Documenta Praehistorica XXXIX (2012) Social complexity and inequality in the Late Neolithic of the Central Balkans: reviewing the evidence Marko Porčic Department of Archaeology, Faculty of Philosophy, University of Belgrade, RS mporcic@f.bg.ac.rs ABSTRACT - The aim of this paper is to review and critically evaluate relevant archaeological evidence regarding recent claims about the social complexity of Late Neolithic societies in the Central Balkans. Theory suggests that the relevant evidence should be related to population size, economic intensification, ranking, and craft specialisation. It is concluded that, although there are indications that inequalities existed and also demographic potential for organisational complexity, there is no unambiguous evidence that institutionalised inequality in the form of complex polities such as chiefdoms or states ever developed. IZVLEČEK - Namen tega članka je pregled in kritična ocena ključnih arheoloških dokazov za nedavno postavljene trditve o družbeni kompleksnosti pozno-neolitskih družb na področju centralnega Balkana. Teoretske diskusije kažejo, da je te dokaze potrebno iskati v velikosti populacij, povečanju proizvodnje, pojavu razredne družbe in obrtni specializaciji. Sklepamo, da kljub kazalcem obstoja družbene neenakosti in demografskega potenciala za organizacijsko kompleksnost ni nedvoumnih dokazov o tem, da se je v tem obdobju kdaj razvila institucionalizirana neenakost v obliki kompleksnih ureditev, kot so poglavarstva ali države. KEY WORDS - social complexity; social evolution; Late Neolithic; Central Balkans; Vinča culture Introduction It is commonly maintained that Late Neolithic and Early Copper Age societies in Europe did not evolve into archaic states of the kind found in Mesopotamia and Egypt. Recently, the question of scale and complexity of Late Neolithic communities in the Central Balkans has been re-opened by several hypotheses about the organisational properties of Late Neolithic (LN) societies in this region: 1) Late Neolithic communities were hierarchically organised (Crnobr-nja 2011; Crnobrnja et al. 2009; Müller 2012) 2) Late Neolithic settlements had urban layouts, with houses organised in regular rows (Crnobrnja 2011; Crnobrnja et al. 2009; Tasic 2008), and 3) specialised pottery production might have been present (Vukovic 2011). These hypotheses generally imply that LN societies were more complex than previously thought and closer in character to the pre-state or even early state societies of the Near East. In previous studies of the social evolution of LN societies in the Central Balkans a trend of social and economic intensification has been detected (Bankoff, Greenfield 1984; Chapman 1981; 1990; Kaiser, Voytek 1983; Tring-ham 1992; Tringham, Krstic 1990), but no one has claimed that LN communities were anything more than relatively egalitarian kinship-based agricultural societies living in villages. From this perspective, the recent claims might indeed imply a different picture of the Late Neolithic societies in the Central Balkans and take us back to the problem of socio-cultural evolution in the European Neolithic. In order to investigate this topic in more detail, the primary aims of this paper are: 1) to establish the relevant conceptual framework and relevant dimensions, based on anthropological and archaeological theories of social complexity; 2) to evaluate LN so- cieties in the Central Balkans on each relevant dimension. With this approach, we should be able to study the issues of socio-cultural evolution more systematically and understand social structure and social change in the LN of the Central Balkans. Understanding the social development of LN societies in Balkans is not merely a matter of 'local' interest. The problem of diverging social trajectories in Europe and the Near East is not a new problem; it goes back to the roots of European social archaeology and to the great architect of European prehistory, Gordon Childe (Sherratt 1989). Childe was interested in explaining the reasons for this divergence, both empirically and theoretically, by exploring the social development of European and Near Eastern societies (Chapman 2009; Childe 1958), and formulating the general mechanisms and archaeological indicators of social evolution towards complex forms of societies such as early states (Childe 1950; Smith 2009). The issue of evolution towards complexity has been further elaborated within anthropology and archaeology (e.g., Blanton et al. 1996; Carneiro 1970; 1986; 2000; Chapman 2003; Feinman, Marcus 1998; Fried 1967; Johnson, Earle 2000; Kosse 1990; 1994; Marcus 2008; Murdock, Provost 1973; Peregrine et al. 2004; Price, Fein-man 1995b; Price, Feinman 2010a; Service 1971; 1975; Yoffee 2005). Therefore, the investigation of this problem in the central Balkans has significant theoretical implications for the comparative investigation of the principles and mechanisms of the social evolution of Neolithic societies in general. Social complexity - conceptual issues The concept of complexity has a long history in anthropological thought. The social-evolutionary theory of the 19th century was based on the assumption that cultures changed from simple to complex, with the most complex end of this dimension being equated with Western Civilization (Spencer 1997; Trigger 1998). Lewis Morgan's (1877) developmental sequence of savagery, barbarism and civilization is a classic example of how complexity was conceptualised in the 19th century. In this context, cultural complexity was equated with progress; it was a value-laden, ethnocentric concept. Neo-evolutionary theory equates an increase in complexity with general evolution as defined by Marshall D. Sahlins and Elman R. Service (1960). The general evolution is seen as a sequence of major evolutionary stages: bands, tribes, chiefdoms and states. In the neo-evolutionary scheme, increase in complexity is mainly seen as an increase in social scale and the degree of hierarchy in society. Attempts have been made to measure complexity by constructing instruments such as socio-cultural complexity scales. Raoul Naroll developed an index of social development, a measure of social evolution and urbanisation, based on an index theory (Naroll 1956). His approach was to use specific indicators which have "low correlations with each other, but high correlations with the phenomenon being measured" (Naroll 1956.691). He used two indicators of social evolution (craft specialisation and organisational ramification) and one indicator of urbanisation (settlement size) to construct a social development index. Robert Carneiro conceptualised complexity as a Gut-tman scale - an additive scale of cultural traits which could be present or absent (Carneiro 1962). Traits were ranked according to their inclusiveness with respect to other traits. A trait which implied the presence (statistically) of several others was ranked higher. If culture A possessed more traits than culture B, it was considered more complex than B. In this way, cultures could be ranked according to complexity. Although Carneiro was criticised for being selective in his choice of cultural traits (Chick 1997), the general validity of his approach was indirectly demonstrated on a diachronic sample which showed that cultural complexity measured as a Guttman scale does increase through time as expected (Peregrine et al. 2004). Perhaps the most famous scale of cultural complexity was devised by George P. Murdock. Murdock basically followed Naroll's approach and defined complexity as a composite score of several correlated dimensions: Writing and records, Fixity of residence, Agriculture, Urbanisation, Technological specialisation, Land transport, Money, Density of population, Level of political integration, and Social stratification (Murdock, Provost 1973). Garry Chick (1997) criticised all these approaches on three grounds: the main problem was that none offered a rigorous definition of the complexity construct - what it is that this construct is supposed to measure. In Murdock's defence, Trevor Denton (2004) replied that the cultural complexity scale was never intended to measure a single construct, but to develop a set of criteria to distinguish between stages of development. The second problem with the complexity scale, noted by both Chick and Denton, was that there was no specified ontology which specified a relationship between other constructs that "cause and are caused by cultural complexity" (Denton 2004.5). In other words, there was no theory of the causal path and mechanism that linked the indicators and constructs of cultural complexity. The third criticism was based on the empirical results of the Principal Components Analysis (PCA) of data from the Standard Cross-Cultural Sample (SCSS) (Murdock, White 1969), with Murdock's complexity items as input variables (Chick 1997). Namely, Chick performed a PCA analysis with Varimax rotation on SCSS data and extracted two factors (Chick 1997. 294). He interpreted this result as evidence that the items chosen by Murdock do not reflect a single construct, but two independent constructs. Denton accepted the results, but maintained that this was an expected outcome if items were conceived as individual interrelated constructs rather than indicators of a single construct (Denton 2004). Denton is right in claiming that PCA results can be interpreted in many ways, but it should be noted that the component loadings shown in Chick's paper (Chick 1997. Tab. 2) do not pertain to the original principal components, but to rotated principal components. It is also puzzling that Chick presented the proportion of variance accounted for by the first two unrotated components and then went on to interpret the rotated solution. What is more, original unrotated component loadings were never shown in Chick's paper. For this reason, I performed a PCA without rotation on the same SCCS data. Percentages of variance accounted for by the first two unrotated principal components are identical to those reported by Chick (Tab. 1). However, a different picture from that Principal Eigenvalue % of Cumulative Component Variance % 1 5.282 52.817 52.817 2 1.452 14.517 67.334 3 0.683 6.827 74.161 4 0.607 6.072 80.232 5 0.538 5.377 85.609 6 0.425 4.248 89.858 7 0.311 3.106 92.964 8 0.301 3.008 95 972 9 0.262 2.620 98 592 10 0.141 1.408 100.000 Tab. 1. PCA of the SCCS data; eigenvalues and explained variance. presented by Chick emerges if one looks at the un-rotated loading matrix (Tab. 2). It is apparent that all items load positively with relatively high values on the first component, which accounts for the proportion of variance being four