Introduction [**]

1Due to the lack of ancient written records, historical-comparative linguistics has grown to be an important way of accessing the early history of Africa. Within the realm of Bantu studies, ongoing internal classifications of the Bantu languages have not only led to changing hypotheses on their evolution and dispersion, but these linguistic theories have also incited and oriented archaeological and cultural historical research in this field of study [1]. As described (and criticized) by several authors, the “Bantu expansion” question has fed interdisciplinary debates since decades [2]. In absence of adequate descriptions of the Bantu languages, scholars could only rarely rely on the most suitable technique for language classification, i.e. the “Comparative Method”. They had to fall back on qualitatively less appropriate methods as “mass comparison” or more frequently “lexicostatistics [3]”. The last method calculates the relatedness between two or more languages from the number of cognate words they share in a standardized list of so-called “basic vocabulary”. Depending on vocabulary that is supposed to be culturally neuter, e.g. body parts, common animal names (bird, dog, fish...) and nature elements (sun, moon, rain, water...), this quantitative approach of lexical data offers us, logically, few insights into the cultural history of African societies. The comparative study of cultural vocabularies is far more telling in this respect. As D. Nurse put it, “terms that refer to social institutions and activities, the spiritual world, the economy, domestic crops and animals, food production and preparation, tools and weapons, flora and fauna are clearly of great significance for historical reconstructions [4]”. Their importance derives from the basic assumption of this more qualitative lexical approach, namely that shared vocabulary testifies shared history. Simply stated, lexical resemblances between languages can be reduced to two kinds of origins. Leaving aside mere coincidence, resembling words are either inherited from a common ancestor or adopted through linguistic contact. In the former case, related languages share a lexical item that can be reconstructed to their ancestor language, which is in the African context most often a hypothetical proto-language. Consequently, it is assumed that the speakers of the ancestor language were familiar with the referent of the reconstructed term, whether it is an object, a technique or a mental concept. In the case of lexical items adopted through linguistic contact, lexical borrowings reveal cultural exchange between peoples, who do not necessarily speak related languages. Hence, the study of loanword patterns may be instructive regarding the introduction and the transmission of cultural items [5]. It is along these lines that developed the method known as “Words and Things”, which originated in early 20^th century Indo-European linguistics [6].

2In the field of African linguistics, the “Words and Things”-method only gained significance from the 1960’s on. Interestingly, historians, and not linguists, turned out to be the first scholars who tried to benefit from the method. This is not so surprising, if one takes into account the general absence of ancient historical documents and the marginal state of archaeological findings at that time. One of the first and most passionate advocates of lexical comparison for the use of historical reconstruction is without any doubt C. Ehret. From his early papers on cattle and sheep breeding until his recent articles on ironworking, both lexical reconstruction and loanword studies have played a crucial role in his attempts to unravel the tangled history of eastern and southern Africa [7]. His writings have heavily influenced historical and archaeological reasoning on this region’s past and his innovating method has provided a model for research. He initiated, so to speak, a school of “linguistic historians” who propagated his approach to other areas of the continent, e.g. the Great Lakes region, the Uele region, the Bantu-speaking Equatorial Rainforest, and more specifically its western part, Tanzania’s Corridor and Central East Tanzania [8]. Despite the importance of their contribution to hypothesis building on the history of the Bantu domain and its neighbouring areas, their methodology has faced serious criticism, in particular in historical linguistic circles. The far-reaching historical accounts, which these “linguistic historians” develop from often scanty and qualitatively poor linguistic data, have baffled “historical linguists”, not at least when it concerns wide-ranging vulgarised literature [9]. Although this linguists’ amazement may partly be explained by their own incapacity to draw a coherent historical narrative from the lexical domains they reconstruct, more is at stake. At the risk of wronging the individual qualities of each of the above-mentioned studies, it can be claimed that the common inconsistencies in linguistic method, which they manifest, jeopardise the historical soundness of the hypotheses they put forward. A critique on C. Ehret’s work, accurately formulated by D. Dalby, can to certain extent be considered as a first common drawback: “Ehret writes with the seeming assumption that his readers, be they linguists or not, are acquainted with the vocabulary and structure of most individual languages between the southern Sudan and Natal. Without such knowledge at one’s fingertips, however, it is impossible to verify the many starred protoforms he is able to produce as evidence (frequently without supporting data or exposition of method) from both the Bantu and non-Bantu fields [10]”. Even if the reconstructions or comparative wordlists are often offered in appendix or published separately [11], the linguistic analysis necessary to distinguish for example inherited from diffused vocabulary or retentions from innovations is not presented. Another methodological drawback is the concentration on one particular region of the Bantu domain, which may lead to certain degree of historical myopia. Ignoring data from regions other that the one at issue often results in a bad estimation of a lexical item’s degree of archaism. The geographic distribution of a term is sometimes larger than supposed. Inversely, the general distribution of a term in one of the major branches of the Bantu group, for instance among the Savannah or eastern Bantu languages, does not necessarily entail its reconstruction to Proto-Bantu, since it may be absent from other major branches, such as the north-western or central Forest Bantu languages. As the examples of the early Bantu verbs meaning “to mould pottery” will illustrate, such is not possible if one does not take into account sufficient comparative data from the entire Bantu area. For at least some scholars, this shortcoming is in part due to the fact they restrain themselves to global lexical reconstruction databases, as M. Guthrie or Y. Bastin et al., which are generally based on a number of test languages believed to be representative for the whole [12]. The distribution maps one can draw from them are often fragmentary and thus not as appropriate for historical reconstruction as the ones based on more detailed databases composed by means of thorough literature study and personal fieldwork. Another side effect of relying on these large-scale lexical reconstruction works is the underestimation of the historical impact of semantic shifts. As I will briefly show in this paper, semantic shifts may have crucial historical inferences, certainly if the different meanings of a verb root or noun stem have geographically complementary distributions. Moreover, some of the above-mentioned historians start from a pre-established historical scheme, which they try to confirm by means of lexical data. Obviously, this may skew the results. C. Ehret, for example, has proved industrious in the collection of evidence supporting his “Central Sudanic theory”, which made him less preoccupied with other evidence [13]. The slightest phonological or semantic resemblance seems to suffice to link a Bantu item to a Central Sudanic item [14]. It is almost always assumed that lexical borrowings were transferred from Central Sudanic to Bantu, but very rarely the other way around. Finally, “linguistic historians” fall without exception back on glottochronology to outline their historical narratives within a framework of absolute dates, despite the fact this method is almost unanimously declined by linguists [15]. As a result, historical linguists tend to be reluctant to accept the “Words and Things”-hypotheses postulated by “linguistic historians”.

