Abstract
Slavery is difficult to ascertain in the archaeological record, especially because of the lack of material evidence. Using the Standard Cross-Cultural Sample of 186 societies, our aim was to find indirect and easily identifiable indicators of the presence of slavery. The results show links between slavery and the expected and familiar domains (e.g., warfare, polygyny, social and political integration) as well as its relationship to metallurgy, which can be considered an innovative finding. This text attempts to explain and give context to the metallurgy relationship with historical examples related to the exploitation of slaves during various stages of the operational chain of metal production. These include raw material extraction, production of charcoal, and construction or reconstruction of smelting furnaces.
Introduction
The institution of slavery has existed for thousands of years. Up to the present day, there is no region in the world (with the exception of Antarctica) where slavery has not occurred at some period of its history (Goody, 1980; Patterson, 1982). Slavery clearly constituted a feature of industrial societies and the archaic world, including hunter-gatherer communities. While in the case of historical and ethnographic societies the recognition of slavery is facilitated by written sources first and eyewitness testimony second, in the context of the preliterate past, the problem is generally not only historical; for disciplines such as anthropology and archaeology, it is also methodological.
However, the presence of slaves in a society may be inferred not only from direct written or material evidence. As shown in a series of intercultural studies (see below), slavery, like other social institutions, does not exist as an isolated cultural phenomenon but is also associated with other manifestations of human society, many of which can already be shown archaeologically. Thus, it is possible to use cross-cultural research (C. R. Ember & Ember, 2009), which, in terms of the study of archaic societies, produces interesting results (for an overview, see M. Ember & Ember, 1995; Peregrine, 2001, 2004).
The aim of this article is to attempt to find correlations between slavery and types of culturally specific domains that are more easily assessed via archaeology than slavery itself and could thereby serve as its archaeological indicators. We will attempt to explore the possibilities of detecting slavery in connection with particular cultural variables such as frequency of warfare, class stratification, level of political integration and perceptions of political leaders’ power, polygyny, bride-price occurrence, contact with significantly different societies, type of subsistence or ecological classification, and specific craft specialization such as metalworking. Rather than testing specific hypotheses, this article aims to explore the above-mentioned variables more broadly. Although the relationship between these domains and slavery has been previously studied, a detailed analysis may yield new results that can be interpretable in a novel and interesting way. Apart from the previously detected correlates of slavery, such as an increased level of warfare (Patterson, 2012), this text will reveal a similar close relationship for metallurgy. As we will show, the correlation between slavery and metallurgy becomes clearer when metallurgy was associated with a directly related activity—mining. As historical and ethnographic sources show, slaves were frequently employed for ore mining and metal processing. It may be inferred that aside from agriculture, mines and metallurgic areas specifically exploited enslaved workers en masse. This conclusion would represent an important clue in archaeology for detecting slavery in the past, because both raw material extraction and metallurgic activities offer useful evidence in the material remains.
A Brief History of the Cross-Cultural Research on Slavery
It is advisable to start with a definition of slavery. Despite the fact that the concept of slavery has been intensively studied by many researchers from around the world, no simple definition exists (see, for example, Allain, 2012). Traditionally, slavery is defined as a condition whereby one person is owned by another person. 1 This definition should normatively define a difference from other forms of absence of freedom such as bondage and serfdom, although, in practice, the boundaries are less sharply defined. Other typical features include the fact that the slave, legally and often socially, does not have a family and is therefore considered an “outsider” or a “socially dead person” (Patterson, 1982).
The institution of slavery not only adopted different forms in different societies but also differed internally within one society (e.g., in ancient Rome, slaves who worked in a mine had a different status than their more educated counterparts who served in the households of wealthy citizens). In the text below, we will adhere to a broader definition, which identifies a slave as an unfree person who serves his or her master.
