132 Cultural Transmission and Archaeology: Issues and Case Studies
COCHRANE, E. E. 2008. Migration and Cultural
Transmission: investigating human movement as an
explanation for Fijian ceramic change, in O'BRIEN, M. J.
(Ed.) Cultural Transmission in Archaeology: Issues and
Case Studies. Washington, D.C., Society for American
Archaeology.
11
Migration and Cultural Transmission:
1nvestiga6ng Human Movement as an Explanation
for Fijian Ceramic Change
Ethan E. Cochrane
P eople may move from home to home during
their lifetimes and in doing so go about their
daily activities in a series of different ecological
and cultural settings. Sometimes people move in
pairs, parents and their children move, or larger
groups relocate. Nonhuman organisms similarly
move about during their lifetimes, including birds
and their seasonal movements, coyotes inhabiting
new dens, and butterflies occupying new tenito-
ries that offer reproductive advantages (Baker
1978). Such relocations are generally referred to
as "migrations," although a range of behavioral
variation is encompassed by this term. Migration
has long been a subject of archaeological inquiry
and disciplinary debate because it is both intuitive
and noncontroversial that migration helps shape
variation in the archaeological record. The ques-
tion is, how do we identify the archaeological sig-
nature(~)of migration? Further, how do we
explain the occurrence of migrations in prehistory
(Anthony 1990; Burmeister 2000; Clark 1994;
Rouse 1986; Snow 1995)?
Migration can be integrated into a cultural
transmission framework, where migration is
defined as the movement of individuals between
local populations that sorts homologous similari-
ties. Local population here is synonymous with
deme in a biological framework, and sorting
denotes the same process referenced by the term
in evolutionary theory, simply the differential dis-
tribution of similarities. These similarities must
be homologous if our explanations are to address
the movement of individuals related through cul-
tural transmission. This definition of migration is
different from the colonization of uninhabited
landscapes and does not specify the intensity or
direction of movement. My goal in this chapter is
to address questions of migration in the simplest
of terms, without the added complexities of multi-
directional movement and movement that occurs
differentially across segments of a population. If
my simple approach works, then we can construct
more precise explanations that incorporate these
aspects of migration more routinely recognized
by anthropological geneticists (Fix 1999).
In an evolutionary context, migration is similar
to gene flow as used by population geneticists,
although they usually refer only to the one-way
movement of individuals (Fix 1999). This defini-
tion of migration follows from the principles of
evolutionary theory used to explain cultural (in
the sense of extended phenotype) change: There
is cultural variation, the variation is transmitted
between individuals, and some variants persist at
the expense of others. From these principles we
can postulate, for example, that some distribu-
tions of transmitted variants are explained by
selection (not necessarily differential human
reproduction [Leonard and Jones 19871) and other
distributions may be explained by drift. We can
also recognize that some cultural similarities will
be homologous, at least in part a result of trans-
mission within a population, and some similarities
will be analogous, possibly a product of conver-
gence in separate populations.
Not surprisingly, the methodological debate
over how to identify migration using artifacts is
approached differently from a transmission per-
spective. Four steps are involved. First, assem-
blages from at least two geographic areas must be
described by artifact classes that measure homolo-
gous similarity. Second, methods such as cladistics
or seriation are used to depict transmission rela-
tionships between classes and assemblages. Third,
the geographic distribution of classes or assem-
blages are is compared with their depicted trans-
mmion relationships, such as in a phylogenetic
tree, to develop hypotheses of the relative likeli-
hood of cultural transmission and possible migra-
tion between local populations. Finally, separate

