| |
Abstract:
Our study investigated whether gender information can
influence the earliest stages of pronoun interpretation, and how it
interacts with referent accessibility. Previous work on this issue
has produced mixed results, with several reading studies showing
little or no sensitivity to gender information during initial
processes of pronoun interpretation (e.g., Garnham et al., 1992;
Greene et al., 1992). By contrast, we present the results of two
eyetracking experiments showing that listeners use gender
information rapidly, and that this information interacts with
accessibility. These findings followed the same pattern as our
off-line measure of the conditions for felicitous pronoun use. In
both experiments, we measured participants' eye movements as they
viewed a picture and listened to a text, deciding whether the
picture matched the text. Each picture contained two familiar
cartoon characters; we manipulated a) whether the characters had
the same gender and b) whether the pronoun referred to the
more-accessible or less-accessible character. The picture made it
clear which character was the intended referent of the pronoun. In
Experiment 1 we manipulated accessibility through order of mention,
as in 1, where Donald is more accessible than Mickey/ Minnie.
(1) Donald is bringing some mail to {Mickey / Minnie}, while a
violent storm is beginning. {He's / She's} carrying an umbrella,
and it looks like they're both going to need it.
Gender and accessibility information were used rapidly to
identify the referent, beginning approximately 400ms after the
onset of the pronoun. We calculated "target advantage" scores
(looks to target minus looks to competitor), for three 200-ms
segments, beginning at the verb onset (approx. 200 ms after the
pronoun onset). In the first segment there was no difference among
conditions. In segments 2 and 3 subjects looked at the target more
than the competitor in all conditions except when the pronoun was
ambiguous and referred to the second character, reflecting a
reliable interaction between the two conditions. In this last
condition, subjects' initial eye movements depended on the point of
disambiguation. A post-hoc analysis revealed that when the verb
provided early disambiguating information, subjects focused more on
the correct target, but when disambiguation didn't come until late,
subjects experienced a temporary garden-path, looking more at the
first-mentioned character.
In Experiment 2 we heightened the difference in accessibility by
adding another pronominal reference to the first-mentioned
character before the critical pronoun ("He/ She's carrying..."), as
in 2.
(2) Donald is bringing some mail to {Mickey / Minnie). He's
sauntering down the hill, while a violent storm is beginning. {He's
/ She's} carrying an umbrella, and it looks like they're both going
to need it.
The results showed that subjects also identified the target
rapidly in the same conditions as in Experiment 1, replicating the
immediate effect of gender. Our manipulation of decreasing the
accessibility of the second-mentioned character resulted in
subjects finding the ambiguous pronominal reference to the
second-mentioned character to be highly infelicitous. This was
evident in the eye movements, which showed that subjects had a
strong preference to look more to the competitor at about 400ms
after the verb. In addition, subjects responded that the text did
not match the picture 48% of the time. This experiment also
replicated the effect of point-of-disambiguation found in
Experiment 1, but the early-disambiguation items merely showed
competition between the target and competitor, rather than an early
target advantage.
|
