This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Journal of Anthropological Archaeology 29 (2010) 62–79

Contents lists available at ScienceDirect

Journal of Anthropological Archaeology

journal homepage: www.elsevier.com/locate/jaa

An evolutionary model of social change in the Middle Ohio Valley: Was social
complexity impossible during the late woodland but mandatory during
the late prehistoric?
Kevin C. Nolan *, Robert A. Cook
Department of Anthropology, The Ohio State University, 4034 Smith Laboratory, 174 W. 18[th] Avenue, Columbus, OH 43210-1106, United States

a r t i c l e i n f o a b s t r a c t

Article history: We present an evolutionary model of social change in the Middle Ohio Valley during the Fort Ancient per-
Received 1 May 2009 iod (AD 1000–1650), primarily relying on an application of Winterhalder’s (1986) and Kelly’s (1995) evo-
Revision received 19 October 2009 lutionary ecology model of cultural responses to environmental variability. We predict changes in social
Available online 25 November 2009
organization, political complexity, and patterns of stylistic similarity. As the environmental background
for our model we employ recently published moisture data. By iteratively applying the synchronic Win-
Keywords: terhalder–Kelly model we are able to predict a sequence of general trends that agree with much current
Evolutionary ecology
interpretation of Fort Ancient development. Specifically, the model leads us to expect increasing village
Social evolution
Environmental variability
size, the development of regional traditions, and the development of leadership roles and a region-wide
Fort Ancient style Horizon around AD 1400. However, our model provides the ability to make testable predictions that
Late Prehistoric period deviate from the accepted cultural history and predicts that the Middle Ohio Valley and the groups
Madisonville Horizon referred to as Fort Ancient were far from homogenous.
Middle Mississippian Ó 2009 Elsevier Inc. All rights reserved.
Social complexity
Middle Ohio Valley
PDSI

Introduction
We present an evolutionary1 model for the development of the
Middle Ohio Valley village agricultural societies collectively referred
to as Fort Ancient (ca. AD 1000–1650) (Carskadden and Morton,
1977; Essenpreis, 1982; Graybill, 1981; Griffin, 1966; Henderson,
1992; Mills, 1906; Prufer and Shane, 1970). Our focus is on explain-
ing increases in social complexity, variability in inter-community
interaction, and changing patterns of material culture similarity
leading up to the Madisonville Horizon (ca. AD 1400/1450), a re-
gion-wide increase in trade and interaction represented by homoge-
nization of style and increasing Mississippian influence (Henderson,
1992).
Ecological and evolutionary theory is absent from most previ-
ous analyses of Fort Ancient social change. Most explanations have
not specifically taken environmental constraints into account in
their models (e.g., Cook, 2008; Pollack and Henderson, 1992; Prufer
* Corresponding author. Fax: +1 614 292 4155.
E-mail address: nolan.117@osu.edu (K.C. Nolan).
1
Throughout this paper we use the term evolutionary to refer to the broad array of
approaches that are derived from Darwinian Theory. Under this umbrella we include
Evolutionary Ecology, Evolutionary Archaeology, Gene-culture Coevolution, among
others. Our presentation relies on each of these to varying degrees.
and Shane, 1970). While climate and other environmental param-
eters have factored into a few analyses (Graybill, 1981; Greenlee,
2002; Kennedy, 2000) it has not often been approached from the
integrated framework offered by evolutionary ecology (EE). In con-
structing our model we draw primarily upon models built under
the umbrella of EE (e.g., Dyson-Hudson and Smith, 1978; Kelly,
1995; Winterhalder, 1986, 1990; Winterhalder and Goland,
1997), but also incorporate insights from evolutionary archaeology
(a.k.a., selectionist archaeology) (e.g., Dunnell, 1996a, 1999; Dun-
nell and Greenlee, 1999; Madsen et al., 1999; Lipo et al., 1997; Rin-
dos, 1980, 1984) and Gene-culture Coevolution (e.g., Boyd and
Richerson, 1985; Soltis et al., 1995; Richerson and Boyd, 2005).
We suggest that by using EE to analyze Middle Ohio Valley Fort An-
cient development we can integrate many of the previous models
of Fort Ancient social change and provide specific, testable predic-
tions that can account for not only the central tendencies (norms),
but also the spatial and temporal variability that tend to be under-
emphasized in previous models.
The study of behavioral variability is crucial to an evolutionary
approach to culture, and patterns of environmental variability can
significantly affect cultural responses. In order to model the social
response of the prehistoric Middle Ohio Valley village agricultural-
ists, we must first reconstruct the environmental context. To be
successful, we need high temporal and spatial resolution. Climate

0278-4165/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jaa.2009.10.004
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
proxy data for the Middle Ohio River Valley is limited. The avail-
able data often have coarse temporal resolution (150 years or
greater Shane et al., 2001). However, recently a source of high-res-
olution climate proxy data has become available. This resolution is
provided by the Palmer Drought Severity Index reconstructions of
Cook et al. (1999, 2004). We use these data as the environmental
background for our model. By aggregating these data into 50-year
periods we can model not only trends in central tendency (mean),
but also trends in environmental variability (spatially and tempo-
rally, i.e., inter-group correlation and standard deviation).
With this environmental background, we apply Winterhalder’s
(1986) model as modified by Kelly (1995) to predict changes in
the prehistoric record. Winterhalder’s (1986) model uses temporal
variation (standard deviation) and spatial variation (inter-group
correlation) in resource abundance. By mapping both means and
standard deviations we can visualize the two axes of Winterhal-
der’s model for the entire Ohio Valley.
Previous models of Fort Ancient social change
The term ‘‘Fort Ancient” was coined by Mills (1906, p. 135) to
connote the similarity of material culture among sites in southern
Ohio (Baum, Gartner, and Fort Ancient specifically), particularly in
contrast with Hopewell artifacts from the same region. Fort An-
cient currently also serves to connote contrast with Middle Missis-
sippian societies. Middle Mississippian refers to those groups that
are often interpreted as ‘‘chiefdoms,” occupying the Middle and
Lower portions of the Ohio Valley (initially defined by Holmes
(1886), and refined for the Fort Ancient region by Griffin (1966)).
Fort Ancient has come to be used in reference to village-based, tri-
bal, maize agriculturalists of the Middle Ohio Valley during the
Late Prehistoric period (Carskadden and Morton, 1977; Essenpreis,
1982; Graybill, 1981; Griffin, 1966; Henderson, 1992; Prufer and
Shane, 1970).
Most previous models of social change among Fort Ancient soci-
eties in the Middle Ohio Valley (ca. AD 1000–1650) can be grouped
into three general categories: (1) external (Griffin, 1966; Prufer and
Shane, 1970); (2) internal (Pollack and Henderson, 1992; Rafferty,
1974); and (3) internal/external (Cook, 2008; Cowan, 1987;
Essenpreis, 1978; Robertson, 1980). Each type of model calls on
different types of causal agents and they were generally formu-
lated within different paradigms ranging from Culture History to
Processualism to recent approaches reconsidering migration and
other historical dimensions alongside general evolutionary models.
External models attributed Fort Ancient social composition to
migration and diffusion, explanations which were commonplace
in the Culture Historical period (Trigger, 2006). Initial understand-
ing of Fort Ancient origins relied primarily on migration (Griffin,
1966), a trend that continued until the end of the Culture Historic
period (Prufer and Shane, 1970). Griffin (1966) argued that Fort
Ancient began as an adaptation originating in the lower Ohio and
Middle Mississippi River valleys that was adapted to a less agricul-
turally productive environment (i.e., the Middle Ohio Valley). Pru-
fer and Shane (1970) furthered this argument by asserting that Fort
Ancient material culture was distinct from Late Woodland groups
who were argued to have occupied a different part of the local
landscape, mainly uplands and rockshelters as opposed to flood-
plains. With limited material for comparison, it was easy to see
major distinctions between Late Woodland and Fort Ancient mate-
rial assemblages (see Essenpreis (1978)). There are some marked
differences between Late Woodland and Late Prehistoric assem-
blages, which are still poorly understand, but all current models
now recognize various degrees of local contributions.
Internal models of Fort Ancient development were common
among Processualists and marked a fundamental shift in interpret-
63
ing culture change. Instead of invoking diffusion and migration the
emphasis was on fit with the local environment, continuity with
preceding groups, and the functional dynamics of settlement sys-
tems (Trigger, 2006). Increasing knowledge about the Late Wood-
land period also helped refine explanatory accounts. The fullest
articulation of the internal-only model of Fort Ancient develop-
ment was provided by Pollack and Henderson (1992), but the mod-
el had been developing for some time (e.g., Essenpreis, 1978;
Graybill, 1981; Rafferty, 1974). Pollack and Henderson (1992) pos-
ited that Fort Ancient was an entirely in situ development repre-
senting a gradual increase in social complexity (based partially
on a model outlined by Johnson and Earle (1987)). Though their
stated aim was to construct a model of social development for
northern Kentucky, this has come to be the most widely cited se-
quence for the whole Fort Ancient territory and their model incor-
porates Ohio Fort Ancient data. Kennedy’s (2000) internal
examination of the general contraction of the entire Fort Ancient
area over time concluded that with the onset of the Little Ice Age
(LIA) and the attendant reduced agricultural productivity, with
populations aggregating in areas with the most productive soils
and access to reliable trade routes along the Middle Ohio River.
Internal/external models have been applied to Fort Ancient
intermittently from the late 1970s and early 1980s (Essenpreis,
1978; Graybill, 1981). These approaches are generally non-typo-
logical and are rooted in what could be considered a more mature
form of Processualism, that consider cultures as systems that do
not develop in isolation (e.g., Crumley, 1979). The most recent
internal/external model (Cook, 2008), builds on these studies by
further taking into account that general processes can couple with
approaches that examine what elements are unique to a society
and the significance of migration in more current sociocultural
models (see Cobb (2005) and Pauketat (2003)).
The initial internal/external model of Fort Ancient social change
was developed by Essenpreis (1978), who argued that Fort Ancient
represented different responses to internal and external stimuli,
with some groups adhering more closely to their Late Woodland
forebears while others shifted their settlement system toward that
of Middle Mississippians with whom they more intensively inter-
acted. The latter was represented by the Madisonville Phase, for
which a key piece of evidence was the platform mounds at the
Marietta site (Essenpreis, 1978). When one of these mounds was
shown to be Middle Woodland in age (Pickard, 1996), many
researchers abandoned this model; however, there is at least one
case of a Fort Ancient temple-style mound (Baum; Mills, 1906). A
combination of Late Woodland and Middle Mississippian ‘‘influ-
ences” on Fort Ancient development has been discussed by various
other researchers as well (e.g., Brose, 1982; Cowan, 1987; Robert-
son, 1980).
Working in the eastern portion of the Fort Ancient region, Gray-
bill (1981) utilized an internal/external model of social change
when he argued that larger settlement sizes and the appearance
of palisades in Late Fort Ancient sites on the eastern periphery of
the Fort Ancient region was in part caused by the effects of the
LIA on northern Iroquoian farmers. Iroquoian populations experi-
encing decreased agricultural productivity associated with a dete-
riorating climate were argued to have initiated raids on their
southern neighbors triggering changes in settlement organization
and distribution, such as consolidation into larger villages sur-
rounded by palisades (Graybill, 1981, p. 168).
In the most recent example of an internal/external Fort Ancient
development model, Cook (2008) focuses on interactions among
people across culture–historical taxonomic boundaries as influenc-
ing the development of social complexity in the Middle Ohio Val-
ley, specifically in the Miami Valleys of southwestern Ohio in
relation to neighboring Middle Mississippian groups. The internal
part of the model looked at the growth of populations and the need
 Author's personal copy
64 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
for higher forms of management to emerge. To this end, it was only
when the SunWatch site doubled in size that a clear authority po-
sition emerged in the village (Cook, 2008). Cook (2008) links this
internal development with a broader consideration of Middle Mis-
sissippian forms of complexity with some being comparable on a
scale previously not considered. Fort Ancient village to Middle Mis-
sissippian village interaction is set within a ‘‘periphery peer”
framework, which adheres well to expectations derived from an
expanded version of Renfrew’s (1986) peer polity interaction mod-
el with the incorporation of subsequent research (see Cook (2008,
p. 3–6)). Cook shows that the structure of Fort Ancient villages is
very similar to small Middle Mississippian villages and forms asso-
ciated with ‘‘emergent Mississippians” in the American Bottom
(Kelly, 1990). It is these types of Mississippian sites that Cook
(2008) argues evolved during peer interactions.
Extant evidence for regional diversification and integration
Two developments across the Middle Ohio Valley and directly
preceding the Fort Ancient period proper occurred during the late
Late Woodland period (AD 800–1000). During this time maize in-
creased in dietary importance (e.g., Greenlee, 2002, Fig. 24; Hart,
1999, Fig. 4) and the large fortified villages of the early Late Wood-
land (AD 400–800) largely disappeared to be replaced by small,
dispersed settlements; unfortunately, domestic architecture has
rarely been discernable (Church, 1987; Church and Nass, 2002;
Pollack and Henderson, 2000b; Seeman and Dancey, 2000, p.
597; Redmond and McCullough, 2000). Several researchers have
indicated that some early Late Woodland sites were circular vil-
lages with central plazas (Cowan, 1987; Pollack and Henderson,
2000b; Rafferty, 1974). There are examples of large villages in
the late Late Woodland (these appear to be concentrated in south-
western Ohio; e.g., Turpin; see Cowan (1986) and Church and Nass
(2002)), but there are no clear cases of plazas. During the late Late
Woodland pottery is mostly undecorated.
We present subregional/drainage summaries of extant data and
interpretation relevant for evaluating our model, namely village
structure, clear indicators of conflict (palisades), evidence of status
differentiation, stylistic distinctions (e.g., pottery design), and
interactions with ‘‘outsiders” (Middle Mississippian- and Iro-
quoian-style house forms and trade items). There are two issues
greatly hampering this exercise. First, site structures are poorly
understood (with a few important exceptions) as few sites have
been extensively excavated. Second, and equally vexing, is that
most sites have very few radiocarbon dates leaving us with a very
poor control over chronology (e.g., most of the Fort Ancient sites in
southwest Ohio have only a few radiocarbon dates), and the ones
that are well dated often span the entire Fort Ancient temporal
spectrum (e.g., Cook, 2007; Drooker, 1997). The following consid-
ers Fort Ancient in commonly agreed upon time periods: Early
(AD 1000/1050–1200/1250), Middle (AD 1200/1250–1400/1450),
and Late (AD 1400/1450–1650/1670), though we recognize the
need for better resolution for fully realizing the model we put for-
ward later (Table 1; see also summaries in Carskadden and Morton,
2000, Fig. 5:3; Drooker, 1997, Fig. 4-3).
In southwestern Ohio, focused on the Little and Great Miami
Rivers, well-structured circular villages with plazas are present
from the beginning of the Fort Ancient period (Cook, 2008; Cowan,
1986, 1987; Drooker, 1997; Essenpreis, 1982), though the timing of
these may be closer to AD 1100 than AD 1000 (Drooker, 1997; see
also Church and Nass (2002)). There is significant variation in vil-
lage sizes. At 3.2 ha the State Line site is more than double that
of the ‘‘typical” village sizes (e.g., SunWatch at 1.4 ha) and four
times the size of smaller villages (e.g., Horseshoe Johnson at
0.4 ha) (see Cook and Schurr (2009)). The larger sites (and State
Line in particular) contain more Mississippian-style houses (wall
trenches) and trade items (pottery) than the smaller ones, leading
Cook (2008) and Cook and Fargher (2008) to conclude that the
State Line site may have served as an entrepot through which Mis-
sissippians moved. Protective palisades have also been noted for
some sites in southwestern Ohio. During the Middle Period, pot-
tery is typified by Anderson wares, which are often decorated with
a variety of guilloche and line-filled triangle motifs (see Fig. 1).
Cowan (1987, p. 12) aptly describes the Middle period in this re-
gion as being ‘‘baroque” in comparison with the initial pottery
styles.
In northern Kentucky, focused on the Kentucky and Licking Riv-
ers, the Early period is characterized as a landscape dotted with
small hamlets (Turnbow and Sharp, 1988) – similar in this sense
with late Late Woodland sites – that aggregated into (sometimes
circular) villages during the Middle period (Pollack and Henderson,
1992; see also Henderson (1998)). There are, however, few recog-
nized early sites. There is no evidence of palisades. There is limited
evidence for external trade during the Early and Middle periods
(Pollack and Henderson, 1992, 2000a). Pottery styles include
Anderson, Fox Farm, and Jessamine types (Pollack and Henderson,
2000a; Turnbow and Henderson, 1992).
Central Ohio, focused on the Scioto and Hocking Rivers, includes
several well-known villages (Blain, Baum, Feurt, Gartner, and
Voss). The Late Woodland-Fort Ancient transition, AD 1000–1200
has been interpreted as reaction to the increased risk associated
with maize agriculture and an increasing need for integrating
mechanisms (Church, 1987; Church and Nass, 2002). During this
period several typical Fort Ancient characteristics emerge: struc-
tured and nucleated settlements; shell-tempered pottery; deco-
rated ceramics. There is limited evidence of Middle Mississippian
trade items. Evidence of inter-community interaction has been lar-
gely derived from decorated ceramics. The Scioto Valley was the
heart of Griffin’s (1966) Baum phase; however, there are marked
stylistic similarities between Griffin’s Baum and Anderson ceram-
ics (Brady-Rawlins, 2007; Graybill, 1981, Fig. 1; Prufer and Shane,
1970, Fig. 1). Prufer and Shane (1970, p. 259) characterized the
Baum phase as the most ‘‘Mississippified” (shell temper and stylis-
tic elaboration) of all Early period phases, but the dating of these
characteristics is later than originally thought (see also Scham-
bach’s (1971) criticism of their use of chronology). Thus the in-
creased Mississippification of the Scioto Valley appears around
the time of increasing elaborative cultural material in other valleys
(Church and Nass, 2002).
Site structure is minimally known for Scioto Valley Fort Ancient
sites. Some Early period sites (e.g., Howard Baum) exhibit a linear
pattern similar to the dispersed household clusters of the late Late
Woodland (Church and Nass’s (2002) ‘‘Transitional Period”). Some
later (Middle/Late period) sites are known to exhibit the typical
Fort Ancient ‘‘doughnut” shaped structure (e.g., Blain (Prufer and
Shane, 1970), Voss (Brady-Rawlins, 2007)). The pottery styles of
the Early and Middle Period, especially from the northern sites,
are characterized as Baum with stylistic affinity to the Anderson
wares of the Miami and Licking drainages. Later and more south-
erly sites are often characterized by Feurt wares. Hocking Valley
Fort Ancient pottery exhibits little variation in style (see discussion
in Prufer and Shane (1970) of Baldwin) and maintains greater sim-
ilarity to the late Late Woodland the Fort Ancient ceramics are
characterized by Baum and Baldwin types.
In Eastern Ohio/West Virginia, focused on the Monongahela
River, circular villages were also present from the beginning of
the sequence (ca. AD 1050) (at least along the Ohio River). These
villages occur in a similar range of sizes as other regions, but they
are more variably structured than in other regions (Graybill, 1981).
In the upper Muskingum Valley the period AD 1275–AD 1350 is
characterized as a ‘‘Late Woodland-Fort Ancient transitional stage”
 Author's personal copy

