ArticlePDF Available

Abstract and Figures

The Atlas of Cultural Evolution provides basic data on the evolution of cultural complexity using the Outline of Archaeological Traditions sample. The Outline of Archaeological Traditions constitutes a sampling universe from which cases can be drawn for diachronic cross-cultural research, an activity I refer to as archaeoethnology. Data for the Atlas were drawn from entries in the Encyclopedia of Prehistory, a nine volume work providing summary information on all cases in the Outline of Archaeological Traditions, thus the Atlas also demonstrates the utility of the Encyclopedia of Prehistory as a basic tool for archaeoethnology. I suggest that a more sophisticated tool for archaeoethnology, the eHRAF Collection of Archaeology, be used to further test and refine the cultural evolutionary trends put forward here.
Content may be subject to copyright.
world cultures
Archaeo home
World Cultures home
J. Patrick Gray, Editor
Contents and How to Use This Issue J. Patrick Gray
Atlas of Cultural Evolution
Peter N. Peregrine
1. Archaeoethnology 1
2. Cultural Evolution 10
3. Toward Explaining Cultural Evolution 22
4. References Cited 32
5. Maps 34
6. Codebook 48
7. Descriptive Statistics 54
8. Correlations 75
World Cultures CD Data Disk William Divale 89
Journal of Comparative and Cross-Cultural Research Vol 14 No 1 Fall 2003
2003 World Cultures 14(1): 1
Contents And How To Use This Issue
This entire issue of WORLD CULTURES is devoted to Peter Peregrine’s Atlas of Cultural
Evolution. The Atlas is a major resource for archaeoethnological research. The data are on
the accompanying CD in the ACE.SAV SPSS file. The CD also contains the MapMaker
Gratis program discussed in Peregrine’s article. The compressed file mmZip.exe is in the
\Map subdirectory of the CD, along with the data files needed to reproduce the maps in the
---J. Patrick Gray
2003 World Cultures 14(1): 2-88
Part B: Live link to
Atlas of Cultural Evolution
Peter N. Peregrine
Department of Anthropology, Lawrence University, Appleton, WI 54911;
The Atlas of Cultural Evolution provides basic data on the evolution of cultural complexity using the Outline of
Archaeological Traditions sample. The Outline of Archaeological Traditions constitutes a sampling universe
from which cases can be drawn for diachronic cross-cultural research, an activity I refer to as
archaeoethnology. Data for the Atlas were drawn from entries in the Encyclopedia of Prehistory, a nine volume
work providing summary information on all cases in the Outline of Archaeological Traditions, thus the Atlas
also demonstrates the utility of the Encyclopedia of Prehistory as a basic tool for archaeoethnology. I suggest
that a more sophisticated tool for archaeoethnology, the eHRAF Collection of Archaeology, be used to further
test and refine the cultural evolutionary trends put forward here.
Comparative research is a necessary tool in evolutionary science. It is only through
comparison that we can identify diversity, and it is the creation and maintenance of diversity
that evolutionary science attempts to understand. Within anthropology, comparative research
is usually called cross-cultural research or ethnology.
The unit of analysis in such research is the culture. What constitutes a culture is rather
loosely defined, but includes sharing a common language, a common economic and socio-
political system, and some degree of territorial continuity. Because any given population
within a culture will show some divergence from the others, a culture is usually represented
by a particular community, and because cultures are always changing, the representative or
focal community is described as of a particular point in time.
Cross-cultural research makes two fundamental assumptions. First, that a culture can be
adequately represented by a single community. And second, that cultures can be compared.
The first assumption is based on the idea that any definition of culture will be broad enough
that any given community in a culture will share fundamental features of behavior and
organization with others similarly defined. The second is based on the uniformitarian
assumption underlying all evolutionary science: if an explanation accurately reflects reality,
“measures of the presumed causes and effects should be significantly and strongly associated
synchronically” (Ember and Ember 1995:88).
Archaeoethnology attempts to extends traditional cross-cultural research in two dimensions.
First, it attempts to add new cases to those which can be used for comparison, and hence
increases the sample size for cross-cultural research (but see section 1.C below). Second, and
perhaps more importantly, archaeoethnology attempts to provide a way to determine whether
the presumed cause of some phenomenon actually precedes its presumed effects. Like all
forms of comparative research, archaeoethnology seeks to identify regular associations
between variables and to test explanations for why those associations exist. Unlike ethnology
using extant or recent cultures, the associations identified through archaeoethnology can be
either synchronic or diachronic, and the explanations for them can be tested both
synchronically and diachronically.
Because of its ability to identify and test explanations diachronically, archaeoethnology is
uniquely suited to exploring both unilinear and multilinear trends in cultural evolution.
Unilinear trends refer to either progressive or regressive changes in societal scale,
complexity, and integration that take place over a long period of time and large geographical
areas. Archaeoethnology can examine change over a long period of time to determine
empirically whether unilinear trends are present, and test explanations for those trends by
determining whether presumed causes actually precede presumed effects. Similarly,
multilinear evolutionary processes, those that create the specific features of different
societies within the larger, unilinear trends, can be tested diachronically to see if presumed
causes precede assumed effects. Such research is perhaps best carried out using the eHRAF
Collection of Archaeology.
The diachronic nature of archaeoethnology also makes it uniquely suited to exploring
patterns of migration, innovation, and diffusion, and to investigating the roles of these
processes in cultural evolution. A synchronic study of a given region might suggest that a
trait diffused through cultures in a region, and perhaps might suggest the source and path of
the diffused trait. Only a diachronic study can demonstrate diffusion empirically, pinpoint
the source of a given trait, and chart the path of its diffusion through time.
In short, the purpose of archaeoethnology is to establish and explain long-term processes of
cultural stability and change.
A. The Outline of Archaeological Traditions
The Outline of Archaeological Traditions (Peregrine 2001a) was designed to serve as a
sampling universe for archaeoethnology. Such a sampling universe must meet several
conditions. First, and perhaps most importantly, the cases included must all be equivalent on
some set of defining criteria, and those criteria must be sensitive enough to variables of
interest that patterns within and among them can be recognized. The Outline of
Archaeological Traditions, as the name suggests, uses “archaeological traditions” as the
units of analysis. These were designed to serve as basic units for archaeoethnology, and are
defined as a group of populations sharing similar subsistence practices, technology, and
forms of socio-political organization, which are spatially contiguous over a relatively large
area and which endure temporally for a relatively long period. Minimal area coverage for an
archaeological tradition can be thought of as something like 100,000 square kilometers;
while minimal temporal duration can be thought of as something like five centuries.
However, these figures are meant to help clarify the concept of an archaeological tradition,
not to formally restrict its definition to these conditions.
Archaeological traditions are not equivalent to cultures in an ethnological sense because, in
addition to socio-cultural defining characteristics, archaeological traditions have both a
spatial and a temporal dimension. Ethnographic cultures are assumed to exist simultaneously
in an “ethnographic present,” and hence lack a temporal dimension. Archaeological
traditions have a temporal dimension. Archaeological traditions are also defined by a
somewhat different set of socio-cultural characteristics than ethnological cultures.
Archaeological traditions are defined in terms of common subsistence practices, socio-
political organization, and material industries, but language, ideology, kinship ties, and
political unity play little or no part in their definition, since they are virtually unrecoverable
from archaeological contexts. In contrast, language, ideology, and cross-cutting ties are
central to defining ethnographic cultures.
The concept of archaeological tradition as it is used in the Outline of Archaeological
Traditions was influenced by, but is also not equivalent to, the concept of archaeological
tradition as used by Gordon Willey and Philip Phillips (1958:37), which they define as “a
(primarily) temporal continuity represented by persistent configurations in single
technologies or other systems of related forms.” The emphasis for Willey and Phillips is on
the temporal dimension, (Willey and Phillips [1958:33] use the concept of “horizon” to
express the spatial dimension of archaeological traditions) and the focus is on technology
(most frequently pottery) rather than broader socio-cultural characteristics. Once again,
archaeological tradition as it is used in the Outline of Archaeological Traditions has both a
spatial and temporal dimension, and is defined primarily by socio-cultural characteristics.
