consistent with power-law structure and inconsistent with alternative statistical models.

Regarding the hypothesis of long occupation spans, multi-century and multi-millenial occupation spans of hunter-gatherer sites are well known archaeologically, including rockshelter and open-air sites alike (Mitchell et al. 2011; Stiger 2001). I focus on open-air sites here, because their temporal persistence cannot be readily explained by the intrinsically localized nature of rockshelters. Several cases in the Andes region of Peru support the hypothesis of long-term reuse of open-air hunter-gatherer sites. Hunter-gatherers occupied the coastal site of La Paloma for

145 over three millenia from 7,800-4,700 cal. B.P. (Benfer 1999). The highlands site of Asana was occupied over more than seven millenia from 10,500-3,500 cal. B.P. (Aldenderfer 1998). North

American hunter-gatherers also appear to have reoccupied settlements over long time spans.

Multiple occupations “are the rule” for hunter-gatherer sites in the Gunnison Basin of Colorado

(Stiger 2001:157). For example, the Tenderfoot site was reoccupied many times with feature dates spanning more than seven millennia from 8,600-1,400 cal. B.P. In the U.S. Southeast, the

Poverty Point site was used for at least three millennia from approximately 6,000-3,000 cal. B.P.

(Ortmann 2010). The nearby Watson Brake site was used for at least 500 years from 5,500-5,000 cal. B.P. (Saunders et al. 1997; Saunders et al. 2005). In southern Africa, evidence suggests that the hunter-gatherer site of Likoaeng was in use over two millenia from 3,300-1,200 cal. B.P.

(Mitchell et al. 2011). Current archaeological evidence therefore seems to suggest that the largest sites in hunter-gatherer settlement systems tend to represent long occupation spans that approach the full temporal span of the settlement systems to which they belong. It is suggested that this incredible persistence of hunter-gatherer sites can be understood as the result of nicheconstruction behavior.

It is worth reiterating that while the model's utility depends in part on its predictive power, it will also depend on how the data compare to predictions derived from alternative models. The current model should be considered a first approximation. It is easy to imagine alternative models to account for the observed archaeological patterns. Taphonomic processes, observer bias, and systemic behaviors are three types of processes that can affect archaeological variation (Grøn 2012; Schiffer 1995). The working model only considers a one-parameter

146 systemic behavior to the exclusion of many other potential processes. I make no claim that the hypotheses tested here are beyond the persistent archaeological problem of equifinality.

However, I am unaware of formal alternatives that predict the structural properties of huntergatherer settlement patterns explored here. In order to advance our understanding of the huntergatherer archaeological record, and ultimately hunter-gatherer behavior, alternative models need to be formalized and put to work in predicting these and other properties of hunter-gatherer behavior and the archaeological record. I suggest that the niche-construction model presented here serves as a parsimonious starting point that accounts for a suite of archaeologically testable hypotheses with few behavioral parameters.

To the extent that the working niche-construction model offers a reasonable account of forager-mobility, several key insights into forager behavior stand to be gained. First and foremost, it would appear that positive-feedback niche construction was not trivial to huntergatherer mobility and land use. Previous hunter-gatherer research has tended to give primacy to external environmental structure in explaining mobility decisions. The results presented here do not undermine the undeniably important role that the natural environment plays in structuring forager mobility, especially at large, inter-patch scales where forager behavior and the associated material record is well accounted for by environmental structure. However, niche-construction behavior may better account for hunter-gatherer settlement-size distributions within a range of environmental contexts.

A second key insight gained from the current study relates the emergence of socioeconomic complexity in human societies. This insight pertains to two areas of research in

147 emergent complexity—domestication and hierarchy. With regard to the former, domestication is known to result from long-term interactions with genetically plastic species (Smith 1995; Smith

2001; Rindos 1983). In the case of agriculture, the species are immobile, and thus it can be presumed that intensive interaction between species also involved intensive interaction with places. The current model may suggest a mechanism by which such long-term interactions with highly localized places on the landscape commence. If infrastructure and materials consistently pull mobile foragers to specific locations on the landscape, long-term interactions with certain plant populations may be catalyzed.

I now turn to the implications for emergent hierarchy. Geographers and anthropologists have long observed that settlement-size variation among sedentary societies is heavy tailed with extremely small settlements extremely common and extremely large settlements extremely rare

(Batty 2008; Decker et al. 2007; Drennan and Peterson 2004; Gabaix 1999; Johnson 1980;

Newman 2005; Zipf 1949). Most scholars seem to agree that settlement-size variation among modern settlements is consistent with power-law and/or lognormal models. Archaeologists have also found that the settlement systems of sedentary agricultural societies also exhibit heavy-tailed statistical structure. Following the lead of central-place economic theory (Christaller 1966), explanation of this statistical structure has tended to rely on complicated economic processes involving various combinations of agricultural production, specialized manufacturing, peerpolity competition, and/or warfare (Brown and Witschey 2003; Flannery 1998; Griffin and

Stanish 2007; Inomata and Aoyama 1996; Johnson 1980; Krugman 1996; Paynter 1982; Griffin

2011). This has led some scholars to suggest that the heavy-tailed statistical structure of

148 settlement-size variation offers a line of evidence for hierarchical settlement structure (Ames

2008; Flannery 1998; Stanish 2003). The findings presented here suggest that the statistical signature of settlement-size hierarchy can actually emerge among hunter-gatherer settlement systems whose populations were unequivocally mobile, not dependent on domesticated resources, without economic specialists, and without organized conflict. The niche-construction model presented here suggests that forager mobility biased by culturally deposited materials is sufficient to generate the pattern.

Importantly, the settlements that comprise hunter-gatherer settlement systems were likely occupied asynchronously in most cases. This is fundamentally different from sedentary societies in which settlements comprising the full range of size variation were/are generally occupied contemporaneously. Nonetheless, the same kinds of processes that led to settlement-size variation found among hunter-gatherer settlement systems may have laid the groundwork for the emergence of synchronous settlement-size hierarchies as populations became less mobile due to population growth, decreased resource productivity (i.e., circumscription), and/or increasing commitment to food production. The difference for sedentary populations is that preferential attachment would lead to a heavy-tailed distribution of contemporaneously-occupied settlements.

The niche-construction model of forager mobility presented here is simple yet generates novel predictions for the structure of hunter-gatherer material distributions. The archaeological test cases offer support for the predictions, and it is hoped that future archaeological and ethnographic research will explore the boundaries of the model and compare the results with those derived from alternative models. For now, the analysis suggests that niche-construction

149 may have played a significant role in structuring the mobility of foragers and in structuring the distribution of their material residues. Such micro- and macro-level behaviors hold potential relevance for models of emergent complexity in human societies.

REFERENCES CITED

Aldenderfer, Mark S.

1989 Archaic Period in the South-Central Andes. *Journal of World Prehistory* 3(2): 117–158.

1998 *Montane Foragers: Asana and the South-Central Andean Archaic*. University of Iowa

Press, Iowa City.

Aldenderfer, Mark S., Nathan M. Craig, Robert J. Speakman, and Rachel Popelka-Filcoff

2008 Four-Thousand-Year-Old Gold Artifacts from the Lake Titicaca Basin, Southern Peru.

*Proceedings of the National Academy of Sciences* 105(13): 5002–5005.

Aldenderfer, Mark S., Edmundo De la Vega M., and Cynthia J. Klink

1996 *Pedestrian and Buried Sites Reconniassance for Early-to-Late Preceramic *

*Archaeological Sites in the Rio Ilave Drainage, Southern Peru: A Report Submitted to the *

*National Geographic Society on the 1995 Field Season of Proyecto Ch’amak Pacha. *

Unpublished National Geographic Society Project Number 5245-94. University of

California, Santa Barbara.

Ames, Kenneth M

2008 The Archaeology of Rank. In *Handbook of Archaeological Theories*, edited by R.

Alexander Bentley and Hebert D.G. Maschner, pp. 487–513. AltaMira Press.

Baker, Paul A., Catherine A. Rigsby, Geoffrey O. Seltzer, Sherilyn C. Fritz, Tim K. Lowenstein,

Niklas P. Bacher, and Carlos Veliz

2001 Tropical Climate Changes at Millennial and Orbital Timescales on the Bolivian

Altiplano. *Nature* 409(6821): 698–701.

Bankes, Steven C.

2002 Agent-based Modeling: A Revolution? *Proceedings of the National Academy of *

*Sciences* 99: 7199–7200.

Batty, Michael

2008 The Size, Scale, and Shape of Cities. *Science* 319(5864): 769–771.

Baxter, Michael

2003 *Statistics in Archaeology*. Hodder Arnold, London.

Beekman, Christopher, and William W. Baden (editors).

2005 *Nonlinear Models for Archaeology and Anthropology: Continuing the Revolution*.

Ashgate, Aldershot.

Benfer, Robert A.

1999 Proyecto de Excavaciones en Paloma, Valle de Chilca, Peru. *Boletin de Arqueologia* 3:

213–237.

150

Bentley, R. Alexander, Matthew W. Hahn, and Stephen J. Shennan

2004 Random Drift and Culture Change. *Proceedings: Biological Sciences* 271(1547): 1443–

1450.

Bentley, R. Alexander, Carl P. Lipo, Harold A. Herzog, and Matthew W. Hahn

2007 Regular Rates of Popular Culture Change Reflect Random Copying. *Evolution and *

*Human Behavior* 28(3): 151–158.

Bentley, R. Alexander, and Hebert D.G. Maschner (editors).

2003 *Complex Systems and Archaeology: Empirical Theoretical Applications*. Foundations of

Archaeological Inquiry. University of Utah Press, Salt Lake City.

Bentley, R. Alexander, and Herbert D.G. Maschner

2008 Complexity Theory. In *Handbook of Archaeological Theories*, edited by R. Alexander

Bentley and Hebert D.G. Maschner, pp. 245–270. AltaMira Press.

