The 40 years between 1970 and 2010 were a period of increasing archaeological research in Mali, West Africa. In the region around the middle reaches of the river Niger, this research documented the remains of a dynamic civilization which, from the last millennium BC onwards, developed urbanism, craft specialization, and long-distance trade (McIntosh 1998), and was highly influential not only in changing narratives about the African past but also in global debates on sociocultural dynamics (McIntosh 1999). One of the characteristics of many settlement sites of the first-early second millennium AD in the region is the high volume of ceramic sherds, which sometimes challenged the capacities of research projects (McIntosh 1995, pp. 132–133). Featuring a large variety of decorative motifs, forms, and production techniques, the pottery of the region has been the subject of numerous studies, both from an archaeological and ethnoarchaeological perspective. Three key points have emerged from this research history. Firstly, establishing chronologically sensitive typologies has been very difficult, with contemporary variability often more significant than diachronic change. Secondly, base forming techniques, which vary throughout the region, are likely to correspond to larger cultural and maybe also linguistic spheres. Thirdly, pottery vessels appear to have been objects of trade, at times over long distances.
In this paper, we consider decorative practices on 12th century AD ceramics of this region. We use a network approach to ask how such practices might have spread, and, even when they are not useful as chronological markers, how they might still inform us about sociocultural dynamics of the ancient Middle Niger. This research aim reflects a long-standing concern of archaeologists in the area, but the network approach we adopt here shifts the perspective from ceramic production as representing sociocultural entities to the processes by which techniques spread. This results in connections, rather than boundaries, as the subject of research. Using network models, we ask what ceramics can tell us about the processes that integrated the Middle Niger as a cultural area despite its diversity, and where cultural exchange might have been facilitated or impeded, and why.
1.1 Middle Niger Ceramics:
In many contemporary societies in the region, pottery is made by occupational specialist groups (Bambara: nyamakala). Though there are exceptions, it is the cultural norm in the wider Middle Niger region that pottery is made by women of groups in which the men are specialists in metallurgy. There is some debate as to how far in the past this specialization developed, but it currently seems likely that it has its roots in the first millennium AD (McIntosh 1998, pp. 177–178), and that it had spread throughout the region by the mid second millennium (Tamari 1991, 1995). There is no doubt that potting and metallurgical experts were of great significance to the economy of the Middle Niger during this period of urbanism and emerging statehood, since they made basic products used in most parts of everyday life, from food production to the religious sphere. The cultural innovations linked to pottery making, both in social terms (endogamous craft specialists) and in technology (pottery and metal products), are a key factor in the development of the Middle Niger civilization (McIntosh 1998).
From the 1st millennium BC onwards, ceramics occur in high numbers on archaeological sites on the Middle Niger, and their appearance, especially their surface treatment, is highly variable. The number of different roulette decorations, painted, incised, and applied decors, but also the variety in form, is very high, especially when attention is paid to the techniques and tools of decoration, rather than exclusively to their effects. Consequently, there has been considerable emphasis on ceramics in archaeological and in ethnoarchaeological work. Early studies (Bedaux 1980; e.g. Curdy 1982; and particularly Bedaux and Lange 1983) noted both regional differences in the appearance of pottery as well as striking widespread similarities in individual forms or attributes. Subsequent work has seen both in-depth studies on ceramics of individual sites (e.g. McIntosh 1995; Schmidt et al. 2005; Gallin 2011; van Doosselaere 2014) as well as studies on the spread of particular decorative practices (Livingston Smith 2007) or shaping techniques (e.g. Sterner and David 2003; Mayor 2011a). Three main features of the wider region’s ceramic record are important to this study: 1) the relative lack of clear temporal change, 2) high contemporary variability, and 3) pottery as exchange items.
1.1.1 Lack of temporal markers:
Susan K. McIntosh’s analyses of the ceramics of Jenné-jeno and surrounding sites resulted in a first detailed pottery chronology on the Middle Niger. In two studies (McIntosh and McIntosh 1980; McIntosh 1995), McIntosh was able to show how some pottery shapes, fabrics and decorations showed clear temporal trends. There was, however, also quite strong stability, particularly in many of the decorative techniques (McIntosh and McIntosh 1980, pp. 233–244). Subsequent studies on other sites have also shown change in ceramic production, and particularly in its decorative aspects, to have been very gradual, and stability to have been the norm (e.g. Schmidt et al. 2005). This contrasts with the archaeological record in many other regions of the world, where the outward appearance of pottery has been far more changeable over time and can therefore be used as chronological markers. This has not been possible for most periods and areas of the Middle Niger region, where decorative aspects tend only to vary in their proportion through time.
