be considered with reference to what the subject has to do. In the experiments to be
reported, the subjects were given Averbach and Coriell's partial-report task and required
to report both the letter probed as well as its intra-array location. First, consider the
conditional probability of correct recall of item identity, given correct recall of its
intra-array location (i.e., p (I|L), where I is the correct recall of item identity and L is the
correct recall of its intra-array location). It has been noted that location information is
accessed via item information if the identify-then-select mode of operation is in force.
That is, p(I|L) represents the proportion of trials in which the correct recall of location is
mediated by correct recall of item identity. Hence, the conditional probability, p(I|L),
may be used as an index of the availability of item information if the dual-buffer model
and the identify-then-select assumption are jointly adopted.
The second conditional probability of interest is the probability of correct recall of
the intra-array location of the probe, given correct recall of item identity (i.e., p[L|I]).
This conditional probability is the proportion of trials in which the correct position
information is still attached to the item. Otherwise, the recall of location will be incorrect
because position information is always acessed via item information in the dual-buffer
model. Hence, the conditional probability, p(L|I), may be used as an index of the
availability of location information. Some expectations in terms of p(I|L) and p(L|i) may
now be derived from the dual-buffer model.
Consider the no-masking condition. First, although its absolute level is uncertain,
the conditional probability of item recall, given location recall (i.e., p[I|L] should be
constant, regardless of the delay of the partial-report probe (see the solid-line function in
the top left panel of Figure 1). The second implication is that the conditional probability
of location recall, given item recall (i.e., p[L|I], should decrease with increases in the
interval between the offset of the stimulus and the onset of the probe, an interval called
the interstimulus interval (ISI), regardless of its absolute level at ISI = 0 ms (see the
solid-line function of the bottom left panel of Figure 1).
According to the dual-buffer model, the effect of backward masking at short mask
delays is different from that at long mask delays. At short delays, the mask interrupts the
features in the feature buffer, thereby bringing about problems in identification. That is,
at short delays, masking increases item errors. If the mask is delayed sufficiently long,
the system will have sufficient time to establish the characters in the character buffer.
Hence, the mask interrupts only location information at the level of the character buffer.
That is, masking increases location errors at long mask delays.
What should be expected from the dual-buffer model if a mask is presented
concurrently with the probe in terms of the two conditional probabilities? It may first be
suggested that in terms of p(I|L), the ISI function should increase with increases in ISI
until an asymptote is reached; the function would remain at the asymptotic level with
further increases in ISI (see the dotted line function in the top left panel of Figure 1).
Second, in terms of p(L|I), the ISI function should remain flat until a certain critical value