times greater than the second. This seems to be consistent with a single construct interpretation, especially given the relatively high Cronbach's alpha value of 0.898. So which interpretation is correct, the one-dimensional or two-dimensional? The results of the PCA without rotation and reliability analysis based on Cronbach's alpha are consistent with the one-dimensional interpretation of cultural complexity as a single construct, but the results of the PCA with rotation are consistent with the two-dimensional interpretation. It should be noted that the application of the rotation algorithm will always result in some solution with more balanced amounts of variance accounted for by rotated factors. From this perspective, it can be claimed that Chick is incorrect in claiming that Murdock's items do not reflect a one-dimensional construct. This problem is analogous to the problem of interpreting the intelligence (IQ) construct: is there a single construct of general intelligence or are there several kinds of intelligence (for an excellent popular account of this problem, see Gould 1996). Murdock's complexity scale does measure something, and does so reliably, but the real issue is whether what it measures has a meaningful theoretical interpretation. In this paper, I will align with Denton's (2004) position that complexity is not a single construct, but a set of several constructs measured by Murdock's items. We can certainly use a single scale as a summary for all other items for some purposes (e.g., demonstrating the reality of an increase in complexity through time, or ranking societies according to their scale), but for Component 1 2 Political integration 0.808 0.123 Social stratification °796 0.209 Density of population 0.777 -0395 Technological specialisation 0.744 0.104 Agriculture 0.743 -0.462 Writing and records 0.713 0.481 Urbanisation 0.704 -0.074 Money 0.693 0.113 Fixity of residence 0.686 -0.613 Land transport 0.576 0.622 Tab. 2. PCA of the SCCS data; loading matrix for the first two components. the purposes of an archaeological investigation of social structure and social change, it is more useful to think of complexity as a multidimensional concept and to evaluate the relevant dimensions independently. This is essentially in accordance with what archaeologists interested in studying social evolution have suggested - the study of past societies on a set of relevant dimensions concurrently (Drennan et al. 2010; Feinman, Neitzel 1984). This kind of approach is best summarised by Robert D. Drennan and colleagues (Drennan et al. 2010.72): "We see no way to avoid recognizing and dealing simultaneously with many dimensions of variability, even though it is conceptually much more difficult than reducing complexity to a few dichotomies or typologies. One way to integrate the information encapsulated in numerous scales of variability is to think in terms of the correlations between dimensions. If high values on some dimensions correspond to consistently high (or low) values on others, these dimensions form 'packages', and knowing that such packages exist gives us patterns to try to make sense of by building theoretical constructs to accountfor them. To the extent that existing theory implies the existence of such packages, they can be sought in the archaeological evidence as a means to evaluate the theoretical models. The former (bottom-up) way of working by no means contradicts the latter (top-down) approach. They are complementary. Both bring our knowledge of what happened in prehistory together with the theoretical notions that help us understand how it came to happen that way; they come together in the act of empirical evaluation of theoretical models." Starting from this framework, the next logical step is to define the relevant dimensions. If the aim is to explore the degree of organisational complexity of LN societies, it makes sense to use dimensions, which correlate with organisational complexity and can be detected archaeologi-cally. A set of 5 dimensions will be considered in this study: demography, social inequality, economic intensification, craft specialisation, and political strategy (the exclusionary-corporate axis as defined by Blanton et al. 1996). Archaeological background The major archaeological phenomenon of the Late Neolithic in the Central Balkans is the Vinča culture, which extended across the region over an area of around 300km2 (Fig. 1), encompassing Central Serbia, Kosovo, southern parts of Vojvodina, Transylvania, Oltenia, eastern parts of Bosnia and northern parts of Macedonia (Brukner 2003; Chapman 1981; Garašanin 1973; 1979; 1982). In calendar years, Vinča culture appears at 5400/5300 BC and continued until 4650/4600 BC (Boric 2009). In general, Vinča settlements subsisted on a mixed economy typical of the temperate European climate (see Barker 1985; Bogaard 2004): cultivation based on cereals (Borojevič 2006; Bottema, Ottaway 1982; van Zeist 2002) and animal husbandry dominated by domestic animals such as cattle, pig, sheep and goat, accompanied by wild species such as red deer, roe deer and wild pig (Blažič 2011; Bökönyi 1988; Di-mitrijevič 2008; Greenfield 1986; Legge 1990; Or-ton 2008; Russell 1993). Sites from this large geographical area shared a similar material culture (e.g., wattle and daub houses, characteristic black