3Does this mean that historical linguists do not believe at all in the merits of the Words and Things-method for the reconstruction of early history ? Not exactly, but it is true that this kind of historical research has aroused little enthusiasm in African linguistic circles. One of the reasons for this lack of interest is the fact that this kind of studies is not considered as “genuine” linguistics. Another significant factor has been raised before, i.e. linguists – and I assume – find it extremely difficult to draw articulate historical conclusions from their comparative studies, because they lack a proper training to do so. Some noble exceptions notwithstanding, the method had not set in yet in African linguistics at the time C. Ehret started his vocabulary based studies [16]. D. Dalby was one of the first linguists to evaluate the historical implications of Guthrie’s lexical reconstructions [17]. The study of P. de Maret and F. Nsuka on iron-working vocabulary in Bantu, actually the collaboration between an archaeologist and a linguist, was innovative in the sense that it relied on a corpus composed by means of a systematic examination of the available ethnographical and linguistic sources and that it lent much significance to the notion of semantic shifts [18]. In the last decade, although still marginally, the number of linguists dedicating their comparative research to the study of cultural vocabularies for the use of African history is slightly growing [19]. This increasing linguists’ devotion to the study of cultural vocabularies possibly points towards a growing consciousness that comparative linguistics, in combination with comparative ethnography and (ethno-)archaeology, may supply invaluable insights into the cultural history of the Bantu domain in particular and the African continent in general. Unlike “linguistic historians” who generally care about the “global” history of one particular region, all of the above-mentioned linguists tend to focus on one specific lexical domain.

4The aim of this article is to present a particular case study of a comparative linguist’s historical approach to cultural vocabularies. I will analyse the pottery vocabulary of the Inner Congo Basin (ICB) languages (§ 2) within the wider framework of a comparative study encompassing the entire Bantu domain [20] (§ 3). As will be demonstrated, the terminology studied is quite distinctive from a common Bantu perspective (§ 5+6), just as is the case for other domains of cultural vocabulary (§ 8). Although I have tried to avoid an all too technical discussion, the linguistic analysis presented in § 5 and § 6 is necessary in order to pay attention to important historical linguistic concepts as inherited vs. borrowed vocabulary, lexical retention vs. innovation and semantic shift. Underlying this approach will be the question what comparative linguistics can tell us – but also what it cannot tell us – on the diachronic evolution of a cultural vocabulary domain, and by extension on the history of the craft industry to which it refers (§ 7). Since the fabrication of pottery is practised by people, these lexically inferred hypotheses unavoidably entail claims on the early human settlement in this part of the equatorial rainforest. Hence, the obtained results will be faced with the (ethno-)archaeological information available for this area (§ 9). Concluding remarks are formulated in § 10.

Bantu languages of the Inner Congo Basin

5In this paper, I will take a close look at the pottery vocabulary of a limited set of Bantu languages, namely the languages of the Inner Congo Basin (Appendix A). The geography of this particular region is characterised by a large network of tributaries of the Congo river, as there are the Lulonga-Lopori/Maringa, Ikelemba and Ruki-Busira-Tshuapa/Momboyo to its east and the Sangha/Likwala-aux-Herbes to its west. This huge aquatic system concurrently forms a natural system of communication highways and linguistic barriers. Apart from some Ubangian languages spoken at its northern fringe, only Bantu languages are spoken in this area. I will essentially focus on the languages situated along the upper stream of the Congo River, crossing the Equatorial rainforest between approximately the present-day RD Congo cities of Mbandaka in the west and Kisangani in the east, and the languages to the south of it.

6These languages belong to zone C of Guthrie’s classification, predomi-nantly to the groups C30, C40, C50, C60, C70 and C80 [21] (see Appendix B). The C10 and C20 languages, situated to the west and north-west of the great Congo river bend are very scarcely documented and will only be referred to as far as the available data allow. Moreover, their genealogical status vis-à-vis the above mentioned languages is not straightforward yet. According to lexicostatistical studies, the ICB languages can be roughly subdivided in two groups of genealogically related languages: the Congo River (CR) languages and the Congo Basin (CB) languages [22]. The CR languages are northernmost and include the C40 and C30 languages, apart from Sakata (C34), Ntomba (C35a) and Bolia (C35b) [23]. The CB languages consist of the rest of the C30 languages and the C60, C70 and C80 languages. The status of the C50 languages is not entirely clear. According to Bastin et al. (1983), they go with the CR languages. In a more recent classification, which is based on far more extensive data, they are closer to the CB languages [24]. Both groups probably stem from a common parent language, which J. Vansina situates in the vicinity of the lower Ubangi and Congo River area [25]. On a deeper time level, the ICB languages belong to Western Bantu, which forms, together with Northwest and East Bantu, one of the major Bantu branches. The other principal Western Bantu subgroups are the South-Western languages, comprising most of Guthrie’s zones K and R and several languages of zones L and Mbala (southwestern DR Congo, western Zambia, southern Angola, Namibia, northern Botswana) and the West-Coastal languages, including Guthrie’s zone H (except Mbala) and a large part of zone B (B40-80) (southwestern DR Congo, Congo, southern Gabon, northern Angola [26]).

