The structure of grammaticalization chains Summary
The data looked at in the preceding sections is summarized in (1). Three kinds of situations or stages of pgn use are distinguished there: “0” signals that pgn’s are not normally used on the relevant word category, “I” means that pgn’s are used in some contexts but not in others, while “II” stands for a predominant or generalized use of pgn’s, even if there may be specific contexts where no pgn is found.
(1)
The data summarized in (17) can be described as an implicational scale in a double sense. First, the three word categories differ in the degree to which they accept pgn’s: adjectives are the most resistant to pgn marking, while demonstratives and personal pronouns are much less so. This allows for a generalization of the following kind: If in one of these languages any of the word categories examined is associated with pgn marking (either of Stage I or Stage II) then all word categories to its right are also associated with pgn marking. (There is one partial exception, in that Naro has stage II for the use of pgn’s on demonstratives but an intermediate stage ”I-II” for personal pronouns.)
Second, the three languages also form an implicational scale relative to one another. Hiechware has the lowest rate of pgn marking, Kxoe a higher one, and Naro the highest rate. Now, if any of the three languages has pgn marking on any of the three word categories (either of Stage I or Stage II), then the languages located below it also have pgn marking of at least the same stage.
What was said about Hiechware, Kxoe, and Naro also applies in some way to other Khoe languages. Although the data available are too scanty to allow for any generalizations, it would seem that the remaining Khoe languages can be located somewhere between Hiechware and Naro along the scale of (1). It would seem, for example, that the languages of the Tshwa group occupy a place close to Hiechware along the chain (cf. Köhler 1962; Vossen 1997); while the data on Ani that Rainer Vossen generously made available to me suggest that this language is located close to Naro, exhibiting a fairly generalized use of pgn marking. For example, in Ani pgn’s are not only found on numerals, they have also spread to ‘proper’ adjectives, e.g. áú-n xà¢m-nà (big 3c.pl lion-3c.pl) ‘big lions’ (Rainer Vossen p.c.).
(1) can be said to reflect a grammaticalization chain, but this chain differs from the ones mentioned earlier in the following respect: While the chains discussed by Heine (1992) and Gildea (1997) both involve various parts of language structure, like semantics, syntax, morphology, and morphophonology, (1) concerns essentially only one part, namely morphology. There is an obvious reason for this restriction: no adequate data exist on other parts of the languages in question. Like the chain examined by Gildea (1997), (1) is an instance of a cross-linguistic chain, whose various members are not confined to one language but are found in a cluster of closely related languages
. Each of the three languages examined here can be said to represent a different use along the grammaticalization chain extending from personal pronoun to gender marker: Hiechware represents the initial use pattern, where there is no gender marking except for third person pronouns, Kxoe is suggestive of a more advanced use, where gender marking has become an obligatory part of third person pronouns and an optional part of demonstrative attributes, and Naro represents the most advanced use pattern, where pgn’s have been generalized on demonstratives, have spread to second and first person pronouns, and have become an optional part of numeral attributes.
(1) has all the formal properties of grammaticalization chains:
a. None of the three attributes of pgn use distinguished in (1) is shared by all languages.
b. No language combines all attributes.
c. Each language has at least one attribute in common with some other language.
d. The endpoints of the chain, that is, the languages Hiechware and Naro, have no attributes in common.
e. Languages next to one another on the chain have more attrib utes in common than non-adjacent languages. Thus, Hiechware and Kxoe, or Kxoe and Naro, share more attributes than Hiech ware and Naro.
f. The peripheral languages Hiechware and Naro have the lowest number of attributes in common, whereas the central language Kxoe has the highest number of shared attributes