Bentley, R. Alexander, and Stephen J. Shennan

2005 Random Copying and Cultural Evolution. *Science* 309(5736): 877–879.

Bettinger, Robert, Peter Richerson, and Robert Boyd

2009 Constraints on the Development of Agriculture. *Current Anthropology* 50(5): 627–631.

Binford, Lewis R.

1982 The Archaeology of Place. *Journal of Anthropological Archaeology* 1(1): 5–31.

2001 *Constructing Frames of Reference: An Analytical Method for Archaeological Theory *

*Building using Hunter-Gatherer and Environmental Data Sets*. University of California

Press, Berkeley.

Bronk Ramsey, Christopher

2013 *OxCal*. Research Lab for Archaeology, Oxford Radiocarbon Accelerator Unit, Oxford.

Brown, Clifford T., Larry S. Liebovitch, and Rachel Glendon

2006 Lévy Flights in Dobe Ju/’hoansi Foraging Patterns. *Human Ecology* 35(1): 129–138.

Brown, Clifford T., and Walter R. T. Witschey

2003 The Fractal Geometry of Ancient Maya Settlement. *Journal of Archaeological Science*

30(12): 1619–1632.

Brown, Clifford T., Walter R. T. Witschey, and Larry S. Liebovitch

2005 The Broken Past: Fractals in Archaeology. *Journal of Archaeological Method and *

*Theory* 12(1): 37–78.

Bruno, Maria C.

2006 A Morphological Approach to Documenting the Domestication of Chenopodium in the

Andes. In *Documenting domestication new genetic and archaeological paradigms*, edited by Melinda A Zeder, Daniel G Bradley, Eve Emshwiller, and Bruce D Smith, pp. 32–45.

University of California Press, Berkeley, Calif.

Christaller, Walter

1966 *Central Places in Southern Germany*. Trans. Carlisle W. Baskin. Prentice-Hall,

Englewood Cliffs N.J.

Cipolla, Lisa M.

2005 Preceramic Period Settlement Patterns in the Huancané-Putina River Valley, Northern

Titicaca Basin, Peru. In *Advances in Titicaca Basin Archaeology-1*, edited by Charles

151

Stanish, Amanda B. Cohen, and Mark S. Aldenderfer, pp. 55–64. Cotsen Institute of

Archaeology at UCLA, Los Angeles.

Clauset, Aaron, Cosma Rohilla Shalizi, and M. E. J Newman

2009 Power-Law Distributions in Empirical Data. *SIAM Review* 51: 1–43.

Craig, Nathan

2011 Cultural Dynamics, Climate, and Landscape in the South-Central Andes During the

Mid-Late Holocene: A Consideration of Two Socio-Natural Perspectives. *Chungará* 43:

367–391.

Craig, Nathan M., Mark S. Aldenderfer, Paul Baker, and Catherine A. Rigsby

2010 Terminal Archaic Settlement Pattern and Land Cover Change in the Rio Ilave,

Southwestern Lake Titicaca Basin, Peru. In *The Archaeology of Anthropogenic *

*Environments*, edited by Rebecca M. Dean. Center for Archaeological Investigations,

Occasional Paper No. 37. Southern Illinois University Center for Archaeological

Investigations, Carbondale.

Decker, Ethan H., Andrew J. Kerkhoff, and Melanie E. Moses

2007 Global Patterns of City Size Distributions and Their Fundamental Drivers. *PLoS ONE*

2(9): e934.

Drennan, Robert D., and Christian E. Peterson

2004 Comparing Archaeological Settlement Systems with Rank-Size Graphs: A Measure of

Shape and Statistical Confidence. *Journal of Archaeological Science* 31(5): 533–549.

Edwards, Andrew M., Richard A. Phillips, Nicholas W. Watkins, Mervyn P. Freeman, Eugene J.

Murphy, Vsevolod Afanasyev, Sergey V. Buldyrev, M. G. E. da Luz, E. P. Raposo, H. Eugene

Stanley, and Gandhimohan M. Viswanathan

2007 Revisiting Levy Flight Search Patterns of Wandering Albatrosses, Bumblebees and

Deer. *Nature* 449(7165): 1044–1048.

Flannery, Kent V.

1998 The Ground Plans of Archaic States. In *Archaic states*, edited by Gary M. Feinman and

Joyce Marcus, pp. 15–57. School of American Research Press, Santa Fe.

Gabaix, Xavier

1999 Zipf’s Law and the Growth of Cities. *The American Economic Review* 89(2): 129–132.

Griffin, Arthur F.

2011 Emergence of Fusion/Fission Cycling and Self-Organized Criticality from a Simulation

Model of Early Complex Polities. *Journal of Archaeological Science* 38(4): 873–883.

Griffin, Arthur F., and Charles Stanish

2007 An Agent-Based Model of Prehistoric Settlement Patterns and Political Consolidation in the Lake Titicaca Basin of Peru and Bolivia. *Structure and Dynamics: eJournal of *

*Anthropological and Related Sciences* 2(2): 1–47.

Grøn, Ole

2012 Our Grandfather Sent the Elk – Some Problems for Hunter-Gatherer Predictive

Modelling. *Quartär* 59: 175–188.

Hahn, Matthew W., and R. Alexander Bentley

2003 Drift as a Mechanism for Cultural Change: An Example from Baby Names.

152

*Proceedings: Biological Sciences* 270: S120–S123.

Hamilton, Marcus J., Bruce T. Milne, Robert S. Walker, and James H. Brown

2007 Nonlinear Scaling of Space use in Human Hunter–Gatherers. *Proceedings of the *

*National Academy of Sciences* 104(11): 4765–4769.

Hastorf, Christine A

2008 The Formative Period in the Titicaca Basin. In *Handbook of South American *

*Achaeology*, edited by Helaine Silverman and William H. Isbell, pp. 545–561. Springer,

New York NY.

Inomata, Takeshi, and Kazuo Aoyama

1996 Central-Place Analyses in the La Entrada Region, Honduras: Implications for

Understanding the Classic Maya Political and Economic Systems. *Latin American *

*Antiquity* 7(4): 291–312.

Jochim, Michael

1981 *Strategies for Survival: Cultural Behavior in an Ecological Context*. Academic Press,

New York.

Johnson, Gregory A.

1980 Rank-Size Convexity and System Integration: A View from Archaeology. *Economic *

*Geography* 56(3): 234–247.

Kelly, Robert L.

1995 *The Foraging Spectrum: Diversity in Hunter-Gatherer Lifeways*. Percheron Press, New

York.

Kelly, Robert L., Todd A. Surovell, Bryan N. Shuman, and Geoffrey M. Smith

2013 A Continuous Climatic Impact on Holocene Human Population in the Rocky

Mountains. *Proceedings of the National Academy of Sciences* 110(2): 443–447.

Kendal, Jeremy, Jamshid J. Tehrani, and John Odling-Smee

2011 Human Niche Construction in Interdisciplinary Focus. *Philosophical Transactions of *

*the Royal Society B: Biological Sciences* 366(1566): 785–792.

Klink, Cynthia J.

2005 Archaic Period Research in the Río Huenque Valley, Peru. In *Advances in Titicaca *

*Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 13–24. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Klink, Cynthia J., and Mark S. Aldenderfer

2005 A Projectile Point Chronology for the South-Central Andean Highlands. In *Advances in *

*Titicaca Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 25–54. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Krugman, Paul

1996 *The Self-Organizing Economy*. Blackwell Publishers, Cambridge.

Lansing, J. Stephen

2003 Complex Adaptive Systems. *Annual Review of Anthropology* 32: 183–204.

Lansing, J. Stephen, and Murray P. Cox

2011 The Domain of the Replicators. *Current Anthropology* 52(1): 105–125.

Lansing, J. Stephen, and Sean S. Downey

153

2009 Complexity and Anthropology. In *Philosophy of Complex Systems*, edited by Cliff

Hooker, 10:pp. 569–601. Handbook of the Philosophy of Science. Elsevier, Boston.

Lee, Jay, and David W. S Wong

2001 *Statistical Analysis with ArcView GIS*. John Wiley, New York.

Mitchell, Peter, Ina Plug, Geoff Bailey, Ruth Charles, Amanda Esterhuysen, Julia Lee Thorp,

Adrian Parker, and Stephan Woodborne

2011 Beyond the Drip-Line: A High-Resolution Open-Air Holocene Hunter-Gatherer

Sequence from Highland Lesotho. *Antiquity* 85(330): 1225–1242.

Morgan, Christopher

2009 Climate Change, Uncertainty and Prehistoric Hunter-Gatherer Mobility. *Journal of *

*Anthropological Archaeology* 28(4): 382–396.

Newman, M. E. J.

2005 Power Laws, Pareto Distributions and Zipf’s Law. *Contemporary Physics* 46(5): 323–

351.

Odling-Smee, F. John, Kevin N. Laland, and Marcus W. Feldman

1996 Niche Construction. *The American Naturalist* 147(4): 641–648.

Ortmann, Anthony L.

2010 Placing the Poverty Point Mounds in Their Temporal Context. *American Antiquity*

75(3): 657–678.

Paynter, Robert

1982 *Models of Spatial Inequality: Settlement Patterns in Historical Archeology*. Academic

Press, New York.

Premo, Luke S.

2006 Agent-based Models as Behavioral Laboratories for Evolutionary Anthropological

Research. *Arizona Anthropologist* 17: 91–113.

2007 Exploratory Agent-based Models: Towards an Experimental Ethnoarchaeology. In

*Digital Discovery: Exploring New Frontiers in Human Heritage*, edited by J.T. Clark and

E.M. Hagemeister, pp. 29–36. Computer Applications and Quantitative Methods in

Archaeology. Archeolingua Press, Budapest.

Reimer, Paula

2013 IntCal13 and Marine13 Radiocarbon Age Calibration Curves 0–50,000 Years cal BP.

*Radiocarbon* 55(4): 1869–1887.

Richerson, Peter J., Robert Boyd, and Robert L. Bettinger

2001 Was Agriculture Impossible during the Pleistocene but Mandatory during the Holocene?

A Climate Change Hypothesis. *American Antiquity* 66: 387–412.

