Hello! Thank you for visiting my virtual poster. If you'd like more information, or if you have any questions/comments, please email me at pianibes@usc.edu or leave a comment at my OSF page (by clicking on the speech bubble on the right upper corner). Thank you!
We use a typological property of Brazilian Portuguese, anaphoric null
objects, to show that resolution of subject/object ambiguities is sensitive
not only to syntactic complexity and plausibility/semantics (e.g. [2]) but
also to anaphoric/coreference information. Due to optionally transitive
verbs (e.g. *read*) allowing both *overt *and *implicit objects* (*IO*s,
ex.2,3), structures like (1) are temporarily ambiguous: ‘the book’ is
initially parsed as the object of ‘read,’ but needs to be re-analyzed when
it’s disambiguated as the subject of ‘fell’ [1,2]. The garden-path arises
because syntactic simplicity (Late Closure) favors parsing the noun as an
object rather than parsing it as a subject and positing an IO. Sentences
like (1) result in lingering misinterpretations (e.g. [2]).
With verbs like ‘read’, English only allows IOs with underspecified
referents (ex.3), not anaphoric ones (ex.4): The missing object of ‘read’
cannot refer to a previously-mentioned book. However, IOs in Brazilian
Portuguese (BP) can have *either* an underspecified *or* discourse-based
anaphoric interpretation (ex.3-4). (Whether these are ellipses or null *pro*
[4,5] is not crucial here).
This means that, in the right context, when BP readers encounter a noun
after an optionally transitive verb in an embedded clause, they could *(a)*
parse the noun as the direct object, *(b)* posit an underspecified IO and
parse the noun as a subject, or–unlike English–*(c)* *posit an anaphoric IO
*that refers to an already-mentioned referent in the discourse and parse
the noun as a subject*.*
Option (a) is preferred over (b), due to Late Closure. Crucially, BP lets
us to test the interaction between syntax and discourse: Do readers go for
the *syntactic route *(noun=object, (a)) or do they pursue a *discourse-based
route *(noun=subject, (c)) (*and be less garden-pathed*) if an antecedent
is available? We test this by manipulating the* presence/absence of an
antecedent. *
To further probe the syntax-discourse relation, we test whether the
postverbal noun’s definiteness modulates the ease of (re)analyzing the verb
as having an anaphoric IO. *Definite* NPs typically represent old/given
information and can (be inferred to) co-refer with a preceding IO, but
*indefinites* are less likely to co-refer with already-evoked entities.
Thus, *discourse considerations* predict garden-path sentences with *definite
NPs are easier to process* at the disambiguation point than ones with
indefinite NPs: Even if the parser was initially garden-pathed, a definite
NP can act as a cue signalling a (coreferring) IO; an indefinite NP
provides no such cue.
*Experiment:* In a self-paced reading study (47 native BP speakers), we
manipulated (i) structural ambiguity (with comma/without), (ii) presence of
potential antecedent (present/absent) and (iii) definiteness of the
ambiguous noun phrase (def/indef); See Table 2 for the design.
*Results:* We analyzed reading times (RTs) at the disambiguating verb and
spillover (Table 2). At the verb and spillover regions, RTs are slower in
ambiguous (commaless) conditions (basic garden-path; p<.001, Fig.1). This
slowdown occurred even when an antecedent was available: There is no sign
that *presence of an antecedent* prevents garden-pathing by biasing people
to assume an anaphoric IO and correctly parse the noun as a subject. In
fact, the effect is in the other direction: conditions with antecedents
yielded *larger* slowdowns (p<.01). Furthermore, sentences with *definite
postverbal NPs *are faster (than indefinites) across all conditions at
various points in the critical area (p’s<.01). This fits with the
discourse-based prediction that people assess whether the subject NP can
co-refer with the IO. If so, processing is facilitated, indicating
re-analysis is not purely syntactic but also sensitive to anaphoric
relations.
For every trial, a *question* (*Did I read the book?*) probed if the NP was
interpreted as an object. Here we see that *presence of an antecedent
facilitates recovery from garden-pathing/decreases lingering
misinterpretations: *“no” responses were more frequent when a potential
antecedent was present (p<.001, Fig.2). This suggests that an antecedent
facilitates positing an *anaphoric* IO and interpreting the NP as a
subject, as predicted. We also see that in *unambiguous*, no-antecedent
conditions, “no” answers are less frequent when the NP is definite
(p<.001): Without an antecedent or a garden-path, a cataphoric relation
between the IO and the NP is preferred over an underspecified referent if
possible (with definites).
Table 1. BP on the right (When helpful for exposition, ∅ is used denote
implicit objects)
1.While the man read *the book* fell on his foot. |While the man read ∅, the
book fell on his foot
2. [obj=overt noun]
The man read the book.
O homem leu o livro.
3. [obj=implicit]
Underspecified referent => read something
The man read ∅.
O homem leu ∅.
4. [obj=implicit]
Anaphoric referent => read the book
*** The man grabbed the bookk and read ∅k.
*ok* O homem pegou o livrok e leu ∅k.
Table 2. Conditions and example target items (critical area: verb +
spillover is underlined.)
5. No antec + Indef
Depois que eu li (,) *um livro* *caiu no meu pé* esquerdo.
*After I read (,) a book fell on my left foot.*
6. No antec + Def
Depois que eu li (,) *o livro* *caiu no meu pé* esquerdo.
*After I read (,) the book fell on my left foot.*
7. Antec+Indef
Depois que eu *recebi uma carta e* li (,) *um livro* *caiu no meu* *pé*
esquerdo.
*After I received a letter and read (,) a book fell on my left foot.*
8. Antec+Def
Depois que *eu **recebi uma carta e* li (,) *o livro* *caiu no meu pé*
esquerdo.
*After I received a letter and read (,) the book fell on my left foot.*