| |
Abstract:
Prosodic information helps mark the information structure of
an utterance. For example, new information is typically accented
and old information deaccented. Listeners are clearly sensitive to
this information during comprehension: Comprehension is slowed by
an accentuation pattern inappropriate for the focus requirements of
an utterance in context (e.g., Birch & Clifton, 1995; Bock
& Mazzella, 1983). It has been proposed that listeners
interpret an accented word as introducing a new discourse entity
and a deaccented word as anaphoric (Bard, Cooper, Kowtko, &
Brew, 1991; Terken & Nooteboom, 1987). However, to our
knowledge, it has not yet been established whether listeners
actually use accent to help circumscribe referential domains.
We monitored eye movements as participants followed spoken
instructions to move objects on a computer screen. On some trials,
two of the four objects shared phonological similarity at onset
("cohort" competitors, e.g., bell and bed). Fixations to potential
lexical competitors were used to determine which potential
referents were considered when listeners encountered an accented or
deaccented noun. Each trial contained two instructions. The first
referred to one of the phonologically similar pictures. The second
could refer to the same object (anaphoric condition, e.g., put the
bell above the triangle [...] now put the bell below the circle),
or to another object (nonanaphoric condition, e.g., put the bed
above the triangle [...] now put the bell below the circle). The
accentual structure of the second instruction was varied, with a
pitch accent either on the target noun (e.g., now put the BELL
below the circle), or on the preposition following the noun (e.g.,
now put the bell BELOW the circle), resulting in a deaccented
noun.
As the second instruction unfolded, participants tended to
initially fixate pictures other than the referent picture of the
first instruction (i.e., the two distractors and the cohort
competitor in the anaphoric conditions, the distractors and the
target in the nonanaphoric conditions). In the anaphoric
conditions, fixations to the cohort competitor--a "new" entity in
the discourse--quickly dropped as the target word unfolded when it
was deaccented, whereas fixations remained high when it was
accented. Conversely, in the nonanaphoric conditions, few fixations
to the cohort competitor--an "old" entity--were observed when the
target word was accented, whereas fixations increased as the target
word unfolded when it was deaccented. This pattern of lexical
competition demonstrates that listeners interpret accented nouns as
introducing new entities and deaccented nouns as anaphoric. Accent
is thus used to circumscribe the set of referential candidates
considered by the listener.
|
