The Puzzle of Pan-African Roots

Table of Contents

• 1. Introduction … 74
• 2. Data tables … 75
• 3. Conclusions … 91
• Abbreviations … 92
• References … 94

Tables

• Table 1: Acronyms for language-specific sources … 92

Acronyms and Conventions

1. Introduction

African language phyla are generally ‘well-behaved’ in the sense that it is relatively easy to assign a given language to a phylum, based on its core lexicon and morphology. Africa has a relatively small number of isolates (Blench 2017), given that it is the centre of dispersal for modern humans. Until recently, it was deemed to have four phyla, Khoisan being counted as a unit. More recently, Khoisan has been split into three independent phyla. Nilo-Saharan remains the most problematic phylum and not all researchers accept either a) that it is a unit or b) that all purported branches actually fit together (Blench 2026). On the other hand, there has been a persistent thread of argument for uniting Nilo-Saharan and Niger-Congo into a single macrophylum, Niger-Saharan which would join the two families (see Blench 1995 and references up to that date). Westermann (1911) in his first version of the Sudan Sprachen put together Nilotic and languages we would regard as Niger-Congo. However, he later seems to have dropped this idea, since he makes no reference to it in his later Sudansprachen (Westermann 1927).

In the last few decades there have been various further attempts to bring together Niger-Congo and Nilo-Saharan, including Gregersen (1972), Creissels (1981), Boyd (1999). Hall & Hall (1974) were the first authors to draw attention to the striking similarities of the ATR vowel systems in Niger-Congo and Nilo-Saharan. Unfortunately, all this is rather inconclusive, despite the massive expansion of data over this period. Personally, although my earlier papers argued for a genetic connection between the two phyla, I now consider the similarities to be the outcome of early borrowing. One reason for that is that the homelands of the two phyla seem to be at opposite sides of Africa, Nilo-Saharan in the Ethio-Sudan borderlands and Niger-Congo in the west of West Africa.

The existence of macrophyla, such as Nostratic or Transeurasian, will always be controversial, in part because it is difficult to disentangle ancient loanwords from true cognates beyond a certain time-depth. So it is with Niger-Saharan; there are certainly common lexical items and strong evidence for typological commonalties, such as ATR vowels. When I wrote my original papers these seemed to me convincing evidence for genetic affiliation. However, I have subsequently come to see that many of the lexical items are also attested in Afroasiatic and more rarely in Khoisan and even the isolates. If so, then some or all of these similarities may be due to ancient loans, although when and where this transmission took place will remain uncertain.

In a first approach to this topic, I focused on crabs, turtles and frogs, at first sight rather unlikely candidates for Pan-Africanisms (Blench 1997). The argument was that, although these are marginal foods at present, this type of aquatic fauna was highly salient at the time when foraging groups were dispersing, and when human beings had yet to learn to fish in open water. This may have something going for it, but subsequent research has revealed some widespread roots which also have Pan-African distributions, but which seem largely arbitrary. Why, for example, should ‘knee’ be so widely attested and not ‘arm’ or ‘leg’?

The purpose of this paper is simply to compile datasets for the most common Pan-African roots and to see what conclusions can be drawn. The data for crabs, turtles and frogs, appropriately updated, is reprised here. §2 consists of the data tables, and the conclusions in §3 ask what emerges from this exercise.

2. Data tables

The data tables have been gradually compiled over many years, as part of a broader exercise to identify common Niger-Congo roots. The attestations are divided by phylum, with isolates cited first. The sources are abbreviated, and the compressed references are keyed in the references at the end. Major proposals for Niger-Congo reconstruction, namely Westermann (1927) [Proto-Western Sudanic] and Mukarovsky (1976-77) [Proto-Western Nigritic] are given although I would not now propose the forms they present. Among other things, they were unaware of the Dogon family of languages, which plausibly provide evidence for the earliest stages of Niger-Congo.

1. come I #ɓwoN-

Commentary: Not attested in Afroasiatic outside Chadic and thus almost certainly a borrowing. Dimmendal (1988:35) notes that the irregular plural *pɔ must be reconstructed to Proto-Nilotic and forms with initial p- appear elsewhere, hinting at a still greater time-depth. The Kru form is unique and Sɛmɛ is isolated, so this may be loan, as is the unique citation in Ubangian. The Bantu form makes it only doubtfully cognate. Westermann (1927: 209) noted that this word frequently shows up as a future auxiliary in Niger-Congo languages. Palatalisation is scattered throughout Niger-Congo but nowhere forms a consistent pattern. The #ba form is so persistent in Niger-Congo that it seems likely that is where it originated. The appearance in Chadic and only two branches of Nilo-Saharan suggest ancient borrowing. There are also parallels in Indo-European, e.g. French venir.

