Semes and Genes in Africa

Corresponding author:

Barry S. Hewlett

Department of Anthropology

Washington State University

Vancouver, WA 98686

360-546-9449 (0)

541-919-3144 (H)

hewlett@vancouver.wsu.edu

This paper has two general aims: (1) to explain the distribution of cultural practices and beliefs across the landscape in Africa; and, (2) to demonstrate how genetic, linguistic and geographic information can be utilized to better understand the nature of culture. We prefer to use the term "semes" to refer to specific culturally transmitted units (e.g., schema, knowledge, or practices) rather than "memes" (Dawkins 1976, Durham 1991, Boyd and Richerson 1985) because the former has linguistic roots in signal and semantics and emphasizes the symbolic nature of culture while the latter term has its roots in mimic which emphasizes the imitative nature of culture (Cavalli Sforza, personal communication). Of course culture has many properties, but the symbolic nature of culture and its transmission is especially characteristic of humans. We focus on cultures that share specific semes and utilize genetic, linguistic and geographic data to evaluate the process that help to explain why disparate cultures have similar semes.

The approach we describe is sometimes call "coevolutionary" or "dual inheritance" because it identifies relationships between genes and culture. We prefer to call the approach "evolutionary cultural anthropology" (Hewlett and Lamb in press) because the major theoretical contributions to date (Cavalli Sforza and Feldman 1981, Boyd and Richerson 1985, Durham 1991) emphasize understanding evolutionary mechanisms and properties of culture, the domain of cultural anthropology. This approach has been criticized because it gives the impression individuals are passive recipients of culture. In this paper we indicate that the acquisition of culture is going to vary across the life-span--active evaluation of semes occurs throughout an individual's life, but substantial modifications of semes are simply more likely to occur as children get older. Human children are clearly more likely to imitate others than are juveniles in our closest living relatives, the chimpanzees, suggesting that particular kinds of imitation are uniquely human (Tomasselo 1999). Good reasons exist for imitating--it reduces the energy it takes to learn everything through trial-and-error (i.e., evaluating the cost and benefit of everything we do). While imitation is central to the human child's experience, this does not mean that this is the only way a child acquires culture; anybody who has been around children knows that children are actively engaged in manipulating and understanding their environments.

Approaches that examine the relationships between biology and culture are also not looked upon favorably in anthropology, especially by cultural anthropologists. This is in part due to the fact that most cultural anthropologists link discussions of biology and culture with discussions of race and culture. The father of American anthropology, Franz Boas, countered rampant racism in the early 1900s by demonstrating that "races" did not have inherently different biologies and cultures. He did not think the processes were connected in any way because biological mechanisms were incredibly slow to change while culture could change very rapidly. Human history, therefore, was interpretable in terms of cultural forces only rather than biological ones. Boas was interested in explaining cultural diversity and rejected the notion that race or biological differences could help explain this diversity. Discussions of race and culture were major components of early anthropology textbooks (Boas 1938, Kroeber 1923) and the fight against racism and ethnocentrism continue to be central tenets in most anthropology courses so it is not surprising that anthropologists have been reluctant to consider relationships between biology and culture. In this paper we try to demonstrate that an understanding of cultural mechanisms can help to understand why biology (genes) and culture can coincide. This does not mean that the biology determines culture. Indeed, culture often determines genes as in the case of adult lactose absorption (Cavalli Sforza and Feldman 1976; Durham 1991). This paper is consistent with a Boasian perspective in that we are interested in trying to understand how particular culture histories are utilized to explain cultural diversity in Africa, but unlike Boas, we utilize biology (genes) and language as tools for interpreting that diversity. This paper extends Boas' and Kroeber's work by identifying specific cultural mechanisms and explanatory models which help to explain cultural diversity and interpret cultural histories.

The paper asks a simple, but enduring anthropological question: "Why do cultures share some semes?" Three broad explanatory models are usually offered: (1) "cultural diffusion" or borrowing or diffusion of the seme from neighbors, (2) local adaptations where individuals develop similar semes to adapt to similar natural and social environments and, (3) "demic diffusion" or the movement of peoples and their semes to a new area. The first explanatory model was a trademark of the Boasian and Kroeberian cultural relativist tradition in anthropology. This tradition de-emphasized the "adaptive" nature of culture (i.e., the impact of natural and social ecologies) and suggested that semes arose primarily through human imagination and mindful play and could take their own courses (culture as sui generis). This perspective is somewhat similar to postmodern and interpretative approaches where culture history (especially contact with colonial peoples) are emphasized and the "scientific" and cross-cultural approaches are rejected as reductionist or not possible because of the uniqueness of each culture (or even the unique experiences of each individual within a culture).

The second explanatory model is common in cultural anthropology and traces its roots to Julian Steward's (1955) cultural ecology. In the 1970s, cultural ecology was modified and called "cultural materialism" by Marvin Harris, and in the 1980s-90s it was modified further and reemerged as "evolutionary ecology". These approaches considered most semes as "adaptive" (i.e., the seme enhanced the Darwinian fitness of the individuals or groups) to particular natural or social environments. Even in these postmodern times, most introductory anthropology textbooks use this adaptationist model to explain culture.