A valid sampling universe must also include all possible cases, and the Outline of
Archaeological Traditions is a catalogue of all known archaeological traditions, covering the
entire globe and the entire prehistory of humankind. The Outline of Archaeological
Traditions begins its coverage with the origins of our genus, Homo, approximately two
million years ago in Africa. Homo spread throughout Eurasia by 500,000 years ago, into
Oceania by 40,000 years ago, and into the Americas by 12,000 years ago. Area coverage for
those regions begins when humans first enter them. The ending date of the Outline’s
coverage also varies by region. In Oceania, the Americas, and Sub-Saharan Africa, coverage
ends at approximately 500 BP with European exploration and initial colonization. In Central
Asia coverage ends with the rise and spread of nomadic states such as the Hsuing-Nu ca.
1500 BP. In Europe coverage ends with the expansion of the Roman Empire ca. 2000 BP. In
China coverage ends with the Shang dynasty, ca. 3100 BP. And in Northern Africa and the
Middle East, coverage ends with the rise of the New Babylonian and Old Kingdom Egyptian
civilizations ca. 3500 BP.
A good sampling universe must also be large enough to allow random samples for
hypothesis tests to be drawn from it, taking into account the loss of cases due to missing
data, yet small enough to allow basic information for stratified or cluster sampling, or for
eliminating cases with specific characteristics. In its current form the Outline of
Archaeological Traditions contains 289 cases--easily large enough to provide a good
universe for sampling. It should be taken as a catalogue in process, which will be continually
revised and updated as new information about human prehistory is generated, and as existing
information is synthesized and reinterpreted. The Outline of Archaeological Traditions is
also small enough that basic information for each case has been collected and published in
the Encyclopedia of Prehistory (Peregrine and Ember, eds. 2001-2002). It is information
from the Encyclopedia of Prehistory that was used to create the data set presented here.
B. The Atlas of Cultural Evolution Data Set
The data set presented and analyzed here is based on a revision of Murdock and Provost's
(1973) ten-item Cultural Complexity Scale. The original scale items were each comprised of
five-point scales, while Table 1.B.1 shows that for the Atlas of Cultural Evolution data set
these variables were recoded into three-point scales (also see Peregrine 2001b). The reason
for this revision was to make coding easier with archaeological data. The five-point scales
required too much inference from the available archaeological record, while the three-point
scales made coding decisions considerably easier. The scale items are summed for each case
to create its total score for Cultural Complexity.
All coding was done by the author from entries for the Encyclopedia of Prehistory as they
were received for pre-publication review and editing. Thus cases were coded in a haphazard
manner. This procedure should have eliminated any bias from coding cases in a
predetermined order (such as oldest to most recent), or systematic inter-coder errors (lack of
reliability). It must be noted that these revised scales have not been evaluated for reliability,
so that if future coding is done to add cases to the data set, a reliability study should be
performed simultaneously. It should also be noted that coding relied exclusively on
information provided in the Encyclopedia of Prehistory entries. Since these were written by
over 200 scholars representing more than 20 foreign nations, it is highly unlikely that any
systematic bias due to a particular theoretical perspective or political orientation is present
(cf. Shanks and Tilley 1992:245). Basic descriptive statistics for the cases are provided in
Part 7.
In addition to the scale items, a number of basic identification and pinpointing variables are
also included in the Atlas of Cultural Evolution data set. These include the tradition name;
start, end, and midpoint dates; locational information; and time-series variables. These are
presented in the Codebook in Part 6. Maps are also provided in Part 5 to show the location of
each archaeological tradition, along with digital files and the MapMaker Gratis software
package, allowing scholars to employ a Geographic Information System in the examination
of these data.
Table 1.B.1--Scales Comprising the Murdock & Provost (1973) Index of Cultural Complexity,
Recoded for Use with Archaeological Cases
Scale 1: Writing and Records
1 = None
2 = Mnemonic or nonwritten records
3 = True writing
Scale 2: Fixity of Residence
1 = Nomadic
2 = Seminomadic
3 = Sedentary
Scale 3: Agriculture
1 = None
2 = 10% or more, but secondary
3 = Primary
Scale 4: Urbanization (largest settlement)
1 = Fewer than 100 persons
2 = 100--399 persons
3 = 400+ persons
Scale 5: Technological Specialization
1 = None
2 = Pottery
3 = Metalwork (alloys, forging, casting)
Scale 6- Land Transport
1 = Human only
2 = Pack or draft animals
3 = Vehicles
Scale 7- Money
1 = None
2 = Domestically usable articles
3 = Currency
Scale 8- Density of Population
1 = Less than 1 person/square mile
2 = 1--25 persons/square mile
3 = 26+ persons/square mile
Scale 9- Political Integration
1 = Autonomous local communities
2 = 1 or 2 level above community
3 = 3 or more levels above community
Scale 10- Social Stratification
1 = Egalitarian
2 = 2 social classes
3 = 3 or more social classes or castes
C. The Atlas of Cultural Evolution and the Standard Cross-
Cultural Sample
It must be emphasized that because the units of analysis are different, it would not be valid to
select cases for comparison from both the Atlas of Cultural Evolution (ACE) and a list of
ethnographic cases, such as the Standard Cross-Cultural Sample (SCCS)(Murdock and
White 1969). To illustrate this point, Table 1.C.1 presents a comparison of descriptive
statistics for the ACE and the SCCS coded on the same ten-item Cultural Complexity Scale
(Murdock and Provost 1973) as revised for archaeological use, while Table 1.C.2 presents
the results of Mann-Whitney U tests. Clearly differences are present.
Table 1.C.1 illustrates that means tend to be higher in the SCCS than the ACE, and Table
1.C.2 demonstrates that rank scores on all but one item of the Cultural Complexity Scale are
significantly higher in the SCCS. This makes sense because the ACE contains cases from
early in human prehistory, when all cases were non-complex; that is, they were low on all
measures of the Cultural Complexity Scale. Because of this, the ACE should be expected to
have scores on each scale item that are significantly lower than the SCCS. It is interesting
that the one scale item where scores were not significantly different was Social Stratification.
This may imply that the SCCS tends to under-represent socially stratified societies in the
contemporary world--a critique that has, indeed, been levied against the sample (e.g.
Otterbein 1976).
Table 1.C.1--Descriptive Statistics for ACE and SCCS
Variable ACE
Mean Std. Deviation SCCS
Mean Std. Deviation
Writing and Records 289 1.15 0.525 186 1.84 0.775
Fixity of Residence 289 2.30 0.856 186 2.48 0.744
Agriculture 289 2.09 0.955 186 2.44 0.812
Urbanization 289 1.71 0.815 186 1.99 0.771
Technological Specialization 289 1.94 0.770 186 2.27 0.787
Land Transport 289 1.35 0.623 186 1.54 0.698
Money 289 1.21 0.555 186 2.10 0.959
Density of Population 289 1.62 0.667 186 2.09 0.843
Political Integration 289 1.80 0.748 186 2.10 0.455
Social Stratification 289 1.77 0.814 186 1.81 0.691
Cultural Complexity 289 16.95 5.979 186 20.65 5.042
Table 1.C.2 -- Mann-Whitney U Statistics for ACE and SCCS
Variable Sample N Mean
Writing and Records ACE 289 191.30 13379.50 0.000
SCCS 186 310.57
Fixity of Residence ACE 289 229.10 24306.00 0.046
SCCS 186 251.82
Agriculture ACE 289 220.80 21905.00 0.000
SCCS 186 264.73
Urbanization ACE 289 219.82 21622.00 0.000
SCCS 186 266.25
Technological Specialization ACE 289 216.64 20704.00 0.000
SCCS 186 271.19
Land Transport ACE 289 224.24 22899.00 0.001
SCCS 186 259.39
Money ACE 289 193.88 14127.500 0.000
SCCS 186 306.55
Density of Population ACE 289 209.26 18570.00 0.000
SCCS 186 282.66
Political Integration ACE 289 214.75 20157.50 0.000
SCCS 186 274.13
Social Stratification ACE 289 233.45 25561.50 0.333
SCCS 186 245.07
Cultural Complexity ACE 289 204.04 17064.00 0.000
SCCS 186 290.76
If we compare the SCCS with the most recent cases in the ACE, those that were in existence
during the last 1000 years, we still find statistically significant differences, as demonstrated
in Table 1.C.3. While there are fewer significant differences (5 of 10 as opposed to 9 of 10
using all ACE cases), it is still clear that the two data sets are quite different.