The first cross-cultural analysis of slavery was implemented more than a hundred years ago by Herman J. Nieboer (1900). He considered the main cause of its formation to be the open/closed aspect of natural resources and other related economic problems. Evsey Domar developed a similar hypothesis that was published 70 years later than Nieboer’s. Looking at data from historical sources about Russian serfdom in the 16th and 17th centuries, he found evidence to support the idea that a high ratio of land–labor predicts forced labor, either in the form of serfdom or slavery. The described theory, later named after its creators as the Nieboer–Domar hypothesis, was soon criticized (Engerman, 1973; Patterson, 1977a, 1977b; Pryor, 1977).
Leonard Hobhouse and his colleagues followed up on Nieboer’s work. In their evolutionary classification of societies, they demonstrated that slavery existed from lower level hunting-gathering societies to societies with fully developed agriculture and is associated with the increasing economic complexity (Hobhouse, Wheeler, & Ginsberg, 1915). George Peter Murdock (1949), the principal proponent of cross-cultural research, also touched marginally on the subject of slavery in his work and documented, as Christiaan Baks also did with his colleagues a few years later (Baks, Breman, & Nooij, 1966), that the presence of slavery, in addition to representing an economic level, is also strongly linked to the existence of social stratification. Economist Frederic Pryor (1977) pointed out that the specific factors that determine the existence of slavery cannot be studied without defining the specific type of unfree slaves. He distinguished slaves representing economic or social capital. Slaves are considered economic capital when they constitute a significant part of the workforce in the society, which, while requiring them to ensure its economy, also strictly segregates them as outsiders. Slaves become social capital if they constitute part of the society that owns them, although they lack certain social and political rights. Typical predictors for slavery as social capital are, in terms of cultural specificities, polygyny along with a stage of economic development.
Orlando Patterson (1977b) specifically suggested the importance of surplus as the decisive factor for the presence of slavery. The same author later emphasized the gender aspect of slavery along with the centrality of the body with regard to slave–master relations (Patterson, 2012). Jack Goody (1980) sought to identify a correlation between subsistence practices and slavery. His analysis, as expected, confirmed that slavery almost never occurred among hunter-gatherers (only in 3% of societies) but that it grew with the development of a productive economy. In horticulturalist communities, slavery occurred only in 17% of societies, while up to 40% of field farmers benefited from unfree labor. Surprisingly, a high proportion of slave labor (34%) was used in societies that subsisted on fishing, and an extremely high percentage of slavery (73%) was found among pastoralists.
Over the last 30 years, the topic of slavery appears to have been neglected in anthropology. However, this was not the case of economic historians (for an overview, see Fenske, 2013). Of the works oriented in this direction, the model presented by Nils-Peter Lagerlöf (2009) is worth mentioning as it is based on a concept similar to the previously mentioned Nieboer–Domar hypothesis. James Fenske (2013) followed up on this theory by examining whether certain geographic aspects can predict factors such as slavery, population density, or land rights.
If we summarize the results of the research history, we can differentiate between two theoretical directions. The first explains the existence of slavery primarily in terms of economic circumstances (technological development, labor shortage, excess of food), while the other ascribes the effect to a wider range of sociocultural phenomena (i.e., polygyny, social stratification, and the frequency of warfare).
Sample and Data
A sample of 186 of the world’s cultures, also referred to as a Standard Cross-Cultural Sample, was chosen for the analysis presented here (Divale, 2000, 2007; Murdock & White, 1969, 2008). 2 This selection is sufficiently stratified and disproportionate to provide a balanced representation of all the cultural regions of the world, thereby minimizing Galton’s problem (C. R. Ember & Ember, 2009). Although the individual cultures were chosen not randomly but judgmentally, J. P. Gray (1996) showed that this sample is likely to show only a minimal degree of autocorrelation distortion.
The data for the presence of slavery are drawn from the study by Orlando Patterson (1982, Appendix B). 3 The definitions and sources of all other variables are provided in the appendix and below in the relevant tables or figures.
For comparison purposes, the analysis was also conducted on a sample of 1,267 societies of the Ethnographic Atlas, as compiled by J. P. Gray (1998). 4 His codebook is also the source of the data provided concerning metalworking and slavery (see the appendix).