Finally, "demic diffusion" refers to people moving to a new area and taking their language and culture with them. People can move for any number of reasons--climatic change, resource availability, warfare, disease or expansion of a territory. Cavalli Sforza et al. (1993) use it specifically to describe a process of repeated expansions of a group that is generated by the development of an innovative technology (e.g., domestication of plants or animals; development of new tools, such as Aurignacian or those of ivory, bamboo, obsidian, bronze or iron; and, means of transportation, such as sleds, rafts, boats and domestic animals) or type of social organization (e.g., type of warfare or institutional innovation). The innovation leads to population growth, migration and intermarriage with those without the technology (i.e., gene flow in the direction of those with new technology). The people take their new technology, genes, language and culture with them.

The semes that move with people may or may not be "adaptive"; many are likely to be neutral. Some semes may be adaptive in some environments or circumstances and neutral or non-adaptive in others. The semes are conserved through specific mechanisms of cultural transmission (see below). The demic diffusion model is seldom used by cultural anthropologists to explain why cultures may share semes, whereas archaeologists utilize this model more frequently (Trigger 1968; Lyman et al. 1997). But archaeologists tend to emphasize the local innovation and adaptation model to explain similarities in semes, in part, because data collection focuses on recovering local cultural and environmental data. Cultural diffusion may be popular among postmodern and ethnohistorical anthropologists because it is so easy to see in today's world-- Western capitalistic culture is rapidly diffusing throughout the world without the movement of Euroamericans. New technologies, such as TVs, films and the Internet enable rapid cultural diffusion.

In order to evaluate the distribution of semes across the landscape it is necessary to identify specific cultural mechanisms of transmission that underlie the three explanatory models described above. Table 1 identifies and

briefly describes four mechanisms. The first is called vertical transmission and is most similar to genetic transmission. Mathematical analysis has shown that semes are highly conserved when transmitted in this way (Cavalli-Sforza and Feldman 1981). Vertical transmission is especially pronounced in infancy and early childhood, in part, due to parent-child proximity and the attachment process. The second is called "group effect" by Cavalli Sforza and Feldman (1973) and "frequency dependent bias" by Boyd and Richerson (1985) and refers to the process whereby individuals acquire semes that occur frequently in the population. Individuals assume that a seme that is frequently observed or expressed is likely to be adaptive. We will refer to this mechanism as "group effect" in Table 1 and the remainder of the text since it was the first to be described. Like vertical transmission this mechanism tends to maintain the status quo. Henrich and Boyd (1998) argue that vertical transmission alone is not sufficient to explain group level conservation of semes and that group effect increases the frequencies of a seme beyond what is expected from vertical transmission. Vertical transmission and group effect are the cultural mechanisms by which semes are conserved in demic diffusion.

The third type of transmission is called horizontal and is based upon epidemiological models of disease transmission. As the frequency of interaction or exposure to an unrelated individual with a seme or disease increases the greater the likelihood a person adopts a seme or catches a disease. This is often the way in which innovations spread from community to community. As the frequency of interactions increases between individuals of different communities the greater the likelihood they adopt aspects of each others' cultures. Three general types of horizontal transmission are usually distinguished: (1) between generations (called oblique); (2) within a generation (the origin of the term "horizontal"); and, (3) one-to-many, which is relatively unique to highly stratified urban industrial societies where teachers, leaders, TV and the Internet transmit information. Culture change with horizontal transmission can be rapid depending upon the frequency of exposure to semes; the one-to-many form is especially conducive to rapid culture change and is common today but was rare in the past. Horizontal transmission is the prime mechanism by which cultural diffusion takes place.

Trial-and-error is the fourth item in Table 1 and is a process that contributes to local innovation and adaptation, the third explanatory model. Individuals observe or hear about alternative semes and critically evaluate the advantages and disadvantages of each. This evaluation may lead to a synthesis of existing semes (i.e., recombination) or the development of an entirely new seme. The innovative seme is, at first, often transmitted horizontally. Trial-and-error takes place from infancy onwards, but a substantial synthetic or innovative seme that is transmitted to others is more likely to emerge in adolescence or early adulthood. The use of trial-and-error by peoples in similar but distant natural and social environments may lead to the development of similar semes, a process similar to evolutionary convergence. Julian Steward (1955) used the term "multilinear evolution " to refer to the method of examining distant cultures with similar semes to determine which aspects of culture were in fact "adaptive".

