Crucial to the valid application of the partial-report task is the assumption that the
select-then-identify strategy (called the "selective readout" by Coltheart, 1980) is adopted
by the subjects. (See also Treisman & Gelade, 1980, for a similar position.) However,
some investigators have argued that the selection takes place after the identities of the
briefly presented items have been established (Dick, 1969, 1971; Merikle, 1980;
Mewhort et al., 1981; Van der Heijden, 1984).
Dick (1969, 1971) and Merikle (1980) obtained data consistent with the
identify-then-select strategy and with Sperling's (1960) partial-report task. However, their
subjects were not given any meaningful training on the partial-report task. Subjects could
use the select-then-identify mode of processing only when they were given at least an
hour's training on the partialreport task (Chow, 1985). The observations made in this
series of experiments support the select-then-identify assumption.
Mewhort et al.'s (1981) dual-buffer model has recently been examined by Van der
Heijden (1984), who is also critical of the orthodox view of iconic memory as
represented by Averbach and Coriell (1961) and Coltheart (1977). Van der Heijden's
(1984) version of a dual-buffer model consists of Store A (in which only precategorical
information is available) and Store B (in which postcategorical information is stored).
The distinguishing marks of Van der Heijden's (1984) and Mewhort et al.'s (1981)
models are as follows:
1. Although only visual features (such as straight lines, angles, and the like) are
allowed in Mewhort et al.'s (1981) feature model, logogens of letters and of words, in
addition to visual features, are also allowed in the precategorical Store A in Van der
Heijden's (1984) model.
2. Selection under the partial-report task constraint is achieved by a process called
precategorical filtering at the level of Store A. In other words, Van der Heijden (1984)
argued for the identify-then-select option. However, this selection is done before Store B
(a store analogous to Mewhort et al.'s, 1981, character buffer).
How can the difference between Van der Heijden's (1984) and the present
conclusions be reconciled? The identify-then-select versus the select-then-identify issue
cannot be discussed meaningfully without first identifying the level of abstraction. For
example, transducing electromagnetic energy into a color sensation, extracting an angle
from the letter A, and identifying the letter A as A are all information processing
activities. However, they differ in sophistication. Traditionally, in the context of the
information-processing approach to visual perception (e.g., Haber & Hershenson, 1980),
transducing a particular frequency of electromagnetic energy into a color sensation is not
treated as an act of categorization. As a methodological assumption, many investigators
treat color as a precategorical dimension when they study iconic memory (e.g., Clark,
1969; Turvey & Kravetz, 1970; von Wright, 1968, 1970). That is, color information is