Refs: W. 209; Ar64: 36; M. 435; E. 563;

2. to split, cut, break #keri

Commentary: This root has almost certainly been loaned very widely in Africa. The occurrence in Chadic is most likely a localised loan, but in Cushitic and Omotic these forms are widespread (see examples under k’er ‘split’ and kaal-ta ‘axe’ in Lamberti & Sottile 1997:411, 435). This lexeme has been proposed as a ‘world etymology’, and the cognate set would presumably then including English ‘cut’ (see Bender 1997:122 for more Nilo-Saharan examples). Temiar (Aslian, Austroasiatic) has cɛr ‘to cut’. Although the first CV seems to be cognate throughout, the diversity of C2 may be explained by reduplication and compounding. The common Manenguba forms probably are cognate if k→s, but the original initial is conserved at least in Akɔɔse as an alternative. The labial-velar is only preserved in Mbe, hence its reconstruction remains doubtful, although the labialisation in Akɔɔse also suggests its presence.

Refs: W. 227+243; Ar64. 43; M. 75+202+203+256+22; G. 280+304+1029+1033+1045; Gr:80; G:97,135,154; B:122, 133; Boyd (1994:106)

3. knee I #kulu

Commentary: A preliminary version of this dataset appears in Blench (1997) and this root appears to have extra-African cognates. Gregersen (1972) treats these as two distinct sets for ‘leg’ and ‘knee’ but they are probably to be put together and the more doubtful cognates discarded. A rather different form, *BU(N)KA is proposed as a ‘world etymology’ in Bengtson & Ruhlen (1994). Dimmendaal (1988:46) reconstructs Proto-Nilotic *kɛlɪ for leg, a semantic change that also took place in Bantoid. BLR3 says the alternation between dúɪ́/núɪ́ ‘est à étudier’, but if this reconstruction is correct, the two are simply different erosional results of an original –nd-. Why this word should be so similar in so many parts of the continent and beyond is hard to explain. English ‘knee’ might be added.

Bender (1997:133) pursues linkages that include a purported PNC root *khon for ‘knee’ and brings in Mende kon ‘head’ because the ‘knee as head of the leg’1. This analysis is not used here. Within Niger-Congo, Biseni is isolated within Ijoid and may thus be a loan from a neighbouring language. East Benue-Congo seems to have a distinctive shift, the loss of the ku- prefix and the development of -ru(ŋ) as the stem. This is retained in Bantoid as least as late as Ejagham. There are two potential cognates in Bantu, the more widespread #-kónò which generally means ‘forearm’ and the less common #–koto/godo which is knee directly. These do not resemble Bantoid, and the Bantu forms may be cognate with other branches of Niger-Congo. It is probable that #-kónò is just a chance resemblance.

Refs: G.:101,123, G. 884; M. 163+304; B:133; B81:.261, Gr.:82,84,

4. knee II #-duŋ-

Commentary: The cognates of Bullom ‘knee’ in South Atlantic show a wide semantic range, for example, Temne ‘leg’ (a-lǝŋk) and ‘arm’ a-loŋk, Baga Koba a-raŋk ‘thigh’. This root is widespread in West Chadic and the #-rum element may have been loaned into Plateau and its prefix re-analysed. Pɛrɛ looks like a composite of the root for ‘knee’ with the root for ‘leg’.

Refs: W. 252; M 109; G.722+729

5. tortoise, turtle #kuru

In the printed table the Attestation column is segmented; the segments are joined here with spaces, preserving each segment exactly.

Commentary: An early version of this table was presented in Blench (1997) where it was argued that the importance of turtles and tortoises in the gathering phase of human history had made this word particularly salient. The diversity of forms attested may reflect the fact that different species may have compound names (see the Kanuri and Aiki forms). Extra-African cognates have been recorded;

The diversity of the forms attested may reflect the fact that different species may have compound names (see the Kanuri and Aiki forms). Turtle/tortoise is the only item of fauna to be widely attested in Niger-Saharan. Greenberg cites parallels from Kordofanian and also Keiga, now classified as Nilo-Saharan. Hoffmann (1970:15–16) points out that this word was also borrowed into Chadic from Benue-Congo, occurring throughout West Chadic and sporadically in Central Chadic. Ironically, he concluded that this might be evidence that these attestations were unconnected. Since Afroasiatic cognates seem only to be recorded in languages near to Niger-Saharan, it is reasonable to conclude that they are ancient loans.

