|
Abstract:
In spoken language, local acoustic information is frequently
congruent with more than one phoneme. Nevertheless, the words of
a sentence are rarely misunderstood; sentence context biases
listeners towards a contextually appropriate interpretation of
acoustically ambiguous words (Borsky, Tuller, and Shapiro, 1997;
Connine, 1987). The purpose of the present study is to
investigate the role of both acoustic and semantic information in
on-line sentence processing. Ten target stimuli forming a
GOAT-to-COAT continuum were created from natural speech by
manipulating a temporal cue for voicing (VOT) that distinguishes
word-initial /g/ from /k/. Stimuli were embedded in biased
sentences such as:
Goat-biased: The busy dairyman forgot[1]to milk the
(goat/coat)[2]in the[3]drafty barn.
Coat-biased: The careful tailor hurried to[1]press the
(goat/coat)[2] in the [3]cluttered attic.
A cross modal lexical decision task (CMLD) was used: subjects
made a word/non-word decision to a visual letter string that
appeared during the uninterrupted auditory presentation of a
sentence. None of the visual probes were related to the
sentences. Response times (RT) to the visual probe were used to
assess comparative processing load for different combinations of
acoustic target stimulus and biased sentence context at positions
[1], [2], and [3]. At the end of the experiment, we tested each
subject's identification of the isolated target stimuli. As
expected, all subjects heard the shortest VOTs as 'goat', the
longest VOTs as 'coat', and mid-range VOTs as both 'goat' and
'coat' on separate trials. The CMLD data showed the following
effects at each probe position.
[1] Pre-target control: There were no significant effects.
[2] Target offset: RTs were significantly greater for a
mid-range VOT stimulus than for either endpoint, regardless of
context.
[3] 450ms after target offset: There was a VOT x Context
interaction due to the effect of context at each endpoint; RTs
were significantly greater when the sentence was biased toward
the opposite endpoint, even though the identification of endpoint
stimuli are presumably consistent with VOT, as in the post-test
identifications as well as in previous identification results for
the same acoustic stimuli embedded in the same biased
sentences(Borsky, et. al., 1997).
The effect of acoustic information at a potentially ambiguous
target word, plus the development of sentence context effects
450ms downstream support an account of auditory sentence
comprehension in which the processing of acoustic and semantic
information each has a different time course.
Borsky, S., Tuller, B., and Shapiro, L.P.(in press). "How to
milk a coat": The effects of semantic and acoustic information on
phoneme categorization.
The Journal of the Acoustical Society of America.
Connine, C.M. (1987). Constraints on interactive processes in
auditory word recognition: The role of sentence context.
Journal of Memory and Language,
26, 527-538.
|
