Multilevel Selection and Political Evolution in the Valley of Oaxaca, 500-100 B.C.
Transcript of Multilevel Selection and Political Evolution in the Valley of Oaxaca, 500-100 B.C.
Multilevel Selection and Political Evolution in the Valley of Oaxaca,500–100 B.C.
Charles S. Spencer and Elsa M. Redmond
Division of Anthropology, American Museum of Natural History, Central Park West at 79th Street, New York, New York 10024-5192
E-mail: [email protected]
Received March 3, 2000; revision received June 30, 2000; accepted August 8, 2000
Journal of Anthropological Archaeology 20, 195–229 (2001)
doi:10.1006/jaar.2000.0371, available online at http://www.idealibrary.com on
Although Monte Albán I in the Valley of Oaxaca (500–100 B.C.) is widely recognized as a period
of major political change, researchers have found it difficult to establish whether the key institu-
tions of the Zapotec state emerged during this or the succeeding Monte Albán II period (100
B.C.–A.D. 200). Also unresolved has been the issue of when the three major subregions of the Oax-
aca Valley (Etla, Tlacolula, Ocotlán/Zimatlán) all became integrated into a single polity under the
rule of Monte Albán, the state capital. This paper presents recent theoretical and empirical contri-
butions that have not yet been brought to bear on the problem of Monte Albán I. Concepts drawn
from multilevel selection theory and evolutionary trend theory are utilized in an analysis of Oaxaca
Valley regional settlement pattern data. The analysis provides a multilevel context for a discussion
of recent survey and excavations at San Martín Tilcajete, the results of which are clarifying the se-
quence of institutional development in Oaxaca. Taken together, the regional analysis and the dis-
coveries at Tilcajete indicate that: (1) the Zapotec state emerged during Late Monte Albán I (300–100
B.C.) in a context of intensifying competition—including violent conflict—among rival polities
within the Oaxaca Valley; and (2) even though the early Zapotec state began a campaign of territo-
rial expansion during Late Monte Albán I, political unification of all three major subregions of the
Valley was not achieved until Monte Albán II. © 2001 Academic Press
Key Words: Oaxaca; Zapotec; state; political evolution.
INTRODUCTION
The prehistoric period known as Monte
Albán I (500–100 B.C.)—hereafter MA I—is
recognized by Oaxaca scholars as a time of
dramatic change (Blanton 1978; Blanton et
al. 1982, 1999; Flannery and Marcus 1990;
Kowalewski et al. 1989; Marcus and Flan-
nery 1996). Often divided into Early MA I
(500–300 B.C., a.k.a. the Monte Albán Ia
phase) and Late MA I (300–100 B.C., a.k.a.
the Monte Albán Ic phase), MA I witnessed a
rapid increase in human population over the
previous Rosario phase (700–500 B.C.) as
well as the development of a more complex
regional settlement hierarchy, at the top of
which was the large center of Monte Albán
(Kowalewski et al. 1989). During the Rosario
phase, Monte Albán had not yet been estab-
lished, and most researchers concur that the
19
three major subregions of the Oaxaca Valley
(Etla, Tlacolula, and Ocotlán/Zimatlán)
were not politically unified but instead were
occupied by at least three independent
chiefly polities (Blanton et al. 1993: 66–69;
Marcus and Flannery 1996: 123–126).
On the other hand, by Monte Albán II (100
B.C.–A.D. 200), it is generally agreed that all
subregions of the Oaxaca Valley had been in-
tegrated into a single polity, the early Za-
potec state, under the rule of Monte Albán
(Marcus and Flannery 1996:172–178). By MA
II, assert Flannery and Marcus (1983:80), the
evidence for statehood in Oaxaca is “clear
and unmistakable.” For example, there is ev-
idence of a four-level regional administra-
tive hierarchy by MA II (Flannery and Mar-
cus 1983:82). MA II also sees “the
appearance of a whole series of clearly rec-
ognizable and functionally distinct public
50278-4165/01 $35.00Copyright © 2001 by Academic Press
All rights of reproduction in any form reserved.
doing so, however, it would be helpful to re-
buildings” (Flannery and Marcus 1976:221),
including the multiroom temple, the royal
palace, the ballcourt, buildings with a spe-
cial military purpose, and others. Flannery
and Marcus (1976: 221) point out that “the
activities carried on in these buildings must
have been very different, presumably reflect-
ing different sociopolitical institutions and
different sets of personnel.” Such institu-
tional diversity is a hallmark of the state as
defined by Wright (1977: 383), “a cultural de-
velopment with a centralized decision-mak-
ing process which is both externally special-
ized with regard to the local processes which
it regulates, and internally specialized in
that the central process is divisible into sepa-
rate activities which can be performed at dif-
ferent places at different times.” By contrast,
a chiefdom is a cultural development with
an administrative organization that is exter-
nally specialized (or centralized) but at the
same time is not internally specialized
(Wright 1977:381). Spencer (1990, 1998) has
argued that, because the internally special-
ized administration of the state is compatible
with the effective delegation of authority, the
state can usually integrate a much larger ter-
ritory than a chiefdom.
Flannery and Marcus (1976:217, 1983:80)
place considerable emphasis on the fact
that the Zapotec royal palace had appeared
by MA II, citing Sanders’s (1974:109) view
that only state rulers have sufficient author-
ity to call upon corvée labor to construct
their own residences. Noting that the
palace and other state institutions seem
clearly in place by MA II, but absent in the
Rosario phase, Flannery and Marcus sug-
gest that this “throws the spotlight on
Monte Albán I” as the “crucial” period for
the rise of the Zapotec state (Flannery and
Marcus 1983:80). Taking a similar position,
Blanton et al. (1999) refer to the changes of
MA I as “the great transformation” in the
Oaxaca Valley. Their regional survey data
196 SPENCER AN
document the patterns of change in popula-
tion size, settlement hierarchies, civic-cere-
monial architecture, and artifact distribu-
tions over the course of MA I; all in all, the
trends “add up to a significant episode of
sociocultural evolution” (1999:68).
Although the past three decades of re-
search have revealed the importance of MA
I, it is fair to say that some significant issues
remain unresolved, in part because rela-
tively few sites of the time period have been
extensively excavated. As Marcus and Flan-
nery (1996:160–161) point out, this lack of
data means that we are still uncertain when
the key institutions of the early Zapotec state
first appeared—was it in Early MA I, Late
MA I, or not until MA II? They also point out
that we do not know precisely when all three
branches of the Oaxaca Valley became politi-
cally unified under the leadership of Monte
Albán (Marcus and Flannery 1996:162–163).
And, while it has been demonstrated that
Monte Albán was involved in extravalley
conquests as early as Late MA I (Redmond
1983; Spencer 1982; Spencer and Redmond
1997), it has been unclear whether the rise of
the Zapotec state was associated with con-
flicts between the Monte Albán polity and
other polities within the Valley of Oaxaca.
In this paper we discuss some recent theo-
retical and empirical contributions that have
not yet been brought to bear on the problem
of MA I development. We first draw upon
recent theoretical work in biology and pale-
ontology concerning multilevel selection
and evolutionary trends in order to address
questions about MA I regional integration
(Heisler and Damuth 1987; McShea 1994;
Wilson and Sober 1994). We then present
new data from ongoing excavations at San
Martín Tilcajete that are clarifying the se-
quence of institutional development within
the Valley of Oaxaca during MA I. Before
D REDMOND
view the evolutionary perspective that un-
derlies the entire paper.
PROCESSUAL EVOLUTIONISM
Over the past two decades, there has
been a notable upsurge of interest in the use
T
of evolutionary models in archaeology
(Barton and Clark 1988; Dunnell 1980, 1989;
Lyman and O’Brien 1998; Rindos 1984,
1989; Rosenberg 1994; Spencer 1997; Teltser
1995). Although unanimity of viewpoints is
lacking among evolutionary researchers,
there is substantial agreement on a few
basic concepts. Widely acknowledged is the
critical role played by selection in processes
of cultural change and stability, as well as
the central importance of variability, the
stuff on which selection acts. Most would
agree that, at any point in time, humans
generate an impressive array of variability
in their cultural behaviors. When the vari-
ant forms come into competition with one
another, it is selection that decides which
variants will persist. In Eldredge’s (1995:34)
apt phraseology, selection is the outcome of
“what worked better than what.”
Far less agreement can be discerned,
however, surrounding the issue of whether
initial variation is always undirected, or
whether it can be directed through the con-
scious, purposeful, premeditated actions of
human beings. In a previous paper, Spencer
(1997) identified a group of evolutionary ar-
chaeologists he called selectionists, who see
culture as consisting of cultural traits, and
define cultural evolution as comprising
shifts in the relative frequencies of these
traits over time. Equally fundamental to se-
lectionist thought is the view that initial
variation in cultural traits is being con-
stantly generated and, moreover, produced
exclusively in an undirected manner. Selec-
tion then operates on this variation, and the
result is patterns of change or stability over
time. Spencer also discussed another group
of evolutionary researchers whom he la-
beled processualists (or processual evolution-ists). It is appropriate to point out that
processual evolutionism is the perspective
we take in the present paper.
Processual evolutionists prefer to see cul-
POLITICAL EVOLUTION IN
ture not as a collection of traits but as a sys-
tem occupied by human actors who group
themselves into nested sets of organiza-
tional levels and pursue conscious, purpo-
sive strategies intended to further their own
interests. The processualist perspective at-
tributes great importance to human agency,
the capacity of human beings to take an ac-
tive role in shaping many of the variant
forms of their individual and group behav-
iors. The significance assigned to human
agency means that initial variation can often
be directed and not exclusively undirected
as the selectionists would have it. Although
processualists recognize that undirected be-
haviors are far from rare, they nonetheless
insist that the ability of humans to design
and carry out novel strategies is a major fea-
ture that distinguishes cultural evolution
from its biological counterpart (Boyd and
Richerson 1985:81–131). From the processu-
alist perspective, biological and cultural
evolution are seen as analogous, not homol-
ogous or identical, processes. We would
argue, however, that there is little to be
gained by pursuing this analogy as an end
in itself; an analogy takes on value when it
is heuristic, stimulating the researcher to
apply novel analyses and interpretations to
his data (Spencer 1991: 145).
Also integral to processual evolutionism
is a conception of cultural systems as hier-
archically organized, with each level com-
posed of components that are themselves
systems of a lower level, a feature that
Crumley (1995) calls the “scalar hierarchy.”
Examples of such levels in a cultural system
would include the household level (com-
posed of individuals), the community level
(composed of households), the polity level
(composed of communities), and the inter-
polity level (composed of separate polities).
Crumley makes an important distinction
between the “scalar hierarchy” and the
“control hierarchy,” in which one of the
components on a particular level in the
scalar hierarchy exerts control over the oth-
ers. An alternative arrangement is the “het-
HE VALLEY OF OAXACA 197
erarchy,” an arrangement whereby the
components are not permanently ranked
with respect to each other, but where indi-
D
vidual components are capable of short-
lived dominance over others depending on
the circumstances. When we examine cul-
tural sequences, important periods of
change are often marked by shifts from the
heterarchical to the hierarchical mode of or-
ganization.
For example, in the transition from autonomous
villages to a chiefdom, intervillage integration
changes from a situation where authority is di-
vided or shared among component villages (heter-
archy) to a situation where intervillage integration
is achieved through centralized authority concen-
trated in a single village, which emerges as the
paramount center of the new chiefly polity (hierar-
chy). A similar process occurs when a state author-
ity emerges and rules over previously autonomous
chiefdoms. (Spencer 1997: 238–239)
How does a control hierarchy evolve in
the first place? A key mechanism is extrapola-tion, “an extension or projection of the inter-
nal model of authority from one social unit
to others on the same level of the scalar hier-
archy” (Spencer 1997: 239). Extrapolation,
when successfully implemented, brings
about the promotion of the social unit that
pursues it to a higher level of the emerging
control hierarchy: a village is promoted to a
chiefly center, a chiefly center is promoted to
a state capital (Flannery 1972). Whether or
not an extrapolation initiative succeeds, of
198 SPENCER AN
course, depends on what competing strate-
replicate themselves if they help the entire
gies are also being pursued and thus is ulti-
mately determined by selection.
MULTILEVEL SELECTION
An important corollary to the assump-
tion that cultural systems have multiple
levels of organization is the proposition
that selection can operate in a multilevel
fashion. With this statement, we invoke the
controversial issue of whether natural se-
lection acts only on individuals or if it can
also act upon social groups—what is often
called the “group selection” debate. Few
theorists are more closely identified with
this debate than David Sloan Wilson and
Elliott Sober (Sober and Wilson 1998; Wil-
son 1989; Wilson and Sober 1989; 1994),
who favor group selection models, and
George Williams (1966; 1992) and Richard
Dawkins (1976, 1982), who oppose them.