1.1.2 Ceramics and social boundaries:
The more striking differences between archaeological ceramics in the area are not diachronic, but contemporary. Pottery from different areas varies in its forming techniques, its shapes (though the most pottery is globular or ovoid in its basic shape), and in decorative practices. Occasionally, as at Tongo Maaré Diabal (Gestrich and MacDonald 2018) or Kumbi Saleh (van Doosselaere 2014), different productions can be clearly separated. The mechanisms underlying these differences have been the subject of extensive ethnoarchaeological and archaeological research by the University of Geneva. Led from the outset by the goal of relating ceramic production to ethnic groups (Gallay et al. 1998, pp. 15–22), this extensive research documented how certain parts of the ceramic production – notably base forming techniques and tools employed in the potting process – could be employed to create typologies that were patterned by the ethnic identity, economic milieu, and ecological zone of the producers (Gallay et al. 1998, pp. 106–107; Mayor 2011b, p. 14). But certain decorative techniques were also found to have a certain correlation to the ethnicity-based typology created by the Geneva team (Mayor 2011b, pp. 40–78). The main limiting factor here, as Mayor stresses, is that “stylistic borrowings are relatively easy and frequent” (Mayor 2011b, p. 78), but given the fact that some of the decors do correlate with ethnic groups — or at least with the types assigned to ethnic groups — Mayor hints at the suggestion that decorative techniques can historically be assigned to one or another group from which they originated (ibid.).
Based on their extensive fieldwork, Mayor’s hypothesis is that pottery making is learnt within stable and relatively impermeable family relationships. This is in keeping with some of Gosselain’s results from Niger, where he has suggested that this initial approach to the shaping of a pot was the most influenced by teaching and learning (Gosselain 2000, p. 193). Since this usually happens from an early age within the family, it makes sense that kinship relationships, and thus often group and linguistic belonging, would influence the propagation of such techniques. These processes of knowledge exchange, limited to endogamous groups and ethnic milieus, are what explains the commonalities that underlie the typologies created by the Geneva group. Exchange outside of these limits is seen as the exception, rather than the norm. Nevertheless, Mayor gives the following scenarios for the diffusion of pottery traits (Mayor 2011b, p. 83): Firstly, when a technique is simple and expedient, thus easy and profitable to learn. Secondly, when an ethnic group is in the minority in a certain area or when communities find themselves at the geographical margins of their ethnic group. Thirdly, rare cases in which the social order is transgressed by marriage outside of the ethnic group. Mayor thus expects that shaping and decorative techniques would have had a single and ethnically distinct origin and that it is “several centuries of borrowing” (ibid.) that have led to the less demarcated situation that the Geneva team observed in the 1990s and 2000s.
1.1.3 Ceramics as trade goods:
Even though most ceramic production appears to have been made for local markets, there is also evidence of pottery being used in trade over longer distances. Historically, the practice of pottery being moved on the Niger by boat and sold widely throughout the region has been documented in an account from the 18th century (Park 1799, p. 219). From the archaeological evidence, it seems as though this was not confined to later periods, but at least occasionally practiced also in the early second millennium and before. Finds of imported ceramics have, for instance, been made at Essouk (Nixon 2017, p. 143,146–147), and at Kumbi Saleh (van Doosselaere 2014, pp. 228–233). We must therefore always consider the possibility that archaeological assemblages do not only reflect local production, and add the indirect contact from long-distance trade to the possible mechanisms of change in ceramic practices.
If the goal of the work by Gallay and Mayor and their teams was to identify the borders between groups in the past, this study asks a different question: how was pottery knowledge transmitted? What were the factors limiting or encouraging the flow of information? We thus begin from a null hypothesis of unlimited sharing, rather than of delimited groups, to see how the results might complement those of the Geneva research teams. Using pottery décor as a relatively easily borrowed attribute, we aim to generate hypotheses on how pottery making knowledge spread along the ancient Middle Niger. Further, we wish to demonstrate the use of network analyses for approaching West African archaeological ceramics and for making such connections visible and explorable. This technique presents a change of perspective on the ceramics of this region, allowing questions of connection to take precedence over those of boundaries.
1.2 Ceramic decors as networks
In recent years, the use of network analysis in archaeology has been growing significantly (see e.g. Knappett 2021). A large number of these studies focus on ceramics, but geographically most important are North America (Hart 2012; Mills et al. 2013; Crabtree 2015; e.g. Barker and Young 2017; Vavrasek 2020) and the Mediterranean (Torrence and Swadling 2008; Brughmans and Poblome 2016; Iacono 2016; de Groot 2019; Sternini 2019). The archaeological records of these regions are amongst the best and longest studied in the world, with fine-grained pottery typochronologies and extensive excavations. To our knowledge, this is the first application of a network approach to archaeological ceramic material from sub-Saharan Africa, where the conditions are somewhat different (McIntosh 1995, pp. 130–131). That is not to say that archaeological pottery on the Middle Niger is poorly studied. Most researchers in the wider area have detailed publications on their ceramic assemblages, efforts have been made at systematizing nomenclature (Haour et al. 2010) and at synthetic overviews (Mayor 2011b). A large body of ethnographic and ethnoarchaeological literature provides the theoretical basis that informs the network model.
