64 ROBERT D. VAN VALIN, JR.
abade "the girls" denotes the primary topical participants, it is not always
pivot; in the third clause from the end it is the undergoer, and there is no
way in Yagaria that the undergoer with a transitive verb can appear as pivot
in the clause, just as in Amele.
Even though the actor will always be pivot with a transitive verb, there
is nevertheless a restricted neutralization of semantic roles for syntactic pur
poses here, because the construction does not distinguish between actor
and undergoer with intransitive verbs. This is what crucially differentiates
languages like Amele from languages like Acehnese in which there is no
restricted neutralization with intransitive verbs. Other languages which
have a primary syntactic pivot in which discourse pragmatic factors have no
effect on the selection of the argument to be pivot with a transitive verb
include Lakhota (Siouan), Choctaw (Muskogean), Cree (Algonquian,
North America), Zapotee (Oto-Manguean, Mexico), Enga, Kewa, Yimas,
Fore (Papua-New Guinea), Mparntwe Arrernte, Warlpiri (Australia),
Chechen, Ingush (Caucasus, USSR), Thai, and Vietnamese.
It must be emphasized that within a single language, different construc
tions may have different pivot selection criteria. That is, in some construc
tions pragmatic considerations play a role in pivot selection, e.g. determin
ing the pivot for topic chains as in (40) and (44) (where both the controller
and the omitted arguments are clause-internal), while in others, e.g.
imperative constructions, they are irrelevant and semantic role considera
tions determine the argument to be pivot. This is the much discussed "split-
subject" phenomena found in Philippine and some ergative languages,
among others.
Thus there are two very different situations regarding selection of the
argument to function as pivot in syntactic constructions in the world's lan
guages. In some constructions in some languages, discourse pragmatic con
siderations influence this selection, while in the majority of languages there
are no constructions in which they play a role.^38 In markedness terms, it
could be described as [±pragmatic influence], with [+pragmatic influence]
the marked case and [—pragmatic influence] the unmarked case. Since
focus (information) structure is a feature of the grammar of all languages,
what this contrast represents is the variable grammaticalization of discourse
pragmatic relations in clause-internal morphosyntax, in particular in clause-
internal relational structure.^39 The two types of syntactic pivot may be
characterized as follows.