The fact that the partial-report probe was presented at the immediate offset of the
array ensured that there was minimal uncertainty regarding which item to recall. It
follows from the dual-buffer model that the availability of location information should
increase with increases in the mask-ISI of the masking array. Yet, there was not any
mask-ISI effect on the availability of location information in the presence of masking.
The absence of an asymptote in terms of the availability of item information as
mask-ISI increased is contrary to Mewhort et al.'s (1981) contention that backward
masking interferes with location information at the level of the feature buffer at short
SOA range. This finding supports the view that backward masking interrupts item
identification (Turvey, 1973).
General Discussion
It has been suggested that the partial-report superiority over whole report is the
result of an artifact brought about by the fact that ". . . measure is heavily biased against
full report . . ." (Mewhort & Butler, 1983, p. 32). This bias is presumably due to the fact
that the nature of response organization is different in whole and partial report (see also
Dick, 1971). Mewhort et al. (1981) also argue that the decline in partial-report superiority
with increases in mask-ISI reflects a progressive loss of useful location information. This
latter claim relies on the demonstration that the systematic decline in partial report is
complemented by a progressive increase in location errors.
The conditional probabilities, p (I|L) and p(L|I), were used as the measures of the
availability of item information and of location information, respectively. These
conditional probabilities were preferred to I (i.e., the correct recall of item identity) and L
(i.e., the correct recall of the item's intra-array location) because the latter two are
ambiguous at the theoretical level. Moreover, specific expectations can be derived from
both the dual-buffer model and the traditional view of iconic memory in terms of these
two conditional probabilities. One likely problem with these measures is that the subjects
might have guessed correctly either the item identity or its location on an unknown
proportion of trials. How would guessing affect these measures?
The numerator of both of these conditional probabilities is the probability of
recalling correctly both the item's identity and its location. The likelihood of getting both
the identity and location of the item correct by guessing may reasonably be assumed to be
negligible when compared to guessing either the identity of the item or its location.
Hence, it is reasonable to assume that guessing may have only a negligible effect on the
numerator of either of the two conditional probabilities. To the extent that either p(L) or
p(I) was inflated by guessing, p(I|L) or p(L|I) would be reduced. This, however, would
not affect the qualitative outcomes (i.e., the relation between the ISI functions with and
without masking) of this study if guessing was independent of the ISI manipulation. The
situation is more complicated if guessing was dependent on the ISI manipulation.