Remerciez-le!

Remerciez @Admin pour avoir partagé cet document gratuitement, de la manière la plus simple, en partageant sur les réseaux sociaux.

69 HOW TO BECOME A ‘KWA’ VERB Larry M. Hyman University of California, Berkeley

69 HOW TO BECOME A ‘KWA’ VERB Larry M. Hyman University of California, Berkeley hyman@socrates.berkeley.edu Within the Niger-Congo family, some languages have very complex, agglutinative verb structure whose head-marking supports multiple objects and unmarked adjuncts. Others have very simplex (often monosyllabic) verb structure, little or no morphology, and an analytical syntax whereby objects and adjuncts are marked by adpositions and/or serial verbs. Assuming the former to be original, in this paper I distinguish three separate changes in the latter that seem, roughly, to correlate: (i) Syntactically, there is a gradual change from multiple to single object marking. (ii) Morphologically, there is the gradual loss of verb suffixes which mark ar- gument structure, e.g. causatives, applicatives, passives, reciprocals. (iii) Phonologically, grad- ual size (and other) prosodic limitations are introduced on the verb. It is this last phenomenon which has received the least amount of attention, but which, I argue, is an important factor in determining the morphological and syntactic developments as well. While I will show that the syntactic changes occurred first, there is an intimate interdependency between the three affected parts of the grammar which has not been sufficiently appreciated. Au sein de la famille Niger-Congo certaines langues ont une structure verbale très com- plexe et agglutinante, dont les marqueurs de tête supportent des objets multiples et des adjoints non-marqués. D’autres ont un système verbal très simple, souvent monosyllabique, n’ayant au- cune morphologie, ou peu, et une syntaxe analytique où les objets et les adjoints sont marqués par des adpositions et/ou des verbes en série. Présumant que les systèmes complexes sont ori- ginaux, je distingue ici trois catégories de changements dans les systèmes simples qui semblent, grosso modo, être en corrélation: (i) Syntaxiquement il s’effectue un changement graduel d’un état où les marques d’objet sont multiples vers un état où elles sont simples. (ii) Morphologi- quement il y a une perte graduelle des suffixes verbaux qui marquent la structure des argu- ments, p.ex. causatifs, applicatifs, passifs, réciproques. (iii) Phonologiquement certaines limitations prosodiques graduelles d’envergure (et autres) viennent toucher le verbe. C’est ce dernier phénomène qui a jusqu’ici reçu le moins d’attention, mais que je propose être un fac- teur important qui influence aussi les développements morphologiques et syntaxiques. Tout en montrant que les changements syntaxiques ont été les premiers, je souligne qu’il existe entre ces trois parties de la grammaire une interdépendance dont on n’a jamais encore suffisamment tenu compte. The present paper is concerned with the historical development of a verb (phrase) typology that is well documented in several branches of Niger-Congo. It finds its most explicit statement in Westermann and Bryan’s (1952:91, 93) characterization in (1) of the properties they expect of a language in the (Old) Kwa subgroup.1 (1) a. “Most Roots (Verb or Noun) are monosyllabic, consisting in CV.” b. “There are no morphological Verb Classes.” c. “There are no verbal derivatives.” d. “There is no Passive Voice.” e. “The Verb Root is invariable.” As a typical case, Westermann (1930:182–183) characterizes the (Western) Kwa lan- guage, Ewe, as in (2). (2) a. “The great majority [of verbs] is monosyllabic.” b. “The disyllabic verbs are either (1) Reduplications of monosyllabic verbs… [or] (2) Compounds of two monosyllabic verbs”, e.g., 1 The internal subgrouping of Niger-Congo languages has undergone considerable modification since Greenberg (1963). See especially Williamson (1989), as well as Williamson (1985), which has inspired my title and this research in general. In addition to the Leipzig meeting, this paper was also presented as a UC Berkeley Linguistics Colloquium and, in part, during the Workshop on Benue-Congo at Berkeley, March 26–27, 2001. 70 Journal of West African Languages XXX.2 (2004) (i) kaka to scatter < ka to scatter bébe to uncover < bé to uncover dzudz¯ to cease < dz¯ to wait (ii) fanyã to knead < fa to knead + nyã to knead The near-invariant, most monosyllabic nature of verb stems can be readily exemplified from other Old Kwa languages such as Yoruba and Nupe: e.g. “All pure verbs in Nupe were probably monosyllabic originally, the other kinds being formed by adding to the verb, a noun, or adverb, or preposition” (Banfield and Macintyre 1915:42). This stands in stark contrast to the structure of the verb stem in Bantu languages, e.g. Yao P.21 (Ngunga 2000), illustrated with the inflectional final vowel -a in (3). (3) a. taam-a sit b. taam-ik-a -ik- (impositive) seat (put in seated position) c. taam-uk-ul-a -ul- (reversive tr.) unseat d. taam-uk-ul-igw-a -igw- (passive) be unseated e. taam-uk-ul-igw-aasy-a -aasy- (causative) cause to be unseated f. taam-uk-ul-igw-aasy-an-a -an- (reciprocal) cause each other to be unseated g. taam-uk-ul-igw-aasy-an-il-a -il- (applicative) cause each other to be unseated for/at As seen, verb stems can be quite long and involve multiple suffixes. Similarly, as seen in (4), Atlantic languages may also have well-developed verbal derivatives occurring in sequence. (4) a. ’o-ma´´-ii yolnde He shut the door. b. ’o-ma´´-it-ii yolnde -t- (reversive) He opened the door. c. ’o-ma´´-it-id-ii jol¼e fuu -d- (comprehensive) He opened all the doors. d. ’o-ma´´-it-id-an-ii =mo jol¼e fuu -an- (dative) He opened all the doors for him. As in Yao, the forms in (4) from Fula (Arnott 1970:367) involve multiple suffixation, the last of which is the general past active suffix -ii. In both Bantu and Atlantic, verb suffixes may have one or more of the three functions in (5) (cf. Peterson 1999, Trithart 1983). (5) a. grammatical function : marking/licensing of argument structure. b. semantic function : marking/licensing of thematic roles, verb semantics, tense/aspect. HYMAN: How to become a ‘Kwa’ verb 71 c. pragmatic function : marking/licensing of topicality/discourse prominence. Comparing Ewe with either Yao or Fula, the natural question to ask is: Why are these related languages so different? To answer this question, we assume, following Givón (1975) and Voeltz (1977), first, that the above Bantu/Atlantic verb-stem struc- ture represents the Proto-Niger-Congo situation; and second, that Niger-Congo lan- guages such as Ewe, Nupe, Yoruba, etc. which conform to the Westermann and Bryan characterization of Kwa in (1) have modified the proto system—most likely in an areal fashion. Evidence in support of this view comes from the fact that similar exten- sions, sometimes cognate with Proto-Bantu, are found, either productively or in relic forms, in many sub-branches of Niger-Congo (Voeltz 1977). Examples are presented in (6). (6) a. Causative -ªsª in Degema [Edoid; Benue-Congo] (Kari 1995:158) t² be burnt → t²-ªsª cause to be burnt tul reach → tul-ese cause to reach kir return → kir-ese cause to return b. Benefactive -rV-/-lI in Igbo [Igboid; Benue-Congo] (Emenanjo 1978, änÔkawa 1999) zÀ buy → zÀ-r² buy for bè cut → bè-re cut for zà sweep → zà-ra sweep for c. Benefactive -d- in Zande [Ubangi] (Boyd 1995:19) kp„ mourir →kp„-d- mourir pour quelqu’un (donc ‘souffrir, se sacrifier’) na pleuvoir →na-d- pleuvoir pour quelqu’un (donc ‘mouiller (comme la pluie)’) gbe tirer →gbe-d- tirer pour soi, pour garder d. Reversive -rV in Banda-Linda [Ubangi] (Cloirec-Heiss 1986:129) vª mettre au chaud pour faire mûrir, couver →v¤r¤ déployer (ailes), ouvrir æe bouillonner, déborder, éructer →æèrƒ se dégonfler, rendre le dernier soupir, enfoncer, descendre e. Multifunctional valence marker -ª in Krahn [Kru] (Bing and Duitsman 1993:99) mu go →mu-ª make go (causative) dbà kill →dbà-ª kill for (applicative) dbà-ª kill with (instrumental) As seen in the last set of examples in (6e), previously distinct extensions can merge and, in this case, have a generalized licensing function of arguments. Given such cognate forms, it is hard not to agree with Voeltz that such a system of verb extensions should be reconstructed at the Proto-Niger-Congo stage. Assuming that such a reconstruction is motivated, the following questions naturally arise: How does the proposed Proto-Niger-Congo structure become a ‘Kwa’ verb? Through what stages does it pass? Why? To answer these questions, a double research strategy is proposed. First, we 72 Journal of West African Languages XXX.2 (2004) can look at comparative Niger-Congo, especially those languages which are at different stages of modifying the original situation. Second, we can examine the fine details in Bantu languages (and ultimately Atlantic, etc.) that maintain the reconstructed verb struc- ture, but with subtle variations. That is, we can focus on the visibly evolved Kwa-type systems themselves or on the ‘seeds for change’ that exist even in languages which appear to be quite conservative. In neither case is the present paper a comprehensive survey. Rather, I draw on materials with which I have greatest familiarity—and which I believe are representa- tive of the phenomena that need to be considered in studying the drift from a Bantu- like to a Kwa-like verb stem. As summarized in (7), this drift potentially involves a realignment in all parts of the grammar. (7) a. Syntax : synthetic > analytic (head-marking > dependent or no marking) b. Morphology : agglutinative > isolating (suffixation > marking by syntactic elements) c. Phonology : free > restricted (unbounded > bounded) To restate, I am assuming that the starting point in Proto-Niger-Congo is one where grammatical relations are marked on, and hence licensed by, the verb; multiple NP uploads/Ingenierie_Lourd/ 2004-hyman-kwa 1 .pdf