Table 1.C.3 -- Mann-Whitney U Statistics for ACE cases 1000 years ago and
Variable Sample N Mean
Writing and Records ACE 77 79.75 3138.00 0.000
SCCS 186 153.63
Fixity of Residence ACE 77 143.16 6301.500 0.068
SCCS 186 127.38
Agriculture ACE 77 128.66 6904.00 0.590
SCCS 186 133.38
Urbanization ACE 77 131.04 7087.00 0.888
SCCS 186 132.40
Technological Specialization ACE 77 120.73 6293.00 0.094
SCCS 186 136.67
Land Transport ACE 77 100.45 4732.00 0.000
SCCS 186 145.06
Money ACE 77 88.25 3792.50 0.000
SCCS 186 150.11
Density of Population ACE 77 113.79 5759.00 0.008
SCCS 186 139.54
Political Integration ACE 77 126.01 6699.50 0.295
SCCS 186 134.48
Social Stratification ACE 77 141.01 6467.50 0.183
SCCS 186 128.27
Cultural Complexity ACE 77 109.52 5430.00 0.002
SCCS 186 141.31
This comparison is not intended to promote the use of one sample over the other; indeed,
quite the opposite. Each sample allows the researcher to generalize to a particular
population. The SCCS is designed to represent the range of variation in the cultures of recent
times, while the ACE is designed to represent the range of variation in the cultures of the
past. And things have changed. The cultures of the past were overall less complex, at least as
measured by these variables, than the cultures of recent times. These differences are present
because cultural evolution has taken place.
D. Conclusion
The ACE provides the first set of coded data for the OAT sample. The OAT itself is the first
statistically-valid sample of archaeological cases for comparative analysis. The OAT can be
thought of as roughly equivalent to the SCCS for comparative research, but the two should
not be used together, as the cases used in each are defined in very different ways. The
importance of the ACE data set is that it provides us with the opportunity to undertake
exercises in archaeoethnology; that is, comparative research on culture, behavior, and
evolution with both geographic and temporal dimensions.
Cultural evolution is conventionally defined as change in societal scale, complexity, and
integration (Blanton et al. 1981:17). Scale refers to the physical size of a society, measured
through population, geographical extent, or, more typically in archaeological and cross-
cultural research, through the size of the largest city (see McNett 1970). Complexity refers to
the number of different roles available in the society. Integration refers to the number of
interconnections between social roles. All three aspects of cultural evolution are captured in
Murdock and Provost’s (1973) Cultural Complexity Scale.
Societal scale is measured through Murdock and Provost’s scale items four (urbanization)
and seven (density of population). Gary Chick (1997) has also argued that societal scale is
one of two underlying factors that comprise the Cultural Complexity Scale, a factor built
from scale items four and seven, along with items two (fixity of residence) and three
(agriculture). Here I refer to this as the Scale Factor.
Societal complexity is measured through Murdock and Provost’s scale items five
(technological specialization) and ten (social stratification). It could be argued that
complexity is also related to items one (writing and records), six (land transport), seven
(money), and nine (political integration), as these typically require specialists. Gary Chick
(1997) has suggested these items form a second underlying factor within the Cultural
Complexity scale, which I refer to as the Technology Factor.
Finally, societal integration is measured through Murdock and Provost’s scale item nine
(political integration), and perhaps through items one (writing and records) and seven
(money). Unfortunately, there is little variation in the ACE on the latter two variables,
and thus they are of little use in the examination of societal integration.
A. Describing Evolutionary Trends
The key question here is whether clear evolutionary trends can be identified through the
ACE data. Table 2.A.1 suggests that evolutionary trends are present. Societal scale,
complexity, and integration are all significantly correlated with two different measures of
time. The first measure, Date, is the midpoint of time range within which a given case
existed. The second measure, Time Series End, is the endpoint of that time range, adjusted to
the nearest millennium. Only the last 12,000 years are included in the Time Series End
correlations. The two variables show correlations in opposite directions because Date is
measured in years before the present, and thus gets larger the farther back one goes into the
past, while Time Series End is a time series count that starts 12,000 years ago and gets larger
as one moves closer to the present.
Table 2.A.1 -- Spearman’s rho Correlation Coefficients Showing the Relationship between
Date and Cultural Complexity
Variable Date p Time Series
Writing and Records -0.065 0.274 -0.060 0.319
Fixity of Residence -0.557 0.000 0.345 0.000
Agriculture -0.484 0.000 0.250 0.000
Urbanization -0.512 0.000 0.313 0.000
Technological Specialization -0.576 0.000 0.367 0.000
Land Transport -0.068 0.249 -0.132 0.029
Money -0.185 0.002 0.108 0.074
Density of Population -0.420 0.000 0.217 0.000
Political Integration -0.555 0.000 0.317 0.000
Social Stratification -0.503 0.000 0.275 0.000
Cultural Complexity -0.572 0.000 0.324 0.000
Technology Factor -0.557 0.000 0.315 0.000
Scale Factor -0.554 0.000 0.317 0.000
Looking more closely at Table 2.A.1 we also see that three of the ten Cultural Complexity
Scale variables do not appear to be strongly correlated with either measure of time--Writing
and Records, Land Transport, and Money. All three are problematic when applied to the
ACE cases, as all measure fairly recent, and relatively rare (until the last century or so)
technological developments. Thus, there is very little variation in any of these variables until
the recent past, which can be readily seen in Figure 2.A.1, 2.A.2, and 2.A.3. These figures
present the means for each variable at 1000-year intervals over the past 12,000 years, and it
is clear that most of the cases for all three were coded 1, even in the recent past.
Figure 2.A.1 -- Mean of Writing and Records by Date
Thousands of Years Ago
Writing and Records
Figure 2.A.2 -- Mean of Transportation by Date
Thousands of Years Ago
Figure 2.A.3 -- Mean of Money by Date
Thousands of Years Ago
One other problem with the ACE data is that once a case achieves writing, it quickly moves
into history and, by definition, out of the archaeological record which the ACE is intended to
capture. Hence, it is not surprising that writing shows an arched curve, for when cases
develop writing they quickly move out of the sample so that over time all that are left are
those cases lacking writing. This trend is perhaps more clearly illustrated with Land
Transport, and is also evident for Money. It should not be surprising that these three
variables are highly inter-correlated, and appear to be directly related to those cases on the
cusp of the historical record. Because of these problems, I will not examine these variables in
Societal Scale
Changes in societal scale are perhaps best measured by the Urbanization and Population
Density variables. Graphs showing the means of these variables at 1000-year intervals for
the last 12,000 years are given in Figures 2.A.4 and 2.A.5. Means have clearly increased
over time, and in a roughly linear fashion; indeed, R-squared values for these two figures are
0.904 and 0.978, respectively. With these results one could argue that societal scale has
increased in a roughly linear fashion over the past 12,000 years. That is, cultural evolution in
terms of societal scale has taken a single, unilineal form (but see section 2.B below).
Figure 2.A.4 -- Mean of Urbanization by Date
Thousands of Years Ago
Figure 2.A.5 -- Mean of Population Density by Date
Thousands of Years Ago
Density of Population
Figure 2.A.6 -- Mean of Scale Factor by Date
Thousands of Years Ago
Scale Factor
Figure 2.A.6 displays means of the Scale Factor plotted at 1000-year intervals. The Scale
Factor, as explained above, is the sum of the Density of Population, Urbanization, Fixity of
Residence, and Agriculture variables. Not surprisingly, the Scale Factor also shows a linear
trend over time, with an R-squared value of 0.976. Again, it appears that societal scale has
increased in a roughly linear fashion over the past 12,000 years.
Societal Complexity
Societal complexity also appears to have increased in a roughly linear fashion over the past
12,000 years, as illustrated in Figures 2.A.7, 2.A.8, and 2.A.9. Figure 2.A.7 shows the mean
values of Technological Specialization plotted at 1000-year intervals, while Figure 2.A.8
shows Social Stratification. Both illustrate linear trends with R-squared values of 0.960 and
0.935, respectively. Figure 2.A.9 shows the mean values for the Technology Factor (which
sums the Technological Specialization, Social Stratification, Writing and Records, Land
Transport, Money, and Political Integration variables) plotted at 1000-year intervals. It, too,
illustrates a linear trend with an R-squared value of 0.949. It should be noted that the "dip" at
1000 years ago evident in each plot is probably due to the more complex cases being
dropped from the sample once they gain writing and become historic (this should be
particularly true in Figure 2.A.9, where the Writing variable is included in the Technology
Figure 2.A.7 -- Mean of Technological Specialization by Date
Thousands of Years Ago
Technological Specialization
Figure 2.A.8 -- Mean of Social Stratification by Date
Thousands of Years Ago
Social Stratification
Figure 2.A.9 -- Mean of Technology Factor by Date
Thousands of Years Ago
Technology Factor
Societal Integration
Finally, a linear trend is also evident in societal integration. Figure 2.A.10 shows the mean
value of Political Integration plotted at 1000-year intervals for the past 12,000 years. Here
the trend has an R-squared value of 0.956.