Rigsby, Catherine A., Paul A. Baker, and Mark S. Aldenderfer

2003 Fluvial history of the Rio Ilave valley, Peru, and its relationship to climate and human history. *Palaeogeography, Palaeoclimatology, Palaeoecology* 194(1-3): 165–185.

Rindos, David

1983 *The Origins of Agriculture: An Evolutionary Perspective*. Academic Press, New York.

Rumold, Claudia U.

2010 Illuminating Women’s Work and the Advent of Plant Cultivation in the Highland

154

Titicaca Basin of South America: New Evidence from Grinding Tool and Starch Grain

Analyses. Unpublished Ph.D. dissertation, University of California, Santa Barbara.

Saunders, Joe W, Rolfe D Mandel, Roger T Saucier, E. Thurman Allen, C. T Hallmark, Jay K

Johnson, Edwin H Jackson, Charles M Allen, Gary L Stringer, Douglas S Frink, James K

Feathers, Stephen Williams, Kristen J Gremillion, Malcolm F Vidrine, and Reca Jones

1997 A Mound Complex in Louisiana at 5400-5000 Years Before the Present. *Science*

277(5333): 1796–1799.

Saunders, Joe W., Rolfe D. Mandel, C. Garth Sampson, Charles M. Allen, E. Thurman Allen,

Daniel A. Bush, James K. Feathers, Kristen J. Gremillion, C. T. Hallmark, H. Edwin Jackson, Jay

K. Johnson, Reca Jones, Roger T. Saucier, Gary L. Stringer, and Malcolm F. Vidrine

2005 Watson Brake, a Middle Archaic Mound Complex in Northeast Louisiana. *American *

*Antiquity* 70(4): 631–668.

Schiffer, Michael B.

1987 *Formation Processes of the Archaeological Record*. 1st ed. University of New Mexico

Press, Albuquerque NM.

1995 *Behavioral Archaeology: First Principles*. University of Utah Press, Salt Lake City.

Smith, Bruce D.

1995 *The Emergence of Agriculture*. Scientific American Library series 54. Scientific

American Library, New York.

2001 Low-Level Food Production. *Journal of Archaeological Research* 9(1): 1–43.

Smith, Eric Alden

1981 The Application of Optimal Foraging Theory to the Analysis of Hunter-Gatherer Group

Size. In *Hunter-Gatherer Foraging Strategies: Ethnographic and Archeological Analyses*, edited by Bruce Winterhalder and Eric Alden Smith, pp. 36–65. University of Chicago

Press, Chicago.

Stanish, Charles

2003 *Ancient Titicaca: The Evolution of Complex Society in Southern Peru and Northern *

*Bolivia*. University of California Press, Berkeley.

Stiger, Mark

2001 *Hunter-Gatherer Archaeology of the Colorado High Country*. University Press of

Colorado, Boulder.

Surovell, Todd A., Nicole M. Waguespack, James H. Mayer, Marcel Kornfeld, and George C.

Frison

2005 Shallow Site Archaeology: Artifact Dispersal, Stratigraphy, and Radiocarbon Dating at the Barger Gulch Locality B Folsom Site, Middle Park, Colorado. *Geoarchaeology* 20(6):

627–649.

The R Foundation

2009 *R Statistical Computing Language*. Wien.

Wilensky, U.

1999 *NetLogo*. Center for Connected Learning and Computer-Based Modeling, Northwestern

University, Evanston.

Winterhalder, Bruce, Douglas J. Kennett, Mark N. Grote, and Jacob Bartruff

155

2010 Ideal Free Settlement of California’s Northern Channel Islands. *Journal of *

*Anthropological Archaeology* 29(4): 469–490.

Winterhalder, Bruce, and R. Brooke Thomas

1978 *Geoecology of Southern Highland Peru: A Human Adaptation Perspective*. University of Colorado, Institute of Arctic and Alpine Research, Boulder.

Yellen, John

1977 *Archaeological Approaches to the Present: Models for Reconstructing the Past*.

Academic Press, New York.

Zeanah, David W.

2002 Central Place Foraging and Prehistoric Pinyon Utilization in the Great Basin. In *Beyond*

*Foraging and Collecting: Evolutionary Change in Hunter-Gatherer Settlement Systems*, edited by Ben Fitzhugh, pp. 231–256. Kluwer Academic/Plenum Publishers, New York.

Zipf, George Kingsley

1949 *Human Behavior and the Principle of Least Effort: An Introduction to Human Ecology*.

Addison-Wesley Press, Cambridge.

156

SUPPLEMENTARY MATERIAL

This supplementary material section provides the original code used to test the hypotheses presented in this dissertation. This includes NetLogo ABM code for the working niche-construction model of forager mobility and R code used in the average nearest neighbor distance analysis.

**NetLogo code for Working Agent-Based Model**

breed [foragers forager] breed [materials material] breed [sites site] foragers-own [

target

homesite] materials-own [

homesite

nnd] ;;nearest neighbor distance sites-own[

homesite

material-count

first-use

last-use

occupation-span

area] to setup

clear-all

reset-ticks

ask patches [set pcolor 101]

set-default-shape foragers "person"

set-default-shape materials "circle"

set-default-shape sites "circle"

create-foragers 1[

set color red

set size 1.5

setxy random-xcor random-ycor

set homesite 1

hatch-materials 1

[set color 46

set size dot-size

hatch-sites 1[

set color white

set size 1

set material-count 1

set first-use ticks

set last-use ticks

]

]

] end to go

ask foragers [

ifelse p-move-to-material >= random 101

[set target [who] of one-of materials

set homesite [homesite] of material target

move-to material target

]

[setxy random-xcor random-ycor

set homesite max [homesite] of materials + 1

hatch-sites 1[

set color white

set size 1

set first-use ticks

]]

hatch-materials 1[

set color 46

set shape "circle"

set size dot-size

rt random 360 fd random-normal 0 jitter]

ask sites with [homesite = [homesite] of forager 0] [

set material-count material-count + 1

]]

ask sites [

if any? foragers with [homesite = [homesite] of myself] [set last-use ticks]

ifelse hide-sites?

[set hidden? True]

[set hidden? False]

157

]

tick

if ticks = stop-tick [stop]

if count sites = stop-sites [stop]

if count sites with [material-count > 1] = stop-areas [stop] end to survey

ask sites [

set occupation-span last-use - first-use + 1

let xmax max [xcor] of materials with [homesite = [homesite] of myself]

let xmin min [xcor] of materials with [homesite = [homesite] of myself]

let ymax max [ycor] of materials with [homesite = [homesite] of myself]

let ymin min [ycor] of materials with [homesite = [homesite] of myself]

set area (xmax - xmin) * (ymax - ymin)

histogram [material-count] of sites

]

ask materials [

set nnd distance min-one-of other materials [distance myself]

] end

**R code for Average Nearest Neighbor Distance analysis**

#Average Nearest Neighbor

##############################################

#synthetic test datasets for proof-of-concept

#area for all test cases equals 1

#dispersed case, a grid dx<-c() for (i in 1:10){

dx<-c(dx,rep(i,10))} dy<-rep(1:10,10) dxy<-data.frame(dx/10,dy/10)

#random case rx<-runif(100,0,1) ry<-runif(100,0,1) rxy<-data.frame(rx,ry)

#clustered case

158

cx<-rnorm(100,.5,.05) cy<-rnorm(100,.5,.05) cxy<-data.frame(cx,cy)

############################################## ann<-function(x,area){

#x can be a vector or nearest-neighbor distances or a table of coordinate pairs in data frame or matrix format.

#area is the user-defined study area if(is.data.frame(x)){

x<-as.matrix(dist(x,upper=T))

} if(is.matrix(x)){

x<-apply(x,1,function(x){min(x[x>0])})

} if(is.vector(x)){

do<-mean(x)

de<-.5/sqrt(length(x)/area)

annr<-do/de

#Monte Carlo simulation

D<-c()

p2s<-c()

pg<-c()

pl<-c()

for (i in 1:100){

synx<-runif(length(x),0,sqrt(area))

syny<-runif(length(x),0,sqrt(area))

syn<-data.frame(synx,syny)

synd<-as.matrix(dist(syn,upper=T))

synd<-apply(synd,1,function(x){min(x[x>0])})

D<-c(D,ks.test(x,synd)$statistic)

p2s<-c(p2s,ks.test(x,synd)$p.value)

}

} hist(x) list(

'expected mean distance'=de,

'observed mean distance'=do,

'nearest neighbor ratio'=annr,

D=mean(D),

'p random'=mean(p2s)

159

)

}

**R Code for Univariate Model Plausibility Tests (Discrete Data)**

source("F:\\R\\mleKSPois.R") source("F:\\R\\mleKSGeom.R") source("F:\\R\\mleKSpld.R") source("F:\\R\\mleKSpldt.R") mpd<-function(x){ plausible<-c() a<-mleKSPois(x) b<-mleKSgeom(x) c<-mleKSpld(x) d<-mleKSpldt(x)

#generate list of plausible models that can be passed to Aic_disc if (a$p>0.1) plausible<-c(plausible,"Poisson") if (b$p>0.1) plausible<-c(plausible,"geometric") if (c$p>0.1) plausible<-c(plausible,"power law") if (d$p>0.1 & d$alpha > 1 & d$alpha < 3) plausible<-c(plausible,"power-law tail")

#show results of Poisson fit cat("Poisson", '\n',

"n: ", a$n, '\n',

"sum: ", a$s,'\n',

"lambda: ", a$lambda, '\n',

"D: ", a$D, '\n',

"p: ", a$p, '\n')

#show results of geometric fit cat("geometric", '\n',

"n: ", b$n, '\n',

"sum: ", b$s,'\n',

"prob: ", b$prob, '\n',

"D: ", b$D, '\n',

"p: ", b$p, '\n')

#show results of power-law fit cat("power law", '\n',

"n: ", c$n, '\n',

160

"sum: ", c$s,'\n',

"alpha: ", c$alpha, '\n',

"xmin: ", c$xmin, '\n',

"D: ", c$D, '\n',

"p: ", c$p, '\n')

