Mewhort et al. (1981) argued that location information is lost faster than item
information in Averbach and Coriell's (1961) bar-probe task. Because there are reasons to
question their choice of indices of loss of item information (i.e., extralist intrusion errors)
and of loss of location information (i.e., intralist intrusion errors), this experiment was
conducted to test Mewhort et al.'s (1981) dual-buffer model with two new indices.
The subjects were tested under both Conditions 1 and 2 of Mewhort et al.'s (1981)
study in this experiment. That is, the partial-report probe and the mask (when applicable)
were presented simultaneously. The ISI values used were 0, 40, 80, 120, and 160 ms. The
duration of the seven-item detection array was 20 ms.
Results
In order to compare the results of this experiment with Mewhort et al.'s (1981)
Conditions l and 2, the subjects' performance was scored in terms of intralist and extralist
intrusion errors. As can be seen from the top left panel of Figure 2, masking increased the
number of intralist intrusions, F(l, 13) _ 52.44; so did increases in ISI, F(4, 52) = 3.56.
Masking also increased the number of extralist intrusions, F(1, 13) = 54.49 (see top right
panel of Figure 2). Although ISI did not have any effect on extralist intrusions, it
interacted significantly with masking, F(4, 52) = 3.86.
As can be seen from the top panel of Figure 3, the subjects' performance in terms
of the correct recall of both item identity and its location declines systematically as ISI
increases under the no-masking condition. Masking reduced performance at short ISI
levels. These observations were confirmed by ANOVA, which showed that the masking
by ISI interaction was significant, F(4, 52) = 5.49. Also significant was the main effect of
masking, F(l, 13) = 110.71. Tukey's HSD (honestly significant difference) test did not
show any significant pairwise difference among the means of the five ISI levels under the
masking condition. Represented in the middle and the bottom panels of Figure 3 are the
conditional probabilities, p(I|L) and p(L|I), respectively.
The middle panel of Figure 3 shows that more item information is available in the
absence of masking than in its presence. As ISI increases, the availability of item
information decreases under the no-masking condition. The ANOVA revealed that the
masking by ISI interaction was significant, F(4, 52) = 6.16; so was the main effect of
masking, F(1, 13) = 87.36. Tukey's HSD test showed that the availability of item
information did not vary as ISI increased under the masking condition, despite the
apparently upward trend shown by the dotted-line function in the middle panel of Figure
3. A similar analysis for the availability of location information data did not reveal any
signficant effect, as may be seen from the bottom panel of Figure 3. That is, the
availability of location information did not vary with the delay of the probe, regardless of
the masking manipulation.