Figure 2.A.10 -- Mean of Political Integration by Date
Thousands of Years Ago
Political Integration
It seems clear that societal scale, complexity, and integration have all increased in a roughly
linear fashion over the past 12,000 years. Thus, there is clear evidence for unilineal trends in
cultural evolution such that societal scale, complexity, and integration all tend to increase
over time. The presence of these unilineal evolutionary trends clearly supports the validity of
cultural evolutionary research, and contradicts critiques made by scholars such as
Goldenwiser (1937), Lowie (1946), Nisbet (1969), and Giddens (1984) that research into
unilineal evolution is invalid because such unilineal trends cannot be demonstrated to exist.
These data illustrate that unilineal evolutionary trends do exist, and their existence begs the
question of why they exist.
Before turning to the question of why unilineal evolutionary trends exist, or rather, how one
might approach an answer to that question, I need to address two other issues: the problem of
autocorrelation and the interesting fact of Old World-New World differences in cultural
B. Autocorrelation
There is an interesting statistical problem in exploring cultural evolution with the ACE data
set, that of autocorrelation. Autocorrelation refers to a situation where two cases are not
statistically independent because changes in one case cause changes in the other. Cultural
evolution itself can be thought of as something of a serial autocorrelation process. Change in
an ancestral society leads to those changes being transmitted to descendants; thus, values of
a variable measuring that change will be serially autocorrelated when viewed over time. For
example, if members of an archaeological tradition develop metalwork, it is likely that
metalworking will be passed on to descendents. The ACE variable TECHNOLO will thus
present serial autocorrelation between the ancestral archaeological tradition and its
descendents, as the development of metalwork is causally linked to the descendent
populations having metalwork.
Autocorrelation is generally regarded as a problem in statistical analyses because it tends to
artificially inflate correlation and regression coefficients (Ostrom 1990:21-26). Thus,
because of autocorrelation, the R-squared value of 0.956 I reported for Figure 2.A.10 is
likely inflated. One way to estimate whether such a value is likely inflated from
autocorrelation is through the Durban-Watson statistic, which examines residuals to
determine whether or not they are randomly distributed (randomly distributed residuals
would suggest no autocorrelation). The Durban-Watson statistic for Figure 2.A.10 is 0.53,
which, not surprisingly, suggests autocorrelation is present (a standard table is used to
determine the significance level of the Durban-Watson statistic given the number of time
periods and variables being used).
Figure 2.B.1 Mean of Political Integration by Date, Differenced to Remove Autocorrelation.
Thousands of Years Ago
Political Integration (Differenced)
The question is, what is the effect of autocorrelation on our understanding or interpretation
of these data? How might autocorrelational effects be interpreted or corrected? One way of
dealing with autocorrelation is to filter out its effects though various algorithms that take
previous values for a given variable and remove or “difference” them from future values.
Figure 2.B.1 shows a plot of political integration with autocorrelation removed by
differencing each time period from the previous time period. The value of R-squared for
these data is 0.324, considerably lower than that for the non-differenced data. However, 1000
B.P. seems to be a rather marked outlier here, and I would argue this is because the general
problem associated with cases in this time period noted earlier (that many of the more
complex cases have entered the historic record and been dropped from the data set) has been
amplified through differencing. Dropping this outlier leads to an R-squared value of 0.762,
which I suggest is a better estimate of the actual value for this figure and perhaps for Figure
A second way of dealing with serial autocorrelation is to focus on individual time periods in
relation to their immediately preceding and following time periods, rather than on the overall
trend. This is the primary approach used in time-series analysis, a set of analytical methods
that is far too large and sophisticated to be dealt with in any detail here (see, e.g., Gottman
1981; McCleary and Hay 1980). Figure 2.B.2 shows a typical time-series graph for the
political integration variable. What it illustrates is essentially what I noted above: there is
significant autocorrelation between each time period and the preceding time period. What is
interesting is that the autocorrelation declines rather sharply, suggesting a fairly strong and
regular process is at work, a process which we might identify simply as cultural evolution.
Figure 2.B.2 Autocorrelation Graph for Political Integration
Political Integration
Lag Number
Confidence Limits
The fact that cultural evolution is a serial autocorrelation process makes me question
whether, indeed, it is a problem we need to address in studying cultural evolution rather than
an expectation that we only concern ourselves with in its absence or when it is violated.
Take, for example, an ordinary bank account. We put $100 in an each month interest
accrues. If we chart the amount of money in the account each month, we discover serial
autocorrelation--the amount in the account each month is correlated with the previous and
future months’ amounts with the specific value equivalent to the interest rate. The increase,
due to interest, is expected, and perhaps uninformative, but it is real and should not be
considered a problem. What might be more interesting, however, is the extent to which the
interest rate changes. The monthly balance will always show strong autocorrelation with the
previous month. But if we factor out that autocorrelation through differencing and look only
at the changes in the rates of increase, that might yield interesting insights into the process of
how interest works.
The exact same thing can be said for cultural evolution. It appears to be a regular process of
serial autocorrelation. If we factor out autocorrelation and look only at the changes that
remain, the exercise might yield interesting results. Figure 2.B.1 shows such results, and
illustrates that, while the evolution of political integration has been an overall linear process,
there have been marked peaks and valleys, as illustrated in Figure 2.B.3. There is a high
outlier at 8,000 years ago, and a low outlier at 6,000 years ago. This may suggest that
political integration took off rapidly around 8,000 years ago and then, over the course of
2,000 years, slowed markedly before resuming a generally linear upward trend. Is this
important? Does it tell us new things about cultural evolution? Perhaps, and I suggest that
because we may gain insights from such questions viewing autocorrelation as part of the
evolutionary process rather than as a statistical problem is by far the best approach.
Figure 2.B.3 Political Integration by Date, Showing Regression Line and Confidence
Thousands of Years Ago
Political Integration (Differenced)
C. New World-Old World Differences in Cultural Evolution
I offered political scientist Claudio Cioffi access to the OAT cases for his Long-Range
Analysis of War (LORANOW) project early in 1997, and he undertook several analyses
using the sample (Cioffi, personal communication). He suggested that cultural evolution
appeared to occur more rapidly in the New World than in the Old. I found this interesting,
since other cross-cultural studies had indicated that North America often produced divergent
results when compared to other world areas (e.g. Ember 1975). In examining the ACE data,
it appears that there are marked differences in cultural evolution between the Old World and
the New World.
Table 2.C.1 illustrates that New World cases appear to score significantly lower on a number
of variables, particularly Writing and Records, Technological Specialization, Land
Transport, Money, Density of Population, Social Stratification, and overall Cultural
Complexity. These differences are not surprising, as only a handful of New World traditions
developed writing or metalwork, and none developed currency or vehicular land transport.
More interesting is the fact that New World traditions have less dense and less stratified
populations. One might question whether these variables are causally linked in some way, a
question I will return to in the next section.
Table 2.C.1 Mann-Whitney U Statistics for New World and Old World Cases
Variable Location N Mean
Writing and Records New World
Old World
9402.5 0.0064
Fixity of Residence New World
Old World
9762.0 0.3898
Agriculture New World
Old World
9131.0 0.0632
Urbanization New World
Old World
10098.0 0.7493
Technological Specialization New World
Old World
8705.0 0.0157
Land Transport New World
Old World
7174.0 0.0000
Money New World
Old World
8444.0 0.0000
Density of Population New World
Old World
8279.0 0.0015
Political Integration New World
Old World
9776.0 0.4206
Social Stratification New World
Old World
8951.0 0.0385
Cultural Complexity New World
Old World
8821.0 0.0343
Figure 2.C.1 illustrates the fact that, while the evolution of overall cultural complexity
occurred later in the New World than the Old World, and never attained the same level, the
evolutionary process towards greater complexity apparently operated at roughly the same
speed in both areas. The evolution of more complex cultures began about 9000 years ago in
the Old World, and took roughly 5000 years to plateau. In the New World, the evolutionary
process began only about 5000 years ago, and was still increasing when the conquest and
subsequent collapse of New World cultures began. So, while beginning more recently than
the Old World, it appears that the evolution of cultural complexity moved at about the same
pace in both the New World and the Old.