#show results of power-law tail fit cat("power-law tail", '\n',

"n: ", d$n, '\n',

"sum: ", d$s,'\n',

"alpha: ", d$alpha, '\n',

"xmin: ", d$xmin, '\n',

"D: ", d$D, '\n',

"p: ", d$p, '\n') return(list(plausible=plausible))

}

**R Code for Univariate Model Plausibility Tests (Continuous Data)**

source("F:\\R\\mleKSnorm.R") source("F:\\R\\mleKSexp.R") source("F:\\R\\mleKSlnorm.R") source("F:\\R\\mleKSplc.R") source("F:\\R\\mleKSplct.R") mpc<-function(x){ plausible<-c() a<-mleKSnorm(x) b<-mleKSexp(x) c<-mleKSlnorm(x) d<-mleKSplc(x) e<-mleKSplct(x)

#generate list of plausible models that can be passed to Aic_disc if (a$p>=0.1) plausible<-c(plausible,"normal") if (b$p>=0.1) plausible<-c(plausible,"exponential") if (c$p>=0.1) plausible<-c(plausible,"lognormal") if (d$p>=0.1) plausible<-c(plausible,"power law") if (e$p>=0.1 & e$alpha > 1 & e$alpha < 3) plausible<-c(plausible,"power-law tail")

#show results of normal fit

161

cat("normal", '\n',

"n: ", a$n, '\n',

"sum: ", a$s,'\n',

"mean: ", a$mean, '\n',

"sd: ", a$sd, '\n',

"D: ", a$D, '\n',

"p: ", a$p, '\n')

#show results of exponential fit cat("exponential", '\n',

"n: ", b$n, '\n',

"sum: ", b$s,'\n',

"rate: ", b$rate, '\n',

"D: ", b$D, '\n',

"p: ", b$p, '\n')

#show results of lognormal fit cat("lognormal", '\n',

"n: ", c$n, '\n',

"sum: ", c$s,'\n',

"meanlog: ", c$meanlog, '\n',

"sdlog: ", c$sdlog, '\n',

"D: ", c$D, '\n',

"p: ", c$p, '\n')

#show results of power-law fit cat("power law", '\n',

"n: ", d$n, '\n',

"sum: ", d$s,'\n',

"alpha: ", d$alpha, '\n',

"xmin: ", d$xmin, '\n',

"D: ", d$D, '\n',

"p: ", d$p, '\n')

#show results of power-law tail fit cat("power-law tail", '\n',

"n: ", e$n, '\n',

"sum: ", e$s,'\n',

"alpha: ", e$alpha, '\n',

"xmin: ", e$xmin, '\n',

"D: ", e$D, '\n',

"p: ", e$p, '\n','\n') return(list(plausible=plausible))

}

162

**R code for Poisson Distribution Plausibility Test**

library(MASS) #(Venables and Ripley 2002) mleKSPois = function(x, limit=2500) { mle<-fitdistr(x, "Poisson") d<-ks.test(x, "ppois", lambda = mle$estimate)$stat

# compute p-value using synthetic data count<-0 for (i in 1:limit) {

syn<-rpois(length(x), lambda = mle$estimate)

synmle<-fitdistr(syn, "Poisson")

synd<-ks.test(syn, "ppois", lambda = synmle$estimate)$stat

if(synd >= d) {count = count + 1}

} return(list( n=length(x), s=sum(x), lambda=as.double(mle$estimate),

D=as.double(d), p=count/limit, loglik=mle$loglik))

}

**R code for Geometric Distribution Plausibility Test**

library(MASS) #(Venables and Ripley 2002) mleKSgeom = function(x, limit=2500) { mle<-fitdistr(x, "geometric") d<-ks.test(x, "pgeom", prob = mle$estimate["prob"])$stat

# compute p-value using synthetic data count<-0 for (i in 1:limit) {

syn<-rgeom(length(x), prob = mle$estimate["prob"])

synmle<-fitdistr(syn, "geometric")

synd<-ks.test(syn, "pgeom", prob = synmle$estimate["prob"])$stat

if(synd >= d) {count = count + 1}

163

} return(list( n=length(x), s=sum(x), prob=as.double(mle$estimate),

D=as.double(d), p=count/limit, loglik=mle$loglik))

}

**R code for Discrete Power-Law Distribution Plausibility Test**

library(poweRlaw) #(Gillespie 2013) mleKSpld<-function(x){

#ensure that all values are considered if zeros present s<-sum(x) adj=F if (min(x)==0){ x<-x+1 adj=T warning("Zero values present. Data + 1 for calculations.")

}

#power-law fit with xmin set to minimum of data xp<-displ$new(x) xp$setXmin(min(x)) params<-estimate_pars(xp) xp$setPars(params$pars)

D=get_KS_statistic(xp) p<-bootstrap_p(xp, xmins=c(min(x),min(x),min(x)))$p #default xmin is computed, therefore only calculating for tail. To calculate for whole distribution, we have to trick it by specifying xmins.

loglik<-dist_ll(xp) xmin<-ifelse(adj==F, min(x), min(x)-1) return(list( alpha=params$pars, xmin=xmin, n=xp$internal$n,

164

s=s,

D=D, p=p, loglik=loglik))

}

**R code for Discrete Power-Law Tail Plausibility Test**

library(poweRlaw) #(Gillespie 2013) mleKSpldt<-function(x){

#ensure that all values are considered if zeros present adj=F if (min(x)==0){ x<-x+1 adj=T warning("Zero values present. Data + 1 for calculations.")

}

#power-law fit with xmin computed, i.e., upper tail xp<-displ$new(x) params<-estimate_xmin(xp) xp$setXmin(params$xmin) xp$setPars(params$pars) p<-bootstrap_p(xp)$p loglik<-dist_ll(xp) xmin<-ifelse(adj==F, params$xmin, params$xmin-1) s<-ifelse(adj==F, sum(x[x>=params$xmin]), sum(x[x>=params$xmin]-1)) return(list( alpha=params$pars, xmin=xmin, n=length(x[x>=params$xmin]), s=s,

D=params$KS, p=p, loglik=loglik))

}

165

**R code for Normal Distribution Plausibility Test**

library(MASS) #(Venables and Ripley 2002) mleKSnorm = function(x, limit=2500) { mle<-fitdistr(x, "normal") d<-ks.test(x, "pnorm", mean = mle$estimate["mean"], sd = mle$estimate["sd"])$stat

# compute p-value using synthetic data count<-0 for (i in 1:limit) {

syn<-rnorm(length(x), mean = mle$estimate["mean"], sd = mle$estimate["sd"])

synmle<-fitdistr(syn, "normal")

synd<-ks.test(syn, "pnorm", mean = synmle$estimate["mean"], sd = synmle$estimate["sd"])

$stat

if(synd >= d) {count = count + 1}

} return(list( n=length(x), s=sum(x), mean=as.double(mle$estimate["mean"]), sd=as.double(mle$estimate["sd"]),

D=as.double(d), p=count/limit, loglik=mle$loglik))

}

**R code for Exponential Distribution Plausibility Test**

library(MASS) #(Venables and Ripley 2002) mleKSexp<-function(x, limit=2500) { mle<-fitdistr(x, "exponential") rate<d<-ks.test(x, "pexp", rate = mle$estimate["rate"])$stat

# compute p-value using synthetic data count<-0 for (i in 1:limit) {

syn<-rexp(length(x), rate = mle$estimate["rate"])

166

mlesyn<-fitdistr(syn, "exponential")

synd<-ks.test(syn, "pexp", rate = mlesyn$estimate["rate"])$stat

if(synd >= d) {count = count + 1}

} return(list( n=length(x), s=sum(x), rate=as.double(mle$estimate),

D=as.double(d), p=count/limit, loglik=mle$loglik))

}

**R code for Lognormal Distribution Plausibility Test**

library(MASS) #(Venables and Ripley 2002) mleKSlnorm = function(x, limit=2500) { mle<-fitdistr(x, "lognormal") d<-ks.test(x, "plnorm", meanlog = mle$estimate["meanlog"], sdlog = mle$estimate["sdlog"])

$stat

# compute p-value using synthetic data count<-0 for (i in 1:limit) {

syn<-rlnorm(length(x), meanlog = mle$estimate["meanlog"], sdlog = mle$estimate["sdlog"])

synmle<-fitdistr(syn, "lognormal")

synd<-ks.test(syn,"plnorm", meanlog = synmle$estimate["meanlog"], sdlog = synmle$estimate["sdlog"])$stat

if(synd >= d) {count = count + 1}

} return(list( n=length(x), s=sum(x), meanlog=as.double(mle$estimate["meanlog"]), sdlog=as.double(mle$estimate["sdlog"]),

D=as.double(d), p=count/limit, loglik=mle$loglik))

}

167

**R Code for Continuous Power-Law Distribution Plausibility Test**

library(poweRlaw) #(Gillespie 2013) mleKSplc<-function(x){

#power law fit to all data (not just tail) xp<-conpl$new(x) xp$setXmin(min(x)) params<-estimate_pars(xp) xp$setPars(params$pars)

D=get_KS_statistic(xp) p<-bootstrap_p(xp, xmins=c(min(x),min(x),min(x)))$p #default xmin is computed, therefore only calculating for tail. To calculate for whole distribution, we have to trick it by specifying xmins.

loglik<-dist_ll(xp) return(list( alpha=params$pars, xmin=xp$xmin, n=xp$internal$n, s=sum(x),

D=D, p=p, loglik=loglik))

}

**R Code for Continuous Power-Law Tail Distribution Plausibility Test**

library(poweRlaw) #(Gillespie 2013)

#power-law fit with xmin computed, i.e., upper tail mleKSplct<-function(x, limit=2500){ xp<-conpl$new(x) params<-estimate_xmin(xp) xp$setXmin(params$xmin) xp$setPars(params$pars) p<-bootstrap_p(xp)$p loglik<-dist_ll(xp) xmin<-params$xmin

168

s<-sum(x[x>=params$xmin]) return(list( alpha=params$pars, xmin=xmin, n=length(x[x>=params$xmin]), s=s,

D=params$KS, p=p, loglik=loglik))

}

**R Code for Akaike Information Criterion Analysis (Discrete Data)**

library(poweRlaw) #(Gillespie 2013)

Aic_disc <- function(x, models=c("Poisson","geometric","power law"), xmin=NULL){ #can specify an xmin parameter for power law tail.

