388 MARY L. NUNES
definitions put forward in configurational analyses. Neither does it include
LFG-like semantically-unrestricted vs. semantically-restricted GFs, where,
in the unmarked case, shared V-vN argument structures are mapped onto
syntagmatically corresponding, but distinct GFs.
In fact, as the discussion thus far has indicated, RRG's point of analyt
ical departure contrasts with both of these approaches in not being based on
structurally parallel clausal and nominal constructions in which a prehead
NP-argument is included in both (e.g. John died/John's death, The enemy
destroyed the city / the enemy's destruction of the city, The city was destroyed
by the enemy / the city's destruction by the enemy, etc.). Such an approach,
presumably intended to inform universally-applicable theories, makes the
fallacious assumption that the double genitive construction is regularly pos
sible in languages other than English. As Comrie (1976) and Comrie and
Thompson (1985) show, however, German is one of the few languages (if
not the only one other than English) which permits the use of two genitive
constructions in a single vNP. Moreover, German has only a minimal,
highly marked ability to use the double genitive form. As the works just
cited relate, numerous languages have possessive adjectives akin to the pos
sessive pronouns in English (e.g. ray, your, their, etc.) which can be used in
vNPs including one genitive construction. However, only English has regu
lar access to a double genitive form. Thus, in basing analyses of the vNP on
double genitive constructions, both the configurational and LFG
approaches assume what from a universal standpoint is a highly marked
frame of reference.
RRG, on the other hand, does not make the double genitive assump
tion. Rather, an RRG analysis of the vNP begins with the A and U assign
ments in the clause and the hierarchically-predictable choice of macrorole
for vN direct-argument linking in the vNP (cf. 8) — a choice which is exhi
bited in the vN [of NP] types of constructions, not necessarily in vNP con
structions which include a genitively-marked prenominal NP. It is from this
point of analytical departure that the analysis sketched above is applied in
2.3 to a body of data and contrasted in its effectiveness with those put for
ward by the other approaches.
2.3 Applying the analysisThe data to which the analysis is applied^14 reflect an effort to include not
only numerous groups of nominals discussed frequently in the literature,