archaeological data sets from the same geographic
areas should be analyzed to construct plausible
accounts relating the likelihood of cultural trans-
mission between local populations to (1) linked
cultural transmission and the movement of indi-
viduals and (2) cultural transmission between local
populations without such movement.
The theoretical issue of how to explain migra-
tion is also approached differently within an evo-
lutionary framework. Ecological studies of animal
migration suggest that natural selection as an
explanation for migration is a good preliminary
hypothesis, as movement and likely habitat
change cost energy (Baker 1978). Evolutionary
studies of nonhuman seasonal migration suggest
that it can often be explained by competition
between conspecifics or species and by inbreed-
ing avoidance (Roff 2002). For example, across
populations in which individuals compete for
resources, fitness advantages may accrue to indi-
viduals through ecological isolation (e.g., niche
specialization). However, if greater relative-fit-
ness advantages accrue to individuals that move
and obtain resources elsewhere, then such move-
ment at a population level can be explained as a
result of selection. Cox (1968) uses this frame-
work to explain migration in birds and argues that
if competition between lineages is important in
the explanation of migration, then migrant line-
ages should show less variation in phenotype
r related to ecology, as the ecological isolating
mechanism of migrant lineages is migration and
not niche specialization. Cox found that migration
tends to occur in groups less able to achieve com-
s plete ecological isolation through other means.
Evolutionary theory presents us with a
process-selection-with which to explain migra-
tion as well as a means by which to test the expla-
nation through comparison of cultural variation
related to ecology. This approach differs from
most archaeological approaches to migration,
which attempt to explain movement by the inten-
tions of past people, largely economic but often
viewed as including a variety of additional cul-
tural motives, so that "each migration has its own
causes" (Burmeister 2000550). This is a decid-
edly different explanatory tack than that provided
by evolutionary theory and a cultural transmission
framework, in which cause is attributed to a set of
processes external to the artifacts studied
(Dunnell 1982; Sellars 1962; Willer and Willer
1974)-primarily selection, transmission, and
Migration and Cultural Transmission / Cochrane 133
innovation (Cochrane 2001). Using the method
developed by Cox (1968) to explain past human
migration requires the definition of migrant cul-
tural lineages and then comparison in terms of
relative ecological specialization.
This chapter explores how migration might be
incorporated into a transmission framework for
explaining cultural variation. The first section
briefly describes the history of migration research
in archaeology with emphasis on developments
relevant to cultural transmission research. Then I
present a case study from the southwest Pacific
Islands, a region where migration-based explana-
tions of cultural change have figured prominently.
The case study is a cladistic analysis of pottery
from Fiji and Vanuatu used to define transmission
patterns indicating migration between the two
archipelagos. The results suggest that homologous
similarity can be measured across the archipelagos
and that migration might explain this similarity.
METHOD AND THEORY OF MIGRATION
EXPLANATIONS IN ARCHAEOLOGY
Although migration as an explanation for archae-
ological variation was invoked by European
scholars as early as the seventeenth cent~~ry, and
by Americanists in the nineteenth century
(Trigger 1989), archaeological methods for identi-
fying migrations did not appear until the early to
mid-twentieth century. In the Americanist tradi-
tion, Spier (1917) was one of the first archaeolo-
gists to suggest that distributions of different arti-
fact classes are a result of the movement of
people. He mapped the temporal distribution of
sites belonging to different time periods as indi-
cated by the presence of chronologically sensitive
ceramic types. The changing locations of sites
placed in temporal sequence demonstrate "that
the center of population has shifted from period
to period" (Spier 1917:300).
Less than a decade later in Europe, Childe was
also tracking human migrations in prehistory.
Building on the work of Kossina (1 91 I), Childe
(1925, 1929) suggested that cultural groups, or
"peoples," can be identified by recurring associa-
tions of artifact types. Artifact types useful for
identifying cultural groups include pots, orna-
ments, burials, and house forms. When a "total
and bodily transference" of these artifact types
occurs from one area to another, a migration had
likely occurred (Childe 1929:~-vi).When only
one or a few artifact types that define a cultural
134 Cultural Transmission and Archaeology: Issues and Case Studies
group appear in a spatially adjacent cultural
group, Childe inferred "some sort of 'relation'
between the respective areas or cultures"
(1929:vi). By relation Childe meant a form of
contact between groups. Some artifact similasities
across cultural groups were explained as the
transmission of adaptive similarities (although not
in those terms), as suggested by Childe's aside
about the unremarkable "spread of an obviously
superior device (e.g., the cut-and-thrust sword)"
(1929:vii). For Childe, then, the identification of
migration in the archaeological record involved
two observations. First, cultural groups or peoples
are identified by a suite of certain kinds of artifact
types that are less likely to be transmitted
between neighboring cultural groups-a perspec-
tive that foreshadowed concepts such as cultural
conservatism and cultural cores (Boyd et al. 1997;
Rosenberg 1994). Second, when these types-
house forms, burial styles-are applied to the
archaeological record and the suite of types iden-
tify chronologically rapid spatial expansions, a
migration has likely taken place. Childe viewed
essentially all cultural similarities as homologous,
suggesting that independent invention is
extremely rase in the cultural realm.
Childe's work provided the model for
European archaeology for the next 30 years
(Trigger 1989), and there was little additional
methodological discussion of artifact classifica-
tion and the identification of migrations. In the
Americas, however, a theoretical and method-
ological literature developed around issues of arti-
fact classification and the construction of artifact
types that measure temporal change (Dunnell
1986; Lyman et al. 1997). The profusion of
chronological types resulting from this work led
to several formal systems for constructing and
arranging types (e.g., Colton and Hasgrave 1937;
Krieger 1944; McKern 1939; Rouse 1939;
Spaulding 1953; Wheat et al. 1958). Ceramic
wares, series, and other aggregate units such as
phases were interpreted to be the product of cul-
turally related peoples, although there was no
explicit justification for such an interpretation,
except perhaps for ethnographic parallels in the
American Southwest (Graves 1998).
Culture historians could describe assemblages
in terms of artifact types and then arrange those
assemblages into chronological order because the
type-frequency distributions they generated meas-
ured the cultural transmission of selectively neu-
tral variation and the operation of drift in finite
populations (Lipo et al. 1997; O'Brien and Lyman
2000a; Tschauner 1994). Culture historians did
not use such terms, instead crafting interpretations
of variation arranged by aggregates of historical
artifact types such as phases and foci (Lyman et
al. 1997). These interpretations, including migra-
tion, were based on the equation of units such as
phases with cultural groups and the application of
cultural processes generalized from ethnology or
through direct historical analogy (e.g., Sears
1961). As Meggers (1955) noted for migration,
this explanation was used when the types
included in a phase or similar unit measured a
rapidly expanding spatial distribution (e.g.,
Ritchie and Dragoo 1960). Thus, Americanist cul-
ture historians superficially identified migration
in a manner similar to European archaeologists-
as rapid spatial expansion of a suite of artifact
types identifying a cultural group.
Archaeologists continue to explore migration as
an explanation of archaeological variation (e.g.,
Anthony 1990; Burmeister 2000; Cameron 1995;
Chapman and Hamerow 1997; Crown 1994; Kirch
1984; Levy and Holl2002; Rouse 1986; Snow
1995; Spriggs 1993; Zedefio 1995). Not surpris-
ingly, recent studies demonstrate that migration is
still identified by the archaeologically sudden
appearance of new artifact types. In an article
heralding the reintroduction of migration into
archaeological explanation, Anthony divides long-
distance migration into "leapfrogging" and
"migration stream" models and suggests that
stream migration will carry regionally
defined artifact types from a circumscribed
home region to a specified destination.
Innovation in the new home might then lead
to a sort of artifactual "founder's effect,"
resulting in rapid stylistic change from what
was in any case a narrowly defined pool of
variability. [1990:903]
There is, however, little discussion of artifact clas-
sification or methods for differentiating between
the above-quoted scenario and ecological con-
straints on artifact variability in a single popula-
tion, to name one plausible alternative explanation.
Recognizing that artifact classification must be
addressed in any attempt to examine migration,
Burmeister (2000) suggests that archaeological
classes describing adaptive cultural variation may
not be suitable for investigating migration. Using

