See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/241644798

The Neural Basis of Motor-Skill Learning

Article  in  Current Directions in Psychological Science · December 1999

DOI: 10.1111/1467-8721.00042

CITATIONS READS
58 157

1 author:

Daniel Willingham
University of Virginia
100 PUBLICATIONS   7,900 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Daniel Willingham on 05 July 2020.

The user has requested enhancement of the downloaded file.

 178 VOLUME 8, NUMBER 6, DECEMBER 1999

In the first motor-control

The Neural Basis of Motor-Skill process, the actor selects the envi-
Learning ronmental goal of the movement.
This process can support motor-
Daniel B. Willingham1 skill learning through the selection
Department of Psychology, University of Virginia, Charlottesville, Virginia of successively more effective
goals. This function corresponds
closely with the everyday use of
different motor-skill functions sub- the word strategy, and hence this
Abstract process is called strategic. For ex-
Recent work indicates that served by different brain areas.
This article provides an overview ample, a bowler faced with a diffi-
motor-skill learning is sup- cult split may try to make the 10
ported by four processes: a of some of these findings. Given
the space restrictions, this article pin strike the 7 pin.
strategic process that selects In the second motor-control
new goals of what to change in focuses on my own point of view,
specifically, on a theory of motor- process, the actor selects spatial tar-
the environment, a perceptual- gets for movements that will
motor integration process that skill learning I have recently pro-
posed (Willingham, 1998). More ec- achieve the environmental goal.
adjusts to new relationships This process is called perceptual-
between environmental stim- umenical reviews are available
(Salmon & Butters, 1995). motor integration. The environmen-
uli and the appropriate motor tal goal is selected in allocentric
response, a sequencing process space (i.e., a coordinate system in
that learns sequences of motor which objects are located relative to
acts, and a dynamic process FOUR PROCESSES
SUPPORTING MOTOR- one another), but the target for
that learns new patterns of movement is selected in egocentric
muscle activations. These four SKILL LEARNING
space (i.e., a coordinate system an-
processes can operate in one of chored on a part of the body).
two modes: an unconscious Motor-skill learning should be Allocentric space depends on vi-
mode, in which one is aware differentiated from motor control. sion, and egocentric space depends
only of the goal of the move- Motor control refers to the process- on proprioception (information
ment, or a conscious mode, in es that support the planning and about the position of the body that
which one consciously con- execution of movements. Motor- comes from receptors in the mus-
trols detailed aspects of the skill learning refers to the increas- cles, tendons, joints, and skin), so
movement. This article pro- ing spatial and temporal accuracy learning becomes necessary when
vides an overview of these of movements with practice. the relationship between them is
four processes and two modes, Recently, a number of researchers changed. For example, spectacles
and describes their neural have proposed that motor-skill made from wedge prisms will mis-
bases. processes may grow directly out of align vision and proprioception,
motor-control processes; in other making motor movements quite in-
Keywords words, motor skill may be nothing accurate, but movements improve
motor skill; learning; motor more or less than the increasingly with practice. Other, less disruptive
control efficient operation of motor-control changes also require learning (e.g.,
processes. the translation between screen lo-
If motor movements could not Figure 1 shows four hypotheti- cations and mouse locations for
be performed more quickly and ac- cal processes that support motor someone using a computer).
curately with practice, getting control. To make a movement, the The third motor-control process
dressed each morning would be a actor2 (a) selects a goal that some- sequences spatial targets for move-
time-consuming affair, and driving thing in the environment be ment. Learning supported by this
to one s office on a highway full of changed, (b) selects spatial targets sequencing process occurs when the
novice motorists would provide for movement that will achieve the actor must make the same se-
more thrills than most of us want at goal, (c) sequences the spatial tar- quence of movements repeatedly.
an early hour. In the past 10 years, gets, and (d) translates the se- For example, a tennis player per-
a great deal has been discovered quence of spatial targets into a pat- fecting a serve attempts to make
about the anatomic structures that tern of muscle activity. How might the same sequence of movements
support motor-skill learning. A key these processes also support each time. Many laboratory tasks
result has been the description of motor-skill learning? psychologists use in their experi-

Published by Blackwell Publishers, Inc.

 CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE 179

ing when the relationship between

egocentric space and muscle move-
ments is poorly represented (e.g.,
fine movements made with the
nonpreferred hand) or when the re-
lationship changes (e.g., because of
disfigurement or normal develop-
ment). This form of motor learning
has been little studied.
These four processes that sup-
Fig. 1. Four processes of motor control. In the strategic process (a), the actor decides
to move a drinking glass (filled square) to a new location (empty square). The spatial port motor control are summarized
locations are described in allocentric space (i.e., relative to the table). In perceptual- in Table 1.
motor integration (b), the spatial locations are translated into egocentric space, in this
case, relative to the location of the shoulder. In the sequencing process (c), the two
spatial locations are sequenced, to ensure that the current location of the glass is
reached first, and then the goal location of the glass. Finally, in the dynamic process
(d), the spatial targets are translated into a pattern of muscle activation to move the UNCONSCIOUS AND
hand to the targets. CONSCIOUS MODES OF
MOTOR CONTROL AND
MOTOR SKILL
ments call for a sequence of move- The fourth process, called dy-
ments to be repeated. For example, namic, translates the sequence of
many tracking tasks3 require sub- egocentric spatial targets into a pat- I have proposed that there are
jects to keep a cursor on a target tern of muscle activation. This two modes in which these four
that moves in a repeating pattern. process could support skill learn- processes can operate, and these

Table 1. The processes that support motor control

Mechanism of
Function in improvement in
Process motor control motor-skill acquisition Example Anatomic locus
Strategic Selects goal of move- Select more effective Hit a lob when Dorsolateral frontal
ment in environmental environmental goals opponent rushes cortex
coordinates the net

Perceptual- Selects spatial target Learning a new relationship Use a racquet to Premotor cortex,
motor or targets for move- between environmental hit a ball instead posterior parietal
integration ment that will fulfill and egocentric space of one’s hand cortex
environmental goal; because of a change in
represented in ego- vision or proprioception,
centric space or an incompatible
stimulus-response mapping

Sequencing Orders spatial targets Learning a repeating Stereotyping the Basal ganglia, supple-
in the correct sequence when the same movements for mentary motor
sequence movement is made a tennis serve cortex
repeatedly

Dynamic Translates egocentric Learning a new relation- Learning fine Spinal interneurons
spatial targets and a ship between egocentric coordination with
pattern of muscle targets and the pattern of nonpreferred hand
firing muscle firing necessary to
move the effector to the
spatial target