Refs: M. 306; G. 321+325+1259+1260; C.:321, Gr.:88, G.:159

6. crab #kala

Commentary: African crabs are highly speciated but can be divided into three categories; marine, freshwater and land crabs. The taxonomy of freshwater crabs is given in Cumberlidge (1999). Material on land crabs is not easily available but they are widespread throughout the continent and are frequently culturally important because of their role in divination systems. Crab divination is reviewed in Blench and Zeitlyn (1990) which shows that the words for ‘spider’ and ‘crab’ are etymologically interconnected in the Bantu borderland because of their comparable significance in divination systems.

There appear to be two intertwined roots, one with r/l/d in C₂ position, the second, less common, with –mb-, in the same slot. All those with a lateral have a copy central vowel, generally –a, whereas the final vowel in type II is much more variable. The table presents a sample of crab names. Mukarovsky (1976:144) adds further Niger-Congo cognates. The Niger-Congo roots are discussed in Williamson & Shimizu (1968:92).

‘Crab’ also has widespread Eurasian cognates (Blench 1997). The table below sets out some attestations and reconstructions that have been proposed for ‘crab’ in Old World language phyla.

Commentary: Clues to the processes at work in African languages can be gained by comparison with Indo-European associations of words for ‘crab’. Indo-European has a root #kar- meaning ‘hard’, which has a complex association with words for ‘crab’2. Latin cancer and Greek karkinos are both derived from reduplications of the original root, the image apparently being the hardness of the crab’s shell. A similar association also exists in Niger-Congo; Westermann (1927:240) reconstructs #kual- for ‘to be hard’ in Proto-Western Sudanic, and #-kal- for ‘crab’.

Refs: W. 230; M. 193; G. 981;

7. skin, hide #kulu

Commentary: A pan-Africanism with apparently related forms in all four major phyla. A preliminary version of this dataset appears in Blench (1997). Greenberg (1963:21) initially identified this root for Niger-Congo and though he quotes Krongo (p. 157), his form does not correspond to that in Reh (1985), which is not evidently cognate. Creissels (1981:316) points out the Songhay cognate and adds further citations for Niger-Congo. Bender (1997:129) gives further examples for Nilo-Saharan, although he includes ‘basket’ in his semantic set. Other commentators include ‘bark’, for example Uduk (Eastern Sudanic) khur ‘bark’. Westermann (1927: 220, 237) separates out two forms which are probably part of a single root. Most Kru languages have reduced to CV, but the nasalisation preserved in Kuwaa provides evidence for the original lateral in C₂ position (Marchese 1983). Nasals are attested in Kru, Gur and Adamawa-Ubangian, which argues for an innovation perhaps at the level of Atlantic-Congo.

Refs: W. 220 + 237; M. 302; G. 392; G.:21, Gr.:84, B:129, E:491

8. fat, oil, grease #morV

Commentary: The assumption here is that the m- is originally a class affix, signifying liquids or mass nouns which have become fused to the stem. Forms for mass nouns with m- affixes correspond to Kordofanian ŋ- classes in other branches of Niger-Congo (Schadeberg 1989). The ‘original’ form (if that has a meaning in this context) would then have been something like #muri or #ŋuri. The second set of forms appears to be a rather strained correspondence, but there is evidence the ŋ- and m- do correspond between Kordofanian and the rest of Niger-Congo (Williamson 1989a). Nonetheless, it is possible there are two roots here. Although this form seems well established enough in Plateau and Kainji to imagine that its source is there and gave rise to Hausa mai, the Hausa form may well have been borrowed back into some languages giving rise to irregularities.

Refs: D. 40, W. 257; Ehret p. 312

9. fire #(dw)isi

Commentary: An old areal word with dispersed reflexes, it is possible some forms are just chance resemblances. Not attested in Khoisan. Only recorded in some subgroups of West Chadic and thus probably a local loanword. If the Akkadian form is genuinely related, then it is tempting to assume this an old Afroasiatic root loaned into Nilo-Saharan and thence into eastern Niger-Congo. Bender (1992:43) reconstructs Proto-Central Sudanic #co, but as an areal loan, so reconstruction is probably not a meaningful exercise. The Niger-Congo attestations are very scattered and these could be chance or borrowing.