the example of barn architecture among early
European settlers in North America, Burmeister
notes that a particular barn form constructed by
southern German immigrants was adopted by
other immigrant groups because "of the function-
ality of its design, which was kept simple and had
multiple applications" (2000:541). The alterna-
tive, argues Burmeister, is to examine variation
associated with a particular cultural lineage,
which can "be found in the material culture of the
internal [nonadaptive] domain" (2000:542).
Burmeister, however, provides no necessary and
sufficient conditions that define variation in the
internal domain.
I would argue that the most important method-
ological step in migration research is to define
artifact classes that measure homologous similar-
ity. Only by constructing such units can we be
sure we are trachng variation within a culturally
related population. Once assemblages are
described with such units, or classes, relationships
between the classes can be defined in seriations,
phylogenetic trees, or other models that depict
transmission patterns. After transmission patterns
have been generated for a set of classes, we can
Figure 11.1. Map of the Pacific Islnncls with major island groups labeled. The islands of Near Oceania have been
inhabited,for approximately 35,000 years, whereas those of Remote Oceania have been occ~ipiedfor a maximum
of only 3,300 years.
Migration and Cultural Transmission / Cochrane 135
develop hypotheses to explain the patterns with
reference to the empirical distribution of classes
in space and time. When migration is the focus of
study, we may try to explain transmission patterns
with reference to the movement of individuals
and compare transmission patterns with independ-
ent archaeological evidence such as assemblage
chronologies, genetic affinities between individu-
als in different regions, and transport of artifacts.
This research agenda is exemplified in the follow-
ing case study.
MIGRATIONS IN THE SOUTH PACIFIC
The South Pacific offers many advantages to
archaeologists examining migration. The various
island groups provide environmentally bounded
locations for archaeological assemblages and the
populations that deposited them. There is also a
long history of research employing migration-
based explanations of cultural variation (e.g.,
Bellwood 1991; Green 1963; Kirch 1996; Kirch
and Green 1987). And, for some areas of the
Pacific, population diversity in terms of biology,
language, and ethnography has been richly
described, giving the archaeologist additional
136 Cultural Transmission and Archaeology: Issues and Case Studies
potential data sets to investigate in migration-
based research.
The west-to-east colonization of the Pacific,
from the Bismarcks to Easter Island (Figure 11.I),
has been substantiated through radiocarbon dating
of cultural deposits and simulated voyaging
(Anderson et al. 2000; Finney 1979; Irwin 1992).
Change in the material culture of an island or
island group is often seen by archaeologists as the
result of the arrival of new human groups (see ch.
4). Perhaps the most well-known group of sup-
posed immigrants are the Lapita peoples, whom
most archaeologists suggest arrived in the already
inhabited circum--New Guinea islands and began
making Lapita pottery circa 3500 B.P. (Kirch
1997). Green summarizes the consensus view,
noting that
the onset of the Lapita horizon is an out-
come of intrusions by tiny populations car-
rying with them common cultural, biologi-
cal and linguistic components. The evidence
suggests that in the first instance, popula-
tions speaking Eastern Malayo-Polynesian
and pre-Broad Oceanic moved into the
cluster of islands that were within the
Bismarck Archipelago and just off-shore of
the already inhabited and much larger main
islands. [2003:1121
The proposed migration of Lapita peoples into
the Pacific is not the only set of possible migra-
tions that archaeologists have investigated. Fiji
has been the focus of considerable debate con-
cerning the identification and explanation of
migration (Hunt 1986), beginning with hypothe-
ses developed by Fornander:
During the first and second centuries of the
Christian era many and properly organized
migrations of the Polynesians into the
Pacific Ocean took place from various
points of the archipelago. . . . [Tlheil- gen-
eral rendezvous during this migratory
period was on the Fiji group, and princi-
pally on the west. . . . [Tlhey were of supe-
rior cultivation to the Papuans then and now
inhabiting that group. . . . [Tlhey stayed
there long enough to introduce a large
amount of their vocables in the Fijian lan-
guage and no inconsiderable part of their
legends and customs. [1969:2 (1878-1885)l
In recent years, scholars have continued to
interpret variation in the Fijian archaeological
record as indicating the arrival of migrant popula-
tions. Best (1984, 2002) and Burley (2003) sug-
gest that some Fijian ceramic change is likely a
result of migrants arriving in the islands. In par-
ticular, Best (2002) states that ceramic changes
that occurred circa 1750 B.P. in Fiji possibly
resulted from the arrival of voyagers from north-
ern Vanuatu. Best argues that ceramic decoration
on vessels from Lakeba Island in eastern Fiji is
similar to that on vessels from Vanuatu.
Additionally, Best observes that obsidian chemi-
cally sourced to northern Vanuatu occurs in the
Lakeba deposits at approximately the same time
(but see Bedford and Clark 2001). Burley (2003),
after examining the ceramic record at the
Sigatoka Sand Dunes on the main island of Fiji,
notes that changes that occurred circa 1500 B.P.
are so striking that any interpretation other than
ethnic group replacement seems unlikely.
Scenarios such as these are possible but diffi-
cult to evaluate because, with few exceptions
(e.g., Allen 1996; Cochrane 2002a, 2002b; Graves
and Cachola-Abad 1996; Sinoto 1962), Pacific
archaeologists have not constructed artifact classi-
fications with the principle goal of measuring
homologous similarity. Instead, the ability of
most classes to track homology is typically
assumed. I examine this assumption below using
pottery from western Fiji and Vanuatu. I first look
at several classification issues and then discuss
how the classes I created were used in a cladistic
analysis.
Classification Issues
Fiji, described as "a sort of 'between place'--a
foyer of exchange and interaction" (Kirch
2000:156), was colonized circa 2900 B.P. by pop-
ulations arriving from the west carrying Lapita
pottery (Anderson and Clark 1999; Frost 1979).
The analysis described here focuses on excavated
and surface ceramic assemblages from the Yasawa
Islands of western Fiji that I collected (Cochrane
2002a, 2004) and on published descriptions of
Vanuatu pottery (Bedford 2000, 2001 ; Bedford
and Clark 2001; Bedford and Spriggs 2000). The
Yasawa Islands were initially inhabited circa 2700
B.P., or approximately two centuries later than the
first occupation of Fiji. The earthenware assein-
blages in the Yasawas are distributed across the
prehistoric sequence, from late Lapita assemblages
 Migration and Cultural Transmission / Cochrane 137

