| |
Abstract:
In languages like German or Russian the case of an ambiguous
NP can be attracted to the case of a morphologically specified
relative pronoun or coindexed pronoun. This happens only if the
latter element intervenes between the NP and the verb which
resolves the local case ambiguity. It can be shown that this case
attraction reflects some aspects of the architecture of the human
parser, and not grammar. We discuss experiments addressing the
question of whether non-coindexed intervening case and number
feature leads to similar results, as exemplified in (1) and
(2).
(1a) ..., dass man Botschafterinnen , ohne mir zuzuhören,
eingeladen hat that one ambassador, without medat to listen
visited has (1b) ..., dass man Botschafterinnen, ohne mich zu
informieren, eingeladen hat that one ambassador, without meacc to
inform visited has
(2a) ..., dass man den Dirigentensg, falls die Minister
zustimmen, einladen sollte that one the conducteursg, if the
ministerpl agree, invited should (2b) ..., dass man den
Dirigenten, falls die Minister zustimmen, gratulieren sollte that
one the conducteurpl, if the ministerpl argee, congratulate
should
We also investigated a structure in which the "unexpected"
event of reanalysis intervenes between the ambiguous NP and the
disambiguating position, as illustrated in (3). Assuming a
default subject preference for the initial NP we expect a
reanalysis effect if the second matrix NP is nominativly marked.
This effect was found in sentences like (3b) but not in (3a)
which has an SOV order in the relative clause. This result shows
that the parser uses the structural information internal to the
relative in determining its strategy of sentence comprehension
for the matrix clause.
(3a) Die Botschafterin, die die Minister gesehen hat, sah
der/den Reporter theamb ambassador, thatsg thepl minister visited
hassg, saw thenom/acc
(3b) Die Botschafterin, die die Minister gesehen haben, sah
der /den Reporter theamb ambassador, thatsg thepl minister
visited havepl, saw thenom/acc
These observations will be discussed in relation to existing
current psycholinguistic models. We will argue for a parallel
processing model, in which the analysis of the material in an
intervening part (e.g. relative clause) of a sentence may
influence the resolution strategy of the initial detected
ambiguity.
|