Wilson and Sober (1994:585) note that up
to the mid-1960s it was relatively common
for biologists and social scientists to con-
ceive of social groups as adaptive units, not
unlike individual organisms in the coordina-
tion of the parts that comprised them. Since
adaptation is a product of selection, such a
view implied that selection could operate on
the level of the group as well as the individ-
ual. Adherents to the group selection posi-
tion argued that individuals engage in
numerous behaviors that are less than opti-
mal from the viewpoint of individual fitness,
but nevertheless enhance the fitness of the
social groups to which they belong; the be-
haviors persist because they are favored by
natural selection operating on the level of
the social group (Wynne-Edwards 1962,
1986). Williams attacked the idea of group
selection in his widely read 1966 book, argu-
ing that group-level adaptations came about
not because of selection acting on the level of
the group but rather as a by-product of
many individual adaptations. To Williams,
all such adaptations had to be explained in
terms of the relative fitness of individuals
within populations; although he agreed that
individual-level adaptations may have
group-level effects, he insisted that these are
not relevant to evolutionary change
(Williams 1966: 92; Wilson and Sober 1994:
590). Williams made selection an even more
specific process when he argued that the
gene is really the basic unit of selection, a
conception that was developed and popu-
larized by Dawkins (1976, 1982). Dawkins
(1976) asserted that even though genes are
the fundamental units of selection, they do
not exist apart from individual organisms;
they are “trapped,” so to speak, along with
other genes in an individual and can only
REDMOND
collection survive; thus individuals become
“vehicles of selection.” It is this property of
the levels of the scalar hierarchy. Consider
“shared fate” that induces “selfish genes” to
cooperate to produce successful organisms
(Dawkins 1976).
Since it lay even farther from the gene
than the individual, the social group seemed
a most unlikely unit of adaptation and selec-
tion to many biologists of the late 1960s and
1970s, and the group selection view fell from
favor. Even so, Wilson kept up an unrelent-
ing campaign in favor of group selection ap-
proaches (Wilson 1975, 1976, 1980, 1983,
1989, 1992), joining Lewontin (1970), El-
dredge (1985), Gould (1980, 1988), and oth-
ers who were of the opinion that selection
could operate at times above the level of the
individual (or the gene). Lewontin (1970) ar-
gued that any grouping of individual units,
from molecules to galaxies, can undergo
evolution if heritable behaviors exist that can
differentially affect the group’s ability to re-
produce itself. In a presidential address to
The Paleontological Society, Gould heralded
“the most important contemporary revision
of evolutionary theory—the insight that se-
lection (and other forces) acts simultane-
ously at several levels of a genealogical hier-
archy, with effects propagating to levels
above and below” (1988:319).
The past 10 years have seen an increase in
the popularity of the group selection view-
point. Researchers are beginning to move
beyond statements of faith to the develop-
ment of methodologies for analyzing multi-
level selection processes (Goodnight et al.
1992; Heisler and Daimuth 1987; Seely 1989;
Sober and Wilson 1998; Wilson and Sober
1994). Wilson and Sober (1994) have pro-
posed a procedure that they believe is us-
able in both biological and sociocultural
contexts, under the assumption that both
the biological and sociocultural worlds are
hierarchically structured. An essential step
in their methodology is the identification of
the “vehicles of selection” in whatever bio-
logical or human situation is being exam-
ined, using the concept of “shared fate.” For
POLITICAL EVOLUTION IN
each measured trait of interest, they ana-
lyze each level of what Crumley (1995)
would call the scalar hierarchy to deter-
mine whether there are fitness differences
among the components of that level; such
differences are usually manifested by vari-
able success in the outcomes of biological or
social competition among components, ex-
pressed in terms of the measured trait (Wil-
son and Sober 1994:591). By focusing on ve-
hicles of selection, they “de-couple” the
concept of organism from the individual
level of the biological hierarchy. To illus-
trate, they envision an extreme example, a
situation where the genes within a single
individual differ more in fitness than do the
individuals within a social group. In this
case, they argue, there will be relatively
strong selection on the level of the genes,
and one could view the genes in this in-
stance as “organisms,” while the individual
becomes a “dysfunctional collection of
genes” (Wilson and Sober 1994:591). Alter-
natively, in situations where individuals in
the same group are “in the same boat” with
respect to fitness, they will behave like the
harmonious “organs” that contribute to the
successful functioning of group-level orga-
nizations, which can be seen as the “organ-
isms” in this case. The point of their analy-
sis is that the relative impact of selection
can shift from one level of the scalar hierar-
chy to another: “In short, the organ–organ-
ism–population trichotomy can be frame-
shifted both up and down the biological
hierarchy” (Wilson and Sober 1994:592).
We contend that selection-level “frame-
shifts” are also an important feature of cul-
tural evolution. A frameshift upward in the
scalar hierarchy can be brought about by
extrapolation, through which one unit in a
given level begins to assert authority over
the others, creating a more inclusive politi-
cal entity. Spencer (1997) has argued that
extrapolation is an expression of agency,
and that the extrapolation initiative itself al-
ters the relative impact of selection among
THE VALLEY OF OAXACA 199
the transition from autonomous villages to
chiefdoms. In the autonomous village con-
D
text, selection on the multivillage (or re-
gional) level will be relatively weaker than
intervillage selection. But, when an emerg-
ing leader in one village seeks to extrapo-
late his authority to other villages, this ini-
tiative will increase the relative importance
of selection on the multivillage level—the
newly centralized multivillage polity will
persist only if the emergent regional leader-
ship can successfully withstand challenges
against the polity (Carneiro 1981; Spencer
1987). Now envision the transition from
chiefdom to state. In a context of indepen-
dent chiefdoms, selection is relatively
stronger among the chiefly polities (i.e., on
the interchiefdom level) than it is above or
below that level. But, when the leadership
of one of those chiefdoms seeks to extend
its control over other chiefdoms, the extrap-
olation initiative will change the pattern of
selection throughout the domain affected
by the strategy. If the extrapolation strategy
succeeds, a new and larger entity, a nascent
state composed of several previous chiefly
polities, emerges as a major unit for future
selection (Carneiro 1970; Marcus and Flan-
nery 1996:157; Redmond 1998; Spencer
1998). The emergence of the multichiefdom
regional state does not signal the end of se-
lection on either the subregional or local
level; rather, state emergence introduces an-
other important level of selection along
with those that are already in operation.
Carneiro (1978, 1992) has argued that the
evolution of complex human societies, over
the long term, has been characterized by
200 SPENCER AN
the following general trend: the successive
amples of researchers who have interpreted
emergence of ever more inclusive levels of
integration and concomitant selection.
EVOLUTIONARY TREND THEORY
Although paleontologists generally inter-
pret a persistent, directional change in
some attribute of a lineage or clade as an
evolutionary trend, McShea (1994) has ar-
gued that not all such trends are alike. He
draws a fundamental distinction between
what he calls “passive” and “driven”
trends. In a passive trend, directional
change is not produced by selection favor-
ing the attribute in question, but rather by
stochastic variation and divergence from an
original boundary condition (1994: 1748).
As an example, he presents Stanley’s (1973)
reinterpretation of certain instances of
Cope’s rule, the well-known tendency for
groups to show an increase in average body
size over time. Stanley argued that mean
body size would increase over time, even in
the absence of selection for larger size, if the
founding species of a clade originated near
the lower limits of viable size, or perhaps
near the boundary of its potential range.
Under such conditions, what appears to be
a directional change in some attribute
might instead be the result of stochastic
change in the attribute’s variance, biased in
the direction of larger size, as the descen-
dants of the founding group undergo mul-
tiple cycles of reproduction, and exhibit
variation that increasingly diverges from
the initial limit or boundary condition (see
also Gould 1988).
A driven evolutionary trend, by contrast,
is the product of a pervasive and persistent
bias reinforced by selection (McShea
1994:1748). The directional change in a
given attribute persists not because of sto-
chastic variation in a constrained or
bounded space, but rather because the
change in question offers adaptive advan-
tages to the individuals or groups possess-
ing higher values for that attribute relative
to competing individuals or groups. Mc-
Shea argues that, while some examples of
Cope’s rule are undoubtedly passive
trends, in other cases an increase in body
size is probably driven by the selective ad-
vantages accruing to those individuals and
groups who tend to have larger bodies than
their competitors. He gives Newell (1949),
Bonner (1988), and McKinney (1990) as ex-
REDMOND
body-size evolution as a driven trend,
while Saunders and Ho (1976) and Bonner
(1988) are among those who have argued
that the evolution of complexity may be fa-
vored by natural selection, and thus qualify
as a driven process in McShea’s terms.
Whenever directionality in a particular at-
tribute conveys selective advantage on the
individuals or groups possessing greater
values for the attribute, the resulting trend
may be considered driven. Moreover, be-
cause a driven trend is powered by selec-
tive advantage, the attribute in question can
undergo much more rapid and dramatic
change than would be the case in a passive
system. McShea acknowledges that most
large-scale trends are “probably quite com-
plex, in the sense that they result from a va-
riety of different causes operating at differ-
ent times and over a variety of different
taxonomic, spatial, and temporal scales”
(1994:1751), and he supports the use of em-
pirical tests designed to measure the degree
to which a given evolutionary trend can be
considered driven or passive. A test he has
found to be generally reliable is the skew-
ness or “subclade” test, which examines a
variable such as body size measured on a
subset of individuals drawn from a
parental population with an overall distrib-
ution for the same variable that is positively
skewed. If the distribution of the subset is
also significantly skewed in the positive di-
rection, then the system is probably exhibit-
ing a driven trend. A particular skewness
coefficient is generally considered statisti-
cally significant when the ratio of skewness
to the standard error of skewness is greater
than 2.0 (SPSS 1998:214).
The distinction between passive and dri-
ven trends takes on additional importance
when we connect it to the concept of multi-
level selection. Drawing upon simulation
studies as well as analyses of paleontological
data, McShea concludes that “the signifi-
cance of the passive–driven distinction lies
in the independence among hierarchical lev-
POLITICAL EVOLUTION IN
els implied by the passive mechanism, a
counterintuitive notion for many” (1994:
1761). In a passive trend, there is a lack of
correlation between adaptive fitness on dif-
ferent levels of the evolving system; selec-
tion processes thus operate independently
on each system level. The converse of this
proposition, we propose, is that driven evolu-tionary trends should be associated with interde-pendent relationships among the hierarchical lev-els of evolving systems. In a driven
evolutionary trend, we would expect the se-
lection processes operating on different lev-
els of the system to be correlated (either pos-
itively or negatively) with one another. For
example, in the case of a negative correla-
tion, the fitness of the group would increase
at the expense of individual-level fitness,
such as when a young warrior gives his life
so his village can prevail over others in com-
bat and attain (as a village) greater regional
power and wealth, a classic example of indi-
vidual altruism. On the other hand, in the
case of a positive correlation, the fitness of
the individual and the group are both en-
hanced, as, for example, when the members
of a village cooperate to construct an irriga-
tion system that could not have been built
without such cooperation and whose use
will be of benefit to all. Such interlevel corre-
lations reinforce the selection processes op-
erating on each level, adding strength and
momentum to directional biases and result-
ing, eventually, in evolutionary trends that
are more driven than passive in character.
An application of McShea’s theory of
evolutionary trends requires some method
for analyzing patterns of selection in a mul-
tilevel organizational context. Just such an
approach has been explored by Heisler and
Damuth (1987), who regard multilevel (or
“group”) selection to have taken place
“whenever an individual’s expected viabil-
ity, mating success, and/or fertility cannot
be accounted for solely on the basis of that
individual’s phenotype, but rather addi-
tional information is required about prop-
erties of the group or groups of which the
THE VALLEY OF OAXACA 201
individual is a member” (1987:584). They
consider a theoretical array of relationships
between a given phenotypic trait and adap-
D
tive fitness in a population consisting of
three subgroups (see Fig. 1). The distribu-
tion in Fig. 1d shows no relationship be-
tween the value of the trait and fitness; the
mean fitness is the same for all three
groups. In Fig. 1a, an increase in fitness is
associated with an increase in the value of
the phenotypic trait, but there are no signif-
icant group selection effects on the trait.
Figure 1b shows the operation of group se-
lection on an altruistic trait; an increase in
the value of the trait is negatively associ-
ated with individual fitness within groups,
but groups with high mean values for the
trait have higher mean fitness than do
groups with lower fitness. Finally, Fig. 1c
depicts multilevel selection in situations
where increases in the trait’s value are asso-
ciated with increases in both group-level
and individual-level fitness. A driven evo-
lutionary trend is expected to be associated
with the pattern shown in Fig. 1b or Fig. 1c;
202 SPENCER AN
tion. Horizontal axes represent values for the charac-
ter, while the vertical axes represent the relative fitness
(redrawn from Heisler and Damuth 1987: Fig. 2). See
the text for explanation.
What quantitative procedures can be
used to assess which of the four idealized
patterns offers the closest fit to a given em-
pirical situation? For this task, Heisler and
Damuth favor a series of multilevel statisti-
cal techniques known collectively as contex-tual analysis. Originally developed for use
in political science research, contextual
analysis seeks to “explain variation in indi-
vidual-level behavior in terms of individ-
ual- and group-level effects” (Boyd and
Iverson 1979:2). Boyd and Iverson present
an array of analytical possibilities, suggest-
ing that the appropriate method in a partic-
ular case will depend both on the problem
one is investigating and on the nature of the
data available for analysis. For example, in
a case where individual differences are
measured on a continuous ratio scale, but
group differences are indicated by a nomi-
nal variable (such as membership vs non-
membership), they recommend the use of
analysis of covariance, or ANCOVA, to de-
tect the existence of group-level effects
(Boyd and Iverson 1979:9–12; see also
Arnold and Fristrup 1982; Goodnight et al.