Of the 186 societies of the Standard Cross-Cultural Sample, Patterson determined 71 (38%) as being slavery present. For the sake of comparison, out of the 1,267 societies that are included in Murdock’s (1967) Ethnographic Atlas summary, at least some form of slavery was determined to be present in 573 (45%) of the societies (Figure 1), with slavery being rare or absent in 524 (41%) of the societies, while in 170 (14%) of the cases, sufficient information was missing (Gray, 1998).
The geographical location of the slave societies that comprise the Ethnographic Atlas (Murdock, 1967).
At this point, it should be noted that slavery does not exist in one single form but can be divided into several categories—For example, hereditary/nonhereditary, small scale/large scale, domestic/chattel, or the slave can simply represent social/economic capital (Kopytoff, 1982; Patterson, 1982; Pryor, 1977). For the purposes of this study, where our goal is to identify any form of slavery, we shall divide the societies into two groups: (a) slavery present and (b) slavery absent.
As stated in the “Introduction” section, the aim of the article is to find a correlation between slavery and variables that have the potential to be archaeologically detectable. Therefore, 22 cultural variables were gradually selected from the large number of variables contained in the Standard Cross-Cultural Sample (see the appendix). The archaeological visibility of some variables is distinct (e.g., evidence of cross-cultural contact, warfare, types of settlement and subsistence, metalworking); in others, material evidence is more difficult to find but is not absent (e.g., polygyny and bride-price).
Results of the Analyses
The analyses showed that there is no specific social feature that constitutes a prerequisite condition for the presence of slavery. This also applies, for example, to contact with another society or endemic warfare. Although there are rarely slaves in culturally isolated societies, even among these societies, slavery can be found. In such cases, slavery may arise, for example, under the influence of intercommunity conflicts that occur within one society (or anthropological culture). The actual presence of frequent warfare is one of the most important indicators of slavery, whose incidence is significantly higher in societies in which a state of war represents a normal part of life. Under these conditions, the slaves became prisoners of war; however, it cannot be concluded that the absence or dearth of warfare points to the absence of slaves (Table 1).
Mutual Correlation of Selected Variables Expressed by Coefficient Kendall’s Tau-b.
p < .05. **p < .01 (two-tailed significance). n = sample size. Correlations between slavery and other variables are in bold.
The correlation between slavery and the “permeability of communities,” which is defined as the number of different obstacles that need to be overcome before entering the innermost part of the settlement (e.g., fences, ditches or doors; Peregrine, 1993, 1994), is interesting from an archaeological point of view. It was found that slavery is infrequent in societies that do not intensively fence in their settlements and that live in simple dwellings (Code Values 1 and 2); however, slavery is frequently found where, by default, the area of the settlement is defined by a barrier and the houses are solid (Code Values 3 and 4; Figure 2). These results, however, are applicable only to nonpastoral societies because pastoralists usually use different strategies to prevent access to and exit from the camp.
Association between slavery and permeability (Cramer’s V = .720; Kendall’s tau-b = .541; p < .01).
Rather than the “permeability of communities,” the presence of horses appears to be an important indicator regarding the difference between nomadic and seminomadic societies (Divale, 2000, Variable 1515). Approximately 64% of societies that are based on impermanent settlements that have horses also have slaves. In other words, this means that nine out of 10 nonsedentary slavery-present societies use horses. A strong relationship between slavery and horses has been documented in areas such as the African Sahel, where cavalry was frequently used for slave raids. In this regard, the increased importing of horses often led to the increased capture of slaves, who were subsequently exchanged for greater horse power, thereby ensuring that the cycle of slavery continued (Haour, 2007).
The existence of social stratification is not a prerequisite of slavery. Approximately a quarter of slave societies are not characterized by the social differences that exist between the free members. Nevertheless, we can still consider social stratification to be one of the most significant predictors of slavery (Table 1).
The political organization of society is also related to stratification. The relationship between this cultural characteristic and slavery was investigated using two data-available parameters: that is, the level of political integration and native perceptions of the leaders’ power (Table 1). Both variables showed a degree of coherence regarding slavery: The level of political organization broadly correlates with the degree of social stratification, and they can be combined into a single explanatory factor. The disadvantage of both these political variables is their difficult archaeological traceability.