Figwe 11.2. Potteryfronl Fiji rmd V a n u a f ~Top:

~ . sherclsfmn Yasawa Islar7ds (slzowz half'size). Bars indicate
vessel interioi: Shei-ds genemllj decrease in age,froin top to bottom rrnclfi-oin left to right. Botfonz: scmple of
vessel dm,~.ingsfronzBedfbi-d (2000, 2001; Bedford and Clark 2001), with scale b a n in ceiztinzetei-s when
included in originals. Vessels generally decrease in age fiorn top to bottom arzdj5onz lejt to righf.

to deposits dating to the nineteenth century
(Figure 11.2). There currently is little evidence for
large-scale craft specialization associated with pot-
tery manufacture in Fiji. Most pottery appears to
have been manufactured and used within small
regions (Dickinson 2001) by a population that was
not functionally integrated into larger production
systems.
Like Fiji, Vanuatu was first colonized by sea-
farers carrying Lapita pottery who sailed into
Remote Oceania leaving behind the intervisible
islands of Near Oceania. The first populations in
Vanuatu arrived circa 3000 B.P. and inhabited
locations at river mouths and bays with access to
marine resources. Archaeologists are still refining
the ceramic sequence of Vanuatu (e.g., Bedford
2000, 2001; Bedford and Clark 2001; Bedford
and Spriggs 2000; Spriggs 2003, 2004), and there
appears to be considerable regional variation
across the archipelago. However, the earliest
deposits on Vanuatu invariably contain Lapita
decorated pottery and plain pottery, largely simi-
lar to assemblages found across the Lapita hori-
zon (Green 2003), including Fiji. After circa 2800
B.P., Lapita decorated pottery no longer was
made on Vanuatu, being replaced by multiple
decorative traditions found throughout the islands
(Bedford and Clark 2001). Apparently an ace-
138 Cultural Transmission and Archaeology: Issues and Case Studies

ramie period existed on some Vanuatu islands
circa 1200-600 B.P. (Bedford 2001).
To investigate the possible effects of migration
on transmission patterns defined from Vanuatu
and Yasawa Islands pottery, the pottery must first
be described using units (classes) designed to
track homology. For Fiji, previous ceramic classi-
fications have been constructed using the princi-
ples of numerical taxonomy. A large nuxber of
observations are made, and statistical grouping
procedures are subsequently used to arrange
empirical specimens into sets (e.g., Best 1984;
Clark 1999; Frost 1974). The meaning of these
groups-that is, the necessary and sufficient con-
ditions for group membership-is often not logi-
cally linked to a particular analytical goal
(Dunnell 1986; O'Brien and Lyman 2000a,
2003a). Further, the process that such units are
measuring is unknown. Obviously, this approach
will not work if we require our units to measure
predominantly homologous similarity.
Using evolutionary theory as a guide, classes
will more likely measure homologous similarity if
the dimensions that define classifications do not
describe functional (Dunnell 1978a) variation.
Dimensions that describe putative functional
variation-for example. variation affecting the
Table 11.1. Description of Modes for Dimensions Describing Shouldered-Vessel Rim Sherds.
performance characteristics of ceramic vessels
(see Bronitsky 1986; O'Brien et al. 1994; Schiffer
et al. 1994)-may track convergent similarities in
multiple, similar selective environments. In this
instance, the similarity of artifacts grouped by
functional classes may be a product of unrelated
transmission lineages in those selective environ-
ments. Of course, functional similarities may also
be explained as transmission within a lineage and
the adaptation of a population in a selective envi-
ronment (O'Brien and Holland 1992), but to
avoid mistaking convergent or parallel similarities
for homologous similarities it seems best in many
cases to construct classes using dimensions that
measure nonfunctional similarities, what Dunnell
(1978a) refers to as "style."Archaeologists have
since linked Dunnell's definition of style to the
distribution of selectively neutral variants in
transmission systems (Lipo 2001; Neiman 1995).
How do we construct classes that measure neu-
tral, homologous similarities? To begin, dimen-
sions that measure homologous similarity should
contain modes (mutually exclusive observations)
whose frequency or presence and absence varies
over time and space and without respect to possi-
ble selective environments. As an example, in
some environments we might expect modes