1992; Heisler and Damuth 1987). Let us
REDMOND
now attempt to apply these evolutionary
in both cases, there is a close correlation (ei-
ther negative or positive) between group
and individual selection.
FIG. 1. Four possible relationships between a phe-
notypic character and fitness in a subdivided popula-
concepts and methods to archaeological
data from Oaxaca.
MONTE ALBAN I IN THE VALLEY OF OAXACA
MA I, as noted earlier, was a time of rapid
population growth and a dramatic increase
in political centralization. Are these devel-
opments appropriately viewed as passive
or driven evolutionary trends? Is there evi-
dence for the operation of multilevel (or
“group”) selection? Can the ascendance of
the regional polity centered at Monte Albán
be understood as the product of coordi-
nated selection on multiple levels of organi-
zation? If so, what were the most important
vehicles of selection, and is there evidence
of competition among them? When did the
key institutions of the Zapotec state first
make their appearance and what was the
selective context of this development? By
which period can we say that the major
subregions of the Valley of Oaxaca were
united into a single polity under the direc-
tion of Monte Albán?
We will approach these questions by ap-
plying the related concepts of evolutionary
trends and multilevel selection to the exten-
sive data set collected by the Valley of Oax-
aca Settlement Pattern Project (Blanton et
al. 1982; Kowalewski et al. 1989), and in
particular the data on productive potential
and population size presented and ana-
lyzed by Nicholas (1989) and Feinman and
Nicholas (1990). In this data set, the infor-
mation for each chronological phase is con-
tained in a separate computer file, and each
record in a particular file contains the data
on production and population for a certain
4-km by 4-km grid square of the valley.
While it is of course true that these grid
squares are arbitrary units imposed on the
valley by the data collectors, the squares are
nevertheless useful for the present analysis
for two main reasons: (1) their small size
makes them relatively “local” in nature;
and (2) they can be combined into succes-
sively more inclusive groupings, yielding a
phase-by-phase, multilevel data set to
which we can apply evolutionary trend the-
ory and the concept of multilevel selection.
In the analysis to follow, we regard individ-
ual grid squares as localities. The localities
can be grouped into the more inclusive
units of subregions, and then grouped again
into the region of the Oaxaca Valley as a
whole, producing a data set with three lev-
els of organization for each prehistoric
phase of interest.
We have drawn subregion boundaries by
taking into account natural topographic di-
visions as well as the hypothetical political
affiliations that have been proposed for each
phase. In the case of the Rosario phase, Blan-
POLITICAL EVOLUTION IN
ton et al. (1993: 69) note that “the valley was
divided into at least three settlement clus-
ters, one in Etla and one south of the central
area, with most of its sites in the Valle
Grande. A third cluster was located at the
center of the Tlacolula arm.” Marcus and
Flannery (1996: 123–126) also see a tripartite
division, and suggest that the relatively un-
occupied area at the valley center was a
buffer zone between the Etla valley polity
(whose first-order center was San José
Mogote) and competing polities in the Tla-
colula and Ocotlán/Zimatlán subregions,
the first-order centers of which were proba-
bly Yegüih and San Martín Tilcajete, respec-
tively (Fig. 2). The Rosario phase subregion
boundaries presented in Fig. 3 reflect these
proposed political groupings, and take into
account natural topographic divisions as
well. The Central subregion as drawn in Fig.
3 corresponds to Kowalewski’s Central Val-
ley Survey area (Kowalewski 1976); it has
very little Rosario settlement and is the loca-
tion of the proposed buffer zone between
Etla and the other branches of the valley.
There are some Rosario phase settlements in
the Northern Valle Grande, and, of course,
several more in the Ocotlán/Zimatlán sub-
region (Fig. 3). Even though the Northern
Valle Grande and Ocotlán/Zimatlán are sep-
arated by a significant range of hills, Marcus
and Flannery suggest they may have been
somewhat affiliated with one another politi-
cally during the Rosario phase
(1996:123–126). If we follow their argument,
we can combine them and thus end up with
four subregions: (1) Etla; (2) Central (a
mostly vacant buffer zone); (3) Tlacolula;
and (4) Ocotlán/Zimatlán, which would in-
clude the Northern Valle Grande (Fig. 3).
For Early MA I and Late MA I, the out-
standing settlement in the entire region is
Monte Albán, situated atop a series of hills in
the valley center (Figs. 4, 5). The establish-
ment of Monte Albán changed the political
landscape dramatically in Oaxaca (Blanton
1978; Marcus and Flannery 1996:142–171).
After 500 B.C., the largest settlement in the
THE VALLEY OF OAXACA 203
Valley was no longer in the middle of the Etla
subregion, as was the case with San José
Mogote in the Rosario phase, but was far
tion of the sites abandoned at the end of
s
more centrally placed with respect to all
three major branches of the Valley. In both
Early MA I and Late MA I, Monte Albán di-
rectly overlooked not only the Central Valley
Survey area, but also the adjacent Northern
Valle Grande. In view of this proximity, we
have combined the Northern Valle Grande
and the Central Survey area into a single
Central subregion for Early MA I and Late
MA I, remembering that the division be-
tween the Central Valley Survey area and the
Northern Valle Grande was based not on a
topographic divide, but instead on the
FIG. 2. The Oaxaca Valley, showing the location
adapted from Flannery 1986: Fig. 3.1).
boundaries of two different survey projects
(Kowalewski 1976; Blanton et al. 1982).
Topographic and political considerations
make it advisable to combine the data from
the Central and Etla subregions for all of MA
I. There is no major topographical division
between the Central and Etla branches, and
there is reason to think that there was a close
political connection between Etla and the
Central subregions at this time. Marcus and
Flannery (1996:139–140) propose that the
founders of Monte Albán probably came
from San José Mogote and other Etla valley
towns and villages, because a high propor-
of archaeological sites mentioned in the text (inset
204 SPENCER AND REDMOND
Rosario phase are in the Central and south-
ern Etla areas. For example, in their excava-
a
tions at San José Mogote, Marcus and Flan-
nery found evidence of Rosario-phase public
buildings and some 60 hectares of Rosario-
phase occupation representing both elite and
nonelite inhabitants; however, there is very
little evidence of Early MA I architecture
here, and no evidence at all of Late MA I ar-
chitecture (Flannery and Marcus 1990). It
seems likely that Monte Albán was serving
as the first-order center for the Etla subregion
FIG. 3. Rosario phase settlements in the Oaxac
from Blanton et al. 1993: Fig. 3.7).
as well as the Central area in Early MA I.
By contrast, both the Tlacolula and
Ocotlán/Zimatlán subregions continued to
have impressive first-order centers in Early
MA I, Yegüih and San Martín Tilcajete, re-
spectively. And, as Marcus and Flannery
(1996:163) note, these centers apparently
grew even larger in Early MA I than they
were in the Rosario phase: “This growth
suggests that Yegüih and Tilcajete may not
have participated in the [Monte Albán
polity]; they might still have been the para-
mount centers of rival polities during
Valley, showing subregion boundaries (redrawn
POLITICAL EVOLUTION IN THE VALLEY OF OAXACA 205
Monte Albán Ia. If so, any smaller commu-
nities near them would have been under
their control, not Monte Albán’s.” Recently,
tic expansion, beginning with an urban relocation
O
Feinman (1998: 128–129) has taken a similar
view of Late MA I, arguing that much of the
southern Valle Grande (the Ocotlán/Zi-
matlán area) and the Tlacolula branch were
not completely integrated into the Monte
Albán polity. Marcus and Flannery have
FIG. 4. Early Monte Albán I settlements in the
drawn from Blanton et al. 1993: Fig. 3.9).
also proposed that the relations among
these autonomous polities became increas-
ingly competitive and hostile between
Early MA I and Late MA I, as Monte Albán
sought to incorporate all parts of the Valley
into its political domain.
We suspect that Monte Albán I was a 400-year
period of spectacular demographic and militaris-
axaca Valley, showing subregion boundaries (re-
206 SPENCER AND REDMOND
at 500 B.C. and proceeding to subjugation of the
entire Valley of Oaxaca by 100 B.C. (Marcus and
Flannery 1996: 165)
O
For the purposes of this paper, we will
view the suggestions of Marcus, Flannery,
and Feinman as propositions consistent
with what we shall call the “Rival Polity
Model” of Oaxaca Valley political organiza-
tion during the Rosario, Early MA I, and
FIG. 5. Late Monte Albán I settlements in the
drawn from Blanton et al. 1993: Fig. 3.12).
Late MA I phases. We will assess the valid-
ity of the Rival Polity Model by proposing
an evolutionary hypothesis that is consis-
tent with the model and also testable with
archaeological data.
Hypothesis 1 (Inter-Subregion CompetitionHypothesis). We would expect inter-subre-
gional selection to increase in relative im-
portance as we proceed from the Rosario
axaca Valley, showing subregion boundaries (re-
POLITICAL EVOLUTION IN THE VALLEY OF OAXACA 207
phase to Early MA I to Late MA I, a reflec-
tion of continuing and intensifying compe-
tition among the hypothesized rival poli-
D
ties. As we shall see, the subregional group-
ing we have offered—Etla/Central, Tla-
colula, and Ocotlán/Zimatlán—permits an
evaluation of this hypothesis through an
analysis of the grid-square data of the Val-
ley of Oaxaca Settlement Pattern Survey.
Before doing so, however, we should de-
termine whether the impressive changes of
the Rosario phase through Late MA I can be
considered a driven trend in McShea’s
terms. Let us examine the distribution of one
key variable of the system over time, the es-
timated archaeological population by local-
ity. According to McShea’s skewness test, an
evolving system is exhibiting a driven trend
if: (1) the overall system shows significantly
positive skewing; and (2) subsets of the sys-
tem also show such skewing. Table 1 pre-
sents measures for skewness, standard error
of skewness, and the ratio of skewness to
standard error of skewness for the grid-
square data set for the Valley as a whole and
by subregion. It is clear that, for all three pe-
riods, the Valley as a whole exhibits a popu-
208 SPENCER AN
lation di
skewed in
tios of ske
Note. As explained in the text, the Oco
Northern Valle Grande for the Rosario ph
MA I, the Northern Valle Grande is group
ness are greater than 2.0 (SPSS 1998:214). We
suggest, however, that the importance of
these results lies not in their statistical signif-
icance per se, but emerges instead when the
figures are used comparatively, to contrast the
different chronological periods in order to
detect trends. In this light it is notable that
there is an increase in the degree of skewing
(as shown both by the skewness measure
and the ratio of skewness to standard error
of skewness) in the Valley as a whole in the
trajectory from the Rosario phase, to Early
MA I, to Late MA I (Table 1).
Table 1 also contains the skewness mea-
sures for the three subregions of the Valley.
With the exception of the Central subregion
in the Rosario phase, all the skewness mea-
sures are positive and statistically signifi-
cant, which suggests that the pattern of
change between the Rosario phase and Late
MA I can indeed be interpreted as a driven
trend (McShea 1994). Once again, the skew-
ness values take on greater importance
when they are used in a comparative con-
REDMOND
er time.
Central
exceed
stribution that is significantly
the positive direction; all three ra-
wness to standard error of skew-
text, to contrast the subregions ov
Note that the values for the Etla/
subregion in each phase greatly
TABLE 1Skewness Measures for the Distribution of Archaeological Population
(All Sites; by Entire Valley and Subregion)
Phase Subregion Skewness SE Skewness Skewness/SES
Rosario Entire valley 5.947 0.343 17.74
Etla 3.995 0.550 7.26
Central 1.902 1.014 1.88
Tlacolula 1.640 0.687 2.39
Ocotlán/Zimatlán 2.249 0.550 4.09
Early MA I Entire valley 8.566 0.251 34.13
Etla/Central 6.225 0.350 17.79
Tlacolula 3.510 0.456 7.70
Ocotlán/Zimatlán 1.606 0.512 3.14
Late MA I Entire valley 10.45 0.212 49.08
Etla/Central 7.122 0.311 22.90
Tlacolula 2.541 0.388 6.55
Ocotlán/Zimatlán 2.228 0.398 5.60
tlán/Zimatlán subregion includes the
ase; however, for Early MA I and Late
ed with the Central subregion.
those for Tlacolula and Ocotlán/Zimatlán
(Table 1). Moreover, if one examines the
pattern of change over time in the Skew-
ness/SES ratio for each subregion, it is the
Etla/Central subregion that shows the most
consistent increase over the three phases in
question (Fig. 6), which would imply that
the Etla/Central subregion was undergoing
a more driven process of change than the
Tlacolula and Ocotlán/Zimatlán subre-
gions. We can draw on these results to pro-
pose a second evolutionary hypothesis:
Hypothesis 2 (Interdependent Selection Hy-pothesis). In line with McShea’s theory of
evolutionary trends, we would expect the
Etla/Central subregion to show evidence of
greater interdependence among levels of
selection than the Tlacolula and Ocotlán/
Zimatlán subregions, and this interdepen-
dence should reach its peak in the last of the
three phases in question, Late MA I.