#Prevent error in powerlaw fitting if xmin is zero and ensure all cases are kept if (is.null(xmin)){

ifelse (min(x)==0, (x<-x+1)&(xmin<-1), xmin<-min(x))

}

#compute Aic list for models aics<-c() models2<-c() if(any(models=="Poisson")){

models2<-c(models2,"Poisson")

#models2 ensures correct ordering of the output

p<-fitdistr(x,"Poisson")

aicp<-AIC(p)

aics<-c(aics, aicp)

} if(any(models=="geometric")){

models2<-c(models2,"geometric")

g<-fitdistr(x,"geometric")

aicg<-AIC(g)

aics<-c(aics, aicg)

} if(any(models=="power law")){

models2<-c(models2,"power law")

169

pl<-displ$new(x)

pl$setXmin(min(x))

pl$setPars(estimate_pars(pl)$pars)

aicpl<- -2*dist_ll(pl)+2*1

aics<-c(aics, aicpl)

}

#compute Aic weights list aic_min<-min(aics) aic_count<-length(aics) sv<-c() if(any(models=="Poisson")){

dp<-exp(1)^(-(aicp-aic_min)/aic_count)

sv<-c(sv,dp)

} if(any(models=="geometric")){

dg<-exp(1)^(-(aicg-aic_min)/aic_count)

sv<-c(sv,dg)

} if(any(models=="power law")){

dpl<-exp(1)^(-(aicpl-aic_min)/aic_count)

sv<-c(sv,dpl)

} s<-sum(sv) weights<-c() weight_list<-list() if(any(models=="Poisson")){

wp<-dp/s

weights<-c(weights,wp)

weight_list<-c(weight_list,"Pois.w"=wp)

} if(any(models=="geometric")){

wg<-dg/s

weights<-c(weights,wg)

weight_list<-c(weight_list,"geom.w"=wg)

} if(any(models=="power law")){

wpl<-dpl/s

weights<-c(weights, wpl)

weight_list<-c(weight_list,"pl.w"=wpl)

}

170

model=models2 aic=round(aics,1) weight=round(weights,2)

} print(data.frame(model,aic,weight)) cat('\n') return(weight_list)

**R Code for Akaike Information Criterion Analysis (Continuous Data)**

library(MASS) #(Venables and Ripley 2002) library(poweRlaw) #(Gillespie 2013)

Aic_cont <- function (x, models=c("normal","exponential","lognormal","power law"), xmin=min(x)){ #can specify an xmin parameter for power law tail.

#compute Aic list for models aics<-c() models2<-c() if(any(models=="normal")){

models2<-c(models2,"normal")

#models2 ensures correct ordering of the output

n<-fitdistr(x,"normal")

aicn<-AIC(n)

aics<-c(aics,aicn)

} if(any(models=="exponential")){

models2<-c(models2,"exponential")

e<-fitdistr(x,"exponential")

aice<-AIC(e)

aics<-c(aics,aice)

} if(any(models=="lognormal"|models=="log-normal")){

models2<-c(models2,"lognormal")

l<-fitdistr(x,"lognormal")

aicl<-AIC(l)

aics<-c(aics,aicl)

} if(any(models=="power law")){

171

models2<-c(models2,"power law")

pl<-conpl$new(x)

pl$setXmin(min(x))

pl$setPars(estimate_pars(pl)$pars)

aicpl<- -2*dist_ll(pl)+2*1

aics<-c(aics, aicpl)

}

#compute Aic weights list aic_min<-min(aics) aic_count<-length(aics) sv<-c() if(any(models=="normal")){

dn<-exp(1)^(-(aicn-aic_min)/aic_count)

sv<-c(sv,dn)

} if(any(models=="exponential")){

de<-exp(1)^(-(aice-aic_min)/aic_count)

sv<-c(sv,de)

} if(any(models=="lognormal"|models=="log-normal")){

dl<-exp(1)^(-(aicl-aic_min)/aic_count)

sv<-c(sv,dl)

} if(any(models=="power law")){

dpl<-exp(1)^(-(aicpl-aic_min)/aic_count)

sv<-c(sv,dpl)

} s<-sum(sv) weights<-c() weight_list<-list() if(any(models=="normal")){

wn<-dn/s

weights<-c(weights,wn)

weight_list<-c(weight_list,"norm.w"=wn)

} if(any(models=="exponential")){

we<-de/s

weights<-c(weights,we)

weight_list<-c(weight_list,"exp.w"=we)

}

172

if(any(models=="lognormal"|models=="log-normal")){

wl<-dl/s

weights<-c(weights,wl)

weight_list<-c(weight_list,"ln.w"=wl)

} if(any(models=="power law")){

wpl<-dpl/s

weights<-c(weights,wpl)

weight_list<-c(weight_list,"pl.w"=wpl)

} model=models2 aic=round(aics,1) weight=round(weights,2) print(data.frame(model,aic,weight)) cat('\n') return(weight_list)

}

**References Cited**

Gillespie, Colin S.

2013 Fitting Heavy Tailed Distributions: The Powerlaw Package.

Venables, W. N, and Brian D Ripley

2002 *Modern applied statistics with S*. Fourth. Springer, New York.

173

174

APPENDIX III:

HUNTER-GATHERERS OF THE EVE OF COMPLEXITY: INVESTIGATIONS AT SORO

MIK'AYA PATJXA, LAKE TITICACA BASIN, PERU, 8000-6700 B.P.

Authors and Affiliations:

W. Randall Haas, Jr.

a

Carlos Viviano Llave b a School of Anthropology, The University of Arizona, Tucson, AZ 85701-0030; b Peruvian

National Register of Archaeologists

Contributions: WRH designed the research; carried out field, laboratory, and analytical work; and wrote the paper. CVL carried out field and laboratory work.

175

ABSTRACT

Recent excavations at the site of Soro Mik'aya Patjxa reveal the earliest securely dated cultural features in the Lake Titicaca Basin, Peru and Bolivia. Radiocarbon assays place site occupation across the Middle-Late Archaic period transition between 8,000-6,700 cal. B.P. This paper draws on the rich material assemblage to identify behaviors among the Basin's last huntergatherers that anticipate later developments in the trajectory to socioeconomic complexity. The data suggest that residentially mobile hunter-gatherers occupied the site repeatedly for more than a millennium. Intensive land-use practices and conspecific competition presaged incipient sedentism, agriculture, and land tenure in subsequent periods.

INTRODUCTION

More than 7,000 years after mobile hunter-gatherer populations colonized the Lake

Titicaca Basin, the state of Tiwanaku emerged around 1,500 cal. B.P. with urban centers, monumental public works, specialized art and craft production, political and religious elites, a calendric system, and large-scale food production (Janusek 2008; Kolata 1993). Understanding the profound socioeconomic transformations that occurred over this interval is a central task of

Andean archaeology (Moseley 1992; Stanish 2001). A comprehensive understanding of the origins of the Tiwanaku requires archaeological study of the emergence of the complex behaviors that define it. Archaeological research to date has shed considerable light on appearance of such behaviors, particularly through survey and excavation of sites dating to the Tiwanaku and

Formative periods, 3,500-1,500 cal. B.P. (Figure 20). Such investigations show that residential

years cal. B.P.

chronology

Pan-Andean

(Rowe & Menzel

1967)

Titicaca Basin

(Janusek 2004; Klink

& Aldenderfer 2005)

Tiwanaku

2000

Middle Horizon

Early

Intermediate

Early Horizon

Formative

Late

Middle key sites

3000

Initial Period Early

4000

Terminal

Preceramic

5000

Late 6000

7000

Archaic

8000 Lithic Middle

9000

10,000 Early

*Figure 20. Chronological and geographic context of Soro Mik'aya Patjxa and other sites discussed in the text. Map projection: UTM19S (WGS 84). *

176

177 sedentism and committed agro-pastoralism—two behaviors of interest here—were firmly in place by at least 3,500 cal. B.P. (Bandy 2006:216; Bruno 2006; Hastorf 2008; Stanish 2003).

More-limited research pushes incipient forms of residential sedentism and resource intensification into the Terminal Archaic Period, sometime between 5,000 and 3,500 cal. B.P.

(Aldenderfer and Flores Blanco 2011; Craig 2011). On one hand, an absence of masonry architecture and ceramic vessels at the Terminal Archaic Period site of Jiskairumoko in the western Basin indicates a greater degree of residential mobility than for the subsequent

Formative periods (Craig 2011; Steadman 1995). On the other hand, the presence pit houses, storage-pit features, and abundant groundstone suggests a greater degree of residential permanence than typically expected for highly mobile hunter-gatherers (Craig 2011; Rumold

2010). Moreover, the presence of nine small (<1 m tall) burial mounds at the site of Kaillachuro suggest a degree of land tenure and territoriality, though we cannot yet speak to the interval over which the interments were made due to limited investigation (Aldenderfer 2005:25). Taken together, these observations have led Mark Aldenderfer and Nathan Craig to characterize the

Titicaca Basin Terminal Archaic as a period of low-level food production (Aldenderfer et al.

2008; Craig 2011:385).

At present, empirical limitations constrain our understanding of the hunter-gatherer behaviors that precipitated low-level food production in the Terminal Archaic Period. Our most robust observations come from three independent settlement-pattern analyses, which indicate that Archaic populations peaked during the Late Archaic Period before declining subsequently in the Terminal Archaic Period (Cipolla 2005; Craig 2011; Klink 2005). Possible ephemeral

178 habitation structures identified at the Late Archaic Period site of Pirco suggest high residential mobility, and minor amounts of groundstone suggest possible seed processing (Craig 2011:375).

