OPTIONAL AND MULTIPLE LONG DISTANCE GENITIVE OF NEGATION

This paper deals with so-called Long Distance Genitive of Negation (LD
GoN), i.e., a case shift of a complement of an infinitival verb from
accusative to genitive when a *higher* verb is negated, e.g.:

(1a) Tomek kazal Janowi calowac Marie-ACC / *Marii-GEN.

    `Tom ordered/asked John to kiss Mary.'

(1b) Tomek nie kazal Janowi calowac Marii-GEN.

    `Tom didn't order/ask John to kiss Mary.'

Although there are a number of analyses of LD GoN, e.g., Dziwirek (1994,
1998) within the Relational Grammar framework, Witkos (1996, 1998) within
GB/Minimalism, and Przepiorkowski and Kupsc (1997) within Head-driven
Phrase Structure Grammar (HPSG), none of them deals with or even notices
two aspects of LD GoN which set it apart from local GoN: the optionality of
LD GoN and the possibility of multiple NPs simultaneously undergoing GoN.
In this paper, we illustrate both aspects and present an HPSG analysis
accounting for them.

Optionality of many cases of LD GoN, although unnoticed or explicitly
denied in the works cited above, is acknowledged by traditional linguists,
and even to some extent sanctioned prescriptively, e.g.:

(2)  Nie moglbys przestac studiowac algebre-ACC / algebry-GEN?

    `Couldn't you stop studying algebra?'

(3)  Jan nie uwazal za stosowne kupowac samochod-ACC / samochodu-GEN.

    `John didn't consider it appropriate to buy a car.'

We adduce many more examples like (2)-(3) and take them as evidence that LD
GoN is in principle optional, although in some instances the genitive is
clearly preferred to the accusative.

Multiple GoN occurs in contexts involving verbs subcategorizing for an
accusative NP and an infinitival verb, which may in turn subcategorize for
an accusative NP.  One such verb, _uczyc_ `teach' is relatively well known,
but there is also a class of constructions involving the light verb _miec_
(lit.: `have') which subcategorizes for a contentive accusative complement
and an infinitival complement, e.g., _miec ochote_ `feel like', cf.:

(4)  Nie mam ochoty-GEN uczyc Marii-GEN pisac wierszy-GEN.

    `I don't feel like teaching Mary how to write poems.'


Although multiple GoN per se is not particularly problematic for the
analyses mentioned above, the existence of examples such as (5), where a
positive verb combines with an accusative NP and a negated infinitival
complement, is a problem for the analysis of Dziwirek, which would predict
that the accusative NP should actually be genitive, and it may be
problematic for Witkos, depending on how various implicit parts of the
analysis are spelled out.

(5)  Mam ochote-ACC nie robic tego-GEN wiecej.

    `I feel like not doing this anymore.'


In this paper, we present an HPSG analysis of GoN which builds on
Przepiorkowski and Kupsc 1997, but accounts for the optionality of LD GoN
and its possible multiplicity.  For lack of space, we present here only the
main points of the analysis, in principle formalizable in many frameworks,
saving technical HPSG details for the full paper.

First, we assume (together with much GB and HPSG literature) a rule
responsible for local Genitive of Negation.  This rule says that any
complement bearing so-called structural (as opposed to lexical or inherent)
case must be realized as accusative when its governing verb is not negated
and as genitive when the verb is negated.

Second, we claim that the so-called Long Distance GoN is actually only
apparently long distance, i.e., we deny that this is a phenomenon separate
from the garden variety local GoN.  Instead, we postulate that in cases of
LD GoN, arguments of lower verbs raise to become arguments of higher verbs,
so that they are in the scope of the local case assignment rule mentioned
above.

Third, the optionality of LD GoN is the result of the optionality of
raising in such infinitival environments.  When the complement stays on the
lower verb, the local rule assigns it the accusative case, when it raises,
the same local rule assigns the genitive case.

Fourth, we independently support this analysis of optional raising in Polish
verb clusters with facts from optional clitic climbing (Witkos 1998) and
optional haplology of the reflexive marker (Kupsc 1999) in such
environments.

Fifth, we show that the analysis correctly accounts for multiple GoN, as
well as for the optionality of GoN on lower arguments in such multiple GoN
environments (not shown here for lack of space).

Finally, we show that this analysis fares better than those of Dziwirek and
Witkos, even if the `clause union' processes they posit are made optional
to account for the optionality of LD GoN.  Minimally, the current analysis
differs from so-modified Dziwirek (1994, 1998) in correctly dealing with
(5), and it differs from (modified) Witkos (1996, 1998) in allowing to
retain the overwhelming generalization that structural case is assigned
(resolved to particular morphological cases) strictly locally; on the
account of Witkos, case assignment domain must be extended in case of LD
GoN.