Comparative Bantu pottery studies

7Pottery is a thankful topic from an interdisciplinary point of view. It is an artefact that unites a high archaeological visibility, a high linguistic salience and a high ethnographical prominence. Earthenware is extremely resistant to decay. Consequently, it has proved to be a major archaeological source of information, certainly in humid forest environments where other materials as wood and metal easily vanish. Cross-linguistically, at least certain aspects of pottery vocabulary tend to be prominent, in the sense that they manifest lexical resemblances, which allow either to reconstruct inherited vocabulary or to retrace loanwords patterns. Moreover, pottery making is a traditional craft that has continuously been practised in many African communities, even if it is rapidly losing ground now. As a result, it is abundantly documented in ethnographic literature. The comparison of the ways in which people currently mould pots may be historically instructive, since pottery traditions consist of a complex mix of inventions, borrowed elements and manipulations that display an amazing propensity to redefinition by individuals and groups [27]. The tools, actions and rules are traits that have been accumulated and transmitted from generation to generation [28]. Pottery studies form thus a rewarding field of interdisciplinary historical research.

8Given its many-sided nature, tackling the history of pottery making in the Bantu domain by lexical means cannot be done without a large and varied corpus. The Comparative Bantu Pottery Vocabulary Database, built up during my PhD-research, constitutes such a corpus. This database results from the gradual accumulation of data from both linguistic and ethnographic sources on the one hand and personal fieldwork in Uganda, Burundi, Tanzania, Malawi and Zambia on the other hand. It contains at present almost 5800 entries from nearly 400 Bantu languages on aspects as diverse as the materials and tools used, the manipulations undertaken and the final products obtained throughout the production process. Thanks to its size, this database is a valuable tool for comparative research, which allows mapping the distribution of individual stems in a fairly detailed way, retracing semantic shifts and reconstructing the evolution of pottery vocabulary at different levels of historical depth. Despite its extent, this database supplies quite heterogeneous information. First of all, some languages are far more documented than other languages. The available vocabulary ranges from a generic term for “pot” or “to make pottery” to an exhaustive listing of the terms related to the different aspects of the production process. Among the ICB languages, the information available for Mongo for instance is much more abundant than for Buja. Another disparity is the fact that the words for some aspects of this lexical domain are more frequently found than others. Nouns referring to types of pots for example are more commonly noted in dictionaries than terms referring to particular manipulations or tools. Given these inconsistencies, one should understand that comparative pottery vocabulary studies might render basic information on the history of this craft, but that it is rather predestined to include little historical detail.

Short note on linguistic method and notation

9In this paper, mainly two types of linguistic data will be shown: actual language data and lexical reconstructions. The former kind of data consists of terms from present-day Bantu languages, as they were noted down in dictionaries, grammars or ethnographical sources. The latter kind of data involves linguists’ abstractions of the lexical resemblances observed between the real language data of several languages. Just like the French cheval, the Spanish caballo and the Italian cavallo go back to the Latin caballus “horse”, the Ngombe mo-lema, Tetela o-tema and Ganda omu-tima are supposed to derive from the Proto-Bantu *–témà “heart”. Unlike for Latin, written testimonies of Proto-Bantu, the common ancestor of the Bantu languages, do not exist. The reconstruction *–témà is thus a hypothetical older form from which the present-day forms can be derived, taking into account the sound correspondences observed between their languages. In the text, lexical reconstructions can be distinguished from real language data by means of the signs * or ° that precede them. Sure proto-Bantu reconstructions are preceded by *, while unsure and/or regional lexical reconstructions are preceded by °. I only present reconstructed noun stem or verb roots, which are indicated by the middle score. Since they have a variable final vowel, verb roots are both preceded and followed by a middle score, e.g. °–gèng– “to mould (pottery)”. As their final vowel is invariable, noun stems are only preceded by a middle score, e.g. *–bègá “pot”. The noun prefixes, e.g. mo–, o–, omu–, that precede them in the present-day languages are not considered in the reconstructions. It is good to realise that hypotheses on the early cultural history of Bantu speaking communities are actually founded on these hypothetical lexical reconstructions, which are derived of course from the real language date.

10This paper also contains certain orthographic signs, which may be unfamiliar to non-linguists. The use of these uncommon characters is necessary to note in a proper way certain features of the Bantu phonological system, in particular vowels and tones. Proto-Bantu had 7 distinct vowels, which different authors represent in different ways. In this paper, the following vowel notation system will be used to represent reconstructions: i e ? a ? o u. The ? or “open e” roughly corresponds to the vowel of the English “red”. The ? or “open o” roughly corresponds to the vowel of the English “rod”. This system deviates from the one used in the most extensive Bantu Lexical Reconstructions database [29], which is also the one used in my previous publications, i.e. i ? ? e a o ? u. The latter system is more accurate from a pure phonetic point of view, but the one used in this paper is seemingly more accessible for an audience missing linguistic training. Moreover, it has the advantage of corresponding to the system most frequently used for the ICB languages and the other Forest Bantu languages, which have a 7 vowels system. However, since none of these languages has a long written tradition, an orthographic standard was not developed yet for all of them, so that the notation modes may differ from one author to the other. In most of the non Forest Bantu languages, however, the Proto-Bantu 7 vowels system was reduced to only 5 distinct vowels. Hence, the more common system i e a o u may be used. In sum, following notation correspondences need to be taken into account while reading the present article:

11Another important feature of Bantu languages is the fact they are tone languages. Two fundamental tones can be distinguished. If they are marked, they are commonly noted with diacritic signs on the vowel: High (á) and Low (à). Most often, only the less frequent tone is marked. A proper notation of tones is crucial, since differences in pitch height involve semantic differences. Simply stated, words with identical vowels and consonants, but with different tones have different meanings, e.g. in Mongo botúli means “blacksmith”, while botulí signifies “misfortune, disaster”. Such vowel and tone differences have not always been suitably marked in scientific literature, as can been observed from several sources cited at the end of this paper, and they are usually still not marked in popular press. Nevertheless, a proper orthographical distinction of these different vowels and tones is crucial for scientific studies that aim at drawing historical conclusions from comparative lexical data. It is indispensable in order to avoid haphazard associations based on apparent lexical resemblances caused by bad notation.