K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79 65

Table 1
Summary of model and extant regional database.

Time Model summary Drainage summaries

period
Spatial Temporal Moisture Response Miami (eI) Kentucky (cK) Scioto (cO) Monongahela Cultural
(wP) period
801 Low/ Moderate Low Small site, residentially mobile, little Some large Some large Mostly small Mostly small Late
moderate exchange sites sites dispersed sites dispersed woodland
sites
851 Low/ Low/ Low Continuation of previous some Increased Increased Increased Increased
moderate moderate exchange south of Ohio River maize maize maize maize
901 Low/ Moderate/ Low/ More rigid boundaries: conflict north, Little Little Little decorated Little
moderate high moderate social control south; increased decorated decorated pottery decorated
residential stability and site size pottery pottery pottery
951 Low/ Low/ Low/ More style and exchange, increased
moderate moderate moderate storage and community pooling
(esp. cK, eI)
1001 Low and Moderate Moderate Conflict (wI, wK), exchange (eI, cK, wP, Circular Small hamlets Linear sites; Circular Early FA
high wV); bifurcated Fort Ancient region villages; large some villages
and small sites structured
1051 Moderate Moderate/ Moderate increased exchange, integration of Fort Mississippian Little evidence Little evidence Little
high Ancient region, conflict in high traits of external of external evidence of
temporal variability regions trade trade external
trade
1101 Low and Moderate Moderate/ cK, eI, wK, wI favor exchange; cO and Little Little Little decorated Little
high low eK some conflict and limited exchange decorated decorated pottery; Baum/ decorated
pottery pottery Baldwin pottery
1151 Moderate/ Moderate/ Moderate Lower valley increased complexity, Palisades (?)
high high increased exchange Mid-upper valley
1201 Low and Moderate/ Moderate Formalized east–west exchange Circular Circular Circular Circular Middle FA
high high villages; large villages; villages villages
and small sites
1251 Variable Moderate/ Moderate Lower valley conflict, Middle valley Anderson wares Anderson, Fox Anderson, Philo, Feurt
high focus on exchange, decrease conflict cO Farm, Baum, Feurt wares
and eK Jessamine wares
wares
1301 Low Moderate/ Moderate Increased conflict and/or larger Mississippian
high exchange networks and increased traits
complexity
1351 Low and Moderate Moderate/ Decreased conflict and increased Palisades Palisades
high low exchange, north–south trade becomes
beneficial

with small villages or hamlets and mixed stone and shell-tempered
pottery (Carskadden and Morton, 2000, p. 160). Carskadden and
Morton (2000) argue that during this period local groups in the
upper portion of the Muskingum valley were ‘‘supplanted” by Fort
Ancient villagers who constructed circular villages around plazas
in a fashion similar to those downstream. Philo Punctate pottery be-
came the norm in this region and is quite distinct from other Fort
Ancient wares (Carskadden and Morton, 2000) (see Fig. 1). There
are few indications of interactions with Middle Mississippians.
After approximately AD 1400/1450 there are fewer known sites
across the Fort Ancient region and they are focused closer to the
Ohio River and exhibit a far more consistent pattern of pottery
attributes (see Fig. 2). These sites are indicated by the presence
of Madisonville Horizon Pottery, which is composed mainly of shell
tempering with minimal decoration, large strap handles and a
more diverse set of vessel forms than during the preceding periods
(Drooker, 1997; Essenpreis, 1988; Henderson, 1992). It has been
argued that village sizes increased significantly by the Late period
and became less organized (e.g., Cowan, 1987; Pollack and Hender-
son, 1992, 2000a); however, recent studies have questioned this
conclusion. Many Late Fort Ancient sites were reoccupied several
times, some of which were villages of the same size and structure
as those common in earlier times (Drooker, 1997; see also Cook
(2008)). Some villages undoubtedly grew in size, but how common
the trend was and the precise organization of these settlements is
not known.
Two trends occurred along with the posited increase in village
size: changes in mortuary patterns and house form, both of which
relate to changes in social structure brought on by integrating lar-
ger local groups. Incipient ranking developed in some villages dur-
ing the Late Fort Ancient (AD 1400–1650) (Cook, 2008; Drooker,
1997), and there are considerably more instances of secondary
and mass burials (maybe due to fusing of disparate populations)
and a more regular occurrence of pottery in graves. Pottery in
graves and mass burial could have developed locally but these
are also common occurrences in neighboring Middle Mississippian
and Iroquoian communities, respectively. House form at some sites
includes both longer houses in the eastern portion (e.g., Hardin
Hanson, 1966) and wall trench houses in the western portion
(e.g., SunWatch Cook, 2008, Stateline Vickery et al., 2000) of the
Middle Ohio Valley. The former is suggestive of extended families
replacing nuclear ones (and possibly coincident with Iroquois
influence) (Cowan, 1987, 1988), while the latter has been used to
support an influx of Middle Mississippian migrants (Cook and
Schurr, 2009).
Palisades have been positively identified at six sites (Bosman,
Buffalo, Neale’s Landing, Slone, South Fort, SunWatch), though
researchers have not often looked for them so we cannot know
how representative this is for the region. If palisades were primar-
ily for defense from threats of violence (real or perceived) it is
interesting to note that they are only present in peripheral areas
suggesting defense from non-Fort Ancient neighbors. The only
clear exception is that they are absent from the lower Miami val-
leys and central Kentucky – areas bordering Mississippians to the
west, which is perhaps related to a more ‘‘open” stance vis a vis
those neighbors in contrast to the Iroquois (Fig. 2). It has been
 Author's personal copy

