| |
Abstract:
The attachment-binding dualism proposed by Hemforth & al.
(2000, in press) claims that relative clause attachment is
determined by a combination of syntactic and anaphoric processes.
While syntactic processes favor the most recent host, anaphoric
processes try to bind the relative pronoun to the most salient
referent, saliency being determined by discourse and thematic
properties of the referents: referents that are direct verbal
arguments and thematic licensers (NP1 in structures without
thematic prepositions like 1 and 2) are considered more salient
than ones that are thematic licensers only (NP2 in 1, 2) which in
turn are more salient than referents that are neither direct verbal
arguments nor thematic licensers (NP3 in 1, 2). Saliency is
supposed to map onto activation: while all referents are supplied
with an equal amount of activation when read, their decay rate
differs with respect to saliency, i.e. the activation of more
salient referents decays more slowly than the activation of less
salient ones. For adjacent RCs following a complex three-site NP
without prepositions (1) this results in the following pattern (cf.
Walter & Hemforth, 1998): NP1 and NP3 are equally active but
more active than NP2. For extraposed RCs (2) this pattern is
different since, because of the intervening verb, additional time
passes between the complex NP and the RC: on reading the extraposed
relative pronoun, NP1 is more active than NP2 and NP3 which do not
differ significantly in activation.
1.. Die Anwältin der Chefin der Firma, die berühmt war
und der Korruption verdächtigt wurde, wurde in der Lokalpresse
erwähnt. Lit: The lawyer (nom, sing, fem) the boss (gen, sing,
fem) the firm (gen, sing, fem) who (nom, sing, fem) very famous was
and of corruption suspected was, was in the local paper mentioned.
(adjacent)
2.. Die Lokalpresse hat die Anwältin der Chefin der Firma
erwähnt, die berühmt war und der Korruption
verdächtigt wurde. Lit: The local paper had the lawyer (nom,
sing, fem) the boss (gen, sing, fem) the firm (gen, sing, fem)
mentioned who (nom, sing, fem) very famous was and of corruption
suspected was. (extraposed)
While these activation patterns can readily explain reading time
data for these structures (cf. Walter & Hemforth, 1998), there
is no direct evidence for activation levels of the referents at the
point of time the relative pronoun is read. We therefore conducted
a unimodal repetition priming experiment to test the activation of
the referents. Participants were presented with sentences like (1)
and (2) that were ambiguous in that the relative pronoun could
refer to either of the three NPs. After reading the relative
pronoun, they were presented with a target that was either
identical to NP1, NP2, or NP3 or unrelated to all three NPs.
Attachment-binding predicts that more active referents should be a
more effective prime than less active ones. Given the activation
levels above, the priming effect for targets identical to NP1 and
NP3 should be stronger than for targets identical to NP2 in the
adjacent conditions (if compared to the unrelated control) and
stronger for NP1 than for NP2 and NP3 in the extraposed cases.
Alternatively, it is possible that repetition priming is not
sensitive to saliency (=activation) of discourse referents but to
the activation of syntactic tree nodes. Assuming more recent tree
nodes are more active than less recent ones, the priming effect
should be stronger for targets identical to more recent hosts.
|