12Hereunder, I comment the pottery related lexical reconstructions, which contribute to hypothesis building with respect to the early human history of the Inner Congo Basin. For each of them, an example table with reflexes, i.e. present-day real language terms that are supposed to stem from these reconstructions, is given in Appendix C. For the notation of these data, I will respect the orthography as it is found in the respective sources, even if it deviates from the notation system used for reconstructions.

Archaic Bantu pottery terminology

13As explained above, some “linguistic historians” manifest a certain degree of historical myopia, because they focus on a certain region and neglect the geographically wider or chronologically deeper context. If one concentrates on a set of cultural vocabulary of one particular group of languages, as I do in this paper, this should be seen against the background of the corresponding vocabulary in more distantly related languages.

14In the case of the ICB languages, this wider framework is in the first place the rest of Western Bantu, with which they form a unit historically distinct from the rest of the Bantu languages. However, in spite of their supposed genealogical unity, no pottery terms have been found that are both common to all Western Bantu groups and totally absent from the other Bantu languages. Although the distribution of certain lexical items is limited to Western Bantu, none of them is sufficiently widespread to be reconstructed to a common Western Bantu ancestor. If Western Bantu languages really form a genealogically distinct unit, this absence of shared pottery vocabulary could point towards a rapid disintegration after their separation from the common Bantu trunk and the development of specific pottery terms. Although not mentioned yet, the ICB languages belong in certain historical classifications to a subgroup called “Forest Bantu”, mainly consisting of – logically – the languages spoken in the forest areas of the Bantu domain [30]. This subset is thus partly overlapping with what I call here “Western Bantu [31]”. Nevertheless, no distinctive pottery vocabulary for the whole of the Forest Bantu languages could be reconstructed either.

15The next comparative level to be considered is the entire Bantu domain. In historical terms, Proto-Bantu is the hypothetical proto-language, which is regarded as the common ancestor of present-day Bantu languages. Comparative research resulted in the reconstruction of a small number of pottery terms to Proto-Bantu. The most clear-cut Proto-Bantu terms are nouns referring to types of pots, i.e. *–bègá and *–jòngó. The former was in all likelihood the generic term standing for the entire category of earthenware pots. It is most often used in that way in contemporary Bantu languages. If its function is mentioned in the literature, it may, according to the language, serve either as a container for liquids or as a vessel to serve or prepare food. The latter term probably designated a cooking pot, since it nearly always assumes this function in present-day languages. Table 1 (in Appendix C) presents some reflexes of these two proto-forms. Their reconstruction is unambiguous, since they have a general distribution among the present-day Bantu languages. Their reflexes occur in all historically relevant Bantu subgroups. Their existence in Proto-Bantu suggests that the speakers of the common Bantu ancestor language used earthenware pots, not only as containers, but also for the preparation of food.

16The speakers of Proto-Bantu did not only use pottery, but they also produced it. Previous studies also came to this conclusion on lexical grounds. M. Guthrie inferred from the reconstruction *–bómb– meaning “to mould pottery” that pot-making was a regular activity of the speakers of Proto-Bantu [32]. Numerous authors adopted *–bómb– as the linguistic evidence for the existence of pottery making in the Proto-Bantu era [33]. Although the historical conclusion as such should not be discarded, the lexical grounds on which it is based should be reviewed. For reasons which I cannot develop in detail here, *–mà– should be reconstructed to Proto-Bantu as the verb designating the fashioning of pottery, and not *–bómb–. The main reason to decline *–bómb– is semantic. Although this verb has a very general distribution in all Bantu subgroups, it is only in certain Savannah Bantu languages that it means “to mould pottery”. In the extremely rare Forest Bantu languages, where this sense is attested, it can be attributed to contact with neighbouring Savannah Bantu languages. Hence, this meaning is in all probability the outcome of a semantic shift, or rather specialisation, posterior to Proto-Bantu. The verb by itself must have existed in the proto-language, but it most likely had the more general sense “to fashion in clay”. For more information on this hypothesis, I refer to previous publications [34]. In any event, this hypothesis shows the importance of semantic shifts for a sound comprehension of language history. Concerning the verb *–mà–, it probably goes back to Proto-Bantu, despite the fact its distribution is limited to Western Bantu. However, it is widely attested in more distantly related non-Bantu languages of West-Africa. This observation suggests that the term is older than Proto-Bantu, which inherited it from an ancestor language, possibly Proto-Benue-Congo, or even Proto-Niger-Congo [35]. In terms of cultural history, this indicates that pottery was certainly not an invention of the Proto-Bantu speakers, but a craft that was practised by their ancestors since many centuries. This linguistically founded hypothesis is in line with archaeological findings [36]. Table 2 presents some reflexes of both verbs.

17The reconstruction of a Proto-Bantu noun for pottery clay, the potter’s basic material, is a puzzle that is too complex to be treated here. The verbs *–mà– and *–bómb– both have a derived noun that signifies “pottery clay”, but none of both is sufficiently widespread to be doubtlessly reconstructed to Proto-Bantu. The problem of °–bómbà “pottery clay” will briefly dealt with further on. The only established reconstruction that is somehow correlated is . Even though this noun refers specifically to pottery clay in certain present-day languages, its sense is usually more general. In most languages, it signifies “clay” or “earth” tout court. Hence, it probably had a broader sense in Proto-Bantu as well. Table 3 presents some reflexes of “soil, earth, clay”.