Copyright ' 1999 American Psychological Society

 180 VOLUME 8, NUMBER 6, DECEMBER 1999

two modes apply to both motor reaching movements without any The dorsolateral frontal cortex
control and motor skill. In the un- attempt to consciously counteract has been implicated in the strategic
conscious mode, the actor is con- the effects of the spectacles. Such process. Positron emission tomog-
scious only of setting the environ- movements are made in the uncon- raphy studies show activation in
mental goal for example, of scious mode, and although per- this region when subjects must se-
wanting to move a glass from one formance will initially be quite lect a movement to execute, relative
location on a table to another. The inaccurate, performance will im- to a condition in which the stimu-
other processes operate outside of prove with training because of lus specifies which movement to
consciousness. In the conscious learning in the perceptual-motor make. Further, patients with frontal
mode, the strategic process not only integration process. The subject lobe lesions have problems in se-
selects the environmental goal, but may, however, obtain explicit, con- lecting actions, although the deficit
also selects and sequences the spa- scious knowledge of the effect of can take different forms: Some pa-
tial targets for movement. Under the spectacles. The subject can then tients select very few actions (i.e.,
typical circumstances, the actor use the conscious mode to select a they are content to simply sit for
employs the unconscious mode target for movement that corrects long periods wherever they are
and is conscious only of wanting for the distorting effects of the placed); others select inappropriate
the glass moved from one location prisms. In this case, performance environmental goals (e.g., trying to
to the other. However, the actor can will improve much more quickly. open a can by pounding it with a
also use the conscious mode and Sequencing tasks, too, may be can-opener); others repeatedly se-
consciously consider the exact loca- learned wholly unconsciously, or lect the same environmental goal
tion in which the glass is to be more rapidly in the conscious (e.g., repeatedly beating an egg and
grasped and the sequence of move- mode. not progressing to the next step in
ments necessary to move the glass. baking a cake).
When the conscious mode is used, The perceptual-motor integra-
the output of the strategic process tion process can be localized to pos-
replaces the output of the sequenc- BRAIN BASIS OF MOTOR- terior parietal cortex and premotor
ing and perceptual-motor integra- SKILL LEARNING cortex. Its locus has been examined
tion processes. using the prism-spectacles task,
This use of the conscious mode even though learning in this task
can play a role in motor-skill learn- Recent evidence shows that the may entail both perceptual-motor
ing. In the prism-spectacles task, four processes described have dis- integration and strategic learning.
wedge prisms shift the visual tinct neural bases. Readers may Perceptual-motor integration learn-
world (often 30 to the right), and wish to refer to Figure 2 for an il- ing may be assessed in isolation by
the subject must point to visual tar- lustration showing some of these a transfer task that measures after-
gets. The subject may simply make locations. effects: If subjects are trained with
the prism spectacles and then asked
to point straight in front of the nose
with eyes closed (so that visual
feedback cannot be used), they
point in the direction opposite to
the prism transformation (e.g., if the
spectacles shifted the visual world
to the right, they point to the left).
This bias occurs because training
with the prism spectacles changes
proprioception subjects feel that
they are pointing straight ahead.
Functional imaging studies using
positron emission tomography and
magnetic resonance imaging show
that the posterior parietal cortex is a
critical site of learning in the prism-
spectacle task (Clower et al., 1996),
and other studies show that pa-
Fig. 2. Brain locations associated with motor-skill learning. See the text for details. tients with damage that spares pos-

Published by Blackwell Publishers, Inc.

 CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE 181

terior parietal cortex and premotor Learning continues in the un-

cortex (such as patients with BRAIN BASIS OF conscious processes all the while
Alzheimer s disease, Huntington s CONSCIOUS AND that the conscious mode is engaged
disease, and Parkinson s disease) UNCONSCIOUS LEARNING and controlling movement. Thus,
show normal aftereffects in this the subject may consciously control
task. These patients also successful- The strategic process selects en- a sequence of movements, but with
ly learn new relationships between vironmental goals for movement sufficient practice, this conscious
stimuli and the appropriate motor and sequences targets for move- control becomes unnecessary be-
responses (e.g., learning to use a ments when subjects respond in the cause the unconscious sequencing
computer mouse or joystick). conscious mode. There is some evi- process will have learned the se-
Learning motor sequences ap- dence that the latter function is quence. This interaction of the two
pears to rely on the basal ganglia supported by the dorsolateral processes would account for the
and supplementary motor cortex. frontal cortex. As noted earlier, the decreasing attentional demands
Patients with basal ganglia abnor- conscious mode can be used to im- observed with practice and the de-
malities due to Huntington s dis- plement a strategy in learning to velopment of automaticity. A recent
ease or Parkinson s disease show adjust to prism spectacles; one can behavioral study supports this
impaired learning of tracking tasks select a target for pointing that idea. Goedert-Eschmann and I
that use a target moving in a re- looks wrong but that adjusts for (Willingham & Goedert-Eschmann,
peating sequence, but they are able the effect of the prisms. Patients 1999) trained subjects in the serial
to learn tracking tasks normally if with frontal lobe lesions are im- response time task. Some subjects
the targets move randomly (such paired in adjusting to prism spec- learned the sequence consciously,
tasks do not require sequence tacles, perhaps because normal and some unconsciously. Next, all
learning). Patients with lesions that subjects spontaneously develop subjects were told that the stimuli
spare the basal ganglia (e.g., pa- conscious strategies for movement, would appear randomly, when in
tients with Alzheimer s disease) whereas frontal patients do not. fact the sequence occasionally ap-
show normal learning of sequenc- In a number of functional imag- peared in an otherwise random
ing tasks. Functional imaging stud- ing experiments, subjects learned trial block. None of the subjects no-
ies also implicate the basal ganglia to tap a particular sequence of fin- ticed the occasional appearance of
and supplementary motor area in ger-to-thumb movements. If the se- the trained pattern, and all subjects
sequence-learning tasks. These quence was learned consciously, showed equivalent unconscious
anatomic areas are consistently ac- activations associated with learn- knowledge of the pattern. The im-
tivated in neurologically intact sub- ing were observed in prefrontal plication is that subjects who had
jects who learn the serial response cortex (which includes the dorso- learned the sequence via the con-
time task4 or a tracking task in lateral frontal cortex and other cor- scious mode had simultaneously
which the target moves in a repeat- tical areas), as well as supplemen- learned it unconsciously.
ing sequence. In one recent study tary motor and premotor cortices.
(Rauch et al., 1997), the amount of If the sequence was learned uncon-
activation in the putamen (one sciously, little or no activation in AREAS OF FUTURE WORK
structure in the basal ganglia) was prefrontal cortex occurred (see
correlated with the amount of se- Willingham, 1998, for a review).
quence learning subjects showed in The frontal activation decreased The basic architecture of the
the serial response time task. with practice, and the decrease neurological substrate of motor-
As noted earlier, there has been began when subjects reported that skill learning is emerging. This ar-
virtually no work examining the they no longer needed to internally chitecture is composed of a number
neural basis of dynamic learning, count the finger taps (Seitz, Roland, of anatomically distinct processes,
but there is evidence indicating the Bohm, Greitz, & Stone-Elander, each performing a different func-
neural basis of the dynamic control 1990). But if subjects were then tion for motor-skill acquisition, and
process is in pools of interneurons5 asked to attend to the process of each rooted in processes of motor
in the spinal cord (see Bizzi, producing this very well learned control. Several outstanding ques-
Giszter, Loeb, Mussa-Ivaldi, & sequence of finger movements, the tions remain.
Saltiel, 1995). Thus, although it is frontal activity returned (Jueptner First, there are other brain struc-
plausible that spinal interneurons et al., 1997). These results strongly tures that appear to contribute to
may also support dynamic learn- suggest that the conscious mode of motor-skill learning, but their func-
ing, the issue has not yet been in- control is supported by the dorso- tion is not yet known. For example,
vestigated. lateral frontal cortex. functional imaging studies fre-