Stability of environment influences the utility of the four mechanisms or processes. When environments change very slowly, adaptive knowledge can be obtained at the level of vertical cultural transmission because only modest updating of knowledge is needed in order to respond to gradual changes in selection pressures. It may be, in fact, difficult to distinguish between genetic and vertical cultural transmission mechanisms that are responsible for a particular behavior since both are vertical and lead to conservation of the seme or gene. The conservation of both culture and genes has probably led some researchers to incorrectly attribute genetic causes to human behaviors. In contrast, where environmental change is very rapid, or when there are sudden environmental shifts, individuals should favor horizontally (within generation) transmitted information and trial-and-error. In such environments, genetic systems will change too slowly to cope, and information from the parental generation is likely to be outdated and error prone.

This paper utilizes genetic, linguistic and geographic data to try and determine which of the three explanatory models (i.e., cultural diffusion, demic diffusion or local adaptation) can be utilized to explain why two or more cultures share a specific seme in Africa. Table 2 summarizes eight patterns of genetic, linguistic and geographic data that

emerge from the three explanatory models. Demic diffusion assumes two or more cultures share semes because they share a common past so genetic and/or linguistic similarities are predicted. To try and control for the effects of cultural diffusion, the cultures should also be geographically distant from each other. Cultures are likely to share semes because of demic diffusion and associated mechanisms of cultural transmission (vertical and group effect) when they exhibit genetic, linguistic and geographic patterns (1), (2) and (3) in Table 2. With these patterns, cultures that share semes are far apart from each other and share genes and/or language.

Cultural diffusion assumes cultures share semes because they regularly interact with each other, so cultures that are geographically close to one another are expected to share more semes. In order to control for other factors, the cultures should not speak the same language and/or share many genes. Patterns (4), (5) and (6) are most likely to demonstrate cultural diffusion and horizontal transmission as cultures are geographically close, and have either different genes or different languages. Pattern (6) is the best measure of cultural diffusion while (4) and (5) are potentially confounded by factors of demic diffusion.

Pattern (8) in Table 2 indicates that cultural similarities could be explained by demic diffusion, cultural diffusion or local adaptation, while pattern (7) is the best for predicting local adaptations as the two or more cultures share semes, but do not share genes or language and are geographically distant.

The Cultures and Types of Distance Measures

In order to conduct this study we were interested in locating cultural (seme), genetic, linguistic and geographic data on the same ethnic groups. The primarily limiting factor in selecting a sample was the availability of genetic data. The Ethnographic Atlas (Murdock 1967; Gray 1999) provides cultural data on over a thousand cultures, Ruhlen (1991) and Grimes (1978) provide linguistic classification data on most of the worlds' languages, and it would be easy to determine geographic distances between any two cultures. We decided to use autosomal genetic markers to calculate genetic distances rather than DNA markers because only a few African populations have been examined for DNA markers. For instance, the most recent study of African mtDNA genetic distances was based upon 20 individuals from 13 ethnic groups (Ingman et al. 2000) and the most recent study of African Y-chromosome genetic distances was based upon 13 individuals from 8 ethnic groups (Underhill et al. 2000). Rather than trying to calculate genetic distances ourselves we first turned to Cavalli Sforza et al.' comprehensive volume (1994), but found that many ethnic groups combined genetic data from similar groups (for instance, Mbuti genetic data combined data on Ituri Aka, Mbuti and Efe; Akamba genetic data combined data on Gusii, Kikuyu and Giriama). Eventually we turned to the genetic data base maintained at Stanford University. Eric Minch, who compiled and was responsible for the database at the time, provided a list of genetic marker data for various groups of African populations. From that list we were able to identify genetic data specific to 42 ethnic groups. Six cases had to be eliminated because corresponding cultural data in the Ethnographic Atlas did not exist.

The remaining 36 cultures provided the basis for all comparisons and are listed and located in Fig. 1. The 36

ethnic groups had data on13.97 loci on average (range 7-26) and 34.5 independent alleles (range 16-74). Genetic distances between a pair of ethnic groups were based upon an average of 22.7 independent alleles (range 14-70). Nei's (1972) method was utilized to calculate the genetic distance for each pair of ethnic groups.

There were a potential of 28 genetic loci with potential alleles. The loci included: A2M (alpha-2-macrogloblulin), ABO, ACP (acid phosphatase), ADA (adenosin deaminase), AK (adenylate kinase), BF (properdine factor B), DIA (diaphorase), Fut2 (fucosyl transferase), FY (Duffy), GC (group specific component), GLA (glyoxalase), HLA-A, HLA-B, HP (haptoglobin), JK (Kidd), LE (Lewis), LU (Lutheran), MNS, P1, PepA (peptidase A), PGD, PGM1, PGM2, PTC tasting, RH, TF (transferrin).

Cultural distances between each pair of societies were calculated utilizing a method similar to Driver and Kroeber's (1932) "G" statistic. Each pair of cultures (36 cultures results in 630 different pairs) was compared on the number of similarities and differences on 42 categories from the Ethnographic Atlas (1967). Each category had several alternative semes and each culture was coded for one seme within each of the 42 categories. If either of the cultures had missing data for a category, no comparison was made in that category. The average number of categories two cultures were compared was 33.2 with a range of 22-42. For instance, both Zulu and Fulani had 30 of the 42 possible categories coded in the Ethnogrpahic Atlas. Of the 30, Fulani and Zulu had the same code (i.e., shared a seme) on 16. For example, for one category called "mode of marriage" both groups were coded as practicing bride wealth rather than bride service, gift exchange or other alternatives. Therefore the Zulu-Fulani comparison had a similarity index of 0.53 (16/30) and cultural distance of 0.47 (1-.53).