Figure 2.C.1 Mean Cultural Complexity by Date for Old World and New World Cases
New World
Old World
Years Before Present
Cultural Complexity
D. Conclusion
The ACE data illustrate clear, unilineal trends in cultural evolution. These trends show
complex and sometimes divergent patterns that suggest the empirical study of cultural
evolution with data like those provided here is rich with potential. Not only are the patterns
themselves potentially of great interest, but methods to elucidate and, perhaps, explain them
also seem a rich area for further research.
The work of anti-evolutionists such as Lowie, Nisbet, and Giddens has been effective, and
largely halted work on identifying and explaining evolutionary trends during much of the
20th century. While scholars such as White (1959), Fried (1967), Service (1975), Harris
(1979), and Boyd and Richerson (1985) developed theoretical frameworks for understanding
unilineal evolution, there were few systematic attempts (beyond those of these scholars
themselves) to evaluate or refine these theories. The ACE provides a first step towards doing
so, as I hope to demonstrate here.
A. Causal Variables and Prime Movers
Three variables included within the ACE data set seem to be repeatedly identified as
underlying cultural change. These are population density, reliance on agriculture, and
technological specialization. Not surprisingly, all three are strongly inter-correlated, and all
three correlate strongly with both cultural complexity and time in years B.P., as shown in
Table 3.A.1. Each has been proposed as something as a “prime mover” underlying cultural
evolution. Population density, for example, has been proposed as the cause of agriculture
around the world (Cohen 1977), and agriculture as the cause of technological innovation
(Harris 1977). These correlations alone suggest such causal relationships may exist, but, as
has been said so often, correlation is not equivalent to causation.
Table 3.A.1 -- Spearman’s rho Correlation Coefficients for Selected Variables
Importance of
0.817 1.0
0.689 0.717 1.0
0.876 0.873 0.892 1.0
Date B.P.
-0.420 -0.484 -0.576 -0.572 1.0
Importance of
The task, then, is to examine how these variables inter-relate, and to determine whether
change in the value of one causes change in the values of the others. Since these data are
diachronic, they should allow us to see whether change in a presumed causal variable
actually proceeded its presumed effects. In other words, one can also examine them as a time
series to see whether changes in one or more of these variables precedes changes in the
others. This ability to examine causal relationships diachronically is one of the unique
strengths of archaeoethnology for identifying and exploring cultural evolution.
Figure 3.A.1 shows a time series plot of population density, agriculture, and technological
specialization. Unfortunately, there does not seem to be a single causal or “prime mover”
variable among the three--each increases at a fairly steady rate, and all tend to increase
together. Using differencing to “de-trend” the plot yields Figure 3.A.2. Again, there seems
no clear “prime mover” here; indeed, the three variables seem remarkably inter-correlated,
although agriculture does seem to lag behind population density and technological
specialization after about 6000 years ago.
Figure 3.A.1 Time Series Plot of Population Density, Agriculture, and Technological
Thousands of Years Ago
Population Density
Tech Specialization
Figure 3.A.2 Time Series Plot of Population Density, Agriculture, and Technological
Specialization, Differenced to De-trend the Series.
Thousands of Years Ago
Population Density
Tech Specialization
The extent of inter-correlation between population density, agriculture, and technological
specialization is shown more clearly in Figures 3.A.3, 3.A.4, and 3.A.5. These are cross-
correlation graphs for the three variables, illustrating the cross-correlation of each with the
others through time. There is a strong and regular relationship between all three. None of
these variables appears to directly cause change in the others, at least not in the time scale
(1000 years) used here. Within that time period, all appear to change together, and none
appears to be a clear “prime mover” of cultural evolution. The lesson here may be that there
are not single “prime mover” variables underlying cultural evolution.
Figure 3.A.3 Cross-Correlation of Population Density and Agriculture
Lag Number
Confidence Limits
Figure 3.A.4 Cross-Correlation of Population Density with Technological Specialization.
Lag Number
Confidence Limits
Figure 3.A.5 Cross-Correlation of Agriculture with Technological Specialization
Lag Number
Confidence Limits
On the other hand, while none of these variables seems a clear “prime mover” of change,
shortening the time period used in the analysis to 100 years or so may allow us to show one
of these variables to be causal in relation to the others. I have not done so here, because I
believe the ACE data are too coarse for such a detailed analysis. The ACE data are able to
illustrate broad patterns of cultural evolution, but may not be refined enough to allow for
close temporal relationship between variables to be resolved. I suggest this is precisely the
type of problem that the more detailed information available through the HRAF Collection
of Archaeology may be able to address.
B. Causal Modeling
The time-series do not appear to provide enough information to determine the causal
relationships between population density, agriculture, and technology. A different method of
examining causal relationships--causal modeling--may provide a means to determine
whether and how these variables effected and perhaps caused change in the others. Causal
modeling is a method used to establish quantitative measures of causal connection between
variables. It does not provide a means to prove that change in one variable causes change in
another, but rather, allows for various assumptions about the possible directions of causality
to be evaluated. In other words, it provides a way to test models of causal connection, but
does not independently identify causality (see Birnbaum 1981).
Figure 3.B.1 shows a simple model for the relationships between population density,
agriculture, and technology. The correlation coefficients are the same as those presented in
Table 3.A.1. Directionality is not illustrated here, because we have yet to identify causal
directions. One way to do so is to examine the partial correlation coefficients of between
these variables when controlling for time, and when controlling for the other variable (an
iterative method often referred to as the Simon-Blalock Technique--see Asher 1983). Partial
correlations are presented in Figures 3.B.2 and 3.B.3.
Figure 3.B.1 Correlations Between Population Density, Agriculture, and Technological
Figure 3.B.2 shows that controlling for date has little effect on the connections between these
variables; that is, variation seems uninfluenced by date. This is not surprising given the time-
series analyses I’ve already presented. The three variables appear to change in unison
regardless of the time period. Figure 3.B.3 is more interesting, as the correlation between
population density and technological specialization drops precipitously when controlling for
agriculture, much more than the other correlations drop when controlling for the third
Figure 3.B.2 Partial Correlations Between Population Density, Agriculture, and
Technological Specialization, Controlling for Date
Figure 3.B.3 Partial Correlations Between Population Density, Agriculture, and
Technological Specialization, Controlling for the Other Variable
Using basic rules of thumb for causal modeling (e.g. Davis 1985), we arrive at the
parsimonious model presented in Figure 3.B.4. First, because the correlation between
population density and technological specialization dropped so precipitously when
controlling for agriculture, we can make the assumption that changes in agriculture may be
causally related to changes in both population density and technological specialization, and
that the correlation between them (without controlling for agriculture) is largely spurious
(but see the discussion below). Second, because we know that agriculture is required to
sustain high population densities, and that ceramics are rare and metal work unknown in
non-agricultural societies, we can assume that agriculture must causally precede at least
some changes in the other two variables. Thus, we end up with a causal model in which
changes in agriculture cause change in population density and change in technological
Figure 3.B.4 A Parsimonious Model of the Relationship Between Population Density,
Agriculture, and Technological Specialization
If change in agriculture causes change in the other two variables, then why is there still a
statistically significant correlation between population density and technological
specialization even when controlling for agriculture? The answer is probably that there is a
strong, underlying variable that is uncontrolled for here. It is not simply date, as controlling
for date does not change the correlation coefficients very much. Rather, it is probably
something we might refer to as cultural evolution--a regular process of change effecting all
the variables and which we have already seen in, for example, Figure 2.B.3. The strength of
that underlying variable--cultural evolution--is significant (R-squared of 0.324 for Figure
2.B.3) and affects all the variables. Hence we must assume that, without controlling for
cultural evolution, there will always remain a strong correlation between two culture
variables on the ACE data set. What we need to look for is not absence of correlation, but
rather significant declines that suggest all other factors aside from cultural evolution have
been controlled.
It appears that agriculture is the more causally important variable of the three we have
examined here, suggesting that both population pressure (e.g. Cohen 1977) and technological
determinism (for example, elements of both Harris 1979 and White 1959) models of cultural
evolution are less satisfactory than models which propose changes in subsistence affecting
changes in other areas of culture (e.g. Steward 1955). It also appears that there is a powerful,
underlying variable not accounted for in any of these models, a variable we might refer to
simply as cultural evolution.
C. Log-Linear Modeling
Log-linear modeling provides an alternative method of examining causal relationships that is
more appropriate for the ordinal data used here. Log-linear modeling is essentially a form of
causal modeling like that used above, but explicitly designed for use with categorical data.
Log-linear modeling allows variables with multiple categories to be used to calculate the
odds (i.e., the ratio of favorable to unfavorable responses) that a change in one variable will
cause a change in the other (Knoke and Burke 1980).
