No abstract

No abstract

No abstract

This paper provides an overview of the variation in connective constructions throughout the Bantu languages. It will be shown that connectives form a category with fuzzy boundaries, for which no definition in terms of necessary and sufficient conditions can be provided. In order to deal with the variation in connective constructions I will adopt an approach inspired by canonical typology, as developed most notably by Corbett (2007). That is, I will define a canonical connective construction and then describe the formal variation found among connective constructions in terms of departures from the canon along several dimensions. Note that a canonical approach makes no claims whatsoever regarding the status of the canonical type as either being frequent or diachronically primary. The canon is simply a starting point for mapping the variation among related constructions. Section 2 introduces the canonical type. Sections 3 to 7 each discuss departures from the canon along a single dimension, viz. departures from the canonical R2 (section 3), from the canonical connective relator (section 4), from the canonical dependency relation (section 5), from the canonical R1 (section 6), and from the canonical arrangement of constituents (section 7). Depending on how one counts, there are between 300 and 600 Bantu languages, spoken in an area south of a line between Cameroon in the west and Kenya in the east. The internal classification of the Bantu languages is problematic, but by and large the Bantu languages can be subdivided in an eastern, a western and a northwestern group. The northwestern Bantu languages, spoken in Cameroon, Gabon, Equatorial Guinea, parts of the two Congos and the Central African Republic, are closest to what is generally accepted to be the Bantu homeland. Not surprisingly, linguistic fragmentation is highest in the northwest.

Using a very large lexical database and generalized additive modeling, this article reveals that labial-velar (LV) stops are marginal phonemes in many of the languages of Northern Sub-Saharan Africa that have them, and that the languages in which they are not marginal are grouped into three compact zones of high lexical LV frequency. The resulting picture allows us to formulate precise hypotheses about the spread of the Niger-Congo and Central Sudanic languages and about the origins of the linguistic area known as the Sudanic zone or Macro-Sudan belt. It shows that LV stops are a substrate feature that should not be reconstructed into the early stages of the languages that currently have them. We illustrate the implications of our findings for linguistic prehistory with a short discussion of the Bantu expansion. Our data also indirectly confirm the hypothesis that LV stops are more recurrent in expressive parts of the vocabulary, and we argue that this has a common explanation with the well-known fact that they tend to be restricted to stem-initial position in what we call C-emphasis prosody.*