264 WILLIAM H. JACOBSEN, JR.
phologically distinct) "consecutive" verb. Clasical Hebrew almost achieves
this in that it can have sequences consisting of a so-called perfect tense fol
lowed by a string of "waw-consecutives" (w- "and" plus the imperfect with
vowel changes). Unfortunately, however, the initial perfect is not obliga
tory and may in many cases be circumstantial in function. At any rate, one
snag in the development of such structures as are here speculatively
posited is the fact that the initial clause of a sentence is the position where
back-reference to a previous sentence may occur in connected discourse.
In that the sentence-initial position is customarily pre-empted for this func
tion it is not available for regular occurrence of a distinctive chain-initial
verb.
Clearly, if we wish to think of these two directionalities as subtypes of
a broader phenomenon of clause chaining, which seems to me a useful thing
to do,^27 we must broaden the requirements from what would apply merely
to regressive clause chaining. This passage suggests right away one respect
in which this must be done: progressive clause chaining must be allowed to
dispense with an initial fully specified verb at times. In Hebrew this hap
pens with certain expressions of time which can replace the initial perfect
(cf. Kiparsky 1968:48). We have seen that such an initial specification can
be absent in narrative contexts in Nootka (section 6.1) and also with the
Vedic injunctive (section 13.2). Similarly, Weimers (1973:364) notes of the
Kpelle consecutive: "In informal narrative, especially folk stories, the same
construction may also be used to refer to the first action in the series as well
as later actions; the reference is assumed to be past time. In a series includ
ing several actions, a sentence-terminal pause may occur, and the next con
struction may still be a consecutive."
Longacre (1985:238-239, 263-284) gives a helpful account and
exemplification of clause chaining.^28 He finds (264-265) three features
which distinguish (regressive) chaining from co-ranking (= coordinate)
structures. One is the distinction between final and non-final verbs, which
translates into that between (dispensable) initial and non-initial verbs in
progressive clause chaining. A second is the marking of switch-reference on
non-final verbs. This is characteristically absent in progresive clause chain
ing, the exception being Lenakel. But not all medial verbs in regressive
chaining languages are so marked either: note, for example, the distinction
between "explicit" and "implicit" (= with and without switch-reference)
chains in (Tucanoan) Guanano (Longacre 1983:199-205, 1985:278-281).
Longacre's third feature of chaining is the marking of relative temporal
relations such as overlap vs. succession (and their metaphorical extensions),