A log-linear model is basically a statement of the expected frequencies in the cells of a
crosstabulation. To assess how well a given model fits the data one determines how well the
cell frequencies expected in the model approximate the observed frequencies, with
goodness-of-fit calculated as odds and odds ratios (often called likelihood ratios). Table
3.C.1 presents a group of log-linear models for population density, technological
specialization and agriculture, along with their associated likelihood ratios (L
). By
convention, interaction between two variables is noted with a * in describing log-linear
models, and non-interaction with a +. Thus, the relationship between population density,
technological specialization, and agriculture presented in Figure 3.B.1 is represented in
Table 3.C.1 by model 2, while the parsimonious causal model presented in Figure 3.B.4 is
represented by model 3.
Table 3.C.1 Log-Linear Models for Population Density, Technological Specialization, and
Model L
df p
1. {D*T*A} 0 0 -
2. {D*A}+{T*A}+{D*T} 11.07 8 0.198
3. {D*A}+{T*A} 34.66 12 0.000
4. {D*A}+{D*T} 41.50 12 0.000
5. {T*A}+{T*P} 104.07 12 0.000
6. {D*A}+T 203.60 16 0.000
7. {T*A}+D 266.17 16 0.000
8. {D*T}+A 273.02 16 0.000
9. {D}+{T}+{A} 435.11 20 0.000
Unlike ordinary evaluation of contingency tables with statistics like chi-squared, where one
usually seeks to find deviations from expected patterns, in log-linear analysis one seeks the
best match with expected patterns. Hence, in looking at a table like 3.C.1, one seeks low
values of L
relative to the degrees of freedom rather than vice-versa. Model 2 has the lowest
value of L
(except for the “saturated” model 1, which is only used as a baseline in
evaluating other models) and highest degrees of freedom. Indeed, it is the only model that
shows non-significant deviance from expected values. However, a critical aspect of log-
linear analysis is the evaluation of alternative models. Although model 2 appears to best fit
the data, it is not the most parsimonious, nor does it match with our theoretical expectations
derived from the causal modeling performed in the section 3.B. To better evaluate the
models, one must examine the changes from model to model in L
and degrees of freedom in
order to determine whether those changes are statistically significant.
Table 3.C.2 shows the change in likelihood ratios (L
) and degrees of freedom for four
model comparisons. The changes are the simple difference in the values of L
and degrees of
freedom for each model, while p can be calculated from a standard chi-squared table using
the values of L
and df. Looking at these it would appear that model 2 may be the most
parsimonious. Model 2 shows a significant change in L
relative to the change in the degrees
of freedom when compared with model 3, and it provides an acceptable fit with the expected
values. On the other hand, the change in L
from model 2 to the “saturated” model 1 is not
statistically significant. Hence, model 2 appears to be the most parsimonious.
Table 3.C.2 Evaluation of the Improvement of Fit for Log-Linear Models for Population
Density, Technological Specialization, and Agriculture.
Model Comparison L
df p
2-1 11.07 8 0.198
3-2 23.59 4 0.000
6-5 99.53 4 0.000
9-8 162.09 4 0.000
Causal modeling suggested that the relationship between population density and
technological specialization was not important and could be dropped from the model. Log-
linear modeling suggests that opposite, that including the interaction between population
density and technological specialization significantly increases the fit of the model despite
the loss of degrees of freedom.
Why is there a difference between the results of these exercises in modeling? The simple
answer may be that the techniques used are different and therefore yield somewhat different
results. A more satisfactory answer probably rests in the data themselves. As I mentioned
earlier, the ACE data are coarse. They were developed to illustrate broad patterns of cultural
evolution, and attempting to use them to differentiate the comparative effects of one variable
on another might be overextending their capabilities. Finally, it may be that the results of the
time-series analyses gave the clearest picture--that these variables mutually affect one
another over short periods of time and it may be impossible to identify a clear causal or
“prime mover” variable among them.
D. Conclusion
Cultural evolution appears to be multi-causal, and as we move towards explaining cultural
evolution, we must avoid the desire to overly simplify what appears to be a complex,
multivariate set of relationships. The Atlas of Cultural Evolution provides a set of data to
begin exploring this complex set of relationships. I hope that the discussion in section 3 has
also provided some background to the various analytical techniques that might be used in
conjunction with the ACE data set to examine cultural evolution. I have used these data to
begin the exploration of cultural evolution through the empirical methods of
archaeoethnology (e.g. Peregrine 2001b), and it is my hope that by providing these data to a
larger audience of researchers, our combined efforts may lead to significant insights into the
patterns and processes of cultural evolution.
Asher, Herbert
1983 Causal Modeling (Second Edition). Thousand Oaks, CA: Sage.
Birnbaum, Ian
1981 Introduction to Causal Analysis in Sociology. London: MacMillian.
Blanton, Richard, Stephen Kowalewski, Gary Feinman, and Jill Appell
1981 Ancient Mesoamerica. Cambridge: Cambridge University Press.
Boyd, Robert and Peter J. Richerson
1985 Culture and the Evolutionary Process. Chicago: University of Chicago Press.
Chick, Gary
1997 Cultural Complexity: Its Concept and Measurement. Cross-Cultural
Research 31(4):275-307.
Cohen, Mark N.
1977 The Food Crisis in Prehistory: Overpopulation and the Origins of
Agriculture. New Haven: Yale University Press.
Davis, James A.
1985 The Logic of Causal Order. Thousand Oaks, CA: Sage.
Ember, Carol R.
1975 Residential Variation among Hunter Gatherers. Behavior Science Research
Ember, Melvin and Carol R. Ember
1995 Worldwide Cross-Cultural Studies and their Relevance for Archaeology. Journal
of Archaeological Research 3:87-111.
Fried, Morton
1967 The Evolution of Political Society. New York: Random House.
Giddens, Anthony
1984 The Constitution of Society. Berkeley: University of California Press.
Goldenweiser, Alexander
1937 Anthropology. New York: F.S. Crofts.
Gottman, John M.
1981 Time-Series Analysis: A Comprehensive Overview for the Social Scientist.
Cambridge: Cambridge University Press.
Harris, Marvin
1977 Cannibals and Kings: The Origins of Culture. New York: Vintage.
1979 Cultural Materialism. New York: Vintage.
Knoke, David and Peter Burke
1980 Log-Linear Models. Thousand Oaks, CA: Sage.
Lowie, Robert
1946 Evolution in Cultural Anthropology: A Reply to Leslie White. American
Anthropologist 48:223-233.
McCleary, Richard and Richard Hay
1980 Applied Time Series Analysis for the Social Sciences. Beverly Hills, CA:
McNett, Charles W.
1970 A Settlement Pattern Scale of Cultural Complexity, in A Handbook of
in Cultural Anthropology (R. Naroll and R. Cohen, eds.). Garden City, NY:
Natural History Press. Pp. 872-886.
Murdock, George Peter and Douglas R. White
1969 The Standard Cross-Cultural Sample. Ethnology 8:329-369.
Murdock, George Peter and Caterina Provost
1973 Measurement of Cultural Complexity. Ethnology: 12:379-392.
Nisbet, Robert A.
1969 Social Change and History: Aspects of the Western Theory of Development.
New York: Oxford.
Ostrom, Charles W.
1990 Time Series Analysis: Regression Techniques (Second Edition). Beverly
Hills, CA: Sage.
Otterbein, Keith F.
1976 Sampling and Samples in Cross-Cultural Studies. Behavior Science Research
Peregrine, Peter N.
2001a Outline of Archaeological Traditions. New Haven: HRAF Press.
2001b Cross-Cultural Comparative Approaches in Archaeology. Annual Review of
Anthropology 30:1-18.
Peregrine, Peter N. and Melvin Ember, eds.
2001- Encyclopedia of Prehistory (9 volumes). New York: Kluwer Academic / Plenum
2002 Publishers.
Service, Elman
1975 Origins of the State and Civilization: The Process of Cultural Evolution.
York: Norton.
Shanks, Michael and Christopher Tilley
1992 Re-Constructing Archaeology: Theory and Practice, 2nd edition. London:
Steward, Julian H.
1955 Theory of Culture Change: The Methodology of Multilinear Evolution.
Urbana: University of Illinois Press.
White, Leslie A.