Unfortunately, a lack of direct dates limit Pirco's interpretative potential. More-direct evidence for Late Archaic subsistence intensification comes from Qillqatani, a stratified rockshelter located along the western margin of the Basin. Faunal data reveal Archaic consumption of wild ungulates with increasing emphasis over time on neonates and low-utility parts of camelids suggesting intensified camelid use (Aldenderfer 1989a:122; Aldenderfer 1989b:146–147).

Beyond population growth and possible resource intensification, we currently can say little about how, if at all, the Titicaca Basin's last hunter-gatherers precipitated early sedentism and food production in the subsequent Terminal Archaic and Early Formative periods (see also

Aldenderfer and Flores Blanco 2011; Bandy 2006:214–216).

Recent excavations at the hunter-gatherer site of Soro Mik'aya Patjxa (SMP, Ilave 95-

259) in the western Titicaca Basin contribute key insights toward filling the empirical gap.

Radiocarbon assays—presented below—place the occupation across the Middle-Late Archaic transition between 8,000 and 6,700 cal. B.P., more than 1,500 years before the posited onset of low-level food production. The dates come from secure pit-feature contexts making them the only securely dated cultural features from unequivocal hunter-gatherer contexts in the Titicaca

Basin and currently the earliest dated cultural features in the Basin.

The evidence presented here suggests that SMP was a residential site used on a nonpermanent basis for more than a millenium. Mobile hunter gatherers appear to have intensively processed and consumed seeds while engaging in resource competition and interpersonal

179 conflict. These intensification behaviors anticipate the emergence of early sedentism and agriculture in subsequent times. The arguments for these inferences are detailed below following a summary of the investigations at SMP.

SUMMARY OF INVESTIGATIONS AT SORO MIK'AYA PATJXA (SMP)

In 2013, we conducted archaeological investigations at the site of Soro Mik'aya Patjxa

(SMP; Ilave 95-259). Mark Aldenderfer and colleagues first recorded the site in 1995 as one of

468 sites in a 41-km 2 pedestrian survey designed to outline the Basin's preceramic prehistory

(Craig 2011). SMP is a high-density artifact scatter covering 2,800 m 2 located at S16º14'2”,

W69º43'30”, 3,860 masl (WGS 84). The site is situated on a relict alluvial terrace in the Ilave

Basin (Rigsby et al. 2003) approximately approximately 1 km north of the modern Huenque

River, 1.5 km south of the nearest bedrock exposure, and 30-km west of the modern shoreline of

Lake Titicaca. The landscape consists of rolling hills and vast grasslands, or *pampas*, that are

grazed and under cultivation today (Figure 21). The site was selected for subsurface investigation

because of its high surface concentration of temporally diagnostic Late Archaic Period artifacts and the local community's interest in hosting and participating in the excavation effort.** **

We excavated approximately 21 m 3 of sediment from 50 m 2 of surface area at SMP

(Figure 22). Fifty meters of 50-cm-wide exploratory trenches were extended through areas of

high surface-artifact concentration and potential for localized burial. When cultural features were encountered, excavation units were extended to define the spatial extents of the features and excavate their contents. The excavations revealed a single, thin cultural stratum (stratum IIa)

180

*Figure 21: Kite aerial view looking northwest across the site of Soro Mik'aya Patjxa, located in a vast altiplano pampa near the center of the Ilave Basin. Site boundary approximated by dashed white line.*

below the modern plow zone (stratum I) and grading down into culturally sterile alluvium

Temporally diagnostic artifacts and radiocarbon dates suggest that site-use occurred almost exclusively in the Middle and Late Archaic periods between 8,000 and 6,700 cal. B.P.

(Table 5). Artifacts were recovered via intensive, systematic surface collection and spatially

controlled sediment screening using 1- and 6-mm meshes. Over 80,000 artifacts were recovered from all contexts. In order of abundance, these artifacts include lithics (*n≈*68,000), bone

(*n≈*8,000), carbon (*n≈*2,000), pigment stone (*n≈*900), groundstone (*n≈*500), and ceramic sherds

*Figure 22: Planimetric and profile maps of Soro Mik'aya Patjxa. Profile elevation is exaggerated 2 x horizontal scale. Geologic strata and cultural features are generalized from composite observations in excavation profiles. Topographic model derived using kite aerial photography and structure-from-motion image processing (Wu 2011).*

181

182

*Table 5: Radiocarbon age determinations by accelerator mass spectrometry (n=17) .*

lab ID material

AA102848 wood

1

AA102854 wood

AA102843 wood provenience area unit level feature

1 61 2 10

7

1

19

55

1

1

14

16 radiocarbon B.P.

age

5891 error

49

5914

5924

35

48

AA102828 wood

AA102851 bark?

AA102859 wood

AA102858 wood

AA102834 parenchyma

AA102855 parenchyma

AA102837 parenchyma

AA102829 parenchyma

AA102835 twig

AA102842 twig

AA102827 grass stem

1

7

7

7

1

7

1

1

1

1

1

25

11

26

23

33

19

48

25

33

52

22

1

1

1

1

2

1

1

1

2

1

1

3

9

18

15

6

14

13

3

6

13

2

5940

5957

5983

5996

6002

6003

6089

6103

6148

6157

6401

49

48

47

51

48

AA102831 grass stem

AA102826 parenchyma

1

1

27

22

1

1

5

2

6458

6631

71

50

AA102838 twig 1 48 1 13 7090 59

1

2

All charred materials from lower levels of feature contexts

OxCal v4.2.3 (Bronk Ramsey 2013); r:5 IntCal13 atmospheric curve (Reimer 2013).

50

49

50

50

49

48

95% calendar B.P.

2 max median

6850 6713 min

6566

6843

6883

6733

6748

6660

6656

6891

6904

6945

6966

6972

6981

7157

7159

7170

7175

7425

7499

7580

8013

6768

6788

6822

6836

6842

6843

6959

6980

7053

7063

7338

7369

7517

7914

6725

6800

6808

6896

6903

7255

6664

6670

6679

6693

6729

7255

7435

7794

(*n*=97). Of the recovered lithics, 233 are temporally diagnostic projectile points, all of which are diagnostic of preceramic Archaic periods (see Klink and Aldenderfer 2005). Ninety four of the projectile points are assigned to one of four of the following Archaic sub-periods: Early Archaic

(*n=*3), Middle Archaic (*n=*20), Late Archaic (*n=*70), and Terminal Archaic (*n=*1). The trace amounts of ceramic sherds were all recovered from the plow zone and stylistically post-date A.D.

1000 3 , indicating a minor intrusion into an otherwise Archaic assemblage.

The excavations revealed thirteen clearly defined cultural pit features, each beginning in cultural stratum IIa and cutting down into the whitish artifact-sterile sandy loam matrix of IIc—

3 Cecilia Chavez Justo, Collasuyo Archaeological Research Institute, identified the ceramic types.

183 likely alluvially redeposited volcanic tuff. The features were readily identified as dark, organicrich sediment dissecting IIc. All of the pits contained ashy fill with flaked stone, bone, charcoal, ochre, and/or groundstone. Eight of the pit features contained between one and three human skeletons for a total of fifteen individuals. One pit contained four large informal groundstone blocks. The remaining four pits contained general fill materials only. Sediment from 19 contexts including 13 features and totaling 307 liters was processed by flotation. The recovered materials have undergone preliminary sorting and identification. Analyses are ongoing.

TERRITORIALITY AND VIOLENCE AT SMP

Population pressure and resource stress is often cited as a catalyst for emergent sedentism and food production in human societies (Rosenberg 1998). Nathan Craig (2011) has suggested that the observed population peak in the Late Archaic Period may have incited the resource stress necessary to expand diet breadth into more costly resource domains, including the wild progenitor species of Andean domesticates. Archaeological evidence for population growth offers critical support for the model, but evidence of resource stress is also required. Rosenburg

(1998:662) suggests that the most unambiguous indicator of prehistoric stress is evidence of violence.

Sixteen human burials were examined for signs of violent trauma at SMP. Nine of the individuals were sufficiently intact to permit mostly complete evaluation of the crania, long bones, and burial contexts. Three of the burials were disturbed but sufficiently intact for partial evaluation of the crania and long bones. The remaining two individuals consisted of highly fragmented crania with little potential for trauma-data recovery.

184

Burial 9, a young-adult male, shows two possible indicators of violent trauma (Figure

23). First, a large, serrated, black andesite dart point was found with the blade under the sternum

and above the right ribs. The stratigraphic relationships along with the absence of postdepositional disturbance indicate that the point was likely inside the flesh of the body near the time of death.

Second, a large hole was observed in the left parietal bone, which was oriented upward in the burial context. The margins of the hole are highly eroded and do not reveal evidence of

*Figure 23: Burial 9, a young-adult male, showing evidence of violent trauma. (a) *

*Perimortem head wound with void in the left parietal bone (left image) and corresponding green-fractured fragment (upper right image) found inside the cranial vault. (b) Projectile point impalement showing the dart and its position below and in direct contact with the dorsal surface of the sternum cortical bone and above the ribs.*

*Note also cranial deformation and deer antler above the right arm in the left image.*

185 perimortem fracture, but the fragment of bone that was dislodged from the void does reveal green-fracture margins. The fragment was found inside the cranial vault, resting on the interior surface of the right temporal bone near the foramen magnum. Because the fragment was protected from weathering, its preservation is exceptional, thus revealing the signature conchoidal-fracture edges and uniform coloration associated with green fracture. These observations suggest perimortem trauma (White and Folkens 2005:50–52).

Cranial deformation provides less direct evidence for competition. Regardless of the proximate, culturally specific reasons for engaging in cranial deformation, an unavoidable outcome is clear and honest signaling of group membership (Blom 2005:17). Such signaling would have been particularly efficacious under conditions of heightened resource competition and territoriality. Five of the crania were sufficiently intact to observe the presence or absence of

artificial deformation. All five exhibit frontal and/or rear deformation (Figure 24). Because the

cranial deformation occurs on both the anterior and posterior sides of the crania and is relatively extreme in some cases, it was likely intentional and not incidental to other behaviors such as cradle boarding (see Blom 2005:8). These observations suggest a heightened interest in the honest signaling of group membership in the Late Archaic Period.