66 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79

Fig. 1. Regional pottery styles from the early and middle Fort Ancient periods. (A) Anderson the Little Miami Valley (copied from Drooker, 1997:Figs. 4–8; see Graybill (1981)
and Griffin (1966)). Anderson-like pottery is common to the Miami and Scioto Valleys (Graybill, 1981; Prufer and Shane, 1970). (B) Feurt Incised sherds from the lower Scioto
Valley (copied from Griffin, 1966: Plate XXII). (C) Philo Punctate sherds from the Muskingum Valley (Carskadden and Morton, 1977). Philo Punctate photos used with
permission of James Morton.

argued that palisades are a later phenomenon (e.g., Buffalo Hanson,
1975; see also Graybill (1981)) but clearly began earlier and were
often rebuilt as villages grew in size (e.g., Slone (Greenlee, 2002),
SunWatch (Cook, 2008)).
Some argue that Middle Mississippian influence occurs mainly
at the onset of the Madisonville Horizon (Pollack and Henderson,
1992, 2000a; Pollack et al., 2002) but others suggest that while cur-
rent evidence most strongly supports this claim, earlier occur-
rences are likely, albeit perhaps fewer in number (Cook, 2008;
Cook and Schurr, 2009; Cowan, 1987; Vickery et al., 2000). There
are simply far too few dates associated with Middle Mississippian
features and artifacts to clarify this important question.
make the investigations of [environmental] variability and buffer-
ing so significant.” Additionally, the investigation of cultural vari-
ability is crucial to a Darwinian evolutionary perspective on
cultural change (Dunnell, 1996a; Dunnell and Greenlee, 1999;
Greenlee, 2002; Madsen et al., 1999; Truncer, 2006). The emphasis
on cultural variability is derived from a population perspective
(materialism) as it relates to Natural Selection. For Natural Selec-
tion to operate there must be heritable variability in the popula-
tion. From this perspective variability is the data (see Lyman and
O’Brien (1998) and O’Brien and Lyman (2000)).
The Winterhalder–Kelly model

Winterhalder (1986) has emphasized the importance of envi-

Evolutionary and ecological perspectives on variability and
interaction
In this section we discuss the general role of environmental and
cultural variability in evolutionary analyses of culture and the spe-
cific models and logic upon which we build our model. Halstead
and O’Shea (1989, p. 5) state that environmental variability ‘‘ex-
ert[s] a strong influence on culture at large, shaping societal orga-
nization and providing the crucial conditions that give rise to social
change and transformation. It is these wider ramifications that
ronmental variability in the structuring of cultural responses and
he specifically isolates two major axes along which environmental
variation influences cultural response: spatial variability (a.k.a.,
interforager correlation) and temporal variability (a.k.a., intrafor-
ager variability; standard deviation). Winterhalder (1986), Fig. 6
arrives at a four class paradigmatic classification of cultural re-
sponses to variation along these axes. Winterhalder is careful to
note that his axes are not composed of two discrete states, but
are instead continuous variables and therefore his four classes
are idealized responses and reality will be much messier.
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
Fig. 2. Location of known fortified villages and Madisonville Horizon sites. Base map after Drooker (2000): Fig. 7:15.
Winterhalder’s (1986) model is (as are most EE models) con-
structed at the conceptual level of the ‘‘individual” forager. How-
ever, there is no theoretical reason that this model could not be
applied to both agriculturalists and groups of organisms. The mod-
el makes no assumptions about subsistence strategy. The key vari-
ables are variability of resource productivity over time (intra-group
standard deviation) and spatial variability in resource productivity
(inter-group correlation). For most EE models, the major difference
between a forager and an agriculturalist would be in the abun-
dance of and distance to certain resources. Agricultural decision
making can be modeled using the same quantitative and qualita-
tive models that have more often been applied to foragers (e.g.,
Keegan, 1986; Keegan and Butler, 1987; Reidhead, 1981; Winterh-
alder and Goland, 1997).2
(Kelly (1995): Chapter 5, Fig. 5-6) has shown how Winterhal-
der’s (1986) model could be adapted to the group level. We refer
to the collective model of group level decision making as the Win-
terhalder–Kelly (W–K) model (Fig. 3). Winterhalder’s (1986) Case
2
We realize this is a simplistic view of the differences in modeling foragers versus
farmers; however, it is beyond our scope to entertain a full discussion of the
applicability of these types of formal models to non-forager situations. We refer the
reader to the cited references for this discussion, especially Winterhalder and Goland
(1997). The point is that the same logic can be applied regardless of ‘‘type” of
subsistence system. One anonymous reviewer specifically referenced the amount of
energy invested in clearing and planting by agriculturalists. This can easily be
incorporated in the calculations of costs versus return. Specifically, the costs of
tending the garden would be added to the handling cost (c.f. search time (Keegan,
1986)) for the cultigens making them more expensive. However, the action of
preparation significantly affects predictability, abundance, and therefore decreases
search time for these resources. Once the cultivated resources join the optimal diet
set they are almost guaranteed to make up a large portion of the diet as the encounter
rate spikes (see Winterhalder and Goland (1997)). This is just a sketch of how to begin
dealing with the changes in resource cost and distribution brought about by a change
in subsistence strategy; however, the main point, that these models are applicable to
farmers, stands.
67
A is characterized by high inter-group correlation of return among
individuals (e.g., foragers, farmers, communities) and high intra-
individual variation (standard deviation). Case A results in carry-
over averaging (household storage) and non-local exchange and/
or migration. Kelly (1995) adds perimeter defense and warfare
(Fig. 3). Under these circumstances we expect little evidence of ex-
change (either in traceable non-perishable goods or in stylistic
elaboration). Additionally, stockades and other signs of direct con-
flict (e.g., skeletal evidence of violent death) would be more
prevalent.
Perimeter defense becomes more likely in cases of resource
abundance (Dyson-Hudson and Smith, 1978). It seems counter-
intuitive that abundance of resources can lead to increases in theft.
However, theft is only a high return activity if there is someone
with a lot of resources. This is related to the justification for sharing
(a.k.a. tolerated theft) of large game meat (for example) among
hunter–gatherers (e.g., Kaplan et al., 1990). In cases characterized
as tolerated theft, the cost of defense is too high for the benefits re-
tained by putting effort into exclusion. With storable agricultural
produce (or livestock Dyson-Hudson and Smith, 1978), protecting
resources and excluding others becomes a viable alternative. If
there are individuals with an abundance of resources, there is the
possibility of a high pay-off for the energy invested in theft (en-
ough to compensate for the risk involved). Additionally, the bene-
fits of exclusion increase if the resources are storable (Kaplan et al.,
1990). Therefore, in situations characterized by W–K Case A and
abundance, the development of hostilities will be most pro-
nounced because of the co-occurrence of a time-transferrable sur-
plus and high likelihood of any one individual falling short of
required resources at a given time (i.e., high intra-group standard
deviation).
Winterhalder’s (1986) Case B is characterized by a high intra-
group standard deviation and low inter-group correlation among
individuals. The predicted response is sharing and/or exchange
 Author's personal copy
68 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
Fig. 3. The Winterhalder–Kelly model. The horizontal axis represents inter-group
correlation (inverse of spatial variability; i.e. low correlation equals high spatial
variability) in food yield and the vertical axis is intra-group variance (aka: temporal
variation) represented by the standard deviation (SD) of the yield. When an
individual’s (group or organism) resource productivity is highly correlated with
individuals nearby there is little to be gained by exchange among individuals. That
is, low spatial variability should lead to low levels of exchange. According to the
model, people are expected to trade with individuals who experience uncorrelated
resource productivity. In general when temporal variability (SD) increases individ-
uals will be most successful with strategies that minimize the likelihood of falling
below a critical minimum amount of resources. Storage is a common solution to
high temporal variability. This four class plot is a generalized model and both axes
are in reality continuous. Figure reproduced after (Kelly (1995): Fig. 5-6; see also
Winterhalder (1986)).
possibly supplemented with storage. Kelly (1995) adds social-
boundary defense and intense, rigid reciprocal exchange. We inter-
pret Kelly’s notion of social-boundary defense to be related to the
increase in exchange. For agriculturalists it is not simply regulating
access to territory but regulating access to surplus. This would in-
clude similar social mechanisms employed by the hunter–gather-
ers that are the focus of Kelly’s efforts (e.g., lineage/moiety
bonds, fictive kinship). In addition we would expect to see the
use of shared, emblemic (sensu Weissner, 1983) symbols to medi-
ate the establishment of alliances and trade relationships. Under
Case B we expect increasing interaction among groups. This should
be manifest materially by stylistic (sensu Dunnell, 1978) elabora-
tion of material culture and stylistic similarities among interacting
groups (Braun, 1991; Lipo et al., 1997) and the presence of non-lo-
cal materials. Winterhalder’s (1986) Case C obtains when return is
predictable (low intra-group standard deviation) and is very likely
the same as neighboring groups (high inter-group correlation);
there is, therefore, either no need or no option for inter- or intra-
individual averaging (exchange or storage, respectively). Kelly
(1995) predicts the development of passive territories with mini-
mal hostilities and occasional migration. In cases characterized
by these conditions there should be limited stylistic elaboration
(see Braun (1991)) and little to no evidence for violent inter-group
conflict (e.g., stockades, projectiles embedded in bone). Finally,
W–K Case D is characterized by a consistent return per individual
(i.e., low intra-group standard deviation) but little inter-group cor-
relation in production with neighbors (i.e., high spatial variability).
In Case D, each group semi-specializes and provides their neigh-
bors with the goods they have in excess and receive in return what
is lacking locally (i.e., differentiated exchange). As indicated, each
of these cases would leave a distinct signature in the archaeological
record; however, it must be remembered that these cases are not
discrete. Both axes are continuous in reality.
When examining diachronic change in archaeological contexts
we cannot simply rely on the W–K model to tell us what to expect.
The W–K model is timeless (i.e., space-like sensu Dunnell, 1996c, p.
74); that is, it does not take account of history in its predictions.
The W–K model is intended to predict cultural response at a single
point in time, or over a short span based on knowledge of the sta-
tistical probability distribution for food returns. While it is true
that use of the standard deviation requires temporal data, the scale
of observation is quite different from the time scale of the archae-
ological record. The W–K model will make a static prediction
regardless of the historical trajectory of the group under consider-
ation. By applying the model iteratively, we create a modeled his-
torical context within which to consider cultural response and
inertia (not considered in the original model). Therefore, we do
not rely solely on the predictions of the W–K model when detailing
our expectations for cultural responses, but draw on a few general
evolutionary models in presenting expectations for prehistoric cul-
tural change in response to changes in environmental variability.
Other evolutionary models
One of the most important shifts in the evolution of human cul-
ture has been the development of complex, hierarchical societies.
According to Dunnell (1996a, p. 92) change in level of socio-polit-
ical complexity is characterized by a shift in the level at which the
selection on cultural variability is primarily focused. As the number
of different, exclusive, and mutually dependent roles in a society
increases (specialization), organisms no longer carry the full com-
plement of information necessary to reproduce the behavioral rep-
ertoire of their cultural group (Dunnell, 1996b, p. 51). In more
complexly organized groups, selection on cultural variability is, un-
der this logic, acting on the whole cultural group and not on the
information contained by individual organisms. This dichotomy
is artificial. There is no quantum leap in level of organization.
The parsing of cultural information into separate but mutually
dependent bundles is likely gradual.
We are interested in cumulative change in cultural entities and
specifically the emergence of social complexity (albeit at an incip-
ient level see Cook (2008)) in the Middle Ohio River Valley. There-
fore, following Dunnell’s logic, our unit of analysis is the
community. According to most interpretations, there is equiva-
lence between a Fort Ancient village/habitation site and the com-
munity (e.g., Essenpreis, 1988; c.f. Essenpreis, 1978). Smaller
sites may represent seasonal uses by some members of those com-
munities (e.g., Brose and White, 1983; Purtill, 1999; Turnbow and
Jobe, 1984), but the possibility of satellites in a small-scale hierar-
chy remains unexplored (see Cook (2008)).
Pooling resources and pooling risk within a group can help en-
sure that organisms decrease their probability of failing to meet
their minimal requirements (Winterhalder, 1990). Additionally,
increasing the size of the community pooling resources can be an
effective solution for managing high temporal variability. For farm-
ing communities this buffering effect can be even more pro-
nounced. One strategy at the household level to minimize risk of
failure is to scatter garden plots across space. This strategy mini-
mizes the likelihood of catastrophic failure for subsistence agricult-
uralists and takes advantage of the inherent spatial variability of the
factors affecting agricultural productivity (Winterhalder, 1990).
Increasing the size of the food producing community pooling re-
sources increases the ability of the group to weather patches of high
temporal variability and would also potentially increase the ability
of community to withstand modest spatial variability on a small
scale. Complex social organizations are well suited to accomplish
this sort of risk minimization (Halstead and O’Shea, 1989).
Once a region experiences the development of group functional
behavior and natural selection on cultural variability is acting (at
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
least partially) directly on aggregates, new evolutionary pressures
may surface. Due to the competitive advantage in defense and (di-
rect or indirect) competition for resources, groups that have the
buffering mechanisms associated with larger populations (produc-
tion effort adapted to stressful periods) will increase in relative fre-
quency over time. That is, communities with effective buffering
mechanisms (e.g., social complexity, centralization, and more
members) will produce daughter communities that are more suc-
cessful in inter-village competition.
This is similar to the reasoning of Richerson et al. (2001) regard-
ing the ‘‘mandatory” nature of agriculture during the Holocene. In a
world of foragers, an energy intensive strategy like agriculture will
be selected against; however, once agriculturalists are on the scene
the selective equation changes. The increase in population associ-
ated with grain agriculture and the inherent instability of the adap-
tation (see also Rindos (1980, 1984)) cause the rapid spread of
agriculture at the expensive of conservative hunter–gatherer groups
(Richerson et al., 2001). The more complex groups and their atten-
dant ability to maintain production and health in bad years are anal-
ogous to Richerson et al.’s (2001) higher reproducing farmers. The
less complex groups are analogous to the foragers with lower popu-
lation densities and lower reproductive rates. This competitive
ratchet (sensu Richerson et al., 2001) would apply pressure towards
complexity (or buffering mechanisms of equivalent function) inde-
pendently of fluctuations in the physical environment. Therefore,
social–political complexity may have been ‘‘mandatory” in some re-
gions adjacent to areas where complexity developed out of need for
high order buffering mechanisms due to environmental variability.
An evolutionary model of Fort Ancient social change
The basis for our regional analysis of environmental variability
is set largely within an evolutionary ecology framework. After pre-
senting the theoretical context, we present our methods (which fo-
cus on fine-grained moisture data), then present the model for the
region of interest. We concluded this section by assessing the fit
with the data outlined above.
The Palmer Drought Severity Index (PDSI) reconstruction cre-
ated by Edward R. Cook and colleagues (1999, 2004). Cook et al.’s
PDSI reconstructions are based on dendroclimatological analyses,
and provide data on moisture availability at an annual scale for
all of North America. The PDSI is ‘‘a widely used measure of relative
drought and wetness over the United States” (Cook et al., 2004, p.
1016; emphasis added) and there is a growing archaeological liter-
ature using the index (e.g., Benson et al. 2007; Cooper, 2008; Stahle
et al., 2007). The PDSI scale ranges from 6 to +6 (arbitrary) using
Fig. 4. Maps of the analytic region. (A) This figure shows the location of Cook et al.’s (2004) interpolated PDSI grid points and the locations of the extant sample of tree-ring
data upon which Cook et al.’s interpolations are based. The reader is referred to the original publications and the NOAA website for more information about how the PDSI
network is calculated and the details of individual tree-ring datasets. (B) The approximate location of relevant Late Prehistoric cultural groups after Cook (2008), Fig. 1.1.
69
‘‘(n)egative (positive) PDSI values [to] indicate dry (wet) periods,
while those near zero presume a state near the average” (Mika
et al., 2005, p. 224). Cook et al. (1999, p. 11,477) found that tree-
ring data correlate best with summer PDSI values and therefore
the reconstructed values are primarily estimations of summer pre-
cipitation for a given year. This is appropriate for our analysis in
the temperate eastern United States as summer precipitation has
the most direct impact on agricultural productivity.
We are not the first to approach the use of PDSI from an evolu-
tionary perspective to evaluate human response to climate induced
subsistence risk. In discussing the effects of climatic variability on
subsistence systems, Larson et al. (1996) state that:
An ideal measure of variability would be one which fluctuates
independently of the average. The standard deviation is
appropriate for a normal distribution, but with data such as pre-
cipitation amounts, for instance, standard deviation often
increases with increasing averages. The Palmer Drought Sever-
ity Index (PDSI) is calculated in such a way as to make it nor-
mally distributed with a mean of 0. As a result, we feel that
the standard deviation provides a good summary of overall var-
iability. From an anthropological viewpoint, the standard devi-
ation can be interpreted as a coefficient of high-frequency
subsistence risk. Periods with the highest standard deviations
will have the greatest variability, implying the greatest amount
of unpredictability for human populations (Larson et al., 1996,
p. 225).
The nature of the PDSI scale and especially the relationship be-
tween the long-term mean and the standard deviation make the
standard deviation of PDSI the most appropriate measure of tem-
poral variability for the W–K model as used here.
The current PDSI reconstructions consist of a 2.5°  2.5° grid of
interpolated (inverse distance weighting, IDW) PDSI values (Cook
et al., 2004). This data is available for querying and download on
NOAA Paleoclimatology’s World Data Center for Paleoclimatology
and the Applied Research Center for Paleoclimatology (http://
www.ncdc.noaa.gov/paleo/pdsi.html). See Fig. 4a for the location
of grid points used (N = 24) in our analysis and the tree-ring data-
sets that are the basis of Cook et al.’s (2004) reconstructions. We
employ the PDSI values as a measure of relative precipitation.
Procedures
We have selected the period AD 801–1400 as our analytical
focus, segmenting the interval into 50-year periods to consider
 Author's personal copy