Historical evolution of ICB pottery vocabulary within a Forest Bantu context

34Which historical narrative can one draw from this comparative study of pottery vocabulary in the ICB languages ? Given the nature of the data considered – a limited lexical domain on which the documentation is quite scanty in most of the languages studied – the historical information obtained as such will necessarily be sketchy. However, if we link the two major characteristics of the Inner Congo Basin pottery terminology, we may sketch a rough pattern of historical evolution. On the one hand, the languages at issue miss almost collectively the most archaic Bantu pottery vocabulary, i.e. the terms reconstructed to Proto-Bantu. On the other hand, there exists a series of pottery related lexical items, which are characteristic of the Inner Congo Basin and refer to more or less similar extra-linguistic realities as the archaic terms. Given their local distribution, the latter are necessarily more recent. Significantly, these indigenous terms are not shared by all ICB languages. We observed a neat split, which coincides approximately with the lexicostatistically based genealogical division between the CB languages and the CR languages. Simply stated, the absence of reflexes of the archaic terms can be interpreted as a rupture of the ICB languages with the Proto-Bantu tradition. Since people still needed vocabulary to talk about pottery, we could presume then that the archaic terms were replaced with more recent ones in the respective ancestors of the two ICB language groups. Nevertheless, the historical reality was probably more complicated. While the absence of the archaic terms indicates undeniably a separation from the Proto-Bantu tradition, it is difficult to imagine that this rupture was accomplished so abruptly, in the sense that a complete set was swapped for a complete other set. This is first of all unlikely, because the Proto-Bantu terms do not correspond one to one to the respective terms of the two Inner Congo Basin sets. While the verb *–mà– “to mould pottery” is quite directly “mirrored” by °–gèng– in the CB languages, such an equivalent has not been observed in the CR languages. It should be said however that the documented verbs referring to moulding with clay in these languages are limited to two: the above-cited Ngombe expression –boma téní “fabriquer des pots [62]” and the Lokele verb –kuka “to mould with clay [63]”. More ample data might reveal a recurrent term shared by several CR languages. Concerning the nouns for types of pots, the Proto-Bantu –jòngó “cooking-pot” seems to have fairly clear-cut corresponding noun stems in both ICB language groups, i.e. °–pògé and °–càcó. Such is not really the case for the other Proto-Bantu noun *–bègá “pot”. In the CB languages its generic function is rather fulfilled by °–pògé, which designates simultaneously a specific kind of cooking pot, while the available data for the CR languages do not permit distinguishing a generic term. Given the rather undefined extra-linguistic reference function of °–bègá (according to the language, it may refer to either a recipient for holding liquids or a common kind of pot for preparing or serving food), it is difficult to determine whether a local term can be considered as an equivalent of – bègá or not. In this respect, any and none of the other Inner Congo Basin nouns designating a kind of pot can be seen as corresponding to –bègá, not at least because their own semantic load does not surface in the barely clarifying dictionary translations. The second reason why it is not possible that the Proto-Bantu lexical set was replaced collectively by the Inner Congo Basin vocabularies relates to the distribution of the Proto-Bantu terms among the Western Bantu languages. Since the ICB languages are part of this larger genealogical unit, it can be safely assumed that a term of Proto-Bantu origin, which occurs in other Western Bantu subsets, especially in the immediately surrounding languages, but not at all in the ICB languages themselves, was replaced after the ICB languages separated from the rest of Western Bantu. Such is certainly the case for *–bègá and *–jòngó, which are well attested in the northwestern part of the Bantu domain, particularly in the immediately surrounding languages. Such is not the case however for and *–mà–. The former has only reflexes in the extreme northwest of the Bantu domain, i.e. in some languages of Guthrie’s zone A. So, it must have been lost before or during the early dispersion of the Western Bantu languages. The latter is represented in the northwest, i.e. in languages of Cameroon and Gabon, in the extreme southwest, i.e. in some Bantu languages of southern Angola and Namibia and finally, in one language in the extreme north-east of RD Congo. Although it is fairly well attested in Western and Forest Bantu, it is totally missing in the language groups surrounding the Inner Congo Basin. Even if it persisted longer than , it may have been replaced before the formation of the ICB ancestor language. On the other hand, inasmuch as data are on hand, no other recurrent verb than °–gèng– has been noted in the adjacent languages. Verbs referring to pot-making are diversified. Hence, the Proto-Bantu verb *–mà– might have been replaced independently in different Western and Forest Bantu groups. The verb °–gèng– of the CB languages could thus be an immediate substitute of *–mà–. At the present stage of knowledge, this is nothing more than an educated guess however. In any event, if the comparative study of ICB pottery vocabulary shows a rupture with the Proto-Bantu tradition, this separation is rather the result of a gradual loss of archaic vocabulary in the course of Western Bantu evolution than a neat break in the ICB ancestor language.

35As already mentioned above, the absence of pottery vocabulary that is shared by all Forest Bantu groups and totally absent from the other Bantu languages, may suggest that their disintegration happened too rapid to develop distinctive terms. The linguistic heterogeneity of the Forest Bantu languages would thus be mirrored in their pottery terminology. This hypothesis seems to be confirmed by the divergent evolution of pottery vocabulary within the ICB languages. Although the CR languages and the CB languages are more closely related to each other than to any other Western Bantu language group, they do not have any inherited pottery terms in common. The few CB languages that have a typical CR term or vice versa most likely obtained it by means of lexical borrowing, but not by inheritance from a common ancestor. In other words, if the CR and CB languages were once united in a common ancestor, this unity did not last long enough to develop a common distinctive pottery vocabulary. This observation is in line with the hypothesis that their ancestral language consisted of two blocks of dialects which, during their spreading, immediately turned into languages themselves as their speakers became relatively isolated from each other [64]. On the contrary, both have a distinctive set of terms, not only distinctive vis-à-vis each other, but also vis-à-vis the other Forest Bantu languages. This indicates that their ancestor languages stayed long enough united to develop distinctive pottery vocabulary before expanding and disintegrating. This gradual falling apart of the ancestral nuclei led to the emergence of the CB and CR languages and the dispersion of the distinctive pottery they inherited.

