that found in Mewhort et al.'s (1981) Condition 1, as may be seen from the middle two
panels of Figure 2.
The fact that the masking array was always present at the immediate offset of the
array, when applicable, ensured that the (SOA). was a constant one, namely, 100 ms. In
Mewhort et al.'s (1981) account, for short SOAS backward masking might disrupt the
location arrangement of the features in the feature buffer, thereby producing item errors.
Hence, masking should reduce the availability of item information because fewer correct
characters are transferred into the character buffer. That is, although the absolute amount
of information in the feature buffer is not affected by the mask, the absolute level of
information is reduced by masking if the character buffer is utilized.
That is, in terms of the availability of item information (the middle left panel of
Figure 4), a flat function across probe-ISI levels is prescribed by the dual-buffer model
for the masking condition in this experiment. This should be the case because (a) the
probe-ISI manipulation here was applicable to the partial-report probe only, and (b) the
delay of the partial-report probe disrupts only the usefulness of the location information
in the dual-buffer model. The function for the no-masking condition is expected to be
parallel to the flat function of the masking condition, but with a higher intercept. The
expectation of having two flat functions was not met. Instead, the availability of item
information decreased with increases in the delay of the partial-report probe in the
absence of masking.
Two alternatives have to be considered. First, the dual-buffer model is not
supported by the results of this experiment. Alternatively, the 100-ms SOA might be
more than sufficient to complete the transfer process (of features from the feature buffer
into the character buffer). However, the second alternative can be rejected in view of the
analysis in terms of p(L|I).
Mewhort et al. (1981) also suggested that backward masking might disrupt the
location information of the characters in the character buffer. Consequently, in terms of
the availability of location information, the probe-ISI function under the masking
condition should be parallel to the probe-ISI function without backward masking, but
with a smaller absolute value. Moreover, both functions should have a negative trend.
However, the two probe-ISI functions in the bottom right panel of Figure 4 are more or
less flat. The absence of any probe-ISI effect on the availability of location information is
contrary to the expectation of Mewhort et al.'s (1981) model that location information is
lost or transposed when the partial-report probe is delayed.
The fact that the presence of the masking array (which stayed on until the subjects had
responded) enhanced the recall of the location information is not surprising. Recall that
the masking array was always at the immediate offset of the stimulus array when
applicable. The masking array effectively extended the availability of the intra-array
location information.