Copyright ' 1999 American Psychological Society

 182 VOLUME 8, NUMBER 6, DECEMBER 1999

quently show activation in the pri- logical analysis in outlining the nipulating a joystick or computer
mary motor cortex, but its role in broad framework of a model. The mouse.
4. In the serial response time task,
skill learning is not well under- detailed mechanisms within each
the subject sees four squares arrayed
stood. It may simply show activa- of the processes remain to be de- horizontally on a computer screen and
tion because of its connections to scribed. rests the index and middle fingers of
the supplementary motor and pre- each hand on response keys. One
motor cortices, or it may make an square becomes filled in black, and the
Recommended Reading subject must press the key correspon-
independent contribution to motor ding to that square, whereupon the
skill. The role of somatosensory Grafton, S.T., Hazeltine, E., & Ivry, square becomes white again, and a new
cortex (which is crucial to proprio- R.B. (1998). Abstract and effector- square is filled in black. The squares be-
ception) also remains obscure. The specific representations of motor come filled in a repeating sequence of
theory described here holds that sequences identified with PET. spatial positions, although the subject
Journal of Neuroscience, 15, is not told this. In a typical experiment,
proprioception is critical because of 9420 9428. the sequence is 12 units long, and noth-
its role in determining egocentric Karni, A., Meyer, G., Rey-Hipolito, ing marks the beginning or end of the
spatial location, and indeed there C., Jezzard, P., Adams, M.M., sequence, so it may appear to the sub-
are data showing impaired motor- Turner, R., & Ungerleider, L.G. ject to be a random stream of stimuli.
(1998). The acquisition of skilled Although many subjects remain un-
skill learning in the face of proprio-
performance: Fast and slow aware that the stimuli are sequenced,
ceptive loss, but this work has only experience-driven changes in steadily decreasing response times
begun. primary motor cortex. Proceed- nevertheless show that they have
A second area of future work ings of the National Academy of learned the sequence unconsciously.
concerns features of movement, Sciences, USA, 95, 861 868. 5. Interneurons connect with mo-
Salmon, D.P., & Butters, N. (1995). toneurons, which in turn directly drive
specifically, force and timing. All of (See References) muscle activity.
the work described in this article Shadmehr, R., & Holcomb, H.H.
has been concerned with the spatial (1997). Neural correlates of
References
aspect of skills. In most tasks, tim- motor memory consolidation.
ing information is confounded Science, 277, 821 825.
Bizzi, E., Giszter, S.F., Loeb, E., Mussa-Ivaldi, F.A.,
Willingham, D.B. (1998). (See & Saltiel, P. (1995). Modular organization of
with spatial information if it is References) motor behavior in the frog s spinal cord. Trends
present at all, but some recent work in Neurosciences, 18, 442 446.
Clower, D.M., Hoffman, J.M., Votaw, J.R., Faber,
shows that subjects can learn tim- T.L., Woods, R.P., & Alexander, G.E. (1996).
ing information on its own. It is Acknowledgments I thank Bobbie Role of posterior parietal cortex in the recali-
Spellman for many helpful suggestions bration of visually guided reaching. Nature,
well established that the cerebel- on previous versions of this article. 383, 618 621.
lum plays a crucial role in timing in Preparation of this article was supported Jueptner, M., Stephan, K.M., Frith, C.D., Brooks,
D.J., Frackowiak, R.S., & Passingham, R.E.
motor control, so one might expect by National Science Foundation Grant
(1997). Anatomy of motor learning: I. Frontal
SBR-9905342.
that the cerebellum is important for cortex and attention to action. Journal of
Neurophysiology, 77, 1313 1324.
learning temporal information in Rauch, S.L., Whalen, P.J., Savage, C.R., Curran, T.,
motor-skill tasks. Skill involved in Notes Kendrick, A., Brown, H.D., Bush, G., Breiter,
H.C., & Rosen, B.R. (1997). Striatal recruitment
force production has also been un- during an implicit sequence learning task as
derstudied. 1. Address correspondence to measured by functional magnetic resonance
Daniel B. Willingham, Department of imaging. Human Brain Mapping, 5, 124 132.
A third area of future work may Salmon, D.P., & Butters, N. (1995). Neurobiology
be clarification of the mechanisms Psychology, 102 Gilmer Hall, Uni- of skill and habit learning. Current Opinion in
versity of Virginia, Charlottesville, VA Neurobiology, 5, 184 190.
within each of the putative process- 22903. Seitz, R.J., Roland, P.E., Bohm, C., Greitz, T., &
es outlined here. I have proposed 2. Actor refers to a person execut- Stone-Elander, S. (1990). Motor learning in
man: A positron emission tomographic study.
that the basal ganglia and supple- ing a motor act. NeuroReport, 1, 57 66.
mentary motor cortex support 3. In a tracking task, a target moves Willingham, D.B. (1998). A neuropsychological
motor-skill learning but by what and the subject pursues it. Typically, theory of motor skill learning. Psychological
the task is administered on a computer. Review, 105, 558 584.
mechanism? How are sequences The target is a circle moving on a
Willingham, D.B., & Goedert-Eschmann, K. (1999).
The relation between implicit and explicit
learned? The work described screen, and the subject attempts to keep learning: Evidence for parallel development.
shows the power of a neuropsycho- a computer cursor on the target by ma- Psychological Science, 10, 531 534.

Published by Blackwell Publishers, Inc.

View publication stats