We will test Hypothesis 1 and Hypothe-
sis 2 by applying the concept of multilevel
selection to the Oaxaca data set. Sober and
Wilson (1998:102–116) give three basic re-
quirements that must be met in order to use
their multilevel selection approach in a
POLITICAL EVOLUTION IN
matlán subregions; the graph indicates that the
Etla/Central subregion shows the most consistent pat-
tern of increase.
among groups,” noting that “at all levels of
the biological hierarchy, units must differ
from one another before they can be sifted
by natural selection” (1998:105). In our
Oaxaca case, if we use an expanded view of
“phenotype” (see Dunnell 1989), we might
consider the phenotypic variation among
localities (and among groups of localities)
to be expressed by the variation in agricul-
tural productivity, which can be expressed
as potential population. Nicholas (1989)
notes that the values for this variable de-
pend not only on environmental factors
such as soil type and rainfall, but also on
agricultural techniques, available labor
supply, and work organization, which are
largely cultural in nature. Both the environ-
mental and cultural factors exhibit varia-
tion among the localities and over time.
Sober and Wilson’s second requirement
is that the phenotypic differences in ques-
tion must be heritable (1998:107). In our cul-
tural evolutionary case, we suggest that the
phenotypic differences can be considered
heritable in the sense that the cultural com-
ponents of agricultural productivity com-
prise practices that are passed down from
generation to generation in a process of cul-
tural transmission (see Boyd and Richerson
1985). Taking a similar position, Sober and
Wilson (1998: 114) note that
The fact that a behavior is transmitted cultur-
ally should not be taken to mean that it is nonher-
itable. Cultural differences between human
groups are often stable over long periods of time
and are faithfully transmitted to descendant
groups. They are heritable in the sense that off-
spring units resemble parent units, which is all
that matters as far as the process of natural selec-
tion is concerned.
Sober and Wilson’s third requirement is
that one must have a way to “determine the
fitness consequences of phenotypic varia-
tion within and among groups” (1998:115).
In our Oaxaca case, the salient question is
THE VALLEY OF OAXACA 209
given situation. Their first requirement is
that one should be able to “determine the
pattern of phenotypic variation within and
FIG. 6. Change over time in the Skewness/SES mea-
sure for the Etla/Central, Tlacolula, and Ocotlán/Zi-
how we should measure the relative “fit-ness” of localities. We suggest that, if we re-
gard agricultural productivity as a pheno-
there still would have been more than
D
the “fitness” measure (the ratio of archaeological pop-
typic trait of a locality, then the “fitness” of
that locality can be estimated by computing
a function, such as a ratio or a linear trans-
formation, of the archaeological populationwith respect to the agricultural productivity(potential population) of the locality. Note
that we are using “locality” as shorthand
for “the human population of the locality.”
The underlying assumption here is that the
relative fitness of the localities will be re-
flected in the population sizes they are able
to sustain, relative to their agricultural po-
tential.
Because McShea’s evolutionary trend
theory deals with the relationship between
driven trends and interdependence among
levels of selection, we will have to consider
the patterns of selection in the subordinate
localities of each subregion separately from
the locality in each subregion that was not
purely “local” in a political sense. Accord-
ing to the Rival Polity Model, each subre-
gion contained a proposed first-order cen-
ter; presumably the center was sustained, in
part, by surplus mobilized from subordi-
nate villages. During the Rosario phase,
these proposed centers are San José Mogote
(Etla/Central subregion), Yegüih (Tlacolula
subregion), and San Martín Tilcajete
(Ocotlán/Zimatlán subregion). During Early
MA I and Late MA I, the three centers are
Monte Albán (Etla/Central subregion),
Yegüih (Tlacolula subregion), and San
Martín Tilcajete (Ocotlán/Zimatlán subre-
gion). Therefore, the following consideration
of patterns of fitness will have two parts re-
flecting these two vantage points: (1) from
the perspective of the first-order centers, and
(2) from the perspective of the subordinate
localities.
First, let us assess the evidence for
change over time in the fitness of the local-
ity that contains the first-order center in
each of the three subregions (Fig. 7). Note
that fitness in this case was computed by
taking the natural logarithm of the ratio of
210 SPENCER AN
archaeological population (based on occu-
pation area) to potential population (based
on agricultural productivity). The graph
shows that the locality containing the first-
order center in the Etla/Central subregion
has a higher fitness in all phases and ex-
hibits a more dramatic increase in fitness
over time than the localities containing the
centers of the Tlacolula subregion or the
Ocotlán/Zimatlán subregion. For both
Early MA I and Late MA I, the actual esti-
mated population at Monte Albán far ex-
ceeds the potential population of its local-
ity; this would have made Monte Albán
highly dependent upon support from the
other localities in its subregion, as Nicholas
(1989) has pointed out.
It is important to recognize that the fact
that Monte Albán was not locally self-suffi-
cient in terms of food production does not
mean that it was “unfit.” Monte Albán not
only maintained its large size but also grew
substantially throughout the four hundred
years of MA I. Clearly, the center was very
proficient at mobilizing material support,
probably in the form of surplus agricultural
production, from the villages within its po-
litical domain. If we assume, conservatively,
that the MA I domain of Monte Albán con-
sisted of just the Etla/Central subregion,
ulation to potential population) for the locality con-
taining the first-order center in each subregion.
REDMOND
FIG. 7. Change over time in the natural logarithm of
enough surplus potential to sustain the en-
tire population of the subregion throughout
MA I, including Monte Albán: the total po-
tential population of the Etla/Central subre-
gion is 83,488 in Early MA I and 116,541 in
Late MA I, comfortably above the total ar-
chaeological population of 11,245 in Early
MA I and 39,382 in Late MA I.
Now, we shall use the methods of contex-
tual analysis (Boyd and Iverson 1979;
Heisler and Damuth 1987) to assess the pat-
terns of selection on the subordinate locali-
ties. The Interdependent Selection Hypoth-
esis predicts that Monte Albán’s growing
reliance upon the other localities of the
Etla/Central subregion would be comple-
mented by increasing dependence in those
localities upon Monte Albán’s subregional
leadership. Such growing dependence
should be manifested by an increase in therelative importance of subregional-level selec-tion upon the subordinate localities. We would
expect selection processes on the subordi-
nate localities to have been more affected
by subregional-level selection in the
Etla/Central subregion (under Monte
Albán’s leadership) than in Tlacolula and
Ocotlán/Zimatlán. In addition, the Inter-
Subregion Competition Hypothesis pre-
dicts that the intensity of inter-subregional
selection should increase between the
Rosario phase and Late MA I.
Let us address these propositions with a
three-step contextual analysis. For each
phase, the first step will be to construct a
separate scatter plot for each subregion. In
doing so we shall use all the localities with
the exception of the locality that contains
the first-order center for each subregion, be-
cause we are attempting to assess the de-
gree to which the subordinate localities are
affected by group selection. On each plot,
the x-axis will represent agricultural pro-
ductivity (expressed in potential popula-
tion), while the y-axis will represent archae-
ological population (as estimated by the
Valley survey team, based on occupation
POLITICAL EVOLUTION IN
area). In the second step, we will conduct a
series of parallel regressions to evaluate the
degree to which the fitness of the localities
in the different subregions can be expressed
as a linear function of the actual population
with respect to the potential population.
The degree to which the regressions are sta-
tistically significant—that is, the signifi-
cance of the slope of the regression—will
reflect the degree to which the distribution
of population throughout the localities of
the subregions corresponds to variation in
the agricultural potential of those localities.
In the third step, we will conduct an analy-
sis of covariance (ANCOVA) to assess the
relative intensity of selection on the level
of the subregion, i.e., the level of organiza-
tion pertaining to the hypothesized rival
polities.
The scatter plots for the Rosario phase,
Early MA I, and Late MA I are presented in
Figs. 8–10. In each plot, individual data
points are identified by symbol as to their
subregion membership. In addition, a sepa-
rate linear regression line is fitted to the
data for each subregion, so that each scatter
plot reflects three different regressions, one
for each subregion. Summary statistics on
the various parallel regressions are pre-
sented in Table 2. It is clear from the rela-
tively low R2 values in Table 2 that a rela-
tively small proportion of the overall
THE VALLEY OF OAXACA 211
variation in archaeological population is
explained by a linear regression on produc-
tive potential. Nevertheless, as we argued
FIG. 8. Scatter plot of archaeological population
(based on occupation area) versus agricultural produc-
tivity (expressed as potential population) for three
subregions during the Rosario phase; the first-order
center in each subregion is not included.
tivity (expressed as potential population) for three
subregions during Early Monte Albán I; the first-order
center in each subregion is not included.
earlier regarding our use of the skewness
values, the regression results are most valu-
able when employed comparatively, to com-
pare the different subregions and phases in
order to elucidate trends.
For the Rosario phase, only the Etla sub-
region exhibits a positive relationship (up-
ward slope) between agricultural produc-
tivity and archaeological population (Fig.
8). The linear regression results (Table 2) in-
dicate that the slope of the Etla regression
line is 0.020, while the slope coefficients of
the Tlacolula and Ocotlán/Zimatlán lines
tivity (expressed as potential population) for three
subregions during Late Monte Albán I; the first-order
center in each subregion is not included.
are both (negative) 0.004. The statistical sig-
actually 0.0) is 0.256 for Etla, which is not
highly significant. Nevertheless, the regres-
sions for the Tlacolula and Ocotlán/Zi-
matlán subregions are even less significant
at 0.847 and 0.692, respectively. The R2
value for Etla (0.088) is also considerably
greater than that for Tlacolula (0.006) or
Ocotlán/Zimatlán (0.012). Thus, while the
relationship between agricultural produc-
tivity and archaeological population is not
strong in any of the three subregions during
the Rosario phase, the greatest tendency to-
ward a linear relationship is seen in the Etla
subregion.
For the Early MA I phase (Fig. 9), the fitted
regression line for the Etla/Central subre-
gion shows a more positive slope than do the
fitted regression lines for Tlacolula and
Ocotlán/Zimatlán. The results of the regres-
sion analysis (Table 2) reveal that the
Etla/Central slope coefficient is 0.081, associ-
ated with a 0.056 significance level. The re-
gression line slope coefficient for the Tla-
colula subregion is 0.018 (0.477 significance
level), while for Ocotlán/Zimatlán the slope
coefficient is 0.040 (0.139 significance level).
Thus, in Early MA I, the linear regressions for
all three subregions have moved in the direc-
tion of stronger statistical significance, but
the Etla/Central subregion exhibits a much
more significant regression than the other
two subregions. Although the R-Square
value associated with the Etla/Central re-
gression (0.081) reveals that just 8% of the
variability in archaeological population is ac-
counted for by the regression on agricultural
productivity, it is noteworthy that the associ-
ated significance level is stronger by a factor
of 4.6 than the strongest significance level ob-
served (Etla subregion) for the Rosario phase
(Table 2). Since this sample does not include
the first-order centers in each subregion, the
Early MA I regression reflects a valley-floor
population distribution in Etla/Central that
corresponds more closely to the interlocality
variability in agricultural productivity than
REDMOND
212 SPENCER ANDFIG. 9. Scatter plot of archaeological population
(based on occupation area) versus agricultural produc-
FIG. 10. Scatter plot of archaeological population
(based on occupation area) versus agricultural produc-
nificance of the regression (i.e., the proba-
bility that the slope of the regression line is
is the case for the previous phase or the other
subregions during the Early MA I phase.
213
cluded from this analysis; also excluded is the Central (buffer) zone during the
For the Late MA I phase (Fig.10), the fitted
regression line shows a more positive slope
for the Etla/Central subregion than we
have seen thus far. The results of the regres-
sion analysis (Table 2) indicate that the coef-
ficient of the slope for Etla/Central is 0.188,
with an associated significance level of
0.002, a highly significant regression. For
Tlacolula, the slope coefficient is 0.046
(0.126 significance level), while for
Ocotlán/Zimatlán the slope coefficient is
0.039 (0.144 significance level). Although it
is true that just 16% of the variation in ar-
chaeological population for Etla/Central is
explained by the regression, the true value
of this analysis lies, as we have already
noted, in using the results in a comparative
context to detect trends over time and
space. For example, if one compares the re-
gression analysis results for Early MA I and
Late MA I (Table 2), one finds that the slope
of the regression line for Etla/Central in-
creases by a factor of 2.3 (from 0.081 to
0.188), the statistical significance becomes
Rosario phase.
stronger by a factor of 28 (from 0.056 to
0.002), and the R2 value increases by a fac-
tor of 1.95 (from 0.081 to 0.158). In addition,
if one compares the Late MA I results for
the three subregions, one finds that: (1) the
slope of the Etla/Central regression line is
more than 4 times greater than the slopes of
Tlacolula and Ocotlán/ Zimatlán; (2) the
statistical significance of the regression line
slope for Etla/Central is more than 63 times
that of the significance of the regressions for
Tlacolula and Ocotlán/Zimatlán; and (3)
the R2 value for Etla/Central is more than
two times greater than those for the other
two subregions (Table 2). Overall, these re-
sults indicate that: (1) the distribution of
Late MA I population through the localities
corresponds more closely to variability in
agricultural productivity in the Etla/Cen-
tral subregion than in Tlacolula and
Ocotlán/ Zimatlán; and (2) in the Etla/Cen-
tral subregion, this correspondence be-
comes closer between Early MA I and Late
MA I. We suggest that this closer correspon-
dence between labor and agricultural po-
tential in Etla/Central would have permit-
ted a relatively more efficient production of
POLITICAL EVOLUTION IN THE VALLEY OF OAXACA
TABLE 2Summary Statistics of Linear Regressions: Archaeological Population on
Agricultural Productivity (Potential Population) by Phase and by Subregion
Phase Subregion R2 Slope Signif. (p)
Rosario
Etla .088 .020 .256
Tlacolula .006 2.004 .847
Ocotlán/Zimatlán .012 2.004 .692
Early MA I
Etla/Central .081 .081 .056
Tlacolula .023 .018 .477
Ocotlán/Zimatlán .124 .040 .139
Late MA I
Etla/Central .158 .188 .002
Tlacolula .068 .046 .126
Ocotlán/Zimatlán .066 .039 .144
Note. As explained in the text, the first-order center in each subregion is ex-
the agricultural surpluses that were needed
to support the subregion’s first-order cen-
ter, Monte Albán.
