Individual Differences in Skilled Performance 511Figure 18.3 Practice and transfer performance in solving logic gate prob-
lems as a function of blocked or random problem presentation. Source:
Carlson and Yaure (1990).
reliance on memory. The inclusion of no-KR trials may also
lead to the development of the sort of internal representation
that is necessary for performers to detect errors on their own.
Whether feedback is intrinsic or extrinsic, it takes time to
process it: KR provided too soon after a trial can interfere
with the processing of intrinsic feedback (Swinnen, 1990;
Swinnen, Schmidt, Nicholson, & Shapiro, 1990).
Practice Schedules
The distinction between performance during practice and
learning as measured with retention or transfer conditions
has proven to be critical in evaluating the results of practice
schedules. For example, massing practice, such that only a
few sessions with many trials of practice are given in place
of more sessions with fewer trials in each session, has been
shown to have detrimental effects during acquisition but
varying effects on learning. Lee and Genovese (1988) noted
that studies with continuous tasks (such as tracking tasks; see
chapter by Heuer) show a small but negative effect of massed
practice on retention. Discrete tasks actually show more
learning when practice is massed.
A dissociation between effects of the scheduling of task
conditions on performance during practice and learning is
also seen when different variations of a task must be learned.
Blocking practice, such that one variation is practiced in one
session and another variation in a different one, has been
shown to lead to better performance than random practice, in
which all variations are possible within a block of practice.
However, learning, as assessed by transfer or retention
tests, is better for the random conditions (see Figure 18.3;
Carlson & Yaure, 1990; V. I. Schneider, Healy, Ericsson, &
Bourne, 1995; Shea & Morgan, 1979). It seems that the need
to recall task requirements on every trial, as in the random
condition, is essential to learning (Battig, 1979; Lee & Magill,
1983).INDIVIDUAL DIFFERENCES
IN SKILLED PERFORMANCEIndividual differences in various abilities have formed the
basis of selection and training research as well as a theo-
retical starting point for characterizing how skill develops.
Theoretically, some models make predictions about which
abilities should explain the most variance in skilled perfor-
mance at different levels of skill acquisition. From a practical
standpoint, the training and selection literature has focused
on determining the abilities that predict success in learning
particular skills.
The general progression from cognitive mediation to
an associative phase to automatic performance (e.g., Fitts,
1962/1990, 1964) forms the basis for Ackerman’s (1988,
1992) account of the relationship between level of skill ac-
quisition and cognitive ability. According to Ackerman, per-
formance in the early, declarative stage of learning a skill is
affected more than later stages by the background knowledge
and general spatial, verbal, and numeric abilities of the
learner. The development of more specific and streamlined
procedures in the associative phase leads to less reliance on
general declarative knowledge. In this stage, as speed and
efficiency develop and the need for conscious mediation
lessens, the dependence on general cognitive abilities is re-
duced and the perceptual speedof the learner, as measured
by tasks such as letter matching and serial response time, be-
comes a more important determinant of performance. Finally,
in the autonomous stage, in which task components have be-
come more automatic and performance is relatively free of
attentional demands, performance will be more subject to the
psychomotor ability of the performer.
Ackerman (1992) tested his model by comparing the cor-
relation between performance and ability at different levels
of skill acquisition in a complex, computerized air traffic con-
trol simulator. The effectiveness of measures of perceptual
ability as a predictor of performance was, as predicted, higher
at higher levels of skill. However, measures of general ability
were also better predictors at high skill levels. It may be that
tasks that require the integration of new information never
become independent of general ability. One reason for this
could be the dependence of such performance on working
memory. Another possibility, suggested by Matthews, Jones,
and Chamberlain (1992), who found that tests of ability in the