Dimension Modes

Rim curvature (C)
Angle 1 ( A l )
Thickness 1 ( T I )
Rim symmetry (S)
First temper abundance
rank (TMI)
Curves C1 and C2 (each measured between inflection point of V2 and upper termination, viewed from
vessel exterior):
1. At least one curve straight
2. Both concave
3. Both convex
4. Exterior concave, interior convex
5. Both S-shaped
6. Exterior S-shaped, interior convex
7. Exterior S-shaped, interior concave
8. Exterior convex, interior concave.
Degrees from line at V2 (0 degrees). perpendicular to central vertical axis (90 degrees), to chord
defined by L l :
1. > 9 0 degrees
2. > 70 degrees and < 90 degrees
3. < 70 degrees
Thickness of sherd (mm) wall at V2 perpendicular to exterior and interior sherd walls:
l.<6mm
2. > 6 and < 10 mm
3.>10mm
Measured in cross section from V2 to top of rim:
1. Parallel
2. ~ x i e r i o expanded
r
3. Interior expanded
4. Interior and exterior expanded
5 . Contracted
6. Exterior expanded and contracted
7. Interior and exterior expanded and contracted
1. Calcium carbonate (C)
2. Ferromagnesian (FM)
3. Quartzo-feldspathic (QF)
4. Lithic fragments (LF)
5. Void
 Migration and Cultural Transmission 1 Cochrane 139

describing rim forms and some ceramic surface
modifications to be distributed in this manner, as
variation in these dimensions does not seem to
offer any differences whose preferential transmis-
sion could be explained as a result of selection
(e.g., Lipo 2001b; Sterling 2001). In contrast, the
distribution of modes describing the thickness of
cooking vessels may be explained as selection
sorting culturally transmitted variation (e.g.,
Braun 1987). Importantly, propositions such as
these are historically contingent and may be falsi-
fied through engineering analyses (e.g., Feathers
1989; O'Brien, et al. 1994; Schiffer et al. 1994)
and the application of population genetics-based
models (e.g., Kohler, et al. 2004; Lipo 2001b;
Neiman 1995; Shennan and Wilkinson 2001) in
particular archaeological sequences. In this sense,
the Yasawa Islands ceramic classifications that
follow are initial hypotheses about the processes
that structure variation (see Cochrane 2002a).
Variation in vessel rim form, surface modifica-
tion, and temper was explored in the Fijian col-
lections to develop the classification presented
here. Archaeologists working in Fiji (e.g., Best
1984; Burley and Dickinson 2004; Clark 1999;
Cochrane 2002a, 2004; Green 1963; Hunt et al.
1999) have recorded variation within these
dimensions across the archipelago's ceramic
sequence, and it appears unrelated to different
selective environments. Testing the selective neu-
trality of these dimensions through engineering
analyses and transmission models is a goal of
future research. Class construction was not car-
ried out with reference to Vanuatu ceramic collec-
tions, as only published pictures of typical
Vanuatu materials were available. Details of the
classificatory process and preliminary evaluation
of the homologous character of the classes in