70 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79

Fig. 5. AD 801–900 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. Interpolation results for AD
801–850 (A–C) and AD 851–900 (D–E). Subfigures A and D present average PDSI distribution for the first and second 50 year period, respectively. The scale for A and D is
based on the scale provided on the data download site (http://www.ncdc.noaa.gov/paleo/pdsi.html; subfigure 3g below). Spatial variation is represented by the steepness of
the isopleth gradients (A and D). That is, tightly packed isopleth lines indicate high variability across space. This allows us to see in which directions from any given point
returns will be correlated. The standard deviation of PDSI for the first and second fifty year period are presented in subfigures B and E, respectively. The resulting W–K model
plot for each period is presented in subfigures C and F. The grid points are represented in the W–K plot as abbreviations for the location of the grid point. Grid point 218 is in
western Indiana (wI); 219 is in western Kentucky (wK); 227 is in eastern Indiana (eI); 228 is in central Kentucky (cK); 236 is in central Ohio (cO); 237 is in eastern Kentucky
(eK); 246 is in western Pennsylvania (wP); 247 is in western Virginia (2). It is important to note that C and F only plot the relative spatial variability and temporal variability
(standard deviation) for the actual grid points. This does not tell the whole story of regional variability on either axis. Quite often peak spatial variability will be found
between the PDSI grid points and it will be necessary to examine subfigures A and D to fully understand the nature of the predicted relations in any one area. It is possible to
create a W–K plot for any point within the interpolated region. The scale for the mean PDSI (A and D) and temporal variation plots (B and E) is the same for each of the
following figures for ease of comparison. Isopleth intervals for all maps is 0.05 PDSI units.

