Roger Blench: Niger-Congo reconstruction

Niger-Congo Reconstruction

Title	Status
Westermann	Unpublished
Niger-Congo map	Unpublished
Niger-Congo: an alternative view	Unpublished
Did early Niger-Congo have triconsonantal roots?	Unpublished
Williamson, Kay 1992. Some Bantu roots in a wider context. Komparative Afrikanistik: Sprach-, geschichts- und literaturwissenschaftliche Aufsatze zu Ehren von Hans G. Mukarovsky anlässlich seines 70. Geburtstags, edited by E. Ebermann, E.R. Sommerauer und K.E. Thomanek, 387-403.	Published

Commentary

Niger-Congo: an alternative view

There are a variety of views concerning the classification and reconstruction of Niger-Congo the world’s largest phylum. It is certainly a matter for regret that there are no regular conferences for Niger-Congo, but its size and complexity makes this difficult to manage. The languages that fall within Niger-Congo remain a matter for debate. There seem to be a number of major methodological points to be made about Niger-Congo, as well as issues concerning the coherence of various branches.

1. You cannot reconstruct a language phylum unless you have good arguments about which language families it includes. The most striking case is Altaic, where one group of scholars produces thousands of reconstructed forms, and another denies that the major branches are even related. The most extreme case for Niger-Congo is Gerrit Dimmendaal’s 2011 book, which rejects numerous established branches and treats them as ‘independent’. No evidence is offered for this so the case is hard to assess. But even more positive assessments may have trouble with Dogon (see below).

2. You also can’t reconstruct a language phylum unless you can also explain the position of unclassifiable or independent branches. Good examples of this are Bɛrɛ, Bijogo, Mpra which look Niger-Congo but which are difficult to assign to known branches.

3. You also can’t reconstruct a language phylum unless you have a convincing model of its internal structure. In the case of Niger-Congo, this is of particular importance. There is an extensive literature discussing the likely noun-classes, verbal extensions, word order and so on of Niger-Congo. But there are, for example, a number of branches where there is no evidence for noun-classes, such as Dogon, Ịjọ, Mande and Kwaalak-Domurik. Is this because they have been lost? If so, it is incumbent on the proponents to demonstrate how this occurred. Dogon, for example, has no labial-velars and no evidence for ATR vowels, which are present almost everywhere else. If they were indeed lost, the proponents of their presence in proto-Niger-Congo phonology should be able to demonstrate sound-correspondences indicating their former existence in proto-Dogon. Needless to say, crucial analytic studies of this type do not exist.

4. If you conflate areal groupings with genetic units, notably erecting ‘Bantoid’ into some sort of family, then again, it will be difficult to make sense of the data.

With these caveats in mind, the following presents a review of existing or proposed groupings as a research tool; without sorting out these rather basic questions, it is difficult to imagine a credible reconstruction of proto-Niger-Congo. Table 1 presents extreme and rather simplified versions of basic positions.

Table 1. Opposing views in historical reconstruction
Reconstructed proto-forms/phonemes should look like real languages spoken today	Reconstructed proto-forms/phonemes should satisfy neatness requirement (all exceptions explained away)
Classifications are purely linguistic and cannot be related to greater time-depths. Bantu is as ‘old’ as other branches because all are part of Niger-Congo	Languages spoken in real time by real people and thus can and should be related to archaeology, genetics
Only cite data that supports reconstructions	Make all data available including material that fails the test. ‘My cognates are your lookalikes’
Lexicostatistics is helpful in understanding genetic relations	Lexicostatistics is useless/ a first approximation
Glottochronology is a valuable tool	Glottochronology is useless
Trees are useless	Trees are helpful
Ordering is ‘just about labels’	Ordering is crucial to methods and results in reconstruction
Lexical and morpheme compilation useless in itself for lack of regular correspondences and inability to identify loanwords	Lexical and morpheme compilation primary tool, despite problems

Table 2 evaluates the evidence for major claimed branches of Niger-Congo;