Refs: Bender (1991:5)

10. #-si dog

Commentary: Although originally cited by Greenberg (1963:120), more complete evidence was marshalled by Bender (1981:258) with attestations in Fur, Sudanic, Kordofanian and possibly Ari [Omotic]. Not attested in Khoesan, Mande and most branches of Afro-Asiatic, which have variants of kVr/n-. This root is extremely widespread in Central Africa and yet does not form a convincing pattern. In Central Sudanic, for example, it is attested in almost every language (Bender 1992:40, 48). Bender separates the roots with initial b-, but it is likely that the two forms go together.

Refs: G:120

11. to fly, jump #pur-

Commentary: The meanings of ‘fly’ and ‘jump’ are regularly intertwined throughout African language phyla. Bender (1997:121) lists more Nilo-Saharan cognates in typically elliptical style. This root presents an intriguing methodological problem. The citations are so similar that they must be related in some way. Is this simply a case of ideophonic convergence, where similar languages come to similar conclusions about the sound of beating wings? Not all languages do this, as many datasets for ‘fly’ are omitted on the grounds of non-cognacy. Similarly, the semantic link with ‘jump’ which seems to be typical for Africa is not generally found elsewhere, where the common association appears to be ‘to flee’. Swadesh (1971) derived a similar form ideophonically as a world gloss, although he spreads the semantic net wider than here. This etymology is reprised in Bengtson & Ruhlen (1994) with a proposed proto-form *par. Western Kwa seems to have undergone a distinctive reduction to *fe.

Refs: Gr.:83, D.:42, W.:275, B: 121, E: 382, Ehret (1987:26)

3. Conclusions

The paper highlights eleven Pan-African roots, which are attested in either three or four phyla and sometimes additionally in isolates. The majority are nouns, although three verbs are included. The most surprising aspect of this is that three words have an almost identical form, #kuru, and a fourth, ‘crab’, #kala, is extremely similar.

There are two possible explanations; either pervasive sound symbolism or a chain of borrowing. If sound symbolism has driven convergence, it has affected selected concepts, rather like the word families in Southeast Asia or the templatic morphology in Semitic. For some roots, such as ‘fly’ or ‘crab’ their presence outside Africa must suggest phonaesthetic explanations. Alternatively, borrowing is responsible for these similarities at a very early period of contact. It is clear there are few roots which are well attested in Khoesan. These may have been borrowed into Nilo-Saharan at the period when Khoesan was further north and in contact with Nilo-Saharan and Afroasiatic. Similarly, many of the attestations in Afroasiatic look like regional borrowings. They are found in Chadic, Omotic and Cushitic and not in Berber-Egyptian-Semitic. Niger-Congo and Nilo-Saharan seem to have been contact at an early period, which may explain shared lexicon, but also typological features such as ATR vowel harmony, plural verbs and possibly alternating prefixes, such as are found in East Sudanic and are common to most branches of Niger-Congo.

Under no circumstances are these resemblances chance; the similarities are too great, and the comparison with the diverse lexicon for closely allied concepts (compare ‘leg’ and ‘knee’) argues that either historical processes are at work, or sound symbolism has created convergence. Further work may provide more coherent explanations.

Abbreviations

1. Principles of deriving acronyms for sources;

1. Single-authored works usually take the first two letters of the surname and add the last two numbers of the year of publication. Thus Childs (2000) becomes Ch00. However, where this might lead to ambiguity, a third letter of the author’s name is added to discriminate between two authors. Thus;

Cr81     Creissels (1981)

but;

Chr33     Christaller (1933)

2. Two- and three-authored works usually take the initial letter of each author’s surname and combine them into a sequece of capital letters, followed by the last two numbers of the year of publication. Thus Jungraithmayr & Ibriszimow (1995) becomes JI95.

3. Single-authored major comparative sources are identified by a single capital letter. These are;

B     Bender (1997)
G     Greenberg (1967-1971)
M     Mukarovsky (1976, 1977)
W     Westermann (1927)

Other publications by these authors are abbreviated in the standard way.

4. Multiply-authored major comparative sources, particularly those in series have slightly idiosyncratic abbreviations derived from the initial letters of the main words in the title.

5. No distinction between electronic and paper formats is made for unpublished works. In view of the difficulties of assigning a year to these materials, they are numbered in sequence, following the abbreviation of the author(s) name(s) according to the principles above.

2. Specific languages: classification and source(s)

Table 1. Acronyms for language-specific sources