Symposium on the Syntax of the Bantu Noun Phrase

Formal universals are generalizations across natural language grammars that stem from the theory of grammar itself, and concern phenomena that the theory can and cannot in principle express. The present paper concerns the phenomenon of word-order variation, where the same grammatical categories can occur in different orders or permutations, either across or within languages. The question is, which permutations of a set of N grammatical elements are allowed in natural grammars? The question has been extensively studied by Cinque and others in the case of the noun phrase consisting of four elements Demonstrative Numerator Adjective Noun, as in English “These five young lads”.

The present paper proposes on the basis of a recent description by Nchare of Shupamem, a Grassfields Bantu language with very free nominal order, that the permissible permutations are all and only the Separable Permutations, whose number grows as the Large Schroeder Number in N. The paper shows that, on the basis of the NP construction, a high degree of confidence can be assigned to its correctness. The hypothesis also makes a strong prediction about which further order(s) will and will not be attested in future.

The paper concludes by briefly discussing some computational implications of this universal for natural language processing.

My talk will begin with issues in the analysis of Bantu noun classes, addressing first a claim of Taraldsen et al (2018) that each member of a pair 1/2, 3/4 etc. is a distinct gender rather than singular or plural of a single gender as in Carstens (1991), Corbett 1987) a.o. Taraldsen et al’s claim is based on facts of agreement with conjoined subjects. After presenting a reanalysis, I turn to the claim that gender is a [+/-interpretable] feature of the categorizer n (Kramer 2015, Fuchs & Van der Wal 2018), providing evidence from the conjunct agreement facts against equating interpretable content directly with genders themselves.

I then move high in the DP, exploring the syntax of Nguni phrase-final particle kuphela – ‘only’ vis-à-vis the LCA (Kayne 1994) and the Final-Over-Final-Condition (Biberauer et al 2014). I show that kuphela’s associations respect a syntactic topography of focus in Nguni and, like English only, a surface c-command requirement (Tancredi 1990) which low copies cannot fulfill. I show that the facts are incompatible with recent antisymmetric approaches to final particles in other languages such as Biberauer et al (2014), Erlewine (2017), and advocate exemption from the LCA for adjuncts/adjunct particles.

Turning to DP architectural questions, in the hierarchical space between kuphela and N right modifiers of Bantu nouns appear. Given kuphela’s evidence for restricting the scope of the LCA, the analytical space is fairly open; I review arguments for symmetric base-generation options vs. roll-up movements.