of the availability of location information, the systematic decline of partial report in the
absence of masking was not matched by a systematic decrease in location information as
ISI increased. On the contrary, the availability of location information was not related to
the delay of either the probe or the mask. Instead, the systematic decline in correct
performance was matched by a systematic decrease in the availability of item information
in the absence of masking.
The theoretical expectations of the orthodox view of iconic memory, as depicted
in the two right-hand panels of Figure 1 are met by the data shown in the middle and
bottom panels of Figure 3. That is, whereas location information does not change, item
information declines systematically with increases in ISI. This is consistent with the
select-then-identify mode of operation implied in Coltheart's (1980) and Treisman and
Gelade's (1980) models.
Experiment 2
This experiment was designed to test the dual-buffer model by examining the
effects of probe delay when the amount of processing time is held constant by using a
fixed stimulus onset asynchrony (SOA). As has been noted earlier although masking
disrupts location information in the dual-buffer model, the manifestation of this effect
may be qualitatively different, depending on whether the disruption occurs in the feature
buffer or the character buffer.
If the (SOA). is short, the disruption of location information is said to take place
in the feature buffer. The stimulus identification process is hence disrupted because some
of the features may be displaced. That is, the availability of item information suffers. This
expectation has been depicted in the middle left panel of Figure 4 in terms of p(I|L). More
specifically, two parallel flat functions are expected because the delay of the probe is said
to have no effect on item information. Given a sufficiently long (SOA). only location
information at the character buffer is interrupted. When coupled with the delay of the
partial-report probe (whose effect is assumed to affect location information only), the
mask is expected to lower the absolute amount of location information. That is, two
parallel functions with a negative trend are expected. This expectation has been depicted
in the bottom left panel of Figure 4. The top left panel of Figure 4 represents the expected
systematic decline of partial-report performance with increases in ISI.
From Mewhort et al.'s (1981) description, it is not possible to determine if an
(SOA). of 100 ms is sufficiently long to complete the transfer of feature information from
the feature buffer to the character buffer. If this (SOA). is not long enough for the process
to be completed, the expectations described in the middle left panel of Figure 4 should
follow. On the other hand, if this (SOA). is of sufficient length, the expectations depicted
in the bottom left panel of Figure 4 should be obtained. The dual-buffer model is in doubt
if neither of these expectations is met.