In Haber and Hershenson's (1980) theoretical account, information about the
location, color, and size of stimuli is fully represented at the first stage of visual
information processing. Moreover, when alphanumeric stimuli are involved, their
representation at the retinal level is transformed into features, such as angles and straight
lines; the persistence of these various types of information constitutes the iconic store.
This view is consistent with Coltheart's notion of informational persistence. In the context
of the partial-report task, a "select-then-identify" strategy is assumed. That is, only the
bundle of features probed is selected for further processing to the item-identity level if the
critical bundle is among those which have not been processed by the time the
partial-report probe is presented. As the partial-report probe is delayed, the selected
bundle becomes less useful because some or all of its features may have decayed.
In contrast, Mewhort et al. (1981) envisage an "identify-then-select" mode of
operation in the partial-report situation. More specifically, they propose that the
processing of visual stimuli requires the postulation of a feature buffer (preidentification)
and a character buffer (postidentification). The selection of the probed item is carried out
in the character buffer. To follow Mewhort et al.'s (1981) terminology, this model will
subsequently be called the dual-buffer model.
Using Averbach and Coriell's (1961) version of the partial-report task, Mewhort
et al. (1981) found that as the decay of the probe was increased, partial-report
performance decreased together with a corresponding increase in location errors while
item errors remained more or less constant. This observation suggested to Mewhort et al.
(1981) that location information suffered when the partial-report cue was delayed. It
further suggested that partial report was superior to whole report because "the
[partial-report superiority] measure is heavily biased against full report. . ." (Mewhort &
Butler, 1983, p. 32).
When given Averbach and Coriell's (1961) task, the subject is presented with a
multi-item array (e.g., MBDHXLG) which is subsequently followed by a partial-report
probe after a predetermined delay. The subject is required to recall only the item
indicated by the probe. That is, the criterion of selection is an intra-array location. With
reference to the array, MBDHZLG, Z is the correct response if Location 5 is probed.
Under such circumstances, there are two kinds of errors. The subject may respond with
an item in the array other than the correct one (e.g., H). This is an intra-array intrusion.
Alternatively, the subject may respond with an item not present in the array (e.g., N).
This is an extra-array intrusion. While acknowledging the possibility of some ambiguity
Mewhort et al. (1981) nonetheless suggest that intra-array intrusions signify the loss of
location information, whereas extra-array intrusions indicate the loss of item information.
Although it is straightforward to identify an intra-array intrusion, it is not easy to
know what it signifies (a point also recognized by Mewhort et al., 1981, and Van der
Heijden, 1984). For example, although Position 5 is probed, the subject may have
mistaken it to be a probe for Position 4. That is, this is a misidentification of the probe's