119I’ll ever give you any money! ), which can be accounted for only by ellipsis of
a negative matrix clause ( You don’t believe ). But it is diffi cult to say whether
there are any limits on which ellipted clauses can acceptably be reconstituted.^26
At stage (3), the clause has achieved greater semantic specifi city and only a
subset of ellipted main clauses can be “restored”: “Restoration of material is
conventionalized to a subset of the grammatically tolerated possibilities,” such
as very general clauses like “It would be nice/ lovely if ....” in the case of
wishes (372– 373).
When the subordinate clause becomes independent, it acquires a “more spe-
cifi c constructional meaning,” and it is “available for pragmatic interpretation”
(374– 375). Heine ( 2012 ) likewise argues that the insubordinated clause is his-
torically derived from a full conditional construction, in course of which change
the apodosis main clause is ellipted via “cooptation ” (§§3, 5; on “cooptation,”
see Section 1.5.5 ). And Lombardi Vallauri ( 2004 ) accounts for the develop-
ment through ellipsis which is brought about by shifts in discourse planning
and dialogic interruption (perhaps interrupted because the conditional sense
seems self- suffi cient).^27
Because the process outlined here goes against the usual direction of
change in grammaticalization – from main clause to subordinate clause, or
from looser to stronger internal dependencies – it is questioned whether this
represents a violation of unidirectionality. Evans points out that the type of
“reanalysis” envisaged here differs from the defi nitions used in grammaticali-
zation approaches (2007: 376n.), but concludes nonetheless that “it is clear
that insubordination goes against the usual direction of change” (2007: 376).
Lombardi Vallauri ( 2004 : 207– 208) argues that the semantic and pragmatic/
illocutionary force of the main clause is incorporated into the if - clause and thus
could be seen as showing strong internal dependencies. But he also notes that
the process of development might be better described as pragmaticalization
(see Section 1.5.3).
Is reconstruction of the deleted apodosis or underlying main clause pos-
sible? Both Evans and Heine admit to certain diffi culties. “[E] xactly which
main clause is restored is determined by a process of conversational implica-
ture” (Evans 2007 : 371), and once constructionalization has occurred it may
not be possible to restore any ellipted material (374). Likewise, reconstruc-
tion is “hypothetical” (Heine 2012 : §3), since in most cases it is not possible
to reconstruct the exact form of the ellipted matrix clause (§5). Importantly,
26 Because the ellipted material would have to be of the most general kind and could not be read-
ily supplied from context, Stirling ( 1999 ) argues against seeing the apodosis as elliptical (cf.
Evans 2007 : 372n. on stricter and looser defi nitions of “ellipsis”).
27 Compare the OED account (s.v. if , conj. and n., def. A7): It suggests that the process of devel-
opment is “aposiopesis,” ‘a rhetorical artifi ce, in which the speaker comes to a sudden halt, as
if unable or unwilling to proceed.’
4.4 If Only