Besides the well-known verbal inverse systems, a few languages display an inverse system in (predicative) relational nouns (R(u,v)) (Amith & Smith-Stark 1994): here, the relative salience of the possessor argument v and noun referent u is evaluated in terms of the hierarchy 1/2 > 3. In contrast to verbal inverse systems, the lower argument (the possessor) is the more salient argument in the direct form. Since the inverse form is not indicated by an inverse marker in most of the languages, direct and inverse form must be disambiguated by the linking pattern, which includes a verbal linker (ACC/ERG) in the inverse form, as the following illustration shows for Guerrero-Nahuatl and Chinook, which I will discuss in my paper.
(1) Guerrero-Nahuatl Chinook
lv lu R(u,v) lv lu R(u,v)
a. direct: P NOM P NOM
b. inverse: NOM ACC ERG NOM
I will argue that the P-linker is excluded in the inverse form in order to avoid ambiguities and that the different distribution of NOM in the inverse pattern of Guerrero-Nahuatl (less salient argument) and Chinook (highest argument) results from different markedness constraints.
I will propose an analysis which is based on the linking approach of Lexical Decomposition Grammar (LDG; Wunderlich 1997), formulated in terms of Correspondence Theory (McCarthy & Prince 1995). I make use of the structural features [+hr] 'there is a higher role' and [+lr] 'there is a lower role' and of the salience feature [+ls] 'there is a less salient argument'. The possessor argument is structurally specified as [+hr]; in the direct form, it is specified as [+ls], as shown in the following:
(2) lv lu R(u,v)
+hr -hr
direct +ls -ls
inverse -ls +ls
The linkers are also specified in terms of the features [hr], [lr] and [ls]: ACC [+hr], ERG [+lr], NOM [ ], P [+hr,+ls]. The correct linking pattern results form an evaluation of the linker assignment in terms of the Faithfulness constraints MAX(+hr) '[+hr] in the input should have a correspondent in the output', MAX(+lr) and MAX(+ls) analogously. The theta-grid (e.g., lv[+hr,+ls] lu[-hr,-ls]) constitutes the input, the linker-annotated structure (e.g., lv[P] lu[NOM]) the output. The presence of morphological linkers violates the markedness constraints *[+hr] 'avoid a [+hr] linker', *[+lr] or *[+ls].
The inverse system in Chinook is analyzed as an ergative split in the possessor paradigm; the possessor argument is lexically specified as [+lr], thus enforcing ERG. Due to the contextualized markedness constraint *[+lr]/+ls 'avoid a [+lr] linker for a [+ls] argument', ERG is excluded in the direct form, whereas it gets preference in the inverse form due to the faithfulness constraints MAX(lex), which requires the visibility of lexical features. *[+lr]/+ls is a special instance of a contextualized markedness constraint that excludes ERG for elements high on the Silverstein hierarchy (Stiebels 2000). The noun referent is invariantly linked by NOM.
The inverse system in Guerrero Nahuatl excludes the P-linker in the inverse form due to the contextualized markedness constraint *[+hr]/-ls, which is a specific instance of a contextualized markedness constraint that excludes ACC for elements low on the Silverstein hierarchy. Guerrero-Nahuatl tolerates a violation of IDENT(hr) 'correspondents should be identically specified for [hr]' due to a higher-ranked constraint that excludes ambiguities between a direct and an inverse reading. Due to *[+hr]/-ls, NOM must always be used for the less salient argument.
References
Amith, Jonathan D. & Thomas C. Smith-Stark. 1994. Predicate nominal and transitive verbal expressions of interpersonal relations. Linguistics 32/3, 511-547.
McCarthy, John and Alan Prince. 1995. Faithfulness and reduplicative identity. In 'Papers in Optimality Theory', University of Massachusetts Occasional Papers 18, Jill N. Beckman, Laura Walsh and Suzanne Urbanczyk (eds.), 249-384.
Stiebels, Barbara. 2000. Linker inventories, linking splits and lexical economy. To appear in Lexicon in focus, Barbara Stiebels and Dieter Wunderlich (eds.). Berlin: Akademie-Verlag.
Wunderlich, Dieter. 1997. Cause and the Structure of Verbs. Linguistic Inquiry 28, 27-68.