pottery and clay figurines), al- Fig. 1. Approximate distribution of Vinča culture. though caution should be taken not to conflate this archaeologically defined entity with anthropological and social entities (political, linguistic or ethnic). Therefore, the term Vinca culture should be understood as a technical label. Reviewing the evidence Demography It has been shown in a number of cross-cultural studies that the organisational complexity (often measured by the number of types of political officials, scale of political integration above the local community, or degree of hierarchy and centralisation) of a society is positively correlated with demographic variables such as population size or the population size of the largest community (Carneiro 1986; 2000; Ember 1963; Feinman 2011; Feinman, Neitzel 1984; Johnson 1982). Gregory Johnson (1982) offered an explanation for the observed empirical pattern: the ability of a group of people to make decisions by consensus or equal participation is limited by group size. Johnson presented the results of small group studies that suggest that the critical group size is six persons. Above this limit, the group is faced with a scalar stress: humans are unable to process information resulting from the interaction of all group members, which leads to inefficiency in the decision-making process. The problem has two potential solutions: sequential hierarchies - the creation of larger basal groups and more levels of decision making (e.g., creating larger households or groups of households which can make decisions by consensus within the group and then send a delegate to negotiate decisions at village level); secondly, simultaneous hierarchies -the centralisation of the decision-making process, which results in the rise of managing elites. It should be noted that the number of levels in a sequential hierarchy can be increased only up to a certain point, whereupon the system becomes ineffective because of the long lines of communication between the lowest and highest levels. From this point, simultaneous hierarchy is to be expected. Efforts have been made to determine the demographic threshold beyond which the organisational change towards complexity (usually toward simultaneous hierarchy) is inevitable. As apparent from Gary M. Feinman's (2011.41.Tab. 3) review, most researchers agree that the critical population size is between 2000 and 3000 people. So in the context of this research, we want to know whether the LN com- munities of the Central Balkans ever reached or crossed this threshold. This is a very conservative test, however, because if the settlement population crossed the threshold, we can be almost certain that hierarchy must have been present. If not, then we have only the absence of evidence, not the evidence of absence, since a community might have consisted of more than a single settlement at the regional level. To answer this question, population size estimates were made for several LN sites in the Central Balkans with relatively good settlement data. Three estimates are used: © A maximum population size estimate, based on the assumption that an entire area of the settlement was in contemporaneous use. This is obviously a problematic assumption in most cases (Porčic 2011). The estimate is derived by multiplying the (estimated) total number of houses by the estimated average household size. The total number of houses is estimated by the simple proportional extrapolation of the number of excavated or surveyed houses to the unexcavated area. For example, if 50% of a site area was detected by magnetometer and 40 houses were detected, the estimated total number of houses for the entire site is 80. The average household size is estimated by dividing the average house floor area by the conversion constant of 7m2/person (Porčic 2012a). © An average population size estimate based on the premise that population size was more or less constant during a settlement's history and that only parts of the settlement were used simultaneously. This estimate is calculated by solving Schiffer's (1976; 1987) discard equation, which describes the accumulation of houses in the archaeological record for the systemic number of houses, assuming that the average use-life of houses was 40 years. This estimate was calculated only for sites where the duration of the settlement is known from 14C dates. © Final population size estimates based on the logistic model of population growth (Porčic 2011). The results are shown in Table 3. Population size estimates suggest that the scalar stress threshold was unambiguously crossed only in the case of Di-vostin and Stubline and only if the least realistic estimation procedure (which assumes that all houses are contemporaneous) is used. It should also be noted that the projection for Divostin is based on 1.17% of the total site area, which makes it highly unreli- able. Therefore, there is no firm evidence that settlement communities crossed the scalar stress threshold unless the majority of houses at Divostin and Stubline were in contemporary use. Social inequality - vertical status differentiation The importance of determining the presence, degree and type of social inequality in a prehistoric society for the study of its social structure and evolution cannot be underestimated. This is best expressed by