assumed to be available in the iconic store. Van der Heijden's (1984) argument can now
be considered.
Van der Heijden (1984) rejected the orthodox view of iconic memory because his
subjects' partial-report performance was affected by the irrelevant stimulus items in the
partial-report display. However, the identities of both the relevant and the irrelevant
stimulus items in his study were defined solely in terms of color, a dimension that is
generally accepted to be a precategorical dimension available at the iconic level.
Moreover, naming a color does not involve building a character from some discrete
features. For this reason, it can be suggested that Van der Heijden's (1984) choice of
stimulus has precluded him from studying iconic memory in the sense of Coltheart's
(1980) informational persistence.
In demonstrating that information at the iconic level is nonassociative,
Wickelgren and Whitman (1970) effectively showed that information in the probed
location was independent of the information in any other location if the choice of
stimulus material permitted the study of informational persistence. In other words, Van
der Heijden's (1984) data are not inconsistent with the orthodox view that selection can
be made at the iconic level if the criterion of selection is color.
Unlike Mewhort et al. (1981), Van der Heijden (I 984) found that as partial-report
performance declined with increases in mask-ISI, both item information and location
information also declined systematically with increases in mask-ISI. There are two
reasons why Van der Heijden's (1984) data are not necessary inconsistent with either
Mewhort et al.'s (1981) results or those of this series of experiments. Relevant to
Mewhort et al.'s (1981) findings is the previous comment that Van der Heijden's (1984)
choice of stimulus material necessarily precluded his subjects from engaging in forming
characters from features, a process necessary for Mewhort et al.'s (1981) task. This
comment is also applicable to the present study. Pertinent only to this series of
experiments is the fact that Van der Heijden (1984) essentially treated intralist and
extralist intrusion errors as indices of item and of location information, respectively. This
is the methodological assumption being questioned here.
Certain derivations have been made from Mewhort et al.'s (1981) dual-buffer
model even though the formulation of the model is not as explicit as expected. The
derivations were based primarily on the identify-then-select assumption. The dual-buffer
model was assessed in terms of two conditional probabilities when (a) the delay of the
partial-report probe was synchronized with the delay of the mask (Experiment 1), (b)
only the delay of the probe was varied when the mask was presented at the immediate
offset of the stimulus (Experiment 2), and (c) only the delay of the mask was varied when
the probe was presented at the immediate offset of the stimulus (Experiment 3). In all
cases, the theoretical expectations of the dual-buffer model were not met. Instead, the
results were consistent with the traditional view of iconic memory, particularly the
selectthen-identify implication.