Other analyses are focused on the relationship between slavery and means of subsistence. Generally speaking, the results shown in Table 2 are hardly surprising. The presence of slaves among hunter-gatherers is rare. Slavery is likewise rare among horticultural farmers. In contrast, it is documented more often among fishermen, especially those engaged in anadromous fishing. In the case of migratory farmers, the presence of slavery depends significantly on the material advancement of the tools used: Societies that have knowledge of metalworking more often have slaves. For pastoralists and hunters, slavery occurs in about half of the cases. In the case of intensive farmers, together with the presence of plows, the probability of slavery decreases slightly. The relationship between slavery and the means of subsistence in accordance with the type of resource base, as categorized by Karen and Jeffrey Paige (1981), is relatively linear (Figure 3).
Association Between Slavery and Type of Subsistence (Cramer’s V = .485; p < .01).
Note. (1) Gathering; (2) Hunting and/or marine animals; (3) Fishing; (4) Anadromous fishing (spawning fish such as salmon); (5) Mounted hunting; (6) Pastoralism; (7) Shifting cultivation, with digging sticks or wooden hoes; (8) Shifting cultivation, with metal hoes; (9) Horticultural gardens or tree fruits; (10) Advanced horticulture, with metal hoes; (11) Intensive agriculture, without plows; (12) Intensive agriculture, with plows (Divale, 2000, Variable 858). Sample size: n = 186 societies.
Association between slavery and subsistence—Resource base classification (Cramer’s V = .314; Kendall’s tau-b = .280; p < .01).
The innovative potential of our study and the most important result was achieved through an analysis of the relationship between slavery and technological specialization, that is, metalworking. Slavery is significantly more frequently present (60.5%) in societies in which metalworking takes place compared with those in which it is absent (17.2%). A similar relationship is also illustrated by an analysis of the 882 societies that are referred to in the Ethnographic Atlas, where slavery is identified as being present in 73% of societies that are engaged in metalworking and in only 24.1% of societies in which this activity is absent (Figure 4).
Association between slavery and metalworking.
Analyses were also carried out using two other variables, both of which are closely related to metallurgy—mining/quarrying and the smelting of ores. In relation to slavery, both variables show a slightly weaker but still statistically significant (p < .001) relationship. Mutual correlations between the individual variables are shown in Table 3.
Correlations Between Slavery, Mining/Quarrying, Smelting, and Metalworking.
Note. Pearson correlations. Sample size: n = 140 societies.
p < .001 (two-tailed significance).
If the indicated relationship between metallurgy and the presence of slaves is valid (and analyses with control variables suggest that this relationship is unaffected by any of the other cultural features; Table 4), from an archaeological point of view, this would represent a significant correlation because evidence of metallurgy is relatively easy to identify. Therefore, in the following section, we will specifically address the interpretation of these correlations.
Correlation Between Slavery and Metalworking With Selected Control Variables.
Note. n = sample size.
Metalworking is constant.
p < .05. **p < .01. ***p < .001.
Among other correlations of cultural variables, our analyses confirm the links between slavery and polygamy and bride-price (cf. Patterson, 2012), features that are virtually unprovable archaeologically.
From the traditional evolutionary perspective, the presence of slavery should be viewed in relation to the advancement of society, that is, its complexity. In societies with a low level of complexity, there is little use for slaves, and acquiring and keeping slaves is therefore difficult. With the increasing complexity of a society, the ability to have slaves and their acquisition also increases. Although there are many ways of measuring cultural complexity, most are strongly correlated (C. R. Ember & Ember, 2009). For this reason, a simple 40-level scale with 10 summed variables, designed by Murdock and Provost (1973b), was chosen, many of which had already been used with regard to some of the analyses that have been referred to above.
The results revealed a statistically significant correlation, although it was not as strong as might have been expected (Figure 5). The least complex societies do not have slaves, while overall, in societies with a low level of complexity (0-9), slaves are found in 22% of cases. The likelihood of slavery increases with the increase of complexity, but in the most complex societies, once again, the rate fell slightly. Within the scale examined, the presence of slavery is therefore only related to certain indicators of complexity—specifically, technological specialization/metallurgy, social stratification, political integration, and agriculture. Indeed, it appears to have no strong correlation to fixed residency, urbanization, land transport, or the existence of money (Table 5).