D
As Blanton et al. (1982:44) have pointed
out, many of the sites in the Etla/Central sub-
region that were first occupied in Late
MA I are located in the piedmont zone,
especially within that part of the valley lying
18–20 km from Monte Albán. This pattern,
they have suggested, resulted from Monte
Albán’s pursuit of the “piedmont strategy,”
which fostered a concentration of agricultural
populations in the piedmont zones around
the capital in order to increase the production
of surpluses (see also Kowalewski et al.
1989:123–124). The results of our parallel re-
gression analyses clearly do not contradict
their suggestions regarding the piedmont
strategy; it could be, however, that this strat-
egy was part of a broader effort to bring the
population distribution of the Etla/Central
subregion into closer alignment with local
variability in agricultural potential.
Although our parallel regressions may
bear a superficial resemblance to those used
by Brumfiel (1976) and Steponaitis (1981) in
their analyses of Basin of Mexico data, our
approach differs from theirs in certain im-
portant ways. In both of their applications,
the different levels of the parallel regres-
sions pertained not to different subregions
(as in our application), but rather to differ-
ent levels of the settlement hierarchy.
Brumfiel (1976) interpreted a significant re-
gression with a high R2 value as reflecting
“population pressure” and used this mea-
sure to assess Carneiro’s (1970) model of
state emergence. When she analyzed two
levels of the settlement hierarchy in the
Late Formative to the Terminal Formative
in her study region, she found that signifi-
cant regressions with relatively high R2 val-
ues characterized both levels of the settle-
ment hierarchy in the Terminal Formative
(which she interpreted as the time of state
emergence). By contrast, her analysis of the
Late Formative data showed a significant
regression only for the upper level of the
214 SPENCER AN
settlement hierarchy. She interpreted these
results as indicating that a condition of
population pressure did not precede and
hence was probably not causal to state
emergence. Steponaitis (1981), on the other
hand, used the vertical differences between
the parallel regressions he calculated for the
various levels of the settlement hierarchy as
a way of gauging the flow of tribute that
presumably moved from the primary pro-
ducers in the smaller settlements of the
Basin to the higher-order centers; such trib-
ute flows, he suggested, allowed the
higher-order centers to grow to sizes that
would have exceeded the carrying capaci-
ties of their immediate catchment areas. In
our present application, each level of the
parallel regressions (Figs. 8–13) pertained
to a different subregion of the Valley of
Oaxaca, and the degree to which each re-
gression had a statistically significant posi-
tive slope reflected the degree to which
greater population size was positively asso-
ciated with greater productive potential.
We then used these results comparatively.
As we pointed out, the Etla/Central subre-
gion exhibited a more significant positive
relationship between productive potential
and population than did the other subre-
gions and, moreover, this statistical rela-
REDMOND
tionship in Etla/Central became increas-
ingly more significant between the Rosario
and Late MA I phases.
FIG. 11. Scatter plot of the natural logarithm of the
archaeological population (based on occupation area)
versus agricultural productivity (expressed as poten-
tial population) for three subregions during the
Rosario phase; the first-order center in each subregion
is not included.
In line with McShea’s evolutionary trend
theory, we would expect the tighter corre-
spondence between productive potential
and population on the level of the localities
in the Etla/Central subregion to be associ-
ated with increasingly strong subregional-
level selection as we proceed from the
Rosario to Early MA I to Late MA I phases.
This expectation, we should note, would
also be consistent with the Rival Polity
Model. Let us test these convergent expecta-
tial population) for three subregions during Early
Monte Albán I; the first-order center in each subregion
is not included.
tivity (expressed as potential population) for three
subregions during Late Monte Albán I; the first-order
center in each subregion is not included.
441–452). Analysis of covariance, or AN-
COVA, is a hybrid of analysis of variance
and regression (Wilkinson et al. 1996: 441).
The method assesses the degree to which
group membership (called a factor, usually a
nominal-scale variable) affects the relation-
ships between two continuously measured
(or ratio-scale) variables, one of which is
called the covariate and the other the depen-dent variable. In our application, the individ-
ual cases are the localities (once again, ex-
cluding the locality that contains the
first-order center in each subregion). The fac-tor is the subregion to which the localities be-
long, while the covariate is the agricultural
productivity (potential population) and the
dependent variable is the archaeological (esti-
mated) population. The ANCOVA will help
us assess whether subregional membership
exerts a significant effect on the relationship
between agricultural productivity and ar-
chaeological population.
Wilkinson et al. (1996:445) point out that
analysis of covariance is most appropriate
in situations where the regression lines for
the different subgroups have the same
slope; the technique essentially evaluates
whether the vertical differences between
the lines are statistically significant. A
glance at our Figs. 8–10 shows that parallel
slopes do not characterize these regres-
sions. However, a natural logarithmic
transformation of the dependent variable
(archaeological population) yields regres-
sion lines that, while not much different for
the Rosario phase (Fig. 11), are much closer
to parallel for Early MA I and Late MA I
(Figs. 12, 13) than was the case in Figs. 9
and 10 for the unmodified dependent vari-
able. If we conduct the ANCOVA on both
the transformed and untransformed data,
we can evaluate the statistical significance
of the results from a comparative perspec-
tive.
Table 3 provides output statistics associ-
THE VALLEY OF OAXACA 215
POLITICAL EVOLUTION INFIG. 12. Scatter plot of the natural logarithm of the
archaeological population (based on occupation area)
versus agricultural productivity (expressed as poten-
tions with another contextual analysis tech-
nique, analysis of covariance (Boyd and
Iverson 1979: 9–12; Wilkinson et al. 1996:
FIG. 13. Scatter plot of archaeological population
(based on occupation area) versus agricultural produc-
ated with each set of ANCOVA analyses.The table contains, for each analysis, the R2
value and the level of statistical significance
216
cultural productivity (potential population). As explained in the text, the first-order
associated with the effects of subregional
membership on the relationship between
archaeological population and agricultural
productivity (expressed as potential popu-
lation). As we have done in earlier analyses,
let us view the statistical results from a
comparative perspective in order to discern
trends. Whether one looks at the ANCOVA
as performed on the unmodified dependent
variable (archaeological population) or the
transformed dependent variable (natural
logarithm of archaeological population),
the overall pattern is the same. The effects
of subregional membership become in-
creasingly significant (i.e., the probability
that the effects are due to chance decline) as
we proceed from the Rosario phase to Early
MA I to Late MA I, peaking during Late
MA I at significance levels of 0.001 and
0.005, in the analyses with the untrans-
formed dependent variable and trans-
formed dependent variable, respectively
(Table 3). The R2 values also show a pattern
of increase between the Rosario phase and
Late MA I, peaking at 0.217 (untransformed
dependent variable) and 0.262 (trans-
formed dependent variable). We suggest
that the multilevel regression lines for Late
MA I (Fig. 13) resemble the pattern (shown
center in each subregion is excluded from
(buffer) zone during the Rosario phase.
in Fig. 1c) that Heisler and Damuth (1987)
have associated with situations where in-
creases in a trait’s value are linked to in-
creases in both group-level and individual-
level fitness. As noted earlier, we would ex-
pect such a pattern of multilevel selection
to be associated with driven evolutionary
trends.
In sum, the results of the parallel regres-
sions and the analysis of covariance are
consistent with both the Interdependent Se-
lection Hypothesis and the Inter-Subregion
Competition Hypothesis: (1) there is evi-
dence of greater interdependence between
the local and subregional levels of selection
within the Etla/Central subregion than
within the Tlacolula and Ocotlán/Zimatlán
subregions, and this interdependence in-
creased over time from the Rosario phase
through Early MA I to Late MA I; and (2)
there is evidence that inter-subregional se-
lection increased in relative importance
over the same period, peaking in Late MA I.
We contend that support is lent to McShea’s
(1994) theory of evolutionary trends and
the Rival Polity Model (Marcus and Flan-
nery 1996: 163; Feinman 1998: 128–129),
from which the two hypotheses were de-
rived.
Our finding that inter-subregional selec-
tion increased between the Rosario phase
and Late MA I would imply that competi-
this analysis; also excluded is the Central
SPENCER AND REDMOND
TABLE 3Summary Statistics of Analysis of Covariance by Phase
Untransformed Transformed
dependent variable dependent variable
(archaeological population) (Log[archaeological population])
Phase R2 Signif (p) R2 Signif (p)
Rosario .037 .514 .095 .17
Early MA I .125 .113 .174 .035
Late MA I .217 .001 .262 .005
Note. The significance level presented is that associated with the effects of subre-
gional membership on the relationship between archaeological population and agri-
tion among subregions became stronger
over this time frame. As a further test of this
implication, we should look for indepen-
dent archaeological evidence of inter-subre-
gional competition, such as the burning of
settlements, the building of fortifications,
and other manifestations of hostilities at
key subregional centers. As it happens, evi-
POLITICAL EVOLUTION IN
dence of this kind has recently been recov-
ered at San Martín Tilcajete, the first-order
center of the Ocotlán/Zimatlán subregion.
SAN MARTIN TILCAJETE: RESEARCHAT A SUBREGIONAL CENTER
In 1993, the authors initiated the Tilcajete
Project, a multiyear program of intensive
mapping, surface collecting, and excava-
tion at three closely related archaeological
sites, El Mogote (SMT-11a), El Palenque
(SMT-11b), and Los Mogotes (SMT-23), all
situated near the present town of San
Martín Tilcajete in the district of Ocotlán,
Oaxaca (Fig. 14). The sites were located in
1978 by the Oaxaca Settlement Pattern Pro-
FIG. 14. The archaeological sites of El Mogote (S
(SMT-23), located near the present town of San Mar
intensive mapping and controlled surface
collecting at the three sites, followed by ex-
tensive excavations in 1995–1999. Three
seasons of excavation have occurred at both
El Mogote and El Palenque, while Los
Mogotes received its first season of excava-
tion in 1999 under the direction of Christina
Elson. Although our work is continuing,
some of the preliminary results bear di-
rectly on the present discussion.
A Rival Polity
At El Mogote (SMT-11a), our intensive
surface collections found evidence of a 25-
ha Rosario phase occupation, less than half
the size of the 60–65 hectares of Rosario
phase occupation at San José Mogote (Mar-
cus and Flannery 1996: 125) but still much
larger than any other Rosario phase settle-
ment in the Ocotlán/Zimatlán region.
THE VALLEY OF OAXACA 217
These data indicate that Tilcajete was al-
e
ld
ject (Blanton et al. 1982). During 1993–1994,
members of the Tilcajete Project carried out
ready a key subregional center by th
Rosario phase, an interpretation that wou
MT-11a), El Palenque (SMT-11b), and Los Mogotes
tín Tilcajete, Ocotlán, Oaxaca.
D
agree with the previously stated views of
other Oaxaca researchers (Blanton et al.
1999: 42; Kowalewski et al. 1989: 80; Marcus
and Flannery 1996:126). During Early MA I,
occupation at El Mogote grew substantially
to cover 52.8 ha (Fig. 15). Our excavations
have shown that the 2.2-ha plaza was laid
out at the onset of Early MA I. Oriented to
17° east of magnetic north (or 25° east of
true north), the plaza contained two
mounds in its center and other mounds
arranged around all four sides. In view of
the substantial occupation and large plaza,
it seems clear that El Mogote continued to
be the first-order center of the Ocotlán/Zi-
218 SPENCER AN
matlán subregion during Early MA I, as
intensive surface collections of the Tilcajete Project
plane map made by the authors in 1993).
