memory set, and the distractors in the visual display were all selected from the
other set. There was a substantial improvement in performance over a total of
2100 trials, and it appeared to reflect the growth of automatic processes.
The most obvious problem with automatic processes is their lack of flexibil-
ity, which is likely to disrupt performance when there is a change in the pre-
vailing circumstances. This was confirmed in the second part of the study just
described. The initial 2100 trials with one consistent mapping were followed by
a further 2400 trials with the reverse consistent mapping. This reversal of the
mapping conditions had a markedly adverse effect on performance; indeed, it
took nearly 1000 trials under the new conditions before performance recovered
to its level at the very start of the experiment!
Shiffrin and Schneider (1977) conducted further experiments in which sub-
jects initially attempted to locate target letters anywhere in a visual display, but
were then instructed to detect targets in one part of the display and to ignore
targets elsewhere in the display. Subjects were less able to ignore part of the
visual display when they had developed automatic processes than when they
had made use of controlled search processes. In general terms, as Eysenck
(1982, p. 22) pointed out: ‘‘Automatic processes function rapidly and in parallel
but suffer from inflexibility; controlled processes are flexible and versatile but
operate relatively slowly and in a serial fashion.’’
Evaluation Shiffrin and Schneider’s (1977) theoretical approach is important,
but it is open to various criticisms. For example, there is a puzzling discrepancy
between theory and data with respect to the identification of automaticity. The
theoretical assumption that automatic processes operate in parallel and place
no demands on capacity means that there should be a slope of zero (i.e. a hori-
Figure 15.9
Response times on a decision task as a function of memory-set size, display-set size, and consistent
versus varied mapping. Data from Shiffrin and Schneider (1977).
386 Michael W. Eysenck and Mark T. Keane