Association between slavery and cultural complexity.
Mutual Correlation of Slavery, Metalworking, and Cultural Complexity (10 Individual Variables and Total Index [Variable 12]) According to Murdock and Provost’s (1973b) Measurement of Cultural Complexity.
Note. Kendall’s tau-b coefficient. Sample size: n = 186 societies (except metalworking, where n = 179). Correlations between slavery and other variables are in bold.
p < .05. **p < .01 (two-tailed significance).
Correlations between the variables were identified using multivariate statistics. The principal components analysis (PCA) method was applied. The results show that slavery has a strong correlation with each of the following variables: frequency of warfare, metalworking (and technological specialization), polygyny, bride-price, a resource base (in the sense of a strong correlation to high resources), political integration, and social stratification (Figure 6). The mutual correlations of these variables are also presented in Table 1.
A graphic representation of the result of the PCA expressing the correlation between slavery and certain selected variables.
The effect of individual independent variables on slavery was also examined by the method of logistic regression (C. R. Ember & Ember, 2009). The result is a model with four dichotomous variables, which are related to slavery at a statistically significant level, even when considered together (Table 6): metallurgy, frequent warfare, social stratification, and polygyny.
A Logistic Regression Analysis With Dichotomized and Statistically Significant Variables.
Note. Sample size: n = 155 societies; overall significance of model: p < .001; Cox & Snell’s R2 = .340; Nagelkerke R2 = .458; overall percentage correctly predicted = 76.1%.
Discussion: Slavery, Metallurgy, and Mining
Although, from the correlations examined, the relationship between slavery and metallurgy is among the strongest, this cannot be explained simply by identifying slaves as metallurgists or smiths. Both the ethnographic and historical evidence suggests that metalworkers (which we collectively identify as metallurgists and smiths) in archaic societies constituted a highly valued group; in some cultures, they even had their own caste status (David, Heimann, Killick, & Wayman, 1989). “Blacksmiths in sub-Saharan Africa occupy confusing social spaces, as if they lived in two conflicting dimensions. They are at once glorified and shunned, feared and despised, afforded special privileges and bounded by special interdictions” (Haour, 2013, p. 87). The work of smiths and metallurgists also required such skill—in terms of both its technological and its spiritual aspects—that the use of unskilled slaves is difficult to justify (Chirikure, 2015; van der Merwe & Avery, 1987). So why should metalworking be so closely associated with slavery?
To answer this question, it is appropriate to focus on the entire chaîne opératoire of the production of metals from above and not associate it only with the actual metallurgy processing. Within the concept of data records of a standard intercultural sample, metalworking evidently comprises a larger set of different types of activities, raw materials, and labor (e.g., Hauptmann, 2006). Metallurgic activity itself was conditional on at least three other activities: the extraction of the raw materials, the production of charcoal, and the construction or reconstruction of furnaces (Huysecom, Augustoni, Godel, & Saparelli, 1997). In all these activities, slaves could apply themselves well, but only for the first activity do we have historical and ethnographic evidence of what occurred—not only in specific case studies but regularly and en masse.
In the Late Iron Age in Europe, for the first time, slave chains and cuffs appeared in such numbers that they are generally considered to be one of the direct material indicators of enslaved persons (Parker Pearson, 2005; Taylor, 2001). In this context, it is also possible to consider possible complementary links between iron chains and bars that kept captives and metallurgical and mining work.
The first clearly documented massive use of slaves for mining activities dates back to the 5th-century
The mining practices were described by Diodorus of Sicily (ca. 90-27
The mass deployment of slaves in mines and in metallurgy continued during the ancient Roman period, when residents abducted from barbarian Europe became forced workers (Briggs, 2003; Mattingly, 2010). A number of Roman authors directly confirm that mining was carried out by slaves. For example, we refer to the largest Roman gold mine in Las Médulas, which at that time employed up to 60,000 workers, or the iron ore, copper, and silver mines located in the Rio Tinto area (Lewis & Jones, 1970; Rickard, 1928).
