514 Procedural Memory and Skill Acquisition
mechanisms that mediate increasing improvements from re-
peated practice trials must exist and play important roles in
the acquisition of expertise. It is not merely exposure to a task
that provides the basis for expert performance, but conscious,
deliberate practice in which feedback is sought and used to
improve performance. Just as it does in skilled performance
in general, memory plays a large role in expertise. As men-
tioned above, experts have both a large, well-organized
knowledge base and, often, specialized procedures for ac-
quiring and storing knowledge. Also noteworthy is the speci-
ficity of the skills experts possess. For example, although
chess players show an exceptional memory for the positions
of chess pieces in various midgame positions, they do not
score any higher on general tests of spatial ability than do
controls (Doll & Mayr, 1987). As Thorndike and Woodworth
(1901a, 1901b, 1901c) argued, there is little evidence for a
“doctrine of formal discipline” (see Higginson, 1931) in
which practice in one difficult skill leads to generalizable
benefits in other domains. The study of expertise allows us to
add that there is little evidence that exceptional abilities are a
necessary prerequisite for the development of expertise.
TRAINING
Training has been a topic of interest to psychologists for the
past hundred years or so. The major question of interest has
been how broad the effects of training can be. As suggested
by instance accounts of skill acquisition, the effects of train-
ing are often quite tightly tied to the training conditions.
Therefore, it is necessary to determine what aspects of the
target environment need to be included in the training. A
related problem is that many skills are too complex to be
learned all at once. Effective training can therefore depend on
learning only certain aspects of a skill at a time. For these rea-
sons, methods of decomposing skills for training and then re-
combining them are needed if effective training programs are
to be designed.
Effective, efficient training programs depend on the iden-
tification of those aspects of the task that are critical for im-
proving skill. These aspects can be identified by interviewing
experts, by determining the characteristics that divide good
performers and bad performers, and through theoretical
analysis of the task. One approach is to emphasize cue-
response relations(Cormier, 1987) and to determine which
cues are necessary for the determination of responses. This
approach has been used in designing simulators for training
complex or dangerous tasks. Building high-fidelity simula-
tors is expensive, so there is pressure on designers to include
only those cues that lead to better transfer to the actual task.
If one can determine the relevant cue-response relations, only
the cues that are necessary need be incorporated into the
simulator. Unfortunately, determining these relations is not
always easy. For example, it has been found that motion cues
can lead to better performance in a flight simulator (e.g.,
Perry & Naish, 1964), but not to better transfer to actual flight
(e.g., Jacobs & Roscoe, 1975). In order to understand this dis-
crepancy, it is necessary to look at the type of motion cues pre-
sented. In general, the presence of disturbance motion cues
(cues associated with outside influences) are more important
for transfer of simulator training to actual flight. However, for
relatively unstable, difficult-to-fly aircraft, maneuver cues
(cues associated with control actions) can be important (for a
review see Gawron, Bailey, & Lehman, 1995).
Most techniques for analyzing tasks start with a
description of the complete human-machine system but focus
on the description, analysis, and evaluation of the perfor-
mance demands placed on the human. For example, the
focus might be on decomposing tasks into their constituent
information-processing requirements, such as the principles,
rules, and goals contained in expert knowledge, the distinc-
tion between automatic and controlled processes, or the
allocation of attention. An example of one such approach is
principled task decomposition(Frederiksen & White, 1989).
This method was used by Frederiksen and White to develop
a training program for the Space Fortress game, a video
game developed by researchers to study complex skill acqui-
sition (Mané & Donchin, 1989), and it is based on task
decomposition, an analysis of human information-processing
requirements, and the characteristics of expert performance.
Frederiksen and White first identified the hierarchical rela-
tionships between skill and knowledge components that
allow the progression from novice to expert performance and
then used this task decomposition to construct training activ-
ities for the component processes as well as their integration.
A comparison of the performance of a group who received
componential training and a control group who practiced the
Space Fortress game showed an initial deficit for the compo-
nential-training group when first transferred to whole-game
performance. However, the componential-training group
quickly overtook the whole-game training group, suggesting
that, after some initial integration of learned skills during
their first experience with the whole game, the specific
knowledge and heuristics taught in the componential training
had benefited learning (see Figure 18.4).
In general, part-task training, such as that used by
Frederiksen and White (1989), has been shown to be an ef-
fective method of training difficult tasks or tasks with inde-
pendent components (Holding, 1965; Wightman & Lintern,
1985). Several methods of part-task training have been