It is surprising that a standard way to calculate linguistic distances does not exist. Regional studies have utilized percentage of cognates from the Swadesh list as the measure of linguistic distance (Jorgensen 1969), but a Swadesh list for each of the 36 cultures did not exist. Current worldwide evaluations of similarities between genetic and linguistic trees are qualitative. Ruhlen (in Chen et al. 1995:599) indicates he has developed a way to evaluate linguistic distances, but indicates "These distances were subjective estimates of their affinities based on available linguistic evidence." The methods are not replicable and the linguistic distance matrix was not published.

Since a Swadesh list was not available for most of the cultures and a linguistic distance method did not exist, we developed a method somewhat similar to that described above for cultural distances. Ruhlen's (1991) classification of languages was utilized to determine the number of linguistic categories that two languages were similar and different. There are potentially 15 levels of classification in African languages and up to 13 classes within each level (usually there are only two or three). For instance the phylum is the highest taxonomic category and there are four phyla in Africa (Nilo-Saharan, etc.). Each culture was given a numeric code for each level classified by Ruhlen. For instance the following gives the linguistic coding between Zulu and Fulani.

The first column represents the phylum and both are in the Niger-Kordofanian phylum, which was coded as (2) out of the possible four; the second column is the second linguistic level which distinguishes between Kordofanian (1) and Niger-Congo (2) and both are in the latter; the third column is the third linguistic level which distinguishes between Mande (1) and Niger-Congo Proper (2), and both are in the latter; the fourth column represents the fourth linguistic level which distinguishes between West Atlantic (1) and Central Niger-Congo (2) and the Fulani fall into (1) and Zulu into (2), and so forth and so on. The Fulani and Zulu share 3 of 15 categories which results in a linguistic similarity of 0.20 and a linguistic distance of 0.80 (1-.20). The phylum was the highest level of possible similarity so if two languages were from different phyla they received a similarity score of zero and a linguistic distance of 1.0 (1.0-0=1.0).

Geographic distances between two cultures were calculated using the Haversine Formula (Sinnott 1984). The formula uses spherical trigonometry to calculate great circle arcs. This "as the crow flies" measure is limited as it does not take into account physical features, such as mountain ranges, rivers, swamps or other things that may help or hinder the movement of peoples or semes.

This paper also utilizes another measure, called the "clustering index", developed in a previous paper (Guglielmino et al. 1995) to quantitatively evaluate the opportunities members of one culture might have of acquiring the seme in question from neighboring cultures. It evaluates density of and geographic proximity to other cultures with the same seme. This index was calculated for the 277 African cultures in the Ethnographic Atlas (EA) that shared particular semes. The index is based upon ratios of distances from nearest neighbors: r = d/dS, where r is the ratio or clustering index for all societies that share a seme, d is the geographic distance to the nearest neighbor and dS is the geographic distance to the nearest neighbor sharing the seme. The clustering index of a particular seme, measures the average ratio of distances for all cultures with that seme. For instance, if a culture practices matrilineal descent and the closest culture in the EA is 10 km away and the closest culture with matrilineal descent is 20 km away, then the clustering index is 0.5. The clustering index for a particular seme is the average index of all cultures with that seme. An index close to 1.0 means very high clustering of a seme--i.e., geographically close cultures share the seme (See Guglielmino et al. 1995 for more details).

Patterns

In a separate paper we will publish (Gugliemino et al. in preparation) genetic, linguistic and cultural trees of the 36 populations, but it is important to briefly summarize here some of the results as they have an impact on the analysis and interpretation of the seme data. First, the genetic tree demonstrates that genetic distances between most African populations are relatively low by comparison to similar trees in the Americas. This is due to frequent intermarriage between ethnic groups, which several ethnographers have noted, especially by comparison to Eurasian cultures (e.g., Goody 1976), and the relatively recent Bantu expansion. The linguistic tree has four distinct branches consistent with the four linguistic phyla in Africa. This was expected because we utilized Ruhlen's classification system, but it supported the usefulness of the linguistic distance methodology described above. The cultural distance tree is of course more complex, in part, because it is influenced more by horizontal transmission that the other two trees (vertical transmission characterizes the transmission of genes and language), but the cultural tree branches did identify three modes of production: hunting-gathering, farming and pastoralism. This suggests that the semes coded by Murdock are often linked to a particular mode of production in Africa.

Table 3. lists the means and standard deviations of the four distance measures and the clustering index.