Table 2. Evidence for major claimed branches of Niger-Congo
Branch	Commentary
Dogon	Dogon is certainly a well-founded and coherent group. But it has no characteristic Niger-Congo features (noun-classes, verbal extensions, labial-velars) and very few lexical cognates. It could equally well be an independent language family.
Ịjọid	The Ịjọ languages constitute a well-founded group, but the membership of Defaka (constituting Ijoid) remains problematic. Defaka has numerous external cognates and might be an isolate or independent branch of Niger-Congo which has come under Ịjọ influence.
Kordofanian	Not a group. See entries below.
Heiban-Talodi	Usually considered a group
Rashad-Kwaalak	Perhaps a group, but the absence of noun-classes in Kwaalak and part of Rashad remains problematic
Tegem-Amira	Tegem [Lafofa] has similarities to Talodi, but a highly divergent lexicon. Provisionally considered and independent branch
Mande	A coherent group
North Atlantic	No strong argument in print for coherence of all members, but likely.
Mel = South Atlantic	A coherent group
Kru	A coherent group
Senufic	A coherent group, previously treated as part of Gur, but no good argument for this.
Gur	Fairly coherent, but the argument that some western Adamawa languages are closer to Gur than those further east is apparently well-founded.
Adamawa	No evidence that all claimed members really form a genetic group. Fali and Daka have been expunged. Much hangs on a typological feature, noun-class suffixes, which must be argued as disappeared in some branches.
Ubangian	Not a group and no evidence yet presented for a particular relation with Adamawa, although geographical proximity makes this likely. Gbaya is either Adamawa-linked or an independent branch of Niger-Congo
Kwa	No argument in print for coherence of all members. See Volta-Niger for discussion of Gbe
Volta-Niger	Previous part-identity as ‘Eastern Kwa’ and ‘Western Benue-Congo’. Proposal by present author to join Yoruboid et al. with Gbe.
Benue-Congo	If treated as the noun-class languages east and north of the Niger, a likely group, but no argument in print for its coherence. Bendi is not Cross River.
Bantoid	Definitely not a group. Present author has argued for a ‘Northern Bantoid’ consisting of Dakoid, Mambiloid and Tikar. The remaining small groups (Grassfields, Tivoid etc.) are independent branches within the Benue-Congo / Bantu borderland.
Bantu	Definitely not a group. This may seem surprising in the light of the published claims to the contrary, but the argument from comparative linguistics which links the highly diverse languages of zone A to a genuine reconstruction is non-existent. Most claimed proto-Bantu is either confined to particular subgroups, or is widely attested outside Bantu proper.

Apart from these groups, there are a number of languages which look Niger-Congo but which cannot easily assigned to any definite genetic group. Of course, if, for example, Kwa or Ubangian falls apart, then the number will be much larger. Table 3 shows a summary of the minor claimed branches of Niger-Congo;

Table 3. Minor claimed branches of Niger-Congo
Branch	Commentary
Mpra	Has western Kwa cognates, and may either be an isolate with borrowings or a highly divergent branch of Kwa. Dead, so no more evidence can be collected.
Bɛrɛ	Has Mande and Kru borrowings, but is not affiliated to either
Ega	Possibly Kwa, but few cognates
Fali	Formerly assigned to Adamawa, evidence is weak.
Bijogo	Formerly assigned to Atlantic on geographical grounds. Hard to place
Ikaan	Noun-classes and concord make it look Benue-Congo, but evidence is weak.
Bangi Me	Formerly assigned to Niger-Congo but improved evidence makes the case weaker.

In the light of this, any tree for Niger-Congo is more a tool for thinking than a design set in stone. Figure 1 presents my most recent version of the Niger-Congo ‘tree’ with all the usual reservations.

Figure 1. Niger-Congo restructured

Figure 2 presents a revised subclassification of Benue-Congo languages, intended to clarify the point about the non-coherence of Bantoid.

Figure 2. Revised subclassification of Benue-Congo languages

Conclusion

Historical linguistics should proceed by evidence-based approaches, not assertion. For all its critics, the comparative method is the only one which has long-term traceability. It is true, that the data now available is vast and moreover is available on the web.