location, not a misidentification of the location of a particular item within the character
buffer. If this is the case, it is inappropriate to consider responding with H in the previous
example as a location error at the level of the character buffer. In other words, it is
necessary to test Mewhort et al.'s (1981) contention with another index of location
information.
At the theoretical level, location information at the level of the character buffer
plays an important part in the dual-buffer model. For example, location information must
be represented in such a way that the relative location of the letters is preserved.
Moreover, the representation must make it possible for ". . . the attention mechanism [to
find] an item in the character buffer by using the probe as a location instruction . . ."
(Mewhort et al., 1981, p. 51). Despite this important role played by location information,
Mewhort et al. (1981) are vague about exactly how this location information is
represented and used. Nonetheless, a particular type of representation of location
information may be derived from their account.
It seems reasonable to suggest that the location information in the character buffer
may assume the form of location tags for the items in the character buffer. That is, it is
not inconsistent with the dual-buffer model to suggest that the location information of an
identified item is tagged to the item in the character buffer. For example, M at Location 1
of MBDHZLG may be represented as M-1 in the character buffer; Z as Z-2, and the like.
When the partial-report probe is presented, the item having a location tag that matches
the location of the partial-report probe is given as the response. That is, the essence of the
identify-then-select mode of operation is that location information is accessed via item
information, a mode of operation contrary to that envisaged in Treisman and Gelade's
(1980) model.
As the partial-report probe is progressively delayed, fewer and fewer location tags
remain useful, either as a result of decay or of their being transposed. For example, Z-5
might become Z-4, whereas H-4 may become H-5. Hence, partial-report performance
suffers, despite the fact that the information about the identities of the items has not
changed. Consequently, responding with H when the item occupying the fifth location of
the array is probed may be categorized as a location error.
The conclusion drawn by Mewhort et al. (1981) is that item information decays
from the character buffer at a much slower rate than location information. It follows from
this conclusion that at any delay of the partial-report probe the probability that item
information is available is higher than that of the availability of location information.
When coupled with the identify-then-select assumption, several testable implications can
be derived for Averbach and Coriell's (1961) partial-report task. However, the dependent
variables to be used in this study have to be discussed first.
What measures can be used to reflect the availability of item and location
information if Mewhort et al.'s (1981) measures are not satisfactory? This question has to