The contextual analysis of the regional
settlement pattern data implied that the
Ocotlán/Zimatlán subregion was not a par-
ticipant in the Monte Albán phenomenon
during Early MA I, but was instead a rival
polity. Our field observations at El Mogote
are consistent with this view. First, neither
the orientation nor the configuration of
Early MA I mounds at El Mogote is similar
to what is known of Monte Albán’s Main
Plaza at this time. Second, given the large
size of the site and its plaza, it is notable that
we found relatively few examples of
the most elaborate Early MA I ceramics
known for Monte Albán (Caso et al. 1967;
REDMOND
Kowalewski et al. 1978). This paucity of the
Marcus and Flannery (1996: 163) have sug-
gested.
fanciest Early MA I ceramics at Tilcajete was
also reported by Blanton et al. (1982: 57). In
FIG. 15. El Mogote (SMT-11a), showing the plaza with major mounds as well as the locations of the
(schematic map, based on a detailed alidade and
T
particular, we found relatively few of the
common Early MA I cremas, such as Types
C.2 and C.4, which Feinman (1982: 188–191)
concluded were produced in the vicinity of
Monte Albán in the Etla/Central subregion
(in contrast to café and gris wares, which
were produced throughout much of the Val-
ley).1,* Most of the Early MA I ceramic as-
semblage at Tilcajete appears to consist of lo-
cally made wares, generically similar but
not identical to contemporaneous ceramics
at Monte Albán. Third, we found much evi-
dence that the El Mogote plaza area was
abandoned in a conflagration at the inter-
face between Early MA I and Late MA I.
Throughout our excavations on the north-
ern and eastern sides of the plaza we ob-
served that the uppermost floors of the
plaza and adjacent buildings were littered
with charcoal as well as burned earth,
adobe, and stone. Within this layer of char-
coal and burned stone we found a few
sherds of Type G.12, a bowl with combed
designs on its interior base; this type was not
otherwise present in the El Mogote plaza
constructions. The occurrence of Type G.12
in these contexts of abandonment is chrono-
logically significant. Caso et al. (1967:25–26)
reported that they first found G.12 sherds in
their Monte Albán Ib deposits, although
they were much more frequent in Monte
Albán Ic. Marcus and Flannery (1996:144)
suggest that MA Ia and MA Ic should be
considered discrete phases (corresponding
to our Early MA I and Late MA I phases, re-
spectively), with MA Ib “serving as the tran-
sition between them.” The first appearance
of Type G.12 (combed bottom), therefore,
probably dates to what we would call the
Early MA I/Late MA I interface.
During Late MA I, the El Mogote plaza
appears to have fallen into disuse. Instead,
a new plaza was built on the opposite side
POLITICAL EVOLUTION IN
of a barranca to the west and some 800 m
uphill, at the El Palenque site (SMT-11b)
(Fig. 16); the vertical difference between the
*See Notes section at end of paper for all footnotes.
two plazas is 30 m. Our intensive collec-
tions and excavations here have shown that
El Palenque was a relatively short-lived oc-
cupation: it was first inhabited in Late MA I
and was abandoned by the early years of
MA II. The Late MA I plaza at El Palenque
has precisely the same orientation, and is
strikingly similar in configuration to, the
Early MA I plaza at El Mogote. Like the ear-
lier plaza, the El Palenque plaza is oriented
17° east of magnetic north and has two
mounds in the middle and other mounds
around all four sides (Fig. 16). The slightly
smaller area of the El Palenque plaza (1.6 ha
vs the 2.2 ha of El Mogote), can probably be
understood as a practical response to the
challenge of building on a narrower pied-
mont ridge in the higher location.
The architectural continuity from El
Mogote to El Palenque, we suggest, reflects
the persistence of a local tradition of plaza
construction from Early MA I to Late MA I.
A reasonable interpretation is that the El
Mogote plaza was abandoned after being
partially burned in a raid, at which point a
decision was made to rebuild the plaza in a
higher, more defensible location. It is note-
worthy that the El Palenque site is pro-
tected by stone walls that traverse the site
along its gradual southern flank (Fig. 16).
An excavated cross section of one of these
walls has shown that it dates to Late MA I.
The continuation of the traditional plaza
layout in a new, more defensible location
probably indicates that the Tilcajete people
withstood the raid, even though they con-
sidered it serious enough to warrant mov-
ing the ceremonial plaza of their center
from El Mogote to a new location uphill.
Population at the site continued to grow.
While it is clear that the El Mogote plaza
was no longer in use in Late MA I, we
found evidence of Late MA I residential oc-
cupation over 43.5 ha of the El Mogote site.
Since there were also 28 hectares of Late
HE VALLEY OF OAXACA 219
MA I occupation at the El Palenque site, the
total habitation area at San Martín Tilcajete
during Late MA I was 71.5 ha.
FIG. 16. El Palenque (SMT-11b), showing the plaza with major mounds as well as the locations of the
t
The analysis of covariance (ANCOVA) of
the regional settlement pattern data indi-
cated that inter-subregional selection inten-
sified between Early MA I and Late MA I. El
Palenque, like El Mogote, shows signs of in-
dependence from Monte Albán. Although
G.12 bowls and other Late MA I diagnostic
types on gris paste are common at El
Palenque, our excavations there have found
only trace amounts of the thin-walled, well-
burnished cremas (such as Types C.6, C.7,
C.13, and C.20) that begin during Late MA I
intensive surface collections of the Tilcajete Projec
plane table map made by the authors in 1994).
at Monte Albán itself (Caso et al.
1967:46–47). Although our analysis is still
ongoing, preliminary results indicate that
the relative frequency of these Late MA I
cremas is even lower than that of the Early
MA I cremas such as Types C.2 and C.4. In
view of the aforementioned likelihood that
fancy crema ware was produced only in the
vicinity of Monte Albán (in contrast to the
gris ware), the paucity of Late MA I cremasat Tilcajete probably means that ceramic ex-
changes between the Monte Albán area and
the Tilcajete area were substantially re-
stricted at this time. Consistent with this in-
terpretation of reduced exchange is the fact
(schematic map, based on a detailed alidade and
220 SPENCER AND REDMOND
that we have recovered much locally avail-
able chert but remarkably little obsidian in
our El Palenque excavations, in spite of the
fact that we have excavated extensively in
ceremonial contexts as well as in elite and
nonelite residential contexts. Because
Monte Albán lay between Tilcajete and the
important obsidian sources of Central Mex-
ico, we suspect that Monte Albán was pre-
venting the flow of traded obsidian from
reaching El Palenque. We suggest that the
barriers that prevented Late MA I cremasand obsidian from reaching El Palenque
were primarily political in nature.
Our El Palenque fieldwork has recovered
evidence indicating that hostilities were
even more intense in Late MA I than in
Early MA I. Despite the fact that El
Palenque was in a defensible location and
was protected by stone walls, the site was
apparently the target of an attack that re-
sulted in a major conflagration and the
complete abandonment of the community.
In our excavation Area I we have excavated
a palatial residence (Structure 7) measuring
16 m 3 16 m that was completely burned
and abandoned. Carbonized roof beams
rested where they had fallen in rooms and
corridors. Whole vessels lay upon floors,
smashed by fallen debris. Moreover, unlike
the partial abandonment that occurred at El
Mogote at the end of Early MA I, when the
El Palenque plaza was burned and aban-
doned, the entire Late MA I residential zone
was abandoned as well.
Evidence of State Institutions
The results of the parallel regressions
(Table 2) were consistent with the proposi-
tion that the Etla/Central subregion during
Late MA I should exhibit the greatest inter-
dependence between levels of selection. In
the subregion where the first-order center
had the greatest need for mobilized sur-
plus, the population in the subordinate lo-
calities was distributed in a way that much
more closely matched the distribution of
POLITICAL EVOLUTION IN
agricultural resources than was the case in
the other two subregions. A reasonable in-
ference from this pattern is that there was
more coordination between the first-order
center and the localities in the Etla/Central
subregion than in Tlacolula and Ocotlán/
Zimatlán. We propose that by Late MA I the
political organization centered at Monte
Albán was beginning to engage in the kinds
of regulatory interventions on the local
level that require the delegation of author-
ity to specialized administrators. Such dele-
gation of authority is compatible with a
centralized administration that is also inter-
nally specialized, i.e., one organized along
the lines of a state (Spencer 1987, 1990, 1998;
Wright 1977, 1984). In line with this propo-
sition, we should expect to find evidence of
key Zapotec state institutions appearing by
Late MA I. Flannery and Marcus have ar-
gued that among the most important of
these institutions are the palace and the
multiroom temple (Flannery 1983, 1998;
Flannery and Marcus 1976, 1990; Marcus
and Flannery 1996: 180–182). We would ex-
pect these institutions to be most evident at
the first-order center of the state polity.
Unfortunately, due to later construction,
it is not yet possible to determine whether
the palace or the multiroom temple existed
at Monte Albán itself during Late MA I
(Flannery and Marcus 1990: 60; Marcus and
Flannery 1996: 165). We can, however, ex-
amine the Late MA I occupation at Tilcajete,
the likely first-order center of an au-
tonomous, rival polity. If the polity cen-
tered at Monte Albán was beginning to de-
velop state institutions by Late MA I, we
might find evidence that leaders of such a
rival polity were choosing to emulate key
aspects of institutional development in the
region’s most precocious and impressive
polity. Their motivation for doing so might
have been not only to attempt to share in
the developing elite culture but also to rein-
force their own authority by adopting novel
institutional features.
In line with this reasoning, we should not
THE VALLEY OF OAXACA 221
be surprised to find evidence of at least
some state institutions at Late MA I Tilca-
jete. Consider Structure 7, the aforemen-
on detailed field drawings made by E. Redmond, J. Sherman, C. Glew, and L. Villamil in 1997–1999;
tioned palatial residence in Area I (Mound
I), on the north side of the El Palenque
plaza (Fig. 17). We have dated Structure 7
securely to Late MA I. Its stone foundations
measured 16 m by 16 m and comprised
eight rooms arranged around an interior
patio; on the western side of the patio was a
two-chambered hearth. The stone founda-
tions of the structure originally supported
walls made of adobe bricks, some of which
were preserved by the fire that coincided
orientation is relative to magnetic north).
with the site’s abandonment. Structure 7 is
very similar in size and complexity to later
Classic period Zapotec palaces at Monte
Albán (Flannery 1983, 1998; Marcus and
Flannery 1996:208–211); however, the Late
MA I date of Structure 7 makes it the earli-
est example of a palatial residence thus far
excavated in the Oaxaca Valley.
In Area G (Mound G), on the eastern side
of the El Palenque plaza, we excavated
Structure 16 (Fig. 18), which consisted of
two large contiguous rooms (one measur-
ing 12.8 m by 2.35 m, and the other 9.8 m by
2.2 m), and two smaller rooms (measuring
222 SPENCER AND REDMOND
FIG. 17. Structure 7 in Area I of El Palenque (SMT-11b), San Martín Tilcajete (schematic plan, based
3.4 m by 2.2 m and 2.75 m by 2.2 m), one at
either end of the major rooms. All four
rooms of Structure 16 (like those of Struc-
detailed field drawings made by E. Redmond and A.
ture 7) had well-preserved stone founda-
tions. Although Structure 16 is somewhat
similar to the multiroom temples that have
been excavated in MA II (and later) con-
texts at San José Mogote and Monte Albán
(Flannery and Marcus 1976; Marcus and
Flannery 1996: 182), Structure 16 dates to
Late MA I. It may therefore be our oldest
excavated example of a multiroom temple
in the Oaxaca Valley. Like Structure 7,
Structure 16 shows signs of having been
burned at the time of its abandonment.
Despite the overall similarity between
Balkansky in 1998; orientation is relative to magnetic
north).
the plaza layouts at El Palenque and El
Mogote, only El Palenque contained exam-
ples of a multiroom temple and a palace. At
El Mogote, our excavations exposed a
three-room high-status residence at Mound
A, on the plaza’s northern side, and a one-
room temple structure at Mound K, on the
plaza’s eastern side (Fig. 15). Both struc-
tures date to Early MA I, and neither corre-
sponds to the architectural forms that have
been linked to the key institutions of the
later Zapotec state (Flannery 1983, 1998;
Flannery and Marcus 1976). The Tilcajete
data therefore provide evidence for the ap-
pearance of state institutions during Late
MA I but not earlier.
Valley-Wide Integration
After the abandonment of El Palenque,
the focus of occupation in the Tilcajete area
moved to Los Mogotes (SMT-23), a hilltop
site about 1 km to the north (Fig. 14). Los
Mogotes has a substantial MA II occupation
but only a small (0.12-ha) plaza surrounded
by three mounds. Elson’s continuing inves-
tigations are aimed, in part, at investigating
whether Los Mogotes became a secondary
center of the MA II state (Elson 1999). Mar-
cus and Flannery (1996: 172–178) have sug-
gested that all major subregions of the Oax-
aca Valley were politically integrated under
Monte Albán’s control by MA II. If so, then
we would expect our contextual analysis of
the regional settlement pattern data to find
evidence of a lessening of inter-subregional
selection (a consequence of less competition
among subregions) between Late MA I and
MA II.