It is important to point out that as the genetic, linguistic and cultural distance measures increase the similarities for those values decrease. For instance, the average genetic distance is 0.038; this means that the peoples in the 36 African cultures share 96.2 percent of their genes for the alleles measured in this study. As mentioned above genetic distances are low by comparison to other parts of the world. The average geographic distance of 29.88 means the two cultures are, on average, 2988 km apart. The measure in degrees can be converted approximately into km by multiplying by 100. Africa is huge and most cultures in this study are at a great distance from one another. Consequently, it is unlikely that geographic distance will be a useful measure of potential cultural diffusion between cultural pairs in this study because the cultures are so far apart from one another. The clustering index for a particular seme, which is based upon all African cultures in the Ethnographic Atlas, will therefore be utilized to evaluate the opportunity for cultural diffusion.

The linguistic distance of 0.88 indicates that two African languages in this sample have, on average, 12 percent in common. While we believe our linguistic distance method is good for evaluating general relative differences in languages, the method exaggerates actual differences between languages. In particular, it exaggerates differences of two languages from two different phyla; they are assigned a linguistic distance of 1.0 and linguistic similarity score of zero. Most linguistic distance scores were based upon two languages from two different linguistic phyla (see Table 4a for number of languages in each phylum) which contributed to relatively high linguistic distances. If Swadesh or similar lists were available for each language, the measure would be more precise. Even languages from different phyla are likely to have some cognates. The method is more precise for evaluating distances between languages within the Niger-Congo group because there are so many taxonomic subdivisons within this group.

Tables 4a-4c summarize the four distance measures for linguistic phyla, natural environments and modes of

subsistence for the 36 ethnic groups. The seme clustering index is not listed as it is not linked to a specific ethnic groups.

Table 4a demonstrates that the Niger-Kordofonian ethnic groups are near the mean for all measures, but show a slight tendency to be closer genetically than the ethnic groups in the other phyla. The measures for the Afro-Asiatic ethnic groups are also near the means for these measures, but are slightly more heterogeneous in genes and culture than ethnic groups in the other three phyla. Khoisan speakers have more in common culturally than do Niger-Kordofonian and Afro-Asiatic speakers. Nilo-Saharan cultures in this study are the most distinct as they are closer genetically and culturally than the cultures in other linguistic groups. While it is possible that this due to being somewhat geographically closer, on average, to each other by comparison to the cultures in other groups, it is more likely to be a result of the relatively recent emergence of this group. The fact that the Sahara is a relatively recent desert and the genes, languages and cultures in this group show low variance makes this especially likely.

Table 4b is a very general evaluation of the impact of ecology on these measures and suggests that ecology does not have much of an influence on any of them. Cultural ecologists might predict that cultures that share similar natural ecologies should be culturally similar, but no pattern emerges from these admittedly limited data. Table 4c examines the three economic modes of production among the groups. The only pattern to emerge is that the three hunter-gatherer groups (San, Mbuti and Hadza) are the most likely to share many elements of culture by comparison to the other modes of production.

In summary: 1) Khoisan and Nilo-Saharan speakers in this study were more likely to have commonalties in culture than were Niger-Kordofonian or Afro-Asiatic speakers; 2) general ecology did not explain cultural similarities in these ethnic groups; and 3) hunter-gatherer groups in this study had more in common culturally than did farmers or pastoralists.

Table 5 provides the list of semes represented by at least three cultures, the number of cultures out of the 36

that have the seme, their respective measures of distance for the cultures that share that seme, a clustering index for each seme, and a quantitative evaluation of which of explanatory model(s) (from Table 2) fits the data for that seme. Semes are placed into six groups (Kinship, Subsistence, etc.) based upon a contingency table in a previous publication in which chi square tests were run between every pair of semes to determine which were related to each other (Gugliemino et al. 1995). For instance, some may think metal working by men should be placed into the sexual division of labor group, but is statistically linked to social stratification because African class systems are often based upon occupation (i.e., blacksmiths in separate class).

The evaluation of which seme fit a particular explanatory model was based upon z-scores for genetic distance, linguistic distance and the clustering index. The z-scores provide a measure of how different the mean score for a particular seme is from the overall mean. The z-score indicates the direction of the seme as well as the strength of the similarity or difference from the overall mean. A negative z-score for the distance measures indicates the cultures are genetically or linguistically close while a positive z-score indicates they are distant in relation to the overall mean distance. A negative z-score for the clustering index implies the cultures that share the seme are relatively distant from each other, while a positive score implies the cultures with the seme are more clustered. Criteria were established for determining the fit between a seme and an explanatory model: 1) if two or three z-scores were below 0.5 it was not possible to determine a model; 2) if two of the three z-scores were above 0.5 it was considered a trend in a model; 3) if one of the z-scores was greater than 1.0 and another greater than 0.5 it was considered a pattern; and, 4) if two z-scores were above 1.0 and the third above 0.6 it was considered a strong pattern. Any z-score below 0.5 was not utilized to determine a model. A z-score of 0.5 was selected as the minimum score to try and determine a pattern because this score would indicate that about 70 percent of the other seme averages would be below or above the overall mean.

