ARGUMENT LINKING IN DERIVED NOMINALS 427- What was not emphasized in that discussion is that RRG's explanation for the ina
bility of some vNPs to pattern with related dative-V clauses in the dative shift does
not require the ad hoc restructuring which plagues configurational treatments of
dative shift. Neither does it entail referring to discriminately available GFs for the
clause and the NP, in conjunction with incompatibly assigned oblique grammati
cal-function markers for particular vN arguments, as is the case with the LFG
treatment of dative shift. Rather, in RRG, it is simply the inability of vNs to take
more than one direct argument which explains why constructions headed by
dative Vs taking two post-verbal direct arguments cannot have nominalized coun
terparts. - In RRG, the passive-voice operation serves to markedly switch the syntactic func
tions of an argument associated with a particular macrorole in the clause. Obvi
ously, the absence of GRs in the NP precludes this operation from directly ex
plaining any syntactic relations in the vNP. - Recall from the discussion in 2.3.2.4 that where ACTs receive an ACM interpre
tation — i.e. where an ACT y argument is interpreted as a delimiter of the activity
— the non-A argument is treated as if it were an U, thus meeting the condition
given in (56). It is important to keep in mind, however, that while no vN argument
can be preposed without meeting the condition, not all arguments meeting the
condition can be preposed where they are arguments of a vN capable of taking
both macroroles. Only with vNs capable of taking just one macrorole does the
condition predict argument preposability. - Due to their "be predicate-adjective" type of construction, no single argument
STAs are included in (57) (or were included in the data base). In the NP counter
parts of such constructions, it is the adjective which serves as the nominalized
nucleus, not the verb: Jeff is happy; the happiness of Jeff; Jeffs happiness. No
ACM examples are included in (57) and (58), as ACMs always take both mac
roroles. - Where an asterisk precedes a parenthesized element, that element is not optional.
In this case, recall from the discussion of Ross's clause-to-noun cline in 1.1 that
whereas gerundive nominals must retain the clausal "subject" NP (cf. 57g),
nominalized verb constructions need not (cf. 57f).
It should also be noted that some noncontrastive action nominals cannot
receive the type of nonprocess interpretation which is available to landing. With
such vN-ing forms, preposing always results in a gerundive nominal interpreta
tion. Breaking, in its ACH sense, is an example:
(i) The stringU broke.
(ii) the breaking of the stringy dir.arg
(iii) ger nom[The * (string's) breaking] was unexpected. - As future discussion will demonstrate, with ACTs like investigation which may be
treated as ACM performance objects, the non-A is not only treated as an U, but
is preposable: the murder's investigation (by Sherlock). Attack is used here pre-