36Subsequent to or in the course of the diversification of these early ICB language communities, this isolation seems to have been interrupted by contacts with neighbouring communities. The pottery related lexical borrowings that could be identified point toward such a development. Lexical exchange is observed between the individual languages of both ICB language groups on the one hand and between these languages and geographically and genealogically more distant languages. These lexical transfers are for the most part nouns designating kinds of pots. The commercial relations, which ICB communities developed in 2^nd half of the 2^nd millennium AD, mainly as a result of the emerging Atlantic trade network, can explain how pots and their names circulated over long distances. Such an explanation is less obvious for the ICB reflexes of °–bómbà “pottery clay”, which are assumed to be loanwords as well.

Observations from other cultural vocabulary domains

37Naturally, we need to refrain from jumping to too far-fetched conclusions on the basis of only one domain of cultural vocabulary. Other lexical domains should be taken into account to face them with the above-sketched historical evolution. As mentioned in the introduction, comparative studies focusing on one particular semantic field are still in their infancy. Nevertheless, I would to like to draw the attention to the comparative studies on woodwork in Bantu, carried out by Annelies Bulkens, a former colleague of mine. Her studies revealed to a certain extent similar tendencies. In her article on mortar terms in Bantu, she has demonstrated that the Proto-Bantu term *–dù /*–nù “mortar (for pounding)” is not attested in the central Forest Bantu languages, while the stem °–bòkà is specific for the forest zone [65]. It has reflexes in present-day languages of Guthrie’s zones A and C. However, these terms manifest several phonological irregularities, which point towards lexical borrowing. As she mentions herself, similar terms were identified in certain Adamawa-Ubangi languages, which forms a possible source of lexical diffusion. Since these languages bridge somehow the geographical gap between Bantu zones A and C, languages of each zone may have adopted Adamawa-Ubangi borrowings independently from each other. The existence of these loanwords in Forest Bantu languages could indicate that Adamawa-Ubangi speaking communities diffused a certain mortar-associated form of agriculture in this region. In this sense, it concerns a totally distinct kind of historical evolution, compared to the evolution I proposed for Forest Bantu pottery vocabulary, for which no loanwords of Adamawa-Ubangi origin could be identified. More convincing similarities are found in her paper on the use of calabashes in the Bantu world [66]. The Proto-Bantu reconstruction *–cópà, of which the reflexes most notably designate bottle-form calabashes or gourds, has the most general distribution throughout the Bantu domain. It is almost completely absent however from the languages of the forest areas. She put forward the regional proto-forms °–béndá and °–kútù as the two major lexical innovations of *–cópà. The first one is attested in the closely related languages of Guthrie’s zones B, C, H and L, while the second is limited to the languages of zone C, more particularly C30, C50, C60, C70 and C80. Moreover, the author has claimed that “in former research on other wooden objects (which has not been published yet, note K.B.), similar distribution patterns for words belonging to the cultural vocabulary have been established. Zone C languages innovated their words for canoe, mortar, pestle and spoon, as they did for calabash [67]”. She also identified the region delimited by the Congo River bend as an innovation centre for nautical terminology, despite the fact that , the Proto-Bantu stem for “canoe”, is still largely attested in the languages of Guthrie’s zone C [68].

Archaeology of the Inner Congo Basin

38The observations on the evolution of pottery terminology are not only interesting in comparison with other domains of cultural vocabulary, but also with respect to archaeological findings in the region. The historical narrative built up so far is linguistically founded. As a result, its chronology is relative in nature. Lexical retentions and innovations have been sequentially ordered with regard to the historical evolution of the Bantu languages. Since I decline the use of glottochronology in the context of African linguistics, absolute dating is impossible. This becomes nonetheless achievable, by approximation to be sure, if the lexically based historical pattern could be linked in a plausible manner to an archaeological sequence. What is more, such an association would give to a certain extent a material face to the reconstructed vocabulary. Although one can hypothesise that an innovation in the pottery terminology signals a change in a certain ceramic tradition, such a deduction is merely tentative until it is materially corroborated.

39As mentioned above, ceramics have proved to be the guiding principle in the study of past cultures in sub-Saharan Africa. Since the humid climate of the Equatorial Rain Forest is fairly unsuitable to conservation, pottery is one of the few artefacts, which enables archaeological research in this ecological belt. Consequently, ceramics constitute a crucial trace of early settlement in the Equatorial Rain Forest. A major and unique contribution to the archaeology of the Inner Congo Basin was realised by the “River Reconnaissance Project”, a large-scale river based exploration, executed during the 1970s and 1980s [69]. The archaeological evidence almost exclusively embodied pottery. The discovery of some very early pottery in the central part of the equatorial forest can be considered as one of the major results of the project [70]. The earliest ceramics were labelled “Imbonga Horizon”, named after their type-site Imbonga on the Momboyo River in the west-central part of the Congo Basin. The Momboyo is a tributary of the Ruki, which is in its turn a tributary of the Congo River. The distribution of this pottery is centred on the left bank of the Congo River itself in the environs of Mbandanka as well as on all rivers directly discharging into it, namely the Ruki, Ikelemba and Lulonga [71]. M.K.H. Eggert attributes the earliest Imbonga pottery to approximately the second half of the first millennium AD, while H.-P. Wotzka is more precise in dating it between 400BC and 100 BC [72]. This earliest pottery was associated with the oil palm (Elaeïs guineensis) and the fruit of the wild Canarium schweinfurthii [73], as it is the case for other sites in West-Central Africa [74]. Although the closest, albeit general, parallels to the Imbonga style known so far are found amongst certain Neolithic and Early Iron Age pottery groups of southern Cameroon, Gabon and the Congo, there are no convincing parallels from regions beyond the western fringes of its presently established distribution [75]. Nevertheless, this Imbonga pottery is the only early pottery that could have an origin external to the Inner Congo Basin, in the sense that arising from the unknown, it is at the beginning of a continuous ceramic sequence, covering several traditions and spanning the last 2400 years approximately. This equatorial tradition does not manifest external influences or discontinuities and seems to be entirely indigenous to the Inner Congo Basin [76]. For that reason, M.K.H. Eggert has postulated that the people who manufactured this pottery are intrusive in the forest, but he does not want to make claims on their linguistic and ethnic affiliation [77]. H.-P. Wotzka is far more explicit in this respect [78]. He tentatively connects the earliest immigration of Bantu-speakers with the appearance of the Imbonga group, after 400 BC. The comparative lexical data, which suggest the development of a distinctive pottery vocabulary in the ancestor of the present-day CB languages, seem to corroborate this hypothesis. Given the location of the early Imbonga ceramics, in the Basin just southwest of the great Congo River bend, it could be assumed that it was fashioned by people speaking the parent language of the present CB languages. This is all the more likely, if one takes into account the fact that, the subsequent pottery groups stemming from the Imbonga group gradually extended further up the tributaries of the Lulonga, Ikelemba and Ruki-Momboyo rivers, roughly between the turn of the Christian era and at least 1600 AD [79]. This eastward advance of succeeding pottery styles clearly marks the spread and, most likely, the growth of farming communities in the equatorial rainforest [80]. The progressive expansion of the initial settlement nucleus led to the dispersal of the ancestor language and hence, to the steady diversification of the present-day CB languages. In sum, a quite obvious correspondence can be observed between the isolated evolution of the ancestral Imbonga pottery tradition and its subsequent daughter traditions on the one hand and the development of distinctive pottery vocabulary in the CB ancestor language followed by its dissemination as a result of the gradual disintegration of the ancestral language community on the other hand.