Logically, the forced labor of slaves in the mines and the subsequent processing of the ore also occurred during the post-Roman period. We are primarily informed about the actual development of this practice from the area controlled by the Muslims (Alexander, 2001; Heck, 1999). The use of slaves in early Christian Europe for these purposes is usually not taken into account, although there is one significant exception: the hypothesis of Martin Ježek, who connected the interpretation of the evidence of metallurgical production in Prague during the 10th to 12th centuries with a written report regarding the presence of slaves in this central location of Bohemia. In connection with the mass production of iron and the enormous demand for labor, it is possible that slaves were present to fill the needs of the market as well as for production (Ježek, 2011).
The modern and industrial eras are both characterized by the massive deployment of slaves on plantations in overseas colonies. Even there, however, the subjugated workers were forced to participate in the mining and processing of minerals (Conrad & Meyer, 1958; Guimarães, de Morais, & de Assis Roedel, 2015). In addition, it would be erroneous to believe that the use of slaves in mining ended with the triumph of abolitionism. In European totalitarian regimes of the 20th century, the forced deployment of unfree persons in mines has occurred periodically (Gordon, 2016). Even now, slave work is associated with the mining of minerals, especially in dysfunctional and collapsed states. Typically, this is related to the mining of rare minerals of the coltan type, of which there is a shortage in the global market and for which there is urgent demand (Crane, 2013).
In light of the brief research, a statistically established cultural link between slavery and metallurgy (which, however, should be understood as a wide scope of activities related to mining and the processing of raw materials) appears to be a real and logically interpretable connection.
Conclusion
The aim of the article was to show the possibility of finding evidence of slavery in archaic and prehistoric societies. From written sources, we know that slavery accompanies humans from the beginnings of recorded history in Mesopotamia (Goody, 1980). In addition, written sources have confirmed existence of slaves in ancient Egypt, Israel, Hittite Anatolia, and the Aegean (Castleden, 1993; Eyre, 2010; Snell, 2011). The presence of slaves is inherently embedded in the subsequent classical Greece and Rome periods, as well as the early and late medieval period in Europe. Documenting slavery in less developed preliterate societies, compared with the above-mentioned historical periods, is much more difficult, because there are no types of archaeological findings that could be wholly interpreted as evidence of the existence of slaves. However, as we attempted to show above, some types of culturally specific parameters that are not devoid of physical clues may, to a greater or lesser extent, imply the presence of slavery. Frequent concurrent cultural features of slavery include warfare, developed social stratification, polygynous marriages associated with payment for the bride, a higher degree of political integration, or specific methods of subsistence. Most of these findings were not surprising and, in the same or a modified form, have already been presented before (e.g., Domar, 1970; Fenske, 2013; Goody, 1980; Hobhouse et al., 1915; Murdock, 1949; Nieboer, 1900, Patterson, 1977, 2012; Pryor, 1977).
From the perspective of preliterate societies, the most important clues that may indirectly indicate the presence of strata of excluded and/or unfree persons appear to be the following: first, the intensity of warfare, which is strongly associated with captives and slavery (Parker Pearson, 2005). Therefore, if we consider the archaeological evidence of military raids (e.g., Golitko & Keeley, 2007; Vencl, 1999), it appears that the conditions for the use of captured people in Europe were set as early as the Neolithic period.
The second correlation, significant from an archaeological point of view, is the above-mentioned wide range of activities associated with the production and processing of metals. If the metallurgy documents are numerous, thereby testifying to intense activity, this could strongly suggest the possible presence of an otherwise virtually invisible social stratum of unfree people. This finding can, for example, support earlier considerations of the use of slaves in mines in Bronze Age Europe (Kristiansen, 1998).
Footnotes
Appendix
Acknowledgements
We would like to thank Orlando Patterson for consulting and providing an updated list of slave societies listed in the Standard Cross-Cultural Sample.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.