Reconstructing the Niger-Congo verb extension paradigm: What's cognate, copied, or renewed?

It is generally assumed that Proto-Niger-Congo (PNC) had a well-developed paradigm of verb-to-verb derivational suffixes known as verb extensions (Voeltz 1977, Hyman 2007). Based on their effect on valence, specific language studies identify three types of extensions: valence increasing (e.g. causative, applicative, associative, instrumental), valence decreasing (e.g. reciprocal, reflexive, decausative, passive, stative) and valence neutral (e.g. intensive, attenuative, pluractional). Sometimes also implicated in the suffix system are inflectional suffixes marking aspect (e.g. (im)perfectivity)). Despite the assumption of such verb extensions at the PNC level, languages within the vast Niger-Congo family of ca. 1200 languages differ considerably: Some have very full paradigms of verb extensions, e.g. many Atlantic languages (Becher 2000) in the West and nearly all Bantu in the East (Meeussen 1967). Others have a limited subset of the above, only a few traces, or perhaps no verb extensions at all (e.g. much of Mande). This paper is concerned with strategies for determining whether the various reflexes of the causative, applicative etc are cognate, i.e. inherited from PNC, are copied directly or indirectly through external contact—or result from renewal via morphological cycles (Heath 1998). As Bendor (2000: 63) has noted, one factor complicating the tasks of establishing cognates and demonstrating geneological relationships is that “grammatical morphemes tend to be small so that similar forms recur even in unrelated phyla.” A notorious case is the affixal [s] or [t] causative found not only within Niger-Congo but also in neighboring, unrelated Afro-Asiatic and elsewhere in Africa, not to mention other parts of the world. Besides phonological issues (e.g. suffixes are often restricted in place of articulation and are subject to change and loss), verb extensions seem often to merge grammatical functions and undergo semantic shifts. The mere presence of phonetically- or semantically-similar verb extensions may not be sufficient, since head-marking on verbs is all four of Greenberg’s (1963) original African phyla (Dimmendaal 2000: 187-8). A suffix-by-suffix approach will therefore have serious limitations. Thus, besides drawing on form-meaning correspondences between individual extensions, focus in this paper will be on similarities vs. differences in the morphological paradigms which determine whether two or more suffixes may co-occur and, if so, in what order. Starting with Bantu, I briefly exemplify the Causative-Applicative-Reciprocal-Passive (CARP) template (Hyman 2003) and then branch out to other parts of Niger-Congo. While CARP is very stable within Bantu, neither it (nor any other template) has been shown to have generality in another subgroup. Good’s (2005) hypothesis that ordering within a template should correspond with the historical order in which the suffixes came into being is hard to reconcile with differences with and within Atlantic (Hyman 2007), while Güldemann (2011) doubts that such suffixal sequences existed even in Proto-Bantu (but see Hyman 2011 for a reply). This study will address both substantive and methodological issues involved in determining the reconstructed PNC system, including the following questions: (i) Beyond phonetic and functional similarities, which are sometimes compromised, how can it be most compelllingly demonstrated that certain Niger-Congo extensions are cognate? (ii) Among those which fail these arguments, how can we determine if their similarity is due to external copying or internal renewal? (iii) How does the methodology judge similarities with the valence-related suffixes found in Afro-Asiatic, Nilo-Saharan, and Khoi?