1959 The Evolution of Culture. New York: McGraw-Hill.
Willey, Gordon R. and Philip Phillips
1958 Method and Theory in American Archaeology. Chicago: University of Chicago
Archaeological traditions have both spatial and temporal dimensions. Each tradition’s
temporal dimension is provided in the ACE data set by a start date, end date, midpoint, and a
set of time series variables (see Section 6). The spatial dimension for each tradition is also
provided, in a crude form, in the ACE data set by an east and north grid coordinate, which
together identify a 1000 km square grid unit within which the midpoint of the geographical
area of the tradition is located. The maps that follow provide a more accurate representation
of the spatial dimension of each archaeological tradition. The maps are included in electronic
form on the accompanying CD, along with the MapMaker Gratis software package with
which the maps can be displayed, printed, and modified.
A. Maps of the World’s Archaeological Traditions
Figure 5.A.1 -- The World’s Archaeological Traditions 2 Million Years Ago
Figure 5.A.2 -- The World’s Archaeological Traditions 500,000 Years Ago
Figure 5.A.3 -- The World’s Archaeological Traditions 100,000 Years Ago
Figure 5.A.4 -- The World’s Archaeological Traditions 50,000 Years Ago
Figure 5.A.5 -- The World’s Archaeological Traditions 40,000 Years Ago
Figure 5.A.6 -- The World’s Archaeological Traditions 30,000 Years Ago
Figure 5.A.7 -- The World’s Archaeological Traditions 20,000 Years Ago
Figure 5.A.8 -- The World’s Archaeological Traditions 12,000 Years Ago
Figure 5.A.9 -- The World’s Archaeological Traditions 11,000 Years Ago
Figure 5.A.10 -- The World’s Archaeological Traditions 10,000 Years Ago
Figure 5.A.11 -- The World’s Archaeological Traditions 9,000 Years Ago
Figure 5.A.12 -- The World’s Archaeological Traditions 8,000 Years Ago
Figure 5.A.13 -- The World’s Archaeological Traditions 7,000 Years Ago
Figure 5.A.14 -- The World’s Archaeological Traditions 6,000 Years Ago
Figure 5.A.15 -- The World’s Archaeological Traditions 5,000 Years Ago
Figure 5.A.16 -- The World’s Archaeological Traditions 4,000 Years Ago. (Note that Southern
Mesopotamia has moved into the historic record and is dropped from the data set.)
Figure 5.A.17 -- The World’s Archaeological Traditions 3,000 Years Ago (Note that Zhou China
and the Greco-Roman world have moved into the historic record and are dropped from the data set.)
Figure 5.A.18 -- The World’s Archaeological Traditions 2,000 Years Ago (Note that much of
Eurasia is historic and not included in the data set.)
Figure 5.A.19 -- The World’s Archaeological Traditions 1,000 Years Ago (Note that much of
Eurasia is historic and not included in the data set.)
B. Interactive Maps Using MapMaker Gratis
Included on the accompanying CD is a copy of the MapMaker Gratis software package, a
simple Geographical Information System (GIS) designed to allow novice users to create and
manipulate maps in sophisticated ways. The ACE maps presented above were produced
using MapMaker Pro, a more sophisticated version of the basic MapMaker package included
here, but a user can display, edit, and print any of these maps with either version. In addition,
MapMaker allows data from the ACE data set to be displayed geographically. One might
produce, for example, a map showing population density values for a given date in the past,
as illustrated in Figure 5.B.1 (next page).
MapMaker Gratis is freeware, but its creators invite users to visit the MapMaker website
( to consider purchasing a full version of the software and to examine
the other mapping resources they have. The MapMaker website includes tutorials and other
user information for the MapMaker software family, as well as links to other mapping sites
and data repositories.
Figure 5.B.1 Population Density Values for the World 6,000 Years Ago.
To use MapMaker Gratis, you need first to install it on your computer. Copy the file
mmZip.exe to your hard disk, then double-click on it and follow the instructions. Files will
automatically be unzipped and installed in c:\MapMaker, although you can specify another
location if you wish. A manual is available and can be downloaded (along with program
updates) from the MapMaker website (
Once MapMaker Gratis is installed, you can use it to display, edit, and print any of the ACE
maps, or indeed, to create your own maps. To open a map, launch MapMaker Gratis and
select “File” and “Open.” You will see a screen like Figure 5.B.2 (next page).
Maps like those presented above are called drawings in MapMaker, and have the extension
.dra. Navigate using the “Currently selected directory” window to the CD which
accompanied this journal and to the \AceMaps directory. MapMaker drawing files for the
ACE maps are located in this directory. Select a drawing file and double-click on it to open
it. A second screen, like that shown in Figure 5.B.3, will appear.
Figure 5.B.2. MapMaker Screen Showing Maps (Drawing Files) Available to Open.
Figure 5.B.3. MapMaker Screen Showing Display and Data Link Options.
MapMaker provides many different ways to display maps and data by linking drawing files
to data files and style files. This screen provides a number of options, only two of which I
will note here. First, to link the ACE maps to the ACE data set, click the button next to
“External data values” and open the file Acemap.dbf in the \AceMaps directory. A screen
like that shown in Figure 5.B.4 will appear. To map a variable, select it in the “Style link
column.” To label the traditions, choose either the name or number variables in the “Object
label link column.”
Second, MapMaker uses style files, with the extension .stl, to create the appearance of a
given map. To make a map look like those presented above, click the button next to “Style
file” and select the Ace.stl file. Finally, you will probably want to select the check box next
to “Layer hit-able with Data Query tool.” Doing so will allow you to click on the geographic
area of an archaeological tradition and view the data associated with it. When you click the
“OK” button your map will be displayed. You will not be able to modify or alter the map. In
order to alter a map you need to load it as a “Live Layer,” but I suggest you become familiar
with MapMaker and read the manual thoroughly before beginning to work with live layers.
Figure 5.B.4. MapMaker Data Link Screen.
Part B: Live link to
Back to top
Back to Archaeo home
World Cultures home
... One proxy for this hypothesis is social stratification (Class). Here, we use the data on emergent stratification among archaeologically known societies collected by Peregrine (39). Another line of scholarship focuses on the length of chains of command, arguing that the more levels of control and command in a hierarchy, the more power accrues to the individuals occupying the top levels, who will favor centralization and state-level institutions that would protect their advantageous position. ...
... It is clear that social complexity has many dimensions or manifestations (65). While several researchers proposed synthetic, integrative measures that capture multiple dimensions of social complexity (2,39,66), a more common approach has been to use a single proxy measure, such as the population size of the largest settlement (2), the number of decision-making levels (67), the number of levels of settlement hierarchy (68), or the extent of controlled territory (63). Others have criticized these approaches on the grounds that these proposed measures focus too much on size and hierarchy (69). ...
... We proxy this hypothesis with social stratification (Class), which was defined by Murdock and Provost (66) and coded by Peregrine (39) for archaeological societies. This variable takes three values: egalitarian (no classes), two classes, and three or more classes. ...
Full-text available
During the Holocene, the scale and complexity of human societies increased markedly. Generations of scholars have proposed different theories explaining this expansion, which range from broadly functionalist explanations, focusing on the provision of public goods, to conflict theories, emphasizing the role of class struggle or warfare. To quantitatively test these theories, we develop a general dynamical model based on the theoretical framework of cultural macroevolution. Using this model and Seshat: Global History Databank, we test 17 potential predictor variables proxying mechanisms suggested by major theories of sociopolitical complexity (and >100,000 combinations of these predictors). The best-supported model indicates a strong causal role played by a combination of increasing agricultural productivity and invention/adoption of military technologies (most notably, iron weapons and cavalry in the first millennium BCE).
... That is, the farming of certain crops may have positively affected past economic outcomese.g., historical adoption of technology, population density, and urbanizationthat might influence present-day attitudes towards technology. To mitigate this, Column (3) controls for technology adoption rates in 1500AD and Column (4) controls for population density in 1500AD, from Acemoglu et al. (2002), and historical urbanization rates in 1000BCE, from Peregrine (2003). ...
... Acemoglu et al. (2001) Urbanization index 1000BCE This is the historical urbanization rate in 1000BCE constructed at the country-level. Peregrine (2003) Historical pathogen prevalence index ...
Recent studies have shown that agricultural legacies can have a lasting effect on cultural formation. However, to date, the literature has not examined how the agricultural origins of culture affect individual preferences for modern technology. This paper addresses this gap by investigating how the agricultural origins of individualist and collectivist cultures have affected individual attitudes towards contemporary science and technology, at the subnational level. Its results suggest that societies that have historically cultivated low labor-intensive crops, which demand individualistic behavior, have developed favorable attitudes towards technology. Conversely, societies that cultivated labor-intensive crops, which required intense cooperation and cohesiveness among farming communities, developed collectivist norms, which in turn led to their exhibiting a lower affinity to, and preference for, technology. This study’s findings advance our understanding of how the diversity of agricultural legacies can explain subnational differences in individual’s attitudes towards modern scientific progress.
