262 WILLIAM H. JACOBSEN, JR.
cation of tense. He moreover drew analogies to several African languages:
(Nilotic) Maasai ("N-tense"), (Bantu) Tswana, Herero, Duala, and Swahili
("narrative tense"), (Kwa) Igbo, and (Chadic) Hausa, in all of which the
full specifications of tense/mood are given only on the first verb of the
sequence. These languages represent three unrelated phyla, and Kiparsky
noted that this is evidently an areal feature of African languages.
Weimers (1973:364-366) gives instances of the same thing as what he
calls the "consecutive" construction, illustrated from three Niger-Congo
languages, again Swahili and Igbo, plus (Mande) Kpelle, indicating that "a
great many [African] languages have a special verbal construction used to
refer to actions after the first in a sequential series". He notes also a "simul
taneous" construction in Kpelle, and he observes that there are no conjunc
tions used in these patterns. Similarly, Dahl (1985:112-114) notes special
narrative forms in Niger-Congo languages in his sample: (Bantu) Kikuyu,
Sotho, Zulu, Setswana, and (Gur) Karaboro, and also brings the Biblical
Hebrew construction into comparison. (He defines narrative discourse
somewhat narrowly, so that it equates to consecutive events related in the
order in which they occurred, rather than as a literary genre.) He
generalizes that narrative stretches typically begin with a verb in some non-
narrative past form, the subsequent verbs taking on the narrative form.
Comrie (1985:102-104) cites (Papuan) Bahinemo (after Longacre 1972:47-
48) beside the Vedic injunctive as instances of (progressive) tense neutral
ization; Bahinemo is similar to Jacaltec in using the present form for the
neutralized category.
Drawing on this small sample for correlations with directionality of
clause chaining, it emerges that only a few of the languages are verb-initial,
such as Nootka, Yana, Hebrew, and Maasai.^26 Most of the African lan
guages, as well as Austronesian Lenakel (fn. 6), show verb-medial (SVO)
word order. Thus the division seems to emerge between verb-final (SOV)
languages, tending toward regressive clause chaining, on the one hand, and
non-verb-final ones (VSO, SVO), tending toward progressive clause chain
ing, on the other. One is reminded of Hawkins's (1983:137) observation:
"The VSO - SVO distinction can now be seen to mirror the rigid - nonrigid
SOV dichotomy." The exceptions are the (nonrigid) Kpelle cited by Wei
mers (although most other Mande languages are SVO, so this may repre
sent a relatively recent change), and apparently also Bahinemo. An excep
tion in the other direction is the Jacaltec case cited by Van Valin (1984:545-
546), which would constitute regressive clause chaining in a VSO language