patterns of spatial and temporal variability. Changes in the precip-
itation regime will be used to predict responses by prehistoric soci-
eties based on the models discussed in the previous section. We
focus our efforts in the Middle Ohio River Valley portion of the re-
gion (Grid Points 218, 219, 227, 228, 236, 237, 246, 247 in Fig. 4a).
However, to avoid unnecessary edge effects the surrounding region
was included in the analysis. After the appropriate grid points were
selected and the data downloaded, the raw PDSI files were com-
piled into separate Excel spreadsheets for each 50-year period.
The tree-ring data and the reconstructed PDSI values do not cover
the whole time period for all grid points. If the data was not avail-
able for the entire 50-year period the grid point was omitted from
the analysis for that period. For the data points used in this analy-
sis, the result is the gradual westward then northward expansion
of the area covered to eventually encompass the American Bottom
and parts of the Mississippi River Valley. The Middle Ohio River
Valley is covered for the entire period of analysis.
An average and standard deviation were computed for each 50-
year period at each grid point used. The spreadsheets were then
imported into ESRI shapefile format using the coordinates provided
with the data download. An analysis mask was used in the time
periods that did not have data for all grid points, and the maximum
extent of inputs was used to limit analysis for the remaining peri-
ods. The data contained in the point shapefiles for each period
were then interpolated (IDW) to predict the expected PDSI values
for areas between the grid points. Interpolations were performed
for both mean and standard deviation for each period. The result-
ing 50-year average and standard deviation maps are displayed in
Figs. 5–10. In what follows we discuss the key aspects of distribu-
tion and variability in precipitation and the predicted cultural
responses.
The model
In this section we detail our model for prediction of social
change in the Middle Ohio Valley. We start 200 years before the
Fort Ancient period to provide a modeled historical context within
which to view subsequent developments, and we stop before the
end of the Fort Ancient period because there are no further clear
shifts in social development. Two concerns (other than the theoret-
ical concerns raised above) should be kept in mind when process-
ing this information and turning it into a diachronic model for
expectations of social responses. First, as we are talking about
the action of selection in response to changes in the modeled
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
selective environment (PDSI interpolations) the predicted changes
are expected to come at an unspecified (and unknown) lag behind
the onset of new selective conditions. Second, in applying the W–K
model it is difficult to know what constitutes a ‘‘high” and ‘‘low”
intra-group standard deviation. Similarly with the proxy for in-
ter-group correlation used here (i.e., inverse of spatial variability),
we have no interval scale measure. We use a scale of temporal var-
iability that overlaps the observed range of standard deviations (1–
2).3 The spatial variability dimension will have to remain fairly dis-
crete and relative with low (i.e., high correlation in returns across
space; widely spaced isopleths), moderate, and high (i.e., low corre-
lation in returns across space; closely spaced isopleths) values. This
will allow us to plot our values against a two dimensional plot sim-
ilar to Kelly’s (1995) Fig. 5-6 reproduced here as Fig. 3 (see also Win-
terhalder (1986), Fig. 6).
In presenting the model we will repeatedly refer to the two axes
of the W–K model. In reference to the modeled distribution of pre-
cipitation, subfigures A and D in Figs. 5–10 display the distribution
of variation in average precipitation for each 50 year period. The
greater the concentration of isopleths (like contours on a topo-
graphic map) in these maps the higher the level of spatial variabil-
ity or the lower inter-community correlation (see Fig. 3). The
orientation of the isopleth gradients can be used to predict the
direction(s) in which exchange is likely to be most beneficial. It
is most beneficial to exchange with neighbors in a direction per-
pendicular (up or down slope) to the isopleths (i.e., trade with
the people whose returns are least like yours).
In subfigures B and E we present the modeled distribution of
standard deviation of PDSI. This is the spatial distribution of the
vertical axis in the W–K model (see Fig. 3). What is important in
subfigures B and E is the actual value of the standard deviation
for any one point on the map. As can be seen by examining subfig-
ures A–E, the exact response predicted at any one location may be
different than that predicted for the PDSI grid points. For expedi-
ency, we have plotted the movement of the available grid points
on the W–K grid for each 50-year period modeled in subfigures C
and D. Subfigures C and D do not capture all, or even the most rel-
evant variability, but they can be used as a guide for the interpre-
tation of the maps in each figure.
AD 801–900
The period AD 801–850 in the Middle Ohio River Valley is one of
the driest intervals under consideration characterized by low pre-
cipitation with low to moderate spatial variability and moderate
temporal variability (Fig. 5a and b). For the region as a whole we
expect a diversified settlement-subsistence system with small
groups scattered about the landscape with only minimal exchange
and periodic relocation of settlements due to unpredictable short-
3 Ohio saw a marked increase in spatial variability. This represents a
In the W–K plots used in this presentation we do not intend the line across the
center to be exactly the same as the horizontal line used by Winterhalder, 1986; Kelly,
1995. The line is a convenient visual landmark for the middle of the range of standard
deviations used. Remember that these axes are continuous and that the cultural
responses will not fit discretely into (for example) Case A. We do not intend this to
represent the cutoff point between the four classes of response. What must be kept in
mind for our presentation are relative positions on the W–K plot and trends through
time in relative spatial and temporal variability. Our W–K plots are centered on the
observed range of variability within the analysis period. If the choice of scale is found
to be inappropriate, this would not invalidate the basis or usefulness of our model,
only the specific predictions outlined. This model can be applied with any appropriate
scale to yield specific predictions about the variables discussed herein. Additionally,
at least one of the reviewers correctly noted that our spatial variability of PDSI does
not cover all that is implied by inter-group correlation axis of the W–K model. We
make no assertion that we have explored every possible application of the logic of the
W–K model in this paper. However, for agriculturalists this captures a significant
aspect of spatial variability in the environment that would condition social response.
Finally, the scale of the maps presented here is not meant to imply that the same data
cannot be used for analysis at finer scales. In fact, the promise for finer-scale analysis
is one of the most important aspects of this model.
71
falls. Perimeter defense is not likely. The shortage of water during
this period makes it unlikely that these late Late Woodland food
producing societies experienced an abundance of crops to the point
that engaging in warfare and defense would have been economical.
The region was still dry during the AD 851–900 interval and we
predict subsistence systems continued experiencing stress
(Fig. 5d). The regions north of the Ohio River move closer to Case
C (Fig. 5f). We expect mobility with passive territories and low lev-
els of exchange. Perimeter defense is even less likely. We expect
slightly more emphasis on trade south of the Ohio River (Fig. 5f).
AD 901–1000
The entire modeled region sees an increase in temporal variabil-
ity during the period AD 901–950 (Fig. 6b). The intensification of
temporal variability amplifies the likelihood of increasingly rigid
boundaries. Inter-group conflict becomes increasingly likely during
this period, especially north of the Ohio River. While the region is
still experiencing a moisture shortage the severity is decreasing,
and an increasing abundance of agricultural produce is likely. More
of a focus on defense of territories is predicted. This should be asso-
ciated with a decrease in frequency of residential moves and, espe-
cially in Kentucky, the development of social-boundary defense
and reciprocal exchange relationships. We expect to see some
non-utilitarian embellishments return to artifact classes such as
ceramics. The development of stylistic embellishments should be
relatively slow because there is not very strong selective pressure
for trade. With greater spatial variability (and therefore greater
ability to alleviate local shortages via exchange) focused along
the trunk of the Ohio River and to the south, we expect these to
be the primary areas of stylistic expression at this time.
The standard deviations of PDSI have consistently been larger in
the western portions of the area (Figs. 5c, f and 6c). The temporal
variability, most acute in central Kentucky (228) and southwestern
Ohio/southeastern Indiana (227), selects for buffering mechanisms
to adjust expenditure of energy to the lower level of carrying
capacity (Dunnell, 1996a, Fig. 4.2; Madsen et al., 1999). Another
(not mutually exclusive) option is a shift in the level of social orga-
nization and/or group size. The formation of a larger community
can combat an unpredictable distribution of temporal variability
by pooling their returns which would reduce the probability of
individuals falling below a critical threshold (aka: risk; see discus-
sion above). We therefore predict an increase in the frequency of
larger sites in the regions hardest hit by the temporal variability
that are also experiencing modest spatial variability.
Temporal variability decreases during the period AD 951–1000
(Fig. 6e). The standard deviations here are still higher than the AD
851–900 period, however. The western end of the study area
would have experienced a decrease in exchange and possibly a de-
crease in residential stability. In contrast, southwestern and central
shift to relatively higher spatial variability for most of the Fort An-
cient region (Fig. 6d–f), favoring increases in exchange and social
mechanisms for the regulation of access to resources. This would
result in greater stylistic elaboration and distinctive distributions
of certain styles.
The temporal variability would be managed by increasing stor-
age (carry-over averaging; Winterhalder, 1986; Kelly, 1995), and
pooling of resources among multiple economic units (Winterhal-
der, 1990). Here we expect to see increases in the size of settle-
ments and increasing organization within settlements. This is
likely to be manifest in the physical footprint of the village and also
in the relationships between people within the village. We expect
to see indications of increasing organizational complexity. The
pressure for increasing size and complexity of communities is fo-
cused around southwestern Ohio and central Kentucky. These
areas should be precocious in their development of social complex-
 Author's personal copy

72 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79

Fig. 6. AD 901–1000 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. Results for AD 901–950
average PDSI distribution (A), standard deviation of PDSI (B), and resulting W–K model plot (C). Results for AD 951–1000 average PDSI distribution (D), standard deviation of
PDSI (E), and resulting W–K model plot (F). See Fig. 5 caption for full description.

Fig. 7. AD 1001–1100 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. New data becomes available
for this period, extending the region covered by our interpolations past the confluence of the Ohio and Mississippi Rivers. Results for AD 1001–1050 average PDSI distribution
(A), standard deviation of PDSI (B), and resulting W–K model plot (C). Results for AD 1051–1100 average PDSI distribution (D), standard deviation of PDSI (E), and resulting
W–K model plot (F). See Fig. 5 caption for full description.

ity and other high order buffering mechanisms. A long period of variability would increasingly select for centralization and
temporal variability coupled with demands of increasing spatial complexity.
 Author's personal copy

K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79 73

Fig. 8. AD 1101–1200 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. Results for AD 1101–1150
average PDSI distribution (A), standard deviation of PDSI (B), and resulting W–K model plot (C). Results for AD 1151–1200 average PDSI distribution (D), standard deviation of
PDSI (E), and resulting W–K model plot (F). See Fig. 5 caption for full description.

Fig. 9. AD 1201–1300 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. Results for AD 1201–1250
average PDSI distribution (A), standard deviation of PDSI (B), and resulting W–K model plot (C). Results for AD 1251–1300 average PDSI distribution (D), standard deviation of
PDSI (E), and resulting W–K model plot (F). See Fig. 5 caption for full description.