Douglas T. Price and Gary M. Feinman (2010b.2): "Social inequality, the organizing principle of hierarchical structure in human society, is manifested in unequal access to goods, information, decision making, and power. Status is the determinant of social position, and status differentiation is the foundation of inequality. A variety of human conditions are used in ordering social hierarchies and in determining status and access. These include age, gender, birth order, class, race, and a number of others. Social inequality is a characteristic of virtually every society on earth today and its history goes back thousands of years. This structure of unequal relations, of status differentiation, is essential to higher orders of social organization and is basic to the operation of more complex societies." Norman Yoffee (2005.35) also considers inequalities in economic production and exchange as a necessary precondition for the development of complex society forms such as states. It should be emphasised that inequalities of various kinds are always present in every human society (Flanagan 1989); they may often arise as a result of chance (Mayhew, Levinger 1976; Mayhew, Schol-laert 1980; Pauketat 1996). So, as Price and Fein-man note (1995a.4), the real question is not whether inequality was present, but what was the degree and Site Population size estimate Maximum Average Final Divostin IIb 8200 1061 868-2684 Gomolava 460 53 70-258 Mali Borak 115 Stubline 4000 Uivar 420 44 90-325 Parta 7b 823 329 Parta 7C-6 1575 630 Tab. 3. Population size estimates. nature of the inequality - e.g., was it institutionalised. This is a traditional issue in social archaeology - to determine the degree of social inequality within a given society. This kind of analysis is usually undertaken with two classes of data - household data and funerary data (Wason 1994). The general aim is to determine the magnitude and pattern of inter-household or inter-individual variation in indicators of social status. In household archaeology, these indicators are house size or house area and the attributes of house inventory. In funerary archaeology, the indicators are the attributes of grave construction and grave goods. The Gini coefficient is often used in anthropology and economics as a formal measure of inequality in the distribution of a currency (Dorfman 1979; Gast-wirth 1972; Smith et al. 2010). The Gini value of 0 indicates that the currency is perfectly equally distributed, while values close to 1 (the highest value that Gini can attain) indicate that the largest share of the currency is held by a minority of the population. Gini coefficients are calculated for two kinds of 'currency': 1) distribution of house floor areas 2) diversity of grave good materials from the only well-recorded LN necropolis in the Central Balkans at Gomolava Ib (Boric 1996; Brukner 1980). The results for household data suggest that differences in house sizes between households were not great (Tab. 4; Fig. 2). However, at Divostin IIb, the correlation between pottery assemblage size1 and house floor area is relatively high and close to being marginally significant at the 0.05 level (r = 0.712, one-tailed p = 0.053, see Fig. 3). More importantly, copper or malachite artefacts are predominantly found in the group of large houses - a single copper bracelet was found in house 14 (McPherron, Srejo-vic 1988). This pattern may be interpreted as indicative of incipient ranking (for a detailed discussion see Porčic 2012b): larger households have a larger labour force available to intensify production and create surpluses which would enable some households to gain an advantage over others. The association of copper items with large households may indicate their higher status. The value of the Gini coefficient for the diversity of grave good materials from Gomolava House data source (McPherron, Srejovič 1988) (Brukner 1980) (Marič 2011) (Crnobrnja et al. 2009) (Schier 2008) (Lazarovici et al. 2001) (Lazarovici et al. 2001) 1 Pot counts for Divostin houses from Porčic (forthcoming); house floor areas from Tripkovic (2009b). Site Gini coefficient Data source Divostin IIa 0.2 McPherron, Srejovic 1988 Divostin IIb °.i5 McPherron, Srejovic 1988 Gomolava Ib 0.14 Brukner 1980 Mali Borak 0.11 Maric 2011 Parta 7a 0.14 Lazarovici et al. 2001 Parta 7b 0.28 Lazarovici et al. 2001 Parta 7c 0.36 Lazarovici et al. 2001 Stubline 0.24 Crnobrnja et al. 2009 Uivar 0.24 Schier 2008 Tab. 4. Gini coefficients for house areas from LN sites in Central Balkans. cemetery is 0.443. An important fact should be mentioned about the Gomolava graves: DNA analysis showed that these were all male individuals of the same patrilineage (Stefanovic 2008). The duration of the cemetery was estimated to be approximately 50 years, or two generations (BoriC2009.227). This is important information, because it may suggest that, regardless of the inequality or equality of the distribution of grave goods, only one lineage might have had access to the burial area within the settlement. Economic intensification Economic intensification is also an important aspect and precondition of complexity. One reason for this is ecological - more food and energy is needed to sustain larger populations. The other reason is social -the elaboration of society and culture requires in- o o ro Ö Lf>