It was not possible to discern an explanatory model for 35 semes (32 %) and 29 of the semes (27%) had two potential explanatory models because one of the distance measures or the clustering index did not reach 0.5. When two possible models are listed in Table 5, the model that is more consistent with the direction of the z-score below 0.5 is listed first. We were able to identify a primary explanatory model for 45 (41 %) of the semes in this study. Given the complex nature of cultural processes, it is not surprising that semes are influenced by several mechanisms and models.

Tables 6 lists the semes with a primary model and Table 7 summarizes the data by seme group. The demic diffusion model explained the greatest number of semes (20) and was especially important for explaining kinship, family and community semes. The data are consistent with recent studies (Gugliemino et al. 1995, Pocklington 1996, Burton et al. 1996, and Jones 1999) which indicate that kinship and social organization in Africa and other culture areas are a result of demic diffusion and expansions of groups of people with particular kinds of kinship and social organization .

The semes explained by demic diffusion and represented by the greatest number of cultures (i.e. >15 cultures) are often thought of as classic features of sub-Saharan African social structure: independent polygynous families with wives in separate dwellings, no marriage with first or second cousins, clan-based neighborhoods and shifting cultivation (i.e., horticulture).

The demic model was also particularly important for explaining political stratification above the community. The data indicate political complexity in Africa is primarily due to expansion of particular peoples rather than cultural diffusion or local innovation and adaptation.

Cultural diffusion explained 12 semes and was especially useful for explaining the distribution of how to build a house and the post-partum sex taboo. Since the clustering index is relatively high for these semes it is also possible that the availability of particular materials in a local ecology may influence the seme (see ecology section below).

Semes that have multiple confounds are equally distributed over all the six seme categories. These semes could be a later stage of demic diffusion in that groups that slowly expanded, shared semes and continued to live in proximity to one another, or the groups always lived next to each other, frequently intermarried and shared semes. Matrilineal clans are a good example of this pattern as there is a well-known "matrilineal belt" across south-central Africa, but we are unable to determine from these data if the distribution of matrilineal clans is due a group of people with matrilineal descent expanding and moving or if it is due to one group developing this descent system and neighbors slowly incorporating it into their cultural system, possibly in order to marry into the group.

Semes linked to local adaptation are particularly interesting because they are the aspects of culture that Julian Steward (1955) was trying to understand with his concept of "multilinear evolution". The four semes listed appear to be variations of demically diffused semes. Particular natural and social conditions have led to the independent development of small (vs. large) extended families, the democratic (vs. hereditary) election of a headman, class elites based upon their control of scarce resources (vs. hereditary classes) and male circumcision in late childhood (vs. adolescence).

Table 8 examines the means of genetic and linguistic distances and the clustering indexes of the 45 semes that

fit into only one explanatory model. As expected, they fit the patterns described in Table 2. Table 8 indicates that the linguistic distances between two groups does not help to distinguish demic versus cultural diffusion; genetic distance and the clustering index are better predictors for understanding these two models. All three measures are important for understanding and predicting the other two explanatory models.

Table 9 examines the relationships between genetic distances, linguistic distances and the clustering indexes. Significant relationships exist between language and culture, genes and culture, and language and the clustering index. The relationship between genes and language is high, but does not reach significance (p=.09). Cultural anthropologists often downplay the relationship between language and culture because they can always point out instances where they clearly do not go together (e.g., Bantu-speaking foragers and farmers in Central Africa both speak the same or similar languages, but the cultures are dramatically different). While there are several exceptions, these admittedly limited data indicate a significant relationship between language and culture in Africa.

Cultural anthropologists argue even more strongly about the lack of a relationship between genes and culture, but again these data indicate otherwise. Hopefully we have made it clear as to why semes and genes may coincide. It is not the result of semes being hard-wired to biology, but due to the fact that both genes and many semes are impacted by the conservative properties of vertical transmission (semes may also be conserved through group effect). Genes are often more conservative than semes due to the lack of horizontal transmission of genes but it is entirely possible that semes may be more conserved over time than genes. For instance, when a population is expanding due to a new innovation (e.g., when Bantu populations grew and expanded as a result of the development of new crops and technologies) language and culture may be more conservative than genes. Women in local foraging cultures may be attracted to male farmers with the innovations because it may increase their access to resources and increase reproductive fitness (called "hypergyny" or women marrying into higher social status). Women and their children become members of the farmers' culture. Consequently, genes in the farmers' group change but the farmers' culture and language are maintained.

The relationship between language and the clustering index indicates that as the proximity between two cultures increases the likelihood that they speak a similar language increases. This is not surprising as most linguistic families and branches tend to be geographically clustered. While cultural difuision may explain some cases, demic diffusion is more likely to explain the clustering of African language families (e.g., expansion of Bantu or Nilotic-speaking peoples).

