510 Procedural Memory and Skill Acquisition
sources of sensory information are integrated to form an
intermodal sensorimotor representation. Performance suffers
if a source of information is removed or added because the
incoming sensory information is then no longer compatible
with the sensorimotor representation. Because specificity
develops with practice, changes in information may result in
greater decrements in performance after extensive practice
than after moderate levels of practice. Thus, whether reliance
on visual information seems to increase could depend on when
such reliance is tested. More recently, Proteau, Tremblay, and
DeJaeger (1998) have suggested that, with practice, the
source of afferent information best suited to ensure optimal
performance progressively dominates other sources of sen-
sory information. The withdrawal of this information will
lead to a deterioration in performance only when its domi-
nance has been firmly established. Thus, withdrawing such a
source of afferent information early in practice will be less
detrimental than doing so later.
FACTORS INFLUENCING SKILL ACQUISITION
Coordination of different effectors, hierarchical control, and
perceptual-motor integration are all necessary for the devel-
opment of skill, but what are the factors that can enhance the
development of skill? Answering this question requires that
we make a distinction between factors that have an effect on
the performance of a task and factors that affect learning, as
measured by retention of the skill or performance on transfer
tasks. Factors that lead to better performance during training
do not necessarily lead to better learning. Bjork (1999) has
argued that immediate performance is based on the retrieval
strengthof newly made memories, whereas learning is based
on what he calls storage strength. He warns that training con-
ditions that support performance by providing a short-term
basis for ready access to correct responses or procedures may
impede the growth of the storage strength necessary to sup-
port long-term performance.
A number of factors have been identified that affect the
rate and extent of learning of motor tasks, and many of these
factors seem to play an equally important role in the learning
of cognitive tasks. Although factors such as the motivation
and ability of the performer have a big influence on the out-
come of practice, the factors that have been most extensively
studied are feedback and practice schedules.
Feedback
There are two major sources of feedback: intrinsic and ex-
trinsic. Intrinsic feedback is feedback that is directly produced
by the response, and this can include proprioceptive, visual,
auditory, and vestibular information. Contrary to an assump-
tion that skilled performance is automatic and therefore in-
creasingly less reliant on feedback, even skilled performance
can be dependent on intrinsic feedback. As suggested by
Proteau’s (1992) specificity-of-practice hypothesis, removing
feedback from a task practiced with feedback can disrupt per-
formance, as can adding visual feedback to a task learned
without such feedback (Elliott & Jaeger, 1988; Proteau et al.,
1992). The important point seems to be that, with practice, a
central representation of the relevant feedback is formed and
that this representation (like the stimulus-response represen-
tations in response-selection tasks) continues to be used in
highly skilled performance. It should be noted, however, that
some studies have found a decreased reliance on feedback. In
one such study, Pew (1966) found evidence that an early re-
liance on visual information in a higher-order tracking task
was replaced by a control strategy that was performed auto-
matically, with only occasional monitoring. One could argue,
however, that performers in Pew’s study learned to use pro-
prioceptive feedback or other information in place of visual
feedback.
Extrinsic feedback is feedback that is added to intrinsic
feedback. It might include hearing a beep when a mistake is
made or when a target is hit, watching a video of one’s own
performance, or viewing a plot of movement dynamics. An
important distinction is between knowledge of results (KR),
in which the outcomes (accuracy or speed) of a movement are
conveyed to the performer, and knowledge of performance,
in which information about the dynamics of movement (tem-
poral or spatial) is provided to the performer. Knowledge of
performance is more effective than KR when the task is more
complex than a simple pointing or tracking task.
It seems reasonable to think that KR will be most effective
when it is provided immediately and on every trial. However,
this is not always the case. For example, Winstein and
Schmidt (1990) found that just as much learning occurred
when KR was provided on 33% of trials in which a complex
movement had to be made as when it was provided on 100%
of the trials. Moreover, decreasing the percentage of trials on
which KR was provided across the training period led to bet-
ter learning. It has also been found that providing a summary
of performance at the end of a block of trials can be more
effective than providing feedback after every trial (Lavery,
1962; Schmidt, Young, Swinnen, & Shapiro, 1989). Schmidt
and colleagues have suggested that the function of feedback
is to guide the performer toward the performance goal. This
guidance hypothesis states that when feedback is provided on
every trial, performers become too dependent on it, which
leads to poorer performance on retention or transfer tests
without the feedback. It may be that the important process
that underlies the benefit for reduced feedback is a greater