OCCUPATIONAL REDUNDANCY AT SMP

Reduced residential mobility is requisite for agricultural production. SMP does not appear to have been a place of year-round residence, but it does appear to have been a place where hunter-gatherers repeatedly resided on a long-term but non-permanent basis. This conclusion follows from the convergence of several lines of evidence.

186

*Figure 24: Two extreme examples of culturally deformed crania at SMP. Above is *

*burial 3. Below is burial 5. See also Figure 23.*

To begin, the sheer quantity and relative of artifacts at SMP indicate high occupation intensity. Eighty-thousand artifacts consisting of five artifact classes—flaked stone, bone, carbon, groundstone, and pigment stone—were recovered from just 1.8 percent of the site. If we project this density to the rest of the 2800-m 2 site, we could expect a total of 3.1 million + 60

thousand artifacts at SMP 4 . The majority of the raw materials recovered are not available in the

4 To estimate the total artifact count for the site, we first examined the density distribution of artifacts in each excavation unit, which cover no more than 1 sq. m each. Akaike information criteria suggests that an exponential distribution with a rate of 8.93x10

-4

offers the best fit (99.9%) to the data over normal, lognormal, and power law alternatives. KS test comparison of the maximum-likelihood estimated (MLE) model parameters with the raw data suggests marginal difference (D = 0.1208, p-value = 0.10), which we consider sufficient for the purposes of estimation. The total artifact count estimates were derived using a random exponential function with 2,800 iterations (2,800 m 2 site) and the MLE estimated rate parameter for the exponential function. This was iterated 10,000 times to generate an error term. The result is a normal distribution of estimates with a mean of 3.1 million and standard deviation of 59 thousand artifacts.

187 immediate vicinity of the site and so their quantities cannot be explained in terms of local quarrying.

SMP's high artifact density in combination with the relatively small spatial extent

(roughly 60 m in diameter) suggests that artifact accumulation was distributed through time as many hours of occupation and not across space as momentary aggregations of many individuals

(see also Appendix I). The wide range of radiocarbon assays further suggests that the site's occupation was spread through time.

Ruling out site growth by contemporaneous aggregation, we are left with two possible proximate explanations for SMP's growth: continuous long-term occupation by a small residential group or high-frequency re-occupation by one or more small groups (sensu Surovell

2009:88). In other words, SMP's residents were either sedentary, occupying the site on a superannual basis, or they were residentially mobile, repeatedly occupying the site on a sub-annual basis.

Two independent lines of material evidence converge to suggest that highly mobile groups occupied SMP and occupation was therefore discontinuous. First, the cultural deposits

occur in a single, thin stratigraphic unit (see Figure 22) suggesting a lack of cultural midden

accumulation that might be expected in the case of permanent occupation. Second, typical material correlates of permanent occupation, including permanent housing and ceramics, are absent. The lack of house foundations may reflect preservation bias due to erosion or nearsurface (~30 cm) plow disturbance. However, the fact that the burial pits evaded disturbance

188 suggests that if substantial pit-house structures were present, we likely would have detected portions of them.

The absence of permanent houses contributes to a recurrent theme in the Titicaca Basin.

Late Archaic surface architecture is unknown from the three major settlement surveys that documented hundreds of Late Archaic sites (Cipolla 2005; Craig 2011; Klink 2005). Excavations at the Late Archaic site of Pirco, which is larger than SMP in terms of areal extent and artifact quantity, did not locate permanent house features despite the absence of mechanical plow disturbance (Craig 2011). Limited test excavations in two Late Archaic sites in the Rio Ramis

Valley also did not encounter permanent house features (Aldenderfer and Flores Blanco 2008).

In sum, SMP reveals a material signature that is most consistent with a high degree of residential mobility—certainly higher than that which is inferred for the subsequent Terminal

Archaic and Early Formative periods when pit houses, masonry architecture, and ceramic technologies were in use. Taking this high degree of mobility in conjunction with evidence for a millenium or more of site use, we conclude that SMP represents many re-occupations by small hunter-gatherer groups. Accordingly, while SMP inhabitants were relatively mobile, they had developed a long-term connection with a highly localized place.

INTENSIVE RESOURCE USE AT SMP

The path to food production first entails intensive use of wild progenitor species. Genetic analysis suggests that potatoes may have been domesticated from wild species in the southcentral Andes, which includes the Titicaca Basin (Spooner et al. 2005). Quinoa (*Chenopodium *

*quinoa*) and kiwicha (*Amaranthus caudatus*) are believed to have Andean origins, possibly in the

189

Lake Titicaca region (Pickersgill 2007:5). If hunter-gatherers domesticated these species in the

Lake Titicaca Basin, we should expect to find material correlates of intensive use of the wild progenitor species prior to domestication. Macrobotanicals offer the clearest intensive plant use; unfortunately, clear macrobotanical nor faunal evidence of food resources was not forthcoming

. Nonetheless, a rich groundstone assemblage and extreme dental attrition suggest that

SMP's inhabitants were intensively harvesting, processing, and consuming wild grains—quite possibly wild *Chenopodium sp*. or *Amaranthus sp*.

Groundstone artifacts comprise a significant portion of the SMP artifact assemblage

(Figure 25). The 532 groundstone artifacts include 91 specimens from the surface and 439 from

subsurface contexts. Twenty-four pieces were recovered from 6 of the 13 secure feature contexts.

Lacking associated macrobotanical remains and pending starch grain analysis (Rumold 2010), direct evidence for the function of the groundstone is currently unavailable. Nonetheless, material, formal, and relational data (Schiffer and Skibo 1997) for the groundstone gives us a sense of the range of possible functions and intensity of use (Buonasera 2013).

Two-hundred and seventy four of the groundstone artifacts were examined for raw material and morphological variation. The planimetric shapes of all of these tools are irregular indicating informal design. However, tabular cross-sectional forms in 75 of the 100 sufficiently intact artifacts suggest intensive use. Two major groups of groundstone artifacts are distinguishable by material. The first group is made from metasedimentary materials ranging

5 BrieAnna Langlie and Dr. Gayle Fritz, paleoethnobotanists at the Washington University in St. Louis, examined a sample of cheno-am seeds and charcoal recovered by flotation from feature contexts. They determined the cheno-am seeds to be unburned and therefore modern, likely having intruded into the features via root and insect disturbances. Some of the charcoal fragments are identified as parenchyma tissue, a type of storage tissue in plants, but little more can be said about their identification without further intensive analysis.

190 dart point fragment

*Figure 25: Groundstone and its context at SMP. Upper left: pit feature with large groundstone boulders (feature 6, max length=1.3 m). Upper right: the most formal example of groundstone at SMP—a volcanic mano/hammerstone with ground surfaces and battering (feature 2). Lower left: a typical sandstone grinding slab placed on the neck of burial 10 in feature 14. Lower right: closeup of the burial 10 sandstone artifact, a typical example of SMP groundstone. Note also black andesite dart points placed on each shoulder of burial 10.*

from friable sandstone to well-cemented quartzite. This group comprises 67 percent (*n*=183) of the examined groundstone artifacts. The grinding surfaces, especially on the sandier, lesscemented materials, tend to be perfectly flat due to intense use, though such wear was likely rapid due to the friable nature of the material. Informal, tabular groundstone made from friable materials tends to be associated with relatively rapid grinding of small seeds (Buonasera 2013).

The second group of groundstone is made from fine-grained volcanic materials.

Compared to the sandstone tools, the volcanic-tool forms show a higher degree of irregularity

191 that is often indistinguishable from locally available alluvial cobbles. The volcanic

mano/hammerstone shown in Figure 25 represents the most formal of the volcanic groundstone

artifacts, yet the natural cobble form remains evident and attests to the overall informal nature of the assemblage.

It also appears that SMP residents invested in the transport and storage of groundstone with the intent to return and re-use the materials. The largest recovered specimen at 3 kg represents a fairly substantial transport investment given that the nearest exposed bedrock and potential quarry is 1.5 km from the site. Moreover, the four largest of the boulders, with a

cumulative mass of 5 kg, were found in a storage pit suggesting intent of future use (see Figure

25). The quantity, ubiquity, and contexts of the artifacts indicate that groundstone was not

incidental to the site but an integral part of the SMP tool assemblage.

Extreme dental attrition in the associated burials offers additional evidence for the imporatnce of ground foods in the diet. Tooth wear was examined in 14 of the 16 recovered

human burials (Figure 26). We focus on molars here because they exhibit the greatest variation in

wear and their known eruption times provide an approximation of adult age, which together with other lines of age evidence can be used to approximate wear rate. In addition to using molar eruption and molar wear for age estimation, cranial suture closures were also used (White and

Folkens 2005:363–371).

192

The dental data in Table 6 show that heavy tooth attrition began at a young age in the

SMP population. Even as adults less than 35 years in age, tooth attrition extended well into the dentin and occasionally into the roots of the first molar. As a mid-adult between the ages of 25-

50, at least one tooth could be expected to be worn to the roots, and by old adulthood (50+) it was likely that most teeth were worn well into the dentin and in many cases into the roots. The high dental-wear rates are likely associated with a high-grit diet, and the abundant groundstone—

*Figure 26: Various degrees of dental attrition shown in four different mandibulae from SMP. Upper left: young adult with approximately 20-percent dentin exposure in first molar (burial 7, feature 10). Upper right: young adult with 100-percent dentin exposure in first molar and approximately 30-percent in second molar (burial 9, feature 13). Lower left: mid adult with 100-percent dentin exposure in the first molar with attrition nearing the root, 80-percent dentin exposure in the second molar, and *

*10-percent dentin exposure in the third molar (burial 11, feature 18). Lower right: mid or old adult showing first and second molars ground to the roots and 100-percent dentin exposed in third molar (burial 16, feature 18).*

193

*Table 6: Dental attrition summary.*

11

12

13 f

8

9

10

5

6

3

4

7

1 d

2 d

**burial feature max. wear a**

1

2 no data no data

**composite molar dentin exposure estimate (%) b**

**M1**

no data no data

**M2**

no data no data

**M3**

no data no data

**age estimate**

no data no data

**c**

3

4

2

2

10

1

1

2

2

1

50

5

100

100

70

1

0 e

100

100

1

0 un-erupted ind.