Figure 11.3. Top: rrm cross sections with ternplates for nzeasurirzg character-Ashown m Table 11.1. More
characters are shown in the figure tlzan are disc~isseclhere (see Cochr-arze 2004). Bottom: sclzematic
representations of character states for rim synzmeti35 rim cuwature, rim thicknes~,and rim angle.
140 Cultural Transmission and Archaeology: Issues and Case Studies
relation to the Yasawa Islands assemblages are
presented in Cochrane (2004). Classes con-
structed from dimensions measuring rim and tem-
per variation were applied to shouldered vessels
in the Yasawa Islands and Vanuatu collections.
The classification includes the dimensions rim
curvature (C), rim angle (A1), rim thickness (TI ),
rim symmetry (S), and the first temper-abundance
rank (TMl [Table 11.1, Figure 11.31). A class def-
inition consists of the modes for each dimension.
For example, class 2321 1 describes a rim with
both curves GI and C2 concave, angled at less
than 70 degrees, a rim thickness between 6 mm
and 10 mm, parallel rim symmetry, and calcare-
ous grains as the most abundant temper type by
rank. This classification produces a total of 2,304
classes, with 142 of the classes containing mem-
bers in the Fijian assemblages. When applied to
the Yasawa Islands assemblages, 14 of the 142
classes have four or more members and a total of
120 rim sherds. By using these 14 classes to ana-
lyze transmission patterns, we are examining only
the most common similarities that may result
from transmission (see O'Brien and Lyman
2003a).
When the classification is applied to the pub-
lished drawings of rim cross sections from
Vanuatu assemblages found in Bedford's publica-
tions (Bedford 2000, 2001 : Bedford and Clark
2001), 18 of the possible 2,304 classes have
members. The pottery depicted in the work of
Bedford and his colleagues appears to be type
specimens or single sherds and vessels that repre-
sent the "mean" of variation found among a col-
lection of similar-looking specimens. This has
four ramifications for the present analysis. First,
we cannot know how Bedford's drawings repre-
sent variation in the underlying ceramic popula-
tion, but we can assume that they depict the most
common forms described by a mix of morpholog-
ical and decorative characteristics. Second, the
type specimens must be treated as rough guides
for the abundance of a particular class in an
assemblage, given that no counts of particular
forms depicted are available. Third, the ability of
the classifications used here to track homologous
similarities cannot be rigorously evaluated. We
must assume at this point that the classification
built from the Fijian materials measures homolo-
gous similarity when applied to the Vanuatu
assemblages. Fourth, the illustrated rim cross sec-
tions do not depict temper variation, one dimen-
sion of the classification used here. However,
Bedford does generalize about temper variation,
indicating that assemblages either contain calcare-
ous-tempered sherds (Bedford and Clark 2001) or
lack any calcareous sand tempers (Bedford 2001).
Comments such as these were used to determine
the appropriate mode in the TM1 dimension.
Cladistic Analyses
After artifacts have been classified, we can use
several methods to define transmission patterns
that might account for class similarities and the
empirical distribution of classes in space and
time. The classification used here was originally
developed as part of a larger project to explain
transmission patterns in the Yasawa Islands
(Cochrane 2004). In that analysis, phylogenetic
transmission patterns were explored with cladisti-
cally derived trees generated using maximum-par-
simony methods in PAUP'"Swofford 1998).
Multiple clades, or groups of related transmission
lineages at several hierarchical levels, were
defined and arrayed against a time line (Figure
11.4). These phylogenetic trees of ceramic classes
(taxa) share similar topologies, but are products
of different outgoups, with no outgroup identified
as more appropriate than another (see O'Brien
and Lyman 2003a for details of cladistic analysis
of archaeological materials). The similar topolo-
gies depict a clade of recent ceramic classes
evolved from a common pool of ancestral varia-
tion-a "common ancestor" in cladistic terminol-
ogy-circa 500 B.P.
On the bottom tree this recent clade shares a
common ancestor with other ceramic classes,
whereas in the top tree the recent clade is derived
directly from the outgroup class. Additional
cladistic analyses of 10 classes with six or more
members produced similar trees. The phyloge-
netic trees presented here do not depict groups of
people associated with different ceramic classes
splitting off from each other in a kind of branch-
ing cultural group evolution (cf. Mace and
Holden 2005). Indeed, there seems little empirical
warrant for associating specific artifact classes
with a particular set of people in this instance.
Instead, these trees depict hypotheses about the
typical route of transmission, linking ceramic
classes in a population within particular spatial
and temporal boundaries. If the transmission pat-
terns depicted in these trees are accurate, we can
suggest processes that may explain characteristics
 Migration and Cultural Transmission 1 Cochrane 141

Figure 11.4. Pliylogenetic

trees representii~g
lzjpthesized relationships
ainong 14 Ynsacva Islands
c ~ m n i i cclasses using different
o ~ ~ t g r o ~Class
~ p s . defznitio~z,~
(rim c u r ~ ~ a t ~LC],
w e rim angle
[ A l l , rim thickness [TI], rim
.syiizmetiy 131, and the first
tenzpei--ahr~nrlnnc.e mnk
[TMI 1) are ,slzoci:iz next to
each rim illustration. These
are 50 percent majoriprule
co~~serisi~s trees genemted
,from 1,974 e q ~ ~ a l l y
pal:siiizonious trees.
Illustrntioiis of classes
(interior of vessel to lep)
convey chaizrcteristics ofthe
class only and clo 1201 depict
actual speci~izens.Gray rinzs
clirplc~ythe FM temper
character state for the
chasacter TMI. Branch
lengths and node positions
depict npproximate
clzroizological origins and
ertiizctions of classes only.