... How do the above-mentioned hypotheses correspond to the statistical correlations between the political complexity and urbanization? For these purposes, we will use The Atlas of Cultural Evolution by Peter Peregrine for 289 primitive societies and archaic states (ACE; Peregrine 2003). In this Atlas, the data from the nine-volume Encyclopedia of Prehistory was summarized Ember 2001, 2002), which involved more than 200 researchers from 20 countries and covers all of the most important and known archeological cultures of the prehistoric epoch. ...
... The chronological scope of the work is 500,000-500 years BCE. Such sample should be fully representative for pursuing both synchronous and diachronous cross-cultural research on prehistoric societies and politogenesis (Peregrine 2003(Peregrine , 2004. ...
Origin of the state is a persistent issue in social sciences and humanities. At present, there are several popular theories on the subject. This chapter discusses some crucial themes—factors of political centralization, stages of statehood, and the role of urbanization in state formation. The chapter deals with the specificity of regional aspects of the rise of complexity and urbanization on the example of two polities of the Russian Far East and Manchuria—the Bohai Kingdom and the Jurchen Empire.
... A key point is that major structural changes often take place without transformational changes in sociopolitical complexity. Nikolay N. Kradin draws on a compilation of cases from the Atlas of Cultural Evolution (Peregrine 2003) to illustrate both the strong relationship between largest community size and sociopolitical complexity, but also the lack of strict thresholds despite the strong general correspondence (see also Feinman 2013). Through an accompanying historical example from the Russian Far East, Kradin advances the thought that, in part, the lack of fit between large settlement sizes (urbanization) and sociopolitical complexity reflects diversity in the ways that different institutions and modes of leadership utilize the built environment. ...
This book explores the deep roots of modern democracy, focusing on geography and long-term patterns of global diffusion. Its geographic argument centers on access to the sea, afforded by natural harbors which enhance the mobility of people, goods, capital, and ideas. The extraordinary connectivity of harbor regions thereby affected economic development, the structure of the military, statebuilding, and openness to the world – and, through these pathways, the development of representative democracy. The authors' second argument focuses on the global diffusion of representative democracy. Beginning around 1500, Europeans started to populate distant places abroad. Where Europeans were numerous they established some form of representative democracy, often with restrictions limiting suffrage to those of European heritage. Where they were in the minority, Europeans were more reticent about popular rule and often actively resisted democratization. Where Europeans were entirely absent, the concept of representative democracy was unfamiliar and its practice undeveloped.
The territory is the most obvious physical embodiment of social relations. Territorial interactions determine neighborhood relations. And these latter are the primary basis of material production, of the economy outside of and prior to the market. Due to the universal, biologically determined, nature of territorial behavior, we must also consider human territorial relations as having grassroots, extra-social basis. Therefore, we almost involuntarily perceive the territorial structure of human societies as an invariant with roots too deep for social analysis. Or we consider it an archaic social institution. As such it has two components: invariable—archaic, and variable—determined by social and political factors. Herein I focus on analyzing the second, variable, part; however, I also constantly keep in mind the invariable component of the territorial structure. These two components create numerous territorial structures, which are based on only several permanent elements. The variance of territorial forms necessitates their typology. The typology of territories is based on the type of local communities distinguished by the degree of spatial isolation and the manner of their emergence and development. I am considering six territorial structures of provincial societies. Ultimately, they are reduced to four types of territories, which differ in all basic characteristics. These four territorial types are: (1) territories of communities existing in spatial isolation and not affected by government impact; (2) territories of communities, which are not isolated, but are developing without significant coercive government impact; (3) territories of communities formed coercively but with an inadequately developed transport infrastructure; (4) territories of communities formed coercively with a well-developed infrastructure due to their location on transport thoroughfares.
It has often been observed that the emergence of states in a region is typically preceded by an earlier transition to agricultural production. Using new data on the date of first state emergence within contemporary countries, we present a global scale analysis of the chronological relationship between the transition to agriculture and the subsequent emergence of states. We find statistically significant relationships between early reliance on agriculture and state age in all sub-samples and also when we use alternative sources of data at different levels of geographical aggregation. A one millennium earlier transition to agriculture among non-pristine states predicts a 317-430 year earlier state emergence. We uncover differences in cases where states were imposed from outside or when they emerged through internal origination. The agriculture-state lag is on average 3.1 millennia in internally originated (including pristine) states, and 2.7 millennia in externally originated states. We also explore some of the mechanisms through which agriculture is believed to have influenced the emergence of states. Our results indicate that the rise of social classes was often an intermediate step towards the presence of early states.
Full-text available
이 연구의 목적은 도킨스가 『만들어진 신』에서 주장한 종교 비판에 과학적 근거가 없다는 사실을 밝히는 것이다. 그는 종교가 자연적인 과정에 의해 진화 했다는 종교진화론을 바탕으로 종교를 비판했다. 본 논문은 그의 비판을 재비 판하기 위해 두 가지 분석을 수행하였다. 첫째로 도킨스의 종교진화론이 미메 틱스(memetics)라는 진화 이론에 의존함을 보였다. 둘째로 미메틱스가 과학 의 수준에 올라서지 못했으며 그의 종교진화론이 성립하기 위해 필요한 과학 적 증거를 제공하지 못함을 보였다. 두 결과를 종합해서 본 연구는 그의 종교 비판에 과학적인 근거가 없다는 결론을 내린다. 도킨스는 과학에 바탕해서 종 교를 비판한다는 견해를 수차례 피력한 바 있다. 그러나 본 연구 결과는 정작 과학적 근거가 없는 것은 그의 종교진화론이라는 것을 보여준다. 도킨스의 주 장은 과학적으로 부실하며 신학적으로 편협한 것이다. 결국 그는 근거없이 과 학과 종교를 대립으로 몰고 간 것이다. 장래 과학과 종교의 대화를 위해 그와 같은 편향된 시각은 지양되어야 한다. The purpose of this study is to reveal that there is no scientific evidence for Dawkins' critique of religion. In this paper, two analyzes were performed. First, it was shown that Dawkins's criticism depends on memetics. Second, it was shown that memetics is not a scientific theory and cannot provide a scientific evidence for Dawkin's theory. Based on both results, this study concludes that there is no scientific basis for his critique of religion. Dawkins has repeatedly expressed the opinion that religion is criticized with science. However, the results of this study show that it is his own opinion that has no scientific basis. Dawkins' argument is scientifically poor and theologically narrow. In the end, he drove science and religion into conflict without rational grounds. Such a biased perspective should be avoided for the future dialogue of science and religion.
Full-text available
Cross-cultural comparative approaches have been used widely in archaeological research, yet to date none seem to have achieved their full potential. Synchronic cross-cultural comparisons have provided a number of material correlates of behavior, as well as a few causal and noncausal associations that allow behavior to be inferred from material remains. However, large areas of material culture, such as ceramics and lithics, have not yet been subject to extensive comparative analysis, and thus large areas of archaeological research that might be aided by synchronic comparative findings have been left unassisted. Diachronic cross-cultural comparisons have been used extensively to chart and analyze cultural evolution. However, these comparisons are typically based on grab-bag samples and only rarely employ statistics to aid in the discovery or testing of evolutionary patterns. New research tools providing a statistically valid sampling universe and information resources for coding archaeological data are being developed to facilitate cross-cultural comparisons.
The various samples and sampling universes available to the researcher about to undertake a cross-cultural study are systematically reviewed. Discussed are the Human Relations Area Files, the "World Ethnographic Sample," the Ethnographic Atlas (both the complete atlas as well as the summary version), the Standard Cross-Cultural Sample, the Quality Control Sample (or HRAF Probability Sample), and the "Standard Eth nographic Sample." Following an assessment of each sample, recom mendations are made as to the conditions and circumstances under which the samples should be utilized. The discussion concludes with recommen dations for drawing a sample from one of the samples or sampling universes.
How do biological, psychological, sociological, and cultural factors combine to change societies over the long run? Boyd and Richerson explore how genetic and cultural factors interact, under the influence of evolutionary forces, to produce the diversity we see in human cultures. Using methods developed by population biologists, they propose a theory of cultural evolution that is an original and fair-minded alternative to the sociobiology debate.