Let us test this expectation of weakening
inter-subregional selection by carrying out
an analysis of covariance (ANCOVA) on all
the sites in the three branches of the Valley,
with the exception of the first-order re-
gional center, Monte Albán. If we subdivide
the localities into the same subregions that
we used for Early MA I and Late MA I, and
use subregional membership as the factor,
THE VALLEY OF OAXACA 223
POLITICAL EVOLUTION INFIG. 18. Structure 16 in Area G of El Palenque (SMT-
11b), San Martín Tilcajete (schematic plan, based on
the natural logarithm of archaeological
population as the dependent variable, and the
agricultural productivity (potential popula-
D
then declines in Monte Albán II. The analysis implies
tion) as the covariate, the resulting AN-
COVA for MA II is associated with a 0.404
level of statistical significance. As depicted
in Fig. 19, this result implies that there was
less intense selection among subregions in
MA II than during Late MA I, Early MA I,
or the Rosario phase. Such a weakening of
inter-subregional selection, we suggest,
was a consequence of the political integra-
tion of all subregions of the Oaxaca Valley
that was achieved by MA II. It is undoubt-
edly significant that this Valley-wide inte-
gration took place immediately after the
phase (Late MA I) that showed the
strongest evidence of inter-subregional se-
lection. Moreover, as we have seen, Late
MA I was also the phase with the strongest
evidence of hostilities at Tilcajete, the subre-
gional center of Ocotlán/Zimatlán. We
therefore propose that Monte Albán made
considerable use of aggressive action as it
that Late Monte Albán I is the phase with the greatest
amount of inter-subregional competition.
224 SPENCER AN
FIG. 19. Plot of ANCOVA significance levels,
Rosario phase through Period II. The significance val-
ues on the y-axis are presented in reverse order to em-
phasize that the intensity of inter-subregional selection
increases as the significance levels decrease. The plot
shows that selection among subregions increases from
the Rosario phase through Late Monte Albán I, and
sought to integrate the major subregions of
the Oaxaca Valley into the early Zapotec
state.
CONCLUSION: THE SHAPE OF THEEARLY ZAPOTEC STATE
Taken together, the findings of the Tilca-
jete Project and the contextual analysis of
the regional settlement pattern data lead to
two major conclusions: (1) the Zapotec state
emerged during Late Monte Albán I, in the
context of Monte Albán’s attempts to sub-
due and incorporate competing, rival poli-
ties in the Oaxaca Valley; and yet, (2) all
three major subregions of the Oaxaca Valley
were not integrated into a single state until
Monte Albán II. These conclusions are con-
sistent with the Rival Polity Model of MA I
political organization in the Oaxaca Valley
(Feinman 1998; Marcus and Flannery
1996:163). During Late MA I, Monte Albán
evidently presided over an early state
polity that included the Etla/Central subre-
gion, while Tlacolula and Ocotlán/Zi-
matlán continued as rival polities that were
also adopting some of the trappings of state
organization, perhaps in response to the de-
velopments at Monte Albán.
It is important to recognize that, even
though the Tlacolula and Ocotlán/Zimatlán
subregions appear to have maintained their
autonomy until MA II, Monte Albán was al-
ready expanding its political domain north-
ward during Late MA I. Abundant evidence
of a hostile Zapotec intrusion was recovered
by the authors in the Cañada de Cuicatlán, a
small canyon some 50 km north of the Etla
branch of the Oaxaca Valley (Redmond 1983;
Spencer 1982; Spencer and Redmond 1997).
Recent radiocarbon analyses date the onset
of the Zapotec domination of the Cañada at
300 B.C., corresponding to the interface be-
tween the Perdido phase (750-300 B.C.) and
the Lomas phase (300 B.C.-A.D. 200); the
new dates thus yield a close alignment of the
Lomas phase with the Late Monte Albán I
and Monte Albán II phases, combined
(Spencer and Redmond 2001). In the Cañada
REDMOND
locality of Dominguillo, the Perdido phase
village site of Llano Perdido was attacked
and burned. The site was completely aban-
doned and the local people were shifted to a
nearby ridge (the site of Loma de La Coy-
otera), where major alterations ensued in
their social, economic, political, and reli-
gious organization. Indigenous institutions
were eliminated and replaced by elements
compatible with Zapotec imperial interests.
A dozen Cañada communities in all received
such treatment at the hands of the Zapotec,
who dominated the canyon throughout the
Lomas phase. The Zapotec aimed military
and diplomatic actions at other regions as
well, eventually subduing a 20,000 square-
kilometer area around the Oaxaca Valley,
from the northern limits of the Cañada to
nearly the Pacific coast (Marcus and Flan-
nery 1996: 206–207); they boasted of their
territorial limits in the “conquest slab” in-
scriptions of Building J, erected in the Main
Plaza of Monte Albán in MA II (Caso 1947;
Marcus 1976, 1980, 1992a: 394–400). To incor-
porate such a large area, Monte Albán of
course needed skilled diplomacy, but, per-
haps more importantly, it also needed to de-
velop a powerful military and an internally
specialized administration capable of man-
aging far-flung tributary provinces through
the effective delegation of authority—in
short, a state form of government. At the
same time, the expansionist strategy would
have brought considerable resources and
prestige to Monte Albán’s ruling elite, thus
underwriting and legitimizing the necessary
military and administrative transformation
(Spencer 1990, 1998).
What is especially relevant to the present
paper is the timing of the Zapotec interven-
tion in the Cañada. In our Lomas phase de-
posits at Loma de La Coyotera we excavated
many examples of Type G.12 (combed bot-
tom) bowls—along with numerous thin-
walled (Late MA I/MA II) cremas such as C.6,
C.7, C. 13, and C.20. By contrast, in our exca-
vations at the earlier village of Llano Per-
dido, we found no examples of the G.12
POLITICAL EVOLUTION IN
(combed bottom) bowl—nor any fragments
of the C.6, C.7, C13, or C.20 types (Spencer
and Redmond 1997:163–164, 183–189, Table
4.2). This probably means that the Cañada in-
tervention occurred at about the same time
as the attack on the El Mogote sector at Tilca-
jete.2 While the entity responsible for both at-
tacks was most likely the Monte Albán polity,
the outcomes of the two actions were quite
different. The attack on El Mogote (at the
Early MA I/Late MA I interface) evidently
did not result in a dramatic alteration of local
institutions, although it did convince the
Tilcajete polity to rebuild their plaza in a
higher, more defensible location at the El
Palenque site. The local population grew,
and a substantial habitation zone was built at
El Palenque, joining the residential sector of
El Mogote (most of which continued in use).
The later attack on El Palenque, which came
at the end of Late MA I, was far more devas-
tating, and the El Palenque plaza along with
the residential sectors at both El Palenque
and El Mogote were completely abandoned.
After this attack, the Ocotlán/Zimatlán sub-
region was integrated into the growing do-
main of the Monte Albán state.
Why did the earlier attack against El
Mogote fail to bring the Ocotlán/Zimatlán
subregion under Monte Albán’s rule, while
the campaign against Llano Perdido and
other Cañada villages met with such suc-
cess? Even though the Cañada is much far-
ther from Monte Albán than the Tilcajete
area (80 km vs 25 km), the lower population
of the Cañada would have made it an easier
target for Monte Albán. The total occupa-
tion area of all 12 Perdido-phase sites in the
Cañada was 37.4 ha (Redmond 1983: 64),
whereas the Early MA I occupation just at
the El Mogote sector of Tilcajete covered
52.8 ha, and additional supporters presum-
ably could have been mustered from
nearby subject villages. The Ocotlán/Zi-
matlán subregion at the end of Early MA I
was simply a tougher foe than the Cañada.
By the end of Late MA I, however, the
forces from Monte Albán—having gained
THE VALLEY OF OAXACA 225
experience in militaristic statecraft and
grown substantially in population size—re-
turned to Tilcajete, and this time they fin-
In fact, since a few G.12 combed bottom sherds
ished the job. They burned the site to the
ground and (evidently) moved the surviv-
ing inhabitants to Los Mogotes (SMT-23),
which then may have become a secondary
center of the Monte Albán II state (Elson
1999; Marcus and Flannery 1996: 173–175).
Recent and ongoing investigations lend
support to the view that Monte Albán did
not expand its territory outward with grad-
ual, concentric regularity (see Marcus 1992b;
Spencer 1982: 256). Rather, the growth of the
Zapotec state was episodic and notably
asymmetric, the result of a series of preda-
tory initiatives that were aimed at areas both
near and far, with varying records of success.
Some distant but weak polities, like the
Cañada, were subjugated before other
226 SPENCER AND
closer—but stronger—ones like Tilcajete
were finally brought under Monte Albán’s
control.
ACKNOWLEDGMENTS
We are grateful to the following organizations for their
generous support of our Tilcajete Project: the National
Science Foundation (SBR-9303129), the Foundation for
the Advancement of Mesoamerican Studies, Inc., the
Heinz Family Foundation (Latin American Archaeology
Program), the National Geographic Society (Committee
for Research and Exploration), the University of Michi-
gan Museum of Anthropology, and the American Mu-
seum of Natural History. Permission to conduct the
Tilcajete fieldwork was granted by the Consejo de Ar-
queología, Instituto Nacional de Antropología e Historia
(INAH); Dr. Mari Carmen Serra Puche and Ing. Joaquín
García-Bárcena have served sequentially as president of
the Consejo de Arqueología during the years of the
Tilcajete Project. We thank Antrop. Eduardo López
Calzada, Director of the Centro INAH Oaxaca, as well as
Dra. Nelly Robles and Arqgo. Raúl Matadamas, also of
the Centro INAH Oaxaca, for their unwavering support
and friendship. Christina Elson, Luca Casparis, Jason
Sherman, Christopher Glew, Laura Villamil, Andrew
Balkansky, Scott Hutson, and Alan Covey have all di-
rected teams of excavators at various times over the past
five field seasons. Special thanks are due to Gary Fein-
man and Linda Nicholas for graciously making avail-
able to us the Oaxaca Settlement Pattern Project’s orig-
inal computer files containing the agricultural
productivity and population data by grid unit for the
Oaxaca Valley. For reading and commenting upon ear-
lier versions of this paper, we thank Kent Flannery,
Joyce Marcus, Gary Feinman, Linda Nicholas, Robert
Carneiro, Christina Elson, and Alan Covey. We are also
grateful to John O’Shea and the anonymous reviewers
REDMOND
for their helpful criticisms and suggestions. Bridget
Thomas, of the Division of Anthropology at the Amer-
ican Museum of Natural History, produced the illus-
trations.
NOTES
1Caso et al. (1967) divided Monte Albán’s pottery
into four basic paste colors, gris (gray), café (brown),
crema (cream), and amarillo (yellow). Within each paste
color, individual pottery types were given numbers,
such as C.2 (“crema type no. 2”).2
were found in the abandonment layer of El Mogote, it
is possible that the attack on Llano Perdido occurred a
bit earlier than the El Mogote raid.
REFERENCES CITED
Arnold, A. J., and K. Fristrup
1982 The theory of evolution by natural selection: A
hierarchical expansion. Paleobiology 8:113–129.
Barton, C. Michael, and Geoffrey A. Clark (Editors)
1998 Rediscovering Darwin: Evolutionary theory andarchaeological explanation. American Anthropo-
logical Association, Archaeological Papers 7.
Blanton, Richard E.
1978 Monte Albán: Settlement patterns at the ancientZapotec capital. Academic Press, New York.
Blanton, Richard E., Stephen A. Kowalewski, Gary M.
Feinman, and Jill Appel
1982 Monte Albán’s hinterland, part I: Prehispanic set-tlement patterns of the central and southern partsof the Valley of Oaxaca, Mexico. University of
Michigan Museum of Anthropology, Memoirs
15.
Blanton, Richard E., Stephen A. Kowalewski, Gary M.
Feinman, and Laura Finsten
1993 Ancient Mesoamerica: A comparison of change inthree regions. Cambridge Univ. Press, Cam-
bridge.
Blanton, Richard E., Gary M. Feinman, Stephen A.
Kowalewski, and Linda M. Nicholas
1999 Ancient Oaxaca: The Monte Albán state. Cam-
bridge Univ. Press, Cambridge.
Bonner, John T.
1988 The evolution of complexity. Princeton Univ.
Press, Princeton.
Boyd, Lawrence H., Jr., and Gudmund R. Iverson
1979 Contextual analysis: Concepts and statistical tech-niques. Wadsworth, Belmont, CA.
Boyd, Robert, and Peter Richerson
1985 Culture and the evolutionary process. Univ. of
Chicago Press, Chicago.
Brumfiel, Elizabeth M.
1976 Regional growth in the eastern Valley of Mex-
ico: A test of the “population pressure” hy-
pothesis. In The early Mesoamerican village,
edited by Kent V. Flannery, pp. 234–249. Acad-
emic Press, New York.
Carneiro, Robert L.
1970 A theory of the origin of the state. Science169:733–738.
1978 Political expansion as an expression of the
principle of competitive exclusion. In Originsof the state: The anthropology of political evolution,
edited by R. Cohen and E. R. Service, pp.
205–223. Institute for the Study of Human Is-
sues, Philadelphia.
1981 The chiefdom: Precursor of the state. In The tran-sition to statehood in the New World, edited by
Grant Jones and Robert Kautz, pp. 37–79. Cam-
bridge Univ. Press, Cambridge.