Ethnographic studies of pottery manufacturing in Africa suggest that the way that pottery is made depends largely on systems of learning and teaching the craft (Gosselain 1998), on local market consumer choices (Dietler and Herbich 1994), or on marriage patterns among manufacturing groups (ibid.). In short, the processes underlying the propagation of technical practices through time and across geographical space are dependent on communication and, more specifically, knowledge exchange. Such exchange occurs between humans and we can therefore understand similarities in the production as evidence that knowledge was shared. However, different elements of the pottery making process are thought to be reflective of different kinds of social interaction. Gosselain (2000) concludes that, on the whole, forming techniques are more stable, reflective of durable cultural ties, sometimes even ethnic groups. On the other end of the spectrum, decorations are thought to be influenced by far more ephemeral contact, in which seeing someone else’s pot might be enough to change a potter’s production, as could indirect contact via the wishes of customers.
Decoration on pottery vessels is a topic that has been central to archaeological debates for a very long time. To construct an overly simplified history of ceramic decors in African archaeology, we might distinguish an early period analogous to other parts of the world, in which decors (as a main constituent of typologies) served to distinguish human groups. This position became more elaborate and changed rationale over time, arguably culminating in Hodder’s “Symbols in Action” (1982). David et al.’s “Why pots are decorated” (1988) did much to remove confidence in this idea and to situate decorative practices as part of chains of production and expressions of often unconscious propagation of social values. For them, “designs on pottery, far from being ‘mere decoration,’ art for art’s sake, or messages consciously emblemic of ethnicity, are low-technology channels through which society implants its values in the individual — every day at mealtimes” (David et al. 1988, p. 379). Over the intervening years, the comparatively strong ethnoarchaeological research in Africa has developed an increasingly complex picture of ceramic décor. Gosselain’s (2011) “pourquoi le decorer”, part review, part programmatic article, gives a good overview : Decorations on pottery occur within cultural constraints and logics, are expressions of individuality, are market-oriented, have practical considerations, are made for visual effect, have profound meanings while being thought of as trivial, are subject to fashion and taste, stand for group membership, mean different things to different people. In short, pottery decoration is not anchored in any stable way to any cultural phenomenon. More importantly for this study, however, Gosselain points out that the propagation, the process of spreading, of decorative practices is a more fruitful topic for study (ibid.: 12). Various decorative practices presuppose various kinds of contact. Some of these, as in the case of a particularly intricate rouletted tool, require the sharing of cultural knowledge and probably reflect shared practices in other cultural domains as well. Others are transmitted more easily, by sight alone. Whatever these mechanisms, there are constraints on transmission and innovation. As Herbich (1987, p. 202) notes, potters innovate and express themselves with reference to other potters they interact with. Therefore, the innovation and spread of decorative practices does not occur freely, but is helped by geographical proximity, social contact, and cultural factors. It is likewise hindered by geographic distance and cultural differences.
This, then, is the basis on which our social network model is conceived. It supposes that co-occurrence of a given practice of decorating pottery is emblematic of a form of sustained social contacts, within the limitations mentioned above. These contacts are not as close as we would expect from a teaching and learning environment, which would be better reflected in forming techniques, but are sustained and meaningful enough for there to have been an exchange of cultural aesthetic values and an exchange of knowledge of how to decorate pottery. These exchanges are therefore what the ties in our network represent.
In choosing sites as the nodes of the network, the data come mainly from contexts in which pottery was consumed, rather than produced. The network agents in the model are therefore communities of consumers. Their ties, based on shared decors on their pottery, are indirect: they must be understood to have been connected via producers (individuals and communities) and possibly also merchant middlemen which remain largely unseen in the archaeological record. Since we have established at least occasional long-distance trade in pottery in our research area (see above), this is clearly a factor limiting the model’s explanatory power. Nevertheless, we can use the work carried out by the Geneva MAESAO project to suggest that the majority of pottery found at a site will generally be locally produced (Mayor 2011b, p. 84), which might reasonably allow some conflation between pottery producers and consumers in the network nodes. A low level of imported ceramics can be adjusted for in the network model, but should not be excluded, as these “weak ties” have been shown to be important for introducing innovation in social networks (Granovetter 1973), including technical innovation in ceramic production (Roux et al. 2018).
The analysis of the network will attempt to generate hypotheses on how such knowledge exchanges were structured, the level of interaction between inferred producers of ceramics, and the overall level and structure of cultural contact in the region. This differs somewhat from the rationales for pottery décor networks in studies from other parts of the world. Hart and Engelbrecht (2012, pp. 324–330) justify their building of networks based on pottery décor similarity by arguing for décors as signals in which both group membership and individual skill are expressed. Iacono (2016), on the other hand, takes the co-occurrence of pottery decoration as analogous to other forms of social encounters, and thus as representative of them, conveying the agency of individuals “simply through sensory recognition” (Iacono 2016, p. 126). We would like to stress here that these differences are in no small part due to the history of research and the nature of the archaeological record in different parts of the world, and that the building of network models on archaeological data always requires a well-theorised foundation appropriate to the historical context of the studied material.