| |
Abstract:
Recognition time for spoken words in tasks like lexical
decision depends on word frequency (e.g., Marslen-Wilson, 1990) and
the number of acoustically similar words in the lexicon, weighted
by frequency (e.g., Luce & Pisoni, 1998). This indicates that
during spoken word recognition, similar words (neighbors) compete
for recognition. However, tasks like lexical decision allow only a
coarse test of similarity metrics and tell us relatively little
about the on-line processing of words, since they yield a single,
post-decisional measurement. We have begun using an artificial
lexicon - which gives us control over distributional
characteristics and allows us to study lexical effects during
learning - and a continuous on-line measure (eye tracking),
allowing us to evaluate similarity metrics over the time course of
spoken words.
We conducted two experiments to study lexical competition for
items varying in frequency and neighbor frequency. Subjects were
trained to recognize the names of sixteen novel objects. The names
were CVCVs (e.g.,
pibu)
with two potential competitors: an onset competitor (e.g.,
pibo)
and a rhyme (e.g.,
dibu).
We tracked eye movements as subjects responded to spoken
instructions to manipulate one of a set of objects (e.g.,
click on the pibu).
This paradigm provides a continuous measure of lexical activation
as a word is spoken (e.g., Allopenna, Magnuson, & Tanenhaus,
1998, used it to measure competition among English words sharing
onsets & rhymes).
In Experiment 1, half the items were high frequency (HF, 7
presentations/training block, with 14 blocks over 2 days) and half
were low (LF, 1 presentation/block). Half the HF and LF items had
HF competitors, and half had LF competitors. After three hours of
training, we found competition effects for onset and rhyme
competitors (with the probability of fixating items over the course
of a word mapping onto emerging similarity), modulated by target
and competitor frequency (e.g., with more competition for HF
competitors, and more again for LF targets). Thus, the paradigm is
sensitive to frequency and neighborhood effects, and shows that
incremental lexical competition effects emerge early in
learning.
In Experiment 2, we examined whether the time course of
recognition is affected by the characteristics of absent
competitors (as in lexical decision). Specifically, we presented HF
items which had HF or LF competitors with three unrelated,
medium-frequency items. Subjects recognized HF items with (absent)
LF competitors more quickly than those with (absent) HF
competitors. Thus, the paradigm is sensitive enough to show that
representations of newly-learned items are competing against
similar items during recognition even when they are not
present.
References
Allopenna, P. D., Magnuson, J. S., & Tanenhaus, M. K.
(1998). Tracking the time course of spoken word recognition using
eye movements: Evidence for continuous mapping models.
Journal of Memory and Language,
38, 419-439.
Luce, P. A., & Pisoni, D. B. (1998). Recognizing spoken words:
The neighborhood activation model.
Ear and Hearing,
19, 1-36.
Marslen-Wilson, W. (1990). Activation, competition, and frequency
in lexical access. In G. T. M. Altmann (Ed.),
Cognitive Models of Speech Processing: Psycholinguistic and
Computational Perspectives,
pp. 148-172. Cambridge, MA: MIT Press.
|