classes in a cladistic analy-

sis with classes from the
Yasawa Islands. To facili-
tate the analysis, classes
with the most empirical
members and associated
with particular temporal
ranges were used. For the
Yasawa assemblages this
includes five classes with
six or more members, and
for the Vanuatu assem-
blages this includes six
classes of the 18 identified
among the drawings of
Vanuatu materials. The six
of the trees having to do with, for instance, clade Vanuatu classes were chosen based on written
origin and extinction. For example, the recent descriptions referring to particular sherd and
clade, whose classes occur in assemblages across vessel drawings as the most prevalent in assem-
the Yasawa Islands, might be a product of envi- blages. The Yasawa Islands and Vanuatu ceramic
ronmental and demographic changes in Fiji that classes used for the combined analysis are listed
influenced cultural transmission (Cochrane 2004; in Table 11.2.
Cochrane and Neff 2006). Ceramic classes were grouped into an early
In a preliminary attempt to investigate popula- period and a late period. The early period con-
tion movement in this region of the southwest tains classes where the majority of their temporal
Pacific, I have combined the Vanuatu ceramic distribution occurred circa 3000-2 100 B .P., and
142 Cultural Transmission and Archaeology: Issues and Case Studies
Table 11.2. Classes Used in the Cladistic Analysis of Yasawa Islands and Vanuatu Pottery.
Class Definition (Modes) Location
Yasawa Islands
Vanuatu
the late period contains classes where the major-
ity of the distribution occurred circa 2100-500
B.P. Class 12211 was used in both analyses.
Using the cladistic procedures described above,
the six early classes are arranged into five equally
parsimonious trees. The outgroup is the class with
the oldest and shortest temporal distribution. Each
tree is described by a consistency index (CI) of
.83 and a retention index (RI) of .80 (Figure 11.5,
top). A similar analysis of late classes produced
three equally parsimonious trees with a CI of .83
and RI of .80 (Figure 11.5, bottom). The outgroup
for the late tree is the class with the oldest and
most restricted temporal range in the Yasawa
Islands ceramic classes. Using the oldest and
most temporally restricted class from Vanuatu
produces the same tree. Although the high CI and
RI values of these trees suggest that they are good
approximations of a true phylogeny (ch. 5), the
small number of classes and invariant character
states make these indexes less accurate descrip-
tions of tree robustness.
To better estimate the ability of these trees to
depict phylogeny, I conducted a bootstrap analy-
sis of the data matrices used to construct the trees.
In bootstrap analysis of cladistic data a new
matrix of classes and their modes is created by
randomly sampling with replacement from the
original data matrix, creating a pseudoreplicate.
This is done many times, and the most-parsimo-
nious trees are then generated from each pseudo-
replicate and combined to create a majority-rule
consensus tree (see Kitching et al. 1998; O'Brien
and Lyman 2003a). Clades found in the consensus
tree constructed from the pseudoreplicates are
less likely to be products of chance mode associ-
ations in the data matrix. Two clades occur in the
50 percent majority-rule consensus tree generated
from 100 bootstrap replicates and are less likely
to be chance constructions: clade 12321-22121-
Temporal Distribution (years B.P.)
12121 in the early time period and clade 2225 1-
2235 1-12351 in the late period.
Clades at different levels in the phylogenetic
hierarchy in the early- and late-period trees con-
tain both Vanuatu and Yasawa Islands ceramic
classes, suggesting that these clades depict trans-
mission within a combined Vanuatu-Yasawa
Islands population. We can examine the geo-
graphic distribution of classes within early and
late clades to determine the relative importance of
transmission, and possibly migration, between the
two archipelagos over time.
The approach is simple and follows a proce-
dure developed by Slatkin and Maddison (1989)
for estimating gene flow among demes. A tree is
constructed from character data that arrange
unique classes from two or more geographic loca-
tions into a series of bifurcating relationships.
Geographic location, though not used to construct
the tree, is then mapped onto each terminal class.
The geographic location associated with nodes in
the tree is determined recursively by assuming
that geographic location is a multistate, unordered
character and is distributed across nodes and ter-
minal classes using a parsimony criterion.
The tree in Figure 11.6 demonstrates the pro-
cedure. Nine terminal classes at the tips of the
tree are described by their geographic character
(location 1 or 2). Moving toward the root of the
tree, we assign geographic states to nodes through
a simple majority-rules procedure. Two terminal
classes, each with geographic character 1, arise
from a node with geographic character I (Figure
11.6, circled). If a node gives rise to a class with
geographic character 2 and another class with
character 1, the node is described by both states 1
and 2 (Figure 11.6, gray area). If a node gives rise
to a class with geographic state 1 and a node with
a state 1 and 2, the deeper node is described by
state 1, according to the majority rule (two 1s ver-
 Migration and Cultural Transmission / Cochrane 143

Figure 11.5. Equally parsimonious phjlogerzetic trees genera fed fi-om early (fop)and late (bottoin) Vanuatu and
Yasawa Islands ceramic classes. Geographic locatiorzs of classes are desigimfed b j 'y'(Fiji) or "v" (Varzuatu)
after each class definition. Nodes are labeled with the geographic character detenvirzed thmzigh the Slntkirz and
Maddison (1989) procedure, with the rz~mberof migmtion events, "s, " given below each tree.

sus one 2). Assigning the deeper node geographic
state 2 would require two changes in the state
(from 2 to 1, and from 2 to 1, 2 ) to account for
the distribution of geographic characters in the
subsequent node and class. When taxa or nodes
sharing a common ancestor do not themselves
share geographic characters, a migration event
must have occurred. In Figure 11.6 two migration
events are depicted at nodes whose characters are
underlined.
144 Cultural Transmission and Archaeology: Issues and Case Studies

/--\

Figure 11.6. Tree showing distribution of geographic chamcters aci-oss terminal classes and nodes (adapted fronz
Slatkin and Maddison 1989).The circled area highlights two ter~ilinalclasses with the snrne geogmplzic clznrncter
arising fi-om a node, and the gray area highlights a tei-nzirzal class urzd a node with diferent geographic chnracte~r
arising from a node. fivo nodes indicating migration events ha~letheir geographic character:^ underlined.