1992 The role of natural selection in the evolution of
culture. Cultural Dynamics 5:113– 140.
Caso, Alfonso
1947 Calendario y escritura de las antiguas culturas
de Monte Albán. In Obras completas de MiguelOthón de Mendizábal, Vol. 1. Mexico.
Caso, Alfonso, Ignacio Bernal, and Jorge Acosta
1967 La cerámica de Monte Albán. Memorias del Insti-
tuto Nacional de Antropología e Historia 13,
Mexico.
Crumley, Carole L.
1995 Heterarchy and the analysis of complex soci-
eties. In Heterarchy and the analysis of complexsocieties, edited by Robert Ehrenreich, Carole
Crumley, and J. E. Levy, pp. 1–5. American An-
thropological Association, Archaeological Pa-
pers 6.
Dawkins, Richard
1976 The selfish gene. Oxford Univ. Press, New York.
1982 The extended phenotype: The long reach of thegene. Oxford Univ. Press, New York.
Dunnell, Robert C.
1980 Evolutionary theory and archaeology. Ad-vances in Archaeological Method and Theory3:35–99.
1989 Aspects of the application of evolutionary the-
ory in archaeology. In Archaeological thought inAmerica, edited by C. C. Lamberg-Karlovsky,
pp. 35–49. Cambridge Univ. Press, Cambridge.
Eldredge, Niles
1985 Unfinished synthesis: Biological hierarchies andmodern evolutionary thought. Oxford Univ.
Press, New York.
POLITICAL EVOLUTION IN
1995 Reinventing Darwin: The great debate at the hightable of evolutionary theory. Wiley, New York.
Elson, Christina
1999 Excavations at Los Mogotes, San Martín Tilca-
jete, Oaxaca: A Terminal Formative subre-
gional center in the Valley of Oaxaca. Project
proposal to the Foundation for the Advance-
ment of Mesoamerican Studies, Inc.
Feinman, Gary M.
1982 Patterns in ceramic production and distribution,
periods Early I through V. In Monte Albán’s hin-terland, part I: The prehispanic settlement patterns ofthe central and southern parts of the Valley of Oax-aca, Mexico, edited by Richard E. Blanton,
Stephen Kowalewski, Gary Feinman, and Jill
Appel, pp. 181–206. University of Michigan Mu-
seum of Anthropology, Memoirs 15.
1998 Scale and social organization: Perspectives on
the ancient state. In Archaic states, edited by
Gary Feinman and Joyce Marcus, pp. 95–133.
School of American Research Press, Santa Fe.
Feinman, Gary M., and Linda M. Nicholas
1990 Settlement and land use in ancient Oaxaca. In
Debating Oaxaca archaeology, edited by Joyce
Marcus, pp. 71–113. University of Michigan
Museum of Anthropology, Anthropological
Papers 84.
Flannery, Kent V.
1972 The cultural evolution of civilizations. AnnualReview of Ecology and Systematics 3:399–426.
1983 The legacy of the Early Urban Period: An eth-
nohistoric approach to Monte Albán’s temples,
residences, and royal tombs. In The cloud peo-ple: Divergent evolution of the Zapotec and Mixteccivilizations, edited by Kent V. Flannery and
Joyce Marcus, pp. 132–136. Academic Press,
New York.
1986 Guilá Naquitz: Archaic foraging and early agricul-ture in Oaxaca, Mexico. Academic Press, Or-
lando.
1998 The ground plans of archaic states. In Archaicstates, edited by Gary M. Feinman and Joyce
Marcus, pp. 15–57. School of American Re-
search Press, Santa Fe.
Flannery, Kent V., and Joyce Marcus
1976 Evolution of the public building in Formative
Oaxaca. In Cultural change and continuity: Es-says in honor of James Bennett Griffin, edited by
Charles Cleland, pp. 205–221. Academic Press,
New York.
1983 The origins of the state in Oaxaca: Editors’ in-
troduction. In The cloud people: Divergent evolu-tion of the Zapotec and Mixtec civilizations,
edited by Kent Flannery and Joyce Marcus, pp.
79–83. Academic Press, New York.
THE VALLEY OF OAXACA 227
1990 Borrón y cuenta nueva: Setting Oaxaca’s archaeo-
logical record straight. In Debating Oaxaca archae-ology, edited by Joyce Marcus, pp. 17–69. Uni-
versity of Michigan Museum of Anthropology,
Anthropological Papers 84.
Goodnight, Charles J., James M. Schwartz, and Lori
Stevens
1992 Contextual analysis of models of group selec-
tion, soft selection, hard selection, and the evo-
lution of altruism. The American Naturalist140:743–761.
Gould, Stephen J.
1980 Is a new and general theory of evolution
emerging? Paleobiology 6:119–130.
1988 Trends as changes in variance: A new slant on
progress and directionality in evolution. Jour-nal of Paleontology 62:319–329.
Heisler, I. Lorraine, and John Damuth
1987 A method for analyzing selection in hierarchi-
cally structured populations. The AmericanNaturalist 130:582–602.
Kowalewski, Stephen A.
1976 Prehispanic settlement patterns of the central partof the Valley of Oaxaca, Mexico. Ph.D. disserta-
tion, Department of Anthropology, University
of Arizona, Tucson.
Kowalewski, Stephen A., Charles Spencer, and Elsa
Redmond
1978 Description of ceramic categories. In MonteAlbán: Settlement patterns at the ancient Zapoteccapital, edited by R. E. Blanton, pp. 167–193.
Academic Press, New York.
Kowalewski, Stephen A., Gary M. Feinman, Laura
Finsten, Richard E. Blanton, and Linda Nicholas
1989 Monte Albán’s hinterland, part II: Prehispanic set-tlement patterns in Tlacolula, Etla, and Ocotlán,the Valley of Oaxaca, Mexico. University of
Michigan Museum of Anthropology, Memoirs
23.
Lewontin, Richard C.
1970 The units of selection. Annual Review of Ecologyand Systematics 1:1–18.
Lyman, R. Lee, and Michael J. O’Brien
1998 The goals of evolutionary archaeology: His-
tory and explanation. Current Anthropology39:615–652.
Marcus, Joyce
1976 Iconography of militarism at Monte Albán and
neighboring sites in the Valley of Oaxaca. In
The origins of religious art and iconography in Pre-classic Mesoamerica, edited by H. B. Nicholson,
pp. 123–139. UCLA Latin American Center,
Los Angeles.
1980 Zapotec writing. Scientific American 242: 50–64.
1992a Mesoamerican writing systems: Propaganda,myth, and history in four ancient civilizations.
Princeton Univ. Press, Princeton.
1992b Dynamic cycles of Mesoamerican states. Na-
228 SPENCER AN
tional Geographic Research and Exploration8:392–411.
Marcus, Joyce, and Kent V. Flannery
1996 Zapotec civilization: How urban society evolved inMexico’s Oaxaca Valley. Thames and Hudson,
London.
McKinney, M. L.
1990 Trends in body-size evolution. In Evolutionarytrends, edited by K. J. MacNamara, pp. 75–118.
Univ. of Arizona Press, Tucson.
McShea, Daniel W.
1994 Mechanisms of large-scale evolutionary
trends. Evolution 48:1747–1763.
Newell, Norman D.
1949 Phyletic size increase, an important trend il-
lustrated by fossil invertebrates. Evolution3:103–124.
Nicholas, Linda
1989 Land use in prehispanic Oaxaca. In MonteAlbán’s hinterland, part II: Prehispanic settlementpatterns in Tlacolula, Etla, and Ocotlán, the Valleyof Oaxaca, Mexico, by Stephen Kowalewski,
Gary Feinman, Laura Finsten, Richard Blan-
ton, and Linda Nicholas, pp. 449–505. Univer-
sity of Michigan Museum of Anthropology,
Memoirs 23.
Redmond, Elsa M.
1983 A fuego y sangre: Early Zapotec imperialism in theCuicatlán Cañada, Oaxaca. University of Michi-
gan Museum of Anthropology, Memoirs 16.
1998 Introduction: The dynamics of chieftaincy and
the development of chiefdoms. In Chiefdomsand chieftaincy in the Americas, edited by Elsa
M. Redmond, pp. 1–17. Univ. of Florida Press,
Gainesville.
Rindos, David
1984 The origin of agricultural systems: An evolution-ary perspective. Academic Press, New York.
1989 Undirected variation and the Darwinian ex-
planation of culture change. In Archaeologicalmethod and theory, edited by Michael Schiffer,
Vol. 1, pp. 1–45. Univ. of Arizona Press, Tuc-
son.
Rosenberg, Michael
1994 Pattern, process, and hierarchy in the evolu-
tion of culture. Journal of Anthropological Ar-chaeology 13:307–340.
Sanders, William T.
1974 Chiefdom to state: Political evolution at Kami-
naljuyú, Guatemala. In Reconstructing complexsocieties: An archaeological colloquium, edited by
Charlotte Moore, pp. 97–116. Supplement to
the Bulletin of the American Schools of OrientalResearch 20.
Saunders, P. T., and M. W. Ho
D REDMOND
1976 On the increase in complexity in evolution.
Journal of Theoretical Biology 63:375–384.
Seeley, T.
1989 The honey bee colony as a superorganism.
Teltser, Patrice (Editor)
1995 Evolutionary archaeology: Methodological issues.
POLITICAL EVOLUTION IN THE VALLEY OF OAXACA 229
American Scientist 77:546–553.
Sober, Elliott, and David S. Wilson
1998 Unto others: The evolution and psychology of un-selfish behavior. Harvard Univ. Press, Cam-
bridge.
Spencer, Charles S.
1982 The Cuicatlán Cañada and Monte Albán: A studyof primary state formation. Academic Press, New
York.
1987 Rethinking the chiefdom. In Chiefdoms in theAmericas, edited by Robert D. Drennan and
Carlos A. Uribe, pp. 369–390. Univ. Press of
America, Lanham, MD.
1990 On the tempo and mode of state formation:
Neoevolutionism reconsidered. Journal of An-thropological Archaeology 9:1–30.
1991 Coevolution and the development of Venezue-
lan chiefdoms. In Profiles in Cultural Evolution:Papers from a Conference in Honor of Elman R. Ser-vice, edited by A. Terry Rambo and Kathleen
Gillogly, pp. 137–165. University of Michigan
Museum of Anthropology, Anthropological Pa-
pers 85.
1997 Evolutionary approaches in archaeology. Jour-nal of Archaeological Research 5:209–264.
1998 A mathematical model of primary state forma-
tion. Cultural Dynamics 10:5–20.
1999 Palatial digs. Natural History 108(2):94–95.
Spencer, Charles S., and Elsa M. Redmond
1997 Archaeology of the Cañada de Cuicatlán, Oaxaca.
American Museum of Natural History, An-
thropological Papers 80.
2001 The chronology of conquest: Implications of
new radiocarbon analyses from the Cañada de
Cuicatlán, Oaxaca. Latin American Antiquity (in
press)
SPSS, Inc.
1998 SYSTAT 8.0: Statistics. SPSS Science Marketing
Department, Chicago.
Stanley, Steven M.
1973 An explanation for Cope’s rule. Evolution27:1–26.
Steponaitis, Vincas
1981 Settlement hierarchies and political complex-
ity in nonmarket societies: The Formative pe-
riod in the Valley of Mexico. American Anthro-pologist 83:320–363.
Univ. of Arizona Press, Tucson.
Wilkinson, Leland, Grant Blank, and Christian Gruber
1996 Desktop data analysis with SYSTAT. Prentice
Hall, Saddle River, NJ.
Williams, George C.
1966 Adaptation and natural selection: A critique ofsome current evolutionary thought. Princeton
Univ. Press, Princeton.
1992 Natural selection: Domains, levels, and challenges.
Oxford Univ. Press, New York.
Wilson, David Sloan
1975 A general theory of group selection. Proceed-ings of the National Academy of Sciences72:143–146.
1976 Evolution on the level of communities. Science192:1358–1360.
1980 The natural selection of populations and communi-ties. Benjamin/Cummings, Menlo Park, CA.
1983 The group selection controversy: History and
current status. Annual Review of Ecology andSystematics 14:159–187.
1989 Levels of selection: An alternative to individu-
alism in biology and the human sciences. So-cial Networks 11:257–272.
1992 Complex interactions in metacommunities,
with implications for biodiversity and higher
levels of selection. Ecology 73:1984– 2000.
Wilson, David S., and Elliott Sober
1989 Reviving the superorganism. Journal of Theoret-ical Biology 136:337–356.
1994 Reintroducing group selection to the human
behavioral sciences. Behavioral and Brain Sci-ences 17:585–654.
Wright, Henry T.
1977 Recent research on the origin of the state. An-nual Review of Anthropology 6:379–397.
1984 Prestate political formations. In On the evolu-tion of complex societies: Essays in honor of HarryHoijer 1982, edited by Timothy K. Earle, pp.
41–77. Undena Press, Malibu.
Wynne-Edwards, V. C.
1962 Animal dispersion in relation to social behavior.
Oliver and Boyd, Edinburgh.
1986 Evolution through group selection. Blackwell Sci-
entific, Oxford.