The paper has provided limited data on the relationship between semes and natural environments. As mentioned above, the natural environment is likely to confound cultural diffusion semes because peoples that live next to each other may share semes as well as natural environments. Ecology may also confound demic diffusion semes as people originating in one natural environment may move to a new location with a similar ecology. In a previous paper (Gugliomino et al. 1995), we examined the relationships between culture and ecology more closely, and found that semes that fit into the demic diffusion model were much less likely to be influenced by ecology in comparison to semes that fit into the cultural diffusion model (6/20 semes under demic model; 9/12 semes under cultural diffusion; chi sq. = 6.11, 1 df, p<.01).

Discussion and Conclusion

In summary, the use of genetic, linguistic and geographic data have provided a better understanding of cultural diversity in Africa. The seme analysis indicated: 1) demic diffusion was important for understanding the distribution of several kinship, family and community semes as well as political stratification semes; 2) cultural diffusion was particularly influential in the distribution of how to build a house and post-partum sex taboo semes; 3) natural and social environments appears to have led to local adaptations and development of small (vs. large) extended families, the democratic (vs. hereditary) election of a headman, class elites based upon their control of scarce resources (vs. hereditary classes) and male circumcision in late childhood (vs. adolescence); and, 4) significant relationships exist between language and culture, genes and culture, and language and the clustering index.

We tried to be cautious in interpreting these data because the number of cultures with genetic data (36) is small. In particular, we limited our discussion/interpretation of specific semes because of the relatively small numbers of cultures represented by some semes. The size and quality (i.e., nuclear and mt DNA) of the genetic database is improving so we hope to conduct more precise studies in the future. Murdock's cross-cultural data have been questioned, but it is admirable that some have recently taken the time to check, extend and improve upon this database (Gray 1999). It would be preferable to utilize emically defined semes, such as myths or beliefs regarding sorcery, and to conduct the study in the field rather than relying on the codes of others, but field studies of this type have not been conducted. Also, semes are not always encoded in language; some semes are experienced directly in social interactions and daily activities. Consequently, it may be necessary to etically define some semes.

But given the limitations the data we have provided some understanding of and explanations for at least some semes in Africa and we also demonstrated that it was possible to utilize genetic, linguistic and geographic (i.e., clustering index) data to better understand cultural diversity and the nature of culture.

Kinship and family semes are very conservative by comparison to other semes and their distribution in Africa (Guglielmino et al. 1995, Pocklington 1996) as well as in other parts of the world (Burton et al. 1996; Jones 1999) appear to be primarily the result of demic diffusion. The conservation is due, in part, to both vertical transmission and group effect. These semes are often transmitted and acquired at an early age and become "marker traits" (e.g., ethnic clothing styles; Boyd and Richerson 1985) that help an individual distinguish in and out groups. Demic diffusion and the associated mechanisms of transmission question the anthropological effort to demonstrate that many of these semes are "adaptive" or "functional" in a particular ecology. While more precise studies of the relationships between semes and ecology are needed, this and our previous study indicate that impact of ecology is limited.

Cultural diffusion and horizontal transmission are of tremendous importance in today's global economy, in part, due to new technologies (TVs, VCRs, computers, international newspapers) that allow rapid dissemination (called one-to-many transmission) of semes. This study suggests a more limited role for cultural diffusion.

As most anthropologists realize semes usually do not evolve as discrete units; they often evolve as part of a "culture complex" or "culture core". We feel that the models, mechanisms and methods described in this paper can help evaluate proposed culture cores or complexes proposed for Africa (e.g., Vansina 1990; Goody 1976), but we feel this would take considerably more analysis and better addressed in a separate paper.

Archaeologists are known for their extensive discussions and distinctions between functional versus stylistic features of artifacts. The implication is that functional features of artifacts are adaptive whereas stylistic features are neutral and better utilized to evaluate cultural evolution. What is stylistic or functional is not always clear and the methods in this paper may assist archaeologists in distinguishing what is stylistic or functional in a particular region of the world.

In conclusion, we hope this paper provides the theoretical, conceptual and methodological tools for others to test and evaluate the relationships between language, culture and genes in any region or continent of the world. The theory and methods are just one way we can come to better understand the histories of human diasporas.

Acknowledgements

We thank Luca Cavalli Sforza, Marc Feldman, Richard Pocklington, Pete Richerson, Monique Bogerhoff Mulder and anonymous reviewers for their very useful comments on earlier drafts of the manuscript. The research was supported, in part, by grants from the Consiglio Nazionale delle Ricerche (CNR) to the Istituto di Genetica Biochimica ed Evoluzionistica (IGBE) in Pavia, Italy and by Italian MPI funds.

References Cited

Boaz, F., ed. 1938. General Anthropology. Boston: DC Heath and Co.

Boyd, R. and P.J. Richerson. 1985. Culture and the Evolutionary Process. Chicago: The University of Chicago Press

Burton, M.L. Moore, C.C., Whiting, J.W.M., and A.K. Romney. 1996. Regions based on social structure. Current Anthropology 37:87-123.