40Worthy of note in the perspective of the indigenous Congo Basin ceramic tradition is the more recent pottery ware produced in the village of Ikenge on the Ruki [81]. The way in which these people manufacture their pots could be seen as a contemporary confirmation of the isolated development of the specific ceramic continuum of the central rainforest, because its is very singular from a comparative technological point of view [82]. The actual modelling of a pot consists of two phases. The upper part of the pot is fabricated by hollowing out a lump of clay, first with the thumb, then with four fingers and finally with the fist in order to obtain the desired diameter, after which the clay is stretched upwards and vaulted in the appropriate way. Subsequently, the potter adds one or several small portions of clay at the bottom of the pot in process. After a short period of drying, the lower part of the pot is shaped by means of a small wooden pestle. By beating the bottom softly, the potter draws the base and the belly of the pot. The exterior is smoothed with a wooden bat. In a comparative study of African pottery fashioning techniques, this particular procedure was set apart as unique of its kind [83]. Except for a certain degree of resemblance with the fashioning mode of the Tiv (Nigeria), the combination of hollowing and stretching out a lump of clay by hammering the bottom has no documented equal in sub-Saharan Africa. Hammering the base of a pot without the use of a depressed or a concave mould as such is already quite rare. This characteristic aspect constitutes an additional element differentiating the Inner Congo Basin from the rest and could be considered as a contemporary outcome of the local uninterrupted ceramic sequence. In default of more specific technological details on the pottery fabrication of other communities in this region, too generalising assumptions need to be avoided however.

41Regarding the distinctive pottery vocabulary developed by the CR languages, finding an archaeological correspondence is more difficult. Not only is the evidence for a separate lexical development less pervasive, the archaeological data to compare with are also less abundant. The Congo River stretch between the mouth of the Lulonga and the modern-day city of Bumba, along which several of the CR languages are spoken, is – to my knowledge at least – quasi undocumented archaeologically. The “River Reconnaissance Project” limited its explorations outside the Inner Congo Basin to the archaeology of region to the west and the northwest of the great Congo River bend. Even if the data are in still shorter supply than in the Inner Congo Basin, this research resulted in the reconstruction of two distinct early ceramic traditions, i.e. the “Batalimo-Maluba Horizon” and the “Pikunda-Munda” horizon, both of which are slightly younger than the “Imbonga Horizon [84]”. The former is concentrated south of the great Ubangi bend and its dating clusters in the first half of the first millennium AD. A partial overlapping with Imbonga pottery cannot be excluded. The latter is attested along the Sangha-Ngoko and Likwala-aux-Herbes rivers and would be somewhat older than the former, although reliable radiocarbon determinations are scarce. These two traditions seem to be independent both towards each other and towards the Imbonga pottery, in the sense that they cannot be integrated into a single coherent ceramic tradition [85]. If the speakers of the CR languages’ ancestor made pots, it seems most likely that they manufactured Batalimo-Maluba pottery [86]. Not only are some of the present-day CR languages, as the Ngiri languages [87], spoken in between the Ubangi and Congo rivers, but the homeland of their parent language has also been located in the same region [88]. However, at the present state of knowledge, it may be wise not to try to infer too much from too little.

42Does there exist any archaeological correspondence for the pottery related lexical borrowings observed in this study ? Regarding the ICB vocabulary of foreign origin, the available archaeological data do not really offer convincing counterparts, since very little pottery can be shown to have been “imported” into the Inner Congo Basin during the time span in question. Conversely, archaeological matches for the vocabulary diffused from the Inner Congo Basin are somewhat more substantial. A rather more lively “export” of ceramics into the Malebo Pool area of southern DR Congo seems to be attested for the Christian era by shreds from Kinshasa that can unequivocally be attributed to the Inner Congo Basin style groups Mbandaka, Bondongo and Nkile [89]. Of these 3 groups, Bondongo has the largest distribution within the Inner Congo Basin. It also covers the largest time span, namely between 1000 and 1400 AD [90]. This dating indicates that the export of pottery from the Inner Congo Basin considerably predates the rise of the Atlantic trade. As most probable explanations for the foreign attestations of these ceramics, H.-P. Wotzka sees the export of pottery as trade ware by itself and the use of earthen pots as travel utensils during southward transports of rainforest products [91]. This river bound dispersion of Inner Congo Basin ceramics confirms to a certain extent the historical hypotheses advanced to explain the diffusion of the irregular reflexes of the noun °–pògé “cooking pot”.