Increasing the size of the pool is not the only way to protect play a role in situations like this. Selection in temporally variable
against unpredictably variable energy returns in the temporal environments favors behaviors that decrease the likelihood of fall-
dimension. Dunnell’s (1996a) waste hypothesis is also likely to ing short in the worst of times. Any behavior that ensures the level
 Author's personal copy
74 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
Fig. 10. AD 1301–1400 interpolated PDSI. The results of the IDW interpolation from Cook et al.’s (2004) PDSI grid points with data for the period. Results for AD 1301–1350
average PDSI distribution (A), standard deviation of PDSI (B), and resulting W–K model plot (C). Results for AD 1351–1400 average PDSI distribution (D), standard deviation of
PDSI (E), and resulting W–K model plot (F). See Fig. 5 caption for full description.
of available effort will be sufficient to feed the organism(s) when
resource returns are at their lowest will increase in frequency in
such environments (Madsen et al., 1999). Increasing social com-
plexity can be a way of accomplishing this. First, the redistribution
of surplus to a centralized store acts as a risk pooling mechanism to
buffer micro-scale spatial variation (Halstead and O’Shea, 1989;
Winterhalder, 1990; see also Church and Nass (2002)). Second,
by requiring production above the normal needs of the economic
unit for tribute the centralized community will be better adapted
to the times when production is low. Viewing increasing complex-
ity in this way allows us to conceive of hierarchy as a form of waste
behavior (sensu Dunnell, 1996a).
All this combines to give us the expectation that community le-
vel activities and management of resources should begin to emerge
during this period in the places experiencing high spatio-temporal
variability in crop yields (i.e., southwestern Ohio, central Ken-
tucky). This is the beginning of the transition to group functional
behavior. It must be remembered that what is important in making
predictions of change in our iterative application of the W–K model
is not just synchronic position on the W–K plot, but long-term
trends as well. For communities near the PDSI grid points in south-
eastern Indiana and central Kentucky, a nearly 200 year history of
higher temporal variability than their contemporaries in other
parts of the study region makes them more likely to move in the
predicted direction.4
4 poral variability would be a significant factor selecting for increas-
One reviewer pointed out that the temporal variability for the period AD 801–850
is higher than that for AD 951–1000 yet similar changes were not predicted. This is an
artifact of the choice of starting point for the model. We have no modeled history for
the first period. For all intents and purposes, the modeled responses for AD 801–850
are timeless (space-like). However, in the subsequent periods we have a modeled
history and now our model is no longer synchronic (a contrast with the original W–K
model). As more time depth becomes available in the reconstructed PDSI record it will
be possible to analyze the Late Woodland record in the same way we have modeled
the Late Prehistoric period.
AD 1001–1100
The area covered by the interpolations expands past the Amer-
ican Bottom during the period AD 1001–1050 (Fig. 7a and b). For
the first time there is no moisture shortage (i.e., PDSI values zero
or greater) anywhere in the region (Fig. 7a). This increases the like-
lihood of perimeter defense (abundance), especially in areas with
high temporal variance (see Fig. 7b). Southeastern Indiana (227),
central Kentucky (228), southwestern Pennsylvania (246), and
western Virginia (247) are characterized by high-moderate spatial
variability. This translates into an increased expectation for ex-
change. We do not, however, expect to see well developed ‘‘rigid
reciprocity” or ‘‘differentiated exchange” (Kelly, 1995, Fig. 5-6;
see Fig. 3).
Central Ohio and eastern Kentucky experience moderate tem-
poral variability and low spatial variability. This period bifurcates
the Fort Ancient area into eastern and western halves. When we
look at the interpolated surface we see that it is essentially central
and south-central Ohio that experienced decreases in spatial vari-
ability (Scioto valley; see Fig. 7a). Most of eastern Kentucky and
western Ohio experience high levels of spatial variability. This
has significant implications for the distribution of styles within
the Fort Ancient area. We predict there was less emphasis on ex-
change by communities in the middle and upper Scioto Valley rel-
ative to the rest of the region.
The western part of the analytical region (Fig. 7c) exhibits the
highest level of temporal variability yet seen. The history of this
variability is uncertain. However, the extreme nature of this tem-
ing levels of social organization and risk buffering mechanisms.
The standard deviations continue to increase until one reaches
southwestern Illinois and southeastern Missouri (American Bot-
tom region; see Fig. 7b). It is probably no coincidence that this area
saw the development of the earliest ‘‘chiefdom” level societies in
eastern North America. In this part of the study region there is lim-
ited expectation of trade during this period (Fig. 7c). We must urge
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
caution in running with these expectations. We have no modeled
historical context of precipitation for this portion of the study area
in our data from which to consider the situation in the American
Bottom during this period and therefore this prediction is timeless
(i.e., space-like, sensu Dunnell, 1996c, p. 76).
The relief of moisture stress seen in the AD 1001–1050 period
continues to the point where the entire western half of the region
is above a PDSI of 0.5, and all but the far southeastern corner reg-
ister positive values during the period AD 1051–1100 (Fig. 7d).
Temporal variance for most sites increases during this period
(Fig. 7e). Eastern Indiana, central and eastern Kentucky and central
Ohio have standard deviations greater than 1.5 and experienced
moderate to high spatial variation with the highest spatial varia-
tion between grid points (Fig. 7e). This reverses the bifurcation of
the Fort Ancient region. We expect increasing levels of exchange
throughout southern Ohio and eastern and central Kentucky.
East–west trade would be the most beneficial during this time per-
iod as sites to the north and south of each other are highly corre-
lated in their returns. The long-term increase in precipitation and
therefore crop abundance again increases the benefits of theft
and therefore perimeter defense (Dyson-Hudson and Smith,
1978). However, we expect there to be a mix of perimeter defense
and reciprocity. Any communities in the area of high spatial varia-
tion between the grid points are likely to have a high balance of
reciprocity mediated relationships with their neighbors to the east
and west. Communities in central Ohio, west-central and extreme
eastern Kentucky, and southeastern Indiana are more likely to tip
the balance towards perimeter defense due to the higher levels
of inter-group correlation with neighbors (see Fig. 5d–f).
There continues to be a steep east–west gradient in temporal
variation, especially from southwest Ohio westward (Fig. 7e). This
pattern culminates in extreme levels of temporal variability in the
Mississippi Valley. Western Indiana into Illinois exhibits very low
spatial variability of precipitation. The variation in precipitation
appears to be concentrated around the Ohio and Mississippi rivers
in the western portion of the region. These conditions would exert
increasing pressure on buffering mechanisms. The low-medium
spatial variability (increasing inter-group correlation of return
with larger areas) in the western portion of the analytic region
would put increasing pressure on societies in these areas to estab-
lish trade contacts with more distant neighboring regions. More
distant trade contacts allow access to more variable (and uncorre-
lated) returns through exchange and therefore greater averaging
ability. Increasing levels of political complexity facilitate long dis-
tance trade. Additionally, the temporal variability would favor sys-
tems that have adapted their energy expenditure capability to the
lowest (and now frequent) levels of production for the region.
AD 1101–1200
The period AD 1101–1150 saw a return to moisture stress (aver-
age PDSI < 0) for the western portion of the region (Fig. 8a). The
Middle and Upper portions of the Ohio valley remain little changed
in the availability of moisture. Interestingly, western Indiana and
western Kentucky exhibit less temporal variability than do eastern
Indiana and central Kentucky (Fig. 8c). This is a relaxation of the
temporal stress on Lower Ohio valley populations (see Fig. 8b).
However, these populations are experiencing very strong pressure
for exchange, especially as this region was experiencing a decrease
in crop abundance (decreased precipitation). Therefore physical
defense of territory is less likely in this part of the region. We pre-
dict an emphasis on social mechanisms of maintaining access to
food (e.g., fictive kinship, reciprocity) within the western portion
of the Ohio Valley. Additionally, east–west trade is predicted to be-
come more frequent.
If we look at Fig. 8a we see that high-to-moderate levels of spa-
tial variability are maintained from south-central Ohio through
75
northern Kentucky (i.e., between PDSI grid points) in a north–south
gradient. This is also true of eastern Indiana and central Kentucky.
Again we see that central Ohio and eastern Kentucky are character-
ized by different regimes of spatial variability than their western
counterparts in the Fort Ancient region. This recurrent difference
would affect the frequency and intensity of interaction with others
outside their immediate environment. This should have stylistic
ramifications: more localized styles and few styles from the west.
The moderate level of spatial variability experienced in the lower
Scioto valley (for example) is in a north–south direction. Therefore,
modest trade along the course of this stream is predicted. The east-
ern Fort Ancient area is much more likely to devote some energy to
territorial defense than their neighbors in central Kentucky and
southwestern Ohio due to depressed spatial variability and a high-
er moisture budget (i.e., increased abundance).
For the period AD 1151–1200 there is a relief of moisture stress
for the whole Ohio River Valley (Fig. 8d). The Lower Ohio Valley
experienced increasing pressure on productive effort (Fig. 8e),
and an increase in social complexity is the predicted solution.
These are continuations of trends from earlier periods. This does
not mean that all villages instantaneously switched to a higher le-
vel of social organization, but that organisms in those communities
that did adopt these behaviors would have fared better nutrition-
ally and therefore likely reproductively. This would engage a ver-
sion of Richerson et al.’s (2001; see discussion above)
competitive ratchet. Within the Fort Ancient region, this pressure
would be most intense in southwestern Ohio and central Kentucky
due both to proximity to societies of higher complexity (see Cook
(2008)) and the consistently higher levels of temporal variability.
At the same time, the Middle and Upper portions of the Ohio Valley
see an increase in temporal variability and either slight increase to
or sustained moderate spatial variability. We predict increasing
importance for exchange, but still pressure for a defensive posture
due to much lower spatial variability than their western neighbors.
AD 1201–1300
During the period AD 1201–1250 the entire region was once
again characterized by sufficient moisture (Fig. 9a). The highest
levels of precipitation were centered on southeastern Indiana
and, southwestern Ohio with steep gradients to the west. The spa-
tial variability of the Lower Ohio Valley increased slightly while the
temporal variability hovered in the same range (Fig. 9b and c). The
communities around eastern Indiana/southwestern Ohio and cen-
tral Kentucky would have experienced greater spatial variability
than their western neighbors, but only moderately. Southwest
Ohio, central Kentucky, and especially south-central Indiana expe-
rienced relatively higher pressure for spatial averaging (exchange)
along an east–west gradient. According to Kelly (1995), Chapter 5
this calls for ‘‘intensive and rigid reciprocity between groups”
and social-boundary defense. We predict more formal and formally
displayed relationships of exchange during this period among
groups of the western Ohio valley.
The long-term pattern of maximum spatial and temporal vari-
ability focused on the Lower Ohio Valley into southwestern Ohio
along an east–west gradient leads us to expect the Fort Ancient
people of southwestern Ohio were interacting most intensely with
their Mississippian neighbors to the west and not with other
groups lumped under the vaguely defined rubric of Fort Ancient.
Eastern Kentucky and Virginia see a slight increase in spatial
variability over the previous period while central Ohio and Penn-
sylvania experience a slight decrease in spatial variability
(Fig. 9c). Temporal variability is essentially unchanged. We predict
a steady emphasis on defense in the northern sites with an increas-
ing role for exchange among the southern sites, especially around
western Virginia.
 Author's personal copy
76 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
There were dramatic changes in the distribution of spatial var-
iability during the period AD 1251–1300 (Fig. 9d). For the Lower
Ohio Valley we predict a shift away from exchange as the primary
means of mediating shortfalls toward an emphasis on territorial
defense (Fig. 9d–f). The sufficient moisture budget (resource abun-
dance) increases the likelihood of conflict in the western portion of
the Ohio Valley.
Southwest Ohio, central Ohio, and eastern Kentucky communi-
ties likely become trading partners. Again the primary direction of
trade is predicted to be east–west. We expect a decrease in the for-
mality of exchange; however, it is possible that inertia will swamp
this temporary lapse in pressure for formalized exchange. An alter-
native option is that the western Fort Ancient communities and
Mississippian groups of the Lower Ohio Valley would introduce
their more rigid practices of exchange to their eastern trading part-
ners. This would be the familiar competitive ratchet. We predict a
decrease in emphasis on defense among central Ohio and eastern
Kentucky groups during this period.
AD 1301–1400
Spatial variability dramatically decreased across the region
from AD 1301–1350 (Fig. 10a). Western Indiana, western and
central Kentucky, and eastern Indiana all experienced very high