0

0 young adult child/adolescent

> mid adult

> mid adult young adult

13

13

14

18

4

16

2

2

1

2

2

0

100

100

80

100

100

1

80

50

1

50

100

1

100

1

1

10

50

1 old adult young adult young adult mid adult

> mid adult child

14

15

4

16

2

2 ind.

ind.

ind.

ind.

ind.

ind.

> young adult

> young adult

16 18 2 100 100 100 > mid adult

**a**

A qualitative ranking of the maximum observed depth of wear for all teeth, where 0=little or no wear, 1=wear into b the dentin, and 2=wear into the roots.

An average of upper, lower, right, and left molars for the given position, M1, M2, or M3. Averages are based on qualitative estimates to the nearest 10%, and a value of 1 is used to indicate observable wear that does not expose

**c**

dentin.

Age classes follow Buikstra and Ubelaker 1994 as follows: child: 3-12 years, adolescent: 12-20 years, young adult:

20-35 years; middle adult: 35-50 years, and old adult: 50+ years. Determinations based on cranial suture closure, d dental eruptions, and relative dental attrition.

e

Highly fragmented cranium in plow zone—not examined.

Roots partially formed, degree of eruption uncertain.

f

Five fully formed deciduous molars were observed, all with light wear. Positions were indeterminate, but the prevalence provides the basis for assuming values of 1 and indicates consumption of stone-ground foods.

especially the sandstone—is a likely source of the grit. Taken together, the abundant groundstone and extreme tooth wear suggest that SMP inhabitants were grinding and consuming wild grains.

With few productive alternatives in a region over 3,800 m in elevation, the grain was possibly the seed of wild *Chenopodium* *sp*. or *Amaranthus sp.*

194

SUMMARY AND CONCLUSION

The ancient Titicaca Basin provides an important case in the study of emergent social complexity in human societies because it is one of few regions where hunter-gatherer behaviors initiated a path to endogenous state formation. Our current understanding of the trajectory remains weighted toward the later Formative and Tiwanku periods. The recent observations at the site of Soro Mik'aya Patjxa contribute to filling a critical gap at the earlier end of the process.

In particular, the field and laboratory results presented here have led to the following inferences for Late Archaic Period behavior at SMP between 8,000 and 6,700 cal. B.P.:

1. Conspecific competition and inter-personal conflict,

2. High residential mobility and high-frequency site re-occupation, and

3. Intensive processing and consumption of wild plant resources—possibly wild

*Chenopodium sp*. or *Amaranthus sp.*

More archaeological observations are needed to satisfactorily test these hypotheses. For example, future analyses of the SMP data will explore hypotheses for lithic-assemblage structure. Additional subsurface investigations at other secure Late Archaic contexts are also sorely needed to provide independent tests. For now, the results presented here suggest that population growth, resource stress, and widening diet breadth characterized hunter-gatherer behavior on the eve of complexity in the Lake Titicaca Basin.

195

REFERENCES CITED

Aldenderfer, Mark S.

1989a Archaic Period “Complementarity” in the Osmore Drainage. In *Ecology, Settlement, *

*and History in the Osmore Drainage, Peru*, edited by Don S. Rice, Charles Stanish, and

Phillip R. Scarr, pp. 101–128. BAR International Series 545. B.A.R., Oxford.

1989b Archaic Period in the South-Central Andes. *Journal of World Prehistory* 3(2): 117–

158.

2005 Preludes to Power in the Highland Late Preceramic Period. *Archaeological Papers of *

*the American Anthropological Association* 14: 13–35.

Aldenderfer, Mark S., Nathan M. Craig, Robert J. Speakman, and Rachel Popelka-Filcoff

2008 Four-Thousand-Year-Old Gold Artifacts from the Lake Titicaca Basin, Southern Peru.

*Proceedings of the National Academy of Sciences* 105(13): 5002–5005.

Aldenderfer, Mark S., and Luis A. Flores Blanco

2011 Reflexiones para Avanzar en los Estudios del Periodo Arcaico en los Andes Centro-Sur.

*Chungara: Revista de Antropología Chilena* 43(1): 531–550.

2008 *Informe Final: Excavaciones en Sitios Arcaico-Formativo de la Cuenca del Ramis, *

*Puno. *Unpublished Presentado al Director Nacional del Instituto Nacional de Cultura,

Lima, Perú.

Bandy, Matthew S.

2006 Early Village Society in the Formative Period in the Southern Lake Titicaca Basin.

*Andean archaeology* 3: 210–236.

Blom, Deborah E.

2005 Embodying borders: human body modification and diversity in Tiwanaku society.

*Journal of Anthropological Archaeology* 24(1): 1–24.

Bronk Ramsey, Christopher

2013 *OxCal*. Research Lab for Archaeology, Oxford Radiocarbon Accelerator Unit, Oxford.

Bruno, Maria C.

2006 A Morphological Approach to Documenting the Domestication of Chenopodium in the

Andes. In *Documenting domestication new genetic and archaeological paradigms*, edited by Melinda A Zeder, Daniel G Bradley, Eve Emshwiller, and Bruce D Smith, pp. 32–45.

University of California Press, Berkeley, Calif.

Buonasera, Tammy Y.

2013 More than acorns and small seeds: A diachronic analysis of mortuary associated ground stone from the south San Francisco Bay area. *Journal of Anthropological Archaeology*

32(2): 190–211.

Cipolla, Lisa M.

2005 Preceramic Period Settlement Patterns in the Huancané-Putina River Valley, Northern

Titicaca Basin, Peru. In *Advances in Titicaca Basin Archaeology-1*, edited by Charles

Stanish, Amanda B. Cohen, and Mark S. Aldenderfer, pp. 55–64. Cotsen Institute of

Archaeology at UCLA, Los Angeles.

Craig, Nathan

196

2011 Cultural Dynamics, Climate, and Landscape in the South-Central Andes During the

Mid-Late Holocene: A Consideration of Two Socio-Natural Perspectives. *Chungará* 43:

367–391.

Hastorf, Christine A

2008 The Formative Period in the Titicaca Basin. In *Handbook of South American *

*Achaeology*, edited by Helaine Silverman and William H. Isbell, pp. 545–561. Springer,

New York NY.

Janusek, John Wayne

2004 Tiwanaku and Its Precursors Recent Research and Emerging Perspectives. *Journal of *

*Archaeological Research* 12(2): 121–183.

2008 *Ancient Tiwanaku*. Cambridge University Press, Cambridge; New York.

Klink, Cynthia J.

2005 Archaic Period Research in the Río Huenque Valley, Peru. In *Advances in Titicaca *

*Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 13–24. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Klink, Cynthia J., and Mark S. Aldenderfer

2005 A Projectile Point Chronology for the South-Central Andean Highlands. In *Advances in *

*Titicaca Basin Archaeology-1*, edited by Charles Stanish, Amanda B. Cohen, and Mark S.

Aldenderfer, pp. 25–54. Cotsen Institute of Archaeology at UCLA, Los Angeles.

Kolata, Alan L.

1993 *The Tiwanaku: Portrait of an Andean Civilization*. Blackwell, Cambridge, MA.

Moseley, Michael

1992 *The Incas and their Ancestors: The Archaeology of Peru*. Revised. Thames and Hudson,

New York N.Y.

Pickersgill, Barbara

2007 Domestication of Plants in the Americas: Insights from Mendelian and Molecular

Genetics. *Annals of Botany*: 1–16.

Reimer, Paula

2013 IntCal13 and Marine13 Radiocarbon Age Calibration Curves 0–50,000 Years cal BP.

*Radiocarbon* 55(4): 1869–1887.

Rigsby, Catherine A., Paul A. Baker, and Mark S. Aldenderfer

2003 Fluvial History of the Rio Ilave Valley, Peru, and Its Relationship to Climate and

Human History. *Palaeogeography, Palaeoclimatology, Palaeoecology* 194(1-3): 165–185.

Rosenberg, Michael

1998 Cheating at Musical Chairs: Territoriality and Sedentism in an Evolutionary Context.

*Current Anthropology* 39(5): 653–681.

Rowe, John Howland, and Dorothy Menzel

1967 *Peruvian Archaeology: Selected Readings*. Peek Publications, Palo Alto, Calif.

Rumold, Claudia U.

2010 Illuminating Women’s Work and the Advent of Plant Cultivation in the Highland

Titicaca Basin of South America: New Evidence from Grinding Tool and Starch Grain

Analyses. Unpublished Ph.D. dissertation, University of California, Santa Barbara.

197

Schiffer, Michael B., and James M. Skibo

1997 The Explanation of Artifact Variability. *American Antiquity* 62(1): 27–50.

Spooner, David M., Karen McLean, Gavin Ramsay, Robbie Waugh, and Glenn J. Bryan

2005 A single domestication for potato based on multilocus amplified fragment length polymorphism genotyping. *Proceedings of the National Academy of Sciences of the United*

*States of America* 102(41): 14694–14699.

Stanish, Charles

2001 The Origin of State Societies in South America. *Annual Review of Anthropology* 30(1):

41–64.

2003 *Ancient Titicaca: The Evolution of Complex Society in Southern Peru and Northern *

*Bolivia*. University of California Press, Berkeley.

Steadman, Lee Hyde

1995 Excavations at Camata: An early ceramic chronology for the Western Titicaca Basin,

Peru. Unpublished Ph.D., University of California, Berkeley, United States -- California.

Surovell, Todd A.

2009 *Toward a Behavioral Ecology of Lithic Technology: Cases from Paleoindian *

*Archaeology*. University of Arizona Press, October 1.

White, T. D, and Pieter A Folkens

2005 *The human bone manual*. Elsevier Academic, Amsterdam; Boston.

Wu, Changchang

2011 *VisualSFM: A Visual Structure from Motion System*.

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project