Cavalli-Sforza, L.L. and M.L. Feldman. 1981. Cultural Transmission and Evolution. Princeton, NJ: Princeton University Press.

Cavalli-Sforza, L.L. Menozzi, P. and A. Piazza. 1993. Demic expansions and human evolution. Science 259: 639-646.

Cavalli-Sforza, L.L., Menozzi, P. and A. Piazza. 1994. The History and Geography of Human Genes. Princeton, NJ: Princeton UP.

Cavalli-Sforza, L.L. and M.W. Feldman. 1973. Models for cultural inheritance. Theoretical Population Biology 4: 42-55.

Cavalli-Sforza, L.L. and M.W. Feldman. 1976. Cultural and biological evolutionary processes, selection for a trait under complex transmission. Theoretical Population Biology 4: 42-55.

Chen, H., Sokal, R.R. and M. Ruhlen. 1995. Worldwide analysis of genetic and linguistic relationships of human populations. Human Biology 67:595-612.

Dawkins, R. 1976. The Selfish Gene. NY: Oxford.

Driver, H. E. and A.L. Kroeber. 1932. Quantitative expression of cultural relationships. University of California Publications in American Archaeology and Ethnology 31:211-256.

Durham, W. 1991. Coevolution. Stanford, CA: Stanford University Press

Goody, J. 1976. Production and Reproduction. Cambridge, UK: Cambridge UP

Gray, J.P. 1999. A Corrected Ethnographic Atlas. World Cultures 10:24-136.

Grimes, Barbara (ed). 1978. Ethnologue. Huntington Beach, CA

Gugliemino, C.R., Viganotti, C., Hewlett, B. and L.L. Cavalli Sforza. 1995. Cultural variation in Africa: Role of mechanisms of transmission and adaptation. Proceedings of the National Academy of Sciences, USA 92:7585-7589.

Henrich, J. and R. Boyd. 1998. The evolution of conformist transmission and the emergence of between-group differences. Evolution and Human Behavior 19:215-242.

Hewlett, B. S. and M.E. Lamb. In press. Integrating Evolution, Culture and Developmental Psychology: Explaining Caregiver-Infant Proximity and Responsiveness in Central Africa and the United States of America. In Between Biology and Culture: Prespectives on Ontogenetic Development, H. Keller, Y. Poortinga and A. Schölmerich, Eds. Cambridge University Press.

Ingman, M., Kaessmann, H., Paabo, S. and U. Gyllensten. 2000. Mitochondrial genome variation and the origin of modern humans. Nature 408:708-712.

Jorgensen, J.G. 1969. Salish Language and Culture: A Statistical Analysis of Internal Relationships, History and Evolution. Bloomington: Indiana University Press

Jones, D. 1999. Culture areas as a product of ancient demic expansions. Paper delivered at annual meeting of the Human Behavior and Evolution Society, Salt Lake City.

Kroeber, A.L. 1923. Anthropology. NY: Harcourt, Brace and Co.

Lyman, R.L., O'Brien, M.J. and R.C. Dunnell. 1997. The Rise and Fall of Culture History. NY: Pleunum Press

Murdock, G.P. 1967. Ethnographic Atlas. University of Pittsburgh Press.

Nei, M. 1972. Genetic Ruhlen, Merritt. 1991. A Guide to the World's Languages. Stanford, CA: Stanford UP.

Pocklington, R. 1996. Population Genetics and Cultural History. MA thesis, Simon Frazer University.

Sinnott, R.W. 1984. Virtues of the Haversine. Sky and Telescope 68: 159.

Steward, Julian. 1955. Theory of Culture Change. Urbana, IL: University of Illinois Press.

Trigger, R.G. 1968. Beyond History: The Methods of Prehistory. NY: Holt, Rinehart and Winston.

Tomasello, M. 1999. The human adaptation for culture. Annual Reviews of Anthropology 28: 509-529.

Underhill, P.A., Peidon, S., Lin, A.A., Jin, L., Passarino, G., Yang, W.H., Kauffman, E., Bonné-Tamir, B., Bertranpetit, J., Francalacci, P., Ibrahim, M., Jenkins, R., Kidd, J.R., Mehdi, S.Q., Seielstad, M.T., Wells, R.S., Piazza, A., Davis, R.W., Feldman, M.W., Cavalli-Sforza, L.L. and P.J. Oefner. 2000. Y Chromosome sequence variation and the history of human populations. Nature Genetics 26: 358-361.

Vansina, J. 1990. Paths in the Rainforests. Madison, Wisconsin: University of Wisconsin Press

Table 1. Relationships between the three explanatory models and mechanisms of cultural transmission

and innovation.

Table 3. General means of the four distance measures and the clustering index.

Table 4a. Genetic, geographic and cultural distances for cultures that share the same language phylum.

Table 4b. Genetic, geographic, cultural and linguistic distances for cultures that share similar natural environments.

Table 4c. Genetic, geographic, cultural and linguistic distance measures for three modes of production.