temporal variation (Fig. 10b) in addition to their high level of in-
ter-group correlation. Again, simple reliance on the W–K model
would be problematic here.
These grid points are predicted by the W–K model (based on our
scales) to transition from emphasis on exchange to emphasis on
defense. However, we cannot forget that the previous 200 years
have been selecting against defense in most of these regions.
Therefore, it is unlikely that we would see a wholesale shift in such
a short period of time from exchange to conflict. We do expect that
inter-village conflict increased during this period. However, the
previously predicted east–west network of trading partners and
the formalized nature of these connections would have opened
up a much larger area over which to average returns. Halstead
and O’Shea (1989) argue that with limited technology for trans-
port, long-distance averaging of this type is unlikely. However, this
region has a prehistoric highway running through it: the Ohio Riv-
er. Water transport would significantly reduce the cost of long dis-
tance travel in the absence of the wheel or draft animals. The
correspondence between the orientation of the Ohio River and
the orientation of the spatial variability gradient would have
encouraged and enabled a much wider trading network with ac-
cess to much more variable precipitation regimes than the radius
reasonably exploitable by foot. The previously predicted increased
level of social complexity would also facilitate this type of spatial
averaging. Along with the increasing level of complexity the selec-
tive pressure will be acting (at least partially) on aggregate cultural
groups.
Spatial variation makes a dramatic return during the period AD
1351–1400. The southern portion of the region still had the luxury
of sufficient moisture, while the northern portion experienced a
deficit (Fig. 10d). This deficit is focused on southwestern Ohio.
The western portion of the Ohio Valley saw a decrease in temporal
variability during this time (Fig. 10e and f). Eastern Indiana, central
and eastern Kentucky, and western and central Ohio saw signifi-
cant increases in spatial variability during this time period. This
was a re-application of the pressure for exchange with a simulta-
neous decrease in the pressure for defense (Fig. 10f). Interestingly
the steep gradient in spatial variability is now oriented primarily
north–south (Fig. 10d). This would encourage trade perpendicular
to the Ohio River in addition to along its course. We predict
increasing homogenization along this new north–south gradient
leading to increased stylistic homogenization across the entire Fort
Ancient area. This does not mean that local styles will cease to ex-
ist, but the spatial distributions of certain motifs are predicted to
have increased.
This leads us to what has been called the Madisonville Horizon
(Drooker, 1997; Essenpreis, 1988; Henderson, 1992). The predic-
tions of our model largely agree with previous interpretations of
this phenomenon and its timing. The pieces of the cultural package
associated with the Madisonville Horizon would develop at differ-
ent times in our model culminating in their recognition as an
aggregate package in the Late Fort Ancient (post ca. AD 1400–
1450) period (e.g., Cowan, 1986; Graybill, 1981; Pollack and Hen-
derson, 1992).
Our model has geographic implications for the development of
the Madisonville Horizon. The initial signs of the homogenization
of material culture should be found in southwest Ohio and
north-central Kentucky (c.f. Prufer and Shane, 1970). We still ex-
pect north–south differences in the distributions of stylistic vari-
ability since for most of the analytical period (AD 801–1400)
little was gained by trading in a north–south direction. It is only to-
wards the end of the AD 1351–1400 period that we expect to see
north and south styles begin to convergence into a region-wide
Horizon.
General fit with extant data
Examination of Table 1 reveals that the extant evidence is lar-
gely in agreement with the predictions outlined above. First, on a
region-wide scale, there is limited stylistic elaboration during the
late Late Woodland and Early Fort Ancient periods with markedly
increasing occurrence and variation of ceramic decoration in the
Middle period. As predicted by the model populations in the Scioto
(cO) and Miami (eI) valleys would be interacting with each other as
evidenced by use of shared symbolic style (Anderson/Baum wares)
with the ‘‘Eastern Periphery” of the Fort Ancient area and beyond
exhibiting different styles (e.g., Philo). Second, the earliest and
most obvious evidence of interaction with Middle Mississippian
populations of the Lower Ohio and Mississippi Valleys is focused
in southwestern Ohio (Miami, eI). Third, the increasing frequency
of large, organized villages in all portions of the Fort Ancient re-
gion, especially by the end of the Middle period, is predicted by
our model. Finally, the region-wide style Horizon (Madisonville
Horizon) and expansion of interaction spheres is predicted in our
final analytical period by a shift in the orientation of spatial vari-
ability/correlation.
Summary and discussion
We have presented a model of cultural response to changes in
precipitation for a series of 50-year periods covering 600 years of
prehistory. We argue that the model developed by Winterhalder
(1986) and modified by Kelly (1995: Chapter 5) is a useful heuristic
device for predicting cultural responses to changing environmental
conditions. This allows us to predict and investigate deviation from
the norms implied by the phases so prevalent in current interpre-
tation in the region. There has been a recent shift in this direction
for Ohio Valley Late Prehistoric studies (Cook, 2008; Hart et al.,
2005; Means, 2007), but variation-obscuring phases still dominate
most current interpretations in the literature. The database does
not currently exist to fully test our model, but the major patterns
of the most popular cultural–historic sequence (Pollack and Hen-
derson, 1992) are largely duplicated (see Table 1). For example,
settlement patterning constituted of smaller sites from the late
Late Woodland (AD 800–1000) into the Early Fort Ancient period
(AD 1000–1200) fit with sufficient moisture availability and a his-
tory of low spatial and moderate temporal variability. The estab-
lishment of social complexity occurs first in southwest Ohio at a
time of moisture stress, again adhering to the expectations of our
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
model. Finally, the previously classified regional homogenization
(Madisonville Horizon) is predicted by a several hundred-year his-
tory of high spatial variability and decreasing moisture availability
along with a shift in the orientation of spatial variability. However,
the specifics of how regional homogenization appears by AD 1400
are distinct in our scenario. One reviewer pointed out that some
current interpretations place the stylistic homogenization of the
Madisonville Horizon later in time than we predict here. This raises
two important issues. First, the current interpretations are based
on typological phases that ignore variability (e.g., Essenpreis,
1978). If our model is valid and a materialist perspective on cul-
tural variability is followed, the transition to a region-wide Horizon
will be via an increasing frequency of the use of certain symbols in
regions outside their origin symbolizing spreading and more regu-
lar and more formalized trade contacts. In addition to the averag-
ing effect of the current systematics, the method used to assign
dates to occupations includes an inherent tendency towards aver-
aging and may not correspond directly with the behavior of inter-
est (see Feathers (2009)). Second, our model makes specific
predictions that differ from current wisdom. This means that our
predictions can be tested against new evidence, or reanalysis of ex-
tant evidence. What is needed are many more dates directly asso-
ciated with the ceramics or trade items of interest. To successfully
do this will require that the investigator(s) take the variation in the
data seriously (see Truncer (2006)). Additionally, direct dates from
the artifacts of interest may be required to fully detangle the com-
plex history of interaction in the Middle Ohio River Valley (again,
see Feathers (2009)). Additionally, our model predicts regional var-
iation in the timing of development towards ‘‘Mississippification”
and the development of the Madisonville Horizon. The model pre-
sented here agrees with much current archaeological interpreta-
tion of the Late Prehistoric period in the Ohio Valley. However,
much additional work is necessary to evaluate the specific predic-
tions of the model, especially where there is disagreement be-
tween our predictions and prevailing interpretations.
Our focus was primarily on the Middle Ohio Valley; however, we
included PDSI grid points beyond this region for two reasons. First,
extending beyond our analytical region was necessary for the inter-
polations to capture spatial trends in moisture availability. Second,
we wanted to make the interpolated moisture surfaces available to
researchers in other regions to evaluate the model predictions
against the data for their regions of interest. Obviously, the nature
of temporal variability of precipitation west of our analytic region
could have played a significant role in the development and contin-
uation of social complexity in the Middle Mississippian heartland.
The patterns predicted by our model have been outlined by the
use of culture–historical phases. Renewed investigation of the sty-
listic variability, settlement structure, and settlement patterns
through time and over space are critical to testing the predictions
of this model; however, the current phases cannot accomplish this.
It has been found in other regions that there is some (first approx-
imation) correspondence between culture–historical types and
patterns of interaction (Lipo et al., 1997), although the inconsistent
(and implicit) methods of their construction limit their usefulness
(Lipo, 2001). What is needed is an explicit, paradigmatic attribute
or attribute-cluster approach (Dunnell, 1971; O’Brien and Lyman,
2000, 2002; Purtill, 2008). That is, we need to limit our classes to
formal definitions and then track form over space and through
time to determine the history of interaction among discrete com-
munities. The current practice of defining units of space–time does
not allow the investigation of subtle change and variability in dis-
tribution and timing. The use of culture–historical phases pre-
cludes an understanding of the evolution of culture in any
meaningful way. There is much work to be done in the Fort Ancient
region to temporally delineate the patterns of behavior alluded to
in this paper.
77
Acknowledgments
The concept for this paper was initially conceived in Kristen
Gremillion’s seminar on Paleodiet. The senior author extends his
thanks to Gremillion for allowing him to explore these topics in
his term paper. This paper has benefited greatly from comments
by Gremillion and four anonymous reviewers. Any remaining
omissions, errors, or logical faults are of course the responsibility
of the authors.
References
Benson, Larry V., Peterson, Kenneth, Stein, John, 2007. Anasazi (Pre-Columbian
Native-American) migrations during the middle-12th and late-13th centuries –
Were they drought induced? Climatic Change 83, 187–213.
Boyd, Robert, Richerson, Peter J., 1985. Culture and the Evolutionary Process.
University of Chicago Press, Chicago.
Brady-Rawlins, Kathleen, 2007. The O.C. Voss Site: Reassessing What We Know
About the Fort Ancient Occupation of the central Scioto Drainage and its
Tributaries. Ph.D. Dissertation, Anthropology, The Ohio State University.
Braun, David P., 1991. Why decorate a pot? Midwestern household pottery, 200 BC–
AD 600. Journal of Anthropological Archaeology 10 (4), 360–397.
Brose, David.S., 1982. The Archaeological investigation of a Fort Ancient community
near Ohio Brush Creek, Adams County, Ohio. Kirtlandia 34, 1–69.
Brose, David.S., White, Nancy.M., 1983. Recent data on Fort Ancient occupation in
the Caesar Creek Valley, southwestern Ohio. West Virginia Archeologist 35, 3–
26.
Carskadden, Jeff, Morton, James, 1977. The Richards Site and Philo Phase of the Fort
Ancient Tradition. The Muskingum Valley Archaeological Survey.
Carskadden, Jeff, Morton, James, 2000. Fort ancient in the central Muskingum Valley
of eastern Ohio: A view from the Philo II site. In: Genheimer, R.A. (Ed.), Cultures
before Contact: The Late Prehistory of Ohio and Surrounding Regions. Ohio
Archaeological Council, Columbus, pp. 158–193.
Church, Flora, 1987. An Inquiry into the Transition from Late Woodland to Late
Prehistoric Cultures in the central Scioto Valley, Ohio, circa A.D. 500 to A.D.
1250. Ph.D. Dissertation, Anthropology, The Ohio State University.
Church, Flora, Nass, John P., 2002. Central Ohio Valley during the late prehistoric:
subsistence-settlement systems’ response to risk. In: Hart, J.P., Reith, C. (Eds.),
Northeast Subsistence-settlement Change, A.D. 700–A.D. 1300, New York State
Museum Bulletin, 496. The University of the State of New York, Albany, pp. 11–
42.
Cobb, Charles R., 2005. Archaeology and the ‘‘Savage Slot”: displacement and
emplacement in the premodern world. American Anthropologist 107, 563–574.
Cook, Edward R., Meko, David M., Stahle, David W., Cleaveland, Malcolm K., 1999.
Drought reconstructions for the continental United States. Journal of Climate
12, 1145–1162.
Cook, Edward R., Woodhouse, Connie A., Mark Eakin, C., Meko, David M., Stahle,
David W., 2004. Long-term aridity changes in the western United States. Science
306, 1015–1018.
Cook, Robert.A., 2007. Single component sites with long sequences of radiocarbon
dates: The SunWatch site and Middle Fort Ancient village growth. American
Antiquity 72, 439–460.
Cook, Robert.A., 2008. SunWatch: Fort Ancient Development in the Mississippian
World. University of Alabama Press, Tuscaloosa.
Cook, Robert A., Fargher, Lane F., 2008. The incorporation of Mississippian traditions
into Fort Ancient societies: a preliminary view of the shift to shell-tempered
pottery use in the middle Ohio Valley. Southeastern Archaeology 27, 222–237.
Cook, Robert A., Schurr, Mark, 2009. Eating between the lines: Mississippian
migration and stable carbon isotope variation in Fort Ancient populations.
American Anthropologist 111, 344–359.
Cooper, Judith Rose, 2008. Bison Hunting and Late Prehistoric Human Subsistance
Economies in the Great Plains. Ph.D. Dissertation, Anthropology, Southern
Methodist University.
Cowan, C. Wesley, 1986. Fort Ancient Chronology and Settlement Evaluation in the
Great Miami River Valley. Volume II: Excavation and Chronology. Report
Submitted to the Ohio Historic Preservation Office by the Cincinnati Museum of
Natural History.
Cowan, C. Wesley, 1987. First Farmers of the Middle Ohio Valley. Cincinnati