Conclusions

43The study of material cultures and cultural history is a field of research that currently is in the sphere of interest of several social sciences. Given the overall absence of ancient written traditions in Africa, this common interest has led to a growing inter-disciplinary use of comparative linguistic data. Within the realm of Bantu studies, not only linguists, but also historians and archaeologists have relied on linguistic methods to develop hypotheses on the past cultures of Bantu speaking communities. The Words and Things-method in particular has received the close attention of non-linguists. What is more, historians were the first to exploit the comparative study of cultural vocabularies to the full. In spite of the undeniable merits of these “linguistic historians” for the reconstruction of early African history, the soundness of their historical conclusions risks sometimes to be undermined by methodological flaws in the use of lexical data. Although parting from similar premises, historical linguists tend to tackle cultural vocabularies differently. The present paper has tried to show a linguist’s historical-comparative approach of one specific lexical domain in a geographically and genealogically well-defined set of languages, namely the pottery vocabulary of the Bantu languages of the Inner Congo Basin. The terminology in question is especially interesting, because it is quite distinctive from a larger Bantu perspective. It misses the most archaic Bantu pottery terms, but has proper alternative terms instead. This particular feature is the outcome of a particular historical formation. As other lexical fields, a language’s pottery vocabulary is historically stratified. It consists of several layers of distinct time depth. As shown in this article, these strata can only be suitably uncovered by examining pottery vocabulary globally as a composite lexical field, instead of considering each of its components individually, and by comparing as many Bantu languages as possible. The Comparative Bantu Pottery Vocabulary Database enables such an approach. Although the presentation of linguistic analyses was reduced to a minimum in this paper, a thorough phonological and semantic analysis has proved to be crucial to distinguish inherited from borrowed vocabulary and lexical retentions from innovations and to take into account the historical importance of semantic shifts. Moreover, pottery as such lends itself admirably to interdisciplinary study, since it combines linguistic salience with archaeological visibility and ethnographical prominence. As a result, the historical narrative drawn from my comparative lexical study could be faced with the historical hypotheses inferred from both archaeological and ethnographical data. Although the historical conclusions acquired are far less impressive than the rich and multifaceted narratives built up by certain “linguistic historians” – I assume once more, the present study may add some new pieces to the intricate historical puzzle of human settlement in the African equatorial rainforest.

44Inasmuch as the ancestors of the present-day Bantu speaking communities in the Inner Congo Basin did not shift from a non-Bantu language to a Bantu language [92], they must have known how to manufacture pottery at the moment of their arrival in the vicinity of the great Congo River bend. Since several pottery related terms – for instance a verb meaning “to mould pottery” – could be reconstructed to Proto-Bantu, we can safely assume that the fashioning of pots already was a regular activity before the first Bantu speaking people left their homeland to settle in the forests and savannahs in the south and the east. Both linguistic and archaeological data suggest that this was even the case before a distinct Proto-Bantu speaking community emerged in the borderland between the present-day Cameroon and Nigeria. Hence, the early ancestors of the ICB Bantu speaking peoples must have exported their ancestral ceramic tradition when they were gradually migrating southwards through the rainforests. The progressive loss of Proto-Bantu vocabulary among the Western Bantu languages possibly indicates a steady innovation of this ancestral tradition. On the other hand, the lack of distinctive Western (or Forest) Bantu pottery vocabulary may be due to the fact that the Western (or Forest) Bantu nucleus disintegrated too rapidly to develop a proper terminology. From an archaeological point of view, such a breakdown is reflected in the existence of several early pottery traditions, which turn up in different parts of West-Central Africa, approximately between the first centuries of the 2^nd millennium BC and the beginning of the Christian era [93]. Although a “genealogical” relatedness between these ceramic groups is not excluded, intermediate traditions linking them up were not discovered (yet). The Imbonga pottery, attested in the Inner Congo Basin from the 2^nd half of the 1^st millennium BC on, represents one of these early forest traditions. The total absence of Proto-Bantu pottery vocabulary in the ICB languages suggests that all of it was lost by the time their ancestor language reached the Inner Congo Basin, or at least before the respective CR and CB ancestors became separate languages and began to expand. The pottery fabricated by that time must have completely moved away from the ancestral Bantu tradition, which is also materially evidenced by the Imbonga pottery. Since the two groups of ICB languages do not have pottery terms in common in spite of their presumed genealogical unity at a certain point of time, it seems likely that their speakers became rapidly isolated from each other. Conversely, the fact that their respective parent languages developed a distinctive pottery vocabulary before they spread out and gradually fell apart suggests a relatively long period of isolated evolution. This must have gone together with the development of distinct ceramic traditions. The genealogical non-relatedness of the earliest pottery groups of the region, i.e. Imbonga, Batalimo-Maluba and Pikunda-Munda, seems to confirm this hypothesis. However, while the Imbonga pottery was in all likelihood manufactured by people speaking the parent language of the CB languages, the association of one of the two other early ceramic traditions with the CR parent language is far less obvious. The gradual spreading of the CB ancestor language from the western part of the Inner Congo Basin to the east, roughly between the turn of the Christian era and the 2^nd half of the 2^nd millennium AD, led to the diversification of the CB languages and the dissemination of the inherited pottery vocabulary. In the archaeological record, this development is testified by the progressive branching out of the ancestral Imbonga tradition along the tributaries of the Congo River eastwards, which resulted in the emergence of the different daughter traditions described by H.-P. Wotzka [94]. Although this equatorial tradition as such does not manifest external influences or discontinuities and seems to be entirely indigenous to the Inner Congo Basin, both lexical and archaeological data testify pottery exchanges with neighbouring areas. The clearest loanword traces (°–pògé “cooking pot”) leaving the Inner Congo Basin lead south- and westward to the Kasai region and the Atlantic coast. Even if the Atlantic trade intensified the circulation of goods along the Congo River and its tributaries, archaeological findings signal the export of Inner Congo Basin pottery from as early as the first centuries of the 1^st millennium AD. Lexical borrowings also suggest external influences on Inner Congo Basin pottery making (°–bómba “pottery clay”), but archaeological correspondences are lacking in this respect.