Application of the method to the early and late
ceramic trees is demonstrated in Figure 11.5. By
moving from the classes at the tips of the tree,
through the nodes, and to the root, we can deter-
mine the number of possible migration events
(labeled "s") necessary to account for the geo-
graphic distribution of classes in the phylogeny.
Four of the five early trees contain two possible
migration events, and one contains one possible
migration event. The late trees contain one possi-
ble migration event. Although these data sets are
small and the classifications are preliminary, the
phylogenetic trees and geographic character
analysis suggests more possible migration events
during the earlier portions of Vanuatu and Yasawa
Islands prehistory than after circa 21 00 B.P.,
when archaeologists (e.g., Best 2002; Burley
2003) suggest migrations from Vanuatu to Fiji
perhaps took place. Using the phrase "possible
migration events" to describe the early and late
trees is shorthand for a more complicated sce-
nario. It is more accurate to state that when the
early and late ceramic classes are arranged in
phylogenetic trees, the geographic distribution of
classes across each set of trees indicates increased
transmission between geographic areas in the
early period relative to the later period. The geo-
graphic distribution of classes could result from
spatial transmission without migration or a com-
bination of this and migration. To explain these
transmission patterns as a result of migration, we
need to compare the patterns with independent
evidence of migration from the two time periods.
Evaluating Proposed Migration Events
Analysis of biological similarities between
individuals in different geographic locations is
one method for evaluating migration hypotheses.
Assuming genetic transmission coincides with
migration, we might expect early skeletal assem-
blages across Vanuatu and Fiji to display more
genetically based similarities than late skeletal
assemblages. Investigating the accuracy of this
expectation requires large, well-dated skeletal
samples, and these are rare in Fiji and Vanuatu for
the beginning of the prehistoric sequence. A much
larger assemblage of Fijian skeletal material
exists for a later time period (circa 2100-1600
B.P.), including a single set of remains from a site
in northern Fiji and over 60 burials from southern
Fiji. Metric and nonmetric analyses of all these
remains (Pietrusewsky 1989; Pietrusewsky et al.
1994) suggest affinities with populations in
Vanuatu and islands to the west, but also with
populations to the east, including those from
Samoa and Tonga. Analysis of post-2100 B.P.
Fijian skeletal material suggests that the single
possible migration event identified in the late-

ceramic phylogenetic trees (Figure 11.5, bottom)
may result from movement of individuals
between Vanuatu and Fiji, but this does not
negate the possibility of concurrent spatial cul-
tural transmission between archipelagic popula-
tions. The paucity of early skeletal material from
either archipelago inhibits comparison across the
early and late periods.
In addition to skeletal analyses, the distribution
of artifacts compositionally sourced to geographic
provenances can also be used to evaluate cladisti-
cally derived migration hypotheses. Petrographic
analysis (Dickinson 2001) of 418 sherds from a
variety of chronological contexts in Vanuatu and
Fiji has identified no instances of pottery move-
ment between the archipelagos (some exotic
sherds in this analysis have yet to be assigned a
provenance). Only the presence of obsidian from
Vanuatu in Fiji that postdates 1750 B.P. indicates
the movement of individuals from Vanuatu to Fiji.
The effect of movement on ceramic similarities is
possibly reflected in the migration event depicted
in the late-period trees constructed from Vanuatu
and Yasawa Islands pottery.
By comparing the results of both skeletal
analyses and artifact provenance work, we can
suggest a plausible interpretation of the migration
events identified in the early and late phyloge-
netic trees. Although the average number of
migration events across the early trees is greater
than that across the late trees, the migration
events in the early trees appear to depict cultural
transmission across space, as there is currently no
evidence suggesting the movement of individuals.
For the late trees, the single migration event
identified seems likely a result of both the move-
ment of individuals between Vanuatu and Fiji and
contemporary cultural transmission across the two
archipelagos after 2100 B.P. The recovery of
Migration and Cuirbrai Tra-5- ss z - Czc--2-5
,.=
--
. .
Vanuatu obsidian in Fiji suggtsrs rhe Frexc:r-rr
direction of movement.
CONCLUSION
I suggest that it is possible to incorporate research
on migration into a cultural transmission frame-
work, but some aspects of migration research
remain problematic. We are limited to construct-
ing only plausible arguments about migration and
artifact similarity because the only empirical evi-
dence of past movements of populations is found
in human biology (but see Chapter 4). We can
infer the movement of individuals by studying
biological similarities as expressed in DNA (e.g.,
Renfrew and Boyle 2000) and skeletal morphol-
ogy (e.g., y'Edynak 1976) as well as ecological
similarities of individuals evident in bone chem-
istry (e.g., Price, et al. 2001). Our observations in
these fields, however, are generated and explained
within a system different from that used to make
observations of material culture similarities.
Although many of the mechanisms and concepts
used in biological and cultural transmission
research are similar, in the end biological trans-
mission, as indicated by DNA similarities, for
example, does not necessitate cultural transmis-
sion between individuals and vice versa.
How, then, can cultural transmission research
aid the study of past human migration? My
answer is that a cultural transmission framework
can be used to define homologous similarities and
lineage continuity across geographic areas instead
of only the general similarities, of unknown
meaning, identified using other frameworks.
Without methods for defining homologous simi-
larities and cultural lineages, the results of migra-
tion research may mistakenly equate like-looking
artifacts with the existence of a culturally related
people in multiple geographic areas.