Museum of Natural History, Cincinnati, Ohio.
Cowan, C. Wesley, 1988. From Pithouse to Longhouse, from Community to Chaos:
Late Prehistoric and Proto-historic Developments in the Middle Ohio Valley.
Paper presented at the Midwest Archaeological Conference, Urbana-
Champaign, Illinois.
Crumley, Carole L., 1979. Three locational models: an epistemological assessment of
anthropology and archaeology. Advances in Archaeological Method and Theory
2, 141–173.
Drooker, Penelope Ballard, 1997. The View from Madisonville. University of
Michigan, Ann Arbor.
Drooker, Penelope Ballard, 2000. Madisonville focus revisited: reexcavating fort
ancient from museum collections. In: Genheimer, R. (Ed.), Cultures before
contact: The Late Prehistory of Ohio and Surrounding Regions. The Ohio
Archaeological Council, Columbus, Ohio, pp. 228–270.
 Author's personal copy
78 K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
Dunnell, Robert C., 1971. Systematics in Prehistory. The Free Press, New York.
Dunnell, Robert C., 1978. Style and function: a fundamental dichotomy. American
Antiquity 43, 192–202.
Dunnell, Robert.C., 1996a. Aspects of the application of evolutionary theory in
archaeology. In: O’Brien, M.J. (Ed.), Evolutionary Archaeology: Theory and
Application. University of Utah Press, pp. 86–97 (1989).
Dunnell, Robert C., 1996b. Evolutionary theory and archaeology. In: O’Brien, M.J.
(Ed.), Evolutionary Archaeology: Theory and Application. University of Utah
Press, pp. 30–67 (1980).
Dunnell, Robert C., 1996c. Science, social science, and common sense: the agonizing
dilemma of modern archaeology. In: O’Brien, M.J. (Ed.), Evolutionary
Archaeology: Theory and Application. University of Utah Press, pp. 68–85 (1982).
Dunnell, Robert C., 1999. The concept of waste in an evolutionary archaeology.
Journal of Anthropological Archaeology 18, 243–250.
Dunnell, Robert C., Greenlee, Diana M., 1999. Late woodland period ‘‘Waste” in the
Ohio River Valley. Journal of Anthropological Archaeology 18, 376–395.
Dyson-Hudson, Rada, Smith, Eric Alden, 1978. Human territoriality: an ecological
reassessment. American Anthropologist 80, 21–41.
Essenpreis, Patricia S., 1978. Fort ancient settlement: differential response at a
Mississippian–Late Woodland interface. In: Smith, B.D. (Ed.), Mississippian
Settlement Patterns. Academic Press, New York, pp. 141–168.
Essenpreis, Patricia S., 1982. The Anderson Village Site: Redefining the Anderson
Phase of the Fort Ancient Tradition of the Middle Ohio Valley. Ph.D. Dissertation,
Anthropology, Harvard University.
Essenpreis, Patricia S., 1988. An introduction to the Fort Ancient cultural complexes
of the Middle Ohio Valley. In: Heilman, J.M., Lileas, M.C., Turnbow, C.A. (Eds.), A
History of 17 Years of Excavation and Reconstruction: A Chronicle of 12th
Century Human Values and the Built Environment. Dayton Museum of Natural
History, Dayton, Ohio, pp. 1–22.
Feathers, James K., 2009. Problems of ceramic chronology in the southeast: does
shell-tempered pottery appear earlier than we think? American Antiquity 74,
113–142.
Graybill, Jeffrey R., 1981. The eastern Periphery of Fort Ancient (A.D. 1050–1650): A
Diachronic Approach to Settlement Variability. Ph.D. Dissertation,
Anthropology, University of Washington.
Greenlee, Diana M., 2002. Accounting for Subsistence Variation among Maize
Farmers in Ohio Valley Prehistory. Ph.D. Dissertation, Anthropology, University
of Washington.
Griffin, James B., 1966. The Fort Ancient Aspect: Its Cultural and Chronological
Position in Mississippi Valley Archaeology. Anthropological Papers No. 28.
University of Michigan, Museum of Anthropology, Ann Arbor (1943).
Halstead, Paul, O’Shea, John, 1989. Introduction: cultural responses to risk and
uncertainty. In: Halstead, P., O’Shea, J. (Eds.), Bad Year Economics: Cultural
Responses to Risk and Uncertainty. New Directions in Archaeology Series. Press
Syndicate, University of Cambridge, New York, pp. 1–7.
Hanson Jr., Lee H., 1966. In: The Hardin Village Site. Studies in Anthropology, vol. 4.
University of Kentucky Press, Lexington.
Hanson Jr., Lee H., 1975. The Buffalo Site, a Late Seventeenth Century Indian Village
Site (46 Pu 31) in Putnam County, west Virginia. Report of Archaeological
Investigations 5. West Virginia Geological Survey, Morgantown.
Hart, John P., 1999. Maize agriculture evolution in the eastern Woodlands of North
America: A Darwinian perspective. Journal of Archaeological Method and
Theory 6, 137–180.
Hart, John P., Nass, John P., Means, Bernard K., 2005. Monongahela subsistence-
settlement change. Midcontinental Journal of Archaeology 30, 327–365.
Henderson, A. Gwynn (Ed.), 1992. Fort Ancient Cultural Dynamics in the Middle
Ohio Valley. Monographs in World Archaeology No. 8. Prehistory Press,
Madison, Wisconsin.
Henderson, A. Gwynn, 1998. Middle Fort Ancient Villages and Organizational
Complexity in Kentucky. Ph.D. Dissertation, Anthropology, University of
Kentucky.
Holmes, William H., 1886. Ancient pottery of the Mississippi Valley. In: Bureau of
American Ethnology 4th Annual Report, 1882–1883. Government Printing
Office, Washington, DC, pp. 367–436.
Johnson, Allen W., Earle, Timothy, 1987. The Evolution of Human Societies: From
Foraging Group to Agrarian State. Stanford University Press, Stanford, California.
Kaplan, Hillard, Hill, Kim, Hurtado, Magdelena, 1990. Risk, foraging and food
sharing among the ache. In: Cashdan, E. (Ed.), Risk and Uncertainty in Tribal and
Peasant Economies. Westview Press, Boulder, pp. 107–143.
Keegan, William F., 1986. The optimal foraging analysis of horticultural production.
American Anthropologist 88, 92–107.
Keegan, William F., Butler, Brian M., 1987. The microeconomic logic of horticultural
intensification in the eastern Woodlands. In: Keegan, W.F. (Ed.), Emergent
Horticultural Economies of the eastern Woodlands. Center for Archaeological
Investigations, Southern Illinois University at Carbondale Occasional Paper No.
7. Board of Trustees, Southern Illinois University, Carbondale, pp. 109–127.
Kelly, J.E., 1990. Range site community patterns and the Mississippian emergence.
In: Smith, B.D. (Ed.), The Mississippian Emergence. Smithsonian Institution
Press, Washington, DC, pp. 67–112.
Kelly, Robert L., 1995. The Foraging Spectrum: Diversity in Hunter–Gatherer
Lifeways. The Smithsonian Institution.
Kennedy, William E., 2000. Interpreting Fort Ancient Settlement Variability.
Master’s Thesis, Anthropology, Kent State University.
Larson, Daniel O., Neff, Hector, Graybill, Donald A., Michaelsen, Joel, Ambos,
Elizabeth, 1996. Risk, climatic variability, and the study of southwestern
prehistory: an evolutionary perspective. American Antiquity 61, 217–241.
Lipo, Carl P., 2001. Community structures among Late Mississippian populations of
the central Mississippi River Valley. In: Hunt, T.L., Lipo, C.P., Sterling, S.L. (Eds.),
Posing Questions for a Scientific Archaeology. Bergin and Garvey, Westport,
Connecticut, pp. 175–216.
Lipo, Carl P., Madsen, Mark E., Dunnell, Robert C., Hunt, Tim, 1997. Population
structure, cultural transmission, and frequency seriation. Journal of
Anthropological Archaeology 16, 301–333.
Lyman, R.Lee, O’Brien, Michael J., 1998. The goals of evolutionary archaeology:
history and explanation. Current Anthropology 39, 615–652.
Madsen, Mark, Lipo, Carl, Cannon, Michael, 1999. Fitness and reproductive trade-
offs in uncertain environments: explaining the evolution of cultural elaboration.
Journal of Anthropological Archaeology 18, 251–281.
Means, Bernard K., 2007. Circular Villages of the Monongahela Tradition. The
University Alabama Press, Tuscaloosa.
Mills, William C., 1906. Baum Prehistoric Village, 15. Ohio State Archaeological and
Historical Society Publications, pp. 45–136.
Mika, J., Sz. Horvath, L. Makra, Z. Dunkel, 2005. Palmer Drought Severity Index
(PDSI) as an indicator of soil moisture. Physics and Chemistry of the Earth 30,
223–230.w.
O’Brien, Michael J., Lyman, R.Lee, 2000. Applying Evolutionary Archaeology: A
Systematic Approach. Kluwer Academic with Plenum Publishers, New York.
O’Brien, Michael J., Lyman, R.Lee, 2002. Evolutionary archeology: current status and
future prospects. Evolutionary Anthropology 11, 26–36.
Pauketat, Timothy, 2003. Resettled farmers and the making of a Mississippian
polity. American Antiquity 68, 39–66.
Pickard, W.H., 1996. 1990 excavations at Capitolium Mound (33Wn13), Marietta,
Washington County, Ohio: a working hypothesis. In: Pacheco, P.J. (Ed.), A View
from the Core, A Synthesis of Ohio Hopewell Archaeology. Ohio Archaeological
Council, Columbus, pp. 274–285.
Pollack, David, Henderson, A. Gwynn, 1992. Towards a model of Fort Ancient
society. In: Henderson, A.G. (Ed.), Fort Ancient Cultural Dynamics in the Middle
Ohio River Valley. Monographs in World Archaeology No. 8. Prehistory Press,
Madison, Wisconsin, pp. 281–294.
Pollack, David, Henderson, A. Gwynn, 2000a. Insights into Fort Ancient culture
change: a view from south of the Ohio River. In: Genheimer, R.A. (Ed.), Cultures
before Contact: The Late Prehistory of Ohio and Surrounding Regions. Ohio
Archaeological Council, Columbus, pp. 194–227.
Pollack, David, Henderson, A. Gwynn, 2000b. Late woodland cultures in Kentucky.
In: Emerson, T.E., McElrath, E.L., Fortier, A.C. (Eds.), Late Woodland Societies:
Tradition and Transformation across the Midcontinent. University of Nebraska
Press, Lincoln, Nebraska, pp. 613–642.
Pollack, D., Henderson, A. Gwynn, Begley, C.T., 2002. Fort Ancient/Mississippian
Interaction on the northeastern periphery. Southeastern Archaeology 21, 206–
220.
Prufer, Olaf H., Shane, Orrin C., 1970. Blain Village and the Fort Ancient Tradition in
Ohio. Kent State University Press, Kent, Ohio.
Purtill, Matthew P., 1999. Evidence for a Late Fort Ancient Fall/Winter occupation in
Southwestern Ohio. North American Archaeologist 20, 105–133.
Purtill, Matthew P., 2008. Early woodland plain-surface pottery from the Mid-Ohio
Valley: two recently excavated assemblages from Ohio and Kentucky.
Midcontinental Journal of Archaeology 33, 41–71.
Rafferty, Janet E., 1974. The Development of the Fort Ancient Tradition in northern
Kentucky. Ph.D. Dissertation, Anthropology, University of Washington.
Redmond, Brian.G., McCullough, Robert.G., 2000. The late woodland to late
prehistoric occupations of central Indiana. In: Emerson, T.E., McElrath, E.L.,
Fortier, A.C. (Eds.), Late Woodland Societies: Tradition and Transformation
across the Midcontinent. University of Nebraska Press, Lincoln, Nebraska, pp.
643–684.
Reidhead, Van A., 1981. A Linear Programming Model of Prehistoric Subsistence
Optimization: A Southeastern Indiana Example. Prehistory Press Research
Series VI No. 1. Indiana Historical Society, Indianapolis.
Renfrew, C., 1986. Introduction. In: Renfrew, C., Cherry, J.F. (Eds.), Peer–Polity
Interaction and Socio-political Change. Cambridge University Press, Cambridge,
pp. 1–18.
Richerson, Peter J., Boyd, Robert, 2005. Not by Genes Alone: How Culture
Transformed Human Evolution. The University of Chicago Press, Chicago.
Richerson, Peter J., Boyd, Robert, Bettinger, Robert L., 2001. Was agriculture
impossible during the pleistocene but mandatory during the holocene? A
climate change hypothesis. American Antiquity 66, 387–411.
Rindos, David, 1980. Symbiosis, instability, and the origins and spread of
agriculture: a new model. Current Anthropology 21, 751–772.
Rindos, David, 1984. The Origins of Agriculture: An Evolutionary Perspective.
Academic Press, Inc., Orlando.
Robertson, J.A., 1980. Chipped Stone and Socio-cultural Interpretations. Master’s
Thesis, Anthropology, University of Illinois at Chicago Circle.
Schambach, Frank F., 1971. Review of blain village and the Fort Ancient tradition in
Ohio by Olaf H. Prufer and Orrin C. Shane, III. American Anthropologist 73,
1402–1404.
Seeman, Mark F., Dancey, William S., 2000. The late woodland period in southern
Ohio: basic issues and prospects. In: Emerson, T.E., McElrath, E.L., Fortier, A.C.
(Eds.), Late Woodland Societies: Tradition and Transformation across the
Midcontinent. University of Nebraska Press, Lincoln, Nebraska, pp. 583–612.
Shane, Linda C.K., Snyder, G.G., Anderson, K.H., 2001. Holocene vegetational and
climatic changes in the Ohio region. In: Prufer, O.H., Pedde, S.E., Meindl, R.S.
(Eds.), Archaic Transitions in Ohio and Kentucky Prehistory. Kent State
University Press, Kent, Ohio, pp. 11–55.
 Author's personal copy
K.C. Nolan, R.A. Cook / Journal of Anthropological Archaeology 29 (2010) 62–79
Soltis, Joseph, Boyd, Robert, Richerson, Peter J., 1995. Can group-functional
behaviors evolve by cultural group selection?: An empirical test. Current
Anthropology 16, 473–494.
Stahle, David W., Fye, Falko K., Cook, Edward R., Daniel Griffin, R., 2007. Tree-ring
reconstructed megadroughts over North America since A.D. 1300. Climatic
Change 83, 133–149.
Trigger, B.G., 2006. A History of Archaeological Thought, second ed. Cambridge
University Press, Cambridge.
Truncer, James, 2006. Taking variation seriously: the case of steatite vessel
manufacture. American Antiquity 71, 157–163.
Turnbow, Christopher A., Jobe, Cynthia E., 1984. The Goolman site: a Late Fort
Ancient winter encampment in Clark County, Kentucky. In: Pollack, D.,
Hockensmith, C., Sanders, T. (Eds.), Late Prehistoric Research in Kentucy.
Kentucky Heritage Council, pp. 25–49.
Turnbow, Christopher A., Sharp, William, 1988. Muir: an early fort Ancient site in
the inner bluegrass. In: Program for Cultural Resource Assessment, University of
Kentucky, Archaeological Research Report No. 165, Lexington.
Turnbow, Christopher A., Henderson, A. Gwynn, 1992. Appendix: ceramis and other
baked clay objects. In: Henderson, A.G. (Ed.), Fort Ancient Cultural Dynamics in
79
the Middle Ohio Valley. Monographs in World Archaeology No 8. Prehistory
Press, Madison, Wisconsin, pp. 113–136.
Vickery, Kent D., Sunderhaus, Theodore S., Genheimer, Robert A., 2000. Preliminary
report on excavations at the Fort Ancient State Line site, 33 Ha 58, in the central
Ohio Valley. In: Genheimer, R.A. (Ed.), Cultures before Contact: The Late
Prehistory of Ohio and Surrounding Regions. Ohio Archaeological Council,
Columbus, pp. 272–328.
Weissner, Polly., 1983. Style and social information in Kalahari San projectile points.
American Antiquity 48, 235–276.
Winterhalder, Bruce, 1986. Diet choice, risk, and food sharing in a stochastic
environment. Journal of Anthropological Archaeology 5, 369–392.
Winterhalder, Bruce, 1990. Open field, common pot: harvest variability and risk
avoidance in agricultural and foraging societies. In: Cashdan, E. (Ed.), Risk and
Uncertainty in Tribal and Peasant Economies. Westview Press, Boulder,
Colorado, pp. 67–87.
Winterhalder, Bruce, Goland, Carol, 1997. An evolutionary ecology perspective on
diet choice, risk, and plant domestication. In: Gremillion, K.J. (Ed.), People Plants
and Landscapes Studies in Paleoethnobotany. University of Alabama Press,
Tuscaloosa, pp. 123–160.